Using fluorophore-conjugated DBPRII protein probes to detect DBPRII-specific cells, we evaluated the capacity of the viral vector and protein/adjuvant platforms to drive a range of B cell responses (gating strategy shown in Supplementary Figure 1A). Within the plasma cell population (CD19+CD27+CD38+; Figure 1A) – which peaked 7 days after final vaccination – significantly higher frequencies of DBPRII-specific cells were observed in protein/adjuvant vaccinees compared to heterologous viral vector vaccinees at FV+7 and FV+14. Within the memory IgG+ B cell population (CD19+CD27+IgG+ [excluding plasma cells]; Figure 1B), frequencies of DBPRII-specific cells were significantly higher in protein/adjuvant vaccinees at all post-vaccination time points. Significant post-vaccination responses were observed within both of these B cell populations with both platforms (intra-trial statistics not shown). Consistent with expected differences in kinetics of short-lived plasma cell (SLPC) and memory B cell responses, DBPRII-specific plasma cells peaked earlier, at FV+7, and then declined by FV+28, whilst frequencies of DBPRII-specific memory IgG+ B cells were better maintained in peripheral blood between FV+7 and FV+28 [with a trend towards a peak at FV+14 (5)].

We next compared DBPRII-responses between the dosing regimens of the protein/adjuvant platform. Here, we observed more robust plasma cell ( Figure 1C) and memory IgG+ B cell ( Figure 1D) responses following delayed booster dosing (PA-D) compared to monthly booster dosing (PA-M). Proliferation – as indicated by intracellular Ki67 staining – was higher in plasma cells than memory B cells across all groups and time points ( Supplementary Figures 1B, C). Differences between platforms and dosing regimens remained when we stratified between activated (CD21-CD27+) and resting (CD21+CD27+) memory IgG+ B cell populations ( Supplementary Figures 2A–D). Interestingly, Ki67 expression was significantly higher in activated memory IgG+ B cells in PA-D at FV+7 compared to PA-M or PA-DB, and at FV+14 compared to PA-M ( Supplementary Figures 1C, D). Very low memory IgA+ B cell responses were detectable (higher in protein/adjuvant than viral vector vaccinees), while responses within the IgM+ memory compartment were negligible ( Supplementary Figures 2E–H).

We sought to validate these protein/adjuvant delayed booster dosing-mediated differences with samples from an independent clinical trial with a different cohort (Tanzanian adults) and antigen (RH5). Here, we observed comparable plasma cell and memory IgG+ B cell kinetics and differences in proliferation ( Figures 1E, F; Supplementary Figure 3). Frequencies of RH5-specific cells within both populations were again higher in delayed booster dosing vaccinees but only reached statistical significance (by Mann Whitney test) for RH5-specific memory B cells at FV+14, likely related to greater intra-group variation, compared to the DBPRII vaccinees. Trends were comparable but not significant when activated and resting memory IgG+ B cells were analysed separately ( Supplementary Figure 4). To note, while background with the RH5 probes was more variable than observed with the DBPRII probes, or as compared to previous studies with the RH5 probe protocol (5, 12), standardisation of RH5 probe gating between all samples and parent populations retains confidence in the data interpretation. No significant RH5-specific post-vaccination responses were observed for either the monthly or delayed boosting groups within the memory IgA+ or IgM+ populations, but further investigation may be of interest as trends appear different to memory IgG+ B cells ( Supplementary Figure 4).

To supplement the more traditional approach above, clustering was next performed with CITRUS to agnostically define B cell populations for further analysis within both the DBPRII and RH5 datasets. For the DBPRII trial samples, 33 clusters were identified in CITRUS within live single (B cell-enriched) lymphocytes. Several sets of clusters had very similar marker expression patterns, resulting in consolidation for FlowJo gating strategies (see Methods). This gave a total of 7 agnostically-defined populations to reanalyse for DBPRII-specific responses ( Table 2). Likewise, 36 clusters were identified in the RH5 trial samples, consolidating to 10 new populations for reanalysis of RH5-specific responses ( Table 2). Populations where significant differences in antigen-specific responses were detected post-vaccination or between dosing regimens are shown in Figure 2 (gating shown in Supplementary Figure 5). DBPRII “Population 2” and RH5 “Population 8” were equivalent, while all other populations identified were unique to the separate trials.

