OBJECTIVE We sought to determine whether the -6 exon 13 T>C polymorphism in the DNA mismatch repair gene hMSH2 modulates susceptibility to acute myeloid leukemia after therapy and particularly after O(6)-guanine alkylating chemotherapy. We also determined the extent of microsatellite instability (MSI) in therapy-related acute myeloid leukemia (t-AML) as a marker of dysfunctional DNA mismatch repair. METHODS Using a novel restriction fragment length polymorphism, verified by direct sequencing, we have genotyped 91 t-AML cases, 420 de novo acute myeloid leukemia cases, and 837 controls for the hMSH2 -6 exon 13 polymorphism. MSI was evaluated in presentation bone marrow from 34 cases using the mononucleotide microsatellite markers BAT16, BAT25, and BAT26. RESULTS Distribution of the hMSH2 -6 exon 13 polymorphism was not significantly different between de novo acute myeloid leukemia cases and controls, with heterozygotes and homozygotes for the variant (C) allele representing 12.2 and 1.6%, respectively, of the control population. However, the variant (C) hMSH2 allele was significantly overrepresented in t-AML cases that had previously been treated with O(6)-guanine alkylating agents, including cyclophosphamide and procarbazine, compared with controls (odds ratio, 4.02; 95% confidence interval, 1.40-11.37). Thirteen of 34 (38%) t-AML cases were MSI positive, and 2 of these 13 cases were homozygous for the variant (C) allele, a frequency substantially higher than in the control population. CONCLUSIONS Association of the hMSH2 -6 exon 13 variant (C) allele with leukemia after O(6)-guanine alkylating agents implicates this allele in conferring a nondisabling DNA mismatch repair defect with concomitant moderate alkylation tolerance, which predisposes to the development of t-AML via the induction of DNA mismatch repair-disabling mutations and high-grade MSI. Homozygosity for the hMSH2 variant in 2 of 13 MSI-positive t-AML cases provides some support for this model.