Professional Documents
Culture Documents
Background: Progress in robotic printing technology tially important advantages in throughput, convenience,
has allowed the development of high-density nucleic and cost.
acid and protein arrays that have increased the through- © 2002 American Association for Clinical Chemistry
put of a variety of assays. We generated protein microar-
rays by printing microbial antigens to simultaneously Assays with the ability to detect in parallel antibodies
determine in human sera antibodies directed against with different specificities have a wide range of potential
Toxoplasma gondii, rubella virus, cytomegalovirus applications in epidemiologic research and vaccine devel-
(CMV), and herpes simplex virus (HSV) types 1 and 2 opment as well as in the diagnosis of allergies and
(ToRCH antigens). autoimmune and infectious diseases. Although ELISA-
Methods: The antigens were printed on activated glass based tests are suitable for this purpose, available assays
slides with high-speed robotics. The slides were incu- can be time-consuming and require large quantities of
bated first with serum samples and subsequently with both sample and reagents, thus limiting their application
fluorescently labeled secondary antibodies. Human IgG for mass screening (1, 2 ). The advent of automated, robust
and IgM bound to the printed antigens were detected by microdeposition technologies has allowed the develop-
confocal scanning microscopy and quantified with in- ment of high-density ordered arrays of molecules (mi-
ternal calibration curves. Both microarrays and commer- croarrays) (3 ). This assay format incorporates key fea-
cial ELISAs were utilized to detect serum antibodies tures, such as true parallelism, miniaturization, and high
against the ToRCH antigens in a panel of characterized throughput, that could overcome most of the current
human sera. ELISA limitations (4 ). Studies using DNA microarrays
Results: The detection limit (mean ⴙ 2 SD) of the have shown that this assay format is ideal for studying the
microarray assay was 0.5 pg of IgG or IgM bound to the expression profiles of thousands of genes simultaneously
slides. Within-slide, between-slide, and between-batch in tissues and organs under different experimental condi-
precision profiles showed CVs of 1.7–18% for all anti- tions (5–7 ). Currently, DNA microarrays are regarded as
gens. Overall, >80% concordance was obtained between one of the most powerful available tools for understand-
microarray assays and ELISAs in the classification of ing the functional relationships between the large reper-
sera; for T. gondii, CMV, and HSV1, concordance ex- toires of genes available in databases generated as a result
ceeded 90%. of genome sequencing projects (8, 9 ).
Conclusions: The microarray is a suitable assay format Ordered arrays of peptides have been used for unrav-
for the serodiagnosis of infectious diseases and can be eling protein–protein and protein–nucleic acid interac-
easily optimized for clinical use. The ToRCH assay tions (10, 11 ). Previous reports have shown that protein
performs equivalently to ELISA and may have poten- arrays printed on membranes can be used to screen
expression libraries for binding to target molecules as well
as to unravel protein binding to either RNA or DNA
1
Department of Biology, Imperial College of Science, Technology and (12–15 ). More recently, arrays have been generated for
Medicine, London SW7 2AZ, United Kingdom. high-throughput screening of recombinant antibodies
2
Dipartimento di Medicina Clinica e Sperimentale Università degli Studi
di Perugia, Via del Giochetto, 006100 Perugia, Italy.
(16 ). These antibody arrays contain thousands of bacterial
*These authors contributed equally to this work. clones, each expressing a different single-chain antibody.
†Address correspondence to this author at: Imperial College of Science, Arrays of proteins covalently bound to glass slides were
Technology and Medicine, Department of Biology, Biomedical Sciences Build-
shown to retain their ability to interact specifically with
ing, 5th Floor, Imperial College Road, London SW7 2AZ, United Kingdom. Fax
44-207-5945439; e-mail acrs@ic.ac.uk. other proteins or with small molecules in solution (17 ).
