Markers of the APC/β-Catenin Signaling Pathway as Potential Treatable, Preneoplastic Biomarkers of Risk for Colorectal Neoplasms

Despite advances in screening and treatment, colorectal cancer (CRC) remains the second leading cause of cancer deaths in the United States (1). The etiology of sporadic CRC is predominately rooted in dietary and lifestyle behaviors (2, 3), suggesting that it may be preventable. The molecular basis of colorectal carcinogenesis is becoming clearer (3), but has yet to be exploited to yield validated, treatable, preneoplastic biomarkers of risk for colorectal neoplasms.

The adenomatous polyposis coli (APC) protein, β-catenin, and E-cadherin are appealing, potentially treatable, preneoplastic biomarkers of risk for colorectal adenomas. Impaired APC function/expression, which is involved in approximately 80% to 90% of sporadic CRCs (4), results in increased cytoplasmic β-catenin which can translocate to the nucleus and bind with T-cell factor (TCF) transcription factors and activate target genes responsible for cell proliferation and differentiation (4). Also, β-catenin is a multifunctioning signaling protein, which, along with α-catenin, binds to the cytoplasmic tail of the calcium-dependent cell adhesion protein E-cadherin, linking E-cadherin to actin filaments and promoting cell adhesion and differentiation. E-cadherin may also antagonize APC/β-catenin signaling by sequestering β-catenin at cell adhesion junctions (5).

In normal colorectal mucosa APC, β-catenin, and E-cadherin are all strongly expressed—APC primarily in the cytoplasm, and E-cadherin and β-catenin primarily at the cell membrane. During the adenoma–carcinoma sequence, APC and E-cadherin expression markedly decrease (although the decrease in E-cadherin tends to occur in later stages; refs. 6–8), and β-catenin expression steadily increases and translocates from the membrane to the cytoplasm and eventually into the nucleus (6).

We know of no other human studies that have investigated APC, β-catenin, and E-cadherin expression in normal colorectal mucosa as potentially treatable, preneoplastic biomarkers of risk for sporadic colorectal neoplasms. We hypothesized that increased APC and E-cadherin expression in the normal colorectal mucosa would be associated with lower, and increased expression of β-catenin would be associated with higher, risk for adenoma. To address this, as reported herein, we conducted a colonoscopy-based case–control study of sporadic colorectal adenoma in which we evaluated normal colorectal mucosa expression of APC, β-catenin, and E-cadherin as potentially modifiable biomarkers of risk for colorectal adenoma.

As previously reported (9–11), the Markers of Adenomatous Polyps II (MAP II) study is a colonoscopy-based case–control study to investigate potentially treatable biomarkers of risk for incident, sporadic colorectal adenomas. Study participants were recruited through the usual scheduling of outpatient elective colonoscopies at Consultants in Gastroenterology, a large private practice gastroenterology group in Columbia, SC. Briefly, eligibility requirements consisted of being 30 to 74 years of age, English speaking, and a resident of the Columbia, SC, area. Exclusion from participation included contraindications to colonoscopic biopsies, previous adenomatous polyps, familial adenomatous polyposis, inflammatory bowel disease, incident CRCs, or a history of cancer other than nonmelanoma skin cancer. Cases were defined as being diagnosed with a first ever adenomatous polyp at their elective outpatient colonoscopy. Controls were persons found to have no colonic neoplasms at the elective outpatient colonoscopy. Hyperplastic polyps were not considered in the criteria for study inclusion or exclusion or in the assignment of case–control status. Over 5 months, 351 patients were identified; of these, 305 (86.6%) were eligible to participate following initial screening, of whom 232 (76%) were successfully contacted and provided informed consent; of these, 203 (87.5%) met final eligibility criteria; of these, 49 (56.3%) were diagnosed with incident adenomas, yielding a final sample size of 49 cases and 154 controls. Five participants were later excluded from dietary analyses because of implausibly low (females <500 kcal/d, males <800 kcal/d) or high (females >3,500 kcal/d, males >4,000 kcal/d) self-reported total energy intake. Because of limited funding, APC expression was evaluated only in 87 participants (42 cases, 45 controls), β-catenin in 77 participants (36 cases, 41 controls), and E-cadherin in 86 participants (40 cases, 46 controls).

