Abstract 3050

Approximately 30% of Caucasian and 70% of African-American hematopoietic stem cell transplant (HCT) patients are unable to find an 8/8 HLA matched unrelated donor. Mismatches at HLA-A, B, C, or DRB1 alleles reduce survival. Therefore, identification and avoidance of high-risk allele combinations and the associated amino acid substitutions (AAS) that negatively impact HCT outcomes may increase access to this treatment option and allow safer utilization of HLA mismatched donors.

Using random forest (RF) analysis, our group has previously reported the AAS associated with 100 day survival (D100S) in single HLA class I mismatched, DRB1 matched recipient-donor pairs. We now extend that analysis to one year outcomes of overall survival (1y OS), disease free survival (1y DFS), transplant related mortality (1y TRM), and acute graft-versus-host disease (aGvHD) grades III-IV using the same clinical variables (recipient age, disease type, disease status, and gender match) and 389 AAS position and types (AASPT). The AASPT were defined by the HLA locus, amino acid position in the HLA class I protein and the actual AAS, e.g. locus: HLA-C, position: 97, type: tryptophan to arginine = C97_WR. Patients (n=2107) received myeloablative (99%) HCT as treatment for ALL, AML, CML, and MDS in early and intermediate stage of disease between 1988 and 2003.

RF analysis, a tree-based method for classification was used to assign an importance score (IS), reflecting the association of each potential predictor variable with the outcome of interest. Logistic regression analyses were performed to determine the magnitude of the effect of each individual AASPT (n=600) relative to 8/8 matched cases (n=1507), adjusted for the four clinical variables.

Using the criteria of n≥10, a relatively high IS (≥5) and a highly statistically significant odds ratio (p<0.01), C 97_WR has a deleterious effect on all outcomes. HLA-C156_RW has a deleterious effect on all outcomes except on 1y DFS. HLA-C80_NK and C77_SN affects 1y OS and 1y DFS; HLA-C11_SA, C116_YS, and C24_AS affect 1y TRM and aGvHD III-IV. Eighteen additional AASPT were associated with a single outcome each. No AASPT at the HLA -B loci met the above criteria, which could be due to the small number of HLA-B mismatched cases (n=88) compared to HLA-A (n=179) and HLA-C (n=333) mismatched cases. Other AASPT conferred high point estimates of relative risk but the number of patients with a mismatch was too small to yield statistical significance.

The most common alleles associated with all of the AASPT listed above are: HLA-C*01:02/02:02; 01:02/15:02; 02:02/01:02; 03:03/04:01; 04:01/16:01; 14:02/15:02; and HLA-A*02:01/02:05; 24:02/24:03; 29:02/30:01; 30:01/30:02; 30:01/32:01.

To understand the potential biological significance of mismatched allele pairs in the molecular context of peptide antigen binding, computational models were constructed for the 10 most important AASPT. In silico screening of ∼500 unique peptides was conducted against each pair to determine differences in the peptide repertoire capable of binding to each mismatched molecule. We found that the mismatched pairs identified by RF as less permissive displayed greater loss in their ability to bind identical peptides compared to other mismatches in the binding groove predicted to be more permissive by RF analysis. Overall, the computational modeling suggests a different affinity and peptide binding repertoire between mismatched HLA molecules.

Results from these analyses indicate that only a small number (6.4%) of AASPT clearly confer adverse outcomes in HCT patients with single HLA class I mismatched unrelated donors, and it is likely that these AASPT are responsible for differential binding of immunogenic peptides. Validation studies in an independent dataset are in progress. Additional prospective studies should be performed to refine HLA matching algorithms in the mismatched setting that may increase donor availability and minimize the adverse effects of donor HLA mismatching.


No relevant conflicts of interest to declare.

Author notes


Asterisk with author names denotes non-ASH members.