Overcoming immunogenic gaps in malaria subunit vaccines by broadening CSP regions targeted

Malaria caused by infection with Plasmodium falciparum leads to more than half a million deaths a year, primarily in young children in Africa (World Health Organization, 2024). Infection is initiated by a mosquito bite; the plasmodial parasite enters through the dermis or blood as a small number of sporozoites (SPZ) (Frischknecht and Matuschewski, 2017; Phillips et al., 2017). SPZ then quickly reach the liver to exponentially expand before release into the blood and clinical infection. Prevention efforts have focused on limiting SPZ infection in the liver (Cockburn and Seder, 2018).

The World Health Organization has recommended two vaccines for malaria prevention, RTS,S and R21 (World Health Organization, 2024). Both are targeted to the P. falciparum circumsporozoite protein (CSP), which covers the exterior of the SPZ (Nussenzweig and Nussenzweig, 1985) and is composed of N-terminal, central repeat, and C-terminal domains. The central repeat domain primarily consists of ∼40 repeats of the amino acids (AAs) NANP, also known as the major repeats, as well as one to four repeats of NVDP, known as the minor repeat region (Enea et al., 1984; Zeeshan et al., 2012). Additionally, there is a junctional region that connects the central repeat domain to the N-terminal domain. Both RTS,S and R21 vaccines present a portion of the NANP central repeat region and the C terminus fused to a hepatitis B surface antigen (HBsAg) (Gordon et al., 1995). RTS,S provides ∼50% protection at 1 year and 36% protection at 4 years (Feng et al., 2022; RTS,S Clinical Trials Partnership, 2015). R21 has ∼75% efficacy after 1 year (Datoo et al., 2024). Efficacy for both RTS,S and R21 requires an initial three-dose series in the first year of life and a booster in the second year of life to maintain high titers of protective antibodies (Abs) (Datoo et al., 2024). The requirement for multiple doses over 2 years and the waning efficacy over time highlight the importance of maintaining a magnitude of Ab responses sufficient to limit infection (White et al., 2015). The induction of higher potency Abs at lower titers or increasing the breadth of Ab responses to other regions across P. falciparum CSP (PfCSP) not contained in the current vaccines provides a potentially promising approach to enhance the efficacy and durability of protection by PfCSP-based vaccines.

Recently, the discovery of mAbs against the junctional and minor repeat regions of PfCSP that are highly effective at preventing malaria infection in humans in the field has validated the importance of targeting these epitopes with vaccines (Kisalu et al., 2018; Wang et al., 2020). Neither of these epitopes is included in RTS,S or R21 (Datoo et al., 2024; Julien and Wardemann, 2019; Tan et al., 2019); specifically, both vaccines display a truncated portion of PfCSP containing 18 central NANP repeats and the C-terminal region (Collins et al., 2017). Recent human clinical trials have demonstrated the high protective efficacy of the anti-junctional CIS43LS and anti-minor L9LS Abs (Wu et al., 2022; Kayentao et al., 2022, 2024; Gaudinski et al., 2021), underscoring the importance of eliciting responses against these epitopes. Both the junctional and minor epitopes contain NANP sequences, highlighting promiscuity in Abs—including L9 and CIS43—that may bind across the major, minor, and junctional regions (Wang et al., 2024; Julien and Wardemann, 2019); Abs capable of binding these regions may, therefore, be elicited by R21 and RTS,S. However, our group previously found that immunofocusing to the junctional region, rather than delivering the full PfCSP protein, was a more effective method for the elicitation of anti-junction responses (Kratochvil et al., 2021); alternative deliveries of junctional epitopes have also demonstrated some protective efficacy (Francica et al., 2021; Jelínková et al., 2021; Tripathi et al., 2026). The efficacy of current vaccines in eliciting responses to these regions remains unknown.

Here, we used a set of transgenic mouse models with B cells expressing mature or germline equivalents of human antimalarial Abs to recapitulate the development of protective responses by these direct specificities of the human B cell repertoire. We assessed the specificity of B cell and Ab responses using immunogens to the major, minor, and junctional regions alone or in combination. We first show that neither the portions of PfCSP included in the R21 vaccine nor full-length PfCSP could activate anti-minor precursor B cells of the highly protective mAb L9 (Wang et al., 2020). To address this limitation, we identified an anti-minor repeat immunofocusing peptide that successfully activated these precursor B cells. This peptide induced longer germinal center (GC) responses than the full-length PfCSP protein, enabling extended somatic hypermutation (SHM), and also promoted significant GC responses and further affinity maturation of high-affinity mature L9. We used bioinformatic, biophysical, and structural analyses to characterize the Ab responses after immunization. Furthermore, combining PfCSP epitopes present in R21 with the immunofocusing peptides elicited balanced and protective B cell and Ab responses to major, minor, and junctional epitopes, offering a promising strategy to enhance malaria vaccine efficacy.

To investigate B cell responses to regions across PfCSP (Fig. 1 A), we created new knock-in (KI) B cell receptor (BCR) mouse models, with B cells bearing the pre-rearranged heavy chain (HC) and light chain (LC) sequences of the inferred germline sequences (iGL) of human anti-malarial Abs to different regions of PfCSP. We generated three new lines using our CRISPR/Cas9 approach (Lin et al., 2018; Wang et al., 2021b): two bearing inferred germline precursors to the anti-major repeat region Abs 311 and 317 (Oyen et al., 2017), iGL_311 and iGL_317, and one containing the anti-minor repeat region Ab L9 (Wang et al., 2020) precursor, iGL_L9. Additionally, we used a previously constructed KI with B cells bearing the anti-junctional region CIS43 (Kisalu et al., 2018) precursor, iGL_CIS43 (Kratochvil et al., 2021) (Fig. 1 A). We isolated B cells from each new line to confirm BCR functionality. Naïve B cells from the iGL_311 and iGL_317 models could bind to a full-length PfCSP probe, and all of the PfCSP-binding B cells bore the HC and LC sequences of iGL_311 or iGL_317 (Fig. 1, B and C). Similarly, naïve B cells from the iGL_L9 model also bound to a PfCSP probe; 96% of those PfCSP⁺ binders expressed the iGL_L9 HC and LC sequences (Fig. 1 D).

To develop models with lower precursor frequencies more relevant to human physiology, we adoptively transferred CD45.2⁺ KI B cells intravenously into congenic CD45.1⁺ WT host mice (Abbott et al., 2018; Dosenovic et al., 2018). To estimate transfer efficiency, we transferred 20,000, 50,000, 100,000, 250,000, and 500,000 CD45.2⁺ B cells and assessed the frequency of PfCSP⁺ CD45.2⁺ cells in the spleen after 1 day (Fig. S1). While exact precursor frequencies for these specificities in humans have not yet been defined, adoptive transfer allowed us to achieve low, stringent ranges (Havenar-Daughton et al., 2018).

To determine whether anti-major (iGL_311, iGL_317), anti-junctional (iGL_CIS43), and anti-minor (iGL_L9) B cells could be activated by the R21 epitope, we used a recombinant protein encoding the same PfCSP sequence—18 NANP repeats and the C-terminal—as R21, without the HBsAg; this is referred to as “truncated PfCSP (R21-epitope)” below. CD45.2⁺ B cells from each KI line were separately adoptively transferred into congenic CD45.1⁺ WT host mice to achieve a precursor frequency of 5 in 10⁵ B cells, and the recipient mice were then immunized with truncated PfCSP (R21-epitope) and Alhydrogel, and the GC response of splenic B cells was analyzed (Fig. 1 E). The frequency of total GC B cells induced by truncated PfCSP (R21-epitope) was not significantly different across adoptive transfer models (means ranged 0.44–1.0%). Immunization with truncated PfCSP (R21-epitope) selectively activated GC responses from the anti-major repeat iGL_311 and iGL_317 CD45.2⁺ B cells; however, iGL_311 was present at higher frequencies in the GC than iGL_317 (28.2 and 9.1% of GC B cells, respectively). In contrast, truncated PfCSP (R21-epitope) failed to elicit GC responses from either the anti-junctional iGL_CIS43 or anti-minor repeat iGL_L9 B cells (Fig. 1, F and G). To substantiate the in vivo data, we measured the apparent affinity between the different germline Abs and truncated PfCSP (R21-epitope) using biolayer interferometry (BLI). iGL_311 showed the highest affinity to truncated PfCSP (<1.0 × 10⁻¹² M), and iGL_317 also displayed measurable affinity (4.07 × 10⁻⁹ M). In contrast, neither iGL_CIS43 nor iGL_L9 displayed detectable binding (Fig. 1 H).

