β-Cateninand ras Oncogenes Detect Most Human Colorectal Cancer¹

Colorectal cancers develop as a result of different combinations of genetic alterations, epigenetic changes, and post-translational modifications (1, 2). Such molecular alterations affect oncogenes, tumor suppressor genes, and the genes encoding enzymes critical for mismatch and excision repairs (1, 2). Epigenetic events are exemplified by gene methylation, which alters the CpG islands frequently distributed in promoter sequences of key regulatory genes (2, 3, 4), polymorphism in genes for detoxification enzymes (5, 6), and altered expression and activity of the molecules involved in intracellular signal transduction (7). Phosphorylation and ubiquitin-mediated proteasomal degradation play key roles in post-translational modification of proteins that regulate function and stability of cellular regulatory molecules (8, 9). Among genes and molecules whose alterations determine the biological characteristics and malignant potential of individual tumors and clinical outcomes, one of the best characterized is K-ras, which encodes a 21kD protein that mediates oncogenic Ras/GTP signaling (10). K-ras has been shown to be a powerful biomarker in molecular diagnosis and risk assessment in colorectal, pancreatic, and lung cancers (11).

Current interest in oncogenes in colorectal cancers has been shifting to β-catenin (12) from the paradigm represented by the K-ras proto-oncogene. β-catenin was identified as a key factor in cell adhesion machinery that cooperatively associates with E-cadherin (13). This molecule is also central in the Wnt/β-catenin/Tcf signaling pathway, which is known to play roles both in normal embryonic development and differentiation and in malignant cellular transformation (14, 15). A recent series of studies in cell lines and gene knockout or transgenic animal models demonstrated that β-catenin becomes susceptible to oncogenic activation by escaping from ubiquitin-mediated proteasomal degradation, a process that includes many regulators and effectors that maintain a low but steady-state level of β-catenin expression (14, 15, 16, 17). Thus, in the oncogenic β-catenin phenotype, its localization is subcellular, shifting from cell membrane to nucleus and/or cytoplasm, as readily detected by immunoblotting and immunocytochemistry materials (12). However, little was known about this oncogene’s activation in clinical cases until our recent study demonstrated the prevalence of β-catenin oncogene in the largest series of clinical colon cancers reported to date (18). In particular, we found that selective oncogenic activation of β-catenin in the tumor invasion front, as represented by its unique expression pattern, is a reliable independent factor that can detect a subset of colon cancer patients highly susceptible to tumor recurrence and thus likely have a less favorable survival rate (18).

In human colorectal cancers, K-ras and β-catenin (CTNNB1) are considered the twin major oncogenes. Common to them is activation of transducing oncogenic signaling cascades. Activation of either pathway results in cell proliferation and inhibition of apoptosis, thereby promoting tumor development and progression (15, 17). Like multidirectional cross-talking among different signaling networks in the cell (7), the two oncogenic cascades’ signaling may have syngeneic or synergistic effects on tumor development and progression (19). The study reported here was undertaken to compare mutational activation of K-ras with distinct patterns of β-catenin activation in the primary tumors of colorectal cancer patients to determine whether the signaling activities of the two oncogenes are related and the influence of such a relationship on clinical and histopathological characteristics of the patients’ tumors, as well as on clinical outcomes.

