Prostate cancer is the most common incident cancer and the second most common fatal cancer among US men. One in six US men will develop this cancer during their lifetime.1 The established risk factors, which are age, family history and race, provide limited insight into who is more likely to develop prostate cancer. Efforts to further define risk factors have focused on biologic processes thought to contribute to prostate cancer etiology, including chronic inflammation. While the source of inflammation in the prostate is unclear, this process is proposed to involve both the innate and adaptive immune systems and the production of numerous reactive species capable of causing mutations and promoting carcinogenesis.2
An essential component of the innate immune system is the family of toll-like receptors (TLR), which are members of the interleukin-1 receptor superfamily and mediate immune responses to foreign pathogens, including bacteria, fungi and viruses. Ten human TLRs have been identified that share significant homology in their cytoplasmic domains and differentially recognize various microbes.3 Upon binding of a microbial ligand, TLRs activate signaling pathways that stimulate cytokine production and other parts of the innate immune response.3 In addition to recognizing ligands derived from foreign microbes, TLRs have been reported to bind numerous endogenous ligands, including heat shock proteins, fibrinogen, messenger RNA and heparin sulfate.4 Thus, the TLR-mediated immune response may be activated in the absence of foreign microbes.
Common polymorphisms in the TLR genes have been studied primarily in relationship to autoimmune diseases and susceptibility to various infections. However, interest in the role of these receptors in cancer has been increasing. Studies of common genetic variants in various populations (reviewed in Ref. 5) and of growth stimulation of cancer cell lines6 suggest that TLR4 plays an important role in the development of H. pylori-associated gastric cancer. Other studies have found that common polymorphisms in TLR2 are associated with increased risk of colorectal cancer,7 gastric cancer8 and lymphoma,9 and a specific haplotype in TLR10 is associated with increased risk of nasopharyngeal cancer.10
Several other studies have investigated these genes in prostate cancer. Three large studies focused on the TLR4 gene,11–13 while 2 investigated a gene cluster that includes the TLR10, TLR1 and TLR6 genes.14, 15 For TLR4, various SNPs in this gene have been associated with increased risk,11 decreased risk12 or increased risk for one SNP and decreased risk for another SNP.13 The role of genetic variants in the TLR6, TLR1 and TLR10 gene cluster with prostate cancer has been investigated in a Swedish case-control population and the Health Professional Follow-up Study (HPFS). Eleven SNPs in this gene cluster were associated with a statistically significant increased risk of prostate cancer14 in the Swedish population but no association was found for these same SNPs using the HPFS population.15 Thus, the studies of association of the TLR4 gene and the TLR6, TLR1 and TLR10 gene cluster with prostate cancer have yielded inconsistent results.
To clarify whether TLR genes influence prostate cancer risk, we have used a large population of 1,414 cases and 1,414 matched controls derived from the American Cancer Society Cancer Prevention Study-II (CPS-II) Nutrition Cohort to investigate associations with common variants in the TLR10-TLR1-TLR6 gene cluster. We expanded the SNP coverage beyond that previously investigated14, 15 and genotyped 28 SNPs in the region of chromosome 4p14 that includes the genes for these 3 TLRs. The association of the 5 haplotype blocks that span this region and interactions with factors that may influence inflammation were also investigated.
Material and methods
Men in our study were enrolled in the American Cancer Society Cancer Prevention II Nutrition Cohort, a prospective study of cancer incidence. Approximately 184,000 US adults between the ages of 50 and 74 completed a mailed questionnaire in 1992 that included questions on demographics, diet and other lifestyle factors. The recruitment and characteristics of this cohort have been described previously.16 Follow-up questionnaires were sent to all living Nutrition Cohort members in 1997 and every 2 years afterwards to update exposure information and to ascertain newly diagnosed cases of cancer. Incident cases reported via questionnaire response were verified through medical records, linkage with state cancer registries or death certificates.16 Between June 1998 and June 2001, blood samples were collected from a subset of Nutrition Cohort participants (21,965 women and 17,411 men), fractionated into serum, plasma, buffy coat and RBCs, and stored in liquid nitrogen vapor phase at −130°C until needed for analysis.
