E-cadherin and adenomatous polyposis coli mutations are synergistic in intestinal tumor initiation in mice.

BACKGROUND & AIMS
Inactivation of the adenomatous polyposis coli (APC) gene is observed at early stages of intestinal tumor formation, whereas loss of E-cadherin is usually associated with tumor progression. Because both proteins compete for the binding to beta-catenin, an essential component of the Wnt signaling pathway, reduction of E-cadherin levels in an Apc mouse model could influence both tumor initiation and progression. In addition, loss or haploinsufficiency of E-cadherin may affect tumorigenesis by altering its cell-adhesive and associated functions.


METHODS
Apc1638N mice were bred with animals carrying a targeted E-cadherin knockout mutation.


RESULTS
Double heterozygous animals showed a significant 9-fold and 5-fold increase of intestinal and gastric tumor numbers, respectively, compared with Apc1638N animals. The intestinal tumors of both groups showed no significant differences in grading and staging. Loss of heterozygosity analysis at the Apc and E-cadherin loci in both intestinal and gastric Apc(+/1638N)/E-cad(+/-) tumors revealed loss of the wild-type Apc allele in most cases, whereas the wild-type E-cadherin allele was always retained. This was supported by a positive, although reduced, staining for E-cadherin of intestinal tumor sections.


CONCLUSIONS
Introduction of the E-cadherin mutation in Apc1638N animals enhances Apc-driven tumor initiation without clearly affecting tumor progression.


See editorial on page 1161.
Background & Aims: Inactivation of the adenomatous polyposis coli (APC) gene is observed at early stages of intestinal tumor formation, whereas loss of E-cadherin is usually associated with tumor progression. Because both proteins compete for the binding to ␤-catenin, an essential component of the Wnt signaling pathway, reduction of E-cadherin levels in an Apc mouse model could influence both tumor initiation and progression. In addition, loss or haploinsufficiency of E-cadherin may affect tumorigenesis by altering its cell-adhesive and associated functions. Methods: Apc1638N mice were bred with animals carrying a targeted E-cadherin knockout mutation. Results: Double heterozygous animals showed a significant 9-fold and 5-fold increase of intestinal and gastric tumor numbers, respectively, compared with Apc1638N animals. The intestinal tumors of both groups showed no significant differences in grading and staging. Loss of heterozygosity analysis at the Apc and E-cadherin loci in both intestinal and gastric Apc ؉/1638N / E-cad ؉/؊ tumors revealed loss of the wild-type Apc allele in most cases, whereas the wild-type E-cadherin allele was always retained. This was supported by a positive, although reduced, staining for E-cadherin of intestinal tumor sections. Conclusions: Introduction of the E-cadherin mutation in Apc1638N animals enhances Apcdriven tumor initiation without clearly affecting tumor progression.
M alignant tumors are the result of accumulation of several genetic alterations. Several lines of evidence indicate that mutations of the adenomatous polyposis coli (APC) gene represent one of the earliest genetic events in intestinal tumor development. First, germline APC mutations are responsible for familial adenomatous polyposis, a dominant condition characterized by multiple colorectal adenomatous polyps in the colon and rectum. 1,2 Germline mutations in the mouse Apc gene also result in a dominant predisposition to the development of multiple tumors in the intestine and elsewhere. [3][4][5][6] In accordance with Knudson's 2-hit hypothesis, somatic mutations at the wild-type APC allele can be detected in most intestinal tumors of both human and murine origin. 5,[7][8][9][10] Lastly, the majority of sporadic colorectal tumors carry APC mutations with similar frequencies between early and more advanced tumor stages. 11 The APC gene encodes a 312-kilodalton protein shown to participate in several cellular processes, including cell cycle regulation, apoptosis, cell adhesion, cell migration, microtubule assembly, and signal transduction. 12 We showed recently that the tumor-suppressing function of APC resides in its capacity to properly regulate intracellular ␤-catenin levels, whereas its COOHterminal domains seem to be less relevant for its function in development and tumorigenesis. 13 ␤-Catenin functions as a signaling molecule within the Wnt signal transduction pathway by associating in the nucleus with members of the lymphoid enhancer factor/T-cell factor family of transcriptional activators, thereby modulating the transcription of target genes. 14 In the absence of a Wnt-signal, APC promotes down-regulation of ␤-catenin, thereby preventing its signaling activity. 15 As a result of APC mutations, ␤-catenin constitutively transduces signals to the nucleus. Accordingly, oncogenic ␤-catenin mutations that make it resistant to proteolytic degradation are found in sporadic colorectal tumors without APC mutations. 16 In addition to its role in the Wnt-signaling pathway, ␤-catenin binds to the cytoplasmic domain of E-cadherin, thus mediating cell-to-cell interactions by linking it to the actin cytoskeleton. [17][18][19][20] Whereas loss of APC function seems to be involved in initiation of intestinal tumor development, inactivation of E-cadherin is generally associated with later stages of tumor progression because disruption of E-cadherin function correlates with increased tumor invasiveness. [21][22][23] Therefore, by reducing the expression of E-cadherin in animals with a targeted Apc mutation, one may expect an increased invasive potential of the developing intestinal tumors. However, because ␤-catenin appears to exist in a state of dynamic equilibrium with E-cadherin and Apc, 24,25 reducing Ecadherin levels may also potentially affect intestinal tumor initiation by increasing the amount of intracellular ␤-catenin available for signaling. Reduction of E-cadherin levels may also contribute to tumorigenesis by affecting its cell-adhesive and associated functions. 26 -29 A possible role for E-cadherin in tumor initiation is supported by the presence of germline mutations in families predisposed to gastric cancer development 30 -32 and mutations in early-stage gastric and breast cancers. [33][34][35] To address the issue of "cross talk" between cell-to-cell adhesion and signal transduction, we have bred the Apc1638N model 4,6 with mice carrying a targeted E-cadherin mutation. 36 The double heterozygous Apc ϩ/1638N /E-cad ϩ/Ϫ mice show a marked increase in tumor number along the upper gastrointestinal tract without any clear modification of tumor invasivity or malignant potential. These results indicate that haploinsufficiency at the E-cadherin locus is synergistic with mutation of the Apc gene in gastrointestinal tumor initiation.

