Genetic polymorphisms in oestrogen receptor-binding sites affect clinical outcomes in patients with prostate cancer receiving androgen-deprivation therapy

Authors


  • Chun-Nung Huang, Shu-Pin Huang and Jiunn-Bey Pao contributed equally to this work.

Bo-Ying Bao, PhD, Department of Pharmacy, China Medical University, 91 Hsueh-Shih Road, Taichung 40402, Taiwan.
(fax: +886-4-22031075; e-mail: bao@mail.cmu.edu.tw).

Abstract

Abstract.  Huang C-N, Huang S-P, Pao J-B, Hour T-C, Chang T-Y, Lan Y-H, Lu T-L, Lee H-Z, Juang S-H, Wu P-P, Huang C-Y, Hsieh C-J, Bao B-Y (Kaohsiung Medical University Hospital, Kaohsiung; Kaohsiung Medical University, Kaohsiung; Taipei City Hospital, Taipei; Kaohsiung Medical University, Kaohsiung; China Medical University, Taichung; National Taiwan University Hospital; Oriental Institute of Technology; National Taiwan University, Taipei; China Medical University Hospital, Taichung, Taiwan). Genetic polymorphisms in oestrogen receptor-binding sites affect clinical outcomes in patients with prostate cancer receiving androgen-deprivation therapy. J Intern Med 2012; 271: 499–509.

Background.  Accumulating evidence indicates that oestrogens have significant direct effects on normal prostate development and carcinogenesis. The majority of the biological activities of oestrogens are mediated through the oestrogen receptor (ER), which functions as a hormone-inducible transcription factor to regulate target gene expression by binding to oestrogen response elements (EREs) in the regulatory regions of target genes. Sequence variants in EREs might affect the ER–ERE interaction and subsequent physiological activities. Therefore, we tested whether common single-nucleotide polymorphisms (SNPs) inside EREs are related to the clinical outcomes of androgen-deprivation therapy (ADT) in men with prostate cancer.

Methods.  We systematically evaluated 49 ERE SNPs predicted using a genome-wide database in a cohort of 601 men with advanced prostate cancer treated with ADT. The prognostic significance of these SNPs on disease progression, prostate cancer-specific mortality (PCSM) and all-cause mortality (ACM) after ADT was assessed using Kaplan–Meier analysis and a Cox regression model.

Results.  Based on multiple hypothesis testing, BNC2 rs16934641 was found to be associated with disease progression; in addition, TACC2 rs3763763 was associated with PCSM, and ALPK1 rs2051778 and TACC2 rs3763763 were associated with ACM. These SNPs remained significant in multivariate analyses that included known clinicopathological predictors. Moreover, a combined genotype effect on ACM was observed when ALPK1 rs2051778 and TACC2 rs3763763 were analysed in combination. Patients with a greater number of unfavourable genotypes had a shorter time to ACM during ADT (P for trend <0.001).

Conclusion.  The incorporation of ERE SNPs into models with known predictors might improve outcome prediction in patients with prostate cancer receiving ADT.

Introduction

Prostate cancer is the most common solid tumour and the second leading cause of cancer death amongst men [1]. Given the indolent nature of many prostatic tumours, a proportion of patients still present aggressive disease and ultimately die from prostate cancer. For many men with metastatic or high-risk localized disease, androgen-deprivation therapy (ADT) is the most commonly used treatment, either alone or in combination with other modalities. Although a dramatic initial response to ADT is common, disease progression frequently occurs at a median of 2–3 years, with a subsequent expected survival of 16–18 months [2]. As a result, there is a need to identify additional biomarkers to improve the prediction of the response to ADT and selection of optimal treatment strategies for high-risk patients.

