Abstract

Background:

Activated androgen receptor binds to androgen-responsive elements (AREs) in genome to regulate target gene transcription and, consequently, mediates physiological or tumorigenic processes of the prostate. Our aim was to determine whether genetic variants in AREs are associated with clinical outcomes after androgen-deprivation therapy (ADT) in prostate cancer patients.

Patients and methods:

We systematically investigated 55 common single-nucleotide polymorphisms (SNPs) in the genome-wide insilico-predicted AREs 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 by Kaplan–Meier analysis and Cox regression model.

Results:

In univariate analysis, two, five, and four SNPs were associated with disease progression, PCSM, and ACM, respectively. After adjusting for known prognostic factors, ARRDC3 rs2939244, FLT1 rs9508016, and SKAP1 rs6504145 remained as significant predictors for PCSM and FBXO32 rs7830622 and FLT1 rs9508016 remained as significant predictors for ACM in multivariate analysis. Moreover, strong combined genotype effects on PCSM and ACM were also observed (Ptrend < 0.001).

Conclusion:

Our results suggest that SNPs in AREs influence prostate cancer survival and may further advance our understanding of the disease progression.

introduction

Androgens play a pivotal role in the development of normal prostate and remain important in prostate cancer. The biological effects of androgens are primarily mediated by binding to the androgen receptor (AR), which belongs to the nuclear receptor superfamily. Upon androgen binding, the androgen–AR complex interacts with coregulators and binds to specific androgen-responsive elements (AREs) in the promoter region of androgen-responsive genes to regulate gene expression and ultimately cell growth in target tissues [1]. Therefore, the goal of androgen-deprivation therapy (ADT) is to reduce androgen levels and make prostate cancer shrink. Although initially effective, ADT eventually fails, leading to prostate cancer progression in a hormone refractory manner [2, 3].

There are a variety of clinicopathologic parameters, such as tumor stage, Gleason score, and prostate-specific antigen (PSA) kinetics, that have been used in clinical practice to help predict which patients have poor prognoses for disease progression or mortality [4–6]. However, the clinical parameters’ prognostic capabilities are still limited and might be improved by the incorporation of other factors, including genetic markers [7]. Given that the essential role of androgen–AR in prostate cancer and that sequence variants within AREs might affect AR–ARE interaction, resulting in altered target gene expression, we carried out a genome-wide search for the most common genetic variation, single-nucleotide polymorphisms (SNPs), within AREs and investigated their prognostic significance on disease progression, prostate cancer-specific mortality (PCSM), and all-cause mortality (ACM) in a cohort of prostate cancer patients receiving ADT.

patients and methods

patient recruitment and data collection

The study population was extended from our hospital-based prostate cancer case–control study, which has been described previously [8–13]. Briefly, patients with diagnosed and pathologically confirmed prostate cancer were actively recruited from three medical centers in Taiwan: Kaohsiung Medical University Hospital, Kaohsiung Veterans General Hospital, and National Taiwan University Hospital. Prostate cancer patients who had been treated with ADT (orchiectomy or luteinizing hormone-releasing hormone agonist, with or without antiandrogen), including those with disease recurrence after local treatments (radical prostatectomy or radiotherapy), were identified and followed up prospectively to evaluate genetic variants as prognostic predictors of clinical outcomes during ADT. Patients were excluded if the clinicopathologic information or follow-up period was insufficient, leaving 601 patients in this cohort. This study was approved by the Institutional Review Board of the three hospitals, and informed consent was obtained from each participant.

Data were collected on patients with disease baseline and clinicopathologic characteristics, as well as three treatment outcomes: time to progression, PCSM, and ACM. The PSA nadir was defined as the lowest PSA value achieved during ADT treatment [6, 14]. Time to PSA nadir was defined as the length of time it took for the PSA value to reach the nadir after initiation of ADT [4]. Disease progression was defined as a serial rise in PSA, at least two rises in PSA (>1 week apart), reaching values greater than the PSA nadir [15]. Initiation of secondary hormone treatment of 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-month intervals. The cause of death was obtained by matching patients’ personal identification numbers with the official cause of death registry provided by the Department of Health, Executive Yuan, Taiwan. As presented in Table 1, 415 (69%) patients exhibited disease progression after ADT, and the median time to progression was 22 months with a mean follow-up of 30.3 months (range, 3–120 months). One hundred and forty-five (24%) patients died, and 101 (17%) patients died of prostate cancer after a mean follow-up of 39 months (range, 3–125 months). The estimated mean times to PCSM and ACM were 138 and 123 months, respectively. The clinical stage, Gleason score, PSA nadir, time to PSA nadir, and treatment modalities before and during ADT were significantly associated (P ≤ 0.006) with time to progression, PCSM, and ACM. Age at diagnosis was only associated with ACM, and the PSA level at ADT initiation was associated with time to PCSM and ACM, but it was not associated with time to progression.

Table 1.

