The Pharmacogenomics Journal (2004) 4, 251–259 & 2004 Nature Publishing Group All rights reserved 1470-269X/04 $30.00 www.nature.com/tpj ORIGINAL ARTICLE

A case-based evaluation of SRD5A1, SRD5A2, AR, and ADRA1A as candidate for severity of BPH

M Klotsman1,2 ABSTRACT 3 In men with a clinical diagnosis of benign prostatic hyperplasia (BPH), CR Weinberg polytomous logistic regression analysis was conducted to evaluate associa- 1,4 K Davis tions between two silent polymorphisms in SRD5A1 (codon positions 30 and CG Binnie5 116), two polymorphisms in SRD5A2 (Val89Leu substitution and C to T KE Hartmann1 transition in intron 1), a trinucleotide (CAG)n repeat in receptor (AR), and an Arg492Cys substitution in ADRA1A and clinical parameters that 1Department of Epidemiology, University of characterize severity of BPH. Candidate selection was based on two North Carolina School of Public Health, Chapel mechanistic pathways targeted by pharmacotherapy for BPH: (1) androgen 2 Hill, NC, USA; Genetics Research, metabolic loci contributing to growth (static obstruction); and (2) GlaxoSmithKline, RTP, NC, USA; 3Biostatistics Branch, National Institute of Environmental factors affecting smooth muscle tone (dynamic obstruction). Polymorphisms Health Sciences, RTP, NC, USA; 4World-Wide in SRD5A2 were not associated with severity of BPH; however, SRD5A1 Epidemiology, GlaxoSmithKline, RTP, NC, USA; polymorphisms were associated with severity of BPH. The process(es) in 5 Discovery Genetics, GlaxoSmithKline, RTP, NC, which these silent single-nucleotide polymorphisms (SNPs) influence BPH USA phenotypes is unknown and additional studies will be needed to assess Correspondence: whether these SNPs have direct functional consequences. The characteriza- Dr M Klotsman, PO Box 13398, Five Moore tion of additional molecular factors that contribute to static and dynamic Dr, RTP, NC 27709-3398, USA. obstruction may help predict response to pharmacotherapy and serve to Tel: þ 1 919 483 5249 identify novel drug targets for the clinical management of BPH. Fax: þ 1 919 315 0311 E-mail: [email protected] The Pharmacogenomics Journal (2004) 4, 251–259. doi:10.1038/sj.tpj.6500248 Published online 11 May 2004

Keywords: BPH; prostate; 5-alpha-; adrenergic alpha-an- tagonists

INTRODUCTION Benign prostatic hyperplasia (BPH) is an extremely common condition among aging men that typically manifests as a constellation of lower urinary tract symptoms (LUTS) which are bothersome, impair psychological and functional well-being, and interfere with activities of daily living. Prevalence estimates for BPH vary, but it is clear that BPH is nearly ubiquitous in older men and is a significant health concern.1 Histologic evidence of BPH was found in almost 90% of men in their 80s,2 and it is estimated that nearly one in four men receive treatment for LUTS by the eighth decade of life.3 Despite significant morbidity caused by BPH, the etiology and determinants of severity of this condition remain poorly understood. Received: 24 November 2003 The current BPH literature suggests that LUTS generally result from static Revised: 27 January 2004 Accepted: 19 February 2004 (mechanical) and/or dynamic component(s). The static component is a Published online 11 May 2004 consequence of the benign enlargement of the prostate, which increases pressure Evaluation of candidate genes for severity of BPH M Klotsman et al 252