We identified 1,414 men who had been diagnosed with prostate cancer between 1992 and 2003 and had no previous history of cancer (except nonmelanoma skin cancer) from the men who provided the blood samples. An equal number of controls were matched to the cases on age (±6 months), race/ethnicity (White, African American, Hispanic, Asian or other/unknown) and date of blood collection (±6 months). Controls were selected from men who were cancer-free at the time of cancer diagnosis of their matched case using risk set sampling.17 Among the cases, 554 men had advanced prostate cancer, which was defined as prostate cancer with stage C or D on the Jewitt-Whitmore staging system and/or Gleason score 7–10 at diagnosis, or men who had prostate cancer as their underlying cause of death.
SNP selection and genotyping
SNPs were selected to attempt to replicate the findings of the previous study14 and to expand the coverage of the TLR10-TLR1-TLR6 gene cluster to thoroughly interrogate all possible common variants in the region. Thus, the SNPs genotyped for our study (Table II) included the 18 investigated by Sun et al.14 and 7 additional nonsynonymous SNPs in the 3 TLR genes not included in the previous study.14 Five additional SNPs identified as haplotype-tagging (ht) SNPs using the method developed by Gabriel et al.18 from HapMap data available in January 2006 were also selected. Genotyping attempts failed for 2 of the SNPs (rs5743816 and rs5743618). Therefore, a total of 28 SNPs were investigated in our study.
Genotyping was done in 2 stages. In the first stage, the genotype of all 28 SNPs were determined in 553 prostate cancer cases and 553 matched controls. These cases included most of the advanced prostate cancer cases (529) and 24 nonadvanced cases. In the second stage, those SNPs found to be associated with prostate cancer at a significance level of p < 0.1 (10 SNPs) were genotyped in the remaining 861 cases and 861 controls.
Genotyping was done using the MassARRAY system (SEQUENOM, San Diego, CA) at the Center for Human Genomics (Wake Forest University School of Medicine). PCR primers were designed using SpectroDesigner software (SEQUENOM). Laboratory personnel were blinded to the case-control status of the samples and 3.5% blinded replicates were randomly included with the samples for quality control of the genotyping. Concordance for these replicates was 100% for all SNPs. The genotyping success rates were all >97% and the genotype distributions of the controls for all but 3 SNPs were in Hardy-Weinberg equilibrium at p > 0.05. Three SNPs in the TLR1 gene (rs4833095, rs5743551 and rs5743604) violated Hardy-Weinberg equilibrium. Careful examination of the raw genotyping results revealed no detectable genotyping error. Because these genotype distributions could arise by chance or could reflect a significant association with the cancer endpoint, the results for these SNPs were not excluded from our study.
The statistical significance of the difference in allele distribution of each polymorphism between cases and controls was calculated using the Chi-square test with 1 degree of freedom. Linkage disequilibrium (LD) measurements (D′ and r2) were made using Haploview. Haplotypes for blocks defined through Haploview18 using HapMap data release phase II from July, 2006 were estimated using the expectation-maximization algorithm implemented in the TAGSNPS program.19 Odds ratios (OR) and 95% confidence intervals (CI) for the association between the TLR SNPs or haplotypes and prostate cancer incidence were determined using unconditional logistic regression. All models were adjusted for age (in single year categories), race (White or other) and date of blood draw (in single year categories). Other covariates that were considered for the analysis were family history of prostate cancer, education, energy intake, diabetes and PSA screening. None of these covariates were included in the logistic model because they did not alter the OR by more than 5%.
