Interaction of Muc2 and Apc on Wnt Signaling and in Intestinal Tumorigenesis: Potential Role of Chronic Inflammation

Somatic mutations of the adenomatous polyposis coli (APC) gene are initiating events in the majority of sporadic colon cancers. A common characteristic of such tumors is reduction in the number of goblet cells that produce the mucin MUC2, the principal component of intestinal mucus. Consistent with these observations, we showed that Muc2 deficiency results in the spontaneous development of tumors along the entire gastrointestinal tract, independently of deregulated Wnt signaling. To dissect the complex interaction between Muc2 and Apc in intestinal tumorigenesis and to elucidate the mechanisms of tumor formation in Muc2^−/− mice, we crossed the Muc2^−/− mouse with two mouse models, Apc^1638N/+ and Apc^Min/+, each of which carries an inactivated Apc allele. The introduction of mutant Muc2 into Apc^1638N/+ and Apc^Min/+ mice greatly increased transformation induced by the Apc mutation and significantly shifted tumor development toward the colon as a function of Muc2 gene dosage. Furthermore, we showed that in compound double mutant mice, deregulation of Wnt signaling was the dominant mechanism of tumor formation. The increased tumor burden in the distal colon of Muc2/Apc double mutant mice was similar to the phenotype observed in Apc^Min/+ mice that are challenged to mount an inflammatory response, and consistent with this, gene expression profiles of epithelial cells from flat mucosa of Muc2-deficient mice suggested that Muc2 deficiency was associated with low levels of subclinical chronic inflammation. We hypothesize that Muc2^−/− tumors develop through an inflammation-related pathway that is distinct from and can complement mechanisms of tumorigenesis in Apc^+/− mice. [Cancer Res 2008;68(18):7313–22]

Muc2, expressed by goblet cells, is the most abundant secreted intestinal mucin, the protein component of the viscous-elastic mucus that protects the intestinal epithelium against mechanical and chemical insults (1). Altered MUC2 expression and/or glycosylation accompany intestinal pathologies, including inflammatory bowel disease (IBD) and colon cancer (2). The importance of MUC2 in intestinal homeostasis is reflected by alterations of cell proliferation, migration, and apoptosis in the mouse intestine upon genetic inactivation of the Muc2 gene (3). Increased proliferation and survival of the epithelial cells in Muc2^−/− mice may be a direct consequence of increased exposure of the cells to the luminal contents and may be an environment for tumor initiation and promotion. Indeed, Muc2-deficient mice develop small and large intestinal and rectal tumors (3). Therefore, loss of Muc2 can be an initiating event in intestinal tumorigenesis.

Intestinal tumorigenesis is most frequently initiated by mutations in APC, a component of the Wnt/β-catenin signaling pathway. APC mutations characterize tumors in familial adenomatous polyposis (FAP), an inherited form of colon cancer, and also 80% of sporadic colon cancers (4–6) affecting the activity of β-catenin/TCF4, resulting in altered intestinal homeostasis characterized by activation and repression of genes that regulate the orderly process of cell maturation along the crypt-lumen axis, leading to tumor development (7).

Consistent with observations that lack of Muc2 can be an independent initiating event, we showed that tumors from Muc2^−/− mice do not have alterations of Wnt/β-catenin/Tcf4 signaling (3). There is, however, evidence of interaction between APC and MUC2 in the early steps of tumorigenesis. For example, in tumors initiated by mutant APC, there is underrepresentation of goblet cells, the cell lineage that expresses MUC2, consistent with repression of HATH1 and CDX2, two positive regulators of MUC2 expression (8, 9), by activated Wnt signaling. Similarly, depletion of goblet cells and mucin reduction occurs in a subset of aberrant crypt foci (ACF; ref. 10), preneoplastic lesions in intestinal tumor development (11). Therefore, to dissect complex interactions between MUC2 deficiency and APC mutation and to elucidate mechanisms of tumor formation in Muc2^−/− mice, we crossed the Muc2^−/− mouse with two mouse models of Apc-initiated intestinal cancer: Apc^1638N/+ (12) and Apc^Min/+ (13). We report that introduction of the mutant Muc2 allele into these strains greatly exacerbated mutant Apc-initiated transformation as a function of mutant Muc2 gene dosage and shifted tumor incidence toward the colon, a phenotype similar to that of Apc^Min/+ mice challenged to mount an inflammatory response (14, 15). Moreover, intestinal epithelial cells of the flat mucosa of Muc2^−/− mice were characterized by a gene expression profile consistent with subclinical levels of chronic inflammation. Thus, we hypothesize that Muc2 deficiency establishes an inflammatory stimulus that modulates mutant Apc-initiated transformation and that Muc2^−/− tumors develop through an inflammation-related pathway distinct from, but interactive with, pathways recruited in tumor formation in Apc^+/− mice.