The subjects of the study were 74 patients with colorectal cancer who underwent surgical treatment in the Department of Surgical Oncology at the Cancer Research Institute Hospital and later at Kanazawa University Hospital between 1998 and 2002. The patients included 41 men and 33 women ranging in age from 33 to 93 years (mean, 68.3 years). The primary tumors analyzed here were distributed as follows: 27 tumors in the cecum and ascending segment, 5 in the transverse segment, 2 in the descending segment, 20 in the sigmoid segment, and 20 in the rectum. The Japanese Classification of Colorectal Carcinoma (20) was used to describe the gross findings and histopathological characteristics of the primary tumors (20). Gross observation showed a protruding appearance (type 1) in 12 tumors, localized and expansive growth with ulceration (type 2) in 60 tumors, and infiltrative proliferation (type 3) in 2 tumors. Histological types of the primary adenocarcinomas were determined as well differentiated in 35 cases, moderately differentiated in 34 cases, and poorly differentiated (including mucinous adenocarcinoma) in 5 cases. Determination of the tumors’ Dukes’ stage at surgery by routine clinical examinations and pathology diagnosis gave the following results: (a) 4 patients in Dukes’ stage A; (b) 36 in Dukes’ B; and (c) 34 in Dukes’ C (Table 1). Distant metastases (i.e., liver and lung) were found in 4 patients at surgery. Postoperative surveillance revealed that 9 patients developed recurrent tumors in distant organs after surgery.

Each patient included in this study signed a written informed consent, and the Institutional Review Board of Kanazawa University Graduate School of Medical Science approved all protocols.

After routine gross observation of the surgical materials, fresh and multiple sets of paired normal and tumor tissues were sampled, immediately snap frozen in liquid nitrogen, and stored at −80°C until examination. Surgical materials were then fixed with neutral-buffered formalin. In each case, paraffin-embedded tissue sections representative of the histopathological characteristics of the tumors were prepared on silica-coated slide glasses for immunohistochemical analysis of β-catenin activation after routine histological examination to determine tumor stage. From each of paired, fresh normal, and tumor samples, genomic DNA was extracted by proteinase K digestion and purified by serial treatments with phenol and chloroform, as described elsewhere (21).

It is well documented that β-catenin translocated in the nucleus functions as an oncogene through transactivation of its target genes and that its transcriptional activity in complex with members of the Tcf/Lef family depends on the abundant presence of particular members of this family (i.e., TCF-4 and LEF-1) and the presence of other proteins (i.e., groucho proteins) in the nucleus (12, 14, 15, 16). Accordingly, we determined activation of β-catenin in the tumor cells on the basis of its NA.⁴

For each case, the most representative section reflecting the major features of the primary colorectal tumor (i.e., histological type and depth of invasion) were selected for immunohistochemical examination to determine the expression of β-catenin, using a mouse monoclonal antibody to β-catenin (Transduction Laboratories, Lexington, KY) for the standard avidin-biotin-peroxidase complex method as described in our previous study (18). As described in the supplier’s instructions, the antibody epitope that binds to β-catenin is the COOH terminus (amino acids 571–781 of mouse β-catenin), which contains a transcription domain, and this antibody is known to cross-react with the human homologue. Working dilutions of the primary antibody to β-catenin and the biotinylated second antibody to mouse IgG (Vector Laboratories, Burlingame, CA) were 1:100 and 1:1000, respectively. The immunoreactive signal was developed using 3′-diaminobenzidine-tetrahydrochloride catalyzed by the avidin-biotin-peroxidase complex (DAKO, Glostrup, Denmark). As a negative control, the primary antibody was replaced by nonimmune IgG1 (DAKO). We immunostained paraffin sections of normal esophageal mucosa as a positive control for membranous expression of β-catenin and of normal colorectal mucosa adjacent to the tumor as an internal positive control.

β-catenin expression was classified into three patterns we recently determined (18) as follows: (a) membranous expression, similar to that in normal colonic or rectal crypts; (b) NAd, defined as cancer cells with β-catenin-positive nuclei distributed throughout the tumor (Fig. 1,A); and (c) NAinv showing NA by cancer cells in the deepest parts of the tumor, whereas cancer cells elsewhere in the same tumor showed only membranous expression (Fig. 1 B). Using these definitions, two well-trained pathologists (A. O. and K. Y.), without knowledge of the presence of K-ras mutations in the tumors or of the patients’ clinical and pathologic parameters, independently reviewed expression patterns of β-catenin in each specimen.