We then deployed the same approach to assess whether full-length PfCSP, which includes the junctional and minor repeat regions, could activate iGL_CIS43 and iGL_L9 B cells as well as the major repeat iGL_317 and iGL_311 precursors (Fig. 1 I). CD45.2⁺ iGL_317 cells were substantially recruited into the GC, comprising 56.4% of GC B cells. iGL_311 cells were also recruited into the GC, but unexpectedly, recruitment was weaker than that of iGL_317. iGL_CIS43 and iGL_L9 cells were also detected within the GC but at significantly lower frequencies (6.7 and 0.5%, respectively). The GC response from iGL_L9 was ∼100-fold lower than that of iGL_317 and 10-fold lower than iGL_CIS43 (Fig. 1, J and K). We next measured the apparent affinity of germline Abs for full-length PfCSP. Both iGL_311 and iGL_317 bound PfCSP, with iGL_311 showing higher affinity (7.68 × 10⁻¹⁰ M and 2.92 × 10⁻⁹ M). In contrast, iGL_CIS43 and iGL_L9 showed no detectable binding (Fig. 1 L).

Collectively, our results demonstrate that both full-length and truncated PfCSP (R21-epitope) can strongly activate anti-major repeat B cells but offer limited activation of junctional or minor repeat B cells.

As the mAb L9 is highly protective in humans, we explored further strategies to robustly activate iGL_L9. Full-length PfCSP, although it contains the minor repeat region that L9 preferentially targets, failed to recruit substantial numbers of iGL_L9 B cells into GCs on day 10 in our adoptive transfer model. To rule out the possibility that this reflected the short-time frame of the initial analysis, we extended the analysis to days 13 and 28 (Fig. S2 A) and again found no significant recruitment (Fig. S2, B and C). As ∼96% of full-length PfCSP consists of major repeats, with the L9-targeted minor repeats making up only ∼4% (Fig. 2 A, left) (Zeeshan et al., 2012), we used an immunofocusing approach, similar to our prior studies showing that a junctional peptide (NPDP19) could increase GC responses and affinity maturation in CIS43 precursors compared with full-length PfCSP (Kratochvil et al., 2021). A set of peptides across the junction of PfCSP containing varying numbers of minor repeat epitopes was made (Fig. 2 A, right). As L9 has been observed to bind more strongly to peptides with two NPNV motifs than to those with one (Wang et al., 2020; Tripathi et al., 2023; Martin et al., 2023), we compared KLH-conjugated peptides containing one (Pep21), two (Pep22), or three (“Pep22 long”) NPNV motifs (Fig. 2, A and B). Of note, Pep21 is the junctional epitope that was defined by CIS43, while Pep22 is the minor epitope defined by L9. GC frequencies were similar across groups (∼5–7%), but 1NPNV (Pep21-KLH) recruited significantly fewer CD45.2⁺ iGL_L9 B cells into GCs (0.71%) than 2NPNV (Pep22-KLH, 5.7%) or 3NPNV (Pep22 long-KLH, 2.19%) (Fig. 2 C and Fig. S2 D). This resulted in significantly higher frequencies of CD45.2⁺ GC B cells in the 2NPNV (0.21%) and 3NPNV (0.13%) groups, but not in the 1NPNV group (0.03%) on day 10 (Fig. 2 D).

Based on its comparative advantage in CD45.2 recruitment, Pep22-KLH was selected for comparison against PfCSP and the major repeat peptide NANP18-KLH (Fig. 2 E and Fig. S2 E) to track initiation of the GC response on days 5–7. Pep22-KLH recruited a higher fraction of CD45.2⁺ iGL_L9 B cells into GCs (9.5%) than PfCSP (2.00%; approximately fivefold lower) or NANP18-KLH (0.98%, ∼10-fold lower) on day 7. This resulted in a significantly higher frequency of CD45.2⁺ GC B cells at day 7 in the Pep22-KLH–immunized group (0.59%) compared with PfCSP (0.014%; ∼40-fold lower) or NANP18-KLH (0.032%; ∼18-fold lower) (Fig. 2 F and Fig. S2 F). Pep22-KLH also promoted greater Ab-secreting cell (ASC) differentiation and class switching (Fig. 2 G and Fig. S2 G). Kinetic analysis out to day 28 revealed that iGL_L9 GCB cells persisted only in the Pep22-KLH group (Fig. 2 H).

The findings with low-affinity iGL precursors raised the question of whether Pep22-KLH could sustain GC responses from high-affinity B cells generated through prior malaria exposure. We generated a BCR KI line expressing high-affinity, mature L9, confirmed PfCSP binding (Fig. S2, H and I), and verified HC and LC identity (Fig. S2 J). L9 B cells were adoptively transferred into WT recipients, which were then immunized with PfCSP, Pep22-KLH, or NANP18-KLH (Fig. 2 I). On day 7, GC frequencies were highest in NANP18-KLH (5.67%), intermediate in Pep22-KLH (3.24%), and lowest in PfCSP (1.25%) (Fig. S2 K, left). In contrast, the proportion of CD45.2⁺ L9 GC B cells was highest following immunization with PfCSP (60.5%), followed by Pep22-KLH (21.4%) and NANP18-KLH (14.5%) (Fig. S2 K, right), resulting in a similar overall frequency of CD45.2⁺ GC B cells across all three groups at day 7 (Fig. 2 J). PfCSP induced a significantly stronger early ASC response than either peptide group at day 6 (Fig. 2 K and Fig. S2 L). Longitudinal tracking (Fig. 2 L) showed that GC frequencies fell over time in all groups, but the persistence of mature L9 B cells within GCs differed sharply. NANP18-KLH–recruited cells were almost absent by day 14, and PfCSP-induced cells fell from 50% at day 7–7.7% by day 28. In contrast, Pep22-KLH maintained L9 CD45.2⁺ cells as 14.8% of GC B cells at day 7 and 22.3% by day 28 (Fig. 2 M and Fig. S2 M); Pep22-KLH recipients were thus the only group in which CD45.2⁺ of GCB cells increased relative to day 7 (Fig. 2 N). Thus, focusing the immune response on the minor repeats produced both strong initial recruitment and sustained GC persistence of low-affinity iGL_L9 B cells, as well as long-term GC retention of high-affinity L9 B cells.

To confirm whether mutations associated with malaria neutralization by L9 (Wang et al., 2020) were induced, we sorted CD45.2⁺PfCSP (probe)⁺ GC B cells on day 28 after Pep22-KLH immunization to perform single-cell BCR sequencing. BCRs primarily contained paired iGL_L9 HC and LC sequences (78%, Fig. 3 A) or paired L9 HC and LC sequences (86%, Fig. 3 B), indicating that Pep22 activated target B cells. The HC mutation rate of the low-affinity iGL_L9 was significantly higher than that of high-affinity L9 (Fig. 3 C), suggesting greater potential for affinity maturation in iGL_L9. We compared post-Pep22-KLH immunization HC and LC mutations from PfCSP⁺iGL_L9 B cells with those in mature L9 to evaluate the induction of “on-track” mutations to residues in their mature equivalents. The HC of mature L9 carries six on-track mutations: T28I and S31T in HCDR1, Y53F and K58I in HCDR2, and Y96F and S99G in HCDR3 (Tripathi et al., 2023; Martin et al., 2023), and we detected high mutation frequencies (4.9–31.7%) at these positions (Fig. 3 D). HCDR1 is critical for the homotypic interaction characteristic of L9, and the key on-track mutations T28I (4.9%) and S31T (4.9%) were observed. The HCDR2 region contributes to antigen binding through W52 of VH3-33 and VH3-30, and nearby on-track mutations Y53F (18.3%) and K58I (7.3%) also emerged. The HCDR3 region plays a role in both antigen binding and homotypic interaction, and Y96F (6.1%) and S99G (1.2%) were detected (Fig. S3 A). The LC of mature L9 contains four on-track mutations: S28F and S31R in LCDR1, and Q90E and N92T in LCDR3 (Fig. S3 B). LCDR1 is essential for both antigen binding and homotypic interaction, with the on-track mutations S28F (5%) and S31R (67.6%) observed. In LCDR3, one of the two on-track mutations, N92T (8.8%), was detected (Fig. S3 C). The S31R mutation, the most critical for both antigen binding and homotypic interaction, was strongly induced by Pep22-KLH immunization (42% at day 14 vs. 9% from PfCSP) and became even more prevalent by day 28 (67.6%) (Fig. S3 D). Furthermore, phylogenetic analysis of iGL_L9 variants revealed significant diversification upon Pep22 immunization (Fig. S3 E). To assess whether the mutations induced by Pep22 immunization contributed to affinity maturation, we selected three iGL_L9 variants carrying on-track mutations and measured their approximate affinities using bead assays. Beads coated with each of the three variants exhibited stronger binding to the Pep22 probe compared with beads coated with iGL_L9 (Fig. S3, F and G), indicating that Pep22-KLH induced affinity maturation of the iGL_L9 B cells.

