|Home | About | Journals | Submit | Contact Us | Français|
Prostate cancer gene 3 (PCA3) is a non-coding gene specifically overexpressed in prostate cancer (PCa) that has great potential as a clinical biomarker for predicting prostate biopsy outcome. However, genetic determinants of PCA3 expression level remain unknown. To investigate the association between genetic variants and PCA3 mRNA level, a genome-wide association study was conducted in 1371 men of European descent in the REduction by DUtasteride of prostate Cancer Events trial. First-voided urine specimens containing prostate cells were obtained after digital rectal examination. The PROGENSA PCA3 assay was used to determine PCA3 score in the urinary samples. A linear regression model was used to detect the associations between (single nucleotide polymorphisms) SNPs and PCA3 score under an additive genetic model, adjusting for age and population stratification. Two SNPs, rs10993994 in β-microseminoprotein at 10q11.23 and rs10424878 in kallikrein-related peptidase 2 at 19q13.33, were associated with PCA3 score at genome-wide significance level (P = 1.22 x 10-9 and 1.06 x 10-8, respectively). Men carrying the rs10993994 “T” allele or rs10424878 “A” allele had higher PCA3 score compared with men carrying rs10993994 “C” allele or rs10424878 “G” allele (β = 1.25 and 1.24, respectively). This is the first comprehensive search for genetic determinants of PCA3 score. The novel loci identified may provide insight into the molecular mechanisms of PCA3 expression as a potential marker of PCa.
Prostate cancer (PCa) is the most prevalent noncutaneous malignancy and the second leading cause of cancer mortality in the United States . Prostate-specific antigen (PSA) is the most commonly used screening tool for PCa. PSA screening has led to progress in early detection . However, elevated serum PSA is regularly detected in benign prostate disorders such as prostatitis and benign prostatic hyperplasia (BPH), resulting in a relatively high rate of unnecessary PSA-triggered biopsies .
Compared to PSA, prostate cancer gene 3 (PCA3) is more specific for PCa. PCA3 is a prostate-specific non-coding mRNA that is over-expressed in prostate tumor tissue and metastases and is not observed in conjunction with benign prostate disorders such as prostatitis or enlarged prostate . Using PCA3, prostate tissue specimens containing less than 10% of cancer cells can be accurately discriminated from noncancer tissue . PCA3 score is generated from the ratio of PCA3 mRNA level to PSA mRNA level, through which PCA3 expression is normalized with the PSA expression used as a housekeeping gene. Recently, several independent studies have shown that PCA3 score has better performance than PSA in predicting prostate biopsy outcome in subjects with elevated PSA (>2.5 ng/ml) [5–8]. PCA3 score was also observed to be indicative of biopsy Gleason score in the REduction by DUtasteride of prostate Cancer Events (REDUCE) trial  and another study in a multidescent population by Nakanishi et al. . Moreover, PCA3 was correlated with prostate tumor size .
Despite the well-characterized clinical use of PCA3, our knowledge of the genetic determinants of PCA3 expression levels is negligible. In this study, we conducted a genome-wide association study (GWAS) to identify SNPs that are associated with the urine level of PCA3 in the REDUCE trial.
Subjects in our GWAS were originally enrolled in the REDUCE trial, a 4-year multicenter, randomized, double-blind, placebo-controlled clinical study to evaluate the effect of dutasteride on the risk of PCa in men. Individuals participating in this trial had to meet the following criteria: 1) age between 50 and 75 years; 2) prostate volume (PV) ≤80 cm3; 3) PSA level of 2.5 to 10 ng/ml for subjects <60 years old and 3 to 10 ng/ml for those ≥60 years; 4) negative biopsy (within 6 months of enrollment) for PCa, high-grade prostatic intraepithelial neoplasia, or atypical small acinar proliferation .
