Design, immunogenicity, and protective efficacy of stabilized PfCSP variants displayed on nanoparticles

We expressed PfCSP so that we could use it as a benchmark antigen in our studies and noticed that the soluble wild-type (WT) protein provided low yield after purification from HEK293E cells (50 µg per liter culture; Fig. 1a, Supplementary Fig. 1a), and was sensitive to cleavage when left overnight at 4 °C⁶⁵. We performed N-terminal sequencing of a degraded protein sample and found it was cleaved between ₆₆KKNSR₇₀ and ₇₁SLGENDD₇₇, within the PEXEL II sequence (₇₀RSLGE₇₄; ref. ⁶⁶) (Supplementary Fig. 1b). Since the ₆₆KKNSR₇₀ sequence preceding the cleavage site resembles a furin cleavage site, we used site-directed mutagenesis to mutate the positively charged residues (Lys and Arg) to serine and alanine (₆₆SSNSA₇₀; Supplementary Fig. 1b), similar to what was done for viral glycoproteins^67,68,69. We expressed this construct, named C25-SAmut, and observed a 400-fold increase in protein expression to 20 mg/L (Supplementary Fig. 1c). However, a dimer peak during SEC was observed, likely due to an unpaired cysteine, and we further mutated C25-SAmut to generate SAmut by mutating the cysteine to serine as previously described⁷⁰. SAmut expressed well (20 mg/L) and eluted during SEC as a single symmetric peak (Supplementary Fig. 1d). We produced a truncated form of this construct, SAmut-5/3 (Fig. 1a), which contains the 3D7 junctional region in place of the full repeat domain, to focus responses to the junctional epitope and minor epitopes targeted by CIS43 and L9 (ref. ¹⁴)_. A panel of mAbs spanning the NTD, the repeat domain, and the CTD all bound SAmut-5/3 as expected, similar to soluble PfCSP (Supplementary Fig. 1e).

a Primary structures of PfCSP variants, including the R21 immunogen. SP, signal peptide; RI, Region I; RIII, Region III; RII + , Region II; GPI, glycosylphosphatidylinositol anchor sequence. Red lines indicate sites of cysteine and PEXEL mutations. b Reducing SDS-PAGE of purified SAmut-5/3 immunogens. c nsEM of the SAmut-5/3 multimers. Micrographs and structural models of each multimer are shown. d Binding of multimers to PfCSP-directed mAbs measured by ELISA. NHP20 is an unpublished antibody isolated from a non-human primate immunized with WT PfCSP that binds to the NTD, CIS43 is a dual binder for the junctional epitope and major repeats, L9 is a dual binder for the minor epitope and major repeats, 311 binds the major repeats, and mAb15 binds the CTD¹⁶. VRC01 is an anti-HIV-1 antibody used as a negative control. e Immunization regimen and details. IV, intravenous; SPZ, sporozoite. f Parasite burden in the liver after challenge with transgenic sporozoites. R21 was used as a benchmark immunogen. Naive refers to uninfected negative control mice and max burden to infected positive control mice. *p < 0.1; **p < 0.01; ***p < 0.001; ****p < 0.0001 as calculated by Kruskal–Wallis test with multiple comparisons.

To evaluate the effect of multimerization on PfCSP immunogenicity, we used the SpyTag-SpyCatcher (ST-SC) “plug-and-display” technology⁷¹ to display SpyTagged SAmut-5/3 on a variety of self-assembling protein nanoparticles—C4b⁷², ferritin⁷³, and I53-50 (ref. ⁵⁸)—which present 7, 24, and 60 copies of SC, respectively. We expressed and purified SAmut-5/3-ST protein and the homomeric SC-nanoparticles (SC-C4b and SC-ferritin) separately and then conjugated the antigen to each SC-nanoparticle (Fig. 1b, c; Supplementary Fig. 1f, g). For the two-component SC-I53-50 nanoparticle, we conjugated SAmut-5/3-ST to the trimeric SC-I53-50A component, assembled the nanoparticle by adding the I53-50B.4PT1 pentamer⁵⁸, and purified the assembled nanoparticle by SEC. SDS-PAGE revealed little unconjugated SC-bearing protein in each case, suggesting highly efficient conjugation (Supplementary Fig. 1f, g), and SEC and negative stain electron microscopy (nsEM) indicated that each nanoparticle immunogen retained the expected size and morphology after conjugation (Fig. 1b, c; Supplementary Fig. 1f, g). All of the mAbs in the panel described earlier bound each nanoparticle by ELISA, establishing that SAmut-5/3 retained its antigenicity when multimerized (Fig. 1d).

