Structural delineation of potent transmission-blocking epitope I on malaria antigen Pfs48/45

Interventions that can block the transmission of malaria-causing Plasmodium falciparum (Pf) between the human host and Anopheles vector have the potential to reduce the incidence of malaria. Pfs48/45 is a gametocyte surface protein critical for parasite development and transmission, and its targeting by monoclonal antibody (mAb) 85RF45.1 leads to the potent reduction of parasite transmission. Here, we reveal how the Pfs48/45 6C domain adopts a (SAG1)-related-sequence (SRS) fold. We structurally delineate potent epitope I and show how mAb 85RF45.1 recognizes an electronegative surface with nanomolar affinity. Analysis of Pfs48/45 sequences reveals that polymorphisms are rare for residues involved at the binding interface. Humanization of rat-derived mAb 85RF45.1 conserved the mode of recognition and activity of the parental antibody, while also improving its thermostability. Our work has implications for the development of transmission-blocking interventions, both through improving vaccine designs and the testing of passive delivery of mAbs in humans.

M alaria is a global health priority, with an estimated 216 million cases worldwide in 2016 alone 1 . The Plasmodium falciparum (Pf) parasite is responsible for most malaria-related mortalities, with over two-thirds occurring in children under 5 years of age. Disrupting the Pf life cycle as the parasite circulates between humans and Anopheles mosquitoes has the potential to reduce infections within communities, and thus reduce illness and death 2 . It was recognized decades ago that antibodies have the potential to inhibit onward transmission of the parasite when passed by malaria-infected humans to the Anopheles mosquito vector [3][4][5] . Transmission-blocking vaccines (TBVs) are based on this principle, and aim to elicit antibodies in humans that can reduce Pf transmission to the vector when mosquitoes ingest these antibodies during feeding 6 . Target proteins for TBV development are located on the surface of gametocytes/gametes (P48/45 7 , P230 8 ) and zygotes/ookinetes (P25, P28) 9,10 , or are expressed within the mosquito midgut (e.g. APN1 11 and FREP1 12 ).
Pfs48/45 is a cysteine-rich surface protein that plays a critical role in male gamete fertility 13 . Antibodies against Pfs48/45 have been shown to prevent parasite development and transmission 14,15 . Because Pfs48/45 is located on gametocytes/ gametes that circulate in humans, naturally occurring antibodies against Pfs48/45 can be elicited in individuals living in malariaendemic areas, and these antibodies have been demonstrated to possess transmission-blocking activity [16][17][18][19][20][21] .
Pfs48/45 is organized into three domains, and epitope mapping has identified at least five predominant sites of antibody recognition that span its primary sequence 22,23 (Fig. 1a). Human, murine, and rat monoclonal antibodies (mAbs) have been derived that target epitopes I, IIb, III, and V 5, [24][25][26] . Of these, mAb 85RF45.1, derived from rats immunized with whole gametocytes followed by fusion and screening by enzyme-linked immunosorbent assay for Pfs48/45 antibodies, was shown to be the most potent at inhibiting parasite transmission to Anopheles mosquitoes 24 . mAb 85RF45.1 is a rat IgG1 that binds to epitope I on Pfs48/45 24 .
A three-dimensional structure of Pfs48/45 remains elusive. Such information would provide useful in devising strategies for improved recombinant expression of immunogens; indeed, monodisperse Pfs48/45 has been challenging to obtain and development of homogeneous and stable recombinant constructs is an area of intense research 7,25,[27][28][29][30] . A molecular understanding of mAb 85RF45.1 binding to Pfs48/45 would also provide the blueprints for the development of next-generation Pfs48/45 immunogens that optimally present this potent epitope to be tested as TBVs. In addition, molecular details of this antibody-antigen interaction could facilitate the development of new antibody interventions. Here, we present the threedimensional structure of the Pfs48/45 6C domain and delineate the potent 85RF45.1 epitope structurally. We also provide molecular details for the development of mAb TB31F, a humanized version of rat mAb 85RF45.1 with improved biophysical properties, now undergoing preclinical development.  (Fig. 1c). The buried surface area of mAb 85RF45.1 on its antigen is extensive (1039 Å 2 ), and primarily contributed by the heavy chain (650 Å 2 ), and less by the light chain (389 Å 2 ) (Supplementary Table 2). Ten discontinuous residues on Pfs48/45 are  contacted by the antibody to form salt bridges or H-bonds,  namely G346, D347, D351, Q355, E365, K394, K413, K414, D415, and K416 (Fig. 1c, insets). Our structural delineation of the antibody-antigen complex corroborates the importance of disulfide bonds and protein conformation for epitope I established by previous biochemical characterization 24 .
