Toll-like receptor genes and their association with colon and rectal cancer development and prognosis†
The contents of this article are solely the responsibility of the authors and do not necessarily represent the official view of the National Cancer Institute.
Toll-like receptors (TLRs) are mediators of inflammation in the gut and potentially important modulators of colon and rectal cancer risk. We use data from a population-based study of incident colon cancer cases (n = 1,555) and controls (n = 1,956) and rectal cancer cases (n = 754) and controls (n = 959). We evaluate genetic variation in TLR2 (six SNPs), TLR3 (four SNPs) and TLR4 (eight SNPs) with risk of developing colon or rectal cancer and survival after diagnosis. TLR3 rs11721827 was associated with rectal cancer (odds ratio [OR] 1.27, 95% confidence interval [CI] 1.02, 1.58 for AC/CC vs. AA genotype, Wald p = 0.035; adjusted p = 0.126); TLR3 rs3775292 and TLR4 rs11536898 were associated with colon cancer (OR 0.68, 95% CI 0.49, 0.95 for GG vs. CC/CG and OR 0.50. 95% CI 0.29, 0.87 for AA vs. CA/CC, respectively; Wald p = 0.023 and 0.015; adjusted p = 0.085 and 0.101, respectively). TLR2 rs7656411 and rs3804099, respectively, interacted with nonsteroidal anti-inflammatory drug (NSAID) use and cigarette smoking to alter risk of colon cancer (adjusted p = 0.034 and 0.077); TLR3 rs11721827 interacted with NSAID use to alter risk of colon cancer (adjusted p = 0.071). TLR3 rs3775292 interacted with dietary carbohydrates to alter colon cancer risk and with dietary carbohydrates and saturated fat to alter rectal cancer risk (adjusted p = 0.064, 0.0035 and 0.025, respectively). Multiple SNPs in TLR2 and TLR4 were associated with colon cancer survival. Although few independent associations with TLR genes were observed, we observed significant interaction of TLR2 and TLR3 with hypothesized lifestyle factors. Interaction with dietary factors remained significant for rectal cancer after adjustment for multiple comparisons.
The role of inflammation in the development of colon and rectal cancer is well documented.1 Toll-like receptors (TLRs) are mediators of inflammation in the gut and may therefore be important modulators of colon and rectal cancer risk.2 TLRs are crucial regulators that help in maintaining the balance between commensal bacteria in the gut and the mucosal immune system.3 Breakdown of the homeostasis in the gut may be a key feature in the pathogenesis of inflammatory bowel disease and possibly of risk of colon and rectal cancer. Several TLRs have been examined in the intestine. TLR3 is expressed in normal intestinal epithelial cells2; TLR4 is upregulated in Crohn's disease and ulcerative colitis; expression of TLR4 and TLR2 is increased in lamina propria macrophages in inflammatory bowel disease.4, 5 TLR4 also induces COX2 and is important for proliferation and apoptosis in response to intestinal injury. A study by Xiao et al.6 showed that decreased TLR signaling resulted in increased intestinal inflammation further suggesting the importance of TLR signaling in colon carcinogenesis.2 Studies also have demonstrated that TLRs play a role in progression of polyps to tumors and that reduced TLR4 expression is associated with increased metastatic potential of colorectal cancers (CRCs).7, 8 TLR5 has been shown to protect the gut from enteric microbes.9
The effects of genetic variation in the TLR genes have been examined in a few studies of Crohn's disease and cancer. A TLR2 GT repeat microsatellite and TLR4 D299G were associated with CRC in a study of 89 cancer cases.10TLR2 and TLR4 were examined in 182 Crohn's disease patients.11 The TLR4 D299G polymorphism was significantly associated with increased risk of Crohn's disease. Single nucleotide polymorphism (SNP) rs3775291 of the TLR3 gene was associated with CRC survival when examined among 614 CRC patients in Germany.12 The impact was greatest among those diagnosed with stage II cancers.
