Interactions of Sequence Variants in Interleukin-1 Receptor–Associated Kinase4 and the Toll-Like Receptor 6-1-10 Gene Cluster Increase Prostate Cancer Risk

Prostate cancer is the most common cancer and second leading cause of death in the Unites States (1) and in Western Europe (2). Many factors contribute to prostate cancer, and genetic susceptibility is among the most important of these (3). Recently, chronic inflammation has been suggested as a causal component in the development of a variety of malignancies, including cancer of the stomach, liver, and prostate (4-7). The cause of inflammation in the prostate remains unclear, but in some cases may reflect a specific response to microbial pathogens. Two of the three reported prostate cancer susceptibility genes (RNASEL and MSR1) are involved in innate immunity, supporting the hypothesis that variations in genes regulating innate immunity may modify both host response to infection and susceptibility to prostate cancer (8, 9).

Recently, we found that sequence variants in four Toll-like receptors (TLR) genes, the TLR6-TLR1-TLR10 cluster, are associated with prostate cancer risk (10). Evidence for association between TLR6-TLR1-TLR10 sequence variants and prostate cancer risk is particularly well supported, as 11 of the 17 variants in this gene cluster are highly significant (10). These findings suggest that innate immunity, particularly the TLR, influences prostate cancer development. The TLR family, recognizing a variety of ligands including pathogens, is involved in pathogen-mediated inflammation (11). The engagement of ligands on these receptors initiates a downstream signaling cascade, leading to the transcription of target genes, including inflammatory cytokines (tumor necrosis factor-α, interleukin-12) and costimulatory molecules.

Interleukin-1 receptor–associated kinases (IRAK) are essential components in the Toll/interleukin-1 receptor (TIR) signaling pathway. Until now, four IRAKs have been identified in mammals. Whereas IRAK1 and IRAK4 are expressed in all tissues, IRAK2 and IRAK-M have a narrower cellular distribution (12-15). A large body of biological evidence suggests that IRAKs play important roles in the TIR signaling pathway. IRAK4 knockout mice are completely nonresponsive to all TIR signaling (16). In contrast, when IRAK1 knockout mice were stimulated with IL-1 and lipopolysaccharide, the response was reduced but was not completely abolished (17, 18), suggesting that IRAK1 is involved in fine-tuning the TIR signaling pathway. Abnormalities in these two kinase genes are likely to be responsible for impaired TIR signaling, resulting in increased innate immunity responses and increased normal inflammatory responses; this, in turn, may contribute to the onset of various inflammatory diseases such as prostate cancer.

Considering the critical role of IRAKs in the TLR signaling pathway, we hypothesized that sequence variants in the IRAK1 and IRAK4 genes are associated with prostate cancer risk. Moreover, considering the intertwined roles of the TLR and IRAK genes in this signaling pathway and the significant association between sequence variants in the TLR6-TLR1-TLR10 gene cluster and prostate cancer risk, we further hypothesized a priori that interactions between variants in TLR6-TLR1-TLR10 gene cluster and these two IRAK genes increase prostate cancer risk.

To test these hypotheses, we systematically evaluated two IRAK genes, IRAK1 and IRAK4, in a large and well-characterized prostate cancer case and control population in Sweden. This is the first published report on the potential interaction between these genes in any disease.

The study base for the first phase of Cancer Prostate in Sweden-1 has been described in detail elsewhere (19). Briefly, Cancer Prostate in Sweden-1 is a large-scale, population-based case-control study in Sweden. Prostate cancer patients were identified and recruited from four of six regional cancer registries in Sweden. The inclusion criteria for case subjects in Cancer Prostate in Sweden-1 were pathology- or cytology-verified adenocarcinoma of the prostate, diagnosed between July 1, 2001, and September 30, 2002. Control subjects were randomly selected from the continuously updated Swedish Population Register and frequency-matched according to age (within 5 years) and geographic origin of the case subjects. In total, 1,444 case and 866 control subjects were recruited. Among all participants, DNA samples and questionnaires were available for 1,383 case and 780 control subjects, representing an 83% and 52% participation rate among all eligible case and control subjects, respectively.

