Anal. Chem. 1999, 71, 3846-3852
Array Biosensor for Simultaneous Identification of Bacterial, Viral, and Protein Analytes Chris A. Rowe,† Leonard M. Tender,† Mark J. Feldstein,† Joel P. Golden,† Stephanie B. Scruggs,‡ Brian D. MacCraith,§ John J. Cras,§ and Frances S. Ligler*,†
Center for Bio/Molecular Science & Engineering, Code 6900, Naval Research Laboratory, Washington, D.C. 20375-5348, Geo-Centers, Inc., Rockville, Maryland 20852, and Physics Department, Dublin City University, Glasnevin, Dublin 9, Ireland
The array biosensor was fabricated to analyze multiple samples simultaneously for multiple analytes. The sensor utilized a standard sandwich immunoassay format: Antigen-specific “capture” antibodies were immobilized in a patterned array on the surface of a planar waveguide and bound analyte was subsequently detected using fluorescent tracer antibodies. This study describes the analysis of 126 blind samples for the presence of three distinct classes of analytes. To address potential complications arising from using a mixture of tracer antibodies in the multianalyte assay, three single-analyte assays were run in parallel with a multianalyte assay. Mixtures of analytes were also assayed to demonstrate the sensor’s ability to detect more than a single species at a time. The array sensor was capable of detecting viral, bacterial, and protein analytes using a facile 14-min assay with sensitivity levels approaching those of standard ELISA methods. Limits of detection for Bacillus globigii, MS2 bacteriophage, and staphylococcal enterotoxin B (SEB) were 105 cfu/mL, 107 pfu/mL, and 10 ng/mL, respectively. The array biosensor also analyzed multiple samples simultaneously and detected mixtures of the different types of analytes in the multianalyte format.
Immunosensors are small, portable instruments for analysis of complex fluids and are designed for ease of use by untrained personnel, rapid assay times, and sensitivity comparable to ELISA methods. In recent years, the fabrication of biosensors able to distinguish multiple analytes in a single sample has become an increasingly well recognized research goal.1 There are several recent reports describing multianalyte immunosensors. Many describe a method for multianalyte sensing but demonstrate the detection of only a single analyte.2-6 Others combine discrete * Corresponding author:
[email protected]. † Naval Research Laboratory. ‡ Geo-Centers Inc. § Dublin City University. (1) Ekins, R.; Chu, R. Clin. Chem. 1993, 39, 369-370. (2) Ekins, R.; Chu, F.; Biggart, E. Clin. Chim. Acta 1990, 194, 91-114. (3) Herron, J. N.; Christensen, D. A.; Wang, H.-K.; Caldwell, K. D.; Janatova, V.; Huang, S.-C. U.S. Patent 5,677,196, 1997. (4) Abel, A. P.; Weller, M. G.; Duveneck, G. L.; Ehrat, M.; Widmer, H. M. Anal. Chem. 1996, 68, 2905-12. (5) Blawas, A. S.; Oliver, T. F.; Pirrung, M. C.; Reichert, W. M. Langmuir 1998, 14, 4243-50.
