- Email: [email protected]
Th2 cytokines using microarrays
Journal of Immunological Methods 261 (2002) 157 – 165 www.elsevier.com/locate/jim
Simultaneous analysis of eight human Th1/Th2 cytokines using microarrays Sun W. Tam 1,2, Rick Wiese 2,3, Sandy Lee, James Gilmore, Krishnanand D. Kumble* Genometrix, Inc., 2700 Research Forest Drive, The Woodlands, TX 77381, USA Received 25 September 2001; received in revised form 12 December 2001; accepted 21 December 2001
Abstract The adaptive immune system induces T cells to change from a naive phenotype to a Th1/Th2 phenotype each of which produce characteristic types of cytokines. Knowledge of whether a specific immune response is Th1 or Th2 is a useful indicator for diseases with basis in immune function disorder. An assay that can rapidly analyze multiple cytokines indicative of these two cell types from small sample quantities can be an extremely useful research and diagnostic tool. Silanized glass slides were printed with multiple arrays of capture antibodies to detect eight different cytokines involved in the Th1/Th2 response along with control proteins for assessing assay performance. Arrays were developed by sequential addition of known antigen amounts, detector antibodies and a fluorescent detection system followed by imaging and quantification. These arrays were used to determine the specificity, sensitivity and reproducibility of the assay and the performance compared with conventional ELISA. This multiplexed assay is able to measure human Th1/Th2 cytokines in sample volumes lower than 20 ml. The assay sensitivity for the eight cytokines range from 0.3 mg/l for IL-4 to 6.4 mg/l for IL-5 which are either comparable to or higher than those reported for conventional ELISA or bead-based multiplex ELISA methods. This assay can be automated to measure expression levels of multiple Th1/Th2 cytokines simultaneously from tens to hundreds of biological samples. This assay platform is more sensitive and has a larger dynamic range as compared to a conventional ELISA in addition to significantly reducing the time and cost of assay. This platform provides a versatile system to rapidly quantify a wide variety of proteins in a multiplex format. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Microarray; Multiplex ELISA; Protein array
Abbreviations: microELISA, microarray format ELISA; IL, interleukin; IFNg, interferon gamma; TNFa, tumor necrosis factor alpha; PBS, phosphate-buffered saline; PBS-T, phosphate-buffered saline containing 0.05% Tween-20; Th, T helper; CV, coefficient of variation. * Corresponding author. Pointilliste, Inc., 2541 Leghorn Avenue, Suite 4, Mountain View, CA 94043, USA. Tel.: +1-650-938-1995; fax: +1-650-938-2624. E-mail address: [email protected] (K.D. Kumble). 1 Present address: Texas Biotechnology Corporation, 7000 Fannin Street, Houston, TX 77030, USA. 2 These authors contributed equally to this work. 3 Present address: Pierce Endogen, Pierce Chemical, 3747 Meridian Road, PO Box 117, Rockford, IL 61105, USA.
1. Introduction Microarrays have been widely used for genotyping and gene expression analysis (Collins, 1999; Lander, 1999; Southern et al., 1999). More recently, proteins were utilized in the same format to detect protein – protein, enzyme – substrate and drug – protein interactions (Mendoza et al., 1999; de Wildt et al., 2000; MacBeath and Schreiber, 2000; Haab et al., 2001; Weise et al., 2001). Here, we report a microarray-based method that utilizes the traditional sandwich ‘‘Enzyme-
0022-1759/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 1 7 5 9 ( 0 1 ) 0 0 5 7 2 - 5
158
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
Linked ImmunoSorbent Assay’’ (ELISA) format to quantify proteins and perform expression profiling. Normally, a sandwich ELISA involves the attachment of a capture antibody to a solid surface, followed by addition of samples containing known or unknown amount of antigen. Subsequently, an enzyme-labeled antibody is introduced for detection, resulting in the formation of a sandwich. Finally, a substrate is added to produce a quantifiable signal via the enzyme-labeled antibody (Douglas and Monteith, 1994; Voller, 1978; Reen, 1994). By comparing the signal intensities from known amounts of antigen, one can derive a standard curve and quantify the unknown amounts of antigen in the samples. In this study, a protein microarray for the examination of Th1/Th2 type cytokines was developed and validated. Traditionally, human T lymphocytes secreting Interferon-gamma (IFNg) are classified as Th1 cells, T lymphocytes producing IL-4 and IL-5 are termed Th2 cells (Cherwinski et al., 1987; Del Prete et al., 1991). Further studies have established that Th1 cells also produce IL-2, while Th2 cells produce IL-6 and IL-13 (Seder and Paul, 1994). IL-10 can potentially be produced by either cell type (Zhai et al., 1999). Naive T cells, also known as Thp cells, can produce cytokines such as IL-2, IL-4, TNFa, IL-13 and IFNg upon activation (Ohshima et al., 1999). Another cell type, Th0, has been described as producing both IL-4 and IFNg (Seder and Paul, 1994). To distinguish these cell types that reflect the status of the immune system for naive, effector and memory T cells, an accurate measurement of all the cytokines are necessary. This cytokine array offers a unique format to simultaneously quantify all the critical Th1/Th2 cytokines-IFNg, IL-2, IL-4, IL-5, IL-6, IL-10, IL-13 and TNFa. The information obtained through the use of this single assay will help understand the interaction and cell surface signaling between T lymphocytes and other immune cell types. Important physiological questions can be addressed by measuring the cytokine levels from both cell culture supernatant or human serum. In this report, we have characterized the specificity, sensitivity and reproducibility of a multiplexed microELISA assay and found it to be comparable with conventional ELISA and other multiplex flowbased quantification methods (Carson and Vignali, 1999).
Fig. 1. Raw image data from microarray ELISA. (A) Array layout. The eight human cytokine capture antibodies include interferon gamma (IFN), interleukin 2 (IL-2), interleukin 4 (IL-4), interleukin 5 (IL-5), interleukin 6 (IL-6), interleukin 10 (IL-10), interleukin 13 (IL-13) and tumor necrosis factor alpha (TNF). The numerals depict print concentration of the antibodies (mg/ml). M1 denotes biotin-Fc. M2 is Cy3 labeled Fc fragment. The anti-hamster antibody in two capture concentrations is indicated as C1 and C2. (B) Representative example of a microarray ELISA. A cocktail of eight cytokines was applied at 500 pg/ml to the array, followed by addition of corresponding detection antibody mixture. C1 and C2 served as negative controls.
