Progressive Changes in Adherens Junction Structure during Intestinal Adenoma Formation in Apc Mutant Mice*

The C57BL/6J-Min/+ (Min/+) mouse bears a mutantApc gene and therefore is an important in vivomodel of intestinal tumorigenesis. Min/+ mice develop adenomas that exhibit loss of the wild-type Apc allele (Apc Min/−). Previously, we found that histologically normal enterocytes bearing a truncated Apc protein (Apc Min/+) migrated more slowly in vivo than enterocytes with either wild-type Apc (Apc +/+) or with heterozygous loss of Apc protein (Apc 1638N). To study this phenotype further, we determined the effect of the Apc Min mutation upon cell-cell adhesion by examining the components of the adherens junction (AJ). We observed a reduced association between E-cadherin and β-catenin in Apc Min/+enterocytes. Subcellular fractionation of proteins fromApc +/+, Apc Min/+, andApc Min/− intestinal tissues revealed a cytoplasmic localization of intact E-cadherin only inApc Min/+, suggesting E-cadherin internalization in these enterocytes. β-Catenin tyrosine phosphorylation was also increased in Apc Min/+enterocytes, consistent with its dissociation from E-cadherin. Furthermore, Apc Min/+ enterocytes showed a decreased association between β-catenin and receptor protein-tyrosine phosphatase β/ζ (RPTPβ/ζ), andApc Min/− cells demonstrated an association between β-catenin and receptor protein-tyrosine phosphatase γ. In contrast to the Apc Min/+ enterocytes,Apc Min/− adenomas displayed increased expression and association of E-cadherin, β-catenin, and α-catenin relative to Apc +/+ controls. These data show that Apc plays a role in regulating adherens junction structure and function in the intestine. In addition, discovery of these effects in initiated but histologically normal tissue (Apc Min/+) defines a pre-adenoma stage of tumorigenesis in the intestinal mucosa.

The adenomatous polyposis coli (APC) 1 protein regulates intestinal cell growth. Loss of APC protein by germ line muta-tion causes familial adenomatous polyposis, an autosomal dominant cancer syndrome characterized by thousands of lower intestinal adenomas. APC loss is also an initiating event in most sporadic colorectal cancer (1,2). The best characterized function of APC is its association with a multiprotein complex that facilitates the degradation of the oncoprotein, ␤-catenin (3,4). APC binds to free cytosolic ␤-catenin together with the scaffolding protein, axin, and glycogen synthase kinase 3␤ (4,5). This kinase phosphorylates ␤-catenin on serine and threonine residues in a domain of the N terminus. These modifications, in turn, initiate the degradation of ␤-catenin by the ubiquitin-dependent proteasome (6). Missense mutations in the CTNNB1 gene that eliminate the glycogen synthase kinase 3␤ substrate residues of ␤-catenin permit tumorigenesis without APC mutation by stabilizing the protein and allowing its downstream signaling activities (7)(8)(9)(10). In tumor cells bearing mutant ␤-catenin or APC proteins, free cytosolic ␤-catenin escapes degradation and binds to Tcf-4, producing a transcriptional regulator that is capable of inducing the expression of growth-promoting genes such as c-Myc and cyclin D1 (8,(11)(12)(13).
The largest cellular pool of ␤-catenin is membrane-associated and engaged in adhesion-mediated signaling via cell-cell contact at the AJ (14). The AJ is a highly dynamic structure composed of the transmembrane protein, E-cadherin, and several members of the catenin family, including ␣-, ␤-, and ␥-catenin, and p120 ctn (reviewed in 15). When bound to E-cadherin, ␤-catenin links adhesion complexes to the actin cytoskeleton in ways that maintain epithelial cell polarity, preserve barrier function, and facilitate cell migration (15). Studies examining the localization of wild-type and cancer-associated truncated APC proteins suggest the involvement of this tumor suppressor in AJ function (reviewed in Ref. 16). By altering the availability of ␤-catenin for engagement with AJ constituent proteins, APC may indirectly modulate epithelial cell adhesion (17). APC was found co-localized with the actin cytoskeleton and with AJs in the epithelial cells of Drosophila embryos consistent with a role in regulating epithelial cell-cell contacts (18,19). Immunohistochemical studies also showed that APC resides in the lateral cytoplasm of intestinal epithelial cells, in regions containing the catenins (20). Tumor-associated APC mutations generally cause chain termination and result in expression of truncated proteins lacking the C terminus (21). By immunofluorescence microscopy, both full-length and truncated APC proteins were associated with the membrane at the apical borders of murine intestinal cells (22), suggesting that both forms of the protein are available to modulate adhesive interactions.