In contrast, despite the long and stable GC response in L9-transferred mice, the overall mutation frequency was very low, and conserved mutations were rarely observed in either HC or LC (Fig. 3, C and E; and Fig. S3 H). Although iGL_L9 AA mutation rates exceeded 20% at seven positions, the highest per-site rate in L9 reached only 17%, and most positions were <10%. L9 nonetheless diversified following Pep22-KLH immunization (Fig. S3 I). To assess whether this diversification translated into affinity maturation, we tested a subset of L9 variants using a bead assay. L9-25 showed reduced affinity, while most variants bound the Pep22 probe with comparable strength with the original L9 (Fig. S3 J). Only a few variants showed improved affinity, and these gains were modest relative to those observed in iGL_L9. Thus, while iGL_L9 variants consistently underwent affinity maturation, the L9 lineage showed a heterogeneous outcome: many variants maintained their original affinity, some lost binding strength, and only a small subset gained affinity—and these gains were smaller than those observed for iGL_L9.

To determine whether longer GC residency might allow for further maturation of high-affinity B cells, we sought to induce a longer GC response by an escalating dose immunization over seven injections from day 0 to day 12 (Tam et al., 2016) (Fig. 3 F). On day 28, the size of the GC was much larger in the escalating dose group (4.4%) than in the group that received a single priming immunization in bolus form (1.4%), likely due to the continuous exposure to antigen. However, the GC CD45.2⁺ fraction was lower in the escalated group (6.4%) than in the bolus group (15.2%), and as a result, there was no significant difference in the frequency of PfCSP⁺ CD45.2 GC B cells between the two groups (Fig. S3, K and L). However, class-switched IgG1⁺ cells had a modestly higher percentage of CD45.2 B cells in the escalating dose group (70.4%) than in the bolus group (61.5%) (Fig. S3 M). PfCSP⁺ CD45.2⁺ GC B cells isolated on day 28 showed no significant difference in mutation frequency between bolus and escalating dose groups (Fig. S3 N). Thus, escalating doses enhanced class-switching but not PfCSP⁺ CD45.2⁺ GC B cell residence or mutation rates.

To ensure antigen availability for GC maintenance, we administered a homologous bolus boost to both groups on day 30 and analyzed the GC response on day 58. GC B cells and CD45.2⁺ ratio in GCs did not differ significantly between post-boost groups (Fig. 3 G and Fig. S3 O). In contrast, the escalating dose + boost group exhibited a significantly higher PfCSP⁺ ratio among CD45.2⁺ GC B cells (42.3%) compared with the bolus + boost group (19%), resulting in the highest frequency of PfCSP⁺ CD45.2⁺ GC B cells among all groups (Fig. 3, G and H). When we isolated PfCSP⁺ CD45.2 GC B cells on day 58 from escalating dose + boost group and examined HC sequences, the mutation frequency was significantly higher on day 58 (HC: 4.29, LC: 2.77 AA mutations) than on day 28 (HC: 1.47, LC: 1.08 AA mutations) from the escalating immunization (Fig. 3, I and J; and Fig. S3 P). Conserved mutations and subclonal expansion were also much more prevalent (Fig. S3 Q). To determine whether the mutations affected affinity, we selected BCR sequences with subclonal expansion, expressed the Abs, and measured their apparent affinity to Pep22. In contrast to day 28 L9 variants, 70% (16 of 23) of the day 58 L9 variants displayed higher affinities to Pep22 than the original L9 (Fig. 3 K and Fig. S3 R).

The finding that some L9 variants gained or lost affinity, together with their underlying sequence diversification, suggested that the repertoire might display broadened reactivity to a wider range of epitopes. We therefore evaluated whether the diversified L9 Abs had acquired affinity for PfCSP variants differing from the 3D7 reference genome (Gardner et al., 2002). To do so, we designed a set of Pep22 variants (Zeeshan et al., 2012) with a range of mutations (Fig. 3 L) and compared the affinity of original L9 against post-bolus day 28 (Fig. 3 M) and escalating + boost day 58 GC B cell variants (Fig. 3 N) using a bead assay. The original L9 displayed similar affinities to Pep22 and three variants with mutations at the V110 or D111 positions (Pep22_Var3, Var4, and Var5) but had lower affinity to other Pep22 variants (Fig. 3 M). L9-08, L9-11, L9-32, and L9-56 displayed slightly higher affinity to Pep22 and all its variants relative to L9. Most L9 variant Abs from day 28 showed similar affinity trends to the original L9, with none displaying significantly superior affinities, though L9-25 exhibited significantly lower affinity (Fig. 3 M). A set of day 58 L9 variants, selected on the basis of higher affinity readings from BLI, were further examined. Most of these also showed higher affinity in the bead assay, and three L9 variants, including L9-B11, C11, and D11, exhibited higher affinity to Pep22 Var2, Var6, and Var7 than did L9 (Fig. 3 N). Thus, high-affinity B cells underwent further affinity maturation when the GC response was extended by boosting, and that prolonged GC reaction enhanced breadth as well as affinity.

As the elicitation of further affinity maturation would likely be more challenging with a higher affinity starting point, to delve deeper into whether Pep22 immunization enhanced the quality of Ab responses, we performed an informatics analysis of 200 Ab sequences isolated from the post-bolus day 28 mature L9 transfer group. This analysis aimed to identify top sequence variants by assessing seven potential features that might correlate with protection, including (1) % identity to the mature L9 Ab, (2) the number of Ab contacts with PfCSP based on modeling of the L9-PfCSP structure (Tripathi et al., 2023), (3) the number of Ab homotypic contacts also based on modeling, (4) the number of AA mutations, (5) the number of modeled van der Waals clashes, (6) the ratio of silent versus AA mutations, and (7) the predicted binding energy (Fig. 4 A). The sieving process identified 59 variant L9-Ab sequences as the top sequence variants, of which 42 could be expressed and purified (Fig. 4 A). We assessed the effectiveness of these 42 Abs in reducing liver burden following an intravenous challenge with transgenic Plasmodium berghei SPZ expressing PfCSP and green GFP-luciferase (GFP-Luc) (Flores-Garcia et al., 2019). Based on the quantitatively assessed reduction in liver burden by the different Ab variants, we selected those that showed the greatest reduction for in-depth structural analysis. This approach allowed us to identify sequence features that correlate with enhanced protective efficacy. Among the 42 Abs, at 50 µg per mouse, ∼14 Abs showed comparable or slightly improved reduction of liver burden relative to the mature WT L9 Ab, indicating highly protective activity (Fig. 4 B, left, and Fig. S4 A). To refine our selection and identify the most promising candidates, we reassessed 24 of these variants at a reduced dose of 25 µg per mouse, separated into three groups (A, B, and C) of eight Abs each, along with the control Ab L9 expressed in EXPI (293) and CHO cells. This secondary screen helped narrow the pool to a subset of Abs, including L9-11 and L9-32, both of which often ranked among those that conferred the greatest reduction in liver burden (Fig. 4 B, right). While protective efficacy was generally correlated between the two doses (Fig. S4 B), distinguishing fine differences among highly potent Abs would require assessment with a substantially higher number of mice to overcome intrinsic variation between animals. Correlations of affinity versus protection revealed that Ab affinity to PfCSP, as measured by BLI, did not correlate significantly; however, higher affinity to peptide 22 (two NPNV repeats) or peptide 21–26 (3 NPNV repeats) did correlate significantly with protection (Fig. 4 C). Moreover, when PfCSP affinities were measured by isothermal titration calorimetry (ITC), multiple binding events were observed, generally governed by at least two dissociation constants (Fig. S4 C). When we assessed correlations with both of these, we observed the affinity associated with the second dissociation constant (KD2) to correlate significantly with protection (Fig. 4 D), indicating an interplay between affinity and protection. Furthermore, in terms of the seven sequence-based features, significant correlations were observed with only two features: mutations on Ab-PfCSP contacts and mutations at homotypic-contact sites, both of which were negatively correlated with improved protection; such negative correlation suggested SHM in these regions to lead to reduced protection (Fig. 4 E), likely because both PfCSP contacts and homotypic contacts have already been substantially optimized by SHM in the mature L9 Ab.

