Clinical Chemistry

April 1 2007, Volume 53, Issue 4 , pp.
543-806
Editorials:
Glen L. Hortin
A New Era in Protein Quantification in Clinical Laboratories: Application of
Liquid Chromatography-Tandem Mass Spectrometry
Clin Chem 2007 53: 543-544.
Current Issues in Laboratory Medicine:

Curtis A. Parvin and Sanford Robbins, III
Evaluation of the Performance of Randomized versus Fixed Time Schedules
for Quality Control Procedures
Clin Chem 2007 53: 575-580. Published online February 22, 2007;
10.1373/clinchem.2006.083311
Molecular Diagnostics and Genetics:

Peng Hou, Zaozao Chen, Meiju Ji, Nongyue He, and Zuhong Lu
Real-time PCR Assay for Ultrasensitive Quantification of DNA-Binding
Proteins
10.1373/clinchem.2006.077503
Claudia Langebrake, Kalle Günther, Jürgen Lauber, and Dirk Reinhardt
Preanalytical mRNA Stabilization of Whole Bone Marrow Samples
10.1373/clinchem.2006.078592
Constance L.H. Lo, Shea Ping Yip, Peter K.C. Cheng, Tony S.S. To, Wilina W.L. Lim,
and Polly H.M. Leung
One-Step Rapid Reverse Transcription–PCR Assay for Detecting and Typing
Dengue Viruses with GC Tail and Induced Fluorescence Resonance Energy
Transfer Techniques for Melting Temperature and Color Multiplexing
10.1373/clinchem.2006.077446
Wolfgang Lalouschek, Georg Endler, Martin Schillinger, Kety Hsieh, Wilfried Lang,
Suzanne Cheng, Peter Bauer, Oswald Wagner, and Christine Mannhalter
Candidate Genetic Risk Factors of Stroke: Results of a Multilocus
Genotyping Assay
10.1373/clinchem.2006.073494
Natale Scaramozzino, Audrey Ferrier-Rembert, Anne-laure Favier, Corinne
Rothlisberger, Stéphane Richard, Jean-Marc Crance, Hermann Meyer, and Daniel Garin
Real-Time PCR to Identify Variola Virus or Other Human Pathogenic
Orthopox Viruses
Clin Chem 2007 53: 606-613. Published online March 1, 2007;
10.1373/clinchem.2006.068635
Evidence-Based Laboratory Medicine and Test

Utilization:
Robert H. Christenson
National Academy of Clinical Biochemistry Laboratory Medicine Practice
Guidelines for Utilization of Biochemical Markers in Acute Coronary
Syndromes and Heart Failure
10.1373/clinchem.2006.079749
NACB WRITING GROUP MEMBERS, Fred S. Apple, Robert L. Jesse, L. Kristin
Newby, Alan H.B. Wu, Robert H. Christenson, NACB COMMITTEE MEMBERS,
Robert H. Christenson, Fred S. Apple, Christopher P. Cannon, Gary Francis, Robert
Jesse, David A. Morrow, L. Kristin Newby, Jan Ravkilde, Alan B. Storrow, Wilson Tang,
Alan H.B. Wu, IFCC COMMITTEE ON STANDARDIZATION OF MARKERS OF
CARDIAC DAMAGE (C-SMCD) MEMBERS, Fred S. Apple, Robert H. Christenson,
Allan S. Jaffe, Johannes Mair, Jordi Ordonez-Llanos, Franca Pagani, Mauro Panteghini,
Jillian Tate, and Alan H.B. Wu
National Academy of Clinical Biochemistry and IFCC Committee for
Standardization of Markers of Cardiac Damage Laboratory Medicine
Practice Guidelines: Analytical Issues for Biochemical Markers of Acute
Coronary Syndromes
10.1373/clinchem.2006.084715
NACB WRITING GROUP MEMBERS, David A. Morrow, Christopher P. Cannon,
Robert L. Jesse, L. Kristin Newby, Jan Ravkilde, Alan B. Storrow, Alan H.B. Wu, Robert
H. Christenson, NACB COMMITTEE MEMBERS, Robert H. Christenson, Fred S.
Apple, Christopher P. Cannon, Gary Francis, Robert L. Jesse, David A. Morrow, L.
Kristin Newby, Jan Ravkilde, Alan B. Storrow, Wilson Tang, and Alan H.B. Wu
National Academy of Clinical Biochemistry Laboratory Medicine Practice
Guidelines: Clinical Characteristics and Utilization of Biochemical Markers
in Acute Coronary Syndromes
10.1373/clinchem.2006.084194
Hemostasis and Thrombosis:

Boris T. Ivandic, Evangelos Giannitsis, Philipp Schlick, Peter Staritz, Hugo A. Katus,
and Thomas Hohlfeld
Determination of Aspirin Responsiveness by Use of Whole Blood Platelet
Aggregometry
10.1373/clinchem.2006.081059
Proteomics and Protein Markers:

Erwin H.J.M. Kemna, Harold Tjalsma, Vladimir N. Podust, and Dorine W. Swinkels
Mass Spectrometry–Based Hepcidin Measurements in Serum and Urine:
Analytical Aspects and Clinical Implications
10.1373/clinchem.2006.079186
Christian Melle, Günther Ernst, Niko Escher, Daniel Hartmann, Bettina Schimmel,
Annett Bleul, Heike Thieme, Roland Kaufmann, Klaus Felix, Helmut M. Friess, Utz
Settmacher, Merten Hommann, Konrad K. Richter, Wolfgang Daffner, Horst Täubig,
Thomas Manger, Uwe Claussen, and Ferdinand von Eggeling
Protein Profiling of Microdissected Pancreas Carcinoma and Identification
of HSP27 as a Potential Serum Marker
10.1373/clinchem.2006.079194
Amelie Plymoth, Claes-Göran Löfdahl, Ann Ekberg-Jansson, Magnus Dahlbäck, Per
Broberg, Martyn Foster, Thomas E. Fehniger, and György Marko-Varga
Protein Expression Patterns Associated with Progression of Chronic
Obstructive Pulmonary Disease in Bronchoalveolar Lavage of Smokers
10.1373/clinchem.2006.076075
John F. Timms, Elif Arslan-Low, Aleksandra Gentry-Maharaj, Zhiyuan Luo, Davy
T’Jampens, Vladimir N. Podust, Jeremy Ford, Eric T. Fung, Alex Gammerman, Ian
Jacobs, and Usha Menon
Preanalytic Influence of Sample Handling on SELDI-TOF Serum Protein
Profiles
10.1373/clinchem.2006.080101
Minna A. Korolainen, Tuula A. Nyman, Paula Nyyssönen, E. Samuel Hartikainen, and
Tuula Pirttilä
Multiplexed Proteomic Analysis of Oxidation and Concentrations of
Cerebrospinal Fluid Proteins in Alzheimer Disease
10.1373/clinchem.2006.078014
Cancer Diagnostics:
Heather L. Beyer, Ryan D. Geschwindt, Curtis L. Glover, Ly Tran, Ingegerd Hellstrom,
Karl-Erik Hellstrom, M. Craig Miller, Thorsten Verch, W. Jeffrey Allard, Harvey I. Pass,
and Niranjan Y. Sardesai
MESOMARKTM: A Potential Test for Malignant Pleural Mesothelioma
10.1373/clinchem.2006.079327
Olga P. Bondar, David R. Barnidge, Eric W. Klee, Brian J. Davis, and George G. Klee
LC-MS/MS Quantification of Zn- 2 Glycoprotein: A Potential Serum
Biomarker for Prostate Cancer
10.1373/clinchem.2006.079681
Lipids, Lipoproteins, and Cardiovascular Risk Factors:

Gregory T. Jones, Andre M. van Rij, Jennifer Cole, Michael J.A. Williams, Emma H.
Bateman, Santica M. Marcovina, Meiying Deng, and Sally P.A. McCormick
Plasma Lipoprotein(a) Indicates Risk for 4 Distinct Forms of Vascular
Disease
10.1373/clinchem.2006.079947
Yoosoo Chang, Seungho Ryu, Eunju Sung, and Yumi Jang
Higher Concentrations of Alanine Aminotransferase within the Reference
Interval Predict Nonalcoholic Fatty Liver Disease
10.1373/clinchem.2006.081257
Renke Maas, Friedrich Schulze, Jens Baumert, Hannelore Löwel, Khatera Hamraz,
Edzard Schwedhelm, Wolfgang Koenig, and Rainer H. Böger
Asymmetric Dimethylarginine, Smoking, and Risk of Coronary Heart
Disease in Apparently Healthy Men: Prospective Analysis from the
Population-Based Monitoring of Trends and Determinants in Cardiovascular
Disease/Kooperative Gesundheitsforschung in der Region Augsburg Study
and Experimental Data
10.1373/clinchem.2006.081893
Drug Monitoring and Toxicology:

Frank T. Peters, Nele Samyn, Thomas Kraemer, Wim J. Riedel, and Hans H. Maurer
Negative-Ion Chemical Ionization Gas Chromatography–Mass Spectrometry
Assay for Enantioselective Measurement of Amphetamines in Oral Fluid:
Application to a Controlled Study with MDMA and Driving Under the
Influence Cases
10.1373/clinchem.2006.081547
Endocrinology and Metabolism:

Santica Marcovina, Ronald R. Bowsher, W. Greg Miller, Myrlene Staten, Gary Myers,
Samuel P. Caudill, Scott E. Campbell, Michael W. Steffes for the Insulin Standardization
Workgroup
Standardization of Insulin Immunoassays: Report of the American Diabetes
Association Workgroup
10.1373/clinchem.2006.082214
Kristina Anna Strnadová, Margareta Holub, Adolf Mühl, Georg Heinze, Rene
Ratschmann, Hermann Mascher, Sylvia Stöckler-Ipsiroglu, Franz Waldhauser, Felix
Votava, Jan Lebl, and Olaf A. Bodamer
Long-Term Stability of Amino Acids and Acylcarnitines in Dried Blood
Spots
10.1373/clinchem.2006.076679
Henri Déchaud, Anne Denuzière, Sabina Rinaldi, Julien Bocquet, Hervé Lejeune, and
Michel Pugeat
Age-Associated Discrepancy between Measured and Calculated Bioavailable
Testosterone in Men
10.1373/clinchem.2006.077362
Antoon J.M. Janssen, Frans J.M. Trijbels, Rob C.A. Sengers, Jan A.M. Smeitink,
Lambert P. van den Heuvel, Liesbeth T.M. Wintjes, Berendien J.M. Stoltenborg-
Hogenkamp, and Richard J.T. Rodenburg
Spectrophotometric Assay for Complex I of the Respiratory Chain in Tissue
Samples and Cultured Fibroblasts
10.1373/clinchem.2006.078873
Automation and Analytical Techniques:

John Middleton and Jeffrey E. Vaks
Evaluation of Assigned-Value Uncertainty for Complex Calibrator Value
Assignment Processes: A Prealbumin Example
10.1373/clinchem.2006.081174
Johanna E.M. Groener, Ben J.H.M. Poorthuis, Sijmen Kuiper, Mariette T.J. Helmond,
Carla E.M. Hollak, and Johannes M.F.G. Aerts
HPLC for Simultaneous Quantification of Total Ceramide,
Glucosylceramide, and Ceramide Trihexoside Concentrations in Plasma
10.1373/clinchem.2006.079012
Duncan C.S. Talbot, Richard M. Ogborne, Tony Dadd, Herman Adlercreutz, Geoff
Barnard, Susanne Bugel, Fortune Kohen, Sandra Marlin, Jerry Piron, Aedin Cassidy, and
Jonathan Powell
Monoclonal Antibody-Based Time-Resolved Fluorescence Immunoassays for
Daidzein, Genistein, and Equol in Blood and Urine: Application to the
Isoheart Intervention Study
10.1373/clinchem.2006.075077
Clinical Immunology:
Santosh K. Ghosh, Thomas A. Gerken, Keith M. Schneider, Zhimin Feng, Thomas S.
McCormick, and Aaron Weinberg
Quantification of Human ß-Defensin-2 and -3 in Body Fluids: Application
for Studies of Innate Immunity
10.1373/clinchem.2006.081430
Other Areas of Clinical Chemistry:
Andrew S. Levey, Josef Coresh, Tom Greene, Jane Marsh, Lesley A. Stevens, John W.
Kusek, Frederick Van Lente for Chronic Kidney Disease Epidemiology Collaboration
Expressing the Modification of Diet in Renal Disease Study Equation for
Estimating Glomerular Filtration Rate with Standardized Serum Creatinine
Values
10.1373/clinchem.2006.077180
Markus Herrmann, Omid Taban-Shoma, Ulrich Hübner, Anette Pexa, Heiko Kilter,
Natalia Umanskaya, Rainer Hans Straub, Michael Böhm, and Wolfgang Herrmann
Hyperhomocysteinemia and Myocardial Expression of Brain Natriuretic
Peptide in Rats
10.1373/clinchem.2006.077859
Technical Briefs:
Zia Fazili, Christine M. Pfeiffer, and Mindy Zhang
Comparison of Serum Folate Species Analyzed by LC-MS/MS with Total
Folate Measured by Microbiologic Assay and Bio-Rad Radioassay
10.1373/clinchem.2006.078451
Hsiao-Mei Wiedmeyer, Kenneth S. Polonsky, Gary L. Myers, Randie R. Little, Carla J.
Greenbaum, David E. Goldstein, and Jerry P. Palmer
International Comparison of C-Peptide Measurements
10.1373/clinchem.2006.081570
Ulrike Haug, Timo Hillebrand, Peter Bendzko, Michael Löw, Dietrich Rothenbacher,
Christa Stegmaier, and Hermann Brenner
Mutant-Enriched PCR and Allele-Specific Hybridization Reaction to Detect
K-ras Mutations in Stool DNA: High Prevalence in a Large Sample of Older
Adults
10.1373/clinchem.2006.078188
Christina Dahl, Karen Grønskov, Lars A. Larsen, Per Guldberg, and Karen Brøndum-
Nielsen
A Homogeneous Assay for Analysis of FMR1 Promoter Methylation in
Patients with Fragile X Syndrome
Clin Chem 2007 53: 790-793. Published online January 26, 2007;
10.1373/clinchem.2006.080762
Letters to the Editor:
Birger Mensel, Ulrike Wenzel, Markus Roser, Jan Lüdemann, and Matthias Nauck
Considerably Reduced Centrifugation Time without Increased Hemolysis:
Evaluation of the New BD Vacutainer® SSTTMII Advance
Clin Chem 2007 53: 794-795.
Helen L. Bailey and Emily M. Chan
Liposomal Amphotericin B Interferes with the Phosphorus Assay on the
Synchron LX 20 Analyzer
Clin Chem 2007 53: 795-796.
Fiona M.F. Lun, Rossa W.K. Chiu, Tak Y. Leung, Tse N. Leung, Tze K. Lau, and Y.M.
Dennis Lo
Epigenetic Analysis of RASSF1A Gene in Cell-Free DNA in Amniotic Fluid
Clin Chem 2007 53: 796-798.
Mariasilvia Tommasi and Silvia Raspanti
Comparison of Calcitonin Determinations by Polyclonal and Monoclonal
IRMAs
Clin Chem 2007 53: 798-799.
Stanley F. Lo, Bernadine Jendrzejczak, and Basil T. Doumas
Total or Neonatal Bilirubin Assays in the Vitros 5,1 FS: Hemoglobin
Interference, Hemolysis, Icterus Index
Clin Chem 2007 53: 799-800.
Geralyn Lambert-Messerlian, Christina Bandera, Elizabeth Eklund, Andrew Neuhauser,
and Jacob Canick
Very High Inhibin A Concentration Attributed to Heterophilic Antibody
Interference
Clin Chem 2007 53: 800-801.
Sarah Molyneux, Michael Lever, Christopher Florkowski, and Peter George
Plasma Total Coenzyme Q9 (CoQ9) in the New Zealand Population:
Reference Interval and Biological Variation
Clin Chem 2007 53: 802-803.
Etienne Cavalier, Pierre Delanaye, Jean-Marie Krzesinski, and Jean-Paul Chapelle
Analytical Variation in Plasma Renin Activity: Implications for the
Screening of Primary Aldosteronism
Clin Chem 2007 53: 803-804.
Katherine Duxbury, Louise Gallagher, and Brian Keevil
The Impact of Simultaneous Measurement of Testosterone and
Androstenedione in Women with Suspected Androgen Excess
Clin Chem 2007 53: 804-805.
Book, Software, and Web Site Reviews:
Roland Valdes, Jr
An A-Z Guide to Pharmacogenomics. Michael G. Catania. Washington DC:
AACC Press, 2006, 58 pp., $24 ($19 AACC members), softcover. ISBN 1-
59425-047-2.
Clin Chem 2007 53: 806.
Emily S. Winn-Deen
Pharmacogenomics, Second Edition. Werner Kalow, Urs B. Meyer, and
Rachel F. Tyndale, eds. Boca Raton, FL: Taylor & Francis Group, 2005, 696
pp., $199.95, hardcover. ISNB 1-57444-878-1.
Clin Chem 2007 53: 806.
Corrections:
Correction
Clin Chem 2007 53: 805.
Editorial
A New Era in Protein Quantification in Clinical Laboratories:
Application of Liquid Chromatography-Tandem Mass Spectrometry
In this issue of Clinical Chemistry, Bondar and coworkers proaches commonly applied in clinical laboratories. Al-
at the Mayo Clinic report on development of a quantita- though the report by Bondar et al. (1 ) conceptually does
tive liquid chromatography-tandem mass spectrometry not break new ground in its analytical approach, it does
(LC-MS/MS) assay for zinc-␣2-glycoprotein (1 ). The assay illustrate how quantitative analysis of proteins by LC-
described may be more notable for the approach em- MS/MS is becoming a practical technique for clinical
ployed rather than for the specific analyte. Unlike most laboratories to implement, using equipment that currently
current clinical laboratory assays of proteins, the ap- is applied to other clinical analyses.
proach used does not rely on the specificity of antibodies There has been some question about how new technol-
for capture and detection. Instead, the assay described by ogies developed for proteomic analysis will impact the
Bondar et al. cleaves the protein into small peptides by clinical laboratory (2, 9 –11 ). Profiling of intact peptides or
proteolysis with trypsin. By use of LC-MS/MS, one of the proteins by 2-dimensional electrophoresis or matrix-
peptide fragments released from zinc-␣2-glycoprotein is assisted laser desorption time-of-flight mass spectrome-
ratioed vs a stable isotope-labeled form of the same ters serves as a discovery tool for top-down proteomics.
peptide. The method serves as an example of an approach Chromatographic separations of proteolytic digests of
to protein quantification that does not require specific proteins analyzed by ion trap or Fourier-transform ion
antibody reagents and may be broadly applicable. cyclotron resonance mass spectrometers offer approaches
The instrument used for this analysis, an LC-triple for bottom-up proteomics. These approaches, however,
quadrupole mass spectrometer, is an increasingly com- present substantial challenges for routine clinical applica-
mon tool in the clinical laboratory. Its range of application and for achieving the usual standards of clinical
tions now includes routine assays for (a) detecting inborn laboratory practice (12 ). In contrast, multiple reaction
errors of metabolism, (b) monitoring therapeutic drugs, monitoring in LC-MS/MS, using triple-quadrupole mass
and (c) quantifying steroid and thyroid hormones. The spectrometers, is a proven clinical laboratory technique
technology provides high selectivity and high throughput for quantitative analysis of molecules in complex matrices
in a clinical laboratory environment. MS/MS usually such as serum and plasma. Application of this technique
offers high specificity, even in complex sample matrices, to protein analysis shows promise for providing high
through selection of a specific precursor ion in the first accuracy and analytical specificity, and should be readily
mass analyzer and selection of a specific fragment ion adaptable to simultaneous multiplex analysis of many
formed during passage of the precursor ion through a proteins. The one major limitation of this technique, from
collision cell. Experience analyzing therapeutic drugs and the standpoint of proteomic analysis, is that it is a targeted
hormones in blood shows the ability of the method to approach, which requires foreknowledge of which pro-
measure components down to nanomole per liter concen- teins to analyze. This characteristic can be a drawback
trations. Considering that each of the 30 most abundant from the standpoint of marker discovery, but not neces-
plasma proteins usually has concentrations well over 1 sarily from the standpoint of a clinical assay, for which
␮mol/L (2 ), many proteins are likely to be sufficiently analytes preferably are known entities.
abundant for analysis by this technique in unfractionated It is becoming increasingly evident that new technolo-
specimens. Low abundance proteins, with concentrations gies evolving from proteomics research are not simply of
⬍1 nmol/L, however, are likely to require some form of academic or research interest. LC-MS/MS employing
fractionation or concentration, such as immunocapture of multiple reaction monitoring is one example of a new
the target protein or peptide (3 ). technology for protein analysis that is likely to have
The report by Bondar et al. (1 ) is but one of numerous substantial practical impact in clinical laboratories. One of
reports describing the application of LC-MS/MS to the the most immediate consequences of the application of
quantification of proteins for research purposes. Most of this technology is the development of new reference
these studies have described approaches for relative methods to standardize protein assays. Two recent exam-
quantification by use of differential isotopic labeling of ples are restandardization of laboratory measures of he-
paired specimens (4 ). A number of reports describe the moglobin A1c (13 ) and C-peptide (14 ). This approach is
absolute quantification of proteins through use of stable likely to be applied progressively to standardization of a
isotope-labeled internal standard peptides (3, 5– 8 ). Such wider range of protein assays, such as for urine albumin
internal standards can readily be prepared, either by using a recently described LC-MS/MS technique (15 ).
chemical synthesis or by in vitro translation. Most of the Because LC-MS/MS measures the abundance of only one
applications developed in research laboratories, however, peptide segment of a protein, expression of the abundance
have relied on capillary LC, nanospray interfaces, and ion of this peptide in units of molarity rather than mass/
trap or other types of mass spectrometers not commonly volume generally is more appropriate, and it may be
found in clinical laboratories (3, 4, 6 – 8 ). These LC- possible to develop chemically-synthesized peptides that
MS/MS techniques may be more challenging to apply as can serve as reference materials for standardization of
routine analytical techniques than the LC-MS/MS ap- protein assays. Finally, the ability to measure the concen-
Clinical Chemistry 53, No. 4, 2007 543

544 Hortin: A New Era in Protein Quantification
tration of any protein of sufficient abundance, without the 9. Diamandis EP. Mass spectrometry as a diagnostic and a cancer biomarker
discovery tool: opportunities and potential limitations. Mol Cell Proteomics
need for purified reference preparations or specific anti- 2004;3:367–78.
bodies, should accelerate the discovery and application of 10. Hortin GL, Jortani SA, Ritchie JC, Valdes RC, Chan DW. Proteomics: a new
measurements of additional proteins for diagnostic diagnostic frontier. Clin Chem 2006;52:1218 –22.
11. Plebani M. Proteomics: the next revolution in laboratory medicine? Clin Chim
applications. Acta 2005;357:113–22.
12. Hortin GL. Can mass spectrometric profiling meet desired standards of
References clinical laboratory practice? [Editorial]. Clin Chem 2005;51:3–5.
13. Sacks D for the ADA/EASD/IDF Working Group of the HbA1c Assay. Global
1. Bondar OP, Barnidge DR, Klee EW, Davis BJ, Klee GG. Quantification of
harmonization of hemoglobin A1c. Clin Chem 2005;51:681–3.
Zn-␣2-glycoprotein (ZAG) by LC-MS/MS: a potential serum biomarker for
14. Rodriguez-Cabaleiro D, Stockl D, Kaufman JM, Fiers T, Thienpont LM.
prostate cancer. Clin Chem 2007;53:673– 8.
Feasibility of standardization of serum C-peptide immunoassays with iso-
2. Hortin GL. The MALDI-TOF mass spectrometric view of the plasma proteome tope-dilution liquid chromatography-tandem mass spectrometry. Clin Chem
and peptidome. Clin Chem 2006;52:1223–37. 2006;52:1193–5.
3. Anderson NL, Anderson NG, Haines LR, Hardie DB, Olafson RW, Person TW. 15. Babic N, Larson TS, Grebe SK, Turner ST, Kumar R, Singh RJ. Application of
Mass spectrometric quantitation of peptides and proteins using stable liquid chromatography-mass spectrometry technology for early detection of
isotope standards and capture by anti-peptide antibodies (SISCAPA). J microalbuminuria in patients with kidney disease. Clin Chem 2006;52:
Proteome Res 2004;3:235– 44. 2155–7.
4. Julka S, Regnier F. Quantification in proteomics through stable isotope
coding: a review. J Proteome Res 2004;3:350 – 63.
Glen L. Hortin
5. Barnidge DR, Goodmanson MK, Klee GG, Muddiman DC. Absolute quantifi-
cation of the model biomarker prostate-specific antigen in serum by LC-
MS/MS using protein cleavage and isotope dilution mass spectrometry. J Department of Laboratory Medicine
Proteome Res 2004;3:644 –52.
National Institutes of Health,
6. Anderson L, Hunter CL. Quantitative mass spectrometric multiple reaction
monitoring assays for major plasma proteins. Mol Cell Proteomics 2006;5: Building 10, Room 2C-407,
573– 88. Bethesda, MD 20892
7. Kuhn E, Wu J, Karl J, Liao H, Zolg W, Guild B. Quantification of C-reactive Fax 301-402-1885
protein in the serum of patients with rheumatoid arthritis using multiple
reaction monitoring mass spectrometry and 13C-labeled peptide standards. E-mail ghortin@mail.cc.nih.gov.
Proteomics 2004;4:1175– 86.
8. Beynon RJ, Doherty MK, Pratt JM, Gaskell SJ. Multiplexed absolute quanti-
fication in proteomics using artificial QCAT proteins of concatenated signa- DOI: 10.1373/clinchem.2006.083857
ture peptides. Nat Methods 2005;2:587–9.
Clinical Chemistry 53:4 Evidence-Based
545–546 (2007) Laboratory Medicine
and Test Utilization
National Academy of Clinical Biochemistry

Laboratory Medicine Practice Guidelines for
Utilization of Biochemical Markers in Acute
Coronary Syndromes and Heart Failure
Preamble ers of Acute Coronary Syndromes. The later of these docu-
The National Academy of Clinical Biochemistry’s (NACB) ments represents a joint effort between the NACB and
Laboratory Medicine Practice Guidelines (LMPG) for use the International Federation of Clinical Chemistry’s
of cardiac markers in coronary artery diseases were Committee on Standardization of Markers of Cardiac
published in July 1999 (1 ). Since production of this initial Damage. The remaining sections involving other clin-
document, numerous published studies and presented ical and analytical issues for utilization of biochemical
data have added significantly to the knowledge base for markers of cardiac injury are available at http://www.
biochemical markers of cardiac injury. This increased aacc.org /AACC/members /nacb/LMPG / OnlineGuide/
knowledge has substantially expanded the scope of rec- PublishedGuidelines.
ommendations for biochemical marker utilization since
the 1999 document, and in particular has required the
inclusion of recommendations regarding biochemical
markers that extend beyond myocardial necrosis. Toward Table 1. American College of Cardiology/American Heart
addressing these advances and their impact on biochem- Association Classifications.
ical marker utilization in clinical practice, the NACB Summary of Indications
appointed a chair and members of an LMPG committee Class I Conditions for which there is evidence
and/or general agreement that a
that was charged with revising and extending the earlier given procedure or treatment is
recommendations by establishing modern guidelines for useful and effective.
Utilization of Biochemical Markers in Acute Coronary Class II Conditions for which there is
Syndromes and Heart Failure. These extended recom- conflicting evidence and/or a
divergence of opinion about the
mendations include Clinical Utilization of Biochemical usefulness/efficacy of a procedure
Markers in Acute Coronary Syndromes (ACS); Analytical or treatment.
Issues of ACS Biochemical Markers; Clinical Utilization of IIa Weight of evidence/opinion is in favor
Biochemical Markers of Heart Failure and Hemodynamic of usefulness/efficacy.
Stress; Analytical Issues of Heart Failure Biochemical IIb Usefulness/efficacy is less well
established by evidence/opinion.
Markers; Point-of-Care Testing and Logistics; and Cardiac
Class III Conditions for which there is evidence
Biomarkers and Other Etiologies. Updated draft revisions and/or general agreement that the
of the Guidelines were prepared and placed for comment procedure/treatment is not useful/
on the NACB World Wide Web site (http: // www.aac- effective and in some cases may
be harmful.
c.org /AACC/members /nacb/LMPG/
Weight of Evidence
OnlineGuide/DraftGuidelines/BioHearFailure/) in Au-
Level of Evidence A Data derived from multiple
gust of 2004. The draft LMPG and suggested revisions randomized clinical trials or
were also presented for public and stakeholder comment appropriately designed studies that
at the October 2004 Arnold O. Beckman Conference titled involved large numbers of patients
Cardiac Markers: Establishing Guidelines and Improving Level of Evidence B Data derived from a limited number of
randomized trials that involved
Results. small numbers of patients, or from
The articles presented here provide the NACB’s up- careful analyses of nonrandomized
dated recommendations for the rapidly evolving topics studies or observational registries.
of Clinical Utilization of Biochemical Markers in Acute Coro- Level of Evidence C Expert consensus was the primary
basis for the recommendation.
nary Syndromes and Analytical Issues for Biochemical Mark-
545
546 Christenson: Preamble NACB Practice Guidelines in ACS
The strength of scientific data supporting each recom- relationships within the 2 years previous to this publica-
mendation is characterized using the scoring criteria tion that may be relevant to this guidelines document. A
adopted from the American Heart Association/Ameri- document of those relationships may be found in the
can College of Cardiology as summarized in Table 1. online Data Supplement at http://www.clinchem.org/
For each recommendation, the designations I, IIa, IIb, content/vol53/issue4.
and III describe the indications, and the upper-case
letters A through C describe the weight of evidence.
Levels of evidence listed in the guidelines were deter- Reference
mined by the full writing committee. Ratings were 1. Wu AH, Apple FS, Gibler WB, Jesse RL, Warshaw MM, Valdes R Jr.
discussed in detail until consensus of the full writing National Academy of Clinical Biochemistry Standards of Laboratory
committee was reached. There was no process for Practice: recommendations for the use of cardiac markers in
coronary artery diseases. Clin Chem 1999;45:1104 –21.
recusal. The members of the committee have reported
all relevant relationships with industry to be published Robert H. Christenson
with these guidelines.
These guidelines were developed utilizing best National Academy of Clinical Biochemistry Laboratory
available evidence, and they incorporated substantial Medicine Practice Guidelines for Utilization of Biochemical
input from acknowledged experts and professional Markers in Acute Coronary Syndromes and Heart Failure
organizations. As such, they represent the current best University of Maryland School of Medicine
practice for utilization of biochemical markers of car- Baltimore, MD
diac injury.
This article has been copublished in the April 3, 2007 online

issue of Circulation.
Financial Disclosures: The National Academy of Clinical © 2007 American Association for Clinical Chemistry and the
Biochemistry Laboratory Medicine Practice Guidelines American Heart Association, Inc.
Committee for Utilization of Biomarkers in Acute Coro-
DOI: 10.1373/clinchem.2006.079749
nary Syndromes and Heart Failure reports all reported
National Academy of Clinical Biochemistry and

IFCC Committee for Standardization of Markers
of Cardiac Damage Laboratory Medicine Practice
Guidelines: Analytical Issues for Biochemical
Markers of Acute Coronary Syndromes
NACB WRITING GROUP MEMBERS
Fred S. Apple,1 Robert L. Jesse,2 L. Kristin Newby,3 Alan H.B. Wu,4 and
Robert H. Christenson5*
NACB COMMITTEE MEMBERS
Robert H. Christenson, Chair; Fred S. Apple; Christopher P. Cannon, Boston, MA;
Gary Francis, Cleveland, OH; Robert Jesse; David A. Morrow, Boston, MA;
L. Kristin Newby; Jan Ravkilde, Aarhus, Denmark; Alan B. Storrow, Nashville, TN;
Wilson Tang, Cleveland, OH; and Alan H.B. Wu
IFCC COMMITTEE ON STANDARDIZATION OF MARKERS OF CARDIAC DAMAGE
(C-SMCD) MEMBERS
Fred S. Apple, Chair; Robert H. Christenson; Allan S. Jaffe, Rochester, MN;
Johannes Mair, Innsbruck, Austria; Jordi Ordonez-Llanos, Barcelona, Spain;
Franca Pagani, Brecia, Italy; Mauro Panteghini, Milan, Italy;
Jillian Tate, Brisbane, Australia; and Alan H.B. Wu
1
Hennepin County Medical Center, Minneapolis, MN.
2
Medical College of Virginia, Richmond, VA.
3
Duke University Medical Center, Durham, NC. I. OVERVIEW OF ANALYTICAL ISSUES FOR
4
University of California at San Francisco, San Francisco, CA. ACUTE CORONARY SYNDROME (ACS)
5
University of Maryland School of Medicine, Baltimore, MD.
All relationships with industry for the guidelines committee are reported BIOMARKERS ............................................................548
online at ⬍http://www.aacc.org/AACC/members/nacb/LMPG/OnlineGuide/ A. Analytical Issues: Background ..........................548
PublishedGuidelines/ACSHeart/heartpdf.htm⬎. II. ANALYTICAL BIOMARKER ISSUES....................548
The materials in this publication represent the opinions of the authors and
committee members, and do not necessarily represent the official position of the
A. Cardiac Troponin Specifications .......................548
National Academy of Clinical Biochemistry (NACB) or the International Federa- B. Cardiac Biomarker Turnaround ........................549
tion of Clinical Chemistry (IFCC). The National Academy of Clinical Biochemistry C. Biomarkers No Longer Recommended for Use in
is the academy of the American Association for Clinical Chemistry. the Context of ACS..............................................549
* Address correspondence to this author at: Director, Rapid Response
Laboratories, University of Maryland School of Medicine, 22 S. Greene St.,
D. Determining Biomarker Decision Cutoff Charac-
Baltimore, MD 21201. Fax 410-328-5880; e-mail rchristenson@umm.edu. teristics for ACS ...................................................549
This article has been copublished in the April 3, 2007 online issue of E. European Society of Cardiology (ESC)/American
Circulation.
College of Cardiology (ACC) Recommenda-
Previously published online at DOI: 10.1373/clinchem.2006.084715
© 2007 American Association for Clinical Chemistry and the American tions ........................................................................550
Heart Association, Inc. III. REFERENCES .............................................................550
547
548 Apple et al.: NACB Practice Guidelines in ACS
I. Overview of Analytical Issues for Acute Coronary mmends a minimum of 120 individuals per group of
Syndrome (ACS) Biomarkers healthy individuals for appropriate statistical deter-
a. Background mination of a normal reference limit cutoff.
In 1999, the National Academy of Clinical Biochemistry
Sex-specific reference limits should be used in
(NACB)6 published the first standards of laboratory prac-
clinical practice for CK-MB mass. For myoglobin,
tice addressing analytical and clinical recommendations
the 97.5th percentile (with sex-specific reference
for use of cardiac markers in coronary artery diseases (1 ).
limits) should be the decision-limit for myocardial
The objectives were to recommend the appropriate im-
injury (Level of Evidence: B).
plementation and utilization of cardiac biomarkers, spe-
2. One decision-limit, the 99th percentile, is recom-
cifically for cardiac troponin (cTn), which had just gained
mended as the optimum cutoff for cTnI, cTnT, and
US Food and Drug Administration (FDA) clearance as
CK-MB mass. ACS patients with cTnI and cTnT
a cardiac biomarker to aid in the diagnosis of acute
results above the decision-limit should be labeled
myocardial infarction (AMI). In 2001, the IFCC Commit-
as having myocardial injury and a high-risk pro-
tee on Standardization of Markers of Cardiac Damage
file (Level of Evidence: B).
(C-SMCD) recommended quality specifications for ana-
3. Assays for cardiac biomarkers should strive for a
lytical and preanalytical factors for cTn assays (2 ). The
total imprecision (%CV) of ⱕ10% at the 99th
objectives were intended for use by the manufacturers of
percentile reference limit. Before introduction into
commercial assays and by clinical laboratories that use
clinical practice, cardiac biomarker assays must be
cTn assays. The overall goal was to establish uniform
characterized with respect to potential interfer-
criteria so that all cTn assays could objectively be evalu-
ences, including rheumatoid factors, human anti-
ated for their analytical qualities and clinical performance.
mouse antibodies, and heterophile antibodies. Pre-
These general principles can also be applied to creatine
analytical and analytical assay characteristics
kinase MB (CK-MB) mass and myoglobin assays by use of
should include biomarker stability (over time and
the analytical recommendations in this document. In this
across temperature ranges) for each acceptable
report, we provide the background for establishing up-
specimen type used in clinical practice and iden-
dated practice guidelines with recommendations addressing
tification of antibody/epitope recognition sites for
analytical issues for cardiac biomarkers based on 8 years of
each biomarker. Analytical and preanalytical spec-
evidence-based medical and scientific observations since the
ifications developed by professional groups such
publication of the initial recommendations (1 ).
as the IFCC should be followed (Level of Evi-
dence: C).
II. Analytical Biomarker Issues 4. Serum, plasma, and anticoagulated whole blood
recommendations: analytical aspects of acs are acceptable specimens for the analysis of car-
biomarkers diac biomarkers. Choice of specimen must be
based on sufficient evidence and the known char-
acteristics of individual biomarker assays (Level of
all class i Evidence: C).
1. Reference decision-limits should be established for
each cardiac biomarker based on a population of
normal, healthy individuals without a known his- a. cTn specifications
tory of heart disease (reference population). For First, in the context of cTn, the epitopes recognized by the
cardiac troponin I (cTnI) and T (cTnT), as well as antibodies must be delineated. Epitopes located on the
for CK-MB mass, the 99th percentile of the refer- stable part of the cTnI molecule should be a priority.
ence population should be the decision-limit for Specific relative responses need to be described for the
myocardial injury. The Clinical Laboratory Stan- following cTnI forms: free cTnI, the I-C binary complex,
dards Institute (CLSI; formerly NCCLS) reco- the T-I-C ternary complex, and oxidized, reduced, and
phosphorylated isoforms of the 3 cTnI forms. The effects
of different anticoagulants on binding of cTnI also need
to be addressed. Second, the source of material used to
calibrate cTn assays, specifically for cTnI, should be
6
Nonstandard abbreviations: NACB, National Academy of Clinical Bio- reported. A cTnI standardization subcommittee of the
chemistry; ACS, acute coronary syndrome; cTn, cardiac troponin; FDA, US
Food and Drug Administration; AMI, acute myocardial infarction; C-SMCD,
AACC in collaboration with the NIST has developed a
Committee on Standardization of Markers of Cardiac Damage; CK-MB, primary reference material (SRM #2921) (3 ). Although
creatine kinase MB; cTnI, cardiac troponin I; cTnT, cardiac troponin T; CLSI, this material demonstrated commutability with only
Clinical Laboratory Standards Institute; POC, point-of-care; TAT, turnaround
50% of current cTn assays, it will be of use in harmonizing
time; MI, myocardial infarction; AHA, American Heart Association; ESC,
European Society of Cardiology; ACC, American College of Cardiology; and cTnI concentrations across different assays (4, 5 ). At
WHF, World Heart Federation. present, it appears that the only way to achieve complete
standardization for cTnI would be for all manufacturers assist in improving myocardial tissue specificity when the
to agree on using the same antibody pairs for all commer- ratio of CK-MB to total CK is greater than previously
cial assays as well as a common reference material for established reference intervals. This concept is empha-
calibration (6, 7 ). The IFCC C-SMCD is currently explor- sized in a statement from the American Heart Association
ing the development of a serum-based secondary refer- (AHA) Council on Epidemiology and Prevention regard-
ence material. For cTnT, as there is only one assay ing case definitions for acute coronary heart disease in
manufacturer, harmonizing between assay generations epidemiology and clinical research studies (12 ). The fol-
has been consistent. Third, manufacturers need to use lowing recommendations were made to allow for a more
methods advocated by the CLSI to characterize detection accurate interpretation of recent trends in ACS during
limit, functional sensitivity, and total imprecision (8, 9 ). implementation of cTn assays and use of the European
Key characteristics for cTn assays include determination Society of Cardiology (ESC)/American College of Cardi-
of the distribution of values in a healthy reference popu- ology (ACC) consensus MI definition (13, 14 ) predicated
lation, the statistical determination of the 99th percentile on cTn: (a) simultaneous use of traditional biomarkers
cutoff for the reference population, and determination of with cTn to determine the performance of new biomark-
the concentration corresponding to the 10% CV (total ers; and (b) use of adjustment factors in databases and
imprecision). Preanalytical factors that should be de- retrospective studies seeking to determine incidence and
scribed include effect of storage time and temperature, trends of MI before and after cTn– derived studies.
effect of glass vs plastic tubes and gel separator tubes, and
the influence of anticoagulants for plasma and whole- d. determining biomarker decision cutoff
blood measurements. As more assay systems are devised characteristics for acs
for point-of-care (POC) testing, identical analytical criteria The 99th percentile of a reference decision-limit (medical
must apply to both central laboratory methodologies and decision cutoff) for cTn assays should be determined in
POC testing systems. When measuring cTn by different each local laboratory by internal studies using the specific
methodologies within the same institution, assay results assay that is used in clinical practice or validating a
should be harmonized or a strategy implemented to avoid reference interval that is based on findings in the litera-
interpretative confusion by clinicians. ture (13, 16 ). Desirable imprecision (expressed as %CV) of
each cTn assay (and CK-MB mass assay) has been defined
b. cardiac biomarker turnaround as ⱕ10% CV at the 99th percentile reference limit (13, 16 ).
Clinicians and laboratorians continue to support a goal Unfortunately, the majority of laboratories have neither
for turnaround times (TATs) ⬍60 min for cardiac bio- the resources to perform adequately powered 99th per-
markers, but the largest study published to date has centile reference studies nor the ability to carry out CLSI
demonstrated that TAT expectations are not being met in protocols to establish total imprecision criteria for the
a large proportion of hospitals (10 ). A College of Ameri- cTn assay that they plan to use in practice (17 ). Therefore,
can Pathologists Q-probe survey study of 7020 cTn and clinical laboratories must rely on the peer-reviewed pub-
4368 CK-MB determinations in 159 predominantly North lished literature to assist in establishing both local refer-
American hospitals demonstrated that the median and ence limits and imprecision characteristics. Caution must
90th percentile TATs for troponin were 74.5 and 129 min, be taken when comparing the findings reported in the
and for CK-MB, 82 and 131 min. In fact, fewer than 25% of manufacturers’ package inserts, which have been cleared
hospitals were able to meet the ⬍60-min TAT, defined as by the FDA, with the findings reported in journals be-
order-to-report time. A separate subanalysis of just POC cause of differences in total sample size, distributions by
testing systems was not reported. Recently published data sex and ethnicity, age ranges, and statistical methods used
have shown that implementation of POC cTn testing can to calculate the 99th percentile reported.
decrease TATs to ⬍30 min in cardiology critical-care and To date, very few in vitro diagnostic companies have
short-stay units (11 ). These data highlight the continued published 99th percentile limits in their package inserts.
need for laboratory services and healthcare providers to There is no established guideline set by the FDA or other
work together to develop better processes to meet a regulatory agencies to mandate a consistent evaluation of
⬍60-min TAT as requested by physicians. the 99th percentile reference limit for cTns. The largest
and most diverse reported reference value study to date
c. biomarkers no longer recommended for use shows plasma (heparin) 99th percentile reference limits
in the evaluation of acs for 8 cTn assays (7 cTnI and 1 cTnT) and 7 CK-MB mass
Use of aspartate aminotransaminase, total lactate dehy- assays (18 ). This study was performed in 696 healthy
drogenase, and lactate dehydrogenase isoenzymes are not adults (ages 18 – 84 years) stratified by sex and ethnicity.
recommended for evaluation of cardiac injury and detec- There was a 13-fold difference between the lowest vs the
tion of myocardial infarction (MI). The use of total CK or highest measured cTnI 99th percentile limit. The lack of
CK-MB activity is an acceptable alternative for evaluating cTn assay standardization (there is no primary reference
cardiac injury in institutions where cTn or CK-MB mass material that is commutable with all commercial methods,
assays are not available or feasible. Total CK can also as noted earlier) and the differences in antibody epitope
550 Apple et al.: NACB Practice Guidelines in ACS
recognition between assays (different assays use different defined in earlier studies, the degree of imprecision at
antibodies, as noted earlier) give rise to substantially these concentrations should be reported.
discrepant concentrations. What is generally recognized,
though, is that as long as one understands the character- e. european society of cardiology/american
istics of an individual assay and does not attempt to college of cardiology recommendations
compare absolute concentrations between different as- An ESC/ACC consensus document along with the
says, clinical interpretation should be acceptable for all AHA/ACC guidelines for differentiating AMI and unsta-
assays. ble angina codified the role of cTn by advocating that the
For CK-MB (as has been recognized for years for total diagnosis of AMI be based on increases of cTnI or T
CK), all assays demonstrate a significant 2- to 3-fold (preferred) or CK-MB mass above the 99th percentile
higher 99th percentile limit for men vs women (18 ). cutoff in the appropriate clinical situation (14, 22 ). The
Further, CK-MB can demonstrate up to 2- to 3-fold higher guidelines recognized the reality that neither the clinical
concentrations for African Americans vs Caucasians— presentation nor the electrocardiogram had adequate clin-
differences attributed to between-race physiological dif- ical sensitivity and specificity. The guidelines do not
ferences in muscle mass. These data led to the class I suggest that all increases of these biomarkers should elicit
recommendation that clinical laboratories should estab- a diagnosis of MI or high-risk profile, only those associ-
lish different CK-MB reference limits based on sex. Labs ated with the appropriate clinical, electrocardiogram, im-
should also consider doing so for ethnic groups. aging, or pathological findings. When cTn increases are
For cTn, expert consensus has emerged in support of not due to acute ischemia, the clinician is obligated to
the 99th percentile as the reference cutoff, in spite of search for another etiology for the elevation (6, 23 ). Up-
whether the total imprecision of the assay is ⱕ10%. This dated guidelines addressing the revised universal defini-
has been supported by a recent study that has demon- tion of MI cosponsored by the Joint ESC-ACC-AHA-
World Heart Federation (WHF) Task Force For The
strated that misclassification of patients who are ruled out
Redefinition of MI will soon be published. This document
using cTn assays with variable imprecision (10%–25%) at
will support and coincide with the recommendations
the 99th percentile does not lead to significant patient
proposed in the current joint NACB IFCC document.
misclassification over serial cTn orders (19 ). Further,
whereas the literature has been enriched with studies
appropriately addressing the total imprecision of cTn
All authors and committee members of the NACB and
assays, as to what the lowest concentration will be to
IFCC C-SMCD groups disclose that they have received
attain a 10% CV, the manufacturers’ package inserts often
either research funding, honoraria, expenses at sponsored
publish imprecision data primarily based on within-run
meetings, or consulting fees from at least one manufac-
or between-day precision. Again, there is no consistent
turer of cTn assays.
regulatory specification regarding precision data that
should be reported in the manufacturers’ package inserts.
To better address day-to-day clinical laboratory practice, Financial Disclosures: The National Academy of Clinical
early findings from an IFCC C-SMCD study demon- Biochemistry Laboratory Medicine Practice Guidelines
strated that the total imprecision for 13 commercial assays Committee for Utilization of Biomarkers in Acute Coro-
[based on a 20-day CLSI protocol (20 )] was unable to nary Syndromes and Heart Failure reports all reported
experimentally achieve a 10% CV at their 99th percentile relationships within the 2 years previous to this publica-
limit. Improved 2nd-generation assays, however, have tion that may be relevant to this guidelines document. A
recently demonstrated 10% CVs at the 99th percentile document of those relationships may be found in the
(21 ). The ultimate goal will be to have all cTn assays attain online Data Supplement at http://www.clinchem.org/
a 10% CV at the 99th percentile reference limit to reduce content/vol53/issue4.
any potential of false-positive analytic results attributable
to imprecision in the low concentration range. References
For clinical trials, to avoid the confusion of multiple 1. Wu AHB, Apple FS, Gibler WB, Jesse RL, Warshaw MM, Valdes R
centers using multiple assays, several approaches are Jr. National Academy of Clinical Biochemistry Standards of Labo-
recommended for cTn testing (15, 16 ). First, analyze all ratory Practice: recommendations for use of cardiac markers in
samples from trial centers in a core, central lab with a coronary artery diseases. Clin Chem 1999;45:1104 –21.
precise, well-defined assay. Second, provide all trial cen- 2. Panteghini M, Gerhardt W, Apple FS, Dati F, Ravkilde J, Wu AH.
Quality specifications for cardiac troponin assays. Clin Chem Lab
ters with the same well-defined, FDA-cleared assay.
Med 2001;39:174 – 8.
Third, uniformly define each center’s assays by using the
3. National Institute of Standards and Technology. Certificate of
99th percentile concentration (assay-dependent), thus reanalysis. Standard reference material 2921: human cardiac tro-
ducing reliance on local laboratory criteria for cTn deci- ponin complex. Gaithersburg, MD: NIST, 2004.
sion cutoffs. Fourth, use a multiple (2- to 3-fold) of the 4. Christenson RH, Duh SH, Apple FS, Bodor GS, Bunk DM, Dalluge
99th percentile. Fifth, if trials decide to use cutoff values J, et al. Standardization of cardiac troponin I assays: round robin
performance of ten candidate reference materials. Clin Chem 15. Newby LK, Alpert JS, Ohman EM, Thygesen K, Califf RM. Changing
2001;47:431–7. the diagnosis of acute myocardial infarction: implications for
5. Christenson RH, Duh SH, Apple FS, Bodor GS, Bunk DM, Pan- practice and clinical investigations. Am Heart J 2002;144:957–
teghini M, et al. Toward standardization of cardiac troponin I 80.
measurements. Part II: Assessing commutability of candidate 16. Apple FS, Wu AHB, Jaffe AS. European Society of Cardiology and
reference materials and harmonization of cardiac troponin I as- American College of Cardiology guidelines for redefinition of
says. Clin Chem 2006;52:1685–92. myocardial infarction: how to use existing assays clinically and for
6. Jaffe AS, Babuin L, Apple FS. Biomarkers in acute coronary clinical trials. Am Heart J 2002;144:981– 6.
disease: the present and the future. J Am Coll Cardiol 2006;48: 17. Clinical and Laboratory Standards Institute. How to define and
1–11. determine reference intervals in the clinical laboratory; approved
7. Panteghini M. Standardization of cardiac troponin I measure- guideline, 2nd ed. CLSI document C28 –A2. 2000.
ments: The way forward? Clin Chem 2005;51:1594 –7. 18. Apple FS, Quist HE, Doyle PJ, Otto AP, Murakami MM. Plasma
99th percentile reference limits for cardiac troponin and creatine
8. Clinical and Laboratory Standards Institute. Protocols for determi-
kinase MB mass for use with European Society of Cardiology/
nation of limits of detection and limits of quantitation; Approved
American College of Cardiology consensus recommendations.
guideline. CLSI document EP17-A. 2004.
Clin Chem 2003;49:1331– 6.
9. Clinical and Laboratory Standards Institute. Evaluation of preci-
19. Apple FS, Parvin CA, Buechler KF, Christenson RH, Wu AHB, Jaffe
sion performance of quantitative measurement methods; ap-
AS. Validation of the 99th percentile cutoff independent of assay
proved guideline, 2nd ed. CLSI document EP5–A2. 2004.
imprecision (%CV) for cardiac troponin monitoring for ruling out
10. Novis DA, Jones BA, Dale JC, Walsh MK. Biochemical markers of myocardial infarction. Clin Chem 2005;51:2198 –200.
myocardial injury test turnaround time: a College of American 20. Panteghini M, Pagani F, Yeo KT, Apple FS, Christenson RH, Dati F,
Pathologists Q-probe Study. Arch Path Lab Med 2004;128:158 – et al. Evaluation of the imprecision at low range concentrations of
64. the assays for cardiac troponin determination. Clin Chem 2004;
11. Apple FS, Chung AY, Kogut ME, Bubany S, Murakami MM. 50:327–32.
Decreased patient charges following implementation of point-of- 21. Christenson RH, Cervelli DR, Bauer RS, Gordon M. Stratus CS
care cardiac troponin monitoring in acute coronary syndrome cardiac troponin I method: performance characteristics including
patients in a community hospital cardiology unity. Clin Chim Acta imprecision at low concentrations. Clin Biochem 2004;37:679 –
2006;370:191–5. 83.
12. Luepker RV, Apple FS, Christenson RH, Crow RS, Fortmann SP, 22. Braunwald E, Antman EM, Beasley JW, Califf RM, Cheitlin MD,
Goff D, et al. Case definitions for acute coronary heart disease in Hochman JS, et al. American College of Cardiology/American
epidemiology and clinical research studies. Circulation 2003;108: Heart Association Task Force on practice guidelines (Committee
2543–9. on the Management of Patients With Unstable Angina). ACC/AHA
13. Jaffe AS, Ravkilde J, Roberts R, Naslund U, Apple FS, Galvani M, guideline update for the management of patients with unstable
Katus H. It’s time for a change to a troponin standard. Circulation angina and non-ST-segment elevation myocardial infarction—
2000;102:1216 –20. 2002: summary article: a report of the American College of
14. Alpert JS, Thygesen K, Antman E, Bassand JP. Myocardial infarc- Cardiology/American Heart Association Task Force on Practice
tion redefined: a consensus document of the Joint European Guidelines (Committee on the Management of Patients With
Society of Cardiology/American College of Cardiology Committee Unstable Angina). Circulation 2002;106:1893–2000.
for the redefinition of myocardial infarction. J Am Coll Cardiol 23. Jaffe AS. Elevations in cardiac troponin measurements: false
2000;36:959 – 69. false-positive. Cardiovasc Tox 2001;1:87–92.
National Academy of Clinical Biochemistry

Laboratory Medicine Practice Guidelines: Clinical
Characteristics and Utilization of Biochemical
Markers in Acute Coronary Syndromes
NACB WRITING GROUP MEMBERS
David A. Morrow,1 Christopher P. Cannon,1 Robert L. Jesse,2 L. Kristin Newby,3
Jan Ravkilde,4 Alan B. Storrow,5 Alan H.B. Wu,6 and Robert H. Christenson7*
NACB COMMITTEE MEMBERS
Robert H. Christenson, PhD, Chair; Fred S. Apple, Minneapolis, MN; Christopher P. Cannon;
Gary Francis, Cleveland, OH; Robert L. Jesse; David A. Morrow; L. Kristin Newby;
Jan Ravkilde; Alan B. Storrow; Wilson Tang, Cleveland, OH; and Alan H.B. Wu
I. OVERVIEW OF THE ACUTE CORONARY SYN- b. Relationship to Clinical Outcomes .......558

DROME........................................................................553 c. Decision-Limits.........................................559
A. Definition of Terms .............................................553 d. Therapeutic Decision-Making................559
B. Pathogenesis and Management of ACS ..........553 2. Natriuretic Peptides ......................................559
II. USE OF BIOCHEMICAL MARKERS IN THE INI- a. Pathophysiology ......................................559
TIAL EVALUATION OF ACS.................................554 b. Relationship to Clinical Outcomes .......559
A. Diagnosis of Myocardial Infarction..................554 c. Decision-Limits.........................................561
1. Biochemical Markers of Myocardial Necro- d. Therapeutic Decision-Making................561
sis......................................................................554 3. Biochemical Markers of Inflammation ......562
2. Optimal Timing of Sample Acquisition ....556 a. Pathophysiology ......................................562
3. Criteria for Diagnosis of MI ........................556 b. Relationship to Clinical Outcomes .......562
4. Additional Considerations in the Use of Bio- c. Decision-Limits.........................................563
markers for Diagnosis of MI .......................557 d. Therapeutic Decision-Making................564
B. Early Risk Stratification ......................................557 4. Biochemical Markers of Ischemia ...............564
1. Biochemical Markers of Cardiac Injury.....558 5. Multimarker Approach.................................565
a. Pathophysiology ......................................558 6. Other Novel Markers....................................565
III. USE OF BIOCHEMICAL MARKERS IN THE MAN-
1
Brigham and Women’s Hospital, Harvard University, Boston, MA.
2
AGEMENT OF NSTEACS ........................................565
Medical College of Virginia, Richmond, VA.
3
Duke University Medical Center, Durham, NC. A. Clinical Decision-Making ...................................565
4
Aarhus University Hospital, Aarhus, Denmark. 1. Biochemical Markers of Cardiac Injury.....565
5
Vanderbilt University, Nashville, TN.
6
2. Other Biochemical Markers .........................567
University of California at San Francisco, San Francisco, CA.
7
University of Maryland School of Medicine, Baltimore, MD. B. Biochemical Marker Measurement After the Ini-
All relationships with industry for the guidelines committee are reported tial Diagnosis ........................................................567
online at ⬍http://www.aacc.org/AACC/members/nacb/LMPG/OnlineGuide/
PublishedGuidelines/ACSHeart/heartpdf.htm⬎. IV. USE OF BIOCHEMICAL MARKERS IN THE MAN-
The materials in this publication represent the opinions of the authors and AGEMENT OF STEMI ..............................................567
committee members, and do not represent the official position of the National
Academy of Clinical Biochemistry (NACB). The National Academy of Clinical
A. Noninvasive Assessment of Reperfusion........567
Biochemistry is the academy of the American Association for Clinical Chemistry. B. Biochemical Marker Measurement After the Di-
*Address correspondence to this author at: Director, Rapid Response agnosis of Acute MI ............................................568
Laboratories, University of Maryland School of Medicine, 22 S. Greene St.,
Baltimore, MD 21201. Fax 410-328-5880; e-mail rchristenson@umm.edu. V. REFERENCES .............................................................568
552
I. Overview of the Acute Coronary Syndrome

a. definition of terms
Acute coronary syndrome (ACS)8 refers to a constellation of
clinical symptoms caused by acute myocardial ischemia
(1, 2 ). Owing to their higher risk for cardiac death or
ischemic complications, patients with ACS must be identi-
fied among the estimated 8 million patients with nontrau-
matic chest symptoms presenting for emergency evaluation
each year in the US (3 ). In practice, the terms suspected or
possible ACS are often used by medical personnel early in the
process of evaluation to describe patients for whom the
Fig. 1. Categorization of acute coronary syndromes.
symptom complex is consistent with ACS but the diagnosis
has not yet been conclusively established (1 ).
Patients with ACS are subdivided into 2 major catego-
ries based on the 12-lead electrocardiogram (ECG) at the majority of patients presenting with ACS, the throm-
presentation (Fig. 1): those with new ST-segment eleva- bus is partially obstructive, or only transiently occlusive,
tion on the ECG that is diagnostic of acute ST-elevation resulting in coronary ischemia without persistent ST-
myocardial infarction (STEMI) and those who present segment elevation (unstable angina or NSTEMI). In the
with ST-segment depression, T-wave changes, or no ECG remaining ⬃30% of patients with ACS (9 ), the intracoro-
abnormalities (non–ST elevation ACS, NSTEACS). The nary thrombus completely occludes the culprit vessel,
latter term (NSTEACS) encompasses both unstable angina resulting in STEMI. Antithrombotic and antiplatelet ther-
and non–ST elevation myocardial infarction (NSTEMI). apies aimed at halting the propagation or recurrence of
This terminology has evolved along clinical lines based on coronary thrombus are central to management of the
a major divergence in the therapeutic approach to STEMI majority of patients across the entire spectrum of ACS
vs NSTEACS (see section IB). Unstable angina and (1, 2, 10 ). The subgroup of patients with STEMI consists of
NSTEMI are considered to be closely related conditions, candidates for immediate reperfusion therapy with either
sharing a common pathogenesis and clinical presentation fibrinolysis or percutaneous coronary intervention (10 ). In
but differing in severity (1 ). Specifically, NSTEMI is contrast, fibrinolysis appears to be harmful in patients
distinguished from unstable angina by ischemia suffi- with NSTEACS (1, 11 ).
ciently severe in intensity and duration to cause irrevers- Including the most common etiology of ACS de-
ible myocardial damage (myocyte necrosis), recognized scribed above, 5 principal causes have been described:
clinically by the detection of biomarkers of myocardial (1) plaque rupture with acute thrombosis; (2) progressive
injury (4 ). mechanical obstruction; (3) inflammation; (4) secondary
unstable angina (e.g., due to severe anemia or hyperthy-
b. pathogenesis and management roidism); and (5) dynamic obstruction (coronary vasocon-
It is important to recognize that ACS is a complex striction) (12 ). It is rare that any of these contributors
syndrome with a heterogeneous etiology, analogous to exists in isolation. Because patients with ACS vary sub-
anemia or hypertension (5 ). Nevertheless, the most com- stantially with respect to the mixture of contributions
mon cause is atherosclerotic coronary artery disease with from each of these major mechanisms, and, as such, are
erosion or rupture of atherosclerotic plaque, exposing the likely to benefit from different therapeutic approaches,
highly procoagulant contents of the atheroma core to characterization of the dominant contributors for an indi-
circulating platelets and coagulation proteins, and culmi- vidual patient can be valuable in guiding their care (12 ).
nating in formation of intracoronary thrombus (6 – 8 ). In With the emergence of newer biomarkers that reflect the
diverse pathobiology of acute ischemic heart disease, their
use as noninvasive means to gain insight into the under-
lying causes and consequences of ACS is being investi-
Received December 27, 2007; accepted December 27, 2007. gated (13 ).
This article has been copublished in the April 3, 2007 online issue of
Circulation.
Commensurate with the heterogeneous pathobiology
Previously published online at DOI: 10.1373/clinchem.2006.084194 of ACS, the risk of subsequent death and/or recurrent
© 2007 American Association for Clinical Chemistry and the American ischemic events also varies widely. As a result, effective
Heart Association, Inc. risk stratification and targeting of therapy is a focus of
8
Nonstandard abbreviations: ACS, acute coronary syndrome; ECG, elec-
trocardiogram; STEMI, ST-elevation myocardial infarction, NSTEACS, non-ST
contemporary clinical management of this condition
elevation ACS; NSTEMI, non-ST elevation myocardial infarction; MI, myocar- (14, 15 ). In addition, among patients with definite ACS,
dial infarction; CK-MB, creatine kinase MB; cTnI, cardiac troponin I; cTnT, early treatment may reduce the extent of myocardial
cardiac troponin T; hs-CRP, high-sensitivity C-reactive protein; BNP, B-type
injury; therefore, rapid diagnosis and initiation of therapy
natriuretic peptide; NT-proBNP, N-terminal pro-BNP; LLD, lower limits of
detectability; CRP, C-reactive protein; IL-6, interleukin-6; IMA, ischemia is also a central tenet of management (1 ). It follows that
modified albumin; and GP, glycoprotein. the objectives of the initial evaluation of patients with
554 Morrow et al.: NACB Practice Guidelines for Biomarkers in ACS
nontraumatic chest pain are 2-fold: (a) to assess the

2. A rapid “rule-in” protocol with frequent early
probability that the patient’s symptoms are related to
sampling of markers of myocardial necrosis
acute coronary ischemia; and (b) to assess the patient’s
maybe appropriate if tied to therapeutic strategies
risk of recurrent cardiac events, including death and
(Level of Evidence: C).
recurrent ischemia (1 ). When applied in conjunction with
the clinical history, physical examination, and interpreta-
class iii
tion of the ECG, cardiac biomarkers are valuable in
achieving both of these objectives. 1. Total CK, CK-MB activity, aspartate aminotrans-
ferase (AST, SGOT), ␤-hydroxybutyric dehydroge-
II. Use of Biochemical Markers in the Initial Evaluation of ACS nase, and/or lactate dehydrogenase should not be
A. diagnosis of myocardial infarction used as biomarkers for the diagnosis of MI (Level
of Evidence: C).
2. For patients with diagnostic ECG abnormalities on
recommendations for use of biochemical
presentation (e.g., new ST-segment elevation), di-
markers for diagnosis of myocardial
agnosis and treatment should not be delayed
infarction (mi)
while awaiting biomarker results (Level of Evi-
class i
dence: C).
1. Biomarkers of myocardial necrosis should be mea-
sured in all patients who present with symptoms
consistent with ACS (Level of Evidence: C).
2. The patient’s clinical presentation (history, physi- 1. biochemical markers of myocardial necrosis
cal exam) and ECG should be used in conjunction Myocardial necrosis is accompanied by the release of
with biomarkers in the diagnostic evaluation of structural proteins and other intracellular macromole-
suspected MI (Level of Evidence: C). cules into the cardiac interstitium as a consequence of
3. Cardiac troponin is the preferred marker for the compromise of the integrity of cellular membranes. These
diagnosis of MI. Creatine kinase MB (CK-MB) by biomarkers of myocardial necrosis include cardiac tropo-
mass assay is an acceptable alternative when cardiac nin I and T (cTnI and cTnT), CK, myoglobin, lactate
troponin is not available (Level of Evidence: A). dehydrogenase, and others (Table 1). On the basis of
4. Blood should be obtained for testing at hospital improved sensitivity and superior tissue-specificity com-
presentation followed by serial sampling with pared with the other available biomarkers of necrosis,
timing of sampling based on the clinical circum- cardiac troponin is the preferred biomarker for the detec-
stances. For most patients, blood should be ob- tion of myocardial injury. The diagnosis of acute, evolv-
tained for testing at hospital presentation and at ing, or recent MI requires (in the absence of pathologic
6 –9 h (Level of Evidence: C). confirmation) findings of a typical rise and/or fall of a
5. In the presence of a clinical history suggestive of biomarker of necrosis, in conjunction with clinical evidence
ACS, the following are considered indicative of (symptoms, or ECG) that the cause of myocardial damage is
myocardial necrosis consistent with MI (Level of ischemia. Because recognition of acute MI is important to
Evidence: C): prognosis and therapy, measurement of biomarkers of
a. Maximal concentration of cardiac troponin ex- necrosis is indicated in all patients with suspected ACS.
ceeding the 99th percentile of values (with opti- Important characteristics of these biomarkers are dis-
mal precision defined by total CV ⬍10%) for a cussed in the remainder of this section.
reference control group on at least 1 occasion In contrast to CK, cTnI and cTnT have isoforms that are
during the first 24 h after the clinical event (ob- unique to cardiac myocytes and may be measured by
servation of a rise and/or fall in values is useful in assays employing monoclonal antibodies specific to
discriminating the timing of injury). epitopes of the cardiac form (16 –19 ). The advantage of
b. Maximal concentration of CK-MB exceeding cardiac troponin over other biomarkers of necrosis has
the 99th percentile of values for a sex-specific been firmly established in clinical studies. Testing for
reference control group on 2 successive samples cardiac troponin is associated with fewer false-positive
(values for CK-MB should rise and/or fall). results in the setting of concomitant skeletal muscle
injury, e.g., after trauma or surgery (16, 20, 21 ) and also
class iib provides superior discrimination of myocardial injury
when the concentration of CK-MB is normal or minimally
1. For patients who present within 6 h of the onset of increased (16, 22, 23 ). Moreover, the association between
symptoms, an early marker of myocardial necrosis an increased concentration of cardiac troponin and a
may be considered in addition to a cardiac tropo- higher risk of recurrent cardiac events in patients with
nin. Myoglobin is the most extensively studied normal serum concentration of CK-MB and suspected
marker for this purpose (Level of Evidence: B). ACS has confirmed the clinical relevance of detecting
Table 1. Properties of biomarkers of myocardial necrosis.

Biochemical Molecular weight, Duration of
marker g/mole Cardiac specific? Advantage Disadvantage elevation
Myoglobin 18 000 No High sensitivity and negative Low specificity in presence of 12–24 h
predictive value. skeletal muscle injury and
Useful for early detection of MI renal insufficiency.
and reperfusion. Rapid clearance after
necrosis.
h-FABP 15 000 ⫹ Early detection of MI Low specificity in presence of 18–30 h
skeletal muscle injury and
renal insufficiency.
CK-MB, mass 85 000 ⫹⫹⫹ Ability to detect reinfarction. Lowered specificity in 24–36 h
assays Large clinical experience. Previous skeletal muscle injury.
gold standard for myocardial
necrosis
CK-MB isoforms 85 000 ⫹⫹⫹ Early detection of MI Lack of availability/ 18–30 h
experience
cTnT 37 000 ⫹⫹⫹⫹ Tool for risk stratification. Not an early marker of 10–14
Detection of MI up to 2 weeks. myocardial necrosis. days
High specificity for cardiac tissue Serial testing needed to
discriminate early
reinfarction.
cTnI 23 500 ⫹⫹⫹⫹ Tool for risk stratification. Not an early marker of 4–7 days
Detection of MI up to 7 days. myocardial necrosis.
High specificity for cardiac tissue Serial testing needed to
discriminate early
reinfarction.
No analytical reference
standards.
Time of first increase for the markers are 1–3 h for myoglobin, 3– 4 h for CK-MB mass, 3– 4 h for cTnT, and 4 – 6 h for cTnI. h-FABP, heart fatty acid– binding protein.
Adapted from Christenson RH and Azzazy HME. Biomarkers of necrosis: past, present and future. In Morrow DA, ed. Cardiovascular Biomarkers: Pathophysiology and
Clinical Management. New York: Humana Press, 2006.
circulating troponin in patients previously classified with tion in skeletal muscle. By virtue of its greater concentra-
unstable angina. An example from one of several studies tion in cardiac vs skeletal myocytes, the MB isoenzyme of
is shown in Fig. 2 (24 –26 ). CK offers an improvement in sensitivity and specificity
When cardiac troponin is not available, the next best compared with total CK. Nevertheless, CK-MB constitutes
alternative is CK-MB (measured by mass assay). Al- 1%–3% of the CK in skeletal muscle, and is present in
though total CK is a sensitive marker of myocardial minor quantities in intestine, diaphragm, uterus, and
damage, it has poor specificity due to its high concentra- prostate. Therefore, the specificity of CK-MB may be
impaired in the setting of major injury to these organs,
especially skeletal muscle. Serial measurements docu-
menting the characteristic rise and/or fall are important
to maintaining specificity for the diagnosis of acute MI.
Alternatives to cardiac injury should be sought when
CK-MB is increased in the presence of a troponin concen-
tration below the 99th percentile. Assays for CK-MB mass
offer superior analytical and diagnostic performance and
thus are strongly preferred to assays for CK-MB activity
(see Analytical Issues for Biomarkers in ACS in separate
guidelines).
Although they are of historical significance, total CK,
lactate dehydrogenase, and aspartate aminotransferase
should no longer be used for the diagnosis of MI because
Fig. 2. Risk of death and recurrent ischemic events among patients they have low specificity for cardiac injury and more
with NSTEACS and normal serial CK-MB with and without increase specific alternative biomarkers of necrosis are available.
baseline concentration of cardiac troponin I (Dimension RxL, Dade Myoglobin shares limitations with these markers due to
Behring). its high concentration in skeletal muscle. However, be-
As discussed in section II-B1.c, the cut point applied in this study is specific to
the assay used. Data from Morrow et al. (67 ). UR, urgent revascularization
cause of its small molecular size and consequent rapid rise
prompted by recurrent ischemia. in the setting of myocardial necrosis, it has retained value
as a very early marker of MI. Clinical studies have shown often clinically very challenging (10 ). Therefore, for most
that the combined use of myoglobin and a more specific patients, blood should be obtained for testing at hospital
marker of myocardial necrosis (cardiac troponin or CK- presentation and at 6 –9 h after presentation (unless the
MB) may be useful for the early exclusion of MI (27, 28 ). timing of symptoms is reliably known) to provide ade-
Multimarker strategies that include myoglobin have been quate clinical sensitivity for detecting MI. Given improve-
shown to identify patients with MI more rapidly than ments in the analytic performance of troponin assays,
laboratory-based determination of a single marker testing up to 6 –9 h after symptom onset is expected to
(29, 30 ). However, this potential advantage of myoglobin deliver optimal sensitivity in most patients. However, in
may be diminished with use of contemporary decision- patients for whom these initial samples are negative and
limits and improving sensitivity of newer troponin assays there is an intermediate or high clinical index of suspicion,
(31 ). CK-MB subforms may also be used as an early rising or in whom plausibly ischemic symptoms have recurred,
indicator of MI (32 ) but are not used today, as there are no repeat testing at 12–24 h should be considered. Among
commercial platforms available. patients without ST elevation, such serial testing increases
the proportion of patients with myocardial injury who are
2. optimal timing of sample acquisition detected from 49% to 68% at 8 h and enhances the
The optimal timing of sample acquisition for measure- accuracy of risk assessment (37 ). More frequent early
ment of biomarkers for the diagnosis of MI derives from testing of cardiac troponin and/or CK-MB, particularly in
both properties of the available biomarkers and patient- combination with myoglobin, may be considered as an
related factors (timing and duration of symptoms relative approach to increase early detection of infarction and to
to presentation and overall probability of ACS). CK-MB facilitate rapid initiation of treatment (38, 39 ). This strat-
begins to rise within 3– 4 h after the onset of myocardial egy has also shown value in some studies for expedited
injury and falls to normal ranges by 48 –72 h (Fig. 3). exclusion of MI (40 ), as has use of the change in markers
Cardiac troponin rises with a time course similar to of necrosis repeated over an interval of 2 h (41, 42 ).
CK-MB but can remain increased for up to 4 –7 days for
cTnI and 10 –14 days for cTnT. The initial release of 3. criteria for diagnosis of mi
cardiac troponin that exists in the cellular cytosol (3%– Detection of increased blood concentrations of biomarkers
8%) followed by the slower dispersion of troponin from of myocardial necrosis in the setting of a clinical syn-
degrading cardiac myofilaments is responsible for this drome consistent with myocardial ischemia is necessary
extended kinetic profile (33 ). In contrast, myoglobin con- for the diagnosis of acute, evolving, or recent MI. Clinical
centration begins to rise as early as 1 h after onset of information from the history and ECG must be integrated
myocyte damage and returns to normal within 12–24 h. with data from measurement of biomarkers in determin-
By virtue of these kinetics, the temporal rise of the ing whether the myocardial necrosis manifested by in-
serum concentration of CK-MB and cardiac troponin creased concentration of these markers is due to myocar-
typically does not permit detection of myocardial necrosis dial ischemia or some other cause (4, 43 ). The tissue
very early (1–3 h) and does not support maximal sensi- specificity of cardiac troponin should not be confused
tivity of these markers until 6 or more hours after the with specificity for the mechanism of injury (e.g., MI vs
onset of MI (34 –36 ). Accurate determination of the timing myocarditis) (44, 45 ). When an increased value is encoun-
of symptom onset is based on patient reporting and is tered in the absence of evidence of myocardial ischemia, a
careful search for other possible etiologies of cardiac
damage should be undertaken.
An increased concentration of cardiac troponin is de-
fined as exceeding the 99th percentile of a reference
control group. Recommendations regarding analytic eval-
uation and performance are described in separate guide-
lines (see Analytical Issues for Biomarkers in ACS). A max-
imal concentration of cardiac troponin exceeding this
decision-limit on at least 1 occasion during the index
clinical event is indicative of myocardial necrosis. Simi-
larly, the diagnostic limit for CK-MB is defined as the 99th
percentile (with acceptable imprecision) in a sex-specific
reference control group. In light of the lower tissue
specificity compared with troponin, it is recommended
that in most situations 2 consecutive measurements of
CK-MB above this decision-limit are required to be con-
Fig. 3. Temporal release of myoglobin, CK-MB, cTnI and cTnT. sidered sufficient biochemical evidence of myocardial
With permission, from Christenson RH, Azzazy HME. Biomarkers of necrosis:
past, present and future. In Morrow DA, ed. Cardiovascular Biomarkers: Patho-
necrosis. Use of total CK for diagnosis of MI is not
physiology and Clinical Management. New York: Humana Press, 2006. recommended. However, in the absence of availability of data
using a troponin or CK-MB assay (mass or activity), when tools remain vital to clinical care. In particular, acute
only total CK values are available, the recommended ST-segment elevation on the ECG in conjunction with a
decision-limit is ⬎2 times the sex-specific upper reference consistent clinical syndrome has a very high positive
limit. A rise and/or fall of CK-MB or total CK provides predictive value for acute STEMI and should prompt
additional evidence supporting the diagnosis of acute MI. initiation of appropriate strategies for coronary reper-
In addition, for values of cardiac troponin between the fusion (10 ). Patients presenting within 6 h of symptom
10% CV and the 99th percentile, as well as for potential onset may not yet have a detectable serum concen-
chronic elevations (e.g., renal failure), the use of a rising tration of biomarkers of necrosis. However, given the
and/or falling pattern is often useful in facilitating the critical relationship between rapid therapy and out-
discrimination of patients with acute events. comes in patients with STEMI, therapy should not be
delayed waiting for confirmatory biomarker measurements.
4. additional considerations in the use of
biomarkers for diagnosis of mi
The criteria for MI recommended in these and other
guidelines (4 ) are based on the principle that any reliably
detected myocardial necrosis, if caused by myocardial b. early risk stratification
ischemia, constitutes an MI. The development of more recommendations for use of biochemical
sensitive and specific biomarkers of necrosis, such as markers for risk stratification in acs
cardiac troponin, has enabled detection of quantitatively
much smaller areas of myocardial injury (46 ). Moreover,
it is likely that future generations of assays for cardiac class i
troponin will push this limit even lower. Elegant histo- 1. Patients with suspected ACS should undergo
logic work in animal models of coronary ischemia has early risk stratification based on an integrated
provided strong evidence that release of CK from cardiac assessment of symptoms, physical exam findings,
myocytes occurs in setting of myocyte necrosis but not in ECG findings, and biomarkers (Level of Evidence:
the setting of reversible myocyte injury. In contrast, data C).
in this regard for cardiac troponin have been mixed (47 ). 2. A cardiac troponin is the preferred marker for risk
Increased concentrations of cTnI and cTnT have been stratification and, if available, should be measured
observed in animal models of ischemia without histologic in all patients with suspected ACS. In patients
evidence of irreversible cellular injury (48 ). Whereas the with a clinical syndrome consistent with ACS, a
potential to miss small amounts of patchy necrosis during maximal (peak) concentration exceeding the 99th
microscopic examination is a significant limitation of all percentile of values for a reference control group
such experimental results, it is also possible that such should be considered indicative of increased risk
release of cardiac troponin into the circulation may result of death and recurrent ischemic events (Level of
from reversible injury to the myocyte cellular membrane Evidence: A).
leading to egress of troponin residing in the cytosol (49 ). 3. Blood should be obtained for testing on hospital
Nevertheless, based on the aggregate evidence to date, the presentation followed by serial sampling with
present guidelines reflect the prevailing consensus opin- timing of sampling based on the clinical circum-
ion (43 ) that any reliably detected elevation of a cardiac stances. For most patients, blood should be ob-
troponin is abnormal and most likely represents necrosis. tained for testing at hospital presentation and at
The committee supports additional investigation to deter- 6 –9 h (Level of Evidence: B).
mine whether current or future generations of assays for
cardiac troponin may detect release of the protein that class iia
occurs during reversible injury due to ischemia without
1. Measurement of high-sensitivity C-reactive pro-
infarction.
tein (hs-CRP) may be useful, in addition to a
Measurement of more than 1 specific biomarker of
cardiac troponin, for risk assessment in patients
myocardial necrosis (e.g., cardiac troponin and CK-MB)
with a clinical syndrome consistent with ACS. The
is not necessary for establishing the diagnosis of myo-
benefits of therapy based on this strategy remain
cardial infarction and is not recommended. The use of
uncertain (Level of Evidence: A).
serial measurements of CK-MB to provide information
2. Measurement of brain-type (B-type) natriuretic
during the management of MI after diagnosis is discussed
peptide (BNP) or N-terminal pro-BNP (NT-
in Section IV-B. Determination of an early marker of
proBNP) may be useful, in addition to a cardiac
necrosis in combination with cardiac troponin may be
troponin, for risk assessment in patients with a
appropriate in some circumstances as described in Section
clinical syndrome consistent with ACS. The bene-
II-A1.
fits of therapy based on this strategy remain un-
Despite the central role for biomarkers of necrosis in
certain (Level of Evidence: A).
establishing the diagnosis of acute MI, other diagnostic
class iib b. Relationship to clinical outcomes

The presence of myocardial necrosis detectable with cre-
1. Measurement of markers of myocardial ischemia,
atine kinase is established as an important prognostic
in addition to cardiac troponin and ECG, may aid
factor in the assessment of patients with ACS (56 ). In
in excluding ACS in patients with a low clinical
addition, the blood concentration of biomarkers of necro-
probability of myocardial ischemia (Level of Evi-
sis shows a consistent graded relationship with the risk of
dence: C).
short- and long-term mortality (57, 58 ). Specifically,
2. A multimarker strategy that includes measure-
among patients with NSTEACS, the concentration of
ment of 2 or more pathobiologically diverse bi-
CK-MB at hospital presentation establishes a gradient of
omarkers in addition to a cardiac troponin may aid
30-day mortality risk from 1.8% in patients with CK-MB
in enhancing risk stratification in patients with a
less than the upper limit of the reference interval to 3.3%
clinical syndrome consistent with ACS. BNP and
for those with a 1- to 2-fold increase above the upper limit
hs-CRP are the biomarkers best studied using this
of the reference interval, to 8.3% among those with
approach. The benefits of therapy based on this
⬎10-fold increase (58 ). The availability of cardiac tropo-
strategy remain uncertain (Level of Evidence: C).
nin has extended the spectrum of detectable myocardial
3. Early repeat sampling of cardiac troponin (e.g.,
injury and further enhanced the clinician’s ability to
2– 4 h after presentation) may be appropriate if
assess risk (24 ). Based on evidence from more than 26
tied to therapeutic strategies (Level of Evidence:
studies, including both clinical trials and observational
C).
studies from community-based cohorts, cardiac troponin
class iii has proven to be a potent independent indicator of the
Biomarkers of necrosis should not be used for routine risk of death and recurrent ischemic events among pa-
screening of patients with low clinical probability of tients presenting with ACS (26 ). In aggregate, the avail-
ACS (Level of Evidence: C). able data indicate an ⬃4-fold higher risk of death and
recurrent MI among patients presenting with suspected
NSTEACS and an increased concentration of troponin
compared with patients with a normal troponin result
(Fig. 4) (26, 59, 60 ). In patients with STEMI, an increased
1. biochemical markers of cardiac injury concentration of troponin at presentation is also associ-
a. Pathophysiology ated with significantly higher short-term mortality
The presence of cardiac troponin in the peripheral circu- (61, 62 ).
lation is indicative of myocardial injury (see Section The prognostic information obtained from measure-
II-A1). Additional pathophysiologic correlates of troponin ment of cardiac troponin is independent of and comple-
elevation have been identified in clinical studies of ACS. mentary to other important clinical indicators of risk
Angiographic data from trials enrolling patients with including patient age, ST deviation, and presence of heart
NSTEACS have shown increased concentrations of tro- failure (57, 61, 63– 66 ). The higher risk of patients present-
ponin to be associated with greater lesion complexity ing with an increased concentration of troponin is also
and severity, more frequent visible thrombus, and more
severely impaired blood flow in the culprit artery (50 –53 ).
In addition, an increased concentration of troponin is
associated with impaired myocardial tissue or “micro-
vascular” perfusion and thus hypothesized to reflect
embolization of platelet aggregates into the distal coro-
nary artery (52 ). Furthermore, increased concentrations
of troponin have been associated with a higher likelihood
of poor outcomes during angioplasty, including very slow
flow (so-called “no reflow”) despite a patent epicardial
artery in a clinical syndrome believed to result from
distal microvascular obstruction (54 ). Advances in the
understanding of the pathobiology of ACS have pointed
toward these phenomena of microembolization and mi-
crovascular obstruction as important mediators of ad- Fig. 4. Risk of death or MI stratified by troponin result in patients with
verse outcomes (55 ). As such, the apparent link between suspected ACS.
Adapted with permission from Braunwald E, et al. American College of Cardiol-
microembolization and release of cardiac troponin may ogy/American Heart Association guidelines for the management of patients with
underlie, at least in part, the strong association between unstable angina and non–ST-segment elevation myocardial infarction: a report of
the American College of Cardiology/American Heart Association Task Force on
this biomarker and subsequent recurrent clinical events Practice Guidelines (Committee on the Management of Patients With Unstable
(52 ). Angina). J Am Coll Cardiol 2000;36:970 –1062.
evident among patients with normal concentrations of 10%CV to ⬍ manufacturer’s suggested diagnostic limit
CK-MB (67 ). As such, cardiac troponin is the preferred for MI), and high (ⱖ suggested diagnostic limit for
biomarker for risk assessment in patients presenting with MI)—revealing a 6-month mortality rate that increased in
suspected ACS. cTnI and cTnT appear to have similar a stepwise fashion compared with patients with negative
value for risk assessment in ACS (26, 68 ). cTnI results [hazard ratio 2.5; 95% confidence interval (CI)
1.4 – 4.4] in the low cTnI group, 3.9 (95% CI 2.3– 6.8) in the
c. Decision-limits intermediate cTnI group, and 6.1 (95% CI 4.2– 8.7) in the
As the lower limits of detection (LLD) have decreased high cTnI group (Fig. 5) (72 ). With future improvements
with incremental improvements in commercially avail- in the analytic performance of available assays, the asso-
able assays for cardiac troponin, the potential prognostic ciation between troponin concentrations at the lower limit
implications of quantitatively modest (“low-level”) in- of detection and outcomes in ACS will require continued
creases in cardiac troponin have attained greater clinical careful evaluation.
relevance. The consensus recommendation is that the
upper limit of normal for cardiac troponin and CK-MB be d. Therapeutic decision-making
defined by the 99th percentile among a reference control The application of cardiac troponin to guide specific
population (69 ). Details regarding the determination of therapeutic choices for patients with ACS is well studied
this cut point and analytic performance of the assay are and is discussed in section IIIA.
discussed elsewhere in these guidelines (see Analytical
Issues for Biomarkers in ACS). 2. natriuretic peptides
When conducted among patients with a compelling a. Pathophysiology
clinical history suggesting ACS (e.g., in clinical trials of BNP and NT-proBNP are released from cardiac myocytes
ACS), prospective analyses have documented that tropo- in response to increases in ventricular wall stress (73 ).
nin concentrations at the low end of the detectable range Wall stress in a chamber is directly related to the diameter
are associated with higher risk of recurrent cardiac events of the chamber and the transmural pressure and inversely
than patients without detectable troponin (66, 70 ). For
related to the thickness of the wall. Therefore, increases
example, in the Treatment with Aggrastat and Determine
both in the diameter of and pressure within the left
Cost of Therapy with an Invasive or Conservative Strat-
ventricle during remodeling after a transmural infarction,
egy (TACTICS)-TIMI 18 study, patients with a baseline
or as a consequence of prior ischemic damage, may
concentration of cTnI in the range immediately above the
contribute to elevation of natriuretic peptides observed in
99th percentile for the assay used in the study (0.1 ␮g/L,
patients with acute MI. In addition, impairment of ven-
CV 20%) were at more than 3-fold higher risk of death or
tricular relaxation and consequent nonsystolic ventricular
recurrent MI than those with cTnI ⬍0.1 ␮g/L (66 ). This
dysfunction is one of the earliest consequences of myo-
observation of the prognostic significance of low-level
cardial ischemia, preceding angina and ST-segment devi-
increase of cardiac troponin has been independently con-
firmed using another assay for cTnI in 2 separate data sets ation. This well-described pathophysiology, together with
from clinical trials (OPUS-TIMI 16 and FRISC II) (70, 71 ), a strong relationship between BNP and NT-proBNP with
as well as within a community-based study (72 ). Specifi- mortality in patients with unstable angina (see below), has
cally, in the latter, patients presenting with chest pain supported the hypothesis that myocardial ischemia can
were stratified into 4 groups according to peak cTnI also elicit the release of BNP in absence of necrosis (74 ).
concentration—negative (⬍LLD), low (ⱖLLD to ⬍99th The concept that ischemia may be an important stim-
percentile, 10%CV), intermediate (ⱖ99th percentile, ulus for BNP synthesis and release is supported by
several lines of evidence. In experimental models of
myocardial infarction, BNP gene transcription is in-
creased both in infarcted tissue and in the surrounding
ischemic but viable myocardium (75 ). Hypoxia has also
been shown to trigger release of BNP (76 ). BNP rises early
after exercise in patients with coronary disease, and the
magnitude of BNP increase is proportional to the size of
the ischemic territory as assessed with nuclear single-
photon emission computed tomography imaging (77 ).
After uncomplicated coronary angioplasty, BNP tran-
siently increases even when cardiac filling pressures re-
main unchanged (78 ). Together, these data provide a
plausible basis to explain the strong association between
Fig. 5. Prognostic implication of low-level troponin elevation in patients
with chest pain suspicious for ACS. BNP and NT-proBNP with mortality in patients with
Data from Kontos, et al. (72 ). URL, upper reference limit supplied by manufac- unstable angina and normal left ventricular systolic
turer. function.
b. Relationship to clinical outcomes ventricular remodeling (92 ). In patients with acute MI, a
In aggregate there are now more than 10 studies showing higher concentration of BNP and NT-proBNP have been
a strong association between BNP or NT-proBNP and shown to predict a greater likelihood of death or heart
outcomes in patients with ACS (Table 2) (79 – 89 ). After failure, independent of other prognostic variables includ-
presentation with transmural infarction, the plasma con- ing left ventricular ejection fraction (80, 81, 83, 93, 94 ).
centration of BNP rises rapidly and peaks at ⬃24 h, with BNP and NT-proBNP are also increased in high-risk
the peak concentration proportional to the size of the MI patients with unstable angina (83, 84, 95 ). When mea-
(90, 91 ). In some patients, particularly those who eventu- sured a median of 40 h after presentation in ⬃1600
ally develop severe heart failure, a second peak may occur patients with NSTEACS, a highly significant graded rela-
after 5 days, likely reflecting the development of adverse tionship between the concentration of BNP and subse-
Table 2. Summary of clinical studies of BNP and NT-proBNP in ACS.

Author, year Study Subjects Marker Follow-up Findings
Arakawa et al., 1996 (79 ) Observational 70 BNP 18 months RR not reported, BNP at admission
independently associated with
mortality.
Darbar et al., 1996 (183 ) Observational 75 BNP 20 months Increase in OR for death by 7.3
(1.9–10.1) per each 10 pmol/L
increase in BNP
Richards et al., 1998 (81 ) Observational 121 NT-proBNP 24 months RR 5.9 (1.8–19) associated with
BNP above vs below median
Crilley and Farrer, 2001 (184 ) Observational 133 BNP 1 year BNP higher in patients who died by
1 year (675 vs 365 pg/mL)
de Lemos et al., 2001 (83 ) Substudy of RCT 1698 BNP 10 months RR 12.5 for mortality in highest vs
(OPUS-TIMI 16) lowest quartile of BNP in
NSTEMI
RR 7.9 for mortality in highest vs
lowest quartile of BNP in
unstable angina
Jernberg et al., 2002 (86 ) Observational 755 NT-proBNP 4 years RR 26.6 for mortality in highest vs
lowest quartile of BNP
Omland et al., 2002 (87 ) Observational 405 NT-proBNP 52 months RR 5.6 for mortality with BNP
above vs below median in
NSTEMI
RR 3.0 for mortality with BNP
above vs below median in
unstable angina
Omland et al., 2002 (85 ) Substudy of RCT 681 NT-proBNP 6 weeks Higher baseline biomarker
(TIMI 11B) concentrations in patients that
died (299 pmol/L) than in
survivors (138 pmol/L)
Morrow et al., 2003 (84 ) Substudy of RCT 1676 BNP 6 months Increased risk of death at 7 days
(TACTICS-TIMI 18) (2.5% vs 0.7%) and 6 months
(8.4% vs 1.8%) in patients with
BNP ⬎80 pg/mL, no interaction
with early invasive strategy
Jernberg et al., 2003 (88 ) Substudy of RCT 775 NT-proBNP 2 years RR 4.1 for mortality in highest
(FRISC II) tertile of BNP compared to
lowest (invasive)
RR 3.5 for mortality in highest
tertile of BNP compared to
lowest (conservative)
James et al., 2003 (89 ) Substudy of RCT 6809 NT-proBNP 1 year RR 10.6 for mortality in highest vs
(GUSTO IV) lowest quartile of BNP
Richards et al., 2003 (94 ) Observational 666 BNP/NTproBNP 3 years RR 3.6 (2.5–53) and 4.9 (2.9–8.2)
for BNP above the median
among those with and without
ejection fraction ⬍40%,
respectively
Heeschen et al., 2004 (97 ) Substudy of RCT 1791 NT-proBNP 30 days RR 2.68 (1.66–4.34) for death or
(PRISM) MI at 30 days in patients with
NT-proBNP ⬎250 pg/mL
OR, odds ratio; RCT, randomized clinical trial; RR, relative risk.
quent risk of short- and long-term mortality was evident

(83 ). The rate of death increased from ⬍1% among
patients with BNP concentrations in the lowest quartile to
15% in those with a BNP concentration in the highest
quartile (P ⬍ 0.0001) (83 ). This finding has been corrobo-
rated in multiple studies of both BNP (83, 84 ) and NT-
proBNP (85, 86, 89 ), including substudies of clinical trials
and observational data from community-based cohorts
(Fig. 6).
Although the plasma concentration of BNP and NT-
proBNP in ACS is associated with older age, female sex,
renal insufficiency, left ventricular dysfunction, clinical
evidence of heart failure, presence of myocardial necrosis, Fig. 7. Mortality risk stratified by BNP concentrations over the range of
and more severe angiographic coronary artery disease, 40 –160 pg/mL (Triage, Biosite).
the prognostic relationship between the biomarkers and The odds ratios (ORs) and ␹2 statistics in the table below the chart are based on
BNP results dichotomized at the lower bound of the range. With permission from
mortality is independent of these other clinical risk indi- Morrow et al. (84 ).
cators (87, 96 ). Importantly, BNP and NT-proBNP iden-
tify patients without systolic dysfunction or signs of heart
failure who are at higher risk of death and heart failure other assays. NT-proBNP has also been evaluated in
and provide prognostic information that is complemen- clinical studies; cut points have been individually derived
tary to cardiac troponin (84, 89 ). within each study, and no specific cut point has yet
undergone separate validation in patients with ACS. The
c. Decision-limits committee encourages additional investigation prospec-
When evaluated in ACS, serum concentrations of BNP tively evaluating the optimal decision-limits for BNP and
and NT-proBNP have a graded relationship with risk for NT-proBNP in ACS, including evaluation of an approach
short- and long-term mortality (84, 89 ). As such, the that incorporates more than one decision-limit to stratify
absolute plasma concentration of BNP or NT-proBNP patients into low, intermediate, and high risk, as well as
carries information with respect to the magnitude of risk, assessment of the need for age- and sex-related decision-
and thus should be considered by the clinician. Neverthe- limits in ACS. It is possible that different decision-limits
less, for convenient clinical use, a decision-limit of 80 should be applied for risk stratification in ACS compared
pg/mL has been validated in patients with high clinical with diagnostic assessment of the patient with shortness
suspicion for ACS using 2 BNP assays and may be used of breath, and that the prognostic decision-limits in ACS
for assays that are similarly calibrated (Fig. 7) (84 ). An will be refined when studied in more heterogeneous
evidence-based approach with specific assays studied and patient populations presenting with suspected ACS. A
validated in clinical studies is thus possible. However, detailed discussion of analytic issues that may impact the
results for specific cut points may not be extrapolated to selection and reporting of decision limits for BNP and
NT-proBNP is presented in separate guidelines (Analytic
Issues in Heart Failure Biomarkers). These, and other issues
discussed below, require additional study before routine
use of BNP and NT-proBNP for risk assessment in ACS
can be recommended.
Whether there is an optimal timing for measurement
also warrants additional investigation. When measured at
admission (86 ), ⬍24 h after symptom onset (84 ), or 2–5
days after the index event, BNP and/or NT-proBNP
maintain prognostic performance (83 ). However, the con-
centrations of natriuretic peptides change over time after
presentation and it is possible that the association with
clinical risk may vary based on the time of ascertainment.
Serial measurements appear to provide additional infor-
mation that may reflect the patient’s risk at presentation
as well as the response to therapy and effects of ventric-
ular remodeling (97–99 ).
Fig. 6. Risk of death in patients with NSTEAC syndrome stratified by
quartile of concentration of NT-proBNP (Elecsys 2010, Roche Diagnos- d. Therapeutic decision-making
tics) at baseline. Few studies have evaluated the effects of specific thera-
With permission from James et al. (89 ). pies on ameliorating the risk associated with increased
BNP or NT-proBNP in ACS (see Section III-A2). Two Increased concentrations of inflammatory biomarkers
studies have evaluated whether BNP/NT-proBNP is such as CRP, serum amyloid A, myeloperoxidase, and
helpful for identifying candidates for early routine refer- interleukin-6 (IL-6) are detectable in a substantial pro-
ral for coronary revascularization (“early invasive strat- portion of patients presenting with ACS, including those
egy”) after ACS. In the first of these studies, patients without evidence of myocyte necrosis (107–112 ). It is
with an increased plasma concentration of BNP experi- plausible that elevation of circulating markers of in-
enced a similar benefit of the early invasive approach flammation during ACS is a manifestation of intensi-
compared to patients with BNP ⬍80 pg/mL (84 ). In the fication of the focal inflammatory processes that contrib-
second, a trend toward greater benefit with the early ute to destabilization of vulnerable plaque. Nevertheless,
invasive strategy was apparent in patients in the highest the precise basis for the relationship between inflamma-
tertile of NT-proBNP (88 ). This latter observation is tory markers and risk in ACS has not been conclusively
supported by a nonrandomized evaluation of patients established. CRP certainly rises as a consequence of the
with increased NT-proBNP who did and did not undergo inflammatory response to myocardial necrosis (113 ).
revascularization (100 ). One study has shown a signifi- However, studies demonstrating elevation of CRP and
cant reduction in the risk of death or new heart failure in IL-6 during ACS in the absence of myocyte necrosis refute
patients with increased BNP treated with intensive statin the position that the rise in these markers is solely a
therapy (101 ). response to necrosis (107, 109, 110 ). CRP has also been
Although convincing data for a strong interaction implicated as a potential direct participant in athero-
between the biomarker and specific therapeutic strategies thrombosis rather than a mere bystander. CRP promotes
do not yet exist for natriuretic peptides as they do for uptake of LDL cholesterol by monocytes, induces the
troponin, BNP and NT-proBNP do assist in an assessment production of tissue factor, activates complement within
of absolute global risk and may therefore still inform arterial plaque, stimulates the expression of adhesion
molecules, and may also recruit monocytes via a mono-
clinical decision-making. For example, owing to the very
cyte-CRP receptor (103 ). Nevertheless, in light of limita-
low mortality rate observed for patients with negative
tions to the experimental data, there remains a need for
troponin results and low concentrations of BNP or NT-
additional investigation of the role of CRP as a potential
proBNP, it has been proposed that less aggressive man-
direct mediator (114 ). Last, the clinical importance of
agement strategies may be employed for such patients
identifying inflammatory activation in ACS may have less
(102 ). In addition, studies of both BNP and NT-proBNP
to do with the particular inciting culprit and more to do
have demonstrated that a decline to a lower concentration
with the widespread presence of vulnerable plaques (115 )
of natriuretic peptides over time after presentation with
and patient-specific responses to inflammatory stimuli
ACS is associated with more favorable outcomes and thus
(116 ).
raised the possibility that natriuretic peptides may be
useful as a tool to monitor the response to preventive b. Relationship to clinical outcomes
interventions (98, 99 ). There have now been more than 12 clinical studies
demonstrating the prognostic capacity of hs-CRP deter-
3. biochemical markers of inflammation mined either at presentation or at discharge after ACS
a. Pathophysiology (Table 3). Data restricted to patients with STEMI are few;
Multiple lines of investigation have converged to impli- in 1 cohort study, patients with increased CRP were more
cate inflammation as a central contributor to plaque likely to suffer complications of acute MI (myocardial
compromise (103 ). Inflammatory processes participate in rupture, left ventricular aneurysm, and death by 1 year)
the earliest stages of atherogenesis in response to insults (117 ). However, in at least 9 studies, multivariable anal-
to the vascular endothelium, as well as to the develop- ysis revealed hs-CRP to be an independent predictor of
ment of the intermediate and mature atheromatous short- and/or long-term outcome among patients with
plaque. Ultimately, inflammatory cells and mediators NSTEACS (59, 60, 118 –125 ). Specifically, measurement of
participate in compromising the protective fibrous cap hs-CRP appears to yield additional prognostic value in
that maintains separation between the highly procoagu- patients with negative testing of cardiac troponins
lant contents of the atheroma core and circulating plate- (109, 124 ) and adds to information obtained from the
lets and coagulation proteins (104, 105 ). Thus, several clinical history and ECG. Several, but not all, studies
mediators of the inflammatory response, including acute- indicate that the relationship between hs-CRP and out-
phase proteins, cytokines, and cellular adhesion mole- come is strongest with respect to mortality with a weaker
cules, have been evaluated as potential indicators of the relationship to recurrent MI (60, 109, 119 ). Whereas hs-
risk of a first acute atherothrombotic event, as well as of CRP is the best studied of the inflammatory markers in
recurrent complications after presentation (106 ). As the the setting of ACS, others such as IL-6 (126, 127 ) and
prototypical acute-phase reactant, C-reactive protein myeloperoxidase (111, 128 ) are also associated with prog-
(CRP) has been the focus of much of the clinical investi- nosis and may eventually prove to add or supercede
gation (107 ). hs-CRP (see section II-B6).
Table 3. Summary of clinical studies of CRP in ACS.
A. NSTEACS
CRP cut point,
Author, year Study Subjects mg/L Follow-up End point, risk relationship for high CRP
Short-term
Liuzzo et al., 1994 (107 ) Observational 31 ⬎3 In-hospital D/MI/RI/UR, 4.5 (1.4–17.5)
Oltrona et al., 1997 (185 ) Observational 140 ⬎10 21 days D/MI/RI, 0.46 (0.19–1.11)
Toss et al., 1997 (119 ) Substudy of RCT 965 ⬎10 5 months D/MI, 1.19 (0.97–1.64)
(FRISC)
Morrow et al., 1998 (109 ) Substudy of RCT 437 ⬎15 14 days Death, 18.3 (2.2–150)
(TIMI 11A)
Rebuzzi et al., 1998 (120 ) Observational 102 ⬎3 3 months MI, 6.0 (1.4–25.3)
Oltrona et al., 1998 (186 ) Observational 91 ⬎3 In-hospital D/MI, 1.94 (0.46–8.3)
Benamer et al., 1998 (134 ) Observational 100 ⬎6 In-hospital D/MI/RI/UR, 0.65 (0.17–2.1)
Ferreiros et al., 1999 (122 ) Observational 105 ⬎15 In-hospital D/MI/RI, 0.83 (0.29–2.4)
3 months D/MI/RI, 2.1 (1.5–3.1)
Bazzino et al., 2001 (187 ) Observational 139 ⬎15 3 months D/MI, 18.6 (4.5–77)
Mueller et al., 2002 (125 ) Observational 1042 ⬎10 In-hospital Death, 4.2 (1.6–10.9)
James et al., 2003 (60 ) Substudy of RCT 7108 ⬎10 1 month Death, 1.2 (1.05–1.4)
(GUSTO IV)
Oltrona et al., 2004 (188 ) Observational 965 ⬎10 1 month D/MI, 2.0 (1.3–3.1)
Long-term
de Winter et al., 1999 (189 ) Observational 156 ⬎5 6 months D/MI/RI, 9.8 (1.5–65)
Heeschen et al., 2000 (59 ) Substudy of RCT 447 ⬎10 6 months Death, 4.7 (1.3–16.9)
(CAPTURE)
Mulvihill et al., 2001 (190 ) Observational 91 ⬎3 6 months D/MI/RI, 9.8 (2.5–38.9)
Bholasingh et al., 2003 Observational 382 ⬎3 6 months D/MI, 5.6 (1.5–22.2)
(191 )
Baldus et al., 2003 (128 ) Substudy of RCT 1090 ⬎10 6 months D/MI, 1.25 (1.02–1.7)
(CAPTURE)
Bodi et al., 2005 (192 ) Observational 515 ⬎11 6 months D/MI, 2.1 (1.2–3.8)
Biasucci et al., 1999 (123 ) Observational 53 ⬎3 1 year D/MI/RI, 4.7 (1.8–12.0)
Lindahl et al., 2000 (124 ) Substudy of RCT 917 ⬎10 3 years Death, 2.5 (1.6–3.9)
(FRISC)
Versaci et al., 2000 (193 ) Observational 62 ⬎5 1 year D/MI/RI, 22.2 (3.1–157)
Mueller et al., 2002 (125 ) Observational 1042 ⬎10 20 months Death, 3.8 (2.3–6.2)
Zebrack et al., 2002 (194 ) Observational 442 ⬎11 3 years D/MI, 2.6 (1.4–4.8)
James et al., 2003 (60 ) Substudy of RCT 7108 ⬎10 1 year Death, 1.5 (1.1–1.9)
Sanchez et al., 2004 (195 ) Observational 83 ⬎5 2 years Death, 4.5 (1.6–12.5)
B. STEMI
Author, year Study Subjects CRP cut point, Follow-up End point, risk relationship for high CRP
mg/L
Short-term
Liuzzo et al., 1994 (107 ) Observational 29 ⬎3 In-hospital RR not provided
Pietila et al., 1996 (196 ) Observational 188 None 6 months RR not provided
Anzai et al., 1997 (117 ) Observational 220 ⬎20 Death, 6.59 (2.7–1.61)
Tommasi et al., 1999 (121 ) Observational 64 ⬎25 1 year D/MI/angina, 3.55 (1.56–8.04)
Nikfardjam et al., 2000 Observational 729 Quintiles 3 years Death, no relationship
(197 )
Oltrona et al., 2004 (188 ) Observational 808 ⬎10 30 days D/MI, 1.9 (1.1–3.2)
Mega et al., 2004 (198 ) Substudy of RCT 483 ⬎15 30 days Death, no relationship
RR, relative risk; RI, recurrent ischemia; UR, urgent revascularization.
c. Decision-limits that the optimal decision limit for ACS is higher than that
The preferred unit for reporting hs-CRP results is mg/L used in candidates for primary prevention (129 ). In 1
(129 ). Multiple decision-limits for hs-CRP, ranging from prospective evaluation of multiple cut points using re-
3–15 mg/L, have been evaluated for risk assessment in ceiver-operating characteristics, 15 mg/L was the optimal
ACS with few comparative studies. Consensus opinion is decision-limit for prediction of a composite of death and
recurrent ischemic events (122 ). A cut point of 10 mg/L ing early after onset of an acute MI for which cTnI/T is
has also been validated in published studies and thus not yet detectable by serum/plasma testing; the remain-
the optimal decision limit remains to be determined der are presenting with acute myocardial ischemia with-
(59, 60, 124 ). When tested 1 or more months after presen- out necrosis (i.e., unstable angina). Discriminating these 2
tation with ACS, use of cut points recommended for groups from patients with chest pain syndrome of an
patients at risk for or with stable coronary artery disease etiology other than coronary ischemia is a major clinical
(low: ⬍1 mg/L; intermediate 1–3 mg/L; high: ⬎3 mg/L) challenge. Thus, a biomarker that reliably detects myocar-
is appropriate (129, 130 ). Additional comparative studies dial ischemia in the absence of necrosis, and/or before
of decision-limits for hsCRP in ACS are likely to be useful. cardiac troponin is increased, has the potential to add
In addition, recognition of differences in the distribution substantially to available clinical tools (143, 144 ).
of hs-CRP based on race and ethnicity may warrant Several biomarkers of myocardial ischemia are under
specific reporting of decision-limits (131–133 ). investigation (144 ). Ischemia-modified albumin (IMA) is
The best timing for measurement of hs-CRP for risk among the most thoroughly studied of these markers and
stratification in ACS remains uncertain. Potential con- has been approved by the US Food and Drug Adminis-
founding by the inflammatory response to necrosis must tration for clinical use (145–148 ). The albumin cobalt-
be considered when samples are drawn late after presen- binding test for detection of IMA is based on the obser-
tation of patients with MI (134, 135 ). Studies with samples vation that the affinity of the N-terminus of human
drawn early after presentation (109, 121 ), at discharge albumin for cobalt is reduced in patients with myocardial
(120, 123 ), and during the convalescent phase of recovery ischemia. Detectable changes in albumin cobalt binding
(ⱖmonths postMI) (130, 136 ) have all demonstrated inde- have been documented to occur minutes after transient
pendent associations with subsequent outcomes. In 2 occlusion and reperfusion of a coronary artery during
comparative studies of samples drawn at admission vs angioplasty and return toward baseline within 6 h (146 ).
discharge, a modest advantage of the predischarge assess- Reduced albumin cobalt binding also occurs in patients
ment was evident (but not statistically heterogeneous) with spontaneous coronary ischemia (145, 147, 149 ), with
(120, 123 ). It is plausible that values of CRP obtained early an abnormal concentration detectable before demonstra-
during the presentation with ACS reflect different patho- ble increase of cardiac troponin (147 ). The precise mech-
physiologic contributors and relationships to risk than anisms for production of IMA during coronary ischemia
those manifest by determination of CRP after resolution of are not known, but have been localized to modifications
the acute-phase response. Data raising the potential value of the N-Asp-Ala-His-Lys sequence of human albumin
of late measurement (ⱖ1 month after ACS) for monitoring and are proposed to be related to production of free
therapy (discussed below) may indicate greater clinical radicals during ischemia and/or reperfusion, reduced
utility to values obtained later rather than early after ACS oxygen tension, acidosis, and cellular alterations such as
(130 ). Additional research aimed at resolving these issues disruption of sodium and calcium pump function
is needed. (146, 150 ).
The clinical specificity of IMA, as well as other poten-
d. Therapeutic decision-making tial markers of ischemia such as unbound free fatty acid
The appropriate therapeutic response to increased mark- (151 ) and whole blood choline (152 ), in the broad popu-
ers of inflammation in patients with ACS is not yet clear. lation of patients with nontraumatic chest pain and sus-
Treatment with hydroxymethylglutaryl (HMG)-CoA re- pected ACS remains an area for further investigation.
ductase inhibitors (statins) is effective in lowering CRP in Increased concentrations of IMA have been demonstrated
patients with recent or prior ACS (137, 138 ). Observations 24 – 48 h after endurance exercise and postulated to relate
from randomized trials of aggressive vs moderate statin to delayed gastrointestinal ischemia (153 ). A deletion
therapy support a possible role for measurement of hs- defect of the N-terminal causing reduced cobalt binding
CRP during follow-up after ACS as a guide for monitor- (a false-positive test for ischemia) has also been reported
ing the success of therapy (130, 139 ). The effect of aspirin (149 ). The concentration of albumin has also been shown
on inflammatory markers is controversial but not likely to to influence albumin cobalt binding in some but not all
impact therapeutic selection, as aspirin therapy is admin- studies (154 ). IMA may be considered for use in conjunc-
istered to all patients with ACS (140 –142 ). It is possible tion with the ECG and cardiac troponin for the diagnostic
that future work investigating more aggressive antiin- assessment of suspected ACS to exclude ACS in patients
flammatory therapies for the acute management of ACS with a low clinical probability (148 ). Available data high-
may lead to a role for inflammatory markers in guiding light the potential for false-positive results when used as
such therapy. a diagnostic tool for ACS. In addition, the concentration of
IMA is no longer increased by 6 –12 h after provoked
4. biochemical markers of ischemia ischemia and thus the negative predictive value may be
Approximately 40%– 60% of patients with definite ACS diminished in patients who do not present early after an
present with an initial troponin concentration below the ischemic event (146 ). Studies of IMA, and other proposed
clinical decision-limit for the assay (64 ). Some are present- tests for ischemia, evaluating the prognostic implications
and/or interaction with specific therapies as well as the cytes during activation in the coronary bed) (111, 128 ) are
kinetics, analytic performance, and underlying patho- newer markers that have shown potential for risk strati-
physiology will be important to defining their clinical fication in ACS. These and other emerging biomarkers
role. that also reflect the underlying pathobiology of athero-
thrombosis are the substrate of ongoing investigation
5. multimarker approach aimed at determining the optimal combination of biomar-
Advances in our understanding of the pathogenesis and kers for characterizing patients with ACS (158 ). Newer
consequences of ACS have stimulated development of technologies that have facilitated proteomic and genomic
new biomarkers and created the opportunity for an ex- strategies for novel marker discovery are likely to extend
panded role of multiple biomarkers in the classification this approach. Careful evaluation of such novel markers
and individualization of treatment (84, 155 ). Accumulat- relative to appropriate use of contemporary tools, avoid-
ing evidence indicates that a multimarker strategy, em- ing limitations to the methodology cited as prevalent in
ploying a pathobiologically diverse set of biomarkers, studies of novel biomarkers, is essential to evaluating
adds to biomarkers of necrosis for risk assessment in ACS their potential to add to clinical use (159 ). In addition,
(13 ). To date, the majority of evidence regarding this collaborative pooled analyses that evaluate the diagnostic
strategy entails newer markers paired with troponin, accuracy and prognostic performance of new and estab-
hs-CRP, and BNP are the most extensively studied. Few lished biomarkers across multiple studies are likely to be
studies have examined strategies incorporating 2 or more useful in the critical assessment of their individual and
markers in addition to troponin (128, 155 ). combined clinical value.
Consistent data from multiple studies indicate that
increased concentrations of CRP and BNP or NT-proBNP III. Use of Biochemical Markers in the Management of
at presentation identify patients who are at higher mor- NSTEACS
tality risk irrespective of whether there is detectable a. clinical decision-making
elevation of troponin (60, 84, 89, 109, 124 ). Thus, applica-
tion of either of these markers along with a biomarker of
necrosis (cardiac troponin) enhances risk assessment (83–
recommendations for the use of biochemical
86, 89, 109, 124 ). Moreover, in one study (with internal
cardiac markers for therapeutic
validation from 2 separate trials), a simple multimarker
decision-making
approach combining each of these markers (BNP, CRP,
class i
Among patients with a clinical history consistent
cTnI) identified a 6- to 13-fold gradient of mortality risk
with ACS, an increased concentration of cardiac
between those without elevation of any marker and those
troponin should prompt application of ACS manage-
in whom all 3 markers were increased (155 ). Additional
ment guidelines for patients with indicators of high
research evaluating this and other strategies for combin-
risk (Level of Evidence: B).
ing 2 or more pathobiologically diverse biomarkers will
clarify the appropriate clinical role for such an approach.
In particular, 2 important issues require exploration. First,
class iii
because the relative risk relationships between the indi- 1. Application of management guidelines for ACS
vidual biomarkers and specific endpoints differ, the opti- should not be based solely on measurement of
mal weighting of each marker for assessment of 1 clinical natriuretic peptides (Level of Evidence: C).
outcome (e.g., mortality risk) may differ from that for 2. Application of management guidelines for ACS
evaluating another outcome (e.g., the risk of recurrent should not be based solely on measurement of
MI). Second, given the present lack of a robust database to CRP (Level of Evidence: C).
guide treatment in response to increased concentrations
of these “novel” markers, more information is needed to
formulate an evidence-based management strategy tied to 1. biochemical markers of cardiac injury
multimarker testing. Nevertheless, as new markers and The recommendation for measurement of cardiac tropo-
therapies are discovered, a multimarker paradigm em- nin in all patients with suspected ACS derives not only
ploying a combination of biomarkers for risk assessment from the importance of biomarkers of necrosis for risk
and clinical decision-making has the potential to improve assessment but also from the established value of cardiac
outcomes for patients with ACS (13 ). troponin, in particular, for therapeutic decision-making.
Consistent with the observation that patients with an
6. other novel markers increased concentration of troponin are more likely to
Other biomarkers such as soluble CD40 ligand, (a marker have complex thrombotic coronary lesions, they also
of platelet activation and potential direct participant in derive greater benefit from more aggressive anticoagu-
plaque destabilization) (156 ), metalloproteinases (en- lant, antiplatelet, and invasive therapies (Figs. 8 and 9). As
zymes that disrupt the integrity of the atheroma’s protec- such, patients with suspected ACS and abnormal tropo-
tive cap) (157 ), and myeloperoxidase (released by leuko- nin results should be treated in accordance with the
Low-molecular-weight heparin
Two studies indicate that potent antithrombotic therapy
with low-molecular-weight heparin offers particular ben-
efit among patients with an increased concentration of
troponin. In the TIMI 11B trial, patients with an increased
serum concentration of cTnI at presentation experienced a
50% reduction in death, MI, or recurrent ischemia at 14
days when treated with enoxaparin compared with un-
fractionated heparin. In contrast, there was no demonstra-
ble advantage of enoxaparin compared with unfraction-
ated heparin in patients without detectable cTnI (67 ). In
Fig. 8. Effect of potent antithrombotic therapy on the risk of death and the Fragmin during Instability in Coronary Artery
recurrent ischemic events. (FRISC) trial, extended treatment with dalteparin (Frag-
Left, effect of the platelet GPIIb/IIIa receptor inhibitor, tirofiban, among patients min) after the initial hospitalization conferred a benefit
with NSTEAC syndrome enrolled in the Platelet Receptor Inhibition in Ischemic only among patients with increased cardiac troponin
Syndrome Management (PRISM) trial. Data from Heeschen et al. (162 ). Right,
effect of the low-molecular-weight heparin, enoxaparin, among patients with (160 ).
NSTEAC syndrome enrolled in the TIMI 11B trial. Data from Morrow et al. (67 ).
Neg, negative; Pos, positive; UR, urgent revascularization prompted by recurrent
ischemia.
Glycoprotein iib/iiia receptor inhibition
Four studies provide evidence for an interaction between
troponin results and the efficacy of potent platelet inhibi-
American Heart Association/American College of Cardi-
tion with intravenous glycoprotein (GP) IIb/IIIa receptor
ology (1 ) and European Society of Cardiology (2 ) guide-
antagonists (161–164 ). In the first of these studies, among
lines for the management of high-risk patients with
patients treated with abciximab for 24 h before percuta-
NSTEACS. These guidelines for the management of ACS
neous intervention, those with an increased concentration
are expected to be dynamic over time as new experience
of troponin experienced a 70% relative reduction in the
and evidence emerge. The reader should recognize that risk of death or MI, while those with negative troponin
the data guiding this recommendation originate from results had no benefit compared with placebo (161 ).
patients with a high clinical probability for ACS. Aggres- Similar results have been obtained with 2 other GPIIb/IIIa
sive treatment with potent antithrombotic therapies and receptor inhibitors (162–164 ). Discordant results from one
early invasive evaluation is often not appropriate for study are notable (165 ). In a trial that tested abciximab as
patients with abnormal troponin results due to mecha- medical therapy in patients being managed conserva-
nisms other than ACS (e.g., myocarditis or sepsis). Data tively (without early coronary angiography) for
regarding the efficacy of specific therapies in patients with NSTEACS, there was no benefit of abciximab, including
increased cardiac troponin are discussed below. among patients with increased concentration of troponin.
These results are not yet well explained, but may derive
from the specific medical strategy and dosing in this trial.
Accordingly, the 2002 update to the American College of
Cardiology/American Heart Association Guidelines for
the Management of Patients with Unstable Angina and
Non–ST-Segment Elevation Myocardial Infarction recom-
mends the use of GPIIb/IIIa receptor antagonists in
patients with increased troponin whether (Class I) or not
(Class IIa, eptifibatide or tirofiban only) early cardiac
catheterization and revascularization are planned (1 ).
Early invasive strategy

The TACTICS-TIMI 18 trial prospectively examined the
value of cardiac troponin for identifying patients who
Fig. 9. Benefit of an early invasive (Inv) vs conservative (Con) manage- would benefit from an early invasive management strat-
ment strategy on the risk of death and new/recurrent MI at 6 months
egy. Among patients with an increased concentration of
in patients with NSTEAC syndrome enrolled in the TACTICS-TIMI 18
trial. troponin at presentation, a strategy of early angiography
The early invasive strategy consisted of routine cardiac catheterization within (4 to 48 h) and revascularization (if appropriate) achieved
48 h of presentation and revascularization when appropriate regardless of a ⬃55% reduction in the odds of death or MI compared
clinical course. The conservative strategy included coronary angiography and
revascularization only when prompted by recurrent spontaneous or provoked
with a conservative management strategy [Fig. 9 (66 )].
ischemia. Data from Morrow et al. (66 ). Neg, negative; Pos, positive. Early angiography and revascularization was not associ-
ated with a detectable benefit in patients who did not by the patient’s clinical status and obtained when recur-
have an increased concentration of troponin. Importantly, rent symptoms consistent with ischemia of sufficient
the advantage of an early invasive strategy was evident duration to cause myocardial necrosis have occurred.
even among patients with the lowest level of troponin Routine measurement of biomarkers of necrosis after
elevation (cTnI 0.1– 0.5 ␮g/L and cTnT 0.01– 0.05 ␮g/L) uncomplicated percutaneous coronary revascularization
(66 ). These data, along with similar results from the may aid in assessment of long-term risk (1 ); however,
FRISC II trial (166 ), support the recommendation for early data with more sensitive markers of necrosis are mixed
angiography in patients with suspected ACS and an (167 ), and the implications for periprocedural manage-
increased concentration of troponin (1 ). ment are uncertain.
IV. Use of Biochemical Markers in the Management of

2. other biochemical markers
STEMI
Consistent and compelling evidence for interactions be-
The diagnosis of STEMI is made by recognition of acute
tween other available biomarkers (e.g., BNP and hs-CRP)
ST-segment elevation (or reciprocal depression) on the
and specific treatment strategies in ACS are not yet
12-lead electrocardiogram. Therefore, appropriate ther-
available (see Section II-B for discussion of individual
apy should be instituted on the basis of a diagnostic ECG
markers/classes). A number of interventions, such as
(See section II-A4) (10 ). Confirmation of myocardial ne-
early treatment with statins and use of GPIIb/IIIa an-
crosis is subsequently made using specific biomarkers of
tagonists, have been shown to reduce the serum concen-
necrosis. In addition to this confirmatory application,
tration of hs-CRP after presentation with ACS and/or
biomarkers may be used for several other purposes in the
in response to percutaneous coronary intervention
management of patients with STEMI.
(137, 138 ). However, testing for a differential impact of
treatment among those with or without higher concentra- a. noninvasive assessment of reperfusion
tions of CRP has been negative (59 ). A substudy of the One of the most challenging decisions in the acute care of
FRISC II trial has demonstrated the potential for greater patients with STEMI is when (and if) to perform urgent
benefit of early invasive management in patients with cardiac catheterization following fibrinolytic therapy. The
evidence of systemic inflammation (increased IL-6) (127 ); pattern of rise and fall of biomarkers of necrosis can assist
however, more data are needed before this application of in a noninvasive assessment of the success of reperfusion
inflammatory biomarkers can be advocated. Similarly, a of the infarct-related coronary artery. In the early experi-
trend toward greater efficacy of early invasive manage- ence with fibrinolytics, it was noted that reperfusion of an
ment has been manifest among patients with a higher occluded artery was accompanied by an abrupt increase
plasma concentration of NT-proBNP (88 ). Additional in serum CK followed by an early peak, findings that
data in this regard are mixed, and more research is were attributed to washout of proteins from injured cells
needed before a role for natriuretic peptides in therapeu- at the time of restoration of blood flow (168, 169 ). Inves-
tic decision-making is clearly defined (84 ). There is some tigators thus recognized that the rate of rise in biomarkers
evidence for promise of novel markers for selection of of necrosis over the first few hours after reperfusion
therapy, such as the use of GPIIb/IIIa receptor antago- therapy provided information regarding patency of the
nists in patients with increased concentrations of soluble infarct-related artery. Myoglobin has attracted the most
CD40 ligand (156 ). attention for this purpose because of its small molecular
size and consequent rapid release (170 –172 ). Rapid wash-
outs of myoglobin, cTnT or cTnI, or CKMB have positive
B. Biochemical Marker Measurement After the Initial predictive values (PPV) ⬎90% for infarct artery patency
Diagnosis (171–174 ).
After the initial diagnosis of unstable angina or NSTEMI However, a number of factors have limited the clin-
is established, measurement of biomarkers is useful for ical application of these findings. First, absence of bio-
updating the initial assessment of risk, qualitative assess- marker washout appears to overestimate the likelihood
ment of the size of infarction, and detection of new or of an occluded artery and cannot accurately distin-
recurrent myocardial injury. See section IV-B for guide- guish slow from normal flow (172, 174 ). Second, the
lines regarding the serial collection of biomarkers of logistical challenges of performing multiple measure-
injury after an initial diagnosis of MI. ments in real time have limited use of this strategy. Last,
For patients in whom the index event is established to with the steady trend toward more frequent use of
be unstable angina, cardiac troponin is the preferred primary angioplasty (where there is direct angiographic
marker for the detection of new infarction. Diagnostic assessment of the artery) for the treatment of STEMI, the
criteria are as described for the index event (section II-A). relevance of this application to contemporary practice is
Repeat sampling of cardiac troponin should be guided diminishing.
b. biochemical marker measurement after the may also be useful in determining the timing of recent MI.
diagnosis of acute mi The committee encourages further investigation of the
recommendations for measurement of kinetics of available troponin assays, as well as concurrent
biochemical markers of cardiac injury after evaluation of troponin and CK-MB for the diagnosis of
the diagnosis of mi early reinfarction. Data directly comparing these biomar-
kers for detection of reinfarction are few and may help
guide deliberation as to whether CK-MB should continue
class i
to have a role in the routine care of patients with acute MI.
1. Once the diagnosis of acute MI is ascertained,
The value of biomarkers of necrosis to discriminate
testing of biochemical markers of injury at a reduced
very early reinfarction (e.g., ⬍18 h) during a period when
frequency (e.g., Q6 –10 h ⫻ 3) is valuable to qualita-
the concentration of these markers is typically still in-
tively estimate the size of the infarction and to
creasing is limited. As discussed in greater detail in
facilitate the detection of complications such as rein-
separate guidelines (Cardiac Biomarkers and Other Etiolo-
farction (Level of Evidence: C).
gies), the diagnosis of very early reinfarction rests pre-
dominantly on clinical grounds (symptoms and electro-
class iia
cardiographic changes). Routine serial acquisition of
2. CK-MB is the preferred marker for detection of
surveillance sampling for biomarkers of necrosis after
reinfarction early after the index event when the
they have returned to the normal range from the index
concentration of cardiac troponin is still increased
event is not recommended.
(Level of Evidence: C).
class iib
3. Cardiac troponin may be used as an alternative to Financial Disclosures: The National Academy of Clinical
CK-MB for detection of reinfarction early after the Biochemistry Laboratory Medicine Practice Guidelines
index event. Serial measurement of troponin is usu- Committee for Utilization of Biomarkers in Acute Coro-
ally necessary to facilitate the discrimination of a new nary Syndromes and Heart Failure reports all reported
increase in concentration (Level of Evidence C). relationships within the 2 years previous to this publica-
tion that may be relevant to this guidelines document. A
During the course of management after the diagnosis document of those relationships may be found in the
of acute MI is ascertained, serial measurements of biomar- online Data Supplement at http://www.clinchem.org/
kers of myocardial necrosis are useful in demonstrating content/vol53/issue4.
the characteristic rise and/or fall that aids in confirming
the diagnosis of MI, providing qualitative information
V. REFERENCES
with respect to infarct size and surveying for ongoing or 1. Braunwald E, Antman EM, Beasley JW, Califf RM, Cheitlin MD,
recurrent myocardial ischemia causing reinfarction. Hochman JS, et al. ACC/AHA 2002 guideline update for the
Among patients admitted with MI, the magnitude and management of patients with unstable angina and non-ST-seg-
temporal course of CK-MB elevation and decline have ment elevation myocardial infarction—summary article: a report
been shown to correlate strongly with infarct size (175– of the American College of Cardiology/American Heart Associa-
177 ). Although experimental and clinical data (using tion task force on practice guidelines (Committee on the Man-
agement of Patients With Unstable Angina). J Am Coll Cardiol
magnetic resonance imaging) demonstrate that cardiac
2002;40:1366 –74.
troponin may provide comparable, if not superior, data
2. Bertrand ME, Simoons ML, Fox KA, Wallentin LC, Hamm CW,
regarding infarct size and reperfusion (178 –181 ), the McFadden E, et al. Management of acute coronary syndromes:
clinical meaning of peak values remains less familiar to acute coronary syndromes without persistent ST segment eleva-
clinicians. Increases in troponin are demonstrable in cases tion; recommendations of the Task Force of the European Society
of early reinfarction (182 ). However, the scope of avail- of Cardiology. Eur Heart J 2000;21:1406 –32.
able evidence is significantly limited compared with that 3. Storrow AB, Gibler WB. Chest pain centers: diagnosis of acute
for CK-MB. In addition, cTnT is known to exhibit a coronary syndromes. Ann Emerg Med 2000;35:449 – 61.
bimodal distribution, and the kinetics for multiple avail- 4. The Joint European Society of Cardiology/American College of
able assays for cTnI have not been studied. Serial testing Cardiology Committee for the redefinition of myocardial infarc-
is usually necessary to discriminate a new increasing tion. Myocardial infarction redefined—a consensus document of
The Joint European Society of Cardiology/American College of
pattern of troponin if the concentration is not known to
Cardiology Committee for the redefinition of myocardial infarc-
have returned to normal. Because CK-MB falls to the tion. J Am Coll Cardiol 2000;36:959 – 69.
reference interval by 48 –72 h, it may aid in the rapid 5. Braunwald E. Unstable angina. A classification. Circulation 1989;
discrimination of reinfarction when symptoms recur be- 80:410 – 4.
tween 72 h and 2 weeks after the index MI, when troponin 6. Fuster V, Badimon L, Badimon JJ, Chesebro JH. The pathogene-
may still be increased from the initial cardiac event. sis of coronary artery disease and the acute coronary syndromes
Measurement of CK-MB in conjunction with troponin (2). N Engl J Med 1992;326:310 – 8.
7. Fuster V, Badimon L, Badimon JJ, Chesebro JH. The pathogene- non-ST elevation acute coronary syndromes: a meta-analysis.
sis of coronary artery disease and the acute coronary syndromes J Am Coll Cardiol 2001;38:478 – 85.
(1). N Engl J Med 1992;326:242–50. 27. Ohman EM, Casey C, Bengtson JR, Pryor D, Tormey W, Horgan
8. Lee RT, Libby P. The unstable atheroma. Arterioscler Thromb JH. Early detection of acute myocardial infarction: additional
Vasc Biol 1997;1859 – 67. diagnostic information from serum concentrations of myoglobin
9. American Heart Association. Heart Disease and Stroke Statis- in patients without ST elevation. Br Heart J 1990;63:335– 8.
tics: 2004 Update. Dallas: American Heart Association, 2004. 28. Brogan GX Jr, Friedman S, McCuskey C, Cooling DS, Berrutti L,
10. Ryan TJ, Antman EM, Brooks NH, Califf RM, Hillis LD, Hiratzka LF, Thode HC Jr, et al. Evaluation of a new rapid quantitative
et al. 1999 update: ACC/AHA guidelines for the management of immunoassay for serum myoglobin versus CK-MB for ruling out
patients with acute myocardial infarction. A report of the Ameri- acute myocardial infarction in the emergency department. Ann
can College of Cardiology/American Heart Association Task Emerg Med 1994;24:665–71.
Force on Practice Guidelines (Committee on Management of 29. Kontos MC, Anderson FP, Hanbury CM, Roberts CS, Miller WG,
Acute Myocardial Infarction). J Am Coll Cardiol 1999;34:890 – Jesse RL. Use of the combination of myoglobin and CK-MB mass
911. for the rapid diagnosis of acute myocardial infarction. Am J
11. The TIMI IIIB Investigators. Effects of tissue plasminogen activa- Emerg Med 1997;15:14 –9.
tor and a comparison of early invasive and conservative strate- 30. Newby LK, Storrow AB, Gibler WB, Garvey JL, Tucker JF, Kaplan
gies in unstable angina and non-Q-wave myocardial infarction. AL, et al. Bedside multimarker testing for risk stratification in
Results of the TIMI IIIB Trial. Circulation 1994;89:1545–56. chest pain units: the Chest pain Evaluation by Creatine Kinase-
12. Braunwald E. Unstable angina: an etiologic approach to manage- MB, Myoglobin, And Troponin I (CHECKMATE) study. Circulation
ment. Circulation 1998;98:2219 –22. 2001;103:1832–7.
13. Morrow DA, Braunwald E. Future of biomarkers in acute coronary 31. Eggers KM, Oldgren J, Nordenskjold A, Lindahl B. Diagnostic
syndromes: moving toward a multimarker strategy. Circulation value of serial measurement of cardiac markers in patients with
2003;108:250 –2. chest pain: limited value of adding myoglobin to troponin I for
14. Maseri A, Rebuzzi AG, Cianflone D. Need for a composite risk exclusion of myocardial infarction. Am Heart J 2004;148:574 –
stratification of patients with unstable coronary syndromes tai- 81.
lored to clinical practice. Circulation 1997;96:4141–2. 32. Puleo PR, Meyer D, Wathen C, Tawa CB, Wheeler S, Hamburg RJ,
15. Cannon CP. Evidence-based risk stratification to target therapies et al. Use of a rapid assay of subforms of creatine kinase-MB to
in acute coronary syndromes. Circulation 2002;106:1588 –91. diagnose or rule out acute myocardial infarction. N Engl J Med
16. Katus HA, Remppis A, Neumann FJ, Scheffold T, Diederich KW, 1994;331:561– 6.
Vinar G, et al. Diagnostic efficiency of troponin T measurements 33. Katus HA, Remppis A, Scheffold T, Diederich KW, Kuebler W.
in acute myocardial infarction. Circulation 1991;83:902–12. Intracellular compartmentation of cardiac troponin T and its
17. Katus HA, Remppis A, Looser S, Hallermeier K, Scheffold T, release kinetics in patients with reperfused and nonreperfused
Kubler W. Enzyme linked immuno assay of cardiac troponin T for myocardial infarction. Am J Cardiol 1991;67:1360 –7.
the detection of acute myocardial infarction in patients. J Mol 34. Mair J, Artner-Dworzak E, Lechleitner P, Morass B, Smidt J,
Cell Cardiol 1989;21:1349 –53. Wagner I, et al. Early diagnosis of acute myocardial infarction by
18. Bodor GS, Porter S, Landt Y, Ladenson JH. Development of a newly developed rapid immunoturbidimetric assay for myoglo-
monoclonal antibodies for an assay of cardiac troponin-I and bin. Br Heart J 1992;68:462– 8.
preliminary results in suspected cases of myocardial infarction. 35. Antman EM, Grudzien C, Sacks DB. Evaluation of a rapid bedside
Clin Chem 1992;38:2203–14. assay for detection of serum cardiac troponin T. JAMA 1995;
19. Wu AH, Valdes R, Jr, Apple FS, Gornet T, Stone MA, Mayfield- 273:1279 – 82.
Stokes S, et al. Cardiac troponin-T immunoassay for diagnosis of 36. Zimmerman J, Fromm R, Meyer D, Boudreaux A, Wun CC,
acute myocardial infarction. Clin Chem 1994;40:900 –7. Smalling R, et al. Diagnostic marker cooperative study for the
20. Adams JE, III, Sicard GA, Allen BT, Bridwell KH, Lenke LG, diagnosis of myocardial infarction. Circulation 1999;99:1671–7.
Davila-Roman VG, et al. Diagnosis of perioperative myocardial 37. Newby LK, Christenson RH, Ohman EM, Armstrong PW, Thomp-
infarction with measurement of cardiac troponin I. N Engl J Med son TD, Lee KL, et al. Value of serial troponin T measures for
1994;330:670 – 4. early and late risk stratification in patients with acute coronary
21. Katus HA, Schoeppenthau M, Tanzeem A, Bauer HG, Saggau W, syndromes. The GUSTO-IIa Investigators. Circulation 1998;98:
Diederich KW, et al. Non-invasive assessment of perioperative 1853–9.
myocardial cell damage by circulating cardiac troponin T. Br 38. Kontos MC, Anderson FP, Schmidt KA, Ornato JP, Tatum JL,
Heart J 1991;65:259 – 64. Jesse RL. Early diagnosis of acute myocardial infarction in
22. Apple FS, Falahati A, Paulsen PR, Miller EA, Sharkey SW. patients without ST-segment elevation. Am J Cardiol 1999;83:
Improved detection of minor ischemic myocardial injury with 155– 8.
measurement of serum cardiac troponin I. Clin Chem 1997;43: 39. Macrae AR, Kavsak PA, Lustig V, Bhargava R, Vandersluis R,
2047–51. Palomaki GE, et al. Assessing the requirement for the 6-hour
23. Ravkilde J, Horder M, Gerhardt W, Ljungdahl L, Pettersson T, interval between specimens in the American Heart Association
Tryding N, et al. Diagnostic performance and prognostic value of Classification of Myocardial Infarction in Epidemiology and Clin-
serum troponin T in suspected acute myocardial infarction. ical Research Studies. Clin Chem 2006;52:812– 8.
Scand J Clin Lab Invest 1993;53:677– 85. 40. McCord J, Nowak RM, McCullough PA, Foreback C, Borzak S,
24. Hamm CW, Ravkilde J, Gerhardt W, Jorgensen P, Peheim E, Tokarski G, et al. Ninety-minute exclusion of acute myocardial
Ljungdahl L, et al. The prognostic value of serum troponin T in infarction by use of quantitative point-of-care testing of myoglobin
unstable angina. N Engl J Med 1992;327:146 –50. and troponin I. Circulation 2001;104:1483– 8.
25. Hamm CW, Braunwald E. A classification of unstable angina 41. Fesmire FM, Fesmire CE. Improved identification of acute coro-
revisited. Circulation 2000;102:118 –22. nary syndromes with second generation cardiac troponin I assay:
26. Heidenreich PA, Alloggiamento T, Melsop K, McDonald KM, Go utility of 2-hour delta cTnI ⬎ or ⫽ ⫹0.02 ng/mL. J Emerg Med
AS, Hlatky MA. The prognostic value of troponin in patients with 2002;22:147–52.
42. Fesmire FM, Christenson RH, Fody EP, Feintuch TA. Delta unstable angina: a comparative analysis. CAPTURE Investiga-
creatine kinase-MB outperforms myoglobin at two hours during tors. Chimeric c7E3 AntiPlatelet Therapy in Unstable angina
the emergency department identification and exclusion of tropo- REfractory to standard treatment trial. J Am Coll Cardiol 2000;
nin positive non-ST-segment elevation acute coronary syn- 35:1535– 42.
dromes. Ann Emerg Med 2004;44:12–9. 60. James SK, Armstrong P, Barnathan E, Califf R, Lindahl B,
43. Jaffe AS, Ravkilde J, Roberts R, Naslund U, Apple FS, Galvani M, Siegbahn A, et al. Troponin and C-reactive protein have different
et al. It’s time for a change to a troponin standard. Circulation relations to subsequent mortality and myocardial infarction after
2000;102:1216 –20. acute coronary syndrome: a GUSTO-IV substudy. J Am Coll
44. Wright SA, Sawyer DB, Sacks DB, Chyu S, Goldhaber SZ. Cardiol 2003;41:916 –24.
Elevation of troponin I levels in patients without evidence of 61. Ohman EM, Armstrong PW, Christenson RH, Granger CB, Katus
myocardial injury. JAMA 1997;278:2144. HA, Hamm CW, et al. Cardiac troponin T levels for risk stratifica-
45. Jaffe AS. Elevations in cardiac troponin measurements: false tion in acute myocardial ischemia. GUSTO IIA Investigators.
false-positives: the real truth. Cardiovasc Toxicol 2001;1:87–92. N Engl J Med 1996;335:1333– 41.
46. Antman EM, Grudzien C, Mitchell RN, Sacks DB. Detection of 62. Ohman EM, Armstrong PW, White HD, Granger CB, Wilcox RG,
unsuspected myocardial necrosis by rapid bedside assay for Weaver WD, et al. Risk stratification with a point-of-care cardiac
cardiac troponin T. Am Heart J 1997;133:596 – 8. troponin T test in acute myocardial infarction. GUSTOIII Investi-
47. Morrow DA. Troponins in patients with acute coronary syn- gators. Global Use of Strategies To Open Occluded Coronary
dromes: biologic, diagnostic, and therapeutic implications. Car- Arteries. Am J Cardiol 1999;84:1281– 6.
diovasc Toxicol 2001;1:105–10. 63. Lindahl B, Venge P, Wallentin L. Relation between troponin T and
48. Chen Y, Serfass RC, Mackey-Bojack SM, Kelly KL, Titus JL, Apple the risk of subsequent cardiac events in unstable coronary artery
FS. Cardiac troponin T alterations in myocardium and serum of disease. The FRISC study group. Circulation 1996;93:1651–7.
rats after stressful, prolonged intense exercise. J Appl Physiol 64. Morrow DA, Rifai N, Tanasijevic MJ, Wybenga DR, de Lemos JA,
2000;88:1749 –55. Antman EM. Clinical efficacy of three assays for cardiac troponin
49. Wu AH, Ford L. Release of cardiac troponin in acute coronary I for risk stratification in acute coronary syndromes: a Thrombol-
syndromes: ischemia or necrosis? Clin Chim Acta 1999;284: ysis In Myocardial Infarction (TIMI) 11B substudy. Clin Chem
161–74. 2000;46:453– 60.
50. Benamer H, Steg PG, Benessiano J, Vicaut E, Gaultier CJ, Aubry 65. Kaul P, Newby LK, Fu Y, Hasselblad V, Mahaffey KW, Christenson
P, et al. Elevated cardiac troponin I predicts a high-risk angio- RH, et al. Troponin T and quantitative ST-segment depression
graphic anatomy of the culprit lesion in unstable angina. Am offer complementary prognostic information in the risk stratifica-
Heart J 1999;137:815–20. tion of acute coronary syndrome patients. J Am Coll Cardiol
51. Heeschen C, van Den Brand MJ, Hamm CW, Simoons ML. 2003;41:371– 80.
Angiographic findings in patients with refractory unstable angina 66. Morrow DA, Cannon CP, Rifai N, Frey MJ, Vicari R, Lakkis N, et al.
according to troponin T status. Circulation 1999;100:1509 –14. Ability of minor elevations of troponin I and T to identify patients
52. Wong GC, Morrow DA, Murphy S, Kraimer N, Pai R, James D, et with unstable angina and non-ST elevation myocardial infarction
al. Elevations in troponin T and I are associated with abnormal who benefit from an early invasive strategy: results from a
tissue level perfusion: a TACTICS-TIMI 18 substudy. Treat Angina prospective, randomized trial. JAMA 2001;286:2405–12.
with Aggrastat and Determine Cost of Therapy with an Invasive or 67. Morrow DA, Antman EM, Tanasijevic M, Rifai N, de Lemos JA,
Conservative Strategy-Thrombolysis in Myocardial Infarction. Cir- McCabe CH, et al. Cardiac troponin I for stratification of early
culation 2002;106:202–7. outcomes and the efficacy of enoxaparin in unstable angina: a
53. Lindahl B, Diderholm E, Lagerqvist B, Venge P, Wallentin L. TIMI 11B substudy. J Am Coll Cardiol 2000;36:1812–7.
Mechanisms behind the prognostic value of troponin T in unsta- 68. Olatidoye AG, Wu AH, Feng YJ, Waters D. Prognostic role of
ble coronary artery disease: a FRISC II substudy. J Am Coll troponin T versus troponin I in unstable angina pectoris for
Cardiol 2001;38:979 – 86. cardiac events with meta-analysis comparing published studies.
54. Matetzky S, Sharir T, Domingo M, Noc M, Chyu KY, Kaul S, et al. Am J Cardiol 1998;81:1405–10.
Elevated troponin I level on admission is associated with adverse 69. Apple FS, Wu AH, Jaffe AS. European Society of Cardiology and
outcome of primary angioplasty in acute myocardial infarction. American College of Cardiology guidelines for redefinition of
Circulation 2000;102:1611– 6. myocardial infarction: how to use existing assays clinically and
55. Topol EJ. Inflammation and embolization in ischemic heart for clinical trials. Am Heart J 2002;144:981– 6.
disease. J Invasive Cardiol 2000;12(Suppl B):2B–7B. 70. Venge P, Lagerqvist B, Diderholm E, Lindahl B, Wallentin L.
56. Sobel BE, Bresnahan GF, Shell WE, Yoder RD. Estimation of Clinical performance of three cardiac troponin assays in patients
infarct size in man and its relation to prognosis. Circulation with unstable coronary artery disease (a FRISC II substudy). Am J
1972;46:640 – 8. Cardiol 2002;89:1035– 41.
57. Antman EM, Tanasijevic MJ, Thompson B, Schactman M, Mc- 71. Morrow DA, Rifai N, Sabatine MS, Ayanian S, Murphy SA, de
Cabe CH, Cannon CP, et al. Cardiac-specific troponin I levels to Lemos JA, et al. Evaluation of the AccuTnI cardiac troponin I
predict the risk of mortality in patients with acute coronary assay for risk assessment in acute coronary syndromes. Clin
syndromes. N Engl J Med 1996;335:1342–9. Chem 2003;49:1396 – 8.
58. Alexander JH, Sparapani RA, Mahaffey KW, Deckers JW, Newby 72. Kontos MC, Shah R, Fritz LM, Anderson FP, Tatum JL, Ornato JP,
LK, Ohman EM, et al. Association between minor elevations of et al. Implication of different cardiac troponin I levels for clinical
creatine kinase-MB level and mortality in patients with acute outcomes and prognosis of acute chest pain patients. J Am Coll
coronary syndromes without ST-segment elevation. PURSUIT Cardiol 2004;43:958 – 65.
Steering Committee. Platelet Glycoprotein IIb/IIIa in Unstable 73. de Lemos JA, McGuire DK, Drazner MH. B-type natriuretic peptide
Angina: Receptor Suppression Using Integrilin Therapy. JAMA in cardiovascular disease. Lancet 2003;362:316 –22.
2000;283:347–53. 74. de Lemos JA, Morrow DA. Brain natriuretic peptide measurement
59. Heeschen C, Hamm CW, Bruemmer J, Simoons ML. Predictive in acute coronary syndromes: ready for clinical application?
value of C-reactive protein and troponin T in patients with Circulation 2002;106:2868 –70.
75. Hama N, Itoh H, Shirakami G, Nakagawa O, Suga S, Ogawa Y, et 90. Morita E, Yasue H, Yoshimura M, Ogawa H, Jougasaki M,
al. Rapid ventricular induction of brain natriuretic peptide gene Matsumura T, et al. Increased plasma levels of brain natriuretic
expression in experimental acute myocardial infarction. Circula- peptide in patients with acute myocardial infarction. Circulation
tion 1995;92:1558 – 64. 1993;88:82–91.
76. Toth M, Vuorinen KH, Vuolteenaho O, Hassinen IE, Uusimaa PA, 91. Arakawa N, Nakamura M, Aoki H, Hiramori K. Relationship
Leppaluoto J, et al. Hypoxia stimulates release of ANP and BNP between plasma level of brain natriuretic peptide and myocardial
from perfused rat ventricular myocardium. Am J Physiol 1994; infarct size. Cardiology 1994;85:334 – 40.
266:H1572– 80. 92. Nagaya N, Nishikimi T, Goto Y, Miyao Y, Kobayashi Y, Morii I, et
77. Marumoto K, Hamada M, Hiwada K. Increased secretion of atrial al. Plasma brain natriuretic peptide is a biochemical marker for
and brain natriuretic peptides during acute myocardial ischaemia the prediction of progressive ventricular remodeling after acute
induced by dynamic exercise in patients with angina pectoris. myocardial infarction. Am Heart J 1998;135:21– 8.
Clin Sci (Colch) 1995;88:551– 6. 93. Richards AM, Nicholls MG, Yandle TG, Ikram H, Espiner EA,
78. Tateishi J, Masutani M, Ohyanagi M, Iwasaki T. Transient in- Turner JG, et al. Neuroendocrine prediction of left ventricular
crease in plasma brain (B-type) natriuretic peptide after percuta- function and heart failure after acute myocardial infarction. The
neous transluminal coronary angioplasty. Clin Cardiol 2000;23: Christchurch Cardioendocrine Research Group. Heart 1999;81:
776 – 80. 114 –20.
79. Arakawa N, Nakamura M, Aoki H, Hiramori K. Plasma brain 94. Richards AM, Nicholls MG, Espiner EA, Lainchbury JG, Troughton
natriuretic peptide concentrations predict survival after acute RW, Elliott J, et al. B-type natriuretic peptides and ejection
myocardial infarction. J Am Coll Cardiol 1996;27:1656 – 61. fraction for prognosis after myocardial infarction. Circulation
80. Omland T, Aakvaag A, Bonarjee VV, Caidahl K, Lie RT, Nilsen DW, 2003;107:2786 –92.
et al. Plasma brain natriuretic peptide as an indicator of left 95. Omland T, de Lemos JA, Morrow DA, Antman EM, Cannon CP,
ventricular systolic function and long-term survival after acute Hall C, et al. Prognostic value of N-terminal pro-atrial and
myocardial infarction. Comparison with plasma atrial natriuretic pro-brain natriuretic peptide in patients with acute coronary
peptide and N- terminal proatrial natriuretic peptide. Circulation syndromes: a TIMI 11B substudy. Am J Cardiol 2002;89:463–5.
1996;93:1963–9. 96. James S, Armstrong P, Califf R, Simoons ML, Venge P, Wallentin
81. Richards AM, Nicholls MG, Yandle TG, Frampton C, Espiner EA, L, et al. Troponin T levels and risk of 30-day outcomes in patients
Turner JG, et al. Plasma N-terminal pro-brain natriuretic peptide with the acute coronary syndrome: prospective verification in the
and adrenomedullin: new neurohormonal predictors of left ven- GUSTO-IV trial. Am J Med 2003;115:178 – 84.
tricular function and prognosis after myocardial infarction. Circu- 97. Heeschen C, Hamm CW, Mitrovic V, Lantelme NH, White HD.
lation 1998;97:1921–9. N-terminal pro-B-type natriuretic peptide levels for dynamic risk
82. Talwar S, Squire IB, Downie PF, McCullough AM, Campton MC, stratification of patients with acute coronary syndromes. Circu-
Davies JE, et al. Profile of plasma N-terminal proBNP following lation 2004;110:3206 –12.
acute myocardial infarction; correlation with left ventricular sys- 98. Lindahl B, Lindback J, Jernberg T, Johnston N, Stridsberg M,
tolic dysfunction. Eur Heart J 2000;21:1514 –21. Venge P, et al. Serial analyses of N-terminal pro-B-type natriuretic
83. de Lemos JA, Morrow DA, Bentley JH, Omland T, Sabatine MS, peptide in patients with non-ST-segment elevation acute coro-
McCabe CH, et al. The prognostic value of B-type natriuretic nary syndromes: a Fragmin and fast Revascularisation during In
peptide in patients with acute coronary syndromes. N Engl J Med Stability in Coronary artery disease (FRISC)-II substudy. J Am Coll
2001;345:1014 –21. Cardiol 2005;45:533– 41.
84. Morrow DA, de Lemos JA, Sabatine MS, Murphy SA, Demopoulos 99. Morrow DA, de Lemos JA, Blazing MA, Sabatine MS, Murphy SA,
L, DiBattiste P, et al. Evaluation of B-type natriuretic peptide for Jarolim P, et al. Prognostic value of serial B-type natriuretic
risk assessment in unstable angina/non-ST elevation MI: BNP peptide testing during follow-up of patients with unstable coro-
and prognosis in TACTICS-TIMI 18. J Am Coll Cardiol 2003;41: nary artery disease. JAMA 2005;294:2866 –71.
1264 –72. 100. James SK, Lindback J, Tilly J, Siegbahn A, Venge P, Armstrong P,
85. Omland T, de Lemos JA, Morrow DA, Antman EM, Cannon CP, et al. Troponin-T and N-terminal pro-B-type natriuretic peptide
Hall C, et al. Prognostic value of N-terminal pro-atrial and predict mortality benefit from coronary revascularization in acute
pro-brain natriuretic peptide in patients with acute coronary coronary syndromes: a GUSTO-IV substudy. J Am Coll Cardiol
syndromes. Am J Cardiol 2002;89:463–5. 2006;48:1146 –54.
86. Jernberg T, Stridsberg M, Venge P, Lindahl B. N-terminal pro 101. Scirica BM, Morrow DA, Cannon CP, Ray KK, Sabatine MS,
brain natriuretic peptide on admission for early risk stratification Jarolim P, et al. Intensive statin therapy and the risk of hospi-
of patients with chest pain and no ST-segment elevation. J Am talization for heart failure after an acute coronary syndrome in
Coll Cardiol 2002;40:437– 45. the PROVE IT-TIMI 22 Study. J Am Coll Cardiol 2006;47:2326 –
87. Omland T, Persson A, Ng L, O’Brien R, Karlsson T, Herlitz J, et al. 31.
N-terminal pro-B-type natriuretic peptide and long-term mortality 102. Rabbani LE. Acute coronary syndromes— beyond myocyte necro-
in acute coronary syndromes. Circulation 2002;106:2913– 8. sis. N Engl J Med 2001;345:1057–9.
88. Jernberg T, Lindahl B, Siegbahn A, Andren B, Frostfeldt G, 103. Libby P, Ridker PM, Maseri A. Inflammation and atherosclerosis.
Lagerqvist B, et al. N-terminal pro-brain natriuretic peptide in Circulation 2002;105:1135– 43.
relation to inflammation, myocardial necrosis, and the effect of 104. Libby P. Molecular bases of the acute coronary syndromes.
an invasive strategy in unstable coronary artery disease. J Am Circulation 1995;91:2844 –50.
Coll Cardiol 2003;42:1909 –16. 105. Morrow DA, Ridker PM. Inflammation in cardiovascular disease.
89. James SK, Wallentin L, Armstrong PW, Barnathan ES, Califf RM, In: Topol E, ed. Textbook of Cardiovascular Medicine Updates.
Lindahl B, et al. N-terminal pro brain natriuretic peptide and other Cedar Knolls: Lippincott Williams & Wilkins, 1999:1–12.
risk markers for the separate prediction of mortality and subse- 106. Blake GJ, Ridker PM. C-reactive protein and other inflammatory
quent myocardial infarction in patients with unstable coronary risk markers in acute coronary syndromes. J Am Coll Cardiol
disease: a GUSTO IV substudy. Circulation 2003;108:275– 81. 2003;41:37S– 42S.
107. Liuzzo G, Biasucci LM, Gallimore JR, Grillo RL, Rebuzzi AG, Pepys with unstable angina predict recurrent instability. Circulation
MB, et al. The prognostic value of C-reactive protein and serum 1999;99:855– 60.
amyloid a protein in severe unstable angina. N Engl J Med 124. Lindahl B, Toss H, Siegbahn A, Venge P, Wallentin L. Markers of
1994;331:417–24. myocardial damage and inflammation in relation to long-term
108. Berk BC, Weintraub WS, Alexander RW. Elevation of C-reactive mortality in unstable coronary artery disease. FRISC Study
protein in “active” coronary artery disease. Am J Cardiol 1990; Group. Fragmin during Instability in Coronary Artery Disease.
65:168 –72. N Engl J Med 2000;343:1139 – 47.
109. Morrow DA, Rifai N, Antman EM, Weiner DL, McCabe CH, Cannon 125. Mueller C, Buettner HJ, Hodgson JM, Marsch S, Perruchoud AP,
CP, et al. C-reactive protein is a potent predictor of mortality Roskamm H, et al. Inflammation and long-term mortality after
independently of and in combination with troponin T in acute non-ST elevation acute coronary syndrome treated with a very
coronary syndromes: a TIMI 11A substudy. Thrombolysis in early invasive strategy in 1042 consecutive patients. Circulation
Myocardial Infarction. J Am Coll Cardiol 1998;31:1460 –5. 2002;105:1412–5.
110. Biasucci LM, Vitelli A, Liuzzo G, Altamura S, Caligiuri G, Monaco 126. Biasucci LM, Liuzzo G, Fantuzzi G, Caligiuri G, Rebuzzi AG,
C, et al. Elevated levels of interleukin-6 in unstable angina. Ginnetti F, et al. Increasing levels of interleukin (IL)-1Ra and IL-6
Circulation 1996;94:874 –7. during the first 2 days of hospitalization in unstable angina are
111. Brennan ML, Penn MS, Van Lente F, Nambi V, Shishehbor MH, associated with increased risk of in-hospital coronary events.
Aviles RJ, et al. Prognostic value of myeloperoxidase in patients Circulation 1999;99:2079 – 84.
with chest pain. N Engl J Med 2003;349:1595– 604. 127. Lindmark E, Diderholm E, Wallentin L, Siegbahn A. Relationship
112. Morrow DA, Rifai N, Antman EM, Weiner DL, McCabe CH, Cannon between interleukin 6 and mortality in patients with unstable
CP, et al. Serum amyloid A predicts early mortality in acute coronary artery disease: effects of an early invasive or noninva-
coronary syndromes: a TIMI 11A substudy. J Am Coll Cardiol sive strategy. JAMA 2001;286:2107–13.
2000;35:358 – 62. 128. Baldus S, Heeschen C, Meinertz T, Zeiher AM, Eiserich JP,
113. Pietila K, Hermens WT, Harmoinen A, Baardman T, Pasternack A, Munzel T, et al. Myeloperoxidase serum levels predict risk in
Topol EJ, et al. Comparison of peak serum C-reactive protein and patients with acute coronary syndromes. Circulation 2003;108:
hydroxybutyrate dehydrogenase levels in patients with acute 1440 –5.
myocardial infarction treated with alteplase and streptokinase.
129. Pearson TA, Mensah GA, Alexander RW, Anderson JL, Cannon
Am J Cardiol 1997;80:1075–7.
RO 3rd, Criqui M, et al. Markers of inflammation and cardiovas-
114. Scirica BM, Morrow DA. Is C-reactive protein an innocent by- cular disease: application to clinical and public health practice:
stander or proatherogenic culprit? The verdict is still out. Circu- a statement for healthcare professionals from the Centers for
lation 2006;113:2128 –34; discussion 2151. Disease Control and Prevention and the American Heart Associ-
115. Buffon A, Biasucci LM, Liuzzo G, D’Onofrio G, Crea F, Maseri A. ation. Circulation 2003;107:499 –511.
Widespread coronary inflammation in unstable angina. N Engl
130. Ridker PM, Cannon CP, Morrow D, Rifai N, Rose LM, McCabe CH,
J Med 2002;347:5–12.
et al. C-reactive protein levels and outcomes after statin therapy.
116. Liuzzo G, Buffon A, Biasucci LM, Gallimore JR, Caligiuri G, Vitelli
N Engl J Med 2005;352:20 – 8.
A, et al. Enhanced inflammatory response to coronary angio-
131. Khera A, McGuire DK, Murphy SA, Stanek HG, Das SR, Vongpa-
plasty in patients with severe unstable angina. Circulation
tanasin W, et al. Race and gender differences in C-reactive
1998;98:2370 – 6.
protein levels. J Am Coll Cardiol 2005;46:464 –9.
117. Anzai T, Yoshikawa T, Shiraki H, Asakura Y, Akaishi M, Mitamura
132. Albert MA, Glynn RJ, Buring J, Ridker PM. C-reactive protein
H, et al. C-reactive protein as a predictor of infarct expansion and
levels among women of various ethnic groups living in the United
cardiac rupture after a first Q-wave acute myocardial infarction.
States (from the Women’s Health Study). Am J Cardiol 2004;
Circulation 1997;96:778 – 84.
93:1238 – 42.
118. Haverkate F, Thompson SG, Pyke SD, Gallimore JR, Pepys MB.
Production of C-reactive protein and risk of coronary events in 133. Ford ES, Giles WH, Mokdad AH, Myers GL. Distribution and
stable and unstable angina. European Concerted Action on correlates of C-reactive protein concentrations among adult US
Thrombosis and Disabilities Angina Pectoris Study Group. Lancet women. Clin Chem 2004;50:574 – 81.
1997;349:462– 6. 134. Benamer H, Steg PG, Benessiano J, Vicaut E, Gaultier CJ,
119. Toss H, Lindahl B, Siegbahn A, Wallentin L. Prognostic influence Boccara A, et al. Comparison of the prognostic value of C-reac-
of increased fibrinogen and C-reactive protein levels in unstable tive protein and troponin I in patients with unstable angina
coronary artery disease. FRISC Study Group. Fragmin during pectoris. Am J Cardiol 1998;82:845–50.
Instability in Coronary Artery Disease. Circulation 1997;96: 135. de Winter RJ, Fischer J, Bholasingh R, van Straalen JP, de Jong
4204 –10. T, Tijssen JG, et al. C-reactive protein and cardiac troponin T in
120. Rebuzzi AG, Quaranta G, Liuzzo G, Caligiuri G, Lanza GA, risk stratification: differences in optimal timing of tests early
Gallimore JR, et al. Incremental prognostic value of serum levels after the onset of chest pain. Clin Chem 2000;46:1597– 603.
of troponin T and C- reactive protein on admission in patients 136. Ridker PM, Rifai N, Pfeffer MA, Sacks FM, Moye LA, Goldman S,
with unstable angina pectoris. Am J Cardiol 1998;82:715–9. et al. Inflammation, pravastatin, and the risk of coronary events
121. Tommasi S, Carluccio E, Bentivoglio M, Buccolieri M, Mariotti M, after myocardial infarction in patients with average cholesterol
Politano M, et al. C-reactive protein as a marker for cardiac levels. Cholesterol and Recurrent Events (CARE) Investigators.
ischemic events in the year after a first, uncomplicated myocar- Circulation 1998;98:839 – 44.
dial infarction. Am J Cardiol 1999;83:1595–9. 137. Ridker PM, Rifai N, Pfeffer MA, Sacks FM, Braunwald E. Long-
122. Ferreiros ER, Boissonnet CP, Pizarro R, Merletti PFG, Corrado G, term effects of pravastatin on plasma concentration of C-reactive
Cagide A, et al. Independent prognostic value of elevated protein. Circulation 1999;100:230 –5.
C-reactive protein in unstable angina. Circulation 1999;100: 138. Kinlay S, Schwartz GG, Olsson AG, Rifai N, Leslie SJ, Sasiela WJ,
1958 – 63. et al. High-dose atorvastatin enhances the decline in inflamma-
123. Biasucci L, Liuzzo G, Grillo R, Caligiuri G, Rebuzzi A, Buffon A, et tory markers in patients with acute coronary syndromes in the
al. Elevated levels of C-reactive protein at discharge in patients MIRACL study. Circulation 2003;108:1560 – 6.
139. Nissen SE, Tuzcu EM, Schoenhagen P, Crowe T, Sasiela WJ, Tsai 156. Heeschen C, Dimmeler S, Hamm CW, van den Brand MJ,
J, et al. Statin therapy, LDL cholesterol, C-reactive protein, and Boersma E, Zeiher AM, et al. Soluble CD40 ligand in acute
coronary artery disease. N Engl J Med 2005;352:29 –38. coronary syndromes. N Engl J Med 2003;348:1104 –11.
140. Ikonomidis I, Andreotti F, Economou E, Stefanadis C, Toutouzas 157. Bayes-Genis A, Conover CA, Overgaard MT, Bailey KR, Chris-
P, Nihoyannopoulos P. Increased proinflammatory cytokines in tiansen M, Holmes DR, Jr., et al. Pregnancy-associated plasma
patients with chronic stable angina and their reduction by protein A as a marker of acute coronary syndromes. N Engl J Med
aspirin. Circulation 1999;100:793– 8. 2001;345:1022–9.
141. Feldman M, Jialal I, Devaraj S, Cryer B. Effects of low-dose 158. Apple FS, Wu AH, Mair J, Ravkilde J, Panteghini M, Tate J, et al.
aspirin on serum C-reactive protein and thromboxane B2 con- Future biomarkers for detection of ischemia and risk stratifica-
centrations: a placebo-controlled study using a highly sensitive tion in acute coronary syndrome. Clin Chem 2005;51:810 –24.
C-reactive protein assay. J Am Coll Cardiol 2001;37:2036 – 41. 159. Jaffe AS, Katus H. Acute coronary syndrome biomarkers: the
142. Kennon S, Price CP, Mills PG, Ranjadayalan K, Cooper J, Clarke need for more adequate reporting. Circulation 2004;110:104 –
H, et al. The effect of aspirin on C-reactive protein as a marker of 6.
risk in unstable angina. J Am Coll Cardiol 2001;37:1266 –70. 160. Lindahl B, Venge P, Wallentin L. Troponin T identifies patients
143. Jesse RL, Kukreja R. Rationale for the early clinical application of with unstable coronary artery disease who benefit from long-term
markers of ischemia in patients with suspected acute coronary antithrombotic protection. Fragmin in Unstable Coronary Artery
syndromes. Cardiovasc Toxicol 2001;1:125–33. Disease (FRISC) Study Group. J Am Coll Cardiol 1997;29:43– 8.
144. Morrow DA, de Lemos JA, Sabatine MS, Antman EM. The search 161. Hamm CW, Heeschen C, Goldmann B, Vahanian A, Adgey J,
for a biomarker of cardiac ischemia. Clin Chem 2003;49:537–9. Miguel CM, et al. Benefit of abciximab in patients with refractory
145. Bar-Or D, Lau E, Winkler JV. A novel assay for cobalt-albumin unstable angina in relation to serum troponin T levels. c7E3 Fab
binding and its potential as a marker for myocardial ischemia: a Antiplatelet Therapy in Unstable Refractory Angina (CAPTURE)
preliminary report. J Emerg Med 2000;19:311–5. Study Investigators. N Engl J Med 1999;340:1623–9.
146. Bar-Or D, Winkler JV, Vanbenthuysen K, Harris L, Lau E, Hetzel 162. Heeschen C, Hamm CW, Goldmann B, Deu A, Langenbrink L,
FW. Reduced albumin-cobalt binding with transient myocardial White HD. Troponin concentrations for stratification of patients
ischemia after elective percutaneous transluminal coronary an- with acute coronary syndromes in relation to therapeutic efficacy
gioplasty: a preliminary comparison to creatine kinase-MB, myo- of tirofiban. Lancet 1999;354:1757– 62.
globin, and troponin I. Am Heart J 2001;141:985–91. 163. Newby LK, Ohman EM, Christenson RH, Moliterno DJ, Harrington
147. Christenson RH, Duh SH, Sanhai WR, Wu AH, Holtman V, Painter RA, White HD, et al. Benefit of glycoprotein IIb/IIIa inhibition in
P, et al. Characteristics of an Albumin Cobalt Binding Test for patients with acute coronary syndromes and troponin T-positive
assessment of acute coronary syndrome patients: a multicenter status: the paragon-B troponin T substudy. Circulation 2001;
study. Clin Chem 2001;47:464 –70. 103:2891– 6.
148. Peacock F, Morris DL, Anwaruddin S, Christenson RH, Collinson 164. Januzzi JL, Chae CU, Sabatine MS, Jang IK. Elevation in serum
PO, Goodacre SW, et al. Meta-analysis of ischemia-modified troponin I predicts the benefit of tirofiban. J Thromb Thromboly-
albumin to rule out acute coronary syndromes in the emergency sis 2001;11:211–5.
department. Am Heart J 2006;152:253– 62. 165. Simoons ML. Effect of glycoprotein IIb/IIIa receptor blocker
149. Bhagavan NV, Lai EM, Rios PA, Yang J, Ortega-Lopez AM, abciximab on outcome in patients with acute coronary syn-
Shinoda H, et al. Evaluation of human serum albumin cobalt dromes without early coronary revascularisation: the GUSTO
binding assay for the assessment of myocardial ischemia and IV-ACS randomised trial. Lancet 2001;357:1915–24.
myocardial infarction. Clin Chem 2003;49:581–5. 166. FRISC II Investigators. Invasive compared with non-invasive
150. Bar-Or D, Curtis G, Rao N, Bampos N, Lau E. Characterization of treatment in unstable coronary-artery disease: FRISC II prospec-
the Co(2⫹) and Ni(2⫹) binding amino-acid residues of the tive randomised multicentre study. Lancet 1999;354:708 –15.
N-terminus of human albumin. An insight into the mechanism of 167. Wu AH, Boden WE, McKay RG. Long-term follow-up of patients
a new assay for myocardial ischemia. Eur J Biochem 2001;268: with increased cardiac troponin concentrations following percu-
42–7. taneous coronary intervention. Am J Cardiol 2002;89:1300 –2.
151. Kleinfeld AM, Prothro D, Brown DL, Davis RC, Richieri GV, 168. Ganz W, Buchbinder N, Marcus H, Mondkar A, Maddahi J,
DeMaria A. Increases in serum unbound free fatty acid levels Charuzi Y, et al. Intracoronary thrombolysis in evolving myocar-
following coronary angioplasty. Am J Cardiol 1996;78:1350 – 4. dial infarction. Am Heart J 1981;101:4 –13.
152. Danne O, Mockel M, Lueders C, Mugge C, Zschunke GA, Lufft H, 169. Rentrop P, Blanke H, Karsch KR, Kaiser H, Kostering H, Leitz K.
et al. Prognostic implications of elevated whole blood choline Selective intracoronary thrombolysis in acute myocardial infarc-
levels in acute coronary syndromes. Am J Cardiol 2003;91: tion and unstable angina pectoris. Circulation 1981;63:307–17.
1060 –7. 170. Christenson R, Ohman E, Topol E, Peck S, Newby L, Duh S, et al.
153. Apple FS, Quist HE, Otto AP, Mathews WE, Murakami MM. Assessment of coronary reperfusion after thrombolysis with a
Release characteristics of cardiac biomarkers and ischemia- model combining myoglobin, creatine kinase-MB, and clinical
modified albumin as measured by the albumin cobalt-binding variables. Circulation 1997;96:1776 – 82.
test after a marathon race. Clin Chem 2002;48:1097–100. 171. Stewart J, French J, Theroux P, Ramanthan K, Solymoss B,
154. Refaai MA, Wright RW, Parvin CA, Gronowski AM, Scott MG, Eby Johnson R, et al. Early noninvasive identification of failed re-
CS. Ischemia-modified albumin increases after skeletal muscle perfusion after intravenous thrombolytic therapy in acute myo-
ischemia during arthroscopic knee surgery. Clin Chim Acta cardial infarction. J Am Coll Cardiol 1998;31:1499 –1505.
2006;366:264 – 8. 172. Tanasijevic MJ, Cannon CP, Antman EM, Wybenga DR, Fischer
155. Sabatine MS, Morrow DA, de Lemos JA, Gibson CM, Murphy SA, GA, Grudzien C, et al. Myoglobin, creatine-kinase-MB and cardiac
Rifai N, et al. Multimarker approach to risk stratification in troponin-I 60-minute ratios predict infarct-related artery patency
non-ST elevation acute coronary syndromes: simultaneous as- after thrombolysis for acute myocardial infarction: results from
sessment of troponin I, C-reactive protein, and B-type natriuretic the Thrombolysis in Myocardial Infarction study (TIMI) 10B. J Am
peptide. Circulation 2002;105:1760 –3. Coll Cardiol 1999;34:739 – 47.
173. Klootwijk P, Langer A, Meij S, Green C, Veldkamp RF, Ross AM, recovering from unstable angina pectoris. Am J Cardiol 2001;
et al. Non-invasive prediction of reperfusion and coronary artery 87:1235–9.
patency by continuous ST segment monitoring in the GUSTO-I 188. Oltrona L, Ottani F, Galvani M. Clinical significance of a single
trial. Eur Heart J 1996;17:689 –98. measurement of troponin-I and C-reactive protein at admission in
174. de Lemos JA, Morrow DA, Gibson CM, Murphy SA, Rifai N, 1773 consecutive patients with acute coronary syndromes. Am
Tanasijevic M, et al. Early noninvasive detection of failed epicar- Heart J 2004;148:405–15.
dial reperfusion after fibrinolytic therapy. Am J Cardiol 2001;88: 189. de Winter RJ, Bholasingh R, Lijmer JG, Koster RW, Gorgels JP,
353– 8. Schouten Y, et al. Independent prognostic value of C-reactive
175. Roberts R, Henry PD, Sobel BE. An improved basis for enzymatic protein and troponin I in patients with unstable angina or
estimation of infarct size. Circulation 1975;52:743–54. non-Q-wave myocardial infarction. Cardiovasc Res 1999;42:
176. Grande P, Hansen BF, Christiansen C, Naestoft J. Estimation of 240 –5.
acute myocardial infarct size in man by serum CK-MB measure- 190. Mulvihill NT, Foley JB, Murphy RT, Curtin R, Crean PA, Walsh M.
ments. Circulation 1982;65:756 – 64. Risk stratification in unstable angina and non-Q wave myocardial
177. Christenson RH, Vollmer RT, Ohman EM, Peck S, Thompson TD, infarction using soluble cell adhesion molecules. Heart 2001;
Duh SH, et al. Relation of temporal creatine kinase-MB release 85:623–7.
and outcome after thrombolytic therapy for acute myocardial 191. Bholasingh R, Cornel JH, Kamp O, van Straalen JP, Sanders GT,
infarction. TAMI Study Group. Am J Cardiol 2000;85:543–7. Dijksman L, et al. The prognostic value of markers of inflamma-
178. Ingkanisorn WP, Rhoads KL, Aletras AH, Kellman P, Arai AE. tion in patients with troponin T-negative chest pain before
Gadolinium delayed enhancement cardiovascular magnetic res- discharge from the emergency department. Am J Med 2003;
onance correlates with clinical measures of myocardial infarc- 115:521– 8.
tion. J Am Coll Cardiol 2004;43:2253–9.
192. Bodi V, Sanchis J, Llacer A, Facila L, Nunez J, Bertomeu V, et al.
179. Ricchiuti V, Sharkey SW, Murakami MM, Voss EM, Apple FS.
Risk stratification in non-ST elevation acute coronary syndromes:
Cardiac troponin I and T alterations in dog hearts with myocardial
predictive power of troponin I, C-reactive protein, fibrinogen and
infarction: correlation with infarct size. Am J Clin Pathol 1998;
homocysteine. Int J Cardiol 2005;98:277– 83.
110:241–7.
193. Versaci F, Gaspardone A, Tomai F, Crea F, Chiariello L, Gioffre
180. Tanaka H, Abe S, Yamashita T, Arima S, Saigo M, Nakao S, et al.
PA. Predictive value of C-reactive protein in patients with unsta-
Serum levels of cardiac troponin I and troponin T in estimating
ble angina pectoris undergoing coronary artery stent implanta-
myocardial infarct size soon after reperfusion. Coron Artery Dis
tion. Am J Cardiol 2000;85:92–5, A8.
1997;8:433–9.
181. Licka M, Zimmermann R, Zehelein J, Dengler TJ, Katus HA, 194. Zebrack JS, Anderson JL, Maycock CA, Horne BD, Bair TL,
Kubler W. Troponin T concentrations 72 hours after myocardial Muhlestein JB. Usefulness of high-sensitivity C-reactive protein
infarction as a serological estimate of infarct size. Heart 2002; in predicting long-term risk of death or acute myocardial infarc-
87:520 – 4. tion in patients with unstable or stable angina pectoris or acute
182. Apple FS, Murakami MM. Cardiac troponin and creatine kinase myocardial infarction. Am J Cardiol 2002;89:145–9.
MB monitoring during in-hospital myocardial reinfarction. Clin 195. Sanchez PL, Morinigo JL, Pabon P, Martin F, Piedra I, Palacios IF,
Chem 2005;51:460 –3. et al. Prognostic relations between inflammatory markers and
183. Darbar D, Davidson NC, Gillespie N, Choy AM, Lang CC, Shyr Y, mortality in diabetic patients with non-ST elevation acute coro-
et al. Diagnostic value of B-type natriuretic peptide concentra- nary syndrome. Heart 2004;90:264 –9.
tions in patients with acute myocardial infarction. Am J Cardiol 196. Pietila KO, Harmoinen AP, Jokiniitty J, Pasternack AI. Serum
1996;78:284 –7. C-reactive protein concentration in acute myocardial infarction
184. Crilley JG, Farrer M. Left ventricular remodelling and brain and its relationship to mortality during 24 months of follow-up in
natriuretic peptide after first myocardial infarction. Heart 2001; patients under thrombolytic treatment. Eur Heart J 1996;17:
86:638 – 42. 1345–9.
185. Oltrona L, Ardissino D, Merlini PA, Spinola A, Chiodo F, Pezzano 197. Nikfardjam M, Mullner M, Schreiber W, Oschatz E, Exner M,
A. C-reactive protein elevation and early outcome in patients with Domanovits H, et al. The association between C-reactive protein
unstable angina pectoris. Am J Cardiol 1997;80:1002– 6. on admission and mortality in patients with acute myocardial
186. Oltrona L, Merlini PA, Savonitto S, Broccolino M, Giarratana G, infarction. J Intern Med 2000;247:341–5.
Pezzano A, et al. Lack of correlation between activation of 198. Mega JL, Morrow DA, De Lemos JA, Sabatine MS, Murphy SA,
hemostatic mechanism and inflammation in unstable angina Rifai N, et al. B-type natriuretic peptide at presentation and
pectoris. J Thromb Thrombolysis 1998;5:169 –73. prognosis in patients with ST-segment elevation myocardial
187. Bazzino O, Ferreiros ER, Pizarro R, Corrado G. C-reactive protein infarction: an ENTIRE-TIMI-23 substudy. J Am Coll Cardiol 2004;
and the stress tests for the risk stratification of patients 44:335–9.
Clinical Chemistry 53:4
575–580 (2007) Current Issues in
Laboratory Medicine
Evaluation of the Performance of Randomized

