Prostate cancer (PC) is the most common malignancy observed in men. It is evident that genetic factors play some important roles in PC etiology. Recently, genome-wide association studies in diverse ethnic groups have identified more than 40 germline variants of various genes or chromosomal loci that are significantly associated with PC susceptibility, including multiple 8q24 loci, prostate-specific genes, metabolic and hormone-related genes, and many regions where no coding gene is annotated. However, there are only a few variants or genes for which biological significance or functions have been elucidated so far. The greatest challenge related to genome-wide association studies loci in prostate genomics is to understand the functional consequences of these PC-associated loci and their involvement in PC biology and carcinogenesis. There have been attempts to determine PC risk estimations by combining multiple PC-associated variants for clinical tests, and these can identify a very minor population with high risk of PC. However, they cannot distinguish risk of aggressive PC from that of non-aggressive PC. Further identification of PC-susceptibility loci in larger genome-wide association studies cohorts and biological insights gained from such functional analyses have the potential to translate into clinical benefits, including the development of reliable biomarkers, risk estimation, and effective strategies for screening and prevention of PC. (Cancer Sci 2012; 103: 607–613)
The most common malignancy observed in men, PC is the second leading cause of cancer-related deaths in Western countries.[1, 2] The prevalence of PC differs markedly among various ethnic groups. For example, African-Americans have the highest prevalence (137 cases per 100 000 persons), while the lowest prevalence is observed in Asian populations (<10 cases per 100 000 persons), suggesting that genetic background might contribute to differences in PC susceptibility. Although the precise mechanisms of prostate carcinogenesis have not been not fully elucidated, it is evident that genetic factors play some important roles in PC etiology. A positive family history of PC has been recognized as one of the most important risk factors for PC, as well as African descent and old age. Studies involving twins have indicated that the contribution of genetic factors to the development of PC is greater than that to the development of other types of common human tumors. However, although Asian populations have the lowest incidence of PC worldwide, its incidence is rapidly increasing in most Asian countries, including Japan. This rise in PC incidence is probably attributable to lifestyle changes including a shift towards a westernized lifestyle and diet, patients’ increased access to the PSA test, and a rapid expansion in the aging population. In fact, the PC risk in Japanese immigrants in Hawaii or Los Angeles has been observed to be significantly increased. Since various genetic, environmental, hormonal, and social factors are involved with prostate carcinogenesis, its etiology is quite heterogeneous and unclear.
Over the past 20 years, genetic research has been conducted to clarify PC genomics and to identify genes responsible for PC susceptibility. Linkage analysis on several hereditary PC families identified some putative candidate genes for hereditary PC, but this approach was not successful for elucidating PC genomics, as other studies have not replicated the findings. There is a growing body of evidence indicating that the genetic basis of PC is much more complicated than initially believed. Recent GWAS have identified more than 40 SNP on various genes or chromosomal loci that are significantly associated with PC susceptibility. This number is much higher than that found for other types of common cancers – a finding that is a distinct characteristic in PC genomics. However, most of the identified SNP are merely markers associated with PC, with only a few variants clarified as affecting prostate carcinogenesis. In order to more clearly understand PC genomics and the underlying biology, functional analysis of these loci or genes identified by GWAS analysis is required, and more loci or genes associated with PC susceptibility must be identified in a wide array of racial/ethnic populations. In this review, we summarize the aspects of PC genomics obtained mainly by the GWAS approach and discuss their underlying biology or potential applications to PC clinics.
Familial PC Segregation and Linkage Analysis for Familial PC
There is strong family clustering of PC in Caucasians, and approximately 10–15% of men with PC have at least one relative who has also been affected by PC. Several studies have consistently shown that a positive family history of PC conveys an OR of ~2.5. Familial segregation of PC is observed with similar OR in different racial/ethnic groups, although the frequency of familial segregation varies among different groups. In Japan and other Asian countries, the morbidity of PC is quite low compared to that in Western countries. Given the low morbidity in the past decades, familial segregation of PC has been observed at a low frequency in Asia. However, the recent expansion of PC incidence in Asia will undoubtedly increase the incidence of familial segregation in the future. The search for PC-susceptibility genes by linkage analysis on familial or hereditary PC offered early hope that the identification of PC-susceptibility genes would be as easily determined as genes for breast cancer and colon cancer susceptibility. Genes including the RNase L (RNASEL) gene in HPC1 on chromosome 1q23-25, the elaC homolog 2 (ELAC2) gene in HPC2 on chromosome 17p, and the macrophage scavenger receptor 1 (MSR1) gene have been identified as PC-susceptibility genes by linkage analyses and mutational search.[7-9] However, most gene linkages were not replicated across studies. The most promising candidate identified so far is BRCA2, which is associated with a 20-fold increased risk for breast cancer relative to the general population, and the evidence points to a more important role of this gene in PC at a younger age. However, this genetic linkage may explain only a small fraction of familial PC because germline mutations of BRCA2 are quite rare in PC patients.