Mice. Muc2/Apc^1638N/+ mice were on the B6/129/Ola mixed background; compound double mutant Muc2/Apc^Min/+ mice were on a congenic C57BL/6 background. Details of procedures are provided in the supplementary data.

Western blot analysis. Cell extracts from intestinal polyps and flat mucosa were prepared by sonicating pulverized tissue in sample buffer. Intestinal epithelial cells were isolated from 3-month-old mice (three mice per genotype; Supplementary Data).

DNA and RNA isolation. Genomic DNA was isolated from frozen tissue using the DNeasy tissue kit (Qiagen). Total RNA was isolated from frozen, pulverized tissue, or purified intestinal epithelial cells, using Trizol (Invitrogen), treated with RQ1 Rnase-free DNaseI (Promega) for 15 min at room temperature.

Microarray analysis. Total RNA was purified from intestinal epithelial cells from the duodenum of age/gender matched Muc2^−/− and wild-type (wt) mice, three mice per genotype, at 3 and 6 months of age, and processed using Affimetrix protocols and Affymetrix mouse 430 arrays (Supplementary Data).

Quantitative reverse transcription–PCR. Total RNA was isolated from purified intestinal epithelial cells from duodenum of each of three Muc2^−/− and three wt 3-month-old, gender-matched mice. RNA was reverse transcribed into cDNA (SuperScriptIII reverse transcriptase; Invitrogen). Amplification used a 7900HT ABI instrument (Applied Biosystems) using SYBER Green Core reagent kit.

Samples were analyzed in duplicate; quantification of relative gene expression level used the standard curve method (User Bulletin #2; ABI), with values normalized to β-actin or ribosomal protein L41 (Rpl41). Primer sequences are in Supplementary Table S1.

In vitro transcription and translation assay. Amplification of DNA with primers specific for the wt Apc allele produced segment 1, spanning nucleotide 1991-5142 in exon 15 of the Apc gene (accession no. M88127). This was used to generate two overlapping fragments (segment 2, codons 677-1234; segment 3, codons 1100-1690), which were used to program the in vitro transcription and translation assay (IVTT) system (TNT T7 quick coupled transcription/translation system; Promega). Products were analyzed by SDS-PAGE and fluorography (16), and mutant clones were characterized by sequencing. Primer sequences are in Supplementary Table S1.

Loss of heterozygosity and microsatellite instability analysis. Ten nanograms of genomic DNA from tumors and normal tissue and serial dilutions (1:2 and 1:4) were tested in triplicate by genomic qPCR amplification of the wt and the mutant Apc allele in separate reactions. In each run, DNA from Apc^1638N/+ mouse liver was assessed to monitor consistency. Relative quantification of wt and mutant Apc allele was determined using the standard curve method (above). The ratio between mutant and wt Apc allele was determined for normal and tumor samples; loss of heterozygosity (LOH) was defined by a mutant/wt value above the maximum value from the compilation of normal tissue samples, as described (15).

Microsatellite instability (MSI) was analyzed by amplifying 50 ng of genomic DNA from paired normal and tumor samples using ³²P end-labeled primers, with PCR products analyzed by electrophoresis on denaturing 6% polyacrylamide gels and bands visualized using the Storm Imaging System or autoradiographic films. Analysis was of five dinucleotide repeats (D7Mit91, D17Mit123, D1Mit36, D15Mit93, D10Mit2) and two mononucleotide repeats (U12235 and a T24 tract in uPAR; ref. 17).

Immunohistochemistry. Sections (5 μm) from formalin-fixed, paraffin-embedded samples were analyzed (Supplementary Data). The list of antibodies is in the supplementary data.

The inactivated Muc2 allele exacerbates the tumor phenotype of the Apc^1638N/+ mouse by a gene dosage-dependent mechanism. Because tumors in Muc2^−/− mice develop through a Wnt-independent pathway, we crossed Muc2^−/− with Apc^1638N/+ mice, a model of intestinal cancer caused by a mutant Apc allele. Introduction of the Muc2 mutation, both in the heterozygous and homozygous state, accelerated tumor development, increasing both tumor incidence and multiplicity (Supplementary Fig. S1; Table 1). Tumors increased at 3 months and progressed with age. At 12 months, heterozygosity and nullizygosity for Muc2 caused almost a doubling and >3-fold increase, respectively, in tumor multiplicity. Muc2^+/− mice lack a detectable phenotype (3). Here, only one of eleven 52-week-old Muc2^+/− mouse had a single small intestinal tumor (Table 1). In contrast, none of the wt mice developed gastrointestinal tumors (not shown). Thus, Muc2 heterozygosity imparted a very low risk of tumors that was enhanced when combined with a stress stimulus, the Apc^1638N/+ background. That Muc2 and Apc mutations were at least additive for tumor development in a compound mutant mouse is a further indication that mutations at these loci operate through distinct mechanisms. In addition, the distribution of tumors was substantially altered; double mutant mice developed tumors along the entire colon, although there were more tumors in the left colon than in the right. An example of tumor histology is shown in Supplementary Fig. S2.