Although there are suggestions of the importance of TLRs in colon and rectal cancer etiology and survival, there is limited information on associations with genetic variation in these genes. It is reasonable to hypothesize that exposure to aspirin/non-steroidal anti-inflammatory drugs (NSAIDs), dietary fat and carbohydrates and cigarette smoking may modify the risk associated with TLR genes. TLR4 has been shown to be required for inducing COX2 expression after intestinal epithelial cell injury.3 Commensal bacteria, which are essential for maintenance of homeostasis in the gut, have been shown to be influenced by dietary fat, especially saturated fats and carbohydrates,13 and TLRs are regulators of this homeostasis. Oxidative stress may enhance intestinal injury and inflammation; cigarette smoking has been shown to influence oxidative stress.14 In our study, we examine colon and rectal cancer risk associated with genetic variation in TLR2, TLR3 and TLR4; we evaluate if NSAID use, dietary intake of fat and carbohydrates and cigarette smoking interact with these genes to alter risk; we assess the association between genetic variation in these TLR genes and survival.
Material and Methods
Two study populations are included. The first, a population-based case–control study of colon cancer, included cases (n = 1,555) and controls (n = 1,956) identified between October 1, 1991 and September 30, 1994 living in the Twin Cities Metropolitan Area, Kaiser Permanente Medical Care Program of Northern California (KPMCP) and a seven-county area of Utah.15 The second study used identical data collection methods as the first study but included population-based cases with cancer of the rectosigmoid junction or rectum (n = 754) and controls (n = 959) who were identified between May 1997 and May 2001 in Utah and KPMCP.16 Eligible cases were between 30 and 79 years old at time of diagnosis, English speaking, mentally competent to complete the interview, no previous history of CRC and no known (as indicated on the pathology report) familial adenomatous polyposis, ulcerative colitis or Crohn's disease. Controls were matched to cases by sex and by 5-year age groups. At KPMCP, controls were randomly selected from membership lists. In Utah, controls 65 years and older were randomly selected from the Health Care Financing Administration lists and controls younger than 65 years were randomly selected from driver's license lists. While in Minnesota, controls were selected from driver's license and state-identification lists. Study details have been reported earlier.15, 16
Interview data collection
Data were collected by trained and certified interviewers using laptop computers. All interviews were audiotaped and reviewed for quality control purposes.17 The referent period for the study was 2 years prior to diagnosis for cases and prior to selection for controls. Detailed information was collected on diet,18 physical activity, medical history, cigarette smoking history, regular use of aspirin and NSAIDs and body size. The Nutrition Coordinating Center (NCC) nutrient database was used to calculate nutrients from reported foods on the extensive diet history questionnaire.
Tumor registry data
Tumor registry data were obtained to determine disease stage at diagnosis and months of survival after diagnosis. Disease stage was categorized using the sixth edition of the American Joint Committee on Cancer (AJCC) staging criteria. Disease staging was done centrally by one pathologist in Utah. Local tumor registries also provided information on patient follow-up including vital status, cause of death and contributing cause of death. Follow-up was obtained for all study participants and was terminated for the Colon Cancer Study in 2000 and for the Rectal Cancer Study in 2007. At that time, all study participants had over 5 years of follow-up.
TagSNP selection and genotyping
All SNPs evaluated were tagSNPs and selected using the following parameters: linkage disequilibrium (LD) blocks were defined using a Caucasian LD map and an r2 = 0.8; minor allele frequency (MAF) > 0.1; range = −1,500 bp from the initiation codon to +1,500 bp from the termination codon and one SNP/LD bin. All markers were genotyped using a multiplexed bead-array assay format based on GoldenGate chemistry (Illumina, San Diego, CA). A genotyping call rate of 99.85% was attained. Blinded internal replicates represented 4.4% of the sample set; the duplicate concordance rate was 100%. Individuals with missing genotype data were not included in the analysis for that specific marker. A description of the genes and corresponding SNPs is available in Supporting Information Table 1; Supporting Information Table 2 details the distribution of the genotypes for cases and controls for all SNPs assessed. All SNPs were in Hardy–Weinberg equilibrium.
