Introduction

Acute myeloid leukemia (AML) is the most common type of hematologic neoplasms in adults, which is characterized by an abnormal expansion of hematopoietic stem/progenitor cells with limited differentiation and excessive proliferation. Conventional cytogenetics analysis finds that approximately 55 % of AML patients exhibit chromosome abnormalities, which can be utilized widely to evaluate the risk status of AML [1]. With the fast development of genetic technology, the discoveries of somatic gene mutations involved in tyrosine kinase signaling, transcriptional, and epigenetic regulation have contributed to the improvement in risk stratification and prognosis prediction of AML [2, 3]. For example, the presence of c-KIT mutations in patients with core-binding factor (CBF) in AML confers a higher risk of relapse [4, 5]. FLT3-ITD mutations are considered to confer a significantly poorer outcome in cytogenetically normal patents [6, 7]. Recently, a new category of genes involved in RNA splicing pathway has been found to be mutated in myeloid neoplasms, most frequently (44–85 %) in myelodysplastic syndrome (MDS), less frequent (26 %) in AML transformed from MDS, and rarely (6.6 %) in de novo AML [8].

SRSF2 (also known as SC35), a member of the splicing factors, belongs to the Ser/Arg-Rich (SR) protein family that plays a crucial role in the recognition of the 3′ splice site during RNA splicing pathway [9]. Hotspot mutations in the position codon 95 of SRSF2 gene have been found in patients with MDS, chronic monomyelocytic leukemia (CMML), primary myelofibrosis (PMF), and AML [8, 10–14]. SRSF2 mutation appears to predict adverse prognosis in MDS and PMF [12–15], while it is not associated with the survival of CMML patients [10, 11]. To date, little has been known about the clinical relevance and prognostic implications of SRSF2 mutations in de novo AML patients [8, 16]. In the current study, we analyzed the frequency of SRSF2 mutations and evaluated its clinical relevance in 249 Chinese AML patients.

Materials and methods

Patients and DNA extraction

Two hundred forty-nine patients with de novo AML were diagnosed according to the French–American–British (FAB) Cooperative Group Criteria and the World Health Organization (WHO) classification [17, 18]. Bone marrow samples were collected at the time of initial diagnosis after the informed consent was written. The bone marrow mononuclear cells (BMNCs) were isolated by density-gradient centrifugation using Ficoll, and subsequent genomic DNA extraction from BMNCs was performed according to the manufacturer’s instructions. This study was approved by the Ethics Committee Board of Affiliated People’s Hospital of Jiangsu University.

Primer design and PCR conditions

Primers for PCR were designed with LightScanner primer design software v1.0 (Idaho Technology, Salt Lake City, Utah) and as follows: 5′-TGCAAATGGCGCGCTAC-3′ (Forward); 5′-GGCGGCTGTGGTGTGAG-3′ (Reverse). PCR was carried out in 25-μl volume in the presence of 1× PCR buffer (Invitrogen, Merelbeke, Belgium), 0.2 mmol/l of each dNTP, 0.5 μmol/l of both forward and reverse primers, 0.8 μmol/l of internal oligonucleotide calibrators [19], 1 × LCgreen Plus (Idaho, Salt Lake City, Utah, USA), 0.75 U hot start DNA polymerase (Takara, Tokyo, Japan), and 50 ng genomic DNA. PCR reactions were performed on a 7300 Thermo cycler (Applied Biosystems, Foster City, CA, USA). The PCR conditions were as follows: 98 °C for 10 min, followed by 40 cycles of 98 °C for 10 s, 63 °C for 30 s, and 72 °C for 30 s, followed by a final 7 min extension step at 72 °C.

High resolution melting analysis

PCR products were transferred to the LightScanner (Idaho Technology Inc. Salt Lake City, Utah, USA) for high resolution melting analysis (HRMA). Plates were heated from 55 to 95 °C with a ramp rate of 0.10 °C/s. The melting curve analysis was carried out by the LightScanner software package with CALL-IT® software (Idaho Technology Inc., Salt Lake City, Utah, USA).

Other nine genes recurrent in AML, including C-KIT, CEBPa, NPM1, FLT3-ITD, RAS, IDH1/IDH2, DNMT3A, SF3B1, and U2AF1 mutations, were detected as reported previously [20–24].

