Prostate-Specific Membrane Antigen Protein Expression in Tumor Tissue and Risk of Lethal Prostate Cancer

Prostate-specific membrane antigen (PSMA) is a type II transmembrane glycoprotein that is highly expressed in the normal prostate epithelium, and to a lesser extent in other tissues such as brain, liver, and kidney (1, 2). PSMA expression is higher in primary prostate tumors and metastatic lesions compared with benign tissue, and is positively associated with tumor grade and stage (3–7). Because of its high expression in malignant prostate tissue, PSMA has been used in immunoscintigraphy to monitor metastatic disease and as a target antigen for immunotherapy (8, 9).

PSMA may also have prognostic utility. Three studies of surgically treated prostate cancer patients showed that high PSMA protein expression in tumor tissue was associated with biochemical recurrence (5–7). Two of these studies found that PSMA overexpression was predictive of biochemical recurrence after multivariable adjustment for clinical parameters, such as tumor stage, grade, and preoperative prostate-specific antigen (PSA) levels (5, 6). However, Minner and colleagues did not find PSMA to be an independent predictor after adjusting for clinicopathologic features (7). High PSMA mRNA expression in preoperative peripheral blood cells, possibly detecting micrometastatic disease, similarly showed a positive association with biochemical recurrence in four prospective studies (10–13), a relationship not observed in the fifth study (14). No studies to date have investigated PSMA expression in relation to prostate cancer–specific mortality.

PSMA functions as a peptidase with both N-acetylated α-linked acidic peptidase and folate hydrolase activity (15, 16). In vitro and in vivo experiments have shown that high PSMA expression activates signaling pathways that promote tumor cell survival and proliferation (17). The association of PSMA with anaphase-promoting complex disrupts cell-cycle checkpoints, induces chromosomal instability, and contributes to aneuploidy (18). In addition, PSMA is negatively regulated by 1α,25-dihydroxyvitamin D₃ (19), a nutrient associated with reduced proliferation in animal models and prostate cancer cell lines (20, 21). Interestingly, androgen deprivation enhances PSMA expression (1, 22), and a role in the development of castration resistance has been hypothesized. Androgens stimulate TMPRSS2:ERG expression, a gene fusion mutation common in human prostate cancer (23), as the TMPRSS2 promoter has an androgen-responsive element, thus providing a potential link between inhibition of PSMA by androgen and ets-related gene (ERG) expression in fusion-positive prostate cancer cells (24). PSMA has also been identified as a regulator of new blood vessel formation (i.e., angiogenesis) in mouse models (25, 26). Although virtually absent from nonprostatic normal tissues, PSMA is expressed in the neovasculature of many solid tumors, thus underscoring its importance in tumor angiogenesis (27–30).

In this prospective study, our main objective was to determine whether tumor PSMA protein expression from primarily radical prostatectomy specimens was an independent predictor of prostate cancer–specific mortality in 902 participants of the Physicians' Health Study (PHS) and Health Professionals Follow-Up Study (HPFS). To identify potential mechanisms of PSMA in disease progression, we also evaluated correlations between PSMA expression and measures of cell proliferation, apoptosis, angiogenesis, and protein expression of vitamin D receptor (VDR), androgen receptor (AR), and ERG in prostate tumor tissue.

This study population of patients with prostate cancer is drawn from participants of the prospective PHS and HPFS studies for whom archival prostate tumor tissue, primarily from radical prostatectomy, was available for biomarker analysis. PHS I and II were randomized, placebo-controlled, double-blind trials for the prevention of cardiovascular disease and cancer. PHS I began in 1982 and evaluated aspirin and β-carotene among 22,071 U.S. male physicians (31); in 1997 PHS II randomized 7,641 physicians from PHS I and 7,000 new physicians to β-carotene, vitamin E, vitamin C, and multivitamins (32). All arms of the PHS I and II have been terminated (33–35), and the PHS continues to be followed annually. The HPFS began in 1986 with 51,529 U.S. male health care professionals (dentists, veterinarians, pharmacists, optometrists, osteopathic physicians, and podiatrists) who are prospectively followed on biennial questionnaires to collect lifestyle and medical information (36). This study was approved by the Partners Healthcare and Harvard School of Public Health Institutional Review Boards.

