This study was nested within two independent clinical trials: R21/Matrix-M^® (NCT03580824; 2019-2022) and RH5.1/Matrix-M^® (NCT04318002; 2021-2023). We additionally compared a subset of the RH5.1/Matrix-M^® (NCT04318002) read-outs to previously published data with an RH5.1/AS01_B clinical trial in a malaria-naïve UK adult population (NCT02927145; 2016-2019). Sample sizes for the analyses were determined by sample availability and are summarised in the Supplementary Table 1. The group sizes shown in Tables 1–3 represent the maximum sample size as defined by trial enrolment.

The R21/Matrix-M^® “VAC073” trial (NCT03580824) was a Phase 1b, open-label, age de-escalation, dose-escalation trial conducted at KEMRI, Kilifi, Kenya, involving adults (18–45 years), children (1–5 years; not included in the analyses presented here), and infants (5–11 months). Vaccines were administered on a 0-1–2 month regimen, with groups varying in the dose of R21 antigen and/or Matrix-M^® adjuvant (Table 1). Full details of the clinical trial can be found in the primary trial publication, including analyses of 1-5-year olds and antibody responses following further booster doses which were outside the scope of this study (25, 27).

The RH5.1/Matrix-M^® “VAC080” trial (NCT04318002) was also a Phase 1b, open-label, age de-escalation dose-escalation trial that was conducted at Ifakara Health Institute, Bagamoyo, Tanzania. The trial recruited healthy adults (18–45 years); not included in the analyses presented here (19), and children (5–17 months) (14). Vaccines were administered on a 0-1–2 month “monthly” or 0-1–6 month “delayed” regimen, with delayed regimen groups varying in terms of the vaccine antigen dose of the final booster (“delayed” versus “delayed fractional”) and pre-exposure to malaria infection (Table 2). Full details of the clinical trial, including analyses of adult participants, can be found in the primary trial publication (14).

To note, in the interest of consistency we use the age group nomenclature as per primary trial publications, i.e. 5-11mo “infants” in the R21/Matrix-M^® trial (NCT03580824) (25) and 5-17mo “children” in the RH5.1/Matrix-M^® trial (NCT04318002) (14). We do not seek to draw any comparisons between these two overlapping age groups in this study.

The RH5.1/AS01_B “VAC063” trial (NCT02927145) was a Phase 1/2a trial conducted in malaria-naïve adults in the UK to determine the safety, immunogenicity, and efficacy of RH5.1/AS01_B against blood-stage controlled human malaria infection (CHMI) model (4). Vaccines were administered on a 0-1-2 “monthly” or 0-1–6 month “delayed fractional regimen” (Table 3). Full details of the clinical trial, including groups not included in this study, can be found in the primary trial publication (4).

All three trials were conducted according to the principles of the Declaration of Helsinki 2013 and Good Clinical Practice. For the RH5.1/Matrix-M^® trial (NCT04318002), guardians of the child participants signed or thumb-printed an informed consent form at the in-person screening visit, and consent was verified verbally before each vaccination. For the RH5.1/AS01_B trial (NCT02927145), all participants signed written consent forms, and consent was checked before each vaccination and prior to CHMI. For the R21/Matrix-M^® trial (NCT03580824) written informed consent for participation in this trial was provided by adult volunteers, whilst the children and infant participants’ parents or legal guardians provided consent for participation in this trial, which allowed the use of the sample and data for the primary trial objective and future exploratory analysis.