After identifying key correlates of protection through informatics screening and affinity analyses, we sought to understand the structural basis of these findings. L9 recognizes a conformational epitope on PfCSP and displays homotypic interactions (Martin et al., 2023; Tripathi et al., 2023), indicating a trimeric binding modality. To examine how specific mutations affect binding and contribute to enhanced protection, we determined cryo-EM structures of antigen-binding fragments (Fabs) of selected L9 variants in complex with PfCSP (Fig. 5 A). In particular, we selected two variants from day 28, L9-11 and L9-21, as L9-11 often trended toward improved protection, and the BLI affinity for L9-21 was two orders of magnitude higher than that of WT L9. We also selected two variants from day 58, L9-F4 and L9-L4, as these had been observed to be particularly expanded on day 58 of the GC in independent mice and showed improved apparent affinity for Pep22 (Fig. S3 R).

We determined cryo-EM structures of L9-11, L9-21, L9-F4, and L9-L4 in complex with PfCSP, and pairwise comparisons with the original L9-CSP structure show that the overall backbone is highly conserved (Cα RMSD ∼0.7–0.8 Å), indicating no substantial global rearrangements. The structure of Fab L9-11 with PfCSP by single-particle cryo-EM was determined to a resolution of 3.5 Å (Fig. 5 B, far left panel, and Fig. S5). There were only three mutations of L9-11 relative to WT-L9: S56G_L, T85S_L, and S30G_H (for clarity, we have added a subscript, H or L, defining whether the mutation is on the HC or LC) (Fig. 5 B, far left panel); all of these mutations reduced the size of the Ab, potentially enabling better induced fit. L9-21 had eight mutations relative to WT-L9 (Q27H_L, S56N_L, S76T_L, S25A_H, S30T_H, S55N_H, N56K_H, and M80L_H), and we determined its structure in complex with PfCSP to a resolution of 4.0 Å (Fig. 5 B, second panel from left, and Fig. S5). Three of these mutations were located at the homotypic interface between Fabs (two from HC and one from LC), and two HC mutations were closer to the PfCSP epitope. In particular, the N57K_H mutation in CDRH2 interacted with Y94_L via a hydrogen bond, and Y94_L had a critical CH-π interaction with P116_CSP (Fig. 5 B second panel).

The Ab L9-F4 has three mutations (M34I_H, M80L_H, and G99R_H), with G99R_H on CDRH3 close to PfCSP. We determined the structure at a resolution of 3.8 Å (Fig. 5 B, third panel from left, and Fig. S5). Structural analysis revealed a side-chain hydrogen bond between G99R_H and the backbone carbonyl of PfCSP, explaining the improved affinity relative to WT-L9 (Fig. 5 B, third panel). Ab L9-L4 has seven mutations (S76T_L, T31N_H, V50A_H, I57T_H, N76T_H, Q81E_H, and A84V_H), with one mutation at the homotypic interface. The structure obtained at a resolution of 4.0 Å (Fig. 5 B, far right panel, and Fig. S5) revealed that T31N_H on CDRH1 formed a network of hydrogen bonds with R94_H and E68_L. T31N_H also formed a critical NH-π interaction with F28_L. As both F28_L and E68_L have been shown to be critical for the stability of the homotypic interface (Tripathi et al., 2023), the introduction of the T31N_H mutation likely further enhances the homotypic interface (Fig. 5 B, far right panel).

Overall, the cryo-EM structures of L9-11, L9-21, L9-F4, and L9-L4 in complex with PfCSP visualized critical mutations leading to improved affinity, as measured by ITC and BLI (Fig. 5 C).

As the peptide affinity of L9 variant Abs did correlate with improved malaria protection, we tested combinations of changes from each of the separate Abs to see if we could further improve both peptide affinity and malaria protective efficacy. As we observed anticorrelations with mutations at peptide contacts or at homotypic Fab contacts, which suggested that SHM at these positions would lead to reduced protection, we assessed combinations with only day 28 variants. Lastly, when examining the kinetics of peptide recognition, we observed variable on- and off-rates. Notably, no natural variants derived from sequencing of B cells ex vivo exhibited both a faster on-rate and a slower off-rate to Pep22 relative to L9 (Fig. 6 A). We chose to focus on making changes to the three variants from day 28, which consistently showed the highest level of malaria protection (L9-11 and L9-32) or had the fastest on-rate (L9-21).

The initial designs utilized L9-11 and L9-32 as templates, and shared mutations observed within the top 15 Abs were added on top of the template Abs. These combination Abs were expressed (Table S1) and assessed for in vivo efficacy. None of these combinations improved the protective efficacy compared with the template Abs (Fig. S4 D). The second set of combinations based on L9-21 was expressed (Fig. 6 B and Table S1), and the affinity to Pep22 was assessed by BLI. From this set, three combinations with L9-21 yielded 4.3–9.6-fold improved affinity (L9-com13, 15, and 16).

While L9-com15 had the highest Pep22 affinity (9.6-fold improved over L9), this was mostly due to an improved on-rate. L9-com16 had more balanced on- and off-rates (Fig. 6 A, middle panel) and was chosen as the template for the next round of optimization. We added additional mutations to L9-com16, identifying two variants, L9-com20 and L9-com23, both with over sixfold improved Pep22 affinity. The analysis of L9-com23 by ITC revealed it to have substantially high affinity (4.3-fold improved KD1, but a similar KD2) to WT L9 (Fig. 6 C). Of these, L9-com23, when assessed at 25 µg/mouse, showed a significant trend toward improved malaria protection relative to WT L9 (Fig. 6 D). However, when L9 and L9-com23 were independently compared directly, this improvement in liver burden was not significantly better than L9 (Fig. 6 E).

Overall, analysis of Ab kinetics led to the design of L9 variants with significantly improved affinity (∼10-fold for L9-com15 and ∼ 6.4–6.5-fold for L9-com23 and L9-com20). However, these improvements did not translate into significantly enhanced in vivo protection, suggesting that higher affinity alone does not significantly enhance protection, which highlights the difficulty in demonstrating further improvements among already highly protective Abs in this model.

Assessing the responses of diverse anti-CSP B cell precursors within a single host better reflects the complex competitive environment of human physiology. To investigate the diversity of anti-CSP epitope responses following immunization, we adoptively co-transferred into WT mice 4 distinct B cell precursors specific for different PfCSP regions—two anti-major repeat clones (iGL_311 and iGL_317), an anti-junctional clone (iGL_CIS43), and an anti-minor repeat clone (iGL_L9), thereby generating a mini-repertoire of human antimalaria precursors. Recipients were then immunized with either (1) full-length PfCSP, (2) the truncated PfCSP (R21-epitope) defined above, or (3) a combination of truncated PfCSP (R21-epitope) with junctional region (NPDP19-KLH) and minor repeat region (Pep22-KLH) peptide immunogens, using Alhydrogel as the adjuvant (Fig. 7 A).

At day 8, GC frequencies were higher in the truncated PfCSP + peptides group (6.3%) compared with truncated PfCSP (2.3%) alone or full-length PfCSP (1.2%). Recruitment of CD45.2⁺ B cells into the GC was highest in the truncated PfCSP group (75%), intermediate in the full-length PfCSP group (40%), and lowest in the truncated PfCSP (R21-epitope) + peptides group (8.4%). As a result, the truncated PfCSP (R21-epitope) group showed a threefold higher frequency of CD45.2⁺ GC B cells (1.8%) compared with the full-length PfCSP (0.5%) or truncated PfCSP + peptides group (0.55%) (Fig. 7, B and C). Thus, truncated PfCSP, with or without the targeted junctional and minor repeat peptides, induced a stronger early GC response than full-length PfCSP from transferred B cells encompassing the major, minor, and junctional iGLs. At day 22, GC B cells remained highest in the truncated PfCSP + peptides group (1.7%), followed by truncated PfCSP (0.75%) and full-length PfCSP (0.65%). However, CD45.2⁺ GC B cells in the truncated PfCSP group were close to undetectable, and only two out of four mice in the full-length PfCSP group had CD45.2⁺ GC B cells. In contrast, the truncated PfCSP + peptides group maintained a markedly higher GC proportion (9.23%) and exhibited the highest frequency of total CD45.2⁺ GC B cells among all groups (0.16%) (Fig. 7, D and E).

