Structural basis of Plasmodium vivax inhibition by antibodies binding to the circumsporozoite protein repeats

Malaria is a global health burden, with Plasmodium falciparum (Pf) and Plasmodium vivax (Pv) responsible for the majority of infections worldwide. Circumsporozoite protein (CSP) is the most abundant protein on the surface of Plasmodium sporozoites, and antibodies targeting the central repeat region of CSP can prevent parasite infection. Although much has been uncovered about the molecular basis of antibody recognition of the PfCSP repeats, data remains scarce for PvCSP. Here, we performed molecular dynamics simulations for peptides comprising the PvCSP repeats from strains VK210 and VK247 to reveal how the PvCSP central repeats are highly disordered, with minor propensities to adopt turn conformations. Next, we solved eight crystal structures to unveil the interactions of two inhibitory monoclonal antibodies (mAbs), 2F2 and 2E10.E9, with PvCSP repeats. Both antibodies can accommodate subtle sequence variances in the repeat motifs and recognize largely coiled peptide conformations that also contain isolated turns. Our structural studies uncover various degrees of Fab-Fab homotypic interactions upon recognition of the PvCSP central repeats by these two inhibitory mAbs, similar to potent mAbs against PfCSP. These findings augment our understanding of host-Plasmodium interactions and contribute molecular details of Pv inhibition by mAbs to unlock structure-based engineering of PvCSP-based vaccines.


Introduction
Malaria is a major public health concern, with an estimated 409,000 deaths in 2019 (World Health Organization, 2020). Human malaria is caused by Plasmodium parasites, with the majority of cases attributed to Plasmodium falciparum (Pf) and Plasmodium vivax (Pv) (Lover et al., 2018). Pv is the predominant Plasmodium spp. in circulation for a majority of countries outside of Africa (~75% of cases in South and North America, ~50% of cases in the Southeast Asia region, and ~30% in the Eastern Mediterranean region; World Health Organization, 2020). Despite overall lower mortality compared to Pf malaria, Pv infection can cause debilitating disease, including fever, myalgia, chronic anemia, reduced birthweight, and increased risk of neonatal death (Alexandre et al., 2010;Bardají et al., 2017;Genton et al., 2008).
Circumsporozoite protein (CSP) is the most abundant protein on the surface of all Plasmodium sporozoites and is necessary for parasite development and infection (Cerami et al., 1992;Ménard et al., 1997;Nguitragool et al., 2017). CSP contains an unusual central region consisting of multiple, short amino acid repeats whose sequence depends on the Plasmodium species (Chenet et al., 2012;Rich et al., 2000;Tahar et al., 1998). The PvCSP central region is composed of nonapeptides GDRA(A/D)GQPA and ANGAGNQPG characteristic of strains VK210 and VK247, respectively (Arnot et al., 1985;Rosenberg et al., 1989; Figure 1A). Unlike the 4-amino acid (aa)-long motifs of Pf and Plasmodium berghei (Pb) CSP, which are rich in asparagine and proline residues, PvCSP repeats are longer and consist primarily of glycine and alanine residues (~50% of all residues). Moreover, ~26% of the PvCSPvk210 central repeat region consists of charged residues, including arginine and aspartic acid residues. Both VK210 and VK247 strains have worldwide distribution (Cheng et al., 2013;Kain et al., 1992;Soares et al., 2020), and VK210 appears to be a major target of the humoral immune response in studied populations (González et al., 2001;Kim et al., 2010;Soares et al., 2020).
Although not much is known about the molecular basis of PvCSP recognition by inhibitory antibodies, a positive statistical association between the level of antibodies against the repeat region of PvCSP and protection has previously been established (Yadava et al., 2014). Anti-PvCSP speciesspecific monoclonal antibodies (mAbs) 2F2 and 2E10.E9 were generated in mice after immunization with radiation-attenuated Pv sporozoites of strains VK210 and VK247, respectively (Nardin et al., 1982). Incubation of Pv sporozoites with mAbs 2F2 and 2E10.E9 results in significantly reduced sporozoite infectivity, and thus despite a limited molecular understanding of their recognition, both antibodies have been valuable research tools in studies of Pv sporozoites (Cabrera-Mora et al., 2015;Gimenez et al., 2017;Miyazaki et al., 2020;Roth et al., 2018;Teixeira et al., 2014). Interestingly, a recent study reported that sporozoites attenuated with low concentrations of mAb 2F2 were significantly reduced in size and had lower DNA content, indicating post-hepatocyte-invasion antibody inhibition of liver stage development (Roth et al., 2018).
Here, we present a detailed molecular analysis of the PvCSP repeat region and its recognition by inhibitory mAbs 2F2 and 2E10.E9. Molecular dynamics (MD) simulations on PvCSP-derived repeat peptides indicate that in the absence of interacting mAbs, the PvCSP repeat is largely disordered. Our structural studies reveal how mAbs 2F2 and 2E10.E9 lock PvCSP repeat peptides in a predominant coiled conformation, with antibody germline-encoded aromatic residues contributing significantly to the antigen contacts. Moreover, we describe how mAb 2E10.E9 engages in head-to-head homotypic interactions when targeting PvCSP in a similar manner as previously described human mAbs against PfCSP (Imkeller et al., 2018;Oyen et al., 2018;Pholcharee et al., 2021) and a murine mAb against PbCSP (Kucharska et al., 2020).