Clinical information, such as tumor-node-metastasis stage (20), Gleason grade, and prostate-specific antigen levels at diagnosis were available from either cancer registries or the National Prostate Cancer Registry for 94% of the case subjects. The clinical characteristics of the study subjects are presented in Table 1. The case subjects were classified as having advanced disease if they met any of the following criteria: T_3/4, N₊, M₊, grade 3, Gleason score sum of 8 to 10, or a prostate-specific antigen level of >50 ng/mL. All other case subjects were classified as having localized disease.

The study received institutional approval at Umeå University and Karolinska Institutet in Sweden, and Wake Forest University in United States. All participants provided written informed consent.

Genotyping was done using the MassARRAY system (Sequenom, Inc., San Diego, CA). For the MassARRAY assay, PCR and extension primers for sequence variants were designed using SpectroDesigner software (Sequenom). The primer information is available at the corresponding author's web site (www.wfubmc.edu/genomics). PCR and extension reactions were done according to the instructions of the manufacturer, and extension product sizes were determined by mass spectrometry.

The IRAK1 and IRAK4 genes are mapped to Xq28 and 12q12, respectively. The length of transcripts for IRAK1 and IRAK4 are ∼9 and 29 kb. IRAK1 is composed of 14 exons, whereas the IRAK4 gene contains 12 exons. The single nucleotide polymorphism (SNP) information for IRAK1 was obtained from the web site http://innateimmunity.net/data/homology (October 2003). The SNP information for IRAK4 was obtained from two web sites, http://pga.gs.washington.edu (October 2003) and http://snpper.chip.org (21). We used two criteria to select SNPs from a large number of SNPs in these Web sites. (a) We selected SNPs with a minor allele frequency of at least 5% at a resolution of one SNP per kb across the genomic region of each gene, including 2.5 kb of the promoter, all exons, introns, and 3′ untranslated region. For the IRAK4 gene, a resolution of 1 SNP per 2 kb was used because of the large size of this gene. (b) We also selected all of the nonsynonymous SNPs in these genes, regardless of resolution.

The haplotypes of the seven SNPs in the IRAK4 gene were estimated using a computer program, PHASE 2.0 (www.stats.ox.ac.uk/mathgen/software.html) based on genotypes of the 780 controls. Haplotype blocks in this region were constructed using a web-based program Haploblockfinder (http://cgi.uc.edu/cgi-bin/kzhang/haploBlockFinder.cgi/). We used the threshold of minimal pair-wise D′ ≥ 0.8 to define haplotype blocks.

Tests of Hardy-Weinberg equilibrium and pairwise linkage disequilibrium for each of the SNPs in IRAK4 were carried out using an exact test (22). Empirical P values for the Hardy-Weinberg equilibrium test were based on 10,000 permutation tests.

For each SNP, allele frequency differences between cases and controls were tested using a χ² test with 1 degree of freedom. The genotype frequency differences were tested using a χ² test with 2 degrees of freedom. Odds ratios (OR) of prostate cancer for the variant-allele carriers versus wild-type allele carriers were estimated by logistic regression after adjusting for age. Association between the haplotypes and prostate cancer risk was done using a likelihood ratio test and a χ² test implemented in SAS/Genetics.

The combined (interaction) effect of the three genes in the TLR6-TLR1-TLR10 cluster and the four IRAK genes on prostate cancer risk was assessed using a logistic regression framework, with adjustment for age. Because of strong linkage disequilibrium among SNPs in the TLR6-1-10 gene cluster, we were able to use only one SNP, TLR1 −6399C/T, which was in the middle of several of the most significant SNPs in this gene cluster, to represent this gene cluster and to ultimately to decrease the number of tests. Based on our previous study of the TLR gene cluster (10), we defined a priori CC/CT as the risk genotypes for TLR1 (TLR positive), and TT as the nonrisk genotype (TLR negative). For IRAK1/4, we assumed that rare allele genotypes are the high-risk genotypes (IRAK positive) and the homozygous common allele is the low-risk genotype (IRAK negative). Consequently, we classified each individual based on their combined TLR and IRAK genotypes.

