ABSTRACT: Celf6 is an RNA binding protein expressed in a subset of neurons in the brain. We conducted an analysis of splicing in RNAseq data derived from Celf6 knockout mice, compared to wildtype littermates. We profiled both total tissue RNAseq from a hindbrain dissection, as well as Translated Ribosome Affinity Purified (TRAP)seq specifically from serotonergic neurons – a population of cells we previously characterized as expressing Celf6. We used MISO for the splicing analysis (PMID: 21057496). MISO uses read evidence found in a sample in a Bayesian model in order to assign a percent spliced in (PSI) value, defining the expected frequency of an alternate event. In the total RNA sample, the number of measurable PSI events was 1817. The most annotated events found were the Skipped Exons (35.5%) , Retained Introns (25 %), and Tandem 3’UTR (alternative poly-adenylation, 17.2%) categories. In general, the Skipped Exons category has the largest number of annotated events in MISO (Figure 1C). Out of 1817 assignable events, 95 showed nominal ANOVA p<0.05 between WT and Celf6 KO with 0 meeting Benjamini-Hochberg FDR <0.05. Alternative event categories had roughly the same distribution in nominally significant differentially abundant alternative events as to the data as a whole. 33/95 nominally significant events were in genes that were putative CELF6 targets, however only two (C1d, Rit2) were in genes matching both CLIP stringency criteria (significant enrichment in both CLIP input & antibody non-specific binding controls). As PSI describes the probability of being either in alternative form A (PSI = 0) or B (PSI = 1), large changes in PSI between one sample to another might indicate a splicing switch between predominant forms. Among the nominally significant events, the majority (88/95, 92.6%) were between -0.2 to 0.2 indicating only slight changes in splice form abundance. There were a few events with larger changes in PSI, with differences between KO and WT greater that 0.2, 5 in alternative polyadenylation, 1 in a Mutually Exclusive Exon event, and 1 in Alternative 5’ Splice Site usage. The largest change of these was in Acox1 (not a putative CELF6 target), peroxisomal straight-chain acyl-CoA oxidase, with a change of -0.62 in the KO (p = 0.037) which may indicate evidence of a switch in usage of a 897 nucleotide 3’UTR in the WT (WT median PSI = 0.89) to a 1660 nucleotide 3’UTR in the KO (KO median PSI = 0.26). However, none of these survive FRD correction. In serotonergic cell mRNA accessed by TRAP, the total measurable events was 2,201 with a distribution of effect sizes similar to that found in total RNA. 112 of these events showed nominal p values < 0.05, with 2 events meeting Benjamini-Hochberg FDR < 0.05, which showed a similar distribution to the data as a whole. 40 nominally significant events were found in putative CELF6 CLIP targets, 5 genes meeting the most stringent criteria (Snap25, Tpm1, Rnf14, Shroom2, App). As with total RNA, most events showed small differences in PSI between WT and KO (-0.2 to 0.2, median ~.01), with only 4 events outside this range. The largest such difference was found in Nmnat2, nicotinamide nucleotide adenylyltransferase 2 (not a putative CELF6 target), again in an alternative 3’ UTR. The WT median PSI of 0.55 indicates two isoforms are similarly abundant: a 3.6 kb 3’UTR and a 400 bp UTR. In the KO, this is shifted towards the shorter UTR (KO median PSI 0.98).
