Anal. Chem. 2000, 72, 2371-2376
Heterogeneous Immunosensing Using Antigen and Antibody Monolayers on Gold Surfaces with Electrochemical and Scanning Probe Detection Yongzhi Dong and Curtis Shannon*
Department of Chemistry, Auburn University, Auburn, Alabama 36849-5312
We report the use of antibody and antigen monolayer immunosurfaces as detection elements in a competitive heterogeneous immunoassay employing either electrochemical or scanning probe detection. Antibody or antigen monolayers were prepared by covalent attachment of the desired immunoreagent to a two-component self-assembled monolayer via amide linkages. More specifically, mixed monolayers of a carboxylic acid-terminated thiol (thioctic acid) and a methyl-terminated thiol (butanethiol) were used to control the surface epitope density. The microscopic structure of the resulting antibody and antigen arrays was characterized by AFM (atomic force microscopy). Individual, surface-confined rabbit IgG antibodies could be directly imaged in contact mode. The average height of the capture antibodies was found to be 7.1 nm; the average antibody diameter, after correcting for tip convolution effects, was determined to be between 7 and 10 nm. The surface epitope density could be varied over approximately 2 orders of magnitude by changing the composition of the mixed monolayer. AFM was also used to characterize the antibody-antigen binding characteristics of these immunosurfaces, and an average binding efficiency of 22.8% was measured for rabbit IgG antibody arrays. In the second part of this study, the electrochemical detection scheme originally developed by Heineman and co-workers was adapted to our system. A calibration data set was measured, and the linear least-squares correlation coefficient (R2) was found to be 0.993. Finally, the electrochemical and scanning probe detection modes were directly compared. We find an excellent correlation between the surface density of antibody-antigen complexes measured by AFM and the electrochemical response of the same immunosurfaces. Due to the highly selective molecular recognition afforded by the immune system, assays involving antibody-antigen reactions are commonly used in chemical, biological, and environmental analysis.1 As the drive to detect such materials at trace levels has increased, enzyme immunoassays (EIA) for a wide variety of (1) (a) Rogers, K. R. Biosens. Bioelectron. 1995, 10, 533-541. (b) Hadas E.; Soussan, L.; Rosen-Margalit, I.; Farkash, A.; Rishpon, J. J. Immunoassay 1992, 13, 231-252. (c) Wittsock, G.; Jenkins, S. H.; Halsall, H. B.; Heineman, W. R. Nanobiology 1998 4, 153-162. (d) Hage, David S. Anal. Chem. 1999, 71, 294R-304R. 10.1021/ac991450g CCC: $19.00 Published on Web 04/29/2000
© 2000 American Chemical Society
materials have been developed for use in clinical, pharmaceutical, and industrial settings. More recently, there has been a shift toward adapting immunoassay concepts to immunosensors that are separation-free and miniaturizable and in which samplehandling steps are kept to a minimum or eliminated entirely.2 To this end, significant effort has been directed toward (1) the development of synthetic strategies that allow receptor molecules and other necessary reagents to be immobilized on a solid substrate in a reproducible and well-defined manner and (2) the development of robust signal transduction schemes that do not compromise the inherent selectivity of EIA. In designing a surface attachment scheme, a balance must be struck between maximizing surface epitope density on one hand, and ensuring that receptor molecules are properly oriented and relatively free of steric constraints and overcrowding on the other.3 Surface attachment schemes can be biologically inspired, as in the case of biotin-avidin binding,4 or chemically inspired, as in schemes based on molecular self-assembly techniques,5 nonspecific interactions with glass or polymer surfaces,6 and electrostatic binding to monolayers or polymers.7 In addition to the direct covalent attachment of intact immunoreagents, it is also possible to adsorb antibody fragments (i.e., Fab′ fragments) to surfaces.8 A wide variety of signal transduction mechanisms have been investigated and developed for immunosensing applications, including electrochemical (amperometric), optical (fluorescence and surface plasmon resonance), radiochemical, capacitive, and piezoelectric methods.9 In addition to these more conventional techniques, scanning probe microscopy, particularly AFM (atomic (2) (a) Keay, R. W.; McNeil, C. J. Biosens. Bioelectron. 