Selective Binding of Polychlorinated Biphenyl Congeners by a

Selective Binding of Polychlorinated Biphenyl Congeners by a Monoclonal ... the dissociation constants (Kd) and on and off rates (kon,koff) for bindin...
0 downloads 0 Views 95KB Size
Download PDF
Anal. Chem. 2001, 73, 5477-5484

Selective Binding of Polychlorinated Biphenyl Congeners by a Monoclonal Antibody: Analysis by Kinetic Exclusion Fluorescence Immunoassay Ya-Wen Chiu,†,§ Qing X. Li,*,‡ and Alexander E. Karu†

Department of Nutritional Sciences, University of California, Berkeley, California 94720, and Department of Molecular Biosciences and Biosystems Engineering, University of Hawaii at Manoa, Honolulu, Hawaii 96822

A previously described monoclonal antibody, S2B1, was highly selective for coplanar (non-ortho-chlorinated) PCB congeners in enzyme immunoassays that measured binding at equilibrium. In the present study, kinetic exclusion fluoroimmunoassay (KinExA) was used to determine the dissociation constants (Kd) and on and off rates (kon, koff ) for binding of various PCB congeners to affinity-purified S2B1 IgG and Fab fragments in solution. This method revealed that mono- and di-ortho-chlorinated PCBs were bound by S2B1, but the on rates were slower, and the off rates faster by 6-60-fold, than with congeners that had no ortho chlorines. Although the sensitivity of immunoassays may be improved by using competing haptens that S2B1 binds more weakly than the parent PCB, the KinExA results demonstrate that congener specificity is an intrinsic property of S2B1 and does not require weaker binding haptens. KinExA also provided new information on the percentage of active binding sites, valence, and effects of buffer, solvent, and biotinylation on S2B1. The advantages and drawbacks of KinExA for measuring antibody-ligand binding are described. Polychlorinated biphenyls (PCBs), originally valuable for many industrial applications, are now known to be among the most hazardous and widely distributed synthetic substances. Commercial PCB formulations, synthesized by controlled chlorination of biphenyl, may be thought of as combinatorial libraries, with various mole fractions of congeners that differ in number and position of the chlorine atoms. Congeners with and without ortho chlorines differ in their most energetically favored conformations because ortho chlorines exert a repulsive force that causes rotation of the phenyl rings about the bond that connects them. Of the 209 congeners possible with 1-10 chlorines, 20 have no ortho chlorines, and their phenyl rings can lie in the same plane.1-3 The coplanar congeners 3,4,3’,4’-tetrachlorobiphenyl (PCB 77) and 3,4,3’,4’,5’-pentachlorobiphenyl (PCB 126) mimic the size and * Corresponding author: (phone) 808-956-2011; (fax) 808-956-5037; (e-mail) [email protected]. † University of California, Berkeley. ‡ University of Hawaii at Manoa. § Current address: Environmental Protection Administration, Heng-Yang Road, Taipei, Taiwan, R.O.C. (1) McKinney, J. D.; Singh, P. Chem.-Biol. Interact. 1981, 33, 271-283. (2) Creaser, C. S.; Krokos, F.; Startin, J. R. Chemosphere 1992, 25, 1981-2008. (3) Egolf, D.; Jurs, P. Anal. Chem. 1990, 62, 1746-1754. 10.1021/ac0102462 CCC: $20.00 Published on Web 10/06/2001

© 2001 American Chemical Society

shape of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD), one of the most toxic organic compounds known. PCBs 77, 126, and 169, and, to a lesser extent, other coplanar congeners, bind to the aryl hydrocarbon receptor (AhR), estrogen receptors, and other proteins. Most of their toxicological effects appear to originate from these binding events.4-7 Noncoplanar PCBs are much more abundant and diverse than the coplanar congeners in commercial PCB formulations, and their toxicity mechanisms are not yet as well defined.8-11 Thus, the ability of congeners to assume the coplanar conformation provides a structural basis for different toxicological modes of action. Congener-specific PCB analytical methods are the most needed tools for contamination surveys, risk assessment, remediation monitoring, and toxicological and epidemiological research.12-15 Congener-specific instrumental analysis is expensive, timeconsuming, and requires skilled personnel, specialized recovery and cleanup methods, and sophisticated data interpretation.2,15,16 Immunoassays and immunoaffinity methods are an attractive alternative. PCB immunoassay kits, validated by the U.S. Environmental Protection Agency (EPA) for particular uses, are commercially available and increasingly used for regulatory purposes.17-20 However, the kits are not congener-specific. (4) Kafafi, S. A.; Afeefy, H. Y.; Ali, A. H.; Said, H. K.; Ad-Elazem, I. S.; Kafafi, A. G. Carcinogenesis (Oxford) 1993, 14, 2063-2071. (5) Kafafi, S. A.; Afeefy, H. Y.; Ali, A. H.; Said, H. K.; Kafafi, A. G. Environ. Health Perspect. 1993, 101, 422-425. (6) Safe, S. CRC Crit. Rev. Toxicol. 1984, 13, 319-396. (7) Safe, S. CRC Crit. Rev. Toxicol. 1994, 24, 87-149. (8) Tilson, H. A.; Kodavanti, P. R. Neurotoxicology 1997, 18, 727-743. (9) Shain, W.; Bush, B.; Seegal, R. Toxicol. Appl. Pharmacol. 1991, 111, 3342. (10) Hansen, L. G. Environ. Health Perspect. 1998, 106, 171-189. (11) Research Triangle Institute. Toxicological profile for polychlorinated biphenyls; U. S. Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and Disease Registry, Atlanta, GA, 1997. (12) McFarland, V. A.; Clarke, J. U. Environ. Health Perspect. 1989, 81, 225239. (13) Wolff, M. S.; Camann, D.; Gammon, M.; Stellman, S. D. Environ. Health Perspect. 1997, 105, 13-14. (14) Kodavanti, P. R. S.; Tilson, H. A. Neurotoxicology (Little Rock) 1997, 18, 425-441. (15) Mes, J.; Conacher, H. B. S.; Malcolm, S. Int. J. Environ. Anal. Chem. 1993, 50, 285-297. (16) Erickson, M. D. Analytical chemistry of PCBs, 2nd ed.; CRC/Lewis Publishers: Boca Raton, FL, 1997. (17) U.S. EPA Office of solid waste and emergency response (OSWER). Screening for polychlorinated biphenyls by immunoassay (Method 4020); 1996; http:// www.epa.gov: 80/epaoswer/hazwaste/test/4020.pdf.

