SRSF2 Mutations Identify a Distinct Subtype of Myeloid Neoplasm Across Myelodysplastic Syndromes and Acute Myeloid Leukemia

Abstract

Splicing factors are the most frequently mutated genes in myeloid neoplasms with myelodysplasia, including myelodysplastic syndromes (MDS), myelodysplastic/myeloproliferative neoplasms (MDS/MPN) and acute myeloid leukemia with myelodysplasia-related changes (AML-MDS). SF3B1 mutations was shown to be strongly associated with disease phenotype with ring sideroblasts, and SF3B1-mutant MDS has been proposed as a distinct disease subtype (Malcovati et al. Blood 2015). Conversely, SRSF2 has been found to be mutated in chronic myelomonocytic leukemia (CMML), and to be associated with multilineage dysplasia and excess blasts in MDS, as well as with clinico-pathological features of secondary AML (Linsdley et al. Blood 2015). In this study, we carried out a comprehensive mutation analysis in a well-annotated prospective cohort of 551 patients with myeloid neoplasms with myelodysplasia, and focused on clinical and molecular features of SRSF2-mutated patients with the aim to characterize patterns of co-occurring mutations and to identify genotype-phenotype correlations.A core panel of 54 genes recurrently mutated in myeloid neoplasms was analyzed using an Illumina HiSeq (Illumina Inc., CA, USA). Multivariable survival analyses were performed by means of Cox proportional hazards regression accounting for left censoring of the observation at the time of mutation assessment. Unsupervised hierarchical clustering analyses were performed in order to identify homogeneous groups of patients with respect to the variables of interest.Overall, 102 patients carrying SRSF2-mutation were identified, including 37 MDS (2 MDS with single lineage dysplasia, 9 with multilineage dysplasia, 6 with ring sideroblasts, 20 with excess blasts), 40 MDS/MPN (35 CMML, 5 MDS/MPN unclassified) and 25 AML-MDS. The median number of co-mutated genes was 2 (range 0 to 5); SRSF2 mutation was the unique mutation detected in 11 cases (5 AML-MDS, 4 MDS/MPN and 3 MDS). The most frequently co-mutated genes were TET2 (31/102 cases, 30%), ASXL1 (28%), IDH1/IDH2 (24%), RUNX1 (21%), STAG2 (14%) and CBL (10%). A significant association was found between co-mutation of TET2 and SRSF2 and myeloid neoplasm with monocytosis (P<.001), whereas STAG2 and IDH1/IDH2 mutations were significantly associated with MDS/AML-MDS phenotype (P=.044 and P=.001, respectively). We applied stepwise logistic regression, and confirmed that co-mutation of TET2 was a highly specific independent predictor of MDS/MPN phenotype within SRSF2-mutated myeloid neoplasms (OR=9.13, P<.001, positive predictive value: 0.85; specificity: 0.89).Then, we carried out uni- and multivariable Cox-regression analyses, and found that SRSF2-mutant MDS/MPN showed a significantly better prognosis compared to MDS (HR=.44, P=.008) and AML-MDS (HR=.47, P<.001). Conversely, having 20% or more bone marrow blasts or exceeding this threshold during follow-up did not affect survival of patients with SRSF2-mutated MDS or AML-MDS (HR=1.56, P=.15), suggesting that the proportion of bone marrow blasts does not recognize distinct subsets within SRSF2-mutated MDS and AML-MDS.Finally, we performed unsupervised hierarchical clustering analyses including current WHO classification criteria and gene mutations. This analysis identified one cluster including patients with MDS/MPN according to current WHO criteria, and a second cluster comprising patients with MDS or AML-MDS, in which a distinct subgroup including SRSF2-mutated patients was identified, irrespective of WHO bone marrow blast cut-offs.In conclusion, these results suggest that SRSF2 mutation is able to identify a distinct subset of myeloid neoplasms at the boundaries of MDS and AML, irrespective of current WHO classification criteria. Conversely, SRSF2-mutant MDS/MPN show different co-mutation patterns and clinical outcomes, supporting a classification in a separate category according to current WHO criteria. The recognition of a distinct entity with a unique molecular basis may significantly improve the diagnosis, classification and treatment of patients with these disorders. DisclosuresNo relevant conflicts of interest to declare.