3846 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
sensing substrates, such as optical fibers,7 capillaries,8 microtiter wells,9 or dipsticks,10 to analyze for multiple analytes simultaneously. Another general approach employs different capture molecules in different regions of a sensing substrate and detects binding using a label-free approach.11,12 These label-free methods continue to be susceptible to problems such as low sensitivity and increased backgrounds due to nonspecific binding. Wadkins et al.13 avoided such problems by using fluorescent tracer antibodies and performing a measurement insensitive to nonspecifically bound proteins (other than the tracer antibody). However, while this example used a single substrate with multiple capture spots, the individual tracer antibodies were added in sequential steps.13,14 Only three reports of multianalyte biosensors use a single sensor substrate and a single detection step employing one,15 two,16 or more17 tracer antibodies: all three investigations detected only protein analytes. The array biosensor is designed to test multiple samples for the presence of any of several analytes. The specific experiments presented here addressed the following issues: (1) Can samples be simultaneously tested for the presence of different classes of analytes? Most biosensors can detect protein analytes, but some are very insensitive to binding of bacteria. The reasons range from optical effects to problems with binding in flow systems. Very little work has been published on the recognition of viruses, possibly (6) Brecht, A.; Klotz, A.; Barzen, C.; Gauglitz, G.; Harris, R. D.; Quigley, G. R.; Wilkinson, J. S.; Sztajnbok, P.; Abuknesha, R.; Gasco´n, J.; Oubin ˜a, A.; Barcelo´, D. Anal. Chim. Acta 1998, 362, 69-79. (7) Bakaltcheva, I. B.; Shriver-Lake, L. C.; Ligler, F. S. Sens. Actuators, B 1998, 51, 46-51. (8) Narang, U.; Gauger, P. R.; Kusterbeck, A. W.; Ligler, F. S. Anal. Biochem. 1998, 255, 13-19. (9) Kakabakos, S. E.; Christopoulos, T. K.; Diamandis, E. P. Clin. Chem. 1992, 38, 338-342. (10) Parsons, R. G.; Kowal, R.; LeBlond, D.; Yue, V. T.; Neagarder, L.; Bond, L.; Garcia, D.; Slater, D.; Rogers, P. Clin. Chem. 1993, 39, 1899-1903. (11) Frebel, H.; Chemnitius, G.-C.; Cammann, K.; Kakerow, R.; Rospert, M.; Mokwa, W. Sens. Actuators, B 1997, 43, 87-93. (12) Berger, C. E. H.; Beumer, T. A. M.; Kooyman, R. P. H.; Greve, J. Anal. Chem. 1998, 70, 703-706. (13) Wadkins, R. M.; Golden, J. P.; Pritsiolas, L. M.; Ligler, F. S. Biosens. Bioelectron. 1998, 13, 407-15. (14) Wadkins, R. M.; Golden, J. P.; Ligler, F. S. J. Biomed. Opt. 1997, 2, 74-79. (15) Silzel, J. W.; Cercek, B.; Dodson, C.; Tsay, T.; Obremski, R. J. Clin. Chem. 1998, 44, 2036-2043. (16) Pritchard, D. J.; Morgan, H.; Cooper, J. M. Anal. Chim. Acta 1995, 310, 251-256. (17) Rowe, C. A.; Scruggs, S. B.; Feldstein, M. J.; Golden, J. P.; Ligler, F. S. Anal. Chem. 1999, 71, 433-439. 10.1021/ac981425v CCC: $18.00
© 1999 American Chemical Society Published on Web 07/24/1999
because there are few analyte-antibody pairs that are convenient to test. The work presented here examines the ability of the biosensor to detect all three types of analytes. (2) Can mixtures of fluorescent tracer antibodies be used effectively? Mixing polyclonal antibodies from different species has been documented to produce cross-species interactions in some cases and might also produce cross-reactive binding against the various complex analytes. Combining the three tracer antibodies also increases the total tracer concentration, potentially increasing the level of nonspecific tracer binding and background fluorescence. For this reason, the blind samples were analyzed on each waveguide with both the tracer mixture and the individual tracer solutions. (3) Can the automated software provide quantitative information on the analyte concentration and the level of analyte or antibody cross-reactivity? Automated software was developed for data analysis, and the blind trial data were reevaluated for quantitative information. (4) Can mixtures of analytes in a single sample be measured? All possible combinations of the three analytes were prepared and each mixture was evaluated. (5) Can multiple samples be analyzed simultaneously? The blind studies analyzed a standard mixture of the three analytes and a blind sample simultaneously on each waveguide. The mixture experiments analyzed four different samples simultaneously. The waveguide with its current flow cell has the capability of measuring six samples simultaneously, with the potential for increasing the number of channels. The array biosensor used in this study is relatively simple in its configuration. It consists of a thermoelectrically cooled CCD camera, a microscope slide (waveguide) coated with vertical columns of capture antibody, a diode laser with a line generator to launch light into the end of the waveguide, and a removable poly(dimethylsiloxane) (PDMS) flow cell for introduction of samples and fluorescent tracer antibody. The array biosensor utilizes evanescent wave excitation to interrogate patterns of fluoroimmunoassays immobilized on the planar waveguide. This system detects and measures analytes in buffer and in a number of physiological fluids17 and is relatively unaffected by nonspecifically bound components from complex samples. The immunoassays described herein demonstrate the ability of the array biosensor to detect and distinguish between different classes of analytes with a considerably shortened turnaround time (approximately 14 min) compared to previous studies.17 The array biosensor was able to identify samples containing bacterial spores, virus particles, or protein antigens in a statistically significant number of blind samples. Three well-characterized analytes were used in this study. Bacillus globigii is a nonpathogenic, Grampositive, sporulating soil bacterium with a worldwide distribution. The viral analyte, MS2, is a small, icosahedral RNA bacteriophage; both its nucleic acid sequence and crystal structure have been determined.18,19 Staphylococcal enterotoxin B (SEB) is one of seven structurally and biologically related superantigen toxins produced by Staphylococcus aureus and is a common cause of food poisoning. (18) Fiers W.; Contreras, R.; Duerinck, F.; Haegeman, G.; Iseretant, D.; Merregaert, J.; Min Jou, W.; Molemans, F.; Raeymaekers, A.; van der Berghe, A.; Volckaert, G.; Ysebaert, M. Nature 1976, 260, 273-80. (19) Valegard, K.; Liljas, L.; Fridgorg, K.; Unge, T. Nature 1990, 345, 36-41.
