Structural basis of epitope selectivity and potent protection from malaria by PfCSP antibody L9

A primary objective in malaria vaccine design is the generation of high-quality antibody responses against the circumsporozoite protein of the malaria parasite, Plasmodium falciparum (PfCSP). To enable rational antigen design, we solved a cryo-EM structure of the highly potent anti-PfCSP antibody L9 in complex with recombinant PfCSP. We found that L9 Fab binds multivalently to the CSP minor (NPNV) repeats, which is stabilized by a novel set of affinity-matured homotypic, antibody-antibody contacts. Molecular dynamics simulations revealed a critical role of the L9 light chain in integrity of the homotypic interface, which likely impacts CSP affinity and protective efficacy. These findings reveal the molecular mechanism of the unique NPNV selectivity of L9 and emphasize the importance of anti-homotypic affinity maturation in protective immunity against P. falciparum. One sentence summary The L9 light chain is crucial for potency by conferring multivalent, high affinity binding to the NPNV minor repeats of PfCSP.


Abstract: 24
A primary objective in malaria vaccine design is the generation of high-quality antibody 25 responses against the circumsporozoite protein of the malaria parasite, Plasmodium falciparum 26 (PfCSP). To enable rational antigen design, we solved a cryo-EM structure of the highly potent 27 anti-PfCSP antibody L9 in complex with recombinant PfCSP. We found that L9 Fab binds 28 multivalently to the CSP minor (NPNV) repeats, which is stabilized by a novel set of affinity-29 matured homotypic, antibody-antibody contacts. Molecular dynamics simulations revealed a 30 critical role of the L9 light chain in integrity of the homotypic interface, which likely impacts 31 CSP affinity and protective efficacy. These findings reveal the molecular mechanism of the 32 unique NPNV selectivity of L9 and emphasize the importance of anti-homotypic affinity 33 maturation in protective immunity against P. falciparum As frequently observed in anti-NPNA major repeat mAbs, many direct antigen contacts are with 99 germline-encoded aromatic residues, which in L9 create a hydrophobic cage surrounding the 100 NPNV motif (Fig. 2B, table S2). In particular, W32 L in CDRL1 stacks closely against the N-101 terminal Asn of the NPNV motif (N1) forming a CH-p bond, while Y94 L in CDRL3 engages the 102 repeat Pro (P2) (Fig. 2C, fig. S3, C and D). L9 also utilizes the strictly conserved IGHV3-33 103 germline residue W52 H in CDRH2, which in all structures of IGHV3-33 mAbs solved to date forms 104 a critical CH-p interaction with P6 of the second NPNA repeat in the NPNA2 epitope (7, 8, 11, 105 20). However, in L9, this role is assumed by Y94 L , and W52 H principally acts to stabilize the 106 Y94 L :P2 interaction through a p-p stacking interaction with the Y94 L side chain ( Fig. S3D-F). 107 108 This paratope structure is distinct from most other IGHV3-33 mAbs targeting both major and 109 minor repeats. In L9, a repositioning of the HC and LC CDR3 loops, along with a rearrangement 110 of W52 H and CDRH2, creates a compact, central CSP binding pocket bounded by each of the HC 111 and LC CDRs (Fig. 2B). A somatically mutated residue, R96 L in CDRL3, is found at the base of 112 the pocket and creates a highly basic cavity (Fig. S3G). This basic binding pocket is nearly fully 113 occupied by the N3 side chain, which forms key H-bonds with R96 L (Fig. 2C), while V4 occupies 114 a hydrophobic cavity at the interface of CDRL1, L3, and CDRH3 ( fig. S3C). With this unique 6 CDR conformation, L9 appears optimally disposed to bind the bulkier minor repeat residue V4, 116 which is the only difference between the NPNA and NPNV epitopes. 117

