Interleukin-1β Promotes Skeletal Colonization and Progression of Metastatic Prostate Cancer Cells with Neuroendocrine Features

The therapeutic management of patients with prostate cancer includes the blockade of androgen receptor (AR) activation and signaling based on androgen-deprivation therapy (ADT; refs. 1, 2) and receptor antagonists (3). This approach is initially remarkably effective but eventually leads to the conversion of the disease to castration-resistant prostate cancer (CRPC). The conversion to CRPC is attributed to the expression of splice variants of the AR and recruitment of alternative signaling pathways that the receptor uses to promote the growth and survival of malignant cells while escaping the effects of a range of inhibitory drugs and hormonal therapies (4, 5). Notably, ADT frequently induces the secondary emergence of highly aggressive prostate phenotypes with neuroendocrine features, including the expression of markers such as chromogranin A and synaptophysin and suppression of PSA (6, 7). Thus, while neuroendocrine prostate cancer (NEPC) is considered an aggressive subtype of the primary tumor (8), prostate disseminated tumor cells (DTC) could acquire NEPC features following ADT and during the most common clinical manifestation of the prostate cancer. As NEPC cells are independent of androgens for their growth, ADT might provide them with a selective survival advantage (9). In fact, the percentage of neuroendocrine cells sharply increases in high-grade and advanced stage prostate tumors upon establishment of ADT (10, 11). More importantly, cancer cells lacking AR and/or PSA expression are frequently detected in bone metastatic lesions among AR+ and PSA+ malignant phenotypes (12). These DTCs with acquired NEPC phenotypes could be very effective in colonizing the bone during the initial stages of metastasis and be responsible for establishing a metastatic niche that would subsequently also support the growth of AR+/PSA+ cancer cells.

Bone-metastatic disease is often fatal for patients with prostate cancer and its treatment remains an unmet medical need. The molecular underpinning for the establishment and progression of secondary bone lesions has been only partially elucidated. A better understanding of the factors regulating bone colonization, particularly the autocrine and paracrine interactions of DTCs with the surrounding stroma, will help to find more effective therapies for the management of metastatic patients. We previously reported that the ability of prostate cancer cells to generate skeletal tumors in animal models correlates with the expression of the platelet-derived growth factor receptor alpha isoform (PDGFRα; refs. 13–15). Here, we show that PDGFRα upregulates the expression of 3 genes that were associated with the occurrence of skeletal metastases in animal models inoculated in the arterial blood circulation with human prostate cancer cells. Among these genes, interleukin (IL)-1β is independently accountable for dictating bone-metastatic behavior and was also detected in human specimens of both primary prostate cancer and bone-metastatic lesions. Finally, the coexpression of IL-1β with the NEPC marker synaptophysin in prostate cells detected in human skeletal lesions corroborates the idea that this cytokine plays a role in the progression of bone-metastatic tumors affecting patients with prostate cancer treated with ADT.

NIH-3T3 and DU-145 cells were obtained from American Type Culture Collection (ATCC) and passaged in our laboratory for less than than 6 months after resuscitation. PC3-N and PC3-ML sublines were derived from the parental PC-3 cell line as previously described (16). Both sublines were tested by Idexx Radil on May 2012 by short tandem repeat-based DNA fingerprinting and confirmed to be of human origin without mammalian interspecies contamination. The alleles for 9 different markers were determined and the genetic profiles of both PC3-ML and PC3-N cells were found identical to the profiles reported for the parental PC3 line deposited with the ATCC. All prostate cancer cells lines were cultured at 37°C and 5% CO₂ in Dulbecco's Modified Eagle Medium (DMEM; Invitrogen) supplemented with 10% FBS (Hyclone) and 0.1% gentamicin (Invitrogen). We cultured all prostate cancer cell lines used in this study for 10 passages and then thawed a new frozen stock to avoid the emergence of genotypic and phenotypic changes (17). Cells were genetically engineered to stably express EGFP using a lentiviral vector (AmeriPharma). Bone marrow-derived human mesenchymal stem cells (MSC; Lonza) were used between passage 5 and 8 and cultured in MSC growth medium [α-MEM (Invitrogen) supplemented with 10% FBS, 1 ng/mL basic fibroblast growth factor (R&D Systems), and 0.1% gentamicin].

