Application Note 4
Epitope Mapping for Binding Site Localization Using Array-Based SPRi Key Highlights
34 mAbs screened for reactivity towards overlapping peptide array of glycoprotein D in a single overnight run
Sorting of linear and conformational binders highly correlative with traditional epitope mapping methods as well as epitope binning studies
Antigens from multiple model species can be incorporated into a single assay to fully leverage multiplex capabilities
Workflow enables screening large numbers of candidates to identify IP-compatible epitopes
Introduction When engineering therapeutic monoclonal antibodies (mAbs), techniques that identify promising candidates early in the drug discovery process often drive the success of a program. Epitope mapping is a technique used to identify the epitope of an antibody–antigen interaction. While epitope binning has emerged as a primary tool for identification of the epitopic diversity of a mAb panel, epitope mapping provides important complimentary information by discerning the regions of interaction on a target. Two common methods of epitope mapping include making an Nmer peptide library with overlapping coverage along the entire primary sequence of the target or using site directed mutagenesis along key residues in the target 1. In this work we have mapped epitopes for a library of mAbs generated against glycoprotein D (gD), the receptor binding glycoprotein of the Herpes Simplex Virus (HSV) that allows for initiation of HSV entry, using a 20mer peptide library with 11 overlapping residues. For the experiments described here, the peptides were first captured on a covalently immobilized biosensor surface. Then, purified mAbs were sequential flowed over the bound 20mer peptide library. These experiments correlated well with previous data for this library of mAbs against gD2-6.
200 μg/ml in running buffer and 110 μl was cycled across the activated surface for 10 min. Reactive esters were quenched by injecting 0.5 M of ethanolamine for 5 min. As a last step prior to undocking the sensor chip, 10% glycerol was passed over the surface for 1 min to improve stabilization of the neutravidin “lawn”.
Capture of biotinylated peptides Overlapping peptides of the HSV Type 2 extracellular domain gD were obtained from Mimotopes (Australia). The 20mer peptides had an 11 residue overlap and an N-terminal biotin. The peptides were prepared to 20 μg/ml in PBS, 0.01% Tween-20 and 70 μl was cycled across the neutravidin sensor surface using the Wasatch Microfluidics Continuous Flow Microspotter (CFM) for 15 min creating a peptide array. Following capture, each spot was washed with PBS, 0.01% Tween-20 for 5 min.
Binding of anti-gD mAbs After redocking the sensor chip in the MX96, the system was primed in PBS, 0.01% Tween-20. Actively
Materials and Methods Neutravidin lawn preparation A CMD200M sensor (Xantec) was docked in the IBIS MX96 SPR imager and primed in a running buffer of 50mM sodium acetate pH 4.5. The surface was activated by injecting a mixture of 0.4 M EDC and 0.1 M sulfo-NHS for 5 min. Neutravidin was prepared to
SPRi epitope mapping assay format
Application Note 4
monitored regions of interest (ROIs) were defined for each location along with interstitial reference spots consisting of neutravidin alone. Mouse monoclonal antibodies known to be reactive to HSV proteins were prepared in running buffer at 100 nM and 110 μl of each was cycled across the sensor surface for 5 min, followed by 2 min of dissociation. Between mAb injections, the surface was regenerated by a 30 s pulse of 50 mM glycine pH 1.5 followed by 100 mM sodium bicarbonate pH 10.
tested, primarily concentrated in the N– and Ctermini of the gD sequence, where gD is known to be less structured and more likely to present linear epitopes.
Data analysis Sensorgrams were referenced and zeroed using the Sprint data pre-processing tool (IBIS). Scrubber2 (BioLogic) was used to extract binding responses for each injection. Microsoft™ Excel was used to build a heat map of binding signals for each mAb.
