Androgen Receptor Variants Mediate DNA Repair after Prostate Cancer Irradiation

The combination of radiotherapy and androgen deprivation therapy (ADT) is superior to radiotherapy alone for treatment of patients with clinically localized, intermediate-risk (PSA, 10–20 ng/mL, cT2 and Gleason sum = 7) and high-risk (PSA ≥ 20 ng/mL, ≥cT3 and Gleason sum ≥8) prostate cancer (1, 2). Clinically, patients receive 2 to 3 months of initial ADT, followed by combined radiotherapy and ADT. High-risk patients may then receive an additional 2 to 3 years of ADT.

Biologically, ADT abrogates the ability of the androgen receptor (AR) to promote the cellular DNA damage repair (DDR; refs. 3, 4). When AR signaling is blocked by ADT, prostate cancer cells cannot efficiently activate DDR. Thus, the combination of ADT and radiotherapy enhances DNA damage and lethality of radiotherapy (3–5).

Despite use of combined ADT and radiotherapy, biochemical disease recurrence is common (up to 50% of high-risk prostate cancer patients; ref. 6). Recurrence is associated with a higher likelihood of metastatic progression and death from prostate cancer (7). Although recurrence may be a function of advanced disease at the time of radiotherapy, the finding of viable radioresistant prostate cancer in salvage prostatectomy specimens suggests that some prostate cancers may be intrinsically resistant to ADT + radiotherapy (8). These data suggest that an understanding of molecular mechanisms of radioresistance may enable more effective radiotherapy for localized prostate cancer.

Emerging evidence supports a role for AR variants (ARV) in the development of resistance of prostate cancer to ADT and second-generation AR-targeting therapies (9–11). ARVs lack the AR ligand–binding domain (LBD) and are not responsive to drugs targeting AR-LBD. ARVs may be formed by AR splicing (plastic) that can be induced by ADT (10), and/or AR gene rearrangements (permanent; 11). ARV expression serves as a predictive biomarker for response to enzalutamide or abiraterone acetate (12). ARVs are nuclear, constitutively active, bind to similar androgen response elements (ARE), as that engaged by full-length AR (ARfl; ref. 13), and interact with similar coregulators as ARfl (14–16). Thus, ARVs may substitute for ARfl, when the ARfl is effectively blocked by ADT (13).

We reasoned that the regimen of 2 to 3 months of initial ADT prior to radiotherapy was sufficient to induce ARVs in prostate cancer prior to radiotherapy (10, 17). Evaluation of the primary prostate tissues from the CALGB 90203 study indicates that neoadjuvant ADT induces both ARfl and ARV expression (M. Gleave, personal communications). Using short-term (48 hours) ex vivo cultures of primary prostate tissues, we noted that enzalutamide was able to induce significantly more AR transcripts containing the AR DBD (reflecting both ARfl and ARV expression) than AR LBD (ARfl expression only; Supplementary Fig. S1A and S1B). Taken together, these data suggest that ADT may induce ARVs in human prostate cancer tissues. We reasoned that these induced ARVs may drive DDR and represent a potential mechanism of resistance to combined ADT and radiotherapy.

22Rv1, VCaP, and PC3 cell lines were obtained from ATCC in 2014 and authenticated by the ATCC using short tandem repeat profiling. C4-2 was a kind gift from Dr. Leland Chung (Cedar-Sinai Medical Center, Los Angeles, CA) in 2013. R1-D567 and R1-AD1 were kind gifts from Dr. Scott Dehm (University of Minnesota, Minneapolis, MN) in 2014 (13). Provenance of C4-2, R1-D567, and R1-AD1 was verified by short tandem repeat profiling in July or August 2016 by Cell Bank Australia. All cell lines underwent regular (approximately every 4 months) testing using a custom PCR-based assay to ensure lack of contamination with the most common Mycoplasma and Acholeplasma strains. All experiments with cell lines were performed within 2 to 4 months (or no more than 20 passages) of resuscitation of frozen cell stocks prepared within three passages of receipt. Cells were irradiated using a 137Cs source (Mark 1-68 irradiator; JL Shepherd and Associates) at a dose rate of 3.47 Gy/minute. NU7441, a specific DNAPKc inhibitor, was obtained from Selleckchem.

AR siRNAs and DNA-PKc smart pool were purchased from Dharmacon.

