Surface plasmon resonance screening of synthetic peptides mimicking the immunodominant region of C-S8c1 foot-and-mouth disease virus

Surface plasmon resonance screening of synthetic peptides mimicking the immunodominant region of C-S8c1 foot-andmouth disease virus Paula Gomes, Ernest Giralt, David Andreu* Department of Organic Chemistry, University of Barcelona, MartõÂi Franque Ás, 1; 08028 Barcelona, Spain Abstract The main antigenic site (site A) of foot-and-mouth disease virus (FMDV, strain C-S8c1) may be adequately reproduced by a 15-peptide with the amino acid sequence H-YTASARGDLAHLTTT-NH 2 (A15), corresponding to the residues 136±150 of the viral protein VP1. The eect of amino acid substitutions within A15 on its antigenicity towards monoclonal antibodies (MAb) raised against antigenic site A, has been studied by means of BIAcore technology, based on surface plasmon resonance (SPR). Although these antigenicities have previously been determined from enzyme-linked immunosorbent assays (ELISA), the SPRbased technique is superior in that it allows a fast and straightforward screening of antigens while simultaneously providing kinetic data of the antigen±antibody interaction. With a view to screening fairly large libraries of individual peptides, we have inverted the typical SPR experiment by immobilizing the MAb on the sensor surface and using peptides as soluble analytes. We report the validation of this approach through the screening of 44 site A peptides, with results generally in good agreement with the relative antigenicities previously determined by competition ELISA. #1999 Elsevier Science Ltd. All rights reserved. Keywords: Biosensor; small peptide analytes; Foot-and-mouth disease virus 1. Introduction The study of biospeci®c macromolecular interactions is central to an understanding of the molecular mechanisms involved in viral infection and may help in the rational design of synthetic vaccines. The antigen±antibody reaction involves a series of molecular events that can be analyzed by dierent biophysical methods to provide qualitative or quantitative data on the binding reaction. Optical spectroscopies, particularly those based on ¯uorescence, have been widely employed so far, but with the introduction of surface plasmon resonance (SPR) in the early '90s, optical evanescent wave biosensors have become increasingly popular for characterization of reversible biospeci®c reactions [1± 9]. SPR is an optical detection method that allows to monitor macromolecular interactions directly, in realtime and in a label-free mode, in contrast with indirect methods such as enzyme-linked immunosorbent assays (ELISA) or ¯uorescence spectroscopies requiring suitable tags. In addition, the ability to provide a realtime analysis means that not only anity data but also kinetic data can be obtained. Another attractive feature is that high purity or large amounts of the biomolecules are not strictly required [1,7±10]. The technique relies on the covalent immobilization of one interactant (the ligand) onto a modi®ed dextran-coated gold surface, which forms one wall of a ¯owcell [11,12]. A solution of the other interactant (the ligate or analyte) is injected over this surface at continuous ¯ow. Monochromatic p-polarized light is directed at the sensor surface and biomolecular binding events are detected as changes in the speci®c angle where SPR creates extinction of light [1,7]. These changes re¯ect refractive index changes in the solution close to the Vaccine 18 (2000) 362±370 0264-410X/99/$ - see front matter #1999 Elsevier Science Ltd. All rights reserved. PII: S0264-410X(99)00206-6 www.elsevier.com/locate/vaccine * Corresponding author. Tel.: +34-93-402-1260; fax: +34-93-4021260. E-mail address: [email protected] (D. Andreu) surface, caused by the increase/decrease of mass due to analyte association/dissociation to the immobilized ligand. Therefore, measuring