Table 1. Antibodies Used for Identification of Bacterial, Viral, and Protein Analytes analyte B. globigii array ELISA
capture antibody
biotin-rabbit Cy5-rabbit anti-B. globigii IgG anti-B.globigii IgG goat antirabbit anti-B. globigii IgG + B. globigii IgG HRP-anti-rabbit IgG
MS2 coliphage array biotin-rabbit anti-MS2 IgG ELISA rabbit anti-MS2 IgG SEB array ELISA
detection antibody
biotin-rabbit anti-SEB IgG goat anti-SEB IgG
Cy5-rabbit anti-MS2 IgG rabbit anti-MS2 IgG + HRP-anti-rabbit IgG Cy5-rabbit anti-SEB IgG monoclonal (SEB-03-B-2A) + HRP-anti-mouse IgG
EXPERIMENTAL SECTION Antibodies. All antibody preparations used in the analyses were generous gifts from Mr. T. O’Brien (Naval Medical Research Center (NMRC), Bethesda, MD) and were provided after purification with protein G or protein A chromatography. The antibody components of the sandwich immunoassays utilized in the array sensor and in the confirmatory ELISA are listed in Table 1. For use in the array biosensor, capture antibodies (Table 1) were biotinylated for attachment to the avidin-modified array sensor surface. The IgG preparations were treated for 30 min with a long-chain derivative of biotin N-hydroxysuccinimidyl ester (EZLink NHS-LC-biotin; Pierce, Rockford, IL) at a 5:1 (biotin:IgG) molar ratio in 50 mM borate, pH 9, containing 40 mM NaCl. Labeled protein was then separated from unincorporated biotin by chromatography on Bio-Gel P10 (Bio-Rad, Hercules, CA). Fluorescent tracer antibodies were labeled with the cyanine dye, Cy5 Bisfunctional Reactive Dye (λex ) 649 nm, λem ) 670 nm; Amersham Life Science Products, Arlington Heights, IL). Antibody labeling was carried out using 3 mg of antibody per vial of dye under the conditions as described above. Fluorescent protein and unincorporated dye were subsequently separated by chromatography on Bio-Gel P10. Dye-to-protein ratios (mole/mole) ranged from 1.5 to 2.5. Blind Samples. A total of 126 blind samples were analyzed using the array sensor. Samples contained one of the analytes described in the introduction, diluted in phosphate-buffered saline (PBS) containing 0.05% Tween 20 (PBST). Tween 20 was included in all buffers, as this detergent has been shown to reduce binding of MS2 coliphage to various matrixes by hydrophobic interactions.20 Sample concentration ranges were chosen to include several below expected detection limits to determine the sensitivity of the immunoassay to the three analytes and very high concentrations to test for nonspecific binding and cross-reactivity. Concentrations for spiked samples ranged from 103 to 108 colonyforming units per milliliter for B. globigii (cfu/mL, 108 cfu ) 430 µg/mL), from 104 to 109 plaque-forming units per milliliter for MS2 (109 pfu/mL ) 43 µg/mL), and from 10 pg/mL to 10 µg/mL for SEB. Six replicate samples were provided for each concentration of analyte; 18 blanks (unspiked samples) were also included for (20) Lytle, C. D.; Routson, L. B. Appl. Environ. Microbiol. 1995, 61, 643-49.
Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
3847
analysis. The composition of each sample was verified using a sandwich ELISA according to standard protocols. Mixed Analytes. For proof-of-concept of simultaneous detection of mixed analytes in a single sample, a series of mixtures was prepared with analytes present in various combinations. The following mixtures were prepared in PBST: 100 ng/mL SEB, 4 × 108 pfu/mL MS2, 2.3 × 105 cfu/mL B. globigii (mix 1); 4 × 108 pfu/mL MS2, 2.3 × 105 cfu/mL B. globigii (mix 2); 100 ng/mL SEB, 2.3 × 105 cfu/mL B. globigii (mix 3); 100 ng/mL SEB, 4 × 108 pfu/mL MS2 (mix 4). Preparation of Arrays. Microscope slides were utilized as waveguides upon which the sandwich fluoroimmunoassays were performed. Glass microscope slides (DaiggerBrand, Daigger, Wheeling, IL) were cleaned by immersion for 30 min in 1 M NaOH, a rinse in deionized water, and treatment with “Piranha” solution (2:1 H2SO4/H2O2), according to standard protocols.21 After exhaustive washing and silane treatment of the slides, NeutrAvidin biotin-binding protein (Pierce) was covalently immobilized onto the slides via the heterobifunctional cross-linker N-succinimidyl4-maleimidobutyrate.17,22,23 This modified form of avidin was found to have reduced nonspecific binding compared to other forms of avidin. Following three rinses to remove unbound NeutrAvidin, the coated slides were stored in PBS containing 0.03% sodium azide at 4 °C until patterned with capture antibodies. Biotinylated capture antibodies were immobilized onto NeutrAvidin-modified waveguides using PDMS flow cells pressed against the waveguide surfaces with a manual chuck.17 Each vertically oriented flow cell channel was filled with a solution containing 20 µg/mL biotinylated anti-SEB, anti-MS2, or anti-B. globigii in PBST and was incubated for 8 h at 4 °C on a rocker. The three capture antibodies were each patterned in two columns to provide duplicate assays (Figure 1). The flow channels were subsequently emptied and rinsed with 1 mL of PBST containing 1 mg/mL bovine serum albumin (PBSTB). Upon removal of the patterning flow guides, antibody-patterned waveguides were rinsed with PBSTB, incubated in storage buffer (50 mM sodium phosphate pH 7.4, 10 mg/mL bovine serum albumin) for 20 min, dried under vacuum, and stored in a desiccator until use. Assay Protocol. A second PDMS flow cell was placed onto the surface of each patterned slide (prewetted) and formed a tight seal as pressure was applied. The channels in the immunoassay flow cell were oriented perpendicular to the stripes of immobilized antibody such that each assay channel crossed the six stripes of immobilized capture antibodies17 (Figure 1). Syringe needles were inserted into both ends of each channel and attached to either a multichannel peristaltic pump (outlet) or a fluid reservoir (inlet). Solutions loaded into the reservoir (a 1-cm3 syringe) were pulled through the channels by the peristaltic pump. Samples were analyzed using the following sandwich immunoassay format: (step I) 1.2-min wash with PBSTB (1 mL) at a flow rate of 0.8 mL/min; (step II) 7-min incubation with sample (0.6 mL) at a flow rate of 0.085 mL/min; (step III) 1.2-min rinse with PBSTB (1.ml) at a flow rate of 0.8 mL/min; (step IV) 3-min (21) Jonsson, U.; Malmquist, M.; Olofsson, G.; Ronnberg, I. Methods Enzymol. 1988, 137, 381-388. (22) Bhatia, S. K.; Shriver-Lake, L. C.; Prior, K. J.; Georger, J. H.; Calvert, J. M.; Bredehorst, R.; Ligler, F. S. Anal. Biochem. 1989, 178, 408-413. (23) Ligler, F. S.; Calvert, J.; Georger, J.; Shriver-Lake, L. S.; Bhatia, S. U.S. Patent 5,077,210, 1991.
