Deletion of the AU-Rich RNA Binding Protein Apobec-1 Reduces Intestinal Tumor Burden in Apc^min Mice Free

Colorectal cancer is the third most prevalent cancer in the United States and the second leading cause of cancer-related deaths (1). Significant advances have been made in our understanding of the molecular genetic events that accompany development of colonic adenomas and their progression through dysplasia to frank adenocarcinoma (2). These include increased understanding of the cross talk between cyclooxygenase (Cox)-mediated prostaglandin (PG) production, specifically prostaglandin E₂ (PGE₂), and the Wnt/β-catenin/T-cell factor signaling pathways that regulate intestinal adenoma formation, particularly in the setting of mutations in the adenomatous polyposis coli (APC) gene, which accompany familial adenomatous polyposis and >80% of sporadic colorectal cancer. The interaction of these pathways is evidenced through studies where genetic deletion of Cox-2 in the Apc^min/+ background decreased intestinal polyp formation by ∼80% by preventing the increase in Cox-2 mRNA, protein, and PGE₂ characteristic of both murine and human intestinal adenomas (3, 4). With the recognition that long-term use of Cox inhibitors, such as aspirin or nonsteroidal anti-inflammatory drugs, decreases colonic adenoma formation and recurrence (5) and colorectal cancer in humans (6), the mechanisms that regulate this dominant genetic pathway have assumed increasing importance.

Cox-2 expression is regulated through multiple mechanisms including alterations in mRNA stability and translational control (7, 8). The 3′ untranslated region (UTR) of Cox-2 mRNA contains several clusters of AUUUA motifs characteristic of class II AU-rich elements, and studies have shown a role for RNA binding proteins including HuR and CUGBP2 in modulating Cox-2 mRNA stability and translation in vitro (9, 10). Recent studies have shown concordant overexpression of both HuR and Cox-2 in human colon cancer tissue, suggesting a role for RNA binding proteins in stabilizing Cox-2 mRNA expression and, in turn, contributing to the progression of colorectal cancer (11). We recently showed that the AU-rich RNA binding protein apobec-1 binds to the first 60 nucleotides of Cox-2 3′-UTR in vitro and stabilizes Cox-2 mRNA in intestinal epithelial cells following radiation injury (8). These studies further showed that the increase in intestinal Cox-2 mRNA expression following lipopolysaccharide pretreatment of irradiated wild-type mice was completely abrogated in apobec-1^−/− animals, suggesting that Cox-2 mRNA may be a target for apobec-1 interaction in vivo (8). Apobec-1 has previously been identified as an RNA binding protein (12, 13) with a high-affinity consensus binding site UUUN(A/U)U embedded in an AU-rich context (14). This consensus binding site has been found within a canonical destabilization element [UUAUU(A/U)(A/U)] located within the 3′-UTR of mRNAs known to be rapidly induced and regulated through alterations in stability (14, 15).

Based on these findings, we tested the hypothesis that apobec-1 deletion in the Apc^min/+ background would abrogate the increase in Cox-2 mRNA expression that predictably accompanies intestinal adenoma formation in this model. Here, we report that the AU-rich RNA binding protein apobec-1 exerts an important and potentially complex role in the development of intestinal adenomas in Apc^min/+ mice.

Animals. C57Bl/6J Apc^min/+ male mice were obtained from The Jackson Laboratory. Apobec-1^−/− mice were backcrossed for more than 10 generations onto the C57Bl/6J strain. Mice were fed a 10% fat diet (Harlan Teklad) and maintained as approved by the Animal Studies Committee of Washington University School of Medicine. Mice were sacrificed at 102 ± 3 days and intestinal polyps examined by an observer blinded to genotype (16). Intestines were fixed with 10% formalin (Sigma) and pinned for scoring by inspection with a Nikon SMZ800 dissecting microscope. Each pinned section was photographed using a Photometrics CooLSNAPcf camera (Imaging Processing Services, Inc.). The circumference of each section was determined and the size of individual polyps quantitated by measuring the width, length, and total area using Metavue software (Molecular Devices).

