Study participants

A total of 874 children aged 12 to 59 months old at first vaccination were included in the study (Fig. 1, Supplementary Fig. 1). Of these, 739 received RTS,S/AS02_A (675 from Manhiça, 64 from Ilha Josina) and 135 received a comparator vaccine (68 from Manhiça, 67 from Ilha Josina) (Table 1).

Participants received either RTS,S/AS02_A or a comparator vaccine administered at month (M) 0, M1 and M2. Blood samples were collected at baseline (M0), 2 weeks after the third dose (M3), and 6.5 months after the third dose (M8). 743 children were enrolled in cohort 1 (Manhiça, low malaria transmission intensity [MTI]) and followed up by passive surveillance to detect clinical malaria episodes as endpoint, and 131 were enrolled in cohort 2 (Ilha Josina, high MTI), where participants received a combination of amodiaquine and sulfadoxine-pyrimethamine 14 days before the third vaccine dose to clear asymptomatic parasitemia, and were followed up using active detection of infection during 4.5 months starting 2 weeks post-vaccination to measure vaccine efficacy against infection.

Immunogenicity to RTS,S vaccine target antigens

A robust PfCSP antibody response was observed among children receiving RTS,S in both trial sites, Manhiça and Ilha Josina (Fig. 2), with significant increases in IgG levels to all three constructs (full length [FL], NANP repeat, C-term). One month (M) post-third vaccine dose (M3), the fold difference in anti-PfCSP FL IgG geometric mean levels in the RTS,S relative to the comparator group was 3935.5 (95% confidence interval [CI] 2065.4–7498.9, p < 0.001) in Manhiça. For NANP repeat and C-term constructs, the fold differences were 602.6 (95% CI 342.8–1056.8, p < 0.001) and 285.8 (95% CI 166.3–490.9, p < 0.001), respectively. In Ilha Josina, the fold-changes were 1285.3 (95% CI 463.5–3564.5, p < 0.001) for PfCSP FL, 462.4 (95% CI 170.6–1253.1, p < 0.001) for NANP and 324.3 (95% CI 120.2–873.0, p < 0.001) for C-term.

a anti-PfCSP IgG levels in Manhiça, b anti-PfCSP IgG levels in Ilha Josina. Boxplots display median IgG levels against PfCSP full length (fl), PfCSP NANP and PfCSP C-terminus (C-term) and the interquartile range (IQR). p-values correspond to the FDR-adjusted from a t-test comparing the IgG levels to each PfCSP construct between treatment groups at month (M) 3. c IgG levels to Hepatitis B surface antigen (HBsAg) in Ilha Josina. Boxplots display median and IQR of IgG levels to HBsAg at each timepoint (month) for each of the age groups (from 12 to 24 [12 < 24] and from 24 to 59 [24 < 59] months). Children in the older group received the Engerix vaccine as a comparator. p-values were obtained from paired t-tests between IgG levels at M0 and M3 and from two-sample t-tests between IgG levels at M3 in each age group in the RTS,S treatment group.

Antibody levels to HBsAg in Ilha Josina had a significant increase from M0 to M3 after RTS,S vaccination in both children younger than 24 months (who had been previously vaccinated with the Engerix vaccine) and children older than 24 months (who had not been previously immunized against hepatitis B virus, because it had only been recently introduced as part of the local EPI programme) (p < 0.001), although M3 IgG levels were significantly higher in the younger children (p = 0.02)(Fig. 2). Among children aged 24 to 59 months, a greater increase (31.7 ± 95% CI 11.38–88.51, p < 0.001) in anti-HBsAg antibody levels was observed after RTS,S vaccination than after Engerix vaccination in the comparator group, consistent with previous analyses²⁰. Thus, RTS,S elicited a stronger antibody response against HBsAg compared to the Engerix-B vaccine, particularly in children with pre-existing hepatitis B immunity.

IgG levels to PfCSP and HBsAg decreased over the subsequent 6 months post-vaccination in both age groups (Fig. 2).

Immunogenicity to vaccine off-target Pf antigens

RTS,S immunization elicited antibody responses against a set of PfBSag not included in the vaccine immunogen. In Manhiça, RTS,S-vaccinated children exhibited a significant increase from M0 to M3 in levels of IgG to PfRh4.2, PfMSP5, PfMSP1 Block (Bl) 2 and PfRh5, with geometric mean fold-changes ranging from 5.06 (95% CI 4.51–5.69, p < 0.001) for PfRh4.2 to 1.22 (95% CI 1.12–1.33, p < 0.001) for PfRh5 (Fig. 3). Conversely, IgG levels to PfEXP1 decreased, with a mean fold-change of 0.79 (95% CI 0.65–0.95, p = 0.002).