Genome-wide binding of RNA pol II (BioLegend, cat. # 920101) and AR (Santa Cruz Biotechnology, N-20, #sc-816) in R1-D567 cells was studied by chromatin immunoprecipitation (ChIP) followed by high-throughput DNA sequencing (ChIP-seq). Following 1% formaldehyde cross-linking for 8 minutes, one tenth volume of 1.25 mol/L glycine was added for 5 minutes at room temperature. Chromatin was sheared to 300 bp (Bioruptor Pico, Diagenode). ChIP antibodies were tagged to protein G magnetic beads with IgG and input DNA as controls. Protein-bound DNA was washed, reverse crosslinked, and eluted using the IPure kit (Diagenode). Purified DNA was analyzed by fluorometry (Qubit 3.0, Invitrogen), and 40% of ChIP-DNA was used for sequencing library using KAPA HyperPrep Kit (KAPA Biosystems) and TruSeq adapters (Integrated DNA Technologies). PCR products were selected using Ampure beads (Beckman Coulter). Quality check of purified libraries was carried out using Bioanalyzer 2100 (Agilent) and sequenced on Illumina NextSeq 500 Sequencer with read length of single-end 75bp.

Following treatment of R1-D567 cells with 2 Gy radiotherapy, cell lysates were incubated with Dynabeads Protein G (Invitrogen) precoupled with AR antibody (Sigma) or an irrelevant IgG control, eluted at 95°C, separated by SDS-PAGE, proteolytically digested overnight with trypsin (Promega) following reduction and alkylation with DTT and iodoacetamide (Sigma-Aldrich). Samples then underwent solid-phase extraction cleanup with Oasis HLB plates (Waters) and were analyzed by LC/MS-MS, using an Orbitrap Fusion Lumos (Thermo Electron) coupled to an Ultimate 3000 RSLC-Nano liquid chromatography systems (Dionex). MS was operated in positive ion mode with a source voltage of 2.2 kV and capillary temperature of 275°C. MS scans were acquired at 120,000 resolution and up to 10 MS/MS spectra were obtained in the ion trap for each full spectrum acquired using higher energy collisional dissociation for ions with charge 2-7. Dynamic exclusion was set for 25 seconds. Raw MS data files were converted to a peak list format and analyzed using the central proteomics facilities pipeline (CPFP), version 2.0.3 (18). Label-free quantitation of proteins across samples was performed using SINQ normalized spectral index Software (18).

Western blotting was performed as described previously, using antibodies to AR (Santa Cruz Biotechnology, N-20, #sc-816, or Santa Cruz Biotechnology, 441, #sc-7305), or AR-V7 (Precision Antibody, #AG10008, 1:1,000), phosphorylated histone H2AX (γ-H2AX; Millipore, 05-636; 1:1,000), actin (Sigma-Aldrich, A2066; 1:5,000), pS2056-DNA-PKc (Abcam, ab18192; 1:400), DNA-PKc (Abcam, ab32566; 1:1,000), and horseradish peroxidase–conjugated donkey anti-rabbit and sheep anti-mouse, AlexaFluor 594 anti-mouse and AlexaFluor 488 anti-rabbit antibodies.

Cells were treated with enzalutamide for 24 hours and then treated with increasing doses of radiation (0, 1, 2, and 3 Gy). Colony formation assay was performed as described previously (3). Colony formation assay was also performed using siRNAs. After 72 hours, cells were plated for colony formation assay and Western blot analysis.

The number of phospho-γH2AX foci (green) and 53BP1 (red) was determined at each time point (average of 200 nuclei), and the percentage of foci remaining was plotted against time to obtain double-strand break (DSB) repair kinetics. Data are represented as mean ± SEM.

Cells were subjected to radiation (10 Gy, 30-minute recovery), harvested, and resuspended in ice-cold PBS (without Mg²⁺ and Ca²⁺). Amount of damaged DNA was visualized by single-cell agarose gel electrophoresis under alkaline conditions following the manufacturer's protocol. Diluted Vista Green DNA dye (100 μL/well) were applied onto slides. Comet images were taken by EVOS fluorescence microscopy (Thermo Fisher Scientific) using a FITC filter.

Cells grown on coverslips were incubated with rabbit polyclonal anti-H2AX (p-S139) antibody (Cell Signaling Technology), followed by incubation with biotinylated swine anti-rabbit IgG (Dako) and streptavidin-conjugated FITC (Dako). Cells were counterstained with 4′,6-diamidino-2-phenylindole (DAPI), analyzed on a Nikon microscope (×40 magnification) equipped with a CCD camera and processed using ImageJ software (NIH, Bethesda, MD). Exposure time, binning, microscope settings, and light source intensity were kept constant. Nuclei were segmented on the basis of DAPI staining and then signal-integrated density of γ-H2AX and 53BP1 staining quantified using ImageJ software. More than 200 γ-H2AX–positive and 53BP1-positive cells were analyzed per experiment per condition.

