Repertoire, function, and structure of serological antibodies induced by the R21/Matrix-M malaria vaccine

Plasmodium falciparum (Pf) malaria is one of the world’s most ancient diseases and remains the deadliest parasitic disease worldwide. The World Health Organization (WHO) estimated there were ∼608,000 malaria-related deaths worldwide in 2022 (Venkatesan, 2024). Although there are vaccine candidates under development that target each stage of the parasite’s life cycle, vaccines targeting the pre-erythrocytic malarial sporozoite phase are the most advanced. This malaria vaccine approach remains an area of intense research due to its potential to entirely circumvent liver and blood-stage infection and thereby prevent disease and onward transmission.

The circumsporozoite protein (CSP), the primary outer coating essential for sporozoite development and hepatocyte invasion, has long been an attractive malaria vaccine target as it is highly immunogenic and capable of inducing a dominant neutralizing titer after infection or vaccination (Hill, 2006; Kappe et al., 2004; Nussenzweig and Nussenzweig, 1989). Earlier in vitro studies have demonstrated antibodies targeting CSP can prevent sporozoites from infecting liver cells (Hollingdale et al., 1984). In murine models, both passive transfer of polyclonal and mAbs against the immunodominant B cell major repeat epitope of CSP and active immunization with constructs containing this epitope confer protection against sporozoite challenge (Foquet et al., 2014; Oyen et al., 2017; Wang et al., 1995; Zavala and Chai, 1990).

Pf CSP (PfCSP) consists of three antigenic domains: an N-terminal junctional domain with a highly conserved pentapeptide sequence, an immunodominant central repeat region consisting of ∼45 tetrapeptides (NANP “major repeat” and NPDP and NVDP “minor repeats”), and concludes with a C-terminal domain. The Asn-Ala-Asn-Pro (NANP) tetrapeptide of the central repeat region constitutes the immunodominant B cell epitope of PfCSP (Dame et al., 1984; Zavala et al., 1983), and IgG1 and IgG3 constitute the majority of the anti-PfCSP serological response (Ubillos et al., 2018). After more than six decades of diligent effort to create an effective preerythrocytic malaria vaccine based on PfCSP, in the last three years, two have been recommended by WHO for use in children living in malaria-endemic areas: RTS,S/AS01 (brand name Mosquirix, by GlaxoSmithKline) and R21/Matrix-M (developed by the University of Oxford’s Jenner Institute and manufactured by the Serum Institute of India).

RTS,S/AS01, recommended by WHO in 2021, is a chimera between the hepatitis B surface antigen and a truncated form of PfCSP that includes a portion of the unstructured and immunodominant NANP-repeating peptide and the C terminus domain (Miura et al., 2024; White et al., 2015; Zavala, 2022). Although several recent clinical trials have demonstrated RTS,S is safe and offers a high degree of protection immediately following vaccination, both the anti-CSP antibody titer and vaccine efficacy against infection wane over time (White et al., 2015; Zavala, 2022).

R21/Matrix-M was added to the WHO recommendation in October 2023 due to its demonstrated high efficacy (Datoo et al., 2024). The availability of R21/Matrix-M complements the ongoing distribution of RTS,S (Nnaji et al., 2024; Schmit et al., 2024) and greatly enhances overall access to malaria vaccines. R21/Matrix-M is a virus-like particle vaccine that targets the same CSP antigen as RTS,S and consists of Pf strain NF54, containing 19 copies of the NANP central repeat, and is formulated with Matrix-M adjuvant. Clinical trials of R21/Matrix-M have demonstrated high vaccine efficacy over one year with a three-dose regimen in 4,644 children aged 5–36 mo. Results indicate efficacy rates of 75% (95% confidence interval [CI] 71–79) at sites with moderate to high seasonal malaria transmission and 68% (95% CI 61–74) at sites with low to moderate perennial malaria transmission (Datoo et al., 2024). At this stage, trials of R21/Matrix-M have not been conducted in areas of high perennial transmission, and published trials do not yet allow for direct comparisons between the efficacy of the two vaccines (Moorthy et al., 2024). Mathematical modeling of three years of follow-up data stemming from an earlier phase 2b trial found that anti-CSP antibody titers satisfy the criteria for an immunological correlate of protection for R21/Matrix-M vaccine efficacy against clinical malaria (Schmit et al., 2024).

Whereas prior studies have examined the molecular features of anti-CSP antibody responses when elicited by natural infection, attenuated Pf whole-sporozoite (PfSPZ) vaccination, or RTS,S/AS01 vaccination, the molecular features of anti-CSP antibody responses following R21/Matrix-M vaccination have yet to be investigated. Until now, human CSP-reactive mAbs have been isolated exclusively by single-cell cloning using peripheral B cells (Imkeller et al., 2018; Kisalu et al., 2018; Murugan et al., 2018; Oyen et al., 2017; Tan et al., 2018; Triller et al., 2017; Williams et al., 2024). In these studies, B cells express germline antibodies or antibodies with low levels of mutation that preferentially express the highly homologous, allelically diverse IGHV3-30 or IGHV3-33 gene segments, which can mediate NANP recognition via a conserved tryptophan residue at position 52 (Trp52) of the germline-encoded complementarity determining region 2 of the heavy-chain variable region (CDR-H2) (Pholcharee et al., 2020). In contrast, the CSP epitopes targeted by the actual circulating plasma Igs (IgG) in vaccinated study subjects—especially those plasma antibodies that are most abundant and thereby play a dominant role in the polyclonal response—remain unknown.

Technological advances now provide a means to directly probe endogenous plasma IgG antibodies (de Graaf et al., 2022; Lavinder et al., 2015; Schulte et al., 2022; Townsend et al., 2024). Here, capitalizing on such advances, we set out to address at the molecular level the PfCSP humoral immune response induced by R21/Matrix-M vaccination in malaria-naïve volunteers, analyzing the plasma IgG antibody response to a NANP₆ peptide (NANP hexamer) after vaccination as well as before and after a controlled human malaria infection (CHMI) challenge. In this study, we show that the R21/Matrix-M malaria vaccine elicits a highly focused IgG humoral response in malaria-naïve individuals, predominantly involving minimally mutated antibodies utilizing IGHV3-30/3-33 genes, consistent with previous observations of B cell responses. Our findings extend this understanding by demonstrating that the IgG response remains both stable in quantity and unaltered in its antibody composition, even following CHMI. Additionally, a key discovery is the cross-reactivity of these vaccine-induced IgG antibodies with the N-terminal junction of the CSP, an epitope not included in the R21 vaccine, indicating epitope spreading within the serological repertoire. Altogether, our data offer a molecular-level dissection of R21/Matrix-M IgG humoral immunity and provide a foundation to explain, in part, the demonstrated vaccine efficacy of R21/Matrix-M.