To determine which clones were recruited into the GC following immunization, we analyzed GC composition by single-cell sorting of CD45.2⁺ GC B cells at both day 8 and day 22. Despite full-length PfCSP containing junctional and minor repeat regions, no iGL_CIS43 cells were detected from that group, and few iGL_L9 cells were observed (2/894 on day 8 and 1/270 on day 22). GC responses were instead dominated by the two anti-major repeat clones, iGL_311 and iGL_317 (Fig. 7 F, left). Truncated PfCSP exclusively induced iGL_311 responses at day 8 and yielded no sequences at day 22 due to the scarcity of CD45.2⁺ GC B cells (Fig. 7 F, center). In contrast, truncated PfCSP + peptide-KLH immunogens to the minor and junctional regions activated all three epitope-specific B cell populations. During the early GC phase, anti-major and anti-minor repeat clones dominated, whereas by the late phase, anti-junctional and anti-major clones were more prevalent (Fig. 7 F, right).

To investigate the affinity maturation of each precursor, we analyzed their mutations and compared them with the corresponding mature Ab sequences to identify on-track mutations from CD45.2⁺ GC B cells in the truncated PfCSP + peptides immunization group at day 22. In iGL_311, a significant S31N mutation in HCDR1—critical for its homotypic interaction—was observed. Additional on-track mutations, including Y59F near HCDR2 and S100T in HCDR3, were also frequently induced. In iGL_CIS43, on-track mutations such as N52K and K59R in HCDR2—key residues for binding to the junctional epitope—were induced at 6.8 and 64%, respectively, along with M34I in HCDR1 and V98L in HCDR3. This mutation pattern closely mirrored our previous findings for iGL_CIS43 (Kratochvil et al., 2021). In iGL_L9, on-track mutations including T28I and S31T in HCDR1, Y53F in HCDR2, and Y96F in HCDR3 were induced, consistent with results from single-clone transfer experiments (Fig. 7 G). Collectively, these data confirm that upon truncated PfCSP + peptide-KLH immunization, each clone acquired on-track mutations supporting affinity maturation toward their corresponding mature Abs despite a competitive co-adoptive transfer environment.

Next, to assess the polyclonal response, including endogenous host responses, we performed serum ELISA analysis using day 22 samples. NANP18, NPDP19, and Pep22 peptides were used to detect anti-major repeat, anti-junctional, and anti-minor repeat responses, respectively. Serum from mice immunized with full-length PfCSP elicited stronger anti-major repeat Ab responses than the truncated PfCSP + peptides group; truncated PfCSP alone fell between these groups. However, anti-junctional and anti-minor repeat responses were significantly higher in the truncated PfCSP + peptides group (Fig. 7 H). In sum, the addition of peptide-KLH immunogens improved junctional and minor repeat Ab responses.

Finally, we tested whether the anti-junctional and anti-minor repeat epitope Abs elicited by the peptide-KLHs could confer improved protection. The current R21/Matrix-M immunization regimen is three doses given 1 mo apart followed by a fourth dose 12 mo after the third. To mimic this longer interval between prime and boost, adoptively co-transferred mice were primed and then boosted at 4–5 mo (Fig. 7 I).

We first compared serum IgG titers before and after the boost. In adoptive transfer recipients, immunization by truncated PfCSP + peptides did not diminish anti-major responses relative to immunization by PfCSP or truncated CSP alone; PfCSP + peptide immunization did, however, significantly increase anti-junction and anti-minor repeat titers; these responses, notably, were observed 3 days pre-boost, 4–5 mo after the initial prime, indicating a durable response to the prime. All titers were elevated after each boost (Fig. 7 J). Immunized mice were then challenged with live P. berghei expressing PfCSP 10 days post-boost, and liver burden was assessed 2 days after challenge (Fig. 7 I). Despite the anti-major IgG titers elicited by all immunizations, significant protective efficacy was observed only in the truncated CSP + peptides group; no significant reduction in liver burden was obtained by any other immunization protocol (Fig. 7 K). Thus, the addition of immunofocusing peptides broadened the Ab response and provided antimalarial protection in vivo.

The recently approved malaria vaccines R21 and RTS,S have shown protective efficacy in infants and young children. Both vaccines mediate protection by inducing Abs against the major NANP repeat epitopes of PfCSP. High anti-NANP titers are required, and protective efficacy is limited relative to other childhood vaccines. Moreover, protection is not durable (Miura et al., 2024). Thus, expanding the breadth of the humoral repertoire to other neutralizing epitopes may be key to effective and long-lived malaria vaccines. Here, in murine models with B cells expressing human BCR precursors to highly protective Abs, we confirmed that an immunogen containing the truncated PfCSP epitopes presented by R21 cannot effectively activate antimalaria B cell precursors to anti-minor repeat or junctional region Abs. Furthermore, the immune response to full-length PfCSP, which includes junctional and minor epitopes, was predominantly directed toward major repeats and was inadequate to elicit stable anti-minor GC responses. An immunofocusing strategy using epitope-specific short peptides, however, induced robust GC responses from both low- and high-affinity anti-minor repeat (L9) precursors. Furthermore, this approach facilitated the affinity maturation of high-affinity anti-minor repeat B cells—suggesting that we may achieve improved protective efficacy even in malaria-experienced populations and allowing us to assess the sequence and structural correlates of anti-minor repeat Ab protection. Finally, by combining the R21-epitope truncated PfCSP with minor and junctional epitope peptide-KLH immunogens, we elicited a balanced, protective humoral response in vivo, suggesting that targeting additional epitopes may enhance current malaria vaccines.

Traditionally, antigen immunogenicity has been evaluated using WT mice. However, murine immune repertoires may not accurately represent those of humans, necessitating BCR-transgenic models. Building upon our previous investigation into the PfCSP junctional epitope (Kratochvil et al., 2021), this study has advanced our preclinical mouse platform by incorporating human Ab precursors of additional PfCSP specificities. Genetically modified mice in which B cells bear human BCRs have been instrumental in HIV vaccinology, facilitating several successful phase 1 clinical trials. The efficacy of the BG505 SOSIP.v4.1-GT1.1 immunogen in phase 1 (Caniels et al., 2025) corresponded well with the prediction of immunogenicity of GT1 variants obtained in genetically modified mice (Caniels et al., 2023; Caniels et al., 2024; Medina-Ramírez et al., 2017). Similarly, the ability of the eOD-GT8 60mer immunogen to activate rare B cell precursors in humans (Leggat et al., 2022) was foreshadowed by its performance in Ig KI mice similar to those used here (Abbott et al., 2018; Dosenovic et al., 2015; Huang et al., 2020; Jardine et al., 2015). The presence of human-derived HC and LC sequences allows for tracking not only B cell activation and differentiation but also the acquisition of specific mutations associated with affinity and breadth. The lines developed here and previously by our group (Kratochvil et al., 2021) allow for the expansion of these genetic technologies to malaria therapeutic and vaccine development. While some results from our new PfCSP mouse models were consistent with observed affinities, Ab-antigen affinity measurements alone were not fully predictive of in vivo outcomes: of note, iGL_311 has higher affinity than iGL_317 to the major repeats, minor repeats, and the junctional region, yet iGL_317 activation after full-length PfCSP immunization was far stronger. Structural features of the BCR–antigen interaction on the B cell surface—including receptor clustering and signaling threshold—may differ between the two clones in ways not captured by solution-phase BLI measurements. These observations, and the collective body of work, underscore the importance of in vivo mouse models with human-derived BCR sequences for immunogenicity studies.

