Genomic Complexity Versus Gene Mutations In Predicting Overall Survival In Adult Acute Myelogenous Leukemia

Genomic complexity as measured through SNP array-based (SNP-A) genomic profiling is a strong negative predictor of survival outcome in adult acute myelogenous leukemia (AML). The recent discovery of multiple novel recurrently mutated genes in AML has led to the development of several prognostic models based on various combinations of genes along with the well-established risk factors of age and karyotype, but these models do not account for the strong effect of SNP-A-based genomic complexity. In this study, we seek to determine the relative importance of AML genomic complexity and gene mutation status on overall survival (OS) in AML.

We employed SNP-A genomic profiling of acquired copy number aberrations (aCNA) and copy neutral LOH (cnLOH) together with sequence analysis of 13 recurrently mutated genes to determine aCNA/cnLOH lesion load and gene mutational status for 156 consecutively collected previously untreated AML patients. AML cell samples were processed with a Ficoll gradient, negatively selected using Miltenyi microbead columns, and then further purified with flow cytometric cell sorting. Processed DNA isolated from highly purified AML blasts and paired buccal DNA was subsequently hybridized to Affymetrix SNP 6.0 arrays. aCNAs were visually identified using the dChip program in paired data displays and corroborated by algorithmic lesion scoring, and cnLOH was detected using internally developed software. Using the same DNA, we resequenced the recurrently mutated exons of NPM1, FLT3, CEBPA, IDH1, IDH2, NRAS, KRAS, and TP53, and all coding exons of RUNX1, ASXL1, TET2, DNMT3A, and BCORL1. Clinical data including age, overall survival, and cytogenetics were ascertained. Cytogenetic risk categories were assigned based on the SWOG criteria. Univariate analyses were performed using Kaplan-Meier estimates of survivor functions, and Cox proportional hazards models were used for multivariate analyses.

At the time of analysis, 119 (76%) patients had died. aCNA/cnLOH were common with ≥1, ≥2, and ≥3 lesions detected in 62%, 38%, and 26% of cases, respectively. Univariate overall survival analysis of all clinical and molecular variables demonstrated a significantly increased OS for mutations of NPM1 (p=0.01) and decreased OS for mutations of TP53 (p<0.001), aCNA/cnLOH load ≥2 vs. <2 (p=0.001), aCNA/cnLOH load ≥3 vs. <3 (p<0.001), age >60 (p<0.001), unfavorable vs. intermediate cytogenetics risk (p<0.001), and intermediate vs. favorable cytogenetic risk (p=0.01). Mutations in the other 11 genes analyzed were not prognostic. Multivariate analysis inclusive of either genomic complexity or gene mutations and always incorporating age- and cytogenetic-based risk categories demonstrated a statistically significant difference in the hazard of death for aCNA/cnLOH load ≥2 (HR 3.13, p=0.03), aCNA/cnLOH load ≥3 (HR 6.53, p=0.001), and TP53 mutations (HR 24.32, p<0.001).

Within the total cohort, 103 patients (66%) were diagnosed with de novo AML. In this group of de novo AML, aCNA/cnLOH were again common with ≥1, ≥2, and ≥3 lesions in 56%, 29%, and 18% of cases, respectively. Statistically significant univariate variables adversely affecting OS included mutations of TET2 (p=0.023) and TP53 (p<0.001), aCNA/cnLOH load ≥2 vs. <2 (p=0.02), aCNA/cnLOH ≥3 vs. <3 (p<0.001), age >60 (p<0.001), unfavorable vs. intermediate cytogenetics risk (p=0.02), and intermediate vs. favorable cytogenetic risk (p=0.04). Similarly, multivariate analysis revealed a statistically significant difference in the hazard of death for aCNA/cnLOH load ≥3 (HR 11.3, p=0.002) and TP53 mutation (HR 74.8, p<0.001).

Genomic complexity is common in adult AML and constitutes a dominant independent predictor of short survival outcomes. Through use of multivariate analysis incorporating genomic complexity and the mutation status of 13 recurrently mutated genes and controlling for age- and cytogenetic-based risk, only genomic complexity and mutations of TP53 emerge as significant independent predictors of short overall survival.

No relevant conflicts of interest to declare.