A synergistic combination of Ig mass spectrometry (Ig-Seq) and B cell receptor (BCR) deep sequencing (BCR-Seq) was used to analyze the molecular composition and relative abundance of circulating anti-NANP IgG antibodies (see Materials and methods). For Ig-Seq proteomics, total IgG purified from volunteer plasma was subjected to affinity chromatography using recombinant NANP polypeptide (Asn-Ala-Asn-Pro hexamer [NANP₆]) representative of the CSP central repeat region. The eluted fraction (“NANP₆-reactive” IgG) and flow-through fraction (“NANP₆-nonreactive” IgG) were analyzed by high-resolution liquid chromatography and tandem mass spectrometry (LC–MS/MS). For BCR-Seq, next-generation sequencing (NGS) of peripheral blood BCR heavy-chain (VH), light-chain (VL), and single B cell–paired VH:VL variable region repertoires was performed. High-confidence IgG peptide spectral matches from the LC-MS/MS could then be identified using these volunteer-specific BCR databases—focusing on VH-derived hypervariable CDR-H3 peptides in particular—to determine the presence and relative abundance of IgG antibodies at both the lineage (family) and clonal (individual) levels. The integration of IgG peptide sequencing (Ig-Seq) with matching VH/VH:VL DNA sequencing (BCR-Seq) (Lavinder et al., 2015; Lavinder et al., 2014; Voss et al., 2021) allowed us to precisely identify and quantify circulating NANP₆-reactive IgG repertoires across a volunteer cohort at the molecular level. This approach also facilitated the cloning and validation of a small panel of recombinant IgG anti-NANP plasma mAbs for subsequent validation and functional assessment.

To evaluate the quality and evolution of the IgG humoral immune response to vaccination, we analyzed peripheral blood mononuclear cells (PBMC) and plasma samples collected from ten malaria-naïve volunteers in two separate phase I/IIa clinical trials, VAC053 (n = 4) and VAC065 (n = 6) (Fig. 1 A) (Venkatraman et al., 2025a; Venkatraman et al., 2025b; Venkatraman et al., 2024). In both trials, three intramuscular vaccinations were administered in 4-wk intervals on days 0, 28, and 56. For BCR-Seq, PBMC (containing antibody-secreting plasmablasts) were collected on day 63 from all 10 volunteers 1 wk after the third and final vaccination on day 56. For Ig-Seq, post-vaccination plasma was collected either on day 63 (VAC053) or on day 84 (VAC065). Unlike VAC053, volunteers in the VAC065 trial were challenged by CHMI with live sporozoite 4 wk after the day 56 final vaccination, with post-vaccination pre-challenge plasma collected on day 84 (referred to as “C₋₁”). VAC065 volunteers were monitored for protection from infection until a final blood draw 5 wk later on day 112 (referred to as “C₊₃₅”). Thus, plasma samples were analyzed after vaccination (VAC053 and VAC065) and after challenge (VAC065).

First, we set out to address fundamental questions about IgG repertoires generated by R21/Matrix-M vaccination: Are they simple or complex in terms of the number of antibody lineages (i.e., families of closely related antibody VH sequences)? Is the distribution of lineages uniform or polarized? In our study the average IgG anti-NANP₆ titer among the 10 volunteers was 38.5 ± 20.3 µg/ml, as determined by ELISA (Table 1). This result indicated a substantial level of circulating IgG, which was sufficient for further analysis using NANP₆ affinity chromatography, and Ig-Seq proteomic deconvolution to identify and quantify IgG lineages. Because this was a “virgin” immune response in malaria-naïve volunteers, these IgG lineages likely derived from the activation of a single naïve B cell, which then clonally diversified through somatic hypermutation (SHM).

In one representative volunteer, the IgG anti-NANP₆ repertoire comprised 21 distinct antibody lineages, each exceeding a relative abundance threshold of >0.5% (Fig. 1 B). Similarly, the size of the IgG anti-NANP₆ repertoire was consistent across VAC053 and VAC065 volunteers, with an average of 20 ± 9 antibody lineages (range: 5–33) contributing to the NANP₆ reactivity (Fig. S1 A; and Tables 1 and S1), similar to other reports of antigen-specific serological repertoires analyzed through proteomic mass spectrometry (Bondt et al., 2021; Townsend et al., 2024). The serological IgG anti-NANP repertoires also displayed a significant degree of polarization; typically, the two to five most prevalent lineages, out of the ∼20 per volunteer, accounted for over 50% of the NANP₆ antigen-reactive titer (Fig. 1 C and Table S1).

Despite IgG repertoires being oligoclonal and polarized, the CDR-H3 region of antibody combining sites—a somatically generated, key determinant of binding specificity for most antibodies (Ippolito et al., 2006; Xu and Davis, 2000)—was diverse in both length (range: 6–26 aa) (Fig. S1 B) and sequence composition (Table S1). Among 197 total IgG lineages derived from the 10 volunteers, there was no CDR-H3 repertoire overlap, nor any apparent repertoire convergence among volunteers according to nonlinear dimensionality reduction (t-distributed stochastic neighbor embedding [tSNE]) analysis, and therefore no evidence for “public” multi-donor anti-NANP₆ antibodies (Fig. S1, C and D).

Contrary to these features of diversity, a common characteristic across all the NANP₆-reactive IgG repertoires was an obvious enrichment for the variable-region gene segments IGHV3-30, IGHV3-33, or their related alleles (Fig. 1 D and Fig. S1 E). This VH family, including alleles with serine, arginine, or tryptophan at position 52 of the CDR-H2 (collectively referred to as “IGHV3-30/33”), encompasses IGHV3-30, IGHV3-30-3, IGHV3-30-5, and IGHV3-33 alleles that share over 96% identity (Tan et al., 2018). On average, IGHV3-30/33 circulating IgG antibodies constituted ∼70% of the NANP₆-reactive repertoire after vaccination (Fig. 1 D). Similarly, most PfCSP repeat mAbs isolated by single-cell cloning of peripheral human B cells are predominantly encoded by IGHV3-30/33 (Julien and Wardemann, 2019). While other IGHV genes played dominant roles in certain individuals in both the VAC053 and the VAC065 cohorts (Fig. 1, C and D), their infrequent usage suggests volunteer-specific, nonpublic motifs. In four of six VAC065 volunteers, non-IGHV3-30/33 genes increased in relative plasma abundance by day 84 after vaccination (MA010, MA013, MA061, and MA067); however, this was not the case for two of six VAC065 volunteers (MA015 and MA024), wherein IGHV3-30/33 dominance continued to prevail. We then compared the total peripheral B cell IgG repertoire from NGS (Fig. 1 E) with the NANP₆-nonreactive plasma IgG repertoire (Fig. 1 F). Averaged over the 10 volunteers, the similarity in IGHV3-30/33 frequency between the BCR-sequencing repertoire (14.7%) and the nonreactive plasma repertoire (10.1%) indicates that the observed ∼70% IGHV3-30/33 bias in the NANP₆-binding plasma IgG repertoire reported in Fig. 1 D must be due to antigen-specific enrichment. In contrast to the IGHV bias, IGHJ gene usage in plasma or the BCR repertoire was highly diversified and non-polarized (Fig. 1 G).

