Presence of a TA Haplotype in the APC Gene Containing the Common 1822 Polymorphism and Colorectal Adenoma

Acquired or inherited mutations in the adenomatous polyposis coli (APC) tumor suppressor gene are causally linked to colorectal cancer. Given the significance of APC in colorectal cancer, we investigated the association between common single-nucleotide polymorphisms (SNP) in the APC gene and the odds of developing metachronous colorectal adenomas as a surrogate measure of colorectal cancer risk. Coding SNPs at codons 486, 1678, 1822, 1960, and 2502 were analyzed in a total of 1,399 subjects who participated in two randomized clinical trials for the prevention of colorectal adenomas. No association was found for any single SNP and the odds of metachronous adenoma. In contrast, a TA haplotype (codons 486 and 1822) was associated with a statistically significant 27% and 26% reduction in the odds of any and nonadvanced metachronous adenoma after adjustment for baseline adenoma characteristics [odds ratio (OR), 0.73; 95% confidence interval (95% CI), 0.59–0.91 and OR, 0.74; 95% CI, 0.57–0.94], respectively. No significant reduction in odds was observed for advanced metachronous lesions. Diplotype analysis revealed a strong gene dose effect with carriers of two alleles containing TT-AA (codons 486 and 1822, respectively) having an 89% lower odds for advanced metachronous adenomas (OR, 0.11; 95% CI, 0.01–0.80) when compared with the common CC-AA diplotype (codons 486 and 1822, respectively). Our findings support an important role for germ-line allele sequence in the APC gene and individual risk of metachronous adenomatous polyps. [Cancer Res 2008;68(14):6006–13]

Germ-line and somatic mutations in adenomatous polyposis coli (APC), a tumor suppressor gene, are common in familial and sporadic forms of colorectal cancer, respectively (1). In addition, APC is an important component of the Wnt signaling pathway, which becomes dysregulated in many cancers, particularly common colorectal tumors. As such, acquired or inherited mutations in the APC gene are recognized as major determinants of colorectal cancer risk.

Under wild-type conditions, APC forms a complex to coordinate β-catenin degradation (2, 3). In the majority of colorectal tumors, mutated APC is unable to interact effectively with β-catenin (4), resulting in nuclear accumulation of β-catenin and activation of key proliferation genes as a common early event in colon carcinogenesis (2, 5, 6). However, unlike the well-characterized pathogenic germ-line mutations in APC and MutY homologue genes found in classic familial adenomatous polyposis (FAP), the consequences of common allele variants in APC in relation to colorectal adenoma or cancer risk are less clear.

A rare single-nucleotide polymorphism (SNP) in APC, creating an isoleucine-to-lysine substitution at codon 1307 and putative hypermutable tract, has been associated with increased risk of colorectal cancer among individuals of Ashkenazi Jewish descent (7, 8). However, results from other studies are less informative with the majority observing no difference in risk of colorectal adenomas or colorectal cancer by the 1307 genotype (9–15). The frequency of the 1307K polymorphism is highest in Ashkenazi Jewish populations (∼6%; refs. 7, 14, 16) and rare in other groups (11, 17, 18). Therefore, risk associated with 1307K is less generalizable across populations, and inconsistent results for this polymorphism may reflect population-specific effects.

Investigations of other rare variants in the highly polymorphic APC gene have been similarly inconsistent but generally indicate a role for low-penetrance variants of APC and colorectal carcinogenesis (14). A missense glutamic acid (E) to glutamine (Q) variant at codon 1317 occurring in 1% to 4% of the population has been investigated as a risk allele (19). Carriage of Q at 1317 has been associated with multiple colorectal adenomas but not with cancer risk (19, 20). Hahnloser and colleagues found no association between 1317Q allele and adenoma or cancer risk using spousal controls. However, when compared with a colonoscopy-negative control group, 1317Q was more prevalent in adenoma and cancers especially among those that retained mismatch repair (19). Subsequent work has generally not supported a major role for the 1317Q in colorectal cancer risk (9, 11, 17, 18, 21) although cumulative and more recent evidence implicate 1317Q in adenoma risk (14).