Results & Discussion Experimental setup Including reagent preparation, time required to make the neutravidin surface was 45 min. Capture of the peptide array required about 15 min to prepare the plate and 20 min to generate the discrete surface spots using the CFM. The injected mAbs required 15 min of preparation and the unattended run across the gD peptide array was 10 hrs using a 5 min association and 2 min dissociation, followed by two regeneration pulses per mAb. Additionally, the stable nature of these peptides allows for the chip to be reused for future studies, eliminating time required to remake the capture lawn.
Figure 1. Se nsorgram s of m A bs inje ct e d across t he gD peptide array. Each cycle along the x-axis represents a unique mAb injection.
Heat map Response unit (RU) intensities for each mAb across the gD peptide array were used to construct a heat map, as shown in Figure 2. Peptides corresponding to the extracellular domain of gD are plotted as columns and injected mAbs are represented as rows. Higher binding responses are represented as reds, while lower responses are represented in blue. For adjacent peptides showing binding, a more intense color indicates a higher binding response and likely
Fully utilizing the spot capacity across the sensor surface would add approximately 20 min to the method for an additional printing step, with no increase in the unattended run time on the MX96. This is due to the “one against many” configuration of the MX96 that allows a single injection to address all 96 sensor locations simultaneously. Utilizing all available sensor spot locations and injecting twice the number of mAbs tested here would still mean the assay could be set up in the afternoon and be complete the following morning.
Sensorgram profiles In total, 34 mAbs were injected across an array of 33 peptides. These mAbs have been extensively characterized for binding and activity in the Cohen/ Eisenberg lab at the University of Pennsylvania and therefore are excellent for validating the SPRi approach to epitope mapping 2-6. In Figure 1 sensorgram profiles over the course of the mAb injections are shown for each spot. Of all mAbs tested, 15 had measurable binding responses against gD peptides. A negative control (mAb C226) was included to monitor non-specific binding. For the gD peptides, 11 were reactive towards the mAbs
Figure 2. He at m ap re pre se nt ing binding of m A bs to array of gD peptides. mAb names are on the y-axis and peptides corresponding to the extracellular domain of gD are along the upper x-axis.
Application Note 4
greater preference for residues focused in that particular sequence. The majority of mAbs showing binding did so against one or two peptides. Two mAbs, BD78 and MC14 however, each bound the same three peptides. Additionally, one of these peptides was located nearer to the N-terminus while the other two peptides were adjacent and localized towards the Cterminus.
Both methods highlight a unique behavior of MC1and support the use of relative binding signals observed by epitope mapping towards differentiating specific sites of binding.
Comparison with epitope binning In an earlier study, epitope binning by competition of mAbs for gD aa1-306 was conducted in the classical format using the same platform as used here for epitope mapping. Broad trends in competitive/noncompetitive profiles showed that mAbs 1D3, 110S, H170, and MC1 all grouped together, demonstrating very similar profiles (Figure 3). Within this group, MC1 was unique in that it did have competitive/noncompetitive behaviors against certain mAbs that the three other members of this community did not. In Figure 4 a dendrogram generated by hierarchical clustering highlights the four mAbs with MC1 being slightly distinct based on its higher point of branching from the other mAbs. Also in Figure 4 are the sensorgrams from the epitope mapping study showing binding to the first two peptides of gD for each of these mAbs. While absolute signals for each mAb may be reflective of a combination of concentration, association kinetics and/or fractional activity, it is interesting that MC1 had a higher relative response for aa10-29 as compared to the other three mAbs. This would imply that although it binds both peptides, more residues that contribute to binding are focused in aa10-29.
Figure 3. Int e grat ion of e pit ope binning and e pit ope mapping data for gD mAbs. Shaded nodes represent groups of mAbs with shared epitope profiles determined by classical sandwich binning. Lines represent blocking relationships between mAbs. Individual mAbs with yellow borders are identified as being linear binders by SPRi; green borders indicate conformational binders.
