Associations of Ki-ras Proto-oncogene Mutation and p53 Gene Overexpression in Sporadic Colorectal Adenomas with Demographic and Clinicopathologic Characteristics

Colorectal carcinoma (CRC) is the second leading cause of cancer death in the U.S. (1). The majority of these CRCs arise sporadically through a multistep process, with the adenoma as the recognized intermediate step in their development in an adenoma-carcinoma sequence (2). Major oncogenes and tumor suppressor genes described thus far in CRC tumorigenesis include alterations in APC and the Wingless/Wnt pathway, Ki-ras, p53, and genes involved in the tumor growth factor-β signaling pathway (TGFβRII and SMAD4; refs. 3-5). A second pathway involves nucleotide mismatch repair genes responsible, when mutated in the germ line, for hereditary nonpolyposis colorectal cancer syndrome, a cancer predisposition syndrome (6). Although APC alterations are common in the earliest lesions (7, 8), the Ki-ras proto-oncogene is also mutated early in this process, with the reported frequency of Ki-ras mutation varying from 10% to 15% in small adenomas (<1 cm) and 30% to 65% in larger adenomas (≥1 cm) and CRCs (5, 9-28).

The Ki-ras gene encodes a plasma membrane–bound GTP binding protein that is a key regulatory component of numerous signal transduction pathways and may mediate the growth of colorectal adenomas (29). Numerous studies have shown that >90% of activating mutations occur in codons 12 and 13 of exon 1 of the gene. G-to-A transition and G-to-T transversion mutations are the most frequently observed mutations in colorectal carcinogenesis (9, 10, 15, 18, 24, 25, 30-32). The p53 gene is a tumor suppressor with critical functions in cell cycle arrest, DNA repair, and apoptosis (33). p53 gene product overexpression occurs with most p53 mutations, and both mutation and overexpression are found late in the CRC tumorigenesis process. p53 alterations occur primarily in the progression of adenoma to carcinoma, with more than half of the carcinomas having abnormal p53 status (18). Overexpression of p53 has been observed in 5% to 32% of colorectal adenomas and in 45% to 67% of carcinomas (15, 17, 19, 34-40). In two small studies, p53 protein expression by immunohistochemistry was significantly associated with villous architecture in one report (15) and with high-grade dysplasia (HGD) in both (15, 19).

The aim of the current study was to determine the associations among Ki-ras and p53 alterations in sporadic colorectal adenomas and demographic and clinicopathologic characteristics in a large case-control study of 1,093 adenomas from 926 individuals.

Archival adenomas were collected at baseline from participants in a blinded, phase III clinical trial of wheat bran fiber supplementation for 3 years with colorectal adenoma recurrence as the end point (Wheat Bran Fiber trial), as previously described (41). Eligibility criteria included age of 40 to 80 years and prior removal of one or more ≥3 mm adenomas. There was a centralized pathology review of all polyps by the study pathologist (A.K. Bhattacharrya). This trial was comprised of individuals with sporadic colorectal adenomas with exclusion of those with two or more first-degree relatives with colorectal cancer or with evidence of syndromic familial CRC. One thousand four hundred and twenty-nine individuals were enrolled, and 1,304 completed the study by undergoing at least one follow-up colonoscopy. The subset of 926 individuals in the current study were selected on the basis of available tissue for molecular characterization and successful generation of data on Ki-ras mutation or p53 expression by immunohistochemistry. Six hundred and thirty nine of the patients were included in a previous study of Ki-ras mutation by another research laboratory (9).

Genomic DNA was extracted from routine formalin-fixed, paraffin-embedded, 5 μm thick, H&E-stained sections from 1,093 adenomas. Adenomatous glands equivalent to ∼10 mm² for each adenoma were microdissected by scraping with a scalpel from tissue slides to achieve ∼80% adenomatous tissue. DNA was extracted after the samples were transferred into 1.5 mL tubes and deparaffinized in xylene. Xylene was removed by vacuum centrifugation and specimens were treated with 50 μL of digestion buffer [0.5% Tween 20, 20 μg proteinase K, 50 mmol/L Trizma base (pH 8.9), and 2 mmol/L EDTA]. The tissue was digested overnight at 56°C, and proteinase K was inactivated at 100°C for 10 minutes.

