ASXL1 and ASXL2 Mutations in Childhood AML Are Strongly Associated with t(8;21) but Do Not Independently Impact on Prognosis: A Report from the Children's Oncology Group and NCI/COG Target Initiative

Increasing use of high-throughput sequencing technologies has enabled the identification of novel somatic gene alterations in cancers, with important implications in understanding the mechanisms of transformation and potentially in improving therapeutic decision-making. In contrast to many solid tumors, the mutational spectrum of acute myeloid leukemia (AML) has proven bland, with relatively few recurrent somatic mutations per tumor. This result appears more pronounced in pediatric AML, where even mutation of genes commonly mutated in adult disease, such as IDH1/2 or DNMT3A, appear at much lower frequencies. Alterations of ASXL1, a regulator of the polycomb-repressive complex, have been identified in adults with myelodysplastic syndromes, myeloproliferative syndromes, and AML. Mutations in the related gene family member, ASXL2, were recently associated with t(8;21) AML in adults and children. In this study, we evaluated the prevalence and prognostic impact of mutations of these epigenetic regulator genes in pediatric AML patients treated on COG studies AAML03P1 and AAML0531.

ASXL1 and ASXL2 (ASXL1/2) were sequenced from genomic DNA isolated by directed capture of the coding sequences of each gene followed by Illumina paired-end sequencing in a total of 769 diagnostic AML specimens. We identified ASXL1/2 mutations in 44/769 pediatric AML cases (5.7%), with ASXL2 altered more frequently (61% of mutated cases) than ASXL1. Consistent with prior reports, the majority of mutations (80%) were nonsense changes involving exons 11-12 of either gene. ASXL1 and ASXL2 mutations were mutually exclusive, with no case demonstrating concurrent mutation of both genes. 

In comparison of the clinical covariates of cases with ASXL1/2 mutation against those without, there were no differences in age, gender, or ethnicity; however we found an increased proportion of ASXL1/2 mutations in patients of Asian race, constituting 15% of those with mutated ASXL1/2 vs. 5% of those without (p=0.018). Patients possessing ASXL1/2 mutations tended to present with lower peripheral WBC and marrow blast counts (median 20K vs 30.8K, p=0.067 and 49% vs 70%, p=0.003, respectively).

In agreement with recent reports, we identified a striking positive association of ASXL1/2 alterations and t(8;21), which was present in 68% of cases with ASXL1/2 vs 12% of those without (p<0.001). More than 25% (30/113) of patients with t(8;21) had a concomitant ASXL1/2 mutation. We similarly found an inverse association with other major cytogenetic subgroups—inv(16): 2% vs 14%, p=0.024; normal cytogenetics: 5% vs 25%, p<0.001; MLL-rearranged: 2% vs 21%, p=0.003. In the evaluation of concurrent molecular changes, ASXL1/2 mutations were inversely associated with NPM1 mutation, with no ASXL1/2 mutated case having a concurrent NPM1 mutation compared to a 9% NPM1 mutation rate in those without alterations of ASXL1/2 (p=0.041). There were no significant associations with FLT3-ITD, CEBPA or KIT mutation status.

We found no differences in the clinical outcomes of patients with ASXL1/2 mutations in comparison to those without. Complete remission rates were similar for both groups at the end of induction courses 1 and 2 (84 vs 75%, p=0.2 and 91 vs 89%, p=1.0, respectively). 5-year EFS and OS was 59 ± 15% vs 50 ± 4% (p=0.305) and 65 ± 15% vs. 66 ± 4% (p=0.851). Although prior reports have suggested the possibility of enhanced relapse risk with ASXL1/2 mutations, the 5-year RR for pediatric patients treated in these trials was similar at 31 ± 16% vs 40 ± 4% (p=0.407). Given the co-occurrence of ASXL1/2 mutations and t(8;21), we also evaluated outcomes of t(8;21) patients (n=113) with or without ASXL1/2 mutations, however no significant differences in 5-year OS, EFS, or RR were observed in this subset analysis.

This work demonstrate that mutations of ASXL1 or ASXL2 are highly prevalent in childhood t(8;21) AML. Although there was no effect on treatment outcomes, this concomitance suggests the possibility of discrete cooperating lesions, even within the group of CBF-translocated AML, which may be amenable to specifically-directed therapy