A Genome-Wide Analysis to Identify Novel Susceptibility Loci for Therapy-Related Acute Myeloid Leukemia

Therapy-related acute myeloid leukemia (t-AML) is a rare but fatal complication of cytotoxic therapy. Whereas sporadic cancer results from interactions between intermittent or low-level environmental exposures and multiple low-penetrance alleles, t-AML results from an acute exposure to potent genotoxic agents. Consequently, we hypothesized that only a small number of relatively higher-penetrance variants would predispose towards t-AML and that, therefore, the association signal from these t-AML risk alleles could be detected even in a modest-sized cohort. To test this, we undertook a pilot genome-wide association study to identify genetic variants associated with t-AML. We genotyped nonmalignant DNA from 80 well-characterized t-AML patients of European descent followed at the University of Chicago and 150 healthy controls using the Affymetrix GeneChip Human Mapping 10K array platform. Following the application of quality control filters, our overall call rate was 98.4%. There was no evidence for population stratification or other forms of bias in our data. Using permutation methods, we first assessed the evidence for an excess of disease-genotype associations in our data over what would have been expected by chance at multiple significance thresholds. Even at nominal significance thresholds, we observed a marked enrichment; for example, we identified 15 SNPs at p < 0.001, whereas we only expected 5.5, and at p < 0.05, we found 329 associations as compared to 278.6 expected by chance. Then, to determine whether this apparent enrichment over chance was significant, we calculated empirical enrichment p-values at each significance threshold. Even at p = 0.05, this enrichment was highly significant (p_enrich = 0.008), indicating that we have sufficient power to identify genetic variants likely to be truly associated with t-AML susceptibility. At p = 0.001, the enrichment for SNPs associated with t-AML over chance was almost 3-fold. The odds ratio for each SNP at this threshold was > 2.0, suggesting that many are likely to be moderately- or highly-penetrant t-AML risk alleles. Nine of the 15 SNPs significant at this threshold are in linkage disequilibrium (LD) with known genes; of these, 4 are intronic. Although none have been previously studied in t-AML, several encode proteins directly implicated in leukemogenesis, including TLE4, a tumor suppressor commonly deleted in a subset of patients with de novo AML; IPMK, a positive regulator of the pro-survival AKT kinase; and the IL-1 family gene cluster, which has been linked to cell proliferation and apoptosis resistance in AML blasts and has also been shown to promote tissue invasion by leukemic cells. To confirm our findings, we validated a subset of top associations in an independent cohort of 70 patients with t-AML and 95 controls. As a complementary approach to this association mapping, we employed a multi-locus method contrasting the extent of LD in cases and controls in 5-SNP sliding windows throughout the genome. We identified a number of plausible candidate genes, including FHIT and ESR1, genes known to be somatically mutated in t-AML. We also performed copy number analysis to identify regions that are recurrently deleted or amplified in cases but not controls. No such alterations were identified. Taken together, these analyses suggest that genetic variants predisposing to cancer are greatly enriched in t-AML as compared to sporadic cancer and can be detected even in a small patient cohort. Given that germline samples from patients with t-AML are extremely scarce, the demonstration that even a low resolution and modest-sized study can yield a compelling list of candidate genes and polymorphic variants is a particularly important step towards the goal of identifying patients at risk for t-AML at the time of their initial diagnosis so that their treatment can be modified to minimize this risk. Furthermore, because cytotoxic therapy is a potent surrogate for the environmental exposures that drive virtually all cancers, t-AML may be a powerful model for the study of gene-environment interactions in cancer. Hence, this study may point towards genetic risk factors not only for t-AML but also a variety of commonly-occurring cancers, given the appropriate environmental exposures.