DNA sequencing

To verify the results of HRMA, a separate PCR was carried out to generate a larger product (212 bp). Primers used as follows: 5′-TTCGCCTTCGTTCGCTTTCA-3′ (forward) and 5′-CCCCTCAGCCCCGTTTACC-3′ (reverse). PCR conditions were similar with that for HRMA except for the annealing temperature at 60 °C. PCR products were directly sequenced on both strands using an ABI 3730 automatic sequencer.

Statistics

Statistical analysis was performed using the SPSS 17.0 software package (SPSS, Chicago, IL, USA). The difference of discrete variables was compared by Pearson chi-square analysis or Fisher exact test. The difference of continuous variables was compared by Mann-Whitney’s U test. Survival was analyzed according to the Kaplan–Meier method. The individual impacts of prognostic factors on OS were estimated using Cox proportional hazards regression. For all analyses, the P values were two-tailed, and less than 0.05 was considered statistically significant.

Results

SRSF2 mutations in AML patients

HRMA was performed to screen mutations in SRSF2. About 3.6 % (9/249) of cases were identified with heterozygous SRSF2 mutations in 249 AML patients (Table 1), which showed complex changes of melting curve shapes, compared to smooth and symmetric melting curves in other 240 wild-type SRSF2 patients (Fig. 1). The identical judgment was demonstrated by two blinded investigators. All HRMA results were further verified by direct sequencing which disclosed one heterozygous P95R (c.284C>G), two heterozygous P95L (c.284C>T), and six heterozygous P95H (c.284C>A) (Fig. 2). The clinical characteristics of SRSF2-mutated patients were shown in Table 2.

Table 1 Distribution of SRSF2 mutations in AML patients
Full size table
Fig. 1
figure 1

Representative results of the HRMA for detecting SRSF2-P95 mutations in AML. Gray lines represented wild-type SRSF2; pale blue and red lines were validated as P95H mutations by Sanger sequencing; green and blue lines were validated as P95L mutations; orange line was validated as P95R mutation

Full size image
Fig. 2
figure 2

Sequence chromatogram of SRSF2-P95 mutation. a wild-type SRSF2; b–g P95H mutation (c.284C>A); h–i P95L mutation (c.284C>T); j P95R mutation (c.284C>G). The rectangle denoted the ninety-fifth codon of SRSF2 gene. The mutated nucleotide was indicated by a red arrow

Full size image
Table 2 The clinical and hematopoietic parameters of nine patients with SRSF2 mutations
Full size table

There was no difference in sex, blood parameters, WHO subtypes, and karyotype classification between cases with and without mutations (P > 0.05, Table 1). However, the patients with SRSF2 mutations were older than those with wild-type SRSF2 (P = 0.014). There was also no significance in SRSF2 mutation between cytogenetically normal (CN) AML (6/114, 5.20 %) and cytogenetically abnormal AML (3/119, 2.52 %) (P = 0.279).

To investigate the prognostic impact of SRSF2 mutation on AML, 173 cases with follow-up data were considered for survival analysis. Although the patients with SRSF2 mutation had the lower rate of complete remission (CR) after induction chemotherapy than those without SRSF2 mutation (28.57 vs. 50.88 %), the difference was not statistically significant (P = 0.249). The median overall survival (OS) of patients with SRSF2 mutations was 4 months (95 % confidence interval = 1.6–6.3 months), obviously shorter than those without SRSF2 mutations (median = 11 months, 95 % confidence interval = 7.4–14.6 months) (P = 0.006, Fig. 3a). Within patients with CN-AML, SRSF2-mutated cases also had a shorter OS than those without SRSF2 mutations (median = 2 vs. 12 months, P = 0.008, Fig. 3b). However, further multivariate analysis including SRSF2 mutation and other six prognostic factors selected by univariate analysis (P < 0.1) showed that significant impact of SRSF2 mutation on OS was lost (P = 0.119, Table 3).

Fig. 3
figure 3

Overall survival of AML patients according to Kaplan–Meier analysis. a Whole AML patients; b CN-AML

Full size image
Table 3 Univariate and multivariate analyses for OS in MDS patients
Full size table

Association of SRSF2 mutations with other mutations

The mutational data of other nine genes recurrent in AML were available (C-KIT, CEBPa, IDH1/IDH2, DNMT3A, NPM1, FLT3-ITD, RAS, SF3B1, and U2AF1) in this cohort of AML. Overall, 44.2 % (110/249) of cases carried at least one gene mutation. Among the nine patients with SRSF2 mutations, three patients simultaneously had the IDH2 mutation, NRAS mutation, and NPM1 mutation, respectively. No association was observed between SRSF2 mutations and other gene mutations (Table 1). It was noted that the most commonly mutated genes in RNA splicing (SF3B1, U2AF1, and SRSF2) were occurred in a mutually exclusive manner.

