Protective antibodies target cryptic epitope unmasked by cleavage of malaria sporozoite protein

Abstract

The most advanced monoclonal antibodies (mAbs) and vaccines against malaria target the central repeat region or closely related sequences within the Plasmodium falciparum circumsporozoite protein (PfCSP). Here, using an antigen-agnostic strategy to interrogate human antibody responses to whole sporozoites, we identify a class of mAbs that target a cryptic PfCSP epitope that is only exposed after cleavage and subsequent pyroglutamylation (pGlu) of the newly formed N-terminus. This pGlu-CSP epitope is not targeted by current anti-PfCSP mAbs and is not included in the licensed malaria vaccines. MAD21-101, the most potent mAb in this class, confers sterile protection against Pf infection in a human liver-chimeric mouse model. These findings reveal a site of vulnerability on the sporozoite surface that can be targeted by next-generation antimalarial interventions.

Main Text:

Malaria is a devastating mosquito-borne disease that is caused by infection with Plasmodium parasites. In 2022 alone, 249 million cases of malaria and over 600,000 deaths were estimated globally, with the majority of these attributed to Plasmodium falciparum (Pf) (1). Malaria is most deadly for children under 5 years of age who account for 78% of deaths within the sub-Saharan African region (1). Infection by the Plasmodium parasite is initiated when the motile sporozoite is injected into the skin during a blood meal of an infected female Anopheles mosquito (2, 3). Sporozoites then breach the dermal capillary network and migrate to the liver via the bloodstream (4–6). Once in the liver, sporozoites may traverse multiple cell types before establishing infection within a hepatocyte (7), where the parasite then undergoes mitotic division and differentiates into exo-erythrocytic merozoites, which infect erythrocytes and initiate the symptomatic blood stage of the life cycle (8). Remarkably, a single sporozoite is sufficient to seed a full infection of a hepatocyte and gives rise to tens of thousands of infectious merozoites (9). Sporozoites represent a bottleneck in the parasite’s life-cycle as small numbers (500–1000) are deposited in the skin during blood meal acquisition and they are non-replicating prior to hepatocyte invasion (10–12). Antibodies delivered passively can prevent infection by neutralizing sporozoites before they invade hepatocytes (13–16) and, thus, sporozoites represent an attractive target for anti-malaria interventions.

The most advanced anti-sporozoite antigen strategies to date all target the P. falciparum circumsporozoite protein (PfCSP), an antigen that coats the sporozoite surface (17, 18). PfCSP contains a central repeat region, flanked by N- and C-terminal regions. The central repeat region consists of tandem, repeating Asn-Ala-Asn-Pro (NANP, major repeat) motifs interspersed with less frequent Asn-Val-Asp-Pro (NVDP, minor repeat) motifs and is directly flanked by the junctional Asn-Pro-Asp-Pro tetrapeptide (NPDP) on the N-terminal end of the repeats (19). The World Health Organization (WHO)-endorsed RTS,S/AS01 (20) and R21/MM (19, 21) vaccines both induce protective antibodies that target the repeat region. Although these vaccines represent a major advance in the fight against malaria, they have only been partially efficacious when tested in malaria-endemic regions (20–22). Sporozoite-neutralizing monoclonal antibodies (mAbs) offer a complementary strategy against malaria and have shown up to 88% protective efficacy against Pf infection when used at high doses in an endemic setting (23, 24). Anti-sporozoite mAbs in clinical development include MAM01 (25), which primarily targets the PfCSP major repeats, and CIS43-LS (14) and L9-LS (16), which cross-react with the major repeats but preferentially target the junction and minor repeats, respectively. Despite the potential abundance of antigenic targets on the sporozoite surface (26), most anti-sporozoite interventions have focused only on these well-defined repeat regions of PfCSP. To address this knowledge gap, we applied an antigen-agnostic approach to survey the landscape of antibody responses to whole Pf sporozoites at the single B cell level. This approach allowed us to screen for antibodies that showed detectable binding to intact sporozoites without restriction to any specific antigen, facilitating the discovery of functional mAbs that target other epitopes on the sporozoite surface.

Antigen-agnostic isolation of mAbs targeting Pf sporozoites

To investigate the antibody response to Pf sporozoites in an antigen-agnostic manner, we screened plasma from Pf sporozoite-exposed individuals for IgG reactivity to intact, freshly dissected Pf sporozoites by high-throughput flow cytometry. Two cohorts with different forms of sporozoite exposure were analyzed: individuals living in a rural community in Mali with exposure to intense, seasonal malaria (n = 843) (27, 28), and malaria-naïve individuals in the United States who were immunized with large numbers of radiation-attenuated sporozoites (n = 98) (29, 30). As expected, most immunized individuals had high levels of sporozoite-specific IgG (Fig. 1A). In contrast, in the Mali cohort, circulating sporozoite-reactive IgG was not universal but was associated with age (Fig. 1, A and B), suggesting that antibodies to sporozoites are acquired slowly with repeated natural infections. To focus our efforts on novel targets on the sporozoite surface, we repeated the plasma IgG screen but included a pre-blocking step with full-length recombinant PfCSP (rPfCSP) to adsorb PfCSP-specific antibodies. Blocking with rPfCSP ablated IgG reactivity to Pf sporozoites in 936 of 941 donors, highlighting the immunodominance of PfCSP on the sporozoite surface (Fig. 1A). However, plasma from one Malian and four immunized individuals retained IgG reactivity to the sporozoite surface after rPfCSP blocking (Fig. 1A). We confirmed that this observation was not due to residual reactivity to rPfCSP (fig. S1A). We hypothesized that these five donors harbored B cells with non-PfCSP specificities, and therefore selected these individuals for mAb isolation.

Fig. 1. Isolation of rare anti-sporozoite mAbs using an antigen-agnostic approach.

(A) Plasma IgG reactivity to Pf sporozoites with (red) and without (grey) pre-blocking of plasma samples by rPfCSP. Bars show mean. Donors with ≥8% reactivity to sporozoites after rPfCSP blocking in the initial screen were subjected to repeat analysis and these validated datapoints are shown in the plot for those donors. The dotted line shows the final 25% cutoff chosen to select positive donors, with down-selected donors enclosed in blue. (B) Correlation of plasma IgG reactivity to whole Pf sporozoites with donor age from the malaria-exposed cross-sectional cohort, p value and correlation coefficient were derived by Spearman’s rank correlation analysis. (C) Schematic of the antigen-agnostic workflow for isolation of mAbs against Pf sporozoites. IgG⁺ MBCs are sorted from PBMC samples and plated at a density of 25–100 cells per well in 384-well plates. The cells are activated and cultured for 9 d, followed by screening of supernatants for binding to Pf sporozoites and PfCSP. B cells of interest (PfSPZ⁺, PfCSP⁻) are transferred to a microfluidics chip for single-cell screening against Pf sporozoites and PfCSP. Single B cells of interest (PfSPZ⁺, PfCSP⁻) are selected for mAb production. PfSPZ, Pf sporozoites. Image produced using Biorender. (D) Number of MBC culture supernatants reactive towards whole Pf sporozoites and rPfCSP. (E) Titration curves for recombinant mAb binding to whole, freshly dissected Pf sporozoites. MAD22-17 bound more poorly to sporozoites than the other mAbs and was excluded from subsequent analysis. CIS43 is a control anti-PfCSP mAb and CV503 is a negative control anti-SARS-CoV-2 mAb. The dotted line shows the binding level with the buffer only control.

We developed a pipeline to screen single B cells from the five donors for reactivity to intact Pf sporozoites (Fig. 1C). This pipeline was divided into two steps to increase the probability of identifying rare B cells of interest. First, supernatants from oligoclonal cultures of 475,250 activated IgG⁺ memory B cells (MBCs) in 384-well plates were assessed for binding to Pf sporozoites and rPfCSP. Despite donor selection for reactivity to non-PfCSP targets, the vast majority of sporozoite-reactive B cell supernatants were also rPfCSP-positive (94.9%, n = 1015) (Fig. 1D), confirming that the strong immunodominance of PfCSP extends to the B cell level. Nevertheless, we identified a small subset of supernatants (5.1%, n = 55) that were sporozoite-reactive but rPfCSP-negative. Of these 55 supernatants, 3 originated from the Malian donor and the rest (52/55) originated from immunized donors. B cells from these wells were individually sorted into nanoliter-volume culture chambers and imaged in real-time to detect secretion of antibodies that bound to the surface of intact Pf sporozoites but not to rPfCSP-coated beads (fig. S1B). Single B cells with this profile were exported for sequencing and production of mAbs as recombinant IgG1. Using this approach, we identified ten mAbs that showed strong binding to Pf sporozoites but no appreciable binding to rPfCSP (Fig. 1E, fig. S1C). Out of the ten mAbs, five (MAD21-17, MAD21-46, MAD21-53, MAD21-95, MAD21-101) were isolated from the same donor (MAD21, a vaccinated donor) and belonged to the same IGHV4-59/IGKV4-1 clonal lineage (table S1). The remaining five mAbs were isolated from two other immunized donors (MAD22 and MAD24) and were encoded by a variety of V genes (IGHV4-39/IGKV1-39, IGHV3-11/IGKV3-11, IGHV3-7/IGLV1-47, IGHV3-33/IGKV2-24). The mAbs bound to 100% of Pf sporozoites in tested samples in an analogous manner to the control anti-PfCSP mAb CIS43, indicating that the target antigen is expressed on each individual sporozoite (Fig. 1E, fig. S1D). We confirmed that the mAbs bound similarly to freshly isolated and cryopreserved sporozoites, enabling us to use cryopreserved sporozoites in downstream binding and Western blot experiments (Fig. 1E, fig. S1E).

Isolated mAbs recognize sporozoite-expressed PfCSP and are functional in vivo

To identify the target antigens of the newly isolated mAbs, we screened them against Pf sporozoite lysate by Western blot. Surprisingly, under both reducing and non-reducing conditions, the mAbs interacted with a protein at the same molecular weight as PfCSP (Fig. 2A, fig. S2, A and B), introducing the possibility that they recognize PfCSP, but only when expressed by sporozoites. Additionally, analysis of binding to sporozoites by an indirect immunofluorescence assay (IFA) revealed that the mAbs bound the sporozoite surface with similar localization as CIS43, a control anti-PfCSP mAb (Fig. 2B). Crucially, the mAbs did not bind to wild-type Plasmodium berghei sporozoites (Pb^WT) but bound strongly to transgenic Pb sporozoites expressing PfCSP (Pb^PfCSP) (Fig. 2C), indicating that the target of these mAbs was sporozoite-produced PfCSP.

Fig. 2. Isolated mAbs exclusively target Plasmodium-expressed PfCSP and reduce liver parasite burden in vivo.

(A) Western blot analysis of Pf sporozoite lysates. Lysates were analyzed by reducing SDS-PAGE and individual blots were probed with representative isolated mAbs (MAD21-101, MAD22-38 and MAD24-01), a control anti-PfCSP mAb (CIS43), and an anti-SARS-CoV-2 spike mAb as an isotype control (CV503). Lysate from 5 × 10⁴ sporozoites was loaded into each well. (B) Immunofluorescence images of Pf sporozoite surface staining with MAD21-101 and MAD22-38, an anti-PfCSP control mAb CIS43, and an isotype control anti-SARS-CoV-2 mAb CV503. Scale bar = 2 μm. (C) Flow cytometry analysis of mAb binding to freshly dissected sporozoites: wild-type P. berghei sporozoites and transgenic P. berghei sporozoites that express full-length PfCSP (Pb^PfCSP). CIS43 was included as a control anti-PfCSP mAb and the dotted line denotes MFI for the buffer control. WT, wild-type. (D and E) In vivo assay to evaluate mAb potency, as measured by liver parasite burden reduction. In this assay, 300 μg of each mAb was delivered 16 h pre-IV challenge with Pb^PfCSP sporozoites that express the luciferase reporter enzyme (Pb^{PfCSP-GFP/FLuc}). Liver parasite burden was measured by in vivo imaging and quantified as luminescence flux signals. The naïve group consisted of mice that were not infected with sporozoites and were used to determine the baseline flux signal, while the “no mAb” group was infected with sporozoites but not administered any mAb. (D) Liver burden reduction from four independent experiments. Data are shown with geometric mean for n = 5 mice per group and statistical significance was determined versus the no mAb condition using one-way ANOVA testing with Dunnett’s Test for multiple comparisons. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 and ns, not significant. (E) Potency of the mAbs calculated based on sporozoite neutralization. Neutralization data are derived from at least 3 independent experiments. Bars show mean with standard deviation.

