The Prostate Cancer TMPRSS2:ERG Fusion Synergizes With the Vitamin D Receptor (VDR) to Induce CYP24A1 Expression-Limiting VDR Signaling

Abstract

A number of preclinical studies have shown that the activation of the vitamin D receptor (VDR) reduces prostate cancer (PCa) cell and tumor growth. The majority of human PCas express a transmembrane protease serine 2 (TMPRSS2):erythroblast transformation-specific (ETS) fusion gene, but most preclinical studies have been performed in PCa models lacking TMPRSS2:ETS in part due to the limited availability of model systems expressing endogenous TMPRSS2:ETS. The level of the active metabolite of vitamin D, 1α,25-dihydroxyvitamin D₃ (1,25D), is controlled in part by VDR-dependent induction of cytochrome P450, family 24, subfamily 1, polypeptide1 (CYP24A1), which metabolizes 1,25D to an inactive form. Because ETS factors can cooperate with VDR to induce rat CYP24A1, we tested whether TMPRSS2:ETS would cause aberrant induction of human CYP24A1 limiting the activity of VDR. In TMPRSS2:ETS positive VCaP cells, depletion of TMPRSS2:ETS substantially reduced 1,25D-mediated CYP24A1 induction. Artificial expression of the type VI+72 TMPRSS2:ETS isoform in LNCaP cells synergized with 1,25D to greatly increase CYP24A1 expression. Thus, one of the early effects of TMPRSS2:ETS in prostate cells is likely a reduction in intracellular 1,25D, which may lead to increased proliferation. Next, we tested the net effect of VDR action in TMPRSS2:ETS containing PCa tumors in vivo. Unlike previous animal studies performed on PCa tumors lacking TMPRSS2:ETS, EB1089 (seocalcitol) (a less calcemic analog of 1,25D) did not inhibit the growth of TMPRSS2:ETS containing VCaP tumors in vivo, suggesting that the presence of TMPRSS2:ETS may limit the growth inhibitory actions of VDR. Our findings suggest that patients with TMPRSS2:ETS negative tumors may be more responsive to VDR-mediated growth inhibition and that TMPRSS2:ETS status should be considered in future clinical trials.

Prostate cancer (PCa) is the most common visceral cancer and the second leading cause of cancer-related death among American men. PCa growth is stimulated by androgens (primarily dihydrotestosterone, DHT), the ligand for the androgen receptor (AR); AR is a primary target for treatment of metastatic PCa (1). In contrast, activation of the vitamin D receptor (VDR) generally results in growth inhibition and/or differentiation (2). Vitamin D is synthesized in the skin or obtained from dietary sources. It is converted to 25-hydroxyvitamin D₃ (25D) in the liver and, finally, to the active VDR ligand, 1α,25-dihydroxyvitamin D₃ (1,25D) in the kidney. 1,25D is inactivated via 24-hydroxylation by cytochrome P450, family 24, subfamily 1, polypeptide1 (CYP24A1), an enzyme which is induced by VDR creating a negative feedback mechanism. Evidence for a role for vitamin D signaling in reducing risk of PCa or as a treatment for PCa is conflicting. Some studies show a correlation between low sunlight exposure and increased risk for PCa (3, 4). Although some studies find a correlation between low levels of circulating 25D and increased PCa risk, others failed to show a correlation or even suggest decreased risk (5, 6). One challenge in correlating 25D levels and VDR activity is the assumption that there are no alterations in vitamin D metabolism in the tumors. There is good evidence in normal prostate and in preclinical models that VDR signaling is growth inhibitory. Low dietary vitamin D levels increase normal mouse prostate cell proliferation in vivo relative to adequate levels of vitamin D (7). Furthermore, the active metabolite of vitamin D (1,25D) and its less calcemic analog (EB1089) inhibit PCa cell growth in vitro and in some prostate xenograft models (8–10). In contrast, DU145 cells, which express high levels of CYP24A1 in response to 1,25D, are refractory to the growth inhibitory effects of 1,25D, unless CYP24A1 expression is reduced (11) or its activity inhibited (12). Despite promising preclinical results, initial studies in humans have shown limited or no efficacy (13, 14).

