Array-Based Multiplexed Screening and Quantitation of Human Cytokines and Chemokines Cheng C. Wang,*,† Ruo-Pan Huang,‡ Martin Sommer,† Henry Lisoukov,† Ruochun Huang,‡ Ying Lin,‡ Thomas Miller,† and Jocelyn Burke† PerkinElmer Life Sciences, 800 Research Parkway, Meriden, Connecticut 06450, and Department of Gynecology and Obstetrics, Emory University, School of Medicine, 1639 Pierce Drive, Atlanta, Georgia 30322 Received March 26, 2002
HydroGel-coated slide is a porous substrate based on a polymer matrix that provides a threedimensional hydrophilic environment similar to free solution suitable for biomolecular interactions. This substrate has been used to develop fluorescence-based multiplexed cytokine immunoassays. Fortythree monoclonal antibodies (mAb) of cytokines and chemokines were printed at a volume of 350 pL per spot using a Packard BioChip Arrayer. For each probe, four replicates were printed at a pitch of 500 µm in the layout of a 13 × 16 pattern on a 12 × 12 mm2 HydroGel pad. Cytokines and chemokines that are captured by the arrayed mAbs are detected by using another biotinylated mAb, following by the addition of a Texas Red-conjugated streptavidin. The fluorescent images of arrays were recorded using a Packard ScanArray 5000 confocal slide scanner and quantitated using Packard QuantArray software. Experiments demonstrated that 43 cytokines and chemokines could be simultaneously screened and quantitated in conditioned culture media, cell lysates, and human plasma. Using this chip, we have examined cytokine expression in breast cancer cells and identified the chemokines associated with human cervical cancers. Keywords: microarrays • protein chip • chemokine • cytokine • multiplex • immunoassay • biochip • parallel
Introduction Cytokines and growth factors mediate a wide range of physiological processes, including hematopoiesis, immune responses, wound healing, and general tissue maintenance.1,2 Cytokines are mostly glycoproteins that play crucial roles in cell-to-cell signaling and have been grouped into separate families according to structural and functional similarities.3,4 Functionally, cytokines have also been classified as being proinflammatory (stimulatory) or anti-inflammatory (inhibitory). Cytokines exhibit pleiotropy and redundancy: a cytokine can have multiple functions, and these functions can be mimicked in part by other cytokines. For example, breast cancer cells have a higher density of high-affinity IL-6 receptors than normal tissue. IL-6 has a pleiotropic effect on breast cancer cells. In human MDA-MB231 breast tumor cells, IL-6 upregulates tumor necrosis factors (TNFa) and IL-1a and IL-1b expression.5 Cytokines can act in sequence, where one cytokine triggers the synthesis or release of one or several other cytokines, which can themselves regulate the synthesis and release of the triggering cytokine as well as other upstream and downstream cytokines. Cytokines can also act in concert and in parallel, where regulatory patterns will depend on the cooperative or antagonistic effects of multiple cytokines in response to a * To whom correspondence should be addressed. E-mail: ccwang62@ yahoo.com. † PerkinElmer Life Sciences. ‡ Emory University. 10.1021/pr0255203 CCC: $22.00
2002 American Chemical Society
neuro-physiological or neuro-immunological process. Interestingly, not all properties of one cytokine will be affected the same way by interactions with other cytokines. Cytokine interaction can be additive, synergistic, or antagonistic.6 Another example is the development of human cervical cancer. It is believed that infection of cervical mucosa cells with highrisk human-papillomavirus (HPV) types such as HPV-16 or 18 caused the cervical cancer. Monocyte chemo attractant protein (MCP)-1, a chemokine, is an important factor involved in the cross talk between mononuclear cells and human papillomavirus (HPV)-infected cervical epithelia.7 Simultaneously measuring multiple cytokines within the same sample will be vital to reveal the complicated cytokine network. Miniaturization and multiplexing with protein microarrays allows for a dramatic reduction in the amount of sample required (typically in the 10-100 µL range), an increase in the number of analytes that can be measured simultaneously, and increased throughput potential. This increased information attained by quantitatively measuring many analytes in parallel maintains relative expression levels within the same sample and may enable determination of relationships between diverse proteins involved in many pathways in response to the same environment.8,9 There have been several recent developments in microarray-based multiplexed cytokine assays. Antibodies spotted into glass-bottom microwell plates10,11 or polystyrene 96-well plates,12 onto glass slides13,14 or membranes,15,16 and antibody-coated beads17 have been used to Journal of Proteome Research 2002, 1, 337-343
337
Published on Web 06/21/2002
research articles
Wang et al.