on the urethra with subsequent obstruction of urine flow. RESULTS Prostate growth is regulated, in part, by a defined androgen- Baseline Distributions dependent pathway,4 which begins with the diffusion of Genetic data were available for 87% of the study participants serum testosterone (TES) into the prostate. In the prostatic (n ¼ 2428 out of 2802) for whom baseline data were available stroma, TES is irreversibly converted to the more active (Table 1). Genotypes were missing primarily because of (DHT) by the 5a-reductase type failed assays or subject refusal to submit a blood sample for II. Converted DHT then enters epithelial cells and selectively genetic analysis. The baseline characteristics for the subset binds to the androgen receptor (AR), a ligand-dependent of men in whom genetic data were unavailable appeared transcription factor, forming a DHT/AR complex. This similar to those of the study group as a whole (data not complex is translocated into the nucleus and promotes shown). prostatic growth by binding to androgen response elements in the regulatory region of target genes5,6 (eg, androgen regulation of prostate-specific antigen (PSA) expression is mediated by AR).7 Table 1 Baseline clinical parameters and genotype distribu- By contrast, dynamic obstruction is caused by dysfunction tions among 2201 Caucasians of smooth muscle tone in the prostatic capsule surrounding Parameter Mean (7SD) the prostate. Prostatic smooth muscle tone varies according to the sympathetic stimulation of a -adrenoceptors, in 1 Age (yr) 66.577.6 a 8–10 particular, the 1A-adrenoceptor subtype. In a healthy PSA (ng/ml) 3.872.0 male, smooth muscle in the capsule contracts to propel AUA-SI 17.276.0 7 prostatic secretions from storage into the prostatic urethra Qmax (ml/s) 10.4 3.6 during ejaculation. In a BPH disease state, increased smooth Post-void residual volume 79775 muscle tone in the outer prostatic capsule is believed to Total prostate volume (cm3) 54.9723.0 bring on LUTS by increasing urethral resistance. Conse- Transition zone volume (cm3) 26.8717.0 quently, the dynamic component of BPH is determined, at Serum testosterone (pg/ml) 392871489 7 least partially, by a -adrenoceptor activity and prostatic Serum dihydrotestosterone 404 202 1A 7 smooth muscle tone.11 Body mass index 27.8 4.1 Systematic screens of genes encoding for 5a-reductase SRD5A1: Silent Arg 30a type I and II (SRD5A1 and SRD5A2), AR, and a -adreno- 1A G/G 625 (28.6%) ceptor (ADRA1A) have identified polymorphisms that may G/C 1104 (50.6%) influence protein production and/or integrity. These poly- C/C 454 (20.8%) morphic loci can be operationally compartmentalized into one of two pathways that contribute to the severity of BPH: SRD5A1: Silent Ala 116a (1) an androgen-dependent pathway, involving 5a-reductase A/A 874 (40.0%) and AR activity, contributing to static obstruction; and (2) a A/G 1004 (46.1%) G/G 304 (13.9%) non-growth-dependent pathway, mediated by a1A-adreno- ceptor, that may be instrumental in the onset of dynamic b obstruction. The polymorphisms evaluated in this study SRD5A2: Val89Leu Val/Val 1110 (50.9%) included two silent polymorphisms in SRD5A1 (codon 12 Val/Leu 885 (40.6%) positions 30 and 116), two polymorphisms in SRD5A2 Leu/Leu 184 (8.4%) 13,14 (Val89Leu and C to T transition in intron 1), a (CAG)n 15,16 repeat in AR, and Arg492Cys substitution in SRD5A2: Intron 1 C-Tb ADRA1A.17,18 C/C 1065 (49.7%) Pharmacotherapy with a-blockers and 5a-reductase inhi- C/T 880 (41.1%) bitors has become commonplace for the medical manage- T/T 196 (9.2%) ment of BPH.19 Alpha blockers alleviate dynamic obstruction by interrupting motor sympathetic adrenergic ADRA1A: Arg492Leu nerve stimulation in the prostate,20 while 5a-reductase Arg/Arg 630 (28.7%) Arg/Cys 1099 (50.1%) inhibitors limit DHT production resulting in involution of Cys/Cys 464 (21.2%) the prostate.21 Given the importance of a-blockers and 5a- reductase inhibitors in the treatment of BPH, this study was AR: (CAG)n repeats undertaken to evaluate associations between polymorph- (CAG)0–18 82 (6.0%) isms in candidate genes whose effects are targeted by (CAG)19–22 970 (43.9%) pharmacotherapy and clinical parameters used to character- (CAG)423 1103 (50.1%) ize the severity of BPH. Further characterization of poly- a morphic loci hypothesized to play a role in static and The SRD5A1 Arg30 and Ala116 SNPs were not strongly linked (correlation, R ¼ 0.6775). dynamic obstruction precipitated by BPH may help to shed bThe SRD5A2 V89L and Intron 1 C-T SNPs appear to be in linkage disequilibrium light on etiologic mechanisms predisposing men to BPH. (correlation, R ¼ 1).

The Pharmacogenomics Journal Evaluation of candidate genes for severity of BPH M Klotsman et al 253

Allele and Genotype Frequencies tion were observed (Table 2). Chi-square statistics were re- Regression analyses performed to evaluate whether candi- calculated for men under the age of 60, to allow for the date genetic markers are associated with severity of BPH possibility that gene-by-gene interaction contributes to an were restricted to Caucasian subjects. The two single- earlier age of onset. In this younger group of men, statistical nucleotide polymorphisms (SNPs) in 5a-reductase type II significance was not reached for any of the gene-by-gene (SRD5A2) are physically close (intron 1 and exon 1) and pairs evaluated (data not shown). were in linkage disequilibrium (LD) (correlation, R ¼ 1). Results based on fitting logistic regression models to relate Concordant SRD5A2 genotype pairs were observed in 93% of three-level genotypes at each polymorphic loci and BPH- study participants, and predominant allele frequencies were related clinical parameters are shown in Table 3. Polymorph- nearly identical (71.2% and 70.3% for 89-Val and intron-C isms in 5a-reductase type II (SRD5A2) did not appear to be alleles, respectively). The SNPs in SRD5A1 were in partial LD associated with the severity of BPH; however, polymorph- (correlation, R ¼ 0.6775). These LD estimates are based on a isms in 5a-reductase type I were associated with severity. case-only population and may not be comparable to LD one The silent Arg30 polymorphism in 5a-reductase type I would observe in the general population. (SRD5A1) was associated with serum TES and DHT levels (P ¼ 0.03 and P ¼ 0.04, respectively), and there was limited Genetic Analysis evidence for an association with maximum urinary flow A total of eight gene-by-gene combinations were evaluated (Qmax; P ¼ 0.09), and post-void residual volume (PVR; to assess the possible synergistic genetic effects of pairs of P ¼ 0.10) (Table 3). Further evaluation of category-specific unlinked genes lying on the androgen metabolic pathway estimates (ie, pair-wise genotype comparisons) revealed an involved in the clinical development of BPH. Under this association between the predominant SRD5A1 Arg30-G/G paradigm, a significant test statistic would indicate that the genotype and a less severe BPH disease state (model fit relative risk of developing BPH associated with one genotype shown in Figure 1). In comparison to predominant homo- is modified by the presence or absence of a particular zygotes (Arg30-G/G), the probabilities of the Arg30-G/C and genotype at another locus. No statistically significant Arg30-C/C genotypes were higher at higher levels of TES departures from a random genotype-by-genotype distribu- (P ¼ 0.04 and P ¼ 0.01, respectively) and DHT (P ¼ 0.02 and

