Splicing factors are the most frequently mutated genes in myeloid neoplasms. SF3B1-mutated myelodysplastic syndromes (MDS) have been recently recognized as a distinct subtype strongly associated with ring sideroblasts. Conversely, SRSF2 has been found to be recurrently mutated in chronic myelomonocytic leukemia (CMML), and predicting worse prognosis in MDS, myelofibrosis (MF) and acute myeloid leukemia (AML).
Hence, SRSF2-mutated neoplasms constitute an heterogenous molecular-defined subgroup, whose clinical characteristics have not been systematically characterized. We aimed at a comprehensive characterization of clinical features and co-mutational determinants of SRSF2-mutated neoplasms across WHO classification boundaries.
We analyzed clinical and molecular characteristics of SRSF2-mutated myeloid neoplasms diagnosed at 4 referral centers. All clinical information was prospectively annotated in clinical chart records and a common panel of 23 recurrently mutated genes in myeloid neoplasms were analysed using Illumina HiSeq sequencing of peripheral blood cell DNA. Multivariate analyses (MVA) accounted for left censoring at mutation assessment time and right-censoring in case of disease-modifying treatments.
Overall, 279 patients carrying SRSF2-mutation were identified, including 84 MDS (49 blast excess, 21 multi-lineage dysplasia, 7 ring sideroblasts, 7 unclassified), 76 MPN (64 primary-MF [PMF], 5 secondary-MF, 3 polycytemia vera, 2 essential thrombocytemia, 2 chronic neutrophilic leukemia), 71 MDS/MPN (59 CMML, 12 MDS/MPN unclassified), 47 AML (38 with MDS-related changes, 9 NOS), 1 blastic plasmacytoid dendritic cell neoplasm. No difference in age was found between WHO subgroups. Male/female ratio was 3.3. Median overall survival (mOS) was 27.3 months (95%CI: 19.9-34.7). The median number of co-mutated genes was 3 (range 0-7). SRSF2 mutation was the unique driver mutation detected in 13 cases. The most frequently co-mutated genes were ASXL1 (38%), TET2 (35%), JAK2 (23%), RUNX1 (17%), IDH2 (15%), CBL (10%), STAG2 (10%), NRAS (8%). We used sequencing data to identify the order of acquisition of SRSF2 mutations in double-mutant patients and found that SRSF2 mutation was acquired first in 80 (28.7%) cases, sub-clonal in 111 (39.8%). Mutational order could not be assessed in 88 (31.5%) cases. In MVA, AML phenotype was independently predicted by RUNX1 (OR 4.2, p < .001), IDH2 (OR 4.3, p = .001), STAG2 mutation (OR 3.2, p = .02). Similarly, co-occurring mutation in JAK2 (OR 77.5, p < .001), MPL (OR 194, p < .001) and ASXL1 (OR 3.1, p = .01) predicted MPN phenotype. MDS/MPN was predicted by TET2 (OR 7.2, p < .001), SETBP1 (OR 4.3, p = .02) mutation and SRSF2-first mutation (OR 3.1, P = .004). Factors predicting OS were blast count (p < .001), white blood cells (p = .009), TET2 (HR .54, p < .001), IDH2 (HR .5, p = .021), NRAS (HR 2.23, P = .017) mutation. Notably, fibrosis, cytogenetics, WHO categories were not associated with different survival in MVA.
In conclusion, SRSF2-mutated myeloid neoplasms are a subset of myeloid neoplasms characterized by bone marrow dysplasia and overall dismal prognosis, whose fibrotic or proliferative phenotype results from a restricted co-mutation pattern. Among MPN, SRSF2 mutation is almost invariably associated with MF, further supporting the concept that PMF is an MDS/MPN disorder. Co-mutational pattern is the best predictor of clinical phenotype, suggesting that recognition of a distinct entity with a unique molecular basis may significantly improve classification and treatment of patients with these disorders.