Splicing Analysis Unveils Convergent Transcriptomic Pathways in U2AF1 S34 and Q157 Mutant Myeloid Neoplasia

Abstract

Splicing factor mutations (SFMT) are frequent in MDS and their pathogenic consequences may include missplicing and changes in alternative splicing (AS). However, divergent sequence-sequence effects on misspliced genes were recorded in U2AF1MT at the two canonical hotspots (S34, Q157). Most of the current studies have been focused on the identification of distinctive features of splicing dysfunction between the two hotspots, unveiling various unique hotspot specific splicing events. Classical instances of missplicing events include those in ATG7,H2AFY and STRAP specific to U2AF1S34 MDS. However, whereas we and others have demonstrated certain unique clinical features of MDS patients carrying S34 vs Q1571, these two alterations produce a generally similar disease phenotype and likely rely on a common transformation mechanism. This conclusion suggests that missplicing events might converge in pathways relevant to MDS pathogenesis, which may be much more widespread than what is believed. While defining a topology of AS events (ASE), our analysis of the global landscape of RNA splicing in myeloid neoplasia (MN)2 also opened the possibility to search for subsets of patients sharing RNA splicing features and gene expression (GE). We now curated RNA-Seq data from diagnostic samples of 1258 MN patients by comparing RNA splicing profiles of clonal primary cells with S34 (n=19) and Q157 (n=26) (VAFs >10%, both) in search for overlapping ASE representing new cooperative mechanisms germane to leukemogenesis. ASE were filtered based on percent spliced in (PSI) (variation threshold > 0.01). Limma analysis for differential GE and PSI was applied. All data were analyzed on the genome assembly GRCh37/hg19. Patients without SFMT plus a cohort of 63 age-matched controls were used to detect exclusive ASE common in U2AF1 S34 and Q157. Pathway analysis showed an enrichment of chromatin remodeling, histone demethylases and signal transduction genes. A total of 17564 ASE were called with 30% (n=5106) of them found to be significant (qvalues ≤ .05 and inclusion level difference ≥ 5%). In-depth analysis of ASE showed that 70% of significant PSI changes resulted from exon skipping (ES) followed by intron retention (17%) and alternative 3′ and 5′ splice site events (14%). No significant mutually exclusive ASE was found. Considering that ES might result in truncated proteins and consequent lack of functional domains, we mapped ASE to protein domains. Top PSI changes which were common between S34 and Q157 were found in chromatin remodeling (EZH2,JMJD1C, ARID2/4B, CREBBP), O-linked β-N-acetylglucosamine transferase (OGT), a partner of TET2, GTPase activating (GNAS, ARHGAP9, HBS1L), RNA processing/ binding (RBM5, DDX21, DHX38,HNRNPM, PRPF3), unfolded protein response (XBP1), and ubiquitination (CUL3, NAE1, UBQLN1) genes. Lesions affecting chromosome 7 either structurally (cytogenetic alterations) or by mutations and CNV (EZH2 among others) are associated with adverse prognosis in MN. In total, 8% of the patients had deletion 7 and 4% carried EZH2MT. EZH2 mRNA expression levels were significantly lower in U2AF1MTvs controls (LFC=-0.88; P= 1e-11). Of note is that only 6 cases carried EZH2MT, of whom none harbored deletion 7. This suggests that EZH2 low expression levels might be due to other mechanisms. Looking at most relevant findings, we identified a significant and single ES (ex12) in EZH2 flanking the pre-SET domain (cysteine rich, CXC: aa503-605) crucial to maintain histone methyltransferase activity. Beside somatic mutations, EZH2 function can be altered by other mechanisms, e.g., cooperative mutations. The presence of an ES possibly suggests that an alternative mechanism of gene inactivation beside mutations in the CXC/SET domain and chromosome loss might play a role. Similarly, a CREBBP ES (ex11) was detected upstream the kinase inducible interacting (KIX) domain, a docking site for transcriptional activators (e.g., MLL, c-Myb sites). Finally, U2AF1MT have been also associated with deletion 20q or with a hyperactive long GNAS isoform. We found indeed an ES in GNAS and other GTPases, further emphasizing the broad dysfunctions in U2AF1 cooperative genes. In conclusion, our study describes convergent splicing features common to both U2AF1 S34 and Q157 to conceptualize the essential mechanisms contributing to clonal expansion due to lesions in U2AF1.