Oestrogens have significant direct and indirect effects on prostate development and have been long suspected in the aetiology of prostatic diseases [3]. Direct effects are mediated through oestrogen receptors (ERs) in the prostate gland (i.e. ERα/ESR1 and ERβ/ESR2). Ligand binding to the ERs induces a conformational change, resulting in transformation to an activated state, and these activated ERs bind to the oestrogen response elements (EREs) within target gene promoters. The ligand/ERE-bound ER complex communicates with the general transcription apparatus via co-regulatory proteins to positively or negatively regulate target gene transcription and ultimately prostate development [4]. ERα is mainly expressed in prostate stromal cells where it is thought to stimulate growth factor release and induce epithelial cell proliferation. By contrast, ERβ is mostly located in the prostatic epithelium and is thought to be a potential ‘brake’ to androgen-driven proliferation [5]. Data from several studies have suggested a relationship between ERs and prostate cancer. ERα has been found in many prostate tumours and is negatively correlated with prognosis [6]. ERβ is commonly expressed in metastatic prostate cancer cells [7]. Given that the sequence variants within EREs might affect ER–ERE interaction resulting in altered target gene expression, it is important to investigate whether the most common genetic variations, single-nucleotide polymorphisms (SNPs), within EREs contribute to the efficacy of ADT in men with prostate cancer. We systematically performed a genome-wide search for SNPs in the putative EREs and found that some were significantly associated with clinical outcomes, disease progression, prostate cancer-specific mortality (PCSM) and all-cause mortality (ACM), in a cohort of patients with prostate cancer receiving ADT.

Materials and methods

Patient recruitment and data collection

Patients with prostate cancer who had been treated with ADT (orchiectomy or luteinizing hormone-releasing hormone agonist with or without anti-androgen treatment), including those with disease recurrence after local treatment, at three medical centres in Taiwan (Kaohsiung Medical University Hospital, Kaohsiung Veterans General Hospital and National Taiwan University Hospital) from 1995 to 2009 were included in this study [8]. Patient data were collected during prospective follow-up and disease progression, PCSM and ACM data were updated most recently in 2010. The prostate-specific antigen (PSA) nadir was defined as the lowest PSA value achieved during ADT treatment [9, 10]. Time to PSA nadir was defined as the time to reach the PSA nadir value after ADT initiation [11]. Disease progression was defined as a serial rise in PSA (i.e. at least two rises in PSA, >1 week apart) greater than the PSA nadir [12]. Initiation of secondary hormone treatment for a rising PSA was also considered as a progression event. Time to progression was defined as the interval in months between the initiation of ADT and progression. In general, patients were followed every month, with PSA tests at 3-monthly intervals. The cause of death was obtained by matching patients’ personal identification numbers with data from the official cause of death registry provided by the Department of Health, Executive Yuan, Taiwan. PCSM was defined as the interval between the initiation of ADT and death from prostate cancer. ACM was defined as the period from initiation of ADT to death from any cause. As the median PCSM and ACM had not been reached, the mean times to PCSM and ACM were estimated using Kaplan–Meier curves. Patients with missing clinicopathological information or an insufficient follow-up period were excluded; a total of 601 patients were included in this cohort. Overall, 145 deaths were identified, 101 of which were because of prostate cancer. This study was approved by the institutional review board of the three hospitals, and informed consent was obtained from all participants.

SNP selection and genotyping

Transcription factors are known to regulate different genes in different cellular contexts [13]; therefore, a genome-wide in silico prediction database of transcription factor-binding sites was used instead of chromatin immunoprecipitation data. We used PReMod (genomequebec.mcgill.ca/PReMod), a genome-wide cis-regulatory module prediction database, to identify computationally all putative EREs in the human genome [14]. PReMod used TRANSFAC (http://www.gene-regulation.com) version 7.2 position weight matrices (PWMs) to score putative transcription factor-binding sites based on how faithfully the human binding site and its orthologues in mouse and rat match the PWM. In addition, the prediction algorithm of PReMod exploits the observation that many known cis-regulatory modules often contain clusters of phylogenetically conserved and repeated transcription factor-binding sites [15] and thus has proven to be more reliable than other algorithms. The PReMod algorithm predicts that a total of 13 737 sites within the human genome are bound by the ER (canonical ER PWM: M00191; consensus: ARGNCANNNTGACCY) [16]. We identified SNPs within EREs by comparing two hexameric half-sites of these putative EREs with HapMap SNPs CHB (Han Chinese in Beijing, China) data in the University of California, Santa Cruz (UCSC) table browser (NCBI35/hg17) [17, 18]. SNPs within low-stringency EREs (occurrence score <34) or with a minor allele frequency of <10% in the HapMap CHB population were excluded, thus 52 SNPs in EREs were initially selected for analysis.