PvCSP repeat peptides are structurally disordered and behave like harmonic springs
Due to the variety of PvCSP repeat sequence motifs, we created an extensive list of peptides for circular dichroism (CD) spectroscopy and MD simulations studies (Supplementary file 1, Figure 1A) to examine the structural propensities of 18-and 27-aa-long motifs from the PvCSPvk210 and PvCSPvk247 repeats. CD spectra of all analyzed peptides were indicative of a lack of secondary structure, with minima at ~200 nm ( Figure 1B).
For a finer dissection of potential minor secondary structure propensities, we performed all-atom MD simulations on 27-aa peptides derived from PvCSPvk210 and PvCSPvk247 ( . All analyzed peptides were highly disordered in solution and adopted a large ensemble of conformations (Figure 2A), similar to peptides derived from PfCSP and PbCSP (Kucharska et al., 2020). Secondary structure was present mainly in the form of transient hydrogen-bonded β-turns ( Figure 2B). The propensity of individual residues to form turns varied from 0 to ~60%, with PvCSPvk210 peptides displaying slightly lower averaged turn propensity (~15-20%) than PvCSPvk247 peptides (~23-26%) ( Figure 2C). PvCSPvk210 peptides containing (GDRAAGQPA) 2 motifs (210-7 and -9), as well as the last repeat of the 247-3 peptide (GNGAGGQAA), were the only motifs with a significant propensity to form helices (>20%).
In order to estimate the elastic modulus of these peptides with intrinsically low secondary structure propensities, peptides were modeled as Hookean springs ( Figure 2D), in which the force needed to extend or compress a spring by some distance is proportional to that distance, and the underlying energy function is quadratic (Equation 1). The PMF or free energy governing changes in the peptide's end-to-end distance was computed and fit to a quadratic function. The excellent data fits (R 2 = 0.91-0.97) suggest that in aqueous solution, peptides derived from the PvCSP repeats behave like harmonic springs. On average, the elastic modulus was ~3-5 cal/(mol Å 2 ), with the highest values (stiffest peptide) observed for peptide 247-2 (4.9 ± 0.2 cal/(mol Å 2 )) and lowest (most flexible peptide) observed for peptide 247-1 (3.2 ± 0.2 cal/(mol Å 2 )).
Next, we co-crystalized the 2F2 Fab with five different peptides derived from PvCSPvk210 (210-1, 210-2, 210-3, 210-4, and 210-5, Supplementary file 1, Figure 3, Figure 3-figure supplement 2) to gain molecular insights into the binding mode and cross-reactivity of mAb 2F2 binding to PvCSPvk210. The crystal structures were solved to resolutions ranging from 1.97 Å to 2.67 Å ( Table 1). 2F2 recognizes the core epitope ( 2 DRA(D/A)GQPAGD 11 ) of all PvCSPvk210 peptides in an almost identical coil conformation peptide backbone root-mean-square deviation (RMSD 0.10-0.28 Å) with two consecutive β-turns observed for residues 7 QPAGD 11 (Supplementary file 2) forming one turn of a 3 10 -helix, thus consistent with the moderate secondary structure propensities observed for the unliganded PvCSPvk210 repeats ( Figure 3B). The co-crystal structures also provide molecular insights into the cross-reactivity of 2F2 to different types of PvCSPvk210 repeat motifs ( 1 GDRA(D/A)GQPA 9 ). The sidechains of D/A 5 point up and away from the 2F2 paratope and do not significantly contribute to the 2F2 Fab-peptide interaction, helping to explain the similar binding affinities to the different peptides containing this variation ( Figure 3E). 2F2 also binds the 210-5 peptide containing a unique repeat C-terminal of the central region ( 1 GDRAAGQPAGNGAG-GQAA 18 ); however, the electron density of residues 12-18 C-terminal of the bound core peptide motif is weak in the co-crystal structure, thus providing limited structural insight into this peptide region and suggesting it does not make strong interactions with 2F2 ( Figure 3-figure supplement  2).
The recognition of PvCSPvk210 peptides by 2F2 is mediated mostly by residues localized in heavy chain complementarity-determining regions (HCDRs) 1, 2, and 3, and kappa chain complementaritydetermining regions (KCDRs) 1 and 3. The PvCSPvk210 one turn of a 3 10 -helix is positioned in the hydrophobic pocket formed by KCDR1 residues of the antibody ( Figure 3C and D, Figure 3-figure supplement 3C, E, and G), and is stabilized by three H-bonds formed between A 9 and the backbone of K.Gly91 and K.Phe96, and G 10 and the sidechain of H.Ser58 ( Figure 3E). The antibody-antigen complex buries 848 Å 2 on the Fab (437 Å 2 on HC and 411 Å 2 on KC) and 1023 Å 2 on the 210-4 peptide. Peptide residues 1 GRDADG 6 are positioned between HCDR1 and 3 and do not interact with the light chain ( Figure 3C and D).