Gross Study and Tissue Processing
All animals were killed between 6 and 9 months of age, after which the entire intestine was removed and opened longitudinally. Sections of approximately 10 cm were spread out flat on filter paper, fixed overnight at 4°C in Notox (Earth Safe Industries, Inc., Bellemead, NJ), and transferred to 70% ethanol. The size and location of macroscopically visible tumors were determined, followed by resection and embedding of lesions in paraffin according to standard procedures. In addition, a number of freshly isolated tumors were snap-frozen in Tissue-tek OCT compound (Sakura B.V., Zoeterwoude, The Netherlands). Cutaneous cysts and desmoids were counted as described previously. 6 Liver and lungs of all animals were investigated for the presence of metastases by macroscopic and microscopic analysis. Tumor numbers were statistically evaluated by the Mann-Whitney exact test.

Histopathologic Classification of Gastrointestinal Tumors
Tumors were classified according to standard criteria, 37 considering a given tumor malignant only if invasion (lamina propria or beyond) was present. The invasion was defined by the presence of stromal fibroblastic reaction (desmoplasia). Tumors were graded according to the extension of glandular differentiation, 38 and staged according to the depth of invasion in the intestinal wall and the presence of lymph node or systemic metastases. Both lungs and liver were embedded in toto, and serial H&E-stained sections were screened for the presence of microscopic metastases by 2 independent observers (S.D.-C. and A.L.).

Tumor DNA Isolation
Tumor DNA isolation from microdissected paraffin or cryosections was performed essentially as previously described. 10 Briefly, six 16-m-thick sections were placed on glass slides, deparaffinized, rehydrated, and briefly stained with hematoxylin. The area containing the tumor cells was subsequently microdissected under a dissection microscope using as reference an H&E-stained 5-m section. The isolated neoplastic tissue was transferred in 250-L extraction buffer (10 mmol/L Tris-HCl [pH 8.0], 100 mmol/L NaCl, 25 mmol/L EDTA, 0.5% sodium dodecyl sulfate, and 300 g/mL proteinase K) and incubated at 55°C for 40 hours. A second aliquot of 60 g of proteinase K was added after 20 hours of incubation. Cellular proteins were removed by a phenol/chloroform and chloroform extraction followed by a 30-minute precipitation at Ϫ20°C with 125 L 7.5 mol/L NH 4 Ac, 20 g glycogen, and 625 L ethanol. The precipitate was resuspended in 50 L TE Ϫ4 (10 mmol/L Tris-HCl [pH 8.0] and 0.1 mmol/L EDTA).

