SF3B1-mutant CMML defines a predominantly dysplastic CMML subtype with a superior acute leukemia–free survival

TO THE EDITOR:

Malcovati et al¹  recently proposed SF3B1-mutant (SF3B1^MT) myelodysplastic syndrome (MDS) to be a distinct nosologic entity defined by cytopenias, somatic SF3B1 mutation, morphologic dysplasia (with or without ring sideroblasts [RSs]), bone marrow (BM) blasts <5%, and peripheral blood (PB) blasts <1%. Select concomitant genetic exclusion criteria are del(5q), monosomy7, inv(3)/abnormal 3q26, complex karyotype, or mutations involving RUNX1 and/or EZH2. Chronic myelomonocytic leukemia (CMML) is an MDS/myeloproliferative neoplasm (MPN) overlap syndrome characterized by sustained PB monocytosis (≥1 × 10⁹/L, ≥10% of white blood cell count [WBC]) further categorized into proliferative (WBC ≥13 × 10⁹/L) and dysplastic (WBC <13 × 10⁹/L) subtypes on the basis of the presenting WBC count.²  Although mutations involving pre-messenger RNA splicing are seen in ∼70% of CMML patients, these mostly involve SRSF2 (50%), with SF3B1 mutations accounting for <10% with no clear prognostic impact.³  We performed this study to compare and contrast SF3B1^MT CMML with SF3B1-wild-type (SF3B1^WT) CMML and SF3B1^MT MDS-RS, focusing on prevalence, phenotypic correlates, epistasis with other mutations or splicing mutations, and survival outcomes.

After approval by the Mayo Clinic (Rochester, MN) and the Moffitt Cancer Center (Tampa, FL) institutional review boards, adult patients with World Health Organization (WHO)–defined CMML and MDS-RS (Mayo Clinic only) were included in the study.²  BM morphology, percentage of RSs, cytogenetics, and 2016 WHO diagnoses were retrospectively reviewed, and all patients underwent targeted next-generation sequencing for myeloid-relevant genes, which were obtained on BM mononuclear cells at diagnosis by previously described methods.⁴  CMML prognostication was performed by the Mayo Molecular Model and the Groupe Francophone des Myelodyplasies (GFM) model, and MDS-RS prognostication was performed by using the Revised International Prognostic Scoring System (IPSS-R).^5-7  The Clinical/Molecular CMML-Specific Prognostic Scoring System (CPSS-Mol) scores could not be calculated because of incomplete data with regard to transfusion dependency. Statistical analyses considered clinical and laboratory parameters obtained at the time of diagnosis or first referral (usually within 6 months of diagnosis). Median time from diagnosis to molecular testing in the Mayo Clinic cohort was 4 months (range, 0-7 months). The Mann-Whitney U test and Fisher’s exact test were used to compare quantitative and qualitative data in subgroups. Kaplan-Meier overall survival (OS) and AML-free survival (AML-FS) estimates and Cox regression models were used for survival analysis. The Benjamini-Hochberg controlling method was used for multiple hypothesis testing.

In all, 819 patients with CMML (40 [5%] with SF3B1^MT]) and 83 with SF3B1^MT MDS-RS were included in the study (Table 1; supplemental Table 1). Patients with SF3B1^MT CMML compared with their WT counterparts were more likely to have lower hemoglobin levels (P = .01), lower WBC counts (P = .009), lower absolute monocyte counts (AMC; P = .007), and higher platelet counts (P = .006). They were also more likely to have BM RSs (P < .001) and JAK2^V617F mutations (P = .03) (Figure 1A) and less likely to have ASXL1 (P = .03) and SRSF2 (P = .02) mutations. Thirty SF3B1^MT CMML patients (75%) were classified as having a dysplastic CMML subtype compared with 399 SF3B1^WT patients (49%) (P = .04). Although there were fewer high-risk karyotypic abnormalities in SF3B1^MT CMML patients, there was no difference between the 2 groups with regard to WHO CMML subtypes. In the Mayo Clinic cohort, 68 SF3B1^WT CMML patients (14%) had BM RSs (>5%). Compared with SF3B1^MT MDS-RS patients, SF3B1^MT CMML patients were more likely to have higher WBC (P = .007), higher AMC (P = .005), lower platelet counts (P = .03), higher PB blast percent (P = .009), and higher BM blast percent (P = .004) and were more likely to have RUNX1 (P = .007) and JAK2^V617F (P = .01) mutations (Figure 1B).

