Genetic Signatures Associated With Clinical Outcome in Acute Myeloid Leukemia

Abstract

Acute myeloid leukemia (AML) is a disease with marked heterogeneity in both response to therapy and survival. Structural genomic variations are of predominant importance to the biology and clinical outcome of patients with AML, and conventional karyotype-based risk classifications are used routinely in clinical decision making in AML. However, almost half of AML patients have normal karyotype, and this subgroup shows various responses to therapy. Recent advances in genome-wide single-nucleotide polymorphism (SNP) analyses have revealed previously unrecognized microdeletions and copy-neutral loss of heterozygosity (cnLOH) in a broad spectrum of human cancers. As AML represents a genetically heterogeneous disease, this technology might prove helpful, especially for cytogenetically normal AML patients. Here, we have analyzed 54 AML blast-derived DNAs with normal karyotypes and 200 healthy controls using Illumina 317K SNP arrays. Chromosome aberrations (≥ 2 Mb) were more frequently observed in AML patients (30%) than healthy controls (4%). Regions of cnLOHs were identified in 26% of patients and chromosome 13 was most frequently affected. Average length of aberrations per individual was also larger in AML patients (14.9 Mb) than healthy controls (0.2 Mb). A total of 4,438 and 2,066 copy number variations (CNVs, < 2 Mb) were detected in AML patients and healthy controls, respectively. The average number and size of CNVs per individual were greater in AML patients (82.2 CNVs and 101.2 Kb) than healthy controls (10.3 CNVs and 65.1 Kb). In further analysis, we showed that the frequency and average size of chromosome aberrations were associated with complete remission attainment in AML patients. These data show the potential of high-resolution SNP analysis for identifying genomic regions of potential pathogenic and clinical relevance in AML. In addition, the genetic association between polymorphisms and chemotherapy outcomes was examined in 103 Korean AML patients and 241 normal controls. A total of 293 SNPs in nineteen candidate genes were genotyped using the Illumina Golden Gate genotyping assay. We selected 105 SNPs in five cytosine arabinose (AraC) metabolism genes (SLC29A1, DCK, NT5C3, CDA and DCTD), 36 SNPs in three drug metabolism genes (NQO1, MTHFR and YMS), and 152 SNPs in eleven candidate genes from our previous GWAS study (NALP8, ADAM15, OR11H6, PLAC2, SPEN, GLCE, PARP15, MCM7, BTBD4, RSAD1, AASDH). The genetic effects of the SNPs on the outcome of chemotherapy were analyzed using logistic regression models and Coxs proportional hazard model. Although none of the SNPs were associated with risk of AML and complete remission rate, several common SNPs were associated with first complete remission (CR1), relapse, CR1 duration and overall survival. NT5C3 and PARP15 SNPs were associated with CR1 rate, one common SNP of NT5C3 (rs3750117) showed significant association with CR1 rate after first course of chemotherapy (P-value=0.00004 and odds ratio=13.04) in AML patients. NT5C3 expression levels were significantly increased in patients with risk allele homozygote of rs3750117. One non-synonymous SNPs in PARP15 (rs12489170, G394R) were also associated with CR1 rate (P-value=0.008 and odds ratio=0.32, respectively). A non-synonymous SNP in MCM7 (rs2070215, N144S) was also associated with risk of relapse (P-value=0.007 and odds ratio=0.37). In our survival analysis, rs2072671, rs12914938 and rs2853741 (SNPs of CDA, GLCE and TYMS genes, respectively) was associated with CR1 duration (P-value=0.004, 0.009 and 0.007), and rs2853741, rs6010659 and rs1534309 (SNPs of TYMS, BTBD4, MCM7, respectively) were also associated with overall survival (P-value=0.04, 0.02 and 0.03). In further statistical analysis, we created a genotype score from the number of risk alleles in three AraC metabolism genes (SLC29A1, NT5C3 and CDA). The genotype score provided a better prediction of CR1 than single allele alone. In our stepwise regression analysis, we can also discriminate the CR1 failure and success group with high accuracy. These data suggest that genotyping the candidate gene polymorphisms may have the potential to identify patients more likely to respond to chemotherapy. We are now close to a time when we will be able to use these prognostic factors and technologies to identify new targets for therapy and to determine who may benefit from that therapy, and ultimately change how we treat individual patients with AML.