versus Fixed Time Schedules for Quality Control
Procedures
Curtis A. Parvin1 and Sanford Robbins III2*
Background: Although minimum regulatory standards vals centered on the desired length of time between QC
exist for determining QC testing frequency, decisions events with fixed time intervals.
regarding when and how to run QC samples are not © 2007 American Association for Clinical Chemistry
standardized. Most QC testing strategies test control
samples at fixed time intervals, often placing the sam- Traditionally, performance evaluation of QC strategies
ples in the same position on an instrument during has focused on the power of QC rules to detect out-of-
subsequent QC events and leaving large gaps of time control error conditions when QC testing is performed.
when control samples are never run, yet patient samples Relatively little has been written regarding the effect of
are being tested. the frequency of QC testing on QC performance, although
Methods: Mathematical derivations and computer CLIA regulations set minimum standards for QC fre-
simulation were used to determine the expected wait- quency (1 ). Specifically, QC testing is required to be
ing time between an out-of-control condition and the performed during “each day of testing” and must follow
next scheduled QC test for various QC testing strate- the manufacturer’s instructions for control testing if they
gies that use fixed or random intervals between QC exceed these requirements.
tests. Until equivalent QC testing has been scientifically
Results: Scheduling QC tests at fixed intervals yields an proven to supplant the need for QC or unless a control
average time between the occurrence of an out-of-con- sample is run with every patient sample, traditional QC
trol error condition and the next scheduled QC test that strategies will leave intervals of time during which pa-
is equal to half of the fixed time interval. This perfor- tients are being tested, but controls are not running. Of
mance was the best among the QC scheduling strategies specific concern is a tendency for most laboratories to run
investigated. Near-optimal performance, however, was controls at relatively fixed intervals that are the same each
achieved by randomly selecting time intervals between day. The net effect of this strategy is to allow certain
QC events centered on the desired expected interval periods of time in the analytical testing process when
length, a method that provides variation in QC testing patient testing is performed, yet QC samples are never
times throughout the day. run.
Conclusions: If the goal is to vary QC testing times Workflow in the laboratory can be broadly separated
throughout the day while maintaining the shortest ex- into batch mode processes and continuous mode pro-
pected length of time between error conditions and the cesses. Most of the theory of how to design a good QC
next scheduled QC test, then a near-optimal QC sched- strategy has been based on batch mode processes, for
uling strategy combines randomly selected time inter- which the batch is often referred to as an analytical run.
For batch mode processes, the 2 main design parameters
that determine the performance of a QC procedure are the
number of QC samples in the analytical run and the QC
1
Department of Pathology and Immunology, Washington University rules that are applied to the control sample results.
School of Medicine, Saint Louis, MO.
2
Department of Pathology, Anne Arundel Medical Center, 2001 Medical
For continuous mode processes, a 3rd important de-
Parkway, Annapolis, MD. sign parameter that must be specified is when to schedule
* Address correspondence to this author at: Department of Pathology, QC testing. During the operation of a continuous mode
Anne Arundel Medical Center, 2001 Medical Pkwy., Annapolis, MD 21401. Fax
process, events will occur that are known to pose an
443-481-4207; e-mail srobbins@aahs.org.
Received November 17, 2006; accepted January 23, 2007. increased risk of system malfunction, such as reagent
Previously published online at DOI: 10.1373/clinchem.2006.083311 change or system maintenance. QC testing is generally
575
576 Parvin and Robbins: Quality Control Schedules
scheduled immediately after these events. During the 1 or more assayed QC samples with subsequent QC rule
operation of a continuous mode process, however, there is evaluation.
also the chance that an out-of-control error condition may The length of fixed intervals between QC events is
occur unpredictably, at any point in time, and adversely denoted I. For QC events randomly placed within fixed
affect subsequent patient samples until the error condi- intervals (strategy 2), the time of the QC event measured
tion is detected and corrected. from the beginning of the interval is denoted ␥I, where ␥
We investigated different strategies for specifying is a random variable with probability distribution g(␥),
when QC testing is performed in situations in which there 0 ⱕ ␥ ⱕ 1. The time to the next QC event for randomly
is no expectation of an increased risk of system malfunc- scheduled QC intervals (strategies 3 and 4) is defined as
tion at any particular time during the day. All strategies I(1 ⫹ ␦) from the current QC event, where ␦ is a random
are required to have the same rate of QC testing (the same variable with mean ⫽ 0 and probability distribution g(␦),
average interval length between QC events). The length of ␦ ⱖ ⫺1.
time that a process is out-of-control after an error condi- A convenient probability distribution to use for the
tion occurs can be divided into 2 segments: the interval random scheduling of QC events is the beta distribution,
from the onset of the error condition until the next denoted Beta(a,b). The beta distribution is defined over
scheduled QC testing time, and the interval from the 1st the range (0,1). When the 2 parameters are equal, the
QC testing time after the error condition occurs until the distribution Beta(a,a) is symmetric, with mean ⫽ 1⁄2 and
error condition is detected. The length of the 1st time variance ⫽ 1/[4(2a ⫹ 1)]. Beta(1,1) is equivalent to a
segment depends solely on the QC scheduling strategy. Uniform(0,1) distribution. As the parameter a increases,
The length of the 2nd time segment is dependent on the the variance of the distribution decreases and the shape of
type of out-of-control error condition, the power of the the distribution looks more like a bell-shaped curve. Fig.
QC rules to detect the error, and the average length of 1 shows examples of the Beta(a,a) distribution for various
values of the parameter a.
time between QC events. Therefore, the primary outcome
Random scheduling for strategy 2 is demonstrated by
measure of interest in this study was the effect of QC
use of ␥ values randomly selected from Beta(1,1) and
scheduling on the expected length of time between the
Beta(3,3) distributions. Random scheduling for strategies
occurrence of an out-of-control error condition and the
3 and 4 is demonstrated with ␦ values that are computed
next scheduled QC testing time.
as ␦ ⫽ w(2x ⫺ 1), where w is a fixed constant between 0
A 2nd outcome measure of interest was the length of
and 1 and x is Beta(1,1) or Beta(3,3). Thus, ␦ will have a
time to the next scheduled QC event as a function of the
mean ⫽ 0 and will be randomly distributed between ⫾w.
time of day. Ideally, the expected length of time to the
The primary performance measure of interest is the
next scheduled QC event (or since the last scheduled QC
expected length of time between the occurrence of an
event) should be independent of the time of day that a
out-of-control error condition and the next scheduled QC
patient sample is tested. The goal is to identify strategies event. It is assumed that the probability of occurrence of
that have near optimal performance with respect to the an out-of-control error condition can be modeled by an
lag time between the occurrence of an out-of-control error exponential distribution with the mean time between
condition and the next scheduled QC testing time while out-of-control error conditions large compared with the
also allowing QC testing to be performed at varying times
throughout the day.
For some laboratory batch mode processes, QC posi-
tion within the batch may play a role similar to that of QC
scheduling over time in a continuous workflow system.
Examples might include batch testing performed with
samples on a multiple-well plate or on a wheel that holds
multiple samples in numbered positions. We focus on
continuous workflow testing systems and do not further
discuss batch mode testing.
Materials and Methods

The different approaches to QC testing schedules that we
evaluated can be categorized into 4 strategies: strategy 1,
QC events scheduled at fixed time intervals; strategy 2,
QC events randomly scheduled within fixed time inter-
vals; strategy 3, QC events scheduled at random intervals;
and strategy 4, 1 QC event scheduled at a random
interval, followed by a series of N QC events scheduled at Fig. 1. Examples of the Beta(a,a) frequency distribution for different
fixed intervals. For purposes of this study, a QC event is values of a.
interval of time between QC events. The probability that Supplement.) Letting TQ represent the time from the
an out-of-control error condition occurs in a particular occurrence of an out-of-control error condition until the
interval between 2 QC events is proportional to the length next scheduled QC event, the expected times for the 4
of the interval. In addition, the position within the partic- general strategies were as follows:
ular interval where the error condition occurs will be
approximately uniformly distributed over the interval (2 ). I
E (TQ) ⫽ (1)
These assumptions are used to mathematically derive and 2
冉冊
confirm by computer simulation estimates for the average
length of time between the random occurrence of an I
E (TQ) ⫽ 1 ⫹ 2 var (␥) (2)
out-of-control condition and the next scheduled QC 2
event.
Computer simulations set the average interval between
QC events at 8 h for all of the evaluated scheduling E (TQ) ⫽
I
2 冉
1 ⫹ var (␦) 冊 (3)
strategies. The probability of an out-of-control error con-
dition is modeled by use of an exponential distribution
with a mean time between occurrences of out-of-control
error conditions set at 30 days. For each simulated trial,
E (TQ) ⫽
I
2冉1⫹
var (␦)
(N ⫹ 1) 冊 (4)
the time of occurrence of the out-of-control error condi- Fixed interval QC scheduling (strategy 1) gives the mini-
tion is randomly generated and QC event times are mum average length of time between the occurrence of an
scheduled until a QC event time follows the time of out-of-control error condition and the next scheduled QC
occurrence of the error condition. The average difference event, which is equal to half of the fixed interval length.
between the time of the out-of-control error condition and The increase in the expected time between an error
the time of the subsequent QC event is computed from condition and the next scheduled QC event for the ran-
100 000 simulated trials. domly scheduled QC strategies depends on the variance
The secondary outcome measure is the average length of the probability distribution used for the random sched-
of time to the next scheduled QC event as a function of uling. A scheduling strategy that randomly selects the
time of day. For a given QC scheduling strategy, QC event time intervals between QC events (strategy 3) has a
times were scheduled over a 365-day period. The length shorter expected time between the occurrence of an out-
of time to the next QC event was calculated at different of-control error condition and the next scheduled QC
time points throughout the day (12:30 AM to 11:30 PM in event than a scheduling strategy that uses the same
1-h increments) for each of the 365 days, and the average probability distribution to randomly place QC events
length of time was computed. within fixed time intervals (strategy 2). The scheduling
strategy that selects a random interval for the next QC
Results event and then follows it with a series of QC events at
The expected times from the occurrence of an error fixed intervals (strategy 4) has an expected time between
condition to the next scheduled QC event for each of the the occurrence of an out-of-control error condition and the
4 basic types of QC testing schedules were derived next scheduled QC event that decreases as the number of
mathematically. The details of the derivations are given in fixed interval QC events following each random interval
the Data Supplement that accompanies the online version increases. Table 1 gives values for E(TQ) for a number of
of this article at http://www.clinchem.org/content/ different QC schedules covering the 4 strategies. In all
vol53/issue4. (References 7–11 are cited in the online Data cases the average interval between QC events is 8 h.
Table 1. Expected time from the occurrence of an error condition to the next scheduled QC event for different QC
scheduling strategies.
QC schedule Expected time from error to next QC event (min) Simulation average (N ⴝ 100 000)
8-h fixed intervals 240.0 240.2
Random within 8 h fixed; Beta(1,1) 280.0 280.2
Random within 8 h fixed; Beta(3,3) 257.1 256.6
Random intervals 8 ⫾ 4 h; Beta(1,1) 260.0 260.1
Random 8 ⫾ 4 h, Beta(1,1) ⫹ two 8 h fixed 246.7 246.3
Assuming that a Beta(a,a) distribution is used to gen- signed so that E(TQ) was 1% greater than E(TQ) for a fixed
erate the random time intervals for strategy 4, then the interval strategy. In panel B, a ⫽ 2.625, w ⫽ 2 h, and N ⫽
expected time between the occurrence of an out-of-control 0. In panel C, a ⫽ 1.583, w ⫽ 4 h, and N ⫽ 5. Fig. 3 shows
error condition and the next scheduled QC event is given the expected length of time to the next scheduled QC
by the following equation: event as a function of time of day for the 3 different QC
冉冊
schedules. The expected wait time to the next scheduled
I w2 QC event is highly dependent on the time of day for the
E (TQ) ⫽ 1⫹ (5)
2 (N ⫹ 1) (2a ⫹ 1) fixed interval schedule, but the 2 random schedules have
expected wait times that are nearly independent of the
This strategy allows for 3 design parameters that can be
time of day.
chosen to balance the increase in E(TQ) and the degree of
variability in the time of day that QC events are sched-
Discussion
uled: the parameter for the Beta distribution (a), the half
The original goal of this study was to devise a QC testing
width of the random interval (w), and the number of fixed
strategy that scheduled QC events at different times
QC intervals following each random interval (N). In
throughout the day and at the same time minimized the
general, increasing the Beta parameter, a, decreasing the
average length of time between an out-of-control condi-
interval half width, w, or increasing the number of fixed
tion and the next scheduled QC event. QC samples are
intervals, N, will decrease E(TQ) and also decrease the rate
generally run at the beginning of each shift or at some
at which QC event times disperse throughout the day.
other time that is convenient to laboratory operation. A
Fig. 2 gives examples of a fixed interval schedule and 2
way to overcome these restrictions is to use sampling
different strategy 4 scheduling schemes that were de-
methods that ensure that the entire system is being tested
with QC. Patient samples may be run at any time of the
day, so it seems logical that control testing should also be
performed at any time that patients might be tested.
Ideally, the length of time from the testing of a patient
sample until QC testing is performed should not depend
on what time of day the patient sample was tested. This
idea appears to be surprisingly novel for QC practices in
a clinical laboratory.
Only a few papers have appeared in the laboratory
medicine literature that address the issue of when to
schedule QC events. Neubauer et al., using a combination
of expert opinions, computer simulations of customary
cost models, and review of literature, looked at the
optimal frequency of QC testing for automatic multichan-
nel analyzers (3 ). They concluded that the optimal num-
ber of samples between controls should be between 30
and 100. Steindel and Tetrault used a College of American
Pathologists Q-Probes study to assess QC practices in
hospital laboratories (4 ). Their study found wide variabil-
ity in QC frequency, QC rule use for data analysis in
determining out-of-control events, and response to out-
of-control events, highlighting the lack of standardization
in the laboratory community with regard to QC proce-
dures. Concluding that a need exists for new approaches
to QC on modern automated analyzers, Steindel and
Tetrault recommended the use of patient results to sup-
plement the infrequent analyses of control materials and
emphasized the need for simplification of QC systems (4 ).
Parvin and Gronowski discussed the arbitrary definition
Fig. 2. Examples of 3 different QC scheduling strategies. of an analytical run in the modern clinical laboratory (5 ).
(A), 8-h fixed intervals. (B), Intervals randomly selected using a Be- They defined performance measures that accounted for
ta(2.625,2.625) distribution scaled to range between 6 and 10 h. (C), An interval different definitions of an analytical run within the overall
randomly selected using a Beta(1.583,1.583) distribution scaled to range
between 4 and 12 h, followed by 5 intervals of 8 h. Examples (B) and (C) are 2 QC strategy and used statistical analysis to evaluate the
different QC scheduling strategies that provide QC testing at different times performance of different analytical run definitions for
throughout the day while maintaining an expected time between the occurrence
of an out-of-control error condition and the next scheduled QC event that is only
detecting when the process is out of control. Their analy-
1% greater than the fixed interval QC scheduling strategy. ses were based on the assumption that an out-of-control
Fig. 3. The average time to the next scheduled QC

event as a function of the time of day (24-h clock) for
the 3 QC scheduling strategies shown in Fig. 2.
Corresponding scheduling strategies in Figs. 2 and 3 share
the same plotting symbols.
condition can occur with equal probability anywhere strategies. We used both statistical modeling and com-
within a run and will remain present until detected. puter simulation to compare a variety of different QC
NCCLS guidelines discuss the concept of an “analytical scheduling strategies: with testing at fixed time intervals,
run” as being defined by the time or number of analyses random time intervals, and a combination of both. Our
for which the measurement process is stable (6 ). The findings demonstrate that fixed time intervals between
document does not define or endorse any specific QC QC events will detect persistent out-of-control situations
strategy for an individual device. It describes the concepts sooner on average than other QC scheduling strategies. If
of Manufacturer’s Recommended Run Length and User’s a laboratory is interested only in minimizing the expected
Defined Run Length, stating that there currently are no time to detect an error condition, then a fixed interval QC
well-accepted methods for establishing run lengths in a scheduling strategy should be used. However, when
more scientific manner. Furthermore, the document other issues are also of concern, such as assuring that the
makes no specific recommendations regarding the loca- expected length of time until QC testing is performed is
tion of control samples within a run, but it does discuss independent of the time of day that a sample is tested,
defining a User’s Defined Run Length as either a batch of then this work shows how a random QC scheduling
samples or a time interval. Several alternative strategies strategy can accomplish this. It should be noted that
for placing control samples within a run are discussed, employing a fixed time interval between QC events that is
but no specific strategy is endorsed. not evenly divisible into a 24-h day will also accomplish
This study was intended to address concerns regarding the goal of varying the time of day of QC events, but in a
limitations in traditional QC scheduling practices. Specif- nonrandom way.
ically, there is a tendency for most laboratories to run QC It might appear that a QC scheduling strategy that
at fixed points in time that do not vary from day to day. randomly places QC events within fixed time intervals
This type of QC testing strategy leaves large gaps of time (strategy 2) and a QC scheduling strategy that randomly
during which patient samples are run but QC testing is selects the length of time to the next QC event (strategy 3)
never performed. Traditional QC testing strategies there- are effectively equivalent. Although both of these strate-
fore are at least theoretically in violation of the spirit of the gies accomplish the goal of randomly distributing QC
CLIA regulations (1 ). CLIA regulations explicitly state events throughout the day, the performance of the 2
“Test control materials in the same manner as patient approaches is clearly different with respect to the ex-
specimens.” Furthermore, the regulation states that the pected length of time between the occurrence of an
control procedures must “Monitor over time the accuracy out-of-control error condition and the next scheduled QC
and precision of test performance that may be influenced event.
by changes in test system performance and environmental To be workable, any strategy that involves changing
conditions, and variance in operator performances.” the time of day when QC is scheduled will require some
Out-of control situations are relatively infrequent type of automated system to remind a technologist when
events. It is not possible to accumulate sufficient real-time to perform QC. Such a system should be easily program-
QC data for comparing the efficacy of different testing mable into existing laboratory instrumentation or labora-
tory information systems if the laboratory community 3. Neubauer A, Wolter C, Falkner C, Neumeier D. Optimizing fre-
adopts this strategy as a standard practice. quency and number of controls for automatic multichannel analyz-
Finally, this work demonstrates the valuable role that ers. Clin Chem 1998;44:1014 –23.
statistical thinking and theory can play in laboratory QC 4. Steindel SJ, Tetrault G. Quality control practices for calcium,
cholesterol, digoxin, and hemoglobin: a College of American
design and evaluation. The insight gained from the theo-
Pathologists Q-probes study in 505 hospital laboratories. Arch
retical derivations of the expected lengths of time from the Pathol Lab Med 1998;122:401– 8.
occurrence of an out-of-control error condition to the next 5. Parvin CA, Gronowski AM. Effect of analytical run length on
scheduled QC event for the various QC scheduling strat- quality-control (QC) performance and the QC planning process.
egies ultimately led to the discovery of a design strategy Clin Chem 1997;43:2149 –54.
that can balance the 2 goals of minimizing the expected 6. NCCLS. Statistical Quality Control for Quantitative Measurements:
time from the occurrence of an out-of-control error con- Principles and Definitions; Approved Guideline—Second Edition.
dition to the next scheduled QC event and enabling the Wayne, PA: NCCLS, 1999.
scheduling of QC events throughout the day. 7. Lorenzen TJ, Vance LC. The economic design of control charts: a
unified approach. Technometrics 1986;28:3–10.
8. Rahim MA. Determination of optimal design parameters of joint X៮
References
and R charts. J Qual Technol 1989;21:21–70.
1. Department of Health and Human Services. Centers for Medicare
and Medicaid Services Centers for Disease Control and Preven- 9. Yu FJ, Jiang LH. Optimization of design parameters for x៮ control
tion. 42 CFR Part 493 Medicare, Medicaid, and CLIA Programs; charts with multiple assignable causes. J Appl Stat 2006;33:
Laboratory Requirements Relating to Quality Systems and Certain 279 –90.
Personnel Qualifications; Final Rule. Fed Regist 2003;68:3653– 10. Evans M, Hastings N, Peacock B. Statistical Distributions, 3rd ed.
3657. New York: John Wiley, 2000:77– 81.
2. Reynolds MR, Amin RW, Arnold JC, Nachlas JA. X៮ charts with 11. Costa AFB. X៮ charts with variable parameters. J Qual Technol
variable sampling intervals. Technometrics 1988;30:181–92. 1999;31:408 –16.
581–586 (2007) Molecular Diagnostics
and Genetics
Real-time PCR Assay for Ultrasensitive

Quantification of DNA-Binding Proteins
Peng Hou,1,2*† Zaozao Chen,1† Meiju Ji,1 Nongyue He,1 and Zuhong Lu1*
Background: The specific binding of proteins to DNA is Conclusions: This technique is customizable and easy
a key step for many cellular activities, such as transcrip- to establish. It has potential applications in research,
tion regulation, DNA replication, recombination, repair, medical diagnosis, and drug discovery.
and restriction. The detection of DNA-binding proteins, © 2007 American Association for Clinical Chemistry
as well as the identification of specific binding sites, is
therefore important to understand gene expression Many proteins with natural, sequence-specific DNA-bind-
mechanisms and cellular function. We describe an uling activity are involved in regulating important cellular
trasensitive method for quantification of DNA-binding processes such as cell growth, differentiation, apoptosis,
proteins. and transformation to cancer (1 ). Transcription factors,
Methods: We combined the common exonuclease III for example, bind to specific promoter sequences and re-
(ExoIII) footprinting assay and real-time PCR for quan- cruit chromatin-modifying complexes and the transcrip-
tification of DNA-binding proteins, for an assay that tion apparatus to initiate RNA synthesis (2– 4 ). The repro-
does not require antibodies against the target proteins. gramming of gene expression that occurs as cells move
Double-strand DNA probes were designed to monitor through the cell cycle, or when cells sense changes in their
the activities of DNA-binding protein. The protein- environment, is effected in part by changes in the DNA-
binding site is at the 5ⴕ end of the forward primer. When binding status of transcription factors. Distinct DNA-bind-
a target protein is present, it will specifically bind to the ing proteins are also associated with origins of DNA repli-
protein-binding site and produce a physical hindrance cation, centromeres, telomeres, and other aspects of genome
to ExoIII, which protects the reverse DNA strand from
maintenance (5, 6 ). Therefore, the quantitative detection of
digestion by ExoIII. The remaining single-strand DNA
DNA-binding proteins, as well as the identification of spe-
template can be quantitatively detected by real-time
cific binding sites, is important to understand gene expres-
PCR. Conversely, in the absence of the target protein,
sion mechanisms and cellular function.
the naked primer regions will be degraded by ExoIII,
Conventional methods for detecting DNA-binding
which then cannot be amplified by real-time PCR.
proteins are electrophoretic mobility shift assay (EMSA)3
Results: We detected the binding of 10 different tran-
(7 ), DNA footprinting (8 ), ELISA (9 ), and Southwestern
scription factors in crude cell extracts. The assay quan-
titatively detected binding at femtomolar concentrations blotting (10 ). They are laborious, time-consuming proce-
of protein. dures that typically involve the use of radioisotopes or
antibodies against the target proteins, which are not
adaptable to high-throughput formats, thus making the
procedures unsuitable for more extensive application. In
1
Chien-Shiung Wu Laboratory, Department of Biological Science and recent years, several techniques have been developed to
Medical Engineering, Southeast University, Nanjing, China.
2
detect DNA-binding proteins, including DNA microarray
Division of Endocrinology and Metabolism, The Johns Hopkins Univer-
sity School of Medicine, Baltimore, MD.
(11–15 ), fluorescence resonance energy transfer (16, 17 ),
†These authors contributed equally to this work. and Exo-dye– based assay (18 ). They are useful tools in
*Address correspondence to this author at: Division of Endocrinology and studying protein-DNA interaction; however, their lack of
Metabolism, The Johns Hopkins University School of Medicine, 1830 East
Monument St., Suite 333, Baltimore, MD 21287. E-mail: phou2@jhmi.edu.
Correspondence may also be addressed to Prof. Zuhong Lu, Chien-Shiung Wu
Laboratory, Department of Biological Science and Medical Engineering,
3
Southeast University, Nanjing 210096, China. Fax 86-25-83619983; e-mail Nonstandard abbreviations: EMSA, electrophoretic mobility shift assay;
zhlu@seu.edu.cn. IPCR, immuno-PCR; ExoIII, exonuclease III; ssDNA, single-strand DNA;
Received August 1, 2006; accepted January 16, 2007. dsDNA, double-strand DNA; TNF-␣, tumor necrosis factor ␣; Ct, threshold
Previously published online at DOI: 10.1373/clinchem.2006.077503 cycle.
581
582 Hou et al.: Ultrasensitive Quantification of DNA-Binding Proteins
sensitivity makes them difficult to apply in detecting trace GGG GAC TTT CCC AGG CTT TTT-3⬘ and 5⬘-GCC TGG
amounts of DNA-binding proteins. GAA AGT CCC CTC AAC TTT TTT-3⬘; the underlined
PCR and other forms of target amplification have enabled sequence represents the NF-␬B– binding sites. The oligo-
rapid advances in the development of powerful tools for nucleotides used as a cold NF-␬B nonspecific competition
detecting and quantifying DNA targets of interest for re- probe were 5⬘-AGT TGA GAT TAC TTT CAC AGG CTT
search, forensic, and clinical applications (19 –21 ). To extend TTT-3⬘ and 5⬘-GCC TGT GAA AGT AAT CTC AAC TTT
the scope of PCR to high-sensitivity detection of proteins, the TTT-3⬘; the underlined sequence represents the nonspe-
immuno-PCR (IPCR) technique was developed in 1992. cific NF-␬B– binding sites. We made competitor probes
IPCR takes advantage of specific antibody–DNA conjugates with 5 protruding bases (T) at each 3⬘ end for protection
(22 ) and has become a well-respected research methodol- from ExoIII digestion. To obtain duplex competitors, we
ogy, having established its general applicability for the mixed 2 oligonucleotides in the same molar ratios at a
sensitive detection of numerous protein antigens, usually final concentration of 50 ␮mol/L in 100 ␮L of 10 mmol/L
providing a 100- to 10 000-fold improvement of the detection Tris-HCl (pH 8.0), 100 mmol/L NaCl, and 1 mmol/L
limit of conventional ELISA (23 ). Nam et al. (24 ) originally EDTA. The mixture was heated for 5 min at 95 °C and
developed a high-sensitivity nanoparticle-based assay for cooled slowly to 25 °C.
protein detection, termed bio-bar-code assay, in which an
amplifier nanoparticle is coloaded to the captured target. quantitative detection of dna-binding
Because only one of the strands is attached covalently to the proteins
nanoparticle, the complementary strand can be released and This assay involves 3 steps: protein–DNA interaction
serves as the surrogate target, which is then detected by (protein binding), ExoIII digestion, and real-time PCR
hybridization to an array (24 ). IPCR and nanoparticle-based amplification and detection. For protein binding, the
bio-bar-code approaches have high sensitivity and could be experiment was performed in binding buffer containing
used for the ultrasensitive detection of target proteins. Both, 10 mmol/L Tris-HCl (pH 7.5), 50 mmol/L NaCl,
however, involve the use of different antibodies against the 0.5 mmol/L EDTA, 0.5 mmol/L dithiothreitol, 3 mmol/L
target proteins. MgCl2, 0.05 g/L poly(dI-dC) (Roche), 10% vol/vol glyc-
Herein, we report a universal real-time PCR assay for erol, 0.5 g/L bovine serum albumin, and 0.05% Nonidet
the ultrasensitive quantification of DNA-binding proteins P40. NF-␬B (rhNF-␬B p50) was purchased from Promega
that does not require antibodies against the target pro- and preincubated with DNA-binding buffer at 37 °C for
teins. The common exonuclease III (ExoIII) footprinting 10 min. HeLa cell nuclear extract stimulated with and
assay is coupled with real-time PCR to monitor the without tumor necrosis factor ␣ (TNF-␣) was purchased
activities of DNA-binding proteins. We used this tech- from Takara Active Motif and preincubated in binding
nique to successfully detect 10 different transcription buffer containing 10 mmol/L Tris-HCl (pH 7.5),
factors in crude cell extracts. This assay allows ultrasen- 50 mmol/L NaCl, 0.5 mmol/L EDTA, 0.5 mmol/L dithio-
sitive and high-throughput quantification of DNA-bind- threitol, 3 mmol/L MgCl2, 0.05 g/L poly(dI-dC) (Roche),
ing proteins and can be used extensively in biomedical 10% vol/vol glycerol, 2 mmol/L sodium phosphate (pH
research. 7.0), 20 ng/␮L HaeIII-cut Escherichia coli DNA (Roche),
and 25 fg/L yeast tRNA (Roche) at 37 °C for 10 min. For
Materials and Methods nuclear extracts, interference may arise in this assay due
preparation of double-strand dna target for to the endogenous nuclease activity upon ExoIII exposure
binding assay (25 ). To avoid this problem, we included sodium phos-
The primers and the single-strand DNA (ssDNA) tem- phate and carrier nucleic acids including poly(dI-dC),
plate used in this study, synthesized using accepted HaeIII-cut E. coli DNA, and yeast tRNA to suppress
phosphoramidate chemistry and purified by HPLC (In- endogenous nuclease activities in the crude extracts
vitrogen), are listed in Table 1 in the Data Supplement (25, 26 ).
that accompanies the online version of this article at After incubation, dsDNA probes with different pro-
http://www.clinchem.org/content/vol53/issue4. Briefly, tein-binding sites were added and incubated at the same
we subjected ssDNA to PCR in the presence of the reverse temperature for 20 min more. At the end of the DNA–
primer and the forward-tailed primer, which contains dif- protein interaction, a sufficient amount of ExoIII was
ferent protein-binding sites at the 5⬘ end (see Table 1 in the added and incubated to digest the dsDNA probes at 37 °C
online Data Supplement). After PCR amplification and pu- for 5 min. The reaction was terminated at 70 °C for 20 min.
rification, we obtained double-strand DNA (dsDNA) probes Real-time PCR was run for quantitative detection of
with different protein-binding sites for binding reactions. DNA-binding proteins by use of a 96-well PCR plate with
a final reaction mixture of 25 ␮L consisting of 600 nmol/L
preparation of duplex competitors for of each primer, 200 nmol/L TaqMan probe (Takara),
competition assay 200 ␮mol/L of each deoxyribonucleoside triphosphate,
We synthesized the following oligonucleotides to be used 3.5 mmol/L MgCl2, and 1⫻ TaqMan Buffer (Takara)
as a cold NF-␬B–specific competition probe: 5⬘-AGT TGA under the following conditions: 95 °C for 10 min, 40 cycles
at 95 °C for 15 s, and 60 °C for 1 min. The primers and Alternatively, the ssDNA template could be produced
TaqMan probe used in this study are listed in Table 1 in when the dsDNA probe was prepared by regular PCR.
the online Data Supplement. ssDNA could not be degraded by ExoIII, which results in
a background signal by real-time PCR.
data analysis
The instrument monitors the increase of the normalized specificity and general applicability
fluorescence reporter signal for each cycle. We applied a To explore more fully the specificity of detection, we
baseline correction, typically using the first 10 to 15 cycles performed competition assays: we made 2 duplex DNA
as background signal. The software calculates the thresh- competitors, one containing a consensus NF-␬B– binding
old cycle (Ct), which represents the first PCR cycle at site (GGGACTTTCC) as a specific competitor and the
which the reporter signal exceeds the signal of the base- other not containing an NF-␬B– binding site as the non-
line determined by the user. The threshold signal was specific competitor. We added the specific and nonspe-
determined to be above the background signals for empty cific competitors to the reaction buffer containing NF-␬B
wells and was set in the linear signal increment phase for a binding reaction and performed ExoIII digestion. The
according to the data obtained for each experiment. The results of competition assays are shown in Fig. 2. The
normalization of the Ct values was achieved as follows. Ct specific competitor decreased the amount of dsDNA
values are inversely proportional to the amount of DNA template (or increased the Ct value) in the presence of
template and, thus, DNA-binding protein concentration, NF-␬B (Fig. 2A; Fig. 1 in the online Data Supplement),
so we subtracted test sample Ct values from negative whereas the nonspecific competitor did not affect the
control Ct values to generate ⌬Ct values. The ⌬Ct values amount of dsDNA template or the Ct value (Fig. 2B; Fig.
increase with increasing amounts of DNA template and 1 in the online Data Supplement). These results provide
proteins. We analyzed each test in triplicate and calcu- strong evidence for the specificity of this assay and also
lated mean values of ⌬Ct to make a calibration curve for indicate that such competition assays could be used to
quantification analysis of target proteins. We obtained rapidly assess the relative binding affinity of the protein
linear regressions of the signals from the enriched sam- to different variants of DNA.
ples against the logarithmic concentrations of proteins. To investigate quantitative detection of DNA-binding
We performed statistical analysis and plotting of the data proteins, we added a series of different concentrations of
with SigmaPlot 8.0 software. NF-␬B to the same molarity dsDNA probe with an
NF-␬B– binding site and performed simultaneous ExoIII
Results and Discussion digestion and real-time PCR. As shown in Fig. 2 in the
mechanism online Data Supplement, the Ct values decreased when
Fig. 1 illustrates this assay. Briefly, we prepared dsDNA the amount of NF-␬B increased from 10 fmol/L to 2
probes by regular PCR, containing the different protein- nmol/L (see Fig. 2, inset, in the online Data Supplement),
binding sites, universal primer regions, and TaqMan and we observed a detection limit of 10 fmol/L NF-␬B
probe region. The protein-binding site is at the 5⬘ end of (see Fig. 2 in the online Data Supplement). Moreover,
the forward primer (Fig. 1A). All sequences used in this linear regression indicated broad dynamic range and
study are summarized in Table 1 in the online Data high linearity (R2 ⫽ 0.999; see Fig. 2 in the online Data
Supplement. We incubated DNA-binding proteins with Supplement), suggesting that, under optimal experimen-
dsDNA probe for a binding reaction. ExoIII was then tal conditions, this assay can be applied for ultrasensitive
added to degrade dsDNA. ExoIII is a modifying enzyme, quantification of DNA-binding proteins.
which exhibits the 3⬘35⬘ exodeoxyribonuclease activity
specific for dsDNA. It can degrade dsDNA from blunt use with nuclear extracts
ends, 5-overhangs, or nicks. In this assay, if a target Nuclear extracts contain many DNA-binding proteins
protein is present, it will specifically bind to the protein- that may interfere with the assay. To illustrate applicabil-
binding site and produce a physical hindrance to ExoIII, ity to nuclear extracts, we performed the assay shown in
protecting the reverse DNA strand from digestion by Fig. 3 in the online Data Supplement in the presence of the
ExoIII (Fig. 1A). The remaining ssDNA template can be HeLa cell nuclear extract. The specific competitor de-
quantitatively detected by use of real-time PCR (Fig. 1B, creased the amount of dsDNA template (or increased the
curve 2). Conversely, in the absence of the target protein Ct value) in the presence of NF-␬B from nuclear extracts
(negative control), the naked primer regions will be (see Fig. 3, A and C, in the online Data Supplement). In
degraded by ExoIII, which cannot be amplified by real- contrast, addition of nonspecific competitor to the same
time PCR (Fig. 1B, curve 3). The positive control is the reaction system did not affect the amount of dsDNA
dsDNA probe without digestion of ExoIII, which shows template or the Ct value (see Fig. 3, B and C, in the online
the original amount of the dsDNA probe (Fig. 1B, curve Data Supplement). Thus nuclear extract– dependent varia-
1). As clearly shown in Fig. 1B, there is a background tion of the Ct value was NF-␬B specific. Although we added
signal in the negative control, which probably results a 100-fold excess of competitor over the amount of DNA
from 2 factors. One is incomplete digestion of ExoIII. probes, a high Ct value was observed (see Fig. 3A in the
Fig. 1. Real-time PCR for quantification of DNA-binding proteins.

(A), schematic representation of the assay. (B), amplification plots. Curves 1–3 indicate the 2 nmol/L double-strand NF-␬B probe without digestion of ExoIII as positive
control, the 2 nmol/L double-strand NF-␬B probe bound by 10 pmol/L NF-␬B and digested by ExoIII, and the 2 nmol/L NF-␬B probe with digestion of ExoIII as negative
control, respectively.
online Data Supplement). For further confirmation, we protected by this specific binding from digestion of ExoIII
added an even higher amount of competitor (800 nmol/L) and be used as the template for further amplification,
to the reaction. The Ct value obtained indicated that a suggesting that the method has high sensitivity for detecting
high degree of amplification was achieved even under these trace amounts of DNA-binding proteins.
conditions (see Fig. 4 in the online Data Supplement). It The high amplification might also be caused by non-
appears that although most of the protein binds to the specific binding. Therefore, we performed additional ex-
competitor, a substantial fraction is still able to bind to the periments to address nonspecific binding of nuclear ex-
specific DNA probe. Therefore, the DNA probe could be tracts to the DNA probe lacking the binding site (see Fig. 5
Fig. 2. Specificity of NF-␬B detection.

(A), amplification plots of real-time PCR for the 2 nmol/L NF-␬B probes in the
presence of 2 nmol/L NF-␬B and increasing amounts of cold specific probe.
(B), amplification plots of real-time PCR for the 2 nmol/L NF-␬B probe in the
presence of 2 nmol/L NF-␬B and increasing amounts of cold nonspecific probe.
The final concentrations of the competitor probe were 0 nmol/L (curve 1), 2
nmol/L (curve 2), 10 nmol/L (curve 3), 50 nmol/L (curve 4), 100 nmol/L (curve
5), and 200 nmol/L (curve 6). ⫹ and ⫺ indicate positive and negative controls,
respectively.
in the online Data Supplement). The results indicated that

the nonspecific binding is very weak, similar to the
negative control, thus providing strong evidence that this
system is stable and works well to quantitatively evaluate
DNA-binding proteins.
To explore quantitative detection of multiple DNA- Fig. 3. Comparison of NF-␬B concentrations between HeLa cell extracts
binding proteins in the nuclear extracts, we selected 10 unstimulated and stimulated with TNF-␣.
important transcription factors as target proteins, in- (A), amplification plots of real-time PCR for the 2 nmol/L NF-␬B probes in the
cluding SP1, AP2, NFAT, SRE, EGR, YY1, AP1, CREB, presence of increasing amounts of nuclear extracts unstimulated with TNF-␣.
(B), amplification plots of real-time PCR for the 2 nmol/L NF-␬B probes in the
OCT1, and NF-␬B; detailed information is summarized in presence of increasing amounts of nuclear extracts stimulated with TNF-␣.
Table 1 in the online Data Supplement. The different (C), Nuclear extract concentration-dependent relative Ct values (⌬Ct). Linear
regression fits to the ⌬Ct values from nuclear extracts unstimulated and
target proteins in HeLa cell nuclear extract were detected stimulated with TNF-␣ were 0.999 and 0.995, respectively. Combining these 2
by adding increasing amounts of nuclear extracts to a linear regression formats, we can observe that the concentration of NF-␬B can be
increased by ⬃8 times when HeLa cells are stimulated by TNF-␣.
series of the same molarity dsDNA probes with different
protein-binding sites in one 96-well PCR plate. The Ct
values decreased with increasing amounts of nuclear vides useful complementary data, detection is still limited
extracts (see Fig. 6 in the online Data Supplement). As to a fraction of the proteins present in crude cell extracts.
clearly shown in Fig. 7 in the online Data Supplement, we Therefore, the development of convenient, specific, sensi-
observed linear regression of signals for the enriched tive assays for detecting DNA-binding proteins remains
samples against the logarithmic concentration of target an extremely important goal. In EMSA, one of the general
proteins, strongly suggesting that this assay could be used DNA-based methods (7 ), the activity of DNA-binding
for high-throughput quantification of DNA-binding pro- proteins is evaluated by measuring the intensity of a
teins in crude cell extracts. retarded band produced by DNA-protein complexes in
Gene expression is only one of the relevant aspects of native polyacrylamide gel electrophoresis. EMSA can
cell physiology, another being the misregulation of pro- offer a sensitive detection for DNA-binding proteins;
tein-mediated cellular signaling pathways. The localiza- however, it is time-consuming and laborious. Moreover,
tion and activity of low-abundance oncogenes such as the procedure requires the use of radioactivity and is not
DNA-binding proteins remain difficult to assess with adaptable to automatic and high-throughput analysis. In
DNA-only approaches. Although proteomic analysis pro- ELISA, another conventional DNA-based method, the
proteins are measured by capturing DNA-binding pro- 5. Kelly TJ, Brown GW. Regulation of chromosome replication. Annu
teins with a dsDNA-coupled plate and then detected Rev Biochem 2000;69:829 – 80.
using ELISA (9 ). Its sensitivity is limited, however, pre- 6. Dutta A, Bell SP. Initiation of DNA replication in eukaryotic cells.
Annu Rev Cell Dev Biol 1997;13:293–332.
venting its utility in detecting trace amounts of DNA-
7. Garner MM, Revzin A. A gel electrophoresis method for quantifying
binding proteins. the binding of proteins to specific DNA regions: application to
In this report, we describe a universal real-time PCR components of the Escherichia coli lactose operon regulatory
coupled with ExoIII protection assay that has broad dy- system. Nucleic Acids Res 1981;9:3047– 60.
namic range and high linearity for the ultrasensitive 8. Galas DJ, Schmitz A. DNAse footprinting: a simple method for the
quantification of sequence-specific DNA-binding pro- detection of protein-DNA binding specificity. Nucleic Acids Res
teins. It does not require antibodies against the target 1978;5:3157–70.
proteins; however, the incomplete digestion of ExoIII has 9. Renard P, Ernest I, Houbion A, Art M, Le Calvez H, Raes M, et al.
Development of a sensitive multi-well colorimetric assay for active
an effect on the detection sensitivity or limit. NFkappaB. Nucleic Acids Res 2001;29:e21.
We used this assay to successfully detect 10 different 10. Bowen B, Steinberg J, Laemmli UK, Weintraub H. The detection of
transcription factors in the crude cell extracts. Addition- DNA-binding proteins by protein blotting. Nucleic Acids Res 1980;
ally, it could be used to quantitatively evaluate transcrip- 8:1–20.
tion factor NF-␬B from the HeLa cell nuclear extract 11. Bulyk ML, Gentalen E, Lockhart DJ, Church GM. Quantifying
unstimulated and stimulated with TNF-␣ (Fig. 3), with a DNA-protein interactions by double-stranded DNA arrays. Nat
⌬Ct of 0.87. According to the calibration curve as de- Biotechnol 1999;17:573–7.
12. Bulyk ML, Huang X, Choo Y, Church GM. Exploring the DNA-binding
scribed in Fig. 2 in the online Data Supplement, we
specificities of zinc fingers with DNA microarrays. Proc Natl Acad
calculated that the concentration of NF-␬B increased 7.41 Sci U S A 2001;98:7158 – 63.
times after the HeLa cell nuclear extracts were stimulated 13. Linnell J, Mott R, Field S, Kwiatkowski DP, Ragoussis J, Udalova
with TNF-␣. IA. Quantitative high-throughput analysis of transcription factor
binding specificities. Nucleic Acids Res 2004;32:e44.
In conclusion, the results presented here provide the 14. Mukherjee S, Berger MF, Jona G, Wang XS, Muzzey D, Snyder M,
results provided strong evidence that this assay has wide et al. Rapid analysis of the DNA-binding specificities of transcrip-
tion factors with DNA microarrays. Nat Genet 2004;36:1331–9.
potential application in research, medical diagnosis, and
15. Egener T, Roulet E, Zehnder M, Bucher P, Mermod N. Proof of
drug discovery. Moreover, the assay should be useful for
concept for microarray-based detection of DNA-binding oncogenes
high-throughput measurement of the DNA-binding activ- in cell extracts. Nucleic Acids Res 2005;33:e79.
ity or affinity of proteins. One application could be in 16. Heyduk T, Heyduk E. Molecular beacons for detecting DNA binding
screening for drugs targeted to a DNA-binding protein. proteins. Nat Biotechnol 2002;20:171– 6.
The technique should also have applications in medical 17. Knoll E, Heyduk T. Unimolecular beacons for the detection of
diagnosis, as a means to determine the presence or the DNA-binding proteins. Anal Chem 2004;76:1156 – 64.
activity of a DNA-binding protein linked to a particular 18. Chen Z, Ji M, Hou P, Lu Z. Exo-Dye-based assay for rapid,
inexpensive, and sensitive detection of DNA-binding proteins.
disease. Our report is proof-of-concept for the assay’s
Biochem Biophys Res Commun 2006;345:1254 – 63.
rapid, ultrasensitive, and high-throughput ability to study 19. Saiki RK, Scharf S, Faloona F, Mullis KB, Horn GT, Erlich HA, et al.
the DNA-binding activities of cellular extracts and puri- Enzymatic amplification of beta-globin genomic sequences and
fied and recombinant proteins. restriction site analysis for diagnosis of sickle cell anemia.
Science 1985;230:1350 – 4.
20. Gibbs RA. Polymerase chain reaction techniques. Curr Opin Bio-
This work was supported by the National Natural Science technol 1991;2:69 –75.
Foundations of China (30600736 and 60121101), the Post- 21. Bustin SA. Quantification of mRNA using real-time reverse tran-
scription PCR (RT-PCR): trends and problems. J Mol Endocrinol
doctoral Foundation of Jiangsu Province, and the Post-
2002;29:23–39.
doctoral Foundation of China (20060390954). We thank 22. Sano T, Smith CL, Cantor CR. Immuno-PCR: very sensitive antigen
Bin Yang for help with supplemental experiments. detection by means of specific antibody-DNA conjugates. Science
1992;258:120 –2.
References 23. Adler M, Wacker R, Niemeyer CM. A real-time immuno-PCR assay
1. Pabo CO, Sauer RT. Transcription factors: structural families and for routine ultrasensitive quantification of proteins. Biochem
principles of DNA recognition. Annu Rev Biochem 1992;61:1053– Biophys Res Commun 2003;308:240 –50.
95. 24. Nam JM, Thaxton CS, Mirkin CA. Nanoparticle-based bio-bar codes
2. Ptashne M, Gann A. Transcriptional activation by recruitment. for the ultrasensitive detection of proteins. Science 2003;301:
Nature 1997;386:569 –77. 1884 – 6.
3. Lee TI, Young RA. Transcription of eukaryotic protein-coding 25. Wu C. Two protein-binding sites in chromatin implicated in the
genes. Annu Rev Genet 2000;34:77–137. activation of heat-shock genes. Nature 1984;309:229 –34.
4. Malik S, Roeder RG. Transcriptional regulation through Mediator- 26. Ng R, Carbon J. Mutational and in vitro protein-binding studies on
like coactivators in yeast and metazoan cells. Trends Biochem Sci centromere DNA from Saccharomyces cerevisiae. Mol Cell Biol
2000;25:277– 83. 1987;7:4522–34.
587–593 (2007) Molecular Diagnostics
and Genetics
Preanalytical mRNA Stabilization of Whole Bone

Marrow Samples
Claudia Langebrake,1* Kalle Günther,2 Jürgen Lauber,2 and Dirk Reinhardt1
Background: Gene expression profiling is a useful tool Gene expression analysis by quantitative reverse-tran-
for cancer diagnosis and basic research. A major limita- scription PCR (Q-RT-PCR)3 is widely used in basic and
tion is that, even during short-term storage of native clinical research of cancer. In patients with leukemia, bone
specimens of peripheral blood or bone marrow (BM) marrow (BM) specimens—rather than peripheral blood
and/or RNA isolation, significant changes of gene ex- (PB) specimens—are the source of material for initial
pression pattern can occur because of gene induction, diagnosis and follow-up evaluation of treatment re-
repression, and RNA degradation. sponse. Nearly all children with leukemia in Germany
Methods: We investigated the effectiveness of a newly (96.9%; http://www.kinderkrebsregister.de) are treated
developed RNA stabilization and preparation system in nationwide multicenter clinical trials with centralized
for BM specimens (PAXgene™ Bone Marrow RNA laboratories for diagnosis and research questions. Antico-
agulated BM specimens are commonly shipped to these
System) over time. We analyzed 256 RNA samples,
laboratories by overnight mail at room temperature. Ow-
processed from 64 BM specimens.
ing to gene induction, repression, and RNA degradation,
Results: Although the overall RNA yield (normalized to
storage of native specimens of PB or BM and method of
1 ⴛ 107 leukocytes) was not different, the RNA prepa-
RNA isolation have significant influence on gene expres-
ration using unstabilized reference samples had an ⬃3 sion (1–3 ). However, the evaluation of gene expression
times higher failure rate. With the PAXgene system, we levels using Q-RT-PCR or microarray technology is essen-
observed significantly higher RNA integrity compared tial for the investigation of the molecular origin of leuke-
with the reference RNA preparation system (P <0.01). In mia (4 – 8 ) as well as for monitoring minimal residual
the stabilized samples, we found very high pairwise disease (9 ).
correlation in gene expression (⌬⌬CT 0.16 – 0.53) for the Acute leukemias are clonal disorders of the hematopoi-
analyzed genes (GATA1, RUNX1, NCAM1, and SPI1) etic system that arise in the BM. The current hypothesis
after 48-h storage compared with immediate preparation for leukemogenesis published by Bonnet and Dick (10 )
of RNA (2 h after BM collection). However, we found assumes a multistep process starting from a leukemic
major differences in half of the analyzed genes using stem cell that undergoes hierarchical differentiation into
the reference RNA isolation procedure (⌬⌬CT 1.07 and leukemic blasts. The aberrant expression of different
1.32). transcription factors that are crucial at specific stages of
Conclusions: The PAXgene system is able to stabilize hematopoiesis is believed to have a dominant role in
RNA from clinical BM samples and is suitable to isolate either generation or maintenance of the malignant clone.
high-quality and -quantity RNA. The RUNX1 gene (runt-related transcription factor 1)4 is
© 2007 American Association for Clinical Chemistry the most frequent target gene of aberrant activity in
human leukemias. It is involved in several chromosomal
translocations, e.g., t(8;21) in childhood acute myeloid
leukemia, t(12;21) in childhood acute lymphoblastic leu-
kemia, and t(3;21) in chronic myeloid leukemia or myelo-
1
Department of Pediatric Hematology and Oncology, Hannover Medical
School, Hannover, Germany.
2
Qiagen GmbH, Hilden, Germany.
3
*Address correspondence to this author at: Medizinische Hochschule Nonstandard abbreviations: Q-RT, quantitative reverse-transcription; RT,
Hannover, Department of Pediatric Hematology and Oncology, AML-BFM reverse-transcription; BM, bone marrow; PB, peripheral blood; RIN, RNA
Study, Carl-Neuberg-Strasse 1, 30625 Hannover, Germany. Fax 49-511-532- integrity number.
4
9029; e-mail: claudia@langebrake.de. Human genes: RUNX1, runt-related transcription factor 1; GATA1,
Received August 16, 2006; accepted January 16, 2007. GATA binding protein 1; SPI1, spleen focus forming virus proviral integration
Previously published online at DOI: 10.1373/clinchem.2006.078592 oncogene 1; NCAM1, neural cell adhesion molecule 1.
587
588 Langebrake et al.: PAXgene Bone Marrow RNA
dysplastic syndrome. Furthermore, a high incidence of after informed consent from each patient or patient’s
RUNX1 point mutations has been revealed. Recently, a guardian. All investigations were approved by the local
3rd mode of RUNX1 involvement in leukemogenesis—an ethics committee. Immediately after collection, we trans-
increased dosage of RUNX1—has been reported (11 ). The ferred 2.5 mL BM to 2 PAXgene Bone Marrow RNA tubes
adhesion molecule NCAM1 (neural cell adhesion mole- containing optimized RNA stabilization solution. After
cule 1) is commonly expressed at the surface of subsets of 2 h (P2) or 48 h (P48) of storage at room temperature after
acute leukemias and furthermore contributes to hemato- collection, the tubes were frozen (⫺20 °C) for a median of
poiesis-supporting capacity of stromal cell lines (12 ). 5 weeks (range, 1–15 weeks). After thawing, we prepared
GATA1 (GATA binding protein 1) is essential for normal RNA by use of the PAXgene Bone Marrow RNA Kit. Two
erythropoiesis, with a crucial role in cell survival and specimens served as reference samples: 0.5 mL BM was
maturation (13 ). Mutations in exon 2 of GATA1, which are transferred into sterile plastic tubes and processed using
translated into a shorter GATA1s protein, are detectable
the QIAamp威 RNA Blood Mini Kit (Qiagen) at the same
in virtually all cases of myeloid leukemia of Down syn-
time points [2 h (R2) and 48 h (R48); Fig. 1]. The time
drome and transient myeloproliferative disease of new-
points were chosen to represent the operational proce-
borns with Down syndrome (14 ). PU.1, which is encoded
dures in many hospitals and laboratories. For the refer-
by the gene SPI1 (spleen focus forming virus proviral
ence protocol, we took into account a lag of 2 h (for
integration oncogene 1), serves as a suppressor of acute
myeloid leukemia, and its down-regulation results in an in-house transportation and sample receipt in the labora-
aggressive form of acute myeloid leukemia (15–17 ). tory) between collection of BM and preparation of RNA in
Recently, a commercially available in vitro diagnostic the laboratory. For the PAXgene system, a minimum
system for the stabilization of RNA in PB specimens precipitation time of 2 h is required before the RNA
(PAXgene™ Blood RNA System, PreAnalytiX) arrived on precipitate can be centrifuged. The 48-h storage at room
the market. This system inhibits RNA degradation and temperature of either the plastic tube with anticoagulated
prevents samples from ex vivo changes in gene expres- BM (reference protocol) or the PAXgene Bone Marrow
sion starting immediately at the time of sample collection RNA tube represents the time interval between collection
(18 ). RNA stabilization is a major prerequisite for reliable of BM and shipping to an external laboratory.
transcript analysis in whole blood samples. An important BM specimens were eligible for evaluation if all 4 RNA
difference between PB and BM that is relevant for RNA preparation procedures (2 PAXgene samples and 2 refer-
stabilization is BM’s significantly higher range in cellular- ence samples) fulfilled the following selection criteria for
ity, especially in diagnostic samples of acute leukemias downstream analysis: (a) RNA concentration ⬎5 mg/L
(⬎500 000 leukocytes/␮L), thus defining a strong need to and (b) RNA integrity number (RIN) ⬎7 or A260/A280
adapt the existing PAXgene Blood RNA system to BM. between 1.9 and 2.1. We measured leukocytes in BM
For research use only, PreAnalytiX has optimized the specimens by use of FlowCount™ Fluorospheres (Beck-
system for BM. The system enables the collection, stabili- man Coulter) and a Cytomics FC 500 flow cytometer
zation, storage, and transportation of human BM speci- (Beckman Coulter).
mens, together with a rapid and efficient protocol for
isolation and purification of intracellular RNA. rna preparation procedure
Our aim was to investigate the influence of preanalyti- We isolated and purified RNA from PAXgene Bone
cal steps of sample handling after aspiration and before Marrow RNA tubes using the PAXgene Bone Marrow
RNA analysis, focusing on RNA yield and integrity and RNA Kit according to manufacturer’s instructions. In
stability of gene expression profiles. We compared the
brief, we harvested nucleic acids by centrifugation of
PAXgene Bone Marrow RNA System with a method
collection tubes. We washed, resuspended, and incubated
without RNA stabilization using standard tubes and
pellets in optimized buffers together with proteinase K.
considered a 2-day storage of BM, reflecting the average
Cell lysates were homogenized and cell debris removed
shipping time in common clinical settings. We used
transcript concentrations of marker genes GATA1, by use of PAXgene Shredder columns. RNA binding
RUNX1, SPI1, and NCAM1, commonly involved in leu- conditions were adjusted with ethanol, and lysates were
kemogenesis, to study the need for stabilization of BM applied to RNA spin columns. After centrifugation
samples in a clinical setting of disease monitoring. through a silica-gel membrane, we removed contaminants
with efficient wash steps, followed by DNase I treatment
Patients, Materials, and Methods to remove traces of bound DNA. After additional wash
patients and samples steps, RNA was released in 80 ␮L elution buffer and
We collected BM specimens (n ⫽ 64), using 5000 IU heat-denatured. We prepared reference RNA samples
heparin per 5 mL BM as anticoagulant, from randomly from unstabilized BM aliquots from sterile standard plas-
selected children with acute leukemia at initial diagnosis tic tubes using the QIAamp RNA Blood Mini Kit as
or during treatment or from children without pathologic described in the handbook, with an elution volume of
hematopoietic findings. BM specimens were obtained 50 ␮L.
Fig. 1. Flowchart of the study design.

Immediately after collection, 2.5 mL anticoagulated
BM was transferred to 2 PAXgene Bone Marrow RNA
tubes containing optimized RNA stabilization solu-
tion. After 2 (P2) and 48 (P48) h of storage at room
temperature after collection, the tubes were frozen
(⫺20 °C) for a median of 5 weeks (range, 1–15
weeks). After thawing, RNA was prepared using the
PAXgene Bone Marrow RNA Kit. Two other speci-
mens of 0.5 mL BM each were transferred into
sterile plastic tubes as reference samples and were
processed using the QIAamp RNA Blood Mini Kit
(Qiagen) at the same time points (R2 and R48).
rna yield and integrity lyzed the expression of IL8 gene (interleukin 8) using the
We diluted aliquots of RNA samples in 10 mmol/L Quantitect SYBR Green RT-PCR Kit (Qiagen) on the
Tris 䡠 Cl, pH 7.5, and quantified them using UV spectros- LightCycler威 2.0 System (Roche Applied Science). We
copy. We also used the buffers in which the RNA were chose IL8 because of its known up-regulation after short-
isolated to zero the spectrophotometer in the final dilution term storage of clinical samples (20 ).
of RNA aliquots to be quantified. We analyzed RNA
integrity from aliquots of 1.5 ␮L RNA by use of RNA 6000 Results
Nano reagents and chips on a Bioanalyzer 2100 device Independent of the leukocyte count of the BM specimens
equipped with BioSizing software version A02.12 (Agilent (median, 17 151 leukocytes/␮L; range, 2800 –546 740), the
Technologies). We applied the software RIN calculation handling of the PAXgene Bone Marrow RNA System was
algorithm to RNA fluorescence profiles after separation of easy and convenient. There was no association between
RNA by capillary gel electrophoresis to establish RNA leukocyte count and the yield of total RNA in either
integrity with a score of 0 to 10 points (low to high RNA preparation method (data not shown). Regarding the
integrity). influence of 2-day storage on IL8 gene expression, we
found a huge difference in the unstabilized samples
q-rt-pcr compared with the stabilized samples [median change in
Q-RT-PCR was performed in a 1-step method with 1 ng gene expression, 14.8-fold (R) vs 0.7-fold (P)]. As shown in
total RNA by use of the Quantitect威 SYBR威 Green re- Fig. 2, the CT values of the R48 samples were consistently
verse-transcription (RT)-PCR Kit and Quantitect Primer lower than the CT values of the R2 samples, representing
Assays (Qiagen) for the detection of GATA1, RUNX1, an increase in gene expression. Although the correlation
NCAM1, and SPI1 and for control gene 18S rRNA. We between CT values of P2 and P48 is not optimal (slope,
chose 18S rRNA because general RNA degradation cor- 0.90; R2 ⫽ 0.64), possibly because of missing normaliza-
relates very well with decreased amounts of 18S rRNA. tion with transcripts of a housekeeping gene, the correla-
We performed all assays in duplicate by use of the ABI™ tion for the reference samples was even worse (slope, 0.32;
7300 Real-Time PCR System (Applied Biosystems). Cal- R2 ⫽ 0.08).
culation was done by relative quantification using the
2⫺⌬⌬CT T method (19 ). Only those data with a CT value yield and integrity of rna
⬍35 (cutoff value) were included in the evaluation. Altogether, 180 RNA samples, processed from 45 BM
specimens, fulfilled the inclusion criteria and were ana-
feasibility study lyzed for yield, integrity, and gene expression profiles.
Before the above study, we performed a feasibility study We excluded 19 specimens because of low-yield or low-
with a subset of 31 BM specimens. Along with the quality RNA samples: P2 (n ⫽ 3), P48 (n ⫽ 5), R2 (n ⫽ 12),
quantification of RNA yield and RNA integrity, we ana- and R48 (n ⫽ 15). In 26 samples, criteria for both yield and
Fig. 2. IL8 gene expression in RNA sam-

ples prepared from BM with 48 and 2 h
of storage using either the PAXgene
system (P) or the reference protocol
without RNA stabilization (R).
The diagrams show the cycle threshold (CT) of
the Q-RT-PCR of IL8 for the samples at 2 (x
axis) and 48 (y axis) h of storage for either
preparation method. The black diagonal line
represents an ideal correlation, i.e., no
change in gene expression between the 2
time points; the dashed line indicates the
calculated regression line (P, y ⫽ 0.90 䡠 x,
R2 ⫽ 0.64; R, y ⫽ 0.31 䡠 x, R2 ⫽ 0.08).There
is a high increase in IL8 expression after 48-h
storage for the RNA prepared by the reference
method compared with the PAXgene method.
quality were not fulfilled, 6 samples were of bad quality, The integrity of the isolated RNA using the PAXgene
and 3 samples (all reference samples) were of insufficient system was significantly higher at both time points com-
yield (1.2, 1.2, and 2.6 mg/L, respectively). pared with the reference system: P2, 8.6 (0.2) vs R2,
The number of invalid samples obtained with the 6.8 (0.4), P ⫽ 0.0003, and P48, 8.1 (0.2) vs R48 6.7 (0.5), P ⫽
reference method was ⬃3 times higher than with the 0.008. We found no statistically significant difference
PAXgene system (27 vs 8 samples), demonstrating higher between R2 and R48, but a slight decrease between P2 and
overall RNA quality and quantity with PAXgene. P48 (P ⫽ 0.048). As shown in Fig. 3B, the spreading of RIN
The RNA yield for each method was measured and is more distinct in the RNA prepared by the reference
normalized to 1 ⫻ 107 leukocytes in the input BM sample. method with regard to the 25% to 75% confidence inter-
As shown in Fig. 3A, the overall yield was similar in the val: R2, 5.2–9.1 and R48, 4.1–9.0 vs P2 8.4 –9.3 and P48,
4 RNA isolation procedures. We found no differences in 7.8 – 8.9. As an example, Fig. 3C illustrates the RNA
RNA yield at 2 and 48 h for each method separately integrity of a complete data set with 4 samples (P2, P48,
(ANOVA). R2, and R48).
Fig. 3. RNA yield and integrity.

(A), RNA yield of BM aliquots with 2 and 48 h
of storage at room temperature obtained us-
ing the PAXgene system (P) and the reference
protocol without RNA stabilization (R) (mean
⫹SE). (B), box plot of RIN values from RNA
samples of BM aliquots with 2 and 48 h of
storage at room temperature obtained using
the PAXgene system (P) and the reference
protocol without RNA stabilization (R) (black
bars, median; gray rectangles, 25th to 75th
percentile; black lines, 5th to 95th percentile;
rhombs, minimum/maximum value). (C), rep-
resentative RNA integrity analysis of 4 RNA
samples of a single BM aliquot with 2 and
48 h of storage at room temperature obtained
using the PAXgene system (P) and the refer-
ence protocol without RNA stabilization (R).
paxgene stabilizes gene expression profiles

We analyzed mRNA levels of different genes after RNA
preparation using the PAXgene procedure at P2 and P48.
Differences of ⌬⌬CT, calculated as ⌬CT(P2) ⫺ ⌬CT(P48) for
the control gene (18S rRNA), were homogeneous within a
tight range: GATA1 0.17 (0.10), RUNX1 0.27 (0.17), NCAM1
0.53 (0.23), and SPI1 0.16 (0.09). This results in a high pair-
wise correlation in gene expression for any gene at P48
compared with P2: GATA1 89 (6)%, RUNX1 83 (10)%,
NCAM1 69 (12)%, and SPI1 89 (6)% (Fig. 4).
gene-dependent change of expression level

using standard rna isolation techniques
Regarding expression levels of the same genes after RNA
preparation from unstabilized BM samples using QIAamp
RNA Blood Kit, the results are more heterogeneous. For 2 of
Fig. 5. Differences of transcript levels between RNA samples prepared
the 4 analyzed genes, the differences in ⌬⌬CT, ⌬CT(R2) ⫺
from BM with the PAXgene system and the reference protocol without
⌬CT(R48), were also within the range of ⫾0.5, whereas for RNA stabilization at same test time point with 2 and 48 h of storage.
the 2 other genes, the differences were ⬎1: GATA1 Data are shown as mean (SE).
1.32 (0.19), RUNX1 0.25 (0.18), NCAM1 1.07 (0.23), and SPI1
⫺0.01 (0.14). The resulting expression levels at R48 for the
latter genes were accordingly reduced to 40 (6)%, P ⬍0.0001 Discussion
(GATA1) and 47 (8)%, P ⫽ 0.005 (NCAM1; Fig. 4). In multicenter clinical trials, a centralized analysis of
patient samples is generally performed to minimize inter-
expression levels are not comparable using laboratory variations by applying a standardized method
different rna isolation protocols to all samples. Little is known, however, about the im-
When we compared the 2 RNA isolation procedures with pact of preanalytical sample handling, particularly for
each other, there were major differences in the expression molecular genetic analyses based on mRNA. High inter-
levels of the analyzed genes. We observed lower mean individual and temporal variations of gene expression
(SD) gene expression at P2 vs R2 for all genes: GATA1 patterns have been demonstrated because of genetic back-
70 (12)%, RUNX1 50 (11)%, NCAM1 41 (8)%, and SPI1 ground, age, sex, and the time of day at which the sample
51 (6)%. At the 48-h time point, the changes in gene was taken (21 ). Other similarly important variables affect-
expression for RUNX1 and SPI1 were similar at 45 (7)% ing gene expression profiles include mode of sample
and 43 (6)%, respectively, whereas for the other 2 genes aspiration, use of different anticoagulants, variable tem-
(GATA1 and NCAM1), the decrease in gene expression perature during storage, and time frame between aspira-
using the reference method resulted in significant discrep- tion and processing of the sample.
ancy: 138 (26)% and 76 (16)% (Fig. 5). We carried out a comparative study of a newly devel-
oped system for RNA stabilization in whole BM speci-
mens (PAXgene Bone Marrow RNA System) and unsta-
bilized controls after different times of storage. The time
delays before RNA preparation were chosen to represent
the average shipping time between clinics and laborato-
ries. Although alterations in gene expression may occur
during a very short period of time, our study was
designed to represent realistic operational procedures in a
hospital, where BM punctures (especially in children) are
often performed in an operating room, and a routine
in-house transport time to the laboratory has to be taken
into account.
The promising data from the feasibility study con-
firmed the known up-regulation of IL8 gene expression in
unstabilized specimens. The PAXgene-stabilized BM
specimens with 48 and 2 h of storage showed high
Fig. 4. Change of gene expression in RNA samples prepared from BM accordance in terms of IL8 gene expression, prompting us
with 48 and 2 h of storage using either the PAXgene system (P) or the to perform a more comprehensive study. The analyses of
reference protocol without RNA stabilization (R). purity (calculated as A260/A280 ratio) and RNA integrity
Data are shown as mean (SE). (calculated by RIN) served as tools to determine the
overall quality of the RNA preparation and purification vealed disease marker transcripts, BM stabilization using
procedure in terms of purity and possible degree of the PAXgene system is recommended.
overall RNA degradation. Although the use of cDNA Different principles of RNA isolation may lead to
microarrays allows genomewide gene expression screen- different gene expression levels: whereas the QIAamp
ing, their value for exact quantification is limited (22 ). For RNA isolation procedure includes erythrocyte lysis, dur-
more detailed information about the stabilization of RNA, ing the PAXgene protocol, the RNA of all BM cells is
Q-RT-PCR was performed to quantify mRNA concentra- isolated. Therefore, it is not suitable to directly compare
tions. The genes to be analyzed were selected because of mRNA expression levels derived from different isolation
their attributes as key regulators during different stages of procedures.
hematopoiesis and leukemogenesis, as well as their more
or less distinct organ specificity: NCAM1 and RUNX1 as In conclusion, the PAXgene Bone Marrow RNA System is
representatives of genes with functions (hematopoiesis, suitable to stabilize, isolate, and purify RNA in high
osteogenesis, cell adhesion) on many cell types (hemato- quality and quantity from clinical BM samples. The ability
poietic cells, neurons, muscles) of various differentiated to preserve gene expression is comparable to that of PB
tissues (embryonic cells, differentiated cells), GATA1 and specimens reported in detail by different researchers with
SPI1 having more circumscribed compartments and cell the PAXgene Blood RNA System.
types (erythropoiesis/megakaryopoiesis, myelopoiesis,
and lymphopoiesis).
There were no major differences in the yield of isolated We thank Carolin Augsburg and Heike Balven for excel-
RNA between the 4 different procedures. Regarding the lent technical assistance and Douglas McGarvey for re-
integrity of the RNA, however, the PAXgene system was view of the manuscript. Conflicts of interest: K.G. and J.L.
superior to the reference method of unstabilized BM are full-time employees and participate in the stock op-
samples at both time points. Handling of samples with tions program of Qiagen GmbH.
very high leukocyte counts was not impaired compared
with samples with normal leukocyte counts. The unex-
References
pectedly high integrity of the RNA samples obtained after
1. Breit S, Nees M, Schaefer U, Pfoersich M, Hagemeier C, Muck-
48 h of storage under unstabilized conditions (R48) might enthaler M, et al. Impact of pre-analytical handling on bone marrow
be explained by the RNA preparation procedure: in mRNA gene expression. Br J Haematol 2004;126:231– 43.
contrast to a direct preparation from whole BM, the 2. Feezor RJ, Baker HV, Mindrinos M, Hayden D, Tannahill CL,
QIAamp procedure used here includes erythrocyte lysis Brownstein BH, et al. Whole blood and leukocyte RNA isolation for
as well as the generation of a cell pellet that contains only gene expression analyses. Physiol Genomics 2004;19:247–54.
intact white BM cells. Therefore, possible RNA degrada- 3. Chai V, Vassilakos A, Lee Y, Wright JA, Young AH. Optimization of
tion products are removed with the supernatant. the PAXgene blood RNA extraction system for gene expression
analysis of clinical samples. J Clin Lab Anal 2005;19:182– 8.
Regarding gene expression profiling, the use of RNA
4. Armstrong SA, Staunton JE, Silverman LB, Pieters R, den Boer ML,
stabilization reagents is still controversial: there are some
Minden MD, et al. MLL translocations specify a distinct gene
reports of differences in gene expression resulting from expression profile that distinguishes a unique leukemia. Nat
poor stability of genes analyzed using either microarray Genet 2002;30:41–7.
technology (23 ) or Q-RT-PCR (24, 25 ). In terms of RNA 5. Bourquin JP, Subramanian A, Langebrake C, Reinhardt D, Bernard
quality and gene expression, however, our results from O, Ballerini P, et al. Identification of distinct molecular phenotypes
BM samples show excellent pairwise correlation between in acute megakaryoblastic leukemia by gene expression profiling.
different incubation times using quantitative RT-PCR and Proc Natl Acad Sci U S A 2006;103:3339 – 44.
are in line with those derived from PB (3, 26, 27 ). The 6. Golub TR, Slonim DK, Tamayo P, Huard C, Gaasenbeek M, Mesirov
PAXgene system is appropriate for quantification of gene JP, et al. Molecular classification of cancer: class discovery and
class prediction by gene expression monitoring. Science 1999;
expression levels in BM samples. For some genes, how- 286:531–7.
ever, the reference method resulted in similar gene ex- 7. Moos PJ, Raetz EA, Carlson MA, Szabo A, Smith FE, Willman C, et
pression levels, so the use of a stabilization reagent does al. Identification of gene expression profiles that segregate pa-
not seem to be essential for every gene. This has also been tients with childhood leukemia. Clin Cancer Res 2002;8:3118 –
shown for the quantification of tissue factor (F3) and 30.
vascular endothelial growth factor, with comparable re- 8. Yeoh EJ, Ross ME, Shurtleff SA, Williams WK, Patel D, Mahfouz R,
sults derived from stabilized and unstabilized samples et al. Classification, subtype discovery, and prediction of outcome
(28 ). Therefore, each researcher must carefully evaluate in pediatric acute lymphoblastic leukemia by gene expression
profiling. Cancer Cell 2002;1:133– 43.
the preanalytical handling of the samples and choose a
9. Viehmann S, Teigler-Schlegel A, Bruch J, Langebrake C, Reinhardt
system for RNA ex vivo stabilization in case of gene
D, Harbott J. Monitoring of minimal residual disease (MRD) by
expression analysis in later project phases with unknown real-time quantitative reverse transcription PCR (RQ-RT-PCR) in
targets at start of study. In studies where RNA isolates are childhood acute myeloid leukemia with AML1/ETO rearrangement.
stored for a long time before analysis of currently unre- Leukemia 2003;17:1130 – 6.
10. Bonnet D, Dick JE. Human acute myeloid leukemia is organized as temperature and storage duration of human whole blood. Clin Lab
a hierarchy that originates from a primitive hematopoietic cell. Nat Haematol 2002;24:337– 41.
Med 1997;3:730 –7. 21. Whitney AR, Diehn M, Popper SJ, Alizadeh AA, Boldrick JC, Relman
11. Osato M, Ito Y. Increased dosage of the RUNX1/AML1 gene: a DA, et al. Individuality and variation in gene expression patterns in
third mode of RUNX leukemia? Crit Rev Eukaryot Gene Expr human blood. Proc Natl Acad Sci U S A 2003;100:1896 –901.
2005;15:217–28. 22. Wang Y, Barbacioru C, Hyland F, Xiao W, Hunkapiller KL, Blake J,
12. Wang X, Hisha H, Taketani S, Inaba M, Li Q, Cui W, et al. Neural et al. Large scale real-time PCR validation on gene expression
cell adhesion molecule contributes to hemopoiesis-supporting measurements from two commercial long-oligonucleotide microar-
capacity of stromal cell lines. Stem Cells 2005;23:1389 –99. rays. BMC Genomics 2006;7:59.
13. Weiss MJ, Orkin SH. GATA transcription factors: key regulators of 23. Thach DC, Lin B, Walter E, Kruzelock R, Rowley RK, Tibbetts C, et
hematopoiesis. Exp Hematol 1995;23:99 –107. al. Assessment of two methods for handling blood in collection
14. Wechsler J, Greene M, McDevitt MA, Anastasi J, Karp JE, Le Beau tubes with RNA stabilizing agent for surveillance of gene expres-
MM, et al. Acquired mutations in GATA1 in the megakaryoblastic sion profiles with high density microarrays. J Immunol Methods
leukemia of Down syndrome. Nat Genet 2002;32:148 –52. 2003;283:269 –79.
15. Cook WD, McCaw BJ, Herring CD, John DL, Foote SJ, Nutt SL, et 24. Thorn I, Olsson-Stromberg U, Ohlsen C, Jonsson AM, Klangby U,
al. PU. 1 is a suppressor of myeloid leukemia, inactivated in mice Simonsson B, et al. The impact of RNA stabilization on minimal
by gene deletion and mutation of its DNA-binding domain. Blood residual disease assessment in chronic myeloid leukemia.
2004;104:3437– 44. Haematologica 2005;90:1471– 6.
16. Stirewalt DL. Fine-tuning PU. 1. Nat Genet 2004;36:550 –1. 25. Kagedal B, Lindqvist M, Farneback M, Lenner L, Peterson C.
17. Rosenbauer F, Wagner K, Kutok JL, Iwasaki H, Le Beau MM, Failure of the PAXgene Blood RNA System to maintain mRNA
Okuno Y, et al. Acute myeloid leukemia induced by graded stability in whole blood. Clin Chem Lab Med 2005;43:1190 –2.
reduction of a lineage-specific transcription factor, PU. 1. Nat 26. Wang J, Robinson JF, Khan HM, Carter DE, McKinney J, Miskie BA,
Genet 2004;36:624 –30. et al. Optimizing RNA extraction yield from whole blood for
18. Rainen L, Oelmueller U, Jurgensen S, Wyrich R, Ballas C, Schram microarray gene expression analysis. Clin Biochem 2004;37:
J, et al. Stabilization of mRNA expression in whole blood samples. 741– 4.
Clin Chem 2002;48:1883–90. 27. Spijker S, van de Leemput JC, Hoekstra C, Boomsma DI, Smit AB.
19. Livak KJ, Schmittgen TD. Analysis of relative gene expression data Profiling gene expression in whole blood samples following an
using real-time quantitative PCR and the 2(-Delta Delta C(T)) in-vitro challenge. Twin Res 2004;7:564 –70.
method. Methods 2001;25:402– 8. 28. Isaksson HS, Nilsson TK. Preanalytical aspects of quantitative
20. Tanner MA, Berk LS, Felten DL, Blidy AD, Bit SL, Ruff DW. TaqMan real-time RT-PCR: Applications for TF and VEGF mRNA
Substantial changes in gene expression level due to the storage quantification. Clin Biochem 2006;39:373–7.
594 –599 (2007) Molecular Diagnostics
and Genetics
One-Step Rapid Reverse Transcription–PCR Assay

for Detecting and Typing Dengue Viruses with GC
Tail and Induced Fluorescence Resonance Energy
Transfer Techniques for Melting Temperature and
Color Multiplexing
Constance L.H. Lo,1 Shea Ping Yip,1 Peter K.C. Cheng,2 Tony S.S. To,1
Wilina W.L. Lim,2 and Polly H.M. Leung1*
Background: Dengue fever is an arthropod-borne infec- limits of DEN-1 to DEN-4 were 1.64 ⴛ 10ⴚ4, 1.05 ⴛ 10ⴚ3,
tion caused by dengue viruses (DVs; DEN-1 to DEN-4). 8.15 ⴛ 10ⴚ4, and 5.80 ⴛ 10ⴚ3 plaque-forming units per
Early diagnosis is critical to prevent severe disease reaction, respectively. The assay had a dynamic range of
progression and the spreading of DV because no vac- 103–108 plaque-forming units/L and could be performed
cine or specific treatment is available; therefore, a rapid in 2 h.
and specific diagnostic assay capable of detecting and Conclusions: A single-tube, 1-step reverse transcrip-
typing all serotypes would be ideal. tion–PCR assay based on Tm and color multiplexing was
Methods: We amplified RNA samples from all 4 DV developed for detecting and typing all 4 DV serotypes.
serotypes and Japanese encephalitis virus with 4 sero- © 2007 American Association for Clinical Chemistry
type-specific forward primers and a universal species-
specific reverse primer. DEN-1 and DEN-3 forward Dengue fever is caused by dengue virus (DV)3 serotypes
primers were labeled at their 5ⴕ ends with BODIPY 1– 4 (DEN-1 to DEN-4) and transmitted by infected female
630/650 and Cy5.5, respectively. DEN-1 and DEN-3 am- mosquitoes (Aedes aegypti and Aedes albopictus) (1 ). More
plicons were detected by their characteristic emission than 50 million cases of dengue fever and dengue hem-
generated from induced fluorescence resonance energy orrhagic fever are reported annually worldwide (2 ). The
transfer. The presence of DEN-2 and DEN-4 amplicons infection became epidemic (⬎700 000 cases) in Southeast
was indicated by SYBR Green I (SGI) signals at specific Asia between 2001 and 2005. In tropical and subtropical
amplicon melting temperatures (Tms). countries, 52% of the population is at risk for the infection
Results: Fluorescence signals with specific emission (3 ). Asymptomatic or flu-like symptomatic dengue fever
wavelengths were obtained from DEN-1 and DEN-3. may delay diagnosis, and the disease may progress to
SGI melting profiles showed a Tm difference between severe dengue hemorrhagic fever and lethal dengue
DEN-2 and DEN-4 of 4.7 °C, which was sufficient for shock syndrome (4, 5 ). Because vaccination and selective
differentiating these 2 serotypes. The primers did not treatment for dengue infection are not available, rapid
amplify the Japanese encephalitis virus. The detection and specific diagnosis during early onset of the disease is
critical and can help prevent progression of the disease
through monitoring and supportive therapy (5, 6 ). Most
1
Department of Health Technology and Informatics, The Hong Kong importantly, early diagnosis can help prevent the spread
Polytechnic University, Hong Kong SAR, China. of DV via mosquito bites from infected patients during
2
Public Health Laboratory Centre, Centre for Health Protection, Depart-
ment of Health, Hong Kong SAR, China.
the viremic phase.
* Address correspondence to this author at: Department of Health Tech-
nology and Informatics, The Hong Kong Polytechnic University, Hung Hom,
Kowloon, Hong Kong SAR, China. Fax 852-2362-4365; e-mail htpolly@
3
inet.polyu.edu.hk. Nonstandard abbreviations: DV, dengue virus; RT, reverse transcription;
Received July 29, 2006; accepted January 25, 2007. Tm, melting temperature; iFRET, induced fluorescence resonance energy
Previously published online at DOI: 10.1373/clinchem.2006.077446 transfer; SGI, SYBR Green I; Ct, threshold cycle.
594
Reverse transcription (RT)–PCR analysis has been used (Table 1; Fig. 1B). Forward primers for DEN-1 and DEN-3
in the diagnosis of flavivirus infections (6, 7 ). DV shares were labeled at the 5⬘ end with BODIPY 630/650 (Molec-
genomic homology with other flaviviruses, and genotypic ular Probes/Invitrogen) and Cy5.5, respectively. We used
variations occur within the same serotype (8, 9 ). Thus, deoxyinosine in the DEN-2 forward primer (D2F) to
species- and serotype-specific detection of DV in the same overcome the genotypic variations within DEN-2. A GC
tube is difficult. Currently, TaqMan assays (Applied Bio- tail was added to the 5⬘ end of the DEN-4 forward primer
systems) are widely used for dengue detection (10-15 ). (D4F) to give a product Tm that was higher than that of
Most of these assays amplify the 4 serotypes in 4 separate DEN-2 (predicted difference, 3.6 °C; Table 1). Because 9 of
tubes (10, 11 ). Only recently has a 1-tube fourplex Taq- 10 consecutive nucleotides upstream of the 3⬘ end of the
Man assay been developed (13 ); however, the 4 pairs of D4F primer were identical among all 4 DV serotypes (Fig.
primers and the 4 distinct fluorophores for dual-labeled 1A, asterisks), we applied a simple strategy of sequence-
probes increase the complexity and cost of the assay. specific PCR for accurate typing of DEN-4 to minimize
Moreover, probe design is less flexible because a perfect nonspecific annealing of primer D4F to non–DEN-4 tem-
match with the target sequence is required. The LightCy- plates. The specificity of primer D4F was enhanced with a
cler approach (Roche Diagnostics), another platform with deliberately destabilizing mismatched G base close to the
high diagnostic value, allows more efficient cycling for 3⬘ end of the primer (18 ). This change ensured that only
rapid detection of DVs (16 ). The detection strategies DEN-4 was amplified under stringent PCR conditions.
provided by the LightCycler methods facilitate multiplex- The amplicons were between 103 and 111 bp in length
ing of melting temperature (Tm; temperature at which (Table 1). A BLAST analysis demonstrated all primers to
50% of double-stranded DNA is denatured) and color in be specific for DV.
the assay (16 ). In this study, we aimed to develop a
single-step LightCycler assay capable of detecting and 1-step rt-pcr assay
typing DVs. Using the Qiagen QIAamp viral RNA extraction re-
agent set, we extracted 24 RNA samples of known DV
Materials and Methods serotypes (n ⫽ 6 for each serotype) and 1 RNA sample
assay design
of Japanese encephalitis virus from cell culture super-
We applied 2 strategies offered by the LightCycler for
natants or clinical serum samples (Table 2). We con-
developing a 1-step assay for detecting and typing DVs.
firmed the presence of DV in the clinical samples with
The 1st strategy was to increase the Tm of an amplicon by
a separate nested RT-PCR assay (19 ). The RT-PCR assay
means of a primer tagged with a GC tail (5⬘-CGC GCC
in this study was performed with the LightCycler RNA
GGC CGC GG-3⬘), which allowed distinguishing between
Amplification Kit HybProbe (Roche Diagnostics) in
DEN-2 and DEN-4 in channel F1 (Tm multiplexing). The
glass capillaries in a LightCycler 1.5 System. The 20-␮L
2nd strategy used induced fluorescence resonance energy
reaction mixture contained the following components:
transfer (iFRET) to detect DEN-1 in channel F2 and DEN-3
in channel F3 (color multiplexing). We incorporated flu- 1⫻ Roche LightCycler RT-PCR reaction mix HybProbe,
orogenic primers specific to DEN-1 and DEN-3 at the ends 5 mmol/L MgCl2, 0.5 ␮mol/L D1– 4R primer, 0.4
of the amplicons by means of the PCR. SYBR Green I (SGI) ␮mol/L D1F primer, 0.3 ␮mol/L each of D2F and D3F
molecules (Molecular Probes/Invitrogen) intercalating primers, 0.5 ␮mol/L D4F primer, 1⫻ SGI, and 0.5 ␮L
with the duplex amplicons acted as an energy donor for RNA sample. The 1⫻ SGI concentration was used as
the acceptor fluorophores tagged to the primers. As a suggested in the protocol of the LightCycler RNA
result, light with a longer wavelength was emitted from Amplification Kit SYBR Green I. Reaction conditions
the excited acceptor fluorophore. During melting-curve were as follows: an RT step at 55 °C for 30 min and
analysis, the progressive rise in temperature diminished denaturation at 95 °C for 1 min, followed by 45 PCR
the interaction between SGI and acceptor fluorophore as cycles of 95 °C for 10 s, 59 °C for 15 s, and 72 °C for 20 s.
the duplex amplicons dissociated. A dramatic reduction The assay ended with a melting-curve analysis: heating
in the fluorescence signal occurred at the Tm of the to 95 °C without a hold, rapid ramping to 59 °C and
amplicon. holding for 30 s, gradual heating to 95 °C at 0.1 °C/s,
and final cooling to 40 °C. The identities of the specific
primer design products were confirmed by agarose gel electrophoresis
We retrieved complete genome sequences of DVs (n ⫽ and DNA sequencing with an ABI 3130 Genetic Ana-
112) from GenBank and aligned them with MAVID (17 ). lyzer (Applied Biosystems). We determined the detec-
The moderately conserved 3⬘ untranslated region was tion limit of each serotype by 10-fold serial dilutions of
selected for primer design (Fig. 1A) with the OLIGO RNA samples extracted from cell culture supernatants
software (Molecular Biology Insights). All primers were with known viral titers. The viral titers for DEN-1,
purchased from Integrated DNA Technologies. The 4 DEN-2, DEN-3, and DEN-4 were 3.27 ⫻ 106, 2.10 ⫻ 108,
serotypes were amplified with 4 serotype-specific for- 1.63 ⫻ 107, and 1.16 ⫻ 108 plaque-forming units/L,
ward primers and a common DV-specific reverse primer respectively.
596 Lo et al.: GC Tail and Induced FRET for Tm and Color Multiplexing
Fig. 1. Primer design for the 1-step RT-PCR LightCycler assay.

(A), alignment of amplified regions of DEN-1 to DEN-4 and positions of forward and reverse primers. Underlined sequences indicate the primer regions, and asterisks
indicate the bases conserved among the 4 serotypes. (B), alignment of primer regions of all primers. Bo, BODIPY 630/650 fluorophore; I, deoxyinosine base; Cy, Cy5.5
fluorophore; Tail, GC tail (5⬘-CGC GCC GGC CGC GG-3⬘). There are 112 aligned DV genome sequences: 31 DEN-1, 52 DEN-2, 25 DEN-3, and 4 DEN-4. The numbers
of viral sequences perfectly matching or possessing 1–2 mismatches with the respective forward primers (D1F to D4F) and the common reverse primer D1– 4R, their
proportions, and the percentages within each serotype are indicated to the right of the aligned sequences. In all forward primers, the 8 bases (AAGCTGTA) highly
conserved in dengue viral genomes are underlined. Of the 112 aligned genome sequences, only 1 DEN-1 and 1 DEN-3 genome differ from this consensus sequence
(by a single base). In the D4F primer, the boldface G base (near the 3⬘ end of the primer and the 3⬘ base of the 8-base consensus sequence) is a deliberate mismatch
with the genome sequence to enhance the DEN-4 specificity of this primer. Most (96%) of the genome sequences show a perfect match with the common reverse
primer, and only 4 genome sequences (4%) show a single mismatch.
Table 1. Primer and amplicon characteristics.

Amplicon
a
Serotype Primer sequence (5ⴕ-3ⴕ) Positions (accession no.) Size, bp Predicted Tm, °C b Observed Tm, °C, % c
DEN-1 D1F: Bo-GGA AGC TGT ACC CTG GTG GT 10 569–10 588 103 84.4 83.3 (0.5), 0.6
D1–4R: GAG ACA GCA GGA TCT CTG GTC 10 651–10 671
(AY277666)
DEN-2 D2F: GAG ATG AAG CTG TAG TCT CIC TG 10 554–10 639 105 82.4 81.0 (0.5), 0.6
(AY858035)
DEN-3 D3F: Cy-AGG GAA GCT GTA CCT CCT TGC A 10 541–10 562 103 84.0 83.9 (0.2), 0.3
(AB189128)
DEN-4 D4F: Tail-GAA GCC AGG AGG A AG CTG TGC T 10 475–10 496 111 86.0 85.7 (0.2), 0.2
(AF326573)
a
Note that the reverse primer, D1– 4R, was common for all serotypes, although the base positions were different in the genome sequences of different serotypes.
Refer to Fig. 1B legend for details.
b
Predicted amplicon Tms were obtained with OLIGO software.
c
Data are presented as the mean (SD), CV. Observed amplicon Tms were the mean of 8 readings measured in separate runs. The Tms were obtained from channel
F1 for DEN-2 and DEN-4, channel F2 for DEN-1, and channel F3 for DEN-3. Although not used for typing, Tms were also obtained from channel F1 for DEN-1 (mean,
83.1 °C; SD, 0.3 °C; CV, 0.4%), and for DEN-3 (mean, 82.9 °C; SD, 0.2 °C; CV, 0.2%).
Table 2. Viral strains analyzed with the 1-step

RT-PCR assay.
No. of strains extracted
Viral culture, Clinical serum,

Virus Serotype nⴝ9 n ⴝ 16
Dengue DEN-1 2 4
DEN-2 2 4
DEN-3 2 4
DEN-4 2 4
Japanese encephalitis 1
Results
The RT-PCR assay specifically detected the 4 DV sero-
types, with no cross-reactivity between the primers (Fig.
2A). The Japanese encephalitis virus was not amplified by
the assay. Twelve (75%) of the 16 clinical serum samples
known to be DV positive were accurately identified and
serotyped. Channel F1 gave melting peaks due to SGI
fluorescence for all 4 serotypes. In particular, the Tm
difference between DEN-2 and DEN-4 was 4.7 °C, which
was sufficient to differentiate these 2 serotypes. DEN-1
and DEN-3 had very similar Tms in channel F1; however,
DEN-1 was characterized by its Tm (83.3 °C) in channel F2
owing to the iFRET between SGI and BODIPY 630/650,
and DEN-3 was characterized by its Tm (83.9 °C) in
channel F3 because of the iFRET between SGI and Cy5.5.
The measurement of Tm was very precise, with within-
run CVs ranging from 0.1% to 0.3% and between-run CVs
ranging from 0.2% to 0.6% (Table 1). The detection limits
in plaque-forming units per reaction and the correspond-
ing threshold cycle (Ct) value for DEN-1 (1.64 ⫻ 10⫺4 and
30.8) and DEN-3 (8.15 ⫻ 10⫺4 and 27.8) were 1 log lower
than those for DEN-2 (1.05 ⫻ 10⫺3 and 33.7) and DEN-4
(5.80 ⫻ 10⫺3 and 30.7). The assay had a dynamic range of
detection (103–108 plaque-forming units/L). The correla-
tion coefficients for the correlation between viral concen-
tration and Ct value were ⫺0.99 for both DEN-1 and
DEN-3, ⫺0.98 for DEN-4, and ⫺0.95 for DEN-2 (Fig. 2B).
Amplification plots are available (see figures in the Data
Supplement that accompanies the online version of this
article at http://www.clinchem.org/content/vol53/
issue4). Amplification efficiency was highest for DEN-1
and lowest for DEN-4, as calculated from the following
equation: efficiency ⫽ 10(⫺1/slope) (20 ).
Discussion
Our assay was based on Tm and color multiplexing for the
detection and typing of DV. The unique amplicon Tm of
each serotype in different channels (F1 to F3) allowed
specific detection of DV. Adding a GC tail improves the Fig. 2. Characteristics of the 1-step RT-PCR LightCycler assay.
detection of single-nucleotide polymorphisms (21, 22 ) (A), derivative melting curves showing the melting peaks characteristic of the 4
and has a marked effect on Tm differentiation for small serotypes detected in 3 LightCycler channels (F1–F3). NTC, no-template control.
Amplicon Tms are measured in channel F1 (SGI fluorescence) for DEN-2 and DEN-4
amplicons (21, 23 ), such as the ⬃100-bp amplicons in this (i), in channel F2 (BODIPY 630/650 fluorescence) for DEN-1 (ii), and in channel F3
study. By contrast, minor base changes within amplicons (Cy5.5 fluorescence) for DEN-3 (iii). In channel F3, inverted melting peaks due to
DEN-2, and sometimes to DEN-1 and DEN-4, may be detected. The reason for these
typically do not produce detectable Tm changes. Because inverted peaks is not known; however, they do not interfere with DEN-3 typing. (B),
the amplicon sequence is highly conserved within the calibration curves for the 4 serotypes. PFU, plaque-forming unit.
598 Lo et al.: GC Tail and Induced FRET for Tm and Color Multiplexing
same DV serotype, a 4.7 °C Tm difference would be able to with those of other available RT-PCR assays for dengue
accommodate the possible minor genotypic variations in diagnosis (7, 13 ).
DEN-2 and DEN-4. Positive controls for all 4 serotypes and a no-template
As expected, the amplicons from DEN-1 and DEN-3 negative control should be run in parallel with test
had very similar Tms (Table 1) and thus could not be samples. This practice not only provides a reference for
distinguished in channel F1. Classically, 2 hybridization interpretation but also helps to monitor the accurateness
probes (1 donor and 1 acceptor) are required for fluores- of DV detection and typing, which is particularly impor-
cence resonance energy transfer to occur. Genotypic vari- tant to recognize if deterioration of reagents has occurred.
ations in the genome of the same serotype, however, make We once observed a systematic difference in results when
it difficult, if not impossible, to design 2 hybridization we used a new batch of reagent sets; this systematic
probes for specific detection and typing. This study com- difference led to a Tm shift in all channels and for all
bined fluorogenic primers and SGI to achieve the same serotypes. The Tm shift was determined to have been
goal via iFRET. In fact, SGI is an effective energy donor for caused by improper storage conditions during delivery.
fluorescence resonance energy transfer, and the fluores- When the samples were retested with another batch of the
cence generated by iFRET is 40 times greater than with the reagent sets that had been delivered under proper storage
classic fluorescence resonance energy transfer (24 ). This conditions, the Tm shift was negligible. Variations be-
assay produced distinct melting peaks due to BODIPY tween different batches of reagent sets have previously
630/650 in channel F2 for DEN-1 and due to Cy 5.5 in been reported and could lead to a 2.5-fold difference in
channel F3 for DEN-3. We chose these 2 less expensive mRNA quantification (28 ).
dyes instead of LC Red 640 and LC Red 705, respectively, The entire assay requires 2 h (40 min for preparation
to reduce costs (25 ). A 1-step LightCycler assay based on plus 80 min for RT-PCR reaction/melting-curve analysis),
the biprobe system and developed to detect only DEN-2 compared with ⱖ3 h for the conventional TaqMan assays
(26 ) is similar in principle to the iFRET assay. (11, 13 ). Our assay is also more economical, because the
primers are used for both detection and typing, only 2
Four (25%) of the serum samples known to be DV
primers have to be labeled with fluorophores, and no
positive were not detected by this assay, but these false-
additional fluorescence probes are required.
negative samples tested either negative or weakly positive
in the 1st-round PCR of a nested RT-PCR assay, which is
a 2-step nonquantitative assay. These results indicated
that the DV titers in the samples might be too low to be This study was supported by Internal Competitive Re-
detected. The RNA sample volume per reaction was search Grant G-YD97 of The Hong Kong Polytechnic
believed and then confirmed to be the major factor University. Purchase of the 3130 Genetic Analyzer was
affecting the sensitivity of our assay. The input sample supported by funds from The Hong Kong Polytechnic
volume for the nested RT-PCR assay was 10-fold larger University (1.55.27.DD02). The DV RNA samples were
than that for our assay. The sensitivity of our assay was supplied by Public Health Laboratory Services Branch,
improved by 12.5% (10 vs 12 of 16 samples) by increasing Centre for Health Protection, Hong Kong SAR, China.
the RNA sample volume from 0.5 to 5 ␮L. This result Disclosures: None declared.
showed that the sample-volume effect became magnified
when the viral load in the serum was low. We suggest the References
use of larger sample volumes for clinical samples. For 1. Moncayo AC, Fernandez Z, Ortiz D, Diallo M, Sall A, Hartman S, et
precious RNA samples from suspected DV-positive pa- al. Dengue emergence and adaptation to peridomestic mosqui-
toes. Emerg Infect Dis 2004;10:1790 – 6.
tients, we suggest a 0.5-␮L input sample volume for
2. World Health Organization Regional Office for South-East Asia.
screening, followed by confirmation of a negative result
Dengue/DHF: introduction. http://www.searo.who.int/en/Section10/
with a larger sample volume. Section332_1103.htm (accessed June 2006).
The present assay gave a PCR efficiency ⬎2, which is 3. World Health Organization Regional Office for South-East Asia.
beyond the maximum theoretical value (efficiency ⫽ 2; Dengue/DHF: situation of dengue/DHF in SEA. http://www.
slope ⫽ ⫺3.3). The PCR efficiency was calculated from the searo.who.int/en/Section10/Section332_1098.htm (accessed
slope of the calibration curve, ⫺2.78 to ⫺2.27 in this assay, June 2006).
which was generated from the Ct values of a dilution 4. Groen J, Koraka P, Velzing J, Copra C, Osterhaus AD. Evaluation of
series of DV RNA. The presence of primer dimers in the six immunoassays for detection of dengue virus-specific immuno-
reaction mixture increased the overall SGI signal, which globulin M and G antibodies. Clin Diagn Lab Immunol 2000;7:
867–71.
in turn lowered the Ct values and yielded a PCR effi-
5. Kautner I, Robinson MJ, Kuhnle U. Dengue virus infection: epide-
ciency ⬎2. Similar findings were noted in a recent study
miology, pathogenesis, clinical presentation, diagnosis, and pre-
that used a Roche Diagnostics reagent set for amplifying vention. J Pediatr 1997;131:516 –24.
Chlamydia pneumoniae in the LightCycler (27 ). Neverthe- 6. Tsai TF. Flaviviruses (yellow fever, dengue, dengue hemorrhagic
less, the sensitivity and specificity of our assay were not fever, Japanese encephalitis, St. Louis encephalitis, tick-borne
affected, because the PCR efficiency compared favorably encephalitis). In: Mandell GL, Bennett JE, Dolin R, eds. Mandell,
Douglas and Benett’s Principles and Practice of Infectious Dis- amplification refractory mutation system (ARMS). Nucleic Acids
eases. Philadelphia: Churchill Livingstone, 2000:1715–36. Res 1989;17:2503–16.
7. Lanciotti RS. Molecular amplification assays for the detection of 19. Lanciotti RS, Calisher CH, Gubler DJ, Chang GJ, Vorndam AV.
flaviviruses. Adv Virus Res 2003;61:67–99. Rapid detection and typing of dengue viruses from clinical sam-
8. Lindenbach BD, Rice CM. Molecular biology of flaviviruses. Adv ples by using reverse transcriptase-polymerase chain reaction.
Virus Res 2003;59:23– 61. J Clin Microbiol 1992;30:545–51.
9. Shurtleff AC, Beasley DWC, Chen JJY, Ni H, Suderman MT, Wang 20. Rasmaussen RP. Quantification on the LightCycler. In: Meur S,
H, et al. Genetic variation in the 3⬘ non-coding region of dengue Wittwer C, Nakagawara K, eds. Rapid Cycler Real-Time PCR:
viruses. Virology 2001;281:75– 87. Methods and Applications. Berlin: Springer-Verlag, 2001:21–34.
10. Warrilow D, Northill JA, Pyke A, Smith GA. Single rapid TaqMan 21. Papp AC, Pinsonneault JK, Cooke G, Sadee W. Single nucleotide
fluorogenic probe based PCR assay that detects all four dengue polymorphism genotyping using allele-specific PCR and fluores-
serotypes. J Med Virol 2002;66:524 – 8. cence melting curves. Biotechniques 2003;34:1068 –72.
11. Laue T, Emmerich P, Schmitz H. Detection of dengue virus RNA in 22. Sheffield VC, Cox DR, Lerman LS, Myers RM. Attachment of a
patients after primary or secondary dengue infection by using the 40-base-pair G ⫹ C-rich sequence (GC-clamp) to genomic DNA
TaqMan automated amplification system. J Clin Microbiol 1999; fragments by the polymerase chain reaction results in improved
37:2543–7. detection of single-base changes. Proc Natl Acad Sci U S A
12. Ito M, Takasaki T, Yamada K, Nerome R, Tajima S, Kurane I. 1989;86:232– 6.
Development and evaluation of fluorogenic TaqMan reverse tran-
23. Liew M, Pryor R, Palais R, Meadows C, Erali M, Lyon E, et al.
scriptase PCR assays for detection of dengue virus types 1 to 4.
Genotyping of single-nucleotide polymorphisms by high-resolution
J Clin Microbiol 2004;42:5935–7.
melting of small amplicons. Clin Chem 2004;50:1156 – 64.
13. Johnson BW, Russell BJ, Lanciotti RS. Serotype-specific detection
24. Howell WM, Jobs M, Brookes AJ. iFRET: an improved fluorescence
of dengue viruses in a fourplex real-time reverse transcriptase
system for DNA-melting analysis. Genome Res 2002;12:1401–7.
PCR assay. J Clin Microbiol 2005;43:4977– 83.
14. Callahan JD, Wu SJ, Dion-Schultz A, Mangold BE, Peruski LF, 25. Yip SP, Lee SY, To SS, Wong ML. Improved real-time PCR assay
Watts DM, et al. Development and evaluation of serotype- and for homogeneous multiplex genotyping of four CYP2C9 alleles
group-specific fluorogenic reverse transcriptase PCR (TaqMan) with hybridization probes. Clin Chem 2003;49:2109 –11.
assays for dengue virus. J Clin Microbiol 2001;39:4119 –24. 26. Tan BH, See E, Lim E, Yap EPH. Development of one-step RT-PCR
15. Houng HS, Chung-Ming Chen R, Vaughn DW, Kanesa-thasan N. for the detection of an RNA virus, dengue, on the capillary
Development of a fluorogenic RT-PCR system for quantitative real-time thermocycler. In: Reischl U, Wittwer C, Cockerill F, eds.
identification of dengue virus serotypes 1– 4 using conserved and Rapid Cycle Real-Time PCR: Methods and Applications: Microbiol-
serotype-specific 3⬘ noncoding sequences. J Virol Methods 2001; ogy and Food Analysis. Berlin: Springer-Verlag, 2002:241– 8.
95:19 –32. 27. Mygind T, Birkelund S, Birkebaek NH, Ostergaard L, Jensen JS,
16. Tan BH, Lim EA, Liaw JC, Seah SG, Yap EP. Diagnostic value of Christiansen G. Determination of PCR efficiency in chelex-100
real-time capillary thermal cycler in virus detection. Expert Rev Mol purified clinical samples and comparison of real-time quantitative
Diagn 2004;4:219 –30. PCR and conventional PCR for detection of Chlamydia pneu-
17. Bray N, Pachter L. MAVID: constrained ancestral alignment of moniae. BMC Microbiol 2002;2:17.
multiple sequences. Genome Res 2004;14:693–9. 28. Bustin SA. Quantification of mRNA using real-time reverse tran-
18. Newton CR, Graham A, Heptinstall LE, Powell SJ, Summers C, scription PCR (RT-PCR): trends and problems. J Mol Endocrinol
Kalsheker N, et al. Analysis of any point mutation in DNA. The 2002;29:23–39.
600 – 605 (2007) Molecular Diagnostics
and Genetics
Candidate Genetic Risk Factors of Stroke: Results

of a Multilocus Genotyping Assay
Wolfgang Lalouschek,1 Georg Endler,2 Martin Schillinger,3 Kety Hsieh,2
Wilfried Lang,4 Suzanne Cheng,5 Peter Bauer,6 Oswald Wagner,2† and
Christine Mannhalter2†*
Background: Epidemiological studies indicate that ge- an odds ratio of 2, and no association with stroke risk
netic factors play a role in the risk of stroke, particularly was statistically significant.
in younger individuals, but the role of single-nucleotide Conclusions: Our results do not indicate a clinically
polymorphisms (SNPs) is controversial. We tested the relevant role of any of the investigated SNPs for stroke
possible association of a number of previously de- risk in individuals hospitalized for ischemic stroke/
scribed SNPs with stroke risk. transient ischemic attack before or at 60 years of age.
Methods: We investigated the prevalence of 60 poly- These results are in accordance with previous meta-
morphisms located in 35 genes in 450 white patients analyses showing at most a very modest or no signifi-
who suffered an acute stroke or transient ischemic cant effect of these SNPs on stroke risk.
attack before the age of 60 years and in 817 healthy © 2007 American Association for Clinical Chemistry
control individuals by a multilocus PCR-based assay.
The controls were randomly selected from attendees of From epidemiological studies it can be concluded that
a health service program. Genetic variations were de- genetic predisposition plays a role in premature stroke
tected by hybridization to nylon strips (Roche Molecu- (1, 2 ). The involvement of several genetic factors reported
lar Systems) containing detection oligonucleotides for to play a role has not yet been definitely proven, however,
the SNPs. We used P values of <0.05 for confirmatory and their contributions to the pathomechanisms of stroke
analysis of the SNPs in the genes for APOE (allele 4), are still poorly understood (3–5 ). Although a large
angiotensin converting enzyme, factor V, prothrombin, amount of data is available associating single-nucleotide
and methylenetetrahydrofolate reductase. To account polymorphisms (SNPs)7 in candidate genes with the risk
for multiple testing we defined a P value of <0.001 as of ischemic stroke, the results for individual polymor-
statistically significant for all exploratory tests. The phisms and genes are controversial. For only a few genes
genes represented in the test panel by more than 1 SNP metaanalyses provide some evidence for a minor contri-
were also evaluated by haplotype analysis. bution to stroke risk, whereas for the majority of poly-
Results: Frequencies of all 60 tested SNPs among pa- morphisms the influence on the odds for stroke are
tients and controls were very similar. No SNP reached unclear (6 ).
We performed a case-control study in which patients
who suffered an acute ischemic cerebrovascular event
[ischemic stroke or transient ischemic attack (TIA)] before
1
University Clinic of Neurology, 2 Clinical Institute of Medical and the age of 60 years were compared with random controls
Chemical Laboratory Diagnostics, 3 Department of Angiology, and 6 Institute participating in a health service program who were free of
for Medical Statistics, Medical University Vienna, Vienna, Austria.
4
Department of Neurology, Hospital Barmherzige Brueder, Vienna, Aus-
a history of vascular events. We evaluated 60 common
tria. variations in 35 genes that have been considered as
5
Department of Human Genetics, Roche Molecular Systems, Inc., Alam- potential genetic risk factors in vascular diseases.
eda, CA.
† These authors contributed equally to this work.
* Address correspondence to this author at: Clinical Institute of Medical
and Chemical Laboratory Diagnostics, Waehringer Guertel 18-20, 1097 Vienna,
7
Austria. Fax 43-1-40400-2096; e-mail christine.mannhalter@meduniwien.ac.at. Nonstandard abbreviations: SNP, single-nucleotide polymorphism; TIA,
Received May 15, 2006; accepted January 23, 2007. transient ischemic attack; IQR, interquartile range; OR, odds ratio; CI, confi-
Previously published online at DOI: 10.1373/clinchem.2006.073494 dence interval; MTHFR, methylenetetrahydrofolate reductase.
600
Patients, Materials and Methods based mainly on associations with cardiovascular

patients and controls diseases.
The study population included 450 consecutive white The SNPs (see Table 1 in the Data Supplement that
patients who suffered an acute ischemic stroke (n ⫽ 358; accompanies the online version of this article at http://
80%) or TIA (n ⫽ 92; 20%) before the age of 60 years [168 www.clinchem.org/content/vol53/issue4) were geno-
female, 282 male; median age 52 years; interquartile range typed by a multilocus PCR-based assay, essentially as
(IQR) 45–57 years] who were documented in the Vienna described (10 ). Briefly, DNA was amplified using 2 cock-
Stroke Registry (7, 8 ). Diagnoses were established clini- tails of biotinylated primer pairs targeting genomic frag-
cally (9 ), and all patients underwent cranial computed ments ranging from 75 to 375 bp in size. Amplified
tomography or magnetic resonance imaging. Patient data fragments within the PCR product pools were detected
were carefully documented according to a standardized colorimetrically after hybridization to sequence-specific
protocol (7 ) that included risk factors, medical history oligonucleotide probes immobilized in a linear array on
(with particular reference to vascular diseases), laboratory nylon membranes (11 ). Probe specificities had previously
investigations, technical investigations, and stroke etiol- been confirmed by sequencing or restriction length poly-
ogy and severity as measured by validated scales. Ex- morphism or allele-specific PCR analysis (10 ). All results
cluded from the study were patients with hemorrhagic were evaluated independently by 2 investigators blinded
stroke or sinus thrombosis; patients with rare causes of to phenotype. In addition to the 60 sites in Table 1 in the
stroke, e.g., arterial dissection; and patients from whom online Data Supplement, 4 very rare SNPs also contained
no written informed consent could be obtained. Reported on the gene strips were genotyped but excluded from
medication intake among the study patients included further analysis: only 3 carriers of APOB arg3500gln and
antihypertensive medication in 167/450 (37%), antidia- no carriers of CETP asp442gly, IVS14 G(⫹1)A, or (⫹3)Tins
betic medication in 44/450 (10%), lipid-lowering medica- were observed in our study population.
tion in 47/450 (10%), antiplatelet agent in 82/450 (18%),
and oral anticoagulant in 11/450 (2%). Control individu- statistical methods
als were 817 randomly selected healthy white individuals Continuous data are given as median and IQR (range
from the area of Vienna [393 female; 424 male; median age from the 25th to the 75th percentile). Discrete data are
44 (IQR 37–53) years], all voluntary participants in a given as counts and percentages. We used ␹2 tests or, if
healthcare program of the city of Vienna. All control appropriate, exact tests to compare groups of categorical
individuals were free of clinically manifest arterial vascu- data. Groups of continuous data were compared by the
lar disease and reported no arterial vascular diseases in Mann–Whitney U-test or, as appropriate, Kruskal–Wallis
1st degree relatives. Medical history, including vascular tests.
risk factors and results of laboratory investigations were Associations between the investigated polymorphisms
documented according to a standardized protocol. and the risk of stroke were analyzed by means of logistic
The study complied with the Declaration of Helsinki regression models adjusted for age (in tertiles) and sex.
and was approved by the local ethics committee. All Regression diagnostics and overall model-fit were ana-
patients and controls gave written informed consent to lyzed according to standard procedures (12 )
participate in the study. Results of the logistic regression models are presented
as odds ratio (OR) and confidence interval (CI). We used
definitions 95% CIs and defined P values of ⬍0.05 for confirmatory
Arterial hypertension was defined as either history of analysis of the SNPs in the genes for APOE (allele 4
arterial hypertension, blood pressure values above 140/90 carriers), factor V, prothrombin, methylenetetrahydrofo-
mmHg (in patients measured 1 week after the qualifying late reductase (MTHFR), and angiotensin-converting en-
event), or intake of antihypertensive medication. Presence zyme (ACE IVS16), for which metaanalyses indicated an
of diabetes mellitus was defined by fasting blood glucose association with stroke risk (6 ). For all exploratory tests
concentrations ⬎6.94 mmol/L, history of diabetes, or we calculated 99.9% CIs for the ORs and defined a P value
treatment with antidiabetic medication. Hyperlipidemia of ⬍0.001 as statistically significant to account for multi-
was defined as fasting total serum cholesterol ⬎5.18 ple testing. Calculations were performed using SPSS for
mmol/L, a history of hyperlipidemia, or intake of lipid- Windows (version 10.0, SPSS Inc). Statistical analyses
lowering medication. were performed by 2 of the authors (M.S. and P.B.).
The study was powered to detect clinically relevant
genotyping differences assuming an OR of 2 as the targeted effect. To
Citrated blood was collected by venipuncture and frozen find this effect by the 2-sided ␹2 test at the level 0.05 (in the
at ⫺20 °C within 3–5 h after sampling. DNA was ex- small group of genes or SNPs with previous evidence),
tracted using the Puregene® DNA Isolation Kit (Gentra the sample sizes of our study provide a power of ⬎80% if
System Inc). The SNPs analyzed in the study are con- the prevalence in the control group is as small as 5% (only
tained on gene strips developed by Roche Molecular the prothrombin variant is less frequent, and for ACE
Systems. Selection of SNPs on the strips by Roche was IVS16 and MTHFR 677 C⬎T even the homozygous geno-
602 Lalouschek et al.: Genetic Risk Factors of Stroke: Multilocus Assay
types are more frequent). If the prevalence in the control study participants, stratified by sex, are given in Tables 2
group is 10%, the power to detect an existing OR of 2 is and 3 in the online Data Supplement.
⬎90% (e.g., MTHFR). To reach a power of 80% for an OR Exclusion of the patients with a TIA (n ⫽ 92) did not
of 1.5 a prevalence of 19% is necessary. For prevalences alter the results (data not shown).
lower than that the study is not sufficiently powered to Being aware of the considerable difference in age
detect small effects. between patients and controls, we performed formal
When applying a significance level of 0.001 for testing testing for interaction between the age, the target poly-
the remaining 56 SNPs on the strip, a power of 80% at an morphisms, and the classification as patient or control to
OR of 2 can be achieved only for a prevalence of 11% or account for this potential confounder; no significant effect
higher (this is the case for the APOE allele 2, the heterozy- modification or change of the model fit was observed by
gous state of 45 SNPs, and the homozygous state of 14 log-likelihood ratio tests (all P ⬎0.1), suggesting that the
SNPs). For a prevalence of 15% the power already exceeds lack of association between these polymorphisms and
90%. For SNPs with prevalences ⬍11% in the heterozy- stroke was not attributable to the younger age of the
gous state the study is not sufficiently powered (in total 8 controls.
exploratory SNPs).
haplotype frequencies
haplotype reconstruction Haplotype frequencies of all genes containing at least 2
Haplotype frequencies in patients and controls were genotyped polymorphisms were estimated and are given
estimated using the Phase 2.1 software (http:// in Table 4 in the online Data Supplement. Because the SEs
archimedes.well.ox.ac.uk/pise/PHASE-simple.html; regis- of all haplotype frequencies were ⬍0.01, these were not
tration required) (13 ), which uses a Bayesian approach. Only indicated separately. We did not find significant differ-
those haplotypes with estimated frequencies ⬎1% were ences in overall haplotype frequencies between patients
further evaluated. Deviations from random distribution of and controls.
haplotypes were calculated by permutation testing within
the Phase 2.1 program. Haplotype reconstruction was per- Discussion
formed by 1 of the authors (G.E.). Epidemiological data suggest an important hereditary
component of stroke risk (1, 14 ). Therefore, the relation
Results between SNPs and ischemic stroke has been examined in
Baseline characteristics of patients and controls are shown many studies. Some of the studies and also some meta-
in Table 1. Although all patients and controls were ⱕ60 analyses indicate an association between the risk of stroke
years old there was a statistically significant age differ- and certain SNPs (15, 16 ). No large-scale study, however,
ence. In addition, patients were more often males and had has unequivocally identified a role of most of the inves-
a higher prevalence of vascular risk factors than did tigated polymorphisms as a risk factor of stroke (3, 5 ). In
controls. our investigation of 60 SNPs located in 35 genes, we did
not find significant associations of specific genotypes or
genotype frequencies haplotypes with an increased or decreased risk for isch-
Most of the investigated polymorphisms were equally emic cerebrovascular events before the age of 60 years.
frequent in patients and controls, yielding ORs close to 1 Many of the polymorphisms included in our study were
(Fig. 1; Table 2 in the online Data Supplement). None of previously reported to be associated with vascular dis-
the tested polymorphisms differed between patients and ease, but in general positive associations were observed in
controls, with statistical significance ⬍0.001 univariately small studies. In the confirmatory analyses of our study
and adjusted for age and sex. Genotype frequencies in all none of the 5 polymorphisms reported to be significantly
Table 1. Baseline characteristics of patients and controls.