simultaneously detect multiple antigens in a miniature format. In general, proteins are more sensitive to their environment than nucleic acids. The hydrophobicity of many glass and plastic surfaces may cause protein denaturation, thus rendering lower sensitivity and high noise-to-signal ratios. Consequently, substrate choice is a major consideration when designing protein microarray experiments. HydroGel-coated slides are based on a specialized polyacrylamide formulation that offers several advantages for protein applications including a high probe binding capacity, very low inherent and assay dependent background, and a hydrophilic environment.18,19 We printed 43 cytokine and chemokine antibodies on a HydroGel-coated slide for simultaneously screening and quantitating cytokines in supernatants of cell culture, conditioned media, cell lysate, and human blood plasma. We also used the chip to profile human cytokine expression of breast tumor cells and to monitor human cytokine expression in cervical tumor samples.
Materials and Methods Standard recombinant human cytokine and chemokine ELISA kits, which include anti-cytokine or chemokine capture antibodies, cytokines or chemokines, and biotinylated anticytokine or chemokine detection antibodies, were purchased from R & D Systems (Minneapolis, MN) and BD Biosciences (San Diego, CA). Anti-bovine IgG (goat) antibody, biotinylated anti-bovine IgG (goat) antibody, Texas Red-conjugated bovine IgG fraction, and Texas Red-conjugated streptavidin were ordered from Rockland Immunochemicals for Research (Gilbertsville, PA). SuperBlock blocking buffer was purchased from Pierce Endogen (Rockford, IL). Fetal bovine serum and DME media were purchased from HyClone Laboratories (Logan, UT). All proteins were reconstituted according to the manufacturers’ recommendations. Stock cytokine solutions and subsequent serial dilutions were prepared using DME culture medium supplemented with 10% fetal bovine serum. All other reagents were purchased from Sigma-Aldrich (St. Louis, MO). Cancer cell lines (MDA-MB-231 cells)20 were obtained from ATCC (Manassas, VA) and grown in Dulbecco’s modified Eagle’s medium containing 10% fetal calf serum and a mixture of glutamine, penicillin, and streptomycin. Patient blood serum was collected in Dr. R. Huang’s laboratory at Emory University. SuperAldehyde glass slides were purchased from Telechem (Sunnyvale, CA), and FAST slides were purchased from Schleicher & Schuell (Keene, NH). HydroGel-coated slides were manufactured in house at Packard Bioscience, now PerkinElmer Life Science (Meriden, CT), with a licensed technology. Antibody Array on HydroGel Substrate. Arrays were printed onto two 12 × 12 mm2 HydroGel pads per slide. Antibody probes were diluted to a concentration of 0.5 mg/mL in PBS (0.14 M NaCl, 0.003 M KCl, 0.01 M sodium phosphate, pH 7.2), and source plates were set up in 384-well conical bottom plates (MJ Research, Waltham, MA). Prior to printing, HydroGelcoated slides were washed three times for 10 min each in dH2O, dried by centrifugation, and further dried in a 40 °C oven for 20 min. The slides were allowed to cool to room temperature for 5 min prior to printing. The antibody probes were printed at a volume of 350 pL per spot using a Packard BioChip Arrayer. For each probe, four replicates were printed at a pitch of 500 µm in the layout illustrated in Figure 1. A series of replicate spots of bovine serum albumin (BSA) in printing buffer served as negative controls. Biotinylated anti-bovine IgG (goat) antibody served as a detection control monitoring the ability of the immunoassay components to form a complex in the 338
Journal of Proteome Research • Vol. 1, No. 4, 2002
Figure 1. Microarray layout. The antibody probes were printed at a volume of 350 pL per spot using a Packard BioChip Arrayer. For each probe, four replicates were printed at a pitch of 500 µm. Bovine serum albumin (BSA) in printing buffer served as negative control. A series of replicate spots of biotinylated antibovine IgG (goat) antibody (10-fold deceasing intervals from A1 to D1) served as a detection control monitoring the ability of the immunoassay components to form a complex in the HydroGel matrix. Texas Red-conjugated IgG was spotted from A13 to D13 (10-fold increasing intervals); the spots served as printing quantity controls. ENA-78 (A11) is set to be a positive control for cytokine assays.