1998, 13, 963-970. (b) Ivnitski, D.; Rishpon, J. Biosens. Bioelectron. 1996, 11, 409-417. (c) Duan, C.; Meyerhoff, M. E. Anal. Chem. 1994, 66, 1369-1377. (d) Blonder, R.; Levi, S.; Tao, G.; Ben-Dov, I.; Willner, I. J. Am. Chem. Soc. 1997, 119, 10467-10478. (3) Kro¨ger, D.; Liley, M.; Schiweck, W.; Skerra, A.; Vogel, H. Biosens. Bioelectron. 1994, 14, 155-161. (4) Pirrung, M. C.; Huang, C.-y.; Bioconjugate Chem. 1996, 7, 317-321. (5) (a) Jones, V. W.; Kenseth, J. R.; Porter, M. D.; Mosher, C. L.; Henderson, E. Anal. Chem. 1998, 70, 1233-1241. (b) Disley, D. M.; Cullen, D. C.; You, H.-X.; Lowe, C. R. Biosens. Bioelectron. 1998, 13, 1213-1225. (c) Rickert, J.; Weiss, T.; Goepel, W. Sens. Actuators, B 1996, B31, 45-50. (6) Bourdillon, C.; Demaille, C.; Moiroux, J.; Save´ant, J.-M. J. Am. Chem. Soc. 1999, 121, 2401-2408. (7) Wybourne, M. N.; Yan, M.; Keana, J. F. W.; Wu, J. C. Nanotechnology 1996, 7, 302-5. (8) Vikholm, I.; Albers, W. M.; Valimaki, H.; Helle, H. Thin Solid Films 1998, 327/328, 643-646. (9) Pierrat, O.; Lechat, N.; Bourdillon, C.; Laval, J.-M. Langmuir 1997, 13, 4112-4118.
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000 2371
force microscopy), has been studied as a possible detector element in microminiaturized immunosensors. For example, Porter and co-workers have shown that the ability of the AFM to sensitively detect small height changes can be exploited to monitor antibodyantigen binding in compositionally patterned antibody arrays.4 In addition to the possibility of using a modified scanning probe device as a detection element, however, scanning probe microscopy also provides a range of fundamental information on the structure of immunointerfaces, allowing these systems to be optimized in a much more systematic way than previously possible. In this paper, we describe the use of antibody and antigen monolayer immunosurfaces as detection elements in a competitive heterogeneous immunoassay with electrochemical or scanning probe detection. The antibody and antigen monolayers were prepared by direct covalent attachment of an immunoreagent to a two-component self-assembled monolayer via amide linkages. To precisely control the surface epitope density, we used mixed monolayers of a chemically reactive, carboxylic acid-terminated thiol (thioctic acid, TA) and a chemically inert, methyl-terminated thiol (butanethiol, BT). We have recently employed a similar approach to create nanometer-scale patterns on surfaces in a massively parallel fashion.10 In the chemical functionalization step, antibodies or antigens are covalently linked to the active (TA) regions of the surface, resulting in the formation of antibody or antigen arrays that are suitable for EIA applications. In the second part of this study, signal transduction by both electrochemistry and AFM was carried out and the two methods were compared. EXPERIMENTAL SECTION Chemicals. Alkaline phosphatase conjugated monoclonal antirabbit IgG (γ-chain-specific), anti-rabbit IgG, alkaline phosphatase conjugated anti-sheep IgG, and rabbit IgG were obtained from Sigma Chemical Co. and used as received. Butanethiol, thioctic acid, 4-aminophenol, 4-nitrophenyl phosphate, 1-ethyl-3-(3-dimethyaminopropyl)carbodiimide hydrochloride (EDC), and Tween-80 were obtained from Aldrich Chemical Co. and used as received. 4-Aminophenol phosphate was prepared from 4-nitrophenyl phosphate as reported in the literature.11 Substrate preparation. Au wire (0.762 mm diameter, 99.999% purity, Alfa) was cleaned by being rinsed sequentially with distilled water, absolute ethanol, piranha solution, distilled water, and absolute ethanol. After being dried in a stream of flowing nitrogen, the clean Au wire was melted in a H2/O2 flame to form a 1.5-2.5 mm diameter ball at the end of the wire. The ball was repeatedly heated and cooled by adjusting its position in the flame until numerous elliptical facets (long axis approximately 200 µm) appeared on the surface of the ball. These Au(111) facets are composed of atomically flat terraces up to 1 µm wide. Prior to characterization or further chemical treatment, the Au ball was annealed in a H2 flame for an additional 15 min. Covalent Immobilization of Immunoreagents. A freshly prepared and cleaned Au ball was soaked in an ethanolic solution of butanethiol (control) or thioctic acid and butanethiol for 24 h to form a self-assembled monolayer, which was used as a platform (10) Hayes, W. A.; Kim, H.; Yue, X.; Perry, S. S.; Shannon, C. Langmuir 1997, 13, 2511-2518. (11) Pierrat, O.; Lechat, N.; Bourdillon, C.; Laval, J.-M. Langmuir 1997, 13, 4112-4118.