Loss of Heterozygosity Analysis at the Apc and E-cadherin Locus
Loss of the wild-type Apc and E-cadherin allele was analyzed by amplifying in a single reaction both the wild-type and targeted alleles. 10 The following primers were used for E-cadherin: 1533 (5Ј-CCTCTCCTTTGACAGGAACCTC-CGT-3Ј) and 1534 (5Ј-CAGCCCAGAGGTGAGCACACT-GATG-3Ј), resulting in a 190 -base pair (bp) product diagnostic of the wild-type allele, and E-cadneo71 (5Ј-GCAAAACCACACTGCTCGAC-3Ј) that in combination with 1533 results in a 198-bp product diagnostic of the targeted allele. For the Apc locus, the following primers were used: Apc-A2 (5Ј-TCAGCCATGCCAACAAAGTCA-3Ј) and Apc-C2 (5Ј-GGAAAAGTTTATAGGTGTCCCTTCT-3Ј), resulting in a 216-bp product diagnostic of the wild-type allele, and Neo3 (5Ј-CACTTCATTCTCAGTATTGTTTTG-3Ј) that in combination with Apc-A2 results in a 227-bp product diagnostic of the targeted Apc1638N allele. A radioactive polymerase chain reaction was performed using 2.5 L of tumor DNA or 2.5 L of normal DNA samples, the latter isolated from normal intestinal or gastric epithelium surrounding the tumor. The amplification was performed in a 20-L volume, containing 10 mmol/L Tris-HCl (pH 8.9), 50 mmol/L KCl, 2.5 mmol/L MgCl 2 , 10% glycerol, 200 g/mL bovine serum albumin, 0.01% gelatin, 0.2 mmol/L of each deoxyadenosine triphosphate, deoxythymidine triphosphate, and deoxyguanosine triphosphate, 0.05 mmol/L deoxycytidine triphosphate, 0.25 Ci [ 32 P]deoxycytidine triphosphate (3000 Ci/mmol), 0.2 U Taq polymerase, and 10 pmol of each primer. The reactions were heated for 5 minutes at 94°C followed by 32 polymerase chain reaction cycles at 94°C for 30 seconds, 55°C for 60 seconds, and 72°C for 90 seconds. Amplified fragments were resolved on a denaturing 6% polyacrylamide gel and dried on paper, and the number of counts per allele was determined on a phosphor imager. Subsequently, the number of counts of the larger allele was divided by the counts of the smaller allele to obtain an allelic ratio. A mean allelic ratio was calculated for all the normal controls. This value was used to generate a comparative ratio Ͼ1.0 by dividing the tumor allelic ratio by the mean normal allelic ratio. A comparative ratio Ͼ1.5 was interpreted as significant, i.e., indicative of loss of the wild-type allele.

Immunohistochemical Analysis
Frozen tissue sections (7 m) were fixed with acetone for 10 minutes and incubated for 30 minutes in phosphatebuffered saline (PBS)/1% bovine serum albumin to block nonspecific binding. Subsequently, sections were incubated for 1 hour with undiluted hybridoma supernatant of the DECMA antibody (Sigma, Zwijndrecht, The Netherlands) recognizing an epitope within residues 407-434 of E-cadherin located close to the membrane proximal part of the extracellular domain. 39,40 Sections were also incubated with a 1:400 dilution of a rabbit polyclonal antibody directed against the COOH terminus of ␤-catenin. 24 After 3 washes with PBS, the primary antibodies were visualized with a fluorescein isothiocyanate-conjugated goat anti-rat IgG antibody and Cy3-conjugated goat anti-rabbit antibody (Jackson ImmunoResearch Laboratories (West Grove, PA).
Immunohistochemical staining of E-cadherin in paraffin sections was performed with a monoclonal antibody (clone 34) raised against the COOH-terminal 150 residues of E-cadherin (Transduction Laboratories, Lexington, KY). Sections were rehydrated, blocked in PBS containing 5% nonfat dry milk, and subsequently incubated for 16 hours at 4°C with the primary antibody diluted 1:100 in PBS/5% nonfat dry milk. After 3 washes in PBS containing 0.2% Tween 20, the sections were incubated for 1 hour in a 1:250 dilution of a horseradish peroxidase-conjugated F(abЈ) 2 fragment goat anti-mouse antibody (Jackson ImmunoResearch Laboratories). The peroxidase reaction was developed with 0.01% hydrogen peroxide in 0.05% diaminobenzidine tetrahydrochloride in 50 mmol/L Tris-HCl (pH 7.6).

