Comprehensive Characterization of Rare Recurrent Mutations in Pediatric AML

Introduction: Sequence variants (mutations) are the cornerstone of adult acute myeloid leukemia (AML) biology, however, their significance has been de-prioritized in younger patients in favor of more well-described structural alterations. We interrogated a genomic database from a large cohort of children and young adults for recurrent variants that, although are rare as single events, contribute to mutational events within families of AML-related genes with shared function and biology.

Methods: Whole genome, targeted capture, transcriptome, and/or Archer panel sequencing were retrospectively employed on available diagnostic specimens from patients with AML enrolled on 4 successive COG trials for de novo pediatric AML (NCT00002798, NCT00070174, NCT01407757, NCT01371981). Mutations with a variant allele frequency of <5% were excluded from analysis. In addition to the common variants, rare but recurrent variants (RRV) were subcategorized into 10 well-described families based on sequence homology, functional domains, and/or interacting partners. These include the additional sex combs-like (ASXL), cohesin, cyclin, E26 transformation-specific (ETS), GATA, isocitrate dehydrogenase (IDH), janus kinase (JAK), MYC network, protein tyrosine phosphatase (PTP) and spliceosome families of genes. Remaining genes were categorized into a single “miscellaneous” group.

Results: Of 2041 patients with genomic data (age: median 10.9 years, range 0.08-29.8 years), 1804 (88.4%) had at least 1 mutation, with the average patient having 3.9 mutations (median 2, range 0-57). When excluding common AML associated mutations (NRAS, FLT3, WT1, KRAS, KIT, NPM1, CEBPA), 1532 (75.1%) were found to have at least 1 RRV, the majority of which (59.2%) were clustered within gene families as described above. The most frequent RRV were within the PTP gene family (9.0%), which included PTPN11, PTPN12, PTPRF, PRPRH, and PTPRJ. RRV within the MYC network (4.1%) were also frequent, and included MYC, MAX, MGA, MYCBPAP, MLXIP and MYC-ITD. Additionally, cohesin RRV (6.5%), including STAG2, STAG3, RAD21, SMC3, SMC1A, and ANGPT1, and ASXL RRV (6.5%) were highly enriched in those with RUNX1::RUNX1T1 (cohesin 12.6%, ASXL1 11.5%, ASXL2 19.8%). Spliceosome RRV, including U2AF1, U2AF2, SF3B1, SRSF2, PRPF8, and ZRSR2 were seen in 2.9%. Other gene families with RRV include GATA (GATA1, GATA2, GATA3; 12.6%), cyclin (CCND1, CCND2, CCND3; 4.6%), ETS (ERG, ETS2, ETV1, ETV6, SPI1, ELF1, ELF4; 3.6%), and JAK (JAK1, JAK2, JAK3; 3.8%).

Evaluation of prognostic relevance of RRV by family provided significant insight into their functional biology, particularly based on their enrichment within RUNX1::RUNX1T1. Patients with RUNX1::RUNX1T1 had dramatically different 5-year event-free survival (EFS) based on the presence or absence of cohesin (83.5% vs 63.3%, p = 0.019) or spliceosome (37.5% vs 66.0%, p = 0.037) family mutations.

In contrast, patients with MYC network mutations had an equivalent EFS compared to those without (43.2% vs 45.6%, P=0.69) but these outcomes became discrepant based on impacted gene. Patients with MYC-ITDs had a superior EFS (69.8% vs 45.8%, p=0.025) while those with MGA mutations had an inferior EFS (25.0% vs 45.8%, p=0.041) compared to those without. Notably, this poor EFS was maintained even in those patients with a concurrent RUNX1::RUNX1T1 (38.5% vs 66.6%, p=0.023). Similarly, EFS within RUNX1::RUNX1T1 patients with ASXL family mutations was dependent on the mutated gene (ASXL1 52.3% vs ASXL2 72.5%, P=0.1).

Conclusions: In this study we demonstrate that, in contrast to current understanding, mutations are quite common in pediatric patients with de novo AML, where the majority of patients have recurrent mutations that impact families of genes with common functionality and are enriched in specific fusion groups. Patients with mutations in the spliceosome and cohesin families had significant but discrepant outcomes within the RUNX1::RUNX1T1 fusion, suggesting that the impact conferred by these mutations is not limited to their co-occurrence profile. In contrast, outcomes for patients with MYC network or ASXL family mutations were contingent on the specific mutated gene, underscoring that not all genes from the same family may confer the same functional or prognostic significance. This categorization should facilitate prognostication, risk stratification and therapeutic interventions.