Received August 24, 2001; accepted October 15, 2001. Protein arrays have also been used to develop compara-
121
122 Mezzasoma et al.: Microarrays for Serodiagnosis
tive fluorescence assays to measure the concentrations of incorporated in the array as a reference signal. T. gondii,
several specific proteins and antibodies in complex solu- CMV, and HSV1 and HSV2 preparations were supplied
tions (18 –21 ). These studies have shown that protein by Radim S.p.A. Rubella virus antigen (grade K1S) was
microarrays could provide a practical means to quantify purchased from Microbix Biosystems Inc. Human IgG
thousands of different protein species in clinical or re- and IgM (reagent grade) were purchased from Sigma
search applications. In spite of these advances, the devel- Chemical Company. Antigens were dissolved in phos-
opment of protein arrays for research and clinical appli- phate-buffered saline [PBS; 1⫻ (0.2 g/L KCl, 1.44 g/L
cations has lagged because of the poor stability of Na2HPO4, 0.24 g/L KH2PO4, 8 g/L NaCl, pH 7.4)] con-
proteins, complex coupling chemistry, and weak detec- taining Tween 20 (0.1 mL/L). Polyvinylpyrrolidone (10
tion signals. High-density protein arrays remain difficult g/L) was added to the human IgG and HSV1 antigen
to generate and to validate for clinical use. solutions, sodium dodecyl sulfate (0.1 g/L) was added to
We used a protein microarray to determine in human the human IgM solution, and sucrose (100 g/L) was
sera the presence of specific antibodies directed against added to the CMV antigen solution. No additives were
parasite and viral antigens, including Toxoplasma gondii, present in the other antigen preparations. Printing was
cytomegalovirus (CMV),3 rubella virus, and herpes sim- performed in a cabinet at 25 °C and 55% humidity. These
plex virus (HSV) types 1 and 2. The clinical performance conditions were constantly monitored by a thermohy-
of this assay was validated with a collection of sera grometer. Printed slides were stored for 12 h inside the
previously characterized with commercial ELISAs for cabinet before removal and storage. Slides were stored in
their reactivity against the selected microbial antigens. boxes at room temperature in the presence of silica gel bags
as desiccant and used within 90 days of being printed.
Materials and Methods
preparation of microarrays serum samples and elisa methods
Microbial antigens as well as human IgG and IgM at Human sera (n ⫽ 60) were supplied by BioMedical
different concentrations were printed on silanized glass Resources Inc. The sera were arranged in five panels on
microscope slides (CEL Associates), using computer-con- the basis of their reactivity against T. gondii, rubella virus,
trolled high-speed robotics (Total Array System; Biorobot- HSV1, HSV2, and CMV as measured by Abbott IMx
ics) (22 ). Samples were transferred from 384 microtiter assay, Sigma enzyme immunoassay (EIA), and Zeus EIA.
plates to glass slides by use of stainless steel solid pins The panels VP6627, VP6628, and VP6678 consisted of two
(200-m diameter). Each pin was estimated to transfer ⬃1 groups of five sera containing either IgG or IgM directed
nL of sample to the slide. Arrays consisted of a 7 ⫻ 7 against T. gondii, rubella virus, and CMV, respectively.
matrix that included five microbial antigens printed in six Each panel also contained five sera selected for their lack
replicates, the IgG and IgM internal calibration curves in of IgG and IgM reactivity against different ToRCH anti-
duplicate, and rabbit muscle myosin as control. The gens. Panel VP6680 contained five sera with IgM directed
fluorophores Alexa 546 and/or Alexa 594 (50 pg) were against HSV1. Panel VP6679 consisted of five positive sera
with IgG against HSV1 and five negative sera.
All sera were analyzed for reactivity against T. gondii,
3
Nonstandard abbreviations: CMV, cytomegalovirus; HSV, herpes sim- rubella virus, HSV1, HSV2, and CMV antigens by the
plex virus; PBS, phosphate-buffered saline; and EIA, enzyme immunoassay. ELISAs (EIA Well) supplied by Radim S.p.A. The assays
Clinical Chemistry 48, No. 1, 2002 123
Fig. 2. Fluorescent scans of array areas printed with increasing concentrations (top to bottom) of human IgG (A; 2, 10, 25, and 50 mg/L) and IgM
(C; 0.4, 2, 4, and 8 mg/L) and incubated with anti-IgG and -IgM secondary antibodies labeled with Alexa 546 and Alexa 594, respectively.
The fluorescence was visualized in a pseudo-color scale (dark blue to white) corresponding to increasing fluorescence. The mean values of duplicate photomultiplier
count (pmc) measurements were plotted as a function of the concentrations of IgG (B) and IgM (D) to generate dose–response curves. The equation and the coefficient
of determination are indicated for each curve.
were performed according to the manufacturer’s instruc- bated for 5 min with Alexa 546-labeled anti-human IgG
tions. monoclonal antibody 2F6 (Radim) suspended in a solu-
tion containing 2⫻ PBS, 10 g/L bovine serum albumin,
processing of microarray slides and 0.1 mL/L Tween 20. To reveal IgM, the slides were
Printed slides were incubated for 1 h at room temperature incubated for 5 min with Alexa 594-labeled goat immu-
with a solution containing 10 g/L bovine albumin in PBS noglobulins directed against the anti-human IgM chain
to block nonspecific antibody binding. An adhesive tape (OEM Concepts Inc.). Before the fluorescence was read in
(Gene-Frame; Abgene Limited) was used to contain sam- the scanner, the slides were washed and dried at 37 °C.