Analytical Chemistry, Vol. 73, No. 22, November 15, 2001 5477

Immunoassay of individual PCB congeners or small groups of congeners poses a very challenging problem in biomolecular recognition. The hydrophobicity and similar shapes and electrostatic charge distributions of PCBs leave very few options for developing congener-specific antibodies and immunoassays by conventional methods.21-23 Antibody engineering is a possible alternative, but more must be learned about PCB-antibody interactions. In 1995, we reported the derivation and properties of a monoclonal hybridoma antibody (MAb), S2B1, that was highly specific for coplanar PCBs 77 and 126 in competition ELISAs (cELISAs).24 More recently, we described the molecular cloning, DNA sequence, and performance in cELISAs of a recombinant Fab fragment (rFab) from S2B1, which was functionally identical to the MAb.25 Sequence comparisons and computational modeling provided a scientifically plausible three-dimensional structure and suggested bonding interactions in the PCB binding site. In the present paper, we describe additional data on the kinetics of PCB congener binding to S2B1 IgG and Fab in solution. These results are proving helpful for testing predictions from the models, such as orientation of PCB 126 in the binding pocket and involvement of framework amino acid side chains in binding (Pellequer, J-L; Roberts, V. A.; et al., in preparation), and genetically engineering variant rFabs with new useful binding properties. Although we might have used any of several classical affinity determination methods, a new type of flow injection fluorescence immunoassay instrument became available at the time we began these experiments. This method and the instrument used for it are known as kinetic exclusion immunoassay (KinExA). Its major advantage is that it measures true liquid-phase equilibrium dissociation constants (Kd) and association rate constants (kon) using very small amounts of antibody. These experiments were done to learn more about the congener selectivity of MAb S2B1, rather than to compare KinExA with other methods. However, KinExA reproducibly gave us the data we sought, and some aspects of its use are presented. MATERIALS AND METHODS Safety Precautions. All PCB-containing solutions, solid wastes, and containers were handled and disposed of as previously described.25 Protective glasses, gowns, and double nitrile gloves were worn. (18) U.S. EPA Region I: New England Quality Assurance Unit Staff, Immunoassay guidelines for planning environmental projects (QA Unit Fact Sheet 10/96; New Immunoassay Guidance), 1996; http://www.epa.gov/region01/measure/ ia/iaguide.html. (19) U.S. EPA. Immunoassay: Principles, Procedures, Advantages, and Limitations. In U.S. EPA Field Analytical Technology Evaluation Encyclopedia (FATE). U.S. EPA Technology Innovation Office, 1999; http://fate.cluin.org/immunoassay_index.asp?techtypeid)45. (20) Hurt, A. Field Analytical Measurement Technologies, Applications, and Selection; California Military Environmental Coordination Committee, California State Water Quality Control Board, Sacramento, CA, 1996. (21) Carlson, R. E. In Immunoanalysis of Agrochemicals: Emerging Technologies; Nelson, J. O., Karu, A. E., Wong, R. B., Eds.; ACS Symposium No. 586; American Chemical Society: Washington, DC, 1995; pp 140-152. (22) Fra´nek, M.; Hruska, K.; Sisa´k, M.; Diblı´kova´, I. J. Agric. Food Chem. 1992, 40, 1559-1565. (23) Goon, D. J. W.; Nagasawa, H. T.; Keyler, D. E.; Ross, C. A.; Pentel, P. R. Bioconjugate Chem. 1994, 5, 418-422. (24) Chiu, Y.-W.; Carlson, R. E.; Marcus, K. L.; Karu, A. E. Anal. Chem. 1995, 67, 3829-3839. (25) Chiu, Y.-W.; Chen, R.; Li, Q. X.; Karu, A. E. J. Agric. Food. Chem. 2000, 48, 2614-2624.

5478 Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

Figure 1. Structure and nomenclature of PCB congeners and hapten derivatives used in this study: Generic PCB structure (top left); IUPAC number and chlorination positions (bottom left); hapten I (top right); 3,4-keto (bottom right).

