Study population
At M21, a total of 709 infants and children from Bagamoyo, Tanzania, and Manhiça, Mozambique, were analyzed between the three vaccine groups that either received the RTS,S booster (R3R) or comparator vaccine (R3C and C3C) at M20 (Table 1; Supplementary Fig. 1). The median age was 26.7 months, and approximately 52% of participants were female. Previous malaria cases within 1-year before M20 occurred in 6.6% of participants; hemoglobin concentrations measured at M20 showed that 29.5% of the participants had anemia ( < 10 g/dL), and 6.2% were malnourished according to weigh-for-age Z-scores (WAZ score <−2).
Table 1 Characteristics of study participants
Immunogenicity of the RTS,S booster dose at M21 and M32
To identify markers of immunogenicity elicited by the RTS,S booster dose, we compared vaccine antigen cytokine concentration ratios from R3R (RTS,S-boosted) versus R3C (RTS,S-unboosted) and R3R versus C3C (Comparator) at M21 and M32. At M21, CSP and HBsAg ratios were significantly higher for IL-2 in the R3R group compared to the C3C group (Fig. 1a) whereas the increase relative to the R3C group was not statistically significant. CSP-stimulated IL-13 ratios were also higher in the R3R compared to the C3C group but not statistically significant after adjustment for multiple testing (P = 0.007, Padj = 0.07). In addition, HBsAg-stimulated cytokine ratios were also significantly higher for IFN-γ, IL-5, and IL-13 in R3R than in C3C, and the booster dose was found to significantly increase IL-17 in the R3R group compared to both R3C and C3C (Fig. 1a). At M32, CSP responses were maintained for IL-2, which were significantly higher in R3R compared to both R3C and C3C; and HBsAg ratios were higher for IL-2, IFN-γ, and IL-5 in the R3R compared to the C3C group (Fig. 1a).
Fig. 1: Comparison of CSP and HBsAg-specific cytokine responses (relative concentrations expressed as ratios of vaccine antigen over background) in antigen stimulated PBMC between R3R (RTS,S-boosted), R3C (RTS,S-unboosted), and C3C (Comparator) vaccine groups.
a Forest plots illustrating fold differences in geometric mean relative concentrations between vaccine groups for all cytokines, one month following the booster at M21 (left) and one year later at M32 (right). Point estimates and whiskers indicate 95% confidence intervals (CI); asterisks (*) indicate significant differences in distributions according to Wilcoxon tests after Benjamini-Hochberg adjustment. b Boxplots showing individual levels and comparing vaccination groups for the five markers that were significantly increased upon vaccination for either CSP- and/or HBsAg-stimulated PBMC samples. The central line in the box represents the median of the relative concentrations in each vaccination group, while the box limits indicate the quartiles.
Box plots displaying individual data points for the primary immunogenicity markers IL-2, IL-13, IFN-γ, IL-17, and IL-5 show an increasing trend for R3C (RTS,S-unboosted) and R3R (RTS,S-boosted) over C3C (Comparator) for both CSP and HBsAg ratios at M21 and M32 (Fig. 1b). Analysis of crude concentrations showed a similar pattern of significantly elevated cytokines following CSP and HBsAg stimulation as was found for the antigen ratios (Supplementary Fig. 2). Correlations between levels of CSP-stimulated versus HBsAg-stimulated primary immunogenicity markers assessed at M21 showed low to moderate positive correlations between the two vaccine antigens in all three vaccine groups (Supplementary Fig. 3).
Factors affecting RTS,S immunogenicity at M21 and M32
To evaluate if there were clinical or demographic factors affecting the above RTS,S immunogenicity markers (IL-2, IL-13, IFN-γ, IL-17 or IL-5), we analyzed the effect modification of age, sex, site, previous malaria exposure, anemia, and WAZ scores on the cytokine increases in R3R (RTS,S-boosted) versus C3C (Comparator) group (Fig. 2). At M21, no statistically significant effect modification by any condition on immunogenicity markers was observed (interaction Padj > 0.05). However, unadjusted analysis showed that age and site were effect modifiers of vaccination immunogenicity, specifically for CSP-specific IL-5 for age (interaction P = 0.02; Padj = 0.32) and IFN-γ (interaction P = 0.005; Padj = 0.14), and IL-17 (interaction P = 0.02; Padj = 0.32) for site (Fig. 2). In stratified analyses, CSP-specific IL-13, IFN-γ, and IL-17 were associated with RTS,S booster for children (P = 0.03, P = 0.006, P = 0.02, respectively) but not for infants (P = 0.78, P = 0.9, P = 0.33, respectively), who had increased CSP-specific IL-5 by vaccination (P = 0.02) (Fig. 2). For HBsAg stimulation, age was an effect modifier of vaccination on IL-17 and IL-13 responses (RTS,S booster induced IL-17 and IL-13 in children but not in infants), although the interaction was not statistically significant after adjustment for multiple testing (interaction P = 0.003; Padj = 0.14 and P = 0.006; Padj = 0.14, respectively) (Fig. 2).
