Multiple Independent Genetic Variants in the 8q24 Region Are Associated with Prostate Cancer Risk

Prostate cancer is a common disease with an underlying genetic component (1-3). Genetic linkage studies have identified numerous putative loci that could harbor prostate cancer susceptibility genes (4, 5) and variant alleles in several candidate genes have been reported to be associated with prostate cancer risk (6), but results have proven difficult to replicate (7). Recently, a genome-wide linkage scan of prostate cancer families in Iceland identified the 8q24 region as a prostate cancer susceptibility locus (8). Multiple alleles in the region, especially the -8 allele of microsatellite marker DG8S737 and the A allele of the single nucleotide polymorphism (SNP) rs1447295, were associated with increased prostate cancer risk in three Caucasian case-control study populations from Iceland, Sweden, and Chicago [overall odds ratio (OR), 1.62, P = 2.7 × 10^-11 for DG8S7373 -8 allele; OR, 1.51, P = 1.7 × 10^-11 for rs1447295 A allele]. The DG8S737 marker was also associated with an increased relative risk in a sample of African Americans from Michigan (OR, 1.60, P = 2.2 × 10^-3). Subsequently, an admixture scan of African Americans confirmed the importance of the 8q24 region in modulating prostate cancer risk (9). The latter study also replicated results for the rs1447295 A allele, but suggested that the association of DG8S737 with risk of prostate cancer might simply reflect the overall difference in allele frequencies between African Americans and Caucasians (10).

The identification of the 8q24 region is notable for successful replication of several markers in studies using distinct populations (11-19). Of note, the original linkage result arose from the exploration of a suggestive linkage signal (maximum lod score of 2.11 at D8S529; ref. 8). The two original markers, DG8S737 and rs1447295, have been studied most widely (20), but several other SNPs in the 8q24 region have also recently been associated with prostate cancer risk. Fine mapping of the 8q24 region with SNPs from genome-wide association studies (12) and follow-up of the admixture study by Haiman et al. (13) have suggested that there are multiple regions within 8q24 that harbor variants altering prostate cancer risk (21). In the present population-based case-control study of Caucasian and African American men, we test the association between prostate cancer and multiple SNPs in three previously described regions of 8q24 and several new SNPs that extend the boundaries of the region. We also evaluate whether the association of SNPs in the 8q24 region varies among men with comparatively less or more aggressive disease or by age at diagnosis, first-degree family history of prostate cancer, or body mass index (BMI).

The study population consists of participants from two population-based case-control studies in Caucasian and African American residents of King County, Washington (studies I and II), which have been described previously (20, 22). Incident cases with histologically confirmed prostate cancer (International Classification of Diseases for Oncology code C61.9) were ascertained from the Seattle-Puget Sound Surveillance, Epidemiology, and End Results cancer registry. In study I, cases were diagnosed between January 1, 1993 and December 31, 1996 and were ages 40 to 64 years at diagnosis. In study II, cases were diagnosed between January 1, 2002 and December 31, 2005 and were ages 35 to 74 years at diagnosis. Overall, 2,244 eligible prostate cancer patients were identified and 1,754 (78.2%) were interviewed. The main reasons for nonresponse were patient refusal (13.9%), physician refusal to allow patient contact (2.1%), patients were too ill to participate (0.9%), or died before interview (1.4%). Blood samples yielding sufficient DNA for genotyping were drawn from 1,457 (83%) cases who completed the study interview.

A comparison group of controls without a self-reported history of prostate cancer, residing in King County, Washington, was identified using random-digit telephone dialing (23). Controls were frequency matched to cases by 5-year age groups and recruited evenly throughout each ascertainment period for cases. During the first step of random-digit dialing, complete household census information was obtained for 94% and 81% of the residential telephone numbers contacted for studies I and II, respectively. A total of 2,448 men were identified who met the eligibility criteria and 1,754 (71.7%) completed a study interview. The main reasons for nonparticipation included refusal (29.1%) or too ill to participate (1.4%). Blood samples were drawn and DNA was prepared from 1,358 (77.4%) interviewed controls using standard protocols (24).