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
2. Materials and methods 2.1. Glass plate preparation and array manufacturing Glass slides containing 16 Teflon delineated wells (Erie Scientific) were cleaned ultrasonically with methanol followed by multiple rinses in distilled water. Oven-dried slides were then silanized by vapor deposition with silane in xylene using a modification of the method described by Falipou et al. (1999). Two different concentrations of each capture antibody diluted in 0.1 mol/l carbonate buffer, pH 9.5 containing 50 ml/l glycerol were printed on the silanized slides using a robotic arrayer (BioChip Arrayer, Packard Instruments). The arrayer was programmed to deposit 2 nl volumes and a center to center distance of 0.4 mm resulting in spot diameters of 0.3 mm. Fig. 1A shows an array map for each of the antibodies spotted including control proteins. Reagents including capture antibodies, biotinylated antibodies and recombinant cytokines
159
were purchased either from BD Pharmingen or BioSource. The biotinylated-Fc protein and Cy3-labeled Fc protein were purchased from Jackson Immunochemical and printed at a concentration of 7 mg/l. These Fc fragments are useful for the orientation and identification of probes within the array. 2.2. MicroELISA assay After overnight storage at 4 C (post-printing), each array was rinsed three times with phosphatebuffered saline containing 0.5 ml/l Tween-20 (PBST), then blocked in Blocker Casein (Pierce Chemical, Rockford, IL) for 1 h at room temperature. Arrays were rinsed three times as before and incubated for 2 h at 37 C in a humidity chamber after application of 20 ml antigen solution in PBS containing 100 ml/l fetal bovine serum (PBS – FBS). Biotinylated detector antibodies were applied to arrays at appropriate concentrations and incubated for 1 h at 37 C in a humidity
Fig. 2. Quantitative representation of cytokine specificity in multiplex micro-ELISA. Two concentrations of the capture antibodies for each cytokine were printed in quadruplicate per well. A single cytokine was applied at 500 pg/ml to each well. Subsequently, a cocktail of the eight detection antibodies was introduced to each well. The array was imaged and quantified after the hybridization for specificity to the antigen.
160
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
chamber, rinsed and streptavidin-R-phycoerythrin in PBS – FBS added to the array. The slides were incubated in the dark for 1 h, followed by successive washes with PBS-T and PBS. After drying, slides were imaged in a ScanArray scanner (Packard Instruments) using the 532 nm laser. Densitometry values for each array element were obtained using OptiQuant
software (Packard Instruments). The background, calculated as the median of pixel intensities from the local area around each spot, was subtracted from the average pixel intensity within each spot to determine signal values for each analyte spot. Spot densitometry values were used to construct calibration curves of densitometry versus antigen concentration for each
Fig. 3. Standard curve comparison between singleplex ( – —), multiplex micro-ELISA ( – ) and standard ELISA (—) for each of the eight cytokines. The averaged correlation coefficients for single-plexed array is 0.980 F 0.014, for multiplexed is 0.944 F 0.050 and for ELISA is 0.975 F 0.021. The graphs represent data obtained for the following antigens (A) IFNg (B) IL-2 (C) IL-4 (D) IL-5 (E) IL-6 (F) IL-10 (G) IL-13 and (H) TNFa. Standard ELISAs were performed according to manufacturer’s instructions. Micro-ELISA was done as described in Materials and methods.
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
161
Fig. 3 (continued).
cytokine, and the cytokine concentration of the unknowns were determined through these curves. 2.3. Standard ELISA ELISA was performed as outlined by the manufacturer (BD PharMingen or BioSource). Briefly, flatbottomed 96-well plates (VWR) were coated with capture antibody (50 ml per well in PBS) overnight at 4 C. Plates were washed three times with PBS-T.
Wells were blocked with 200 ml of PBS –FBS for 1 h at room temperature. Standards and samples were titrated accordingly and added into the rinsed plate (50 ml per well) and incubated for 2 h at room temperature. Plates were washed five times before addition of biotinylated detection antibody (50 ml per well in PBS – FBS) and incubated for 1 h at room temperature. Finally, the plates were washed seven times before introduction of ABTS substrate in citrate phosphate buffer with hydrogen peroxide. Fully de-
162
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
Table 1 Sensitivity of micro-ELISA for each cytokine Cytokine
Sensitivity of microELISA (pg/ml)
Sensitivity of bead-based assay (pg/ml)
IFNg IL-2 IL-4 IL-5 IL-6 IL-10 IL-13 TNFa
0.55 0.58 0.27 6.40 2.80 1.07 3.87 3.07
15.6 6.6 6.5 2.8 NA 4.7 NA 4.3
NA = data not available.
veloped plates were inactivated with 0.1 M H2SO4 and read at 405 nm with a reference at 470 nm, on a microtiter plate reader (SpectraMax).
3. Results 3.1. Assay development In this assay, capture antibodies against IFNg, IL-2, IL-4, IL-5, IL-6, IL-10, IL-13 and TNFa were initially printed onto a silanized glass surface (array map in Fig. 1A). The capture antibodies were printed at two different concentrations to avoid saturation of sample signals. This provides a unique advantage over conventional ELISA that often requires repeating the assay for concentrated samples. The array also contained control elements to monitor assay performance. They include biotin-Fc, Cy3-Fc, and anti-hamster IgG. Biotin-Fc measures the performance of the detection system The Cy3-Fc printing control allows for the visualization and quantification of protein surface binding without any reagent hybridization. It may also be used to normalize signals across different arrays (Fig. 1B). Finally, addition of hamster IgG during the antigen incubation step assesses the success of the assay and is an indicator for the integrity of other assay reagents (capture antibody status, detector molecules). Additionally, the hamster IgG could also be used as a normalization marker to account for well to well and slide to slide variations. Unlike conventional ELISA that has no reagent controls, the control elements in this assay help detect reagent problems during different steps of the assay and assure proper quality control.