Fig. 2: Effect modification of clinical and demographic variables on RTS,S-induced immunogenicity markers following the booster dose at M21.
Forest plots illustrating fold differences in geometric mean relative concentrations (ratios) of CSP- and HBsAg-stimulated cytokine responses between R3R (RTS,S-boosted) and C3C (Comparator) groups stratified by age, sex, site, prior malaria exposure (1-year before M20), anemia, and WAZ scores, for the primary RTS,S immunogenicity markers (IL-2, IL-13, IFN-γ, IL-17 or IL-5). Open and closed diamonds representing conditions within each factor are defined in the bottom panel showing sample size. Point estimates and whiskers indicate 95% CI. No significant effect modification (interaction) in a regression model was found following Benjamini-Hochberg adjustment. Bar plots representing the number of study participants within each group are shown at the bottom.
At M32, no immunogenicity effect modifiers were identified for CSP stimulations (Supplementary Fig. 4). However, age and site were associated with significantly different RTS,S-induced HBsAg IL-2 responses (Supplementary Fig. 4), though significance was lost after adjusting for multiple comparisons. RTS,S booster dose induced higher IL-2 concentrations in infants than children at M32 (interaction P = 0.006, Padj = 0.16) and higher in Bagamoyo than Manhiça (interaction P = 0.007; Padj = 0.17).
Malaria exposure before M21 was not associated with CSP-specific responses in non-vaccinated participants (C3C), suggesting absence of naturally acquired immunity for CSP cellular responses (Supplementary Fig. 5).
Correlates of malaria risk
The association of the main immunogenicity markers at M21 with the risk of malaria during the following 11 months was assessed through Cox models including the interaction with vaccination group (Fig. 3a). None of the CSP and HBsAg primary immunogenicity markers were associated with risk of malaria between the different vaccine groups (interaction), nor was there a direct association within any of these groups (stratified association) (Fig. 3a; Supplementary Fig. 6). For the other markers, CSP-stimulated RANTES showed a tendency towards increasing malaria risk with more RTS,S vaccination (interaction P = 0.02, Padj = 0.5) (Supplementary Fig. 6). For CSP, no association was found for any marker with malaria regardless of the vaccine group (main effect).
Fig. 3: Association of the primary immunogenicity marker levels at M21 and hazard of malaria during the following 11 months in the R3R (RTS,S-boosted) and R3C (RTS,S-unboosted) vaccine groups.
Hazard ratios indicate the factor by which the hazard is estimated to increase or decrease for one standard deviation change in the cytokine relative concentrations (ratio). Association of the primary immunogenicity markers with hazard of malaria in (a) children and infants together; (b) children only; and (c) infants only. Significance was tested with a Wald test on the corresponding Cox regression coefficient used to obtain hazard ratios. No significant differences were found following Benjamini-Hochberg adjustment.
For HBsAg responses, increased IL-10, macrophage inflammatory protein (MIP-1α) and TNF levels in R3R (RTS,S-boosted) showed tendencies for lower risk of malaria (P = 0.045 and P = 0.040 and P = 0.072, respectively) but were not significant after adjustment. HBsAg-induced monocyte chemoattractant protein (MCP-1) was significantly associated with lower risk of malaria for main effect (regardless of vaccine group), though significance was lost after adjustment (P = 0.01; Padj = 0.160) (Supplementary Fig. 6).