2F2 binds to PvCSPvk210 peptides using both germline and somatically hypermutated residues. Nine germline-encoded aromatic residues form significant van der Waals interactions with the peptides, contributing a total of ~385 Å 2 of buried surface area (BSA; Figure 3E, Figure 3figure supplement 3C, E, and G). To accommodate arginine residues present in the sequence of the PvCSPvk210 peptides, the 2F2 paratope has an overall electronegative potential (Figure 3-figure supplement 4A). R 3 and Q 7 play a central role in mediating the Fab-peptide interactions, forming six H-bonds with both heavy and light chain residues of the antibody ( Figure 3E), and contributing ~107 Å 2 and ~ 156 Å 2 of BSA, respectively. residue i and N-H of residue i + 3 is shown as a gray line. (C) Secondary structure propensity at each residue, averaged over 20 replicas and computed using the Dictionary of Secondary Structure for Proteins (DSSP) criteria (Nagy and Oostenbrink, 2014). (D) Elastic modulus of peptides computed from MD simulations. The reversible work or free energy (ΔG) for extension and compression of the peptide is plotted as a function of equilibrium end-to-end distances (d EE ) (solid line). The data is fitted to the quadratic function of elastic potential energy (black dashed line). For each peptide, the estimated values of k (elastic modulus), d 0 (equilibrium d EE ), and R 2 (regression coefficient to indicate quality of fit) are shown. Shading represents standard error of mean.
The online version of this article includes the following figure supplement(s) for figure 2:     In the 2E10.E9 Fab-247-2 co-crystal structure, two Fabs bind to one peptide, which is in agreement with the 2:1 stoichiometry established by ITC ( Figure 4A). The core epitope of PvCSPvk247 peptides contains eight residues ( 3 GAGNQPGA 10 ) and adopts a similar coil conformation in all analyzed peptides when bound to 2E10.E9 (peptide backbone RMSD 0.48-1.0 Å), with only isolated turns ( Figure 4B, Supplementary file 2). Although 2E10.E9 Fab binds peptides 247-2 and 247-3 containing the first (EDGAGNQPG) and the last repeat (ANGAGGQAA) motifs, the electron density of residues unique to these repeats was absent in the co-crystal structures obtained. This suggests that the antibody does not interact extensively with these variable residues that are outside the well-resolved conserved core ( Figure  2E10.E9 interacts with the PvCSPvk247 peptides using HCDRs 1, 2, and 3, and KCDR1 and 3, with the N-terminal part of the peptide positioned between HCDR1 and 3. Six germline-encoded aromatic residues, including K.Tyr32, K.Tyr92, K.Tyr94, K.Phe96, H.Tyr32, and H.Trp50, play a central role in peptide recognition, forming three H-bonds with G 9 , Ala 10 , and A 13 , and contributing 176 Å 2 BSA to the relatively small paratope of this Fab (444 Å 2 total BSA on the 2E10.E9 paratope; 588 Å 2 total BSA on peptide 247-3) (Figure 3-figure supplement 3B, D, and F). Interestingly, HCDR3 H.Cys98 and H.Cys100 make a disulfide bond that positions H.Gly99 in an ideal position to form an H-bond with residue G 9 of the peptide and mimic the stacking effect provided by aromatic sidechains ( Figure 4E). Residue N 6 of the PvCSPvk247 peptides is central to the interaction, forming four H-bonds with HCDR2 residues H.Thr30, H.Asn52, and H.Ser52A ( Figure 4E).

Homotypic 2E10.E9 Fab-Fab interactions upon PvCSPvk247 repeat binding
In the 2E10.E9 Fab-247-2 peptide co-crystal structure where two Fabs bind to one peptide, we observed multiple contacts between the two 2E10.E9 Fabs ( Figure 5). Indeed, the two 2E10.E9 Fabs interact in a head-to-head binding mode at an ~146 o angle ( Figure 5A). Contacts between the two indicates residues resolved in the corresponding X-ray crystal structures. Bottom panel: comparison of the conformations of PvCSP210 peptides in X-ray crystal structures. PvCSPvk210 peptides are colored from navy to light blue, with the residues adopting one turn of a 3 10 -helix depicted in pink. (C) Top and side views of the 210-4 peptide (light blue) in the binding groove of the 2F2 Fab shown as surface representation (heavy chain [HC] shown in green and kappa chain [KC] shown in white). (D) Comparison of the conformations adopted by the core epitope of peptides 210-1, 210-2, 210-3, 210-4, and 210-5 when bound to 2F2. (E) Detailed interactions between Fab 2F2 and peptide 210-4. H-bonds and salt bridges are shown as black dashes, peptide 210-4 is shown in light blue, HC is shown in green, and KC is shown in gray. Fab residues are annotated with H or K letters to indicate heavy and kappa light chain, respectively.
The online version of this article includes the following figure supplement(s) for figure 3:      Table 1. X-ray crystallography data collection and refinement statistics.   Fabs are mostly symmetric and involve mainly HCDR2 of both Fab A and B, as well as the KCDR3 of Fab A ( Figure 5A and B  Figure 5B). Comparison of the 2E10.E9 variable gene sequences to the inferred germline precursor (IGHV9-3 and IGKV8-19) reveals that only one of the residues involved in Fab-Fab contacts has been somatically hypermutated (H.Ser52A), making this homotypic Fab-Fab interaction primarily germline-encoded in the context of binding its repeating epitope ( Figure 5C).