Subjects in both studies completed in-person interviews conducted by trained male interviewers using standardized questionnaires. The questions pertained to the period up to the date of prostate cancer diagnosis for cases and a similar, randomly preassigned reference date for controls, which approximated the distribution of cases' diagnosis dates. Information was collected on family structure and cancer history, medical history, and social and demographic factors. All study procedures were approved by the Fred Hutchinson Cancer Research Center Institutional Review Board and written informed consent was obtained from all study participants before participation. Clinical information on cases, including Gleason score, tumor stage, and serum prostate-specific antigen (PSA) level at diagnosis, was obtained from the cancer registry.

Twenty-six SNPs were selected for genotyping in previously defined regions 1, 2, and 3 and the c-MYC gene. The majority of SNPs were selected based on published reports on 8q24 variants associated with cancer or were highly correlated (r² > 0.8) with SNPs that have been reported previously (http://cgems.cancer.gov/), aiming to minimize linkage disequilibrium (LD) between SNPs and focus on those with a minor allele frequency of at least 5%. SNPs in c-MYC were selected using the SeattleSNPs genome variation server (http://gvs.gs.washington.edu/GVS/index.jsp). The Applied Biosystems SNPlex Genotyping System was used to genotype SNPs in individual DNA samples and proprietary GeneMapper software was used for allele assignment (http://www.appliedbiosystems.com). Discrimination of the specific SNP allele was carried out with the Applied Biosystems 3730xl DNA Analyzer and is based on the presence of a unique sequence assigned to the original allele-specific oligonucleotide. Quality control included genotyping of 140 blind duplicate samples distributed across all genotyping batches. One SNP was monomorphic (rs1326634), one SNP was genotyped in only study II samples (rs2384921; agreement was 99% based on 82 blind duplicates), and two SNPs failed on the genotyping platform (rs2290840 and rs7818201). For the remaining 22 SNPs, there was 100% agreement between blinded samples. Each batch of DNA aliquots genotyped incorporated similar numbers of case and control samples, and laboratory personnel were blinded to the case-control status of samples.

Departure from Hardy-Weinberg equilibrium was assessed for each SNP separately, by race, in controls using the χ² test. Pairwise LD was estimated between SNPs, also by race, based on D′ and r² statistics calculated in controls (Table 1), using Haploview software version 4.0 (ref. 25; available from the Broad Institute at http://www.broad.mit.edu/mpg/haploview/). Individual region boundaries within 8q24 are those defined by Haiman et al. (13) as follows: region 1 from 128.54 to 128.62 Mb, region 2 from 128.14 to 128.28 Mb, and region 3 from 128.47 to 129.54 Mb.

Unconditional logistic regression models were used to estimate ORs and 95% confidence intervals (95% CI) to measure the association between individual SNP genotypes and prostate cancer risk (26), as implemented in SAS version 9.1.3. Polytomous regression models were used to generate ORs and 95% CIs for the association between SNP genotypes and cases stratified by disease aggressiveness (less versus more based on a composite variable including Gleason score, stage, and PSA) compared with controls. More aggressive cases were those with a Gleason score of ≥7 (4 + 3) or regional or distant stage or a PSA level ≥20 ng/mL at diagnosis. Both codominant (additive) and dominant genetic models were considered for each variant allele depending on the distribution of genotypes. Likelihood ratio–based test statistics were used to identify statistically significant associations between individual SNP genotypes and prostate cancer risk, by comparing the full model containing the SNP genotypes with the reduced model without the SNP, with a two-sided P value of 0.05, unadjusted for multiple comparisons, considered significant.