Congenic adoptive transfer from models with pre-rearranged B cells bearing BCRs of known specificities enables the establishment of calibrated precursor frequencies within an intact murine repertoire. Alternative approaches, such as recombining mouse models (Tian et al., 2016) and Kymab mice (Sok et al., 2016), are also available and have contributed to the development of candidate HIV immunogens. However, early recombining mice had an unphysiologically high number of relevant precursors, while Kymab mice averaged only 1.3 precursors per animal for eOD-GT8, thus underestimating its later performance. More recent rearranging models have thus been designed to produce more physiological frequencies (Luo et al., 2023). As precursor frequency is a crucial determinant of immunogenicity (Abbott et al., 2018; Dosenovic et al., 2018), the ability to precisely titrate frequencies is critical to replicate human physiology. Although the precise frequencies of anti-major, anti-minor, and anti-junction precursors have not yet been determined in the human population, the extremely low frequencies we deployed here make the new anti-major and anti-minor repeat mouse models described here quite stringent.

The flexibility offered by adoptive transfer from multiple KI lines into a single congenic host enabled the examination of the relationship between protective efficacy and combined B cell immune responses. Although both full-length PfCSP and truncated PfCSP induced activation of anti-major iGL_311 and iGL_317 precursors, these responses did not translate into significant protective efficacy. The avidity of the genuine R21 vaccine presentation, as opposed to the monomer here, may affect this response. Alternatively, the lack of protection may be due to the timeline: the 4+ month gap after the prime may have allowed anti-major titers to wane too substantially to be protective on their own. Finally, as the skin is the locus of activity for some anti-central–repeat Abs (Aliprandini et al., 2018), mosquito-bite models of protection may have a lower protective threshold for similar Abs. However, it is clear that after defined IV delivery, achieving robust protection required inclusion of anti-junction (iGL_CIS43) and anti-minor (iGL_L9) responses. In addition, anti-junction and anti-minor IgG titers were durable and were significantly boosted 4–5 mo after priming, suggesting that peptide-KLH immunogens can elicit substantial long-lived plasma cell and memory B cell responses even after a single immunization. Together, these findings demonstrate that our model is a robust platform for evaluating the protective potential of malaria vaccine candidates and identifying B cell clones that are functionally associated with protection.

As several potent Abs against the junctional and minor repeat regions have been reported, and as B cells targeting the same epitope may compete and limit antigen accessibility (Dvorscek et al., 2024; Schaefer-Babajew et al., 2023; Tas et al., 2022), eliciting neutralizing Abs against diverse PfCSP epitopes may improve protective efficacy. Although full-length PfCSP contains epitopes capable of activating L9 or CIS43 precursors, it failed to sustain a stable GC response. Truncated PfCSP (R21-epitope) also failed to elicit these protective Abs at all, although the responding B cell repertoire in actual vaccination may differ due to avidity effects, since, in contrast to the R21-epitope monomeric truncated PfCSP used in this study, R21 is a HBsAg-linked multimeric immunogen (Collins et al., 2017). In contrast, our epitope-specific peptide-KLH multimer immunogen design increases the molecular abundance of junctional and minor repeat epitopes relative to PfCSP, reducing competitive suppression. A single immunization with peptide-KLH immunogens alongside the R21-epitope truncated PfCSP was sufficient to elicit a significant GC response and affinity maturation in anti-junction and anti-minor B cells, while maintaining a response to the major repeats. Recent reports have also described vaccine designs specifically targeting anti-junctional and anti-minor repeat B cells (Francica et al., 2021; Jelínková et al., 2021; Jelínková et al., 2022; Langowski et al., 2025; Ludwig et al., 2023; Tripathi et al., 2026), collectively highlighting the importance of eliciting responses against these regions. The use of a Pep22-KLH boost also enhanced anti-minor repeat responses in our system. Abs generated during the primary immune response to PfCSP have been observed to mask the major repeat region during subsequent responses (McNamara et al., 2020); because the junctional and minor repeat regions also contain NANP sequences, the primary response may also interfere with anti-junctional and anti-minor repeat responses upon secondary immunization. Consequently, extended or repeat exposures to PfCSP may be required to expand past the major repeat–biased response; the supraphysiological doses of irradiated PfSPZ that successfully isolated these anti-minor and anti-major Abs may fall in this category (Kisalu et al., 2018; Wang et al., 2020). The application of immunofocusing to boost immunogen cocktails may also avert this major repeat bias.

Vaccination is not the only mechanism of antimalarial prophylaxis, as evinced by the clinical trials of L9LS and CIS43LS (Lyke et al., 2023; Wu et al., 2022; Kayentao et al., 2022; Kayentao et al., 2024; Gaudinski et al., 2021). We anticipated that variants with higher affinity than mature L9 would exhibit greater in vivo potency and that a significant decrease in affinity (compared with L9) would be associated with a clear reduction in in vivo potency; indeed, most of the variants we identified with affinity two to four times higher than L9 exhibited similar levels of in vivo potency. However, the correlations we observed between affinity and protection appeared to hold only when assessed over a large range. Thus, while lower affinity than that of L9 often showed reduced protection, higher affinities did not guarantee increased protection in vivo. Instead, the specific impact of the variant mutation on PfCSP binding in the context of the SPZ is likely more critical for in vivo efficacy; this aligns with previous findings where higher affinity did not correlate with increased in vivo potency for major repeat Abs that exhibit high in vivo potency (Williams et al., 2024), although improved efficacy could result from carefully selected junctional region Abs (Banach et al., 2022; Reveiz et al., 2025). With the L9 variants, we analyzed on-rate/off-rates to see if we could gain insight into how affinity related to improved protection. However, the L9 variants with significantly higher affinity did not display significantly higher protective efficacy than L9. Therefore, further improvement would require not only research on the relationship between affinity and protection, but also the structural changes induced by Abs in PfCSP, and their relationship with liver cell invasion and attachment.

Cross-reactivity against the major repeat, minor repeat, and junctional region may be a key factor for protective efficacy. The current RTS,S and R21 vaccines were designed based on the finding that Abs targeting the major repeat region exhibit protective efficacy, and the level of anti-major repeat Abs in the serum has shown correlation with clinical efficacy (Datoo et al., 2024). However, Abs recognizing both the major repeat and either the minor repeat or junctional regions have higher protective efficacy than those recognizing only the major repeat (Murugan et al., 2020). Notably, R21 vaccination has been shown to elicit Abs that cross-react with the junctional region despite the absence of this epitope in the immunogen, suggesting that recipients of the R21 vaccine retain the capacity to mount high-affinity responses against non-targeted epitopes (McDaniel et al., 2025). Consistent with this broader recognition potential, L9, as observed through ITC analysis, exhibited two-step binding to PfCSP, indicating an affinity for both the minor and major repeat regions. F10, an Ab that shares a common ancestor with L9, shows a lower affinity for the major repeat region and significantly lower in vivo protection efficacy (Wang et al., 2022). The mechanism underlying the superior efficacy of Abs targeting both regions remains unclear; however, the affinity for the major repeat region likely provides a strong binding effect owing to the avidity of the repeated sequences, facilitating PfCSP recognition. Just ahead of the junctional region is a region rich in positively charged AA, which plays a crucial role in heparan sulfate proteoglycan (HSPG) binding and is expected to be important for the role of PfCSP in liver invasion (Ancsin and Kisilevsky, 2004; Coppi et al., 2007; Frevert et al., 1993); Abs against the junctional or minor repeat regions may therefore interfere with HSPG binding, thereby preventing liver invasion, although this requires experimental validation.

The addition of short peptide-based immunofocusing immunogens may enhance the suboptimal vaccine efficacy of PfCSP epitopes included in R21 by activating a broader antimalaria response. In a previous study, immunization with R21 followed by passive transfer of anti-junctional or anti-minor repeat Abs exhibited added efficacy in an in vivo challenge model (Wang et al., 2021a) and a recent study demonstrated that tandem delivery of a CIS43-based junctional vaccine in mice improved protection at low doses of R21 (Tripathi et al., 2026); we might therefore expect synergy between R21 and immunofocusing immunogens. Given the variability of the human B cell clonal landscape, controlled human malaria infection studies would be the next logical step to validate these findings. Overall, this study demonstrates that including the junctional and minor repeat PfCSP epitopes, potentially in the form of immunofocusing peptides, may enhance the current generation of malaria vaccines.

iGL_L9, L9, iGL_311, and iGL_317 BCR-KI (both HC and LC) mice were generated at the animal facility of the Gene Modification Facility (Harvard University) using methods we previously published (Lin et al., 2018; Wang et al., 2021b). Subsequent colony breeding and experiments were conducted at the Ragon Institute of Mass General Brigham, MIT, and Harvard. Seven-to nine-week-old male B6.SJL-Ptprca Pepcb/BoyJ CD45.1 host mice (RRID:IMSR_JAX:002014) were purchased from The Jackson Laboratory, acclimated for 1 wk, and then used for adoptive transfer experiments. All animal experiments were conducted in accordance with protocols (2016N000286 and 2016N000022) approved by the Institutional Animal Care and Use Committee (IACUC) of Harvard University and the Massachusetts General Hospital, an Association for Assessment and Accreditation of Laboratory Animal Care (AAALAC) international accredited facility.

