SRSF2 mutations in CCUS, MDS and AML: a cross-entity study Open Access

TO THE EDITOR:

The most frequent spliceosome mutations in myeloid neoplasms (MN) occur in SF3B1, SRSF2, ZRSR2, and U2AF1.¹ Although SF3B1 mutations in myelodysplastic neoplasms (MDS) are defining for a specific World Health Organization entity, all 4 spliceosome genes are defining for myelodysplasia-related acute myeloid leukemia (AML-MR).² Although SRSF2 mutations are known to be associated with poor prognosis and higher-risk forms of MDS and AML,³ they can already be detected in myeloid precursor lesions like clonal cytopenia of undetermined significance (CCUS).^4,SRSF2 mutations are also regarded as high-risk mutations within myeloproliferative neoplasms^5,6 and are frequently detected in chronic myelomonocytic leukemia, where they typically co-occur with TET2 mutations.^7,8 Although SF3B1 mutations in the context of MN are very well characterized,^9,10 comprehensive studies analyzing the role of the also frequently-mutated splicing factor SRSF2 in MN especially across the continuum from CCUS to MDS to AML are missing so far.

Here, we present a cross-entity study comparing comutations, chromosomal aberrations, and survival in patients with SRSF2-mutated CCUS, MDS, and AML. The cohort comprised of 1664 patients (median age, 72 years; 42% female; 222 CCUS, 698 MDS, 744 AML; supplemental Table 1). Diagnoses were established following WHO-HAEM5: World Health Organization Hematology (Haematology) 5th Edition² (supplemental Methods; supplemental Table 1; supplemental Figure 1). Samples were analyzed by cytomorphology, immunophenotyping, chromosome banding analysis, and whole genome sequencing evaluating mutations in 56 genes (supplemental Methods). To evaluate clonal hierarchy variant allele frequency (VAF) estimates were considered using a VAF difference cutoff of ≥5%.¹¹ 

SRSF2 mutations were detected in 11% (25/222) of CCUS, in 15% (102/698) of MDS, and in 15% (110/744) of patients with AML (supplemental Figure 1; supplemental Table 2), in line with other studies.^3,12 No SRSF2 mutations were found in MDS-5q and MDS-biTP53 (Figure 1A), whereas the majority (76%) of SRSF2-mutated MDS presented with increased blasts (<5% vs ≥5% blasts, 7% vs 26%; P < .001) indicating a strong association with advanced disease stage as shown previously.³ In AML, SRSF2 mutations were mainly observed in AML-MR (36%; 73/131), but rarely in AML with defining genetic alterations (8%; 38/504; Figure 1B).

Regarding clinical parameters, SRSF2-mutated patients mainly (48%-64%) showed bicytopenia across all entities and often pancytopenia in MDS and AML (supplemental Figure 2). In CCUS SRSF2 mutations were significantly associated with bicytopenia (64%; supplemental Figure 3). SRSF2-mutated MDS more often had bi- and pancytopenia than SRSF2 wildtype MDS reflecting advanced disease stages (supplemental Figure 4). In MDS, anemia plus thrombocytopenia were significantly associated with SRSF2 mutations, also shown previously.¹³ 

As SRSF2 mutations were reported to be associated with monocytic phenotypes in MN,¹⁴ we analyzed the percentage of bone marrow monocytes. Although there were no differences between SRSF2-mutated and wildtype patients in CCUS and AML, SRSF2-mutated MDS showed significantly higher monocyte levels compared to wildtype patients (median, 3.5% vs 1.5%; P < .001; supplemental Figure 5). Notably, SRSF2/TET2 comutated MDS showed a significantly higher fraction of monocytes (supplemental Figure 5), in line with another study.¹⁵ A similar trend was observed in peripheral blood of CCUS patients (supplemental Figure 5D). In line with the recently published molecular taxonomy of MDS¹⁶ SRSF2 mutations were significantly associated with biallelic TET2 mutations and monocytosis, and STAG2 comutations with neutropenia (supplemental Results).

Immunophenotypic analysis showed a tendency toward a higher proportion of myeloid progenitor cells with an increase in aberrations in SRSF2-mutated cases (mainly reduced CD45 expression), whereas cytomorphologic analysis showed a tendency toward reduced erythropoiesis with fewer dysplastic features in SRSF2-mutated cases (supplemental Results).