The Pb^PfCSP parasites, which also carry a dual GFP/Luciferase reporter system, are routinely used in a standardized mouse infection model to assess in vivo efficacy of PfCSP-specific mAbs (31). Using this model, we evaluated the 10 mAbs and found that six mAbs reduced liver parasite burden by more than 50% following intravenous sporozoite challenge (Fig. 2, D and E, fig. S2C). MAD21-101 was the most potent, providing 97.5% neutralization at a 300 μg dose and MAD22-38 was the most potent mAb outside the MAD21 clonal lineage (Fig. 2, D and E). Therefore, we focused on MAD21-101 and MAD22-38 for downstream analysis. When the mAbs are analyzed collectively, we referred to them as MAD21-101-type mAbs.

Two potential mechanisms by which antibodies inhibit sporozoite activity in vivo, include reducing their motility, or blocking development of liver-stage parasites in hepatocytes by inhibiting the sporozoite traversal or invasion processes that are required for productive infection. To investigate these possibilities, we tested a subset of MAD21-101-type mAbs for their ability to reduce motility in a trail gliding assay or inhibit Pf liver stage development within primary human hepatocytes in vitro using the in vitro inhibition of liver stage development assay (ILSDA). Inhibition of sporozoite motility by MAD21-101 was modest at 37.2% for the highest dose tested (100 μg/mL) (fig. S3, A and B). In contrast, 100 μg/mL of MAD21-101 and MAD22-38 substantially reduced liver stage parasite development in the ILSDA, with MAD21-101 providing 88.8% inhibition and performing similarly to an equivalent dose of CIS43 that conferred 94.1% inhibition (fig. S3C). Thus, the mechanism of protection conferred by the mAbs is likely dependent on their ability to inhibit sporozoite invasion or traversal activity that leads to productive infection.

MAD21-101-type antibodies target sporozoite-cleaved and pyroglutamated PfCSP

We hypothesized that the lack of mAb reactivity for rPfCSP results from a post-translational modification performed by sporozoites that was absent in rPfCSP produced in mammalian HEK293 cells. Therefore, we evaluated MAD21-101 and MAD22-38 binding to rPfCSP produced by cell lines of diverse bacterial and eukaryotic origin (Escherichia coli, Lactococcus lactis, Pichia pastoris, Sf9, wheat germ), reasoning that one of these cell lines may produce PfCSP with post-translational modifications that more closely mimic native Plasmodium expression. However, we did not observe binding to any form of rPfCSP tested (tables S2 and S3). Next, we tested MAD21-101 and MAD22-38 for binding to a series of 15-mer peptides spanning the entire protein. The mAbs did not bind to any of these constructs or peptides, which was surprising given the likely linearity of the target epitope based on the Western blot results with sporozoite lysate (Fig. 2A, table S2). These findings suggested that the mAbs require a sporozoite-specific modification of PfCSP that was absent in the recombinant or chemically synthesized constructs, and that further analysis would have to be performed with sporozoite-expressed PfCSP.

To identify the PfCSP region recognized by MAD21-101 and MAD22-38, we performed ELISAs with transgenic Pb sporozoite lines expressing PbCSP with specific regions deleted or substituted with corresponding regions from PfCSP. We confirmed that neither mAb bound to wild-type Pb sporozoites and, furthermore, we found that in-substitution of the PfCSP N-terminus or C-terminus did not restore binding, suggesting that these regions do not contain the target epitope (Fig. 3A, fig. S4A). As expected, MAD21-101 and MAD22-38 bound to Pb sporozoites expressing full-length PfCSP (Fig. 3B). However, binding was completely ablated upon deletion of the junction amino acids A⁹⁸DGNPDP¹⁰⁴, which lie immediately C-terminal to the highly conserved region I motif K⁹³LKQP⁹⁷ (Fig. 3B). Furthermore, insertion of A⁹⁸DGNPDP¹⁰⁴ (alongside four NANP repeating units) into full-length PbCSP rescued mAb binding, while insertion of a stretch of 12 NANP repeats alone did not (Fig. 3B). Collectively, these findings indicated that the target epitope of MAD21-101 and MAD22-38 partially overlaps with A⁹⁸DGNPDP¹⁰⁴. We searched for sporozoite-specific modifications to this region and did not find evidence of parasite glycosylation at this site. However, previous studies reported that an unidentified Plasmodium cysteine protease cleaves PfCSP at the highly conserved region I, which would leave the A⁹⁸DGNPDP¹⁰⁴ sequence close to the new N-terminus and potentially expose an epitope for mAb binding (32). We determined whether MAD21-101-type mAbs bind only to cleaved PfCSP by running a Western blot under conditions that would resolve cleaved and uncleaved protein. Indeed, MAD21-101 and MAD22-38, as well as the other mAbs isolated in this study, had no detectable reactivity to uncleaved PfCSP (~50kDa) and bound only to the cleaved form of the protein (~43kDa) (Fig. 3C, fig. S4B). MAD21-101 only immunoprecipitated cleaved PfCSP from sporozoite lysate, further confirming its specificity for cleaved protein (fig. S4C). The identification of MAD21-101 gave us the opportunity to investigate whether individual sporozoites can display cleaved and uncleaved PfCSP simultaneously on their surface. To answer this question, we stained Pf sporozoites with 5D5, a mAb that binds to the PfCSP N-terminus (and thus binds only to uncleaved PfCSP), along with MAD21-101 (which binds only to cleaved PfCP). The majority of sporozoites could be bound simultaneously by both mAbs, indicating that each sporozoite can display both forms of PfCSP at the same time (fig. S4D).

Fig. 3. MAD21-101-type mAbs target the newly formed, pyroglutamate-modified N-terminus of PfCSP after sporozoite cleavage.

(A) Schematic of PfCSP depicting the highly conserved region I (brown), the junction (beige) and central major NANP repeats (sky blue) interspersed with minor NVDP repeats (black), which are flanked by the N- and C-terminus (pink and green, respectively). Below, the sequence of region I, junction and initial repeats are shown. (B) ELISA analysis of mAb binding to lysate from Pb sporozoites that express full PfCSP (Pb^PfCSP) or PfCSP lacking key junction region residues (Pb^PfCSP-JRKO), and chimeric Pb sporozoites that express the PfCSP junction and NANP repeats (Pb^PfJNANP4) or PfCSP NANP repeats only (Pb^PfNANP12). Sequences and domains in orange are derived from PfCSP, while sequences and domains in blue are derived from PbCSP. SARS-CoV-2 spike mAb CV503 was included as an isotype control. (C) Western blot analysis of mAb binding to cleaved and uncleaved forms of PfCSP from Pf sporozoite lysate. Lysates were analyzed by SDS-PAGE under reducing conditions and individual blots were probed with MAD21-101, MAD22-38 and MAD24-01. Control mAbs that bind to uncleaved PfCSP only (5D5) or both cleaved and uncleaved PfCSP (CIS43) were included. CV503 IgG1 was included as an isotype control and lysate from 5 × 10⁴ sporozoites was loaded into each well. (D) Representative titration curves and AUC heat map showing mAb binding to N-terminally truncated PfCSP peptides. AUC, area under the curve. (E) AUC heat map showing mAb binding to PfCSP peptides with either an N-terminal Gln⁹⁶ or pGlu⁹⁶ residue. (F) An equimolar mixture of peptides with N-terminal Gln⁹⁶ or pGlu⁹⁶ was immunoprecipitated with MAD21-101 or CV503 (isotype control) and analyzed by LC-MS. The mAb-depleted supernatants were analyzed in addition to eluates. Shown are extracted ion chromatograms for the peptides. Retention times are aligned. Peptide abundance is normalized and offset for clarity. Peptide identity was confirmed from fragment spectra.

Our low-resolution PfCSP peptide screen (4-residue shift between peptides) did not reveal mAb binding to peptides that should cover the N-terminus of cleaved PfCSP, assuming cleavage occurs at region I (table S2). Therefore, we hypothesized that the mAbs are highly specific to the precise N-terminus generated by sporozoite cleavage, which was presumably not represented in this low-resolution peptide assay. While a recent model proposes the PfCSP cleavage site to be between Lys⁹⁵ and Gln⁹⁶ in region I (33), the exact location of cleavage has not been experimentally determined. Therefore, we assessed binding to a panel of peptides with single residue truncations across region I to more precisely mimic potential sporozoite cleavage sites. Except for MAD24-52, all of the mAbs bound only when glutamine (Gln⁹⁶) was presented as the N-terminal residue (Fig. 3D). However, the binding of most MAD21-101-type mAbs, including MAD22-38, was surprisingly weak given their strong binding to sporozoites (Fig. 1E), suggesting that the Gln⁹⁶ peptide did not fully recapitulate the native epitope on the sporozoite surface. It appears that parasite glutaminyl cyclase is active in Plasmodium sporozoites and likely converts P. berghei CSP Gln⁹², the equivalent of PfCSP Gln⁹⁶, to pyroglutamic acid (pGlu) to enhance immune evasion in mosquitoes (33). Therefore, we compared mAb binding to PfCSP peptides bearing either an N-terminal Gln⁹⁶ or pGlu⁹⁶ residue. Binding was markedly enhanced to peptides with N-terminal pGlu⁹⁶ (Fig. 3E, fig. S5A), indicating that this is the likely native target of the mAbs.

N-terminal glutamine has been shown to spontaneously convert to pyroglutamate in solution, even in the absence of glutaminyl cyclase (34). Indeed, when we used liquid chromatography-mass spectrometry (LC-MS) to analyze the synthetic Gln⁹⁶ peptide employed in the mAb binding assay described above (QPADGNPDPNANP-Ahx-Lys[biotin]), we found that 5.4% of the N-terminal residue had already undergone conversion to pGlu⁹⁶, even though the peptide had been stored at −20°C since reconstitution (fig. S5B). Therefore, we investigated whether the observed mAb reactivity to the Gln⁹⁶ peptide was due to spontaneously converted pGlu⁹⁶, rather than true binding to Gln⁹⁶. To distinguish between binding to the two peptides, we immunoprecipitated a mixture of Gln⁹⁶ and pGlu⁹⁶ peptides with MAD21-101 and analyzed the bound and eluted peptides by LC-MS, which unequivocally distinguished between the two isoforms. MAD21-101 bound strongly to the pGlu⁹⁶ peptide but exhibited no binding to the Gln⁹⁶ peptide (Fig 3F), indicating that MAD21-101 is highly selective for pGlu-CSP.

To determine whether the pGlu⁹⁶ epitope can be detected in sporozoites, we used LC-MS to analyze PfCSP from Pf salivary gland sporozoites, starting with published datasets. In the published Pf salivary gland sporozoite global (35) and surface proteomes (26), PfCSP is highly abundant and peptides beginning with Gln⁹⁶ (i.e., arising from cleavage of region I between Lys⁹⁵ and Gln⁹⁶) are detected in every sample. Furthermore, it was observed that virtually all of the Gln⁹⁶ had been converted to pGlu⁹⁶. However, it is not possible to determine from these datasets whether the pGlu⁹⁶ epitope was present in vivo, for two reasons: first, the peptides analyzed in these studies were generated by digestion with trypsin, which cleaves C-terminal to Lys and Arg residues, so we do not know whether the Gln⁹⁶ peptides were produced by trypsin digestion or if they arose from sporozoite processing of the K⁹³LKQP⁹⁷ motif. Second, we do not know what proportion of the pGlu observed was present in vivo as opposed to forming spontaneously during sample handling. To circumvent these issues, we employed a different approach to obtain PfCSP peptides. In published sporozoite proteomes, the repeat units within PfCSP were observed to be highly heat labile (26, 35). We therefore hypothesized that the region of PfCSP immediately N-terminal to the repeats could be liberated non-enzymatically by heat alone. To this end, frozen sporozoite pellets were resuspended in aqueous buffer and briefly heated to induce non-enzymatic protein fragmentation. Acetonitrile was then added, causing intact protein to precipitate while released peptides remained in the supernatant, which was recovered and analyzed by LC-MS. We observed peptides with N-terminal pGlu⁹⁶ of various lengths that arose from heat-induced cleavage of the major and minor repeat motifs (fig. S5C). Importantly, in 100% of these peptides the N-terminal Gln⁹⁶ had been converted to pGlu. To confirm that these observations were not artifacts due to sample handling, we subjected rPfCSP to the same protocol. While we detected peptides containing the intact K⁹³LKQP⁹⁷ motif, we did not observe any peptides with N-terminal Gln⁹⁶ or pGlu⁹⁶, suggesting that this protocol does not induce cleavage between Lys⁹⁵ and Gln⁹⁶ (fig. S5C). To quantify the extent to which sample handling induces spontaneous conversion of N-terminal Gln to pGlu, a synthetic Gln⁹⁶ peptide (QPADGNPDPNANP) was subjected to the same protocol as the sporozoites and rPfCSP. The heating and peptide extraction process increased the relative proportion of spontaneously converted pGlu from 2.1% to 11.5%, but the majority (88.5%) of the peptide N-terminus remained as native Gln (fig. S5D), confirming that proteomic sample handling alone does not account for the fact that the N-terminal Gln⁹⁶ is 100% converted to pGlu in salivary gland sporozoites. Taken together, these data show that in PfCSP of Pf salivary gland sporozoites, the K⁹³LKQP⁹⁷ motif can be cleaved at Lys⁹⁵ and the N-terminus, Gln⁹⁶, is converted to pGlu⁹⁶.