The median age for SF3B1^MT CMML patients was 74.5 years, and 62% were males. The median SF3B1^MT variant allele frequency (VAF) burden in SF3B1^MT CMML was 40.5% compared with 41% for SF3B1^MT MDS-RS (Figure 1C). Mutant SF3B1-associated amino acid changes included 23 (48%) K700E, 6 (12.5%) H662Q, 3 (6.2%) K666N, 2 (4%) G740E, 2 (4%) K554E, and 1 each (2%) for K669N, K741E, E622D, Y623C, Y768N, E283K, H516Y, Y623C, Y625H/L, N626S, and Y765N (3 patients had 2 SF3B1 mutations) (Figure 1D). Nine SF3B1^MT CMML patients (22.5%) had >1 splicing mutation (supplemental Table 2): 5 SRSF2^P95H/L, 3 SF3B1, and 1 ZRSR2. The VAFs of concurrent splicing mutations were similar (codominant) in 2 patients (SF3B1^K700E 42%/SRSF2^P95H 45% and SF3B1^G669Q 43%/SRSF2^P95L 42%). In 5 patients, the SF3B1 mutations were dominant, and the second splicing mutation represented subclones (VAFs not available in 2 patients). In 3 SF3B1^MT CMML patients, no other driver mutations were identified. Common concurrent mutations included 17 (42.5%) TET2, 8 (20%) RUNX1, 7 (17.5%) DNMT3A, 7 (17.5%) JAK2^V617F, and 5 (12.5%) SRSF2 (supplemental Figure 1); 10 patients (25%) had clonal cytogenetic abnormalities. U2AF1 mutations were not seen in either SF3B1^MT CMML or in SF3B1^MT MDS-RS patients. Seven SF3B1^MT CMML patients had concurrent JAK2^V617F mutations, with 4 (57%) having a proliferative CMML subtype. None of these patients had ASXL1 or oncogenic RAS pathway mutations, and the median VAF for JAK2^V617F was 20% (range, 5%-52%). Within the limitations of a smaller data set, JAK2^V617F and SF3B1 co-mutated CMML patients had a trend toward higher WBC and higher platelet counts without any significant differences in OS and AML-FS compared with SF3B1^MT CMML patients. In SF3B1^MT MDS-RS, frequencies of common hotspot amino acid changes included K700E, 50.6%; H662Q, 10.8%; E662D, 9.6%; and K666N, 4.8% (Figure 1D), with common concurrent mutations being TET2 (25%), DNMT3A (20%), and ASXL1 (13%).

At the last follow-up, among SF3B1^MT CMML, SF3B1^WT CMML, and SF3B1^MT MDS-RS patients, 18 (45%), 461 (56%), and 56 (68%) deaths and 6 (14%), 159 (18%), and 2 (2.4%) leukemic transformations were documented, respectively. The median OS for SF3B1^MT CMML patients was 57.3 months (95% confidence interval [CI], 35.8 months to not reached) compared with 32.9 months for SF3B1^WT CMML patients (95% CI, 30-37.3 months; P = .059). SF3B1^MT MDS-RS patients had a trend toward a better median OS (median OS, 91.7 months; 95% CI, 69.8-115.6 months; P = .058) compared with SF3B1^MT CMML patients and a clear survival advantage over SF3B1^WT CMML patients (P < .001; Figure 1E). The median AML-FS for SF3B1^MT CMML patients was 42.3 months (95% CI, 30.5 months to not reached) compared with 23.5 months (95% CI, 20.8-26.4 months; P = .012) for SF3B1^WT CMML patients Figure 1E). SF3B1^MT MDS-RS patients had a better AML-FS than CMML patients in both categories (91.7 months; 95% CI, 69.8-115.6 months; P = .012 for both categories).