Table 2 P-values generated from cross-tabulations to assess joint gene-by-gene effects of unlinked genes

- Polymorphisms SRD5A1: Arg30 SRD5A1: Ala116 SRD5A2: Val89Leu SRD5A2: C T AR: (CAG)n

SRD5A1: Arg30 SRD5A1: Ala116 Linked SRD5A2: Val89Leu 0.39 0.24 SRD5A2:C-T 0.34 0.28 Linked AR: (CAG)n 0.79 0.79 0.77 0.60

Table 3 Observed associations between continuous clinical parameters and candidate polymorphic genesa

b c Genetic marker Prostate vol. TZV PSA Qmax PVR Testosterone DHT

5a-reductase (Type I) Silent Arg30 0.31 0.60 0.48 0.09 0.10 0.03 0.04 Silent Ala116 0.67 0.64 0.40 0.04 0.01 0.43 0.41

5a-reductase (Type II) Val89Leu 0.99 0.87 0.91 0.40 0.24 0.17 0.83 C-T 0.97 0.82 0.44 0.02d 0.47 0.29 0.87

Androgen receptore (CAG)n 0.14 0.36 0.32 0.65 0.34 0.04 0.08 a1A-adrenoceptor Arg492Cys 0.65 0.07 0.30 0.98 0.87 0.47 0.05 aReported main-effect P-values, from unadjusted models, test the overall contribution of clinical parameters towards predicting observed genotypes. bTransition zone volume. cPost-void residual volume dNot clinically significant; mean values were comparable across C/C, C/T, and T/T genotypes (10.273.6, 10.373.3, 11.074.2, respectively). e (CAG)n categorized as p18, 18–23, and X23 repeats.

www.nature.com/tpj Evaluation of candidate genes for severity of BPH M Klotsman et al 254

Figure 1 Associations between serum DHT (panel a), TES (panel b), Qmax (panel c), and post-void residual volume (panel d) and 5a- reductase type-I Arg30 silent polymorphism. All reported P-values compare predominant homozygotes (referent) to heterozygotes and variant homozygotes. Mean values (7SD) are also shown.

P ¼ 0.03, respectively). A ratio of TES to DHT was also older men (Figure 2). Comparing men with long-repeat analyzed to further assess whether SRD5A1 polymorphisms lengths ((CAG)X23) to those with medium length repeats affect androgen , but no associations were ((CAG)19–23) showed the probability of (CAG)X23 to be detected. Additionally, pair-wise comparisons between the strongly associated with high serum TES levels (P ¼ 0.004), two homozygous groups showed that the Arg30-G/G but less so with elevated DHT levels (P ¼ 0.01). Results for genotype was associated with slightly faster maximum men with short repeat lengths ((CAG)o19) were difficult to urinary flow (P ¼ 0.05) and a small decrease in post-void interpret because of the limited sample size (n ¼ 83). Similar residual volume (P ¼ 0.04) (Figure 1). Further analysis did results were obtained using a dichotomous (CAG) classifica- not identify associations between the joint distribution of tion scheme encountered in the literature;22 longer repeats clinical predictors and SRD5A1 Arg30 genotypes. (CAG422 repeats) were associated with elevated TES and The Ala116 SNP was associated with maximum urine flow DHT levels (P ¼ 0.003 and P ¼ 0.01, respectively).

(P ¼ 0.04) and post-void residual volume (P ¼ 0.01). Plotting An SNP in the a1A-adrenergic receptor was evaluated the fitted genotype probabilities of Ala116 against Qmax and because the product of this gene is hypothesized to PVR revealed a trend similar to that observed for Arg30 contribute to dynamic obstruction. While not on the

(which is in partial LD with Ala116); the homozygous androgen pathway, a weak association between a1A-adrener- variant Ala116-G/G differed from predominant homozy- gic receptor Arg492Cys polymorphism and DHT was found gotes and heterozygotes with respect to Qmax and PVR (P ¼ 0.05); however, this association did not persist (P ¼ 0.15) (pairwise Ala116 ÀA/A vs ÀG/G comparisons P ¼ 0.01 and after excluding an outlier with serum DHT concentration of P ¼ 0.003, respectively). 4210 ng/ml. Evaluation of the other clinical predictor Complex associations between AR (CAG) repeats and variables did not identify outliers. Spline regression analysis, serum androgen levels were observed. Associations were not performed to more thoroughly examine potential dose– observed among men less than 60 years of age, while response relationships, did not improve the fit to the data of genotype/phenotype associations were identified among any of the models considered.

The Pharmacogenomics Journal Evaluation of candidate genes for severity of BPH M Klotsman et al 255

Figure 2 Associations between serum androgen levels and AR (CAG)n repeat length. The probability of (CAG) length in relation to serum TES and serum DHT among Caucasian subjects 60 years and younger (panels a and b, respectively) and Caucasian subjects older than 60 years (panels c and d, respectively).