Genomic DNA was extracted from peripheral blood using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA, USA) and stored at −80 °C until the time of study. Genotyping was performed as described previously [19] using Sequenom iPLEX matrix-assisted laser desorption/ionization time-of-flight mass spectrometry technology at the National Genotyping Center, Academia Sinica, Taiwan. The average genotype call rate for these SNPs was 99.4%, and the average concordance rate was 99.8% amongst 54 blind duplicated quality control samples. Any SNP that did not conform to the Hardy–Weinberg equilibrium (< 0.005) or had a genotyping call rate of <90% was removed (= 3). Thus, a total of 49 SNPs were selected for further statistical analysis.

Real-time reverse transcription-polymerase chain reaction (RT-PCR)

Total RNA was extracted from LNCaP, PC-3 and DU 145 cells using Trizol (Invitrogen, Carlsbad, CA, USA). We carried out reverse transcription with the High Capacity cDNA Reverse Transcription Kit (ABI, Foster City, CA, USA) and PCR amplifications with Smart Quant Green Master Mix (Protech, Taipei, Taiwan) on an MJ mini and MiniOpticon real-time PCR detection system (Bio-Rad, Hercules, CA, USA). PCR was performed as follows: initial HotStart activation at 95 °C for 15 min, and 40 cycles of denaturation at 95 °C for 15 s, annealing and extension at 60 °C for 1 min. Primer sequences were ESR1, sense 5′-AAGTATGGCTATGGAATC-3′ and antisense 5′-TCGTTA TGTCCTTGAATA-3′; ESR2, sense 5′-ATACATACC TTCCTCCTAT-3′ and antisense 5′-TTCCAAGTTAGTGACATT-3′; basonuclin 2 (BNC2), sense 5′-AAGACAGACTCAGATATAAGG-3′ and antisense 5′-ATGAATG CTGGAAGGATT-3′; transforming acidic coiled-coil containing protein 2 (TACC2), sense 5′-CTCAGAATGGAAAGATAAATATG-3′ and antisense 5′-TGTTCGTCCTCTATCATC-3′; alpha-kinase 1 (ALPK1), sense 5′-CCATCCACAATACTACTGA-3′ and antisense 5′-CGACCACAACATCTCTAT-3′; glyceraldehyde-3-phosphate dehydrogenase, sense 5′-TCACCACCATGGAGAAGGC-3′ and antisense 5′-GCTAAGCAGTTGGTGGTGCA-3′. The quantification of each sample relative to the LNCaP sample was calculated using the 2−ΔΔCT method [20]. The expected sizes and the absence of nonspecific amplification products were confirmed by agarose gel electrophoresis and melting curve analysis.

