You are on page 1of 3

Anal Bioanal Chem (2007) 387: 525–527

DOI 10.1007/s00216-006-0687-8

TRENDS

Zhengzheng Pan . Daniel Raftery

Comparing and combining NMR spectroscopy and mass


spectrometry in metabolomics

Published online: 12 August 2006


# Springer-Verlag 2006

Introduction largely limited to volatile compounds. The developments


in LC (liquid chromatography)–MS and CE (capillary
The rapidly expanding field of metabolomics has been electrophoresis)–MS have significantly broadened the
driven in recent years by advances in the analytical applicability of MS-based metabolomics [8], although the
methods of high-resolution nuclear magnetic resonance need to separate or purify the sample before it is directed
(NMR) spectroscopy and mass spectrometry (MS), and into the mass analyzer can make the analysis time-
their combination with multivariate statistics [1, 2]. consuming. In addition, the selectivity of GC/LC–MS for
Widespread application of metabolomics has been classes of analytes provides both benefits and complica-
achieved in the areas of system biology, drug discovery, tions. Tandem MS and so-called accurate mass (time of
pharmaceutical research, early disease detection, toxicol- flight) methods are most often used to validate the
ogy, food and nutrition science and others [3–6]. The identities of unknown analytes. Fourier transform (FT)
potential scientific payoff is also driving the development MS provides an alternative approach with extremely high
of new methodologies that promise to fuel a further resolution and a mass accuracy better than 1 ppm [9].
expansion of the field. In addition, because of the However, the high cost of the instrument has limited its use
complementary analytical features of NMR and MS, to date. Current challenges for MS-based metabolomics
opportunities for leveraging both methods are being include the development of more robust methods for
considered which will create more comprehensive meta- chromatographic separation, data reduction methods to
bolic profiling. While only a few studies to date have deal with the enormous file sizes, and of course the
combined both analytical techniques with statistical anal- reduction of matrix effects, including ion suppression, that
ysis [7], this approach has exciting potential for the field. can cause widely varying signal intensities.
Compared with MS, NMR spectroscopy yields rela-
tively low-sensitivity measurements, with limits of detec-
Current state of metabolomics tion on the order of 10 μM or a few nmol at high fields
using new cryoprobes. Nevertheless, NMR-based meta-
The high sensitivity (typically pg level) of MS detection bolic profiling can be performed successfully because
makes it an important method for measuring metabolites in NMR is highly quantitative and reproducible. These
complex biosamples. While electrochemical methods properties are especially important when metabolomic
provide excellent sensitivity, MS methods are more data are analyzed by multivariate statistical methods [10],
popular because they allow for reliable metabolite identi- and they compensate for the limited sensitivity of NMR.
fication. Since the 1970s, chromatographic methods have Furthermore, NMR sensitivity does not depend on metab-
been used to separate complex mixtures of metabolites for olite pKa or hydrophobicity, which makes NMR an
improved analysis and identification. While GC (gas excellent choice for broad-based analyses (see Fig. 1),
chromatography) and later GC–MS methods have been and suitable for samples in different conditions. Recent
used for quantitative metabolic profiling, GC–MS is clinical examples of this approach include promising
studies involving the detection of ovarian cancer and
inborn errors of metabolism [11, 12]. Progress has been
made to improve the resolution and sensitivity of NMR
Z. Pan . D. Raftery (*) spectroscopy by means of the J-resolved experiment [13]
Department of Chemistry, Purdue University,
560 Oval Drive, and other 2-D methods. Using advanced high-throughput
West Lafayette, IN 47907, USA NMR methodology, up to 500 samples can be measured
e-mail: raftery@purdue.edu within one day with the assistance of flow-injection probes
526

