APC E1317Q Is Not Associated with Colorectal Cancer in a Population-Based Case-Control Study in Northern Israel

Germ line truncating mutations in the APC (adenomatous polyposis coli) gene (5q21-22) cause familial adenomatous polyposis, which is characterized by thousands of polyps during adolescence and early-onset colorectal cancer. Several putative low-penetrance susceptibility alleles of APC, such as APC E1317Q, have been investigated in affected individuals with a family history of disease and in populations at risk. The APC E1317Q missense variant was first described by White et al. (1) in two of four siblings in an Ashkenazi Jewish colorectal cancer family. APC E1317Q has been reported to confer an increased risk of colorectal adenomas and cancer in some (2-5) but not all (6-10) case-control studies. To clarify the divergent findings from previous reports, the present study investigates the relationship between E1317Q and colorectal cancer in a large population-based case-control study of colorectal cancer in northern Israel.

Participants were recruited as a part of the Molecular Epidemiology of Colorectal Cancer study. Residents of the Northern and Haifa districts of Israel who were diagnosed with histopathologically confirmed, incident colorectal cancer between March 31, 1998 and April 1, 2004 were eligible as cases. Population-based controls were identified from the Clalit Health Services database and were individually matched to cases by year of birth, gender, primary clinic location, and ethnicity (Jewish versus non-Jewish). A detailed description of the ascertainment, eligibility, and participation rates has been previously published (11).

Genotyping for the known susceptibility variant APC I1307K was done as described elsewhere (12) using an allele-specific oligohybridization assay. Genotyping for APC E1317Q used an allele-specific oligohybridization and oligonucleotides corresponding to the wild-type or mutant sequence at codon 1317, 5′-AGGTCAGCTGAAGATCCT-3′ and 5′-AGGTCAGCTCAAGATCCT-3′, respectively. The assay was validated using 10 subjects (3 carriers and 7 noncarriers) by sequencing in forward and reverse, with 100% correspondence. Eighty subjects were repeated during genotyping, also with 100% correspondence.

Conditional and unconditional logistic regression were used for analyzing the presence or absence of APC E1317Q (SAS Institute, Cary, NC, 1999). Potential confounding factors (Ashkenazi versus non-Ashkenazi ethnicity, family history of colorectal cancer in a first-degree relative, weekly use of nonsteroidal anti-inflammatory drugs/aspirin for at least 1 year, and APC I1307K) were included in a multivariable model. Data regarding colonoscopy, sigmoidoscopy, or bowel surgery on the control subjects were obtained through interview. Fixed effect (Mantel-Haenszel) summary odds ratios (OR) and Woolf's test of heterogeneity were calculated using the rmeta library in R (http://www.r-project.org).

The data were comprised of 1,834 matched pairs from 1,970 cases and 1,994 controls with available DNA for genotyping (Table 1). The proportion of subjects of Ashkenazi Jewish ethnicity was higher in cases than controls, which is consistent with the higher incidence rates of colorectal cancer in Ashkenazi than Sephardi Jews in Israel. Colorectal cancer cases were slightly more likely to carry the variant E1317Q allele than controls [OR, 1.15; 95% confidence interval (95% CI), 0.65-2.02]. The point estimates of the ORs differed when stratified by ethnicity (OR_Ashkenazi, 1.75; 95% CI, 0.74-4.15; OR_{non-Ashkenazi}, 0.87; 95% CI, 0.39-1.91), but a test of interaction was not significant (P = 0.24). Considering only the matched pairs, E1317Q was not strongly associated with risk of colorectal cancer (OR, 1.09; 95% CI, 0.61-1.95). Adjusting for potential confounding factors did not appreciably change the original estimate (data not shown).

It is possible that cases with microsatellite unstable tumors or controls with undiagnosed polyps who carry the E1317Q polymorphism may attenuate the OR in this sample. E1317Q was equally frequent in microsatellite stable cases (N = 1,288, 1.17%) and controls (1.15%; OR, 1.01; 95% CI, 0.53-1.95). Three hundred ninety-eight controls reported to have undergone colonoscopy, sigmoidoscopy, or bowel surgery and were polyp-free. When only considering polyp-free controls and all cases in the Molecular Epidemiology of Colorectal Cancer study (N = 1,970), the OR for E1317Q was 0.87 (95% CI, 0.39-1.91). No controls who reported polyps carried the E1317Q mutation (0 of 87).

We completed a meta-analysis of seven previous studies, in addition to our data, to summarize the association among E1317Q, colorectal cancer, and adenomas (Table 2; Fig. 1). Two of the three studies reporting a positive association use the same group of controls reported by White et al. group (3, 4), and the other study (5) uses some of the cases in the Lamlum and Frayling studies. The Lamlum et al. study also uses controls and five cases from the Frayling et al. study. A case was defined as colorectal cancer or adenoma, and controls were included without regard to polyp status. The results were essentially the same when analyzed with and without inclusion of adenoma cases. The summary OR for these eight studies was 1.35 (95% CI, 0.92-1.97; Woolf's test of heterogeneity, P = 0.14). However, when excluding the one positive association of Lamlum et al. and calculating a summary OR for the remaining replication studies, the summary OR is 1.14 (95% CI, 0.77-1.77; Woolf's test of heterogeneity, P = 0.86). Exclusion of the present study, by far the largest, did not change this estimate (OR, 1.14; 95% CI, 0.65-2.01; Woolf's test of heterogeneity, P = 0.77). Woolf's test of heterogeneity did not indicate a significant difference between any of the studies.

Due to the low frequency of the APC E1317Q allele in these populations, there is modest power (78.0%) to detect a 2-fold increase in risk, and it would be difficult to discern smaller risks. Despite this, it is still unlikely that this variant predisposes individuals to a clinically meaningful increase in risk of colorectal cancer, although weak associations could not be excluded, especially within the Ashkenazi population. However, the point estimates do not significantly change when considered in the context of known risk factors for colorectal cancer or when restricting the sample to microsatellite stable cases. When only controls who are reported to be polyp-free are used as a comparison group, E1317Q is weakly protective (OR, 0.87) in the Molecular Epidemiology of Colorectal Cancer study, arguing against attenuation of the OR due to misclassification of control subjects. It is notable that the summary OR obtained from the meta-analyses is not significant (1.35; 95% CI, 0.92-1.97), and exclusion of the initial Lamlum et al. study and the present analysis yields essentially the same OR as we find with our data (OR_summary, 1.14; 95% CI, 0.65-2.01; OR_{this study}, 1.15; 95% CI, 0.65-2.02). APC E1317Q seems to confer little or no risk of colorectal cancer in an Israeli population, and this study highlights the importance of large studies to replicate associations of candidate alleles.