Before colonoscopy, patients completed mailed questionnaires about dietary, lifestyle, and other potential risk factors for CRC. Physical activity was assessed using a modified Paffenbarger questionnaire (12). Diet and nutritional supplement intakes were assessed using a Willett Food Frequency Questionnaire (13, 14).

Participants were asked to abstain from aspirin use seven days before colonoscopy. Colonoscopy of all participants was done in the usual manner following a 12-hour fast and polyethylene glycol bowel cleansing preparation. Six approximately 1-mm thick “pinch” biopsy specimens were taken from the normal-appearing rectal mucosa 10 cm above the anus. On 20% of participants, biopsies from the mid-sigmoid and proximal ascending colon were also collected. No biopsies were taken within 4.0 cm of a polypoid lesion. Biopsies were placed onto a strip of bibulous paper and immediately placed in PBS, oriented, transferred to 10% normal-buffered formalin for 24 hours, and then transferred to 70% ethanol. Then, within a week, the biopsies were processed and embedded in paraffin blocks (2 blocks of 3 biopsies per colon site per participant). For any polyp removed at colonoscopy, colon site, in vivo size, and shape were recorded, and histologic information was further reviewed by the study index pathologist using standard criteria (15).

For each person, for each colon site, 5 slides with 4 levels of 3-μm thick biopsy sections taken 40-μm apart were prepared for each antigen, yielding a total of 20 levels for each antigen. Antigen retrieval was conducted by placing the slides in a preheated Pretreatment Module (Lab Vision Corp.) with 100× citrate buffer (pH 6.0; DAKO S1699, DAKO Corp.) and steaming them for 40 minutes. Following antigen retrieval, slides were immunohistochemically processed in a DAKO Automated Immunostainer (DAKO Corp.) using a labeled streptavidin-biotin method for APC (Calbiochem, OP80; 1:70 dilution), β-catenin (Transduction Laboratories 610154; 1:300 dilution), and E-cadherin (Zymed 33-4000; 1:50 dilution). No slides were counterstained. After processing, slides were coverslipped with a Leica CV5000 Coverslipper (Leica Microsystems, Inc.). Each staining batch contained positive and negative control slides, which were treated identically to the patients' slides except that antibody diluent was used rather than primary antibody on the negative slides.

A detailed description of the protocol used to quantify biomarker labeling optical densities (“biomarker expression”) in normal colon crypts was previously described (16). Briefly, a “scorable” crypt was defined as an intact crypt extending from the muscularis mucosa to the colon lumen (17). Before “scoring,” the negative and positive control slides were checked for staining adequacy. The major equipment and software for the image analysis procedures included: personal computer, light microscope (Olympus BX40; Olympus Corporation) with appropriate filters and attached digital light microscope camera (Polaroid DMC Digital Light Microscope Camera; Polaroid Corporation), digital drawing board, ImagePro Plus image analysis software (Media Cybernetics, Inc.), our in-house developed plug-in software for colorectal crypt analysis, and Microsoft Access relational database software (Microsoft Corporation).

Evaluation of biomarker expression consisted of the same technician, who was blinded to case–control status, cleaning all slides, selecting the 2 of the 3 biopsies with the most scorable crypts per biopsy, creating background correction images for each slide scored, capturing 16-bit grayscale images of crypts at 200× magnification, and tracing the border of the “hemicrypt” (one half of the crypt). The program then divided the outlined hemicrypt into equally spaced segments that corresponded to the average width of colonocytes and measured the optical density of the labeling across the entire hemicrypt and within each segment, adjusting for the background (Fig. 1A–D). The technician then repeated this process for the adjacent hemicrypt and proceeded to the next crypt, level, biopsy, and/or slide. The goal was to score 16 to 20 hemicrypts on 2 of 3 biopsies per biomarker per colon site.