B cells were isolated from CD45.2 BCR KI mice using the Pan B Cell Isolation Kit II (Miltenyi Biotec). The isolated B cells were stained with probes and Abs, and precursor frequency was calculated using a BD FACSymphony S6. The B cells were then diluted in PBS to a volume of 200 μl and adoptively transferred into B6.SJL-Ptprca Pepcb/BoyJ host mice (CD45.1) via tail vein injection. One day later, the mice were immunized. Sample sizes were derived from prior work with this model type (Kratochvil et al., 2021); no mice were excluded from analysis.

Protein or Peptide-KLH antigen (GenScript), 10 mg, was diluted in 100 μl of PBS and mixed with 2% Alhydrogel (InvivoGen) in a 1:1 volume ratio. This mixture was incubated on a rotator for 30 min and then administered intraperitoneally to mice (10 µg in 200 μl per mouse). All animal experiments were conducted in accordance with the approved protocols by the IACUC of Harvard University and Massachusetts General Hospital.

Individual spleens were ground on a 70-µm filter to isolate single cells, followed by the removal of RBCs using ACK lysis buffer. The RBC-free cells were then incubated with Live/Dead Blue and FC block, diluted in PBS containing 2% FBS (FACS buffer), at 4°C for 10 min and subsequently washed. The biotinylated probe was first incubated with fluorochrome-streptavidin at a 4:1 M ratio at 4°C for 20 min, then diluted in FACS buffer to a probe concentration of 50 nM and incubated with the cells at 4°C for 30 min before washing. For surface staining, Abs were diluted in FACS buffer at a ratio of 1:100 to 1:400 and incubated with the cells at 4°C for 30 min, followed by two washes. The prepared cells were analyzed or sorted using a BD FACSymphony S6. Single cells were dry-sorted into a 96-well PCR plate, quickly frozen on dry ice, and stored in a −80°C deep freezer for future sequencing. FACS data were analyzed using FlowJo10.

First, cDNA synthesis was performed using SuperScript III Reverse Transcriptase (Invitrogen) according to the manufacturer’s instructions. Subsequently, HC and LC PCRs were conducted separately, each performed twice in a nested format. The PCR was carried out using 10× PCR buffer (QIAGEN), HotStarTaq enzyme (QIAGEN), 10 mM dNTPs (Thermo Fisher Scientific), and a cocktail of IgG or IgK-specific primer sets (von Boehmer et al., 2016) diluted in water. The PCR products were electrophoresed on E-Gels 96 2% with SYBR Safe (Thermo Fisher Scientific), and the PCR bands were visualized using a ChemiDoc imaging system (Bio-Rad). Following this, the PCR plate was sent to Genewiz, where Sanger sequencing was performed using the HC reverse primer from PCR-2 (5′-GCTCAGGGAARTAGCCCTTGAC-3′) and the LC reverse primer (5′-TGG​GAA​GAT​GGA​TAC​AGT​T-3′) to confirm the BCR sequence.

For the mixed adoptive transfer model, single-cell RNA sequencing was performed by generating NGS libraries from sorted cells with the Chromium Next GEM Single Cell 5′ Reagent Kit v2 (10x Genomics), following the manufacturer’s protocol. The libraries were pooled and sequenced on a NextSeq 2000 instrument (Illumina). Sequencing reads were processed and analyzed using Cell Ranger software (version 8, 10x Genomics) with a customized reference database. Sequencing data were excluded if they failed standard quality thresholds, such as multiple peaks in Sanger sequencing.

96-well plates were coated with 25 μl/well of PfCSP (1 µg/ml) diluted in PBS, and incubated overnight at 4°C. The next day, the coating solution was discarded, and plates were washed three times with 0.05% Tween 20 in PBS (tPBS). 100 μl/well of blocking buffer (tPBS with 3% BSA) was added prior to incubation for 1 h at room temperature on an orbital shaker. The blocking solution was discarded, and the plates were washed three times with tPBS. Mouse serum diluted (1:150) in tPBS with 0.5% BSA was then added (60 μl/well), and a threefold serial dilution was performed prior to incubation overnight at 4°C. The next day, the unreacted solution was discarded, and plates were washed three times with tPBS. Secondary AP-conjugated anti-mouse IgG Ab (Cat# 115-055-003; Jackson ImmunoResearch Labs; RRID: AB_2338528) diluted 1:1,000 in tPBS with 0.5% BSA was added (50 μl/well) and plates incubated for 1 h at room temperature on an orbital shaker. Unreacted Ab solution was discarded, and plates washed five times with 100 μl tPBS and once with dH₂O. pNPP dissolved in dH2O was added (50 μl/well) and plates incubated for 30 min at room temperature. The reaction was stopped by adding 50 μl of 3N NaOH. Optical density was measured at 450 nm using a BioTek Synergy Neo2 microplate reader (Agilent).

Protein A beads (10 μl) were diluted in 1 ml of PBS and washed. Then, 1 µg of Ab was diluted in PBS to a total volume of 50 μl and incubated with the beads for 30 min at room temperature to capture the Ab. Afterward, 0.5 µg of rat IgG2c isotype control Ab (Cat# 400701; BioLegend; RRID:AB_326567) was added and incubated for 10 min at room temperature to block nonspecific binding. The beads were washed with PBS, then incubated with 10 nM of the prepared probe for 30 min at room temperature to allow probe binding. Following two washes, the probe binding intensity of the Ab-captured beads was analyzed using a BD FACSymphony S6. FACS data were analyzed using FlowJo10.

Ab variable HC and LC sequences were codon optimized, synthesized, and cloned into a VRC8400 (CMV/R expression vector)-based IgG1 vector as previously described (Kong et al., 2019). The variants were expressed by transient transfection in Expi293 cells (Thermo Fisher Scientific) using Turbo293 transfection reagent (SPEED BioSystems) according to the manufacturer’s recommendation. 50 μg of plasmid encoding HC and 50 μg of plasmid encoding LC variant genes were mixed with the transfection reagents, added to 100 ml of cells at 2.5 × 10⁶/ml, and incubated in a shaker incubator at 120 rpm, 37°C, 9% CO₂. At 5 days after transfection, cell culture supernatant was harvested and purified with a Protein A (GE Healthcare) column. The Ab was eluted using IgG Elution Buffer (Thermo Fisher Scientific) and were brought to neutral pH with 1 M Tris-HCl, pH 8.0. Eluted Abs were dialyzed against PBS overnight before use.

The sample for cryo-EM was prepared by mixing Fab to PfCSPm at 2:1 M ratio. The complex was incubated overnight at 4°C and flash frozen until ready for use. To prepare grids of the complex, 3 μl of protein sample was applied to freshly glow-discharged (easiGlow) C-flat grids (Protochips, CF1.2/1.3-3Au). Blotting was done using a Vitrobot Mark IV (Thermo-Fisher), with a 5-s blotting time and 8 pN blotting force at 6°C in 100% humidity. Grids were vitrified by plunging into liquid ethane and stored in liquid nitrogen before examination by cryo-EM. Images were recorded on a Glacios TEM (Thermo Fisher Scientific) at 200 kV and recorded at 36,000× magnification with a defocus range of −0.3 to −2.2 µm on K3 direct electron detector (Gatan) in super-resolution mode.

