Mutational and Transcriptomic Landscape of AML with Core-Binding Factor Rearrangements

Background Acute myeloid leukemias (AML) with t(8;21)(q22;q22);RUNX1-RUNX1T1 and inv(16)(p13.1q22) or t(16;16)(p13.1;q22);CBFB-MYH11 are recurrent genetic entities commonly designated as core binding factor (CBF) AML. Both subgroups have distinct gene expression signatures and are characterized by recurrent mutations in KIT, FLT3, and RAS pathway genes. More recently, ASXL1 and ASXL2 mutations have been identified in t(8;21) AML. The TCGA study has analyzed a limited number of AML with CBF rearrangements, but to date more comprehensive CBF AML cohorts have not been unbiasedly explored by next-generation sequencing. Therefore, we have performed RNA sequencing of 415 genetically diverse AML specimens, including 48 CBF AML samples. In this analysis, we compared the mutational profile and transcriptomic landscape of both CBF subgroups to that of non-CBF AML.

Methods Analysis of mutations and gene expression was performed as previously described (Lavallée et al, Nature Genetics, doi:10.1038/ng.3371). Mutations in all genes that are recurrently mutated in hematological malignancies are reported. In order to identify novel acquired recurrent mutations in CBF subgroups, genes with variants in ≥ 3 samples were systematically confirmed by Sanger sequencing of non-tumoral DNA.

Results Genes mutated in the t(8;21) cohort are: KIT (8/20, 40%), FLT3 and ASXL2 (4/20 each, 20%), ASXL1, NRAS,  ZBTB7A, TET2, SMC1A (3/20 each, 15%), DNMT3A (2/20), and JAK2, SMC3, STAG2, WT1, and CSF3R (1/20 each).  Mutations in inv(16) AML are found in the following genes: KIT (14/28, 50%), NRAS (12/28, 43%), FLT3 (8/28, 29%), PRRC2B (3/28, 11%), KRAS (2/28, 7%) and BCORL1, DNMT3A,  GATA2 and NF1 (1/28). The most frequent mutations were found in activated signaling genes (KIT, NRAS, KRAS, FLT3, JAK2, CSF3R), identified in 14/20 (70%) and 25/28 (89%) of t(8;21) and inv(16) AML samples respectively. 38% of mutated samples contained 2 to 5 such mutations, and the sum of their variant allele frequencies never exceeded ~50%, suggesting that each mutation occurs in a different subclone. This result supports the hypothesis that these mutations and CBF fusion genes are strong collaborators in AML. Several novel observations emerged from these analyses. First, we identified 2 frameshift and 1 missense novel acquired mutations in ZBTB7A, which are specific to the t(8;21) subgroup (3/20 vs 1/395, p = 0.0004). ZBTB7A encodes a transcription factor of the POK/ZBTB family and other genes encoding this family of transcription factors, such as BCL6 and PLZF, are rearranged in hematological malignancies. Second, we established that ASXL2 mutations are very rare (2/395) in t(8;21) negative samples and thus specifically associated to RUNX1-RUNX1T1 fusions (p < 0.0001). Third, mutually exclusive mutations in cohesin complex genes (SMC1A, SMC3 and STAG2) are frequent in t(8;21) AML (5/20, 25%). Lastly, a novel acquired PRRC2B A1506S missense mutation was identified in 3 inv(16) AML samples. PRRC2B, a gene with poorly described functions, was the only non-activated signaling gene recurrently mutated in this subgroup.

Using the most significantly and differentially expressed genes, we identified signatures of 145 and 127 genes specific to t(8;21) and inv(16) groups, respectively. 78% and 81% of these genes have not been previously described in gene set enrichment analyses of CBF AML, and are potential novel CBF diagnostic markers or genes that are functionally related to CBF fusions. Using gene signatures and principal component analyses (PCA), CBF subgroups homogeneously clustered together with one sole exception: a sample harboring a t(16;21);RUNX1-CBFA2T3 unambiguously grouped with t(8;21) specimens. The rare but recurrent RUNX1-CBFA2T3 chimeric proteins are known to share similar structural characteristics with RUNX1-RUNX1T1, and our observations now unify the transcriptomic networks of these 2 genetic entities. We also identified and characterized 8 additional RUNX1 fusions in our AML cohort, including 6 novel fusions, which share a different transcriptomic profile compared to RUNX1-RUNX1T1 positive samples, thereby suggesting that they might have distinct functional consequences.

Conclusion Our comprehensive RNA sequencing analysis substantially contributes to a better understanding of mutations and gene expression profiles in CBF AML, and reports a unity between RUNX1-RUNX1T1 and RUNX1-CBFA2T3 genetic networks.