continuously the SPR angle provides a full record (a sensorgram) of the progress of the biospeci®c interaction. Our research has been focused for some time on synthetic peptide vaccines against foot-and-mouth disease virus (FMDV), which causes the economically most important disease of cattle [13±18]. The high genetic variability of FMDV is responsible for the antigenic diversity among FMDV isolates and even within isolate subpopulations [13,16±19]. This fact is an important obstacle in the design of eective synthetic epitope-based vaccines. Ideally, such vaccines should include multiple epitopes to prevent the selection of resistant mutants in the ®eld, and the eects of amino acid substitutions in infection-involved epitopes should be well characterized at the molecular level. We have directed our attention to the various overlapping linear B-cell epitopes within antigenic site A [13±15,17], located on the G±H loop of FMDV. This site can be adequately reproduced by peptide A15 (HYTASARGDLAHLTTT-NH 2 ), corresponding to residues 136±150 of viral protein VP1. We have previously described the eect of systematic single-residue changes on the antigenicity of site A [18]. Since we wished to characterize these antigenic responses not only at equilibrium but also in kinetic terms, SPR was the method of choice. However, given the large number of variant peptides, the only practical way to perform the screening consists of monoclonal antibody (MAb) immobilization and injection of peptide analytes. As direct detection of analytes smaller than 5000 Da is usually unfeasible with standard BIAcore protocols [20,21], we have carried out an extensive optimization of this approach using 44 of the 240 A15 single mutants as analytes and anti-site A MAb SD6 as ligand. The generally good agreement found between biosensor data and the anities known from ELISA [18] validates our analytical method. 2. Experimental procedures 2.1. Materials MAb SD6 stock solution (1.6 mg/ml in PBS, pH 7.3) was obtained as described [14] by Dra. Nuria Verdaguer (CSIC, Barcelona). The antibody was desalted and buer-exchanged on Nap-5 Sephadex G25 columns (Pharmacia) and the eluate quantitated by the Pierce BCA protein assay method (Pierce). The synthesis and puri®cation of A15 analogues has been previously described [18]. Consumables for the BIAcore 1000 instrument (CM5 certi®ed sensor chips, HBS buer, the amine coupling reagents N-hydroxysuccinimide (NHS) and N-ethyl-N'-dimethylaminopropyl-carbodiimide (EDC), ethanolamine) as well as the BIAevaluation 3.0.1 software were all from Biosensor AB (Uppsala, Sweden). 2.2. Biosensor experiments 2.2.1. M ab pre-concentration and immobilization SD6 solutions (100 and 50 mg/ml, in either 10 mM acetate, pH 5.0 or 5 mM maleate, pH 6.0) were separately injected (30 ml) at 5 ml/min over a non-activated sensor surface in order to determine which gave the most ecient pre-concentration of Mab into the dextran matrix. Three SD6 surfaces were prepared using the standard amine coupling procedure as described by the manufacturer. Each carboxymethyl surface was activated with a 35 ml injection (at 5 ml/min) of a NHS/EDC mixture and SD6 was then coupled at three dierent densities by exposing each surface to 35 ml injections of SD6 in 10 mM acetate buer pH 5.0 at concentrations of 50, 5 and 3 mg/ml, respectively. Nomenclature FMDV foot-and-mouth disease virus SPR surface plasmon resonance ELISA enzyme-linked immunosorbent assay MAb monoclonal antibody NHS N-hydroxysuccinimide EDC N-ethyl-N'-dimethylaminopropylcarbodiimide RU resonance units k s apparent kinetic rate constant k a association kinetic rate constant k d dissociation kinetic rate constant Rbiosensor response (in RU) at a given time R max maximum response (in RU) for a given surface capacity R immob ligand immobilization response R eq extrapolated response (in RU) for the equilibrium A analyte L ligand C A analyte concentration K A association equilibrium constant w 2 chi squared MW molecular weight P. Gomes et al. / Vaccine 18 (2000) 362±370 363 Unreacted activated groups were blocked by a 30 ml injection of ethanolamine and remaining non-covalently bound molecules were washed o in a regeneration step with a 3-min pulse of 100 mM HCl. The ®nal densities of immobilized SD6 were 8000, 1700 and 800 RU, respectively (1000 RU corresponding to a 1 ng/mm 2 surface density). 