2372
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
for further chemical modifications of the surface. In the twocomponent systems, the total analytical concentration of thiol was 1 mM. After the self-assembly step, the specimen was emersed from solution, thoroughly rinsed with absolute ethanol, and dried in a stream of flowing nitrogen. Next, the specimen was immersed in a 1 mM solution of EDC in acetonitrile for 5 h. After being emersed from the EDC solution, rinsed thoroughly with acetonitrile, and dried in nitrogen, the Au ball was soaked in a 5 µg/ mL antibody (or antigen) solution at 4 °C for 24 h. The electrolyte used was 0.135 M NaCl and 1% (v/v) Tween-80. Finally, the modified Au substrate was soaked in a 5% (v/v) ethaneamine solution (in 0.1 M borate buffer) to react with any unreacted surface carboxyl groups, rinsed with copious amounts of distilled water, allowed to air-dry, and imaged by AFM. Immunoassay Procedure. To carry out the electrochemical immunoassay, a rabbit IgG antigen immunosurface was prepared as described above. This specimen was then transferred into a solution containing a mixture of anti-rabbit IgG and alkaline phosphatase labeled anti-rabbit IgG in which the molar ratios of labeled to unlabeled antibody ranged from 0 to 1. In all cases, the total antibody concentration was 5 µg/mL, and the buffer was 0.135 M NaCl + 1% (v/v) Tween-80. The samples were refrigerated for 5 h, thoroughly rinsed with distilled water, and allowed to air-dry prior to AFM analysis. After imaging, an electrochemical assay was carried out using the procedure first described by Heineman.12 Briefly, the substrate was transferred into a smallvolume electrochemical cell containing 200 µL of a solution that was 2 mM 4-aminophenol phosphate, 0.05 M carbonate buffer (pH 9.6), and 1 mM MgCl2. The alkaline phosphatase label catalyzes the conversion of 4-aminophenol phosphate to 4-aminophenol, which is easily detected by cyclic voltammetry. Peak oxidative currents were recorded as a function of the ratio of labeled to unlabeled anti-rabbit IgG to construct the calibration data set. AFM Imaging. All AFM experiments were performed on a Park Scientific Instruments Autoprobe CP scanning probe microscope in contact mode. The tips used were obtained from Digital Instruments Co. with a force constant of 0.12 N/m. The samples were scanned at approximately 2 Hz on a 5 µm piezoelectric scanner. The scan sizes and other experimental details are provided in the figure captions. Electrochemistry. All cyclic voltammetry experiments were performed in a homemade cell with a total volume of 400 µL. A standard three electrode configuration was used in which the Au ball was the working electrode, a Pt wire was the counter electrode, and Ag/AgCl was the reference electrode. All potentials are referenced to Ag/AgCl. The electrochemical cell was completely masked with aluminum foil to isolate the system from light. Before and after sample transfer, the solution was thoroughly degassed with Ar and then isolated from air with a blanket of Ar. RESULTS AND DISCUSSION Preparation and Characterization of Surface-Confined Antibody Monolayers. The chemistry used to covalently link immunoreagents to Au substrates is summarized in Scheme 1.13 Briefly, a mixed monolayer of BT and TA is self-assembled onto a clean Au substrate for 24 h. Following the self-assembly step, (12) Kaneki, N.; Xu, Y.; Kumari, A.; Halsall, H. B.; Heineman, W. R. Anal. Chim. Acta 1994, 287, 253-258. (13) Duan, C.; Meyerhoff, M. E. Mikrochim. Acta 1995, 117, 195-206.
