Introduction

Myelodysplastic syndromes (MDS) are a heterogeneous group of clonal hematopoietic stem cell disorders leading to abnormal blood production and that have a variable risk of progression to acute myeloid leukemia (AML) [1,2,3].

In recent years, large-scale analysis using next-generation sequencing (NGS) has made it possible to identify recurrent genetic alterations, thereby improving our knowledge of MDS pathogenesis [4,5,6,7,8]. More than 80% of MDS patients harbor at least one mutation, affecting genes from a variety of functional groups: splicing machinery (SF3B1, SRSF2), DNA methylation (DNMT3A, TET2), transcription factors (TP53, RUNX1), chromatin modification (ASXL1, EZH2), RAS pathway (KRAS, NRAS), cohesin complex (STAG2, RAD21), kinases (JAK2, FLT3), and/or DNA repair (ATM, BRCC3) [4, 9,10,11].

Several clinical and biological implications of specific mutations have been demonstrated. Some of these gene mutations have been associated with morphological and clinical features such as complex karyotypes (TP53), excess proportions of bone marrow blast (RUNX1, NRAS) or ring sideroblasts (SF3B1), and with the prognosis for leukemia-free and overall survival [12, 13]. Mutations in TP53, U2AF1, RUNX1, SRSF2, IDH2, CUX1, ASXL1, and BCOR genes are associated with significantly worse leukemia-free survival [5]. Those in TP53, EZH2, ETV6, RUNX1, and ASXL1 are predictors of poor overall survival (OS), while mutations in SF3B1 are associated with a better outcome in MDS patients [6, 14]. Most of these studies have analyzed single mutations, but little is known about the impact of concurrent mutations on the outcome of MDS patients [15].

It is common for more than one mutation to be present in MDS. Papaemmanuil and colleagues clearly demonstrated the variability of gene mutation frequencies in a large series of MDS, whereby 40% of cases had 2 or 3 mutated genes and up to 10% of patients presented 4 to 8 oncogenic point mutations. All these features are associated with a more complex disease and have an adverse effect on OS [4, 5]. SF3B1 encodes a core component of the RNA splicing machinery. NGS studies have revealed that approximately 30% of MDS cases have a mutation of the SF3B1 gene, with a particularly high prevalence (> 90%) in the MDS with ring sideroblasts subtype (MDS-RS), as reflected in the most recent 2017 WHO classification [3, 13, 14]. Furthermore, SF3B1 mutations in low-risk MDS patients were associated with good prognostic parameters [13]. Although no significant affinity of SF3B1 with common mutational genes other than DNMT3A and JAK2 has been found, various gene mutations co-existing with SF3B1 have been described [4, 5].

Nevertheless, and given that the majority of MDS patients carry multiple genetic alterations, the well-known, better OS and leukemia-free survival referred to SF3B1 mutations in low-risk MDS patients may be worse. Recent studies in low-risk MDS–RS patients have highlighted the adverse influence of coexisting DNMT3A and ASXL1-SF3B1 mutations on clinical outcome [15, 16].

However, information about the influence of other gene mutations co-existing with SF3B1mut in MDS patients is scarce. Detailed molecular characterization of these groups would allow a better stratification of patients within the low-risk MDS categories as well as a better choice of treatment for these patients. The aim of this study was to analyze, by means of NGS, the presence of mutations associated with SF3B1 mut and to evaluate the prognostic value in a large series of very low/low/intermediate-risk MDS patients.

Materials and methods

Patients

The mutational profiles of 324 MDS patients, diagnosed in our center between 1999 and 2017, were analyzed. Diagnosis was based on the World Health Organization (WHO) criteria [3, 17]. Patients with refractory anemia with excess of blasts (RAEB) and MDS associated with isolated del(5q) were excluded from the study.

SF3B1 mutations were detected in 135 patients (42%). Detailed analysis of clinical parameters and cytogenetic findings was performed to facilitate risk stratification, according to the International Prognostic Scoring System (IPSS) and the revised IPSS (IPSS-R) [18, 19]. Finally, a total of 102 MDS cases with SF3B1 mutations and very low, low, and intermediate IPSS-R scores were included. The study was approved by the Local Ethical Committee (Comité Ético de Investigación Clínica, Hospital Universitario de Salamanca), and written informed consent was obtained for each patient according to the guidelines of the Declaration of Helsinki.