Common Variants of Candidate Genes
The most likely candidate genes for PC susceptibility are those involved in the metabolism of testosterone and pathway of androgens, because the growth of prostate epithelium and PC cells is mostly dependent on testosterone and AR pathway. A number of genetic variants of components in the AR pathway have been examined in case-control studies, including the following: 5α-steroid reductase 2, encoding an enzyme that converts testosterone to a more potent androgen dihydrotestosterone; cytochrome P-450 17alpha (CYP17), encoding the enzyme that regulates critical steps in testosterone biosynthesis; and the 3 beta-hydroxysteroid dehydrogenase (HSD3B) family, encoding enzymes involved in dihydrotestosterone metabolism. Furthermore, many case-control studies evaluated genes involved in the metabolism of environmental carcinogens, such as glutathione s-transferase theta-1 (GSTT1) and n-acetyltransferase 2 (NAT2). Vitamin D metabolites have an anti-proliferative and pro-differentiating effect on PC, and low levels of sunlight exposure and vitamin D, which is synthesized from sunlight exposure, are indicated to be associated with an increased risk of PC. Numerous studies have investigated the genetic association between variants of the vitamin D receptor and PC risk. Although these genes have strong biological support, consistent and replicable PC risks associated with these candidate genes have not been achieved.[12, 14]
GWAS Approach to Identify New Loci Associated with PC Susceptibility
The GWAS approach is typically based on a case-control design in which several hundred thousands SNP across the human genome are genotyped and used to scan for germline variants associated with disease risk. Such associations are consistent with the “common disease–common variant” hypothesis, which posits that genetic influences on susceptibility to common diseases such as PC are attributable to a number of common germline variants. For PC, GWAS have been remarkably successful in identifying more than 40 common genetic variants or loci,[16-29] which are summarized in Table 1. The Manhattan plot of a PC GWAS indicates many peaks of −log10 (P-value) across the genome, including extremely high multiple peaks (low P-value for the association) in chromosome 8q24 (Fig. 1). Subsequent replication and follow-up studies have identified further associated variants showing genome-wide significance, such as P-value <1 × 10−7 or 5 × 10−8 after adjustment for multiple testing. Recent meta-analyses combining several GWAS cohorts in multiple ethnic groups (more than 50 000 samples) identified additional PC-susceptibility genes or variants, and an ongoing larger meta-analysis across the world is expected to identify additional PC-susceptibility loci.
Table 1. Summary of loci that are associated with prostate cancer susceptibility
However, it remains unclear whether these genetic variants are associated with PC aggressiveness. Unlike other types of cancer, PC is a disease that develops at an older age, and many indolent PC cases in older individuals (≥80 years old) are subject to a watching strategy without any PC treatment. Autopsy studies have shown that the rate of latent PC in men may be about 20–30%. Hence, distinguishing between indolent PC and aggressive PC, which requires treatment, is one of the most important questions in PC research and clinics today. A recent meta-analysis showed that rs4054823 at 17p12 is associated with aggressive PC, but there is no annotated gene around this SNP and no explanation for its biological association with PC aggressiveness. Most PC variants are located in the introns and intergenic regions (Table 1), and there are few variants of which biological significance is shown to be associated with PC. Below we review several PC-susceptibility loci or genes of which the functions or biological significance in prostate carcinogenesis have been elucidated or speculated.
Chromosome 8q24 region
Many GWAS in several ethnic groups have reported strong and consistent associations of PC susceptibility with multiple genetic variants at chromosome 8q24.[16-19, 23, 24] This region contains various independent PC-susceptibility loci within a ~1-Mb segment, and some of them were found to be significantly associated with other types of cancer, including colorectal cancer,[33, 34] breast cancer, ovarian cancer, and bladder cancer. However, no gene has been annotated in this ~1-Mb region, and its biological significance in cancer remains unclear. There are at least five separate loci (Region 1–3 or Block 1–5) associated with PC susceptibility in this region (Fig. 2).