Motion correction, contrast transfer function estimation, particle picking, extraction, 2D classification, ab initio model generation, 3D refinements, and local resolution estimation were carried out in cryoSPARC 3.3.1 (RRID:SCR_016501) (Punjani et al., 2017). The 3D reconstructions were performed using C1 symmetry for both classes. The coordinates of the L9-PfCSP structure, PDB entry (8EK1), were employed as an initial model for fitting the sharpened cryo-EM map of the L9-PfCSP structures (Table S2). Manual and automated model building were iteratively performed using Coot (RRID:SCR_014222) (Emsley and Cowtan, 2004) and real space refinement in Phenix (Adams et al., 2010) to accurately fit the coordinates to the electron density map. MolProbity (RRID:SCR_014226) (Davis et al., 2004) and EMRinger (Barad et al., 2015) were used to validate geometry and check structure quality. UCSF ChimeraX (Goddard et al., 2018) was used for map-fitting cross-correlation calculation (Fit-in-Map tool) and for figure preparation.

Variants were designed iteratively based on BLI measurements of association (kon) and dissociation (koff) rates against PfCSP and peptide antigens. Initial L9 variants (including day 28 isolates) were profiled to identify mutations associated with slower dissociation and/or faster association. Mutations showing anticorrelation with protection, particularly at peptide-contact or homotypic Fab–Fab interface positions, were deprioritized. In the first round, mutations enriched among higher-performing variants were recombined onto selected templates (L9-11, L9-32, and L9-21) to generate combination designs. In the second round, L9-com16, selected based on balanced kinetic improvements, was used as the template for further optimization by introducing additional mutations associated with favorable kinetic profiles. All variants were expressed and screened by BLI, and those with improved kinetics were advanced for further characterization.

Ab binding affinity to various ligands was measured using BLI on an Octet Red384 instrument (FortéBio) with streptavidin capture biosensors (FortéBio) in solid black tilt-well 96-well plates (Geiger Bio-One). Assays were performed with agitation at 30°C. Biotinylated peptides were diluted to 0.1 μg/ml in (PBS + 1% BSA) and were immobilized for 3 s to reduce the density (<0.01 nm) on the biosensor to avoid avidity. This step was followed by a 60-s baseline in buffer (PBS + 1% BSA). Association with IgG (serially diluted from 1,000 to 1.3 nM was done for 240 s, followed by a dissociation step in buffer for 1,200 s. In all Octet measurements, parallel correction to subtract systematic baseline drift was carried out by subtracting the measurements recorded for a loaded sensor incubated in PBS. Data analysis was carried out using Octet software, version 9.0. Experimental data were fitted globally with a 1:1 Langmuir model of binding for all the antigens.

Female C57Bl/6 mice (RRID: IMSR_JAX:000664) of 6–8 wk of age were purchased from The Jackson Laboratories. All animals were maintained and cared for in accordance with AAALAC Standards. All studies were approved by the Animal Care and Use Committee at Johns Hopkins University (JHU), protocol numbers MO21H417 and MO24H329.

Transgenic SPZ in P. berghei expressing PfCSP, GFP, and luciferase reporter gene, used in all studies, have been previously described (Flores-Garcia et al., 2019). Briefly, 5-day-old adult Anopheles stephensi mosquitoes (maintained at the insectary at JHU, Malaria Research Institute) were allowed to feed on Swiss Webster mice carrying 2–3% parasitemia. Then, 20–22 days after murine blood meal, transgenic SPZ were collected from salivary glands and used within 60 min for IV infection for liver burden studies.

To assess the protective efficacy of anti-PfCSP mAbs in vivo, mAbs were diluted in sterile PBS at the desired dilution. Mice were then passively immunized IV in the tail vein. 2 h later, animals were challenged IV with 2,000 P. berghei transgenic SPZ that express full-length PfCSP and GFP-Luc enzyme. 42 h after parasite challenge, parasite load in the liver was measured by bioluminescence in an in vivo imaging system (IVIS Spectrum, Perkin Elmer). Mice were injected intraperitoneally with 100 μl of d-luciferin (30 mg/ml) and immediately anesthetized with isoflurane for 5 min prior to IVIS. Groups of five anesthetized mice were placed in the imager, and the radiance measurements were recorded by the live imager software, version 4.5.1. The total flux reading for each mouse was recorded individually. Background reading was verified for each study with mice that received only the d-luciferin substrate.

Significant differences were calculated with Mann–Whitney’s t test (between two groups) assuming non-normal distribution, one-way, or two-way ANOVA (among three or more groups) with Tukey’s multiple-comparisons test applied throughout. P values are shown as no significant difference (ns) P > 0.05, *P < 0.05, **P < 0.01, ***P < 0.001, and ****P < 0.0001. All P value analyses were calculated using GraphPad Prism 10 (RRID: SCR_002798).

Fig. S1 establishes adoptive transfer precursor frequencies and relates to Fig. 1. Fig. S2 shows extended iGL_L9 and L9 B cell responses to immunogens, related to Fig. 2. Fig. S3 shows the SHM and affinity of Ab variants produced by those responses, related to Fig. 3. Fig. S4 provides additional in vivo and biophysical characterization of postimmunization L9-derived Abs, related to Fig. 4, 5, and 6. Fig. S5 provides supporting details for cryo-EM, related to Fig. 5 and Table S2. Table S1 details L9-variant mAb expression, and Table S2 details cryo-EM statistics and deposition information.

KI mouse models are available from F.D. Batista on request under a standard material transfer agreement with Mass General Hospital. BCR sequences are available at GenBank: PZ498966–PZ499037 and PZ498895–PZ498965. PDB deposition numbers are listed in Table S2.

We would like to thank Ian A. Wilson and Andrew B. Ward of The Scripps Research Institute for generously sharing the sequences of iGL_311 and iGL_317 for mouse model generation. We would also like to thank the members of the Flow Cytometry Core at the Ragon Institute, as well as Anastasia Yandulskaya-Blue at the Ragon Institute Scientific Editing Platform.

This work was supported by the National Institutes of Health (NIH) National Institute of Allergy and Infectious Diseases grants R01 AI168114 (to F.D. Batista) and R01 AI151178 (to F.D. Batista); the National Research Foundation of Korea’s Basic Science Research Program NRF-2021R1A6A3A14044219 (to J.-H. Koo); and flexible funding from the Ragon Institute of Mass General Brigham, MIT, and Harvard (to F.D. Batista). This project has been funded in whole or in part with Federal funds from the National Cancer Institute, NIH, task order no. 75N91019F00135 under contract no. 75N9101900024. This research was also supported in part by the Intramural Research Program of the NIH. The contributions of the NIH authors are considered Works of the United States Government. The findings and conclusions presented in this paper are those of the authors and do not necessarily reflect the views of the NIH or the US Department of Health and Human Services.

Author contributions: Ja-Hyun Koo: conceptualization, data curation, formal analysis, funding acquisition, investigation, methodology, project administration, resources, software, validation, visualization, and writing—original draft, review, and editing. Prabhanshu Tripathi: conceptualization, formal analysis, investigation, methodology, visualization, and writing—original draft, review, and editing. Yevel Flores-Garcia: formal analysis, investigation, and methodology. Kazutoyo Miura: formal analysis, methodology, and writing—review and editing. Ganchimeg Bayarsaikhan: investigation and methodology. Chen-Hsiang Shen: software. Stephanie R. Weldon: writing—original draft, review, and editing. Sven Kratochvil: investigation, methodology, and writing—review and editing. Thavaleak Prum: Investigation, Ali A. Albowaidey: methodology. Ye-Ji Kim: validation. Jordan R. Ellis-Pugh: data curation and investigation. Quynh Anh Phan: investigation. Jennifer Suurbaar: investigation and methodology. Robert Sifa Onjiko: methodology and validation. Misook Choe: resources. Gordon A. Dale: software and writing—review and editing. Haotian Lei: data curation, investigation, methodology, software, and validation. Nicholas C. Morano: investigation. Lais Da Silva Pereira: investigation. Marlon Dillon: resources. Michael F. Bender: investigation. Mariah Lofgren: investigation. Baoshan Zhang: conceptualization. Theodore C. Pierson: funding acquisition, resources, supervision, and writing—review and editing. Lawrence Shapiro: funding acquisition and supervision. Tongqing Zhou: funding acquisition, resources, and supervision. Usha Nair: resources. Azza H. Idris: supervision. Fidel Zavala: conceptualization, data curation, and methodology. Peter D. Kwong: conceptualization, funding acquisition, project administration, resources, supervision, and writing—original draft, review, and editing. Robert A Seder: conceptualization, funding acquisition, supervision, and writing—review and editing. Facundo D. Batista: conceptualization, funding acquisition, project administration, supervision, and writing—original draft, review, and editing.