Mutational analysis

Mutational screening of genomic non-amplified DNA from bone marrow (BM) or peripheral blood (PB) cells was performed. A customized myeloid panel of 117 MDS-related genes was applied (Supplementary Table S1). NGS was carried out on a NextSeq sequencing platform (Illumina, San Diego, CA, USA) following Illumina’s standard protocol for Nextera Rapid Capture Enrichment. Sequencing data were analyzed by applying an in-house informatic pipeline that uses different software tools to perform quality assessment, alignment, and variant calling (Trimmomatic, FastQC, NGSQCToolkit, BWA, GATK, VarScan, SAMTools, ANNOVAR). The Integrative Genomics Viewer (IGV, Broad Institute) was used to evaluate variants visually.

After analysis, only those variants of good quality (Q > 30), supported by ≥ 100 total and ≥ 10 mutated reads, with a variant allele frequency (VAF) ≥ 3%, located in exonic or splicing regions, and which generate an amino acid change were considered. In addition, already reported polymorphisms (SNPs) (dbSNP144, 1000-genomes Project, ExAC, ESP-6500, when MAF ≥ 1%) and sequencing artifacts (internal laboratory database where distortions in data processing caused by a measuring instrument, already seen in previous NGS analysis in our laboratory, are annotated and stocked as a reference for further analysis) were discarded. For variant interpretation and oncogenic potential evaluation, COSMIC and ClinVar databases, SIFT, PolyPhen-2, and Mutation Taster predictors were used (Supplementary Fig. 1).

Statistical analysis

Numerical variables were summarized as the median and range; categorical variables were summarized as the frequency and percentage of subjects in each category. Group differences were examined with Student’s t test or the Mann-Whitney U test, respectively, for normally and non-normally distributed continuous variables. Overall survival was measured from the time of diagnosis to the time of last follow-up or death from any cause. Survival curves were generated using the Kaplan-Meier method and differences were assessed with the log-rank test. For multivariate analysis, Cox proportional hazards models were constructed, adjusting for potential confounding covariates. Significance of all statistical tests was defined as a value of p < 0.050. Analyses were carried out using the IBM SPSS version 22.0 statistical package (IBM Corp., Armonk, NY).

Results

One hundred and two patients with a mutation in SF3B1 gene were included as lower-risk cases according to the IPSS-R: very low (34.4%), low (60.4%), or intermediate (5.2%). Regarding the WHO 2017 classification, 1% of patients had MDS with single-lineage dysplasia (MDS-SLD), 5.2% had MDS with multilineage dysplasia (MDS-MLD), 40.6% had MDS with ring sideroblasts (MDS-RS) with single-lineage dysplasia (MDS-RS-SLD), and the remaining 53.1% had MDS-RS with multilineage dysplasia (MDS-RS-MLD). The median age was 76 years (range: 41–90 years). More detailed clinical features of the cohort are summarized in Table 1.

Table 1 Clinical features of 102 MDS patients
Full size table

Characterization of SF3B1 mutations in MDS patients

One hundred and seven SF3B1 mutations were found in 102 MDS patients. The median VAF was 33.6% (range 5.72–64.00%). All mutations were missense, heterozygous, and, except for three (K748E, Y623F, and L536V), had been previously reported. The most frequent mutation was K700E (47/107, 43.9%) followed by K666R (12/107, 11.2%) and E622D (9/107, 8.4%). All SF3B1 mutational variants detected are illustrated in Fig. 1. Notably, five of the 102 MDS patients presented two SF3B1 mutations. The double variants of four patients were located in exon 14 (E622D and Y623F; H662Q and K666T; and two cases with E622V and T663I), while the double SF3B1 mutation in the other patients affected exons 14 and 15 (R625H and K700E) (detailed information about all mutations is shown in Supplementary Table S2).