Tumor Size and Numbers in the Gastrointestinal Tract of
To study the interaction between cell-to-cell adhesion and signal transduction in the gastrointestinal epithelium, we have bred Apc1638N mice 4,6 with mice carrying a targeted E-cadherin mutation. 36 The gastrointestinal tracts of the compound Apc ϩ/1638N /E-cad ϩ/Ϫ mice were analyzed for the presence of tumors and compared with their Apc ϩ/1638N , E-cad ϩ/Ϫ , and wild-type littermates. All animals were killed between 6 and 9 months of age (mean age for all groups, 7.5 months). As expected, no tumors were observed in wild-type (n ϭ 10) and E-cad ϩ/Ϫ (n ϭ 9) animals, whereas Apc ϩ/1638N mice (n ϭ 17) had an average of 1.24 tumors per animal in the small intestine. A dramatic effect on tumor multiplicity along the gastrointestinal tract was observed in the compound heterozygous Apc ϩ/1638N /E-cad ϩ/Ϫ animals ( Table  1). The number of small intestinal tumors in these mice (n ϭ 18) showed a significant 9-fold increase (average, 11.3 tumors per animal) compared with the Apc ϩ/1638N group (P Ͻ 0.0001, Mann-Whitney test). A similar increase was observed for gastric tumors. Only 4 of 11 Apc ϩ/1638N mice had a single gastric tumor, but all 14 investigated Apc ϩ/1638N /E-cad ϩ/Ϫ animals had developed at least 1 tumor. The average number of macroscopically detectable gastric lesions was increased by more than  6 were not affected by the E-cadherin mutation (data not shown).
In contrast with the increase in tumor numbers, tumor size was not clearly affected by the E-cadherin haploinsufficiency. The Apc ϩ/1638N tumors (n ϭ 21) had grown to an average diameter of 2.32 Ϯ 0.83 mm, which is similar to the 2.01 Ϯ 0.83 mm observed in 124 tumors derived from 11 Apc ϩ/1638N /E-cad ϩ/Ϫ animals. As shown in Figure 1, also the distribution in tumor size between both groups is comparable, whereby the tumors of the Apc ϩ/1638N /E-cad ϩ/Ϫ animals even show a slight tendency toward smaller tumors. Hence, these results are indicative of an effect of E-cadherin haploinsufficiency on gastrointestinal tumor multiplicity in the Apc1638N model, rather than on the rate at which these tumors grow.

Morphology and Malignant Potential of the Gastrointestinal Tumors
Detailed histologic analysis of 15 Apc ϩ/1638N small intestinal tumors showed evidence of stromal invasion in all cases. The same was observed in 117 of 124 Apc ϩ/1638N / E-cad ϩ/Ϫ tumors. Most of the lesions found in both genotypes can therefore be regarded as carcinomas. Submucosal invasion was observed in 10 of 15 Apc ϩ/1638N tumors and 62 of 124 Apc ϩ/1638N /E-cad ϩ/Ϫ tumors. No diffuse growth pattern characteristic for a loss of adhesive function was observed in any of the tumors. Serial sectioning of the liver and lungs could not detect any metastases, neither in the Apc ϩ/1638N group nor in the group with additional E-cadherin mutation. Although we cannot exclude the presence of micrometastases, the absence of any identifiable metastasis in the double heterozygous animals with the 9-fold higher tumor burden suggests that this is an uncommon event. Therefore, both the degree of local invasion and the metastatic potential of the Apc1638N tumors are not affected by the Ecadherin mutation.
All gastric tumors were located in the fundus and pyloric region. The lesions of both groups showed variable proportions of gland-like structures, indicating an intestinal phenotype without any evidence of diffuse tumor growth. Examples of gastric lesions are shown in Figure 2.