Self-reported, incident cases of prostate cancer arising in the PHS (1983–2004) and HPFS (1986–2001) were confirmed by medical record and pathology report review by study investigators. In rare cases, prostate cancer diagnoses were identified on death certificates and confirmed by medical record, pathology report, and death certificate review. To ascertain clinical characteristics and disease-specific treatments or outcomes, information on tumor stage, PSA at diagnosis, body mass index (BMI), and metastases events during follow-up was collected from medical record and pathology report review, and from questionnaires sent to prostate cancer survivors (2004 onward). Pathologic tumor stage was available for 90% of patients, whereas the remaining had clinical stage information (n = 89) or were missing (n = 2). More than 97% of tumor specimens were re-reviewed by a study pathologist (M. Fiorentino and R. Flavin) to achieve uniformity of scoring, and the remaining were assigned clinical Gleason score. Cause of death was assigned via review of medical records and death certificates for the vast majority of participants, and secondarily via information from family. We defined lethal disease as death from prostate cancer or distant metastases (to bone or other organs, excluding lymph nodes) during follow-up. A total of 95 lethal events occurred: 29 in PHS and 66 in HPFS. We analyzed a composite of biochemical recurrence and lethal prostate cancer (n = 231) as a secondary endpoint, using the first recorded event as the event date. Biochemical recurrence was defined as PSA above 0.2 ng/mL after surgery sustained over two measures (when abstracted from medical records), or a report of biochemical recurrence by the participant or treating physician.

Formalin-fixed, paraffin-embedded archival tumor tissue specimens were obtained from the hospital pathology departments; 95% were from radical prostatectomy procedures and the remaining were from the transurethral resection of the prostate (TURP). Our pathologist reviewed all available slides to provide standardized Gleason grading and for identification of the areas of tumor tissue for tissue microarray construction blinded to outcome status (37). For this project, we used nine tissue microarrays constructed from areas of the dominant tumor nodule or highest Gleason grade, with at least three tumor cores (0.6 mm) sampled from each patient.

Protein expression of PSMA was ascertained on 5 μm sections of the tissue microarrays (pathologist: S.P. Finn). Antigen retrieval was by microwave in citrate buffer (3 × 5 minutes). We used a primary mouse monoclonal antibody (Clone E36, M3620; Dako) with 1:100 dilution for 60 minutes after treatment with a peroxidase block (Dako). An anti-mouse secondary antibody was applied, followed by a counterstain with hematoxylin (Sigma-Aldrich). PSMA expression was quantified using the Ariol platform (Genetix Corp.), a semiautomated, quantitative image analysis system, and defined as staining intensity (scale, 0–255) multiplied by percentage area staining positive (scale, 0%–100%) for a given tumor field on each tissue microarray core. All nine microarrays were stained in the same batch, and positive and negative controls were included according to the antibody manufacturer's instructions.

Cellular proliferation was assessed on 5 μm sections of the tissue microarrays using rabbit polyclonal anti-Ki67 antibody (Vector Laboratories), diluted at 1:2,000 with citrate-based antigen retrieval solution (pathologist: S.P. Finn). Ki67 staining was visualized using the Ariol platform (Genetix Corp.), and quantified as the percentage of positively stained nuclei in the tumor region of each core. Apoptosis was evaluated on 5 μm sections of the tissue microarrays using the ApopTag Peroxidase In Situ Kit (Chemicon International) according to the manufacturer's instructions, and defined as the percentage of tumor cells undergoing apoptosis (pathologist: M. Fiorentino; ref. 38).

VDR expression was calculated on 5 μm sections of the tissue microarrays using rabbit polyclonal anti-VDR antibody (Santa Cruz Biotechnology) at a dilution of 1:600 as previously described (pathologist: R. Flavin; ref. 37). VDR expression was quantified as a combination of percentage area that was positively stained and staining intensity using CRi Vectra, a semiautomated, quantitative image analysis system (CRi). AR expression was calculated on 5 μm sections of the tissue microarrays using rabbit polyclonal anti-AR antibody (Upstate/Millipore) at a dilution of 1:100 (pathologist: S.P. Finn). Mean intensity (scale, 0–255) of AR staining in the nucleus of tumor cells in a given core was calculated using the Ariol platform (Genetix Corp.). ERG expression was calculated on 5 μm tissue microarray sections (91% of patients) and prostate tissue block sections (9% of patients), using rabbit monoclonal anti-ERG antibody (Clone ID: EPR3864; Epitomics, Inc.) at a dilution of 1:100. Tumor specimens were evaluated individually by a study pathologist (R. Lis). The presence of ERG staining in the vascular endothelium served as a positive internal control, with ERG assessment restricted to cores in which the positive internal control was observed. A patient was considered positive for tumor ERG expression if ERG staining was observed within prostate cancer epithelial cells of at least one tissue microarray core.

