Multivalency drives the neutralizing activity of antibodies against the Plasmodium falciparum circumsporozoite protein

The repeat region of the Plasmodium falciparum circumsporozoite protein (CSP) is a major vaccine antigen because it can be targeted by parasite neutralizing antibodies; however, little is known about this interaction. We used isothermal calorimetry and X-ray crystallography to analyze the binding of the Plasmodium-neutralizing 2A10 antibody to CSP. Strikingly, we found that the repeat region of CSP is bound by multiple antibodies and that this multivalent interaction drives the affinity of this antibody. Because the CSP protein can cross-link multiple B cell receptors (BCRs) we hypothesized that the B cell response might be T-independent. However, by sequencing the BCRs of CSP-repeat specific cells we found that these cells underwent somatic hypermutation and affinity maturation indicative of a T-dependent response. Interestingly, the BCR repertoire of responding B cells was limited suggesting that the structural simplicity of the repeat may limit the breadth of the immune response. Author Summary Vaccines aim to protect by inducing the immune system to make molecules called antibodies that can recognize molecules on the surface of invading pathogens. In the case of malaria, our most advanced vaccine candidates aim to make antibodies that recognize the circumsporozoite protein molecule on the surface of the invasive parasite stage called the sporozoite. In this report we use X-ray crystallography to determine the structure of CSP-binding antibodies at the atomic level. We use other techniques such as isothermal titration calorimetry to examine how this antibody interacts with the CSP molecule. Strikingly, we found that each CSP molecule could bind 6 antibodies. This finding has implications for the immune response and may explain why high titers of antibody are needed for protection. Moreover because the structure of the CSP repeat is quite simple we determined that the number of different kinds of antibodies that could bind this molecule are quite small. However those antibodies can become quite high affinity as a result of a process called affinity maturation that allows the body to learn how to make improved antibodies specific for pathogen molecules. These data show that while it is challenging for the immune system to recognize and neutralize CSP, it should be possible to generate viable vaccines targeting this molecule.

Attempts to obtain a crystal structure of a complex between 2A10 F AB and the 2 1 7 (NANP) 6 peptide were unsuccessful; unlike binary Ab-Ag interactions, in which the 2 1 8 Ab will bind to a single epitope on an antigen and produce a population of structurally 2 1 9 homogeneous complexes that can be crystallized, in this interaction we are dealing with an intrinsically-disordered peptide, the presence of multiple binding sites 2 2 1 (epitopes) on the peptide and the possibility that more than one 2A10 F AB domain can 2 2 2 bind the peptide, i.e. it is difficult to obtain a homogeneous population of complexes, interactions with the binding-loops, caused the crystals to dissolve, again suggesting 2 2 6 that the heterogeneity of the peptide and the presence of multiple epitopes produces 2 2 7 disorder that is incompatible with crystal formation. Although it was not possible to obtain a crystal structure of the 2A10-2 3 3 (NANP) 6 peptide complex, the accurate structures of the 2A10 F AB fragment, the 2 3 4 (NANP) 6 peptide, and the knowledge that antibodies seldom undergo significant 2 3 5 conformational changes upon antigen binding [28], allowed us to model the interaction, which we tested using site directed mutagenesis. Computational modeling 2 3 7 of Ab-Ag interactions has advanced considerably in recent years and several performed site directed mutagenesis of residues predicted to be important for binding.

