Myelodysplastic syndromes (MDS) and myelodysplastic/myeloproliferative neoplasms (MDS/MPN) are heterogeneous groups of chronic myeloid neoplasms characterized by clonal hematopoiesis, varying degrees of cytopenia or myeloproliferative features with evidence of myelodysplasia and a propensity to acute myeloid leukemia (AML).1 In recent years, a number of novel gene mutations, involving TET2, ASXL1, DNMT3A, EZH2, IDH1/2, and c-CBL, have been identified in adult cases of chronic myeloid neoplasms, which have contributed to our understanding of disease pathogenesis.2, 3, 4, 5, 6, 7 However, these mutations are rare in pediatric cases, with the exception of germline or somatic c-CBL mutations found in 10–15% of chronic myelomonocytic leukemia (CMML) and juvenile myelomonocytic leukemia (JMML),8 highlighting the distinct pathogenesis of adult and pediatric neoplasms.9

Recently, we reported high frequencies of mutations, involving the RNA splicing machinery, that are largely specific to myeloid neoplasms, showing evidence of myeloid dysplasia in adult.10 Affecting a total of eight components of the RNA splicing machinery (U2AF35, U2AF65, SF3A1, SF3B1, SRSF2, ZRSR2, SF1 and PRPF40B) commonly involved in the 3′ splice-site (3′SS) recognition, these pathway mutations are now implicated in the pathogenesis of myelodysplasia.10 To investigate the role of the splicing-pathway mutations in the pathogenesis of pediatric myeloid malignancies, we have examined 165 pediatric cases with AML, MDS, chronic myeloid leukemia (CML) and JMML for mutations in the four major splicing factors, U2AF35, ZRSR2, SRSF2, and SF3B1, commonly mutated in adult cases.

Bone marrow or peripheral blood tumor specimens were obtained from 165 pediatric patients with various myeloid malignancies, including de novo AML (n=93), MDS (n=28), CML (n=17) and JMML (n=27), and the genomic DNA (gDNA) was subjected to mutation analysis (Supplementary Table 1). The status of the RAS pathway mutations for the current JMML series has been reported previously (Supplementary Table 2).11, 12 Nineteen leukemia cell lines derived from AML (YNH-1, ML-1, KASUMI-3, KG-1, HL60, inv-3, SN-1, NB4 and HEL), acute monocytic leukemia (THP-1, SCC-3, J-111, CTS, P31/FUJ, MOLM-13, IMS/MI and KOCL-48) and acute megakaryoblastic leukemia (CMS and CMY) were also analyzed for mutations. Peripheral blood gDNA from 60 healthy adult volunteers was used as controls. Informed consent was obtained from the patients and/or their parents and from the healthy volunteers. We previously showed that for U2AF35, SRSF2 and SF3B1, most of the mutations in adult cases were observed in exons 2 and 7, exon 1, and exons 14 and 15, respectively.10 Therefore, we confirmed mutation screening to these ‘hot-spot’ exons. In contrast, all the coding exons were examined for ZRSR2, because no mutational hot spots have been detected. Briefly, the relevant exons were amplified using PCR and mutations were examined by Sanger sequencing, as previously described.10 The Fisher’s exact test was used to evaluate the statistical significance of frequencies of mutations for U2AF35, SF3B1, ZRSR2 or SRSF2 in adult cases and pediatric cases. This study was approved by the Ethics Committee of the University of Tokyo (Approval number 948-7).

No mutations were identified in the 28 cases with pediatric MDS, which included 13 cases with refractory anemia with excess blasts, 5 with refractory cytopenia of childhood, 2 with Down syndrome-related MDS, 2 with Fanconi anemia-related MDS, 2 with secondary MDS and 4 with unclassified MDS. Similarly, no mutations were detected in 93 cases with de novo AML or in 17 with CML, as well as 19 leukemia-derived cell lines. Our previous study in adult patients showed the frequency of mutations in U2AF35, SF3B1, ZRSR2 or SRSF2 to be 60/155 cases with MDS without increased ring sideroblasts and 8/151 de novo AML patients, emphasizing the rarity of these mutations in pediatric MDS (P<5.0 × 10−6) and AML (P<0.02) compared with adult cases. We found mutations in two JMML cases, JMML 4 and JMML 17. JMML 4 carried a heterozygous U2AF35 mutation (R156M), whereas JMML 17 had a 6-bp in-frame deletion (c.518-523delAAGTCC) in SRSF2 that resulted in deletion of amino acids S170 and K171 (Figure 1). Both nucleotide changes found in U2AF35 and SRSF2 were neither identified in the 60 healthy volunteers nor registered in the dbSNP database (http://www.ncbi.nlm.nih.gov/projects/SNP/) or in the 1000 genomes project, indicating that they represent novel spliceosome mutations in pediatric cases.