The pGlu-CSP epitope is distinct from previously identified PfCSP sites

To determine the minimal epitope required for binding, we screened the pGlu-CSP-specific mAbs against a series of peptides that contained sequential C-terminal truncations. We found that the major (NANP) and minor repeat (NVDP) at the C-terminal end of the peptide were not required for binding (Fig. 3E and fig. S6) and that the minimal epitope for most mAbs, including MAD21-101 and MAD22-38, was pGlu⁹⁶PADGNP¹⁰² (fig. S6). We repeated the peptide screen with previously isolated anti-PfCSP mAbs that target the major repeats (317 (36)), minor repeats (L9 (16)), junction (CIS43(14)), and N-terminus (5D5 (37)). None of these mAbs bound to the pGlu⁹⁶PADGNP¹⁰² sequence, with the mAb that bound the closest epitope, CIS43, targeting a more C-terminal epitope that required the extension of the sequence beyond the first major repeat unit (N¹⁰⁵ANP¹⁰⁸) (fig. S6).

MAD21-101-type antibodies bind a hydrophobic and aromatic pocket to stabilize pGlu-modified PfCSP

To determine the structural basis for the selectivity of MAD21-101-type antibodies to the epitope that we discovered and was produced by pGlu modification of cleaved PfCSP, we determined X-ray crystallographic structures of the antigen binding fragments (Fab) of MAD21-101, MAD22-38, and MAD24-01 in complex with an 18aa peptide corresponding to PfCSP pGlu⁹⁶-Asn¹¹³ at resolutions of 1.46Å, 1.99Å, and 1.82Å, respectively (table S4) (note, Kabat numbering for Fab residues is used throughout). In all three structures, the Fabs bind the peptide with the N-terminus pointing inwards into the antigen-binding groove (Fig. 4A); the N-terminal pGlu-Pro motif is buried in a pocket formed by multiple CDRs of the antibody heavy and light chains (Fig. 4A, inset). In the MAD21-101 structure, peptide electron density can be observed for the entire 18aa peptide, forming a buried surface area (BSA) of 609Ų (Fig. 4B). The C-terminal Val¹⁰⁹-Asn¹¹³ protrudes out of the binding site and interacts with an adjacent Fab that is related by crystallographic symmetry. However, the disorder in the peptide C-terminus indicated by the weak electron density suggest these are fortuitous contacts in the crystallographic lattice due to close proximity of adjacent Fabs and are not part of the native paratope (fig. 4B). In the structures of MAD22-38 and MAD24-01, clear peptide electron density is visible for residues pGlu⁹⁶-Asp¹⁰³, in agreement with the minimal epitope determined by peptide truncation experiments (fig. S6A), and forms BSA of 319Ų and 351Ų, respectively (Fig. 4, C and D). This insertion of the newly cleaved and cyclized N-terminus into the antibody combining site is consistent with the requirement for PfCSP cleavage for antibody recognition and binding. To achieve selectivity for the N-terminal pGlu⁹⁶ compared to the unmodified Gln⁹⁶ or PfCSP cleavage at alternative residues, MAD21-101 interacts extensively with the pGlu-Pro motif using a vast network of aromatic and hydrophobic residues (Tyr50^H, Phe100E^H, Tyr27D^L, Tyr32^L, Tyr91^L_, Tyr92^L, and Leu93^L; n.b. the antibody residue numbers are not superscripted to differentiate them from the CSP residues, and L is light chain and H is heavy chain) forming a cage-like arrangement surrounding the pGlu-Pro, where the aromatic residues make hydrophobic and CH/π interactions. Further stabilization of the peptide N-terminus is achieved through hydrogen bonding to the secondary carbonyl and amide of the N-terminal pGlu⁹⁶ via polar groups (Tyr50^H hydroxyl, Ala100F^H backbone carbonyl) (Fig. 4E). A similar binding mode is seen in the MAD22-38 Fab-peptide complex, where the pocket for pGlu-Pro is formed by a set of aromatic and hydrophobic residues from multiple CDRs of both heavy and light chains (Tyr35^H, Tyr58^H, Ile95^H, Trp97^L and Tyr100D^H), and with additional coordination of the N-terminal pGlu carbonyl by Ser50^H hydroxyl and Trp97^L indole nitrogen (Fig. 4F). A more minimal binding pocket is seen in the MAD24-01 Fab-peptide complex, formed by the HCDR3 and LCDR3 loops, with the pGlu-Pro motif sandwiched between Trp100A^H, Val100B^H, Trp94^L, and Leu96^L with additional polar interactions by nearby Fab side chains (Arg91^L) and backbone (Val100B^H) (Fig. 4G). We aligned the Fab sequences and their respective germline V genes with the per-residue BSA formed by peptide binding and interactions coordinating the bound peptide (fig. S7). Several key binding residues forming the core binding pocket are germline-encoded in each of the three Fabs, with somatic hypermutation clustering around contact regions indicative of affinity maturation towards this specific target to further increase binding contacts. Therefore, despite their different donor source and distinct V_H/V_L combinations, MAD21-101-like Abs employ analogous features to produce hydrophobic and aromatic binding pockets that accommodates the pGlu-Pro motif, thereby achieving their common selectivity and specificity to this novel PfCSP epitope.

Fig. 4. Structural basis of MAD21-101-like antibody epitope recognition and specificity.

(A) Overall view of X-ray structures of MAD21-101, MAD22-38, and MAD24-01 in complex with a peptide corresponding to PfCSP residues pGlu⁹⁶-Asn¹¹³. The Fab surface shown in gray, bound peptide as a yellow backbone cartoon representation, framework regions in gray cartoons, and CDR1, CDR2, and CDR3 of Fab heavy and light chains are shown in dark and light green, blue, and purple, respectively. Inset showing the bound N-terminus in a pocket at the heavy-light chain interface, with pGlu-Pro motifs represented as yellow sticks with the electron density map (2Fo-Fc) in blue mesh (contoured to 1σ). Bound peptide of (B) MAD21-101, (C) MAD22-38, and (D) MAD24-01 shown in yellow sticks and electron density map (2Fo-Fc) in blue mesh (contoured to 1σ). Per-residue peptide buried surface area (Ų) at the peptide-Fab interface shown as yellow bars. (E-G) Binding interactions between the N-terminal peptide residues pGlu⁹⁶-Asp⁹⁹ to the hydrophobic pockets of (E) MAD21-101, (F) MAD22-38, and (G) MAD24-01, with interacting residues shown as sticks and hydrogen bonds in black dashed lines.

Although showing similar binding to their epitope, MAD21-101-like antibodies display a broad range of liver burden reduction in vivo (Fig. 2D and fig. S2C). Previously, the binding affinities of antibodies targeting CSP has been correlated with protection (38–40). To determine if this was the case for the pGlu-CSP epitope, we performed biophysical analysis of the binding of 7 Fabs to their PfCSP epitope peptide by isothermal titration calorimetry (ITC) (fig. S8A). The highly protective MAD21-101 produces the strongest binding with an equilibrium dissociation constant (Kd) of 3.8 nM, MAD22-38 with Kd of 111 nM, and non-protective MAD24-01 produces the lowest binding affinity (850 nM) to the target peptide. Across all Fabs tested, antibody affinity measured by ITC exhibits a positive correlation to liver-burden reduction with R² of 0.753 (fig S8B), and a strong correlation R² = 0.952 was observed for MAD21-lineage antibodies (fig S8C). In the MAD21-101 structure, the core epitope of the bound peptide is stabilized by a type-I β-turn between the backbone carbonyl of pGlu⁹⁶ and the amide nitrogen of Asp⁹⁹, while MAD22-38 and MAD24-01 bind the peptide in their respective binding pockets in an extended conformation. Although the enthalpy of peptide binding is comparable in both MAD21-101 and MAD22-38 (fig S8A), the lower entropic penalty of binding for MAD21-101 compared to MAD22-38 contributes to a 30-fold higher affinity for MAD21-101, potentially by decreasing conformational sampling required for peptide binding in solution. Stabilization of secondary structural elements is commonly observed in potent PfCSP-binding mAbs (38), and antibody stabilization of different CSP peptide conformations has been associated with functional differences (41). Similar to other CSP epitopes, the specific conformation of pGlu-CSP stabilized by MAD21-lineage antibodies may facilitate the high-affinity binding observed for MAD21-101 and contribute to its protection in the mouse model.

The pGlu-CSP epitope is conserved in global Pf isolates

To determine the degree of conservation of the pGlu-CSP epitope, we examined sequences from 16,441 globally isolated Pf parasites based on the MalariaGEN Pf7 database (42) and compared them to the canonical Pf3D7 sequence Q⁹⁶PADGNPDPNANP¹⁰⁸. Only two nucleotide polymorphisms that led to amino acid changes, both at location Ala⁹⁸, were identified. A small number of sequences (five out of 16,441) had an Ala⁹⁸Val substitution, while most sequences (9,901 out of 16,441) had an Ala⁹⁸Gly substitution, which is consistent with previous reports (43–45) (fig. S9A). We tested whether the MAD21-101-type mAbs retain the ability to bind to the pGlu-CSP epitope with this Ala⁹⁸Gly mutation (Q⁹⁶PGDGNPDPNANP¹⁰⁸), while also examining binding to peptides with sequential alanine substitutions to determine key epitope residues. Most mAbs retained binding to the Ala⁹⁸Gly peptide (Fig. 5, A and B). Consistent with the minimal epitope results and our crystal structures, residues that affected binding by MAD21-101-type mAbs were located within the pGlu⁹⁶PADGNP¹⁰² sequence (fig. S6, Fig. 5A). The key binding residues for MAD21-101 were the conserved residues pGlu⁹⁶, Pro⁹⁷ and Asp⁹⁹, while MAD22-38 also required Gly¹⁰⁰ (Fig. 5B, fig. S9B). Although Ala⁹⁸Gly did not affect binding of most mAbs, the rarer Ala⁹⁸Val substitution reduced binding by all the mAbs to different degrees, with MAD24-01 being the least affected by the mutation, followed by MAD21-101 (fig. S9C). We also confirmed that the mAbs did not bind to a peptide containing an N-terminal pGlu followed by a series of NANP repeats (fig. S9D), suggesting that the MAD21-101-type mAbs bind specifically to the pGlu-CSP sequence and not the N-terminal pGlu residue alone. This finding is consistent with our crystal structures showing specific hydrogen-bonding interactions and significant buried surface area produced by most residues across the minimal epitope. Collectively, these findings suggest that MAD21-101-type mAbs target a conserved PfCSP-specific epitope that includes a critical N-terminal pGlu residue.

Fig. 5. MAD21-101 binds to a new conserved epitope within PfCSP and provides sterile protection against Pf sporozoite infection.

