Genomic and Epigenomic Characterization of Acute Myeloid Leukemia 

Get Permission

In a study by the Cancer Genome Atlas Research Network reported in The New England Journal of Medicine, genomes of 200 adult cases of de novo acute myeloid leukemia (AML) were analyzed by whole-genome sequencing (n = 50) or whole-exome sequencing (n = 150) to identify mutations and relationships between mutation patterns and epigenetic phenotypes involved in AML pathogenesis.1 The investigators identified at least one driver mutation in nearly all samples and found that “a complex interplay of genetic events contributes to AML pathogenesis in individual patients.”

Small Number of Mutations

AML genomes were found to contain fewer mutations than most other adult cancers. A total of 2,315 single nucleotide variants and 270 small insertions and deletions were found in the coding region of the genome, with an average of only 13 mutations per sample. Of these, an average of 5 were in genes that are recurrently mutated in AML. In three samples without recurring coding mutations, fusion events already known to initiate AML were found.

Additional analysis showed that samples with MLL fusions had the smallest number of recurrent coding mutations, suggesting that MLL fusions require fewer cooperating mutations than other AML-initiating events. Similarly, lower numbers of recurring mutations were found in the group of samples containing PML-RARA fusions. Greater numbers of recurrent mutations were found in groups containing RUNX1-RUNX1T1 fusions or the combination of high-risk cytogenetic profile and TP53 mutation.

A total of 23 genes were significantly mutated and another 237 were mutated in two or more samples. The significantly mutated genes included those well established as relevant to AML development (eg, DNMT3A, FLT3, NPM1, IDH1, IDH2, and CEBPA) along with some only recently implicated (eg, U2AF1, EZH2, SMC1A, SMC3). Some of the mutations commonly found in AML (eg, in DNMT3A, NPM1, CEPBA, IDH1/2, and RUNX1) were mutually exclusive of transcription-factor fusions, suggesting that these mutations may have roles in initiation of AML that are similar to the roles of fusion genes.

Distribution of Mutations in Gene Categories

Among the 200 samples, all but one had nonsynonymous mutations in at least one of nine categories that are considered to be important to AML pathogenesis, including transcription-factor fusions in 18% of cases, the gene encoding nucleophosmin (NPM1) in 27%, tumor-suppressor genes in 16%, DNA methylation-related genes in 44%, signaling genes in 59%, chromatin-modifying genes in 30%, myeloid transcription factor genes in 22%, cohesin-complex genes in 13%, and spliceosome-complex genes in 14%.

Overall, FLT3 mutations were found in 56 samples, and an additional 62 samples had mutations in genes coding other kinases, phosphatases, or RAS family proteins, although most of these genes contained mutations in only one to three samples (except for KIT, KRAS, NRAS, and PTPN11). Although it has been suggested that an activating mutation in a gene encoding a signaling protein might be required for AML pathogenesis, a total of only 59% of samples had mutations in a gene encoding a signaling protein.

Analysis of mutual exclusivity among sets of genes revealed a number of patterns. The strongest patterns of exclusivity were exhibited by: (1) the transcription-factor fusion genes, NPM1, RUNX1, TP53, and CEPBA; (2) mutations in FLT3 and genes encoding other tyrosine kinases, serine-threonine kinases, protein tyrosine phosphatases, and RAS family proteins; and (3) mutations in ASXL1 and in genes encoding components of the cohesin complex, other myeloid transcription factors, and other epigenetic modifiers. The finding of mutual exclusivity within particular biologic classes—eg, genes encoding the cohesins, proteins of the spliceosome, signaling proteins, and histone-modifying proteins—suggests that one mutation in these pathways may be sufficient for AML pathogenesis.

Novel Subtype of AML?

Analysis of pairwise relationships between mutations showed that the most prominent co-occurrence was among mutations in FLT3, DNMT3A, and NPM1. Samples with co-occurrence of these mutations were strongly associated with distinct clusters of mRNA expression, miRNA expression, and DNA methylation loss pattern, suggesting that the presence of mutations in these three genes represents a novel subtype of AML. Strong exclusivity was exhibited by PML-RARA, MYH-11-CBFB, and MLL-containing fusion genes vs mutations in NPM1 and DNMT3A; and RUNX1 and TP53 mutations were strongly exclusive of FLT3 and NPM1 mutations.

Additional findings when mRNA and miRNA expression data were integrated with clinical and mutation data included the previously reported finding that differentiation states of AML samples are highly correlated with expression signatures. In addition, samples with PML-RARA fusions had distinct mRNA and miRNA signatures that were also highly correlated with a specific DNA methylation signature. All transcription-factor fusions were correlated with specific mRNA expression signatures, and PML-RARA, RUNX1-RUNX1T1, and some MLL fusions were associated with specific miRNA signatures.

Notable findings on DNA methylation analysis, in addition to the observation of a characteristic signature for samples with co-occurring FLT3, DNMT3A, and NPM1 mutations, included extensive loss of DNA methylation in some samples with MLL fusions and specific patterns of gain or loss in samples with CEBPA mutations or PML-RARA, RUNX1-RUNX1T1, or MYH11-CBFB fusions.

The investigators noted, “This data set will be available to provide a framework for future studies that pertain to the molecular classification of patients with AML. The identification of many potentially important relationships among recurrently mutated AML genes and pathways provides a comprehensive foundation for an understanding of the genetic rules of pathogenesis.” ■

Disclosure: The study was funded by the National Institutes of Health.


1. The Cancer Genome Atlas Research Network: Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia. N Engl J Med May 1, 2013 (early release online). 

Related Articles

COMMENTARY: Molecular Landscaping of Acute Myeloid Leukemia: Are We Relearning the Past or Informing the Future?

Acute myeloid leukemia (AML) is a clinically and molecularly heterogeneous disease.1 This concept has been supported by more than 4 decades of studies showing distinct outcomes of subsets of patients that differ in age, disease type (primary vs secondary vs therapy-related), and cytogenetic and...