More than 50% of PCa contain a chromosomal rearrangement that fuses the androgen-regulated promoter of transmembrane protease serine 2 (TMPRSS2) (21q22) to the coding region of erythroblast transformation-specific (ETS) transcription factors, with the TMPRSS2:ETS-related gene (ERG) fusion as the most common form by far (15–17). This typically results in a large increase (>100-fold) in ERG expression. Depletion of ERG in TMPRSS2:ERG positive VCaP cells reduces invasiveness and cell growth in vitro, orthotopic tumor growth in vivo, and decreases ERG target gene expression, including v-myc avian myelocytomatosis viral oncogene homolog (c-Myc) and genes associated with invasiveness (18–20). Multiple subtypes of TMPRSS2:ERG fusions containing various combinations of 5′ TMPRSS2 regions and exons fused with different truncated 3′ ERG transcripts have been identified (18, 21). VCaP cells contain the most common subtype, type III, in which the noncoding TMPRSS2 exon 1 is fused to ERG exon 4.

Most preclinical studies of vitamin D action have been performed in models lacking TMPRSS2:ERG. Surprisingly, activation of VDR induces TMPRSS2 and TMPRSS2:ERG expression in VCaP cells independent of AR signaling (22), but high levels of 1,25D inhibit VCaP cell growth. Nonetheless, this raises the question of whether the response to vitamin D signaling differs in TMPRSS2:ERG-containing tumors. Some ETS factors cooperate with VDR to induce rat CYP24A1 (23); the human promoter also contains a predicted ETS binding site (24). Thus, the high levels of ETS factors might cause aberrant induction of CYP24A1 causing local inactivation of 1,25D and reducing the activity of VDR. We have tested this hypothesis and have found that ERG synergizes with 1,25D to induce high levels of CYP24A1 causing metabolism of 1,25D. Second, we asked whether administration of EB1089, a CYP24A1-resistant 1,25D analog, would overcome the actions of TMPRSS2:ERG to inhibit xenograft growth as previously observed in the TMPRSS2:ERG negative LNCaP xenograft model (9).

Materials and Methods

Cell culture and reagents

VCaP PCa cells, a gift from Dr Kenneth Pienta (University of Michigan) and HEK293 cells (American Type Culture Collection [ATCC]) were maintained in DMEM (Invitrogen) with 10% fetal bovine serum (FBS) (Sigma). LNCaP PCa cells were purchased from ATCC and cultured in RPMI 1640 medium (Invitrogen) with 10% FBS. VCaP cells were derived from a vertebral metastasis, express wild-type AR, the TMPRSS2:ERG translocation, and are androgen dependent in vivo and in vitro (15, 25, 26). LNCaP cells were derived from a lymph node metastasis, express a mutant AR (T877A) with broadened ligand specificity, contain a genomic translocation of the entirue ETV1 region, and are androgen dependent in vitro and in vivo (26–28). DU145 PCa cells were purchased from ATCC and cultured in MEM (Invitrogen) with 10% FBS. Inducible LNCaP TMPRSS2:ERG type VI+72 stable cell lines were generated using the Gateway ViraPower T-REx Lentiviral Expression System (Invitrogen). The TMPRSS2:ERG type VI+72 coding sequence from a previously described plasmid (18) was cloned into a Gateway destination vector (catalog A11141; Invitrogen), lentiviral particles were generated according to the manufacturer's protocol, transduced into LNCaP cells stably expressing Tet repressor protein, and cultured in RPMI 1640 medium with 10% FBS, 300-μg/mL Geneticin (G418) (Invitrogen), and 3-μg/mL blasticidin (Invitrogen) for the double selection of both Tet repressor protein and TMPRSS2:ERG type VI+72 positive cells. EB1089 was initially provided by LEO Pharmaceuticals and more recently by Cougar Biotechnology, 1,25D was obtained from Selleck, and DHT was purchased from PerkinElmer. Stock solutions were prepared in ethanol and stored in the dark at −80°C. All other reagents were molecular biology reagent grade unless otherwise indicated.