In the DBPRII trial clusters, significant differences were observed between monthly and delayed protein/adjuvant dosing in “Population 1” (CD19+CD20+CD21+CD27+CD138-CD38-IgM-IgA-IgG+; Figure 2B; Supplementary Figure 5) and trends between monthly and delayed or delayed/booster (PA-DB) dosing in “Population 2” (CD19-CD20-CD21-CD27-CD138-CD38+IgM-IgA-IgG+; Figure 2C; Supplementary Figure 5). While “Population 1” is a subset of the CD19+CD21+CD27+IgG+ resting memory population analysed above (Supplementary Figure 2), “Population 2” was not included in the previous analysis. Responses within other clusters and frequencies of each cluster within total live (B cell-enriched) lymphocytes are shown in Supplementary Figure 6.

In the new RH5 trial populations, significant responses were observed in “Population 12” (Figure 2E; Supplementary Figure 5; CD19+CD20+CD21+CD27+CD138-CD38+IgM-IgA-IgG+; again, a subset of the CD19+CD21+CD27+IgG+ resting memory in Supplementary Figure 4) in both monthly and delayed booster protein/adjuvant dosing, with higher responses in delayed boosting vaccinees. Significant differences in post-vaccination RH5-specific responses were not detected between groups in the remaining populations (Supplementary Figure 7). No significant responses were observed within “Population 8” (CD19-CD20-CD21-CD27-CD138-CD38+IgM-IgA-IgG+; equivalent “Population 2” in the DBPRII trial) in the delayed boosting vaccinees.

We have previously published ELISA data on serum anti-DBPII total IgG (against the Sal I strain), with an emphasis on comparison of FV+14 responses. There, we observed significantly higher titres with a delayed protein/adjuvant dosing regimen compared to viral vectors (11). Now, we extend these analyses to compare the impact of platform and regimen on different isotypes/subclasses, (Figures 3A–F; Supplementary Figure 8), durability of serum antibody (Figures 3G, H), and immunodominance of subdomain 3 (sd3; Supplementary Figure 8). The isotype and subclass analyses showed IgG1, IgG3, IgA, and IgA1 responses in both platforms, while IgG4 was detectable solely in the protein/adjuvant vaccinees. No statistically significant post-vaccination IgM responses were observed within individual groups (Supplementary Figure 8), while no detectable IgG2 or IgA2 was observed in any sample (not shown). The protein/adjuvant platform also induced higher IgG1, IgG3, IgA, and IgA1 (and IgG4) responses compared to viral vectors (Supplementary Figures 8A–F). Within the protein/adjuvant platform, the DBPRII-specific IgG1, IgG4, and IgA1 response was significantly higher in delayed dosing vaccinees (Figures 3A, C, E) but comparable between regimens for IgG3 and IgA (Figures 3B, D). Median IgM responses were higher in monthly dosing but not statistically significant (Figure 3F). Interestingly, we also observed increased anti-DBPRII serum total IgG durability in delayed dosing regimens with both vaccine platforms; a significantly higher fold-change in total serum IgG was observed > 3 months following the peak response in delayed booster vaccinees (VV-D, PA-D; median = 0.45) compared to monthly dosing vaccinees (VV-M, PA-M; median = 0.15; Figure 3G). Enhanced serum durability over > 3 months was also observed with IgG1 in the isotype and subclass analyses (Figure 3H). Analyses with even later serum time points (e.g., 6 months or 1 year after final vaccination) were not feasible in the current study but would be of interest to understand longer-term durability. Finally, delayed dosing with PvDBPII appears to have no impact on the immunodominance of the sd3 region (Supplementary Figure 8). Equivalent serological analyses from the RH5.1/Matrix-M™ trial will be the focus of a separate report (Silk et al., manuscript in preparation).