Immunohistochemical studies, bromodeoxyuridine labeling, and apoptotic index. Swiss rolls of the intestine were fixed in 10% formalin and paraffin embedded. Four-micron sections were used to assess histomorphology following staining with H&E. Flat lesions were identified where the base diameter was at least twice the height. Immunostaining of Cox-2 within small intestinal adenomas was done on formalin-fixed, paraffin-embedded samples. The slides were rehydrated and then microwaved for 15 min in 10 mmol/L citrate buffer (pH 6.0) before immersion with a 1:100 dilution of goat anti–Cox-2 antibody (Santa Cruz Biotechnology) followed by incubation with a secondary antigoat biotinylated antibody and streptavidin-peroxidase detection. Tumor proliferation was measured by bromodeoxyuridine (BrdUrd) incorporation. Mice were injected with BrdUrd (120 mg/kg, i.p.) and fluorodeoxyuridine (12 mg/kg; Sigma, MO) and sacrificed 2 h later. Paraffin sections of 4 μm were stained with a mouse monoclonal anti-BrdUrd monoclonal antibody (DAKO) followed by horseradish peroxidase–linked anti-mouse immunoglobulin G (IgG). BrdUrd-positive cells were counted at ×400 magnification and expressed as positive cells per tumor. Alternatively, mitotic figures were counted in H&E-stained sections of intestinal adenomas to obtain a mitotic index. At least 20 adenomas per mouse and 8 mice per genotype were evaluated. Immunostaining for terminal deoxyribonucleotidyl transferase–mediated dUTP nick end labeling (TUNEL) was done using an APOP Kit POD (Roche) as described by the manufacturer. The apoptotic index determined by TUNEL staining was confirmed against the number of pyknotic nuclei in stained serial sections.

Western blotting. Tissue extracts were prepared in denaturing sample buffer as previously described (15), separated by SDS-PAGE, and transferred onto a polyvinylidene difluoride membrane (Millipore). The membranes were probed sequentially with a 1:200 dilution of goat anti–Cox-2 (M19, Santa Cruz Biotechnology), a 1:500 dilution of rabbit anti–apobec-1, or a 1:50,000 dilution of rabbit anti–heat shock protein 40 (Hsp40; StressGen Biotechnologies). Proteins of interest were visualized with enhanced chemiluminescence reagents (GE Healthcare) and scanned using ImageQuant software (Amersham Biosciences).

PGE₂ analysis. Normal tissue and tumor samples from individual (five per genotype) animals were excised individually and flash frozen. Cleared homogenates were prepared (17) and analyzed using a PGE₂ monoclonal immunoassay kit (Cayman Chemical).

Adenovirus apobec-1 experiments. HCA-7 colon cancer cells (generously provided by S. Kirkland, Imperial College London, United Kingdom) were plated (18) onto six-well dishes and infected with increasing concentrations of adenoviral particles encoding either rat apobec-1 (19) or β-galactosidase (Gene Transfer Vector Core, University of Iowa, Iowa City, IA) diluted into serum-free medium. The next day, medium was removed and 2 mL of fresh complete DMEM (Cellgro, Mediatech, Inc.) were added to the cells for another 24 h. Following overnight culture, the cells were treated with complete growth medium supplemented with actinomycin D (5 μg/mL; Sigma; ref. 14). Total RNA was isolated at the indicated intervals and mRNA was quantified by real-time PCR using SYBR Green. All values were standardized to 18S rRNA.

Real-time PCR primers. The primers used were as follows: C-myc, 5′-AGCAACAACCGCAAGTGCT-3′ and 5′-GTGTCCGCCTCTTGTCGTT-3′; Cox-1, 5′-CCAGAACCAGGGTGTCTGTGT-3′ and 5′-GTAGCCCGTGCGAGTACAATC-3′; Cox-2, 5′-GGTGTCCCTTCACTTCTTTCAATG-3′ and 5′-TCTGGAGTGGGAGGCACTTG-3′; peroxisome proliferator–activated receptor (PPARδ); 5′-CAACGCACCCTTTGTCATCC-3′ and 5′-TTCTCTGCCTGCCACAGTGT-3′; EP2, 5′-GCAACATCAGCGTTATCCTCAA-3′ and 5′-AATCCGCAGCGGCTTCTT-3′; EP4, 5′-GCCCGGGAGTTAAAGGAGAT-3′ and 5′-TTTCAGTCAGGTCTGGCAGGTA-3′; tumor necrosis factor α (TNFα), 5′-GACCCTCACACTCAGATCATCTTCT-3′ and 5′-CCACTTGGTGGTTTGCTACGA-3′; vascular endothelial growth factor (VEGF), 5′-TGTGCAGGCTGCTGTAACG-3′ and 5′-GCATGATCTGCATGGTGATGTT-3′; granulocyte macrophage colony-stimulating factor (GM-CSF), 5′-GAAGCATGTAGAGGCCATCAAAG-3′ and 5′-CTTCTACCTCTTCATTCAACGTGACA-3′; epidermal growth factor receptor (EGFR), 5′-TCATGCGAAGACGTCACATTG-3′ and 5′-GAGGTTCCACGAGCTCTCTCTCT-3′; 18S, 5′-CGGCTACCACATCCAAGGAA-3′ and 5′-GCTGGAATTACCGCGGCT-3′.