Fig. 1
figure 1

SF3B1 mutations distributed in patients included in the study and classified by WHO 2017 MDS subtype

Full size image

Co-occurrence of SF3B1 mutations with other gene mutations

Co-occurrence of SF3B1 mutations with those of other genes was observed in 82 of 102 patients (80.4%). A total of 192 concomitant mutations, involving 51 genes, were found in the 102 SF3B1mut patients, with a median of two additional mutations per sample (range: 0–5), (Supplementary Table S2). Patients with one mutation in addition to that of SF3B1 were the most frequent type (n = 26, 25.5%), followed by two and three concurrent mutations (n = 23, 22.5%, and n = 19, 18.6%, respectively). Cases with four and five concomitant SF3B1 mutations were infrequent (n = 7, 6.9% and n = 8, 7.8%, respectively). The most frequently mutated genes co-occurring with SF3B1 were as follows: TET2 (n = 40, 39.2%), DNMT3A (n = 26, 25.5%), SRSF2 (n = 11, 10.8%), CDH23 (n = 6, 5.9%), and ASXL1, CUX1, KMT2D (n = 5, 4.9% each). Mutations were also present in the BCOR, NUP98, SMC3, SETBP1, and STAG2 genes, although at low frequencies (n = 3, 2.9% each) (Fig. 2).

Fig. 2
figure 2

Mutations in 51 genes co-occurring with SF3B1 in 102 very low, low, and intermediate-risk MDS patients

Full size image

The great heterogeneity of the overall distribution of concomitant mutations is illustrated in Fig. 3. The addition of a single mutation to the SF3B1 mutation usually involves the TET2 or DNMT3A genes. However, the addition of two mutations concomitant with SF3B1 mutations did not show a clear pattern of distribution or evolution.

Fig. 3
figure 3

Mutational landscape of 102 MDS. Each column represents one patient. Patients are grouped by number of additional SF3B1 mutations. Each line represents one gene. Genes are grouped by function. Red circles indicate genes that reduce overall survival of MDS patients.VAF: ≥ 3 < 5% ≥ 5 < 10% > 10%. Gene function: splicing, DNA methylation, chromatin modification, transcription factors, activated signaling, cohesin complex, other. Cytogenetics: very good, good, intermediate, NA. IPSS-R: very low, low, intermediate, NA

Full size image

The effect of the frequency of co-occurring SF3B1 mutations

In the next step of the study, the influence of the number of concurrent mutations was analyzed, taking the clinical characteristics and overall survival (OS) into account. In general, no differences were observed in the clinical characteristic of the patients affected by between one and five of the additional SF3B1 mutations, (Supplementary Table S3). However, in the univariate analysis of the series, after a median follow-up of 54 months and a median OS of 70 months, we found that the presence of at least two mutations concomitant with SF3B1 had a more adverse effect on survival compared with those with SF3B1 + fewer than two additional mutations (median of 54 vs. 87 months, respectively, p = 0.007) (Fig. 4A). The negative impact of at least two additional SF3B1 mutations remained a significant factor in the multivariate analysis (Table 2).

Fig. 4
figure 4

Overall survival of MDS SF3B1mut patients by frequency of concomitant mutations (median 87 vs. 54 months) (A). Co-occurrence of SRSF2mut (median 75 vs. 27 months) (B). IDH2mut (median 75 vs. 11 months) (C). BCORmut (median 71 vs. 11 months) (D). NUP98mut (median 71 vs. 27 months) (E). STAG2mut (median 71 vs. 11 months) (F)

Full size image
Table 2 Univariate and multivariate Cox proportional hazard model for clinical features and number of additional mutations to SF3B1
Full size table

Several co-occurring SF3B1 gene mutations reduce OS of MDS patients

Frequently mutated genes overlapping with SF3B1 were analyzed (Fig. 2). There were no differences in the clinical features between groups of SF3B1mut patients with wild type or mutated forms of the most frequently mutated genes (data not provided). We observed that the presence of somatic mutations in other genes was able to modify the good prognosis of patients with isolated SF3B1mut. Thus, in the univariate analysis, co-occurrence of SF3B1 with SRSF2 mutations was associated with shorter OS than isolated SF3B1 mutations with SRSF2 wild type (median of 27 vs. 75 months, respectively; p = 0.001) (Fig. 4B). However, the negative impact of SRSF2 mutation was not retained as a significant term in the multivariate analysis (data not shown). Furthermore, although small groups of patients are involved, a similar adverse effect was observed with concomitant IDH2 mutations (median OS of 11 vs. 75 months, respectively; p = 0.001) (Fig. 4C) and BCOR mutations (median OS of 11 vs. 71 months, respectively; p = 0.036) (Fig. 4D). Interestingly, SF3B1 with NUP98, and SF3B1 with STAG2 also had a negative effect on patient prognosis (medians of 27 and 11 vs. 71 months, respectively; p = 0.008 and p = 0.002) (Fig. 4E and F). Very few patients bore these mutations, so they were not included in the multivariate analysis.