Angiogenesis markers were assessed on 5 μm serial sections of the individual prostate tissue blocks in the HPFS cohort only. One to nine blocks with cancer were evaluated per case by a study pathologist as previously described by Mucci and colleagues (39). Endothelial cell marker CD34 protein expression was visualized using immunohistochemistry (QBEND10 primary mouse monoclonal antibody; Biogenex) and imaged using Image ProPlus 4.5 software (Media Cybernetics), a semiautomated image analysis platform. Angiogenesis markers were defined as the following: microvessel density, that is, the number of vascular structures in a high-power field (×200); vessel area in μm²; vessel diameter in μm; and vessel irregularity, that is, the irregularity of the vessel lumen calculated as the perimeter²/4 · Π × area, where a value of 1.0 indicates a perfect circle and values >1.0 indicate increasing irregularity. Measurements were averaged over the total tumor area evaluated for each patient. Smaller vessel area and diameter, and less regular vessel shape were associated with development of lethal prostate cancer in this cohort (39).

Analyses were based on the 902 men (n = 346 from PHS; n = 556 from HPFS) for whom PSMA expression was measured. The average value of each biomarker was calculated across all cores or tumor sections for a given patient. We compared age at diagnosis, clinical parameters, and BMI across quartiles of PSMA expression using ANOVA for normally distributed continuous variables, Kruskal–Wallis test for non-normally distributed continuous measures, and χ² tests for categorical variables.

Cox proportional hazards regression was used to calculate multivariable HRs and 95% confidence intervals (CI) for the association between PSMA expression and lethal prostate cancer. Follow-up time was calculated from the date of diagnosis to development of distant metastases, death from prostate cancer, or censored at death from another cause or end of follow-up (January 2009 or last date of contact for PHS; April 2012 for HPFS), whichever occurred first. We adjusted for tissue microarray (indicator variables) to account for staining variation across microarrays, and age at diagnosis (continuous), in all models. We further adjusted for Gleason score (2 to 6, 3 + 4, 4 + 3, 8 to 10) and PSA at diagnosis (<4, 4 to <10, ≥10 ng/mL, missing) to test whether PSMA expression was an independent predictor of lethal prostate cancer risk. We also examined these associations stratified by tumor stage (T1–T2, N0–Nx, M0–Mx vs. T3–T4 or N1 or M1), Gleason score (2 to 7 vs. 8 to 10), and ERG expression (absent, present). Violation of the proportional hazards assumption was tested by creating interaction terms between PSMA quartiles and follow-up time; the addition of the interaction terms to the model including PSMA quartiles, age at diagnosis, and tissue microarray, was not statistically significant (Wald test P = 0.21; 3 degrees of freedom), thus the assumption was satisfied. Because PSMA is negatively correlated with androgen levels (1, 22), we also performed a sensitivity analysis excluding the 57 patients who received any type of neoadjuvant or adjuvant hormone therapy ± 1 year from the date of radical prostatectomy or TURP. To test the association between PSMA expression and the composite endpoint of biochemical recurrence and lethal disease, follow-up time was calculated from date of diagnosis to date of recurrence, distant metastases, or death from prostate cancer; patients without a recurrence were censored at death from another cause or end of follow-up.

We examined correlations between PSMA expression and tumor biomarkers (proliferation index, apoptotic index, AR expression, VDR expression, and angiogenesis measures) using partial Spearman rank correlations, adjusted for age at diagnosis, and tissue microarray. PSMA expression across categories of ERG expression (absent, present) was evaluated using analysis of covariance (ANCOVA), adjusted for age at diagnosis and tissue microarray.

Analyses were conducted using SAS system software (version 9.2; SAS Institute). All P values were two-sided and considered statistically significant if less than 0.05.

Among the 902 patients with prostate cancer, mean age at diagnosis was 65.8 years with an average follow-up time of 13.2 years (Table 1). Higher PSMA expression was associated (P < 0.01) with increasing age, higher Gleason score, and higher PSA at diagnosis, and modestly associated (P = 0.07) with higher tumor stage. Mean tumor PSMA expression among all patients was 43.9 with an interquartile range (IQR) of 10.5 to 70.7. PSMA expression (mean ± SD) was similar between the cohorts (44.7 ± 36.8 for PHS and 43.5 ± 35.7 for HPFS), and between prostatectomy and TURP specimens (44.2 ± 36.1 and 39.6 ± 36.4, respectively). PSMA staining in the tumor was membranous and cytoplasmic (Fig. 1).