Statistical analysis. Statistical analysis was done using GraphPad Prism 4 software (GraphPad Software, Inc.). In most cases, Student's t test was used to determine P values. For those experiments in which the data did not follow a Gaussian distribution, P values were determined with the Mann-Whitney U test. All values are reported as mean ± SE.

Apobec-1 deletion reduces small intestinal adenoma burden in the Apc^min/+ background. Apc^min/+ mice crossed into apobec-1^−/− mice showed a 64% decrease in small intestinal adenoma incidence (Fig. 1A), with a regional decrease evident in all portions of the small intestine (Fig. 1B). The proximal to distal gradient of adenoma distribution seen in Apc^min/+ mice was abrogated in the compound Apc^min/+ apobec-1^−/− strain (Fig. 1B). In addition, there was a decrease in polyp area per section per mouse from proximal to distal intestine in the compound Apc^min/+ apobec-1^−/− mice (i.e., fewer, smaller polyps), consistent with the suggestion of reduced tumor initiation (Fig. 1C). One carcinoma in situ was observed in 98 adenomas examined, but in keeping with previous observations in Apc^min/+ mice, the majority of tumors in both groups were adenomas with no high-grade dysplasia or frank carcinoma (20). However, Apc^min/+ apobec-1^−/− mice showed a significantly higher proportion (22%) of flat intestinal polyps compared with Apc^min/+ mice (7%; Fig. 1D). This shift in morphologic pattern is similar to that reported in adenomatous polyps from other compound lines of Apc^min/+ mice, including crosses into the Cox-2 knockout, EP₂, or EP₄ receptor knockout background and also in rofecoxib-treated Apc^Δ716 mice (21–23). Examination of proliferation rates and mitotic indices revealed a significant decrease in compound Apc^min/+ apobec-1^−/− mice (Fig. 2A and B), which was accompanied by a significant increase in apoptosis (Fig. 2C). These findings together suggest that eliminating apobec-1 expression alters the abundance, morphology, and growth characteristics of intestinal adenomas in Apc^min/+ mice.

As previously noted (21), the number of colonic polyps was generally <2 per mouse, with no significant differences noted between genotypes. Approximately 35% of colon polyps from both genotypes analyzed exhibited carcinoma in situ. However, there were no significant differences in either the histologic features or size of polyps between Apc^min/+ and Apc^min/+ apobec-1^−/− mice (data not shown).

Apobec-1 deletion suppresses the induction of Cox-2 gene expression and PGE₂ production in the Apc^min/+ background. There was a ∼10-fold increase in Cox-2 mRNA abundance in adenomas from the parental Apc^min/+ strain, which was abrogated in the compound Apc^min/+ apobec-1^−/− mice (Fig. 3A,, left). We recently showed that apobec-1 binds to the 3′-UTR of Cox-2 mRNA in vitro and stabilizes expression of Cox-2 mRNA in a heterologous cell system (8). Herein we show that Cox-2 mRNA can be recovered by immunoprecipitation of adenoma extracts (from Apc^min/+ mice) using anti–apobec-1 IgG (Fig. 3A,, right), suggesting that apobec-1 binds Cox-2 mRNA in vivo. Cox-2 protein abundance was also reduced in adenomas from the compound Apc^min/+ apobec-1^−/− mice compared with adenomas from parental Apc^min/+ mice (Fig. 3B). Immunohistochemical staining of Cox-2 revealed a distribution in both epithelial and subepithelial lamina propria compartments in tumor-bearing samples from Apc^min/+ mice (Fig. 3C). There was a qualitative decrease in Cox-2 immunoreactivity in the compound Apc^min/+ apobec-1^−/− mice and staining appeared to be confined to the subepithelial compartment (Fig. 3C). These variables of decreased Cox-2 expression in the compound Apc^min/+ apobec-1^−/− mice were accompanied by a reduction in tissue PGE₂ levels (Fig. 3D). Taken together, these findings suggest that apobec-1 deletion decreases Cox-2 mRNA and protein expression, which in turn abrogates signaling through prostaglandin-dependent pathways, resulting in suppressed intestinal adenoma initiation and progression.