118
Another unique property of L9 is the ability to "crosslink" two NPNV motifs within the minor 119 repeat region of PfCSP, which improves binding affinity (6). Our cryo-EM structure reveals that 120 L9 achieves this through multivalent Fab binding to sequential NPNV repeats stabilized by an 121 extensive antibody-antibody, or homotypic, interface between adjacent Fabs (Fig. 3A). Homotypic 122 interactions have now been identified in several anti-NPNA mAbs and appear to be a characteristic 123 feature of the IGHV3-33 antibody family (8,11,(18)(19)(20). Importantly, we demonstrate L9 as the 124 first non-NPNA targeting anti-PfCSP mAb to also utilize homotypic interactions, suggesting that 125 both the major and minor PfCSP repeats can facilitate their development. 126

127
The L9 homotypic interface is distinct from that observed in NPNA-specific IGHV3-33 mAbs, 128 which is generally conserved and derives primarily from the heavy chain (11, 19) ( fig. S5). In 129 contrast, L9K contributes numerous critical homotypic contacts, and total BSA in the interface is 130 evenly distributed between heavy and light chains (905Å 2 and 839Å 2 , respectively) (Fig. 3, E and 131 F). In the cryo-EM structure, L9K of FabC packs tightly against L9H of FabB, and extensive polar 132 and hydrophobic contacts are made between CDRL1 and the LC framework region 3 of FabC 133 (LFR3) with HFR1, CDRH1, and CDRH3 of FabB ( Fig. 3A; fig. S4; table S3). The interface 134 between FabB and FabA is nearly identical. Importantly, several residues mediating critical 135 homotypic interactions (Fig. 3, B to D) correlate with somatic hypermutation of the germline 136 IGHV3-33 and IGKV1-5 genes ( Fig. 3E and F; fig. S5). Four somatically mutated residues in L9K, 7 F28 L and R31 L in CDRL1, and E68 L and H70 L in LFR3, account for the majority of BSA 138 contributed by the LC to the homotypic interface (Fig. 3E). 139 140 E68 L lies at the core of the homotypic interface in L9, where it forms a key salt bridge with the 141 germline-encoded R94 H of CDRH3B ( Fig. 3B; fig. S4). In L9H, R94 H forms a conserved interaction 142 with Y102 H to stabilize the base of CDRH3; thus E68 L may also indirectly impact antigen binding 143 through stabilization of the CDRH3 loop in the adjacent Fab. F28 L coordinates a series of p-p 144 stacking interactions in the opposing CDRH1B (Y32 H ) and CDRH3B (F96 H and F100c H ) while also 145 packing against the E68 L side chain. This pi network culminates in a cation-p bond between R31 L 146 from CDRL1C and F100c H from the opposing CDRH3B (Fig. 3C). On the other side of the 147 homotypic interface from E68 L , a mutated framework residue H70 L forms a hydrogen bond with 148 the side chain of Q1 H in FabA in addition to multiple van der Waals contacts with CDRH1B (Fig.  149   3D). Each of these homotypic contacts are not encoded in the germline sequence, and none directly 150 contact rsCSP ( fig. S3, A and B). These findings provide strong evidence for affinity maturation 151 to optimize antibody-antibody binding, which may in turn enhance CSP avidity and protective 152 efficacy, as we have shown recently for multiple NPNA-specific IGHV3-33 mAbs (11). 153

154
The four somatic mutations in L9K are atypical: F28 L , E68 L , and H70 L are observed in less than 155 1% percent of all human IGKV1 light chain sequences, while R31 L is observed in only 2% (Fig.  156 S5A) (21). Strikingly, F28 and H70 also correspond to two of the five amino-acid differences 157 between mature L9 and the chimeric L9 mAb F10K/L9H (S28 and D70 in F10K). As F28 and H70 158 both mediate key homotypic interactions in L9, which would likely be lost in F10K, these residues 159 may explain the functional differences of F10K/L9H from L9, namely (1) reduced avidity to CSP 8 minor repeats, (2) loss of the ability to bind two adjacent NPNV repeats, and (3) significantly 161 reduced protection in vivo (p<0.001) (6). 162

