Distinct genetic architectures and environmental factors associate with host response to the γ2-herpesvirus infections

Abstract

Kaposi's sarcoma-associated herpesvirus (KSHV) and Epstein-Barr Virus (EBV) establish life-long infections and are associated with malignancies. Striking geographic variation in incidence and the fact that virus alone is insufficient to cause disease, suggests other co-factors are involved. Here we present epidemiological analysis and genome-wide association study (GWAS) in 4365 individuals from an African population cohort, to assess the influence of host genetic and non-genetic factors on virus antibody responses. EBV/KSHV co-infection (OR = 5.71(1.58-7.12)), HIV positivity (OR = 2.22(1.32-3.73)) and living in a more rural area (OR = 1.38(1.01-1.89)) are strongly associated with immunogenicity. GWAS reveals associations with KSHV antibody response in the HLA-B/C region (p = 6.64 × 10^-09). For EBV, associations are identified for VCA (rs71542439, p = 1.15 × 10^-12). Human leucocyte antigen (HLA) and trans-ancestry fine-mapping substantiate that distinct variants in HLA-DQA1 (p = 5.24 × 10^-44) are driving associations for EBNA-1 in Africa. This study highlights complex interactions between KSHV and EBV, in addition to distinct genetic architectures resulting in important differences in pathogenesis and transmission.

Fig. 1. Inter-individual variability in IgG antibody responses to KSHV and EBV (n = 4365 individuals).

a Distribution of anti-KSHV antibody levels (Log₁₀ MFI) for K10.5 (min = 0, max = 28,514), K8.1 (min = 0, max = 26,468) and ORF73 (min = 0, max = 29,308). b Distribution of anti-EBV antibody levels (Log₁₀ MFI) for EAD (min = 0, max = 22,747), EBNA-1 (min = 0, max = 24,696), VCA (min = 11, max = 24,754). Seropositivity to all antigens and median (centre) plus IQR displayed in the text box.

Fig. 2. Predictors of KSHV IgG antibody response.

Odds ratios (ORs) and 95% confidence intervals (CI) of significant associations (p < 0.006) between factors and IgG variable as determined based on serologies in the 4365 individuals from the GPC. a Predictors of IgG serostatus. ORs and 95% CI were estimated following a multivariate logistic regression analysis with serostatus (seropositive vs seronegative) as response variable. b Predictors of IgG response level. ORs and 95% CI were estimated following a multivariate linear regression analysis with log₁₀-transformed MFI as response variable. UQ urbanicity quartile. Dots represent the ORs and are coloured by IgG variable: KSHV (seropositivity to any antigen)—purple, ORF73—red, K8.1—yellow, K10.5—light blue; lines represent the 95% confidence intervals. Full results from this analysis are provided in Supplementary Data 1.

Fig. 3. Predictors of EBV IgG antibody response.

Odds ratios (ORs) and 95% confidence intervals (CI) of significant associations (p < 0.006) between factors and IgG variable as determined based on serologies in the 4365 individuals from the GPC. a Predictors of IgG serostatus. ORs and 95% CI were estimated following a multivariable logistic regression analysis with serostatus (seropositive vs seronegative) as response variable. b Predictors of IgG response level. ORs and 95% CI were estimated following a multivariable linear regression analysis with log₁₀-transformed MFI levels as response variable. UQ urbanicity quartile. Dots represent the ORs and are coloured by IgG variable (EBV (seropositivity to EBNA-1 or VCA)—blue, EBNA-1—magenta, VCA—light green, EAD—orange), lines represent the 95% confidence intervals. Full results from this analysis are provided in Supplementary Data 2.

Fig. 4. Regional association plots for lead SNPs associated with anti-KSHV IgG levels p < 5 × 10⁻⁸), n = 4365 individuals.

a Association on chromosome 6 in HLA class I region with anti-K10.5 IgG (rs111664408, p = 6.64 × 10⁻⁰⁹). b Association on chromosome 10 in the ACADSB region with anti-K10.5 IgG (rs10794590, p = 4.59 × 10⁻⁰⁸). c Association on chromosome 5 in GDNF region with anti-ORF73 IgG (rs114429578, p = 4.60 × 10⁻⁰⁸). GWAS performed using linear mixed model accounting for kinship in GEMMA. The lead SNPs are labelled and coloured in purple. LD (r²) was calculated based on Ugandan SNP genotypes.

Fig. 5. Regional association plots for lead SNPs associated with anti-EBV IgG levels (p < 1 × 10⁻⁸), n = 4365 individuals.

a Lead association on chromosome 6 in HLA-DQA1 with anti-EBNA-1 IgG (rs9272371, p = 3.60 × 10⁻⁴⁴). b Secondary association on chromosome 6 in HLA-DRA with anti-EBNA-1 IgG (rs3129867, p_cond = 2.78 × 10⁻¹¹). c Lead association on chromosome 6 in the HLA-DQB1 with anti-VCA IgG (rs71542439, p = 1.15 × 10⁻¹²). GWAS performed using linear mixed model accounting for kinship in GEMMA. The lead SNPs are labelled and coloured in purple. LD (r²) was calculated based on SNP genotypes in PLINK.