Mutations in codons 12 and 13 of the K-ras gene were detected by mismatched primer-mediated PCR-RFLP analysis using restriction enzymes MvaI (Takara, Tokyo, Japan) and BglI (TOYOBO, Kyoto, Japan) specific for RFLP in codons 12 and 13, respectively. Briefly, 200 ng of genomic DNA extracted from each tumor were PCR amplified with each set of mismatched primers designed to enable the RFLP analysis of the two codons (21, 22). When PCR products were separated on native PAGE after digestion with the restriction enzymes, PCR products encoding the wild-type and mutant sequences were distinguished as 114- and 143-bp fragments, respectively, in the case of codon 12, and 125- and 157-bp fragments, respectively, in the case of codon 13 (Fig. 2; Refs. 22 and 23). Mutant alleles detected by PCR-RFLP were then selectively amplified by the enriched PCR (21). To determine the types of mutations, the enriched mutant products were sequenced using a Dye Terminator Sequencing Kit and an ABI PRISM 310 Genetic Analyzer (Perkin-Elmer ABI, Foster City, CA), according to the manufacturer’s protocol.

In every run of PCR-RFLP analysis, DNAs for wild-type and mutant controls were amplified in parallel with DNA samples. The wild-type DNA in codons 12 and 13 was obtained from human placenta (Sigma, St. Louis, MO). The DNA for positive control of codon 12 mutant was derived from the human colon adenocarcinoma cell line SW480 (American Type Culture Collection, Rockville, MD) that has been found to harbor a homozygous mutation in codon 12 (24). For a positive control of codon 13 mutant, we used DNA samples obtained from a clinical colon cancer confirmed to have a codon 13 mutation in our previous study (23). In cases in which mutations were found, we repeated the steps in the analysis from PCR amplification to sequencing at least two times.

Statistical analysis of all data obtained in this study was based on the χ² test. A P < 0.05 was used in determining statistically significant differences.

In the 74 colorectal cancers available for analysis, membranous expression of β-catenin similar to its localization in nonneoplastic colorectal crypts was found in 35 cases (47%) and NA in 39 cases (53%). The latter group included 20 cases (27%) showing the NAd pattern and 19 cases (26%) with NAinv pattern of β-catenin expression (Fig. 1, A and B). Compared with the findings of our previous study (18), although the incidence of tumors showing NA (both NAd and NAinv patterns) was similar, the frequency of tumors with the NAinv expression pattern was greater in the present series. The latter was partly attributed to the difference in tumor site distribution, because rectal cancers were included in the present study but not the earlier one (18). Mutational activation of K-ras proto-oncogene was detected in the tumors in 31 (42%) of 74 patients. In these cases, mutations in codons 12 and 13 were detected in 27 and 4 cases, respectively.

The next step was to relate the presence of the K-ras mutation and different patterns of β-catenin activation to clinical and histopathological characteristics of the patients (Table 1). There was no significant correlation between the presence of mutational activation of K-ras and any clinicopathological parameters. However, certain correlations were found between the different patterns of β-catenin activation and clinicopathological characteristics. Strikingly, there was a significantly higher association of the NAinv activation pattern with advanced tumor stage (P = 0.0005) and with the development of distant metastases (P = 0.0064). Additionally interesting were statistically significant differences in the incidence of β-catenin activation according to tumor site in the large bowel. We divided the large bowel anatomically into three segments: (a) the right side of the colon, including cecum, ascending, and transverse colon; (b) the left side of the colon, including descending and sigmoid colon; and (c) the rectum. β-catenin was more frequently activated in tumors of the left side of the colon and rectum than in those of the right side of the colon. Although no statistical correlation was found between β-catenin activation patterns and histological types of the primary tumors, none of three mucinous adenocarcinomas showed NA of β-catenin.

