Predicting relevant immunogenetic disparities

In this issue of Blood,Crivello et al report a novel approach to characterize HLA-DPB1 mismatches between donor and recipients that are better tolerated than others after hematopoietic cell transplantation (HCT) from unrelated volunteer donors.¹ 

The identification of a histocompatible donor is critical to the success of allogeneic HCT. Given the limited availability of HLA-matched sibling donors, currently unrelated donors facilitate the majority of the transplant procedures. Although unrelated donor or cord blood options are available for nearly all patients, ∼25% of whites and up to 80% of other racial and ethnic groups have no full match for 8/8 HLA-A, HLA-B, HLA-C, and HLA-DRB1 alleles.²  Under current transplant approaches, single-locus mismatch at HLA-A, HLA-B, HLA-C, or HLA-DRB1 is associated with increased risk of severe acute graft-versus-host disease (GVHD) and mortality.³  Additionally, mismatch at HLA-DQ confers additional risk for GVHD, and growing evidence supports that certain HLA-DPB1 mismatches increase mortality.

Several efforts are underway to mitigate these risks. Investigators are testing new strategies to overcome the HLA-mismatch barrier, and in parallel, much work is done to identify those types of mismatch associated with greater or less risk. Those that appear to confer no additional risk above HLA-matched donors have been termed “permissive” and those with increased risk “nonpermissive.” Extensive work has identified allele mismatches and amino acid substitutions with differential risks, and these have highlighted the importance of mismatches altering HLA peptide-binding residues.^4-7 

Fleischhauer and colleagues have led efforts to examine differential risk of HLA-DPB1 mismatches. Initial modeling based on cross-reactivity patterns of T cells toward 23 HLA-DPB1 alleles led to a T-cell epitope (TCE) classification. TCE groups were defined by relative immunogenicity, and donor-recipient HLA-DPB1 mismatches were defined as nonpermissive if they involved alleles from different TCE groups or permissive if from the same TCE group.^8,9  Subsequent large registry analyses demonstrated permissively HLA-DPB1–mismatched unrelated donor-recipient pairs with outcomes similar to HLA-DPB1 allele-matched cases, and nonpermissive HLA-DPB1–mismatched pairs with significantly increased mortality compared with permissively HLA-DPB1–mismatched pairs.^2,5  These findings are highly relevant to unrelated donor selection strategies, as ∼80% are HLA-DPB1 mismatched. However, the major limitation of this system is that the molecular basis for this TCE classification is not well understood.

In subsequent work, this group mutated amino acids in 10 positions of HLA-DPB1*09:01 and measured the impact on T-cell allorecognition from unrelated individuals.¹⁰  Alloresponse was affected by amino acid substitution at multiple peptide binding residues (positions 9, 11, 35, 55, 69, 76, and 84), and this relative response was translated into a functional distance score (FDaa) for each amino acid. Individual HLA-DPB1 alleles were given a functional distance score (FDallele) based on the sum of their relevant FDaa scores, and these FDallele scores highly correlated with previously defined TCE groups. These data supported the concept that HLA-DPB1 TCE groups and their allied FDallele scores reflect structural differences that influence the T-cell alloresponse and opened the possibility of classifying any HLA-DPB1 allele based on the FDallele score.

In the current analysis from Crivello et al, the difference in FDallele scores (ΔFD) between HLA-DPB1–mismatched unrelated donor-recipient pairs (n = 379) was explored as a determinant of outcome after HCT. These donors were otherwise matched at HLA-A, HLA-B, HLA-C, HLA-DRB1, and HLA-DQB1 (10/10 match). The population underwent first HCT for acute myelogenous leukemia, acute lymphoblastic leukemia, or myelodysplastic syndrome between 2005 and 2014 and received either myeloablative or reduced-intensity conditioning and predominantly peripheral blood stem cell transplantations. GVHD prophylaxis was cyclosporine/methotrexate with or without anti–thymocyte globulin. Nonconservative amino acid substitutions at peptide-binding residues had significantly higher median FDaa scores. The ΔFD score was highly concordant with TCE classification, and those with ΔFD > 2.665 had significantly worsened overall survival and event-free survival on multivariate analysis. These intriguing data support that the position and type of amino acid substitution resulting from donor-recipient HLA-DPB1 allele mismatch (akin to previous data in class I mismatches) affect unrelated donor HCT outcome.

Contrary to what one would have expected, no significant effect of ΔFD was seen for the outcomes of acute and chronic GVHD or nonrelapse mortality, and paradoxically, the risk of malignancy relapse was increased with high ΔFD. Several next steps are warranted. First, verification of these findings in larger and independent patient populations is needed. Additional refinements to the current model may be facilitated by such larger studies. These may clarify the association between ΔFD and levels of recipient HLA-DPB1 expression, GVHD, and malignancy relapse. Second, similar functional distance-based approaches may bring new insight to the functional significance of mismatches at other class I and II HLA loci. Finally, integration of the FDallele classification system into current TCE-based donor selection strategies (regarding HLA-DPB1), as well as larger efforts to understand the relative importance of multiple competing permissive mismatches and other non-HLA donor factors, is needed to inform current best practices in donor selection. Taken together, these efforts hold significant promise to optimize donor selection to improve unrelated HCT outcome.