3848 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
Figure 1. Idealized pattern of fluorescence for three analytes tested. Vertically oriented stripes of biotinylated capture antibodies were patterned onto NeutrAvidin-coated microscope slides as indicated above the figure. Samples and tracer antibodies were applied using PDMS flow chambers with horizontally oriented channels. Sample assayed in lane 1 (positive control) contained 2.3 × 105 cfu/mL B. globigii, 4.6 × 105 pfu/mL MS2, and 10 ng/mL SEB; this lane was interrogated using a cocktail containing 20 µg/mL each of Cy5-labeled anti-B. globigii, anti-MS2, and anti-SEB IgGs. Unknown sample was applied to lanes 2-5; these lanes were interrogated using either the tracer antibody cocktail (above, lane 2) or 20 µg/mL individual antigenspecific tracer antibody (lanes 3-5, indicated).
incubation with tracer antibody (0.4 mL) at a flow rate of 0.085 mL/min; (step V) 1.2-min wash with PBSTB (1 mL) at a flow rate of 0.8 mL/min. Following the final rinse, the channels were emptied and the PDMS flow cell template was removed. The slides were rinsed with deionized water and dried under nitrogen before imaging with the CCD-based optical readout device. To validate the use of antibody cocktails as the tracer reagent in multianalyte assays, four immunoassays were performed simultaneously on each waveguide for each blind sample: a multianalyte assay (channel 2) and three single-analyte assays (channels 3-5) as shown in Figure 1. The multianalyte assay utilized a cocktail of 20 µg/mL each of Cy5-labeled anti-SEB, antiMS2, and anti-B. globigii tracer antibodies to interrogate channel 2. Assays specific for single analytes were performed in channels 3-5 using 20 µg/mL each of the individual tracer antibody. As the multianalyte assay and assays specific for individual analytes were performed in parallel and on the same waveguide, results could be directly compared. In a separate study, analyses of analyte mixtures were performed using the tracer cocktail only. Imaging Optics. Patterns of fluorescence on the sensor surface were interrogated using a CCD-based optical readout device. Light was launched from a 635-nm, 12-mW diode laser (Lasermax, Rochester, NY) into the end of the waveguide, resulting in evanescent excitation of surface-bound fluorophores. A series of lenses was utilized to provide uniform excitation in the sensing area of the waveguide as discussed elsewhere.24 A two-dimensional array of graded index (GRIN) lenses25 (Nippon Sheet Glass, Sumerset, NJ) imaged the fluorescent array in the sensing region of the waveguide onto a thermoelectrically cooled (24) Feldstein, M. J.; Golden, J. P.; Rowe, C. A.; MacCraith, B. D.; Ligler, F. S. J. Biomed. Microdevices, 1999, 1, 139-153. (25) Golden, J. P. U.S. Patent 5,827,748, 1998.
CCD array (Spectra Source, Westlake Village, CA). Two longpass filters (OG-0665, Schott Glass, Duryea, PA) and one bandpass filter (S40-670-S, Corion, Holliston MA) were installed to eliminate excitation and scattered light prior to imaging.17,24 Due to the wide range of analyte concentrations encountered in this study, exposure times on the CCD camera were varied according to the fluorescence intensity of the signal generated at each array element. Twelve seconds was the standard exposure time used for imaging most samples. However, for samples with strong signals, times as short as 2 s were used, and for weak signals, images were collected for as long as 24 s. It was experimentally determined that fluorescent signals from the array elements increased in proportion to increasing exposure time (data not shown). It was therefore possible to compare fluorescent signals directly from samples imaged for different exposure times. The fluorescent images of blind samples were visually scrutinized for a fluorescent pattern consistent with the presence of a single analyte; conclusions of positive detection were based on this visual inspection. Limits of detection were defined as the concentration at which 100% of the samples were correctly identified. The images were also retrospectively analyzed using data acquisition software developed at NRL.24 Fluorescence intensities for each square and adjacent background intensities were extracted from each image. The mean fluorescent signal (MFS) for each spot was calculated by subtracting the background pixel intensities on each side of the square from the mean pixel intensity within the square.24 For comparison of MFS obtained at different analyte concentrations, statistical analyses were performed using a two-sample unpaired Student’s t-test, assuming unequal sample sizes