Introduction

Plasmodium vivax accounts for majority of malaria cases outside sub-Saharan Africa, where P. falciparum is more predominant1. For many decades, P. vivax malaria was considered to be a ‘benign’ infection. However, recent studies have reported significant incidence of severe symptoms in P. vivax malaria2,3. Ineffective control of P. vivax could pose an important public health threat to communities in which P. vivax is endemic or at risk of emerging. Current efforts to control malaria are less effective in reducing P. vivax compared to P. falciparum due to the unique biology of P. vivax4. For example, the dormant liver stage of P. vivax known as the hypnozoite can reactivate and cause blood stage infection weeks, months or even years after the initial infection. This dormant stage, which cannot be detected, greatly contributes to P. vivax prevalence5. In addition, P. vivax gametocytes appear early in the blood stage, so the parasite can be transmitted even before the first symptoms appear and treatment is administered to the patient6. An effective vaccine that can prevent P. vivax infection, reduce blood stage replication and protect against disease could greatly help efforts to control and eventually eliminate P. vivax.

During the blood stage, the invasive form of P. vivax known as the merozoite repeatedly infects and multiplies within reticulocytes causing the clinical symptoms of malaria. The invasion of reticulocytes is mediated by the interaction of the P. vivax Duffy binding protein (PvDBP) and the Duffy antigen receptor for chemokines (DARC)7,8. The amino-terminal cysteine-rich region II of PvDBP (PvDBPII), serves as the binding domain of this invasion ligand9. Importantly, Duffy-negative individuals remain largely resistant to P. vivax infection7. Numerous reports indicate PvDBPII is highly polymorphic10,11,12, suggesting that it is under intense immune pressure. However, the binding residues in PvDBPII that make contact with DARC are highly conserved13,14,15. The binding residues, which include positively charged as well as hydrophobic amino acids, assemble on the PvDBPII surface to form a receptor recognition site that is fully exposed to neutralization by antibodies15. Moreover, since the binding residues are conserved, antibodies that target them are predicted to be strain-transcending and capable of neutralizing diverse P. vivax strains. Indeed, upon natural exposure a small percentage of individuals develop high titre binding inhibition antibodies against PvDBPII that are cross-reactive16,17. Importantly, the presence of such high titre binding inhibitory antibodies against PvDBPII is associated with reduced risk of P. vivax infection16,17. However, the low frequency of individuals in the population with such binding inhibitory antibodies indicates that natural exposure is not effective at eliciting protective anti-PvDBPII antibodies.

Pre-clinical studies demonstrated that immunization of animals with recombinant PvDBPII readily elicits high titre binding inhibitory antibodies that cross-react with multiple variants18,19. A vaccine based on PvDBPII could thus potentially elicit binding inhibitory antibodies that limit blood-stage growth of diverse P. vivax isolates. To date, two delivery platforms developed for PvDBPII have been tested in clinical trials. These include recombinant PvDBPII protein formulated with adjuvants such as glucopyranosyl lipid adjuvant-stable emulsion (GLA-SE)20 and PvDBPII delivered by two viral vectors, namely, replication-deficient chimpanzee adenovirus serotype 63 (ChAd63) for priming followed by modified vaccinia virus Ankara (MVA) for boosting21. The vaccine candidate PvDBPII/GLA-SE was shown to be safe and no related adverse events were associated with the vaccine20. In addition, PvDBPII-specific antibodies were elicited and shown to inhibit DARC-binding by diverse PvDBPII variants20. In a Phase I trial with UK adult volunteers, the viral-vectored PvDBPII showed similar safety and immunogenicity results21. Recently, PvDBPII protein reformulated in Matrix-MTM adjuvant (PvDBPII/Matrix-MTM) from Novavax, and PvDBPII expressed by the same viral vectors, ChAd63/MVA, were tested in parallel in a Phase I/IIa blood stage challenge clinical trial using a heterologous P. vivax isolate to evaluate efficacy22. Interestingly, volunteers who were vaccinated with PvDBPII/Matrix-MTM in a delayed dosing schedule showed significant reduction in the parasite multiplication rate (PMR) compared to unvaccinated volunteers. None of the other groups in the trial, including PvDBPII delivered by viral vectors, showed any reduction in PMR. This result provided proof-of-concept that immunization with PvDBPII can elicit immune responses that can impair in vivo growth of a heterologous P. vivax strain. Here, we report extended analyses of the antibody responses, including functional analysis of anti-PvDBPII antibodies, PvDBPII-antibody binding kinetics, recognition of native PvDBP in P. vivax schizonts as well as binding to C1q and Fc gamma receptors, to identify immune correlates of protection following immunization with vaccine candidates based on PvDBPII.

Results

Functional analysis of antibody response to PvDBPII in a challenge trial

A schematic describing the challenge trials to evaluate efficacy of PvDBPII administered using two approaches for vaccine antigen delivery, namely, recombinant PvDBPII SalI protein adjuvanted with Matrix-MTM, PvDBPII/Matrix-MTM, and PvDBPII SalI delivered by viral-vectors, ChAd63 followed by MVA (ChAd63/PvDBPII and MVA/PvDBPII)22 is shown in Supplementary Fig. 1.