In our main trial report, we showed higher frequencies of IFN-γ-producing effector memory (CD45RA-CCR7-) CD4+ T cells at FV+14 in viral vectors compared to protein/adjuvant vaccinees, with no significant differences observed between monthly and delayed dosing regimens with the latter platform (11). Here, we extended these analyses to include further time points as well as IL-2/TNF-α/IL-5/IL-13 intracellular cytokine detection (in addition to IFN-γ), allowing a more nuanced comparison of responses between vaccine platform and regimen. Looking first at all DBPRII-specific effector memory CD4+ T cells – based on the secretion of any cytokine following stimulation with the DBPRII peptide pool (Table 3) – we observed significantly higher frequencies in protein/adjuvant vaccinees at FV+7 compared to viral vector vaccinees but no difference between dosing regimens (Figures 4A, B). Moving from this measure of the magnitude of the DBPRII-specific CD4+ T cell response to more qualitative read-outs, within the same effector memory CD4+ T cell population, we next assessed total Th1 (defined as IFN-γ and/or IL-2 and/or TNF-α) and Th2 responses (IL-5 and/or IL-13). Both platforms induced Th1 cytokine production and while there were no significant differences between platforms in the total Th1 response (Figure 4C), viral vector vaccinees trended towards higher IFN-γ production versus higher IL-2 with protein/adjuvant vaccinees (Supplementary Figure 9). No significant differences between protein/adjuvant dosing regimens were detected when Th1 cytokines were analysed together (Figure 4D) or individually (data not shown). In contrast, the protein/adjuvant vaccines drove higher frequency total Th2 responses, comparable across different dosing regimens (Figures 4E, F). Trends were similar for both IL-5 and IL-13, although the magnitude of DBPRII-specific IL-5-producing cells was higher (Supplementary Figure 9).

Finally, we similarly quantified the effector memory CD8+ T cell response. Here, Th1 responses were observed only within the viral vector platform, whether by total Th1 or by individual cytokine (Supplementary Figure 10).

To confirm the biological relevance of increased B cell immunogenicity in delayed dosing vaccinees, we performed Spearman correlations with in vivo growth inhibition of blood-stage parasites following P. vivax CHMI (comparisons of parasite growth dynamics by group have previously been reported (11)). The magnitude of the circulating DBPRII-specific memory IgG+ B cell response correlated strongly with IVGI (Figure 5A), as did the frequency of DBPRII-specific plasma cells (Figure 5B). Since IVGI is calculated over a relatively short time frame following parasite inoculation, we were also interested in assessing the relationship between these B cell responses and the fold change in serum antibody between the peak (FV+14) and a later time point (C+96, approximately 16-weeks later; Figure 3G). DBPRII-specific IgG+ memory B cells again correlated strongly with durable serum anti-DBPRII IgG responses (Figure 5C), while no association was observed with peak FV+7 plasma cells (Figure 5D). Neither IVGI nor serum antibody maintenance correlated with DBPRII-specific Th2 effector memory CD4+ T cell responses (data not shown), suggesting a lesser role for T cell immunogenicity in DBPRII vaccine-mediated protection.