(A) Heat map showing binding of MAD21-101-type mAbs to pGlu-PfCSP peptides bearing alanine or glycine (when original residue was alanine) substitutions. Colors are based on area under the curve (AUC) binding relative to binding to the wild-type peptide. The red rectangle highlights the common A98G substitution. (B) Binding of MAD21-101 to pGlu-CSP peptides bearing alanine or glycine (when original residue was alanine) substitutions. Bars represent binding in AUC versus binding to the wild-type peptide sequence and the red line denotes 50% binding versus the wild-type sequence. (C) Binding of MAD21-101-type and previously identified mAbs (317 and 224) to the PfCSP-based R21 vaccine. Analysis performed using an Meso Scale Discovery (MSD) assay. An anti-COV-2 spike IgG1 (CV503) was included as an isotype control. (D) Parasitemia in human liver-chimeric (FRG-huHep) mice after mosquito bite challenge with Pf sporozoites. An anti-Pfs25 IgG1 (mAb 1245) was included an isotype control and the red dotted line represents the limit of detection for the qRT-PCR assay (5 parasites/mL). Statistical significance was determined versus the isotype control using a mixed-effects model with Šídák’s multiple comparisons test. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001 and ns, not significant.

pGlu-CSP accounts for residual plasma IgG reactivity to sporozoites

To determine the contribution of pGlu-CSP-specific antibodies to the overall sporozoite antibody response, we assessed the effect of adsorbing plasma antibodies using combinations of the pGlu-CSP peptide (pGluPADGNPDPNANP) and rPfCSP. In particular, we focused on the five donors that had residual sporozoite reactivity after rPfCSP blockade, as we were interested in whether the residual reactivity was exclusively due to pGlu-CSP or if non-PfCSP antigen reactivity contributed to a portion of that signal. As expected, blocking plasma from each donor with pGlu-CSP peptide alone had little effect on reducing sporozoite reactivity, due to strong remaining reactivity from conventional anti-rPfCSP antibodies (fig. S10). Upon blocking with rPfCSP alone, three donors (VRC_1A2, VRC_1F7 and VRC_1I6) showed a reduction in sporozoite reactivity while the other two (VRC_2B4 and M9_G09) did not, likely because the latter donors had a higher concentration of residual antibodies that could still bind to sporozoites (fig. S10). Importantly, for all five donors, blocking the plasma with a combination of pGlu-CSP and rPfCSP completely ablated sporozoite reactivity, confirming that antibodies directed to the pGlu-CSP epitope account wholly for the residual sporozoite reactivity observed after PfCSP blocking (fig. S10). Taken together with the initial screening results (Fig. 1A), these findings suggest that all observed polyclonal IgG reactivity to the sporozoite surface in the 941 malaria-exposed and sporozoite-immunized donors could be attributed to PfCSP, when the new pGlu-CSP epitope was also taken into account.

MAD21-101 does not bind to the R21 malaria vaccine and confers sterile protection in an in vivo model of Pf infection

Given the projected widespread use of the R21 and RTS,S malaria vaccines, a desirable feature of a mAb being developed for malaria prevention is non-interference between the mAb and these vaccines (46). Therefore, we tested whether the MAD21-101-type mAbs bind to the R21 malaria vaccine, which has the same PfCSP sequence as RTS,S. We compared these mAbs to mAb 317, a potent NANP-specific mAb (36), as well as mAb 224, the precursor to MAM01, which is being developed as a clinical candidate (25). Unlike mAbs 317 and 224, the MAD21-101-type mAbs showed no binding to R21 (Fig. 5C).

Since the unveiling of the pGlu-CSP epitope relies on parasite processing, an in vivo model that uses wild-type Pf sporozoites may be more appropriate for mAb evaluation, as transgenic Pb^PfCSP parasites may process non-native PfCSP differently from wild-type parasites. Therefore, we turned to a model that utilizes human liver-chimeric mice (FRG huHep) that are permissive to infection with Pf sporozoites by mosquito bite, after which parasitemia can be evaluated using a highly sensitive qRT-PCR approach (47). We confirmed that mice administered an isotype control mAb (48) and subsequently infected with Pf sporozoites were highly parasitemic at days 7 and 9 after infection (median 2.1 × 10⁶ parasites/mL blood on day 7, and 7.4 × 10⁵ parasites/mL blood on day 9). In contrast, all five mice that received 300 μg of MAD21-101 were parasite-negative on day 7, and only one mouse had parasitemia just above the limit of detection on day 9 (Fig. 5D). MAD21-101 was also capable of providing partial protection at lower doses: two out of four mice that received 100 μg of MAD21-101 were parasite-negative on days 7 and 9 (50%), while two out of six mice that received 30 μg MAD21-101 were protected on both days (33%). These findings indicate that MAD21-101 can confer sterile protection against Pf sporozoite infection in vivo.

Discussion

The malaria protein PfCSP is the main component of the two WHO-endorsed malaria vaccines and the target of the most advanced mAbs in clinical development (14, 16, 20, 22, 25). Three PfCSP epitopes targeted by potent neutralizing mAbs have been identified to date: (i) the major NANP repeats, (ii) the minor NVDP repeats, and (iii) the NPDP junction. The three sites have similar sequences centered around the same (N/D)PNANPN(V/A) core motif (15, 41), resulting in extensive cross-reactivity between mAbs targeting these sites. The commonality of these epitopes is highlighted by the finding that the malaria vaccine RTS,S contains the major repeats but not the junction or minor repeats, and yet can elicit mAbs that bind with high affinity to the two latter sites (25). A key goal of ongoing work in this field is to identify next-generation mAbs that target conserved sporozoite antigens or epitopes that are distinct from these sites. In this study, we identify a fourth target of potent anti-PfCSP mAbs and the first non-repeat target. Unlike the three established sites, this pGlu-CSP site is only exposed after sporozoite processing of PfCSP, which entails cleavage of the N-terminus and the conversion of the new N-terminal Gln⁹⁶ residue to an atypical amino acid, pGlu⁹⁶. PfCSP cleavage has been linked to sporozoite infection of hepatocytes and shown to occur at an indeterminate site within the highly conserved region I (K⁹³LKQP⁹⁷) (32, 37, 49). A recent study on pyroglutamylation of Plasmodium proteins proposed a model wherein PfCSP cleavage occurs between Lys⁹⁵ and Gln⁹⁶, followed by conversion of Gln⁹⁶ to pGlu⁹⁶ (33). However, the details of cleavage, including the identity of the parasite protease, the exact cleavage location, and the mechanism of how cleavage contributes to hepatocyte invasion, have not been experimentally verified. Our study provides several pieces of evidence that illuminate this process. First, it provides the first direct experimental evidence that PfCSP cleavage occurs between Lys⁹⁵ and Gln⁹⁶, validating the model described above (33). Interestingly, while the aforementioned study suggested that pyroglutamylation of Plasmodium proteins enables immune evasion in mosquitos, our work suggests that pyroglutamylation at this specific PfCSP site can increase parasite recognition by neutralizing human antibodies. These findings point to a potential tension between the need for Plasmodium sporozoites to evade both mosquito and human immune defenses, such that the evolution of a parasite protein to evade one host may lead to increased vulnerability in the other. Second, this study clarifies the long-standing observation that sporozoites dissected from mosquito salivary glands have a mixture of at least two clearly distinct PfCSP species (50), by providing evidence that the lower MW species has been cleaved at region I (K⁹³LK|QP⁹⁷) and has already undergone conversion to pGlu⁹⁶. The identification of pGlu-CSP-specific mAbs offers a new resource for the research community to further understand the process of PfCSP cleavage and, in particular, how this process is important for sporozoite invasion of hepatocytes.

The level of protection provided by the pGlu-CSP mAbs, particularly in the Pf in vivo infection model, suggest that this class of mAbs could be a valuable addition to the malaria prevention arsenal. To augment protection against malaria in certain vulnerable populations, the strategy of administering an anti-sporozoite mAb before or after RTS,S/AS01 or R21/MM vaccination is being considered (NCT06461026). For example, infants might benefit from a long-acting anti-sporozoite mAb shortly after birth since they are only eligible to receive RTS,S/AS01 or R21/MM at 5 months of age, a potentially cost-effective strategy given currently estimated mAb production costs of US $5–10 per dose for infants based on their low weight (51). For this strategy to be successful, it is essential that the mAb not interfere with vaccination through sequestration or epitope masking (46). The RTS,S/AS01 and R21/MM vaccines, which are being deployed widely in sub-Saharan Africa in 2024, do not contain the pGlu-CSP epitope, and accordingly, we found that pGlu-CSP mAbs do not bind the R21/MM vaccine (which shares the identical PfCSP sequence with RTS,S), which offers a potential advantage over other anti-PfCSP mAbs.

The pGlu-CSP site is also a candidate for a next-generation pre-erythrocytic vaccine. The identification of a Malian donor with circulating polyclonal antibodies against pGlu-CSP suggest that natural infection, as well as whole sporozoite vaccination, can elicit antibodies against pGlu-CSP. However, in both cohorts, donors with detectable antibodies against pGlu-CSP were uncommon. This observation could be due to the lower copy number of this epitope (maximum one pGlu-CSP compared to ~40 NANP repeats per molecule of PfCSP), with the immunodominant NANP repeats also potentially drawing antibody responses away from other regions of PfCSP (52). Therefore, a subunit pGlu-CSP vaccine that focuses on this epitope may be more appropriate to elicit a strong antibody response to this site. A similar approach has been taken with RH5, a blood-stage malaria antigen that is weakly immunogenic during natural infection but elicits a strong antibody response as a subunit vaccine (53, 54).

More broadly, this study highlights the benefits of an antigen-agnostic approach to study antibody responses to infectious pathogens, particularly unicellular pathogens with complex proteomes such as P. falciparum. A key advantage of this approach is the potential to identify uncharacterized antigens or epitope features that are not reproduced by recombinant expression, including post-translational modifications. The identification of a different protective mAb target within PfCSP, likely the most widely studied protein in malaria (certainly the most well-characterized target of malaria antibodies), highlights the benefits of this strategy to make discoveries even with well-studied proteins. Given the complexity of the P. falciparum proteome, it is striking that within the limits of detection of our assay, all observed plasma IgG reactivity to the sporozoite surface after infection or sporozoite vaccination could be attributed to PfCSP, when the pGlu-CSP epitope was considered. Whether this is also true for other antibody isotypes remains to be determined. Beyond malaria, the antigen-agnostic strategy deployed in this study could be expanded to other unicellular pathogens with relatively uncharacterized surface antigen profiles such as Mycobacterium tuberculosis (55). The identification of new functional antibodies and their associated targets could be powerful tools to aid the development of interventions against infectious pathogens.

Limitations and future directions

While we have shown that MAD21-101 can provide sterile protection against Pf sporozoite infection and provided evidence for activity at the hepatocyte stage, a deeper functional characterization of these mAbs with a wider range of assays remains to be performed. For instance, the in vitro motility assay results should be confirmed with an in vivo assay evaluating sporozoite activity in the skin to determine the relative contributions of the mAbs in the skin versus the liver or circulation (56, 57). We also plan to perform comparative in vivo protection experiments with other clinical candidate mAbs at a range of doses to determine their limits of protection. With the knowledge gained from this study, we are designing more sensitive tools to probe the B cell repertoires of naturally exposed or vaccinated individuals to further investigate the antibody response to this cryptic site.

Materials and methods

Study cohort

Plasma samples and peripheral blood mononuclear cells (PBMCs) were obtained from a cohort of 843 naturally exposed individuals from Kalifabougou, Mali (27, 28) and a cohort of 98 individuals from the United States who received the Sanaria PfSPZ vaccine of irradiated P. falciparum sporozoites (29, 30). Samples used in this study were already stored from the previously established cohorts. All participants provided informed consent for their blood products to be used for research purposes. No randomization was applied to the analysis of participants’ plasma or PBMC samples, but all samples were anonymized before being used in this study. Ethical approval for the Kalifabougou cohort study was obtained from the Ethics Committee of the Faculty of Medicine, Pharmacy and Dentistry at the University of Sciences, Technique and Technology of Bamako, and the Intramural Institutional Review Board (IRB) of the National Institute of Allergy and Infectious Diseases, National Institutes of Health (NIH IRB protocol number: 11IN126; https://clinicaltrials.gov/; trial number NCT01322581). Ethical approval for the VRC312 (https://clinicaltrials.gov; NCT01441167) and VRC314 (https://clinicaltrials.gov; NCT02015091) cohort studies was obtained from the NIAID IRB.

Preparation of antigen coated beads

Multiplexed panels of antigen coated beads were prepared using streptavidin beads pre-labelled with individual intensities of PE-channel fluorophore (Spherotech, SVFA-2558-6K and SVFB-2558-6K) or FITC-channel fluorophore (Spherotech, SVFA-2552-6K and SVFA-2552-6K). Beads were conjugated by incubation with 10 μg/mL each of the following biotinylated antigens for 1 h at room temperature: recombinant PfCSP (rPfCSP, Genscript), (NANP)₉, C-terminus or N-terminus or CD4. Additionally, N-terminally biotinylated peptides spanning the entire KLK(pGlu/Q/A)PADGNPDPNANPNVDP sequence were commercially synthesized (Genscript) and reconstituted in DMSO (Corning, 25-950-CQC). Peptides were diluted to 0.1 μg/mL in 0.5% BSA in PBS and directly conjugated to the PE or FITC-labelled beads by incubation for 1 h at room temperature. Protein or peptide-conjugated beads were pelleted at 2500 rpm for 3 min and washed once with 130 μL of 0.5% BSA. Excess streptavidin sites on the beads were blocked by incubation with 10 μg/mL of CD4 for 1h, followed by two washes with 130 μL 0.5% BSA per wash. Antigen-coated beads were combined to generate multiplexed panels as required.