In addition to these rare allele variants, a commonly occurring aspartic acid (D) to valine (V) polymorphic variant at codon 1822 located in the middle of the β-catenin down-regulatory domain was originally reported as a mutation for nonclassic FAP (22). Subsequently, the 1822V variant was reclassified as a common APC polymorphism (frequency ≈0.22) that lacked clinical significance (23), a conclusion that was challenged by Macdonald and Wallis who suggested that the 1822 minor variant might act as a low-penetrance allele (23). Slattery et al. (24) found no significant associations between colon cancer risk and carriage of the D1822V variant. Similarly, no main effects of the 1822 variant have been observed in subsequent studies for colorectal adenoma or colorectal cancer (24–26). Additional analyses of gene by environment interaction, however, hint at the risk-modifying effects of the 1822 variant in the presence of dietary fat (24) or hormone replacement therapy (25).

Whereas a handful of studies have measured individual polymorphisms in the APC gene and risk for colorectal adenoma and colorectal cancer, including those highlighted above, there has been no investigation of the role of common APC haplotypes and colorectal adenoma or cancer risk, which might explain the mostly inconsistent observations. Cui and colleagues investigated the role of APC haplotypes and risk of schizophrenia. Building on the hypothesis that high activity APC in the brain confers risk for schizophrenia and might explain the previously observed lower cancer risk among patients with schizophrenia (27–29), Cui and colleagues reported that the carriage of a CAT haplotype that altered codons 486, 1678, and 1960 was elevated among subjects with schizophrenia. Cui and colleagues concluded that the high prevalence of the CAT haplotype might reflect higher APC activity and explain the observed lower risk of cancer among these individuals (27–29). As yet, no functional studies have been conducted to specifically address the influence of these allelic variations on protein stability or activity. If present in linkage with each other or other unstudied polymorphisms, then the sequence content of APC alleles (i.e., haplotype) may explain, in part, study-to-study inconsistencies.

Overall, the associations between APC polymorphisms and colorectal neoplasia remain unresolved. No studies to date have reported on the contextual importance of cocarriage of these SNPs and their role as genetic risk factors for sporadic colorectal cancers or adenoma. The purpose of this study was to investigate the association between APC polymorphisms that influence codons 486, 1678, 1822, 1960, and 2502 and metachronous colorectal adenomas with emphasis on the association between commonly occurring APC exonic SNPs and risk of advanced, large tubulovillous/villous lesions.

Study population. Subjects were drawn from participants of the Wheat Bran Fiber (WBF) and Ursodeoxycholic Acid (UDCA) trials conducted at the University of Arizona (30–32), as described elsewhere. Briefly, the purpose of the WBF study was to determine the effect of a high-fiber versus a low-fiber cereal supplement on colorectal adenoma recurrence among participants in a randomized, double-blind clinical trial (31). Study participants were between the ages of 40 and 80 y, with at least one histologically confirmed colorectal adenoma removed no more than 3 mo before entry into the study (31). A total of 1,310 participants completed the trial by having had a follow-up colonoscopy (33); no effect of the high fiber supplement on colorectal adenoma recurrence was detected. The UDCA trial was a phase III, double blind, placebo controlled study designed to determine the effect of UDCA as a preventive agent for recurrent colorectal adenomas (32). Subjects were between 40 and 80 y old and had a colorectal adenoma removed no more than 6 mo before enrollment in the study (32). A total of 1,192 participants completed the trial (32), although no effect of UDCA on adenoma recurrence was observed. Both the WBF and UDCA trials were approved by the University of Arizona Human Subjects Committee and Institutional Review Board.

Variables that differed between the two studies that were used in the pooled analyses included calcium intake, family history of colorectal cancer, any metachronous adenoma, advanced metachronous adenoma, number of baseline adenomas, and location of baseline adenoma. Total dietary fat intake, smoking, and study differed between the final population and the total study population at the P = 0.05 level.