Whereas Apc^1638N/+ mice have a low penetrance tumor phenotype in the colon (three distal colonic tumors in 44 mice studied), ∼70% of Apc^Min/+ mouse develop about two tumors per mouse in the distal colon and rectum by 120 days. Therefore, we also crossed the Muc2^−/− mouse with the Apc^Min/+ mouse. Due to early mortality, sacrifice was at 75 days of age. The phenotype of double mutant Muc2^−/−; Apc^Min/+mice was qualitatively similar to that of Muc2^−/−; Apc^1638N/+ mice, although remarkably more severe (Table 2). There was a carpet of tumors in the distal colon of all compound mutant mice (Supplementary Fig. S3), with only a few tumors in the proximal colon and a significant increase in tumor number in the small intestine of Muc2^−/−; Apc^Min/+mice. For Muc2^+/−; Apc^Min/+ mice, survival was longer and mice were sacrificed at either 75 or 120 days of age. Similar to Muc2^+/−;Apc^1638N/+ mice, the Apc^Min/+ phenotype was affected by Muc2 haploinsufficiency, reflected in a less prominent, yet significant, increase of colonic tumors at 75 (Table 2) and 120 days (data not shown). In the small intestine, tumor burden increased at 75 days (Table 2) and became more pronounced in older mice (not shown). Thus, introduction of the Muc2 mutation modulated the tumor phenotype of mouse models carrying a mutant Apc allele, irrespective of the nature of the Apc mutation, and the extent of this modulation was Muc2 gene dosage dependent.

In Muc2^−/− mice, 50% of colonic tumors were carcinomas, but in Muc2^−/−; Apc^1638N/+ mice, only 17% (17 of 102) were carcinomas and 83% were adenomas. All Muc2^+/−; Apc^1638N/+ colonic tumors were adenomas. In Muc2^−/−; Apc^Min/+ colonic tumors, there was a continuum of proliferative lesions that included hyperplasia, dysplasia, and adenoma. Thus, although the colon of double mutant mice developed many more tumors, they rarely presented as carcinomas, possibly due to increased mortality of double mutant mice from tumor burden, also indicated by the shorter life span of Muc2^−/−; Apc^Min/+mice, thus precluding progression to carcinoma as a function of time.

Characterization of tumors of Muc2^−/−; Apc^1638N/+ mice, and compound double mutant mice. We first determined whether tumors from compound double mutant mice exhibited altered Wnt signaling, as in Apc^1638N/+ tumors, or lacked deregulated Wnt, as in Muc2^−/− tumors. Tumors from Muc2^+/−; Apc^1638N/+ and Muc2^−/−; Apc^1638N/+ mice exhibited increased accumulation of β-catenin in the cytoplasm and in the nucleus (Fig. 1A), suggesting that these tumors had an altered Apc pathway. This was confirmed by elevated levels of active (dephosphorylated) β-catenin in tumors of compound mutant mice compared with adjacent normal mucosa, similar to overexpression of active β-catenin in Apc^1638N/+ tumors. In contrast, there was little difference in levels of active β-catenin in Muc2^−/− tumors and adjacent normal mucosa (Fig. 1B). Quantitative data of active β-catenin relative to total β-catenin in tumors and adjacent normal tissue in Muc2^−/− and Muc2^−/−; Apc^1638N/+ mice are shown in Supplementary Fig. S4.

The fundamental role of cyclooxygenase 2 (Cox2) in intestinal tumorigenesis in the context of deregulated Wnt signaling is well established (18). Western blot analysis of paired tumor-normal samples showed robust elevation of Cox2 levels in tumors of Muc2^−/−, Muc2^+/−; Apc^1638N/+, Muc2^−/−; Apc^1638N/+, and Apc^1638N/+ mice (Fig. 1B). Immunohistochemical analysis confirmed that, as previously shown in the Apc^Min/+ and interleukin-10^−/− (IL-10^−/−) mice (19, 20), Cox2 positivity was detected in tumor infiltrating stromal cells, but not in the flat mucosa of the Muc2^−/−, Apc^1638N/+ mice (Supplementary Fig. S5). Thus, elevation of Cox2 expression in intestinal tumors is a common feature of these genotypes and is not linked to the etiology and earliest event of tumor initiation but may represent convergence on a common mechanism.