Reagents and PCB Reference Standards. Spectrograde organic solvents and deionized glass-distilled water were used throughout. All reagents were obtained from Aldrich Chemical Co. (Milwaukee, WI), Fisher Scientific (Pittsburgh, PA), or Sigma Chemical Co. (St. Louis, MO) unless otherwise indicated. PCB congeners, > 99% purity, from AccuStandard, Inc. (New Haven, CT) were used to prepare reference solutions of 200 ppm (PCBs 77 and 126) in 2-propanol and 100 ppm (PCBs 35, 52, and 70) in methanol. These were stored at 4 °C in glass vials with Teflonlined screw caps. The detecting antibody used for KinExA measurements was fluorescein isothiocyanate (FITC)-conjugated goat antimouse IgG H+L chain antibody (Catalog No. 115-095003, Jackson Immunoresearch Labs, West Grove, PA). Aliquots were diluted to 0.5 mg/mL in 50% glycerol and stored at -20 °C. Buffers for KinExA and Enzyme Immunoassay (ELISA). Beads for KinExA were adsorptively coated with hapten-BSA or hapten-cytochrome c conjugates as described below. The coating buffer was 0.015 M Na2CO3-0.035 M NaHCO3-0.003 M NaN3, pH 9.6, Blocking buffer and diluent were 0.1% bovine serum albumin (BSA)-0.1% NaN3n-0.005% Tween 20-10% methanol in phosphate-buffered saline (PBS).25 Washing buffer was PBS containing 0.1% BSA-0.1% NaN3-0.005% Tween 20, pH 7.4. Direct cELISAs were done in the streptavidin-biotin format as previously described,24 with PBS-0.01% gelatin-0.005% Tween 20 as blocking buffer, and this buffer with 5% (v/v) methanol as diluent. This Tween concentration was one-tenth of that used in most immunoassays. As previously reported, higher concentrations of Tween 20 adversely affect assays with MAb S2B1.24 Haptens and Conjugates. The previously described hapten derivatives, 6-[(3,3′,4-trichlorobiphenyl-4-yl)oxy]hexanoic acid (hapten I) and 4-(3,4-dichlorobenzoyl)butyric acid (3,4-keto),24 were newly synthesized and purified, and their structures were verified.25 Portions were converted to active esters and conjugated to BSA and horseradish peroxidase (HRP) as before.24 Figure 1 shows the general structure of PCBs, the hapten derivatives, and the chlorination patterns of the congeners used in this study. Purification and Biotinylation of S2B1 IgG and Fab Fragments. S2B1 IgG was affinity-purified on protein A-Sepharose, IgG concentration was determined from the absorbance at 280 nm, Fab fragments were prepared by papain digestion of affinitypurified S2B1 IgG using an Immunopure Fab preparation kit (Pierce Chemical Co., Rockford, IL), and portions of MAb and

Fab were biotinylated, all as previously described.24,26 Aliquots were stored at -70 °C. Adsorption of Hapten-BSA Conjugates to Beads. Poly(methyl methacrylate) (PMMA) beads and polystyrene-12% divinylbenzene (PS-DVB) beads of 98-µm diameter were obtained from Sapidyne Instruments, Inc. (Boise, ID). Dry beads (200 mg) in 1.5-mL Eppendorf tubes were suspended and settled twice in 1 mL of 6 N NaOH and then three times successively in glass distilled water, PBS, and coating buffer. They were then suspended in 1 mL of coating buffer containing 0.2 mg of hapten I-BSA or 3,4-keto-BSA and rolled at room temperature for 3 h. After the conjugate solution was discarded, the beads were washed three times with PBS, then resuspended in blocking buffer, and rolled again at room temperature for 1 h. Coated bead preparations in blocking buffer were stable for 7-10 days at 4 °C. All experiments were done with 3,4-keto-BSA-coated PS-DVB beads, unless otherwise indicated. On the day of use, 200-mg aliquots of coated beads were uniformly suspended in 29 mL of PBS to provide the concentration required for KinExA. For each sample analyzed, a new bed of hapten conjugate-coated beads exactly 4 mm high (to match the width of the excitation beam) was deposited over the mesh trap in the capillary flow cell. The bed height proved very important for reproducibility and, thus, was checked with a small metal template before each run. KinExA Principles and Operation. KinExA is a flow injection system designed to measure the unoccupied binding sites after an antibody and analyte are allowed to react in solution. A schematic of the device and principles of operation were described on the manufacturer’s Web site (www.sapidyne.com) and by Blake et al.27,28 The antibody-analyte mixture is pumped through a capillary flow cell of 1.6-mm inside diameter, at the bottom of which is a bed of beads, 4 mm deep, coated with hapten-protein conjugate, retained on a very fine mesh. The beads lie in the path of a fluorescence excitation beam and an emission detector. As the sample, in about 50 µL, is passed through the beads, antibodies with unoccupied binding sites are bound. This occurs so quickly (on the order of 240 ms when the flow rate is 0.5 mL/min) that there is negligible time for an antibody molecule to release from the hapten and reassociate with soluble ligand. The test antibody attached to the beads is exposed to excess fluorescent-labeled goat anti-mouse secondary antibody and then unbound secondary antibody is washed away, and the fluorescence that remains on the beads (directly proportional to the amount of primary antibody) is measured. When these steps are completed, the beads are automatically pumped out and replaced with a new aliquot for the next measurement. All experiments were done on a first-generation instrument generously loaned by Dr. Steve Lackie (Sapidyne Instruments, Inc., Boise, ID). The KinExA was controlled, and data were acquired through an IBM-compatible personal computer with Intel 486 or later microprocessor, running Microsoft Windows. The operating parameters (volume, flow rate, sequence, time interval) (26) Harlow, E.; Lane, D. Antibodies: A Laboratory Manual; Cold Spring Harbor Laboratory; Cold Spring Harbor, NY, 1988. (27) Blake, D. A.; Khosraviani, M.; Pavlov, A. R.; Blake, R. C. In Immunochemical Technology for Environmental Applications; Aga, D. S., Thurmann, E. M., Eds.; ACS Symposium Series 657; American Chemical Society: Washington, DC, 1997; pp 49-60. (28) Blake, R. C., 2nd; Pavlov, A. R.; Blake, D. A. Anal. Biochem. 1999, 272, 123-134.

were configured and run by Sapidyne’s control software. Experimental data were imported into a Microsoft Excel spreadsheet. Experimental Procedures. Total protein was determined by UV spectrophotometry or by micro-BCA assay (Pierce Chemicals, Rockford, IL), using crystalline BSA as a standard. The percentage of active binding sites was estimated from a Scatchard plot and also from an iterative nonlinear fit of the percent of unoccupied sites versus site concentration. To determine Kd, various amounts of antibody and PCB analyte were mixed and allowed to reach equilibrium for 30 min at room temperature, and then the solutions were passed through the KinExA’s beads to capture and quantify the unoccupied binding sites. An estimate for Kd was calculated by iterative least-squares fit of the parameters [A0], [B0], Sig0%, and Sig100%, corrected for the percentage of functional binding sites. To measure kon, a fixed amount of antibody and various concentrations of analyte were mixed by simultaneous injection, and the mixture was passed through a tube of known length, and through the haptenated beads, at a predetermined rate. Each Kd determination was made with five different PCB concentrations in triplicate, i.e., 15 separate runs. The kon determinations were made with 8-12 concentrations of PCB congeners in triplicate, ranging from 0 to 100% binding site occupancy. The koff values were calculated; koff ) konKd.29 Data Analysis. Analyses and curve fitting were done in MathCad Plus 6.0 (MathSoft, Inc., Cambridge, MA), on a Macintosh PowerPC computer, using worksheets developed by Sapidyne. The calculations and theoretical basis have been extensively described in the literature.29-31 Three types of determinations were made: estimated percentage of functional binding sites, equilibrium dissociation constant (Kd), and kon. The instrument’s fluorescence signal was related to antibody and ligand concentrations kon

