Associations between Dietary Intake and Ki-ras Mutations in Colon Tumors: A Population-based Study¹

Although it is well established that cancer in humans occurs as a process involving multiple and variable events and stages, little is known about how these events relate to variability in outcome. Variability in colon tumors is reflected in the diversity in genetic alterations observed in tumors themselves, as well as by differences in age at diagnosis and in age and site-specific risk factors(1, 2, 3). Specific exposures that constitute risk factors for colon cancer could contribute to the heterogeneity of acquired genetic alterations observed in colon tumors. Therefore, it is reasonable to hypothesize that diet and other lifestyle and environmental factors are associated with some mutations but not with others.

Mutations in the Ki-ras gene were among the first linked to the pathogenesis of colon cancer (4). Ki-rasmutations are thought to be early events in colon cancer pathogenesis because they are observed in both adenomas and carcinomas(5, 6, 7). Most studies estimate that 30–50% of adenomas and carcinomas have Ki-ras mutations; ∼90% of Ki-ras mutations in colon tumors occur in codons 12 and 13(7, 8, 9).

Data to support specific associations between diet and genetic mutations in tumors are limited. In one case-control study of 108 sporadic colorectal cancers, low levels of calcium and high levels of monounsaturated fat were associated with increased odds of having Ki-ras+ tumors (10). Another study of adenomas observed that high levels of dietary plus supplemental folate were associated with decreased odds of having Ki-ras+ adenomas;dietary folate without supplements was not associated with Ki-ras+ adenomas (11). Because previous diet and Ki-ras mutations in colon tumors have included few people, it has not been possible to evaluate dietary associations with specific types of Ki-ras mutations. However, the study by Martinez et al. (11) did evaluate specific types of mutations with dietary factors and did not observe significant differences in dietary associations with specific types of mutations.

In this study, we used data collected as part of a large population-based case-control study of sporadic colon cancer to evaluate associations between dietary intake and Ki-rasmutations. Because we have data available on 1428 colon cancer cases and 2406 controls, we were able to evaluate dietary associations with the overall presence of any Ki-ras mutation in the tumor as well as with specific types and location of mutations. Although we focus on distinct pathways that have been hypothesized as being involved in colon cancer, including bile acids and dietary fat, DNA methylation, carcinogen detoxification, and insulin-related factors(12, 13, 14, 15), we also explore other dietary associations with Ki-ras. Because little information exists about the role of diet and distinct genetic mutations in tumors, this information also can be viewed as hypothesis generating.

Study participants were African-American, white, or Hispanic and were from either the Kaiser Permanente Medical Care Program of Northern California, an eight-county area in Utah (Davis, Salt Lake, Utah,Weber, Wasatch, Tooele, Morgan, and Summit counties), or the Twin Cities Metropolitan area in Minnesota. Eligibility criteria for cases included diagnosis with first-primary incident colon cancer (ICD-O Ed. 2, codes 18.0 and 18.2–18.9) between October 1, 1991 and September 30,1994; between 30 and 79 years of age at time of diagnosis; and mentally competent to complete the interview. Cases with adenocarcinoma or carcinoma of the rectosigmoid junction or rectum (defined as the first 15 cm from the anal opening), with known familial adenomatous polyposis, ulcerative colitis, or Crohn’s disease, were not eligible. Of all cases asked to participate, 75.6% participated.

Controls, in addition to the eligibility criteria for cases, could never have had a previous colorectal tumor. Controls were selected from eligibility lists for the Kaiser Permanente Medical Care Program,driver’s license lists for Minnesota, and random-digit-dialing,driver’s license lists, or Health Care Finance Administration lists for Utah. These methods have been described in detail(16). Of all controls selected, 63.7% participated.

Dietary intake data were ascertained using an adaptation of the validated CARDIA diet history questionnaire (17, 18). Participants were asked to determine which foods were eaten (using brand names of food items such as fast foods, cookies, crackers, and cereals, when possible), the frequency with which foods were eaten, and the type of fat used in preparation of foods. Three-dimensional food models were used to help participants estimate their usual serving size. Cue cards were used to provide a consistent prompt to help identify foods within broad categories. For categories within which many types of food might have been eaten (such as breakfast cereal),participants were asked to report the three most commonly eaten items. Detailed information was also obtained on foods eaten as additions to other foods (such as sugar added to cereal); standard amounts of additions were assigned per unit of the food item they accompanied. Nutrients were calculated using the Minnesota Nutrition Coordinating Center’s nutrient database version 19.