Another common characteristic of NANP₆-reactive plasma antibodies was a significant reduction in SHM among those using the IGHV3-33 gene segment. Antibody lineages not using IGHV3-33 showed SHM rates approximately threefold higher than those using IGHV3-33 (6.2% ± 4.8 versus 2.1% ± 1.9; Fig. 1 H). In support of the hypothesis that NANP recognition is fundamentally enhanced by amino acid encoded in the germline of the IGHV3-33 gene segment (CDR-H2 Val50 and Trp52, in particular) (Wahl and Wardemann, 2022), it was found that NANP₆-nonreactive IGHV3-33 antibodies were nearly three times more mutated compared with NANP₆-reactive antibodies (6.0% ± 3.8 versus 2.1% ± 1.9, respectively). Of note, nonreactive antibodies had a higher number of replacement mutations at positions Val50 and Trp52 (Fig. 1, I and J). Altogether, our results illustrate that the antibody features of the serological repertoire are consistent with findings from the analysis of IgG receptors isolated by B cell cloning showing that IGHV3-30/33 genes (IGHV3-33 primarily) confer binding to the CSP NANP major repeat without reliance on any exact CDR-H3 structure, IGHJ sequence, or acquired somatic mutations during B cell evolution.

Mosquito bites delivering sporozoite-stage malaria parasites have traditionally served as a model for testing preerythrocytic stage vaccines. Here, to gain a deeper understanding of the stability and quality of IgG anti-NANP antibodies induced by R21/Matrix-M vaccination, Anopheles stephensi mosquitoes and the chloroquine-sensitive NF54 strain (3D7 clone) of Pf were used for CHMI. To evaluate the immune response to vaccination and sporozoite challenge, we analyzed the IgG repertoire after vaccination in n = 5 volunteers at day 84 (C₋₁) and then monitored for protection from infection until a final blood draw 5 wk later at day 112 (C₊₃₅). Fig. 2, A and B show plasma antibody repertoires induced by R21/Matrix-M vaccination remained stable despite challenge, with a protected (Fig. 2 A) and a non-protected (Fig. 2 B) volunteer displaying consistent IgG lineage composition and relative abundance profiles for their NANP₆-reactive repertoires. The high degree of correlation between pre- and post-challenge repertoires is shown in scatter plots representing the relative abundance of each identified lineage at the time of plasma collection. The relative abundance of detectable NANP₆-reactive IgG antibodies in protected volunteer MA-1010 (Fig. 2 A), for instance, was strikingly correlated lineage-for-lineage and nearly indistinguishable before and after challenge (Pearson correlation, r² = 0.98). In non-protected subject MA-1015, the pre- and post-challenge IgG repertoires were similarly highly correlated at r² = 0.88 (Fig. 2 B). A strong degree of correlation (r² > 0.6) held for all five volunteers (Fig. S2). 35 days, or approximately two IgG1 half-lives, separated the pre- and post-challenge time points, suggesting that the volunteers’ static IgG repertoires are maintained by established, dynamic pools of antibody-secreting cells.

Given the static composition of IgG lineages before and after challenge with sporozoite, we wondered whether this molecular feature might also be reflected at the bulk serological level. Using a larger VAC065 cohort of 17 volunteers, we observed NANP₆ titers in both protected and non-protected groups showed no boosting effect after sporozoite challenge (Fig. 2, C and D). In fact, collectively, titers declined by an average of 27% between days C_-1 and C₊₃₅ over the 5-wk period between challenge and plasma collection. No significant differences were observed before challenge or after challenge between the absolute (µg/ml) or the relative (%) changes in titer when comparing protected and non-protected individuals. Upon reexamination of the subset of five volunteers analyzed by Ig-Seq, we note that persistent antibodies, defined as those present before and after challenge, accounted for over 95% of the total NANP₆-reactive repertoire (Fig. 2 E), with <5% of plasma antibodies after challenge being detectable only at the second time point (Fig. 2 F), which could explain in part the static or else decreased titers measured by bulk serology.

Cross-reactive antibodies induced by R21/Matrix-M target multiple CSP epitopes PfCSP features a “junction region” (JR) that connects the N-terminal domain to the central NANP repeat region. In the Pf reference isolate 3D7 (PfCSP_3D7), the JR starts with the sequence NPDP, followed by three interspersed NANP and NVDP repeats, with NPDP being unique to this region. Studies have shown that mAbs targeting the JR epitope in both mice and humans can be protective (Kayentao et al., 2022; Kayentao et al., 2024; Wang et al., 2020). B cell repertoire analyses have identified antibodies specific to the NANP sequence, and those cross-reactive with the JR are encoded by IGHV3-30/33 genes, particularly IGHV3-33 paired with IGKV1-5 (reviewed in Julien and Wardemann [2019]). In contrast, antibodies specific to the NVDP sequence and those cross-reactive with both NVDP and NPDP utilize various IGHV gene segments (Kisalu et al., 2018; Murugan et al., 2020). Notably, all JR-binding mAbs exhibit some level of cross-reactivity with NANP (Murugan et al., 2020). Despite the R21 vaccine containing only a truncated form of PfCSP with 19 NANP repeats and the C terminus—lacking the N terminus, as well as the NPDP and NVDP repeats of the JR (Collins et al., 2017; Stoute et al., 1997)—we found R21/Matrix-M could induce JR-reactive plasma antibodies in all five VAC065 volunteers tested at day 84 after vaccination (Fig. 3). The depletion of pre-challenge (C₋₁) and post-challenge (C₊₃₅) polyclonal IgG using NANP₆ affinity chromatography led to a significant reduction in JR-binding signals, as measured by indirect ELISA with a peptide encompassing the JR epitope (JR peptide: KQPADGNPDPNANPNVDPN).