N-linked glycosylation improves expression of Pfs48/45 6C. The Pfs48/45 6C construct used for structure determination was expressed in HEK293 mammalian cells, thus in a system capable of post-translational modifications. The primary amino acid sequence of Pfs48/45 6C encodes for two putative N-linked glycosylation sites when expressed in mammalian cells: N299 (NFS) and N303 (NVS), both located near the N terminus of the construct. Our crystal structure revealed that N303 is indeed glycosylated, and harbors a NAG residue residual from Endoglycosidase H (Endo H) treatment (Fig. 3a). In fact, NAG303 forms H-bonds with D390 and D391 of the neighboring loop and buries approximately 150 Å 2 of surface area on the two loops that connect strands a-a′ and strands e-f (Fig. 3a). On the other hand, N299, which immediately follows C298, is not glycosylated, and instead its sidechain plays an important structural role in the strand a-loop-strand a′ motif, mediating H-bonds with S301 (Fig. 3a). To determine whether N-linked glycosylation played a role in the expression of Pfs48/ 45 6C in our HEK293 system, we created single and double mutants removing the N-linked glycosylation sequon by introducing S301A and S305A mutations. Expression tests showed a substantial decrease in recovered protein levels when removing N303 glycosylation, with the double mutant yielding almost no Pfs48/45 6C (Fig. 3b). N-linked glycosylation was also critical to obtain a monomeric and monodisperse Pfs48/45 6C sample, as observed by size-exclusion chromatography (Fig. 3c). Importantly, the N303 N-linked glycan is distally located from the 85RF45.1 epitope. N-linked glycosylation resulting from expression in mammalian cells therefore favors production of well-folded Pfs48/45 6C, without affecting exposure of potent epitope I.
Humanization of mAb 85RF45.1. Next, we sought to humanize mAb 85RF45.1, introducing as much human sequence as possible while retaining sequences that preserve the binding characteristics of the antibody. The closest human V germlines to the parental rat sequence were found to be IGLV6-57 (66.2% for the light chain) and IGHV3-7 (84.0% for the heavy chain). Three humanized light chains and three humanized heavy chains were designed, all exceeding our threshold humanness scores (Supplementary Figure 2). In all cases, CDRs remained unchanged. Three humanized light and heavy chains were combined to create a total of nine fully-humanized variant antibodies (Supplementary Figure 2). All three light chain combinations within the HC2 group bound the antigen similarly and the HC2 group affinities were higher than the HC1 and HC3 groups (Supplementary Figure 2). The HC2-LC1 (hereafter named TB31F, Fig. 5a) was selected for further characterization based on high binding activity and high human-like scores.
To characterize the ability of mAb TB31F to recognize Pfs48/ 45, we solved the crystal structure of TB31F Fab in complex with Pfs48/45 6C to 2.6 Å resolution ( Fig. 5b  resolutions, respectively, revealed a strikingly similar paratope configuration (Fig. 5c). The all-atom rmsd between the two unliganded variable domains is 0.88 Å, primarily arising from differences in loop conformations within the framework regions (Fig. 5c). These structures also revealed how the variable region is optimally configured for binding Pfs48/45 even when unliganded (all-atom rmsd 85RF45.1 bound/unbound : 1.52 Å; TB31F bound/unbound : 1.25 Å). Despite this high structural similarity, amino acid changes during humanization resulted in electrostatic surfaces that are considerably different between the 85RF45.1 and TB31F variable regions, indicating the attainment of antibody resurfacing (Fig. 5d). Importantly, both 85RF45.1 Fab and TB31F Fab bound the antigen with nanomolar binding affinities (   Figures 5 and 6). These desirable biophysical properties for TB31F, combined with its retained nanomolar binding affinity and potent activity to inhibit parasite transmission, make this molecule an attractive biologic for malaria interventions.