Boxplots show median and interquartile range (IQR) of IgG levels against each of the Pf-blood-stage antigens at each time point (Month [M]). p-values are adjusted by false discovery rate (FDR) from t-tests comparing IgG levels at M3 between RTS,S and comparator groups, and from paired t-tests comparing IgG levels at M0 and M3 in RTS,S vaccinees.

Comparing IgG levels at M3 between vaccination groups, children in Manhiça had significantly higher antibody levels against PfRh4.2 (5.31, 95% CI 3.45–8.15, p < 0.001) and PfMSP5 (6.08, 95% CI 3.07–12.05, p < 0.001) in the RTS,S group. A trend of higher IgG levels to PfRh5 was also observed, although this difference did not reach statistical significance (1.57, 95% CI 1.1–2.25, p = 0.089) (Fig. 3). In Ilha Josina, similar observations were noted for PfRh4.2 (3.06, 95% CI 1.5–6.27, p = 0.02) (Supplementary Fig. 2).

Post-vaccination levels of IgG to off-target PfBSag had a low correlation with levels of IgG to PfCSP constructs (Supplementary Fig. 3).

Effect of age and site on RTS,S immunogenicity

At M3, RTS,S-vaccinated children from Manhiça (N = 675) compared to those from Ilha Josina (N = 64) had lower IgG levels to PfCSP constructs (Supplementary Fig. 4), though these differences were not statistically significant (0.58 95% CI [0.26–1.27], p = 0.17 for NANP; 0.63 95% CI [0.34–1.16], p = 0.32 for C-term).

In Manhiça, older children (N = 569) had lower geometric mean antibody levels compared to younger children (N = 174), with fold differences of 0.46 (95% CI 0.27–0.78, p = 0.016) for NANP repeat and 0.54 (95% CI 0.32–0.90, p = 0.038) for C-term (Supplementary Fig. 5). In Ilha Josinha older children also exhibited lower antibody responses, with fold differences of 0.29 (95% CI 0.05–1.65) for NANP and 0.53 (95% CI 0.09–3.06) for C-term, though this difference did not reach significance (p = 0.21 and p = 0.47, respectively).

Effect of baseline Pf exposure on RTS,S immunogenicity

Analyzing correlations between PfCSP or PfBSag IgG levels prior to vaccination (M0), and PfCSP IgG levels post-vaccination (M3), we observed positive moderate associations between IgG to PfEXP1, PfMSP2 or PfCSP constructs at baseline and PfCSP IgG response at M3 (Fig. 4). In contrast, baseline IgG levels to other PfBSag showed very low associations, if any, with post-vaccination PfCSP responses (Supplementary Fig. 6). For baseline PfBSag antibodies that are considered good markers of naturally acquired Pf exposure, correlation coefficients ranged from 0.22 (PfEXP1 at M0 vs. PfCSP C-term at M3, p < 0.001) to 0.31 (PfMSP2 at M0 vs. PfCSP NANP at M3, p < 0.001) (Fig. 4).

Correlations between IgG levels against PfCSP NANP repeat and PfCSP C-terminus (C-term) at month (M) 3 and IgG to PfEXP1, PfMSP2, PfCSP NANP and C-term at M0, assessed by the Pearson’s correlation coefficients. Displayed p-values are adjusted by FDR. a Outcome variable is M3 anti-PfCSP NANP IgG levels. b Outcome variable is M3 anti-PfCSP C-term IgG levels.

Pre-vaccination anti-PfCSP IgG levels correlated positively with post-vaccine antibodies to the same construct (0.44 for PfCSP NANP, p < 0.001; 0.45 for PfCSP C-term, p < 0.001) (Fig. 4). These results did not change when adjusting for age at baseline. The positive effect of previous malaria exposure on RTS,S antibody responses could potentially be mediated by PfCSP-specific acquired immunity, through boosting by vaccination. To determine it, we regressed post-vaccination PfCSP-antibody responses on baseline anti-PfCSP, adjusting for baseline exposure, measured as the sum of IgG levels to PfEXP1 and PfMSP2, and for age. The previously observed correlations between pre-vaccination anti-PfEXP1 or PfMSP2 levels and post-vaccination anti-PfCSP IgG levels disappeared. In contrast, those involving pre-vaccination anti-PfCSP levels were maintained (all adjusted p-values for PfCSP FL, NANP and C-term were <0.001 when the outcome was PfCSP NANP and when it was PfCSP C-term) (Fig. 5).