PSMA protein expression in tumor tissue was associated with a 2.4-fold (95% CI, 1.3–4.5) increased risk of lethal prostate cancer comparing the highest to lowest quartile, adjusting for age at diagnosis, and tissue microarray (Table 2). This positive association was stronger among patients with nonadvanced stage disease (HR_{Quartile(Q)4 vs. 1}, 4.3; P_trend < 0.01), lower Gleason score ≤7 tumors (HR_{Q4 vs. 1} = 4.6; P_trend < 0.01), as well as those with ERG-positive tumors (HR_{Q4 vs. 1} = 3.5; P_trend < 0.01). No associations with lethal cancer were found in men with advanced stage disease (P_trend = 0.27), poorly differentiated (Gleason score ≥8) tumors (P_trend = 0.39), or ERG-negative tumors (P_trend = 0.35). After further adjustment for Gleason score and PSA at diagnosis, the associations between PSMA expression and lethal prostate cancer were attenuated for overall (P_trend = 0.76), nonadvanced (P_trend = 0.61), Gleason score ≤7 (P_trend = 0.51), and ERG-positive (P_trend = 0.88) prostate cancer, and all were nonsignificant.

Among all 902 patients, associations of clinical parameters and risk of lethal prostate cancer were: age at diagnosis (per 5-year increase; HR, 1.2; 95% CI, 1.0–1.4); Gleason score (HR_{3+4 vs. 2–6}, 1.4; 95% CI, 0.5–4.5; HR_{4+3 vs. 2–6}, 4.1; 95% CI, 1.4–12.0; HR_{8–10 vs. 2–6}, 7.7; 95% CI, 2.7–21.9); PSA at diagnosis (HR_{4–9.9 vs. <4}, 1.5; 95% CI, 0.3–6.2; HR_{≥10 vs. <4}, 2.8; 95% CI, 0.7–11.8); tumor stage (HR_{T3 vs. T1–T2}, 1.7; 95% CI, 1.1–2.8; HR_{T4/N1/M1 vs. T1–T2}, 5.1; 95% CI, 2.9–9.1); mutually adjusted for all four parameters.

In the model adjusting for age at diagnosis and tissue microarray, effect estimates were slightly stronger after excluding patients who had received neoadjuvant or adjuvant hormone therapy: HR_{Q2 vs. 1}, 2.14 (95% CI, 1.03–4.44), HR_{Q3 vs. 1}, 2.01 (95% CI, 0.96–4.21), HR_{Q4 vs. 1}, 3.20 (95% CI, 1.60–6.39), P_trend < 0.01. Similar to the main analysis, results were attenuated and nonsignificant after further adjusting for Gleason score and PSA at diagnosis: HR_{Q2 vs. 1}, 1.78 (95% CI, 0.84–3.80), HR_{Q3 vs. 1}, 1.72 (95% CI, 0.80–3.72), HR_{Q4 vs. 1}, 1.38 (95% CI, 0.67–2.86), P_trend = 0.92.

Compared with the primary outcome of lethal prostate cancer, the association between PSMA expression and the composite outcome of biochemical recurrence and lethal disease was weaker and nonsignificant: HR_{Q2 vs. 1}, 0.90 (95% CI, 0.61–1.33), HR_{Q3 vs. 1}, 1.26 (95% CI, 0.87–1.82), HR_{Q4 vs. 1}, 1.24 (95% CI, 0.86–1.78), P_trend = 0.09, adjusting for age at diagnosis and tissue microarray; and HR_{Q2 vs. 1}, 0.75 (95% CI, 0.50–1.12), HR_{Q3 vs. 1}, 0.89 (95% CI, 0.61–1.31), HR_{Q4 vs. 1}, 0.68 (95% CI, 0.46–1.01), P_trend = 0.13, after further adjusting for Gleason score and PSA at diagnosis.

Tumors with high PSMA expression showed significantly lower protein expression of VDR and AR, and absence of ERG protein expression, among all patients (Table 3). High PSMA expression was also significantly correlated with markers of angiogenic activity, including higher microvessel density, smaller vessel area, smaller vessel diameter, and irregular shape. With the exception of ERG expression, the correlations between PSMA and other tumor biomarkers did not retain statistical significance in poorly differentiated tumors. No correlations were found for proliferation or apoptotic indices among all patients or within subgroups.

