Subclass and avidity of circumsporozoite protein specific antibodies associate with protection status against malaria infection

Abstract

RTS,S/AS01 is an advanced pre-erythrocytic malaria vaccine candidate with demonstrated vaccine efficacy up to 86.7% in controlled human malaria infection (CHMI) studies; however, reproducible immune correlates of protection (CoP) are elusive. To identify candidates of humoral correlates of vaccine mediated protection, we measured antibody magnitude, subclass, and avidity for Plasmodium falciparum (Pf) circumsporozoite protein (CSP) by multiplex assays in two CHMI studies with varying RTS,S/AS01B vaccine dose and timing regimens. Central repeat (NANP6) IgG1 magnitude correlated best with protection status in univariate analyses and was the most predictive for protection in a multivariate model. NANP6 IgG3 magnitude, CSP IgG1 magnitude, and total serum antibody dissociation phase area-under-the-curve for NANP6, CSP, NPNA3, and N-interface binding were also associated with protection status in the regimen adjusted univariate analysis. Identification of multiple immune response features that associate with protection status, such as antibody subclasses, fine specificity and avidity reported here may accelerate development of highly efficacious vaccines against P. falciparum.

Fig. 1. Distinct vaccine elicited antibody responses.

The distribution of Spearman’s rank correlation coefficients is shown between measurements within and between assay types, with the percent of correlation coefficients above 0.75 indicated (a). A heatmap of all correlation values, is also shown. This heatmap is symmetric about the diagonal, with inter-assay subclass-serum correlations shown in the upper right and lower left quadrants, intra-assay subclass correlations shown in the upper left quadrant, and intra-assay serum correlations shown in the lower right quadrant (b).

Fig. 2. NANP repeat specificities, antibody subclass and avidity measurements associate with protection status.

Odds ratios (circles) and 95% confidence intervals (error bars) obtained from logistic regression models fit individually to each immune measurement. Regression models were fit using both studies combined and were adjusted for regimen arm (Protection ~ Regimen + Measurement). Effects were considered significant and measurements are shown here if raw p-value < 0.05 and FDR-adjusted p-value < 0.2 (**p < 0.005 and FDR-adjusted p < 0.05, * p < 0.05 and FDR-adjusted p < 0.2).

Fig. 3. Three IgG antibody subclass measurements individually associate with protection from sporozoite infection in RTS,S/AS01 studies.

Antibody responses on the day of challenge (DOC) were compared for the protected (P) and infected (I) groups of RTS,S/AS01 vaccinees from the Ad35.CS.01 prime (ARR, green circles), standard dose (RRR, blue triangles), and delayed fractional dose (RR_r, orange squares) regimens from two different Phase 2a studies. The top antibody subclass measurements identified were CSP specific IgG1 binding responses (a, p = 0.017 and FDR-adjusted p = 0.091), NANP6 specific IgG1 binding responses (b, p = 0.001 and FDR-adjusted p = 0.021), and NANP6 specific IgG3 binding responses (c, p = 0.037 and FDR-adjusted p = 0.117). N = 25 for MAL-068 ARR, N = 21 for MAL-068 RRR, N = 16 for MAL-071 RRR, and N = 30 for MAL-071 RR_r. The lower and upper hinges of the boxplots correspond to the 25th and 75th percentiles, with a line at the median. The lower and upper whisker extends from the box hinges to the smallest and largest values, respectively, which are within 1.5 * IQR of the hinge (where IQR, the inter-quartile range, is equal to the distance between the 25th and 75th percentiles).

Fig. 4. Four total serum antibody measurements individually associate with protection from sporozoite infection across RTS,S/AS01 studies.

Antibody responses were compared on the day of challenge (DOC) for the protected (P) and infected (I) groups of RTS,S/AS01 vaccinees from the Ad35.CS.01 prime (ARR, green circles), standard dose (RRR, blue triangles), and delayed fractional dose (RR_r, orange squares) regimens from two different Phase 2a studies. The top total serum antibody measurements identified were CSP (a, p = 0.003 and FDR-adjusted p = 0.034), NANP6 (b p = 0.021 and FDR-adjusted p = 0.091), NPNA3 (c, p = 0.025 and p-adjusted = 0.091), and N-interface (d, p = 0.016 and FDR-adjusted p = 0.091) specific serum antibody AUC_diss. N = 25 for MAL-068 ARR, N = 21 for MAL-068 RRR, N = 16 for MAL-071 RRR, and N = 30 for MAL-071 RR_r. Data shown for MAL-071 were previously reported and are displayed here in combination with MAL-068. The lower and upper hinges of the boxplots correspond to the 25th and 75th percentiles, with a line at the median. The lower and upper whisker extends from the box hinges to the smallest and largest values, respectively, which are within 1.5 * IQR of the hinge (where IQR, the inter-quartile range, is equal to the distance between the 25th and 75th percentiles).

Fig. 5. Frequency of putative protective immune measurements in computational models.

Immune measurements are ranked by the proportion of cross-validation LASSO penalized models in which they appear (a). Blue bars represent measurements which were found to be associated with protection by univariate analyses. The most frequent measurements, IgG1 NANP6 magnitude and IgG1 CSP AI, also appear in the LASSO penalized model with λ equal to lambda.min, the value of the tuning parameter corresponding to the minimum cross-validation misclassification error. IgG1 NANP6 magnitude appeared in the LASSO penalized model with λ equal to lambda.1se, the largest value of λ such that misclassification error was within 1 standard error of the minimum. The ROC curves for these models are shown (b) and had AUC values of 0.788 (λ = lambda.min, red curve) and 0.779 (λ = lambda.1se, blue curve). The coefficients for the included measurements are shown along the entire path of λ values (c), with lambda.min and lambda.1se indicated by the vertical red and blue lines, respectively. Each colored line corresponds to either a regimen indicator, RR_r or ARR (0 or 1 value, where 1 = yes), or an immune measurement. Variables included in the model with λ = lambda.min are represented by thicker lines. Smaller values of λ correspond to less penalization and will result in more measurements included in the regression model.