As described earlier22, volunteers received priming immunizations with PvDBPII/Matrix-MTM and ChAd63/PvDBPII in January 2020. The trial was halted in March 2020 due to the Covid-19 pandemic. Twelve volunteers in the PvDBPII/Matrix-MTM arm (n = 12) received the priming dose and one booster dose of PvDBPII/Matrix-MTM at 1 month, whereas ten volunteers from the viral vectors arm (n = 10) had only received the priming dose of ChAd63/PvDBPII by the time the trial was put on halt in March 2020. The trial was restarted in April 2021 by which time several volunteers had dropped out of the trial. Six volunteers who received the first 2 monthly doses of the recombinant PvDBPII/Matrix-MTM in January and February 2020 (days 0 and 28), got the final boost after 14 months (day 440) and underwent Controlled Human Malaria Infection (CHMI) with P. vivax blood-stage parasites (clonal strain PvW1)23 in May 2021. We refer to this group as PvDBPII/M-M (0-1-14 months). Two volunteers, who received the ChAd63/PvDBPII prime in January 2020 (day 0), received an additional ChAd63/PvDBPII followed by MVA/PvDBPII at 17 and 19 months (days 530 and 586), respectively. We designated this group as VV-PvDBPII (0-17-19 months). New volunteers were enrolled in the trial and four of them received PvDBPII/Matrix-MTM in a regimen of 0, 1 and 2 months (days 530, 558 and 586) and three others received ChAd63/PvDBPII followed by MVA/PvDBPII scheduled at 0 and 2 months (days 530 and 586). These short regimen groups (with vaccines given as originally intended) are referred to as groups PvDBPII/M-M (0-1-2 months) and VV-PvDBPII (0–2 months), respectively. Volunteers from PvDBPII/M-M (0-1-2 months), VV-PvDBPII (0-17-19 months) and VV-PvDBPII (0–2 months) groups underwent CHMI in October 2021. Regardless of the vaccine regimen, all volunteers underwent CHMI 2–4 weeks after the last immunization. Seven individuals who did not received any vaccine were included as the control group in the CHMI conducted in May 2021 and four nonvaccinated volunteers underwent CHMI in October 2021 (infectivity controls). Both CHMI challenges involved infection with blood-stage parasites of the heterologous P. vivax monoallelic isolate W1 (or PvW1)23. Parasite growth was evaluated in immunized and control unvaccinated groups by RT-PCR at different time points during the CHMI to determine vaccine efficacy22.

As reported earlier22, only the PvDBPII/M-M (0-1-14 months) group showed a significant reduction of the PMR compared to infectivity controls (median PMR of 3.2 for PvDBPII/M-M (0-1-14 months) group and median PMR of 6.8 for controls, p = 0.002). This represented a 53% reduction of the median PMR in PvDBPII/M-M (0-1-14 months) group compared to controls. No other vaccine group had a significant impact on the PMR. The PMR data from the challenge trial22 is summarized for clarity in Table 1.

Table 1 In vivo parasite multiplication rate (PMR) of vaccinated groups and controls following blood stage challenge with P. vivax PvW1.
Full size table

ELISA titres for recognition of the homologous PvDBPII SalI domain and PvW1 (the heterologous PvDBPII domain from the P. vivax strain used for CHMI) by sera from immunized volunteers were determined through the course of the trial (Fig. 1a). Anti-PvDBPII antibody titres peaked 2 weeks after final boost in all vaccinated volunteers. The antibody titres of sera in the PvDBPII/M-M (0-1-14 months) group increased significantly after the delayed second boost at day 440 for both PvDBPII SalI and PvW1. In case of PvDBPII/M-M (0-1-2 months), VV-PvDBPII (0-17-19 months) and VV-PvDBPII (0–2 months) groups, there was no significant increase after the second monthly boost. The anti-PvDBPII antibody titres in group PvDBPII/M-M (0-1-14 months), which was the only group to show a significant PMR reduction, remained stable after second boost and challenge with only a marginal decline over a 4-month period after the delayed second boost (Fig. 1a). In the groups immunized with viral-vectored PvDBPII, the anti-PvDBPII antibodies reached a peak after MVA boost prior to challenge (Fig. 1a).

Fig. 1: ELISA reactivity and binding inhibitory activity of sera from volunteers after immunization with protein-in-adjuvant or viral-vectored PvDBPII vaccines.
figure 1

a ELISA titres for recognition of PvDBPII SalI by sera from volunteers at different timepoints over the course of the trial reported as median and range for each vaccine group are shown. Black arrows show the time points of the immunizations. Shaded grey areas represent the period after blood stage P. vivax challenge. b ELISA recognition titres and binding inhibition titres for PvDBPII variants SalI and PvW1 for all volunteers at C-2. Medians are shown in horizontal bars, **p < 0.01, Kruskal–Wallis test with Dunn’s correction for multiple comparisons.

Full size image

Next, we compared the antibody responses at C-2, 2 days before challenge, against PvDBPII SalI and PvW1. Sera from the immunized volunteers had similar ELISA recognition titres for PvDBPII SalI and PvW1 in all the groups (Fig. 1b). Also, at C-2, the antibody titres of group PvDBPII/M-M (0-1-14 months) for recognition of PvDBPII were not significantly different from titres of groups PvDBPII/M-M (0-1-2 months) or VV-PvDBPII (0-17-19 months) for either PvDBPII SalI or PvW1. The only statistically significant differences were found for antibody titres for recognition of both PvDBPII SalI and PvW1 between groups PvDBPII/M-M (0-1-14 months) and VV-PvDBPII (0–2 months) (Fig. 1b). Control sera from volunteers that did not receive any immunization were seronegative for recognition of PvDBPII SalI and PvW1 (arbitrary units, AU ≤ 1) at all time points including C-2.

We also assessed the ability of anti-PvDBPII sera to block DARC-receptor binding by PvDBPII SalI and PvW1. Sera from infectivity controls at C-2 and day 0 sera from volunteers collected prior to immunization showed DARC-binding inhibition of ≤30% for PvDBPII SalI and PvW1 at the lowest serum dilution of 1:10 that was tested in the binding inhibition assay. Binding inhibition for all vaccine groups at C-2 tested at 1:10 dilution was greater >90%. Anti-sera of PvDBPII/M-M (0-1-14 months) group with delayed boost of PvDBPII/Matrix-MTM tended to have the highest binding inhibition titres at C-2 (Fig. 1b). However, only comparison between PvDBPII/M-M (0-1-14 months) and VV-PvDBPII (0–2 months) reached statistical significance. There was no difference in binding inhibition titres for PvDBPII SalI and PvW1 in all vaccinated groups at C-2.