To validate this at the molecular level, Ig-Seq analysis of pre-challenge IgG from one volunteer (MA-1010) using two separate pulldowns (NANP₆ and JR) demonstrated that the JR-reactive repertoire overlapped with the NANP₆-reactive repertoire lineage-for-lineage. Thus, in this volunteer, the JR-binding antibody lineages were a subset of the broader anti-NANP₆ repertoire. Of the seven JR cross-reactive lineages identified in MA-1010 (meeting quality criteria and exceeding 0.5% relative abundance), there was diverse IGHV gene usage, with only one lineage employing an IGHV3-30/33 family member (VH3-30). These seven lineages exhibited minimal somatic mutation in their IGHV gene segments (average 2.6%). The predominant lineage (VH3-7 at 47.5% relative abundance) had merely a single Asn52 mutation in its CDR-H2. Asn52 was previously identified as a key interacting residue in the dual-reactive mAbs Ab668 and CIS42 targeting NANP/JR (Kisalu et al., 2018; Oyen et al., 2020).

Overall, these results suggest the NANP polypeptide of the central repeat region in the R21/Matrix-M vaccine is sufficient to instigate the generation and/or evolution of diverse, near-germline IgG antibodies that are cross-reactive with the N-terminal junction—even though the JR is absent from the vaccine construct.

From VAC053 immunized volunteers, a small panel of nine PfCSP plasma mAbs (VH:VL pairs) were isolated from day 63 (day 7 after third vaccination) peripheral blood B cells. These nine plasma mAbs mapped with high confidence to proteomic CDR-H3 peptides of NANP₆-reactive IgG lineages identified by Ig-Seq proteomic analysis (and circulating at >0.5% relative abundance). mAbs were expressed recombinantly as full-length IgG1 and first tested for NANP₆ peptide-binding specificity by ELISA. Anti-sporozoite binding and inhibitory activity were also initially evaluated and compared with two previously characterized mAbs derived from B cells of naturally exposed individuals in Africa (Triller et al., 2017); these two mAbs, 663 and 580, can inhibit Pf sporozoite traversal in hepatocytes in vitro and recognize transgenic Plasmodium berghei (Pb) parasites expressing PfCSP (PfCSP@PbCSP).

Three positive IgG plasma mAbs were prioritized for full, replicate downstream experimentation (Table 2). All three antibodies use IGHV3-30/33 genes and represent a range of serological abundances, extremes of somatic mutation, and diverse expression of Ig light chains: MA1 (13% abundance; 1.6% SHM; IGLV1-40), MA6 (0.6%; 1.7%; IGKV1-5), and MA8 (19.6%; 8.2%; IGKV1-16). One negative IGHV1-69 mAb (MA5) was also carried forward as a control reagent. All three IGHV3-30/33 mAbs (MA1, MA6, and MA8) were NANP specific and did not bind the JR peptide according to conventional ELISA and a novel flow cytometry assay (Fig. S3 and data not shown).

MAbs MA1, MA6, and MA8 bound strongly to WT 3D7 sporozoites in an immunofluorescence assay (IFA) and to transgenic (PfCSP@PbCSP) sporozoites (Fig. 4 A). The functional capacity of MA1, MA6, and MA8 was then assessed using an in vitro model of liver-stage development using cultured human hepatocytes to measure inhibition of sporozoite invasion (ISI) activity with PfCSP@PbCSP. All mAbs tested, including 580 and 663, were able to block hepatocyte invasion, with MA1 and MA8 demonstrating a reduction in infection of 35% and 52%, respectively, while 580, 663, and MA6 all demonstrated more than 85% ISI activity (Fig. 4 B). To assess the ability of plasma antibodies induced by R21/Matrix-M to confer protection in vivo, we replicated the Triller et al. (2017) experimental scheme and included mAbs 580 and 663 as references. The model employs a transgenic strain of Pb, where the endogenous PbCSP is replaced with full-length PfCSP (PfCSP@PbCSP), enabling the assessment of PfCSP mAbs within the context of natural infection in a normal mouse, as rodents are permissive for Pb infection (Espinosa et al., 2017). Female C57BL/6 mice received 400 μg of antibody i.p. and were challenged the following day with 5,000 Pb-PfCSP sporozoites.

Beginning 5 days after sporozoite injection, mice were monitored for protection against liver-stage malaria, indicated by reduction in time to develop blood-stage parasitemia or complete absence of blood-stage parasites. In two independent experiments, MA1 and 580 sterilely protected 33% and 50% of mice, respectively, compared with naïve control mice, while MA5, as expected, did not demonstrate protective immunity in this model (Fig. 4, C–E). mAb 580 demonstrated reduced efficacy (50% protection) in our model compared with 72% in the original experiment by Triller and colleagues. MA6 and 663 protected 92% and 100% of challenged mice, respectively, demonstrating potent inhibition of malaria infection. Of note, ISI appeared more closely related to in vivo protection than IFA or binding to NANP (Fig. 4, B and E).

The IGHV3-33 germline-encoded tryptophan at position 52 in CDR-H2, which can also arise in some IGHV3-30 alleles associated with anti-NANP responses, plays a crucial role in antigen binding, as previously observed in several PfCSP repeat–reactive antibodies identified following immunization with sporozoites or vaccination with RTS,S/AS01 (Pholcharee et al., 2020). To further explore the molecular nature of near-germline IGHV3-33 antibodies generated by R21/Matrix-M vaccination, we selected mAb MA6 for biophysical and structural studies (Fig. 5). Isothermal titration calorimetry experiments performed with the antigen-binding fragment (F_ab) of MA6 and a representative peptide of the PfCSP repeat region (Ac-NPNANPNANPNA-NH₂) resulted in a calculated binding affinity of ∼400 nM. Structural studies have clarified that anti-repeat–binding motifs are in fact DPNA, NPNV, and NPNA, which result from the combination of major and minor repeats (Dyson et al., 1990; Pholcharee et al., 2020). The MA6 F_ab in complex with the peptide crystallized in space group P21 and diffracted x-rays to a resolution of 1.8 Å. The molecular replacement solution contained two F_ab–peptide complexes in the asymmetric unit. Electron density for most of the peptide was well defined, allowing us to unambiguously place the first nine amino acids. The electron density for the remaining three C-terminal peptide residues was weaker, and the corresponding residues could not be placed with confidence and were omitted from the model.