In comparison to conventional ELISA, this procedure requires no overnight pre-incubation of capture antibody on wells leading to considerable savings in time. 3.2. Assay specificity A fully multiplexed assay for cytokine measurement involves mixing all the cytokine standards and detection antibodies. Although this approach is necessary to analyze multiple cytokines simultaneously, any cross-reactivity or interference between reagents and samples would result in assay failure. Each of the capture and detection antibodies has been prescreened for specific reactivity with their respective antigens. Each cytokine antibody pair demonstrates the anticipated specificity to its antigen when added individually (data not shown) or in a multiple detector antibody cocktail (Fig. 2). No cross-reactivity is observed among the cytokines, even at concentrations as high as 500 pg/ml. In Fig. 2, the intensity of the signal from each cytokine varies, reflecting the unique affinity of the capture and detection antibodies to their respective antigen. 3.3. Assay sensitivity The sensitivity of the assay was evaluated, based on the ‘‘minimum detectable concentration’’ or MDC. MDC is defined as the concentration extrapolated from the standard curve of the zero standard of an ELISA plus two standard deviations. This is a theoretical limit of assay sensitivity (Moran and Brown, 1997). The average of eight individual values for each cytokine were determined from arrays incubated with known amounts of cytokine ranging from 10 to 1000 pg/ml
Table 2 Reproducibility of data across slides Cytokine
Average DLU
SD
CV
IFNg IL-2 IL-4 IL-5 IL-6 IL-10 IL-13 TNFa
789,900 774,886 1,738,500 116,418 971,367 498,570 326,752 140,720
107,384 104,330 232,371 13,648 36,527 40,430 30,306 5630
0.14 0.13 0.13 0.12 0.04 0.08 0.09 0.04
SD is standard deviation and CV is the coefficient of variation.
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
163
Table 3 Linearity of cytokines from cell culture mixture Cytokine
Spike concentration (pg/ml)
Observed concentration (pg/ml)
Actual dilution
Observed dilution
IFNg
900 450 112 56 0
1,022.41 538.95 138.17 48.58 0.08 Slope = 0.93 794.06 436.46 132.99 43.04 0.21 Slope = 0.95 762.45 530.93 221.86 38.35 11.51 Slope = 0.99 1098.92 670.85 182.32 52.51 9.12 Slope = 0.96 1078.81 513.76 164.17 45.76 3.83 Slope = 0.93 964.18 565.15 183.15 51.3 0.7 Slope = 0.96 851.14 521.98 168.37 57.51 3.92 Slope = 0.96 1368.56 455.04 127.8 65.12 41.45 Slope = 0.83
Undiluted 1:2 1:8 1:16
1: 1.9 1: 7.4 1: 21
Undiluted 1:2 1:8 1:16
1: 1.8 1: 6 1: 18.5
Undiluted 1:2 1:8 1:16
1: 1.4 1: 3.4 1: 19.8
Undiluted 1:2 1:8 1:16
1: 1.6 1: 6 1: 21
Undiluted 1:2 1:8 1:16
1: 2.1 1: 6.6 1: 23.6
Undiluted 1:2 1:8 1:16
1: 1.7 1: 5.3 1: 18.8
Undiluted 1:2 1:8 1:16
1: 1.6 1: 5 1: 14.8
Undiluted 1:2 1:8 1:16
1: 3 1: 10.7 1: 21
IL-2
IL-4
IL-5
IL-6
IL-10
IL-13
TNFa
900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0
along with a negative control (0 pg/ml of antigen). Average signal values expressed as ‘‘Digital Light Units’’ or DLU were used to plot standard curves
(Fig. 3A to H). Subsequently, the respective MDC was extrapolated from an individual standard curve of the selected cytokine and expressed in pg/ml (Table 1).
164
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
Overall, the sensitivity of the cytokine ranges from 0.55 to 6.4 pg/ml in our assay system. It compares favorably with reported values for a multiplex flow cytometry-based assay (Human Th1/Th2 cytokine CBA kit, PharMingen). Reproducibility across wells and slides was evaluated by performing a multiplex microELISA experiment in which two wells from each of three slides were assayed with 200 pg/ml of the antigen mixture. Each well had four spots per antibody and the results shown in Table 2 are expressed as the mean of 24 data points. The sum of values from three different slides was calculated to determine the extent of slide variations. Briefly, well to well variations range from 5% to 18%. The variations among slides range from 4% to 14%. 3.4. Comparison of microELISA and conventional ELISA for cytokine quantification Known amounts of each of the eight cytokines were assayed using conventional ELISA and microarray ELISA to compare their relative sensitivity and dynamic range. Fig. 3A to H represents standard curves plotted from data obtained from these assays. The correlation coefficient of the curves were greater than 0.9, except for IL-13. First, the ELISA was compared to individual microELISA (singleplex) where a single cytokine was analyzed in the absence of other cytokines but in the presence of all the detection antibodies (Fig. 3). For each of the cytokines on the array, greater dynamic range and similar linearity were achieved with the microELISA as compared to conventional ELISA. The dynamic range for microELISA ranges from 2.5 to 3 logs for each of the antigens, while the conventional ELISA has only 1.5 to 2 log dynamic range. The multiplexed format, in which all eight cytokines and biotinylated secondary antibodies were added in the same well also showed greater dynamic range and comparable linearity as the standard ELISA (Fig. 3). 3.5. Assay validation The ability of the multiplexed microELISA assay to simultaneously quantify recombinant cytokines in cell culture media was tested. Known amounts of cytokines were spiked into tissue culture media (RPMI plus 100 ml/l fetal bovine serum) to determine the linearity
and correlation of extrapolated values. Four different concentrations were spiked—900, 450, 112 and 56 pg/ml. For each of the cytokines, the observed dilution is close to the expected dilution; namely 1:2, 1:8 and 1:16 (Table 3). The slopes of the regression line from the observed concentration for most cytokines are greater than 0.9.
