|Home | About | Journals | Submit | Contact Us | Français|
Whereas truncating germline mutations of the adenomatous polyposis coli (APC) gene give rise to familial adenomatous polyposis, missense polymorphisms of APC may confer a weaker risk for colorectal cancer.
We sequenced the entire open reading frame of the APC gene and tested for two common MYH mutations in a population-based series of patients with colorectal cancer and 5 to 99 adenomas. Missense adenomatous polyposis coli alterations identified in this colorectal cancer multiple-polyp population were analyzed in a population-based series of patients with colorectal cancer and healthy control subjects.
Germline APC or mutY human homologue (MYH) alterations were identified in 16 of 39 colorectal cancer-multiple polyp patients. Four missense APC gene alterations (S130G, E1317Q, Dl822V, G2502S) were observed in 13 individuals and 3 additional patients carried presumed pathogenic (APC Y94X, biallelic MYH Y165C and heterozygous MYH G382D) mutations. When independently assessed in 971 patients with colorectal cancer and 954 healthy control subjects, none of the identified missense APC alterations conferred a significantly increased risk for colorectal cancer, odds ratio (95 percent confidence intervals): S130G=3.1 (0.29–32.25), E1317Q= 1.08 (0.59–2.74), G2502S= 1 (0.65–1.63), D1822V (heterozygous)=0.79 (0.64–0.98), D1822V (homozygous) =0.82 (0.63–1.27).
Germline missense APC alterations observed in 33 percent of patients with multiple colorectal neoplasms seemed to play a limited role in colorectal cancer risk when independently assessed by a population-based, case-control analysis.
Approximately 20 percent of patients with colorectal cancer (CRC) or adenomatous polyps have a significant family history of CRC and as many as three-quarters of these familial cases demonstrate a dominant inheritance pattern.1,2 Familial adenomatous polyposis (FAP) is caused by truncating germline mutations in the adenomatous polyposis coli (APC) tumor suppressor gene,3 whereas hereditary nonpolyposis colorectal cancer (HNPCC) is associated most commonly with germline mutations in the MLH1, MSH2, MSH6, and PMS2 mismatch repair genes.4 These two syndromes account for <1 percent and approximately 2 percent of CRC cases, respectively.4 More recently, biallelic germline missense mutations of the MYH base excision repair gene have been associated with the development of polyposis and CRC,5 whereas monoallelic germline MYH mutations seem to be associated with a more modest increase in risk for CRC or polyps.6 Thus, to date, the majority of cases in which a familial clustering of CRC has been observed cannot be explained by any known genetic predisposition.4
More than 95 percent of patients with FAP have truncating germline APC mutations that result in the premature termination of protein translation.7 Interestingly, germline mutations in APC display a genotype-phenotype correlation because the location of a truncating mutation within the APC gene has an effect on the age of onset of disease, the severity of polyposis, and the pattern of extracolonic manifestations.4,8 In the attenuated adenomatous polyposis coli (AAPC or AFAP) syndrome, truncating mutations at the extreme 5′ and 3′ ends of the gene are associated with a “milder” form of FAP, with the development of fewer polyps (<100) and a later age of onset of CRC.9
In contrast to the rare, highly penetrant phenotypes conferred by truncating APC mutations in FAP and AAPC, the presence of more common, less penetrant predisposition alleles provides an attractive and potentially more common mechanism for CRC risk in individuals who have a significant family history of CRC. In Support of this hypothesis, recent meta-analyses have demonstrated moderately increased CRC risk in carriers of the TGFβR-1(6A), HRAS1*VNTR, and APC I1307K alleles.10
Although most previously characterized pathogenic APC mutations result in a truncated protein product,11 it is plausible that other alterations in the gene may influence development of CRC. The relationship between APC gene dysfunction and the development of CRC is complex. The genotype-phenotype relationship of APC mutations in FAP, the possibility of a haploinsufficiency model of tumorigenesis,7,12 and the relationship between the APC I1307K polymorphism and polyps and CRC all imply that simple loss of APC tumor-suppressor function is inadequate to fully describe the role of APC in CRC. Thus, missense APC polymorphisms may influence CRC risk by altering the functional, structural, and/or localization characteristics of the protein, thereby altering cellular physiology.
Although the possibility that nontruncating alterations in APC may contribute to CRC risk has been explored, previous studies have analyzed only a portion of the APC gene,13-15 or the entire APC gene in relatively small populations of cancer or polyp patients with few if any unaffected control subjects.16-19 In one moderate-sized, hospital-based series, no statistically significant APC alleles were identified in 91 high-risk Swedish CRC cases or in 247 sporadic CRC cases and 476 control subjects.20 In contrast to these searches of the entire APC gene, a few case-control series have evaluated CRC risk associated with specific alleles, such as APC D1822V21 or E1317Q,13,22-28 and in the case of E1317Q results of these efforts remain inconclusive.
