Myelodysplastic syndrome (MDS) is a heterogeneous hematologic disease with greater variability in clinical features and treatment outcome. The pathogenesis of the development of MDS is not quite clear, but may be related to accumulated genetic aberrations . Recently, somatic mutations of genes encoding core components of the RNA splicing machinery, such as SF3B1, SRSF2, U2AF1, and ZRSR2, were identified in MDS patients [2, 3]. Splicing factor 3B subunit 1 (SF3B1), encoded by SF3B1, is a core component of U2 small nuclear ribonucleoprotein (U2 snRNP) and plays a role in premessenger RNA splicing and associated transcription . The incidence of somatic mutations of the SF3B1 gene is especially high in MDS with ring sideroblasts (RS): 64%–83% in refractory anemia with RS (RARS) and 33%–76% in refractory cytopenia with multilineage dysplasia and RS (RCMD-RS), a subtype in the 2001, but not 2008, WHO classification [2, 5-8].
The prognostic implication of SF3B1 mutations in MDS remains controversial. In some studies, SF3B1 mutation was identified as a better prognostic factor [5, 6], whereas in others, conflicting results were found [7-10]. In addition, to the best of our knowledge, the report concerning sequential analysis of SF3B1 mutations to evaluate their stability during MDS progression is lacking in literature. In this study, sequential analyses of SF3B1 mutations in 417 samples from 142 patients showed the original mutations of the SF3B1-mutated patients were retained while none of the patients without the mutation at diagnosis acquired a novel one during follow-ups.
A total of 479 adult patients diagnosed as having de novo MDS according to French–American–British (FAB) Cooperative Group Criteria who had cryopreserved bone marrow (BM) cells and complete clinical and laboratory data for analyses were enrolled. Among them, the disease of 372 patients fulfilled the criteria of MDS according to the 2008 WHO classification . The study was approved by the Institutional Review Board of the National Taiwan University Hospital and written informed consent in accordance with the Declaration of Helsinki was obtained from all participants. Sequential analyses of SF3B1 mutations were also performed in 417 samples from 142 patients, including 51 patients with leukemic transformation.
Mutation analyses of SF3B1 were performed on BM cells by polymerase chain reaction and direct sequencing as previously reported . Mutation analyses of other 16 relevant genes, including Class I mutations, such as FLT3/ITD , NRAS , KRAS , JAK2 , and PTPN11 mutations, Class II mutations, such as MLL/PTD , RUNX1, and WT1  mutations, mutations of genes involving in epigenetic modifications, such as ASXL1, IDH1 , IDH2 , including R140 and R172 mutations, DNMT3A , and EZH2  mutations, splicing machinery mutations, such as U2AF1 , SRSF2 , and SF3B1 mutations and SETBP1  mutations, were performed as previously described.
Bone marrow cells were collected directly or after 1–3 days of unstimulated culture and the metaphase chromosomes were banded by the G-banding as described earlier .
TA cloning analysis
For the patients with discrepancy of the mutation status of the SF3B1 in paired samples, TA cloning was performed in the samples without detectable mutant by direct sequencing. The DNA spanning the mutation spots of SF3B1 detected at either diagnosis or during subsequent follow-ups was amplified and the PCR products were then cloned into the Taq polymerase-amplified (TA)-cloning vector pGEM®-T Easy (Promega, Madison, WI) . Direct sequencing was then performed on the selected clones.
The discrete variables of patients with and without SF3B1 mutations were compared using the chi-square test. Fisher exact test was used if the expected values of contingency tables were smaller than 5. Mann–Whitney test method was used to compare continuous variables and medians of distributions. The OS was measured from the first diagnosis date to the date of last follow-up or death from any cause. Kaplan–Meier estimation was used to plot survival curves and log-rank tests were used to test the differences between groups. Cox proportional hazard regression analysis was used to investigate the independent prognostic factors for OS. A P-value <0.05 was considered statistically significant. We used the SPSS 17 software (SPSS Inc., Chicago, IL) and Statsdirect (Cheshire, England, UK) to perform all statistical analyses.
The SF3B1 mutations were detected in 48 (10%) of 479 patients studied (Supporting Information Table 1). The most common mutation was K700E (n = 33), followed by K666N (n = 5) and R625C (n = 2) (Fig. 1). All other mutations were detected in one person each. The SF3B1 mutations in all these patients were heterozygous. If only the 372 MDS patients based on the 2008 WHO classification were analyzed, the frequency of SF3B1 mutations was 10.8%.