Foods were grouped into categories of red meat; processed meats(including hot dogs, luncheon meat, and sausage); eggs; low-fat dairy products (including milk, yogurt, and cheese); fruit (including fresh,frozen, and canned fruits); vegetables; cruciferous vegetables; whole grains; and refined grains. Food groups were developed by assigning each food item a standard serving size. The number of standard servings consumed from a food group were summed for each individual. A dietary glycemic index was created so that dietary carbohydrates could be weighed by their metabolic effect (19). A mutagen index was created that accounted for the amount of meat eaten, how it was prepared (i.e., well done, medium rare, and rare), and if it were precooked in a microwave.

Environmental exposure data were collected by trained and certified interviewers (20). The referent period for the study was the calendar year, ∼2 years prior to the date of diagnosis (cases) or selection (controls). Information was collected on demographic factors such as age, sex, and center; physical activity (16); body size, including usual adult height and weight 2 and 5 years prior to diagnosis; use of aspirin and/or nonsteroidal anti-inflammatory drugs;cigarette smoking history; and medical and reproductive history including use of hormone replacement therapy.

Methods for obtaining tumor tissue and extracting DNA have been discussed (21). Mutations in codons 12 and 13 of the Ki-ras gene were detected by PCR amplification and sequencing of exon 1 using DNA obtained from paraffin-embedded tissue blocks (22). Primers were tailed with universal primer and reverse primer for sequencing. PCR products were sequenced using prism Big Dye terminators and cycle sequencing with Taq FS DNA polymerase. DNA sequence was collected and analyzed on an ABI prism 377 automated DNA sequencer. We considered as mutations only those bp changes that were verified by sequencing in both directions (figure in Ref. 23). Tumor DNA was available from 1836 cases. Of these, 1428 were interviewed and had valid study data as well as Ki-ras data. We were able to successfully PCR amplify and sequence DNA from ∼95% of tumors we received to determine Ki-ras mutational status.

The distribution of specific types of mutations in tumors was determined. Logistic regression models, using two comparison groups,were fit to the data. A case-control comparison was conducted to estimate the relative risk of developing disease with specific genetic mutations. Evaluation of case-control data enables the evaluation of risk similar to that presented in the traditional case-control study. Controls were compared with cases with and without Ki-rasmutations in their tumors. Logistic regression models also were fit with the dichotomous dependent variables as either “no disease” versus “Ki-ras+ disease,” or “no disease” versus “Ki-ras− disease.” We also assessed associations of Ki-ras+ compared with Ki-ras−. The purpose of the “case-case” comparison was to evaluate etiological heterogeneity of the risk factors under study. For these analyses, the logistic regression models were fit with a dichotomous dependent variable (Ki-ras+ or Ki-ras−). We also assessed specific types of mutations that were relatively common. We evaluated lifestyle exposures with mutations in codon 12, mutations in codon 13, transversions, transitions, and three of the most common specific types of mutations: G→A of the second base of codon 12(2G→A), G→T of the second base of codon 12 (2G→T), and G→A of the second base of codon 13 (5G→A). All nutrients were evaluated using the density measure, or the amount per 1000 kcal. Categorization of nutrients was based on the distribution of nutrient values in the control population. Assessment of linear trend was done using continuous variables in logistic regression models.

Thirty-two % of tumors had Ki-ras mutations (Table 1). Of these, the majority were in codon 12 (80.7%). The most common type of mutation was a G→A transition on the second base of codon 12, which occurred in 10.5% of all colon tumors and represented 32.8% of all Ki-ras mutations.

Assessment of dietary fat did not show differences in association based on Ki-ras+ versus Ki-ras− tumor status (Table 2). Differences in risk were similar when comparing controls with individuals in either tumor group, i.e., those Ki-ras+ and those Ki-ras−. Few differences in Ki-ras mutational status were observed for dietary factors in DNA methylation pathways or insulin-related pathways. Cruciferous vegetables showed a statistically significant association with Ki-ras+ tumors. People who reported low levels (<0.37 servings/week) of these vegetables were less likely to have Ki-ras mutations when compared with those who ate more than four servings of cruciferous vegetables a week(OR,³ 0.6; 95% CI,0.4–0.9; P linear trend, 0.01).