Epithelial homeostasis depends on the dynamic regulation of cadherin-catenin adhesion complexes at the AJ. The intestinal epithelium is normally maintained by the robust proliferation of stem cells situated in crypts. The progeny of stem cells differentiate as they migrate to the tips of the villi, where they eventually senesce and are exfoliated into the intestinal lumen. To mediate the processes of embryonic development, wound healing, and crypt-villus migration, the AJ must be efficiently disassembled and reassembled (reviewed in Ref. 23). The importance of the AJ to epithelial homeostasis is illustrated by the observation that loss of AJ structure correlates with tumor formation, invasion, and metastasis (24 -26).
Targeted gene disruption in mice shows that ␤-catenin is essential for maintenance of AJ activity, as this protein is required for the structural organization of the embryonic ectoderm (27). A high expression of E-cadherin reduces the nuclear localization and transcriptional potential of ␤-catenin (14,28). The interactions of ␤-catenin and E-cadherin, and consequently downstream signal transduction, are also regulated by tyrosine phosphorylation of AJ constituents including ␤-catenin (29 -31). ␤-Catenin can be phosphorylated via c-Src family tyrosine kinases (SFKs) such as Fer (32) or epidermal growth factor receptor (EGFR) (33). In addition, receptor protein-tyrosine phosphatases (RPTPs) can dephosphorylate ␤-catenin, thus promoting cell-cell adhesion and migration (34). The consequence of ␤-catenin tyrosine phosphorylation is its dissociation from the AJ (14,33,35) which, in turn, can cause loss of AJ structure (36,37). Under certain conditions of AJ loss, E-cadherin may be internalized and returned to the membrane of epithelial cells as adhesion contacts re-form (36,37).
The importance of appropriate adhesive interactions to optimal enterocyte migration was demonstrated by the targeted expression of E-cadherin in the murine small intestine (38,39). Previous studies in our laboratory showed that Apc mutation also affects enterocyte migration in the intestinal mucosa of Min/ϩ mice. The germ line of the Min/ϩ mouse contains a chain terminating mutation in Apc, and Min/ϩ mice develop multiple intestinal adenomas that exhibit loss of the wild-type Apc allele (Apc Min/Ϫ ). We found that Min/ϩ non-tumor enterocytes (Apc Min/ϩ ) displayed an ϳ25% reduced migration rate when compared with the Apc ϩ/ϩ littermate controls and the functionally hemizygous Apc 1638N enterocytes (40). This result suggested Apc genotype-dependent alterations in intestinal cell-cell adhesion. To characterize further the effect of the Apc Min allele upon enterocyte adhesion, we examined the changes in AJ structure that occur during the progression from wild-type (Apc ϩ/ϩ ) to slowly migrating Apc Min/ϩ enterocytes and to adenoma cells (Apc Min/Ϫ ). We extracted enterocytes from the mouse small intestine using a method that preserves cellcell and cell-ECM contacts and therefore provides an accurate evaluation of the in vivo status of adhesion complexes (41).
Here we show that diminished ␤-catenin-E-cadherin association and internalization of E-cadherin occurs in the non-tumor enterocytes from Min/ϩ mice when compared with their wildtype littermates and to Apc Min/Ϫ adenomas. These changes suggest that the Min mutation alters the membrane-associated pool of ␤-catenin in adult animals in ways that reduce cell-cell adhesion and slow enterocyte migration but that nevertheless maintain the histologically normal appearance of the mucosa. The data further support the view that the Min mutation yields a dominant negative effect on Apc function.
Animal Maintenance and Enterocyte Isolation-Female Apc ϩ/ϩ and Apc Min/ϩ mice were fed AIN-76A diet and tap water ad libidum. Similar growth and food intake occurred in the two groups. At 110 days of age all mice were euthanized by CO 2 inhalation, and their intestinal tracts were removed, opened, and washed with cold phosphate-buffered saline (PBS). Tumors were excised, pooled, and frozen in liquid N 2 . Enterocytes from the ileum were isolated in the following manner. Sections of proximal small intestine were opened longitudinally, washed with PBS, and inspected for evidence of tumor involvement by examination under ϫ 3 magnification. Enterocytes were removed from 1-to 2-cm segments of tumor-free ileum by lightly scraping the mucosal surface with the edge of a microscope slide. The material obtained by scraping was then washed in cold PBS, and the resulting pellet was frozen by immersion in liquid nitrogen. Frozen cell pellets were stored at Ϫ70°C until the time of assay. Enterocytes prepared in this manner also contained some lamina propria and fibroblasts, although the absorptive enterocytes and goblet cells of villi substantially outnumbered these cell types (Fig. 1). Serial sections underestimate the ratio between enterocytes and stromal cells, which were shown by three-dimensional anal- yses to exceed other cell types by Ͼ100-fold (42,43). These preparations, therefore, provided a favorable signal-to-noise ratio for studying the characteristics of post-mitotic, differentiated enterocytes in their in vivo condition.