Co-occurrence of SF3B1 and clonal hematopoiesis indeterminate potential mutations

The presence of mutations in some genes considered CHIPs (clonal hematopoiesis indeterminate potentials) co-occurring with SF3B1 mutations was analyzed. More than half of MDS patients (57.8%, N = 59) displayed mutations in TET2, DNMT3A, and/or ASXL1 (Fig. 3). The comparison of the most relevant clinical and biological charactersitics, such as age, levels of hemoglobin, platelets and neutrophils, bone marrow percentage of blast and ring sideroblast, and OS, did not differ between the low-risk MDS patients with TET2 or DNMT3A as unique mutations associated with SF3B1 and SF3B1 mutations as the only abnormality (median OS of 70 and 87 months, respectively). Therefore, CHIP mutations had no effect on SF3B1-mutated cases (Table 3 and Fig. 5).

Table 3 Comparison of median values between patients with vs. without co-occurrence of SF3B1 mutation with CHIP-related genes mutations
Full size table
Fig. 5
figure 5

Overall survival of MDS SF3B1mut patients with respect to co-occurrence with mutations in CHIP genes (TET2 or DNMT3A) (median 87 vs. 70 months)

Full size image

Discussion

SF3B1 is one of the most frequently mutated genes in MDS and the presence of mutations in this gene is associated with a favorable outcome in MDS patients without an excess of blasts [4, 13]. However, information about the appearance and influence of other concurrent mutations in prognosis is scarce, although they are important topics since at least 40% of patients with MDS have at least two mutations [4, 5, 20]. In the present study, we analyzed the co-occurrence of mutations in SF3B1 and other frequent mutations in myeloid-related genes, the consequences of these concomitant mutations for the clinical and biological phenotype, and the impact on prognosis in a cohort of 102 MDS low-risk patients without an excess of blasts. Our study confirmed that a complex mutational SF3B1 status (at least two associated mutations) is associated with a dismal prognosis in low-risk MDS patients.

To ensure the consistency of our analyses, we included only those low-risk patients who were categorized as very low, low, and intermediate risk according to the IPSS-R. Clinical features did not differ from those typical of low-risk MDS patients with respect to age, gender, and WHO classification [19, 21, 22]. Patients with excess blasts were excluded from the study because the frequency of SF3B1 mutations in this population was low and, due to an adverse outcome, predefined in this subset of patients [4, 5, 23]. In addition, and based on another recognized entity, with a clear genotype-phenotype relationship and prognosis, patients with MDS with an isolated 5q deletion were excluded from the study [24, 25].

The MDS subtype most frequently found in our cohort was MDS-RS (93.7%) according to the WHO’s 2017 classification. The percentage of RS ranged between 0 and 95%, with the majority of cases (83.3%) containing more than 15% of RS [26, 27]. The 2017 WHO classification expanded the MDS-RS group to include cases with an RS level between 5 and 15% and confirmed the SF3B1 mutation, which involved 4.9% of the cases in our cohort (Table 1) [3]. Six patients could not be classified because their RS frequency was not known. The increase in the proportion of patients diagnosed with MDS-RS following the 2017 WHO reclassification has been noted in other studies [28].

The extent of SF3B1 mutations among MDS-RS patients was similar to that reported in previous studies. All the mutations found were missense and heterozygous, with a change from lysine to glutamic acid in codon 700 (K700E) being the most frequent [14, 29]. Apart from three variants (K748E, Y623F, and L536V), all were already known to occur in hematological cancer, and at similar frequencies (Fig. 1) [30,31,32]. Interestingly, five of 102 patients had double SF3B1 mutations (E622D and Y623F; H662Q and K666T; two cases with E622V and T663I and R625H and K700E). Double mutations occur rarely in the SF3B1, although they have been reported [14, 20].