To identify germline APC missense alterations that may alter the risk of CRC, we have sequenced the entire open reading frame of the APC tumor suppressor gene in a population-based series of 39 individuals who did not fulfill clinical criteria for FAP, but in whom a CRC resection included evidence of ≥5 synchronous adenomatous polyps. To account for the potentially confounding contribution of MYH alterations in this population, we also assessed for the germline status of the two common MYH mutations.6 Finally, to specifically determine relative risk, all nonsynonymous APC allele alterations identified in our original screen of patients with CRC and attenuated polyposis were then assessed in a large population-based series of 971 patients with CRC and 954 healthy control subjects.
The OFCCR is one of six sites of the National Institute of Health/National Cancer Institute supported Cooperative Family Registries for Colorectal Cancer Studies. Detailed characteristics of the OFCCR have been published previously.29 In brief, incident CRC patients aged 20 to 74 years, diagnosed between July 1, 1997 and June 30, 1998, and aged 20 to 70 years, diagnosed between July 1, 1998 and June 30th 2000, were identified by the population-based Ontario Cancer Registry and pathology reports reviewed by clerical staff. Other nonadenocarcinoma histologies (such as squamous-cell cancer, carcinoid, and sarcoma), recurrent and in situ malignancies, and cases of classical FAP were excluded. Permission was sought from primary physicians to contact the patient. Each patient was asked to complete a family history questionnaire from which a pedigree was constructed and the familial risk of each case was classified as high, intermediate, or low risk by using predetermined criteria.29 Family history data were verified by genetic counselors. All high-risk and intermediate-risk cases and a 25 percent random sample of sporadic cases were selected to participate in the OFCCR, including completing a personal risk factor questionnaire and a diet questionnaire. Relatives of the proband also were invited to participate in the OFCCR. Blood samples and paraffin blocks of tumor tissues were collected.
Incident cases from the OFCCR who were identified as having multiple synchronous polyps on pathology reports from CRC resections were selected for detailed pathology review. Polyps with villous, tubulovillous, tubular, and serrated histologic architecture were included in the polyp count; hyperplastic and inflammatory polyps were excluded. Cases with 5 to 99 adenomatous polyps in their colectomy specimen pathology report were selected for analysis. These subjects were excluded from the subsequent population-based CRC case-control analysis.
Controls were obtained from two sources: 1) using a random selection of residential telephone numbers in Ontario, subjects without any history of CRC (n = 533 controls) were matched by five-year age and sex to the distribution of OFCCR cases, and 2) using a similar strategy, subjects in Ontario without any history of breast cancer were matched to a similar population-based breast cancer registry, the Ontario Familial Breast Cancer Registry (n =421 controls).30,31 Family history questionnaires, diet, and personal risk factor questionnaires were administered to controls and blood samples were obtained.
As per local institutional review board approval, this research was conducted with all patients’ informed, written consent.
Lymphocytes were isolated from venous blood samples by Ficoll separation and genomic DNA extracted by Phenolchloroform or Qiagen® prep kit (Qiagen, Valencia, CA). The entire open reading frame of APC was divided into 28 PCR fragments for the purposes of sequencing. Exons 1–14, including all intron-exon boundaries were amplified individually in single PCR fragments (primer sequences and PCR conditions available on request). Exon 15 was divided into 14 overlapping segments of approximately 550 bp each. Aliquots of PCR products for each patient sample were electrophoresed on 2 percent agarose gels. Samples that ran as a single band on agarose were purified using Qiagen® PCR purification kit (Qiagen) and eluted in 30 μl of elution buffer. For aliquots with nonspecific bands, the entire sample was rerun on 2 percent agarose. The desired band was then excised, extracted using Qiagen® gel extraction kit (Qiagen) and eluted in 30 μl of elution buffer. Manual sequencing of each PCR product was performed using Thermosequenase® (Amersham, Piscataway, NJ) and 33P-labeled di-deoxy dNTP terminators.32 Each product was sequenced from the forward and reverse direction and loaded onto a denaturing urea-acrylamide gel.33 Gels were run at 60W for 1.45 hours and 4.5 hours (if necessary), dried onto filter paper, and exposed to film. All nonsynonymous alterations were confirmed by sequencing the corresponding region in both the forward and reverse directions. Additionally, each patient was screened for two common MYH gene variants, Y165C and G382D, by dHPLC and sequence analysis as previously described.6 In carriers of either of these mutations, the entire MYH gene was analyzed for other, possibly biallelic, mutations by dHPLC and sequence analysis as previously described.6