Genotyping. Polymorphisms at codons 486, 1678, and 1960 of APC were selected because of the potential protective effect against cancer, identified in schizophrenic subjects carrying a CAT haplotype reported by Cui and colleagues (29). Additionally, the common SNPs at codons 1822 (MAF 0.225) and 2502 (MAF 0.017) were investigated because of prior colorectal cancer risk associations (24). SNPstream (Beckman Coulter, Inc.), a fluorescence-based single base extension assay, was used to genotype SNPs in codons 486, 1678, 1960, and 2502 (Sequence Detail, Supplementary Table S1) according to the manufacturer's recommended protocols. PCR reaction mixture was prepared with a final concentration of 1× PCR Buffer II, 5 mmol/L MgCl₂, 90 μmol/L of each deoxynucleotide triphosphate mix, 50 nmol/L of each primer pool, 2 ng genomic DNA/5 μL, and 0.1 unit Taq Gold/μL. The extension reaction was prepared with SNPware Extension Dilution Buffer, 5 μmol/L each SNPware C/G Extension Primer Mix, SNPware C/G 20× Extension Mix, and SNPware DNA polymerase. Standard SNPstream thermocycle conditions were applied in both reactions. After extension, the reaction wells were washed with Wash Buffer 1 and dried. Hybridization solution was prepared with SNPware Hybridization Solution and SNPware Hybridization Additive added to the reaction wells, incubated, washed with Wash Buffer 2, and dried. The dried plates were then placed in the SNPstream Imager for imaging.

TaqMan Assay-by-DesignSM primer ID no. C_3162935_20 (Applied Biosystems) was used to genotype codon 1822, which did not convert successfully to SNPstream. PCR reaction mixture was prepared using 2× TaqMan Universal PCR Master Mix with No AmpErase UNG, 20× SNP Genotyping Assay Mix, and 3 ng of DNA according to the manufacturer's instructions. The reaction was then thermocycled under the following cycling conditions: 10 min at 95°C followed by 40 cycles of 15 s at 92°C and 1 min at 60°C. On completion of thermocycling, the plate was placed in an ABI Prism 7700 Sequence Detector (Applied Biosystems) for analysis. The T allele complementary strand probe contained the sequence 5′-CAGAATTTGAAAAATAATTCCAAGG[A/T]CTTCAATGATAAGCTCCCAAATAAT-3′ and was labeled with FAM reporter dye on the 5′ end. The complementary probe for the A allele was labeled with VIC reporter dye on the 5′ end.

Genotyping was successfully done in 666 of the participants in the WBF study and in 946 of the participants of the UDCA. Participants for whom blood samples were not available were not genotyped (n = 970; 39%). After exclusion of participants with no recurrence data (n = 260), discordant genotype calls (n = 8), failed genotype calls (n = 142), or haplotype probability scores of <0.9 (n = 89), as identified from PHASE 2.1 (34, 35) as described below, the final study population included 1,399 subjects.

The genotyping success rate was 96.1% for codon 486, 96.3% for codon 1678, 98.3% for codon 1822, 97.6% for codon 1960, and 98.1% for codon 2502. Genotype data were given a pass if they met the following criteria: distributed into three statistically significant clusters, failed to generate genotypes from all 181 blank wells, and showed Mendelian consistency for Coriell CEPH trios and showed >97% concordance among lab blinded replicates. Any samples that exhibited low fluorescence intensity or generated an ambiguous genotype call were classified as failed and not included in the data. Genotype frequencies for all the polymorphisms were found to be in Hardy-Weinberg equilibrium.

Haplotype determination. Haplotypes were derived using PHASE 2.1⁷

with default iteration settings (34, 35). Haploview v3.2 was used for Hardy-Weinberg and D′ determination.⁸ All SNPs were found to have a D′ >0.90. Individuals with a haplotype probability of <0.90, as determined by PHASE, were dropped from the analysis. Haplotypes that contained less than eight participants, including cases and controls, were dropped from the analysis.

Statistical analysis. Statistical analyses were conducted using STATA version 9.0. Covariates that are considered a priori as potential confounders were evaluated in a preliminary model of the association between SNPs, haplotypes, diplotypes, and adenoma recurrence. These included age, intake of dietary fat, energy, alcohol, dietary fiber, calcium, body mass index (BMI), aspirin use, history of smoking, family history of colon cancer, gender, race, history of previous polyps, and study (WBF versus UDCA). Additional baseline adenoma characteristics such as histology size, number, and location were included because these variables have been shown to positively predict the recurrence of adenomas in this study population (36). Stepwise regression was used to determine which covariates were significant in the model (P < 0.10) for inclusion in the final model. Odds ratios (OR) were calculated for crude (age and gender only) as well as the fully adjusted models presented. There were no major differences in the point estimates with the observed associations between the crude and adjusted models; therefore, only the fully adjusted models are presented.