Cryopreserved PBMCs from the RH5.1/Matrix-M^® and R21/Matrix-M^® vaccine trials were thawed using pre-warmed R10 media (RPMI [R0883, Sigma] supplemented with 10% heat-inactivated FCS [60923, Biosera], 100U/ml penicillin/0.1mg/mL streptomycin [P0781, Sigma], and 2mM L-glutamine [G7513, Sigma]). Thawed PBMC were washed twice and rested in R10 for 1h at 37 °C with benzonase (70664-10KUN, EMD Millipore). The PBMCs were enriched for B cells using the Human Pan-B cell Enrichment Kit [19554, StemCell]. Enriched B cells were then stained with viability dye FVS780 (565388, BD Biosciences), followed by incubation with Trustain (422302, Biolegend) to block non-specific Fc receptor staining. Subsequent surface staining of B cells was performed with anti-human CD19-BV786 (563325, BD Biosciences), IgG-BB515 (564581, BD Biosciences), IgM-BV605 (562977, BD Biosciences), CD27-PE-Cy7 (560609, BD Biosciences), CD21-BV711 (563163, BD Biosciences) IgA-PerCP-Vio700 (130-113-478, Miltenyi), CD38-BV480 (566137, BD Biosciences), and CD138-APC-R700 (566050, BD Biosciences). Two fluorophore-conjugated NANP or RH5 probes, for R21/Matrix-M^® and RH5.1/Matrix-M^® trial samples respectively, were also included. The preparation of the RH5 probes has been published previously (6, 19). In brief, monobiotinylated RH5 was produced by transient co-transfection of HEK293F cells with a plasmid encoding BirA biotin ligase and a plasmid encoding a full-length RH5. The RH5 plasmid was based on ‘RH5-bio’ (a gift from Gavin Wright; University of York, York, United Kingdom; Addgene plasmid 47780; http://n2t.net/addgene:47780;RRID: Addgene_47780; (39). RH5-bio was modified prior to transfection to incorporate a C-tag for subsequent protein purification. The monobiotinylated NANP peptide used for probe generation was commercially obtained (CPD70698, Thinkpeptides, ProImmune). Two probes per antigen were freshly prepared for each experiment, by incubation of monobiotinylated RH5 or NANP with streptavidin-PE (S866, Invitrogen) or streptavidin-APC (17-4317-82, Biolegend) at an approximately 4:1 molar ratio to facilitate tetramer generation. Probes were then centrifuged at 13000rpm (max microfuge speed) for 10min at room temperature to remove aggregates. After surface staining, enriched B cells were permeabilised and fixed with Transcription Factor Buffer Set (562574, BD Biosciences) to enable intracellular staining. The fixed cells were stained with anti-human Ki67-BV650 (563757, BD Biosciences; data not shown) and probes at 1/10 dilution as compared to concentrations used for surface staining. The cells were washed and stored at 4 °C until acquisition.

B cells were acquired on a Fortessa X20 flow cytometer with FACSDiva8.0 (BD Biosciences). Data were analysed using FlowJo (v10.8; Tree Star Inc.). Samples were excluded from analysis if there were < 50 cells in the parent population. NANP- or RH5-specific IgG+ B cells were identified as RH5/APC+RH5/PE+ or NANP/APC+NANP/PE+ cells within various live memory B cell populations as shown in Supplementary Figure 1.

Technical issues with our CSP stimulation meant we were unable to investigate frequencies of CSP-specific cTfh cells within total memory cTfh cells from available PBMC in the R21/Matrix-M^® trial (NCT03580824). Instead, we used a different read-out as a proxy for vaccine-induced cTfh cell responses 7 days after the final vaccination in the primary three-dose series (V3 + 7), total CXCR5+PD1+ cTfh cells.

The ex vivo total cTfh assay was adapted from previously published methodology (37). In brief, cryopreserved PBMC from the R21/Matrix-M^® clinical trial (NCT03580824) were thawed, washed twice in R10 as above, and then rested for a further 3h at 37 °C. PBMC were then stained with viability dye Live/Dead Aqua (L34957, ThermoFisher) and washed before surface staining with anti-human PD1-BV650 (329950, Biolegend), CD45RA-eF450 (304123, Biolegend), CXCR5-FITC (356913, Biolegend), CCR6-PE (G034E3, G034E3), CD4-APC-eF780 (47-0049-42, eBioscience), CD3-AF700 (56-0038-82, eBioscience) and CXCR3-APC (353708, Biolegend) at 37°C. Cells were fixed with 1% paraformaldehyde (43368, Alfa Aesar) and stored at 4 °C until acquisition on the same day.

The cells were acquired on a Fortessa X20 flow cytometer with FACSDiva8.0 (BD Biosciences). Data were analysed using FlowJo (v10.8; Tree Star Inc.). Samples were excluded from analysis if there were < 50 cells in the parent population. Memory CD4+ T cells were identified as live CD4+ CD45RA- cells, and cTfh were defined as CXCR5+PD1+ within the memory CD4+ T cells as shown in Supplementary Figure 2.