Table 1. Comparison of Clinical Manifestation and Laboratory Features Between MDS Patients with and Without SF3B1 Mutation
Total (n = 479)
SF3B1 mutated (n = 48, 10.0%)
SF3B1 wild (n = 431, 90.0%)
CMML, chronic myelomonocytic leukemia; FAB, French-American-British classification; IPSS, international prognosis scoring system; IPSS-R, revised IPSS; MDS-U, unclassified; NA, not applicable; RA, refractory anemia; RARS, refractory anemia with ring sideroblasts; RAEB, refractory anemia with excess blasts; RAEBT, refractory anemia with excess blasts in transformation; RCUD, refractory cytopenia with unilineage dysplasia; RCMD, refractory cytopenia with multilineage dysplasia; WHO, World Health Organization.
Number of patients (% of patients with or without SF3B1 mutation in the subgroup).
International prognosis scoring system (IPSS): low, 0; intermediate (INT)-1, 0.5–1; INT-2, 1.5–2; and high, ≥ 2.5.
Revised IPSS: Very low, ≦1.5; Low, >1.5–3; intermediate (INT), >3–4.5; High, >4.5–6; and Vey high, >6.
Comparison of clinical characteristics between MDS patients with and without SF3B1 mutations is shown in Table 1. The SF3B1 mutations were closely associated with older age (median, 73 years vs. 65 years, P = 0.001), higher platelet count (median, 149.5 × 103/μL vs. 71 × 103/μL, P < 0.001), and lower lactate dehydrogenase (LDH) level (median, 418 U/L vs. 499 U/L, P = 0.038) at diagnosis. Patients with RARS by the FAB classification had the highest incidence of SF3B1 mutations (61.8%), while those with RA had the lowest incidence (4%). If only MDS patients based on the WHO classification were counted, those with RARS and RCMD-RS had the highest incidence of SF3B1 mutations (65% and 57.1%, respectively), whereas those with RCMD had the lowest incidence (2.1%). The SF3B1 mutations were also associated with lower risk MDS, including low and intermediate-1 (int-1) risk by International Prognostic Scoring System (IPSS) (14.4% vs. 3.3%, P = 0.001), or very low and low risk by revised IPSS (IPSS-R) (19.4% vs. 6.4%, P < 0.001).
Association of SF3B1 mutations with cytogenetic and other molecular genetic alterations
Chromosome data were available in 447 patients at diagnosis, including 44 with SF3B1 mutations and 403 without (Table 2). Clonal chromosomal abnormalities were detected in 194 patients (43.4%). Among the 44 SF3B1-mutated patients (Supporting Information Table 1), thirty had normal karyotype, two had trisomy 8, two had 20q deletion, two had isolated 5q deletion, and one had isolated loss of chromosome Y. SF3B1 mutations were closely associated with good-risk cytogenetics (12.8% vs. 5.1%, P = 0.008) and isolated 5q deletion (100% vs. 9.4%, P = 0.0095), but inversely related to poor-risk cytogenetics (2.4% vs. 11.6%, P = 0.01, Table 2).
Table 2. Comparison of Cytogenetic Changes Between MDS Patients with and Without SF3B1 Mutationa
Total (n = 447)
SF3B1 mutated (n = 44, 9.8%)
SF3B1 wild (n = 403, 90.2%)
Cytogenetic data were available in 447 patients among total cohort including 44 patients with SF3B1 mutation and 403 without the mutation.
Good, normal karyotype, isolated -Y, del(5q) or del(20q); poor, complex (≧ 3 abnormalities) or chromosome 7 anomalies; intermediate, other abnormalities.
As the sole abnormality.
Number of patients (% of patients with or without SF3B1 mutation in the subgroup).
To investigate the interaction of gene mutations in the pathogenesis of adult MDS, we also analyzed the mutation status of 16 other molecular genetic alterations in these MDS patients. Thirty (62.5%) of the 48 patients with SF3B1 mutations had concurrent cytogenetic abnormalities (n = 9) or other molecular gene mutations (n = 21), most commonly DNMT3A mutation (n = 13) followed by RUNX1 mutation (n = 4), at diagnosis (Supporting Information Table 1). Patients harboring SF3B1 mutations had a significantly higher incidence of DNMT3A mutation (27.1% vs. 7.9%, P < 0.001, Table 3), but lower incidence of ASXL1 mutation (8.3% vs. 25.2%, P = 0.009) and SRSF2 mutation (2.1% vs. 14.6%, P = 0.015) than those with wild-type SF3B1.
Table 3. Comparison of Other Genetic Alterations Between MDS Patients with and Without the SF3B1 Mutation
Percentage of patients with the other gene mutation
Sequential study of SF3B1 mutations during clinical follow-ups
SF3B1 mutations were studied sequentially in 417 samples from 142 patients, including 17 patients with SF3B1 mutations at diagnosis and 125 patients without the mutation. With the exception of two patients (patients 7 and 37) who lost the original SF3B1 mutation at remission status following allogeneic hematopoietic stem cell transplantation (allo-HSCT), all remaining 15 SF3B1-mutated patients retained the same mutations during follow-ups (Table 4). Three (patients 38, 39, and 47) of the nine patients who had disease evolution acquired a novel RUNX1 mutation and other two (patients 7 and 12) had cytogenetic evolution at disease progression.
