Interactions of Peroxisome Proliferator–Activated Receptor γ and Diet in Etiology of Colorectal Cancer

The peroxisome proliferator–activated receptor γ (PPARγ) is one of a group of ligand-activated nuclear receptors that are responsible for regulation of glucose and lipid homeostasis as well as cell differentiation and apoptosis. The importance of PPARγ in the colon is implied by variation of expression of PPARγ within the distal and proximal colon (1, 2) as well as by increased levels in more differentiated epithelial cells (1, 3-6). Free fatty acids are known to bind to PPARγ with polyunsaturated fatty acids having the higher binding affinity than saturated or monounsaturated fatty acids (7, 8) and with greater activation stemming from oxidized polyunsaturated fatty acids than unoxidized fatty acids (9).

The 12 proline-to-alanine (Pro12Ala) substitution polymorphism in PPARγ produces proteins with lower activity (10, 11). Presence of the Pro12Ala variant polymorphism is reported to be associated with lower body mass index (BMI), improved insulin sensitivity, and a reduced risk for type 2 diabetes (12-20). Therefore, it is possible that PPARγ is associated with risk for colorectal cancers through insulin-related mechanisms. However, PPARγ's role in modulating other genes (21), potentially different coactivators with specific biological activity (22, 23), and potential activation from elevated levels of cyclooxygenase-2 (a common feature of colorectal cancer; refs. 24, 25) or B-catenin and TCF4 as a result of upstream APC mutations (26, 27) adds to the reason for exploring associations with colon and rectal cancer but may make interpretation complex.

Activation of PPARγ seems positive for colon health (9) by restricting the S-phase entry into the cell cycle and thereby inhibiting the proliferation of malignant cells including colorectal carcinomas; however, data to date do not uniformly support such a role. A small study (n = 55 cases) suggested that mutation with loss-of-function may be associated with colon cancer (28), but two studies failed to find any mutations in the PPARγ gene in 397 cancer specimens of differing tissue origin (29, 30). Two animal studies suggested that activation of PPARγ was associated with progression, rather than prevention, of cancer (31, 32); one reported an increase in colon tumors with activation of PPARγ in mice on high fat diets, suggesting that PPARγ might mediate the influence of a high fat diet on colon cancer (32).

Little information is available regarding interactions of PPARγ genotypes and dietary exposures in risk for colon and rectal cancer. A small case-control study from Spain suggested that the Pro12Ala variant of PPARγ was more protective among individuals with low vitamin A intake (33); α and γ tocopherol have been noted to up-regulate expression of PPARγ mRNA in SW 480 colon cells (34).

We investigated nutrients that could interact with PPARγ based on the pathways that PPARγ is involved in. Dietary fat and sugars could be associated with insulin resistance, whereas antioxidants could interact with PPARγ in the regulation of inflammation. Thus, we evaluated whether the associations of fats, antioxidants, specific foods, and the combined effects expressed as dietary patterns (35) on risk for colon and rectal cancer were modified by the Pro12Ala variant of PPARγ using data from two (colon and rectal) case-control studies conducted in Northern California, Utah, and the Twin City Metropolitan area of Minnesota (colon cancer study only).

Two study populations from the Kaiser Permanente Medical Care Program of Northern California (KPMCP), the state of Utah, and the Twin City Metropolitan area of Minnesota (colon cancer study only) are included in these analyses. All eligible cases within these defined geographic areas were identified and recruited for the colon cancer study. The first study includes incident cases and controls from a population-based case-control study of first primary colon cancer (International Classification of Diseases for Oncology, 2nd edition codes 18.0, 18.2-18.9) diagnosed between October 1, 1991 and September 30, 1994 conducted in all three geographic areas (36). Cases with a first primary tumor in the rectosigmoid junction or rectum were identified between May 1997 and May 2001 in Utah and KPMCP. Case eligibility was determined by the Surveillance Epidemiology and End Results Cancer Registries in Northern California and in Utah and the Minnesota Surveillance System (colon cancer cases only). In both studies, cases were identified using rapid-reporting systems. For the colon cancer study, the median time from diagnosis to interview was 131 days overall (126 days at KPMCP, 154 days in Minnesota, and 109 days in Utah). The median time from diagnosis to interview was longer for the rectal cancer study, primarily because of different levels of permission needed before contacting patients; at KPMCP, the median days from diagnosis to interview was 154 and for Utah was 183.

