Integrative Genome-Wide Analysis of RNA Binding and Splicing Reveals Complex Loss and Gain of Function Alterations By SRSF2 P95 Mutations in Myelodysplasia

Abstract

Specific splicing Factor (SF) mutations are recurrent and mutually exclusive in hematopoietic diseases.Mutations in the splicing factor SRSF2 occur in nearly 40% of patients with CMML, 14% of MDS, and 19% of secondary AML, and portend a poor prognosis. SRSF2 binds to exonic splicing enhancers (ESEs), thereby affecting exon inclusion or exclusion. We have recently shown that SRSF2 mutations result in altered RNA binding affinity and specificity via in vitro structure-function studies, specifically isothermal titration calorimetry and nuclear magnetic resonance (NMR) modeling. While wildtype (WT) SRSF2 binds the consensus RNAs 5'-SSNG-3' equally well (S=G/C; N=A/T/C/G), mutant (MUT) SRSF2 shows increased affinity specifically for 5'-CCNG-3' and 5'-GCNG-3', due to structural changes in the N- and C-termini of the RRM (Kim et al, Cancer Cell, 2015).To determine how mutations in SRSF2 alter its function in vivo, we generated hematopoietic cell lines with inducible expression of Flag-tagged SRSF2WT or SRSF2MUT. We performed "High-Throughput Sequencing after UV-Cross-linking and RNA Immunoprecipitation" (HITS-CLIP) to identify specific RNAs bound by WT versus MUT SRSF2 and performed RNA deep sequencing (RNA-Seq) on the same cells to correlate RNA binding and splice events.HITS-CLIP reads were highly enriched in protein-coding RNAs (83%) and specifically exonic RNA sequences (>90%), consistent with the known function of SRSF2 as an ESE-binding protein and emphasizing the specificity of HITS-CLIP in detecting in vivo SRSF2-mRNA interactions. To determine whether the differences in in vitro binding affinities determined by isothermal calorimetry held true in vivo, we selected those RNA reads differentially bound by WT and MUT SRSF2 and analyzed the distribution of RNA consensus motifs within these regions. The 5'-CCNG-3' motif was significantly enriched over the 5'-GGNG-3' motif in SRSF2MUT bound RNAs, whereas the 5'-GGNG-3' motif was significantly enriched in SRSFWTbound reads.Crosslinking followed by reverse transcription results in single nucleotide deletions at the protein-RNA crosslink site, allowing determination of the exact protein binding site and resolution of the SRSF2-mRNA interaction at the single-nucleotide level. To compute the RNA binding consensus in an unbiased fashion we therefore performed MEME (Multiple Em for Motif Elicitation) analysis of only those reads with nucleotide deletions. We observed enrichment for CCNG in MUT-bound motifs, whereas WT-bound motifs were GA-rich, as has been previously shown (Pandit et al, Mol Cell, 2013). In addition, first analyses utilizing deletions to discern whether SRSF2 ESEs are located within specific regions of the exon suggest that while WT SRSF2binding maps to all exon regions (5', middle, 3'), more MUT SRSF2 binding occurs in the 5' region of exons.We next sought to identify RNA targets that are both differentially bound and alternative spliced in SRSF2 WT versus MUT cells. Interestingly, gene enrichment analysis within bound and alternatively spliced transcripts using "ToppGene" (https://toppgene.cchmc.org/enrichment.jsp), revealed significant enrichment for genes involved in mRNA binding, RNA processing, and RNA splicing in both our HITS-CLIP and transcriptome data sets. Strikingly, similar enrichment is found via RNA sequencing in primary patient samples with WT vs MUT SRSF2. To validate the correlation between RNA-binding and splicing, we verified several alternative splice events predicted to have differential transcript abundance in the WT vs MUT HITS-CLIP data set via exon specific RT-PCR. These splice events result in either altered protein function via in-frame inclusion/exclusion of an alternative exon, or altered protein abundance via frame-shift splice changes that lead to the introduction of a premature stop codon, slating transcripts for degration by non sense mediated decay.In summary, changes in consensus motif preference, emphasize the complex gain and loss of function nature of SRSF2 mutations, further underlining the highly context-specific SRSF2MUT-mRNA interactions. Additionally, alternative splicing of transcripts encoding other splicing factors may be the link between some of the mutually exclusive splicing factor mutations in MDS and provide a common pathway in the pathology and treatment of splicing factor mutant myeloid malignancies. DisclosuresNo relevant conflicts of interest to declare.