Apobec-1 expression alters Cox-2 mRNA stability in human colorectal cancer cells. To establish more directly that apobec-1 expression influences Cox-2 gene expression in a relevant target cell, we undertook an adenovirus-mediated gain-of-function experiment in HCA-7 human colon cancer cells, a well-differentiated adenocarcinoma cell line in which Cox-2 is expressed but endogenous apobec-1 expression is below detection limits (Fig. 4A; refs. 18, 24). Adenovirus-apobec-1 (Ad-apobec-1) infection at increasing doses into HCA-7 cells was accompanied by a progressive increase in Cox-2 protein expression (Fig. 4A). We next undertook mRNA turnover studies to pursue the mechanism for this increase in Cox-2 expression. HCA-7 cells were infected with either Ad-apobec-1 or a β-galactosidase control adenovirus and Cox-2 mRNA turnover was examined following actinomycin D–induced transcriptional arrest. The results indicate that Ad-apobec-1 infection is accompanied by stabilization of Cox-2 mRNA, with its half-life increasing from ∼83 to ∼120 min (Fig. 4B). These findings show a plausible biological interaction between this candidate RNA binding protein and a putative target within colon cancer cells, where Cox-2 and apobec-1 genes are known to be coexpressed (25, 26).

Alterations in mRNA expression of other potential apobec-1 targets. Apobec-1 is an AU-rich RNA binding protein with a consensus binding site motif UUUN[A/U]U, which is present not only within the region immediately downstream of the edited C of its canonical target apolipoprotein B (apoB) but also in the 3′-UTR of other transcripts (Fig. 5A). Many of these candidate target mRNAs showed decreased abundance in adenoma tissue from the compound Apc^min/+ apobec-1^−/− mice (Fig. 5B–D). These findings raise the possibility that other targets of apobec-1, beyond Cox-2 mRNA, may be regulated as a result of deletion of this AU-rich RNA binding protein. It is tempting to speculate that the effects of apobec-1 deletion might involve altered stability of these other target mRNAs in the setting of intestinal adenoma development, but this remains to be formally tested.

A further feature of these findings concerns the function of apobec-1, beyond the site-specific deamination of its canonical target, apoB_C6666 (27). Apobec-1 has been shown to mediate site-specific C-to-U RNA editing of the tumor suppressor gene NF1 within tumors of patients with neurofibromatosis (28, 29) and, more recently, to exhibit deoxycytidine deamination activity toward sequences within viral and bacterial DNA, findings that point to a broad substrate activity profile whose origins may reside in an evolutionarily conserved function in innate immunity (30, 31). These findings raised the possibility that an effect of apobec-1 on tumor initiation or progression might be associated with alterations in C-to-U (or dC-to-dT) editing of one or more of the targets to which apobec-1 binds. However, sequencing of Cox-2 mRNA immunoprecipitated with apobec-1 from adenomas of Apc^min/+ mice revealed no C-to-U alterations (data not shown). In addition, we examined dC hotspots in p53, including those recently identified to exhibit site-specific deamination preferences in database searches (32). None of the sites examined revealed evidence of C-to-U changes in p53 RNA from adenomatous tissue of Apc^min/+ mice, although one of the four templates examined in an in vitro assay (33, 34) revealed low levels (∼7%) of dC deamination (Supplementary data S1). These findings raise the possibility that under situations where apobec-1 is overexpressed, such as transgenic overexpression where dysplasia and carcinoma are recognized phenotypes (35), promiscuous editing of tumor suppressor genes and/or stabilization of tumor promoting transcripts (such as c-myc mRNA) may occur (14, 36, 37). That said, we found no evidence for apobec-1 protein overexpression in extracts prepared from adenomas of Apc^min/+ mice compared with control isogenic mouse small intestine (data not shown).