by the classical binding equation, [A] + [B] y\ z [AB], where the k off

concentrations of analyte, active antibody binding sites, and antibody-analyte complex are [A], [B], and [AB], respectively. Kd, is defined as koff/kon. Therefore, [AB] ) [B][A]/Kd. At equilibrium, the total analyte concentration [A0] ) [A] + [AB], and the total concentration of functional binding sites [B0] ) [B] + [AB]. The experimental data are the fluorescence signals, in millivolts, measured at various values of [A]. The equation that relates the instrument readings to molecular species and kinetic parameters is signal )

Sig100% - Sig0% (([Bo] - Kd - [Ao]) + 2[Bo]

x[Bo]2 + 2[Bo]Kd - 2[Bo][Ao] + Kd2 + 2[Ao]Kd + [Ao]2) + Sig0%

where signal is the instrument signal for [B] and Sig0% is the signal when [B] ) 0, i.e., the baseline nonspecific background when no primary antibody is bound to the haptenated beads. Sig100% is the maximum signal when no soluble analyte-antibody complex was (29) Day, E. D. Advanced Immunochemistry, 2nd ed.; Wiley-Liss: New York, 1990; pp 298-299. (30) Steward, M. W.; Steensgard, J. Antibody affinity: Thermodynamic aspects and biological significance; CRC Press: Boca Raton, FL, 1983; pp 5-11. (31) Glass, T. R.; Lackie, S. Kinetic Exclusion Fluoroimmunoassay (KinExA); 1997; http://www.sapidyne.com/.

Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

5479

formed, and all of the active antibody ([B0]) is captured on the beads.29-31 For calculation of kon, the Kd, molar concentration of active binding sites, and the time allowed for the analyte and antibody to react were all known. The data were fitted using an iterative nonlinear least-squares algorithm, as in the Kd determination. The 95% confidence limits for [B0] (concentration of active binding sites), Kd, and kon, were computed by varying the best-fit value and reoptimizing the remaining variables at each point, using an algorithm developed by Sapidyne.32 RESULTS AND DISCUSSION The PCB-MAb S2B1 interaction is particularly interesting from the standpoint of structural biology, because binding may involve a conformational change of the ligand, part of the antibody, or both. In this study, the primary target analytes were the coplanar PCBs 77 and 126, as defined by the specificity of MAb S2B1 and its proteolytic Fab fragments in cELISAs.24,25 The hapten derivatives used as competitors were the structures chemically defined in Materials and Methods, depicted in Figure 1, and abbreviated as hapten I and 3,4-keto for consistency with our earlier papers. The experiments described here were done (a) to obtain quantitative data for the differences in binding of PCB congeners and hapten derivatives that resulted in the high selectivity for coplanar congeners in ELISAs and (b) to compare the binding properties of whole IgG and Fab fragments and choose an optimal hapten derivative for panning display phage to select rFabs. Most present-generation immunoassay methods required a polyclonal antiserum or MAb that was empirically found to have strong binding and high selectivity. By contrast, kinetic analysis has become essential to support the recent advances in antibody engineering, selection from combinatorial libraries, multianalyte immunoassays, discrimination within large groups of chemical analogues such as PCBs, and sensor applications that involve temporal as well as end point binding. The KinExA method was attractive for two reasons. First, only nanogram-to-microgram amounts of antibody, analyte, and hapten conjugate were needed for each measurement; sample volumes could be as small as 50 µL. Second, KinExA directly measured bimolecular, solution-phase binding kinetics, without the expense and artifacts possible when antibody or hapten conjugate is immobilized in surface plasmon resonance (SPR) and resonant mirror instruments such as BIAcore and IASYS.33-39 Technical Aspects of KinExA Use. Some standard KinExA protocols had to be modified for experiments with the PCBhapten conjugates and hydrophobic PCBs. PMMA beads remained monodisperse when coated with hapten I-BSA, but they clumped if coated with 3,4-keto-BSA. PS-DVB beads did not (32) Lackie, S.; Glass, T. Kd mathematics; 2000; www.sapidyne.com/kdmath.html. (33) Nieba, L.; Krebber, A.; Plu ¨ ckthun, A. Anal. Biochem. 1996, 234, 155-165. (34) Schuck, P. Biophys. J. 1996, 70 (A212), 1230-1249. (35) Schuck, P. Annu. Rev. Biophys. Biomol. Struct. 1997, 26, 541-566. (36) Edwards, P. R.; Maule, C. H.; Leatherbarrow, R. J.; Winzor, D. J. Anal. Biochem. 1998, 263, 1-12. (37) Malmborg, A.-C.; Borrebaeck, C. A. K. J. Immunol. Methods 1995, 183, 7-13. (38) Lowe, P. A.; Clark, T. J.; Davies, R. J.; Edwards, P. R.; Kinning, T.; Yeung, D. J. Mol. Recognit. 1998, 11, 194-199. (39) Adamczyk, M.; Moore, J. A.; Yu, Z. Methods (Orlando, Fla.) 2000, 20, 319328.