Clinicopathologic characteristics of the study population and analyses of factors that predicted disease progression, PCSM, and ACM during ADT

Variable Na (%) Disease progression
 
PCSM
 
ACM
 
 
Median (95% CI), months PEstimated mean (95% CI), months PEstimated mean (95% CI), months P 
All patients 601 22 (20–24)  138 (132–145)  123 (116–130)   
Age at diagnosis, years 
    Median (IQR) 73 (67–79)       
    ≤73 320 (53.2) 21 (18–24) 0.219 141 (133–150) 0.280 135 (126–144) 0.001  
    >73 281 (46.8) 25 (21.29)  127 (118–137)  105 (94–115)   
Clinical stage at diagnosis 
    T1/T2 189 (31.7) 25 (20–30) 0.005 145 (138–153) <0.001 130 (119–140) <0.001  
    T3/T4/N1 184 (30.8) 25 (22–28)  149 (138–160)  138 (127–150)   
    M1 224 (37.5) 17 (15–19)  110 (99–121)  95 (83–106)   
Gleason score at diagnosis 
    2–6 194 (33.0) 26 (22–30) 0.006 154 (146–163) <0.001 141 (130–151) <0.001  
    7 180 (30.6) 25 (22–28)  134 (124–145)  116 (103–130)   
    8–10 214 (36.4) 17 (15-19)  108 (96-121)  97 (85-109)   
PSA at ADT initiation, ng/ml 
    Median (IQR) 35.0 (11.4–129)       
    <35 287 (49.6) 25 (22–28) 0.083 146 (139–153) <0.001 132 (123–141) <0.001  
    ≥35 292 (50.4) 19 (15–23)  117 (108–127)  103 (93–113)   
PSA nadir, ng/ml 
    Median (IQR) 0.18 (0.01–1.33)       
    <0.2 301 (50.8) 31 (28–34) <0.001 159 (152–166) <0.001 145 (136–154) <0.001  
    ≥0.2 292 (49.2) 14 (12–16)  110 (100–119)  95 (85–105)   
Time to PSA nadir, months 
    Median (IQR) 10 (5–18)       
    <10 293 (49.4) 10 (9–11) <0.001 121 (111–131) <0.001 105 (94–116) <0.001  
    ≥10 300 (50.6) 33 (30–36)  150 (142–157)  136 (127–146)   
Treatment modality 
    ADT as primary treatment 333 (55.7) 21 (18–24) 0.001 131 (122–140) 0.003 114 (105–124) <0.001  
    ADT for post  RP PSA failure 68 (11.4) 25 (22–28)  118 (105–131)  113 (99–127)   
    ADT for post  RT PSA failure 18 (3.0) 12 (10–14)  110 (88–131)  84 (50–117)   
    Neoadjuvant/adjuvant ADT with RT 125 (20.9) 29 (24–34)  133 (126–140)  126 (116–136)   
    Others 54 (9.0) 14 (12–16)  103 (85–121)  89 (71–107)   
Variable Na (%) Disease progression
 
PCSM
 
ACM
 
 
Median (95% CI), months PEstimated mean (95% CI), months PEstimated mean (95% CI), months P 
All patients 601 22 (20–24)  138 (132–145)  123 (116–130)   
Age at diagnosis, years 
    Median (IQR) 73 (67–79)       
    ≤73 320 (53.2) 21 (18–24) 0.219 141 (133–150) 0.280 135 (126–144) 0.001  
    >73 281 (46.8) 25 (21.29)  127 (118–137)  105 (94–115)   
Clinical stage at diagnosis 
    T1/T2 189 (31.7) 25 (20–30) 0.005 145 (138–153) <0.001 130 (119–140) <0.001  
    T3/T4/N1 184 (30.8) 25 (22–28)  149 (138–160)  138 (127–150)   
    M1 224 (37.5) 17 (15–19)  110 (99–121)  95 (83–106)   
Gleason score at diagnosis 
    2–6 194 (33.0) 26 (22–30) 0.006 154 (146–163) <0.001 141 (130–151) <0.001  
    7 180 (30.6) 25 (22–28)  134 (124–145)  116 (103–130)   
    8–10 214 (36.4) 17 (15-19)  108 (96-121)  97 (85-109)   
PSA at ADT initiation, ng/ml 
    Median (IQR) 35.0 (11.4–129)       
    <35 287 (49.6) 25 (22–28) 0.083 146 (139–153) <0.001 132 (123–141) <0.001  
    ≥35 292 (50.4) 19 (15–23)  117 (108–127)  103 (93–113)   
PSA nadir, ng/ml 
    Median (IQR) 0.18 (0.01–1.33)       
    <0.2 301 (50.8) 31 (28–34) <0.001 159 (152–166) <0.001 145 (136–154) <0.001  
    ≥0.2 292 (49.2) 14 (12–16)  110 (100–119)  95 (85–105)   
Time to PSA nadir, months 
    Median (IQR) 10 (5–18)       
    <10 293 (49.4) 10 (9–11) <0.001 121 (111–131) <0.001 105 (94–116) <0.001  
    ≥10 300 (50.6) 33 (30–36)  150 (142–157)  136 (127–146)   
Treatment modality 
    ADT as primary treatment 333 (55.7) 21 (18–24) 0.001 131 (122–140) 0.003 114 (105–124) <0.001  
    ADT for post  RP PSA failure 68 (11.4) 25 (22–28)  118 (105–131)  113 (99–127)   
    ADT for post  RT PSA failure 18 (3.0) 12 (10–14)  110 (88–131)  84 (50–117)   
    Neoadjuvant/adjuvant ADT with RT 125 (20.9) 29 (24–34)  133 (126–140)  126 (116–136)   
    Others 54 (9.0) 14 (12–16)  103 (85–121)  89 (71–107)   

P < 0.05 are in boldface.

a

Column subtotals do not sum to 601 for number of patients due to missing data.

b

P values were calculated using the log-rank test.