HydroGel matrix. Tips were washed in buffered surfactant and sonicated between aspirates to eliminate carryover. Following printing, HydroGel-coated slides were incubated overnight at 25 °C in a humidified chamber (65% humidity). Slides were then washed three times for 30 min each in PBS + 0.5% Tween 20, pH 7.2 (PBST), to remove any unbound probe and stored overnight in Pierce Endogen SuperBlock buffer at 4 °C. Immunoassays. Printed slides were rinsed three times in buffered surfactant wash buffer (PBST) to remove blocking solution and then rinsed one time with PBS. After being dried with low speed centrifugation, arrays were incubated with the target samples. All incubations were carried out in a 50 µL volume using Frame-Seal gasketed hybridization chambers (MJ Research, Waltham, MA) to prevent cross-contamination of microarrays. Incubations were carried out for 1 h at room temperature in gasketed and sealed arrays on a rotating shaker. Slides were then washed two times for 15 min each in wash buffer (PBST) followed by a 5 min wash in PBS, pH 7.2, to remove excess surfactant and dried by brief centrifugation. A 50 µL aliquot of a cocktail of all 43 biotinylated detection antibodies at levels optimized for this system was applied to the slides, and they were incubated for 1 h at room temperature followed by washing and centrifugation as in the previous step. Finally, the slides were incubated for 30 min at room temperature in an excess of Texas Red-conjugated streptavidin (50 µL). Slides were again washed and dried by the above procedure, followed by an additional 5 min rinse in PBS, to remove detergent prior to imaging. Imaging and Data Analysis. Arrays were imaged in the Texas Red channel using a Packard ScanArray 5000 confocal slide scanner. Optimal laser power and PMT settings were determined such that the most intense spots for each analyte series were at approximately 80% saturation. Within an experiment,
Screening and Quantitation of Human Cytokines and Chemokines
all slides of the same substrate type were scanned using the same settings. Images were quantitated using Packard QuantArray software, and data analysis was carried out in Microsoft Excel. Ratio metric analysis and scatter plot were achieved by overlapping two images from top and bottom arrays on a slide. The mean DLU per pixel values for each array were normalized using the signal obtained from the detection antibody control and the values were baseline subtracted using the signals from the no-cytokine present control. The data were plotted to determine dynamic range and dose response using Statistica software (StatSoft, Tulsa, OK). The slope of the linear portion of each curve and the coefficient of determination (r2) were determined by trimming the curves and analyzing the resulting linear portion of each curve by linear regression using SigmaPlot software (SPSS, Chicago, IL). ELISA. Conventional ELISA was performed according to the manufacturer’s instructions. Ninety-six-well MaxiSorp plates from Nalge Nunc International (Rochester, NY) were coated overnight at 4 °C using 50 µL of 8 µg/mL capture antibodies. BSA (1%) in PBS was used as a blocking buffer. A 100 µL portion of conditioned media and different concentrations of standard cytokines were added to each well in duplicate. Unbound materials were washed out with PBST. A 100 µL portion of 1 µg/mL biotinylated anti-cytokine detection antibody was added to each well. The plates were incubated for 1 h at room temperature. After washing, 100 µL of streptavidin-HRPconjugated antibodies was added to each well, and incubation was continued for 30 min at room temperature. Extensive washing followed. After incubation with substrate solution containing ethylbenzthiazoline sulfonate, the o.d. at 405 nm for each well was measured in a microplate reader.