While bone marrow aspirates were not taken in the DBPRII clinical trials, we had the rare opportunity to investigate bone marrow plasma cells with the RH5.1/Matrix-M™ adult vaccinees – the first bone marrow analyses in the context of malaria vaccination (13). Here, we observed a trend to higher frequencies of RH5-specific cells within total bone marrow B cells in the protein/adjuvant delayed dosing regimen (Figure 6A), which correlated strongly with matched time point serum anti-RH5 IgG (Figure 6B). We were also interested to see whether the frequencies of RH5-specific cells within circulating B cell populations of interest correlated with vaccine-specific seeding of the bone marrow. We therefore performed additional Spearman correlation analyses between the frequency of RH5-specific cells within “Population 12” (CD19+CD20+CD21+CD27+CD138-CD38+IgA-IgM-IgG+; the B cell population with greatest differences between regimens) at both matched time points (i.e., V1 or FV+28; Figure 6C) or with FV+14 (Figure 6D). We observed strong correlations between RH5-specific circulating memory B cells and bone marrow plasma cells at both time points. Similar results were obtained if correlations were performed with RH5-specific CD19+CD27+IgG+ B cells (Figure 1F; data not shown). RH5-specific bone marrow B cells did not correlate with circulating RH5-specific CD27+CD38+ (plasma) cells at FV+7 (data not shown).

This study focused on the comparison of immune responses between groups vaccinated with the Plasmodium vivax antigen DBPRII with different platforms and dosing regimens [Table 1, (11)]. In brief, two Phase I/IIa vaccine efficacy trials (NCT04009096 and NCT04201431) were conducted in parallel at a single site in the UK (Centre for Clinical Vaccinology and Tropical Medicine, University of Oxford). NCT04009096 was an open-label trial to assess the ChAd63 and MVA viral-vectored vaccines encoding PvDBPII (VV-PvDBPII), while the NCT04201431 trial assessed the protein vaccine PvDBPII in Matrix-M™ adjuvant from Novavax (PvDBPII/M-M). NCT04009096 viral vectors were administered at 0 and 2 months (VV-M), or in a delayed dosing regimen, at 0, 17, and 19 months (VV-D). For NCT04201431, the protein/adjuvant was administered monthly (0, 1, 2 months; PA-M), or in a delayed dosing regimen, at 0, 1, 14 months (PA-D). A subset of vaccinees from the protein/adjuvant delayed dosing regimen returned for an additional booster at 19 months (PA-DB). ChAd63-PvDBPII was administered at a dose of 5×10¹⁰ viral particles, MVA-PvDBPII at 2×10⁸ plaque forming units, and PvDBPII protein at 50μg mixed with 50μg Matrix-M™. Delayed regimens were due to trial halts during the pandemic. Eligible vaccinees were healthy, Duffy-positive, malaria-naïve adults, aged 18 to 45 years. Trials were approved by the UK National Health Service Research Ethics Services (REC; references 19/SC/0193 and 19/SC/0330) and the UK Medicine and Healthcare products Regulatory Agency (MHRA; reference CTA 21584/0414/001-0001 and CTA 21584/0418/001-0001). All vaccinees gave written informed consent.

This study also includes analyses of samples from a further Phase Ib clinical trial with Plasmodium falciparum vaccine candidate RH5.1 (50µg) in Matrix-M™ adjuvant (50µg) in a malaria-endemic setting in Tanzania (NCT04318002). RH5.1/Matrix-M™ was administered either monthly, at 0, 1, and 2 months (M), or in a delayed booster regimen, at 0, 1, and 6 months (D). The final vaccination in the delayed booster regimen was given at a fractionated RH5.1 antigen dose of 10µg, rather than 50µg (the dose of Matrix-M™ was not fractionated and remained at 50 µg). Eligible vaccinees were healthy adults (negative for malaria by blood smear at screening), aged 18 to 45 years. The trial was approved by the Tanzanian Medicines and Medical Devices Authority (reference TMDA0020/CTR/0006/01), the National Institute for Medical Research (references NIMR/HQ/R.8a/Vol.IX/3537 and NIMR/HQ/R.8c/Vol.1/1887), the Ifakara Health Institute Institutional Review Board (reference IHI/IRB/No:49-2020), and the Oxford Tropical Research Ethics Committee (reference 9-20). All vaccinees gave written informed consent.