Muc2 gene dosage is linked to the molecular mechanisms of inactivation of the wt Apc allele. In tumors from Apc^1638N/+ mice, the wt Apc allele is lost in the great majority of cases, or, less frequently, inactivated by somatic mutations that produce a truncated Apc protein unable to drive β-catenin degradation, resulting in β-catenin accumulation, and nuclear translocation. Similarly, we observed nuclear accumulation of β-catenin in tumors from Apc^1638N/+ mice that were either heterozygous or homozygous for mutant Muc2 (Fig. 1A). We therefore ascertained the status of the wt Apc allele by in vitro transcription-translation (IVTT), interrogating Apc exon 15 spanning codons 664-1690, the region homologous to the mutation cluster region of human APC, which preferentially accumulates such mutations (16). In this assay, a chain terminating mutation yields a peptide smaller than that encoded by the wt allele. Representative examples of size fractionated IVTT generated polypeptides are shown in Supplementary Fig. S6, with results summarized in Table 3. None of the 30 Muc2^−/− tumors harbored mutant Apc polypeptides (Supplementary Fig. S6A), confirming that mutated Apc was not present in Muc2^−/− tumors, and consistent with the lack of nuclear β-catenin accumulation (above; ref. 3). However, 50% of the tumors (12 of 24; Table 3) from Muc2^+/−; Apc^1638N/+ mice had truncating Apc mutations, as shown by shorter peptides (Supplementary Fig. S6A and B, T7-12). All but one mutation were in segment 2 (T12 in Supplementary Fig. S6B). Surprisingly, <15% of Muc2^−/−; Apc^1638N/+ tumors had truncating Apc mutations (Table 3). This difference in frequency of truncating mutations in tumors between Muc2^+/−; Apc^1638N/+ and Muc2^−/−; Apc^1638N/+ mice was significant (P < 0.01, two-sided χ² test).

All PCR products that generated mutant peptides in the IVTT assay were cloned and sequenced; data are summarized in Table 3 and Supplementary Table S2. The majority of Apc mutations in the Muc2^+/−; Apc^1638N/+ tumors were base substitutions (8 of 12), with six of these transitions (6 of 12). Interestingly, of these six transitions, four were C-to-T transitions, and half occurred at CpG islands coding for Arg⁸⁵⁴, a mutational hotspot in compound mismatch repair (MMR)–deficient/Apc^1638N/+ mice (ref. 16; Supplementary Table S2). The remaining two base substitutions were G-to-T mutations at codon 866. The remaining four tumors from Muc2^+/−; Apc^1638N/+ mice with mutated Apc alleles harbored frame shift mutations due to insertions/deletions of a single nucleotide in mononucleotide repeats. In contrast, all Apc truncating mutations of the four Muc2^−/−; Apc^1638N/+ tumors were base substitutions, two of which were C-to-T transitions and two were G-to-T transversions (Supplementary Table S2). Thus, not only were there fewer Apc mutations in tumors from the Muc2^−/−; Apc^1638N/+ mice compared with tumors from the Muc2^+/−; Apc^1638N/+ mice, but the mutation spectrum differed between these genotypes.

The spectrum of mutations in the Muc2^+/−; Apc^1638N/+ tumors was reminiscent of that in tumors of compound double mutant MMR-deficient/Apc^1638N/+ mice (16). We, therefore, analyzed paired normal/tumor DNA from mice of the three different genotypes for the presence of MSI, diagnostic for MMR defects, using seven different markers (Materials and Methods). None of the tumors analyzed, including those characterized by the presence of truncating mutation in the wt Apc allele, displayed MSI (an example is shown in Supplementary Fig. S7). These data suggest that tumors from Muc2^−/− and compound double mutant Muc2; Apc^1638N/+ mice were not MSI, a form of genetic instability that can have a causative role in tumor development, despite the fact that heterozygous inactivation of Muc2 increases mutation frequency of the wt Apc allele in the compound mutant mice.

LOH is the most frequent mechanism of inactivation of the wt Apc allele in tumors of Apc^1638N/+ mice (21). Thus, we investigated whether tumors that did not show truncating mutation of Apc were instead characterized by LOH. We determined, by qPCR, the ratio between mutant and wt Apc alleles using primers that amplify only the mutant or the wt Apc allele in DNA from paired tumor and normal tissues, as well as from normal spleen, liver, and tails. Data were analyzed as described (15). Control DNAs from normal tissue displayed a median value of the ratio between mutant and wt allele of 1.12 and a maximal value of 1.35. Thus, a ratio of mutant/wt Apc allele of >1.35 was considered positive for LOH. Among tumors that had mutational inactivation of the wt Apc allele, we generally did not detect LOH (Fig. 2), with only 5 of 16 tumors showing LOH (31%; Table 4). However, LOH was not significantly enriched in the tumors that were not characterized by mutation in the wt Apc allele. Of 26 tumors analyzed that did not exhibit detectable Apc mutation, only 10 displayed LOH (38%). As controls, 83% (five of six) of tumors from Apc^1638N/+ mice exhibited LOH, in agreement with published data (21). Interestingly, when we analyzed the data as a function of Muc2 mutation, we found that the majority of tumors (65%) from Muc2^+/−; Apc^1638N/+ mice either had mutation or LOH of wt Apc. However, in tumors from Muc2^−/−; Apc^1638N/+ mice, only 36% (8 of 22) of tumors had either mutation or LOH of wt Apc, despite the fact that our molecular analyses showed that tumors in compound double mutant mice develop through deregulation of Wnt signaling (Fig. 1A and B and Supplementary Fig. S4).