Once missense polymorphisms were identified and selected for case-control analyses, genotyping was performed using a 5′nuclease (Taqman®) assay. All probes and primers were designed using the Primer Express® software package (PE Biosytems, Foster City, CA) and synthesized by Operon (Qiagen). Allele specific probes were designed to bind to the wild-type and variant alleles at each locus and labeled with a different 5′ fluorescent label. Wild-type probes were labeled with HEX and variant probes were labeled with 6-carboxy-flourescein (FAM). Both probes were labeled on the 3′ end with Blackhole® quenchers (Operon®). DNA from each sample was diluted in duplicate into 96-well master plates. Duplicates of these master plates were made for each assay by aliquoting 15 ng of DNA from each well into a skirted-frosted 96-well plate suitable for the Taqman assay along with duplicates of the appropriate controls. Each reaction included 15 ng genomic DNA, 30 pmol of each primer, 12.5 pmol of each probe, and 5 μl of Quantitect PCR Probe master mix (Qiagen) in a final volume of 10 μl. Fluorescence was measured by using the ABI PRISM 7900HT Sequence Detection system. The genotype of each sample was then determined by using the graphic view of relative fluorescent intensities from the Sequence Detection System Version 2.0 (Applied Biosystems®).
All means for continuous, normally distributed values are reported as means±standard deviations. Median values for nonnormally distributed variables are reported as medians and ranges. All categorical values (such as genotype, gender, family history, and epidemiologic variables) were compared by chi-squared or Fisher’s exact test. Ordinal data (polyp and tumor numbers) were compared by Mann-Whitney U test and Kruskall-Wallis test. Continuous variables were compared by ANOVA and Student’s t-test.
During the first two years of recruitment by the OFCCR (July 1, 1997 to June 30, 1999), we identified 112 patients with CRC whose pathology reports were classified as having multiple synchronous polyps. The original pathology reports were re-reviewed, and only adenomatous polyps were included in the subsequent polyp count, whereas hyperplastic or inflammatory polyps were excluded. Thirty-nine cases were identified with 5 to 99 synchronous adenomatous polyps in their colectomy specimen pathology report and who had provided a blood sample to the OFCCR by January 1, 2002. The characteristics of these patients with CRC and multiple polyps are provided in Table 1. The mean age (±SD) was 61.5 ± 9.4 years. Fourteen cases had a history of two or more synchronous or metachronous CRCs based on review of the personal history questionnaire. All had five or more histologically confirmed adenomatous polyps. Fifteen individuals had more than 5 polyps, and two had more than 50 polyps.
Manual sequencing of the entire 8,535 bp open reading frame of APC was performed in these 39 CRC-multiple adenoma patients. A total of five nonsynonymous germline APC single nucleotide variants were detected, including one predicted to lead to protein truncation and four predicted to encode amino acid alterations. A single patient was found to have a nonsense mutation, Y94X, which results in a tyrosine (TAT) to stop (TAA) substitution. This 59-year-old patient had two synchronous CRCs, >50 adenomatous polyps, and a strong family history of CRC. Given the phenotypic characteristics of this case as well as the nature of the mutation, this subject has been classified as an unrecognized case of AAPC. The Y94X mutation has not been previously reported in the FAP or APC databases.34
A summary of all the missense variants identified in patients with CRC and 5 to 99 adenomas is shown in Table 2. A previously unreported missense change, S130G, was observed at the 5′ end of the gene in a single patient, the result of an A to G (AGT>GGT) substitution. This variant was found in a patient who had CRC and seven synchronous adenomatous polyps at age 64 years. There was no history of CRC in first-degree or second-degree relatives of the S130G carrier. A glycine to serine substitution at codon 2502 was identified in two multiple polyp patients. One of the G2502S carriers developed 2 synchronous cancers with 11 adenomatous polyps at age 53 years and did not have a family history of CRC. The other G2502S carrier developed CRC at age 63 years and had five adenomatous polyps. We detected this polymorphism in two of this patient’s relatives, one of whom had a history of polyps. The G2502S polymorphism has been observed in previous studies.20 The E1317Q variant was observed in two multiple polyp patients. This variant has been described previously.13,15,22-28 Finally, at codon 1822, we observed seven aspartate/valine heterozygotes and one valine/valine homozygote. This polymorphism has been described previously.21
All 39 multiple adenoma CRC patients were screened for MYH Y165C and G382D mutations. One subject was homozygous for the MYH Y165C mutation and had two synchronous colorectal cancers and 40 adenomatous polyps diagnosed at age 50 years. We also identified a heterozygous MYH G382D carrier who had two metachronous colorectal cancers at ages 53 and 59 years and more than 10 synchronous adenomatous polyps in each resection specimen. No other MYH mutations were detected in this subject by dHPLC analysis of the entire MYH coding region, including splice sites.6 None of the previously described APC variants were detected in these two patients.
