Genomic Landscape of Pediatric CBF-AML By SNP-Array Karyotyping and Extensive Mutational Analysis

Background. Core binding factor (CBF) acute myeloid leukemia (AML) includes AML with t(8;21) and inv(16) leading to RUNX1-RUNX1T1 or CBFB-MYH11 fusion genes. These recurrent genetic abnormalities are both associated with disruption of genes encoding subunits of the CBF, a heterodimeric transcription factor involved in hematopoiesis. Although the fusion proteins appear to be crucial for the leukemogenic process, considerable experimental evidence indicates that they are not sufficient to induce AML on their own. Due to their high sensitivity to chemotherapy with high complete remission rates and their relatively favorable outcome, CBF-AML is considered to have a good prognosis. Nonetheless, about 30-40% of these patients relapse after standard intensive chemotherapy. In this context, identification of additional genetic or molecular abnormalities could allow better understanding of CBF-AML leukemogenesis, prediction of clinical outcome and identification of novel therapeutic targets.

Methods. This study focuses on 73 patients with CBF-AML [43 t(8;21) and 30 inv(16)-AML] enrolled in the pediatric trial ELAM02. Single nucleotide polymorphism array (SNP-A) was performed for all patients using Cytoscan® HD arrays according to the manufacturer instructions. In order to distinguish somatic from constitutional SNP-A lesions, we excluded known copy number abnormalities (CNA) if there was >50% overlap with variants from public database, except for breakpoints-related alterations. Interstitial uniparental disomies (UPD) <10 Mb and telomeric UPD <5 Mb were considered as constitutional and excluded from the analysis. Additionally, extensive mutational analysis (including 45 genes frequently reported to be mutated in myeloid malignancies) were performed for 37 and 25 patients with t(8;21) and inv(16)-AML respectively. Two different technologies of next generation sequencing (NGS) were used, allowing direct validation.

Results. Among the 73 cases, 145 SNP-A lesions were found in 58 patients (81%) with a median of 2 lesions per case (range, 0-8). CNA was more frequent (84 losses, 47 gains) than UPD (n=14). No significant difference was noted between the number of CNA and UPD in inv(16) and t(8;21)-AML. Small lesions were common at breakpoints involved in the t(8;21) and inv(16) (respectively 4/43 and 6/30). Additional recurrent CNA mostly involved entire chromosomes, chromosomal arms or large chromosomal regions. Del(9q) and loss of sex chromosome were restricted to t(8;21)-AML (respectively 6/43 and 20/43). Trisomy 22 was restricted to inv(16)-AML (2/30). Other recurrent CNA included trisomy 8 (3/43 vs 1/30) and gains of 13q (2/43 vs 1/30) in both subtypes, gains of 1q and del(2q) in t(8;21)-AML (each 2/43). Del(7q) was among the most common aberrations regardless of subtype (7/43 and 7/30). The minimally deleted region of 7q contained 57 genes including MLL3 and EZH2. Additionally, we found focal deletions of IKZF1 in one patient, NF1 in another and 3 deletions of CCDC26. Except for known mutations (KIT, RAS, FLT3), NGS did not reveal any other alterations in inv(16)-AML. By contrast, t(8;21)-AML was marked by the frequency of mutations in ASXL1/2 (8%/24%) and cohesin genes SMC1A, SMC3, RAD21, STAG2, NIPBL (27% combined). Mutations were also detected in epigenetic-related genes EZH2 (5%), TET2 (8%), IDH1/2 (5%) and WT1(11%).

Conclusions. SNP-A karyotyping of 73 pediatric CBF-AML revealed several recurrent alterations, with differing distribution between the 2 subgroups. Moreover, t(8;21) and inv(16)-AML appeared to have distinct mutational profiles, leading us to consider them separately for future studies. We recently reported high frequency of ASXL mutations in t(8;21)-AML and their absence in inv(16)-AML (Micol, Duployez and Boissel et al, Blood 2014). We now report high frequency of mutations in cohesin genes with the same distribution. Recent description of functional relations between cohesin and polycomb proteins, together with our results, suggest an important pathway in t(8;21) leukemogenesis. Concurrent ASXL and cohesin mutations were found in several patients, suggesting they could cooperate in some cases. Interestingly, ASXL mutations were exclusive of del(7q), suggesting that disruption of the ASXL-associated proteins MLL3 and EZH2 could be of great interest in the physiopathology of t(8;21)-AML. Finally, correlations with clinical outcome are in progress.