Lysate Preparation, Immunoprecipitation (IP), and Immunoblot (IB) Analysis-Cold lysis buffer (38) was added to the tumors or enterocytes pooled from two mice of the same genotype. Unless otherwise indicated, the proteasome inhibitor, ALLN (10 mM), was added to the lysis buffer. Cells and tumors in lysis buffer were separately homogenized by 10 strokes in chilled Dounces. All subsequent steps for protein isolation were performed at 4°C. Lysates were centrifuged for 10 min at 12,000 ϫ g, and protein concentrations of supernatants were determined by Lowry assay. Normalized aliquots of each lysate were placed in Laemmli buffer (67 mM Tris-HCl, pH 6.8, 2% SDS, 10% glycerol, 0.01% bromphenol blue, and 0.05% ␤-mercaptoethanol (v/v)) and stored at Ϫ70°C. In parallel, portions of each lysate were pre-cleared to remove immunoglobulins by mixing with protein G beads at 4°C for 4 h, followed by centrifugation and removal of the beads. After preclearing, the same amount of protein from each lysate was reacted with primary antibody for 1 h. Immune complexes were precipitated for 12 h with protein G beads. The beads were washed, and the bound proteins were released in 50 l of Laemmli buffer. The entire amount of each IP sample or the total cell lysate samples were resolved by 7.5% SDSpolyacrylamide gel electrophoresis. Procedures for IB analyses were as detailed (42). Washed membranes were stripped by incubation at 65°C for 20 min in 68 mM Tris-HCl, pH 6.8, 10% SDS, and 0.01% ␤-mercaptoethanol before re-probing. All experiments were repeated at least twice using independently prepared lysates.
Immunohistochemistry-Hematoxylin and eosin staining of formalin-fixed, paraffin-embedded tissue was performed, and tissues were examined by light microscopy to document histology. Serial sections of mid-small intestinal mucosa from Apc ϩ/ϩ and Apc Min/ϩ animals were deparaffinized and rehydrated. Endogenous peroxidases were quenched in 3% H 2 O 2 , and the slides were rinsed in PBS. The sections were then incubated at 25°C for 1 h with a rat anti-mouse E-cadherin antibody (clone ECCD-2). Detection was accomplished using the LSAB 2 system using a biotinylated anti-rat antibody at 1:300 dilution.
Membrane Fractionation-Enterocytes and tumor cells isolated as described above were washed twice in Tris-buffered saline, pH 7.4, containing 1 mM CaCl 2 . Pelleted cells were frozen in liquid N 2 and stored at Ϫ70°C. At the time of assay, cell pellets were placed in 1 ml of a hypotonic lysis buffer without EDTA, NaCl, and detergent (10 mM Tris-HCl, pH 7.8, 1 mM CaCl 2 , 5 mM KCl, plus proteasome, protease, and phosphatase inhibitors as described (40)). Cells were allowed to thaw and swell on ice for 15 min, before being homogenized with 30 strokes. Cell suspensions were centrifuged for 5 min at 1,700 rpm. The pellet containing debris, nuclei, and unlysed cells were discarded; the supernatants were centrifuged at 100,000 ϫ g for 1 h. The subsequent high speed centrifugation supernatants (S100) were retained; the pellets (P100) were suspended in lysis buffer containing 1% Triton X-100 and centrifuged as above. The supernatants (detergent-soluble P100) were retained; the pellets (detergent-insoluble P100), suspended in Laemmli sample buffer, were also retained. Aliquots of each fraction were removed for protein determination by Lowry assay. Normalized amounts of the fractionated proteins were resolved by 10% SDS-PAGE and IB analyses performed.

E-cadherin-␤-Catenin Complexes Were Decreased in Apc Min/ϩ Enterocytes and Increased in Apc Min/Ϫ Adenoma Cells-Because
the migration of Apc Min/ϩ enterocytes is decreased relative to that of their wild-type littermates (Apc ϩ/ϩ ) (40), we predicted that the AJ structure and/or function of Apc Min/ϩ intestinal cells is also altered. Furthermore, we expected that changes in the AJ would also be found in Apc Min/Ϫ tumor cells, since these cells may have lost the ability to migrate altogether (14,15,40,44). We examined the overall expression of E-cadherin in Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ adenoma cells by IB analysis. As shown in Fig. 2A, equivalent expression of E-cadherin was present in Apc ϩ/ϩ and non-tumor Apc Min/ϩ tissues; however, expression of this protein was increased in Apc Min/Ϫ adenoma cells. To examine the association between E-cadherin and ␤-catenin in the tissue samples, ␤-catenin was immunoprecipitated from tissue lysates prepared with the proteasome inhibitor, ALLN. An IB of the precipitated proteins detected the association of E-cadherin with ␤-catenin in Apc ϩ/ϩ cells (Fig. 2B). This experiment showed a significantly increased association of E-cadherin with ␤-catenin in the Apc Min/Ϫ adenoma cells. In Apc Min/ϩ enterocytes, however, a decreased association between these proteins was evident.