For the DBPRII clinical trial, antigen-specific total IgG, IgG3, IgG4, IgA, IgA1, and IgM titres were determined by standardised ELISA in accordance with published methodology (20). Nunc MaxiSorp™ flat-bottom ELISA plates (44-2404-21, Invitrogen) were coated overnight with 2µg/mL (for total IgG titres) or 5µg/mL (for IgG1, IgG3, IgG4, IgA, IgA1, and IgM titres) of DBPRII SalI protein or 2µg/mL of sd3 protein in PBS. DBPRII protein was produced as previously described (11), while sd3 protein was produced by transient transfection of Expi395F cells with a plasmid encoding a monoFc, DBP Subdomain 3 (sequence as per UniProt P22290 PVDR residues P387-S508), and a C-terminal c-tag. The monoFc was cleaved using TEV protease and sd3 was purified by affinity chromatography (c-tag) and size exclusion chromatography. Plates were washed with washing buffer composed of PBS containing 0.05% TWEEN® 20 (P1379, Sigma-Aldrich) and blocked with 100µL of Starting Block™ T20 (37538, ThermoFisher Scientific). After removing the blocking buffer, standard curve and internal controls were diluted in blocking buffer using a pool of high-titre vaccinee plasma or serum specific to each antigen and isotype or subclass being tested, and 50µL of each dilution was added to the plate in duplicate. Test samples were diluted in blocking buffer to a minimum dilution of 1:50 (or 1:100 for total IgG) and 50µL was added in triplicate. Plates were incubated for 2h at 37°C (or 20°C for total IgG) and washed in washing buffer. An alkaline phosphatase-conjugated secondary antibody was diluted at the manufacturer’s recommended minimum dilution for ELISA in blocking buffer. The antibody used was dependent on the isotype or subclass being assayed and were as follows: IgG-AP (A3187, Thermo Scientific), IgG1 Fc-AP (9054-04, Southern Biotech), IgG3 Hinge-AP (9210-04, Southern Biotech), IgG4 Fc-AP (9200-04, Southern Biotech), IgA-AP (2050-04, Southern Biotech), IgA1-AP (9130-04, Southern Biotech), and IgM-AP (2020-04, Southern Biotech). Then, 50µL of the secondary antibody dilution was added to each well of the plate and incubated for 1h at 37°C (or 20°C for total IgG). Plates were developed using PNPP alkaline phosphatase substrate (N2765, Sigma-Aldrich) for 1-4h at 37°C (or approximately 15 minutes at 20°C for total IgG). Optical density at 405 nm was measured using an ELx808 absorbance reader (BioTek) until the internal control reached an OD405_nm of 1. The reciprocal of the internal control dilution giving an OD405_nm of 1 was used to assign an AU value of the standard. Gen5 ELISA software v3.04 (BioTek) was used to convert the OD405_nm of test samples into AU values by interpolating from the linear range of the standard curve fitted to a four-parameter logistics model. Any samples with an OD405_nm below the linear range of the standard curve at the minimum dilution tested were assigned a minimum AU value according to the lower limit of quantification of the assay. For the assessment of IgG2 and IgA2 responses, no anti-DBPRII IgG2 or IgA2 samples were available for standard curve generation. Responses were measured on plates coated with 5µg/mL DBPRII. Four wells were also coated with RH5.1 protein for development control wells. Each sample was tested in duplicate with six negative control serum samples and two development control serum samples on the RH5.1 coated wells from a previous RH5.1/AS01 vaccine trial (4). Secondary antibodies used were IgG2 Fd-AP (9080-04) and IgA2-AP (9140-04). The assay was carried out as above and plates were developed for 2-4h at 37°C.