Central repeat flexibility upon binding of inhibitory antibodies to fulllength PvCSP
Next, we characterized the binding of mAbs 2F2 and 2E10.E9 to full-length recombinant PvCSPvk210 and PvCSPvk247. Both mAbs exhibit fast associations to their respective PvCSP sequences, as revealed in biolayer interferometry (BLI) experiments ( Figure 6A and C). However, mAb 2E10.E9 displays a relatively fast dissociation compared to mAb 2F2, which contributes to the lower overall binding affinity of this mAb to PvCSPvk247 compared to a higher binding affinity for the mAb 2F2-PvCSPvk210 interaction. Next, ITC measurements indicated that, as expected, both 2F2 and 2E10.E9 Fabs bind PvCSP with high stoichiometry indicative of multiple Fab copies interacting with a single PvCSP molecule. 2F2 Fab recognizes PvCSPvk210 with ~10 times higher affinity (0.242 µM) compared to the 2E10. E9 Fab-PvCSPvk247 interaction (2.21 µM), corroborating the binding kinetics data ( Figure 6B and D). Size-exclusion chromatography coupled with multiangle light scattering (SEC-MALS) characterization of the Fab-PvCSP complexes revealed high binding stoichiometry with a molecular weight of ~522 kDa and ~463 kDa for the 2F2 Fab-PvCSPvk210 and 2E10.E9 Fab-PvCSPvk247 complexes, respectively ( Figure 6E). These sizes correspond to approximately 10 2F2 Fab's bound to one molecule of PvCSPvk210 and to approximately 9 2E10.E9 Fab's bound to one molecule of PvCSPvk247. Although both ITC and SEC-MALS confirm the assembly of large complexes formed by multiple Fab's binding to one PvCSP molecule, the exact Fab:PvCSP stoichiometry that ensues from these independent analyses is slightly different between the two techniques, which we attribute to the difficulty in obtaining precise concentration measurements for recombinant PvCSP and to the two experiments being performed at distinct concentrations.
To investigate a possible structural ordering of the PvCSP central repeat as might be induced by the binding of multiple 2F2 and 2E10.E9 Fab's, we performed negative stain electron microscopy (NS EM) and electron cryomicroscopy (cryo-EM) analyses of the SEC-purified Fab-PvCSP complexes ( Figure 6-figure supplement 1). 2D class average images from the negative stain micrographs of the 2E10.E9 Fab-PvCSPvk247 complex revealed multiple 2E10.E9 Fabs spaced tightly against each other ( Figure 6-figure supplement 1A, left panels). However, cryo-EM analysis of the same 2E10.E9 Fab-PvCSPvk247 complex indicated that Fab 2E10.E9 does not form regular, spiral assemblies with CSP, presumably because this type of complex would not accommodate the symmetric, head-to-head interactions between 2E10.E9 Fabs that were observed in the 2E10.E9 Fab-247-2 peptide crystal structure ( Figure 5, Figure 6-figure supplement 1A, right panels).
Interestingly, 2D classes from the negative stain micrographs of the 2F2 Fab-PvCSPvk210 complex indicated multiple conformational states, and thereby suggest that the PvCSPvk210 central repeat in the corresponding X-ray crystal structures. Bottom panel: comparison of the conformations of PvCSP247 peptides in X-ray crystal structures, with peptides 247-2, 247-3, and 247-4 depicted in yellow, orange, and teal, respectively. (C) Top and side views of the 247-3 peptide (orange) in the binding groove of the 2E10.E9 Fab shown as surface representation (heavy chain [HC] shown in blue and kappa chain [KC] shown in white). (D) Comparison of the conformations adopted by the core epitope of peptides 247-2, 247-3, and 247-4 when bound to 2E10.E9. (E) Detailed interactions between Fab 2E10.E9 and peptide 247-3. H-bonds are shown as black dashes, peptide 247-3 is shown in orange, HC is shown in green, and KC is shown in gray. The Fab residues are annotated with H or K letters to indicate heavy and kappa light chain, respectively.
The online version of this article includes the following figure supplement(s) for figure 4:    (Figure 6-figure supplement 1B, left  panels). Cryo-EM analysis of the same 2F2 Fab-PvCSPvk247 complex showed similar conformational heterogeneity ( Figure 6-figure supplement 1B, right panels). These results suggest that the 2F2 Fabs may not be stabilized by appreciable inter-Fab contacts as observed for 2E10.E9 Fabs. To gain a better understating of the conformational flexibility observed for the 2F2 Fab-PvCSPvk210 complex, we collected NS EM data of the 2F2 Fab bound to a PvCSPvk210-derived peptide of sufficient length to accommodate binding of two Fabs (peptide 210-10, Supplementary file 1). EM class averages from the micrographs of the 2F2 Fab-210-10 peptide complex displayed high variability in the angles between the two Fabs, ranging from ~30° to ~170° ( Figure 7A). These data indicate that the complex is highly flexible, and that the two Fabs are likely not forming extensive stabilizing inter-Fab homotypic contacts upon binding the repeat peptide. This mode of binding is in contrast to other Fab-peptide complexes known to form extensive homotypic interactions (2E10.E9 Fab-247-2 [ Figure 7B], 3D11 Fab-NPNDx2 [Kucharska et al., 2020;Figure 7C], and 1210 Fab-NANP 5 [Imkeller et al., 2018; Figure 7D, Supplementary file 1]). Complexes that form homotypic contacts showed lower flexibility than the 2F2 Fab-210-10 peptide complex, with either one or two distinct 3D classes present.