A permutation procedure was used to account for the effect of multiple testing. Pairs of case-control labels and ages were permuted to approximate the distribution of the age-adjusted P values under the null hypothesis. Ages and case-control labels were permuted together to preserve any relationship that may exist between age and case-control status and allow age-adjusted P values to be calculated for each permutation that are consistent with the original analysis. For each permutation, codominant and dominant models were fit for all SNPs and the minimum of the P values kept for each SNP. The P values were ordered to approximate the null distribution of the order statistics for the P values, that is, minimum P value, second smallest P value, etc. The original P values were also ordered and permutation P values were calculated by comparing the ordered P values to the null distribution for the appropriate order statistic. Permutation P values can be interpreted as the probability of observing a P value less than or equal to what was observed for the given order statistic under the null hypothesis of no association with disease for any of 23 SNPs. For example, the minimum P value was compared with the null distribution for the minimum P value and the corresponding permuted P value can be interpreted as the probability of the minimum P value being less than or equal to the observed minimum P value under the null hypothesis. The same is true for the second smallest P value, the third smallest P value, etc. The permutation approach to approximating the null distribution of the order statistics will be valid regardless of any correlation between the SNPs. A SNP was considered to be significantly associated with prostate cancer risk if the nominal P value and the permuted P value were both <0.05. In Results, we report unadjusted P values.

After consideration of SNP genotypes individually, all SNPs remaining significant after adjustment by permutation were included in a stepwise selection model using Akaike's Information Criterion to select the most parsimonious model (27). SNPs that were significant (nominal P < 0.05) after adjustment for each other were included in the final model (28). Variant alleles were tested under a dominant genetic model and both forward and backward selection models were compared, with equivalent results. A similar approach was used to test for SNP-SNP interactions, where all possible pairwise interactions of independently significant SNPs from the first stepwise procedure were included in a second stepwise selection. Haplotype frequencies and measures of association were estimated separately for African Americans and Caucasians using Hplus version 2.5 (ref. 29; available from http://qge.fhcrc.org/hplus/). All models were adjusted for age at reference date.

Other potential confounding factors, including first-degree family history of prostate cancer, prostate cancer screening history (digital rectal examination and PSA), and BMI (<25, 25-29.9, ≥30), were examined to see if such factors changed the risk estimates by ≥10%. To test whether such factors modified estimates of risk associated with SNP genotypes, ORs and 95% CIs were calculated for stratified models. If risk estimates differed across levels of these secondary factors, the interaction was then tested formally by including an interaction term in the model with the SNP genotype. The reduced model, with main effects only, was compared with the full model containing the interaction term using a likelihood statistic-based test.

The study population was predominantly Caucasian (Table 2). In comparison with controls, a higher proportion of cases were African American, had a first-degree family history of prostate cancer, and reported having a PSA screening test within the 5-year period before reference date. No significant differences were detected between cases and controls for education or BMI. The majority of cases had diagnostic PSA values of 4 to 9.9 ng/mL, localized stage disease, and Gleason scores of 5 or 6.

All 23 SNPs analyzed were in Hardy-Weinberg equilibrium in Caucasian controls. However, among African American controls, the distribution of genotypes for rs2384921 and rs6983561 was significantly different from that expected (P = 0.0002 and 0.02, respectively). For both SNPs, this may be related to the small number of samples from African Americans or due to chance. In addition, genotyping data were only available for men from one of the two studies for SNP rs2384921. Genotyping error is another reason for Hardy-Weinberg equilibrium violation but seems unlikely given the high level of agreement observed in blind duplicate samples.