4 8
Our model predicted that the interaction with (NANP) 6 would be mainly between In the light chain (Fig. 3A,B), Y38 is predicted to be one of the most pocket that buries a proline residue and is within hydrogen bonding distance, via its 2 5 5 hydroxyl group, to a number of backbone and side-chain groups of the peptide. Loss of this side-chain abolished binding. Y56 also forms part of the same proline-binding 2 5 7 pocket as Y38, and loss of this side-chain also resulted in an almost complete loss of 2 5 8 binding. R109 forms a hydrogen bond to an asparagine residue on the side of the helix; mutation of this residue to alanine results in a partial loss of binding. Y116 is 2 6 0 located at the center of the second proline-binding pocket; since loss of the entire 2 6 1 side-chain through an alanine mutation would lead to general structural disruption of 2 6 2 the F AB fragment, we mutated this to a phenylalanine (removing the hydroxyl group), 2 6 3 which led to a significant reduction in binding. Finally, S36A was selected as a indicated that had no effect on (NANP) n binding. Within the heavy chain (Fig. 3C,D), mutation of N57 to alanine led to 2 6 8 complete loss of binding, which is consistent with it forming a hydrogen bond to a 2 6 9 side-chain asparagine but also being part of a relatively well packed region of the 2 7 0 binding site that is mostly buried upon binding. T66 is located on the edge of the binding site and appears to provide hydrophobic contacts through its methyl group interfaces are known to contribute primarily to specificity rather than affinity [30]. Specifically, the cost of desolvating charged residues such as glutamate is not 2 7 8 compensated for by the hydrogen bonds that may be formed with the binding partner. Y37 is located outside the direct binding site in the apo-crystal structure; the loss of  The multivalency of the CSP repeat region As shown in Fig. 3A and 3C the binding mode of the F AB fragment to the 2 8 7 (NANP) 6 peptide is centered on two proline residues from two non-adjacent NANP- in the core of the F AB antigen binding site, into hydrophobic pockets formed by Tyr38 2 9 0 and Tyr56 of the light chain and the interface between the two chains. In contrast, the 2 9 1 polar asparagine residues on the sides of the helix are involved in hydrogen binding interactions with a number of polar residues on the edge of the binding site, such as 2 9 3 N57 of the heavy chain. Due to the twisting of the (NANP) 6 repeat, the binding 2 9 4 epitope of the peptide is 2.5-3 alternate NANP repeats, with a symmetrical epitope 2 9 5 available for binding on the opposite face (Fig. 4A). Thus, this binding mode is where we observed a stoichiometry of two 2A10 F AB fragments per (NANP) 6 peptide. To investigate whether this binding mode was also compatible with the indication 2 9 9 from ITC that ~10.7 2A10 F AB fragments, or six antibodies (containing 12 F AB 3 0 0 domains) could bind the CSP protein (Table 1), we extended the peptide to its full 3 0 1 length. It is notable that the slight twist in the NANP helix results in the epitope being 3 0 2 offset along the length of the repeat region, thereby allowing binding of ten 2A10 F AB complex and lead to greater affinity (Table 1). Thus, the initially surprising 3 1 0 stoichiometry that we observe through ITC appears to be quite feasible based in the We next set out to determine the implications of our structure for the B cell repeats with streptavidin conjugated phycoerythrin (PE) or allophycocyanin (APC).

0
We used BALB/C mice as this was the strain originally used to generate 2A10. populations, which had class switched ( Fig. 5A and B). In contrast, the number of showed that most B cells present at this time-point were GL7 + CD38indicating that 3 4 4 they are germinal center B cells ( Fig. 5B and D). Given that T cells are required to We next set out to determine the diversity of the B cell response to CSP.

5 1
While the repeat structure of CSP is hypothesized to induce a broad polyclonal immunized mice as well as from 3 naïve mice from which we bulk sorted B cells as controls. We obtained usable sequences from 3 of the 4 mice for both the heavy chain regions dominated the immune response (Fig. 6A). The V regions identified (IGHV1-3 6 2 20; IGHV1-26; IGHV1-34 and IGHV5-9) were generally shared among the mice. As populations. This analysis formally demonstrated that the diversity of the antigen 3 6 6 specific B cells was significantly lower than the diversity of the repertoire in naïve 3 6 7 mice (Fig. 6B). We further found that each V region was typically associated with the typically associated with J4, IGHV5-9 with J4 while in different mice IGHV1-34 was 3 7 0 variously paired with J1 or J4 (Fig. 6C). Similar results were obtained for the kappa IGKV1-117 and IGKV14-111 ( Fig. 6D and E). The V regions were typically paired 3 7 3 with the same J regions even in different mice (Fig. 6F) high throughput sequencing approach is that the degenerate primers only amplified 3 7 7 ~70% of the known IGHV and IGKV sequences in naive mice, suggesting that we published antibody sequences (S2 and S3 Table) that include IGHV-1-20, IGKV5-45 and IGKV1-110 reveals that we are likely capturing the bulk of the antibody diversity.

8 1
Together these data suggest that the number of B cell clones responding to CSP may 3 8 2 be limited, potentially reducing the ability of the immune system to generate effective 3 8 3 neutralizing antibodies. While it is clear that CSP is the target of neutralizing antibodies it has been 3 8 8 suggested that CSP might induce large T-independent responses at the expense of that would be indicative of B cells specific for CSP entering the germinal center.