To evaluate the effect of multimerization and antigen copy number on immunogenicity and protection, we immunized groups of 10 C57BL/6 mice with a constant molar dose of antigen at weeks 0 and 3 using ALFQ adjuvant⁷⁴ (Fig. 1e). ALFQ is a liposome-based adjuvant with a synthetic monophosphoryl lipid A analog (3D-PHAD^®) and QS-21, similar to the composition of the AS01 adjuvant used with the RTS,S vaccine. We used R21 as a benchmark immunogen, also adjuvanting it with ALFQ as we could not access RTS,S/AS01B (Mosquirix^TM) or R21/Matrix-M for these studies. We note that although R21 is typically adjuvanted with Matrix-M (Novavax; ref. ⁷⁵), it has also been shown to be protective in combination with liposomal adjuvants containing QS-21 similar to ALFQ⁷⁶. We intravenously challenged the mice at week 6 with 2000 transgenic P. berghei parasites that express PfCSP in place of endogenous PbCSP and GFP/luciferase for measuring liver burden (Pb-PfCSP-GFP/LUC)⁷⁷. A clear trend in liver burden measurements taken 2 days after challenge suggested that increased antigen copy number may improve protection (Fig. 1f). However, none of the differences between the C4b, ferritin, or I53-50 nanoparticles or soluble antigen were statistically significant. Among the immunogens tested, only R21 and the I53-50 nanoparticle displaying ~60 copies of SAmut-5/3 significantly reduced liver burden compared to an unimmunized control group (“max burden”). R21 alone conferred significantly higher protection than the soluble SAmut-5/3. The superior performance of R21 may be due to the different antigen it displays (Fig. 1a) or other features of the HBsAg particle.

Design and characterization of nanoparticle immunogens comprising the CSP junctional region

Given that the I53-50-based immunogen showed good protection in our initial study and this nanoparticle has proven capable of displaying a wide variety of antigens^60,61,62,78, we selected I53-50 as a platform for iterative CSP-based nanoparticle vaccine design. Our overall aims were to evaluate the contribution of each region of CSP to immunogenicity and protection, and to evaluate whether variants of the repeat region could increase the likelihood of eliciting protective antibodies that bind the junctional and minor epitopes that are not present in RTS,S and R21 (refs. ^14,15,16). We used genetic fusion for these studies rather than SC-ST conjugation to generate well-defined immunogens and to simplify our workflow by eliminating the need for a conjugation step. We began by genetically fusing the truncated repeat region and CTD of CSP found in RTS,S and R21 (i.e., the “RT” antigen) to I53-50A, the trimeric component of I53-50 (Fig. 2a, b). In vitro assembly of RT-I53-50A with I53-50B.4PT1 followed by preparative SEC yielded monodisperse nanoparticles of the expected size and morphology (Supplementary Fig. 2). We then replaced RT with a series of antigens that included the full NTD or Region I (RI), comprised several different variations of the central repeat region, and lacked the CTD (CSP A-H, Supplementary Table 1). Though the CTD contains T cell helper epitopes^79,80, we explored whether it could be excluded because it has been shown that antibodies targeting it are weakly neutralizing or do not inhibit parasite traversal/development^81,82. These designs all expressed but were prone to aggregation except for CSP F, the only design which contained a truncated N-terminal domain comprising only the RI (Fig. 2b). We were able to successfully assemble and characterize CSP F nanoparticles that closely resembled RT-I53-50 (Supplementary Fig. 2). We then tested whether similarly truncating the N-terminal domain in the other proteins would improve their solution properties, but the new constructs (CSP A2-H2; Supplementary Table 1) also aggregated and were not pursued further.

a Models of the trimeric RT-I53-50A (RT in dark gray and blue, I53-50A in light gray) and pentameric I53-50B (orange) components, and an assembled RT-I53-50 nanoparticle. b Schematics of junctional region antigens. Each antigen was genetically fused to I53-50A. c Immunization regimen and details of the study. d Serum antibody titers against SAmut, determined by ELISA using sera obtained 1–2 weeks after the primary and third immunizations. Statistical significance was calculated by one-way ANOVA with multiple comparisons. e Peptide mapping ELISAs using pooled sera from each group, measured using mesoscale discovery (MSD) -multi-spot assay system. f Parasite burden in the liver after challenge with transgenic sporozoites. RT-I53-50 was used as the benchmark immunogen. *p < 0.1; **p < 0.01; ***p < 0.001; ****p < 0.0001 as calculated by Kruskal–Wallis test with multiple comparisons.