Figure 4. (A) Dendrogram representation of epitope binning relationships for gD mAbs. Highlighted in yellow are mAbs MC1, 1D3, H170, and 110S which grouped similarly in the assay. The higher branch point of MC1 reflects its nuanced competition profile vs. the other mAbs it clusters with. (B) Sensorgram profiles for mAbs injected across the first two gD N-term peptides. The relative response of MC1 was higher for aa10-29 as compared to the other mAbs, further emphasizing unique epitope.
Application Note 4
Linear vs. conformational epitopes With the exception of C226, all mAbs tested against this peptide array are known binders to HSV gD, however approximately half showed no binding to any of the peptides present, consistent with previous studies identifying these mAbs as being conformational binders 3-5. Figure 5 compares findings from a combination of ELISA and dot blots with those using the SPRi approach. Both strategies show excellent agreement with only linear mAbs being detected by SPRi as would be expected. In this sample set, the dissociation rates observed by SPRi were relatively slow and likely helped maintain the correlation with ELISA. For instances with rapidly dissociating mAb signals, there is the possibility of false negatives in formats such as ELISA when insufficient mAb is present at the measured endpoint
of the assay. The continual monitoring of binding by SPRi circumvents this issue. A challenge with epitope mapping using peptide arrays is that conformational binders are anticipated to not bind in this format, as was seen with the gD mAbs. While regions of amino acids involved in binding cannot be determined for conformational binders, lack of binding would in the very least exclude the possibility of these mAbs being linear binders, particularly if they are known to bind native antigen. Interestingly, in epitope binning studies performed for these same mAbs as shown in Figure 3, there was no evidence of linear and conformational mAbs grouping to the same bin, which suggests mapping using peptide arrays can delineate unique regions of epitope for each mAb, even when no binding is observed in this assay due to secondary structure requirements. In other words, this data suggests selecting linear and conformational mAbs by epitope mapping will likely translate to discrete epitopes in the context of the full length, native antigen.
Alternate assay configurations The assay format described here utilized an array of 33 peptides derived from a single antigen. However, the remaining 62 sensor locations could be utilized in a number of different configurations to maximize information generated per assay. One approach would be to include peptide arrays from model species antigens such as rodents and primates. Cross -reactivity and localization of binding regions would then be determined in parallel. The inclusion of additional species’ peptide arrays on the sensor chip adds minimal time to the assay setup (approximately 20 min of additional capture versus a single print) while yielding valuable information on binding behaviors of the mAbs being investigated. Along these same lines, another option to increase the data gleaned from a single experiment would be to fix peptides from different antigens on the sensor chip and inject mAbs selective for each. Mapping would then be achievable for multiple mAbs against multiple targets in a single run. As peptide arrays typically can handle harsh regeneration conditions, suitable conditions should not be challenging to achieve with multiple antigen arrays immobilized.
Figure 5. Re lat ionship of e pit ope m apping using SPRi or ELISA/dot blot showing excellent agreement. NA = unknown. ELISA and dot blot data adapted from references 3-5.
Lastly, while the workflow presented here evaluates antibodies against an array of antigen peptides, this approach is readily amenable to any configuration where surface immobilized peptides are presented to proteins in solution. The peptide array could be derived from a native ligand with receptor injected across the surface. Alternatively, discrete peptides for either therapeutic or tool purposes could be fixed to the surface to probe binding behaviors towards a target in solution. In this instance, additional characteristics such as binding kinetics could be explored in detail utilizing a format that tracks up to 96 interactions simultaneously.
Application Note 4
Comparison with other epitope localization methodologies Among the many criteria in effectively selecting a therapeutic candidate mAb is navigating the intellectual property (IP) space common for high profile targets. Detailed characterization of engaged epitopes is limited to later stages of discovery when smaller sets of candidates are more amenable to existing methodologies. Approaches such as hydrogen–deuterium exchange mass spectrometry and x-ray crystallography are excellent for discerning the involvement of specific residues, but come with higher sample requirements, restricted throughput, and frequently at a point when programs have already invested heavily in a very limited number of leads.
