Acute GvHD of the Gut Is Associated with Minor Histocompatibility Antigens Cross-Reactive Against Gut-Tropic Viral Epitopes

Background: The gastrointestinal system (GI) is often affected by acute Graft-versus-Host-Disease (aGvHD), and severe manifestations of aGvHD of the gut portend a poor prognosis after allogeneic hematopoietic cell transplantation (allo-HCT). Therefore, new tools to predict – and possibly mitigate – GI aGvHD risk are urgently needed. In the context of HLA-matched allo-HCT, T cell alloreactivity is directed against minor histocompatibility antigens (mHAgs), polymorphic peptides resulting from donor-recipient (D-R) disparity at sites of genetic polymorphisms. We have previously developed an analytic framework to systematically identify mHAgs, based on integration of polymorphism detection by whole-exome sequencing (WES) of germline DNA from D-R pairs together with organ-specific expression and HLA class I epitope prediction (Cieri, Nat Biotechnol in press). Building on the clinical observation that viral reactivations often coincide with the development of GvHD, we hypothesized that sequence homology of alloreactive peptides with gut-tropic viral epitopes (i.e., from adenovirus [ADV], CMV and EBV) may contribute to the pathophysiology of GI aGvHD.

Methods: We devised a computational algorithm to quantify the extent of cross-reactivity across the patient-specific mHAg landscape, inferred from analysis of WES of allo-HCT D-R pairs. Complete viral proteomes for ADV, CMV and EBV were retrieved from UniProt (available strains/virus: 51, 14 and 6, respectively) to create in silico all possible 8-11mers (n=1,108,265) that were then subjected to HLA class I epitope prediction, resulting in a catalog of 184,889 viral epitopes across 105 HLA class I alleles. Predicted mHAgs from WES analysis of 220 HLA-matched D-R pairs were filtered for expression in GI-resident non-hematopoietic cells, as identified by single cell RNA-Seq datasets of GI tissue from healthy and post-transplant individuals. Homology with predicted viral epitopes was defined by either: (i) sliding window, wherein the mHAg and viral antigen were required to share ≥6 consecutive amino acids; or (ii) skipping window, wherein the Levenshtein distance between the mHAg and viral antigen was ≤2 amino acids, irrespective of their position within the epitope sequence. Predicted mHAgs were considered cross-reactive only if displaying sequence homology with viral epitopes presented on the same HLA restriction.

Results: Across the 220 HLA-matched allo-HCT D-R pairs analyzed (for recipients with AML/MDS), the median number of cross-reactive GI-specific mHAgs was 37 (range: 13 – 87). Patients who developed severe GI aGvHD (grade III-IV, n=11) had a higher load of cross-reactive GI-specific mHAgs than patients who did not (P = 0.038). We considered the impact of diagnosis, prognostic risk score, status at transplant, conditioning regimen, graft source, donor age, and CMV status against median cross-reactive mHAg load on risk for GI aGvHD. Across these patients, only cross-reactive mHAg load > median was associated with increased risk of developing severe GI acute GvHD (HR = 4.96, 95% CI: 1.13 – 19.44, P = 0.03). To functionally verify cross-reactivity, we focused on the GI-specific mHAg repertoire of 2 patients in our cohort who experienced severe GI aGvHD and shared the common HLA-A*03:01, B*07:02 and C*07:02 haplotype. We first challenged T cells from healthy donors with the same haplotype with synthetic peptides corresponding to predicted viral epitopes. After 2 rounds of stimulation, T cells were screened for antigen specificity by dextramer staining. Of the 21 peptides tested, 4 (1 HLA-B*07:02, 3 HLA-C*07:02; 2 ADV- and 2 EBV-derived epitopes) elicited viral-specific responses. T cells were then co-cultured with autologous antigen-presenting cells pulsed with the viral epitopes or cross-reactive mHAgs, and T cell activation was assessed by CD137 upregulation. For all 4 viral epitopes, viral-specific T cells upregulated CD137 also in response to the cross-reactive GI-specific mHAgs, thereby suggesting their potential for cross-reactivity.

Conclusion: Overall, our findings indicate cross-reactivity against gut-tropic viruses as a key contributor to GI aGvHD pathophysiology. Molecular characterization of D-R pairs to quantify cross-reactivity provides a promising tool for pre-transplant prognostication and could facilitate the design of personalized post-HCT treatments to minimize this highly morbid condition.