Reliability control was conducted by selecting samples of previously analyzed slides (10%) to be reanalyzed by the technician. The technician was blinded to the selection. Intrareader reliability for APC, E-cadherin, and β-catenin was >0.90 throughout.

Serum 25(OH) vitamin D₃ [25(OH)D₃] was measured using liquid chromatography/tandem mass spectrometry, as previously described (18).

All statistical analyses were done using Statistical Analysis System software (version 9.2; SAS Institute, Inc.). The characteristics of the cases and controls were compared using the t test for continuous variables and Fisher's exact test for categorical variables.

The mean labeling optical density (“expression”) of each biomarker on each study participant was calculated by summing the biomarker's expression for all analyzed crypts and dividing by the total number of analyzed crypts. Biomarker expression was standardized to adjust for possible staining batch effects by dividing an individual's mean biomarker expression by the mean biomarker expression of the staining batch in which the individual's sample was included. As a sensitivity analysis of our batch standardization, we used a robust method that conditioned on batch and dichotomized on the batch-specific median in controls (19). Measures of biomarker expression in functional distinct crypt zones selected a priori were the upper 40% of the crypts (differentiation zone), the lower 60% of the crypts (proliferation zone), and the distribution index (Φh, defined as the ratio of the upper 40% of the crypts to the whole crypt).

The distributions of batch-standardized APC, E-cadherin, and β-catenin labeling optical density along the full length of the crypts were graphically plotted and evaluated using the LOESS procedure. First, each hemicrypt was standardized to 50 sections. Then, the average of each section across all crypts was predicted by the LOESS model separately for cases and controls by colon site. The results were graphically plotted along with the smoothing line.

Potential confounders were evaluated on the basis of the biologic plausibility of their being associated with the biomarker of interest and with colorectal adenomas. All nutrients were energy-adjusted using the residual regression method (20). Continuous variables were dichotomized on the basis of their sex-specific median in the controls.

The ratio of Φh APC to β-catenin expression (Φh APC expression/β-catenin expression) was calculated to investigate the association of combined Φh APC and β-catenin expression (APC/β-catenin score) with adenoma status, adenoma characteristics, and risk factors for colorectal neoplasms. We hypothesized that the APC/β-catenin score would be inversely associated with having an adenoma. E-Cadherin was not included in the APC/β-catenin score because during carcinogenesis, malfunctioning regulation of β-catenin by APC occurs earlier than E-cadherin downregulation (21). As a sensitivity analysis of the APC/β-catenin score, we also calculated normalized z-scores for Φh APC and β-catenin expression [z = (x − μ)/σ, where x is the biomarker value, μ is the mean biomarker expression in controls, and σ is the SD of the biomarker expression in controls] and then the β-catenin z-score was subtracted from the Φh APC z-scores (APC/β-cat z-score).

Logistic regression was used to calculate ORs with 95% confidence intervals (CI) to evaluate associations of batch-standardized expressions of APC, β-catenin, and E-cadherin, and of the APC/β-catenin score (hereafter referred to as “the biomarkers”) with adenoma; for these analyses the biomarkers were dichotomized on the basis of the mean values in the controls. For analyses of differences in mean expression of the biomarkers according to risk factors for colon cancer, given the small sample size and similarity of results across cases and controls, cases and controls were combined in analysis of covariance (ANCOVA) models with inverse probability weighting according to the relative proportions of cases and controls measured for a given biomarker. However, because there were differences in the biomarkers by nonsteroidal anti-inflammatory drug (NSAID) use according to case–control status, for this analysis, cases and controls were analyzed separately. The results of the associations between E-cadherin and risk factors for CRCs are not presented as no apparent difference in E-cadherin expression between adenoma cases and controls was found. All models were adjusted for age and sex. Because of limited sample sizes for the ascending and sigmoid colon, only the results for the rectal mucosa are presented in the tables.

Selected characteristics of the study population are presented in Table 1. Relative to controls, cases, on average, tended to be older and to have lower serum 25(OH)D₃ concentrations, greater total energy, fat, and processed meat intakes, and lower total folate and calcium intakes. Cases also tended to be less likely to have a history of CRC in a first-degree relative and to regularly take a NSAID; however, only total energy consumption statistically significantly differed between cases and controls. When restricted to cases and controls available for specific biomarkers, the case–control differences did not appreciably change (data not shown).