and variances. Comparisons between lanes (e.g., tracer antibody cocktail lane versus individual antibody lanes) on the same waveguide were performed using the paired Student’s t-test. Safety Considerations. Cleaning of slides was performed in a chemical hood by personnel wearing acid-resistant gloves and appropriate personal protective gear. Silanization procedures utilizing (3-mercaptopropyl)trimethoxysilane were carried out in a glovebag placed in a chemical hood. All solutions containing analyte were handled by personnel wearing gloves and appropriate personal protective gear (lab coat, safety goggles). All equipment, benchtops, etc., exposed to these solutions were disinfected with a 20% bleach solution and were rinsed with distilled water. Silicone tubing attached to peristaltic pumps (through which samples were pumped) and PDMS flow cell modules were rinsed with 20% bleach and distilled water after each experiment. Analyte solutions were also treated with bleach (final concentration of 20%) and were rinsed down the sink with excess water. Contaminated disposables (test tubes, pipet tips, used sensor substrates) were placed in biohazard bags and later incinerated. RESULTS AND DISCUSSION Detection of Bacterial, Viral, and Protein Analytes. Actual patterns of fluorescence for blind samples containing B. globigii, MS2, and SEB are shown in Figure 2. Fluorescence was observed in the antigen-specific spots, provided that a complementary tracer antibody was utilized, either alone (lanes 3-5) or as part of the antibody mixture (lane 2). The array biosensor was able to identify the analytes accurately in the process of performing simultaneous assays for a bacterium, a virus, and a toxin: i.e., row 2 always
showed the same result as rows 3-5. For the identifications performed using the tracer antibody mixtures, limits of detection for the array biosensor were similar or identical to those obtained with confirmatory ELISAs (Table 2). Table 2 shows the percentage of correctly identified samples as a function of antigen concentration. The limit of detection for B. globigii was 105 cfu/mL and for MS2 was 107 pfu/mL, identical to that obtained with the confirmatory ELISAs. The limit of detection for SEB determined using the array biosensor was 10 ng/mL, 1 order of magnitude above that of ELISA. The differences in affinities between the detection antibodies used in the ELISA method (mouse monoclonal) and those used in the array biosensor (rabbit polyclonal) were potentially responsible for this result. All of the samples at concentrations above the limit of detection (60 of 126) were correctly identified as positives, yielding 0% false negatives. One sample below the detection limit gave a false positive for a different agent, yielding 0.8% false positives. The ELISA also exhibited a 0.8% false positive ratio. Automated Image Analysis. Following the development of software for automated image analysis, the data from the 126 blind samples were reevaluated. Mean fluorescent signals were determined for each spot in the array for all samples. After corrections were made for exposure time (using 12 s as the standard exposure), these values were plotted as a function of antigen concentration (Figure 3). A significant increase in MFS above background values (p < 0.05) was observed at 105 cfu/mL B. globigii (panel A), at 107 pfu/mL MS2 (panel B), and at 10 ng/ mL SEB (panel C). The fluorescent signals at those concentrations could be clearly observed by the researchers and corresponded to the limits of detection originally obtained by visual inspection of the images. Although the MFS increased at concentrations above the limits of detection, these increases were not linearly proportional to concentration: i.e., 10-fold increases in MFS were not observed for each order of magnitude increase in concentration. This nonlinearity may potentially be due to saturation of higher affinity recognition molecules attached to the sensor surface, especially at high antigen concentrations. The multianalyte assay in lane 2 of the blind analyses utilized a mixture of tracer antibodies whose final IgG concentration was 3 times higher than each of the individual antibodies used in the parallel assays (lanes 3-5). To quantify the effect of the cocktail, fluorescence intensities for background regions (in the channel between spots), antigen-specific spots, and irrelevant spots (sites of nonspecific binding of analyte to antibodies other than the analyte-specific antibody) were compared between the cocktail lane (lane 2) and the single antibody lane (lane 3, 4, or 5), using a paired Student’s t-test. In blank