Generation of P. falciparum salivary gland sporozoites

Standard laboratory procedures were employed to maintain the asexual cultures of Pf NF54 strain. Briefly, the parasites were cultured in vitro in type O+ erythrocytes by daily monitoring of parasitemia and changes of RPMI 1640 supplemented with 50μM hypoxanthine, aerated in a gas mixture of 5% CO₂, 5% O₂ and 90% N₂. Gametocyte cultures were set up from the healthy asexual cultures at 1% parasitemia and 5% hematocrit. The cultures were maintained for 14 days with daily media changes to initiate gametocyte development. Upon evaluation of maturation of stage 5 gametocytes on day 14, female Anopheles stephensei mosquitoes (4–7 days old) were fed with cultures at 0.3% gametocytemia and 50% hematocrit using membrane feeding apparatus (59). The fed mosquitoes were maintained for 17 days at 27°C and 75% humidity and were provided with a 10% w/v dextrose solution and 0.05% w/v p-aminobenzoic acid (PABA) in water. The oocysts inside the midguts of the mosquitoes were checked on day 7 post feeding. Supplemental feed using 50:50 mixture of blood and serum was carried out on day 7. The mosquitoes were dissected on day 17 to collect the salivary glands and purify the sporozoites. The salivary glands from the mosquitoes were collected on day 17 and homogenized in a glass tube to release the sporozoites.

Generation of P. berghei salivary gland sporozoites

Inbred 10–15-week-old female BALB/c mice were purchased from the Charles River Laboratories (Wilmington, MA, Strain code: 028). Transgenic sporozoites from P. berghei expressing GFP-luciferase and chimeric sporozoites from P. berghei expressing different CSP versions were generated as previously described (31, 60). Anesthetized P. berghei infected BALB/c mice, at 3–4% parasitemia, were used to feed 4 – 7 day old female Anopheles stephensi mosquitoes. The infected mosquitoes were kept at 19 – 21 °C and 80% humidity and fed a 10% sucrose solution. The infection of the midgut was assessed by light microscopy 10 – 11 days post-infection using mercurochrome (0.05% w/v) in water to visualize the oocysts. Salivary glands were dissected 19 – 26 days post-infection after killing the mosquitoes by freezing for 5 min at −20°C and dipping them in 70% ethanol for 1 – 2 min. Salivary glands were collected in PBS for cryopreservation or in media with 2% FBS for mouse challenges. Salivary glands were homogenized by passaging them in a tuberculin 1cc syringe 15 times and then passed through a 40 μm Mini-Strainer (Pluriselect-USA). Sporozoites were counted in a hemocytometer.

Purification of sporozoites

Freshly dissected salivary gland sporozoites were loaded onto a 3 mL cushion of 17% w/v solution of Accudenz (Accurate Chemical, AN7050) dissolved in distilled, deionized water. Sporozoites were separated from mosquito debris by density gradient centrifugation for 25 min at 2500 ×g and room temperature (acceleration = 1, no brake) as described before (61). Purified sporozoites were harvested from the interphase layer and pelleted at 16,000 ×g for 5 – 10 min at room temperature. For cryopreservation, purified sporozoites were washed with 1 mL of 0.5% w/v bovine serum albumin (Sigma, A7030) and re-suspended in a 75% v/v solution of Cryostor (C3124) in RPMI 1640 media (Gibco, 42401-018) to a final density of 2.5 × 10⁶ per mL. 100 μL aliquots of sporozoites were transferred to freezing containers and held at −80°C overnight to achieve freezing at a rate of −1°C per minute. The cryopreserved sporozoites were transferred to liquid nitrogen for long term storage.

Plasma blocking and screening against Pf sporozoites and recombinant CSP

Dilutions of vaccinee plasma (1-in-1000) and naturally exposed plasma (1-in-333) were prepared in 0.5% BSA in PBS. For the blocked condition, plasma was mixed and incubated with 300 μg/mL of recombinant Pf circumsporozoite protein for 2 h at room temperature with constant shaking at 300 rpm. For the unblocked condition, plasma was instead incubated with an equivalent volume of 0.5% BSA under the same conditions. 10 μL of blocked or unblocked plasma was mixed at a 1:1 volume ratio with either rPfCSP-coated beads or 1500 – 3000 sporozoites per well in 96-well, V-bottom plates. Plasma samples were incubated with sporozoites or beads for 30 min followed by two washes with 130 μL of 0.5% BSA. 30 μL of 2.5 μg/mL Alexa Fluor 647-conjugated, goat anti-human IgG (Jackson ImmunoResearch, 109-606-170, RRID:AB_2337902) was added to each well and the samples were incubated for 30 min to detect mAb binding, followed by two washes with 130 μL of 0.5% BSA. After the final wash, samples were resuspended in 0.5% BSA and acquired on the iQue flow cytometer (Sartorius). For all incubation steps, beads were maintained at room temperature and sporozoites were maintained at 4°C. For all wash steps, beads were pelleted at 1500 ×g for 3 min and sporozoites were pelleted at 3000 ×g for 5 min. Donors were down-selected for validation by repeat screening if reactivity of ≥ 8% was detected for their rPfCSP-blocked plasma during the initial screen. Based on the validation screening results, 5 donors with confirmed rPfCSP-blocked plasma reactivity greater than 25% were down-selected for mAb isolation studies.

Memory B cell isolation from PBMCs

Cryopreserved PBMCs were thawed in media supplemented with benzonase (Novagen, 71205-3) and stained with the following panel: Aqua Live/Dead (Invitrogen, L34957), CD14-BV510 (BioLegend, 301842, RRID:AB_2561946), CD3-BV510 (BioLegend, 317332, RRID:AB_2561943), CD56-BV510 (BioLegend, 318340, RRID:AB_2561944), CD19-ECD (Beckman Coulter, IM2708U, RRID:AB_130854), CD21-BV711 (BD, 563163, RRID:AB_2738040), IgA-Alexa Fluor 647 (Jackson ImmunoResearch, 109-606-011, RRID:AB_2337895), IgD-PE-Cy7 (BD, 561314, RRID:AB_10642457), and IgM-PerCP-Cy5.5 (BD, 561285, RRID:AB_10611998), CD27-Alexa Fluor 488 (BioLegend, 393204, RRID:AB_2750089) and CD38-APC-Cy7 (BioLegend, 303534, RRID:AB_2561605). The cells were sorted using the BD FACS Aria II and gated on live CD19⁺CD14⁻CD3⁻CD56⁻IgM⁻IgD^-IgA⁻ (IgG⁺ memory B cells).

Oligoclonal culture supernatant screen

25–100 IgG⁺ MBCs were co-cultured with 3000 irradiated 3T3-CD40L feeder cells (62, 63) per well of a 384-well plate in IMDM (Gibco, 31980-030) supplemented with 10% HI-FBS (Gibco, 10438-026), 100 ng/mL IL21 (Gibco, PHC0211), 0.5 μg/mL R848 (Invivogen, tlrl-r848) and 1X Mycozap (Lonza, VZA-2021). Cultures were maintained at 37°C and 5% CO₂ to allow for expansion of oligoclonal MBC cultures and to stimulate antibody secretion. On day 9 of culture, 10 μL of the culture supernatant was mixed with an equal volume of rPfCSP-coated beads and 1500 – 3000 Pf sporozoites per well in 384-well, V-bottom plates and incubated for 30 min followed by two washes with 60 μL of 0.5% BSA in PBS per wash. Samples were then incubated with 20 μL of 2.5 μg/mL Alexa Fluor 647-conjugated, goat anti-human IgG (Jackson ImmunoResearch, 109-606-170, RRID:AB_2337902) for 30 min followed by two 60 μL washes of 0.5% BSA per wash. After the final wash, samples were resuspended in 0.5% BSA and acquired on the iQue flow cytometer. For all incubation steps, beads were maintained at room temperature and sporozoites were maintained at 4°C. For all wash steps, beads were pelleted at 1500 ×g for 3 min and sporozoites were pelleted at 3000 ×g for 5 min. Wells with IgG reactivity towards Pf sporozoites and no accompanying reactivity towards rPfCSP (SPZ⁺/rPfCSP⁻ reactivity) were down-selected for monoclonal B cell isolation.

Isolation of SPZ⁺/rPfCSP⁻ specific B cells (Beacon assay)

On day 10 of culture, MBCs from wells with PfSPZ⁺/rPfCSP⁻ supernatant reactivity were pooled and washed in MACS buffer (PBS supplemented with 0.5% w/v BSA in PBS and 2 mM EDTA). Approximately 23,000 cells were loaded onto an OptoSelect 11k chip (Berkeley Lights) and OEP light cages were applied to sort single B cells into nanoliter-volume pens (nanopens) on the chip. Isolation of PfSPZ⁺/rPfCSP⁻ specific B cells was carried out in a two-step assay. In the first step of the assay, freshly dissected and Accudenz-purified Pf sporozoites were stained with 1-in-10,000 SYBR Green for 30 min on ice and washed three times post-staining with 1 mL of 0.5% BSA per wash. Sporozoites were mixed with 3.0 – 3.9 μm silica beads (Spherotech, SIP-30-10), re-suspended in 2.5 μg/mL goat anti-human IgG-Alexa Fluor 647 (Jackson ImmunoResearch, 109-606-170, RRID:AB_2337902) and immobilized in the channels of the OptoSelect 11k chip (Berkeley Lights). Binding of secreted antibody to the surface of whole Pf sporozoites was detected in the CY5 channel by capturing images at 6 min intervals over a 30 min time course. In the second step of the assay, the sporozoites were replaced with 6 – 8 μm streptavidin beads (Spherotech, SVP5-60-5) coated with 10 μg/mL rPfCSP and antibody binding was detected as before. Individual B cells secreting PfSPZ⁺/rPfCSP⁻ mAbs were exported directly into Dynabeads mRNA DIRECT lysis buffer (Life Technologies, 61011) in 96-well plates. Plates were sealed with Microseal foil film (BioRad, MSF1001) and immediately frozen on dry ice before transferring to −80°C for long-term storage.

Immunoglobulin sequence analysis and recombinant mAb expression

Heavy and light chain sequences were amplified from exported B cells by cDNA synthesis and RT-PCR as previously described (16, 64, 65). V_H and Vλ/Vκ sequences, gene usage, nucleotide and amino-acid mutations and CDR3 sequences were analyzed using Geneious Prime (Version 2023.0.4, http://www.geneious.com/) and the International Immunogenetics Information System database (IMGT, http://www.imgt.org/, analysis performed in 2021) (66). Unique pairs of V_H and Vλ/Vκ sequences were commercially cloned into IgG1 or light chain expression vectors and expressed as recombinant mAbs (Genscript). mAbs were also expressed in-house by transient transfection of Expi293 cells (Gibco, A14527) using Polyethylenimine Max transfection reagent (Polysciences Inc, 24765). During antibody production, an extra codon was inadvertently added to the MAD22-38 light chain, resulting in an extra residue in an external loop in framework 3 (F61, see fig. S7) at a location distant from the paratope (which is thus unlikely to affect binding or the structure of the antibody). This version of MAD22-38 was used in all experiments and its sequence has been uploaded to GenBank (accession number PQ382881). The original version of the MAD22-38 light chain does not include the ‘ttc’ codon at position 181 in the GenBank sequence. Table S1 shows percent mutation analysis on the original sequence. 6 days post-transfection, the expression culture supernatants were clarified for 10 min at 3500 ×g followed by filtration through a 0.22 μm filter (Millipore, SLGP033RS). Recombinant IgG mAbs were purified using HiTrap Protein A columns (Cytiva/GE Healthcare Life Sciences, 17040303) or Protein A Sepharose Fast Flow resin columns (Cytiva/GE Healthcare Life Sciences, 17127903).

Monoclonal antibody binding to antigen-labelled beads by flow cytometry

mAb titrations were prepared in 0.5% BSA in PBS. 10 μL of mAb titrations were mixed at a 1:1 volume ratio with multiplexed antigen-coated beads per well in 96-well V-bottom plates. mAb samples were incubated with the beads for 30 min at room temperature followed by two washes with 130 μL of 0.5% BSA. 30 μL of 2.5 μg/mL Alexa Fluor 647-conjugated, goat anti-human IgG (Jackson ImmunoResearch, 109-606-170, RRID:AB_2337902) was added to each well and the samples were incubated for 30 min at room temperature to detect mAb binding, followed by two washes with 130 μL of 0.5% BSA. For all wash steps, beads were pelleted at 1500 ×g for 3 min. After the final wash, samples were resuspended in 20 μL 0.5% BSA and acquired on the iQue flow cytometer (Sartorius).