In order to examine the stability of anti-PvDBPII antibodies, we analyzed correlations in antibody recognition titres and binding inhibition titres at time points C-2, C + 56 and C + 96 (Supplementary Fig. 2). The antibody recognition titres and binding inhibition titers were highly correlated between time points C-2 and C + 56 or C + 96.

PvDBPII-specific antibodies recognize native antigen in P. vivax schizonts

The ability of anti-PvDBPII sera to recognize native PvDBP in P. vivax schizonts was determined by IFA. All sera from vaccinated volunteers collected on day C-2 showed apical staining in merozoites within mature P. vivax schizonts (Fig. 2a). Day 0 control sera collected prior to immunization (D0) did not show any specific signal in IFA with P. vivax schizonts (Fig. 2a). Sera collected at C-2 from the PvDBPII/M-M (0-1-14 months) group had the highest reactivity at dilutions of 1:500 and 1:5,000 (Fig. 2b). Day C-2 sera from the VV-PvDBPII (0-17-19 months) group had higher signal than sera collected at C-2 from groups PvDBPII/M-M (0-1-2 months) and VV-PvDBPII (0-2 months) at dilution of 1:5,000. Sera at C-2 from PvDBPII/M-M (0-1-2 months) group had stronger reactivity than VV-PvDBPII (0-2 months) at 1:500 (Fig. 2b). Based on signal intensity, we can order the groups from highest to lowest serum reactivity in IFA as follows: PvDBPII/M-M (0-1-14 months) > VV-PvDBPII (0-17-19 months) > PvDBPII/M-M (0-1-2 months) > VV-PvDBPII (0-2 months).

Fig. 2: Reactivity of sera from immunized volunteers with P. vivax schizonts.
figure 2

a Representative images of P. vivax schizonts incubated with sera from an individual of the PvDBPII/M-M (0-1-14 months) group at dilution 1:5000. Sera at time point C-2 showed apical staining to PvDBP (green) in P. vivax schizonts compared to sera collected prior to initial immunization (D0). b Signal intensity in IFA for day C-2 and D0 sera from different groups tested at different dilutions. The fluorescence intensity of each sample was scored as: +++, for high intensity; ++, medium intensity; +, low intensity; and -, no signal. Some IFA samples that were not done are designated as ND.

Full size image

Immune correlates for PMR reduction

To explore the immune processes that are associated with the PMR outcome, we measured diverse antibody functions and determined their correlation to PMR reduction calculated for each individual as: 100 – (test PMR x 100)/(PMR of unvaccinated controls). Multivariate analysis was performed with all the variables for both PvDBPII alleles SalI and PvW1 using a frequently-used machine learning tool, the classification algorithm Boruta24, in order to select the features that significantly contribute to the PMR outcome and rank them according to their contribution or importance in predicting the PMR outcome.

Sera from timepoint C-2 was used to evaluate i) the PvDBPII-binding kinetics of the polyclonal antibodies (affinity, association and dissociation constants or KD, ka and kdis respectively); ii) antibody isotypes and subclasses; iii) binding capacity for complement component C1q and Fc gamma receptors or FcγRs; iv) avidity; and v) ELISA recognition titres and binding inhibition titres for PvDBPII. Both datasets corresponding to either antibody functions specific to PvDBPII SalI (Fig. 3a) or to PvDBPII PvW1 (Fig. 3c) showed significant correlations between most of the variables. We observed positive correlations between C1q binding, FcγR binding, isotypes, antibody recognition titres and binding inhibition titres in both datasets (Fig. 3a, c). Negative correlations were observed for the affinity and dissociation constants, KD and kdis, in the PvDBPII SalI dataset (Fig. 3a), whereas all three PvDBPII-antibody binding kinetic constants in the PvDBPII PvW1 dataset showed negative correlations with the rest of the dataset (Fig. 3c). The IgG3 and IgM readouts did not show correlations to most of the features of both datasets (Fig. 3a, c).

Fig. 3: Variable correlations and feature selection of the antibody functions measured in the study.
figure 3

a Correlations between the anti-PvDBPII SalI antibody functions are shown. Correlation coefficients (shown in a double red-blue gradient) for each comparison were calculated using Spearman’s rank correlation tests. p values for each significant correlation are indicated, *p < 0.05 **p < 0.01 ***p < 0.001 ****p < 0.0001. b Importance plot for classification of variables specific to PvDBPII SalI that significantly contribute to the PMR reduction using the Boruta algorithm. The importance of each variable is defined as the Z-score of the mean decrease accuracy (normalised permutation importance). Blue boxes correspond to the minimal, average, and maximum Z-scores of shadow features. The variables that contribute significantly (green) or not (red) to the PMR reduction are shown. Boxplots show median Z-score (horizontal bar), interquartile range (boxes), range (whiskers), and outliers (open circles). c Correlations between the PvDBPII PvW1 antibody functions are depicted as in a. d Importance plot for classification of variables specific to PvDBPII PvW1 are depicted as in b.

