Abstract
Graft-versus-leukemia (GVL) effect is considered to reduce relapse rate due to eradication of residual leukemia cells after allogeneic hematopoietic stem cell transplantation (HSCT). Segregation it from graft-versus-host disease (GVHD) has been main issue clinically. We recently clarified 16 high-risk HLA mismatch combinations and eight high-risk specific amino acid substitution positions for severe acute GVHD in six HLA loci. In the current study, we clarified HLA allele mismatch combinations and amino acid substitution positions associated with GVL effect. Consecutive 4643 patients transplanted for hematological malignancy (ALL, AML, CML, MDS, MM and ML) with T cell replete marrow from a serologically HLA-A, -B and -DR antigen-matched donor through Japan Marrow Donor Program were registered in this cohort study. All HLA-A, -B, -C, -DRB1, -DQB1 and -DPB1 alleles were retrospectively typed. The effect of HLA locus mismatch in allele level, the HLA allele mismatch combinations in HLA six loci and amino acid substitution positions on reduced relapse rate was analyzed using a multivariable competing risk regression model. As results (1) Mismatches of HLA-C (Odds ratio (OR)=0.69; p<0.0001) and HLA-DPB1 (OR=0.78; p<0.0001) were strongly reduced leukemia relapse, and HLA-A (OR=0.99; p=0.9), HLA-B (OR=0.98; p=0.91), HLA-DRB1 (OR=0.93; p=0.54) and HLA-DQB1 (OR=1.06; p=0.54) were not. (2) Total 10 HLA mismatch combinations were significantly associated with GVL effect; four in HLA-C allele (donor Cw*0303- patient Cw*1502 (n=25) OR=0.23, Cw*0102-Cw*1402 (n=16) OR=0, Cw*0801-Cw*0102 (n=10) OR=0 and Cw*1402-Cw*0304 (n=23) OR=0), six in HLA-DPB1 allele (DP*0402-DP*0201 (n=66) OR=0.41,?DP*0501-DP*0201 (n=351) OR=0.7,?DP*0501-DP*0401 (n=53) OR=0.45,?DP*0501-DP*0402?(n=121) OR=0.59, DP*0901-DP*0201 (n=50) OR=0.38 and DP*1301-DP*0201 (n=21) OR=0), but none in HLA-A, -B, -DRB1 and -DQB1 allele. Except two of four combinations in HLA-C, the other two in HLA-C and all six in HLA-DPB1 were different from high-risk one for severe acute GVHD. (3) Specific amino acid substitution at positions 9, 99, 156 in HLA-C molecule was elucidated as significant factors responsible for GVL effect and one of three was different from substitutions responsible for severe acute GVHD. As for HLA-DPB1, no significant association between the positions of specific amino acid substitution and GVL were found. In conclusion, large scale comprehensive analysis made it possible to identify 4 HLA-C and 6 HLA-DPB1 mismatch combinations responsible for GVL effects, some of which are different from one responsible for acute GVHD. Responsible amino acid substitutions on specific position were also elucidated in HLA-C, but not in HLA-DPB1. These findings suggest that donor selection according to these results could segregate GVL from acute GVHD, therefore these strategies might be beneficial for the selection of suitable donor for HSCT. And that, we speculate that the molecular base of GVL caused by the HLA-DPB1 mismatch might be different from that in HLA-C.
Discriminative modelling of context-specific amino acid substitution probabilities