For both studies, eligibility included being between 30 and 79 years of age at time of diagnosis, English speaking, mentally competent to complete the interview, no previous history of colorectal cancer (36), and no known (as indicated on the pathology report) familial adenomatous polyposis, ulcerative colitis, or Crohn's disease. Participants were eligible for the colon cancer study if they were non-Hispanic White, Hispanic, or African American; Asian and Native American populations also were included in the rectal cancer study.

Controls were frequency matched to cases by sex and by 5-year age groups. At the KPMCP, controls were randomly selected from membership lists and, in Utah, controls ages ≥65 years were randomly selected from social security lists and controls ages <65 years were randomly selected from driver's license lists. A total of 794 rectal cancer cases and 1,001 matched controls and 1,577 colon cancer cases and 1,971 matched controls are included in the analyses presented. Response rates for the rectal study were 65.2% for cases and 65.3% for controls; cooperation rates, or the number of people who participated from those who we were able to contact was 73.2% for cases and 68.8% for controls. For the colon cancer study, response rates were 64.0% for cases and 64.0% for controls.

Interviews were conducted by trained and certified interviewers and collected using laptop computers. The interview took ∼2 hours. Quality control methods used in the rectal cancer study were the same as those used in the colon cancer study; these methods have been described in detail previously (37).

Dietary intake was ascertained using an adaptation of the CARDIA diet history (38-40). Participants were asked to recall for the calendar year occurring 2 years before their cancer diagnosis (cases) or selection into the study (controls), the frequency that foods were eaten, serving size, and if fats were added in the preparation. Food intake data were created with the Nutrition Data System for Research, nutrient data using the Minnesota Nutrition Coordinating Center nutrient Database version 4.04_30©2003, Regents of the University of Minnesota. Dietary patterns were developed using factor analysis with eigenvalues of >1.25 used to limit the number of factors and create meaningful factors (41). Component scores were created by combining observed variables with weights proportional to the factor loading using a method previously described (42). Tertiles of nutrient intake (in appropriate unit per day) were determined by the sex-specific distribution of intake in controls and are expressed as high, medium, and low intake. The Western dietary pattern loaded heavily (factors with loadings of over 0.30) on processed meats, red meat, fast-food meat, eggs, butter (men only), margarine, potatoes, high-fat dairy foods (men only), legumes, refined grains, added sugar (men only), sugar drinks (men only), and sugar desserts. The prudent diet pattern loaded heavily on all types of fruits and vegetables, whole grains, fish, and poultry (43). Mutagen index was calculated as the frequency of red meat, poultry, and fish consumption prepared by frying, broiling, baking, or barbecuing plus the use of drippings from red meat, poultry, or fish, multiplied by the usual doneness of red meat, poultry, and fish (1 = rare, 2 = medium rare, 3 = medium well, and 4 = well done) and the microwave factor (1 = microwave never used or used for thawing, 0.75 = sometimes used, 0.50 = often used, and 0.25 = always used). A higher index reflects higher intake of potentially mutagenic compounds.

DNA was extracted from whole blood. The Pro12Ala (C > G) polymorphism of PPARγ was assessed using a Taqman assay. Primer and probe oligonucleotides were obtained from Applied Biosystems (Foster City, CA), Assays on Demand (assay id C_1129864_10). Briefly, 20 ng of genomic DNA were amplified in a 17-μL reaction containing 8.5 μL of 2× Taqman Universal PCR Master Mix and 0.72 μL of C_1129864_10 (pparg pro12 ala) 20× Assay on Demand mix. PCR amplification was done using the BIO-RAD iCylcer under the following conditions: 95°C for 10 minutes followed by 40 cycles of 95°C for 15 seconds and 62°C for 1 minute. Real-time fluorescence was collected from each sample and analyzed using version 3 of the iCycler IQ Real-Time detection software.

Height and weight were measured at the time of interview. The BMI of weight (kg)/height² (m) was calculated for men and women using reported height and weight data from 2 years ago. Weight and height were also reported for the 2 and 5 years before interview. Physical activity data were collected using a physical activity questionnaire that has been described elsewhere (44).