5480

Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

clump when coated with 3,4-keto-BSA. Clumping causes uneven bead distribution, inconsistent exposure of ligand for antibody binding, and unacceptable variance in baseline and maximum fluorescence (Sig100%) values. The amount of beads automatically delivered to the flow cell from the stirred reservoir can be adjusted by varying the rate and duration of withdrawal, but only if the suspension is monodisperse. Several pre-and postexperimental procedures proved essential to maintain precision and reproducibility. These included a degassing cycle to clear air bubbles from the tubes, a procedure to prime the rotary valves, a washing cycle between replicates and new samples, a “night wash” program to automatically wash the entire system 30 times, and a “morning rinse” to condition the system with PBS 3-10 times before starting a series of measurements. To prevent carry-over of antibody and ligand, all runs were begun with the most dilute samples and progressed to those with higher concentrations. An additional wash was included between measurements of the most concentrated sample from one datum set and the most dilute sample of the next set. In some experiments, nonspecific binding in the flow cell caused the baseline fluorescence to drift upward in each consecutive measurement. When it exceeded 0.1 V, experiments were interrupted, and the system was washed three times each with neat methanol, 0.1 M NaOH, and PBST. In the prototype instrument we used, the range of measurable kon values was limited by the available sample injection flow rates (0.25-0.75 mL/min) and the flow path (length of tubing) between the mixing chamber and the flow cell. The kon determinations in this paper were done using a tube with a minimum transit time of 12 s. Fortunately, the kon measurements for S2B1 MAb and Fab could be obtained by adjusting the antibody and ligand concentrations so that equilibrium was not reached in 12 s. Other antibody-ligand combinations that we tested had kon values that approached or exceeded 108 M-1 s-1. At the dilutions necessary to bring those complexes into a measurable time frame, the fluorescence signal was too weak to use (Bell, C. W.; Karu, A. E., unpublished). The present version of the instrument (KinExA 3000) has much broader measurement ranges: 10-2-10-13 M for Kd (the reciprocal of affinity), and 103-109 M-1 s-1 for kon. Within these bounds, koff values from 10-6 to 1.0 s-1 can be calculated.31 Analysis of PCB Binding. For all KinExA measurements, fluorescence in the bead bed was continuously recorded as a function of time for each sample. Plots of the raw data resulted in “KinExAgrams” essentially identical to the “sensorgrams” produced by BIAcore (SPR) and IASYS (resonant mirror) instruments.37,38 These plots provided real-time information on baseline drift, monodispersity of the beads, and other factors that affect data quality. Figure 2 shows the superimposed KinExAgrams for equilibrated mixtures of MAb S2B1 and PCB 126. The signal used for analysis was the difference between the baseline (t ) 100180 s) and the steady-state fluorescence after excess probe was washed away (t ) 420-500 s). Figure 2 also shows that the net signal was linearly proportional to the amount of S2B1 IgG (i.e., active binding site) over the entire concentration range used in these experiments. The kinetic analyses were done in three steps: estimation of the percentage of active binding sites, determination of Kd, and measurement of kon. All methods for determining affinity and rate

Figure 2. Example of KinExA output (KinExAgram). Superimposed fluorescence responses from five equilibrium mixtures of S2B1 IgG and various amounts of PCB 126, used for determination of Kd. The baseline was established from 0 to 200 s, as unoccupied S2B1 was captured by the immobilized haptens. Fluorescein-labeled secondary antibodies were added over the next 200 s, followed by a wash to remove unbound secondary antibodies. The signal (delta) used for calculations was the difference between the average values from 450 to 550 s and the baseline.

Figure 3. Scatchard plot for estimation of active antibody. The crosses are datum points (average of triplicates), and the solid line is the theoretical best fit. The ordinate is the ratio of bound/free of S2B1 IgG, the slope is 1/Kd, and the X intercept indicates that nearly 100% of the binding sites in this sample are active.

constants are affected by the percentage of binding sites that remain functional after purification, storage, or chemical modification.30,40,41 This percentage was estimated by comparing results when either the PCB concentration or the protein concentration was used as the reference. The theoretical maximum concentration of binding sites was based on total protein concentration. KinExA data were collected on equilibrium solutions containing a fixed amount of S2B1 IgG or Fab (13.3 nM, equal to 2 µg of IgG/mL), with no PCB, a saturating amount of PCB 77 or 126, and five to eight intermediate concentrations in triplicate. A Scatchard plot (the ratio of unoccupied to occupied antibody versus the amount of occupied antibody) was linear, as expected for a MAb binding a single small-molecule analyte (Figure 3). The x-axis intercept gave the moles of active binding sites per mole of MAb or Fab, where each molecule has two binding sites. The active binding site concentration was also estimated from a nonlinear fit of the percent of unoccupied sites versus site (40) Karlsson, R.; Fagerstam, L.; Nilshans, H.; Persson, B. J. Immunol. Methods 1993, 166, 75-84. (41) Adamczyk, M.; Mattingly, P. G.; Shreder, K.; Yu, Z. Bioconjugate Chem. 1999, 10, 1032-1037.