Men who carried presumed low-risk genotypes of both genes (−,−) were used as the reference group. The remaining individuals are grouped into three test groups (dummy variables). Men who were TLR negative and IRAK positive comprised group 1 (−,+), TLR positive and IRAK negative group 2 (+,−), and TLR positive and IRAK positive group 3 (+,+). We tested for both additive and multiplicative interaction effects (23) using three dummy variables (groups) in a logistic regression model. In brief, the expected effect of (+,+) considering an additive model is equal to the sum of the effects of (−,+) and (+,−); and the expected effect of (+,+) considering a multiplicative model is equal to the multiplicative product of the effects of (−,+) and (+,−). Additive or multiplicative interaction effects are suggested to be present if the observed effect of (+,+) is larger than the expected value considering additive or multiplicative models. All statistical tests were done using SAS/computer software (version 9.1, SAS Institute, Cary, NC).

Based on the two criteria described in Materials and Methods, we selected 16 SNPs in the IRAK1 gene, including all five of the reported nonsynonymous changes: S532L, C203S, F196S, R194H, and T113I. We also selected 16 SNPs in the IRAK4 gene, including all three of the reported nonsynonymous changes: S98R, H390R, and A428T. We genotyped these 32 SNPs among 96 randomly selected controls. Nineteen SNPs were monomorphic in our population, with 11 of these in IRAK1 and eight in IRAK4, and thus were excluded from additional study. We also excluded other two SNPs for which PCR results remained inconsistent even after repeated attempts at optimization. Haplotypes of IRAK1 can be directly evaluated because this gene resides on the X-chromosome. For the seven SNPs we examined in the IRAK4 gene, we identified two haplotype blocks: the first six SNPs were contained within one block and only the last SNP, 20791A/G, was located in the second block.

We genotyped 11 SNPs, including four in IRAK1 and seven in IRAK4 in all 1,383 and 780 controls whose DNA samples were available at the time of this study. All seven of the SNPs in IRAK4 were in Hardy-Weinberg equilibrium among case and control groups, respectively (all P > 0.05).

The distribution of genotype frequencies among cases and controls are presented in Table 2. No statistically significant differences in allele or genotype frequencies between cases and controls were observed for any of these SNPs.

For the IRAK1 gene at Xq28, 10 haplotypes were observed, four of which are common (frequency > 0.01; Table 3). However, no significant differences in haplotype frequencies between cases and controls were observed. For the IRAK4 gene at 12q12, six common haplotypes were inferred in this population (Table 3). Again, we did not observe significant differences in the haplotype frequencies between cases and controls.

When we examined interactions between the TLR gene cluster and IRAK1/4 on prostate cancer risk, we found a multiplicative interaction between a SNP in IRAK4 (7987G/C) and TLR1. Men who were TLR positive and IRAK positive (+,+) were ∼10 times more likely to have prostate cancer when compared with the reference group [−,−; OR, 9.68; P = 0.03; 95% confidence interval (95% CI), 1.27-73.96; Fig. 1B]. The excess risk for men in this group (group 3) was larger than the product of the risk for group 1 (OR, 1.47) and 2 (OR, 1.33), suggesting a multiplicative interaction (Table 4). However, this risk estimate for this joint effect has wide confidence intervals due to small cell count.

We also found an expected additive effect between a TLR1 SNP and an IRAK4 SNP (652T/C). Men that were TLR positive and IRAK positive (+,+) had a significant 59% increased prostate cancer risk when compared with the reference group (−,−; OR, 1.59; P = 0.02; 95% CI, 1.06-2.40; Table 4; Fig. 1A). In contrast, we found no significant excess of risk for the joint effect among men in group 1 (−,+; OR, 1.24; P = 0.11; 95% CI, 0.95-1.63). The increased risk of 35% (OR, 1.35; P < 0.01; 95% CI, 1.08-1.69) among men in group 2 (+,−) was statistically significant. The risk in the TLR-positive and IRAK-positive group (+,+) is equal to the sum of the excess risks in groups 1 (−,+) and 2 (+,−), suggesting an expected additive effect considering the additive model.