ERG activity assays

VI+72 TMPRSS2:ERG LNCaP cells were plated at 250 000 cells/well in 6-well plates, and 1 μg of 8xPal luciferase plasmid reporter (TK luciferase plasmid containing eight copies of the DNA-binding site GCAGGAAGCA from the rat stromelysin promoter) (41), a gift of Dr Arthur Gutierrez-Hartmann (University of Colorado), was transfected using Lipofectin (Invitrogen) according to the manufacturer's protocol. Cells were incubated with vehicle or various doses of doxycycline for 24 hours, rinsed with PBS (Invitrogen), pelleted, and lysed with 1× Reporter Lysis buffer (Promega) for 30 minutes. Luciferase activity was measured using luciferase assay reagent (Promega); protein concentrations were determined using the Bradford assay dye (Bio-Rad).

Small interfering RNA (siRNA) transfection

VCaP cells were plated at 150 000 cells/well in 6-well plates and allowed to adhere for 48 hours. Sense strand sequences for the 2 independent siRNAs targeting ERG were as follows: siRNA number 1, 5′-CCAUCUCCUUCCACAGUGCtt-3′ and siRNA number 2, 5′-AGCCUUACAAAACUCUCCAtt-3′. ERG siRNAs were custom synthesized as Silencer Select Custom siRNA from Ambion, and 10 pmol siRNA/well of ERG siRNA and/or Ambion Silencer Select Negative Control number 1 siRNA (catalog 4390843) was transfected for 6 hours using Lipofectamine 2000 reagent (Invitrogen) according to the manufacturer's protocol.

Quantitative RT-PCR (qRT-PCR)

RNA was isolated using TRIzol reagent (Invitrogen) according to the manufacturer's protocol. VCaP cells were plated at 400 000 cells/well in 6-well plates, and LNCaP cells were plated at 100 000 cells/well in 6-well plates for RNA expression studies. Tumor samples were immersed in TRIzol, homogenized, and total RNA was isolated using the RNeasy Mini kit (QIAGEN) according to the manufacturer's protocol. Isolated RNAs were converted into cDNA with amfiRivert Platinum cDNA synthesis Master Mix according to the manufacturer's protocol (GenDEPOT). qRT-PCR was performed using SYBR green PCR Master Mix (Applied Biosystems) on a 7900 Fast Real-Time PCR system (Applied Biosystems). Primer sequences used for calcium channel, voltage-dependent, L type, alpha 1D subunit (CACNA1D), c-Myc, CYP24A1, ERG, TMPRSS2, and TMPRSS2:ERG were described previously (22, 29). Primers for plasminogen activator tissue (PLAT) were described by others (19). Primers for transient receptor potential cation channel, subfamily V, member 6 (TRPV6) were 5′-AGAGCCGAGATGAGCAGAAC-3′ and 5′-CAGGGCCTGGACATCATTA-3′.

Mammalian 2-hybrid assay for measurement of 1,25D

HEK293 cells plated at 200 000 cells/well in 12-well plates were transfected with 10-ng pCMV-AD-hVDR (prey), 30-ng pCMV-BD-hRXRα (bait), 500-ng pFR-Luc reporter gene (gifts of Dr Peter Jurutka) (30) and 100-ng pCR 3.1 β-galactosidase expression plasmid using Lipofectamine 2000. Three hours after transfection, cells were treated directly with 1,25D, conditioned medium, or various control media. After 24 hours, the cells were lysed, and luciferase and β-galactosidase activity were measured. The data are reported as luciferase/β-galactosidase activity expressed as relative luciferase activity.

Western blot analysis

Cells were washed in PBS, pelleted, and protein extracted by 3 freeze/thaw cycles in 1× Reporter Lysis buffer (Promega) containing 0.4M NaCl and protease and phosphatase inhibitors (GenDEPOT). The lysates were centrifuged, and supernatants were used for the SDS-PAGE Western blotting. Western blot analyses for ERG (catalog sc-354; Santa Cruz Biotechnology, Inc), c-Myc (catalog Ab32072; Abcam), VDR (GTX72715 from Genetex), AR (AR441 described previously, see Ref. 31), tubulin (catalog 05–661; Millipore), and actin (catalog MA5–11869; Pierce Thermo Fisher Scientific) were performed as previously described (22, 29).