Full size image

PMR reduction showed a significant correlation in the PvDBPII SalI dataset with antibody recognition titre, binding inhibition titre, kdis, IgG1, IgG2, IgG4, IgA, C1q binding and FcγR binding (Fig. 3a). In case of the PvDBPII PvW1 dataset, the PMR reduction showed significant correlations with the variables mentioned above for PvDBPII SalI dataset, KD and avidity (Fig. 3c). We proceeded to identify the variables that have a significant contribution to the PMR reduction outcome in both datasets. The Boruta algorithm selected as important the same 11 readouts in both datasets (Fig. 3b and d). These important variables include binding inhibition titres, FcγR binding and C1q binding, IgG1, IgG2, IgG4, kdis and antibody titres. Binding inhibition titre appeared as the variable with the highest importance in the PvDBPII SalI dataset (Fig. 3b). In case of the PvDBPII PvW1 dataset, this variable also had high importance in the dataset (Fig. 3d). Correlations of the PvDBPII SalI and PvW1 binding inhibition titres to PMR reduction showed high significance (p ≤ 0.0014) (Fig. 4a). This feature was also highly correlated when both PvDBPII alleles were compared (Fig. 4b), indicating that immunization with PvDBPII SalI elicits antibodies that cross-react with PvDBPII PvW1. We modeled PMR reduction using a Random Forest regression with binding inhibition titres specific to the two PvDBPII alleles. The importance of the two features as a PMR reduction predictor was similar and no statistical difference was observed (Fig. 4c), indicating both binding inhibition titers specific to SalI and PvW1 contribute equally to predict the PMR outcome.

Fig. 4: Correlation of binding inhibition titre and PMR reduction.
figure 4

a Individual correlations of the PMR reduction and binding inhibition titres for PvDBPII SalI or PvW1. (b) Correlation between binding inhibition titres for PvDBPII SalI and PvW1. Correlations shown in a and b were calculated using Spearman’s rank correlation tests. Correlation coefficients and p values for each comparison are shown. c Importance plot for the binding inhibition titres specific to PvDBPII SalI and PvW1 using Random Forest regression of the PMR reduction.

Full size image

Correlations to PMR reduction for the rest of the variables considered as important by the Boruta algorithm in the PvDBPII SalI and PvW1 datasets are shown in Supplementary Figs. 3 and 4, respectively. All these correlations to PMR reduction were statistically significant (Supplementary Figs. 3 and 4). The unimportant variables of both datasets are also shown in Supplementary Figs. 5 and 6 (PvDBPII SalI and PvW1 datasets, respectively). In case of the variables specific to PvDBPII SalI, IgA was significantly correlated to PMR reduction but it was not considered as an important predictor by the Boruta algorithm (Supplementary Fig. 5). In case the variables specific to PvDBPII PvW1, avidity, KD and IgA also correlated to the PMR reduction but they were not significant predictors as per the Boruta algorithm (Supplementary Fig. 6). Next, we compared individually the 10 variables selected as important by the Boruta algorithm between the PvDBPII SalI and PvW1 datasets, with the exception of the binding inhibition titres, already shown in Fig. 4B. We observed significant correlations for all the variables (p ≤ 0.001) (Supplementary Fig. 7), indicating strong cross-reactivity of polyfunctional PvDBPII-specific antibodies that correlate with PMR reduction.

Discussion

Two vaccine platforms, recombinant protein-in-adjuvant and viral vectors were used to deliver the leading vivax malaria vaccine candidate PvDBPII to volunteers in a challenge trial to evaluate their efficacy22. A PMR reduction of 53% was observed in the volunteers who received PvDBPII/Matrix-MTM in a delayed dosing regimen of 0, 1 and 14 months, group PvDBPII/M-M (0-1-14 months)22. No impact on the PMR was found in groups that received PvDBPII/Matrix-MTM with monthly regimen of 0, 1 and 2 months or viral-vectored PvDBPII at 0 and 2 months or 0, 17 and 19 months, groups PvDBPII/M-M (0-1-2 months), VV-PvDBPII (0–2 months) and VV-PvDBPII (0-17-19 months), respectively22. Here, we performed exploratory analysis to evaluate the antibody responses in vaccinated volunteers to identify immune correlates associated with PMR reduction.

All vaccinated volunteers seroconverted after immunizations with both vaccine platforms. Sera collected 2 days before challenge (day C-2) were analyzed for polyfunctional antibody responses. Day C-2 sera from the PvDBPII/M-M (0–1–14 months) group tended to have higher anti-PvDBPII recognition titres, binding inhibition titres and reactivity to native PvDBP in P. vivax schizonts by IFA compared to other groups. These findings are in accordance with studies showing improved antibody responses and neutralizing activity with delayed dosing regimens in vaccines targeting P. falciparum blood-stage antigen PfRH525,26 and SARS-Cov-227,28. This improved immunogenicity in delayed dosing regimens may arise from increased B cell populations. Indeed, the exploratory analyses of the PvDBPII-specific cellular responses conducted in the volunteers of this trial showed that memory B cells (CD19 + CD27+ IgG+) and plasma cells (CD19 + CD27 + CD38+) were significantly higher in delayed dosing regimens 7 days after final boost with both vaccine platforms used here29. In addition, these two B cell populations are significantly correlated to PMR reduction29. Extending the time between immunizations thus substantially improves antibody responses to PvDBPII potentially translating into greater efficacy.

The administration of the delayed dosing regimen of PvDBPII/Matrix-MTM [PvDBPII/M-M (0-1-14 months) group] resulted in a PMR reduction of 53% compared to unvaccinated controls. To our knowledge, this PMR reduction is the highest observed for any blood-stage malaria vaccine tested in CHMI25,30,31,32,33,34. The highest PMR reduction observed in CHMI following immunization with a blood-stage antigen thus far was 17% for the P. falciparum blood-stage antigen, PfRH525. Immunization with other P. falciparum blood-stage antigens such as PfMSP142 and PfAMA1 did not elicit significant parasite growth inhibition in vivo30,31,32,33,34. A malaria vaccine should protect against infection by diverse strains. Immunization with the PvDBPII SalI variant elicits cross-reactive antibodies capable of inhibiting DARC-binding by multiple PvDBPII variants in pre-clinical18,19 and clinical20,21 studies. PvDBPII-specific antibodies inhibit binding to DARC by both PvDBPII alleles, SalI and PvW1, with similar efficiency (Fig. 1B). The essential residues of PvDBPII that bind to DARC are conserved in the highly divergent PvW1 sequence13,14,15. Additional analyses are required to determine if the binding inhibitory antibodies target these conserved DARC binding residues. If that is the case, we would expect that heterologous P. vivax parasites such as PvW1 would be as susceptible for growth inhibition as the homologous strain SalI. A challenge trial with homologous parasites would probably result in a similar PMR reduction as seen in this heterologous challenge trial22. Further trials involving heterologous and homologous P. vivax challenge strains will reveal the extent of the cross-reactive protection elicited by PvDBPII immunization.