Discussion
Previous knowledge of inhibitory mAbs against the PfCSP repeats suggest that an in-depth molecular understanding of PvCSP could facilitate the design of next-generation Pv biomedical interventions. Indeed, molecular characterization of hundreds of human mAbs induced either by natural infection (Triller et al., 2017), Pf sporozoite immunization (Imkeller et al., 2018;Kisalu et al., 2018;Murugan et al., 2020;Tan et al., 2018;Wang et al., 2020), or RTS,S/AS01 vaccination (Oyen et al., 2017;Pholcharee et al., 2021) revealed important insights into vaccine design and antibodies as prophylactics. These include preferential mAb binding to the conserved core epitope (N/D)PNANPN(V/A) (Murugan et al., 2020) and potential correlations of protection with binding affinity and recognition of epitopes with secondary structural motifs of type I β-and pseudo 3 10 -turns (Pholcharee et al., 2021). Moreover, a subset of potent mAbs elicited by whole sporozoite vaccination was shown to not only bind to NANP repeats, but also to a junctional epitope positioned between the N terminus and the central repeat domain of PfCSP, which is absent in the RTS,S vaccine (Kisalu et al., 2018;Tan et al., 2018;Wang et al., 2020). Our understanding of CSP targeting by neutralizing mAbs is however almost exclusively limited to PfCSP. Despite the prominence of Pv malaria morbidity worldwide, a molecular understanding of PvCSP and how its central repeat is recognized by inhibitory antibodies has remained scarce.
Here, we performed CD spectroscopy and MD simulations on PvCSP repeat peptides to better understand the unliganded structure of this inhibitory antibody target. Our data show that all analyzed peptides are disordered in solution and any secondary structure observed is local and transient. This result is similar to what has been previously described for the Pf and Pb CSP repeat region (Dyson et al., 1990;Kucharska et al., 2020). Interestingly, the sequence composition of the CSP repeats from these three Plasmodium species is substantially different; whereas PfCSP and PbCSP are asparagine and proline-rich, PvCSP is predominantly alanine and glycine-rich. In addition, the repeating modulus for PfCSP and PbCSP contains four amino acids, whereas the repeating modulus of PvCSP contains nine amino acids. These comparisons underscore a diversity of amino acid features associated with low structural propensity in repeating sequences.
Interestingly, calculations of elastic modulus demonstrated that PvCSP peptides behave like harmonic springs, meaning that the stretching or compressing of the peptides is proportional to the force applied to them. Elastic properties are well characterized in a vast array of proteins of different functions, including elastin, spider silk, or mussel byssus (Gosline et al., 2002). The high degree of conformational disorder and the resulting elastic properties of PvCSP peptides likely stem from the combination of the low complexity periodic nature of the sequence as well as the particular amino acid composition of this region. A high combined proline and glycine content was shown to distinguish the sequence of self-assembled elastomeric proteins such as elastin, resilin, and elastic spider silks, amongst many others, from that of proteins prone to amyloid formation (Rauscher et al., 2006). Both proline and glycine are 'amyloid breakers' because of their low propensity to form regular secondary structure such as the extended β-sheets found at the core of amyloid fibrils (Parrini et al., 2005;Williams et al., 2004). As such, a high proline and glycine content enables self-assembled elastomers to avoid the amyloid fate while remaining disordered even in the assembled or phase-separated state. In turn, high structural disorder enables these proteins to function as entropic springs, whereby the large conformational entropy of the polypeptide chain provides at least part of the driving force for elastic recoil. PvCSP peptides have glycine and proline contents of ~22-33% and ~7-11%, respectively, and thus are predicted to fall within the transition region separating amyloidogenic from elastomeric proteins, the latter of which include elastin domains and insect resilin (Rauscher et al., 2006). Moreover, the elastic modulus of PvCSP peptides (~3-5 cal/(mol Å 2 )) is commensurate with the elastic modulus of peptides modeled after human elastin (~9 cal/(mol Å 2 )), which requires extensibility and elasticity for its physiological function (Reichheld et al., 2021). This analogy supports the observation that CSP is inherently elastic and suggests that achieving a high degree of conformational disorder may be an essential aspect of CSP function on the surface of sporozoites. Interestingly, recent in vitro experiments demonstrated that PbCSP repeats also have elastic properties, which are lost when the repeat sequence is scrambled (Balaban et al., 2021). The link between the elasticity of CSP repeats and CSP function is still not completely understood; yet, it appears that biophysical properties of the CSP repeats are necessary for maintaining sporozoite motility (Balaban et al., 2021;Coppi et al., 2011), even though CSP itself is not a motor protein.