2.2.2. Optimization of the experimental parameters A few sets of experiments, using A15 as analyte, were run on the three SD6 surfaces at dierent peptide concentrations (ranging from 1 to 2440 nM) and ¯ow rates (5 and 60 ml/min). A 2.44 mM solution of peptide in 100 mM acetic acid was used as stock, from which assay samples were prepared by 1000-fold and subsequent serial dilutions in HBS. All experiments were done in HBS running buer at 258C, using the `kinjection' mode in order to obtain accurate kinetic data by decreasing sample dispersion at the injection plugs. To generate the sensorgrams peptide injections during 7 min followed by 6 min of dissociation in running buffer were used, prior to the regeneration cycle (2-min pulse with 100 mM HCl). 2.2.3. Kinetic analysis of A15 analogues Once suitable experimental conditions for kinetic screening of site A peptides had been established (see Section 4), a systematic analysis of 43 A15 analogues was performed, in order to compare biosensor kinetic data with anities previously determined by competition ELISA. This analysis was done on a medium density SD6 surface (1700 RU of immobilized MAb), at peptide concentrations ranging from ca. 2500 to 150 nM. Stock peptide solutions were quantitated by amino acid analysis. Sensorgrams were generated at a 60 ml/min ¯ow rate and `kinjections' consisted on a 90 s association step, followed by a 240 s dissociation in running buer. A 90 s pulse of 100 mM HCl was applied to regenerate the surface at the end of each cycle. A pentadecapeptide, A15scr, containing the constituient amino acids of A15 in a random sequence (HRAGTATTLADLHYST-NH 2 ) was injected under the same conditions as a control for instrument artifacts such as refractive index changes and non-speci®c binding. 2.3. Data analysis Biosensor data were prepared, modelled and ®tted by means of the BIAevaluation 3.0.1. software. Calculations are carried out by numerical integration [22] and global curve ®tting is done by non-linear least-squares analysis [23] applied simultaneously to the entire data set [10,24]. The quality of the ®tted data can be evaluated by comparison between the modeled and the experimental sensorgrams, as well as by statistical parameters such as the chi squared, given by Eq. (1), where r f is the ®tted value at a given point, r x is the experimental value at the same point, nis the number of data points and pis the number of degrees of freedom. w2 X n 1 rfÿrx2 nÿp1 The reaction models and correspondent dierential rate equations tested are presented in Table 1. 3. Results and discussion 3.1. Optimization of the experimental setup In our ®rst approach, we tried to overcome the detection problems to be expected from the low molecular weight (MW) of our analytes by using a very dense SD6 surface (8 ng/mm 2 ) and high peptide concentrations at a low ¯ow rate (5 ml/min). The sensorgrams generated under these conditions (Fig. 1a) could not be ®tted to the 1:1 bimolecular model (Table 1) as inferred from the high and non-random residuals obtained (Fig. 1b) and from the fact that the association rate constant was concentration-dependent (Fig. 1c). Fitting the data to the heterogeneous ligand model (Table 1), the residuals were lower and randomly distributed, but the values for k a1 and k a2 were still concentration-dependent (data not shown). Since previous X-ray diraction crystallographic studies have suggested that SD6 binds