The immune correlates of protection against P. vivax are not clearly defined. Here, in addition to determining antibody recognition and binding inhibition titres, we evaluated polyfunctional antibody responses elicited by PvDBPII SalI immunization to identify immune parameters that are associated with PMR reduction. We found that in addition to antibody recognition and binding inhibition titres, PMR reduction is correlated with kdis, IgG1, IgG2, IgG4, IgA, C1q binding and FcγR binding for antibodies against both PvDBPII SalI and PvW1 variants (Fig. 3a, c). Using the Boruta classifying algorithm, we obtained the same 11 variables that significantly contribute to PMR reduction in both PvDBPII SalI and PvW1 datasets (Fig. 3b, d). Although a statistical analysis cannot definitively prove that a biomarker is causally associated with protection rather than just correlated, the variable that showed the highest importance in the PvDBPII SalI dataset was the binding inhibition titre (Fig. 3b), which also had high importance in the PvDBPII PvW1 dataset (Fig. 3d). The binding inhibition titres specific to the two PvDBPII variants were highly correlated (Fig. 4b) and modeled PMR reduction similarly when combined (Fig. 4c). This may indicate the significant value of PvDBPII-DARC binding inhibition titres as a P. vivax protection predictor. The high correlation between PvDBPII binding inhibition titres for SalI and PvW1 (Fig. 4b) reflects the cross-reactivity of these high titre binding inhibitory antibodies. Children residing in a malaria-endemic area of Papua New Guinea (PNG), who developped high-titre cross-reactive binding inhibitory antibodies to PvDBPII had a reduced risk of P. vivax infection and lower parasite densities16. However, the IC50 binding inhibition titres of these naturally acquired antibodies was ~1:2016. In comparison, day C-2 sera of the PvDBPII/M-M (0-1-14 months) group showed 50-100 times higher binding inhibition titres with IC50 of ~1:1000 to 1:2000 (Fig. 1b). Given that PvDBPII immunization elicits substantially higher binding inhibition titres compared to naturally acquired binding inhibitory antibodies that are associated with protection, one can expect immunization-induced antibodies to protect against P. vivax infection. The P. vivax blood-stage challenge model allows one to determine if a vaccine elicits antibodies that have a biological effect to reduce parasite growth in vivo. Whether the PMR reduction of 53%22 translates to protection against clinical disease remains to be determined. Protection against clinical P. vivax malaria will need to be determined in field trials in a malaria-endemic area.

In addition to binding inhibitory antibodies, ELISA recognition titres and presence of anti-PvDBPII-specific IgG1, IgG2 and IgG4 significantly contributed to PMR reduction. Consistent with our results, total IgG as well as IgG1 specific to PvDBPII SalI were associated with a reduced risk of malaria in exposed children from PNG35. In addition, the He et al. study35 reported undetectable IgG3 levels for PvDBPII SalI similar to the very low levels of IgG3 detected here (Supplementary Fig. 5). However, IgG3 against PvDBPII allele AH, the most frequent allele in this region, was detected in the PNG field study35. IgG3 against PvDBPII allele AH was more predominant in older children and correlated with protection35. Other studies in malaria-endemic areas showed that there is a higher prevalence of IgG3 specific to merozoite surface proteins in P. vivax36 and P. falciparum37,38 with age. This evidence suggests that class-switching to antigen-specific IgG3 may develop after multiple P. vivax infections but not after a limited number of immunizations with PvDBPII. The IgG2 and IgG4 subclasses, which are rarely found upon natural exposure to both P. falciparum and P. vivax infections, were significant predictors of PMR reduction according to our feature selection analysis. These IgG subclasses do not bind FcγRs as efficiently as IgG1 and IgG339. The immune effector mechanisms induced by anti-PvDBPII specific IgG2 and IgG4 antibodies remain to be identified.

The binding of anti-PvDBPII antibodies to FcγRs was also a significant predictor of PMR outcome. The cellular responses mediated by these receptors could play an important role in parasite growth reduction and protection against clinical symptoms as seen for P. falciparum. After a Phase II trial showing that vaccination with RTS,S (based on the P. falciparum circumsporozoite protein, PfCSP) protects against P. falciparum infection in malaria-naïve volunteers40, multivariate analysis revealed that FcγRIIIa and antibody-dependent cellular phagocytosis (ADCP) were the strongest correlates of protection in a sporozoite-based CHMI41. Binding of anti-PfCSP antibodies to C1q was also found to correlate with RTS,S-elicited protection41. Largely unknown in P. vivax, complement-fixing antibodies of diverse blood-stage antigens in P. falciparum have been correlated with protection upon natural exposure42. Our data suggests that complement and FcγRs may also contribute to P. vivax growth reduction.