concentration, using Sapidyne’s standard Kd analysis program. Various preparations of S2B1 IgG and proteolytic Fab had from 60 to 90% active binding sites. All subsequent experiments were done with one batch of affinity-purified S2B1 IgG that had approximately 75% active sites and one batch of proteolytic Fab S2B1 with approximately 67% active sites. These correction factors reduced the Kd and kon estimates by 10-17%. Effects of Blocking Protein and Buffer Composition on KinExA. For direct cELISAs with biotinylated S2B1 IgG captured on streptavidin-coated microwells, the optimal diluent and blocking buffer was PBS with 0.005% Tween 20 and 0.01% gelatin.24 When this buffer was used in the KinExA, the baseline increased by approximately 0.035 V per run. This drift was partially eliminated by using 0.1% BSA instead of gelatin. In many immunoassay formats, surfactants such as Tween 20 are commonly added to PBS at 0.05% (v/v) to reduce nonspecific binding. However, diluents containing 0.01% (v/v) or more Tween 20 significantly reduced PCB binding by MAb S2B1 in cELISAs.24 KinExA was affected even more; the highest usable concentration of Tween 20 was 0.005%. The optimum diluent and wash buffer consisted of PBS with 10% methanol, 0.1% BSA, and no surfactant. This kept the background stable within 0.05-0.1 V over many consecutive runs. Solvent Effects. Immunoassays for very hydrophobic analytes such as PCBs generally require a minimal amount of watermiscible organic solvent, surfactant, or both to keep the analytes soluble. However, larger amounts of these solubilizers often degrade assay performance. We previously found that addition of 5-10% methanol (MeOH) or dimethyl sulfoxide (DMSO) to the diluent changed the selectivity as well as sensitivity of direct cELISAs with MAb S2B1.24 When the diluent contained 10% DMSO instead of 5% methanol, the half-maximal inhibition (I50) by PCB 35 was 80 nM, rather than 1240 nM. Similar effects occurred in the KinExA. In the standard diluent, PBS-10% methanol-0.1% BSA, 0.05 mL of a 10 nM solution of S2B1 IgG, captured on beads coated with 3,4-keto-BSA, gave a net signal of 0.8 V. This was reduced by approximately 80, 87.5, or 90% if the methanol was replaced by 10, 15, or 20% DMSO, respectively, in the antibody capture step. However, substitution of 10% DMSO for the 10% methanol in the diluent improved the Kd for PCB 35 roughly 6-fold, from 2.2 ( 0.01 to 0.37 ( 0.001 nM. Thus, in both cELISA and KinExA, DMSO in the diluent weakened binding of S2B1 to the hapten but increased the affinity for the coplanar congener PCB 35. Hapten Preference. In KinExA, the Kd of a particular antibody-analyte pair should be independent of the hapten used to trap unoccupied antibody on the beads. To test this, the Kd for PCB 77 binding by S2B1 IgG was determined in separate experiments using PMMA beads coated with hapten I-BSA and PS-DVB beads coated with 3,4-keto-BSA (PMMA beads coated with 3,4-keto-BSA clumped, as noted above). The difference in Kd values, 2.8 ( 0.05 and 1.7 ( 0.03 nM, respectively, was probably not significant with respect to between-assay variation with separate preparations of antibody and beads coated with either hapten conjugate. Unfortunately, this comparison could not be made using S2B1 Fab, because it bound 3,4-keto-BSA-coated PSDVB beads too weakly to allow Kd measurements. This was consistent with our previous observation that S2B1 proteolytic and Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

5481

Table 1. PCB Binding Characteristics of MAb S2B1 Kd (nM)

PCB no.

chlorine positions

KinExAa

ELISAb

kon (M-1 s-1)a × 104

koff (s-1)

77 126 35 70 52

3,4,3′,4′-tetrachloro 3,4,3′,4′,5′-pentachloro 3,4,3′-trichloro 2,5,3′,4′-tetrachloro 2,5,2′,5′-tetrachloro

1.7 ( 0.03 2.5 ( 0.01 2.2 ( 0.01 66 ( 0.6 2300 ( 20

24 31 1240 6900 ncc

150 ( 4 120 ( 6 120 ( 6 22 ( 1.4 2.7 ( 0.05

0.003 0.003 0.0026 0.014 0.06

a Data are corrected values and are best-fitted for 95% confidence limit. b Direct cELISA with biotinylated S2B1 immobilized on streptavidincoated microwells, and 3,4-keto-HRP as competitor, as described in Chiu et al.24 c nc, no competition in cELISA by up to 5 ppm of a PCB congener.

Figure 4. Summary data for Kd of S2B1 IgG with PCB 77. A fixed amount of S2B1 IgG was mixed with 0, 2, 4, 6, 8, 10, 12, 14, 16, and 100 ppb of PCB 77 in triplicate, as described in Materials and Methods (panel A, left). The plot of binding site occupancy versus antibody concentration (panel A, right) shows the actual data (diamonds; average of triplicate measurements) and the theoretical best-fit curve (solid line). The residual error of the best-fit values for Kd and active antibody concentration are shown in panels B and C, respectively.

recombinant Fabs did not bind to microplate wells coated with 3,4-keto conjugates of BSA and cytochrome c.25 Kinetics of PCB Congener Binding. Table 1 summarizes the Kd and kon values determined by KinExA, and the koff values calculated from them, for MAb S2B1. Also included are Kd values estimated from I50 data in a previously reported direct cELISA,24 as described by Mu¨ller.42 These are for relative comparison only. The absolute values cannot be compared because direct cELISAs measure competition of soluble analyte and hapten for binding to antibody immobilized on a solid phase, and the results depend on the affinity for the competing hapten-enzyme conjugate (3,4keto-HRP in this experiment). An experiment analogous to direct ELISA could have been done by KinExA. For example, the binding of 3,4-keto-HRP to beads coated with MAb S2B1 IgG could have been measured with a fluorescent anti-HRP conjugate as the reporter. However, this was peripheral to our main objectives, which were (a) to determine whether KinExA would detect binding of mono- or di-ortho-chlorinated PCB congeners to S2B1, when ELISA could not, and (b) to select a hapten conjugate for panning S2B1 rFab displayed on M13 phagesa procedure analogous to indirect ELISA. PCBs 77, 126, and 35 have no ortho chlorines, PCB 70 has one, and PCB 52 has two. The Kd and kon values determined by (42) Mu ¨ ller, R. J. Immunol. Methods 1980, 34, 345-352.