Unconditional logistic regression models were used to estimate the associations of foods, food patterns, and nutrients to risk of colon and rectal cancer by PPARγ genotype. We assessed the PPARγ genotype as most common PP and any variant PA or AA. Most associations of dietary variables and risk for colon and rectal cancer by PPARγ polymorphism were not significantly different by sex; therefore, to increase power and simplify data presentation, we pooled colon and rectal cancer cases by sex and present the exception in the text. We assessed dietary data by determining risk across tertiles, or cutoffs used for previous articles (45) depending on the distribution of the data in controls for men and women separately. In these models, the following variables were included: age at selection, BMI, physical activity, energy intake, and dietary calcium. We tested interaction, or effect modification on both the additive scale (46) using the Relative Excess Risk due to Interaction and on the multiplicative scale to better understand the relationship of the data. Both tests are appropriately applied to relative risks or odds ratios (OR; i.e., the additive test is not a test of risk differences, it is a test of additive effects of risk ratios; ref. 47). Brennan (48) describes the applicability of both types of interaction to carcinogenic processes and provides a discussion of the assumptions of both.

Participants in the colon cancer study were slightly older than those in the rectal cancer study (Table 1). Participants were more often male and White than female or other ethnicities. BMI was slightly greater among colon cancer cases than controls, but similar between rectal cancer cases and controls. Nonsteroidal anti-inflammatory agents were less commonly used by colon cancer and rectal cancer cases than controls. Leisure time vigorous physical activity was less among colon cancer cases than controls but similar between rectal cancer cases and controls. Energy intake was slightly greater among colon and rectal cancer cases than controls, whereas calcium and fiber intakes were similar among cases and controls in both studies. The PP PPARγ genotype was present in roughly 77% of the population studied, the PA genotype in 21% of the population studied, and the AA genotype in 1.5% of the population studied.⁵

The associations of dietary macronutrients and colon and rectal cancer were not statistically significantly modified by the PPARγ genotype (Table 2). Risk of rectal cancer was increased among those with the PA or AA genotype and low intake of monounsaturated fatty acids, although not statistically significantly (OR, 1.78; 95% CI, 1.10-2.88; P for relative excess risk for interaction = 0.14). No interactions between total fat, saturated, monounsaturated, or polyunsaturated fats and PPARγ genotype on risk of colon or rectal cancer were detected either when fatty acids were analyzed as density (per 1,000 cal) or as absolute amount (g/d). We found no interaction of the PPARγ genotype with linoleic acid, docosahexaenoic acid, or eicospentaenoic acid on risk for colon or rectal cancer. Although an increase was seen in risk for rectal cancer among individuals with low total or animal protein intake and the PA or AA genotype (data not shown), the interaction with PPARγ genotype was not statistically significant. Further adjustment for nonsteroidal anti-inflammatory drugs (NSAIDS) in addition to other covariates did not materially alter the risk estimates.

We examined whether the association of antioxidants differed by PPARγ genotype (Table 3). Risk for rectal cancer was increased among those with the PA or AA genotype who consumed low levels of tocopherol, but the interaction was not statistically significant. Interactions with the specific forms of vitamin E, α, γ, and δ tocopherol, were not statistically significant. High intakes of lutein were associated with a decreased risk of colon cancer among those with the PA or AA genotype. The associations of β-carotene were not statistically significantly altered by PPARγ genotype. Additional adjustment for NSAIDS resulted in a significantly reduced risk of colon cancer among those with a high β-carotene intake and the PA or AA genotype (OR, 0.69; 95% CI, 0.51, 0.93) but did not modify any of the other antioxidant associations. Stratification by BMI category, NSAID, or ibuprofen use did not materially alter these findings.

When we examined food consumption and food pattern interaction with PPARγ genotype, the risk of colon cancer was lower among individuals both with the PA or AA genotype and either the lowest intake of refined grain or a high score on the prudent diet (Table 4). Although the risk for colon cancer was reduced among those with a low score for Western diet and the PA or AA genotype, the interaction was not statistically significant. Risk for rectal cancer was increased with high mutagen index among individuals with the PA or AA genotype. Stratification by BMI category, NSAID, or ibuprofen use did not alter these findings.