These results suggested an effect of the Apc Min allele on the AJ structure of intestinal epithelial cells. To characterize this effect further, we determined the subcellular location of Ecadherin and ␤-catenin by fractionating Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ cells. The cells were lysed without detergent; nuclei were removed, and then the lysate samples were centrifuged at high speed to obtain cytosolic (S100) and membrane (P100) fractions. The membrane fractions were further separated by treatment with detergent to yield membrane-soluble and membrane-insoluble fractions. IB analyses were separately performed to detect both the intracellular (clone 36) and extracellular (DECMA-1) domains of E-cadherin. As shown in Fig. 3, E-cadherin was detected by both antibodies in the P100 membrane-associated (detergent-soluble) and membrane integral (detergent-insoluble) fractions in all three samples. The localization of E-cadherin in these two epithelial cell compartments was described in previous studies (45)(46)(47). Attachment of Ecadherin to the actin cytoskeleton via ␣and ␤-catenin renders the complex insoluble in detergent (45)(46)(47). Interestingly, this experiment showed an increase in both the extra-and intracellular components of E-cadherin in the cytosol of Apc Min/ϩ enterocytes (Fig. 3). Internalization of E-cadherin is promoted in epithelial cells grown under conditions that do not permit stable cell-cell contact, such as low cell density (36,37). These results are therefore consistent with a significant loss of AJ FIG. 2. E-cadherin expression and ␤-catenin association are altered during Apc-associated tumorigenesis. IB analysis of E-cadherin expression was performed using lysates from Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ cells. Proteins (50 g) were resolved by 10% SDS-PAGE, electroblotted to a membrane. The top portion of the membrane was probed with clone 36 anti-E-cadherin antibody, and the bottom portion was probed with antiactin antibody (A). B, IP of the same lysates (300 g of protein) using anti-␤catenin antibody followed by IB analysis using clone 36 anti-E-cadherin antibody.
The band corresponding to 120-kDa Ecadherin in a HeLa cell lysate serves as a standard. Ig heavy chain (HC) bands (ϳ50 kDa) serve as internal loading controls. structure in the Apc Min/ϩ enterocytes. This result was confirmed using immunohistochemistry to detect the location of E-cadherin in sections obtained from Apc ϩ/ϩ and Apc Min/ϩ small intestine (Fig. 4). E-cadherin antibody demonstrates prominent staining of the lateral membranes in the Apc ϩ/ϩ enterocytes (Fig. 4A) that is lost in the Apc Min/ϩ tissue (Fig.  4B).
Cell fractionation also showed that ␤-catenin was located primarily in the membrane detergent-soluble P100 fraction of all three cell types (Fig. 3). The Apc Min/Ϫ adenomas lack functional Apc and are therefore unable to degrade free cytosolic ␤-catenin (48,49); therefore, we expected to find increased amounts of cytosolic ␤-catenin in the Apc Min/Ϫ cells. Because the fractionation shown in Fig. 3 was performed using cells treated with a proteasome inhibitor, the increase in cytosolic ␤-catenin in the Apc Min/Ϫ adenomas could not be detected. To compare the effect of Apc-mediated degradation on ␤-catenin levels in Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ tissues, we measured ␤-catenin expression by IB using whole cell lysates prepared with and without ALLN (Fig. 5). Bands of 92 kDa, the size of ␤-catenin, were present in the Apc Min/ϩ and Apc ϩ/ϩ samples at similar low intensities (Fig. 5, arrows, left). In the Apc Min/Ϫ tumor sample, however, a broad intense band was seen in this location, indicating a lack of ␤-catenin degradation following loss of the wild-type Apc allele. As a control, the activity of the proteasome was blocked with ALLN. In this experiment, Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ cells contained similar high levels of ␤-catenin (Fig. 5, right), consistent with the results seen in Fig. 3. Thus, while we did not assess the nuclear pool of ␤-catenin in our samples, we conclude that in these normal tissues and benign adenomas, ␤-catenin is mainly associated with the membrane.