Cell lysates were obtained and SDS-PAGE and Western Blot analysis conducted as previously described (18) with few modifications. Membranes were blotted with antibodies targeting IL-1β (SC-7884, Santa Cruz), actin (A-2066, Sigma-Aldrich), COX-2 (ab15191, Abcam), elafin (SC-20637, Santa Cruz), and GAPDH (D16H11, Cell Signaling Technology). Primary antibody binding was detected using a horseradish peroxidase-conjugated secondary antibody (Pierce; Thermo Scientific). Chemiluminescent signals were obtained using SuperSignal West Femto reagents (Pierce) and detected with the Fluorochem 8900 imaging system and related software (Alpha Innotech, ProteinSimple).

Conditioned media were obtained according to ref. 19. In brief, 7.5 × 10⁵ PC3-ML cells were plated in 15 mL of DMEM supplemented with 10% FBS and 0.1% gentamycin and cultured for 5 days. The medium from each dish was then collected and centrifuged at 2,000 rpm for 10 minutes and then used fresh as described below.

For bone cell treatment experiments, MSCs were plated at least 48 hours before treatment; when 70% confluent, cells were incubated in a 1:1 mixture of conditioned medium and MSC growth medium for 48 hours. To pharmacologically induce the overexpression of COX-2 in MSCs, cells were exposed to 0.1 ng/mL of IL-1β (R&D Systems). Cells were preincubated with the IL-1R inhibitor Anakinra (Amgen) at a 10 μg/mL concentration for 30 minutes before being exposed to IL-1β.

The concentrations of IL-1β and CXCL6 were measured by ELISA following the manufacturers' protocols. In brief, same numbers of cells (5 ×10⁵) were plated in 35 mm culture dishes; the next day, the medium was replaced with 1 mL of DMEM supplemented with 10% FBS and 0.1% gentamycin and cultured for 24 hours. The supernatants were then collected, the adherent cells in each dish measured again, and IL-1β or CXCL6 protein concentrations measured using Quantikine kits (R&D Systems). The ELISA data were normalized to the number of cells counted in each dish when the supernatants were collected.

Virus containing Mission TRC lentiviral vectors shRNA (Sigma-Aldrich) with sequence 5′-CGGCCAGGATATAACTGACTT-3′ were used to knockdown IL-1β expression. Subconfluent cell cultures were infected overnight in the presence of 8 μg/mL polybrene (Millipore). The successfully infected cells were selected for the ability to proliferate in media containing 600 μg/mL of G418 (Invitrogen) and protein expression was validated by Western blot analysis using an antibody against IL-1β. Cells transduced with a TRC lentiviral vector carrying a noncoding short-hairpin RNA (shRNA) were used in control experiments.

To prepare the IL-1β–overexpressing retrovirus, a mixture of pLXSN vector containing 50 ng of IL-1β plasmid and 8 μL of Lipofectamine 2000 (Invitrogen) was incubated at room temperature for 30 minutes. The transfection mix was transferred to phoenix cells that were approximately 70% confluent. After 16 hours, the transfection medium was replaced with growth medium containing 10% serum and the virus was harvested at 38 hours after transfection. The virus-containing medium was pooled, centrifuged at 44,000 rpm for 20 minutes, and the supernatant was used to infect PC3-N and DU-145 cells. Cells were selected for the ability to proliferate in media containing G418 (0.6 mg/mL) and the infection was further validated by Western blot analysis using an antibody against IL-1β. Cells transduced with an empty pLXSN vector were used in control experiments.

Five-week-old male immunocompromised mice (CB17-SCRF) were obtained from Taconic and housed in a germ-free barrier. At 6 weeks of age, animals were anesthetized with 100 mg/kg ketamine and 20 mg/kg xylazine administered by intraperitoneal route and successively inoculated in the left cardiac ventricle with cancer cells [5 × 10⁴ in 100 μL of serum-free DMEM/F12 (Invitrogen)]. Cell inoculation was conducted using a 30-gauge needle connected to a 1 mL syringe. The delivery of the cell suspension in the systemic blood circulation was validated by the coinjection of blue fluorescent 10 μm polystyrene beads (Invitrogen-Molecular Probes). Animals were randomly assigned to different experimental groups and sacrificed at specified time-points following inoculation. Organs were harvested and prepared as described below and tissue sections inspected blindly for metastatic lesions. The homogeneous and numerically consistent distribution of the beads in adrenal glands and lungs collected at necropsy and inspected by fluorescence microscopy were used as discrimination criteria for the inclusion of animals in the studies.