The structure revealed that the peptide lies within a groove formed primarily by the heavy chain, with CDR-H3 residues forming one side of the groove and CDR-H1 and -H2 residues forming the other side (Fig. 5 A). In total, there are 701 Ų of buried surface area on the peptide, 406 Ų buried on the heavy chain, and 110 Ų buried on the light chain, with the latter interface composed entirely of the CDR-L3. CDR-H3 residue Trp100B forms Van der Waals interactions with the peptide, as do CDR-H2 residues Trp52, Tyr52A, and Tyr58. On the CDR-L3, only the sidechains of Tyr94 and Trp96 substantially contribute to the binding interface with the peptide.

In addition to the hydrophobic interactions, numerous hydrogen bonds are formed between the peptide and antibody. Several are water mediated, including those involving the sidechains of CDR-H2 residues Trp52 and Tyr58. The sidechains of CDR-H3 residues Asp95 and Ser100C both directly hydrogen bond to the peptide, but otherwise all remaining hydrogen bonds involve antibody main chain atoms. This allows for sequence variability in the antibody to be accommodated and likely explains the lack of CDR-H3 sequence conservation among similar antibodies.

Several structures of NANP₆-reactive antibodies in complex with peptides have been determined, including antibodies 311 and 1210 (Imkeller et al., 2018; Murugan et al., 2018; Oyen et al., 2017). Of these, antibody 1210, which was obtained from a volunteer immunized with the Sanaria PfSPZ vaccine, is most like MA6, with 89% sequence identity between the heavy chains and 96% sequence identity between the light chains. Despite this overall high sequence similarity, the MA6 CDR-H3 is substantially different from the 1210 CDR-H3, with only 5 of 14 residues in common. Yet these antibodies bind the peptide in an almost identical conformation, resulting in a root mean square deviation (RMSD) of 0.23 Å for the first eight C_α atoms. X-ray crystallography of CSP-binding antibodies has previously revealed a key structural requirement for Trp52 in the CDR-H2 of the IGHV3-33 gene element; this feature is conserved in human mAbs 1210, MGG4, and 311, whether elicited by live sporozoites or RTS,S immunizations, as has been reviewed (Oyen et al., 2017; Pholcharee et al., 2021; Scally and Julien, 2018). Compared to other structures, the MA6 Trp52 sidechain is flipped such that the pyrrole ring of the indole, rather than the benzene ring, packs against Pro6 of the peptide. While mAb1210 and MA6 have nearly identical CDR-K3 sequences, they differ at a key residue previously identified (Imkeller et al., 2018) as critical for inter-F_ab interactions: mAb1210 contains Ser93, which enables hydrogen bonding with CDR-H3, whereas MA6 retains a germline Asn at this position. Nonetheless, MA6 may still form an inter-F_ab hydrogen bond via its Ser99 residue in CDR-H3. Collectively, the structure of MA6 in complex with a NPNA₃ peptide and its comparison to the peptide-bound 1210 structure provide a molecular basis for the dominance of near-germline IGHV3-33 NANP₆-reactive antibodies and their ability to accommodate diversity in their CDR-H3s.

The robust efficacy rates observed with R21/Matrix-M at 12 mo in areas with year-round malaria transmission (Datoo et al., 2024), along with promising preliminary data at 42 mo in seasonal transmission sites, indicate that R21 is a highly efficacious vaccine with significant potential to reduce the burden of malaria. Our molecular-level analysis of the R21-generated IgG plasma antibody repertoire—showing global similarity to previously reported RTS,S B cell repertoires and validating the robust potency of these antibodies—support this conclusion.

A significant finding in this study is the dominance of NANP repeat–specific antibodies with minimal SHM in the plasma IgG repertoire induced by the R21/Matrix-M malaria vaccine. Like RTS,S-induced B cell repertoires (Oyen et al., 2017; Williams et al., 2024), the IgG plasma repertoires elicited by R21/Matrix-M also are predominantly built upon germline or near-germline IGHV3 family genes (IGHV3-30/3-33). Through the cloning, functional, and structural analysis of prevalent IgG anti-NANP antibodies, we demonstrated how the R21-generated plasma antibody, MA6, inhibits sporozoite cell invasion in cell culture, curtails parasitemia, protects in a transgenic mouse model, and mediates NANP recognition through a canonical CDR-H2 Trp52 structural motif encoded by IGHV3-33. Since the MA6 VH region differs from the germline IGHV3-33 gene segment by only two amino acids in its framework 2 region—neither of which interacts with NPNA₃ in the F_ab–peptide crystal structure—we speculate that MA6 originated from the naïve B cell pool and may have seeded the IgG-secreting plasma cell repertoire via a noncanonical maturation pathway. Our data suggest that repeat-specific antibodies targeting the NANP region may not undergo the extensive SHM typically seen in more mature antibody responses, instead favoring an innate-like, early stage antibody pool that is rapidly mobilized following vaccination. This speculation aligns with both recent and past observations based on RTS,S and whole-sporozoite immunizations (Murugan et al., 2018; Murugan et al., 2020; Williams et al., 2024). Furthermore, the MA6 VL region is encoded by unmutated IGKV1-5 and contains an 8-aa long CDR-K3, akin to the public NANP-reactive “VH3-33/Vk1-5/KCDR3:8” antibody class repeatedly observed in CSP B cell repertoires (Murugan et al., 2018; Tan et al., 2018; Williams et al., 2024). Our data thus establish a connection between CSP/NANP immune responses in the naïve B cell compartment and their evolution into secreted, prevalent IgG circulating in the blood.

Notably, our study also demonstrated that CHMI challenge does not significantly boost the plasma IgG response. Despite the challenge, the quantity and composition of the NANP-reactive antibody repertoire remained stable. This finding, although counterintuitive, highlights the robust and static nature of the antibody response generated by the R21/Matrix-M vaccine. Since the elicitation of neutralizing IgG antibodies correlates with protection by vaccination for most licensed vaccines (Plotkin, 2010), and IgG anti-CSP titers meet the criteria for an immunological correlate of protection for the R21/Matrix-M vaccine (Schmit et al., 2024), we propose that the molecular profile of circulating antibodies elicited by R21/Matrix-M, as identified by Ig-Seq proteomic analysis, must reflect this fundamental biology. In this context, the temporal stability and static composition of R21-generated antibodies up to 5 wk after a sporozoite challenge infection underscore the vaccine’s ability to elicit an effective sporozoite-neutralizing response. Nevertheless, titer alone could not be correlated with protection, as no significant differences were observed before challenge or after challenge in absolute (µg/ml) or relative (%) changes in titer when comparing protected and non-protected individuals. This aligns with findings from the larger phase 1–2a CHMI trial (Venkatraman et al., 2025a; Venkatraman et al., 2025b), which reported no significant difference in peak NANP-specific IgG concentrations by challenge outcome.