Discussion

Recurrent somatic mutations involved in RNA splicing machinery have been firstly reported in myelodysplasia by Yoshida et al. [8], suggesting that genetic alternations of the critical splicing components (mainly including SF3B1, U2AF1, and SRSF2 genes) could contribute to the pathogenesis of disease. Later, increasing studies investigated these mutations in MDS and other hematologic diseases. In the present study, we identified a very low frequency of SRSF2 mutations (3.6 %) in Chinese de novo AML patients. The low frequency was not caused by the methodology as the maximal sensitivity of HRMA could reach 10 % in a background of wild-type DNA, which was higher than those of direct DNA sequencing (25 %) and all HRMA results were further confirmed by direct DNA sequencing [25]. Besides, our result was in accordance with another two papers reported (0.7 and 5.6 %, respectively) [8, 16], implicating that SRSF2 mutation is a rare event in AML patients.

All SRSF2 mutations were heterozygous. In the previous studies, we have investigated the mutations in another two critical splicing factors SF3B1 and U2AF1 genes (1.1 and 2.5 %, respectively) [20, 21]. In total, 6.8 % (17/249) of AML patients harbored the three splicing genes mutations and we observed the same results in other studies that the three splicing factors mutated in mutually exclusive manner [8, 11, 14], suggesting that they might not affect RNA splicing and progression of disease cooperatively.

The presence of SRSF2 mutations was significantly correlated with older age in our AML cohorts, which was in accordance with the observations in CMML, PMF, and MDS [10, 13, 15]. However, no correlations in sex, blood parameters, FAB subtypes, and karyotype classification were noted between cases with and without mutations. The prognostic impact of SRSF2 mutations in MDS remains obscure: Makishima et al. reported that SRSF2 mutations had worse survival in low-risk MDS but not in the entire cohort of MDS patients [14]. Thol et al. reported that SRSF2 mutations were correlated with inferior survival and a higher rate of progression to AML [12]. Wu et al. proposed that SRSF2 mutation was associated with shorter OS but not an independent prognostic factor in MDS or in low-risk MDS and not associated with disease progression [15]. Our previous study also confirmed the adverse impact of SRSF2 mutation on survival in MDS [20]. In PMF, SRSF2 mutations predicted inferior OS and leukemia-free survival (LFS) [13]. Although Herold et al. identified SRSF2 mutations in 81 % of AML patients with isolated trisomy 13 (+13) [26], no case with +13 was found in our group. So far, no prognostic relevance has been found in AML. Our data revealed that the presence of SRSF2 mutations was significantly associated with worse OS according to univariate analysis; however, multivariate analysis did not disclose the association of SRSF2 mutation with prognosis. Clonal hematopoiesis with somatic mutations including SRSF2 mutation has been identified as an increasingly common event in aged people and is associated with increased risks of hematologic cancer [27–29]. Furthermore, a recent study has revealed that SRSF2 mutation contributes to myelodysplasia by altering SRSF2’s RNA binding activity and thereby altering the exon recognition to drive abnormal splicing of key hematopoietic regulators [30]. These results suggest that SRSF2 mutation is an early driver mutation involved in leukemogenesis but not a prognosis-associated event.

Previous studies have shown that in CMML SRSF2 mutations are frequently co-occurred with other mutations, such as TET2, IDH1/IDH2, RAS, RUNX1, and ASXL1 mutations whereas mutually exclusive with EZH2 mutation. Besides, patients with RUNX1 mutations overlapped with SRSF2 mutations have a favorable survival [10, 13–15]. Among the nine patients with SRSF2 mutations in our study, three patients simultaneously had IDH2 mutation, NRAS mutation, and NPM1 mutation, respectively. No significant association of SRSF2 mutation with them due to the low frequency in AML.

In summary, SRSF2 mutations are rarely mutated and associated with older age in Chinese AML patients. Furthermore, SRSF2 mutation might not be an independent prognostic event in AML.