Epithelial cells of Muc2^−/− flat mucosa display a transcriptional signature reflecting inflammation. The colonic tumor phenotype of compound double mutant Muc2/Apc mice is reminiscent of Apc^Min/+ mice treated with dextran sodium sulfate (DSS; ref. 14),⁷

a model of intestinal inflammation, and in compound Smad3^−/−; Apc^Min/+ mice in which inactivation of Smad3 triggers inflammation (15). This suggested that lack of Muc2 might provide an inflammatory stimulus. Therefore, to investigate whether absence of Muc2 causes low level, chronic, subclinical inflammation that alters intestinal homeostasis, we compared gene expression profiles of epithelial cells isolated from the duodenum of wt mice or Muc2^−/− mice, at 3 and 6 months of age. At 3 months, of 41,530 sequences assayed, the expression pattern of 951 (∼2%), corresponding to 669 distinct genes, were altered significantly in Muc2^−/− compared with wt mice, with more genes repressed (525) than induced (144) in the Muc2^−/− flat mucosa (partial list in Supplementary Table S3). The most highly down-regulated genes included members of the families of phase I and II detoxifying enzymes involved in xenobiotic metabolism, including the cytochrome P450 (Cyp) family (Cyp1a1, Cyp2a10, Cyp3a11, and Cyp4a10), greatly down-regulated in the small intestine of both 3-month-old and 6-month-old Muc2^−/− mice, as well as glutathione-S-transferases, UDP–glucoronosyl transferases, sulfotransferase, and Nqo1. Although these enzyme complexes are generally induced as a component of an adaptive defense response to environmental stress, they can be repressed in chronic inflammation (reviewed in ref. 22). In addition, genes of the antioxidant response were down-regulated, such as catalase and Glrx1, with the notable exception of glutathione reductase and glutathione peroxidase 2 (Gpx2) that are involved in the inactivation of hydroperoxides and regeneration of reduced glutathione (GSH); however, Gclm, the modifier subunit of glutamate-cysteine ligase, the rate limiting enzyme in synthesis of GSH, was significantly down-regulated. The repression of several of these genes was confirmed by quantitative reverse transcription–PCR (qRT-PCR; Fig. 3A).

Interestingly, and consistent with these data in the Muc2^−/− mice, it was recently shown (23) that in a rat model of IBD, there was a robust inhibition of several Cyp members, with evidence that endotoxins of commensal bacteria contribute to these effects. In this regard, a role of commensal bacteria in developing a reactive intestinal mucosa in Muc2^−/− mice can be inferred from elevation of several genes expressed by Paneth cells that have antibacterial activity, including angiogenin4 (Ang4), matrylisin (Mmp7), secretory phospholipase A2 (Pla2g2a and Pla2g5), and RegIIIγ (Fig. 3B), the latter recently shown to have specific bactericidal activity against Gram-positive bacteria (24). Moreover, levels of RegIIIγ are decreased in mice deficient for Retnlb/Relmβ, a member of the resistin family specifically expressed in intestinal goblet cells, implying positive regulation of RegIIIγ by Retnlb (25). Accordingly, we detected increased levels of Retnlb in Muc2^−/− mice (Fig. 3B). The importance of Relmβ in intestinal inflammation is underscored by demonstration that its inactivation protects mice from DSS-induced colitis (25). Additional members of the C-type lectin family expressed in Paneth cells were induced, such as Pap1 and Mbl2 (Fig. 3B). Furthermore, lipocalin 2, which regulates bacterial growth by iron sequestration (26), was greatly up-regulated in Muc2^−/− mucosa. Consistent with up-regulation of genes of the innate immune response, we detected elevated expression of two members of the CC chemokine family (Ccl6 and Ccl28) expressed by intestinal epithelial cells to promote antimicrobial immunity (27, 28), as well as Ifn-γ.