Results Development of Human Cytokine Arrays. (1) Array Construction. The monoclonal antibodies of 43 cytokines and chemokines were printed in a 4 columns by 13 rows format as indicated in Figure 1. Each antibody probe was printed in quadruplicate. Cytokines and chemokines are arranged into several functional groups. Group 1 (A2-A10) belongs to CC, C, and CX3C chemokines, group 2 (A11, A12, and B2-B5) is CXC chemokines, group 3 (B6-B11) contains hematopoietic factors, group 4 (B12) is a chronic inflammatory mediator, group 5 (C2-C8) is composed of growth and angiogenic factors, group 6 (C9, C10) belongs to the IL-6 family, group 7 (C11, C12) represents the TNF family, group 8 (D2, D3) is IL-1s, and group 9 (D4-D11) consists of immunomodulating interleukins. There are several controls built within the chip. Positive cytokine control is located at A11 (ENA-78), negative control used is BSA at a printing concentration of 1 mg/mL (D12), printing control is achieved by directly printing various concentrations of Texas Red-conjugated bovine IgG (row 13), and detection is monitored by direct printing of different concentrations of biotinylated goat anti-bovine IgG (row 1). (2) Substrate Characterization. The performance of HydroGel-coated slides were compared to that of two other commercially available substrates in a cytokine model assay. The substrates chosen for comparison were an aldehyde-derivatized glass substrate (SuperAldehyde glass slides) and a threedimensional nitrocellulose polymer-coated slide (FAST slides). The experiment was performed essentially the same as described above with the exceptions that cytokine targets were applied in PBS at concentrations ranging from 1.6 pg/mL to
research articles
Figure 2. Comparison of standard curves derived for IL-1β (pg/ mL) performed with multiplexed assays on varied substrates. A multiplex assay was performed with six cytokines using arrays printed on HydroGel-coated slides and on two commercially available substrates. Cytokine concentrations applied were from 1.6 pg/mL to 50 ng/mL in PBS. Detected signal is reported as DLU per pixel. Curve A (filled circles) derived from results using arrays on HydroGel-coated slides. Curve B (open circles) derived from results using arrays on aldehyde-derivatized glass slides. Curve C (filled squares) derived from results using arrays on nitrocellulose-coated slides. Results show a broad dynamic range for assay performed with HydroGel-coated slides and narrowed dynamic range on other substrates owing to decreased response at the lower extremes of the curves.
50 ng/mL in 0.4 Log steps, and the blocking and washing protocols were adjusted to fit the manufacturers’ recommendations. For imaging, the laser power and PMT settings of the ScanArray 5000 confocal slide scanner were optimized for each substrate. Compared to the HydroGel-coated slide (curve A in Figure 2), the dynamic range of the aldehyde-derivatized glass slide (curve B in Figure 2) was short of the lower concentration points due to the capture capacity. While the nitrocellulose-coated slide (curve C in Figure 2) performs similar to HydroGel-coated slides at the higher target concentrations, its sensitivity is much lower due to high levels of inherent fluorescent background. The data from IL-1β demonstrated that the higher probe loading capacity of HydroGel substrate in comparison to the two-dimensional substrate, coupled with the low inherent fluorescent background of HydroGel-coated slides, results in superior overall performance. All six cytokines assayed exhibit the same trends. This may represent an important feature for applications analyzing proteins that have a broad expression range due to either physiological or pathological causes. (3) Standard Curves. We previously demonstrated the utility of the HydroGel substrate in a model fluorescent immunoassay in microarray format. Six more standard curves were generated by spiking cytokines into cell culture media supplemented with fetal bovine serum (FBS). To approximate a typical cytokine immunoassay on tissue culture supernatants, the 6-fold multiplex cytokine assay was run using cytokine-spiked DME medium supplemented with 10% fetal bovine serum. The experiment was set up such that two arrays (on one HydroGelcoated slide) were incubated with a mix of all six cytokines at Journal of Proteome Research • Vol. 1, No. 4, 2002 339
research articles
Wang et al.
Figure 3. Standard curves for individual cytokines derived from serum-based assays performed in multiplex on HydroGel-coated slides. Slides were prepared and assays were performed as stated in the Materials and Methods. Images were collected using a ScanArray 5000 microarray scanner, and data were normalized to control spots and then baseline subtracted. Plots of a full range of data for six cytokines show a sigmoid curve of dose response typical of assays from this system.