Scheme 1. Formation of Antibody Monolayers on Au Surfaces Using Self-Assembly and Surface Derivitization Stepsa
a Antigen arrays are prepared in an analogous fashion. Details are given in the Experimental Section.
the carboxylic acid end groups of adsorbed TA are chemically activated by treating the monolayer with EDC (1-ethyl-3-(3dimethyaminopropyl)carbodiimide hydrochloride), which promotes the formation of amide bonds between adsorbed TA and the exposed lysine residues on the surface of the IgG (antibody or antigen) that is to be immobilized. This procedure allows precise control of the surface epitope density because the surface concentration of TA can be adjusted by varying the [BT]/[TA] ratio in the soaking solution. Contact mode AFM images of arrays of anti-rabbit IgG antibodies on two different TA/BT mixed monolayers are shown in Figure 1A,B. Both of these images are characterized by an atomically smooth surface covered with approximately spherical features that average 7.1 nm in height and 52.4 nm in diameter. The overall diameter of IgG antibodies has been estimated from X-ray experiments to be ca. 10 nm.14 The height of the features measured by AFM is in excellent agreement with this estimate, as well as with previous AFM measurements of adsorbed IgG.15 On the other hand, because of tip convolution effects, the apparent diameter of the features is significantly greater than what would be expected assuming an approximately spherical tertiary structure for the antibody. Contact mode AFM overestimates the lateral dimension of features that are similar in size to the tip. Several methods have been proposed for taking these tip convolution effects into account.16 When the features are not in direct contact with each other, as is the case here, the correction is particularly straightforward. In our experiments, the radius of curvature of (14) Silverton, J. W.; Navia, M. A.; Davies, D. R. Proc. Natl. Acad. Sci. U.S.A. 1977, 74, 5140-5144. (15) Feng, C.-d.; Yu, D. M.; Hesketh, P. J.; Gendel, S. M.; Stetter, J. R. Sens. Actuators B 1996, 35, 431-434. (16) Butt, H.-J.; Gerharz, B. Langmuir 1995, 11, 4735-4741.
the tip is larger than the lateral dimension of the feature being measured,17 and the apparent feature diameter is given by eq 1.18
dAFM ) 4(rtiprfeature)1/2
(1)
In this equation rtip is the tip radius of curvature, rfeature is the nominal feature radius, and dAFM is the measured feature diameter. Using a value of rtip ) 40 nm,17 and the values rfeature ) 3.5 nm (half the experimental antibody height) or rfeature ) 5 nm (from ref 14), we calculate dAFM ) 47.3 nm or dAFM ) 56.7 nm, respectively. These values are in excellent agreement with our measurements. On the basis of these considerations, we therefore assign the features in these images as single adsorbed antibodies; the few larger features we observe (a minority species on the surface) are assigned as arising from small antibody clusters. More important, it is also clear from Figure 1A,B that the number density of antibodies on the surface increases as the concentration of TA in the soaking solution used to form the mixed, self-assembled monolayer increases, as expected if the antibodies are only covalently linked to adsorbed TA. A plot of antibody density as a function of monolayer composition for a more complete data set is shown in Figure 1C. It is clear from this plot that the surface epitope density can be varied over approximately 2 orders of magnitude simply by changing the composition of the soaking solution used to form the mixed monolayer. The adsorption isotherm of the TA/BT mixed monolayers, and the quantitative relationship between the surface concentration of TA and the immobilized antibody density, is reported elsewhere.19 To address the possibility of nonspecific adsorption of antibodies onto the surface, a control experiment was carried out in which a pure BT SAM was subjected to the same chemical treatment as the TA/BT mixed monolayers (i.e., as outlined in Scheme 1). AFM images of this sample reveal a featureless, atomically flat surface. No evidence of the nonspecific adsorption of significant amounts of biological material was observed. While some nonspecific adsorption may indeed occur, (17) By imaging atomic steps on Au(111), we calculate a tip radius of curvature of approximately 40 ( 10 nm for the tips used in these experiments, which is in good agreement with the manufacturer’s specifications (20 nm e rtip e 60 nm). (18) In our experiments, the tip is almost always larger than the features being imaged. This situation is illustrated in the accompanying diagram, where the larger circle represents the radius of the AFM tip and the smaller circle represents the radius of a spherical particle adsorbed on the surface. From the geometry of the problem, it is clear that we can solve for rAFM using the Pythagorean theorem and multiplying by 2 to obtain dAFM, yielding eq 1.