4. Discussion We have described a new multiplexed ELISA assay for cytokine quantification using microarrays. A major advantage of the assay is the amount of information that can be obtained from sample volumes under 20 ml while a conventional ELISA requires up to 100 ml per well. This translates into a 40-fold savings in sample required to measure eight cytokines. Substantial savings in time is achieved in processing samples since eight cytokines can be monitored simultaneously. This assay would be beneficial to measure protein levels in samples of limited availability such as needle biopsies, providing a useful tool for exploring important biological questions. Besides cost and time, specificity, sensitivity and reproducibility are other factors that are important for a quantitative assay. Our data clearly showed specificity in both singleplex and multiplex format. No cross reactivity of the detector antibodies was observed, this is likely due to the added benefit of using monoclonal antibodies during this step in the multiplex microarray design. The assay sensitivity, based on minimum detectable concentration, is at least comparable to commercially available bead-based multiplex assay systems. Moreover, the microarray assay outperforms the beadbased assay by roughly 30-fold greater sensitivity for some of the cytokines. Although an average of 9.6% CV for slide reproducibility is higher than those reported for some multiplexed ELISAs, it does not affect the biological validation results significantly (Huang et al., 2001). For each of the cytokines tested, intra-slide CV decreased with increasing concentration of antigen up to 50 pg/ml after which the CV remained at around 10%. Another advantage of the microarray platform described here is the ability to analyze tens to hundreds of samples simultaneously. Each of the 16 wells in the glass slide has similar printed arrays. This format can also been extended to a 96-well glass plate,
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
allowing for 96 samples to be analyzed on a single plate. This platform can be easily automated with commercially available fluid handling robotics. An important aspect that impacts data analysis is the metric used to quantify spot signal intensity. Typical microarray spots are likely to have variable morphology (size, shape) dependent on the type of printing mechanism (CV for printing with the BioChip Arrayer was less than 10%). The spots also have microheterogeneity that reflect surface irregularities impacting intra-spot pixel intensities. The commonly used metric for measuring spot intensities ‘‘sum of pixels’’ can result in incorrect differences in signal for the same concentration of antigen dependent on the characteristics of the printed antibody spot. Measuring the average pixel intensity of the spot (as described under Materials and methods) resulted in data that provided better quantification of the antigens being measured. Unlike conventional ELISA, which requires a standard curve to be performed with each 96-well plate, for a given print batch of microarrays a standard curve need be plotted only once and used for quantification of data from samples analyzed on all the other arrays from that batch. In addition, the incorporated controls in each array allow for data to be normalized between different samples thereby accounting for any variations that may occur from one array to another. Good correlation in cytokine levels measured using conventional ELISA and the microELISA was observed in Jurkat cells stimulated with PHA and PMA (Huang et al., 2001). Our current assay system addresses some of the challenges of conventional ELISA by the ability to simultaneously measure multiple antigens in shorter experimental time. One can extend the platform to examine immunoglobulins, chemokines, growth factors or soluble receptors. This assay can provide an excellent tool for research or clinical diagnostic as well as a valuable platform in high throughput drug screening for identification of lead drug targets and candidates. References Carson, R.T., Vignali, D.A., 1999. Simultaneous quantitation of 15 cytokines using a multiplexed flow cytometric assay. J. Immunol. Methods 227, 41. Cherwinski, H.M., Schumacher, J.H., Brown, K.D., Mosmann, T.R., 1987. Two types of mouse helper T cell clone. III. Further differ-
165
ences in lymphokine synthesis between Th1 and Th2 clones revealed by RNA hybridization, functionally monospecific bioassays, and monoclonal antibodies. J. Exp. Med. 166, 1229. Collins, F.S., 1999. Microarrays and macroconsequences. Nat. Genet. 21, 2 [Editorial]. de Wildt, R.M., Mundy, C.R., Gorick, B.D., Tomlinson, I.M., 2000. Antibody arrays for high-throughput screening of antibody – antigen interactions. Nat. Biotechnol. 18, 932. Del Prete, G.F., De Carli, M., Mastromauro, C., Biagiotti, R., Macchia, D., Falagiani, P., Ricci, M., Romagnani, S., 1991. Purified protein derivative of Mycobacterium tuberculosis and excretory – secretory antigen(s) of Toxocara canis expand in vitro human T cells with stable and opposite (type 1 T helper or type 2 T helper) profile of cytokine production. J. Clin. Invest. 88, 346. Douglas, A.S., Monteith, C.A., 1994. Improvements to immunoassays by use of covalent binding assay plates. Clin. Chem. 40, 1833. Falipou, S., Chovelon, J.-M., Martelet, C., Margonari, J., Cathignol, D., 1999. New use of cyanosilane coupling agent for direct binding of antibodies to silica supports. Physicochemical characterization of molecularly bioengineered layers. Bioconjugate Chem. 10, 346. Haab, B.B., Dunham, M.J., Brown, P.O., 2001. Protein microarrays for highly parallel detection and quantitation of specific proteins and antibodies in complex solutions. Genome Biol. 2, 4.1, RESEARCH0004. Huang, J.X., Mehrens, D., Wiese, R., Lee, S., Tam, S.W., Daniel, S., Gilmore, J., Shi, M., Lashkari, D., 2001. High-throughput genomic and proteomic analysis using microarray technology. Clin. Chem. 47, 1912. Lander, E.S., 1999. Array of hope. Nat. Genet. 21, 3 [Editorial]. MacBeath, G., Schreiber, S.L., 2000. Printing proteins as microarrays for high-throughput function determination. Science 289, 1760. Mendoza, L.G., McQuary, P., Mongan, A., Gangadharan, R., Brignac, S., Eggers, M., 1999. High-throughput microarray-based enzyme-linked immunosorbent assay (ELISA). BioTechniques 27, 778. Moran, R.F., Brown, E.N., 1997. The smallest concentration. Clin. Chem. 43, 856. Ohshima, Y., Yang, L.P., Avice, M.N., Kurimoto, M., Nakajima, T., Sergerie, M., Demeure, C.E., Sarfati, M., Delespesse, G., 1999. Naive human CD4+ T cells are a major source of lymphotoxin alpha. J. Immunol. 162, 3790. Reen, D.J., 1994. Enzyme-linked immunosorbent assay (ELISA). Methods Mol. Biol. 32, 461. Seder, R.A., Paul, W.E., 1994. Acquisition of lymphokine-producing phenotype by CD4+ T cells. Annu. Rev. Immunol. 12, 635. Southern, E., Mir, K., Shchepinov, M., 1999. Molecular interactions on microarrays. Nat. Genet. 21, 5. Voller, A., 1978. The enzyme-linked immunosorbent assay (ELISA) (theory, technique and applications). Ric. Clin. Lab. 8, 289. Wiese, R., Belosludtsev, Y., Powdrill, T., Thompson, P., Hogan, M., 2001. Simultaneous multi-analyte ELISA performed on a microarray platform. Clin. Chem. 47, 1451. Zhai, Y., Ghobrial, R.M., Busuttil, R.W., Kupiec-Weglinski, J.W., 1999. Th1 and Th2 cytokines in organ transplantation: paradigm lost? Crit. Rev. Immunol. 19, 155.