Epitope mapping by SPRi can address regions of approximately 10 amino acids in length for binding by mAbs. While this does not point to specific residues implicated in binding, it does enable substantially more upfront screening and selection of candidates that may avoid regions of established IP. Though conformational mAbs are a challenge when mapping using peptide arrays, identifying sites for linear binders likely eliminates these regions for conformational mAbs, as was shown by the lack of overlap between linear and conformational mAbs in this study. Additionally, mAb sample requirements of 1 to 2 ug and the relative ease of sourcing biotinylated peptide arrays for an antigen make the reagent requirements of epitope mapping minimal in the context of choosing a suitable subset of leads at the earliest stages possible.
Conclusion Here, we reintroduce epitope mapping using high-throughput Surface Plasmon Resonance imaging (SPRi) as another tool enabling the rapid identification of leads that target functional epitopes which when run early in drug discovery can provide valuable information for the success of a program. In this study, a library of 34 mAbs were flowed over 33 peptides to investigate linear epitopes. The project, completed in 12 hours, generated 1,155 sensorgrams and only required less than 45 min of hands on set up time. We propose with such a short experimental setup and run time that researchers can now conduct these experiments earlier in the drug discovery process. Further, when epitope mapping results are combined with both epitope binning and functional assays, that researchers now have the tools to make much better decisions early in the drug discovery process. Additional opportunities exist by testing multiple antigens in parallel such as the reactivity of human versus mouse or injection of proteins other than mAbs such as a native ligand receptor versus the ligand peptide library. Expansion of these multiplexing approaches to explore epitope granularity during screening will inevitably improve screening results and reduce the high drug attrition rates currently observed with mAbs.
Acknowledgments Special thanks to the Cohen and Eisenberg groups at the University of Pennsylvania for providing mAbs and peptides used in these studies as well as extensive knowledge regarding the complex binding relationships of gD and anti-gD mAbs.
REFERENCES 1. 2. 3. 4. 5. 6. 7.
Baerga-Ortiz, A., et al. Epitope mapping of a monoclonal antibody against human thrombin by H/D-exchange mass spectrometry reveals selection of a diverse sequence in a highly conserved protein. Protein Sci. 11, 1300–1308 (2002). T. M. Cairns, et al. Patient-specific neutralizing antibody responses to herpes simplex virus are attributed to epitopes on either gD, gB, or both and can be type-specific. J. Virol., Jun. 2015. J. C. Whitbeck, et al. Repertoire of epitopes recognized by serum IgG from humans vaccinated with herpes simplex virus 2 glycoprotein D. J. Virol., vol. 88, no. 14, pp. 7786–95, Jul. (2014). E. Lazear, J. C., et al. Antibody-Induced Conformational Changes in Herpes Simplex Virus Glycoprotein gD Reveal New Targets for Virus Neutralization. J. Virol., vol. 86, no. 3, pp. 1563–76, Feb. 2012. V. J. Isola, et al. Fine mapping of antigenic site II of herpes simplex virus glycoprotein D.. J. Virol., vol. 63, no. 5, pp. 2325–34, May 1989. G. H. Cohen, et al. Localization and synthesis of an antigenic determinant of herpes simplex virus glycoprotein D that stimulates the production of neutralizing antibody. J. Virol., vol. 49, no. 1, pp. 102–8, Jan. 1984. Y. N. Abdiche. High-Throughput Epitope Binning Assays on Label-Free Array-Based Biosensors Can Yield Exquisite Epitope Discrimination That Facilitates the Selection of Monoclonal Antibodies with Functional Activity. PLoS One, vol. 9, no. 3, p. e92451, Jan. 2014. 5 Wasatch Microfluidics, Inc. 825 N 300 W Suite C309 Salt Lake City, UT 84103 www.microfl.com 801.532.4486