the same concentration. Nine concentrations (5 pg/mL to 50 ng/mL in 0.5 log steps) of each cytokine were tested to generate standard curves. In addition, as a negative control, one slide was incubated with DME medium supplemented with 10% fetal bovine serum only, followed by processing with the detection antibody mixture and Texas Red-conjugated streptavidin as described above. Standard curves produced with HydroGel-coated slides demonstrate a dynamic range of up to 3 orders of magnitude and observed detection limits in the picogram/milliliter range for five of the six cytokines tested in a model multiplexed array system. A standard curve for each cytokine was generated in Figure 3, and the detection limits for each cytokine were calculated. The IL-6 assay demonstrated sensitivity at 2 pg/ mL, followed by TNFa at 6 pg/mL. IFNg (16 pg/mL) and IL-1b (17 pg/mL) are almost identical. IL-2 has the sensitivity at 20 pg/mL. IL-13 has demonstrated a slightly higher detection limit at 698 pg/mL owing to the performance of antigen-antibody pairs. The microarray-based cytokine assays on HydroGelcoated slides resulted in an observed sensitivity of less than 20 pg/mL for five out of six cytokines tested. This level of sensitivity results from the high degree of reproducibility of the data (CV% is less than 10%) and the low baseline values of the probes on HydroGel substrate. Ongoing efforts to optimize relative levels of antigen-antibody pairs suggest that a 4- to 5-fold reduction in the detection limit is attainable without signal amplification. The high loading capacity of HydroGelcoated slide substrate, with retained protein function, allows the researcher to choose from a broader range of capture antibody deposition than other commercially available substrates. This in turn enhances flexibility in choosing relative levels of matched pairs for multiplexed analysis and may enable 340
Journal of Proteome Research • Vol. 1, No. 4, 2002
Table 1. Summary of the Theoretical Limits of Detection, Slope, and Coefficient of Determination (r2 Value) for Each Cytokine
y-intercept slope R2
TNF-R
IL-1β
IL-2
IL-6
IL-13
IFN-γ
1.8 1.21 0.97
1.7 1.11 0.99
0.92 1.02 0.96
1.8 0.89 0.98
0.01 1.26 0.92
1.2 0.90 0.95
the use of antigen-antibody pairs that would not be supported by other systems. The combination of sensitivity and broad dynamic range would allow for the discrimination between the healthy and diseased states associated with these cytokines and quantitation of cytokine level present in the pathological state. (4) Performance. (a) Linear Range. Serial dilutions from a stock solution were performed to prepare the range of cytokine concentrations used for generating standard curves. For each concentration, two arrays on one slide were used. Each experiment included a cytokine-free negative control and a detection control (biotinylated anti-bovine IgG). The endpoints of each graph approximate the boundaries of the dynamic range for the detection of the target cytokines. The observed dynamic ranges ranged from a low of at least 2 Logs for IL-13 to a high of at least 3.0 Logs for IL-1B, IL-2, IL-6, and IFN-R (Figure 3). The standard curves were trimmed and the linear portion reflecting the dynamic range of the assay was used to calculate the slope of the dose response. The coefficient of determination (r2) was determined by trimming the curves and analyzing the resulting portion of each curve by linear regression (Table 1). The slope of the linear regression equation best represents the doseresponse for each cytokine. As summarized in Table 1, all six cytokines demonstrated pronounced dose-responses, having
Screening and Quantitation of Human Cytokines and Chemokines
slopes ranging from 0.89 to 1.26 and coefficient of determination (r2) ranging from 0.92 to 0.99. These dose-responses are suitable for quantitating the levels of the cytokines from an experimental sample when applied to these standard curves. (b) Cross Reaction. Unlike standard singleplex methods of analyte detection, assays performed in multiplex are potentially complicated by possible cross reactivity from multiple capture and detection antibody pairs used in conjunction. To rule out this possibility, each cytokine was applied to an array individually and the signals of the unrelated probes were measured. The arrays tested for cross reactivity showed no signal above background for any of the cytokines, suggesting that cross reactivity is not a factor in this system (data not shown). A similar measure of cross reactivity between the multiplex components is to compare standard curves for an analyte derived with microarrays when the analyte is alone in the capturing step or when it is in a mixture with other analytes. This was done with IL-6 as the detected analyte with the sixplex cytokine immunoassay chip in the presence of all six biotinylated antibodies. The resulting dose-response curves (Figure 4) exhibit little difference in dose-response or dynamic range, indicating essentially no interference from the other antibodies in the multiplexed system. However, multiplexed assays bear lower detection limits compared with singleplexed assays. We checked with all six cytokines, and results for other cytokines are similar (data not shown). (c) Variability. The intra-variability was determined by comparing the signal from 16 duplicate spots in the same array. The variability of spots from different slides (inter-variability) was determined from four different arrays. The CV (standard deviation divided by average) was usually less than 10%, suggesting the reliability of our system (data not shown). Tracing Human Cytokine Expression in