Cells grown on coverslips were fixed with methanol at −20°C for 10 minutes followed by a 5-minute extraction in 0.3% Triton X-100 in PBS. In situ proximity ligation was performed using a Duolink Detection Kit in combination with anti-mouse PLUS and anti-rabbit MINUS PLA probes, according to the manufacturer's instructions (Sigma-Aldrich Duolink). Nuclear foci were imaged using a Nikon E600 Eclipse microscope equipped with a 40× lens. The number of nuclear foci/cell was quantified using ImageJ. More than 50 cells were analyzed per experiment per condition.

Total RNA was isolated using RNeasy Mini Kit according to the manufacturer's instructions (Qiagen) and reverse-transcribed using the cDNA Synthesis Kit (Bio-Rad). Quantitative PCR used iQ SYBR Green Supermix (Bio-Rad) on an iCycler thermal cycler (Bio-Rad) with a denaturation step at 95°C for 3 minutes, followed by 40 cycles of amplification at 95°C for 30 seconds, 60°C for 30 seconds, and 72°C for 60 seconds. Primer sequences are provided in Supplementary Fig. S1C. Fold changes in mRNA expression levels were calculated using the comparative C_t method.

Male athymic nude mice (nu/nu, 5–6 weeks old) were injected (1 × 10⁶ cells in 100 μL 50% Matrigel,) subcutaneously into the right posterior flanks. Mice were castrated when tumor size reached 200 mm³. Treatment groups (5 animals/group) included untreated control (treated with 0.9% saline), those treated with radiation (2 Gy/d, 2 days), or combined treatment with NU7441 and ionizing radiation (IR). NU7441 (25 mg/kg) was administered intraperitoneally 2 hours before focal IR. All experiments were conducted under Institutional Animal Care and Use Committee of UTSW-approved guidelines for animal welfare.

Numerical data were analyzed using parametric and nonparametric statistical tests, namely Student t test and split-plot (mixed-design) ANOVA test. Significance thresholds are indicated in figure legends.

To evaluate the ability of either ARfl or ARVs alone to mediate DDR, we used transcription activator–like effector nuclease genetically engineered R1-AD1 and R1-D567 cell lines derived from CWR-R1 prostate cancer cells (Supplementary Fig. S2A; refs. 13, 19). Whereas R1-AD1 cells express only ARfl, R1-D567 cells only express the constitutively active ARV, ARv567es (due to genomic deletion of AR exons 5–7; ref. 19). Following IR, we noted induction of DSBs in both cell lines, as evidenced by foci of γ-H2AX and 53BP1. Similar DNA repair kinetics were noted in both R1-AD1 and R1-D567 cells, with complete resolution of IR-induced γ-H2AX and 53BP1 foci within 24 hours (Fig. 1A). The combination of enzalutamide and IR delayed and decreased DNA repair in R1-AD1 cells (30%–40% γ-H2AX foci at 24 hours; Fig. 1B) but not in R1-D567 cells (<5% γ-H2AX foci at 24 hours; Fig. 1A). Depletion of ARfl or ARv567es, using siRNA (Supplementary Fig. S2B), altered DNA repair kinetics in R1-AD1 and R1-D567 cells, respectively (Supplementary Fig. S2C). Our use of enzalutamide rather than ADT in prostate cancer cells is both to achieve a more complete AR blockade and to reflect the evolving clinical paradigm with earlier use of second-generation anti-androgens. Overall, these data suggest that a cell line expressing ARVs alone can mediate DDR and that ARV-mediated DDR is not affected by enzalutamide.

Following IR, clonogenic survival of R1-D567 cells was not affected by enzalutamide (Fig. 1C). Importantly, clonogenic survival of 22Rv1 cells, which express both ARfl and ARv7 (Supplementary Fig. S2A), was not affected neither by enzalutamide (Fig. 1C) nor by CRISPR-mediated knockdown of ARfl or ARv7 (Supplementary Fig. S2D and S2E). In contrast, the combination of IR and enzalutamide decreased clonogenic survival of both R1-AD1 and C4-2 cells (which express only ARfl), in comparison with IR alone (Fig. 1D). These data validate prior findings of a synergism between ADT and radiotherapy in mediating DNA repair in prostate cancer cells expressing ARfl and suggest that prostate cancer cells that express ARVs either alone or in combination with ARfl do not exhibit this same synergism.