The presence of disorder in all analyzed PvCSP repeat peptide structures was also appreciated in our X-ray crystallography data, where inhibitory mAbs 2F2 and 2E10.E9 were found to recognize their epitopes in an induced coil conformation, with only one turn of a 3 10 -helix and isolated turns as secondary structure motifs observed in both instances. Binding of the CSP repeat by inhibitory antibodies has been shown to induce a range of conformations in this intrinsically disordered region for Pf and Pb, with the occasional presence of type I β-turns and pseudo 3 10 -turns often but not always linked with antibody-mediated protection (Imkeller et al., 2018;Kisalu et al., 2018;Kucharska et al., 2020;Murugan et al., 2020;Oyen et al., 2017;Pholcharee et al., 2020;Pholcharee et al., 2021;Tan et al., 2018;Triller et al., 2017).
PfCSP-and PbCSP-reactive antibodies have been shown to cross-react to a varying extent with the predominant repeat motif and neighboring sequences of subtle variance within the repeat region, for example, the PfCSP junction (Julien and Wardemann, 2019;Kisalu et al., 2018;Murugan et al., 2020;Tan et al., 2018). Cross-reactivity is a feature of human antibodies encoded by various Ig-gene combinations and was also observed in murine antibodies against the PfCSP repeat (mAb 2A10; Zavala et al., 1983) and against the PbCSP repeat (3D11; Kucharska et al., 2020;Yoshida et al., 1980). Our data indicate that both mAbs 2F2 and 2E10.E9 are cross-reactive as they bind to different repeat motifs of their respective PvCSP variants with similar affinity and in identical conformations. mAbs 2F2 and 2E10.E9 are, however, not cross-reactive to different strains of Pv (Nardin et al., 1982) due to the significantly different CSP sequences of strains PvCSPvk210 and PvCSPvk247. Although both mAbs 2F2 and 2E10.E9 are inhibitory, only mAb 2F2 was reported to induce aberrations in size and in DNA content of sporozoites, indicating a distinct mechanism of attenuation (Roth et al., 2018), possibly due to the higher affinity to PvCSP and slower dissociation rate of mAb 2F2 compared to mAb 2E10.E9.
Both mAbs 2F2 and 2E10.E9 recognize their respective epitopes using germline-encoded and somatically hypermutated residues. The difference in affinities between mAbs 2E10.E9 and 2F2 and their respective antigen might partially stem from the low lumber of somatic hypermutations of mAb 2E10.E9 (1 in HC, 6 in KC) compared to 2F2 (14 in HC, 4 in KC). Specifically, mAbs 2F2 and 2E10.E9 use nine and six germline-encoded aromatic residues, respectively, to mediate contacts with their core epitopes. Germline-encoded mAb 2E10.E9 residue H.Trp50 forms extensive van der Waals interactions with N 6 and P 8 residues, analogous to IGHV3-33 germline-encoded H.Trp52 in several human mAbs, including mAbs MGG4 (Tan et al., 2018), 1210 (Imkeller et al., 2018), 311 (Oyen et al., 2017), and other antibodies induced by sporozoite immunization (Murugan et al., 2020) or RTS,S vaccination (Pholcharee et al., 2021;Figure 3-figure supplement 3D and F). As both human and murine mAbs appear to depend on germline-encoded aromatic residues for CSP recognition, it is likely that CSP repeats prime the mammalian immune system to select antibodies from germline genes with already optimally positioned aromatic residues.
MAb 2E10.E9 displays homotypic Fab-Fab interactions, as was previously described for mAbs 1210 (Imkeller et al., 2018), 311 , and 239 and 399 (Pholcharee et al., 2021) against PfCSP and murine mAb 3D11 against PbCSP (Kucharska et al., 2020). 2E10.E9 Fabs interact in a head-to-head binding mode, forming symmetric interactions similar to those observed in the case of antibodies 1210 (Imkeller et al., 2018) and 399 (Pholcharee et al., 2021). Interestingly, mAb 2E10. E9 does not appear to bind longer PvCSPvk247 peptides with higher affinity compared to shorter PvCSPvk247 peptides ( Figure 3A), in contrast to what has been described for other mAbs that recognize the PfCSP and PbCSP repeats and form homotypic interactions (Imkeller et al., 2018;Kucharska et al., 2020;Pholcharee et al., 2021). Unlike mAbs 1210 (Imkeller et al., 2018) and 3D11 (Kucharska et al., 2020), the residues forming Fab-Fab contacts in mAb 2E10.E9 are almost exclusively germlineencoded, similar to mAb 399 (Pholcharee et al., 2021). Combined, these results suggest that forming Fab-Fab interactions is ubiquitous among anti-CSP mAbs targeting different Plasmodium species. Analysis of human mAbs isolated after immunization with whole Pf sporozoites indicated that NANP homotypic antibody interactions promote activation and strong clonal expansion of PfCSP-reactive B-cells (Imkeller et al., 2018). It appears that homotypic interactions may emerge from B-cell receptor clustering on the surface of B cells (Imkeller et al., 2018); however, a direct link between homotypic interactions in