The APC/β-catenin score in rectal crypts of controls was 14.3% greater (P = 0.02) than in cases, and having a higher APC/β-catenin score in rectal crypts was inversely associated with adenomas (OR, 0.40; 95% CI, 0.14–1.14; Table 2). This association did not appreciably change with multivariate adjustment.

APC expression in cases and controls along the full lengths of colorectal crypts was lowest at the crypt bases, and, beginning at approximately the upper 40th percentile of crypts, expression sharply increased toward the crypt apex (Fig. 2A). Mean total APC expression in rectal crypts did not appreciably differ between cases and controls (data not shown); however, Φh APC was 3.2% lower in cases than in controls and modestly inversely associated with adenomas (OR, 0.75), but these findings were not statistically significant and did not appreciably change with multivariate adjustment (Table 2). The findings for APC expression in the ascending and sigmoid colon were similar to those for the rectum (data not shown).

β-Catenin expression in cases and controls was relatively uniform along the full lengths of rectal crypts (Fig. 2B). Mean total β-catenin expression in rectal crypts was 3.0% higher in cases than in controls and modestly positively associated with being a case (OR, 1.36), but these findings were not statistically significant and did not appreciably change with multivariate adjustment (Table 2). Similar case–control differences were found in specific functional zones of crypts (data not shown).

E-Cadherin expression in cases and controls, similar to β-catenin expression, was relatively uniform along the full lengths of rectal crypts (Fig. 2C). Mean total E-cadherin expression in rectal crypts was 0.7% greater in controls than in cases and modestly positively associated with adenomas (OR, 1.18), but these findings were not statistically significant and did not appreciably change with multivariate adjustment (Table 2). The findings for E-cadherin expression in specific functional zones of crypts and for other colon sites were similar to the findings for total E-cadherin expression in the rectum (data not shown).

The APC/β-catenin score tended to be lower among men (11.1%), participants with a positive family history of CRC in a first-degree relative (10.7%), higher levels of physical activity (14.2%), higher waist-to-hip ratio (WHR; 11.0%), and higher intakes of total fat (14.8%) and processed meats (14.3%); and higher among participants with serum 25(OH)D₃ concentrations >27 ng/mL (17.7%) and greater dietary fiber (13.3%) and total folate intakes (10.5%). However, only the findings for physical activity, 25(OH)D₃ concentrations, total fat intake, and processed meat intake were statistically significant. In cases, regular NSAID use was associated with a higher APC/β-catenin score (11.8%), but in controls, regular NSAID use was statistically significantly associated with a lower APC/β-catenin score (20.2%; Table 3).

Φh APC expression tended to be lower among participants with a positive family history of CRC in a first-degree relative (12.8%), higher levels of physical activity (9.8%), and higher intakes of total fat (9.4%) and processed meats (7.3%); and higher among participants with greater dietary fiber (15.6%) and total calcium (11.3%) and folate intakes (12.4%). However, only the findings for physical activity and intakes of total fat, calcium, fiber, and folate were statistically significant. In controls, regular NSAID use was statistically significantly associated with lower Φh APC expression (11.0%; Table 3).

β-Catenin expression tended to be higher among men (5.2%), participants ≥55 years of age (6.9%), former or current smokers (5.7%) or alcohol consumers (15.7%), and those with higher levels of physical activity (5.9%), higher WHR (8.1%), and greater processed meat intakes (8.3%); and lower among participants with higher total energy intakes (9.0%). However, only the findings for alcohol exposure, WHR, and total energy and processed meat intakes were statistically significant. In controls, regular NSAID use was statistically significantly associated with higher β-catenin expression (10.6%; Table 3).

Alternative methods (19) noted above for accounting for staining batch yielded nearly identical associations between the biomarkers and adenoma. Also, the APC/β-catenin z-score produced results (OR, 0.35; 95% CI, 0.12–0.99) similar to those using the APC/β-catenin score. There were no apparent differences in findings using weighted and unweighted ANCOVA analyses.