Effect modification of the association between the SNPs and prostate cancer by a history of benign prostatic hyperplasia (BPH, reported no or mild symptoms consistent with BPH in 1997 vs. moderate to severe symptoms), use of nonsteroidal antiinflammatory drugs (NSAID, non vs. current users), regular use of NSAIDs (non vs. users of 15 + times a month) and BMI (<25 vs. ≥25) was evaluated using the likelihood ratio test. Statistical significance was determined by calculating a p-value for the difference in the trends among genotypes for each stratum. Reported p-values were two-sided and those ≤0.05 were defined as being statistically significant. p-values were corrected for multiple comparisons using the Bonferroni method where indicated.
Our study population consisted of 1,414 prostate cancer cases and 1,414 controls. The cases and controls were similar in age (a matching factor), BMI, number of ever smokers and ethanol consumption (Table I). The cases were more likely to have a family history of prostate cancer than were the controls. Among the cases, 35 died from prostate cancer, 73% were classified as Gleason score 5–7 and 69% had PSA levels between 4 and 50 ng/ml.
Table I. Characteristics of the CPS-II Nutrition Cohort Participants Used in this Study Based on Information Reported at Baseline in 1992 or at Time of Diagnosis (for Tumor Characteristics)
Cases (N = 1,414)
Controls (N = 1,414)
Among drinkers, which were 1,330 (94%) cases and 1,329 (94%) controls.
Prostate cancer with stage C or D on the Jewitt-Whitmore staging system and/or Gleason score 7–10 at diagnosis, or men who had prostate cancer as their underlying cause of death.
The TLR10-TLR1-TLR6 gene cluster covers ∼54 kb on chromosome 4p14. This gene cluster and the locations of the 28 SNPs genotyped in our study within the cluster are shown in Figure 1. More details of the position of each SNP in the genes, any amino acid change, and the frequency of the minor allele in the cases and controls are listed in Table II. The 10 SNPs marked with “3” were genotyped in all 1,414 cases and 1,414 controls and include 6 for which the difference in minor allele frequency had unadjusted p-values less than 0.05. None of the other SNPs appeared to be associated with prostate cancer risk based on this statistic.
Table II. Characteristics and Genotype Frequencies of the TLR Gene Polymorphisms Analyzed in this Study Among Cases and Controls
The multivariate-adjusted associations of the 6 SNPs for which the difference in prevalence of the minor allele between cases and controls was statistically significant are shown in Table III. All of these variants, which include 2 nonsynonymous SNPs in TLR10, 2 nonsynonymous SNPs, 1 intronic SNP and 1 SNP in the 5′UTR of TLR1, were associated with a decreased risk of prostate cancer. The reduction in risk ranged from 36 to 41% for the TLR1 SNPs (for the homozygous variant genotype for all cases), and was 22% for both TLR10 SNPs. The ORs for each SNP were decreased proportionally to the number of minor alleles, suggesting that the polymorphisms functioned in a codominant manner. Restricting the analyses to cases with advanced prostate cancer yielded similar results as obtained when all cases were considered.
Table III. Association of SNPs in TLR Genes with Incidence of Prostate Cancer in all Cases and Advanced Cases
ORs are adjusted for age, race/ethnicity and date of blood draw.
0.84 (0.72, 0.98)
0.95 (0.77, 1.17)
0.78 (0.61, 0.99)
0.68 (0.48, 0.96)
0.84 (0.72, 0.98)
0.96 (0.78, 1.18)
0.78 (0.61, 0.99)
0.69 (0.49, 0.97)
0.90 (0.77, 1.05)
1.03 (0.84, 1.27)
0.64 (0.47, 0.86)
0.59 (0.39, 0.91)
0.79 (0.66, 0.93)
0.86 (0.69, 1.07)
0.59 (0.38, 0.91)
0.58 (0.31, 1.07)
0.82 (0.70, 0.97)
0.93 (0.75, 1.15)
0.63 (0.42, 0.93)
0.69 (0.40, 1.19)
0.90 (0.77, 1.06)
1.03 (0.84, 1.27)
0.67 (0.50, 0.91)
0.68 (0.45, 1.03)
The genotyped SNPs included a number which were in LD and fell into 5 haplotype blocks defined with the HapMap project data by Haploview using the method developed by Gabriel et al.18 (Fig. 1). The association of various haplotypes for Blocks 1 and 2 in TLR10 and Block 4 in TLR1 with prostate cancer risk are shown in Table IV. Results for Blocks 3 and 5 are not shown because only two or one, respectively, of the SNPs genotyped were within those blocks. For Block 1, the ATCTC haplotype, which contains the variant alleles of the I369L and N241H SNPs from TLR10, was associated with a statistically significant reduced risk of prostate cancer when compared to the most common haplotype for this block (OR = 0.62; 95% CI: 0.41, 0.95). The CCCTCCG haplotype in Block 4, which contains the variant alleles of the N248S, S26L, rs5743595 and rs5743551 SNPs in TLR1, was also associated with a statistically significant reduced risk when compared to the most common haplotype (OR = 0.79; 95% CI: 0.68, 0.91). None of the common haplotypes for Block 2 were associated with altered risk of prostate cancer.