ples/reagents within the array area. Serum samples were The secondary antibodies, monoclonal antibody 2F6 and
diluted 1:200 in 2⫻ PBS containing 10 g/L bovine serum the goat immunoglobulins, were labeled with succinimi-
albumin and 0.1 mL/L Tween 20 and allowed to react dyl Alexa 546 and 594, respectively, and purified accord-
with the array for 15 min at room temperature in a humid ing to the instructions provided by the manufacturer
chamber. To reveal IgG bound to the printed antigens, the (Molecular Probes Inc.). The efficiency of the labeling
slides were washed five times (1 mL each time) with PBS procedure was analyzed by measuring the molar ratio of
containing 0.1 mL/L Tween 20 and subsequently incu- dye to protein.
124 Mezzasoma et al.: Microarrays for Serodiagnosis
Fig. 3. Fluorescent scans of antigen arrays incubated with samples from VP6678/1 (A), VP6678/15 (B), VP6628/2 (C), and VP6627/8 (D) serum
panels, which differ according to their reactivities in ELISA in the presence of IgG directed against the ToRCH antigens.
Serum antibodies were detected by incubating the slides with Alexa 546-labeled secondary antibody directed against human IgG. The intensity of fluorescence emitted
by the fluorophore was visualized in a pseudo-color scale (dark blue to white).
data collection and analysis printed on amino-silane-activated glass slides. This sur-
The slides were analyzed with a GSI 5000 scanner. Images face was chosen on the basis of experiments indicating
were generated with the ScanArrayTM software provided that proteins, such as antibodies and albumin, bound with
by GSI Lumonics and quantified using the QuantArrayTM high efficiency and reproducibility to amino-silane-acti-
software provided by the same company (23 ). Human vated glass slides (data not shown). Each antigen was
IgG and IgM dose–response curves were fitted with a printed in six replicates within a 7 ⫻ 7 matrix (Fig. 1). The
linear curve fit (Excel; Microsoft). The amounts of IgG and array was designed to contain a negative control (rabbit
IgM in the sera were determined by interpolating the myosin) and the fluorophores Alexa 546 and Alexa 594 as
photomultiplier counts collected at the microbial antigen sources of reference signal. Internal calibration curves
spots with the IgG and IgM internal calibration curves. were generated by printing, in duplicate, 1 nL of immu-
noglobulin solution containing increasing concentrations
Results of IgG or IgM (2–50 and 0.4 – 8 mg/L, respectively). The
array design and analysis of IgG and IgM reactivities of the secondary antibodies against the nega-
calibration curves tive control (myosin) were used as the first point of the
Purified human IgG and IgM as well as T. gondii, CMV, calibration curves. Regression analysis demonstrated that
rubella virus, HSV1, and HSV2 antigen preparations were linear dose–response curves were obtained by plotting the
Clinical Chemistry 48, No. 1, 2002 125
Fig. 4. Fluorescent scans of antigen arrays incubated with samples from VP6678/1 (A), VP6678/15 (B), VP6628/2 (C), and VP6627/8 (D) serum
panels.
Serum antibodies were detected by incubating the slides with Alexa 594-labeled secondary antibody directed against human IgM. The fluorescence emitted by the
fluorophore was visualized in a pseudo-color scale (dark blue to white).
photomultiplier counts against the amounts of immuno- antigens. These sera were certified for the presence or
globulins printed on the slides (Fig. 2). Both IgG and IgM absence of antibodies directed against distinct ToRCH
calibration curves showed similar slopes and coefficients antigens on the basis of the Abbott IMx assay, the Sigma
of determination (r2) close to 1. These experiments also EIA, or the Zeus EIA. Radim ELISAs for ToRCH antigens
revealed that the cross-reactivities of the anti-IgM goat were used to confirm the reactivities of all sera. Fluores-
serum and monoclonal antibody 2F6 against human IgG cent scans of microarray slides incubated with represen-
and IgM, respectively, were negligible. tative sera of the different panels showed that this assay
allows the detection of specific antibodies directed against
analysis of serum reactivity against arrayed microbial antigens in accordance with ELISA reactivity
antigens data (Figs. 3 and 4). In general, sera that failed to show a
Human sera from panels VP6627, VP6628, VP6678, positive reaction in ELISA did not react against the
VP6680, and VP6679 were used to investigate the ability corresponding microbial antigens in the microarray assay,
of the microarray assay to reveal the presence or absence whereas fluorescent signals were usually associated with
of IgG and IgM directed against the different ToRCH positive reactions in the ELISAs (Figs. 3 and 4). No
126 Mezzasoma et al.: Microarrays for Serodiagnosis
Fig. 5. Mean photomultiplier count (pmc) measurements collected from the arrayed antigens (colored symbols), incubated with samples from
VP6678/1 (A), VP6678/15 (B), VP6628/2 (C), and VP6627/8 (D) serum panels, were interpolated to the dose–response curves generated by
plotting the fluorescent intensity as a function of printed amounts of IgG (blue diamonds).
