JAK2 Dependent GM-CSF Signaling Triggers SRSF2 Phosphorylation: A Potential Cause Of Mutation-Independent RNA Splicing Aberrancy In CMML

SRSF2 is a RNA-binding protein frequently mutated in CMML tasked with marking exon-intron boundaries during alternative splicing. SRSF2 contains a domain consisting of serine-arginine repeats that are dynamically phosphorylated and comprise a major regulatory mechanism in RNA capture. It has been previously demonstrated that EGFR signaling can elicit the phosphorylation of SRSF2 in an AKT dependent manner, resulting in functional alterations in exon usage of SRSF2 target genes. We previously reported that GM-CSF hypersensitivity is a reproducible feature of CMML. Because AKT is a known signaling intermediate of GM-CSF, we hypothesize that aberrant GM-CSF signaling could induce mutation-independent splicing abnormalities in CMML via phosphorylation of SRSF2. To test this hypothesis, we selected CMML samples that displayed GM-CSF hypersensitivity but lacked mutations by Sanger sequencing in 5 splicing-related genes commonly mutated in myeloid malignancies (SRSF2, SF3B1,U2AF1, ZRSF2,and PRPF40B) and labeled this group as CMML ‘splicing WT’. A second sample set included CMML specimens with isolated SRSF2 mutation (SRSF2 ‘splicing MT’). Bone marrow mononuclear cells from the SRSF2 ‘splicing WT’ group (n=4), CMML ‘splicing MT’ group (n=4), and normal controls (n=4) were lymphodepleted (CD3 and CD19) to enrich for the clonal myeloid population. RNA was extracted and whole transcriptome RNA sequencing and exon arrays were performed to compare the relative global exon usage of CMML ‘splicing WT’, CMML ‘splicing MT’, and normal controls. Briefly, the human exon 1.0 ST array was used for exon array analysis and splice index was calculated with the Exon/Gene Array Analyzer Web Tool (Gellert Bioinformatics 2009). RNA sequencing data was generated and aligned with TopHat2 (Kim Genome Bio 2013) while read counts and exon usage assignments were made with DexSeq (Anders Genome Res 2012). Splice indices determined by exon array demonstrated statistically significant exon usage differences between CMML ‘splicing MT’ and normal controls consistent with previously published results (Figure 1a). Interestingly, when CMML ‘splicing WT’ was compared to normal controls, a statistically significant difference in exon usage was also identified (Figure 1b). Further, when ‘splicing MT’ and ‘splicing WT’ were compared by splice index, differences in exon usage differences were dramatically reduced indicating similar splicing abnormalities in the CMML ’splicing WT’ group (1c). Euclidian hierarchical clustering at the gene- and exon-level separated normal controls from CMML groups but failed to separate ‘splicing MT’ from ‘splicing WT’, supporting the notion that mutation-independent splicing aberrancy may characterize CMML (1d). Global patterns in exon usage were also tested by whole transcriptome RNA-sequencing and were similar to those described above identified by exon array splice index.