Co-occurrence analyses of the SF3B1 gene with other myeloid-related genes revealed that only 20 patients (19.3%) carried a mutation of SF3B1 as the only mutation. Previous studies had reported that more than 40% of cases of MDS had at least two mutations, which underlines the need for a detailed and complete molecular analysis to gain a better knowledge of the molecular pattern and risk to our patients, especially those with low-risk MDS [4, 5]. The isolated SF3B1 mutation appeared as a minority group in our cohort and in other studies, with a tendency to be higher (40%) among the MDS-RS cases [4, 20]. Those discrepancies may have arisen because the studies used different NGS methods and VAF cut-offs. Martin et al. used the amplicon-based technique to analyze regions of 39 genes, while we used a custom-capture strategy to interrogate exonic regions of 117 genes. Martin et al. chose a VAF cut-off of 5% whereas the threshold was 3% in the present study. Nevertheless, the majority group of patients in both series bore additional mutations that coexisted with those of SF3B1.

The TET2 and DNMT3A genes were the most frequently mutated, co-occurring with SF3B1 in 39.2% and 25.5% of cases, respectively, similar to what was noted in previous series [4, 5, 16]. Mutations in SRSF2 were the next most frequently observed. Although these are quite frequent in MDS or MDS-RS patients, they rarely co-occur with SF3B1 [8, 33]. In our previous study of splicing genes in MDS-RS patients, we identified double splicing gene mutations and noted their possible influence on the outcome of patients. This prompted us to investigate in more detail the group of patients in which SF3B1 and SRSF2 occur jointly [14]. As a consequence, we found that patients with the SRSF2 + SF3B1 mutation were another of the most frequently occurring types (10.8%) in our series. This is a novel observation; double splicing gene mutations have previosuly been considered to be mutually exclusive [4, 5]. The other gene mutations known in MDS were quite similar to those previously reported in the MDS and MDS-RS subtypes, as shown in Fig. 2 [4, 5, 20].

Previous studies suggested that a higher frequency of driver mutations worsens the clinical outcome in these patients [5, 34]. Our study confirms that SF3B1 with at least two additional mutations is an independent prognostic factor associated with shorter survival, even in these low-risk MDS patients. This could be interpreted as meaning that a highly heterogeneous mutational status gives rise to a more complex disease.

We now consider the role of the co-occurrence of particular gene mutations with SF3B1. Previous studies suggested that the presence of some gene mutations in low-risk SF3B1mut MDS patients could modify clinical features and prognosis. Malcovati et al. found that in SF3B1mut patients, RUNX1 mutations were significantly associated with worse OS and a higher cumulative incidence of disease progression [13]. Similary, Martin et al. demonstrated the same pattern in RARS patients with SF3B1 and DNMT3A mutations [15]. As far as we know, these findings have not been confirmed in other cohorts. Our own inability to corroborate these findings could be due to an insufficient number of SF3B1mut with RUNX1mut patients and the greater heterogeneity of our study cohort compared with other series [15]. Nevertheless, we found a negative impact on prognosis when there was co-occurrence of SF3B1 with SRSF2 and/or BCOR and/or IDH2 and/or NUP98 and/or STAG2 mutations. With the exception of SRSF2, the small number of affected cases available to us means that these findings need to be confirmed by the analysis of a larger cohort of patients.

In 2014, NGS studies in healthy individuals revealed that around of 10–20% of older people (≥ 65 years of age) carried somatic mutations in hematopoietic disorder–related genes, making hematopoiesis clonal. The majority of the variants occurred in three genes: DNMT3A, TET2, and ASXL1. The clonal hematopoiesis of indeterminate potential (CHIP) of healthy people helped create an increased risk of subsequent hematological cancer and of the appearance of cardiovascular diseases [35, 36]. In the current study, DNMT3A, TET2, and ASXL1 were among the most frequently mutated genes. No effect of co-occurrence with SF3B1 was seen in relation to clinical features and prognosis, although, according to previous findings, their existence could be crucial for initiating the clonal hematopoiesis that evolves to become MDS disease [35,36,37]. These results demonstrate the importance and utility of the NGS technique not only for monitoring patients’ disease course but also for making it possible to detect molecular changes (clonal hematopoiesis) early on in otherwise healthy people, thereby increasing the opportunity to improve the follow-up of receptors and immediately determine whether the disease has evolved.

In summary, our study demonstrated that the co-ocurrence of additional myeloid mutations is very frequent (80.4%) in very low/low/intermediate-risk SF3B1-mutated MDS patients. The presence of at least two additional mutations is associated with shorter survival and the presence of specific mutations concomitant with those of SF3B1 (SRSF2, IDH2, BCOR, NUP98, and/or STAG2). The comprehensive mutational study of low-risk MDS patients with SF3B1 mutation is essential if they are to receive a more accurate prognosis.