5482 Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

KinExA for PCBs 77, 126, and 35 were nearly identical. By comparison, PCBs 70 and 52 had weaker affinities, 5-56-fold slower kon, and 5-23-fold faster koff. The data in Table 1 reflect the two greatest advantages of KinExA, compared with cELISA; KinExA measures binding kinetics in solution, and it is considerably more sensitive. In cELISAs, PCB concentrations as high as 5 µg/mL (5 parts per million) were required to demonstrate that MAb S2B1 had roughly 1% cross-reactivity with PCB 70 and negligible cross-reactivity with PCB 52.24 KinExA gave reproducible Kd and kon estimates at much lower PCB concentrations. The Kd values for PCBs 70 and 52 corresponded to 19 and 672 ng/mL (parts per billion), respectively. Error Estimates. Figure 4 shows the Kd analysis for PCB 77 binding by MAb S2B1 and error estimates that indicate the quality of the data. Panel A shows the data and best-fit curve. Panels B and C are plots of the residual errors in the fitted Kd and the active binding site estimates, respectively. The nonlinearity and sharp minimums indicate how sensitive the affinity determination is to the percentage of active binding sites and other experimental conditions. Figure 5 is the summary for determination of kon and its estimated error. Panel A demonstrates how the kon estimate becomes less accurate as the concentration of antibody used in the experiment approaches the Kd. Panel B shows the upper and lower 95% confidence levels for kon, and steep increase of residual

Figure 5. Example of results for kon determinations. A fixed amount of MAb S2B1 was injected and mixed with solutions of PCB 77 to give final concentrations of 0, 5, 10, 15, 25, 35, 60, and 100 ppb, in triplicate. Panel A shows the input data and calculated results at left. The plot at upper right shows the actual data (diamonds), the theoretical best-fit curve (solid line), and the response predicted if the solutions had reached equilibrium (dashed line). Panel B shows the residual error and 95% confidence limits of the best-fit kon value.

error on both sides of the best fitted value. Other data quality statistics included the ratio of antibody concentration to Kd ([Ab]/ Kd) in Kd determinations and how far from equilibrium the conditions were in kon determinations. A ratio of [Ab]/Kd > 1 indicates that the estimate of binding site concentration was accurate, but the Kd determination was less so. A ratio of [Ab]/ Kd < 1 indicates that the active binding site concentration estimate was less reliable, but the Kd value was more accurate, given the available data. Antibody Valence. Theoretically, Kd determinations by KinExA should be independent of antibody valence (the number of binding sites per molecule of antibody that can function under the particular experimental conditions).28 A separate experiment was conducted to compare the Kd values of S2B1 whole IgG and Fab for PCB 77. Equilibrium assays were performed with hapten1-BSA-coated PMMA beads. Whole S2B1 IgG (10 nM) and proteolytic Fab (30 nM) gave Kd values of 2.8 ( 0.05 and 1.6 ( 0.03 nM, respectively, corrected for the percentage of active binding sites. These values, with small intra-assay differences, were close enough to each other, and to the result in Table 1, to expect that Fab and intact IgG would perform similarly in cELISAs and other applications. KinExA can be used to verify the quality of chemically modified antibodies. A portion of affinity-purified S2B1 IgG was biotinylated and gave highly reproducible, sensitive results in direct cELISAs for more than one year.24 The Kd values for binding of PCB 35 by unmodified and biotinylated IgG were 2.2 and 0.68 nM, respectively, using beads coated with 3,4-keto-BSA hapten conjugate. Thus, PCB 35 binding by the biotinylated IgG was not significantly

different, or may have been slightly better, within experimental error. CONCLUSIONS The initial strategy for developing MAbs specific for coplanar PCB congeners required two components: an immunizing hapten that most closely mimicked coplanar congeners and a competitor hapten that retained features essential for recognition, but would be bound more weakly than the target PCBs in competition assays.21,24 In cELISAs, which are based on binding at equilibrium, MAb S2B1 showed no detectable binding of noncoplanar (monoand di-ortho-chlorinated) PCBs. The KinExA analyses showed that representative noncoplanar congeners were transiently bound by S2B1 in solution, but the off rates were faster and on rates were slower than those of the coplanar congeners by 5-60-fold. This establishes that S2B1 is intrinsically congener-selective, and its potential applications are not limited to competitive binding assay formats. However, use of weaker binding competitor haptens increased the sensitivity for coplanar congeners, i.e., resulted in lower I50 values in cELISAs.24 KinExA and cELISA revealed different aspects of PCB binding by S2B1. Ortho-chlorinated PCBs 70 and 52, which gave no observable binding in cELISAs, gave measurable, reproducible values in KinExA. The coplanar 3,4,3’-trichlorobiphenyl (PCB 35) was only about 2-10% cross-reactive, compared to PCBs 77 and 126, in a direct cELISA, yet all three of these congeners behaved similarly in KinExA. Comparison of Kd values for one batch of biotin-conjugated and unconjugated IgG demonstrated that kinetic measurements are useful for evaluating the functionality of Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