Assessment of dietary factors by age showed that among people diagnosed prior to age 65 (data not shown in Table), there was an increase in risk of having Ki-ras mutations in tumors if high levels of dietary fat were consumed (OR, 1.7; 95% CI, 1.1–2.8). Among older people, low levels of intake of vitamin B₁₂ (OR,0.6; 95% CI, 0.4–1.0) and cruciferous vegetables (OR, 0.6; 95% CI,0.4–0.8) were associated with a reduced risk of Ki-rasmutations in tumors. High levels of alcohol were associated with a reduced risk of Ki-ras mutations in tumors in men (OR, 0.7;95% CI, 0.5–1.0). Low cruciferous vegetable intake was associated with a reduced risk of having Ki-ras mutations in tumors to a greater extent in women (OR, 0.5; 95% CI, 0.3–0.8) than in men (OR,0.9; 95% CI, 0.6–1.4).

Assessment of specific types of mutations (Table 3) support three distinct disease pathways. The 2G→A mutations were associated with dietary factors involved in the DNA methylation process; high levels of mutagen index(not shown in the Table) also were associated with a reduced risk of having a 2G→A mutation (OR, 0.6; 95% CI, 0.3–0.9; Plinear trend = 0.02). Interestingly, high levels of alcohol and low levels of dietary folate, vitamin B₆, and B₁₂ were associated with a reduced risk of having this type of Ki-ras mutation. A 5G→A mutation (second base of codon 13) was directly associated with levels of carbohydrate, refined grain intake, and glycemic index;low levels of cruciferous vegetable intake and high levels of processed meat intake were associated with reduced risk of a 5G→A mutation (for cruciferous vegetables: OR, 0.4; 95% CI, 0.2–1.0; P linear trend = 0.74; for processed meat: OR, 0.4; 95% CI,0.2–0.8; P linear trend = 0.14). Dietary fats were directly associated with 2G→T mutations. Low levels of lutein in the diet (data not shown in the Table) were associated with a reduced risk of 2G→T mutations (OR, 0.7; 95% CI, 0.3–1.4; P linear trend = 0.02).

Those who consumed low levels of cruciferous vegetables were less likely to have codon 12 Ki-ras mutations (OR, 0.7; 95% CI,0.5–1.0; P linear trend = 0.02). High levels of carbohydrate and glycemic index were associated with a greater risk of having a Ki-ras+ mutation on codon 13 (OR, 2.1; 95% CI,1.0–4.5; P linear trend = 0.02 and OR, 2.0;95% CI, 0.8–4.7; P linear trend = 0.34,respectively).

Low levels of intake of cruciferous vegetables and dietary factors associated with DNA methylation were associated with a reduced risk of having a transition type of Ki-ras mutation (for cruciferous vegetables: OR, 0.6; 95% CI, 0.4–0.9; for all dietary DNA methylation factors: OR, 0.7; 95% CI, 0.4–1.1). High levels of energy (OR, 1.8;95% CI, 0.1–3.0; P linear trend = 0.60),fat (OR, 1.5; 95% CI, 0.9–2.4; P linear trend = 0.04), monounsaturated fat (OR, 1.3; 95% CI, 0.8–2.0; P linear trend = 0.06), and eggs (OR, 1.5;95% CI, 1.0–2.4; P linear trend = 0.07)were associated with a greater risk of having a transversion Ki-ras mutation. High levels of carbohydrate were associated with a greater likelihood of not having a transversion Ki-ras mutation (OR, 0.5; 95% CI, 0.3–0.9; Plinear trend = 0.10).

We observed few dietary factors that were associated with Ki-ras mutations overall. Clusters of dietary factors that may influence distinct disease pathways were associated with the three most common types of Ki-ras mutations. Folate and dietary factors possibly involved in DNA methylation pathways (13)were associated with 2G→A mutations; carbohydrates, glycemic index,and refined grains were associated with 5G→A mutations; and dietary fats were associated with 2G→T mutations. Low levels of cruciferous vegetables were associated with fewer Ki-ras mutations. Cruciferous vegetables are inducers of the phase II metabolizing enzyme, glutathione S-transferase, which is involved in multiple biological mechanisms (14). Other dietary factors were inconsistently associated with Ki-ras mutations (see Appendix 1).