DISCUSSION 30% more active than variant 89-Leu/Leu.25,26 Observa- The notion that lower urinary tract symptoms arise from tional studies have examined the risk of either a static or dynamic component is supported by a associated with the Val89Leu SNP, but results have var- growing body of BPH literature and clinical experience; yet, ied14,24,27–29 and no conclusive genotype/phenotype asso- the molecular epidemiologic evidence supporting either of ciations can be drawn. In the present study, we did not find these pathways is scant. To improve the understanding of associations between SRD5A2 polymorphisms and severity molecular factors contributing to this condition, associa- of BPH. Based on these findings and previous studies tions between candidate gene loci and severity of BPH were examining this SNP, we conclude that the Val89Leu SNP evaluated. Candidate gene selection was based on mechan- plays, at most, a minimal role in the natural history of BPH. istic pathways that provide a plausible etiology for the Further study of additional SRD5A2 polymorphisms (eg, a involvement of androgen metabolic loci contributing to functional SNP at codon 49)27,30 is needed to rule out prostate growth (static obstruction) and genes encoding for SRD5A2 as a genetic determinant of BPH severity. factors affecting smooth muscle tone (dynamic obstruction). SRD5A1 has not been extensively studied in relation to We evaluated the association between severity of BPH and BPH because SRD5A2 is the predominant found in polymorphisms in SRD5A1 and SRD5A2, two genes involved the prostate.31 Studies have shown SRD5A1 expression levels in the conversion of TES to DHT. As DHT is evidently a more to be elevated in BPH tissue,32 but associations between active prostate mitogen23 with greater affinity for the AR, SRD5A1 polymorphisms and BPH have not been reported. genetic factors that could affect 5a-reductase expression or We identified associations between severity of BPH (though activity were hypothesized to be associated with the severity not for prostate volume) and a silent polymorphism in of BPH. As previously reported,24 the frequency of the SRD5A1 at codon 30, thus further implicating SRD5A1 in SRD5A2-Val89Leu polymorphism shows marked differences BPH. Specifically, the variant Arg30 allele was associated among ethnic groups; however, the phenotypic conse- with higher serum androgen concentrations, impaired Qmax, quences associated with this SNP are uncertain. In vitro and increased post-void urinary retention volume. Also, a studies have shown that the predominant 89-Val/Val is over silent polymorphism at codon 116 was associated with Qmax

www.nature.com/tpj Evaluation of candidate genes for severity of BPH M Klotsman et al 256

and PVR. The process(es) in which these silent SNPs The lack of a consensus definition for BPH is a significant influences BPH phenotypes is unknown and additional hindrance to the epidemiologic study of BPH. Neither the studies will be needed to assess whether these SNPs have clinical measures that should be used to describe the direct functional consequences (ie, altered expression levels, properties of BPH, nor the normal ranges for these measures reduced protein activity). One likely explanation is that the have been agreed upon. The analytic approach used silent SRD5A1 SNPs are in LD with undefined causative circumvented this problem by using continuous clinical allele(s). Alternatively, it is plausible that the silent SNPs predictor variables to assess genotype/phenotype associa- impact upon the rate of translation, thus affecting the tions. While it was not possible to identify disease suscept- conversion of DHT to TES in the . Verifying the ibility loci, for example by comparing non-symptomatic involvement of SRD5A1 in the clinical phenotype of BPH men with those that have BPH, this study identified genetic would reinforce the rationale behind the use of dual 5a- factors that may play a role in the severity of BPH among a reductase inhibitors for the treatment of this condition. symptomatic population. Multiple testing, which can inflate For their activity to manifest in the prostate, the type I error rate and lead to spurious findings, is another must first bind to the AR and form an AR–androgen complex potential limitation. While it is plausible that the associa- that acts to transduce growth signals from the cellular tions identified for polymorphisms in SRD5A1 and AR are environment to the nucleus. Given the importance of AR in false positives, we feel this to be unlikely given that the modulation of genes involved in cell proliferation and genotype/phenotype associations held across multiple phe- differentiation, the (CAG)n microsatellite has been intensely notypes (eg, the silent Arg30 SNP was associated with higher 22 studied, in particular with respect to prostate cancer. Both serum androgen concentrations, impaired Qmax, and in- in vitro and in vivo studies have demonstrated that CAG creased post-void urinary retention volume). repeat length is inversely related to transactivation activity Overall, our results, taken together with previous reports, of AR,33–36 with additional evidence implicating AR in BPH demonstrate that the molecular determinants of BPH severity coming from population-based observational stu- severity are complex. Further studies are needed to more dies. Using data from the Physicians’ Health Study, it was clearly define prostatic growth pathways and to identify determined that the risk of surgery for BPH increased factors that increase smooth muscle tone in the prostatic linearly with decreasing CAG repeat length (OR ¼ 1.76, capsule. The characterization of additional molecular factors