Factors affecting correlates of malaria risk
In stratified analyses by age cohort (Fig. 3b, c; Supplementary Fig. 7a, b), in infants and the R3C (RTS,S-unboosted) group, CSP IL-2 was associated with risk of malaria (Hazard Ratio (HR) = 3.3, confidence intervals (CI) (1.3, 8.2)), but this was not significant when adjusting for multiple comparisons (P = 0.009, Padj = 0.5) and CI were very wide (Fig. 3c). Likewise, in R3C for HBsAg, IL-2 and IL-13 were associated with increased risk of malaria, but significance was lost after adjustment (P = 0.01, Padj = 0.5; P = 0.02, Padj = 0.6, respectively) (Fig. 3c). For infants, IL-10 was significantly associated with a higher risk of malaria regardless of vaccination (main effect P = 0.0005, Padj = 0.02), an effect that also reached unadjusted significance in the R3R (RTS,S-boosted) group (P = 0.02, Padj = 0.5) (Supplementary Fig. 7b).
Stratified correlates analysis by site showed a significant effect modification of CSP-stimulated IL-2 on the risk of malaria by the different vaccination groups in Manhiça before adjusting for multiple comparisons (interaction P = 0.02, Padj = 0.8), with a stronger association with lower risk in the R3R (RTS,S-boosted) group, but not in Bagamoyo (Supplementary Fig. 8a, b). In line with these observations, CSP-specific IL-2 levels were higher in controls than malaria cases in the R3R group in Manhiça (unadjusted P = 0.0356) (Supplementary Fig. 8c), whereas there were no differences in Bagamoyo (P = 0.597). No other primary immunogenicity markers were found to have significant associations with malaria risk for either vaccine antigen in either of the two sites (Supplementary Fig. 8a, b).
Correlation between RTS,S-induced cytokines and antibody responses at M21 and M32
For total anti-CSP NANP IgG antibody levels measured by ELISA, no statistically significant correlations were found with any of the primary immunogenicity markers at M21 (Fig. 4a); however, a significant inverse correlation of RANTES with anti-CSP IgG was observed (Fig. 4b). Total anti-CSP NANP IgG had a significant moderate positive correlation with IL-2 at M32 (Fig. 4c).
Fig. 4: Correlations of anti-NANP IgG levels with CSP-stimulated cytokines in RTS,S vaccinees.
a Primary immunogenicity markers measured at M21, (b) RANTES measured at M21, and (c) primary markers measured at M32. Antibody levels shown as EU/mL and cytokines as log10 ratios. Spearman rho and P-values are shown in the plots with a linear regression trend line. R3R (RTS,S-boosted) and R3C (RTS,S-unboosted) have been pooled, n = 30 at M21 and n = 28 at M32.
Further correlations were performed using anti-CSP full length (CSP FL), anti-NANP, and anti-C-terminus (Cterm) IgG subclass levels measured by Luminex in a previous study showing that the booster dose induced IgG, IgG1, IgG3, and IgG4, but not IgG2, which had relatively unchanged levels following the primary vaccination until M32. At M21, for the primary immunogenicity markers, we only observed a significant moderate positive correlation for IL-17 with anti-NANP IgG2 levels (Supplementary Fig. 9a). IL-15 had a moderate significant inverse correlation with anti-CSP FL IgG4 at M21 (Supplementary Fig. 9b). At M32, a significant moderate inverse correlation was found for IL-2 and IL-13 with anti-NANP IgG2, and IL-2 and IL-5 with IgG4 (Supplementary Fig. 9c, d). For the remaining primary markers and IgG subclasses against CSP antigens, there were no significant correlations (Supplementary Fig. 10). RANTES had moderate significant positive correlations with anti-CSP FL and Cterm IgG and IgG1, and anti-NANP IgG4 at M32 (Supplementary Fig. 11).
Regarding HBsAg cytokine responses at M21, IL-17 had significant moderate positive correlations with HBsAg IgG1, and IL-2, IL-5 and IL-13 with IgG2 (Supplementary Fig. 12). At M32, positive correlations were found between IL-17, IL-5, and IL-13 with anti-HBsAg IgG and IgG1, also IL-17 and IL-5 with IgG2, and IL-5 with IgG4 (Supplementary Fig. 13). Additionally, some HBsAg-specific cytokine concentrations correlated positively with CSP antibodies at M21, mainly involving IL-17 and IL-2 responses correlating with protective IgG1 and IgG3 against the CSP FL and NANP (Supplementary Figs. 14, 15) and IL-17, IL-5 and IL-13 correlating with Cterm IgG3, and IL-17 with Cterm IgG1 (Supplementary Figs. 16). At M32, fewer significant correlations between HBsAg-specific cytokine responses and CSP antibodies were found (Supplementary Fig. 17).