In vitro growth assay

VCaP cells were plated at 100 000 cells/well in 6- or 12-well plates, and LNCaP cells were plated at 25 000 cells/well in 6-well plates. Cells were treated with ethanol (EtOH) or the indicated amounts of 1,25D or EB1089. The medium and hormones were replenished every 3 days. Cells were washed with Hanks' buffered salt solution (Invitrogen), lifted from plates with 0.25% EDTA Trypsin (Invitrogen), placed in Isoton II Diluent (Beckman) with 3 drops of Zapoglobin II Lytic reagent (Beckman), and counted using a Coulter particle counter Z1 (Beckman).

VCaP xenograft study

VCaP cells stably expressing luciferase (VCaP-luc) were generated by transducing VCaP cells with a previously described pCDH-CMV-MCS-EF1-Puro lentiviral construct (catalog CD510B-1; System Biosciences) encoding luciferase (18). VCaP-luc cells were maintained in DMEM, 10% FBS, and 0.4-μg/mL puromycin (Sigma). One million VCaP-luc cells were orthotopically injected into 6-week-old severe combined immunodeficient (SCID) (NCI-Charles Rivers Laboratories) male mouse prostate. One week later, mice were weighed, and tumor take was measured by anesthetizing mice with isoflurane (Butler Animal Health Supply) and imaging 5 minutes after the ip injection of firefly D-luciferin dissolved in deionized water (catalog XR-1001; Caliper LifeSciences) (25 mg/kg) using the IVIS imaging system (Xenogen). Mice were divided into 2 groups, and 50 μL of sesame oil as vehicle (Sigma) or 0.5-μg EB1089/kg mouse dissolved in 50 μL of sesame oil were given every other day via oral gavage. Tumor growth was measured weekly using the IVIS imaging system. Mice were killed 5 weeks after initiation of treatment. Blood was collected via cardiac puncture. Body mass and tumor mass were recorded, tumors and kidneys were snap frozen in liquid nitrogen and stored at −80°C. All animal experiments were conducted in accord with accepted standards of humane animal care, as outlined in the Ethical Guidelines and approved by the Baylor College of Medicine (BCM) Institutional Animal Care and Use Committee.

Serum isolation and calcium analysis

Blood was placed in Eppendorf tubes, incubated on ice for 20 minutes, and spun in a microcentrifuge at 5000 rpm for 3 minutes. The serum was stored at −20°C. Serum calcium was measured by the BCM Comparative Pathology Laboratory in the Center for Comparative Medicine.

Statistics

GraphPad Prism 5 software was used to analyze statistical significance. The normal distribution within groups was tested using the Kolmogorov-Smirnov normality test. Box plots were used to show the distribution of the data if the dataset did not pass the normality test. To determine the statistical differences, nonparametric Mann-Whitney test was used if the dataset did not pass the normality test and an unpaired t test was used if the dataset passed the normality test, and 2 groups were compared. Statistical differences between means of multiple treatments within 1 group were analyzed by one-way ANOVA with Bonferroni's post hoc test. Statistical differences for experiments with 2 independent variables were analyzed by two-way ANOVA with Bonferroni's post hoc test. Unless noted otherwise, all in vitro experiments were performed a minimum of 3 independent times, with 1 representative experiment shown. Bars represent the average ± SEM of 3 independent samples.

Results

ERG depletion in VCaP cells reduces 1,25D-mediated CYP24A1 induction but does not reduce other VDR target genes

To determine whether the TMPRSS2:ERG fusion regulates CYP24A1 expression, 2 independent siRNAs targeting ERG were used to reduce ERG expression in VCaP cells, and resulting changes in gene expression were measured. The ERG siRNAs reduced ERG protein and RNA expression compared with the control siRNA-transfected cells (Figure 1, A and B). Furthermore, ERG depletion reduced CACNA1D RNA expression, a known ERG target gene (Figure 1C) (19). ERG depletion also greatly reduced 1,25D-mediated induction of CYP24A1 (partial depletion of ERG resulted in >80% reduction of CYP24A1) without significantly altering basal levels; this suggests that ERG cooperates with VDR to hyperinduce CYP24A1 expression (Figure 1D). The effect of ERG on 1,25D-mediated gene induction appears to be CYP24A1 specific, because we did not observe the same effect on 1,25D-mediated induction of TMPRSS2 or TRPV6 (Figures 1, E and F).