163
To test this hypothesis, and to understand the role of homotypic contacts in L9K in general, we 164 used molecular dynamics simulations to characterize WT L9 and a series of L9K variants. L9K 165 residues were reverted to either the germline IGKV1-5 gene (R31S, E68G) or to the L9K precursor 166 F10K (F28S, H70D). We first compared the free energy landscapes of the CDR loops of individual 167 Fv domains unbound to rsCSP ( Fig. 4; fig. S6). We find that the R31S, E68G and H70D mutations 168 in L9K result in a broader conformational space and additional highly probable minima compared 169 to the WT L9 Fv, indicating that these residues are critical for determining the shape and the 170 conformational flexibility of the paratope (Fig. 4, B and C; fig. S6). These minima correspond to 171 a substantial shift away from the binding competent conformation in combination with a higher 172 conformational entropy, suggesting a decrease in stability and/or binding affinity (Fig. 4D). 173 Importantly, when combined (R31S-E68G-H70D), these mutations significantly destabilize the 174 homotypic interface (table S4; p<0.001), substantiating their key role in mediating homotypic 175 interactions. Interestingly, the H70D single mutant stabilizes the homotypic interface (table S4), 176 suggesting the germline E70 or F10K D70 may have initialized the evolution of homotypic 177 interactions during L9 maturation. Unlike other LC mutants, the F28S Fv reveals a similar 178 conformational space and diversity in the CDR loops compared to the WT L9 Fv. However, F28S 179 leads to formation of a new intramolecular salt bridge between residues R31 L and E68 L , with 180 simultaneous loss of the intermolecular salt bridge between E68 L and R94 H and the cation-p bond 181 between R31 L and F100c H (Fig. 4A). Thus, in addition to direct homotypic interactions, F28 acts 182 indirectly through E68 L and R31 L to further stabilize antibody-antibody binding. This is reflected 183 in the significantly decreased interaction energies of the homotypic interface in the F28S mutant 184 relative to WT L9 (table S4) and is visualized in Movie S1. To understand the molecular basis of 185 key functional differences between L9 and F10, we next modelled the F10 chimeras in the context 186 of the trimeric Fab-rsCSP complex. Compared to WT L9 and L9K/F10H, the homotypic interface 187 is strongly destabilized in F10K/L9H (table S4). This suggests that F10K/L9H would not bind 188 multivalently to the minor repeats and would have overall reduced binding affinity, which is 189 consistent with our previous functional data on this chimera (6). Five residues differ between L9K 190 and F10K: F28S, L33V, P40A, H70D, and E90Q (Fig. 3E). We find that the F28S mutation alone 191 accounts for ~80% of the destabilization of the homotypic interface observed with F10K/L9H 192 compared to WT L9, while the H70D single mutant and the L33V-P40A-E90Q triple mutant Fvs 193 both slightly increase stability of the complex (table S4). Taken together, these data suggest that 194 the dramatic destabilization seen in MD simulations of the F10K/L9H chimera is primarily the result 195 of the F28S mutation. Therefore, this rare mutation in L9K (S28F), and the network of homotypic 196 contacts it mediates, may underlie the key functional differences between L9 and F10K/L9H. 197 198 Overall, this study reveals the structural basis for the extraordinary selectivity and binding affinity 199 of L9 for the NPNV minor repeats and highlights the critical role of L9K for both functions. We 200 find that rare, somatically mutated residues in L9K mediate extensive homotypic contacts between 201 adjacent L9 Fabs and thus multivalent binding to adjacent NPNV motifs. These contacts 202 underscore the requirement of at least two NPNV motifs for high affinity CSP binding by L9 (1000 203 nM vs 13 nM for CSP peptides with one and two NPNV, respectively) (6); Based on our recent 204 finding that affinity-matured homotypic interactions in three potent NPNA-specific IGHV3-33 205 mAbs are critical for both high NPNA avidity and protective efficacy (11), it is likely that L9K-206