Importantly, some of these changes in the duodenum were also detected in epithelial cells of Muc2^−/− colonic flat mucosa. We found that 30% of the gene expression changes associated with the Muc2^−/− duodenum (14 of the 46 genes listed in Supplementary Table S3) overlap with similar changes in the colon of Muc2^−/− mice. Of these, Ang4 and Reg1 changed in opposite direction. Additionally, two genes, Intelectin a and Defensin5, that showed no changes in the duodenum of Muc2^−/− mice, were greatly down-regulated in the colon of Muc2^−/− mice. Interestingly, both genes are normally expressed in the small intestine providing defense against microbes. The partial overlap of the changes that we detected in duodenum and colon of Muc2^−/− mice most likely reflects distinct patterns of gene expression that characterizes these different segments of the normal intestinal tract.

Therefore, despite the absence of significant infiltration of inflammatory cells and of obvious inflammation-related histopathology under the conditions used in this study, these data suggest that the mucosa of Muc2^−/− mice reflects an inflammatory response to low levels of inflammation.

To better characterize inflammation in Muc2^−/− mice, we determined the basal level of expression of chemokines/cytokines in the supernatants of whole colon and duodenum tissue explants. Using an antibody array panel, we detected increased secretion of inflammatory cytokines in the supernatant of the colon and the duodenum of Muc2^−/− mice compared with wt counterparts, as shown in Fig. 3C. The Muc2^−/− duodenum was characterized by a robust response as illustrated by increased levels of Mig/CXCL9, MCP1, IL-1α, and IL-1ra, in addition to C5a and IL-16, the latter also induced in the colon, albeit at lower levels. IL-23, instead, was uniquely expressed in the Muc2^−/− colon. Of note, Mig/CXCL9, an Ifn-γ responsive gene, was secreted by Muc2^−/− duodenum that uniquely showed up-regulation of Ifn-γ mRNA (Fig. 3B); furthermore, both IL-16 and IL-23 have been specifically linked to Crohn's disease (29, 30). Thus, these data clearly show that a reactive intestinal mucosa is associated with Muc2 deficiency, reflecting a limited inflammatory response.

To confirm that the exacerbation of the tumor phenotype we observed in compound double mutant Muc2/Apc mice was due to an inflammatory stimulus provided by Muc2 deficiency, we investigated whether the pattern of gene expression of intestinal epithelial cells of compound double mutant Muc2^−/−; Apc^1638N/+ mice was similar to that of Muc2^−/− animals by qRT-PCR analysis of a subset of genes that were modulated in the Muc2^−/− flat mucosa (Fig. 3A and B). Supplementary Fig. S8 shows that all the genes we tested were similarly regulated in double mutant Muc2^−/−; Apc ^1638N/+ epithelial cells, as in Muc2^−/− mice. These data indicate that Muc2 deficiency in the Apc^1638N/+ background also results in a low-grade inflammation reflected by the gene expression profile and that this most likely contributes to the more severe tumor phenotype in double mutant mice.

We reported that genetic inactivation of the Muc2 gene causes spontaneous development of tumors along the entire gastrointestinal tract without deregulating Wnt signaling. However, complex interaction between APC, a component of Wnt signaling fundamental to colon cancer initiation, and the MUC2 gene is documented not only by data showing a reduction of goblet cells and the mucins they produce in tumors but that reduced representation of goblet cells characterizes a subset of ACF, called mucin-depleted foci that are precancerous lesions (10). Thus, to further dissect interactions between MUC2 deficiency and altered Wnt signaling and to elucidate mechanisms of tumor formation in Muc2^−/− mice, we crossed the Muc2^−/− mouse with two mouse models of Apc initiated intestinal cancer, Apc^1638N/+ and Apc^Min/+. We observed an exacerbated tumor phenotype, shown by accelerated kinetics and increased tumor frequency in compound double mutant mice in which the Muc2 mutation was present with either the Apc¹⁶³⁸ or Apc^Min allele. Most impressive, in compound Muc2/Apc mutant mice, there was a shift in tumor location toward the colon, which was particularly severe in Muc2^−/−; Apc^Min/+mice, characterized by a high density of tumors in the distal colon. The fewer proximal colon tumors support the hypothesis that the mouse distal colon is more permissive than the proximal for tumor formation (15, 31). Tumor frequency in the colon was strongly dependent on Muc2 gene dosage, both in compound double mutant Muc2^+/−; Apc^1638N/+ and Muc2^+/−; Apc^Min/+ (Tables 1 and 2).