We found no apparent differences of expression of APC, β-catenin, or E-cadherin or the APC/β-catenin score among cases according to adenoma subtype, shape, location, or multiplicity (data not shown).

To the best our knowledge, this is the first report to quantify and characterize the distribution of APC, β-catenin, and E-cadherin expression in the normal colorectal mucosa in incident, sporadic, colorectal adenoma cases, and healthy controls. Our results suggest that persons with higher APC combined with lower β-catenin expression in the normal colorectal mucosa may be at lower risk for incident sporadic colorectal neoplasms. Also, our results suggest that these biomarkers may be associated with dietary and lifestyle risk factors for colorectal neoplasms, suggesting that these biomarkers may be modifiable. These findings, which support previous findings suggesting that molecular phenotypic differences in the normal-appearing colorectal mucosa may be associated with increased risk of colorectal neoplasms (9–11) and are modifiable (16, 22–25), are relevant because there are currently no validated treatable, preneoplastic biomarkers of risk for colorectal neoplasms.

In the present study having a higher proportion of APC expression in the upper 40% (differentiation zone) of colorectal crypts (Φh APC) was consistently, but not statistically significantly, greater in controls than in cases at all colon sites. These results are consistent with reports that in CRC cases (26) and those at increased risk for the disease (27, 28), the colorectal crypt proliferation zone (the lower 60%) expands upwardly into the upper 40% of crypts. That E-cadherin expression did not appreciably differ between cases and controls and cases tended to have lower Φh APC expression may suggest that the excess β-catenin expression found in cases could induce a greater proliferative potential than in controls. This idea is further supported by the significant inverse association of the APC/β-catenin score with adenoma. We suggest that the APC/β-catenin score may reflect the potential of β-catenin to activate proliferative genes in colorectal crypts.

A family history of CRCs and increased age are well-established risk factors for colorectal neoplasms (3). Consistent with this, we found a lower APC/β-catenin score and Φh APC in people with a history of CRC in a first-degree relative, and increased β-catenin expression in older participants. Consistent with the slightly higher risk of CRC in men than in women (29), we found men to have a lower APC/β-catenin ratio.

The etiology of CRC is heavily influenced by modifiable dietary and lifestyle behaviors (2, 3); however, we are unaware of any in vivo investigations of associations of dietary and lifestyle behaviors with APC and β-catenin expression in the normal colorectal mucosa. Increased physical activity and NSAID use are consistently reported to reduce risk of colorectal neoplasms (29). Unexpectedly, we found a statistically significant lower APC/β-catenin score among people with higher physical activity, and the association between regular NSAID use and biomarker expression differed according to adenoma status. Among sporadic adenoma cases, regular NSAID use was associated with a higher APC/β-catenin score, but among controls, regular NSAID use was associated with a lower APC/β-catenin score (P_interaction = 0.03). We are unaware of a biologically plausible explanation for these findings, which may have been due to chance. Considering the concomitant higher E-cadherin expression (data not shown), it is plausible that the higher β-catenin expression observed with higher physical activity and NSAID use could represent β-catenin localized to the cell membrane and the cytoplasmic tail of E-cadherin and not higher levels of cytoplasmic or nuclear β-catenin. Elevated body fatness is a well-established risk factor for colorectal neoplasms, and consistent with these observations, we found a lower APC/β-catenin score among people with a larger WHR and body mass index. In line with reports that smoking and alcohol consumption are associated with increased proliferation (30, 31), we found that people who formerly or currently consumed alcohol or smoked had higher β-catenin expression. Higher intakes of total fat and processed meats, both of which are hypothesized to increase risk for colorectal neoplasms (2, 32), were associated with lower APC/β-catenin scores. We found a positive association between serum 25(OH)D₃ and the APC/β-catenin score, which is consistent with growing evidence, suggesting that higher circulating 25(OH)D₃ reduces risk for colorectal neoplasm (33, 34). Higher folate and fiber intakes were associated with a higher APC/β-catenin score. In mice, a diet deficient in folate and other B vitamins reportedly decreased APC expression and increased β-catenin expression (35, 36); and at least one animal study reported that pectin reduced β-catenin expression (37).