9 3
Taking advantage of the fact that our kappa chain primers capture the entire V-J 3 9 4 sequences of the antibodies we sequenced we asked: 1) if the kappa chains shared and 2) if the mutations were conserved between different mice indicative of directed showed that these had a much higher degree of mutation than bulk B cells from naïve 3 9 9 mice, demonstrating SHM in the CSP-specific antibodies (Fig. 7A). We further  To directly test if CSP-binding antibodies undergo affinity maturation we binding by the mAb compared to the gAb (Fig. 7C). To identify the specific mutations that were important we introduced the that were shared with the 27E antibody which has previously been found to be 4 2 6 clonally related to 2A10 having been isolated from the same mouse and which shares 4 2 7 the same germline heavy and light chains as the 2A10 mAb [18]. We found that two most of the gain in binding (Fig. 7C). The effect of these antibodies appeared to be 4 3 0 additive rather than synergistic as revealed by experiments in which we introduced 4 3 1 these mutations simultaneously (Fig. 7D). A further mutation close to the light chain N59I and T67F all gave modest increases in binding (Fig. 7E). Collectively our data response. We found that the avidity of 2A10 for the rCSP molecule was in the 4 4 8 nanomolar range, which was much higher than the affinity previously predicted from 4 4 9 competition ELISAs with small peptides [20,21]. This affinity is a consequence of the repertoire to the CSP repeat was limited, perhaps due to the relative simplicity of the 4 5 7 target epitope. However, these antibodies have undergone affinity maturation to Using ITC we determined the affinity of 2A10 for rCSP to be 2.7 nM, which 4 6 1 is not unusual for a mouse mAb. However it is a much higher affinity than that 4 6 2 predicted from competition ELISAs that predicted a micro-molar affinity [20,21].

6 3
However, these competition ELISAs were performed with short peptides rather than between the F AB binding to the peptide or full-length CSP and that of the antibody 4 6 7 appears to be driven by a more favorable enthalpy of binding. It is likely that 4 6 8 additional stabilizing interactions between adjacent F AB domains, which is consistent with the structural model, contribute to this. One caveat of these data is that we used a 4 7 0 slightly truncated repeat, however it is likely that longer repeats will have further 4 7 1 stabilization of the interaction that could result in even higher affinity interaction 4 7 2 between CSP and binding antibodies. Our data provide important insights into the requirements for sporozoite- Indeed, this may be an underestimate and it may be that full-length CSP can antibody to evade this response [37]. It has also been suggested that the CSP repeat Our results uncovering how neutralizing antibodies bind to CSP has several 4 9 3 implications for understanding the development of the immune response to CSP.

9 4
Notably the finding that the CSP molecule can be bound by multiple antibodies/B cell high, but relatively short-lived, titers of anti-CSP antibodies [4,38], which would be The finding of a limited repertoire in the BCR sequences specific for the 5 0 9 (NANP) n repeat contradicts previous suggestions that the response to CSP might be 5 1 0 broad and polyclonal [33]. One explanation for this limited antibody diversity is that that the (NANP) n repeat shares structural similarity with a self-antigen as is clear what this self-antigen might be. One potential outcome of this finding is that if 5 2 0 each B cell clone has a finite burst size this may limit the magnitude of the overall B 5 2 1 cell response. One area for future investigation is to determine the binding modes and limiting their ability to make neutralizing antibodies. This may explain why, while there is a broad correlation between ELISA tires of antibodies to the CSP repeat and While our work has been performed with mouse antibodies, there are major difference between the two species is the considerably more diverse heavy chain 5 3 5 CDR3 regions that are found in human antibodies [43]. In terms of our sequence data, it may be that humans may have a more diverse antibody repertoire, not least because 5 3 7 as larger individuals they may have a greater diversity of naïve B cells specific CSP.

3 8
However, it is notable that all 4 human monoclonal antibodies described to date from repeat epitope. These data support previous suggestions that CSP may be a suboptimal target for vaccination. However, we also find that CSP binding antibodies the IGKV-IGKJ pairings in a representative naïve mouse and 3 immune mice. (NANP) 9 measured by ELISA.