Patients, n ⴝ 450 Controls, n ⴝ 817 P
Age median, IQR 52 (45–57) 44 (37–53) ⬍0.001
Female, % 168 (37) 393 (48) ⬍0.001
Hypertension, % 264 (59) 228 (28) ⬍0.001
Diabetes mellitus, % 92 (20) 13 (2) ⬍0.001
Current cigarette smoking, % 241 (54) 230 (28) ⬍0.001
Hyperlipidemia, % 320 (71) 544 (67) 0.098
Cholesterol 关median (IQR)兴, mmol/L 5.7 (4.9–6.4) 5.6 (4.9–6.3) 0.441
BMI 关median (IQR)兴 26.9 (24.2–30.1) 24.6 (22.3–27.0) ⬍0.001
BMI, body mass index. Numbers in parentheses represent percentages of the total or ranges of mean values. For percentages, the significances do not differ from
those found for absolute numbers. The same is the case for the ranges.
Fig. 1. Graphical representation of

ORs and 95% CIs or 99.9% CIs for the
investigated polymorphisms (domi-
nant and recessive model).
associated with stroke in metaanalyses reached signifi- lations (sample size calculations for the confirmatory
cance at the level 0.05. Notably, the CIs for the association analyses are given in Table 2 in the online Data Supple-
between these 5 SNPs with stroke risk found in our study ment). Previous metaanalyses showed ORs well below 2
overlap with those of published metaanalyses (6, 15, 16 ). for these SNPs. Given the limited sample size of our
Our study was underpowered to reach statistical signifi- population, we cannot exclude such small effects of these
cance for the genotype distributions found in our popu- SNPs on stroke risk.
604 Lalouschek et al.: Genetic Risk Factors of Stroke: Multilocus Assay
In the exploratory part of our study, applying a signif- duced stringent criteria for statistical significance (P
icance level of 0.001 gave a power of 80% at an OR of 2 for ⬍0.001), except for those SNPs for which previous meta-
any SNP with a prevalence of 11% or higher (this is the analyses indicated a significant association with stroke.
case for the APOE 2 allele, the heterozygous state of 45 Polymorphisms associated with stroke risk with ORs ⬍2.0
SNPs, and the homozygous state of 14 SNPs). For a or those that contribute to stroke risk only in combination
prevalence of 15% (44/55 in the heterozygous and 12/55 with external risk factors may have been missed. Our
in the homozygous state) the power exceeded 90%. For study is in accord with previously published data dem-
SNPs with lower frequencies (8 SNPs in the heterozygous onstrating at most a very modest effect of several SNPs on
state) the power was lower depending on the exposition the risk of stroke.
rates of the polymorphism in the control group. For sake Our findings underline the fact that stroke is a poly-
of completeness, however, we included in the statistical genic disorder (20 ). The effects of gene– gene as well as
calculation all results obtained in the study. The observed gene– environment interactions on the predisposition to
effect sizes were close to 1.0 for all tested SNPs, a result cerebrovascular disease are still under investigation. In-
indicating that there were no systematic differences be- deed, such interactions have recently been described (21 ).
tween patients and controls in the genetic background, In our own studies performed on the same patient pop-
i.e., through selection bias. Our study was not powered to ulation, we found an effect of the SCNN1A trp493arg
reach statistical significance for small differences between polymorphism on stroke risk in relation to age and sex
patients and controls observed in our population, e.g., for (22 ). We also observed that the effect of factor V Leiden as
the APOC(-482) polymorphism with a prevalence of car- well as the 20210 G⬎A prothrombin mutation on stroke
riers of the T-allele of 53% vs 51% (OR 1.17) a sample size risk is related to sex and smoking status (23 ). In the
of 21 308 in each population would be necessary to reach present study we did not evaluate interactions between
statistical significance at the 0.001 level; for the CETP_405 single genotypes and sex, age, or external factors.
polymorphism with a prevalence of VV carriers of 5% vs Our control population was randomly included al-
10% (OR 2.1) a sample size of 946 in each population though it was not entirely population based and was self
would be necessary. selected. Individuals were recruited in different parts of
Polymorphisms in the gene for 5-lipooxygenase acti- Vienna and during different times of the year. Through
vating protein were recently reported to be associated the information obtained regarding history of vascular
with stroke risk (17, 18 ), but because we began our study diseases, we identified and excluded as controls members
before this genetic variant was identified, analysis of this of families with cerebrovascular disease in 1st degree
SNP is not included. relatives. Although this selection process might have led
Three quarters of strokes happen after the age of 60 to a certain degree of overselection, the genetic similarities
years. Age is the most important nonmodifiable risk between patients and controls indicated that overselection
factor of stroke. In patients who suffer a stroke at a did not occur.
younger age than the majority of stroke patients, genetic Our study provides no evidence for a strong (i.e., OR
influences are probably more important than in older ⬎2) effect of the investigated SNPs on stroke risk before
patients (2, 19 ). We therefore analyzed the role of SNPs in 60 years of age in the investigated population. Our study
individuals younger than 60 years. was underpowered to reach significance for smaller ef-
A recent metaanalysis of studies of ischemic stroke fects, however, because they were also reported for some
patients, including studies with older patients, addressed SNPs in previous metaanalyses. The possibility that 1 or
several of the variations that we evaluated: factor V more of the investigated polymorphisms influences stroke
arg506gln (FVL), MTHFR 677C/T, factor II 20210G/A, risk under certain conditions is not precluded, however.
ACE insertion/deletion (I/D), apolipoprotein E, glycop- Detailed analyses of our data may reveal hitherto unde-
rotein IIIa leu33pro, endothelial nitric oxide synthase tected gene– environment interactions for 1 or more of
glu298asp, plasminogen activator inhibitor-1 4G/5G, fac- these polymorphisms.
tor VII (-323)ins10 (A1/A2), and lipoprotein lipase
asn291ser. The authors concluded that the 4 most studied In conclusion, our analysis of 60 candidate gene polymor-
factors (FVL, MTHFR 677 C/T, F 2 20210 G/A, ACE I/D) phisms showed no significant association (OR ⱖ2.0) with
were significantly associated with stroke (6 ). The effects the risk of stroke or TIA in a large population of patients
were modest, however, with ORs of 1.24 for MTHFR, 1.33 with ischemic cerebrovascular events before the age of 60
for FVL, 1.44 for prothrombin G20210A, and 1.21 for ACE years.
I/D.
The results obtained in our large sample of patients
with ischemic cerebrovascular events before the age of 60 We thank M. Janisiw, M. Reisinger, and D. Haering for
years show that none of the investigated polymorphisms excellent technical assistance. We also thank M. Grow and
reaches an OR ⬎2.0 or has a strong and clinically relevant A. Silbergleit for their expertise and efforts in developing
impact on stroke risk in this age group. To correct for the the genotyping reagents used, and we thank V. Brophy
testing of multiple genes in one investigation we intro- for constructive criticism during manuscript preparation.
The Vienna Stroke Registry (VSR) is supported by re- 9. Whisnant JP, Basford JR, Bernstein EF, Cooper ES, Dyken ML,
search grants from the Medizinisch-Wissenschaftlicher Easton JD, et al. Special report from the National Institute of
Fonds des Bürgermeisters der Bundeshauptstadt Wien Neurological Disorders and Stroke: classification of cerebrovascu-
lar diseases III. Stroke 1990;21:637–76.
(project numbers 1540, 1829, and 1970), the Jubiläums-
10. Cheng S, Grow MA, Pallaud C, Klitz W, Erlich HA, Visvikis S, et al.
fonds der Oesterreichischen Nationalbank (project num- A multilocus genotyping assay for candidate markers of cardio-
bers 6866, 7115, 8281, and 9344), and the Austrian Science vascular disease risk. Genome Res 1999;9:936 – 49.
Foundation (P13902-MED). The VSR is also supported by 11. Saiki R, Walsh P, Levenson C, Erlich H. Genetic analysis of
the Wiener Krankenanstaltenverbund. Furthermore, the amplified DNA with immobilized sequence-specific oligonucleotide
VSR is sponsored by an unrestricted educational grant probes. Proc Natl Acad Sci U S A 1989;86:6230 – 4.
from Sanofi-Synthelabo and Bristol-Myers Squibb. These 12. Hosmer D, Lemeshow S. Applied Logistic Regression. New York:
John Wiley & Sons, Inc., 1989.
sponsoring companies were involved in neither original
13. Stephens M, Smith NJ, Donnelly P. A new statistical method for
concepts and systematic review of existing trial evidence, haplotype reconstruction from population data. Am J Hum Genet
nor in the design, the choice of investigators, the control of 2001;68:978 – 89.
allocation schedule, the conduct of the trial, the collection 14. Flossmann E, Schulz UG, Rothwell PM. Systematic review of
and monitoring of the data, the analysis and interpreta- methods and results of studies of the genetic epidemiology of
tion, or the writing and approval of the report. S.C. is an ischemic stroke. Stroke 2004;35:212–27.
employee of Roche Molecular Systems, which provided 15. McCarron MO, Delong D, Alberts MJ. APOE genotype as a risk
factor for ischemic cerebrovascular disease: a meta-analysis.
genotyping reagents at no cost under a collaborative Neurology 1999;53:1308 –11.
agreement for this study. 16. Sharma P. Meta-analysis of the ACE gene in ischaemic stroke.
J Neurol Neurosurg Psychiatry 1998;64:227–30.
References 17. Helgadottir A, Manolescu A, Thorleifsson G, Gretarsdottir S,
1. Hassan A, Sham PC, Markus HS. Planning genetic studies in Jonsdottir H, Thorsteinsdottir U, et al. The gene encoding 5-lipoxy-
human stroke: sample size estimates based on family history genase activating protein confers risk of myocardial infarction and
data. Neurology 2002;58:1483– 8. stroke. Nat Genet 2004;36:233–9.
2. Schulz UG, Flossmann E, Rothwell PM. Heritability of ischemic 18. Lohmussaar E, Gschwendtner A, Mueller JC, Org T, Wichmann E,
stroke in relation to age, vascular risk factors, and subtypes of Hamann G, et al. ALOX5AP gene and the PDE4D gene in a Central
incident stroke in population-based studies. Stroke 2004;35: European population of stroke patients. Stroke 2005;36:731– 6.
819 –24. 19. Jerrard-Dunne P, Cloud G, Hassan A, Markus HS. Evaluating the
3. Hassan A, Markus HS. Genetics and ischaemic stroke. Brain genetic component of ischemic stroke subtypes: a family history
2000;123:1784 – 812. study. Stroke 2003;34:1364 –9.
4. Markus H, ed. Stroke Genetics. Oxford: Oxford University Press, 20. Carr FJ, McBride MW, Carswell HV, Graham D, Strahorn P, Clark
2003. JS, et al. Genetic aspects of stroke: human and experimental
5. Humphries SE, Morgan L. Genetic risk factors for stroke and studies. J Cereb Blood Flow Metab 2002;22:767–73.
carotid atherosclerosis: insights into pathophysiology from candi- 21. Pezzini A, Del Zotto E, Bazzoli E, Akkawi NM, Padovani A, Assanelli
date gene approaches. Lancet Neurol 2004;3:227–35. D, et al. Synergistic effect of apolipoprotein E polymorphisms and
6. Casas JP, Hingorani AD, Bautista LE, Sharma P. Meta-analysis of cigarette smoking on risk of ischemic stroke in young adults.
genetic studies in ischemic stroke: thirty-two genes involving Stroke 2004;35:438 – 42.
approximately 18,000 cases and 58,000 controls. Arch Neurol 22. Hsieh K, Lalouschek W, Schillinger M, Endler G, Reisinger M,
2004;61:1652– 61. Janisiw M, et al. Impact of alphaENaC polymorphisms on the risk
7. Vienna Stroke Registry. Vienna Stroke Study Group. The Vienna of ischemic cerebrovascular events: a multicenter case-control
Stroke Registry: objectives and methodology. Wien Klin Wochen- study. Clin Chem 2005;51:952– 6.
schr 2001;113:141–7. 23. Lalouschek W, Schillinger M, Hsieh K, Endler G, Tentschert S,
8. Lalouschek W, Lang W, Mullner M. Current strategies of secondary Lang W, et al. Matched case-control study on factor V Leiden and
prevention after a cerebrovascular event: the Vienna stroke regis- the prothrombin G20210A mutation in patients with ischemic
try. Stroke 2001;32:2860 – 6. stroke/TIA up to the age of 60 years. Stroke 2005;36:1405–9.
606 – 613 (2007) Molecular Diagnostics
and Genetics
Real-Time PCR to Identify Variola Virus or Other

Human Pathogenic Orthopox Viruses
Natale Scaramozzino,1 Audrey Ferrier-Rembert,1 Anne-laure Favier,1
Corinne Rothlisberger,1 Stéphane Richard,1 Jean-Marc Crance,1 Hermann Meyer, 2
and Daniel Garin1*
Background: Variola virus (family Poxviridae, genus smallpox and other human pathogenic orthopoxvirus
Orthopoxvirus) and the closely related cowpox, vaccinia, infections.
and monkeypox viruses can infect humans. Efforts are © 2007 American Association for Clinical Chemistry
mounting to replenish the smallpox vaccine stocks,
optimize diagnostic methods for poxviruses, and de- The WHO officially declared smallpox eradicated from
velop new antivirals against smallpox, because it is the world in 1980 (1 ). Efforts are underway, however, to
feared that variola virus might be used as a weapon of replenish the vaccine stocks and develop new drugs and
bioterrorism. diagnostic methods for this virus because of the possibil-
Methods: We developed an assay for the detection of ity that the variola virus will be used as a weapon of mass
variola virus DNA. The assay is based on TaqMan destruction. The family Poxviridae and the subfamily
chemistry targeting the 14-kD protein gene. For the 1st Chordopoxvirinae (poxviruses of vertebrates) contain 8
stage of the assay we used genus consensus primers and genera. Variola virus and the closely related cowpox virus
a mixture of 2 probes (14-kD POX and 14-kD VAR) (CPV)3, vaccinia virus (VV), and monkeypox virus (MKP)
spanning the 14-kD protein-encoding gene for detection can all infect humans and are classified in the genus
of all human pathogenic orthopoxviruses. We then Orthopoxvirus. These DNA viruses are large brick-shaped
tested positive samples with the specific orthopoxvirus- particles that vary in size (length, 220 – 450 nm; width,
specific probe 14-kD POX to identify monkeypox, cow- 140 –260 nm); the genome of variola virus consists of 186
pox, and vaccinia viruses and with the 14-kD VAR kb of linear double-stranded DNA (2 ). DNA-based tech-
probe to identify variola viruses. The assay was estab- niques have been developed for the early detection and
lished on 4 different PCR cycler platforms. It was identification of several orthopoxviruses. These methods
assessed in a study with 85 different orthopoxvirus include PCR, nucleic acid sequencing, and restriction
species and strains that included variola, camelpox, fragment length polymorphism analysis (2–9 ). Sequenc-
cowpox, monkeypox, and vaccinia viruses at concentra- ing and restriction fragment length polymorphism analy-
tions ranging from 100 ng/L to 1 ␮g/L.
sis provide high specificity, but only PCR amplification
Results: The assay detected as little as 0.05 fg of DNA,
offers high sensitivity and the needed lower limit of
corresponding to 25 copies of DNA, and enabled differen-
detection. Indeed, PCR methods allow rapid detection
tiation of variola virus from the other orthopoxviruses.
and identification of many viruses in clinical and environ-
Conclusions: This real-time PCR assay provides a rapid
mental samples (10 –12 ). Several PCR-based methods
method for the early detection and differentiation of
with gel-based analyses of PCR products have been
described for the identification of orthopoxviruses (4, 6 –
9, 13, 14 ) but they require laborious and time-consuming
1
Laboratoire de Virologie, Centre de Recherches du Service de Santé des postPCR processing. One of the most promising ap-
Armées (CRSSA) Emile Pardé, Grenoble, France.
2
Bundeswehr Institute of Microbiology, Munich, Germany.
* Address correspondence to this author at: 24 avenue des Maquis du
Grésivaudan, BP87; 38702 Grenoble, France. Fax ⫹33-476-63-69-06; e-mail
3
Daniel.Garin@wanadoo.fr. Nonstandard abbreviations: CPV, cowpox virus; VV, vaccinia virus;
Received February 14, 2006; accepted January 30, 2007. MKP, monkeypox virus; FCS, fetal calf serum; TCID 50, 50% tissue culture
Previously published online at DOI: 10.1373/clinchem.2006.068635 inefective dose; Ct, threshold cycle; LOD, limit of detection.
606
proaches for rapid and sensitive diagnosis is the real-time known gene sequences of the 14-kD protein from different
5⬘ nuclease PCR assay that includes the TaqMan assay. orthopoxvirus species (open reading frame A30L with the
We report the development of a real-time TaqMan variola virus India strain as reference according to the
assay based on the gene encoding the 14-kD protein for nomenclature). The gene sequence alignments of selected
detection of orthopoxviruses in clinical material. representative human pathogenic orthopoxviruses and
camelpox virus, the most closely related to variola virus
Materials and Methods (18 ) are shown in Fig. 1. Probes were designed in an area
viruses, clinical samples, and dna preparation that shows several bases specific for the virus group
To calibrate the method, we first used DNA isolated from vaccinia– cowpox–monkeypox viruses, differentiating
CPV (Brighton strain), VV (Copenhagen, Lister and MVA them from variola virus. Primers and probes were syn-
strains), varicella-zoster virus (VZV, LES 2770 and PER thesized by Genset Oligos. Probes contained 6-carboxy-
2762 strains), and monkeypox (Zaı̈re-96), and a synthe- fluorescein at the 5⬘-end and 6-carboxytetramethylrho-
sized variola plasmid. CPV and VV were produced on damine linked to a phosphate residue at the 3⬘-end.
Vero cells (ATCC CCL-81), which were propagated in
medium 199 glutamax (M199) with 5% heat inactivated plasmid preparation
fetal calf serum (FCS), penicillin, and streptomycin at Positive DNA controls from members of the genus Or-
37 °C in a 5% CO2 atmosphere. VZV was grown on thopoxvirus were cloned with the pGEM®-T Easy Vector
MRC-5 cells (ATCC CCL-171) in RPMI 1640 glutamax Cloning Kit (Promega Corp.) from the 14-kD protein gene
with 5% FCS, penicillin, and streptomycin at 37 °C in a 5% of vaccinia (GenBank accession no. M35027), cowpox
CO2 atmosphere. Virus titers were determined by the 50% (X75158), and monkeypox (Z99061) viruses. The respec-
tissue culture infective dose (TCID50) method (16, 17 ). tive PCR amplicons were then inserted into separate
Viral DNA was extracted from 100 ␮L of tissue culture plasmid vectors. Identity of the inserted sequence was
suspensions by use of the automated MagNAPure LC威 verified by sequencing. The vaccinia virus (Lister strain)
device with the DNA Isolation Kit I (both from Roche plasmid was named pVACV-List143, the cowpox virus
Diagnostics) according to the manufacturer’s instructions. (Brighton strain) plasmid named pCPXV-V162, and the
To obtain clinical samples, we infected mice via the monkeypox virus Zaire strain plasmid named pMPXV-
intranasal route with CPV (Brighton strain). Spleen and A29L.
lungs were homogenized with 1 mL RPMI 1640 with 4 mL A positive control for variola virus (X69198) was gen-
FCS/L, freeze-thawed 3 times, and centrifuged for 10 min erated by multisite-directed mutagenesis from the gene
at 750g. Viral DNA was extracted from supernatants as encoding the 14-kD protein of the cowpox virus (permis-
described above. The viral titer of the supernatants was sion from WHO was granted by Dr Guénaël Rodier,
determined in plaque-forming units and genomic quanti- n°S2-180-4, July 1, 2003) using the QuikChange Multisite-
fication in DNA copies per milliliter. Directed Mutagenesis Kit (Stratagene). The cowpox gene
We then evaluated the TaqMan assay using 85 differ- was partially mutated so that this recombinant positive
ent orthopoxvirus strains from both our in-house labora- control plasmid (pVARV-A30L) could be distinguished
tory strain collection and from the WHO collaborating from variola DNA by sequencing in the case of false-
center for variola and other poxviruses at the CDC positive results due to plasmid contamination of the
(Atlanta, GA). A smallpox diagnostic panel compiled by sample. The sequences of the primers for the multisite-
Inger Damon’s laboratory at the CDC included 12 differ- directed mutagenesis were 1 (5⬘-GCT CTT AGA GTT TCA
ent variola virus isolates (VAR), 13 different strains and GCG TGA TTT TCC AAC-3⬘), 2 (5⬘-CTA AAT AGA ACG
isolates of camelpox, cowpox (CPV), ectromelia, herpes, TCA TCA TTG CGT TTA CAA CAC TTT TC-3⬘), 3
monkeypox, Staphylococcus aureus, vaccinia (VV), and (5⬘-GTT GTT ACA TTA GTA ATT TTT TTT TCC AAA
varicella-zoster viruses (VZV). The concentrations of TTA GTT AGT CGT TG-3⬘), 4 (5⬘-GAG AGT TTC TTC
DNA used in the test panel ranged from 1 ng/L to 1 ␮g/L ATT ATT GTC TCC ATC GGC TTT AAC AAT TGC TTC
and included total viral and cellular DNA from cell G-3⬘), and 5 (5⬘-GAA TTG CAA GAT CAT CAT CTC CAG
lysates and crude material as well as purified viral DNA. GGA AAA GAG TTC-3⬘).
pcr primers, target sequences, and fluorogenic 5⬘ nuclease pcr assay

probes The 5⬘ nuclease PCR assay and amplification conditions
The sequences of orthopox consensus genus primers GF were optimized according to the standard protocols used
(5⬘-GCC AGA GAT ATC ATA GCC GCT C-3⬘), and GR in our laboratory by adjusting primers, probes, and MgCl2
(5⬘-CAA CGA CTA ACT AAT TTG GAA AAA AAG concentrations as well as the thermal cycling tempera-
AT-3⬘) and orthopoxvirus consensus genus probe 14-kD tures and duration. The 14-kD POX and 14-kD VAR
POX (5⬘-TTT TCC AAC CTA AAT AGA ACT TCA TCG probes were tested either in the same PCR tube or
TTG CGT T-3⬘), and specific variola virus 14-kD VAR separately in different PCR tubes in the same PCR run.
probe (5⬘-TTT TCC AAC CTA AAT AGA ACG TCA TCA The assay was assessed on 4 different PCR cycler plat-
TTG CGT T-3⬘) were designed after alignment of all forms as follows: the reactions were carried out in 30-␮L
608 Scaramozzino et al.: RT-PCR Identification of Variola Virus
Fig. 1. Sequence alignment of the

first 150 bases of the 14-kD pro-
tein gene of Variola virus (VAR,
India strain) and of other orthopox-
virus locating primers (GF and GR)
and probes (14-kD POX and 14-kD
VAR) used in the assay.
14-kD VAR is the variola-specific probe
and 14-kD POX is the orthopoxvirus
consensus genus probe. The se-
quences were obtained from EMBL/
GenBank, accession numbers: X69198
(VAR India, A30L), NC000900 (VAR Gar-
cia, A31L), M35027 (vaccinia virus,
Copenhagen strain, A27L), X75158
(cowpox virus, Brighton strain, V162),
Z99061 (monkeypox virus, Zaire A29L),
and X75156 (camelpox virus, cp-1,
144L).
total volume with the SmartCycler instrument (Cepheid) equivalent) of purified pVARV-A30L plasmid as a posi-
and in 20 ␮L with the LightCycler (Roche Diagnostics), tive control sample, as well as 1 no-template control, and
the MX 4000 (Stratagene), and the 7000 SDS (ABI Prism) uninfected cellular extracts as negative controls. Baseline
instruments. PCR reactions were performed with Light- fluorescence was given by the respective program of the
Cycler-FastStart DNA Master Hybridization Probes for different PCR cyclers. The threshold cycle (Ct) was de-
both the LightCycler and the SmartCycler instruments fined for each sample as the number of cycles necessary
and TaqMan Universal PCR Master Mix for the 7000 SDS for the fluorescence emitted by the sample to cross above
and the MX 4000 instruments. LightCycler concentrations baseline fluorescence. A sample was considered negative,
were 0.5 ␮mol/L for each primer, 0.5 ␮mol/L for each below the limit of detection (LOD) of the assay, if the
TaqMan probe, and 4 mmol/L for MgCl2. TaqMan con- emitted fluorescence was found to be below baseline on 2
centrations were 0.3 ␮mol/L for each primer and 0.2 separate occasions (i.e., the sample contained ⬍25 gene
␮mol/L for each TaqMan probe, and the MgCl2 concen- copies). For comparison, the curves were classified as
tration was already optimized by the manufacturer. For high or low according to the Fmax value, defined as the
all reactions, 5 ␮L of template DNA was added. Thermal maximum fluorescence emitted by the positive control
cycling for the LightCycler and the SmartCycler instru- after a specific PCR run. High curves were between Fmax
ments was performed as follows: 1 cycle at 95 °C for 10 and Fmax/3 and low curves were below Fmax/3 but
min, followed by 45 cycles each of 95 °C for 15 s, followed above baseline fluorescence. The LOD of the assay was
by 62 °C for 60 s. For the 7000 SDS and the MX 4000 determined on the LightCycler instrument. For the 14-kD
instruments, thermal cycling was performed as follows: 2 POX probe (orthopoxvirus consensus genus probe) the
min at 50 °C, 10 min at 95 °C, followed by 40 cycles each LOD was determined from serial dilutions of both
of 95 °C for 15 s, followed by 60 °C for 60 s. Data pCPXV-V162 and pVACV-List143. For the 14-kD VAR
acquisition and analysis were carried out with the respec- probe (variola specific) the LOD was determined from
tive platform company data analysis software: LightCy- serial dilutions of pVARV-A30L.
cler (version 3.5.3), Cepheid SmartCycler (version 1.2d),
ABI Prism 7000 SDS (version 1.0.1), and MX 4000 (version Results
3.01). limit of detection and dynamic range of the assay
All PCR reactions were performed on coded samples The LOD was 0.05 fg DNA for both pVACV-List143 and
with at least 1 well that contained 0.05 fg (25 DNA copies pCPXV-V162 plasmid, which represents ⬃25 copies of the
14-kD protein-encoding gene (Fig. 2A). The LOD was also

0.05 fg of pVARV-A30L plasmid DNA (Fig. 2B). The Figs.
show the expected linearity. The dynamic range (Fig. 2C
and 2D) was 8 orders of magnitude and represented ⬃25
to 2.5 ⫻ 108 copies for pVACV-List143 and 14-kD POX
(Fig. 2C) and pVARV-A30L with 14-kD VAR (Fig. 2D).
Similar results were obtained on the 7000 SDS instrument
(data not shown) and with other orthopoxviruses plas-
mids (e.g., MKP, CPV).
With purified DNA, the LOD with the LightCycler
instrument was 5 fg per run and 200 fg for the MX4000
instrument. With crude cell extracts, the LOD with the
Fig. 3. Detection of variola virus DNA (d) and other orthopoxviruses
LightCycler instrument was 50 fg per run and 1 pg for the
[monkeypox (a), variola (b) and cowpox (c) viruses] with the mixed
MX4000 instrument, corresponding to ⬃1 TCID50 dose. probes 14-kD POX and 14-kD VAR.
The different observed Cts depend on the different amount of viruses present in
selectivity (specificity) of the assay the sample. Baseline fluorescence and the respective Cts for each virus are
A representative curve obtained with variola, monkey- shown on the graph.
pox, cowpox and vaccinia virus with the 2 probes mixed

in the same PCR mixture is shown in Fig. 3. Each virus with the 2 probes used separately was necessary to
produced a single curve with a Ct different from the differentiate variola from other orthopoxviruses (Fig. 4
others. Nonpoxvirus samples were always negative. and Table 1). A high curve with the 14-kD VAR probe
Analysis of the curves generated by the virus samples compared to 14-kD POX was considered to indicate the
Fig. 2. LOD of the assay expressed in copy numbers (Fig. 2A and 2B).
Eight-fold serial dilutions of purified plasmid DNA (100 pg/␮L to 0.01 fg/␮L) were tested on the LightCycler instrument. (A), pVACV-List143 plasmid with the TaqMan
probe 14-kD POX; (B), pVARV-A30L plasmid with the TaqMan probe 14-kD VAR. Each curve represents the mean fluorescence value of 5 different PCR runs. Since 5
␮L of DNA was used for each concentration in each PCR run, the corresponding number of copies were 2.5 ⫻ 101, 2.5 ⫻ 102, 2.5 ⫻ 103, 2.5 ⫻ 104, 2.5 ⫻ 105,
2.5 ⫻ 106, 2.5 ⫻ 107, and 2.5 ⫻ 108.
Dynamic range of the assay (Fig. 2C and 2D).

Regression analysis was performed on the Ct against log copy numbers produced from a total of 24 samples of purified DNA plasmids with concentrations ranging from
0.01 fg/␮L (25 copies) to 100 pg/␮L (2.5 ⫻ 108 copies). (C), pVACV-List143 plasmid with the TaqMan probe, 14-kD POX. The actual and predicted values are shown
with the linearity coefficient (r ⫽ ⫺1.00; Error ⫽ 0.0719). (D), pVARV-A30L plasmid with the TaqMan probe, 14-kD VAR. The actual and predicted values are shown
with the linearity coefficient (r ⫽ ⫺1.00; Error ⫽ 0.0632).
intranasal infection with CPV, there were ⬃12 ⫻ 106 DNA

copies in the lungs, 9.5 ⫻ 102 in the spleen, and no copies
were detected in blood.
Discussion
Smallpox eradication renders the variola virus one of the
most feared weapons in biological warfare or bioterrorism
scenarios (19 –21, 22 ). It is crucial that reliable diagnostic
tests be developed for the early detection and confirma-
tion of suspected variola contamination as a first line of
defense. Apart from the variola virus, a number of other
orthopoxviruses and viruses belonging to other families
cause diseases with similar symptoms (fever and papu-
lovesicular rash). Differentiation between these viruses is
an important part of any early smallpox detection system,
so that specific measures to counteract the spread of the
disease can be put in place. Timely action depends on
early diagnosis and rapid characterization of the virus
responsible for the contamination (23 ). Several PCR-
based assays have been designed for the identification of
orthopoxviruses, but they require analysis to determine
the sizes of the PCR products from gel electrophoresis
(4, 6 –9, 13, 14 ) after restriction endonuclease digestion.
Fig. 4. Detection of variola, monkeypox, cowpox and vaccinia viruses Although these methods provide high specificity and
with the probes, 14-kD VAR (variola specific) and 14-kD POX, used allow species and sometimes strain differentiation, they
separately.
are very labor-intensive and time-consuming and thus
The variola virus is easily differentiated from the other orthopoxviruses. The
different observed Cts depends on the different amount of viruses present in may delay diagnosis. Newer approaches that use species-
each tube. (A), with 14-kD VAR (variola-specific), the variola virus produces a high specific oligonucleotide hybridization on a microchip (24 )
curve (fluorescence superior to Fmax /3), from which the Fmax is calculated,
whereas the other orthopoxviruses produce either a low curve (monkeypox virus;
and fluorescence resonance energy transfer probes on the
fluorescence inferior to Fmax /3) or a negative result (cowpox and vaccinia LightCycler instrument (25, 26 ) have also been described.
viruses; fluorescence inferior to baseline). (B), with 14-kD POX, the variola virus
produces a low curve (fluorescence inferior to Fmax/3) whereas the other
Another approach, the 5⬘ nuclease PCR or the TaqMan
orthopoxviruses produce high curves (fluorescence superior to Fmax/3). assay, first developed by Holland et al. (27 ) and improved
by Lee et al. (28 ), allows for the simultaneous amplifica-
presence of the variola virus, and a high curve obtained tion and detection of nucleic acids in real time. Some
with the 14-kD POX probe compared to 14-kD VAR the assays targeting the 14-kD protein gene (29 ) or the rpo18
presence of other human pathogenic orthopoxviruses, gene (30 ) and relying on the concept of melting temper-
such as monkeypox, cowpox or vaccinia viruses, in hu- ature analyses correctly discriminate nonvariola or-
man clinical samples. thopoxviruses from variola virus, but they can not be
The assay was evaluated on 85 samples of orthopoxvi- used on all real-time PCR platforms.
rus genomic DNA. DNA genomes of herpes and varicella- We describe a real-time PCR assay that targets the
zoster viruses, BSC40 cells, staphylococcus aureus, and 14-kD protein gene and allows easy and rapid diagnosis
human genomic DNA were not detected (Table 2). With of all human pathogenic orthopoxviruses by use of a
the 14-kD VAR probe, variola virus DNA was detected mixture of 2 probes in a first-screening test without
only with the 12 variola strains tested. The same assay modifying the performance of the test. If the sample is
specificity was obtained with the 4 tested PCR instru- positive, variola virus can then be differentiated from
ments. Using the LightCycler instrument the method also monkeypox, cowpox, and vaccinia viruses by using the
successfully detected orthopoxvirus species in lungs, same probes in separate runs with LightCycler, 7000 SDS,
spleen, and blood of mice infected with CPV: 21 days after MX 4000, and SmartCycler instruments. Although Taq-
Table 1. Keys to analysis of results and conclusion.

Curve with 14-kD POX probe Curve with 14-kD VAR probe Conclusion
Negative Negative Absence of orthopoxvirus
Positive–high Negative, or positive–low Orthopoxvirus other than variola virus
Negative or Positive low or high Positive–high Variola virus
Negative curve: curve fluorescence below baseline fluorescence. Positive curve: curve fluorescence above baseline fluorescence. Positive–low: fluorescence below
Fmax/3. Positive– high: fluorescence above Fmax/3 (see Fig. 4).
Table 2. Identification of orthopoxviruses with the 14-kD POX and 14-kD VAR TaqMan probes used separately.
Species Strains Curve with 14-kD POX Curve with 14-kD VAR Results
Variola virus India 7125 – ⫹, H VAR
7124 – ⫹, H VAR
66–39 ⫹, H ⫹, H VAR
68–258 ⫹, H ⫹, H VAR
Afghanistan Variolator 4 ⫹, L ⫹, H VAR
Heidelberg ⫹, H ⫹, H VAR
Higgins – ⫹, H VAR
Kembula – ⫹, H VAR
v74–227 Congo 9 ⫹, L ⫹, H VAR
MS Lee ⫹, L ⫹, H VAR
Brazil Garcia ⫹, H ⫹, H VAR
Shahzaman – ⫹, H VAR
Monkeypox virus v96-I-16 ⫹, H ⫹, L Orthopoxvirus No VAR
v79-I-005 ⫹, H ⫹, L Orthopoxvirus No VAR
AP1–1 ⫹, H ⫹, L Orthopoxvirus No VAR
AP-6 ⫹, H ⫹, L Orthopoxvirus No VAR
Gabon ⫹, H ⫹, L Orthopoxvirus No VAR
MSF#2 ⫹, H ⫹, L Orthopoxvirus No VAR
009/01 ⫹, H ⫹, L Orthopoxvirus No VAR
Vaccinia virus Copenhagen ⫹, H – Orthopoxvirus No VAR
Dryvax ⫹, H – Orthopoxvirus No VAR
Bern ⫹, H – Orthopoxvirus No VAR
EP Marina ⫹, H – Orthopoxvirus No VAR
BP-1 ⫹, H – Orthopoxvirus No VAR
Utrecht ⫹, H – Orthopoxvirus No VAR
CVA ⫹, H – Orthopoxvirus No VAR
Levaditi ⫹, H – Orthopoxvirus No VAR
Lister ⫹, H – Orthopoxvirus No VAR
MVA ⫹, H – Orthopoxvirus No VAR
v70-I-260 ⫹, H – Orthopoxvirus No VAR
Cowpox virus Brighton ⫹, H – Orthopoxvirus No VAR
EP-3 ⫹, H – Orthopoxvirus No VAR
Rat Moscow ⫹, H – Orthopoxvirus No VAR
Catpox 5 ⫹, H – Orthopoxvirus No VAR
OPV 89–5 ⫹, H – Orthopoxvirus No VAR
Sweden II ⫹, H – Orthopoxvirus No VAR
Biber ⫹, H – Orthopoxvirus No VAR
427 ⫹, H – Orthopoxvirus No VAR
v00-I-20 (Swedish) ⫹, H – Orthopoxvirus No VAR
Camelpox virus 2379 ⫹, H – Orthopoxvirus No VAR
Saudi ⫹, H – Orthopoxvirus No VAR
Mauritania ⫹, H – Orthopoxvirus No VAR
CP-5 ⫹, H – Orthopoxvirus No VAR
Ectromelia virus Moscow ⫹, H – Orthopoxvirus No VAR
MP-3 ⫹, H – Orthopoxvirus No VAR
Silberfuchs ⫹, H – Orthopoxvirus No VAR
Raccoonpox virus Herman ⫹, H – Orthopoxvirus No VAR
Table 2. Continued
Species Strains Curve with 14-kD POX Curve with 14-kD VAR Results
Varicella-zoster virus Webster – – Absence of Orthopoxvirus
LES 2770 – – Absence of Orthopoxvirus
PER 2762 – – Absence of Orthopoxvirus
Herpes simplex virus 1 HFEM – – Absence of Orthopoxvirus
Human genomic DNA supT – – Absence of Orthopoxvirus
BSC40 cells – – Absence of Orthopoxvirus
Staphylococcus aureus 2 – – Absence of Orthopoxvirus
Water NAa – – Absence of Orthopoxvirus
Results were reported after comparing the respective amplification curves of the samples listed in this table with each probe (see Table 1 for keys to results analysis
and Fig. 4 for representative curves).
a
NA, not applicable; –, negative sample; ⫹, H: Positive sample, high curve; ⫹, L: Positive sample, low curve; VAR: variola virus
Man technology cannot generate melting curves to enable none of the nonorthopoxviruses, including varicella-zos-
comparison of the 2 amplification product reactions, sim- ter and herpes viruses as well as Staphylococcus aureus.
ilar relevant information can be easily obtained by the The assays carried out on the SmartCycler instrument
simple comparison of the amplification curves obtained and on the 7000SDS instrument showed comparable sen-
with each of the 2 probes. In the gene area targeted by sitivity and specificity. These assays showed lower sensi-
both probes, we observed 3 mismatches (Fig. 1). Guanine tivity with the MX 4000 instrument, which could be
at position 62 is specific for variola virus and adenine at explained by the fact that the initial development of the
position 68 is observed in the majority of variola strains. assay was standardized on the LightCycler and the 7000
When these 2 bases are present at their respective posi- SDS instruments without specific adaptation for the
tions, the variola-specific 14-kD VAR probe is able to MX4000 instrument. It is possible that the sensitivity of
hybridize with a high efficiency: the fluorescence intensity the assay on this PCR platform could be improved with
is higher than that observed with the 14-kD POX probe. time, as could that of the SmartCycler, for which the assay
However, if only 1 specific base is present (as in the has not been fully evaluated for sensitivity.
Alastrim variola strain, Garcia), then both probes are able
to hybridize the DNA template. On the other hand, when In conclusion, we describe a consensus orthopoxvirus
thymidine at position 62 and guanine at position 68 are PCR assay with 2 built-in steps that allow the 2nd stage of
targeted specifically by the orthopoxvirus consensus the assay to differentiate variola virus from other human
probe 14-kD POX, it is possible to conclude that the pathogenic orthopoxviruses with only 2 probes. Positive
sample is from another orthopoxvirus. Guanine at posi- control plasmids have a different gene sequence from the
tion 59 is specific for camelpox virus, and this mismatch variola virus gene sequence and can be provided to any
could explain the lack of sensitivity obtained with the hospital or research institution. In contrast to fluorescence
14-kD POX probe with this virus. This was not considered resonance energy transfer technology, the TaqMan tech-
to be a problem because the objective of the test was the nology can be used on all real-time PCR platforms now
detection of orthopoxviruses in human clinical samples, available. As part of a prompt defense against bioterror-
thus making the camelpox virus, which is not pathogenic ism, the assay we describe must be followed by a more
in humans, less relevant for the purpose of the assay. detailed analysis with other specific PCR smallpox re-
However, the method perfectly detected and differenti- agent sets (26, 31 ) to customize vaccination and treatment
ated variola from monkeypox virus (endemic in Africa), to the contaminant virus, because it is possible that a
from cowpox virus (endemic in Europe but responsible bioattack could be carried out not only with wild-type
for only limited cutaneous lesions in immunocompetent virus but also with genetically modified pathogenic or-
people), and also from vaccinia virus, thus helping to thopoxviruses (32 ).
ascertain the diagnosis of complications of a smallpox
vaccination in clinical practice (e.g., eczema vaccinatum).
The probes were also able to detect ectromelia virus but Grant/funding support: This work was supported by
with a lower sensitivity (data not shown). The detection grants from the Service de Santé des Armées (SSA), the
limit assessed with DNA plasmids was 0.05 fg, corre- Délégation Générale pour l’Armement (DGA) and asso-
sponding to ⬃25 gene copies for each probe. The 2 probes ciation ARAMI.
perfectly detected the orthopoxviruses in all 99 samples of Financial disclosures: none declared.
the WHO repository panel that included variola virus- Acknowledgements: We thank the CDC Poxvirus Pro-
infected cells and tissues, purified variola virus DNA, and gram, and specifically acknowledge Inger Damon,
the DNA of 13 different strains and isolates of camelpox, Miriam Laker, and Joanne Patton for design of the exper-
cowpox, ectromelia, monkeypox, vaccinia viruses, and iment, preparation of the reagents, and facilitation of the
variola nucleic acid testing. This work was done in 16. Crance JM, Gratier D, Guimet J, Jouan A. Inhibition of sandfly fever
conjunction with the Global Health Security Action Sicilian virus (Phlebovirus) replication in vitro by antiviral com-
Group’s laboratory exercise. We also thank Danielle Gra- pounds. Res Virol 1997;148:353– 65.
17. Reed LJ, Muench H. A simple method of estimating fifty percent
tier, Josette Guimet, Henri Blancquaert, and Mary O’Brien
end points. Amer J Hyg 1938;27:493–7.
for their technical assistance and Bernard Souberbielle for 18. Gubser C, Smith GL. The sequence of camelpox virus shows it is
critically reviewing the manuscript. most closely related to variola virus, the cause of smallpox. J Gen
Virol 2002;83:855–72.
References 19. Berche P. The threat of smallpox and bioterrorism. Trends Micro-
1. Henderson DA. The history of smallpox eradication. Henry E biol 2001;9:15– 8.
Sigerist Suppl Bull Hist Med 1980;:99 –114. 20. Henderson DA, Inglesby TV, Bartlett JG, Ascher MS, Eitzen E,
2. Massung RF, Liu LI, Qi J, Knight JC, Yuran TE, Kerlavage AR, et al. Jahrling PB, et al. Smallpox as a biological weapon: medical and
Analysis of the complete genome of smallpox variola major virus public health management. Working Group on Civilian Biodefense.
strain Bangladesh-1975. Virology 1994;201:215– 40. JAMA 1999;281:2127–37.
21. O’Toole T. Smallpox: an attack scenario. Emerg Infect Dis 1999;
3. Esposito JJ, Knight JC. Orthopoxvirus DNA: a comparison of
5:540 – 6.
restriction profiles and maps. Virology 1985;143:230 –51.
22. Breman JG, Henderson DA. Diagnosis and management of small-
4. Loparev VN, Massung RF, Esposito JJ, Meyer H. Detection and
pox. N Engl J Med 2002;346:1300 – 8.
differentiation of old world orthopoxviruses: restriction fragment
23. Shchelkunov SN, Safronov PF, Totmenin AV, Petrov NA,
length polymorphism of the crmB gene region. J Clin Microbiol
Ryazankina OI, Gutorov VV, et al. The genomic sequence analysis
2001;39:94 –100.
of the left and right species-specific terminal region of a cowpox
5. Mackett M, Archard LC. Conservation and variation in Orthopoxvi- virus strain reveals unique sequences and a cluster of intact ORFs
rus genome structure. J Gen Virol 1979;45:683–701. for immunomodulatory and host range proteins. Virology 1998;
6. Meyer H, Neubauer H, Pfeffer M. Amplification of ‘variola virus- 243:432– 60.
specific’ sequences in German cowpox virus isolates. J Vet Med B 24. Laasri M, Chizhikov V, Mikheev M, Shchelkunov S, Chumakov K.
Infect Dis Vet Public Health 2002;49:17–9. Detection and discrimination of orthopoxviruses using microarrays
7. Neubauer H, Pfeffer M, Meyer H. Specific detection of mousepox of immobilized oligonucleotides. J Virol Methods 2003;112:67–
virus by polymerase chain reaction. Lab Anim 1997;31:201–5. 78.
8. Neubauer H, Reischl U, Ropp S, Esposito JJ, Wolf H, Meyer H. 25. Espy MJ, Cockerill IF, Meyer RF, Bowen MD, Poland GA, Hadfield
Specific detection of monkeypox virus by polymerase chain reac- TL, et al. Detection of smallpox virus DNA by LightCycler PCR.
tion. J Virol Methods 1998;74:201–7. J Clin Microbiol 2002;40:1985– 8.
9. Ropp SL, Jin Q, Knight JC, Massung RF, Esposito JJ. PCR strategy 26. Ibrahim MS, Kulesh DA, Saleh SS, Damon IK, Esposito JJ,
for identification and differentiation of small pox and other or- Schmaljohn AL, et al. Real-time PCR assay to detect smallpox
thopoxviruses. J Clin Microbiol 1995;33:2069 –76. virus. J Clin Microbiol 2003;41:3835–9.
10. Garcia S, Crance JM, Billecocq A, Peinnequin A, Jouan A, Bouloy 27. Holland PM, Abramson RD, Watson R, Gelfand DF. Detection of
M, et al. Quantitative real-time PCR detection of Rift Valley fever specific polymerase chain reaction product by utilizing the 5⬘-3⬘
virus and its application to evaluation of antiviral compounds. exonuclease activity of thermus acquaticus DNA polymerase. Proc
J Clin Microbiol 2001;39:4456 – 61. Natl Acad Sci U S A 1991;88:7276 – 80.
28. Lee LG, Connell CR, Bloch W. Allelic discrimination by nick-
11. Garin D, Peyrefitte C, Crance JM, Le Faou A, Jouan A, Bouloy M.
translation PCR with fluoregenic probes. Nucleic Acids Res 1993;
Highly sensitive Taqman PCR detection of Puumala hantavirus.
21:3761– 6.
Microbes Infect 2001;3:739 – 45.
29. Olson VA, Laue T, Laker MT, Babkin IV, Drosten C, Shchelkunov
12. Scaramozzino N, Crance JM, Jouan A, DeBriel DA, Stoll F, Garin D. SN, et al. Real-time PCR system for detection of orthopoxviruses
Comparison of flavivirus universal primer pairs and development and simultaneous identification of smallpox virus. J Clin Microbiol
of a rapid, highly sensitive heminested reverse transcription-PCR 2004;42(5):1940 – 6.
assay for detection of flaviviruses targeted to a conserved region 30. Nitsche A, Ellerbrok H, Pauli G. Detection of orthopoxvirus DNA by
of the NS5 gene sequences. J Clin Microbiol 2001;39:1922–7. real-time PCR and identification of variola virus DNA by melting
13. Meyer H, Pfeffer M, Rziha HJ. Sequence alterations within and analysis. J Clin Microbiol 2004;42:1207–13.
downstream of the A-type inclusion protein genes allow differen- 31. Kulesh DA, Baker RO, Loveless BM, Norwood D, Zwiers SH,
tiation of Orthopoxvirus species by polymerase chain reaction. Mucker E, et al. Smallpox and pan-orthopox virus detection by
J Gen Virol 1994;75:1975– 81. real-time 3⬘-minor groove binder TaqMan assays on the Roche
14. Mukinda VB, Mwema G, Kilundu M, Heymann DL, Khan AS, LightCycler and the Cepheid Smart Cycler Platforms. J Clin Micro-
Esposito JJ. Re-emergence of human monkeypox in Zaire in 1996. biol 2004;42(2):601–9.
Monkeypox Epidemiologic Working Group. Lancet 1997;349: 32. Jackson RJ, Ramsay AJ, Christensen CD, Beaton S, Hall DF,
1449 –50. Ramshaw IA. Expression of mouse interleukin-4 by a recombinant
15. Rodriguez JF, Smith GL. IPTG-dependent vaccinia virus: identifica- ectromelia virus suppresses cytolytic lymphocyte responses and
tion of a virus protein enabling virion envelopment by Golgi overcomes genetic resistance to mousepox. J Virol 2001;75:
membrane and egress. Nucleic Acids Res 1990;18:5347–51. 1205–10.
614 – 619 (2007) Hemostasis and
Thrombosis
Determination of Aspirin Responsiveness by Use

of Whole Blood Platelet Aggregometry
Boris T. Ivandic,1†* Evangelos Giannitsis,1† Philipp Schlick,1 Peter Staritz,1
Hugo A. Katus,1 and Thomas Hohlfeld2
Background: Insufficient platelet inhibition is associ- Conclusions: Impedance aggregometry may prove use-
ated with an increased cardiovascular risk in up to 30% ful to study aspirin responsiveness, and incubation with
of patients taking regular doses of aspirin. We describe ASA may help to identify nonresponders and classify
an assay to study aspirin responsiveness. resistance.
Methods: We performed impedance aggregometry on © 2007 American Association for Clinical Chemistry
diluted whole blood with 1 mg/L collagen and 0.5
mmol/L arachidonic acid (AA). We measured thrombox- Platelet activation and aggregation play critical roles in
ane B2 (TXB2) by RIA. We examined 66 healthy control cardiovascular disease, triggering acute coronary syn-
individuals, 144 aspirin users with stable coronary ar- drome (ACS)3 and stroke. Aspirin (acetylsalicylic acid,
tery disease (CAD), and 245 CAD patients treated with ASA) has become a cornerstone of the therapy of cardio-
aspirin and clopidogrel. Nonresponsive samples were vascular disease, reducing the rate of adverse clinical
incubated with excess DL-lysinmonoacetylsalicylic acid. events in high-risk patients by ⬃25% (1 ). ASA irreversibly
Results: Assay imprecision (CV) was 9.8% and 8.2% at inhibits cyclooxygenase-1 (COX-1) activity in circulating
mean (SD) 6-min impedance of 13.7 (2.8) ⍀ and 13.6 (2.3) platelets, and as a consequence, reduces the synthesis of
⍀ for collagen and AA, respectively. Collagen induced thromboxane A2 (TXA2), a potent platelet activator (2, 3 ).
stronger aggregation (P ⴝ 0.0199) in women [n ⴝ 28, 14.6 Measurements of platelet activation and aggregation as
(2.4) ⍀] than in men [n ⴝ 38, 13.1 (2.9) ⍀], even after well as of urinary TXA2 metabolite concentrations have
sample incubation with 0.1 mmol/L acetylsalicylic acid revealed, however, that TXA2 synthesis may not be inhib-
(ASA) or 1 ␮mol/L terbogrel, a combined inhibitor of ited sufficiently in all ASA users. These ASA nonre-
thromboxane synthase and receptors. The sex associa- sponders drew attention to a phenomenon, referred to as
tion persisted in aspirin users, but not if clopidogrel was ASA resistance, that occurs in 5% to 40% of all treated
also taken. A 6-min impedance >8 ⍀ with collagen
patients (3, 4 ). Increased platelet turnover, drug interac-
(mean ⴚ 2 SD of the controls) was taken as evidence of
tions, increased activity of other COX isoforms, genetic
nonresponsiveness, particularly if incubation with ASA
polymorphisms, and platelet hyperreactivity may all ex-
did not inhibit aggregation further (>2 ⍀). Compared
plain ASA resistance to some extent (3, 5 ). ASA resistance
with AA, collagen identified more nonresponsive sam-
is associated with a 2- to 4-fold increased risk of recurrent
ples among aspirin users (15%) and CAD patients who
cardiovascular events (6 – 8 ) and is an independent risk
also received clopidogrel (10%). Incubation with ASA
factor in patients presenting with ACS (9 –11 ).
improved inhibition of aggregation in 70% of samples and
We describe a simple assay to study ASA responsive-
consistently reduced TXB2 formation during aggregation.
ness in diluted whole blood by use of impedance aggre-
gometry. We incubated nonresponsive samples with ex-
cess dl-lysinmonoacetylsalicylic acid to differentiate
1
Department of Medicine III, University of Heidelberg, Germany. pharmacokinetic and pharmacodynamic types of ASA
2
Institute of Pharmacology and Clinical Pharmacology, University of resistance.
Duesseldorf, Germany.
†
These authors have contributed equally to this study.
* Address correspondence to this author at: Innere Medizin, Abt. III,
Universitätsklinikum Heidelberg, Otto-Meyerhof-Zentrum, Im Neuenheimer
Feld 350, 69120 Heidelberg, Germany. Fax ⫹49-6221-56-4866; e-mail Boris. 3
Nonstandard abbreviations: ACS, acute coronary syndrome; ASA, ace-
Ivandic@med.uni-heidelberg.de tylsalicylic acid; COX-1, cyclooxygenase-1; TXA2, thromboxane A2; CAD,
Received September 30, 2006; accepted January 19, 2007. coronary artery disease; PCI, percutaneous coronary intervention; AA, arachi-
Previously published online at DOI: 10.1373/clinchem.2006.081059 donic acid; and TXB2, thromboxane B2.
614
patients (180 men and 65 women; ages 29 to 90 years,

median 68 years) who underwent percutaneous coronary
intervention (PCI). The use of glycoprotein IIb-IIIa inhib-
itors was an exclusion criterion. PCI patients received 500
mg intravenous ASA and an oral loading dose of 0 to 600
mg clopidogrel 12 to 24 h before blood sampling. Therapy
was continued with 100 mg ASA and 75 mg clopidogrel,
both orally, once a day. We collected whole blood with a
butterfly cannula (21-gauge) into 10-mL plastic syringes
containing 1 mL of 0.106 mol/L trisodium citrate (Citrate
Monovette; Sarstedt).
aggregometry
Electrical aggregometry measures the impedance between
a pair of electrodes immersed in diluted whole blood. The
increase in impedance (⍀) is associated with the amount
of platelet aggregates deposited on the electrodes after the
addition of a platelet agonist (12 ). Details of the aggre-
gometry method have been reported (13 ). We started
aggregation by addition of collagen or arachidonic acid
(AA), final concentrations 1 mg/L and 0.5 mmol/L,
respectively. AA was provided as dried lipid and was
dissolved with a bovine serum albumin buffer supplied
by the manufacturer (Chrono-Log). We briefly flushed
50-␮L aliquots of solubilized AA with nitrogen and stored
them at ⫺20 °C until use. We incubated aliquots of
citrated whole blood with 0.1 mmol/L ASA (lysine salt;
Bayer) at room temperature for 20 to 30 min and mea-
sured aggregation. We incubated additional aliquots with
1 ␮mol/L terbogrel (Thomae), an equipotent inhibitor of
thromboxane synthase and thromboxane receptors.
measurement of thromboxane b2
We measured concentrations of TXB2, the stable hydroly-
sis product of TXA2, before and after incubation with
ASA. We first analyzed diluted whole blood samples by
use of collagen-induced aggregometry. We then retrieved
the samples from the aggregometer and centrifuged them
Fig. 1. Representative results of impedance aggregometry in a diluted immediately to obtain diluted plasma, which was stored
whole blood sample from a healthy participant before (upper curve) at ⫺20 °C until measurement of TXB2 by use of a previ-
and after (lower curve) incubation with ASA. ously described RIA (14 ). Briefly, to prepare calibrators,
The addition of collagen (A) and AA (B) is indicated by an arrow. The 6-min
impedance results of aggregation (dashed line) are given for each aggregation
we removed TXB2 from control plasma by reversed-phase
curve. C18 column (Bakerbond 7025– 00) and added TXB2 (Cal-
biochem) to the plasma. We mixed 0.3-mL aliquots of
Materials and Methods appropriately diluted (10- to 1000-fold) samples and stan-
patients dard calibrators with 0.05 mL antiserum (polyclonal,
This study was approved by the Institutional Review rabbit) and 0.5 mL PBS [137 mmol/L NaCl, 2.7 mmol/L
Board in accordance with the Declaration of Helsinki. We KCl, 10 mmol/L Na2HPO4O (dibasic, anhydrons), 2
obtained blood samples from 3 groups: healthy volun- mmol/L KH2PO4O (monobasic, anhydrons) in H2O; ad-
teers and blood donors (38 men and 28 women; ages 16 to justed to pH 7.4 by HCl] containing 2 g/L gelatin and 40
61 years, median 29 years) who denied taking any anti- ␮g/L [3H]TXB2. After 12-h incubation, we extracted un-
platelet medication within 10 days before sampling; 144 bound tracer with activated charcoal and measured
outpatients with stable coronary artery disease (CAD) bound radioactivity by scintillation counting. Cross-reac-
(114 men and 30 women; ages 28 to 90 years, median 66 tivity of the assay with numerous prostaglandin deriva-
years) who took 100 mg ASA once a day; and 245 CAD tives was ⬍3%.
616 Ivandic et al.: Aspirin Resistance and Whole Blood Aggregation
Fig. 2. Box plots of 6-min impedance results of collagen-

induced aggregometry in diluted whole blood samples
from healthy blood donors (A) and CAD patients on ASA
(B) and ASA and clopidogrel (C).
Samples were analyzed before and after incubation with ASA or
terbogrel. Impedance results of samples from men and women
are illustrated by filled and open boxes, respectively.
statistical analysis men and 6 women; ages 22 to 58 years, median 34 years)

Statistical significance was calculated by use of the Mann– before and after incubation with ASA and terbogrel (Fig.
Whitney U-test and Fisher exact test for continuous nu- 2). We also examined CAD patients who exhibited plate-
merical and categorical data, respectively, allowing an let aggregation within the reference interval in spite of
␣-error ⬍0.05. All tests were calculated by use of the treatment with ASA alone (11 men and 5 women; ages 52
StatView 5.0.1 software package (SAS Institute) to 83 years, median 67 years) or ASA and clopidogrel (23
men and 5 women; ages 39 to 91 years, median 65 years).
Results Incubation with ASA reduced 6-min impedance consid-
Impedance increased upon addition of the platelet ago- erably (⬎2 ⍀) in all groups. Terbogrel caused even more
nists collagen or AA to a diluted whole blood sample inhibition, suggesting incomplete COX-1 inhibition by
from a healthy individual (Fig. 1). If citrate-anticoagulated ASA or COX-1–independent TXA2 formation. Notably,
blood was first incubated with ASA, then aggregation the sex-related difference found in the healthy cohort
was markedly reduced with collagen and suppressed persisted after incubation with ASA and terbogrel (Fig.
with AA (6-min impedance ⬍3 ⍀) (Fig. 1). We measured 2A). This difference was also present in CAD patients
samples from 8 participants 4 to 5 times during a 4-h treated with ASA (Fig. 2B).
period after sampling to determine the imprecision of the Next, we examined whole blood aggregation in 144
assay (CV). The mean CVs were 9.8% and 8.2% for outpatients with stable CAD who were treated with 100 mg
collagen and AA, respectively. The mean (SD) 6-min of ASA daily. If collagen was used as a platelet agonist,
impedance of the healthy cohort was 13.7 (2.8) ⍀ for mean (SD) of 6-min impedance was 8.7 (4.6) ⍀ in all patients
aggregometry with collagen. The reference interval, de- and 8.4 (4.6) ⍀ and 10.0 (4.3) ⍀ in male and female patients,
rived from the mean and 2 SD, was 8.1 to 19.3 . Conse- respectively. The distribution of impedance results in men
quently, an ASA user was considered a potential nonre- differed from the relatively symmetric distribution in
sponder if his or her sample exhibited a 6-min impedance women by a small additional peak at around 2 to 4 ⍀ (Fig.
⬎8 ⍀. Mean (SD) 6-min impedance was higher in women 3). This peak may have accounted for the sex-related differ-
[14.6 (2.4) ⍀] than in men [13.1 (2.9) ⍀]. This difference ence, which was not significant (P ⫽ 0.0660). If we used AA
was significant (P ⫽ 0.0199), but we found no correlation as a platelet agonist, aggregation was completely sup-
between age and impedance results. We examined AA- pressed (⬍3 ⍀) in 126 (88%) patients. Regression plots of
induced aggregation in only 8 men and 13 women of the 6-min impedance results obtained with AA and collagen
healthy cohort. The mean (SD) was 13.6 (2.3) ⍀, and there showed 2 different groups of nonresponders (Fig. 4A): 1
was no sex- or age-related difference. group was characterized by increased impedance with col-
The main focus of this work was to examine the lagen and suppressed aggregation with AA; in the other
suitability of collagen-induced whole blood aggregation group, impedance results of AA and collagen appeared to
to study the pharmacologic effects of ASA. To determine correlate. Thus, collagen-induced aggregometry identified
the extent to which COX-1 inhibition and TXA2 receptor more nonresponders than AA-induced aggregometry. Any
blockade inhibit whole blood aggregation, we analyzed sample exhibiting a 6-min impedance ⬎8 ⍀ was analyzed
diluted whole blood samples from 18 blood donors (12 again after incubation with ASA. A measurable reduction
Fig. 3. Distributions of 6-min impedance results of collagen-induced

aggregometry in diluted whole blood samples from 114 male (A) and
30 female (B) CAD patients treated with ASA. Fig. 4. Results of AA- and collagen-induced aggregometry in CAD
patients on ASA (A) and ASA and clopidogrel (B) therapy.
Results of aggregometry correlated linearly in a subgroup of samples (filled
(⬎2 ⍀) was achieved in 40 of 56 men and 5 of 19 women, circles).
suggesting that the full inhibitory potential of ASA was not

achieved in about one third of the male patients and three
quarters of the female patients (pharmacokinetic type of whereas 7 samples exhibited a pharmacodynamic type of
resistance). Conversely, 21 CAD patients (15%, 16 men and resistance. ASA resistance was independent of age, diag-
5 women) maintained an aggregation with collagen ⬎8 ⍀ nosis (stable angina, unstable angina, non–ST-segment
despite incubation with ASA (pharmacodynamic type of elevation myocardial infarction, or ST-segment elevation
resistance). myocardial infarction), and clopidogrel loading dose, but
Even in the presence of double antiplatelet therapy, it was associated with female sex (P ⫽ 0.0301). Other factors
may be desirable to test for ASA resistance. We examined previously implicated in ASA resistance showed no sig-
whole blood aggregation in 245 CAD patients who re- nificant association: a familial history of myocardial in-
ceived ASA and clopidogrel. This double antiplatelet farction, ACS while on ASA therapy, diabetes mellitus, or
therapy inhibited collagen-mediated aggregation more concomitant therapy with angiotensin converting enzyme
than ASA alone. Impedance results ranged from 0 to 18 ⍀ inhibitors, angiotensin II receptor blockers, or statins.
and exhibited a skewed distribution [skewness 1.7; me- We hypothesized that the formation of TXB2 during
dian 2 ⍀; mean (SD) 3.4 (3.6) ⍀]. If we used AA as a whole blood aggregometry correlated with ASA respon-
platelet agonist, aggregation was suppressed in 218 of 225 siveness. Therefore, we examined TXB2 concentrations in
patients tested (97%). Again, the nonresponders detected samples from 155 of 245 CAD patients on double anti-
by both agonists were only partly identical (Fig. 4B). In platelet therapy (Fig. 5). TXB2 concentrations were signif-
collagen-induced aggregometry, 24 of 245 samples (10%) icantly higher in nonresponsive samples (n ⫽ 15, 1 patient
exhibited a 6-min impedance ⬎8 ⍀ and were considered with pharmacodynamic resistance) than in responsive
nonresponsive. Incubation with ASA reduced impedance samples (n ⫽ 140) before (P ⬍0.0001) and also after (P ⫽
to ⬍8 ⍀ in 17 samples (pharmacokinetic resistance), 0.0142) incubation with ASA.
618 Ivandic et al.: Aspirin Resistance and Whole Blood Aggregation
aggregation, when a patient is already taking ASA. This

assay is potentially suitable for point-of-care use by nurs-
ing staff in coronary care units or catheterization labora-
tories. In contrast to light transmission aggregometry, this
assay does not involve centrifugation steps. Thus, the
turnaround time for detecting potential nonresponders is
10 min and is extended by another 30 min only if a
nonresponsive sample requires further characterization
by in vitro incubation.
Incubation with ASA or terbogrel revealed that colla-
gen-mediated aggregation depended largely, but not ex-
clusively, on TXA2 synthesis by COX-1 activity. Other
commercially available COX-independent inhibitors
(such as BM 567) may be used instead of terbogrel.
Interestingly, inhibition of COX-1— by ASA either in vitro
or in vivo—revealed that collagen-induced aggregation
was stronger in women than in men, although this
difference was not statistically significant in the small
number of studied samples. The difference was not de-
pendent on TXA2 and disappeared if patients were also
treated with clopidogrel. This sex difference may have
contributed to the failure of ASA to lower the risk of
myocardial infarction or cardiovascular death in nearly
40 000 women in the Women’s Health Study (26 ).
Many authors have employed AA instead of collagen
Fig. 5. Box plots of TXB2 measurements in 155 CAD patients on ASA as an agonist for aggregometry. Although AA is a sub-
and clopidogrel.
strate of COX-1, significant aggregation with AA in ASA
Responders (n ⫽ 140) and nonresponders (n ⫽ 15) are depicted by white and
gray bars, respectively. users does not specifically indicate a failure of ASA to
inhibit COX-1 activity, because TXA2 may have other
origins (5 ). The use of collagen as an alternative platelet
Discussion agonist may offer certain advantages over AA. Prepara-
Widespread use of antiplatelet drugs has created a de- tion, handling, storage, and costs favor collagen. In the
mand for simple assays to determine their pharmacolog- presence of low concentrations of collagen, aggregation is
ical effects. Many investigators have suggested arbitrary sensitive to inhibition by ASA (27 ). In whole blood,
cutoffs for nonresponsiveness, which are usually derived collagen stimulates TXA2 synthesis from blood cell mem-
from the difference in maximal light transmittance of brane– derived AA in addition to binding to known
optical platelet aggregometry measured before and after collagen receptors, including the glycoprotein GpVI/Fc
the start of ASA therapy. Optical aggregometry requires receptor ␥-chain complex. Collagen receptor-dependent
expert personnel and time-consuming centrifugation platelet activation may be influenced by sex and may also
steps to obtain platelet-rich and -poor plasma. Platelet- explain the differences in aggregation response obtained
rich plasma is an artificial milieu deficient in giant platelet with collagen and arachidonic acid. This hypothesis
subspecies as well as erythrocytes and leukocytes, which awaits further investigation by the use of more specific
are regarded as critical modulators of platelet function in receptor ligands such as collagen-related peptides. Our
vivo (15 ). Erythrocytes may increase the release of AA preferred approach to use collagen as a platelet agonist
and the formation of activating eicosanoids (16 –18 ). Tech- integrates all relevant sources of TXA2 formation and may
nical difficulties and the requirement to perform aggre- also detect TXA2-independent platelet hyperreactivity, an
gometry twice— before and after initiation of ASA thera- important cause of ASA failure (28 ).
py—may have prevented the widespread clinical use of A limitation of whole blood aggregometry was the
optical aggregometry. specific assessment of ASA responsiveness in patients
Alternatively, we explored the use of impedance ag- with double antiplatelet therapy. This was a particular
gregometry, because it reportedly offers a higher sensitiv- challenge, because clopidogrel also affected collagen-in-
ity for antiplatelet drug effects and platelet hyperactivity duced aggregation to some extent. However, the results of
than optical aggregometry (19 ). Since its introduction by TXB2 analysis suggested a relatively high specificity of
Cardinal and Flower in 1980 (12 ), impedance aggregom- collagen-induced aggregometry. We defined nonrespon-
etry has been thoroughly evaluated and improved (20 – siveness as a 6-min impedance ⬎8 ⍀ (mean ⫺ 2 SD of the
25 ). We describe here a simple assay to study ASA healthy cohort) for patients with single or double anti-
responsiveness by a single measurement of whole blood platelet therapy. We understand, however, that a classifi-
cation based on results of impedance aggregometry in 11. Alexander JH, Harrington RA, Tuttle RH, Berdan LG, Lincoff AM,
ASA-naı̈ve individuals may be inappropriate for ASA users. Deckers JW, et al. Prior aspirin use predicts worse outcomes in
ROC analysis of whole blood aggregometry and clinical patients with non-ST-elevation acute coronary syndromes. PURSUIT
Investigators. Platelet IIb/IIIa in Unstable angina: Receptor Suppres-
endpoints is required to define appropriate response thresh-
sion Using Integrilin Therapy. Am J Cardiol 1999;83:1147–51.
olds after examination of various larger patient cohorts on 12. Cardinal DC, Flower RJ. The electronic aggregometer: a novel
single or double antiplatelet therapy. Until future studies device for assessing platelet behaviour in blood. J Pharmacol
can provide these data, we suggest incubating samples from Methods 1980;3:1350 – 8.
potentially nonresponsive patients with 0.1 mmol/L ASA to 13. Ivandic BT, Schlick P, Staritz P, Kurz K, Katus HA, Giannitsis E.
determine the maximal inhibition that can theoretically be Determination of clopidogrel resistance by whole blood platelet
achieved. For comparison, 2 min after intravenous adminis- aggregometry and inhibitors of the P2Y12 receptor. Clin Chem
2006;52:383– 8.
tration of a 500-mg dose of ASA, the plasma peak concen- 14. Schror K, Seidel H. Blood-vessel wall arachidonate metabolism and
tration reaches 0.15 mmol/L (drug prescription informa- its pharmacological modification in a new in vitro assay system.
tion). The issue of maximal inhibition is critical, because Naunyn Schmiedebergs Arch Pharmacol 1988;337:177– 82.
ASA requires a ⬎95% inhibition of its pharmacological 15. Bouchard BA, Tracy PB. Platelets, leukocytes, and coagulation
target, COX-1, for optimal efficacy. In addition, incubation [Review]. Curr Opin Hematology 2001;8:263.
with ASA was also helpful to differentiate between pharma- 16. Santos MT, Valles J, Marcus AJ, Safier LB, Broekman MJ, Islam N, et
al. Enhancement of platelet reactivity and modulation of eicosanoid
cokinetic and pharmacodynamic types of resistance. This
production by intact erythrocytes. A new approach to platelet activa-
may be useful for deciding whether a nonresponder should tion and recruitment. J Clin Invest 1991;87:571– 80.
increase the dose of ASA (pharmacokinetic resistance) or 17. Valles J, Santos MT, Aznar J, Marcus AJ, Martinez-Sales V,
rather switch to an alternative antiplatelet drug (pharmaco- Portoles M, et al. Erythrocytes metabolically enhance collagen-
dynamic resistance), such as clopidogrel. induced platelet responsiveness via increased thromboxane pro-
duction, adenosine diphosphate release, and recruitment. Blood
1991;78:154 – 62.
18. Valles J, Santos MT, Aznar J, Osa A, Lago A, Cosin J, et al.
Acknowledgements: We thank Dr. Cornelia Wolf and Dr.
Erythrocyte promotion of platelet reactivity decreases the effec-
Albrecht Leo for access to samples from blood donors. We tiveness of aspirin as an antithrombotic therapeutic modality: the
are indebted to Christa Dewald, Barbara Calvo, and effect of low-dose aspirin is less than optimal in patients with
Irmhild Rüter for excellent technical assistance. This work vascular disease due to prothrombotic effects of erythrocytes on
is dedicated to Professor M.A.A. Kratzer, M.D. platelet reactivity. Circulation 1998;97:350 –5.
19. Dyszkiewicz-Korpanty AM, Frenkel EP, Sarode R. Approach to the
assessment of platelet function: comparison between optical-
References
based platelet-rich plasma and impedance-based whole blood
1. Collaborative overview of randomised trials of antiplatelet thera-
platelet aggregation methods. Clin Appl Thromb Hemost 2005;
py—I: Prevention of death, myocardial infarction, and stroke by
11:25–35.
prolonged antiplatelet therapy in various categories of patients.
20. Johnson MF, Davis RB. Effects of blood dilution on electrical
Antiplatelet Trialists’ Collaboration. BMJ 1994;308:81–106.
impedance aggregometry. Thromb Res 1986;42:855–7.
2. Cattaneo M. Aspirin and clopidogrel: efficacy, safety, and the
21. Muller MR, Salat A, Pulaki S, Stangl P, Ergun E, Schreiner W, et al.
issue of drug resistance [Review]. Arterioscler Thromb Vasc Biol
Influence of hematocrit and platelet count on impedance and
2004;24:1980 –7. [Epub 2004 Sep 23].
reactivity of whole blood for electrical aggregometry. J Pharmacol
3. Mason PJ, Freedman JE, Jacobs AK. Aspirin resistance: current Toxicol Methods 1995;34:17–22.
concepts [Review]. Rev Cardiovasc Med 2004;5:156 – 63. 22. Muller MR, Schreiner W, Wohlfahrt A, Salat A, Wolner E. The
4. Hankey GJ, Eikelboom JW. Aspirin resistance [Review]. Lancet influence of sample age on collagen-induced platelet aggregation
2006;367:606 –17. in whole blood. Thromb Res 1990;60:477– 87.
5. Patrono C. Aspirin resistance: definition, mechanisms and clinical 23. Riess H, Braun G, Brehm G, Hiller E. Critical evaluation of platelet
read-outs [Review]. J Thromb Haemost 2003;1:1710 –3. aggregation in whole human blood. Am J Clin Pathol 1986;85:50 – 6.
6. Grotemeyer KH, Scharafinski HW, Husstedt IW. Two-year follow-up of 24. Sweeney JD, Labuzetta JW, Fitzpatrick JE. The effect of the
aspirin responder and aspirin non responder. A pilot-study including platelet count on the aggregation response and adenosine
180 post-stroke patients. Thromb Res 1993;71:397– 403. triphosphate release in an impedance lumi-aggregometer. Am J
7. Eikelboom JW, Hirsh J, Weitz JI, Johnston M, Yi Q, Yusuf S. Aspirin- Clin Pathol 1988;89:655–9.
resistant thromboxane biosynthesis and the risk of myocardial infarc- 25. Sweeney JD, Hoernig LA, Michnik BS, Fitzpatrick JE. Whole blood
tion, stroke, or cardiovascular death in patients at high risk for aggregometry: influence of sample collection and delay in study
cardiovascular events. Circulation 2002;105:1650 –5. performance on test results. J Clin Pathol 1989;92:676 –9.
8. Gum PA, Kottke-Marchant K, Welsh PA, White J, Topol EJ. A 26. Ridker PM, Cook NR, Lee IM, Gordon D, Gaziano JM, Manson JE,
prospective, blinded determination of the natural history of aspirin et al. A randomized trial of low-dose aspirin in the primary
resistance among stable patients with cardiovascular disease. prevention of cardiovascular disease in women. N Engl J Med
J Am Coll Cardiol 2003;41:961–5. 2005;352:1293–304. Epub 2005 Mar 7.
9. Antman EM, DeMets D, Loscalzo J. Cyclooxygenase inhibition and 27. Weber AA, Przytulski B, Schanz A, Hohlfeld T, Schror K. Towards a
cardiovascular risk. Circulation 2005;112:759 –70. definition of aspirin resistance: a typological approach. Platelets
10. Antman EM, Cohen M, Bernink PJ, McCabe CH, Horacek T, 2002;13:37– 40.
Papuchis G, et al. The TIMI risk score for unstable angina/non-ST 28. Yee DL, Sun CW, Bergeron AL, Dong JF, Bray PF. Aggregometry
elevation MI: A method for prognostication and therapeutic deci- detects platelet hyperreactivity in healthy individuals. Blood 2005;
sion making. JAMA 2000;284:835– 42. 106:2723–9. [Epub 005 Jun 21].
620 – 628 (2007) Proteomics and
Protein Markers
Mass Spectrometry–Based Hepcidin Measurements

in Serum and Urine: Analytical Aspects and
Clinical Implications
Erwin H.J.M. Kemna,1* Harold Tjalsma,1 Vladimir N. Podust,2 and
Dorine W. Swinkels1
Background: Discovery of the central role of hepcidin cycles and storage conditions, but less influenced by
in body iron regulation has shed new light on the diurnal variation, than is serum hepcidin.
pathophysiology of iron disorders. Information is lack- Conclusion: SELDI-TOF MS can be used to measure
ing on newer analytical approaches to measure hepcidin hepcidin in both serum and urine, but serum requires a
in serum and urine. Recent reports on the measurement standardized sampling protocol.
of urine and serum hepcidin by surface-enhanced laser- © 2007 American Association for Clinical Chemistry
desorption/ionization time-of-flight mass spectrometry
(SELDI-TOF MS) necessitate analytical and clinical The iron-regulating peptide hepcidin is produced by
evaluation of MS-based methodologies. hepatocytes and secreted into plasma (1–3 ). Increased
Methods: We used SELDI-TOF MS, immunocapture, iron stores and inflammation induce hepcidin expression
and tandem MS to identify and characterize hepcidin in (3 ), whereas suppression occurs during hypoxia and
serum and urine. In addition to diagnostic application, anemia (4 ). Although recognition of the central role of
we investigated analytical reproducibility and biologi- hepcidin in body iron regulation has increased our un-
cal and preanalytical variation for both serum and urine derstanding of the pathophysiology of iron disorders
on Normal Phase 20 and Immobilized Metal Affinity (5, 6 ), few investigative tools are available; these include
Capture 30 ProteinChip arrays. We obtained samples an immunodot method for measuring urinary hepcidin
from healthy controls and patients with documented (7 ) and a urinary hepcidin assay based on surface-en-
iron-deficiency anemia, inflammation-induced anemia, hanced laser-desorption/ionization time-of-flight mass
thalassemia major, and hereditary hemochromatosis. spectrometry (SELDI-TOF MS)3 for differentiating various
Results: Proteomic techniques showed that hepcidin-20, disorders of iron metabolism (8 ). A SELDI-TOF MS–
-22, and -25 isoforms are present in urine. Hepcidin-25 in based serum assay is under development for measuring
serum had the same amino acid sequence as hepcidin-25 hepcidin in serum (9 ), but this method requires optimi-
in urine, whereas hepcidin-22 was not detected in se- zation and evaluation to facilitate clinical investigation of
rum. The interarray CV was 15% to 27%, and interspot hepcidin measurement in serum compared with urine
CV was 11% to 13%. Preliminary studies showed that and the biological variation of hepcidin in body fluids. In
hepcidin-25 differentiated disorders of iron metabolism. this study, we used SELDI-TOF MS, immunocapture, and
Urine hepcidin is more affected by multiple freeze-thaw tandem MS to identify and characterize hepcidin in serum
and urine.
1
Department of Clinical Chemistry, Radboud University Nijmegen Med-
ical Center, Nijmegen, The Netherlands.
2 3
Ciphergen Biosystems, Inc., Fremont, CA. Nonstandard abbreviations: SELDI-TOF, surface-enhanced laser desorp-
* Address correspondence to this author at: Department of Clinical Chem- tion/ionization time-of-flight; MS, mass spectrometry; OMIM, Online Mende-
istry 441, Radboud University Nijmegen Medical Centre, P.O. Box 9101, 6500 lian Inheritance in Man; NP20, normal phase 20 ProteinChip array; IMAC30,
HB Nijmegen, The Netherlands. Fax 31 (0) 24-3541743; e-mail: e.kemna@akc. Immobilized Metal Affinity Capture 30 ProteinChip array; EAM, energy-
umcn.nl. absorbing matrix; ACN, acetonitrile; TFA, trifluoroacetic acid; TIC, total ion
Received August 29, 2006; accepted January 10, 2007. current; MS/MS, tandem mass spectrometry; DTT, dithiothreitol; PBS, phos-
Previously published online at DOI: 10.1373/clinchem.2006.079186 phate-buffered saline; PBS-Tx, phosphate-buffered saline with Triton X-100.
620
Materials and Methods We collected samples randomly between December

study participants 2005 and June 2006, at no specified time of day, except
Study participants included a control population of from the endotoxemia patients, from whom samples were
healthy volunteers (laboratory personnel), hereditary collected according to a tight time schedule (10, 13 ). Urine
hemochromatosis (HFE4 C282Y homozygous) patients and blood samples were centrifuged immediately after
cross-sectionally selected from a family screening pro- collection, divided into aliquots to avoid multiple freeze-
gram (various stages of phlebotomy treatment), iron thaw cycles, and stored at ⫺80 °C. We performed the
deficiency anemia patients (Hb ⱕ8.3 mmol/L men, ⱕ7.3 hepcidin assay within 2 months after collection.
mmol/L women; mean corpuscular volume ⱕ80 fL; fer-
ritin ⱕ10 ␮g/L), and thalassemia major patients treated
with chelation therapy. The patients were recruited by laboratory measurements
their physicians during outpatient clinic visits (all in Using an Abbott Aeroset analyzer, we measured total
Radboud University Nijmegen Medical Centre, Nijmegen, serum iron and latent iron binding capacity by the ascor-
The Netherlands, except for the thalassemia major pa- bate/FerroZine colorimetric method, urine creatinine by
tients, who were in Ospedale Sant’Eugenio, Rome, Italy). enzymatic/colorimetric detection (Roche Diagnostics),
Endotoxemia samples from volunteers injected with lipo- and C-reactive protein by immunologic agglutination
polysaccharide were selected from a human endotoxemia detection with latex-coupled polyclonal anti–C-reactive
project (10 ). Written informed consent was obtained from protein antibodies (Abbott Laboratories). We quantified
all study participants, according to the Declaration of serum ferritin by a solid-phase, 2-site chemiluminescent
Helsinki. Characteristics of the study participants are immunometric assay (Immulite 2000 and 2500, Diagnostic
shown in Table 1. In addition, collaborators from sev- Products Corp.) and routine hematology characteristics
eral medical centers in The Netherlands provided urine by use of a Sysmex XE-2100 analyzer.
and serum samples from 3 patients with distinct forms
of hereditary hemochromatosis not associated with the
hemochromatosis (HFE) gene [Online Mendelian In- seldi-tof ms
heritance in Man (OMIM) type 2a, homozygous G320V We performed nonblinded hepcidin measurements by
HJV variation (11 ); OMIM type 4, N144H-caused ferro- use of SELDI-TOF MS as previously reported (8 ). In brief,
portin disease (12 )]. Characteristics of these patients are a 5-␮L sample was applied to Immobilized Metal Affinity
shown in Table 1 in the Data Supplement that accompa- Capture 30 (IMAC30), Normal Phase 20 (NP20), or cation
nies the online version of this article at http://www. exchange ProteinChip arrays (CM10), all equilibrated
clinchem.org/content/vol53/issue4. with appropriate buffers according to the manufacturer’s
instructions (Ciphergen Biosystems). Loaded arrays were
incubated in a humidity chamber for 30 min and then
4
Human gene: HFE, hemochromatosis. washed according to the manufacturer’s instructions and
Table 1. Characteristics of participants who provided sample material for assay validation.a
Hereditary
Control Endotoxemia (model) Iron deficiency anemia Thalassemia majorb hemochromatosisc
n 20 28 6 5 14
Sex (male:female) 9:11 NAd 3:3 1:4 10:4
Age, years 44 (24–62) NA 47 (14–61) 35 (25–44) 56 (23–82)
Hemoglobin, mmol/L 8.7 (7.8–11.0) NA 6.7 (5.0–8.3) 6.3 (6.1–7.3) 9.3 (7.5–12.0)
MCV, fL 88 (84–93) NA 77 (76–79) 86 (78–88) 90 (83–95)
Serum iron (Fe), ␮mol/L 19 (9–38) 18 (7–38) 6 (4–7) 52 (47–62) 22 (10–48)
Fe/TIBC, TS, % 35 (14–73) 40 (23–79) 8 (5–10) 98 (93–100) 56 (21–100)
Ferritin, ␮g/L 60 (11–191) 120 (31–233) 6 (6–10) 784 (272–1710) 64 (28–1361)
C-reactive protein, mg/L ⬍5 9 (⬍5–33) ⬍5 ⬍5 ⬍5
Urinary hepcidin-25 (NP20), 0.52 (0.09–2.97) 6.16 (1.27–19.78) 0.01 (0.01–0.02) 0.21 (0.04–0.98) 0.15 (0.02–1.25)
M Int/mmol Cr
Urinary hepcidin-25 (IMAC30), 1.45 (0.52–7.83) 12.97 (1.95–30.81) 0.10 (0.01–0.19) 0.59 (0.24–2.96) 0.66 (0.08–4.50)
M Int/mmol Cr
Serujm hepcidin-25 (IMAC30), 4.38 (0.58–9.95) 12.69 (4.11–27.33) 0.41 (0.35–0.60) 0.28 (0.18–2.42) 1.60 (0.19–9.39)
M Int/L
a
Data are median (range).
b
Cross-sectional selection of thalassemia major patients with variable transfusion history and treated with different iron chelators (desferrioxamine and/or
deferiprone).
c
Cross-sectional selection of homozygous C282Y patients from family study; 11 patients are under phlebotomy treatment.
d
NA, not available; TIBC, total iron binding capacity; M Int/mmol Cr, megaintensity/mmol creatinine.
622 Kemna et al.: Urine and Serum Hepcidin Measurements by MS
air-dried for 15 min. Finally, 1 ␮L energy-absorbing identification of peptides by tandem mass

matrix (EAM), made up of a 12.5 g/L solution of sinapinic spectrometry
acid in 500 mL/L acetonitrile (ACN) containing 5 mL/L We analyzed peptides of interest for the presence of
trifluoroacetic acid (TFA), was applied twice onto each disulfide bonds. Aliquots of the 30% ACN fractions were
spot surface with the use of polymer-free polypropylene air-dried on an NP20 array. A solution containing 10
pipette tips and allowed to air-dry for 5 min. Mass-to- mmol/L dithiothreitol (DTT) in 50 mmol/L ammonium
charge (m/z) spectra were generated using a Ciphergen bicarbonate, and the sample loaded arrays, was preheated
Protein Biology System IIc TOF mass spectrometer at laser to 70 °C. Then we loaded 10-␮L aliquots of the DTT
intensity 180 (NP20 and CM10) or 175 (IMAC30); detector solution onto the spots and air-dried them at 70 °C. After
sensitivity 9; high mass to acquire 50 kDa; optimization complete evaporation of solutions and cooling to room
interval 1500 to 10 000 Da. External mass calibration was temperature, we applied EAM. We acquired single MS
performed with a mixture of synthetic human hepcidin-25 spectra for unreduced and DTT-reduced samples by use
peptide (2789.4 Da, Peptides International), hepcidin-22, of a Q-STARXL MS/MS (Applied Biosystems) equipped
and hepcidin-20 (2436.1 and 2191.8 Da, respectively; with a Ciphergen PCI-1000 ProteinChip Interface. We
kindly provided by E. Nemeth, University of California, used the reduced samples to acquire tandem mass spec-
Los Angeles). Peak annotation was performed with Ci- trometry (MS/MS) spectra. We subjected the ions of
phergen ProteinChip Software version 3.2.0. From unpub- interest (m/z values of 2198 for hepcidin-20, 2442 for
lished experiments we found that for serum and urine, hepcidin-22, and 2796 for hepcidin-25) to collision-in-
normalization to total ion current (TIC) did not improve duced dissociation and submitted the results to the data-
the hepcidin measurements and was therefore not applied base-mining tool Mascot (Matrix Science) for peptide
in this study. Urine hepcidin measurements showed a identification.
relationship with TIC and total peptide content, which
immunocapture
was predominated by hepcidin under the applied exper-
We captured hepcidin from urinary samples by use of
imental conditions and instrumental settings (data not
Protein G coupled to IDM beads (Ciphergen Biosystems)
shown). Consequently, normalization of urine hepcidin
and polyclonal rabbit antihepcidin serum (generous gift
values to TIC leads to loss of differentiation between
from E. Nemeth and T. Ganz, University of California, Los
samples. In contrast, due to the relatively stable protein
Angeles). We performed all incubation steps at room
content of serum samples, the serum hepcidin concentra-
temperature. Protein G beads were first incubated with
tions did not significantly change upon normalization to
rabbit antiserum diluted 20 times in 0.01 mol/L phos-
TIC (unpublished observations). Although the reabsorp-
phate-buffered saline (PBS; Sigma-Aldrich Chemie BV)
tion and excretion characteristics of hepcidin are unclear,
supplemented with 0.1% Triton X-100 (PBS-Tx). We
we used urinary creatinine to normalize all peak intensi- washed the beads 3 times with PBS-Tx to remove un-
ties for hepcidin-25 in urine. Normalization on creatinine bound serum proteins and resuspended the beads in
is a prerequisite for comparison of urine hepcidin mea- urine, diluted 20 times in PBS containing a final concen-
surements because it is the best method to correct for the tration of 0.25 mol/L NaCl and 0.1% Triton X-100 (PBS0.25-
highly fluctuating concentration differences between Tx). We washed the beads 6 times with PBS0.25-Tx to
urine samples. Relative concentrations were expressed as remove unbound and nonspecifically bound proteins.
mega-intensity units per millimole of creatinine. Relative Finally, we eluted bound proteins with 500 mL/L ACN
concentrations of serum hepcidin-25 were expressed as containing 3 mL/L TFA. To obtain profiles of Protein
mega-intensity units per liter. Information on the assay G– captured proteins, we applied eluates to a CM10 array
can be found on: www.UMCN.NL/Hepcidin. equilibrated with 0.1 mol/L ammonium acetate (pH 3)
and incubated them for 30 min in a humidity chamber. To
purification of peptides from serum and urine obtain reference spectra, we diluted untreated urine sam-
We first fractionated serum samples by use of spin ples once in equilibration buffer before on-spot incuba-
columns containing Q HyperD F resin (Pall Corp.), eluted tion. Spots were washed, allowed to air-dry, and followed
the flow-through fraction by centrifugation, and eluted by sinapinic acid application.
bound proteins with buffers of pH 9, 7, 5, 4, and 3. We
analyzed fractions by use of IMAC30 arrays. Every serum analytical assay characteristics
fraction or urine sample was adjusted to final concentra- We performed spot-to-spot precision for hepcidin-25 on
tions of 5% ACN and 0.5% TFA and bound to PLRP-S NP20 and IMAC30 ProteinChip array with 2 human urine
beads (Polymer Laboratories, Varian). Bound proteins samples with medium and high hepcidin concentrations.
were eluted with 5%, 10%, 20%, 30%, 40%, 50%, and 60% Both samples were applied onto the first 6 positions of the
ACN solutions containing 0.1% TFA. We detected pro- 8-spot array, followed by addition of the EAM. The last 2
teins in eluted fractions by profiling 1 ␮L of each fraction spots were used for control samples. For serum, we
on an NP20 array using Protein Biology System IIc MS. followed the same procedure for a single sample.
We used the same urine and serum samples for a can be verified in both urine and serum with the use of a
chip-to-chip reproducibility study. We collected single purified synthetic reference human hepcidin peptide,
measurements for serum and urine on NP20 and IMAC30 whereas hepcidin-22 can be identified only in urine
array types from randomized chip positions for 10 days (8 (Fig. 1A). This supports the hypothesis that both the 20-
days for serum application on NP20 arrays). Every day a and 25–amino acid peptides are secreted in the circulation
new sample aliquot was thawed, and freshly prepared after production in hepatocytes, whereas the 22–amino
EAM was applied. From these data, we calculated means, acid peptide merely is a urinary degradation product of
SDs, and CVs. hepcidin-25 (14 ).
We measured linearity of peak intensities by dilution In addition to mass tracing, we investigated immuno-
of the urine and serum samples used for the reproduc- capture of all 3 known forms of hepcidin from urine.
ibility tests with human urine or serum (dilution samples) Immunoaffinity assays using polyclonal rabbit antihepci-
from a patient with hepcidin concentrations below those din antibodies [the same as those used in hepcidin dotblot
detectable by SELDI-TOF MS. Immediately before sample assay (7 )] showed that the 3 peaks annotated as hepcidin-
application, we diluted 1 to 18 ␮L sample in a polypro- 25, -22, and -20 can be specifically captured from a human
pylene microcentrifuge container to a 20 ␮L end volume urine sample of a healthy individual (Fig. 1B). MS/MS
with the dilution sample. After pipette mixing, 5 ␮L of the analysis of peptides in a urine sample from an endotox-
diluted sample was spotted on the array. emia patient confirmed that the peaks with m/z of 2198,
We created standard curves of synthetic hepcidin-25 2442, and 2796 corresponded to hepcidin-20, -22, and -25,
for both serum and urine applications. After dissolving respectively (see Fig. 1A in the online Data Supplement).
the lyophilized peptide in distilled water (100 ␮mol/L), Similarly, we identified the m/z 2796 peak in serum as
the peptide was diluted 4500-fold with dilution sample hepcidin-25. Notably, single MS spectra of urine and
for serum or urine (22.2 nmol/L). Dilutions of 18, 16, 12, serum samples displayed a mass shift of ⫹8 Da for all
hepcidin forms after reduction with DTT, as exemplified
8, 4, 2, and 1 ␮L of the 22.2 nmol/L solution with the
for urine hepcidin-25 (see Fig. 1B in the online Data
dilution sample to the final volume of 20 ␮L yielded a
Supplement). This observation confirms the presence of 4
standard curve ranging from 22.2 to 1.11 nmol/L.
disulfide bonds in these peptides, which is a typical
characteristic of hepcidin (15 ). Finally, the MS/MS spec-
preanalytical and biological interferences tra for serum and urine peptides with m/z of 2796
We evaluated the influence of multiple freeze-thaw cycles displayed mostly the same fragment ions, strongly sug-
for both serum and urine using IMAC30 arrays. Five sera gesting that both parent ions corresponded to the same
and 4 urine samples from different participants and peptide hepcidin-25 (see Fig. 1C in the online Data
different intensity concentrations underwent 4 freeze- Supplement).
thaw cycles. After every cycle, we analyzed hepcidin
batchwise by single measurement. analysis of serum and urine hepcidin-25 on
We studied the existence of a circadian rhythm for np20 and imac30 proteinchip arrays
hepcidin in serum and urine by performing a 24-h time Because the roles of hepcidin-20 and -22 in iron metabo-
course in 3 healthy volunteers (1 man, 2 women). Blood lism are unclear (2, 14 ), hepcidin-25 is used for optimiza-
and urine were collected every 3 h for analysis of routine tion of serum and urine hepcidin measurements. To
iron measurements and serum and urine hepcidin evaluate the effect of ProteinChip array type on the
(IMAC30 array), starting at 6:00 AM with fasting blood performance of the hepcidin-25 assay, we analyzed 73
and urine samples, after which the fasting state was urine and serum sample pairs from controls and patients
ended. All samples were processed as described above with various iron metabolism disorders using the previ-
and stored at ⫺80 °C before analysis (batchwise by single ously reported NP20 and IMAC30 arrays (8, 9 ). Although
measurement). hepcidin concentrations measured using NP20 arrays are
lower than those measured with IMAC30 arrays, both
arrays correlate for the urinary application (R ⫽ 0.928, P
statistic analysis ⬍0.0001; Pearson correlation; Fig. 2A). Sensitivity was
We performed statistical analyses with GraphPad Prism higher with the use of IMAC30 compared with the NP20
software, version 4.0. Pearson correlation was used for arrays, likely because of highly specific binding of hepci-
comparison studies. P values ⬍0.05 were considered din to the IMAC-Cu surface. In contrast to the IMAC30
significant. arrays, NP20 arrays did not bind detectable concentrations
of serum hepcidin-25; therefore, we could not perform
Results correlation analyses for serum and urine on NP20 arrays. In
hepcidin identification and characterization contrast, IMAC30 arrays showed a significant correlation
The presence of hepcidin-25 in human urine (8 ) and (R ⫽ 0.822, P ⬍0.0001; Pearson correlation; Fig. 2B).
serum (9 ) has been demonstrated. Our current investiga- Chip-to-chip variation of the urine application for
tion revealed that, in addition to hepcidin-25, hepcidin-20 IMAC30 ranged from 22% at a high intensity of 48 to
Fig. 1. Hepcidin identification and

characterization.
(A), SELDI-TOF MS profiles in serum and
corresponding urine samples from a
participant with an inflammation re-
sponse after lipopolysaccharide injec-
tion. The annotated peak masses
matched to the masses in the reference
peptide mix that contained purified syn-
thetic human hepcidin-20, -22 (E. Nem-
eth, University of California, Los Ange-
les), and -25 (Peptides International).
(B), immunocapture of urine hepcidin.
Urine samples from a healthy individual
were incubated with Protein G beads
loaded with either a polyclonal antihep-
cidin rabbit antibody or control rabbit
IgG. Immunocaptured proteins were an-
alyzed by SELDI-TOF MS, and protein
profiles were compared with the protein
profile of an untreated urine sample
(upper panel). Peaks corresponding to
3 forms of urine hepcidin (8 ) and those
corresponding to rabbit IgGs are indi-
cated. A peak with m/z 2436 that is
present in untreated urine, but not
present in the captured fractions, is
marked (*), as it shows that the capture
of the 3 hepcidin forms was specific.
Notably, on-spot incubation of the same
urine sample with antihepcidin antibod-
ies bound to Protein G-loaded arrays did
not yield detectable hepcidin concentra-
tions (our unpublished observations).
27.5% at a lower intensity of 20 (Table 2), whereas NP20 patient (single measurements). Fig. 2, A and B, in the
variation reached 34% for the same samples. Precision online Data Supplement shows a high degree of linearity
was also better for urine hepcidin measurements using for both urine and serum (R ⫽ 0.993 and 0.971, respectively).
the IMAC30 chip (CVs 11% and 13%) compared with the Intensity values ⬎55 seem to deviate from linearity, perhaps
NP20 chip (CVs of 16% and 23%; Table 2). because of MS detector saturation. Therefore, an intensity
For serum, the reproducibility was strongly affected by value of 55 was considered to be the upper level of detection.
differences in protein binding capacity of both arrays. The construction of a calibration curve with synthetic
Intensities measured from the same sample differed, from human hepcidin-25 showed that both NP20 and IMAC30
an intensity of 5 with relatively high CVs on NP20 array arrays are capable of producing a 6-point calibration
to an intensity of 50 with low CVs on IMAC30 array. We curve based on a urine matrix (see Fig 2, C and D, in the
checked the linearity of the urinary and serum hepcidin online Data Supplement). Again, intensity values ⬎55
application on IMAC30 array by measuring a dilution deviated from linearity and therefore were considered as
series of urine and serum samples from an inflammation exceeding the upper level of detection. Construction of
Table 2. Reproducibility of hepcidin-25 analysis by SELDI-

TOF MS in NP20 and IMAC30 chip arrays in both serum
and urine.
NP20 IMAC30
m/z Intensity m/z Intensity

Urine application
Chip-to-chipa
Level I
Mean 2788.48 13.10 2788.65 19.84
SD 0.94 4.37 1.96 5.46
CV, % 0.03 33.4 0.07 27.5
Level II
Mean 2788.37 41.55 2788.45 47.84
SD 0.89 10.00 1.64 10.50
CV, % 0.03 24.1 0.06 22.0
Spot-to-spotb
Level I
Mean 2787.13 15.23 2788.53 17.83
SD 1.00 2.62 1.55 1.97
CV, % 0.04 16.5 0.06 11.0
Level II
Mean 2787.88 40.61 2788.24 40.89
SD 0.53 9.47 1.00 5.25
CV, % 0.02 23.3 0.04 12.8
Serum application
Chip-to-chipc
Level I
Mean 2788.40 5.26 2788.02 51.96
SD 1.41 1.73 1.56 7.76
CV, % 0.05 32.9 0.06 14.9
Spot-to-spotb
Level I
Mean 2787.54 4.27 2788.33 43.89
SD 0.60 1.14 0.59 5.06
CV, % 0.02 26.6 0.02 11.5
a
Ten-day single measurement with randomized chip position.
b
One chip (6 replicates).
c
Eight-day single measurement (NP20 array) and 10-day single measurement
(IMAC30 array) with randomized chip position.
diagnostic application
We categorized urine and serum sample pairs into 5
groups after clinical diagnosis. Fig. 3 shows that, despite
Fig. 2. Detection of serum and urine hepcidin-25 on NP20 and IMAC30 the variation within each group and a slight overlap,
arrays.
differentiation between inflammation-induced and iron
Correlation of urinary hepcidin-25 values as measured by NP20 and IMAC30
ProteinChip arrays (A) and coupled serum and urine hepcidin-25 concentrations deficiency anemia is clear for both serum and urine
measured on IMAC30 array (B). The solid line represents the regression line, and hepcidin concentrations measured on IMAC30 array. The
the dotted line indicates the 95% CI of the regression line. It should be noted that
the apparent increased scatter in absolute variation on high hepcidin values
results are comparable with the results obtained with
compared with low values, as presented in panel A of this figure, is because NP20 and consistent with previous reports (7, 8 ). The
absolute concentrations are used in this plot. The relative average for the high broad ranges of the thalassemia group and the hemochro-
values is not more scattered than the low-end values (data not shown), which is
in concordance with the CV values shown in Table 2. As it is currently unknown matosis group are in accordance with the individual
whether and how much hepcidin is reabsorbed in the tubulus, we cannot indicate heterogeneity in genetic makeup, extent of anemia, and
whether this aspect contributes to the scatter in panel B of this figure.
treatment (16 –18 ).
the same concentration interval on IMAC30 array in a In this study, we found serum and urine hepcidin
serum matrix showed hepcidin-25 only for a concentra- concentrations close to zero in 2 brothers treated for
tion of 22.2 nmol/L and above (results not shown). juvenile hemochromatosis (OMIM type 2a) and values in
preanalytical and biological determinants

that influence hepcidin-25 measurements in
serum and urine
Decreases in hepcidin concentrations in serum samples
due to multiple freeze-thaw cycles is of minor importance
compared with those in urine (see Fig. 3 in the online
Data Supplement). Preliminary results from urine sam-
ples stored at ⫺80 °C and ⫺20 °C during a 6-month
course showed that (a) hepcidin was stable only at ⫺80 °C
and (b) addition of protease inhibitor phenylmethylsulfo-
nyl fluoride (19 ) had no beneficial effect (unpublished
observations).
Hepcidin concentrations in serum follow a clear circa-
dian rhythm, such that the concentrations were 2- to
6-fold higher at 1500 than at 0600. Urinary concentrations,
however, show a more blunted response, and thus less
Fig. 3. Urinary and serum hepcidin-25 concentrations in selected

clinical populations.
Urinary hepcidin-25 concentrations measured on NP20 array (A) and IMAC30
array (B) and serum hepcidin-25 measured on IMAC30 array (C). LPS, volunteers
injected with lipopolysaccharide (6 h after injection); IDA, iron deficiency anemia
patients; TM, thalassemia major patients; HH(YY), C282Y homozygous hemo-
chromatosis patients. Box plots show 25th and 75th percentile with median for
every group. Error bars represent minimum and maximum values. Characteristics
of each population are shown in Table 1.
Fig. 4. Circadian influence on urinary and serum hepcidin-25
concentrations.
Urinary and serum hepcidin concentrations and serum iron concentration in 3
healthy volunteers during a 24-h time course (participants A and C, premeno-
the upper level of the reference range for a patient treated pausal women; participant B, man). Circadian influence is seen on all 3
variables. Individual differences are visible on amplitude and frequency pattern
for ferroportin disease (OMIM type 4; see Table 1 in the and relative concentration difference between urinary and serum hepcidin. Note
online Data Supplement). These data confirm the poten- that in participant A, C-reactive protein concentrations increased to 10 mg/L in
the last 6 h of the time course, which suggests that an inflammatory process
tial of hepcidin analysis in prescreening for the presence may have contributed to the persistent increase in hepcidin and the decrease in
of non-HFE hemochromatosis. serum iron during this period.
diurnal variation. The course of serum iron shows an excretion from the blood into the urine. Construction of a
inverse association with serum and urine hepcidin values calibration curve by addition of the same amount of
(Fig. 4). hepcidin-25 to serum or diluted urine confirms this effect,
leading to speculation that a binding protein in serum
Discussion prevents the binding of free hepcidin on the chip surface
The IMAC30 array– based urinary hepcidin assay corre- at low concentrations (27 ). To circumvent differences in
lated significantly with the previously described urinary binding characteristics or possible matrix interferences,
hepcidin MS application using NP20 arrays (8 ) and methodologic approaches such as plasma fractionation (27 )
showed greater sensitivity and reproducibility, resulting or the use of magnetic reversed-phase beads (28 ) have to be
in improved analytical performance. The increase in pro- investigated for utility on hepcidin measurements.
tein-binding capacity of the IMAC30 array makes it Our reported SELDI-TOF MS method detected all 3
particularly suitable for serum analysis and measurement isoforms of hepcidin, improved urinary hepcidin analysis,
of low hepcidin concentrations in urine. and enabled measurement of serum hepcidin by IMAC30
The total assay variation includes analytical variation, array. We show for the first time a high correlation of
biological or diurnal variation, and preanalytical varia- concentrations measured in corresponding serum and
tion, all contributing to the widening of ranges of the urine samples. Selection of the optimal body fluid for
clinical utility clusters. In the future, we plan to routinely analysis, however, is influenced by preanalytical, ana-
perform duplicate measurements and investigate isotope lytical, and biological variations that effect serum and
dilution (20 ) or hepcidin derivatives as internal standards urine differently. Urine is more vulnerable to multiple
to reduce analytical variability. freeze-thaw cycles and storage temperatures other than
Biological variation in hepcidin concentrations due to a ⫺80 °C but less influenced by diurnal variation. Serum
circadian rhythm correlated inversely with daily varia- is more susceptible to circadian variation, and therefore
tions in serum iron concentrations (21 ), in accordance standardization of sampling time is needed in clinical
with previous reports that transferrin receptors 1 and 2 studies with serum. Serum is likely to be more sensitive
on the outer hepatocyte membrane act as sensors of than urine, however, for monitoring short-term kinetics
circulating iron and TS, thereby linking low serum iron of body hepcidin concentrations. Therefore, the speci-
with increased hepcidin synthesis (22, 23 ). Circadian men of first choice depends on study design and prac-
rhythm may also lead to important variation in outcome tical aspects such as availability of materials and equip-
if sampling time is not standardized, thereby contribu- ment. The ability to measure hepcidin in both serum and
ting to the wide variation in hepcidin concentrations of urine confirms that the learning process on hepcidin
the control population (Fig. 3C). Sampling according to characteristics, kinetics, and clinical utility has only just
protocol led to decreased variation in hepcidin concentra- begun.
tions within the lipopolysaccharide group. The low vari-
ation within the iron deficiency anemia group is likely to
be due to the dominant down-regulating influence of an
iron-deficient erythropoiesis on hepcidin that prevents We thank Elizabeta Nemeth and Tomas Ganz (University
circadian increases. of California, Los Angeles) for kindly providing samples
Variation in hepcidin results is also attributable to of synthetic hepcidin-20 and -22 and polyclonal antihep-
preanalytical conditions. According to our preliminary cidin rabbit serum. We thank Henk Engel (Isala Klinieken,
results, urine samples are more susceptible than serum to Zwolle, The Netherlands), as well as Omer Njajou, Cor-
variation from multiple freeze-thaw cycles, which should nelia van Duijn, and Leon Testers (Erasmus Medical
be avoided, and urine samples should be stored at ⫺80 °C Center, Rotterdam, The Netherlands), for providing sam-
as soon as possible (24, 25 ). ple material of 2 HJV G320V patients and a FPN N144H
Measurement of serum and urine hepcidin by the same patient, respectively, and Peter Pickkers for sharing the
technique, and under the same circumstances, enabled endotoxemia samples. We thank Giuliana Zanninelli (Os-
comparison of serum with urinary values. Although their pedale Sant’Eugenio, Rome, Italy) for selection and col-
association was strong, confirming the previously re- lection of sample material of thalassemia major patients.
ported correlation between hepcidin mRNA expression in We also thank Mirian Janssen, Esther Jacobs, Lammy
liver cells and urinary hepcidin excretion (26 ), differences Elving, and all other physicians involved in selection and
in serum and urine composition prohibit absolute com- collection of samples.
parison of analytical characteristics such as binding com-
petition on the array surface and flight behavior during
SELDI-TOF MS analysis (27 ). Aspects such as ionization
References
efficiency or ion suppression also can play a role and 1. Krause A, Neitz S, Magert HJ, Schultz A, Forssmann WG, Schultz-
should be investigated in future studies. Meanwhile, Knappe P, et al. LEAP-1, a novel highly disulfide-bonded human
differences in specimen behavior preclude calculations of peptide, exhibits antimicrobial activity. FEBS Lett 2000;480:147–
the kinetics of hepcidin release by hepatocytes and its 50.
2. Park CH, Valore EV, Waring AJ, Ganz T. Hepcidin, a urinary that is involved in iron uptake and hereditary hemochromatosis.
antimicrobial peptide synthesized in the liver. J Biol Chem 2001; J Biol Chem 2002;277:37597– 603.
276:7806 –10. 16. Papanikolaou G, Tzilianos M, Christakis JI, Bogdanos D, Tsimirika
3. Pigeon C, Ilyin G, Courselaud B, Leroyer P, Turlin B, Brissot P, et K, MacFarlane J, et al. Hepcidin in iron overload disorders. Blood
al. A new mouse liver-specific gene, encoding a protein homolo- 2005;105:4103–5.
gous to human antimicrobial peptide hepcidin, is overexpressed 17. Waalen J, Nordestgaard BG, Beutler E. The penetrance of hered-
during iron overload. J Biol Chem 2001;276:7811–9. itary hemochromatosis. Best Pract Res Clin Haematol 2005;18:
4. Nicolas G, Bennoun M, Porteu A, Mativet S, Beaumont C, Grand- 203–20.
champ B, et al. Severe iron deficiency anemia in transgenic mice 18. Kearney SL, Nemeth E, Neufeld EJ, Thapa D, Ganz T, Weinstein
expressing liver hepcidin. Proc Natl Acad Sci U S A 2002;99: DA, et al. Urinary hepcidin in congenital chronic anemias. Pediatr
4596 – 601. Blood Cancer 2007;48:57– 63.
5. Ganz T. Hepcidin in iron metabolism. Curr Opin Hematol 2004; 19. Thongboonkerd V, McLeish KR, Arthur JM, Klein JB. Proteomic
11:251– 4. analysis of normal urinary proteins isolated by acetone precipita-
6. Swinkels DW, Janssen MCH, Bergmans J, Marx JJM. Hereditary tion or ultracentrifugation. Kidney Int 2002;62:1461–9.
hemochromatosis: genetic complexity and new diagnostic ap- 20. Kema IP, Meiborg G, Nagel GT, Stob GJ, Muskiet FAJ. Isotope
proaches. Clin Chem 2006;52:950 – 68. dilution ammonia chemical ionization mass fragmentographic
7. Nemeth E, Valore EV, Territo M, Schiller G, Lichtenstein A, Ganz T. analysis of urinary 3-o-methylated catecholamine metabolites:
Hepcidin, a putative mediator of anemia of inflammation, is a type rapid sample clean up by derivatisation and extraction of lyophi-
II acute-phase protein. Blood 2003;101:2461–3. lized samples. J Chromatogr B Biomed Appl 1993;617:181–9.
8. Kemna E, Tjalsma H, Laarakkers C, Nemeth E, Willems H,
21. Witte DL, Crosby WH, Edwards CQ, Fairbanks VF, Mitros FA.
Swinkels D. Novel urine hepcidin assay by mass spectrometry.
Practice guideline development task force of the College of
Blood 2005;106:3268 –70.
American Pathologists: hereditary hemochromatosis. Clin Chim
9. Tomosugi N, Kawabata H, Wakatabe R, Higuchi M, Yamaya H,
Acta 1996;245:139 –200.
Umehara H, et al. Detection of serum hepcidin in renal failure and
22. Frazer DM, Anderson GJ. The orchestration of body iron intake:
inflammation by using ProteinChip system. Blood 2006;108:
how and where do enterocytes receive their cues? Blood Cell Mol
1381–7.
Dis 2003;30:288 –97.
10. Pickkers P, Dorresteijn MJ, Bouw M, van der Hoeven JG, Smits P.
In vivo evidence for nitric oxide-mediated calcium-activated potas- 23. Ganz T, Nemeth E. Regulation of iron acquisition and iron distri-
sium-channel activation during human endotoxemia. Circulation bution in mammals. Biochim Biophys Acta 2006;1763:690 –9.
2006;114:414 –21. 24. Schaub S, Wilkins J, Weiler T, Sangster K, Rush D, Nickerson P.
11. Papanikolaou G, Samuels ME, Ludwig EH, MacDonald ML, Urine protein profiling with surface-enhanced laser-desorption/
Franchini PL, Dube MP, et al. Mutations in HFE2 cause iron ionization time-of-flight mass spectrometry. Kidney Int 2004;65:
overload in chromosome 1q-linked juvenile hemochromatosis. Nat 323–32.
Genet 2004;36:77– 82. 25. Klasen IS, Reichert LJM, de Kat Angelino CM, Wetzels JFM.
12. Njajou OT, de Jong G, Berghuis B, Vaessen N, Snijders PJLM, Quantitative determination of low and high molecular weight
Goossens JP, et al. Dominant hemochromatosis due to N144H proteins in human urine: influence of temperature and storage
mutation of SLC11A3: clinical and biological characteristics. time. Clin Chem 1999;45:430 –2.
Blood Cell Mol Dis 2002;29:439 – 43. 26. Detivaud L, Nemeth E, Boudjema K, Turlin B, Troadec MB, Leroyer
13. Kemna E, Pickkers P, Nemeth E, van der Hoeven H, Swinkels D. P, et al. Hepcidin levels in humans are correlated with hepatic iron
Time-course analysis of hepcidin, serum iron, and plasma cyto- stores, hemoglobin levels, and hepatic function. Blood 2005;106:
kine levels in humans injected with LPS. Blood 2005;106: 746 – 8.
1864 – 8. 27. Hortin GL. The MALDI-TOF mass spectrometric view of the plasma
14. Ganz T. Hepcidin. A regulator of intestinal iron absorption and iron proteome and peptidome. Clin Chem 2006;52:1223–37.
recycling by macrophages. Best Pract Res Clin Haematol 2005; 28. Vilanueva J, Shaffer DR, Philip J, Chaparro CA, Erdjument-Bromage
18:171– 82. H, Olshen AB et al. Differential exoprotease activities confer
15. Hunter HN, Fulton DB, Ganz T, Vogel HJ. The solution structure of tumor-specific serum peptidome patterns. J Clin Invest 2006;
human hepcidin, a peptide hormone with antimicrobial activity 116:271– 84.
Protein Markers
Protein Profiling of Microdissected Pancreas

Carcinoma and Identification of HSP27 as a
Potential Serum Marker
Christian Melle,1 Günther Ernst,1 Niko Escher,1 Daniel Hartmann,2
Bettina Schimmel,1 Annett Bleul,1 Heike Thieme,1 Roland Kaufmann,3 Klaus Felix,2
Helmut M. Friess,2 Utz Settmacher,3 Merten Hommann,3 Konrad K. Richter,3
Wolfgang Daffner,3 Horst Täubig,4 Thomas Manger,4 Uwe Claussen,1 and
Ferdinand von Eggeling1*
Background: Patients with pancreatic adenocarcinomas specificity of 84% for the recognition of pancreatic
have a poor prognosis because of late clinical manifes- cancer.
tation and the tumor’s aggressive nature. We used pro- Conclusions: The detection of DJ-1 and HSP27 in pure
teomic techniques to search for markers of pancreatic defined tissue and the retrieval of HSP27 in serum by
carcinoma. antibody-based methods identifies a potential marker
Methods: We performed protein profiling of microdis- for pancreatic cancer.
sected cryostat sections of 9 pancreatic adenocarcinomas © 2007 American Association for Clinical Chemistry
and 10 healthy pancreatic tissue samples using Protein-
Chip technology (surface-enhanced laser desorption/ Pancreatic cancer is a formidable challenge in oncology
ionization). We identified proteins by use of 2-dimen- and has the lowest 5-year survival rate (⬃2%) of any solid
sional gel electrophoresis, peptide fingerprint mapping, cancer (1 ). Only 10% of patients present with a potentially
and immunodepletion and used immunohistochemistry curable tumor. To gain a chance of combating this cancer
for in situ localization of the proteins found. We used type, we must elucidate early tumorigenic processes. In
ELISA to quantify these proteins in preoperative serum current practice, cancer antigen 19-9 assays and imaging
samples from 35 patients with pancreatic cancer and 37 techniques are not optimal for detecting small pancreatic
healthy individuals. lesions. Improved understanding of DNA/RNA alter-
Results: From among the differentially expressed sig- ations and protein concentrations, in combination with
nals that were detected by ProteinChip technology, we the development of high-throughput, sensitive tech-
identified 2 proteins, DJ-1 and heat shock protein 27 niques, could lead to the discovery of a panel of biomar-
(HSP27). We then detected HSP27 in sera of patients by kers that will enable aggressive therapy while tumors are
use of ELISA, indicating a sensitivity of 100% and a still curable (2 ).
Surface-enhanced laser desorption/ionization (SELDI)5
is a proteomic high-throughput technique that uses chro-
matographic surfaces to retain proteins depending on their
1
Core Unit Chip Application, Institute of Human Genetics and Anthro- physicochemical properties, followed by direct analysis via
pology, Medical Faculty at the Friedrich Schiller University Jena, Jena, Ger-
many.
time-of-flight mass spectrometry (MS) (3 ). This technique
2
Department of General Surgery, University of Heidelberg, Heidelberg, requires only a small amount of sample, making it ideal for
Germany. small biopsies or microdissected tissue (required to produce
3
Department of General, Visceral and Vascular Surgery, Medical Faculty the homogeneous tissue samples typically used in cancer
at the Friedrich Schiller University Jena, Jena, Germany.
4
SRH Wald-Klinikum Gera, Department of General and Visceral Surgery,
Gera, Germany.
* Address correspondence to this author at: Institut für Humangenetik und
5
Anthropologie, CUCA, 07740 Jena, Germany. Fax 0049-0-3641-935518; e-mail Nonstandard abbreviations: SELDI, surface-enhanced laser desorption/
fegg@mti.uni-jena.de. ionization; MS, mass spectrometry; 2-DE, 2-dimensional gel electrophoresis;
Received August 29, 2006; accepted January 11, 2007. HSP, heat shock protein; TBS, Tris-buffered saline; IHC, immunohistochemis-
Previously published online at DOI: 10.1373/clinchem.2006.079194 try.
629
630 Melle et al.: Protein Profiling in Pancreas Carcinoma
research). Microdissected tissue material, free of contaminat- washing steps with water, we applied 2 ⫻ 0.5 ␮L sinap-
ing and unwanted tissue components, is extremely impor- inic acid (saturated solution in 0.5% trifluoroacetic acid/
tant for producing clean data for biomarker identification in 50% acetonitrile). We performed mass analysis by use of a
cancer diagnostics and in elucidating clonal heterogeneity of ProteinChip Reader (series 4000; Ciphergen) according to
tumors. We were able to show in a previous study that the an automated data collection protocol. Spectra were nor-
detection of differentially expressed proteins was possible malized with total ion current and cluster analysis of the
only in pure microdissected samples (4 ). In the case of detected signals. We calculated respective P values for
pancreatic cancer, the tumor cells have to be separated from healthy pancreatic tissue and pancreatic carcinoma tissue
all surrounding tissue constituents. This separation can be with CiphergenExpress (version 3.0). We selected normal-
done only with an extremely precise technique such as ized spectra with signals between 2.5 and 20 kDa for low
laser-based microdissection. Laser-based microdissection range and 20 and 200 kDa for high range, exhibiting a
has been combined with ProteinChip technology to identify signal-to-noise ratio of at least 10, and analyzed them with
protein markers in other cancers (5–11 ). the Mann–Whitney U-test for nonparametric data sets.
In this study, we used ProteinChip technology to
analyze pure microdissected populations of cells from two-dimensional gel electrophoresis
healthy exocrine pancreatic tissue and the central and We prepared samples for 2-dimensional gel electrophore-
peripheral areas of pancreatic adenocarcinomas to detect sis (2-DE) directly from surgical material of pancreatic
discriminating specific protein profiles. tumor and corresponding healthy pancreatic tissue as-
sessed by a pathologist. Proteins were isolated and 2-DE
Materials and Methods performed as described (12 ). In brief, isoelectric focusing
laser microdissection was carried out on a Multiphor II (Amersham) using 7-cm
We obtained 9 pancreatic central tumor areas (pT2/pT3), immobilized pH gradient strips in a pI interval of 3 to 10.
matched healthy pancreatic samples (n ⫽ 10), and 9 Vertical sodium dodecyl sulfate–polyacrylamide gel elec-
pancreatic tumor margins [mean age 61.1 (SD 6.2) years] trophoresis was performed in a Novox MiniGel system
after surgical resection with informed consent at the (Invitrogen) using 4% to 12% Bis-Tris Zoom™ gel (In-
Department of General and Visceral Surgery of the vitrogen). The gels were stained with Simply Blue Safe
Friedrich Schiller University Jena, Germany. The samples Stain (Enhanced Coomassie; Invitrogen).
were collected fresh, snap-frozen in liquid nitrogen, and
stored at ⫺80 °C. We categorized tumor specimens ac- in-gel digestion
cording to their WHO classification. We compared protein patterns of the 2-DE gels from
Laser microdissection was performed with a laser healthy pancreatic and tumor tissue and excised consis-
microdissection and pressure catapulting microscope tently differentially expressed proteins as well as ⬃95
(Palm) as described elsewhere (12 ). Briefly, we microdis- additional spots. In-gel digestion of proteins was per-
sected ⬃3000 to 5000 cells each on native air-dried, formed as described (12 ). In brief, excised gel pieces were
unstained cryostat tissue sections in a maximum of 20 to destained and dried. After rehydration and digestion with
30 min. We extracted proteins by incubating with 10 ␮L 10 ␮L of a trypsin solution (0.02 g/L; Promega) at 37 °C
lysis buffer (100 mmol/L Na-phosphate, pH 7.5, overnight, we applied supernatants directly on a NP20
5 mmol/L EDTA, 2 mmol/L MgCl2, 3 mmol/L 2-␤- ProteinChip array (Ciphergen). An empty gel piece un-
mercaptoethanol, 1 mL CHAPS, 500 ␮mol/L leupeptin, derwent the same treatment as a control. After addition of
and 1 mmol/L phenylmethylsulfonyl fluoride) for 30 min the matrix (CHCA; Ciphergen), we analyzed peptide
on ice. After centrifugation (15 min; 13 000g), the super- fragment masses by use of the ProteinChip reader. The
natant was immediately analyzed or frozen in liquid spectra for the peptide-mapping experiments were
nitrogen for a maximum of 1 day. externally calibrated using 5 proteins including Arg8-
vasopressin (1082.2 Da), somatostatin (1637.9 Da), dy-
profiling of tissues norphin (2147.5 Da), ACTH (2933.5 Da), and insulin
We analyzed the protein lysates from microdissected ␤-chain (3495.94 Da). We identified proteins using the
tissues (central tumor, tumor margin, and healthy tissue) fragment masses generated through trypsin digestion by
on both strong anion exchange arrays (Q10) and weak searching in a publicly available database (ProFound;
cation exchange arrays (CM10; Ciphergen Biosystems) as http://prowl.rockefeller.edu/prowl-cgi/profound.exe).
described (12 ). We preincubated array spots in a wash-
ing/loading buffer containing 100 mmol/L Tris-buffer, immunodepletion assay
pH 8.5, with 0.02% Triton X-100 (for Q10 arrays) and We incubated 2 ␮L antihuman heat shock protein 27
100 mmol/L Tris-buffer, pH 4.5, with 0.02% Triton X-100 (HSP27) polyclonal antibody (SP5105P; Acris) or anti-
(for CM10 arrays) followed by application of 2 ␮L sample human DJ-1 monoclonal antibody (ab11251; Abcam) with
extract on ProteinChip arrays, which were incubated at 10 ␮L protein A-agarose (Sigma-Aldrich) for 15 min on
room temperature for 90 min in a humidity chamber. ice. Pellets were generated by centrifugation, and the
After washing 3 times with the same buffer and 2 final supernatants were discarded. The pellets were washed
twice with a buffer containing 20 mmol/L HEPES, pH 7.8, western blot

25 mmol/L KCl, 5 mmol/L MgCl2, 0.1 mmol/L EDTA, Identification of HSP27 and DJ-1 was verified by Western
and 0.5 mL NP-40. We then incubated 5 ␮L of a lysate blot. We subjected 30 ␮g crude extract of pancreatic tumor
from laser-dissected pancreatic tumor with each pellet for tissue to 12% sodium dodecyl sulfate–polyacrylamide gel
45 min on ice. We incubated 5 ␮L of the lysate with electrophoresis and electrophoretically transferred it to
protein A-agarose without the specific antibody, as a polyvinylidene difluoride membrane (Bio-Rad). Mem-
negative control, for 45 min on ice. After incubation, branes were incubated overnight at 4 °C with a 1:1000
samples were cleared by centrifugation, and 2 ␮L of each dilution of anti-HSP27 antibody (SP5105P; Acris) or a
supernatant was analyzed by use of ProteinChip arrays. 1:1000 dilution of anti–DJ-1 antibody (ab11251; Abcam;
diluted in 20 g/L milk powder in TBS containing 05 mL/L
immunohistochemistry Tween 20) and for 3 h with the corresponding secondary
We placed 8-␮m cryostat sections of pancreatic cancer antibody. Both HSP27 and DJ-1 were detected by alkaline
tissue and adjacent healthy tissue on slides, air-dried phosphatase reaction. We estimated band intensities for
them for ⬃60 min at 20 °C, and fixed them in paraformal- both proteins by visual inspection.
dehyde as described (11 ). After fixation, slides were
treated in the microwave at 80 watts (3 ⫻ 3 min) in Results
10 mmol/L citric acid, pH 6.0, to inhibit endogenous protein profiling of central pancreatic
peroxidase activity. We rinsed them twice with Tris- tumor, tumor margin, and adjacent healthy
buffered saline (TBS), pH 7.4, and incubated them over- tissue
night at 4 °C in a humidity chamber with a corresponding We excised tissue areas corresponding to ⬃3000 to 5000
primary polyclonal antibody against HSP27 (SP5105P; cells per probe by use of laser microdissection and pres-
Acris) or a primary antihuman DJ-1 monoclonal anti- sure catapulting microscope. In this way, we successfully
body (ab11251; Abcam). We rinsed the slides 3 ⫻ 10 min collected 28 samples in total (10 healthy pancreatic sam-
in TBS and used the Vectastain Elite ABC reagent set ples, 9 central pancreatic tumors, and 9 tumor margins).
(Vector Laboratories) and the Jenchrom pxbl reagent set All protein lysates from the microdissected tissues were
(MoBiTec) according to the manufacturer’s instructions to applied to both the Q10 arrays and the CM10 arrays and
visualize antibody localization. Negative controls were analyzed on a ProteinChip Reader Series 4000. The SELDI
incubated with the labeled secondary antibody only. measurements of all tissue samples detected up to 340
Sections cut in parallel to the immunohistochemistry peaks in the 2.5- to 200-kDa interval, with normalized
(IHC)-treated sections were stained by hematoxylin and intensities. After evaluation with CiphergenExpress, a
eosin for better identification of different tissue areas. IHC number of these peaks were found to be significantly
staining was evaluated by a pathologist. different between pancreatic carcinomas and healthy pan-
creatic tissue samples (Table 1).
quantification of hsp27 by elisa
In addition to the tissue samples, we independently tested identification of signals
a set of serum samples from 35 patients (pancreatic tumor; To separate protein lysates, we subjected histologically
pT2/pT3), taken before surgery at the Department of checked pancreatic tumor pieces and biopsies derived
General and Visceral Surgery of the Friedrich Schiller
University Jena and at the Department of General Sur- Table 1. Significantly different signals that separate
gery, University of Heidelberg. The samples were imme- pancreatic carcinoma and adjacent healthy pancreatic
diately divided into aliquots and frozen at ⫺80 °C. Serum tissue detected on Q10 arrays (anion exchanger) and
samples from healthy donors (n ⫽ 37) were obtained with CM10 arrays (cation exchanger).
the same protocol, divided into aliquots, and frozen at Signal up-regulated Molecular
⫺80 °C. The mean (SD) age for the tumor patients was in tissue mass, Da Array surface
61.3 (8.1) years, and for the control volunteer group, 45.6 Healthy 7.961 CM10
(15.4) years. The set did not include any sera from patients Carcinoma 11.171 Q10
whose samples were used for ProteinChip array analysis. Healthy 15.147 CM10
We measured serum HSP27 concentrations by use of Healthy 15.898 CM10
an appropriate ELISA (Sigma-Aldrich) in duplicate, ac- Carcinoma 19.939a Q10
cording to the manufacturer’s instructions. We measured Carcinoma 22.538 CM10
ELISA plates on a microtiter plate reader (MRX II; Dynex Carcinoma 22.749a CM10
Technology) at 450 nm and calculated concentration of Healthy 25.083 Q10
Healthy 25.916 Q10
the respective protein in serum according to a calibra-
Carcinoma 66.731 CM10
tion curve. We calculated P values by 1-sided t-test,
Carcinoma 66.738 Q10
constructed ROC curves for HSP27 serum concentration
Carcinoma 134.259 Q10
by plotting sensitivity vs 1-specificity, and calculated the a
Signals representing DJ-1 and HSP27.
areas under the ROC curves.
from healthy pancreatic tissue to 2-DE (see Fig. 1 in the

Data Supplement that accompanies the online version of
this article at http://www.clinchem.org/content/vol53/
issue4). Numerous protein spots showing differential
expression were observed. Approximately 95 protein
spots in the interval of ⬃10 to 150 kDa were excised from
the gels, and we analyzed peptide fingerprints of the
tryptic-digested spots by use of SELDI time-of-flight MS.
In this way, we were able to identify 29 proteins by
database searching (ProFound; http://prowl.rockefeller.
edu/prowl-cgi/profound.exe; see Table 1 in the online Data
Supplement).
One of these identified proteins, DJ-1 (see Fig. 1, spot 2,
in the online Data Supplement), matched well in molec-
ular mass with a significantly differentially expressed
signal detected in prior protein profiling using SELDI.
This signal of ⬃20 kDa was detected on Q10 arrays and
showed an increased expression in samples derived from
pancreatic tumor. Presence of the spot discriminated
significantly between central pancreatic carcinoma and
tumor margin and healthy adjacent pancreatic tissue (P ⫽
3.66 ⫻ 10⫺2) as well as between pancreatic carcinoma and
healthy pancreatic tissue (2.08 ⫻ 10⫺2). Another signifi-
cantly different signal possessing an m/z of nearly 23 kDa
matched well to a protein identified as HSP27 (see Fig. 1,
spot 13, in the online Data Supplement). This significantly
different signal was up-regulated in pancreatic carcinoma
tissue compared with healthy pancreatic tissue (2.68 ⫻
10⫺2) as detected on CM10 arrays in prior protein Fig. 1. Immunodepletion assay of microdissected pancreatic tumor
profiling. tissue.
We double-checked that DJ-1 and HSP27 match the (A), peak (19937.79 Da) representing DJ-1 was detectable in the negative
differentially expressed peaks at 19.9 and 22.7 kDa by use control and clearly decreased in the depleted control. (B), peak (22740.34 Da)
representing HSP27 was detectable in the negative control but not in the
of ProteinChip analysis with immunodeplete assays, us- corresponding depleted probe.
ing microdissected pancreas carcinoma tissue as starting
material. Analyses showed that the peaks corresponding
to DJ-1 and HSP27 were reduced. In the negative control To further confirm that the localized DJ-1 and HSP27
without the specific antibody, these peaks were clearly are identical to the peaks found by ProteinChip analysis,
detectable (Fig. 1). areas of similar size from tumorous and healthy tissue
that were previously analyzed in IHC were obtained by
characterization by immunohistochemistry tissue laser microdissection. In protein lysates from the
and western blot pancreatic tumor fraction, we detected signals identical in
To further characterize the identified markers and to mass to the interesting peaks obtained with the initial
localize DJ-1 and HSP27 in tissue sections, we examined SELDI-MS analysis on proper arrays (Q10 and CM10). In
their expression in several pancreatic tissue samples by the protein lysate from IHC-negative areas, these peaks
immunohistochemistry using specific antibodies against were absent (Fig. 3). We also examined by Western blot
DJ-1 and HSP27. Negative controls without the primary the expression levels of DJ-1 and HSP27 in lysates derived
antibody or without any antibody had no signal. Both from an independent set of pancreatic tissue specimens
healthy pancreatic cells and malignant tumor cells dem- (see Fig. 3 in the online Data Supplement). We detected
onstrated cytoplasmic signals for DJ-1 and HSP27 in all strong signals corresponding to DJ-1 and HSP27 in the
tissue samples examined, but in every case with higher majority of analyzed probes.
signal intensities in tumor cells (Fig. 2). Quantitative
differences between the expression of these interesting analyses of hsp27 in serum
proteins in healthy pancreatic cells and malignant tumor In addition to the tissue samples, we quantified HSP27
cells were as clear as in ProteinChip array results. Fur- in a sample set of sera derived from pancreatic cancer
thermore, we carried out IHC assays to a number of patients and controls (n ⫽ 72) by use of an independent
additional proteins, including PEBP, cystatin B, and cy- ELISA analysis. The concentration of HSP27 in serum
clophilin A (see Fig. 2 in the online Data Supplement). from tumor patients was found to be significantly higher
Fig. 2. IHC of DJ-1 (A) and HSP27 (B) visualized by

microscopy (magnification ⫻400).
Tissue derived from pancreatic carcinoma with increased signal
intensity in carcinoma structures.
than in serum from controls (P ⬍0.001). The median for

controls was 0.76 ␮g/L, and for patients with pancreatic
carcinoma, 2.93 ␮g/L (Fig. 4). We constructed ROC curves
for HSP27 serum concentration, resulting in a sensitivity
of 100% at a specificity of 84% and a cutoff of 1.33 ␮g/L.
The area under the ROC curve was calculated as 0.985
(Fig. 4).
Discussion
Despite enormous efforts, relevant markers useful for
screening have been established in only a few tumor types
(13, 14 ), and no studies have found markers for early
detection of pancreatic cancer (15–19 ). 2-DE, especially in
combination with microdissection, seems an appropriate
tool (19 ), but proteins in the peptide interval, as well as
those of high hydrophobicity or of extreme isoelectric
points, are difficult to separate and hence are typically
neglected, resulting in a loss of potentially interesting
proteins. Additionally, the sensitivity is low compared
with MS.
In this study, we used protein-profiling SELDI MS and
2-DE to find biomarkers that might enable earlier tumor
detection. Only a small number of protein-profiling stud-
ies in pancreas tumor have so far been performed using
SELDI technology (20 –22 ). Our study improves on this
approach by using samples of pure microdissected cells
derived from central pancreatic tumor areas, tumor mar-
gin, and adjacent healthy tissue. We detected a small
number of signals that discriminated well between the 3
sample groups. For identification of these signals, we Fig. 3. Reanalysis of IHC-treated tissue sections by ProteinChip
separated histologically checked tissue specimens from technology.
pancreatic tumors and healthy pancreatic tissue using Areas of similar size of healthy and tumorous tissue that were applied in IHC
2-DE followed by analysis of peptide mass fingerprints were microdissected and analyzed on proper ProteinChip arrays. (A), signal (*)
with a molecular mass of ⬃20 kDa representing DJ-1 was detectable in protein
using SELDI MS. We excised 95 protein spots from those lysate derived from pancreatic tumor tissue on a Q10 array. This signal is absent
that were obviously different in 2-DE gels and, using the in the protein lysate from the healthy tissue fraction. (B), signal (*) corresponding
to HSP27 was detectable in IHC-positive tissue derived from pancreatic tumor
methodology described in Melle et al. (23 ), identified 29 tissue on a CM10 array in contrast to the IHC-negative tissue from healthy
proteins unequivocally. Two proteins identified in this pancreas samples, where the signal was absent.
PARK7 gene is associated with an autosomal recessive,

early onset Parkinson disease (24 ). Recent reports show
that PARK7 is overexpressed in a number of cancer types,
including breast, lung, and prostate. It partly obtains its
transforming activity by an RNA helicase named Abstrakt
(25, 26 ). In primary breast cancer samples, DJ-1 negatively
regulates the PTEN tumor suppressor and thus produces
overexpressed hyperphosphorylation of PKB/Akt and
increased cell survival (27 ). In a proteomic analysis of
gastric cancer, DJ-1 was detectable only in metastatic
tumor tissue vs nonmetastatic tumor tissue and healthy
gastric tissue (28 ). Based on this fact, it seems likely that
DJ-1 contributes to the metastatic potential of a tumor.
HSP27 is a powerful molecular chaperone whose main
function is to prevent the aggregation of nascent and
stress-accumulated misfolded proteins. It is able to inter-
act directly with various components of the tightly regu-
lated programmed cell death machinery, upstream and
downstream of the mitochondrial events, and seems to
play a role in the proteasome-mediated degradation of
selected proteins. HSP27 is associated with poor progno-
sis in gastric, liver, and prostate carcinoma and osteosar-
Fig. 4. ROC curve for HSP27 serum concentration for patients with comas (29, 30 ). Data concerning the prognosis potential of
pancreatic carcinoma and controls. HSP27 in the above cancer types are conflicting because a
The area under the curve was 0.985 [95% confidence interval (CI) 0.965–1]. At
a cutoff of 1.33 ␮g/L (arrow), the sensitivity was 100% (CI 90.11% to 100%) and
recent study showed that in gastric cancer, HSP27 was not
specificity 84% (CI 68.86% to 92.35%). The median concentrations for patients detectable in metastatic tumors and could be found only
and controls were 2.93 and 0.76 ␮g/L, respectively. Inset, box plot of serum in samples derived from nonmetastatic tumors (28 ). To
concentrations of HSP27 for controls and pancreatic carcinoma patients for an
independent sample set. The lines inside the boxes denote the medians. The date, only a few studies are available that report an
boxes represent the interval between the 25th and 75th percentiles, and the association of differential expression of HSP27 and pan-
whiskers indicate the interval between maximum and minimum.
creatic carcinoma, and the results of these investigations
are partly conflicting (31, 32 ).
manner, DJ-1 and HSP27, matched well to different sig- Whereas our study identified DJ-1 and HSP27 as
nals found in prior protein-profiling assays. Both proteins potential new biomarkers for the early detection of pan-
were up-regulated in pancreatic cancer and discriminated creatic cancer, further studies with larger sample sizes
well between central tumor and tumor margin vs healthy using cancerous and healthy tissue acquired by noninva-
tissue and between central tumor and healthy tissue. We sive sampling methods are required.
confirmed the identities of both proteins in immuno-
depletion assays and further characterized them by im- In conclusion, we show that a proteomic procedure com-
munological techniques. In an independent set of pancre- posing tissue microdissection, protein profiling by
atic tumor tissue specimens that had not been assessed ProteinChip technology, separation and identification of
earlier by ProteinChip technology, we also detected interesting proteins by 2-DE, peptide mass fingerprinting,
strong signals corresponding to DJ-1 and HSP27 by West- and SELDI MS as well as confirmation of these proteins
ern blot analysis. We also analyzed the identified HSP27 using immunological techniques is able to identify and
specifically in serum by a corresponding ELISA, with characterize differentially expressed proteins that could
exactly the same results for HSP27 as found in prior serum markers for pancreatic carcinoma. The clinical
protein profiling. relevance of these findings will require further study.
Our evidence suggests that the concentration of DJ-1 is
increased in pancreatic carcinomas and that this increase
distinguishes pancreatic tumor tissue from adjacent
healthy tissue. DJ-1 is a conserved protein, coded by the This work was supported by a grant of the German
gene PARK7 (Parkinson disease 7),6 that is reported to be Federal Ministry of Education and Research and the
involved in diverse cellular processes including cellular Interdisciplinary Center for Clinical Research, Jena,
transformation, control of protein–RNA interaction, oxi- Germany.
dative stress response, and control of male infertility. The
References
1. Jemal A, Murray T, Ward E, Samuels A, Tiwari RC, Ghafoor A, et al.
6
Human gene: PARK7, Parkinson disease 7. Cancer statistics, 2005. CA Cancer J Clin 2005;55:10 –30.
2. Goggins M. Molecular markers of early pancreatic cancer. J Clin 17. Yu KH, Rustgi AK, Blair IA. Characterization of proteins in human
Oncol 2005;23:4524 –31. pancreatic cancer serum using differential gel electrophoresis and
3. Tang N, Tornatore P, Weinberger SR. Current developments in tandem mass spectrometry. J Proteome Res 2005;4:1742–51.
SELDI affinity technology. Mass Spectrom Rev 2004;23:34 – 44. 18. Chen R, Yi EC, Donohoe S, Pan S, Eng J, Cooke K, et al. Pancreatic
4. Melle C, Ernst G, Schimmel B, Bleul A, Thieme H, Kaufmann R, et cancer proteome: the proteins that underlie invasion, metastasis,
al. Discovery and identification of alpha-defensins as low abun- and immunologic escape. Gastroenterology 2005;129:1187–97.
dant, tumor-derived serum markers in colorectal cancer. Gastro- 19. Shekouh AR, Thompson CC, Prime W, Campbell F, Hamlett J,
enterology 2005;129:66 –73. Herrington CS, et al. Application of laser capture microdissection
5. Wright GL, Cazares LH, Leung SM, Nasim S, Adam BL, Yip TT, et combined with two-dimensional electrophoresis for the discovery
al. Proteinchip® surface enhanced laser desorption/ionization of differentially regulated proteins in pancreatic ductal adenocar-
(SELDI) mass spectrometry: a novel protein biochip technology for cinoma. Proteomics 2003;3:1988 –2001.
detection of prostate cancer biomarkers in complex protein mix- 20. Rosty C, Christa L, Kuzdzal S, Baldwin WM, Zahurak ML, Carnot F,
tures. Prostate Cancer Prostatic Dis 1999;2:264 –76. et al. Identification of hepatocarcinoma-intestine-pancreas/pan-
6. von Eggeling F, Davies H, Lomas L, Fiedler W, Junker K, Claussen creatitis-associated protein I as a biomarker for pancreatic ductal
U, et al. Tissue-specific microdissection coupled with ProteinChip adenocarcinoma by protein biochip technology. Cancer Res 2002;
array technologies: applications in cancer research. Biotech- 62:1868 –75.
niques 2000;1066 –70. 21. Koopmann J, Zhang Z, White N, Rosenzweig J, Fedarko N,
7. Cazares LH, Adam BL, Ward MD, Nasim S, Schellhammer PF, Jagannath S, et al. Serum diagnosis of pancreatic adenocarci-
Semmes OJ, et al. Normal, benign, preneoplastic, and malignant noma using surface-enhanced laser desorption and ionization
prostate cells have distinct protein expression profiles resolved by mass spectrometry. Clin Cancer Res 2004;10:860 – 8.
surface enhanced laser desorption/ionization mass spectrometry. 22. Honda K, Hayashida Y, Umaki T, Okusaka T, Kosuge T, Kikuchi S,
Clin Cancer Res 2002;8:2541–52. et al. Possible detection of pancreatic cancer by plasma protein
8. Escher N, Spies-Weisshart B, Kaatz M, Melle C, Bleul A, Driesch D, profiling. Cancer Res 2005;65:10613–22.
et al. Identification of HNP3 as a tumour marker in CD4⫹ and 23. Melle C, Osterloh D, Ernst G, Schimmel B, Bleul A, von Eggeling F.
CD4⫺ lymphocytes of patients with cutaneous T-cell lymphoma. Identification of proteins from colorectal cancer tissue by two-
Eur J Cancer 2006;42:249 –5 dimensional gel electrophoresis and SELDI mass spectrometry.
9. Cheung PK, Woolcock B, Adomat H, Sutcliffe M, Bainbridge TC, Int J Mol Med 2005;16:11–7.
Jones EC, et al. Protein profiling of microdissected prostate tissue
24. Hod Y. Differential control of apoptosis by DJ-1 in prostate benign
links growth differentiation factor 15 to prostate carcinogenesis.
and cancer cells. J Cell Biochem 2004;92:1221–33.
Cancer Res 2004;64:5929 –33.
25. Nagakubo D, Taira T, Kitaura H, Ikeda M, Tamai K, Iguchi-Ariga
10. Kwapiszewska G, Meyer M, Bogumil R, Bohle RM, Seeger W,
SM, et al. DJ-1, a novel oncogene which transforms mouse
Weissmann N, et al. Identification of proteins in laser-microdis-
NIH3T3 cells in cooperation with ras. Biochem Biophys Res
sected small cell numbers by SELDI-TOF and Tandem MS. BMC
Commun 1997;231:509 –13.
Biotechnol 2004;4:30.
26. Sekito A, Taira T, Niki T, Iguchi-Ariga SM, Ariga H. Stimulation of
11. Melle C, Ernst G, Schimmel B, Bleul A, Kaufmann R, Hommann M,
transforming activity of DJ-1 by Abstrakt, a DJ-1-binding protein. Int
et al. Characterization of pepsinogen C as a potential biomarker
J Oncol 2005;26:685–9.
for gastric cancer using a histo-proteomic approach. J Proteome
Res 2005;4:1799 – 804. 27. Kim RH, Peters M, Jang Y, Shi W, Pintilie M, Fletcher GC, et al.
12. Melle C, Ernst G, Schimmel B, Bleul A, Koscielny S, Wiesner A, et DJ-1, a novel regulator of the tumor suppressor PTEN. Cancer Cell
al. Biomarker discovery and identification in laser microdissected 2005;7:263–73.
head and neck squamous cell carcinoma with ProteinChip® tech- 28. Chen J, Kahne T, Rocken C, Gotze T, Yu J, Sung JJ, et al. Proteome
nology, two-dimensional gel electrophoresis, tandem mass spec- analysis of gastric cancer metastasis by two-dimensional gel
trometry, and immunohistochemistry. Mol Cell Proteomics 2003; electrophoresis and matrix assisted laser desorption/ionization-
2:443–52. mass spectrometry for identification of metastasis-related pro-
13. van’t Veer LJ, Dai HY, van de Vijver MJ, He YDD, Hart AAM, Mao teins. J Proteome Res 2004;3:1009 –16.
M, et al. Gene expression profiling predicts clinical outcome of 29. Garrido C, Schmitt E, Cande C, Vahsen N, Parcellier A, Kroemer G.
breast cancer. Nature 2002;415:530 – 6. HSP27 and HSP70: potentially oncogenic apoptosis inhibitors.
14. Sidransky D. Emerging molecular markers of cancer. Nat Rev Cell Cycle 2003;2:579 – 84.
Cancer 2002;2:210 –9. 30. Ciocca DR, Calderwood SK. Heat shock proteins in cancer:
15. Fukushima N, Koopmann J, Sato N, Prasad N, Carvalho R, Leach diagnostic, prognostic, predictive, and treatment implications.
SD, et al. Gene expression alterations in the non-neoplastic Cell Stress Chaperones 2005;10:86 –103.
parenchyma adjacent to infiltrating pancreatic ductal adenocarci- 31. Gangarosa LM, Grove PS, Chang WWL. Heat shock proteins
noma. Mod Pathol 2005;18:779 – 87. and pancreatic adenocarcinoma. Gastroenterology 1999;116:
16. Segara D, Biankin AV, Kench JG, Langusch CC, Dawson AC, A1127.
Skalicky DA, et al. Expression of HOXB2, a retinoic acid signaling 32. Lu Z, Hu L, Evers S, Chen J, Shen Y. Differential expression
target in pancreatic cancer and pancreatic intraepithelial neopla- profiling of human pancreatic adenocarcinoma and healthy pan-
sia. Clin Cancer Res 2005;11:3587–96. creatic tissue. Proteomics 2004;4:3975– 88.
Protein Markers
Protein Expression Patterns Associated with

Progression of Chronic Obstructive Pulmonary
Disease in Bronchoalveolar Lavage of Smokers
Amelie Plymoth,1* Claes-Göran Löfdahl,1 Ann Ekberg-Jansson,2 Magnus Dahlbäck,3
Per Broberg,4 Martyn Foster,5 Thomas E. Fehniger,4 and György Marko-Varga4
Background: We modeled the expression of proteins in term smokers may be detected in smokers susceptible to
baseline bronchoalveolar lavage (BAL) samples from a progression of COPD disease, before disease is clini-
asymptomatic 60-year-old lifelong current smokers or cally evident.
healthy never-smokers, who were reevaluated after 6 to © 2007 American Association for Clinical Chemistry
7 years to record clinical outcome.
Methods: Applying a technology toolbox consisting of Chronic cigarette smoking is a major cause of lethal
replicate 2-dimensional gel separations, image annota- diseases such as cancers, cardiovascular diseases, pulmo-
tion, and mass spectrometry identification, we cata- nary hypertension, chronic obstructive pulmonary dis-
logued a global set of proteins that were differentially ease (COPD), and pneumonia and harms nearly every
expressed in individuals by presence, absence, and organ of the body. Cigarette smoke contains more than
intensity scores. 4000 substances, many of which have a toxic effect on cells
Results: By use of multivariate analysis, we selected a (1 ). Within the respiratory tract, in addition to the xeno-
subset of proteins that accurately separated smokers biotic effects associated with oncogene expression and
from never-smokers based on composite scoring. Fol- malignant transformation, smoking drives a chronic in-
low-up after 6 to 7 years identified a group of individ- flammatory state resulting in the release of mediators
uals who had progressed to chronic obstructive pulmo- such as proteases that cause the destruction of ground
nary disease (COPD), Global Initiative for Chronic matrix and tissue components (2 ).
Obstructive Lung Disease stage 2. The baseline BAL Little information is available regarding the effects of
samples of these eventual COPD patients shared a smoking on the proteome of the respiratory tract, and
distinct protein expression profile that could be identi- only a few studies have directly compared global patterns
fied using partial least-squares discriminant analysis. of protein expression in the respiratory tract of smokers
This pattern was not observed in BAL samples of and nonsmokers. To date, 2-dimensional polyacrylamide
asymptomatic smokers free of COPD at 6- to 7-year gel electrophoresis (2-D PAGE) and liquid chromatogra-
follow-up. phy platforms have been used to study clinical samples
Conclusions: Our model suggests that certain patterns including bronchoalveolar lavage (BAL), nasal lavage,
of protein expression occurring in the airways of long- and whole lung tissue (3– 6 ). These studies have shown
that smokers typically show increased concentrations of
inflammatory and redox proteins, as well as decreases in
1
Department of Respiratory Medicine and Allergology, Lund University
surfactants and Clara cell secretory proteins, compared
Hospital, Lund, Sweden. with nonsmokers. These observations are in line with
2
Sahlgrenska University Hospital, Gothenburg, Sweden.
3
Department of Discovery Medicine and Epidemiology, AstraZeneca,
Lund, Sweden.
4 6
Department of Biological Sciences, AstraZeneca R&D, Lund, Sweden. Nonstandard abbreviations: COPD, chronic obstructive pulmonary dis-
5
Department of Safety Assessment, AstraZeneca R&D, Loughborrough, ease; 2-D PAGE, two-dimensional polyacrylamide gel electrophoresis; BAL,
United Kingdom. bronchoalveolar lavage; MS, mass spectrometry; FEV1, forced expiratory
* Address correspondence to this author at: Fax ⫹46-46-146793; e-mail volume in 1 s; DLCO, carbon monoxide lung diffusion capacity; GOLD, Global
amelie.plymoth@med.lu.se. Initiative for Chronic Obstructive Lung Disease; SSP, standard spot number;
Received July 7, 2006; accepted January 18, 2007. PCA, principle component analysis; and PLS-DA, partial least-squares dis-
Previously published online at DOI: 10.1373/clinchem.2006.076075 criminant analysis.
636
previous studies showing that gene expression patterns