Modeling of the association between SNPs, haplotypes, and diplotypes was done using logistic regression, with the end points of any metachronous adenoma, nonadvanced metachronous adenoma, and advanced metachronous adenoma. Advanced lesions included adenomas ≥1 cm and/or adenomas with tubulovillous or villous histology, and/or a diagnosis of colorectal cancer, the latter of which includes five cancers. Those with no metachronous lesions were used as the reference group for all analyses. For investigations of the effect of single SNPs and haplotypes, the most common allele and haplotype served as the reference groups, respectively. We first conducted each of these analyses in the WBF and UDCA populations separately. Similar trends in the association between haplotype and recurrence were observed in both populations. In addition, a test for heterogeneity of genotypes between WBF and UDCA was conducted using a likelihood ratio test with P = 0.10. No statistically significant differences were found between these populations. Therefore, the data were pooled to achieve more precise estimates.

Baseline characteristics of the genotyped participants with a history of adenoma used in the pooled analyses of metachronous adenoma are presented in Table 1. The observed frequency of the genotyped SNPs in APC was similar to those expected for a European-Caucasian population (Table 2) and, thus, did not indicate a bias in their distribution in subjects with a history of colorectal adenoma.

As shown in Table 3, each SNP was analyzed to determine if there was a main effect of each on the odds of any, advanced, or nonadvanced metachronous adenoma independent of information at the other SNPs. We found no significant association between any individual SNP and any metachronous lesion, nor was a relationship observed when lesions were stratified into nonadvanced or advanced lesion categories.

The associations between APC haplotypes and metachronous lesions are presented in Table 4. In each analysis, the most common haplotype served as the reference group. When all SNPs (group 1) were included in the generation of the APC haplotypes (n = 5 SNPs), a TGACC haplotype (codons 486, 1678, 1822, 1960, and 2502), occurring with a frequency of 0.13, was found to be associated with a statistically significant 27% reduction in the odds of any metachronous adenoma [OR, 0.73; 95% confidence interval (95% CI), 0.57–0.94], whereas a TAATC haplotype showed a similar but nonsignificant reduction in the odds of any metachronous lesion (OR, 0.69; 95% CI, 0.46–1.04) when also compared with the most common CAATC haplotype.

Based on the apparent association of the T nucleotide at codon 486 and A nucleotide at codon 1822 and the lower odds of metachronous adenoma in the initial analysis of the group 1 haplotype, we collapsed the haplotype analysis to include only the polymorphisms at codons 486 and 1822. When collapsed, the TA-containing haplotype (frequency = 0.19) was associated with a statistically significant decreased odds of any metachronous adenoma (OR, 0.73; 95% CI, 0.59–0.91), which was of equal magnitude to that observed for the TGACC and TAATC variants of group 1. We then explored whether the TA haplotype (codons 486 and 1822) had differential effects on nonadvanced and advanced lesions when present with all SNPs (group 1) or in the truncated group 2 haplotype (Table 5). The TGACC haplotype was associated with a statistically significant reduction in the odds for advanced adenomas (OR, 0.63; 95% CI, 0.42–0.94) whereas this association was slightly attenuated for nonadvanced lesions (OR, 0.78; 95% CI, 0.59–1.03). The collapsed TA group 2 haplotype was associated with a significantly reduced odds for nonadvanced adenomas (OR, 0.74; 95% CI, 0.57–0.94), with a similar but nonstatistically significant effect observed for advanced lesions (OR, 0.76; 95% CI, 0.54-1.07).

To evaluate the effect of gene dose of the 486 and 1822 APC SNP variants on the outcomes of metachronous adenoma, we conducted an analysis of diplotypes restricted to the two highest interest SNPs (486 and 1822) used to generate the group 2 haplotype (Table 6). Using the most common CC-AA diplotype as the reference, carriage of the CT-AT diplotype was associated with an increased odds for nonadvanced adenomas (OR, 1.46; 95% CI, 1.05–2.01), whereas carriers of TT-AA showed a statistically significant inverse association for any metachronous and for advanced lesions, with the strongest effect observed with an 89% reduction in the odds for advanced metachronous adenomas (OR, 0.11; 95% CI, 0.01–0.80).