Statistical analyses were performed using SAS® version 9.2 (SAS Institute, Cary, NC). We report odds ratios (ORs) and 95% confidence intervals (95% CIs) assessed from adjusted multiple logistic regression models adjusting for age, center, ethnicity (approximately 90% Caucasian, 5% African American and 5% Hispanic) and sex which were matching variables for the original study. Inheritance models were selected based on initial review of results from the additive model. When risk estimates indicated that a recessive or dominant model was more appropriate that model was used. If the MAF was too low (less than 0.1 for evaluation of main effects; less than 0.15 for evaluation of interactions), then a dominant model was assigned. Analysis for interaction was based on tagSNPs within each TLR gene, unadjusted for other SNPs. Diet and lifestyle variables were selected because of their biological plausibility for involvement in this candidate pathway. Interactions between TLR genes and hypothesized exposures associated with inflammation and oxidative stress (i.e., recent aspirin or NSAID use) and cigarette smoking (recent or not recent smoker), or their effect on commensal bacteria in the gut (saturated fat and carbohydrates). Aspirin/NSAID use and cigarette smoking were categorized into two levels to maximize power. Nutrients were determined from the University of Minnesota Nutrition Data System Nutrient Database that was used to covert reported foods into nutrients. Tertiles of dietary intake per 1,000 calories were based on the sex-specific distribution in controls. For colon cancer, the cut points for carbohydrates were 260 and 366 g for men and 205 and 282 g for women; for saturated fat, the cut points were 23.8 and 38.6 g for men and 17.7 and 28.1 g for women. The corresponding cut points for the rectal cancer study for carbohydrates were 271 and 290 g for men and 220 and 318 g for women and were 24.4 and 41.7 g for men and 20.6 and 32.6 g for women for saturated fat. p values for interaction were determined using a likelihood-ratio test comparing a full model that included an interaction term with a reduced model without an interaction term.
Survival months were calculated based on month and year of diagnosis and month and year of death or date of last contact. Associations between SNPs and risk of dying of CRC within 5 years after diagnosis were evaluated using Cox proportional hazards models to obtain multivariate hazard rate ratios (HRRs) and 95% CIs. We adjusted for age at diagnosis, study center, ethnic group, sex, tumor molecular phenotype and AJCC stage to estimate HRRs.
Adjusted multiple-comparison p values (pACT), taking into account tagSNPs within the gene, were estimated using the methods by Conneely and Boehnke19via R version 2.11.0 (R Foundation for Statistical Computing, Vienna, Austria). Wald p values from the main effect models and interaction p values based on likelihood-ratio tests were used for estimates of multiple comparisons. We consider a pACT of <0.20 as being potentially important given the candidate pathway approach and the need to consider both type 1 and type 2 errors. We believe that findings at this level would merit replication.
We detected few statistically significant associations between TLR2, TLR3 and TLR4 and colon and rectal cancer (Table 1). The only main association that was significant for colon cancer was with TLR4, where the AA genotype of rs11536898 was inversely associated with colon cancer risk (OR 0.50, 95%CI 0.29, 0.87). After adjustment for multiple comparisons, the pACT was 0.101 compared to the Wald p value from the unadjusted model of 0.0148. TLR3 rs11721827 was significantly associated with rectal cancer (OR 1.27 95% CI 1.02, 1.58); the adjusted pACT value was 0.127 while the unadjusted values was 0.0351.
Table 1. Summary of SNPs
Evaluation of interaction between diet and lifestyle factors showed significant interaction for both TLR2 and TLR3 and NSAID use and for TLR2 and cigarette smoking for colon cancer (Table 2). Lower risk was associated with having recently used aspirin/NSAIDs in the presence of the variant allele of TRL2 rs7656411 and TLR3 rs11721827; genotype did not influence risk among those who did not recently use aspirin/NSAID. The CC genotype of TLR2 rs3804099 reduced risk associated with cigarette smoking but has no impact among nonsmokers. TLR3 rs3775292 interacted with both total carbohydrates for colon and rectal cancer and with saturated fat for rectal cancer. Although there was significant interaction for both colon and rectal cancer with this polymorphism, the direction of the association was different. The CG/GG genotypes were associated with reduced risk of rectal cancer at high levels of carbohydrate intake and slight nonsignificant increased risk at low levels of intake, while the opposite was true for colon cancer. Similar patterns of association were observed for saturated fat intake and rectal cancer.