Molecular analysis of tumors from mice of different genotypes showed that, contrary to tumors in Muc2^−/− mice, Wnt signaling was altered in tumors of double mutant mice similarly to mutant Apc alone, reflected by nuclear accumulation of β-catenin, a hallmark of activated Wnt signaling, and increased levels of active, nonphosphorylated β-catenin. These data strongly suggest that deregulation of the Wnt pathway is the dominant mechanism of tumor formation in compound double mutant Muc2/Apc mice. However, it would be expected that the mechanism of tumor development, irrespective of initiating insult, would converge on obligatory steps. In fact, we previously showed that c-Myc was elevated in Muc2^−/− tumors, as it is in mutant Apc initiated tumors (3), and here, we showed that Cox2 levels were induced in all tumors and that this expression was confined to the stromal compartment of the tumors. Thus, both c-Myc and Cox2 elevation seem to be obligatory steps for tumor formation regardless of etiology or differences in early events.

In tumors of Apc^1638N/+ mice, the wt Apc allele is most frequently inactivated by LOH (21). Increased frequency of inactivation by mutation, however, is detected in tumors of Apc^1638N/+ mice also deficient in components of the MMR system (16). Interestingly, we found that the molecular mechanism of inactivation of the wt Apc allele in tumors of compound double mutant mice is linked to Muc2 gene dosage. Half of the Muc2^+/−; Apc^1638N/+ tumors exhibited truncating mutations in the wt Apc allele, the majority being point mutations, 50% of which were C-to-T transitions, the remaining being single-nucleotide insertions or deletions. Increased point mutations and an enrichment of C-to-T transitions have been described in Msh6^−/− mice that are deficient in MMR. However, our studies in Muc2^−/− and Muc2/Apc^1638N/+ tumors showed that, regardless of mutational status of the Apc allele, tumors did not exhibit instability at the homopolymeric or microsatellite loci assayed. Therefore, our data strongly suggest that MSI is not an early event in tumor formation in these mice.

Tumors that did not have Apc mutations exhibited LOH, and these two mechanisms were usually mutually exclusive, with the exception of three tumors that showed both mutational inactivation and LOH. These three samples may have contained more than one tumor or may have been polyclonal (32).

In contrast, when the Muc2 mutation was homozygous in Muc2^−/−; Apc^1638N/+ mice, there was considerable reduction of tumors with mutational inactivation of Apc compared with Muc2^+/−; Apc^1638N/+ tumors, yet surprisingly 72% (13 of 18) of tumors that were Apc mutation–negative were also LOH-negative. However, these Muc2^−/−; Apc^1638N tumors exhibited deregulated Wnt signaling, suggesting that activation of β-catenin occurred either through inactivation of Apc by mutation at other sites in the gene, by epigenetic mechanisms or through alterations at another locus. As regards the latter, an obvious candidate, as a direct target, is β-catenin itself, observed in human tumors, albeit infrequently, to be mutated at potential phosphorylation sites through which Gsk3-β kinase targets β-catenin for degradation (33). However, we did not detect β-catenin mutations in DNA from tumors negative for both truncating mutation of Apc and LOH at hotspots in exon 3 identified in rodent intestinal tumors induced by AOM (34). There are, however, additional mechanisms that result in β-catenin stabilization (35, 36), and the extent to which they play a role in these mice remains under investigation.

A wealth of experimental evidence has established the importance of inflammation in cancer initiation and progression (37, 38), a phenotype that might be expected in the Muc2^−/− mice with a compromised mucous barrier. Indeed, Muc2^−/− mice displayed enhanced mucosal permeability (a 4-fold increase compared with wt mice; data not shown), a defect that has been detected preceding and predicting relapse in IBD patients and prior to the onset of chronic immune-mediated histopathology in some mouse models of IBD (39, 40). Furthermore, although the Muc2^−/− mouse on a C57BL/6 background exhibited no obvious intestinal inflammation in the barrier facility where our mice are housed, we recently reported that Muc2-deficient mice, on a congenic 129SV background, housed elsewhere, spontaneously developed early colitis that aggravates with age (41). We therefore hypothesize that Muc2 deficiency causes a low-grade inflammation important in tumorigenesis and in exacerbating tumor formation initiated by Apc mutation. Moreover, the substantial increase in colonic tumor burden in compound double mutant Muc2/Apc mice is reminiscent of the increased colon tumor phenotype reported for Apc^Min/+ mice treated with DSS that induces intestinal inflammation and in compound Smad3^−/−; Apc^Min mice in which inactivation of Smad3 triggers an inflammatory response (15). Consistent with this, we identified a gene expression pattern in the flat mucosa of Muc2^−/− mice characteristic of low level inflammation. Most important, one third of the gene expression changes in the duodenum of Muc2^−/− mice that are associated with inflammation are similarly changed in the colon, further supporting our hypothesis that lack of Muc2 generates low levels of inflammation. Changes included decreased expression of detoxifying phase I and II genes involved in drug biotransformation and elimination. The importance of these defense mechanisms in probability of tumor formation is underscored by the fact that many of these intestinal detoxification and glutathione biosynthetic enzymes are induced by phytochemicals, as well as chemoprotective pharmacologic agents, and, furthermore, that natural dietary compounds that induce phase I and II detoxifying genes, as well as antioxidant genes in the flat mucosa, decrease tumor formation in Apc^Min/+ mice (42). Thus, reduced levels in the normal intestinal tract of Muc2^−/− mice likely contribute to tumorigenesis by compromising the ability of intestinal cells to mount an effective response to endogenous and exogenous stresses.