Table 1. Changes of lung function measurements of
are permanently altered in bronchial epithelial cells and in
smokers at follow-up.
the lung tissue of smokers who develop COPD (7, 8 ). In a
FEV1/FVC FEV1, % predicted
recent study of BAL using liquid chromatography cou-
pled with on-line linear ion trap quadrupole mass spec- COPD patient no. Visit 1 Visit 2 Visit 1 Visit 2 ⌬a
trometry (MS), we identified 481 high-to-low abundant 243 70 61 94 70 ⫺24

proteins (6 ). These proteins were differentially expressed 301 61 46 86 64 ⫺22
in BAL samples obtained from age-matched smokers 326 69 62 84 64 ⫺20
501 67 65 67 66 ⫺1
compared with never-smokers.
621 64 54 88 64 ⫺24
In this study, we further characterized the global
817 64 53 83 79 ⫺4
proteome of the BAL samples from the smokers and
855 58 54 73 62 ⫺11.0
never-smokers studied above. Mean 65 56 82 67 ⫺15.1
Asymptomatic smokers
Materials and Methods Mean
study population 76 73 98 94 ⫺3.4
We recruited 60-year-old men from the randomized pop- The initial study (visit 1) took place during February 1994 to July 1995. The
ulation study “Men born in Göteborg in 1933” [see Fig. 1 follow-up examinations (visit 2) took place in 2000 and 2001. Seven of the
in the Data Supplement that accompanies the online version smokers developed moderate COPD and 13 of the smokers remained asymp-
of this article at http://www.clinchem.org/content/vol53/ tomatic at follow-up. a Numerical difference: visit 2–visit 1. FVC, forced vital
capacity.
issue4] (9 –11 ); a subset of the study population of 879
individuals volunteered to undergo fiber-optic bronchos-
copy at the age of 60. This group included 47 men: 29
asymptomatic chronic smokers (14 light and 15 heavy significant airway obstruction as seen in the decline in
smokers) and 18 healthy never-smokers. At the time of FEV1/forced vital capacity quotients and FEV1 (% pre-
BAL sampling, all men in the study were evaluated dicted) scores, compared with the other smokers.
clinically as being healthy, because they had not sought
medical attention for respiratory disease. Total lung ca-
pacity values did not differ significantly between smokers bal sampling
and never-smokers (Fig. 1); however, there was a trend The studies were approved by the ethics committees of
for decreased lung function values in both the light and the Sahlgrenska Hospital (Gbg M 117-01) and Lund
heavy smokers according to forced expiratory volume in University Hospital (LU 689 – 01) in Sweden. Participants
1 s (FEV1) and carbon monoxide lung diffusion capacity provided informed consent. BAL sampling was carried
(DLCO). out under typical conditions (10 ), and the BAL fluid was
At 6 to 7 years after the original BAL sampling, the immediately transported on ice to the laboratory for
study cohort was asked to participate in a follow-up. processing and storage at ⫺80 °C (10 ). The protein con-
Seven of the 29 smokers had developed moderate COPD centrations in the recovered BAL were measured by use of
[Global Initiative for Chronic Obstructive Lung Disease Coomassie plus protein assay reagent (Pierce) with bo-
(GOLD) stage 2] (12, 13 ). None of the never-smokers had vine serum albumin as a reference (14 ). The total protein
developed either mild or moderate COPD. Three partici- concentration of the recovered BAL samples was not
pants had died of nonrespiratory complications. As significantly different between individuals, but recovery
shown in Table 1, the patients who developed COPD had of BAL was lower in the smoking cohort (6 ).
Fig. 1. Lung function measurements of

the never-smokers and smokers from
the initial study.
The lung function measurements show no
significant difference in total lung capacity
(TLC) values between smokers and never-
smokers; however, there is a trend for de-
creased lung function values in both the light
and the heavy smokers in the measurements
of FEV1 and DLCO. *, **The definition of light
smokers was 1–15 cigarettes smoked/day
with a median number of 22 pack-years
(range 9 – 45); heavy smokers, ⱖ15 ciga-
rettes smoked/day with a median number of
45 pack-years (range 31–79). Pack-years are
calculated by multiplying the number of packs
of cigarettes smoked per day by the number
of years the person has smoked.
638 Plymoth et al.: Protein Expression Patterns Associated with COPD
tissue sampling and histology (a) reduce variability in sample selection, collection, and
We took peripheral bronchial biopsy specimens from each preparation and (b) address the range of biological varia-
of the groups with an alligator forceps from the main tions in clinical presentation for comparison to protein
carina between the right and left lung as has been expression patterns.
described (15 ). The sections were stained using conven- We were able to evaluate approximately two thirds of
tional hematoxylin and eosin staining. the bronchial tissue biopsies. Two prominent histological
features, the condition of the epithelium and the degree
2-d gel electrophoresis of inflammation, distinguished the mucosal compartment
We generated 2-D PAGE protein maps according to our of never-smokers from smokers. As an example, com-
previously developed procedure (16 ). BAL fluid contain- pared to never-smokers [see Fig. 2A in the online Data
ing 100 ␮g protein was desalted by ultramembrane filtra- Supplement], smokers as a group showed a broad phe-
tion and loaded using a pI 4 –7 immobilized pH gradient notypic variation of the bronchial epithelium ranging
strip. We performed quantitative analysis with the image from hyperplasia, metaplasia, and hypertrophy [see Fig.
software PDQuest™ version 6.0 (Bio-Rad) (16, 17 ). 2B in the online Data Supplement] to sloughing and
Reference gels for each group were synthesized by regeneration [see Fig. 2C in the online Data Supplement].
hierarchic clustering of images from the individual mem- Current smokers generally showed increased signs of
ber gels of each group and combined to create a synthetic diapedesis and inflammation throughout the submucosa
master gel of all protein spots. For further description, see [see Fig. 2D in the online Data Supplement] and the elastic
text in the online Data Supplement. lamina [see Fig. 2C in the online Data Supplement], as
reported (15 ). Both lung function measurements and
standard spot number presence rates histology evaluations therefore indicated that the range of
To associate standard spot number (SSP) attributes be- features observed in the current-smoker groups were in
tween the never-smoking and smoking groups, we first variance with the never-smokers. This suggests ongoing
filtered the global protein set to exclude proteins that episodes of pathogenesis despite the lack of clinical com-
were not frequently expressed by either group. We scored plaints or symptoms.
each gel image for each SSP of the global set of 944
identifications in terms of presence, absence, and aggre- global protein expression maps and
gate pixel density. annotations
Using a stringency cutoff demanding at least 30% We measured global protein expression patterns in BAL
presence of a given SSP in any single group, we analyzed samples separated by 2-D PAGE. In group comparisons,
818 SSP identifications in detail. We calculated mean distributions of BAL proteins observed on raw gels from
scores of SSP presence/absence and aggregate density for never-smokers (Fig. 2A, B) varied qualitatively and quan-
each group. These means were further used to compare titatively from protein patterns observed for light smokers
the relative differences in expression levels between the (Fig. 2C, D) and heavy smokers (Fig. 2E, F). Both the
groups by dividing the mean SSP density of the smoking never-smoking group and the smoking groups expressed
group by the SSP density of the never-smoking cohort. proteins in their BAL samples that were absent or not
frequently observed in the respective group gels. We
statistical calculations synthesized reference gels for each group (Fig. 2G never-
We performed principal components analysis (PCA) and smokers, Fig. 2H heavy smokers) by hierarchic clustering
partial least-squares discriminant analysis (PLS-DA) with of images from the individual member gels of each group.
Simca-P version 10.0.4 (Umetrics AB). PCA was used for All individual gels were combined to create a synthetic
unsupervised views of the data. PLS-DA was used to master gel of all protein spots detected (Fig. 2I). The
build classifiers of clinical category (18 ). The 2 methods synthetic master gel of all 47 samples contained a global
provide 1 measure, R2, of model fit, and another, Q2, of set of 944 SSP identifications. The zoomed image of a
the predictability of the model. representative gel region [see Fig. 3A in the online Data
In the hierarchical clustering method, data were nor- Supplement], displaying medium-to-low abundant pro-
malized by subtracting the mean intensity and dividing tein spot expression levels, reveals a high degree of
by the SD separately for each SSP. For further description, annotation differences between smokers [see Fig. 3, B
see text in the online Data Supplement. and C in the online Data Supplement] and never-smokers
[see Fig. 3, D and E in the online Data Supplement].
Results Software rendering of pixel density (z-axis) into 3-D
strategy for sample standardization and image profiles improves visualization of the relative
clinical characterization presence/absence.
To identify and model individual variations in the protein We constructed a model to compare differences be-
expression profiles present in complex clinical samples tween the smoking and never-smoking SSP expression
such as bronchoalveolar lavage, we adopted a strategy to patterns, applying differing set points that spanned be-
Fig. 2. Representative 2-D gels of the BAL proteome from never-smokers and smokers.
The distribution of BAL proteins on the raw gels from never-smokers (A, B) varies qualitatively and quantitatively from the protein patterns for light smokers (C, D) and
heavy smokers (E, F). Reference gels for each group (G, never-smokers; H, heavy smokers) are synthesized by hierarchic clustering of images from the individual
member gels of each group. (I) All individual gels combined to create a synthetic master gel of all protein spots detected.
tween the ⱖ30% and ⱖ90% presence rates. Several ques- both never-smokers and smokers. The expression densi-
tions were posed: What proportion of SSP identities is ties decreased to between ⫺4- and ⫹4-fold differences of
regulated in smokers compared with never-smokers? the mean density ratios for smokers compared with
What is the relationship between presence rate and fold never-smokers (Fig. 3B). Approximately 2% of the SSP
expression factor? Are certain sets of SSPs associated with identities were down-regulated in smokers (density ratio
smoking? ⬍– 4). Among the remaining 13% protein identities, we
Fig. 3A displays a distribution plot that maps SSP observed a bimodal, highly induced expression pattern in
count (y-axis), relative fold differences in individual SSP the smoking group. This relative trend was maintained
expression levels between smokers and never-smokers over the interval of low-to-high presence rates.
(x-axis), and individual SSP distribution at various points
of presence between 30% (n ⫽ 818 SSPs) and ⱖ90% (n ⫽ statistical distributions and separations of
189 SSPs) (z-axis). The fold-change interval in the model ssp profiles
was ⫺20 ⱕX ⱕ20, where X ⬍⫺20 is set to ⫺20 and X ⬎20 We found that each individual expressed a partial subset
is set to 20 to capture the observed variances in mean SSP of the global SSP set, in part because of the intrinsic
expression in both groups. Very few SSP identities (n ⫽ biological variations observed between groups as de-
69) were present in 100% of individuals in either of the scribed by our independent measurements, such as histo-
groups. logical examination of bronchial biopsies (see Results and
By use of the model, we could easily define modal the online Data Supplement). Conventional nonparamet-
distributions of protein expression at the level of individ- ric statistical comparisons of the stringency cutoff set at
ual SSPs. This enabled us to categorize expressions as 70% presence rate (406 SSPs), using parameter presence,
up-regulated, down-regulated, or not regulated when we absence, and abundance scoring, showed that significant
compared smoking groups to the never-smoking group. differences in protein expression patterns occurred be-
Irrespective of the presence rate, ⬃85% of the SSP anno- tween never-smokers and smokers (Mann–Whitney P
tation identities showed similar expression patterns in ⬍0.001). Separation of heavy smokers from light smokers
Fig. 3. Model to compare differences

between never-smoking and smoking
BAL proteome expression patterns.
A, Distribution plot that maps SSP count
(y-axis), relative fold differences in individual
SSP expression levels between smokers and
never-smokers (x-axis), and individual SSP
distribution at various points of presence
between 30% (n ⫽ 818 SSPs) and ⱖ90%
(n ⫽ 189 SSPs) (z-axis). B, 85% of the SSP
identities show similar expression patterns
in both never-smokers and smokers; 2% of
the SSP identities are down-regulated in
smokers (density ratio ⬍– 4), and 13% are
highly induced in the smokers (density ratio
⬎4).
based on individual SSP scores required the application of heavy smokers (red). In this analysis, there is some degree
further statistical tools. of spontaneous separation; in particular, the never-smok-
We applied PCA and PLS-DA to the SSP dataset ers stand out as a group.
parameters. Using PCA analysis of 406 protein identities Having established the group dynamics of the dataset
expressed by at least 70% of the never-smoking or smok- of 406 SSPs, we tested at what level of predictive accuracy
ing cohorts, we unambiguously (P ⬍0.001) and accurately we could assign any given individual to a specific group
(R2 ⫽ 0.84) separated smokers from never-smokers based designation based solely on their individual SSP profile.
on composite protein expression phenotype (Fig. 4A). A The Q2 measures of group predictability were 0.78, 0.54,
progression of dimensional features obliquely separates and 0.69, respectively, for never, light, and heavy smok-
the never-smokers (green) from light smokers (blue) to ers. In this analysis, 1.00 defines perfect prediction. These
Fig. 4. Unsupervised and supervised analysis of the BAL

protein identities expressed by the never-smoking or smoking
cohorts.
(A), Principal component analysis of the 406 BAL protein identities
expressed by at least 70% of either the never-smoking or smoking
cohorts separates the never-smokers (green) from the light smokers
(blue) and the heavy smokers (red). The encircled patients were found
to have developed moderate COPD in the follow-up study. (B), by
reanalyzing the smokers using successive rounds of PLS-DA analysis,
the Q2 scores for predicting the COPD clinical outcome rise significantly
from a negative value using all 406 protein identities to a predictability
of 0.8 applying 100 protein identities. (C), PLS-DA on this set of 100
SSP identities separates all 7 patients who developed moderate COPD
(black) from the other smokers (red) by both T1 and T2 dimensions. The
ellipse represents the 95% confidence region based on Hotelling T2.
results validated that the qualifiers used as dimensions in because of its utility to address secreted and extracellular
these comparisons of SSP datasets were significantly proteins present within the central and descending air-
related to a specific group character. ways of smokers and never-smokers. The technology we
adopted has general applicability to a wide variety of
clinical outcome at 6 to 7 years follow-up clinical samples, including plasma, serum, urine, sputum,
Seven of the 29 light and heavy smokers were found to and other solubilized tissue components such as targeted
have developed moderate COPD (GOLD stage 2, Fig. 4A, cells obtained by laser capture microscopy.
encircled) in the follow-up study. None of the never- Using standardized 2-D gel technology, we observed
smokers developed either mild or moderate COPD. Using that the protein components present in BAL are variably
the set of 406 SSPs expressed at a 70% presence rate, we expressed in individuals. The set of SSP identities found
were unable to separate the protein expression profiles of in each BAL sample was quantitatively scored to establish
the 7 COPD patients from the other 22 heavy or light hierarchal relationships between different SSPs for each of
smokers using the Q2 predictability measure (R2 ⬍ 0.8). the 2 groups. Multivariate analysis provided us with the
We then tested whether a subset of the 406 SSPs might be means to relate not only singular SSP characteristics such
more significantly associated with clinical outcome by as presence and relative expression level values in com-
reanalyzing the 29 smokers by successive rounds of parison to group behavior, but also the same parameters
PLS-DA analysis, utilizing 200, 100, 50, 25, or 10 SSP to all of the other SSPs in the dataset.
components. As shown in Fig. 4B, the Q2 scores for We then asked what independent predictors among
predicting the COPD clinical outcome rise significantly all SSPs characterize a particular group. By using pres-
from a negative value using all 406 components to a high ence rate stringency rules that could be applied indepen-
predictability of 0.8 applying 100 SSPs. We analyzed the dently to one or another of the study groups, we allowed
expression profiles of all 29 smokers using PLS-DA anal- for extreme changes in absolute factors of expression. We
ysis on this set of 100 SSP identities. We found that all 7 found that smoking individuals lost the expression of a
COPD progressors could be well separated from the other small subset of identities present in never-smokers and
smokers by both T1 and T2 dimensions (Fig. 4C). acquired a high level of expression of SSP identities ab-
We further analyzed the set of 100 differentially ex- sent in never-smokers. We thus demonstrated that lifelong
pressed SSP identities using 2-D hierarchal clustering smokers develop significantly different protein expression
(Fig. 5). Our application reveals order both among the patterns in their respiratory tract from those of age-matched
samples and within the SSPs. The 7 eventual COPD never-smokers.
patients (indicated by blue arrows) segregated together Diagnostic datasets are often composed of groups of
on the left side of the dendrogram except for 1 individual signals representing the various transition states of nu-
(patient 855) whose clinical characteristics were not dis- merous singular proteins. Because of the broad dynamic
similar from those of the other patients (Table 1). range of protein concentrations present in any given
sample volume, it is beneficial to take advantage of
Discussion whatever associations of regularity exist between individ-
In this report, we present a strategy for relating the ual protein identities, sets of proteins within given clinical
measurement of sets of proteins within a clinical sample samples, or given clinical presentations. To accomplish
with phenotypes of clinical presentation. To accomplish this degree of segregation, it is necessary to link a specific
this, we used an interdisciplinary toolbox that included phenotype of the clinical presentation to that exclusive
protein separation, identification platforms, and statistical protein set. However, validation of models based on
methods for multivariate analysis. These methods al- associating proteins to disease can best be achieved by
lowed us to group the men in the study by individual relating these values to clinical outcome.
presence/absence and intensity scores for each of the Because the study material was collected in 1993, when
differentiated protein spot annotations. the asymptomatic men were 60 years old (and many of
The key component for pursuing such a study was the the participants agreed to be clinically evaluated at fur-
careful selection of a clinical material that could be ther time points), we were able to pose questions relating
normalized for study, in terms of being associated with eventual clinical outcome to the status of BAL protein
the biology of interest and also for sampling variability. expression 6 –7 years earlier. Over this time period, 7 of
Our study material is unique: age-matched men all born the smokers had clinically progressed to moderate
in 1933, living in 1 city, differing by lifelong smoking COPD (GOLD stage 2). We were unable to segregate this
history, and compared by clinical function measurements COPD subset from the other smokers using the PLS-DA
and histological assessment at the same relative time model based on the presence/absence and intensity
points. By reducing the variability of the input samples, scores of the global set of 406 proteins expressed at the
we could concentrate on refining the technology and stringency rate of 70% presence. As a cross-validation for
analysis platforms to standardize the quantification and further analysis, we applied repetitive cycles of filtering
normalization methods in support of our need for an on this global set to obtain a limited protein set shared by
unbiased assessment. We have chosen to analyze BAL the eventual COPD patients, yet not observed by the other
Fig. 5. Hierarchical clustering of the set of 100 differentially expressed SSP identities from the 29 smokers from Fig. 4B and C.
The x-axis shows the similarity by individual to individual and the y-axis by the expression behavior characteristics of the individual SSP identities. The progressing COPD
patients (arrows) segregate together on the left side of the dendrogram except for 1 individual (patient 855). The color scale describing low-to-high expressions (green
to red) clearly separates into 4 main trees and branches described on the horizontal (A, B) and vertical (C, D) axes. Individuals progressing to COPD are indicated with
blue arrows.
lifelong smokers. In particular, the hierarchal clustering of protein expressions in BAL, as defined by presence,
analysis of 100 identities by presence/absence and inten- absence, and intensity, can be associated with different
sity scores independently clustered 6 of the 7 participants rates of disease progression in individuals sharing risk
in 1 quadrant, demonstrating the commonality in protein behavior for eventual disease development, such as age,
expression for this subset of eventual COPD patients smoking, and geographical location.
compared with the other smokers. Thus, the set of 100 SSP The traditional approach for characterizing proteomes
identities associated with COPD outcome was shown to is to use reductional methods that identify and charac-
be composed of statistically related proteins that occurred terize the individual components by MS and then differ-
at either much lower or much higher concentrations than entially relate that proteome to a biological question or
in the other smokers. We conclude that certain patterns endpoint. Individual MS identities act as important an-
notation landmarks for comparison within or between scale liquid chromatography and mass spectrometry. Clin Chem
studies but also for establishing a biological context to the 2006;52:671–9.
model. Our results indicate that the information com- 7. Spira A, Beane J, Shah V, Liu G, Schembri F, Yang X, et al. Effects
of cigarette smoke on the human airway epithelial cell transcrip-
bined in large datasets of individual SSP presence/
tome. Proc Natl Acad Sci U S A 2004;101:10143– 8.
absence and abundance scores has sufficient power to 8. Ning W, Li CJ, Kaminski N, Feghali-Bostwick CA, Alber SM, Di YP,
accurately predict, identify, and assign a stable pheno- et al. Comprehensive gene expression profiles reveal pathways
type. By use of multivariate algorithms, we showed that related to the pathogenesis of chronic obstructive pulmonary
both study cohorts maintained a high degree of overall disease. Proc Natl Acad Sci U S A 2004;101:14895–900.
predictability of group association. This result is impor- 9. Rosengren A, Orth-Gomer K, Wedel H, Wilhelmsen L. Stressful life
tant, as it implies that stable predictive phenotypes of events, social support, and mortality in men born in 1933. [see
comment]. BMJ 1993;307:1102–5.
protein expression can be determined and used to segre- 10. Ekberg-Jansson A, Andersson B, Bake B, Boijsen M, Enander I,
gate individuals even in clinical cohorts that consist of Rosengren A, et al. Neutrophil-associated activation markers in
asymptomatic and clinically healthy participants. This healthy smokers relates to a fall in DL(CO) and to emphysematous
segregation can be accomplished despite the requirement changes on high resolution CT. Respir Med 2001;95:363–73.
of assigning actual protein identities to each of the com- 11. Vikgren J, Boijsen M, Andelid K, Ekberg-Jansson A, Larsson S,
ponents within the clinical proteome. Bake B, et al. High-resolution computed tomography in healthy
smokers and never-smokers: a 6-year follow-up study of men born
in 1933. [erratum appears in Acta Radiol. 2004 May;45(3):
following 663]. Acta Radiol 2004;45:44 –52.
The study was supported by the Swedish Heart Lung 12. Global IfCOLD. www.goldcopd.com (Accessed September 2006).
Foundation. 13. Pauwels RA, Buist AS, Ma P, Jenkins CR, Hurd SS. Global strategy
for the diagnosis, management, and prevention of chronic ob-
References structive pulmonary disease: National Heart, Lung, and Blood
1. Services USDoHaH. The Health Consequences of Smoking: A Institute and World Health Organization Global Initiative for
Report of the Surgeon General. Vol. Atlanta, Georgia: U.S. Depart- Chronic Obstructive Lung Disease (GOLD): executive summary.
ment of Health and Human Services, Centers for Disease Control Respir Care 2001;46:798 – 825.
and Prevention, National Center for Chronic Disease Prevention 14. Bradford MM. A rapid and sensitive method for the quantitation of
and Health Promotion, Office on Smoking and Health, 2004b. microgram quantities of protein utilizing the principle of protein-
dye binding. Anal Biochem 1976;72:248 –54.
2. Churg A, Wright JL. Proteases and emphysema. Curr Opin Pulm
15. Amin K, Ekberg-Jansson A, Lofdahl CG, Venge P. Relationship
Med 2005;11:153–9.
between inflammatory cells and structural changes in the lungs of
3. Wattiez R, Falmagne P. Proteomics of bronchoalveolar lavage
asymptomatic and never smokers: a biopsy study. Thorax 2003;
fluid. J Chromatogr B Analyt Technol Biomed Life Sci 2005;815:
58:135– 42.
169 –78.
16. Plymoth A, Lofdahl CG, Ekberg-Jansson A, Dahlback M, Lindberg
4. Lindahl M, Ekstrom T, Sorensen J, Tagesson C. Two dimensional H, Fehniger TE, et al. Human bronchoalveolar lavage: biofluid
protein patterns of bronchoalveolar lavage fluid from non-smok- analysis with special emphasis on sample preparation. Proteom-
ers, smokers, and subjects exposed to asbestos. Thorax 1996; ics 2003;3:962–72.
51:1028 –35. 17. Malmstrom J, Larsen K, Hansson L, Lofdahl CG, Norregard-Jensen
5. Marko-Varga G, Lindberg H, Lofdahl CG, Jonsson P, Hansson L, O, Marko-Varga G, et al. Proteoglycan and proteome profiling of
Dahlback M, et al. Discovery of biomarker candidates within central human pulmonary fibrotic tissue utilizing miniaturized
disease by protein profiling: principles and concepts. J Proteome sample preparation: a feasibility study. Proteomics 2002;2:394 –
Res 2005;4:1200 –12. 404.
6. Plymoth A, Yang Z, Lofdahl CG, Ekberg-Jansson A, Dahlback M, 18. Eriksson J, Chait BT, Fenyo D. A statistical basis for testing the
Fehniger TE, et al. Rapid proteome analysis of bronchoalveolar significance of mass spectrometric protein identification results.
lavage samples of lifelong smokers and never-smokers by micro- Anal Chem 2000;72:999 –1005.
645– 656 (2007) Proteomics and
Protein Markers
Preanalytic Influence of Sample Handling on

SELDI-TOF Serum Protein Profiles
John F. Timms,1,2* Elif Arslan-Low,1 Aleksandra Gentry-Maharaj,1 Zhiyuan Luo,3
Davy T’Jampens,4 Vladimir N. Podust,4 Jeremy Ford,1 Eric T. Fung,4 Alex Gammerman,3
Ian Jacobs,1 and Usha Menon1
Background: High-throughput proteomic methods for there was little effect of clotting time, storage method, or
disease biomarker discovery in human serum are prom- transit time. Certain proteins (TTR, ApoCI, and trans-
ising, but concerns exist regarding reproducibility of ferrin) were unaffected by handling, but others (ITIH4
results and variability introduced by sample handling. and hemoglobin ␤) displayed significant variability.
This study investigated the influence of different pre- Conclusions: Changes in preanalytical handling vari-
analytic handling methods on surface-enhanced laser ables affect profiles of serum proteins, including pro-
desorption/ionization time-of-flight mass spectrometry posed disease biomarkers. Proteomic analysis of sam-
(SELDI-TOF MS) protein profiles of prefractionated ples from serum banks collected using less stringent
serum. We investigated whether older collections with protocols is applicable if all samples are handled
longer sample transit times yield useful protein profiles, identically.
and sought to establish the most feasible collection © 2007 American Association for Clinical Chemistry
methods for future clinical proteomic studies.
Methods: To examine the effect of tube type, clotting Analysis of serum proteomes holds great promise for
time, transport/incubation time, temperature, and stor- identifying novel cancer markers for screening, diagnosis,
age method on protein profiles, we used 6 different and prognosis (1–3 ). Most patients find venipuncture
handling methods to collect sera from 25 healthy volun- tolerable, and it is standard practice to monitor disease
teers. We used a high-throughput, prefractionation strat- progression or response to therapy by collecting serial
egy to generate anion-exchange fractions and examined blood samples. Recently, various proteomics-based ap-
their protein profiles on CM10, IMAC30-Cu, and H50 proaches coupled with advanced bioinformatics have
arrays by using surface-enhanced laser desorption/ion- been used to identify putative disease biomarkers in
ization time-of-flight mass spectrometry. patient serum and plasma, and several studies have
Results: Prolonged transport and incubation at room identified biomarkers that improve the positive predictive
temperature generated low mass peaks, resulting in value of disease detection (4, 5 ). The cancer antigen
distinctions among the protocols. The most and least CA125, a diagnostic and prognostic marker for ovarian
stringent methods gave the lowest overall peak vari- cancer, is increased in only 85% of patients with ovarian
ances, indicating that proteolysis in the latter may have cancer and 50% of those with early stage disease (6 ), and
been nearly complete. For samples transported on ice may also be increased in benign disease. Research using
new technologies is focused on developing a marker that
1
will outperform CA125 or can be used in combination to
Translational Research Laboratory, Institute of Women’s Health, Univer-
sity College London, London, United Kingdom.
increase the performance of the current test.
2
Cancer Proteomics Group, Ludwig Institute for Cancer Research, London Several different methods based on mass spectrometry
Branch, London, United Kingdom. (MS)5 have been applied in the search for cancer biomar-
3
Department of Computer Science, Royal Holloway College, University of kers [reviewed in (1, 7 )]. Surface-enhanced laser desorp-
London, London, United Kingdom.
4
Ciphergen Biosystems, Inc., Fremont, CA.
5
* Address correspondence to this author at: Translational Research Labo- Nonstandard abbreviations: MS, mass spectrometry; SELDI-TOF MS,
ratory, Institute of Women’s Health, University College London, Huntley surface-enhanced laser desorption/ionization time-of-flight mass spectrome-
Street, London, WC1E 6DH, UK. Fax 44-207-6796334; e-mail jtimms@ try; UKCTOCS, United Kingdom Collaborative Trial of Ovarian Cancer
ludwig.ucl.ac.uk. Screening; RT, room temperature; IMAC, immobilized metal affinity capture;
Received September 15, 2007; accepted January 15, 2007. PCA, principal components analysis; ITIH4, inter-␣-trypsin inhibitor heavy
Previously published online at DOI: 10.1373/clinchem.2006.080101 chain H4; ApoCI, apolipoprotein C-I; TTR, transthyretin.
645
646 Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
tion/ionization time-of-flight (SELDI-TOF) MS (8, 9 ) has sealed and stored at ⫺80 °C. Time from venipuncture to
been used extensively for serum profiling. In this method, centrifugation was 30 h for each sample. Additional
high-throughput mass profiling with laser desorption/ samples from the same volunteers were collected in
ionization MS instrumentation is performed on sample Becton Dickinson red-top tubes, allowed to clot at RT for
proteins bound selectively to chip surfaces with different 60 min, subjected to transport/storage on wet ice for 2 h
chemical properties. Spectral patterns are then compared before centrifugation, transferred to straws, and stored at
across samples to find discriminating masses or changes ⫺80 °C (protocol 2a; Yellow; YE) (14 ). A 3rd protocol used
in peak intensities. Initial enthusiasm about these new a 5 min clotting time at RT, followed by transport/storage
technologies has been somewhat tempered by questions on wet ice for 3 h before centrifugation, transfer to straws,
on the robustness of class discriminating algorithms and and storage (protocol 2b; Gray; GY). Three variants of
method reproducibility (7, 10, 11 ). Increasing evidence protocol 2b were used to prepare cryovials for storage
that sample collection and processing can affect protein instead of straws (protocol 2c; Cryovial; CR), with trans-
profiles and the ability to differentiate between disease port/storage on wet ice for 6 h instead of 3 h (protocol 2d;
and control samples has cast further doubt on the validity Orange; OR) and with transport/storage for 3 h at RT
of some studies (12, 13 ). Transit time, storage conditions, instead of on wet ice (protocol 2e; White; WH). Handling
clotting time, and tube type can all affect serum profiles, protocols are detailed in Table S1 of the Data Supplement
irrespective of true biological variation (14 –17 ). It is likely that accompanies this article at (http://www.clinchem.
that such introduced differences are primarily driven by org/content/vol53/issue4). Although 25 ⫻ 6 protocols
proteolysis, although other variables may contribute, would generate 150 individual serum samples for analy-
such as agglutination or differential adhesion of serum sis, because of insufficient material only 13 protocol 2c
polypeptides to tube walls. These findings have raised samples were available.
concerns about using samples for case-control studies
from older collections, where samples were collected and sample preparation
transported for different times at ambient temperatures.
Sample preparation details are provided as Supple-
Many of these collections are unique, with samples pre-
mental Data. Briefly, samples were thawed, randomized,
dating cancer diagnosis. One such collection is the United
and triplicate 25 ␮L aliquots placed into 96-well plates.
Kingdom Collaborative Trial of Ovarian Cancer Screening
After denaturation in urea and dilution, the samples at
(UKCTOCS), in which 202 638 postmenopausal women
pH 9 were put in filter plates containing rehydrated
from 13 centers in the UK were randomized to screening
QHyperD威 F resin (Pall Corporation) and incubated with
vs controls; this serum bank will eventually have 500 000
shaking. Unbound material was removed on a vacuum
samples, including serial samples from 50 000 women (see
manifold as fraction 1 (FR1), and proteins were eluted in
www.ukctocs.org.uk). The collection protocol for this trial
a step-wise fashion by decreasing pH (FR2, pH 7; FR3, pH
allows blood samples to stand on the clot for 24 –56 h
before processing. Thus, if proteomic technologies are 5; FR4, pH 4; and FR5, pH 3) with a final organic solvent
used for biomarker discovery from such collections, it is elution (FR6). The 6 fractions were applied to CM10
imperative to compare samples collected and processed (weak cation-exchange) and IMAC30-Cu (immobilized
using these less stringent protocols with those collected in metal affinity capture) arrays in 96-sample bioprocessors
accordance with protocols that involve immediate trans- (Ciphergen Biosystems). FR6 samples were also applied
port on ice. to H50 (hydrophobic) arrays. Chip preparation, sample
We examined the impact of diverse serum handling application, and matrix application were performed ac-
protocols on protein profiles observed by SELDI-TOF MS. cording to the manufacturer’s instructions. All liquid
We also sought to determine the least variable and most handling steps were performed on an Aquarius work-
clinically feasible handling method for prospective serum station (Tecan).
collections for proteomic studies.
seldi-tof ms data acquisition and processing
Materials and Methods Details of SELDI-TOF MS data acquisition and processing
sample collection and handling are provided as Supplemental Data. Briefly, spectra were
This study was approved by the local ethics committee acquired on an externally calibrated ProteinChip威 System
and written informed consent was obtained from all Series 4000 instrument, using 2 laser intensities for acqui-
donors. Serum samples were collected at Barts and the sition of low (2.5–20 kDa) and high (20 –200 kDa) mass
London NHS Trust on 2 consecutive days from 25 range data. Spectra were processed (baseline subtraction,
healthy, postmenopausal women randomized to the deionzing, normalization, spectral alignment, and peak
CA125 arm of UKCTOCS. Protocol 1 (Green; GN) used detection) with CiphergenExpress software, as described
the existing UKCTOCS protocol (see www.ukctocs. in Supplemental Data. Peak numbers were recorded for
org.uk) with samples collected in Greiner gel tubes, the different handling methods and fraction types. We
allowed to clot, centrifuged at room temperature (RT), examined the differences between the handling methods
divided into aliquots, and placed in straws that were heat by principal components analysis (PCA), hierarchical
clustering, and examination of P value by use of mean role of the sponsors

peak intensities from triplicate samples. Median variances Ciphergen Biosystems participated in the study design,
were also calculated as descriptors of trends for different donation of reagents, interpretation of results, and writing
collection/handling methods. of this article, but were not involved in donor selection.
Approval of the paper by Ciphergen Biosystems before its
peak identification submission was not required.
Proteins were enriched by liquid chromatography and
ultrafiltration followed by SDS-PAGE. Peptides ⬍5 kDa Results
were identified by direct sequencing by tandem mass Serum samples were processed 6 different ways (see
spectrometry (see below). Proteins ⬎5 kDa were purified Table S1 in the online Data Supplement) to assess the
by SDS-PAGE and stained using a Colloidal Blue Staining effects of time and temperature before storage, clotting
Kit (Invitrogen). Selected bands were excised, and one time, and storage tube type on SELDI-TOF MS profiles. To
quarter of each was extracted using 50% formic acid/25% improve coverage, samples were prefractionated using
acetonitrile/15% isopropanol/10% water and reanalyzed strong anion-exchange chromatography, and fractions
by SELDI-TOF MS to confirm matching of the stained further spotted onto different SELDI arrays (CM10,
band with the peak of interest. The remainder was in-gel IMAC30-Cu2⫹ and H50) generating 13 different condi-
digested with trypsin and analyzed by tandem MS using tions for SELDI-TOF MS analysis. All steps were auto-
a Q-STAR威 XL equipped with a PCI-1000 ProteinChip mated using robotics to improve reproducibility and
Interface (Ciphergen Biosystems). MS/MS spectra were throughput. The number of peaks (average of triplicate
submitted to the database mining tool Mascot (Version samples) in each fraction type was first compared across
2.1.2; Matrix Science) and searched against the updated the different handling methods (Table 1). Based on previ-
SwissProt or NCBInr databases with the following search ous work, and after preliminary analysis, only fractions
parameters: trypsin, allowing up to 2 missed cleavages (or FR1, 4 and 6 were considered since the others gave either
semitrypsin if the trypsin search was not successful); low peak numbers (FR2) or similar patterns of peaks to
peptide tolerance ⫾50 ppm; MS/MS tolerance ⫾ 0.3 Da; those in FR4 (FR3 and FR5). For most fractions, the CM10
peptide charge ⫹1. Peak identifications were also con- surface gave more peaks than the IMAC30 surface, with
firmed using data from previous publications (3, 18 –31 ). FR1 (anion exchange flow-through) on the CM10 chip
All are abundant, ubiquitously expressed serum proteins surface yielding the most peaks overall. The organic
or proteolytic fragments. elution (FR6) gave the highest number of peaks on all chip
Table 1. Numbers of spectral peaks by fraction type and collection method.

Mass range Protocol FR1 CM10 FR1 IMAC30 FR4 CM10 FR4 IMAC30 FR6 CM10 FR6 IMAC30 FR6 H50 Total
2.5–20 kDa ALL 117 71 64 52 82 71 77 534
20–200 kDa ALL 56 57 48 51 48 70 67 397
2.5–200 kDa ALL 173 128 112 103 130 141 144 931
2.5–20 kDa CR 70 23 30 25 48 49 60 305

GN 80 45 39 31 51 45 61 352
GY 64 24 28 24 48 48 58 294
OR 65 29 35 27 54 46 58 314
WH 69 30 43 27 54 48 56 327
YE 74 27 39 28 56 42 58 324
20–200 kDa CR 39 34 32 35 41 45 42 268
GN 44 34 34 33 43 41 40 269
GY 36 29 34 32 45 39 39 254
OR 39 30 34 29 46 40 42 260
WH 39 29 32 32 46 39 40 257
YE 40 29 33 32 44 39 42 259
2.5–200 kDa CR 109 57 62 60 89 94 102 573
GN 124 79 73 64 94 86 101 621
GY 100 53 62 56 93 87 97 548
OR 104 59 69 56 100 86 100 574
WH 108 59 75 59 100 87 96 584
YE 114 56 72 60 100 81 100 583
Peaks were picked in CiphergenExpress software using the criteria outlined in Materials and Methods. Peak numbers by fraction type, molecular weight range, and
collection protocol are shown.
types. The individual collection methods gave similar participant and protocol. To test the null-hypothesis that
numbers of peaks in each fraction and on each chip type, there is no difference between protocols, P values were
although the least stringent protocol 1 (GN) gave an calculated for all 160 peaks using the Wilcoxon sign test.
appreciably higher number in the low mass range, while The Pmin (minimum P value of all 160 values) was then
protocols 2b (GY) and 2c (CR) gave the lowest peak used to find a measure of agreement between pairs of
numbers (Table 1). protocols by calculating the corresponding “conserva-
We next conducted PCA to assess how samples and tive” P value according to the formula: min (n*Pmin;1),
handling methods grouped together. Protocol 1 (GN) was where n ⫽ 160 and the term “min” means that the
the most distinctive method, with most volunteer samples minimum of 2 numbers Pmin and 1 is taken; the word
grouping together and away from samples collected using “conservative” refers to the fact that the probability of the
the other methods (Fig. 1). This was true for all fraction/ P value not exceeding epsilon is at most epsilon, for any
chip surface combinations, but only in the low (2.5– epsilon between 0 and 1. To create a summary table to
20 kDa) mass range. A likely explanation for this separa- characterize the combination of all 13 fraction/chip types,
tion is the extended transport/storage time used in this the smallest P value in the fractions for each protocol pair
protocol. In support of this, partial separation was ob- was taken and adjusted according to the formula using
served between protocols 2b (GY; 3 h on ice) and 2e n ⫽ 13 (Table 2). Using this approach, protocols 2b (GY),
(WH; 3 h at RT), suggesting that temperature before 2c (CR), and 2d (OR) were most similar and protocol 1
centrifugation is a major factor influencing spectral pat- (GN) most dissimilar, in agreement with the PCA. How-
terns (data not shown). Importantly, there was no differ- ever, there was no significant difference between proto-
ences among the protocols, using either mass range, when cols 1 and 2c, (P ⫽ 0.276), although protocol 2c had the
samples were transported on ice for 3 vs 6 h (GY vs OR), smaller number of 13 samples, making it difficult to make
when samples were clotted for 60 min vs 5 min (YE vs reliable conclusions. This may also explain why all other
GY), or when samples were strawed vs aliquoted into protocols showed a strong agreement with protocol 2c.
cryovials for storage (GY vs CR) (data not shown). Proto- Notably, there was a significant difference between pro-
cols were also compared based on P values. For this, the tocols 2a (YE) and 2b (GY), suggesting that clotting time
preprocessed 160 most frequent peaks were selected and does influence the protein profiles, a finding which was
a median intensity value (n ⫽ 3) obtained for each study not apparent from the PCA.
Fig. 1. Protocol comparison by princi-

ple components analysis.
PCA was performed using Ciphergen Ex-
press software to compare samples col-
lected using the different protocols. In the
example shown, peaks from FR1 CM10
2.5–20 kDa were used for the analysis. The
circles denote the clustering of most proto-
col 1 (GN) samples. A similar clustering of
protocol 1 samples was found for all frac-
tion/chip surface combinations, but only in
the low (2.5–20 kDa) mass range. Note that
the colors used in the Fig. differ from those
used for protocol labeling.
Table 2. Protocol comparison based on Pmin values.

Protocol 1 (GN) 2a (YE) 2b (GY) 2c (CR) 2d (OR) 2e (WH)
1 (GN) 1 0.000135 0.00027 0.276123 6.74E-05 6.74E-05
2a (YE) 0.000135 1 0.000576 0.276123 0.000144 0.043129
2b (GY) 0.00027 0.000576 1 0.723633 0.265808 0.000288
2c (CR) 0.276123 0.276123 0.723633 1 0.507813 0.276123
2d (OR) 6.74E-05 0.000144 0.265808 0.507813 1 7.21E-05
2e (WH) 6.74E-05 0.043129 0.000288 0.276123 7.21E-05 1
Wilcoxon sign test P values were calculated for pairs of protocols by using the median intensity values (n⫽3) for the preprocessed 160 most frequent peaks. The
minimum P value (Pmin) was then used to find a measure of agreement between pairs of protocols by calculating the corresponding “conservative” P value according
to the formula: min (n ⫻ Pmin;1), where n⫽160. To create the summary table shown, to characterize the combination of all 13 fraction/chip types, the smallest P value
in the 13 fractions/chip types for each protocol pair was taken and adjusted according to the formula, using n⫽13.
We next analyzed peak variances to assess the general thyretin (13.9 kDa), which were relatively stable across the
stability of the handling methods. Data were calibrated different handling methods. In contrast, hemoglobin ␣
internally based on known peaks, and median intensities (15126.36 Da) and ␤ (15867.28 Da), and fibrinogen ␣
and SDs taken for each peak across the 25 study partici- fragments 3262.47 Da and 5904.22 Da and their modified
pants. These values were used to calculate a coefficient of forms, displayed lower intensities in protocols 1 (GN) and
variance (SD/median intensity) for each protocol and 2e (WH) (Table 3). Several peaks, including those repre-
fraction type. Median variance values were also calcu- senting the major form of platelet factor 4 [7765.10 Da; see
lated and compared across protocols to give an overall (30 )], showed altered intensity in protocol 2b (GY) vs 2a
measure of variability. Protocol 2a (YE) gave the lowest (YE) and 1 (GN), but not the other protocols, suggesting
median variances in both mass ranges, followed by pro- that their final serum concentration is affected by clotting
tocols 1 (GN) and 2e (WH), suggesting that these were the time (highlighted in bold in Table 3).
most stable methods (Fig. 2A and B). This was corrobo-
rated by the observation that these methods gave the Discussion
highest numbers of peaks with a median variance ⬍1.0 One of the main objectives of this study was to select an
(Fig. 2C). optimal protocol for serum collection and handling for a
Selection of peaks for identification was based on large case-control study, which would be feasible for both
altered intensity and variance across protocols, with em- clinical collection and protein profiling. Variance analysis
phasis on differences between protocols 1 (GN) and 2a showed that protocol 2a (YE) gave the lowest overall
(YE); considered the least and most stringent protocols, variance when all peaks were considered, and is therefore
respectively. Identifications were in agreement with pre- the preferred serum collection/handling method. This
vious studies, where available (Table 3). Examples of two method, while clinically feasible, requires rapid transit of
peaks that displayed increased intensity in protocol 1 samples on ice to the laboratory for processing and
samples, but not those of protocol 2a, are shown in Fig. 3. freezing, and imposes additional logistic and resource
Peak 4286 Da (Fig. 3A) was identified as the 4281.78 Da burdens on clinical studies. No obvious effects on overall
fragment of inter␣-trypsin inhibitor heavy chain H4 peak number, median intensity, or variance were appar-
[ITIH4) (see (22 )]. Two other ITIH4 fragments (3157.58 Da ent between the various protocols (GY, YE, CR, OR)
[see (27 )] and 3955.48 Da [see (31 )] and their methionine- where samples were placed on ice. This indicates that
oxidised forms were also identified, with all displaying transport times of up to 6 h, and different storage methods
increased intensity in protocol 1 samples (Table 3). An (straws vs cryovials), have little effect on serum protein
8144 Da peak appeared to be a superposition of two profiles as long as samples are on ice. However, the
peaks. In most protocols the 8144 Da peak represented an altered intensity of some peaks (e.g., platelet factor 4 at
8141.59 Da form of platelet factor 4 (30 ), with an alterna- 7765.18 Da) did show that altering the clotting time has
tively cleaved signal sequence (data not shown). In pro- some effects. Perhaps not surprisingly, PCA showed the
tocol 1, the up-regulated peak 8126 Da corresponded to a greatest separation for protocol 1 over other methods.
C-terminal-truncated fragment of C3a anaphylatoxin Separation was representative for all 2.5–20 kDa spectra
[8126.52 Da; see (3, 23 )] (Fig. 3C). A C1 inhibitor C- on all arrays, and was not apparent in the 20 –200 kDa
terminal fragment (4152.87 Da), an albumin N-terminal range, in accordance with increased peak numbers and
fragment (3156.59 Da), and peaks corresponding to neu- intensities in the lower mass range. There was minor
trophil defensins 1, 2, and 3 were also noticeably in- separation attributable to transport/storage at RT vs ice
creased in protocol 1 samples. Other identifications (GY vs WH). Thus, although strict control on sample
were apolipoprotein C-I (ApoCI; 6330.59 Da) and its handling protocols has been proposed to reduce variabil-
SPA adduct, a truncated form of ApoCII (8204.17 Da), ity, these data show that sample collection protocols can
albumin dimer (138 kDa), transferrin (79 kDa), and trans- be more flexible with regard to clotting times and trans-
Fig. 2. Variance analysis.

(A), median peak variances (standard devi-
ation/median peak intensity) by protocol
and fraction type. Numbers in italics are
greater than an arbitrary cutoff value of 0.6.
(B), graphical representation of median
peak variances for the low (left) and high
(right) mass ranges. (C), number (#) and
percentage (%) of peaks where median
peak variance (SD/median peak intensity)
was ⬍1.0.
port; transport for 3 h at RT or up to 6 h on ice can be used Many of these biobanks are unique because they are
in clinical studies with limited impact on variability. The associated with long follow-up, and contain samples
critical issue is that all individual samples must be treated stored many years before disease diagnosis.
exactly the same. The greater number of peaks (often with lower vari-
The stringent protocol with 1 h clotting and 3 h ances and higher peak intensities) in the low mass range
transit/storage on ice (YE) has been shown to give the for protocol 1 samples suggests that proteolysis in these
most reproducible results in a serum profiling analysis samples has gone to completion. A similar trend was also
performed using automated magnetic bead-based pre- apparent in protocol 2e (WH) samples, which were incu-
fractionation and MALDI-TOF MS (14 ). However, our bated for 3 h at RT before storage. These data are in
study showed that despite transit/storage at RT for 30 h, agreement with a previous SELDI-TOF MS study show-
protocol 1 (GN) also gave a relatively low overall vari- ing increases in certain peaks with time between veni-
ance. This finding is critical, because it establishes that puncture and sample processing, with some overlap with
samples collected in older studies with longer transit the peaks identified here (15 ). In particular, the increased
times at RT can be used in case-control studies for novel intensities of the ITIH4, C3a, and C1 inhibitor fragments
biomarkers as long as all samples were handled similarly. and their modified forms in protocol 1 samples provides
Table 3. Selected peaks displaying differential intensity and variance across handling protocols.
Median SD/Median
Cluster MW (int
Fraction type index cal) Identification Refs CR GN GY OR WH YE CR GN GY OR WH YE
F1 CM102.5– 6 2770.5 0.15 0.02 0.15 0.22 0.18 0.20 1.66 5.14 1.12 0.75 0.97 0.53
20 kDa 8 2851.3 0.13 0.51 0.05 0.06 0.10 0.11 2.11 0.52 3.05 2.44 2.54 1.68
9 2870.9 1.01 6.27 0.77 1.20 2.02 1.98 2.63 0.54 2.05 1.10 1.23 0.99
11 2954.7 5904.22 Da, 2⫹ ion 1.83 0.44 1.49 1.93 1.78 1.67 0.56 1.16 0.50 0.47 0.56 0.53
14 3075.2 0.04 0.52 0.05 0.06 0.14 0.12 6.12 0.62 3.01 2.32 1.77 1.63
17 3141.3 0.19 0.54 0.14 0.20 0.22 0.24 1.61 0.65 1.56 0.87 0.94 1.13
18 3159.0 ITIH4 fragment, (27 ) 0.61 2.78 1.10 0.89 1.15 1.13 1.70 0.50 0.88 1.03 0.86 0.66
3157.58 Da
19 3193.2 1.51 0.02 1.65 1.84 1.67 1.65 0.93 28.3 0.62 0.72 0.54 0.53
21 3264.5 Fibrinogen ␣ frag, (31 ) 2.42 0.01 2.11 2.30 2.72 2.45 0.77 70.3 0.64 0.72 0.50 0.56
3262.47 Da
26 3423.0 0.29 0.08 0.25 0.21 0.12 0.13 0.47 1.16 0.59 0.68 0.87 0.59
29 3489.1 Neutrophil defensin 3, (21, 24 ) 0.11 0.32 0.12 0.06 0.17 0.10 0.85 0.91 0.98 1.09 0.69 1.03
3486.1 Da
33 3955.8 ITIH4 fragment, (31 ) 0.80 4.16 0.88 0.76 0.86 0.92 1.88 0.48 2.53 2.55 1.18 1.14
3955.46 Da
34 3969.9 3955.46 Da Met-ox 0.36 1.13 0.60 0.47 0.30 0.43 1.78 0.78 4.20 4.06 3.89 1.43
35 4072.7 0.17 0.79 0.08 0.07 0.20 0.22 1.13 0.62 1.82 1.43 0.66 0.61
37 4137.6 0.25 1.71 0.46 0.49 0.77 0.61 4.01 0.40 1.05 0.78 0.67 0.87
39 4286.0 ITIH4 fragment, (22 ) 0.42 1.66 0.97 0.84 0.50 0.84 3.43 0.75 3.74 3.58 3.57 1.64
4281.78 Da
43 4677.0 0.22 0.98 0.12 0.11 0.23 0.28 2.36 0.92 2.15 1.36 1.10 1.31
47 5074.4 0.12 0.42 0.11 0.11 0.17 0.18 2.42 0.71 1.17 1.48 1.08 0.95
54 5492.2 0.21 0.05 0.17 0.18 0.20 0.19 0.43 2.20 0.66 0.79 0.73 0.56
55 5764.6 0.06 0.47 0.04 0.05 0.11 0.12 4.18 0.78 2.76 1.63 1.46 1.55
Clinical Chemistry 53, No. 4, 2007
57 5905.3 Fibrinogen ␣ frag, (25 ) 13.38 3.01 10.06 13.96 14.38 12.43 0.54 1.39 0.52 0.48 0.44 0.50
5904.22 Da
58 5914.0 5904.22 Met-ox 7.96 1.93 6.80 8.39 8.16 7.87 0.48 1.55 0.62 0.34 0.26 0.30
60 6114.0 5904.22 SPA adduct 1.29 0.31 0.98 1.49 1.36 1.30 0.58 1.38 0.58 0.46 0.54 0.53
63 6637.8 ApoCI, 6630.59 Da (3, 29 ) 13.96 10.63 14.97 14.56 12.53 12.14 0.24 0.32 0.35 0.28 0.28 0.20
64 6843.7 6630.59 Da, 1.62 1.15 1.61 1.59 1.34 1.41 0.28 0.37 0.38 0.30 0.30 0.20
SPA adduct
75 8144.0 C3a truncated, (3, 23 ) 0.82 5.07 0.76 0.55 1.11 1.25 1.52 0.52 1.28 0.81 0.61 0.57
8126.52 Da
76 8344.0 8126.52 Da, 0.13 0.40 0.11 0.15 0.13 0.16 0.79 0.48 0.76 0.60 0.50 0.41
SPA adduct
109 15158.0 Hb alpha, (3, 22 ) 0.15 0.06 0.31 0.33 0.10 0.22 0.52 1.99 0.54 0.49 2.05 0.40
15126.36 Da
F1 CM1020– 20 26573.0 Transferrin, 3⫹ ion 0.05 0.25 0.04 0.04 0.06 0.07 2.06 1.13 4.55 1.19 1.04 1.57
200 kDa 40 79510.0 Transferrin (20, 29 ) 3.18 2.93 2.65 2.67 2.81 2.66 0.38 0.22 0.38 0.29 0.32 0.19
43 102333.0 0.06 0.05 0.05 0.05 0.05 0.05 0.34 0.14 0.17 0.23 0.23 0.14
50 138323.0 Albumin dimer 0.22 0.23 0.23 0.23 0.23 0.24 0.56 0.21 0.28 0.26 0.30 0.16
651
Table 3. Continued.
652
Median SD/Median
Cluster MW (int
F1 IMAC 2.5– 10 2863.6 1.58 5.62 1.39 1.33 2.27 1.44 0.80 0.52 1.41 0.51 1.24 0.48
20 kDa 11 2950.9 1.37 0.35 1.10 1.22 1.26 1.33 0.41 1.33 0.69 0.54 0.49 0.40
12 3156.0 ITIH4 fragment, (27 ) 0.84 2.47 0.62 0.78 1.02 1.00 1.04 0.53 1.34 0.44 1.09 0.46
3157.58 Da
16 3394.8 0.25 1.30 0.26 0.27 0.37 0.45 2.33 0.89 1.73 1.34 2.11 0.62
19 3956.1 ITIH4 fragment, (31 ) 0.62 4.86 0.54 0.63 0.78 0.58 4.62 0.87 1.40 3.85 3.34 5.35
3955.46 Da
20 3975.8 3955.46, Met-ox 0.40 3.14 0.59 0.41 0.32 0.43 4.50 0.63 0.91 5.69 6.63 6.90
24 4284.1 ITIH4 fragment, (22 ) 0.51 1.58 0.50 0.63 0.73 0.67 1.59 2.04 1.58 4.44 3.01 5.50
4281.78 Da
25 4297.5 4281.78, Met-ox 0.31 0.91 0.50 0.42 0.37 0.27 1.99 1.76 1.09 2.31 2.16 4.42
30 5767.6 0.19 0.57 0.09 0.11 0.19 0.11 0.95 0.80 1.78 1.02 1.79 1.52
31 5905.5 Fibrinogen ␣ frag, (25 ) 6.02 1.12 5.77 6.80 8.30 7.11 0.80 1.30 0.74 0.89 0.57 0.46
5904.22 Da
32 5921.4 5904.22 Da Met-ox 3.64 0.80 3.14 3.23 4.22 4.00 0.66 0.94 0.86 0.71 0.60 0.40
33 5970.1 0.39 0.20 0.53 0.46 0.59 0.60 1.01 0.64 0.67 0.85 0.62 0.48
34 6113.8 5904.22 Da SPA adduct 0.53 0.13 0.63 0.63 0.93 0.86 0.83 1.97 0.66 0.92 0.54 0.41
38 6850.8 ApoCI, SPA adduct 0.74 0.52 0.76 0.92 0.84 0.99 0.46 0.79 0.40 0.36 0.40 0.29
39 7159.7 0.30 0.64 0.16 0.17 0.23 0.19 0.51 0.68 0.73 0.96 1.19 0.71
46 8131.5 C3a truncated, 8126.52 (3, 23 ) 0.53 3.11 0.34 0.34 0.62 0.33 3.15 0.69 0.85 0.79 0.79 1.43
Da
47 8159.1 0.33 1.85 0.31 0.36 0.81 0.74 2.52 0.63 0.80 1.43 0.58 0.58
56 11436.0 0.22 0.10 0.21 0.15 0.19 0.17 0.41 0.70 3.21 0.35 0.49 0.30
68 15913.0 Hb beta, (3, 22 ) 0.19 0.08 0.33 0.35 0.13 0.31 0.40 0.83 0.49 0.58 1.35 0.30
15867.27 Da
69 16073.0 0.13 0.04 0.21 0.22 0.07 0.19 0.40 1.12 0.56 0.57 1.54 0.32
F1 IMAC20– 1 20037.0 0.03 0.10 0.03 0.03 0.03 0.04 1.10 0.78 1.23 0.93 0.74 0.58
200 kDa 24 26560.0 Transferrin, 3⫹ ion 0.06 0.19 0.05 0.05 0.06 0.05 0.97 0.98 0.65 0.45 0.87 0.60
27 30461.0 0.05 0.06 0.08 0.07 0.07 0.07 0.72 0.36 0.32 0.52 0.38 0.22
31 39771.0 Transferrin, 2⫹ ion (29 ) 0.29 0.37 0.35 0.37 0.40 0.38 0.37 0.23 0.16 0.20 0.24 0.20
41 79249.0 Transferrin (29 ) 2.05 2.79 2.37 2.70 3.17 2.86 0.38 0.22 0.25 0.21 0.26 0.22
42 89545.0 0.16 0.18 0.16 0.16 0.18 0.17 0.30 0.32 0.21 0.29 0.23 0.27
Tims et al.: Sample Handling Effects on Serum Proteomic Profiles
44 101944.0 0.06 0.05 0.05 0.05 0.05 0.05 0.20 0.28 0.25 0.27 0.26 0.18
45 106330.0 Transferrin ion 0.03 0.02 0.03 0.03 0.02 0.02 0.47 0.50 0.33 0.23 0.36 0.25
51 137052.0 Albumin dimer 0.21 0.21 0.24 0.22 0.21 0.21 0.31 0.32 0.26 0.34 0.29 0.25
F4 CM10 2.5– 19 3817.1 1.13 1.01 1.11 0.99 1.01 1.05 0.42 0.29 0.41 0.40 0.32 0.22
20 kDa 26 4965.7 0.17 1.77 0.13 0.13 0.53 0.28 1.11 0.54 1.32 1.25 0.85 0.87
27 5070.4 0.31 0.06 0.24 0.27 0.41 0.42 0.67 2.07 1.04 0.64 0.41 0.34
29 5755.2 0.41 0.13 0.40 0.39 0.51 0.82 0.42 1.06 0.35 0.38 0.50 0.28
35 6955.1 Transthyretin, cysteinylated, 0.78 1.07 0.98 0.89 0.95 0.96 0.29 0.21 0.35 0.26 0.22 0.24
2⫹ ion
36 7045.8 Transthyretin, 0.70 0.74 0.68 0.68 0.76 0.78 0.32 0.16 0.24 0.28 0.18 0.18
glutathionylated, 2⫹ ion
Table 3. Continued.
Median SD/Median
Cluster MW (int
38 7568.1 Hb alpha, 2⫹ion 0.23 0.06 0.16 0.25 0.07 0.14 0.51 1.31 0.91 0.51 1.74 0.68
54 13795.0 Transthyretin (26, 28 ) 0.55 0.52 0.69 0.64 0.68 0.67 0.33 0.20 0.29 0.29 0.19 0.18
55 13915.0 Transthyretin, cysteinylated (26, 28 ) 1.82 1.95 1.87 2.06 2.08 2.11 0.36 0.18 0.34 0.28 0.20 0.15
56 14091.0 Transthyretin, (26, 28 ) 1.61 1.90 1.70 2.00 2.08 2.09 0.46 0.26 0.35 0.30 0.27 0.21
glutathionylated
57 14289.0 Transthyretin SPA adduct 0.35 0.42 0.39 0.45 0.48 0.45 0.32 0.24 0.30 0.27 0.22 0.25
F4 IMAC 2.5– 7 2864.5 1.62 3.97 1.86 1.76 1.87 1.52 0.84 0.58 0.50 0.46 0.48 0.50
20 kDa 10 3156.5 Albumin, N-term frag, (3 ) 2.01 4.70 1.73 1.80 2.28 2.17 0.42 0.43 0.48 0.37 0.36 0.28
3156.59 Da
12 3370.5 Neutrophil defensin 2, (24 ) 0.50 2.94 0.42 0.44 0.74 0.21 0.90 0.45 1.49 1.10 0.84 2.63
3371.01 Da
14 3815.1 3.04 2.49 2.31 2.73 2.43 2.74 0.22 0.31 0.36 0.34 0.26 0.20
15 3881.5 0.48 0.99 0.25 0.17 1.37 1.04 0.98 0.77 1.95 3.54 0.53 0.66
16 4125.6 0.76 1.89 1.02 0.94 0.91 0.94 0.99 0.30 0.75 0.83 0.68 0.63
19 4801.6 0.36 1.53 0.57 0.33 0.54 0.58 1.33 0.44 0.82 1.36 1.19 0.76
20 5754.9 0.45 0.16 0.39 0.44 0.55 0.91 0.55 1.24 0.59 0.59 0.55 0.30
27 6951.3 Transthyretin, cysteinylated, 2.37 2.54 2.32 2.30 2.45 2.35 0.20 0.17 0.28 0.25 0.18 0.18
2⫹ ion
28 7042.9 Transthyretin, 1.63 1.78 1.68 1.71 1.72 1.70 0.26 0.15 0.30 0.18 0.17 0.13
glutathionylated, 2⫹ ion
29 7146.8 0.51 0.63 0.54 0.54 0.52 0.52 0.70 0.23 0.46 0.36 0.32 0.21
30 7760.2 Platelet factor 4, (30 ) 1.49 4.48 1.47 1.22 4.06 3.72 0.83 0.43 0.51 0.73 0.55 0.46
7765.18 Da
40 13747.0 Transthyretin (26, 28 ) 1.44 1.26 1.42 1.43 1.51 1.47 0.19 0.24 0.32 0.24 0.19 0.16
41 13888.0 Transthyretin, cysteinylated (26, 28 ) 4.60 4.72 4.40 4.68 4.59 4.65 0.19 0.17 0.31 0.23 0.17 0.13
Clinical Chemistry 53, No. 4, 2007
42 14072.0 Transthyretin, (26, 28 ) 3.91 4.09 3.90 3.95 4.27 4.11 0.23 0.20 0.32 0.23 0.19 0.16
glutathionylated
43 14275.0 Transthyretin SPA adduct 1.19 1.12 1.13 1.19 1.25 1.18 0.29 0.24 0.34 0.24 0.23 0.22
47 15884.0 Hb beta, 15867.27 Da (3, 22 ) 0.48 0.12 0.41 0.34 0.15 0.24 0.44 0.58 0.44 0.48 1.35 0.37
F6 CM10 2.5– 14 3371.4 Neutrophil defensin 2, (21, 24 ) 1.43 6.33 1.42 1.32 2.35 1.78 0.66 0.43 0.67 0.71 0.49 0.52
20 kDa 3371.01 Da
15 3442.0 Neutrophil defensin 1, (21, 24 ) 0.69 6.38 0.55 0.47 1.45 0.80 0.96 0.70 0.83 0.97 0.89 0.75
3442.09 Da
16 3473.6 0.51 0.00 0.43 0.45 0.17 0.28 0.71 127 1.07 1.16 1.41 1.75
17 3489.6 Neutrophil defensin 3, (24 ) 0.28 2.21 0.24 0.23 0.57 0.34 0.67 0.98 1.12 0.84 0.71 1.26
3486.1 Da
19 4051.4 0.21 1.04 0.12 0.19 0.45 0.30 1.08 0.59 1.10 0.59 0.59 0.68
20 4153.3 C1 inhibitor C-term frag, 0.82 2.13 0.59 0.73 0.76 0.70 0.78 0.83 1.05 0.70 0.74 0.76
4152.87 Da
43 7931.8 Hb beta, 15867.27 Da, 2⫹ 0.20 0.05 0.16 0.14 0.07 0.12 0.56 0.58 0.56 0.69 0.85 0.45
ion
44 8207.2 ApoCII, truncated, 8204.17 0.36 0.37 0.38 0.37 0.39 0.34 0.43 0.43 0.46 0.44 0.42 0.45
72 15148.0 Hb alpha, 15126.36 Da (26, 28 ) 0.39 0.08 0.22 0.32 0.11 0.18 0.60 0.57 0.86 0.70 0.93 0.64
653
evidence that these degradation products are generated as
Internally calibrated mass, peak identification, references, median peak intensity, and median variance (SD/median intensity) across different handling protocols are shown. Median intensity values in bold indicate
0.39
0.70
0.66
1.82
0.82
0.36
0.60
0.61
0.30
0.19
1.03
0.44
YE
a result of increased proteolysis due to extended trans-
1.12
port/storage. Conversely, full-length hemoglobin ␣ and ␤
0.48
0.65
1.00
0.75
0.56
0.54
1.20
0.56
0.23
1.18
0.42
displayed decreased intensities in protocols 1 (GN) and 2e
WH
(WH), suggesting that they may be subject to degradation.

Similarly, the decreases in fibrinogen ␣ fragments 3262.47
0.71
0.61
0.63
1.26
0.80
0.62
1.44
0.91
0.53
0.34
0.58
0.60
OR
and 5904.22 Da were consistent with further degradation

SD/Median
to smaller undetected forms. It is harder to explain the

increased levels of the neutrophil defensins in the proto-
0.58
0.75
0.57
2.51
0.91
0.50
0.90
1.16
0.54
0.28
0.92
0.48
GY
col 1 samples. These disulfide bond-containing molecules

are resistant to proteolysis, so an indirect mechanism
must account for their altered intensities. Several other
0.48
0.53
0.86
0.71
0.80
0.54
0.67
0.62
0.82
0.33
4.84
1.02
GN
protein forms did not change significantly with collection

method, revealing them to be relatively stable serum
0.53
0.59
0.58
1.50
0.80
0.63
1.24
0.51
0.32
0.27
0.80
0.84
markers.
CR
It has been suggested that many candidate disease

biomarkers identified in SELDI-TOF MS profiling exper-
0.66
1.33
0.89
0.28
0.26
0.52
0.08
0.14
0.04
0.05
0.03
0.32
YE
iments are abundant acute-phase reactants, and are thus

secondary effects of the diseased state (7 ). For example,
serum transthyretin (TTR) is a known marker for nutri-
0.63
1.59
1.26
0.46
0.36
0.48
0.11
0.08
0.03
0.04
0.02
0.25
WH
tional status and the inflammatory acute-phase response.

In ovarian cancer, TTR was identified as a potential early
diagnostic marker, with decreased TTR levels reported in
0.46
0.99
0.49
0.11
0.27
0.20
0.03
0.20
0.05
0.05
0.04
0.28
OR
the sera of ovarian cancer patients compared with con-

Median
trols, without differences in its microheterogeneity

0.47
0.86
0.48
0.08
0.28
0.22
0.03
0.21
0.04
0.05
0.03
0.27
Table 3. Continued.
(20, 32, 33 ). Notably, TTR and its modified forms were

GY
unaffected by the different handling conditions used here,

and displayed relatively low variances across this healthy
2.89
4.40
5.11
2.93
0.80
0.41
0.11
0.06
0.01
0.02
0.01
0.16
peaks that change due to altered clotting time. Median variance values shaded gray and in italics are ⬎1.0.
GN
cohort. Thus, it would appear that TTR is relatively stable,

making it a more robust disease biomarker. Similarly,
SELDI-TOF MS was previously used to detect ApoCI and
0.44
0.85
0.59
0.16
0.31
0.19
0.04
0.29
0.06
0.06
0.04
0.32
CR
transferrin as classifiers of ovarian, colorectal, and other

cancers (20, 34 ), and our data show that they are also
(21, 24 )
(21, 24 )
(21, 24 )
stable under the conditions tested.

(3, 22 )
Refs
In a recent study, the putative acute-phase protein

(30 )
(3 )
ITIH4 was shown to be extensively proteolytically pro-

cessed, and its fragmentation patterns associated with
different disease conditions (27, 31 ). Fragmentation was
Platelet factor 4, 7765.18
generally consistent with cleavages by endoproteases,

Hb beta, 15867.27 Da
Neutrophil defensin 2,
Albumin, N-term frag,
followed by exoproteases, and the observed fragments

Identification
4152.87 Da Met-ox
were reported to change little under different assay con-

Albumin, 2⫹ ion
3156.59 Da
3371.01 Da
3442.09 Da
ditions or processing procedures. An up-regulated cleav-

3486.1 Da
age fragment of ITIH4 was also shown to enable differ-

entiation of patients with ovarian cancer from healthy
Da
controls or patients with benign pelvic masses (32 ). Our

data provide evidence that ITIH4 is relatively unstable,
with the generation of fragments increasing in serum
3153.6
3371.4
3440.4
3483.5
4170.6
7759.6
8143.5
15853.0
82171.0
83506.0
32526.0
33453.0
MW (int
maintained at RT for prolonged periods. Hemoglobin ␤

cal)
has also been identified as a putative ovarian cancer

biomarker (3, 20 ), but appears from our study to be
relatively unstable in serum. With this in mind, ITIH4
Cluster
index
16
19
20
21
23
40
42
68
51
52
27
28
fragments or hemoglobin ␤ may not make robust disease

biomarkers unless strict precautions are taken with sam-
F6 IMAC 2.5–
Fraction type
F6 IMAC20–
F6 H50 20–
ple handling. Future work will involve additional MS/

200 kDa
200 kDa
20 kDa
MS-based identification of the unknown peaks that are

discriminatory for the different collection methods. This
will allow the assessment of their usefulness as potential
Fig. 3. Example spectra.

(A), example spectra showing peak 4286
Da (inter␣-trypsin inhibitor heavy chain H4
(ITIH4) fragment, 4281.78 Da) of FR1
CM10 across protocols. Median intensi-
ties and standard deviations for all sam-
ples are also shown. The peak displayed
a highest intensity in protocol 1 (GN) and
may be generated as a result of increased
proteolysis due to the extended trans-
port/storage at RT used in this protocol.
(B), example showing peak intensities for
8144 Da of FR1 CM10, also increased in
protocol 1. (C), the 8144 Da peak is a
superposition of 2 peaks representing a
C-terminal-truncated form of C3a anaphy-
latoxin (8126.52 Da), abundant in proto-
col 1 (GN) samples, and an N-terminal
variant of platelet factor 4 (8141.59 Da;
PF4). The spectra shown have not been
normalized.
disease biomarkers where they have been identified in have been handled in a similar manner. Cases and con-
other studies. trols should be matched for transport time and an assess-
Our work establishes that the proteomic analysis of ment made when proteolysis in these samples reaches
samples from established serum banks, where samples completion. Biomarker discovery using a proteomic ap-
were not collected in accordance with more stringent proach in such case-control sets, such as UKCTOCS, will
protocols, can be used for proteomic biomarker studies. involve stable biomarkers rather than labile proteins. For
The key factor is that all samples in the collection should future studies, the key variable during specimen collec-
tion will be transport on ice, and it does not seem to 18. Ruetschi U, Zetterberg H, Podust VN, Gottfries J, Li S, Hviid
matter if transport times are then 3 or 6 h. Simonsen A, et al. Identification of CSF biomarkers for frontotem-
poral dementia using SELDI-TOF. Exp Neurol 2005;196:273– 81.
19. Fung ET, Yip TT, Lomas L, Wang Z, Yip C, Meng XY, et al.
Classification of cancer types by measuring variants of host
Financial support from The Eve Appeal and Ciphergen response proteins using SELDI serum assays. Int J Cancer
Biosystems Inc. is gratefully acknowledged. 2005;115:783–9.
20. Kozak KR, Su F, Whitelegge JP, Faull K, Reddy S, Farias-Eisner R.
References Characterization of serum biomarkers for detection of early stage
1. Wulfkuhle JD, Liotta LA, Petricoin EF. Proteomic applications for ovarian cancer. Proteomics 2005;5:4589 –96.
the early detection of cancer. Nat Rev Cancer 2003;3:267–75. 21. Albrethsen J, Bogebo R, Gammeltoft S, Olsen J, Winther B, Raskov
2. Xiao Z, Prieto D, Conrads TP, Veenstra TD, Issaq HJ. Proteomic H. Upregulated expression of human neutrophil peptides 1, 2 and
patterns: their potential for disease diagnosis. Mol Cell Endocrinol 3 (HNP 1–3) in colon cancer serum and tumours: a biomarker
2005;230:95–106. study. BMC Cancer 2005;5:8.
3. Engwegen JY, Gast MC, Schellens JH, Beijnen JH. Clinical pro- 22. Koomen JM, Shih LN, Coombes KR, Li D, Xiao LC, Fidler IJ, et al.
teomics: searching for better tumour markers with SELDI-TOF Plasma protein profiling for diagnosis of pancreatic cancer reveals
mass spectrometry. Trends Pharmacol Sci 2006;27:251–9. the presence of host response proteins. Clin Cancer Res 2005;
4. Ludwig JA, Weinstein JN. Biomarkers in cancer staging, prognosis 11:1110 – 8.
and treatment selection. Nat Rev Cancer 2005;5:845–56. 23. Li J, Orlandi R, White CN, Rosenzweig J, Zhao J, Seregni E, et al.
5. Chatterjee SK, Zetter BR. Cancer biomarkers: knowing the present Independent validation of candidate breast cancer serum biomar-
and predicting the future. Future Oncol 2005;1:37–50. kers identified by mass spectrometry. Clin Chem 2005;51:2229 –
6. Brioschi PA, Irion O, Bischof P, Bader M, Forni M, Krauer F. Serum 35.
CA 125 in epithelial ovarian cancer: a longitudinal study. Br J 24. Melle C, Ernst G, Schimmel B, Bleul A, Thieme H, Kaufmann R, et
Obstet Gynaecol 1987;94:196 –201. al. Discovery and identification of alpha-defensins as low abun-
7. Diamandis EP. Mass spectrometry as a diagnostic and a cancer dant, tumor-derived serum markers in colorectal cancer. Gastro-
biomarker discovery tool: opportunities and potential limitations. enterology 2005;129:66 –73.
Mol Cell Proteomics 2004;3:367–78. 25. Nomura F, Tomonaga T, Sogawa K, Ohashi T, Nezu M, Sunaga M,
8. Merchant M, Weinberger SR. Recent advancements in surface- et al. Identification of novel and downregulated biomarkers for
enhanced laser desorption/ionization-time of flight-mass spec- alcoholism by surface enhanced laser desorption/ionization-mass
trometry. Electrophoresis 2000;21:1164 –77. spectrometry. Proteomics 2004;4:1187–94.
9. Petricoin EF, Liotta LA. SELDI-TOF-based serum proteomic pattern 26. Schweigert FJ, Wirth K, Raila J. Characterization of the microhet-
diagnostics for early detection of cancer. Curr Opin Biotechnol erogeneity of transthyretin in plasma and urine using SELDI-
2004;15:24 –30. TOF-MS immunoassay. Proteome Sci 2004;2:5.
10. Baggerly KA, Morris JS, Edmonson SR, Coombes KR. Signal in 27. Song J, Patel M, Rosenzweig CN, Chan-Li Y, Sokoll LJ, Fung ET, et
noise: evaluating reported reproducibility of serum proteomic al. Quantification of fragments of human serum inter-alpha-trypsin
tests for ovarian cancer. J Natl Cancer Inst 2005;97:307–9. inhibitor heavy chain 4 by a surface-enhanced laser desorption/
11. Diamandis EP. Serum proteomic profiling by matrix-assisted laser ionization-based immunoassay. Clin Chem 2006;52:1045–53.
desorption-ionization time-of-flight mass spectrometry for cancer 28. Terazaki H, Ando Y, Suhr O, Ohlsson PI, Obayashi K, Yamashita T,
diagnosis: next steps. Cancer Res 2006;66:5540 –1. et al. Post-translational modification of transthyretin in plasma.
12. Coombes KR, Morris JS, Hu J, Edmonson SR, Baggerly KA. Serum Biochem Biophys Res Commun 1998;249:26 –30.
proteomics profiling–a young technology begins to mature. Nat 29. Ward DG, Suggett N, Cheng Y, Wei W, Johnson H, Billingham LJ,
Biotechnol 2005;23:291–2. et al. Identification of serum biomarkers for colon cancer by
13. Karsan A, Eigl BJ, Flibotte S, Gelmon K, Switzer P, Hassell P, et al. proteomic analysis. Br J Cancer 2006;94:1898 –905.
Analytical and preanalytical biases in serum proteomic pattern 30. Vermeulen R, Lan Q, Zhang L, Gunn L, McCarthy D, Woodbury RL,
analysis for breast cancer diagnosis. Clin Chem 2005;51: et al. Decreased levels of CXC-chemokines in serum of benzene-
1525– 8. exposed workers identified by array-based proteomics. Proc Natl
14. Villanueva J, Philip J, Chaparro CA, Li Y, Toledo-Crow R, DeNoyer Acad Sci USA 2005;102:17041– 6.
L, et al. Correcting common errors in identifying cancer-specific 31. Villanueva J, Shaffer DR, Philip J, Chaparro CA, Erdjument-Bro-
serum peptide signatures. J Proteome Res 2005;4:1060 –72. mage H, Olshen AB, et al. Differential exoprotease activities
15. Banks RE, Stanley AJ, Cairns DA, Barrett JH, Clarke P, Thompson confer tumor-specific serum peptidome patterns. J Clin Invest
D, Selby PJ. Influences of blood sample processing on low- 2006;116:271– 84.
molecular-weight proteome identified by surface-enhanced laser 32. Zhang Z, Bast RC, Jr., Yu Y, Li J, Sokoll LJ, Rai AJ, et al. Three
desorption/ionization mass spectrometry. Clin Chem 2005;51: biomarkers identified from serum proteomic analysis for the
1637– 49. detection of early stage ovarian cancer. Cancer Res 2004;64:
16. Baumann S, Ceglarek U, Fiedler GM, Lembcke J, Leichtle A, Thiery 5882–90.
J. Standardized approach to proteome profiling of human serum 33. Gericke B, Raila J, Sehouli J, Haebel S, Konsgen D, Mustea A, et
based on magnetic bead separation and matrix-assisted laser al. Microheterogeneity of transthyretin in serum and ascitic fluid of
desorption/ionization time-of-flight mass spectrometry. Clin Chem ovarian cancer patients. BMC Cancer 2005;5:133.
2005;51:973– 80. 34. Engwegen JY, Helgason HH, Cats A, Harris N, Bonfrer JM,
17. Findeisen P, Sismanidis D, Riedl M, Costina V, Neumaier M. Schellens JH, et al. Identification of serum proteins discriminating
Preanalytical impact of sample handling on proteome profiling colorectal cancer patients and healthy controls using surface-
experiments with matrix-assisted laser desorption/ionization enhanced laser desorption ionisation-time of flight mass spec-
time-of-flight mass spectrometry. Clin Chem 2005;51:2409 –11. trometry. World J Gastroenterol 2006;12:1536 – 44.
657– 665 (2007) Proteomics and
Protein Markers
Multiplexed Proteomic Analysis of Oxidation and

Concentrations of Cerebrospinal Fluid Proteins in
Alzheimer Disease
Minna A. Korolainen,1,2* Tuula A. Nyman,3,4 Paula Nyyssönen,1,2
E. Samuel Hartikainen,1,2 and Tuula Pirttilä1,2
Background: Carbonylation is an irreversible oxidative D– binding protein, apolipoprotein A-I, and ␣-1-antit-
modification of proteins that has been linked to various rypsin exhibited higher extents of carbonylation in men.
conditions of oxidative stress, aging, physiological dis- Conclusions: None of the brain-specific proteins exhib-
orders, and disease. Increased oxidative stress is thus ited carbonylation changes in probable AD patients
also considered to play a role in the pathogenesis of compared with age-matched neurological controls
age-related neurodegenerative disorders such as Alzhei- showing no cognitive decline. The carbonylation status
mer disease (AD). In addition, it has recently become of proteins differed between women and men. Two-
evident that the response mechanisms to increased oxi- dimensional multiplexed oxyblotting is applicable to
dative stress may depend on sex. Several oxidized car- study both the concentrations and carbonylation of
bonylated proteins have been identified in plasma and cerebrospinal fluid proteins.
brain of AD patients by use of 2-dimensional © 2007 American Association for Clinical Chemistry
oxyblotting.
Methods: In this pilot study, we estimated the concen-
Oxidative stress is an integral part of physiological func-
trations and carbonylation of the most abundant cere-
tions in all organisms living in an aerobic environment.
brospinal fluid proteins in aging women and men, both
An increase in oxidative stress, especially in the brain, is a
AD patients suffering from mild dementia and individ-
part of normal aging and is related directly to decreased
uals exhibiting no cognitive decline. Oxidized carbony-
neurological activities and inversely to lifespan (1, 2 ). In
lated proteins were analyzed with 2-dimensional multi-
plexed oxyblotting, mass spectrometry, and database addition, oxidative stress may be an essential feature in a
searches. variety of neurodegenerative diseases, including Alzhei-
Results: Signals for ␤-trace, ␭ chain, and transthyretins mer disease (AD)5 (3 ). In AD, increased oxidative stress
were decreased in probable AD patients compared with has been associated with aggregation of proteins, calcium
controls. The only identified protein exhibiting an in- dysregulation, mitochondrial malfunction, chronic in-
creased degree of carbonylation in AD patients was ␭ flammation, altered antioxidant function, and accumula-
chain. The concentrations of proteins did not generally tion of redox-active metals (3–5 ). Accumulating evidence
differ between men and women; however, vitamin also points to significant sex differences with respect to
oxidative stress and antioxidant defense systems, includ-
ing the response to antioxidant treatment (6 – 8 ). Taken
1 together, the factors that are known to be related to an
Department of Neurology, University of Kuopio and Kuopio University
Hospital, Kuopio, Finland. increase in oxidative stress, such as age, and differences in
2
Brain Research Unit, Clinical Research Centre/ Mediteknia, University of antioxidant response mechanisms, such as sex, may also
Kuopio, Kuopio, Finland.
3
Centre for Biotechnology, University of Turku and Åbo Akademi Uni-
versity, Turku, Finland.
4
Finnish Institute of Occupational Health, Helsinki, Finland.
5
*Address correspondence to this author at: Department of Neurology, Nonstandard abbreviations: AD, Alzheimer disease; ROS, reactive oxy-
University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland. Fax 358-17- gen species; 2-D, 2-dimensional; 2-DE, 2-D gel electrophoresis; CSF, cerebro-
162048; e-mail Minna.Korolainen@uku.fi. spinal fluid; MMSE, Mini-Mental State Examination; IPG, immobilized pH
Received August 9, 2006; accepted January 15, 2007. gradient; DBP, vitamin D– binding protein; ␣-1-AT, ␣-1-antitrypsin; TTR,
Previously published online at DOI: 10.1373/clinchem.2006.078014 transthyretin.
657
658 Korolainen et al.: Carbonylated CSF Proteins in AD
influence the risk for an individual to develop AD composition of CSF partially reflects cerebral metabolic
(3, 9 –11 ). changes and enables screening of ongoing pathophysio-
Oxidative stress occurs as a result of an imbalance logical processes in brain. Approximately 20% of the
between the generation and scavenging of reactive oxy- proteins in CSF are considered to be brain derived,
gen species (ROS) (12 ). This increase in the amount of although they may not be brain specific (27 ). Physiologi-
ROS may be oxidatively damaging to a variety of macro- cally, proteins that are small and/or highly lipophilic can
molecules, leading to fatal cellular effects. Proteins are be passively transported across the blood–CSF barrier
considered to be the major targets of oxidative stress. In from the peripheral circulation. Alternatively, proteins
general, the complexity of oxidative modifications of can be actively transported across the blood–CSF barrier
proteins is still not well understood (12 ). Some modifica- via pinocytosis. Because CNS is normally protected from
tions may be considered reversible and beneficial against the peripheral circulation, the presence of some blood-
attacks caused by ROS; however, some modifications, derived macromolecules in CSF may indicate a disruption
such as carbonylation, are irreversible and may thus be of blood–CSF barrier that is thought to occur during aging
destructive, resulting in alterations in the structure and as well as in some neurodegenerative disorders (17, 28 ).
function of proteins. Carbonyl groups are introduced to Based on the current knowledge of the CSF proteome
proteins by oxidizing amino acid residue side-chain hy- and oxidation of proteins in AD, the aim of this pilot
droxyls into ketone or aldehyde derivatives (13 ). A vari- study was to find out whether alterations in carbonylation
ety of oxidative pathways lead to carbonylation of pro- status of proteins may also be detected in CSF. By use of
teins (12 ). Carbonyl groups can be introduced to proteins 2-D multiplexed oxyblotting, we measured the concentra-
by direct oxidation of lysine, arginine, proline, and thre- tions and carbonylation of major CSF proteins in aging
onine residues, or from the cleavage of peptide bonds by women and men, both neurological controls with no
␣-amidation pathway or by oxidation of glutamyl resi- cognitive decline and AD patients (29 ). These proteins
dues. ROS can also react with other molecules, such as were further identified by in-gel digestion, mass spec-
sugars (glycation/ glycoxidation) and lipids (lipid peroxi- trometry, and database searches.
dation), resulting in generation of reactive carbonyl de-
rivatives and aldehydes, which may then react with Materials and Methods
proteins, followed by formation of protein-bound carbon- study population and samples
yls. Measurement of protein carbonylation is thought to We obtained lumbar CSF samples, collected as a part of
be a good estimation for the extent of oxidative damage of the ongoing biomarker project, from 11 probable AD
proteins associated with various conditions of oxidative patients and 8 neurological control patients showing no
stress, aging, physiological disorders, and disease (14, 15 ). cognitive decline (Table 1). All participants and caregivers
The traditional approach in proteomics, 2-dimensional of dementia patients gave informed consent for participa-
(2-D) gel electrophoresis (2-DE) in combination with mass tion in the study. The study was approved by the local
spectrometry, enables simultaneous identification and ethics committees of University of Kuopio and Kuopio
quantification of multiple protein isoforms (16 ). A num- University Hospital. The diagnosis of probable AD was
ber of studies have reported differences in the cerebrospi- established according to National Institute of Neurologi-
nal fluid (CSF) proteome of AD patients compared with cal and Communicative Disorders and Stroke Alzhei-
controls (17–20 ). Two-dimensional oxyblotting is a mer’s Disease and Related Disorders Association criteria
method that permits examination of both the concentra- (30 ). We used the Mini-Mental State Examination
tions and carbonylation of multiple proteins by slightly (MMSE) to assess the severity of AD (31 ).
modifying 2-DE and combining it with immunoblotting. We performed lumbar puncture according to a stan-
This method can be used to assess the concentrations of dardized protocol. Samples that contained visible blood
oxidative stress via the increase in the formation of were excluded. CSF samples were immediately centri-
protein-bound carbonyls, which are measured based on fuged at 4000g for 10 min. Supernatants were removed,
the covalent reaction of the carbonylated protein side divided into aliquots, and immediately frozen and stored
chains with 2,4-dinitrophenylhydrazine followed by de- at ⫺70 °C until use. In addition, we routinely measured
tection with a specific antidinitrophenyl antibody (21 ). protein content in the supernatant after lumbar puncture
The amount of protein-bound carbonyls is increased in for all samples. Two patients (Table 1, nos. 4 and 12) had
AD brains (14 ), and several oxidized carbonylated pro- abnormal (⬎600 mg/L) protein concentrations in CSF,
teins have recently been characterized by 2-D oxyblotting indicating a possible disruption of the blood–CSF barrier.
in AD plasma and brain (5, 22–25 ).
CSF is a secreted product of choroid plexus and a 2-d oxyblotting
highly specific repository of cellular byproducts, metabo- We measured total protein content by use of a DC protein
lites, neurotransmitters, and proteolytic fragments (26 ). assay reagent set according to the manufacturer’s instruc-
CSF is also connected with the extracellular space in brain, tions (Bio-Rad). Individual CSF samples containing 1 mg
and is protected from the peripheral circulation thanks to protein were precipitated with ice-cold acetone (1:2, vol/
the highly regulated blood–CSF barrier. Therefore, the vol), incubated overnight at ⫺20 °C, and centrifuged
Table 1. Patient demographics.a

Patient ApoE ⑀4 Age, Duration of cognitive MMSE
no. Diagnosis Sex allele years impairment, years score Clinical history
1 Control M ⫺ 61 29 Depression, medication for hypertension
2 Control M ⫹ 52 ND Alcoholism and diabetes, medication for
bipolar disorder
3 Control M ⫺ 65 22 Essential tremor, medication for
depression
4 Control M ⫹ 69 25 Multisystem atrophy, medication for
diabetes and atherosclerosis
5 Control M ⫺ 57 ND Schizophrenia with medication
6 Control M ⫹ 55 29 No medication, nonsmoker, nondrinker
7 Control/MCI F ⫺ 79 21 Stable MCI, medication for chronic
epilepsy and hypertension
8 Control F ⫺ 79 24 Medication for hypertension,
atherosclerosis, and essential tremor
9 MCI/AD F ⫺ 66 0 26 Transient ischemic attack, mild
depression, medication for
hypertension, conversion to AD 4
years after lumbar puncture
10 AD M ⫹ 77 5 14 Despite evidence of dementia, healthy
without any medication
11 AD M ⫹ 90 1 18 Mild hypothyroidism, chronic myeloid
leukemia
12 AD M ⫹ 71 4 23 Dementia-related epilepsy treated with
sodium-valproate, asthma,
hypertension,
13 AD F ⫺ 79 2 20 Medication for hypertension and
atherosclerosis, type 2 diabetes
treated with diet
14 MCI/AD F ⫹ 68 2 21 Despite evidence of mild dementia,
healthy without any medication,
conversion to AD 3 years after lumbar
puncture
15 AD F ⫺ 54 1 22 Smoker, estrogen treatment
16 AD F ⫹ 74 1 16 Despite evidence of mild dementia,
healthy without any medication
17 MCI/AD F ⫹ 78 3 24 Mild medication for depression,
conversion to AD 1 year after lumbar
puncture
18 AD F ⫹ 78 2 22 Medication for type 2 diabetes and
atherosclerosis
19 AD F ⫹ 74 3 23 Medication for atherosclerosis and
hypercholesterolemia
a
AD, probable Alzheimer disease; ApoE ⑀ 4 allele, 1 or 2 copies of apolipoprotein E ⑀ 4 allele; MCI, mild cognitive impairment; ND, not defined. Score range, 0 –30.
(5000g at 4 °C) for 15 min. The precipitates were dissolved an electrotransfer to Hybond-P polyvinylidene difluoride
in rehydration buffer containing 7 mol/L urea, 2 mol/L membranes (Amersham Biosciences) in Towbin transfer
thiourea, 4% CHAPS, 20 mmol/L dithiothreitol, and 1% buffer by use of a Protean II Transfer apparatus (Bio-Rad).
3–10 nonlinear immobilized pH gradient (IPG) buffer We detected proteins as described (29 ). Briefly, we stained
(Amersham Biosciences). We performed isoelectric focus- proteins with Sypro Ruby membrane stain according to the
ing 18-cm 3–10 nonlinear IPG gels (Amersham Bio- manufacturer’s instructions (Bio-Rad) followed by the detec-
sciences) with an IPGphor apparatus (Amersham Bio- tion of hydrazone groups with an antidinitrophenyl anti-
sciences). The samples were focused until 32 500 body (Dako, V0401). We detected fluorescence signals with
volt-hours. fluoroimager Storm 860 (Amersham Biosciences).
To detect carbonylated proteins, IPG gels were treated
with 2,4-dinitrophenyl hydrazine (Sigma) and equili- image analysis
brated according to the in-strip derivatization method We analyzed light intensities of detected proteins on
described by Conrad et al. (32 ). In the 2nd dimension, membranes by use of ImageMaster 2D Elite, version 4.1
proteins were separated on 12.5% sodium dodecyl sulfate (Amersham Biosciences). Two-dimensional images of rel-
polyacrylamide gel with Protean II (Bio-Rad) followed by ative concentrations and carbonylation of proteins were
analyzed separately. The relative concentrations of pro- addition, the largest spot in Fig. 1 represents human
teins are given as normalized values. To compare the serum albumin, which was heavily oxidized. Human serum
concentrations of proteins, the normalized value was albumin was not identified by mass spectrometry in this
justified since the total amount of protein loaded to each study, nor did we measure its carbonylation or concentra-
gel was the same. The extents of protein-bound carbonyls, tions because of its high abundance and low resolution. The
however, are given as light intensity values, since 2-D most abundant identified carbonylated proteins in the CSF
oxyblots could not be normalized according to the total were vitamin D– binding protein (DBP), ␣-1-antitrypsin
CSF carbonylation. In other words, the equal amount of (␣-1-AT), transthyretin (TTR), ␤-trace, apolipoprotein A-I,
total protein was not expected to exhibit the same degree and apolipoproteins E and J (Fig. 1 and Table 2). Although
of carbonylation in individuals. We transferred values of some of the proteins, including apolipoproteins E and J,
light intensities to SPSS software to calculate the degree of were identified as oxidized proteins, they were not reliably
carbonylation. matched during image analysis owing to individual varia-
tion and insufficient resolution in 2-D images and therefore
protein identification not further quantified. These proteins are indicated as “not
We performed 2-DE as described above, except that IPG quantified” in Table 2. We matched and quantified carbony-
gels were not treated with 2,4-dinitrophenyl hydrazine. lation status and concentrations of a total of 22 protein
We manually cut out protein spots from the Sypro Ruby– isoforms in probable AD patients vs neurological controls
stained 2-D gels on a ultraviolet light table. Proteins were exhibiting no cognitive decline, and in women vs men
reduced and alkylated before in-gel digestion with tryp- (Fig. 1 and Table 2).
sin (Sequencing Grade Modified trypsin; Promega), and In patients with probable AD, there was a significant
we analyzed the resulting peptides with automated nano– decrease in the protein concentrations of ␤-trace, TTR
light chromatography/tandem mass spectrometry essen- isoforms, ␭ chain, and 1 unidentified protein compared
tially as described (25 ). The nano–light chromatography/ with controls (Table 2). Generally, the degrees of car-
tandem mass spectrometry runs were performed using bonylation were similar in probable AD and control
either Ultimate™ capillary LC system, Famos autosam- samples. ␭ Chain and 1 unidentified protein exhibited
pler, and Switchos II (Dionex/LC Packings) coupled to increased carbonylation in probable AD (Fig. 1 and Table
QStarPulsar (ABI/MDS-SCIEX) or CapLC coupled to 2, proteins 17 and 18).
Q-TOF Global (Waters/Micromass). Database searches Because our study participants were not sex-matched,
were performed using the publicly available Mascot we also wanted to look for possible differences in the
search engine (http://www.matrixscience.com) against carbonylation and protein concentrations between men
the NCBInr database. and women. Amounts of proteins did not generally differ
between men and women, apart from 1 isoform of ␤-trace
statistical analysis [mean (SE) 390 (130) in all women vs 843 (204) in all
We used the Mann–Whitney U-test to compare normal- men; P ⫽ 0.04; Fig. 1, protein 25]. However, there were
ized volumes of light intensities of the amount of protein- differences in the degrees of carbonylation between
bound carbonyls and the amount of each protein. The women and men. The degrees of carbonylation were
degree of carbonylation was calculated as the ratio of significantly increased in all men compared with all
protein-bound carbonyls to the amount of protein. The women for DBP, apolipoprotein A-I, and ␣-1-AT (Fig. 1
sensitivity for the detection of protein spots is limited and Table 3, proteins 4, 8, 9, 10, and 22). In probable AD
when using image analysis software. For a few spots, zero men compared with probable AD women, DBP exhibited
values were obtained for light intensities although these a higher degree of carbonylation (Fig. 1 and Table 3,
proteins were visible by eye. Because we could not obtain protein 4). The degrees of carbonylation did not vary
background values by image analysis, we replaced zero significantly between control women and men.
values with an empirical minimum value for the calcula-
tion of the degree of carbonylation. The empirical mini- Discussion
mum was the lowest light intensity value of 0.01 detected Carbonylation of several abundant proteins has been
by image analysis. characterized in plasma and brain of AD patients in
All protein identification scores obtained using Mascot different stages compared with defined controls (5, 22–
search engine were significant (P ⬍0.05). 25 ). Therefore, the major aim of our study was to deter-
mine whether carbonylation changes in abundant pro-
Results teins can also be detected in CSF. Altogether, we were
In this pilot study, we found 30 oxidized carbonylated able to assess the carbonylation and concentrations of a
protein spots in the CSF of probable AD patients and total of 22 abundant CSF proteins in probable AD patients
controls and identified 27 of them. In 3 spots, the protein and neurological control patients exhibiting no cognitive
amount was too low for identification, even when spots decline. In general, the extent of carbonylation did not
were pooled from several 2-DE gels before in-gel diges- vary between probable AD patients and controls; how-
tion (Fig. 1 and Table 2, proteins 16, 17, and 18). In ever, 2 proteins did exhibit a higher extent of carbonyla-
Fig. 1. Two-dimensional images of the most abundant proteins stained with Sypro Ruby and major carbonylated proteins detected by
immunoblotting in the CSF of a female AD patient (Table 1, no. 15) and a male control patient (Table 1, no. 5).
Numbers of spots correspond to proteins listed in Tables 2 and 3. AD, probable AD; C, control.
tion in probable AD patients, showing that differences in been shown to deposit in epithelial basement membrane
carbonylation status of proteins are detectable in CSF. It of choroid plexus in AD brain (38 ). It is possible that the
must be noted that in this study, we selected probable AD increased carbonylation and the decreased amount of ␭
patients suffering from mild dementia with relatively chain in CSF may reflect the oxidative stress present in
short durations of clinical impairment. Alterations in CSF probable AD. However, the origin of ␭ chain in CSF is
protein oxidation status may vary according to disease difficult to ascertain, and it may infiltrate to CSF directly
stage and be more pronounced in patients suffering from from blood.
more severe dementia. The prevalent trend we found of the decrease in the
Differences in protein oxidation may provide a diag- protein concentrations of ␤-trace and TTR is consistent
nostic tool in neurodegenerative diseases. A recent study with previous proteomic studies, although we were not
showed that a peptide whose mass corresponded to able to confirm the prominent decreases in the concentra-
oxidized ␤-amyloid 1– 40 could be used to differentiate tions of apolipoprotein A-I (19, 20 ). The similar findings
patients with Lewy body dementia from patients with increase the credibility of the results in the different
Parkinson disease dementia (33 ). In our 2-D study, the studies. Proteomic studies are difficult to compare with
only identified protein exhibiting increased carbonylation each other because of insufficient characterization of the
in probable AD was blood-derived ␭ chain. In addition, study material, the small number of patients involved in
protein concentrations of ␭ chain were significantly de- studies, and variations in experimental designs. The se-
creased in probable AD patients compared with controls. lection of patients and experimental design differed be-
␭ Chains, also known as immunoglobulin ␭ light chains, tween our studies as well (19, 20 ). For example, Puchades
are involved in systemic and localized primary immuno- et al. (20 ) measured the concentrations of CSF proteins
globulin-related (amyloid light chain or AL) amyloidosis in non–age-matched elderly people with MMSE scores of
(34, 35 ). Interestingly, immunohistopathological studies 29 to 30 and older probable AD patients, in contrast to
have indicated that ␭ chains may be associated with our age-matched neurological controls showing no cog-
amyloid deposits (36, 37 ), whereas IgG and IgM have nitive decline (with MMSE scores between 21 and 29) and
Table 2. List of the most abundant oxidized CSF proteins including concentrations and carbonylation of proteins that were
reliably quantified and matched with each other.a
Amount of protein, light intensity Degree of carbonylation
Carbonylated Control, AD, AD vs Control, AD, AD vs

No. proteins Acc. no. mean (SE) mean (SE) P controlsb mean (SE) mean (SE) P controlsb
I. Significant changes
17 ␭ Chain precursor 33395 250 (78) 103 (57) 0.02 2 0.59 (0.14) 1199.15 (668.73) 0.03 1
18 Unidentified 44 (17) 2 (2) 0.01 22 1309.80 (930.85) 2646.28 (1328.936) 0.04 2
25 ␤-Trace 18028972 934 (223) 366 (105) 0.03 2 0.48 (0.16) 790.62 (602.40) 0.10 ⫺
29 TTR 14719497 848 (218) 243 (139) 0.02 2 8519.97 (8517.62) 17 102.46 (7314.15) 0.16 ⫺
30 TTR 999653 3591 (718) 1522 (492) 0.03 2 1.46 (0.45) 2.46 (1.01) 0.56 ⫺
II. Nonsignificant changes
1 DBP 21730549 3232 (424) 2859 (332) 0.41 ⫺ 1.25 (0.29) 1.28 (0.26) 0.87 ⫺
2 DBP 22219267 4872 (953) 3676 (459) 0.30 ⫺ 1.36 (0.49) 1.25 (0.42) 0.93 ⫺
3 DBP 22219267 3596 (665) 3815 (557) 0.79 ⫺ 1.23 (0.21) 0.89 (0.11) 0.16 ⫺
4 DBP 72105 2193 (428) 2306 (316) 0.82 ⫺ 1.34 (0.29) 0.82 (0.14) 0.12 ⫺
5 ␣-1-AT 1942629 977 (193) 695 (130) 0.16 ⫺ 1.06 (0.17) 1.10 (0.28) 0.56 ⫺
7 ␣-1-AT 7245932 2322 (271) 2704 (318) 0.51 ⫺ 0.19 (0.03) 0.15 (0.05) 0.19 ⫺
8 ␣-1-AT P01009 2777 (521) 2856 (252) 0.80 ⫺ 0.40 (0.18) 0.25 (0.07) 0.41 ⫺
9 ␣-1-AT P01009 5918 (554) 3638 (360) 0.16 ⫺ 0.20 (0.05) 0.26 (0.11) 0.56 ⫺
10 ␣-1-AT P01009 4389 (613) 4770 (701) 0.32 ⫺ 0.08 (0.02) 0.10 (0.05) 0.93 ⫺
16 Unidentified 1211 (274) 604 (234) 0.10 ⫺ 0.47 (0.0.17) 0.43 (0.09) 0.68 ⫺
20 ␤-Trace 18028972 6662 (687) 4946 (461) 0.07 ⫺ 0.30 (0.06) 0.30 (0.06) 0.93 ⫺
21 ␤-Trace 18028972 4410 (472) 3439 (508) 0.32 ⫺ 0.28 (0.05) 1.29 (1.02) 0.93 ⫺
22 Proapolipoprotein A-I 178775 2350 (425) 1740 (265) 0.19 ⫺ 0.06 (0.02) 0.05 (0.02) 0.41 ⫺
23 Proapolipoprotein A-I 178775 1368 (207) 898 (194) 0.19 ⫺ 0.02 (0.01) 0.04 (0.02) 0.46 ⫺
26 Cu/Zn-superoxide 1237406 102 (35) 72 (44) 0.20 ⫺ 534.88 (366.20) 2281.25 (1040.58) 0.12 ⫺
dismutase
27 Unidentified 762.84 (220) 442 (179) 0.19 ⫺ 0.34 (0.11) 673.04 (459.12) 0.48 ⫺
28 TTR 339685 13 271 (1684) 10 811 (1001) 0.25 ⫺ 1.50 (0.32) 1.34 (0.14) 0.93 ⫺
III. Proteins that were not quantified
6 ␣ 1-AT precursor 7245932 NQ NQ NQ ⫺ NQ NQ NQ ⫺
11 Apolipoprotein E3 178853 NQ NQ NQ ⫺ NQ NQ NQ ⫺
12 Apolipoprotein E3 1942471 NQ NQ NQ ⫺ NQ NQ NQ ⫺
13 Apolipoprotein E 338305 NQ NQ NQ ⫺ NQ NQ NQ ⫺
14 Apolipoprotein E 178853 NQ NQ NQ ⫺ NQ NQ NQ ⫺
15 Apolipoprotein J 338305 NQ NQ NQ ⫺ NQ NQ NQ ⫺
19 ␤-Trace 410564 NQ NQ NQ ⫺ NQ NQ NQ ⫺
24 Proapolipoprotein A-I 178775 NQ NQ NQ ⫺ NQ NQ NQ ⫺
a
Degree of carbonylation, light intensities of protein-bound carbonyls divided by light intensities of the amount of corresponding protein spots. P value obtained by
Mann–Whitney U-test. No., number of protein spots; Acc. no, identification of proteins indicated as NCBI or SWISS-PROT accession numbers; AD, probable AD; NQ, not
quantified owing to individual variation, unreliable matching, or insufficient resolution.
b
Arrows indicate the concentrations in probable AD compared with controls: –, no difference; 2 or 1, P ⱕ0.05; 22, P ⱕ0.01.
probable AD patients suffering from mild dementia. In We also found significant differences in carbonylation of
addition, the concentrations of proteins were measured proteins between men and women regardless of diagno-
on a Sypro Ruby–stained 2-D gel (20 ) in contrast to our sis. Men had increased carbonylation of an isoform of
Sypro Ruby–stained 2-D membranes. The congruencies in DBP, an isoform of apolipoprotein A-I, and ␣-1-AT.
the results obtained in our studies support the usefulness Increased carbonylation of several isoforms of ␣-1-AT in
of 2-DE when studying CSF proteome profiles in AD plasma from AD patients compared with healthy nonde-
patients. mented controls has been reported (22 ). However, we did
CSF is composed mainly of soluble proteins. Therefore, not find any difference in the carbonylation of ␣-1-AT in
the decrease in the amounts of several brain proteins in CSF between probable AD patients and neurological
CSF may be due to decreased production and/or in- controls. Several possible reasons may explain differences
creased aggregation in brain that may occur during the between our studies, such as the selection of study par-
early stages of AD. Notably, decreased concentrations of ticipants, e.g., numbers of women and men. For example,
proteins in CSF may also be a secondary event to in- our study participants were not sex-matched; based on
creased oxidative stress, since excessive carbonylation is our findings on sex differences in the carbonylation of
generally thought to lead to enhanced aggregation of proteins, this may have influenced our results when
proteins. comparing protein carbonylation between AD patients
Previous studies have revealed sex-specific differences and controls. Moreover, variations in experimental design
related to oxidative stress and antioxidant defenses (6 – 8 ). may have contributed to the differences in the results. For
Table 3. Differences in the degrees of carbonylation between men and women.a

Degree of carbonylation
Women, n ⴝ 10, Men, n ⴝ 9,

No. Carbonylated proteins mean (SE) mean (SE) P Men vs womenb
4 DBP 0.67 (0.09) 1.45 (0.25) 0.01 11
7 ␣ -1-AT precursor 0.14 (0.05) 0.21 (0.03) 0.09 ⫺
8 ␣ -1-AT precursor 0.16 (0.04) 0.49 (0.15) 0.01 11
9 ␣ -1-AT precursor 0.04 (0.01) 0.15 (0.06) 0.02 1
10 ␣ -1-AT precursor 0.14 (0.04) 0.34 (0.12) 0.05 1
22 Proapolipoprotein A-I 0.03 (0.02) 0.08 (0.02) 0.04 1
Control women, n ⴝ 2, Control men, n ⴝ 6, p Men vs Women
mean (SEM) mean (SEM)
4 DBP 1.00 (0.09) 1.45 (0.38) 0.18 ⫺
7 ␣ -1-AT precursor 0.19 (0.07) 0.20 (0.04) 1.00 ⫺
8 ␣ -1-AT precursor 0.12 (0.04) 0.50 (0.23) 0.10 ⫺
9 ␣ -1-AT precursor 0.03 (0.003) 0.10 (0.03) 0.10 ⫺
10 ␣ -1-AT precursor 0.09 (0.05) 0.24 (0.07) 0.10 ⫺
22 Proapolipoprotein A-I 0.02 (0.02) 0.07 (0.02) 0.31 ⫺
AD women, n ⴝ 8, AD men, n ⴝ 3, p Men vs Women
mean (SEM) mean (SEM)
4 DBP 0.59 (0.09) 1.44 (0.06) 0.01 11
7 ␣ -1-AT precursor 0.12 (0.06) 0.22 (0.07) 0.15 ⫺
8 ␣ -1-AT precursor 0.17 (0.05) 0.47 (0.15) 0.07 ⫺
9 ␣ -1-AT precursor 0.05 (0.01) 0.25 (0.19) 0.10 ⫺
10 ␣ -1-AT precursor 0.15 (0.04) 0.55 (0.37) 0.22 ⫺
22 Proapolipoprotein A-I 0.03 (0.02) 0.10 (0.04) 0.10 ⫺
a
Degree of carbonylation, light intensities of protein-bound carbonyls divided by light intensities of the amount of corresponding protein spot. P value obtained by
Mann–Whitney U-test. No., number of protein spots.
b
Arrows indicate a comparison of the degree of oxidation between men and women: ⫺, no difference, 1, P ⱕ0.05, 11, P ⱕ0.01.
example, Choi et al. (22 ) detected the concentrations of exploration of the deep CSF proteome presents a number
plasma proteins by staining one 2-DE gel and electro- of challenges. For example, the protein content in blood is
transferred plasma proteins from another 2-D gel to a much higher than in CSF, and thus even a minor contam-
membrane for oxyblotting. In contrast, we applied a ination during collection may influence the results (39 ).
multiplexed oxyblotting approach that enables the detec- Therefore, in CSF proteomic studies as well as in our
tion of both the concentrations and carbonylation of study, the possibility of blood contamination is common
proteins using a single membrane (29 ). Carbonylation and difficult to control. Although we excluded samples
changes in CSF proteome may not be analogous to those that contained visible blood and centrifuged the samples,
in plasma even if matching the selection criteria of study and all patients except 2 had CSF protein concentrations
participants and experimental designs. Finally, we found in the normal interval, suggesting that they did not have
that DBP was more carbonylated in probable AD men any major damage to the blood– brain barrier, minor
compared with probable AD women. Taken together, our blood contaminations cannot be excluded. Moreover, ex-
results indicate that the extent of protein carbonylation amination of the brain-derived deep CSF proteome would
may vary between men and women, and that the impact require either the removal of the most abundant proteins
of sex on protein oxidation needs to be further evaluated. (albumin and immunoglobulins) that account for ⬃80% of
In addition, our results emphasize the importance of the total protein content in CSF or prefractionation of the
having sex-matched patients when studying protein car- sample (40 ). One problem with the removal of abundant
bonylation in disease states. CSF proteins is the concomitant omission of some addi-
Recent studies suggest that oxidized proteins may offer tional less abundant proteins. A further limitation is that
a diagnostic tool for differentiation of neurodegenerative the volume of CSF needed for the studies of less abundant
diseases (33 ). This is the first proteomic study on carbony- proteins using 2-DE is high. Nevertheless, our results
lation of CSF proteins using 2-D multiplexed oxyblotting. indicate that there are differences in carbonylation sta-
Only 2 proteins, ␭ chain and 1 unidentified protein, tuses of CSF proteins and further studies are warranted. It
exhibited a higher extent of carbonylation in probable AD is likely that additional changes in CSF protein carbony-
patients compared with controls. Hereafter, it will be lation will be revealed by methodologically modifying
important to go deeper into the proteome and to study 2-D oxyblotting or by using quantitatively more sensitive
carbonylation of less abundant CSF proteins. Yet the methods.
We thank Seija Hynynen for her skillful technical assis- 15. Levine RL, Williams JA, Stadtman ER, Shacter E. Carbonyl assays
tance. The study was supported by the National Tech- for determination of oxidatively modified proteins. Methods Enzy-
nology Agency of Finland (Grant 201495), the Academy of mol 1994;233:346 –57.
16. Görg A, Weiss W, Dunn MJ. Current two-dimensional electrophore-
Finland, Kuopio University Hospital (EVO Grant
sis technology for proteomics. Proteomics 2004;4:3665– 85.
5772722), University of Kuopio, the Nordic Centre of 17. Mattila KM, Pirttilä T, Blennow K, Wallin A, Viitanen M, Frey H.
Excellence, the Kuopio University Foundation, the Finn- Altered blood-brain-barrier function in Alzheimer’s disease? Acta
ish Cultural Foundations of Northern and Southern Savo, Neurol Scand 1994;89:192– 8.
the Farmos Research Foundation, the Foundation for 18. Choe LH, Dutt MJ, Relkin N, Lee KH. Studies of potential
Advanced Technology of Eastern Finland, Helsingin Sa- cerebrospinal fluid molecular markers for Alzheimer’s disease.
nomain 100-vuotissäätiö Foundation, Maire Taponen Electrophoresis 2002;23:2247–51.
19. Davidsson P, Westman-Brinkmalm A, Nilsson CL, Lindbjer M,
Foundation, and the Alzheimer Research Society of
Paulson L, Andreasen N, et al. Proteome analysis of cerebrospinal
Finland. fluid proteins in Alzheimer patients. Neuroreport 2002;13:611–5.
20. Puchades M, Hansson SF, Nilsson CL, Andreasen N, Blennow K,
References Davidsson P, et al. Proteomic studies of potential cerebrospinal
1. Navarro A, Sanchez Del Pino MJ, Gomez C, Peralta JL, Boveris A. fluid protein markers for Alzheimer’s disease. Brain Res Mol Brain
Behavioral dysfunction, brain oxidative stress, and impaired mito- Res 2003;118:140 – 6.
chondrial electron transfer in aging mice. Am J Physiol Regul Integr 21. Nakamura A, Goto S. Analysis of protein carbonyls with 2,4-
Comp Physiol 2002;282:R985–92. dinitrophenyl hydrazine and its antibodies by immunoblot in
2. Navarro A, Emerit J, Edeas M, Bricaire F. Mitochondrial enzyme two-dimensional gel electrophoresis. J Biochem (Tokyo) 1996;
activities as biochemical markers of aging, neurodegenerative 119:768 –74.
diseases and oxidative stress. Mol Aspects Med 2004;25:37– 22. Choi J, Malakowsky CA, Talent JM, Conrad CC, Gracy RW. Identi-
48. fication of oxidized plasma proteins in Alzheimer’s disease.
3. Mariani E, Polidori MC, Cherubini A, Mecocci P. Oxidative stress in Biochem Biophys Res Commun 2002;293:1566 –70.
brain aging, neurodegenerative and vascular diseases: an over- 23. Yu HL, Chertkow HM, Bergman H, Schipper HM. Aberrant profiles
view. J Chromatogr B Analyt Technol Biomed Life Sci 2005;827: of native and oxidized glycoproteins in Alzheimer plasma. Pro-
65–75. teomics 2003;3:2240 – 8.
4. Pereira C, Agostinho P, Moreira PI, Cardoso SM, Oliveira CR. 24. Choi J, Rees HD, Weintraub ST, Levey AI, Chin LS, Li L. Oxidative
Alzheimer’s disease-associated neurotoxic mechanisms and neu- modifications and aggregation of Cu/Zn superoxide dismutase
roprotective strategies. Curr Drug Targets CNS Neurol Disord associated with Alzheimer’s and Parkinson’s diseases. J Biol
2005;4:383– 403. Chem 2005;280:11648 –55.
5. Butterfield DA. Proteomics: a new approach to investigate oxida- 25. Korolainen MA, Goldsteins G, Nyman TA, Alafuzoff I, Koistinaho J,
tive stress in Alzheimer’s disease brain. Brain Res 2004;1000: Pirttilä T. Oxidative modification of proteins in the frontal cortex of
1–7. Alzheimer’s disease brain. Neurobiol Aging 2006;27:42–53.
6. Mendoza-Núñez VM, Sánchez-Rodriguez MA, Retana-Ugalde R, 26. Romeo MJ, Espina V, Lowenthal M, Espina BH, Petricoin EF 3rd,
Vargas-Guadarrama LA, Altamirano-Lozano MA. Total antioxidant Liotta LA. CSF proteome: a protein repository for potential biomar-
levels, gender, and age as risk factors for DNA damage in ker identification. Expert Rev Proteomics 2005;2:57–70.
lymphocytes of the elderly. Mech Ageing Dev 2001;122:835– 47. 27. Reiber H. Dynamics of brain-derived proteins in cerebrospinal
7. Schuessel K, Leutner S, Cairns NJ, Muller WE, Eckert A. Impact of fluid. Clin Chim Acta 2001;310:173– 86.
gender on upregulation of antioxidant defence mechanisms in 28. Skoog I, Wallin A, Fredman P, Hesse C, Aevarsson O, Karlsson I,
Alzheimer’s disease brain. J Neural Transm 2004;111:1167– 82. et al. A population study on blood-brain barrier function in 85-year-
8. Liu H, Harrell LE, Shenvi S, Hagen T, Liu RM. Gender differences olds: relation to Alzheimer’s disease and vascular dementia.
in glutathione metabolism in Alzheimer’s disease. J Neurosci Res Neurology 1998;50:966 –71.
2005;79:861–7. 29. Korolainen MA, Goldsteins G, Alafuzoff I, Koistinaho J, Pirttilä T.
9. Jorm AF, Korten AE, Henderson AS. The prevalence of dementia: a Proteomic analysis of protein oxidation in Alzheimer’s disease
quantitative integration of the literature. Acta Psychiatr Scand brain. Electrophoresis 2002;23:3428 –33.
1987;76:465–79. 30. McKhann G, Drachman D, Folstein M, Katzman R, Price D, Stadlan
10. Turner RS. Alzheimer’s disease in man and transgenic mice: EM. Clinical diagnosis of Alzheimer’s disease: report of the
females at higher risk. Am J Pathol 2001;158:797– 801. NINCDS-ADRDA Work Group under the auspices of Department of
11. Rocca WA, Hofman A, Brayne C, Breteler MM, Clarke M, Copeland Health and Human Services Task Force on Alzheimer’s Disease.
JR, et al. Frequency and distribution of Alzheimer’s disease in Neurology 1984;34:939 – 44.
Europe: a collaborative study of 1980 –1990 prevalence findings: 31. Folstein MF, Folstein SE, McHugh PR. “Mini-mental state”: a
the EURODEM-Prevalence Research Group. Ann Neurol 1991;30: practical method for grading the cognitive state of patients for the
381–90. clinician. J Psychiatr Res 1975;12:189 –98.
12. Dalle-Donne I, Aldini G, Carini M, Colombo R, Rossi R, Milzani A. 32. Conrad CC, Choi J, Malakowsky CA, Talent JM, Dai R, Marshall P,
Protein carbonylation, cellular dysfunction, and disease progres- et al. Identification of protein carbonyls after two-dimensional
sion. J Cell Mol Med 2006;10:389 – 406. electrophoresis. Proteomics 2001;1:829 –34.
13. Berlett BS, Stadtman ER. Protein oxidation in aging, disease, and 33. Bibl M, Mollenhauer B, Esselmann H, Lewczuk P, Klafki HW,
oxidative stress. J Biol Chem 1997;272:20313– 6. Sparbier K, et al. CSF amyloid-beta-peptides in Alzheimer’s dis-
14. Smith MA, Sayre LM, Anderson VE, Harris PL, Beal MF, Kowall N, ease, dementia with Lewy bodies and Parkinson’s disease de-
et al. Cytochemical demonstration of oxidative damage in Alzhei- mentia. Brain 2006;129:1177– 87.
mer disease by immunochemical enhancement of the carbonyl 34. Hamidi Asl K, Liepnieks JJ, Nakamura M, Benson MD. Organ-
reaction with 2,4-dinitrophenylhydrazine. J Histochem Cytochem specific (localized) synthesis of Ig light chain amyloid. J Immunol
1998;46:731–5. 1999;162:5556 – 60.
35. Schröder R, Linke RP. Cerebrovascular involvement in systemic AA 38. Serot JM, Bene MC, Faure GC. Comparative immunohistochemical
and AL amyloidosis: a clear haematogenic pattern. Virchows Arch characteristics of human choroid plexus in vascular and Alzhei-
1999;434:551– 60. mer’s dementia. Hum Pathol 1994;25:1185–90.
36. Eikelenboom P, Stam FC. Immunoglobulins and complement 39. You JS, Gelfanova V, Knierman MD, Witzmann FA, Wang M, Hale
factors in senile plaques: an immunoperoxidase study. Acta JE. The impact of blood contamination on the proteome of
Neuropathol (Berl) 1982;57:239 – 42. cerebrospinal fluid. Proteomics 2005;5:290 – 6.
37. Kametani F, Tanaka K, Sato M, Haga S, Saitoh T, Shinoda T, et al. 40. Righetti PG, Castagna A, Antonucci F, Piubelli C, Cecconi D,
A monoclonal antibody Hy20 –54-16 –3L to lambda (␭) light chain Campostrini N, et al. Proteome analysis in the clinical chemis-
of human immunoglobulin reacts with amyloid in Alzheimer’s try laboratory: myth or reality? Clin Chim Acta 2005;357:123–
disease brain. Neurosci Lett 1990;117:62–7. 39.
666 – 672 (2007) Cancer Diagnostics
MESOMARKTM: A Potential Test for Malignant

Pleural Mesothelioma
Heather L. Beyer, Ryan D. Geschwindt, Curtis L. Glover, Ly Tran,
Ingegerd Hellstrom, Karl-Erik Hellstrom, M. Craig Miller, Thorsten Verch,
W. Jeffrey Allard,* Harvey I. Pass, and Niranjan Y. Sardesai
Background: Soluble mesothelin-related peptides (SMRP) Conclusions: The MESOMARK assay is analytically
have been reported to be potential biomarkers for ma- robust and may be useful for the detection and manage-
lignant pleural mesothelioma (MPM). We report analyt- ment of mesothelioma.
ical and preliminary clinical studies of MESOMARKTM, © 2007 American Association for Clinical Chemistry
a quantitative assay for SMRP.
Methods: The MESOMARK assay is a 2-step immu- Malignant mesothelioma is a highly aggressive neoplasm
noenzymatic assay in an ELISA format with a 6-point with poor prognosis. Mesothelioma accounted for ⬃1.5%
calibration curve (0 –32 nmol/L). We assessed analytical of the 171 900 total lung and bronchus malignancies in the
imprecision, analyte stability, and analytical interfer- US in 2003 (1, 2 ), and epidemiological studies have estab-
ences. We measured SMRP by this assay in 409 appar- lished exposure to asbestos fibers as the primary cause
ently healthy individuals (reference interval study), 177 (3–5 ). Although malignant mesothelioma remains a rela-
patients with nonmalignant conditions, and 500 cancer tively uncommon malignancy in the US, it continues to be
patients, including 88 with MPM. an important cause of mortality in numerous areas world-
Results: The limit of detection was 0.16 nmol/L. At 2–19 wide, e.g., England, Wales, continental Europe, and
nmol/L, intraassay imprecision (CV) was 1.1%–5.3%, Australia.
and total imprecision was 4.0%–11.0%. The mean dilu- Because the disease is asymptomatic in early stages
tion recovery for 5 samples was 109% (range, 99%– and definitive diagnosis is difficult, detection typically
113%). No interference was seen from added bilirubin occurs at a late stage, and recurrence rates, even in
(200 mg/L), hemoglobin (500 mg/L), triglycerides (30 patients with surgically resected tumors, are very high.
g/L), chemotherapeutic agents, or other tested sub- Treatment efficacy is routinely assessed by clinical symp-
stances. Recombinant mesothelin was stable in serum toms and costly radiologic imaging techniques with lim-
upon freeze/thaw at ⴚ70 °C and upon storage for at least ited sensitivity and specificity (6 ). Thus, there has been an
7 days at 2– 8 °C. The 99th percentile of the reference increasing need for a simple diagnostic blood test for
group was 1.5 nmol/L [95% confidence interval (CI), screening asbestos-exposed patients as well as for moni-
1.2–1.6 nmol/L; n ⴝ 409], and mean SMRP was signifi- toring response to treatment.
cantly higher in sera from patients with MPM (7.5 The MESOMARK™ assay measures soluble molecules
nmol/L; 95% CI, 2.8 –12.1 nmol/L; n ⴝ 88). SMRP was that are related to the mesothelin/megakaryocyte poten-
increased in 52% and 5% of MPM patients and asbestos- tiating factor (MPF)1 family of proteins and recognized by
exposed individuals, respectively. Concentrations in the monoclonal antibody OV569 (7 ). Mesothelin and MPF
other nonmalignant and malignant conditions were are synthesized together as a 69-kDa precursor from
similar to those in healthy controls. which an N-terminal 31-kDa fragment is released as MPF,
and mesothelin is the 40-kDa C-terminal membrane-
bound fragment. Immunohistochemical studies revealed
Research and Development Division, Fujirebio Diagnostics, Inc., Malvern,
PA.
* Address correspondence to this author at: Research and Development
1
Division, Fujirebio Diagnostics, Inc., 201 Great Valley Pkwy., Malvern, PA Nonstandard abbreviations: MPF, megakaryocyte potentiating factor;
19355. Fax 610-240-3803; e-mail allardj@fdi.com. SMRP, soluble mesothelin-related peptides; HAMA, human antimouse anti-
Received August 31, 2006; accepted December 22, 2006. body; RF, rheumatoid factor; AUC, area under the curve; CI, confidence
Previously published online at DOI: 10.1373/clinchem.2006.079327 interval.
666
tissue expression of mesothelin in a variety of tissues, the primary calibrators and controls based on absorbance
most notably in mesothelioma, lung, ovarian, endome- measurement in the MESOMARK assay, allowing for
trial, and pancreatic tumors (8 ). At least 3 known variants traceability to the reference preparation.
of the mesothelin family have been reported in the liter-
ature (9, 10 ). Although early studies indicated that me- samples
sothelin was not soluble (11 ), recent work demonstrated Serum samples for use in analytical studies were sup-
variants 1 and 3 to be soluble, with variant 1 being the plemented with recombinant OV569-reactive antigen,
predominant form detected in serum (10, 12 ). Variant 3 whereas serum samples for use in the clinical studies were
contains a 3⬘ frameshift leading to an extended C-termi- not supplemented. The latter were collected from appar-
nus, and variant 2 contains a 24-bp insertion and can be ently healthy volunteers (n ⫽ 409), as well as from
detected only at the mRNA level. patients with several malignant and nonmalignant con-
Despite the expression of membrane-bound mesothe- ditions. Malignant conditions included mesothelioma
lin in different tumors, these soluble mesothelin-related (n ⫽ 88), ovarian cancer (n ⫽ 111), lung cancer (n ⫽ 174),
peptides (SMRP) have proven to be a promising cancer colon cancer (n ⫽ 50), pancreatic cancer (n ⫽ 52), and
biomarker in the sera of patients with tumors of mesothe- endometrial cancer (n ⫽ 25). Nonmalignant conditions
lial origin (7, 10 ). We report here analytical validation included hypertension (defined as a blood pressure
studies of the MESOMARK assay, establishing assay ⬎150/90 mmHg; n ⫽ 100), exposure to asbestos (n ⫽ 61),
performance and its potential utility in a clinical patient and endometriosis (n ⫽ 16). Clinical samples were col-
population. lected retrospectively based on reported diagnosis and
blinded to the operator. Invalid results were addressed
Materials and Methods by retests of the sample in question. These samples were
mesomark assay obtained from commercial sources (Biochemed) in 2004
The MESOMARK assay (Fujirebio Diagnostics, Inc.) uses except for samples from mesothelioma patients and
antibody 4H3, which binds to mesothelin variants 1, 2, asbestos-exposed individuals, which were kindly sup-
and 3, and antibody 569, which binds to variant 1 and 3 plied by Dr. Harvey Pass (Karmanos Cancer Institute,
(12 ). We performed the assay according to the manufac- Wayne State University, Detroit, MI). Patients with a
turer’s instructions. Briefly, patient serum samples were histologically confirmed diagnosis of mesothelioma seen
diluted 1:101 with the assay diluent provided and applied at the Karmanos Cancer Institute gave informed consent
in duplicate to a microwell plate precoated with anti- to have serum, plasma, and normal and tumor tissue
body 4H3. After incubation for 1 h at room temperature, samples taken on the day of operative intervention for
plates were washed, and antibody OV569-HRP conju- their tumors. These cases were performed both at the
gate was added for 1 h. After a wash step, TMB substrate Karmanos Cancer Institute and the National Cancer Insti-
was added to the reaction wells for 15 min, and then tute in Bethesda, Maryland, between 1995 and 2003. All
100 ␮L of stop solution was added. The absorbance at blood samples were drawn before anesthesia or before
450 nm was used to quantify the SMRP concentrations by surgery in the clinic, when the patient was examined by
comparison with a 6-point calibration curve established the physician. The asbestos-exposed population consisted
by a 4-parameter logistic fit using Softmax Pro software of patients seen at the Center for Environmental Medicine
(Molecular Devices). Unless otherwise noted, all studies in Southfield, Michigan, who, after giving informed con-
were performed with representative reagent sets from 2 sent, donated urine and serum. These asbestos-exposed
independent lots. Four operators were trained during a patients also filled out a demographics questionnaire and
5-day, 10-run training study before the start of analytical gave permission for analysis of their pulmonary function
MESOMARK tests. tests and radiographic images. All patients provided
informed consent, and all procedures and protocols were
primary antigen and assay standardization approved by the institutional review board.
MESOMARK values are expressed as nmol/L and are
related to a reference preparation maintained by Fujirebio analytical methods
Diagnostics, Inc. The reference preparation is a recombi- Analytical Sensitivity. Calibrator A (the assay diluent) was
nant antigen (reactive with both OV569 and 4H3) pro- assayed in replicates of 25 for each reagent set lot, and the
duced with a stably transfected cell line, OV569-, immu- mean absorbance plus 2 SD was determined and com-
noaffinity purified, and quantified by amino acid analysis. pared with a calibration curve (prepared in duplicate).
This cell line has been described in detail elsewhere (10 ). The detection limit (also called the limit of the blank) was
Antigen concentrations were measured in nanomoles per calculated based on the linear segment connecting the A
liter. Primary calibrators and controls were referenced to and B calibrators.
this preparation and ranged from 0 to 32 nmol/L to cover
the majority of the expected range in patient populations. Imprecision. Within-run and total imprecision values were
Subsequent lots of antigen (used to manufacture the evaluated according to NCCLS Protocol EP5-A (13 ). Two
reagent set calibrators and controls) were matched against replicates each of 3 panels were assayed in 2 separate runs
668 Beyer et al.: MESOMARK Test for Malignant Pleural Mesothelioma
on each of 20 days, at 2 separate clinical sites. The antigen stability studies

low-end imprecision was evaluated at a single clinical Serum Processing. We tested 30 serum samples supple-
site by assaying low-end panels (40 runs in 20 days). mented with 2 different concentrations of OV569-reactive
Panels consisted of defibrinated human plasma sup- antigen and 20 nonsupplemented samples on days 0, 3,
plemented with OV569-reactive antigen (range, 1.26 – and 7 (intermediate storage at 2– 8 °C). Values were
19.02 nmol/L). Data from the study were analyzed with compared over the time course. Two different collection
the Analyze-It software package (Analyze-it Software, procedures were also compared (red top tubes vs serum
Ltd.). separator tubes; Becton Dickinson)
Dilution Linearity. Dilution linearity was assessed with a Sample Freeze/Thaw. Aliquots from 12 of the above samples
single reagent set lot and modeled after the NCCLS and 1 mesothelioma patient sample were subjected to
protocol EP6-A (13 ). Five serum samples from apparently 1–10 freeze/thaw cycles at ⫺70 °C, and SMRP values for
healthy individuals were supplemented with OV569- the frozen/thawed samples were compared with the fresh
reactive antigen to ⬎25 nmol/L, followed by dilutions sample values.
ranging from 1:1.1 to 1:20. Expected and observed SMRP
values were compared for each dilution. Storage Temperature. Five serum separator tubes of blood
were collected from each of 20 healthy volunteers. One
Recovery. To each of 5 sera we added OV569-reactive tube of each set was processed on the day of collection,
antigen at concentrations covering the range of the cali- and the serum was stored at ⫺70 °C as the comparison
bration curve. Using the same reagent set lot, we com- (presumably stable) material. The remaining blood tubes
pared observed and expected sample values. were stored unprocessed at 2– 8 °C or at 37 °C and pro-
cessed on days 1, 2, 4, and 7 followed by storage at ⫺70 °C
Interference Studies. All interference studies were per- until further analysis. All samples were analyzed in 1
formed using 1 reagent set lot. Potentially interfering batch.
compounds were added at final concentrations ⬃10-fold
statistical analyses of clinical samples
higher than the expected peak plasma concentrations
Nonparametric ROC analyses were performed comparing
(Cmax) to aliquots of 5 independent serum samples con-
the MESOMARK results in the mesothelioma patients to
taining OV569-reactive antigen, whereas control samples
those of the healthy volunteers alone and to the benign
were treated with the appropriate vehicle solution.
and other nonmalignant conditions combined, and the
Chemotherapeutics included Gemzar (gemcitabine-
areas under the curves (AUCs) were calculated. The 95%
HCl) and Alimta, both purchased from Eli Lilly and
confidence interval (CI) was also determined for the
Company, and carboplatin and cisplatin (cis-diammine-
AUCs. Equality of the median MESOMARK assay values
platinum dichloride), both purchased from Sigma
in the healthy volunteers and the different stages and
Chemical.
histological subtypes of the mesothelioma patients were
Naturally occurring serum components included he- compared using a nonparametric k-sample test and the ␹2
moglobin (5 g/L), triglycerides (Liposyn, Abbott Diagnos- statistic.
tic Division; 30 g/L), bilirubin (Sigma; 200 mg/L), and
added protein (1.5% bovine serum albumin and 0.5% Results
bovine gamma globulin, Sigma). Human hemoglobin was analytical studies
isolated from whole blood cellular pellets that were lysed Within-run imprecision (CV) of the MESOMARK assay
with an equivalent volume of water followed by centrif- ranged from 1.1% to 6.4% between the 2 clinical sites
ugation to remove cellular debris. tested (total imprecision, ⱕ11.0%; see Table 1 in the Data
To assess human antimouse antibody (HAMA) and Supplement that accompanies the online version of this
rheumatoid factor (RF) interference, 1 serum specimen article at http://www.clinchem.org/content/vol53/issue4),
collected from an otherwise healthy volunteer, 10 speci- and the limit of detection (limit of the blank) was 0.16
mens that were positive for HAMA, and 5 specimens that nmol/L for both lots tested (see Table 2 in the online Data
were positive for RF were supplemented with 3 different Supplement). Dilution linearity was demonstrated across
concentrations of OV569-reactive antigen. All samples 1–27 nmol/L, and linear regression analysis resulted in
were assayed, and the observed values of the supple- the following: observed ⫽ 0.99(expected) ⫹ 0.77; R2 ⫽
mented samples were compared with the nonsupple- 0.986. The total mean percent recovery of added antigen
mented samples. in serum samples (see Table 3 in the online Data Supple-
A potential prozone/hook effect was tested by adding ment) across all of the dilutions tested was 107%, and
high concentrations of recombinant antigen to 3 serum individual recoveries ranged from 94% to 127%.
samples (5991, 7881, and 10 291 nmol/L) followed by After addition of potential interferents, mean results
2-fold serial dilutions to 1:1024. were 96%–101% of expected (Table 1).
exposed individuals (Fig. 1). Median MESOMARK values

Table 1. Interfering substances.a
in the preoperative samples from mesothelioma patients
Measured concentration
Test as % of expected were significantly higher than values in samples from
Test condition concentration concentration (range)b healthy individuals (Table 2; P values ⱕ0.0001). In MPM,
Native serum components results showed no clear relationship to stage of MPM or
Bilirubin 200 mg/L 100 (96–102) its histologic type (Tables 2 and 3).
Hemoglobin 500 g/L 100 (85–111) ROC curves (Fig. 2) comparing normal individuals
Total protein 120 g/L 99 (94–105) with preoperative mesothelioma samples yielded an AUC
Triglycerides 30 g/L 101 (95–108) of 87.4% (95% CI, 82.8%–92.0%), whereas a comparison of
Chemotherapeutic agents asbestos-exposed patients with preoperative mesotheli-
Cisplatin 1.53 g/L 96 (85–110) oma samples yielded an AUC of 80.6% (95% CI, 73.7%–
Carboplatin 0.34 g/L 101 (93–107)
87.4%). An ROC curve comparing preoperative mesothe-
Alimta 1.3 g/L 99 (89–106)
liomas vs all nonmesothelioma samples (healthy, benign,
Gemcitabine 300 ␮mol/L 99 (92–108)
and other cancers) yielded an AUC of 81.0% (95% CI,
Potentially interfering clinical
condition 75.5%– 86.5%; data not shown).
HAMAc 1281 ␮g/L 99 (86–112) As shown in Table 4, 99% of serum collected from
RFd 288 kIU/L 105 (100–112) apparently healthy individuals demonstrated SMRP val-
a
Different potential interfering substances that may be found in patient blood
ues ⱕ1.5 nmol/L, whereas SMRP values were ⬎1.5
were added together with recombinant antigen into serum samples. nmol/L in 52% of mesothelioma patients. In other cancers
b
For native serum components and chemotherapeutic agents, value ⫽ 100 䡠 (ovarian, pancreatic, colon, and endometrial), MESO-
measured value with test condition/measured value without test condition. For MARK values were increased in ⬍10% of the samples
HAMA and RF samples, values ⫽ 100 䡠关(measured value of sample with added
tested. In 83% of samples from lung cancer patients and
HAMA or RF sample with added antigen) ⫺ (measured original value of
sample)兴/关(measured value of normal serum with added antigen) ⫺ (measured 95% of samples from asbestos-exposed individuals, SMRP
original value of normal serum)兴. was ⱕ1.5 nmol/L. Statistical performance of MESO-
c
Pool of 10 HAMA-positive samples. MARK at the chosen cutoff was determined in compari-
d
Pool of 5 RF-positive samples. son with different patient groups (Table 4).
Discussion
None of the 3 serum samples tested exhibited this Our analytical studies indicate that the MESOMARK
effect (data not shown). Most importantly, the highest assay is robust, with freedom from interference from a
concentration of analyte tested for the prozone effect range of potential interferents; an analyte that is stable
(10 291 nmol/L) was ⬎60-fold greater than the highest (e.g., with freeze/thaw cycles and processing); assay
measured value in a mesothelioma patient sample (170 imprecision (total CV) ⱕ11%; detection limit (0.16
nmol/L) to date. nmol/L) well below the upper limit of the reference
antigen stability
Freeze/thaw studies (12 samples) yielded mean values of
89%–98% of the “fresh control” samples (⫺70 °C sample)
across 10 freeze/thaw cycles, with no trends observed
(see Table 4 in the online Data Supplement). For samples
stored at 2– 8 °C for up to 7 days, mean recoveries across
all samples tested (n ⫽ 51) were 94% and 92% on days 3
and 7 of testing, respectively (data not shown).
Neither a storage temperature of 37 °C nor hemolysis
resulting from storing unprocessed blood samples over
extended periods of time (at 2– 8 °C and 37 °C) had a
significant effect on measured SMRP. For the 10 samples
in each set at 2– 8 °C (37 °C), mean recoveries were as
follows, as a percentage of the initial values: day 1, 99%
(96%); day 2, 99% (87%); day 4, 101% (80%); day 7, 95%
(103%). No trends were observed, and the variations in
recoveries were within the imprecision profile for the
assay.
Fig. 1. MESOMARK values across a cross-section of samples.

clinical validation
Samples from patients with different diagnosed cancers were tested for SMRP
SMRP was higher in 88 mesothelioma patients than in concentrations using the MESOMARK assay. Ca, cancer; Exp, exposure; Pre-op
other patient groups and controls, including 61 asbestos- mesos, preoperative mesothelioma samples.
Table 2. SMRP concentrations in healthy controls and mesothelioma patients.a

Diagnosis n MESOMARK, nmol/L Individuals with MESOMARK within indicated intervals, %
Mean (SD) Median <1.5 nmol/L 1.5–3.0 nmol/L 3.1–10.0 nmol/L >10.0 nmol/L
Healthy 409 0.4 (0.4) 0.3 99 1.2 0.0 0.0
Mesothelioma 88 7.5 (21.9) 1.6 48 23 16 14
Mesothelioma histology
Epithelioid 59 7.0 (15.6) 1.9 41 25 19 15
Biphasic 21 2.1 (2.7) 1.0 67 14 14 4.8
Sarcomatoid 8 25.1 (59.2) 1.6 50 25 0.0 25
Mesothelioma stage
Stage I 14 5.7 (17.0) 1.5 64 29 0.0 7.1
Stage II 22 11.9 (35.9) 1.8 46 23 18 14
Stage III/IV 52 6.1 (14.3) 1.9 44 21 19 15
a
Samples from mesothelioma patients were collected before surgical intervention.
interval; linearity to a concentration (27 mmol/L) nearly For example, increased SMRP concentrations were re-
20 times the upper limit of the reference interval; and ported in an Australian mesothelioma patient population,
essentially quantitative recovery of added analyte. which included patients who had been occupationally
The clinical findings in our study are similar to those exposed to asbestos (7 ). Interestingly, of 7 patients with
reported by others using different assays (7, 14, 15 ). In our increased SMRP values, 3 were later diagnosed with
study, 83% of lung cancer patients and 95% of asbestos- mesothelioma, and 1 was found to have lung cancer (7 ).
exposed individuals exhibited serum SMRP concentra- Increased SMRP concentrations were found in sera from
tions below the upper limit of the reference interval patients with mesothelioma of the epithelial subtype, but
(defined as the 99th percentile value of a distribution of not in sera from patients with sarcomatoid mesothelioma
normal healthy individuals). In contrast, 52% of mesothe- in this study and in studies by other groups (7, 16 ). In
lioma patients had SMRP values above the same cutpoint.
Table 3. Demographic data for healthy controls and

preoperative mesothelioma patients.
Mesothelioma patients,
Healthy controls preoperative samples
Characteristic (n ⴝ 409) (n ⴝ 88)
Sex
Male, % 246 (60) 70 (80%)
Female, % 163 (40%) 18 (20%)
Age at diagnosis, years
Mean (SD) 46 (12) 64 (10)
Median 45 64
Range 31 to 80 39 to 84
Race
Caucasian
Black
Smoking history, %
Never (no) 251 (61)
Past
Current (yes) 108 (27)
Unknown 50 (12) 88 (100)
Stage, %
I 14 (16)
II 22 (25)
III 49 (56)
IV 3 (3)
Histology, % Fig. 2. ROC curve for healthy vs preoperative (pre-op) mesotheliomas.
Epithelioid 59 (67) ROC curve for the ability of the MESOMARK assay to differentiate between
Sarcomatoid 8 (9) healthy volunteers (n ⫽ 409) and patients with mesothelioma (n ⫽ 88). The table
Biphasic 21 (24) shows the threshold value required for the MESOMARK assay and the sensitivity
(and the 95% CI for the sensitivity) at the 5 different specificity levels.
Table 4. Concentrations and diagnostic accuracy of SMRP in subgroups of patients and controls.
A. The percentage of study participants with MESOMARK values within the selected intervals are indicated. Percentage
sums of the individual groups may not add up to 100% due to rounding influences.
%
Number of study
participants <1.5 nmol/L 1.5–3.0 nmol/L 3.1–10.0 nmol/L >10.0 nmol/L
Apparently healthy
Females 163 99 1 0 0
Males 246 99 1 0 0
Malignant conditions
Mesothelioma 88 48 23 16 14
Ovarian cancer 111 93 5 2 0
Lung cancer 174 83 11 2 4
Colon cancer 50 94 6 0 0
Pancreatic cancer 52 90 8 2 0
Endometrial cancer 25 96 4 0 0
Nonmalignant conditions
Hypertension 100 88 12 0 0
Asbestos-exposed individuals 61 95 5 0 0
Endometriosis 16 100 0 0 0
B. Diagnostic specificity of SMRP at the 99th percentile concentration in diverse patient populations. At this
concentration, the sensitivity for the detection of mesothelioma was 57% (95% CI 46%–67%).
Sample population % Specificity at a 1.5 nmol/L cut-off (95% CI)
Mesothelioma vs normal/healthy 88/409 98.8 (97.2–99.6)
Mesothelioma vs asbestos-exposed 88/61 95.1 (86.3–99.0)
Mesothelioma vs tested cancers 88/412 88.3 (84.9–91.3)
Mesothelioma vs lung cancer 88/174 82.2 (75.7–87.6)
Mesothelioma vs nonmalignant 88/177 90.1 (45.8–67.3)
addition, in longitudinal studies increased SMRP concen- ings, SMRP does have potential as a biomarker for detec-
trations were detected 12– 48 months before detection of tion of mesothelioma in a high-risk, asbestos-exposed
mesothelioma by conventional means (2 of 8 patients; population.
unpublished data). Although these longitudinal studies Because mesothelioma is a rare disease, the use of
included only a limited number of patients, our results SMRP as a screening assay may also require the addition
might indicate the potential of SMRP as a marker for of other markers to increase specificity, for example in an
monitoring response to treatment. Such a biomarker asbestos-exposed population. Candidate markers include
would be beneficial because current imaging methods osteopontin, which was recently reported to be increased
have limited sensitivity (6 ) and are costly. There were no in mesothelioma patients (18 ), the absence of increased
reports of established biomarkers for mesothelioma pub- carcinoembryonic antigen (6 ), or a combination with
lished at the time of the study. Additional studies, how- CYFRA21–1, which could aid in differential diagnosis
ever, are ongoing to compare SMRP with other serologic (19 ). A marker panel requires further studies that are
markers such as osteopontin, CYFRA21-1, or CA125. currently ongoing. Other studies are under way to con-
Preliminary data are presented elsewhere (17 ). firm diagnostic accuracy as established in this preliminary
study.
In summary, serum concentrations of SMRP are higher
in patients with mesothelioma than in healthy persons, References
and can be reliably measured by the MESOMARK assay. 1. Connelly RR, Spirtas R, Myers MH, Percy CL, Fraumeni JF Jr.
Data are remarkably consistent across different patient Demographic patterns for mesothelioma in the United States.
populations and different laboratories and indicate poten- J Natl Cancer Inst 1987;78:1053– 60.
2. Walker AM, Loughlin JE, Friedlander ER, Rothman KJ, Dreyer NA.
tially important clinical utility of the SMRP assay
Projections of asbestos-related disease 1980 –2009. J Occup
in mesothelioma diagnostics. Limitations of SMRP in- Med 1983;25:409 –25.
clude its increase in patients with renal failure, hyper- 3. Carbone M, Kratzke RA, Testa JR. The pathogenesis of mesothe-
tension, and certain other tumors such as ovarian cancer. lioma. Semin Oncol 2002;29:2–17.
Although additional tests and demographic informa- 4. Carbone M, Rdzanek MA. Pathogenesis of malignant mesotheli-
tion may be needed to address some of these shortcom- oma. Clin Lung Cancer 2004;5(Suppl 2):S46 –50.
5. Craighead JE, Mossman BT. The pathogenesis of asbestos- tical approach. NCCLS document EP6-A. Wayne, PA: NCCLS,
associated diseases. N Engl J Med 1982;306:1446 –55. 2003.
6. Sterman DH, Albelda SM. Advances in the diagnosis, evaluation, 14. Creaney J, Robinson BW. Detection of malignant mesothelioma in
and management of malignant pleural mesothelioma. Respirology asbestos-exposed individuals: the potential role of soluble me-
2005;10:266 – 83. sothelin-related protein. Hematol Oncol Clin North Am 2005;19:
7. Robinson BW, Creaney J, Lake R, Nowak A, Musk AW, de Klerk N, 1025– 40.
et al. Mesothelin-family proteins and diagnosis of mesothelioma. 15. Hassan R, Remaley AT, Sampson ML, Zhang J, Cox DD, Pingpank
Lancet 2003;362:1612– 6. J, et al. Detection and quantitation of serum mesothelin, a tumor
8. Ordonez NG. Application of mesothelin immunostaining in tumor marker for patients with mesothelioma and ovarian cancer. Clin
diagnosis. Am J Surg Pathol 2003;27:1418 –28. Cancer Res 2006;12:447–53.
9. Muminova ZE, Strong TV, Shaw DR. Characterization of human 16. Scherpereel A, Grigoriu B, Conti M, Gey T, Gregoire M, Copin MC,
mesothelin transcripts in ovarian and pancreatic cancer. BMC et al. Soluble mesothelin-related peptides in the diagnosis of
Cancer 2004;4:19. malignant pleural mesothelioma. Am J Respir Crit Care Med
10. Scholler N, Fu N, Yang Y, Ye Z, Goodman GE, Hellstrom KE, et al. 2006;173:1155– 60.
Soluble member(s) of the mesothelin/megakaryocyte potentiating 17. Stieber P, Hatz R, Holdenrieder S, Hofmann K, Schalhorn A.
factor family are detectable in sera from patients with ovarian Diagnostische relevanz der SMRP beim malignen mesotheliom.
carcinoma. Proc Natl Acad Sci U S A 1999;96:11531– 6. The XXXIII Meeting of the International Society of Oncodevelop-
11. Chang K, Pastan I. Molecular cloning of mesothelin, a differenti- mental Biology and Medicine. Rhodes, Greece. Tumour Biology
ation antigen present on mesothelium, mesotheliomas, and ovar- 2006; 27(Suppl 1):1–113.
ian cancers. Proc Natl Acad Sci U S A 1996;93:136 – 40. 18. Pass HI, Lott D, Lonardo F, Harbut M, Liu Z, Tang N, et al.
12. Hellstrom I, Raycraft J, Kanan S, Sardesai NY, Verch T, Yang Y, et Asbestos exposure, pleural mesothelioma, and serum osteopon-
al. Mesothelin variant 1 is released from tumor cells as a tin levels. N Engl J Med 2005;353:1564 –73.
diagnostic marker. Cancer Epidemiol Biomarkers Prev 2006;15: 19. Paganuzzi M, Onetto M, Marroni P, Filiberti R, Tassara E, Parodi S,
1014 –20. et al. Diagnostic value of CYFRA 21–1 tumor marker and CEA in
13. National Committee for Clinical Laboratory Standards. Evaluation pleural effusion due to mesothelioma. Chest 2001;119:1138 –
of the linearity of quantitative measurement procedures: a statis- 42.
673– 678 (2007) Cancer Diagnostics
LC-MS/MS Quantification of Zn-␣2 Glycoprotein:

A Potential Serum Biomarker for Prostate Cancer
Olga P. Bondar, David R. Barnidge, Eric W. Klee, Brian J. Davis, and George G. Klee*
Background: Zn-␣2 glycoprotein (ZAG) is a relatively Zn-␣2-glycoprotein (ZAG)1 is a glycoprotein with a mo-
abundant glycoprotein that has potential as a biomarker lecular mass of ⬃41 000 Da and a crystal structure similar
for prostate cancer. We present a high-flow liquid chro- to that of a class I major histocompatibility complex (1, 2 ).
matography–tandem mass spectrometry (LC-MS/MS) Biochemically, ZAG stimulates lipid degeneration in adi-
method for measuring serum ZAG concentrations by pocytes and appears to be involved in cachexia, a wasting
proteolytic cleavage of the protein and quantification of syndrome that can affect people with cancer, AIDS, and
a unique peptide. other terminal illnesses (3, 4 ). ZAG appears naturally in
Methods: We selected the ZAG tryptic peptide 147EI- most body fluids, such as blood (5 ), sweat (6 ), seminal
PAWVPEDPAAQITK162 as the intact protein for quan- fluid (7 ), breast cyst fluid (8 ), cerebrospinal fluid (9 ), and
tification and used a stable isotope-labeled synthetic urine (10 ) and is also found in secretory epithelial cells of
the liver and the gastrointestinal tract (11 ).
peptide with this sequence as an internal standard.
Previous studies employing techniques such as immu-
Standards using recombinant ZAG in bovine serum
nohistology and 2-dimensional electrophoresis have re-
albumin, 50 g/L, and a pilot series of patient sera were
ported that ZAG is overexpressed in certain malignant
denatured, reduced, alkylated, and digested with tryp-
tumors and thus may serve as a potential cancer biomar-
sin. The concentration of ZAG was calculated from a ker (12, 13 ). ZAG quantification in serum by immunoas-
dose–response curve of the ratio of the relative abun- say found circulating concentrations ranging from 40
dance of the ZAG tryptic peptide to internal standard. mg/L in healthy individuals to 120 mg/L in some dis-
Results: The limit of detection for ZAG in serum was eased persons (14 ). In this study we used liquid chroma-
0.08 mg/L, and the limit of quantification was 0.32 mg/L tography–tandem mass spectrometry (LC-MS/MS), at a
with a linear dynamic range of 0.32 to 10.2 mg/L. high LC flow rate commonly used in the clinical labora-
Replicate digests from pooled sera run during a period tory (250 ␮L/min), combined with proteolysis, to quan-
of 3 consecutive days showed intraassay imprecision tify ZAG in serum. We also determined whether ZAG
(CV) of 5.0% to 6.3% and interassay imprecision of 4.4% could be used as a specific biomarker for prostate cancer
to 5.9%. Mean (SD) ZAG was higher in 25 men with (PCa).
prostate cancer [7.59 (2.45) mg/L] than in 20 men with
nonmalignant prostate disease [6.21 (1.65) mg/L, P ⴝ Materials and Methods
0.037] and 6 healthy men [3.65 (0.71) mg/L, P ⴝ 0.0007]. recombinant zag protein standard
Conclusions: This LC-MS/MS assay is reproducible and The recombinant ZAG protein used in this study was a
can be used to evaluate the clinical utility of ZAG as a gift from Dr. Bjorkman (Division of Biology, Howard
cancer biomarker. Hughes Medical Institute, California Institute of Technol-
© 2007 American Association for Clinical Chemistry ogy). The protein was generated in CHO cells by trans-
fection with a ZAG vector using Lipofectamine 2000
(Invitrogen) (2 ). The recombinant ZAG stock concentra-
tion was found to be 1.2 g/L as determined by BCA total
protein analysis (Pierce). The purity of the protein was
Department of Laboratory Medicine and Pathology, Mayo Clinic and
Mayo Foundation, Rochester, MN.
* Address correspondence to this author at: Mayo Clinic and Mayo
Foundation, 200 First St. SW, Rochester, MN 55905. Fax 507-284-4542; e-mail 1
Nonstandard abbreviations: ZAG, zinc-␣2 glycoprotein; LC-MS/MS,
klee.george@mayo.edu. liquid chromatography–tandem mass spectrometry; PCa, prostate cancer;
Received September 7, 2006; accepted January 18, 2007. BSA, bovine serum albumin; LOQ, limit of quantification; LOD, limit of
Previously published online at DOI: 10.1373/clinchem.2006.079681 detection.
673
674 Bondar et al.: LC-MS/MS for ZAG
estimated to be ⬎95% by sodium dodecyl sulfate–polyac- of the LC-MS/MS conditions used can be found else-
rylamide gel electrophoresis (Fig. 1A, inset). Standard where (15 ). The tryptic peptide from ZAG found to have
curves were generated by diluting the recombinant ZAG the greatest response by this method was 147EI-
into 5% bovine serum albumin (BSA) in phosphate- PAWVPFDPAAQITK162 (hereafter referred to as
buffered saline (10 mmol sodium phosphate, 150 mmol tpZAG147–162). The sequence EIPAWVPFDPAAQITK
sodium chloride, pH 7.2). was found to be specific to Chain D, Human Zinc-〈-2-
Glycoprotein (accession no. 1ZAG_D, gi:7246026) via a
serum digestion BLAST search, suggesting that this sequence could be
All digests were performed on 125 ␮L aliquots of serum. used to specifically quantify ZAG. This same tryptic
No prior purification or removal of high-abundance pro- peptide for ZAG was recently described by Anderson and
teins such as serum albumin was performed. Samples Hunter (16 ).
were first denatured with 6 mol/L urea and followed by
reduction in 15 mmol/L dithiothreitol for 60 min at 40 °C. lc-ms/ms conditions
Samples were then alkylated with 50 mmol/L iodoacet- Absolute quantification was performed using a CTC
amide for 60 min in the dark at room temperature. After Analytics HTC PAL autosampler (LEAP Technologies), a
reduction and alkylation we decreased the urea concen- Shimadzu 10-AD binary pumping system, and an API
tration by diluting the samples to 490 ␮L with 50 mmol/L 5000 triple quadrupole mass spectrometer (Applied Bio-
hydroxymethylaminoethane, pH 8, containing 10 systems). For each run we injected a total of 10 ␮L of
mmol/L CaCl2. We then added 1 mg of l-(tosylamido-2- sample onto a 50 ⫻ 2.1 mm TARGA C18 column (Higgins
phenyl) ethyl chloromethyl ketone-treated trypsin (Sig- Analytical) at a flow rate of 250 ␮L/min, with a total run
ma-Aldrich) and digested the sample overnight at 37 °C. time of 30 min. The gradient used consisted of solvent A
(water, 4 mmol/L ammonium acetate, 1 g/L formic acid)
lc-ms/ms and solvent B (methanol, 4 mmol/L ammonium acetate, 1
A Waters Q-TOF Premier quadrupole time-of-flight mass g/L formic acid) starting at 5% B for 2 minutes, ramping
spectrometer was used to identify and acquire the relative to 95% B over 24 min, holding at 95% B for 2 min, back to
abundances of all the tryptic peptides generated from the 5% B in 1 min, and then holding to 30 min. The API 5000
digest of recombinant ZAG. A more thorough explanation instrument source parameters were CAD: 10, CUR: 30,
Fig. 1. (A), Precursor ion spectra of

ZAG trypsin digest.
Inset in (A) shows 1-dimensional so-
dium dodecyl sulfate–polyacrylamide
gel electrophoresis gel of recombinant
ZAG used for optimizing protein cleav-
age and as a standard for the absolute
quantification of ZAG from human se-
rum (lane 1, protein molecular mass
standards; lane 2, 5 ␮g recombinant
ZAG loaded and stained with Coomas-
sie Blue). (B), MS/MS spectrum of
tpZAG147–162. This tryptic relatively
high molecular weight 1088.7 Da. The
primary transition y*10 892.531088.7
and the secondary transition Y7
892.553728.5 have been monitored.
GS1: 40, GS2: 40, IS: 5500, TEM: 600, DP: 225, and EP: 5. specimens from the Mayo Foundation Prostate Special-
Using a 200 ms dwell time the doubly charged precursor ized Program in Research Excellence (SPORE). These
ion for the ZAG tryptic peptide tpZAG147–162 at m/z 892 specimens were collected in 2002–2006, according to a
was selected in Q1 and 3 singly charged transitions were protocol approved by Mayo Clinic Institutional Review
monitored in Q3: the y-ion PFDPAAQITK at m/z 1088 (CE: Board (#1937-00).
40, CXP: 40), the y-ion PAAQITK at m/z 728.4 (CE: 40,
CXP: 40), and the internal fragment ion PAW at m/z 355.2 calculations and statistics
(CE: 50, CXP: 26). We defined the limit of quantification (LOQ) for ZAG as
the response for digested recombinant ZAG added to 50
stable isotope-labeled peptide internal g/L BSA matrix that gave a signal-to-noise value of 10.
standard The limit of detection (LOD) for ZAG was defined as the
The stable isotope peptide 147EIPAWVPEDPAAQITK162 concentration of standard that gave a signal-to-noise
was synthesized in the Mayo Proteomics Research Center value of 3. We used 2-tailed Student t-tests (SAS software)
on an ACT 396 Multiple Peptide Synthesizer (Advanced to compare the ZAG concentrations in the 3 pilot study
ChemTech), using recommended procedures for 1,3-di- groups.
isopropylcarbodiimide activation and coupling. Stable
isotope-labeled proline (5 13C, 1 15N-Fmoc-Proline, Iso- Results
tech) was coupled in the peptide sequence at positions 7 The most critical step in the protein cleavage approach to
and 9 to give a total molecular mass shift of ⫹12 Da from quantification of a protein is the selection of a cleaved
the nonlabeled peptide and a monoisotopic molecular peptide that will provide adequate analytical specificity
mass of 1775.92 Da. We added 2 ␮L of the internal and sensitivity. Selection of the tryptic peptide
standard in concentration of 2 nmol/mL (3.6 ␮g/mL) to tpZAG147–162 for ZAG quantification was determined
10 ␮L of sample before injection into the LC-MS/MS. empirically by performing LC-MS/MS on tryptic digests
of recombinant ZAG. The most abundant tryptic peptide
location of selected peptide in 3-dimensional observed was tpZAG147–162 (Fig. 1A). Fortuitously, this
structure of zag tryptic peptide has 3 proline residues that produce strong
To coordinate the MS/MS measurements with potential fragment ions with m/z values of 355.2, 728.5, and 1088.7
immunoassay measurement systems, we specifically Da (Fig. 1B). All ZAG quantification was performed using
looked for a peptide sequence that was on the exterior of the y10 ⫽1088.7 fragment ion response since this transition
the 3-dimensional structure of ZAG. This exterior location had the best signal-to-noise and LOQ for ZAG from
would more likely be an immunologic binding site for human serum. Also fortuitous was the location of this
appropriately targeted antisera, rather than an internal peptide on the external part of the molecule (see Fig. 1 in
site. The IT7V Zn-␣-2-glycoprotein:baculo-ZAG PEG 200 the Data Supplement that accompanies the online version
structure was downloaded from the RCSB Protein Data of this article at http://www.clinchem.org/content/vol53/
Bank (1 ). The structure was determined using x-ray issue4).
diffraction with a 1.95 Å resolution. The DeepView/ Quantification was performed using a standard curve
Swiss-PdbViewer 3.7 program (2 ) was used to visualize prepared by serial dilutions of recombinant ZAG added
and highlight the epitope site 147EIPAWVPEDPAA- to 50 g/L BSA in phosphate-buffered saline as a surrogate
QITK162 (17, 18 ) serum matrix. Fig. 2 shows 3 ion chromatograms: (a) the
digest of recombinant ZAG added to 50 g/L of BSA at a
study subjects concentration of 1.3 mg/L, (b) ZAG tryptic peptide pro-
We used serum samples from 3 pilot groups of men in this duced after the digest of patient serum, and (c) the stable
study. The 1st group included healthy men (n ⫽ 6, ages 46 isotope-labeled tpZAG147–162 added to the ZAG stan-
to 58 years), the 2nd group, men with nonmalignant dard after digestion. We added 2 ␮L of the internal
prostate biopsy results (n ⫽ 20, ages 59 to 83 years), and standard in concentration of 2 ␮mol/L (3.6 mg/L) to 10
the 3rd group, men with PCa (n ⫽ 26, ages 56 to 84 years). ␮L of sample. A comparison of the response for equimolar
The men with nonmalignant prostate biopsy results had amounts of ZAG vs the internal standard peptide was
pathology reports of normal (n ⫽ 8), inflammation (n ⫽ performed by generating a standard curve, with each
4), prostate intraepithelial neoplasia (n ⫽ 6), and atypical point in the curve having equimolar amounts of ZAG and
acinar proliferation (n ⫽ 2). Median follow-up of the 29 internal standard. Linear regression analysis of this curve
men with PCa was 26.5 months (range 3.8 – 43.6 months). resulted in y ⫽ 1.054x ⫺ 0.283 with an R2 ⫽ 0.9965. These
The prostate-specific antigen values were ⱕ4 ␮g/L (2 results demonstrated that ZAG was completely digested
patients); 4.1–10 ␮g/L (6 patients); 10.1–20 ␮g/L (7 pain the 50 g/L BSA matrix, as the regression line slope was
tients); and ⬎20 ␮g/L (11 patients). Gleason scores were ⬃1.0 and the intercept reflected minimal background
available for 25 patients at the time of initial diagnosis; interference. After standard curves were found to be
36% had a total Gleason score of 6, 32% a score of 7, 16% linear and reproducible, patient and normal donor serum
a score of 8, and 16% a score of 9 or 10. We obtained the samples were digested and analyzed.
Table 1. Assay performance characteristics.