Exon 1 of Ki-ras was amplified in 50 μL using 2 μL of 1:10 genomic DNA, 1× PCR buffer, 2 mmol/L MgCl, 0.8 mmol/L deoxynucleotide triphosphate mix, 2.5 units of AmpliTaq Gold, and 20 pmol/L of primers: K-ras (sense) 5′-GGCCGGTAGTGTATTAACCTTATGTGTGACAT-3′ and K-ras (antisense) 5′-CCGCGGCCGGCGGCCAAAACAAGATTTACCTCTATTGTTGG-3′.

PCR amplification conditions were denaturation at 95°C for 10 minutes, 14 cycles (95°C × 20 seconds, 59 ± 0.5°C/cycle × 60 seconds, 72°C × 60 seconds), 25 cycles (95°C × 20 seconds, 52°C × 60 seconds, 72°C × 60 seconds), and a final extension at 72°C × 10 minutes. Cycling conditions were done on a GeneAmp PCR System 9700 (ABI PRISM, Foster City, CA). The quality of the product was checked on a 2.5% agarose gel.

Five microliters of the PCR product was purified by mixing with 2 μL of Exo/Sap, incubated at 37°C for 15 minutes, and inactivated at 80°C for 15 minutes. DNA sequencing was done in 20 μL using 2 μL purified PCR product, 4 μL ABI PRISM BigDye Terminator version 3.0 cycle sequencing kit, and 10 pmol/L of forward primer with the following cycling conditions: initial denaturation at 95°C × 5 minutes followed by 25 cycles of 95°C × 20 seconds, 52°C × 60 seconds, and 60°C × 60 seconds. Following spin column purification (Princeton Separations, Princeton, NJ), the reaction products were sequenced by capillary electrophoresis using an ABI PRISM 3700 DNA Analyzer.

Immunohistochemistry for the p53 gene product was done with DO7 antibody (Cayman Chemical, Ann Arbor, MI) following antigen retrieval with target unmasking fluid (Kreatech, Amsterdam). Deparaffinized, rehydrated slides were immersed in target unmasking fluid and heated for 10 minutes at 90°C. Routine avidin-biotin complex technique (Vectastain, Burlingame, CA), diaminobenxidine chromagen, and methyl green counterstain were used. A set of positive cases with known p53 mutations and a negative irrelevant antibody were included in each staining run. p53 immunohistochemistry was considered to have overexpression when 30% or more of adenomatous, but not nonneoplastic, nuclei in all or part of the adenoma were stained (42).

Colorectal adenomas were classified by site as follows: distal colonic adenomas were located in the splenic flexure through the sigmoid colon and included the rectosigmoid; proximal colonic adenomas were located in the cecum through the transverse colon; and rectal adenomas were located within the rectum, defined as the first 12 cm above the anal verge. Histology was classified into the following categories: tubular, tubulovillous (25-75% villous component), and villous (>75% villous component). The degree of dysplasia was classified as either low-grade (mild to moderate) or high-grade (marked to carcinoma in situ). Adenomas were classified as advanced if they had any of the following characteristics: tubulovillous or villous histology, HGD, or ≥1 cm in size. A positive family history was defined as one or more first-degree relatives with colorectal cancer. A history of polyps prior to the baseline adenoma was self-reported.

All statistical analyses were done using SAS (Cary, NC) or STATA Corp. (College Station, TX) software. Individuals were categorized as positive or negative for Ki-ras mutation and/or p53 overexpression if at least one adenoma was positive for the alteration. The relationships between Ki-ras mutation and p53 overexpression and the various individual characteristics were evaluated by using both univariate and multivariate logistic regression models. The multivariate analysis models were adjusted for potential confounders as noted in the individual tables of results. Potential confounding variables were included in multivariate models if they changed the crude measure of association by ≥10%. Odds ratios (OR) and 95% confidence intervals (CI) were determined. Two-sample t tests were used for continuous variables and χ² for categorical variables. P trends were assessed for the association of age and number of adenomas.

Table 1 shows that there were no significant differences for any of the risk factors of interest between the subset of cases we used for molecular analyses and the study population of the Wheat Bran Fiber clinical trial, with the exception of a higher proportion of advanced adenomas in our subset. Although not significant, our subset included slightly more adenomas that were large, tubulovillous, and had HGD, which contributed to the significantly higher proportion of advanced adenomas that were defined by at least one of the three characteristics.