95% CI (1.2, 2.7) for men with (CAG)p19 relative to men that contribute to static and dynamic obstruction may serve 37 with (CAG)X25). Similar results were obtained in the to identify novel drug targets for the clinical management of Health Professionals Follow-Up Study cohort, which used a BPH. similar (CAG) classification scheme; the odds ratio for BPH 15 surgery or an enlarged prostate was 1.92 (95% CI 1.2, 3.0). METHODS Also, the prevalence of BPH increased with age and shorter Study Population CAG repeat lengths, leading the authors of this study to Baseline data were available for 2802 men diagnosed with conclude that variability in CAG length influences the moderate-to-severe benign prostatic hyperplasia recruited development of symptomatic BPH. A third study, conducted throughout the United States to participate in randomized, among Japanese prostatectomy patients (176 BPH cases and placebo-controlled clinical trials. These trials, previously 41 controls), modeled resected prostate weight against CAG described in detail,38 had identical inclusion and exclusion repeat length. Using this methodology, a statistically criteria to allow for a preplanned pooled analysis of the data. significant trend for larger adenoma size with short CAG Briefly, to be eligible for inclusion, participants had to be at repeat length was identified, leading these authors to least 50 years of age, have a minimum American Urologic postulate that shorter CAG alleles represent a genetic risk Association symptom index (AUA-SI) score of 12, a prostate 16 factor that promotes the growth of BPH. volume 430 cm3, and maximum urinary flow rates of less In contrast to the observational studies described above, than 15 ml/s. Participants were excluded from the study if direct associations between candidate gene loci and risk of they had ever used 5a-reductase inhibitors or alpha-receptor BPH were not evaluated in the present study because a blockers within 4 weeks of screening. Prior to study suitable control population was unavailable. Rather, an enrollment, the aims of the study were fully explained and application of logistic regression modeling, which under informed consent was obtained from each participant. The general conditions yields maximum likelihood estimates study protocol was reviewed and approved by the appro- that are consistent, asymptomatically efficient, and asymp- priate Institutional Review Boards. tomatically normal, was used to identify associations between genetic loci and clinical measures marking the Ascertainment of Clinical Predictor Variables severity of BPH. Using this analytic approach, longer (CAG) Demographic and clinical severity indicators for BPH and its repeats were found to be associated with elevated androgen complications were collected for each subject according to a levels. The higher probability of observing long (CAG) standardized protocol. All statistical analysis was performed repeats at higher serum androgen levels may represent a on baseline measures collected among Caucasians prior to systemic response to decreased AR activity associated with receiving placebo or experimental drug. longer (CAG) repeats (ie, increased androgen bioavailability Subjects’ total prostate volume (TPV) and transition zone to counteract reduced AR activity). volume (TZV) were measured by transrectal ultrasound

The Pharmacogenomics Journal Evaluation of candidate genes for severity of BPH M Klotsman et al 257

(TRUS), an established method of measuring prostate size SRD5A1 silent Ala116 polymorphism, a A to G transition in that has a high level of reproducibility for volume measure- exon 2, was recognized by NspI.12 The Arg492Cys poly- 39,40 ments. The anteroposterior, cephalocaudal, and trans- morphism in a1A-adrenergic receptor was characterized by verse diameters of the prostate were obtained to calculate digestion with PstI. Results were visually obtained by the prostate volume using the following formula: pC6(an- evaluating 3% agarose gels. (CAG)n repeat length in AR teroposterior  cephalocaudal  transverse). The TZV was was established by direct sequencing at a central laboratory measured because this region is more prone to hyperplasia (GlaxoWellcome Inc., Research Triangle Park). Genotypes and more directly impinges on the prostatic urethra. were entered into the database independently by two Pressure-flow studies were performed to provide informa- individuals. tion on urethral resistance, bladder capacity, residual urine volume, and urinary flow rate. The peak urine flow (Qmax) Analysis was measured using a Dantec Uroflow Meter (URODYN Allele frequencies were estimated by allele counting. 1000) with a Thompson filter (a minimum voided volume of Departures from Hardy–Weinberg equilibrium (HWE) were 4125 ml was required to qualify the measurement). Sub- assessed by a w2 goodness-of-fit test based on the difference jects’ post-void residual volume was measured suprapubi- between expected (calculated from the allele frequencies) cally by an ultrasound immediately after every urinary flow and observed genotype frequencies. All genetic analyses measurement. were performed using the HelixTree software package (Golden Helix, Bozeman, Montana). Ascertainment of Clinical laboratory Tests The strict study eligibility criteria yielded a study group Blood samples were collected for each participant. All consisting of only BPH patients. Consequently, a case-only hematological and biochemical measurements, including analysis was appropriate for assessing etiologic gene-by-gene PSA, TES, and DHT, were conducted by a central laboratory interaction.44 Here we define the interaction on a multi- in a blinded manner. plicative scale, considering that interaction is present between two unlinked genes if there is departure from Genotyping equality of the genotype relative risks at one locus across All genotyping was based on whole blood collected from genotypes at the other locus. We considered gene-by-gene subjects who consented to genetic testing. DNA extraction interaction of all pairs of unlinked genes lying on the same was performed using the Puregene DNA isolation kit at a androgen-dependent prostatic growth pathway by con- central laboratory (LabCorp, Research Triangle Park, NC, structing a series of genotype-by-genotype tables (three by USA). Subjects were tested for a total of five SNPs and one three contingency tables) and testing the independence.45 trinucleotide repeat. Polymorphisms evaluated include: a C An overall w2 (d.f. ¼ 4) was calculated for each pair to test the to T transition in intron 1 (C to T) and a Val to Leu null hypothesis that the joint effects were multiplicative substitution at amino-acid position 89 (Val89Leu) in 5-alpha (statistical significance was two-tailed and defined as reductase type II41 (HUGO nomenclature: -5-alpha- Po0.05). Similar to the single-locus test described by Nielsen reductase, alpha polypeptide 2, SRD5A2; Sequence Accession et al,46 this is a valid test for interaction in the absence of ID: NM_000348); two silent SNPs in 5-alpha reductase type population LD. A hallmark of many genetic-based disorders I(12) (HUGO nomenclature: steroid-5-alpha-reductase, al- is an earlier age of onset (eg, the occurrence of cancer at an pha polypeptide 1, SRD5A1; Sequence Accession ID: unusually young age, relative to the typical age for the type NM_001047) at codons 30 (Arg30) and 116 (Ala116); an of cancer, is suggestive of an inherited predisposition).47