The SE of replicate measurements was ⬍10%. The equation and the coefficient of determination are indicated for each curve.
reactivity was detected against the myosin negative con- when comparing slides from different batches (Table 1).
trol, thus demonstrating the absence of nonspecific inter- These values were similar to ELISA-generated data, as
ference in the assay. The r2 values for both the IgG and stated in the instruction manual for the ELISA. The
IgM curves in these arrays were 0.97– 0.99, indicating that detection limit of the microarray assay was determined by
the internal calibration curves generated under these assessing the reactivity of the sera against 20 replicates of
experimental conditions showed a high reproducibility rabbit myosin printed on slides. The detection limit,
and a linear dose–response relationship (Figs. 5 and 6). defined as mean photomultiplier counts of the negative
A set of human sera showing different reactivities in control plus 2 SD interpolated on the IgG calibration
the ELISA (low, medium, and high), was used to monitor curve, was 0.5 pg.
the performance of the microarray assay for each ToRCH
antigen. This analysis revealed that increasing concentra- microarray assay for multianalyte
tions of IgG directed against the microbial antigens gen- serodiagnosis
erated fluorescence signals with precision profiles that We compared the microarray assay and Radim ELISA for
showed in most cases a CV ⬍10% within and between their ability to determine in human sera the presence or
slides (Table 1). Slightly higher CV values were observed absence of IgG reactive against the ToRCH antigens. This
Clinical Chemistry 48, No. 1, 2002 127
Fig. 6. Mean photomultiplier count (pmc) measurements collected from the arrayed antigens (colored symbols), incubated with samples from
VP6678/1 (A), VP6678/15 (B), VP6628/2 (C), and VP6627/8 (D) serum panels, were interpolated to the dose–response curves generated by
plotting the fluorescent intensity as a function of printed amounts of IgM (blue diamonds).
The SE of replicate measurements was ⬍10%. The equation and the coefficient of determination are indicated for each curve.
analysis could not be extended to serum IgM because we to these criteria, the microarray assay classified 19.6% of
lacked a sufficient number of characterized positive sera samples as positive, 76.8% as negative, and 3.6% as
for accurate comparison with the ELISA. Notably, the equivocal for the presence of IgG directed against T.
antigen microarray and the Radim ELISA shared the same gondii; corresponding ELISA data were 21.4% positive,
antigen and antibody reagents, facilitating the compari- 69.7% negative, and 8.9% equivocal (Table 2). For CMV,
son of the two diagnostic assays. The cutoff values of the the microarray classified 64.3% of samples as positive,
distinct serum reactivity determinations within the mi- 32.1% as negative, and 3.6% as equivocal; corresponding
croarray assay were calculated using reference sera that, ELISA data were 66.1% positive, 32.1% negative, and 1.8%
according to information that accompanied the Abbot equivocal. For rubella virus the microarray classified
IMx, Sigma EIA, and Radim ELISA, lacked antibodies 17.9% of samples as positive, 75% as negative, and 7.1% as
directed against T. gondii, rubella virus, CMV, HSV1, or equivocal; ELISA classified 87.5% as positive, 1.8% as
HSV2. These values incorporated the 95th percentile of negative, and 10.7% as equivocal (Table 2).
these reference sera. The reactivities of the serum samples For HSV1, the microarray classified 80.3% of samples
were classified as either positive or negative according to as positive, 17.9% as negative, and 1.8% as equivocal;
the presence of IgG concentrations above or below the corresponding ELISA data were 78.6% positive, 19.6%
cutoff value. Sera with IgG concentrations between ⫾ 10% negative, and 1.8% equivocal. For HSV2, the microarray
of the cutoff value were regarded as equivocal. According classified 22.2% of samples as positive, 66.7% as negative,
128 Mezzasoma et al.: Microarrays for Serodiagnosis
Table 1. Within-slide, between-slide, and between-batch precision profiles of the microarray assay.