(19) Dong, Y.; Shannon, C., Auburn University. Unpublished work.
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
2373
Figure 1. Capture antibody density as a function of self-assembly conditions. (A, top left) [TA]/[BT] ) 0.2. (B, top right) [TA]/[BT] ) 1.0. Both images were obtained in contact mode under ambient conditions. (C, bottom) Antibody surface density vs [TA]/[BT] for the complete data set. The line is a guide to the eye.
it is clear that the covalently attached antibodies are the majority species observed in Figure 1A,B. Other evidence to be discussed shortly also supports this conclusion. Characterization of Antibody-Antigen Binding. AFM was also used to characterize the binding of antigen to the surfaceconfined capture antibodies. A representative image of an antirabbit IgG antibody array after exposure to rabbit IgG is shown in Figure 2A. For comparison, an image of the surface prior to antigen exposure is shown in Figure 2B. Two clearly identifiable types of features are observed in the image shown in Figure 2A. The smaller features are characterized by an average height and diameter indistinguishable from those observed in the antibody arrays prior to exposure to antigen, Figure 2B. The larger features, in contrast, are characterized by an average height of 13.0 nm and an average diameter of 72.5 nm. Significantly, the larger features are almost exactly twice the height of the smaller features, as expected for the formation of an antigen-antibody complex, since rabbit IgG and anti-rabbit IgG are similar in size. This is illustrated quite clearly by the line scans shown in Figure 2C. On this basis, we assign the larger features as arising from antibody2374
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
antigen complexes and the smaller features as uncomplexed capture antibodies. From the analysis of numerous images obtained under identical conditions, we calculate that approximately 22.8% of the adsorbed antibodies are complexed upon exposure to antigen solution. There are two reasonable explanations for the observed binding efficiency. First, given that there are roughly 60 available lysine residues on the surface of IgG, it is clear that there is a distribution of antibody orientations within a given antibody array. Clearly, some of these orientations are unsuitable for binding. Second, some of the antibodies may lose biological activity upon adsorption.20 Heterogeneous Competitive Immunoassay. To independently assess the biological activity of the immunosurfaces and to validate the use of AFM for the detection of antibody-antigen binding in these systems, we sought to adapt a conventional competitive immunoassay using labeled and unlabeled analyte to our system and compare the results to those obtained by AFM. To accomplish this goal, the experiment was altered slightly: (20) Vikholm, I.; Albers, W. M. Langmuir 1998, 14, 3865-3872.
Figure 2. AFM analysis of antibody-antigen binding at an anti-rabbit IgG antibody array. (A, top left) Contact-mode image after exposure of a freshly prepared antibody array to a rabbit IgG solution. (B, top right) An image of the surface before exposure for comparison purposes. Note the different gray scales for the two images. See part C for quantitative height information. (C, bottom) Line scans selected from the images in (A) and (B) showing evidence of antibody-antigen coupling. The average fraction of capture antibodies that are complexed for this system is 0.228.