Arg to Cys substitution at codon 492 in a1A-adrenergic Accordingly, this analysis was repeated for men less than 60 receptor42 (OMIM nomenclature: Alpha-1A-adrenergic re- years old to consider whether gene-by-gene interaction ceptor, ADRA1A; OMIM access number: 104221); and a contributed to risk of BPH. 43 (CAG)n repeat in the AR (HUGO nomenclature: AR; Polytomous logistic regression analysis was conducted to Sequence Accession ID: M20132). For presentation purposes, evaluate associations between genotypes and clinical para- the more frequently observed allele at each locus is referred meters used to characterize the severity of BPH. The to as the ‘predominant’ allele, while ‘variant’ refers to the probability of observing a locus-specific genotype was less commonly observed allele. estimated by polytomous logistic regression, based on All SNPs were analyzed in a blinded manner by validated selecting values for the model parameters that maximize PCR-RFLP procedures at a central laboratory (GlaxoWell- the likelihood of the observed data. Thus, we modeled Pr come Inc., Research Triangle Park, NC, USA). PCR for all (genotype ¼ i|CP1,CP2, y,CPN) where i indexes the genotyping was performed in 96-well specimen plates with genotype and the CPn are clinical predictor variables. In negative controls (no template DNA) and positive controls applying the logistic model in this manner, overall two (DNAs of known genotype). Approximately 5 ml PCR product degrees-of-freedom w2 tests can be used to assess associations was used for each subsequent enzyme digestion. The between clinical variables and the three-level genotypes SRD5A2 Val89Leu polymorphism and C to T transition were corresponding to the bi-allelic candidate polymorphic loci recognized by the restriction SunI and PstI, (predominant homozygotes, heterozygotes, and variant respectively. The SRD5A1 silent Arg30 polymorphism, a G homozygotes). The likelihood ratio test for a given clinical to C transition in exon 1, was recognized by HinfI.12 The measure is based on comparing the fit of such a model with

www.nature.com/tpj Evaluation of candidate genes for severity of BPH M Klotsman et al 258

