Background: C-to-U RNA editing is mediated by APOBEC1 and two RNA binding proteins (A1CF and RBM47) in WT C57 mice. We found reduced polyp burden in Apcmin/+ mice crossed into Apobec1–/– mice, whose mechanisms are unknown. Here we explored the relative importance of epithelial APOBEC1 and RBM47 expression and identified novel RNA targets and mechanisms of regulation. Approach: Apcmin/+ mice were crossed into Rbm47 intestinal knockout (Apc-RIKO) and intestinal human Apobec1 transgenic mice (Apc-A1tg) to examine gain- and loss-of-function (GOF/LOF) of C-to-U RNA editing. RNA from Apcmin/+ polyp and uninvolved mucosa (UM) was pooled from >5 mice, deep sequenced and variants Sanger-sequenced to verify C-to-U RNA editing. Validated targets were examined in patients with familial adenomatous polyposis (FAP) or sporadic colorectal cancer (CRC). Findings: 36 novel, APOBEC1-dependent editing sites were identified in the 3’UTR of 29 RNAs in Apcmin/+ mice. 6 were edited only in polyp, 12 only in UM and 18 in both UM and polyp (Fig1a); 37 of 57 editing sites previously identified in C57 WT mice were validated in Apc min/+ mice. An RNA sequence composition /folding prediction model examining shared C57+Apc min/+ RNAs revealed that editing efficiency correlated with the AU content of Apc min/+ but not C57-specific targets. 23/29 APOBEC1 RNA targets were RBM47-dependent. QPCR analysis of Apobec1, Rbm47 and A1cf RNAs in C57 vs Apc min/+ mice showed reduced Rbm47 and increased A1cf in Apcmin/+ polyp, suggesting that editing efficiency is regulated via APOBEC1/RBM47/A1CF stoichiometry. GOF Apc-A1tg mice showed increased polyp burden and size (Fig1b) with only 21/36 sites edited (6 higher frequency, 4 lower frequency and 11 similar frequency vs Apcmin/+ mice); LOF Apc -RIKO mice showed larger polyps but similar overall burden (Fig1b), again suggesting a stoichiometric requirement for APOBEC1/RBM47. 3/29 RNAs showed altered expression in LOF Apc min/+ Apobec1–/– and Apc-RIKO (Flnb, Herc6, Casp8ap2). Flnb RNA (95 % edited in UM and polyps) was >50% edited in individual tumors while individual mice showed regional variability in proximal and mid intestine but consistently ~90% editing in distal region (Fig1c). We observed reduced Flnb mRNA and protein in Apc-RIKO mice (Fig2a) with decreased Flnb RNA half-life suggesting that loss of RBM47 promotes tumor growth via altering Flnb RNA stability (Fig2a). TCGA of 524 colorectal adenocarcinoma patients revealed a subset with reduced RBM47 and FLNB RNAs (Fig2b). FNLB mRNA was also decreased in CRC samples, correlating with RBM47 RNA expression (Fig2c). Conclusions: We identified 36 novel C-to-U RNA editing sites across 29 targets in Apc min/+ mice that exhibit potential roles in intestinal polyposis in mouse and human tissues with cofactor (RBM47) dependence influencing altered mRNA stability and tumorigenesis.

Figure 1. a. Venn diagram showing Apc min/+ -specific 36 APOBEC1-dependent newly identified editing sites. Out of 57 previously identified editing sites in WT C57b/6j intestine, 37 were also edited in both uninvolved mucosa (UM) and/or tumor of Apc min/+ mice. b. Gross morphology of polyps in jejunum of Apc min/+, Apc min/+ Apobec1–/–, Apc min/+ Apobec1 +/Tg and Apc min/+ RIKO mice.Plyp number and size are shown for each genotype. * p<0.05, *** p< 0.001, **** p<0.0001. c. Top. Flnb RNA editing frequency in individual tumors of 3 separate mice. Bottom. Editing frequency of Flnb RNA in pooled tumors (3-4) isolated from proximal, mid and distal intestine of 7 separate Apc min/+ mice.

Figure 2. a. Left, Flnb RNA is significantly downregulated in Apc min/+ RIKO uninvolved and tumor tissue. * p<0.05, *** p<0.001. Middle, FLNB expression is reduced in Apc min/+ RIKO compared to Apc min/+ mice. ACTIN is used as loading control (n=3 per genotype). Right, RNA decay assay showing reduced Flnb RNA half-life upon loss of RBM47. b. TCGA survey of 524 colorectal adenocarcinoma patients for RBM47 and FLNB showing a subset of patients with both reduced RBM47 and FLNB RNA (red square). c. Left, FLNB RNA is significantly downregulated in CRC patients * p<0.05 (n=14 uninvolved-tumors pairs). Right, positive correlation between FLNB and RBM47 RNAs in CRC uninvolved and tumor tissues, by Q-PCR analysis.