The AIM assay was adapted from previously published methodology (21). In brief, cryopreserved PBMC from the RH5.1/Matrix-M^® clinical trial (NCT04318002) were thawed as above and washed twice in R10 before resting for 3h at 37°C with benzonase (70664-10KUN, EMD Millipore). This was followed by 22h stimulation with either R10 medium alone as a negative control, 2.5μg/peptide/mL of a RH5 peptide pool (20-mer peptides spanning full-length RH5 protein, overlapping by 10 amino acids, total of 50 peptides [NeoScientific] (21, 40) or 1μg/mL Staphylococcal enterotoxin B (SEB; S-4881, Sigma; positive control). Following stimulation, PBMC were surface stained as per the ex vivo cTfh cell panel above as well as anti-human CD134 (OX40)-BUV395 (743286, BD Biosciences), CD25-PE-Cy7 (335824, BD Biosciences), and CD69- PE-Cy5 (555532, BD Biosciences). Finally, the cells were washed and then fixed with 1% paraformaldehyde (43368, Alfa Aesar) and stored at 4 °C until acquisition on the same day.

Samples were acquired on a Fortessa X20 flow cytometer with FACSDiva8.0 (BD Biosciences). Data were analysed using FlowJo (v10.8; Treestar). Samples were excluded from analysis if there were <50 cells in the parent population. RH5-specific cells within memory CD4+ or cTfh cells were defined as CD25+OX40+ or CD25+CD69+ as shown in Supplementary Figures 2–4. Background responses to medium alone were subtracted from RH5-specific responses in matched samples.

Standardised ELISAs were used to quantify the magnitude of the anti-NANP and anti-RH5 serum IgG pre- and post-vaccination with R21/Matrix-M^® and RH5.1/Matrix-M^® vaccines, respectively. Anti-RH5.1 total IgG ELISAs were performed against full-length RH5 protein (RH5.1) using standardised methodology as previously described (14, 26). Anti-NANP total IgG antibodies were measured against a synthetic peptide comprising 6 repeats of the amino acid sequence NANP (NANP6), as previously described (25). In both trials, the reciprocal of the test sample dilution giving an OD₄₀₅ of 1.0 in the standardised assay was used to assign an ELISA unit value of the standard. A standard curve and Gen5 ELISA software v3.04 (BioTek, UK) were used to convert the OD₄₀₅ of individual test samples into AU. In RH5 vaccinees, the AU were converted into μg/mL following generation of a conversion factor by calibration-free concentration analysis as reported previously (40).

Data was analysed using GraphPad Prism version 10.1 for Windows (GraphPad Software Inc., California, USA). Comparisons between different time points/groups were done with Mann-Whitney U tests and/or a Kruskal-Wallis test with Dunn’s correction for multiple comparisons. Spearman’s rank correlation was used to test associations. All p values (or adjusted p values for tests with Dunn’s correction) are two-tailed and were considered significant at the value of < 0.05. Only significant results are annotated on figures but all statistical tests performed (including those with non-significant results) are detailed in figure legends.

Within the R21/Matrix-M^® clinical trial (NCT03580824), NANP probes were used to identify circulating NANP-specific IgG+, IgM+, and IgA+ memory B cells (mBC; defined as shown in Figure 1A; Supplementary Figure 1) in PBMC samples at baseline (Day 0 (D0)) and 28 days after the final vaccination in the primary three-dose series (V3 + 28). The groups from NCT03580824 included in this analysis are summarised in Table 1. In brief, this comprises Group 1A/B (adults; 10µg R21/50µg Matrix-M^® at 0-1–2 months), Group 3A/C (infants; 5µg R21/25µg Matrix-M^® at 0-1–2 months), Group 3B/D (infants; 10µg R21/50µg Matrix-M^® at 0-1–2 months), and Group 3E (infants; 5µg R21/50µg Matrix-M^® at 0-1–2 months). Significant NANP-specific responses were observed within total, resting, or activated memory IgG+, IgA+, and IgM+ B cell populations between D0 and V3 + 28 within all adult and infant groups, with the only exception of resting IgA+ memory B cells in Groups 3B/D and 3E, activated and resting IgM+ in Group 3B/D and 3A/C, respectively (Supplementary Figure 5). Of the different mBC subsets, the highest median responses were observed within the activated IgG+ memory subset (Figure 1B; Supplementary Figure 5).