to FMDV C-S8c1 through a single Fab fragment [14,18], we expected a 1:1 SD6/ A15 interaction. Therefore, the inconsistency of the kinetic data was attributed to the high SD6 density employed, which generated signi®cant heterogeneity in ligand accessibility and orientation, as well as favoured diusion-controlled delivery of analyte to the most hindered SD6 molecules [25]. Analysis of a broad range of A15 concentrations at the same ¯ow rate but over less dense (1.7 and 0.8 ng/mm 2 ) SD6 surfaces did Table 1 Reaction schemes and dierential rate equations tested. Symbols are described in nomenclature. Subscripts 1 and 2 refer to the parallel binding events in two dierent ligand subpopulations One-to-one reaction: A+L \AL dR/dt=k a C A (R max ÿR)ÿk d R Parallel reactions (ligand heterogeneity): A+L 1 \AL 1 A+L 2 \AL 2 dR 1 /dt=k a1 C A (R max1 ÿR)ÿk d1 R dR 2 /dt=k a2 C A (R max2 ÿR)ÿk d2 R P. Gomes et al. / Vaccine 18 (2000) 362±370364 Fig. 1. Biosensor analysis of A15 injected over a 8 ng/mm 2 SD6 surface, at a 5 ml/min buer ¯ow rate: (a) sensorgrams (78, 156, 305, 610 and 1220 nM); (b) residuals (approximately 212 RU); (c) dependence of k a on analyte concentration. P. Gomes et al. / Vaccine 18 (2000) 362±370 365 Fig. 2. Biosensor analysis of A15 injected over a 1.7 ng/mm 2 SD6 surface, at a 60 ml/min buer ¯ow rate: (a) sensorgrams (156, 305, 610, 1220 and 2440 nM); (b) residuals (approximately 20.5 RU); (c) linearity of k s on analyte concentration; (d) R eq versus analyte concentration plot. P. Gomes et al. / Vaccine 18 (2000) 362±370366 not improve the results, the former surface producing equally inconsistent data and the latter being useless given the extremely low signal-to-noise ratios (data not shown). Therefore, and using again a medium-density SD6 surface (1.7 ng/mm 2 ) and high peptide concentrations (150 to 2440 nM), we decided to increase the buer ¯ow rate up to 60 ml/min to obviate the diusion-controlled kinetics which seemed the most probable source for the persisting deviations from the expected Langmuirian behaviour. Indeed, consistent and apparently reliable data were now obtained, with a random distribution of residuals within an interval of ca. 20.5 RU, except for injection plugs (Fig. 2b). Linearity of k s versus peptide concentration (Eq. (2)) over the 32fold concentration range was observed, thus k a was concentration independent (Fig. 2c). kskaCAkd2 Modeled and experimental curves were virtually superimposable (Fig. 2a) and a w 2 smaller than 1 was obtained. Data self-consistency was further con®rmed by the total agreement between the values for the equilibrium association constant, K A , as obtained from the k a /k d ratio or from the plot of R eq versus peptide concentration (Eq. (3), Fig. 2d) for a 1:1 interaction. From this same plot, a value of R max close to that predicted from the immobilization response, R immob (Eq. (4)) was obtained, thus indicating that the MAb was fully active. Req KACARmax KACA13 Rmax Rimmob MWL MWA 4 Table 2 Kinetic data of the interaction between MAb SD6 and 44 site A peptides, corrected for bulk eects (see text). Qualitative relative antigenicities obtained in ELISA competition assays [18] are represented, with a black box corresponding to IC 50 >100, a dark grey box to IC 50 =30 to 100, a light grey box to IC 50 =5 to 30 and a white box to IC 50 < 5. no, no interaction; nr, non-reliable data P. Gomes et al. / Vaccine 18 (2000) 362±370 367 3.2. Systematic analysis of site A peptides Having found suitable experimental conditions for the kinetic analysis of the A15/SD6 interaction, we applied a similar protocol to the systematic screening of 43 other A15 analogues. Reduction of `kinjection' times (90 s for association, 240 s for dissociation) lowered sample, buer and time consumption, with no loss in sensorgram quality or kinetic information. The consistency and accuracy of the ®tted kinetic