All experiments were conducted in accordance with NIH guidelines for the humane use of animals. All animal protocols were approved by the Drexel University College of Medicine Institutional Animal Care and Use Committee (Philadelphia, PA).

Bones and soft-tissue organs were collected and fixed in 4% paraformaldehyde solution (Electron Microscopy Sciences) for 24 hours and then transferred into fresh formaldehyde for an additional 24 hours. Soft tissues were then placed either in 30% sucrose for cryoprotection or 1% paraformaldehyde for long-term storage. Bones were decalcified in 0.5M EDTA (Fisher Scientific) for 7 days followed by incubation in 30% sucrose. Tissues were maintained at 4°C for all aforementioned steps and frozen in optimum cutting temperature medium (Sakura Finetek) by placement over dry-ice chilled 2-methylbutane. Serial sections of 80 μm thickness were obtained using a Microm HM550 cryostat. Femur and tibia in each knee joint were cut entirely through, resulting in approximately 30 sections per specimen made available for analysis.

Fluorescent images of skeletal metastases were acquired using a Zeiss AX70 microscope (Carl Zeiss) connected to a Nuance Multispectral Imaging System. Digital images were analyzed and processed with the Nuance Software (v. 2.4). Microscope and software calibration for size measurement was conducted using a TS-M2 stage micrometer (Oplenic Optronics).

Total RNA was purified with Qiagen RNeasy Mini Kit (Qiagen). RNA quality control for each set of samples was conducted using a BioAnalyzer (Agilent). Two rounds of amplification were used according to the Affymetrix Two-cycle Amplification protocol using 25 ng for total RNA. Aliquots of 15 μg of amplified biotinylated RNA were hybridized to 1.0 Human Gene ST arrays (Affymetrix). Arrays were scanned using the GeneChip Scanner 3000 (Affymetrix). The Robust Multichip Analysis (RMA) algorithm was applied to all array data (20). GeneSpring software version 11.5 was used to filter and complete the statistical analysis. To analyze the microarray data, CEL files were loaded to GeneSpring, and probeset summarization was conducted using the RMA 16 algorithm. For each probe, the median of the log-summarized values from all the samples was calculated and subtracted from each sample. After processing and normalization, the resulting 28,869 genes included in the 1.0 Human Gene ST arrays were filtered to remove very low or saturated signal values. Each entity was filtered on raw data by percentile with an upper cutoff of 100 and a lower cutoff 20. The resulting new entities were then subjected to statistical analysis using an unpaired t test with a P value fixed at 0.05. Finally, a higher stringency filter was applied to the resulting entities using a 2.0 fold-change cutoff and the Benjamin–Hochberg multiple testing corrections. The microarray data were submitted to the gene expression omnibus (GEO) data repository and can be accessed with the number GSE43332.

The oncomine database (available online) was searched for IL-1β, CXCL6, and PI3 genes. The datasets containing expression data for each gene were filtered to display upregulation in prostate cancer versus normal prostate tissue with P < 0.05. If more than one dataset passed the filters, we conducted a meta-analysis to obtain a P value.

Commercially available human tissue microarrays (TMA; PR956, PR8010, PR483, PR751) contained 192 prostate tissue cores and were obtained from US Biomax. Two additional existing TMAs containing 35 deidentified human prostate cancer specimens as well as 7 deidentified bone tissue specimens with metastatic prostate cancer were obtained from the archives of the Department of Pathology at Drexel University College of Medicine.

Immunohistochemical detection was conducted using antibodies against IL-1β (ab2105, AbCam), PSA (ER-PR8, Cell Marque), and Synaptophysin (SP11, Ventana) all diluted 1:50 on formalin-fixed paraffin-embedded sections. The staining conditions using the BenchMark ULTRA IHC/ISH Staining module were as follows: antigen retrieval (pH 8.1) using CC1 reagent 64 minutes, followed by primary antibody incubation for 40 minutes at 37°C, and then staining with the XT, Ultraview Universal DAB Detection Kit. Interpretation and scoring was conducted by 2 clinical pathologists (F.U. Garcia and M.I. Lioni) using light microscopy. Staining intensities were scored as follows 0, no staining; 1, weak staining; 2, moderate staining; and 3, strong staining. Only samples that showed 40% or more of cellular staining were used for the analysis.