that of a model with all the same clinical predictors except Campbell’s Urology, 7th edn. W.B. Saunders Company: Philadelphia the one being tested. Based on this w2 test, the P-value 1998; pp 1381–1402. quantifies the strength of the evidence for an association 5 Ross RK, Pike MC, Coetzee GA, Reichardt JK, Yu MC, Feigelson H et al. Androgen metabolism and prostate cancer: establishing a model of between the genotype and the clinical predictor, adjusting genetic susceptibility. Cancer Res 1998; 58: 4497–4504. for the values of the other clinical predictors. 6 Chang C, Saltzman A, Yeh S, Young W, Keller E, Lee HJ et al. Androgen Initially, polytomous-outcome logistic regression models receptor: an overview. Crit Rev Eukaryot 1995; 5: were constructed by modeling the three-category genotype 97–125. 7 Gsur A, Preyer M, Haidinger G, Zidek T, Madersbacher S, Schatzl G et al. at each locus against each continuous clinical predictor to Polymorphic CAG repeats in the androgen receptor gene, prostate- separately maximize each likelihood. If the predictive specific antigen polymorphism and prostate cancer risk. Carcinogenesis 2002; 23: 1647–1651. contribution of a given clinical predictor, CPn, approached statistical significance (P 0.10), then category-specific 8 Walden PD, Gerardi C, Lepor H. Localization and expression of the o alpha1A-1, alpha1B and alpha1D-adrenoceptors in hyperplastic and parameters were estimated to explore which contrasts (ie, non-hyperplastic human prostate. J Urol 1999; 161: 635–640. pair-wise contrasts between genotype groups; d.f. ¼ 1) were 9 Marshall I, Burt RP, Chapple CR. Noradrenaline contractions of human producing the apparent effect. These pair-wise comparisons prostate mediated by alpha 1A-(alpha 1c-) adrenoceptor subtype. Br J (predominant homozygotes vs heterozygotes and variant Pharmacol 1995; 115: 781–786. 10 Andersson KE. Prostatic and extraprostatic alpha-adrenoceptors— 2 homozygotes) were made by evaluating pair-specific w contributions to the lower urinary tract symptoms in benign prostatic statistics and corresponding P-values (ie, test statistics hyperplasia. Scandinavian Journal of Urology & Nephrology. Scand J corresponding to the specific genotype category pair) to Urol Nephrol 1996; 179(Suppl): 105–111. test the null hypothesis that the relative frequencies of 11 Pennefather JN, Lau WA, Mitchelson F, Ventura S. The autonomic and sensory innervation of the smooth muscle of the prostate gland: a 0 genotype ¼ ivsgenotype ¼ i are the same over the observed review of pharmacological and histological studies. J Auton Pharmacol range of the given clinical explanatory variable, CPn.In 2000; 20: 193–206. other words, a significant test (defined as Po0.05) indicates 12 Jenkins EP, Hsieh CL, Milatovich A, Normington K, Berman DM, Francke that CP is related to the relative genotype frequencies U et al. Characterization and chromosomal mapping of a human steroid n 5 alpha-reductase gene and pseudogene and mapping of the mouse Pr(genotype ¼ i), suggesting the presence of a phenotype/ homologue. Genomics 1991; 11: 1102–1112. genotype association. Violations of constancy of genotype 13 Makridakis NM, di Salle E, Reichardt JK. Biochemical and pharmacoge- probabilities were visualized by plotting the estimated netic dissection of human steroid 5 alpha-reductase type II. Pharmaco- predicted probability of genotype ¼ i against each predictor. genetics 2000; 10: 407–413. 14 Febbo PG, Kantoff PW, Platz EA, Casey D, Batter S, Giovannucci E et al. To further characterize potential dose–response relation- The V89L polymorphism in the 5alpha-reductase type 2 gene and risk of ships between genotypes and predictors, those variables that prostate cancer. Cancer Res 1999; 59: 5878–5881. appeared to be associated with genotype were also modeled 15 Giovannucci E, Platz EA, Stampfer MJ, Chan A, Krithivas K, Kawachi I using quadratic splines.48 Splines were constructed by et al. The CAG repeat within the androgen receptor gene and benign prostatic hyperplasia. Urology 1999; 53: 121–125. introducing knots at the 25th and 75th percentiles of the 16 Mitsumori K, Terai A, Oka H, Segawa T, Ogura K, Yoshida O et al. CP distribution. Model fit was assessed by using a likelihood Androgen receptor CAG repeat length polymorphism in benign ratio test to compare spline models to those with a logistic- prostatic hyperplasia (BPH): correlation with adenoma growth. Prostate linear structure. 1999; 41: 253–257. 17 Shibata K, Hirasawa A, Moriyama N, Kawabe K, Ogawa S, Tsujimoto G. Step-up modeling, based on the initial logistic models fit, Alpha 1a-adrenoceptor polymorphism: pharmacological characteriza- was used to evaluate whether the fit to the data was tion and association with benign prostatic hypertrophy. Br J Pharmacol improved by including more of the clinical variables as 1996; 118: 1403–1408. predictors. Models fit with multiple predictor variables were 18 Buscher R, Herrmann V, Insel PA. Human adrenoceptor polymorphisms: evolving recognition of clinical importance. Trends Pharmacol Sci 1999; compared to the reduced models by likelihood ratio tests. 20: 94–99. All analyses were performed using the SAS and JMP 19 Barry MJ, Roehrborn CG. Benign prostatic hyperplasia. BMJ 2001; 323: statistical software packages (SAS Institute, Cary, NC, USA). 1042–1046. 20 Narayan P, Tewari A. Overview of alpha-blocker therapy for benign prostatic hyperplasia. Urology 1998; 51(Suppl 4A): 38–45. DUALITY OF INTEREST 21 Bartsch G, Rittmaster RS, Klocker H. Dihydrotestosterone and the Drs M Klotsman, K Davis, and CG Binnie are employees of concept of 5alpha-reductase inhibition in human benign prostatic hyperplasia. Eur Urol 2000; 37: 367–380. GlaxoSmithKline. 22 Nelson KA, Witte JS. Androgen receptor CAG repeats and prostate cancer. Am J Epidemiol 2002; 155: 883–890. REFERENCES 23 Wright AS, Thomas LN, Douglas RC, Lazier CB, Rittmaster RS. Relative 1 Verhamme KMC, Dieleman JP, Bleumink GS, van der Lei J, Sturkenboom potency of testosterone and dihydrotestosterone in preventing atrophy M. Incidence and prevalence of lower urinary tract symptoms and apoptosis in the prostate of the castrated rat. J Clin Investig 1996; suggestive of benign prostatic hyperplasia in primary care—The 98: 2558–2563. Triumph project. Eur Urol 2002; 42: 323–328. 24 Lunn RM, Bell DA, Mohler JL, Taylor JA. Prostate cancer risk and 2 Berry SJ, Coffey DS, Walsh PC, Ewing LL. The development of human polymorphism in 17 hydroxylase (CYP17) and benign prostatic hyperplasia with age. J Urol 1984; 132: 474–479. (SRD5A2). Carcinogenesis 1999; 20: 1727–1731. 3 Jacobsen SJ, Jacobson DJ, Girman CJ, Roberts RO, Rhodes T, Guess HA 25 Makridakis N, Ross RK, Pike MC, Chang L, Stanczyk FZ, Kolonel LN et al. et al. Treatment for benign prostatic hyperplasia among community A prevalent missense substitution that modulates activity of prostatic dwelling men: the Olmsted County study of urinary symptoms and steroid 5alpha-reductase. Cancer Res 1997; 57: 1020–1022. health status. J Urol 1999; 162: 1301–1306. 26 Ross RK, Bernstein L, Lobo RA, Shimizu H, Stanczyk FZ, Pike MC et al. 4 Partin AW, Coffey DS. The molecular biology, endocrinology, and 5-alpha-reductase activity and risk of prostate cancer among Japanese physiology of the prostate and seminal vesicles. In: Walsh PC (ed) and US white and black males. Lancet 1992; 339: 887–889.

