DNMT3A Mutations Associate with Shorter Survival and Modulate the Prognostic Impact of Mutated NPM1: an Analysis Based on Comprehensive Mutational Screening of 660 AML Patients Treated on German AML Cooperative Group (AMLCG) Trials

Background: Mutations in DNA methyltransferase 3A (DNMT3A) are common in acute myeloid leukemia (AML), affecting ~20% of patients (pts) and 30-40% of those with cytogenetically normal (CN-) AML. Although several groups have investigated their prognostic relevance, most studies focused on younger adults (<60 years [y]), and their results were inconsistent. Moreover, there is conflicting data regarding possible differences between mutations affecting the 'hotspot' codon R882 and other variants. We therefore performed comprehensive mutational analyses in 660 younger and older (>=60 y) AML pts treated on German AML Cooperative Group (AMLCG) protocols, and studied the association between DNMT3A mutations and outcomes.

Patients and Methods: We analyzed pretreatment blood or bone marrow specimens from 660 adult AML pts who received intensive induction chemotherapy on two consecutive phase III trials (AMLCG-1999, n=388, and AMLCG-2008, n=272; median age, 57y, range, 18-86y). Sequence variants in DNMT3A exons 7-23 and other genes known to be mutated in myeloid neoplasms were analyzed by multiplexed amplicon resequencing (Agilent Haloplex). Sequencing was performed on an Illumina MiSeq instrument using 2x250bp paired-end reads. Variants were classified as known/putative driver mutations, variants of unknown significance, or known germline polymorphisms based on published data including dbSNP, the Catalogue Of Somatic Mutations In Cancer (COSMIC) and The Cancer Genome Atlas (TCGA). Cytogenetic analyses were performed centrally.

Results: We identified 223 DNMT3A mutations in 207/660 pts (31%), including 180/449 pts (40%) with intermediate-risk cytogenetics according to the MRC classification (P<.001). Missense mutations affecting codon R882 were found in 114 pts, other missense mutations in 59, and truncating mutations (nonsense SNVs or frame shift variants) in 43. Nine pts had >1 type of DNMT3A mutation. DNMT3A mutations tended to be more frequent in older compared to younger pts (35% vs. 28%, P=.08) and were associated with female sex (38% vs 26% in males; P<.001), higher leukocyte counts (P=.008) and higher marrow blast percentages (P=.005). In the entire cohort, mutated DNMT3A associated with shorter relapse-free survival (RFS, hazard ratio [HR], 1.64, P<.001) and shorter overall survival (OS; HR, 1.26; P=.02). Outcomes were similar for pts with DNMT3A codon R882 mutations, other missense mutations, or truncating mutations. Shorter RFS and OS of DNMT3A-mutated pts was also observed in the subgroup with intermediate-risk cytogenetics (RFS: HR, 1.62; P=.002 and OS: HR, 1.34; P=.02). DNMT3A mutations associated with inferior outcomes in younger pts (RFS: HR, 1.58; P=.02 and OS: HR, 1.55; P=.005), while in older pts, no significant impact of mutated DNMT3A as a single marker on RFS or OS was observed. Due to the strong association of DNMT3A mutations (which appear to be prognostically unfavorable) with mutated NPM1 (an established favorable risk marker), we studied the four subgroups defined by the combination of both mutations. DNMT3A mutations associated with shorter RFS (Fig. A) in pts with mutated NPM1 as well as in those with wild-type NPM1, and with shorter OS in NPM1-mutated pts (Fig. B). When we considered the prognostically favorable 'molecular low risk' genotype (i.e., CN-AML with mutated NPM1 without FLT3 internal tandem duplication [ITD]), DNMT3A mutations associated with shorter RFS (Fig. C) and a trend for shorter OS (Fig. D) in pts with this combination, and with significantly shorter RFS and OS in the remaining ('high molecular risk') CN-AML pts. Finally, in a multivariate model adjusting for other clinical and genetic risk factors, mutated DNMT3A remained a significant risk factor for shorter RFS (HR, 1.44; P=.01) and OS (HR, 1.26; P=.04).

Conclusion: In our cohort of intensively treated AML pts covering a broad age range, we found that DNMT3A mutations associate with inferior survival and modulate the prognostic impact of mutated NPM1, confirming data recently reported by the MRC group (Gale et al., J Clin Oncol 33:2072). In contrast to this and other published reports, we observed no outcome differences between different types of DNMT3A mutations. Information on DNMT3A mutation status further refined the risk stratification of CN-AML based on the NPM1 mutated / FLT3-ITD negative genotype, supporting a role for DNMT3A mutations as a prognostic marker.