Prediction of differential Gag versus Env responses to a mosaic HIV-1 vaccine regimen by HLA class I alleles

Abstract

HLA class I variation has the strongest effect genome-wide on outcome after HIV infection, and as such, an understanding of the impact of HLA polymorphism on response to HIV vaccination may inform vaccine design. We sought HLA associations with HIV-directed immunogenicity in the phase 1/2a APPROACH vaccine trial, which tested vaccine regimens containing mosaic inserts in Ad26 and MVA vectors, with or without a trimeric gp140 protein. While there were no HLA allelic associations with the overall cellular immune response to the vaccine assessed by ELISpot (Gag, Pol, and Env combined), significant associations with differential response to Gag compared to Env antigens were observed. Notably, HLA class I alleles known to associate with disease susceptibility in HIV natural history cohorts are associated with stronger Env-directed responses, whereas protective alleles are associated with stronger Gag-directed responses. Mean viral loads determined for each HLA allele in untreated individuals correlated negatively with the strength of the Gag response minus the Env response in Black vaccinees based on both ELISpot and CD8⁺ T cell ICS responses. As the association of T cell responses to conserved Gag epitopes with lower viral load in untreated individuals is well established, our data raise the possibility that the Ad26.Mos.HIV vaccine may induce more effective cellular responses in those with HLA alleles that confer improved virologic control in untreated HIV infection.IMPORTANCENo vaccine tested to date has shown sufficient efficacy against HIV infection. A vaccine that induces robust responses in one individual may fail to do so in another individual due to variation in HLA class I genes, loci central to the immune response. Extensive data have shown the strong effect of HLA variation on outcome after HIV infection, but very little is known about the effect of such variation on HIV vaccine success. Here, we identify a link between the effect of HLA variation on HIV disease outcome and immune responses to an HIV vaccine. HLA variants associated with better HIV control after infection also induce stronger responses against the HIV Gag protein relative to the Env protein after vaccination. Given the virologic control conferred by responses to Gag in natural history of HIV infection, these data suggest that HLA alleles conferring protection after HIV infection may also support a more effective cellular response to HIV vaccination.

Keywords: HIV proteins; HIV vaccine; HLA class I.

Fig 1

HLA class I alleles marginally associated with ELISpot response to HIV antigens after vaccination of combined Black APPROACH participants (N = 155). (A) Percent variance of ELISpot response to tested antigens captured by principal components. The first component, capturing 56% of the total variance, provides a good measurement of the magnitude of overall response. (B) Vector plot showing contribution of antigen responses to PC1 and PC2. The X and Y coordinates of the vectors represent the PC1 and PC2 contributions, respectively. All antigens contribute positively to PC1 (all vectors point to the right, indicating a positive contribution), whereas Gag and Pol antigens (green and blue, respectively) contribute positively to PC2 (vectors point up), while Env antigens (red) contribute negatively to PC2 (vectors point down). Vectors in the plot are shorter than 1 due to contributions to PCs beyond PC1 and PC2. (C–F) QQ plots of associations of HLA class I and II alleles with PCs of vaccine response, comparing the distribution of P values expected by random chance (X-axis) and the observed distribution of P values (Y-axis). Associations significant after correction will appear well above the diagonal line. (C and D) QQ plots showing lack of an association of HLA class I (C) and class II (D) alleles with PC1. (E) QQ plot showing association of HLA class I alleles with PC2 of the antigen responses. (F) QQ plot showing no association of HLA class II alleles with PC2 of the antigen responses. (C–F) Alleles shown in red have a negative correlation with antigen response, indicating that they are associated with a weaker response, while alleles shown in blue have a positive correlation with antigen response, indicating that they are associated with a stronger response.

Fig 2

HLA class I alleles associated with differential response to Gag versus Env antigens (“Gag – Env” variable), and Gag antigens alone. (A–D) QQ plots of HLA allelic associations with the “Gag – Env” variable for PC2 based on (A) ELISpot responses in combined Black participants (N = 155), (B) ELISpot responses in White participants (N = 81), (C) ICS CD8⁺ T cell responses in combined Black participants (N = 172), and (D) ICS CD8⁺ T cell responses in White participants (N = 82) from the APPROACH trial. (E and F) QQ plots of HLA allelic associations with Gag responses in PC2 based on (E) ICS CD8⁺ T cell responses in combined Black participants (N = 172) and (F) ICS CD8⁺ T cell in White participants (N = 82). Alleles shown in red associate with a weaker response, whereas those shown in blue associate with a stronger response.

Fig 3

Schematic overview of the approach taken to determine the relationship between HLA allelic contributions to “Gag-Env” score and virologic control.(A) The allele “Gag – Env” score is the average of the “Gag – Env” scores of all APPROACH participants carrying a given HLA-B allele (dominant model). Alleles are depicted as red, yellow, or blue balls. (B) An imputed “Gag – Env” score for each subject in the PLWH cohort is calculated as the average of the “Gag – Env” scores of the two HLA-B alleles (red, blue, or yellow) carried by that subject. The mVL for each subject is also determined. A correlation between the mVL (Y-axis) and imputed “Gag-Env” score (X-axis) for each subject is tested (each dot represents one person; left plot). (C) The mVL for each allele is calculated as the average of the mVL of all subjects in the PLWH cohort carrying that allele (red, blue, or yellow; dominant model). The allele mVL is plotted on the Y-axis and the allele “Gag-Env” score (as determined from the APPROACH data) is plotted on the X-axis (each dot represents means of data generated from subjects carrying a given HLA allele; right plot).

Fig 4

Correlations between mVL and imputed means of the “Gag – Env” variables in natural history cohorts of Black PLWH as a function of specific HLA class I alleles. (A and B) Correlations between HIV mVL and the imputed mean of the “Gag – Env” scores for the two HA-B alleles carried by each individual in the natural history cohort (N = 1411), where the imputed scores were derived from (A) ELISpot and (B) ICS CD8⁺ T cell response data in the APPROACH trial. Each circle represents a single individual. (C and D) Correlations between mVL for individuals grouped based on the presence of distinct HLA-B alleles (dominant model) and the mean of the imputed “Gag – Env” scores for those HLA-B alleles as derived from (C) ELISpot and (D) ICS CD8⁺ T cell response data in the APPROACH trial. P values shown are empirical P values from a permutation test. The size of the circles represents the number of individuals in the analysis and there are two circles for each allele: the pink circles represent the number of APPROACH participants in the analysis from whom “Gag-Env” score was generated and the blue circles represent the number of PLWH cohort participants from whom mVL was generated.