Monoclonal antibody binding to sporozoites by flow cytometry

Sporozoites were pre-stained with 1-in-10,000 SYBR Green for 30 min on ice and washed three times post-staining with 1 mL of 0.5% BSA in PBS per wash. mAb titrations were prepared in 0.5% BSA in PBS. 10 μL of mAb titrations were mixed at a 1:1 volume ratio with 1500 – 3000 sporozoites per well in 96-well V-bottom plates, and incubated for 30 min at 4°C followed by three washes with 130 μL of 0.5% BSA. 30 μL of 2.5 μg/mL Alexa Fluor 647-conjugated, goat anti-human IgG (Jackson ImmunoResearch, 109-606-170, RRID:AB_2337902) was added to each well and the samples were incubated for 30 min at 4°C to detect mAb binding, followed by three washes with 130 μL of 0.5% BSA. After the final wash, samples were resuspended in 10 μL 0.5% BSA and acquired on the iQue flow cytometer (Sartorius). For all wash steps, sporozoites were pelleted at 3000 ×g for 5 min.

For the co-staining experiment, 20 μL of 100 μg/mL 5D5-Alexa Fluor 647 was mixed at a 1:1 volume ratio with 50,000 SYBR Green-stained sporozoites in 96-well V-bottom plates (final 5D5 concentration 50 μg/mL) and pre-incubated for 15 min at 4°C. After the pre-incubation step, 20 μL of 60 μg/mL MAD21-101 was added to the wells without washing (final MAD21-101 concentration 20 μg/mL) and an additional incubation step of 30 min at 4°C was carried out. For single staining conditions, 20 μL of individual fluorophore-labelled test mAbs (100 μg/mL 5D5-Alexa Fluor 647 or 40 μg/mL MAD21-101-Dylight 405 – for final 5D5 concentration 50 μg/mL, or final MAD21-101 concentration 20 μg/mL) or isotype control mAbs (40 μg/mL CV664-Alexa Fluor 647; 40 μg/mL CV664-Dylight 405) were mixed at a 1:1 volume ratio with 50,000 sporozoites in 96-well V-bottom plates and incubated for 30 min at 4°C. After incubation with mAbs, sample wells were washed three times with 130 μL of 0.5% BSA per wash. For all wash steps, sporozoites were pelleted at 3000 xg for 5 min. Samples were then resuspended in 50 μL 0.5% BSA and acquired on the iQue flow cytometer (Sartorius).

PfCSP peptide binding by ELISA

Half-area plates (Corning, 3690) were coated overnight with 100 μL per well of 1 μg/mL PfCSP peptides or 5 μg/mL rPfCSP in PBS. For rPfCSP, CSP 4/38 was produced in Lactococcus lactis (67), CSP 1 and CSP 2 were produced in Escherichia coli, and CSP 3 and CSP 4 were produced in Pichia pastoris (68) as previously described. Plates were blocked with 100 μL per well of 1% BSA (w/v in PBS) for 1 h at room temperature followed by incubation for 1 h at room temperature with 25 μL per well of 10 μg/mL MAD21-101, MAD22-38, anti-PfCSP mAbs (CIS43, L9, 5D5) and an isotype control mAb (VRC01). For detection of antibody binding, plates were incubated for 1 h at room temperature with 25 μL per well of 1-in-500 alkaline phosphatase (AP)-conjugated goat anti-human IgG (Southern Biotech, 2040-04, RRID:AB_2795643). Plates were washed four times with 0.05% Tween-20 (v/v in PBS) (Sigma, P7949) between each step. After the final wash, 50 μL of p-nitrophenyl phosphate (p-NPP) substrate (Sigma, N2765) was added to each well and the plates were developed for 30 min at room temperature, after which absorbance values were measured at 405 nm.

Proteome microarray analysis

Full P. falciparum 3D7 proteome microarrays were generated as previously described (69). The protein microarray used in this study was produced by Antigen Discovery, Inc (ADI) and encompassed 8871 full-length or fragmented Pf proteins representing 5234 protein-coding genes and covering ~99% of the proteome. Each open reading frame (ORF) sequence was amplified by PCR and inserted into the vector pXT7 by recombination in E. coli to establish a library of partial or complete coding DNA sequences. Proteins were expressed using a coupled E. coli cell-free in vitro transcription and translation (IVTT) system (Rapid Translation System, Biotechrabbit, Berlin, Germany, BR1400201) and spotted onto nitrocellulose-coated glass AVID slides (Grace Bio-Labs Inc., Bend, OR) using an Omni Grid Accent robotic microarray printer (Digilabs Inc. Hopkinton, MA). Each expressed protein included a 5’ poly-histidine (His) epitope and 3’ hemagglutinin (HA) epitope. Proteome microarray chip printing and protein expression were quality checked by probing random slides with anti-His and anti-HA monoclonal antibodies fluorescently labeled and quantifying spot signals using a microarray scanner. The mAbs were diluted to 0.05 μg/mL, 0.5 μg/mL and 5 μg/mL and preincubated with 20% DH5α E. coli lysate for 30 min and incubated on the proteome microarrays overnight at 4°C on a rocker. Bound IgG was detected with DyLight650 anti-human IgG (Bethyl Laboratories, Montgomery, TX, A80-104D5, RRID:AB_10634506). Washed and dried microarray chips were scanned, and the spot and background signal intensities (SI) were exported into R package for analysis. A separate mini protein microarray containing the CSP and select other P. falciparum proteins was fabricated with varying cell-free IVTT expression systems: the standard E. coli IVTT, cell-free insect expression from two vendors with and without additional protease inhibitor (TnT^® T7 Insect Cell Extract Protein Expression System, Promega, Madison, WI, USA, L1101; RTS 100 Insect Membrane Kit, Biotechrabbit, BR1401502) and cell-free wheat germ expression with and without additional protease inhibitor (RTS 100 Wheat Germ Kit, Biotechrabbit, BR1402501). This chip was probed with mAbs to determine if superior binding could be achieved using eukaryotic cell-free expression systems that allow for post-translational modifications. Data were normalized by dividing the foreground spot signal intensity (SI) by the negative control spot (IVTT without ORF) SI and transforming values by the base 2 logarithm. Thus, normalized data represent the log2 signal-to-noise ratio, where a value of 0 represents specific antibody SI equal to the background, 1.0 represents twice the background, 2.0 represents 4-fold over background, and so forth. Antigens with antibody binding at least twice the background, or normalized SI of 1.0, were considered specific mAb binding hits.

Sporozoite Western blot

Freshly dissected, Accudenz-purified Pf sporozoites were pelleted at 8000 xg and resuspended to a density of 5 × 10⁶ per mL in a low-protein binding Eppendorf tube. For non-reducing SDS-PAGE, sporozoites were resuspended in 4X NuPAGE LDS sample buffer only (Invitrogen NP0007) and for reducing SDS-PAGE the sample buffer was supplemented with 1% v/v beta-mercaptoethanol (Sigma, M-7154). Samples were denatured by heating at 95°C for 5 min with constant shaking at 300 rpm. For crude blots, proteins were resolved by electrophoresis alongside Precision Plus Protein Dual Color Standards (Bio-Rad, 1610374) on a 4–12% Bis-Tris NuPAGE gradient polyacrylamide gel (Invitrogen, NP0321BOX and NP0323BOX) at 120 V for 2 h. For separating cleaved and un-cleaved CSP, proteins were resolved on a 12% Bis-Tris NuPAGE polyacrylamide gel (Invitrogen, NP0349BOX and NP0343BOX) on ice at 150 V for 5.5 h. Resolved proteins were transferred to a nitrocellulose membrane (Invitrogen, LC2001) on ice at 30 V for 1 h followed by a 10 min wash with 0.2% v/v Tween-20 in PBS, and membranes were then blocked overnight in SuperBlock Blocking Buffer (ThermoFisher Scientific, 37515). After blocking, membranes were first incubated with primary antibody at room temperature for 1 h then incubated with 1-in-2000 HRP-conjugated anti-human IgG secondary (Sigma, NA933V, RRID:AB_772208) diluted in blocking buffer for 1 h. Membranes were washed three times with 0.2% v/v Tween-20 in PBS after each antibody incubation. After the final wash, membranes were developed by incubating with SuperSignal West Pico PLUS Chemiluminescent Substrate (ThermoFisher Scientific, 34580) for 5 min at room temperature and imaged on the ChemiDoc system (BioRad).

Indirect immunofluorescence assay of whole sporozoites

The wells of an 8-well chamber slide (Cellvis, C8–1.5H-N) were incubated with 50 μg/mL Poly-D-Lysine (Gibco, A3890401) for 1 h at room temperature then washed three times with 300 μL PBS per wash. After the final wash, the coated chamber slide was left uncovered in the laminar hood to air dry for 2 h, then stored at 4°C. Freshly dissected P. falciparum sporozoites were clarified by Accudenz density gradient centrifugation and resuspended in 0.5% BSA in PBS. 5 × 10⁴ sporozoites were seeded into each well of the coated chamber slide and allowed to air dry for 4 h in the laminar flow hood. Sporozoites were fixed in 4.0% v/v paraformaldehyde (Electron Microscopy Services, 15710) in PBS for 10 min at room temperature, washed three times with PBS, then blocked by incubating in 4% BSA for 30 min at 4°C. The blocking solution was replaced with Alexa Fluor 647-conjugated antibodies (MAD21-101, MAD22-38, CIS43, and CV503) diluted to 10 μg/mL in 2% BSA, and samples were stained for 1h at 4°C. After staining, the samples were washed three times with PBS followed by counterstaining with 10 μg/mL 4′,6-diamidino-2-phenylindole (DAPI, Invitrogen, MP01306) for 5 min at 4°C. The fixed parasites were imaged within 24 h using a Zeiss LSM 880 with Airyscan equipped with a Zeiss plan apochromat 63x / 1.4 NA oil immersion objective. Airyscan processing was performed with Zeiss Zen software and images were colorized using ImageJ (Version 1.54g, National Institutes of Health, USA).

Intravenous challenge with Pb-PfCSP sporozoites

Female 6–7-week-old B6(Cg)-Tyr^c−2J/J (B6 albino) mice were purchased from The Jackson Laboratory (Bar Harbor, ME), and kept in the animal care facility at the National Institute of Allergy and Infectious Diseases (NIAID). Animal experiments were approved by the NIAID Animal Care and Use Committee, under protocols Animal Study Protocols VRC-17-702 and LIG 21. Liver burden experiments were performed as previously described (31). Briefly, Anopheles stephensi mosquitoes were infected with a transgenic P. berghei strain expressing full-length P. falciparum CSP and a dual reporter GFP-luciferase (Pb^{PfCSP-GFP/Luc}) construct. Mosquito salivary glands were dissected for sporozoite isolation in a 2% FBS solution (v/v in 1X HBSS) or Leibovitz’s L-15 medium (Sigma-Aldrich). Mice were injected intravenously via the tail vein with 300 μg of antibody resuspended in 200 μL of sterile 1X PBS, and subsequently infected through the same route with 2000 Pb^{PfCSP-GFP/Luc} sporozoites resuspended in 200 μL of 2% FBS solution (v/v in 1X HBSS) or Leibovitz’s L-15 medium. At 40 – 42 h post-infection, mice were anesthetized and injected with either 200 μL of 15 mg/mL or 150 μL of 30 mg/mL D-luciferin intraperitoneally. Parasite luminescence was measured in the IVIS Spectrum in vivo imaging system (Revvity, Inc.) using ten 15-second exposures with a 1-minute interval between them. The total flux was expressed in photons/second.

Sporozoite ELISA

P. berghei ANKA sporozoites expressing regular, full-length PbCSP (Pb^WT), and transgenic Pb sporozoites expressing full-length PfCSP (Pb^PfCSP), PfCSP without the junction region (Pb^PfCSP-JRKO), or PbCSP including the PfCSP N-Term (Pb^{Pf N-term}), C-Term (Pb^{Pf C-term}), junction plus NANP₄ (Pb^PfJNANP4) or NANP₁₂ (Pb^PfNANP12) epitopes, were utilized for ELISA experiments. Sporozoites were dissected from salivary glands of Anopheles stephensi infected mosquitoes in HBSS media. 5000 sporozoites were seeded per well of ELISA plates (Immuno Plates MaxiSorp, Thermo Scientific Nunc, 442404) and frozen at −80°C, followed by 3 freeze and thaw cycles. After this step, plates were defrosted, washed three times with PBS, and blocked with PBS supplemented with 1% BSA. After blocking, plates were washed again with PBS. Serial dilutions of the anti-CSP mAbs or control mAbs were tested against sporozoites immobilized on the ELISA plates. Following 1 h incubation at room temperature with varying concentrations (0.008–1.00 μg/mL) of CSP mAbs or control mAbs, plates were washed and incubated with 100 μL per well of 250 ng/mL peroxidase-labeled goat anti-human IgG antibody (Jackson ImmunoResearch, 109-035-088, RRID:AB_2337584). After washing, samples were incubated with ABTS Peroxidase substrate (KPL, Gaithersburg, MD) for 15 min and read in a plate reader at 405 nm.