The Tyrosine-phosphorylated Form of ␤-Catenin Was Upregulated in Non-tumor Apc Min/ϩ Tissue-The signaling potential of ␤-catenin at cell-cell adhesion sites is modulated by tyrosine phosphorylation, and distinct modification sites cause its dissociation from E-cadherin and ␣-catenin (50). Because we observed a reduced association of ␤-catenin with E-cadherin in Apc Min/ϩ intestinal cells and the opposite effect in Apc Min/Ϫ adenomas, we predicted that the level of tyrosine-phosphorylated ␤-catenin would be increased in the Apc Min/ϩ cells and decreased in the adenomas. To examine steady-state levels of tyrosine-phosphorylated ␤-catenin in enterocytes, we performed IP and IB analyses using lysates of Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ adenoma cells prepared in the presence of ALLN. Cell lysates were immunoprecipitated with 4G10 anti-phosphotyrosine antibody, followed by IB analysis using anti-␤-catenin antibody. As shown in Fig. 6A, the relative level of tyrosinephosphorylated ␤-catenin was increased in Apc Min/ϩ enterocytes when compared with Apc ϩ/ϩ intestine. Interestingly, two bands of ϳ92-95 kDa appeared in the lane containing the non-tumor Apc Min/ϩ cell lysate (arrow). Only the faster mobility band was evident in the lane containing the wild-type (Apc ϩ/ϩ ) and Apc Min/Ϫ lysates. This result was confirmed by the reciprocal experiment in which the blot of samples precipitated with anti-␤-catenin antibody was probed with 4G10 (Fig. 6B). The intensity of the ϳ92-Da band in the lane containing the nontumor Apc Min/ϩ cell lysate was roughly twice that of the corresponding band from the Apc ϩ/ϩ lysate, and the expression of tyrosine-phosphorylated ␤-catenin was again lower in the tu- Immunohistochemistry performed using anti-E-cadherin antibody shows prominent membrane staining for E-cadherin in Apc ϩ/ϩ enterocytes (A) that is substantially decreased in Apc Min/ϩ tissue (B).

FIG. 3. Comparison of E-cadherin and ␤-catenin localization in Apc ؉/؉ , Apc Min/؉ , and Apc Min/؊ enterocyte lysates following cell fractionation.
Intestinal cell scrapings and adenomas were lysed in hypotonic buffer without detergent. Cell fractionation was performed as detailed under "Experimental Procedures." IB analysis used 50 g of protein for the E-cadherin blots (clone 36 and DECMA-1) and 10 g for the ␤-catenin (clone 14) one. mor cells. These data suggested that non-tumor Apc Min/ϩ enterocytes may contain ␤-catenin that was phosphorylated at two different tyrosine residues, whereas Apc ϩ/ϩ and Apc Min/Ϫ cells contain this protein modified at a single tyrosine.
Altered Associations of ␤-Catenin with RPTPs in Apc Min/ϩ and Apc Min/Ϫ Cells-Tyrosine phosphorylation of adhesion proteins is controlled by the coordinated activities of kinases and phosphatases. We therefore determined whether the association of ␤-catenin with RPTPs could be demonstrated in enterocytes and whether differences in the degree of this association would match the differences in the tyrosine-phosphorylated ␤-catenin observed in these cells. IP/IB analyses were performed using antibodies specific for RPTP␤/, RPTP␥, and PTP LAR. In the case of PTP LAR, no association with ␤-catenin was detected in the enterocytes, although expression of this phosphatase was observed by IB of total cell lysates from all three cell types (data not shown). An association of ␤-catenin with RPTP␤/ and RPTP␥ in Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ cells was observed as shown in Fig. 7. IB analyses of the overall expression of RPTP␤/ and RPTP␥ are presented in Fig. 7, A and C, respectively. The ectodomains of these receptor proteins are processed in a manner that leaves the phosphatase-containing cytoplasmic portion anchored in the membrane and the soluble extracellular portion non-covalently attached (51). Arrows indicate the extracellular soluble fragment (top) and the intracellular phosphatase fragment (bottom). The processed forms of these RPTPs are also apparent in positive control cell lysates (Jurkat and A431). In all three tissues, the expression of these phosphatases appears invariant and thus is unaffected by the Apc genotype. Consistent with the increased tyrosine phosphorylation of ␤-catenin in non-tumor Apc Min/ϩ enterocytes, however, IP/IB analyses showed a reduced association of ␤-catenin with RPTP␤/ (Fig. 7B). The same results were obtained with a second anti-RPTP␤/ antibody, MAB5210 (data not shown). In addition, consistent with the reduced tyrosine phosphorylation of ␤-catenin in Apc Min/Ϫ adenoma cells, associations of this protein with both RPTP␤/ and RPTP␥ were detected (Fig. 7, B and D). No association of these phosphatases with E-cadherin could be demonstrated in these tissues when blots of Fig. 7, B and D, were stripped and re-probed using clone 36 antibody (data not shown).
In epithelial cells, EGFR can tyrosine-phosphorylate ␤-catenin (33), and we therefore explored the possibility that ␤-catenin is phosphorylated by EGFR in our specimens. An IP of the lysates was performed using antibody against EGFR (clone 13), followed by IB with ␤-catenin antibody. This experiment showed minimal or no association between ␤-catenin and EGFR (data not shown). Stripping and reprobing the IB with the anti-EGFR antibody confirmed that the 180-kDa EGFR protein was precipitated. The expression of EGFR in the Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ cells was relatively low and invariant among the samples (data not shown). In addition, Src kinase activity was assessed by IP of the cell lysates using antibodies against c-Src (SRC2) and phospho-Src (Tyr-416) (1: 200 antibody dilution). By this method, no differences in Src expression or phosphorylation were observed between the Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ cells (data not shown). Thus, we have no evidence that the increased expression of tyrosinephosphorylated ␤-catenin in the Apc Min/ϩ enterocytes involved up-regulation of these kinases.