We compared K-ras mutations and the different patterns of β-catenin activation in the primary tumors to determine the presence of interaction or cross-talking between the ras- and β-catenin-mediated signaling cascades in clinical colorectal cancers. Contrary to the in vitro results (12, 15), there was an inverse, although statistically weak (P = 0.046), correlation between mutational activation of K-ras and oncogenic activation of β-catenin. The sites or types of K-ras mutation did not affect this correlation (data not shown). It is intriguing, however, that most (78%) of the clinical colorectal cancers showed activation of either the K-ras or β-catenin oncogene (Table 2).

An analysis of the influence of mutational activation of K-ras and pattern of oncogenic β-catenin activation on tumor stage is presented in Table 3. Interestingly, of a small number of patients showing both K-ras activation and the NAinv pattern of β-catenin activation, 7 of 8 patients were in Dukes’ stage C at surgery (P = 0.0727). As summarized in Table 4, 4 of these 8 patients, including 1 (case 5) with ascending colon cancer in Dukes’ stage B at surgery, developed simultaneous and/or metachronous metastases to the liver and lungs. Another patient (case 4) had a second primary cancer of the right ureter 4 years and 8 months after wide resection of the colon. On the other hand, of 11 patients with tumor showing wild-type K-ras and NAinv pattern of β-catenin activation, 4 developed distant metastases (clinical data not shown in Table 4).

The present study highlights the relationship between the two major oncogenes, K-ras and β-catenin, and their roles in, and influence on, the development and progression of clinical colorectal cancers. Contrary to the suggestion of previous in vitro studies (25, 26, 27, 28), our results pointed to a seemingly independent activation of K-ras and β-catenin in human colorectal cancers but also displayed a trend toward a statistically inverse correlation of the frequencies of activation of the two oncogenes. To put it another way, activation of either oncogene could detect most (∼80%) of the cancer patients. The latter finding suggests that detecting these oncogenes in clinical materials/samples may have application to molecular diagnosis of colorectal cancers. The most striking finding was that synchronous activation of the two oncogenes identified exclusively a small subset of patients with advanced Dukes’ stage tumors, in which metastases to distant organs are frequently seen.

On the basis of accumulating evidence in vitro for interactions of effectors downstream to the respective signal transduction cascade (14, 15, 16, 17, 26, 27, 28), one may expect the eventual demonstration of a link between ras-mediated and Wnt/β-catenin-signaling pathways. One of the possible mechanisms that could link them was reported to be tyrosine phosphorylation of β-catenin by activated ras (25). It is known that colorectal cancers express several growth factors and their corresponding receptors, including, e.g., epidermal growth factor and epidermal growth factor receptor (or c-cerbB-2) and hepatocyte growth factor and hepatocyte growth factor receptor (or c-met; Refs. 29 and 30). Similar to the effects of these growth factors on β-catenin (31, 32), the effect of the activated ras oncogene product promotes tyrosine phosphorylation of β-catenin and disrupts organization of a complex of cell adhesion molecules, thereby leading to shifting subcellular localization of β-catenin (25). Alternatively, a possible mechanism might be ras activation of protein kinase B/Akt, a known inhibitor of glycogen synthase kinase-3β via its phosphorylation (33). This mechanism is attributed to the role of glycogen synthase kinase-3β in targeting β-catenin for ubiquitination and subsequent proteasomal degradation (15, 16). Considering more complex mechanisms integrated in the process of colorectal tumor development and progression (1, 2), an absence of straightforward interaction or transactivation between these signaling pathways is conceivable and no doubt occurs in clinical cancers. However, this does not imply that the two oncogenic signaling pathways never interact or display connections in vivo in human colorectal cancers.