Ki-ras mutation was found in 14.7% of individuals and p53 gene product overexpression was found in 7.0% of those tested. As we have previously reported (10, 43), age was positively associated with the prevalence of Ki-ras mutations (Table 2). In addition, individuals with rectal adenomas were twice as likely to have Ki-ras mutations compared with those with distal colonic adenomas (Table 2). Univariate analyses of the relationships among adenoma size, degree of dysplasia, histology, and Ki-ras mutation showed positive associations with point estimates similar to those reported in the literature (15-18, 44-51). However, because these characteristics are highly correlated, multivariate analysis was required to assess the independent effect of each and showed that the presence of tubulovillous histology (OR, 4.19; 95% CI, 2.62-6.68) or of villous histology (OR, 5.45; 95% CI, 2.28-13.03) was strongly related to Ki-ras mutation as compared with tubular adenomas. The ORs for large (≥1 cm) adenomas and those with HGD were largely attenuated and therefore no longer statistically significant in multivariate models, although the OR for HGD was borderline statistically significant (P = 0.07). Individuals with an advanced adenoma had an ∼4-fold increase in the likelihood of a Ki-ras mutation (OR, 3.96; 95% CI, 2.54-6.18).

Table 3 shows the distribution of the Ki-ras mutation type, the affected codon, and the corresponding amino acid change. Among all mutations, 48.6% were a G-to-A transition, 42.8% were a G-to-T transversion, and 8.0% were a G-to-C transversion. Eighty-one percent (81.2%) of Ki-ras mutations were located in codon 12 and only 18.8% were located in codon 13. In addition, the types of mutations differed in the codons. In codon 12, G-to-T transversion was the most common mutation (41.3%) followed by G-to-A transition (31.9%) and G-to-C transversion (7.2%). By contrast, in codon 13, G-to-A transition was by far the most common mutation (16.7%) with only 1.4% a G-to-T transversion and 0.7% a G-to-C transversion. In codon 12, the most frequent change was a G-to-T mutation in the second nucleotide position of the codon, and the remainder were as follows: 23.2% G-to-A in the second position, 8.7% G-to-T in the first position, and 8.7% G-to-A in the first position. The corresponding amino acid changes were from a glycine to valine, aspartic acid, cysteine, and serine, respectively. In codon 13, the most frequent mutation was G-to-A in the second position, resulting in an aspartic acid substitution for glycine.

Table 4 summarizes Ki-ras mutation type by selected demographic and clinicopathologic characteristics. Mutations in codon 12 and G-to-A transition mutations were associated with older age. G-to-T transitions were more frequent in the rectum, as were codon 12 mutations. Mutations in codons 12 and 13, and mutations of the G-to-T and G-to-A type were all significantly more frequent in adenomas that had a villous component, were larger, or had HGD (i.e., advanced adenomas).

The associations between the characteristics of interest and p53 overexpression are presented in Table 5. A self-reported history of a previous polyp was inversely associated with p53 overexpression (OR, 0.35; 95% CI, 014-0.88). In regard to the adenoma characteristics, although substantial attenuation occurred in multivariate analyses, large size, presence of HGD, and villous histology were all positively and significantly associated with p53 overexpression. The strongest association with a single feature was with large size of an adenoma (OR, 5.25; 95% CI, 2.39-11.50). An even stronger association was seen for advanced adenomas with p53 overexpression (OR, 7.20; 95% CI, 3.01-17.22).

Our data showed that individuals with a Ki-ras mutation were ∼2.5 times as likely to also have p53 overexpression (OR, 2.46; 95% CI, 1.36-4.46). We further investigated the associations between the variables of interest in our analyses and the presence of either Ki-ras mutation or p53 overexpression or both (Table 6). Among the total population, 19.3% of the individuals were positive for either Ki-ras mutation or p53 overexpression, whereas only 1.4% were positive for both. Individuals with adenomas that had both a Ki-ras mutation and p53 overexpression were older than those without these alterations (P = 0.04). There was an 11% prevalence of a previous polyp among individuals with adenomas that had both Ki-ras mutation and p53 overexpression but was much higher among those lacking both alterations (40%, P < 0.0001). Although rectal adenomas were more likely to be positive for both Ki-ras mutation and p53 overexpression, this result was largely driven by the association with Ki-ras mutation (Table 2). In regard to the other adenoma characteristics, larger size, the presence of HGD, and villous histology were all more likely to be associated with both Ki-ras mutation and p53 overexpression. Ninety-four percent of adenomas positive for both Ki-ras mutation and p53 overexpression were advanced, whereas less than half of those negative for both were advanced (P ≤ 0.0001).