New methods

While the current technologies used in metabolomics are


already fueling a growing number of applications, new
approaches are certain to accelerate this trend. For MS,
new atmospheric pressure ionization methods dramatically
simplify the sample treatment and introduction process. For
example, desorption electrospray ionization (DESI) MS
makes it possible to measure biosamples without pre-
separation or sample pretreatment [16]. DESI utilizes
charged droplets to ionize chemical species from a surface
located outside the mass analyzer. A somewhat similar
approach is used by the DART (direct analysis in real time)
method [17]. These direct measurements significantly
shorten the experimental time compared to GC– or LC–
MS analysis. DESI spectra can be used as inputs for
statistical analysis for metabolomics studies, and initial
studies have been carried out to provide fast and
comprehensive measurements of urine samples from
tumor-bearing mice [18] and the detection of inborn errors
Fig. 1 A schematic comparison of NMR spectroscopy and mass of metabolism [19]. However, two potential concerns for
spectrometry as a function of different metabolite characteristics this approach are the overshadowing of small signals by
larger ones (similarly encountered in NMR) and ion
and automated liquid handlers. The detection limit can also suppression, which could limit their utility for broad-based
be decreased to tens of nanograms through the use of metabolomics.
microcoil probes [14]. However, the complexity inherent in New NMR methods have focused on improving the
biosamples generates a large number of peaks in a small resolution in complex spectra. A relatively simple approach
chemical shift range (∼10 ppm) in the 1H-NMR spectrum. is to fractionate the hydrophilic and hydrophobic compo-
Therefore, peaks generated by different species have more nents, which improves the resolution somewhat [20]. As an
chance of overlapping with each other in the spectrum, alternative approach, the use of the selective TOCSY (total
especially in the aliphatic region. As a result, potentially correlation spectroscopy) experiment [21] allows one to
important compounds present at smaller concentrations are focus on certain target molecules with much improved
often overshadowed by larger peaks, and are thus less resolution and sensitivity. NMR detection in this case is
likely to be analyzed. focused on J-coupling in the system, i.e., hydrogen atoms
Statistical pattern recognition methods, when combined linked by covalent bonds in a single molecule. Using
with either MS or NMR, create enormous opportunities for selective TOCSY, important metabolites, which may
metabolomics research beyond simple data reduction appear as low-concentration species, can be emphasized
methods. Among a variety of statistical methods employed, so as to achieve a better discriminatory ability when
principal component analysis (PCA) has proven to be performing metabolomics analyses [22]. Selective TOCSY
suitable for differentiating sample groups on the basis of can be multiplexed to improve throughput using Hadamard
metabolite composition. As an unsupervised method, PCA methods [23].
requires no knowledge of the nature of samples prior to the An exciting recent development is the combination of
data analysis, and any significant difference between the NMR and MS data, which allows improved identification
groups of samples will be detected, whether or not this of unknown analytes and creates an opportunity to expand
difference is important to the study. Other methods are the scope of metabolomics research. A statistical correla-
described as “supervised” because classes of samples are tion tool has been developed by Nicholson and coworkers
identified before the mathematical analysis begins. These to interrelate the metabolites from NMR and UPLC-MS
include partial least squares descriminant analysis (PLS- [7]. Different metabolites, including some possible bio-
DA) and soft independent modeling of class analogies markers, measured with NMR and UPLC-MS individually
(SIMCA). These methods have been used for the classi- were linked by their mutual correlation values and then
fication of spectroscopic results, and are especially useful illustrated by a corresponding contour plot. This correla-
for detecting markers which differentiate the pre-identified tion approach has been applied to the study of inborn errors
classes of samples, such as diseased and undiseased. In of metabolism using the unique combination of NMR and
addition, variable stability scaling (VAST) and orthogonal DESI-MS [19]. Correlation of several known biomarkers
signal correction (OSC) methods have been reported to identified by NMR spectroscopy with a group of
improve NMR-based metabolomics sample classification metabolites seen in the DESI–MS results allowed a new
[15]. set of metabolites to be studied and rationalized in terms of
527