We then tested a new family of designs that included the CTD, as it appeared to be important for the solution properties of I53-50A fusion proteins displaying CSP-derived antigens. Four designs contained RI, the junctional region, and increasing numbers of major repeats (up to 17; CSP P, Q, R, and J), while a fifth was identical to CSP J except that it lacked the RI (CSP K; Fig. 2b). All of these constructs were successfully expressed in E. coli, purified, and assembled into nanoparticles by mixing with I53-50B.4PT1 pentamer. In each case, analytical SEC, dynamic light scattering (DLS), and nsEM revealed monodisperse populations with the expected morphology, and binding of mAbs specific to various regions of PfCSP showed the expected patterns (Supplementary Fig. 2). Interestingly, similar constructs that contained the entire repeat region of PfCSP failed to express (CSP L and M; Supplementary Table 1). In summary, we generated a series of nanoparticle immunogens displaying various CSP-derived antigens that would allow us to evaluate the contribution of each region of the protein to immunogenicity and protection.

Immunogenicity of nanoparticle immunogens comprising the CSP junctional region

To evaluate immunogenicity and the protection afforded by this series of nanoparticle immunogens, groups of 10 mice were injected intramuscularly with 3 µg of each nanoparticle formulated with ALFQ, followed by two homologous boosts given 3 weeks apart (Fig. 2c). As benchmarks, mice were immunized with RT-I53-50 nanoparticles and non-assembled RT-I53-50A trimers. Serum was collected 1–2 weeks after each immunization and measured anti-SAmut titers by ELISA (Fig. 2d). Although the RT-I53-50 nanoparticles elicited significantly higher levels of antigen-specific antibodies than the RT-I53-50A trimer and CSP F nanoparticles after a single immunization, all groups had similar titers in the anti-SAmut ELISA after three immunizations.

To determine the epitopes targeted by vaccine-elicited antibodies, we conducted a peptide binding ELISA using pooled sera from each group of mice. We used a series of overlapping peptides, spanning the repeat region and CTD of PfCSP, which we refer to as peptides 20–61 and C-term (Fig. 2e; ref. ¹⁴). As expected, sera from mice vaccinated with immunogens containing the CTD (i.e., all except CSP-F) showed a strong response to this domain. Also as expected, serum antibodies from RT-I53-50-vaccinated mice had a strong preference for binding to NANP-containing peptides 27 (NVDPNANPNANPNAN), 29 (NANPNANPNANPNAN), and 61 (NANPNANPNANPNKN), but did not bind well to peptides containing the junctional epitope or minor repeats. Sera from mice that received immunogens containing the junctional region displayed more balanced binding across the set of peptides, although with varying magnitudes that roughly correlated with the total number of repeats in each immunogen. For example, CSP P elicited weak responses against repeat peptides while CSP Q and R showed stronger binding across all peptides tested, including peptide 20, which spans the junctional epitope (PADGNPDPNANPNVD). CSP F, J, and K, which included more copies of the major repeat, showed stronger binding that was more balanced than RT-I53-50 but skewed more toward the major repeats (i.e., peptides 27 and 29) than CSP Q and R.

Six weeks after the second boost, we challenged the mice intravenously with 2000 sporozoites and measured liver burden 2 days later by IVIS (Fig. 2f). Although all of the immunogens other than CSP P and Q provided significantly better protection than the max burden control group, only CSP K reached the same level of statistical significance as RT-I53-50. Considered together with the peptide ELISA data (Fig. 2e), these results indicate that immunogens with higher major repeat content (i.e., RT-I53-50 and CSP F, J, K, and R) induced better protection than those focused only on the junctional region or containing a reduced number of major repeats (i.e., CSP P & Q). Furthermore, the data show that the non-assembling RT-I53-50A trimer confers less protection than the RT-I53-50 nanoparticle, and that pre-challenge anti-SAmut ELISA titers alone cannot be used to reliably predict protection^83,84. Overall, our data demonstrate that we were able to modulate the epitope specificities of vaccine-elicited antibodies by displaying various CSP-derived antigens on I53-50, but that none of the novel immunogens was able to induce better protection than our benchmark RT-I53-50 nanoparticle immunogen.