Table IV. Association of TLR Haplotypes with Prostate Cancer (All Cases) Based on Presence of a Variant Haplotype Versus the Major Haplotype
We investigated whether the significant associations with both the TLR10 and TLR1 SNPs was likely to reflect a common haplotype (D′ > 0.87) or independent associations using a haplotype based on the 6 significantly associated SNPs. The haplotype containing the minor alleles of the 6 associated SNPs (CCCTCG) was associated with a statistically significant decreased risk of prostate cancer (OR = 0.77, 95% CI: 0.67, 0.89) when compared to the major haplotype. Having a single copy of this inclusive haplotype resulted in a similar decreased risk (OR = 0.80, 95% CI: 0.67, 0.94) whereas carrying 2 copies of the CCCTCG haplotype resulted in greater decreased risk (OR = 0.58, 95% CI: 0.37, 0.91, global p-value = 0.025). In both the haplotype analysis and that for individual SNPs, there was a dose-response in the association for the CCCTCG haplotype, consistent with it acting in a codominant manner.
Inflammation and the relative levels of pro- and anti-inflammatory cytokines can be altered by nonpathogen-related factors, including BPH, use of NSAIDs and elevated BMI. However, none of these factors were found to modify the association of the TLR SNPs with prostate cancer risk (data not shown).
Six SNPs in the TLR10-TLR1-TLR6 gene cluster were found to be significantly associated with a reduced risk of prostate cancer. The 4 SNPs in the TLR1 gene, rs4933095 (N248S), rs5743596 (S26L), rs5743595 and rs5743551 share a fairly high degree of LD (D′ > 0.98, r2 between 0.60 and 0.96) and appear to represent a single association. Likewise, the 2 SNPs in TLR10, rs11096955 (I369L) and rs11096957 (N241H), are in complete LD (D′ = 1, r2 = 1.0) and are likely to reflect a common single allele. The possibility that the TLR1 and TLR10 associations were not independent was suggested by the considerable co-occurrence of the minor alleles and was confirmed by the finding that a single haplotype consisting of just the 6 significantly associated SNPs was associated with decreased risk of the same magnitude. Thus, we found that a single haplotype within the TLR10-TLR1-TLR6 gene cluster is associated with a decreased risk of prostate cancer.
Similar associations were found regardless of whether all cases or only advanced cases were considered, suggesting that the genetic variation influenced the initial development of prostate cancer rather than the progression of this disease. Risk decreased as the number of variant alleles increased, consistent with a codominant mechanism for this genetic effect. The magnitude of the association ranged from OR = 0.62 to 0.81, depending on which SNP or haplotype block was analyzed, and was statistically significant, with p-values ranging from 0.001 to 0.024 (Table I). Global p-values for the associated haplotype blocks were also significant (p = 0.049 for Block 1, p = 0.019 for Block 4 and p = 0.025 for the inclusive block). Bonferroni correction of these results for false positives resulting from multiple comparisons increases the p-values from 0.028 to 0.672 (correcting for 28 SNPs) and the global p-values to 0.25 (Block 1), 0.113 (Block 4) and 0.15 (inclusive block). Thus, the association for the TLR1 S26L (rs5743596) SNP remains statistically significant after conservative correction for multiple comparisons.