Next, we collapsed the data to determine if specific haplotype subgroups in addition to the TA group differentially influenced the risk of metachronous adenoma (Table 5). First, we measured the effect of the CAT (codons 486, 1678, and 1960) haplotype (group 3) because of the postulated lower cancer risk in this haplotype group by Cui and colleagues (29). No significant association was observed between this haplotype group and metachronous adenoma in our population. In addition, a GCC haplotype (codons 1678, 1960, and 2502) was also evaluated separately due to a marginal protective effect with cocarriage of these SNPs in group 1. The GCC haplotype of group 4 showed a marginally significant 23% reduction in the odds for advanced metachronous adenoma (OR, 0.77; 95% CI, 0.59–1.00) compared with the common CAT haplotype, which is consistent with the protective effect seen in TGTCC, TGACC, and CGACC for advanced lesions that are likely driven by the protective effects of the T at codon 486 and A at codon 1822 noted above and in strong linkage with the GCC group 4 haplotype. There was no significant effect for any metachronous adenoma or nonadvanced metachronous adenoma in group 4 haplotypes.

We observed no significant associations between any of the individual APC SNPs studied (codons 486, 1678, 1822, 1960, and 2502) and the odds of metachronous colorectal adenomas, including no main effect of the previously studied D1822V variant (24–26). In contrast, haplotype analysis that included all five coding SNPs in the APC gene revealed a protective effect from metachronous colorectal adenomas in carriers of the minor synonymous 486 (T nucleotide variant) when combined with the major A1822 variant (codes for aspartic acid). Carriers of any T at codon 486 and any A at codon 1822 haplotype had a significant 27% lower odds of any metachronous adenoma that was observed for advanced and nonadvanced adenomas. These data support a general protective effect of the T at codon 486 and A at codon 1822 from the development of new adenomas. Diplotype analyses support a strong gene dose effect of the two protective variants, with the odds of advanced metachronous adenoma dramatically lower among individuals homozygous for both polymorphisms (TT at codon 486 and AA at codon 1822). Analyses of the contribution of the other SNPs suggested no major contribution to the odds of metachronous adenoma that could not be explained by strong linkage to the SNPs at 486 and 1822.

Given the critical role of APC as a regulator of β-catenin and Wnt signaling in colorectal tumorigenesis, clarifying the role of common allelic variation in APC and colorectal cancer risk is of potential relevance for the identification of risk modifying alleles (2). APC is an obvious candidate gene for the study of common allelic variation and disease risk. Nonsynonymous changes in the genetic sequence that result in amino acid changes have been of particular interest and have received the most attention (14). Variation at 1822 and 2502 involves nonsynonymous polymorphisms,⁹

with the A-to-T polymorphism in 1822 resulting in an aspartic acid to valine amino acid change⁹ in the middle of the β-catenin down-regulatory domain of the protein. This is of potential functional significance because aspartate is an acidic amino acid (37) whereas valine is a nonpolar amino acid (38). The C-to-T polymorphism at codon 2502 results in a glycine-to-serine change.⁹ Glycine is a nonpolar amino acid (38) whereas serine is a polar amino acid (39). Group 2 of our haplotype analyses was found to have a statistically significant association with metachronous adenomatous lesions that seem to be dependent on the occurrence of the aspartic acid allele when it co-occurs with a synonymous tyrosine coding variant at 486. It is unclear at present why the protective effect of the 1822D variant is restricted to the alleles that carry the 486T. The effect of the 486T synonymous or “silent” SNP in the context of the 1822D may be the result of linkage with an as yet unidentified variant that does change the amino acid sequence and perhaps binding to β-catenin. Alternatively, the “so-called” silent SNPs might differentially influence the rate of protein synthesis based on differences in the available pool of the cognate tRNA or actual codon context (40). Shifts in sequence to rare codon use have been shown to influence the rate of translation, which in turn affects protein folding with the third base having the largest influence (41). In the case of the silent SNP at 486, the nucleotide C-to-T change occurs in the 3rd base position of the codon for tyrosine, yielding the preferred sequence for the UAU tyrosine tRNA and perhaps enhanced protein synthesis and folding with higher available APC protein (42). Additional studies that include in vitro characterization, in silico analysis, and resequencing efforts are necessary to clarify the relation between these specific variants in the APC gene and protein levels and activity, including effects on binding to β-catenin.