Importantly, Suscovich and collaborators found that IgA is a highly important predictor of RTS,S-induced protection41. Similarly, multivariate analysis of immune responses in a Phase I/IIa trial with blood-stage P. falciparum vaccine PfRH5 revealed that IgA responses were the variables with the greatest importance that contribute to PMR reduction in malaria-naïve individuals25. Contrary to this, our data suggest that anti-PvDBPII IgA was not a significant predictor even though it was individually correlated to PMR reduction (Supplementary Figs. 5 and 6). In addition, avidity and isotype IgM did not significantly contribute to PMR reduction. After volunteers were immunized with RTS,S and underwent CHMI, IgM was found to be a susceptibility marker for parasite infection as it was more predominant in infected volunteers41. In case of avidity, it also increased in malaria-naïve volunteers who received the delayed third dose of PfRH5 compared to a monthly dosing regimen25. We observed a similar trend of higher avidity of anti-PvDBPII antibodies in delayed dosing regimens compared to shorter dosing schedules of both vaccine platforms, groups PvDBPII/M-M (0-1-14 months) vs PvDBPII/M-M (0-1-2 months) and groups VV-PvDBPII (0-17-19 months) vs VV-PvDBPII (0–2 months). However, our feature selection algorithm showed that avidity was not an important variable for P. vivax PMR reduction unlike the case for anti-PfRH5 avidity and P. falciparum25. In addition, two field trials analyzing the vaccine efficacy of RTS,S in children found that anti-PfCSP avidity was not statistically associated with protection43,44.

In terms of binding kinetics of PvDBPII and anti-PvDBPII antibodies, the affinity constant, KD, and association constant, ka, were not selected as important by the feature selection algorithm in both datasets. Instead, the algorithm identified that the dissociation constant, kdis, was a significant contributor to PMR reduction. The kdis for anti-PvDBPII polyclonal antibodies indicates the stability of antibodies bound to PvDBPII. The ability of anti-PvDBPII antibodies to remain bound to PvDBPII could lead to better neutralizing and effector activities resulting in significant PMR reduction.

The data presented here identify key functional properties of PvDBPII-specific antibodies elicited by immunization that predict PMR reduction following P. vivax blood-stage challenge. PvDBPII-binding inhibition titres showed high importance for the prediction of PMR reduction, though other features such as antibody titre, kdis, IgG1, IgG2, IgG4 and binding to C1q and FcγRs may also contribute to P. vivax protection.

PvDBPII/Matrix-MTM delivered in a delayed dosing regimen has shown significant reduction in PMR. Extensive analysis of immune responses reported here has identified for the first time functional antibody immune correlates. These immune correlates together with anti-PvDBPII binding inhibition titres should be evaluated in subsequent trials and could guide the clinical development of a high-efficacy P. vivax vaccine based on PvDBPII.

Methods

Study design

A full description of the Phase I/IIa clinical trial is described elsewhere22. Serum samples were obtained from a series of three Phase I/IIa clinical trial protocols (called VAC069, VAC071 and VAC079) conducted in parallel and designed to evaluate the efficacy of immunization with PvDBPII vaccine candidates including His-tag free recombinant PvDBPII SalI formulated with Matrix-MTM adjuvant and viral-vectored PvDBPII SalI delivered by ChAd63 and MVA followed by blood stage challenge with the Thai P. vivax clinical isolate PvW1 to evaluate efficacy22.

Expression of recombinant PvDBPII and DARC-Fc

Recombinant PvDBPII SalI (the vaccine candidate) and PvDBPII PvW1 (the binding domain of PvDBP from the P. vivax isolate PvW1 used for CHMI) were produced as previously described19,45,46,47. Briefly, synthetic genes encoding PvDBPII SalI and PvW1 with C-terminal 6-His tags that were codon optimized for expression in E. coli, were cloned into the pET28a (+) vector (GenScript) and the resultant plasmids were transformed into E. coli strain BL21(DE3) pLysS (C6060, Thermo Fisher). PvDBPII was expressed by fed-batch fermentation, cells were lysed, PvDBPII was solubilized from inclusion bodies under denaturing conditions and purified by nickel-charged nitrilotriacetic acid (Ni-NTA) affinity chromatography (17524802, Cytiva). Recombinant PvDBPII was refolded by rapid dilution method, dialyzed and finally purified by cation exchange (SP Sepharose column, 17115201, Cytiva) and gel filtration (Superdex 200 column) chromatography (28-9893-35, Cytiva). The final monomeric recombinant PvDBPII was stored at −80 °C.

A plasmid encoding DARC‐Fc was generated by ligating the first 60 codons of human DARC (FyB allele) to sequences encoding the Fc region of human IgG1 in the mammalian expression vector pCDM848. The plasmid containing the human tyrosylprotein sulfotransferase-2 (TPST-2)48 was co-transfected with the DARC-Fc plasmid into HEK293T cells (CRL-3216, ATCC). Recombinant DARC-Fc was purified from culture supernatants by protein G affinity chromatography. The final recombinant protein was stored at −80 °C.

ELISA

Nunc MaxiSorp ELISA plates (439454, Thermo Fisher) were coated overnight with 100μl of recombinant PvDBPII variant SalI or PvW1 proteins (1 μg/ml) in carbonate-bicarbonate buffer (C3041, Sigma) at 4 °C. Next day, plates were washed three times with 0.05% Tween (P1379, Sigma) in PBS (PBS/T) and blocked with 200 μl of 5% non-fat milk (Regilait) PBS/T for 1 h at 37 °C. Test and reference anti-PvDBPII sera (reference serum consisted of a pool of sera collected from volunteers at two weeks after the final boost in a previous PvDBPII Phase I trial20) were diluted in 2.5% milk (initial dilution 1:200) and 100 μl was added per well in duplicate and incubated for 1 h at 37 °C. After washing, bound antibodies were detected by adding 100 μl of horseradish peroxidase-conjugated anti-human IgG rabbit antibodies (A8792, Sigma) at dilution 1:4,000 and incubated for 1 h at 37 °C. The assay was developed at room temperature by using 100 μl of the two-component chromogenic substrate for peroxidase detection, TMB (3,3′,5,5′-tetramethylbenzidine, 5120-0047, Life Sciences), for 5 min and the reaction was stopped with 100 μl of 1 M phosphoric acid (H3PO4, 695017, Sigma). The optical density was measured immediately at a wavelength of 450 nm (OD450). The reference serum was assigned 200 arbitrary units, AU, and the standard curve from the reference serum was used to fit a four-parameter logistic model using ADAMSEL version 3.0 software (Ed Remarque© 2021). OD450 values were later converted to AU using the standard curve and antibody titres of test sera were reported in AU for each sample.