ADT, androgen-deprivation therapy; PCSM, prostate cancer-specific mortality; ACM, all-cause mortality; PSA, prostate-specific antigen; IQR, interquartile range.

SNP selection and genotyping

Since transcription factors have been known to regulate different genes in different cellular contexts [16], we used PReMod (genomequebec.mcgill.ca/PReMod), a genome-wide cis-regulatory module prediction database, to identify computationally all putative AREs in the human genome [17]. PReMod used TRANSFAC version 7.2 position weight matrices (PWMs) to score putative transcription factor-binding sites based on how faithfully the binding site in human and its orthologs 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 factors' binding sites [18] and thus has proven to be more reliable than other algorithms. The PReMod algorithm predicts that a total of 6590 sites within the human genome are bound by the AR (canonical AR PWM: M00481; consensus: GGTACANNRTGTTCT) [19]. We identified SNPs within AREs by comparing two hexameric half-sites of these putative AREs with HapMap SNPs CHB (Han Chinese in Beijing, China) data in the UCSC table browser (NCBI35/hg17) [20, 21]. SNPs with a minor allele frequency <10% in the HapMap CHB population were excluded, thus leaving 59 SNPs in AREs were initially selected for analysis.

Genomic DNA was extracted from peripheral blood using the QIAamp DNA Blood Mini Kit (Qiagen, Valencia, CA) and stored at −80°C until the time of study. Genotyping was carried out as described previously [11] using Sequenom iPLEX matrix-assisted laser desorption/ionization–time of flight (MALDI–TOF) 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% among 54 blind duplicated quality control samples. Any SNP that did not conform to the Hardy–Weinberg equilibrium (P < 0.005), or fell below a genotyping call rate of 90%, was removed (N = 4). Thus, a total of 55 SNPs were selected for further statistical analysis.

statistical analysis

Patients’ clinicopathologic characteristics were summarized as the number and percentage of patients or the median and interquartile range of values. The continuous factors were dichotomized at the median value within the cohort, with the exception of the PSA nadir, which was dichotomized at 0.2 ng/ml because of its correlation with disease progression and PCSM [5, 6]. 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. The associations of 55 individual SNPs and clinicopathologic characteristics with time to progression, PCSM, and ACM were assessed using the Kaplan–Meier analysis with log-rank test. Multivariate analyses to determine the interdependency of genotypes and other known prognostic factors, such as age at diagnosis, clinical stage, Gleason score, PSA at ADT initiation, PSA nadir, time to PSA nadir, and treatment modality, were carried out using the Cox proportional hazards regression model. Based on our study population (N = 601), we were able to detect a hazard ratio of 1.6, for a minor allele frequency of 0.2, with over 0.8 statistical power for disease progression, PCSM, and ACM. As we were testing 55 SNPs, false-discovery rates (q values) were calculated to determine the degree to which the tests for association were prone to false positives [22]. The q value estimates the proportion of results declared interesting (i.e. significant at the nominal P value of 0.05) that is actually false. For example, we carried out 55 statistical tests to assess the associations between SNPs and PCSM with an α of 0.05 and then about three tests on average would produce P <0.05 by chance alone. Five SNPs were associated with PCSM with nominal P <0.05 and all had q ≤0.108, indicating that of all the results declared interesting (five SNPs), 10.8% of these were likely to be false (less than one false positive). No thresholds for the selection of a particular q value have been established in candidate gene investigations. However, it has been suggested that a threshold <0.20 would be considered reasonable [23]. The q values were estimated using R q value package (http://genomics.princeton.edu/storeylab/qvalue/) to analyze the observed distribution of P values from the log-rank test for 55 SNPs. The Statistical Package for the Social Sciences software version 16.0.1 (SPSS Inc., Chicago, IL) was used for other statistical analyses. A two-sided P value of <0.05 was considered statistically significant.

results

Univariate log-rank tests of the 55 SNPs in AREs regarding time to progression, PCSM, and ACM after ADT were summarized in supplemental Table 1, available at Annals of Oncology online. In the first analysis, ACTN2 rs4659711 and NR2F1 rs10074743 were associated with disease progression (nominal P ≤ 0.044), and all had a false-discovery rate (q value) of ≤0.562 (Table 2). Five SNPs were associated with PCSM with nominal P ≤0.025 and q ≤0.108 (ARRDC3 rs2939244, XRCC6BP1 rs12824706, FLT1 rs9508016, PSMD7 rs328318, and SKAP1 rs6504145; Table 3). Four SNPs, FBXO32 rs7830622, XRCC6BP1 rs12824706, FLT1 rs9508016, and FLRT3 rs4140562, exhibited significant effects on ACM (nominal P ≤ 0.038), and all had a q value of ≤0.243 (Table 4).

Table 2.