Trail gliding assay

Mosquito infection with P. falciparum NF54 was performed as previously described (59). Infected mosquitoes were maintained for 15 days at 25°C and 80% humidity and were provided with 10% sucrose solution. Freshly dissected P. falciparum salivary gland NF54 sporozoites in HBSS/2% BSA (Sigma, A7888) pH 7.4 were mixed with the indicated antibody in HBSS. Sporozoites were pre-incubated for 30 minutes at 20°C and 15,000 sporozoites/well were added to a 96 well glass bottom plate (Greiner, 655892) pre-coated with 5 μg/mL of mAb 2A10, which is specific for the PfCSP repeat region (70). The plate was centrifuged for 5 minutes at 200 × g and incubated for 1 h at 37°C. Wells were fixed in 4% paraformaldehyde in PBS, blocked with 1% BSA in PBS (pH 7.4) and stained with biotinylated mAb 2A10 in 1% BSA in PBS (pH 7.4) for 1 h at room temperature, followed by detection with Alexa Fluor 488 streptavidin (Invitrogen) diluted at 1:500 in PBS for 1 h at room temperature. Samples were preserved in a 9:1 glycerol / PBS solution at 4°C and imaging was performed on 25 positions per well (5 × 5, 500 μm apart) by using ImageXpress Micro XLS Widefield high-content analysis system (Molecular Devices) with 40X Plan fluor objective. Acquired images were processed using Cell Profiler software (version 3.0.0) to measure area occupied by trail (70).

In vitro inhibition of liver stage development assay (ILSDA)

The inhibition of liver stage development assay was performed as previously described (71) with some modifications. The day before the in vitro sporozoite infection, Nunc^™ Lab-Tek^™ 8 well chamber slides (ThermoFisher Scientific, 177410) were coated by incubating at 37°C for 1 h with Bovine PureCol^® Type I Collagen Solution (Advanced Biomatrix, 5005) at 100 μg/mL v/v in sterile tissue culture grade water. Coated slides were rinsed once with PBS and air dried for 10 – 15 min at room temperature. Cryopreserved primary human hepatocytes (PHH, BioIVT, lot ZHL) were thawed using INVITROGRO^™ HT thawing medium (BioIVT, Z990005) supplemented with 1X Penicillin-Streptomycin-Glutamine (ThermoFisher Scientific, 10378016) and 10 μg/mL of Fungin^™ (Invivogen, ant-fn) per manufacturer’s instructions. 200,000 hepatocytes were seeded per chamber on the collagen coated slides and slides were gently, manually shaken to ensure homogeneous cell distribution. The cells were incubated at 5% CO₂ and 37°C for 24 h. On the day of in vitro infection, mosquitoes infected with Pf (NF54 strain) were killed in 70% isopropyl alcohol, immediately washed in 1x PBS and shipped on ice in RPMI medium. Upon receipt, mosquitoes were washed four times with INVITROGRO^™ HI medium supplemented with the above antimicrobials. INVITROGRO^™ HI medium was used throughout the rest of the assay for mosquito dissection, sample dilutions and media changes. Live sporozoites were isolated from mosquito salivary glands and sporozoites were incubated with each of the test and control mAbs at room temperature for 20 min. Untreated sporozoites incubated with media only were used as infectivity controls. After 20 min incubation, sporozoitem-Ab mixture was added to the one-day old PHH cultures (after removing the medium) at a final concentration of 25,000 sporozoite/well and 100 μg/mL of each mAb/well in duplicate. The culture slides were centrifuged at 300 ×g for 1 min and incubated at 5% CO₂ and 37°C for 3 h. After the incubation, medium was aspirated to remove unbound sporozoites and replaced with 300 μL of INVITROGRO^™ HI medium. Cultures were maintained 96 h with daily media changes. To prepare the qRT-PCR standards, cultured PHHs were incubated with 100 μL of 0.05% Trypsin-EDTA at 5% CO2 at 37°C for 7 min, PHH was harvested into low protein binding 1.5 mL tubes by gently scraping the wells and spiked with serially diluted sporozoites (3-fold dilutions for 6 dilutions starting from 4860 sporozoites/well). Subsequently, sporozoite-spiked hepatocytes were briefly vortexed to mix and centrifuged at 5000 rpm for 5 min. Supernatant was aspirated, and the cell pellet was frozen at −20°C for 4 days until automated RNA extraction. 96 h post-infection, total RNA was extracted from the cultured hepatocytes using the Rneasy 96 QIAcube HT kit (Qiagen) per manufacturer’s instructions. Pf 18S ribosomal RNA (rRNA) copy number per sample was determined by quantitative real-time PCR (qRT-PCR) using the standard curve method. Hepatocyte cultures were washed thrice with 1X PBS, then hepatocytes were directly lysed with Buffer RLT supplemented with beta-mercaptoethanol and the lysed cells were harvested by gently scraping the wells. The lysates were vortexed to ensure complete lysis of the cells, total RNA was extracted, and the RNA concentration was measured using QIAxpert high-speed microfluidic UV/VIS spectrophotometer (Qiagen). 100 ng of total RNA from each test condition and standards were transcribed to cDNA using High-Capacity cDNA Reverse Transcription Kit (ThermoFisher Scientific, 4368813) per manufacturer’s instructions and stored at 4°C overnight. The following day, qRT-PCR was performed in technical duplicates of each sample and in triplicates of each standard. The following reaction mixture (total volume of 20μL) was used to detect parasite 18S rRNA: 5 μL of cDNA template, TaqMan Fast Universal PCR Master Mix (1X, no AmpErase UNG kit), Custom TaqMan probe : 5′ – 6FAM - CAG GTC TGT GAT GTC C – MGBNFQ - 3′ (0.25 μM final concentration), Unlabeled Sequence detection primer 18s Forward, 5’ - TAA CAC AAG GAA GTT TAA GGC AAC A - 3’ (0.9 μM final concentration), Sequence detection Primer 18s Reverse, Sequence: 5’ - CGC GTG CAG CCT AGT TTA TCT - 3’ (0.9 μM final concentration) (ThermoFisher Scientific), in 7500 FAST Real-Time PCR System instrument (Applied Biosystems^™/ThermoFisher Scientific). Default thermal cycling conditions for TaqMan^™ Fast Universal PCR Master Mix (no AmpErase^™ UNG) (7500 or 7500 Fast system) was used for the qRT-PCR (Enzyme activation :95°C, 20 sec, 1 cycle and denature:95°C, 3 sec, Anneal/Extend:60°C, 30 sec, repeat 40 cycles). Pf 18S rRNA copy number for each sample was automatically determined by the 7500 Software v2.3 using the standard curve. Percentage inhibition of in vitro Pf liver-stage parasite development was calculated using the Pf 18S rRNA copy number of the test mAb compared to the untreated condition. Formula for percentage inhibition = [1 − (Pf 18S rRNA copy number of the test mAb / Pf 18S rRNA copy number of the untreated condition)] × 100.

Fab mutagenesis, expression and purification

For biophysical studies, overlapping primer mutagenesis was used to introduce a stop codon between the C_H1 and hinge region of the IgG sequence, producing WT Fab expression constructs. For X-ray crystallography, light-chain residues T¹⁹⁷HQGLSSPV were mutagenized to a two-residue shorter T¹⁹⁷-QGTTS-V by overlapping primer mutagenesis to promote crystal packing and facilitate crystallization (72). Fab expression for both biophysical and structural studies was performed in ExpiCHO mammalian cells (Gibco, A29127). Heavy and light chains of the Fab were prepared with the Takara Bio Inc. Nucleobond Maxi/Midi purification kits and transfected Expifectamine CHO reagent according to the manufacturer’s protocol. Cells were cultured at 37°C, shaking at 125 rpm, 8% CO₂, for 18–24 h and transferred to 32°C, 125 rpm, 8% CO₂ for a further 10 – 12 days. Transfection enhancer was added on day 1, and transfection feed medium was added on days 1 and 5. Cell supernatants were harvested by centrifugation at 4,000–6,000 ×g and filtered through a 0.22 μm or 0.45 μm filter. Clarified supernatants were loaded onto a HiTrap^™ protein G 5 mL column (Cytiva), washed with 5 column volumes PBS, eluted with 0.1M glycine solution (pH 2.8–3.0), and neutralized with 1M Tris solution (pH 8.0–9.0). Relevant elution fractions were further purified by size-exclusion chromatography on a HiLoad^™ Superdex200 or Superdex75 16/600 column equilibrated with TBS. For ITC, Fab MAD24-01 was recombinantly expressed (Genscript) and purified by size-exclusion chromatography on a HiLoad^™ Superdex75 16/600 column equilibrated with PBS.

Crystallization and structure determination

For crystallization, Fab purified in TBS was prepared at 10 mg/mL with a 10-fold molar excess of synthetic peptide pGluPADGNPDPNANPNVDPN-NH₂ (Innopep). Crystallization screening was conducted with the high-throughput robotic CrystalMation system (Rigaku) at TSRI by sitting drop vapor diffusion, with many conditions producing crystals for all Fabs. For the datasets shown, MAD21-101 crystals were grown in 2.0 M ammonium sulfate, 0.2 M sodium chloride, 0.1 M cacodylate pH 6.5, MAD22-38 crystals were grown in 20% PEG 3350, 0.2 M ammonium chloride, and MAD24-01 crystals were grown in 20% PEG 3350 and 0.2 M potassium fluoride. Crystals were cryoprotected in mother liquor supplemented with 30% ethylene glycol and stored in liquid nitrogen until data collection. Datasets were collected at Brookhaven BNL FMX beamline at a wavelength of 0.97934 Å by rotating crystals for a full 360°, with diffraction data collected every 0.2°, producing 1800 frames per dataset. Diffraction data were processed using HKL2000 (73) and structures were determined through molecular replacement in Phaser (74) using a variable domain search model produced by AbPred (75) and constant domain search model for a kappa light chain Fab. Structures were refined in REFMAC5 (76) and PHENIX (77), and further built and refined in Coot (78). Data collection and refinement statistics are tabulated in table S4.

Heat-induced fragmentation of CSP and acetonitrile extraction peptides.

For proteomic studies, Accudenz-purified sporozoites were collected from the gray band at the interface and transferred to a 1.5 mL microcentrifuge tube, pelleted at 16,000 ×g for 10 min, washed three times in PBS to remove the residual Accudenz, and stored at −80°C. The three samples analyzed consisted of 9.4, 4.0, and 4.2 ×10⁶ sporozoites. Frozen sporozoite pellets were resuspended in 200 mM ammonium bicarbonate (ABC) and water to achieve a final volume of 20 μL of 100 mM ABC. The synthetic peptide QPADGNPDPNANP (resuspended at 5 mg/mL in DMSO and stored at −20°C) was diluted 100-fold in water, then this solution was diluted 100-fold (final concentration of 500 pg/μL) in 0.1% TFA for analysis of untreated peptide, or in 100 mM ABC prior to heat treatment. To 19 μL of recombinant PfCSP (0.44 mg/mL in PBS) was added 1 μL 2M ABC. To induce protein fragmentation, 20 μL of sample was incubated 10 min at 95°C in a thermomixer at 1200 rpm. The samples were briefly cooled on ice, after which 20 μL of acetonitrile (ACN) was added. The samples were vortexed for 5 min at room temperature at 1200 rpm, then centrifuged for 5 min at 20,000 × g to pellet the precipitate. The supernatant was recovered, transferred to a 1.5 mL microcentrifuge tube, and dried in a vacuum centrifuge at 45°C. The extracted peptides were resuspended in 20 μL 0.5% TFA prior to analysis by LC-MS. Injection volumes for LC-MS analysis were 15 – 18 μL for sporozoites, 1 μL (nominally 10 pmol) for PfCSP, and 1 μL (nominally 500 pg) of peptide. Unless otherwise noted, all solid reagents were from Sigma; solvents and acids were Optima LC-MS grade from Fisher Scientific. Water and LC mobile phases were LC-MS grade from Honeywell Burdick & Jackson. Microcentrifuge tubes were Eppendorf Protein LoBind.