Simultaneous analysis of eight human Th1/Th2 cytokines using microarrays Sun W. Tam 1,2, Rick Wiese 2,3, Sandy Lee, James Gilmore, Krishnanand D. Kumble* Genometrix, Inc., 2700 Research Forest Drive, The Woodlands, TX 77381, USA Received 25 September 2001; received in revised form 12 December 2001; accepted 21 December 2001
Abstract The adaptive immune system induces T cells to change from a naive phenotype to a Th1/Th2 phenotype each of which produce characteristic types of cytokines. Knowledge of whether a specific immune response is Th1 or Th2 is a useful indicator for diseases with basis in immune function disorder. An assay that can rapidly analyze multiple cytokines indicative of these two cell types from small sample quantities can be an extremely useful research and diagnostic tool. Silanized glass slides were printed with multiple arrays of capture antibodies to detect eight different cytokines involved in the Th1/Th2 response along with control proteins for assessing assay performance. Arrays were developed by sequential addition of known antigen amounts, detector antibodies and a fluorescent detection system followed by imaging and quantification. These arrays were used to determine the specificity, sensitivity and reproducibility of the assay and the performance compared with conventional ELISA. This multiplexed assay is able to measure human Th1/Th2 cytokines in sample volumes lower than 20 ml. The assay sensitivity for the eight cytokines range from 0.3 mg/l for IL-4 to 6.4 mg/l for IL-5 which are either comparable to or higher than those reported for conventional ELISA or bead-based multiplex ELISA methods. This assay can be automated to measure expression levels of multiple Th1/Th2 cytokines simultaneously from tens to hundreds of biological samples. This assay platform is more sensitive and has a larger dynamic range as compared to a conventional ELISA in addition to significantly reducing the time and cost of assay. This platform provides a versatile system to rapidly quantify a wide variety of proteins in a multiplex format. D 2002 Elsevier Science B.V. All rights reserved. Keywords: Microarray; Multiplex ELISA; Protein array
Abbreviations: microELISA, microarray format ELISA; IL, interleukin; IFNg, interferon gamma; TNFa, tumor necrosis factor alpha; PBS, phosphate-buffered saline; PBS-T, phosphate-buffered saline containing 0.05% Tween-20; Th, T helper; CV, coefficient of variation. * Corresponding author. Pointilliste, Inc., 2541 Leghorn Avenue, Suite 4, Mountain View, CA 94043, USA. Tel.: +1-650-938-1995; fax: +1-650-938-2624. E-mail address: [email protected] (K.D. Kumble). 1 Present address: Texas Biotechnology Corporation, 7000 Fannin Street, Houston, TX 77030, USA. 2 These authors contributed equally to this work. 3 Present address: Pierce Endogen, Pierce Chemical, 3747 Meridian Road, PO Box 117, Rockford, IL 61105, USA.
1. Introduction Microarrays have been widely used for genotyping and gene expression analysis (Collins, 1999; Lander, 1999; Southern et al., 1999). More recently, proteins were utilized in the same format to detect protein – protein, enzyme – substrate and drug – protein interactions (Mendoza et al., 1999; de Wildt et al., 2000; MacBeath and Schreiber, 2000; Haab et al., 2001; Weise et al., 2001). Here, we report a microarray-based method that utilizes the traditional sandwich ‘‘Enzyme-
0022-1759/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved. PII: S 0 0 2 2 - 1 7 5 9 ( 0 1 ) 0 0 5 7 2 - 5
158
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
Linked ImmunoSorbent Assay’’ (ELISA) format to quantify proteins and perform expression profiling. Normally, a sandwich ELISA involves the attachment of a capture antibody to a solid surface, followed by addition of samples containing known or unknown amount of antigen. Subsequently, an enzyme-labeled antibody is introduced for detection, resulting in the formation of a sandwich. Finally, a substrate is added to produce a quantifiable signal via the enzyme-labeled antibody (Douglas and Monteith, 1994; Voller, 1978; Reen, 1994). By comparing the signal intensities from known amounts of antigen, one can derive a standard curve and quantify the unknown amounts of antigen in the samples. In this study, a protein microarray for the examination of Th1/Th2 type cytokines was developed and validated. Traditionally, human T lymphocytes secreting Interferon-gamma (IFNg) are classified as Th1 cells, T lymphocytes producing IL-4 and IL-5 are termed Th2 cells (Cherwinski et al., 1987; Del Prete et al., 1991). Further studies have established that Th1 cells also produce IL-2, while Th2 cells produce IL-6 and IL-13 (Seder and Paul, 1994). IL-10 can potentially be produced by either cell type (Zhai et al., 1999). Naive T cells, also known as Thp cells, can produce cytokines such as IL-2, IL-4, TNFa, IL-13 and IFNg upon activation (Ohshima et al., 1999). Another cell type, Th0, has been described as producing both IL-4 and IFNg (Seder and Paul, 1994). To distinguish these cell types that reflect the status of the immune system for naive, effector and memory T cells, an accurate measurement of all the cytokines are necessary. This cytokine array offers a unique format to simultaneously quantify all the critical Th1/Th2 cytokines-IFNg, IL-2, IL-4, IL-5, IL-6, IL-10, IL-13 and TNFa. The information obtained through the use of this single assay will help understand the interaction and cell surface signaling between T lymphocytes and other immune cell types. Important physiological questions can be addressed by measuring the cytokine levels from both cell culture supernatant or human serum. In this report, we have characterized the specificity, sensitivity and reproducibility of a multiplexed microELISA assay and found it to be comparable with conventional ELISA and other multiplex flowbased quantification methods (Carson and Vignali, 1999).
Fig. 1. Raw image data from microarray ELISA. (A) Array layout. The eight human cytokine capture antibodies include interferon gamma (IFN), interleukin 2 (IL-2), interleukin 4 (IL-4), interleukin 5 (IL-5), interleukin 6 (IL-6), interleukin 10 (IL-10), interleukin 13 (IL-13) and tumor necrosis factor alpha (TNF). The numerals depict print concentration of the antibodies (mg/ml). M1 denotes biotin-Fc. M2 is Cy3 labeled Fc fragment. The anti-hamster antibody in two capture concentrations is indicated as C1 and C2. (B) Representative example of a microarray ELISA. A cocktail of eight cytokines was applied at 500 pg/ml to the array, followed by addition of corresponding detection antibody mixture. C1 and C2 served as negative controls.
