A Refined Model of HLA-DP Permissiveness Improves Stratification of Acute Graft-Versus-Host Disease Risks after Unrelated Hematopoietic Cell Transplantation: A Retrospective Study from the CIBMTR

Abstract

Introduction: Permissive HLA-DPB1 mismatches defined by the T-cell epitope (TCE) model are an established selection criterion for unrelated donors in allogeneic hematopoietic cell transplantation (alloHCT) (Dehn et al. Blood 2019). Based on biological evidence, the TCE model has classified HLA-DPB1 alleles into at least three functional groups, one of which (TCE group 3; TCE3) houses a large number of alleles with different structural and functional characteristics. We have recently shown that structurally close HLA-DP allotypes have similar peptide-binding motifs and share a significant proportion of their immunopeptidomes, the latter being fundamental for permissiveness (Meurer & Crivello et al. Blood 2021). Hence, we hypothesized that HLA-DPB1 mismatches involving alleles that encode structurally distant allotypes within TCE3 could be less permissive than those involving alleles that encode structurally closer allotypes, and thus have a differential impact on clinical outcomes.Methods: Multidimensional scaling techniques were used to compare 28 polymorphic positions (amino acids 8-215) among 51 alleles present in a cohort of 5,140 10/10 matched patient-donor pairs who received a first alloHCT for AML, ALL, or MDS between 2008-2017. Based on these analyses, TCE3-permissive mismatches (N=2,216) were further stratified into those involving structurally close or more distant combinations and compared with HLA-DPB1-matched (N=785) and non-permissively mismatched (N=2,023) pairs. These models were tested in parallel to the “classical” TCE model considering permissive mismatches (N=2,332) as a whole, to determine their association with overall survival (OS), disease-free survival (DFS), treatment-related mortality (TRM), primary disease relapse, and acute (a) and chronic (c)GVHD. Kaplan-Meier analysis and log-rank testing were used to compare the median OS and DFS. The incidences of GVHD, relapse and TRM were compared using competing risks and Gray's test. The effect of HLA-DPB1 mismatch on time-to-event outcomes was modelled by Cox regression after adjusting for confounders and testing for the proportional hazards assumption.Results: Within TCE3, we identified a subgroup of 4 frequent and structurally as well as functionally closely related alleles (i.e. DPB1*02:01, 04:01, 04:02, 23:01) that form a separate cluster (Figure 1A). These “core” alleles have similar bound-peptide motifs (van Balen et al. J Immunol 2020) and can be distinguished from other alleles in TCE3 in terms of the strength of in vitro alloreactive responses elicited from permissive responders (Meurer et al. Front Immunol 2018). Moreover, principal component analysis identified the HLA-DP 84-87 DEAV/GGMP motif as a major factor driving structural variability among TCE3 alleles (not shown). We used these observations to stratify TCE3 permissive mismatches in the allo-HCT cohort into “core” (N=930) and “non-core” (N=1,286) or into DEAV/GGPM-matched (N=1,209) and mismatched (N=1,007) pairs (Figure 1B).Multivariable analysis confirmed the association of aGVHD2-4 for the classical TCE model of non-permissive mismatching (p<.0001) and revealed a trend in DEAV/GGPM and “core”/“non-core” TCE3-permissive models. When compared to HLA-DPB1 allele matched pairs the risks of aGVHD2-4 increased progressively with “core” TCE3-permissive (HR 1.12 [0.98-1.28]; p=0.1012), “non-core” TCE3-permissive (HR 1.24 [1.06-1.46]; p= 0.0082), and non-permissive mismatches (HR 1.32 [1.16-1.50]; p<.0001) (Figure 1C, “core” vs. “non-core” HR 0.90 [0.80-1.01]; p=0.062). Similar albeit less significant results were obtained with the DEAV/GGPM model. The “core”/“non-core” TCE3 model was also associated with TRM with alloHCT from “core” TCE3-permissive donors showing lower risks of TRM than “non-core” TCE3-permissive (HR 0.82 [0.70-0.96]; p=0.0118) and non-permissive donors (HR 0.78 [0.68-0.88]; p=0.0002).Conclusion: Our results suggest that structural differences within TCE3 that reflect functional divergence and differential immunogenicity of alleles in this group associate with the risks of aGVHD and TRM after alloHCT. Hence, within the population of 10/10 matched donors, selection of “core” TCE3-permissive donors might reduce patient morbidity after transplantation. [Display omitted] DisclosuresPaczesny: Medical University of South Carolina: Patents & Royalties: inventor on the ST2 bispecific antibody patent application. Lee: AstraZeneca: Research Funding; Incyte: Research Funding; Janssen: Other; Kadmon: Research Funding; Novartis: Membership on an entity's Board of Directors or advisory committees, Research Funding; Pfizer: Research Funding; Syndax: Research Funding; Takeda: Research Funding; National Marrow Donor Program: Membership on an entity's Board of Directors or advisory committees; Amgen: Research Funding.