Immunoprecipitation of synthetic peptides

The synthetic peptides QPADGNPDPNANP and pGluPADGNPDPNANP (resuspended at 5 mg/mL in DMSO and stored at −20°C) were diluted 100-fold in water, then 13.1 μL of the Gln peptide and 12.9 μL of the pGlu peptide were combined and diluted to 1000 μL in PBS to achieve a final concentration of 500 fmol/μL. Four μg of Protein G-functionalized magnetic beads (GenScript, L00274) were washed four times in PBS, then 20 μL of MAD1-101 mAb or CV503 (isotype control) at 5 mg/mL in PBS was added. The beads were incubated for 1 h at room temperature with axial rotation then kept on ice until use, then washed three times with PBS and resuspended at 0.25 mg/mL. For each reaction, 1 μg of beads was added to a 1.5 mL microcentrifuge tube and washed once with PBS, then 10 μL (5 pmol) of peptide solution was added and incubated for 1 h at room temperature with axial rotation. The depleted supernatant was removed and diluted to 50 μL in 0.5% TFA. The beads were washed three times with PBS, then peptides were eluted by adding 50 μL of 0.2% TFA and incubating 5 min at room temperature with axial rotation. The supernatant was recovered and transferred to a new tube. Both the depleted supernatant and eluate samples were centrifuged 2 min at 20,000 ×g to settle any residual magnetic beads. Undepleted sample was prepared by diluting the peptide mixture five-fold in 0.5% TFA. For the undepleted and depleted samples, 1 μL (nominally 100 fmol) of sample was analyzed by LC-MS. The eluted sample was desalted with C18 tips (Pierce, 88513) to deplete the mAb that co-eluted from the Protein G beads. Ten μL of eluate was loaded onto a tip and eluted with 5 μL of 40% ACN, to which was added 45 μL of 0.1% TFA. Ten μL of each eluate (nominally 400 fmol) was analyzed by LC-MS.

Mass spectrometry data generation and analysis

All mass spectrometry data, including extended methods, raw data, results, and search parameters, have been deposited at the ProteomeXchange Consortium (79) via the PRIDE (80) partner repository with the dataset identifier PXD052186. Peptide spectrum matches (PSM) have been assigned Universal Spectrum Identifiers (USI) (81) and may be visualized using the tool Quetzal (https://proteomecentral.proteomexchange.org/quetzal/). LC was performed with an EASY-nLC 1000 (Thermo Fisher Scientific, USA) using a vented trap set-up. The trap column was a PepMap 100 C18 (Thermo Fisher Scientific, 164946) with 75 μm i.d. and a 2 cm bed of 3μm 100 Å C18. The analytical column was an EASY-Spray column (ThermoFisher Scientific, ES903) with a 75 μm i.d. and a 50 cm bed of 2μm 100 Å C18 beads, operated at 55°C. The LC mobile phases consisted of buffer A (0.1 % v/v formic acid in water) and buffer B (0.1 % v/v formic acid in ACN). The separation gradient, operated at 300 nL/min, was 4% B to 28% B over 15 min for synthetic peptides and over 60 min for heat-induced fragmentation samples. Data-dependent acquisition (DDA) was performed with a Thermo Fisher Scientific Orbitrap Eclipse. Raw mass spectrometry files were converted to mzML using msConvert version 3.0.19106 (82) (Proteowizard). Data were searched and analyzed on a Slurm 19.05.5 cluster running under Ubuntu 20.04. MS2 spectra were searched with Comet version 2023.01 rev. 2 (83) and analyzed with the Trans Proteomic Pipeline (TPP) version 7.0.0 (84). The sporozoite samples were searched against a protein FASTA database assembled from the reference P. falciparum 3D7 (85, 86) database (PlasmoDB version 67 (87, 88)) the reference Anopheles stephensi (89) database (Vectorbase version 67 (88, 90)), and the common Repository of Adventitious Proteins (www.thegpm.org/cRAP). The recombinant CSP data were searched against a database comprising the protein sequence and cRAP. De Bruijn decoy protein entries were generated using a tool in the TPP. Two decoys were created for each real entry, denoted DECOY0 and DECOY1. The synthetic peptides were searched against their respective sequences only. Pertinent search parameters included the following: no enzyme was specified (“cut everywhere”); no static modifications were specified (since Cys residues were not alkylated); and variable modifications of +15.994915 at Met (oxidation) and −17.026549 at Gln (conversion to pyro-Glu) were allowed. Because the biotinylated Gln⁹⁶ peptide contained two non-standard residues at the C-terminus (a 6-amino hexanoic acid (Ahx) linker and a biotinyl-lysine, i.e., biocytin), custom residues were specified in the Comet params file: 113.08406 for Ahx and 354.17256 for biocytin, denoted as X and B, respectively, in the FASTA database. PSMs for the sporozoite samples were analyzed with PeptideProphet using non-parametric models. The number of missed cleavages (NMC) and number of tryptic termini (NTT) models were disabled since the peptides were not produced with trypsin. DECOY0 entries were used to train the models and then assigned a probability of 0, while DECOY1 entries were assigned probabilities and used to estimate false discovery rates (FDR). PSMs with PeptideProphet probabilities corresponding to a decoy-estimated FDR < 1.0% were taken for further consideration. For the recombinant PfCSP samples, a 1% FDR was estimated using the Comet Expect score and the DECOY0 entries. The PSMs of the synthetic peptides were visually inspected with Quetzal to confirm their identities and the conversion of N-terminal Gln to pGlu.

Antibody sequence and structure analysis

Antibody V_H lineages were determined using IgBLAST (NCBI), germline sequences for V_H genes obtained from IMGT, and CDR boundaries determined using AbYsis. Buried surface areas were calculated with MS (91) using a 1.4 Å probe radius, and binding interactions were determined using PDBePISA server of EMBL-EBL. Structure figures were generated in Mac PyMol (Schrödinger) and graphical data were plotted using GraphPad Prism.

Isothermal titration calorimetry (ITC)

ITC experiments were performed on a MicroCal Auto-iTC200 (GE Healthcare). Synthetic lyophilized peptide pGluPADGNPDPNANP-NH2 (Innopep) was reconstituted in PBS to a stock solution of ~4 mM. To avoid buffer mismatch, Fab were extensively dialyzed against PBS, and dialysis buffer was used to dilute both Fab and peptide during sample preparation. Fabs were placed in the cell at a concentration of ~10 μM, with peptides placed in the syringe at a concentration of ~100 μM for MAD21-101 and ~150 μM for all other Fabs. ITC experiments were conducted in triplicate (N = 3) at 25°C and consisted of 1 × 0.5 μl injection and 31 × 1.2 μl injections at a rate of 0.5 μl/s, at 180 s intervals, and reference power of 5 μCal. Baseline measurements were conducted with peptide against a cell containing buffer only and subtracted from the integrated data. As is common practice, the first data point was excluded from analysis. Outliers resulting from aberrant injections were removed following baseline correction. Fitting of the integrated titration peaks was performed using a single-site binding model on the Origin 7.0 software and analysis of kinetic parameters was performed using GraphPad Prism.

Bioinformatic analysis of global PfCSP sequences

The variant call format file containing variant data for P. falciparum chromosome 3 was downloaded from the Pf7 database (42). Variants in the region of interest that also had a FILTER value of PASS were extracted using bcftools software (92). These high-confidence variants were imported into the R v4.2.2 statistical software (93) and processed with the package ComplexUpset v1.3.3 (94, 95) to generate the upset plot.

Flow cytometry analysis of mAb binding to mutated PfCSP peptides

C-terminally biotinylated peptides spanning the pGluPADGNPDPNANP sequence and bearing alanine or glycine (when original residue was alanine) substitutions at each residue position were commercially synthesized (Genscript). The pGluPVDGNPDPNANP peptide was also commercially synthesized (Genscript). Peptides were reconstituted in DMSO (Corning, 25-950-CQC), diluted to 0.1 μg/mL in 0.5% BSA in PBS and directly conjugated to the FITC-labelled beads as described above. Peptide-labelled beads were combined to generate multiplexed panels as required. mAbs were titrated and screened for binding to the peptide-labelled beads as described above.

MSD analysis of mAb binding to R21 vaccine

A Multi Array 96-well plate (Meso Scale Discovery, MSD) was coated with 1μg/ml R21 in 1x PBS for 2h at 37°C. The plate was then washed 5 times with 1X PBS with 0.05% tween 20. After washing, the plate was blocked for 1h with MSD Blocker A (MSD, R93AA). The mAbs were applied to the coated plate after diluting to a starting concentration 5μg/ml in Diluent 100 buffer (MSD, R50AA) and then serially diluted 5-fold with 8 dilutions and incubated for 1 h. Plates were washed and Sulfo-Tag Detection antibody anti-human IgG (MSD, R32AJ, RRID:AB_2905663) diluted to 0.5 μg/mL in diluent 100 was added to the plate and incubated for 1 h. The plate was washed to remove unbound detection antibody and read on MSD Sector Imager 600mm after the addition of MSD Read Buffer T (MSD, R92TC).

Mosquito bite challenge of FRG huHep model by Pf sporozoites

Mosquito bite challenge studies were carried out as described previously (96). Mosquitos infected with P. falciparum (strain NF54) were provided by the Johns Hopkins Malaria Research Institute Insectary Core. Mouse experiments were approved by the Oregon Health and Sciences University (OHSU) Institutional Animal Care and Use Committee (IACUC) (protocol IP00002077). Female NOD mice (≥5 months) were repopulated with human hepatocytes and humanization of the resulting mouse line (FRG huHep) was confirmed by measurement of human albumin levels in mouse serum (Yecuris Inc., Beaverton, OR). 24 h prior to the challenge, mice were injected intravenously with the indicated doses of test mAbs or the isotype control mAb 1245 (48). On the day of the challenge, each mouse was anesthetized under isoflurane for 10 min and caged with 5 infectious mosquitos to allow for infection by mosquito bite. Five days post-infection (p.i.), mice were injected intraperitoneally (i.p.) with human red blood cells (BloodWorks Northwest, Seattle, WA, USA). On days 7 and 9 p.i., peripheral blood was collected from each mouse, lysed and parasitemia was determined by Pf 18S rRNA qRT-PCR as previously described (97).

Depleting plasma reactivity using rPfCSP and pGlu-CSP peptides

Dilutions of vaccinee plasma (1-in-1000), naturally exposed plasma (1-in-333), MAD21-101 (0.1 μg/mL) and CIS43 (0.1 μg/mL) were prepared in 0.5% BSA in PBS. For the rPfCSP blocked condition, plasma and control mAbs were mixed with 300 μg/mL of rPfCSP only. For the pGlu-CSP blocked condition, plasma and control mAbs were mixed with 1 – 150 μg/mL pGluPADGNPDPNANP peptide only. For the double blocked condition, plasma and control mAbs were mixed with 300 μg/mL of rPfCSP supplemented with 1 – 150 μg/mL pGluPADGNPDPNANP. All conditions were incubated for 2 h at room temperature with constant shaking at 300 rpm. 20 μL of each plasma or control mAb condition was mixed with 3000 Pf sporozoites in a 96-well, V-bottom plate and incubated for 30 min at 4°C. Sporozoites were pelleted at 3000 ×g for 5 min and samples were washed twice with 130 μL of 0.5% BSA per wash. Samples were incubated with 30 μL of 2.5 μg/mL Alexa Fluor 647-conjugated goat anti-human IgG secondary for 30 min (Jackson ImmunoResearch, 109-606-170, RRID:AB_2337902), followed by two washes with 130 μL of 0.5% BSA. After the final wash, samples were resuspended in 0.5% BSA and acquired on the iQue flow cytometer.

Data analysis and statistics

Flow cytometry data were analyzed using FlowJo 10.9.0 (BD) and GraphPad Prism 10 (GraphPad Software, Massachusetts USA). One-way ANOVA with Dunnet’s multiple comparisons or the mixed-methods model with Šídák’s multiple comparisons were applied for statistical analyses of in vitro and in vivo functional assays. All statistical analyses were carried out using GraphPad Prism 10. N is stated in figure legends throughout for multiple experimental replicates, otherwise n = 1.

Supplementary Material

Supplementary Material

Figs. S1 to S10

Tables S1 to S4

References (59–97)

Data and materials availability:

All data associated with this manuscript are available in the main text or the supplementary materials. Data presented in the figures are deposited online at Dryad (58). Crystal structures have been deposited into the Protein Data Bank with PDB IDs (9C79 for MAD21-101, 9C7D for MAD22-38, 9C7F for MAD24-01). Antibody sequences are deposited in GenBank with accession numbers PQ382870–PQ382889. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD052186. Materials described in this manuscript are available through a materials transfer agreement (MTA) with the National Institute of Allergy and Infectious Diseases.