We describe decreased risk of colon cancer among those with the PPARγ PA or AA genotype and with low intake of refined grains or a high score on the prudent diet. We also found reduced risk of colon cancer among individuals with the PPARγ PA or AA genotype and the highest intake of lutein and with the highest intake of β-carotene (after adjustment for NSAIDS use). Risk of rectal cancer was significantly greater among individuals with the PPARγ PA or AA genotype and a high mutagen index. These results support the idea that consumption of a prudent diet may be useful in prevention of colon cancer. Diets with a low mutagen index may be more important in prevention of rectal cancer among those with the PPARγ PA or AA genotype.

Previously, we reported an inverse association of specific antioxidants with colon (49) and rectal cancer (50). The present results support modest inverse associations of antioxidants with risk for colon and rectal cancer that seem more evident among those with the PA or AA PPARγ genotype. A significant decrease in risk of colon cancer with high lutein intake was evident in individuals with PA or AA genotype. Additionally, we observed a significant decrease in risk of colon cancer in individuals with PA or AA genotype with high β-carotene intake and further adjustment for NSAIDS. Rectal cancer risk seemed to increase with decreasing intake of lutein and β-carotene among individuals with the PA or AA genotype, but the interaction was not significant. Although the biological plausibility of an interaction of vitamin E and PPARγ genotype is supported by the report that α and γ tocopherols increased PPARγ mRNA expression in colon cancer cells (34), we found no significant increase in the risk for colon or rectal cancer among those with any variant of PPARγ and low intake of total tocopherols.

We investigated the influence of foods and dietary patterns on risk differed for colon and rectal cancer by PPARγ genotype. We previously reported increased risk of colon cancer among those with a score for high western diet pattern and a family history of colon cancer and a direct association of intake of refined grain with risk of rectal cancer (35). Colon but not rectal cancer risk was decreased among those with the PA or AA genotype and low refined grain intake or high prudent diet score. These associations fit the pattern that we observe of insulin resistance–related pathways and factors in the etiology of colon cancer but not rectal cancer (51) and the decreased risk for diabetes, lower BMI, and increased insulin sensitivity with the PA or AA genotype (12-20). However, unpublished data from this same study do not support an association with obesity. Thus, it is not clear whether nutrients and/or these dietary patterns operate as ligands for PPARγ or whether they operate directly in anticarcinogenic pathways that are more apparent in the presence of lower PPARγ activity.

Study strengths include the large number of cases and controls in both colon and rectal studies. Despite the large sample size, the relatively low frequency of the polymorphisms combined with stratification results in somewhat limited power to detect three-way interactions of diet, polymorphisms, and colon and rectal cancer risk. We trained and certified interviewers and used detailed questionnaires and rigorous quality control methods that may have helped reduce inaccurate recall of study subjects. As in any study, we cannot rule out the possibility of recall bias. Multiple comparisons increase the risk of spurious findings; therefore, findings should be replicated in other study samples. Additionally, because all data were obtained for a reference date 2 years before cancer diagnosis, it is possible that cases were asked to recall a time period that is not relevant in the etiology or that is post-cancer development. In previous analyses, we had assessed dietary associations by stage at diagnosis and did not observe differences in association for those diagnosed when at stage IV and those diagnosed at stage 1 (52).

These data suggest that PPARγ genotype modifies the association of a prudent diet (direct association) and refined grains (inverse association) with risk for colon cancer. In light of the association of the common PPARγ ProAla12 polymorphism with lower BMI, decreased risk of type 2 diabetes, and increased insulin sensitivity (12-20), it is possible that the associations with colon cancer may operate through or in conjunction with insulin-related pathways. However, unpublished data from this same study indicate that there is no association with obesity or waist-to-hip ratio. Therefore, it seems equally likely that antioxidants or other nutrients found at high levels in the prudent diet and depleted from refined grains may be the operable substance acting independently. Additionally, it seems more likely that the association with mutagen index and rectal cancer operates through cell cycle regulation or inflammation than insulin-related mechanisms. Thus, although there are multiple biological pathways in which PPARγ genotype may modify associations with diet and colon and rectal cancer, further research is needed to determine the relevant mechanisms.

We thank the contributions of Michael Hoffman, Thao Tran, and Kazuko Yakumo for genotyping and Joan Benson, Sandra Edwards, Roger Edwards, Leslie Palmer, Donna Schaffer, and Judy Morse for data collection and analysis components of the study.