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
2. Materials and methods 2.1. Glass plate preparation and array manufacturing Glass slides containing 16 Teflon delineated wells (Erie Scientific) were cleaned ultrasonically with methanol followed by multiple rinses in distilled water. Oven-dried slides were then silanized by vapor deposition with silane in xylene using a modification of the method described by Falipou et al. (1999). Two different concentrations of each capture antibody diluted in 0.1 mol/l carbonate buffer, pH 9.5 containing 50 ml/l glycerol were printed on the silanized slides using a robotic arrayer (BioChip Arrayer, Packard Instruments). The arrayer was programmed to deposit 2 nl volumes and a center to center distance of 0.4 mm resulting in spot diameters of 0.3 mm. Fig. 1A shows an array map for each of the antibodies spotted including control proteins. Reagents including capture antibodies, biotinylated antibodies and recombinant cytokines
159
were purchased either from BD Pharmingen or BioSource. The biotinylated-Fc protein and Cy3-labeled Fc protein were purchased from Jackson Immunochemical and printed at a concentration of 7 mg/l. These Fc fragments are useful for the orientation and identification of probes within the array. 2.2. MicroELISA assay After overnight storage at 4 C (post-printing), each array was rinsed three times with phosphatebuffered saline containing 0.5 ml/l Tween-20 (PBST), then blocked in Blocker Casein (Pierce Chemical, Rockford, IL) for 1 h at room temperature. Arrays were rinsed three times as before and incubated for 2 h at 37 C in a humidity chamber after application of 20 ml antigen solution in PBS containing 100 ml/l fetal bovine serum (PBS – FBS). Biotinylated detector antibodies were applied to arrays at appropriate concentrations and incubated for 1 h at 37 C in a humidity
Fig. 2. Quantitative representation of cytokine specificity in multiplex micro-ELISA. Two concentrations of the capture antibodies for each cytokine were printed in quadruplicate per well. A single cytokine was applied at 500 pg/ml to each well. Subsequently, a cocktail of the eight detection antibodies was introduced to each well. The array was imaged and quantified after the hybridization for specificity to the antigen.
160
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
chamber, rinsed and streptavidin-R-phycoerythrin in PBS – FBS added to the array. The slides were incubated in the dark for 1 h, followed by successive washes with PBS-T and PBS. After drying, slides were imaged in a ScanArray scanner (Packard Instruments) using the 532 nm laser. Densitometry values for each array element were obtained using OptiQuant
software (Packard Instruments). The background, calculated as the median of pixel intensities from the local area around each spot, was subtracted from the average pixel intensity within each spot to determine signal values for each analyte spot. Spot densitometry values were used to construct calibration curves of densitometry versus antigen concentration for each
Fig. 3. Standard curve comparison between singleplex ( – —), multiplex micro-ELISA ( – ) and standard ELISA (—) for each of the eight cytokines. The averaged correlation coefficients for single-plexed array is 0.980 F 0.014, for multiplexed is 0.944 F 0.050 and for ELISA is 0.975 F 0.021. The graphs represent data obtained for the following antigens (A) IFNg (B) IL-2 (C) IL-4 (D) IL-5 (E) IL-6 (F) IL-10 (G) IL-13 and (H) TNFa. Standard ELISAs were performed according to manufacturer’s instructions. Micro-ELISA was done as described in Materials and methods.
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
161
Fig. 3 (continued).
cytokine, and the cytokine concentration of the unknowns were determined through these curves. 2.3. Standard ELISA ELISA was performed as outlined by the manufacturer (BD PharMingen or BioSource). Briefly, flatbottomed 96-well plates (VWR) were coated with capture antibody (50 ml per well in PBS) overnight at 4 C. Plates were washed three times with PBS-T.
Wells were blocked with 200 ml of PBS –FBS for 1 h at room temperature. Standards and samples were titrated accordingly and added into the rinsed plate (50 ml per well) and incubated for 2 h at room temperature. Plates were washed five times before addition of biotinylated detection antibody (50 ml per well in PBS – FBS) and incubated for 1 h at room temperature. Finally, the plates were washed seven times before introduction of ABTS substrate in citrate phosphate buffer with hydrogen peroxide. Fully de-
162
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
Table 1 Sensitivity of micro-ELISA for each cytokine Cytokine
Sensitivity of microELISA (pg/ml)
Sensitivity of bead-based assay (pg/ml)
IFNg IL-2 IL-4 IL-5 IL-6 IL-10 IL-13 TNFa
0.55 0.58 0.27 6.40 2.80 1.07 3.87 3.07
15.6 6.6 6.5 2.8 NA 4.7 NA 4.3
NA = data not available.
veloped plates were inactivated with 0.1 M H2SO4 and read at 405 nm with a reference at 470 nm, on a microtiter plate reader (SpectraMax).
3. Results 3.1. Assay development In this assay, capture antibodies against IFNg, IL-2, IL-4, IL-5, IL-6, IL-10, IL-13 and TNFa were initially printed onto a silanized glass surface (array map in Fig. 1A). The capture antibodies were printed at two different concentrations to avoid saturation of sample signals. This provides a unique advantage over conventional ELISA that often requires repeating the assay for concentrated samples. The array also contained control elements to monitor assay performance. They include biotin-Fc, Cy3-Fc, and anti-hamster IgG. Biotin-Fc measures the performance of the detection system The Cy3-Fc printing control allows for the visualization and quantification of protein surface binding without any reagent hybridization. It may also be used to normalize signals across different arrays (Fig. 1B). Finally, addition of hamster IgG during the antigen incubation step assesses the success of the assay and is an indicator for the integrity of other assay reagents (capture antibody status, detector molecules). Additionally, the hamster IgG could also be used as a normalization marker to account for well to well and slide to slide variations. Unlike conventional ELISA that has no reagent controls, the control elements in this assay help detect reagent problems during different steps of the assay and assure proper quality control.
