Large-scale statistical mapping of T-cell receptorβsequences to Human Leukocyte Antigens

Abstract

Interactions between diverse sets of T-cell receptors (TCRs) and peptides presented by human leukocyte antigens (HLAs) are the foundation of the adaptive immune system but population-level analysis of TCR-HLA interactions is lacking. Here we use the TCRβ repertoire of 4,144 HLA-genotyped subjects to associate ∼10^6 public TCRs (i.e., TCRs observed in multiple subjects) with specific HLAs, providing a new window into the functional characteristics of HLAs. We find that the vast majority of these HLA-associated public TCRs are specific to unique HLA allotypes, not allelic groups, and to the paired α-β heterodimer of class II HLAs though we observe some exceptions and also that the breadth of the TCR response is proportional to HLA zygosity. Identification of public HLA-specific TCRs permits highly accurate imputation of 248 class I and II HLAs from the TCRβ repertoire alone. Notably, 45 HLA-DP and -DQ heterodimers cannot be imputed due to a lack of associated TCRs, despite high representation in our training set. Gene linkage analysis indicates these heterodimers primarily arise from trans-complementation resulting in non-functional α-β pairs. Cell sorting, clonal expansion, and comparisons between SARS-CoV-2-exposed and -naive populations suggest that public class I and class II HLA-associated TCRs we identify are primarily expressed on CD8+ and CD4+ memory T cells, respectively, which are responding to a mix of common antigens. Our results recapitulate fundamental immunology, provide critical new insights into the functionality of HLAs, and demonstrate the power and potential of population-level TCR repertoire sequencing.

Identifying immune signatures of common exposures through co-occurrence of T-cell receptors in tens of thousands of donors

“…We consider a small subset of the TCRs associated with one of two Class II HLAs (DRB1*07:01 or DRB1*05:01) as well as with one of two exposures (CMV or SARS-CoV-2) using methods we will describe below. We visualize the occurrence of these TCRs within the repertoires of T-DETECT donors determined to have one or both HLAs using our HLA imputation models (Emerson et al, 2015a; Zahid et al, 2024) and to have one or both exposures using our previously described diagnostic models (Emerson et al, 2015b; Snyder et al, 2020b).…”

“…We used a “pseudolabeling” approach to expand our database of HLA-associated TCRs beyond the 2,904,747 TCRs previously described (Zahid et al, 2024). We used our previously-described (Emerson et al, 2015b; Zahid et al, 2024) HLA inference models to infer donor status with respect to 131 HLAs for 27,606 donors from our T-DETECT cohort. We then identified TCRs associated with each HLA using the imputed HLA types of the repertoires using the same approach with which we originally associated TCRs with HLAs using genotyped HLA status.…”

“…We used a “pseudolabeling” approach to expand our database of HLA-associated TCRs beyond the 2,904,747 TCRs previously described (Zahid et al, 2024). We used our previously-described (Emerson et al, 2015b; Zahid et al, 2024) HLA inference models to infer donor status with respect to 131 HLAs for 27,606 donors from our T-DETECT cohort.…”

“…Next, we applied those HLA-imputation models to a much larger pool of donor repertoires of unknown HLA (27,606 of our T-DETECT donors) to infer those donors’ HLA types. Finally, we used those imputed HLA types to identify 3,805,455 TCRs having strong statistical association with one or more HLAs (one-sided Fisher’s Exact Test p < 1e-4) using the previously described L1LR method (Zahid et al, 2024).…”

“…Similarly, by identifying TCRs with higher prevalence in subjects expressing a particular HLA as compared to subjects not expressing that HLA, sets of TCRs may be associated to specific HLAs. Using TCRβ repertoires from 4,144 HLA genotyped subjects, Zahid et al associate ∼106 public TCRs (i.e., TCRs observed in multiple subjects) to hundreds of common HLAs (Zahid et al, 2024). They show that these sets of TCRs are enriched for T cells with specific HLA restriction and build models to impute donor expression of hundreds of HLAs with high sensitivity and specificity.…”