Among the IgG antibodies involved in the vaccine response, we found that R21/Matrix-M generates antibodies targeting the CSP JR, in addition to the CSP NANP repeat region encoded by the vaccine immunogen, representing a subset of the broader anti-NANP₆ repertoire. Although further studies are needed, these cross-reactive antibodies may belong to the same class as the prophylactic mAbs CIS43 and L9, which have demonstrated high efficacy in advanced clinical field trials (Kayentao et al., 2022; Kayentao et al., 2024). On the other hand, we speculate that a new class of IgG antibodies may have been missed in our analysis of R21/Matrix-M. A recent report has identified a cryptic CSP epitope—pGlu-CSP at the region I proteolytic cleavage site—on the Pf sporozoite surface as a target of protective antibodies. Notably, these potent anti-sporozoite mAbs do not bind the PfCSP central repeat region found in RTS,S/AS01 and R21/Matrix M—suggesting they are unlikely to interfere with these vaccines and might be elicited only by native sporozoites, whether as an inactivated or attenuated vaccine, or by natural infection.

For the IGHV3-30/33 NANP-specific mAbs that were tested for in vivo protection, we selected two “germline” mAbs with similar minimal SHM (MA1 and MA6, ∼1.7% SHM; Table 2) to explore the relationship between binding affinity and in vivo efficacy. Notably, MA6, with nearly one order of magnitude greater binding affinity than MA1, provided the strongest protection against parasite challenge. While our data suggest that germline antibodies with higher binding affinity may offer superior in vivo protection, future studies will be needed to directly compare the affinity and in vivo efficacy of germline antibodies with mature antibodies that contain substantial somatic mutations. Such comparisons could provide further insights into the effectiveness of the R21/Matrix-M vaccine.

There are limitations to our study. Due to the separate PCR amplification of IgM versus IgG/IgA and the nonquantitative nature of our PCR method, we are unable to accurately quantify the distribution of IgM, IgG, and IgA in our BCR-Seq libraries. We plan to address this limitation in future studies by incorporating scRNA-seq. Because our universal IgG PCR primer binds to a conserved sequence common to all IgG subclasses, it cannot distinguish between the four Fc subtypes—thereby preventing us from assessing their individual contributions to the plasmablast or serological repertoire. All cloned mAbs were expressed exclusively in IgG1 format, and we do not know whether the endogenous VH region originated from an IgG1 or from one of the other IgG subclasses (IgG2, IgG3, or IgG4).

Applying Ig-Seq antibody proteomics to analyze circulating IgG produced by the R21/Matrix-M malaria vaccine at the molecular level demonstrates the power of this technique for evaluating preerythrocytic vaccines. Furthermore, Ig-Seq offers a promising approach for unraveling lingering serological enigmas in malarial disease, such as the sporadic production of autoantibodies during acute infection (Crompton et al., 2014; Donati et al., 2004). It may also shed light on the molecular composition of circulating immune complexes implicated in cerebral malaria and severe malarial anemias (Pleass, 2009) or could offer insights into the unexpected polyclonal synergism recently described for neutralizing and non-neutralizing antibodies elicited by the strain-transcendent RH5/Matrix-M vaccine (Barrett et al., 2024; Silk et al., 2024). These insights expand our understanding of these key interactions and inform the development of future next-generation malaria vaccines.

Plasma and PBMC samples were collected from 10 healthy malaria-naïve United Kingdom adult volunteers. For four volunteers, blood was drawn 7 days (day 63) following their third and final (day 56) vaccination with R21/Matrix-M (VAC053; clinicaltrials.gov identifier NCT02572388; [Venkatraman et al., 2024]). For six volunteers, blood was drawn 28 days (day 84) following their third and final (day 56) vaccination. (VAC065; clinicaltrials.gov identifier NCT02905019; [Venkatraman et al., 2025a; Venkatraman et al., 2025b]). The six VAC065 volunteers also had blood drawn 35 days (day 112) following a day 85 challenge infection by mosquito-bite transfer of Pf 3D7 sporozoites. Both clinical trials were conducted by the Jenner Institute at the University of Oxford and approved by the UK National Research Ethics Service, Committee South Central–Berkshire (16/SC/0261), the Medicines and Healthcare Products Regulatory Agency (21584/0360/001-0001), and the Oxford University Clinical Trials and Research Governance team.

Frozen PBMCs were thawed at 37°C, resuspended in Roswell Park Memorial Institute medium (RPMI)-1640 (Thermo Fisher Scientific) supplemented with 10% FBS and 20 U/ml deoxyribonuclease I (Roche), concentrated by centrifugation (300 × g for 15 min at 20°C), and allowed to recover in 4 ml media at 37°C for 30 min. The cells were then split: half were dedicated to BCR heavy chain transcript sequencing, and half were used to identify the native VHVL repertoire.

Cells were diluted with 10 ml of cold buffer (PBS supplemented with 0.5% BSA and 2 mM EDTA) and pelleted via centrifugation (300 × g for 15 min at 4°C). The supernatant was decanted, and the cells were resuspended in 1 ml of TRIzol reagent (Thermo Fisher Scientific). RNA was extracted using RNeasy (Qiagen), and first-strand cDNA was synthesized from 500 ng total RNA using SuperScript IV (Invitrogen). IgG, IgA, and IgM heavy chain repertoires were amplified using a multiplex primer set (Ippolito et al., 2012) and sequenced by 2 × 300 paired-end Illumina MiSeq.

Total B cells were isolated from PBMCs using the Human Memory B Cell Isolation Kit (Miltenyi Biotec) with an LD column according to the manufacturer’s instructions. B cells were then coemulsified with poly-dT magnetic beads and lysis buffer using a custom-designed flow-focusing device as described (McDaniel et al., 2016). VH and VL amplicons from single B cells were amplified, stitched together using overlap-extension RT-PCR, and sequenced by 2 × 300 Illumina MiSeq.

1–2 ml plasma from each volunteer was purified with Protein G Plus agarose affinity chromatography (Thermo Fisher Scientific), and IgG F(ab′)₂ fragments were generated with IdeS at a 1:50 ratio at 37°C for 2 h. NANP₆-reactive F(ab′)₂ fragments were isolated by affinity chromatography using NeutrAvidin resin (Pierce) coupled to biotinylated NANP₆ peptide (ABclonal) and eluted in 100 mM glycine (pH 2.7).