In comparison to conventional ELISA, this procedure requires no overnight pre-incubation of capture antibody on wells leading to considerable savings in time. 3.2. Assay specificity A fully multiplexed assay for cytokine measurement involves mixing all the cytokine standards and detection antibodies. Although this approach is necessary to analyze multiple cytokines simultaneously, any cross-reactivity or interference between reagents and samples would result in assay failure. Each of the capture and detection antibodies has been prescreened for specific reactivity with their respective antigens. Each cytokine antibody pair demonstrates the anticipated specificity to its antigen when added individually (data not shown) or in a multiple detector antibody cocktail (Fig. 2). No cross-reactivity is observed among the cytokines, even at concentrations as high as 500 pg/ml. In Fig. 2, the intensity of the signal from each cytokine varies, reflecting the unique affinity of the capture and detection antibodies to their respective antigen. 3.3. Assay sensitivity The sensitivity of the assay was evaluated, based on the ‘‘minimum detectable concentration’’ or MDC. MDC is defined as the concentration extrapolated from the standard curve of the zero standard of an ELISA plus two standard deviations. This is a theoretical limit of assay sensitivity (Moran and Brown, 1997). The average of eight individual values for each cytokine were determined from arrays incubated with known amounts of cytokine ranging from 10 to 1000 pg/ml
Table 2 Reproducibility of data across slides Cytokine
Average DLU
SD
CV
IFNg IL-2 IL-4 IL-5 IL-6 IL-10 IL-13 TNFa
789,900 774,886 1,738,500 116,418 971,367 498,570 326,752 140,720
107,384 104,330 232,371 13,648 36,527 40,430 30,306 5630
0.14 0.13 0.13 0.12 0.04 0.08 0.09 0.04
SD is standard deviation and CV is the coefficient of variation.
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
163
Table 3 Linearity of cytokines from cell culture mixture Cytokine
Spike concentration (pg/ml)
Observed concentration (pg/ml)
Actual dilution
Observed dilution
IFNg
900 450 112 56 0
1,022.41 538.95 138.17 48.58 0.08 Slope = 0.93 794.06 436.46 132.99 43.04 0.21 Slope = 0.95 762.45 530.93 221.86 38.35 11.51 Slope = 0.99 1098.92 670.85 182.32 52.51 9.12 Slope = 0.96 1078.81 513.76 164.17 45.76 3.83 Slope = 0.93 964.18 565.15 183.15 51.3 0.7 Slope = 0.96 851.14 521.98 168.37 57.51 3.92 Slope = 0.96 1368.56 455.04 127.8 65.12 41.45 Slope = 0.83
Undiluted 1:2 1:8 1:16
1: 1.9 1: 7.4 1: 21
Undiluted 1:2 1:8 1:16
1: 1.8 1: 6 1: 18.5
Undiluted 1:2 1:8 1:16
1: 1.4 1: 3.4 1: 19.8
Undiluted 1:2 1:8 1:16
1: 1.6 1: 6 1: 21
Undiluted 1:2 1:8 1:16
1: 2.1 1: 6.6 1: 23.6
Undiluted 1:2 1:8 1:16
1: 1.7 1: 5.3 1: 18.8
Undiluted 1:2 1:8 1:16
1: 1.6 1: 5 1: 14.8
Undiluted 1:2 1:8 1:16
1: 3 1: 10.7 1: 21
IL-2
IL-4
IL-5
IL-6
IL-10
IL-13
TNFa
900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0 900 450 112 56 0
along with a negative control (0 pg/ml of antigen). Average signal values expressed as ‘‘Digital Light Units’’ or DLU were used to plot standard curves
(Fig. 3A to H). Subsequently, the respective MDC was extrapolated from an individual standard curve of the selected cytokine and expressed in pg/ml (Table 1).
164
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
Overall, the sensitivity of the cytokine ranges from 0.55 to 6.4 pg/ml in our assay system. It compares favorably with reported values for a multiplex flow cytometry-based assay (Human Th1/Th2 cytokine CBA kit, PharMingen). Reproducibility across wells and slides was evaluated by performing a multiplex microELISA experiment in which two wells from each of three slides were assayed with 200 pg/ml of the antigen mixture. Each well had four spots per antibody and the results shown in Table 2 are expressed as the mean of 24 data points. The sum of values from three different slides was calculated to determine the extent of slide variations. Briefly, well to well variations range from 5% to 18%. The variations among slides range from 4% to 14%. 3.4. Comparison of microELISA and conventional ELISA for cytokine quantification Known amounts of each of the eight cytokines were assayed using conventional ELISA and microarray ELISA to compare their relative sensitivity and dynamic range. Fig. 3A to H represents standard curves plotted from data obtained from these assays. The correlation coefficient of the curves were greater than 0.9, except for IL-13. First, the ELISA was compared to individual microELISA (singleplex) where a single cytokine was analyzed in the absence of other cytokines but in the presence of all the detection antibodies (Fig. 3). For each of the cytokines on the array, greater dynamic range and similar linearity were achieved with the microELISA as compared to conventional ELISA. The dynamic range for microELISA ranges from 2.5 to 3 logs for each of the antigens, while the conventional ELISA has only 1.5 to 2 log dynamic range. The multiplexed format, in which all eight cytokines and biotinylated secondary antibodies were added in the same well also showed greater dynamic range and comparable linearity as the standard ELISA (Fig. 3). 3.5. Assay validation The ability of the multiplexed microELISA assay to simultaneously quantify recombinant cytokines in cell culture media was tested. Known amounts of cytokines were spiked into tissue culture media (RPMI plus 100 ml/l fetal bovine serum) to determine the linearity
and correlation of extrapolated values. Four different concentrations were spiked—900, 450, 112 and 56 pg/ml. For each of the cytokines, the observed dilution is close to the expected dilution; namely 1:2, 1:8 and 1:16 (Table 3). The slopes of the regression line from the observed concentration for most cytokines are greater than 0.9.