Design, characterization, and immunogenicity of non-native CSP-repeat nanoparticles

We next designed a series of I53-50A trimers bearing non-native repeat-based antigens to attempt to further focus the vaccine-elicited immune response towards the junctional region or minor repeats (Fig. 3a, b). Each antigen in the series comprised 18 total repeats, always ending with a major repeat to provide a native-like junction with the C-terminal domain. We designed constructs that displayed alternating forms of the CSP junctional region (CSP Y, Z), alternating junctional-major or minor-major repeats (CSP W, X), completely non-native sequences that included junctional or minor repeats only (CSP U, V) as well as a tandem junctional-minor-major repeat antigen (CSP β). All of these I53-50A fusion proteins were expressed in E. coli, purified using IMAC and SEC, and mixed with I53-50B to generate nanoparticle immunogens. Analytical SEC, DLS, and nsEM again indicated the formation of monodisperse I53-50-based nanoparticle immunogens (Supplemental Fig. 3). Antigenicity was characterized by ELISA and showed that most of the mAbs in our panel bound each of the immunogens, though most notably a decrease or loss in binding for CIS43 was observed for CSP U, V, W, and β (Supplemental Fig. 4a).

a Models of the CSP X-I53-50A trimer (CSP X in green, dark gray, and blue; I53-50A in light gray), I53-50B (orange), and an assembled CSP X-I53-50 nanoparticle. b Schematics of non-native CSP-repeat antigens. Each antigen was genetically fused to I53-50A. c Immunization regimen and details of the study. d Serum antibody titers against SAmut, determined by ELISA using sera obtained after the second and third immunizations. Statistical significance was calculated by one-way ANOVA test with multiple comparisons. e Parasite burden in the liver after challenge with transgenic sporozoites. RT-I53-50 was used as the benchmark immunogen. *p < 0.1; **p < 0.01; ***p < 0.001; ****p < 0.0001 as calculated by Kruskal–Wallis test with multiple comparisons.

Like our previous immunization study, we immunized groups of 10 mice three times intramuscularly with 3 µg of RT-I53-50 or each non-native repeat nanoparticle formulated with ALFQ adjuvant (Fig. 3b, c). In this instance, each group of mice received nearly the same number of moles of nanoparticle immunogen because of their nearly identical molecular weights. Anti-SAmut CSP titers measured by ELISA after the second and third immunizations showed that RT-I53-50 induced the highest CSP-specific antibody titers (Fig. 3d). CSP Z, which comprised two tandem repeats of the entire junctional region of PfCSP, elicited the highest SAmut CSP-specific antibody titers among the non-native repeat nanoparticles. As before, the mice were challenged with 2000 sporozoites six weeks after the second boost and parasite load in the liver was measured (Fig. 3e). RT-I53-50, CSP-X, and CSP-Z were the only three immunogens that significantly reduced liver burden compared to the max burden control group, with RT-I53-50 providing the best protection. These results may be explained by the fact that these immunogens contained more native-like repeat cadences, while the others had higher proportions of non-native sequences of repeats such as NANP-NPDP or NPDP-NVDP (Fig. 3b).

To map the epitopes targeted by antibodies elicited by these three protective immunogens, we used a variation of the peptide binding ELISA in which serum antibody binding to a combined peptide 20–23 or a major repeat peptide coated on the ELISA plate (PADGNPDPNANPNVDPNANPNVDPNAN and (NANP)₉, respectively) competed with binding to free peptides pre-incubated with the pooled sera (Supplementary Fig. 4b, c). The peptides used for competition (peptides 20–61) spanned the repeat region of PfCSP. We first gauged the performance of the assay using three mAbs that bind the major, junctional, and minor repeat regions (mAb4 [ref. ¹⁴], CIS43, and L9, respectively) and found that the preferred peptide epitopes of each mAb competed with its binding to the plated antigens as expected. Specifically, mAb4 bound (NANP)₉ more strongly than peptide 20–23 and exhibited the greatest reduction in binding in the presence of major repeat-containing peptides (peptides 27, 29, 61), while L9 bound peptide 20–23 more strongly and showed the greatest reduction in signal in the presence of peptides containing minor repeats (peptides 20, 21, 22, 23, 43, and 44). CIS43 bound both peptides strongly as expected¹⁴ and showed a clear and consistent rank-ordering of peptide competition (peptide 21 > 20, 23 > 27 > 22, 43 > 44 > 29 > 61). Based on these mAb benchmarking data, we concluded that the peptide competition assay provided a sensitive readout of epitope specificity. Serum antibodies elicited by CSP X and CSP Z bound both ELISA antigens roughly equivalently, whereas the sera from mice receiving RT-I53-50 clearly bound (NANP)₉ more strongly than peptide 20–23 as expected. For all three immunogens, the rank-ordering of competing peptides was consistent across both ELISA antigens. Anti-RT-I53-50 sera were most strongly competed by peptides containing higher numbers of major repeats (peptides 27, 29, and 61) and less so by peptides from the junctional region (peptides 20, 21, 22, and 23). By contrast, peptides containing minor repeats and those from the junctional region most effectively prevented CSP X- and Z-elicited antibodies from binding to the plated antigens, respectively. For both of the latter two immunogens, the major repeat peptides (peptides 29, 43, and 61) provided the weakest competition.