The association between PSMA expression and lethal prostate cancer among all patients, adjusted for age at diagnosis and tissue microarray, remained statistically significant after further adjustment for VDR (HR_{Q4 vs. 1}, 2.16; 95% CI, 1.14–4.11; P_trend = 0.03; n = 812), AR (HR_{Q4 vs. 1}, 2.31; 95% CI, 1.25–4.29; P_trend < 0.01; n = 860), or ERG expression (HR_{Q4 vs. 1}, 2.41; 95% CI, 1.28–4.53; P_trend < 0.01; n = 880). Among HPFS patients with measured angiogenesis markers (microvessel density, vessel area, vessel diameter, and vessel irregularity), higher PSMA expression was nonsignificantly associated with lethal disease (HR_{Q4 vs. 1}, 2.45; 95% CI, 0.92–6.49; P_trend = 0.19; n = 414), adjusting for age at diagnosis and tissue microarray. This association was attenuated after further adjusting for all four markers (HR_{Q4 vs. 1}, 1.65; 95% CI, 0.60–4.54; P_trend = 0.75), or any of the markers individually (data not shown).

In a large cohort of prostate cancer patients with over 13 years of average follow-up, PSMA protein expression in tumor tissue was positively associated with risk of lethal disease, but this association was not independent of clinical parameters. Thus, our study does not support the clinical utility of PSMA expression as a strong candidate biomarker for lethal prostate cancer among surgically treated patients. After considering additional markers of disease aggressiveness, we found that PSMA expression likely captures, in part, malignant features of Gleason grade and tumor angiogenesis.

Three prior studies of PSMA protein expression in prostate tumor tissue have reported positive associations with risk of biochemical recurrence (5–7). Minner and colleagues followed 1,426 patients with prostate cancer for up to 12 years and noted a borderline significant association for high versus low PSMA expression in radical prostatectomy tissue and PSA recurrence (7). Similar to our study, the association did not remain statistically significant after multivariable adjustment for clinical parameters. A smaller study of 136 patients (61% with organ-confined tumors) who underwent radical prostatectomy found that PSMA overexpression was associated with biochemical recurrence, even after multivariable adjustment for clinicopathological parameters (6). A third study of 93 patients (43% with lymph-node positive disease at surgery) found a significant positive association between PSMA expression and biochemical recurrence after adjusting for extraprostatic extension, though the estimates adjusted for additional clinical parameters is not presented (5). Our results may differ from these studies as more than 70% of our patients were diagnosed with nonadvanced stage tumors, our specimens were re-reviewed by a study pathologist for uniformity of Gleason score, and PSA levels at diagnosis were included in the multivariable models. Because PSMA expression has been positively correlated with these clinicopathologic features, it is unclear whether the positive findings from other studies would persist after accounting for all these factors. Furthermore, Ross and colleagues used the 7E11 anti-PSMA antibody, which recognizes the internal domain of PSMA (6), whereas the other two prior studies (5, 7) and our current study used clone 3E6, which recognizes the extracellular domain. Finally, our results may differ as our study was the first to assess lethal disease as the primary endpoint, whereas all prior studies evaluated time to biochemical recurrence.

We previously showed that a greater number of smaller and more poorly formed vessels within the prostate tumor were strong predictors of lethal disease (39). Our current study supports that PSMA is indicative of increased tumor angiogenesis, and after adjusting for these markers, the association of PSMA expression with lethal prostate cancer was markedly attenuated. This is consistent with the prior observation of PSMA being expressed in the endothelial cells of certain solid tumor neovasculature, including prostate cancer, renal cell carcinoma, transitional cell carcinoma of the bladder, gastric cancer, and colorectal cancer (27–30). Also, a small study of LNCaP tumors grown in nude mice found a strong positive correlation between protein expression of PSMA and VEGF, a signal protein that stimulates angiogenesis (40).

PSMA seems to be regulated by androgens, in that PSMA expression in prostate tumors is highest in hormone-deprived states, and is repressed in response to testosterone (1, 22). We found that higher PSMA expression was correlated with lower AR expression in prostate tumor tissue, though we did not have a measure of circulating testosterone levels at the time of surgery in our study. We also found that PSMA expression was lower in tumors that expressed ERG, which is supported by the prior finding that TMPRSS2-ERG fusion negatively regulated PSMA expression in LNCaP cells (24). In addition, the association between PSMA expression and lethal prostate cancer in our study was limited to ERG-positive tumors, suggesting that the link between PSMA and disease progression may depend on the molecular subtype of the tumor. Further studies are warranted to better understand the mechanisms by which PSMA, AR, and the TMPRSS2-ERG fusion may interact to influence prostate carcinogenesis.