Genotyping frequencies and the association of genotype with disease progression during ADT

Gene SNP Genotype No. of patients No. of events Median (months) Pq HR (95% CI) P
ACTN2 CC/CT 498 347 21 0.044 0.562 1.00  
    rs4659711 TT 83 53 26   0.98 (0.72–1.33) 0.897 
NR2F1 CC 516 350 24 0.004 0.220 1.00  
    rs10074743 CT/TT 78 62 16   1.30 (0.98–1.74) 0.069 
Gene SNP Genotype No. of patients No. of events Median (months) Pq HR (95% CI) P
ACTN2 CC/CT 498 347 21 0.044 0.562 1.00  
    rs4659711 TT 83 53 26   0.98 (0.72–1.33) 0.897 
NR2F1 CC 516 350 24 0.004 0.220 1.00  
    rs10074743 CT/TT 78 62 16   1.30 (0.98–1.74) 0.069 

P < 0.05 are in boldface.

a

P values were calculated using the log-rank test.

b

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

ADT, androgen-deprivation therapy; SNP, single-nucleotide polymorphism; HR, hazard ratio; 95% CI, 95% confidence interval; PSA, prostate-specific antigen.

Table 3.

Genotyping frequencies and the association of genotype with PCSM during ADT

Gene SNP Genotype No. of patients No. of events Estimated mean (months) Pq HR (95% CI) P
ARRDC3 AA/AT 487 74 137 0.012 0.065 1.00  
    rs2939244 TT 111 27 124   2.07 (1.29–3.31) 0.002 
XRCC6BP1 CC/CT 491 92 134 0.004 0.043 1.00  
    rs12824706 TT 105 148   0.51 (0.24–1.07) 0.075 
FLT1 GG 462 67 143 0.004 0.043 1.00  
    rs9508016 GA/AA 135 33 114   1.82 (1.17–2.83) 0.008 
PSMD7 AA 238 52 125 0.011 0.065 1.00  
    rs328318 AG/GG 359 48 145   0.82 (0.55–1.23) 0.335 
SKAP1 CC 439 83 123 0.025 0.108 1.00  
    rs6504145 CT/TT 159 17 152   0.55 (0.32–0.96) 0.035 
No. of unfavorable genotypes presentc        
    0  102 150 <0.001  1.00  
    1  330 49 142   1.69 (0.81–3.53) 0.166 
    2  147 34 114   3.12 (1.45–6.73) 0.004 
    3  21 65   6.05 (2.30–15.9) <0.001 
Ptrend <0.001 
Gene SNP Genotype No. of patients No. of events Estimated mean (months) Pq HR (95% CI) P
ARRDC3 AA/AT 487 74 137 0.012 0.065 1.00  
    rs2939244 TT 111 27 124   2.07 (1.29–3.31) 0.002 
XRCC6BP1 CC/CT 491 92 134 0.004 0.043 1.00  
    rs12824706 TT 105 148   0.51 (0.24–1.07) 0.075 
FLT1 GG 462 67 143 0.004 0.043 1.00  
    rs9508016 GA/AA 135 33 114   1.82 (1.17–2.83) 0.008 
PSMD7 AA 238 52 125 0.011 0.065 1.00  
    rs328318 AG/GG 359 48 145   0.82 (0.55–1.23) 0.335 
SKAP1 CC 439 83 123 0.025 0.108 1.00  
    rs6504145 CT/TT 159 17 152   0.55 (0.32–0.96) 0.035 
No. of unfavorable genotypes presentc        
    0  102 150 <0.001  1.00  
    1  330 49 142   1.69 (0.81–3.53) 0.166 
    2  147 34 114   3.12 (1.45–6.73) 0.004 
    3  21 65   6.05 (2.30–15.9) <0.001 
Ptrend <0.001 

P < 0.05 are in boldface.

a

P values were calculated using the log-rank test.

b

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

c

Unfavorable genotypes refer to TT in ARRDC3 rs2939244, GA/AA in FLT1 rs9508016, and CC in SKAP1 rs6504145.

PCSM, prostate cancer-specific mortality; ADT, androgen-deprivation therapy; SNP, single-nucleotide polymorphism; HR, hazard ratio; 95% CI, 95% confidence interval; PSA, prostate-specific antigen.

Table 4.