Table 4. Sequential Studies in the 19 MDS Patients who had SF3B1 Mutations at Diagnosis and follow-upsa
AML, acute myeloid leukemia; CMML, chronic myelomonocytic leukemia; HSCT, hematopoietic stem cell transplantation; N, normal cytogenetics; ND, no data; RA, refractory anemia; RAEB, refractory anemia with excess blasts; RAEBT, refractory anemia with excess blasts in transformation; UPN, unique patient number.
The 123 patients without SF3B1 mutation at both diagnosis and during sequential follow-ups were not shown in this table. MDS entity with bone marrow blasts of 20–29% was subclassified as RAEBT according to the FAB classification and that with bone marrow blasts more than 30% as AML, MDS.
Interval between the two successive status.
The SF3B1 mutation in these two patients (patients 49 and 50) was detected by TA cloning, but not direct sequencing.
Following Azacitidine treatment the disease status became RA in patient 43.
In contrast, among the 125 patients who had no SF3B1 mutations at diagnosis, two (patients 49 and 50) acquired SF3B1 mutations during follow-ups. Because direct sequencing might not be sensitive enough to detect low level of SF3B1 mutant, we therefore did TA cloning of the samples obtained from these two patients at diagnosis. Truly, SF3B1 mutation could be detected in one of eight clones in the sample from patient 49 and in one of seven clones in the sample from patient 50. In other words, SF3B1 mutants already existed early at diagnosis in these two patients and expanded during disease evolution. Further, among 125 SF3B1-wild patients, 63 had disease progression during subsequent follow-ups and 14 of them acquired novel mutations of other genes: 7 acquired an NRAS mutation, 4, a RUNX1 mutation, and one each, KRAS, ASXL1, and JAK2 mutation, respectively.
Impact of SF3B1 mutations on clinical outcome
With a median follow up duration of 43.3 months (range 0.1–250.7 months), univariate analysis showed mutations of ASXL1, IDH2, RUNX1, NRAS, EZH2, and SRSF2 mutation, as well as older age, poor-risk cytogenetics and high IPSS score were associated with poorer OS (Table 5). There was no difference in OS between SF3B1-mutated and -wild patients with MDS defined either by the FAB or the 2008 WHO classifications (39.7 months vs. 28.7 months, P = 0.475 and 83.6 months vs. 35.7 months, P = 0.610, respectively; Fig. 2A,B). The SF3B1 mutations also had no influence on time to AML transformation (Fig. 2C,D). Subgroup analyses in patients with RARS or RCMD-RS, in whom SF3B1 mutation occurred most frequently, did not find any difference in survival and time to AML transformation between patients with and without SF3B1 mutations (Fig. 2E,F). Similarly, there was no prognostic impact of SF3B1 mutation in the MDS patients with lower IPSS or IPSS-R score.
Table 5. Univariate Analysis of Clinical Parameters and Molecular Alterations on Overall Survival
No. of patients
Overall survival months (median ± s.d.)
IPSS, international prognostic scoring system; NA, not applicable.
IPSS, Higher-risk group: intermediate 2 and high risk; Lower-risk group, low and intermediate 1 risk.
In this study, SF3B1 mutations were detected in 10% and 10.8% of MDS patients defined either by the FAB or the 2008 WHO classification, respectively. The majority of SF3B1-mutated patients had concurrently other genetic alterations, most commonly DNMT3A mutations, or chromosomal abnormalities at diagnosis. All SF3B1 mutations detected at diagnosis remained unchanged during clinical follow-ups unless allo-HSCT was performed, and none of the SF3B1-wild patients acquired such mutation at disease progression.
The incidence of SF3B1 mutations in our cohort was lower than that reported (20% by Papaemmanuil et al.  and 28% by Malcovati et al. ), probably due to the lower frequency of RARS among Asian patients with MDS, compared with Western population [19, 23, 25] However, the incidence of SF3B1 mutations in RARS or RCMD-RS was similar to those reported (64–83% in RARS, 33–76% in RCMD-RS) [2, 5-8]. The prognostic impact of SF3B1 mutations in MDS remains inconclusive . In some studies, the SF3B1 mutations were associated with better prognosis in total MDS cohort , and in subgroups of patients with MDS-RS, including RARS, RCMD-RS, and RAEB-RS . However, in other studies, SF3B1 mutation showed no prognostic implication in MDS patients [7, 8]. In our study, SF3B1 mutations did not have prognostic implication on both OS and leukemia transformation. Subgroup analyses did not reveal survival difference between the two groups with and without SF3B1 mutations, either. There are some possible explanations. First, the patients harbored the mutation were distinctly older (median, 73 years vs. 65 years, P = 0.001) and age had striking prognostic impact on survival in our study. Second, SF3B1 mutation was closely associated with DNMT3A mutation, which was associated with worse OS and more rapid progression to AML in MDS patients . Third, the prognostic effect may be skewed by different treatment modalities such as demethylating agents, intensive chemotherapy or allo-HSCT. Moreover, because of the heterogeneity of treatment in MDS patients and retrospective analyses in nature, the treatment could not be clarified clearly in most studies published. More studies are needed to clarify the clinical implication of SF3B1 mutations in MDS.