Association between ␤and ␣-Catenin Was Altered during Apc-associated Tumor Formation-␣-Catenin is an actin-bundling protein that links the transmembrane cadherin to the actin cytoskeleton indirectly via association with ␤or ␥-catenin at the AJ (33,52,53). The interaction between ␣-catenin and ␤-catenin in the various cell types was therefore examined. In the slowly migrating Apc Min/ϩ enterocytes, increased expression of ␣-catenin was observed (Fig. 8A), although its associa- FIG. 5. IB analysis of ␤-catenin using total cell lysates prepared in the presence and absence of the proteasome inhibitor, ALLN. Without inhibition of the proteasome, full-length ␤-catenin was present at low levels in Apc ϩ/ϩ and non-tumor Apc Min/ϩ tissue (3rd and 4th lanes from left), but this protein was greatly increased in the Apc Min/Ϫ cells (5th lane from left). With ALLN, the amount of ␤-catenin was the approximately the same in each cell type (6th to 8th lanes from left). All lanes were loaded with 30 g of protein.
FIG. 6. Expression of tyrosine-phosphorylated ␤-catenin in Apc ؉/؉ , Apc Min/؉ , and Apc Min/؊ cells. Lysates from Apc ϩ/ϩ , non-tumor Apc Min/ϩ , and Apc Min/Ϫ cells were prepared with ALLN (10 mM) and then immunoprecipitated with the anti-phosphotyrosine antibody, 4G10. IB was then performed using clone 14 anti-␤-catenin antibody (A). Reciprocal experiment in which IPs of lysates prepared in parallel used clone 14 anti-␤-catenin antibody, and the IB was performed with 4G10 (B). The Ig heavy chain (HC) bands serve as loading controls. IPs used 300 g of protein each. tion with ␤-catenin was not significantly different than that of wild-type enterocytes (Fig. 8B). The overall expression of ␣-catenin was up-regulated in the Apc Min/Ϫ adenoma cells (Fig. 8A), a result consistent with the increased expression of E-cadherin in the same samples ( Fig. 2A). The association of ␣-catenin with ␤-catenin was also markedly increased in Apc Min/Ϫ adenoma cells (Fig. 8B). The increased binding between these two catenins and E-cadherin in adenoma cells is consistent with the complex being associated with the actin cytoskeleton and therefore detergentinsoluble (Fig. 3). P120 ctn , a cadherin-associated SFK, binds directly to E-cadherin at the AJ (reviewed in Ref. 54). P120 ctn is phosphorylated in response to ligand stimulation of receptor tyrosine kinases (55) and may regulate cadherin-mediated adhesion. To determine whether p120 ctn was affected by the Min mutation and as a further test of the specificity of the results presented in Figs. 2-8, we examined the expression and phosphorylation of p120 ctn in Apc ϩ/ϩ , Apc Min/ϩ , and Apc Min/Ϫ adenoma cells. As shown in Fig. 9, both the overall expression of p120 ctn and its tyrosine phosphorylation status were unchanged throughout these tumor progression stages. As a control, the anti-phospho-tyrosine IBs were stripped and re-probed with anti-p120 ctn antibody. These re-probed blots showed that intact p120 ctn proteins were present in all of the samples (data not shown). DISCUSSION In the epithelium, the signaling pathways controlling cell migration are integrated with those regulating cell proliferation and survival. These processes are altered during tumorigenesis, resulting in a cell that escapes normal growth controls. In the majority of human colon carcinomas and in the germ line of patients with familial adenomatous polyposis, early tumor formation is associated with APC protein truncation (56,57). A truncated form of Apc protein together with the full-length product of a wild-type Apc allele is found in the histologically normal small intestine of the Min/ϩ mouse, and Apc truncation with loss of heterozygosity at the second Apc allele is characteristic of adenomas from Min/ϩ mice (48,49). The phenotype of this animal, therefore, illustrates two stages of tumor progression, i.e. Apc Min/ϩ enterocytes and Apc Min/Ϫ adenoma cells. The earliest tumor-associated changes, found in Apc Min/ϩ cells, are alterations of adhesion. Although normal-appearing by his- FIG. 7. Comparisons of the expression and associations with ␤-catenin of RPTP␤/ and -␥ in Apc ؉/؉ , Apc Min/؉ , and Apc Min/؊ enterocyte lysates. IB using clone 12 anti-RPTP␤/ and 50 g of protein of the three cell samples (A, left). A similar analysis of RPTP␥ using goat antibody M -18 (C, right). The bottom portions of these blots were separately probed for ␤-actin as loading controls. IP of 500 g of protein from each cell lysate using clone 12 anti-RPTP␤/ (B, left) or M-18 anti-RPTP␥ antibody (D, right). Ig heavy chain (HC) bands serve as loading controls. The processed extracellular (120 kDa) and phosphatase (ϳ75 kDa) fragments of these RPTPs are indicated by arrows. These specific bands were not detected when these blots were stripped and re-probed with mouse Ig as a negative control (not shown). The processed RPTP forms are also present in the positive control Jurkat (A) and