Among the individual SNPs, 14 were significantly associated with risk of prostate cancer in Caucasians based on a codominant model, with nominal P values < 0.05 (Table 3). In African Americans, 5 SNPs were significantly associated with prostate cancer risk with nominal P values < 0.05. After adjusting for multiple comparisons, all nominally significant SNPs in Caucasians and African Americans remained significant at permuted P values < 0.05. Proceeding along the chromosome from the most centromeric to the most telomeric 8q24 region, 5 SNPs in region 2 (13) were significantly associated with prostate cancer risk in Caucasians. The most statistically significant finding in this region was for rs1016343 (Table 3), with an OR of 1.9 (95% CI, 1.3-2.7) for men with the rare homozygote genotype (P < 0.00005). Two other SNPs in region 2 (rs6983561 and rs16901966) were significantly associated with prostate cancer risk in both Caucasians and African Americans. In Caucasians, SNP rs6983561 was associated with a 1.8-fold increase in risk among men carrying any variant C allele (dominant model 95% CI, 1.4-2.4, P < 0.00005), and similar results (dominant model OR, 1.8) were observed for rs16901966 among men carrying any variant G allele (P = 0.0001). However, rs6983561 and rs16901966 are in nearly perfect LD, that is, D′ = 1 and r² = 0.98. In African Americans, the variant C allele in rs6983561 was associated with a significant 4-fold elevation in prostate cancer risk (95% CI, 2.0-8.0). Two significant associations for more centromeric SNPs were identified, including rs1456310 (dominant model OR, 0.82; 95% CI, 0.70-0.97 in Caucasians; OR, 2.4; 95% CI, 1.1-5.0 in African Americans) and rs979200 (dominant model OR, 0.79; 95% CI, 0.7-0.9 in Caucasians; no association with this variant was observed in African Americans), located 35 and 906 kb upstream from the boundary of region 2 as described previously by Haiman et al. (13). Interestingly, the rare T allele of rs1456310 was associated with a lower risk of prostate cancer in Caucasians (OR, 0.7) but a higher risk of disease in African Americans (OR, 2.4).

Continuing along 8q24 regions from the centromere toward the telomere, in 8q24 region 3, three SNPs were significantly associated with risk in Caucasians. The most significant SNP in this region, rs6983267 (P = 0.002), conferred a 30% decrease in risk of disease for carriers of the rare homozygote genotype (95% CI, 0.5-0.9). In African Americans, rs7000448 was the only significant SNP in region 3 and was associated with a significant elevation in risk of prostate cancer (OR, 3.6; 95% CI, 1.5-8.9 for men with the T allele).

Next, in 8q24 region 1, three SNPs were significantly associated with prostate cancer risk in Caucasians, including the polymorphism identified as the marker most strongly associated with prostate cancer in the initial studies by Amundadottir et al. (8) and Freedman et al. (9). This SNP, rs1447295, was associated with a 2.1-fold increase in risk of prostate cancer in Caucasian men homozygous for the rare A allele relative to men homozygous for the more common C allele (95% CI, 1.1-4.0). Two SNPs in strong LD with rs1447295, rs10090154 and rs7837688, produced similar positive associations in Caucasians. The rs1447295 SNP was not significantly associated with prostate cancer risk among African Americans.

Telomeric to region 1, we genotyped three SNPs, including two in c-MYC. SNP rs4645959 encodes a missense change in MYC protein (Asn²⁶Ser) and was not significantly associated with prostate cancer risk in either Caucasians or African Americans. However, rs3891248, located in the first intron, was significantly associated with a reduced risk of prostate cancer in Caucasians (OR, 0.55; 95% CI, 0.31-0.98 in men with the variant AA genotype relative to the common TT genotype; P = 0.04). No association with this latter SNP was observed in African Americans. Interestingly, the allele distribution for rs3891248 was substantially different in African Americans from that observed in Caucasians.