Avidity

PvDBPII SalI or PvW1 pre-coated ELISA plates were incubated with test samples diluted to give an OD450 of ∼2.0. After sample incubation, descending concentrations of the chaotropic agent sodium thiocyanate (NaSCN, 251410, Sigma) (7 M to 0 M in PBS) were added (100 μl) and incubated for 15 min at room temperature. Plates were washed with PBS/T and reaction was developed to detect bound IgGs as per ELISA protocol. OD450 values were plotted versus NaSCN concentration and fitted in a four-parameter logistic model. The NaSCN concentration that resulted in a 50% reduction of the OD450 was used as a measure of the avidity (IC50).

Isotyping

PvDBPII SalI or PvW1 pre-coated ELISA plates were incubated with test samples diluted 1:100 in duplicates. After washing, the following antibodies were added for detection (100 μl at dilution 1:1,000): biotin-conjugated mouse monoclonal anti-human IgG1 (B6775, Life Technologies); biotin-conjugated mouse monoclonal anti-human IgG2 (B3398, Life Technologies); biotin-conjugated mouse monoclonal anti-human IgG3 (B3523, Sigma); biotin-conjugated mouse anti-human IgG4 (B3648, Sigma); peroxidase-conjugated goat polyclonal anti-human IgA α-chain (A0295, Sigma); and peroxidase-conjugated goat polyclonal anti-human IgM μ-chain (401905, Millipore). After 1 h incubation at 37 °C, plates were washed and avidin-peroxidase (A7419, Sigma) was added (100 μl at dilution 1:5,000), except for the IgA and IgM wells, to which 2.5% milk was added. After 1 h incubation at 37 °C, the reaction was developed like in ELISA protocol. The OD450 was used to evaluate the IgG subclass, IgA or IgM of each sample.

ELISA-based PvDBPII-DARC Binding Inhibition Assay

The binding of PvDBPII to DARC was analyzed in presence or absence of anti-PvDBPII antibodies using an ELISA based format as described earlier48. Briefly, recombinant DARC-Fc (1 μg/ml) was coated on to Nunc MaxiSorp ELISA plates overnight at 4 °C in carbonate-bicarbonate buffer. Next day, the plate was blocked for 2 h at 37 °C using 2% non-fat milk PBS/T. Recombinant PvDBPII SalI or PvW1 with concentrations in the range of 0.8–25 ng/ml (with OD450 of ~1 at 25 ng/ml) were used to generate the PvDBPII standard curve using a four-parameter logistic curve. Serum samples were analyzed at dilutions 1:10 to 1:2,430. Each serum dilution was incubated with 25 ng/ml PvDBPII at 37 °C for 30 min. The reaction mixture was added to wells in duplicate and incubated at 37 °C for 1 h. PvDBPII bound to DARC was probed with anti-PvDBPII polyclonal rabbit sera (generated in house) at 37 °C for 1 h and detected with peroxidase-conjugated anti-rabbit IgG secondary antibody (A6154, Sigma) at 37 °C for 1 h. The assay was developed as described in ELISA protocol. The amount of bound PvDBPII was estimated by converting OD450 values to protein concentrations using the PvDPBII standard curve. The interpolated protein concentration values were used to calculate percent binding (%) for each serum sample dilution. Then, the % binding inhibition for each serum dilution was calculated as follows: % Binding Inhibition = 100 - % Binding. The plot of % Binding Inhibition versus serum dilution was used to find the serum dilution at which 50% binding inhibition (IC50) is achieved. Three independent replicates were averaged to determine the median IC50.

Immunofluorescence assay (IFA)

A clinical isolate collected from a P. vivax malaria patient in Cambodia during field surveys by Institut Pasteur du Cambodge was used for IFA and PvDBPII sequence was determined by Sanger sequencing as described before15. An informed written consent was obtained from patient prior to enrollment. All procedures were carried out in strict accordance with relevant guidelines and regulations. The P. vivax clinical isolate was matured to schizont stage in culture. Schizonts were purified on Percoll and used to prepare slides for use in immunofluorescence assays. Slides were fixed and frozen at −70 °C in presence of desiccant. Frozen slides of P. vivax schizonts were thawed at room temperature for 30 min. Slides were blocked with 5% bovine serum albumin (BSA, A7030, Sigma) in PBS for 30 min at 37 °C and probed with test sera diluted in 2.5% BSA at 1:200, 1:500 and 1:5,000 for 30 min at 37 °C, followed by three washes with PBS. A mixture of Alexa Flour 488-conjugated goat anti-human IgG (H + L) secondary antibodies (A11013, Themo Fisher) at 1:500 and Hoeschst 33342 solution (62249, Thermo Fisher) at 1:20,000 was added and incubated for 30 min at 37 °C. After washing, slides were treated with anti-Fade (H-1000-10, Vector Laboratories) and visualized on Leica DM 5000B Microscope (Leica Microsystems). The fluorescence intensity of each sample was scored as follows: +++, for high intensity; ++, medium intensity; +, low intensity; and -, no signal.