Recovery in normal male pool
Added, ␮g Measured, mg/L % Recovery

0 2.98
3 5.76 93%
6 9.8 114%
Intra- and interassay precision
Intraassay Interassay
Male pool Female pool Male pool Female pool

Replicates 5 5 15 15
Mean, mg/L 3.27 3.42 3.26 3.38
SD, mg/L 0.16 0.22 0.14 0.20
CV, % 4.99 6.30 4.36 5.92
lated, and digested separately and then analyzed using

our LC-MS/MS method over a period of 3 consecutive
Fig. 2. Chromatograms showing the response from the LC-MS/MS
days. The results of this analysis (Table 1) demonstrated
analysis of the ZAG tryptic peptide tpZAG147–162 produced after the
digest of 1.3 mg/L of recombinant ZAG added to a 50 g/L BSA matrix the reproducibility of our method; 10 separate digests
(A), ZAG tryptic peptide tpZAG147–162 produced after the digest of performed and run over 3 consecutive days showed CVs
patient serum (B), along with the stable isotope-labeled internal ⬍7%. The ZAG concentration from normal male sera was
standard (C) 2 ␮L of the internal standard in concentration of 2 between 2.6 and 4.7 mg/L with an average concentration
nmol/mL (3.6 mg/L) was added to 10 ␮L of ZAG tryptic digest before of 3.65 mg/L. The average concentration of ZAG in men
injection to LC-MS/MS (C).
with nonmalignant prostate disease was 6.21 mg/L, and
the average ZAG concentration in men with PCa was 7.59
The trace for the PCa patient (Fig. 2B) shows the mg/L. Student t-tests performed on the 3 sample sets
superior signal-to-noise afforded by the y10 transition for comparing samples of normal, nonmalignant disease, and
the tpZAG147–162 peptide. Controls run without supple- PCa samples showed statistically significant increases in
menting ZAG or stable isotope-labeled tpZAG147–162 the concentration of ZAG across these groups (Table 2)
into 50 g/L BSA matrix showed low background response when analyzed by our LC-MS/MS method.
for the y10 transition, as did controls for the internal
standard transition in human serum digests where no Discussion
internal standard was added. Since ZAG free human The method presented here for the quantification of ZAG
serum was not available, standard addition of recombi- in serum by LC-MS/MS is based on the technique first
nant ZAG into pooled normal male sera was performed to
further confirm the origin of the peak observed in serum
digests. Recoveries were also determined by adding re-
combinant ZAG to pooled normal male sera. Two differ-
ent supplemented concentrations were evaluated, 3 mg/L
(within normal range) and 6 mg/L (mean value for
nonmalignant prostate disease). The results are shown in
Table 1. In addition, a 5-point dilution series was made
from a nonmalignant prostate serum sample by adding 50
g/L BSA matrix. Linear regression analysis of the dilution
series, plotted as the known concentration vs the calcu-
lated concentration, was found to be y ⫽ 0.8112x ⫹ 2.063
with an R2 ⫽ 0.9695. Fig. 3 shows a typical standard curve
used for quantification, demonstrating the linearity asso-
ciated with recombinant ZAG added to the BSA matrix
over a range of 0.33 to 10.4 mg/L. The LOQ was 0.32
mg/L and the LOD was 0.08 mg/L. This LOD translates
into ⬃20 fmol of tpZAG147–162 loaded on column for a
10 ␮L injection. Fig. 3. Linear regression analysis of a standard curve composed of
To evaluate intra- and interassay imprecision, we pre- recombinant ZAG added to a 50 mg/L BSA matrix over a range of 0.33
pared 2 serum pools from normal male and normal to 10.4 mg/L.
female sera. Aliquots of these pools were reduced, alky- (Figure represents 1 typical run.)
ing or prognostic value of ZAG, and the LC-MS/MS assay

Table 2. Comparison of serum ZAG values in normal
described in this paper should be a valuable tool for
individuals, patients with nonmalignant prostate disease,
conducting these studies.
and patients with PCa.
Serum ZAG concentration,
mg/L
This study was partially supported by SPORE in Prostate
Group Number Average SD P values
Cancer Grant P50CA 91956 from the National Cancer
Normal 6 3.65 0.71 } 0.0013
Nonmalignant prostate
disease
PCa
20
25
6.21
7.59
1.65
2.45 } 0.0007 } 0.037

Institute, NIH.
References
1. Delker SL, West AP Jr, McDermott L, Kennedy MW, Bjorkman PJ.
Crystallographic studies of ligand binding by Zn-alpha2-glycopro-
described by Barr et al. (19 ) and modified by our group tein. J Struct Biol 2004;148:205–13.
and others for use in quantifying proteins from serum 2. Sanchez LM, Chirino AJ, Bjorkman P. Crystal structure of human
ZAG, a fat-depleting factor related to MHC molecules. Science
(16, 20 ). Recently, Anderson and Hunter described the
1999;283:1914 –9.
same tryptic fragment of ZAG protein (tpZAG147–162)
3. Bing C, Bao Y, Jenkins J, Sanders P, Manieri M, Cinti S, et al.
that can be used for identification of ZAG among 53 Zinc-alpha2-glycoprotein, a lipid mobilizing factor, is expressed in
plasma proteins ranging in abundance from albumin adipocytes and is up-regulated in mice with cancer cachexia. Proc
(⬃70 g/L) down to fibronectin (⬃1 mg/L) (16 ). We Natl Acad Sci U S A 2004;101:2500 –5.
performed absolute quantification of ZAG by using pure 4. Russell ST, Tisdale MJ. The role of glucocorticoids in the induction
recombinant protein for our standards, and a stable of zinc-alpha2-glycoprotein expression in adipose tissue in cancer
isotope-labeled synthetic peptide with the same sequence cachexia. Br J Cancer 2005;92:876 – 81.
as tryptic fragment tpZAG147–162 as an internal stan- 5. Burgi W, Schmid K. Preparation and properties of Zn-alpha 2-gly-
coprotein of normal human plasma. J Biol Chem 1961;236:
dard. The concentration of ZAG in serum is relatively
1066 –74.
high (our findings showed ⬃3– 8 mg/L), making it pos-
6. Nakayashiki N. Sweat protein components tested by SDS-poly-
sible to quantify the protein by LC-MS/MS directly from acrylamide gel electrophoresis followed by immunoblotting. To-
digests of serum, without purification or depletion of hoku J Exp Med 1990;161:25–31.
high-abundance proteins. We were also able to use a high 7. Burgi W, Simonen S, Baudner S, Schmid K. Unusually high
LC flow rate of 250 ␮L/min and an electrospray source, concentrations of Zn alpha 2-glycoprotein and the lack of alpha
both of which are routinely used in clinical laboratories. 2HS-glycoprotein in human ejaculates. Clin Chem 1989;35:
We anticipate that the method will be a valuable resource 1649 –50.
for determining the concentration of ZAG in larger co- 8. Sanchez LM, Vizoso F, Diez-Itza I, Lopez-Otin C. Identification of
horts of patients to determine the utility of ZAG as a the major protein components in breast secretions from women
with benign and malignant breast diseases. Cancer Res 1992;52:
cancer biomarker. 95–100.
If follow-up LC-MS/MS cohort studies demonstrate 9. Davidsson P, Nilsson CL. Peptide mapping of proteins in cerebrospi-
the ZAG protein has clinical utility as a cancer biomarker, nal fluid utilizing a rapid preparative two-dimensional electrophoretic
development of an immunoassay may be desirable. The procedure and matrix-assisted laser desorption/ionization mass
position of the tpZAG147–162 fragment on the periphery spectrometry. Biochim Biophys Acta 1999;1473:391–9.
of the protein’s tertiary structure suggests that this pep- 10. Hirai K, Hussey HJ, Barber MD, Price SA, Tisdale MJ. Biological
tide may also provide an accessible epitope for antibody evaluation of a lipid-mobilizing factor isolated from the urine of
binding and facilitate future assay development. This cancer patients. Cancer Res 1998;58:2359 – 65.
LC-MS/MS assay, based on an external peptide, could 11. Tada T, Ohkubo I, Niwa M, Sasaki M, Tateyama H, Eimoto T.
Immunohistochemical localization of Zn-alpha 2-glycoprotein in
potentially be used as a reference method for immunoas-
normal human tissues. J Histochem Cytochem 1991;39:1221– 6.
says, especially if the immunoassays also target the same
12. Hale LP, Price DT, Sanchez LM, Demark-Wahnefried W, Madden
external region of the ZAG molecule illustrated in Fig. 1 in JF. Zinc alpha-2-glycoprotein is expressed by malignant prostatic
the online Data Supplement. epithelium and may serve as a potential serum marker for
ZAG has been identified as a biomarker associated prostate cancer. Clin Cancer Res 2001;7:846 –53.
with multiple diseases, including prostate, bladder, and 13. Irmak S, Tilki D, Heukeshoven J, Oliveira-Ferrer L, Friedrich M,
breast cancer (3, 6, 7 ). Therefore, ZAG concentrations by Huland H, et al. Stage-dependent increase of orosomucoid and
themselves are unlikely to be a specific screening test for zinc-alpha2-glycoprotein in urinary bladder cancer. Proteomics
any targeted disease, such as PCa. Nonetheless, ZAG may 2005;5:4296 –304.
be a valuable biomarker to increase sensitivity for early 14. Ekman R, Johansson BG, Ravnskov U. Renal handling of Zn-
alpha2-glycoprotein as compared with that of albumin and the
disease detection if it is combined with other more tissue retinol-binding protein. J Clin Invest 1976;57:945–54.
specific biomarker markers, such as prostate-specific an- 15. Barnidge DR, Tschumper RC, Jelinek DF, Muddiman DC, Kay NE.
tigen. ZAG also may have utility for differentiating more Protein expression profiling of CLL B cells using replicate off-line
aggressive forms of cancer. However, a large number of strong cation exchange chromatography and LC-MS/MS. J Chro-
measurements will be needed to further define the screen- matogr B Analyt Technol Biomed Life Sci 2005;819:33–9.
16. Anderson L, Hunter CL. Quantitative mass spectrometric multiple 19. Barr JR, Maggio VL, Patterson DG Jr, Cooper GR, Henderson LO,
reaction monitoring assays for major plasma proteins. Mol Cell Turner WE, et al. Isotope dilution–mass spectrometric quantifica-
Proteomics 2006;5:573– 88. tion of specific proteins: model application with apolipoprotein A-I.
17. Guex N, Peitsch MC. SWISS-MODEL and the Swiss-PdbViewer: an Clin Chem 1996;42:1676 – 82.
environment for comparative protein modeling. Electrophoresis 20. Barnidge DR, Goodmanson MK, Klee GG, Muddiman DC. Absolute
1997;18:2714 –23. quantification of the model biomarker prostate-specific antigen in
18. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H, et serum by LC-Ms/MS using protein cleavage and isotope dilution
al. The Protein Data Bank. Nucleic Acids Res 2000;28:235– 42. mass spectrometry. J Proteome Res 2004;3:644 –52.
Clinical Chemistry 53:4 Lipids, Lipoproteins,
679 – 685 (2007) and Cardiovascular
Risk Factors
Plasma Lipoprotein(a) Indicates Risk for 4 Distinct

Forms of Vascular Disease
Gregory T. Jones,1* Andre M. van Rij,1 Jennifer Cole,2 Michael J.A. Williams,1
Emma H. Bateman,1 Santica M. Marcovina,3 Meiying Deng,1 and Sally P.A. McCormick2
Background: Increased lipoprotein(a) [Lp(a)] concentra- This association was not altered by exposure to standard
tions are predictive for coronary artery disease (CAD). lipid-lowering therapy.
The risk conferred by Lp(a) for other types of vascular © 2007 American Association for Clinical Chemistry
disease compared with CAD has not been investigated
within a single population. This study aimed to inves- Atherosclerosis, a chronic condition characterized by the
tigate Lp(a) risk association for 4 different types of formation of lipid-rich plaques within the walls of me-
vascular disease (including CAD) within a predomi- dium and large arteries (1, 2 ), underlies many forms of
nantly white population. vascular disease. The development of vascular disease is
Methods: We used an Lp(a) ELISA that measures Lp(a) dependent on multiple genetic and environmental deter-
independently of apolipoprotein(a) size to measure minants (2 ), and there is heavy reliance on large multi-
plasma Lp(a) in patients [384 CAD, 262 peripheral vas- factorial association studies to identify individual risk
cular disease, 184 ischemic stroke (stroke), 425 abdomi- factors (3 ). Among the most prominent group of risk
nal aortic aneurysm] and 230 disease-free controls. We factors identified by this approach have been the plasma
then conducted association studies with logistic regres- lipoproteins as measured by total serum cholesterol (TC),4
sion, integrating the potential confounding effects of triglyceride (TG), LDL cholesterol (LDL-C) and HDL
age, sex, diabetes, plasma lipids, and a history of previ- cholesterol (HDL-C). These have proven to be reliable
ous hypertension, hypercholesterolemia, and smoking. predictors of vascular disease risk in many large clinical
Results: Multivariate analyses with Lp(a) concentra- cohorts (4, 5 ). Despite the advent of effective dietary (6 )
tions of >45 nmol/L (the 75th percentile value for and drug treatments (7 ) to modify lipoprotein concentra-
controls) as the clinical cutoff showed increased Lp(a) tions, however, studies show that vascular disease re-
concentrations to be a risk factor for all disease groups, mains prevalent among individuals whose plasma li-
with adjusted odds ratios ranging from 1.96 [95% confi- poproteins have reached target concentrations (8 ).
dence interval (CI) 1.24 –3.08] for CAD to 2.33 (95% CI Lipoprotein(a) [Lp(a)] is a plasma lipoprotein that
1.39 –3.89) for PVD. The risk conferred by Lp(a) ap- consists of an LDL molecule covalently linked to the
peared to be independent of other confounders, includ- plasminogen “look-alike”, apolipoprotein(a) [apo(a)] (9 ).
ing exposure to statin/fibrate therapies. Similar odds Unlike other plasma lipoproteins, Lp(a) is very resistant
ratios and CIs between disease groups indicated that to change by diet or lipid-lowering drugs (10 ). The
increased Lp(a) conferred a similar risk for all groups concentrations of Lp(a) in human plasma vary from 0 to
studied. ⬎300 nmol/L, show ethnic differences (11 ), and appear to
Conclusions: Lp(a) constitutes a stable risk factor of be strictly genetically controlled, largely by a well-char-
similar magnitude for 4 major forms of vascular disease. acterized size polymorphism in the apo(a) gene (12 ).
Although the exact physiological role of Lp(a) has not
been definitively established, substantial evidence indi-
1
Departments of Medical and Surgical Sciences, and 2Biochemistry, Uni- cates that Lp(a) is both atherogenic and thrombogenic in
versity of Otago, Dunedin, New Zealand.
3
Department of Medicine (SMM), Northwest Lipid Research Laboratories,
University of Washington, Seattle, WA.
* Address correspondence to this author at: Vascular Research Group,
4
Department of Medical and Surgical Sciences, Dunedin School of Medicine, Nonstandard abbreviations: TC, total serum cholesterol; TG, triglyceride;
University of Otago, PO Box 913, Dunedin 9054, New Zealand. Fax 64-3-474- LDL-C, LDL cholesterol; HDL-C, HDL cholesterol; Lp(a), lipoprotein(a);
7622; e-mail greg.jones@otago.ac.nz. apo(a), apolipoprotein(a); Mab, monoclonal antibody; CAD, coronary artery
Received September 11, 2007; accepted January 15, 2007. disease; CI, confidence interval; PVD, peripheral vascular disease; AAA,
Previously published online at DOI: 10.1373/clinchem.2006.079947 abdominal aortic aneurysm.
679
680 Jones et al.: Plasma Lp(a) Concentrations and Vascular Disease
nature (13 ). The high heritability of plasma Lp(a) concen- tomography or magnetic resonance imaging of cerebral
trations and their lack of responsiveness to environmental infarction. AAA patients examined in this study had
influences suggest a potential value of Lp(a) as a stable infrarenal aneurysms ⬎5 cm in maximum anteroposterior
risk factor for vascular disease. diameter, as determined by ultrasound scan. All patients
Increased plasma Lp(a) has been shown to be an with PVD, CAD, or stroke underwent abdominal ultra-
independent risk factor for many forms of vascular dis- sound examination to identify concurrent AAA and were
ease, including coronary artery disease (CAD), (14 –16 ) excluded if the maximum anteroposterior aortic diameter
peripheral vascular disease (PVD) (17–19 ), ischemic was ⬎2.5 cm. All patients with disease of undetermined
stroke (20 –22 ), and abdominal aortic aneurysm (AAA) etiology were excluded. We recruited controls from local
(23, 24 ). The status of Lp(a) as a risk factor has been Otago community groups, with inclusion criteria of age
controversial, however, with several large studies show- ⬎55 years, no history of ischemic heart disease, including
ing no apparent association (25–28 ). The general consen- angina pectoris, PVD, stroke (including transient ischemic
sus, at least for CAD, is that Lp(a) is an independent risk attack), or AAA and being currently in good general
factor as established by a recent metaanalysis of 27 health. All study participants completed a questionnaire
prospective CAD cohorts involving 5436 cases (29 ). to ascertain demographic risk factors, including age, sex,
Most studies evaluating Lp(a) as a risk factor have history of hypertension and hyperlipidemia, current med-
included just 1 form of vascular disease. Differences in the ication, concurrent disease such as ischemic heart disease,
methods used to measure Lp(a) (10 ), the type of data PVD, stroke, and diabetes. An experienced clinical re-
analysis, and the population size and ethnicity preclude search technician assisted participants with questionnaire
evaluation of the relative importance of Lp(a) as a risk completion. Smoking habits (current and past) were as-
factor for different forms of vascular disease across these sessed and the number of pack years (20 cigarettes per
studies. This study aimed to investigate the risk conferred day for 1 year) was calculated.
by Lp(a) for 4 distinct forms of vascular disease (CAD,
PVD, stroke, and AAA) within a predominantly white blood collection and lipid measurements
population by using the consensus reference method for We obtained venous blood samples from all participants
Lp(a) (30 ) that is unaffected by bias from apo(a) isoform with vacuum tubes containing EDTA. Plasma was sepa-
size. rated from cells and aliquots stored at ⫺80 °C until Lp(a)
and lipid concentrations were measured. Storage times
Materials and Methods before measurement ranged from 3 months for stroke
study participants samples to a maximum of 5 years for the AAA samples
Vascular disease patients, mainly of white origin (⬎97%), (mean 2.6 years). A subset of plasma samples from each
were recruited from the Otago region of New Zealand as group was subjected to apo(a) Western blot analysis (32 )
previously described (31 ) and compared with a healthy to check for sample degradation. We determined Lp(a)
elderly control group from the same geographical region. concentrations by a double-sandwich ELISA as previ-
All participants provided written informed consent, and ously described (30 ), using monoclonal antibody (MAb)
the study was undertaken with the approval of the a-6 as the capture antibody and MAb a-1 as the detection
Regional Ethics Committee. A total of 1255 patients were antibody. This assay detects all apo(a) isoforms on an
examined, in whom vascular disease consisted of coro- equivalent molar basis, and is considered to be an accu-
nary artery disease (CAD, n ⫽ 384), ischemic peripheral rate method for Lp(a) measurement (33 ). The limit of
vascular disease (PVD, n ⫽ 262), ischemic stroke (stroke, quantification of this assay was 2 nmol/L at a 1:200
n ⫽ 184), and abdominal aortic aneurysm (AAA, n ⫽ 425). sample dilution. We measured plasma total cholesterol
Patients classified as having CAD all had angiographi- and triglycerides using Roche Diagnostics enzymatic
cally proven coronary artery stenosis of ⱖ50% of the methods. HDL-C was measured with Roche HDL-C Plus
vessel internal diameter in at least 1 vessel. The percent- reagents. LDL-Cholesterol was calculated using the Fried-
age diameter stenosis was visually estimated as the max- wald formula (34 ).
imum percentage reduction in the vessel diameter ex-
pressed as a percentage of the angiographically normal statistical analyses
adjacent vessel. Peripheral vascular disease was defined We performed statistical analysis with StatView version
as significant stenosis in multiple segments, including 5.01 (SAS Institute). The ␹2 test was used to compare
clinical symptoms such as claudication, pain during rest, observed with expected frequencies. ANOVA tests were
or tissue loss. The diagnosis of PVD was further con- used to compare lipid variables of vascular disease pa-
firmed with a resting ankle-brachial index ⬍0.7, pulse tients with those of controls. Because the Lp(a) distribu-
volume recordings, arteriography, and/or duplex arterial tion was nongaussian, we used the Mann–Whitney U-test
scan. Stroke patients had clinical symptoms consisting of to compare median Lp(a) concentrations for each patient
rapidly developing signs of a focal or global disturbance group with those of controls. We compared the distribu-
of cerebral function, with symptoms lasting more than tion of plasma Lp(a) by plotting the concentrations
24 h. The diagnosis was confirmed by computerized against percentiles for each group. Relative risk was
estimated in terms of odds ratio and corresponding 95% hypercholesterolemic compared with the controls, al-
confidence interval (CI). Risk assessment models used though the AAA group did have a mean total cholesterol
either Lp(a) concentrations ⬍45 nmol/L (the 75th percen- of 5.6 mmol/L, which could be considered hypercholes-
tile value for controls) or ⬍4.5 nmol/L (the 25th percentile terolemic. This result was most likely due to the lower
value for controls) as cutoff values for the reference rate of lipid-lowering treatments in the AAA group.
group. We conducted a secondary assessment using 75 Indeed, total and LDL-C concentrations were significantly
nmol/L as the high cutoff so that comparison could be lower in the CAD, PVD, and stroke groups, likely a direct
made with previously published results from the Fra- reflection of the higher incidence of lipid-lowering treat-
mingham study control population (13 ). An assessment of ment in these groups compared with the AAA and control
the risk conferred by having an Lp(a) concentration groups. HDL-C concentrations were significantly de-
⬍2 nmol/L was also performed using controls with Lp(a) creased in CAD and AAA patients (Table 1).
concentrations ⬎2 nmol/L as the reference group. We The integrity of samples for Lp(a) measurement, as
used multiple logistic regression to test interactive effects checked by apo(a) Western blotting, showed little evi-
of variables, producing adjusted analyses assessing Lp(a) dence of apo(a) degradation among subsets of samples
concentration as a risk factor for CAD, PVD, stroke, and from each disease group (see Fig. 1 in the Data Supple-
AAA. Significant or suggestive (P ⬍ 0.15) univariate ment in the Data Supplement that accompanies the online
confounders of either patient group or Lp(a) concentra- version of this article at http://www.clinchem.org/
tion were identified and applied to a multivariate regres- content/vol53/issue4). Within the AAA patient popula-
sion model. The resulting winnowed model included age, tion (which had the largest range of sample storage
sex, history of hypertension, hypercholesterolemia, diabe- times), there was no significant difference in Lp(a) con-
tes, total and HDL cholesterol, and smoking history. centrations between short- (⬍6 months) and long-term
Because of interacting effects with other modeled vari- (⬎4 years) stored samples.
ables, a separate analysis was conducted to determine the Plasma Lp(a) concentrations were significantly higher
confounding influence of exposure to lipid-lowering in patients within all 4 vascular disease groups compared
therapies. with controls (Table 1). A plot of Lp(a) concentrations
across percentiles demonstrated a nongaussian distribu-
Results tion, which was highly skewed toward lower concentra-
Analyses of demographic profiles highlighted signifi- tions. A divergence between the Lp(a) concentrations for
cantly higher frequencies of history of hypertension, hy- the control group and that for all vascular disease groups
percholesterolemia, and smoking in all vascular disease was apparent from the 50th percentile (17 nmol/L in
patients compared with controls without vascular disease controls, Fig. 1). The 75th percentile of the control group
(Table 1). The frequency of diabetes was significantly corresponded to an Lp(a) concentration of 45 nmol/L. The
increased in patients with CAD, PVD, and stroke, but not commonly applied risk-associated concentration of 75
patients with AAA, compared with controls (Table 1). The nmol/L (10, 33 ) corresponded to approximately the 80th
lipid profiles among the vascular disease groups were not percentile of controls in the current study. The proportion
Table 1. Demographic profiles of the vascular disease populations.

Controls CAD PVD Stroke AAA
Variable n ⴝ 230 n ⴝ 384 n ⴝ 262 n ⴝ 184 n ⴝ 425
Sex, male, % 33.8 72.7a 59.2a 63.7a 77.7a
Age, years 70.3 (6.9) 64.0 (9.7) 71.6 (9.1) 71.9 (10.0) 71.7 (7.6)
Hypertension, % 31.0 51.3a 64.9a 68.5a 55.3a
Hypercholesterolemia, % 18.6 52.2a 43.2a 55.5a 30.5c
Lipid-lowering treatment 22.4 94.6a 65.7a 68.1a 37.0a
(statin/fibrate), %
Diabetes, % 8.7 20.3a 22.9a 23.9a 5.9
Smoking, pack years 9.8 (18.3) 19.6 (25.8)a 30.6 (33.8)a 21.2 (26.8)a 29.2 (27.9)a
Ischaemic heart disease 0 100 26.7 34.7 42.8
Total-C, mmol/L 5.5 (1.4) 4.3 (0.9)a 5.0 (1.1)a 4.9 (1.2)a 5.6 (1.3)
LDL-C, mmol/L 3.3 (1.2) 2.2 (0.8)a 2.6 (1.0)a 2.8 (1.1)a 3.4 (1.2)
HDL-C, mmol/L 1.3 (0.4) 1.1 (0.3)a 1.3 (0.4) 1.2 (0.4) 1.1 (0.4)a
Triglycerides, mmol/L 2.0 (0.7–⫺9.0) 1.8 (0.5–7.8) 2.0 (0.7–8.0) 1.7 (0.5–5.3) 2.0 (0.5–11.2)
Lp(a), nmol/L median, range 17.0 (⬍2–285) 25.5 (⬍2–471)a 28.8 (⬍2–539)a 30.6 (⬍2–386)a 29.9 (⬍2–583)a
Lp(a), above 45 nmol/L, % 24.7 40.6b 41.5b 39.1b 39.8b
Lp(a), above 75 nmol/L, % 19.5 34.6b 31.5b 30.4c 29.3c
Lp(a), below 2 nmol/L, % 17.3 17.5 12.7 10.9 11.7c
Results are expressed as means (1SD), except for Lp(a) concentrations, which are expressed as medians and range. a, P ⬍0.002; b, P ⬍0.005; c, P ⬍0.05.
risk were determined using multivariate logistic regres-

sion. Adjusted odds ratios of 1.96, 2.33, 2.00, and 2.12 were
obtained for the CAD, PVD, stroke, and AAA groups
respectively, using the ⬎45 nmol/L cutoff (Table 2).
Similarly, adjusted odds ratios of 2.06, 2.86, 2.50, and 2.54
were obtained for the CAD, PVD, stroke, and AAA
groups, respectively, using the 4th (⬎45 nmol/L) vs 1st
(⬍4.5 nmol/L) control quartiles (Table 3). All analyses
showed considerable overlap in odds ratios and confi-
dence intervals between disease groups, and there was no
significant difference in odds ratio between the 4 vascular
disease groups. Adjustment for known confounders had
very little effect on the odds ratio. The independence of
the Lp(a) and vascular disease association, from exposure
to lipid-lowering therapy (statin or fibrate medications)
was specifically examined. Although Lp(a) odds ratios
(⬎45nmol/L) were slightly reduced when modeled with
lipid-lowering therapy, the association of Lp(a), and vas-
cular disease remained significant, both in the combined
vascular disease group (odds ratio 1.76, 95% CI 1.3–2.4, P
⬍0.0005) and the specific disease subgroups. Within sub-
groups, patients on lipid-lowering therapy did not have
significantly different Lp(a) concentrations compared
with untreated individuals.
A significant number study participants within the
control (17.3%) and vascular disease groups (10.9%–
17.5%) had Lp(a) concentrations below the sensitivity (⬍2
nmol/L) of the assay (Table 1). Only the AAA group had
a significantly smaller proportion of individuals with
Fig. 1. Distribution of Lp(a) concentrations in the control and vascular Lp(a) ⬍2 nmol/L than the control group. According to
disease groups.
analysis of the adjusted odd ratios (Table 4), having an
Individual Lp(a) concentrations are plotted against percentiles of Lp(a) for the
control and 4 vascular disease groups. The control quartile values are indicated, Lp(a) ⬍2 nmol/L had no significant effect on the risk of
with the control 75th percentile corresponding to an Lp(a) of 45 nmol/L. Note vascular disease for any of the disease groups.
that the commonly used 75 nmol/L cutoff corresponds to approximately the 80th
percentile.
of patients with Lp(a) ⬎45 nmol/L (or 75 nmol/L) was Table 2. Logistic regression for Lp(a) >45 nmol/L and risk
significantly greater in all vascular disease groups com- of various forms of vascular disease.
pared with controls (Table 1). Because ⬃5% more vascular Odds ratio for Lp(a)
disease patients were included in the “at risk” group with >45 nmol/L 95% CI P value
the 45 nmol/L cutoff, this cutoff was used for subsequent Unadjusted
risk analyses. CAD 2.89 1.29–3.22 ⬍0.0001
The confounding effects of demographic variables on PVD 2.17 1.47–3.20 ⬍0.0001
Lp(a) concentrations were assessed. Sex was not signifi- Stroke 1.96 1.29–2.99 ⬍0.002
cantly associated with Lp(a) concentrations in controls or AAA 2.02 1.41–2.88 0.0001
in PVD, stroke, or AAA patients. Female CAD patients, All vascular disease 2.06 1.50–2.84 ⬍0.0001
patients
however, had significantly higher Lp(a) concentrations
Adjusted
than their male counterparts (medians 28.8 vs 23.9
CAD 1.96 1.24–3.08 ⬍0.004
nmol/L, Mann–Whitney U-test P ⬍ 0.04). A previous
PVD 2.33 1.39–3.89 ⬍0.002
history of hypercholesterolemia was significantly associ-
Stroke 2.00 1.22–3.26 ⬍0.01
ated with increased Lp(a) concentrations within the CAD
AAA 2.12 1.37–3.29 ⬍0.001
patient group (medians 18.4 vs 45.4 nmol/L, Mann– All vascular disease 2.03 1.37–2.98 ⬍0.0005
Whitney U-test P ⫽ 0.0002). Across all disease groups and patients
controls, diabetic individuals had lower Lp(a) concentra- The reference population were control individuals free of vascular disease,
tions than persons without diabetes (medians 19.9 vs 27.6 with Lp(a) ⬍45 nmol/L. Odds ratios were expressed with 95% CI. A P value
nmol/L, Mann–Whitney U-test P ⬍0.01). ⬍0.05 was considered significant. The adjusted model accounted for age, sex,
The adjusted odds ratios for increased Lp(a) concen- history of hypertension, hypercholesterolemia, diabetes, total and HDL choles-
terol, and smoking history.
trations (⬎45 nmol/L) as an indicator of vascular disease
to evaluate Lp(a) as a risk factor for the development of

Table 3. Logistic regression for Lp(a) 4th quartile and risk
multiple forms of vascular disease within the same pop-
of various forms of vascular disease.
ulation. The population studied was recruited from the
Odds ratio for Lp(a)
in 4th Quartile 95% CI P value same geographical area and was of uniform ethnicity.
Unadjusted Lp(a) measurements were performed with an assay that
CAD 1.98 1.26–3.13 ⬍0.004 recognizes apo(a) size isoforms on a equal basis and is
PVD 2.29 1.39–3.77 ⬍0.002 considered to be the consensus reference method for Lp(a)
Stroke 2.36 1.35–4.13 ⬍0.003 measurement (33 ).
AAA 2.25 1.43–3.54 0.0005 The demographics of Lp(a) within the population
All vascular disease 2.18 1.47–3.25 0.0001 studied here were similar to those reported for other
patients white populations, i.e., a nongaussian distribution that
Adjusted was highly skewed toward low concentrations. The me-
CAD 2.06 1.14–3.71 ⬍0.02 dian Lp(a) concentration of 17 nmol/L in our control
PVD 2.86 1.43–5.72 ⬍0.004 group was similar to the 20 nmol/L value reported in a
Stroke 2.50 1.28–4.86 ⬍0.01
large study of American whites (35 ). The 75th percentile
AAA 2.54 1.41–4.56 0.002
value for controls was 45 nmol/L (Fig. 1), which is lower
All vascular disease 2.37 1.43–3.92 ⬍0.001
patients than the commonly used cutoff concentration of 75
nmol/L derived as the 75th percentile value from the
The reference population was controls with Lp(a) in the first quartile (⬍4.5
nmol/L). Odds ratios were expressed with 95% CI. A P value ⬍0.05 was Framingham study control population (13 ). We used
considered significant. The adjusted model accounted for age, sex, history of ⬎45 nmol/L as the clinical cutoff for increased Lp(a) in
hypertension, hypercholesterolemia, diabetes, total and HDL cholesterol, triglyc- this study, because the greatest divergence between the
erides, and smoking history. control and disease groups occurred at this concentration.
Furthermore, at this concentration rather than 75 nmol/L,
Discussion 5% more of the vascular disease patients were placed
The majority of studies establishing Lp(a) as a vascular in the at risk group, a feature that may be of clinical
disease risk factor have done so in populations with CAD. benefit with respect to more aggressively treating patients
The association with other forms of vascular disease is for other risk factors. After adjustment for known risk
less clear, although some large cohorts have established factors, the risk conferred by Lp(a) ⬎45 nmol/L ranged
Lp(a) as a risk factor for stroke (21, 22 ). There are fewer from 1.96 to 2.33 across all forms of vascular disease
studies evaluating Lp(a) as a risk factor for PVD and studied. Interestingly, similar adjusted odds ratios (all
AAA, and many are limited by small study populations vascular disease patients, odds ratio 2.0, 95% CI 1.31–3.03,
(17, 19, 23 ). The many differences in population size, P ⬍ 0.002) were obtained if the commonly applied
ethnicity, and method of Lp(a) measurement make it ⬎75 nmol/L (33 ) cutoff was used. Slightly higher odds
difficult to assess the value of Lp(a) as a risk factor across ratios, ranging from 2.06 to 2.86, were derived from
different vascular diseases. Indeed, many studies have comparison of the upper vs lower quartiles. Notably,
used assays that are sensitive to apo(a) isoform size, there was no significant difference in risk ratios between
therefore introducing some bias in the Lp(a) measure- disease groups for all 3 analyses, indicating a similar risk
ment. The present study is the first cross-sectional study conferred by increased Lp(a) across all vascular disease
groups.
A large proportion of individuals with Lp(a)
Table 4. Logistic regression for Lp(a) <2 nmol/L and risk ⬍45 nmol/L had concentrations ⬍2 nmol/L. We were
of various forms of vascular disease. originally concerned that a high number ⬍2 nmol/L
Odds ratio for
Lp(a) Null 95% CI P value
individuals in the ⬍45 nmol/L reference group may have
Unadjusted
accentuated the risk ratio analysis. However, an analysis
CAD 1.01 0.66–1.55 0.97
to establish the effect of having an Lp(a) ⬍2 nmol/L on
PVD 0.69 0.42–1.14 0.15 vascular disease risk showed no significant effect across
Stroke 0.58 0.33–1.04 0.07 all groups.
AAA 0.63 0.40–0.99 ⬍0.05 Lp(a) shows both atherogenic and thrombogenic prop-
Adjusted erties. Atherogenic properties include a high affinity for
CAD 1.03 0.56–1.83 0.91 extracellular matrix proteins (36 ) and an ability to accu-
PVD 0.65 0.32–1.34 0.25 mulate oxidized phospholipids (37 ), which promote in-
Stroke 0.57 0.29–1.15 0.12 flammation, and thrombogenic properties center around
AAA 0.68 0.37–1.24 0.21 the ability of the apo(a) protein to inhibit plasminogen
The reference populations was controls with Lp(a) ⬎2 nmol/L. Odds ratios are activation (38 ). The etiologies of the various forms of
expressed with 95% CI. A P value ⬍0.05 was considered significant. The vascular disease studied here involve atherosclerotic and
adjusted model accounted for age, sex, history of hypertension, hypercholester- thrombotic processes, although the relative importance of
olemia, diabetes, total and HDL cholesterol, and smoking history.
these processes might be expected to vary between the
different vascular diseases. For example, in the case of forms of vascular disease in some patients. The main
AAA, there are pronounced proteolytic and inflammatory finding of the study is that increased Lp(a) is a risk factor,
components associated with the breakdown and remod- with similar magnitudes of effect, for CAD, PVD, stroke,
eling of the aortic wall compared with aortic atheroscle- and AAA. The risk is not attenuated by other known risk
rotic occlusive disease (39, 40 ). Results from our study, factors. We conclude that Lp(a) is a stable marker for
however, indicate that any possible difference in the assessing the risk of all major forms of vascular disease.
pathogenic role of Lp(a) is not associated with any signif-
icant changes in relative risk ratios between vascular
disease groups. This finding does not preclude possibility This work was supported by grants from Lottery Health
that Lp(a) and other atherogenic or thrombogenic risk and an Otago Research Grant. We would like to thank
factors may interact in some populations. For example, Jean-Ha Beck for technical assistance and Dr Graeme
some studies have documented large increases in risk Hammond-Tooke for assistance with patient recruitment
ratios for stroke when Lp(a) concentrations are combined of the stroke patients.
with certain thrombogenic risk factors such as factor V
Leiden and antithrombin III deficiency (41 ). References
Although adjusting for known risk factors did not 1. Ross R. The pathogenesis of atherosclerosis: a perspective for
significantly alter the risk relationship for Lp(a), there the 1990s. Nature 1993;362:801–9.
were some interactions between Lp(a) and other vascular 2. Lusis AJ. Atherosclerosis. Nature 2000;407:233– 41.
3. Kannel WB, D’Agostino RB, Sullivan L, Wilson PW. Concept and
disease risk factors worthy of mention. Of note, the
usefulness of cardiovascular risk profiles. Am Heart J 2004;148:
presence of diabetes was associated with a lower concen- 16 –26.
tration of Lp(a) in the controls and all vascular disease 4. Castelli WP, Anderson K, Wilson PW, Levy D. Lipids and risk of
groups, with the exception of stroke. Our study did not coronary heart disease. The Framingham Study. Ann Epidemiol
have the power to dissect this relationship fully, because 1992;2:23– 8.
some groups (AAA patients and controls) had low num- 5. Assmann G, Cullen P, Schulte H. The Munster Heart Study
bers of diabetics. This relationship, however, is consistent (PROCAM). Results of follow-up at 8 years. Eur Heart J 1998;19
Suppl A:A2–11.
with a recent study of 587 CAD patients, which showed
6. Jenkins DJ, Kendall CW, Marchie A, Faulkner DA, Wong JM, de
that type 2 diabetes patients within the cohort had signif- Souza R, et al. Effects of a dietary portfolio of cholesterol-lowering
icantly lower concentrations of Lp(a) than those without foods vs lovastatin on serum lipids and C-reactive protein. JAMA
diabetes (42 ). The mechanism by which diabetes de- 2003;290:502–10.
creases Lp(a) is unknown, but merits further investiga- 7. Randomised trial of cholesterol lowering in 4444 patients with
tion. A highly significant relationship was also observed coronary heart disease: the Scandinavian Simvastatin Survival
in CAD patients between Lp(a) and a history of hyper- Study (4S). Lancet 1994;344:1383–9.
8. Linsel-Nitschke P, Tall AR. HDL as a target in the treatment of
cholesterolemia (typically increased LDL). This probably
atherosclerotic cardiovascular disease. Nat Rev Drug Discov
reflects the higher risk for CAD when both Lp(a) and LDL 2005;4:193–205.
are increased, as has been reported in other studies 9. Utermann G. The mysteries of lipoprotein (a). Science 1989;246:
(14, 43 ), hence there is a selection bias for these individ- 904 –10.
uals. The relationship between Lp(a) and LDL cholesterol 10. Marcovina SM, Koschinsky ML, Albers JJ, Skarlatos S. Report of
was not significant in the current analysis because many the National Heart, Lung, and Blood Institute Workshop on
patients were already on lipid-lowering (predominantly Lipoprotein(a) and Cardiovascular Disease: recent advances and
future directions. Clin Chem 2003;49:1785–96.
statin) treatment before recruitment. The rates of lipid-
11. Sandholzer C, Hallman DM, Saha N, Sigurdsson G, Lackner C,
lowering therapies varied considerably across the vascu- Csaszar A, et al. Effects of the apolipoprotein (a) size polymor-
lar disease groups examined in this study. This facilitated phism on the lipoprotein (a) concentration in 7 ethnic groups. Hum
analysis of a possible association between such treatment Genet 1991;86:607–14.
and Lp(a) concentrations. Significantly, given the now 12. Boerwinkle E, Leffert CC, Lin J, Lackner C, Chiesa G, Hobbs HH.
widespread use of such agents, this study confirms Lp(a) Apolipoprotein(a) gene accounts for greater than 90% of the
variation in plasma lipoprotein(a) concentrations. J Clin Invest
as a predictive lipid marker of vascular disease risk,
1992;90:52– 60.
regardless of exposure to standard lipid-lowering 13. Boffa MB, Marcovina SM, Koschinsky ML. Lipoprotein (a) as a risk
therapy. factor for atherosclerosis and thrombosis: mechanistic insights
The strengths of this study include a relatively large from animal models. Clin Biochem 2004;37:333– 43.
study population of uniform ethnicity and the use of an 14. Luc G, Bard JM, Arveiler D, Ferrieres J, Evans A, Amouyel P, et al.
apo(a) isoform-insensitive method for measuring Lp(a). A Lipoprotein (a) as a predictor of coronary heart disease: the
weakness of this study was the likely presence of overlap PRIME Study. Atherosclerosis 2002;163:377– 84.
15. Seman LJ, DeLuca C, Jenner JL, Cupples LA, McNamara JR,
between disease groups. Although the presence of AAA
Wilson PW, et al. Lipoprotein (a)-cholesterol and coronary heart
was excluded from other vascular disease groups, and disease in the Framingham Heart Study. Clin Chem 1999;45:
patients were recruited into disease groups based on the 1039 – 46.
clinical symptoms and diagnosis associated with their 16. Berg K, Dahlen G, Christophersen B, Cook T, Kjekshus J, Ped-
initial event, we cannot discount the coexistence of other ersen T. Lp(a) lipoprotein level predicts survival and major coro-
nary events in the Scandinavian Simvastatin Survival Study. Clin chemical measurements of lipoprotein(a). Clin Chem 1995;41:
Genet 1997;52:254 – 61. 246 –55.
17. Valentine RJ, Kaplan HS, Green R, Jacobsen DW, Myers SI, 31. Jones GT, Harris EL, Phillips LV, van Rij AM. The Methylenetetra-
Clagett GP. Lipoprotein (a), homocysteine, and hypercoagulable hydrofolate Reductase C677T Polymorphism Does Not Associate
states in young men with premature peripheral atherosclerosis: a with Susceptibility to Abdominal Aortic Aneurysm. Eur J Vasc
prospective, controlled analysis. J Vasc Surg 1996;23:53– 61, Endovasc Surg 2005;30:137– 42.
discussion-3. 32. Marcovina SM, Hobbs HH, Albers JJ. Relation between number of
18. Cheng SW, Ting AC, Wong J. Lipoprotein (a) and its relationship to apolipoprotein(a) kringle 4 repeats and mobility of isoforms in
risk factors and severity of atherosclerotic peripheral vascular agarose gel: basis for a standardized isoform nomenclature. Clin
disease. Eur J Vasc Endovasc Surg 1997;14:17–23. Chem 1996;42:436 –9.
19. Widmann MD, Sumpio BE. Lipoprotein (a): a risk factor for 33. Marcovina SM, Albers JJ, Scanu AM, Kennedy H, Giaculli F, Berg
peripheral vascular disease. Ann Vasc Surg 1993;7:446 –51. K, et al. Use of a reference material proposed by the International
20. Peng DQ, Zhao SP, Wang JL. Lipoprotein (a) and apolipoprotein E Federation of Clinical Chemistry and Laboratory Medicine to
epsilon 4 as independent risk factors for ischemic stroke. J Car- evaluate analytical methods for the determination of plasma
diovasc Risk 1999;6:1– 6. lipoprotein (a). Clin Chem 2000;46:1956 – 67.
21. Milionis HJ, Filippatos TD, Loukas T, Bairaktari ET, Tselepis AD, 34. Friedewald WT, Levy RI, Fredrickson DS. Estimation of the concen-
Elisaf MS. Serum lipoprotein (a) levels and apolipoprotein (a) tration of low-density lipoprotein cholesterol in plasma, without
isoform size and risk for first-ever acute ischaemic nonembolic use of the preparative ultracentrifuge. Clin Chem 1972;18:499 –
stroke in elderly individuals. Atherosclerosis 2006;187:170 – 6. 502.
35. Marcovina SM, Albers JJ, Wijsman E, Zhang ZH, Chapman NH,
22. Jurgens G, Taddei-Peters WC, Koltringer P, Petek W, Chen Q,
Kennedy H. Differences in Lp (a) concentrations and apo (a)
Greilberger J, et al. Lipoprotein(a) serum concentration and apo-
polymorphs between Black and White Americans. J Lipid Res
lipoprotein(a) phenotype correlate with severity and presence of
1996;37:2569 – 85.
ischemic cerebrovascular disease. Stroke 1995;26:1841– 8.
36. Marcovina SM, Koschinsky ML. Lipoprotein(a) as a risk factor for
23. Norrgard O, Angquist KA, Dahlen G. High concentrations of Lp (a)
coronary artery disease. Am J Cardiol 1998;82:57U– 66U; discus-
lipoprotein in serum are common among patients with abdominal
sion 86U.
aortic aneurysms. Int Angiol 1988;7:46 –9.
37. Tsimikas S, Brilakis ES, Miller ER, McConnell JP, Lennon RJ,
24. Sofi F, Marcucci R, Giusti B, Pratesi G, Lari B, Sestini I, et al. High
Kornman KS, et al. Oxidized phospholipids, Lp (a) lipoprotein, and
levels of homocysteine, lipoprotein (a) and plasminogen activator
coronary artery disease. N Engl J Med 2005;353:46 –57.
inhibitor-1 are present in patients with abdominal aortic aneu-
38. Marcovina SM, Koschinsky ML. Evaluation of lipoprotein(a) as a
rysm. Thromb Haemost 2005;94:1094 – 8.
prothrombotic factor: progress from bench to bedside. Curr Opin
25. Ridker PM, Hennekens CH, Stampfer MJ. A prospective study of Lipidol 2003;14:361– 6.
lipoprotein (a) and the risk of myocardial infarction. Jama 1993; 39. Thompson RW, Holmes DR, Mertens RA, Liao S, Botney MD,
270:2195–9. Mecham RP, et al. Production and localization of 92-kilodalton
26. Wang W, Hu D, Lee ET, Fabsitz RR, Welty TK, Robbins DC, et al. gelatinase in abdominal aortic aneurysms. An elastolytic metallo-
Lipoprotein(a) in American Indians is low and not independently proteinase expressed by aneurysm-infiltrating macrophages.
associated with cardiovascular disease. The Strong Heart Study. J Clin Invest 1995;96:318 –26.
Ann Epidemiol 2002;12:107–14. 40. Saito S, Zempo N, Yamashita A, Takenaka H, Fujioka K, Esato K.
27. Lindholt JS, Heegaard NH, Vammen S, Fasting H, Henneberg EW, Matrix metalloproteinase expressions in arteriosclerotic aneurys-
Heickendorff L. Smoking, but not lipids, lipoprotein (a) and mal disease. Vasc Endovascular Surg 2002;36:1–7.
antibodies against oxidised LDL, is correlated to the expansion of 41. Nowak-Gottl U, Junker R, Hartmeier M, Koch HG, Munchow N,
abdominal aortic aneurysms. Eur J Vasc Endovasc Surg 2001;21: Assmann G, et al. Increased lipoprotein (a) is an important risk
51– 6. factor for venous thromboembolism in childhood. Circulation
28. Ridker PM, Stampfer MJ, Hennekens CH. Plasma concentration of 1999;100:743– 8.
lipoprotein(a) and the risk of future stroke. JAMA 1995;273: 42. Saely CH, Koch L, Schmid F, Marte T, Aczel S, Langer P, et al.
1269 –73. Lipoprotein(a), type 2 diabetes and vascular risk in coronary
29. Danesh J, Collins R, Peto R. Lipoprotein (a) and coronary heart patients. Eur J Clin Invest 2006;36:91–7.
disease. Meta-analysis of prospective studies. Circulation 2000; 43. Suk Danik J, Rifai N, Buring JE, Ridker PM. Lipoprotein (a),
102:1082–5. measured with an assay independent of apolipoprotein (a) iso-
30. Marcovina SM, Albers JJ, Gabel B, Koschinsky ML, Gaur VP. Effect form size, and risk of future cardiovascular events among initially
of the number of apolipoprotein(a) kringle 4 domains on immuno- healthy women. JAMA 2006;296:1363–70.
686 – 692 (2007) and Cardiovascular
Risk Factors
Higher Concentrations of Alanine

Aminotransferase within the Reference Interval
Predict Nonalcoholic Fatty Liver Disease
Yoosoo Chang,1 Seungho Ryu,2* Eunju Sung,3 and Yumi Jang4
Background: In nonalcoholic fatty liver disease weight participants who were still within the reference
(NAFLD), increased alanine aminotransferase (ALT) interval of ALT and GGT at all follow-up examinations.
concentrations are considered to be a consequence of Conclusions: In apparently healthy, nondiabetic Ko-
hepatocyte damage. We performed a prospective study rean men, increased ALT concentration, even within the
to examine the association between ALT within its reference interval, was an independent predictor of
reference interval and risk for subsequent development incident NAFLD.
of NAFLD. © 2007 American Association for Clinical Chemistry
Methods: The study cohort comprised 5237 healthy men
without diagnosed NAFLD and without increases of Hepatic steatosis unrelated to excessive alcohol consump-
either ALT (>35 U/L) or ␥-glutamyltransferase (GGT; tion is termed nonalcoholic fatty liver disease (NAFLD).5
>40 U/L) above the reference intervals. We assessed NAFLD encompasses the entire spectrum of liver condi-
alcohol intake via self-reporting (questionnaire) and tions, ranging from simple steatosis through steatohepa-
performed biochemical tests for liver and metabolic titis to advanced fibrosis and cirrhosis (1 ). Concurrent
function and abdominal ultrasonography. We used the with the worldwide epidemic of obesity, NAFLD is con-
Cox proportional hazards model to calculate the ad- sidered to be the most common cause of unexplained
justed hazard ratios (aHRs) in the model for NAFLD. abnormal results of liver function tests (2 ). Although
Results: During 13 276.6 person-years of follow-up over hepatic steatosis was long believed to be a benign disease,
a 4-year period, 984 new incident cases of NAFLD NAFLD has recently gained much interest (3 ). Percentage
developed. We adjusted for age, weight change, body hepatic fat has been reported to be a feature highly
mass index, glucose, blood pressure, triglycerides, HDL associated with insulin resistance (4, 5 ). Furthermore,
cholesterol, smoking, alcohol consumption, regular ex- several studies have suggested an association between
ercise, homeostasis model assessment of insulin resis- NAFLD and features of the metabolic syndrome, includ-
tance, C-reactive protein, and incident diabetes. Com- ing dyslipidemia and obesity, thereby stressing the asso-
pared with an ALT concentration of <16 U/L, aHR ciation with insulin resistance as an important feature of
values (95% confidence intervals) for ALT concentra- NAFLD (6, 7 ). Currently, NAFLD is recognized as a
tions were 1.53 (1.18 –1.98), 1.66 (1.29 –2.13), 1.62 (1.26 – pathogenic factor of insulin resistance and type 2 diabetes
2.08), and 2.21 (1.73–2.81) for ALT concentrations of (3 ).
16 –18, 19 –21, 22–25, and 26 –34 U/L, respectively. This Several prospective epidemiological studies have dem-
relationship remained significant even among normal- onstrated that increased concentrations of hepatic enzyme
in serum, even within the reference interval, may be
related to increased risk of type 2 diabetes and the
metabolic syndrome, as well as death (8 –10 ). Among the
1
Health Screening Center, 2 Department of Occupational Medicine, and hepatic enzymes aspartate aminotransferase (AST), ala-
3
Department of Family Medicine, Kangbuk Samsung Hospital, Sungkyunk-
wan University School of Medicine, Seoul, Korea.
4
Department of Radiology, Asan Medical Center, University of Ulsan
College of Medicine, Ulsan, Korea.
5
* Address correspondence to this author at: Kangbuk Samsung Hospital, Nonstandard abbreviations: NAFLD, nonalcoholic fatty liver disease;
108 Pyung dong, Jongro-Gu, Seoul, Korea 110-746. Fax 82-2-2001-2626; e-mail AST, aspartate aminotransferase; ALT, alanine aminotransferase; GGT, ␥-glu-
sh703.yoo@samsung.com. tamyltransferase; ALP, alkaline phosphatase; HOMA-IR, homeostasis model
Received October 3, 2006; accepted January 8, 2007. assessment of insulin resistance; CRP, C-reactive protein; BMI, body mass
Previously published online at DOI: 10.1373/clinchem.2006.081257 index; CI, confidence interval; aHR, adjusted hazard ratio.
686
nine aminotransferase (ALT), and ␥-glutamyltransferase The NAFLD-free cohort thus comprised 5888 men.
(GGT), ALT is most closely related to liver fat accumula- They were reexamined at the same hospital annually or
tion (11 ). In cross-sectional studies (12 ), participants with biennially until July 2006. After excluding 651 men who
NAFLD often have increased circulating concentrations of did not complete their follow-up examinations, 1222
ALT. Paradoxically, the complete spectrum of NAFLD participants with annual health examinations (mean
was reported in patients with normal ALT activity, even follow-up 3.02 years, SD 1.14 years) and 4015 with bien-
after the cutoff value was decreased to ⱕ19 U/L (13 ). nial health examinations (mean follow-up 2.39 years, SD
Moreover, increased ALT not associated with fatty liver 1.01 years) were successfully followed and observed for
was observed frequently in obese participants (14 ). ALT the development of NAFLD. Altogether, these 5237 men
actually is a glucogenic enzyme, and increased ALT has were included in the final analysis, and their mean (SD)
been demonstrated to be an indicator of impaired insulin follow-up period was 2.54 (1.08) years. This study was
signaling, which might not necessarily be associated with approved by the Institutional Review Board at Kangbuk
liver injury due to hepatic steatosis (3, 15 ). To date, Samsung Hospital and was in accordance with the prin-
although an increased ALT concentration is considered a ciples of the Helsinki II Declaration.
consequence of hepatocyte damage in NAFLD (3 ), it is
unclear what underlies the relationship between ALT and measurements
NAFLD. Of more recent interest, an inverse correlation Initial health examinations performed in 2002 included a
between ALT concentrations and adiponectin concentra- medical history, physical examination, questionnaire
tions has been demonstrated (16 –18 ). As these previous about health-related behavior, anthropometric measure-
observational findings are consistent with a biological link ments, and biochemical measurements. Medical history
between ALT and development of NAFLD, we performed and history of prescription drug use were assessed by the
a prospective study to test whether higher concentrations examining physicians. All participants were asked to
respond to a questionnaire on health-related behavior.
of ALT within its reference interval predict future NAFLD
Questions about alcohol intake included the frequency of
in apparently healthy men in Korea.
alcohol consumption on a weekly basis and the usual
amount consumed on a daily basis. We considered per-
sons reporting that they smoked to be current smokers. In
study population
addition, participants were asked about their weekly
We conducted a prospective cohort study of nondiabetic
frequency of physical activity such as jogging, bicycling,
Korean men at a large semiconductor manufacturing
and swimming that lasted long enough to produce
company and its 13 affiliates. All employees participate in
perspiration.
either annual or biennial health examinations, as required
Fasting blood samples were drawn from an antecubital
by Korea’s Industrial Safety and Health Law. The study
vein after participants had fasted for ⱖ12 h. We measured
population included men ⱖ40 years old who underwent
fasting serum glucose, total cholesterol, triglycerides,
an annual comprehensive health examination and men 30 LDL cholesterol, HDL cholesterol, GGT, ALT, AST, and
to 39 years old who underwent a biennial comprehensive alkaline phosphatase (ALP) concentrations enzymatically
health examination. In 2002, 15 347 workers, ages 30 to 59 with an automatic analyzer (Advia 1650, Bayer Diagnos-
years, participated in the comprehensive health examina- tics). We measured fasting serum glucose with the hex-
tions at a university hospital in Seoul, Korea. We excluded okinase method; total cholesterol and serum triglycerides
9462 men based on the following exclusion criteria that with enzymatic colorimetric tests; LDL cholesterol with
might influence insulin resistance or ultrasonography the homogeneous enzymatic colorimetric test; HDL cho-
findings of the liver as a result of other liver disease lesterol with the selective inhibition method (Bayer Diag-
(because some individuals met more than 1 exclusion nostics); and insulin concentrations with immunoradio-
criterion, the following data total more than 9462): (a) 27 metric assays (Biosource), with intra- and interassay CVs
had a history of a malignancy; (b) 16 had a history of of 4.7% to 12.2%. We estimated insulin resistance by use of
cardiovascular disease; (c) 125 reported current use of the homeostasis model assessment (HOMA-IR), as de-
antihyperlipidemics; (d) 279 had fasting glucose concen- scribed by Matthews et al. (19 ). We analyzed high-
trations ⱖ7.0 mmol/L or current use of blood glucose– sensitivity C-reactive protein (CRP) by use of particle-
lowering agents; (e) 2498 reported an alcohol intake ⱖ20 enhanced immunonephelometry with the BN System
g/day; (f) 5053 had fatty liver based on ultrasonography; (Dade Behring). The minimum detectable CRP concentra-
(g) 6928 had (i) a positive serologic finding for hepatitis B tion was 0.175 mg/L after 1:20 sample dilution. The
or C virus, (ii) chronic liver disease or liver cirrhosis based Korean Association of Quality Assurance for Clinical
on ultrasonography, (iii) increased ALT (ⱖ35 U/L) or Laboratories assessed the quality control of the labora-
GGT (ⱖ40 U/L), or (iv) a reported history of known liver tory, both internally and externally, on a regular basis.
disease, including viral, genetic, autoimmune, and drug- Because the reference interval and cutoff value of ALT
induced liver disease; and (h) 337 had missing data in are controversial (20 ), increased serum ALT and in-
medical histories or serum aminotransferase concentrations. creased serum GGT were defined as values in their
688 Chang et al.: ALT and Nonalcoholic Fatty Liver Disease
respective highest quartiles of our study population (ⱖ35 tion. Incidence densities were compared by calculating
U/L for serum ALT and ⱖ40 IU/L for serum GGT) (21 ). hazard ratios with the 95% confidence interval (CI). We
The reference interval used at the Kangbuk Samsung used the Cox proportional hazards model to calculate
Hospital for serum concentrations of ALT in men was 0 to each adjusted hazard ratio (aHR) in the model for
40 U/L. NAFLD. The data were first adjusted for age alone and
Trained nurses obtained sitting blood pressure read- then for the multiple covariates. In the multivariate mod-
ings with a standard mercury sphygmomanometer. The els, we included the variables (age, BMI, weight change,
1st and 5th Korotkoff sounds were used to estimate fasting serum glucose, systolic blood pressure, triglycer-
systolic and diastolic blood pressure. Height and weight ides, HDL cholesterol, HOMA-IR, CRP, smoking, alcohol
were measured with the participants wearing a light- consumption, and regular exercise) that might confound
weight hospital gown and no shoes. Body mass index the relationship between the serum ALT and NAFLD. For
(BMI) was calculated as the patient’s weight (in kilo- linear trends of risk, the number of quintiles was used as
grams) divided by the square of the patient’s height (in a continuous variable and tested on each model. Analyses
meters). The rate of weight change was calculated as were done with SPSS version 13.0 (SPSS Inc.). All the
follows: weight change ⫽ (last weight ⫺ baseline reported P values were 2-tailed, and those ⬍0.05 were
weight)/follow-up period (years). considered to be statistically significant.
The diagnosis of fatty liver was based on the results of
abdominal ultrasonography with a 3.5-MHz transducer Results
(Logic Q700 MR, GE). Ultrasound studies were carried At baseline, the mean (SD, range) age of the 5237 partic-
out by 3 experienced radiologists who were unaware of ipants was 36.6 years (4.8, 30 –59); the BMI was 22.7 kg/m2
the aims of the study and blinded to laboratory values. (2.4, 15.6 –33.3). Of the 5888 eligible participants at base-
Images were captured in a standard fashion with the line, 651 who did not have a follow-up examination by the
patient in the supine position, right arm raised above the end of July 2006 were more likely to be current smokers
head. All ultrasonographic images were stored in the than the remaining participants; all other variables were
image server and also taken with instant film for later not different between the participants lost to follow-up
inspection by the radiologists and physicians. Of 4 known and those with successful follow-up.
criteria [hepatorenal echo contrast, liver brightness, deep During 13 276.6 person-years of follow-up, 984 new
attenuation, and vascular blurring (22 )], a diagnosis of incident cases of NAFLD developed. In contrast to partic-
fatty liver required the participant to have hepatorenal ipants without incident NAFLD, those with incident
contrast and liver brightness. Based on computer-gener- NAFLD were slightly older and more likely to have the
ated random samples among the stored images, there was metabolic syndrome and incident diabetes. As expected,
excellent agreement on fatty liver diagnosis between the 3 BMI, glucose, blood pressure, lipid profiles, and hepatic
radiologists (agreement 99%, ␬ 0.98). enzymes, except ALP, were significantly different be-
The ATP III proposed the following 5 abnormalities to tween the 2 groups. In addition, HOMA-IR and CRP were
define the metabolic syndrome (23 ): (a) abdominal obe- higher for those with incident NAFLD (data not shown).
sity; (b) high fasting glucose: ⱖ6.1 mmol/L; (c) hypertri- The baseline characteristics of the study sample accord-
glyceridemia: triglycerides ⱖ1.69 mmol/L; (d) low HDL ing to the quintiles of serum ALT concentrations are
cholesterol: ⬍1.04 mmol/L; and (e) high blood pressure: shown in Table 1. Tests for differences of variables across
ⱖ130/85 mmHg. Because waist measurements were not the quintiles of serum ALT found that BMI, glucose, blood
available for the entire study sample, we substituted BMI pressure, lipid profiles, CRP, and HOMA-IR showed a
ⱖ25 kg/m2 for all participants as an index of obesity, linear trend in relation to serum ALT concentration, even
because this cutoff has been proposed for the diagnosis of within its reference interval, whereas age, current smok-
obesity in Asian people (24 ). Individuals with at least 3 of ing, alcohol consumption, regular exercise, and incident
the 5 abnormalities were considered to have the metabolic diabetes did not.
syndrome. An increase across the serum ALT quintiles predicted
the incidence of NAFLD in a graded and dose-responsive
statistical analysis manner (Table 2). After adjustment for age, weight
The ␹2-test and 1-way ANOVA were used to analyze change, BMI, glucose, systolic blood pressure, triglycer-
statistical differences among the characteristics of the ides, HDL, smoking, alcohol consumption, regular exer-
study participants at the time of enrollment in relation to cise, HOMA-IR, CRP, and incident diabetes, and in com-
serum ALT concentrations within the reference interval. parison with concentrations ⬍16 U/L, aHR values (95%
Serum ALT was categorized into the following quintiles: CI) for ALT concentrations of 16 –18, 19 –21, 22–25, and
⬍16, 16 –18, 19 –21, 22–25, and 26 –34 U/L. Incidence 26 –34 U/L were 1.53 (1.18 –1.98), 1.66 (1.29 –2.13), 1.62
density was expressed as the number of cases divided by (1.26 –2.08), and 2.21 (1.73–2.81), respectively (P for trend
the person-years from baseline until development of ⬍0.001), in overall participants. The relationship of ALT
NAFLD, by assuming a date of diagnosis in the middle of with incident NAFLD remained significant even after
the follow-up period or until the final physical examina- further adjustment for serum GGT or AST. Similar asso-
Table 1. Baseline characteristics of the study participants by concentration of ALT within the reference interval.a
ALT, U/L
<16 16–18 19–21 22–25 26–34 P for trend

n 1189 1014 1018 978 1038
Age, years 36.2 (4.7) 36.8 (4.9) 36.6 (4.8) 36.9 (5.0) 36.5 (4.8) 0.111
BMI, kg/m2 21.7 (2.2) 22.4 (2.2) 22.7 (2.3) 23.3 (2.2) 23.7 (2.3) ⬍0.001
Fasting serum glucose, mmol/L 4.91 (0.47) 4.94 (0.46) 4.95 (0.48) 5.02 (0.49) 5.00 (0.48) ⬍0.001
Systolic blood pressure, mmHg 111.8 (11.4) 113.4 (12.2) 113.7 (11.8) 114.9 (12.1) 114.7 (11.9) ⬍0.001
Diastolic blood pressure, mmHg 72.3 (9.4) 73.1 (9.3) 73.4 (9.5) 74.3 (9.6) 74.3 (9.4) ⬍0.001
Total cholesterol, mmol/L 4.81 (0.76) 4.91 (0.83) 5.04 (0.81) 5.10 (0.81) 5.18 (0.84) ⬍0.001
HDL cholesterol, mmol/L 1.45 (0.31) 1.39 (0.29) 1.38 (0.28) 1.36 (0.30) 1.33 (0.28) ⬍0.001
LDL cholesterol, mmol/L 2.85 (0.68) 2.93 (0.73) 3.01 (0.70) 3.06 (0.68) 3.12 (0.72) ⬍0.001
Triglyceride, mmol/L 1.04 (0.81–1.38) 1.13 (0.87–1.54) 1.23 (0.94–1.65) 1.28 (0.99–1.72) 1.36 (1.01–1.85) ⬍0.001
CRP, mg/L 0.30 (0.20–0.70) 0.40 (0.20–0.70) 0.30 (0.20–0.70) 0.40 (0.20–0.80) 0.45 (0.30–0.90) ⬍0.001
Insulin, pmol/L 39.4 (32.5–49.2) 42.2 (34.0–53.1) 43.5 (34.9–55.3) 46.2 (37.1–60.0) 47.6 (37.3–62.7) ⬍0.001
HOMA-IR 1.24 (1.02–1.59) 1.32 (1.04–1.73) 1.36 (1.09–1.77) 1.50 (1.16–1.93) 1.52 (1.17–2.01) ⬍0.001
Current smoker, % 44.7 41.3 41.6 39.7 44.6 0.658
Light drinker, %b 24.1 25.0 26.5 25.4 24.6 0.696
Regular exerciser, %c 49.4 50.7 51.2 54.0 50.3 0.298
Metabolic syndrome, % 1.3 2.9 3.4 4.6 7.2 ⬍0.001
Incident diabetes, % 0.8 0.5 1.4 0.5 0.5 0.890
a
Data are mean (SD) or median (interquartile range) unless otherwise noted.
b
Ethanol, 10 –20 g per day.
c
One time or more per week.
ciations were also observed in the stratified subgroup pants (BMI ⬍23 kg/m2). The ratio of baseline ALT to AST
analyses according to drinking habit (alcohol drinking of also predicted the incidence of NAFLD, although again
⬍10 g per day or nondrinker), nonobese (BMI ⬍25 this association was weaker than the gradient for ALT.
kg/m2), and even normal weight (BMI ⬍23 kg/m2). Neither AST nor ALP was significantly related to the
Moreover, even in participants without any features of the incidence of NAFLD.
metabolic syndrome, any increase of serum ALT, despite
being within the reference interval, continued to predict Discussion
the incidence of NAFLD. In this prospective study of apparently healthy, nondia-
To explore whether the risk for NAFLD in relation to betic Korean men, increasing ALT concentration, even
serum ALT within its reference interval was mediated by within its reference interval, was an independent predic-
the subsequent increase of serum ALT and serum GGT, tor of incident NAFLD, irrespective of various potential
we fit an additional model, excluding participants who confounders, including BMI, alcohol consumption, CRP,
showed an increase of serum ALT ⱖ35 U/L and serum HOMA-IR, and the metabolic syndrome components of
GGT ⱖ40 U/L at follow-up. Even after these exclusions, fasting glucose, lipids, and blood pressure. Moreover, this
the linear association between NAFLD and the serum relationship of ALT on incident NAFLD remained even
ALT quintiles within the reference interval remained after adjustment for incident diabetes and weight gain.
statistically significant (P for trend ⬍0.001). In addition, The strength of this study was the large sample size,
even after exclusion of participants who reported ethanol which allowed us to identify the effect among stratified
intake of ⱖ20 g/day only at follow-up, the relationship of subgroup analyses. Even in normal-weight participants,
ALT on incident NAFLD remained statistically significant modestly increased ALT continued to predict incident
(P for trend ⬍0.001). During follow-up, 4 new incident NAFLD, as was apparent among our participants whose
cases of hepatitis B virus (serologically positive) were ALT remained within the reference interval at follow-up.
found, and no cases of hepatitis C virus were found. To our knowledge, this is the first study to demonstrate
Furthermore, these 4 new incident cases of hepatitis B a relationship between ALT and incident NAFLD based
virus did not have NAFLD at follow-up. on ultrasonography, although several cross-sectional
Baseline GGT also predicted the incidence of NAFLD, studies have already demonstrated that higher ALT, even
but this association was weaker than the gradient for within the currently accepted normal reference interval, is
ALT, and the associations across GGT quintiles seemed to associated with NAFLD (25 ). A previous study also
be nonlinear (models 4 and 5 in Table 3). Furthermore, showed that ALT appears to have associations with both
serum GGT was not significantly related to the incidence hepatic insulin resistance and later decline in hepatic
of NAFLD in nondrinkers and normal-weight partici- insulin sensitivity (26 ). Moreover, a recent study has
Table 2. Adjusted hazard ratios of incidence of nonalcoholic fatty liver disease in relation to ALT concentrations within the
reference interval
ALT, U/L
<16 16–18 19–21 2–25 26–34 P for trend

All participants (n ⫽ 5237)
Cases 102 156 188 237 301
Person-years 3226.9 2639.1 2605.7 2400.3 2404.6
ID (per 1000 person-years)a 31.6 59.1 72.2 98.7 125.2
aHR, 95% CI
Model 1b 1.00 1.82 (1.42–2.33) 2.23 (1.76–2.84) 3.11 (2.39–3.80) 3.89 (3.11–4.87) ⬍0.001
Model 2c 1.00 1.56 (1.21–2.01) 1.70 (1.33–2.18) 1.70 (1.33–2.18) 2.29 (1.81–2.89) ⬍0.001
Model 3d 1.00 1.56 (1.21–2.00) 1.68 (1.32–2.15) 1.68 (1.32–2.15) 2.26 (1.79–2.86) ⬍0.001
Model 4e 1.00 1.53 (1.18–1.98) 1.67 (1.30–2.15) 1.62 (1.26–2.08) 2.20 (1.73–2.80) ⬍0.001
Model 5f 1.00 1.53 (1.18–1.98) 1.66 (1.29–2.13) 1.62 (1.26–2.08) 2.21 (1.73–2.81) ⬍0.001
Ethanol ⬍10 g/day (n ⫽ 3925)
aHR (95% CI), Model 5 1.00 1.60 (1.18–2.17) 1.74 (1.30–2.33) 1.96 (1.47–2.60) 2.36 (1.79–3.11) ⬍0.001
Nondrinkers (n ⫽ 1119)
aHR (95% CI), Model 5 1.00 1.60 (0.92–2.81) 1.54 (0.88–2.67) 2.70 (1.59–4.60) 2.53 (1.51–4.22) ⬍0.001
BMI ⬍25 kg/m2 (n ⫽ 4333)
aHR (95% CI), Model 5 1.00 1.51 (1.12–2.04) 1.60 (1.20–2.13) 1.89 (1.42–2.52) 2.13 (1.61–2.80) ⬍0.001
BMI ⬍23 kg/m2 (n ⫽ 3226)
aHR (95% CI), Model 5 1.00 1.53 (1.05–2.24) 1.61 (1.12–2.29) 1.64 (1.13–2.39) 1.92 (1.34–2.76) 0.001
Without metabolic syndrome (n ⫽ 5037)
aHR (95% CI), Model 5 1.00 1.47 (1.13–1.92) 1.63 (1.26–2.10) 1.54 (1.19–2.00) 2.29 (1.79–2.92) ⬍0.001
No metabolic syndrome traits (n ⫽ 2795)
aHR (95% CI), Model 5 1.00 2.05 (1.37–3.06) 1.80 (1.21–2.69) 2.32 (1.54–3.48) 2.07 (1.39–3.07) 0.001
ALT ⬍35 U/L and GGT ⬍40 U/L during the
study period (n ⫽ 2959)
aHR (95% CI), Model 5 1.00 1.69 (1.10–2.59) 2.27 (1.50–3.45) 2.55 (1.68–3.87) 3.07 (2.04–4.63) ⬍0.001
Ethanol ⬍20 g /day during the study period
(n ⫽ 4535)
aHR (95% CI), Model 5 1.00 1.67 (1.25–2.22) 1.79 (1.35–2.36) 1.85 (1.40–2.45) 2.42 (1.85–3.16) ⬍0.001
a
ID, incidence density.
b
Model 1: adjustment for age.
c
Model 2: model 1 plus adjustment for weight change, fasting serum glucose, loge triglyceride, HDL cholesterol, BMI, systolic blood pressure, smoking, exercise,
and alcohol intake.
d
Model 3: model 2 plus adjustment for HOMA-IR.
e
Model 4: model 3 plus adjustment for CRP.
f
Model 5: model 4 plus adjustment for incident diabetes.
demonstrated that subtle alterations in glucose tolerance NAFLD was independent of the degree of adiposity and
and lipid metabolism exist in those patients with higher that ALT, but not GGT, predicted the development of
ALT, and that this is not necessarily accompanied by NAFLD, even in normal-weight participants (BMI ⬍23
hepatic steatosis (27 ). In our prospective study, even ALT kg/m2). Our results appear to agree with the findings of
concentrations in the upper portion of the reference several previous studies that ALT is more closely associ-
interval predicted NAFLD, and they did so in a dose- ated than either AST or GGT with both hepatic insulin
dependent manner. Therefore, although the mechanism resistance and later decline in hepatic insulin sensitivity
through which serum ALT is related to the risk for (26 ). GGT might also be involved in the pathogenesis of
NAFLD remains to be elucidated, ALT might be not only NAFLD through other mechanisms, such as oxidative
an indicator of liver injury due to hepatic steatosis but stress (28 ).
also an early indicator of impaired insulin signaling. With respect to other risk factors, our findings support
Among the hepatic enzymes in the present study, an association between weight gain and NAFLD (6 ). In
serum ALT concentration was more closely associated addition, the link between ALT and other confounders,
with the development of NAFLD than either AST or GGT such as triglycerides, CRP, and HOMA-IR could be re-
concentrations. This finding could be explained by the lated to incident NAFLD. Even after the potential con-
higher specificity of ALT for liver injury (11 ), as well as by founders in the present study were adjusted for, the
the contribution of ALT as a glucogenic enzyme. We have predictive ability of ALT persisted across every ALT
shown that the predictive effect of ALT on incident category compared with the referent group of the lowest
Table 3. Adjusted hazard ratios of incidence of nonalcoholic fatty liver disease in relation to GGT concentrations within the
reference interval.
GGT, U/L
<15 15–17 18–21 22–26 27–39 P for trend

Total participants (n ⫽ 5237)
Cases / person-years 144 / 3,333.0 134 / 2,432.7 229 / 2,888.0 229 / 2,320.1 248 / 2,302.8
ID (per 1000 person-years)a 43.2 55.1 79.3 98.7 107.7
aHR, 95% CI
Model 1b 1.00 1.26 (0.99–1.59) 1.81 (1.47–2.22) 2.22 (1.80–2.74) 2.38 (1.93–2.92) ⬍0.001
Model 2c 1.00 1.12 (0.88–1.42) 1.40 (1.13–1.74) 1.46 (1.18–1.82) 1.23 (0.98–1.54) 0.019
Model 3d 1.00 1.12 (0.89–1.42) 1.38 (1.11–1.70) 1.44 (1.16–1.80) 1.21 (0.96–1.52) 0.030
Model 4e 1.00 1.16 (0.91–1.48) 1.37 (1.10–1.71) 1.47 (1.17–1.84) 1.18 (0.93–1.50) 0.064
Model 5f 1.00 1.16 (0.91–1.48) 1.37 (1.10–1.72) 1.47 (1.18–1.85) 1.18 (0.93–1.50) 0.058
Ethanol ⬍10 g /day (n ⫽ 3925)
aHR (95% CI), Model 5 1.00 1.17 (0.89–1.54) 1.53 (1.19–1.95) 1.60 (1.24–2.06) 1.29 (0.99–1.68) 0.010
Non-drinkers (n ⫽ 1119)
aHR (95% CI), Model 5 1.00 1.24 (0.79–1.93) 1.25 (0.82–1.91) 1.52 (0.94–2.45) 0.96 (0.58–1.60) 0.662
BMI ⬍25 kg/m2 (n ⫽ 4333)
aHR (95% CI), Model 5 1.00 1.10 (0.83–1.46) 1.48 (1.14–1.92) 1.57 (1.21–2.05) 1.37 (1.05–1.79) 0.003
BMI ⬍23 kg/m2 (n ⫽ 3226)
aHR (95% CI), Model 5 1.00 1.05 (0.72–1.53) 1.38 (0.97–1.95) 1.44 (1.00–2.07) 1.38 (0.95–1.98) 0.026
Without metabolic syndrome
(n ⫽ 5037)
aHR (95% CI), Model 5 1.00 1.15 (0.90–1.48) 1.34 (1.06–1.69) 1.46 (1.16–1.84) 1.20 (0.93–1.53) 0.044
No metabolic syndrome traits
(n ⫽ 2795)
aHR (95% CI), Model 5 1.00 1.21 (0.83–1.76) 1.54 (1.09–2.19) 1.68 (1.16–2.43) 1.13 (0.76–1.66) 0.157
ALT ⬍35 U/L and GGT ⬍40 U/L
during the study period
(n ⫽ 2959)
aHR (95% CI), Model 5 1.00 1.25 (0.87–1.78) 1.38 (0.97–1.95) 1.51 (1.06–2.16) 1.41 (0.91–2.20) 0.032
Ethanol ⬍20 g /day during the
study period (n ⫽ 4535)
aHR (95% CI), Model 5 1.00 1.08 (0.83–1.41) 1.36 (1.07–1.73) 1.485 (1.16–1.89) 1.19 (0.92–1.53) 0.045
a
ID, incidence density.
b
Model 1: adjustment for age.
c
Model 2: model 1 plus adjustment for weight change, fasting serum glucose, loge triglyceride, HDL cholesterol, BMI, systolic blood pressure, smoking, exercise,
and alcohol intake.
d
Model 3: model 2 plus adjustment for HOMA-IR.
e
Model 4: model 3 plus adjustment for CRP.
f
Model 5: model 4 plus adjustment for incident diabetes.
quintile. Therefore, increased ALT, despite remaining NAFLD. In addition, because ultrasonography cannot
within the reference interval, might be an important accurately differentiate steatosis from fibrosis (29 ), in
preclinical marker of NAFLD, possibly as a component of the present study incident NAFLD on ultrasonography
the early phase of NAFLD. might have represented the simple hepatic steatosis as
Our study had several limitations. First, NAFLD was well as another NAFLD spectrum condition, such as
not assessed by biopsy. Although in the diagnosis of nonalcoholic steatohepatitis, defined as steatosis plus
liver steatosis ultrasonography is a useful method with any stage of fibrosis or steatosis plus inflammation (31 ).
a reasonable sensitivity and specificity (29, 30 ), it may A second possible limitation is that serum ALT during
underestimate the actual rate of NAFLD, because ultra- follow-up was not included in the analysis. However,
sonography changes do appear at a hepatocyte fat an association between ALT and incident NAFLD re-
content of ⱖ15% to 30% (29 ). Another possible expla- mained even after exclusion of participants who
nation could be that a slightly increased serum ALT showed an increase of serum ALT ⱖ35 U/L and serum
concentration might reflect subclinical (or ultrasonog- GGT ⱖ40 U/L at follow-up. Finally, the information on
raphy-undetectable) early fatty changes in the liver alcohol drinking was self-reported and thus likely to
(hepatic steatosis), which predate the overtly detectable have been underreported. Nevertheless, the association
NAFLD, i.e., ALT might be a preclinical marker of of ALT with incident NAFLD in our study remained
even after exclusion of participants with ethanol intake 13. Kunde SS, Lazenby AJ, Clements RH, Abrams GA. Spectrum of
of ⬎20 g/day first reported at follow-up. NAFLD and diagnostic implications of the proposed new normal
range for serum ALT in obese women. Hepatology 2005;42:650 – 6.
In conclusion, in apparently healthy, nondiabetic Korean 14. Sakugawa H, Nakayoshi T, Kobashigawa K, Nakasone H, Kawa-
kami Y, Yamashiro T, et al. Alanine aminotransferase elevation
men, higher ALT concentrations within the reference not associated with fatty liver is frequently seen in obese Japa-
interval were an independent predictor of incident nese women. Eur J Clin Nutr 2004;58:1248 –52.
NAFLD. This relation remained significant even among 15. O’Brien RM, Granner DK. Regulation of gene expression by
normal weight participants still within the reference in- insulin. Biochem J 1991;278:609 –19.
terval of ALT and GGT at all follow-up examinations, 16. Targher G, Bertolini L, Scala L, Poli F, Zenari L, Falezza G.
irrespective of potential confounders. Further studies on Decreased plasma adiponectin concentrations are closely asso-
the relation between ALT within its reference interval and ciated with nonalcoholic hepatic steatosis in obese individuals.
NAFLD might help to elucidate the underlying mecha- Clin Endocrinol Oxf 2004;61:700 –3.
nism of the relationship between ALT and increased 17. Yokoyama H, Hirose H, Ohgo H, Saito I. Inverse association
between serum adiponectin level and transaminase activities in
NAFLD.
Japanese male workers. J Hepatol 2004;41:19 –24.
18. Lopez-Bermejo A, Botas P, Funahashi T, Delgado E, Kihara S,
Ricart W, et al. Adiponectin, hepatocellular dysfunction and insulin
We thank Dr. Eunock Oh, Dr. Kyungsoo Cha, and Dr. sensitivity. Clin Endocrinol Oxf 2004;60:256 – 63.
Eunmi Jung at Kangbuk Samsung Hospital (Seoul, Korea) 19. Matthews DR, Hosker JP, Rudenski AS, Naylor BA, Treacher DF,
for help with the revision. Turner RC. Homeostasis model assessment: insulin resistance
and ␤-cell function from fasting plasma glucose and insulin
References concentrations in man. Diabetologia 1985;28:412–9.
1. Angulo P. Nonalcoholic fatty liver disease. N Engl J Med 2002; 20. Ratziu V, Imbert-Bismut F, Messous D, Poynard T. The elusiveness
346:1221–31. of “normal” ALT in fatty liver. Hepatology 2004;39:1172.
2. Daniel S, Ben-Menachem T, Vasudevan G, Ma CK, Blumenkehl M. 21. Ryu S, Chang Y, Kim DI, Kim WS, Suh BS. gamma-Glutamyltrans-
Prospective evaluation of unexplained aminotransferase levels in ferase as a predictor of chronic kidney disease in nonhypertensive
the United States. Am J Gastroenterol 1999;94:3010 – 4. and nondiabetic Korean men. Clin Chem 2007;53:71–7.
3. Schindhelm RK, Diamant M, Dekker JM, Tushuizen ME, Teerlink T, 22. Kojima S, Watanabe N, Numata M, Ogawa T, Matsuzaki S.
Heine RJ. Alanine aminotransferase as a marker of non-alcoholic Increase in the prevalence of fatty liver in Japan over the past 12
fatty liver disease in relation to type 2 diabetes mellitus and years: analysis of clinical background. J Gastroenterol 2003;38:
cardiovascular disease. Diabetes Metab Res Rev 2006;22:437– 43. 954 – 61.
4. Marchesini G, Brizi M, Morselli-Labate AM, Bianchi G, Bugianesi E, 23. Expert Panel on Detection Evaluation, and Treatment of High Blood
McCullough AJ, et al. Association of nonalcoholic fatty liver Cholesterol in Adults. Executive summary of the third report of the
disease with insulin resistance. Am J Med 1999;107:450 –5. National Cholesterol Education Program (NCEP) Expert Panel on
5. Ryysy L, Hakkinen AM, Goto T, Vehkavaara S, Westerbacka J, Detection, Evaluation, and Treatment of High Blood Cholesterol in
Halavaara J, et al. Hepatic fat content and insulin action on free Adults (Adult Treatment Panel III). JAMA 2001;285:2486 –97.
fatty acids and glucose metabolism rather than insulin absorption 24. WHO Western Pacific Region, IASO and IOTF. The Asia-Pacific
are associated with insulin requirements during insulin therapy in Perspective: Redefining Obesity and Its Treatment. Sydney, Aus-
type 2 diabetic patients. Diabetes 2000;49:749 –58. tralia: Health Communications Australia Pty. Ltd., 2000.
6. Hamaguchi M, Kojima T, Takeda N, Nakagawa T, Taniguchi H, Fujii 25. Prati D, Taioli E, Zanella A, Della Torre E, Butelli S, Del Vecchio E,
K, et al. The metabolic syndrome as a predictor of nonalcoholic et al. Updated definitions of healthy ranges for serum alanine
fatty liver disease. Ann Intern Med 2005;143:722– 8. aminotransferase levels. Ann Intern Med 2002;137:1–10.
7. Marchesini G, Brizi M, Bianchi G, Tomassetti S, Bugianesi E, Lenzi
26. Vozarova B, Stefan N, Lindsay RS, Saremi A, Pratley RE, Bogardus
M, et al. Nonalcoholic fatty liver disease: a feature of the
C, et al. High alanine aminotransferase is associated with de-
metabolic syndrome. Diabetes 2001;50:1844 –50.
creased hepatic insulin sensitivity and predicts the development
8. Wannamethee SG, Shaper AG, Lennon L, Whincup PH. Hepatic
of type 2 diabetes. Diabetes 2002;51:1889 –95.
enzymes, the metabolic syndrome, and the risk of type 2 diabetes
27. Burgert TS, Taksali SE, Dziura J, Goodman TR, Yeckel CW, Papadem-
in older men. Diabetes Care 2005;28:2913– 8.
etris X, et al. Alanine aminotransferase levels and fatty liver in
9. Sattar N, Scherbakova O, Ford I, O’Reilly DS, Stanley A, Forrest E,
childhood obesity: associations with insulin resistance, adiponectin,
et al. Elevated alanine aminotransferase predicts new-onset type
and visceral fat. J Clin Endocrinol Metab 2006;91:4287–94.
2 diabetes independently of classical risk factors, metabolic
syndrome, and C-reactive protein in the west of Scotland coronary 28. Ceriello A, Motz E. Is oxidative stress the pathogenic mechanism
prevention study. Diabetes 2004;53:2855– 60. underlying insulin resistance, diabetes, and cardiovascular dis-
10. Kim HC, Nam CM, Jee SH, Han KH, Oh DK, Suh I. Normal serum ease? The common soil hypothesis revisited. Arterioscler Thromb
aminotransferase concentration and risk of mortality from liver Vasc Biol 2004;24:816 –23.
diseases: prospective cohort study. BMJ 2004;328:983. 29. Siegelman ES, Rosen MA. Imaging of hepatic steatosis. Semin
11. Tiikkainen M, Bergholm R, Veehkavaara S, Rissanen A, Hakkinen Liver Dis 2001;21:71– 80.
AM, Tamminen M, et al. Effects of identical weight loss on body 30. Joseph AE, Saverymuttu SH, al-Sam S, Cook MG, Maxwell JD.
composition and features of insulin resistance in obese women Comparison of liver histology with ultrasonography in assessing
with high and low liver fat content. Diabetes 2003;52:701–7. diffuse parenchymal liver disease. Clin Radiol 1991;43:26 –31.
12. Clark JM, Brancati FL, Diehl AM. The prevalence and etiology of 31. Neuschwander-Tetri BA, Caldwell SH. Nonalcoholic steatohepati-
elevated aminotransferase levels in the United States. Am J tis: summary of an AASLD Single Topic Conference. Hepatology
Gastroenterol 2003;98:960 –7. 2003;37:1202–19.
693–701 (2007) and Cardiovascular
Risk Factors
Asymmetric Dimethylarginine, Smoking, and Risk

of Coronary Heart Disease in Apparently Healthy
Men: Prospective Analysis from the Population-
Based Monitoring of Trends and Determinants in
Cardiovascular Disease/Kooperative
Gesundheitsforschung in der Region Augsburg
Study and Experimental Data
Renke Maas,1 Friedrich Schulze,1 Jens Baumert,2 Hannelore Löwel,2
Khatera Hamraz,1 Edzard Schwedhelm,1 Wolfgang Koenig,3* and Rainer H. Böger1
Background: An increased plasma concentration of the 2.33; P ⴝ 0.282) for the top vs the bottom tertile of the
endogenous nitric oxide synthase inhibitor asymmetric ADMA distribution. In cases and controls, lower
dimethylarginine (ADMA) predicts adverse clinical out- ADMA plasma concentrations were observed in smok-
come in patients with coronary heart disease. We investi- ers. Analysis of ADMA-associated risk in smokers and
gated the association between plasma concentrations of nonsmokers separately revealed substantial differences:
ADMA and risk in initially healthy smoking and non- the adjusted relative risk for future coronary events (top
smoking men. vs bottom tertile of the ADMA distribution) was 0.48
Methods: Participants for this nested case-control study (95% CI 0.16 –1.46; P ⴝ 0.198) in smokers and 2.40 (95%
came from the population-based Monitoring of Trends CI 1.14 –5.08; P ⴝ 0.021) in nonsmokers. Exposure of
and Determinants in Cardiovascular Disease/Kooperative human endothelium-derived EAhy 926 cells to tobacco
Gesundheitsforschung in der Region Augsburg study. smoke enhanced expression of the ADMA metabolizing
ADMA was measured by liquid chromatography–tandem enzyme dimethylarginine dimethylaminohydrolase 2
mass spectrometry in 88 men with incident coronary and reduced ADMA concentration.
events (fatal and nonfatal myocardial infarction and sud- Conclusions: In apparently healthy men, increased
den cardiac death) and 254 age-matched controls, with a ADMA predicts the risk for coronary events in nonsmok-
median (interquartile range) follow-up of 6.2 (3.3–7.9) years. ers, but not in smokers. This may be explained in part by
Results: After adjustment for potential confounders, the an alteration of ADMA metabolism by tobacco smoke.
relative risk for a future coronary event was 2.00 [95% © 2007 American Association for Clinical Chemistry
confidence interval (CI) 1.27–3.16; P ⴝ 0.003] for smok-
ers compared with nonsmokers and 1.35 (95% CI 0.78 – Accumulating evidence is linking asymmetric dimethyl-
arginine (ADMA),4 an endogenous inhibitor of all major
isoforms of nitric oxide synthase (NOS), to human disease
1
Institute of Experimental and Clinical Pharmacology and Toxicology,
University Hospital Hamburg-Eppendorf, Hamburg, Germany.
2
GSF, Research Center for Environment and Health, Institute of Epidemi-
4
ology, Neuherberg, Germany. Nonstandard abbreviations: ADMA, asymmetric dimethylarginine; NOS,
3
Department of Internal Medicine II–Cardiology, University of Ulm nitric oxide synthase; CHD, coronary heart disease; MONICA, Monitoring of
Medical Center, Ulm, Germany. Trends and Determinants in Cardiovascular Disease; KORA, Kooperative
*Address correspondence to this author at: Department of Internal Medicine Gesundheitsforschung in der Region Augsburg; BMI, body mass index; CSC,
II–Cardiology, University of Ulm Medical Center, Robert-Koch Strasse 8, D-89081 cigarette smoke condensate; DDAH, dimethylarginine dimethylaminohydro-
Ulm, Germany. Fax 49-731-500-45021; e-mail wolfgang.koenig@uniklinik-ulm.de. lase; TC, total serum cholesterol; HDL-C, HDL cholesterol; CRP, C-reactive
Received October 18, 2006; accepted January 19, 2007. protein; LDH, lactate dehydrogenase; GFR, glomerular filtration rate; MDRD,
Previously published online at DOI: 10.1373/clinchem.2006.081893 Modification of Diet in Renal Disease; HR, hazard ratio; CI, confidence interval.
693
694 Maas et al.: ADMA and Coronary Events in Healthy Men
(1, 2 ). Increased ADMA plasma concentrations are found pants (89 cases, 267 controls). We decided not to include
in various clinical settings ranging from renal failure and women because of their very low event rate. We excluded
liver failure to atherosclerosis, hypertension, and im- from the analysis 14 men (1 case, 13 controls) with missing
paired glucose tolerance (3-6 ). Moreover, increase of data for ADMA or any of the other considered variables.
ADMA has been identified as an independent risk factor Therefore, the study population of the present report is
for progression of atherosclerosis, cardiovascular death, based on 88 cases and 254 event-free controls ages 35 to 74
and total mortality in patients with coronary heart disease years.
(CHD) (7-9 ) and renal failure (10, 11 ) and in critically ill The outcome variable was a combination of incident
patients (12 ). In the only large prospective study that fatal or nonfatal acute myocardial infarction and sudden
included patients with and without CHD, all smokers cardiac death. According to the MONICA manual (14 ),
(29% of patients) had been excluded from the main the diagnosis of a major nonfatal myocardial infarction
statistical analysis, and a significantly increased risk for was based on acute symptoms, cardiac enzymes, and
those in the highest tertile of plasma ADMA distribution typical electrocardiogram changes. Deaths from cardio-
was mainly confined to men with a previous history of vascular causes were validated by autopsy reports, death
CHD (7 ). Data are still lacking that clearly demonstrate certificates, chart review, and information from the last
that an increase of ADMA is also associated with death treating physician.
and cardiovascular events in patients without CHD or
organ failure. Moreover, the underlying mechanisms re- survey methods
sponsible for the confounding effect of smoking on All participants completed a standardized questionnaire,
ADMA-associated risk remain to be elucidated. We there- including medical history, lifestyle, and drug intake.
fore conducted a nested case-control study to assess the Blood pressure, body height (m), body weight (kg), body
potential relevance of ADMA as a risk factor for coronary mass index (BMI; kg/m2), smoking habits, and alcohol
events in smoking and nonsmoking men without a his- consumption (g/day) were determined as described (16 ).
tory of CHD.
We assessed leisure-time physical activity on a 4-level
graded scale for winter and summer (none, ⬍1, 1–2, and
⬎2 h/week) and calculated the number of years of
study design and participants
education from the highest level of formal education
We set up this study as a prospective case-control study
completed (17 ). Actual hypertension was defined as sys-
within the population-based Monitoring of Trends and
tolic blood pressure ⱖ140 mm Hg and/or diastolic blood
Determinants in Cardiovascular Disease (MONICA)
pressure ⱖ90 mm Hg, being aware of having hyperten-
Augsburg surveys, conducted in 1989/1990 (S2) and
sion, or taking antihypertensive medication.
1994/1995 (S3). The MONICA Augsburg project was part
of the multinational WHO MONICA project (13, 14 ), for
which men and women from a general population of
cell culture experiments
more than 280 000 inhabitants of a mixed urban/rural We maintained human endothelial cells (EAhy 926 cells, a
area were randomly invited to participate (response rates hybrid human cell line derived from fusion of human
for participation in the S2 and S3 surveys were 76.9% and umbilical vein endothelial cells and A549 carcinoma cells,
74.9%, respectively). The study was approved by the local a kind gift of Dr. Edgell (University of North Carolina,
ethics committee, and all participants provided written Chapel Hill, NC) (18 ) in DMEM at 37 °C in a humidified
informed consent. atmosphere containing 50 mL CO2/L. The cigarette
Altogether, 9796 men and women ages 25 to 74 years smoke condensate (CSC) used in the cell culture experi-
participated in the 2 independent cross-sectional surveys ments was prepared from the University of Kentucky
(S2, 4940 participants; S3, 4856 participants). In the frame- reference cigarette 2R4F. We collected the particulate
work of the Kooperative Gesundheitsforschung in der phase of smoke on a Cambridge filter pad by use of a
Region Augsburg (KORA), vital statistics were assessed smoking machine under standard conditions prescribed
for all sampled persons in the 2 survey populations. by ISO4387:2000. The smoke particulate matter was dis-
Patients with incident acute coronary events before the solved in dimethyl sulfoxide at 10 g/L and stored frozen
75th year of age were identified through the population- at ⫺80 °C in separate vials. On the day of the experiment,
based MONICA/KORA coronary event registry (15 ). The each vial of CSC solution was opened and diluted in
median (interquartile range) follow-up time of the study serum-free cell culture medium to desired concentration.
population was 6.2 (3.3–7.9) years. Eighty-nine men with- Nicotine concentrations observed at CSC concentrations
out prevalent CHD or diabetes mellitus at baseline devel- of 1 and 10 mg/L matched the concentration interval
oped an incident acute coronary event (S2, 78 cases; S3, 11 found in plasma of smokers. We assessed ADMA libera-
cases). For each case, we randomly selected 3 age- and tion, RNA expression, and cytotoxicity in 6-well plates
survey-matched control individuals without an incident after treatment of cells for 48 h with CSC. Control cells
acute coronary event during follow-up from the 2 survey were treated with medium containing an equivalent
populations, resulting in a study sample of 356 partici- amount of dimethyl sulfoxide.
taqman real-time quantitative reverse CRP concentration was markedly skewed, and we there-
transcription-pcr analysis fore calculated the geometric mean (geometric SD) from
Total RNA was extracted from EAhy 296 cells by RNAzol the log-transformed data. We tested differences in mean
(Wak-Chemie) and reverse-transcribed (Superscript II, values of continuous variables for statistical significance
Invitrogen) by use of random hexamers. We quantified by t-test and differences in proportions by ␹2 test. We
mRNA expression of dimethylarginine dimethylamin- assessed associations among continuous variables by
ohydrolase 1 (DDAH1) and DDAH2 by use of the Ap- Pearson correlation coefficient r.
plied Biosystems ABI Prism 7900 HT system (TaqMan). We used Cox proportional hazards analysis to assess
We carried out TaqMan reactions in 384-well plates the independent risk for the occurrence of a 1st coronary
according to the manufacturer’s instructions (Applied event. Study participants were grouped into tertiles of
Biosystems) using premade probes for DDAH1 (Probe plasma ADMA concentrations, and the risk was calcu-
Hs00201707_m1 generating an amplicon of 77 bp from the lated relative to the bottom tertile. Results are presented
NM_012137.2 transcript) and DDHA2 (probe as hazard ratios (HRs) together with their 95% confidence
Hs00203889_m1 generating an amplicon of 85 bp from the interval (CI). First, crude HRs were calculated (model 1).
NM_013974.1 transcript) and glyceraldehyde-3-phos- Results were then adjusted for age (continuous) and
phate dehydrogenase as an endogenous control (probe survey (S2 or S3; model 2) and for age (continuous),
Hs99999905_m1 generating an amplicon of 122 bp from survey (S2, S3), education years (⬍12, ⱖ12 years), smok-
the M33197.1 transcript). We performed relative quantifi- ing status (yes, no), alcohol consumption (0 g/d, 1–39.9
cation of gene expression using the ⌬⌬CT method as g/d, ⱖ40 g/d), obesity (BMI ⬍30.0 kg/m2, ⱖBMIⱖ30.0
described in the user guide for the ABI Prism 7900 HT kg/m2), physical activity (inactive, active), actual hyper-
system. tension (no, yes), TC/HDL-C ratio (continuous), and GFR
(continuous; model 3). To further evaluate the interaction
laboratory procedures of ADMA and smoking status with regard to incident
We collected a nonfasting venous blood sample from all coronary events, we calculated adjusted HRs according to
participants in a supine resting position. We measured smoking status and tertile of ADMA, with never-smoking
total serum cholesterol (TC) and HDL cholesterol and low ADMA tertile as reference. For a test of trend, we
(HDL-C) by routine enzymatic methods. Samples for coded ADMA tertiles with their median values and
measurement of C-reactive protein (CRP) and ADMA repeated the Cox regression described above. Moreover,
were stored at ⫺70 °C until analysis. We measured serum to test for possible modifications of the ADMA effect on a
CRP concentrations with a high-sensitivity immunoradio- coronary event by risk factors, we included interaction
metric assay (range, 0.05 to 10 mg/L) as described (19 ). In terms of ADMA tertiles and the parameter under concern
cell culture supernatants, we measured ADMA and l- in the multivariate Cox regression models. Cell culture
arginine by ultrasensitive liquid chromatography–tandem data were compared by ANOVA followed by Dunnett
mass spectrometry as previously validated and described multiple comparison test.
in detail (20 ). We measured plasma concentrations of All significance tests were 2-tailed, and probability
ADMA by use of a commercial ELISA reagent set (DLD) values ⬍0.05 were considered statistically significant. All
(21 ). The lower limit of detection for the ADMA ELISA analyses were performed with the Statistical Analysis
was ⬍0.05 ␮mol/L; intra- and interassay CVs were 4.5% System (version 8.2, SAS Institute) and Prism4 (GraphPad
and 8.3%, respectively. Cross-reactivities with SDMA and Software).
l-arginine were 1.2% and ⬍0.02%, respectively. The cor-
relation coefficient of the plasma ADMA values obtained Results
by ELISA and liquid chromatography–tandem mass spec- baseline characteristics
trometry (n ⫽ 29) was 0.984. Baseline characteristics of cases and controls are shown in
We measured oxidized LDL by use of a competitive Table 1. The median (interquartile range) time to a 1st
ELISA (Mercodia). We measured lactate dehydrogenase coronary event was 2.9 (1.2–5.1) years. Men who experi-
(LDH) by use of a commercial reagent set (cytotoxicity enced an event were more frequently smokers or had
detection reagent set; Roche Diagnostics) and protein hypertension compared with controls. Likewise, signifi-
concentration by use of the Bradford protein assay (Bio- cantly higher concentrations of CRP, TC (and TC/HDL-C
Rad). We estimated glomerular filtration rate (GFR) ac- ratio), and oxidized LDL were measured in cases com-
cording to the abbreviated Modification of Diet in Renal pared with controls, whereas the difference in BMI
Disease (MDRD) formula (22 ). reached only borderline significance. Cases and controls
All analyses were run in a blinded fashion. did not differ significantly in age (matching variable),
educational level, physical activity, alcohol intake, dia-
statistical analysis stolic blood pressure, or HDL-C. Mean (SD) plasma
We computed means or proportions for baseline demo- ADMA concentrations in participants who experienced
graphic and clinical characteristics for men with and an event and in controls were also similar: 0.80 (0.22) and
without an incident coronary event. The distribution of 0.79 (0.21) ␮mol/L, P ⫽ 0.72.
Table 1. Demographic and clinical characteristics of men with (cases) and without (controls) incident coronary event.a
Characteristics Cases Controls P value
n 88 254
Age, years 61.1 (8.6) 61.2 (8.4) 0.898
Education, ⬍12 yearsb 76.1 75.2 0.860
Body mass index, kg/m2 28.6 (4.0) 27.7 (3.7) 0.058
Actual hypertensionb 76.1 56.7 0.001
Systolic blood pressure, mmHg 146.0 (22.3) 138.8 (18.2) 0.007
Diastolic blood pressure, mmHg 82.6 (13.2) 81.9 (10.7) 0.671
Physical activityb 31.8 29.5 0.686
Smoking statusb 0.013
Never smoker 21.6 31.9
Former smoker 42.0 46.9
Current smoker 36.4 21.3
Alcohol intakeb 0.746
0 g/day 18.2 16.9
0.1–39.9 g/day 48.9 53.5
ⱖ40 g/day 33.0 29.5
TC, mmol/L 6.6 (1.1) 6.3 (1.1) 0.014
HDL-C, mmol/L 1.3 (0.4) 1.3 (0.4) 0.262
TC/HDL-C 5.5 (1.6) 5.0 (1.8) 0.023
Oxidized LDL, U/L 109.8 (32.1) 93.2 (28.0) ⬍0.001
CRP, mg/Lc 2.8 (3.2) 1.7 (3.0) 0.001
Creatinine, ␮mol/L 97.2 (26.5) 97.2 (17.7) 0.198
GFR, mL/mind 75.5 (18.7) 77.4 (17.3) 0.368
ADMA, ␮mol/L 0.80 (0.22) 0.79 (0.21) 0.721
a
Data are arithmetic mean (SD), P value from t-test, unless noted otherwise.
b
%, P value from ␹2 test.
c
Geometric mean (geometric SD), P value from t-test after log transformation (n ⫽ 316, 235 controls, 81 cases).
d
Calculated according to the abbreviated MDRD formula.
associations and correlations between adma and high ADMA tertiles were 0.61, 0.77, and 0.97 ␮mol/L,
and cardiovascular risk factors respectively.
Correlations between ADMA and cardiovascular risk When all 342 men were included, irrespective of smok-
factors are shown in Table 2. ADMA and age were ing status, the trend toward a higher HR for an incident
positively correlated (r ⫽ 0.28, P ⬍0.001 in controls and coronary event in the intermediate and the top tertile vs
r ⫽ 0.41, P ⬍0.001 in cases), whereas there was an inverse the bottom tertile of the ADMA distribution did not reach
correlation of ADMA and GFR (r ⫽ ⫺0.25, P ⬍0.001 in statistical significance in any of the Cox regression models
controls and r ⫽ ⫺0.326, P ⫽ 0.002 in cases). We observed (Table 3). In the fully adjusted model (3 ) the HR for the
no statistically significant correlations of ADMA with BMI top ADMA tertile was 1.35 (95% CI 0.78 –2.33; P ⫽ 0.282).
or with systolic or diastolic blood pressure, TC, HDL-C, In the same model, the relative risk of a future coronary
oxidized LDL, or CRP. In both groups (cases and con- event was 2.00 (95% CI 1.27–3.16; P ⫽ 0.003) for smokers
trols), differences in ADMA concentrations between compared with nonsmokers. By contrast, analysis of
smokers and nonsmokers were observed: mean (SD) ADMA-associated risk for smokers and nonsmokers sep-
ADMA concentrations among cases were 0.69 (0.17) and arately revealed striking differences (Table 3). The ad-
0.87 (0.23), respectively (P ⬍0.001), and among controls, justed relative risk comparing the top tertile to the bottom
0.74 (0.20) and 0.80 (0.22), respectively (P ⫽ 0.067). tertile of the ADMA distribution was 0.48 (95% CI 0.16 –
1.46; P ⫽ 0.198) for smokers, but it was 2.40 (95% CI
relative risks for coronary events associated 1.14 –5.08; P ⫽ 0.021) for nonsmokers. Smoking was the
with increase of adma in relation to smoking only risk factor with a significant interaction with ADMA
status values (P ⫽ 0.031). In further analysis, adjusting for CRP
To assess the risk of an incident coronary event according (n ⫽ 316), the HRs were similar, with HRs (95% CIs) of
to baseline concentrations of ADMA, we calculated Cox 0.48 (0.14 –1.66) for smokers and 2.35 (1.10 –5.03) for
proportional hazard models (Table 3). For this purpose, nonsmokers.
ADMA concentrations were categorized into tertiles To explore the interaction of ADMA and smoking in
(0.36 – 0.70, 0.71– 0.85, and 0.86 –2.30 ␮mol/L). Median more detail, we assessed HRs after grouping the partici-
ADMA plasma concentrations in the low, intermediate, pants according to smoking status and tertile of ADMA
Table 2. Correlation and association between ADMA and cardiovascular risk factors in cases and controls.a
ADMA
Risk factor Cases P value Controls P value

Age 0.414 ⬍0.001 0.284 ⬍0.001
BMI ⫺0.051 0.640 ⫺0.005 0.933
Systolic blood pressure ⫺0.065 0.546 0.040 0.523
Diastolic blood pressure ⫺0.112 0.297 ⫺0.079 0.209
TC ⫺0.041 0.702 ⫺0.058 0.362
HDL-C ⫺0.021 0.844 ⫺0.093 0.142
TC/HDL-C ⫺0.070 0.517 0.038 0.551
Creatinine 0.198 0.002 0.229 0.032
GFRb ⫺0.326 0.002 ⫺0.250 ⬍0.001
Oxidized LDL 0.054 0.395 ⫺0.078 0.486
Log CRPc 0.056 0.621 0.047 0.473
Actual hypertensiond 0.606 0.944
No 0.78 (0.22) 0.79 (0.21)
Yes 0.81 (0.23) 0.79 (0.22)
Physical activityd 0.816 0.848
No 0.81 (0.24) 0.79 (0.19)
Yes 0.79 (0.19) 0.80 (0.27)
Smokerd 0.001 0.165
Never 0.88 (0.24) 0.80 (0.20)
Former 0.85 (0.22) 0.81 (0.23)
Current 0.69 (0.17) 0.74 (0.20)
a
Data are Pearson correlation coefficient r unless noted otherwise.
b
Calculated according to the abbreviated MDRD formula.
c
81 cases, 235 controls.
d
Mean (SD).
(Fig. 1). Using as reference nonsmoking men with plasma smoke concentrate, whereas changes in expression did
ADMA in the lowest tertile (n ⫽ 77), the complex inter- not reach statistical significance at lower concentrations
action of smoking and ADMA with respect to coronary (Fig. 2B). Expression of DDAH1 was not significantly
risk became even more evident. In accordance with the altered (all P ⬎0.05). None of the concentrations of CSC
initial analysis in nonsmoking men, the risk of suffering a was associated with significant increases in LDH activity
coronary event increased with tertiles of ADMA [HR 1.63 (as indicator of cytotoxicity; Fig. 2C).
(0.77–3.46) and 2.37 (1.14 – 4.95), respectively, for men in
the 2nd (n ⫽ 88) and 3rd (n ⫽ 91) ADMA tertiles]. By Discussion
contrast, the coronary risk in the middle and top ADMA The major finding of this prospective population-based
tertiles did not further increase in current smokers [HR study of 342 men (88 incident cases and 254 age-matched
2.25 (0.87–7.79) in the 2nd ADMA tertile (n ⫽ 24) and HR controls) without a history of CHD or diabetes mellitus at
2.36 (0.81– 6.86) in the 3rd ADMA tertile (n ⫽ 17)]. baseline is that increase of ADMA is an independent risk
Smoking status was associated with the highest risk in factor for future fatal and nonfatal coronary events in
patients with low ADMA concentrations (n ⫽ 45) [HR nonsmoking men but not in smoking men. With respect to
4.49 (2.08 –9.68)]. ADMA and smoking, our findings point to a significant
interaction of smoking and ADMA-associated cardiovas-
effects of tobacco smoke on adma liberation cular risk. Furthermore, our cell culture experiments
and ddah expression in EAhy 926 cells indicate that CSC may enhance degradation of ADMA by
Incubation of EAhy 926 cells for 48 h with 1.0 and 10.0 upregulation of DDAH2.
mg/L CSC resulted in declines of the ADMA concentra-
tion in the cell culture medium (normalized to cellular smoking and adma
protein) by 28.2% and 24.8%, respectively (Fig. 2A), So far, the effect of smoking on ADMA has been discussed
whereas very low concentrations of CSC had no effect. quite controversially (23 ). In a previous study in elderly
The ADMA/l-arginine ratio declined with increasing men with a high cardiovascular risk profile, significantly
concentrations (1.0 and 10.0 mg/L) of CSC, by ⫺18.2% lower plasma ADMA concentrations were no longer
and ⫺22.0%, respectively (both P ⬍0.05). statistically significant after correction for baseline vari-
Expression of DDAH2 mRNA was augmented by ables and morbidity (24 ). Another recent study by Schna-
87.6% after 48 h of incubation with 10.0 mg/L tobacco bel et al. (9 ) found higher ADMA concentrations in
Table 3. Relative risk for the incidence of a coronary event estimated by HR with 95% CIs.
n, Cases/ Low ADMA, Intermediate ADMA, P, intermediate High ADMA, P, high vs P for
Tertile of ADMA controls 0.36–0.70 ␮mol/L 0.71–0.85 ␮mol/L vs low ADMA 0.86–2.30 ␮mol/L low ADMA trend
Model Reference value
All participants 88/254
Unadjusted 1 1.09 (0.65–1.83) 0.742 1.27 (0.77–2.11) 0.356 0.354
Adjusted for age and survey 1 1.02 (0.60–1.73) 0.953 1.20 (0.70–2.04) 0.505 0.488
Multivariable adjustmenta 1 0.97 (0.56–1.67) 0.901 1.35 (0.78–2.33) 0.282 0.262
Current smokers 32/54
Unadjusted 1 0.76 (0.33–1.73) 0.506 0.72 (0.27–1.92) 0.505 0.439
Multivariable adjustmentb 1 0.39 (0.15–1.04) 0.060 0.48 (0.16–1.46) 0.198 0.150
Former smokers 37/119
Unadjusted 1 1.12 (0.46–2.69) 0.808 1.74 (0.78–3.91) 0.179 0.148
Never smokers 19/81
Unadjusted 1 5.05 (1.09–23.44) 0.039 4.43 (0.94–20.86) 0.060 0.086
Never smokers and former smokers 56/200
Unadjusted 1 1.83 (0.88–3.82) 0.109 2.28 (1.12–4.64) 0.023 0.025
a
HRs for Cox regression models adjusting for age, survey, education level, alcohol consumption, obesity, physical activity, hypertension, TC/HDL-C ratio, GFR, and
smoking status.
b
HRs for Cox regression models adjusted for age, survey, education level, alcohol consumption, obesity, physical activity, hypertension, TC/HDL-C ratio, and GFR.
smoking compared with nonsmoking men and women in investigators is that ADMA plasma concentrations are
a population of patients with preexisting CHD. The most regulated by multiple factors and that in health and
plausible explanation for apparent discrepancies regard- disease the relative contributions of these mechanisms to
ing the effects of smoking on ADMA reported by different ADMA concentrations may be quite different (1, 25 ).
Fig. 1. HRs for a 1st coronary event in

men categorized in tertiles of plasma
ADMA concentration and smoking sta-
tus.
Cox regression model adjusted for age,
survey, education level, alcohol consump-
tion, obesity, physical activity, hyperten-
sion, and TC/HDL-C ratio. §, reference (low
ADMA tertile and nonsmokers).
Most data available today suggest that degradation of

ADMA by DDAH, rather than asymmetrical methylation
of l-arginine residues and proteolysis, plays the key role
in the regulation of ADMA concentrations (1, 25 ). Both
isoforms of DDAH have been shown to be sensitive to
oxidative stress (26 ), and ADMA concentrations tend to
rise under conditions of oxidative stress (27 ). Hence,
differences in systemic oxidative stress in patients with
and without underlying cardiovascular disease may ex-
plain some of the apparent discrepancies.
A previous analysis of the MONICA Augsburg 1994/
1995 cohort found a strong association between smoking
and various markers of systemic inflammation in men
(28 ). Unlike oxidative stress, inflammation is associated
with lower ADMA concentrations (29 ). Although in our
study there was no formal statistical interaction between
ADMA and CRP, smoking-induced inflammation may
have contributed to lower ADMA concentrations.
This study was the first to investigate a complete (lipid
and water soluble) extract of cigarette smoke; previous
studies reporting divergent results on ADMA and DDAH
expression used only the water-soluble fraction of ciga-
rette smoke (30 ) or nicotine at cytotoxic levels (31 ) or
investigated ADMA concentrations in regenerated endo-
thelial cells from carotid arteries obtained 6 weeks after
endothelial removal in rabbits chronically exposed to
nicotine (32 ).
A further explanation for divergent observations may
be that CSCs from different cigarette brands were found
to differ in their effect on the expression of a subset of
⬎1000 genes (33 ).
smoking, adma, and coronary risk

This is the first prospective study investigating the asso-
ciation of ADMA and cardiovascular risk in a nested
case-control design in a large cohort of primarily healthy
participants. All previous studies on ADMA involved
patients with preexisting CHD (6, 9 ) or organ failure
(11, 12, 34 ) or mixed cohorts (7 ). In the whole cohort on
which the present study was based (n ⫽ 3022), the
incidences of fatal and nonfatal coronary events per
100 000 persons for current smokers, former smokers, and
Fig. 2. EAhy 926 cells were cultured for 48 h with different concentra-
men who never smoked were 723, 578, and 399, respec- tions (0.0, 0.1, 1.0, and 10.0 mg/L) of cigarette smoke concentrate.
tively. With this in mind, it is striking that in the present (A), Concentrations of ADMA in conditioned medium expressed as mmol/g total
study we observed little or no additive effect of smoking cellular protein. CSC significantly reduced accumulation of ADMA in the condi-
on risks in the middle and top ADMA tertiles (Fig. 1). tioned medium. Data are mean and SE of 12 different experiments and
expressed relative to control. A significant reduction of ADMA was observed with
Even more surprising is the observation that in men with 1.0 and 10.0 mg/L cigarette smoke concentrate, P ⬍0.01 and P ⬍0.001,
low ADMA values, smoking was associated with the respectively. (B), Relative expression of DDAH2 mRNA in EAhy 926 cells
quantified by TaqMan reverse transcription PCR using glyceraldehyde-3-phos-
highest HR. Based on our in vitro data (which suggest that phate dehydrogenase as internal standard. Data are mean and SE of expression
cigarette smoke may enhance metabolism of ADMA) a relative to control (⌬⌬CT). Data are mean and SE of 12 different experiments. A
significant increase of DDAH2 expression (P ⫽ 0.025 vs reference) was
simple explanation for this apparent paradox may be that observed with 10.0 mg/L CSC. (C), Cytotoxicity of CSC was assessed as LDH
with respect to ADMA-associated risk, smoking may liberation from cultured cells in % of LDH liberation induced by Triton X-100
(positive control). Data are mean and SE of 6 different experiments. No
present a classical confounding factor. High cigarette significant differences were detected.
consumption may lower ADMA concentrations while
contributing itself to a higher incidence of cardiovascular Upregulation of DDAH expression should not raise too
events via multiple other mechanisms. much enthusiasm in smokers, though. Smoking clearly is
associated with a higher risk for cancer, and promotion of R.H.B. and R.M. have filed patents related to NOS
NO-mediated angiogenesis could have catastrophic ef- inhibitors.
fects when occurring in tumors. Indeed, recent studies
suggest that DDAH activity and expression are increased References
in human tumors (35 ) and may enhance tumor growth 1. Vallance P, Leiper J. Cardiovascular biology of the asymmetric
dimethylarginine: dimethylarginine dimethylaminohydrolase path-
(36 ). Thus, induction of DDAH expression may be an
way. Arterioscler Thromb Vasc Biol 2004;24:1023–30.
additional mechanism linking smoking and cancer. 2. Böger RH. When the endothelium cannot say ‘NO’ anymore.
ADMA, an endogenous inhibitor of NO synthase, promotes cardio-
adma and other risk factors vascular disease [Editorial]. Eur Heart J 2003;24:1901–2.
Of all risk markers evaluated (including CRP and oxi- 3. Tsikas D, Rode I, Becker T, Nashan B, Klempnauer J, Frölich JC.
dized LDL), we found a positive correlation of ADMA Elevated plasma and urine levels of ADMA and 15(S)-8-iso-
PGF2alpha in end-stage liver disease. Hepatology 2003;38:1063– 4.
concentrations only with age and a negative correlation
4. Miyazaki H, Matsuoka H, Cooke JP, Usui M, Ueda S, Okuda S, et
only with GFR. Conflicting results regarding the correla- al. Endogenous nitric oxide synthase inhibitor: a novel marker of
tion or association of ADMA and other risk markers have atherosclerosis. Circulation 1999;99:1141– 6.
been previously noted (1 ) and have been attributed at 5. Kielstein JT, Böger RH, Bode-Böger SM, Frölich JC, Haller H, Ritz E,
least in part to differences in concomitant diseases and et al. Marked increase of asymmetric dimethylarginine in patients
risk factors. In contrast to most previous studies, we with incipient primary chronic renal disease. J Am Soc Nephrol
2002;13:170 – 6.
investigated a rather healthy, unselected, population-
6. Stühlinger MC, Abbasi F, Chu JW, Lamendola C, McLaughlin TL,
based study cohort. Negative findings in our population Cooke JP, et al. Relationship between insulin resistance and an
certainly do not preclude significant correlations in high- endogenous nitric oxide synthase inhibitor. JAMA 2002;287:
er-risk populations with more extreme deviations from 1420 – 6.
the normal range of BMI or blood pressure. In this respect, 7. Valkonen VP, Paiva H, Salonen JT, Lakka TA, Lehtimaki T, Laakso
J, et al. Risk of acute coronary events and serum concentration of
it is of interest to note that in healthy men an infusion of
asymmetrical dimethylarginine. Lancet 2001;358:2127– 8.
ADMA (leading to a more than 20-fold increase of plasma 8. Lu TM, Ding YA, Lin SJ, Lee WS, Tai HC. Plasma levels of
ADMA concentration) resulted in an increase of the mean asymmetrical dimethylarginine and adverse cardiovascular events
arterial blood pressure by as little as 4.5 mm Hg (37 ). after percutaneous coronary intervention. Eur Heart J 2003;24:
1912–9.
strengths and limitations of the study 9. Schnabel R, Blankenberg S, Lubos E, Lackner KJ, Rupprecht HJ,
Espinola-Klein C, et al. Asymmetric dimethylarginine and the risk
The major strengths of the present study are its popula-
of cardiovascular events and death in patients with coronary artery
tion-based design and the length of follow-up. We took all disease: results from the AtheroGene Study. Circ Res 2005;97:
incident cases from 2 random samples of the general e53–9.
population and age-matched controls from the same 10. Zoccali C, Benedetto FA, Maas R, Mallamaci F, Tripepi G, Malatino
source. Also, we analyzed only the hard endpoints fatal LS, et al. Asymmetric dimethylarginine, C-reactive protein, and
carotid intima-media thickness in end-stage renal disease. J Am
and nonfatal myocardial infarction and sudden cardiac
Soc Nephrol 2002;13:490 – 6.
death. Nonetheless, there are limitations of the present 11. Zoccali C, Bode-Böger S, Mallamaci F, Benedetto F, Tripepi G,
study that need to be considered. Whereas the sample size Malatino L, et al. Plasma concentration of asymmetrical dimethyl-
in this event-based nested case-control design was clearly arginine and mortality in patients with end-stage renal disease: a
adequate to detect or exclude clinically meaningful differ- prospective study. Lancet 2001;358:2113–7.
ences in outcome for the primary comparison, sample size 12. Nijveldt RJ, Teerlink T, Van Der Hoven B, Siroen MP, Kuik DJ,
Rauwerda JA, et al. Asymmetrical dimethylarginine (ADMA) in
for the subgroup analysis of smokers and nonsmokers critically ill patients: high plasma ADMA concentration is an
was small. This resulted in relatively wide CIs, indicating independent risk factor of ICU mortality. Clin Nutr 2003;22:23–
limited power of the subgroup analyses, especially with 30.
regard to negative findings and absolute effect sizes. 13. WHO-MONICA Project Principal Investigators. The World Health
Organization MONICA Project (Monitoring Trends and Determi-
nants in Cardiovascular Disease): a major international collabora-
In conclusion, in an apparently healthy population with a
tion. J Clin Epidemiol 1988;41:105–14.
low to moderate cardiovascular risk, we found no signif- 14. WHO MONICA Project. WHO MONICA project: objectives and
icant relationship between plasma ADMA concentration design. Int J Epidemiol 1989;18(Suppl 1):29 –37.
and long-term cardiovascular outcome for the cohort at 15. Loewel H, Lewis M, Hoermann A, Keil U. Case finding, data quality
large. This apparent lack of association could be attrib- aspects and comparability of myocardial infarction registers:
uted at least in part to a substantial interaction of smoking results of a south German register study. J Clin Epidemiol
1991;44:249 – 60.
status, plasma ADMA concentration, and cardiovascular
16. Hense HW, Filipiak B, Döring A, Stieber J, Liese A, Keil U. Ten-year
risk. When accounting for this interaction, our data actu- trends of cardiovascular risk factors in the MONICA Augsburg
ally provide first evidence that ADMA is an independent region in southern Germany: results from 1984/1985, 1989/
risk marker in healthy nonsmoking men. 1990, and 1994/1995 surveys. CVD Prevention 1998;1:318 –
27.
17. Koenig W, Sund M, Döring A, Ernst E. Leisure-time physical activity 27. Sydow K, Münzel T. ADMA and oxidative stress. Atheroscler Suppl
but not work-related physical activity is associated with decreased 2003;4:41–51.
plasma viscosity: results from a large population sample. Circu- 28. Fröhlich M, Sund M, Lowel H, Imhof A, Hoffmeister A, Koenig W.
lation 1997;95:335– 41. Independent association of various smoking characteristics with
18. Edgell CJ, McDonald CC, Graham JB. Permanent cell line express- markers of systemic inflammation in men. Results from a repre-
ing human factor VIII-related antigen established by hybridization. sentative sample of the general population (MONICA Augsburg
Proc Natl Acad Sci U S A 1983;80:3734 –7. Survey 1994/95). Eur Heart J 2003;24:1365–72.
19. Hutchinson WL, Koenig W, Fröhlich M, Sund M, Lowe GD, Pepys 29. Zoccali C, Maas R, Cutrupi S, Pizzini P, Finocchiaro P, Cambareri
MB. Immunoradiometric assay of circulating C-reactive protein: F, et al. Asymmetric dimethyl-arginine (ADMA) response to inflam-
age-related values in the adult general population. Clin Chem mation in acute infections. Nephrol Dial Transplant 2006 [Epub
2000;46:934 – 8. ahead of print] doi:10.1093/ndt/gfl719.
20. Schwedhelm E, Tan-Andresen J, Maas R, Riederer U, Schulze F, 30. Zhang WZ, Venardos K, Chin-Dusting J, Kaye DM. Adverse effects
Böger RH. Liquid chromatography-tandem mass spectrometry of cigarette smoke on NO bioavailability: role of arginine metabo-
method for the analysis of asymmetric dimethylarginine in human lism and oxidative stress. Hypertension 2006;48:278 – 85.
plasma. Clin Chem 2005;51:1268 –71. 31. Jiang DJ, Jia SJ, Yan J, Zhou Z, Yuan Q, Li YJ. Involvement of
21. Schulze F, Wesemann R, Schwedhelm E, Sydow K, Albsmeier J, DDAH/ADMA/NOS pathway in nicotine-induced endothelial dys-
Cooke JP, et al. Determination of asymmetric dimethylarginine function. Biochem Biophys Res Commun 2006;349:683–93.
(ADMA) using a novel ELISA assay. Clin Chem Lab Med 2004;42: 32. Hamasaki H, Sato J, Masuda H, Tamaoki S, Isotani E, Obayashi S,
1377– 83. et al. Effect of nicotine on the intimal hyperplasia after endothelial
22. Levey AS, Greene T, Kusek JW, Beck GJ, Group MS. A simplified removal of the rabbit carotid artery. Gen Pharmacol 1997;28:
equation to predict glomerular filtration rate from serum creatinine 653–9.
[Abstract]. J Am Soc Nephrol 2000;11:A0828. 33. Jorgensen ED, Dozmorov I, Frank MB, Centola M, Albino AP.
23. Kielstein JT, Peter C, Adams MC. Cigarettes and ADMA: the Global gene expression analysis of human bronchial epithelial
smoke hasn’t cleared yet. Hypertension 2006;48:E20. cells treated with tobacco condensates. Cell Cycle 2004;3:1154 –
24. Eid HM, Arnesen H, Hjerkinn EM, Lyberg T, Seljeflot I. Relationship 68.
between obesity, smoking, and the endogenous nitric oxide 34. Mallamaci F, Tripepi G, Cutrupi S, Malatino LS, Zoccali C. Prog-
synthase inhibitor, asymmetric dimethylarginine. Metabolism nostic value of combined use of biomarkers of inflammation,
2004;53:1574 –9. endothelial dysfunction, and myocardiopathy in patients with
25. Maas R. Pharmacotherapies and their influence on asymmetric ESRD. Kidney Int 2005;67:2330 –7.
dimethylarginine (ADMA). Vasc Med 2005;10(Suppl 1):S49 –57. 35. Kostourou V, Robinson SP, Whitley GS, Griffiths JR. Effects of
25. Leiper JM, Santa Maria J, Chubb A, MacAllister RJ, Charles IG, overexpression of dimethylarginine dimethylaminohydrolase on
Whitley GS, et al. Identification of two human dimethylarginine tumor angiogenesis assessed by susceptibility magnetic reso-
dimethylaminohydrolases with distinct tissue distributions and nance imaging. Cancer Res 2003;63:4960 – 6.
homology with microbial arginine deiminases. Biochem J 1999; 36. Kostourou V, Robinson SP, Cartwright JE, Whitley GS. Dimethyl-
343:209 –14. arginine dimethylaminohydrolase I enhances tumour growth and
26. Murray-Rust J, Leiper J, McAlister M, Phelan J, Tilley S, Santa angiogenesis. Br J Cancer 2002;87:673– 80.
Maria J, et al. Structural insights into the hydrolysis of cellular 37. Kielstein JT, Impraim B, Simmel S, Bode-Böger SM, Tsikas D,
nitric oxide synthase inhibitors by dimethylarginine dimethylamin- Frölich JC, et al. Cardiovascular effects of systemic nitric oxide
ohydrolase. Nat Struct Biol 2001;8:679 – 83. [Erratum in Nat synthase inhibition with asymmetrical dimethylarginine in hu-
Struct Biol 2001;8:818.] mans. Circulation 2004;109:172–7.
702–710 (2007) Drug Monitoring and
Toxicology
Negative-Ion Chemical Ionization Gas

Chromatography–Mass Spectrometry Assay for
Enantioselective Measurement of Amphetamines in
Oral Fluid: Application to a Controlled Study with
MDMA and Driving Under the Influence Cases
Frank T. Peters,1 Nele Samyn,2 Thomas Kraemer,1 Wim J. Riedel,3 and
Hans H. Maurer1*
Background: Enantioselective analysis of amphetamine exception of MDEA, analytical recoveries, repeatability,

(AM), methamphetamine (MA), 3,4-methylenedioxyam- and intermediate precision were within required limits.
phetamine (MDA), 3,4-methylenedioxymethamphet- The analyte concentrations and enantiomer ratios in the
amine (MDMA), and 3,4-methylenedioxyethylamphet- application samples correlated only weakly with corre-
amine (MDEA) helps interpret toxicological results. sponding published plasma data.
Methods have been described for various matrices, but Conclusions: This sensitive, reliable, and fast GC-
so far not for oral fluid, a matrix of increasing impor- NICI-MS assay enantioselectively measures AM, MA,
tance in testing for drugs of abuse, especially in the MDA, and MDMA in oral fluid samples. Prediction of
context of driving under the influence of drugs (DUID). plasma concentrations and enantiomer ratios from re-
Methods: After dilution with 200 ␮L carbonate buffer spective oral fluid data is not possible.
(pH 9), oral fluid samples (10 –50 ␮L) were derivatized © 2007 American Association for Clinical Chemistry
with S-heptafluorobutyrylprolyl chloride. The resulting
diastereomers were extracted into 100 ␮L of cyclohex- Amphetamine (AM)4, methamphetamine (MA), and the
ane, separated by gas chromatography (HP-5MS col- amphetamine-derived designer drugs 3,4 –methylenedioxy-
umn), and detected by mass spectrometry in the nega- amphetamine (MDA), 3,4-methylenedioxymethamphet-
tive-ion chemical ionization mode (GC-NICI-MS). The amine (ecstasy, MDMA), and 3,4-methylenedioxyethylam-
method was validated and applied to samples from a phetamine (MDEA) are widely abused recreational drugs.
controlled study with MDMA and from authentic DUID AM and MA are potent stimulants (1 ), whereas the so-called
cases. entactogens MDA, MDMA, and MDEA increase alertness
Results: The derivatized AM, MA, MDA, MDMA, and and endurance while causing a sense of euphoria and
MDEA enantiomers were well separated from each greater sociability (2 ). Undesired effects include tachycar-
other. The method was linear from 5–250 ␮g/L per dia, hypertension, increased muscle tension, hyperpyr-
enantiomer of MDA and from 25–1250 ␮g/L per enan- exia, and blurred vision (1, 2 ), and many severe or fatal
tiomer of AM, MA, MDMA, and MDEA. With the intoxications have been described (1, 2 ). Evidence from
animal experiments and studies comparing recreational
users with nonusers further suggest that these drugs may
1
Department of Experimental and Clinical Toxicology, Saarland Univer- cause irreversible neurotoxicity (2, 3 ), although this is still
sity, Homburg (Saar), Germany.
2
National Institute of Criminalistics and Criminology, Brussels, Belgium.
3
Experimental Psychopharmacology Unit, Maastricht University, Maas-
4
tricht, The Netherlands Nonstandard abbreviations: AM, amphetamine; MA, methamphetamine;
* Address correspondence to this author at: Institute of Pharmacology and MDA, 3,4-methylenedioxyamphetamine; MDMA, 3,4-methylenedioxymeth-
Toxicology, Homburg, D-66421, Germany. Fax ⫹49-6841-16-26051; e-mail amphetamine; MDEA, 3,4-methylenedioxyethylamphetamine; DUID, driving
hans.maurer@uniklinikum-saarland.de. under the influence of drugs; GC-MS, gas chromatography-mass spectrometry;
Received October 10, 2006; accepted January 23, 2007. S-HFBPCl, S-heptafluorobutyrylprolyl chloride; IS, internal standard; and
Previously published online at DOI: 10.1373/clinchem.2006.081547 NICI, negative-ion chemical ionization.
702
a matter of debate (4, 5 ). In recent years, these drugs have S-heptafluorobutyrylprolyl chloride (S-HFBPCl) was syn-
also become increasingly important in the context of thesized as described (29 ).
driving under the influence of drugs (DUID) (6 ). Unde-
sired effects such as altered sensory perception, attention, Biosamples. We collected blank oral fluid samples used for
and risk-taking in decision-making behavior, as well as validation experiments from drug-free volunteers. Resid-
psychomotor effects, can impair the ability to safely ual oral fluid samples (n ⫽ 33) previously collected for
operate a vehicle (7, 8 ). DUID testing were used in this study. All personally
In the last few years, oral fluid has become increasingly identifying information was removed from the samples.
important in drug analysis. Its main advantage is the The oral fluid samples from the controlled study had been
relatively easy and noninvasive sample collection, which collected from 12 study participants 1, 2, 3, 4, and 5 h after
can be closely supervised, thus preventing adulteration or administration of 75 mg racemic MDMA (33 ). Oral fluid
substitution of samples (9, 10 ). A number of reports have samples were not available from all volunteers at each
recently been published on analysis of AM, MA, MDA, sampling time; the numbers of samples were as follows:
MDMA, and/or MDEA in oral fluid samples using im- n ⫽ 10 (1 h), n ⫽ 10 (2 h), n ⫽ 11 (3 h), n ⫽ 12 (4 h), and
munoassays (11–13 ), gas chromatography-mass spec- n ⫽ 11 (5 h). All samples were unstimulated oral fluid
trometry (GC-MS) (14 –21 ), liquid chromatography with samples (0.2– 0.5 mL) collected by spitting into a dry
fluorescence detection (22 ), or liquid chromatography– polypropylene tube. They were frozen immediately after
tandem mass spectrometry (23–25 ). Although informa- collection and stored without preservatives at ⫺20 °C
tion on enantiomer composition in clinical and forensic until achiral analysis. After thawing, the oral fluid sam-
samples can help interpret analytical results, the chirality ples were centrifuged to eliminate food residues and cell
of the above drugs has been neglected. The S-(⫹)-enanti- debris, and the clean supernatant was recovered for
omers of the drugs are known to be more potent than the analysis. The remainder of these clean oral fluid speci-
mens was stored at ⫺20 °C until enantioselective analysis.
respective R-(⫺)-enantiomers, and their toxicokinetics are
known to be different (2, 26 –30 ). Furthermore, not only
sample preparation
are AM and MA abused as stimulants, but both are also
We diluted 50-␮L aliquots of oral fluid with 200 ␮L
metabolites of several (racemic or optically pure) thera-
aqueous carbonate buffer (35 g/L sodium bicarbonate
peutic drugs, or are even used as such themselves (27, 31 ).
and 15 g/L sodium carbonate, pH 9). After adding 25 ␮L
Hence, their enantioselective analysis can provide the
methanolic solution of the racemic internal standards (IS)
information needed for differentiation of illicit and ther-
MDA-d5 (0.2 mg/L), AM-d11, MA-d5 and MDMA-d5 (1.0
apeutic drug use. In the case of MDMA, enantiomer ratios
mg/L each), and 6 ␮L derivatization reagent (0.1 mol/L
(R vs S) in plasma samples increase steadily over time,
S-HFBPCl in dichloromethane), we sealed the reaction
which might be useful for estimating the time since
vials and left them on a rotary shaker at ambient temper-
ingestion (26, 30 ). Numerous procedures have been pub- ature for 30 min. We added 100 ␮L cyclohexane to the
lished for enantioselective analysis of these analytes in reaction vials, resealed them, and placed them on a rotary
various biological matrices, but so far not in oral fluid. shaker for 1 min. After phase separation by centrifugation
The only exception is a method by Matin et al. (32 ) that (10 000g for 1 min), we transferred the cyclohexane phase
was limited to analysis of AM enantiomers and based on to autosampler vials. We injected 3-␮L aliquots into the
now-outdated analytical equipment. GC-MS system. Samples with concentrations exceeding
Therefore, the first aim of this study was to develop the highest points of the calibration curves were reana-
and validate an assay for simultaneous measurement of lyzed using 10-␮L volumes. If the results still exceeded
concentrations and ratios of AM, MA, MDA, MDMA, and the calibration interval after this reduction of sample
MDEA enantiomers in oral fluid. The second aim was to volume, approximate concentrations were estimated by
apply the assay to samples from a controlled study with extrapolation.
MDMA and from authentic DUID cases and compare the
results to published plasma data (28, 30 ). enantioselective quantification
The prepared samples were analyzed using an Agilent
Materials and Methods Technologies 6890 Series GC system combined with an
materials Agilent 5973 network mass selective detector, an Agilent
Chemicals and reagents. We obtained methanolic solutions 7683 series injector, and an Agilent enhanced Chem
(1000 mg/L) of racemic AM-d11 and MA-d5 and methan- Station G1701CA, v. C.00.00 21-Dec-1999. GC conditions
olic solutions (100 mg/L) of racemic MDA-d5 and were as follows: splitless injection mode, split vent
MDMA-d5 from Promochem; racemic hydrochlorides of opened after 1.5 min; 5% phenylmethyl siloxane column
AM, MA, MDA, MDMA, and MDEA from Lipomed; (HP-5MS; 30 m ⫻ 0.25 mm internal diameter, 250 nm film
sodium bicarbonate from Fluka; and all other chemicals thickness); injection port temperature, 280 °C; carrier gas,
from E. Merck. All chemicals were of analytical grade or helium; flow rate, 1 mL/min; column temperature, 100 °C
the highest purity available. The derivatization reagent raised to 190 °C at 30 °C/min, to 230 °C at 5 °C/min, to
704 Peters et al.: Enantioselective Analysis of Amphetamines in Oral Fluid
245 °C at 30 °C/min, to 260 °C at 5 °C/min, and to 310 °C enantiomer ratios (R vs S) for MDMA at each sampling
at 30 °C/min. Negative-ion chemical ionization (NICI)-MS time to the theoretic ratio of 1.0 for racemic mixtures by
conditions were as follows: transfer line heater, 280 °C; NICI, 1-tailed t tests. In addition, we compared the concentra-
methane (2 mL/min); source temperature, 150 °C; solvent tions and ratios of the analytes in the oral fluid samples to
delay, 7.5 min; selected-ion monitoring mode. The time the corresponding reported results in plasma samples
windows, selected ions, and electron multiplier voltage (28, 30 ). We studied correlations between the enantiomer
offsets are listed in Table 1. concentrations and their ratios (R vs S) by plotting the
Quantification was performed via peak area ratios results for the oral fluid samples against those obtained
(analyte vs IS) using linear weighted (1/x2) least-squares for plasma samples.
calibration curves. For all analytes except S-(⫹)-MA and
the 2 enantiomers of MDEA, the corresponding enanti- Results
omers of the deuterated analogs were used as IS. R-(⫺)- sample preparation and gc-nici-ms
MA-d5 was also used as IS for S-(⫹)-MA, and the enanti- quantification
omers of MDMA-d5 were also used as IS for the
We achieved baseline separation of the S-heptafluorobu-
corresponding enantiomers of MDEA.
tyrylprolyl (S-HFBP) derivatives of all 5 enantiomer pairs
within a 16-min program. The peaks corresponding to
assay validation
S-(⫹)-MDMA and R-(⫺)-MDEA were not fully resolved
We validated the method with respect to selectivity,
but could be differentiated because of their different MS
linearity, analytical recovery, precision, stability, and limit
of quantification. A detailed description of the validation properties. The observed noise was generally low, allow-
experiments is provided in the supplemental data [see ing detection of all enantiomers, even for sample volumes
Text 1 in the Data Supplement that accompanies the ⱕ50 ␮L. Supplemental data Fig. 1 shows the merged
online version of this article at http://www.clinchem. selected ion chromatograms of a derivatized and ex-
ort/content/vol53/issue4]. tracted blank sample and a derivatized and extracted
calibration sample containing 50 ␮g/L of each enantiomer
application of MDA and 250 ␮g/L of each enantiomer of AM, MA,
The 54 samples from the controlled study and the 33 MDMA, and MDEA.
samples from authentic DUID cases were assayed as
described above. We compared the oral fluid concentra- assay validation
tions of the R-(⫺)- and S-(⫹)-enantiomers of MDMA in A detailed description of the validation studies and
individual study samples visually and by 2-tailed paired observed analytical properties of the assay is provided in
t tests at each sampling time. We compared the mean supplemental data, Text 2 and Tables 1–3.
Table 1. Time windows, analytes, selected ions (with relative intensities), and electron multiplier voltage (EMV) offsets
used in the selected-ion monitoring method.
Analyte Target ion Qualifier ion 1 Qualifier ion 2
a
Time window, min m/z Relative intensity, % m/z Relative intensity, % m/z Relative intensitya, % EMV offset, V
7.50–9.50 R-(⫺)-AM-d11 399 100 379 92 439 44 0
R-(–)-AM 388 100 368 83 428 45
S-(⫹)-AM-d11 399 100 379 89 439 54
S-(⫹)-AM 388 100 368 76 428 54
9.50–12.00 R-(⫺)-MA-d5 407 100 387 49 447 23 400
R-(⫺)-MA 402 100 382 45 442 19
S-(⫹)-MA-d5 407 100 387 46 447 20
S-(⫹)-MA 402 100 382 41 442 20
12.00–13.50 R-(⫺)-MDA-d5 437 100 417 111 477 72 600
R-(⫺)-MDA 432 100 412 108 472 61
S-(⫹)-MDA-d5 437 100 417 114 477 78
S-(⫹)-MDA 432 100 412 115 472 68
13.50–16.17 R-(⫺)-MDMA-d5 451 100 431 45 491 28
R-(⫺)-MDMA 446 100 426 44 486 22
S-(⫹)-MDMA-d5 451 100 431 43 491 27
S-(⫹)-MDMA 446 100 426 48 486 25
R-(⫺)-MDEA 460 100 480 84 500 28
S-(⫹)-MDEA 460 100 480 92 500 24
a
Relative intensityion ⫽ (intensityion/intensitytarget ion) ⫻ 100%.
assay application missing values for R-(⫺)-MDA, enantiomer ratios (R vs S)