A number of studies have investigated the presence of Ki-ras and p53 alterations in colorectal adenomas, although these have generally addressed populations of 100 or fewer individuals (15-19, 30, 34-37, 44, 45, 47-49, 51-56). The current report of 1,093 adenomas from 926 individuals is the largest study in sporadic colorectal adenomas of the association of clinicopathologic and demographic characteristics with Ki-ras mutations and p53 overexpression.

The reported frequencies for mutation of Ki-ras in colorectal adenomas have varied considerably from 15% to 75%, with higher frequencies in larger adenomas, adenomas containing a focus of cancer, and individuals with prior CRC (10, 15-19, 34, 45, 46, 48, 49, 51, 54, 57, 58). This large variation is likely due to a number of factors, including different methods for Ki-ras mutation analysis, population differences, and selection bias. In our study of sporadic colorectal adenomas, the rate of Ki-ras mutation was 14.7% and that of p53 gene product overexpression was 7.0%. In univariate analyses of the adenoma characteristics, mutation of Ki-ras was associated with larger adenoma size (≥1 cm), the presence of villous histology, and HGD, but when multivariate models were applied, only villous histology remained independently associated with Ki-ras mutation. Likewise, in a previous analysis of 639 individuals from this study population done by a different laboratory, univariate analyses found Ki-ras mutations to be more common in adenomas that were larger, had more villous architecture, and had HGD, whereas multivariate analysis showed that Ki-ras mutation was independently associated with villous histology and HGD, but not adenoma size (10). This previous report differs from the current study in that adenomas 0.5 cm or larger were selected for analysis, whereas the current study did not select samples based on adenoma size. Several smaller studies, generally using univariate analyses, have investigated the relationships among Ki-ras mutation and histology, size, and grade of dysplasia. Many, but not all, show a significant association between Ki-ras mutation and adenoma size (16, 17, 44, 45) or more advanced histology (15, 18, 46-51, 58). HGD has also been associated with Ki-ras mutation in some (17, 44), but not all of these studies (16, 19).

In addition, we found Ki-ras mutations in adenomas to be independently associated with older age. Some reports have shown this relationship in both colorectal adenomas (10, 43) and carcinomas (24, 48), whereas other studies have not seen this relationship and have even found lower rates of mutation in older individuals (15, 24).

Using multivariate analyses, we found that Ki-ras mutations were significantly more common in rectal adenomas compared with those in the colon. We previously reported a similar but nonsignificant association (10). Some studies of CRC have also reported higher Ki-ras mutation frequencies in rectal cancers compared with other locations (9, 21), whereas others have not found an association (20, 24). In a large population-based study of CRC, Ki-ras mutations were higher in proximal compared with distal colonic cancers, but cancers of the rectum were excluded in that study (27, 59).

The majority of studies of CRC (9, 24, 27, 32, 60) and colorectal adenomas (10, 19, 61), including the present study, found G-to-T transversion and G-to-A transition mutations in codon 12 to be the most common alteration. To our knowledge, the finding of a significant relationship between older age and Ki-ras codon 12 mutations and G-to-A transitions in colorectal adenomas has not been reported previously. G-to-A transitions can result from the formation of guanine adducts in DNA, which are normally repaired by the O⁶-methylguanine DNA methyltransferase (MGMT) enzyme. In colorectal tumorigenesis, the MGMT gene is often silenced through methylation of the promoter, resulting in the accumulation of G-to-A mutations (62). In a recent study from our group (63), MGMT promoter methylation in the normal-appearing colorectal mucosa adjacent to a CRC was associated with older age (≥66 years). In addition, MGMT methylation was also associated with G-to-A mutations in the Ki-ras gene in the CRCs. Taken together, these results support the concept that a mucosal field defect in DNA repair develops with advancing age and may contribute to colorectal tumorigenesis.