data obtained for the whole set of site A peptides were in every aspect similar to those described for A15 under the same conditions. The stability of SD6 surfaces to the repeated acidic regeneration cycles allowed us to screen the entire set over the same surface, without any detectable loss in MAb activity, thus providing reliable comparison among the dierent peptides. The scrambled 15-mer sequence A15Scr had no apparent speci®c binding, but gave rise to a high bulk response, as observed for the other peptides analysed. Therefore, the curves obtained for each site A peptide were corrected by subtraction of the corresponding A15Scr sensorgrams. A discrete overall improvement in data quality (lower w 2 values) accompanied this correction and a general increase in K A was obtained. This anity increase re¯ected simultaneously an increase in k a and a k d decrease (data not shown). These ®ndings are consistent with the higher probability of analyte non-speci®c retention within the dextran matrix when high concentrations are used [8,27]. This binding was probably a `response averaging' of the biospeci®c union with the non-speci®c analyte imprisonment. The corrected biosensor data for the 44 site A peptides analysed are shown in Table 2, in which relative antigenicities derived from ELISA [18] are also represented. The general agreement observed between both techniques validates our analytical method for the antigenicity screening of low MW peptides using biosensor technology. 4. Concluding remarks This investigation has served three major purposes: 1. It shows that BIAcore technology is a powerful tool to monitor and characterize antigen±antibody interactions even when low MW analytes are used. This provides a new basis for fast and straightforward antigenicity ranking of small synthetic peptide antigens. 2. It emphasizes the need for careful experimental setup and optimization to obviate the problems usually found in quantitative kinetic biosensor analysis. 3. It demonstrates the correspondence between peptide±antibody anities as obtained by BIAcore or ELISA experiments. Although evanescent wave biosensors are successful [1±9] in that they provide simple and fast (`yes' or `no') answers about the binding of analytes to a ligand, the quantitative data obtained from them are subject to several potential artifacts [7,8,25±28]. The most common sources for such artifacts are heterogeneity in ligand accessibility and orientation, as well as masstransport limitations. Methods for diagnosis and evaluation of these artifacts, such as global analysis of data sets generated in dierent surface capacities or self-consistency tests, are extensively discussed in the literature [24±32]. Nevertheless, these control tests, though helpful to ensure internally consistent data, provide no critical information about the validity of the kinetic parameters as absolute values. We minimized the undesired artifacts by adapting our experimental setup to high buer ¯ow rate and the minimum surface capacity possible and thus obtained self-consistent kinetic and equilibrium data, further con®rmed by previous ELISA experiments [18]. However, caution is still advisable in the evaluation of such kinetic data. Firstly, one cannot totally compare the events taking place in a biosensor experiment with those occurring in physiologic media. It must not be forgotten that although no labels are required in biosensor analysis, the ligand is still somehow `labelled' by its attachment to a sensor chip. Secondly, while ELISA provides a valuable check of the reliability of biosensor data, one cannot write o the possibility that mass-transport aects actual k a and k d values by a similar factor, thus providing thermodynamic constants apparently consistent with equilibrium experiments. In fact, a preliminary analysis of the A15/SD6 interaction using IAsys biosensor (a resonant mirror evanescent wave