Although our findings suggest an association with decreased risk, they do not agree with the conclusions of either of the previous reports of associations of SNPs in the TLR10-TLR1-TLR6 gene cluster with prostate cancer.14, 15 The main findings of our study and the other 2 studies for the 3 SNPs that we found significantly associated with prostate cancer risk that were genotyped in the other studies are summarized in Table V. All 3 SNPs were found to be associated with decreased risk in the HPFS but were interpreted as showing no association because the results were not statistically significant.15 However, as suggested by the authors of that study, this may be because the study population was not large enough to detect a statistically significant association. In fact, the distribution of the minor alleles of these 3 SNPs in the HPFS cases and controls is the same as we found, indicating that the previous results are similar to our present findings.
Table V. Comparison of the Findings for the Association of TLR10-TLR1-TLR6 SNPS with Prostate Cancer Risk
Results listed under CPS-II refer to the present study, those under HPFS refer to the Health Professionals Follow-Up Study findings15 and those under CAPS refer to the Cancer Prostate in Sweden findings14.
ORs are adjusted for age, race/ethnicity and date of blood draw.
ORs are adjusted for age.
0.85 (0.72, 0.99)
0.78 (0.61, 0.98)
0.79 (0.63, 1.01)
0.91 (0.63, 1.30)
AC + CC
1.25 (1.04, 1.50)
0.84 (0.72, 0.99)
0.79 (0.63, 1.01)
0.90 (0.63, 1.28)
AC + CC
1.20 (1.00, 1.43)
0.91 (0.78, 1.07)
0.95 (0.75, 1.19)
0.71 (0.53, 0.96)
0.76 (0.48, 1.18)
AG + GG
1.29 (1.06, 1.56)
The findings of the Swedish case-control study (CAPS), which were statistically significant, suggest that these SNPs are associated with increased risk of prostate cancer.14 This could indicate that our findings are due to chance. Alternatively, unknown differences in the study populations could contribute to the disparate findings. Both study populations consisted of Caucasian men of similar ages, although the stage of prostate cancer of the Swedish cases was somewhat more advanced than our cases. Comparison of the minor allele frequencies of the SNPs genotyped by both studies indicated that the occurrence of the minor alleles among cases was fairly similar while there was a greater difference between the controls in the 2 studies. What could be responsible for this difference between controls in our and the Swedish population is unclear.
The association of selected SNPs in the TLR10-TLR1-TLR6 gene cluster with prostate cancer has also been investigated by the Cancer Genetics Markers of Susceptibility (CGEMS) project, for which the results of a genome-wide association study involving 1,177 prostate cancer cases and 1,105 controls from the prostate, lung, colorectal and ovary (PLCO) study are available online (http://cgems.cancer.gov/).20 This study of >500,000 SNPs through the genome found no evidence for an association of 19 SNPs in this gene cluster, including the TLR10 N241H (rs11096957) SNP, which is one of the SNPs we found associated with a statistically significant decrease in prostate cancer risk. As with the study of Swedish cases and controls,14 why our findings are different from those of the CGEMs study is not clear. However, follow-up genotyping with independent populations has shown that a number of SNPs significantly associated with prostate cancer risk were not identified as such by the initial CGEMs scan.21 Thus, it could be that the association of the TLR10-TLR1-TLR6 gene cluster was also missed in this whole genome study.