E-cadherin Is Expressed in Tumors of Both Genotypes
We have previously shown that in the majority of Apc1638N intestinal tumors, the wild-type copy of Apc is somatically lost. 10,41,42 To assess whether loss of the wild-type E-cadherin allele underlies tumor formation in double heterozygous Apc ϩ/1638N /E-cad ϩ/Ϫ animals, 16 intestinal and 16 gastric tumors were investigated by LOH analysis. Twelve of 16 intestinal tumors tested showed loss of the wild-type Apc allele. Gastric tumors showed loss of the wild-type Apc allele in 9 of 16 cases with borderline results in 2 other tumors. In contrast, all gastric and intestinal tumors showed retention of the wild-type E-cadherin allele (Figure 3).
Immunohistochemical analysis was performed on cryosections of tumors from both genotypes with the rat monoclonal antibody DECMA, directed against residues 407-434 of E-cadherin. 39,40 Expression could be detected in all tumors, although the staining intensity was clearly reduced compared with the normal intestinal mucosa (Figure 4). The staining pattern was similar for tumors of both genotypes with no obvious quantitative differences. This was also the case for a rabbit polyclonal ␤-catenin antibody (Figure 4).
Both the LOH and immunostaining results with the DECMA antibody suggest that wild-type E-cadherin is still expressed in Apc ϩ/1638N /E-cad ϩ/Ϫ tumors. However, it is plausible that a truncating mutation in the second half of the protein disrupts E-cadherin function. Therefore, 10 paraffin-embedded Apc ϩ/1638N /E-cad ϩ/Ϫ intestinal tumors were stained with a monoclonal antibody raised against the COOH-terminal 150 residues of Ecadherin. All tumors had retained immunoreactivity, strongly suggesting that wild-type E-cadherin is still present in the majority of tumors (data not shown).