their relationship to the disease and their impact on related 4. Fiehn O, Kopka J, Dormann P, Altmann T, Trethewey RN,
metabolic pathways. Willmitzer L (2000) Nat Biotechnol 18:1157–1161
5. Waters NJ, Holmes E, Williams A, Waterfield CJ, Farrant RD,
Nicholson JK (2001) Chem Res Toxicol 14:1401–1412
6. Goodacre R, Vaidyanathan S, Dunn WB, Harrigan GG, Kell
Outlook DB (2004) Trends Biotechnol 22:245–252
7. Crockford DJ, Holmes E, Lindon JC, Plumb RS, Zirah S, Bruce
SJ, Rainville P, Stumpf CL, Nicholson JK (2006) Anal Chem
A key goal of metabolomics is the selection of a set of 78:363–371
reliable biomarker metabolites that can be used to identify 8. Van der Greef J, Smilde AK (2005) J Chemometr 19:376–386
and follow changes in systems biology. To this end, more 9. Brown SC, Kruppa G, Dasseux JL (2005) Mass Spectrom Rev
sensitive, reliable, faster, cheaper, and easier analytical 24:223–231
10. Nicholson JK, Wilson ID (1989) Prog Nucl Magn Reson Sp
tools to perform metabolomics research are highly 21:449–501
desirable, and thus provide an impetus for instrument 11. Constantinou MA, Papakonstantinou E, Spraul M, Sevastiadou
development. New statistical tools, such as advanced S, Costalos C, Koupparis MA, Shulpis K, Tsantili-Kakoulidou
multivariate analyses and correlation methods, will likely A, Mikros E (2005) Anal Chim Acta 542:169–177
be developed and applied to different types of samples, 12. Odunsi K, Wollman RM, Ambrosone CB, Hutson A, McCann
SE, Tammela J, Geisler JP, Miller G, Sellers T, Cliby W, Qian F,
including various biofluids and even tissues, etc. In the Keitz B, Intengan M, Lele S, Alderfer JL (2005) Int J Cancer
longer run it may also be possible to correlate metabo- 113:782–788
lomics more extensively with proteomics and genomics, 13. Viant MR (2003) Biochem Bioph Res Co 310:943–948
although this has been an overwhelmingly difficult 14. Lacey ME, Subramanian R, Olson DL, Webb AG, Sweedler JV
(1999) Chem Rev 99:3133–3152
problem to date. Ultimately, the real challenge will be to 15. Keun HC, Ebbels TMD, Antti H, Bollard ME, Beckonert O,
place discovered biomarkers in their biochemical contexts Holmes E, Lindon JC, Nicholson JK (2003) Anal Chim Acta
in terms of pathways and processes. For the present, 490:265–276
however, advanced and correlated NMR- and MS-based 16. Cooks RG, Ouyang Z, Takats Z, Wiseman JM (2006) Science
311:1566–1570
analyses will definitely move the field of metabolomics 17. Cody RB, Laramee JA, Durst HD (2005) Anal Chem 77:
into a new era. 2297–2302
18. Chen H, Pan Z, Talaty N, Raftery D, Cooks RG (2006) Rapid
Commun Mass Spectrom 20:1577–1584
References 19. Pan Z, Gu H, Talaty N, Chen H, Hainline BE, Cooks RG,
Raftery D (2006) Anal Bioanal Chem (in press)
20. Serkova N, Fuller TF, Klawitter J, Freise CE, Niemann CU
1. Lindon JC, Holmes E, Nicholson JK (2004) Prog Nucl Magn (2005) Kidney Int 67:1142–1151
Reson Sp 45:109–143 21. Sharman GJ (1999) Chem Comm:1319–1320
2. Wagner S, Scholz K, Donegan M, Burton L, Wingate J, Volkel 22. Sandusky P, Raftery D (2005) Anal Chem 77:2455–2463
W (2006) Anal Chem 78:1296–1305 23. Kupce E, Freeman R (2003) J Biomol NMR 25:349–354
3. Watkins SM, German JB (2002) Curr Opin Mol Therap 4:
224–228

You might also like