By using this large molecularly annotated cohort of CMML and SF3B1^MT MDS-RS patients, we defined SF3B1^MT CMML to be an infrequent (5%) but unique CMML category predominantly composed of dysplastic CMML subtypes, with similarities to MDS-RS in the form of median SF3B1 VAFs, frequencies of SF3B1 mutational hotspots (K700E was the most common) and IPSS-R risk stratification. As seen in MDS-RS, SF3B1 mutations were also found to occur alone in SF3B1^MT CMML and seemed to be secondary to other oncogenic mutations in a minority of cases. Although there was a higher frequency of JAK2^V617F and RUNX1 mutations in SF3B1^MT CMML compared with SF3B1^MT MDS-RS, these mutations are not enough to fully explain biological and phenotypic differences between the 2 entities. In addition, although there was a numeric trend toward a higher frequency of TET2 and SRSF2 mutations (known to skew hematopoiesis toward monocytosis) in SF3B1^MT CMML, this did not reach statistical significance.

SF3B1^MT CMML patients had a significant AML-FS advantage compared with their WT patient counterparts, with a trend toward a better OS. SF3B1^MT MDS-RS patients had a better OS compared with SF3B1^WT CMML patients and had a clear AML-FS advantage compared with both CMML subtypes. Conversely, SF3B1^MT CMML demonstrated key differences from SF3B1^WT CMML and from CMML in general by having a lower-than-expected frequency of ASXL1 mutations and a higher-than-expected frequency of JAK2^V61F mutations and by demonstrating frequent concurrent splicing mutations. Truncating ASXL1 mutations are seen in 40% of CMML patients and in 10% to 15% of SF3B1^MT MDS-RS patients, and they had a negative impact on survival.^1,6,7  Only 4 (10%) of 40 SF3B1^MT CMML patients in our cohort had an ASXL1 mutation, suggestive of less aggressive disease biology. The association of SF3B1^MT CMML with a higher-than-expected frequency of JAK2^V617F mutations remains to be elucidated but is reminiscent of MDS-RS, in which subsequent acquisition of JAK2^V617F confers proliferative features resulting in MDS/MPN-RS thrombocytosis.⁸  Mutational co-expression of JAK2^V617F and SF3B1 in CMML confers proliferative features to what is otherwise a predominantly dysplastic CMML subtype. Splicing mutations are frequent in myeloid neoplasms and, in general, are mutually exclusive secondary to their synthetic lethal interactions and convergent effects.⁹  We recently documented a prevalence of 0.85% for ≥2 splicing mutations in 4231 patients with myeloid neoplasms, with ∼50% being in the same cell; the distribution of double mutants deviated from the single mutants with selection against the most common alleles, SF3B1^K700E and SRSF2^P95H/L.¹⁰  In our study, 9 (22.5%) of 40 SF3B1^MT CMML patients had concurrent splicing mutations, and 5 patients had coexisting SF3B1 and SRSF2 mutations, with 3 (75%) demonstrating involvement of the most common alleles, SF3B1^K700E and SRSF2^P95H/L (K700E 23%/P95H 45%, K700E 32%/P95L 16%, and K700E 48%/P95H 2.8%). This unique epistatic interaction deserves further exploration.

In summary, we define SF3B1^MT CMML as a CMML subtype with predominant dysplastic features, with a low frequency of ASXL1 mutations, higher frequency of JAK2^V61F mutations, concurrent splicing mutations, and a superior AML-FS. Given its infrequent occurrence, unlike in SF3B1^MT MDS-RS because of the lack of a clear impact on OS, further validation is needed in a larger cohort of patients before this is recognized as an independent nosological entity.