Oral fluid samples from the controlled study. We applied the could be calculated for only 12 samples, in which they
described GC-MS method to 54 oral fluid samples col- were 0.29 – 0.87.
lected during a controlled study with 12 volunteers. Fig. 1
shows plots of MDMA enantiomer concentrations (A) and Oral fluid samples from authentic DUID cases. We applied
ratios (B) vs sampling time for the 9 volunteers from the described GC-MS method to 33 oral fluid samples
whom samples were collected at all time points. Concen- collected during routine road patrols. In Fig. 2, merged
trations of R-(⫺)-MDMA for all participants and sampling selected ion chromatograms are shown for 2 derivatized
times (including those excluded in Fig. 1) were from and extracted oral fluid samples from DUID cases. The 1st
26.9 –2273 ␮g/L, and those of S-(⫹)-MDMA were from sample (A) contained AM, MDA, and MDMA, whereas
⬍25–1567 ␮g/L. Peak oral fluid concentrations were the 2nd sample (B) contained MDA, MDMA, and MDEA.
reached 1– 4 h after ingestion. In all samples, concentra- AM was present in 11 samples. Concentrations of
tions of R-(⫺)-MDMA exceeded those of S-(⫹)-MDMA, R-(⫺)-AM were 47.8 to ⬃7300 ␮g/L, with a median of 400
with statistically significant differences at all sampling ␮g/L. Concentrations of S-(⫹)-AM were 49.3 to ⬃6700
times (P ⬍0.05). The enantiomer ratios (R vs S) were from ␮g/L, with a median of 386 ␮g/L. The AM enantiomer
1.04 –1.92, increased steadily over time, and were ⬎1.0 for ratios (R vs S) were 0.97–1.33, with a mean of 1.10. MA
all sampling times (P ⬍0.001). was detected in only 1 oral fluid sample. Concentrations
The concentrations of R-(⫺)-MDA were ⬍5 ␮g/L in 42 of R-(⫺)-MA and S-(⫹)-MA were 104 and 84.7 ␮g/L,
of 54 samples, with a maximum concentration of 21.4 respectively, with the enantiomer ratio (R vs S) being 1.23.
␮g/L. S-(⫹)-MDA was present at concentrations ⬍5 ␮g/L MDA was present in 30 samples. In 4 samples, concentra-
in only 15 samples. For this enantiomer, the maximum tions of R-(⫺)-MDA were ⬍5 ␮g/L. The highest mea-
concentration was 74.7 ␮g/L. Because of the numerous sured concentration and the median concentration for this
enantiomer were 336 and 30.9 ␮g/L, respectively. Con-
centrations of S-(⫹)-MDA were ⬍5 ␮g/L (in 3 samples) to
979 ␮g/L, with a median concentration of 104 ␮g/L. The
enantiomer ratios (R vs S) for MDA were 0.26 – 0.52, with
a mean of 0.31. MDMA was detected in 30 oral fluid
samples. The minimum concentration of R-(⫺)-MDMA
was 40.2 ␮g/L and the maximum concentration was
⬃37 000 ␮g/L, with a median concentration of 2523 ␮g/L.
S-(⫹)-MDMA concentrations were 36.8 to ⬃26 000 ␮g/L,
with a median concentration of 1749 ␮g/L. After exclu-
sion of the results from the 3 samples for which only
approximate concentrations were measured, the R vs S
ratios of MDMA enantiomers were 1.09 –1.98. MDEA was
detected in only 1 sample. The merged selected ion
chromatogram of this sample, after derivatization and
extraction, is shown in Fig. 2B. The concentrations of
R-(⫺)-MDEA and S-(⫹)-MDEA were ⬃840 and 540 ␮g/L,
respectively, with an enantiomer ratio of ⬃1.6.
Comparison of plasma and oral fluid data. We studied corre-

lations between enantiomer concentrations and their ra-
tios (R vs S) by plotting the results for oral fluid samples
against those for corresponding plasma samples (28, 30 ).
In Fig. 3, plots are shown for the concentrations of
R-(⫺)-AM (A), S-(⫹)-AM (B), and their R vs S ratios (C),
after exclusion of the results of 1 sample for which only
approximate values were available. The ratio of the con-
centrations in oral fluid vs plasma were 2.4 – 43.2 for
R-(⫺)-AM and 3.0 – 42.6 for S-(⫹)-AM. In Fig. 4, the oral
fluid concentrations of R-(⫺)-MDMA (A), S-(⫹)-MDMA
(B), and their R vs S ratios (C) are plotted against the
corresponding data obtained from plasma samples. The
Fig. 1. Oral fluid enantiomer concentrations (A) and ratios (B) of MDMA
vs time after administration of 75 mg racemic MDMA.
correlation was too weak to allow reliable estimations of
The data points represent means and SEs of the n ⫽ 9 datasets for which
plasma concentrations from oral fluid concentrations.
samples were available at all time points. Oral fluid/plasma ratios were 2.6 – 46.3 for R-(⫺)-MDMA
Fig. 2. Merged selected ion chromatograms of oral fluid samples from authentic DUID cases after derivatization with S-HFBPCl.
The dotted lines separate the different time windows of selected-ion monitoring. (A), sample containing 768 ␮g/L R-(⫺)-AM (peak 2), 629 ␮g/L S-(⫹)-AM (peak 4), 65.9
␮g/L R-(⫺)-MDA (peak 10), 161 ␮g/L S-(⫹)-MDA (peak 12), 3947 ␮g/L R-(⫺)-MDMA (peak 14), and 2564 ␮g/L S-(⫹)-MDMA (peak 16). Peaks 9 and 11 correspond
to 250 ␮g/L R-(⫺)-MDA-d5 and S-(⫹)-MDA-d5, respectively. Peaks 1, 3, 5, 7, 13, and 15 correspond to 1250 ␮g/L R-(⫺)-AM-d11, S-(⫹)-AM-d11, R-(⫺)-MA-d5,
S-(⫹)-MA-d5, R-(⫺)-MDMA-d5, and S-(⫹)-MDMA-d5, respectively; only 0.01 mL used for analysis. (B), sample containing 15.4 ␮g/L R-(⫺)-MDA (peak 10), 45.2 ␮g/L
S-(⫹)-MDA (peak 12), 281 ␮g/L R-(⫺)-MDMA (peak 14), 210 ␮g/L S-(⫹)-MDMA (peak 16), 844 ␮g/L R-(⫺)-MDEA (peak 17), and 545 ␮g/L S-(⫹)-MDEA (peak 18).
Peaks 9 and 11 correspond to 50 ␮g/L R-(⫺)-MDA-d5 and S-(⫹)-MDA-d5. Peaks 1, 3, 5, 7, 13, and 15 correspond to 250 ␮g/L R-(⫺)-AM-d11, S-(⫹)-AM-d11, R-(⫺)-MA-d5,
S-(⫹)-MA-d5, R-(⫺)-MDMA-d5, and S-(⫹)-MDMA-d5, respectively.
and 3.5– 49.8 for S-(⫹)-MDMA. As can be seen in Fig. 4C, enantioselective liquid chromatography-tandem mass
the correlation of enantiomer ratios (R vs S) was good as spectrometry method published by Wood et al. (23 ) and
long as enantiomer ratios in plasma were ⱕ2, but above less than the sample volumes of 100 –1000 ␮L required by
this ratio the scatter increased considerably. other nonenantioselective chromatographic methods (14 –
21, 24, 25 ). The sample preparation was simple, consisting
Discussion of direct derivatization without any previous extraction
We measured the concentrations and ratios (R vs S) of steps. This was possible because oral fluid samples consist
AM, MA, MDA, MDMA, and MDEA enantiomers in oral mainly of water and contain only low concentrations of
fluid samples and compared the concentrations to those protein and mucin, and because derivatization with S-
reported for plasma (28, 30 ). The availability of sufficient HFBPCl works well under aqueous alkaline conditions
sample volume is critical in oral fluid analysis. Sample (28, 29 ). With the described temperature program, enan-
volumes of 50 ␮L or less were sufficient to perform our tioseparation of all analytes was achieved within a run-
assay, comparable to the volume required for the non- time of only 16 min, which is relatively short for separa-
Fig. 3. Oral fluid concentrations of R-(⫺)-AM (A), S-(⫹)-AM (B), and their
Fig. 4. Oral fluid concentrations of R-(⫺)-MDMA (A), S-(⫹)-MDMA (B),
R vs S ratios (C) vs corresponding plasma data (28, 30 ).
and their R vs S ratios (C) vs corresponding plasma data (28, 30 ).
tion of 5 enantiomer pairs. This is ⬃1 min longer than our

method for analysis of AM and MA enantiomers in The fragment ions of MDEA-d5, with the exception of the
plasma (29 ) and shorter than our method for analysis of molecular ion, had the same m/z values as those of
MDA, MDMA, and MDEA enantiomers in plasma (28 ). MDMA, causing interference with the quantification of
S-(⫹)-MDMA could not be fully resolved from R-(⫺)- the latter. While the alternative use of MDMA-d5 as an IS
MDEA, however, precluding the use of MDEA-d5 as an IS. for MDEA worked well for enantioselective analysis of
MDEA in plasma samples (28 ), it did not work in the reduced sample volumes, so only approximate concentra-
present method, as illustrated by the unacceptable analyt- tions could be measured.
ical recovery and precision data for MDEA [see Tables 2 Navarro et al. (21 ) studied oral fluid vs plasma ratios
and 3 in the online Data Supplement]. Extending the for total MDMA and concluded that oral fluid concentra-
chromatographic run-time or choosing a different GC tions might be helpful to estimate plasma concentrations,
column with respect to phase material or dimension but also cautioned that these conclusions were obtained
might solve this problem; however, we decided not to under controlled conditions that might be different for
modify the assay because of the low prevalence of MDEA authentic samples. When we compared the enantiomer
(34 ) and the inconvenience of an extended analysis time concentrations and ratios of the above analytes in oral
or the need to change the column, which can be used for fluid samples to plasma samples taken at the same time
enantioselective analysis of the same analytes in plasma from the same individuals (28, 30 ), very large ranges of
samples. oral fluid:plasma ratios were observed. In our opinion,
As previously observed (28, 29 ), the ionization proper- these wide ranges for oral fluid:plasma ratios preclude
ties of the S-HFBP derivatives of the primary amines were any reliable prediction of plasma concentration from oral
better than those of the secondary amines. We adjusted fluid data. In the case of AM, the same is true for the
for this difference by modifying the electron multiplier enantiomer ratios (R vs S), for which only a weak associ-
voltage during the run (Table 1). The chosen calibration ation could be found. Therefore, the only real interpreta-
intervals were in good agreement with the calibration tive value of enantioselective analysis of AM in oral fluid
intervals of nonenantioselective chromatography-based samples should be the possibility to differentiate between
methods for analysis of the same analytes in oral fluid illegal or legitimate ingestion of AM and/or one of its
(14 –17, 20 –25 ) and covered the majority of observed precursor compounds, as has been described previously
concentrations in real samples, as determined in the for other matrices (27, 29, 31, 39, 40 ). For MDMA, enan-
present study and reported in the literature (24, 34 –36 ). tiomer ratios (R vs S) in plasma have been described as
Much higher oral fluid concentrations have also been being potentially helpful for assessing the time since
reported, however, particularly for AM and MDMA. The ingestion of MDMA (26, 30 ). Unfortunately, the correla-
ability to use reduced sample volumes to achieve analyte tion between the enantiomer ratios of MDMA in oral fluid
concentrations within the intervals of the calibration and plasma samples was not strong enough to draw
curves was demonstrated using 10 ␮L of a quality control similar conclusions for the enantiomer ratios of MDMA in
sample with a concentration above the calibration oral fluid samples. As can be seen in Fig. 4C, the correla-
interval. tion is rather good as long as enantiomer ratios in plasma
Freeze-thaw stability is usually tested over 3 cycles are ⱕ2. Above that, the scatter increases, making reliable
(37 ). We tested it over 6 cycles [see Text 1 and 2 in the estimations impossible.
online Data Supplement], because oral fluid samples are
sometimes deliberately frozen and thawed before analysis
to reduce viscosity, which markedly facilitates handling We thank Gabi Ulrich, Armin A. Weber, and Andreas H.
(9 ). Ewald for their help. The controlled study was supported
The findings of oral fluid concentrations of R-(⫺)- financially by the Dutch Ministry of Transport, Transpor-
MDMA significantly exceeding those of S-(⫹)-MDMA at tation Research Center AVV. No financial disclosures
all sampling times during the controlled study, and declared.
steadily increasing enantiomer ratios (R vs S) are similar
to results reported for plasma samples (26, 28, 30, 38 ). The References
low concentrations of MDA enantiomers, particularly 1. Logan BK. Amphetamines: an update on forensic issues [Review].
R-(⫺)-MDA, in these oral fluid samples are in agreement J Anal Toxicol 2001;25:400 – 4.
with plasma data, in which R-(⫺)-MDA was also the 2. Kalant H. The pharmacology and toxicology of “ecstasy” (MDMA)
and related drugs [Review]. CMAJ 2001;165:917–28.
minor MDA enantiomer (26, 28, 30, 38 ). This is also in
3. McCann UD, Ricaurte GA. Amphetamine neurotoxicity: accom-
agreement with low concentrations of AM, the corre- plishments and remaining challenges [Review]. Neurosci Biobe-
sponding demethylation product of MA, after controlled hav Rev 2004;27:821– 6.
administration of MA (16, 35 ). 4. Easton N, Marsden CA. Ecstasy: are animal data consistent
In the 33 oral fluid samples collected during routine between species and can they translate to humans? [Review].
road patrols, AM as well as MDMA and its metabolite J Psychopharmacol 2006;20:194 –210.
MDA were most frequently detected, with 11 and 30 5. Gouzoulis-Mayfrank E, Daumann J. The confounding problem of
occurrences, respectively. There was a large distribution polydrug use in recreational ecstasy/MDMA users: a brief over-
view [Review]. J Psychopharmacol 2006;20:188 –93.
of total AM, MDMA, and MDA concentrations in these
6. Moeller MR, Kraemer T. Drugs of abuse monitoring in blood for
samples, which is in accordance with the findings of control of driving under the influence of drugs [Review]. Ther Drug
Wylie et al. (34 ). In some of the samples, the concentra- Monit 2002;24:210 –21.
tions of AM and MDMA enantiomers were so high they 7. Logan BK. Methamphetamine and driving impairment [Review]. J
did not fall within the calibration interval after using Forensic Sci 1996;41:457– 64.
8. Logan BK, Couper FJ. 3,4-Methylenedioxymethamphetamine 23. Wood M, de Boeck G, Samyn N, Morris M, Cooper DP, Maes RA,
(MDMA, ecstasy) and driving impairment [Review]. J Forensic Sci et al. Development of a rapid and sensitive method for the
2001;46:1426 –33. quantitation of amphetamines in human plasma and oral fluid by
9. Kintz P, Samyn N. Use of alternative specimens drugs of abuse in LC-MS-MS. J Anal Toxicol 2003;27:78 – 87.
saliva and doping agents in hair [Review]. Ther Drug Monit 24. Wood M, Laloup M, Fernandez MM, Jenkins KM, Young MS,
2002;24:239 – 46. Ramaekers JG, et al. Quantitative analysis of multiple illicit drugs
in preserved oral fluid by solid-phase extraction and liquid chro-
10. Maurer HH. Advances in analytical toxicology: Current role of liquid matography-tandem mass spectrometry. Forensic Sci Int 2005;
chromatography-mass spectrometry for drug quantification in 150:227–38.
blood and oral fluid [Review]. Anal Bioanal Chem 2005;381: 25. Wylie FM, Torrance H, Anderson RA, Oliver JS. Drugs in oral fluid
110 – 8. Part I. Validation of an analytical procedure for licit and illicit drugs
11. Cooper G, Wilson L, Reid C, Hand C, Spiehler V. Validation of the in oral fluid. Forensic Sci Int 2005;150:191– 8.
Cozart Amphetamine Microplate EIA for the analysis of amphet- 26. Fallon JK, Kicman AT, Henry JA, Milligan PJ, Cowan DA, Hutt AJ.
amines in oral fluid. Forensic Sci Int 2006;159:104 –12. Stereospecific analysis and enantiomeric disposition of 3,4-meth-
12. Laloup M, Tilman G, Maes V, Boeck GD, Wallemacq P, Ramaekers ylenedioxymethamphetamine (Ecstasy) in humans. Clin Chem
J, et al. Validation of an ELISA-based screening assay for the 1999;45:1058 – 69.
detection of amphetamine, MDMA and MDA in blood and oral 27. Kraemer T, Maurer HH. Toxicokinetics of amphetamines: metab-
olism and toxicokinetic data of designer drugs, of amphetamine,
fluid. Forensic Sci Int 2005;153:29 –37.
methamphetamine and their N-alkyl derivatives [Review]. Ther
13. Crouch DJ, Walsh JM, Flegel R, Cangianelli L, Baudys J, Atkins R. Drug Monit 2002;24:277– 89.
An evaluation of selected oral fluid point-of-collection drug-testing 28. Peters FT, Samyn N, Lamers C, Riedel W, Kraemer T, de Boeck
devices. J Anal Toxicol 2005;29:244 – 8. G, et al. Drug testing in blood: validated negative-ion chemical
14. Scheidweiler KB, Huestis MA. A validated gas chromatographic- ionization gas chromatographic-mass spectrometric assay for
electron impact ionization mass spectrometric method for meth- enantioselective determination of the designer drugs MDA,
ylenedioxymethamphetamine (MDMA), methamphetamine and MDMA (Ecstasy) and MDEA and its application to samples
metabolites in oral fluid. J Chromatogr B Analyt Technol Biomed from a controlled study with MDMA. Clin Chem 2005;51:1811–
Life Sci 2006;835:90 –9. 22.
15. Fucci N, De Giovanni N, Chiarotti M. Simultaneous detection of 29. Peters FT, Kraemer T, Maurer HH. Drug testing in blood: validated
negative-ion chemical ionization gas chromatographic-mass spec-
some drugs of abuse in saliva samples by SPME technique.
trometric assay for determination of amphetamine and metham-
Forensic Sci Int 2003;134:40 –5.
phetamine enantiomers and its application to toxicology cases.
16. Schepers RJ, Oyler JM, Joseph RE Jr, Cone EJ, Moolchan ET, Clin Chem 2002;48:1472– 85.
Huestis MA. Methamphetamine and amphetamine pharmacoki- 30. Peters FT, Samyn N, Kraemer T, de Boeck G, Maurer HH.
netics in oral fluid and plasma after controlled oral methamphet- Concentrations and ratios of amphetamine, methamphet-
amine administration to human volunteers. Clin Chem 2003;49: amine, MDA, MDMA, and MDEA enantiomers determined in
121–32. plasma samples from clinical toxicology and driving under the
17. Gunnar T, Ariniemi K, Lillsunde P. Validated toxicological determi- influence of drugs cases by GC-NICI-MS. J Anal Toxicol 2003;
nation of 30 drugs of abuse as optimized derivatives in oral fluid 27:552–9.
by long column fast gas chromatography/electron impact mass 31. Musshoff F. Illegal or legitimate use? Precursor compounds to
spectrometry. J Mass Spectrom 2005;40:739 –53. amphetamine and methamphetamine [Review]. Drug Metab Rev
2000;32:15– 44.
18. Yonamine M, Tawil N, Moreau RL, Silva OA. Solid-phase micro-
32. Matin SB, Wan SH, Knight JB. Quantitative determination of
extraction-gas chromatography-mass spectrometry and head-
enantiomeric compounds. I: Simultaneous measurement of the
space-gas chromatography of tetrahydrocannabinol, amphet-
optical isomers of amphetamine in human plasma and saliva
amine, methamphetamine, cocaine and ethanol in saliva
using chemical ionization mass spectrometry. Biomed Mass
samples. J Chromatogr B Analyt Technol Biomed Life Sci 2003;
Spectrom 1977;4:118 –21.
789:73– 8. 33. Samyn N, de Boeck G, Wood M, Lamers CTJ, De Waard D,
19. Toennes SW, Steinmeyer S, Maurer HJ, Moeller MR, Kauert GF. Brookhuis KA, et al. Plasma, oral fluid and sweat wipe ecstasy
Screening for drugs of abuse in oral fluid: correlation of analysis concentrations in controlled and real life conditions. Forensic Sci
results with serum in forensic cases. J Anal Toxicol 2005;29: Int 2002;128:90 –7.
22–7. 34. Wylie FM, Torrance H, Seymour A, Buttress S, Oliver JS. Drugs in
20. Mancinelli R, Gentili S, Guiducci MS, Macchia T. Simple and oral fluid Part II. Investigation of drugs in drivers. Forensic Sci Int
reliable high-performance liquid chromatography fluorimetric pro- 2005;150:199 –204.
cedure for the determination of amphetamine-derived designer 35. Drummer OH. Review: Pharmacokinetics of illicit drugs in oral
drugs. J Chromatogr B Biomed Sci Appl 1999;735:243–53. fluid. Forensic Sci Int 2005;150:133– 42 [Review].
36. de la Torre R, Farre M, Navarro M, Pacifici R, Zuccaro P, Pichini S.
21. Navarro M, Pichini S, Farre M, Ortuno J, Roset PN, Segura J, et al.
Clinical pharmacokinetics of amfetamine and related substances:
Usefulness of saliva for measurement of 3,4-methyl-
monitoring in conventional and non-conventional matrices [Re-
enedioxymethamphetamine and its metabolites: correlation with
view]. Clin Pharmacokinet 2004;43:157– 85.
plasma drug concentrations and effect of salivary pH. Clin Chem 37. Peters FT, Maurer HH. Bioanalytical method validation and its
2001;47:1788 –95. implications for forensic and clinical toxicology: a review [Review].
22. Concheiro M, de Castro A, Quintela O, Lopez-Rivadulla M, Cruz A. Accred Qual Assur 2002;7:441–9.
Determination of MDMA, MDA, MDEA and MBDB in oral fluid using 38. Pizarro N, Farre M, Pujadas M, Peiro AM, Roset PN, Joglar J, et al.
high performance liquid chromatography with native fluorescence Stereochemical analysis of 3,4-methylenedioxymethamphet-
detection. Forensic Sci Int 2005;150:221– 6. amine and its main metabolites in human samples including the
catechol-type metabolite (3,4-dihydroxymethamphetamine). Drug monitoring in adult-attention deficit hyperactivity disorder treat-
Metab Dispos 2004;32:1001–7. ment. J Anal Toxicol 2005;29:682– 8.
39. Nystrom I, Trygg T, Woxler P, Ahlner J, Kronstrand R. Quantitation 40. George S, Braithwaite RA. Using amphetamine isomer ratios to
of R-(⫺)- and S-(⫹)-amphetamine in hair and blood by gas chro- determine the compliance of amphetamine abusers prescribed
matography-mass spectrometry: an application to compliance dexedrine. J Anal Toxicol 2000;24:223–7.
711–716 (2007) Endocrinology and
Metabolism
Standardization of Insulin Immunoassays: Report of

the American Diabetes Association Workgroup
Santica Marcovina,1 Ronald R. Bowsher,2,3 W. Greg Miller,4 Myrlene Staten,5
Gary Myers,6 Samuel P. Caudill,6 Scott E. Campbell,7 and Michael W. Steffes8*
for the Insulin Standardization Workgroup
Background: Circulating insulin concentration in serum proinsulin. For 9 of 10 assays, the cross-reactivity of des
or plasma provides important information for the esti- (64, 65) proinsulin exceeded 40%. Overall, most assays
mation of insulin secretion and insulin resistance. Cur- had acceptable imprecision and specificity for insulin.
rently, lack of standardization of insulin assays hinders Conclusion: The discordance in test results for commer-
efforts to achieve consistent measures for treatment cial insulin reagent sets is likely multifactorial and will
guidelines. require a continuing effort to understand the differ-
Methods: A Workgroup convened by the American ences and achieve the desired consistency and harmo-
Diabetes Association evaluated 12 different commercial nization among commercial immunoassays.
insulin methods from 9 manufacturers. © 2007 American Association for Clinical Chemistry
Results: The within-assay CVs ranged from 3.7% to
39.0%, with 7 of 10 assays having a CV <10.6%. The The prevalence of diabetes mellitus has increased mark-
among-assay CVs ranged from 12% to 66%, with a edly during the past decade, with projections for contin-
median value of 24%. A common insulin reference ued increase worldwide. Most type 2, and some type 1,
preparation did not change the among-assay CV and patients may benefit therapeutically from more specific
failed to improve harmonization of results among as- information regarding their insulin secretion, as well as
says. Results from 6 of 10 assays agreed within the total insulin resistance. The current definitions of insulin resis-
error of 32% that is allowable based on biological tance and hyperinsulinemia are problematic, due largely
variability criteria. Seven of 10 assays recovered insulin to the fact that immunoassays for measuring circulating
added to a serum pool within 15.5% of the expected insulin have not been standardized across manufacturers
concentration. In 9 of 10 methods, there was <2% and clinical laboratories. To date, this issue has limited
cross-reactivity with intact human proinsulin, and 8 of our ability to make data comparisons for different clinical
10 methods had <3% cross-reactivity with split (32, 33) studies and to develop uniform clinical practice
guidelines.
Highly standardized and specific assays for measuring
circulating insulin are essential to achieve a consistent
1
Northwest Lipid Metabolism and Diabetes Research Laboratory, Univer- measure of insulin resistance and insulin secretion among
sity of Washington, Seattle, WA.
2
LINCO Diagnostic Services, St. Charles MO. different healthcare facilities worldwide. Despite the
3
B2S Consulting, Beech Grove, IN. awareness of analytical differences (1 ) and the potential
4
Department of Pathology, Virginia Commonwealth University, Rich- cross-reactivity of proinsulin and/or proinsulin-related
mond, VA.
5
National Institute of Diabetes and Digestive and Kidney Diseases, NIH,
peptides in insulin immunoassays, there is no consensus
Bethesda, MD. on the analytical requirements (quality specifications) for
6
Division of Laboratory Services, National Center for Environmental the assays. To address these issues, the American Diabetes
Health, Centers for Disease Control and Prevention, Atlanta, GA.
7
Association (ADA)9 in conjunction with the National
American Diabetes Association, Alexandria, VA.
8
Department of Laboratory Medicine and Pathology, Medical School,
Institute of Diabetes and Digestive and Kidney Diseases
University of Minnesota, Minneapolis, MN.
*Address correspondence to this author at: Department of Laboratory
Medicine and Pathology, Medical School, University of Minnesota, Minneap-
9
olis, MN. Fax 612-273-3489; e-mail steff001@umn.edu. Nonstandard abbreviations: ADA, American Diabetes Association; SI,
Received October 30, 2006; accepted December 27, 2007. Système International; PRMI, proposed reference material for insulin; hPI,
Previously published online at DOI: 10.1373/clinchem.2006.082214 human proinsulin.
711
712 Marcovina et al.: Standardization of Insulin Immunoassays
and the CDC convened an international work group in degree of cross-reactivity with known concentrations of
2004 to evaluate the specificity of different assays, to intact human proinsulin (hPI) and the proinsulin interme-
establish guidelines for assay acceptability, and to de- diate metabolites, split (32, 33) hPI [B–C junctional
velop a standardization program to achieve uniform, cleaved metabolite] and des (64, 65) hPI des [A–C junc-
accuracy-based values. In this paper, we report the find- tional cleaved metabolite] added to aliquots of the low-
ings and conclusions from 2 phases of this standardiza- insulin serum pool that had been supplemented with
tion effort. PRMI to a final concentration of 240 pmol/L (40 mIU/L)
to ensure that the insulin was at an easily measured
Materials and Methods concentration. In addition, each participant prepared di-
A complete description of materials and methods, includ- lutions of PRMI using the assay’s recommended diluent
ing statistical analysis, is available in the Data Supplement to evaluate parallelism of insulin recovery from the di-
that accompanies the online version of this article at luent series to the recovery from the series of serum pools
http://www.clinchem.org/content/vol53/issue4). A brief to which insulin had been added.
description is provided here.
The ADA Workgroup agreed on 2 major points: (a) the Results
standardization effort would primarily target commer- comparability of results among assays
cially available assays that were verified to measure In phase 1, all the manufacturers’ product calibrators
insulin with negligible cross-reactivity with proinsulin were reported to have assigned values traceable to the
and cleaved intermediate products of proinsulin conver- WHO 1st international reference preparation 66/304. One
sion to insulin, and (b) insulin concentrations would be assay was excluded from the evaluation, because it was
reported in Système International (SI) units (pmol/L) not compatible with plasma samples. The remaining
rather than the traditional units based on insulin biologic methods claimed that plasma was an acceptable sample.
activity. Highly purified, recombinant human insulin Results for 6 and 10 samples were excluded for assays 4
provided by Novo Nordisk and Eli Lilly and Company and 8, respectively, because they were below the assay’s
with a potency of 6.0 nmol/IU or 28.7 IU/mg pure insulin lower limits of quantification (12 and 18 pmol/L).
was used as the basis for SI molar units of insulin, Fig. 1 shows the among-assay CV for each of the 40
assuming a molecular mass of monomeric insulin of 5808
patient samples (0.6 to 1008 pmol/L) first using manufac-
Da. A multiplication factor of 6.0 was used to convert
turer’s product calibrator and then PRMI as a common
concentrations in mU/L to pmol/L (1, 2 ).
calibrator for 6 assays that were verified in phase 2 to be
specific for insulin (see Table 4 below, 1– 6). Results were
study design
excluded for assay 7 because it was not specific for
The following manufacturers participated in at least 1 of
insulin; for assays 8 and 9 because they did not participate
the 2 phases of the investigation: Abbott, Japan; Bayer
in phase 2, which evaluated specificity for insulin; for
HealthCare, USA; Beckman Coulter Inc., USA; Dako
assay 10 because it was not able to use PRMI as a
Cytomation, Ltd., United Kingdom; Diagnostic Products
calibrator; and for assays 11 and 12 because they did not
Corporation, USA; Diagnostic Systems Laboratories,
participate in phase 1. As expected, the imprecision, when
USA; Linco Research Inc., USA; Mercodia, Sweden; and
expressed as CV, was greatest at low insulin concentra-
Roche, Germany. An assay from Tosoh Bioscience Inc.,
Japan, was evaluated in-house at the University of Wash-
ington by one of the authors. Two manufacturers used
more than one assay. The participating manufacturers
requested that their respective performance outcomes
remain anonymous; consequently, each assay was as-
signed an arbitrary number for reporting results.
For the first phase of the study, heparin plasma ali-
quots from 40 donors (verified to have received no
exogenous insulin) were measured in duplicate on 3
separate days by each assay using the manufacturer’s
product calibrators and a highly purified recombinant
insulin preparation as a proposed reference material for
insulin (PRMI). To prepare working calibrators the PRMI
was diluted in each reagent set manufacturer’s recom-
mended calibrator diluent.
For phase 2 of the study, the recoveries of known Fig. 1. Among-assay CV from ANOVA on results from 40 patient
concentrations of insulin (PRMI) added to a low-insulin samples for 6 assays that were specific for insulin.
CVs are shown for results using the manufacturer’s product calibrator (open
serum pool were measured by each assay. The specificity circles) and PRMI as a common calibrator (triangles). The overall mean value
of each assay in phase 2 was assessed by determining the from all insulin specific assays is displayed on the abscissa.
tions. Use of PRMI as a common calibrator did not affect

Table 1. Within-assay CVs.a
the overall imprecision. The volunteers’ test sample re-
Insulin CV, %
sults were divided into 2 groups of 20, corresponding to
concentrations below and above ⬃69 pmol/L to isolate Assay <69 pmol/L >69 pmol/L
the influence of differences in imprecision at lower and 1 11.0 9.5
higher concentrations on the evaluation of the effect of 2 7.4 10.4
PRMI as a common calibrator. For the lower concentration 3 4.4 3.7
group, the among-assay SDs using manufacturer’s cali- 4 11.2b 12.3
brators and PRMI as common calibrator were 11.3 and 9.9 5 27.9c 16.5
pmol/L, respectively (F-test P ⫽ 0.270), and for the higher 6 4.2 6.7
concentration group were 62.2 and 78.7 pmol/L, respec- 7 21.5 6.2
8 39.0d 16.8
tively (F-test P ⫽ 0.136).
9 5.7 7.3
Harmonization of results among the methods was
10 5.7 8.7
evaluated by determining the number of results, mea-
a
Determined from the within assay SD from AVOVA on replicate results for
sured using the manufacturer’s recommended calibration,
patient samples, and the median value within each group. Two groups of n ⫽ 20,
that were within a total error allowance of ⫾25.6% from corresponding to concentrations below and above ⬃69 pmol/L, were used
the mean value for all insulin-specific assays. The 25.6% because the CVs were higher (as expected) at lower insulin concentrations.
criterion was obtained by adjusting the 32.0% total error b
This assay had a lower measurement limit of 12 pmol/L, and results for 6
criterion for a single reportable result (see Discussion) for samples were not available in the low concentration range.
c
the replication used in phase 1 by dividing the impreci-
samples had missing values for some replications in the 12–24 pmol/L range.
sion component in the equation by the square root of 3. d
Six of 10 assays (1, 2, 6, 7, 8, and 10) had ⬎87% of patients’ samples were not available in the low concentration range.
sample results within the acceptable total error range. All
results that were discrepant for this group had mean
concentrations ⱕ27.0 pmol/L, except 1 result for assay 8 ratios for each assay. The slopes for regression of the
with a mean concentration of 68.4 pmol/L. The other 4 measured vs the expected insulin concentrations for the
assays (5, 3, 4, and 9) had poorer agreement, with 5%, PRMI diluted with manufacturer’s recommended diluent
55%, 62% and 77% of results, respectively, within the (Fig. 2B) were compared with those for the PRMI added to
acceptable total error range. a serum pool (Fig. 2A). For this comparison, only the
PRMI dilutions that were in the same approximate con-
imprecision of insulin assays centration range as the samples with PRMI added to
Within-assay CVs for the results for the 40 patient sam- serum (60 – 600 pmol/L) were used. Table 3 shows the
ples when each assay was calibrated with the respective results for comparison of slopes. The slopes were signifi-
manufacturer’s product calibrators are shown in Table 1. cantly different for all assays (P ⬍0.001 to 0.050). Slopes
The median value within each group was used as the with an apparently small magnitude difference yet having
denominator to calculate CV from the within-assay SD. a strong statistical significance occurred because of very
small standard errors for 1 or both of the slopes.
insulin recovery and parallelism
In phase 2 of the evaluation, standard additions of the influence of protease inhibitor addition on
PRMI were made to a serum pool with low insulin insulin quantification
concentration (40.2– 65.4 pmol/L by the different assays). Because proinsulin and its cleaved metabolites have been
Linear regression was performed between the expected reported to have limited stability in human serum (per-
and recovered insulin concentrations for the 120 to 600 sonal communication, Ronald Bowsher), the protease
pmol/L data points for each assay in Fig. 2A to estimate inhibitor Pefabloc SC (AEBSF) was added to the serum
a mean recovery. The 60 pmol/L data were not used in aliquots before supplementation to promote stability of
the regression due to the wider dispersion of results for the proinsulin-related peptides. One serum aliquot was
most methods at this concentration. Seven assays recov- not supplemented with AEBSF and served as a control for
ered insulin within 15.5% of the added amounts, and 3 the impact of the protease inhibitor. The addition of
assays had recoveries of 64%–79% of the added amounts AEBSF decreased results of 5 assays by 0%–5%, 3 others
(Table 2). by 5%–10%, and 1 assay by 13.6%, but increased results
Phase 2 also included each assay manufacturer making for a final assay by 24.4% (from 218 to 272 pmol/L).
dilutions of PRMI with the assay’s recommended diluent
to the same approximate concentrations used for product cross-reactivities of intact proinsulin and
calibrators in that assay. The diluted PRMI specimens cleaved proinsulin metabolites
were then measured with that assay calibrated according The cross-reactivities of intact proinsulin (hPI) and the
to the manufacturer’s usual practice. Fig. 2B shows the proinsulin metabolites des (64, 65) hPI [A–C junctional
recovery of the concentrations expected from the dilution cleaved metabolite] and split (32, 33) hPI [B–C junctional
Table 3. Comparison of slopes between insulin recovered

from PRMI diluted with the recommended assay diluent,
and recovered from PRMI added to a serum pool.
Assay P nb PRMI diluted (SE)a PRMI added (SE)a
1 0.029 5 0.99 (0.023) 1.11 (0.037)
2 0.001 5 0.88 (0.019) 1.17 (0.051)
3 ⬍0.001 2 1.01 (0.005) 0.88 (0.013)
4 ⬍0.001 3 1.10 (0.016) 0.77 (0.018)
5 0.050 3 1.12 (0.018) 1.00 (0.045)
6 ⬍0.001 4 0.71 (0.004) 0.91 (0.013)
10 ⬍0.001 2 0.82 (0.002) 0.76 (0.037)
11 ⬍0.001 4 1.03 (0.013) 0.63 (0.012)
12 0.001 3 0.81 (0.009) 0.88 (0.008)
a
SE is standard error of the slope.
b
n is the number of points used to determine the slope for the recovery from
PRMI diluted samples. n ⫽ 6 for each assay for the recovery from PRMI added
to the serum pool.
assay’s response produced by the added proinsulin ex-

ceeded the upper limit of quantification.
The pattern and extent of cross-reactivity were similar
for intact hPI and the split (32,33) hPI junctional metabo-
lite in all assays but 1 (assay 7). Assay 7 differed from the
others in that it displayed a high level of cross-reactivity
from split (32, 33) hPI but negligible cross-reactivity with
both intact hPI and des (64, 65) hPI. In all cases, except
assay 7, the degree of cross-reactivity was greater for the
des (64, 65) hPI junctional metabolite than split (32, 33)
hPI. This finding occurred although the serum pool was
Fig. 2. Recovery of insulin. supplemented with des (64, 65) hPI at one tenth the molar
(A), known amounts of PRMI were added to a serum pool. (B), known amounts of
PRMI were diluted with the diluent recommended by each manufacturer for the
respective assay. Complete recovery is 1.0. The key shows the symbols for the
different assays, numbered as in the text. Table 4. Cross-reactivities of insulin-related peptides in
insulin assays.a
Estimated cross-reactivities of proinsulin (hPI) and related
cleaved metabolite] in each assay are summarized in peptides, %b
Table 4. For 8 of 10 assays, intact hPI produced estimated Assay Intact hPI Split (32, 33) hPI Des (64, 65) hPI
cross-reactivities of no more than 2.3%. However, for 1 0.7 1.3 61.2
assay 11, the cross-reactivity from intact hPI had an 2 0.8 1.5 67.2
estimated potency ⬎27% that of insulin. The extent of 3 ⫺1.8 ⫺2.2 41.9
cross-reactivity could not be estimated exactly, as the 4 0.8 1.0 54.0
5 1.8 2.1 68.8
6 0.1 0.3 42.2
Table 2. Insulin recovered from PRMI added to a serum pool. 7 0.2 98.1 2.5
Recovery fraction (95% confidence interval) at: 10 0.1 0.0 41.2
11 ⬎27.0c ⬎27.0c 20.5
Assay 60 pmol/L 600 pmol/L
12 2.3 2.6 39.8
1 1.020 (1.013–1.028) 1.105 (1.097–1.112) a
As described in Materials and Methods, portions of a serum pool were
2 0.977 (0.972–0.983) 1.154 (1.149–1.160) supplemented with 6000 pmol/L intact hPI, 6000 pmol/L split (32, 33) hPI, and
3 0.872 (0.871–0.873) 0.885 (0.884–0.886) 600 pmol/L Des (64, 65) hPI.
b
4 0.765 (0.763–0.767) 0.781 (0.779–0.783) Estimated cross-reactivity was calculated as the difference in the interpo-
5 1.092 (1.080–1.105) 0.998 (0.985–1.011) lated results in the presence of proinsulin-related peptide vs the control ⫻ 6 to
6 0.917 (0.914–0.919) 0.919 (0.917–0.922) convert the difference to pmol/L. The difference was then divided by the
proinsulin peptide concentration and multiplied by 100 to obtain the estimated
7 1.064 (1.062–1.066) 0.997 (0.995–0.998)
percent cross-reactivity.
10 0.789 (0.788–0.789) 0.772 (0.771–0.772) c
An exact value of percent cross-reactivity could not be estimated as the
11 0.770 (0.744–0.795) 0.638 (0.612–0.663) assay’s results after addition of intact hPI and split (32, 33) hPI exceeded the
12 0.989 (0.986–0.991) 0.883 (0.881–0.886) upper limit of the analytical measurement range.
concentration of either intact hPI or split (32, 33) hPI. For should be a priority to provide a metrologically appropri-
8 of 10 assays, the approximate extent of cross-reactivity ate basis to evaluate the accuracy of routine methods.
for the des (64, 65) hPI junctional metabolite was ⱖ40%. The among-assay CVs for the patient samples were the
same when the assays were calibrated either with the
Discussion manufacturer’s product calibrator or with PRMI diluted
quality specifications, performance in the manufacturer’s recommended diluent. Because a
characteristics, and harmonization of assays common calibrator is expected to improve the agreement
An allowable total error for a single reportable insulin among results, this observation suggests the PRMI diluted
measurement can be based on the average biological in this manner was not commutable with native patient
variability of insulin concentrations and the generally samples.
accepted premise that a desirable measurement error The suitability of diluting PRMI in the manufacturer’s
should not cause more than ⬃12% error in the ability to recommended diluent was evaluated for parallelism of
categorize an individual result as belonging to a popula- results compared with those for PRMI added to a serum
tion for which a distribution-based reference interval has pool. The slopes for recovered vs expected insulin for
been established (3 ). The within-individual biological vari- each type of sample were statistically different for all of
ability for insulin has been reported as 21.1% CV, and the the assays examined. For some assays, the manufacturer’s
within-group variability (for nondiseased individuals) has recommended diluent to prepare working concentra-
been reported as 58.3% CV (3 ). From these biological vari- tions of the PRMI caused a response that was clearly not
ability values for insulin, one can derive a desirable allow- parallel to that when the diluent was a serum pool. For
able measurement bias of ⫾15.5%, imprecision of 10.6% CV, other assays, the differences between slopes were small
and total analytical error of 32.0% for a single reportable enough that the impact on clinical interpretation of results
result at concentrations within or near the normal reference might be acceptable in relation to the desirable bias goal
interval, ⬃15–120 pmol/L (2.5–20 mIU/L), and presumably of ⫾15.5% described earlier. A potential limitation to this
for values at other concentrations. evaluation was that a single serum pool was used for the
For 7 assays, recovery of added insulin was within assessment, and an interfering substance could have been
15.5% of the expected values, suggesting that assay cali- present that affected the methods differently. Another
bration was acceptable and reasonably uniform over the potential limitation was that for some methods, the con-
concentration range evaluated. Three assays had lower centration range of samples from dilution of the PRMI
recoveries (64%–77%), suggesting their calibration was matched the concentrations of product calibrators rather
not adequate for clinical comparisons of results with those than those of the PRMI added to the serum pool, produc-
from other methods. The biases of these assays were ing fewer points to construct the slopes for comparison in
reasonably proportional over the concentration range and the concentration range 60 – 600 pmol/L. However, the
could be corrected by having the calibration traceable to results are persuasive that PRMI of the formulation used
an appropriate reference system (Fig. 2A). in this study will not be effective to improve the standard-
Harmonization of results among assays can be evalu- ization of insulin results. The finding regarding the lack of
ated from the patient samples measured in phase 1. benefit from using a common reference standard agrees
Because no reference measurement procedure exists for with the finding from a previous study that found little
insulin, we used the overall mean value from all insulin- improvement in variation among research-grade insulin
specific assays to assess agreement among the commercial assays when a common standard was used for calibration
methods. The allowable total error derived from biologi- (1 ), suggesting that commutability with clinical samples
cal variability was used as the criterion for agreement of must be carefully validated for any proposed reference
results with the mean value for each sample. Harmoniza- material.
tion assessment identified 6 of 10 assays with acceptable
agreement among results. Of the 8 assays that were used specificity of methods for insulin
in both phases 1 and 2, 5 had concordant conclusions Differences in specificity for proinsulin-related peptides
regarding harmonization and consistency of results, with may represent an important source of intermethod vari-
4 having acceptable performance and 1 having unaccept- ability for insulin assays. Nine of 10 assays had low
able performance. Two assays had acceptable bias based cross-reactivity with intact proinsulin, and 8 of 10 had low
on recovery of insulin from a serum pool but had only cross-reactivity with split (32, 33) hPI. The assays with
55% and 5% of the patient samples, respectively, that unacceptably high cross-reactivity for these peptides are
met the harmonization criterion. One assay had accept- not suitable for use when a specific insulin measurement
able harmonization results but a 78% recovery of in- is needed, because the circulating concentrations of these
sulin added to a serum pool. The discrepant conclusions molecules may be sufficiently high to cause misinterpre-
from the harmonization and recovery evaluations may tation of the result, a situation that is likely to be prob-
be related to the inadequacy of a mean of assays as an lematic for samples from patients with insulin resistance
appropriate target value for assay agreement. Investiga- who are known to have increased circulating concentra-
tion of a reference measurement procedure for insulin tions of proinsulin and split (32, 33) hPI cleaved forms (4 ).
All assays, except 1, had high cross-reactivity with des traceability to a common reference system to ensure that
(64, 65) hPI. This cross-reactivity is less important for results can be compared among different assays and over
clinical interpretation, however, because the circulating extended time periods.
concentrations of this metabolite are usually low relative
to intact and split (32, 33) hPI (5–7 ). It is clear from this
study that a common strategy for commercial vendors to We thank Eli Lilly Co. and Novo Nordisk for the generous
produce insulin-specific immunoassays would utilize an- provision of pure peptides to complete the work.
tisera that recognize the free N-terminal region of the
A-chain of insulin. This epitope is absent in proinsulin References
peptides that possess an intact A-C junctional region, such 1. Robbins DC, Andersen L, Bowsher R, Chance R, Dinesen B, Frank
as intact proinsulin and split (32, 33) hPI. B, et al. Report of the American Diabetes Association’s task force
on standardization of the insulin assay. Diabetes 1996;45:242–5.
2. Volund A, Brange J, Drejer K, Jensen I, Markussen J, Ribel U, et al.
In conclusion, we found that several commercial assays,
In vitro and in vivo potency of insulin analogues designed for clinical
but not all, measured insulin with acceptable imprecision use. Diabet Med 1991;8:839 – 47.
and cross-reactivities. In agreement with the conclusions 3. Fraser CE. Biological Variation: From Principles to Practice. Wash-
of the earlier ADA task force (1 ), our findings suggest that ington, DC: AACC Press, 2001.
the source of discrepancies in results among commercial 4. Haffner SM, Mykkanen L, Valdez RA, Stern MP, Holloway DL,
methods is likely multifactorial and not explainable by a Monterrosa A, et al. Disproportionately increased proinsulin levels
single analytical performance characteristic. Furthermore, are associated with the insulin resistance syndrome. J Clin Endo-
crinol Metab 1994;79:1806 –10.
not all insulin assays have acceptable performance char-
5. Given BD, Cohen RM, Shoelson SE, Frank BH, Rubenstein AH,
acteristics at concentrations as low as 12 pmol/L, as is Tager HS. Biochemical and clinical implications of proinsulin con-
needed for clinical use. version intermediates. J Clin Invest 1985;76:1398 – 405.
Improvement in harmonization of results will require 6. Bowsher RR, Wolny JD, Frank BH. A rapid and sensitive RIA for the
an ongoing effort to achieve traceability to a reference measurement of proinsulin in human serum. Diabetes 1992;41:
system for measuring insulin. The work group is investi- 1084 –90.
gating alternative preparations for insulin reference ma- 7. Ostrega D, Polonsky K, Nagi D, Yudkin J, Cox LJ, Clark PMS, et al.
Measurement of proinsulin and intermediates: Validation of immu-
terials. An isotope dilution mass spectrometry high-level
noassay methods by HPLC. Diabetes 1995;44:437– 40.
reference measurement procedure for insulin is also being 8. Cabaleiro DR, Stockl D, Kaufman JM, Fiers T, Thienpont LM.
investigated to establish calibration traceability for rou- Feasibility of standardization of serum C-peptide immunoassays
tine methods (8 ). Finalization of this reference system with isotope-dilution liquid chromatography-tandem mass spec-
will enable all assays specific for insulin to establish trometry. Clin Chem 2006;52:1193– 6.
Metabolism
Long-Term Stability of Amino Acids and

Acylcarnitines in Dried Blood Spots
Kristina Anna Strnadová,1,2 Margareta Holub,2 Adolf Mühl,2 Georg Heinze,3
Rene Ratschmann,2 Hermann Mascher,4 Sylvia Stöckler-Ipsiroglu,5
Franz Waldhauser,2 Felix Votava,1 Jan Lebl,1 and Olaf A. Bodamer2*
Background: Dried blood filter cards, collected for Conclusions: Estimation of the annual decrease of me-
newborn screening, are often stored for long periods of tabolites may allow for the retrospective diagnosis of
time. They may be suitable for the retrospective diag- inborn errors of metabolism in filter cards that have
nosis of inborn errors of metabolism, but no data are been stored for long periods of time.
currently available on the long-term stability of amino © 2007 American Association for Clinical Chemistry
acids and acylcarnitine species.
Methods: We analyzed amino acids and acylcarnitines Since the advent of newborn screening for phenylketonu-
by tandem mass spectrometry in 660 anonymous, ran- ria more than 35 years ago, neonatal screening programs
domly selected filter cards from 1989 through 2004. We have included many additional inborn errors of metabo-
assessed long-term stability of metabolites by linear lism (1, 2 ). In particular, electrospray ionization tandem
regression and estimated annual decrease of concentra- mass spectrometry (ESI-MS/MS)6 has added a number of
tion for each metabolite. different metabolites to the analytical repertoire. With
Results: Concentrations of free carnitine increased by ⬃100% sensitivity and specificity, ESI-MS/MS has be-
7.6% per year during the first 5 years of storage and come the predominant technique for neonatal screening
decreased by 1.4% per year thereafter. Alanine, arginine, of dried blood on filter cards for inborn errors of metab-
leucine, methionine, and phenylalanine decreased by olism, including amino acidopathies, fatty acid oxidation
6.5%, 3.3%, 3.1%, 7.3%, and 5.7% per year, respectively. disorders, and organic acidurias (3 ).
Acetylcarnitine, propionylcarnitine, citrulline, glycine, Filter cards from each newborn infant are stored for up
and ornithine decreased by 18.5%, 27.4%, 8.1%, 14.7%, to 10 years as a source of valuable material for epidemi-
and 16.3% per year during the first 5 years, respectively; ological studies and forensic purposes (4 ). In infants who
thereafter the decline was more gradual. Tyrosine de- have died unexpectedly, retrospective diagnosis may be
creased by 1.7% per year during the first 5 years and important to provide the family with genetic counseling
7.9% per year thereafter. We could not analyze medium- (5–12 ).
and long-chain acylcarnitine species because of low The diagnosis of inborn errors of metabolism by ESI-
physiological concentrations. MS/MS is based on the quantitative analysis of amino
acids and acylcarnitines (2, 3 ). The concentrations of me-
tabolites in affected infants exceed the respective popula-
1
Department of Paediatrics, 3rd Faculty of Medicine, Charles University, tion-dependent cutoff limits. In addition, metabolite con-
Prague, Czech Republic.
2
Department of General Pediatrics, University Children’s Hospital Vi- centrations depend on metabolic status, dietary intake (2 ),
enna, Vienna, Austria. and maturity of the infants (13, 14 ), which may pose a
3
Section of Clinical Biometrics, Core Unit for Medical Statistics and diagnostic challenge. Time-dependent degradation of me-
Informatics, Medical, University of Vienna, Vienna, Austria.
4
Pharm-analyt Laboratory GmbH, Baden, Germany.
tabolites during long-term storage of dried blood filter
5
Children’s Hospital British Columbia University, Vancouver, British cards may contribute to metabolite variability. To the best
Columbia, Canada. of our knowledge, however, few data are currently avail-
*Address correspondence to this author at: Department of General Pedi-
atrics, Division of Biochemical and Pediatric Genetics, University Children’s
Hospital Vienna, Währinger Gürtel 18-20, A-1090 Vienna, Austria. Fax 43-1-
406-3484; e-mail olaf.bodamer@meduniwien.ac.at.
6
Received August 3, 2006; accepted January 8, 2007. Nonstandard abbreviations: ESI, electrospray ionization; MS/MS, tan-
Previously published online at DOI: 10.1373/clinchem.2006.076679 dem mass spectrometry.
717
718 Strnadová et al.: Long-Term Stability of Analytes in Dried Blood Spots
able regarding the long-term stability of amino acids and functions. Their concentrations were derived from the
acylcarnitines in dried blood filter cards. ratio of the ion intensity of the substance to the corre-
Consequently, we decided to investigate the long-term sponding internal standard. We included low and high
stability of these metabolites in dried blood filter cards controls with each batch of samples (i.e., per 96-well
stored for ⬎10 years. Our objectives were to identify plate). All samples and controls were analyzed in a single
simple mathematical models to describe the time-depen- run sequence. A total of 7 runs were performed to analyze
dent change of metabolite concentrations while adjusting all 60 samples.
for random variation in subsequent years.
statistical analysis
Materials and Methods Because the data were nongaussian in distribution, all
samples data were described by median and interquartile range.
We retrieved 60 dried blood filter cards, without con- Before further statistical computations, we transformed
scious bias, from infants born in January of the years 1989, the concentrations of acylcarnitines and amino acids to a
1991, 1993, 1995, 1997, 1999, 2000, 2001, 2002, 2003, and logarithmic scale. We assessed the effect of time of storage
2004 (total n ⫽ 660) from the Neonatal Screening Archives on the concentration of each metabolite by 2 linear regres-
in Prague, Czech Republic. We chose samples 1 at a time, sion models. In the 1st model, we assumed a constant
not as a sequence of 60, to avoid sampling bias. Samples decrease of concentration over the period of 15 years. This
were handled appropriately before shipment by courier to assumption may oversimplify the actual rate of change for
the screening laboratory in Vienna. some metabolites. Thus, in a 2nd model, we introduced a
Before June 2003, blood was collected between the 5th breakpoint after 5 years of storage, allowing for 2 different
and 7th days of life; after that time, between the 3rd and constant decreases before and after 5 years of storage. The
4th days of life. This minor change in infant age does not breakpoint was not selected by means of statistical signif-
have any effect on metabolite concentrations (15 ). icance but rather was determined in advance such that
We stored the filter cards (Schleicher & Schuell 2992) at there were equal amounts of data available before and
ambient temperature in a dry environment. The cards after the breakpoint. Among all possible choices for the
were sent from the respective birth clinics to the screening breakpoint, our choice statistically guaranteed the most
center through regular public mail in sealed and bagged precise results. To simplify interpretation of the results,
envelopes; therefore, the effects of humidity, temperature, we refrained from refining our regression models any
and light should be negligible. None of the filter cards further. We verified the adequacy of the 2nd model by
was autoclaved. testing the required additional regression parameter for
statistical significance. The parameter estimates from the
acylcarnitine and amino acid analysis regression models could be interpreted as yearly declines
We analyzed all samples (16 ) within 3 weeks by the same in concentration, measured on the log scale. Thus, we
tandem mass spectrometer (MS Wallac). We punched obtained the annual percent decrease of concentration
3-mm discs (DBS Puncher, Wallac) from the center of the (L%) from the parameter estimates (B) by retransforma-
dried blood spots and transferred the discs to a 96-well tion, L% ⫽ (exp(B) ⫺ 1) 䡠 100%. The log transformation
microtiter plate (Greiner). Subsequently, we added 100 ␮L applied to the original concentrations ensured that the
methanol extraction solution, mixed with stable isotope– residuals (observed minus predicted values) were ap-
deuterated internal standards (CIL). The concentrations of proximately normally distributed with equal variance. P
the internal standards were as follows: 2H9-carnitine 0.8 values ⬍0.05 were considered statistically significant. We
␮mol/L; 2H3-acetylcarnitine 0.2 ␮mol/L; 2H3-propionyl- used the SAS System v. 8.2 (SAS Institute Inc.) for
carnitine, 2H3-butyrylcarnitine, 2H9-isovalerylcarnitine, statistical calculations.
2
H3-octanoylcarnitine, and 2H9-myristoylcarnitine 0.04
␮mol/L; 2H3-palmitoylcarnitine 0.1 ␮mol/L; 15N,213C- Results
Gly 12.5 ␮mol/L; and 2H4-Ala, 2H8-Val, 2H3-Leu, 2H3-Met, With the exception of free carnitine (C0) and Val, all
2
H5-Phe, 13C6-Tyr, 2H3-Asp, 2H3-Glu, 2H2-Orn, 2H2-Cit, metabolite concentrations showed a decrease (Table 1).
and 13C-Arg-HCl 2.5 ␮mol/L. The plate was sealed and C0 increased during the first 5 years of storage, with the
mixed. The samples were centrifuged, and the eluate largest increase occurring during the 1st year of storage,
was evaporated under vacuum. We then incubated the during which the concentration rose by ⬃40%. After
samples with butanol and 10% acetylchloride. The re- storage for more than 5 years, C0 began to gradually
sulting solution of butyl-esters of acylcarnitines and decrease. Two different constant decreases/increases per
amino acids was again evaporated and subsequently year were assumed before and after 5 years of storage.
reconstituted in 100 ␮L water:acetonitrile:formic acid Thus, we obtained 2 curves fitting the data by linear
(2500:2500:1). The source block temperature was 140 °C, regression (Fig. 1). The concentration increased by 7.6%
nebulizer gas 160 L/h, and desolvation gas 518 L/h. We (95% CI, 5.8 –9.3) per year during the first 5 years of
measured acylcarnitines by positive precursor ion scan of storage and decreased by 1.4% (0.5–2.4) per year thereaf-
m/z 85 and amino acids by different neutral loss scan ter. Val may be regarded as stable, since no significant
Table 1. Estimated change per year of free carnitine (C0),

acetylcarnitine (C2), propionylcarnitine (C3), and amino
acids during the first 5 years of storage (years of
collection, 1999 –2004) and between 5 and 15 years of
storage (years of collection, 1989 –1999).
Decrease [or increase]
per year Decrease per year
(1999–2004), (1989–1999),
Metabolite $ (95% CI) % (95% CI)
Free carnitine (C0) [7.6 (5.8–9.3)]a 1.4 (0.4–2.3)
Acetylcarnitine (C2) 18.5 (16.4–20.6) 7.5 (6.5–8.4)
Propionylcarnitine (C3) 27.4 (23.7–31.2) 7.8 (6.2–9.4)
Valine No change No change
Alanine 6.5 (5.8–7.2) 6.5 (5.8–7.2)
Arginine 3.3 (2.9–3.7) 3.3 (2.9–3.7)
Leucine 3.1 (2.5–3.6) 3.2 (2.5–3.6)
Methionine 7.3 (6.7–8) 7.3 (6.7–8)
Phenylalanine 5.7 (5.0–6.3) 5.7 (5.0–6.3)
Fig. 2. Valine.
Citrulline 8.1 (6.5–9.7) 1.5 (0.8–2.3)
No significant change of Val concentration was observed. A horizontal line fitting
Glycine 14.7 (12.8–16.6) 4.4 (3.6–5.3) the data was obtained by linear regression (see Fig. 1 in the online Data
Ornithine 16.3 (14.2–18.4) 3.5 (2.6–4.5) Supplement for phenylalanine). The x axis (logarithmic) represents the year of
blood collection, the y axis the concentration in ␮mol/L.
Tyrosine 1.6 (⫺1.5–4.9) 7.9 (6.2–9.6)
a
Increase rather than decrease.
The concentrations of C2, C3, Cit, Gly, and Orn de-
creased significantly during the 1st 5 years of storage and
change of concentration was found during the time period more gradually thereafter (Table 1, Fig. 3, and Fig. 2 in the
tested (Fig. 2 and Table 1). online Data Supplement). The absolute values for all
The concentration changes for several amino acids, concentrations are also shown in a table in the online Data
including Ala, Arg, Leu, Met, and Phe (see Fig. 1 in the Supplement.
Data Supplement that accompanies the online version of In contrast, the concentration of tyrosine was relatively
this article at http://www.clinchem.org/content/vol53/ stable, with an average annual decrease of 1.7% (⫺1.5 to
issue4), may be described by a simple exponential model, 4.9) during the first 5 years of storage. After ⬎5 years of
as the estimated decrease per year was constant during storage, the estimated annual decrease was 7.9% (6.2–9.6).
the entire 15-year period (Table 1). The estimated percen- Most acylcarnitines, including hexanoylcarnitine (C6),
tile decrease per year for Ala, Arg, Leu, Met, and Phe was octanoylcarnitine (C8), and long-chain acylcarnitines, had
6.5% (5.8 –7.2), 3.3% (2.9 –3.7), 3.1% (2.5–3.6), 7.3% (6.7– to be excluded from our study, because their physiologic
8.0), and 5.7% (5.0 – 6.3), respectively (Table 1).
Fig. 3. Acetylcarnitine.
Fig. 1. Free carnitine. Two straight lines fitting the acquired data were obtained by linear regression
Two straight lines fitting the data were obtained by linear regression. The x axis (see Fig. 2 in the online Data Supplement for propionylcarnitine). The x axis
(logarithmic) represents the year of blood collection, the y axis the concentration (logarithmic) represents the year of blood collection, the y axis the concentration
in ␮mol/L. in ␮mol/L.
concentrations are just above the detection limit of the evaluated using artificially enriched dried blood spots.
ESI-MS/MS. We also had to eliminate Asp and Glu from The stability of C6 and C8 acylcarnitine species, using
the study because their mass spectra overlap when ana- blood spots enriched with C6 and C8 and stored in sealed
lyzed by MS/MS. plastic bags at 4 °C, has previously been established. After
4.5 years, the concentration decreases of C6 and C8 in
Discussion blood were 17% and 15%, respectively (22 ). Retrospective
Residual dried blood spot samples remaining after new- analysis of metabolites that are increased in inborn errors
born screening constitute an important resource for health of metabolism would also allow evaluation of their sta-
professionals, scientists, and epidemiologists (17 ). Data bility in dried blood spots following prolonged storage.
on metabolite long-term stability would be valuable for Dried blood filter cards from children with malonic
retrospective diagnosis of inborn errors of metabolism to aciduria were reanalyzed at different time intervals for up
aid in genetic counseling of affected families (5 ). We to 900 days, and malonylcarnitine was found to decrease
investigated amino acid and acylcarnitine concentrations, exponentially, with a half-life of 248 days (23 ).
using ESI-MS/MS, in dried blood filter cards stored for Madira et al. (24 ) demonstrated a 50% to 60% decrease
up to 15 years. of phenylalanine during the first 5 years of storage at
A number of studies have examined the stability of room temperature, although according to their graph, a
genomic or viral DNA, thyroid-stimulating hormone, decrease of ⬃20% occurred during the first 4 years, which
17-␣ progesterone, or vitamin A (18 –21 ). Although these is similar to our estimate of ⬃25% decrease after the same
studies used a similar approach to investigate long-term storage time. It is difficult to conceive why there should be
stability, the biochemical properties of the investigated a decrease of 30% during year 5. The article did not report
compounds are inherently different from acylcarnitine the actual number of samples tested, nor whether samples
species and/or amino acids in dried blood spots. Conse- had been autoclaved. To our knowledge no other studies
quently the results cannot be compared with those of our
addressing the issue of long-term stability of metabolites
study.
have been published.
Interestingly, only free carnitine concentrations
According to our findings, Phe, Ala, Arg, Leu, and Met
showed an increase after storage, due to hydrolysis of
decrease exponentially, with a constant decrease during
esterified acylcarnitines (22, 23 ). Whereas we were able to
15 years of storage (Table 1 and the table in the online
demonstrate an ⬃40% increase in free carnitine after 1
Data Supplement). Methionine is the least stable of these
year of storage, other investigators observed, on average,
amino acids, which corresponds to the previous report of
a 20% increase during the same time period (22, 23 ). The
Chace et al. (25 ). These authors investigated the stability
difference may result from different storage conditions of
of amino acids in dried blood filter cards at 37 °C and
the filter cards. After 5 years of storage, concentrations of
found that the initial concentrations of Phe, Leu, Tyr, and
free carnitine gradually declined, likely because of re-
duced availability of the hydrolyzable esterified acylcar- Val decreased by 15% to 17%, whereas that of Met
nitines. Short-chain acylcarnitine species (C2, C3) fol- decreased by 24% after 30 days of storage (25 ). More
lowed a different pattern than free carnitine, as hydrolysis detailed information on concentration changes during
continued (Table 1). The concentration of C3, and the storage was not provided (25 ).
ratios of C3/C2 and C3/C0, are important for the diag- No significant decrease of Val was observed during the
nosis of propionic aciduria or methylmalonic aciduria. observation period, whereas the concentrations of Leu/Ile
The cutoff value for C3 in the Austrian screening labora- fell by an estimated 3.1% per year. At this rate of decrease,
tory is currently set at 3.5 ␮mol/L. Concentrations ob- the initial concentration would have to be at least 630
served in infants with propionic aciduria/methylmalonic ␮mol/L to exceed the 400 ␮mol/L cutoff after 15 years of
aciduria typically exceed 10 ␮mol/L (2 ). Accounting for a storage. Because the observed concentration range for
mean decrease of 27.4% per year during the first 5 years, samples from children with maple syrup urine disease
a retrospective diagnosis of propionic aciduria is still (319 to 3650 ␮mol/L) overlaps with the normal values,
feasible after 3 years of storage. Values observed in some false-negative results may be obtained during neonatal
forms of methylmalonic aciduria, or defects of intracellu- screening, but in the majority of cases a retrospective
lar cobalamin metabolism, range from 6 to 10 ␮mol/L (2 ), diagnosis would still be feasible, even after 15 years of
indicating that concentrations in samples older than 2 storage (26, 27 ).
years may not be flagged as abnormal without correction The concentration of citrulline decreased by ⬃8.1% per
for storage time. year during the first 5 years and by 1.5% per year
Because of their low physiologic concentrations, most thereafter. Consequently, the initial concentration should
medium and long acylcarnitine species, as possible indi- exceed 103 ␮mol/L to exceed the 60 ␮mol/L cutoff value
cators for fatty acid oxidation defects, could not be after 15 years of storage. These values may be observed in
evaluated in approximately two thirds of the neonatal infants with classic citrullinemia, which would most
screening population (24 ). Alternatively, the stability of likely be detectable after prolonged storage without cor-
medium- and/or long-chain acylcarnitine species can be rection for storage time.
Other factors may affect the results obtained from 9. Nuoffer JM, de Lonlay P, Costa C, Roe CR, Chamoles N, Brivet M,
testing of dried blood spot filter cards. Limited freezer et al. Familial neonatal SIDS revealing carnitine-acylcarnitine
capacity and other constraints may prevent screening translocase deficiency. Eur J Pediatr 2000;159:82–5.
10. Treacy EP, Lambert DM, Barnes R, Boriack RL, Vockley J, O’Brien
centers from storage of filter cards at lower temperatures.
LK, et al. Short-chain hydroxyacyl-coenzyme A dehydrogenase
The concentrations of most amino acids and acylcar- deficiency presenting as unexpected infant death: a family study.
nitines vary considerably among the normal newborn J Pediatr 2000;137:257–9.
population, depending on dietary intake, time of blood 11. Boles RG, Buck EA, Blitzer MG, Platt MS, Cowan TM, Martin SK.
collection, hematocrit, and position of the punch within Retrospective biochemical screening of fatty acid oxidation disor-
the dry blood spot (2, 28 ). The concentration of some ders in postmortem livers of 418 cases of sudden death in the
metabolites may be affected by overlapping spectra with first year of life. J Pediatr 1998;132:924 –33.
12. Poplawski NK, Ranieri E, Harrison JR, Fletcher JM. Multiple
other metabolites (2 ). Last, the concentration of free
acyl-coenzyme A dehydrogenase deficiency: diagnosis by acyl-
carnitine may be unpredictably affected by hydrolysis of carnitine analysis of a 12-year-old newborn screening card. J Pe-
acylcarnitines during sample preparation (22 ). diatr 1999;134:764 – 6.
13. Chace DH, Pons R, Chiriboga CA, McMahon DJ, Tein I, Naylor EW, et
al. Neonatal blood carnitine concentrations: normative data by
electrospray tandem mass spectrometry. Pediatr Res 2003;53:
This study was initiated by the Middle European Society 823–9.
for Pediatric Endocrinology (MESPE) study group. The 14. Meyburg J, Schulze A, Kohlmüller D, Pöschl J, Linderkamp O,
study material (newborn screening cards) was provided Hoffmann GF, et al. Acylcarnitine profiles of preterm infants over
by the Czech Neonatal Screening Archive in Prague the first four weeks of life. Pediatr Res 2002;52:720 –3.
within the framework of the research project MSM 15. Cavedon CT, Bourdoux P, Mertens K, Van Thi HV, Herremans N, de
0021620814; laboratory facilities were provided by the Laet C, et al. Age-related variations in acylcarnitine and free
carnitine concentrations measured by tandem mass spectrome-
Biochemical Genetics and National Neonatal Screening try. Clin Chem 2005;51:745–52.
Laboratories in Vienna, Austria. K.A.S. was a recipient of 16. Carpenter KH, Wiley V. Application of tandem mass spectrometry
financial support from PRO INFANTIBUS. to biochemical genetics and newborn screening. Clin Chim Acta
2002;322:1–10.
17. Bodamer OA, Mitterer G, Maurer W, Pollak A, Mueller MW,
References Schmidt WM. Evidence for an association between mannose-
1. Zytkovicz TH, Fitzgerald EF, Marsden D, Larson CA, Shih VE, binding lectin 2 (MBL2) gene polymorphisms and pre-term birth.
Johnson DM, et al. Tandem-mass spectrometric analysis of Genet Med 2006;8:518 –24.
amino, organic, and fatty acid disorders in newborn dried blood 18. Chaisomchit S, Wichajarn R, Janejai N, Chareonsiriwatana W. Stabil-
spots: a two year summary from the New England Newborn ity of genomic DNA in dried blood spots stored on filter paper.
Screening Program. Clin Chem 2001;47:1945–55. Southeast Asian J Trop Med Public Health 2005;36:270 –3.
2. Chace DH, Kalas TA, Naylor EW. Use of tandem mass spectrom- 19. Waite KV, Maberly GF, Eastman CJ. Storage conditions and
etry for multianalyte screening of dried blood specimens from stability of thyrotropin and thyroid hormones on filter paper. Clin
newborns. Clin Chem 2003;49:1797– 817. Chem 1987;33:853–5.
3. Schulze A, Lindner M, Kohlmüller D, Olgemöller K, Mayatepek E, 20. Torok D, Muhl A, Votava F, Heinze G, Solyom J, Crone J, et al. Stability
Hoffmann GF. Expanded newborn screening for inborn errors of of 17alpha-hydroxyprogesterone in dried blood spots after autoclav-
metabolism by electrospray ionization-tandem mass spectrome- ing and prolonged storage. Clin Chem 2002;48:370 –2.
try: results, outcome, and implications. Pediatrics 2003;111: 21. Oliver RW, Kafwembe EM, Mwandu D. Stability of vitamin A
1399 – 406. circulating complex in spots of dried serum samples absorbed
4. Therrell BL, Hannon WH, Pass KA, Lorey F, Brokopp C, Eckman J, onto filter paper. Clin Chem 1993;39:174 –5.
et al. Guidelines for the retention, storage, and use of residual 22. Chace DH, Hillman SL, Van Hove JLK, Naylor EW. Rapid diagnosis
dried blood spot samples after newborn screening analysis: of MCAD deficiency: quantitative analysis of octanoylcarnitine and
statement of the Council of Regional Networks for Genetic Ser- other acylcarnitines in newborn blood spots by tandem mass
vices. Biochem Mol Med 1996;57:116 –24. spectrometry. Clin Chem 1997;43:2106 –13.
23. Santer R, Fingerhut R, Lässker U, Wightman PJ, Fitzpatrick DR,
5. Chace DH, DiPerna JC, Mitchell BL, Sgroi B, Hofman LF, Naylor
Olgemöller B, et al. Tandem mass spectrometric determination
EW. Electrospray tandem mass spectrometry for analysis of
of malonylcarnitine: diagnosis and neonatal screening of mal-
acylcarnitines in dried postmortem blood specimens collected at
onyl-CoA decarboxylase deficiency. Clin Chem 2003;49:
autopsy from infants with unexplained cause of Death. Clin Chem
660 –2.
2001;47:1166 – 82.
24. Madira WM, Xavier F, Stern J, Wilcox AH, Barron JL. Determination
6. Rinaldo P, Stanley CA, Hsu BY, Sanchez LA, Stern HJ. Sudden and assessment of the stability of phenylalanine and tyrosine in
neonatal death in carnitine transporter deficiency. J Pediatr 1997; blood spots by HPLC. Clin Chem 1992;38:2162–3.
131:304 –5. 25. Chace DH, Adam BW, Smith SJ, Alexander JR, Hillman SL, Hannon
7. Harpey JP, Charpentier C, Coudé M, Divry P, Paturneau-Jouas M. WH. Validation of accuracy-based amino acid reference materials
Sudden infant death syndrome and multiple acyl-coenzyme A in dried-blood spots by tandem mass spectrometry for newborn
dehydrogenase deficiency, ethylmalonic-adipic aciduria, or sys- screening assays. Clin Chem 1999;45:1269 –77.
temic carnitine deficiency. J Pediatr 1987;110:881– 4. 26. Chace DH, Hillman SL, Millington DS, Kahler SG, Roe CR, Naylor
8. Howat AJ, Bennett MJ, Variend S, Shaw L, Engel PC. Defects of EW. Rapid diagnosis of maple syrup urine disease in blood spots
metabolism of fatty acids in the sudden infant death syndrome. from newborns by tandem mass spectrometry. Clin Chem 1995;
BMJ 1985;290:1771–3. 41:62– 8.
27. Rashed MS, Bucknall MP, Little D, Awad A, Jacob M, Alamoudi M, 28. Holub M, Tuschl K, Ratschmann R, Strnadova K, Sperl W, Muhl A,
et al. Screening blood spots for inborn errors of metabolism by et al. Influence of hematocrit and localisation of punch in dried
electrospray tandem mass spectrometry with a microplate batch blood spots on levels of amino acids and acylcarnitines measured
process and a computer algorithm for automated flagging of by tandem mass spectrometry. Clin Chim Acta 2006;373:27–
abnormal profiles. Clin Chem 1997;43:1129 – 41. 31.
Metabolism
Age-Associated Discrepancy between Measured

and Calculated Bioavailable Testosterone in Men
Henri Déchaud,1,2,3* Anne Denuzière,2 Sabina Rinaldi,4 Julien Bocquet,3
Hervé Lejeune,5 and Michel Pugeat1,6
Background: Bioavailable testosterone (BT) concentra- Aging in healthy men is generally associated with a
tion is considered the best marker for evaluating testic- significant decrease in the concentration of circulating
ular function in men. The decrease of BT in older men is androgens (1– 4 ). There is evidence that both testicular
more pronounced than the decrease in total testosterone and adrenal sources of androgens are involved in this
because of the parallel increase in sex hormone– binding progressive decline. In contrast to menopause, a process
globulin (SHBG) concentrations. Measurement of BT is associated with irremediable ovarian depletion in folli-
therefore crucial for the diagnosis of hypoandrogenism cles, aging of testicular function is marked by decreased
in the aging male population. capacity to produce androgens and sperm in response to
Methods: We compared BT concentrations measured by gonadotropin stimuli (1 ). In addition, contrary to the
a specific RIA after ammonium sulfate precipitation gonadotropin status observed in postmenopausal women,
(BTmeas) with those obtained by theoretical calculations aging men have no apparent increase in gonadotropin
(BTcal) in plasma samples from 694 young men (14 to 49 concentrations (1 ). Therefore, the dilemma for the clini-
years old) and 51 older men (50 to 81 years old). We cian is to recognize clinical symptoms of mild hypogo-
based theoretical calculations on Vermeulen’s simpli- nadism, such as decreased libido, erectile dysfunction,
fied mass equation using total testosterone and SHBG reduced presence of facial and pubic hair, gynecomastia,
concentrations. reduced muscular mass, and physical and psychological
Results: BTcal and BTmeas correlated significantly in asthenia, and relate them to androgen decline (5– 8 ).
young (Pearson r ⴝ 0.87) and aging (r ⴝ 0.89) men, but Measuring total testosterone to identify hypogonadism
the BTcal:BTmeas ratio differed markedly between the 2 in older men is inappropriate, as stated in the official
groups (2.28 vs 3.48; P <0.001). recommendations of the International Society for the
Conclusions: In men, there is an age-associated discrep- Study of Aging Males. Indeed, according to the free
ancy between calculated and measured BT concentra- hormone hypothesis, the active form of testosterone avail-
tions. We suggest some hypotheses for the discrepancy, able for tissues is the fraction of circulating testosterone
but additional studies will be performed to finally that is not bound to proteins (9 ). Equilibrium dialysis (10 )
elucidate this difference in results and to determine the and ultrafiltracentrifugation (11 ) are reference methods
most appropriate method for BT measurements in older for measuring free testosterone; however, they are tech-
men. nically unsuitable for routine assay. It has been proposed,
© 2007 American Association for Clinical Chemistry but not experimentally demonstrated, that the concentra-
tion of testosterone not tightly bound to sex hormone–
binding globulin (SHBG)7 could be considered the “bio-
1
available” fraction of circulating testosterone (12 ). In
INSERM ERM 0322 Hôpital Debrousse, Lyon, France.
2
Université de Lyon, Université Lyon 1, ISPB, Lyon, France,
older men, the decrease in bioavailable testosterone (BT)
3
Service de Radioanalyse, Centre de Médecine Nucléaire, and 6 Fédération concentrations is more marked than the decrease in total
d’Endocrinologie and Hôpital Neurologique et Cardiologique (Hospices Civils testosterone (T) concentrations because of the concomi-
de Lyon), Bron, France. tant increase of SHBG concentrations with age (3, 4 ). A
4
International Agency for Research on Cancer, Lyon, France.
5
Département de Médecine de la Reproduction, Hôpital Edouard, Herriot,
and InSERM-INRA U418, Hôpital Debrousse, Lyon, France.
* Address correspondence to this author at: Service de Radioanalyse,
7
Centre de Médecine Nucléaire, Hôpital Neurologique et Cardiologique, 59 Nonstandard abbreviations: SHBG, sex hormone– binding globulin; BT,
boul. Pinel, Bron 69394, France. E-mail henri.dechaud@chu-lyon.fr. bioavailable testosterone; T, testosterone; DHEA, dehydroepiandrosterone;
Received July 28, 2006; accepted January 23, 2007. DHEAS, dehydroepiandrosterone sulfate; CBG, corticosteroid-binding globu-
Previously published online at DOI: 10.1373/clinchem.2006.077362 lin.
723
724 Déchaud et al.: Bioavailable Testosterone and Aging Men
reliable measurement of the concentration of BT and the 3.7%, 4.3%, and 3.7% for concentrations of 0.17, 1.0, 2.6,
establishment of an interval of values under which an 6.9, 15.6, and 26.0 nmol/L, respectively, with interbatch
individual could be considered hypogonadal are neces- CVs of 7.1%, 7.9%, and 7.1% for T concentrations of 1.8,
sary for an ultimate prescription of hormonal therapies. 3.5, and 21.3 nmol/L, respectively. This method has been
BT can be directly measured in a serum sample by validated by gas chromatography/mass spectrometry
cautious ammonium sulfate precipitation of SHBG-linked measurements (23 ).
testosterone. Methods based on the preincubation of We measured SHBG with a commercially available
plasma samples with tritiated T measure the percentage immunoradiometric assay (SHBG-IRMA; Cis-Bio Interna-
of BT, which can then be multiplied by the total concen- tional), with mean intra- and interbatch CVs ⬍6%.
tration of T to calculate the absolute concentration of BT We measured BT as described (15 ). In brief, we treated
(13, 14 ). Alternatively, BT can be directly measured in the plasma samples with ammonium sulfate (50% of satura-
supernatant fraction after ammonium sulfate precipitation) at 4 °C, and after centrifugation, we measured tes-
tion by use of a highly specific RIA for testosterone (15 ). tosterone in supernatant by RIA after organic extraction
Ammonium sulfate precipitation has been used routinely and diatomaceous earth (Celite) purification. Interassay
in many reference laboratories, although technical con- CVs were 11.7%, 8.7%, 8.9%, 7.0%, 8.7%, and 9.4% for BT
cerns have limited its universal application. To overcome concentrations of 0.13, 0.90, 1.6, 4.8, 7.8, and 14.4 nmol/L,
this difficulty, it has been proposed that BT concentrations respectively.
be calculated by measuring total testosterone and SHBG We measured DHEA and DHEAS with RIA using
3
and using the values in a law mass equation model H-DHEA and 3H-DHEAS as radioactive markers. We
(16, 17 ). Many laboratories, confronted with increasing separated free and bound antigen by use of dextran-
demand for BT measurements, are currently using the coated charcoal. We performed DHEAS measurements
simplified mass law equation published by Vermeulen et directly on diluted plasma, whereas we measured DHEA
al. (17 ). This calculation assumes that T and SHBG by RIA after liquid extraction and diatomaceous earth
concentrations are the main covariant of BT and assumes (Celite) chromatography.
that albumin concentrations and affinity constants of
SHBG and albumin for testosterone can be predefined. theoretical calculations
Different algorithms for the determinations of BT have The following equation was used to calculate BT, as
been published and the results compared (18 –20 ), show- suggested by Vermeulen et al. (17 ):
ing large differences between the results of different
fT ⫽ (T ⫺ [N ⫻ fT])/Ks(SHBG ⫺ T ⫹ [N ⫻ fT])
algorithms (21 ).
In this study, to identify age-dependent discrepancies from which we obtained
in BT and potential relevant mechanisms, we compared
BT ⫽ fT ⫹ AT
the concentrations of BT that were measured by use of a
routine ammonium sulfate precipitation assay (15 ) to where T ⫽ molar concentration of total T, fT ⫽ molar
calculated BT using the Vermeulen mass action law model concentration of free T, BT ⫽ molar concentration of BT,
(17 ) in 2 populations of young and aging men. SHBG ⫽ molar concentration of SHBG, Ka ⫽ affinity
constant of albumin for T (⫽ 3.6 ⫻ 104 L/mol), Ks ⫽
Materials and Methods affinity constant of SHBG for T (1.0 ⫻ 109 L/mol), N ⫽ Ka
study participants ⫻ Ca ⫹ 1 (where Ca ⫽ albumin concentration), and AT ⫽
We recruited a population of 694 men 14 to 49 years old molar concentration of albumin-bound T (⬇ Ka ⫻ Ca ⫻
and 51 older men, 50 to 81 years old, who were referred to fT).
the Department of Reproductive Medicine (Hospices Concentrations of BT obtained by use of this calcula-
Civils de Lyon, France) for infertility or endocrinology tion were compared to those obtained by the Vermeulen
aging evaluation between 1990 and 2004. For each patient, equation (available at http://www.issam.ch/freetesto.htm).
concentrations of total testosterone, SHBG, and BT were
measured in plasma from the blood samples collected sensitivity analysis
during their visit to the hospital. We performed a first comparison between calculated BT
In a complementary study, T, SHBG, BT, dehydroepi- (BTcal) and measured BT (BTmeas) by comparing BTcal vs
androsterone (DHEA), and dehydroepiandrosterone sul- BTmeas values of 745 men while varying the respective
fate (DHEAS) were measured in a single run in plasma affinity constants of SHBG and albumin for binding T
samples from 14 patients younger than 35 and 13 patients (from 0.6 to 2.0 ⫻ 109 L/mol for SHBG and from 0.5 to
older than 60. 5.0 ⫻ 104 L/mol for albumin) in the theoretical
calculations.
laboratory measurements
We measured total T with an in-house RIA after organic statistics
extraction and diatomaceous earth (Celite) chromatogra- We performed t tests on mean concentrations and calcu-
phy (22 ). Mean intrabatch CVs were 4.4%, 3.3%, 2.2%, lations of Pearson correlation coefficients between BTcal
and BTmeas with Microsoft Excel. P values ⬍0.05 were Discussion

considered statistically significant. This study shows that measuring BT or calculating it by
use of mass law equations achieves different results.
Results Using the ammonium sulfate precipitation assay (15 ), we
The mean concentrations of T, SHBG, BTmeas, and BTcal found that BTmeas and BTcal concentrations had very high
and mean ratios of BTcal:BTmeas for the 745 men who were Pearson correlation coefficients. As reported in other
part of the study are presented in Table 1. Concentrations studies (17, 18, 24 ), however, BTcal gives values that are
of BT obtained by theoretical calculations were substan- generally twice as high as those of BTmeas, and these are
tially higher than those obtained by our ammonium even higher in young men compared with older men. In
sulfate precipitation assay, with a BTcal:BTmeas ratio ⬎2. our study of 745 patients, men ⬎50 years old had BTcal:
The difference in concentrations of BTcal between younger BTmeas ratios 2-fold higher than younger men (3.48 vs
(age ⬍50 years) and older (age ⬎50 years) men was of 2.28; P ⬍0.001). We also observed this difference in our
borderline significance (P ⫽ 0.08) because of a very mild complementary study on 27 men, in which older men had
nonsignificant decrease in T concentrations and a signifi- ratios of BTcal:BTmeas 2-fold higher than younger men. The
cant increase in SHBG concentrations. In contrast, the results of our complementary study, although prelimi-
concentrations of BTmeas were much higher in men ⬍50 nary, are of interest because we measured samples from
years old than in men ⬎50, and the difference was highly both young and old men within the same analytical batch,
significant (P ⬍0.001). Interestingly, the BTcal:BTmeas ratio underlining the fact that the differences observed in
decreased significantly (P ⬍0.001) from 3.48 for men ⬎50 absolute concentrations between BTcal and BTmeas in the 2
to 2.28 for men ⬍50 years old. Pearson correlation coeffi- different populations are real and not an artifact of
measurement imprecision or a change in the methodology
cients, comparing BTcal vs BTmeas in men ⬍50 (r ⫽ 0.87)
over time.
and ⬎50 (r ⫽ 0.89), were highly significant (P ⬍0.001).
Some hypotheses may explain the discrepancy in ab-
However, the slope of the correlations between BTcal and
solute concentrations between BTcal and BTmeas. Limita-
BTmeas in the 2 populations was different (y ⫽ 1.49x ⫹ 2.13
tions may exist in the accuracy of measuring BT by serum
in men ⬍50 and y ⫽ 1.94x ⫹ 2.24 in men ⬎50 years old;
ammonium sulfate precipitation. Indeed, the association
Fig. 1).
constant of SHBG for T is slightly higher than at 37 °C
The consequence of varying values of the binding
(25 ), and because the precipitation of the SHBG-bound T
affinity constants of albumin (Ka) and SHBG (Ks) for
complex by ammonium sulfate was performed at 4 °C,
testosterone, within the interval of values published in the this may cause a decrease in the fraction of BT. Moreover,
literature, is shown in Fig. 2. When Ks and Ka were 1.6 ⫻ during this step, a fraction of albumin can be precipitated
109 and 1.0 ⫻ 104 L/mol, respectively, the concentrations by ammonium sulfate, which might decrease the concen-
of BTcal and BTmeas were virtually identical and BTcal: tration of BT. Because the testosterone binding to cortico-
BTmeas was close to 1. steroid-binding globulin (CBG) is not negligible, as em-
Mean testosterone, SHBG, DHEA, and DHEAS concen- phasized by Nisula et al. (26 ), the small part of CBG-
trations in 14 men ⬍35 years old and 13 men ⬎60 years bound T that is precipitated by ammonium sulfate might
old are given in Table 2. In older men, mean testosterone also contribute to the decrease in the final result of
concentrations were significantly lower and mean SHBG measured BT. During the precipitation step, T could
concentrations significantly higher than in younger men. partially dissociate from albumin, and get fixed on the
In addition, older men had much lower mean DHEA and precipitate, contributing as well to an overall decrease in
DHEAS concentrations than younger men. As expected, BTmeas. These methodological deficiencies have been care-
BTcal and BTmeas were higher in younger than in older fully checked previously (15 ), however, and are unlikely
men. The BTcal:BTmeas ratio in older men was twice as to explain a 2-fold difference between measured and
high as in younger men (4.70 vs 2.09; P ⬍0.001). In calculated BT.
addition, BTcal:BTmeas was inversely correlated with the When BT is calculated from total testosterone and
concentration of DHEA (Pearson r ⫽ ⫺0.503) and DHEAS SHBG concentrations, several assumptions are made. The
(r ⫽ ⫺0.626). binding affinity constants for testosterone used by Ver-
Table 1. Mean (SD) concentrations of total T, SHBG, BTmeas, BTcal, and BTcal:BTmeas according to age in men.a
n T, nmol/L SHBG, nmol/L BTmeas, nmol/L BTcal, nmol/L BTcal:BTmeas
All patients 745 14.48 (6.92) 29.28 (14.52) 3.60 (2.06) 7.60 (3.55) 2.37 (0.92)
Men ⬍50 years old 694 14.52 (6.82) 29.00 (14.25) 3.70 (2.04) 7.67 (3.51) 2.28 (0.72)
Men ⱖ50 years old 51 14.00 (8.23) 33.72 (17.35) 2.27 (1.86) 6.65 (4.02) 3.48 (1.99)
P 0.61 0.02 ⬍0.001 0.08 ⬍0.001
a
P values compare men ⬍50 with men ⱖ50 years old.
Fig. 1. Correlation between calculated

and measured BT in the study
population.
meulen et al. (17 ) were 1.0 ⫻ 109 L/mol for SHBG (Ks) and Our study shows that BTcal:BTmeas had a tendency to
3.6 ⫻ 104 L/mol for albumin (Ka), although a broad range increase and eventually double when young and older
of affinity constants have been reported (18 ). Calculating men were compared. This discrepancy is of clinical rele-
BT using higher Ks values and lower Ka values reconciled vance, because with similar concentrations of total T and
BTcal vs BTmeas, with a BTcal:BTmeas ratio that tends to 1. SHBG, BTmeas is lower in old men than in young men,
Our calculation shows that BTcal:BTmeas is highly sensitive whereas BTcal is similar. This difference in BTcal:BTmeas
to the effect of protein binding affinity constants for T. between young and old men in our study population was
Although exact values of Ks and Ka are difficult to observed when BTcal was calculated by using the algo-
establish, we might speculate that the Ks and Ka used in rithm by Vermeulen as well as when using the algorithm
the equation of Vermeulen tend to overestimate the true by Morris et al. (21 ), with a significant (P ⬍0.001) decrease
value of BT. Indeed, different theoretical calculations can in BTcal:BTmeas from 2.38 for men older than 50 to 1.44 for
give different results depending on the algorithm used men younger than 50 years. Although an increase in Ks
(21 ). For example, when using the algorithm by Morris et with age has been suggested (27 ) and some mutations in
al. (19 ), the mean BTcal on the study population was 4.58 SHBG have been reported (28 ), they are unlikely to
nmol/L, whereas it was 3.60 for BTmeas and 7.60 when account for such a large discrepancy.
BTcal was estimated by using the algorithm by Vermeulen Because we did not measure albumin concentration in
et al. (17 ) (results not shown). our patients, we cannot exclude the possibility that a
Fig. 2. Sensitivity analyses of BTcal:

BTmeas while changing Ka and Ks be-
tween 0.5 and 5 ⫻ 104 and 0.6 and
2.0 ⫻ 109 L/mol, respectively.
Table 2. Mean (SD) concentrations of T, SHBG, BTmeas, BTcal, BTcal:BTmeas, DHEA, and DHEAS in 27 men.
n T, nmol/L SHBG, nmol/L BTmeas, nmol/L BTcal, nmol/L BTcal:BTmeas DHEA, nmol/L DHEAS, ␮mol/L
Men ⬍35 years old 14 17.02 (5.66) 30.58 (13.11) 4.26 (0.82) 8.62 (1.65) 2.09 (0.58) 19.96 (8.58) 9.55 (3.65)
Men ⱖ60 years old 13 12.66 (5.98) 49.62 (15.06) 1.07 (0.47) 4.55 (1.96) 4.70 (1.75) 4.67 (2.05) 2.40 (1.72)
P (t-test) 0.63 1.88 ⫻ 10⫺3 4.36 ⫻ 10⫺12 5.40 ⫻ 10⫺6 1.73 ⫻ 10⫺5 1.05 ⫻ 10⫺5 1.01 ⫻ 10⫺6
decrease in albumin concentration in older men may This work was supported by a grant from Hospices Civils
contribute to lower BT, although Vermeulen (29 ) recently de Lyon. We thank G. Galland, M.P. Raibaud, and A.
reported that for a decrease in albumin concentrations Diogon for technical assistance and V. Raverot, K. Ho-
from 43 to 35 g/L, the BTcal would diminish by only 10%. geveen, and I. McGill for text revision.
Interestingly, Cooke et al. (30 ) reported some discrep-
ancy between BTcal and BTmeas concentrations according References
to cortisol concentrations and suggested that lipids or 1. Kaufman JM, Vermeulen A. The decline of androgen levels in
other molecules could interact with albumin to decrease elderly men and its clinical and therapeutic implications. Endocr
the number, affinity, or disposal of albumin steroid- Rev 2005;26:833–76.
binding sites. In this hypothesis, testosterone would be 2. Labrie F, Belanger A, Cusan L, Gomez JL, Candas B. Marked
decline in serum concentrations of adrenal C19 sex steroid
displaced from CBG to SHBG, inducing an increase of the precursors and conjugated androgen metabolites during aging.
ratio BTcal:BTmeas. J Clin Endocrinol Metab 1997;82:2396 – 402.
An increase in various T-binding proteins of weak 3. Harman SM, Metter EJ, Tobin JD, Pearson J, Blackman MR.
affinity constant (CBG, orosomucoid), because they may Longitudinal effects of aging on serum total and free testosterone
be partially precipitated by ammonium sulfate, might also levels in healthy men: Baltimore Longitudinal Study of Aging. J Clin
contribute to discordant BTcal and BTmeas in older indi- Endocrinol Metab 2001;86:724 –31.
viduals and therefore contribute to the decrease of the 4. Feldman HA, Longcope C, Derby CA, Johannes CB, Araujo AB,
BTmeas concentrations. Coviello AD, et al. Age trends in the level of serum testosterone
and other hormones in middle-aged men: longitudinal results from
It has also been suggested that an important cause for the Massachusetts male aging study. J Clin Endocrinol Metab
the decline in BTmeas vs BTcal concentration in aging men 2002;87:589 –98.
is the lowering of adrenal androgens that bind signifi- 5. Kratzik CW, Schatzl G, Lunglmayr G, Rucklinger E, Huber J. The
cantly to SHBG. Because the affinity constants of SHBG impact of age, body mass index and testosterone on erectile
for DHEA and ⌬5-androstenediol are not negligible (66 ⫻ dysfunction. J Urol 2005;174:240 –3.
106 and 1.5 ⫻ 109 L/mol, respectively), it has been 6. Muller M, Aleman A, Grobbee DE, de Haan EH, van der Schouw YT.
predicted that more than 10% of the SHBG binding sites Endogenous sex hormone levels and cognitive function in aging
are occupied by these adrenal androgens in young men men: is there an optimal level? Neurology 2005;64:866 –71.
7. Tsujimura A, Matsumiya K, Matsuoka Y, Takahashi T, Koga M,
(31 ). In agreement with previous studies, we observed a
Iwasa A, et al. Bioavailable testosterone with age and erectile
highly significant decline in DHEA/DHEAS in older men dysfunction. J Urol 2003;170:2345–7.
(1– 4 ). Therefore, in older men, some of these sites on 8. Yaffe K, Lui LY, Zmuda J, Cauley J. Sex hormones and cognitive
SHBG may be available for binding T, consequently function in older men. J Am Geriatr Soc 2002;50:707–12.
decreasing the concentration of BTmeas, whereas the con- 9. Ekins R. The free hormone hypothesis and measurement of free
centration of BTcal would remain unchanged. In our hormones. Clin Chem 1992;38:1289 –93.
study, by including DHEA in the mass action equation, 10. Forest MG, Rivarola MA, Migeon CJ. Percentage binding of testos-
we found that BTcal increased with increasing concentra- terone, androstenedione and dehydroisoandrosterone in human
tions of DHEA. This influence of adrenal androgen, plasma. Steroids 1968;12:323– 43.
however, was observed for ⌬5-androstenediol using con- 11. Hammond GL, Nisker JA, Jones LA, Siiteri PK. Estimation of the
percentage of free steroid in undiluted serum by centrifugal
centrations reported in the literature (data not shown). ultrafiltration-dialysis. J Biol Chem 1980;255:5023– 6.
This influence of adrenal androgens is further supported 12. Pardridge WM. Serum bioavailability of sex steroid hormones. Clin
by the results of our complementary analysis in which the Endocrinol Metab 1986;15:259 –78.
ratio BTcal:BTmeas was inversely correlated with the con- 13. Tremblay RR, Dube JY. Plasma concentrations of free and non-
centration of DHEA. TeBG bound testosterone in women on oral contraceptives.
Contraception 1974;10:599 – 605.
In summary, there is substantial discordance between 14. Loric S, Guechot J, Duron F, Aubert P, Giboudeau J. Determination
absolute concentrations of BTmeas and BTcal, suggesting of testosterone in serum not bound by sex-hormone-binding
globulin: diagnostic value in hirsute women. Clin Chem 1988;34:
that a simplified law of mass action cannot predict the
1826 –9.
variations in steroid hormone distribution in serum. Fur-
15. Dechaud H, Lejeune H, Garoscio-Cholet M, Mallein R, Pugeat M.
ther investigation should explore their physiological rel- Radioimmunoassay of testosterone not bound to sex-steroid-
evance in aging men and contribute to a consensual binding protein in plasma. Clin Chem 1989;35:1609 –14.
approach for BT measurement as a tool for therapeutic 16. Sodergard R, Backstrom T, Shanbhag V, Carstensen H. Calcula-
decisions. tion of free and bound fractions of testosterone and estradiol-17
beta to human plasma proteins at body temperature. J Steroid [3,4 –13C]testosterone as an internal standard. J Chromatogr
Biochem 1982;16:801–10. 1985;339:233– 42.
17. Vermeulen A, Verdonck L, Kaufman JM. A critical evaluation of 24. Emadi-Konjin P, Bain J, Bromberg IL. Evaluation of an algorithm for
simple methods for the estimation of free testosterone in serum. calculation of serum “bioavailable” testosterone (BAT). Clin Bio-
J Clin Endocrinol Metab 1999;84:3666 –72. chem 2003;36:591– 6.
18. Giton F, Fiet J, Guechot J, Ibrahim F, Bronsard F, Chopin D, 25. Lutz RA, Lutz-Ewan L, Weder HG. Further studies on the temper-
Raynaud JP. Serum bioavailable testosterone: assayed or calcu- ature dependence of the binding of testosterone to human
lated? Clin Chem 2006;52:474 – 81. pregnancy plasma proteins. Steroids 1973;21:423–31.
26. Nisula BC, Loriaux DL, Wilson YA. Solid phase method for
19. Morris PD, Malkin CJ, Channer KS, Jones TH. A mathematical
measurement of the binding capacity of testosterone-estradiol
comparison of techniques to predict biologically available testos-
binding globulin in human serum. Steroids 1978;31:681–90.
terone in a cohort of 1072 men. Eur J Endocrinol 2004;151:
27. Morley JE, Perry HM, III. Androgen treatment of male hypogonad-
241–9.
ism in older males. J Steroid Biochem Mol Biol 2003;85:367–73.
20. Ly LP, Handelsman DJ. Empirical estimation of free testosterone
28. Hogeveen KN, Cousin P, Pugeat M, Dewailly D, Soudan B,
from testosterone and sex hormone-binding globulin immunoas-
Hammond GL. Human sex hormone-binding globulin variants
says. Eur J Endocrinol 2005;152:471– 8.
associated with hyperandrogenism and ovarian dysfunction. J Clin
21. de Ronde W, van der Schouw YT, Pols HA, Gooren LJ, Muller M, Invest 2002;109:973– 81.
Grobbee DE, et al. Calculation of bioavailable and free testoster- 29. Vermeulen A. Reflections concerning biochemical parameters of
one in men: a comparison of 5 published algorithms. Clin Chem androgenicity. Aging Male 2004;7:280 –9.
2006;52:1777– 84. 30. Cooke RR, McIntosh JE, Murray-McIntosh RP. Effect of cortisol on
22. Rinaldi S, Dechaud H, Biessy C, Morin-Raverot V, Toniolo P, percentage of non-sex-hormone-bound steroid: implications for
Zeleniuch-Jacquotte A, et al. Reliability and validity of commer- distribution of steroids on binding proteins in serum. Clin Chem
cially available, direct radioimmunoassays for measurement of 1996;42:249 –54.
blood androgens and estrogens in postmenopausal women. Can- 31. Dunn JF, Nisula BC, Rodbard D. Transport of steroid hormones:
cer Epidemiol Biomarkers Prev 2001;10:757– 65. binding of 21 endogenous steroids to both testosterone-binding
23. Sabot JF, Deruaz D, Dechaud H, Bernard P, Pinatel H. Determina- globulin and corticosteroid-binding globulin in human plasma.
tion of plasma testosterone by mass fragmentography using J Clin Endocrinol Metab 1981;53:58 – 68.
729 –734 (2007) Endocrinology and
Metabolism
Spectrophotometric Assay for Complex I of the

Respiratory Chain in Tissue Samples and
Cultured Fibroblasts
Antoon J.M. Janssen, Frans J.M. Trijbels, Rob C.A. Sengers,† Jan A.M. Smeitink,
Lambert P. van den Heuvel, Liesbeth T.M. Wintjes,
Berendien J.M. Stoltenborg-Hogenkamp, and Richard J.T. Rodenburg*
Background: A reliable and sensitive complex I assay is cies, the new assay detected the complex I deficiencies
an essential tool for the diagnosis of mitochondrial in both muscle and fibroblasts.
disorders, but current spectrophotometric assays suffer Conclusions: This spectrophotometric assay is repro-
from low sensitivity, low specificity, or both. This ducible, sensitive, and specific for complex I activity
deficiency is mainly due to the poor solubility of coen- because of its high rotenone sensitivity, and it can be
zyme-Q analogs and reaction mixture turbidity caused applied successfully to the diagnosis of complex I
by the relatively high concentrations of tissue extract deficiencies.
that are often required to measure complex I. © 2007 American Association for Clinical Chemistry
Methods: We developed a new spectrophotometric as-
say to measure complex I in mitochondrial fractions and Complex I (NADH:ubiquinone oxidoreductase, EC
applied it to muscle and cultured fibroblasts. The 1.6.5.3) is the first complex of the oxidative phosphoryla-
method is based on measuring 2,6-dichloroindophenol tion system. It is the entry point for electrons into the
reduction by electrons accepted from decylubiquinol, respiratory chain by oxidation of NADH and transport of
reduced after oxidation of NADH by complex I. The electrons to coenzyme-Q10. Complex I also has proton-
assay thus is designed to avoid nonspecific NADH transporting activity over the inner mitochondrial mem-
oxidation because electrons produced in these reactions brane to the intermembrane space. With a relative molec-
are not accepted by decylubiquinone, resulting in high ular mass of ⬃980 000, it is the largest complex of the
rotenone sensitivity. respiratory chain. Complex I consists of 45 subunits
Results: The assay was linear with time and amount of (identified so far), forming a characteristic L-shaped con-
mitochondria. The Km values for NADH and 2,6- figuration (1 ). The hydrophilic peripheral arm stretches
dichloroindophenol in muscle mitochondria were 0.04 out into the mitochondrial matrix and catalyzes the
and 0.017 mmol/L, respectively. The highest complex I NADH oxidation and electron transport. The hydropho-
activities were measured with 0.07 mmol/L decylubiqui- bic membrane arm is embedded in the inner mitochon-
none and 3.5 g/L bovine serum albumin. The latter was drial membrane and contains the proton-transport activ-
an essential component of the reaction mixture, increas- ity. A deficiency of complex I is probably the most
ing the solubility of decylubiquinone and rotenone. In frequently encountered cause of mitochondrial disease,
patients with previously diagnosed complex I deficien- and mutations in several nuclear DNA– encoded and
mitochondrial DNA– encoded subunits have been de-
scribed to date (2 ). In addition, mutations in mitochon-
drial tRNAs, such as the m.3243A⬎G mutation in the
Department of Pediatrics and Laboratory of Pediatrics and Neurology, The mitochondrial tRNALEU(UUR), usually result in complex I
Nijmegen Centre for Mitochondrial Disorders at the Radboud University
Medical Centre, Nijmegen, The Netherlands.
deficiency (2 ). The most commonly used technique for
† Dr. Sengers died on April 6, 2006. measuring complex I is a spectrophotometric assay mea-
* Address correspondence to this author at: Laboratory of Pediatrics and suring rotenone-sensitive NADH oxidation at 340 nm in
Neurology, UMC St. Radboud, Geert Grooteplein zuid 10, 6525 GA Nijmegen,
tissue homogenate or mitochondria-enriched fractions
The Netherlands. Fax 31-24-3618900; e-mail R.Rodenburg@cukz.umcn.nl.
Received August 28, 2006; accepted January 18, 2007. from cultured fibroblasts (3, 4 ); however, the sensitivity
Previously published online at DOI: 10.1373/clinchem.2006.078873 and specificity of these assays are not optimal. One reason
729
730 Janssen et al.: Spectrophotometric Analysis of Complex I
for this deficiency is the poor solubility of coenzyme-Q complex i assay

analogs. In addition, owing to the low sensitivity, rela- In the new complex I assay, DCIP is used as a terminal
tively high concentrations of tissue extract are often electron acceptor. Complex I oxidizes NADH, and the
required to detect complex I, resulting in turbidity of electrons produced reduce the artificial substrate decylu-
reaction mixtures. Here we describe a new sensitive and biquinone that subsequently delivers the electrons to
specific assay for complex I that is suitable for diagnostic DCIP. The reduction of DCIP can be followed spectropho-
purposes. tometrically at 600 nm. As the electrons produced by
other NADH-dehydrogenases are not accepted by decy-
Materials and Methods lubiquinone (3 ), reduction of DCIP is almost completely
materials caused by complex I activity, resulting in very high
Coenzyme Q1, decylubiquinone, 2,6-dichloroindophenol rotenone-sensitive activity.
(DCIP),1 rotenone, and antimycin-A were obtained from We measured complex I spectrophotometrically at 600
Sigma. NADH and bovine serum albumin (BSA; fraction nm in an incubation volume of 1.0 mL containing 25
V, fatty acid free) were obtained from Roche. All other mmol/L potassium phosphate, 3.5 g/L BSA, 60 ␮mol/L
chemicals were of the highest purity commercially avail- DCIP, 70 ␮mol/L decylubiquinone, 1.0 ␮mol/L antimy-
cine-A, and 0.2 mmol/L NADH, pH 7.8. Decylubiquinone
able. Spectrophotometric assays were performed on a
and antimycine-A were dissolved in dimethyl sulfoxide
Perkin-Elmer Lambda 25 UV/Vis spectrophotometer
(17.5 mmol/L and 1.0 mmol/L, respectively). We pre-
with an automatic water thermostattable 8-cell holder in
pared a stock solution of 80 g/L BSA in 5 mmol/L
disposable semimicro 10-mm acryl-cuvettes from Sarstedt.
potassium phosphate buffer, pH 7.4. Because BSA is a
critical component of the complex I assay, we measured
preparation of muscle-mitochondria and
the concentration spectrophotometrically at 280 nm (A280
mitochondria-enriched fractions from
1 g/L BSA ⫽ 0.667). The stock solution was diluted to 70
cultured skin fibroblasts g/L and stored in 1-mL aliquots at ⫺30 °C. Of this
Human muscle tissue (musculus quadriceps or musculus solution, 50 ␮L was added to a final reaction volume of 1
semitendinosus) was homogenized as described (5 ). Frac- mL. We preincubated an aliquot of 2.5 to 10 ␮L mitochon-
tions of 200 to 300 ␮L of a frozen supernatant of muscle drial suspension from muscle or 20 ␮L mitochondria-
homogenate (centrifuged at 600g and 2 °C for 10 min) enriched fraction from fibroblasts at 37 °C in 960 ␮L
were thawed at 2– 4 °C and centrifuged (10 min at 14 000g incubation mixture without NADH. After 3 min, we
and 2 °C) in an Eppendorf 5402 centrifuge. We carefully added 20 ␮L of 10 mmol/L NADH and measured the
removed the 14 000g supernatant and the fluffy layer with absorbance at 30-s intervals for 4 min at 37 °C. After 4
an Eppendorf pipettor. The mitochondrial pellet was min, we added 1.0 ␮L rotenone (1 mmol/L in dimethyl
resuspended in 150 ␮L of 10 mmol/L Tris, pH 7.6, frozen sulfoxide) and measured the absorbance again at 30-s
in 25-␮L aliquots in liquid nitrogen, and kept at ⫺80 °C. intervals for 4 min.
We cultured human skin fibroblasts in M199 medium Complex I activity was expressed as mU/U complex II,
(Gibco) supplemented with 20% (vol/vol) fetal calf se- mU/U complex IV, or U/g protein, in which 1 U complex
rum. Aliquots of 10 to 15 ⫻ 106 cells were washed with I activity equals 1 ␮mol DCIP reduced per min. Fibro-
ice-cold phosphate-buffered saline, frozen in liquid nitro- blasts from 6 patients and muscle tissue samples from 3
gen, and kept at ⫺80 °C until use. For isolation of mito- patients with a diagnosed complex I deficiency were used
chondria-enriched fractions, the pellets were thawed at to measure complex I to demonstrate the applicability of
2– 4 °C and suspended in 2.9 mL ice-cold 10 mmol/L Tris, the assay to the diagnosis of complex I deficiency. The
pH 7.6. We disrupted the cells mechanically with a 5-mL controls and patients have been described (5 ).
glass/Teflon Potter-Elvehjem homogenizer (clearance,
0.025 mm), 8 strokes at 1800 rpm in melting ice. After complex ii assay
homogenization, we added 0.6 mL ice-cold 1.5 mol/L During isolation of mitochondria from frozen samples of
sucrose and centrifuged the homogenate (10 min at 600g muscle or fibroblasts, part of the citrate synthase leaks out
and 2 °C) in a Sorval-RC2B centrifuge. The 600g superna- of the mitochondria and is lost, so citrate synthase cannot
tant was centrifuged again (10 min at 14 000g and 2 °C), be used as a mitochondrial marker enzyme. For that
and the resulting supernatant was carefully removed. The reason, we used complex II and complex IV as mitochon-
mitochondrial pellet was resuspended in 0.5 mL of 10 drial marker enzymes. We measured complex II spectro-
mmol/L Tris, pH 7.6, frozen in 50-␮L aliquots in liquid photometrically at 600 nm as described, with some mod-
nitrogen, and kept at ⫺80 °C. ifications (6 ). The 1.0-mL incubation volume contained 80
mmol/L potassium phosphate, 1 g/L BSA, 2 mmol/L
EDTA, 0.2 mmol/L ATP, 10 mmol/L succinate, 0.3
mmol/L potassium cyanide (KCN), 80 ␮mol/L DCIP, 50
1
Nonstandard abbreviations: DCIP, 2,6-dichloroindophenol; BSA, bovine ␮mol/L decylubiquinone, 1 ␮mol/L antimycine-A, and 3
serum albumin; KCN, potassium cyanide. ␮mol/L rotenone, pH 7.8. We preincubated an aliquot of
5 ␮L mitochondrial suspension from muscle or 10 ␮L

mitochondria-enriched fraction from fibroblasts at 37 °C
in the incubation mixture without KCN and succinate.
After 10 min, we added KCN and succinate to start the
reaction and measured the absorbance at 1-min intervals
for 5 min at 37 °C. Blanks were measured in the presence
of 5 mmol/L malonate that was added before preincuba-
tion. Decylubiquinone (10 mmol/L) was dissolved in
dimethyl sulfoxide. Antimycine-A (1 mmol/L) and rote-
none (3 mmol/L) were dissolved in ethanol. For both
assays, we used a molar absorptivity at 600 nm of 19.1
(mmol/L)⫺1 cm⫺1 for DCIP.
complex iv and protein assays

We measured complex IV activity as described (7 ) and
protein according to Lowry et al. (8 ).
Results
Lineweaver–Burk plots revealed Km values of 0.04
mmol/L for NADH and 0.017 mmol/L for DCIP (data not
shown). The pH optimum was 7.8 (data not shown). In a
concentration series experiment with decylubiquinone,
the highest complex I activity was at 0.07 mmol/L decy-
lubiquinone (data not shown). We tested coenzyme Q1 as
an alternative ubiquinone analog and found that in the
presence of 0.07 mmol/L, the activity was ⬃80% of that
measured with decylubiquinone, with similar rotenone
sensitivities. Therefore, we continued with decylubiqui-
none as electron acceptor for complex I in the reaction
mixture. Although in most complex I assays Mg2⫹ and
KCN are added, the latter to inhibit nonspecific NADH
dehydrogenase activity (3 ), we observed no influence of
Mg2⫹ and KCN in our assay (data not shown).
In most complex I assays, pretreatment of the mito-
chondria, such as repeated freezing and thawing or son-
ication, is necessary to disrupt the mitochondrial mem-
brane. In our assay, a simple osmotic shock in 10 mmol/L Fig. 1. (A), effect of BSA on relative rotenone-sensitive complex I
Tris 䡠 Cl, pH 7.6, followed by a single freeze-thaw cycle, activity in muscle mitochondrial suspensions.
was sufficient to measure optimal complex I activity. Complex I activity measured in the absence of BSA was set to 100%. Mitochon-
drial protein content in each incubation was 0.3 g/L. Data are the mean values
Repeated freezing and thawing did not improve this of 2 experiments performed with different mitochondrial preparations, and all
result, and sonication even decreased complex I activity incubations were performed in duplicate. (B), effect of BSA on the percentage of
rotenone-sensitive activity measured in the complex I assay. Complete inhibition
(data not shown). by rotenone corresponds to 100% rotenone sensitivity. The percentages calcu-
Complex I activity was linear for at least 4 min, with lated were derived from the incubations described in (A).
sample amounts varying between 0.25 ␮g protein (con-
taining 0.9 mU complex IV) and 3 ␮g protein (11 mU placed by coenzyme Q1 in the reaction mixture (data not
complex IV; see Data Supplement that accompanies the shown).
online version of this article at http://www.clinchem.org/ The IC50 for rotenone in muscle mitochondria, which
content/vol53/issue4). could be measured only in the presence of BSA, was 13.5
Because BSA is essential for measuring optimal com- nmol/L (mean of 2 results; range, 10 –17 nmol/L; Fig. 2).
plex I activity (3 ), we studied the effect of BSA on The presence of BSA in the reaction mixture is required
complex I activity in muscle mitochondria. We measured not only for rotenone sensitivity of the assay (probably by
optimal complex I activity in the presence of BSA concen- solubilizing rotenone), but also for solubilization of decy-
trations between 3.2 and 3.9 g/L, whereas the percentage lubiquinone, as illustrated by the fact that in the absence
rotenone sensitivity of complex I was near 100% at BSA of BSA, we observed an orange/yellow layer on the
concentrations between 2.1 and 4.9 g/L (Fig. 1). Similar surface of the reaction mixture after centrifugation at
results were obtained when decylubiquinone was re- 14000g. By spectrophotometric analysis at 278 nm, we
complex I activity in these samples was 607 (47) U/L (n ⫽

10; CV 8%), 795 (30) U/L (n ⫽ 6; CV 4%), and 888 (49) U/L
(n ⫽ 6; CV 5%).
To compare this method to the method described by
Fischer et al. (3 ), we used the 2 methods to measure
complex I in mitochondria from muscle and fibroblasts.
The mean (SD) activity measured in muscle mitochondria
using the method of Fischer et al. (3 ) was 31% (5%; n ⫽ 7;
range, 26%–38%; paired t-test P ⫽ 0.0005) and in mito-
chondria from fibroblasts was 17% (3%; n ⫽ 46; range,
10%–26%; paired t-test P ⫽ 7 ⫻ 10⫺29) of the activities
measured with our method, showing that the new
method is 3-fold (for muscle) and more than 5-fold (for
fibroblasts) more sensitive than the method of Fischer et
al. (3 ).
We assessed the specificity of the complex I assay by
measuring its rotenone sensitivity. The mean (SD) rote-
Fig. 2. Inhibition of complex I activity by rotenone measured in muscle
none sensitivity in mitochondria from control muscle (n ⫽
mitochondria.
Complex I was measured in the absence (open circles) or presence (filled circles)
17) was 95% (5%), and in mitochondria from control
of 3.5 g/L BSA and with the rotenone concentration interval as indicated. fibroblasts (n ⫽ 46), it was 82% (9%).
Rotenone-sensitive complex I activity (U/L mitochondrial suspension) is shown. We tested whether the new complex I assay could be
Reaction mixtures contained 1.3 mg/L protein. Using nonlinear regression
analysis, the mean (SD) IC50 for rotenone was 13.5 (3.5) nmol/L. applied to cruder muscle preparations than those used in
the experiments described above. In 5 control muscle
found that ⬃70% of the amount of decylubiquinone samples, the mean (SD) complex I activity in 600g super-
added to the mixture was present in this layer. natants was 46% (6%; range, 43%–51%) of that in equiv-
Intraassay imprecision (CV) was between 2% and 8%, alent amounts of mitochondrial fractions, and the rote-
and interassay imprecision was between 2% and 11% none sensitivity was 71% (14%; range, 48%– 86%). Finally,
(Table 1). we tested whether the new complex I assay could be
To test the long-term reproducibility of the assay, applied to the diagnosis of complex I deficiencies in both
complex I was measured repeatedly in 3 muscle mito- muscle and fibroblasts. First, we measured control values
chondrial samples over a period of 6 months. Mean (SD) for mitochondria from muscle and fibroblasts (Table 2).
Table 1. Intra- and interassay imprecision for complex I in mitochondrial fractions from muscle and cultured fibroblasts.a
Complex I (U/L)
Intraassay imprecision Experiment 1 Experiment 2 Experiment 3 Mean SD CV, %

Muscle mitochondria
10 ␮L 240 267 262 256 14 6
10 ␮L, 1:1 dilution 248 234 276 253 21 8
10 ␮L, 1:2 dilution 252 249 241 247 6 2
Fibroblast mitochondria
20 ␮L 117 131 124 124 7 6
20 ␮L, 1:1 dilution 124 118 130 124 6 5
20 ␮L, 1:2 dilution 132 129 123 128 5 4
Interassay imprecision
Muscle mitochondria 1 2 3
10 ␮L 251 256 263 257 6 2
10 ␮L, 1:1 dilution 280 252 280 271 16 6
10 ␮L, 1:2 dilution 249 243 297 263 30 11
Fibroblast mitochondria
20 ␮L 124 131 110 122 11 9
20 ␮L, 1:1 dilution 124 134 116 125 9 7
20 ␮L, 1:2 dilution 129 147 123 133 13 9
a
Complex I activities are expressed as U/L undiluted sample. Intra- and interassay imprecision were determined with undiluted, 1:1 diluted, and 1:2 diluted
mitochondrial fractions from muscle and cultured fibroblasts. Intraassay imprecision was determined by measuring complex I activities in 3-fold on the same day, and
interassay imprecision was determined by measuring complex I activities on 3 different days. All incubations were performed in duplicate. The protein contents of the
undiluted muscle and fibroblast mitochondrial fractions were 0.26 and 0.88 g/L, respectively.
ples and cultured fibroblasts. In addition to complex I,

Table 2. Control values for complex I in mitochondrial
muscle tissue and cultured fibroblasts contain several
fractions from muscle and cultured fibroblasts.a
nonmitochondrial NADH-oxidizing dehydrogenases.
Complex I
Therefore, the use of tissue homogenates in complex I
mU/U CII mU/U C IV assays results in a relatively high rate of rotenone-insen-
Muscle sitive NADH oxidation that interferes with the sensitivity
Mean (n ⫽ 17) 1140 343 of the complex I assay. Another disadvantage of measur-
SD 180 63 ing muscle homogenate is turbidity of the incubation-
Observed range 783–1497 270–475 mixture, which interferes with the spectrophotometric
Mean (2 SD) 1140 (360) 343 (126)
assay. For these reasons, Brooks and Krähenbühl (9 )
Fibroblasts
developed a radiochemical assay for complex I in muscle
Mean (n ⫽ 46) 1161 1100
by measuring 3H2O production from [4B-3H]–NADH
SD 237 245
oxidation, based on the stereospecificity of complex I for
Observed range 720–1708 678–1675
Mean (2 SD) 1161 (474) 1100 (490)
the 4B hydrogen atom of NADH.
a
Our new assay uses no radioactivity, is suitable for the
All incubations were performed in duplicate.
diagnostic analysis of complex I in fibroblasts and muscle
tissue, and uses DCIP as a final electron acceptor. DCIP
We examined fibroblasts from 6 patients carrying varia- has a molar absorptivity that is ⬃3 times higher than that
tions in different complex I genes and suffering from a of NADH: the molar absorptivity at 600 nm of DCIP is
previously established complex I deficiency (5 ). Using the 19.1 (mmol/L)⫺1 cm⫺1, whereas the molar absorptivity at
new method, the enzyme deficiency could be confirmed 340 nm of NADH is 6.2 (mmol/L)⫺1 cm⫺1. Compared
in all 6 patients. In 3 of the patients, we measured complex with the method described by Fischer et al. (3 ) that
I in muscle and also confirmed the deficiency in this measures NADH oxidation, 3- to 5-fold more complex I
tissue. The method showed lower results in all tested activity is measured using our method.
patients with complex I deficiencies than in any control DCIP has been used by others in a complex I assay
subjects (Table 3). (10 ), but that assay was not suitable for diagnostic pur-
poses because of nonlinearity of the absorbance with time.
Discussion In our assay, the addition of an optimal concentration of
At present, the diagnosis of complex I deficiency is BSA to the reaction mixture, combined with the use of
usually established using complex I assays that are based isolated mitochondrial preparations instead of crude sam-
on the spectrophotometric measurement of rotenone- ple homogenates, resulted in a complex I activity that was
sensitive NADH oxidation in patient-derived tissue sam- linear in time and had high rotenone sensitivity. BSA is
Table 3. Complex I in muscle and fibroblasts of 6 patients with a previously established complex I deficiency.
Complex 1 by our methodb
Complex I by the method of
Patienta mU/U CII mU/U C IV Fischer et al. (3 ), mU/U CSc
Muscle
1 154 48 14
3 510 144 16
5 465 157 24
Control mean (2 SD) 1140 (360) 343 (126) 85 (40)
Observed range 783–1497 270–475 53–163
n 17 17 43
Fibroblasts
1 (NDUFS2) 255 179 29
2 (NDUFS4) 240 145 64
3 (NDUFS7) 484 459 65
5 (NDUFS7) 438 467 26
7 (NDUFV1) 426 416 85
8 (MT-ND2) 339 333 42
Control mean (2 SD) 1161 (474) 1100 (490) 188 (104)
Observed range 720–1708 678–1675 110–260
n 46 46 14
a
Patient numbering is the same as in Janssen et al. (5 ).
b
Complex I was measured in mitochondrial fractions.
c
Complex I was measured in 600g supernatants of muscle and mitochondrial fractions from fibroblasts.
essential, as it facilitates the solubilization of both rote- References

none and decylubiquinone. Rotenone and decylubiqui- 1. Carroll J, Fearnley IM, Skehel JM, Shannon RJ, Hirst J, Walker JE.
none are both hydrophobic and are practically insoluble Bovine complex I is a complex of 45 different subunits. J Biol
in water; rotenone strongly and reversibly binds to BSA Chem 2006;281:32724 –7.
(11 ). Direct binding to BSA probably also plays a role in 2. Janssen RJ, Nijtmans LG, van den Heuvel LP, Smeitink JA.
Mitochondrial complex I: structure, function and pathology. J In-
the solubilization of decylubiquinone, as it is known that
herit Metab Dis 2006;29:499 –515.
BSA reversibly binds molecules with long alkyl chains
3. Fischer JC, Ruitenbeek W, Trijbels JMF, Veerkamp JH, Stadhoud-
(12 ). ers AM, Sengers RCA, et al. Estimation of NADH oxidation in
The mean (SD) complex I activities measured in fibro- human skeletal muscle mitochondria. Clin Chim Acta 1986;155:
blasts and expressed on complex IV activity were similar 263–73.
to those measured by Kramer et al. (4 ): 1100 (245; n ⫽ 46) 4. Kramer KA, Oglesbee D, Hartman SJ, Huey J, Anderson B, Magera
vs 1200 (170; n ⫽ 15), respectively. The mean (SD) MJ, et al. Automated spectrophotometric analysis of mitochondrial
rotenone-sensitive activity in our assay was 82% (9%; n ⫽ respiratory chain complex enzyme activities in cultured skin fibro-
46) compared with 30% (range, 15%–50%) in the assay blasts. Clin Chem 2005;51:2110 – 6.
described by Kramer et al. (4 ). The mean (SD) rotenone 5. Janssen AJM, Trijbels JMF, Sengers RCA, Wintjes LTM, Ruitenbeek
W, Smeitink JAM, et al. Measurement of the energy-generating
sensitivity measured in digitonin- and Percoll-treated
capacity of human muscle mitochondria: diagnostic procedure and
fibroblasts as described by Chretien et al. (13 ), 86 (19%; application to human pathology. Clin Chem 2006;52:860 –71.
n ⫽ 22), was similar to our results. The radiochemical 6. Rustin P, Chretien D, Bourgeron T, Gérard B, Rötig A, Saudubray
enzyme assay of Brooks and Krähenbühl (9 ) gave a JM, et al. Biochemical and molecular investigations in respiratory
slightly lower rotenone sensitivity of 60%– 80%. As both chain deficiencies. Clin Chim Acta 1994;228:35–51.
Chretien et al. (13 ) and Brooks and Krähenbühl (9 ) 7. Cooperstein SJ, Lazarow A. A microspectrophotometric method for
measured complex I in crude lysates and expressed the determination of cytochrome oxidase. J Biol Chem 1951;189:
activities on protein base, it was not possible to directly 665–70.
compare the complex I activities measured by these 2 8. Lowry OH, Rosebrough NJ, Farr AL, Randell RJ. Protein measure-
methods with our results. ment with the Folin phenol reagent. J Biol Chem 1951;193:265–
75.
The method we developed is also suitable for measure-
9. Brooks H, Krähenbühl S. Development of a new assay for complex
ments of complex I in 600g supernatants from muscle. The I of the respiratory chain. Clin Chem 2000;46:345–50.
sensitivity and specificity appeared to be lower than 10. Jewess PJ, Devonshire AL. Kinetic microplate-based assays for
observed with mitochondrial fractions. inhibitors of mitochondrial NADH:ubiquinone oxidoreductase
This complex I assay has clear advantages over the (complex I) and succinate:cytochrome c oxidoreductase. Anal
commonly used assays because it is nonradioactive; Biochem 1999;272:56 – 63.
shows high sensitivity, precision, and rotenone sensitiv- 11. Panov A, Dikalov S, Shalbuyeva N, Taylor G, Sheror T, Greenamyre
ity; and can be performed on a simple spectrophotometer. JT. Rotenone model of Parkinson disease: multiple brain mito-
The method is applicable to the analysis of muscle sam- chondria dysfunctions after short term systemic rotenone intoxi-
cation. J Biol Chem 2005;280:42026 –35.
ples and is also suitable for measuring complex I in
12. Spector A, Fletcher J, Ashbrook J. Analysis of long-chain free fatty
cultured fibroblasts.
acid binding to bovine serum albumin by determination of step-
wise equilibrium constants. Biochemistry 1971;10:3229 –32.
13. Chretien D, Bénit P, Chol M, Lebon S, Rötig A, Munnich A, et al.
This work was supported by a grant from the European Assay of mitochondrial respiratory chain complex I in human
Union Framework Program 6 (MITOCIRCLE). Financial lymphocytes and cultured skin fibroblasts. Biochem Biophys Res
disclosure: None declared. Commun 2003;301:222– 4.
735–741 (2007) Automation and
Analytical Techniques
Evaluation of Assigned-Value Uncertainty for

Complex Calibrator Value Assignment Processes:
A Prealbumin Example
John Middleton1* and Jeffrey E. Vaks2
Background: Errors of calibrator-assigned values lead to augments existing methods to allow estimation of un-
errors in the testing of patient samples. The ability to certainty in complex processes.
estimate the uncertainties of calibrator-assigned values © 2007 American Association for Clinical Chemistry
and other variables minimizes errors in testing pro-
cesses. International Organization of Standardization For a clinical measuring system, the purpose of calibration
guidelines provide simple equations for the estimation and calibrator value assignment is to provide for diag-
of calibrator uncertainty with simple value-assignment nostically accurate patient test results. Along with the
processes, but other methods are needed to estimate reference ranges and clinical cutoffs, accurate patient test
uncertainty in complex processes. results provide data for valid medical decisions and
Methods: We estimated the assigned-value uncertainty effective treatment. We studied value-assignment pro-
with a Monte Carlo computer simulation of a complex cesses for the Beckman Coulter, Inc. SYNCHRON® Pre-
value-assignment process, based on a formalized de- albumin (transthyretin) immunoassay. Prealbumin (PAB)3
scription of the process, with measurement parameters is an important marker for protein malnutrition and is
estimated experimentally. This method was applied to used for assessment of patient nutritional status. Early
study uncertainty of a multilevel calibrator value as- detection of, and response to, protein malnutrition can
signment for a prealbumin immunoassay. shorten patient hospital stays and improve outcomes (1 ).
Results: The simulation results showed that the compo- To support global quality improvements in clinical
nent of the uncertainty added by the process of value diagnostics, the European Union has created the In Vitro
transfer from the reference material CRM470 to the Diagnostics Medical Devices Directive, which requires
calibrator is smaller than that of the reference material that calibrator products offered in Europe have defined
itself (<0.8% vs 3.7%). Varying the process parameters traceability and calibrator uncertainty (2 ). In addition,
in the simulation model allowed for optimizing the International Organization of Standardization (ISO) doc-
process, while keeping the added uncertainty small. The ument 17511 provides guidance to manufacturers on how
patient result uncertainty caused by the calibrator un- to establish assigned-value traceability (3 ). For determin-
certainty was also found to be small. ing calibrator-assigned value uncertainty, the ISO docu-
Conclusions: This method of estimating uncertainty is a ment refers to the Guide to the Expression of Uncertainty in
powerful tool that allows for estimation of calibrator Measurement (GUM) (4 ). Unfortunately, the GUM does
uncertainty for optimization of various value assign- not describe uncertainty analysis methods for complex
ment processes, with a reduced number of measure- value-transfer processes that involve multilevel calibra-
ments and reagent costs, while satisfying the require- tors with correlated errors. One approach to estimating
ments to uncertainty. The new method expands and the uncertainty of value assignment in such situations is
by simultaneous building of multiple lots of calibrators
and calculating SDs of the assigned value for each cali-
1
brator level. This approach, although it is conceptually
Department of Clinical Chemistry Development, Beckman Coulter Inc.,
Fullerton, CA.
2
Math/Stats Group, Beckman Coulter, Inc., Fullerton, CA.
* Address correspondence to this author at: Department of Clinical Chem-
3
istry Development, Beckman Coulter Inc., Fullerton, CA. 92821. Fax 714-961- Nonstandard abbreviations: ISO, International Organization of Standard-
3740; email jsmiddleton@beckman.com. ization; PAB, prealbumin; IWC, internal working calibrator; PC, production
Received October 3, 2006 ; accepted January 15, 2007. calibrator; VA, value assignment; GUM, Guide to the Expression of Uncer-
Previously published online at DOI: 10.1373/clinchem.2006.081174 tainty in Measurement.
735
736 Middleton and Vaks: Calibrator Assigned Value Uncertainty
simple and enables unbiased estimation, is prohibitively that is a modification of the stochastic approach. Both the
expensive. A more feasible approach is Monte Carlo Beckman Coulter and stochastic approximation methods
simulation of the process. assume that the logarithm of the signal vs the reciprocal
Aside from fulfilling In Vitro Diagnostics Medical dilution ratio (for narrow signal ranges) is approximately
Devices compliance requirements, the clinical utility of linear. In our modified method, however, rather than
estimating and reporting uncertainty in diagnostic assays doing a number of calculations over a number of itera-
has been debated (5, 6 ). We found Monte Carlo simulations, the value-assignment testing and calculations are
tions to be useful in the design and optimization of the performed in 1 step.
value-assignment process, because this method allows for To compare the 2 value-assignment methods, we used
changing simulated process parameters such as testing the same dilutions and data from the stochastic approxi-
replication and number of instruments. In essence the mation evaluation. For the modified approach, to deter-
process of PAB calibrator value assignment consists of mine the PAB concentration for the highest-level (level 8)
transferring the value from the reference material calibrator, (1/dilution) vs ln(signal), a linear equation
CRM470 (7, 8 ) to a multilevel set of internal working was fitted to data (Fig. 1) in a narrow interval around the
calibrators (IWC), and then from the IWCs to the end-use ln (signal) obtained with CRM470. The fitted line was
calibrators during their production. The value-assignment used to estimate dilution of the highest-level calibrator
process transfers the value from CRM470 to the produc- needed to obtain the same signal as the one obtained with
tion calibrator with additional uncertainty that is much CRM470. The assigned value of the highest-level calibra-
smaller than that of the published uncertainty of CRM470 tor is equal to the reciprocal of the dilution of the
itself. highest-level calibrator multiplied by the assigned value
of CRM470 (243 mg/L). This analysis is done for each of
Materials and Methods the 3 instruments used for data collection. The final value
assay method assigned to the highest level PAB IWC was the average
The Beckman Coulter reagent used to measure PAB across the instruments. The assigned values of the lower
concentration is a turbidimetric method. In the analytical concentrations are determined by use of the respective
reaction, serum PAB combines with a specific antibody to dilution ratios applied to the assigned value of the highest
form light-scattering antigen-antibody complexes. The concentration calibrator.
system monitors the changes in absorbances at several
wavelengths and uses a nonlinear calibration curve ob- production calibrator value assignment
tained with the multilevel calibrator to calculate the The production calibrators form a 6-level calibrator with
concentrations of PAB in patient samples. PAB concentrations ranging from 0 mg/L to ⬃600 mg/L.
For the value assignment of the production calibrators,
formulation of calibrators and standards 18 measurements per concentration of the production
Delipidized human serum (9 ) containing endogenous calibrator and the IWC were performed on each of 4
PAB was used to formulate the IWC and production SYNCHRON CX instruments with each of 2 reagent lots
calibrators. High PAB concentrations were achieved by (144 measurements per test sample). Within a reagent lot,
ultrafiltration concentration, and lower concentrations the data were averaged over all the instruments, and an
were created through dilutions of the concentrated ultra- IWC standard curve was determined by fitting a 5-param-
filtrate with phosphate buffered saline (PBS). For the IWC,
the dilution rations were quantified gravimetrically. So-
dium azide at 1 mL/L is added as an antimicrobial. The
upper limit of the analytical range for the PAB assay is
600 mg/L, which determines the requirement for the
highest concentration of the production calibrator. To
transfer reference values to the production calibrator from
the IWC, the latter spans the range from 0 to 700 mg/L.
iwc value assignment

The IWC are directly traceable to CRM470 through a
value-assignment process. The highest concentration cal-
ibrator is directly value assigned using CRM470, and then
the values of the lower concentrations are determined
from dilution ratios. Fig. 1. Reciprocal dilution (of IWC level 8) vs natural logarithm of the
We initially considered a stochastic approximation signal.
approach to value assignment (10 ). However, because of The IWC level 8 assigned value is calculated using the best fit line and the
CRM470 signal. The natural log of the CRM470 signal is input into the line
the complexity of this method and accumulation of dilu- equation to calculate a reciprocal dilution value. That value is multiplied by the
tion errors with each iteration, we developed a method CRM470 assigned value (24.3 mg/L) to obtain the IWC level 8 assigned value.
Fig. 2. Summary diagram of the un-

certainty assessment process.
The main areas depicted are the actual
standard and calibrator formulation and
value assignments, the simulation parame-
ter extraction process, and the Monte Carlo
simulation process.
eter logit function to the averages of instrument responses U tot ⫽ 冑U std

2
⫹ U iwc
2
⫹ U pc
2
. (2)
vs assigned concentrations:
Kc However, for multilevel value transfers with cross-
R ⫽ R0 ⫹ ⫺ a ⫺ bln(Conc) c (1) correlated errors of assigned values of calibrator levels,
共1 ⫹ e ) .
deriving closed form equations for total uncertainty for
This standard curve was then used to calculate produc- each calibrator level is practically impossible. There-
tion calibrator PAB concentrations. The assigned values of fore, we evaluated the uncertainties of the IWC, as well
the production calibrators are the average concentrations as of the production calibrators, by Monte Carlo simu-
across 2 reagent lots, 4 instruments, and 18 replicates per lation of the value transfer process. To be performed
reagent lot per instrument. correctly, the simulation process must be well de-
scribed, including accurate estimation of error contri-
Estimation of Uncertainty butions and the correlations of the errors in various
In estimating the uncertainty of the values assigned to the
stages of the process. In addition, all the variance/
calibrators, it is useful to consider the traceability chain
covariance components used in the simulation are
(See Fig. 1 in the Data Supplement that accompanies the
evaluated from the real value assignment process data.
online version of this article at http://www.clinchem.
The concepts of Monte Carlo simulation are described
org/content/vol53/issue4). Each process step in the
in (11 ). The program for the simulation for this study
traceability chain is considered when determining the
total uncertainty of the production calibrator. At the was developed in S-Plus on a PC platform (12 ); it is
beginning of the chain is the international protein refer- included in the Supplemental Appendix. A distribution
ence material CRM470, with an uncertainty, Ustd, charac- of the assigned values is generated through simulations
terized by the SD of 9.0 mg/L, as published in the of 1000 value assignment events. The SD of these
CRM470 product insert. Uiwc is the uncertainty in the simulated assigned values is the assigned value uncer-
assigned values for the IWC. Lastly, Upc denotes the tainty. The 95% confidence intervals of the uncertainty
uncertainties of the production calibrator assigned values. were determined as 2.5th and 97.5th percentiles of the
If these were single calibrator level process steps with SDs estimated from 40 runs of 1000 simulated value
uncorrelated additive error variance components, then assignment events. This simulation procedure aug-
the uncertainty of each step could be combined into 1 ments the method applicable to simple value assign-
value representing the total uncertainty (4 ): ment processes described in GUM (4 ).
plied with the (P ⫻ N) matrix of random numbers with

Table 1. PAB IWC assigned values (mg/L).
independent gaussian distribution with zero means and
A. IWC level 8 results.a
unit variances, then divided by the diagonal of the same
Instrument serial no. BCI method Stochastic method (10 )
covariance matrix (11 ). Above, P is the number of calibra-
CX S/N 611 714.4 713.3
CX S/N 1114 720.5 725.3
tor levels and N is the number of instruments. Items 3.3 to
CX S/N 3044 729.7 735.6
3.9 (Fig. 2) refer to the Monte Carlo simulation of the
Average 721.5 724.7 value-assignment events. The value-assignment calcula-
SD 7.7 11.2 tions, using the simulated instrument signals, are per-
formed as they were in the actual calibrator value assign-
B. IWC assigned values based on level 8 assigned value ments. For the IWC value-assignment calculations the
and dilution ratiob natural log of the signal data for the IWC dilutions is
Mass IWC, Mass diluent, Assigned
LVL Mi Md Values, C
regressed with the reciprocal dilution factors. The IWC set
1 n/a n/a 0.0
point estimate is based on the dilution required to match
2 18.67 242.92 50.1
the CRM470 signal. For the PC value assignment, a
3 37.33 224.26 100.4 concentration vs signal standard curve is created from the
4 74.70 188.25 200.7 IWC data by curve fitting. By inputting the production
5 112.01 151.28 301.7 calibrator signal data into this standard curve, the calibra-
6 149.31 115.07 402.2 tor concentrations are calculated.
7 197.87 67.93 533.0 The gravimetric error contributions during formula-
8 279.00 0.00 721.5 tion of the IWC and value assignment were much less
IWC level 8 density (Ds) 1.03 than 0.01%, a value that was insignificant compared to the
Diluent density (Dd) 1.00 3.72% total error of the assigned values. Therefore we
a
Prealbumin level 8 (highest-level) Internal Working Calibrator assigned-value ignored the contribution of errors of gravimetric dilution.
estimates by instrument for the BCI value-assignment method and the stochastic The uncertainty of the CRM470 standard, which was
approximation method (10 ). Of course the 2 methods give the same values included in the simulations, was the major source of IWC
within the uncertainty of the estimates.
b
uncertainty because with the described process the error
The mass values (grams) of the level 8 IWC and the diluent were converted
to volumes by using their respective densities (g/mL). Assigned values, C, were of transferring the reference value to the IWC is very
calculated from the level 8 assigned value and the dilution ratios by using the small. The uncertainty of IWC was included in the simu-
formula below. Mi is the mass of IWC level 8, Md is the mass of diluent, Ds is the lations of the value assignment of the production calibra-
density of serum, and Dd is the density of diluent. tors. Simulation of 1000 value assignment events was
M iD s
C ⫽ 72.15 repeated 40 times. The pooled SDs and the nonparametric
M iD s ⫹ M dD d
95% confidence intervals obtained from these simulation
results characterize the uncertainties. The 95% confidence
intervals of the uncertainties were set as the nonpara-
iwc and production calibrator uncertainties metrically estimated 2.5th and 97.5th percentiles of the 40
The entire uncertainty evaluation process is flow-charted groups of 1000 simulated assigned values for each level of
in Fig. 2. Items 1.1 through 1.7 refer to the actual IWC and the calibrator.
production calibrator formulation and value-assignment
processes previously described. During the preliminary
assessment of the value-assignment events for the IWC
and the production calibrator, it was determined that
instrument-to-instrument variation, as well as inter-
sample covariation of the errors, were significant and thus
must be accounted for in the Monte Carlo simulations.
Items 2.3 and 2.6 in Fig. 2 refer to the extraction of
covariance structure and variance components of the
instrument signals from the actual value-assignment data.
Here instrument signal averages obtained with the
CRM470 dilutions, IWC, and the production calibrator
samples were calculated. Also, the instrument-to-instru-
ment and within-instrument variance components of in-
strument signals were estimated with variance compo-
nents analysis (13 ) along with the covariance matrix of
instrument signals for various calibrator levels. To impose Fig. 3. Histogram of 1000 simulated assigned values for level 8 of the
the above covariation structure on the simulated instru- Internal Working Calibrator.
The standard reference deviation of this data is an estimate of uncertainty of the
ment responses for various calibrator levels, the Cholesky IWC assigned value. The lower IWC level uncertainties are determined by
decomposition of the (P ⫻ P) covariance matrix is multi- multiplying the high level uncertainty by the dilution ratio.
Table 2. PAB IWC and production calibrator uncertainties (mg/L).

A.
IWC1 IWC2 IWC3 IWC4 IWC5 IWC6 IWC7 IWC8
Assigned Value 0 50.1 100.5 200.8 301.9 402.5 533.3 722.0
Uncertainty 0 ⫾1.87 ⫾3.74 ⫾7.48 ⫾1.12 ⫾1.50 ⫾1.99 ⫾2.69
Uncertainty, % 0 ⫾3.725 ⫾3.725 ⫾3.725 ⫾3.725 ⫾3.725 ⫾3.725 ⫾3.725
95% Confidence intervala 0 ⫾0.17 ⫾0.35 ⫾0.69 ⫾1.04 ⫾1.39 ⫾1.84 ⫾2.49
B. Production calibrator total uncertaintyb

CAL1 CAL2 CAL3 CAL4 CAL5 CAL6
Assigned value 0 74.4 154.0 290.6 429.4 573.5
Uncertainty 0 ⫾2.78 ⫾5.75 ⫾10.89 ⫾16.14 ⫾21.65
Uncertainty, % 0 ⫾3.743 ⫾3.735 ⫾3.746 ⫾3.758 ⫾3.774
95% Confidence interval 0 ⫾0.23 ⫾0.54 ⫾1.04 ⫾1.53 ⫾1.86
IWC assigned values with uncertainties.
a
The 95% confidence half-interval in the uncertainty estimates is determined nonparametrically from the minimum and maximum of the 40 estimates (excluding the
highest and lowest estimates accounting for 5% of the distribution). The percent uncertainty is constant across the levels as the lower levels are calculated from the
highest level and the appropriate dilution factor, with dilution errors being negligible. An assigned value of zero implies zero uncertainty because the sample is not
human serum based and is PAB free. The confidence interval is expressed in absolute terms (mg/L).
b
Tabulated are results from 40 sets of 1000 simulated value-assignment events. The 95% confidence half-interval (in mg/L) of the uncertainties is determined from
the minimum and maximum of the 40 estimates (excluding the highest and lowest estimates accounting for 5% of the distribution). An assigned value of zero implies
zero uncertainty because the sample is not human serum based and is PAB free. The confidence interval is expressed in absolute terms (mg/L).
This method enables estimation of the calibrator uncer- dilutions that do not match those of the reference material
tainties; however, because of the multilevel nonlinear are discarded. Here, however, the difference is not statis-
calibration, it is not obvious how these uncertainties tically significant based on F-test with 2 degrees of free-
propagate to patient result uncertainty. To evaluate these dom. The assigned values of the lower levels of the IWC
effects, recoveries of patient sample concentrations were were calculated using the dilution ratios of the highest-
simulated. In this simulation the calibrator-assigned val- level calibrators.
ues varied according to the previously determined uncer- Simulation of the IWC value assignment event was
tainties. Then, using target responses for various concen- repeated; 1000 value assignments were simulated 40 times
trations, patient results were calculated with various yielding 40 sets of 1000 assigned values. A histogram of 1
calibrations using those variable assigned values. Calcu- of the sets of 1000 assigned values is shown in Fig. 3. The
lated patient result SDs showed the effect of the calibrator pooled SD of these assigned-value sets, SD ⫽ 26.9 mg/L,
uncertainty. This method allows for establishing require- is the estimate of the assigned-value uncertainty, Uiwc, for
ments to the calibrator uncertainty. the highest-level IWC. The uncertainties, based on dilu-
tion ratio, for all 8 levels of the IWC are summarized in
Results Table 2.
The IWC value-assignment results are shown in Table 1. For the production calibrator value assignment, mea-
During the value-assignment evaluations, multiple dilu- surements of the IWC were taken to create a standard
tions of the highest level IWC were made such that their curve, from which the assigned values of the produc-
signals were near the signal of CRM470, and a linear fit tion calibrators were calculated (See Fig. 2 in the Data
was made to find the dilution necessary to obtain the Supplement that accompanies the online version of this
same signal as the reference material (Fig. 1). Once the article at http://www.clinchem.org/content/vol53/issue4).
dilution was determined, the assigned value of the The pooled SD of the 40 sets of 1000 simulated value-
highest-level calibrator was calculated by dividing the assignment events is the uncertainty. By using this ap-
assigned value of the reference material by the dilution proach, the production calibrator uncertainty includes the
factor used to match the highest-level calibrator. Table 1 IWC and CRM470 components and the error of value
shows the 3 assigned values calculated for the 3 different transfer from IWC. The final results of the value assign-
instruments, and the average of those values as the final ment, as well as the uncertainty analysis, are summarized
assigned value for the highest level IWC. For comparison, in Table 2. The relative uncertainties over all levels of the
the results using the stochastic approximation method production calibrator are ⬃3.75%.
(10 ) are also included. Our approach, because it uses all Because the uncertainty component associated with
the data from many dilutions to estimate the dilution of the production calibrator value assignment was rather
the level 8 IWC, might provide for smaller uncertainty of small, we assessed the effect of reducing the amount of
the assigned values than does stochastic approximation testing. Rather than making 144 measurements (4 instru-
(SD of 7.7 vs 11.2 mg/L), in which the signal data from ments, 36 replicates per sample), we made 48 measure-
Table 3. Uncertainty (in mg/L) as a function of the number of measurements.