In the REDUCE trial, a total of 1649 individuals of European descent (885 in the placebo and 764 in the dutasteride treatment arms) had available PCA3 measurement. Among them, 162 of 885 subjects (18.3%) in the placebo and 116 of the 764 subjects (15.2%) in the dutasteride arm were diagnosed with PCa in the 4-year follow-up. The remaining 1371 individuals (723 in the placebo arm and 648 in the dutasteride arm) with negative prostate biopsy were included for further GWAS analysis. Characteristics of the 1371 individuals are summarized in Table W1.
First-voided urine specimens containing prostate cells were obtained after digital rectal examination before year 2 and 4 biopsies. The PROGENSA PCA3 Assay was used to measure PCA3 and PSA mRNA levels in the urinary samples at an independent laboratory (Caris Life Sciences/Molecular Profiling Institute, Phoenix, AZ) .The firstavailable PCA3 and PSA value (at year 2 or year 4) was used in our analysis.
Genotyping was conducted using an Illumina HumanOmniExpress BeadChip system (729,755; single nucleotide polymorphisms [SNPs] included) at the Center for Cancer Genomics, Wake Forest University School of Medicine. Samples with a genome-wide call rate ≥95% were included in subsequent GWAS analysis. For SNPs, the quality control criteria were as follows: minor allele frequency >0.01, genotype call rate >95%, and Hardy-Weinberg Equilibrium >0.001. The remaining 586,462 SNPs were used to impute the genotypes of SNPs that were not directly genotyped in the genome, by IMPUTE software  using combined data of the 1000 Genomes low-coverage pilot project and HapMap3 data as references (www.1000genomes.org and www.hapmap.org). The same quality control and a posterior probability of >90% was applied to imputed SNP calls.
Personalized PCA3 score and cutoff were calculated for each individual. First, a linear regression model was fitted, where log-transformed PCA3 scores were treated as outcomes and the number of alleles associated with higher PCA3 scores was treated as covariates. Genetic effects were calculated for three genotypes (aa, Aa, and AA), using the fitted values. Relative allelic effects (percentage of increase per allele) were calculated by dividing the fitted values of aa and Aa. The relative risk to the population for each genotype was then computed on the basis of the relative allelic effect and genotypic frequency. Second, assuming a multiplicative model, the combined relative genetic effect was calculated by multiplying the relative genotypic effect for each SNP relative to the general population. Third, a personalized PCA3 score for each subject was determined by dividing the raw PCA3 score by the combined relative genetic effect, while the personalized cutoff for each subject was generated by multiplying a uniform cutoff of 35 by the combined relative genetic effect.
We calculated PCA3 score as follows: PCA3 mRNA copy number/PSA mRNA copy number x 1000. PSA mRNA copy number here was used to correct for the number of prostate cells present in urine sample. PCA3 score was log transformed to approximate a normal distribution for further analysis (Figure W1).
A linear model implemented in PLINK software package  was used to detect the association of each SNP (genotyped and imputed) with PCA3 score under an additive genetic model in 1371 subjects negative for PCa biopsy. Principal component analysis was conducted to detect potential population stratification by EIGENSTRAT software . Age and the top eigenvector that indicates ancestral heterogeneity within a group of individuals were adjusted as covariates in multiple regression analysis. Genome-wide significance level was set at a P value of 1 x 10-8. For regions containing more than one significant variant (P <10-6), multiple linear regression analysis was applied to test the independence of each SNP, adjusting for the most significant SNPs, as well as all other covariates.
Urine mRNA levels of PCA3, PSA (copies per milliliter, expressed as PCA3CPM and PSACPM, respectively), as well as the resulting PCA3 scores are summarized in Table W1. Although the median values of PCA3CPM and PSACPM in the dutasteride arm differed from those in the placebo arm (P < .0001), PCA3 scores were not significantly different between dutasteride and placebo arms (Wilcoxon rank sum test, P = .72).
In the present GWAS, a total of 3,076,666 SNPs (genotyped and imputed) were analyzed for association with log-transformed PCA3 score in 1371 subjects (placebo and dutasteride arms; Figure 1). The inflation factor of GWAS results based on a quantile-quantile (Q-Q) plot analysis was modest (λ = 1.01; Figure W2). Therefore, the reported P values are not corrected for genomic inflation.