A total of 971 population-based CRC cases and 954 control subjects were genotyped for the four nontruncating APC alterations identified in our CRC-multiple polyp cohort. Characteristics of cases and controls are shown in Table 3. There were more women than men among the control subjects because approximately half of our controls were recruited through the Ontario Familial Breast Cancer Registry and were matched to breast cancer cases.31 In addition, the CRC cases had stronger family histories of cancer compared with the control subjects at least in part because the OFCCR oversampled subjects with positive family histories of colorectal and other cancers.29
Genotyping results are shown in Table 4. The S130G variant seems to be rare and was found in three cases (0.3 percent) and one control (0.1 percent). Although potentially conferring an odds ratio of approximately 3.0, the current study was not powered to detect statistically significant differences in allele frequencies of such rare variants. Nonetheless, when the frequency of the S130G allele was compared among all three groups (2.6 percent CRC-multiple polyps vs. 0.3 percent CRC cases vs, 0.1 percent control subjects), a significant association (P=0.01, test for trend) with increasingly severe phenotype was observed. No other phenotypic associations were apparent when comparing S130G carriers and noncarriers (data not shown).
The E1317Q variant was observed at similar frequencies in both cases (1.8 percent) and controls (1.9 percent). Similarly, 4.6 percent of patients and 4.9 percent of control subjects were heterozygous for the G2502S variant, whereas one CRC patient and two control subjects were homozygous for this variant.
Heterozygous (36.4 vs. 32 percent) and homozygous (6.2 vs. 5.6 percent) carriers of the D1822V polymorphism were more common among control subjects than cases, respectively, Thus, D1822V heterozygous carrier status was associated with an odds ratio of 0.79 (95 percent confidence intervals (CI), 0.64–0.98; P=0.04) and homozygous carrier status with an odds ratio of 0.82 (95 percent CI, 0.63–1.27; P=0.35). If a dominant model is assumed, the combined heterozygous and homozygous variant genotypes are associated with an odds ratio of 0.82 (95 percent CI, 0.64–0.97; P=0.024). The Dl822V genotypes observed were in Hardy-Weinberg equilibrium in the control population, because their observed frequencies were not statistically different from the expected frequencies by chi-squared test (P=0.87). D1822V carrier status was not associated with a significant case-case difference in age of diagnosis of CRC with a mean age of diagnosis of 59.7, 60.5, and 60.2 years for wild-type, heterozygous, and homozygous variant carriers, respectively (P=0.43). Furthermore, multiple adenomatous polyps were observed in similar frequencies among wild-type and heterozygous D1822V carriers (3.7 and 3.9 percent, respectively). However, we did not observe any homozygous variant cases with multiple polyps. Furthermore, we did not observe any consistent relationship between the protective effect of the D1822V genotype and the age or dietary fat intake proftles of patients and control subjects (data not shown).
Approximately one in five patients diagnosed with a colorectal neoplasm has a family history of CRC, implying a significant hereditary contribution to CRC risk.1 Although significant progress has been made in characterizing highly penetrant forms of hereditary CRC, such as FAP and HNPCC and MYH-associated polyposis,4 we have postulated that some of the remaining hereditary cases may be the result of less penetrant genetic variants. The APC gene has been described previously as the “gatekeeper” of the colorerectal epithelium, and truncating mutations of this gene are observed in the germline of individuals with FAP, and somatic mutations are found in most adenomatous polyps and CRCs.35 Missense variants of the APC gene represent more subtle and more common genetic changes, which may have varying impact on protein stability, structure, and function, leading to an increased, or even decreased, risk of neoplastic transformation and progression. The present study is the largest effort to date to both identify potential disease-associated germline APC substitution mutations and to ascertain risk association of these alterations in an independent case-control series. Furthermore, it is the first such study performed that used population-based patients and control subjects. Finally, it is one of the few studies to date that estimates the potential risk of APC alleles in a case series separate from the cases in which they were initially detected. This separate retesting avoids bias introduced where the same cases are used for both hypothesis generation (to detect an inventory of relatively rare, potential disease-related alleles) and hypothesis testing (to estimate the case-control risk of these alleles ).36 Despite the exclusion of patients with phenotypic FAP by the OFCCR, 1 of 39 of the population-based cases with CRC and multiple synchronous adenomatous polyps did represent a case of AAPC (APC Y94X) and another case was caused by biallelic MYH mutations. A further CRC-multiple polyp patient was a heterozygous MYH G382D carrier, supporting our recent hypothesis of a possible codominant transmission pattern for MYH gene mutations.6
Our data suggest that the APC Sl30G variant may represent a CRC-risk allele. However, because this allele is very rare, it is likely that functional characterization, rather than case-control data, will be needed to further investigate this association.