Statistical analysis

Patients’ clinicopathological characteristics were summarized as the number and percentage of patients. Continuous factors were dichotomized at their median value within the cohort, with the exception of the PSA nadir, which was dichotomized into two groups with a cut-off at 0.2 ng mL−1, because of its correlation with disease progression and PCSM [9, 21]. The associations between clinicopathological characteristics and disease progression, PCSM and ACM were assessed using chi-squared tests. Kaplan–Meier curves and the log-rank test were used to analyse the association between each SNP and disease progression, PCSM and ACM. The heterozygous and rare homozygous genotypes were collapsed in the analysis if the frequency of the rare homozygote was low (<3%) or if the homozygous and heterozygous genotypes had the same direction of effect. A two-sided P-value of ≤0.05 was considered nominally significant. As we were testing 49 SNPs, multiple hypothesis testing using the Benjamini–Hochberg [22] method was applied to control the false discovery rate (FDR). Associations were deemed significant at the FDR <0.20 level. A multivariate Cox proportional hazards regression model including age at diagnosis, clinical stage, Gleason score, PSA at ADT initiation, PSA nadir, time to PSA nadir and treatment modality as covariates was fitted to re-evaluate the association between significant SNPs and disease progression, PCSM and ACM, considering the imbalances in the distributions of baseline characteristics. Based on our study population (= 601), we were able to detect a hazard ratio (HR) of 1.6, for a minor allele frequency of 0.2, with over 0.8 statistical power for disease progression, PCSM and ACM. R software version 2.12.2 and the Statistical Package for the Social Sciences software version 16.0.1 (SPSS Inc., Chicago, IL, USA) were used for statistical analyses.

Results

Table 1 summarizes the clinicopathological characteristics of our study population. Of the 601 patients with prostate cancer receiving ADT, 415 (69.3%) had disease progression during the mean follow-up of 30.3 months (range, 3–120 months). The median time to progression was 22 months [95% confidence interval (CI), 20–24 months]. During the mean follow-up of 39 months (range, 3–125 months), 145 (24.2%) patients died and an additional 101 (16.8%) died of prostate cancer. The estimated mean times to PCSM and ACM were 138 (95% CI, 132–145) and 123 (95% CI, 116–130) months, respectively. Clinical stage, PSA nadir, time to PSA nadir and treatment modalities before and during ADT were significantly associated ( 0.022) with time to progression, PCSM and ACM. Age at diagnosis was only associated with ACM, and the Gleason score and PSA level at ADT initiation were only associated with time to PCSM and ACM, but not time to progression.

Table 1.   Clinicopathological characteristics of the study population and analyses of factors that predicted disease progression, PCSM and ACM during ADT
VariableDisease progression, n (%)aPCSM, n (%)aACM, n (%)a
Progression-freeProgressionP*AliveDeceasedP*AliveDeceasedP* 
  1. ADT, androgen-deprivation therapy; PCSM, prostate cancer-specific mortality; ACM, all-cause mortality; PSA, prostate-specific antigen.

  2. aSubtotals are less than 601 for the total number of patients, 415 for the number with disease progression, 101 for cases of PCSM and 145 for ACM, owing to missing data.

  3. *P-values were calculated using the chi-squared test.

  4. Values of < 0.05 are in bold.

All patientsa184 (30.7)415 (69.3) 499 (83.2)101 (16.8) 455 (75.8)145 (24.2) 
Age at diagnosis, years
 ≤73 90 (48.9)228 (54.9)0.173268 (53.7)51 (50.5)0.555258 (56.7)61 (42.1) 0.002
 >74 94 (51.1)187 (45.1)231 (46.3)50 (49.5)197 (43.3)84 (57.9)   
Clinical stage at diagnosis
 T1/T2 70 (38.3)118 (28.6)0.006176 (35.5)13 (12.9)<0.001163 (36.1)26 (17.9)<0.001
 T3/T4/N1 61 (33.3)123 (29.8)163 (32.9)21 (20.8)152 (33.6)32 (22.1)   
 M1 52 (28.4)172 (41.6)157 (31.7)67 (66.3)37 (30.3)87 (60.0)   
Gleason score at diagnosis
 2–6 64 (35.4)129 (31.8)0.594174 (35.7)20 (20.0)<0.001160 (36.0)34 (23.8)<0.001
 7 56 (30.9)124 (30.5)161 (33.0)19 (19.0)148 (33.3)32 (22.4)   
 8–10 61 (33.7)153 (37.7)153 (31.4)61 (61.0)137 (30.8)77 (53.8)   
PSA at ADT initiation, ng mL−1
 <35100 (55.2)185 (46.7)0.057262 (54.8)24 (24.0)<0.001242 (55.5)44 (31.0)<0.001
 ≥35  81 (44.8)211 (53.3) 216 (45.2)76 (76.0)194 (44.5)98 (69.0)  
PSA nadir, ng mL−1
 <0.2113 (63.8)187 (45.1)<0.001281 (57.0)20 (20.0)<0.001264 (58.7)37 (25.9)<0.001
 ≥0.2  64 (36.2)228 (54.9)212 (43.0)80 (80.0)186 (41.3)106 (74.1)   
Time to PSA nadir, months
 <10  72 (40.7)220 (53.0)0.006228 (46.2)65 (65.0)0.001204 (45.3)89 (62.2)<0.001
 ≥10105 (59.3)195 (47.0)265 (53.8)35 (35.0)246 (54.7)54 (37.8)   
Treatment modality
 ADT as primary treatment107 (58.5)226 (54.6)0.022264 (53.1)69 (68.3)0.004233 (51.4)100 (69.0)<0.001
 ADT for post-RP PSA failure 27 (14.8) 40 (9.7) 62 (12.5) 6 (5.9) 60 (13.2)  8 (5.5)   
 ADT for post-RT PSA failure 2 (1.1) 16 (3.9) 15 (3.0) 3 (3.0) 13 (2.9)  5 (3.4)   
 Neoadjuvant/adjuvant ADT with RT 38 (20.8) 87 (21.0)115 (23.1)10 (9.9)111 (24.5) 14 (9.7)   
 Others 9 (4.9) 45 (10.9) 41 (8.2)13 (12.9) 36 (7.9) 18 (12.4)   