When R1-D567 subcutaneous xenografts were subject to short-course radiotherapy, their rate of growth slowed down initially but rebounded quickly (Supplementary Fig. S3A). A similar profile of response was noted with R1-AD1 subcutaneous xenografts (Supplementary Fig. S3B). We observe that R1-D567 xenografts grow faster than matched R1-AD1 xenografts; this difference becomes even more striking upon IR treatment. This suggests that ARVs may confer enhanced radioresistance as compared with ARfl. Using siRNA, we noted that decreased expression of either ARfl or ARv7 had minimal impact on clonogenic survival following IR in 22Rv1 cells (Fig. 2A). Knockdown of both ARfl and ARv7 was needed to enhance the effect of IR in clonogenic survival studies in 22Rv1 cells (Fig. 2A). These data indicate that inhibition of both forms of AR is needed to effectively increase radiotherapy-mediated cell kill in 22Rv1 cells.

Proximity ligation assays (PLA) indicated that ARv567es was recruited within minutes of IR to sites of DNA damage, as evidenced by its interaction with γ-H2AX foci in R1-D567 cells (Fig. 2B). Every irradiated (2 Gy) R1-D567 cells displayed >5 foci/cell. Double immunofluorescence staining indicated overlap between most γ-H2AX foci and ARv567es staining within the nucleus of R1-D567 cells (Supplementary Fig. S4A). No change was noted in AR interaction with a known coregulator ARID1A following IR (Fig. 2B). Similarly, PLA studies in 22Rv1 cells indicate that both ARv7 and ARfl were recruited to DNA damage sites (γ-H2AX) upon IR (Supplementary Fig. S4B). These data suggest that ARfl and ARV can be recruited to sites of DNA damage.

Evaluation of global RNA polymerase II (Pol II) binding via ChIP-Seq analyses indicated that IR dramatically increased Pol II promoter occupancy in R1-D567 cells [Fig. 2C and D; the GEO accession number for RNA polymerase II and AR ChIP-seq data reported in Fig. 2C and D is GSE102164 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE102164)]. Quantitative RT-PCR validated radiotherapy-induced expression of genes involved in homologous recombination and nonhomologous end joining (NHEJ) pathways (Supplementary Fig. S4C). We were intrigued by our observation that AR in R1-D567 cells bound to <100 sites in the genome. A majority of AR-bound sites in R1-D567 cells were associated with ARE motifs, indicating true AR-binding sites. However, evaluation of AR ChIP-seq indicated a decrease in ARv567es binding upon IR treatment in R1-D567 cells (Supplementary Fig. S5A and S5B). These data indicate that although IR globally induces transcription in R1-D567 cells, the ARVs are unlikely to be the primary drivers of this transcriptional program. Pathway analysis of genes differentially occupied by Pol II upon IR treatment indicated upregulation of steroid hormone receptor pathways involving AR, glucocorticoid, estrogen, and aldosterone (Supplementary Fig. S5C). Interestingly, IR treatment did not increase the levels of AR at 16 hours. Our integrative analyses suggested that IR treatment upregulates androgen signaling at a pathway level rather than a gene level, as no single component of the pathway was drastically upregulated (Supplementary Fig. S5B). Rather, moderate upregulation of a subset of individual components contributed to increase in the pathway flux (Supplementary Fig. S5C). These data also explain why multiple steroid hormone receptor pathways (e.g., AR and glucocorticoid) could operate temporally during longitudinal progression of prostate cancer in response to enzalutamide.