Genotyping frequencies and the association of genotype with ACM during ADT

Gene SNP Genotype No. of patients No. of events Estimated mean (months) Pq HR (95% CI) P
FBXO32 TT 441 98 121 0.027 0.230 1.00  
    rs7830622 TC/CC 158 46 112   1.69 (1.18–2.43) 0.004 
XRCC6BP1 CC/CT 491 126 119 0.038 0.243 1.00  
    rs12824706 TT 105 18 131   0.82 (0.49–1.36) 0.446 
FLT1 GG 462 101 127 0.012 0.230 1.00  
    rs9508016 GA/AA 135 43 101   1.52 (1.05–2.21) 0.028 
FLRT3 GG 297 59 131 0.027 0.230 1.00  
    rs4140562 GT/TT 299 86 106   1.29 (0.91–1.82) 0.153 
No. of unfavorable genotypes presentc      
    0  350 74 123 <0.001  1.00  
    1  207 53 119   1.25 (0.86–1.80) 0.243 
    2  43 18 62   3.33 (1.94–5.72) <0.001 
Ptrend <0.001 
Gene SNP Genotype No. of patients No. of events Estimated mean (months) Pq HR (95% CI) P
FBXO32 TT 441 98 121 0.027 0.230 1.00  
    rs7830622 TC/CC 158 46 112   1.69 (1.18–2.43) 0.004 
XRCC6BP1 CC/CT 491 126 119 0.038 0.243 1.00  
    rs12824706 TT 105 18 131   0.82 (0.49–1.36) 0.446 
FLT1 GG 462 101 127 0.012 0.230 1.00  
    rs9508016 GA/AA 135 43 101   1.52 (1.05–2.21) 0.028 
FLRT3 GG 297 59 131 0.027 0.230 1.00  
    rs4140562 GT/TT 299 86 106   1.29 (0.91–1.82) 0.153 
No. of unfavorable genotypes presentc      
    0  350 74 123 <0.001  1.00  
    1  207 53 119   1.25 (0.86–1.80) 0.243 
    2  43 18 62   3.33 (1.94–5.72) <0.001 
Ptrend <0.001 

P < 0.05 are in boldface.

a

P values were calculated using the log-rank test.

b

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

c

Unfavorable genotypes refer to TC/CC in FBXO32 rs7830622 and GA/AA in FLT1 rs9508016.

ACM, all-cause mortality; ADT, androgen-deprivation therapy; SNP, single-nucleotide polymorphism; HR, hazard ratio; 95% CI, 95% confidence interval; PSA, prostate-specific antigen.

To assess the predictive effects of these SNPs beyond the clinical features to influence outcomes after ADT, we carried out a multivariate analysis, adjusting for age at diagnosis, clinical stage, Gleason score, PSA level at ADT initiation, PSA nadir, time to PSA nadir, and treatment modality. After adjusting for these predictors, no statistical association was observed between any SNPs and disease progression (P > 0.050; Table 2). However, three (ARRDC3 rs2939244, FLT1 rs9508016, and SKAP1 rs6504145) and two (FBXO32 rs7830622 and FLT1 rs9508016) polymorphisms remained as significant predictors for time to PCSM and ACM, respectively (P ≤ 0.035; Tables 3 and 4).

A gene–dosage effect on PCSM and ACM was found when three genetic loci of interest related to PCSM and two genetic loci of interest associated with ACM were analyzed in combination (log-rank P < 0.001, Tables 3 and 4 and Figure 1, left). The time to PCSM and ACM decreased as the number of unfavorable genotypes increased, and the combined genotype remained as a significant predictor after adjusting for other clinical factors (P for trend < 0.001, Tables 3 and 4).

Figure 1.

Kaplan–Meier curves of (A) time to PCSM during ADT for patients with zero, one, two, or three unfavorable genotypes at the three genetic loci of interest and (B) time to ACM during ADT for patients with zero, one, or two unfavorable genotypes at the two genetic loci of interest, measured in all patients (left), in patients without distant metastasis (middle), or in patients with distant metastases (right). Numbers in parentheses indicate the number of patients.

Figure 1.

Kaplan–Meier curves of (A) time to PCSM during ADT for patients with zero, one, two, or three unfavorable genotypes at the three genetic loci of interest and (B) time to ACM during ADT for patients with zero, one, or two unfavorable genotypes at the two genetic loci of interest, measured in all patients (left), in patients without distant metastasis (middle), or in patients with distant metastases (right). Numbers in parentheses indicate the number of patients.

To further evaluate the clinical relevance of these SNPs in AREs, we risk-stratified patients with respect to their metastatic status at the initiation of ADT. The combined genotypes still had significant effects on PCSM in patients with or without distant metastases (P ≤ 0.019, Figure 1A, middle and right). This additional information further supports an enhancement of risk prediction for PCSM in both low- and high-risk patients, whereas the incorporation of combined genotypes only leads to better risk prediction for ACM in the high-risk patients with distant metastases (P < 0.001, Figure 1B, right), not in patients without metastases (P = 0.253, Figure 1B, middle).

discussion

It has been well documented that AR mediates essential processes in normal and cancerous prostate development and still persists in the majority of patients with hormone refractory disease [24–26]. The classical activation mechanism of AR involves the binding of androgen and other coregulators to form the androgen–AR complex, which binds to AREs in the promoter regions of androgen target genes, regulating their transcription and consequently mediating androgen’s physiological or tumorigenic effects. Thus, sequence variants within AREs might affect AR–ARE interactions, and it is meaningful to assess the associations between common genetic variants in AREs and prostate cancer progression. The SNPs we genotyped were based on genome-wide-predicted AREs and, therefore, provided a unique opportunity to examine comprehensively putative AREs without depending on a prior hypothesis. Our study took advantage of these hybrid designs of candidate androgen-responsive genes and genome-wide approaches. Furthermore, this study is one of the largest carried out in advanced prostate cancer for which ADT was the main therapeutic intervention. We found that 4 of 55 SNPs in AREs were significantly associated with PCSM or ACM in prostate cancer patients receiving ADT, remaining so in multivariate analyses after controlling for other known clinicopathologic risk factors (age, clinical stage, Gleason score, PSA level at ADT initiation, PSA nadir, time to PSA nadir, and treatment modality). Moreover, strong combined genotype effects on PCSM and ACM were also observed. To our knowledge, no previous studies have investigated these associations.