Table 2. Interaction between TLR polymorphisms and risk of colon cancer
Both TLR2 and TLR4 were associated with survival after diagnosis with colon cancer but not with rectal cancer (Table 3). TLR2 rs5743704 and rs5743708 were associated with poorer survival (HRR 1.89, 95% CI 1.26, 2.83 and HRR 1.74, 95% CI 1.12, 2.70, respectively). These two SNPs were not in LD, with R2 values of <0.01. The CC genotype of rs1554973 of TLR4 was associated with better survival (HRR 0.55; 95% CI 0.31, 0.98) after diagnosis with colon cancer. Stratification by disease stage at time of diagnosis into those with AJCC stages 1 and 2 and those with stages III or IV showed two significant associations (data not shown in table). TLR2 rs5743704 CA/AA genotypes were associated with worse survival when diagnosed with stages III or IV (HRR 2.10; 95% CI 1.23, 3.59), while there was no impact on survival among those diagnosed at stages I or II. TLR3 rs3775292 CG/GG genotypes were associated with slightly better survival if diagnosed at more advanced disease stage (HRR 0.74; 95% CI 0.56, 0.99), while there was no impact on survival among those with disease stages I or II at diagnosis (HRR 0.96; 95% CI 0.55, 1.67). There were no significant associations with survival after diagnosis with rectal cancer for any of the TLR genotypes.
Table 3. Associations between TLR genes and survival after diagnosis with colon or rectal cancer
Few associations were observed between TLR genes and risk of developing colon or rectal cancer. Associations were restricted primarily to colon cancer, although one SNP in TLR3 showed a modest unadjusted significant association with rectal cancer. The rs11536898 of TLR4 was associated with reduced risk of colon cancer; TLR2 rs7656411 and TLR3 rs11721827 interacted with NSAIDs use while TLR2 rs3804099 interacted significantly with cigarette smoking. TLR3 rs3775292 interacted with carbohydrate and saturated fat intake. Both TLR2 and TLR4 polymorphisms were associated with survival after diagnosis with colon cancer.
Both NSAID use and cigarette smoking interacted with genetic variation in TLR2 and TLR3 as hypothesized. Although most studies have focused on TLR4 and have shown that it is required for the induction of COX2 in response to injury,3 TLR2 and TLR3 also have been shown to have a protective effect against intestinal injury. Given that NSAIDs are a COX2 inhibitor, it is possible that the influence of TLRs that stem from genetic variation in these genes may be influenced by the presence of COX2 inhibitors such as NSAID use. Smoking cigarettes can influence cellular oxidative stress; studies have further shown that oxidants in cigarette smoke inhibit pathogen recognition receptor function that may involve TLRs.14 TLR2 also has been shown to provide a signal that initiates inflammatory response in mice in a state of oxidative stress.5 Our observed interaction between cigarette smoking and TLR2 polymorphisms is biologically plausible; we observed that in the presence of cigarette smoke, variation in the TLR2 gene influenced risk of colon cancer.