As ARV localized to the sites of DNA damage, we hypothesized that evaluation of the ARV interactome could elucidate its mechanisms of action. Using unbiased immunoprecipitation-mass spectrometric (IP-MS) analysis, we identified DNA-PKc as the top binder of ARv567es in R1-D567 cells regardless of IR treatment (Fig. 3A). These data were supported by coimmunoprecipitation studies demonstrating that IR enhanced interaction between endogenous ARfl or ARv567es and DNA-PKc in R1-AD1 and R1-D567 cells, respectively (Fig. 3B). PLA also indicated increased interaction between DNA-PKc and ARfl or ARVs upon IR in prostate cancer cells (Fig. 3C). Mapping studies indicated differential IR-induced interactions between ARfl and ARVs with DNA-PKc: the WHTLF domain appears to be critical for ARV-DNA-PKc interaction, while the AR-LBD may also play a role in ARfl interaction with DNA-PKC (Supplementary Fig. S6A–S6C). These data are supported by the ARΔ1 mutant to interact with DNA-PKc and prior mapping studies indicating AR-LBD and DNA-PKc interaction (20). Importantly, enzalutamide abolished IR-induced interaction between ARfl and DNA-PKc but not ARVs and DNA-PKc interaction (Fig. 3C). Furthermore, IR-mediated induction of ARv567es interaction with DNA-PKc was blocked by a selective DNA-PKc inhibitor, NU7441 (Fig. 3D). In contrast, the interaction between ARv567es and ARID1A was not influenced by IR or NU7441 (Fig. 3D). Taken together, these data indicate that IR enhances the interaction between DNA-PKc and ARfl or ARV.

Comet assays indicated that pretreatment with NU7441 significantly increased DNA fragmentation following IR in R1-D567 cells (Fig. 4A). These data were supported by persistence of residual DSBs after IR in NU7441-treated R1-D567 cells, as evaluated by γ-H2AX and 53BP1 staining (Fig. 4B, shown for γ-H2AX). Mechanistic evaluation indicated that DNA-PKc inhibition with either chemical inhibition (Supplementary Fig. S7–S7C) or siRNA-mediated knockdown blocked NHEJ repair (Supplementary Fig. S7D, E). Importantly, NU7441 enhanced the effect of IR and enzalutamide on prostate cancer cell survival, as shown by clonogenic survival assay in R1-D567, 22Rv1, and R1-AD1 cells (Fig. 4C). The combination of IR, NU7441, and enzalutamide was more potent than IR and enzalutamide or IR and NU7441 in R1-AD1 and 22Rv1 cells. In R1-D567 cells, NU7441 was not additive with enzalutamide but potently decreased survival after IR (Supplementary Fig. S8A). Thus, DNA-PKc inhibition was associated with increased IR-induced DNA damage and decreased cell survival in prostate cancer cells regardless whether they express ARfl alone (Supplementary Fig. S8B), ARV alone (Supplementary Fig. S8C), or both ARfl and ARVs (Supplementary Fig. S8D). These in vitro results suggest that DNA-PKc inhibition can enhance the effect of IR on prostate cancer cells, irrespective of ARfl and ARV expression status.

To validate these findings in vivo, we established subcutaneous R1-D567 xenografts in Nu/Nu mice. When tumors reached 200 mm³, mice were castrated and then were treated with two daily doses of NU7441 or vehicle, followed by 2 Gy IR. The combination of IR and NU7441 was more effective in slowing the R1-D567 tumor growth than IR alone (Fig. 4D). Although DNA DSBs were almost completely repaired within 24 hours in the vehicle-treated irradiated tumors, NU7441-treated irradiated tumors still had significant foci of DDR, suggesting that DNA damage was still present (Fig. 4D). These data support our hypothesis that inhibition of DNA-PKc can effectively synergize with IR in ARV-expressing prostate cancer cells. Overall, our data support a role for combined IR and AR and DNA-PKc inhibition to optimize outcomes of patients undergoing radiotherapy (Supplementary Fig. S8 B–S8D).

In prostate cancer cells that express AR, radiotherapy-induced DDR can be driven by ARfl-mediated signaling. Combined ADT and radiotherapy decrease DDR and increase radiotherapy-mediated cell kill. In this study, we demonstrate using genetically engineered prostate cancer cells that ARVs mediate DDR in response to IR. We have shown that in prostate cancer cells that express ARVs, the combination of radiotherapy and ADT is not more effective than radiotherapy in blocking DDR. These data imply that the expression of ARVs in prostate cancer cells may lead to resistance to combined radiotherapy and ADT. As ADT can rapidly induce ARVs in prostate cancer cells, our study identifies a potential molecular mechanism of resistance to the current paradigm of upfront ADT, followed by radiotherapy and a therapeutic strategy to abrogate this.