The c-MYC proto-oncogene is located at approximately 200-kb downstream, and recent studies have indicated that one of the loci at 8q24 (rs6983267 represents Region 1/Block 4) could be associated with the Wnt signal pathway in colorectal cancer and with c-MYC expression in several cancers.[39, 40]In vitro and in vivo experiments indicated that these risk loci at 8q24 contain possible enhancers that could regulate and interact with the c-MYC promoter.[41-43] However, these studies did not show that the associated SNP at 8q24 could affect c-MYC expression in colorectal cancer or PC, and their association with MYC expression in PC and other cancers is not conclusive.
A GWAS performed in a Japanese population indicated that two independent loci are strongly associated with PC: Region 1 and Region 2 (Fig. 2). The association of its centromeric region (Region 2: Chr8: 128.14–128.28 Mb) with PC is likely to be stronger in Japanese and Africans than in other populations.[18, 27, 45] We identified SNPs on Region 2 between rs1456315 and rs7463708 to be most significantly associated with PC susceptibility (P = 2.00 × 10−24, OR = 1.74). This region was transcribed as the ~10-kb intronless non-coding RNA, termed prostate cancer non-coding RNA1 (PRNCR1), and its expression was upregulated in PC cells and its precursor lesions. Interestingly, knockdown of PRNCR1 by siRNA attenuated the viability of PC cells and the transcriptional activity of AR, implicating its important roles in PC carcinogenesis probably through the AR pathway. Furthermore, a preliminary study reported that the variant on Region 2 was correlated with serum testosterone levels. Given that Region 2 is associated only with PC, but not with breast cancer or colorectal cancer, its functional consequence is different from those of other loci at the 8q24 region, and Region 2 is likely to be involved with the AR pathway specifically.
An SNP at 10q11 (rs10993994) was validated to be associated with GWAS analyses in several populations.[21, 22, 27] It is located at the 5′ region of the MSMB gene, encoding β-microseminoprotein. Fine mapping and functional analysis demonstrated that rs10993994 is strongly associated with PC risk, and that its risk allele has a major effect on the transcriptional activity of the proximal promoter of MSMB.[48, 49] MSMB is synthesized by epithelial cells in the prostate gland and is secreted into seminal plasma. It has been suggested to be a potential tumor suppressor, as MSMB expression progressively decreases during development of PC from early to late stages. Moreover, loss of MSMB expression is associated with PC recurrence or poor survival. Interestingly, fine mapping in this region detected a second loci located in the NCOA4 gene, which is known to enhance AR transcriptional activity in PC;[48, 49] however, its association with PC is low.
Several GWAS analyses showed consistent association of rs1512268 at 8p21 with PC susceptibility (P-value = 4.3 × 10−11, OR = 1.34).[26, 27] This SNP represents a 57.6-kb linkage disequilibrium block, in which the NKX3.1 gene is solely annotated.NKX3.1 is exclusively expressed in the prostate; quantitative RT-PCR analysis showed that its expression in the normal prostate is significantly lower in subjects with the haplotype carrying the risk allele than that in subjects with the non-risk haplotype. Fine mapping and EMSA/luciferase assays indicated that rs11781886, which is absolutely linked with marker SNP rs1512268 in the 5′-UTR of NKX3.1, is a functional variant and can regulate the proximal promoter activity of the NKX3.1 gene.NKX3.1 is an androgen-regulated homeobox gene that plays a key role in the regulation of growth and differentiation of prostate epithelium in the normal prostate, and it is involved in generating a stem cell population that functions during prostate regeneration. With regard to prostate carcinogenesis, NKX3.1 maps to a region that is frequently subject to loss of heterozygosity in human PC,[54, 55] and it has been recognized to function as a tumor suppressor of prostate. However, germline or somatic mutations of NKX3.1 are not observed frequently, and common variants around NKX3.1 are likely to reduce NKX3.1 expression or change its function in the prostate, leading to PC susceptibility.