The reports concerning the association of SF3B1 mutation with other genetic alterations in the development of MDS are scanty. SF3B1 mutation was found to co-occur frequently with DNMT3A mutation in patients with lower risk IPSS . In this study, we noted SF3B1 mutation was closely correlated with DNMT3A mutation but negatively with SRSF2 and ASXL1 mutations. The concurrent mutations of SF3B1 and DNMT3A could be found not only in patients with lower risk MDS but also those with higher risk MDS (Supporting Information Table 1). The interaction of SF3B1 and DNMT3A mutations may play an important role in the development of MDS. Besides, RUNX1 mutation was the second most common accompanied genetic alteration in SF3B1-mutated patients at diagnosis. Intriguingly, three of nine SF3B1-mutated MDS patients who were serially analyzed and had disease progression acquired RUNX1 mutation, whereas none acquired DNMT3A mutation, at disease progression (Table 4). The findings implied that RUNX1 mutation may play a role not only in the development but also progression of SF3B1-mutated MDS, whereas DNMT3A mutation may play a little role in disease progression.
No study regarding sequential analysis of SF3B1 mutation during the clinical course of MDS has been reported in literature till now. In a study of SF3B1 mutations in chronic lymphocytic leukemia (CLL) , Rossi et al. found 5% of newly diagnosed CLL patients and 17% of fludarabine-refractory patients had this gene mutation. Serial samples of three SF3B1-mutated patients were analyzed. SF3B1 mutations were acquired when CLL had high grade transformation or became fludarabine-refractory in two of them. The authors suggested that SF3B1 mutations might occur during disease evolution of CLL. To the best of our knowledge, this study is the first to evaluate the dynamic change of SF3B1 mutation during disease progression in a large cohort of patients with MDS. We found that all SF3B1-mutated patients studied retained the same mutation during sequential follow-ups with the exception of two patients who lost SF3B1 mutations in remission status after HSCT. On the contrary, 123 of 125 patients without SF3B1 mutation at diagnosis remained SF3B1-wild during the clinical courses except the two who acquired SF3B1 mutation during follow-ups; however, the mutation could be detected in the initial sample by TA cloning technique, a method more sensitive than direct sequencing. Together, these findings showed that SF3B1 mutations were quite stable during clinical course and may play little role in the progression of MDS. Interestingly, the most frequent mutation that was acquired during disease progression in SF3B1-wild patients was NRAS mutation (7 out of 63), followed by RUNX1 mutation (4 out of 63), quite different from that in SF3B1-mutated patients in whom RUNX1 mutation was the most common mutation acquired (3 of 9) during disease evolution. It seemed that NRAS mutation may play a more important role than RUNX1 mutation in disease progression of SF3B1-wild patients, while the opposite was true for SF3B1-mutated patients. However, further studies will be needed to clarify the point.
In summary, SF3B1 mutations were present in 10% of MDS patients, and were positively associated with higher platelet count, older age, lower LDH level, RARS subtype, and DNMT3A mutation. SF3B1 mutations had no prognostic implication on OS and time to AML transformation. The sequential study showed SF3B1 mutations were stable during the clinical course and may be used as a marker to monitor treatment response. The most common mutation acquired during disease progression in SF3B1-mutated patients was RUNX1 mutation, whereas that in SF3B1-wild patients was NRAS mutation.
C.-C.L was responsible for literature collection, data management and interpretation, statistical analysis and manuscript writing; H.-A.H. was responsible for study design and plan, literature collection, data management and interpretation, statistical analysis, and manuscript writing; C.-Y.L was responsible for statistical analysis and interpretation of the statistical findings; Y.-Y.K., L.-I.L. was responsible for mutation analysis and interpretation; C.-Y.C., W.-C.C., M.-Y., S.-Y.H., J.-L.T., B.-S.K., S.-C.H., S.-J.W., S.-C.C., W.T., and Y.-C.C. contributed patient samples and clinical data; M.-H.T., C.-F.H., Y.-C.C., C.-Y. L., F.-Y.L. and M.-C.L. performed the gene mutation and chromosomal studies and H.-F.T. planned, designed, coordinated the study over the entire period and wrote the manuscript.