4. Discussion We have described a new multiplexed ELISA assay for cytokine quantification using microarrays. A major advantage of the assay is the amount of information that can be obtained from sample volumes under 20 ml while a conventional ELISA requires up to 100 ml per well. This translates into a 40-fold savings in sample required to measure eight cytokines. Substantial savings in time is achieved in processing samples since eight cytokines can be monitored simultaneously. This assay would be beneficial to measure protein levels in samples of limited availability such as needle biopsies, providing a useful tool for exploring important biological questions. Besides cost and time, specificity, sensitivity and reproducibility are other factors that are important for a quantitative assay. Our data clearly showed specificity in both singleplex and multiplex format. No cross reactivity of the detector antibodies was observed, this is likely due to the added benefit of using monoclonal antibodies during this step in the multiplex microarray design. The assay sensitivity, based on minimum detectable concentration, is at least comparable to commercially available bead-based multiplex assay systems. Moreover, the microarray assay outperforms the beadbased assay by roughly 30-fold greater sensitivity for some of the cytokines. Although an average of 9.6% CV for slide reproducibility is higher than those reported for some multiplexed ELISAs, it does not affect the biological validation results significantly (Huang et al., 2001). For each of the cytokines tested, intra-slide CV decreased with increasing concentration of antigen up to 50 pg/ml after which the CV remained at around 10%. Another advantage of the microarray platform described here is the ability to analyze tens to hundreds of samples simultaneously. Each of the 16 wells in the glass slide has similar printed arrays. This format can also been extended to a 96-well glass plate,
S.W. Tam et al. / Journal of Immunological Methods 261 (2002) 157–165
allowing for 96 samples to be analyzed on a single plate. This platform can be easily automated with commercially available fluid handling robotics. An important aspect that impacts data analysis is the metric used to quantify spot signal intensity. Typical microarray spots are likely to have variable morphology (size, shape) dependent on the type of printing mechanism (CV for printing with the BioChip Arrayer was less than 10%). The spots also have microheterogeneity that reflect surface irregularities impacting intra-spot pixel intensities. The commonly used metric for measuring spot intensities ‘‘sum of pixels’’ can result in incorrect differences in signal for the same concentration of antigen dependent on the characteristics of the printed antibody spot. Measuring the average pixel intensity of the spot (as described under Materials and methods) resulted in data that provided better quantification of the antigens being measured. Unlike conventional ELISA, which requires a standard curve to be performed with each 96-well plate, for a given print batch of microarrays a standard curve need be plotted only once and used for quantification of data from samples analyzed on all the other arrays from that batch. In addition, the incorporated controls in each array allow for data to be normalized between different samples thereby accounting for any variations that may occur from one array to another. Good correlation in cytokine levels measured using conventional ELISA and the microELISA was observed in Jurkat cells stimulated with PHA and PMA (Huang et al., 2001). Our current assay system addresses some of the challenges of conventional ELISA by the ability to simultaneously measure multiple antigens in shorter experimental time. One can extend the platform to examine immunoglobulins, chemokines, growth factors or soluble receptors. This assay can provide an excellent tool for research or clinical diagnostic as well as a valuable platform in high throughput drug screening for identification of lead drug targets and candidates. References Carson, R.T., Vignali, D.A., 1999. Simultaneous quantitation of 15 cytokines using a multiplexed flow cytometric assay. J. Immunol. Methods 227, 41. Cherwinski, H.M., Schumacher, J.H., Brown, K.D., Mosmann, T.R., 1987. Two types of mouse helper T cell clone. III. Further differ-
165
ences in lymphokine synthesis between Th1 and Th2 clones revealed by RNA hybridization, functionally monospecific bioassays, and monoclonal antibodies. J. Exp. Med. 166, 1229. Collins, F.S., 1999. Microarrays and macroconsequences. Nat. Genet. 21, 2 [Editorial]. de Wildt, R.M., Mundy, C.R., Gorick, B.D., Tomlinson, I.M., 2000. Antibody arrays for high-throughput screening of antibody – antigen interactions. Nat. Biotechnol. 18, 932. Del Prete, G.F., De Carli, M., Mastromauro, C., Biagiotti, R., Macchia, D., Falagiani, P., Ricci, M., Romagnani, S., 1991. Purified protein derivative of Mycobacterium tuberculosis and excretory – secretory antigen(s) of Toxocara canis expand in vitro human T cells with stable and opposite (type 1 T helper or type 2 T helper) profile of cytokine production. J. Clin. Invest. 88, 346. Douglas, A.S., Monteith, C.A., 1994. Improvements to immunoassays by use of covalent binding assay plates. Clin. Chem. 40, 1833. Falipou, S., Chovelon, J.-M., Martelet, C., Margonari, J., Cathignol, D., 1999. New use of cyanosilane coupling agent for direct binding of antibodies to silica supports. Physicochemical characterization of molecularly bioengineered layers. Bioconjugate Chem. 10, 346. Haab, B.B., Dunham, M.J., Brown, P.O., 2001. Protein microarrays for highly parallel detection and quantitation of specific proteins and antibodies in complex solutions. Genome Biol. 2, 4.1, RESEARCH0004. Huang, J.X., Mehrens, D., Wiese, R., Lee, S., Tam, S.W., Daniel, S., Gilmore, J., Shi, M., Lashkari, D., 2001. High-throughput genomic and proteomic analysis using microarray technology. Clin. Chem. 47, 1912. Lander, E.S., 1999. Array of hope. Nat. Genet. 21, 3 [Editorial]. MacBeath, G., Schreiber, S.L., 2000. Printing proteins as microarrays for high-throughput function determination. Science 289, 1760. Mendoza, L.G., McQuary, P., Mongan, A., Gangadharan, R., Brignac, S., Eggers, M., 1999. High-throughput microarray-based enzyme-linked immunosorbent assay (ELISA). BioTechniques 27, 778. Moran, R.F., Brown, E.N., 1997. The smallest concentration. Clin. Chem. 43, 856. Ohshima, Y., Yang, L.P., Avice, M.N., Kurimoto, M., Nakajima, T., Sergerie, M., Demeure, C.E., Sarfati, M., Delespesse, G., 1999. Naive human CD4+ T cells are a major source of lymphotoxin alpha. J. Immunol. 162, 3790. Reen, D.J., 1994. Enzyme-linked immunosorbent assay (ELISA). Methods Mol. Biol. 32, 461. Seder, R.A., Paul, W.E., 1994. Acquisition of lymphokine-producing phenotype by CD4+ T cells. Annu. Rev. Immunol. 12, 635. Southern, E., Mir, K., Shchepinov, M., 1999. Molecular interactions on microarrays. Nat. Genet. 21, 5. Voller, A., 1978. The enzyme-linked immunosorbent assay (ELISA) (theory, technique and applications). Ric. Clin. Lab. 8, 289. Wiese, R., Belosludtsev, Y., Powdrill, T., Thompson, P., Hogan, M., 2001. Simultaneous multi-analyte ELISA performed on a microarray platform. Clin. Chem. 47, 1451. Zhai, Y., Ghobrial, R.M., Busuttil, R.W., Kupiec-Weglinski, J.W., 1999. Th1 and Th2 cytokines in organ transplantation: paradigm lost? Crit. Rev. Immunol. 19, 155.