Pediatric non-Down syndrome acute megakaryoblastic leukemia is characterized by distinct genomic subsets with varying outcomes

Abstract

Acute megakaryoblastic leukemia (AMKL) is a subtype of acute myeloid leukemia (AML) in which cells morphologically resemble abnormal megakaryoblasts. While rare in adults, AMKL accounts for 4-15% of newly diagnosed childhood AML cases. AMKL in individuals without Down syndrome (non-DS-AMKL) is frequently associated with poor clinical outcomes. Previous efforts have identified chimeric oncogenes in a substantial number of non-DS-AMKL cases, including RBM15-MKL1, CBFA2T3-GLIS2, KMT2A gene rearrangements, and NUP98-KDM5A. However, the etiology of 30-40% of cases remains unknown. To better understand the genomic landscape of non-DS-AMKL, we performed RNA and exome sequencing on specimens from 99 patients (75 pediatric and 24 adult). We demonstrate that pediatric non-DS-AMKL is a heterogeneous malignancy that can be divided into seven subgroups with varying outcomes. These subgroups are characterized by chimeric oncogenes with cooperating mutations in epigenetic and kinase signaling genes. Overall, these data shed light on the etiology of AMKL and provide useful information for the tailoring of treatment.

Figure 1

Pediatric and Adult Non-DS-AMKL are Genomically Distinct. Distributions of recurrent chromosome translocations and GATA1 mutations in pediatric and adult non-DS-AMKL. p value according to Pearson’s Chi squared test.

Figure 2

Gene Expression Analysis Confirms Genomic Subgroups. (a) Unsupervised clustering of patients using the expression values of the 100 most variant genes. (b) t-SNE visualization. (c) Expression of myeloid HOXA9 target genes most highly associates with gene expression in HOXr AMKL. Global association between AMKL subgroup and HOXA9 target gene expression was estimated using global test. Contributions of genes to the overall association are indicated by the height of the bars. Bars are ranked with the most significantly correlated genes on the left, and colored according to the subgroup with the highest association. Samples with less than 60% tumor purity were not included in this analysis; as a result no RBM15-MKL1 cases are represented.

Figure 3

Cooperating Mutations in Pediatric Non-DS-AMKL. (a) Recurrent genes in diagnostic and relapsed specimens targeted by SNV/Indel mutations. Genes for which three or more cases carried a lesion are shown. (b) Frequency of copy number alterations for cases SNP array data and/or paired whole exome sequencing data. The outer track indicates the chromosomal location. Amplification events are shown in red, deletions in blue, and copy neutral loss of heterozygosity are shown in orange. Total number of cases carrying the event are shown, tracks do not correspond to a patient sample. (c) Non-random associations between genomic AMKL subgroup and cooperating mutation. Circos plot showing co-occurrence in patients at diagnosis between grouped (n) cooperating mutations (top) and AMKL subgroup (bottom). n, number of genes within cooperating gene sets: CTCF/Cohesin: CTCF, STAG2, STAG3, SMC1A, NIPBL, SMC3, RAD21; JAK-STAT: JAK1, JAK2, JAK3, STAT5B; RAS: NRAS, KRAS, PTPN11. Global association p value of 2.8×10⁻⁸ (i.e. probability of random distribution) is estimated according to global test using a multinomial regression model. Individual associations: *, p <0.05; **, p<0.01, Fisher exact test.

Figure 4

Clinical Outcomes in Pediatric Non-DS-AMKL. (a) Probability of overall survival of pediatric non-DS-AMKL patients stratified according to the molecular subgroup. CBFA2T3-GLIS2 (n=16); KMT2Ar (n=15); NUP98-KDM5A (n=10); HOXr (n=13); RBM15-MKL1 (n=9); GATA1 (n=8); Other (n=16). Medium follow up time: 89 months. (b) Probability of event free survival of pediatric non-DS-AMKL patients stratified according to molecular subgroup. (c) Probability of cumulative incidence of relapse or primary resistance.