Two SNPs, rs10993994 in the β-microseminoprotein (MSMB) gene at 10q11.23 and rs10424878 in the kallikrein-related peptidase 2 (KLK2) gene at 19q13.33, reached a genome-wide significant P value of 1.22 x 10-9 and 1.06 x 10-8, respectively (Table 1), adjusting for age and the top eigenvector for potential population stratification in an additive model. Seventeen SNPs with a P value smaller than 1 x 10-6 are presented in Table W2. Several other SNPs in the MSMB gene region or KLK2/kallikrein-related peptidase 3 (KLK3) gene cluster were significant in the P value range of 1 x 10-6 to 1 x 10-8 (Figure 2, A and B). After adjusting for the most significant SNPs and all other covariates, none of the other SNPs in the MSMB or KLK2/KLK3 regions remained independently associated with PCA3 score. No significant association between PCa risk and the 183 SNPs within or surrounding the PCA3 gene at 9q21 (78,519,174–78,642,285, NCBI build 36) was detected.
The effects of rs10993994 and rs10424878 on PCA3 score were then evaluated. As shown in Table 1, the least square mean values of PCA3 score were 13.35, 16.82, and 20.11 for the rs10993994 CC, CT, and TT carriers, respectively (β = 1.25 for “T” allele). The least square mean values of PCA3 score were 13.68, 16.61, and 20.76 for rs10424878 GG, AG, and AA carriers, respectively (β = 1.24 for “A” allele).
In addition, we examined the associations of the two top SNPs with PCa risk, baseline PV, and International Prostate Symptom Score (IPSS; Table W3). The SNP rs10993994 was significantly associated with PCa risk (odds ratio [OR] = 1.20, P = .02), baseline PV (β = 0.03, P = .01), and IPSS (β = -0.27, P = .05). The SNP rs10424878 was not significantly associated with PCa risk (OR = 1.08, P = .35) or baseline IPSS (β = 0.04, P = .76) but was associated with baseline PV (β = -0.02, P = .02).
Moreover, a personalized cutoff value of PCA3 score was estimated for each individual. The estimated personalized cutoff ranged from 24.9 to 60.6. The estimated personalized cutoff was 24.9 for subjects carrying both the rs10993994 CC genotype and the rs10424878 GG genotype, while it was 60.6 for subjects carrying both the rs10993994 TT genotype and the rs10424878 AA genotype (Figure W3).
In the present GWAS, conducted among 1371 European descent men with negative prostate biopsies, the strongest overall effects on PCA3 score were observed for two independent genetic variants, rs10993994 in MSMB at 10q11.21 (P = 1.22 x 10-9) and rs10424878 in KLK2 at 19q13.33 (P = 1.06 x 10-8). Presently, this is the first discovery of genetic determinants based on a genome-wide study for PCA3, one of the most valuable early detection biomarkers for PCa.
SNP rs10993994 is located 57 bp upstream of exon 1 of the MSMB gene, which encodes MSMB, one of the major constituents of seminal plasma . MSMB is a member of immunoglobulin binding factor family and can act as a tumor suppressor by binding to cell surface receptors and regulating prostate cell apoptosis through mitogen-activated protein (MAP) kinase/AKT signaling . Previous studies discovered that the expression of MSMB is much higher in normal and benign prostate tissues than in tumor tissues  and urinary MSMB level decreases during tumorigenesis .
As the most thoroughly studied genetic variant in the MSMB gene region, rs10993994 has been reported to be associated with PCa  and this association appears to be independent of age at diagnosis or tumor grade . Functional investigation revealed that the minor allele “T” of rs10993994 was causally associated with lower expression of MSMB, because replacement of wild-type C to T altered a cAMP response element-binding protein transcription binding site, resulting in reduced promoter activity . Fine mapping of the MSMB promoter region did not identify other SNPs having major effects regulating MSMB expression . Taken together, these findings suggest rs10993994 as a potential independent causal genetic variant for PCa.