A431 (C) lysates. Although the full-length proteins were present in the control lysates, very little of these forms was detected in the mouse intestinal cells (data not shown). The data of A and B were exactly reproduced using MA5210, another antibody for RPTP␤/, but is not shown. tology, these enterocytes migrate slowly (40), have increased residence time in the intestine (40), and demonstrate decreased AJ structural integrity. It is not until the second Apc allele is lost that the cells acquire a hyperproliferative phenotype, as evidenced by up-regulation of cyclin D1 expression in adenomas from Min/ϩ mice (58,59). In Apc Min/Ϫ adenoma cells, we found that the association of both ␣and ␤-catenin with Ecadherin was increased. This state of augmented AJ structure occurred in association with a presumed increase in ␤-catenin/ Tcf4-mediated gene expression caused by loss of wild-type Apc at the adenoma stage. Similar conditions were found in intestinal epithelial cells expressing a mutant form of ␤-catenin that was protected from degradation because it lacked the serine-threonine phosphorylation sites (60). When this condition was examined in vivo, enterocytes with increased ␤-catenin/Tcf-mediated gene transcription also demonstrated increased E-cadherin expression and a 4-fold elevation in proliferation within the murine crypts (39).
The tyrosine phosphorylation of catenins regulates their interaction with adhesion complexes and consequently modulates the activity of the AJ. We found that increased tyrosine phosphorylation of ␤-catenin was present in Apc Min/ϩ enterocytes and occurred together with reduced ␤-catenin-E-cadherin binding and reduced enterocyte migration (40). In multiple experimental systems, tyrosine phosphorylation of ␤-catenin was associated with decreased cadherin-dependent adhesion (14,33,35). Reaction of recombinant c-Src in vitro with ␤-catenin produced tyrosine phosphorylation of two residues, Tyr-86, located near the N terminus, and Tyr-654, found in the final armadillo repeat (31). In addition, modification of ␤-catenin at Tyr-654 prevented the interaction of ␤-catenin with E-cadherin in Caco-2 colon cancer cells (31). It remains to be determined whether the increased phosphorylation of Apc Min/ϩ enterocytes occurred at one or both of these sites. Our data shows that decreased ␤-catenin-E-cadherin association occurred in Apc Min/ϩ enterocytes together with the appearance of full-length E-cadherin in the cytoplasm. Taken together, these results suggest that, in slowly migrating Apc Min/ϩ enterocytes, decreased ␤-catenin-E-cadherin binding leads to increased E-cadherin internalization and may therefore alter the cycling of E-cadherin from membrane to cytoplasm as adhesion contacts form and re-form.
Insight into the mechanisms of early Apc-associated intestinal tumorigenesis can be obtained by comparing the differences in AJ structure between Apc Min/ϩ enterocytes and Apc Min/Ϫ adenoma cells. In the Apc Min/Ϫ tumor cells, in addition to decreased tyrosine-phosphorylated ␤-catenin and increased ␤-catenin-E-cadherin binding, we also observed increased association between ␤-catenin and ␣-catenin. ␣-Catenin is an actin bundling protein that joins the cadherin-associated adhe-sion complexes to the cytoskeleton (33,53). We found that E-cadherin, ␤-catenin, and ␣-catenin were assembled at the membrane of Apc Min/Ϫ adenoma cells in a manner that suggests tight adhesion of the tumor cells, a condition associated with reduced turnover of these proteins (61,62). ␣-Catenin binds to the N-terminal domain of ␤-catenin (63), whereas E-cadherin binds to the C terminus (28,37). In view of the reduced binding between E-cadherin and ␤-catenin in Apc Min/ϩ enterocytes, it is reasonable to suggest that differential tyrosine phosphorylation of ␤-catenin separately affects the interactions of these proteins. By this model, the lack of augmented ␣-catenin/␤-catenin association in Apc Min/ϩ enterocytes is consistent with the increased tyrosine phosphorylation of ␤-catenin observed in these cells (Fig. 6).
Tyrosine phosphorylation of ␤-catenin is modulated by several factors, including SFKs, receptor tyrosine kinases, and specific transmembrane and non-receptor tyrosine phosphatases. We have not yet identified an Src kinase specifically activated in Apc Min/ϩ enterocytes, although likely candidates include c-Src, c-Yes, Fyn, and Fer (29,30,64). EGFR can also tyrosine-phosphorylate ␤-catenin (35), as activation of this growth factor receptor produced ␤-catenin tyrosine phosphorylation and dissociation of the actin cytoskeleton from the AJ in breast cancer cells (33). In breast epithelial cells, signaling via EGFR also induced tyrosine phosphorylation of p120 ctn , a condition not seen in Apc Min/ϩ or Apc Min/Ϫ cells (Fig. 5). Our data suggest that neither EGFR nor c-Src association was responsible for the increased tyrosine-phosphorylated ␤-catenin in Apc Min/ϩ enterocytes.