A Japanese GWAS identified that rs339331 at 6q22 was significantly associated with PC susceptibility (P-value = 1.6 × 10−12, OR = 1.22), and this association was validated in an African-American cohort.[27, 57] This SNP represents a 200-kb linkage disequilibrium block including GPRC6A and RFX6 genes.RFX6 belongs to the RFX transcriptional factor family and is expressed almost exclusively in pancreatic islets. In contrast, GPRC6A encodes an orphan G-protein-coupled receptor that is highly expressed in the Leydig cells of the testis, and Gprc6a-null mice have been shown to exhibit male feminization and a metabolic syndrome with increased circulating levels of estradiol and reduced levels of testosterone. These hormones critically contribute to the initiation and progression of PC, and genetic variations at the GPRC6A/RFX6 locus may affect PC susceptibility by altering GPRC6A-mediated sex hormone production and metabolic pathways. Furthermore, the expression of GPRC6A was upregulated in PC and GPRC6A ligand-enhanced PC cell proliferation.Gprc6a deficiency in the transgenic adenocarcinoma of the mouse prostate model significantly retarded PC progression and improved survival of Gprc6a (−/−)/transgenic adenocarcinoma of the mouse prostate model mice.
Diabetes-associated genes: HNF1B, JAZF1, and THADA
It is noteworthy that GWAS on PC identified several loci or SNP that were validated to be inversely associated with 2DM. Epidemiological studies suggest that men with 2DM, compared to non-diabetic men, are less likely to develop PC. Several studies have determined that variants of the HNF1B gene at 17q12, which predispose haplotype-carrying subjects to 2DM, were associated with a decreased risk of PC.[21, 61] The gene HNF1B encodes a transcription factor that was previously known to be mutated in individuals with maturity-onset diabetes of the young type 5, and it likely plays a key role in the development and function of the pancreas and kidney. It regulates the expression of numerous genes in these tissues, but it is unknown whether it alters levels of the various metabolic and hormonal factors that may influence PC risk in diabetic men. Interestingly, this variant at HNF1B was also found to be associated with risk of endometrial cancer in women, which is also thought to be dependent on some metabolic and hormone factors, similar to PC in men. A GWAS showed variants of JAZF1 at 7p15 were significantly associated with 2DM and with a decreased risk of PC. Encoding a nuclear protein with three zinc fingers, JAZF1 functions as a transcriptional repressor and may be involved in lipid metabolism in the liver and adipocytes. Moreover, JAZF1 variants have been associated with human height, supporting a role for this gene in the regulation of growth and metabolism. Thus, it appears possible that genetic variations in JAZF1 and HNF1B may influence PC risk by altering the levels of hormones or growth-related factors previously suggested to be associated with 2DM. Variants of THADA at 2p21 were identified to be associated with 2DM as well as PC risk,[26, 27, 67] and THADA was identified as a genomic rearranged gene in thyroid adenomas. However, its functional association with 2DM and PC is completely unknown. Consistent with the epidemiological data, the risk alleles of 2DM in JAZF1 and HNF1B are protective for PC and vice versa. Diet or lifestyle is now proposed to be one of the critical factors of PC development, and in fact, PC incidence in Japan and other Asian countries is rapidly growing partly owing to the prevalence of Western lifestyles, including a high-fat and/or high-calorie diet. This interesting observation may illustrate a biological phenomenon that connects food metabolism with prostate carcinogenesis.
PSA Test and PSA Quantitative Trait Loci Analysis
Encoded by the kallikrein 3 gene, PSA is produced exclusively in prostate epithelium and is restrictively regulated by androgens. It is synthesized in healthy prostate tissues, benign prostate diseases and PC of all grades and stages. As such, it is not a specific marker for PC, but it is a prostate-specific marker. The serum PSA test is the most common screening method for PC, but serum PSA levels are affected by a number of factors such as prostate volume and age. In addition, PSA levels differ in among racial and ethnic groups. Even after adjusting for prostate volume, PSA levels are higher in men of African descent than in Caucasian men. As PSA screening has become more widely used, it has generated increasing controversy over the threshold level of PSA that should indicate the need for invasive prostate biopsy for PC diagnosis. In general, 10~ ng/mL PSA indicates positive PC. However, when the PSA level is 4–10 ng/mL in a “gray” zone, the positive rate of PC in a prostate biopsy is only 20–25%. Even men with <4 ng/mL PSA are found to have PC in 20% of biopsies, and there is a current debate about the cutoff value for PSA. Gudmundsson et al. reported results from a GWAS conducted on variations in serum PSA values. They identified six genome-wide significant loci associated with serum PSA levels. Of these, four loci were previously associated with PC risk (kallikrein 3 at 19q13, MSMB at 17q12, HNF1B at 17q12, and 15p15 locus near the telomerase reverse transcriptase [TERT] gene). New loci at 10q26 and 12q24 were associated with basal PSA levels but not with PC risk in an Icelandic population. It is unknown how these variants can affect PSA level. Genetic variants affecting basal PSA levels may influence the clinical recommendations of a prostate screening or biopsy, but the improvement from incorporating these variants is modest for the PSA test.