Figure 1.

ERG depletion in VCaP cells preferentially reduces 1,25D-mediated CYP24A1 induction. ERG expression was reduced using 2 independent siRNAs targeting ERG. A, VCaP cells were transfected with control or ERG-specific siRNAs; after 24 hours, cells were treated with EtOH (E) or 100nM 1,25D (D) for an additional 24 hours, and ERG and actin protein levels were detected by Western blotting. B, VCaP cells were transfected and treated as in A, RNA was extracted, and ERG RNA levels were measured by qRT-PCR and normalized to 18S RNA. C, CACNA1D (ERG target gene) expression. D–F, VDR target gene expression (**, P < .01; ***, P < .001).

1,25D-dependent levels of CYP24A1 in VCaP cells are similar to those in DU145 cells and sufficient to inactivate much of the available 1,25D

The level of CYP24A1 in PCa cell lines is very variable. LNCaP cells have extremely low basal levels, and induced levels are low despite a large fold induction by 1,25D (32). In contrast, DU145 cells express unusually high basal levels of CYP24A1, which are increased by 1,25D (8, 32, 33). This CYP24A1 expression in DU145 cells is sufficient to metabolize 1,25D and reduce responsiveness of the cells to 1,25D (11, 12). We found that VCaP cells have induced levels of CYP24A1 RNA comparable with those of 1,25D-treated DU145 cells (Figure 2A). To determine whether 1,25D metabolism is sufficient to limit 1,25D-mediated gene expression, we treated VCaP cells with suboptimal levels of 1,25D (30nM) or the CYP24A1-resistant analog, EB1089, for 96 hours or for the final 24 hours and measured CYP24A1 RNA expression as a measure of VDR transcriptional activity (Figure 2B). Treatment for 24 hours yielded higher levels of CYP24A1 than 96 hours of treatment. In contrast, gene induction by EB1089 was somewhat higher at 96 hours compared with 24 hours, suggesting that CYP24A1 metabolizes a significant amount of the 1,25D (Figure 2B). To more directly test for metabolism of 1,25D, we used a previously described mammalian 2-hybrid assay, in which hormone-dependent dimerization of a Gal4-DNA binding domain VDR fusion protein with retinoid x receptor α (RXRα) linked to an activation domain yields activation of a luciferase reporter that is proportional to the concentration of 1,25D in the medium (30). As shown in Figure 2C, activity in response to 1,25D is dose dependent. To assess residual 1,25D in the medium of the cells from Figure 2B, we added 100 μL of conditioned medium from each of the 24- and 96-hour treated samples to transfected HEK293 cells in 1.5 mL of medium. This dilution would result in a final concentration of approximately 2nM if none of the 1,25D were metabolized. As shown in Figure 2C, the residual 1,25D in the 96-hour-treated samples yielded about 20% of the activity relative to the 24-hour incubation time point (conditioned medium). As controls for loss of 1,25D due simply to incubation at 37°C, a final concentration of 30nM 1,25D was placed in 6-well plates containing DMEM and 10% FBS without cells (incubated control) and incubated for 96 hours in parallel with the samples in Figure 2B. The activity of the incubated control medium was compared with freshly prepared 30nM 1,25D in DMEM and 10% FBS (control). There was no loss of activity after 96 hours in incubated control medium relative to the freshly diluted control; the levels of activity in the 24-hour sample suggested minimal degradation at 24 hours but a very substantial reduction in hormone at 96 hours.

Figure 2.