Antibody affinity by biolayer interferometry (BLI)

Experiments were performed on an Octet RED 384 instrument (Fortebio) at 25 °C with shaking at 1,000 round per minute (rpm). All assays were conducted in standard Greiner black 96-well microtiter plates (655209, Greiner) in a volume of 120 μl/well. Buffer consisting of PBS with 1 mg/ml BSA was used for baselines, dissociation steps and to dilute recombinant proteins. Sera were diluted in PBS to achieve final concentrations of IgGs in the range of 30 to 0 nM. This optimal range gives a sensogram with even spacing between the binding signals at different concentrations. The method was set as follows: NTA biosensors (18-5101, Sartorius) were hydrated for 10 min in PBS and regenerated (3 cycles, 30 s each) with 10 mM glycine pH 1.5 (410225, Sigma), followed by another regeneration with 50 mM ethylenediaminetetraacetic acid (EDTA) (EDS, Sigma). Biosensors were activated with 10 mM nickel sulfate (NiSO4, 656895, Sigma) for 180 s and later dipped in buffer (120 s). PvDBPII variant SalI or PvDBPII PvW1 were immobilized via His tag at 5 μg/ml for 600 s. A mammalian cytosolic protein, peroxiredoxin 6 (PRDX6)49, was loaded at 3 μg/ml to pre-charged NTA biosensors for 600 s as a reference biosensor (negative control). Loaded biosensors were tested for binding to test sera in the following steps: baseline (60 s in buffer), association step (600 s in serum dilutions) and dissociation step (600 s in buffer). Two wells containing only buffer instead of sera were assigned as reference wells. Signals from reference wells and reference biosensors were later subtracted. Affinity, association and dissociation constants (KD, ka and kdis, respectively) were calculated by Data Analysis HT software version 10 (Fortebio) using kinetic analysis and a 1:1 binding model.

Binding analysis to complement component C1q and Fc receptors by BLI

Experiments were performed on an Octet HTX instrument (Fortebio) at 25 °C with shaking at 1,000 rpm. All assays were conducted in standard black 384-well microtiter plates (781209, Greiner) and volume was 120 μl/well. The following recombinant proteins diluted in PBS containing 1 mg/ml BSA (PBS-BSA) were used to determine binding to sera: C1q (ab282858, Abcam) at 12 nM; the Fc gamma receptor (FcγR) Ia (FcγRIa), FcγRIIa, FcγRIIb, and FcγRIIIa (all from Sino Biological: 10256-H08H, 10374-H08H1, 10259-H08H and 10389-H08H1, respectively) at dilutions 19 nM, 1.7 μM, 3.7 μM and 2.3 μM, respectively. The method was set as follows, including the reagents from the AR2G Reagent kit (18-5095, Sartorius): Amine-reactive second-generation (AR2G) biosensors (18-5092, Sartorius) were activated as per manufacturer’s protocol in 20 mM 1-Ethyl-3-(3-Dimethylamino propyl) Carbodiimide, hydrochloride (EDC) and 10 mM sulfo N-hydroxysuccinimide (s-NHS) solution for 300 s. PvDBPII SalI domain, PvDBPII PvW1 recombinant protein or PRDX6 (reference biosensor) were immobilized at 20 μg/ml in 10 mM acetate buffer pH 6 for 600 s. Loaded biosensors were quenched with 1 M ethanolamine pH 8.5 for 300 s. Later, biosensors were regenerated (3 cycles, 30 s each) with 10 mM glycine pH 1.5 and dipped into wells containing serum diluted in PBS-BSA at 1:100 for 600 s. After an incubation of 900 s in PBS-BSA to stabilize the signals, biosensors were tested for binding to C1q and FcγRs in the following steps: baseline (120 s in PBS-BSA), association step (600 s in C1q and FcγR dilutions) and dissociation step (600 s in PBS-BSA). Wells containing PBS-BSA alone were assigned as reference wells. Signals from reference wells and reference biosensors were later subtracted. Binding to C1q and FcγRs was measured as the wavelength shift signal or response (nm) at the end of the association step. All the data were analyzed using Octet BLI Analysis Studio 13.0 software.

Statistics

Because of the complex structure of the data, we used frequently-used statistical tests for analysis of univariate data, and machine learning algorithms for analysis of multivariate data. Comparisons among vaccinated groups and timepoints in antibody and binding inhibition titres were performed using a Kruskal–Wallis test with Dunn’s correction for multiple comparisons. Differences between SalI and PvW1 were tested with pairwise comparisons with Bonferroni’s multiple comparison tests. Correlations between variables were determined by Spearman rank test. All statistical tests were two-sided and a p value < 0.05 was considered significant. Modeling of PMR reduction including only the binding inhibition titres specific to both PvDBPII alleles was performed by Random Forest regression. Data analysis and graphs were performed using GraphPad Prism version 9.3.1 (GraphPad Software Inc.). The correlation plot of all variables and Random Forest regression were generated in R version 4.2.2 and RStudio version 2022.12.0 + 353.

Feature selection was performed using a Boruta algorithm24,50. The Boruta algorithm is a wrapper method built around a random forest classifier that performs a top-down search for relevant features, while progressively eliminating irrelevant features, by comparing the importance of original features with the importance achievable at random (shadow features, estimated using permuted copies of the original features). To measure the importance of the immunological readouts that significantly contribute to the PMR outcome, we applied the Boruta algorithm using R version 4.2.2 and RStudio version 2022.12.0 + 353.

Ethics Statement

Sera from volunteers participating in the three clinical trials conducted at the University of Oxford, UK were sent to Institut Pasteur, Paris under a Material Transfer Agreement. The clinical trial studies had received ethical approval from UK National Health Service Research Ethics Services, (VAC069: Hampshire A Research Ethics Committee, Ref 18/SC/0577; VAC071: Oxford A Research Ethics Committee, Ref 19/SC/0193; VAC079: Oxford A Research Ethics Committee, Ref 19/SC/0330). The vaccine trials were approved by the UK Medicines and Healthcare products Regulatory Agency (VAC071: EudraCT 2019-000643-27; VAC079: EudraCT 2019-002872-14). The trials are registered under the following ClinicalTrials.gov numbers: VAC069 NCT03797989, VAC071 NCT04009096, VAC079 NCT04201431. The research conducted at the Institut Pasteur received approval from the Institutional Review Board (IRB) of the Institut Pasteur (Research Ref IRB2022-03).

Reporting summary

Further information on research design is available in the Nature Research Reporting Summary linked to this article.