IgG serum reactivity
Low Medium High Low Medium High Low Medium High Low Medium High Low Medium High
Within slide (n ⫽ 6)
pmca 30.0 61.0 111.5 8.1 27.5 73.2 16.7 30.0 50.7 6.3 77.0 124.5 4.9 14.8 49.3
SD 0.87 4.50 1.90 0.51 0.48 6.33 0.60 1.69 2.93 0.30 3.72 5.89 0.19 0.57 1.19
CV,b % 2.9 7.4 1.7 6.4 1.7 8.6 3.6 5.6 5.8 4.8 4.8 4.7 3.9 3.9 2.4
Between slides (n ⫽ 12)
pmc 28.3 66.3 101.7 7.5 25.6 76.8 16.8 26.7 52.4 6.0 79.6 128.7 4.6 14.5 51.8
SD 2.07 7.32 12.40 0.70 2.54 9.57 0.45 3.89 2.99 0.39 6.44 7.09 0.48 0.63 2.81
CV,% 7.3 11 12 9.3 9.9 12 2.6 15 5.7 6.5 8.1 5.5 10 4.4 5.4
Between batches (n ⫽ 18)
pmc 27.8 66.3 99.7 7.8 26.1 76.0 15.8 27.4 50.9 6.0 76.2 130.8 4.3 14.6 53.1
SD 2.44 7.39 10.48 0.76 2.72 9.38 1.74 5.06 3.50 0.31 7.70 11.31 0.64 0.61 3.04
CV,% 8.8 11 10 9.7 10 12 11 18 6.9 5.2 10 8.6 15 4.2 5.7
a
pmc, photomultiplier count.
b
CVs were determined for photomultiplier counts collected from the arrayed antigens, each incubated with three serum samples classified on the basis of their
reactivities in ELISA as negative, equivocal, or positive. The CVs were calculated as the SD of the photomultiplier counts divided by the mean.
and 11.1% as equivocal; corresponding ELISA data were wells. In these assays, matrix problems (differences be-
18.5% positive, 51.9% negative, and 29.6% equivocal (Ta- tween the calibrator and sample matrices) occur fre-
ble 2). quently and represent a known source of bias (24 ). In
contrast, in our microarray format any given signal de-
comparison of test protocols veloped within the arrayed antigens is interpolated to a
Comparison of microarray assay and ELISA test protocols calibration curve processed under the same physicochem-
(Table 3) indicated that the major advantage of the mi- ical conditions (e.g., time, temperature, protein content,
croarray test is that regardless of the number of analytes fat content, and bilirubin concentration) as the antibody
determined on a slide, the time-to-result and costs were reactivity being determined. Under these conditions, the
the same. With ELISA formats, each analyte is determined dose–response curve will compensate for variability in the
in a separate assay, thereby increasing time and costs. sample physicochemical conditions, thus minimizing ma-
trix effects.
Discussion The use of a dedicated microarray robot coupled with
Our data indicate that a protein microarray assay with optimized printing solutions and glass substrates allowed
indirect fluorescence detection can be used to determine us to generate high-quality fluorescence data suitable for
the presence or absence of specific antibodies directed a study aimed at assessing the clinical performance of the
against various microbial antigens in human sera. The ToRCH microarray assay. For this purpose the microarray
presence of internal dose–response IgG and IgM calibra- assay and several commercially available ELISAs were
tion curves allowed us to demonstrate that antibodies compared for their ability to reveal IgG directed against
from different serum samples bound to printed antigens the selected microbial antigens in a set of characterized
can be quantified with high reproducibility. In addition, human sera. Notably, both the microarray assay and the
the calibration curves provided unique advantages over ELISAs used in the comparative study shared the same
current immunoassay protocols in which the calibration antigen preparations and antibody reagents. This analysis
curves and the samples are processed in separate tubes or revealed that the microarray assay is able to identify
Table 2. Comparison of microarray and ELISA for their ability to detect IgG directed against ToRCH antigens in human sera.
Serum reactivity, n (%)
20. Wiese R, Belosludtsev Y, Powdrill T, Thompson P, Hogan M. 22. Schena M, Shalon D, Davis RW, Brown PO. Quantitative monitor-
Simultaneous multianalyte ELISA performed on a microarray plat- ing of gene expression patterns with a complementary DNA
form. Clin Chem 2001;47:1451–7. microarray. Science 1995;270:467–70.
21. Wiltshire S, O’Malley S, Lambert J, Kukanskis K, Edgar D, 23. ScanArrayTM and QuantArrayTM operating manuals, Ver. 2.0. Wa-
Kingsmore SF, Schweitzer B. Detection of multiple allergen- tertown, MA: GSI Lumonics, 2000.
specific IgEs on microarrays by immunoassay with rolling circle 24. Wood W. “Matrix effects” in immunoassays. Scand J Clin Lab
amplification. Clin Chem 2000;46:1990 –3. Invest 1991;51:105–12.