samples (containing no analyte), the lane interrogated with tracer cocktail showed significantly higher backgrounds and nonspecific binding than lanes interrogated with individual antibodies (p < 0.05). The difference in background was greatest when comparing the cocktail lane (lane 2) and the lane interrogated with anti-SEB IgG (lane 5, p < 0.001). This difference may be due to some nonuniformity in illumination from the diode laser: background intensities from the lower half of the control slides were consistently lower than those from the upper half of these slides (data not shown). Although the blank samples showed a significant difference in background and nonspecific binding between the tracer cocktail Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
3849
Figure 2. Comparison of idealized and real fluorescence images for single-analyte assays. Shown are images of samples containing the highest analyte concentration tested (10 µg/mL SEB, 109 pfu/mL MS2, 108 cfu/mL B. globigii (Bg)) and samples containing analyte at the limit of detection (10 ng/mL SEB, 107 pfu/mL MS2, 105 cfu/mL Bg). Immobilized capture antibodies were patterned and assays performed as indicated by Figure 1. The multianalyte assay (lane 2) performed equally well, with respect to analyte specificity and sensitivity, as the corresponding three single-analyte assays (lanes 3-5). In each case, the two appropriate analyte-specific elements of lane 2 fluoresce, as do the two appropriate elements in lanes 3, 4, or 5. Table 2. Percent Positive Assays for the Array Biosensor and ELISA at Each Antigen Concentration MS2
B. globigii conc (cfu/mL) 108 107 106 105 104 103
array 100 100 100 100 8.3 0
SEB
ELISA
conc (pfu/mL)
array
ELISA
conc (ng/mL)
array
ELISA
100 100 100 100 50 0
109
100 100 100 17 0 0
100 100 100 67 0 0
10000 100 10 1 0.1 0.01
100 100 100 17 17 0
100 100 100 100 0 0
108 107 106 105 104
lane and the individual tracer lanes, this trend was not consistent when samples containing analytes were analyzed. This inconsistency may be due to antigen carry-over between samples, increased background and signal at the irrelevant spots due to nonspecific binding of analytes (see example of MS2, below), and the number of samples at each concentration (6 samples at each analyte concentration versus 18 blank samples). Comparison of the MFS for the cocktail lane and the parallel individual antibody lanes supported the observation that the cocktail lane and the appropriate individual antibody lane always showed the same result. At analyte concentrations above the stated detection limits, the MFS in the cocktail lane were generally the 3850 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
same as those of the corresponding parallel assay (p < 0.05) with three exceptions. Only one pair of analyte-specific spots (10 ng/ mL SEB) gave a higher signal in the cocktail lane than in the single antibody lane (p < 0.025). Two samples showed a higher nonspecific binding to the anti-MS2 spot in the cocktail lane versus the appropriate individual antibody lane (10 µg/mL SEB and 105 cfu/mL B. globigii, p < 0.01). In all of the other 27 cases, the MFS for specific and nonspecific binding in the cocktail lane were not significantly different from those in the appropriate individual assay (p < 0.05). These results confirm that an antibody cocktail may be used in place of individual antibodies for quantitation as well as for multianalyte detection and quantification.
Figure 4. Proof-of-concept for detecting mixtures of analytes. Different mixtures of the three different types of analytes flowed through each lane, followed by a cocktail of tracer antibodies, using the protocol described in the Experimental Section. Analyte concentrations were approximately 10-fold higher than the detection limit. Lane 1 contained SEB, MS2, and B. globigii (mix 1; see Experimental Section), lane 2 contained MS2 and B. globigii (mix 2), lane 3 contained SEB and B. globigii (mix 3), and lane 4 contained SEB and MS2 (mix 4).
Figure 3. Detection of analytes. Mean fluorescent signals versus antigen concentration. MFS were calculated by subtracting background intensities in the areas adjacent to each spot from the fluorescent intensities within the spots. These values were then corrected for exposure time (normalized to the standard 12-s imaging time) and plotted as a function of antigen concentration. Values shown are mean of the MFS for six replicate samples at each concentration (SEM. Panel A shows data for SEB samples, Panel B for MS2, and Panel C for B. globigii (Bg). The data shown were obtained from lane 2 using mixtures of tracer antibodies.