For each sample, the flow-through (nonreactive to NANP₆) and NANP₆-reactive elution were separately denatured in 1:1 (vol/vol) trifluoroethanol, reduced with 5 μl dithiothreitol (110 mM) at 55°C for 45 min, and alkylated with 32 mM iodoacetamide by incubating in the dark at room temperature (RT) for 30 min. Samples were diluted 10× into 50 mM Tris (pH 8.0) and digested with trypsin (1:10 trypsin/protein) for 3 h at 37°C. Formic acid (0.1%) was used to quench the reaction. The solution was then concentrated, desalted using a C18 Hypersep SpinTip (Thermo Fisher Scientific) according to the manufacturer’s instructions, and submitted for LC-MS/MS.

Samples were submitted to the University of Texas (UT) Austin center for biomedical research (CBRS) Proteomics Facility for protein identification by LC-MS/MS using the Dionex Ultimate 3000 RSLCnano LC coupled to the Thermo Orbitrap Fusion. A 2-cm long × 75-µm I.D. C18 trap column was followed by a 75-µm I.D. × 25-cm long analytical column packed with C18 3-µm material (Thermo Acclaim PepMap 100). The FT-MS resolution was set to 120,000, and 3-s cycle time MS/MS were acquired in ion trap mode.

SEQUEST (Proteome Discoverer 1.4, Thermo Fisher Scientific) with previously described settings (Williams et al., 2017) was used to search the spectra against a patient-specific protein database constructed from the full-length VH and VL sequences, Ensembl human protein-coding sequences, and common contaminants (https://maxquant.org). Peptide-spectrum matches (PSMs) were filtered with Percolator (Proteome Discoverer 1.4) at a false discovery rate of <1% and filtered for average mass deviations <1.5 parts per million. Peptides mapping to the CDR-H3 region of unique antibody lineages were grouped, and the relative abundances of the corresponding peptide matches were determined by the sum of the extracted ion chromatogram peak areas of the respective precursor ions, as previously described (Lavinder et al., 2014; Lee et al., 2016; Voss et al., 2021).

We chose mAbs that demonstrated high-confidence identification of CDR-H3 peptides in plasma via LC-MS/MS, evidenced by high abundance (indicated by the extracted ion chromatogram area) and extensive peptide coverage of the VH, particularly in the hypervariable CDR regions. Selected antibody sequences were purchased from Twist Bioscience as gene fragments cloned into a customized pcDNA3.4 vector (Invitrogen) containing human IgG1 Fc regions, IgK1 Fc regions, or IgL2 Fc regions. VH and VL plasmids were transfected into 30 ml cultures of Expi293F cells (Invitrogen) at a 1:2 ratio and incubated at 37°C and 8% CO₂ for 7 days. The supernatant containing secreted antibodies was collected following centrifugation (1,000 × g for 10 min at 4°C), neutralized, and filtered. Antibodies were isolated using Protein G Plus agarose (Thermo Fisher Scientific) affinity chromatography, washed with 20 column volumes of PBS, eluted with 100 mM glycine-HCl, pH 2.7, and neutralized with 1 M Tris-HCl, pH 8.0. The antibodies were then concentrated and buffer exchanged into PBS using 10,000 MWCO Vivaspin centrifugal spin columns (Sartorius).

The amino acid sequences for MA6 heavy and light chain variable domains were codon optimized and synthesized as gBlock gene fragments (Integrated DNA Technologies). Gene fragments were cloned into the mammalian expression plasmid pVRC8400 with an HRV 3C protease cleavage site inserted within the hinge region of the heavy chain sequence. Plasmids were used to co-transfect 0.5 L FreeStyle 293-F cells (Invitrogen) with 0.167 mg heavy chain and 0.083 mg light chain DNA using polyethylenimine 25K (Polysciences, Inc.). 6 days after transfection, cells were harvested and centrifuged at 5,400 × g before supernatant containing soluble protein was concentrated and buffer exchanged into PBS using a 30-kDa tangential flow filtration cassette (PALL). The sample was then passed over Pierce Protein A Plus agarose resin (Thermo Fisher Scientific) in a gravity flow column. Eluate was processed with HRV 3C protease and passed over a gravity flow column packed with CaptureSelect IgG-CH1 resin to purify the MA6 F_ab. A purification polishing step was then performed by passing the eluate over a Superdex 200 10/300 GL gel filtration column (Cytiva) in crystallization running buffer (2 mM Tris, pH 8.0 and 200 mM NaCl). F_ab peak fractions were concentrated to 11.88 mg/ml using an Amicon 30,000 molecular weight cutoff (MWCO) centrifugal spin column (EMD Millipore). NPNA₃ peptide (ABclonal) was solubilized in crystallization buffer (2 mM Tris, pH 8.0 and 200 mM NaCl) to a stock concentration of 500 µM.

F_ab and NPNA₃ peptide were combined at a 1:12 M ratio and concentrated to 12.5 mg/ml. Crystals were produced using the sitting-drop vapor diffusion method by mixing 0.1 μl of MA6 F_ab–NPNA₃ complex with 0.05 μl of reservoir solution containing 27% PEG 3350 and 3% 2-methyl-2,4-pentanediol (MPD). The crystal was flash-cooled in liquid nitrogen directly from the crystallization drop. Diffraction data were collected to 1.8 Å from a single crystal at SBC beamline 19ID (Advanced Photon Source, Argonne National Laboratory).

Data were indexed and integrated in iMOSFLM (Battye et al., 2011) before being merged and scaled using Aimless (Evans and Murshudov, 2013). Molecular replacement was performed in Phaser (McCoy et al., 2007), and the model was subjected to multiple rounds of building and refinement in Coot (Emsley and Cowtan, 2004) and PHENIX (Adams et al., 2002), respectively. Data collection and refinement statistics can be found in Table 3. SBGrid (Morin et al., 2013) compiles all the crystallography processing and refinement programs used in this study.

Chambered microscope slides coated with either WT 3D7 Pf or transgenic Pb expressing PfCSP sporozoites were stored at −80°C until use. Slides were brought to RT and then fixed for 15 min in 4% paraformaldehyde. After washing twice in PBS for 5 min, slides were blocked for 1 h in casein. Slides were washed as before, and 10 μl of serum sample diluted 1:100 in casein was added to each well. Slides were incubated for 30 min at RT in a humidity chamber, then wells were individually washed with PBS three times for 5 min. Secondary antibody (anti–IgG-Alexa Fluor 488) was diluted 1:800 in casein, and 15 μl was added to each well. Slides were incubated for 30–45 min in a humidity chamber at RT protected from light. After a final wash, slides were rinsed in distilled water and left to dry before mounting with DAPI-containing media. Slides were set overnight at 4°C before being examined under a Leica DMI3000 B microscope. Images were captured in QCapturePro software using bright-field illumination, GFP, and DAPI filters at set exposure levels. ImageJ software was used to measure the median fluorescent intensity for five sporozoites in each well, and an average was taken.