In our study, rs10993994 in MSMB was associated with PCA3 score, a risk factor for PCa. The T allele, associated with a decrease in MSMB promoter activity and an increase in PCa risk, was associated with an increase in urinary PCA3. Since rs10993994 was associated with PCa risk in the placebo arm of REDUCE trial (OR = 1.20, P = .02), one potential scenario is, even in this population of men screened for PCa, men with the T allele would be more likely to have (undetected) PCa and thus have a higher urinary PCA3. However, if this was the explanation, one would expect other PCa risk SNPs (e.g., SNPs at 8q24 and HNF1B) to be associated with urinary PCA3. Except for rs10993994, the majority of the other 32 known PCa risk-associated SNPs  were not significantly associated with PCA3 score (P > .0016, Bonferroni corrected for 32 tests), and only two of these documented PCa risk SNPs, rs10486567 and rs1512268, have P values slightly smaller than .0016 (.0005 and .0010, respectively). Therefore, it is unlikely that rs10993994 affects PCA3 score through this mechanism.
Given that rs10993994 in MSMB is also associated with serum PSA levels, we also considered the possibility that rs10993994 affects PCA3 score indirectly through PSA. Additional association analysis of PCA3 score in which urinary PSA level was adjusted as a covariate showed that urinary PSA level had little effect on PCA3 score (P = .84). Moreover, rs10993994 is not statistically related to serum PSA level among the 1371 subjects in our study (P = .37, adjusted for age and the top eigenvector). Therefore, the association between rs10993994 and PCA3 score found in our study is most likely to be independent of PSA level in urine or in serum.
Since rs10993994 is located within an androgen receptor binding region  and androgen signaling plays an important role in the pathogenesis of PCa , it is possible that rs10993994 has an effect on PCA3 score through the androgen action pathway. In the present study, 648 of the total 1371 subjects were in the dutasteride arm. Dutasteride is a 5α-reductase inhibitor that reduces systemic and tissue levels of 5α-dihydrotestosterone, the principal androgen in the prostate . Therefore, we compared the effects of rs10993994 on PCA3 score in the placebo and dutasteride arms separately. The β values of the T allele in an additive model (adjusted for age and the top eigenvector) in 723 placebo subjects and 648 dutasteride subjects were 1.27 and 1.24, respectively, with P values of 2.5 x 10-6 and 6.7 x 10-5. The effects of rs10993994 on PCA3 score were quite similar in the two arms. On the basis of these results, rs10993994 may not affect PCA3 score through the androgen pathway.
There is still another scenario that may explain the association between rs10993994 and PCA3 score. There may be some currently unknown downstream relationship whereby the level of MSMB expression directly affects the production of PCA3 mRNA or its ability to get into the urine. However, in vitro studies with PCA3-producing cell lines and other functional studies are required to address this hypothesis.
The second SNP associated with PCA3 score, rs10424878, is about 2kbupstream of KLK2, the coding gene of KLK2. As a member of kallikrein gene family, KLK2 is a serine protease and can activate urokinase-type plasminogen activator, cleave binding proteins of insulin-like growth factor, and cleave the preform of PSA to the mature enzyme, participating in a proteolytic cascade, which has been associated with tumorigenesis and metastasis of PCa . In addition, PCa tumor tissues have a higher KLK2 expression level than benign tissues [27,28] and patients with PCa were also found to have higher serum KLK2 [29,30].
Genetic variants in KLK2 were found to be associated with PCa in candidate gene association studies , although no SNPs in KLK2 have been confirmed or identified by GWAS. In our GWAS on PCA3 score, rs10424878 and another nine SNPs in KLK2 and five SNPs in KLK3 reached a P value of 1 x 10-6. However, the associations between all of the other 14 KLK2/KLK3 SNPs and PCA3 score no longer existed after adjusting for rs10424878 (P > .05), indicating that rs10424878 is the only independent SNP in KLKs that is associated with PCA3 score. Unfortunately, little is known about the pathways or functional relationships between the KLK2 and PCA3 gene. Additional association analysis of PCA3 score in which urinary PSA level was adjusted as a covariate showed that urinary PSA level had little effect on PCA3 score (P = .89). We also assessed serum PSA level with rs10424878, and no association was found (P = .12, adjusted for age and the top eigenvector). Overall, the effect of rs10424878 on PCA3 scores was unlikely to be mediated through serum PSA levels.