We analyzed number and size of skeletal metastases between 2 experimental groups using a 2-tailed Student t test and between multiple groups using a one-way ANOVA test. A value of P ≤ 0.05 was deemed significant. The results of TMA staining were subjected to χ² analysis and plotted in a contingency table.

We conducted genome-wide comparative transcriptome analyses of human prostate cancer cell lines that differ in PDGFRα expression and propensity to establish tumors in the skeleton of animal models. First, we examined genes that were differentially regulated in the highly bone-metastatic PC3-ML cells and their low-metastatic counterpart PC3-N cells (16). Both sublines were derived from the PC3 parental cell line, which was originally obtained from a skeletal lesion in a patient with grade 4 metastatic prostate adenocarcinoma treated with ADT (21). These cells lack AR and PSA and their androgen-independent status is associated with the expression of neuroendocrine markers (22). We have previously reported that PC3-ML cells directly inoculated into the arterial circulation of severe combined immunodeficient (SCID) mice generate large skeletal lesions in more than 90% of animals (Supplementary Fig. S1A and S1C; ref. 14). These cells express high levels of PDGFRα, in contrast to PC3-N cells, which show low metastatic potential in the same animal model and express significantly lower levels of the PDGFRα (13). Microarray data analysis revealed that 16 genes were differentially expressed between high-metastatic PC3-ML and low-metastatic PC3-N cells (Supplementary Fig. S2A). Because we previously found that the overexpression of PDGFRα in PC3-N cells induces a bone-metastatic behavior identical to that of PC3-ML cells (14, 23), we investigated the genes differentially regulated between PC3-N and PC3-N(Rα) (Supplementary Fig. S2B). This approach identified 7 genes that were similarly upregulated in highly metastatic PC3-ML and PC3-N(Rα) cells as compared with low metastatic PC3-N cells (Supplementary Fig. S2C and S2D). These results were significantly strengthened by the analysis of DU-145 cells, which were isolated from a brain rather than a skeletal metastatic lesion in a patient with prostate cancer (24, 25). We have previously shown that DU-145 cells lack PDGFRα (13) and fail to survive longer than 3 days as DTCs after homing to the mouse bone marrow (Supplementary Fig. S1A and S1B; ref. 14). Interestingly, and in contrast to PC3-N cells, the exogenous expression of PDGFRα did not promote metastatic bone-tropism of DU-145 cells in our model (data not shown). Consistent with this observation, PDGFRα expression in DU-145 cells upregulated 5 genes that did not overlap with the 7 putative prometastatic genes identified in PC3 cells (Supplementary Fig. S3), suggesting their lack of involvement in the bone-metastatic behavior of prostate cancer cells. To refine these findings and compensate for the inherent genetic disparity of PC3 and DU-145 cells, we isolated 2 single-cell progenies from PC3-ML cells. When tested in our animal model, both PC3-ML clone 1 and clone 3 were highly bone-metastatic (Supplementary Fig. S4). A comparative analysis of the genes differentially regulated between these 2 clonal cell lines and our newly identified 7 gene set resulted in a final cohort of 3 upregulated genes: the inflammatory cytokine IL-1β, the chemokine CXCL6, and the leukocyte protease inhibitor elafin. These 3 genes consistently correlated with both PDGFRα expression and aggressive bone-metastatic behavior in our model. Proteomic approaches validated the transcriptome analysis and confirmed the data relative to IL-1β (Supplementary Fig. S5) as well as CXCL6 and elafin (Supplementary Fig. S6). These results were corroborated by mining prostate cancer datasets publically available through the Oncomine repository, showing that IL-1β, CXCL6, and elafin are significantly upregulated in tumors as compared with normal prostate tissues (Table 1, top). Furthermore, a meta-analysis indicated a strong association of both IL-1β and CXCL6 with prostate cancer with Gleason scores (≥7; Table 1, bottom). In light of these observations, we screened human tissue arrays including 227 cases of prostate adenocarcinoma for IL-1β protein expression and correlated signal intensities with the Gleason score attributed to each tissue specimen (Fig. 1). This approach validated the results from the Oncomine analysis and conclusively shows that prostate tumors with intermediate and high Gleason scores, which have the highest propensity to metastasize (26, 27), express increased levels of IL-1β as compared with tumors with Gleason scores (<7) or normal tissues (Supplementary Table S1). Remarkably, IL-1β inhibits the expression of both PSA (28) and AR (29) in prostate cancer cells, thus reproducing features observed in PC3-ML cells as well as NEPC cells either primarily or as a consequence of ADT (7, 30). Therefore, we reasoned that IL-1β could be a crucial player in the establishment of skeletal secondary lesions by prostate cancer. To challenge this hypothesis, we used a preclinical animal model of metastasis and used shRNA to deplete IL-1β in PC3-ML cells to levels of expression and secretion comparable with those observed in PC3-N cells (Fig. 2A and B). The resulting PC3-ML(sh-IL-1β) cells were delivered in the systemic arterial circulation of mice euthanized 4 weeks later and showed significantly impaired metastatic abilities. The inspection of femora and tibiae of inoculated animals showed that PC3-ML and PC3-ML(sh-IL-1β) cells produced bone metastases in a comparable number of animals (Fig. 2C). However, the lesions generated by PC3-ML(sh-IL-1β) cells were 70% smaller than those observed in mice inoculated with PC3-ML cells expressing endogenous levels of IL-1β. (Fig. 2D and E). It is plausible that silencing IL-1β in combination with one or both of the other 2 genes identified in this study provides superior inhibition of metastatic progression than silencing IL-1β alone. Our laboratory is currently actively pursuing the preclinical validation of this paradigm.