In terms of differences between groups, the comparisons we were most interested in within the R21/Matrix-M^® trial were: Group 1A/B versus Group 3B/D (adults versus infants; same doses of R21 and Matrix-M^®), Group 3A/C versus Group 3E (25µg versus 50µg Matrix-M^® in infants; same dose R21), and Group 3B/D versus Group 3E (5µg versus 10µg R21 in infants; same dose of Matrix-M^®). Post-vaccination NANP-specific B cell frequencies were significantly higher in infants than adults (receiving the same 10μg R21/50μg Matrix-M^® regimen; Groups 1A/B and 3B/D) within total memory IgG+ and IgA+ B cells, as well as within activated and resting memory IgG+ subpopulations (activated memory IgG+ Figure 1C; V3 + 28 medians: 0.19% Group 1A/B adults, 0.56% Group 3B/D infants p = 0.0002; Supplementary Figure 6).

There were no significant differences by age within any of the NANP-specific memory IgM+ B cell populations (Supplementary Figure 6). Within the infant groups, an increase in Matrix-M^® dose from 25µg (Group 3A/C) to 50µg (Group 3E) with the same R21 dose was weakly associated with increased activated memory IgG+ B cells (Figure 1D; V3 + 28 medians: 0.28% Group 3A/C 25µg Matrix-M^®, 0.56% Group 3E 50µg Matrix-M^®, p = 0.0294). Increasing the R21 antigen dose showed no significant difference in the NANP-specific mBC populations (Group 3B/D versus Group 3E; data not shown). When pooling all groups analysed, frequencies of NANP-specific cells within memory IgG+ B cells correlated with matched time point V3 + 28 serum anti-NANP IgG (activated memory IgG+ Figure 1E; Supplementary Figure 7), but not late time point serum anti-NANP IgG titres (V3 + 1yr infants, V3 + 2yr adults; data not shown). To note, these differences were no longer significant when groups were analysed separately.

While there were no significant post-vaccination differences in memory CD4+ T cells within total T cells in all groups (Figure 2A), there were weak trends to increased frequencies of CXCR5+PD1+ cTfh cells within the memory CD4+ T cell population and this was significant in adults (Figure 2B; Group 1A/B adults D0 median: 2.47% vs V3 + 7 median: 3.31% p = 0.0376). Few baseline samples were assayed in the infant groups; thus we were not able to meaningfully compare pre- and post-vaccination responses given baseline frequencies of these subsets were not negative (Figures 2A, B).

In terms of differences between adults and infants, adult vaccinees had significantly higher frequencies of memory CD4+ T cells post-vaccination as compared to infant vaccinees (Figure 2C V3 + 7 medians adults 1A/B 42% vs infants 3B/D median 17.8% p = 0.0001). Within this CD4+ memory population, we observed no difference in frequencies of CXCR5+PD1+ cTfh cells in infants compared to adults (Figure 2D). There were also no significant differences in the frequencies of memory CD4+ T cells or CXCR5+PD1+ cTfh cells between the infant groups that received different R21 or Matrix-M^® doses (data not shown). However, we did observe higher frequencies of cTfh1 and lower frequencies of cTfh17 in infants as compared to adults within the CXRC5+PD1+ cTfh population, suggesting some differences in the cTfh cell repertoire (Figure 2E; Tfh1 medians adults 1A/B 11.25% vs infants 3B/D 27.50% p = 0.0021, and Tfh17 medians adults 1A/B 19.70% vs infants 3B/D median 7.91% p = 0.0001). No correlation was observed between the frequencies of CXCR5+PD1+ cTfh cells at V3 + 7 and serum antibody at V3 + 28 or late time point (~2 years post-vaccination; data not shown).