The escalating increase of APC expression in colorectal crypts beginning at approximately the junction between the proliferation and differentiation zone (Fig. 2A) supports the results from previous studies that documented an important role for APC in inhibiting proliferation and promoting differentiation (4). E-cadherin adhesion depends on the availability of extracellular calcium. A calcium gradient from low to high is hypothesized starting at the crypt base and extending to its luminal surface (38). The fairly uniform distribution of E-cadherin in crypts suggests that cell adhesion within colorectal crypts may depend more on the increasing extracellular calcium concentration as cells migrate toward the crypt apex, as opposed to increasing E-cadherin expression. The similar distribution patterns of E-cadherin and β-catenin may imply that most β-catenin in normal colorectal crypts is bound to the cytoplasmic tail of E-cadherin.

The primary limitation of this pilot study was the small sample size. Because of limited resources, we could evaluate rectal biopsies on only a subset of study participants, and ascending and sigmoid colon biopsies on an even smaller subset. However, despite our limited sample size, we found a statistically significant difference in the APC/β-catenin score between cases and controls. We also found statistically significant associations of the expression of APC and β-catenin and the APC/β-catenin score with plausible risk factors for colorectal neoplasms, although given the small sample size and multiple comparisons these results should be interpreted cautiously. Despite much of our immunohistochemical procedures being automated, batch variability inevitably is a source of measurement error; however, our method of standardizing for batch variability is an efficient method of addressing it (19). We did not specifically evaluate nuclear or cytoplasmic β-catenin because in the normal colorectal mucosa little or no nuclear and cytoplasmic β-catenin expression would be expected; however, it is plausible that total β-catenin expression is positively correlated with nuclear β-catenin expression.

The strengths of this study include the following: (i) that it is, to our knowledge, the first evaluation of components of the APC/β-catenin signaling pathway in the normal colorectal mucosa as potentially treatable, preneoplastic biomarkers of risk for incident, sporadic colorectal adenoma; (ii) the automated immunostaining and novel image analysis software to quantify the crypt distribution of the expression of APC, β-catenin, and E-cadherin; (iii) the assignment of case–control status based on colonoscopy, minimizing the chances of misclassification; and (iv) the assessment of dietary and lifestyle behaviors before case–control assignment to reduce recall bias.

In summary, the results of this pilot study suggest that lower APC expression, especially the proportion of APC in the upper 40% of crypts, combined with higher β-catenin expression in the normal colorectal mucosa may be associated with increased risk for incident, sporadic colorectal adenoma. Our results do not support E-cadherin as a preneoplastic biomarker of risk for colorectal adenoma. Also, we found that the APC/β-catenin score and APC and β-catenin expression may be associated with modifiable plausible dietary and lifestyle risk factors for colorectal neoplasms, suggesting that the biomarkers may potentially be modifiable. These results, taken together with our previous reports (9–11, 16, 22–25), provide support for conducting further, larger investigations to evaluate APC and β-catenin as potentially treatable, preneoplastic biomarkers of risk for colorectal neoplasms.

The Fullerton Foundation, the Georgia Cancer Coalition, and the Franklin Foundation had no influence on the design of the study; the collection, analysis, and interpretation of the data; the decision to submit the manuscript for publication; or the writing of the manuscript. No potential conflicts of interest were disclosed.

Conception and design: R.M. Bostick

Development of methodology: W.D. Flanders, R.M. Bostick

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T.U. Ahearn, M.E. Seabrook, R.M. Bostick, A. Shaukat

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): T.U. Ahearn, W.D. Flanders, R.M. Bostick, A. Shaukat

Writing, review, and/or revision of the manuscript: T.U. Ahearn, A. Shaukat, W.D. Flanders, R.M. Bostick

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): T.U. Ahearn

Study supervision: R.M. Bostick

The study was supported by The Fullerton Foundation, Georgia Cancer Coalition Distinguished Scholar award (to R.M. Bostick), and the Franklin Foundation.

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.