To further define the role of IL-1β in skeletal metastasis, we conducted complementary experiments in which this cytokine was exogenously overexpressed in prostate cancer cells with demonstrated inability to progress to macroscopic bone lesions. We first studied PC3-N cells, which routinely produce small lesions in only 20% of animals inspected at 3 weeks after inoculation and regress thereafter (Supplementary Fig. S1) and (14). After homing to the skeleton from the blood circulation, PC3-N(IL-1β) cells were able to fully progress into tumors comparable in number and size to the lesions produced by PC3-N(Rα) cells (Fig. 3; refs. 14, 15). More importantly, analogous results were obtained with DU-145 cells, which are widely reported to lack bone-tropism in mouse models (31). We have previously shown that this lack of metastatic behavior is caused by the inability of these cells to survive for more than 3 days after homing to the bone marrow (14). DU-145 cells do not endogenously express IL-1β (28,32); interestingly, upon overexpression of this cytokine (Fig. 4A and B) these cells generated skeletal metastases in 40% of mice examined at 4 weeks after inoculation (Fig. 4C). Although these lesions were smaller in size than the skeletal tumors produced by PC3-N(IL-1β) cells after the same time interval (Fig. 4D and E), these data provide compelling evidence that, in addition to potentiating the weak bone-tropism of PC3-N cells, IL-1β can induce de novo metastatic behavior in prostate cancer cells. Furthermore, the increase in tumor area observed for DU-145 (IL-1β) lesions detected 4 weeks compared with those at 2 weeks after inoculation (Fig. 4D) clearly indicates that IL-1β promotes both survival and proliferation of metastatic cells in the skeleton.

Despite originating from different metastatic sites in patients with prostate cancer, PC3 and DU-145 cells both express neuroendocrine markers (33). The acquisition of a neuroendocrine phenotype is frequently induced by the ADT commonly used for patients with advanced prostate adenocarcinoma (9, 34, 35) and is also observed in transgenic animal models of prostate cancer upon castration (36). These observations would suggest that the convergence of ADT-induced neuroendocrine transdifferentiation and increased expression of IL-1β underpins the propensity of prostate cancer cells to colonize the skeleton and eventually progress to secondary bone lesions. To test this model, we used an ex vivo analysis of skeletal lesions obtained from patients with advanced prostate cancer and found that all specimens stained positively for IL-1β. Interestingly, the intensity of the signal showed an inverse correlation with PSA expression (Fig. 5A), and two specimens also stained positively for synaptophysin (Fig. 5B). These results provide strong support for a role of IL-1β overexpression in bone-metastatic growth of prostate cancer cells and suggest a frequent association of this cytokine with evident NEPC features of skeletal metastases from prostate adenocarcinoma.