Our findings suggest an unexpected role for apobec-1, an RNA-specific cytidine deaminase that, until recently, was thought to function exclusively in the context of C-to-U mRNA editing of the apoB and NF1 transcripts. Our new findings point to an important and unanticipated role for apobec-1 in intestinal tumorigenesis, which conceivably involves mediating alterations in mRNA stability of AU-rich targets including Cox-2. The cumulative evidence from previous studies (8) as well as the current in vivo immunoprecipitation experiments and the gain-of-function approaches with Ad-apobec-1 expression support the proposal that apobec-1 functions to bind and stabilize Cox-2 mRNA and thereby promotes expression of Cox-2 protein. In the absence of apobec-1, the stability of Cox-2 mRNA (as well as other AU-rich RNA candidate targets) is decreased, leading to reduced expression of Cox-2 protein and, in turn, decreased PGE₂ production. These changes were predicted to abrogate adenoma formation in the Apc^min/+ model of intestinal tumorigenesis, and the current findings in the compound Apc^min/+ apobec-1^−/− mice support this prediction while raising additional considerations that merit further discussion.

The current findings show that apobec-1 deletion in the Apc^min/+ background is associated with reduced expression of Cox-2 mRNA and protein, as well as reduced PGE₂ levels in adenomas of the compound Apc^min/+ apobec-1^−/− mice. An unanswered question, however, concerns the cell-specific compartment(s) in which this modulation occurs. The expression of endogenous apobec-1 protein in murine tissues is extremely low, precluding unambiguous detection using immunohistochemical approaches (38, 39). However, apobec-1 mRNA is detectable in a wide range of rodent tissues (40), suggesting that functional expression of this RNA binding protein in the intestine is not restricted to the enterocyte-specific lineage of its canonical target, apoB mRNA. By contrast, the expression of Cox-2 is readily detectable in both intestinal epithelial cells (41) and subepithelial lamina propria cells (23, 42, 43). Accordingly, it is formally possible that the observed effects of apobec-1 deletion on Cox-2 expression may reflect changes that take place in either the epithelial or subepithelial compartments. Further exploration of the effects of apobec-1 deletion in modulating Cox-2 expression will require the development of additional lines of deletor mice to localize the cellular populations in which apobec-1-RNA interactions contribute to the observed phenotype.

Another consideration with regard to the pathways involved in the abrogation of polyposis in the compound Apc^min/+ apobec-1^−/− mice concerns the complexity of downstream events following disruption of Cox-2 expression and decreased PGE₂ signaling. For example, studies have shown that either genetic disruption or pharmacologic inhibition of PGE₂ receptors, specifically EP2 and EP4 (23, 44), not only abrogated small intestinal polyposis in Apc^min/+ mice but also revealed an important feed-forward loop in which PGE₂ itself stimulated the expression of Cox-2 as a result of signals that are transduced through one or both EP receptors. The 3′-UTR of EP4 contains an apobec-1 consensus binding site embedded within an AU-rich context (Fig. 5A), and our findings revealed a decrease in EP4 mRNA abundance in adenomas from Apc^min/+ apobec-1^−/− mice. These findings raise the possibility that the decrease observed in Cox-2 gene expression in the compound Apc^min/+ apobec-1^−/− mice may, in part, reflect a decrease in EP4 expression. Along these lines, there was a 42% reduction in intestinal polyp burden in EP4^−/− mice crossed into the Apc^Δ716 background (44) and an accompanying shift toward flat polyp morphology, which are features shared with both Cox-2^−/− Apc^Δ716 mice (21) as well as the compound Apc^min/+ apobec-1^−/− mice reported in our study. The magnitude of reduction in the polyposis burden in compound Apc^min/+ apobec-1^−/− mice (64%) was less than observed in the Cox-2^−/− Apc^Δ716 mice reported earlier (84%; ref. 21), suggesting that there may be additional considerations with regard to the pathways involved, particularly because Cox-2 expression was decreased, but not eliminated, in adenomas from compound Apc^min/+ apobec-1^−/− mice. It would be of interest to examine the combinatorial effects of deletion of both Cox-2 and apobec-1 genes in the Apc^min/+ background to examine the possibility that targets of apobec-1 other than Cox-2 might be altered.