Dietary components may have an important role in regulating gut microflora. Intake of a high fat, especially high saturated fat and high carbohydrate diet that influence gut microflora can induce inflammation, oxidative stress and endotoxins as well as TLR expression.20 Intestinal microflora can stimulate the immune system through TLR and negative regulation of TLRs is essential for maintaining intestinal homeostasis.21 We observed that TLR3 rs3775292 significantly interacted with dietary carbohydrates for both colon and rectal cancer, while the same SNP also interacted significantly with saturated fat for rectal cancer. Given the same polymorphism was associated with colon and rectal cancer for both carbohydrates and saturated fat, we believe that these observations should be replicated in other studies. It is not clear why this interaction would be seen only for TLR3 genetic variation. It is possible that as TLR3 is expressed in normal colon tissue it may play an important role at the early stages of tumor initiation. If that were the case, this would imply that dietary factors such as high intake of carbohydrates and saturated fat may function as initiators in the carcinogenic process. Studies have shown that TLR4 promotes the development of colitis-associated tumors.4, 22 Other studies have shown that TLR4 is associated with metastatic status of CRC.7 TLR4 expression has been shown in normal colon epithelium and CRC cell lines, while a loss of expression has been shown to increase metastases.7 Other studies have shown that TLR2 expression increased as stage increased from normal tissue to stage III tumors, while TLR3 expression decreased.8 While the evidence is somewhat limited, there is support that TLRs may influence survival. TLR3 rs3775291 of the TLR3 gene was associated with CRC survival among those diagnosed at stage II in a study of 614 CRC patients in Germany.12 This finding was not replicated in our patient population. We did observe that TLR3 rs3775292 CG/GG genotypes were associated with slightly improved survival among those diagnosed with more advanced colon cancer. However, our data support that genetic variation in TLR2 and TLR4 influence survival for colon cancer but not rectal cancer. Genome Wide Association Studies (GWAS) studies have not identified TLR genes among their top few hits; however, of the four independent associations detected, three were for the homozygote rare variant and they may not have the ability to detect these associations. In addition, GWAS have generally not taken into account the influence of diet and lifestyle factors; our data suggest that interaction with these factors is needed to define the risk associated with genes. These findings need replication in other studies.
The study has many strengths; the dataset is rich in exposure data, genetic data, tumor marker data and survival information. In addition, we are able to evaluate colon and rectal cancer separately given the vast literature that suggest that these are two separate diseases.16, 23–26 Because of this, we have the ability to thoroughly evaluate TLR to determine not only broad independent associations but also to help identify molecular pathways through which they may operate. In doing this, we made many comparisons, and thus, we determined p values that took into account these comparisons. The study is hypothesis driven, examining factors that may influence genetic susceptibility. However, a limitation is lack of information on functionality of polymorphisms examined in our study. Our results provide insight for follow-up studies that evaluate these polymorphisms to determine their functionality. Study limitations include our inability to evaluate rare SNPs that may be associated with colon or rectal cancer. Although the SNP selection criteria we used to haplotype the gene provided coverage at those specified levels in the methods, it is possible that we do not completely cover the gene given our lack of assessment of extremely rare variants. In addition, we were limited to markers that were designable on our selected platform. In developing our custom platform, we identified approximately 200 genes along a candidate pathway that we hypothesized as important. Our analysis approach has been to evaluate smaller components of the pathway and thus not adjust for genes that we have not analyzed that are on the platform. We believe that this approach is legitimate and correlates with approaches in epidemiology that do not adjust for every item on a questionnaire when a more focused analysis is being conducted. We believe that avoiding false negatives is important and that replication of potentially important findings is needed by other studies.
In summary, we provide links between TLR genes which have been hypothesized as important to the carcinogenic process of CRC. Our data suggest that genetic variation in TLR2, TLR3 and TLR4 may influence the development of colon cancer as well as influence of survival after diagnosis with colon cancer. As hypothesized, these associations were influenced by recent use of NSAIDs and cigarettes. Dietary carbohydrates and saturated fat appeared to influence risk associated with these genes for both colon and rectal cancer. Replication of these findings is warranted given the biological plausibility of the association and the interaction with diet and lifestyle factors that can modify these genetic effects. Further, we believe that our study provides an important mechanistic link that could be the starting point for further studies, and important SNPs in these genes could possibly serve as biomarkers for decision making in CRC treatment.
The authors would like to acknowledge the contributions of Dr. Bette Caan, Donna Schaffer and Judy Morse at the Kaiser Permanente Medical Care Program in Oakland, CA; Sandra Edwards, Roger Edwards and Leslie Palmer at the University of Utah and Drs. Kristin Anderson and John Potter at the University of Minnesota for data management and collection.