Univariate analysis using Kaplan–Meier curves and the log-rank tests of 49 SNPs in the putative EREs showed that time to progression, PCSM and ACM were nominally associated with five, five and seven polymorphisms, respectively (Table S1). SPRED2 rs17030511, GNPDA2 rs1398263, BNC2 rs16934 641, ZNF521 rs7238440 and ZNF507 rs17691999 were associated with disease progression during ADT (nominal  0.041; Table 2). However, after adjusting for FDR at the <0.20 level, only BNC2 rs16934641 was found to be associated with disease progression (FDR-adjusted = 0.012). To adjust for potential confounders, a multivariate Cox proportional hazards model, adjusting for age at diagnosis, clinical stage, Gleason score, PSA at ADT initiation, PSA nadir, time to PSA nadir and treatment modality, was used. BNC2 rs16934641 remained significantly associated with time to progression (adjusted = 0.017). Kaplan–Meier survival curves and log-rank tests showed that BNC2 rs16934641 CT/TT genotypes were significantly associated with shorter time to progression compared with the CC genotype (< 0.001; Table 2 and Fig. 1a left).

Table 2.   Genotyping frequencies and the association between genotype and disease progression during ADT
Gene
SNP
GenotypeNo. of patientsNo. of eventsMedian (months)P*FDRHR (95% CI)P**
  1. ADT, androgen-deprivation therapy; FDR, false discovery rate; HR, hazard ratio; 95% CI, 95% confidence interval; PSA, prostate-specific antigen.

  2. *P-values were calculated using the log-rank test.

  3. **HRs were adjusted for age, clinical stage, Gleason score, PSA at ADT initiation, PSA nadir, time to PSA nadir and treatment modality.

  4. FDR <0.20 are in bold.

SPRED2CC313216200.0260.363  
 rs17030511CT/TT28119524    
GNPDA2CC199141210.0270.363  
 rs1398263CT/TT38626523    
BNC2CC46331624<0.0010.0121.000.017
rs16934641CT/TT1319617 1.35 (1.05–1.72)  
ZNF521CC/CT523355230.0410.363  
 rs7238440TT725716    
ZNF507AA/AT524356220.0360.363  
 rs17691999TT665220    
Figure 1.

 Kaplan–Meier curves of (a) time to progression during androgen-deprivation therapy (ADT) stratified by BNC2 rs16934641, (b) time to prostate cancer-specific mortality during ADT stratified by TACC2 rs3763763 and (c) time to all-cause mortality during ADT for patients with 0, 1 or 2 unfavourable genotypes at the two genetic loci of interest; measured in all patients (left), in patients without distant metastasis (middle) and in patients with distant metastases (right). Numbers in parentheses indicate the number of patients.