In summary, our data show that although nanoparticles displaying non-native CSP repeat cadences can focus responses toward junctional and minor repeat epitopes, they do not elicit antibodies that reduce liver burden to the same extent as the same nanoparticle displaying the RT antigen.

Immunogenicity and protection afforded by mosaic and cocktail nanoparticle immunogens

Previous analyses have indicated that the most potently protective anti-CSP mAbs tend to be those that bind the major repeats with high affinity while also cross-reacting with the junctional and minor epitopes^15,33,85. To explore whether we could elicit protective levels of such cross-reactive antibodies by vaccination, we conducted another mouse immunogenicity and challenge study in which we compared a series of mosaic and cocktail nanoparticle immunogens based on CSP X, CSP Z, and RT-I53-50. Several studies over the last few years have evaluated mosaic nanoparticle immunogens that co-display multiple antigenic variants on the same nanoparticle for their ability to elicit B cell and antibody responses of greater breadth than monovalent nanoparticles or mixtures thereof (“cocktails”)^{63,72,73,86,87,88,89,90,91,92,93,94,95,96}. Two-component assemblies like I53-50 facilitate the production of mosaic nanoparticles since multiple antigens can be co-displayed by simply adding I53-50B pentamer to mixtures of multiple different antigen-bearing I53-50A trimers^{63,72,73,86,87,88,89,97,98}.

We generated mosaic nanoparticles co-displaying RT and CSP X, RT and CSP Z, or CSP X and Z at 50% valency (i.e., 30 copies) each, as well as a mosaic nanoparticle co-displaying all three antigens at 33% valency (i.e., 20 copies) by adding a molar equivalent of I53-50B pentamer to appropriate mixtures of antigen-bearing I53-50A trimer components. We also made corresponding cocktail immunogens (i.e., RT + X, RT + Z, and RT + X + Z) by individually assembling and purifying each monovalent nanoparticle and then mixing them together. Analytical SEC, DLS, and nsEM indicated that the mosaic and cocktail nanoparticle immunogens assembled as intended (Supplementary Fig. 5).

Following our previous immunization regimens, we administered 3 µg of each mosaic or cocktail nanoparticle immunogen to groups of 10 mice intramuscularly with ALFQ adjuvant, followed by two boosts (Fig. 4a, b). We again included RT-I53-50 as a benchmark immunogen as well as naive and max burden control groups. Ten days post-prime, mice were bled and anti-SAmut CSP titers were measured by ELISA (Fig. 4c). All vaccine groups had similar anti SAmut CSP titers post-prime. Due to restrictions imposed during the COVID-19 pandemic, the study was put on hold after the primary immunization, resulting in an interval of 25 weeks between the prime and the first boost. Six weeks after the second boost, mice were challenged IV with 2000 sporozoites and parasite liver load was measured by IVIS (Fig. 4d). Despite having similar post-prime anti-SAmut CSP titers, the immunogens conferred various reductions in liver burden compared to the mock-immunized control. Unfortunately, due to COVID-19 related restrictions, immunogenicity data post-boost was not collected, preventing direct correlation of boost immunogenicity with observed liver burden. RT-I53-50 was again the most protective immunogen with the lowest liver burden (p < 0.0001), while the mosaic RT/X nanoparticle performed second-best (p < 0.001) and the cocktail RT + X nanoparticle performed third-best (p < 0.01). One potential reason for immunogens based on RT and CSP X being more protective compared to those based on CSP Z may be that they contain a higher proportion of native-like repeat cadences (e.g., NVDP-NANP and NANP-NANP vs. NANP-NPDP) and thus more protective epitopes, an interpretation that is also supported by our previous study (Fig. 3). Our data did not allow us to distinguish between the mosaic and cocktail immunogens; more detailed studies would be required to determine whether differences exist in the B cell and antibody responses elicited by each.