Three multiple linear regression models are presented, each using one baseline anti-PfCSP IgG measurement (full-length [FL], NANP, or C-terminus [C-term]) as the predictor, with outcomes defined as the post-vaccination IgG levels at month (M) 3 against NANP (blue triangles) and C-term (red circles). These models were adjusted for baseline malaria exposure, measured as the sum of anti-PfEXP1 and anti-PfMSP2 IgG levels at M0, and age at first vaccination. Each forest plot displays the estimated effect sizes (multiplicative effect in the dependent variable per a 10-fold increase in the predictor) and their 95% confidence intervals.

We also analyzed the relationship between pre-existing (M0) anti-HBsAg IgG levels and post-vaccination PfCSP IgG levels, and age at first vaccination, but found no significant correlation (Supplementary Fig. 7).

Association of post-vaccination IgG levels and clinical malaria protection

To evaluate the relevance of post-vaccination PfCSP antibodies on protection against clinical malaria over a 6-month follow-up period, we took into account the heterogeneity of Pf exposure at baseline, as well as age. For the primary analysis including cohort 1 children (Manhiça), higher levels of IgG to all PfCSP constructs –including the C-term fragment– induced by RTS,S immunization, were significantly associated (p < 0.05) with a lower hazard of a first or only malaria episode, adjusted by baseline PfEXP1 and PfMSP-2 IgG levels, and by age (Table 2). These results were replicated using multiple malaria episodes as the endpoint. In the multivariable models, Pf exposure at baseline independently correlated with higher malaria risk subsequently (Table 2). In secondary analysis including both cohort 1 and 2 vaccinated children, higher levels of anti-PfCSP IgG also significantly correlated with a lower risk of a first or only clinical malaria event for FL (p = 0.042), NANP repeat (p = 0.022) and C-term (p = 0.027); Table 2 shows the results adjusting by age, site and Pf exposure at baseline. The associations between M3 IgG levels to the NANP repeat or C-term regions and protection were not significant if adjusted by M3 IgG levels to the other respective PfCSP construct, which means that we could not detect an independent association (data not shown).

Finally, we assessed the effect of PfBSag antibodies at M3 on malaria protection in both vaccine groups together, adjusting by their M0 IgG levels and by age at baseline (Table 3). Including Manhiça children only, higher levels of IgG to PfBSag showing an off-target profile were associated with a reduced hazard of having a first or only clinical malaria episode, with statistical significance for PfMSP5 (HR 0.81, 95% CI 0.69–0.96, p = 0.013) and PfMSP1 Bl2 (HR 0.75, 95% CI 0.58–0.96, p = 0.024), and borderline significance for PfRh4.2 (HR 0.80, 95% CI 0.63–1.01, p = 0.064) and PfRh5 (HR 0.77, 95% CI 0.58–1.01, p = 0.060). Adjusting by M3 PfCSP IgG levels, PfMSP1 Bl2 maintained its significant association with protection, and a trend continued to be observed for PfMSP5 and PfRh5 (Table 3). As expected, M0 levels of IgG to all PfBSag, used for adjustment as markers of exposure, positively correlated with malaria risk in the models. Including all study children, significant associations with a lower risk of having a first or only malaria episode were observed for levels of IgG to PfMSP5 (HR 0.78, 95% CI 0.68–0.90, p = 0.0005), PfMSP1 Bl2 (HR 0.72, 95% CI 0.59–0.87, p = 0.0006), PfRh5 (HR 0.74, 95% CI 0.59–0.94, p = 0.011), PfRh4.2 (HR 0.71, 95% CI 0.60–0.84, p = 8.77e-5), and PfEBA140 (HR 0.78, 95% CI 0.64–0.94, p = 0.010). Adjusting by site and M3 CSP IgG levels (Table 3), significant associations were maintained for IgG levels to PfMSP5 and PfMSP1 Bl2 and borderline significant for IgG levels to PfRh5, PfRh4.2 and PfEBA140. No protective associations were found for M3 IgG levels to antigen markers of Pf exposure or PfEMP1.