The secondary tropism of metastatic tumors is the result of compatibility between DTCs and the tissue microenvironment of the colonized organs (37, 38). Because of the stimulatory effect exerted by IL-1β on the bone-resorption activity of osteoclasts (39), a plausible scenario would include this cytokine supporting secondary skeletal lesions by promoting bone matrix turnover and increasing the availability of trophic factors for the disseminated cancer cells. Indeed, this mechanism is commonly targeted by bisphosphonates and RANKL inhibitors in the clinical management of patients with metastatic breast and prostate cancer (40, 41). Because we and others have shown that osteoclasts are not involved in these early phases of bone marrow colonization (42–14), the newly identified prometastatic role of IL-1β might be exerted through either autocrine trophic stimulation of cancer cells, or a more complex paracrine recruitment of surrounding bone stromal cells other than osteoclasts. In the latter scenario, IL-1β would stimulate cells of the bone stroma and induce them to reciprocate with increased or ex novo production of trophic factors that would support the survival and growth of DTCs (44). In a recent study, Weinberg and colleagues have shown that IL-1β derived from carcinoma cells increases the expression of the COX-2 enzyme in MSCs of the tumor stroma and the consequent secretion of prostaglandin E₂ (PGE₂). The effect of PGE2 is exerted through a direct paracrine stimulation of the cancer cells and the autocrine induction of cytokine secretion from MSCs (19). Thus, we sought to ascertain whether a similar model could be applied to the prometastatic role of IL-1β revealed by our studies. To this end, we exposed human MSCs to media conditioned by PC3-ML cells and measured the effects on COX-2 expression after 48 hours. We concluded that the increase in COX-2 observed in MSCs was induced by the IL-1β secreted by PC3-ML cells, as the IL-1R antagonist Anakinra (45) was able to completely prevent it (Fig. 6). On the basis of this observation, further experiments are being conducted in our laboratory to conclusively establish the role of COX-2 and PGE₂ in the bone metastatic progression of prostate cancer cells.

Our study emphasizes the importance of early survival of DTCs mediated by IL-1β for successful lodging and initial colonization of the bone. These events are particularly relevant for the seeding of additional tumors by circulating tumor cells dislodged from preexisting lesions and commonly detected in the peripheral blood of patients with metastatic disease (46–48). Thus, disrupting the functional interactions between IL-1β and its receptors, most likely IL-1R, would substantially attenuate the progression of prostate cancer at the skeletal level and possibly reduce the secondary involvement of other organs.

Notably, therapeutics that target either IL-1β or IL-1R are currently available and prescribed for skeletal inflammatory conditions of non-neoplastic etiology such as rheumatoid arthritis (45, 49, 50). The evidence provided by our study should lead to the repositioning of these drugs in the clinic and rapidly translate into novel strategies for treating existing metastatic skeletal lesions as well as preventing ongoing seeding of additional lesions both in bone and visceral organs.

No potential conflicts of interest were disclosed.

Conception and design: Q. Liu, K. Shahriari, A. Fatatis

Development of methodology: Q. Liu, M.R. Russell, K. Shahriari, F.U. Garcia

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Q. Liu, M.R. Russell, K. Shahriari, F.U. Garcia

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Q. Liu, M.I. Lioni, F.U. Garcia, A. Fatatis

Writing, review, and/or revision of the manuscript: Q. Liu, K. Shahriari, D.L. Jernigan, M.I. Lioni, F.U. Garcia, A. Fatatis

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Q. Liu M.R. Russell, D.L. Jernigan, A. Fatatis

Study supervision: A. Fatatis

The authors thank Dr. Olimpia Meucci (Drexel University College of Medicine) for input and commentary and gratefully acknowledge the expertise and advice with the gene-expression analyses provided by Dr. Paolo Fortina (Dept. of Cancer Biology, Thomas Jefferson University and Kimmel Cancer Center, Philadelphia, PA), and the excellent technical assistance of Shankar Addya and Kathryn Scott (Cancer Genomic Microarray Facility, Kimmel Cancer Center, Philadelphia, PA).

This work was supported by the Prostate Cancer Program of the Department of Defense with the grant W81XWH-09-1-0724 (A. Fatatis).

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.