VCaP cells express high levels of CYP24A1-inducing metabolism of 1,25D. A, LNCaP cells, VCaP cells, and DU145 cells were cultured for 72 hours, treated with vehicle (EtOH) or 100nM 1,25D (1,25D) for 24 hours, RNA was extracted, and CYP24A1 RNA levels were measured using qRT-PCR and normalized to 18S RNA. B, VCaP cells were allowed to adhere for 48 hours, and medium was replenished. The 96-hour group was then treated with vehicle (EtOH), 30nM 1,25D, or 1nM EB1089. After 72 hours, the 24-hour group was treated with vehicle (EtOH), 30nM 1,25D, or 1nM EB1089. After another 24 hours, both 96- and 24-hour treatment groups were harvested, RNA was extracted, and CYP24A1 RNA levels were measured using qRT-PCR and normalized to 18S RNA (*, P < .05; **, P < .01; ***, P < .001). C, To determine whether VCaP cells reduce the level of 1,25D in their medium, HEK293 cells plated at 200 000 cells/well were transfected with 10-ng pCMV-AD-hVDR (prey), 30-ng pCMV-BD-hRXRα (bait), 500-ng pFR-Luc reporter gene, and 100-ng β-galactosidase using Lipofectamine 2000. Three hours after transfection, cells were treated with 1,25D at the indicated final concentrations, 100 μL of conditioned medium from 2B (left) in a total of 1.5-mL medium (conditioned medium), 100 μL of medium from 6-well plates incubated in parallel with the samples in 2B in the absence of cells (incubated control), or with 100 μL of medium to which ethanol or 30nM 1,25D was added immediately before use (control). Standards were done in triplicate, and all other samples were from 3 independent wells. After 24 hours, the HEK293 cells were harvested and luciferase and β-galactosidase activity were measured. The data are reported as luciferase/β-galactosidase activity expressed as relative luciferase activity.

VCaP cells require higher levels of 1,25D to induce optimal growth inhibition than do LNCaP cells

The finding that 1,25D is metabolized by VCaP cells suggests that the activity of 1,25D might be reduced by its metabolism. Thus, we compared the effect of various doses of 1,25D on growth of VCaP cells with that of EB1089, a synthetic VDR agonist reported to be resistant to CYP24A1 degradation. As shown in Figure 3A, much lower doses of EB1089 are required to induce growth inhibition. However, this analog also has a higher affinity for VDR. Therefore, we compared responses of LNCaP cells (Figure 3B), which have extremely low levels of CYP24A1 (32) to those of VCaP cells. The direct comparisons of the responses of the 2 lines are shown as % maximal growth inhibition in Figure 3, C and D. The LNCaP cells are more sensitive to 1,25D. Although LNCaP cells are maximally growth inhibited at 3nM–10nM 1,25D, VCaP cells require higher levels to reach maximal growth inhibition (Figure 3C). LNCaP cells are somewhat more responsive to EB1089, but both achieve more than 90% growth inhibition by 1nM EB1089. This suggests that the metabolism of 1,25D in the VCaP cells is sufficient to reduce the response to added 1,25D.

Figure 3.

VCaP cells are less sensitive to 1,25D than are LNCaP cells. A, VCaP cells plated at 100 000 cells/well in 6-well plates were incubated with the indicated concentrations of 1,25D and EB1089, medium and hormone were replenished on day 3 and 6 and cells counted on day 9. B, LNCaP cells plated at 25 000 cells/well were treated with the indicated concentrations of 1,25D and EB1089, medium and hormone replenished on day 3, and the cells counted on day 7. C, Comparison of the % maximal growth inhibition in response with the indicated doses of 1,25D. In each line, the difference in cell number between ethanol-treated cells, and the highest dose was set at 100% maximal growth inhibition. D, Identical to C except that the EB1089 data were plotted.

Expression of type VI+72 TMPRSS2:ERG in LNCaP cells synergizes with 1,25D to hyperinduce CYP24A1 but does not increase other VDR target genes

One limitation in studying the role of TMPRSS2:ERG is the lack of endogenous TMPRSS2:ERG fusion containing PCa cell lines other than VCaPss. Therefore, we generated doxycycline-inducible type VI+72 TMPRSS2:ERG-overexpressing LNCaP cells. The type VI+72 TMPRSS2:ERG fusion protein is an alternatively spliced form that is slightly larger than the type III found in VCaP cells. Doxycycline treatment increased ERG RNA expression, but as expected, 1,25D had no effect on expression (Figure 4A). Dose-dependent ERG protein expression was obtained in response to doxycycline treatment with no detectable ERG in parental LNCaP cells (Figure 4B). The levels achieved with 10-ng/mL doxycycline are substantially lower than the levels of the endogenous type III isoform in VCaP cells (Figure 4B). Thus, the response in this line may be an underestimate of the effect in PCa patients with tumors expressing the TMPRSS2:ERG fusion gene. To confirm that the overexpressed ERG protein arising from the type VI+72 TMPRSS2:ERG fusion is transcriptionally active, we transfected cells with an 8XPal-luciferase plasmid, which contains 8 repeating ETS transcription binding sites linked to the luciferase-coding region and confirmed that type VI+72 ERG increased 8XPal-luciferase expression substantially (Figure 4C).