For the RH5 clinical trial, serum antibody levels to full-length RH5 protein (RH5.1) were assessed by standardised ELISA methodology as previously described (4, 21). In brief, the reciprocal of the test sample dilution giving an optical density of 1.0 at 405nm (OD405_nm) was used to assign an ELISA unit value of the standard. The standard curve and Gen5 software v3.04 (Agilent) were then used to convert the OD405_nm of test samples to arbitrary units (AU). Responses are reported in µg/mL using a conversion factor from AU generated by calibration-free concentration analysis (CFCA), as previously reported (21).

A single 10mL bone marrow aspirate was collected in EDTA per vaccinee in the RH5 trial. Bone marrow mononuclear cells (BMMNC) were purified from aspirates by density centrifugation on Lymphoprep (1114545, Axis Shield) following passage through a 70µm nylon cell strainer (542070, Greiner Bio-One Ltd). BMMNC were cryopreserved for future use in FCS (S1810, Biosera) with 10% DMSO (D2650, Sigma). Samples were subsequently thawed and enriched for B cells using a Pan B Cell Enrichment Kit (19554, Stemcell) for the detection of RH5-specific plasma cells with an antibody-secreting cell (ASC) ELISPOT. In brief, enriched bone marrow B cells were aliquoted onto MAIP ELISpot Plates (MAIPS4510, Millipore) coated with 5µg/mL RH5 protein, PBS (negative control), or 50µg/mL polyvalent goat anti-human immunoglobulin (positive control; H1700, Caltag). Plates were incubated for 16-18h at 37°C prior to cell removal and incubation with anti-human IgG conjugated to alkaline phosphatase (γ-chain specific; 401442, Calbiochem). Finally, plates were developed with BCIP/NBT (M0711A, Europa Bioproducts) and read on an AID ELISPOT Plate Reader (AID). Note, for consistency, all ASCs are referred to as plasma cells throughout this report.

In vivo growth inhibition (IVGI) has been reported elsewhere for the DBPRII trial (11). In brief, vaccinees underwent controlled human malaria blood-stage infection 2-4 weeks after their final vaccination and in parallel with unvaccinated infectivity controls. Controlled blood-stage infection was initiated by intravenous injection of blood infected with the PvW1 clone of P. vivax. Participant blood parasitemia was measured in real time by qPCR of the 18S ribosomal RNA gene (22). IVGI was calculated for each vaccinee as the percentage reduction in parasite multiplication rate (PMR) relative to the mean PMR of the unvaccinated controls. PMR was modelled for each vaccinee based on log₁₀ transformed qPCR data of the P. vivax 18S ribosomal RNA gene, using a mean of three replicate qPCR results for each vaccinee per time point (11, 22, 23). Further details of the IVGI/PMR methodology in this trial can be found in the Supplementary Appendix of the primary trial report (11).

Comparisons were performed between regimens with (two-tailed) Mann-Whitney tests or a Kruskal-Wallis test with Dunn’s correction for multiple comparisons (GraphPad Prism v9). A p-value of < 0.05 was considered statistically significant.

Specifics of the CITRUS analyses are outlined in further detail below. For all data, relevant statistical tests and sample sizes are specified in the figure legends.

Raw fcs files with file-internal compensation (i.e. acquisition-defined) from the B cell flow cytometry assays with DBPRII or RH5 trial samples were uploaded separately into Cytobank. Live single (B cell-enriched) lymphocytes were gated and then analysed with CITRUS using equal event sampling. All fluorophore channels were used for clustering with the exception of the probes and the viability stain. All groups and time points were run in a single CITRUS analysis per trial. Median expression values for each fluorophore for each CITRUS-defined cluster were exported per sample. Average expression values across all samples for each fluorophore were then calculated per cluster to define FlowJo gating strategies. Clusters with shared gating strategies were combined into new populations (Table 2) for re-analysis in FlowJo. Ki67 was included in the CITRUS clustering, but median expression values did not facilitate defining a dichotomous grating strategy and, thus, Ki67 was not utilised in the population definitions. Samples were excluded from analysis if there were <50 cells in the parent population.