ALPK1 rs2051778, SKAP2 rs212837, TACC2 rs3763763, SKAP1 rs7209855 and KLHL14 rs12970312 showed nominal associations with time to PCSM ( 0.048; Table 3). After adjusting for the FDR, only TACC2 rs3763763 was associated with PCSM. Patients carrying at least one A allele of TACC2 rs3763763 remained at a significantly higher risk of PCSM after adjusting for clinical predictors (HR, 2.01; 95% CI, 1.31–3.08 for CA/AA compared with the CC genotype; = 0.001; Table 3 and Fig. 1b left).

Table 3.   Genotyping frequencies and the association between genotype and prostate cancer-specific mortality during ADT
Gene
SNP
GenotypeNo. of patientsNo. of eventsEstimated mean (months)P*FDRHR (95% CI)P**
  1. ADT, androgen-deprivation therapy; FDR, false discovery rate; HR, hazard ratio; 95% CI, 95% confidence interval; PSA, prostate-specific antigen.

  2. *P-values were calculated using the log-rank test.

  3. **HRs were adjusted for age, clinical stage, Gleason score, PSA at ADT initiation, PSA nadir, time to PSA nadir and treatment modality.

  4. FDR <0.20 are in bold.

ALPK1GG448661440.0340.417  
 rs2051778GC/CC14632114    
SKAP2CC209451220.0190.310  
 rs212837CT/TT38554144    
TACC2CC292371340.0040.1961.000.001
rs3763763CA/AA307631312.01 (1.31–3.08)   
SKAP1GG436831230.0180.310  
 rs7209855GA/AA16217152    
KLHL14GG477741370.0480.417  
 rs12970312GA/AA11926128    

Seven SNPs (NR4A2 rs2691786, ALPK1 rs2051778, FBXO32 rs7830622, TACC2 rs3763763, AATF rs9330247, SKAP1 rs7209855 and KLHL14 rs12970312) were associated with ACM with nominal significance ( 0.050; Table 4). Based on our prespecified FDR, ALPK1 rs2051778 and TACC2 rs3763763 were significantly associated with ACM (FDR-adjusted = 0.098). After adjusting for clinical factors, the associations between ALPK1 rs2051778 and TACC2 rs3763763 and time to ACM remained significant (adjusted  0.007). A combined genotype effect on ACM was observed when these two SNPs were analysed in combination, and the HRs for ACM during ADT increased as the number of unfavourable genotypes increased (P for trend <0.001; Table 4 and Fig. 1c left).

Table 4.   Genotyping frequencies and the association between genotype and all-cause mortality during ADT
Gene
SNP
GenotypeNo. of patientsNo. of eventsEstimated mean (months)P*FDRHR (95% CI)P**
  1. ADT, androgen-deprivation therapy; FDR, false discovery rate; HR, hazard ratio; 95% CI, 95% confidence interval; PSA, prostate-specific antigen.

  2. *P-values were calculated using the log-rank test.

  3. **HRs were adjusted for age, clinical stage, Gleason score, PSA at ADT initiation, PSA nadir, time to PSA nadir and treatment modality.

  4. aUnfavourable genotypes refer to GC/CC in ALPK1 rs2051778 and CA/AA in TACC2 rs3763763.