Single SNP association tests identified several variants for which a statistically significant relationship to prostate cancer risk was observed. To determine whether a variant was independently associated with disease, all SNPs that were individually significantly associated with risk (permuted P < 0.001) were entered into a single multivariate model. Using Akaike's Information Criterion to select the most parsimonious final model, this approach identified those SNPs that were independently associated with prostate cancer risk, taking into account the effects of other SNPs (Table 4). Stepwise forward selection with Akaike's Information Criterion was carried out separately for Caucasians and African Americans based on results shown in Table 3. Identical results were obtained from backward stepwise selection. For these analyses, results did not differ when highly correlated SNPs (r² > 0.8) were excluded from the initial model. In Caucasians, the initial model containing all 14 individually significant SNPs was reduced to a final model with 5 SNPs that were independently and significantly associated with prostate cancer risk. These SNPs include variants in each of the 8q24 subregions (rs10090154, rs1016343, rs6983561, and rs7837328 located in regions 1, 2, 2, and 3, respectively), plus rs979200, which is centromeric to region 2. In African Americans, of the 5 SNPs individually associated with prostate cancer risk in single SNP tests (Table 3), 3 remained significant in the final stepwise regression model, suggesting independent associations for each of these SNPs. SNP rs6983561 was independently associated with prostate cancer risk, as it was in Caucasians, and so were two additional SNPs (rs13254738 and rs7000448 in regions 2 and 1, respectively).

SNPs that were found to be independently associated with prostate cancer risk were included in haplotypes to examine the joint effect of variant alleles on risk in our population (Table 5). Interestingly, the haplotype expected to be associated with the highest relative risk of prostate cancer among Caucasians, that is, rs979200 G, rs1016343 T, rs6983561 C, rs7837328 A, and rs10090154 T alleles, is not estimated to occur in Caucasian controls, although the frequency in Caucasian cases is expected to be 0.07%. In Caucasians, haplotype GCAAT is significantly associated with disease risk with an OR of 2.2 (95% CI, 1.3-3.9) and carries two alleles associated with increased risk, the G allele of rs979200 and the A allele of rs7837328. These two alleles identify the three haplotypes in Caucasians with the largest point estimates of disease risk. Among African Americans, the AAC haplotype with all low-risk alleles is the most common (10% cases and 22.6% controls). When men with the high-risk CCT haplotype (40.9% cases and 21.0% controls) are compared with men with the AAC haplotype, there was a statistically significant 4.4-fold increase in prostate cancer risk (95% CI, 1.8-10.5).

Evaluation of the five independently significant 8q24 SNPs in Caucasians and three in African Americans with prostate cancer risk within categories defined by first-degree family history of prostate cancer, prostate cancer aggressiveness, Gleason score, stage of disease, or BMI in polytomous models did not reveal any appreciable heterogeneity of risk estimates (data not shown). Among Caucasian cases, only rs16902124 was significantly associated with age at diagnosis. Patients having any A allele were diagnosed, on average, 1.7 years younger than GG homozygous cases (58.4 versus 60.1 years, respectively; P = 0.02). Among African Americans, the largest difference in age at diagnosis was observed for rs7837328, where, on average, cases with any A allele were diagnosed at age 58.8 years versus GG homozygous cases who were diagnosed at age 64.0 years (P = 0.005). The age of diagnosis for African Americans also differed significantly across genotypes of SNPs rs1456310, rs10505476, and rs1447295, with carriers of variant alleles being diagnosed at earlier ages (data not shown).

Multiple polymorphisms across the 8q24 region have been significantly associated with prostate cancer risk in both case-control association and genetic linkage studies. This study confirms the previous findings for several 8q24 SNPs reported in populations of western European descent and extends the centromeric boundary of the 8q24 region by identifying variant alleles in two SNPs almost 1 Mb centromeric of region 2 (rs979200 and rs1456310) that are significantly associated with risk of prostate cancer. Consistent with previously published studies (8, 13, 14, 16, 18), our results confirm an increase in prostate cancer risk for the rs1447295 SNP (OR, 2.1) and two SNPs in strong LD with that SNP (rs10090154 and rs7837688). We also previously reported strong LD between rs1447295 and the microsatellite DG8S7373 in the 8q24 region (15). In the current analyses, the largest significant effect we observed in Caucasians was for rs7837688, with a >2-fold increase in the relative risk of prostate cancer (OR, 2.2; 95% CI, 1.1-4.1) when comparing variant TT homozygotes to common GG homozygotes. Significant findings in Caucasians for rs6983267 (13, 18) and rs6983561 (13) were also replicated, but not those for rs13257438 or rs7000448. The allele frequencies for rs13254738 and rs7000448 in our data were not significantly different from those reported previously (ref. 13; 0.33 cases and 0.31 controls and 0.36 cases and 0.36 controls for rs13254738 and rs7000448, respectively). We found no evidence to support the notion that the 8q24 SNPs are preferentially associated with more aggressive prostate cancer as defined by Gleason score, stage, and PSA at diagnosis. Neither was there a difference in the associations detected across strata of family history of prostate cancer, BMI, or smoking history.