Of the 55 SNPs evaluated, ARRDC3 rs2939244, FLT1 rs9508016, SKAP1 rs6504145, and FBXO32 rs7830622 exhibited significant associations with PCSM or ACM after adjusting for all clinical predictors. Of these, FLT1 rs9508016 was significantly associated with both PCSM and ACM during ADT. The closest known gene to rs2939244 is the arrestin domain-containing 3 (ARRDC3) gene. Although rs2939244 is located 395-kb upstream of ARRDC3, rs2939244 could still regulate ARRDC3 expression through enhancers, which have been demonstrated to be able to regulate target genes from >1 Mb away. ARRDC3 has been classified as an α-arrestin, which contains structural homology to the arrestin family, suggesting a role in the internalization and regulation of membrane proteins [27]. A recent study has found that ARRDC3 is preferentially lost in some breast cancers, and overexpression of ARRDC3 repressed cancer cell proliferation, migration, invasion, and in vivo tumorigenicity through the degradation of a cell surface adhesion molecule, β-4 integrin, associated with aggressive tumor behavior [28]. ARRDC3 was up-regulated in the muscle cells of male AR knockout mice [29], implying that androgens might negatively regulate ARRDC3 expression and increase the risk of aggressive prostate cancer. rs9508016 is intronic to FLT1, fms-related tyrosine kinase 1, which encodes a member of the vascular endothelial growth factor receptor (VEGFR) family. Vascular endothelial growth factor (VEGF) is vital to physiological and pathological angiogenesis, largely by activating its receptors, FLT1 and kinase insert domain receptor (KDR). Most biological functions of VEGF are mediated via KDR, whereas the role of FLT1 remains to be clarified. Recently, the expression of FLT1 has been correlated with a higher Gleason grade, pathological stage, and microvessel density in the premalignant and malignant tissues when compared with normal prostatic glands [30]. Consistent with the prediction of ARE in FLT1, a dose-dependent increase in the expression of FLT1 by androgen treatment implies that FLT1 might be an androgen target gene [31]. Several small molecule inhibitors targeting VEGFRs have been shown to exhibit some activity in prostate cancer, suggesting that the use of antiangiogenesis agents in prostate cancer might be promising [32, 33]. rs6504145 is an intronic SNP in src kinase-associated phosphoprotein 1 (SKAP1) that has strong homology to the src oncogene at the C-terminal end. SKAP1 was shown to be involved in the regulation of the Ras oncogenic pathway and mitotic progression, specifically at the metaphase-to-anaphase transition [34, 35]. A genome-wide association study also identified an ovarian cancer susceptibility locus in SKAP1 [36]. rs7830622 is located in the intron of the F-box protein 32 (FBXO32) gene, which encodes a member of the F-box protein family and constitutes a subunit of the phosphorylation-dependent ubiquitin protein ligase complex [37]. FBXO32 was originally reported to be involved in muscle atrophy, but recent findings suggest that FBXO32 is a novel apoptosis regulator [38]. FBXO32 was downregulated in some ovarian cancer cell lines, and restoration of FBXO32 sensitized platinum-resistant ovarian cancer cells to cisplatin by upregulating apoptosis. Furthermore, ovarian cancer patients with high FBXO32 methylation levels exhibited significantly shorter progression-free survival times than did those with low methylation levels [39]. The expression level of FBXO32 was decreased in mice after orchidectomization, and its expression could be restored by both estrogen and androgen treatments of orchidectomized mice [40], indicating that FBXO32 might be an androgen-regulated gene.

Taken together, using a genome-wide approach to study the common variants in AREs, our research reveals multiple unappreciated pathways affecting survival after ADT, such as ARRDC3 in membrane receptor desensitization, FLT1 in angiogenesis, SKAP1 in cell cycle regulation, and FBXO32 in apoptosis. The genetic variants identified in this study might not only encourage further investigation of the implicated signaling pathways on ADT but also produce clinically useful biomarkers. However, the current findings are hypothesis generating, and the results reported here are limited by analyzing the small number of patients in a genetic subset and multiple comparisons. In addition, our homogeneous Chinese Han population might make our findings less applicable to other ethnic groups. Larger studies in the future are warranted to confirm our findings in other ethnical cohorts, and more thorough functional analyses will clarify the role of these SNPs/genes in the course of prostate cancer progression.

funding

National Science Council, Taiwan (NSC-98-2320-B-039-019-MY3 to BB and NSC-99-2314-B-037-018-MY3 to SH).

disclosure

The authors have declared no conflicts of interest.

We thank the National Genotyping Center of National Research Program for Genomic Medicine, National Science Council, Taiwan, for technical support.