  5. FDR <0.20 are in bold.

NR4A2GG/GC5561411200.0190.310  
 rs2691786CC383144    
ALPK1GG448951310.0040.0981.000.006
rs2051778GC/CC14647941.71 (1.17–2.50)   
FBXO32TT441981210.0270.319  
 rs7830622TC/CC15846112    
TACC2CC292561190.0040.0981.000.007
rs3763763CA/AA307881171.62 (1.14–2.31)   
AATFTT433961280.0480.319  
 rs9330247TC/CC16448106    
SKAP1GG4361141090.0500.319  
 rs7209855GA/AA16230136    
KLHL14GG4771081230.0470.319  
 rs12970312GA/AA11936110    
No. of unfavourable genotypes presenta
 0 22035127<0.001 1.00 
 1 30785118 2.00 (1.32–3.04)0.001 
 2 732588 2.65 (1.53–4.60)<0.001 
       P-trend<0.001

For further relevant clinical translation of the ERE SNP data, we stratified high-risk patients based on their metastatic status at the initiation of ADT. BNC2 rs16934641 particularly had a significant effect on disease progression in patients without distant metastasis compared to the patients with distant metastases (= 0.002; Fig. 1a middle). TACC2 rs3763763 particularly had a significant effect on PCSM in patients with distant metastases compared to the patients without distant metastasis (= 0.017; Fig. 1b right). ALPK1 rs2051778 and TACC2 rs3763763 combined had a significant effect on ACM in patients either with or without distant metastases ( 0.008; Fig. 1c middle and right). These data support the notion that the addition of ERE SNPs to the known predictors improved risk stratification and might help guide treatment decisions.

To gain an initial indication of these candidate genes in prostate cancer, we examined the expression levels of ESR1, ESR2, BNC2, TACC2 and ALPK1 in three commonly used human prostate cancer cell lines, LNCaP, PC-3 and DU 145, using real-time RT-PCR (Fig. 2). We were able to detect mRNA for all five genes in all three prostate cancer cell lines. The threshold cycle numbers shown on the bars in Fig. 2 reflect how many PCR cycles are required for the sample fluorescence to reach the threshold level; therefore, the sample containing more target transcripts approaches the threshold level at a lower cycle number. ESR2 appeared to be the major form of ER expressed in these prostate cancer cells, which is consistent with previous reports [23]. When considering gene expression relative to that of LNCaP cells, the endogenous receptors, including ESR1 and ESR2, were expressed at higher levels in PC-3 and DU 145 than in LNCaP cells. The expression of endogenous BNC2 transcripts correlated with ER status, that is, it was higher in PC-3 and DU 145 than in LNCaP, whereas the expression of TACC2 and ALPK1 transcripts was negatively correlated with ER status amongst the three cell lines. BNC2, TACC2 and ALPK1 have been shown to contain putative EREs in their promoter regions; this supports the biological plausibility of our findings that all three genes might be regulated by ERs although oppositely [14]. Thus, SNPs in the EREs of these genes might affect gene expression and contribute to prostate cancer progression.

Figure 2.

 The mRNA expressions of endogenous ESR1, ESR2, BNC2, TACC2 and ALPK1 in human prostate cancer cell lines. Total RNAs were prepared from LNCaP, PC-3 and DU 145 cells, and gene expressions were analysed using real-time RT-PCR. Relative gene expression represents the fold changes in gene expression relative to LNCaP cells set at 1.0. Numbers on the bars represent the threshold cycle of the gene in each cell line. Data are expressed as the mean ± SE of three independent experiments.

Discussion

Accumulating evidence has demonstrated that oestrogens and their corresponding receptor, ER, play crucial roles in prostate cancer development and progression. Because the biological functions of oestrogen are mainly mediated through ERs by binding to EREs and regulating target gene expression, the influence of common genetic variants in the EREs on prostate cancer progression should be extensively explored. The SNPs we examined in the present study were based on EREs predicted using a genome-wide database; therefore, they provided a unique opportunity to comprehensively evaluate their clinical significance without depending on a prior hypothesis. Here, we show that SNPs in EREs affect prognosis in patients with prostate cancer receiving ADT.