Another possible impact of rs10424878/KLK2 on PCA3 score could be through the androgen action pathway, since KLK2 is highly responsive to androgens. As mentioned, androgen signaling is important for PCa pathogenesis  and dutasteride could greatly reduce 5α-dihydrotestosterone levels . Similar to rs10993994, we assessed the effect of rs10424878 on PCA3 score separately in the placebo and dutasteride arms. The β values of “A” allele in an additive model (adjusted for age and the top eigenvector) in 723 placebo subjects only and 648 dutasteride subjects only are 1.22 and 1.26, respectively, with P values of 2.06 x 10-4 and 1.13 x 10-5. The effects of rs10424878 on PCA3 score were quite similar in the two arms. Therefore, it is unlikely that rs10424878 affects PCA3 score through the androgen pathway. Additional biologic studies are needed to elucidate the mechanism through which rs10424878 impacts urine PCA3 scores.
On the basis of genetic discovery, we estimated personalized PCA3 scores for 1649 subjects with PCA3 measurements, including 278 subjects with positive prostate biopsy results and 1371 with negative biopsy results. We compared the area under curve (AUC) between the model with the raw PCA3 score and the model with the personalized PCA3 score for predicting prostate biopsy outcome. The AUCs for both models were 0.681 and 0.682, respectively. We did not observe a statistically significant difference in AUCs between the two models, which may due to the limited number of genetic determinants identified for the PCA3 score and the relatively small effect of each variant. Although the personalized PCA3 scores showed limited improvement in AUC compared with the raw PCA3 scores, the detection rate of positive prostate biopsy was higher for men with 35 and higher for the personalized PCA3 score (31.0%, 120/388) compared with the detection rate for men with 35 and higher for the raw PCA3 score (28.3%, 119/421). Overall, improvement in clinical use of the personalized PCA3 score was limited in the current study. However, it may improve if more genetic variants for PCA3 score are identified in future studies.
Two advantages of this study were its prospective design and the stringent inclusion and exclusion criteria. Nevertheless, several limitations need to be addressed. Although the most stringent Bonferroni correction was applied to control false positive findings, our study did not include a confirmation stage in an independent population, due to the difficulty in identifying additional studies with available germline DNA, PCA3 measurements, and prostate biopsy outcomes. Another limitation is that only two SNPs were identified to be associated with PCA3 score at a genome-wide level in this study. Therefore, further discoveries and validations across populations are warranted.
In conclusion, this study represents the first effort to investigate genetic determinants of PCA3 score, which may advance the understanding of PCA3 score. More importantly, the genetic determinants identified might be helpful to facilitate the utilization of personalized cutoff of PCA3 score as a marker for PCa prediction.
We thank the patients enrolled in REDUCE (sponsored by GSK) who provided consent and genetic samples that enabled this study, and the clinicians who contributed their expertise in recruiting study patients for the REDUCE clinical study. Dave Pulford, Jennifer Aponte, Jon Charnecki, and Mary Ellyn Volk participated in consent reconciliation and sample management to enable genetic sample selection for inclusion and genotype determination. Karen King provided data management support for this project. We appreciate the assistance of Lauren Marmor in coordinating the support of the Avodart Collaborative Research Team.
1This work was supported by a National Cancer Institute RC2 grant (CA148463) and a research contract by GlaxoSmithKline (GSK) to J.X. L.D.C. and R.S.R. were GSK employees during this study and hold stock in GSK. J.G. is an employee of Hologic Gen-Probe, which supported PROGENSA PCA3 Assay testing under contractual agreement with GSK. J.X. certifies that all conflicts of interest, including specific financial interests and relationships and affiliations relevant to the subject matter or materials discussed in the manuscript (e.g., employment/affiliation, grants or funding, consultancies, honoraria, stock ownership or options, expert testimony, royalties, or patents filed, received, or pending), have been disclosed.