A related consideration is that although there was a striking reduction in adenoma burden in the compound Apc^min/+ apobec-1^−/− mice, it is not possible to conclude that this effect is mediated exclusively through Cox-2–dependent mechanisms. For example, there was a reduction in mRNA abundance in several other AU-rich targets, including EP4 (noted above) as well as PPARδ, EGFR, and c-myc (Fig. 5). A decrease in mRNA expression of any of these targets could plausibly impose a dominant loss-of-function effect on the polyposis phenotype observed. For example, studies using EP4-specific agonists showed enhanced proliferation of HCA-7 human colorectal cancer cells whereas administration of a specific EP4 antagonist reduced the formation of aberrant crypt foci and abrogated tumor burden in Apc^min/+ mice (23). It is possible therefore that apobec-1 deletion works, in part, to reduce the tumor burden in the Apc^min/+ background through effects on the expression of one or more prostaglandin receptors. In addition, it is known that PGE₂ stimulation via EP4 activates downstream signaling events through mechanisms that include EGFR phosphorylation (45). The current findings show that EGFR mRNA abundance was significantly reduced in tumors from compound Apc^min/+ apobec-1^−/− mice. Accordingly, it is not unreasonable to predict that reduced expression of EP4 would result in reduced activation of EGFR, and that this effect, coupled with the reduced expression of EGFR, would lead to a synergistic reduction in the downstream signaling events following PGE₂ production. Further studies will be required to resolve these possible considerations. These possible interactions become particularly relevant in relation to the effects on c-myc because recent findings indicate that conditional deletion of c-myc in the Apc^min/+ background, in fact, rescues the loss-of-function phenotype associated with Apc loss (46). Accordingly, the effects of apobec-1 deletion on the expression of many of these targets might conceivably work in a combinatorial manner that magnifies or reduces the net effects on the observed phenotype depending on the cell-specific context.

Previous studies showed that transgenic overexpression of apobec-1 in the livers of rabbits and mice resulted in dysplasia and malignant tumor development in association with extensive C-to-U RNA editing of a developmentally regulated target, NAT1 (36). It is worth pointing out that apobec-1 is a member of a multigene family of nucleoside deaminases that includes activation-induced deaminase (47) as well as other APOBEC-related genes with documented roles in retroviral restriction and innate immunity (48). Forced transgenic overexpression of activation-induced deaminase in mice (49) led to the development of dysplasia and malignant tumors in association with mutations in known tumor suppressor genes. More recently, infection with Helicobacter pylori has been shown to up-regulate the expression of activation-induced deaminase in human gastric epithelium in association with progressive accumulation of p53 mutations, although it remains unclear on whether the mechanisms of this association include alterations at the of p53 DNA or RNA level (50). We were unable to find evidence for mutations at the known C-to-T hotspots in p53 in adenomas from (apobec-1 wild-type) Apc^min/+ mice, but it is possible that examination at later time points may be more informative because these changes tend to be acquired late in the process of adenoma progression. Alternatively, it is quite possible that overexpression of apobec-1 would be required for such promiscuous C-to-U RNA editing events. We have previously shown that apobec-1 overexpression occurs in human colorectal cancer and the precedent certainly exists for a gain-of-function role in tumor promotion (26, 46). Along these lines, it would be of interest to examine some of the chemical models of murine colon carcinogenesis in some of the genetic strains outlined above to determine if deletion of apobec-1 has a role in the acquisition of somatic mutations known to occur in this setting.

A final consideration in discussing the possible effect of these findings concerns the implications for colorectal cancer pathogenesis in humans. As alluded to in the introduction, several lines of evidence suggest an important role for the Cox-2-PGE₂ signaling pathway, most significantly the results from randomized trials that have revealed chemopreventive effects of regular aspirin use in patients whose tumors overexpress Cox-2 (6). These findings, coupled with the results from the current study, raise the intriguing possibility that genetic interactions that reduce Cox-2 expression may be relevant to human colorectal cancer pathogenesis and prevention. In this regard, it will be particularly important to validate the cell-specific patterns of expression of both Cox-2 as well as putative modifier genes, and such analyses merit serious consideration. With this in mind, the current findings illustrate in principle the possibility of therapeutic modulation of Cox-2 gene expression in trans through effects mediated by targeted alterations in AU-rich RNA binding proteins.

Grant support: NIH grants HL-38180, DK-56260, DDRCC DK-52574 (Morphology Core; to N.O. Davidson), T32 DK-07130 (J.O. Henderson), DK-61261, DK-46122 (D.C. Rubin), and DK-64798 (R.D. Newberry).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank Dr. Terry Riehl for assistance in the PGE₂ assays, Keely McDonald for aid in the preparation of intestinal cell populations, Jonathan Lake and Connie Wang for help in the statistical analysis of polyp formation and development. Ad-apobec-1 was amplified and purified by the Vector Core of the University of North Carolina DDRCC, supported by grant DK 34987.