ISI was assessed as previously described (Bowyer et al., 2018; Rodríguez-Galán et al., 2017). Human hepatoma cells (HC04 or Huh7) cultured in RPMI with 10% FBS were added to 96-well plates at 30,000 cells/well and left to settle overnight at 37°C, 5% CO₂. Viable GFP-labeled Pb sporozoites expressing PfCSP at the Pb CSP locus (Pb PfCSP@PbCSP) were obtained by dissecting infected A. stephensi mosquitoes. Dissected salivary glands were pooled into RPMI 1640. Culture medium was aspirated from the hepatoma cells, then 100 μl of serum diluted 1:5 in RPMI with 10% FBS, and 100 μl of sporozoite solution (10,000 sporozoites) were added to each well, giving a 10% final serum concentration. Samples were tested in duplicate: “hepatoma only” wells and positive control wells containing hepatoma cells and sporozoites, but no serum, were included. Baseline (pre-vaccination) and post-vaccination samples were run for each volunteer tested. After incubation for 20–26 h at 37°C, media was aspirated and plates were washed with 90 μl/well DPBS. Cells were trypsinized, resuspended in 65 μl DPBS with 1% BSA, and acquired immediately using a BD LSR II and FACSDiva v6.2. DAPI stain was added to each sample just before acquisition. Data were analyzed in FlowJo software v10.6 (TreeStar, Inc.).

Mice were used in accordance with the UK Animals (Scientific Procedures) Act under project license number P9804B4F1 granted by the UK Home Office following ethical review by the University of Oxford Animal Welfare and Ethical Review Board. Animals were group housed in individually ventilated cages under specific pathogen–free conditions, with constant temperature and humidity and lighting on a 13:11 (7 am to 8 pm) light–dark cycle. Each mouse received 400 µg of mAb i.p. injection the day before challenge with 5,000 Pb-PfCSP sporozoites.

Pb sporozoites expressing PfCSP under the control of the PbCSP promoter (Pb-PfCSP) (Rodríguez-Galán et al., 2017) were isolated from salivary glands of female A. stephensi mosquitoes around 21 days after feeding on a Pb-PfCSP blood stage infected donor mouse. Salivary glands were homogenized, sporozoites counted under phase contrast microscopy, and 5,000 Pb-PfCSP sporozoites injected s.c. into recipient mice. Mice were monitored daily from day 5 onward by taking a thin blood film and staining with 5% Giemsa (Sigma-Aldrich) to screen for the presence of schizonts within the red blood cells. Parasitemia was calculated as the percentage of infected red blood cells per microscope field (100× objective), with at least five fields counted per mouse per day. Using linear regression, the time to 1% parasitemia was calculated based on the y-intercept and slope of the line.

Nonparametric Mann–Whitney U test was used to compare two groups. Two-tailed Kruskal–Wallis test followed by Dunn’s multiple comparisons test was used to compare groups of three or more. Kaplan–Meier survival analysis was tested by the log-rank test. Sample sizes for each experiment were based on a power calculation and/or previous experience of the mouse models. Data analysis and visualization were performed using GraphPad Prism, Microsoft Excel, Microsoft PowerPoint, Adobe Illustrator, Immunarch, and R Studio. P values <0.05 were considered statistically significant.

Fig. S1 shows the profile of NANP₆-reactive IgG repertoires across all 10 volunteers after vaccination, indicating number of IgG lineages per volunteer, frequency of CDR-H3 lengths for all lineages, dissimilarity of VH repertoires according to tSNE analysis, and frequency of IGHV gene use by volunteer. Fig. S2 shows the composition and correlation between NANP6 pre-challenge and post-challenge IgG repertoires among VAC065 volunteers. Fig. S3 shows the epitope specificity of MA6 mAb by flow cytometry. Table S1 shows the R21 Ig-Seq repertoire summaries for each subject.

Sequence data are openly available in a public repository. Raw Illumina sequencing reads for natively paired VH:VL and VH-only BCR sequences (BCR-Seq) have been deposited at NCBI Sequence Read Archive under BioProject ID PRJNA1216750 and are publicly available as of the date of publication. Raw LC-MS/MS data for plasma IgG proteomics (Ig-Seq) have been deposited to the MassIVE proteomics repository (dataset MSV000096964) and are publicly available as of the date of publication. The mAbs have been deposited to GenBank with accession numbers PV667658–PV667665. Coordinates for the MA6 Fab in complex with NANP3 peptide have been deposited to the Protein Data Bank under accession no. 9CR9. All other data are presented in the figures and supplementary materials.

NGS was performed by the Genomic Sequencing and Analysis Facility at UT Austin, Center for Biomedical Research Support, RRID#: SCR_021713. Mass spectrometry was performed by the Biological Mass Spectrometry Facility at UT Austin. The authors acknowledge the Texas Advanced Computing Center at UT Austin for providing high-performance computing resources that have contributed to the research results reported within this paper.

This work was funded by PATH’s Center for Vaccine Innovation and Access, through a grant awarded to PATH by the Bill & Melinda Gates Foundation (INV-007217), under a collaborative agreement with The Jenner Institute at the University of Oxford (K.J. Ewer) and UT Austin (G.C. Ippolito). This work was partly funded by Welch Foundation grant number F-0003-19620604 (J.S. McLellan). Open Access funding provided by the University of Oxford.

Author contributions: J.R. McDaniel: conceptualization, data curation, formal analysis, investigation, methodology, software, validation, visualization, and writing—original draft, review, and editing. W.N. Voss: formal analysis, investigation, visualization, and writing—original draft, review, and editing. G. Bowyer: conceptualization, data curation, formal analysis, investigation, visualization, and writing—review and editing. S.A. Rush: formal analysis, investigation, visualization, and writing—original draft. A.J. Spencer: formal analysis, investigation, methodology, resources, and writing—review and editing. D.G. Bellamy: investigation. M. Ulaszewska: resources. J. Goike: investigation. S. Gregory: project administration. C.R. King: conceptualization, funding acquisition, investigation, and writing—review and editing. J. McLellan: investigation, resources, supervision, validation, visualization, and writing—review and editing. A.V.S. Hill: funding acquisition, project administration, resources, supervision, and writing—review and editing. G. Georgiou: conceptualization, funding acquisition, and supervision. K.J. Ewer: formal analysis, funding acquisition, resources, supervision, visualization, and writing—original draft, review, and editing. G.C. Ippolito: conceptualization, formal analysis, funding acquisition, investigation, methodology, project administration, supervision, validation, visualization, and writing—original draft, review, and editing.