Genomic and epigenomic landscapes of adult de novo acute myeloid leukemia

Abstract

Background: Many mutations that contribute to the pathogenesis of acute myeloid leukemia (AML) are undefined. The relationships between patterns of mutations and epigenetic phenotypes are not yet clear.

Methods: We analyzed the genomes of 200 clinically annotated adult cases of de novo AML, using either whole-genome sequencing (50 cases) or whole-exome sequencing (150 cases), along with RNA and microRNA sequencing and DNA-methylation analysis.

Results: AML genomes have fewer mutations than most other adult cancers, with an average of only 13 mutations found in genes. Of these, an average of 5 are in genes that are recurrently mutated in AML. A total of 23 genes were significantly mutated, and another 237 were mutated in two or more samples. Nearly all samples had at least 1 nonsynonymous mutation in one of nine categories of genes that are almost certainly relevant for pathogenesis, including transcription-factor fusions (18% of cases), the gene encoding nucleophosmin (NPM1) (27%), tumor-suppressor genes (16%), DNA-methylation-related genes (44%), signaling genes (59%), chromatin-modifying genes (30%), myeloid transcription-factor genes (22%), cohesin-complex genes (13%), and spliceosome-complex genes (14%). Patterns of cooperation and mutual exclusivity suggested strong biologic relationships among several of the genes and categories.

Conclusions: We identified at least one potential driver mutation in nearly all AML samples and found that a complex interplay of genetic events contributes to AML pathogenesis in individual patients. The databases from this study are widely available to serve as a foundation for further investigations of AML pathogenesis, classification, and risk stratification. (Funded by the National Institutes of Health.).

Figure 1. Characterization of Mutations

Panel A shows the numbers of verified, recurrent tier 1 mutations in each of 200 samples obtained from patients with AML, organized according to important cytogenetic and mutational findings. For each set of data, the middle horizontal line indicates the mean, and the shaded area indicates ±1 SD. P values are shown for the groups that had significant differences from the mean number of recurrent tier 1 mutations in all samples. NK denotes normal karyotype. Panel B shows significantly mutated genes, as identified by the MuSiC analysis suite, and the number of samples with each mutation. Panel C shows the number of discrete clusters of mutations with distinct variant allele frequencies (VAFs) for each of 50 samples that underwent whole-genome sequencing. Each discrete VAF cluster represents a founding clone or a subclone derived from it.^, Samples with one clone have only a founding clone, those with two clones have a founding clone and one subclone, those with three clones have a founding clone and two subclones, and so forth. Exome sequencing defined too few mutations to accurately define subclones. Each sample contained evidence of a single founding clone, and most had one or more subclones derived from the founding clone. The French–American–British (FAB) subtypes of the samples are designated. (See Table 1 for FAB subtypes of AML.)

Figure 2. Organization of Mutations into Categories of Related Genes

Shown are somatic, nonsynonymous mutations in individual genes and sets of genes, grouped into nine categories, including one single-gene category, as labeled on the left. Of the 200 samples evaluated, 199 (>99%) had at least one mutation in one of the listed genes or sets. Blue boxes indicate mutations that are exclusive across all categories; green boxes, mutations that co-occur in the same sample across different categories; and orange boxes, mutations that co-occur in the same sample in the same category. Computational analysis with the use of the Dendrix++ algorithm identified three significant, mutually exclusive groups of genes, annotated on the right as groups A, B, and C. The cytogenetic risk for each patient is shown at the bottom of the chart. Additional information about data in this figure is provided in Tables S17 through S20 in the Supplementary Appendix. Ser–Thr denotes serine–threonine, TF transcription factor, and Tyr tyrosine.

Figure 3. AML Gene Fusions

Panel A is a plot created with the use of Circos software showing in-frame (green) and out-of-frame (orange) gene fusions detected in the AML cohort in the Cancer Genome Atlas (TCGA) with the use of Trans-ABySS software. Ribbon widths are proportional to the frequency of a fusion event. Chromosomes are individually colored and are arranged clockwise from chromosome 1 to X, starting with chromosome 1 at 12 o’clock. No rearrangements involved the Y chromosome. The frequencies of in-frame and out-of-frame gene fusions are shown in Panels B and C, respectively. For gene names shown in red, one of the partner genes in that fusion was found to be mutated in at least one other AML sample from this data set. On the basis of chromosomal aberrations and genomic variants annotated in the Mitelman database from the Cancer Genome Anatomy Project (CGAP) (http://cgap.nci.nih.gov/Chromosomes/Mitelman), all previously identified gene fusions are shown in blue, a single known polymorphic fusion is shown in green, and all novel events are shown in red.

Figure 4. Unsupervised RNA and miRNA Expression Patterns

Shown are unsupervised consensus clusters for data obtained with the use of messenger RNA sequencing (Panel A) and microRNA (miRNA) sequencing (Panel B). Shown from top to bottom are RNA abundance heatmaps, with each messenger RNA or miRNA centered on its mean; atypical members of each group (shown in black), which have silhouette widths below 0.9 of the group’s maximum width; a silhouette-width profile (i.e., a dimensionless metric that reflects how well samples fit into compact and distinct clusters) that was calculated from the consensus membership matrix; and covariates (e.g., FAB subtypes), with P values for association corrected for multiple testing, at the far left and far right (see the Methods section in the Supplementary Appendix). B-H denotes Benjamini–Hochberg multiple-testing correction. The numbers refer to the silhouette-width profiles for which P values are provided. One asterisk denotes P<0.05, two asterisks P<0.01, and three asterisks P<0.001. The color scales for both heatmaps reflect mean-normalized log₂ abundances, with RPKM (reads per kilobase of exon model per million mapped reads) for RNA-sequencing data and log₂ RPM (reads per million) for miRNA-sequencing data. The scale-bar numbers (−2.5 for least abundant to 2.5 for most abundant) indicate the range of log₂ mean-centered abundance values in the heatmaps. Cytogenetic-risk profiles are shown at the bottom of the chart.

Figure 5. Unsupervised Analysis of DNA Methylation at Extremes of CpG Density

DNA-methylation values for specific CpG residues are shown as a proportion, ranging from 0% (unmethylated, in blue) to 100% (fully methylated, in red), for unsupervised clustering of CpG-dense regions of the genome (Panel A) and of CpG-sparse regions (Panel B). Covariates are shown below the corresponding samples. Data for CD34+CD38− bone marrow cells, promyelocytes, neutrophils, and monocytes from three healthy volunteers are plotted to the left of the data for 192 AML samples in each panel. CpG islands and shores are annotated in dark green and light green, respectively, in the space between the normal and AML samples. CpG density was computed as the ratio of observed to expected CG dinucleotides in a 3-kb window, as described by Saxonov et al. The 1000 most variable loci among those falling into the top and bottom 5% according to CpG density are plotted in Panels A and B, respectively. Cytogenetic-risk profiles are shown at the bottom of the chart.