Donor HLA-DQ genetic and functional divergence affect the control of BK polyoma virus infection after kidney transplantation

Abstract

BK polyomavirus (BKPyV) infection remains a major concern after kidney transplantation, increasing the risk of graft loss in the absence of specific antiviral agent now available. Here, we investigated the impact of HLA diversity on the control of posttransplant BKPyV replication. High HLA evolutionary divergence (HED) at the DQ locus in the donor was an independent predictor of BKPyV-free outcome. More generally, we highlighted the protective effect of highly divergent pairs of HLA-DQ heterodimers corresponding to heterozygous HLA-DQα01/non-DQα01 combinations. We then defined a functional divergence metrics assessed by the similarity of peptide-binding motifs between pairs of HLA-DQ molecules. Greater functional divergence correlated with the size of the BKPyV-derived DQ-bound immunopeptidome and a lower risk of BKPyV reactivation, thus providing a molecular basis for the observed genetic differences. Together, these data provide evidence for a direct link between donor HLA-DQ genetic and functional divergence, diversity of the DQ-bound immunopeptidome, and control of viral infection, likely reflecting stronger antiviral T cell responses.

Fig. 1.. Donor HLA-DQ diversity is associated with control of posttransplant BKPyV infection.

(A to C) BKPyV cumulative incidence postkidney transplantation in patients (A) above or below median age, (B) with number of HLA mismatch above or below 6, and (C) with donor HED-DQAB above or below median. (D) Distribution of donor HED-DQA (left) and HED-DQB (right) values. (E and F) Donor HED-DQA (F) and HED-DQAB (G) in individuals with two alleles of the DQA1*01 supertype, an allele of the DQA1*01 supertype together with a non-DQA1*01 allele, or with two non-DQA1*01 alleles (i.e., of the DQA1*02, 03, 04, 05, or 06 supertype). (G) BKPyV cumulative incidence postkidney transplantation in patients with donors showing DQα01/DQα01, DQα01/non-DQα01, or non-DQα01/non-DQα01 pairs. P values from log-rank tests [(A) to (C) and (H)] are indicated on each graph. MM, mismatch.

Fig. 2.. Association of HLA-DQ genetic and motif divergence with the BKPyV-derived immunopeptidome.

(A) Principal components (PC) analysis on the binding specificities from HLA-DQ molecules (outer colors: HLA-DQα chain, inner colors: HLA-DQβ chain). Squares: DQα01 group; triangles: non-DQα01 groups. (B) Example of motif divergence computed for the HLA-DQ molecules from two donors. Motif divergence is first computed between each pair of stable HLA-DQ heterodimers from the donor (for donor 1, this corresponds to a single comparison; donor 2 has four stable HLA-DQ heterodimers, resulting in six pairwise comparisons). Values from each pairwise comparison are indicated (arrows). The final motif divergence value assigned to the donor corresponds to the maximum value and is indicated in bold. (C) BKPyV cumulative incidence postkidney transplantation in patients stratified based on donor’s HLA-DQ motif divergence. (D to I) Number of predicted DQ-bound peptides (15-mers overlapping the BKPyV proteome) determined for each donor of the entire cohort. (D) Number of unique peptides from the indicated BKPyV proteins predicted to bind the HLA-DQ molecules of each donor. (E) Number of unique predicted binders normalized to the length of the corresponding protein. (F and G) Correlation between the total number of predicted BKPyV-derived DQ-bound peptides and the HED-DQAB (F) or motif divergence (G) scores for each donor. Spearman’s rank correlation coefficients (R) are indicated. (H) Total number of predicted BKPyV-derived peptide binders in donors from indicated groups. (I) BKPyV cumulative incidence according to the number of VP2 peptides predicted to bind donors’ HLA-DQ molecules (using the optimal cutoff threshold, i.e., 60 peptides). P values from log-rank [(C) and (I)] or Mann-Whitney (H) tests are indicated on each graph. ns, not significant; **P < 0.01 and ****P < 0.0001.

Fig. 3.. Time-varying causal inference: Effect of HLA divergence on the risk of posttransplant BKPyV reactivation.

(A) Multivariable analyses using Cox proportional hazard including HLA-DQ divergence (assessed by DQα combinations) and all variables statistically significant in the univariable analysis. Adjusted hazard ratios (95% confidence interval) are indicated for each variable with corresponding P values. (B) Target trial emulation (as described in table S2) including all covariates of the Cox model was performed using the g-method, to estimate the causal effect of donor HEDs on the BKPyV reactivation. Donor HEDs for indicated loci were compared by segregating individuals into two groups based on the median for each HED. N = 511 patients with no missing data in all variables were included in the analysis.