One study that evaluated dietary factors and Ki-rasmutations observed that a high intake of calcium decreased the risk of having a Ki-ras mutation, whereas a high intake of monounsaturated fat increased the risk of these mutations(10). In that study, there were 108 colorectal cancer cases, and thus estimates of association were less precise than those presented here. We did not observe an association between dietary calcium that was specific to tumors with Ki-ras mutations. We did observe direct associations with monounsaturated fat for specific types of mutations, as mentioned earlier, although associations with overall Ki-ras mutations were much weaker. High levels of dietary heterocyclic amines influence Ki-rasmutations in rats (24). In this study, we observed that both the intake of processed meats and dietary mutagen index, which was created to discriminate between people with varying levels of heterocyclic amines in their diets, were associated with specific types of Ki-ras mutations.

Perhaps one of the most intriguing observations from this study is that, for several dietary factors, particularly those involved in DNA methylation pathways, the results observed are different from those reported recently for adenomas (11). In the study of adenomas, only folate supplements were associated with reducing risk of Ki-ras mutations, whereas dietary folate and alcohol were not. Other dietary factors involved in the DNA methylation pathway were not reported in the work by Martinez et al.(11). It was hypothesized in that report that when the availability of methyl donors was low, a G→A transition would be generated (11). In this study, we observed that dietary factors involved in the DNA methylation pathway were associated with G→A transitions. However, low levels of intake of nutrients involved in this pathway, including folate, vitamin B₆, and vitamin B₁₂, and high levels of intake of alcohol were associated with a greater likelihood of being Ki-ras− rather than having a G→A mutation. In this study population, we observed that folate and vitamins B₆ and B₁₂ were inversely associated with colon cancer in the total population(25). Because the overall association between these nutrients and cancer was inverse and these same nutrients were directly associated with this specific Ki-ras mutation, it is likely that these dietary factors associated with DNA methylation are involved in other disease pathways. Our results do suggest that associations with adenomas and tumors are different, and that dietary factors associated with DNA methylation act differently at the initiation versus promotion stage of the carcinogenic process. The sequence of genetic alterations in adenomas and tumors may have important significance as we unravel how dietary factors contribute to the carcinogenic process.

Two other pathways, distinct from the pathway involving DNA methylation, appear to exist. One of these involves 2G→T Ki-ras mutations and high levels of dietary fat; the other involves 5G→T Ki-ras mutations and high levels of carbohydrates, glycemic index, and refined grains. Specific chemical or molecular mechanisms involved can only be speculative. However, this information may provide insight into how Ki-ras functions. These observations also may explain differences in dietary associations observed between epidemiological studies.

It is possible that our findings do not have biological meaning but are the result of bias associated with misclassification. Given that Ki-ras mutations are early events in the carcinogenic process, it is possible that dietary changes have been made so that the referent date of 2 years prior to diagnosis is not reflective of usual adult diet. Likewise, those for whom data are less likely to be available are those with tumors at a more advanced disease stage. We observed both in our collection of tumor tissue (17) and in our collection of dietary data (26) that individuals with more advanced disease are less likely to participate in the study. In our attempts to estimate selection bias associated with exclusion of these individuals from the dietary collection segment of the study, we did not observe specific associations by disease stage(26). However, we have observed that those with more advanced tumors are more likely to have Ki-ras mutations in codon 12.

There are several strengths to our study. Our study is population based rather than cases derived from select populations of high-risk individuals or hospital-based studies. Given the large sample size, we were able to evaluate specific types of Ki-ras mutations in tumors, which provides more insight into distinct disease pathways. Although not all cases had exposure data, the distribution of Ki-ras mutations was not different for those with and without interview data; there does not appear to be selection bias involving Ki-ras mutation status.

Because there was limited information to guide our analyses, we evaluated dietary factors that had been associated with colon cancer along three pathways that have been hypothesized as being important to colon cancer. Because we evaluated numerous dietary factors, findings could be the result of chance. However, we believe that clustering of similar dietary factors as a group into what could represent distinct disease pathways lends support for the observed associations being real. Repeating these analyses in other large study populations will help to distinguish between these possibilities.

In summary, our results suggest that dietary intake influences distinct Ki-ras mutations that could represent distinct disease pathways, rather than Ki-ras mutations overall. These results, together with previous research, suggest that there may be differences between Ki-ras mutations in adenomas and Ki-ras mutations in colon tumors. In reading these results,however, it should be kept in mind that although some factors, such as dietary folate, were shown to increase the likelihood of specific Ki-ras mutations, these nutrients were associated with a decreased risk of developing colon cancer.

a 

We acknowledge the contributions and support of Melanie Nichols,Kristen Gruenthal, Margaret Robertson, and Heather Linn at the University of Utah’s DNA sequencing core facility and of Dr. Bette Caan and Leslie Palmer for the data collection efforts of this study.