Breast Tumor Cells. We have achieved simultaneous assays of multiple cytokines spiked into DME medium with 10% FBS. To validate performance in real samples, we applied the chip to human estrogen-receptor-negative MDA-MB231 breast tumor cells. Characterization of cytokine expression in breast cancer cell lines would reveal the relationship between serum IL-6 levels with poor prognosis in breast cancer.5 In the laboratory, we separated the cancer cells from their culture media. Cell lysate was screened using the top array; GCSF, IL-8, and Il-1β were detected. For the bottom array, cell culture media was applied. GCSF and IL-8 were detected, but IL-1β was missing. Instead, IL-6 was strongly presented. As seen in Figure 5, we overlapped the two images from the top and bottom arrays and did a scatter plot. In the scatter plot, the spots above the correlation line were the cytokines in cell culture media, and the spots below the line indicated the cytokines inside of cells. The results suggested the sequence of cytokine induction and expression in breast cancer cells. As seen in the plot, at the time of assays IL-1β was just expressed since it was inside of cells while IL-6 had already been released into cell culture media. Monitoring Human Chemokine Expression in Cervical Cancer. In earlier studies, the experimental model of a negative regulatory loop between the expression of the HPV oncogenes E6/E7 and the MCP-1 gene in vivo was proposed.7 Since macrophages and macrophage-derived cytokines appear to be important in the transcriptional regulation of high-risk types of human papillomaviruses (HPV), monitoring MCP-1 expression by in situ hybridization (ISH) in histologically distinct stages of cervical intra-epithelial neoplasms (CIN), cervical cancer, and non-HPV-associated cases of erosive endocervicitis
research articles
Figure 4. Comparison of standard curves generated for IL-6 using arrays on HydroGel-coated slides with target cytokine applied either in multiplex (filled circles) or singleplex (open circles). All six biotinylated detection antibodies were present in both immunoassays. Assay was performed as stated in the Materials and Methods with cytokines spiked into DME medium with 10% FBS. Cytokine concentrations ranged from 50 pg/mL to 50 ng/mL. Curves show similar dose response and detection range except lower detection limits in the multiplex form.
indicated that dysregulation of MCP-1 gene expression may represent an important step during HPV-linked carcinogenesis, allowing the escape of virus-positive cells from local immune response.21 We decided to test the hypothesis using our cytokine chip. Several cervical cancer blood samples were compared with normal blood samples for chemokine expression. The cervical cancer blood was applied to the top array while the normal control blood was applied on bottom array in a HydroGel-coated slide. After the assays, two images were overlapped, and a scatter plot was generated (Figure 6). In highgrade dysplasia (CIN II) patients, expression of MCP-1 decreased to 33%, and MCP-2 was expressed at a 40% level of the normal samples. Other cytokines and chemokines showed little change during carcinogenesis. Our results support the hypothesis that HPV oncoproteins are negative regulators of MCP-1 transcription. Quantitation of Cytokines in Human Serum. Measuring cytokines in biological fluids and tissues is challenging since these mediators are involved in many pathological manifestations of inflammatory, infectious, and immunological diseases. Developing an accurate and sensitive method for the measurement of cytokines in body fluids is an absolute prerequisite for the proper use of these mediators in clinical practices. Many factors contribute to the complexity of cytokines quantitation.22 To further validate the chip, we compared the quantitative data obtained from chip-based assays against 96-well plate based ELISA assays. The same patient blood serum was subjected to two assay systems. One had the traditional 96-well plate based ELISA, and the other was measured with our cytokine multiplex array assay. The results obtained indicate the two assay systems performed very closely (Table 2). Journal of Proteome Research • Vol. 1, No. 4, 2002 341
research articles
Wang et al.
Figure 5. Composed image of two arrays (left). Two identical 43 cytokine arrays were applied separately with cell lysate (red) and cell supernatant (green) from same MDA-MB231 breast cancer cell culture. The yellow spots represent the occurrences of cytokines in both samples. Red spots indicate the cytokines expressed in cell lysate only, and green spots represent the released cytokines in cell culture media. Ratio metric analysis of overlapped images revealed IL-6 is in cell supernatant and IL-1β is presented inside of the cell at the time of assays. Table 2. Quantitation of Cytokines Using Both Classical Well-Based ELISA and Array-Based Multiplex Assay
96-well ELISA (pg/mL) array multiplex assay (pg/mL) % difference
Figure 6. Ratio metric plot of cervical intra-epithelial neoplasms (CIN-II) blood against normal human blood. There are two 12 × 12 mm2 HydroGel pads on a slide. Identical arrays were printed on both pads. Patient plasma was applied to the top array, and plasma from normal persons was applied on bottom array. After incubation and washing, both arrays were subjected to the same postincubation immuno-detection. Images from top and bottom arrays were overlapped, and a ratio metric plot was generated. The spots circled were the ratios between patient and normal blood samples. The expression of MCP-1 decreased to 33% and MCP-2 to 40% respectively in patients.