In contrast, Gpx2 (an intestine specific selenoprotein) and glutathione reductase were robustly up-regulated in the epithelium of Muc2^−/− mice. These enzymes are induced during the antioxidant response (43). Thus, their up-regulation in the Muc2 mutant mouse suggests that the normal appearing mucosa in these mice is indeed under oxidative stress. Collectively, these data suggest that, in the flat mucosa of Muc2^−/− mice, there are perturbations in the detoxifying/antioxidant response resulting in increased exposure of epithelial cells to genotoxic insult that contributes to tumorigenesis.

The mucus barrier is the first line of defense that physically separates underlying epithelial cells and the intestinal microbiota, establishing a link between mucins, innate immunity, and inflammatory response (44). In agreement, in Muc2-deficient mice, we detected induction of genes of the innate immune system, the majority of which are expressed by Paneth cells (45), including up-regulation of secretory phospholipase groups IIA and V, Pla2g2a and Pla2g5, that share antibacteriocidal properties. Interestingly, the Pla2g2a gene, mutated in C57BL/6 mice, is a modifier of the Apc^Min phenotype and also a modifier of the Muc2^−/− tumor phenotype, as a functional Pla2g2a allele confers resistance to tumor development to Muc2^−/− mice,⁸

further supporting the hypothesis that impaired mucosal protection increases risk of tumor development. Moreover, these results indicate that stress to the mucosa may, at least in part, stem from increased exposure to microbial populations that occupy the intestinal lumen. In this regard, we documented a robust increase of several members of the C-type lectin family in Muc2^−/− mice, genes with well-established bacteriocidal effects (46), and elevation of Ccl6 and Ccl28, members of the CC chemokine family expressed by epithelial cells that promote antimicrobial immunity. Interestingly, despite the increased expression of antibacterial polypeptides and the increased mucosal permeability, we did not detect bacterial translocation to mesenteric lymph nodes,⁹ in agreement with the absence of histopathologic manifestations.

The role of mucins in the cross-talk that determines the inflammatory response is further emphasized by our data demonstrating increased release of inflammatory cytokines in ex vivo explants of whole colon and duodenum of Muc2^−/− compared with wt mice. Intriguingly, IL-23, shown to be linked to Crohn's disease by inducing a T-helper type 17 proinflammatory response (47, 48), was significantly enhanced in the colon of Muc2^−/− mice, possibly explaining the aggravated colonic phenotype of compound double mutant Muc2/Apc mice. Furthermore, our data showing the concomitant induction of IL-1α and its receptor antagonist IL-1ra suggest that the intestinal tract has complex regulatory mechanisms that maintain immune homeostasis and that the level of inflammation is linked to the extent to which the balance between inflammatory stimuli and the host compensatory response is subverted.

In conclusion, we emphasize that the Muc2 mutant mouse model, alone and in combination with an Apc mutation, has unique physiopathologic features useful for dissecting complex relationships among tumor development, environmental stresses on the mucosa, and chronic inflammation. In relation to the development of human colon cancer, we have noted the depletion of globlet cells and the mucins they produce in ACF and therefore that this can contribute to focal development and progression of disease. Interestingly, we found that Muc2 haploinsufficiency may impart a modest increased risk of tumor development, as illustrated by the detection of one tumor in the small intestine of a single 52-week-old mouse, a risk that becomes greatly exacerbated in the mutant Apc background. This is important because it suggests that variations in MUC2 could modulate tumor development by affecting the landscape of the intestinal tract. In this regard, it was reported that polymorphisms in the number of tandem repeat region (variable number of tandem repeats) of the MUC2 gene (49) that have the potential to contribute to modulating the protective barrier established by mucins were not linked to susceptibility to inflammation (50). In contrast, our data suggest that variation of MUC2 amount may contribute to the risk for colon cancer development.

S. Scherer: Employment, Hoffmann-La Roche Ltd. The other authors disclosed no potential conflicts of interest.

Grant support: NIH grants DK058245 and U54CA100926, Cancer Center grant P30-13330, and University of Minnesota Academic Health Center grant (R.T. Cormier).

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 D. Reynolds for the help in the generation and interpretation of the microsatellite instability data, S. Nasser for histopathology, and M. Lesser and E. Livote for statistical analysis.