The data from the cocktail lane (lane 2) can be used to characterize the levels of antibody cross-reactivity (Figure 3). Panels A and C show very little fluorescence in the spots other than those relevant for the identification of SEB and B. globigii, respectively. Panel B, however, shows that there is significant signal generated on the anti-SEB and anti-B. globigii spots in the presence of MS2. Examination of the images from the singleanalyte assays shows that the irrelevant spots are fluorescent only in the presence of anti-MS2 tracer antibody. Thus it appears that the MS2 itself sticks nonspecifically to any capture IgG as opposed to MS2 cross-reactivity in the anti-SEB and anti-B. globigii antibodies. At concentrations above 107 pfu/mL, the MFS for the MS2-specific spots were significantly greater than the MFS for
the irrelevant spots in both the cocktail lanes and the anti-MS2 lane (p < 0.025). Thus there was no confusion as to the identity of the analyte in the sample. In the future, the amount of nonspecific binding to expect relative to the amount of specific binding for MS2 could be factored into a data analysis program to simplify the identification of mixed analytes. Analysis of Analyte Mixtures. Figure 4 illustrates the proofof-concept of simultaneous identification of analytes in mixtures using the multianalyte assay. Samples containing the four possible mixtures of the three analytes were assayed simultaneously using a mixture of the three tracer antibodies and the same 14-min assay protocol. The correlation between the idealized waveguide image and the real waveguide image demonstrates the utility of the multianalyte assay. The automated data analysis showed MFS values above the threshold limit (set at 2500 MFS units) only for the appropriate spots, even though low levels of nonspecific fluorescence were visible to the eye. Only one of the values for MS2 (spot 1C, Figure 4) was below the threshold limit, but the other MS2-specific spot in the same lane had a MFS above the threshold limit. This finding further reinforces the value of being able to perform multiple assays on a single waveguide. CONCLUSIONS The array biosensor demonstrated the capability of performing simultaneous assays for very different types of analytes on a single substrate. To our knowledge, this is the first report of a sensor that can simultaneously measure three such diverse analytes. Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
3851
Furthermore, the biosensor performed the assays quickly and in a format that is amenable to automation and portability. Simultaneous immunoassays performed on the array biosensor correctly detected and identified bacterial, viral, and protein analytes in 126 blind samples, each containing a single analyte. Limits of detection and sensitivities were similar to those obtained with confirmatory ELISAs but with significantly shorter assay times (14 min versus several hours). Mixing the three fluorescent tracer antibodies in a single assay provided data identical to those obtained by running three individual assays each with a single tracer. While the images provided correct identification when perused by eye, automated image analysis provided reliable quantitation and better discrimination between signal and fluorescence due to nonspecific binding of the analyte to the “wrong” capture antibody. In addition to identifying single analytes in blind samples, the array biosensor was capable of discriminating each analyte in mixtures of two or three analytes. Furthermore, multiple samples could be analyzed simultaneously for multiple analytes. These experiments exemplify a significant reduction in the complexity of performing multianalyte assays. The planar waveguide provides a vehicle not only for simultaneously assaying a sample for multiple analytes but also for assaying multiple samples in parallel. Combining several analyte-specific reagents into a single solution reduces both the assay time and complexity of the fluidics
3852 Analytical Chemistry, Vol. 71, No. 17, September 1, 1999
(compared to sequential addition of the reagents). This approach may be scaled up to accommodate a potentially unlimited number of analytes for which antibodies can be generated. Cross-reactivity can be characterized and quantified. Simple pattern recognition can be used to sort out mixtures of analytes. The image analysis has been successfully automated and will be followed by the automation of the fluidics for increased ease of use and reproducibility of quantitative determinations. ACKNOWLEDGMENT The authors thank T. O’Brien and J. Aldrich at NMRC for the purified antigens and antibodies and for their invaluable assistance during the blind trials. This work was supported by the Office of Naval Research and the Department of Defense. C.A.R. and M.J.F. were supported by postdoctoral fellowships from the American Society for Engineering Education and the National Research Council, respectively. The views expressed here are those of the authors, and do not represent those of the U.S. Navy, the U.S. Department of Defense, or the U.S. Government.
Received for review December 29, 1998. Accepted June 14, 1999. AC981425V