References

1.
Heinlein
CA
Chang
C
Androgen receptor in prostate cancer
Endocr Rev
 , 
2004
, vol. 
25
 (pg. 
276
-
308
)
2.
Pienta
KJ
Bradley
D
Mechanisms underlying the development of androgen-independent prostate cancer
Clin Cancer Res
 , 
2006
, vol. 
12
 (pg. 
1665
-
1671
)
3.
Walczak
JR
Carducci
MA
Prostate cancer: a practical approach to current management of recurrent disease
Mayo Clin Proc
 , 
2007
, vol. 
82
 (pg. 
243
-
249
)
4.
Choueiri
TK
Xie
W
D'Amico
AV
, et al.  . 
Time to prostate-specific antigen nadir independently predicts overall survival in patients who have metastatic hormone-sensitive prostate cancer treated with androgen-deprivation therapy
Cancer
 , 
2009
, vol. 
115
 (pg. 
981
-
987
)
5.
Hussain
M
Tangen
CM
Higano
C
, et al.  . 
Absolute prostate-specific antigen value after androgen deprivation is a strong independent predictor of survival in new metastatic prostate cancer: data from Southwest Oncology Group Trial 9346 (INT-0162)
J Clin Oncol
 , 
2006
, vol. 
24
 (pg. 
3984
-
3990
)
6.
Stewart
AJ
Scher
HI
Chen
MH
, et al.  . 
Prostate-specific antigen nadir and cancer-specific mortality following hormonal therapy for prostate-specific antigen failure
J Clin Oncol
 , 
2005
, vol. 
23
 (pg. 
6556
-
6560
)
7.
Bao
BY
Pao
JB
Huang
CN
, et al.  . 
Polymorphisms inside microRNAs and microRNA target sites predict clinical outcomes in prostate cancer patients receiving androgen-deprivation therapy
Clin Cancer Res
 , 
2011
, vol. 
17
 (pg. 
928
-
936
)
8.
Huang
SP
Chou
YH
Wayne Chang
WS
, et al.  . 
Association between vitamin D receptor polymorphisms and prostate cancer risk in a Taiwanese population
Cancer Lett
 , 
2004
, vol. 
207
 (pg. 
69
-
77
)
9.
Huang
SP
Huang
CY
Wang
JS
, et al.  . 
Prognostic significance of p53 and X-ray repair cross-complementing group 1 polymorphisms on prostate-specific antigen recurrence in prostate cancer post radical prostatectomy
Clin Cancer Res
 , 
2007
, vol. 
13
 (pg. 
6632
-
6638
)
10.
Huang
SP
Huang
CY
Wu
WJ
, et al.  . 
Association of vitamin D receptor FokI polymorphism with prostate cancer risk, clinicopathological features and recurrence of prostate specific antigen after radical prostatectomy
Int J Cancer
 , 
2006
, vol. 
119
 (pg. 
1902
-
1907
)
11.
Huang
SP
Huang
LC
Ting
WC
, et al.  . 
Prognostic significance of prostate cancer susceptibility variants on prostate-specific antigen recurrence after radical prostatectomy
Cancer Epidemiol Biomarkers Prev
 , 
2009
, vol. 
18
 (pg. 
3068
-
3074
)
12.
Huang
SP
Ting
WC
Chen
LM
, et al.  . 
Association analysis of Wnt pathway genes on prostate-specific antigen recurrence after radical prostatectomy
Ann Surg Oncol
 , 
2010
, vol. 
17
 (pg. 
312
-
322
)
13.
Huang
SP
Wu
WJ
Chang
WS
, et al.  . 
p53 Codon 72 and p21 codon 31 polymorphisms in prostate cancer
Cancer Epidemiol Biomarkers Prev
 , 
2004
, vol. 
13
 (pg. 
2217
-
2224
)
14.
Kwak
C
Jeong
SJ
Park
MS
, et al.  . 
Prognostic significance of the nadir prostate specific antigen level after hormone therapy for prostate cancer
J Urol
 , 
2002
, vol. 
168
 (pg. 
995
-
1000
)
15.
Ross
RW
Oh
WK
Xie
W
, et al.  . 
Inherited variation in the androgen pathway is associated with the efficacy of androgen-deprivation therapy in men with prostate cancer
J Clin Oncol
 , 
2008
, vol. 
26
 (pg. 
842
-
847
)
16.
Zeitlinger
J
Simon
I
Harbison
CT
, et al.  . 
Program-specific distribution of a transcription factor dependent on partner transcription factor and MAPK signaling
Cell
 , 
2003
, vol. 
113
 (pg. 
395
-
404
)
17.
Ferretti
V
Poitras
C
Bergeron
D
, et al.  . 
PReMod: a database of genome-wide mammalian cis-regulatory module predictions
Nucleic Acids Res
 , 
2007
, vol. 
35
 (pg. 
D122
-
D126
)
18.
Bulyk
ML
Computational prediction of transcription-factor binding site locations
Genome Biol
 , 
2003
, vol. 
5
 pg. 
201
 