The hypothesis that gene-to-gene interactions play a role in susceptibility to common human diseases is not new (24). Multiple modifier genes, not just an individual gene, are considered responsible for the onset of many common diseases, a theory supported by the ubiquitous biomolecular interactions among gene transcripts and biochemical pathways (25). Functional studies have defined interactions between TLRs, adaptors, and kinases (11). We observed a gene-to-gene interaction between sequence variants in IRAK4 and TLR1; when combined with the risk genotype at TLR6-1-10, one SNP in IRAK4 conferred a multiplicative risk of prostate cancer. Our study is the first evaluation of gene-to-gene interaction effects among these key genes in the TIR signaling pathway in relation to prostate cancer. These results strengthen the hypothesis that inflammation is etiologically relevant in prostate cancer and that several genes in the TIR signaling pathway, instead of a single gene, influence this process.

Four key features of this study that strengthen our findings are the comprehensive evaluation of SNPs in kinase and adaptor genes, the use of a relatively homogeneous study population, a large sample size, as well as a well-defined a priori hypothesis. The comprehensive evaluation of 11 SNPs in IRAK1 and IRAK4 allows us to capture most of the common genetic variations in these two genes. The relatively homogeneous study population decreases the degree of locus and allelic heterogeneity and, therefore, increases our ability to detect genetic association and minimizes the potential for population stratification that can occur in case control study designs. The large sample size provides greater confidence in the validity of all results except for the multiplicative interaction effect. Finally, the a priori hypotheses of this study are well supported from functional studies.

However, the positive interaction finding of our study should be interpreted with caution. Gene-to-gene interaction studies generally suffer from several limitations, including inflated type I error due to multiple tests and small cell counts as outcomes of many potential genotype combinations. For example, even in our study where we considerably decreased the number of potential tests by fixing on one variant that was implicated from our previous single-gene study (TLR1) and only varied the genotyped variants in the IRAK genes, there were still 33 tests for interaction effects. A further complication of multiple tests in genetic studies is the difficulty of accurately correcting for multiple tests. For this reason, we were not able to adjust for multiple tests in this study, and our results should be interpreted within this context. Our study also suffered from small cell counts in several tests, even with the large overall sample size. The small cell counts could lead to chance significance or an unstable risk estimate. For example, our major finding of a 9-fold increased risk for the interaction was supported by only 14 cases and 1 control in the cells that were both TLR positive and IRAK positive (+,+). Although caution should be exercised when interpreting this statistically significant result, we cannot exclude the possibility that this statistical interaction reflects an underlying biological interaction between these two SNPs. Our results warrant confirmation in even larger studies.

The interaction and additive effects between the IRAK4 SNPs (7987G/C and 652T/C) and the TLR1 SNP −6399C/T may be a direct result of biological interactions between these specific SNPs, or, alternatively, may indirectly reflect other SNPs that are in linkage disequilibrium with these genotyped SNPs. TLR1 −6399C/T is located in the promoter region and may potentially influence transcription of the TLR1 gene. Those two IRAK4 SNPs reside in introns and the nucleotide substitutions in those sites may alter gene splicing. Abnormal expression of TLR1 and IRAK4 directly or indirectly related to these SNPs, may lead to altered responses in the TIR signaling pathway. This, in turn, may affect inflammatory responses that have been implicated in the development of prostate cancer (7). Considering the intriguing genetic findings and the biologically plausible relationship between these genes, replication in other large studies are warranted.

One weakness of our study is the unbalanced number of cases and controls due to the significantly low ratio of controls to cases (0.56). As described in Materials and Methods, this is a population-based and frequency-matched (age and geographic) case-control study. The smaller number of controls (n = 780), compared with cases (n = 1,383), reflects a lower response rate among controls (52%). One of the potential problems associated with a low response rate among controls could be a selection bias where control subjects who have a family history of prostate cancer are more likely to participate in the study. This potential bias, together with the decreased statistical power associated with a smaller number of control subjects, may bias the results toward the null. Although this poses some concern, we feel it does not alter the interpretation of the current study because associations for interaction effect have been observed.

In summary, our results support the hypothesis that interaction effects between sequence variants of IRAK4 and TLR1 contribute to the development of prostate cancer. This study, together with our previous observations of association with TLR1-6-10 and TLR4 sequence variants, provides further evidence that inflammation plays an important role in the initiation and progression of prostate cancer.