biosensor where a stirred cuvette, instead of a ¯owcell, is claimed to overcome mass-transport limitations) showed virtually the same anity constant but slightly higher kinetic rate constants [33]. Further experiments in progress include using other anti-site A MAbs as ligands and a high MW analyte (a 1:1 conjugate of A15 with a carrier protein) for real-time competitive analysis [20]. These experiments will hopefully provide further information on the validity of our methodology. Acknowledgements We thank Dr. Maria Luz Valero for supplying the synthetic pentadecapeptides, Dr. Esteban Domingo and Dr. Mauricio Mateu (CBM-CSIC, Madrid) for the ELISA competition experiments and Dr. Nuria P. Gomes et al. / Vaccine 18 (2000) 362±370368 Verdaguer for supplying the MAb SD6. We are grateful to the Scienti®c-Technical Services of the University of Barcelona for providing the BIAcore 1000 instrument, BIAevaluation 3.0.1 software and all the BIAcerti®ed reagents and materials. Work supported by grants from the European Union (FAIR PL97-3577), DGICYT (Spain, grants PB94-0845 and PB97-0873) and Generalitat de Catalunya (CERBA). P.G. thanks the Fundac°ao Calouste Gulbenkian (Lisbon, Portugal) for her Ph.D. grant and the Oporto University (Portugal) for a temporary leave from teaching duties. References [1] Fagerstam LG, Frostell-Karlsson A, Karlsson R, Persson B, Ronnberg I. Biospeci®c interaction analysis using SPR detection applied to kinetic binding site and concentration analysis. J Chromatogr 1992;597:397±410. [2] Brigham-Burke M, Edwards JR, O'Shannessy DJ. Detection of receptor±ligand interactions using SPR: model studies employing the HIV-1 gp120/CD4 interaction. Anal Biochem 1992;205:125±31. [3] Altschuh D, Dubs MC, Weiss E, Zeder-Lutz G, Van Regenmortel MHV. Determination of kinetic constants for the interaction between a monoclonal antibody and peptides using SPR. Biochemistry 1992;31:6298±304. [4] Van Regenmortel MHV, Altschuh D, Pellequer JL, RichaletSe Âcordel P, Saunal H, Wiley JA, Zeder-Lutz G. Analysis of viral antigens using biosensor technology. Methods Comp Methods Enzymol 1994;6:177±87. [5] Lasonder E, Schellekens GA, Koedijk DGAM, Damhof RA, Welling-Wester S, Feijlbrief M, Scheer AJ, Welling GW. Kinetic analysis of synthetic analogues of linear-epitope peptides of glycoprotein D of herpes simplex virus type I by SPR. Eur J Biochem 1996;240:209±14. [6] Zeder-Lutz G, Zuber E, Witz J, Van Regenmortel MHV. Thermodynamic analysis of antigen±antibody binding using biosensor measurements at dierent temperatures. Anal Biochem 1997;246:123±32. [7] Schuck P. Use of SPR resonance to probe the equilibrium and dynamic aspects of interactions between biological macromolecules. Ann Rev Biophys Biomol Struct 1997;26:541±66. [8] Garland PB. Optical evanescent wave methods for the study of biomolecular interactions. Q Rev Biophys 1996;29(1):91±117. [9] Lakey JH, Ragget EM. Measuring protein±protein interactions. Curr Opin Struct Biol 1998;8:119±23. [10] Malmqvist M, Karlsson R. Biomolecular interaction analysis: anity biosensor technologies for functional analysis of proteins. Curr Opin Chem Biol 1997;1:378±83. [11] Lofas S, Johnsson B. A novel hydrogel matrix on gold surface in SPR sensors for fast and ecient covalent immobilization of ligands. J Chem Soc Chem Commun 1990;1526±1528. [12] Johnsson B, Lofas S, Lindquist G. Immobilization of proteins to a carboxylmethyl dextran modi®ed gold surface for biospeci- ®c interaction analysis in SPR sensors. Anal Biochem 1991;198:268±77. [13] Carren Äo C, Roig X, Cairo ÂJ, Camarero J, Mateu MG, Domingo E, Giralt E, Andreu D. Studies on antigenic variability of C strains of FMDV by means of synthetic peptides and monoclonal antibodies. Int J Peptide Protein Res 1992;39:41±7. [14] Verdaguer N, Mateu MG, Andreu D, Giralt E, Domingo E, Fita I. Structure of the major antigenic loop of FMDV complexed with a neutralizing