the context of soluble antibodies and sporozoite inhibition is still to be established. mAbs 2F2 and 2E10.E9 recognize core PvCSP epitopes of 10 and 8 residues, respectively. Interestingly, potent inhibitory mAbs against PfCSP and PbCSP typically recognize epitopes of similar length (8-10 residues; Imkeller et al., 2018;Kucharska et al., 2020;Murugan et al., 2020;Pholcharee et al., 2021), despite shorter repeating motifs for PfCSP and PbCSP (4-aa) compared to PvCSP (9-aa). Given the similar antibody epitope lengths, in addition to similar structural propensities for underlying antigenic motifs and induction of antibody homotypic interactions when recognizing closely spaced repeating epitopes, we suggest that similar CSP-based vaccine design approaches could be applied across different Plasmodium species. Nevertheless, the lack of Plasmodium species cross-reactivity for the potent inhibitory mAbs described thus far suggests that different pre-erythrocytic vaccines or a multicomponent CSP-based vaccine will likely be required for broad coverage against the different Plasmodium species causative of human malaria morbidity and mortality worldwide.  Best and Hummer, 2009;Best and Mittal, 2010;Lindorff-Larsen et al., 2012;MacKerell et al., 1998;Piana et al., 2011 https://www.charmm. org/charmm/?CFID= 66837e22-4ee5-47ba-bcbf-b4b385c2397 e&CFTOKEN=0; RRID:SCR_014892
MD simulations GROMACS 2016.5 (Abraham et al., 2015Berendsen et al., 1995) was used to perform all-atom molecular simulations of the following seven peptides: 210-6, 210-7, 210-8, 210-9, 247-1, 247-2, and 247-3 (Supplementary file 1). The CHARMM22* (Best and Hummer, 2009;Best and Mittal, 2010;Lindorff-Larsen et al., 2012;MacKerell et al., 1998;Piana et al., 2011) force field and the CHARMM-modified TIP3P (TIPS3P) explicit water model (Jorgensen et al., 1983) were used for all simulations. PyMOL (Schrödinger, 2015) was used to design the peptides with acetylated N-terminus and amidated C-terminus. The peptides were first collapsed from their arbitrary extended state without any solvent (in vacuo) under NVT conditions. The last conformations from simulations in vacuo were used to initiate equilibrium simulations in water, with 20 replicas for each peptide. Peptides were solvated with water and 0.15 M NaCl in a rhombic dodecahedral box, with a side length of 4.0 nm. Periodic boundary conditions were applied, and energy minimization was carried out using the steepest descent algorithm. Lennard-Jones and short-range electrostatic interactions were computed with a cutoff of 9.5 Å. Long-range electrostatic interactions were computed with Particle-Mesh Ewald summation (Darden et al., 1993;Essmann et al., 1995), using a fourth-order interpolation and a grid spacing of 1.2 Å. All bonds were constrained using the LINCS algorithm (Hess, 2008). The system was brought to the specified temperature of 300 K and pressure of 1 atm under NPT conditions. 10 ns NPT simulations were performed with velocity-rescaling temperature coupling (Bussi et al., 2007) and Berendsen pressure coupling (Berendsen et al., 1984). Finally, simulations were carried out under NPT conditions with Parrinello-Rahman pressure coupling (Parrinello and Rahman, 1981) for 300 ns. The integration step was 2 fs, and atomic coordinates were recorded every 100 ps.

MD simulation analyses
Visual molecular dynamics (VMD) (Humphrey et al., 1996) was used to create snapshots of peptides, while all plots were created with Matplotlib (Hunter, 2007). The first 225 ns were excluded for computation of secondary structure propensities per residue, as well as for H-bonding contact maps. Secondary structure was assigned using the Python package MDTraj (McGibbon et al., 2015), which uses the DSSP algorithm (Nagy and Oostenbrink, 2014). An in-house script was used to compute H-bonding contact maps. Forward peptide-peptide H-bonds form between C=O of residue i and N-H of residue i + n. A forward H-bond was identified if the donor-acceptor distance (r ON ) and the hydrogen-donor-acceptor angle (θ) were less than 3.5 Å and 37° for n = 2 (γ-turn), 4.9 Å and 66° for n = 3 (β-turns), 4.5 Å and 60° for n = 4 (α-turn) and 3.5 Å and 40° for n ≥ 5. A reverse turn is formed between N-H of residue i and C=O of residue i + n, and identified if r ON < 3.5 Å and θ < 60° (except for n = 3, for which θ < 40°). From a histogram of equilibrium end-to-end distances (d = distance between Cα of the first and last residues), the probabilities at each given d (P(d)) were used to compute a free energy profile (ΔG(d)) of the peptide's end-to-end distances. The resulting potential of mean force (PMF) was then fitted to the quadratic function of elastic potential energy: where k B is the Boltzmann constant, T is the absolute temperature, k represents the stiffness or elastic modulus of the peptide, and d 0 is the equilibrium end-to-end distance.