Risk Estimation by Multiple SNP
Various GWAS have identified more than 40 variants or loci significantly associated with PC risk, but each has limited use in the assessment of PC risk in an individual because each genetic marker confers a modest effect (OR: 1.1–1.8) (Table 1). To develop a clinical test using the risk-associated variants, Zhen et al. combined five variants for PC risk assessment plus a family history of PC and confirmed a cumulative and significant association with PC. However, several subsequent studies using ≥5 SNPs indicated that the AUC in the ROC curve is approximately 0.6,[73-75] and these SNP panels may have limited clinical utility. In general, the AUC of the PSA test is approximately 0.7. If the risk estimation model for PC screening is applied in addition to the PSA test, the AUC should be larger than for the PSA test. Otherwise, the risk estimation model should focus on individuals in the gray zone or with PSA levels in the normal range – individuals in whom the efficiency of PSA testing to detect PC is questionable. In contrast, the risk estimation model, combined with family history, using 14 SNP indicates the absolute risk for PC and can select a very small population (~1%) with a high risk of PC (40–50% absolute lifetime risk). This frequency and magnitude of PC risk are comparable to that for breast cancer risk prediction among individuals with a BRCA1 or BRCA2 mutation in the general population. Distribution of the OR of PC risk estimated by combining multiple SNPs indicates that a small percentage of men can have more than three times of the OR of PC risk (Fig. 3). It is significant to focus on men with such a high PC risk because risk factors other than PC family history and older age have not yet been established and frequency of PC family history is very rare, particularly in some ethnic groups such as the Japanese and other Asians. Men with a high PC risk may pursue preventive procedures, lifestyle intervention, or targeted chemoprevention such as finasteride and dutasteride (5α-steroid reductase inhibitors), which have been shown to reduce PC risk by more than 20%. Another issue with these risk estimation models is that they cannot distinguish risk of aggressive PC from that of non-aggressive PC, and therefore, may exacerbate the potential problems of overdiagnosis and overtreatment. The benefits of these risk estimation models for PC need further careful evaluation. The commercialized deCODE genetic test (deCODE genetics, Reykjavik, Iceland) for PC risk assessment uses 25 SNPs, and in Caucasians, the test can identify about 15% of men who have double the average PC risk and 5% of men who have triple the average risk. This test may be used to complement the standard clinical risk screening methods. However, this algorithm is only applicable to Caucasians, and construction of an ethnic-specific risk estimation model for PC using multiple SNPs is required.
Prostate cancer is a heterogeneous disease both in epidemiological and genetic aspects. Analyses by GWAS have identified more than 40 variants associated with PC risk, but the functions or biological significance of only a few variants or genes have been elucidated so far. The greatest challenge of the “post-GWAS” era is to understand the functional consequences of these loci and to attempt to increase our understanding of PC biology and prostate carcinogenesis. By combining multiple variants, PC risk estimations have been attempted for clinical tests and can identify a very small proportion population with high risk of PC. However, they cannot distinguish risk of aggressive PC from that of non-aggressive PC. Further identification of additional PC-susceptibility loci by larger GWAS analyses in diverse ethnic groups may provide biological insights that can be translated to clinical benefits. These potential benefits include the development of reliable biomarkers, risk estimation, and effective strategies for screening and preventing PC.
We are grateful to the members of the BioBank Japan Project and the Rotary Club of Osaka-Midosuji District 2660, Rotary International in Japan for supporting our study. This work was conducted as a part of the BioBank Japan Project, which was supported by the Japanese Ministry of Education, Culture, Sports, Sciences and Technology, research grant #22390306 (H. Nakagawa) from the Japan Society for the Promotion of Science, the Princess Takamatsu Cancer Research Fund, and the Takeda Science Foundation.
The authors have no conflicts of interest.
Type 2 diabetes mellitus
G protein-coupled receptor, family C, group 6, member A