Production calibrator value assignment uncertaintya
A. Testing Protocol of 144 Measurements
Calibrator level 2 3 4 5 6
Assigned value, mg/L 74.3 153.9 290.4 429.3 573.2
Uncertainty, mg/L 0.43 0.66 1.99 3.65 5.72
Relative uncertainty, % 0.574 0.428 0.684 0.850 0.997
B. Testing protocol of 96 measurements

Calibrator level 2 3 4 5 6
Assigned value, mg/L 74.3 153.9 290.4 429.2 573.5
Uncertainty, mg/L 0.49 0.78 2.39 4.29 7.07
Relative uncertainty, % 0.660 0.507 0.822 1.000 1.234
Component of patient result uncertainty due to calibrator value assignmentb

C. Testing Protocol of 144 Measurements
PAB concentration, mg/L 50 180a 280 380a 500
Result uncertainty, mg/L 0.97 1.55 2.44 2.83 2.62
Reference range, % 4.83 7.74 12.21 14.17 13.12
D. Testing Protocol of 96 Measurements

PAB concentration, mg/L 50 180a 280 380a 500
Result uncertainty, mg/L 1.11 1.80 2.80 3.30 3.23
Reference range, % 5.55 9.00 14.02 16.50 16.17
a
Tabulated are results from 1000 simulated value-assignment events. The number of measurements per sample was reduced from 144 to 96. Comparison of the
2 testing configurations showed a small increase of the uncertainty of transferring the reference material value to the production calibrators.
b
Results for 5 different PAB concentrations from 1000 simulated value-assignment events. The percentage of reference range entries is based on width of 20 mg/L
of the reference range 18 to 38 mg/L. Because the uncertainties resulting from calibrator value assignment for both testing protocols are small compared with the
width of the reference range, the risk of patient misdiagnosis because of value assignment error is low.
ments (4 instruments, 12 replicates per sample) were dilutions of the highest-level IWC were made such that
made. The results of those simulations are summarized in the signals were near that of CRM470 when assayed. The
Table 3. Here the production calibrator value assignment relationship between the dilutions and the signals allows
was isolated from the previous events such that the error for the determination of the highest-level IWC assigned
contribution from this event alone may be considered. value, while gravimetric dilutions introduce no signifi-
Lastly, the effect of calibrator uncertainty on patient result cant errors to lower levels of the IWC.
uncertainty is provided in Table 3. The final relative total uncertainties of the PAB pro-
duction calibrators were all ⬃3.75%, insignificantly high-
Discussion er than the relative uncertainty of the reference material.
Our results demonstrate that that the process of transfer- The relative uncertainties of the assigned values of all
ring PAB values from the reference material to the pro- levels of the production calibrator, excluding the uncer-
duction calibrators contributes little to the total error. The tainty of CRM470, were all ⬍0.8%. So, the major part of
assigned-value uncertainty of CRM470 is 243 (9.0) mg/L, the uncertainty of the BCI PAB production calibrators is
with a relative error of 3.70%. Through the value trans- comprised of the uncertainty of the reference material,
fer process from this reference material to the IWC, the because the value-transfer process adds very little to the
highest-level IWC had an assigned value of 722 (26.9) total uncertainty. The exceedingly small contribution to
mg/L with a relative uncertainty of 3.72%. The compo- the total error of the IWC and production calibrator value-
nent of uncertainty of the IWC assigned value, con- transfer steps, although counterintuitive, is partially of
tributed by the process of transferring the value from the result of taking the square root of the sum of squared
reference material, is negligible. So, the described value- individual contributions per equation (2). In addition, the
assignment process allows for a precise transfer of values relatively large number of measurements reduces the
from a single-level international standard CRM470 to a uncertainty. In this instance the reduction is ⬃10-fold,
multiple level IWC. Unlike the stochastic approximation the square root of one hundred. As shown in Table 3,
approach (10 ), which relies on determining the assigned reducing the amount of testing of the production calibra-
values by an iterative process involving many tests and tor by 33% reduces the cost of the calibrator but has little
calculations, the described approach is simpler and pro- effect on the total calibrator uncertainty or on the assay
vides for at least as good calibrator uncertainty. Multiple precision.
The question of clinical usefulness can be addressed in 4. International Organization for Standardization. Guide to the Ex-
relation to patient result uncertainties due to calibrator pression of Uncertainty in Measurement, Geneva: ISO, 1995;:
uncertainties (Table 3). To be clinically useful, the errors 101pp.
5. Krouwer J. Critique of the Guide to the Expression of Uncertainty in
of the patient results caused by the calibrator uncertain-
Measurement Method of Estimating and Reporting Uncertainty in
ties must be small relative to the PAB reference range of Diagnostic Assays. Clin Chem 2003;49(11):1818 –21.
180 –380 mg/L (14 ). At the lower and upper limits of the 6. Kristiansen J. The Guide to Expression of Uncertainty in Measure-
reference range, uncertainties of the patient results caused ment Approach for Estimating Uncertainty: An Appraisal. Clin
by the uncertainties of the calibrator are 1.6 and 2.8 mg/L, Chem 2003;49(11):1822–9.
respectively, with 96 measurements per sample. These 7. Whicher JT, Ritchie RF, Johnson AM, Baudner S, Bienvenu J,
values as percentages of the width of the reference range Blirup-Jensen S, et al. New International Reference Preparation for
Proteins in Human Serum (RPPHS). Clin Chem 1994;40(6):
are 0.9% and 1.6%, respectively, small relative to the 934 – 8.
reference range. Therefore, the risk of patient misdiagno- 8. Johnson AM, Sampson EJ, Blirup-Jensen S, Svendsen PJ. Recom-
sis due to calibrator uncertainty is small. mendations for the Selection and Use of Protocols for Assignment
of Values to Reference Materials. Eur J Clin Chem Clin Biochem
The contributions to this work by Phebe Shaw, Phil Martz, 1996;34:279 – 85.
Art Kessner, and Steve Alter were invaluable. 9. Vance DE, Weinstein DB, Steinberg D. Isolation and Analysis of
Lipoproteins Secreted by Rat Liver Hepatocytes. Biochim Biophys
Acta 1984;792:39 – 47.
References 10. Schlain B. A Stochastic Approximation Method for Assigning
1. Beck FK, Rosenthal TC. Prealbumin: A Marker for Nutritional Values to Calibrators. Clin Chem 1998;44(4):839 – 48.
Evaluation. Am Fam Physician 2002;65(8):1575– 8. 11. Rubenstein RY. Simulation and the Monte Carlo Method. New
2. “Council Directive 98/79/EC of the European Parliament and of York: John Wiley, 1981.
the Council of 27 October 1998 on In Vitro Diagnostic Medical 12. Venables WN, Ripley BD. Statistics and Computing. Modern
Devices”, Official Journal of the European Union L331 (December Applied Statistics with S-Plus. New York: Springer-Verlag, 1994.
12, 1998). 13. Corbeil RR and Searle SR. Restricted maximum likelihood (REML)
3. International Organization for Standardization. In vitro diagnostic estimation of variance components in the mixed model. Techno-
medical devices-measurement of quantities in biological samples- metrics 1976;18:31–38.
metrological traceability of values assigned to calibrators and 14. Synchron CX Chemistry Information Manual. Beckman Coulter
control materials (17511). Geneva: ISO, 2003:23 pp. reorder no. 249595.
742–747 (2007) Automation and
HPLC for Simultaneous Quantification of Total

Ceramide, Glucosylceramide, and Ceramide
Trihexoside Concentrations in Plasma
Johanna E.M. Groener,1* Ben J.H.M. Poorthuis,1 Sijmen Kuiper,1
Mariette T.J. Helmond,1 Carla E.M. Hollak,2 and Johannes M.F.G. Aerts1
Background: Simple, reproducible assays are needed ␮mol/L for CTH. Plasma concentrations of GlcCer were
for the quantification of sphingolipids, ceramide (Cer), higher in Gaucher disease patient samples and of CTH
and sphingoid bases. We developed an HPLC method in Fabry disease patient samples.
for simultaneous quantification of total plasma concen- Conclusions: HPLC enables quantification of total Cer,
trations of Cer, glucosylceramide (GlcCer), and cer- GlcCer, and CTH in plasma and is useful for the
amide trihexoside (CTH). follow-up of patients on therapy for Gaucher or Fabry
Methods: After addition of sphinganine as internal disease.
calibrator, we extracted lipids from 50 ␮L plasma. We © 2007 American Association for Clinical Chemistry
deacylated Cer and glycosphingolipids by use of micro-
wave-assisted hydrolysis in methanolic NaOH, fol- Sphingolipidoses are inherited lysosomal storage disor-
lowed by derivatization of the liberated amino-group ders caused by lysosomal hydrolase deficiency leading to
with o-phthaldialdehyde. We separated the derivatized storage of specific glycosphingolipids. Two forms of ther-
sphingoid bases and lysoglycosphingolipids by HPLC apies are available. In enzyme replacement therapy
on a C18 reversed-phase column with a methanol/water (ERT),3 the missing enzyme is administered to the patient
mobile phase (88:12, vol/vol) and quantified them by intravenously. The enzyme is taken up by cells by recep-
use of a fluorescence detector at ␭ex 340 nm and ␭em tor-mediated endocytosis, which leads to reduction of the
stored material. Currently, this therapy is used in the
435 nm.
treatment of Gaucher disease type 1 (1, 2 ) and Fabry
Results: Optimal conditions in the Solids/Moisture Sys-
disease (3, 4 ). Treatment of Niemann-Pick type B is under
tem SAM-155 microwave oven (CEM Corp.) for the
development. In substrate reduction therapy (SRT), the
complete deacylation of Cer and neutral glycosphingo-
synthesis of glycosphingolipids is reduced by partial
lipids without decomposition were 60 min at 85%
inhibition of glucosylceramide (GlcCer) synthase by de-
power, fan setting 7. Intra- and interassay CVs were
oxynojirimycin analogs (5 ), leading to reduction of the
<4% and <14%, respectively, and recovery rates were
glycosphingolipid pool and eventually to reduction of the
87%–113%. The limit of quantification was 2 pmol (0.1 substrate load for lysosomal degradation by sphingolipid
pmol on column), and the method was linear over the hydrolases. This therapy is applied to a subgroup of
interval of 2–200 ␮L plasma. In samples from 40 healthy patients with Gaucher disease type 1 for whom ERT is not
individuals, mean (SD) concentrations were 9.0 (2.3) available (6 ). Because SRT is nonspecific, it could in
␮mol/L for Cer, 6.3 (1.9) ␮mol/L for GlcCer, and 1.7 (0.5) principle be applied to all sphingolipidoses in which the
degradation of GlcCer-derived sphingolipids is impaired.
In contrast to the enzymes administered in ERT, the drug
used for SRT can cross the blood– brain barrier and reach
Departments of 1 Medical Biochemistry and 2 Internal Medicine, Univer-
sity of Amsterdam, Academic Medical Center, Amsterdam, The Netherlands.
target cells in the brain and thus could be used in the
*Address correspondence to this author at: Department of Medical Bio-
chemistry, University of Amsterdam, Academic Medical Center, P.O. Box
22700, 1100 DE Amsterdam, The Netherlands. Fax 31-20-6915519; e-mail
3
j.e.groener@amc.uva.nl. Nonstandard abbreviations: ERT, enzyme replacement therapy; SRT,
Received September 1, 2006; accepted January 26, 2007. substrate reduction therapy; GlcCer, glucosylceramide; Cer, ceramide; CTH,
Previously published online at DOI: 10.1373/clinchem.2006.079012 ceramide trihexoside; OPA, o-phthaldialdehyde.
742
treatment of sphingolipidosis with involvement of the microwave oven

central nervous system (7 ). SRT trials are under way for We used the Solids/Moisture System SAM-155 (CEM
Niemann-Pick type C, juvenile GM2-gangliosidosis, and Corp.) with a power range of 0% to 100% of full power
Gaucher disease type 3. (630 W), an exhaust fan, and a rotating Teflon tray with 36
To monitor the effect of these therapies, glycosphingo- tube holes.
lipids in plasma must be measured. In addition to moni-
toring the primary storage products, it is important to sample preparation and hplc
measure possible changes in ceramide (Cer) concentrations We added 1 nmol C18-sphinganine, as the internal cali-
and other intermediates or products in glycosphingolipid brator, to 50 ␮L plasma. Lipids were extracted with 2 mL
metabolism, particularly in SRT. Cer is an intermediate in chloroform/methanol (1:1, vol/vol) for 30 min at room
synthesis of glycosphingolipid, sphingomyelin, sphingosine, temperature, followed by centrifugation for 5 min at
1500g. We added 1 mL chloroform and 0.75 mL 0.73%
and Cer-1-P, compounds that are increasingly recognized
NaCl to the clear extract to obtain phase separation
as mediators of cellular processes. They participate in
according to Folch et al. (25 ) and removed the lower
various cell-surface–related processes such as cell differ-
chloroform layer. We washed the upper phase once with
entiation, transmembrane signaling, cell recognition, and
1 mL chloroform, and the combined chloroform layers
toxin binding (8, 9 ).
were thoroughly evaporated to dryness. Using 0.5 mL
Several techniques are available to measure Cer and, in freshly prepared 0.1 mol/L NaOH in methanol, we deacy-
particular, neutral glycosphingolipids. These techniques lated the lipids in 12 ⫻ 100 mm borosilicate glass tubes
include thin-layer chromatography with densitometry or (Schott GL14) with polytetrafluorethylene-lined screw
immunochemical detection to HPLC and mass spectrom- caps in the CEM microwave oven. For reasons of repro-
etry (10 –21 ). Most of these methods, however, do not ducibility and safety, 36 tubes filled with 0.5 mL of 0.1
allow the simultaneous measurement of Cer and neutral mol/L NaOH in methanol were always present. To max-
glycolipids, or they require extensive calibration and imize deacylation and minimize decomposition, the opti-
expensive equipment (as in the case of mass spectrome- mal condition for this microwave oven proved to be an
try). Here we describe a simple method for the simulta- exposure time of 60 min at 85% power and a fixed setting
neous quantification of Cer and neutral glycosphingolip- of 7 for the exhaust fan (see Results).
ids in plasma and its application for monitoring GlcCer After hydrolysis, the samples were cooled to room
and Cer in patients with Gaucher disease on ERT and SRT temperature inside the microwave oven. We transferred
and for ceramide trihexoside (CTH) and Cer in patients 50 ␮L of the deacylated lipid mixture into autosampler
with Fabry disease on ERT. The method is based on the vials and derivatized the lipids for 30 min with the
conversion of glycosphingolipids and Cer to lysoglyco- addition of 25 ␮L OPA reagent [5 mg OPA, 0.1 mL
sphingolipids and sphingoid bases by microwave-medi- ethanol, 5 ␮L 2-mercaptoethanol, and 10 mL 3% (wt/vol)
ated saponification, as pioneered by Taketomi et al. (22 ) boric acid adjusted to pH 9.0], essentially as described by
and discussed by Itonori et al. (23 ). We optimized this Merrill et al. (24 ). The final pH of the mixture during OPA
deacylation procedure and quantified the sphingoid bases derivatization was pH 10.7. We separated OPA-derivat-
and lysoglycosphingolipids after derivatization with o- ized sphingoid bases and lysoglycosphingolipids by use
phthaldialdehyde (OPA) by HPLC with fluorometric de- of a Waters HPLC system consisting of a 600 Controller,
717plus Autosampler, and 474 Scanning Fluorescence
tection (24 ).
Detector with an Altima BDS C18 3␮, 150 ⫻ 4.6 mm
reversed-phase column. The mobile phase was methanol/
water 88:12 (vol/vol). We quantified the OPA-derived
materials
lysocompounds with a fluorescence detector at ␭ex 340 nm
We obtained C18-sphinganine, sphingosine, lyso-GlcCer,
and ␭em 435 nm. Peak identification was based on com-
lyso-Cer, lyso-CTH, Cer, GlcCer, lactosylceramide, and
parison of the retention times with those of authentic
CTH from Avanti Polar Lipids; methanol pro analysi and calibrators of the lysocompounds (sphingosine, sphinga-
chloroform pro analysi from Merck KGaA; and OPA from nine, lyso-GlcCer, and lyso-CTH) and those of Cer,
Sigma. GlcCer, and CTH after microwave-assisted deacylation.
Peak integration was performed with Waters Millennium
plasma samples software. All samples were run in duplicate, and 2 refer-
We obtained samples from patients referred to the Lyso- ence samples were included in every run.
somal Outpatient Clinic of the Academic Medical Center
in Amsterdam for assessment of severity of disease or quantification, recovery, and imprecision
institution of ERT or SRT. Control samples were obtained studies
from healthy volunteers. Informed consent was provided We used C18-sphinganine as the internal calibrator for
according to the Declaration of Helsinki. Plasma samples quantification. The limit of detection was defined as the
were stored at ⫺20 °C until use. amount of sphinganine producing a signal-to-background
744 Groener et al.: Quantification of Neutral Glycosphingolipids and Cer
mol/L NaOH in methanol, placed in the microwave oven,

and exposed to different levels of energy for different time
periods. Fig. 1 shows the effect of time and energy output
on the hydrolysis of Cer and GlcCer at 2 different fan
settings. At settings 6 and 7, hydrolysis time was set at 30,
60, and 90 min and the power was set at 75%, 85%, and
95%. At the higher setting of the fan (Fig. 1, A and B),
deacylation of Cer was complete after 30 min (Fig. 1A)
and did not vary significantly among the 3 power settings
and 3 time periods. GlcCer deacylation (Fig. 1B) was
optimal at 85% and 95% power but was clearly lower at
75% power. The curve also shows that the optimal time
for deacylation of GlcCer was between 60 and 120 min; at
shorter time periods, deacylation was incomplete. At a
lower fan setting of 6, which results in a higher temper-
ature and overall higher energy input, a decrease in the
amount of sphingosine and lyso-GlcCer was noted at
Fig. 1. Time course of the formation of sphingosine and lyso-GlcCer longer time periods (Fig. 1, C and D), suggesting decom-
from Cer and GlcCer in plasma at different fan and power settings. position of the (glyco)sphingolipids. Therefore 60 min at
(A, B), fan setting 7. (C, D), fan setting 6. (A, C), formation of sphingosine from 85% power and fan setting 7 was selected as the optimal
Cer at different power settings. (B, D), formation of lyso-GlcCer from GlcCer at
different power settings.
condition for the alkaline hydrolysis of Cer and GlcCer.
ratio of 3. Plasma normally contains very low concentra-

tions of sphinganine, which do not affect the quantifica-
tion (see Results). We measured recovery by adding
known amounts of Cer, GlcCer, and CTH to 2 different
plasma specimens. Because Cer, GlcCer, and CTH consist
of a mixture of different molecular species, we prepared
calibrator solutions of these compounds used for the
recovery studies using an estimated average molecular
weight and exactly measured each concentration by quan-
tification after deacylation. Recoveries were calculated for
each concentration. To calculate within-assay imprecision,
we measured plasma concentrations 10 times in the same
assay series; between-assay imprecision was calculated
using results from the same control and Fabry plasma
samples analyzed on 10 different days. We checked the
linearity of the method by serial 25-fold dilution of a
Gaucher plasma and a Fabry plasma sample in phos-
phate-buffered saline in the lower concentration interval,
and a 4-times-higher volume of plasma than used in the
standard assay in the higher concentration interval. We
defined the limit of quantification of Cer, GlcCer, and
CTH with the current method as the amount that gave a
signal-to-noise ratio of 5.
Results
microwave-assisted deacylation
Because our goal was to develop an assay suitable for a
large set of samples, we selected an industrial microwave
oven (CEM Solids/Moisture System SAM-155) with a
rotating tray with 36 tube holes. We first determined the
optimal microwave conditions—i.e., conditions for com-
Fig. 2. Example HPLC chromatograms of plasma from a control, a
plete deacylation with minimal decomposition—for the patient with Gaucher disease, and a patient with Fabry disease.
alkaline hydrolysis of Cer and GlcCer in lipid extracts of Peak identification: internal calibrator (IS), sphinganine; 1, sphingosine; 2,
plasma. Lipid extracts were redissolved in 0.5 mL of 0.1 lyso-GlcCer; 3, tentative lysolactosylceramide; 4, lyso-CTH.
The settings of the microwave oven for the optimal amounts added, and the recoveries were 87% to 113%
hydrolysis of CTH were found to be the same as for (Table 1).
GlcCer (data not shown). It should be emphasized that We tested the linearity of Cer, GlcCer, and CTH
these settings are specific for the CEM Solids/Moisture measurement after serial dilution of a plasma sample
System SAM-155 microwave oven, filled with 36 Schott from a Gaucher patient with high GlcCer concentrations
GL14 tubes containing 0.5 mL 0.1 mol/L NaOH in and a plasma sample from a Fabry patient with high CTH
methanol. concentrations. In the Gaucher plasma, the concentrations
The chromatograms for plasma from a control and pa- were Cer 6.49, GlcCer 20.74, and CTH 1.52 ␮mol/L. In the
tients with Gaucher and Fabry disease are shown in Fig. 2. Fabry plasma, the concentrations were Cer 7.22, GlcCer
The chromatograms show small unidentified peaks that do 6.20, and CTH 7.90 ␮mol/L. The amount of lipids mea-
not interfere with those of the (glyco)sphingolipids. sured was linear with the volume of plasma over the
Sphingoid bases and lysoglycosphingolipids are quan- whole interval between 2 and 200 ␮L. For individual
tified with this method, so it is important to know the lipids, the correlation coefficient was ⬎0.99 (Fig. 3).
amount of lysoglycosphingolipids and sphingoid bases The intraassay CV (n ⫽ 10) measured in Gaucher
originally present in the plasma samples. To measure
plasma and Fabry plasma was ⬍4% for Cer, GlcCer, and
these, we applied lipid extracts to the column without
CTH. The interassay CV (n ⫽ 10) was ⬍14%.
deacylation and with and without the addition of the
internal calibrator sphinganine. We found no detectable
application of the method
amounts of lysoglycosphingolipids. We detected very low
To assess the suitability of this method for the follow-up
concentrations of sphingosine, ⬍3% of the sphingosine
of Gaucher and Fabry patients during therapy, we inves-
present after deacylation of Cer and sphinganine. The
tigated samples from 4 Gaucher patients, 2 on ERT and 2
amount of sphinganine was ⬍2% of the amount of the
internal calibrator. on SDT, and 2 hemizygous Fabry patients on ERT. We
compared the concentrations in plasma of the patients
validation of the method before therapy (Fig. 4, t ⫽ 0) with those of plasma from 40
The limit of detection achieved with calibrators applied controls. Before start of therapy, GlcCer was higher in
directly to the HPLC, defined as peak signal-to-noise ratio patients with Gaucher disease than in controls [mean
⬎3, was 20 fmol. Routinely, only 2% to 5% of the extracted concentration of GlcCer in plasma from controls was 6.3
glycolipids were injected onto the HPLC column. The (1.9) ␮mol/L (n ⫽ 40)]. On therapy, GlcCer decreased. Cer
limit of quantification obtained with serial dilution of concentrations were within the reference interval [mean
plasma, defined as signal-to-noise ratio of 5, was 2 pmol concentration of Cer in plasma from controls was 9.0 (2.3)
present in the lipid extract (0.1 pmol on the column). In ␮mol/L (n ⫽ 40)]. No change in Cer concentrations was
practice, 2 ␮L plasma gave a signal-to-noise ratio ⬎5 for noticed during the 24 months of therapy, either ERT or
all components measured. SRT (Fig. 4). In patients with Fabry disease, increased
We measured recoveries of Cer, GlcCer, and CTH by concentrations of CTH were measured before therapy (t ⫽
standard addition of these lipids to plasmas. The amounts 0 in Fig. 4) [mean concentration of CTH in plasma from
of Cer, GlcCer, and CTH showed a linear function of the controls was 1.7 (0.5) ␮mol/L (n ⫽ 40)]. On therapy, the
Table 1. Recovery studies.a

Plasma Added solution, ␮L b
CTH GlcCer Cer
c c
Measured Expected Recovery Measured Expected Recovery Measured Expected Recoveryc
Control
0 1.3 6.0 8.6
5 2.4 2.7 89 6.5 7.0 93 12.3 11.7 106
10 3.7 4.0 93 8.2 9.6 85 15.7 14.7 106
25 7.3 7.9 92 13.3 15.0 89 26.2 24.1 109
50 14.3 14.5 98 21.0 23.9 88 40.6 39.6 103
Fabry
0 7.1 6.4 6.4
5 7.2 8.4 85 6.6 8.2 80 10.3 9.5 109
10 9.4 9.7 97 9.0 10.0 90 14.3 12.6 113
25 14.0 13.7 103 15.0 15.4 97 25.5 21.9 116
50 20.7 20.3 102 23.0 24.3 95 42.5 37.4 114
a
Data are expressed in ␮mol/L and are the mean of duplicate measurements.
b
Methanolic solution with 31.0 ␮mol/L Cer, 17.9 ␮mol/L GlcCer, and 13.2 ␮mol/L CTH.
c
Calculated as percentage of measured compared with expected.
746 Groener et al.: Quantification of Neutral Glycosphingolipids and Cer
Fig. 3. Linearity of the measurement of Cer, GlcCer, and CTH in plasma

ranging in volume from 2 to 200 ␮L.
(A), Gaucher plasma: filled diamonds, Cer (R2 ⫽ 0.9967); open squares, GlcCer
(R2 ⫽ 0.9934); filled triangles, CTH (R2 ⫽ 0.9888). B, Fabry plasma: filled
diamonds, Cer (R2 ⫽ 0.9994); open squares, GlcCer (R2 ⫽ 0.9996); filled
triangles, CTH (R2 ⫽ 0.9987).
concentrations of CTH in plasma of Fabry patients de-

creased (Fig. 4). Fig. 4. Effect of therapeutic intervention on plasma GlcCer and Cer
concentrations in Gaucher patients and on CTH and Cer concentra-
tions in Fabry patients.
Discussion (A, B), Gaucher patients 1 and 2 on ERT. C and D, Gaucher patients 3 and 4 on
We developed an HPLC method for simultaneously SRT: filled circles, GlcCer; open circles, Cer. E and F, Fabry patients 1 and 2 on
quantifying Cer and neutral glycosphingolipids in large ERT: open squares, CTH; filled squares, Cer.
numbers of plasma samples. The method is based on the
microwave-assisted hydrolysis of glycosphingolipids and
Cer (22, 23 ). By critically examining the efficiency of the
microwave-assisted hydrolysis of Cer, GlcCer, and CTH, (13, 30, 31 ). These data illustrate the suitability of the new
we established optimal conditions for deacylation. The method for the follow-up of patients with Gaucher and
procedure depends on achieving complete and reproduc- Fabry disease during therapy. Data on Cer concentrations
ible deacylation of glycosphingolipids and Cer without in human plasma are scarce. Our data are comparable to
decomposition. Because we were interested in a method those published (32–34 ).
suitable for measurements of large numbers of samples, We obtained total concentrations of Cer and the glyco-
we used a laboratory-grade instrument supplied with a sphingolipids of interest, independent of their molecular
Teflon tray suitable for 36 tubes. The power setting, fan fatty acid composition. Recently, mass spectrometric anal-
setting, and time are optimal for this microwave oven. In ysis of sphingolipids has been described (16 –21 ). Mass
addition, the oven should always be filled with 36 tubes spectrometric analysis renders information and quantifi-
containing 0.5 mL of 0.1 mol/L NaOH in methanol. cation of molecular species of a particular glycosphingo-
Optimal conditions for deacylation would need to be lipid; however, quantification is more difficult since each
established for different microwave ovens. sample should ideally be supplemented with appropriate
We show the effect of therapy on the concentrations of calibrators for each molecular form. In addition, mass
GlcCer and Cer in plasma from 4 Gaucher disease pa- spectrometry requires expensive equipment and special-
tients. Our data for plasma GlcCer in plasma from con- ized expertise, which are not always available in clinical
trols and Gaucher patients compared well with those chemistry laboratories. Our HPLC fluorescence detection
obtained with thin-layer chromatography and electro- method provides an alternative approach for simulta-
spray ionization–tandem mass spectrometry (26 –30 ). In neous quantification of neutral glycosphingolipids and
the 4 patients, GlcCer concentrations decreased during Cer in plasma and can be used to monitor plasma
both ERT and SRT. In the 2 Fabry patients, concentrations concentrations of Cer, GlcCer, and CTH in patients with
of CTH decreased upon therapy, as has been reported Gaucher disease and Fabry disease during therapy.
References 18. Mills K, Eaton S, Ledger V, Young E, Winchester B. The synthesis

1. Beutler E, Grabowski GA. Gaucher disease. In: Scriver CR, Beau- of internal standards for the quantitative determination of sphin-
det AL, Sly WS, Valle D, eds. The Metabolic and Molecular Bases golipids by tandem mass spectrometry. Rapid Commun Mass
of Inherited Disease, 8th ed. New York: McGraw-Hill, 2001:3635– Spectrom 2005;19:1739 – 48.
68. 19. Levery SB. Glycosphingolipid structural analysis and glycosphin-
2. Barton NW, Brady RO, Dambrosia JM, Di Bisceglie AM, Doppelt golipidomics. Methods Enzymol 2005;405:300 – 69.
SH, Hill SC, et al. Replacement therapy for inherited enzyme 20. Merrill AH Jr, Sullards MC, Allegood JC, Kelly S, Wang E. Sphin-
deficiency: macrophage-targeted glucocerebrosidase for Gau- golipidomics: high-throughput, structure-specific, and quantitative
cher’s disease. N Engl J Med 1991;324:1464 –70. analysis of sphingolipids by liquid chromatography tandem mass
3. Desnick RJ, Ioannou YA, Eng CM. ␣-Galactosidase A deficiency: spectrometry. Methods 2005;36:207–24.
Fabry disease. In: Scriver CR, Beaudet AL, Sly WS, Valle D, eds. 21. Suzuki Y, Suzuki M, Ito E, Goto-Inoue N, Miseki K, Iida J, et al. A
The Metabolic and Molecular Bases of Inherited Disease, 8th ed. convenient structural analysis of glycosphingolipids using MALDI-
New York: McGraw-Hill, 2001;3733–74. QIT-TOF mass spectrometry with increased laser power and cool-
4. Schiffmann R, Kopp JB, Austin HA 3rd, Sabnis S, Moore DF, ing gas flow. J Biochem (Tokyo) 2006;139:771–7.
Weibel T, et al. Enzyme replacement therapy in Fabry disease: a 22. Taketomi T, Hara A, Uemura K, Kurahashi H, Sugiyama E. A
randomized controlled trial. JAMA 2001;285:2743–9. microwave-mediated saponification of galactosylceramide and
5. Platt FM, Neises GR, Dwek RA, Butters TD. N-Butyl-deoxynojirimy- galactosylceramide I3-sulfate and identification of their lysocom-
cin is a novel inhibitor of glycosphingolipid biosynthesis. J Biol pounds by delayed extraction matrix-assisted laser desorption
Chem 1994;269:8362–5. ionization time-of-flight mass spectrometry. Biochem Biophys Res
Commun 1996;224:462–7.
6. Cox T, Lachmann R, Hollak CE, Aerts J, van Weely S, Hrebicek M,
23. Itonori S, Takahashi M, Kitamura T, Aoki K, Dulaney JT, Sugita M.
et al. Novel oral treatment of Gaucher’s disease with N-butyl-
Microwave-mediated analysis for sugar, fatty acid, and sphingoid
deoxynojirimycin (OGT918) to decrease substrate biosynthesis.
compositions of glycosphingolipids. J Lipid Res 2004;45:574 –
Lancet 2000;355:1481–5.
81.
7. Aerts JMEG, Hollak CEM, Boot RG, Groener JEM, Maas M.
24. Merrill AH Jr, Wang E, Mullins RE, Jamison WCL, Nimkar S, Liotta
Substrate reduction therapy of glycosphingolipids storage disor-
DC. Quantitation of free sphingosine in liver by high-performance
ders. J Inherit Metab Dis 2006;29:448 –55.
liquid chromatography. Anal Biochem 1988;171:373– 81.
8. Hannun YA, Obeid LM. The ceramide-centric universe of lipid
25. Folch J, Lees M, Sloane Stanley GH. A simple method for the
mediated cell regulation: stress encounters of the lipid kind. J Biol
isolation and purification of total lipids from animal tissues. J Biol
Chem 2002;277:25847–50.
Chem 1957;226:497–509.
9. Hakomori S. Structure, organization, and function of glycosphin-
26. Nilsson O, Hakansson G, Dreborg S, Groth CG, Svennerholm L.
golipids in membrane. Curr Opin Hematol 2003;10:16 –24.
Increased cerebroside concentration in plasma and erythrocytes
10. Cremesti AE, Fisch AS. Current methods for the identification and in Gaucher disease: significant differences between type I and
quantitation of ceramides: an overview. Lipids 2002;35:937– 45. type III. Clin Genet 1982;22:274 –9.
11. Svennerholm S, Svennerholm L. The separation of neutral blood 27. Erikson A, Åström M, Månsson JE. Enzyme infusion therapy of the
serum glycolipids by thin-layer chromatography. Biochim Biophys Norrbottnian (type 3) Gaucher disease. Neuropediatrics 1995;26:
Acta 1963;70:432– 41. 203–7.
12. Berna L, Asfaw B, Conzelmann E, Cerny C, Ledvinova J. Determi- 28. Erikson A, Forsberg H, Nilsson M, Åström M, Månsson JE. Ten
nation of urinary sulfatides and other lipids by combination of years’ experience of enzyme infusion therapy of Norrbottnian (type
reversed-phase and thin-layer chromatographies. Anal Biochem 3) Gaucher disease. Acta Paediatrica 2006;95:312–7.
1999;269:304 –11. 29. Whitfield PD, Nelson P, Sharp PC, Bindloss CA, Dean C, Raven-
13. Zeidner KM, Desnick RJ, Ioannou YA. Quantitative determination scroft EM, et al. Correlation among genotype, phenotype, and
of globotriaosylceramide by immunodetection of glycolipid-bound biochemical markers in Gaucher disease: Implication for the
recombinant verotoxin B subunit. Anal Biochem 1999;267:104 – prediction of disease severity. Mol Genet Metab 2002;75:46 –55.
13. 30. Vance DE, Krivit W, Sweeley CC. Concentration of glycosylceram-
14. Ullman MD, Pyeritz, Moser HW, Wenger DA, Kolodny EH. Applica- ides in plasma and red cells in Fabry’s disease, a glycolipid
tion of “high-performance” liquid chromatography to the study of lipidosis. J Lipid Res 1969;10:188 –92.
sphingolipidosis. Clin Chem 1980;26:1499 –503. 31. Boscaro F, Pieraccini G, la Marca G, Bartolucci G, Luceri C, Luceri
15. Pons A, Popa J, Portoukalian J, Bodennec J, Ardail D, Kol O, et al. F et al. Rapid quantitation of globotriaosylceramide in human
Single-step gas chromatography-mass spectrometry analysis of plasma and urine: a potential application for monitoring enzyme
glycolipid constituents as heptafluorobutyrate derivatives with a replacement therapy in Anderson-Fabry disease. Rapid Commun
special reference to the lipid portion. Anal Biochem 2000;284: Mass Spectrom 2002;16:1507–14.
201–16. 32. Samuelson K. Identification and quantitative determination of
16. Fujiwaki T, Yamaguchi S, Tasaka M, Sakura N, Taketomi T. ceramide in human plasma. Scan J Clin Lab Invest 1971;27:371–
Application of delayed extraction-matrix-assisted laser desorption 80.
ionization time-of-flight mass spectrometry for analysis of sphin- 33. Sugita M, Iwamori M, Evans J, McCluer RH, Dulaney JT, Moser
golipids in pericardial fluid, peritoneal fluid and serum from HW. High performance liquid chromatography of ceramides: appli-
Gaucher disease patients. J Chromatogr B Analyt Technol Biomed cation to analysis in human tissues and demonstration of cer-
Life Sci 2002;776:115–23. amide excess in Farber’s disease. J Lipid Res 1974;15:223– 6.
17. Nelson BC, Roddy T, Araghi S, Wilkens D, Thomas JJ, Zhang K, et 34. Drobnik W, Liebisch G, Audebert F-X, Fröhlich D, Glück T, Vogel P,
al. Globotriaosylceramide isoform profiles in human plasma by et al. Plasma ceramide and lysophosphatidylcholine inversely
liquid chromatography-tandem mass spectrometry. J Chromatogr correlate with mortality in sepsis patients. J Lipid Res 2003;44:
B Analyt Technol Biomed Life Sci 2004;805:127–34. 754 – 61.
748 –756 (2007) Automation and
Monoclonal Antibody-Based Time-Resolved

Fluorescence Immunoassays for Daidzein,
Genistein, and Equol in Blood and Urine:
Application to the Isoheart Intervention Study
Duncan C.S. Talbot,1* Richard M. Ogborne,1 Tony Dadd,1 Herman Adlercreutz,4
Geoff Barnard,3 Susanne Bugel,5 Fortune Kohen,2 Sandra Marlin,1 Jerry Piron,1
Aedin Cassidy,6 and Jonathan Powell1
Background: Time-resolved fluorescence immunoas- 0.911– 0.994) across the concentration range observed in
says (TR-FIAs) for phytoestrogens in biological samples the Isoheart study (50 mg/day isoflavones). The concen-
are an alternative to mass spectrometric methods. These trations of urinary daidzein and genistein observed
immunoassays were used to test urine and plasma during intervention demonstrated good compliance,
samples from individuals in a dietary intervention trial and a corresponding increase in serum daidzein and
aimed at determining the efficacy of dietary isoflavones genistein confirmed bioavailability of the isoflavone-
in reducing the risk of coronary heart disease in post- rich foods; 33 of the 117 volunteers (28.2%) were classi-
menopausal women. fied as equol producers on the basis of their urinary
Methods: We established murine monoclonal TR-FIA equol concentration (>936 nmol/L), and significant dif-
methods for daidzein, genistein, and equol. These as- ferences in the numbers of equol producers were ob-
says could be performed manually or adapted to an served between Berlin and the 3 other European cohorts
automated analyzer for high throughput and increased studied.
accuracy. Analysis of urine was conducted on nonex- Conclusions: The validated monoclonal TR-FIA meth-
tracted samples. Blood analysis was performed on non- ods are applicable for use in large-scale human phy-
extracted samples for daidzein, whereas genistein and toestrogen intervention studies and can be used to
equol required diethyl-ether extraction. monitor compliance, demonstrate bioavailability, and
Results: Comparison of monoclonal TR-FIA, commer- assess equol producer status.
cial polyclonal antibody– based TR-FIA, and gas chro- © 2007 American Association for Clinical Chemistry
matography–mass spectrometry showed correlations (r,
Interest in the role of phytoestrogens in human health has
increased during the past decade, with numerous dietary
1
interventions and epidemiological studies investigating
Unilever Corporate Research, Sharnbrook, United Kingdom.
2
Department of Biological Regulation, The Weizmann Institute of Science, the potential health benefits of these compounds, partic-
Rehovot, Israel. ularly in relation to cardiovascular and bone health and
3
Centre for Veterinary Science, Department of Veterinary Medicine, cancer prevention (1–3 ). A recent consensus based on the
University of Cambridge, Cambridge, United Kingdom.
4
Folkhalsan Research Center, Institute for Preventive Medicine, Nutrition
available data concluded that although the use of soy
and Cancer and Division of Clinical Chemistry, University of Helsinki, products and soy isoflavones may be beneficial for bone,
Helsinki, Finland. cardiovascular risk, and hot flushes, the benefits demon-
5
Department of Human Nutrition, The Royal Veterinary and Agricultural strated in available studies are subtle. Adequately pow-
University, Frederiksberg, Denmark.
6
School of Medicine, Health Policy & Practice, University of East Anglia,
ered human intervention studies are needed to demon-
Norwich, United Kingdom. strate the benefits of encapsulated isoflavones or
* Address correspondence to this author at: Unilever Corporate Research, isoflavone-fortified foods on clinical outcomes, such as
Colworth Park, Sharnbrook, Bedfordshire MK44 1LQ, United Kingdom. Fax
incidence of heart disease and bone fractures (2 ). Large-
44-0-1234-248010; e-mail Duncan.Talbot@Unilever.com.
Received June 20, 2006; accepted January 23, 2007. scale studies are hampered by the lack of availability of
Previously published online at DOI: 10.1373/clinchem.2006.075077 rapid, validated assays for assessment of these com-
748
pounds; current validated assays rely on expensive and Materials and Methods
laborious methods such as gas chromatography–mass preparation of monoclonal antibodies
spectrometry (GC-MS),7 which is considered the reference Preparation of the daidzein monoclonal antibody has
method (4, 5 ) for assessment of these compounds and previously been described (20 ). Genistein– bovine serum
offers good sensitivity and excellent specificity, although albumin (BSA) immunoconjugates were synthesized by
HPLC (generally with coulometric detection) has gained addition of 2.2 g/L genistein to 0.1 mol/L carbonate
acceptance because of reduction in the number of re- buffer (pH 9.6) and solubilized with 0.2 mol/L NaOH.
quired purification steps (6, 7 ). Liquid chromatography– Genistein was mixed with 10 g/L BSA in carbonate buffer
mass spectrometry is an excellent method, but the high before addition of formaldehyde (66 ␮mol/L in MES/
costs of instrumentation limit its use in laboratories. These NaCl) and the reaction stirred at 38.2 °C for 24 h. Reaction
products were purified through a PD10 column. The
methods require sample extraction and have relatively
conjugate was eluted in phosphate buffered saline (7
low throughput.
mmol/L Na2HPO4, 3.7 mmol/L H2NaPO4, and 0.15
Immunoassays offer the potential advantages of small
mol/L NaCl, pH 7.4), supplemented with 0.05% sodium
sample volume, direct sample testing, speed, simplicity,
azide, and analyzed by size-exclusion chromatography at
low cost, throughput, and availability of automated the carrier protein maximum wavelength (280 nm) and
equipment. The development of rapid immunoassays for the hapten maximum wavelength (262 nm). The ratio of
assessment of both parent compounds and intestinal incorporation was 18.7:1 (genistein:BSA). A genistein-
metabolites has allowed biological effects in relation to ovalbumin screening conjugate was prepared in the same
isoflavone metabolism to be addressed in intervention manner as the genistein-BSA immunoconjugate, yielding
trials, an area of particular interest because of increasing an incorporation ratio of 1:8 to 1:12. Carboxymethyl equol
evidence that the clinical efficacy of isoflavones in hu- immunoconjugate was synthesized by addition of 23.4 mg
mans depends on the production of a gut metabolite, equol to a round bottomed flask containing 14 mL HPLC-
equol. Equol is a microbial metabolite of daidzein (8 ), and grade dry n-propanol, followed by addition of 100 mg dry
its production in humans is highly variable; only 30%– sodium that was cut into small pieces. After dissolution of
50% of any given population group can produce equol sodium, 150 mg dry bromoacetic acid was added. The
after ingestion of soy foods (9 –11 ). The capacity to reaction was refluxed for 8.5 h and cooled, and then 1 mL
produce equol is associated with circulating reproductive water was added and the precipitate was collected. The
hormones (9, 12 ), inversely associated with mammo- precipitate was extracted into methanol (0.6 mL), insolu-
graphic density (13, 14 ), and positively associated with ble products were discarded, and the filtrate was evapo-
bone mineral density in response to isoflavone interven- rated. The residue was chromatographed on silica gel 60.
tion (15, 16 ). Equol production may enhance the action of Elution with methanol:CHCl3:HOAc (89.7:10:0.3) yielded
isoflavones because it has a lower affinity for serum the desired monoaddition product of equol (1.5 mg), with
proteins and a greater affinity for estrogen receptors and an RF of 0.2 in the solvent system. In the same system
equol showed an RF of 0.9. The electron mass spectrum of
exhibits antioxidant activity greater than that of genistein
the carboxy derivative of equol showed a peak at 298.9
and daidzein (17–19 )
(M-1), corresponding to a compound with a formula of
As part of the European Union Isoheart project to
C17O5H15.
investigate the potential cardioprotective effects of iso-
Carboxymethyl equol was conjugated to KLH via a
lated isoflavones in healthy postmenopausal women, we 2-step reaction. The reactive equol derivative (2 mg) was
developed and validated methods for the assessment of dissolved in 200 ␮L dry dioxane and 1 mg N-hydroxysuc-
the isoflavone metabolites daidzein, genistein, and equol. cinimide, and then 2 mg carbodiimide (2 mg) was added
When the Isoheart project commenced, commercial phy- to the reaction mixture. After an overnight reaction at
toestrogen assays were not available, but an appropriate room temperature, urea was formed, indicating that an
monoclonal antibody against daidzein was available (20 ). active ester of equol was formed. The active ester of equol
We therefore prepared monoclonal antibodies to 2 other (0.6 mg) was then used in the next step without further
key isoflavone metabolites, genistein and equol. Our aim purification. KLH (5 mg) was dialyzed twice against
was to use validated phytoestrogen immunoassays to phosphate-buffered saline (7mmol/L Na2HPO4, 3.7
demonstrate bioavailability and monitor compliance mmol/L H2NaPO4, 0.15 mol/L NaCl, pH 7.4, pH 7.4). To
within the Isoheart trial. this solution (0.9 mL) we added 50 ␮L of 1 mol/L sodium
phosphate, followed by drop-wise addition of 100 ␮L of
the active ester. The reaction mixture was stirred for 2 h at
room temperature and then dialyzed against phosphate-
buffered saline and stored at ⫺20 °C until use.
7
Nonstandard abbreviations: GC-MS, gas chromatography–mass spec-
We immunized 6- to 8-week-old female Balb/c mice
trometry; BSA, bovine serum albumin; DELFIA, dissociation-enhanced lan-
thanide fluorescence immunoassay; TR-FIA, time-resolved fluorescence im- subcutaneously and intraperitoneally with ⬃50 ␮g of each
munoassay; CI, confidence interval. immunogen. Booster injections were given on day 14, and
750 Talbot et al.: Monoclonal TR-FIAs for Phytoestrogens
sera from these mice were tested on day 18. Best respond- buffer (250 mL; Cat. No. 1244 –111), wash buffer (250 mL;
ers were boosted at 2-week intervals, and 3 days before Cat. No. 1244 –114), and enhancement solution (250 mL;
splenectomy were given a final (intravenous and intra- Cat No. 1244 –105).
peritoneal for equol and genistein, respectively) boost We added 25 ␮L duplicate samples of urine (diluted in
with immunogen alone. Throughout the immunization assay buffer, 1/25 daidzein, 1/5 genistein, or 1/10 equol),
schedule, the mice were treated humanely in accordance serum (daidzein), or extracted serum (genistein and
with strict U.K. Government Home Office regulations. equol), calibrator (Daidzein, Sigma D7802; Genistein,
Ethical review board meetings were held regularly, and Sigma G6649; and Equol, Indofine Chemical Co. 02-1268),
the animal facilities were visited on a quarterly basis by and QC samples to appropriate wells of an antimouse
U.K. Government Home Office inspectors. plate. QC material was prepared by diluting stock solu-
The mouse spleens were perfused with culture me- tions of 3 known amounts of analyte in phosphate-
dium, and the resulting lymphocyte suspension (⬃107 buffered saline ⫹ 0.1% ovalbumin. These concentrations
cells per spleen) was combined with SP2/0 myeloma cells
in a 1:1 ratio. After centrifugation at 200g for 5 min, the
supernatant was discarded and the pellet was loosened.
To this mixed pellet we added 1 mL 45% polyethylene Table 1. Assay performance.
glycol 3000 (Fluka AG). The cells were centrifuged gently A. Intraassay CV was determined by analyzing 25
for 3 min, and after a total contact time of 8 min, the replicates of standards and determining the concentration
polyethylene glycol was gradually diluted and removed using Multicalc.
Concentration, nmol/L Intraassay CV, %
by addition of culture medium and further centrifugation.
Daidzein assay
The cells were finally suspended in culture medium
31.4 4.9
containing hypoxanthine and azaserine and dispensed
157.2 3.0
into 10 48-well plates. Hybridomas were visible at day 3, 786 2.2
and a culture medium change was performed on day 7. Urinary genistein assay
On day 10 supernatants from all fusion wells were 231 13.4
screened by enzyme immunoassay using genistein- 925 8.8
ovalbumin or equol-ovalbumin at 5 mg/L, with detection 3700 3.9
using antimouse antibody conjugated to alkaline phos- 14 800 5.3
phatase. Confirmation of positive clones was performed Serum genistein assay
by inhibition assay using free genistein or equol, with 92.5 8.3
doubling dilutions starting at 20 mg/L. Clones of interest 462.5 4.0
were recloned to ensure monoclonality and then grown to 2313 2.6
a volume of 8 L in RPMI, supplemented with 5% donor Equol
horse serum using stirred fermentors or roller bottles, 19.5 6.7
97.5 2.9
centrifuged and resuspended in serum-free medium,
390 5.0
grown for 2–3 days, and allowed to die. After filtration
1950 4.6
and concentration, purification was by protein A chroma-
tography. The concentration of the purified antibody was B. Interassay CV was determined by analyzing QC samples
determined by measuring the absorbance at 280 nm in duplicate, 2 runs per day, for 20 days (n ⴝ 40) and
determining the concentration using Multicalc.
[concentration (g/L) ⫽ absorbance 280 nm/1.4]. The
Concentration, nmol/L Interassay CV, %
concentrations of the selected purified antibodies were as
Daidzein assay
follows: daidzein, 0.78 g/L; equol, 6.8 g/L, serum 15.7 8.6
genistein, 1.01 g/L; and urinary genistein, 2.98 g/L. 94.3 4.1
Candidate antibodies were then evaluated in the dissoci- 2358 6.8
ation-enhanced lanthanide fluorescence immunoassay Urinary genistein assay
(DELFIA) immunoassay format. 1480 15.3
7400 15.8
delfia immunoassay methods 22 200 18.5
The Perkin-Elmer DELFIA time-resolved fluorescence im- Serum genistein assay
munoassay (TR-FIA) was selected because of the sensitiv- 55.5 11.7
ity, specificity, and wide linear range offered by this 1480 6.8
immunoassay compared with conventional enzyme im- 7400 15.6
munoassays. The DELFIA reagents and instruments were Equol assay
obtained from Perkin-Elmer Life Sciences. These included 9.75 9.5
58.5 6.7
yellow, low-fluorescence antimouse IgG– coated plates
975 9.2
with nonremovable wells (Cat. No. AAAND-0003), assay
were selected to appropriately represent the range of To assess the limit of detection and intraassay variabil-
analyte likely to be encountered with each assay. These ity, we included a standard curve comprising 6 calibrators
samples were termed QC low, medium, and high, respec- and 3 QC samples in each assay (in duplicate). Twenty-
tively, and the low, medium, and high concentrations five replicates of each calibrator were run as samples on 3
were as follows: for daidzein, 15.75, 63.6, and 2340 occasions, and the mean concentration of each was deter-
nmol/L (4, 24, and 600 ␮g/L); for equol, 9.25, 55.5, and mined (intraassay) using Multicalc (Perkin-Elmer). The
925 nmol/L (2.5, 15, and 250 ␮g/L); for serum genistein, limit of detection of the assays was defined as the con-
58.5, 1560, and 7800 nmol/L (15, 400, and 2000 ␮g/L); for centration that was 3 SDs above the mean of the zero
urinary genistein, 1560, 7800, and 23 400 nmol/L (400, standards, giving a ⬎99% confidence value. The limit of
2000, and 6000 ␮g/L). Standard dilutions were prepared quantification for each assay was determined as the
from 1 g/L stocks in 96% ethanol in 0.1 mol/L phosphate- concentration that was 10 SDs above the mean of the zero
buffered saline, pH 7.2, containing 0.1% ovalbumin standards. To determine interassay CVs, low, medium,
(Sigma A5378), and stored frozen at ⫺20 °C. Six standards and high QC samples for each assay were analyzed in
were used per assay, including a zero. The highest and duplicate, on 2 analytical runs per day, for 20 days (n ⫽
lowest standards were 3930 nmol/L (1000 ␮g/L) and 6.2 40), and the concentrations determined with Multicalc.
nmol/L (1.6 ␮g/L) for daidzein; 1950 nmol/L (500 ␮g/L) Assay specificity was assessed by measuring cross-reac-
and 3.9 nmol/L (1.0 ␮g/L) for equol; 1156.25 nmol/L tivity with a panel of relevant phytoestrogens dissolved in
(312.5 ␮g/L) and 18.5 mol/L (5.0 ␮g/L) for serum assay buffer. Cross-reactivity was defined as 100(X/Y)%,
where X is the mass of homologous phytoestrogen and Y
genistein; and 44 400 nmol/L (12 000 ␮g/L) and 231.5
is the mass of heterologous phytoestrogen required to
nmol/L (62.5 ␮g/L) for urinary genistein. To each well we
produce 50% inhibition of binding of the Europium-
added 100 ␮L tracer [daidzein 1/24 000 (daidzein-eu-
labeled phytoestrogen.
ropium) genistein 1/4000 (genistein-ovalbumin-eu-
ropium), and equol, 1/10 000 (equol-ovalbumin-eu-
extraction of blood samples before analysis by
ropium)] diluted in assay buffer incorporating 0.4 U/L
delfia genistein and equol immunoassay
B-glucuronidase (Sigma G7396) and incubated the plates The extraction method was the diethyl ether method
with shaking for 60 min at room temperature. The en- described previously (21 ), and was used in the commer-
zyme treatment is an integral part of the assay protocol. cial (Labmaster Ltd.) DELFIA immunoassay.
We then added 100 ␮L antibody (daidzein, 4E4 25 ␮g/L;
urinary genistein, 6547:3 65 ␮g/L; serum genistein, 6066:1 gc-ms and labmaster tr-fia determinations
200 ␮g/L; or equol, 6588:1 20 ␮g/L) diluted in assay GC-MS and Labmaster TR-FIA analyses were performed
buffer, to each well, after which the plate was incubated, at the laboratory of Herman Adlercreutz according to
with shaking, for an additional hour at room temperature. previously published methods (4, 5, 21–23 ).
The plate was washed 6 times with 400 ␮L of wash buffer
(wash concentrate diluted 1/25 in MilliQ water), 200 ␮L the isoheart study and blood/urine samples
of enhancement solution was added, and the plate was for phytoestrogen analysis
shaken for a final 5 min before reading of the fluorescence. The Isoheart intervention study was a placebo-controlled,
Assays were performed manually using a Perkin-Elmer 2 ⫻ 8-week randomized crossover, with an 8-week wash-
DELFIA plate shaker (1296-003) and washer (1296-026) out period. Study participants consumed 2 cereal bars per
and were read on a Victor2 1420 multilabel counter or day, which provided 50 mg/day isoflavones (genistein:
automated for high-throughput testing on an daidzein ratio, 2:1), or the same quantity of cereal bars
AutoDELFIA. containing no added isoflavones (placebo). Twenty-four-
Table 2. Specificity of the monoclonal antibody based DELFIA phytoestrogen assays.a

Cross-reactivity, Cross-reactivity, urinary Cross-reactivity, serum Cross-reactivity, equol
Cross-reactivity with daidzein (clone 4E4), % genistein (clone 6547:3), % genistein (clone 6066:1), % (clone 6588:1), %
Apigenin ⬍0.02 ⬍2 ⬍0.2 0.3
Daidzein 100 12 18.8 ⬍0.04
Dihydrodaidzein 9.9 ⬍2 0.8 2
17B-Estradiol ⬍0.02 ⬍2 ⬍0.2 ⬍0.04
Equol ⬍0.02 ⬍2 ⬍0.2 100
Genistein 1.6 100 100 0.3
Glycitein ⬍0.02 ⬍2 ⬍0.2 ⬍0.04
Luteolin ⬍0.02 ⬍2 ⬍0.2 ⬍0.04
Quercitin ⬍0.02 ⬍2 ⬍0.2 ⬍0.04
a
The listed compounds were tested in assay buffer for their cross-reactivity with the 4 selected monoclonal antibodies as described in Materials and Methods.
Fig. 1. Correlation of Unilever TR-FIAs

for the detection of urinary daidzein,
genistein, and equol with GC-MS and
Labmaster TR-FIA using Deming re-
gression analysis as described in Ma-
terials and Methods.
Slopes and intercepts with 95% CIs are
given. Dotted line, 1:1 correlation; solid
line, best fit.
hour urine samples were collected at the beginning and SAS for graphical presentation. Deming regression was
end of each arm of the study, and serum samples at weeks used to fit straight lines to the data because, in contrast to
0, 4, and 8 of each arm. The study design and protocol ordinary linear regression, Deming regression realistically
were approved by the University of Reading, INRAN assumes that the measurements by both methods are
Rome, DIFE Berlin, and the Municipal Ethical Committee subject to random errors. Linnet weighting was applied
of Copenhagen and Frederiksber. After receiving a de- because it takes into account nonconstant analytical vari-
tailed explanation of the study procedure, each partici- ances for both methods. Because of the large proportion of
pant gave written informed consent. nonproducers in these data we did not use weighting for
the urinary equol comparison with GC-MS. We used a
statistical analysis jackknife algorithm to calculate 95% confidence intervals
Method Validator (Philip Maquis, 1999; available at (CIs) for slopes and intercepts. Comparison between
http://www.MultiQC.com) was used for analysis and monoclonal TR-FIA and Labmaster TR-FIA was made
Fig. 2. Correlation of Unilever TR-FIAs

for detection of serum daidzein,
genistein, and equol with GC-MS and
Labmaster TR-FIA using Deming re-
gression analysis as described in Ma-
terials and Methods.
Slopes and intercepts with 95% CIs are
given. Equol GC-MS values are omitted due
to small numbers being available for analy-
sis. Dotted line, 1:1 correlation; solid line,
best fit.
with noncorrected Labmaster TR-FIA data (raw assay nmol/L (equol). The intra- and interassay CVs are sum-
data before application of a correction factor to GC-MS marized in Table 1.
units). Assay specificity is summarized in Table 2, which
shows the cross-reactivity of each monoclonal phytoestro-
Results gen antibody for a panel of similar compounds. Fig. 1
assay performance shows the correlation between the established urinary
The limits of detection of the TR-FIA assays were deter- monoclonal TR-FIA assays, commercial (Labmaster Ltd.)
mined to be 3.9 nmol/L (daidzein), 88.8 nmol/L (urinary polyclonal TR-FIAs, and gold-standard GC-MS determi-
genistein), 8.7 nmol/L (serum genistein), and 2.2 nmol/L nations. Fig. 2 shows the correlation between the assay
(equol). The limits of quantification (LOQ) of the TR-FIA methods for comparison of nonextracted serum samples
assays were 11.8 nmol/L (daidzein), 319.3 nmol/L (uri- in the daidzein monoclonal TR-FIA assay and extracted
nary genistein), 41.1 nmol/L (serum genistein), and 4.3 serum samples in the monoclonal serum genistein and
measured by monoclonal TR-FIA, for each volunteer in

the Isoheart intervention study. Equol producers were
defined as having a urinary equol concentration ⬎960
nmol/L (21 ). Volunteers with low urinary equol concen-
trations also showed low serum equol concentrations
(⬍15 nmol/L). On the basis of their urinary equol con-
centration, 33 of the 117 volunteers (28.2%) were classified
as equol producers. The Fisher exact test revealed a
general imbalance between the proportions of equol pro-
ducers in the different centers (P ⫽ 0.0057). The most
obvious cause of this is the difference between Berlin (50%
equol producers) and the other centers (19% equol pro-
ducers, P ⫽ 0.0014). Although Rome (11% equol produc-
ers) appears to be different from Reading (26% equol
producers) and Copenhagen (21% equol producers), no
evidence for imbalance is observed between the 3 centers
overall (P ⫽ 0.33), and there is no evidence for imbalance
between Rome and Reading/Copenhagen combined (24%
equol producers, P ⫽ 0.24; Table 3).
Fig. 3. Equol producer status of Isoheart volunteers. Discussion

We analyzed 8-week intervention urine and serum samples by AutoDELFIA assay Current methods for the measurement of daidzein,
as described in Materials and Methods. Producer status defined as urinary equol
concentration ⬎960 nmol/L (solid vertical line) and serum equol concentration genistein, and equol in clinical trial samples rely predom-
⬎40 nmol/L (solid horizontal line). inantly on GC-MS, a labor-intensive and expensive tech-
nique. To increase throughput, we raised mouse mono-
equol assays with Labmaster and GC-MS results. Unfor- clonal antibodies against these compounds, developed
tunately, insufficient samples tested by GC-MS had high high-throughput AutoDELFIA assays, and validated
enough equol concentrations for a robust correlation to be them against GC-MS and a commercially available immu-
established. noassay reagent set (Labmaster). We subsequently used
the assays to measure compliance, bioavailability, and
isoheart study: clinical analysis equol production in the Isoheart trial.
The validated assays were used to test urine and serum The monoclonal TR-FIA assays could be performed
samples from the Isoheart intervention study. Mean (SE) either manually or on the AutoDELFIA which, as well as
urinary daidzein and genistein concentrations for the 117 offering improved precision, facilitated automated high-
volunteers on the study were 393 (45) nmol/L and 731 throughput testing of up to 12 plates (39 samples/plate)
(52) nmol/L, respectively, during the placebo phase and of bar-coded samples at a time. Alternatively, 4 plates
12038 (714) nmol/L and 13399 (631) nmol/L, respectively, (4 ⫻ 39 samples) could be tested for daidzein, genistein,
during the supplemented phase of the trial. Mean (SE) and equol in 1 run. Because of the lower phytoestrogen
serum daidzein and genistein concentrations rose from 38 concentrations measured in blood than in urine, serum
(1) nmol/L and 9 (2) nmol/L during the placebo phase to (daidzein), or extracted serum/plasma (genistein and
139 (5) nmol/L and 283 (21) nmol/L, respectively, during equol), undiluted samples were tested. The AutoDELFIA,
the intervention arm of the study. however, enabled preparation of urinary dilutions from 1⁄2
Fig. 3 shows plots of week 8 urinary equol concentra- to 1/100 depending on levels expected/observed in par-
tion against week 8 extracted serum equol concentration, ticular studies. The selected urinary genistein antibody
Table 3. Equol producer status by study center.a

Isoheart Study Center, frequency (%)
Total number of producers
Equol producer status Berlin, Germany Copenhagen, Denmark Reading, U.K. Rome, Italy or nonproducers
Nonproducer 17 (50%) 22 (78.6%) 20 (74.1%) 25 (89.3%) 84 (71.8%)
Producer 17 (50%)b 6 (21.4%) 7 (25.9%) 3 (10.7%) 33 (28.2%)
Total number of study participants 34 28 27 28 117
analyzed per study center
a
Urinary equol concentrations were analyzed by AutoDELFIA assay as described in Materials and Methods. Equol producer defined as urine equol concentration
⬎960 nmol/L. Upper number per cell represents number of study participants defined as producer or nonproducer per study center. Percentages are proportion of
producers or nonproducers per study center.
b
Significantly different proportion of equol producers compared with other study centers (P ⫽ 0.0057).
(6547:3), suitable for assaying intervention study samples pared here were across the intervention study range,
at a single urine dilution, had a broader assay range but facilitating robust correction across this concentration
was less sensitive than the selected monoclonal antibody range.
for blood genistein measurement (6066:1). The daidzein Urinary equol concentrations (in a 24-h collection)
(4E4) and equol (6588:1) monoclonal antibodies were were considered to be a better indicator of producer status
sensitive, exhibited broad assay range, and were used for than serum equol concentrations, because the timings for
both urine and blood testing. In the Isoheart intervention cereal bar consumption and bleeding times were not
study, genistein analysis was performed on 1 of 5 urine standardized, and analysis required extraction, with as-
samples, whereas urinary daidzein and equol analysis sociated losses. The plot of urinary equol concentrations
were performed on 1 of 25 and 1 of 10 urine samples, clearly splits the volunteers into “producers” and “non-
respectively. producers.” If the urinary equol concentrations were
By incorporating B-glucuronidase in the DELFIA assay converted to GC-MS units, only 6 borderline results (3
buffer (pH 7), this enzyme treatment could be performed Berlin, 1 Copenhagen, and 2 Reading) would then fall
as an integral part of the urinary daidzein, genistein, below the 936 nmol/L cutoff and be considered as non-
equol, and (nonextracted) serum daidzein assays. For producers. A statistically significant difference was noted
blood genistein and equol, B-glucuronidase and sulfatase in the number of equol producers in Berlin compared
treatments were performed in pH 5 acetate buffer, an with the other study centers (P ⫽ 0.0057). This observa-
appropriate pH for both enzymes, before diethyl ether tion is worthy of further investigation and may reflect
extraction (21, 22 ). differences in dietary habits or antibiotic treatment (24 ).
Assay validation showed correlations ranging from r ⫽ Work is ongoing to test this hypothesis by analysis of
0.911 to r ⫽ 0.994 across the range of urine and serum completed food questionnaires from the study volunteers.
intervention study samples with GC-MS and Labmaster
TR-FIA. As previously observed (21, 23 ), higher urinary
phytoestrogen concentrations were obtained by TR-FIA The development of the genistein and equol DELFIA
than GC-MS. This result could be due to cross-reactivity assays, assay validation, and testing of the bioavailability
of the antibodies, although their specificity appeared and intervention study samples was carried out within
generally good (Table 2). Losses due to the number of the EU project Isoheart no. QLK1-CT-2001-00221. We
procedures involved in the GC-MS method (e.g., purifi- particularly acknowledge the other members of the Iso-
cation and derivatization, whereas TR-FIA used untreated heart consortium, especially those involved in running
urine) may also have contributed to losses, despite the the bioavailability and intervention studies at the 4 study
application of a correction based on the yield of a radio- centers (DIFE, Berlin, KVL, Copenhagen, Reading Univer-
label. Correlations between monoclonal TR-FIA and Lab- sity, U.K., and INRAN, Rome). Reagent availability: Re-
master TR-FIA were good for all 3 analytes (Figs. 1 and 2). searchers interested in obtaining the genistein and equol
The slope for equol was close to 1 for urine and serum. antibodies should contact the corresponding author. Re-
Daidzein, and to a greater extent genistein, however, were searchers interested in the daidzein antibody should
overreported in urine compared with the Labmaster TR- contact F.K.
FIA. The genistein monoclonal TR-FIA has a greater
clinical range and higher limit of detection than the References
Labmaster TR-FIA assay and thus was selected for use in 1. Setchell KDR, Cassidy A. Dietary isoflavones: biological effects
compliance monitoring. These characteristics may have and relevance to human health. J Nutr 1999;129:758S– 67S.
led to a greater difference in the assay comparison results 2. Cassidy A, Albertazzi P, Lise Nielsen I, Hall W, Williamson G,
at higher urinary genistein concentrations. The slope of Tetens I, et al. Critical review of health effects of soyabean
isoflavones in postmenopausal women. Proc Nutr Soc 2006;65:
the correlation between monoclonal TR-FIA and Labmas-
76 –92.
ter TR-FIA for the serum genistein assay suggests under-
3. Adlercreutz H. Phyto-estrogens and cancer. Lancet Oncol 2002;3:
recovery, which may be attributable to the extraction 364 –73.
procedure resulting in losses in the monoclonal serum 4. Adlercreutz H, Fotsis T, Bannwart C, Wahala K, Brunow G, Hase T.
genistein assay, although further work is required to Isotope dilution gas chromatographic-mass spectrophotometric
confirm this. Comparison of the monoclonal and poly- method for determination of lignans and isoflavones in human
clonal TR-FIA results for extracted serum equol, however, urine, including identification of genistein. Clin Chim Acta 1991;
showed equivalence when the same extracted samples 199:263– 8.
were used as for serum genistein. 5. Adlercreutz H, Fotsis T, Lampe J, Wahala K, Makela T, Brunow G,
The established correlation equations could be used to et al. Quantitative determination of lignans and isoflavanoids in
plasma of omnivores and vegetarian women by isotope dilution
correct monoclonal TR-FIA values to GC-MS units, as gas chromatography-mass spectrophotometry. Scand J Clin Lab
previously described in reports on the use of the Labmas- Invest Suppl 1993;53:5–18.
ter polyclonal TR-FIA (21–23 ). Whereas the Labmaster 6. Nurmi T, Adlercreutz H. Sensitive HPLC method for profiling
correction factors were established with low urinary daid- plasma phytoestrogens using coulometric electrode array detec-
zein and genistein concentrations (23 ), the samples com- tion. Anal Biochem 1999;274:110 –7.
7. Nurmi T, Voutilainen S, Nyyssonen K, Adlercreutz H, Salonen JT. 16. Fujioka M, Uehara M, Wu J, Adlercreutz H, Suzuki K, Kanazawa K.
Liquid chromatography method for plant and mammalian lignans Equol, a metabolite of daidzein, inhibits bone loss in ovariecto-
in human urine. J Chromatogr B Analyt Technol Biomed Life Sci mised mice. J Nutr 2004;134:2623–7.
2003;798:101–10. 17. Shutt DA, Cox RI. Steroid and phytoestrogen binding to sheep
8. Setchell KDR, Clerici C, Lephart ED, Cole SJ, Heenan C, Castellani uterine receptors in vitro. J Endocrinol 1972;52:299 –303.
D, et al. S-equol, a potent ligand for estrogen receptor beta, is the
exclusive enantiomeric form of the soy isoflavone metabolite 18. Hodgson JM, Croft KD, Puddey IB, Mori TA, Beilin LJ. Soyabean
produced by human intestinal bacterial flora. Am J Clin Nutr isoflavonoids and their metabolic products inhibit in vitro lipopro-
2005;81:1072–9. tein oxidation in serum. J Nutr Biochem 1996;7:664 –9.
9. Cassidy A, Bingham S, Setchell KDR. Biological effects of a diet 19. Arora A Nair MG, Strasburg GM. Antioxidant activities of isofla-
rich in isoflavones on the menstrual cycle of premenopausal vones and their biological metabolites in a liposomal system. Arch
women. SAm J Clin Nutr 1994;60:333– 40. Biochem Biophys 1998;356:133– 41.
10. Lampe JW, Li SS, Potter JD, King IB. Serum beta-glucuronidase 20. Kohen F, Lichter S, Gayer B, DeBoever J, Lu LJW. The measure-
activity is inversely associated with plant-food intakes in humans. ment of the isoflavone daidzein by time resolved fluorescent
J Nutr 2002;132:1341– 4.
immunoassay: a method for assessment of dietary soya expo-
11. Rowland I, Faughnan M, Hoey L, Wahala K, Williamson G, Cassidy
sure. J Steroid Biochem Mol Biol 1998;64:217–22.
A. Bioavailability of phyto-estrogens. Brit J Nutr 2003;89:S45–58.
12. Cassidy A, Bingham S, Setchell KDR. Biological effects of isofla- 21. Brouwers E, L’Homme R, Al-Maharik N, Lapcik O, Hampl R,
vones in young women: importance of the chemical composition Wahala K, et al. Time-resolved fluoroimmunoassay for equol in
of soyabean products. Brit J Nutr 1995;74:587– 601. plasma and urine. J Steroid Biochem Mol Biol 2003;84:577– 88.
13. Ozasa K, Nakao M, Watanabe Y, Hayashi K, Miki T, Mikami K. Serum 22. Wang GJ, Lapcik O, Hampl R, Uehara M, Al-Maharik N, Stumpf K,
phytoestrogens and prostate cancer risk in a nested case-control et al. Time-resolved fluoroimmunoassay of plasma daidzein and
study among Japanese men. Cancer Sci 2004;95:65–71. genistein Steroids 2000;65:339 – 48.
14. Akaza H, Miyanaga N, Takashima N, Naito S, Hirao Y, Tsukamoto
23. Uehara M, Lapcik O, Hampl R, Al-Maharik N, Makela T, Wahala K,
T. Comparisons of percent equol producers between prostate
et al. Rapid analysis of phytoestrogens in human urine by time-
cancer patients and controls: case-controlled studies of isofla-
resolved fluoroimmunoassay; J Steroid Biochem Mol Biol 2000;
vones in Japanese, Korean and American residents. Jpn J Clin
72:273– 82.
Oncol 2004;34:86 –9.
15. Setchell KDR, Brown NM, Lydeking-Olsen E. The clinical impor- 24. Axelson M, Setchell KD. The excretion of lignans in rats: evidence
tance of the metabolite equol: a clue to the effectiveness of soy for an intestinal bacterial source for this new group of compounds.
and its isoflavones. J Nutr 2002;132:3577– 84. FEBS Lett 1981;123:337– 42.
757–765 (2007) Clinical Immunology
Quantification of Human ␤-Defensin-2 and -3 in

Body Fluids: Application for Studies of Innate
Immunity
Santosh K. Ghosh,1 Thomas A. Gerken,2 Keith M. Schneider,1 Zhimin Feng,1
Thomas S. McCormick,3 and Aaron Weinberg1*
Background: Human ␤-defensins (hBDs) are epithelial ing masking effects of anionic molecules. This approach
cell-derived antimicrobial and immunoregulatory cat- may therefore be applicable for quantifying these pep-
ionic peptides. Our objective was to establish an analyt- tides in health and disease.
ical tool to quantify inducible hBD-2 and -3 in body © 2007 American Association for Clinical Chemistry
fluids.
Methods: We developed sandwich ELISAs using com- Antimicrobial peptides have been identified as key com-
mercially available capture and detection antibodies ponents in innate host defense and as important contrib-
and determined optimal assay conditions (with 250 utors in maintaining health at mucosal barriers. Human
mmol/L CaCl2) to overcome masking by endogenous ␤-defensins, a family of epithelial cell-derived cationic
components of body fluids. We used recombinant hBD peptides (4 –5 kDa), exhibit both antimicrobial and immu-
as calibrators and for recovery testing. nomodulatory properties (1–3 ). Four human ␤-defensins
Results: hBD-2 and -3 detection limits were ⬃75 ng/L have been identified. Human ␤-defensin 1 (hBD-1)4 is
and ⬃3 ␮g/L, respectively. Mean (SD range) values in constitutively expressed, whereas hBD-2 and -3 are induc-
saliva samples from healthy donors (n ⴝ 60) were ible (2, 4, 5 ). All 3 peptides can be isolated from mucosal
9.5 (1.2–21) ␮g/L for hBD-2 and 326 (50 –931) ␮g/L for sites of the body. An hBD-4 transcript has been described
hBD-3. We did not detect hBD-3 in suction blister fluid (6 ), but the peptide has not yet been isolated.
(BF; n ⴝ 10) or bronchoalveolar lavage (BAL; n ⴝ 5) from hBDs have demonstrated activity against gram-posi-
healthy participants. We detected low hBD-2 peptide tive and -negative bacteria, mycobacteria, fungi, and
concentrations in BF and BAL, 0.16 (0.03– 0.32) and certain enveloped viruses at low micromolar concentra-
0.04 (0 – 0.049) ␮g/g total protein, respectively. We ob- tions (7, 8 ). We recently showed that hBDs have anti-
served no correlation of hBD-2 in BF and saliva or BAL retroviral activity by inhibiting HIV-1 infectivity of im-
and saliva from the same person. In vaginal swabs from munocompetent cells (9, 10 ). In addition, hBDs can
healthy women (n ⴝ 2), mean hBD-2 and -3 concentra- enhance adaptive immunity by acting as adjuvant and
tions were 3.42 and 103 ␮g/g total protein, respectively. chemoattracting T cells, immature dendritic cells, B cells,
Cervicovaginal lavage from the same women contained neutrophils, and macrophages (11–13 ). With new infor-
mean concentrations of 1.46 and 55.5 ␮g/g total protein. mation emerging about these pluripotent peptides and
Conclusion: These ELISA assays can measure inducible their role in mucosal protection, diagnostic tools to quan-
hBD peptide concentrations in body fluids by overcom- tify hBD-2 and -3 in body fluids and tissues are essential
to better associate hBD expression with disease predispo-
sition and progression.
1
Department of Biological Sciences, Case School of Dental Medicine; Previous methods of quantifying hBDs in body fluids
2
Departments of Pediatrics and Biochemistry, Case Western Reserve involved acid extraction followed by slot blot assays
University, Cleveland, OH.
3
Department of Dermatology, University Hospitals of Cleveland and Case
(14, 15 ), semiquantitative Western analysis (16 –18 ), or
Western Reserve University, Cleveland, OH.
* Address correspondence to this author at: School of Dental Medicine,
Case Western Reserve University, Cleveland, OH 44106-4905. Fax 216-368-
0145; e-mail aaron.weinberg@case.edu. 4
Nonstandard abbreviations: hBD, human ␤ -defensin; BF, blister fluid;
Received October 9, 2006; accepted January 16, 2007. BAL, bronchoalveolar lavage; CVL, cervicovaginal lavage; OSM, ovine sub-
Previously published online at DOI: 10.1373/clinchem.2006.081430 maxillary gland mucin.
757
758 Ghosh et al.: Human ␤-Defensins in Body Fluids
RIA (19 ). ELISA assays have become one of the most etry. We have described the use of these peptides in
popular biomedical methods for quantifying proteins in previous work (9 ).
biological samples because, in addition to their sensitivity
and specificity, they are simpler and less costly than other elisa
analyses. The main objective of the present study was to We coated 96-well immunoplates (MaxiSorp™; Nunc)
develop a sensitive and reproducible analytical tool to with 50 ␮L anti– hBD-2 or – hBD-3 antibodies from differ-
measure hBD-2 and hBD-3 peptide concentrations and ent vendors (see Table 1 in the Data Supplement that
use it to quantify these peptides in saliva and other body accompanies the online version of this article at http://
fluids. www.clinchem.org/content/vol53/issue4), diluted to
1 mg/L in 0.05 mol/L carbonate buffer, pH 9.6, 4 °C, for
Materials and Methods 18 h. Subsequently, we blocked the wells with 200 ␮L of
sample collection 1% bovine serum albumin in PBS at room temperature for
Individuals who contributed saliva samples were chosen 10 min. After washing 3 times with 200 ␮L PBS containing
at random, with equal representation of male and female 1 mL/L Tween 20, we incubated 100 ␮L/well of test
participants and reflective of all age groups, from infants samples at room temperature for 60 min. The plates were
to the elderly. We did not exclude anyone for smoking or washed 3 times with PBS containing 1 mL/L Tween 20,
taking medications. We collected unstimulated saliva and wells were incubated at room temperature with
samples from infants and young children by use of sterile 50 ␮L secondary antibody diluted to 0.2 mg/L in PBS
disposable pipettes. Adults were asked to expectorate plus 1 mL/L Tween 20 for 30 min. Plates were washed
directly into sterile tubes. Saliva samples were then trans- 3 times with PBS plus 1 mL/L Tween 20 and filled with
ferred into sterile vials, centrifuged at 10 000g at 4 °C for 50 ␮L/well streptavidin-peroxidase (Roche Diagnostics;
20 min, and stored at ⫺70 °C until use. We obtained 1:10 000 in PBS plus 1 mL/L Tween 20). Plates were
blister fluids (BF) from the Skin Diseases Research Center then incubated at room temperature for an additional
(Department of Dermatology, Case School of Medicine 30 min, washed 3 times as described above, and incu-
and University Hospitals of Cleveland). Blisters were bated with 2,2⬘-azino-bis-3-ethylbenzthiazoline-6-sulfonic
generated by applying suction blister cups onto forearm acid (Roche Diagnostics) in the dark at room temperature
sites; after ⬃90 –120 min, the vacuum was released and for 20 min. Absorbance was measured at 415 nm with a
fluid was aspirated from each blister using a 23-gauge microplate reader (Bio-Rad Model 680).
needle. Fluids were microcentrifuged (8000g) for 5 min, Unless otherwise mentioned, during the validation
and the supernatants were frozen at ⫺70 °C. We collected and standardization process we performed ELISA assays
bronchoalveolar lavage (BAL) fluids from healthy partic- in 1⫻ PBS (pH 7.3) using antibody pairs from Peprotech.
ipants as described (20 ). We collected female genital To measure defensin concentrations in body fluids, we
secretions from premenopausal women visiting the Metro performed ELISA using 250 mmol/L CaCl2 (final concen-
Health Medical Center of Case Western Reserve Univer- tration). We quantified hBDs by simultaneous ELISA runs
sity by 2 different procedures. In the first procedure (in 250 mmol/L CaCl2) using recombinant hBDs as
(vaginal swab), we used sterile dry swabs to collect calibrators.
genital secretions on the endocervix. The swabs were
mucin isolation and modification
gently applied on the cervical os, and a slight pressure
Ovine submaxillary gland mucin (OSM) was purified
was applied by partly rotating the swabs, without any
from frozen glands as described (23 ), omitting the
mucosal trauma. The swab samples were rapidly inserted
hydroxyapatite chromatographic step and including pro-
into 1 mL PBS (137 mM NaCl, 10 mM phosphate, 2.7 mM
tease inhibitors (Chelex 100 and phenyl-methane-sulfonyl-
KCl; pH 7.2) and stored at ⫺70 °C until use. After vaginal
fluoride) in the initial stages of purification. Enzymatic
swab sampling, we collected cervicovaginal lavage (CVL)
desialylization of OSM (giving a-OSM) was performed as
by use of a standardized 60-s vaginal washing with 10 mL
described by Gerken and Dearborn (23 ) using neuramin-
PBS (pH 7.2) as described (21 ). Informed consent was
idase from Clostridium perfringens (Sigma). [13C]NMR
obtained for all sample donors.
spectra confirmed the purity of the isolated mucin and
complete removal of the sialic acid after neuraminidase
generation of recombinant hbd-2 and -3
treatment (23 ). Note that native OSM contains exclusively
We produced recombinant human BD-2 from the infec-
the disaccharide ␣-NeuNAc 2– 6 ␣-GalNAc-O-Ser/Thr
tion of Sf21 cells with baculovirus constructs (a gift from
and is therefore one of the most heavily sialylated mucins.
T. Ganz, UCLA) as described (22 ). We produced recom-
binant human BD-3 using an hBD-3–His tag fusion con- Results
struct, generated by PCR and cloned into pET-30c (2 ). We evaluation of elisa for detection of hbd-2 and
confirmed the identity and purity of rhBD-2 and -3 by -3 by use of different capture and detection
acid urea–polyacrylamide gel electrophoresis migration, antibodies
N-terminal amino acid sequencing, and matrix-assisted We investigated the sensitivity of commercially available
laser desorption/ionization time-of-flight mass spectrom- antibodies against hBD-2 and -3 (see Table 1 in the online
Data Supplement). Affinity-purified polyclonal capture hBD-2 and hBD-3 were far lower than the values obtained
antibodies (antihuman BD-2 and BD-3; Peprotech) and with serially diluted recombinant hBDs (r ⬎0.99). This
biotinylated anti-hBD detection antibodies (⫺2 and ⫺3; nonlinearity led us to investigate the possibility of inter-
Peprotech) provided the highest sensitivity in our ELISA ference by other molecules present in saliva.
formats (data not shown). Using this pair of antibodies,
absorbance at 415 nm against the respective calibrators masking effect of saliva in the detection of
(rhBD-2, 0.075–9.6 ␮g/L; rhBD-3, 3.0 –192 ␮g/L) exhibited hbd-2 and -3
a linear correlation coefficient of r ⬎0.99 for both de- We performed ELISA measurements of pooled (n ⫽ 3)
fensins (Fig. 1). The limit of detection (limit of the blank) serially diluted saliva (1:2 to 1:128), with or without
for the assay, defined as the mean of the buffer control addition of rhBD-2 (20 pg) or rhBD-3 (5 ng). We observed
(PBS) ⫹ 3SD, was 75 ng/L for hBD-2 and 3 ␮g/L for significantly lower detection of added hBDs in the pres-
hBD-3. We observed no cross-reactivity between the 2 ence of saliva compared with hBDs in PBS, even when the
defensin peptides (Fig. 1). dilution was as low as 1:128. From the calculated percent
masking, it became apparent that salivary masking of
detection of hbd-2 and -3 in saliva both hBDs decreased with increased dilution of the saliva,
We and others have detected hBD-1 and hBD-2 in saliva and that the percent masking of hBD-3 was greater than
using Western blot analysis (14, 16 ). Here we demonstrate that of hBD-2 (Fig. 3). These results indicate that masking
that Western blot analysis can detect hBD-3 in normal agents act differentially on the 2 defensin peptides.
human saliva (Fig. 2).
We analyzed samples from 3 healthy individuals. Both role of mucin in masking hbd-2 and -3 elisa
hBD-2 and -3 were detectable in all 3 samples. When signals
serially diluted saliva samples were used, however, non- Because large anionic glycoproteins, i.e., mucins, are
linearity in ELISA readouts for both hBDs was observed abundantly present in saliva (24 ) and in other epithelial
(Fig. 1). Linear correlation coefficient (r) values for both cell-derived body fluids (25, 26 ), we conducted ELISAs of
Fig. 1. ELISA detection of hBD-2 (A) and -3 (B) in serially

diluted saliva samples.
Saliva samples were serially diluted as indicated with PBS. For
comparison of linearity, serially diluted recombinant hBDs were
used. ELISA readings for hBD-3 using hBD-2 calibrators (A) and for
hBD-2 using hBD-3 calibrators (B) are also incorporated (black
squares).
fixed concentrations of hBD-2 and -3 in the presence of

serially diluted purified salivary mucins to determine the
involvement of these molecules in masking hBD detec-
tion. Because the sensitivity of the 2 defensin assays
varied, we used different fixed amounts of each, 20 pg
hBD-2 and 5 ng hBD-3, and added purified mucin (OSM)
in wt:wt ratios with the defensins (Fig. 4A). The results
demonstrated concentration-dependent mucin-associated
masking of hBD-2 and -3. To establish if anionic sialic acid
residues of mucin are involved in the masking effect, we
conducted ELISAs of hBDs in the presence of neuramin-
idase-treated mucin, i.e., asialo-mucin. Fig. 4B shows a
decrease of ⬃25%–30% in masking of both hBDs by
asialo-mucin compared with untreated mucin (Fig. 4A),
confirming a role for sialic acid residues in masking hBD
signals in the ELISAs.
Fig. 2. Western blot analysis of hBD-3 in saliva. optimization of the elisa for detection of hbd-
Whole saliva (100 ␮L) was centrifuged (10 000g, 4 °C, 10 min); supernatant was 2 and -3 in body fluids
lyophilized, reconstituted in 20 ␮L of sample buffer, separated on 12% Tris-HCl We explored different strategies to overcome the mask-
gels (Bio-Rad) along with recombinant hBD-3 (50 ng; positive control), and
transferred to a polyvinylidene difluoride membrane. The membrane was blocked ing. We examined the effect of pH on the detection of
with 5% skim milk in TBS containing 0.05% Tween for 1 h, followed by overnight hBDs in saliva by assaying a pooled saliva sample (n ⫽ 3)
incubation with anti-hBD-3 primary antibody (Peprotech). The membrane was
washed, incubated with horseradish peroxidase-conjugated secondary antibody using a pH interval of 5.5–7.8 using 100 mmol/L sodium
(Bio-Rad), and visualized with SuperSignal West Femto Maximum Sensitivity phosphate. Detection of the salivary hBDs was slightly
Substrate (Pierce).
Fig. 3. Masking effect of saliva in the detection of

hBD-2 (A) and hBD-3 (B) using ELISA.
Serially diluted (1:2 to 1:128 with PBS) saliva was en-
riched with recombinant hBDs (20 pg hBD-2 or 5 ng
hBD-3), and percent masking of the respective hBDs in
our ELISA was measured from absorbance readings of
saliva enriched with hBDs and hBDs alone. The formula
used was % masking ⫽ [100% ⫺ (ASaliva⫹rhBDs ⫺ ASaliva)/
(ArhBDs ⫺ Anegative control) ⫻ 100], where A is absorbance.
The absorbance reading of PBS was the negative control.
Fig. 4. Masking effect of mucins and asialo-

mucins in the detection of hBDs using ELISA.
Serially diluted mucin in PBS (A) or serially diluted
asialo-mucin in PBS (B) was enriched with 20 pg
rhBD-2 and 5 ng rhBD-3, respectively. The absor-
bance (A) of the positive control (20 pg for rhBD2;
5 ng for rhBD3) was set at 100% detection. The
percent masking by either mucin or asialo-mucin
was calculated by the following formula: % mask-
ing ⫽ [100% ⫺ (ArhBDs⫹mucins ⫺ Anegative control)/
(ArhBDs ⫺ Anegative control) ⫻ 100]. The absorbance
reading of PBS was the negative control.
better in acidic pH, but acidic pH alone did not improve assays for detection of hBDs in body fluids in the presence
hBD detection in saliva to any significant degree (see of 250 mmol/L CaCl2.
Fig. 1 in the online Data Supplement). We then examined
the effect of monovalent and divalent cations at various intra- and interassay precision
ionic strengths on the ELISAs to detect hBDs in saliva. We Using the hBD-2 calibrator (1 ␮g/L) and pooled saliva (8.2
found that divalent cations (Mg2⫹, Ca2⫹) were better than ␮g/L hBD-2), the intraassay CVs were 4.8% and 6.06%,
monovalent cations (Na⫹) (Fig. 5) and that 250 mmol/L respectively (n ⫽ 20), and interassay CVs were 5.81%
CaCl2 optimized the detection of both hBD-2 and hBD-3 and 7.63%, respectively (n ⫽ 8). With an hBD-3 calibrator
equally well (see Fig. 2 in the online Data Supplement). (50 ␮g/L) and the pooled saliva (627 ␮g/L hBD-3), the
We further compared the ELISA signals from neuramin- intraassay CVs were 4.6% and 6.7%, respectively (n ⫽ 20),
idase-pretreated saliva in PBS to the ELISA signals from and the interassay CVs were 5.31% and 8.12%,
saliva in 250 mmol/L CaCl2 and observed that 250 respectively.
mmol/L CaCl2 was best at enhancing the hBD-2 and
hBD-3 signals. analytical recovery of the calibrator
Indeed, when purified mucin was incubated with the Recoveries of exogenously added recombinant hBD-2 (50,
respective hBDs in the presence of 250 mmol/L CaCl2, we 100, and 200 ng/L) from saliva, BF, BAL, and CVL
were able to detect hBD-2 and -3 to virtually 100% samples ranged from 82%–107%, 88%–105%, 81%–107%,
(Fig. 5C). We therefore performed subsequent ELISA and 84%–103%, respectively. The percentage recoveries of
Fig. 5. (A,B), Effect of inorganic salts on

ELISA readouts.
A saliva sample was mixed (1:1) with respective
inorganic salts (final concentration 100 mmol/L) as
shown, and 100 ␮L from each condition was used
to screen hBD-2 and -3 concentrations by ELISA.
Triplicate assays were performed and results are
expressed as the mean and SE. (C), ELISA readouts
for hBD-2 and hBD-3 preincubated with serially
diluted mucin and assayed in the presence of 250
mmol/L CaCl2. HBDs and mucin were preincubated
in 250 mmol/L CaCl2, followed by the ELISA assay.
The y axis represents the percent of ELISA readout
obtained when comparing absorbance for hBDs in
the presence or absence of mucin.
exogenously added recombinant hBD-3 (2.5, 5, and 10 were then compared with salivary hBD-2 concentrations
␮g/L) from saliva, BF, BAL, and CVL samples ranged (in ␮g/g total salivary proteins) from corresponding
from 89% to 104%, 83% to 99%, 86% to 109%, and 87% to samples (Fig. 6B). BF concentrations of hBD-2 did not
107%, respectively. correlate with the salivary concentrations of hBD-2 pep-
tides. Similar analysis of the hBDs in BAL from healthy
measurement of hbd-2 and -3 concentrations participants (n ⫽ 5) showed the absence of hBD-3 and low
in saliva from healthy individuals concentrations of hBD-2 (0 – 49 ng/g total proteins). We
Saliva from 60 healthy individuals was analyzed for the also compared the concentration of hBD-2 in BAL with
presence of hBD-2 and -3. Concentrations of hBD-2 that in corresponding saliva samples (Fig. 6C). BAL
ranged from 1.2 to 21.1 ␮g/L (mean, 9.48; median, 3.28), concentrations of hBD-2 also did not correlate with the
whereas concentrations of hBD-3 ranged from 50 to salivary hBD-2 peptide concentrations. Unlike BF and
931 ␮g/L (mean, 325.77; median, 253) (Fig. 6A). BAL, we could detect hBD-3, along with hBD-2, in CVL
and vaginal swabs from healthy women (n ⫽ 2; mean
identification of hbd concentrations in hBD-2 in CVL, 1.42 ␮g/g total proteins; in vaginal swab,
blister fluids, bal, vaginal swabs, and cvl 3.42 ␮g/g total proteins; mean hBD-3 in CVL, 55 ␮g/g
from healthy individuals total proteins; in vaginal swab, 103 ␮g/g total proteins).
We analyzed BF samples from healthy participants (n ⫽
10) by ELISA for the presence of both hBD-2 and hBD-3 Discussion
peptides. Although we could not detect hBD-3 in the BF We developed ELISAs for quantifying hBD-2 and hBD-3
samples (limit of detection of assay, 3.0 ␮g/L), we found in body fluid samples. During assay development, it
hBD-2 in all the samples (30 –320 ng/g total protein). The became clear that salivary components partially mask the
hBD-2 concentrations (in ␮g/g total BF proteins) in BF detection of both hBD-2 and -3, with the effect being
Fig. 6. (A), Box-plot representation of salivary hBD-2 and

hBD-3 peptide concentrations in healthy participants
(n ⫽ 60).
Each saliva sample was mixed (1:1) with CaCl2 (final concentra-
tion 250 mmol/L), and 100 ␮L from each sample was assayed for
the presence of the respective hBD. Each sample was run in
duplicate. Results were obtained by running hBD-2 and hBD-3
calibrators with each assay, respectively. *Outliers. (B), ELISA
measurements of hBD-2 concentrations in BFs (n ⫽ 10) and
matched saliva from healthy donors. BF (75 ␮L) from each sample
was mixed with 25 ␮L CaCl2 (final concentration 250 mmol/L)
and assayed for hBD-2. C, Determination of hBD-2 in BAL (n ⫽ 5)
samples and matched saliva from healthy donors. BAL samples
were lyophilized, dialyzed, and reconstituted in 250 mmol/L
CaCl2, and 100 ␮L from each sample was assayed for hBD-2.
Salivary hBD-2 was measured as described (A). Results are
calculated as the ratio of hBD-2 peptide compared with total
protein (measured by Bio-Rad DC Protein Assay) per sample.
greater for hBD-3. This difference may be due to stronger shielding the interactions between hBDs, sialic acid resi-
electrostatic interactions between negatively charged dues, and other negatively charged moieties.
moieties and hBD-3, since hBD-3 is more positively The almost 40-fold difference between hBD-2 and
charged (⫹11) than hBD-2 (⫹6) (27 ). Moreover, previous hBD-3 concentrations in saliva from healthy oral cavities
studies with hBD-1 (⫹4 net charge) (28 ) demonstrated is consistent with our observations that hBD-3 is more
that mucins also mask the detection of hBD-1 in slot-blot highly expressed than hBD-2 in healthy oral epithelium
assays but not in Western blots (14 ). We therefore sur- (data not shown). The concentrations of salivary hBD-3
mised that negatively charged mucins, found in mg% that we found are in line with those reported by others
(mg/100 mL) quantities in saliva (24 ), are masking hBD-3 using extraction and slot blot procedures (15 ). The large
more than hBD-2 due to the greater net positive charge of standard deviations in our data and those reported by Tao
this peptide. Moreover, hBD-3 includes a greater percent- et al. (15 ) suggest large interindividual variability in
age of total surface area available for electrostatic inter- hBD-2 and -3 expression. This suggestion is consistent
actions compared with hBD-2 (28 ). In addition, our cal- with findings that hBD genes are polymorphic in copy
culation of polar surface area using GetArea 1.1 software numbers and that high copy numbers correlate with high
(29 ) was greater for hBD-3 (40% of total surface area) than levels of mRNA (37 ). Finally, being able to quantify
hBD-2 (30% of total surface area), which further supports salivary hBD-2 and hBD-3, which primarily reflect release
the greater masking of hBD-3 vs hBD-2. from the oral epithelium where they would most likely be
To assess further the contribution of a mucin-depen- expressed at high concentrations, could provide a means
dent electrostatic interaction in masking hBD detection, to determine the degree of fitness of the mucosal epithe-
we found that asialo-mucin significantly diminished the lium toward microbial challenges.
masking effect. The sialic acid residue is generally located Results for normal BFs also demonstrated the presence
on the terminal position of carbohydrate chains of glyco- of hBD-2, albeit at low concentrations. Ortega et al. (38 )
demonstrated the absence of hBD-2 in burned blister
proteins, and sialic acid-mediated antigen masking has
fluids. The presence of hBD-2 peptide in normal blister
been described (30 ). Moreover, sialic acid–mediated anti-
(this report) and its absence in BF from burned sites (38 )
gen masking by steric hindrance has been reported in the
support reverse-transcription PCR results showing re-
3-fucosyl-N-acetyl-lactosamine antigen (31 ). Thus, in ad-
duced or no expression of hBD-2 mRNA in burned skin
dition to the possible role of electrostatic and polar
compared with unburned skin (39 ). In contrast to hBD-2,
interaction of mucins with defensins, the role of the sialic
we could not detect hBD-3 in BFs, perhaps because of the
acid residue itself in masking the ELISA signal cannot be
low sensitivity of our ELISA for hBD-3 compared with
excluded.
hBD-2.
Mucin forms a viscoelastic gel that coats epithelial
Previous semiquantitative Western analysis showed
surfaces and is affected by pH (32 ) and ion content (33 ).
the presence of hBD-2 in BAL samples of patients with
We observed that pH changes and ionic strength affect the cystic fibrosis (⬃15 ␮g/L) or bronchiolitis obliterans (⬃1.3
recovery of the ELISA signal for both hBD-2 and -3 in ␮g/L), but not in healthy individuals (17, 18 ). Our ELISA
saliva. Recovery of signal is better in acidic pH and best in method, however, can detect the presence of hBD-2 (⬃0.4
the presence of 250 mmol/L CaCl2. Moreover, the persis- ng/L) in BAL samples from healthy individuals, a clear
tence of hBD masking in the presence of asialo-mucins, advantage over semiquantitative Western blots. More-
albeit at reduced levels, supports possible macromolecu- over, our ELISA is able to detect both hBD-2 and hBD-3 in
lar organization or viscoelastic properties of mucin in cervicovaginal lavage fluids (1.46 and 55.5 ␮g/g total
contributing to hBD masking. We cannot rule out the proteins) and vaginal swabs (3.42 and 103 ␮g/g total
possibility that other negatively charged salivary compo- proteins) from healthy women. This is the first reported
nents, aside from mucins, are also involved. These could documentation of these peptides in female genital tract
include, but are not limited to, calprotectin present in secretions.
mg/L quantities (34 ). Nevertheless, recovery of recombi- With new and exciting information promoting hBDs as
nant hBDs in the presence of inorganic salts supports the important agents in mucosal defense, these assays should
involvement of electrostatic interactions, as we obtained help to determine if individuals expressing low amounts
the best recovery using divalent cations rather than mono- of these peptides may be inherently more susceptible to
valent cations. Electrostatic interactions alone cannot fully mucosal infections. Moreover, do infectious diseases have
explain recovery of the hBD signal, since asialylation of an affect on mucosal hBD peptide concentrations? Inter-
saliva or use of other divalent cations other than calcium estingly, Sun et al. (40 ) showed diminished hBD-2 peptide
does not completely unmask the hBD signal in our expression in HIV-positive oral mucosa compared with
ELISA—near-complete recovery of signal was obtained healthy controls. The ability to now measure inducible
only with calcium. This unique calcium-dependent phe- hBD peptide concentrations in body fluids, in a manner
nomenon could be due to calcium-induced changes in the conducive for screening samples from multiple body
intrinsic viscosity of saliva (35 ) and macromolecular con- sites, sets the stage for epidemiological assessment of the
traction/folding of mucins (36 ), or calcium could be role of hBDs in numerous infectious diseases.
We thank Dr. Richard Silver, Division of Pulmonary 15. Tao R, Jurevic RJ, Coulton KK, Tsutsui MT, Roberts MC, Kimball
Critical Care, and Department of Medicine, University JR, et al. Salivary antimicrobial peptide expression and dental
Hospitals of Cleveland and Case Western Reserve Uni- caries experience in children. Antimicrob Agents Chemother
2005;49:3883– 8.
versity, for providing bronchoalveolar lavage samples,
16. Mathews M, Jia HP, Guthmiller JM, Losh G, Graham S, Johnson
and Dr. Richard Beigi, Department of Obstetrics and GK, et al. Production of beta-defensin antimicrobial peptides by
Gynecology, Metro Health Medical Center, Cleveland, for the oral mucosa and salivary glands. Infect Immun 1999;67:
providing cervicovaginal lavage and vaginal swab sam- 2740 –5.
ples. This work was supported by National Institutes of 17. Chen CI, Schaller-Bals S, Paul KP, Wahn U, Bals R. Beta-defensins
Health grants RO1DE16334 (A.W.), RO1 DE17334 (A.W.), and LL-37 in bronchoalveolar lavage fluid of patients with cystic
RO1DE15510 (A.W.), R01CA78834 (T.A.G.), and fibrosis. J Cyst Fibros 2004;3:45–50.
18. Ross DJ, Cole AM, Yoshioka D, Park AK, Belperio JA, Laks H, et al.
P30AR39750 (T.S.M.).
Increased bronchoalveolar lavage human beta-defensin type 2 in
bronchiolitis obliterans syndrome after lung transplantation.
Transplantation 2004;78:1222– 4.
References 19. Hiratsuka T, Mukae H, Iiboshi H, Ashitani J, Nabeshima K,
1. Harder J, Bartels J, Christophers E, Schroder JM. A peptide Minematsu T, et al. Increased concentrations of human beta-
antibiotic from human skin. Nature 1997;387:861. defensins in plasma and bronchoalveolar lavage fluid of patients
2. Harder J, Bartels J, Christophers E, Schroder JM. Isolation and with diffuse panbronchiolitis. Thorax 2003;58:425–30.
characterization of human beta -defensin-3, a novel human induc- 20. Technical recommendations and guidelines for bronchoalveolar
ible peptide antibiotic. J Biol Chem 2001;276:5707–13. lavage (BAL): report of the European Society for Pneumology Task
3. Bowdish DM, Davidson DJ, Hancock RE. Immunomodulatory prop- Group. Eur Respir 1989;561– 85.
erties of defensins and cathelicidins. Curr Top Microbiol Immunol 21. Bélec L, Meillet D, Lévy M, Georges A, Tévi-Bénissan C, Pillot J.
2006;306:27– 66. Dilution assessment of cervicovaginal secretions obtained by
4. Krisanaprakornkit S, Weinberg A, Perez CN, Dale BA. Expression vaginal washing for immunological assays. Clin Diag Lab Immunol
of the peptide antibiotic human beta-defensin 1 in cultured 1995;57– 61.
gingival epithelial cells and gingival tissue. Infect Immun 1998; 22. Valore EV, Park CH, Quayle AJ, Wiles KR, McCray PB Jr, Ganz T.
66:4222– 8. Human beta-defensin-1: an antimicrobial peptide of urogenital
5. Garcia JR, Jaumann F, Schulz S, Krause A, Rodriguez-Jimenez J, tissues. J Clin Invest 1998;101:1633– 42.
Forssmann U, et al. Identification of a novel, multifunctional 23. Gerken TA, Dearborn DG. Carbon-13 NMR studies of native and
beta-defensin (human beta-defensin 3) with specific antimicrobial modified ovine submaxillary mucin. Biochemistry 1984;23:1485–
activity: its interaction with plasma membranes of Xenopus oo- 97.
cytes and the induction of macrophage chemoattraction. Cell 24. Rayment SA, Liu B, Offner GD, Oppenheim FG, Troxler RF.
Tissue Res 2001;306:257– 64. Immunoquantification of human salivary mucins MG1 and MG2 in
6. Garcia JR, Krause A, Schulz S, Rodriguez-Jimenez FJ, Kluver E, stimulated whole saliva: factors influencing mucin levels. J Dent
Adermann K, et al. Human beta-defensin 4: a novel inducible Res 2000;79:1765–72.
peptide with a specific salt-sensitive spectrum of antimicrobial 25. Gipson IK, Ho SB, Spurr-Michaud SJ, Tisdale AS, Zhan Q, Torlak-
activity. FASEB J 2001;15:1819 –21. ovic E, et al. Mucin genes expressed by human female reproduc-
tive tract epithelia. Biol Reprod 1997;56:999 –1011.
7. Yadava P, Zhang C, Sun J, Hughes JA. Antimicrobial activities of
26. Thornton DJ, Howard M, Khan N, Sheehan JK. Identification of two
human beta-defensins against Bacillus species. Int J Antimicrob
glycoforms of the MUC5B mucin in human respiratory mucus.
Agents 2006;28:132–7.
Evidence for a cysteine-rich sequence repeated within the mole-
8. De Smet K, Contreras R. Human antimicrobial peptides: de-
cule. J Biol Chem 1997;272:9561– 6.
fensins, cathelicidins and histatins. Biotechnol Lett 2005;27:
27. Bauer F, Schweimer K, Kluver E, Conejo-Garcia JR, Forssmann
1337– 47.
WG, Rosch P, et al. Structure determination of human and
9. Quinones-Mateu ME, Lederman MM, Feng Z, Chakraborty B, murine beta-defensins reveals structural conservation in the
Weber J, Rangel HR, et al. Human epithelial beta-defensins 2 and absence of significant sequence similarity. Protein Sci 2001;
3 inhibit HIV-1 replication. Aids 2003;17:F39 – 48. 10:2470 –9.
10. Feng Z, Dubyak GR, Lederman MM, Weinberg A. Cutting edge: 28. Schibli DJ, Hunter HN, Aseyev V, Starner TD, Wiencek JM, McCray
human beta defensin 3–a novel antagonist of the HIV-1 coreceptor PB Jr, et al. The solution structures of the human beta-defensins
CXCR4. J Immunol 2006;177:782– 6. lead to a better understanding of the potent bactericidal activity of
11. Yang D, Chertov O, Bykovskaia SN, Chen Q, Buffo MJ, Shogan J, HBD3 against Staphylococcus aureus. J Biol Chem 2002;277:
et al. Beta-defensins: linking innate and adaptive immunity 8279 – 89.
through dendritic and T cell CCR6. Science 1999;286:525– 8. 29. Fraczkiewicz R, Braun W. Exact and efficient analytical calculation
12. Tani K, Murphy WJ, Chertov O, Salcedo R, Koh CY, Utsunomiya I, of the accessible surface areas and their gradients for macromol-
et al. Defensins act as potent adjuvants that promote cellular and ecules. J Compu Chem 1998;19:319 –33.
humoral immune responses in mice to a lymphoma idiotype and 30. Schauer R. Sialic acids and their role as biological masks. Trends
carrier antigens. Int Immunol 2000;12:691–700. Biochem Sci 1985;10:357– 60.
13. Oppenheim JJ, Biragyn A, Kwak LW, Yang D. Roles of antimicrobial 31. Howie AJ, Brown G. Effect of neuraminidase on the expression of
peptides such as defensins in innate and adaptive immunity. Ann the 3-fucosyl-N-acetyllactosamine antigen in human tissues. J Clin
Rheum Dis 2003;62(Suppl 2):ii17–21. Pathol 1985;38:409 –16.
14. Sahasrabudhe KS, Kimball JR, Morton TH, Weinberg A, Dale BA. 32. Verdugo P. Goblet cells secretion and mucogenesis. Annu Rev
Expression of the antimicrobial peptide, human beta-defensin 1, Physiol 1990;52:157–76.
in duct cells of minor salivary glands and detection in saliva. J 33. Beeley JA. Fascinating families of proteins: electrophoresis of
Dent Res 2000;79:1669 –74. human saliva. Biochem Soc Trans 1993;21:133– 8.
34. Cuida M, Brun JG, Tynning T, Jonsson R. Calprotectin levels in oral variation of a beta-defensin antimicrobial-gene cluster. Am J Hum
fluids: the importance of collection site. Eur J Oral Sci 1995;103: Genet 2003;73:591– 600.
8 –10. 38. Ortega MR, Ganz T, Milner SM. Human beta defensin is absent in
35. Raynal BD, Hardingham TE, Sheehan JK, Thornton DJ. Calcium- burn blister fluid. Burns 2000;26:724 – 6.
dependent protein interactions in MUC5B provide reversible 39. Milner SM, Ortega MR. Reduced antimicrobial peptide expression
cross-links in salivary mucus. J Biol Chem 2003;278:28703–10. in human burn wounds. Burns 1999;25:411–3.
36. Chernick WS, Barbero GJ. Studies on human tracheobronchial and 40. Sun L, Finnegan CM, Kish-Catalone T, Blumenthal R, Garzino-
submaxillary secretions in normal and pathophysiological condi- Demo P, La Terra Maggiore GM, et al. Human beta-defensins
tions. Ann N Y Acad Sci 1963;106:698 –708. suppress human immunodeficiency virus infection: potential role
37. Hollox EJ, Armour JA, Barber JC. Extensive normal copy number in mucosal protection. J Virol 2005;79:14318 –29.
766 –772 (2007) Other Areas of
Clinical Chemistry
Expressing the Modification of Diet in Renal

Disease Study Equation for Estimating Glomerular
Filtration Rate with Standardized Serum
Creatinine Values
Andrew S. Levey,1 Josef Coresh,2 Tom Greene,3 Jane Marsh,2 Lesley A. Stevens,1*
John W. Kusek,4 and Frederick Van Lente5 for Chronic Kidney Disease Epidemiology
Collaboration
Purpose: We sought to reexpress the 4-variable Modifi- from Beckman CX3 assay results during the MDRD
cation of Diet in Renal Disease (MDRD) Study equation Study in 1989 –1991 (r2 ⴝ 0.9987 in 253 samples). Com-
for estimation of glomerular filtration rate (GFR) using bining these factors, standardized Scr ⴝ 0.95 ⴛ original
serum creatinine (Scr) standardized to reference MDRD Study Scr. The reexpressed 4-variable MDRD
methods. Study equation for Scr (mg/dL) is GFR ⴝ 175 ⴛ stan-
Methods: Serum specimens included creatinine refer- dardized Scrⴚ1.154 ⴛ ageⴚ0.203 ⴛ 1.212 (if black) ⴛ 0.742
ence materials prepared by the College of American (if female), and for Scr (␮mol/L) is GFR ⴝ 30849
Pathologists (CAP), traceable to primary reference ma- ⴛ standardized Scrⴚ1.154 ⴛ ageⴚ0.203 ⴛ 1.212 (if black) ⴛ
terial at the NIST, with assigned values traceable to 0.742 (if female) [GFR in mL 䡠 minⴚ1 䡠 (1.73 m2)ⴚ1].
isotope dilution mass spectrometry (IDMS), a calibra- Conclusion: When the calibration of Scr methods is
tion panel prepared by the Cleveland Clinic Research traceable to the Scr reference system, GFR should be
Laboratory (CCRL), and frozen samples from the estimated using the MDRD Study equation that has
MDRD Study. Split specimens were measured at the been reexpressed for standardized Scr.
CCRL using the Roche enzymatic and Beckman CX3 © 2007 American Association for Clinical Chemistry
kinetic alkaline picrate assays.
Results: Roche enzymatic assay results on CAP samples To facilitate detection, evaluation, and management of
were comparable to IDMS-assigned values. Beckman chronic kidney disease (CKD),6 guidelines recommend
CX3 assay results in 2004 –2005 were significantly higher that clinical laboratories compute and report estimated
than but highly correlated with simultaneous Roche glomerular filtration rate (GFR) using estimating equa-
enzymatic assay results (r2 ⴝ 0.9994 on 40 CCRL sam- tions such as that derived from the Modification of Diet in
ples) and showed minimal but significant upward drift Renal Disease (MDRD) Study. Many clinical laboratories
have begun this practice (1– 6 ). Proper use of GFR esti-
mating equations requires a known calibration of the
serum creatinine (Scr) assays (7, 8 ). Calibration to a single
1
Division of Nephrology, Tufts-New England Medical Center, Boston, standardized Scr based on gold standard methods has
MA.
2
Johns Hopkins Medical Institutions, Baltimore, MD.
been widely recommended (3, 8, 9 ). The College of Amer-
3
Department of Quantitative Health Sciences, Cleveland Clinic Founda- ican Pathologists (CAP) has prepared fresh-frozen Scr
tion, Cleveland, OH. reference materials, traceable to a primary reference ma-
4
National Institute of Diabetes, Digestive and Kidney Diseases, Bethesda,
MD.
5
Department of Clinical Pathology, Cleveland Clinic Foundation, Cleve-
6
land, OH. Nonstandard abbreviations: CKD, chronic kidney disease; GFR, glomer-
* Address correspondence to this author at: Division of Nephrology, ular filtration rate; MDRD, Modification of Diet in Renal Disease; Scr, serum
Tufts-New England Medical Center, 750 Washington St., Box 391, Boston, MA creatinine; CAP, College of American Pathologists; IDMS, isotope dilution
02111. Fax 617-636-8329; e-mail lstevens1@tufts-nemc.org. mass spectrometry; CCRL, Cleveland Clinic Research Laboratory; LC, liquid
Received July 26, 2006; accepted February 2, 2007. chromatography; LN24, Creatinine Accuracy Calibration Verification/Linear-
Previously published online at DOI: 10.1373/clinchem.2006.077180 ity Survey LN24; NKDEP, National Kidney Disease Education Program.
766
terial at the NIST, with assigned values based on isotope assay, the assay that was used during the MDRD Study
dilution mass spectrometry (IDMS), for use as trueness and is currently in use in the CCRL. Third, with frozen
controls to verify the traceability of results from routine samples from the MDRD Study, the Beckman CX3 assay
methods (10, 11 ). was adjusted for drift over the past decade. The study was
To date, the relationship of the MDRD Study creatinine approved by the institutional review boards of all partic-
assay to standardized creatinine was unknown, although ipating institutions.
results of smaller studies suggested the CX3 rate Jaffe
assay had a small positive bias compared with an IDMS reference samples from cap
reference method (10, 11 ). The purposes of this report are CAP 2004 LN24 samples 01 through 07 were prepared
to describe procedures for calibrating the Scr assay at the from a female-only donor pool, so the creatinine value
Cleveland Clinic Research Laboratory (CCRL) where the would be slightly lower than a mixed-sex pool (10, 11, 13 ).
MDRD Study samples were measured and to reexpress Sample 02 is the base female serum pool. Sample 07 had
the 4-variable (modified) MDRD Study equation (6 ) for reagent-grade creatinine added to bring the creatinine to
use with standardized Scr. ⬃352 ␮mol/L (4 mg/dL). Samples 03 through 06 were
prepared by gravimetric admixing of samples 02 and 07.
Materials and Methods Sample 01 was prepared from sample 02 by gravimetric
study design dilution with 0.01 mol/L phosphate buffered saline [pre-
Chronic Kidney Disease Epidemiology Collaboration is a pared by adding one packet of Sigma P-3813 phosphate-
research group sponsored by the National Institute of buffered saline, pH 7.4, to 1 L of deionized water (per-
Diabetes, Digestive and Kidney Diseases to develop and sonal communication, Mary Zimmer, January 22, 2007)].
validate improved estimating equations for GFR by pool- Creatinine concentrations in samples LN24-02 and
ing data from research studies and clinical populations. LN24-07 were value-assigned by the NIST using an LC-
Pooling data requires calibration of Scr assays of collabo- IDMS method. NIST-assigned values for samples 02 and
rating laboratories to creatinine assays currently used at 07 were 65.032 and 353.32 ␮mol/L (0.7390 and 4.0150
the CCRL. The reference standard method for measure- mg/dL), respectively. The creatinine values in the other
ment of Scr, IDMS, is extremely labor intensive, allowing pools were computed based on the known admixtures.
assay of only very few samples. Therefore, we followed
recommendations to establish a calibration hierarchy (12 ). calibration panel
The hierarchy starts with a primary calibrator that mate- We developed a calibration panel of pooled human sera to
rializes the SI units used for expression of creatinine establish the calibration relationship between the CX3
measurements, i.e., the NIST SRM 914a. This primary (2004) and Roche enzymatic methods across the range of
calibrator is used to directly calibrate the gas chromatog- Scr for use in patient samples. The calibration panel was
raphy and liquid chromatography (LC) IDMS reference developed at the CCRL from pooled patient sera from the
methods. From that point on, all subsequent calibrations Cleveland Clinic. The calibration panel included 40 refer-
or validations of the calibration status of lower-order ence sera (20 aliquots of 1.0 mL each frozen at ⫺70 °C)
methods are done via split-sample comparisons at CCRL pooled from at least 10 mixed-sex donors known to have
(Fig. 1). First, CAP Creatinine Accuracy Calibration Ver- Scr values covering the full range of 177 to 442 ␮mol/L
ification/Linearity Survey LN24 (LN24) samples were (0.5 to 5.0 mg/dL).
used as trueness controls to verify that the calibration of Briefly, serum pools were constructed by combining
the Roche enzymatic method was traceable to LC-IDMS within 2 h of collection excess clear serum obtained from
and correctly implemented in the CCRL. Second, values apparently healthy patients and patients with CKD as
measured with the Roche enzymatic assay on a calibration soon as routine testing was completed. Sera were refrig-
panel of 40 native pooled sera prepared by the CCRL were erated and combined according to the creatinine concen-
compared with values assigned using the Beckman CX3 tration to achieve a final pool volume of ⱖ25 mL. The
Fig. 1. Steps and results of calibration of the MDRD Study samples to creatinine reference materials.
Stepwise progression by which the Beckman Synchron CX3 (CX3) instrument used at the time of the MDRD Study (CX3 in 1989 –1991) was compared with creatinine
reference standards measured using IDMS technology. CAP samples with IDMS-assigned values were measured on the Roche enzymatic method and showed
equivalence. The CCRL calibration panel was then measured on the Roche instrument and the CX3 machine in 2005. This showed a regression slope of 0.906. Frozen
samples from the MDRD study (n ⫽ 253) were reassayed in 2004 on the CX3 instrument. This showed a regression slope of 1.046. Multiplication of slopes (1.00
䡠 0.906 䡠 1.046) yielded a calibration factor relationship of the original MDRD Study creatinine values to IDMS-traceable reference material of 0.95.
768 Levey et al.: Modification of Diet in Renal Disease Study Equation
serum pools were mixed by gentle inversion and filtered; The Beckman modified kinetic rate alkaline picrate
their combined creatinine concentrations were measured, (Jaffe) reaction was performed with the Beckman Syn-
and then serum was apportioned into separate 1.0 mL chron CX3 method during the MDRD Study and in 2004.
aliquots and frozen at ⫺70 °C. A set of 20 aliquots was Measurement of the picrate-creatinine complex formation
thawed and analyzed in triplicate in 3 separate runs on was taken at 520 and 560 nm at 25.6 s after sample
the same day. This process was repeated on a 2nd set of 20 introduction. The Beckman CX3 demonstrated CVs of
aliquots on a subsequent day. Each sample was analyzed 2.6% and 5.6% at creatinine values of 571 and 78.7 ␮mol/L
with the Roche enzymatic and CX3 assays after verifica- (6.46 and 0.89 mg/dL), respectively, in 2004 (n ⫽ 390) and
tion that the methods were and remained within internal 2.4% and 3.8% at creatinine values of 359 and 93.7 ␮mol/L
quality control limits. The runs also included a CAP (4.06 and 1.06 mg/dL), respectively, in 2005 (n ⫽ 165).
sample C-02 from the 2003 C mailing (11 ) prepared CAP proficiency test results (n ⫽ 25, 2004 C01 to 2005
according to Clinical and Laboratory Standards Institute C10) for these 2 methods at the CCRL demonstrated a
37A (14 ) as a validated reference material for each 20 mean percentage bias to the peer method target mean of
pooled specimens. Runs were deemed acceptable if the 1.01% (range, 3.46%–5.11%) and 1.79% (range, 2.09%–
concentration value for this material remained within 1 9.49%) for the Modular P and CX3, respectively. All
SD of the peer method mean. results were within acceptance criteria. The range of
Assigned values for the calibration panel were based creatinine values for these challenges was 61.9 to 663
on the Roche enzymatic assay. This method was selected ␮mol/L (0.7 to 7.5 mg/dL).
because it is free from most interferences and its calibra-
tion is traceable to IDMS (15 ). In addition, as shown later, computations
this method was verified to recover values assigned by The mean values of repeated measurements were used for
NIST to the CAP trueness controls. all computations. Linear regression slopes and intercepts
were obtained for each comparison of assays. Intercepts
stored samples from the mdrd study that were very small and nonsignificant (P ⬎0.05) were
Samples from the MDRD Study were fresh frozen at dropped from the regression. The final calibration rela-
⫺70 °C without thawing until 2004. The MDRD Study tionship was obtained by combining point estimates for
equation was derived using samples from the 1st baseline slopes and intercepts into a single new equation. The final
visit (B0) during 1989 –1991. The 253 samples used in the creatinine calibration factor was rounded to 2 significant
current study were from the 3rd baseline visit (B3, ⬃3 digits. The SE for the final calibration factor was com-
months later) and selected sequentially from the serum puted with the delta method (16 ). Ordinary least-squares
repository. regressions were used instead of errors-in-variables re-
gressions because correlations were ⬎0.993. These ex-
creatinine assays at the cleveland clinic tremely high correlations indicated the measurement er-
foundation research laboratory ror variance was a very small fraction of the total variance
The Roche enzymatic assays were performed on a Roche/ in this calibration setting, at which samples spanned a
Hitachi P module automated analyzer with Creatinine wide range of creatinine values and assays were averaged
Plus enzymatic assay reagents. The enzymatic method is across repeated measurements. The final creatinine cali-
based on the determination of sarcosine after conversion bration factor was then incorporated into the constant in
of creatinine with creatininase, creatinase, and sarcosine the MDRD Study GFR estimating equation.
oxidase. This assay demonstrated CVs of 1.8% and 2.0% at
creatinine values of 518 and 78.7 ␮mol/L (5.86 and 0.89 Results
mg/dL), respectively, in 2004 (n ⫽ 194) and 1.1% and standardization of scr assays
1.6% at creatinine values of 340 and 88.4 ␮mol/L (3.84 and Use of CAP sample for verification of traceability of Roche
1.00 mg/dL), respectively, in 2005 (n ⫽ 409). enzymatic assay to IDMS. Results of assays of the LN24
Table 1. Scr concentration in LN24 survey samples measured by Roche enzymatic assay.a
LN24 survey sample
01 02 03 04 05 06 07
NIST-assigned value, ␮mol/L 44 65 123 181 239 297 354
2004 (LN24 pilot), mean of 8 samples, ␮mol/L 45 71 125 186 244 301 358
2005 (LN24A), mean of 4 samples, ␮mol/L 45 69 124 184 239 299 354
Mean of all samples, ␮mol/L 45 70 125 185 241 300 356
Difference between NIST-assigned value and 2 5 2 4 3 3 2
mean of all samples, ␮mol/L
a
Results are rounded to nearest 1 ␮mol/L.
survey materials on the Roche enzymatic assay are shown MDRD Study samples in 2004 revealed a change over the
in Table 1. The Roche assay recovered the NIST-assigned past decade (Fig. 4). The regression showed a slope of
values for the LN24 trueness controls, confirming that the 1.037 (0.008) and intercept of 2.53 (2.12) ␮mol/L [0.029
Roche method was calibrated to be traceable to IDMS. (0.024) mg/dL], with an intercept P value of 0.23. After
Assays of the LN24 survey materials in 2004 and 2005 the small and nonsignificant intercept value was dropped,
showed nearly identical regression parameters (Fig. 2). slope was 1.046 (0.0024), r2 ⫽ 0.9986.
For all samples combined, the intercept (SE) was ⫺3.01
(0.88) ␮mol/L [⫺0.035 (0.01) mg/dL; P ⫽ 0.02], and the computation of the idms-traceable
slope was 1.00 (0.004), r2 ⫽ 0.9999. Assay results for calibration factors and reexpression of the
samples 01 to 07 were all within 6% of the LN24 survey mdrd study equation
assigned values [with all values ⬎88 ␮mol/L (1.0 mg/dL) The general method for calibrating the CCRL assays to the
assaying within 2%]; Statistically, the data suggest that IDMS assay at NIST is given in Table 2, equation 1. For all
3.01 ␮mol/L (0.035 mg/dL) could be subtracted from the 3 comparisons, intercepts for the regression were taken to
Roche enzymatic assay to get the best comparability to be zero, so the final calibration was derived by multipli-
NIST-assigned values. Because this difference is small and cation of regression slopes (1.0 䡠 0.906 䡠 1.046 ⫽ 0.95; Fig.
its inclusion would complicate calculations and the final 1). The SE of this correction factor, calculated with the
equation, however, this difference was omitted, and the delta method and assuming the Roche enzymatic method
Roche enzymatic values were used without correction to to be equivalent to the gold standard method, was 0.005.
assign reference values. The published MDRD Study equation (Table 2, equation
2) was reexpressed by substitution of equation 1 for the
Standardization of Beckman Synchron CX3 method to Roche term Scr (Table 2, equation 3). Sensitivity analyses retain-
enzymatic assay using the CCRL calibration panel. The Beck- ing intercepts and retaining additional significant digits to
man Synchron CX3 and Roche enzymatic methods were the final calibration factors to compute estimated GFR did
highly correlated (Fig. 3). The regression of the Roche not show clinically meaningful differences (see the Data
enzymatic on the Beckman Synchron CX3 method Supplement that accompanies the online version of this
showed a slope (SE) of 0.915 (0.009) and intercept of ⫺2.30 article at http://www.clinchem.org/content/vol53/issue4).
(2.21) ␮mol/L [⫺0.026 (0.025) mg/dL]. The P value for the
intercept was 0.31. After the small and nonsignificant Discussion
intercept was dropped, the slope (SE) was 0.906 (0.004; r2 Variability among clinical laboratories in calibration of Scr
⫽ 0.996). assays is an important limitation in the use of GFR
estimating equations. Variation in calibration introduces
Adjustment of Beckman Synchron CX3 for drift over time by error in GFR estimates, especially at high GFRs (17 ), and
re-assay of MDRD Study samples. Analysis of 253 frozen may account in part for the recent reports of widely
varying performance of the MDRD Study equation in
populations with higher GFRs (18 –38 ). In particular, the
bias at high GFR levels appears greater among studies in
which the Scr assay was not calibrated (18 ). This source of
error can be overcome by recalibration of the clinical
laboratory creatinine assay to the creatinine assay values
of the research laboratory in which the estimating equa-
tion was developed. Calibration of clinical laboratory
assay values obtained at a specific research laboratory is
not practical for widespread implementation of reporting
GFR estimates, however.
The National Kidney Disease Education Program
(NKDEP) has initiated a creatinine standardization pro-
gram to improve and normalize Scr results used in esti-
mating equations (8 ). After creatinine reference materials
that are traceable to higher order reference standards are
developed, a proficiency testing system will be used to
enable ongoing monitoring of calibration among clinical
laboratories. This process is expected to be completed by
2008. Reexpression of the MDRD Study equation based on
standardized assays will enable implementation of report-
Fig. 2. Roche enzymatic (enz.) assay vs NIST-assigned creatinine
values using LC-IDMS on CAP LN24 survey samples.
ing estimated GFR in clinical practice using calibrated Scr
Linear regression intercept ⫺3.01 ␮mol/L (⫺0.035 mg/dL), SE 0.88 ␮mol/L assays, thereby overcoming this limitation to the current
(0.01 mg/dL); slope 1.00, SE 0.004; r2 ⫽ 0.9999). use of GFR estimating equations.
Fig. 3. Roche enzymatic (enz.) vs Beck-

man Synchron CX3 assays using cali-
bration panel.
(A), intercept set to zero since it was not
significant; slope 0.906, SE 0.004, r2 ⫽
0.9994. (B), Bland–Altman plot.
Using the 2004 CAP LN24-A survey samples, we found reassay of original MDRD Study specimens on the Roche
the Roche enzymatic assay is comparable to IDMS across enzymatic assay. Nonetheless, our approach is robust.
a range of Scr values from ⬃0.5 to 4.0 mg/dL. Note that First, the CCRL calibration panel was prepared with
CAP LN24 survey samples were used as recommended rigorous techniques, and the MDRD Study stored samples
for trueness control, validating the Roche assay, rather were collected at the same time as the samples used to
than as calibration materials (8 ). Using the CCRL calibra- develop the equation. Second, several factors make mul-
tion panel and adjusting the Beckman CX3 assay for drift tiplication of the regression slopes a maximum likelihood
since the MDRD Study, we calibrated the Scr assay at the estimator with high efficiency. The 2 regressions have
MDRD Study laboratory and reexpressed the 4-variable extremely high correlations (r2 ⫽ 0.987 and 0.994), making
MDRD Study equation for use with creatinine methods the loss of efficiency very small. These extremely high
traceable to an IDMS reference measurement procedure correlations over a wide range of creatinine show the
(8 ). linear relationship between the different assays. The SEs
Based on the Roche method’s performance in the CAP for all comparisons performed in this study were minute.
LN24, creatinine results from the Roche enzymatic Omission of the intercepts is justified by their small magni-
method were considered to be traceable to IDMS values. tude and absence of a statistically or clinically meaningful
These results have consistently showed the Roche enzy- effect on GFR estimates in sensitivity analyses. The final
matic method to give results in agreement with IDMS IDMS-traceable calibration factor of 0.95, relating original
target values. The LN24 samples were verified by the MDRD Study Scr measures to standardized creatinine, is our
NKDEP Laboratory Working Group to have results that best approximation. Regardless of study design, storage
were commutable with those for native clinical samples effects on specimens are possible, but the small difference
for the Roche enzymatic creatinine method (personal between results on thawed MDRD Study specimens assayed
communication, Greg Miller, December 12, 2006). From a a decade later and CAP samples assayed a year later
practical perspective, the CAP LN24 is the only material suggests that any storage effects are small.
currently available for use as a trueness control and is The 2003 CAP survey of 5624 clinical laboratories,
reasonable to use for a clinical laboratory verification. using a fresh-frozen serum reference material with an
These methods may not fully account for measurement assigned value of 79.38 ␮mol/L (0.902 mg/dL) by IDMS,
error. Our study design includes a 2-step approach to showed a range of method-dependent bias across labora-
IDMS-traceable calibration of the MDRD Study samples, tories from an underestimate of 5.28 ␮mol/L (0.06 mg/
using combined results from 2 regressions based on split dL) to an overestimate of 27.28 ␮mol/L (0.31 mg/dL)
samples rather than results of a single regression based on (10 ). This finding suggests that in many laboratories,
Fig. 4. Results for Beckman Synchron

CX3 assays performed at the Cleve-
land Clinic Research Laboratory with
MDRD Study samples during 2004 vs
during the MDRD Study.
(A), intercept set to zero since it was not
significant, slope 1.046, SE 0.002, r2 ⫽
0.9870. (B), Bland–Altman Plot.
trueness control materials from the CAP (LN24 Survey),

Table 2. Summary of equations for use of standardized Scr
clinical laboratories can establish and maintain IDMS-
assays.a
traceable calibrated Scr assays and use reexpressed esti-
Equation 1. Calibration to standardized Scr assay
mating equations, such as the 4-variable MDRD Study
Standardized Scr ⫽ A ⫹ B ⫻ Original MDRD Study Scr ⫽ 0.95 ⫻
Original MDRD Study Scr equation, to report GFR estimates (Table 2, equation 3).
Equation 2: Published MDRD Study equation before calibration Data presented here and in the 2003 CAP survey suggest
GFR ⫽ 186 ⫻ Original MDRD Study Scr⫺1.154 ⫻ age⫺0.203 ⫻ the Roche enzymatic method meets these criteria. Until
1.212 (if black) ⫻ 0.742 (if female) improved equations are developed, it may be appropriate
Equation 3: Reexpression of MDRD Study equation after IDMS- to report a specific numeric result only for estimated GFR
traceable calibrationb ⬍60 mL 䡠 min⫺1 䡠 (1.73 m2)⫺1, as is recommended by cur-
GFR ⫽ 186 ⫻ (Standardized Scr/0.95)⫺1.154 ⫻ age⫺0.203 ⫻ rent guidelines (4 ). Although differentiating higher esti-
1.212 (if black) ⫻ 0.742 (if female)
mates may be useful in research studies, higher values can
⫽ 175 ⫻ Standardized Scr ⫺1.154 ⫻ age⫺0.203 ⫻ 1.212 (if
black) ⫻ 0.742 (if female) be reported as “⬎60 mL 䡠 min⫺1 䡠 (1.73 m2)⫺1” for clinical
a
Standardized Scr traceable to isotope dilution mass spectrometry at NIST.
reports.
Units: GFR, mL 䡠 min⫺1 䡠 (1.73 m2)⫺1; Scr, mg/dL; age, years; race, African
American or non–African American. A is the intercept and B is the slope of the
final regression relationship of Scr using the reference assay vs Scr using the
original MDRD Study or clinical laboratory assay.
We acknowledge assistance from John Eckfeldt, PhD, and
b
Using standard international units for Scr (␮mol/L), equation 3 is as follows: Jillian Celi. This study was supported by grants UO1 DK
GFR ⫽ 30849 ⫻ Standardized Scr⫺1.154 ⫻ age⫺0.203 ⫻ 1.212 (if black) ⫻ 0.742 053869, UO1 DK 067651, and UO1 DK 35073. Presented in
(if female). abstract form at the Annual Meeting of the American
Society of Nephrology in Philadelphia on November 11,
2005. Chronic Kidney Disease Epidemiology Collabora-
standardization of Scr assays will lead to decreased re- tion investigators include the following: from Tufts-New
ported Scr concentrations, requiring redefinition of the England Medical Center, Andrew S. Levey, MD, Lesley A.
reference range. Without reexpression of estimating equa- Stevens, MD, MS, Christopher H. Schmid, PhD, and Lucy
tions, lower reported values for Scr would increase GFR Zhang, MS; from Cleveland Clinic Foundation, Tom
estimates. Reexpression of the MDRD Study equation Greene, PhD, Frederick VanLente, PhD, and Liang Li,
according to the standardized assay enables consistent PhD, from John Hopkins University, Josef Coresh, MD,
interpretation of estimated GFR by use of this equation. PhD, MHS, Jane Marsh, PhD, Brad Astor, PhD, MPH, and
Use of other GFR estimating equations will require reex- Elizabeth Selvin, PhD MPH; from University of Pennsyl-
pression of the equations with standardized Scr. The effect vania, Harold I. Feldman, MD, MSCE, and J. Richard
of standardization of Scr assays on urine creatinine results Landis, PhD; and from the National Institute of Diabetes
has not been studied. Thus, it is not possible at this time and Digestive and Kidney Diseases, John W. Kusek, PhD,
to determine the effect of standardization on measure- Paul W. Eggers, PhD, Thomas H. Hostetter, MD, and
ments and estimates of creatinine clearance. Josephine P. Briggs, MD.
Creatinine calibration is only 1 limitation of the current
estimating equations. Despite calibration, performance of
the MDRD Study equation in populations with higher References
GFRs appears worse than in populations with lower 1. Levey AS, Coresh J, Balk E, Kausz AT, Levin A, Steffes MW, et al.
National Kidney Foundation practice guidelines for chronic kidney
GFRs. Possible reasons for decreased accuracy include
disease: evaluation, classification, and stratification. Ann Intern
reduced creatinine generation attributable to loss of mus- Med 2003;139:137– 47.
cle mass or decreased protein intake, especially in elderly 2. National Kidney Foundation. K/DOQI clinical practice guidelines
and chronically ill persons, greater measurement error for chronic kidney disease: evaluation, classification, and stratifi-
and biological variation in GFR at higher GFR levels, and cation. Am J Kidney Dis 2002;39(Suppl 1):S1–266.
limitations of generalizing equations developed in popu- 3. Levey AS, Eckardt KU, Tsukamoto Y, Levin A, Coresh J, Rossert J,
lations with CKD to populations without CKD (38 ). The et al. Definition and classification of chronic kidney disease: a
Chronic Kidney Disease Epidemiology Collaboration is position statement from Kidney Disease: Improving Global Out-
addressing these issues by use of pooled analysis of comes (KDIGO). Kidney Int 2005;67:2089 –100.
individual patient data and IDMS-traceable calibration of 4. National Kidney Disease Education Program. Creatinine Standard-
stored specimens by standardized methods. ization Program. http://www.nkdep.nih.gov/labprofessionals/index.
htm (accessed February 9, 2006).
NKDEP currently suggests that laboratories should
5. Levey AS, Bosch JP, Lewis JB, Greene T, Rogers N, Roth D. A more
report estimated GFR using the original 4-variable MDRD
accurate method to estimate glomerular filtration rate from serum
Study equation (Table 2, equation 2), even without IDMS- creatinine: a new prediction equation. Ann Intern Med 1999;130:
traceable calibration of Scr assays, recognizing lesser ac- 461–70.
curacy, especially at levels of GFR ⬎60 mL 䡠 min⫺1 䡠 (1.73 6. Levey AS, Greene T, Kusek J, Beck G. A simplified equation to
m2)⫺1 (17 ). With the availability of appropriate IDMS- predict glomerular filtration rate from serum creatinine [Abstract].
traceable calibrator materials from NIST (SRM 967) and J Am Soc Nephrol 2000;11:155A.
7. Coresh J, Astor B, McQuillan G, Kusek J, Greene T, Van Lente F, et 24. Hallan S, Asberg A, Lindberg M, Johnsen H. Validation of the
al. Calibration and random variation of the serum creatinine assay Modification of Diet in Renal Disease Formula for estimating GFR
as critical elements of using equations to estimate the glomerular with special emphasis on calibration of the serum creatinine
filtration rate. Am J Kidney Dis 2002;39:920 –9. assay. Am J Kidney Dis 2004;44:84 –93.
8. Myers G, Miller W, Coresh J, Fleming J, Greenberg N, Greene T, et 25. Lin J, Knight E, Hogan. ML, Singh A. A comparison of prediction
al. Recommendations for improving serum creatinine measure- equations for estimating glomerular filtration rate in adults without
ment: a report from the laboratory working group of the National kidney disease. J Am Soc Nephrol 2003;14:2573– 80.
Kidney Disease Education Program. Clin Chem 2006;52:5–18.
26. Bostom A, Kronenberg F, Ritz E. Predictive performance of renal
9. Mathew TH, Australasian Creatinine Consensus Working Group.
function equations for patients with chronic kidney disease and
Chronic kidney disease and automatic reporting of estimated
normal serum creatinine levels. J Am Soc Nephrol 2002;13:
glomerular filtration rate: a position statement. Med J Aust
2140 – 4.
2005;183:138 – 41.
10. Miller W, Myers G, Ashwood E, Killeen A, Wang E, Thienpont L, et 27. Zuo L, Ma Y, Wang M, Zhou Y, Xu G, Wang H. Application of
al. Creatinine measurement: state of the art in accuracy and glomerular filtration rate estimating equations in Chinese patients
interlaboratory harmonization. Arch Pathol Lab Med 2005;129: with chronic kidney disease. Am J Kidney Dis 2005;45:463–72.
297–304. 28. Gaspari F, Ferrari S, Stucchi N, Centemeri E, Carrara F, Pellegrino
11. Miller G, Eckfeldt J. A Creatinine Accuracy Calibration Verification M, et al. Performance of different prediction equations for esti-
Participant Summary. College of American Pathologists 2005:1–2. mating renal function in kidney transplantation. Am J Transplant
12. ISO Guide 17511. In vitro diagnostic medical devices: measure- 2004;4:1826 –35.
ment of quantities in biological samples: metrological traceability 29. Lamb E, Webb M, Simpson D, Coakley A, Newman D, O’Riordan S.
of values assigned to calibrators and control materials. Geneva: Estimation of glomerular filtration rate in older patients with
International Organization for Standardization, 2003;1–23. chronic renal insufficiency: is the modification of diet in renal
13. Miller WG. Specimen materials, target values and commutability disease formula an improvement? J Am Geriatr Soc 2003;51:
for external quality assessment (proficiency testing) schemes. 1012–7.
Clin Chim Acta 2003;327:25–37. 30. Vervoort G, Willems H, Wetzels J. Assessment of glomerular
14. NCCLS. Preparation and validation of commutable frozen human filtration rate in healthy subjects and normoalbuminuric diabetic
serum pools as secondary reference materials for cholesterol patients: validity of a new (MDRD) prediction equation. Nephrol
measurement procedures: approved guideline. NCCLS document Dial Transplant 2002;17:1909 –13.
C37-A. Wayne, PA: NCCLS, 1999;1– 8.
31. Skluzacek P, Szewc R, Nolan C, Riley D, Lee S, Pergola P.
15. Roche Diagnostics. Creatinine Plus [Package Insert]. Indianapolis,
Prediction of GFR in liver transplant candidates. Am J Kidney Dis
IN: Roche Diagnostics; 1– 4.
2003;42:1169 –76.
16. Cox C. Delta method. In: Armitage P, Colton T, eds. Encyclopedia
of Biostatistics, Vol. 2. New York: Wiley, 1998:1125–7. 32. Ibrahim H, Mondress M, Tello A, Fan Y, Koopmeiners J, Thomas
17. Murthy K, Stevens LA, Stark PC, Levey AS. Variation in serum W. An alternative formula to the Cockcroft-Gault and the Modifi-
creatinine assay calibration: a practical application to glomerular cation of Diet in Renal Diseases formulas in predicting GFR in
filtration rate estimation. Kidney Int 2005;68:1884 –7. individuals with type 1 diabetes. J Am Soc Nephrol 2005;16:
18. Stevens LA, Levey AS. Clinical implications for estimating equa- 1051– 60.
tions for glomerular filtration rate [Editorial]. Ann Intern Med 33. Rigalleau V, Lasseur C, Perlemoine C, Barthe N, Raffaitin C, Liu C,
2004;141:959 – 61. et al. Estimation of glomerular filtration rate in diabetic subjects:
19. Poggio E, Wang X, Greene T, Van Lente F, Hall P. Performance of Cockcroft formula or Modification of Diet in Renal Disease study
the MDRD and Cockcroft-Gault equations in the estimation of equation? Diabetes Care 2005;28:838 – 43.
glomerular filtration rate in health and in chronic kidney disease. 34. Poge U, Gerhardt T, Palmedo H, Klehr H-U, Sauerbruch T, Woitas
J Am Soc Nephrol 2005;16:459 – 66. RP. MDRD equations for estimation of GFR in renal transplant
20. Lewis J, Agodoa L, Cheek D, Greene T, Middleton J, O’Connor D, recipients. Am J Transplant 2005;5:1306 –11.
et al. Comparison of cross-sectional renal function measurements 35. Grubb A, Bjork J, Lindstrom V, Sterner G, Bondesson P, Nyman U.
in African-Americans with hypertensive nephrosclerosis and of A cystatin C-based formula without anthropometric variables
primary formulas to estimate glomerular filtration rate. Am J estimates glomerular filtration rate better than creatinine clear-
Kidney Dis 2001;38:744 –53. ance using the Cockcroft-Gault formula. Scand J Clin Lab Invest
21. Rule A, Larson T, Bergstralh E, Slezak J, Jacobsen S, Cosio F. 2005;65:153– 62.
Using serum creatinine to estimate glomerular filtration rate:
36. Fehrman-Ekholm I, Skeppholm L. Renal function in the elderly
accuracy in good health and in chronic kidney disease. Ann Intern
(⬎70 years old) measured by means of iohexol clearance, serum
Med 2004;141:929 –37.
creatinine, serum urea and estimated clearance. Scand J Clin Lab
22. Gonwa T, Jennings L, Mai M, Stark P, Levey A, Klintmalm G.
Invest 2005;38:73–7.
Estimation of glomerular filtration rates before and after ortho-
topic liver transplantation: evaluation of current equations. Liver 37. Coresh J, Stevens LA. Kidney function estimating equations:
Trans 2004;10:301–9. where do we stand? Curr Opin Nephrol Hyper 2006;15:276 – 84.
23. Froissart M, Rossert J, Jacquot C, Paillard M, Houillier P. Predic- 38. Stevens LA, Coresh J, Greene T, Levey AS. Assessing kidney
tive performance of the MDRD and Cockcroft-Gault equations for function: measured and estimated glomerular filtration rate. New
estimating renal function. J Am Soc Nephrol 2005;16:763–73. Eng J Med 2006;354:2473– 83.
773–780 (2007) Other Areas of
Clinical Chemistry
Hyperhomocysteinemia and Myocardial Expression

of Brain Natriuretic Peptide in Rats
Markus Herrmann,1 Omid Taban-Shoma,1 Ulrich Hübner,1 Anette Pexa,3
Heiko Kilter,2 Natalia Umanskaya,1 Rainer Hans Straub,4 Michael Böhm,2 and
Wolfgang Herrmann1*
Background: Hyperhomocysteinemia (HHcy) has been 308 (192– 429) pg/mg protein, P ⴝ 0.12]. In the Meth
linked to impaired left ventricular function and clinical group, BNP expression was comparable to that of con-
class in patients with chronic heart failure. We hypoth- trols [200 (159 –235) vs 225 (186 –263) pg/mg protein, P ⴝ
esized that HHcy stimulates myocardial brain natri- 0.32]. The percentage of perivascular and interstitial
uretic peptide (BNP) expression and induces adverse collagen and mast cell infiltration were comparable in
left ventricular remodeling. all groups, indicating no adverse cardiac remodeling.
Methods: We randomized 50 rats into 5 groups. Groups Conclusion: Three months of intermediate HHcy stim-
Co1 and Co2 (controls) received a typical diet. Groups ulated increased cardiac BNP expression that was not
Meth, Hcy1, and Hcy2 were fed the same diet supple- accompanied by adverse cardiac remodeling.
mented with 2.4% methionine, 1% homocystine, and 2% © 2007 American Association for Clinical Chemistry
homocystine, respectively. After 12 weeks, we measured
total plasma homocysteine (tHcy) and BNP in plasma Hyperhomocysteinemia (HHcy) and chronic heart failure
and tissue, and we performed histomorphometric (CHF) are frequent problems of elderly individuals and
analyses. often coincide (1 ). Recent epidemiologic studies from
Results: All animals had comparable baseline body Europe and the US consistently found that the frequency
weight [mean (SD) 234 (26) g] and total circulating Hcy of HHcy was ⬎25% in persons ⱖ55 years old (2 ) and
[4.7 (1.7) ␮mol/L]. After 12 weeks of treatment, total increased continuously with age. CHF affects 5 million
circulating Hcy increased in Meth, Hcy1, and Hcy2 [27.3 patients in the US, with ⬃550 000 newly diagnosed cases
(8.8), 40.6 (7.0), and 54.0 (46.0) ␮mol/L, respectively] and each year (3 ). Hospital discharges for CHF increased from
remained unchanged in Co1 and Co2. Serum BNP 399 000 in 1979 to 1 093 000 in 2003, and the 2003 overall
significantly increased in 1 of 10 animals in Meth, 3 of death rate for CHF was 19.7% (3 ). A similar situation
10 animals in Hcy1, and 3 of 10 animals in Hcy2. Median exists in Europe (4 ). Consequently, prevention of CHF by
(25th–75th percentile) BNP tissue concentrations in identifying and modifying risk factors is a major issue.
Hcy1 and Hcy2 were 55% higher than in the correspond- Previous analyses found hypertension, smoking, dia-
ing controls [Co1 vs Hcy1, 225 (186 –263) vs 338 (262– 410) betes mellitus, obesity, and advancing age to be the most
pg/mg protein, P ⴝ 0.05; Co2 vs Hcy2, 179 (107–261) vs important risk factors for CHF (5 ). Recent clinical and
experimental studies suggest that HHcy is also a risk
factor for CHF (1, 6, 7 ). The prevalence of HHcy in CHF
patients is 50%– 60%, depending on the cutoff applied
Departments of 1 Clinical Chemistry and Laboratory Medicine and 2 De-
partment of Internal Medicine III, University Hospital of Saarland, Homburg/
(8, 9 ). Results from the prospective Framingham Heart
Saar, Germany. Study demonstrated an almost doubled incidence of CHF
3
Institute of Physiology, Medical Faculty Carl Gustav Carus, Technical in individuals with a total circulating Hcy (tHcy) above
University Dresden, Dresden, Germany.
4
Department of Internal Medicine I, University Hospital of Regensburg,
Regensburg, Germany.
* Address correspondence to this author at: Wolfgang Herrmann, Abtei-
5
lung für Klinische Chemie und Laboratoriumsmedizin / Zentrallabor, Univer- Nonstandard abbreviations: HHcy, hyperhomocysteinemia; CHF,
sitätsklinikum des Saarland, D-66421 Homburg/Saar, Germany. Fax ⫹496841 chronic heart failure; Hcy, homocysteine; tHcy, total circulating homocysteine;
1630703, E-mail kchwher@uniklinik-saarland.de. NT-proBNP, N-terminal pro-BNP; BNP, brain natriuretic peptide; Co1, Co2,
Received August 14, 2006; accepted January 17, 2007. control group 1 and 2; Meth, methionine group; Hcy1, homocystine 1% group;
Previously published online at DOI: 10.1373/clinchem.2006.077859 Hcy2, homocystine 2% group.
773
774 Herrmann et al.: Homocysteine and BNP
the sex-specific median (1 ). In previous studies from our animals were anesthetized using Ketavet 0.1% (Pharmacia
group, analyzing ⬎1000 individuals, we found a consis- GmbH) and Rompun 2% (Bayer Healthcare). After the
tent relation of tHcy with the severity of CHF (6, 9 ). The final blood sampling, hearts and lungs were isolated,
closest relation could be observed between tHcy and explanted, and rinsed thoroughly to remove adherent
N-terminal pro– brain natriuretic peptide (NT-proBNP). A blood. The left ventricles were isolated, and the left
recent study from Korea confirmed this finding (10 ). In ventricular weight and lung weight were measured using
animal experiments from the group of Joseph and a Sartorius Genius scale (Sartorius AG). Left ventricles
Kennedy (11, 12 ), chronic intermediate HHcy induced were divided into 4 aliquots. One aliquot was fixed with
cardiac hypertrophy, pathological remodeling (11, 13, 14 ), 4% formalin and used for histomorphometric analyses.
and diastolic as well as systolic dysfunction (11, 12 ). The remaining 3 aliquots were shock-frozen with liquid
These findings provide strong evidence for a causal role nitrogen and stored at ⫺80 °C until measurement. One of
of HHcy in the pathogenesis of left ventricular these aliquots was used for the quantification of tissue
dysfunction. BNP concentration.
Brain natriuretic peptide (BNP) and NT-proBNP are
laboratory markers of CHF with a high negative predic- blood characteristics
tive value (15–17 ). They are increased in systolic and We used heparin plasma samples to measure tHcy, BNP,
diastolic dysfunction (18, 19 ) and have been found to be folate, vitamin B12, and creatinine. tHcy was analyzed
useful for differential diagnosis of dyspnea in CHF and with an enzymatic fluorescence polarization immuno-
pulmonary diseases (20 ). Especially in early disease assay on an Axsym analyzer (Abbott). Intra- and interas-
stages, NT-proBNP seems to be more sensitive than say CVs of this method were 4.5% and 5.1%, respectively.
echocardiography (21, 22 ). Moreover, measurements of We quantified BNP by use of the rat BNP-45 ELISA
BNP and NT-proBNP are free from subjective influences reagent set (Gentaur), using a polyclonal antibody spe-
and have good reproducibility (23 ). In this study, we cific for BNP-45. BNP-45 represents the circulating (ac-
investigated myocardial BNP expression and cardiac re- tive) form of BNP in rats and is equivalent to BNP-32
modeling in healthy adult rats after 3 months of moderate in humans. Intra- and interassay CVs of the BNP ELISA
or intermediate HHcy. were 5.3% and 8.5% at concentrations of 100 ng/L. For
this assay, the recovery rate from plasma is 89%, and
Materials and Methods the reference range of healthy untreated rats is below
animals and study design the detection limit of 10 ng/L. Folate and vitamin B12
After 7 days of acclimatization, 50 10- to 12-week-old were detected with chemiluminescence immunoassays
female Wistar rats (180 to 292 g; Charles River GmbH) on an ACS Centaur analyzer (Bayer Healthcare). Intra-
were randomized into 5 groups of 10 animals each: 2 and interassay CVs were 4.0% and 4.4% at a concentra-
control groups (Co1 and Co2) and 3 groups with HHcy. tion of 208 ng/L for vitamin B12, and 5.3% and 5.5% at
Animals were maintained on a 12 h:12 h light-dark cycle a concentration of 5.3 ␮g/L for folate. We measured
in our institutional stable with free access to food and creatinine by use of a colorimetric assay using the alka-
water. HHcy was induced by 3 different dietary regimens. line picrate method on a Hitachi 917 analyzer (Roche
The first HHcy group (Meth) received a diet that was Diagnostics).
enriched with 24 g methionine (Sigma) per kg food
(corresponds to 2.4% methionine). The 2 other groups bnp tissue concentration
were fed with homocystine-enriched diets, containing One deep frozen tissue aliquot per ventricle was thawed
10 g (Hcy1) or 20 g (Hcy2) homocystine (Sigma) per kg and homogenized with a mortar. The homogenized tissue
food (corresponding to 1% and 2% Hcy). Except for the was dissolved in phosphate buffered saline (PAA Labo-
added methionine and homocystine, all diets were of ratories; 0.2 g/L KCl, 0.2 g/L KH2PO4, 8.0 g/L NaCl,
identical composition. Diets were purchased from Al- 1.15 g/L Na2HPO4), and we measured the total protein
tromin. Because of limited space in the stable, we per- concentration using a Bradford assay (Bio-Rad). We quan-
formed experiments in 2 consecutive runs. The first run tified BNP with the same method that was previously
comprised the groups Co1, Meth, and Hcy1; in the second described for the quantification in plasma. The recovery
run, we analyzed Co2 and Hcy2. The study was approved rate from tissue is 85% with this assay. The stability of
by the Institutional Animal Care and Use Committee of tissue BNP-45 at ⫺20 °C is 3 months. Because tissue
the Saarpfalz-Kreis. aliquots had different volumes and masses, BNP values
We monitored body weight, food consumption, and were expressed relative to the protein content of each
fluid intake weekly. We performed capillary blood sam- aliquot.
pling from the retrobulbar plexus before and at the end of
the 12-week feeding period by use of heparin-coated histomorphometry
capillaries. Immediately after blood sampling, plasma We performed morphometric analyses as previously de-
was separated by centrifugation (10 min at 1500g) and scribed by Southern et al. (24 ). After dehydration with
stored at ⫺80 °C until analysis. For blood sampling, increasing alcohol concentrations, aliquots were embed-
ded in paraffin, and 5-␮m sections from 2 locations 100 mmol/L ascorbic acid (Merck), 1.3 mmol/L calcium
␮m apart were prepared. After removal of the paraffin, (Merck), 100 000 IU/L penicillin, and 100 mg/L strepto-
sections were stained with picrosirius red (Sigma), which mycin (Invitrogen Gibco).
is specific for fibrillar collagen (25 ). During the first 2 h of superfusion, we superfused
To avoid subjective influences on morphometric re- tissue slices with culture medium without further stimu-
sults, sections were blinded before analysis. We prepared lation (wash-in period). From 2–13 h, Hcy-conditioned
digital images from 15 microscopic fields per section and medium was applied continuously at concentrations of 0,
analyzed 2 sections per ventricle, 1 per location. The 10, 50, and 200 ␮mol/L to modulate BNP secretion. Each
interstitial collagen fraction was calculated as the percent- Hcy concentration was applied to 8 tissue slices per rat (8
age of red-colored interstitial regions relative to the total slices in 3 rats ⫽ 24 total).
area. For the quantification of perivascular collagen, we We collected superfusate aliquots for 15 min at 2, 3, 5,
analyzed 5 vessels per section with a diameter of 50 –200 7, 9, 11, and 13 h and immediately froze them at ⫺80 °C
␮m and expressed perivascular collagen relative to the until measurements. We quantified BNP with the same rat
vessel luminal area. Serial measurements of the same BNP-45 ELISA reagent set (Gentaur) as used for serum
section revealed a variation of 1% for the analysis of and tissue samples. After the end of the superfusion
perivascular and interstitial collagen. period, tissue slices were homogenized and protein con-
Previous work by Joseph et al. showed a protective role centration was measured using the BCA-Protein assay
of mast cells in Hcy-induced cardiac remodeling (26 ). (Pierce). BNP secretion was expressed as picograms per
Therefore, we also analyzed myocardial mast cell infiltra- microgram protein.
tion using toluidine blue staining (27 ). Sections from both
locations were rehydrated and incubated for 30 min in statistics
0.001% toluidine blue solution (Sigma). In 2 sections per Based on results of a Kolmogorov–Smirnov test, all vari-
ventricle, 1 per location, mast cells (blue) were counted ables, except of tissue BNP, showed gaussian distribution.
and expressed relative to the analyzed area. Results of the descriptive statistics are expressed as mean
(SD). Group comparisons of gaussian-distributed vari-
bnp expression in ex vivo superfused rat ables were performed using 1-way ANOVA with least
myocardium significant difference post hoc test. Tissue BNP, which
To confirm the BNP results in vivo, we performed super- was not gaussian distributed, was compared using a
fusion experiments with freshly prepared rat myocar- Kruskal–Wallis or a Mann–Whitney test. All calculations
dium from three 10- to 12-week-old Wistar rats (mean were performed using the SPSS 11.0 software package
weight 260 g; Charles River) as described previously by (SPSS Inc.).
Jeron et al. (28 ). Briefly, after explantation of the heart and
isolation of the left ventricle, we prepared tissue slices Results
(0.35 mm thick) from the left ventricle (48 slices per heart) animal model
and transferred them immediately to minisuperfusion At the beginning of the study, animals had comparable
chambers (volume 80 ␮L). We performed superfusion for baseline characteristics. Mean (SD) body weight was 234
13 h at a temperature of 37 °C and a flow rate of 66 (26) g and tHcy was 4.7 (1.7) ␮mol/L. After 12 weeks of
␮L/min (1 slice per chamber, 48 chambers in parallel) treatment, Meth, Hcy1, and Hcy2 animals exhibited sig-
with culture medium containing RPMI 1640 (Sigma), 25 nificant HHcy (Table 1). The Meth group developed
mmol/L HEPES (Roth), 5% fetal calf serum (PAA Labo- moderate HHcy, and the Hcy1 and Hcy2 groups had
ratories), 30 ␮mol/L mercaptoethanol (Sigma), 0.57 intermediate HHcy, with the highest Hcy concentrations
Table 1. Characterization of experimental groups and controls at the end of the study.
Variable Co1 Meth Hcy1 Co2 Hcy2
n 10 10 10 7 8
Body weight, g 306 (20) 259 (17)a 322 (17) 345 (28) 254 (28)b
Change in body weight, g 63 (19) 53 (15) 85 (14) 100 (28) 8.5 (25)b
Hcy, ␮mol/L 4.6 (1.5) 27.3 (8.8)a 40.6 (7.0)a 7.0 (1.6) 54.0 (46.0)c
Folic acid, ␮g/L 53 (10) 56 (9) 55 (4) 46 (16) 64 (13)c
Vitamin B12, ng/L 1180 (174) 1132 (248) 1016 (158) 1392 (157) 1177 (353)
Creatinine, ␮mol/L 45.1 (9.7) 56.6 (13.3) 48.6 (11.5) 47.7 (2.7) 36.2 (5.3)c
HR, bpm 405 (26) 401 (43) 393 (27) 391 (23) 413 (32)
Systolic blood pressure, mm Hg 136 (12) 131 (10) 142 (10) 144 (8) 124 (8)
Data are mean (SD). Three of 10 animals from group Co2 had to be excluded because of spontaneously increased Hcy levels ⬎10 ␮mol/L. Two of 10 animals from
group Hcy2 had to be excluded because Hcy concentrations at the end of the treatment were ⬍10 ␮mol/L.
a
P ⬍0.001 vs Co1; b P ⬍0.001 vs Co2; c P ⬍0.05 vs Co2.
in the Hcy2 group. We excluded 3 animals from the Co2

group because they exhibited spontaneously increased
Hcy concentrations ⬎10 ␮mol/L. We excluded 2 animals
from the Hcy2 group because their Hcy concentration at
the end of the treatment was ⬍10 ␮mol/L.
Plasma creatinine was within reference intervals in all
animals [50.4 (9.7) ␮mol/L]. Moreover, Meth, Hcy1, and
Hcy2 groups did not show higher plasma creatinine than
controls (Table 1). All animals exhibited high plasma
concentrations of folate and vitamin B12 that did not
differ between groups (Table 1). Body weight was signif-
icantly lower in Meth and Hcy2 compared with controls,
owing to reduced food intake in these groups. Heart rate
and blood pressure did not differ between groups.
bnp expression
At the end of the study, plasma BNP was below the
detection limit of 10 ng/L in all controls. One Meth Fig. 2. Box plot of median (25th–75th percentile) tissue BNP expres-
animal, 3 Hcy1 animals, and 3 Hcy2 animals showed an sion in relation to total protein concentration after 12 weeks of
increase in circulating BNP (Fig. 1). Because many of the feeding.
animals in the HHcy groups exhibited undetectable BNP Co1, control group 1st run; Co2, control group 2nd run; Meth, methionine group;
Hcy1, homocystine 1% group; Hcy2, homocystine 2% group. *P ⱕ0.05 vs Co1,
concentrations at the end of the study, the calculation of $
P ⫽ 0.12 vs Co2.
mean values was not meaningful. Therefore, we catego-
rized animals as BNP positive and negative. The number
of BNP-positive animals increased with higher circulating Because most of the animals had undetectable BNP
Hcy concentrations (Fig. 1). Introducing categorized BNP plasma concentrations, we quantified tissue concentra-
values in a cross table, Pearson ␹2 test revealed a signifi- tions of BNP in the myocardium (Fig. 2). For animals in
cant relation between circulating Hcy and plasma BNP (␹2 the Hcy1 and Hcy2 groups, BNP tissue concentrations
⫽ 7.89, P ⫽ 0.048). Pooling all controls and all HHcy were ⬃55% higher than in controls. In the Hcy1 group,
animals in 2 groups, we confirmed a higher number of this difference was significant (P ⫽ 0.05). Group compar-
BNP-positive animals in the HHcy group (␹2 ⫽ 5.03, P ⫽ ison between Co2 and Hcy2 revealed only a trend (P ⫽
0.025). 0.12) because of larger interindividual variation of BNP
tissue concentrations in the Hcy2 group.
To confirm the results described above and to demon-
strate causality between circulating Hcy and BNP expres-
sion, we performed superfusion experiments stimulating
freshly prepared rat myocardium slices with increasing
concentrations of Hcy. The mean BNP secretion during
the 11 h of Hcy superfusion increased significantly with
higher Hcy concentrations in the superfusion medium
(Fig. 3).
left ventricular remodeling