The negative correlation we observed between VDR and PSMA expression is consistent with Serda and colleagues, who reported that 1α,25-dihydroxyvitamin D₃ downregulated PSMA expression in LNCaP cells (19). We previously reported an inverse association between VDR expression and prostate cancer progression in this patient cohort (37). In the current study, PSMA expression was associated with lethal prostate cancer independently of VDR levels in the age- and tissue microarray–adjusted models, suggesting that PSMA and VDR may act through different mechanisms to influence disease progression. Indeed, vitamin D has been shown to exert antiproliferative and proapoptotic effects on prostate tumors (20, 21, 41), whereas we found no correlation between PSMA and indices of proliferation or apoptosis.

Limitations of our study include potential misclassification of PSMA protein expression due to assay and detection variability, though any bias is likely nondifferential as study pathologists were blinded to outcome status. Also, we had low statistical power to detect associations among subgroups of patients with small numbers of outcomes. Furthermore, we used mainly prostatectomy tissue with the majority of patients having organ-confined disease, thus it is unknown whether our findings would be generalizable to PSMA expression measured in biopsy specimens. Our study has several notable strengths. We were the first to evaluate the association between PSMA expression and lethal disease within two large, established cohort studies with long-term and complete follow-up among patients with prostate cancer. In addition, the patients were well-characterized with respect to clinical and pathologic measures, including re-review of Gleason scores.

In our study of 902 U.S.-based patients with prostate cancer, PSMA protein expression measured in prostate tumor tissue was associated with progression to lethal disease, but not independent of clinical predictors. Our results suggest that PSMA is an indicator of increased tumor angiogenesis, and through this pathway, increased risk of prostate cancer progression. Overall, our findings do not support the clinical utility of tumor PSMA expression as a predictor of lethal disease among patients who undergo radical prostatectomy, though it is unknown how this biomarker may perform in biopsy specimens from patients who choose other treatment modalities such as active surveillance or radiation.

No potential conflicts of interest were disclosed.

Conception and design: J.L. Kasperzyk, L.A. Mucci

Development of methodology: J.L. Kasperzyk, S.P. Finn, R. Flavin, S.K. Clinton, M. Loda, L.A. Mucci

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): J.L. Kasperzyk, S.P. Finn, R. Flavin, M. Fiorentino, R. Lis, W.K. Hendrickson, S.K. Clinton, H.D. Sesso, E.L. Giovannucci, M. Loda, L.A. Mucci

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): J.L. Kasperzyk, M. Fiorentino, R. Lis, W.K. Hendrickson, E.L. Giovannucci, M.J. Stampfer, M. Loda, L.A. Mucci

Writing, review, and/or revision of the manuscript: J.L. Kasperzyk, S.P. Finn, R. Flavin, R. Lis, S.K. Clinton, H.D. Sesso, E.L. Giovannucci, M.J. Stampfer, L.A. Mucci

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): R. Lis, W.K. Hendrickson, H.D. Sesso

Study supervision: H.D. Sesso, L.A. Mucci

The authors thank the participants and staff of the PHS and HPFS for their valuable contributions. In addition, they also thank the following state cancer registries for their help: AL, AZ, AR, CA, CO, CT, DE, FL, GA, ID, IL, IN, IA, KY, LA, ME, MD, MA, MI, NE, NH, NJ, NY, NC, ND, OH, OK, OR, PA, RI, SC, TN, TX, VA, WA, WY. The Dana-Farber/Harvard Cancer Center Tissue Microarray Core Facility constructed the tissue microarrays in this project, and Chungdak Li for her expert tissue microarray construction.

This work was supported by the NIH [grant numbers CA34944, CA40360, CA097193, CA055075, CA131945, CA136578, CA141298, CA09001 (to J.L. Kasperzyk), HL26490, HL34595]; the American Institute for Cancer Research, the Department of Defense Prostate Cancer Research Program (grant number W81XWH-12-1-0072) and Dana-Farber Cancer Institute Mazzone Awards Program to J.L. Kasperzyk; Dana-Farber/Harvard Cancer Center SPORE in Prostate Cancer (grant number 5P50CA090381-08); and the Prostate Cancer Foundation to L.A. Mucci and S.P. Finn.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.