2F2 Fab and 2E10.E9 Fab expression and purification
Variable light and heavy chains of mAb 2F2 and 2E10.E9 antibody genes were sequenced from the hybridomas (BEI Resources MRA-184 and MRA-185, respectively; Applied Biological Materials Inc). Sequenced regions were gene synthesized and cloned (GeneArt) into custom pcDNA3.4 expression vectors immediately upstream of human Igκ and Igγ1-C H 1 domains. pcDNA3.4-Fab KC and Fab HC plasmids were co-transfected into HEK 293F cells for transient expression using FectoPRO DNA transfection reagent (Polyplus). Cells were cultured in Gibco FreeStyle 293 Expression Medium for 6-7 days and subsequently purified via a combination of KappaSelect affinity chromatography (GE Healthcare), cation exchange chromatography (MonoS, GE Healthcare), and size-exclusion chromatography (Superdex 200 Increase 10/300 GL, GE Healthcare).

Cell lines
HEK 293F cells (Thermo Fisher Scientific 12338026) and mAb 2F2 and 2E10.E9 hybridoma cell lines (BEI Resources MRA-184 and -185, respectively) were authenticated and validated to be mycoplasmafree by their respective commercial entities.

Biolayer interferometry
BLI (Octet RED96, ForteBio) experiments were conducted to determine the binding kinetics of 2F2 and 2E10.E9 IgG and Fab to recombinant PvCSPvk210 and PvCSPvk247. PvCSPvk210 or PvCSPvk247 was diluted to 10 µg/mL in kinetics buffer (PBS, pH 7.4, 0.01% [w/v] BSA, 0.002% [v/v] Tween-20) and immobilized onto Ni-NTA biosensors (ForteBio). Subsequently, biosensors were dipped into wells containing dilutions of either 2F2 or 2E10.E9 IgG or Fab in kinetics buffer. For measurement of the dissociation rate, tips were immersed back into kinetics buffer after association. All data were analyzed using ForteBio's Octet Data Analysis software 9.0.0.6, and curves were fitted to a 2:1 binding model given the presence of multiple epitopes of slightly different sequence composition for these mAbs within a single PvCSP molecule.

Isothermal titration calorimetry
ITC experiments were performed with an Auto-iTC200 instrument (Malvern) at 25°C. Titrations were performed with 2F2 or 2E10.E9 Fab in the syringe in 15 successive injections of 2.5 µl. Full-length recombinant PvCSPvk210 and PvCSPvk247, and PvCSP-derived peptides (Supplementary file 1) were added to the calorimetric cell. All proteins and peptides were diluted in Tris-buffered saline (TBS; 20 mM Tris pH 8.0, and 150 mM NaCl). Full-length recombinant PvCSP was diluted to 5 µM and titrated with Fab at 240-400 µM. All PvCSP-derived peptides were diluted to 4-8 µM and titrated with 2F2 Fab at 100-125 µM or 2E10.E9 Fab at 180-220 µM. Experiments were performed at least in duplicates, and the mean and standard error of the mean are reported. ITC data were analyzed using the Micro-Cal ITC Origin 7.0 Analysis Software according to a 1:1 binding model.
Data were collected at the 23-ID-D or 23-ID-B beamlines at the Argonne National Laboratory Advanced Photon Source. All datasets were processed and scaled using XDS (Kabsch, 2010). The structures were determined by molecular replacement using Phaser (McCoy et al., 2007). Refinement of the structures was performed using phenix. refine (Adams et al., 2010) and iterations of refinement using Coot (Emsley et al., 2010). Access to all software was supported through SBGrid (Morin et al., 2013). Prediction of secondary structure of the co-crystallized peptides was performed with Stride (Heinig and Frishman, 2004). Fab-peptide and Fab-Fab contacts were analyzed using the PDBePisa server (Krissinel and Henrick, 2007). The detection of intramolecular H-bonds in peptides was performed with PyMOL (Schrödinger, 2015).

Cryo-EM data collection and image processing
2F2 Fab-PvCSPvk210 and 2E10.E9 Fab-PvCSPvk247 complexes were purified via Superose 6 Increase 10/300 GL chromatography (GE Healthcare) and concentrated to 0.5 mg/mL. 3 µL of the sample was deposited on homemade holey gold grids (Marr et al., 2014), which were glow-discharged in air for 15 s before use. Samples were blotted for 12.0 s and subsequently plunge-frozen in a mixture of liquid ethane and propane (Tivol et al., 2008) using a modified FEI Vitrobot (maintained at 4°C and 100% humidity). Data collection was performed with a FEI Tecnai F20 microscope operated at 200 kV with a K2 camera (Gatan Inc). A calibrated 34,483× magnification, resulting in a pixel size of 1.45 Å, and defocus range between 1.5 and 2.8 µm were used for data collection. Exposures were fractionated as movies of 30 frames with a total exposure of 35 electrons/Å 2 . A total of 269 movies were obtained for the 2E10.E9 Fab-PvCSPvk247 complex and 169 movies for the 2F2 Fab-PvCSPvk210 complex. Image processing was carried out in cryoSPARC v2 (Punjani et al., 2017). Initial specimen motion correction, exposure weighting, and CTF parameters estimation were done using patch-based algorithms. 100,133 and 87,287 particle images were extracted from micrographs of 2E10.E9 Fab-PvCSPvk247 and 2F2 Fab-PvCSPvk210 complex, respectively, and subjected to 4-5 rounds of 2D classification.