In addition to our analyses of the licensed malaria vaccine R21/Matrix-M^®, we were also interested in the vaccine antigen-specific cTfh and B cell responses to the leading blood-stage malaria vaccine candidate, RH5.1/Matrix-M^®. The groups from NCT04318002 included in this analysis are summarised in Table 2. In brief, this comprises Group 2A (children; 10µg RH5.1/50µg Matrix-M^® at 0-1–2 months), Group 2B (children; 10µg RH5.1/50µg Matrix-M^® at 0-1–6 months), Group 2C (children; 50µg RH5.1/50µg Matrix-M^® at 0–1 months and 10µg RH5.1/50µg Matrix-M^® at 6 months), and Group 2D (“high exposure” children; 50µg RH5.1/50µg Matrix-M^® at 0–1 months and 10µg RH5.1/50µg Matrix-M^® at 6 months).

Using cryopreserved samples from the RH5.1/Matrix-M^® trial, we were able to employ an AIM assay to report on RH5-specific cTfh frequencies. We first evaluated the baseline (Day 0; D0) and post-vaccination cTfh cell responses 7 days after the final vaccination in the primary three-dose series (V3 + 7) within total memory CD4+ T cells and total memory cTfh cells (as defined in Supplementary Figures 2–4). RH5-specific responses were detected within total memory CD4+ T cells in all groups (Figure 3A) and within total memory cTfh cells in all delayed regimen groups (Figure 3B). Trends were very similar when reporting frequencies of RH5-specific memory cTfh cells within CXCR5+PD1+ cells (Figure 3C). However, there were no significant bulk changes in the frequencies of total CXCR5+PD1+ cTfh cells within the memory CD4+ T cell population post-vaccination, except in the high malaria-exposed group (Supplementary Figure 8). Within cTfh1 (Figure 3D), cTfh2 (Figure 3E), and cTfh17 (Figure 3F) subsets (as defined in Supplementary Figure 2), RH5-specific responses were only detected consistently across all groups within the cTfh2 population (Figure 3E).

With respect to differences between groups, the comparisons we were interested in within the children vaccinees in the RH5.1/Matrix-M^® trial were Group 2A versus 2B (monthly versus delayed regimen), Group 2B versus 2C (delayed versus delayed fractional regimen), and Group 2C versus 2D (low versus high prior exposure to malaria; same delayed fractional regimen). Interestingly, within this RH5-specific cTfh cell data set, it was only in comparison between the different malaria pre-exposure groups that we observed differences in the magnitude of the RH5-specific cTfh cell response. Specifically, we observed higher frequencies of RH5-specific cells within the total memory CXCR5+ cTfh (Figure 3B V3 + 7 median 2D 0.48% vs 2C 0.15% p = 0.04), total memory CXCR5+PD1+ cTfh (Figure 3F V3 + 7 median 2D 0.88% vs 2C 0.19% p = 0.004), and memory cTfh2 (Figure 3D V3 + 7 median 2D 0.38% vs 2C 0.12% p = 0.001) populations in the “high exposure” Group 2D as compared to “low exposure” Group 2C who received the same delayed fractional regimen. Additionally, RH5-specific responses detected within total memory CD4+ T cells in the delayed regimen group (2B) were significantly higher than the delayed fractional regimen group (2C) post vaccination (Figure 3A V3 + 7 median 2B 0.32% vs 2C 0.19%, p = 0.03). In contrast to the ex vivo cTfh assay described for R21/Matrix-M^®, no significant correlations were observed between frequencies of RH5-specific cTfh cells and anti-RH5.1 serum IgG at V3 + 28 (data not shown).

Within the RH5.1/Matrix-M^® trial (NCT04318002), RH5 probes were used to identify circulating RH5-specific IgG+, IgM+ and IgA+ memory B cells (defined as shown in Figure 4A; Supplementary Figure 1) in PBMC samples at baseline (Day 0; D0) and 28 days after the final vaccination in the primary three-dose series (V3 + 28). Significant RH5-specific total, resting, or activated memory IgG+ B cell responses were observed at V3 + 28 in all groups (total memory IgG+ Figure 4B; Supplementary Figures 9A, B). In contrast to the NANP analyses in R21/Matrix-M^® trial (Supplementary Figure 5), but consistent with previously published adult analyses from the same trial (19), no significant IgM+ or IgA+ responses were detected post vaccination (Supplementary Figures 9C, D). However, at baseline, adults had a background signal with the RH5 probes in the IgM+ and IgA+ memory B cell subsets.