Taking advantage of the combination of methods to determine candidate responsive genes and genome-wide approaches, we identified three genetic variants, BNC2 rs16934641, TACC2 rs3763763 and ALPK1 rs2051778, that are significantly associated with disease progression, PCSM or ACM in patients with prostate cancer treated with ADT. Of these, TACC2 rs3763763 was significantly associated with both PCSM and ACM during ADT. BNC2 encodes an evolutionarily conserved DNA-binding zinc-finger protein, presumably as a regulatory protein of gene transcription [24] It is abundantly expressed in developing periurethral tissues and may play a role in the differentiation of spermatozoa and oocytes [25]. Furthermore, BNC2 maps to chromosome 9p22, a region with frequent homozygous deletions in glioblastoma and oesophageal adenocarcinoma [26, 27]. An ovarian cancer susceptibility locus in BNC2 was also identified by a genome-wide association study [28]. BNC2 expression was decreased in cancer cells, and stable expression caused cancer cell growth arrest, suggesting that BNC2 might function as a tumour suppressor during cancer development [26, 29]. TACC2 encodes a protein that concentrates at centrosomes throughout the cell cycle and has been implicated in tumorigenesis [30]. TACC2 was found to be downregulated with increasing malignancy of breast tumours and neuroblastomas [31, 32]. Overexpression of TACC2 in tumorigenic breast cancer cells reduced their malignant phenotypes both in vivo and in vitro, suggesting that TACC2 might function as a putative tumour suppressor [31]. ALPK1 is a member of the alpha-kinase family, which displays limited sequence similarity to conventional protein kinases. ALPK1 was implicated in epithelial cell polarity and exocytic vesicular transport towards the apical plasma membrane by phosphorylating myosin IA, an apical vesicle transport motor protein [33]. Although ALPK1 has been found to be one of the signature genes to predict chemotherapeutic response to paclitaxel in patients with breast cancer [34], its role in cancer currently remains largely unknown. The expression of BNC2 was upregulated in wild-type mice aortas after 17β-oestradiol treatment, but it was not changed in ERα knockout mice aortas [35], indicating that BNC2 might be a direct oestrogen target gene. Despite the lack of direct evidence, both TACC2 and ALPK1 have been implicated in female cancers such as breast and ovarian cancers [31, 34], suggesting that these genes might also be regulated by oestrogen. No SNP is simultaneously associated with prostate cancer progression and mortality after ADT. The reasons for this need further investigation, but our results suggest that different biological pathways and distinct underlying risk factors might be involved during prostate cancer progression and mortality. Thus, a locus associated with prostate cancer progression might not necessarily be associated with mortality.

To our knowledge, this study represents the most extensive analysis of ERE SNPs in one of the largest cohorts of patients with advanced prostate cancer for whom ADT was the main therapeutic intervention. Although we cannot rule out the possibility of false-positive findings given the number of tests performed, this is an exploratory study, with the aim of generating new hypotheses to determine the efficacy of ADT, which needs to be validated in independent populations. In addition, our findings in this homogeneous Chinese Han population might not be generalizable to other ethnic groups. Regardless of these potential limitations, we have identified genetic variants in BNC2, TACC2 and ALPK1 in association with clinical outcomes after ADT. There was also a cumulative effect on ACM following ADT of combinations of genotypes across the two loci of interest. Moreover, these SNPs retained their association with the efficacy of ADT after adjusting for multiple hypothesis testing and known clinicopathological predictors (age at diagnosis, clinical stage, Gleason score, PSA level at ADT initiation, PSA nadir, time to PSA nadir and treatment modality), suggesting that these host genetic factors add information above and beyond currently used predictors. Our results revealed novel pathways that govern response to ADT and might serve as potential biomarkers in an outcome prediction model to guide individual therapeutic interventions.

Conflict of interest statement

The authors have no conflicts of interest to declare.

Acknowledgements

This work was supported by the National Science Council (NSC), Taiwan (NSC-98-2320-B-039-019-MY3, NSC-99-2314-B-037-018-MY3, and NSC-100-2314-B-039-009-MY3), China Medical University (CMU99-COL-13), and Kaohsiung Medical University Hospital (KMUH99-9R12). We thank the National Genotyping Center of the National Research Program for Genomic Medicine, NSC, Taiwan, for technical support.

Ancillary