Discussion Protein chips are fabricated in much the same way as DNA microarrays, but getting all the elements to work is far more difficult than with DNA arrays.23,24 Proteins bind to their targets based on the three-dimensional shape of each other, as well as a myriad of chemical interactions. Soluble proteins found in blood typically have water-friendly hydrophilic groups near their surface; thus, a biochip surface that is chemically treated to bind hydrophilic proteins will be preferred. The degree of multiplexing that can be achieved on microarrays will in part be dependent on the performance of antigen-antibody pairs on the substrate. Currently, flat surface based chemical-treated 342
Journal of Proteome Research • Vol. 1, No. 4, 2002
IL-6
TNFa
96 97 1.04
52 47 -9.62
glass slides and nitrocellulose-coated slides have been used to spot proteins.25,26 Thus, we compared HydroGel with these two substrates. As the data indicate, HydroGel offers several advantages. The porous nature of the substrate enables probes, analytes, and detection molecules to form the necessary molecular interactions to generate stable complexes in immunoassays, and the high loading capacity and hydrophilic environment of the HydroGel slides may give an advantage over other microarray substrates as well. The three-dimensional substrate offers much better fluid performance. For example, two-dimensional substrates have some difficulties in handling cell lysate samples because of the high viscosity of cell debris (personal observation at Emory University) while HydroGel performs well. The structure of the HydroGel substrate is also sufficiently open for efficient washing of excess analytes and detecting antibodies so that background noise and nonspecific binding activities are low. This contributes to the substrate’s ability to achieve limits of detection comparable to what is commonly reached in traditional assays. This combination of physical attributes is unique to HydroGel-coated slides, making them an excellent choice for protein microarray applications. Keeping proteins active during the arraying process is another valued asset of our protein arraying system. We achieved these by adopting two protein-arraying tools. A noncontact piezoelectric robotic dispenser, Packard BioChip Arrayer, is used to produce arrays. The small drops (350 pL) are squeezed out gently without touching (pin tool based arrayer) or heating (inkjet based arrayer) of protein samples, which ensures the proteins are not denatured while printing. The precision on the controlled volume delivery accounts for the accuracy in quantitation assays. Keeping proteins in their native environments on slides is another unique feature of HydroGel-coated slides. In addition to its three-dimensional
research articles
Screening and Quantitation of Human Cytokines and Chemokines
network structure, HydroGel contains a protein stabilizer, which keeps the antibodies stable and hydrated. Additionally, the use of exceedingly small samples is a critical factor for drug discovery and diagnostics. The minimum sample requirement is a critical factor. Immunoassays are assays that use the binding of antibodies to a ligand to detect specific interactions, resulting in quantitation of that ligand. Immunoassays can be highly specific, quick, and simple to carry out and require little expensive machinery. These assays are typically carried out in a microplate format, in which each target analyte has to be measured in a separate reaction. Consequently, ELISAs require milliliter volumes of sample to analyze multiple analytes. Microarray based parallel multiplex immunoassays enable simultaneous detection of many analytes from a small volume (50 microliters or less) of sample fluid. The use of exceedingly small quantities of capture antibody immobilized in microspots facilitates ambient analyte assay conditions. Under these assay conditions, the fractional occupancy of capture antibody is independent of sample volume. Analyte concentrations are thus measured with greater sensitivity and potentially greater speed.27 Although some of the time-consuming steps to conduct ELISA assays can be carried out in an automated system, parallel analysis of several analytes for the same sample and at the same time is still a challenge. Here we have demonstrated that HydroGel-coated slides perform very well immunoassays in a microarray format using fluorescent sandwich detection. The value of the miniaturized assay in multiplexed format is several fold. Traditional ELISA methods of detection may use 100 µL or more of sample to measure a single analyte. The demonstrated assay uses less sample volume (50 µL), and smaller gel pads may decrease the required volume further. In addition, the ability to measure multiple analytes from a single sample reduces the need to prepare samples from multiple sources of cell culture or experimental animals. Most importantly, measurements of analytes made in parallel maintain the expression relationships of the analytes within an experiment. The alternative to parallel analysis is to perform separate assays for each analyte of interest, which generally would require samples to be prepared from different sources making it difficult to conclude direct correlations in relative expression levels. The activities of many proteins are diverse such that it is common that a family of proteins participates in both divergent and overlapping pathways. This is especially true of the cytokines and chemokines to the extent that it is often difficult to classify individual members of this group by function. Cytokines are playing an ever-increasing role in the treatment of human disease and in drug discovery.28-30 Because of the large number of samples available and the variety of drug libraries requiring analysis, the tests used to characterize
cytokines must be appropriate for both high-throughput screening and relative quantitative analysis.