antibody: direct involvement of the Arg±Gly±Asp motif in the interaction. EMBO J 1995;14(8):1690±6. [15] Mateu MG, Camarero JA, Giralt E, Andreu D, Domingo E. Direct evaluation of the immunodominance of a major antigenic site of FMDV in a natural host. Virology 1995;206:298±306. [16] Feigelstock DA, Mateu MG, Valero ML, Andreu D, Domingo E, Palma EL. Emerging FMDV variants with antigenically critical amino acid substitutions predicted by model studies using reference viruses. Vaccine 1996;14(2):97±102. [17] Mateu MG, Valero ML, Andreu D, Domingo E. Systematic replacement of amino acid residues within an Arg±Gly±Aspcontaining loop of FMDV and eect on cell recognition. J Biol Chem 1996;271(22):12814±9. [18] Verdaguer N, Sevilla N, Valero ML, Stuart D, Brocchi E, Andreu D, Giralt E, Domingo E, Mateu MG, Fita I. A similar pattern of interaction for dierent antibodies with a major antigenic site of FMDV: implications for intratypic antigenic variation. J Virol 1998;72(1):739±48. [19] Domingo E, Mateu MG, MartõÂnez MA, Dopazo J, Moya A, Sobrino F. Genetic variability and antigenic diversity of footand-mouth disease virus. In: Kurstak K, Marusyk RG, Van Regenmortel MHV, editors. Applied Virology Research, vol. 2. Plenum Publishing Corporation. 1990. p. 233±265. [20] Karlsson R. Real-time competitive kinetic analysis of interactions between low-molecular-weight ligands in solution and surface-immobilized receptors. Anal Biochem 1994;221:142±51. [21] Karlsson R, Stahlberg R. SPR detection and multispot sensing for direct monitoring of interactions involving low-molecularweight analytes and for determination of low anities. Anal Biochem 1995;228:274±80. [22] Morton TA, Myszka DG, Chaiken IM. Interpreting complex binding kinetics from optical biosensors: a comparison of analysis by linearization, the integrated rate equation and numerical integration. Anal Biochem 1995;227:176±85. [23] O'Shannessy DJ, Brigham-Burke M, Soneson KK, Hensley P, Brookes I. Determination of rate and equilibrium binding constants for macromolecular interactions using SPR: use of nonlinear least squares analysis methods. Anal Biochem 1993;212:457±68. [24] Karlsson R, Falt A. Experimental design for kinetic analysis of protein±protein interactions with SPR biosensors. J Immunol Methods 1997;200:121±33. [25] O'Shannessy DJ, Winzor DJ. Interpretation of deviations from pseudo-®rst-order kinetic behavior in the characterization of ligand binding by biosensor technology. Anal Biochem 1996;236:275±83. [26] Hall DR, Cann JR, Winzor DJ. Demonstration of an upper limit to the range of association rate constants amenable to study by biosensor technology based on SPR. Anal Biochem 1996;235:175±84. [27] Schuck P, Minton AP. Analysis of mass transport-limited binding kinetics in evanescent wave biosensors. Anal Biochem 1996;240:262±72. [28] Oddie GW, Gruen LC, Odgers GA, King LG, Kortt AA. Identi®cation and minimization of non-ideal binding eects in BIAcore analysis: ferritin/anti-ferritin Fab'interaction as a model system. Anal Biochem 1997;244:301±11. [29] Bowles MR, Hall DR, Pond SM, Winzor DJ. Studies of protein interactions by biosensor technology: an alternative approach to the analysis of sensorgrams deviating from pseudo-®rst-order kinetic behavior. Anal Biochem 1997;244:133±43. [30] Edwards PR, Leatherbarrow RJ. Determination of association rate constants by an optical biosensor using initial rate analysis. Anal Biochem 1997;246:1±6. [31] Myszka DG, Morton TA, Doyle ML, Chaiken IM. Kinetic P. Gomes et al. / Vaccine 18 (2000) 362±370 369 analysis of a protein±antigen interaction limited by mass transport on an optical biosensor. Biophys Chem 1997;64:127±37. [32] Schuck P, Minton AP. Kinetic analysis of biosensor data: elementary tests for auto-consistency. Trends in Biochem Sci 1996;21:458±60. [33] Gomes P, Giralt E, Andreu D, unpublished results. P. Gomes et al. / Vaccine 18 (2000) 362±370370