The Pharmacogenomics Journal Evaluation of candidate genes for severity of BPH M Klotsman et al 259

27 Jaffe JM, Malkowicz SB, Walker AH, MacBride S, Peschel R, Tomaszewski 37 Giovannucci E, Stampfer MJ, Chan A, Krithivas K, Gann PH, Hennekens J et al. Association of SRD5A2 genotype and pathological characteristics CH et al. CAG repeat within the androgen receptor gene and incidence of prostate tumors. Cancer Res 2000; 60: 1626–1630. of surgery for benign prostatic hyperplasia in US physicians. Prostate 28 Latil AG, Azzouzi R, Cancel GS, Guillaume EC, Cochan-Priollet B, 1999; 39: 130–134. Berthon PL et al. Prostate carcinoma risk and allelic variants of genes 38 Roehrborn CG, Boyle P, Nickel JC, Hoefner K, Andriole G. Efficacy and involved in androgen and metabolism pathways. Cancer safety of a dual inhibitor of 5-alpha-reductase types 1 and 2 (dutaste- 2001; 92: 1130–1137. ride) in men with benign prostatic hyperplasia. Urology 2002; 60: 29 Yamada Y, Watanabe M, Murata M, Yamanaka M, Kubota Y, Ito H et al. 434–441. Impact of genetic polymorphisms of 17-hydroxylase cytochrome P-450 39 Collins GN, Raab GM, Hehir M, King B, Garraway WM. Reproducibility (CYP17) and steroid 5alpha-reductase type II (SRD5A2) genes on and observer variability of transrectal ultrasound measurements of prostate-cancer risk among the Japanese population. Int J Cancer 2001; prostatic volume. Ultrasound Med Biol 1995; 21: 1101–1105. 92: 683–686. 40 Grossfeld GD, Coakley FV. Benign prostatic hyperplasia: clinical over- 30 Makridakis NM, Ross RK, Pike MC, Crocitto LE, Kolonel LN, Pearce CL et view and value of diagnostic imaging. Radiol Clin North Am 2000; 38: al. Association of mis-sense substitution in SRD5A2 gene with prostate 31–47. cancer in African-American and Hispanic men in Los Angeles, USA. 41 Thigpen AE, Davis DL, Milatovich A, Mendonca BB, Imperato-McGinley Lancet 1999; 354: 975–978. J, Griffin JE et al. Molecular genetics of steroid 5 alpha-reductase 2 31 Harris G, Azzolina B, Baginsky W, Cimis G, Rasmusson GH, Tolman RL et deficiency. J Clin Investig 1992; 90: 799–809. al. Identification and selective inhibition of an isozyme of steroid 5 42 Loftus SK, Shiang R, Warrington JA, Bengtsson U, McPherson JD, alpha-reductase in human scalp. Proc Natl Acad Sci USA 1992; 89: Wasmuth JJ. Genes encoding adrenergic receptors are not clustered on 10787–10791. the long arm of human 5. Cytogenet Cell Genet 1994; 67: 32 Iehle C, Radvanyi F, Gil Diez de Medina S, Ouafik LH, Gerard H, Chopin 69–74. D et al. Differences in steroid 5alpha-reductase iso-enzymes expression 43 Brown CJ, Goss SJ, Lubahn DB, Joseph DR, Wilson EM, French FS et al. between normal and pathological human prostate tissue. J Steroid Androgen receptor locus on the human X chromosome: regional Biochem Mol Biol 1999; 68: 189–195. localization to Xq11-12 and description of a DNA polymorphism. Am J 33 Chamberlain NL, Driver ED, Miesfeld RL. The length and location of Hum Genet 1989; 44: 264–269. CAG trinucleotide repeats in the androgen receptor N-terminal domain 44 Goldstein A, Andrieu N. Detection of interaction involving identified affect transactivation function. Nucleic Acids Res 1994; 22: 3181–3186. genes: available study designs. Monogr Natl Cancer Inst 1999; 26: 34 Choong CS, Kemppainen JA, Zhou ZX, Wilson EM. Reduced androgen 49–54. receptor gene expression with first exon CAG repeat expansion. Mol 45 Yang Q, Khoury M, Sun F, Flanders W. Case-only design to measure Endocrinol 1996; 10: 1527–1535. gene–gene interaction. Epidemiology 1999; 10: 167–170. 35 Mhatre AN, Trifiro MA, Kaufman M, Kazemi-Esfarjani P, Figlewicz D, 46 Nielsen DM, Ehm MG, Weir BS. Detecting marker-disease association Rouleau G et al. Reduced transcriptional regulatory competence of the by testing for Hardy–Weinberg disequilibrium at a marker locus. Am J androgen receptor in X-linked spinal and bulbar muscular atrophy. Nat Hum Genet 1998; 63: 1531–1540. Genet 1993; 5: 184–188. 47 Lindor N, Greene M, Program tMFC. The concise handbook of family 36 Kazemi-Esfarjani P, Trifiro MA, Pinsky L. Evidence for a repressive cancer syndromes. J Natl Cancer Inst 1998; 14: 1039–1071. function of the long polyglutamine tract in the human androgen 48 Greenland S. Dose-response and trend analysis in epidemiology: receptor: possible pathogenetic relevance for the (CAG)n-expanded alternatives to categorical analysis (comment). Epidemiology 1995; 6: neuronopathies. Hum Mol Genet 1995; 4: 523–527. 356–365.

www.nature.com/tpj