We also report for the first time a higher frequency of G-to-T mutations in rectal adenomas compared with distal colonic adenomas or proximal colonic adenomas. This relationship has been previously described in CRC (9), although other studies of CRC have not supported this finding (14, 27, 64, 65). The rectum is exposed to very different environmental and biological factors compared with the colon. Possible explanations for an increase in G-to-T mutations in the rectum include exposure to differing fecal content within the colorectum (9), fermentation of carbohydrates (66), production of volatile fatty acids (24), and exposure to specific carcinogens (67).

The reported prevalence of p53 overexpression, as shown by immunohistochemistry, has varied, with the majority of studies showing that adenomas fall into the 5% to 38% range (15, 17, 19, 34-37, 39, 40, 53, 55, 56, 68). More than 50% of adenomas containing cancer (35) and 37% to 70% of cancers (19, 26, 37, 39, 40, 45, 53, 56, 69-72) have been found to have p53 overexpression. Interpretation of p53 immunohistochemistry is a common difficulty in generalizing studies and although the correlation with p53 mutation is far from perfect, p53 overexpression has been used as a surrogate for mutation in numerous studies (15, 19, 39, 40, 53, 55, 56, 68, 71). The cut point used to consider p53 immunohistochemistry overexpressed in the current study was very similar to that reported by Rashid et al. (15) who used the same 30% or greater level for considering an adenoma positive. In our study, p53 overexpression was independently associated with large adenoma size, villous histology, HGD, and advanced adenoma, and to our knowledge, this is the first and largest study in individuals with sporadic colorectal adenomas to show this association. In a previous study, comprised of a much smaller population, there was a significant association of p53 overexpression with villous histology but not size (15), whereas even smaller studies have found an association with HGD (19, 36, 55, 56, 68). The association of p53 overexpression with advanced adenomas was stronger than that observed for Ki-ras mutation, thus supporting the role of p53 in the later stages of CRC tumorigenesis.

Our multivariate analysis found that a history of a previous polyp was inversely associated with p53 overexpression. We report an interesting, albeit nonsignificant, reduction in p53 overexpression in individuals reporting a family history of CRC (OR, 0.40; 95% CI, 0.14-1.19). Our study excluded individuals with two or more family members with a history of CRC, which complicates the evaluation of this relationship. Two previous studies in CRC found an inverse association between family history of CRC and p53 mutation (73) or overexpression (74) compared with p53-negative tumors. These results in CRC provide supportive evidence that our finding of an inverse relationship between family history and p53 overexpression in colorectal adenomas was not due to chance. Several studies noted that tumors from individuals with hereditary nonpolyposis colorectal cancer rarely exhibit p53 alterations (74, 75), suggesting that CRC and adenomas from individuals with a family history of CRC develop through a mechanism that is p53-independent. In our study, there were no known cases of hereditary nonpolyposis colorectal cancer included. Results lead us to speculate that this association may indicate the possibility of other uncharacterized inherited predisposition pathways that do not require p53 mutation for progression. Our multivariate analysis also found that a history of a previous polyp was inversely associated with p53 overexpression. This result provides evidence against the involvement of the p53 gene in the mucosal field defect that contributes to colorectal tumorigenesis and that can be manifested by metachronous development of multiple polyps.

A major strength of the current study is that we report the largest study to date of Ki-ras mutations and p53 overexpression in a population of individuals with sporadic colorectal adenoma (10, 15-17, 19, 45). Additionally, our study population was well characterized, and we did not select adenomas for larger adenoma size, or select individuals based on risk factors such as HGD or a personal history of cancer, as previous studies have done (15-17, 19, 45). Our study design increases the generalizability of our findings. The limitations of our study include its cross-sectional design and the homogeneity of our study population with regard to race/ethnicity and other characteristics.

In summary, the important findings of our study are that colorectal adenomas with Ki-ras mutations were more likely to also overexpress p53, and furthermore, that individuals with an adenoma that had both a Ki-ras mutation and p53 overexpression were older, lacked a previous history of a polyp, and were more likely to have an advanced adenoma compared to those without these alterations. We also report, for the first time in this population, that specific types of Ki-ras mutations were associated with older age and rectal location. Our findings in this large cross-sectional study support the notion that both Ki-ras mutation and p53 alterations are important in the progression to more advanced adenomas. Our data suggests that the molecular pathogenesis of the sporadic adenoma-carcinoma sequence differs by anatomic location in the large bowel, age of the individual, and a mucosal predisposition state that is evidenced by the occurrence of multiple polyps.