Interestingly, the present study shows the tumor site-specific prevalence of β-catenin activation, whereas there was no significant difference in the incidence of K-ras mutations in the tumors in different sites of the large bowel. When we divided the large bowel anatomically into three parts, the results indicated more frequent activation of β-catenin—both NAd and NAinv patterns—in tumors of the recto-sigmoid segments than those of the right side of the colon. This is partly because subjects of the present study had both colon and rectal cancers, unlike the study we reported previously (18). Evidence has been accumulating that molecular mechanisms involved in carcinogenesis and molecular phenotypes differ in the tumors arising in the proximal and distal segments of the large bowel (1). The presence of different molecular pathways to colorectal tumorigenesis is exemplified by the fact that cancers of mutator phenotypes preferentially occur in the proximal (right side) of the colon, whereas the adenoma-carcinoma sequence phenotype represents cancers in the distal (left side) colon and rectum (34, 35, 36). A minor but intriguing observation in the current study is infrequent activation of the β-catenin oncogene in poorly differentiated and mucinous adenocarcinomas, although the number of patients analyzed was too small for statistical significance. This may be a consequence of the fact that mucinous adenocarcinoma is more frequently found in the proximal segment of the colon (37). Along with variations in their evolution as well as molecular and morphological phenotypes, clinical characteristics and behaviors of large bowel tumors differ according to whether they arise in the proximal or distal segments (38). From this perspective, differential activation of the β-catenin oncogene is among the molecular alterations that determine tumor phenotypes and malignant potential, as well as clinical outcome.

Regardless of the possible presence of cross-links between the K-ras and β-catenin signaling pathways, combined analysis of the two oncogenes could detect most (78%) colorectal cancers in the present study. This analysis will provide a powerful tool for early diagnosis and screening of colorectal cancers (39, 40) when it is applied to clinical samples that reflect the entirety of the large bowel, i.e., stool, colon washings, or peripheral blood. Mutational activation of K-ras is frequently and effectively documented in colorectal (40–50%), pancreas (70–90%), and lung (25–50%) cancers (10, 41). A line of studies demonstrated that cancers of these types share clinical characteristics of rapid increase in incidence and high mortality because of difficulty in diagnosis at early stages and molecular biological traits of frequent activation by oncogenic signaling, exemplified by K-ras activation, in the process of tumor development. As in a previous study (42), our current preliminary analysis showed no oncogenic activation or loss of β-catenin in pancreas cancers.⁵ In the case of lung cancer, several studies of β-catenin alteration in the tumor have focused on impairment of its physiological function as a component of the cell adhesion complex cooperating with E-cadherin, other types (α- and γ-) of catenins, and members of the cytokeratin family (43, 44, 45, 46). Thus, the evidence currently available supports the hypothesis that β-catenin acts as an oncogene in colorectal cancer, poses as a bystander in pancreas cancer, but behaves as a tumor suppressor in lung cancer. Given that K-ras and β-catenin have such different roles in different cancers, a combination of these genetic markers in clinical samples could differentiate patients with and potentially individuals who are at risk of developing colorectal, pancreas, or lung cancers.

Clinically, the present study has reproduced earlier findings in our preliminary study (18) as to the importance of the specific pattern (NAinv) of β-catenin activation in the primary tumors in relation to assessment of the malignant potential of colorectal cancers. Most striking in the current study was the ability of synchronous detection of activated K-ras and the specific pattern (NAinv) of β-catenin activation to identify a group of patients with colorectal cancer who are intractable when given standard therapy. Both oncogenes mediate their specific signaling cascades so as to result in increased cell proliferation and inhibition of apoptosis (15, 17). An example of cooperation between the two oncogenes (47) has been demonstrated for Myc; a set of reports shows that ras enhances its stability (48) and that β-catenin transactivates its transcription (49). Therefore, regardless of the possible presence of cross-linking between the two oncogenes, synchronous activation of both may exert syngeneic and/or synergistic effects in enhancing invasion and metastasis of colorectal cancers. In all, comparative and combined analysis of the two major oncogenes is promising in molecular diagnosis and determining the malignant potential of colorectal cancers.

We thank all staff members of the Divisions of Diagnostic Molecular Oncology and Surgical Oncology for providing materials, clinical and pathologic data, and critical comments and discussions, Atsuko Shimizu-Kaneda for excellent technical assistance, and Michael Meyer (New York, NY) for editorial assistance.