The relative weight of the left ventricle in Meth and Hcy1
groups was comparable to the corresponding controls.
Only Hcy2 showed a significant increase, by 15% (Fig.
4A). Similarly, lung weight, as a measure of interstitial
fluid, did not differ from controls in the Meth and Hcy1
groups (Fig. 4B). Only the Hcy2 group exhibited a signif-
icantly higher lung weight, indicating congestion.
Because ventricular and lung weights are relatively
raw measures, we performed morphometric analyses of
interstitial and perivascular collagen. Interstitial collagen
Fig. 1. Distribution of BNP plasma concentrations after 12 weeks of was 2.8%–5.1% and was not different between groups
feeding. (Fig. 5A). Perivascular collagen was comparable in Co1,
Œ plasma BNP ⬍10 ng/L, E increased plasma BNP. Co1, control group 1st run;
Co2, control group 2nd run; Meth, methionine group; Hcy1, homocystine 1%
Co2, Meth, and Hcy1 groups, at 2.0%– 2.1% (Fig. 5B). The
group; Hcy2, homocystine 2% group. Hcy2 group exhibited a perivascular collagen of 2.7%,
Discussion
The main finding of the present study was a significant
induction of cardiac BNP expression by chronically in-
creased tHcy concentrations. Higher BNP concentrations
were not accompanied by an adverse remodeling of the
left ventricle, the main source of BNP. The causality
between HHcy and increased BNP expression was dem-
onstrated by ex vivo superfused left ventricular tissue
slices, which were stimulated with various Hcy concen-
trations. HHcy was induced by 3 different models. Ac-
cording to current concepts, in humans HHcy can be
classified as moderate (12–30 ␮mol/L), intermediate (30 –
100 ␮mol/L), or severe (⬎100 ␮mol/L) (29, 30 ). Based on
this classification, moderate HHcy was induced in the
Meth group by the addition of 2.4% methionine to the
typical diet. A comparable model of moderate HHcy has
been described by Woo et al. (31 ). The mean tHcy in Meth
was comparable with the findings of Woo et al. (27.3 vs
25.3 ␮mol/L). Intermediate HHcy was created with a
homocystine-enriched diet (1%), as published by Joseph
et al. (12, 14 ). tHcy was similar to tHcy values reported by
Joseph et al. [40.6 vs 32.7 ␮mol/L (12 ) and vs 47.1 ␮mol/L
Fig. 3. Mean BNP secretion [95% confidence interval (CI)] of freshly (14 )]. To further increase tHcy, we augmented the homo-
prepared rat myocardium slices during 13 h of superfusion with cystine concentration (2%) in the diet of the 3rd group.
Hcy-enriched RPMI 1640 medium. This diet caused a mean tHcy of 54.0 ␮mol/L; however,
The experiment was performed with 3 animals and 8 slices in each animal. For animals of the Hcy2 group exhibited large interindividual
each Hcy concentration, 24 tissue slices (8 per animal) were analyzed. The P
value above the horizontal line represents the result of the 1-way ANOVA variations in circulating Hcy, and 3 animals had to be
including all experimental groups. **P ⬍0.01 vs 0 ␮mol/L Hcy (least significant excluded because they did not show increased Hcy con-
difference post hoc test).
centrations. Compared with the corresponding controls
(Co2), weight gain was significantly lower, indicating a
⬃35% higher than the corresponding control Co2 (Fig. reduced tolerance to this diet and suggesting that a
5B). Because of large interindividual variation, however, homocystine concentration of 2% is probably the maxi-
this difference was not statistically significant (P ⫽ 0.265). mum that can be administered.
In this study, mean mast cell infiltration was 11–15 This study demonstrates a significant stimulation of
cells/mm2 and was comparable in all groups [see Fig. 1 in BNP expression, a sensitive marker of myocardial dys-
the Data Supplement that accompanies the online ver- function and increased wall stress (21, 22 ), with a high
sion of this article at http://www.clinchem.org/content/ negative predictive value (15–17 ) by intermediate HHcy
vol53/issue4]. (Hcy1 and Hcy2), but not by moderate HHcy (Meth). This
Fig. 4. Relative heart (A) and lung (B)

weights after 12 weeks of feeding.
Co1, control group 1st run; Co2, control
group 2nd run; Meth, methionine group;
Hcy1, homocystine 1% group; Hcy2, homo-
cystine 2% group. **P ⬍ 0.01 vs Co1;
$$
P ⬍ 0.01 vs Co2.
Fig. 5. Histomorphometric analysis of

indicators of left ventricular remodel-
ing after 12 weeks of feeding.
A, Percentage of interstitial collagen. B,
percentage of perivascular collagen. Co1,
control group 1st run; Co2, control group
2nd run; Meth, methionine group; Hcy1,
homocystine 1% group; Hcy2, homocystine
2% group.
BNP-inducing effect was demonstrated in plasma as well Hcy and BNP expression was demonstrated by the ex
as in myocardial tissue and confirms the results of 2 vivo superfusion of freshly prepared rat myocardial slices
previous studies, in which we demonstrated a significant with Hcy-enriched medium. Hcy concentrations of 50 and
positive relation between Hcy and NT-proBNP in CHF 200 ␮mol/L caused a significant increase of BNP secretion
patients (6, 9 ). The changes in BNP were not accompanied by 60% and 140%, respectively.
by adverse cardiac remodeling, as indicated by lack of What is the explanation for the absence of BNP induc-
change in interstitial and perivascular collagen in all tion in the Meth group? An obvious reason could be the
groups. Moreover, the number of mast cells did not differ lower tHcy in these animals. However, even if tHcy was
between controls and HHcy groups. The discrepancy significantly lower than in the Hcy1 and Hcy2 animals, it
between BNP expression and cardiac remodeling is not was clearly above control concentrations. At baseline, all
surprising, since BNP probably represents a more sensi- controls had tHcy ⬍10 ␮mol/L. Therefore, a mean tHcy
tive indicator to detect cardiac changes. Increases of BNP of 27.3 ␮mol/L corresponds to at least triple normal
and NT-proBNP can frequently be found in diastolic values. A more probable explanation is that methionine
dysfunction (18, 19, 32 ) in which the ejection fraction is and homocystine caused HHcy by different biochemical
normal, as well as in mild systolic dysfunction (21, 22 ). In mechanisms with different consequences. Homocystine is
addition, modern concepts consider the heart to be a the disulfide of 2 Hcy molecules. Addition of homocystine
multifunctional and interactive organ that is part of a to the diet probably increases tHcy directly and leads to a
complex network (33, 34 ). Several studies have shown product excess in methionine metabolism, inhibiting
that BNP has antifibrotic and cytoprotective properties transmethylation and thereby the utilization of methio-
(35, 36 ). Obviously, these counterregulatory actions of nine. In contrast, a high intake of methionine induces a
BNP are present before clinical symptoms and before substrate excess in methionine metabolism that is proba-
adverse cardiac remodeling. bly accompanied by increased transmethylation activity.
Contrary to histomorphometry, BNP assessment is also Because hypomethylation is a major mechanism of ad-
free from subjective modification. The sensitivity of cur- verse Hcy effects (37, 38 ), this biochemical difference
rently available rat BNP assays is limited, however. The between the diets could explain the absent BNP induction
ELISA used in the current study could not quantify in the Meth group. However, this explanation is still not
plasma BNP in control animals (all control animals had proven.
undetectable plasma BNP concentrations). Consequently, Contrary to previous work by the group of Joseph et al.
a reference interval could not be defined. Because of the (11–14, 26 ), we could not find any morphologic evidence
limited sensitivity of the rat BNP assay, we could not see for adverse cardiac remodeling. This result was surpris-
potential changes of plasma BNP below the detection ing, because we used exactly the same dietary model and
limit. Therefore, the measurement of tissue BNP has to be the same methods to detect cardiac remodeling. The
considered more reliable, since all animals exhibited con- increase of perivascular and interstitial collagen observed
centrations above the detection limit and in the linear by Joseph et al. was 50%–100% in normotensive rats (12 )
range of the assay. Hcy1 and Hcy2 groups showed a 55% and 100%–500% in hypertensive rats (14 ). Both studies
increase of BNP tissue expression, which supports the included fewer than 10 animals per group. Our investi-
result in plasma. Unchanged BNP tissue concentrations gation comprised 28 HHcy animals and 17 controls. These
were observed in the Meth group. The causality between relatively large numbers lend our study a strong impact.
At present, we cannot find any explanation for the dis- 15. Hobbs FD, Davis RC, Roalfe AK, Hare R, Davies MK. Reliability of
crepancy between the results by Joseph et al. and ours. N-terminal proBNP assay in diagnosis of left ventricular systolic
Because Joseph et al. confirmed their findings in several dysfunction within representative and high risk populations. Heart
2004;90:866 –70.
studies (11–14, 26 ), there is no reason to doubt them.
16. Nielsen LS, Svanegaard J, Klitgaard NA, Egeblad H. N-terminal
However, our experiments were performed in 2 indepen- pro-brain natriuretic peptide for discriminating between cardiac
dent runs, and all variables analyzed consistently sug- and non-cardiac dyspnoea. Eur J Heart Fail 2004;6:63–70.
gested no relevant cardiac remodeling after 12 weeks of 17. McDonagh TA, Holmer S, Raymond I, Luchner A, Hildebrant P,
HHcy. These conflicting findings must be clarified by Dargie HJ. NT-proBNP and the diagnosis of heart failure: a pooled
further studies. analysis of three European epidemiological studies. Eur J Heart
Fail 2004;6:269 –73.
18. Yamaguchi H, Yoshida J, Yamamoto K, Sakata Y, Mano T, Akehi
N, et al. Elevation of plasma brain natriuretic peptide is a hallmark
Supported by the Kompetenznetzwerk Herzinsuffizienz of diastolic heart failure independent of ventricular hypertrophy.
by the BMBF (TP11) to Michael Böhm and Heiko Kilter. J Am Coll Cardiol 2004;43:55– 60.
19. Lubien E, DeMaria A, Krishnaswamy P, Clopton P, Koon J,
References Kazanegra R, et al. Utility of B-natriuretic peptide in detecting
1. Vasan RS, Beiser A, D’Agostino RB, Levy D, Selhub J, Jacques PF, diastolic dysfunction: comparison with Doppler velocity record-
et al. Plasma homocysteine and risk for congestive heart failure in ings. Circulation 2002;105:595– 601.
adults without prior myocardial infarction. JAMA 2003;289: 20. Mueller C, Scholer A, Laule-Kilian K, Martina B, Schindler C, Buser
1251–7. P, et al. Use of B-type natriuretic peptide in the evaluation and
2. Herrmann W, Quast S, Ullrich M, Schultze H, Bodis M, Geisel management of acute dyspnea. N Engl J Med 2004;350:647–54.
J. Hyperhomocysteinemia in high-aged subjects: relation of B- 21. Mueller T, Gegenhuber A, Dieplinger B, Poelz W, Haltmayer M.
vitamins, folic acid, renal function and the methylenetetrahydrofo- Capability of B-type natriuretic peptide (BNP) and amino-terminal
late reductase mutation. Atherosclerosis 1999;144:91–101. proBNP as indicators of cardiac structural disease in asymptom-
3. Thom T, Haase N, Rosamond W, Howard VJ, Rumsfeld J, Manolio atic patients with systemic arterial hypertension. Clin Chem
T, et al. Heart disease and stroke statistics–2006 update: a report 2005;51:2245–51.
from the American Heart Association Statistics Committee and 22. Hammerer-Lercher A, Ludwig W, Falkensammer G, Muller S,
Stroke Statistics Subcommittee. Circulation 2006;113:e85–151. Neubauer E, Puschendorf B, et al. Natriuretic peptides as markers
4. Cleland JG, Khand A, Clark A. The heart failure epidemic: exactly of mild forms of left ventricular dysfunction: effects of assays on
how big is it? Eur Heart J 2001;22:623– 6. diagnostic performance of markers. Clin Chem 2004;50:1174 –
5. Kenchaiah S, Narula J, Vasan RS. Risk factors for heart failure. 83.
Med Clin North Am 2004;88:1145–72. 23. Collinson PO, Barnes SC, Gaze DC, Galasko G, Lahiri A, Senior R.
6. Herrmann M, Kindermann I, Muller S, Georg T, Kindermann M, Analytical performance of the N terminal pro B type natriuretic
Bohm M, et al. Relationship of plasma homocysteine with the peptide (NT-proBNP) assay on the Elecsys 1010 and 2010
severity of chronic heart failure. Clin Chem 2005;51:1512–5. analysers. Eur J Heart Fail 2004;6:365– 8.
7. Kennedy RH, Melchert RB, Joseph J. Cardiovascular effects of 24. Southern FN, Cruz N, Fink LM, Cooney CA, Barone GW, Eidt JF, et
hyperhomocysteinemia in conscious unrestrained rats. Am J Hy- al. Hyperhomocysteinemia increases intimal hyperplasia in a rat
pertens 2006;19:94 –7. carotid endarterectomy model. J Vasc Surg 1998;28:909 –18.
8. Ventura P, Panini R, Verlato C, Scarpetta G, Salvioli G. Hyperho- 25. Junqueira LC, Bignolas G, Brentani RR. Picrosirius staining plus
mocysteinemia and related factors in 600 hospitalized elderly polarization microscopy, a specific method for collagen detection
subjects. Metabolism 2001;50:1466 –71. in tissue sections. Histochem J 1979;11:447–55.
9. Herrmann M, Mueller S, Kindermann I, Guenther L, Koenig J, 26. Joseph J, Kennedy RH, Devi S, Wang J, Joseph L, Hauer-Jensen M.
Boehm M, et al. Plasma B-vitamins and their relation to the Protective role of mast cells in homocysteine-induced cardiac
severity of chronic heart failure. Am J Clin Nutr 2007;in print. remodeling. Am J Physiol Heart Circ Physiol 2005;288:H2541–
10. Cho SE, Sook HK, Shin GJ, Chung WS. The methylenetetrahydro- 45.
folate reductase C677T gene mutation is associated with hyper- 27. Roberts IS, Brenchley PE. Mast cells: the forgotten cells of renal
homocysteinemia, cardiovascular disease and plasma B-type fibrosis. J Clin Pathol 2000;53:858 – 62.
natriuretic peptide concentrations in Korea. Clin Chem Lab Med 28. Jeron A, Kaiser T, Straub RH, Weil J, Riegger GA, Muders F.
2006;44:1070 –5. Myocardial IL-6 regulation by neurohormones: an in vitro superfu-
11. Devi S, Kennedy RH, Joseph L, Shekhawat NS, Melchert RB, sion study. Brain Behav Immun 2003;17:245–50.
Joseph J. Effect of long-term hyperhomocysteinemia on myocar- 29. Durand P, Prost M, Loreau N, Lussier-Cacan S, Blache D. Impaired
dial structure and function in hypertensive rats. Cardiovasc Pathol homocysteine metabolism and atherothrombotic disease. Lab
2006;15:75– 82. Invest 2001;81:645–72.
12. Joseph J, Joseph L, Shekhawat NS, Devi S, Wang J, Melchert RB, 30. Stanger O, Herrmann W, Pietrzik K, Fowler B, Geisel J, Dierkes J,
et al. Hyperhomocysteinemia leads to pathological ventricular et al. DACH-LIGA Homocystein (German, Austrian and Swiss
hypertrophy in normotensive rats. Am J Physiol Heart Circ Physiol Homocysteine Society): Consensus paper on the rational clinical
2003;285:H679 – 86. use of homocysteine, folic acid and B vitamins in cardiovascular
13. Joseph J, Washington A, Joseph L, Kennedy RH. Hyperhomocys- and thrombotic diseases: guidelines and recommendations. Clin
teinaemia-induced atrial remodeling in hypertensive rats. Clin Exp Chem Lab Med 2003;41:1392– 403.
Pharmacol Physiol 2004;31:331–7. 31. Woo CW, Siow YL, O K. Homocysteine activates cAMP-response
14. Joseph J, Washington A, Joseph L, Koehler L, Fink LM, Hauer- element binding protein in HepG2 through cAMP/PKA signaling
Jensen M, et al. Hyperhomocysteinemia leads to adverse cardiac pathway. Arterioscler Thromb Vasc Biol 2006;26:1043–50.
remodeling in hypertensive rats. Am J Physiol Heart Circ Physiol 32. Celinski R, Cholewinski W, Stefaniak B, Tarkowska A. Relationship
2002;283:H2567–74. between plasma BNP levels and left ventricular diastolic function
as measured by radionuclide ventriculography in patients with cytes: importance of cyclic GMP. Cardiovasc Res 2003;57:515–
coronary artery disease. Nucl Med Rev Cent East Eur 2004;7: 22.
123– 8. 36. D’Souza SP, Davis M, Baxter GF. Autocrine and paracrine actions
33. Clerico A, Recchia FA, Passino C, Emdin M. Cardiac endocrine of natriuretic peptides in the heart. Pharmacol Ther 2004;101:
function is an essential component of the homeostatic regulation 113–29.
network: physiological and clinical implications. Am J Physiol
37. Zaina S, Lindholm MW, Lund G. Nutrition and aberrant DNA
Heart Circ Physiol 2006;290:H17–29.
34. Schulze CP, Gielen S, Adams V, Linke A, Möbius-Winkler S, Erbs S, methylation patterns in atherosclerosis: more than just hyperho-
et al. Muscular levels of proinflammatory cytokines correlate with mocysteinemia? J Nutr 2005;135:5– 8.
a reduced expression of insulin like growth factor-I in chronic heart 38. James SJ, Melnyk S, Pogribna M, Pogribny IP, Caudill MA. Eleva-
failure. Bas Res Cardiol 2006;98:267–74. tion in S-adenosylhomocysteine and DNA hypomethylation: poten-
35. Rosenkranz AC, Woods RL, Dusting GJ, Ritchie RH. Antihypertro- tial epigenetic mechanism for homocysteine-related pathology.
phic actions of the natriuretic peptides in adult rat cardiomyo- J Nutr 2002;132(Suppl 8):2361S– 6S.
Technical Briefs
Comparison of Serum Folate Species Analyzed by LC- surveys: the new isotope-dilution liquid chromatogra-
MS/MS with Total Folate Measured by Microbiologic phy–tandem mass spectrometry method (LC-MS/MS)
Assay and Bio-Rad Radioassay, Zia Fazili, Christine M. (2, 3 ) and the traditional Lactobacillus casei microbiologic
Pfeiffer,* and Mindy Zhang (Division of Laboratory Sci- assay (MA) (4, 5 ). The LC-MS/MS measures 5 folate
ences, National Center for Environmental Health, Centers species: 5-methyltetrahydrofolic acid (5CH3THF), FA,
for Disease Control and Prevention, Atlanta, Georgia; 5-formyltetrahydrofolic acid (5CHOTHF), tetrahydro-
* address correspondence to this author at: Division of folic acid (THF), and 5,10-methenyltetrahydrofolic acid
Laboratory Sciences, National Center for Environmental (5,10CH⫽THF). The MA and BR measure total folate
Health, Centers for Disease Control and Prevention, 4770
(TFOL). The 1st objective of this study was to compare the
Buford Highway, NE, Mail Stop F55, Atlanta, GA 30345;
LC-MS/MS method with the traditional MA. The 2nd
fax 770-488-4139, e-mail CPfeiffer@cdc.gov)
objective was to provide equations for converting results
Background: The Bio-Rad QuantaPhase II radioassay from the assay currently used for the NHANES (BR) to
(BR), used for 25 years to measure total folate (TFOL) the new assay for future analyses (MA or LC-MS/MS).
concentrations for the National Health and Nutrition Pristine serum samples (n ⫽ 237) collected as part of
Examination Survey (NHANES), will be discontinued the NHANES from 1999 to 2004 and stored at ⫺70 °C
in 2007. Liquid chromatography–tandem mass spectrome- were analyzed between December 2005 and February
try (LC-MS/MS) or a microbiologic assay (MA) will be 2006 for folate species by the LC-MS/MS and for TFOL by
used in the future. the MA. The BR has analyzed aliquots of the same
Methods: We measured folate species by LC-MS/MS samples for TFOL from 1999 to 2004. These samples are
and TFOL by MA and BR in 327 serum samples. not representative of the US population. They were se-
Results: LC-MS/MS measured 5-methyltetrahydrofolic lected to cover a wide range of TFOL concentrations, as
acid (5CH3THF; 82%), folic acid (FA; 8%), 5-formyltetra- measured by the BR assay. The second set of samples
hydrofolic acid (5CHOTHF; 6%), tetrahydrofolic acid consisted of 100 pristine serum samples obtained from a
(THF; 4%), and 5,10-methenyltetrahydrofolic acid blood bank from January to March 2006. These samples
(5,10CHⴝTHF; 0%). The sum of the folate species cor- were analyzed by all 3 assays in March and April 2006.
related well with TFOL measured by MA (R2 ⴝ 0.97) Data were analyzed with SAS (version 9; SAS Institute)
and BR (R2 ⴝ 0.91). Compared with LC-MS/MS results, and Microsoft Excel with a clinical statistical analysis
MA and BR values were significantly lower (ⴚ6% and plug-in (Analyse-it; Analyse-it Software). Two NHANES
ⴚ29%, respectively); however, these differences were samples were excluded as outliers because of an ex-
concentration dependent. The MA almost completely tremely high concentration of either 5CHOTHF or FA.
recovered folates added to serum samples except for FA The sample sets were analyzed first separately and then
[69% (3%)] and THF [36% (10%)]. The BR underrecov- together. We used the sum of the folate species deter-
ered 5CH3THF [61% (9%)] and 5CHOTHF [38% (14%)] mined by the LC-MS/MS and TFOL determined by the
and overrecovered 5,10CHⴝTHF [234% (32%)]. Multiple MA or BR to compare the methods. Results are presented
linear regression models with log-transformed data only for the combined set because we found no differ-
yielded a good fit for converting BR data to MA or ences between the 2 data sets other than the concentration
LC-MS/MS data and MA data to LC-MS/MS data. ranges (see Fig. 1 in the Data Supplement that accompa-
Conclusions: The good correspondence between the nies the online version of this Technical Brief at http://
sum of folate species determined by LC-MS/MS and www.clinchem.org/content/vol53/issue4). Methods were
TFOL determined by MA makes these 2 assays inter- compared by least-squares regression after log-transforma-
changeable. The BR produces much lower results, on tion of the data to account for nongaussian distribution. The
average, probably because of 5CH3THF underrecovery. concentration-dependent relationship between these assays
The conversion equations provided could be used for prompted us to develop multiple linear regression models
future NHANES time trend analyses. that used a dummy binary variable (0, 1), IND, to account
© 2007 American Association for Clinical Chemistry for intercept differences and an interaction variable (IND ⫻
log10 BR or IND ⫻ log10 MA) to account for differences in
The serum concentration of folate is an important marker slope over 2 discrete concentration ranges (ⱕ45 vs ⬎45
of nutritional status and has been measured for ⬎30 years nmol/L for the BR and ⬍50 vs ⱖ50 nmol/L for the MA).
as part of the National Health and Nutrition Examination The fit of the models was evaluated by comparing the sum
Survey (NHANES). The Bio-Rad QuantaPhase II radioas- of squared residuals (SSR) to the predicted residual sum of
say (BR) has been used since 1991 but will be discontinued squares (PRESS). We tested the recoveries of the MA and BR
in 2007. The introduction of mandatory folic acid (FA) methods (n ⫽ 2 days) by adding each of the 5 folate
fortification in 1998 has contributed to appreciable in- calibrators (Merck Eprova) at 10 nmol/L to a serum pool
creases in serum folate concentrations in the US popula- (21.8 nmol/L TFOL by LC-MS/MS).
tion (1 ); continued monitoring through NHANES is re- The mean concentrations for TFOL, 5CH3THF, and FA
quired. Two assays have been discussed for future were higher in the NHANES set than in the blood bank

782 Technical Briefs
set, whereas the mean 5CHOTHF concentration was concentrations greater than ⬃50 nmol/L TFOL (see Fig.
higher in the blood bank set (Table 1). In the combined set, 2B in the online Data Supplement). After log-transforma-
the major folate forms were 5CH3THF (82%) and FA (8%). tion, the R2 was 0.95 for both method comparisons. The
Most samples had only low FA (⬍2.5 nmol/L) and conversion equations were as follows: log10 LC-MS/
5CHOTHF (⬍10 nmol/L) concentrations. Approximately MS ⫽ 0.1436 ⫹ (1.0435 ⫻ log10 BR) ⫹ (0.51 ⫻ IND) ⫺
half the samples had detectable THF concentrations; only (0.3218 ⫻ IND ⫻ log10 BR), (SSR, 2.0119; PRESS, 2.0682);
10% had THF concentrations ⬎10 nmol/L. log10 MA ⫽ 0.0504 ⫹ (1.0958 ⫻ log10 BR) ⫹ (0.6358 ⫻
Mean and median TFOL concentrations measured by IND) ⫺ (0.4105 ⫻ IND ⫻ log10 BR), (SSR, 1.7889; PRESS,
LC-MS/MS and MA were generally in agreement, but BR 1.8432). These equations were with IND ⫽ 0 for BR results
values were much lower. The sum of folate species and ⱕ45 nmol/L and with IND ⫽ 1 for BR results ⬎45
TFOL determined by the LC-MS/MS and the MA, respec- nmol/L (Fig. 1, B and C).
tively, were highly correlated (R2 ⫽ 0.97), but the latter In 2005 the NIST released a new standard reference
produced slightly lower results, corresponding to a small material for homocysteine and folate in human serum,
but significant negative concentration-dependent differ- SRM 1955 (6 ). We found similar good agreement for
ence of ⫺5.94 nmol/L (95% confidence interval, ⫺7.49 to TFOL between our LC-MS/MS and MA methods
⫺4.39 nmol/L) and a relative difference of ⫺6%. The (level 1, 6.0 vs 5.6 nmol/L; level 2, 13 vs 14 nmol/L;
multiple linear regression model was developed for an level 3, 41 vs 44 nmol/L). TFOL concentrations obtained
MA cutoff of 50 nmol/L because of appreciably increased by the BR were 25%– 40% lower (level 1, 4.5 nmol/L;
FA concentrations above this concentration. The FA con- level 2, 10 nmol/L; level 3, 25 nmol/L).
centration [mean (SD)] at ⬍50 nmol/L was 0.8 (1.4) The lower response of the BR compared with the MA
nmol/L, corresponding to 3% of the TFOL, whereas it was has been known for many years (7 ). The BR was initially
38 (71) nmol/L at ⱖ50 nmol/L, corresponding to 20% of marketed (QuantaPhase I) with calibrator concentrations
the TFOL (see Fig. 2A in the online Data Supplement). to match MA performance, but questions raised by Levine
After log-transformation, R2 was 0.96 for the multiple (8 ) prompted introduction of the QuantaPhase II assay in
linear regression equation: log10 LC-MS/MS ⫽ 0.1439 ⫹ 1993 with spectrophotometrically verified FA calibrator
(0.9193 ⫻ log10 MA) ⫹ (0.0048 ⫻ IND) ⫹ (0.029 ⫻ IND ⫻ concentrations, resulting in a 30% downward shift in
log10 MA), with IND ⫽ 0 for MA results ⬍50 nmol/L and TFOL concentrations. One potential cause for lower TFOL
IND ⫽ 1 for MA results ⱖ50 nmol/L (Fig. 1A). The SSR concentrations by the BR is underrecovery of certain
(1.4042) agreed with the PRESS (1.4377). folate forms. Our recovery experiments with serum sam-
We also obtained good correlation between TFOL ples showed satisfactory recovery of FA [91% (10%)] and
obtained by LC-MS/MS or MA and TFOL obtained by THF [106% (27%)], underrecovery of 5CH3THF [61%
the BR (R2 ⫽ 0.91 for both comparisons); the BR produced (9%)] and 5CHOTHF [38% (14%)], and overrecovery of
much lower results than either the LC-MS/MS or the MA. 5,10CH⫽THF [234% (32%)]. Others have also reported
The mean negative concentration-dependent difference lower BR recoveries of 5CH3THF (60%) (9 ). We obtained
was 29%. The multiple linear regression model was de- satisfactory MA recoveries for 5CH3THF [88% (9%)],
veloped for a 45 nmol/L cutoff for the BR, mainly because 5CHOTHF [120% (9%)], and 5,10CH⫽THF [101% (7%)].
this value represents the upper end of the calibration FA and THF recoveries were 69% (3%) and 36% (10%),
curve, but also because of appreciably increased FA respectively. Because we calibrate the MA with 5CH3THF,
Table 1. Descriptive statistics for serum folate concentrations measured by LC-MS/MS, MA, and BR methods.a
Folate species by LC-MS/MS, nmol/L Total folate, nmol/L
5CH3THF FA 5CHOTHF THF 5,10CHⴝTHF LC-MS/MS MA BR

Combined set (n ⫽ 325)
Mean (SD) 54.4 (40.3) 13.5 (45.2) 2.54 (3.60) 2.54 (4.49) ⬍LOD 73.0 (70.9) 67.0 (62.6) 56.5 (70.0)
Median 40.9 0.81 1.20 0.49 ⬍LOD 47.4 43.5 29.2
Range 5.86–266 0.00–561 ⬍LOD-33.4 ⬍LOD-28.9 ⬍LOD 8.20–642 7.78–572 6.94–642
NHANES set (n ⫽ 225)
Mean (SD) 67.0 (42.0) 18.9 (53.4) 1.87 (3.65) 3.54 (5.07) ⬍LOD 91.3 (78.0) 82.9 (68.8) 72.4 (78.8)
Median 62.3 1.04 0.96 1.83 ⬍LOD 71.0 66.8 42.9
Range 5.86–266 ⬍LOD-561 ⬍LOD-33.4 ⬍LOD-28.9 ⬍LOD 8.20–642 7.78–572 6.97–642
Blood bank set (n ⫽ 100)
Mean (SD) 26.0 (12.4) 1.40 (4.72) 4.04 (2.97) 0.30 (0.47) ⬍LOD 31.7 (15.2) 31.4 (15.7) 20.8 (11.0)
Median 23.7 0.55 3.59 ⬍LOD ⬍LOD 27.9 27.8 18.8
Range 7.61–72.0 0.15–43.6 0.29–13.0 ⬍LOD-2.39 ⬍LOD 12.8–102 9.00–99.0 6.94–93.0
a
Limit-of-detection (LOD) values are as follows: FA, 0.07 nmol/L; 5CHOTHF, 0.05 nmol/L; THF, 2.5 nmol/L; and 5,10CH ⫽ THF, 0.7 nmol/L.
which produces a slightly higher response curve than tween this assay and the LC-MS/MS or MA. The lower MA
FA, we expect to underrecover FA. The lower recovery of recovery of FA is probably the reason for the increased
THF, probably due to oxidative loss of THF during difference between LC-MS/MS and MA results at higher
the 42-h incubation, might be improved by increasing the TFOL concentrations.
ascorbic acid concentration of the medium. Differential
recovery of different folate forms by the BR is likely the We conclude that the new LC-MS/MS method agrees well
reason for the concentration-dependent relationship be- with the traditional MA method. The small difference in
Fig. 1. Least-squares regression plots for TFOL measured in the

combined sample set by 3 methods: LC-MS/MS vs MA (A),
LC-MS/MS vs BR (B), and MA vs BR (C).
The BR or MA is used as the reference point (n ⫽ 325). The concentration-
dependent relationship between these assays prompted our development
of a multiple linear regression model that accounts for a change in slope
and/or intercept over 2 discrete concentration ranges (ⱕ45 vs ⬎45 nmol/L
by the BR and ⬍50 vs ⱖ50 nmol/L for the MA). The fit of the model was
evaluated by comparing the SSR with the PRESS.
results between these 2 methods is not clinically relevant. International Comparison of C-Peptide Measurements,
Irrespective of the assay used for future NHANES moni- Hsiao-Mei Wiedmeyer,1 Kenneth S. Polonsky,2 Gary L. Myers,3
toring, population reference ranges will change to higher Randie R. Little,1* Carla J. Greenbaum,4 David E. Goldstein,1
values; however, MA-determined cutoff values for defi- Jerry P. Palmer,5 (1 Departments of Pathology & Anatom-
ciency (10 ) could be directly applied to the LC-MS/MS ical Sciences and Child Health, University of Missouri-
because of its excellent agreement with the MA. Some Columbia School of Medicine, Columbia, MO; 2 Depart-
advantages of the LC-MS/MS compared with the MA are ment of Medicine, Washington University School of
Medicine, St. Louis, MO; 3 Centers for Disease Control
that it provides information on the different folate species
and Prevention, Division of Environmental Health Labo-
in addition to TFOL and it is less prone to interferences
ratory Sciences, Centers for Environmental Health (F25),
such as antibiotics. BR underrecovery of 5CH3THF, the
Chamblee, GA; 4 Benaroya Research Institute, Seattle,
main circulating form of folate, is likely the major reason WA; 5 University of Washington and VA Medical Center,
for its lower results. A model will be required to convert Seattle, WA; * address correspondence to this author at:
results from the old or the new NHANES method for time Diabetes Diagnostic Laboratory, M767, Departments of
trend analysis. Our model provides an excellent fit over a Pathology & Anatomical Sciences and Child Health, Uni-
wide concentration range. This information may also be versity of Missouri School of Medicine, 1 Hospital Dr.
useful to the international community in that national Columbia, MO 65212; fax 573-884-8823, e-mail:
data generated with the MA can now potentially be LittleR@health.missouri.edu)
compared with US reference ranges.
Background: C-peptide measurement has been widely
used as a marker of insulin secretion in patients with
diabetes. We assessed the comparability of C-peptide
We thank Joseph Jacobson (Battelle, Columbus, OH) results obtained with different methods and by differ-
and Irene Williams (Centers for Disease Control and ent laboratories and determined whether C-peptide re-
Prevention, Atlanta, GA) for technical assistance with the sults could be harmonized by normalization with a
LC-MS/MS and BR methods.
WHO reference reagent or with plasma.
Methods: We sent 16 different heparin plasma samples
References
to 15 laboratories in 7 countries. The samples were
1. Pfeiffer CM, Caudill SP, Gunter EW, Osterloh J, Sampson EJ. Biochemical analyzed with 10 different assay methods. A WHO
indicators of B vitamin status in the U.S. population after folic acid C-peptide standard was also sent to each laboratory and
fortification: results from the National Health and Nutrition Examination
Survey 1999 –2000. Am J Clin Nutr 2005;82:442–50. used to determine the feasibility of normalizing results.
2. Pfeiffer CM, Fazili Z, McCoy L, Zhang M, Gunter EW. Determination of folate To assess the impact of calibrator matrix on the compa-
vitamers in human serum by stable-isotope-dilution tandem mass spectrom- rability of results, we also used the mean results of all
etry and comparison with radioassay and microbiologic assay. Clin Chem
2004;50:423–32. laboratories for 4 of the samples to normalize the
3. Fazili Z, Pfeiffer CM. Measurement of folates in serum and conventionally remaining sample results.
prepared whole blood lysates: application of an automated 96-well plate Results: Between-laboratory variability increased with
isotope-dilution tandem mass spectrometry method. Clin Chem 2004;50:
2378 – 81. increasing C-peptide concentrations. Normalization of
4. O’Broin S, Kelleher B. Microbiological assay on microtitre plates of folate in results with WHO reference reagents did not improve
serum and red cells. J Clin Pathol 1992;45:344 –7. comparability, but normalization with samples signifi-
5. Molloy A, Scott JM. Microbiological assay for serum, plasma, and red cell cantly improved comparability among laboratories and
folate using cryopreserved, microtiterplate method. Methods Enzymol 1997;
281:43–53. methods. The 95% confidence interval estimate for the
6. Satterfield MB, Sniegoski LT, Sharpless KE, Welch MJ, Hornikova A, Zhang SD for the lab/method effect (0.0 – 0.061) using sample-
N-F, et al. Development of a new standard reference material: SRM 1955 normalized values did not overlap with the 95% CI
(homocysteine and folate in human serum). Anal Bioanal Chem 2006;385:
612–22. estimate with the raw data (0.090 – 0.225).
7. Raiten DJ, Fisher KD, eds. Assessment of folate methodology used in the Conclusions: C-peptide results generated by different
Third National Health and Nutrition Examination Survey (NHANES III, 1988 – methods and different laboratories do not always agree,
1994). J Nutr 1995;125:1371S–98S.
8. Levine S. Analytical inaccuracy for folic acid with a popular commercial
especially at higher concentrations of C-peptide. These
vitamin B12/folate kit [Letter]. Clin Chem 1993;39:2209 –10. data support the concept of using a single laboratory for
9. Blackmore S, Pfeiffer C, Hamilton MS, Lee A. Recoveries of folate species multisite studies and support efforts to harmonize C-
from serum pools sent to participants of the UK NEQAS Haematinics
Scheme in February and March 2004. Clin Chim Acta 2005;355:S459.
peptide measurements by use of calibrators prepared in
10. Life Sciences Research Office. Assessment of the folate nutritional status the sample matrix.
of the U.S. population based on data collected in the Second National © 2007 American Association for Clinical Chemistry
Health and Nutrition Survey, 1976 –1980. Prepared for the Center for Food
Safety and Nutrition, Food and Drug Administration. Rockville, MD: Federa-
tion of American Societies for Experimental Biology, 1984. Human C-peptide provides an accurate assessment of
residual beta-cell function and thus has been widely used
Previously published online at DOI: 10.1373/clinchem.2006.078451 as a marker of insulin secretion in patients with diabetes
(1, 2 ). Some studies have also suggested that C-peptide is
biologically active (3 ) and may play a role in preventing ampoule of WHO IRR 84/510 standard (10 ␮g) in 25 mL
and possibly reversing some chronic complications of of distilled water (8, 9 ). The resulting solution (400 ng/mL
type 1 diabetes (4, 5 ). C-peptide is also important in the or 132.4 nmol/L) was shipped on dry ice along with the
diagnosis of insulinoma/endogenous hyperinsulinemia samples. Each laboratory serially diluted the standard solu-
(6 ). tion with their own assay buffer to make 4 preparations that
Despite the fact that measurement of C-peptide has were then analyzed in each analytical run.
become increasingly important, the accuracy and be- To assess the impact of the calibrator matrix on the
tween-laboratory comparability of C-peptide results have comparability of results, 4 of the samples were used to
not been thoroughly evaluated. In 2002, the National normalize the results from the remaining 12 samples. The
Institute of Diabetes and Digestive and Kidney Diseases mean of all laboratories’ results for each sample was used
(NIDDK) organized a C-peptide standardization commit- as the assigned value for each of these sample calibrators.
tee and funded an international comparison study of To compare the effect of normalization with WHO stan-
C-peptide assays to assess the comparability of C-peptide dards and normalization with samples, normalization
results within and among laboratories involved in diabe- was applied to the same 12 samples for both analyses. The
tes-related long-term research studies. The goal of the 4 samples used as sample matrix calibrators were there-
study was to assess the degree of comparability of C- fore not included in the normalization with WHO stan-
peptide results among methods and laboratories and to dards.
determine whether C-peptide results could be normalized We performed all data analyses with Excel and SAS.
to make comparing and combining data from different Most laboratories in this study reported C-peptide results
laboratories and studies more feasible. The comparison in nanograms per milliliter, but a few reported results in
study was coordinated by a central laboratory at the SI units (nmol/L). All C-peptide results were converted to
University of Missouri-Columbia and was overseen by SI units (nmol/L) before data analyses. Because scatter
the C-peptide Standardization Committee. The members plots of the data indicated increasing variability in re-
of the Committee and the participating laboratories are sponses with increasing C-peptide, the model assumed
listed in Appendix A. constant CV. Weighted regression methods were used
The study was conducted by the Diabetes Diagnostic with weights inversely proportional to the variance. We
Laboratory (DDL) at the University of Missouri-Colum- performed iterative re-weighting because the variance
bia. The protocols were approved by the University of had to be estimated from the modeled regression results.
Missouri Health Sciences Institutional Review Board and For each laboratory/method, 2nd-degree equations
all participants gave written informed consent. Three were fitted by use of laboratory/methods values for the
laboratories used more than one assay for C-peptide WHO standards on y and the “true” values for the
analysis, which resulted in a total of 20 laboratory meth- standards on x; weights based on constant CV were used
ods being evaluated. Five laboratories used the Linco RIA; in the weighted regression analyses. In all but one case the
1 additional laboratory used Linco C-peptide antibody coefficient for the squared term was not significantly
with in-house isotope and precipitating reagents. Four different from zero, and in that case the fitted quadratic
laboratories used the Diagnostic Product Corp. (DPC) curve visually showed minimal curvature. Therefore, we
Immulite chemiluminescence immunoassay system, and 3 performed normalization with simple linear relationships.
laboratories used the DPC RIA kit. Two laboratories used The weighted regression analyses were run again using
Tosoh Bioscience AIA-600II, an automated immunoassay the normalized values and the same analyses were per-
system. The other methods—Liaison Immunoluminomet- formed when we used samples as calibrators.
ric assay (ILMA) (DiaSorin), Biochem Pharma RIA, Visual inspection of the data showed that one labora-
Shionogi RIA, Dako ELISA, and PerkinElmer Time-re- tory’s results (using the DPC Immulite method) did not
solved fluoroimmunoassay (FIA)—were each used by 1 show consistent increases with increasing C-peptide con-
laboratory. centrations and was therefore considered an outlier (Fig.
Heparin plasma samples were collected from 8 nondi- 1a). This laboratory’s results were therefore excluded
abetic volunteers after they had fasted overnight and 60 from all further evaluation.
min after they had consumed a standard meal (Boost™, The assay survey of participating laboratories showed
Novartis Medical Nutrition). The samples were shipped that 6 of the C-peptide assay methods were already
frozen on dry ice and kept at ⫺70 °C in the laboratories calibrated to WHO standard IRR lot 84/510 (DPC RIA,
before analysis. Laboratories were instructed to analyze DPC Immulite, Linco RIA, Tosoh AIA, DiaSorin ILMA,
the specimens on specified days, in the same manner as and in-house RIA). Two of the laboratories reported that
they would analyze clinical specimens, and to provide a their methods were not calibrated to WHO standards
single result for each specimen. (PerkinElmer FIA, Shionogi RIA), and no information was
Along with patient specimens, the laboratories received available for 2 of the methods (Dako ELISA, Biochem
a WHO C-peptide reference reagent (standard) to assess Pharma RIA).
its commutability and to determine the feasibility of C-peptide results before and after normalization with
normalizing results to this standard (7 ). This standard WHO standard (Fig. 1A and 1B) showed that the 95%
was prepared in the central laboratory with a vial of confidence interval (CI) estimate for the SD for the lab/
C-peptide stock solution into which was dissolved 1 method effect (0.172– 0.456) overlapped with the 95% CI
Fig. 1. Between-laboratory/method comparison of C-peptide in heparin plasma.

Results are shown before normalization (A), after normalization with WHO standard (B), and after normalization with sample matrix calibrators (C). Specimens are shown
in order of C-peptide concentration. Each different symbol represents a different method. The Q symbol shown only in (A) shows results for the laboratory/method that
was considered an outlier. This laboratory’s data was not included in the normalization analyses shown in (B) and (C).
estimated with the raw data (0.090 – 0.225). This result Recently, isotope-dilution liquid chromatography-mass
suggests that WHO normalization was ineffective in re- spectrometry has been proposed as a reference measure-
ducing the variability of C-peptide results within and ment procedure for serum C-peptide (14, 15 ). Further
among the lab/methods. studies are needed to determine if this method may be
As with the WHO normalization, weighted linear re- useful in standardizing measurements of plasma/serum
gression analyses were used for each of the assay methods C-peptide.
for normalization with the 4 patient samples. Results (Fig. It is important that comparability of C-peptide results
1C) showed that the 95% CI estimate for the SD for the between laboratories be addressed before the initiation of
lab/method effect (0.0 – 0.061) using normalized values large-scale trials involving multiple laboratories perform-
did not overlap with the 95% CI estimated with the raw ing C-peptide analyses. The data presented here support
data (0.090 – 0.225). This result shows that normalization the concept of using a single core laboratory for such
with sample calibrators was effective in reducing variabil- studies. Our data also illustrate the need for efforts to
ity of C-peptide results. standardize C-peptide measurements by use of materials
The importance of measuring C-peptide has increased prepared in sample matrix.
significantly in recent years with the evidence from the
DCCT that higher C-peptide concentrations are associated Appendix
with improved hemoglobin A1c concentrations, less hy- Members of the C-Peptide Standardization
poglycemia, and less retinopathy and nephropathy (10 ).
Committee and Participating Laboratories
Furthermore, stabilization of C-peptide concentrations is
being used as a measurable endpoint for immunomodu-
NIDDK C-Peptide Standardization Committee. David Gold-
latory trials in type 1 diabetes (11, 12 ). Our data show that
stein (University of Missouri), Gary Myers (Centers for
a wide variety of C-peptide assay methods are being used
Disease Control and Prevention), Jerry Palmer (University
by laboratories worldwide, and that the C-peptide results
of Washington), Kenneth Polonsky (Washington Univer-
generated by the different methods, and even in some sity), Judith Fradkin (NIDDK), Lisa Spain (NIDDK), Ran-
cases by different laboratories using the same method, are die Little (University of Missouri), Hsiao-Mei Wiedmeyer
not always comparable. This finding has important impli- (University of Missouri).
cations for studies in which C-peptide results from differ-
ent laboratories involved in multicenter studies are com- Participating Laboratories. Anette Ziegler, Kerstin Koc-
pared or combined for data analyses and for clinicians zwara, Diabetes Research Institute Munich (Germany);
following patients over time if samples are tested at Paolo Pozzilli, University Campus Bio-Medico (Italy);
different laboratories. Charlotte Becker, University Hospital Malmö (Sweden);
Normalization using specimens achieves a modest im- Ezio Bonifacio, San Raffaele Hospital (Italy); Merete
provement in comparability but the harmonization may Frandsen, Thomas Mandrup-Poulsen, Steno Diabetes
be inadequate to fulfill the requirements of clinical trials. Center (Denmark); Anders Isaksson, Mona Landin-Ols-
Our data reaffirm the difficulty in achieving standardiza- son, Lund University (Sweden); Armando Mendez, Linda
tion of C-peptide measurements because, like many Jones, University of Miami (FL); Jean Bucksa, Vicky
polypeptides, C-peptide does not appear to behave the Makky, University of Minnesota Medical Center, Fair-
same way in pure calibrators as it does in patient samples, view (MN); Veronica Luzzi, Gene Sherrow, Washington
i.e., the commutability of pure standards is poor (13, 14 ). University (MO); Liz Rinehart, Linco Diagnostic Services,
Inc. (MO); Jon Nakamoto, Anne Caston-Balderrama, Mutant-Enriched PCR and Allele-Specific Hybridiza-
Quest Diagnostics Nichols Institute (CA); Vinod Gaur, tion Reaction to Detect K-ras Mutations in Stool DNA:
Northwest Lipid Metabolism and Diabetes Research Lab- High Prevalence in a Large Sample of Older Adults,
oratory, University of Washington (WA); Alethea Tennill, Ulrike Haug,1* Timo Hillebrand,2 Peter Bendzko,3 Michael
University of Missouri (MO); Akira Shimada, Taro Ma- Löw,1 Dietrich Rothenbacher,1 Christa Stegmaier,4 and Her-
ruyama, Keio University (Japan); Spiros Fourlanos, Royal mann Brenner1 (1 Division of Clinical Epidemiology and
Melbourne Hospital (Australia). Aging Research, German Cancer Research Center, Heidel-
berg, Germany; 2 AJ Innuscreen GmbH, Berlin, Germany;
3
Invitek GmbH, Berlin, Germany; 4 Saarland Cancer Reg-
Grant/funding support: This project was funded by the istry, Saarbrücken, Germany; * address correspondence to
Centers for Disease Control and Prevention through a this author at: German Cancer Research Center (DKFZ),
funding contract (No.200200409985), supported by the Division of Clinical Epidemiology and Aging Research,
NIDDK Special Statutory Funding Program for Type 1
Bergheimer Strasse 20, 69115 Heidelberg, Germany; fax
Diabetes Research.
Disclosures: None declared 49-(0)6221-54-8142, e-mail u.haug@dkfz.de)
Acknowledgments: We thank Judith Fradkin and Lisa
Spain for their support. We also thank Alethea Tennill, Background: Testing for mutant K-ras in stool has been
Curt Rohlfing, and Donghua Huang of the Diabetes proposed for the detection of pancreatic and colorectal
Diagnostic Laboratory, University of Missouri, for their cancer (CRC). Different analytical techniques have been
technical assistance with this project, as well as Richard developed, but studies of this biomarker in the general
Madsen, University of Missouri, for statistical consulta- population are lacking. We investigated the prevalence
tion. and potential determinants of mutant K-ras in stool in a
large sample of unselected older adults and assessed the
References
1. Polonsky, KS, Licinio-Paixao J, Given BD, Pugh W, Galloway RJ, Karrison T, et
association with colonoscopic findings.
al. Use of biosynthetic human C-peptide in the measurement of insulin Methods: In stool samples from 875 older adults (age
secretion rates in normal volunteers and type 1 diabetic patients. J Clin range 50 –75 years) participating in a large-scale popula-
Invest 1986;77:98 –105.
2. Kjems LL, Volund A, Madsbad S. Quantification of beta-cell function during tion-based cohort study, we used mutant-enriched PCR
IVGTT in Type II and non-diabetic subjects: assessment of insulin secretion and allele-specific hybridization reaction to analyze
by mathematical methods. Diabetologia 2001;44:1339 – 48.
3. Wahren J, Jornvall H. C-peptide makes a comeback. Diabetes Metab Res mutations in codons 12 and 13 of the K-ras gene. We
Rev 2003;19:345–7. assessed the association between mutant K-ras in stool
4. Johansson JW, Wallberg-Henriksson H, Linde B, Fernqvist-Forbes E, Zierath
JR. C-peptide revisited–new physiological effects and therapeutic implica- and risk factors for gastrointestinal cancer sites, exocrine
tions. J Intern Med 1996;240:115–24. pancreatic insufficiency determined by fecal pancreas
5. Johansson BL, Borg K, Fernqvist-Forbes E, Odergren T, Remahl S, Wahren J.
C-peptide improves autonomic nerve function in IDDM patients. Diabetologia
elastase 1, and colonoscopic findings.
1996;39:687–95. Results: The overall prevalence of mutant K-ras in stool
6. Saddig C, Bender R, Starke A. A new classification plot for the C-peptide was 8% (95% confidence interval 6%–10%). There was a
suppression test. JOP 2002;3:16 –25.
7. Miller WG, Myers GL, Rey R. Why commutability matters. Clin Chem tentative association between increased fecal pancreas
2006;52:553– 4. elastase 1 and mutant K-ras in stool (P ⴝ 0.09). Patients
8. WHO synthetic C-peptide reference material 84/510, National Institute for
Biological Standards and Control (UK). with advanced colorectal neoplasia diagnosed within 2
9. Bristow AF, Gaines Das RE. WHO International reference reagents for human years after stool collection (24 with advanced adenomas,
proinsulin and human insulin C-peptide. J Biol Stand 1988;16:179 – 86.
10. Steffes MW, Sibley S, Jackson M, Thomas W. Beta-cell function and the
7 with CRC) all tested negative.
development of diabetes-related complications in the diabetes control and Conclusion: The proposed assay identifies mutant K-
complication trial. Diabetes Care 2003;26:832– 6. ras in stool at a higher prevalence than has been
11. Diabetes Prevention Trial–Type 1 Diabetes Study Group. Effects of insulin in
relatives of patients with type 1 diabetes mellitus. N Engl J Med 2002;346: reported for other analytical techniques. Our findings
1685–91. do not support the use of this assay for CRC screening,
12. European Nicotinamide Diabetes Intervention Trial (ENDIT) Group. European
Nicotinamide Diabetes Intervention Trial (ENDIT): a randomized controlled but its potential use for early detection of pancreatic
trial of intervention before the onset of type 1 diabetes. Lancet 2004;363: cancer (in combination with other markers) requires
925–31.
13. Stockl D, Franzini C, Kratochvilla J, Middle J, Ricos C, Thienpont LM. Current
further investigation.
stage of standardization of measurements of specific polypeptides and © 2007 American Association for Clinical Chemistry
proteins discussed in light of steps needed towards a comprehensive
measurement system. Eur J Clin Chem Clin Biochem 1997;35:719 –32.
14. Rodrı́guez-Cabaleiro D, Stöckl D, Kaufman JM, Fiers T, Thienpont LM. Given that somatic mutations of the K-ras gene are ob-
Feasibility of standardization of serum C-peptide immunoassays with iso- served in ⬃40% of colorectal cancers (CRCs) (1–3 ) and in
tope dilution-liquid chromatography-tandem mass spectrometry. Clin Chem
2006;52:1193–5.
⬎80% of pancreatic cancers (4 ), mutant K-ras in stool
15. Rogatsky E, Balent B, Goswami G, Tomuta V, Jayatillake H, Cruikshank G, et specimens has been proposed as a potential component of
al. Sensitive quantitative analysis of C-peptide in human plasma by 2-di- marker combinations aimed at the early detection of these
mensional liquid chromatography-mass spectrometry isotope-dilution assay.
Clin Chem 2006;52:872–9. cancers (5–10 ). Conclusions drawn from existing studies
must be viewed cautiously, however, because most of the
Previously published online at DOI: 10.1373/clinchem.2006.081570 study populations were highly selective and rather small
(5–10 ), and different analytical techniques were used and
may have led to different detection rates (5 ). For example, Fecal pancreas elastase 1 was measured in nonproc-
assays based on the amplification of DNA by mutant- essed stool samples with a commercially available ELISA
enriched PCR are supposed to result in higher rates of (ScheBo® Biotech AG). The results were categorized ac-
K-ras-positive samples than other assays (11 ). Although cording to an established cutoff for exocrine pancreatic
higher detection rates may optimize sensitivity, another insufficiency of ⬍200 ␮g elastase 1/g stool (13 ).
important question is whether specificity may be affected, All laboratory analyses were done in a blinded fashion.
i.e., whether diminutive amounts of mutant K-ras in stool Overall, 875 participants of the ESTHER study were
can also be detected in average-risk people or, preferen- included in the statistical analyses (in 19 participants stool
tially, in people with certain risk factors (e.g., as proposed DNA was not amplifiable). Characteristics of the study
for smokers) (12 ). To clarify these questions, which are population are shown in Table 1A. Group A (n ⫽ 535) and
highly relevant in regard to potential population-wide B (n ⫽ 340) did not differ from each other or from the
testing, large-scale investigations in the screening setting whole study population of the ESTHER study, except that
are needed. more participants of group A reported a 1st-degree rela-
We investigated prevalence and potential determinants tive with CRC.
of mutant K-ras in stool (analyzed by mutant-enriched Mutations in the K-ras gene were identified in 70 of 875
PCR and allele-specific hybridization reaction) in a large stool samples, i.e., the overall prevalence was 8% (95%
sample of unselected older adults and assessed the asso- confidence interval, 6%–10%). In group A and B, the
ciation with colonoscopic findings. prevalence was 7% and 10%, respectively. Type and
The study was carried out in the context of the ESTHER frequency of identified K-ras gene mutations are detailed
study (epidemiological investigations of the chances of in Table 1B. For 70% of samples that tested positive for
preventing, recognizing early, and optimally treating mutant K-ras, enough DNA material was left for gene
chronic diseases in an elderly population), a large-scale sequencing, which always confirmed the result of the
population-based cohort study. Details of ESTHER, which hybridization assay.
has been approved by local and regional ethics commit- Regarding potential determinants, there was a tenta-
tees, have been reported elsewhere (13 ). Briefly, 9953 tive, albeit not statistically significant (P ⫽ 0.09), associa-
tion with exocrine pancreatic insufficiency (Table 1A).
inhabitants of southwest Germany ages 50 to 75 years
In group A, 441 participants underwent complete
were recruited by their general practitioner in the context
colonoscopy, reaching the cecum. Advanced colorectal
of a general health examination. After informed consent
neoplasia was present in 31 participants (including 24
was obtained, all participants were asked to mail a stool
patients with advanced adenomas, i.e., adenomas that
sample to the study center and to fill out a standardized
were at least 1 cm in diameter, adenomas with villous
questionnaire. In addition, medical data were recorded
components, or high-grade dysplasia, and 7 patients with
from the patients’ charts. As part of the 2-year follow-up
invasive CRC). None of these participants tested positive
(response rate 96%), information was collected on
for mutant K-ras in stool. The same applies to the 25
colonoscopies performed since baseline recruitment. participants bearing hyperplastic polyps. Among partici-
Overall, stool samples from 894 ESTHER study partic- pants with nonadvanced adenomas (n ⫽ 50) and with
ipants were analyzed for mutant K-ras. Participants were unspecified polyps (i.e., without histological examination,
divided into 2 subgroups. Group A consisted of persons n ⫽ 35), K-ras mutations were found in 1 and 3 patients,
who underwent colonoscopy between baseline and 2-year respectively. The highest rate of mutant K-ras (7.5%,
follow-up and for whom the colonoscopic findings were 22/293) was detected in participants showing a negative
available. Group B was quasi-randomly selected, with the result at colonoscopy.
only selection criterion being a high amount of initially Testing a large sample of unselected older adults for
collected stool, which was assumed to be a random mutant K-ras in stool with the described assay yielded a
variable. This selection criterion was chosen to allow prevalence of this marker of ⬃8% in the whole sample as
additional measurement of fecal pancreas elastase 1 and well as among individuals with negative colonoscopy
subsequent comparison with the K-ras status. results. Gene sequencing (performed for 70% of positive
On arrival at the laboratory, stool samples were stored results) always confirmed the result of the hybridization
at ⫺70 °C. After baseline recruitment was finished, DNA assay.
from 200 mg stool was extracted using the Invisorb® Spin With the use of this assay the observed prevalence of
Stool DNA Kit (Invitek). To analyze codons 12 and 13 of mutant K-ras in stool was much higher than that observed
the K-ras gene, 5 ␮L of the DNA extraction product (stool with alternative analytical techniques. Covering the same
DNA dissolved in elution buffer), derived from ⬃5 mg of potential mutations but using another assay, the largest
initially collected stool, was used for the analyses. Stool study in the field, which was aimed at the evaluation of
DNA was amplified by mutant-enriched PCR, and an fecal DNA testing for the detection of colorectal neo-
allele-specific hybridization reaction assay was carried plasms, reported a prevalence of only 1.5% (22/1423)
out according to a published protocol (8, 14 ). If enough among persons with negative colonoscopy results (15 ).
DNA material was left from the stool sample initially Given that the latter study applied somewhat stricter
designated for K-ras analysis, positive test results were inclusion criteria compared with the ESTHER study,
confirmed by gene sequencing. potential selection bias should also be considered when
Table 1. Results of testing for mutant K-ras in stool among 875 participants of the ESTHER study.
A. Occurence of mutant K-ras in stool according to age, sex, body mass index, cigarette smoking status, alcohol
consumption, 1st-degree relative with colorectal or pancreatic cancer, and fecal pancreas elastase 1 concentration.
No. of subjects, No. of subjects with K-ras mutation,
n (column-%), n ⴝ 875 n (row-%), n ⴝ 70 P valuea
Age group, years
50–54 137 (16) 11 (8)
55–59 148 (17) 16 (11)
60–64 256 (29) 20 (8)
65–69 211 (24) 15 (7)
70–75 123 (14) 8 (7) P ⫽ 0.70
Sex
Female 467 (53) 38 (8) P ⫽ 0.87
Male 408 (47) 32 (8)
Body mass index, kg/m2
ⱕ25 252 (29) 21 (8) P ⫽ 0.91
⬎25 to 30 392 (45) 30 (8)
⬎30 222 (26) 19 (9)
Alcohol consumption, g/week
None 207 (26) 19 (9) P ⫽ 0.56
⬍60 244 (31) 17 (7)
60–140 211 (27) 22 (10)
⬎140 122 (16) 9 (7)
Cigarette smoking status P ⫽ 0.76
Never smoker 420 (49) 32 (8)
Former smoker 313 (36) 23 (7)
Current smoker 127 (15) 12 (9)
1st-degree relative with
Colorectal cancer 102 (12) 6 (6) P ⫽ 0.40
Pancreatic cancer 13 (1) 1 (8) P ⫽ 1.00
Neither of themb 762 (87) 63 (8)
Levels of fecal pancreas elastase 1
(determined in group B, n ⫽ 336)
⬍200 ␮g/g stool 39 (12) 7 (18) P ⫽ 0.09
ⱖ200 ␮g/g stool 297 (88) 27 (9)
B. Results of the analyses for mutant K-ras in stool: base-pair changes, resulting amino acid changes, and absolute
frequency of the mutations.
Base pair changec Amino acid changed Absolute frequency
Mutations on codon 12
2G to C Gly12 to Ala 18
2G to T Gly12 to Val 17
2G to A Gly12 to Asp 6
1G to T Gly12 to Cys 1
1G to A Gly12 to Ser 1
Mutations on codon 13
5G to A Gly13 to Asp 21
5G to C and 6C to T Gly13 to Val 3
Mutations on codon 12 and codon 13
2G to T and 5G to A Gly12 to Val, Gly13 to Asp 2
1G to A and 5G to A Gly12 to Ser, Gly13 to Asp 1
a
Based on ␹2 statistics; if expected cell numbers were ⬍5, the Fisher exact test was used.
b
Reference group.
c
1G and 2G are the first 2 bases of codon 12, 5G is the 2nd base of codon 13, 6C is the 3rd base of codon 13 (wild-type).
d
Changed codon indicated by superscript.
comparing the findings between both studies. However, indicated that selection bias is unlikely to explain the
additional analyses applying comparably strict inclusion observed large differences in prevalence of mutant K-ras
criteria to ESTHER study participants (data not shown) in stool. As already suggested by other authors (11 ), an
alternative explanation may be that assays based on References

mutant-enriched PCR result in a higher rate of K-ras- 1. Bos JL, Fearon ER, Hamilton SR, Verlaan-de Vries M, van Boom JH, van der
Eb AJ, et al. Prevalence of ras gene mutations in human colorectal cancers.
positive samples. This suggestion is further supported by Nature 1987;327:293–7.
a study by Kopreski et al., in which 27% of 105 individuals 2. Capella G, Cronauer-Mitra S, Pienado MA, Perucho M. Frequency and
with negative colonoscopy results had positive results for spectrum of mutations at codons 12 and 13 of the c-K-ras gene in human
tumors. Environ Health Perspect 1991;93:125–31.
mutant K-ras in plasma samples tested with an assay 3. Scott N, Quirke P. Molecular biology of colorectal neoplasia. Gut 1993;34:
based on mutant-enriched PCR (16 ). 289 –92.
To estimate the potential of the proposed assay for 4. Bos JL. ras oncogenes in human cancer: a review. Cancer Res 1989;49:
4682–9.
screening, an important question is whether it enables 5. Minamoto T, Mai M, Ronai Z. K-ras mutation: early detection in molecular
identification of individuals who currently bear colorectal diagnosis and risk assessment of colorectal, pancreas, and lung cancers: a
or pancreatic neoplasms or who are at increased risk of review. Cancer Detect Prev 2000;24:1–12.
6. Haug U, Brenner H. New stool tests for colorectal cancer screening: a
developing neoplasms at these organs in the future. The systematic review focusing on performance characteristics and practical-
lack of associations between important risk factors for the ness. Int J Cancer 2005;117:169 –76.
respective cancer sites and mutant K-ras in stool provides 7. Caldas C, Hahn SA, Hruban RH, Redston MS, Yeo CJ, Kern SE. Detection of
K-ras mutations in the stool of patients with pancreatic adenocarcinoma and
more rebutting than supporting evidence regarding the pancreatic ductal hyperplasia. Cancer Res 1994;54:3568 –73.
value of the proposed assay to identify high-risk individ- 8. Berndt C, Haubold K, Wenger F, Brux B, Muller J, Bendzko P, et al. K-ras
uals. In regard to colorectal disease, none of the individ- mutations in stools and tissue samples from patients with malignant and
nonmalignant pancreatic diseases. Clin Chem 1998;44:2103–7.
uals diagnosed with advanced adenomas or CRC within 2 9. Wenger FA, Zieren J, Peter FJ, Jacobi CA, Muller JM. K-ras mutations in
years after stool collection tested positive. In the most tissue and stool samples from patients with pancreatic cancer and chronic
important study for comparison, 16% of CRC cases (95% pancreatitis. Langenbecks Arch Surg 1999;384:181– 6.
10. Lu X, Xu T, Qian J, Wen X, Wu D. Detecting K-ras and p53 gene mutation from
confidence interval, 5%–34%) and ⬃4% of advanced ade- stool and pancreatic juice for diagnosis of early pancreatic cancer. Chin Med
noma cases (95% confidence interval, 3%–7%) tested pos- J 2002;115:1632– 6.
itive for mutant K-ras in stool at the time of diagnosis (15 ). 11. Lohr M, Kloppel G, Maisonneuve P, Lowenfels AB, Luttges J. Frequency of
K-ras mutations in pancreatic intraductal neoplasias associated with pan-
According to these figures, 2 positive findings would creatic ductal adenocarcinoma and chronic pancreatitis: a meta-analysis.
have been expected among the 31 cases with advanced Neoplasia 2005;7:17–23.
adenomas/CRC in our study. Although the analytical 12. Berger DH, Chang H, Wood M, Huang L, Heath CW, Lehman T, et al.
Mutational activation of K-ras in nonneoplastic exocrine pancreatic lesions
sensitivity of the assay used in our study (i.e., the proba- in relation to cigarette smoking status. Cancer 1999;85:326 –32.
bility of detecting mutant K-ras in stool if present) could 13. Rothenbacher D, Löw M, Hardt PD, Klor HU, Ziegler H, Brenner H. Prevalence
be improved through stool quantity and storage condi- and determinants of exocrine pancreatic insufficiency among older adults:
results of a population-based study. Scand J Gastroenterol 2005;40:697–
tions, these measures are expected to go along with a 704.
decrease in diagnostic specificity (i.e., the probability that 14. Berndt C, Wolf G, Schroder G, Bebenroth M, Oehlschlegel K, Hillebrand T, et
al. A microplate assay for K-ras genotyping. Eur J Clin Chem Clin Biochem
healthy people test negative). Given that for CRC, K-ras 1996;34:837– 40.
has been proposed for parallel testing in combination 15. Imperiale TF, Ransohoff DF, Itzkowitz SH, Turnbull BA, Ross ME. Fecal DNA
with further markers, the validity of such a multiple versus fecal occult blood for colorectal-cancer screening in an average-risk
population. N Engl J Med 2004;351:2704 –14.
marker panel is highly vulnerable to decreased specificity
16. Kopreski MS, Benko FA, Borys DJ, Khan A, McGarrity TJ, Gocke CD. Somatic
of any component. mutation screening: identification of individuals harboring K-ras mutations
Mutant K-ras in stool may also play a role for the with the use of plasma DNA. J Natl Cancer Inst 2000;92:918 –23.
detection of pancreatic cancer. Sequence testing in com- 17. Arvanitakis M, Van Laethem JL, Parma J, De Maertelaer V, Delhaye M,
Deviere J. Predictive factors for pancreatic cancer in patients with chronic
bination with additional markers has been suggested to pancreatitis in association with K-ras gene mutation. Endoscopy 2004;36:
distinguish patients with pancreatic cancer from patients 535– 42.
with benign pancreatic disease (8, 17 ). In this context, the
high detection rate of the proposed assay may be a minor Previously published online at DOI: 10.1373/clinchem.2006.078188
problem or even be advantageous provided that the right
specific combination of markers is found. Whether the
combination of fecal pancreas elastase 1 as a marker of
exocrine pancreatic insufficiency and mutant K-ras, which
are both independent predictive factors for pancreatic A Homogeneous Assay for Analysis of FMR1 Promoter
cancer development in patients with chronic pancreatitis Methylation in Patients with Fragile X Syndrome, Chris-
(17 ), may be useful remains to be clarified by further tina Dahl,1 Karen Grønskov,2 Lars A. Larsen,3 Per Guldberg,1
studies. and Karen Brøndum-Nielsen2* (1 Institute of Cancer Biology,
Danish Cancer Society, Copenhagen, Denmark; 2 The
In conclusion, our study results do not support the use of Kennedy Institute-National Eye Clinic, Glostrup, Den-
the proposed assay for CRC screening. Its potential for mark; and 3 Wilhelm Johannsen Centre for Functional
early detection of pancreatic cancer (in combination with Genome Research, Department of Medical Biochemistry
other markers) requires further investigation. and Genetics, University of Copenhagen, Denmark; * ad-
dress correspondence to this author at: The Kennedy
Institute-National Eye Clinic, Gl. Landevej 7, DK-2600
We greatly appreciate the laboratory work done by Glostrup, Denmark; fax 45-43-43-11-30, e-mail kbn@
Wiebke Hiesener. kisoe.org)
Background: Fragile X syndrome is caused by the ex- The FMR1 allele with a CGG repeat number of ⬎200 is
pansion of a CGG trinucleotide repeat at the 5ⴕ untrans- known as a full mutation and is associated with hyper-
lated region of the fragile X mental retardation 1 gene methylation of the repeat region and the adjacent pro-
(FMR1). When expanded to >200 repeats (full muta- moter CpG island. It is generally recognized that pro-
tion), the repeat region and the adjacent promoter CpG moter hypermethylation causes transcriptional silencing
island become hypermethylated, rendering FMR1 tran- of FMR1 and that the pathophysiological background for
scriptionally inactive. Conventional molecular diagno- disease manifestations is the absence or deficit of the
sis of fragile X syndrome involves determination of the fragile X mental retardation protein (FMRP) (1 ). The
theory of methylation-coupled silencing is supported by
CGG repeat number by Southern blot analysis.
rare reports of healthy individuals with FMRP expression
Methods: A homogeneous methylation-specific melting
from full-mutation FMR1 alleles without promoter hyper-
curve analysis (MS-MCA) assay for methylation status
methylation (2 ).
of the FMR1 promoter region was developed on the Because the lack of definitive clinical diagnostic criteria,
LightCycler platform. Genomic DNA was treated with molecular tests are important for detection of individuals
sodium bisulfite, and a region containing 8 CpG sites with fragile X syndrome. The gold standard is Southern
was amplified in the presence of SYBR Green I, using blot analysis, which requires large amounts of DNA and
primers that do not differentiate between methylated is labor-intensive and time-consuming. Numerous PCR-
and unmethylated FMR1 molecules. After amplifica- based tests have been developed for screening proce-
tion, the samples were melted at 0.05 °C/s, and fluo- dures, which estimate the number of CGG repeats (3, 4 ).
rescence melting curves were recorded. We studied The major disadvantage of these methods is that their
samples, previously characterized by Southern blot specificity for detecting full-mutation and large premuta-
analyses, from 10 female and 10 male donors with tion alleles is only moderate, owing to the common failure
normal numbers of CGG trinucleotide repeats, 9 male of conventional PCR to amplify long repetitive regions
donors who were premutation carriers, 4 male donors with a very high G⫹C content. The Clinical Molecular
who carried both a premutation and a full mutation, and Genetics Society (CMGS) best practice guidelines recom-
25 patients with fragile X syndrome. mend prescreening by PCR amplification of the CGG
Results: Samples from all 20 male patients with fragile repeat and subsequent Southern blot analysis of samples
X syndrome showed a high melting peak corresponding that fail to amplify (males) or show a single allele (fe-
males). A disadvantage of this strategy is the inability to
to fully methylated FMR1, whereas samples from
reliably distinguish between healthy males and males
healthy males showed a single low melting peak corre-
who are mosaic for normal and full-mutation alleles.
sponding to unmethylated FMR1. Of 24 samples from
Alternative PCR-based tests to determine FMR1 methyl-
affected males, 9 (38%) showed 2 melting peaks, sug- ation status and/or repeat length have been developed
gesting that cellular methylation mosaicism is common (5– 8 ); these rely on treatment of DNA template with bisul-
in fragile X syndrome. fite to convert unmethylated cytosine, but not methylated
Conclusions: MS-MCA allows rapid and reliable iden- cytosine, into uracil before PCR amplification (9 ). However,
tification of fragile X syndrome in male patients. these methods all require a 2-step procedure, comprising
© 2007 American Association for Clinical Chemistry initial PCR amplification followed by product analysis,
which implies the risk of carry-over contamination.
Fragile X syndrome is the most common inherited cause We have developed a homogeneous assay for fragile X
of mental retardation, with a frequency of 1 in 4000 males syndrome on the LightCycler platform, based on detec-
and 1 in 6000 females (1 ). The mental impairment in tion of hypermethylated FMR1 alleles by methylation-
affected individuals ranges from learning disabilities to specific melting curve analysis (MS-MCA). This method
autism and severe mental retardation and may be accom- resolves differentially methylated DNA sequences on the
panied by a variety of physical and behavioral character- basis of differences in melting temperature after treatment
istics, making the clinical diagnosis difficult. of DNA with bisulfite (10 ). The basic steps, which can
At the genomic level, fragile X syndrome is associated be integrated by means of a thermal cycler coupled with
with the expansion of a naturally occurring CGG trinu- a fluorometer, involve nondiscriminatory amplification of
cleotide tandem repeat in the promoter region and 5⬘ methylated and unmethylated alleles in the presence of
untranslated region of the fragile X mental retardation 1 SYBR Green I, followed by step-wise increase of the
gene (FMR1) at Xq27.3 (1 ). The number of CGG repeats is temperature under continuous fluorescence monitoring to
highly variable, and 4 allelic forms of FMR1 have been examine the melting properties of the PCR products. The
defined: normal alleles have ⬍50 CGG repeats and are derived melting peaks provide a graphic profile of the
stable upon transmission from generation to generation; entire pool of DNA molecules in the sample and can be
intermediate (or “gray-zone”) alleles have 50 –58 CGG used to differentiate among 4 different methylation states:
repeats and may show some instability; premutations (a) unmethylated alleles generate a single low melting
have between 59 and ⬃200 repeats and are liable to peak, (b) fully methylated alleles generate a single high
further expand upon transmission; and the alleles in melting peak, (c) a mixture of unmethylated and fully
individuals with fragile X syndrome have ⬎200 repeats. methylated alleles generate both the low and high melting
peaks, and (d) heterogeneously methylated alleles gener- single melting peak with a Tm ⬃4 °C higher (Fig. 1A), and
ate a broadened melting top located between the low and DNA from the healthy female showed both melting peaks
high melting peaks. The latter state indicates molecular because 1 allele was methylated with X-chromosome
mosaicism, in which the content and distribution of inactivation (Fig. 1B).
methylated cytosine residues differ among different al- We blindly tested DNA from 58 sample donors, includ-
leles in the same sample. ing 10 females and 10 males with normal numbers of CGG
Genomic DNA was treated with sodium bisulfite, as trinucleotide repeats, 9 male premutation carriers, 4 mo-
previously described (10 ). To establish an MS-MCA assay saic males carrying both a premutation and a full muta-
for detection of FMR1 hypermethylation, we designed a tion, and 20 male and 5 female patients with fragile X
primer set that spans 8 CpG sites within the FMR1 syndrome. These samples had previously been examined
promoter and tested it on bisulfite-treated DNA from 2 by Southern blot analysis. The samples were donated for
healthy persons, a male and a female, and in vitro- investigation with informed consent from the patients/
methylated DNA (CpGenome Universal Methylated parents. All samples from healthy females showed both
DNA, Chemicon). The primers were 5⬘-GTG AAG TGG low and high melting peaks, corresponding to the un-
TTT TAG TGT TTA TAT T-3⬘ (2541) and 5⬘-CTC AAA methylated and fully methylated FMR1 alleles, respec-
AAC TAC CCT CCA CC-3⬘ (2645R), which amplify a tively. Samples from healthy male donors showed a single
105-bp region of the FMR1 promoter 5⬘-CpG island (Gen- low melting peak, whereas samples from all 24 male
Bank Accession No. X61378). The underlined nucleotides patients with fragile X syndrome showed a high melting
indicate mismatches within the primer sequences repre- peak corresponding to fully methylated FMR1 (Fig. 1C).
senting CpG sites in the FMR1 promoter sequence. PCR The 4 samples from male donors who carried both full-
was carried out using the LightCycler 1.0 instrument mutation and premutation alleles showed both low and
(Roche) in 10-␮L reaction mixtures containing 5 pmol of high melting peaks (Fig. 1C). In addition, of the 20
each primer, 3 mmol/L MgCl2, 1⫻ FastStart DNA Master samples from fragile X male patients who carried a full
SYBR Green I (Roche), and bisulfite-modified DNA. Re- mutation, 5 samples showed 2 melting peaks but were not
actions were started by initial denaturation at 95 °C for 10 identified as mosaics by Southern blot analysis. The size
min, followed by 34 cycles at 95 °C for 5 s, 65 °C for 10 s, of the low melting peak in the mosaic samples from male
and 72 °C for 15 s. Melting curve analysis was performed donors showed some interrun variation, probably due to
immediately after amplification by measuring the fluores- stochastic amplification. In total, 9 of the 24 samples from
cence of SYBR Green I during a temperature transition affected males (38%) contained both unmethylated and
from 60 °C to 95 °C at 0.05 °C/s. Fluorescence data were fully methylated FMR1 alleles, suggesting that cellular
converted into melting peaks using the LightCycler Soft- FMR1 methylation mosaicism is common in fragile X
ware 4.05. DNA from the healthy male showed a single syndrome. In previous studies, mosaicism in affected
melting peak with an apparent melting temperature (Tm) males has been detected with highly variable frequencies,
of ⬃77 °C (Fig. 1A), in vitro-methylated DNA showed a ranging from 12% (11 ) to 41% (12 ), which probably
reflects differences in assay resolution and/or the tissues
analyzed (13 ).
In our study, all 5 samples from female patients with
fragile X syndrome showed both low and high melting
peaks and thus could not be distinguished from samples
from healthy female donors (Fig. 1D). This is in agreement
with other studies showing that blood lymphocytes from
female patients with fragile X syndrome hold both methyl-
ated and unmethylated FMR1 alleles (14, 15 ). We found that
samples from male premutation carriers had a single low
melting peak similar to that seen in samples from those of
healthy males in previous studies (Fig. 1D), consistent with
previous studies showing that premutation FMR1 alleles are
unmethylated (16, 17 ). None of the 58 samples showed
melting peaks with intermediate Tms, suggesting that mo-
lecular mosaicism at the FMR1 locus is rare and, accord-
ingly, the distribution of FMR1 promoter methylation is
Fig. 1. MS-MCA analysis of the FMR1 promoter region.
bimodal (18 ).
Bisulfite-treated DNA was amplified in the presence of SYBR Green I using
primers that do not differentiate between methylated and unmethylated FMR1 As with other methods relying on treatment of DNA
alleles. The melting characteristics of the PCR products were determined directly with bisulfite, incomplete conversion of cytosine to uracil
in the PCR tube by continuous fluorescence monitoring during a temperature
transition at 0.05 °C/s. (A), the melting peaks for in vitro-methylated DNA (M)
may be a problem and may cause false-positive results
and DNA from a healthy male (U) have Tms of ⬃81 °C and ⬃77 °C, respectively. (19 ). However, MS-MCA of the FMR1 promoter in DNA
Also shown are melting peaks for DNA from (B) a healthy female donor; (C) 2 from healthy males consistently showed a single low-
male patients with fragile X syndrome, including 1 whose DNA shows a mosaic
pattern (thick line); and (D) a female patient with fragile X syndrome (thick line) melting peak, suggesting that the protocol used in this
and a male donor whose DNA shows a promutation carrier pattern. study led to complete bisulfite conversion. Another po-
tential problem associated with MCA is that resolution of 4. O’Connell CD, Atha DH, Jakupciak JP, Amos JA, Richie K. Standardization of
PCR amplification for fragile X trinucleotide repeat measurements. Clin
Tm differences using the LightCycler may be reduced Genet 2002;61:13–20.
when more samples are melted at the same time (20 ). 5. Panagopoulos I, Lassen C, Kristoffersson U, Aman P. A methylation PCR
Nevertheless, the Tm difference between methylated and approach for detection of fragile X syndrome. Hum Mutat 1999;14:71–9.
unmethylated FMR1 sequences was so large (⬃4 °C) that 6. Zhou Y, Law HY, Boehm CD, Yoon CS, Cutting GR, Ng IS, et al. Robust fragile
X (CGG)n genotype classification using a methylation specific triple PCR
interpretation of the melting peaks was unproblematic assay. J Med Genet 2004;41:e45.
even when a full rotor was analyzed. 7. Zhou Y, Lum JM, Yeo GH, Kiing J, Tay SK, Chong SS. Simplified molecular
In summary, closed-tube resolution of methylated and diagnosis of fragile X syndrome by fluorescent methylation-specific PCR and
GeneScan analysis. Clin Chem 2006;52:1492–500.
unmethylated FMR1 alleles on the basis of differences in 8. Boyd VL, Moody KI, Karger AE, Livak KJ, Zon G, Burns JW. Methylation-
Tm after bisulfite conversion was 100% effective in iden- dependent fragment separation: direct detection of DNA methylation by
tifying fragile X syndrome in male patients. Because the capillary electrophoresis of PCR products from bisulfite-converted genomic
DNA. Anal Biochem 2006;354:266 –73.
assay reads both methylated and unmethylated alleles 9. Clark SJ, Harrison J, Paul CL, Frommer M. High sensitivity mapping of
simultaneously, mosaicism was also readily identified in methylated cytosines. Nucleic Acids Res 1994;22:2990 –7.
DNA from male donors. On the other hand, the assay did 10. Worm J, Aggerholm A, Guldberg P. In-tube DNA methylation profiling by
fluorescence melting curve analysis. Clin Chem 2001;47:1183–9.
not enable us to detect fragile X syndrome in DNA from 11. Rousseau F, Heitz D, Biancalana V, Blumenfeld S, Kretz C, Boue J, et al.
female donors or permutation carrier status in the male Direct diagnosis by DNA analysis of the fragile X syndrome of mental
donors because the methylation patterns were similar to retardation. N Engl J Med 1991;325:1673– 81.
12. Nolin SL, Glicksman A, Houck GE Jr, Brown WT, Dobkin CS. Mosaicism in
those of healthy individuals. In addition, for some male fragile X affected males. Am J Med Genet 1994;51:509 –12.
patients who had fragile X with mosaicism, DNA patterns 13. MacKenzie JJ, Sumargo I, Taylor SA. A cryptic full mutation in a male with a
were indistinguishable from those of healthy females. Nev- classical fragile X phenotype. Clin Genet 2006;70:39 – 42.
ertheless, considering that the main indication to test for 14. Heine-Suner D, Torres-Juan L, Morla M, Busquets X, Barcelo F, Pico G, et al.
Fragile-X syndrome and skewed X-chromosome inactivation within a family:
FMR1 abnormalities is mental retardation in male patients, a female member with complete inactivation of the functional X chromo-
the MS-MCA assay could find widespread use as an initial some. Am J Med Genet A 2003;122:108 –14.
screening procedure due to its low cost, simplicity, and 15. Martinez R, Bonilla-Henao V, Jimenez A, Lucas M, Vega C, Ramos I, et al.
Skewed X inactivation of the normal allele in fully mutated female carriers
closed-tube format. determines the levels of FMRP in blood and the fragile X phenotype. Mol
Diagn 2005;9:157– 62.
16. Genc B, Muller-Hartmann H, Zeschnigk M, Deissler H, Schmitz B, Majewski
F, et al. Methylation mosaicism of 5⬘-(CGG)(n)-3⬘ repeats in fragile X,
We thank Vibeke Ahrenkiel for expert technical assis- premutation, and normal individuals. Nucleic Acids Res 2000;28:2141–52.
tance. This study was supported by the Danish Medical 17. Hornstra IK, Nelson DL, Warren ST, Yang TP. High resolution methylation
analysis of the FMR1 gene trinucleotide repeat region in fragile X syndrome.
Research Council and the Neye Foundation. Hum Mol Genet 1993;2:1659 – 65.
18. Stöger R, Kajimura TM, Brown WT, Laird CD. Epigenetic variation illustrated
by DNA methylation patterns of the fragile-X gene FMR1. Hum Mol Genet
1997;6:1791– 801.
References 19. Dahl C, Guldberg P. DNA methylation analysis techniques. Biogerontology
1. Warren ST, Sherman SL. The fragile X syndrome. In: Scriver CR, Beaudet AL, 2003;4:233–50.
Sly WS, Valle D, eds. The Metabolic and Molecular Bases of Inherited 20. Gundry CN, Vandersteen JG, Reed GH, Pryor RJ, Chen J, Wittwer CT.
Disease. New York: McGraw-Hill, 2001:1257– 89. Amplicon melting analysis with labeled primers: a closed-tube method for
2. Smeets HJ, Smits AP, Verheij CE, Theelen JP, Willemsen R, van de Burgt I, differentiating homozygotes and heterozygotes. Clin Chem 2003;49:396 –
et al. Normal phenotype in two brothers with a full FMR1 mutation. Hum Mol 406.
Genet 1995;4:2103– 8.
3. Larsen LA, Grønskov K, Nørgaard-Pedersen B, Brøndum-Nielsen K, Hasholt
L, Vuust J. High-throughput analysis of fragile X (CGG)n alleles in the normal
and premutation range by PCR amplification and automated capillary elec- Previously published online at DOI: 10.1373/clinchem.2006.080762
trophoresis. Hum Genet 1997;100:564 – 8.
Letters
Considerably Reduced Centrifugation stored at room temperature for at No differences in gel integrity or
Time without Increased Hemolysis: least 30 min to allow complete coag- wandering were noted in the tubes
Evaluation of the New BD Vacutainer® ulation. All centrifugation was per- tested. Data from phases 1 and 2 are
SST™II Advance formed in 1 centrifuge at 21 °C (Ro- presented together because there
tana 46 RSC Robot, rotor diameter were no marked differences in values
To the Editor: 180 mm, 90 °C Hettich Zentrifugen), from each phase (P ⬎0.05) except for
which is part of an automated pre- LD because of delayed centrifugation
Short in-laboratory turnaround time
analytical system (StreamLAB, Dade (P ⬍0.001). Values for analytes from
is important, and reducing centrifu-
Behring). To avoid frequent changes the paired tubes spun with the CC or
gation time can shorten this phase of
in centrifugation settings, phase 1 AC mode were not significantly dif-
the preanalytical process. The new
(n ⫽ 104) and phase 2 (n ⫽ 104) of the ferent (P ⬎0.05 for all) (Fig. 1). Pass-
BD Vacutainer® SST™II Advance
study were performed in a crossover ing and Bablock regression (4 )
(Becton Dickinson, article number
design. In phase 1, the 1st specimen showed AC ⫽ 1.00x ⫹ 0.00 for all
367955) has a semisolid, thyxotrophic
of each pair was centrifuged within (except LD where AC ⫽ 1.01x ⫺4.09).
acrylic gel with chemical composi-
60 min by the CC mode and the 2nd We concluded that for these new
tion identical to that of the previous
centrifuged in the AC mode within tubes the AC mode did not increase
SST™II tube. In the new Advance 120 min of arrival in the laboratory. the rate of hemolysis.
tube, however, the gel has been ex- In phase 2, the 2 centrifugation set- Based on these results, we
tended to 1 side of the tube, forming tings were reversed. All analyses switched our routine centrifugation
a gel nose with a larger initial contact were completed within 3 h. mode from CC (13 min) to AC (5
surface that is supposed to accelerate In all tubes, the mean (SD) thick- min) to decrease processing time by 8
gel movement during centrifugation ness of the gel layer was 10 (1) mm. min. To confirm that the AC mode
to allow shorter centrifugation time
(1 ). We compared gel barrier forma-
tion and hemolysis rates of the new
Vacutainer SST™II Advance under a
conventional centrifugation (CC)
mode (13 min, 1700g; acceleration
and braking time, 55 s) vs an accel-
erated centrifugation (AC) mode (5
min, 3000g; acceleration and braking
time, 39 s) to determine whether this
new Vacutainer system allows re-
duced centrifugation time while
maintaining sample quality.
In vitro hemolysis can be caused
by mechanical stress inflicted by
higher g-forces, which might artifi-
cially increase several serum analytes
and cause interference in some labo-
ratory assays (2 ). Therefore we mea-
sured lactate dehydrogenase (LD),
aspartate aminotransferase (ASAT),
and potassium in paired serum sam-
ples as indicators of hemolysis, using
a Dimension RxL Max (3 ). As a more
sensitive and specific marker of he-
molysis, free hemoglobin was mea-
sured immunologically on a Dade
BN II system (all reagents and ana-
lyzers from Dade Behring). We used
routine blood samples from inpa-
Fig. 1. Percentage difference of free hemoglobin (fHb; n ⫽ 206), potassium (K; n ⫽ 208), LD
tients and outpatients at the Greifs-
(n ⫽ 206), and ASAT (n ⫽ 207) in sample pairs, which were drawn in parallel from patients and
wald University Hospital who had
centrifuged sequentially.
laboratory orders requiring 2 serum The solid lines represent the mean difference between the CC and AC modes, the dashed lines represent
tubes. The university ethical board mean SD (1.96). Up to 2 data points with very high values were removed from the figures so that number of
did not require approval by the hu- data points differs from 208. In phase 1 (open circles), CC (13 min 1700g;, acceleration and braking time,
55 s) was performed within 60 min of arrival in the laboratory, whereas AC (5 min, 3000g; acceleration and
man study committee for this study. braking time, 39 s) was performed within 120 min. In phase 2 (closed circles), AC was performed within 60
Before centrifugation, samples were min of arrival in the laboratory, and CC was performed within 120 min.
794 Clinical Chemistry 53, No. 4, 2007

did not increase hemolysis, we exam- Birger Mensel serum samples produced phospho-
ined hemolysis (H)-indices before Ulrike Wenzel, rus values within or near the refer-
and after implementing the AC Markus Roser, ence interval when tested on other
mode on a large number of samples. Jan Lüdemann, chemistry instruments including the
H-indices were determined photo- Matthias Nauck* Vitros 950 (Johnson and Johnson)
metrically at 405 and 700 nm (5 ). and Advia (Bayer) analyzers (Table
Retrospective examination of H-indi- Institute of Clinical Chemistry 1). Lipemic indices on the Synchron
ces from 42 124 samples processed in and Laboratory Medicine LX20 for samples from patient A
the CC mode was compared with Ernst Moritz Arndt University were within the reference interval of
H-indices from 30 483 samples pro- Greifswald, Germany 0 –3 but were increased (7– 8) for
cessed in the AC mode. The distribu- samples from patient B. Serum tri-
tions of the different grades of hemo- glyceride concentrations for patients
lysis were identical for the 2 groups *Address correspondence to this au- A and B were 1.0 and 6.1 mmol/L,
(see Table 1 in the Data Supplement thor at: Matthias Nauck, University Hos- respectively. Treatment of samples
pital Greifswald, Institute of Clinical
that accompanies the online version Chemistry and Laboratory Medicine, Fer- from both patients with Lipoclear
of this Letter to the Editor at http:// dinand Sauerbruchstrasse, D-17487 Grei- (Iris Sample Processing) to reduce
www.clinchem. fswald, Germany. Fax 49-3834-86-5502; turbidity was associated with a
org/content/vol53/issue4), with no e-mail matthias.nauck@uni-greif- marked decrease in phosphorus re-
swald.de.
increase in the frequency of hemo- sults on the Synchron LX20 analyzer
lyzed samples during operation in DOI: 10.1373/clinchem.2006.079582 (Table 1). According to the product
the AC mode. We also investigated insert, neither Intralipid nor endoge-
the frequency of clot occurrences nous hyperlipidemia (lipemic index
before and after the introduction value of 8) interfere with the Beck-
of the AC mode by examining the man PHOSm assay (1 ). These obser-
frequency of the error code “process- Liposomal Amphotericin B Interferes vations suggest that alterations in
ing error” on the Dimension RxL with the Phosphorus Assay on the sample turbidity related to liposomal
Max, as a surrogate indicator for clot Synchron LX 20 Analyzer amphotericin B rather than increased
detection. The clot detection rate triglyceride values may have contrib-
dropped from 0.09% to 0.05% (P To the Editor: uted to the false increases of phos-
⬍0.0001). Thus, the AC mode did not We report an interference caused by phorus values.
result in increased clot formation on liposomal amphotericin B (AmBisome, A previous study demonstrated
the analyzers. Astellas Pharma) in the Synchron LX that falsely increased phosphorus
In conclusion, the BD Vacutainer 20 (Beckman Coulter) phosphorus as- values may occur in patients with
SST™II Advance allows fast and reli- say (PHOSm). Two inpatients treated plasma cell dyscrasias and lym-
able serum separation within 5 min with liposomal amphotericin B for phoreticular malignancies associated
at 3000g without change in sample invasive mucormycosis were noted with abnormal immunoglobulin syn-
quality regarding gel barrier forma- to have significantly increased serum thesis (2 ). Immunofixation studies of
tion, separation, and hemolysis. The phosphorus concentrations. Serum patient B revealed an IgG ␬ parapro-
AC settings can be recommended for phosphorus values from patient A tein; however, patient A was nega-
routine use with the SST™II Advance ranged from 0.9 mmol/L to 3.1 tive for a paraprotein. Given the ab-
tubes to shorten turnaround times. mmol/L compared with pretreat- sence of paraprotein in patient A
ment values of 0.7 to 0.9 mmol/L along with the appearance of hyper-
We thank Becton Dickinson for pro- (reference interval, 0.8 –1.5 mmol/L). phosphatemia beginning at the start
viding sample tubes free of charge. Patient B was transferred to our in- of AmBisome therapy, paraprotein
stitution after 23 days of amphoteri- interference is a less likely explana-
References cin treatment and had phosphorus tion for the hyperphosphatemia in
1. Product information. Comparison of BD Vacu-
tainer® SST™ II Advance Tubes with BD Vacu- values ranging from 1.4 to 3.9 these patients. The manufacturer has
tainer® SST™ plus tubes for selected clinical mmol/L. Both patients had slightly also reported positive interference
chemistry analytes. 2004. with the Synchron LX20 phosphorus
increased but stable serum creatinine
2. Yucel D, Dalva K. Effect of in vitro hemolysis on
25 common biochemical tests. Clin Chem concentrations (53.0 –185.6 ␮mol/L) assay for hemolyzed samples and
1992;38:575–7. and did not exhibit symptoms of from nafcillin and rifampin (1 ), fac-
3. Frank JJ, Bermes EW, Bickel MJ, Watkins BF.
Effect of in vitro hemolysis on chemical values
hyperphosphatemia. One attending tors that were not present in either of
for serum. Clin Chem 1978;24:1966 –70. physician suspected a possible inter- the cases we describe. It is theoreti-
4. Passing H, Bablok W. Comparison of several ference in the phosphorus assay and cally possible that the LX 20 method
regression procedures for method comparison
studies and determination of sample size. Part
contacted the clinical laboratory to was reading correctly but the other
II. J Clin Chem Clin Biochem 1984;22:431– 45. initiate investigation. methods had low recovery. Given
5. Vermeer HJ, Thomassen E, de Jonge N. Auto- In contrast to the markedly in- the lack of clinical signs or symptoms
mated processing of serum indices used for
interference detection by the laboratory infor- creased phosphorus values from the of hyperphosphatemia in these pa-
mation system. Clin Chem 2005;51:244 –7. Synchron LX20 analyzer, the same tients, however, the most likely ex-
796 Letters
from the placenta (3 ). Although the

Table 1. Serum phosphorus values obtained on selected chemistry analyzers.
cells in amniotic fluid are widely
Synchron LX 20ⴙ
Synchron LX 20, Lipoclear, Advia, Vitros, known to have originated from fetal
mmol/L mmol/L mmol/L mmol/L urogenital, respiratory, and alimen-
Patient A 3.1 1.6 1.7 1.5 tary tracts, skin and amnion, the or-
Patient B 3.4 0.9 1.0 0.8 igin of cell-free fetal DNA in amni-
otic fluid remains unclear (2 ). It has
been speculated that the latter might
originate from the same sources as
planation for the results is interfer- rently available include liposomal amniotic fluid cells or the placenta
ence in the LX 20 phosphorus assay. doxorubicin (Doxil®, Ortho Biotech (2 ). Here, we applied the recent de-
The occurrence of severe hyper- Products) and liposomal morphine velopments in epigenetics to this
phosphatemia during treatment with sulfate (Depodur®, Endo Pharma- area of research (4, 5 ).
high-dose liposomal amphotericin ceuticals). Liposomal formulations of The promoter of RASSF1A has
was reported in a pediatric patient other drugs are currently in clinical been shown to be hypermethylated
(3 ), but the possibility of pseudohy- trials. in the placenta but essentially un-
perphosphatemia caused by interfer- methylated in maternal blood cells
ence of liposomal amphotericin B in and other fetal tissues (4 ). Chan
References
the phosphorus assay was not ad- et al. (5 ) adopted hypermethylated
1. PHOSm package insert. Beckman Coulter, Inc.
dressed. In this pediatric case, the 2005. RASSF1A as a placental-specific epi-
phosphorus assay was performed on 2. Larner AJ. Pseudohyperphosphatemia. Clin Bio- genetic marker and developed an
chem 1995;28:391–3. assay for its detection in maternal
a Synchron LX20 analyzer (Chris
3. Jain A, Butani L. Severe hyperphosphatemia plasma as a sex-independent fetal
Jarvinen, personal communication, resulting from high-dose liposomal amphoteri-
January 20, 2006). It may be worth- cin in a child with leukemia. J Pediatr Hematol DNA marker. The assay involved
while to investigate the potential for Oncol 2003;25:324 – 6. digestion by BstUI, a methylation-
AmBisome interference on other an- sensitive restriction enzyme, to
alyzers that use a phosphomolybdate Helen L. Bailey1 eliminate the background maternal
ultraviolet assay from other manu- Emily M. Chan2 plasma unmethylated RASSF1A DNA
facturers. The Johnson and Johnson 1
contributed by the maternal blood
Vitros analyzer and the Bayer Advia Department of Laboratory Medicine cells. Hypermethylated RASSF1A se-
analyzer also use a phosphomolyb- University of California San Francisco quences in maternal plasma, previ-
date method but use different reac- San Francisco, CA 94110 ously confirmed to carry the fetal ge-
tion timing and wavelength mea- 2
notype (5 ), would be resistant to
surements than the Synchron LX20, Clinical Laboratory digestion and therefore quantifiable
which might minimize susceptibility University of California San Francisco by real-time PCR. Another assay tar-
to AmBisome interference. Medical Center geting unmethylated ␤-actin (ACTB)
AmBisome consists of amphoteri- San Francisco, CA 94110 was included to control for the com-
cin B intercalated into a unilamellar pleteness of enzyme digestion. We
bilayer vesicle consisting of phos- * Address correspondence to this au-
adopted this assay system to deter-
pholipids and cholesterol and mea- thor at: The Department of Laboratory mine whether cell-free DNA in amni-
suring ⬍100 nm in diameter. We Medicine, University of California San otic fluid is derived from the placenta.
postulate that biodegradation of the Francisco, 505 Parnasus Ave., M-580, San Fourteen pregnant women (17–21
Francisco, CA 94143-0102; fax 415-206- gestational weeks) clinically indi-
liposomal vehicle may lead to inter- 6996; e-mail Helen.bailey@ucsfmedctr.org.
ference due to light scatter or precip- cated for amniocentesis were re-
itation that affects the absorbance DOI: 10.1373/clinchem.2006.078840 cruited from the Prince of Wales
measurements. Hospital with informed consent and
In patients receiving liposomal institutional ethical approval. Mater-
amphotericin B, interference in the nal blood was collected into EDTA-
Synchron LX20 phosphorus assay containing tubes before amniocente-
may lead to unnecessary treatment Epigenetic Analysis of RASSF1A Gene sis, and plasma was harvested by
with phosphorus-binding agents. in Cell-Free DNA in Amniotic Fluid centrifugation at 1600g for 10 min at
These cases illustrate the importance 4 °C, with the plasma portion recen-
of evaluating the potential role of To the Editor: trifuged at 16 000g for 10 min at 4 °C.
assay interferences in unexplained Cell-free fetal nucleic acids exist not Two milliliters of amniotic fluid were
abnormal laboratory results. The cur- only in the maternal circulation (1 ) drawn and centrifuged at 16 000g for
rent observations should also moti- but also in amniotic fluid samples 10 min at 4 °C. The supernatant was
vate testing of other liposomal-based (2 ). Evidence suggests that DNA in removed and filtered by a 0.22-␮m
pharmaceutical agents for similar in- maternal plasma that bears the fetal filter to obtain the cell-free portion.
terference. Liposomal agents cur- genotype is predominantly derived The cell pellet was washed once with
natures in determining the origin of

Table 1. Quantification of RASSF1A sequences after BstUI digestion in amniotic
cell-free fetal DNA in amniotic fluid.
fluid and maternal plasma DNA samples of 2nd trimester pregnancies.
Cell-free DNA in amniotic fluid is
Fractional concentration of RASSF1A after BstUI digestion, %
mainly derived from the fetus
Case Amniotic fluid Amniotic fluid Maternal proper. Conversely, amniotic fluid is
no. cell-free fraction cellular fraction plasma unlikely to be a significant contribu-
1 0.00 0.00 5.04 tor of fetal DNA in maternal plasma.
2 1.30 0.00 3.35
3 0.00 1.09 6.33
4 0.59 0.17 11.53
5 0.00 0.10 2.36 Grant/funding support: This work
6 0.82 0.28 10.55 was supported by a Central Alloca-
7 0.07 0.16 0.68 tion Grant (CUHK 01/03C) from the
8 0.00 0.37 12.33 Research Grants Council of the Hong
9 0.00 0.37 4.03 Kong Special Administrative Region
10 0.31 0.00 2.34 (China). Y.M.D.L. is supported by
11 1.57 0.26 8.72 the Chair Professorship Scheme of
12 0.00 0.00 7.16 the Li Ka Shing Foundation.
13 0.56 0.13 8.31 Financial disclosures: F.M.F.L.,
14 0.36 0.02 1.48 R.W.K.C., and Y.M.D.L. have filed
patent applications on aspects of fe-
tal nucleic acid analysis from mater-
nal plasma.
phosphate buffered saline and resus- maternal plasma were comparable to
pended in 200 ␮L of PBS [7.4 (1x) the previous data (5 ). Complete BstUI
liquid (Gibco), consisting of 1.06 digestion was confirmed in all samples References
mmol/L monobasic potassium phos- by the absence of postdigestion ACTB 1. Lo YMD, Corbetta N, Chamberlain PF, Rai V,
phate, 155.17 mmol/L sodium chlo- DNA (data not shown). Sargent IL, Redman CW, et al. Presence of fetal
DNA in maternal plasma and serum. Lancet
ride, and 2.97 mmol/L dibasic so- Our previous data showed that 1997;350:485–7.
dium phosphate at pH 7.4]. For each RASSF1A was hypermethylated only 2. Bianchi DW, LeShane ES, Cowan JM. Large
case, DNA was extracted from 1.6 in the placenta and not other fetal amounts of cell-free fetal DNA are present in
mL of maternal plasma, 400 ␮L of tissues (4 ). The proportion of diges- amniotic fluid. Clin Chem 2001;47:1867–9.
cell-free amniotic fluid, and 200 ␮L tion-resistant RASSF1A in maternal 3. Bianchi DW. Circulating fetal DNA: its origin and
diagnostic potential-a review. Placenta 2004;
of the resuspended cell pellet with plasma was similar to the fractional 25(Suppl A):S93–S101.
the QIAamp DNA Blood Mini Kit concentrations of circulating fetal 4. Chiu RWK, Chim SSC, Wong IHN, Wong CSC,
(Qiagen). All of the DNA samples DNA previously determined using Lee WS, To KF, et al. Hypermethylation of
RASSF1A in human and rhesus placentas. Am J
were divided into 2 aliquots and Y-chromosomal markers (1 ). These Pathol;in press.
subjected to either no treatment or findings suggest that fetal DNA in 5. Chan KCA, Ding C, Gerovassili A, Yeung SW,
BstUI digestion followed by real- maternal plasma is predominantly Chiu RWK, Leung TN, et al. Hypermethylated
time PCR (5 ). derived from the placenta. On the RASSF1A in maternal plasma: a universal fetal
DNA marker that improves the reliability of
In all cases RASSF1A was detected other hand, cell-free fetal DNA con- noninvasive prenatal diagnosis. Clin Chem
with and without BstUI digestion in centration in amniotic fluid was re- 2006;52:2211– 8.
the maternal plasma and cell-free ported to be ⬎100-fold higher than
and cell-pellet portions of the amni- that of maternal plasma (2 ). Yet, Fiona M.F. Lun1,2
otic fluid. The median concentrations here we show that the proportion Rossa W.K. Chiu1,2
after BstUI digestion were 120, 54, of digestion-resistant RASSF1A was Tak Y. Leung3
and 124 copies/mL in the 3 corre- some 30-fold lower in the cell-free Tse N. Leung3
sponding sample types, respectively. portion of amniotic fluid than ma- Tze K. Lau3
The results were further expressed as ternal plasma. These observations Y.M. Dennis Lo1,2*
a fraction of the total RASSF1A DNA suggest that the placenta is unlikely
concentration obtained without BstUI to be the main source of amniotic Departments of
1
digestion, and corresponded to 5.69% fluid cell-free DNA. The majority of Chemical Pathology
(interquartile range: 2.36%– 8.72%), RASSF1A sequences in amniotic and 3 Obstetrics and Gynaecology
2
0.19% (interquartile range: 0%– 0.59%), fluid, either from the cell-free or cel- Centre for Research into
and 0.15% (interquartile range: 0%– lular portions, were digestible by Circulating Fetal Nucleic Acids
0.28%) in the maternal plasma and the BstUI and thus were unmethylated, Li Ka Shing
cell-free and cellular portions of amni- and were probably derived from the Institute of Health Sciences
otic fluid, respectively (Table 1). The nonplacental part of the fetus. The Chinese University
absolute and fractional concentrations Our study has demonstrated the of Hong Kong, Shatin,
of digestion-resistant RASSF1A in use of tissue-specific epigenetic sig- New Territories, Hong Kong SAR
798 Letters
* Address correspondence to this

Of the 43 serum samples, 2 dem- measurements (y axis) against their
author at: Department of Chemical Pa- onstrated a hook effect. One CT se- mean (x axis) allows detection of lack
thology, Room 38023, 1/F, Clinical Sci- rum sample analyzed by the poly- of individual agreement. Near agree-
ences Building, Prince of Wales Hospital, clonal assay was 699 ng/L undiluted ment appeared for CT concentrations
30 –32 Ngan Shing Street, Shatin, Hong and 6564 ng/L after 1:100 dilution; ⬍40 ng/L. For higher values, the
Kong SAR. Fax ⫹852-2194 -6171; e-mail
loym@cuhk.edu.hk. the same sample, measured by the monoclonal assay tended to produce
monoclonal assay, was ⬎1500 ng/L lower results than the polyclonal as-
DOI: 10.1373/clinchem.2006.084350 undiluted and 2968 after 1:20 dilu- say. In addition, the higher the mean
tion. A 2nd sample analyzed with value, the greater the difference be-
both assays gave polyclonal assay tween the measurements (Fig. 1).
values of 510 ng/L undiluted and Analysis by Passing-Bablok regres-
Comparison of Calcitonin 30236 ng/L after 1:100 dilution, and sion showed a significant deviation
Determinations by Polyclonal and monoclonal assay values of 634 ng/L from linearity (P ⬍0.01). A possible
Monoclonal IRMAs undiluted and 3225 ng/L after 1:100 explanation is that at higher concen-
dilution. The hook effect seems less trations of mature monomeric CT,
To the Editor: likely to occur in a 1-step method polymeric forms and other immuno-
Serum calcitonin (CT) has important when 2 monoclonal antibodies are reactive fragments increase in the
diagnostic and prognostic value in used, probably because of mono- circulation. In these circumstances,
patients with medullary thyroid car- specificity of the signal antibody (2 ). polyclonal antibodies might yield
cinoma (MTC). Although CT can be The 2 samples with the hook effect higher values than those measured
routinely quantified by a variety of were excluded from the subsequent by the monoclonal assay.
2-site immunoassays, different as- analysis because they were deemed The lack of a linear relationship
says may yield very different results. to be outliers in a nonparametric between the 2 assays, particularly at
This variability, particularly at low distribution (Statistica for Windows higher concentrations, precludes in-
CT concentrations, led to the recom- 98, Statsoft Inc.). terchangeability of results between
mendation to use reliable and sensi- The remaining 41 serum samples the 2 methods. Furthermore, a strik-
tive assays in the screening of MTC exhibited CT concentrations of 2.87– ingly discordant value (Fig. 1, circled
and C-cell hyperplasia (1 ). 344 ng/L (median, 46.8 ng/L) by the sample) was obtained in a patient
To highlight CT variability at high polyclonal assay and 1.72–148 ng/L with CT concentrations of 157 and
concentrations, we measured 43 se- (median, 31.6 ng/L) by the monoclo- 1.5 ng/L by the polyclonal and the
rum samples by use of a polyclonal nal assay. To compare the 2 methods monoclonal assays, respectively.
IRMA and by a monoclonal IRMA. we used the Bland-Altman represen- The patient was later diagnosed
The single-step polyclonal assay tation (3 ), which focuses on the mean with cholangioadenocarcinoma. Some
used 2 different goat antibodies (as and variability of differences be- monoclonal immunometric assays
the capture and labeled antibodies), tween pairs of measurements: a scat- cross-react with high concentrations
which were specific for well-defined ter plot of the difference between the of procalcitonin (4 ). The sample was
regions of the CT molecule, with no
cross-reactivity with parathyroid
hormone, thyroid-stimulating hor-
mone, CT gene–related peptide, por-
cine CT, or salmon CT. The reagent
set was used according to the manu-
facturer’s instructions (Calcitonin
Assay, Scantibodies Laboratory,
Inc.). The concentrations of calibra-
tors ranged from 10 to 1000 ng/L.
Expected values were ⱕ9.9 ng/L for
women and ⱕ17.0 ng/L for men. The
single-step monoclonal assay uses 2
antibodies with no cross-reactivity
with procalcitonin. The reagent set
was used according to manufactur-
er’s instructions (CIS Biointerna-
tional). The concentrations of calibra-
tors ranged from 10 to 1500 ng/L,
with a reference interval of ⬍10 Fig. 1. Variability of differences (CT measured by the polyclonal assay minus CT measured by the
ng/L. Both assays were calibrated monoclonal assay) between pairs of measurements using the Bland and Altman analysis in 41
against the 2nd Calcitonin Interna- serum samples with normal and high CT.
tional Standard 89/620. Mean serum CT difference, 46.1 ng/L.
analyzed for procalcitonin (Liaison sum of unconjugated (Bu) and conju- ter thawing, the hemolysate was cen-
Brahms PCT), yielding a result of gated (Bc) bilirubins] methods on the trifuged to remove the stroma. The
0.88 ␮g/L, slightly higher than the Vitros 5,1 FS analyzer and assessed Hb concentration was measured
cutoff value (0.58 ␮g/L). The cross- the reliability of the analyzer’s hemo- with the CELL-DYN 4000 analyzer.
reactivity with procalcitonin might lysis and icterus index values by We prepared a stock solution of Bu
be more likely in our sample mea- comparing them to measured con- (NIST SRM 916) in pooled human
sured with a polyclonal assay. Other centrations of bilirubin and Hb in serum samples from healthy volun-
interferences cannot be excluded. test specimens. We also compared teers as described previously (1 ); dilu-
The 2 CT assays give markedly TBIL with NBIL values in specimens tions were made with the same pooled
different results. The introduction of from neonates 1–14 days old and sera.
an internationally agreed upon stan- obtained estimates of the extent of A solution of Bc was prepared by
dard would contribute to optimizing hemolysis in these specimens. adding to pooled human sera a mix-
the CT assay, thereby providing a We prepared a hemolysate from ture of bilirubin mono- and diglucu-
more reliable tool for the treatment EDTA blood drawn from volunteers. ronide isolated from human bile (2 );
of patients with MTC. After centrifugation, the cells were the solution was dispensed into vials,
washed 5 times with physiologic sa- lyophilized, and stored at ⫺70 °C.
References line, diluted with deionized water, Vials containing the pooled human
1. Bieglmayer C, Scheuba C, Niederle B, Flores J, and stored at ⫺20 °C overnight. Af- sera were also lyophilized. The com-
Vierhapper H. Screening for medullary thyroid
carcinoma: experience with different immunoas-
says for human calcitonin. Wien Klin Wochen-
schr 2002;114:267–73.
2. Leboeuf R, Langlois MF, Martin M, Ahnadi CE,
Fink GD. “Hook effect” in calcitonin immunora-
diometric assay in patients with metastatic med-
ullary thyroid carcinoma: case report and review
of the literature. J Clin Endocrinol Metab 2006;
91:361– 4.
3. Bland JM, Altman DG. Statistical methods for
assessing agreement between methods of clin-
ical measurement. Lancet 1986;8:307–10.
4. D’Herbomez M, Leclerc L, Vantyghem MC, Four-
rier F, Proye C, Wemeau JL. Clinical evaluation of
a new sensitive calcitonin assay: study of spec-
ificity. Clin Chim Acta 2001;311:149 –55.
Mariasilvia Tommasi*
Silvia Raspanti
Nuclear Medicine Unit

Department of Clinical
Physiopathology
University of Florence
Florence, Italy
*Address correspondence to this au-

thor at: Sezione di Medicina Nucleare,
Dipartimento di Fisiopatologia Clinica,
V. le Morgagni 85, 50134 Firenze, Italy.
Fax 39055-4224392; e-mail m.tommasi@
dfc.unifi.it.
DOI: 10.1373/clinchem.2006.083733
Total or Neonatal Bilirubin Assays in

the Vitros 5,1 FS: Hemoglobin
Interference, Hemolysis, Icterus Index
To the Editor: Fig. 1. Hb interference in the NBIL method.

We evaluated hemoglobin (Hb) in- (A), pooled human sera enriched with Bu, Bc, and Hb were analyzed with the NBIL method. y⫺y unconjugated
bilirubin; ●⫺● conjugated bilirubin. (B), Bland-Altman bias plot: effect of Hb on the TBIL method at various
terference in the total bilirubin (TBIL) bilirubin concentrations. y⫺y baseline TBIL value, 4 mg/L; ●⫺● baseline TBIL value, 51 mg/L; »⫺» baseline
and neonatal bilirubin (NBIL) [the TBIL value, 124 mg/L; Œ⫺Œ baseline TBIL value, 229 mg/L.
800 Letters
position of bilirubin fractions as de- index corresponds very closely to the total bilirubin. II. Method of Jendrassik-Grof. Clin
Chem 1980;26:26 –9.
termined by HPLC was ␦-bilirubin, Hb concentration in mg/dL. 5. Doumas BT, Wu TW. The measurement of biliru-
6%; bilirubin monoglucuronide, 34%; In the presence of both Bc and Bu, bin fractions in serum. CRC Crit Rev Clin Lab
bilirubin diglucuronide, 57%; and the icterus index varied between 92% Sciences 1991;28:415– 45.
6. Langbaum ME, Farber SJ, Rosenthal P. Auto-
Bu, 3%. The lyophilized materials and 100% of the measured TBIL val- mated total and neonatal bilirubin values in
were rehydrated on the day of use; ues of the test solutions. In the pres- newborns: is a distinction clinically relevant?
dilutions were made with pooled ence of only Bu the agreement was Clin Chem 1992;38:1690 –3.
sera. not as good. Numerically the index
Three solutions of Bu (concentra- corresponds to milligrams per decili- Stanley F. Lo1,2*
tions 51–229 mg/L) and the pooled ter bilirubin. Bernadine Jendrzejczak2
sera were enriched with 4 concentra- In 92 blood specimens from neo- Basil T. Doumas1
tions of hemolysate (concentrations nates 1–14 days old, mean values for 1
1.43– 4.65 g/L). Three solutions of Bc TBIL and NBIL were 102.8 mg/L and Department of Pathology
and the rehydrated pooled sera were 105.5 mg/L, respectively; the regres- Medical College of Wisconsin
enriched with 4 concentrations of he- sion equation was Milwaukee, WI
molysate and a constant amount of
y(NBIL) ⫽ 1.050 ⫻ (TBIL) ⫺ 2.4; 2
Department of Pathology
Bu. The presence of Bu is required
Children’s Hospital of Wisconsin
for obtaining results with the NBIL Sy/x⫽ 5.4. Milwaukee, WI
method. Contrary to our experience,
Rosenthal et al. (3 ) reported that the This result confirms a previous
presence of Bc is required for ob- report (6 ) that differences between Address correspondence to this au-
taining Bu results with the NBIL the 2 methods are negligible. thor at: Department of Pathology/
method. Our results indicate that the TBIL MS701, Children’s Hospital of Wisconsin,
and NBIL methods provide the same 9000 West Wisconsin Ave., Milwaukee,
All specimens were analyzed with WI 53201. Fax 414-266-2779; e-mail
the Vitros 5,1 FS analyzer for TBIL results in blood specimens from stanlo@mcw.edu.
and NBIL. The TBIL assay is a diazo healthy neonates; thus laboratories
method that measures the sum of should use the method they prefer. DOI: 10.1373/clinchem.2006.082586
Bu, Bc, and ␦-bilirubin. The NBIL In cholestasis TBIL may be higher
method is based on direct spectro- than NBIL because the latter does not
photometry and measures Bu and measure ␦-bilirubin, which, because
the sum of bilirubin mono- and di- of its long half-life (17 days), could
glucuronide (Bc), but does not mea- Very High Inhibin A Concentration
persist long after hepatitis has sub-
sure ␦-bilirubin. sided or obstruction has been re- Attributed to Heterophilic Antibody
Hb interference with the NBIL Interference
lieved and obscure the clinical pic-
method was negligible (Fig. 1A). In ture in patients recovering from
the TBIL method (Fig. 1B), Hb in- To the Editor:
hepatobiliary cholestasis. We believe,
terference was positive at low bili- however, that the NBIL method is Maternal serum screening for Down
rubin concentrations and negative at syndrome is commonly performed
superior to TBIL and should be used
high concentrations, a result attribut-
for all specimens regardless of age in the 2nd trimester using ␣ feto-
able to a combination of chemical (4 ) protein (AFP), unconjugated estriol
because it is free from Hb interfer-
and spectral (5 ) interference. In the (uE3), human chorionic gonadotro-
ence and has the advantage of detect-
latter, the absorbance of Hb at low pin (hCG), and inhibin A. Concen-
bilirubin concentrations overcoming cholestasis. We also believe that
trations of each marker are com-
pensates for decreased absorbance the “⬍14-day” restriction recom- bined with maternal age to calculate
caused by the destruction of the azo- mended by the manufacturer, which a patient-specific risk of fetal Down
bilirubin, whereas at high bilirubin discourages use of the NBIL method, syndrome. In cases of Down syn-
concentrations Hb does not compen- is unnecessary. drome, inhibin A concentration is,
sate for this decrease. on average, approximately twice as
The hemolysis index, expressed as References high as in unaffected singleton preg-
1. Doumas BT, Poon PKC, Perry BW, Jendrzejczak
a percentage of the measured Hb B, McComb RB, Schaffer R, Hause LL. Candi- nancies (1 ). Second trimester mater-
concentration, varied from 92%–107% date reference method for determination of total nal serum inhibin A is also increased
in the test solutions, a very good bilirubin in serum: development and validation. in twin pregnancies [1.99 multiples
Clin Chem 1985;31:1779 – 89.
agreement. At high Bu concentra- 2. Lucassen J. The diazo reaction of bilirubin and of the median (MoM) (1 )] and in
tions (⬎200 mg/L) there was no re- bilirubin diglucuronide. PhD thesis, Univ. of Utre- Turner syndrome with hydrops
cht 1961.
sult for this index. In 92 blood spec- 3. Rosenthal P, Keefe MT, Henton D, Cheng M.
(3.91 MoM; (2 )). Markedly increased
imens from neonates 1–14 days old, Ektachem and unconjugated bilirubin measure- inhibin A has been observed in preg-
the index was ⬍15– 49 in 49 speci- ments [Letter]. J Clin Chem Clin Biochem 1989; nancies with complete hydatidiform
27:829.
mens, 50 –99 in 15, 100 –199 in 14, and 4. Shull BC, Lees H, Li PK. Mechanism of inter- mole [4 –7 MoM; (3 )]. Increased in-
200 –250 in only 2. Numerically this ference by hemoglobin in the determination of hibin A may also be seen in non-
pregnant women with ovarian can-

Table 1. Inhibin A concentrations before and after heterophilic blocking
cer (4 ).
We describe a woman having 2nd treatment in prenatal screening samples having an initial inhibin A
trimester serum inhibin A concentra- concentration >5 MoM.
tion 80 times that expected for ges- Inhibin A concentration Inhibin A concentration
Inhibin A, before blocking reagent, after blocking reagent,
tational age. The patient was seen at MoM ng/L ng/L
16 weeks of pregnancy for routine 80.0 13 214 ⬍100a
maternal serum screening. Her re- 6.0 647 171a
sult indicated a low risk for Down 5.4 808 ⬍100a
syndrome (1 in 3300) and no fur- a
Greater than 30% difference between sample inhibin A value before and after heterophilic blocking
ther action was recommended. AFP, reagent pretreatment.
uE3, and hCG were 1.25, 1.00, and
0.74 MoM, respectively, but a mark-
edly increased inhibin A was noted previous screens, 26 samples had before results are reported, to avoid
(13 214 ng/L). The pregnancy and an apparent inhibin A ⬎5 MoM unnecessary concern.
delivery were otherwise unremark- (0.3%), and 2 of 23 tested (3 had
References
able. Because of the patient’s in- insufficient volume) showed a sig- 1. Wald N. Down’s syndrome. In: Wald N and Leck
creased inhibin, she was referred to nificant decrease in results after I, eds. Antenatal and Neonatal Screening. UK:
oncology 9 months postpartum to heterophilic antibody blocking Oxford University Press, 2000:85–115.
2. Lambert-Messerlian GM, Saller DN Jr, Tumber
explore the possibility of an ovarian treatment (Table 1). MB, French CA, Peterson CJ, Canick JA. Sec-
tumor. At that time, her inhibin A Maternal serum concentrations ond-trimester maternal serum inhibin A levels
was again markedly higher (61 362 in fetal trisomy 18 and Turner syndrome with
of the placental secretory products and without hydrops. Prenat Diagn 1998;18:
ng/L) than expected during the nor- inhibin A and hCG are moderately 1061–7.
mal menstrual cycle (10 –160 ng/L). correlated (r ⫽ 0.2– 0.4) (5 ) in the 3. Lambert-Messerlian G, Pinar H, Rubin LP,
Ultrasound and computed tomo- DePaepe ME, Tantravahi U, Steinhoff MM, et
2nd trimester. In this dataset, se- al. Second trimester maternal serum markers
graphic scans of the abdomen and rum hCG MoM values were ⱖ1.8 in twin pregnancy with complete mole: report
pelvis were normal. among those samples having an of 2 cases. Pediatr Dev Pathol 2005;
8:230 – 4.
Inhibin A was measured with the inhibin A MoM ⱖ5.0, with the 4. Robertson DM, Cahir N, Burger HG, Mamers P,
assay from Diagnostic Systems Lab- exception of 2 patients (1.25 and Groome N. Inhibin forms in serum from post-
oratories. Intra- and interassay im- menopausal women with ovarian cancers. Clin
0.71 MoM hCG) that had falsely Endocrinol [Oxf] 1999;50:381– 6.
precision (CV) was ⬍15%, and the increased inhibin A. Furthermore, 5. Haddow JE, Palomaki GE, Knight GJ, Foster
lower limit of detection was 10 ng/L. concentrations of AFP and uE3 DL, Neveux LM. Second trimester screening
The presence of heterophilic anti- for Down’s syndrome using maternal serum
were also unremarkable in these dimeric inhibin A. J Med Screen 1998;5:
body interference was assessed using samples. 115–9.
blocking tubes from Scantibodies Heterophilic antibody interfer-
Laboratory, Inc. A variation in result ence, which has been well docu- Geralyn Lambert-Messerlian1*
⬎30% (2 SD, 95% confidence) after mented in various immunoassays, Christina Bandera2
blocking treatment was considered a most often results in artificially Elizabeth Eklund1
significant change. increased concentrations, and when Andrew Neuhauser3
To examine possible heterophilic unrecognized can lead to unneces- Jacob Canick1
interference in other prenatal screen- sary medical intervention. Immu- 1
ing samples with a high inhibin A noassays can incorporate passive Departments of Pathology and
concentration, we searched labora- blocking solutions to prevent inter- Laboratory Medicine, and
2
tory records compiled during a ference, but particular antibody af- Obstetrics and Gynecology
2-year time period (n ⫽ 9079). An finities may lead to persistence of Women and Infants Hospital and
inhibin A value ⬎5 MoM was con- some heterophilic, interfering anti- Brown Medical School
sidered unusually high (higher than bodies. The impact of an artificially Providence, RI
the median level in twins or preg- increased inhibin A on Down syn- 3
Westerly Hospital
nancies affected by Down or Turner drome screening results is likely
Westerly, RI
syndrome). This study was approved to be minimal, however, because
by the Institutional Review Board for multiple markers are used in risk
Human Studies at Women and In- calculation, never inhibin A alone, * Address correspondence to this au-
fant’s Hospital. and truncation limits of the MoM thor at: Geralyn Lambert-Messerlian, Di-
In the case in question, inhibin A are routinely implemented. Never- vision of Prenatal and Special Testing,
Women and Infants Hospital, 70 Elm
was ⬍10 ng/L (at 1:100 dilution) theless, prenatal screening labora- Street, 2nd floor, Providence, Rhode Is-
after heterophilic antibody block- tories may want to consider using land 02903. Fax 401-276-7882; gmesserl@
ing, indicating that the very high heterophilic antibody blocking re- wihri.org.
result was attributable to interfer- agents as a routine protocol for
DOI: 10.1373/clinchem.2006.082735
ence. In a review of more than 9000 isolated very high inhibin A results,
802 Letters
Plasma Total Coenzyme Q9 (CoQ9) in in the morning after a 10-h overnight Nonparametric statistics were used
the New Zealand Population: fast, at least 1 week apart, over a to determine the reference interval,
Reference Interval and Biological 2-month period. and outliers were included in the
Variation We determined the effect of CoQ10 analysis. Inter- and intraindividual
supplementation on plasma CoQ9 variations were reported as CVs, cor-
Coenzyme Q9 (CoQ9) in human concentrations using the same 10 in- relation as the Pearson correlation
plasma may originate as a product of dividuals. After baseline blood sam- coefficient, and comparisons by the
incomplete CoQ10 biosynthesis or ples were taken, we administered a Mann–Whitney rank-sum test. Statis-
from the diet. The estimated dietary CoQ10 supplement (Q-Gel威, Tishcon tical significance was inferred when
CoQ9 intake is 0 –1.3 ␮mol/day, pri- USA) as a single, nominal 150-mg P ⬍0.05.
marily from cereals and fats (1, 2 ), dose. A vegetarian breakfast and There was no significant difference
but this is unreliable because many lunch were provided. We obtained a in total CoQ9 between males and
food items contain levels below the 2nd blood sample 6 h after adminis- females (Table 1).
detection limit (1 ). tration of the supplement. CoQ9 and CoQ10 were significantly
Some methods for estimating hu- The plasma CoQ9 and CoQ10 as- correlated (r ⫽ ⫹0.577; P ⬍0.001) in
man plasma CoQ10 use CoQ9 as an say, adapted from Tang et al. (3 ), the reference cohort, suggesting that
internal standard; thus, intraindi- measured total CoQ9 and CoQ10 si- CoQ9 could be made during CoQ10
vidual variation in the concentration multaneously. We measured total biosynthesis by omission of 1 isopre-
of endogenous CoQ9 is a potential cholesterol and direct LDL-choles- noid unit. However, the observed
source of error. terol on all samples, using enzymatic correlation was also consistent with a
We measured total CoQ9 in human methods (Aeroset, Abbott Laborato- hypothesis that CoQ9 is a metabolite
plasma, determined a reference in- ries), with coefficients of variation of CoQ10 catabolism.
terval, examined the biological vari- (CVs) of 1.6% and 1.2%, respectively. The intra- and interindividual
ation, and documented the influence Blood samples were collected in variations in CoQ9 were 20.9% and
of CoQ10 supplementation on plasma lithium heparin, centrifuged within 35.5%, respectively. The index of in-
CoQ9 concentrations. 1 h of collection, and plasma was dividuality was 0.6 for CoQ9, calcu-
Self-reportedly healthy volunteers stored protected from light: at lated as within-subject variation/be-
(n ⫽ 193; men/women, 82/113) were ⫺30 °C until analysis (maximum tween-subject variation, whereas
enrolled for determination of a refer- storage time, 112 days) for the refer- that for CoQ10 was 0.4 (4 ).
ence interval for CoQ9. Mean age of ence interval study; and at ⫺80 °C Plasma CoQ9 increased after sup-
participants was 45 years (range 18 – until analysis (maximum storage plementation with CoQ10, with the
82). No participants reported taking time, 150 days) for the biological median (2.5–97.5 percentiles) CoQ9
CoQ10 supplements. variation and effect of CoQ10 sup- concentration at baseline and after
Biological variation of CoQ9 was plementation on plasma total CoQ9 supplementation at 17.17 (9.23–36.81)
determined in healthy adult male concentration studies. These studies and 26.94 (14.49 –39.55) nmol/L, re-
volunteers (n ⫽ 10) who did not were approved by the Canterbury spectively (P ⫽ 0.052). The mean
smoke and did not take vitamin or Ethics Committee, Christchurch, (SD) concentration of CoQ9 ingested
CoQ10 supplements or medications New Zealand, and written informed with the CoQ10 supplements was
in the 4 weeks before the study. consent was obtained from all 2.71 (0.99 ␮g; n ⫽ 6). The mean (SD)
Median age of the participants was participants. CoQ9 concentration of the vegetarian
23.5 years (range 21–28). We took 7 Statistical analysis was carried out breakfast was approximately 134 (63
blood samples from each participant using SigmaStat and SPSS software. ␮g) (1, 2 ). This increase in plasma
Table 1. The median (reference intervala) for total CoQ9, total CoQ10, the CoQ9 to CoQ10 ratio, the CoQ9 to total
cholesterol ratio, and the CoQ9 to LDL-cholesterol ratio.
Females Males P Valueb Total Cohort
Total CoQ9, nmol/L 19.92 (8.79–43.47) 21.19 (8.34–71.80) P ⫽ 0.350 20.70 (8.81–46.97)
Total CoQ10, ␮mol/L 0.85 (0.47–1.71) 0.97 (0.49–2.10) P ⫽ 0.006 0.90 (0.47–1.80)
CoQ9 to CoQ10 ratio, mmol/mol 24.40 (12.06–55.45) 22.55 (9.54–38.79) P ⫽ 0.192 23.90 (11.36–49.84)
Total cholesterol, mmol/L 5.47 (3.80–8.50) 5.70 (3.39–8.54) P ⫽ 0.979 5.59 (3.59–8.48)
CoQ9 to total cholesterol, ␮mol/mol 3.77 (1.57–8.90) 4.03 (1.58–9.26) P ⫽ 0.316 3.85 (1.58–8.90)
CoQ10 to total cholesterol, mmol/mol 160 (90–243) 170 (120–290) P ⬍0.001 160 (100–270)
LDL-cholesterol, mmol/L 2.96 (1.47–5.26) 3.28 (1.59–4.89) P ⫽ 0.350 3.07 (1.49–5.03)
CoQ9 to LDL-cholesterol, ␮mol/mol 6.86 (2.69–18.95) 7.07 (2.93–19.31) P ⫽ 0.916 6.89 (2.90–18.77)
CoQ10 to LDL-cholesterol, mmol/mol 290 (147–486) 305 (200–560) P ⫽ 0.699 290 (159–523)
a
Reference interval is the 2.5–97.5 interfractile interval.
b
P value for comparison between males and females.
CoQ9 after supplementation with analysis. Using the RENCTK reagent

* Address correspondence to this au-
CoQ10 concurred with animal data in thor at: Sarah Molyneux, Biochemistry set (Diasorin) according to the man-
mice and rats (5 ). Unit, Canterbury Health Laboratories, ufacturer’s instructions, 4 well-
The plasma ratio of CoQ9 to CoQ10 P.O. Box 151, Christchurch 8140, New trained and experienced technicians
was significantly lower after CoQ10 Zealand. Fax 64-3-3640889; e-mail sarah. (with an average work experience of
molyneux@cdhb.govt.nz.
supplementation (P ⫽ 0.022), with 26.8 years) performed the tests along
mean (SD) ratio before and after sup- DOI: 10.1373/clinchem.2006.083568 with routine PRA determinations.
plementation being 22.56 (4.88 The results are presented in Table 1.
mmol/mol) and 16.66 (5.66 mmol/ Given the effect of analytical vari-
mol), respectively. ation on the mean values obtained
Our confirmation that CoQ9 is a and using the highest value possible
normal constituent of human plasma Analytical Variation in Plasma Renin at 37 °C with the lowest value at 4 °C
make it less than ideal for use as an Activity: Implications for the and vice versa, PRA values would
internal standard, leading to underes- Screening of Primary Aldosteronism range from 0.44 to 0.81 ␮g/L/h for
timation of CoQ10 by 1% to 7%. This sample 1 and from 0.58 to 0.89 ␮g/
complicates the CoQ10 assay design To the Editor: L/h for sample 2. If these values
and calls for further investigation. The Editorial by Stowasser and Gor- occurred in patients with slightly in-
don recently published in Clinical creased supine aldosterone (e.g., 200
ng/L), with an acceptable CV of 10%
Chemistry is commendable (1 ). We
for aldosterone measurement, the
fully agree with their reservations on
The authors acknowledge Endolab ARR would range from 22 to 50 for
the use of commercial assays by in-
for collaboration in the reference in- patient 1 and from 20 to 38 for pa-
experienced laboratories and their
terval studies, Tishcon for donation tient 2. Therefore, even with major
point that patient management deci-
of the Q-Gel® CoQ10 supplement, sions should not be based on a single
differences in the criterion for a
and the National Heart Foundation raised ARR (2 ), the width of the grey
measurement of the ratio of aldoste-
of New Zealand for funding. zone around the threshold value al-
rone to plasma renin activity (ARR). lows a patient result to be indiscrim-
We would like to reinforce this point inately considered as “normal” or
References by demonstrating the important in-
1. Weber C, Bysted A, Hølmer G. The coenzyme “pathologic” based on a single
Q10 content of the average Danish diet. Int J
fluence of analytical variation on sample.
Vitam Nutr Res 1997;67:123–9. ARR measurement results. This approach, it may be argued, is
2. Mattila P, Lehtonen M, Kumpulainen J. Compar- Plasma renin activity (PRA) is purely analytical, and a slight degra-
ison of in-line connected diode array and electro- evaluated as the difference in angio-
chemical detectors in the high-performance liq- dation of angiotensin I over time
uid chromatographic analysis of coenzymes Q9 tensin I production by the enzyme at could have had an important impact
and Q10 in food materials. J Agric Food Chem 37 °C and endogenous angiotensin I, on the low values measured in this
2000;48:1229 –33. estimated by a determination of an-
3. Tang PH, Miles MV, DeGrauw A, Hershey A,
study. Moreover, endogenous angio-
Pesce A. HPLC analysis of reduced and oxidized
giotensin I at 4 °C. Therefore the CV tensin I concentrations are below the
coenzyme Q10 in human plasma. Clin Chem of the assay is determined by the first quantifiable point of the curve
2001;47:256 – 65. variations of the measurements at and perhaps should not be reported.
4. Molyneux SL, Florkowski CM, Lever M, George 37 °C and 4 °C.
PM. Biological variation of coenzyme Q10. Clin Nevertheless, many studies pub-
Chem 2005;51:455–7. To evaluate the impact of these lished on screening for primary aldo-
5. Kwong LK, Kamzalov S, Rebrin I, Bayne A-CV, assay variations on the ratio results, steronism either did not consider
Jana CK, Morris P, et al. Effects of coenzyme Q10 we performed 5 PRA determinations PRA analytical variation or underes-
administration on its tissue concentrations, mi-
tochondrial oxidant generation, and oxidative during a 5-week period on 2 samples timated it because activities in the
stress in the rat. Free Radic Biol Med 2002;33: with activities of ⬃0.6 and ⬃0.7 ␮g/ samples used for estimating be-
627–38. L/h. The samples were divided into tween-assay variation were much
aliquots and kept at ⫺80 °C until higher than 0.6 or 0.7 ␮g/L/h, lead-
Sarah Molyneux1,2*
Michael Lever1,2
Christopher Florkowski1
Peter George1 Table 1. Plasma renin activity results for 5 measurements of 2 samples.
Angiotensin I produced after an Endogenous
1 incubation of 1 h at 37 °C angiotensin I PRA result
Biochemistry Department
Canterbury Health Laboratories Sample 1
Mean (SD), ␮g/L/h 0.69 (0.14) 0.06 (0.05) 0.63 (0.15)
Christchurch, New Zealand
CV, % 20.1 84.2 23.7
2 Sample 2
Department of Chemistry
Mean (SD), ␮g/L/h 0.88 (0.12) 0.14 (0.03) 0.73 (0.10)
University of Canterbury
CV, % 14.1 22.3 13.5
Christchurch, New Zealand
804 Letters
ing to lower CVs (3 ). Even if accu- The Impact of Simultaneous its impact, we collated data on 1436
racy of the assay can be improved for Measurement of Testosterone and women from samples received in the
values ⬍1 ␮g/L/h by incorporating Androstenedione in Women with laboratory for assessments of their
a long incubation (18 h) (4 ), PRA Suspected Androgen Excess androgen status between June 2005
determination is and will remain and October 2006. Local ethics com-
very difficult because of the impor- Liquid chromatography tandem mittee approval was received for this
tant effects of pH, time, and temper- mass spectrometry (LC-MS/MS) is study. The upper limits of our refer-
ature conditions and the combined becoming more widely available in ence intervals were 1.6 nmol/L for
variation of 2 angiotensin I deter- clinical laboratories, allowing the de- testosterone, 4.7 nmol/L for andro-
minations, sometimes at very low velopment of methods that measure stenedione, and 6.5 for FAI [FAI ⫽
concentrations. multiple analytes simultaneously. (testosterone/SHBG) ⫻ 100] (2 ).
Our results emphasize that ARR Moving away from traditional anal- Of the 1436 participating women,
determined on a single blood sample ysis of single analytes may have un- 876 showed no increases of testoster-
should not be used for primary aldo- expected consequences, however, as one, androstenedione, or FAI. The
steronism screening. The value of we report here. samples from the remaining 560
performing a 2nd ARR determina- An assessment of biochemical an- women had at least 1 increased an-
tion to reduce the uncertainty associ- drogen status in women is com- drogen (Fig. 1).
ated with a single ARR measurement monly requested when a patient pre- Simultaneous assays provide more
should be evaluated in a prospective sents with clinical features of information and have faster turn-
study. Nevertheless, we believe that hyperandrogenism, such as hirsut- around times. The provision of 2
ism or acne, and also in the investi- steroid values, at no extra expense,
because of analytical variation a grey
gation of infertility. Conditions that allows the clinician to bring more
zone will always exist around the
characteristically show androgen ex- confidence to the acceptance or rejec-
clinical threshold.
cess include polycystic ovarian syn- tion of diagnoses associated with an-
drome (PCOS), congenital adrenal drogen excess. Here, testosterone,
hyperplasia (CAH), and androgen- FAI, or both were increased in the
References
1. Stowasser M, Gordon RD. Aldosterone assays:
secreting tumors (1 ). samples of 396 women (27.5% of the
an urgent need for improvement. Clin Chem Our previous protocol for the as- total). When tested with the old pro-
2006;52:1640 –2. sessment of androgen status in fe- tocol, before LC-MS/MS, these pa-
2. Kaplan NM. Cautions over the current epidemic males was to measure testosterone tients would then have undergone
of primary aldosteronism. Lancet 2001;357:
953– 4. before 2nd- and 3rd-line tests mea- measurement of androstenedione
3. Schwartz GL, Turner ST. Screening for primary suring sex hormone binding globulin concentration. Therefore, these pa-
aldosteronism in essential hypertension: diag- (SHBG) to calculate the free andro- tients received faster service with the
nostic accuracy of the ratio of plasma aldoste- gen index (FAI) (2 ) and andro- simultaneous assay because the an-
rone concentration to plasma renin activity. Clin
Chem 2005;51:386 –94. stenedione, respectively. drostenedione concentration had al-
4. Sealey JE, Laragh JH. Radioimmunoassay of Since June 2005, an LC-MS/MS ready been measured.
plasma renin activity. Semin Nucl Med 1975;5: assay for the simultaneous measure- An unexpected finding was the
189 –202.
ment of testosterone and andro- identification of a group of 164 pa-
stenedione in samples from women tients (11.4% of the total) who had an
Etienne Cavalier1*
has been in routine use. To evaluate increased androstenedione concen-
Pierre Delanaye2
Jean-Marie Krzesinski2
Jean-Paul Chapelle1
Departments of
1
Clinical Chemistry and
2
Nephrology and Hypertension
University Hospital of Liege
University of Liege
Liege, Belgium
* Address correspondence to this au-

thor at: Service de Chimie Médicale, Cen-
tre Hospitalier Universitaire de Liège,
Domaine du Sart-Tilman, B-4000 Liège,
Belgique. Fax 32-4-3667691; e-mail
Etienne.cavalier@chu.ulg.ac.be.
DOI: 10.1373/clinchem.2006.083600 Fig. 1. Venn diagram to show the distribution of results of the 560 samples with 1 or more
increased androgen measurements.
tration without increased testoster- tients meet 2 of the following 3 crite- ularity, and additional work could
one or FAI. In these patients, the ria (5 ): include the measurement of 17-hy-
mean androstenedione concentration 1. Oligo- or anovulation. droxyprogesterone in such patients
was 5.8 nmol/L, with a range of 4.8 2. Clinical and/or biochemical signs to further investigate this hypothesis.
nmol/L to 15.3 nmol/L. This group of hyperandrogenism.
of patients would not have been 3. Polycystic ovaries on ultrasound. References
identified using the old protocol and 1. Azziz R, Sanchez LA, Knochenhauer ES, Moran
thus no diagnostic questions would Specific guidance as to what actu- C, Lazenby J, Stephens KC, et al. Androgen
excess in women: experience with over 1000
have been raised. The finding of this ally constitutes biochemical hyperan- consecutive patients. J Clin Endocrinol Metab
group of patients raises several ques- drogenism is hard to find, however, 2004;89:453– 62.
and it is difficult to say whether or 2. Marthur RS, Moody LO, Landgrebbe S, William-
tions—most pertinently, is this status son HO. Plasma androgens and sex hormone
a variant of normal, or is it of clinical not an increased androstenedione binding globulin in the evaluation of hirsute
significance? concentration without increased tes- patients. Fertil Steril 1981;35:29 –37.
tosterone or increased FAI should be 3. Burger HG. Androgen production in women.
Androstenedione is often consid- Fertil Steril 2002;77:S3–5.
ered a proandrogen because it re- considered biochemical hyperandro- 4. Abraham GE. Ovarian and adrenal contribution
quires conversion to testosterone to genism. to peripheral androgens during the menstrual
cycle. J Clin Endocrinol Metab 1974;39:340.
exert its androgenic effects (3 ). This The specific constituents of in-
5. The Rotterdam ESHRE/ASRM-sponsored Con-
characteristic, together with the fact creased biochemical androgen action sensus Workshop Group: revised 2003 con-
that androstenedione concentrations remain to be elucidated. The use of sensus on diagnostic criteria and long-term
health risks related to polycystic ovary syn-
increase in the 2nd half of the men- sensitive, functional, androgen-de- drome. Fertil Steril 2004;81:19 –25.
strual cycle (4 ), may suggest that an pendent parameters measured under
isolated increase in androstenedione physiological conditions may enable Katherine Duxbury*
is of no clinical significance. By vir- identification and ranking of the Louise Gallagher
tue of its much decreased binding to most important predictors for hy- Brian Keevil
SHBG, however, androstenedione perandrogenism. Until these in vivo
has higher bioavailability in serum data are available, the role of andro- University Hospital
than does testosterone and thus has gens other than testosterone in the of South Manchester
important androgenic potential. An- diagnosis of biochemical hyperan- NHS Foundation Trust
drostenedione concentrations up to drogenism is unknown. Manchester, United Kingdom
15.3 nmol/L were seen in the group It is also important to consider
of patients we identified, and this whether an isolated increased andro- * Address correspondence to this au-
finding likely warrants further expla- stenedione value may have a partic- thor at: Department of Clinical Bio-
nation. ular clinical significance. A possibil- chemistry, Clinical Sciences Building,
Biochemical hyperandrogenism is ity is that some of the patients may Wythenshawe Hospital, Southmoor Rd,
Manchester M23 9LT, United Kingdom.
a positive indicator for conditions have nonclassical CAH. Nonclassi- Fax 0161-291-2927; e-mail k.j.duxbury.
that feature androgen excess. For ex- cal, or late-onset, CAH may share 00@cantab.net.
ample, the current consensus for a many of the features of PCOS such as
DOI: 10.1373/clinchem.2006.083782
diagnosis of PCOS requires that pa- acne, hirsutism, and menstrual irreg-
Correction
In the editorial entitled “Clinical Urinary Peptidomics: Learning to Walk Before We Can Run”, by Anthony
G.W. Norden, Pedro Rodriguez-Cutillas, and Robert J. Unwin Lindmans (Clin Chem 2007;53:375– 6; DOI:
10.1373/clinchem.2006.084038), the first reference contained the wrong page numbers. The correct citation
for reference 1, by Fiedler et al, is Clinical Chemistry 2007;53:421–28. The printer regrets the error.
DOI: 10.1373/clinchem.2007.088401
Book, Software, and Web Site Reviews
An A-Z Guide to Pharmacogenomics. or in routine clinical practice. Never- large differences in the ability of var-
Michael G. Catania. Washington DC: theless, for PGx to be effective in ious peoples to respond to different
AACC Press, 2006, 58 pp., $24 ($19 medical practice, it must be applied drugs.
AACC members), softcover. ISBN by practicing physicians and other Because there are today only a
1-59425-047-2. healthcare providers involved in pre- handful of well-characterized exam-
scribing, monitoring, or dispensing ples of genes affecting drug re-
What is an allele or a genetic variant, of medications. In addition, we must sponse, the book contains significant
and how is it related to drug dosing? not forget or underestimate the need
What do the acronyms ADMET and duplication of information among
to provide this knowledge to other the chapters, and the same informa-
RFLP mean? What is a CYP450 or its important users of this information
genetic variant? For those uninitiated tion is presented in slightly differing
such as individuals involved as 3rd- formats by multiple authors. Some of
into the new realm of “pharmaco- party payers or in vitro diagnostic
genomics”, for whom the meanings the chapters focus on techniques that
manufacturers, as well as regulators are used primarily for target discov-
of these words are not clear or need of those products. In my opinion, all
refreshing, this little book, packed ery and information management,
of these practitioners will greatly while others are clinically oriented.
with useful definitions, will be in- benefit from this little book of valu-
valuable. Even those already involved The editors should have done more
able definitions—it is simple to read, to create a coherent text instead of
in this discipline may benefit from the and I recommend it.
100-plus definitions and cross-refer- simply stringing together a set of
ences provided in this new text just chapters written by multiple authors
Roland Valdes, Jr
released by AACC Press. Words are with different writing styles. One
important, and definitions, once stan- wishes for better copy editing as
University of Louisville
dardized, provide an important basis well, because of the numerous typo-
School of Medicine
for mutual understanding. graphical and grammatical errors
Department of Pathology
The field of pharmacogenetics that appear throughout the book.
and Lab Medicine
(PGx)—and more broadly, pharma- Louisville, KY 40202 Setting aside these structural criti-
cogenomics—is relatively new to lab- cisms, the highlights of the book in-
oratory medicine, and the buzz sur- DOI: 10.1373/clinchem.2006.080507 clude a well-written opening chapter
rounding this application in clinical by the senior editor giving a histori-
settings is growing so rapidly that cal perspective on the field, excellent
a compendium of definitions is most tables on details surrounding poly-
welcome. Dr. Catania takes the Pharmacogenomics, Second Edi- morphisms in a variety of relevant
pharmacogenomics lexicon from A tion. Werner Kalow, Urs B. Meyer, genes and their clinical effects in
(ADMET) to Z (Zenobiotic) and pro- and Rachel F. Tyndale, eds. Boca chapters 5 and 6, a wonderful review
vides not only definitions of words but Raton, FL: Taylor & Francis Group,
of the pharmacogenetics of cardiac
also explanations of some important 2005, 696 pp., $199.95, hardcover.
ion channels, and an excellent over-
fundamental processes, which will be ISNB 1-57444-878-1.
view of SNP analysis methods. Peter
greatly appreciated by those just be- This text covers the range of applica- Ray provides insight into the rela-
coming familiar with PGx as a disci- tions from drug discovery and selec- tionship between pharmacogenetic
pline. My only recommendation in tion to clinical utility and represents testing and genetic testing for mono-
reading through this interesting text is an updated and enriched edition of genic disorders. The new chapter on
that it omits consideration of the use of the original book. New chapters cov- haplotyping provides an easy-to-un-
the nomenclature “genetic variant” ering human drug metabolizing en- derstand description of the concept
and not “mutation” as these may ap- zymes, pharmacogenetics of cardiac of haplotype and of how knowledge
ply to PGx. ion channels, the FDA’s perspective from the recently completed Hap-
The relatively new application of on pharmacogenetics, metabolomics, Map program will impact the field of
PGx principles to medical practice haplotype structure and the HapMap pharmacogenetics.
combines the disciplines of genetics project, epigenetics, and a review of Overall, I would recommend using
and pharmacology and thus poses known disease associations have this book as a reference text, rather
challenges to healthcare practitioners been added to the previous edition. than as a teaching text or as an intro-
who, during their training, may not Most of the older chapters have been duction to the field.
have been exposed to concepts com- successfully updated; however, the
bining 2 disciplines that originally chapter by Glen Miller on technolo- Emily S. Winn-Deen
seemed unrelated but are now inex- gies for pharmacogenetic analysis is
tricably linked. For example, terms significantly out of date. Ethical is- Cepheid
such as allele, genetic variant, RFLP, sues were dealt with very lightly— Sunnyvale, CA
phenotyping, surrogate marker, and issues of race and ethnicity were ad-
theranostics are not commonly ap- dressed only in the context of how DOI: 10.1373/clinchem.2006.079350
plied in routine laboratory medicine population divergence can lead to
806 Clinical Chemistry 53, No. 4, 2007

Clinical Chemistry

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Clinical Chemistry

Uploaded by

Copyright:

Available Formats

April 1 2007, Volume 53, Issue 4 , pp.

Current Issues in Laboratory Medicine:

Molecular Diagnostics and Genetics:

Evidence-Based Laboratory Medicine and Test

Hemostasis and Thrombosis:

Proteomics and Protein Markers:

Lipids, Lipoproteins, and Cardiovascular Risk Factors:

Drug Monitoring and Toxicology:

Endocrinology and Metabolism:

Automation and Analytical Techniques:

Clinical Chemistry 53, No. 4, 2007 543

National Academy of Clinical Biochemistry

This article has been copublished in the April 3, 2007 online

National Academy of Clinical Biochemistry and

National Academy of Clinical Biochemistry

I. OVERVIEW OF THE ACUTE CORONARY SYN- b. Relationship to Clinical Outcomes .......558

I. Overview of the Acute Coronary Syndrome

nontraumatic chest pain are 2-fold: (a) to assess the

Table 1. Properties of biomarkers of myocardial necrosis.

class iib b. Relationship to clinical outcomes

Table 2. Summary of clinical studies of BNP and NT-proBNP in ACS.

quent risk of short- and long-term mortality was evident

Table 3. Summary of clinical studies of CRP in ACS.

Early invasive strategy

IV. Use of Biochemical Markers in the Management of

Evaluation of the Performance of Randomized

Materials and Methods

Fig. 3. The average time to the next scheduled QC

Real-time PCR Assay for Ultrasensitive

Fig. 1. Real-time PCR for quantification of DNA-binding proteins.

Fig. 2. Specificity of NF-␬B detection.

in the online Data Supplement). The results indicated that

Preanalytical mRNA Stabilization of Whole Bone

Fig. 1. Flowchart of the study design.

Fig. 2. IL8 gene expression in RNA sam-

Fig. 3. RNA yield and integrity.

paxgene stabilizes gene expression profiles

gene-dependent change of expression level

One-Step Rapid Reverse Transcription–PCR Assay

Fig. 1. Primer design for the 1-step RT-PCR LightCycler assay.

Table 1. Primer and amplicon characteristics.

Table 2. Viral strains analyzed with the 1-step

Viral culture, Clinical serum,

Candidate Genetic Risk Factors of Stroke: Results

Patients, Materials and Methods based mainly on associations with cardiovascular

Table 1. Baseline characteristics of patients and controls.

Fig. 1. Graphical representation of

Real-Time PCR to Identify Variola Virus or Other

and Daniel Garin1*

pcr primers, target sequences, and fluorogenic 5⬘ nuclease pcr assay

Fig. 1. Sequence alignment of the

14-kD protein-encoding gene (Fig. 2A). The LOD was also

pox, cowpox and vaccinia virus with the 2 probes mixed

Dynamic range of the assay (Fig. 2C and 2D).

intranasal infection with CPV, there were ⬃12 ⫻ 106 DNA

Table 1. Keys to analysis of results and conclusion.

Determination of Aspirin Responsiveness by Use

patients (180 men and 65 women; ages 29 to 90 years,

Fig. 2. Box plots of 6-min impedance results of collagen-

statistical analysis men and 6 women; ages 22 to 58 years, median 34 years)

Fig. 3. Distributions of 6-min impedance results of collagen-induced

suggesting that the full inhibitory potential of ASA was not

aggregation, when a patient is already taking ASA. This

Mass Spectrometry–Based Hepcidin Measurements