From a stepwise multivariate model, five SNPs were retained with each being independently associated with prostate cancer risk in Caucasians (Table 4). These SNPs are associated with as much as a 1.6-fold increase in the relative risk of disease (95% CI, 1.1-2.1 for rs6983561). Our results are consistent with the inclusion of rs6983561 and rs10090154 in the multivariate model of Haiman et al. (13), but not their inclusion of rs7000448 or rs13254738. Interestingly, these two latter SNPs are independently associated with risk in our multivariate SNP model for African Americans (Table 4).

Several studies, including recently available data from the Cancer Genetic Markers of Susceptibility Study, have reported the strength of the associations for rs1447295 and rs6983267 with risk of disease (14, 17, 18). In our data, both of these SNPs were significantly associated with risk of prostate cancer in single SNP tests, but neither was included in our multivariate models as other SNPs in strong LD with these two were more significant in our data set. However, the earlier data are consistent with results presented here, as rs10090154 and rs7837328 were both independently associated with risk in Caucasians, rs10090154 is in strong LD with rs1447295, and rs7837328 is in strong LD with rs6983267 (Table 1). Thus, they serve as proxies or “tags” for one another. Indeed, when rs1447295 and rs6983267 are substituted in the multivariate SNP model of Table 4, results are similar. When haplotypes are constructed from the SNPs found to be independently associated with risk, prostate cancer risk increases as the number of alleles associated with higher risk increases, e.g., compare GTAAC and GCAAT, each with 3 risk alleles, with ACAGC, GCAGC, and ACAAC, each with ≤1 risk allele. Combinations of independent SNPs more than double the relative risk estimates of disease (95% CI, 1.26-3.87) in men carrying the GCAAT haplotype compared with the most common haplotype (Table 5).

The relative risk estimates in African Americans were of greater magnitude than those observed in Caucasians, with several significant ORs above 3.5 (rs13254738, rs6983561, and rs7000448), although the limited sample size implies the need for caution in interpreting these data. Five SNPs were significantly associated with prostate cancer risk in models with single SNPs, with rs6983561 and rs16901966 potentially reflecting the same association signal, as they are in modest LD (r² = 0.51). This is similar to the results of Haiman et al. (13), who reported significant associations for rs13254738, rs6983561, and rs7000448 for African Americans from Michigan. In contrast, no significant associations were identified for rs10090154 in these data, but allele frequencies differed between study populations (23.3% cases and 16.7% controls in our study compared with 12.3% and 8.9% for cases and controls in Haiman et al.). No significant association was detected for rs1447295, which is consistent with several earlier studies (8, 9, 12, 19). At an α level of 0.05 and the number of independent tests conducted in Caucasians, at least one false-positive result would be expected. The probability of finding as many independent significant results as were observed in Caucasians is <1 × 10^-13, that is, 11 SNPs with r² < 0.8 had significant nominal P values < 0.05. Among African Americans, the probability of observing as many independent significant results as were observed, that is, four SNPs with r² < 0.8, is 1 × 10^-4. These probabilities are calculated based on the assumption that the null hypothesis is true. As the 8q24 region has been identified previously as a region of interest with multiple variants associated with risk in other studies, this assumption may not be valid for all SNPs tested here. As an additional measure to correct for multiple comparisons, P values were obtained from permutation testing. All of the SNPs initially identified as significant at the nominal P < 0.05 level in both Caucasians and African Americans remained significant based on the permuted P values that account for multiple comparisons.