Currently, it is not clear if altered TLR function should increase or decrease cancer risk. TLR1 and TLR6 recognize bacterial triacylated and diacylated lipoproteins, respectively, both when heterodimerized with TLR2.3 No ligands have been identified for TLR10. The discovery that TLRs mediate the enhanced immune responses induced by many immunoadjuvants22, 23 has lead to the suggestion that increased TLR activation may stimulate anticancer immunity.24, 25 Thus, enhanced TLR activity could inhibit carcinogenesis while decreased activity could allow cancer cells to escape immune detection and elimination. Alternatively, TLR activation may promote carcinogenesis by creating a proinflammatory environment that enhances tumor growth and chemoresistance26 and/or result in chronic inflammation-induced immunosuppression that allows cancer progression.24 In this case, decreased activity would minimize chronic inflammation and decrease the possibility of cancer development. Further basic and clinical research into the effects of TLR activation by either microbial or endogenous ligands on these pro- and anti-carcinogenic processes will be necessary before accurate predictions of how variation in TLR action will influence cancer risk can be made.
Determining which SNP, if any of those we studied, is responsible for our findings is complicated by the fact that 4 of the 6 SNPs associated with decreased risk of prostate cancer result in amino acid substitutions. The TLR proteins have 3 domains: an ectodomain that resides either outside the cell or in the cytoplasm (depending on the cellular location of the receptor) that contains multiple leucine-rich repeats (LRR) involved in pathogen recognition and binding, a transmembrane domain and a toll/interleukin-1 receptor (TIR) domain which interacts with signaling machinery to stimulate the immune response.27 The 4 nonsynonymous SNPs are located in the LRR of their TLR.28 Thus, it is possible that one or more of these SNPs influence the ability of the receptor to bind the pathogens they normally recognize.28 Only one of these 4 SNPs has been studied independent of the others to determine its effect on TLR. The TLR1 N248S SNP (rs4833095) did not alter expression of the receptor by cells transfected with the gene engineered to contain only this variant.29 However, the influence of this SNP on TLR1 function was not tested, so the possibility that the N248S replacement in TLR1 influences prostate cancer risk cannot be eliminated.
In addition to the 4 nonsynonymous SNPs, we must consider the possibility that other SNPs that was not genotyped in our study but are in LD with these variants are the functional polymorphisms responsible for the reduced prostate cancer risk. A nonsynonymous SNP in TLR1 (rs5743618, I602S), which is located in the transmembrane domain of the receptor, has been found to diminish the response to cells homozygous for this variant to bacterial stimuli.29, 30 This SNP, which causes a defect in the intracellular trafficking of the TLR1 that prevents the receptor from getting to the cell surface where it functions,30 is reported to be in LD with the TLR1 N248S SNP (rs4833095).29 Therefore, the TLR1 I602S SNP may be the variant responsible for the associated observed for the SNPs which it is in LD with. Proving this will require animal experiments with TLR1 engineered to contain only this SNP because it will not be found independent of the other potentially functional SNPs in humans.
No modification of the association of the 6 SNPs with reduced risk of prostate cancer was found for several factors expected to influence inflammation, including elevated BMI, NSAID use and presence of BPH in 1997. While the nested case-control design provided us with prospectively collected information on these factors, the sample size of our study limited the exploration of effect modification by these factors to only 2 groups. This may have resulted in insufficient difference between the groups to observe significant results.
Strengths of our study include the high quality genotyping, which minimizes genetic misclassification, the large study population, which is the largest used to investigate the TLR genes in prostate cancer to date and the nested case-control design, in which covariate information is collected prior to prostate cancer diagnosis and minimizes recall bias. The SNP coverage through the TLR10-TLR1-TLR6 gene cluster was more thorough than in previous studies but it did not permit the identification of a causative SNP or SNPs. A limitation to the study is a lack of understanding of how changes in TLR activity influence chronic inflammation and carcinogenesis that would facilitate the prediction of how SNPs that alter receptor action would affect these processes.
In summary, we found a significant association of a common genetic variant within the TLR10-TLR6-TLR1 gene cluster with decreased risk of prostate cancer. However, while these findings are statistically significant and consistent for a large number of SNPs in high LD across genes, they do not clarify the importance of these inflammation-related genes in prostate cancer because of the disparity with previous findings. Thus, further study of these genes in additional populations as well as exploration of the effect of TLR activation on carcinogenesis is needed to determine the role of TLR SNPs in prostate cancer risk.