instead of preparing antibody arrays as described in the previous sections, antigen arrays were synthesized using the same chemical approach. By adopting this strategy, we were able to employ commercially available enzyme-labeled antibodies (i.e., anti-rabbit IgG) in our competitive EIA. In addition, this experiment highlights the general nature of our synthetic approach. In the interest of experimental simplicity we based our competitive heterogeneous EIA with electrochemical detection on the scheme developed a number of years ago by Heineman et al.19 The enzyme label used by these workers was alkaline phosphatase, which catalyzes the hydrolysis of p-aminophenol phosphate (PAPP, enzyme substrate) to p-aminophenol (PAP). PAP is a well-characterized electroactive compound that, perhaps more importantly, can be detected under conditions where the antigen array is stable to electrochemical cycling (i.e., without employing strongly oxidizing conditions). The competitive assay was carried out using test solutions containing concentrations of alkaline phosphatase conjugated
monoclonal anti-rabbit IgG (Ab*) and anti-rabbit IgG (Ab) such that [Ab*]/[Ab] varied from 0.0 to 1.0. Representative cyclic voltammograms for [Ab*]/[Ab] ) 1.0 and [Ab*]/[Ab] ) 0.0 are shown in Figure 3A,B, respectively. In Figure 3A, a single pair of quasi-reversible voltammetric waves centered at E°′ ≈ -0.090 V are observed. This couple is the characteristic voltammetric signature of PAP; the details of this voltammetry have been discussed at length in the literature and will not be considered further here.21 It is important to note, however, that the cathodic peak current is slightly smaller than the anodic peak current due to a followup reaction (EC mechanism); therefore, only the anodic peak current density was used for quantitation. In addition, when no labeled antibody was present (i.e., [Ab*]/[Ab] ) 0), a small reductive peak at -0.196 V is nevertheless still observed, Figure 3B. This is due to the slow hydrolysis of small amounts of PAPP to PAP in aqueous solutions, and has been observed previously.19 (21) (a) Niwa, O.; Xu, Y.; Halsall, H. B.; Heineman, W. R. Anal. Chem. 1993, 65, 1559-1563.
Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
2375
Table 1. EIA Calibration Data [Ab*]/ [Ab] + [Ab*]
ipa (µA cm-2)
[Ab*]/ [Ab] + [Ab*]
ipa (µA cm-2)
0.1 0.25 0.4 0.5
3.38 8.06 11.2 15.9
0.6 0.75 0.85 1
19.2 24.4 26.4 30.2
Figure 4. Correlation of electrochemical response (peak current density) and antibody-antigen complex surface density (as measured by AFM) for a rabbit IgG antigen array.
1) to the number density of Ab-Ag complexes detected by AFM, as a function of epitope density (controlled by changing [TA]/ [BT] in the self-assembly step). The correlation of these two measurements is very good, as shown in Figure 4, and indicates that the density of antibody-antigen complexes measured by AFM is representative of the immunosurface as a whole. Figure 3. Heterogeneous immunoassay with electrochemical detection. Electrochemical response of rabbit IgG antigen arrays as a function of [Ab*]/[Ab]. (A, top left) [Ab*]/[Ab] ) 1.0. (B, top right) [Ab*]/ [Ab] ) 0.0. (C, bottom) Electrochemical response after exposure of an anti-rabbit IgG array to alkaline phosphatase conjugated anti-sheep IgG. All measurements were made in 0.05 M carbonate buffer (pH 9.6) containing 1 mM MgCl2. The scan rate was 40 mV/s and the electrode area 0.197 cm2.
A complete calibration data set for a rabbit IgG antigen array is presented in Table 1. The linear least-squares correlation coefficient (R2) is 0.993. As a control experiment, a rabbit IgG antigen array was exposed to a solution containing alkaline phosphatase conjugated anti-sheep IgG. The resulting cyclic voltammetry is shown in Figure 3C, and indicates no binding to the capture antigen. Comparison of Electrochemical and AFM Detection. Finally, experiments were carried out to address whether the electrochemical and AFM signal transduction modes gave comparable results. Specifically, we compared the peak currents measured in an electrochemical assay (in which [Ab*]/[Ab] )
2376 Analytical Chemistry, Vol. 72, No. 11, June 1, 2000
CONCLUSIONS A simple surface-chemical route to structurally well-defined antibody and antigen monolayers on electrode surfaces has been developed, and the resulting immunosurfaces have been characterized. In addition to its simplicity, the key features of our approach include the ability to control the surface epitope density over a fairly wide range and to use scanning probe microscopy in the design and optimization of more efficient EIA detectors. Finally, two independent detection modes, one based on AFM and the other on voltammetry, were directly compared and found to be in very good agreement. ACKNOWLEDGMENT The support of this research by the Auburn University Dean’s Research Initiative program and the National Science Foundation is gratefully acknowledged. We thank Professor Eric Bakker for helpful discussions during the early phases of this work. Received for review December 20, 1999. Accepted March 9, 2000. AC991450G