Figure 1
figure 1

Novel U2AF35and SRSF2 mutations detected in JMML cases. (a) Left panel: sequence chromatogram of a heterozygous mutation at R156 in N-terminal zinc-finger motifs of U2AF35 detected in a JMML case (JMML 4) is shown. Mutated nucleotides are indicated by arrows. Right panel: illustration of functional domains and mutations of U2AF35. Red arrow heads indicate hot-spot mutations at S34 and Q157 detected in the adult cases.10 Blue arrow head indicates the missense mutation at R156. (b) Left panel: sequence chromatogram of a 6-bp in-frame deletion (c.518-523delAAGTCC) in SRSF2 detected in JMML 17 is shown. Mutated nucleotides are indicated by arrows. Right panel: illustration of functional domains and mutations of SRSF2. Red arrow head indicates hot-spot mutation at P95 frequently detected in the adult cases.10 Blue arrow head indicates a 6-bp in-frame deletion leading to deletion of S170 and K171.

Full size image

U2AF35 is the small subunit of the U2 auxiliary factor (U2AF), which binds an AG dinucleotide at the 3′SS, and has an essential role in RNA splicing.13 With the exception of a single A26V mutation found in a case of refractory cytopenia with multilinage dysplasia, all the U2AF35 mutations reported in adult myeloid malignancies involved one of the two hot spots within the two zinc-finger domains, S34 and Q157, which are highly conserved across species, suggesting the gain-of-function mutations.10 In JMML 4, the R156M U2AF35 mutation affects a conserved amino acid adjacent to Q157, suggesting it may also be a gain-of-function mutation, leading to aberrant pre-mRNA splicing possibly in a dominant fashion.

SRSF2, better known as SC35, is a member of the serine/arginine-rich (SR) family of proteins.14 SRSF2 binds to a splicing-enhancer element in pre-mRNA and has a crucial role not only in constitutive and alternative pre-mRNA splicing but also in transcription elongation and genomic stability.14 All mutations thus far identified in adult cases exclusively involved P95 within the intervening sequence between the N-terminal RNA-binding domain and the C-terminal RS domain.10 This region interacts with other SR proteins, again suggesting that the P95 mutation may result in gain-of-function.10 This proline residue is thought to determine the relative orientation of the two flanking domains of SRSF2, and a substitution at this position could compromise critical interactions with other splicing factors necessary for RNA splicing to take place. In contrast, the newly identified 6-bp in-frame deletion in JMLL 17 results in two conserved amino acids, S170 and K171, within the RS domain. Although it may affect protein–protein interactions, the functional significance of this deletion remains elusive.

JMML is a unique form of pediatric MDS/MPN characterized by activation of the RAS/mitogen-activated protein kinase signaling pathway; in 90% of cases, there are germ line and/or somatic mutations of NF1, NRAS, KRAS, PTPN11 and CBL.8 Although JMML shares some clinical and molecular features with CMML, its spectrum of gene mutations suggests that it is a neoplasm distinct from CMML.15 This was also confirmed by the current results that the splicing-pathway mutations are rare in JMML, whereas they are extremely frequent (∼60%) in CMML.10 Although the two JMML cases carrying the splicing-pathway mutations had no known RAS-pathway mutations, both the pathway mutations frequently coexisted in CMML.8

To summarize, no mutations of SF3B1, U2AF35, ZRSR2 or SRSF2 are found in pediatric MDS and AML. In our study, except for ZRSR2, mutations were examined focusing on the reported hot spots in adult studies, raising a possibility that we may have missed some mutations occurring in other regions. However, these hot spots represent evolutionally conserved amino acids and have functional relevance, it is unlikely that the distribution of hot spots in children significantly differs from adult cases and as such, we could safely conclude that mutations of SF3B1, U2AF35, ZRSR2 and SRSF2 are rare in myeloid neoplasms in children. Finally, mutations of U2AF35 and SRSF2 may have some role in the pathogenesis of JMML, although further evaluations are required.