Discussion
Complete loss of E-cadherin function is generally associated with the acquisition of an invasive phenotype. [21][22][23] However, inactivation has also been observed in early-stage gastric and lobular breast cancers characterized by a diffuse growth pattern. [33][34][35] We now show that heterozygosity for an inactivating E-cadherin mutation in Apc1638N animals results in a strong increase of the number of gastrointestinal tumors without clearly affecting malignant potential. Interestingly, no LOH at the E-cadherin locus was observed in the corresponding tumors, and E-cadherin expression was still detectable by immunohistochemistry. No diffuse growth pattern was observed in any of the tumors, as is frequently the result of a complete inactivation of E-cadherin function. [43][44][45] Thus, introduction of the E-cadherin mutation in Apc1638N animals mainly seems to affect tumor formation rather than tumor progression.
A number of not mutually exclusive mechanisms can be envisaged to explain this increase in tumor multiplicity. First, E-cadherin haploinsufficiency could enhance Apc-driven tumorigenesis by increasing the pool of cytoplasmic ␤-catenin available for Wnt signaling. Defects in the Wnt signaling pathway underlie intestinal tumor formation, as indicated by the presence of mutations in either APC or ␤-catenin in most colorectal cancers. 7, 16 We have recently shown that (1) the tumor-suppressing activity of Apc lies in its ability to prevent ␤-catenin to signal through the Wnt pathway, 13 and (2) that Apc mutations are selected upon their capacity to inactivate the ␤-catenin down-regulating activity in intestinal tumorigenesis in the mouse. 46 Tumor formation is most efficient if all Apc domains involved in ␤-catenin regulation are removed. In other words, once both copies of Apc become mutated, a stronger ␤-catenin/Wnt signal to  the nucleus increases the chance of intestinal tumor formation.
Although ␤-catenin has distinct roles in cell adhesion and signaling, [47][48][49] several investigations have shown that cadherins may act as modulators of the Wnt signal-ing pathway. Increasing the level of cadherin in Xenopus embryos depleted the cytoplasmic pool of ␤-catenin and interfered with Wnt signal-mediated axis formation. 47,50 Similar studies in Drosophila showed that overexpression of cadherin mimicked the wingless phenotype observed when Armadillo, the Drosophila homologue of ␤-catenin, is mutated. This defect could be rescued by increasing Armadillo levels. 48 In colorectal cancer cells, expression of cadherin derivatives inhibited ␤-catenin-mediated transactivation of a Wnt reporter. 51,52 The opposite effect, i.e., increasing the signaling pool of ␤-catenin by reducing cadherin levels, has also been shown. 53 Haploinsufficiency for cadherin in Drosophila diminished the wingless defect of an intermediate mutant of Armadillo, presumably by releasing some of the normally membrane-bound Armadillo and making it available for signaling. Thus, a considerable amount of experimental evidence supports a substantial role for cadherins in modulating Wnt signal-related processes. Accordingly, the reduced amount of E-cadherin in Apc ϩ/1638N /E-cad ϩ/Ϫ animals could result in an enlarged intracellular pool of ␤-catenin, thus enhancing Wnt signaling and the likelihood of tumor formation to occur. The increased signaling activity most likely becomes apparent only once Apc is temporarily inactivated by an extracellular Wnt signal or when both copies of Apc become mutated. That Apc inactivation still appears to be a rate-limiting step for tumor development is supported by the loss of the wild-type Apc allele in 9 -11 of 16 gastric and 12 of 16 intestinal Apc ϩ/1638N /E-cad ϩ/Ϫ tumors, similar to LOH frequencies previously observed in Apc1638N tumors on various genetic backgrounds. 10,41,42 Moreover, heterozygosity at the E-cadherin locus itself is not sufficient to trigger tumorigenesis, as confirmed by the lack of tumors in E-cad ϩ/Ϫ mice. However, no clear increase in ␤-catenin levels was observed in the Apc ϩ/1638N /E-cad ϩ/Ϫ tumors in comparison with Apc ϩ/1638N tumors, although it should be noted that the immunofluorescence method used is not appropriate for quantitative comparisons. Further experimentation is needed to show that the reduced E-cadherin levels affect Wnt signaling. E-cadherin participates in a complex network of interactions between adhesion receptors, cytoskeletal proteins, and several signaling pathways. 26 -29 Therefore, changes at adherens complexes may result in alterations of cell shape and polarity, proliferation, cell motility, attachment to the extracellular matrix, and cytoskeletal organization. This was also shown by the introduction of a dominant negative N-cadherin transgene in the murine intestinal tract. 54 Expression of the transgene interfered with normal cadherin function, resulting in a reduced epithelial barrier function, altered proliferation, migration, and apoptosis along the crypt-villus axis, leading to inflammatory bowel disease and development of adenomas. In the normal intestinal mucosa of Apc ϩ/1638N /Ecad ϩ/Ϫ animals, no such abnormalities were observed, suggesting that sufficient E-cadherin function is available to support its normal functions. Nevertheless, we cannot exclude that an undetected change in cell-to-cell adhesion and subsequent changes in cellular functions may be responsible for the observed increase in tumor numbers.
The 5-fold increase of gastric tumors in Apc ϩ/1638N /Ecad ϩ/Ϫ mice is interesting in view of the recent observation suggesting that germline mutations at the human E-cadherin gene are responsible for familial diffuse gastric cancer. 30 -32 Somatic E-cadherin mutations have also been observed at high frequency in sporadic gastric cancers of the diffuse type 34,44,55,56 and in invasive, lobular breast carcinomas. 35,43,45 In most cases, these mutations are accompanied by a loss of the second allele. Thus, as a consequence of the complete loss of E-cadherin function, there seems to be a morphologic shift to a diffuse growth pattern. No mutations have been detected in intestinaltype gastric lesions 55,57,58 and colorectal cancers. 59 Nevertheless, a high proportion of colorectal and some intestinal-type gastric cancers show reduced expression of E-cadherin. [57][58][59][60][61][62] The latter has been explained by a number of mechanisms, including transcriptional down-regulation either by promoter methylation or binding of transcriptional repressors such as Snail, which are upregulated in tumor cells. 23,63,64 This resembles the situation of the gastrointestinal tumors of the Apc ϩ/1638N /Ecad ϩ/Ϫ animals, in which no diffuse growth pattern has been observed while retaining E-cadherin immunoreactivity at reduced levels, which suggests that inactivation of only one E-cadherin allele contributes to tumor development.
At 6 -9 months of age, many Apc ϩ/1638N /E-cad ϩ/Ϫ animals showed severe signs of morbidity as a result of the high tumor multiplicity. This precluded analysis of the tumor phenotype at later time points. However, at the time of analysis, no obvious differences in histopathologic appearance were observed between the tumors of both genotypes, as also confirmed by the similar tumor size and degree of invasion. No E-cadherin loss was observed among Apc ϩ/1638N /E-cad ϩ/Ϫ tumors, and E-cadherin expression in the tumors was still detectable by immunohistochemistry. Therefore, the level of E-cadherin in Apc ϩ/1638N /E-cad ϩ/Ϫ mice seems to suffice to prevent further progression toward a more invasive tumor phenotype compared with the Apc ϩ/1638N controls.
However, besides a complete inactivation of E-cadherin, the acquirement of a metastatic phenotype requires the functional dysregulation of additional proteins. 65 In conclusion, introduction of the E-cadherin mutation in Apc1638N animals affects tumor initiation in the gastrointestinal tract rather than progression toward malignancy. This is most likely a result of the overlapping functions of E-cadherin in controlling both Wnt signal transduction and cell adhesion.