The delayed booster regimen induced substantially higher frequencies of RH5-specific B cells as compared to the monthly booster regimen across total, activated, and resting memory IgG+ populations (activated memory, Figure 4C; V3 + 28 medians: 0.40% Group 2A monthly, 4.68% Group 2B delayed, p = < 0.0001; Supplementary Figures 9E, F). Within activated memory IgG+ B cells, delayed boosting also appeared to induce a superior response than delayed fractional boosting (V3 + 28 median: 2.90% Group 2C delayed fractional, p = 0.0156; Figure 4D). There was no impact of high malaria pre-exposure on responses by any of the B cell populations measured (activated memory, Figure 4E; other data not shown). With respect to the relationship to serum antibody, RH5-specific responses within both total and activated memory IgG+ B cells correlated with matched time point serum anti-RH5.1 IgG (activated memory, Figure 4F; r = 0.48, p = 0.001; Supplementary Figure 9G). Unlike with NANP responses described above, these RH5-specific B cell responses continued to correlate strongly with serum IgG out to V3 + 2yr (Figure 4G; r = 0.58, p = 0.001, Supplementary Figure 9H).

Given these data on the strength of the RH5-specific B cell and serum antibody response in the context of delayed booster dosing of RH5.1/Matrix-M^®, we contextualised these findings with previously published data on RH5.1. The anti-RH5.1 serum IgG kinetics from baseline to V3 + 2yr were replotted for both RH5.1/Matrix-M^® Tanzanian children vaccinees (Figure 5A; (14)) that have been the focus of the RH5 component of this study, as well as the Tanzanian RH5.1/Matrix-M^® adult vaccinees from the same trial, alongside RH5.1/AS01_B adult vaccinees from an earlier UK clinical trial (Figure 5B; NCT04318002/ (4)).

While delayed boosting with Matrix-M^® in Tanzanian children or delayed fractional boosting with AS01_B in UK adults results in higher anti-RH5.1 serum IgG at V3 + 2yr as compared to monthly regimens, we did not see the same effect in Tanzanian adults receiving delayed fractional boosting with Matrix-M^®. Only the Tanzanian children who received the delayed regimen (2B) with Matrix-M^® and the UK adults who received the delayed fractional regimen (G3) with AS01_B maintained median anti-RH5.1 IgG concentrations above 50 µg/mL at V3 + 2yr (Figure 5C; see original trial publications for detailed comparisons of antibody responses between groups within each clinical trial) (4, 14). However, the mechanism behind these high V3 + 2yr anti-RH5.1 titres appears to be different: in the delayed fractional dosing AS01_B UK adults, the serum IgG percentage durability was approximately four-fold higher than in any other regimen (median = 44.5% of peak anti-RH5.1 µg/mL still present at V3 + 2yr versus < 10% in all other groups; Figure 5D). Conversely, the high V3 + 2yr titres in Tanzanian children seemed linked to a more robust peak response in the first instance, not an improvement in serum antibody persistence. Consistent with this, both V3 + 28 B cell and serum IgG responses correlate with late time point V3 + 2yr serum antibody (Figure 4F) but not percentage durability (Figure 5E) in the Tanzanian children. In contrast, the V3 + 28 B cell response in UK adults with AS01_B correlates with both late time point V3 + 2yr serum antibody and percentage durability (Figure 5F).

There is perhaps a yet to be understood interaction between adjuvant, age, antigen-specific B cell activation and serum antibody persistence. However, we also note that these are small sample sizes and that the median timing of the V3 + 2yr visit was earlier in the G2 and G3 RH5.1/AS01_B UK groups as compared to the RH5.1/Matrix-M^® Tanzanian groups due to differences in operational use of approved sampling windows during the respective clinical trials (NCT04318002 G2/G3 median = V3 + 510 days [range 386-826]; NCT04318002 median = V3 + 681 days [range 675-698]). Further work is therefore needed to validate these findings, ideally with larger sample sizes and – for the flow cytometry – head-to-head comparisons within the same assays.