Of all 237 SRSF2 mutations, 232 (98%) were located in the hotspot region Pro95. The median VAF was similar across all entities (CCUS, 44%; MDS, 47%; AML, 47%; supplemental Figure 6), indicating that SRSF2 mutations are early events in leukemogenesis. Biallelic SRSF2 alterations were detected by whole genome sequencing only in 1 SRSF2-mutated patient (copy-neutral loss of heterozygosity). SRSF2-mutated cases showed a significantly higher number of mutations compared to wildtype cases across all entities (supplemental Figure 7). The number of comutations in SRSF2-mutated cases increased from CCUS to MDS and AML (median, 1, 2.5, and 3, respectively; supplemental Figure 7). Across all SRSF2-mutated cases ASXL1, TET2, RUNX1, and IDH2 were most frequently comutated (Figure 2A; supplemental Figure 8) as shown previously.^3,15,17 The most frequent comutations in CCUS were TET2 (44%) and ASXL1 (28%), in MDS TET2 (46%) and ASXL1 (41%), and in AML ASXL1 (44%) and RUNX1 (36%). Within AML-MR SRSF2 mutations were mutually exclusive with mutations in MR-defining genes SF3B1 and EZH2. Overall, comutations in other splicing genes or TP53 (supplemental Results; supplemental Table 3) were rare across all entities similar to previous studies.^3,15,17 SRSF2 mutation as the sole mutation was rare (6%; 5 CCUS, 6 MDS, and 3 AML; normal karyotype in 4 CCUS and 3 MDS), indicating that SRSF2 mutations require a second hit to result in overt MN as also suggested by others.¹⁵ In the majority of cases (CCUS, 80%; MDS, 75%; AML, 65%) SRSF2 mutations were present in the primary clone (supplemental Figure 9; supplemental Table 4). In cases where SRSF2 mutations were subclonal, the most commonly mutated gene in the primary clone was TET2 in CCUS and ASXL1 in MDS and AML (supplemental Figure 9). Cytogenetic abnormalities in SRSF2-mutated cases were rare in CCUS (12%) and MDS (25%), but frequent in AML (56%) (supplemental Results; supplemental Figure 10). SRSF2-mutated MDS and AML often showed a normal karyotype or trisomies, mostly trisomy 8, whereas complex karyotype was rare (supplemental Results).

SRSF2 mutations were associated with inferior overall survival (OS) in MDS (2.8 vs 5.6 years; P < .001, Figure 2B; multivariate analysis: hazard ratio, 1.41; P < .001; supplemental Table 5) in line with Bernard et al¹⁸ including SRSF2 as independent prognostic factor in MDS risk stratification. There was no prognostic relevance of SRSF2 mutations within AML-MR (supplemental Table 6) presumably attributed to the generally poor prognosis of AML-MR. In SRSF2-mutated MDS the number of comutations (≥3) was an independent negative prognostic factor (Figure 2C; supplemental Table 7). Within AML-MR ≥2 MR mutations did not affect OS in contrast to Tazi et al¹⁹ (presumably due to differences in cohort sizes), however, SRSF2/ASXL1–mutated AML-MR showed a slightly shorter OS (not significant) than other MR comutations (supplemental Figure 11). Due to the low number of SRSF2-mutated CCUS cases no prognostic markers could be identified in this subgroup.

Taken together, our data showed that SRSF2 mutations are already present with high VAFs at early stages in the development of MN, suggesting SRSF2 as strong driver. Accordingly, the most frequent comutations in SRSF2-mutated patients ASXL1, RUNX1, TET2, and IDH2 showed similar frequencies across disease stages, with SRSF2 mutations mainly in the primary clone. An SRSF2 mutation was an independent prognostic factor for inferior OS in MDS. The overall poor prognosis implies close monitoring of SRSF2-mutated patients even before the onset of overt disease. Limitations of our study include the retrospective design and the short follow-up data of CCUS patients. Thus, prospective studies are needed to validate the crucial role of SRSF2 mutations during disease progression and to identify patients at risk as early as possible.

Acknowledgments: The authors thank all coworkers at the MLL Munich Leukemia Laboratory for their dedicated work. The authors also thank all physicians for providing samples and caring for patients as well as collecting data.

Contribution: I.S. and C.H. designed the study and interpreted the data, and were responsible for chromosome banding and fluorescence in situ hybridization analyses; S.O. and S. Huber analyzed the data; S. Huber wrote the manuscript; M.M., G.H., and S. Hutter were responsible for the molecular and bioinformatic analyses; V.E. was responsible for the immunophenotyping; T.H. was responsible for the cytomorphologic analyses; and all authors read and contributed to the final version of the manuscript.

Conflict-of-interest disclosure: C.H. and T.H. declare part ownership of MLL Munich Leukemia Laboratory. S. Huber, S.O., S. Hutter, V.E., M.M., G.H., and I.S. are employed by the MLL Munich Leukemia Laboratory.

Correspondence: Isolde Summerer, MLL Munich Leukemia Laboratory, Max-Lebsche-Platz 31, 81377 Munich, Germany; email: isolde.summerer@mll.com.