Discussion
Pfs48/45 is a prominent target for the development of biomedical interventions that can interrupt the transmission of Pf and contribute to the decrease of malaria incidence. Here, we report the three-dimensional structure of the Pfs48/45 6C domain and delineate the epitope targeted by one of the most potent transmission-blocking mAbs yet described, 85RF45.1. Pfs48/45 6C adopts an SRS fold that consists of two β-sheets formed by a mix of parallel and anti-parallel β-strands. This fold has been proposed to derive from an ephrin-like precursor during evolution and is predicted to be adopted by at least 14 proteins in Pf 33,34 . Correspondingly, proteins of known structures that have the highest structural homology are parasitic proteins of this s48/ 45 family for which structures have been determined: Pf12 (Pf), Pf41 (Pf), and Sporosag (Toxoplasma gondii). It is thought that these glycosylphosphatidylinositol (GPI)-anchored parasitic proteins play an important role in cell adhesion and immune evasion, although no exact function has yet been assigned to Pf12 and Pf41 34 . Our structure also reveals that the electrostatic surface of Pfs48/45 6C is highly electronegative. Whether this property is associated with its function in male gamete fertility 13 and involved in ligand binding or scaffolding in a multi-protein complex 35 will be an important area of further studies. It will also be important to determine whether the potency of mAb 85RF45.1 is associated with disruption of the Pfs48/45 epitope I in protein-protein interactions.
To obtain the crystal structure of the 85RF45.  30 might also emanate from such favorable inter-domain stabilizing effects, possibly at the site occupied by the NAG residue here. Thus, our three-dimensional structures of Pfs48/45 6C now provide an opportunity for protein engineering to further stabilize the antigen, enhance its expression, and develop TBV immunogens that will optimally present epitope I to preferentially elicit potent transmission-blocking antibodies. mAb 85RF45.1 is one of the most potent antibodies identified against any transmission-blocking target, with an IC 80 of approximately 1-3 μg/mL in standard membrane feeding assay (SMFA) experiments. We also now reveal that its epitope is largely conserved: mAb 85RF45.1 can accommodate all sequence polymorphisms previously reported and still bind with nanomolar affinity. However, whether this antibody can block field isolates containing these polymorphisms remains to be determined. mAb 85RF45.1 represents a unique tool to assess the efficacy of antibodies to reduce Pf transmission and malaria incidence in affected communities. For this purpose, optimizing the developability of this rat-originating mAb was an important objective. Here, we show how humanization of mAb 85RF45.1 into mAb TB31F retained the exact molecular mode of recognition of Pfs48/45, its high-affinity binding to the antigen, and its potent transmission-reducing activity in SMFA. In addition, humanized TB31F Fab shows improved thermostability characteristics with~10°C higher T m and T agg compared to its ratderived counterpart. Thus, our data provide a salient example of how humanization can present an opportunity to simultaneously improve surface and biophysical profiles of non-human-derived mAbs, and will contribute to ongoing efforts to improve multiobjective antibody optimization functions 40,41 . mAb TB31F now offers a unique opportunity to test the efficacy of a Pfs48/45-based transmission-blocking intervention in humans. The mAb has entered good manufacturing production and preclinical safety testing. Phase I trials are planned to assess safety and tolerability. Pending a positive outcome, this will be followed by efficacy testing with the goal of establishing the level of mAb required to block parasite transmission from human to mosquito, as determined by direct skin feeding (infection prevalence endpoint). Comparison of these data to those from the laboratory-based SMFA, which employs an infection intensity (as opposed to infection prevalence) endpoint and is routinely used to inform go/no-go criteria in pre-and early-clinical testing, will enable "back-validation" of the SMFA to better inform its use as an early development stage-gate for measuring the efficacy of candidate vaccines and antibodies. Such results will also provide a benchmark for performance of future vaccines and validate the epitope studied here as an important target. Humanization of mAb 85RF45.1. Humanized antibodies were designed by creating multiple hybrid sequences that fuse select parts of the parental antibody sequence with the human framework sequences. Humanness scores, representing how human-like the antibody variable region sequence is, were calculated according to Gao et al. 42 . Based on the method employed, heavy chain scores of 79 or above were considered human-like; for lambda light chains, a score of 84 or above was considered humanlike. Full-length antibody genes for all nine heavy/light chain pairs were constructed by cloning the synthesized variable region sequences into expression vectors that contained the human IgG1 or lambda constant region genes.

Data availability
The crystal structures reported in this manuscript have been deposited in the Protein Data Bank: 6E62, 6E63, 6E64, 6E65. The authors declare that all other data supporting the findings of this study are available within the article and its Supplementary Information files are available from the authors upon request.  1  10  20  30  40  50  60  70  80  90  100  110   1  10  20  30  40  50  60  70  80  90  100  110  120  85RF45.1-HC   TB31F-LC   TB31F-