5483

chemically modified antibodies. Our limited data were consistent with results on other antibody-antigen combinations reported by Adamczyk et al., using a SPR instrument.41 At present, the most widely used instruments for measuring ligand-binding kinetics include those based on surface plasmon resonance (BIAcore), resonant mirror technology (IASYS), and isothermal titration calorimetry (ITC). The BIAcore and IASYS require immobilization of antibody or ligand and are subject to diffusion and rebinding artifacts.33-36,38 ITC is gaining use as a means of getting true solution-phase binding kinetics and thermodynamic data.43-45 These instruments are all more expensive to purchase, maintain, and operate than the current model of the KinExA. Despite these shortcomings, the instrumental methods are more accurate and require less manual work than cELISAs. Several ways to get approximate Kd estimates from solid-phase radioimmunoassays and ELISAs have been published over many years.42,46-53 By far the most accurate method to estimate Kd from an ELISA is to measure the amount of unoccupied antibody remaining in the liquid phase after equilibration with various amounts of analyte.52 The procedure is similar to what occurs in KinExA, but more complicated and less reproducible. The primary advantages of KinExA are the small amounts of antibody needed (e100 ng in 50 µL per sample), the ability to measure the percentage of active antibody, analysis of true solution-phase Kd and kon with good error estimates over a very wide range on one instrument, and the well-documented, user(43) Murphy, K. P.; Freire, E.; Paterson, Y. Proteins: Struct. Funct. Genet. 1995, 21, 83-90. (44) Indyk, L.; Fisher, H. F. Methods Enzymol 1998, 295, 350-364. (45) Jelesarov, I.; Leder, L.; Bosshard, H. R. Methods (Orlando, Fla.) 1996, 9, 533-541. (46) Van Heyningen, V.; Brock, D. J. H.; Van Heyningen, S. J. Immunol. Methods 1983, 62, 147-153. (47) Nimmo, G.; Lew, A.; Stanley, C. M.; Steward, M. W. J. Immunol. Methods 1984, 72, 177-187. (48) Loomans, E. E.; Roelen, A. J.; Van Damme, H. S.; Bloemers, H. P.; Gribnau, T. C.; Schielen, W. J. J. Immunol. Methods 1995, 184, 207-217. (49) Goldberg, M. E.; Djavadi-Ohaniance, L. Curr. Opin. Immunol. 1993, 5, 278281. (50) Hardy, F.; Djavadi-Ohaniance, L.; Goldberg, M. J. Immunol. Methods 1997, 200, 155-159. (51) Steward, M. W.; Lew, A. M. J. Immunol. Methods 1985, 78, 173-190. (52) Friguet, B.; Chaffotte, A. F.; Djavadi-Ohaniance, L.; Goldberg, M. E. J. Immunol. Methods 1985, 77, 305-319. (53) Azimzadeh, A.; Van Regenmortel, M. H. J. Immunol. Methods 1991, 141, 199-208. (54) Blake, D. A.; Blake, R. C.; Khosraviani, M.; Pavlov, A. R. Anal. Chim. Acta 1998, 376, 13-19. (55) Blake, D. A.; Chakrabarti, P.; Khosraviani, M.; Hatcher, F. M.; Westhoff, C. M.; Goebel, P.; Wylie, D. E.; Blake, R. C. I. J. Biol. Chem. 1996, 271, 2767727685. (56) Chakrabarti, P.; Hatcher, F. M.; Blake, R. C., 2nd; Ladd, P. A.; Blake, D. A. Anal. Biochem 1994, 217, 70-75. (57) Khosraviani, M.; Blake, R. C., 2nd; Pavlov, A. R.; Lorbach, S. C.; Yu, H.; Delehanty, J. B.; Brechbiel, M. W.; Blake, D. A. Bioconjugate Chem. 2000, 11, 267-277. (58) O’Connell, K. P.; Valdes, J. J.; Azer, N. L.; Schwartz, R. P.; Wright, J.; Eldefrawi, M. E. J. Immunol. Methods 1999, 225, 157-169. (59) Eldefrawi, M. E.; Azer, N. L.; Nath, N.; Anis, N. A.; Bangalore, M. S.; O’Connell, K. P.; Schwartz, R. P.; Wright, J. Appl Biochem Biotechnol. 2000, 87, 25-35. (60) Rogers, K. R.; Kohl, S. D.; Riddick, L. A.; Glass, T. Analyst 1997, 122, 11071111. (61) Craig, L.; Sanschagrin, P. C.; Rozek, A.; Lackie, S.; Kuhn, L. A.; Scott, J. K. J. Mol. Biol. 1998, 281, 183-201.

5484

Analytical Chemistry, Vol. 73, No. 22, November 15, 2001

friendly control and data analysis software. When reagents and reaction conditions are optimal, KinExA data are very reproducible, with coefficients of variation less than 5%. Disadvantages include the empirical choice of beads, haptens, and hapten presentation conditions, tedious procedures to prime and clean the system before and after a series of runs, and inability to measure kon when the analyte and antibody concentrations needed for measurable signals would reach equilibrium in less than 2 s. As with all currently available instruments for binding analysis, the KinExA can be run as a flow injection device, but it can only measure one sample at a time. A full analysis cycle for one replicate, including the steps before and after the actual measurement, takes about 15 min. To date, the KinExA has been used to measure binding kinetics of MAbs that bind metals,27,54 metal chelates,55-57 opiate metabolites in urine,58,59 chlorophenoxy herbicides,60 and peptide libraries.61 A detailed comparison of eight procedural and data quality criteria with the nine most common methods and instruments for measuring ligand binding kinetics may be found on the Sapidyne web site (http://www.sapidyne.com/systcomp.html). In previous-generation hybridoma and immunoassay technologies, hapten design was basically the only way to evoke antibodies with ideal characteristics for recognizing small molecules that have many analogues. MAb development was, and still is slow, expensive, and influenced by unpredictable biological factors. In those circumstances, it was more practical to focus on optimizing hapten structure and immunoassay formats for the available antibodies, as we did to produce S2B1. With the rapid advances in antibody engineering and combinatorial library production guided by computational modeling, target ligands and mimics are still needed for antibody selection but are no longer required to evoke antibody production. Higher-throughput methods for kinetic analysis are critical for development of recombinant antibodies and other ligand binding molecules, and instrumentation similar to KinExA is likely to evolve rapidly soon. ACKNOWLEDGMENT We thank T. Glass and S. Lackie (Sapidyne Instruments, Boise, ID) for generously supplying the PMMA and PS-DVB beads, instruments, related software, training, and valuable advice throughout the experiments. The interpretations and conclusions herein are not necessarily those of Sapidyne, Inc., and the authors are solely responsible for any errors. C.W. Bell and T.E. Chin also provided valuable discussion and helped to develop experimental protocols. Portions of this work were included in Y.-W.C.’s Ph.D. dissertation, filed at U.C. Berkeley in April 2000. The project was supported by a grant from the Consumer and Environmental Protection Division of the Alameda County, California, District Attorney’s Office. A.E.K. was an investigator in the NIEHS Environmental Health Sciences Center at the University of California, Berkeley (Grant ES01896, Bruce N. Ames, Director).

Received for review February 28, 2001. Accepted August 13, 2001. AC0102462