a Immunization regimen and details of the study. b Schematics depicting the antigenic composition of each immunogen: either mosaic nanoparticles with 33 or 50% valency of each antigen, or groups (cocktails) of monovalent nanoparticles. c ELISA endpoint titer of each immunogen to SAmut-coated plates post-prime. Statistical significance was calculated by one-way ANOVA test with multiple comparisons. d Parasite burden in the liver after challenge with transgenic sporozoites. RT-I53-50 was used as the benchmark immunogen. *p < 0.1; **p < 0.01; ***p < 0.001; ****p < 0.0001 as calculated by Kruskal–Wallis test with multiple comparisons.

Comparison of R21 and RT-I53-50 adjuvanted with ALFQ

As we repeatedly observed that RT-I53-50 conferred better protection than immunogens containing junctional or minor repeats, we compared it against R21 in a head-to-head immunogenicity and challenge study. We also included an I53-50 nanoparticle displaying full-length SAmut CSP as a genetic fusion to determine if inclusion of the NTD and the entire repeat region improved protection. A comparator group received soluble (i.e., non-particulate) SAmut CSP to control for the effect of multivalent display. The SAmut-CSP-I53-50A trimer was expressed in E. coli, purified by immobilized metal affinity chromatography (IMAC) and SEC, and assembled with I53-50B to form nanoparticles. The resultant SAmut-CSP-I53-50 nanoparticles were purified by SEC to remove residual components, and the purified assemblies were evaluated by DLS and analytical SEC, both of which indicated monodisperse nanoparticles of the expected size and morphology (Supplementary Fig. 6).

Groups of 10 mice were immunized intramuscularly three times with a 3 μg total protein dose of R21, SAmut CSP, SAmut-CSP-I53-50, or RT-I53-50 formulated in ALFQ (Fig. 5a, b). We also included a group that received 300 μg of CIS43 2 hours prior to infection as a fully protective control. Anti-SAmut CSP ELISA using sera obtained 2 weeks after the final immunization revealed that R21, RT-I53-50, and SAmut-CSP-I53-50 induced similar levels of anti-CSP antibodies, all three of which were lower than monomeric SAmut CSP (Fig. 5c). Following challenge, immunization with R21, RT-I53-50, and monomeric SAmut CSP all significantly reduced liver burden compared to the max burden control group, with R21 and RT-I53-50 exhibiting the greatest levels of liver burden reduction using this immunization regimen (Fig. 5d).

a Groups and doses used. b Immunization regimens and details for the three-dose and two-dose experiments. c ELISA endpoint titer for each immunogen in the three-dose study to SAmut-coated plates after the second boost and before the challenge. Statistical significance was calculated by one-way ANOVA test with multiple comparisons. d Parasite burden in the liver after three immunizations followed by challenge with transgenic sporozoites. For the CIS43 group, 300 μg of CIS43 was administered 2 h prior to infection. *p < 0.1, **p < 0.01, ***p < 0.001, ****p < 0.0001 as calculated by Kruskal–Wallis with multiple comparisons test compared to infected (max burden) control. e ELISA endpoint titer for each immunogen in the two-dose titration study to SAmut-coated plates after the prime and boost immunizations. Statistical significance was calculated by one-way ANOVA test with multiple comparisons. f Parasite burden in the liver after two immunizations followed by challenge with transgenic sporozoites. *p < 0.1; **p < 0.01; ***p < 0.001; ****p < 0.0001 as calculated by Kruskal–Wallis test with multiple comparisons compared to max burden control.

To attempt to resolve potential differences in the levels of protection provided by R21 and RT-I53-50, we conducted another study in which we lowered the total protein dose to 2, 0.4, and 0.08 μg and reduced the number of immunizations to two (Fig. 5b). At each dose, R21 elicited higher levels of anti-SAmut CSP serum antibody titers post-prime and -boost, with comparable titers between R21 at 0.08 μg and RT-I53-50 at 2 μg (corresponding to 0.03 and 0.66 μg of RT antigen, respectively; Fig. 5e). This result was consistent with the parasite challenge results, with R21 at 2 μg being the only group other than the CIS43 mAb-treated control that significantly reduced liver burden compared to the mock-immunized group (Fig. 5f). Altogether, our data show that although immunization with RT-I53-50 and R21 in ALFQ induces equivalently robust protection after 3 doses, R21 is more protective at lower doses and using a reduced number of immunizations.