Importantly, we show that although IR globally induces transcription in R1-D567 cells, that the ARVs are unlikely to be the primary drivers of this transcriptional program. Our data suggests that the observed effects of ARVs in terms of clonogenic cell survival and radiation resistance are likely due to a direct physical interaction with DNA-PKc. In contrast, we have evidence indicating that ARfl influences DNA repair both directly through its interaction with DNA-PKc and indirectly through its control of the expression of DNA repair genes. Thus, in prostate cancer cells expressing both ARfl and ARV (e.g., 22Rv1), enzalutamide will inhibit the transcriptional regulation and perhaps DNA repair by ARfl, but not the DNA repair functions mediated by ARVs. Our data suggest that transcriptional regulation by AR may be less important to enzalutamide resistance in metastatic castration-resistant prostate cancer (CRPC), with this increased DNA repair function being more critical. We speculate that the role of AR amplification in CRPC is actually to promote DNA repair as well as regulate transcription.

Overall, our data indicate that ARVs and ARfl interact with DNA-PKc and that IR enhances this interaction. We show that DNA-PKc is critical for these DNA repair mechanisms following IR and that the combined blockade of AR and DNA-PKc enhances IR-mediated cell kill, regardless of whether the prostate cancer cell expresses ARfl alone, ARVs alone, or ARfl and ARVs together. DNA-PKc inhibitors block AR–DNA-PK interaction, disrupt DNA repair, and enhance radiotherapy-mediated cell kill. These data establish the importance of DNA-PKc in IR-mediated DDR and establish the rationale for a potential clinical trial combining drugs targeting AR and DNA-PKc in combination with radiotherapy for patients with clinically localized prostate cancer. Although the optimal multimodal strategy incorporating antiandrogens, DNA-PKc blockade, and radiotherapy needs to be clinically validated, our findings would suggest the incorporation of the DNA-PKc antagonist immediately prior to radiotherapy, to enhance cell kill.

F.Y. Feng is the founder of PFS Genomics, reports receiving a commercial research grant from Varian, and is a consultant/advisory board member for Dendreon, EMD Serono, Ferring, Janssen, Medivation/Astellas, and Sanofi. J.S. de Bono has received speakers bureau honoraria from AstraZeneca and is a consultant/advisory board member for Astellas, AstraZeneca, Bayer, Genentech/Roche, Merck Serono, and Orion. No potential conflicts of interest were disclosed by the other authors.

Conception and design: Y. Yin, K. Xu, J.S. de Bono, B.P.C. Chen, G.V. Raj

Development of methodology: Y. Yin, R. Li, S.G. Ramanand, R. Hutchinson, S. Burma, R.S. Mani, G.V. Raj

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Yin, R. Li, K. Xu, Sentai Ding, J. Li, G. Baek, S.G. Ramanand, Sam Ding, Z. Liu, X. Li, R. Hutchinson, J.S. de Bono, R.S. Mani, B.P.C. Chen, G.V. Raj

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Yin, R. Li, K. Xu, Sentai Ding, J. Li, G. Baek, Y. Gao, M.S. Kanchwala, R. Hutchinson, X. Liu, S.L. Woldu, C. Xing, F.Y. Feng, S. Burma, J.S. de Bono, S.M. Dehm, R.S. Mani, B.P.C. Chen, G.V. Raj

Writing, review, and/or revision of the manuscript: Y. Yin, K. Xu, Sam Ding, G. Baek, Y. Gao, M.S. Kanchwala, R. Hutchinson, S.L. Woldu, C. Xing, N.B. Desai, F.Y. Feng, J.S. de Bono, S.M. Dehm, R.S. Mani, B.P.C. Chen, G.V. Raj

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Y. Yin, K. Xu, G. Baek, Y. Gao, M.S. Kanchwala, J.S. de Bono, G.V. Raj

Study supervision: Y. Yin, C. Xing, J.S. de Bono, B.P.C. Chen, G.V. Raj

We thank K. Luby-Phelps for confocal imaging.

This work was supported by the following grants: W81XWH-15-1-0128 (to Y. Yin), W81XWH-14-1-0556 (to Y.Yin), W81XWH-13-2-0093 (to G.V. Raj), and W81XWH-12-1-0288 (to G. V. Raj) from the Department of Defense, 1R01CA200787-01 (to G.V. Raj), R01CA174777 (to S.M. Dehm), RO1CA149461, RO1CA197796 and R21CA202403 (to S. Burma), R01CA166677 (to B.P. Chen), UL1TR001105 (to C. Xing), and R00CA160640 (to R. S. Mani) from the NIH, NNX16AD78G from the National Aeronautics and Space Administration (to S. Burma), and the Mimi and John Cole Prostate Cancer Fund (to G.V. Raj).

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.