Novel HLA class I associations with HIV-1 control in a unique genetically admixed population

Abstract

Associations between HLA class I alleles and HIV progression in populations exhibiting Amerindian and Caucasian genetic admixture remain understudied. Using univariable and multivariable analyses we evaluated HLA associations with five HIV clinical parameters in 3,213 HIV clade B-infected, ART-naïve individuals from Mexico and Central America (MEX/CAM cohort). A Canadian cohort (HOMER, n = 1622) was used for comparison. As expected, HLA allele frequencies in MEX/CAM and HOMER differed markedly. In MEX/CAM, 13 HLA-A, 24 HLA-B, and 14 HLA-C alleles were significantly associated with at least one clinical parameter. These included previously described protective (e.g. B*27:05, B*57:01/02/03 and B*58:01) and risk (e.g. B*35:02) alleles, as well as novel ones (e.g. A*03:01, B*15:39 and B*39:02 identified as protective, and A*68:03/05, B*15:30, B*35:12/14, B*39:01/06, B*39:05~C*07:02, and B*40:01~C*03:04 identified as risk). Interestingly, both protective (e.g. B*39:02) and risk (e.g. B*39:01/05/06) subtypes were identified within the common and genetically diverse HLA-B*39 allele group, characteristic to Amerindian populations. While HLA-HIV associations identified in MEX and CAM separately were similar overall (Spearman's rho = 0.33, p = 0.03), region-specific associations were also noted. The identification of both canonical and novel HLA/HIV associations provides a first step towards improved understanding of HIV immune control among unique and understudied Mestizo populations.

Figure 1

Comparison of HLA class I allele frequencies between the Mestizo MEX/CAM cohort (n = 3213) and the mainly Caucasian HOMER cohort (n = 1622). Allele frequencies (2n) were calculated using the HLA Analysis tool from Los Alamos HIV Database (https://www.hiv.lanl.gov); all HLA AF > 0.001 in at least one cohort are shown here. AF were compared using Fisher’s exact test, with multiple tests addressed using q-values. Significant differences (p < 0.05, q < 0.2) are denoted by a star.

Figure 2

HLA class I haplotype structures and linkage disequilibrium in the MEX (panel A) and CAM (panel B) cohorts. HLA loci are stacked vertically, with each orange tile representing a specific HLA subtype, and with segments connecting linked alleles on adjacent loci. The height of each tile and the thickness of each segment correspond to HLA allele and haplotype frequencies, respectively. The most frequent HLA allele pairs (two-loci) found to be in linkage disequilibrium are highlighted in green (PF > 0.10) and blue (PF < 0.10 and > 0.02); less frequent pairs (PF < 0.02) are shown in grey. Frequently linked (PF < 0.10 and > 0.02) HLA-A and HLA-C allele pairs were also found in our cohorts including A*33:01/C*08:02, A*29:02/C*16:01, A*68:01/C*03:04, and A*68:03/C*07:02 in the MEX cohort, and A*02:06/C*07:02, A*24:02/C*01:02, A*24:02/C*03:05, and A*68:03/C*07:02 in the CAM cohort (not shown in the figure; see Supplementary Tables S4 and S6).

Figure 3

HLA associations with HIV pVL in the MEX/CAM and HOMER cohorts. Associations between HLA subtypes and pVL were investigated for HLA alleles with frequency equal or greater than 5 in HIV-1 clade B-infected ART-naïve individuals from the MEX/CAM (n = 3213) and predominantly Caucasian HOMER (n = 1622) cohorts. Associations between HLA alleles and pVL were evaluated using the Mann-Whitney U test, with multiple tests addressed using q-values. Significant (p < 0.05, q < 0.2) associations are highlighted in blue. Boxes denote median, 25^th and 75^th percentile, and whiskers represent the 10–90^th percentile of pVL distributions of individuals expressing each HLA-B allele (left panels) and HLA-A/C allele (right panels). HLA alleles are ordered by their pVL median and the number (n) of individuals expressing each HLA allele is shown. Red vertical lines denote plasma viral load medians for each cohort.

Figure 4

HLA-HIV associations in Mesoamerican cohorts using 5 HIV clinical parameters (univariable analysis). Associations between the expression of HLA class I alleles and 5 HIV clinical parameters (pVL, CD4 count, Z-score, CD4% and CD4/CD8 ratio) were investigated for alleles with frequency equal or greater than 5 in HIV-1 clade B-infected ART-naïve individuals from the pooled MEX/CAM cohort (A), only in MEX cohort (B) or only in CAM cohort (C). Associations were evaluated using the Mann-Whitney U test and multiple tests were addressed using q-values. Boxplots of only significant (p < 0.05, q < 0.2) HLA-HIV associations are shown. Alleles are grouped by HLA-HIV score ( + 5 to −5), then ordered by the median of Z-score, pVL, CD4 count, %CD4 and CD4/CD8 ratio. Protective and risk alleles are shaded with progressively deeper green and orange colors, respectively. Boxes denote the median, 25^th and 75^th percentile of the HIV clinical parameter of interest; whiskers represent the 10–90^th percentile. The number (n = ) of individuals expressing each HLA allele is shown. Blue vertical lines denote cohort median values for each parameter.

Figure 5

Multivariable analysis of HLA-HIV associations in Mesoamerican cohorts using 5 HIV clinical parameters. An independent linear regression model (GLM) was constructed for each HLA allele and clinical parameter, while adjusting for gender, age, geographical origin (country or region) and the presence of HLA alleles with p < 0.001 in the Mann-Whitney univariable analyses for each clinical parameter; see Supplementary Table S9 for specific HLA alleles adjusted for in each model). Coefficients and 95% confidence intervals (CI) of significant (p < 0.05, q < 0.2) associations are shown. Alleles are grouped by HLA-HIV score (+5 to −5), then ordered by the coefficient of Z-score, pVL, CD4 count, %CD4 and CD4/CD8 ratio. Protective and risk alleles are shaded with progressively deeper green and orange colors, respectively. The number (n = ) of individuals expressing each HLA allele is shown. Blue vertical lines denote a coefficient equal to zero.

Figure 6

Correlation of 5-parameter HLA-HIV association scores within and between cohorts Scatter plot of univariable (Mann-Whitney) and multivariable HLA-HIV scores in the combined MEX/CAM (A), MEX (B) and CAM (C) cohorts. Panel (D) shows a scatter plot of univariable HLA-HIV scores between MEX and CAM cohorts. Correlation between HLA-HIV scores were determined using Spearman’s rank test. Random jittering was used to prevent dots being superimposed. For A, B & C panels only HLA alleles with at least one association with a HIV clinical parameter were considered.