Acknowledgment. Our gratitude is extended to Q. Wang, F.-E. Wiedmer, and J. Horn for their contributions to HydroGel slide production and array printing, R. Fisler for the business development, F. Corden, K. Woodward, D. Shen, and F. Wang for valuable discussions, and F. Witney for his support of the project. References (1) Carding, S. R.; Hayday, A. C.; Bottomly, K. Immunol. Today 1991, 12 (7), 239-4. (2) Murphy, P. M. N. Engl. J. Med. 2001, 345 (11), 833-5. (3) Minasian, E.; Nicola, N. A. Protein Seq. Data Anal. 1992, 5 (1), 57-64. (4) Kelso, A. Immunol. Cell Biol. 1998, 76 (4), 300-17. (5) Haverty, A. A.; Harmey, J. H.; Redmond, H. P.; Bouchier-Hayes, D. J. Surg. Res. 1997, 69 (1), 145-9. (6) Begley, C. G.; Nicola, N. A. Blood 1999, 93 (5), 1443-7. (7) Kleine-Lowinski, K.; Gillitzer, R.; Kuhne-Heid, R.; Rosl, F. Int. J. Cancer 1999, 82 (1), 6-11. (8) Zhu, H.; Snyder, M. Curr. Opin. Chem. Biol. 2001, 5 (1), 40-5. (9) Vignali, D. A. J. Immunol. Methods 2000, 243 (1-2), 243-55. (10) Wiese, R.; Belosludtsev, Y.; Powdrill, T.; Thompson, P.; Hogan, M. Clin Chem 2001, 47 (8), 1451-7. (11) Mendoza, L. G.; McQuary, P.; Mongan, A.; Gangadharan, R.; Brignac, S.; Eggers, M. Biotechniques 1999, 27 (4), 778-80, 7826, 788. (12) Moody, M. D.; Van Arsdell, S. W.; Murphy, K. P.; Orencole, S. F.; Burns, C. Biotechniques 2001, 31 (1), 186-90, 192-4. (13) Schweitzer, B.; Wiltshire, S.; Lambert, J.; O’Malley, S.; Kukanskis, K.; Zhu, Z.; Kingsmore, S. F.; Lizardi, P. M.; Ward, D. C. Proc. Natl. Acad. Sci. U.S.A. 2000, 97 (18), 10113-9. (14) Haab, B. B.; Dunham, M. J.; Brown, P. O. Genome. Biol. 2001, 2 (2), RESEARCH0004. (15) Huang, R. P. J. Immunol. Methods 2001, 255 (1-2), 1-13. (16) Huang, R. P.; Huang, R.; Fan, Y.; Lin Y. Anal. Biochem. 2001, 294 (1), 55-62. (17) Carson, R. T.; Vignali, D. A. J. Immunol. Methods 1999, 227 (12), 41-52. (18) Broude, N. E.; Woodward, K.; Cavallo, R.; Cantor, C. R.; Englert, D. Nucleic Acids Res. 2001, 29 (19), E92. (19) Mirzabekov, A. Methods Mol. Biol. 2001, 170, 17-38. (20) Huang, R. P.; Fan, Y.; de Belle, I.; Niemeyer, C.; Gottardis, M. M.; Mercola, D.; Adamson, E. D. Int. J. Cancer 1997, 72 (1), 102-9. (21) Riethdorf, S.; Riethdorf, L.; Richter, N.; Loning, T. Pathobiology 1998, 66 (6), 260-7. (22) Bienvenu, J.; Monneret, G.; Fabien, N.; Revillard, J. P. Clin. Chem. Lab. Med. 2000, 38 (4), 267-85. (23) Kodadek, T. Chem. Biol. 2001, 8, 105-115. (24) Service, R. F. Science 2001, 294, 2080-2082. (25) MacBeath, G.; Schreiber, S. L. Science 2000, 289 (5485), 1760-3. (26) Stillman, B. A.; Tonkinson, J. L. Biotechniques 2000, 29 (3), 6305. (27) Ekins, R. P. J. Pharm. Biomed. Anal. 1989, 7 (2), 155-68. (28) Foster, J. R. Int. J. Exp. Path 2001, 82, 171-192. (29) House, R. V. Methods 1999, 19, 17-27. (30) Mire-Sluis, A. R. TIBTECH 1999, 17, 319-325.
PR0255203
Journal of Proteome Research • Vol. 1, No. 4, 2002 343