19.
Roche
PJ
Hoare
SA
Parker
MG
A consensus DNA-binding site for the androgen receptor
Mol Endocrinol
 , 
1992
, vol. 
6
 (pg. 
2229
-
2235
)
20.
Karolchik
D
Hinrichs
AS
Furey
TS
, et al.  . 
The UCSC Table Browser data retrieval tool
Nucleic Acids Res
 , 
2004
, vol. 
32
 (pg. 
D493
-
D496
)
21.
Thorisson
GA
Smith
AV
Krishnan
L
, et al.  . 
The International HapMap Project Web site
Genome Res
 , 
2005
, vol. 
15
 (pg. 
1592
-
1593
)
22.
Storey
JD
Tibshirani
R
Statistical significance for genomewide studies
Proc Natl Acad Sci U S A
 , 
2003
, vol. 
100
 (pg. 
9440
-
9445
)
23.
Benjamini
Y
Yekutieli
D
Quantitative trait Loci analysis using the false discovery rate
Genetics
 , 
2005
, vol. 
171
 (pg. 
783
-
790
)
24.
Sadi
MV
Walsh
PC
Barrack
ER
Immunohistochemical study of androgen receptors in metastatic prostate cancer. Comparison of receptor content and response to hormonal therapy
Cancer
 , 
1991
, vol. 
67
 (pg. 
3057
-
3064
)
25.
van der Kwast
TH
Schalken
J
Ruizeveld de Winter
JA
, et al.  . 
Androgen receptors in endocrine-therapy-resistant human prostate cancer
Int J Cancer
 , 
1991
, vol. 
48
 (pg. 
189
-
193
)
26.
Chodak
GW
Kranc
DM
Puy
LA
, et al.  . 
Nuclear localization of androgen receptor in heterogeneous samples of normal, hyperplastic and neoplastic human prostate
J Urol
 , 
1992
, vol. 
147
 (pg. 
798
-
803
)
27.
Alvarez
CE
On the origins of arrestin and rhodopsin
BMC Evol Biol
 , 
2008
, vol. 
8
 pg. 
222
 
28.
Draheim
KM
Chen
HB
Tao
Q
, et al.  . 
ARRDC3 suppresses breast cancer progression by negatively regulating integrin beta4
Oncogene
 , 
2010
, vol. 
29
 (pg. 
5032
-
5047
)
29.
MacLean
HE
Chiu
WS
Notini
AJ
, et al.  . 
Impaired skeletal muscle development and function in male, but not female, genomic androgen receptor knockout mice
FASEB J
 , 
2008
, vol. 
22
 (pg. 
2676
-
2689
)
30.
Pallares
J
Rojo
F
Iriarte
J
, et al.  . 
Study of microvessel density and the expression of the angiogenic factors VEGF, bFGF and the receptors Flt-1 and FLK-1 in benign, premalignant and malignant prostate tissues
Histol Histopathol
 , 
2006
, vol. 
21
 (pg. 
857
-
865
)
31.
Sieveking
DP
Lim
P
Chow
RW
, et al.  . 
A sex-specific role for androgens in angiogenesis
J Exp Med
 , 
2010
, vol. 
207
 (pg. 
345
-
352
)
32.
Dror Michaelson
M
Regan
MM
Oh
WK
, et al.  . 
Phase II study of sunitinib in men with advanced prostate cancer
Ann Oncol
 , 
2009
, vol. 
20
 (pg. 
913
-
920
)
33.
Ryan
CJ
Stadler
WM
Roth
B
, et al.  . 
Phase I dose escalation and pharmacokinetic study of AZD2171, an inhibitor of the vascular endothelial growth factor receptor tyrosine kinase, in patients with hormone refractory prostate cancer (HRPC)
Invest New Drugs
 , 
2007
, vol. 
25
 (pg. 
445
-
451
)
34.
Fang
L
Seki
A
Fang
G
SKAP associates with kinetochores and promotes the metaphase-to-anaphase transition
Cell Cycle
 , 
2009
, vol. 
8
 (pg. 
2819
-
2827
)
35.
Kosco
KA
Cerignoli
F
Williams
S
, et al.  . 
SKAP55 modulates T cell antigen receptor-induced activation of the Ras-Erk-AP1 pathway by binding RasGRP1
Mol Immunol
 , 
2008
, vol. 
45
 (pg. 
510
-
522
)
36.
Goode
EL
Chenevix-Trench
G
Song
H
, et al.  . 
A genome-wide association study identifies susceptibility loci for ovarian cancer at 2q31 and 8q24
Nat Genet
 , 
2010
, vol. 
42
 (pg. 
874
-
879
)
37.
Li
HH
Kedar
V
Zhang
C
, et al.  . 
Atrogin-1/muscle atrophy F-box inhibits calcineurin-dependent cardiac hypertrophy by participating in an SCF ubiquitin ligase complex
J Clin Invest
 , 
2004
, vol. 
114
 (pg. 
1058
-
1071
)
38.
Tan
J
Yang
X
Zhuang
L
, et al.  . 
Pharmacologic disruption of Polycomb-repressive complex 2-mediated gene repression selectively induces apoptosis in cancer cells
Genes Dev
 , 
2007
, vol. 
21
 (pg. 
1050
-
1063
)
39.
Chou
JL
Su
HY
Chen
LY
, et al.  . 
Promoter hypermethylation of FBXO32, a novel TGF-beta/SMAD4 target gene and tumor suppressor, is associated with poor prognosis in human ovarian cancer
Lab Invest
 , 
2010
, vol. 
90
 (pg. 
414
-
425
)
40.
Svensson
J
Moverare-Skrtic
S
Windahl
S
, et al.  . 
Stimulation of both estrogen and androgen receptors maintains skeletal muscle mass in gonadectomized male mice but mainly via different pathways
J Mol Endocrinol
 , 
2010
, vol. 
45
 (pg. 
45
-
57
)

Author notes

These authors contributed equally as senior authors.