Through the modification of AJ-associated proteins, phosphatases may regulate adhesion-mediated signaling. AJ functions are altered by c-Src, a signal transduction protein that is activated by tyrosine phosphatases (reviewed in Ref. 65). Phosphatase activity may control adhesion protein cell surface distribution, including protein movement into and out of lipid rafts (51). Although the targets of many phosphatases are as yet uncharacterized, gene knockout studies indicate that they have very distinct biological effects that are likely to be tissuespecific (66 -71). The association of phosphatases with the AJ has not been described previously. In addition, the appropriate interaction of ␤-catenin with tyrosine phosphatases in postmitotic cells whose differentiated function requires them to be mobile is not known. A number of different phosphatases interact with ␤-catenin, including cytosolic PTPs such as PTP␣ (72), hPTP (73), and PTP LAR (50), and the receptor proteintyrosine phosphatase, RPTP␤/ (34, 74 -76). RPTP␤/ was first characterized in neural tissue, where it regulates cell migration, adhesion, and neurite outgrowth (50). In unstimulated cells, RPTP␤/ is intrinsically active and controls the tyrosine FIG. 9. Comparisons of the expression and tyrosine phosphorylation of p120 ctn in Apc ؉/؉ , Apc Min/؉ , and Apc Min/؊ enterocyte lysates. IB of the intestinal cell lysates (50 g of protein) in which the top portion was probed with clone clone 98 antibody against p120 ctn , and the bottom portion of of the membrane was probed with ␤-actin as a loading control (A). B, IP of lysates (500 g of protein) using 4G10 anti-phosphotyrosine antibody followed by IB using clone 98. Ig heavy chain (HC) bands serve as internal loading controls. phosphorylation status of ␤-catenin (34). In our enterocyte assays, even through we do not demonstrate phosphatase activity directly, the decreased association of RPTP␤/ with ␤-catenin observed in Apc Min/ϩ enterocytes is consistent with increased levels of tyrosine-phosphorylated ␤-catenin and decreased cell migration. The expression of RPTP␤/ was reported previously (77) to be restricted to the nervous system, whereas RPTP␥ is ubiquitously expressed. Under the assay conditions reported here, IP studies using antibody specific for RPTP␥ showed significantly increased association of this phosphatase with ␤-catenin in the Apc Min/Ϫ adenoma cells (Fig. 7D). This association may account for the diminished level of tyrosinephosphorylated ␤-catenin observed in these tumors.
The data presented here do not define the exact link between APC truncation and modulation of AJ structure. APC and E-cadherin do not directly interact, although both of these proteins serve as scaffolds for catenin binding (78). APC binds to the core region of ␤-catenin, which is composed of 12 armadillo repeats, an amino acid sequence motif that is common to catenins and APC (78,79). The Min mutation at codon 850 produces a 95-kDa protein truncated at the C terminus. This truncated protein retains the homodimerization domain and armadillo repeats of APC but lacks both the constitutive and kinase-regulated ␤-catenin binding regions (17,80). In Apc Min/ϩ tissues, however, a dimer containing both fulllength and truncated APC may bind ␤-catenin, possibly creating conditions where its tyrosine phosphorylation is facilitated. This argument is supported by the lack of a second tyrosine phosphorylation site in the Apc Min/Ϫ tumor cells, as these cells retain the truncated APC but have lost the fulllength protein (49). Alternatively, the Min truncation may prohibit the assembly of a functional degradation complex, causing an accumulation of the ␤-catenin-Tyr(P)-654 species. Finally, PTPs bind to ␤-catenin at the armadillo repeats (50), and it is possible that dephosphorylation is inhibited when truncated APC is associated with ␤-catenin-Tyr(P)-654.
These experiments show successive changes in E-cadherin-␤-catenin association and the relative levels of tyrosine-phosphorylated ␤-catenin, as enterocytes progress from Apc ϩ/ϩ to Apc Min/ϩ to Apc Min/Ϫ , and illustrate several important concepts governing tumor development in the setting of Apc loss. First, they provide in vivo data showing that progressive alterations in AJ structure occur during adenoma formation. These changes suggest that there is an initial reduction of cell-cell adhesion in Apc Min/ϩ enterocytes, followed by an eventual increase in AJ formation in the Apc Min/Ϫ adenoma cells. These experiments also show that the tumor-promoting effect of the Min mutation involves a defect in cadherin-mediated adhesion, indicating a dominant negative effect of the truncated APC protein produced from the Min allele.