Figure 4.

Overexpression of type VI+72 TMPRSS2:ERG in LNCaP cells synergizes with 1,25D to induce CYP24A1. A, Inducible type VI+72 TMPRSS2:ERG LNCaP cells were pretreated with 10-ng/mL doxycycline (Dox) for 24 hours and treated with vehicle (EtOH) or 100nM 1,25D for another 48 hours. RNA was extracted, ERG RNA levels were measured using qRT-PCR and normalized to 18S RNA. B, Inducible type VI+72 TMPRSS2:ERG LNCaP cells were treated with the indicated amounts of Dox, VCaP cells and parental LNCaP cells were treated with EtOH (E) or 100nM 1,25D (D) for 24 hours, and ERG and actin protein levels were detected by Western blotting. C, Inducible type VI+72 TMPRSS2:ERG LNCaP cells were transiently transfected with the ETS responsive reporter, 8xPal luciferase plasmid, treated with the indicated amount of Dox for 24 hours, and luciferase activity was measured and normalized to total protein concentration (**, P < .01, significant effect of 0-ng/mL vs 10-ng/mL Dox). D–F, RNA isolated as described in A was used to measure CYP24A1, TMPRSS2, and TRPV6. (*, P < .05; **, P < .01; ***, P < .001). G, Twenty-five micrograms of lysates from control LNCaP VI+72 TMPRSS2:ERG cells or 10-ng/mL Dox-treated cells treated with vehicle or 100nM 1,25D for 24 hours, and 25 μg of lysates from VCaP cells were run on an sodium dodecyl sulfate (SDS) gel and immunoblotted to detect VDR.

ERG did not alter basal levels of CYP24A1 but synergized with 1,25D to hyperinduce CYP24A1 RNA (Figure 4D). In contrast, it did not enhance 1,25D-mediated induction of TMPRSS2 or TRPV6 (Figure 4, E and F), nor did it increase levels of VDR protein (Figure 4G).

Expression of type VI+72 TMPRSS2:ERG synergizes with 1,25D to induce CYP24A1 at both low and high concentrations of 1,25D

Induction of CYP24A1 is a feedback mechanism to limit the levels of 1,25D. Higher levels of 1,25D are required to induce CYP24A1 than many other target genes. To examine ERG-dependent changes as a function of hormone concentration, type VI+72 TMPRSS2:ERG LNCaP cells were treated with vehicle or 10-ng/mL doxycycline and various doses of 1,25D. ERG overexpression (Figure 5A) synergized with activated VDR to induce CYP24A1 expression even at doses as low as 1nM; 10nM 1,25D yields more induction than is achieved with 1μM 1,25D in the absence of ERG (note the differences in the Y axes in the 2 panels in Figure 5B). Higher levels of hormone yielded higher levels of CYP24A1; thus, there was little change in the dose response to 1,25D, although induction at each dose of 1,25D was greatly increased. Consistent with the depletion studies in VCaP cells, overexpression of ERG had little effect on the induction of the other VDR target genes, TMPRSS2 and TRPV6 (Figure 5, C and D).

Figure 5.

ERG preferentially increases efficacy of CYP24A1 induction. Inducible type VI+72 TMPRSS2:ERG LNCaP cells were pretreated with 10 ng/mL of doxycycline (Dox) for 24 hours and treated with vehicle (EtOH) or the indicated amounts of 1,25D for another 48 hours. RNA was extracted, RNA levels were measured using qRT-PCR and normalized to 18S RNA. A, Expression of ERG RNA. B, Expression of CYP24A1 RNA. C and D, TMPRSS2 and TRPV6 RNA expression (**, P < .01; ***, P < .001).