Consistent with Cui and colleagues (29), the CAT haplotype (group 3) was found to be the most common (0.57) in our study. Within the group 3 haplotype, the TAT carriers had a nonsignificant lower odds of any metachronous adenoma that was largely driven by a statistically significant 43% reduced odds for nonadvanced adenoma types. This observation is likely explained by the presence of the T at codon 486 discussed above. These results are counter to the hypothesis proposed by Cui that the CAT haplotype in APC would confer higher APC activity and thus explain lower cancer risk in patients with schizophrenia (29).

There are a number of limitations to this study. Our population consists only of individuals who have a history of at least one adenoma who are followed postpolypectomy in an intervention trial for new, metachronous adenomas. Thus, we are only able to assess the influence of the APC variants on the risk of developing a new adenoma among an adenoma-prone population. The results from this data are not generalizable to incidence of colorectal adenoma or colorectal cancer. Furthermore, we did not evaluate the entire allelic structure of APC but rather focused on variants of potential interest. A major strength of this study is the prospective nature and comprehensive collection of lifestyle data.

Our results highlight the potential importance of the joint effects of adjacent SNPs and suggest that additional efforts to comprehensively evaluate the role of common APC “allele” sequence variation as a potential low-penetrance risk locus for the development of colorectal neoplasia through a gene-environment interaction are needed. The results from our study are intriguing because they support previous conclusions that these common and rare allele variants of APC likely influence colorectal carcinogenesis (7, 8, 22). A number of studies support the importance of variation in APC on colorectal adenoma with weak to no association with colorectal cancer (9, 10, 12, 13, 15, 17, 18, 23). These data suggest that modest acting variants might act as predisposing variants for adenoma development but are insufficient alone as risk factors for cancer. This is consistent with the relatively common nature of the adenoma and the rare nature of colorectal cancer. Additional studies of APC gene variants and environmental factors may improve our understanding of the importance of these variants as modifiers of colorectal cancer risk as suggested by the studies of Slattery and colleagues (24) and Tranah and colleagues (25).

The risk of metachronous adenoma in polyp formers is strongly associated with a number of baseline adenoma and patient characteristics including size and multiplicity of lesions. These characteristics inform on the timing of follow-up colonoscopy in this at-risk population. In our adjusted analysis, the T at codon 486 and A at codon 1822 haplotype remained significant as a modifier of risk even after accounting for the known risk factors for the development of adenoma after polypectomy. Future analyses will include modeling the joint effects of baseline factors and APC haplotypes/diplotypes and adenoma recurrences in this population, which may improve the identification of individuals at highest risk for metachronous lesion and add to recommendations for colonoscopy interval.

Our results suggest that cocarriage of specific polymorphisms in the APC gene can confer protection against metachronous adenomatous lesions. Furthermore, there seems to be a protective role for synonymous polymorphic changes when present with particular nonsynonymous polymorphisms. Whereas previous studies have shown the value of individual SNP analysis in assessing the odds of disease (7, 9–13), our work and that of Cui and colleagues (29) suggest that evaluating allele structure is critical in the understanding of the role of APC polymorphic variation and colorectal adenoma and perhaps susceptibility to cancer. Further evaluation of these findings would serve to expand the understanding of the role of these polymorphisms and haplotypes in the etiology of colorectal cancer. The functional significance of these and other polymorphisms in APC and how these may affect the interaction of APC with proteins in the Wnt signaling pathway are needed.

No potential conflicts of interest were disclosed.

Grant support: Specialized Program of Research Excellence in Gastrointestinal Cancer grant CA95060, National Cancer Institute/NIH Colon Cancer Prevention Program Project grant PO1 CA41108, and Technology and Research Initiative Fund award from the University of Arizona and Arizona State University.

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.