Findings from this study extend the boundaries of the 8q24 region where significant associations with prostate cancer risk have been found. SNPs rs979200 and rs1456310, both significantly associated with disease, are located 88 and 40 kb centromeric to rs10086908, as reported in Zheng et al. (18). These three SNPs are located over 120 kb distant from 8q24 region 2 (128.14-128.28 Mb), the most centromeric of the three 8q24 regions identified so far, and are not in LD with known SNPs in other regions. These SNPs may identify what Zheng et al. suggest may be a fourth 8q24 region in Caucasians. At the other margin of 8q24, rs3891248 (Table 3) identifies a potentially new telomeric boundary, which is also notable for being located in the first intron of c-MYC (IVS1-355). Although several recent studies have examined SNPs in or near c-MYC (12, 13, 17, 18), none has reported significant associations. However, these studies did not report investigating rs3891248 and this SNP does not appear to be in significant LD with any known SNP. This result is interesting as the c-MYC oncogene has been hypothesized as playing a role in prostate cancer based on its amplification in prostate tumors (30) and its location in the 8q region most frequently gained in the genome of prostate tumors (31).

Our study has several strengths and limitations. The population-based study design, the sample size, the availability of information on potential confounders and effect modifiers, and the clinical information on prostate cancer cases are strengths. Limitations include the modest number of African Americans, which reflects demographics of the Seattle-Puget Sound area. Because the genotypes of SNPs in the 8q24 region seem unlikely to be related to study participation, the potential for selection bias is reduced. In addition, although not all interviewed cases and controls provided blood samples for genotyping, no significant differences in demographic (cases and controls) or clinical features of disease (cases) were observed between men who provided blood samples and those who did not.

In summary, results reported here and elsewhere confirm that multiple genetic polymorphisms in the 8q24 region are associated with prostate cancer susceptibility. Multiple independent subregions have been discerned and multiple genetic markers in these regions apparently contribute independently to disease risk. This is not a gene-rich region of the genome, although several potential transcripts and genes of unknown function are present. Current interest in the 8q24 region may motivate further characterization of loci, such as PVT1, which is telomeric to c-MYC and which has been recently implicated in breast cancer progression (32). In the meantime, the underlying biological mechanism(s) driving the positive associations for SNPs in this region will be challenging to uncover. Others have speculated that these multiple independent variants are likely to underlie a common biological mechanism and may influence the regulation of a local gene (13, 18). The presence of c-MYC in the distant genomic landscape presents a tempting yet challenging candidate gene. Aside from the significant association of rs3891248 with prostate cancer risk, the distance between the most telomeric SNP reported previously as significantly associated with prostate cancer risk and c-MYC is >200 kb. It is not beyond reason to speculate that polymorphisms this far upstream may influence expression of c-MYC. First, the largest segments of DNA that have been engineered upstream of a c-MYC reporter transgene fail to recapitulate the full profile of c-MYC expression in vivo (33). Secondly, genes with long-range enhancers located as far upstream as 8q24 is from c-MYC have been reported previously (34-36). Our finding of a significant association between an intronic SNP in c-MYC and prostate cancer is intriguing and suggests that the boundaries of the region may extend farther than thought previously. The combined efforts of the prostate cancer research community will be required to fully understand the role of the 8q24 region to development of this disease and the identification of the biological mechanism(s) involved.

No potential conflicts of interest were disclosed.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

We thank the men who participated in these studies and whose help made this work possible, Dr. Robert Eisenman for helpful discussions about c-MYC, Dr. Meredith Yeager for help with SNP selection in 8q24, and Drs. Bo Johanneson and Brandon Pierce for helpful discussions.