APC El317Q has been reported previously as a CRC-risk associated allele13,23 and was observed in two of our patients with CRC and multiple-polyps. However, similar to other smaller association studies,15,22,24-28 our population-based data revealed that this allele was carried by a similar numbers of patients (1.8 percent) and control subjects (1.9 percent), providing strong evidence that this variant is not associated with an independent increased risk for CRC. Our large population-based, case-control series has sufficient power to exclude a significant odds ratio of 2.2 for an allele with the prevalence of APC E1317Q (alpha=0.05, power=0.8), and as such a smaller effect of this allele cannot be fully excluded. This underscores the importance of large population-based series to characterize relatively rare alleles with modest putative disease association and highlights the potentially erroneous conclusions of previous small studies.13,23 Similarly, our case-control data were powered to exclude significant CRC-risk with an odds ratio of more than 1.7 for the G2502S variant.
D1822V is the most common missense APC variant described to date.37 We observed that the heterozygous and homozygous variant genotypes were more common in control subjects than in patients (36.4 and 6.2 percent vs. 32 and 5.6 percent) respectively. This data yielded an OR of 0.79 (95 percent CI, 0.64–0.98) for heterozygotes and 0.82 (95 percent CI, 0.63–1.27) for homozygous variants, indicating that this variant likely confers a protective effect against CRC that is inherited in an autosomal dominant fashion. Our results are consistent with those previously published by Slattery and colleagues.21 However, in contrast to this previous study, we did not observe any consistent relationship between the protective effect of D1822V specific to patients who reported a low-fat diet or at an advanced age.
The findings of our large, population-based study in combination with those of previously published smaller case series14-16,18-20,22,24-27 suggest that it is highly unlikely that germline APC gene missense mutations confer a common, increased risk for CRC. The exception to this seems to be the APC 11307K variant,10 identified almost exclusively in individuals of Ashkenazi Jewish origin and mechanistically leading to an increased rate of truncating somatic mutations of the APC gene.
Interestingly, previous studies similar to ours have not revealed common, risk-associated missense alleles of the MSH2 or MLH1 genes in CRC28 or the BRCA1 gene in patients with breast or ovarian cancer.38,39 The fact that nonsense germline mutations of the APC, MSH2, MLH1, and BRCA1 genes cause highly penetrant inherited cancer risk, yet presumably functionally deleterious germline missense alleles are rarely associated with even a weak cancer risk is difficult to reconcile. If a mutation of the APC gene were skewed to an “all or none” functional consequence, then we would not expect to identify moderate penetrance missense “hypomorphs.” However, in this scenario, one would expect missense alleles to frequently account for cases of FAP and they do not.16,18 Nonetheless, the results of our study imply that the majority of genetic risk accounting for familial and even seemingly sporadic CRCs is likely to arise from as yet uncharacterized genes or, more likely, more complex gene-gene (digenic) or gene-environment interactions and not from monogenic inherited missense mutations of the “gatekeeper” APC gene.
Funding Support: S.P.C is a Research Fellow of the National Cancer Institute of Canada supported with funds from the Canadian Cancer Society. R.G. is a Charles H. Hollenberg Senior Fellow/Clinician Scientist supported with funds provided by Eli-Lilly Canada, Cancer Care Ontario, and the Canadian Institutes of Health Research. This work was supported by the National Cancer Institute, National Institutes of Health under RFA # CA-95–011, the Ontario Registry for Studies of Familial Colorectal Cancer (U01 CA074783), and through cooperative agreements with members of the Colon Cancer Family Registry (CFRs) and P.I.s, the Canadian Breast Cancer Research Initiative (grant # 9181 to J.A.K) and the National Cancer Institute of Canada (grant # 13304 to S.G.).
The content of this manuscript does not necessarily reflect the views or policies of the National Cancer Institute or any of the collaborating centers in the CFRs, nor does mention of trade names, commercial products, or organizations imply endorsement by the United States Government or the CFR.