EB1089 inhibits growth of VCaP cells in vitro but does not inhibit growth of orthotopic VCaP-luc tumors

The studies above suggest that VDR signaling in TMPRSS2:ETS fusion gene positive tumors may be limited by induction of CYP24A1. Thus, a CYP24A1-resistant 1,25D analog may be needed to fully activate VDR and to inhibit growth in vivo. We found that EB1089 inhibits LNCaP xenograft growth without inducing hypercalcemia (9). We previously reported (22) that EB1089 and 1,25D inhibit the growth of VCaP cells in vitro, and Figure 3 shows the dose dependence of this response in comparison with LNCaP cells. To monitor tumor growth, we transduced VCaP cells with a lentivirus encoding luciferase (VCaP-luc cells) and confirmed that these cells remain responsive to EB1089 in vitro (Figure 6A). Male SCID mice were orthotopically injected with VCaP-luc cells, separated into equivalent groups based on initial luciferase activity 1 week after injection, given vehicle (sesame oil) or EB1089 (0.5 μg/kg) by oral gavage on alternate days, and tumor growth was monitored via measuring luciferase activity. There was no loss of body weight as a result of treatment. Initial average body weights were 20.5 ± 0.5 g for the vehicle group and 22.2 ± 0.3 g for the EB1089 group, and average final body weights were 22.2 ± 0.6 g for the vehicle group and 23.5 ± 0.4 g for the EB1089 group. EB1089 did not inhibit tumor growth in vivo measured by monitoring the luciferase activity (Figure 6B) or by tumor mass measurements at the time of killing (Figure 6C). To assess whether sufficient EB1089 was given to activate VDR signaling in the mice as expected, mouse kidney CYP24A1 RNA expression was measured. EB1089 induced mouse CYP24A1 RNA expression (24-fold and P < .001) (Figure 6D). The serum calcium level was elevated (P < .001), although it remained within the normal range for SCID mice (Figure 6E), suggesting that EB1089 activated VDR signaling systemically without causing hypercalcemia. To assess the response of the tumors to EB1089, we examined changes in gene expression. EB1089 induced TMPRSS2:ERG RNA expression as well as the RNA of its reported target genes, PLAT and c-Myc (Figure 6, F–H), and the c-Myc target gene, E2F transcription factor 1 (E2F1) (Figure 6I) (19, 20). EB1089 did not significantly induce tumor levels of CYP24A1 (Figure 6J). However, EB1089 significantly induced a more sensitive VDR target gene, TRPV6, even though gene expression was extremely variable in the EB1089-treated animals (Figure 6K). That higher levels of VDR activation are required to induce CYP24A1 than TRPV6 and that TMPRSS2 is most sensitive to 1,25D in LNCaP cells is shown in Figure 5. To test whether there is a similar difference in sensitivity to EB1089 in VCaP cells, dose-dependent induction of TMPRSS2 and CYP24A1 mRNA in VCaP cells was measured (Figures 6, L and M). Figure 6N depicts % maximal induction of gene expression as a function of EB1089 treatment and shows that TMPRSS2 is induced at lower level of EB1089 than is CYP24A1. This suggests that the tumors, which show modest induction of TMPRSS2:ERG but not of CYP24A1, may have taken up relatively low levels of EB1089 despite the apparently higher systemic responses.

Figure 6.

EB1089 inhibits VCaP cell growth in vitro but not in vivo. A, VCaP-luc cells were treated with vehicle control (EtOH) or 10nM EB1089 for the indicated time periods, and cells were counted with a Coulter counter (***, P < .001, significant effect of EtOH vs EB1089). B, Average luciferase measurement of all mice (n = 9 control and n = 12 EB1089) from the same time point as C. C, The tumor mass at the time of killing. D, RNA was extracted from kidneys, and CYP24A1 was measured using qRT-PCR and normalized to 18S RNA (***, P < .001, significant effect of vehicle vs EB1089). E, Serum was isolated from blood collected at the time of killing, and serum calcium levels were measured (***, P < .001, significant effect of vehicle vs EB1089), F–K, RNA was extracted from VCaP-luc tumors, and levels were measured by qRT-PCR and normalized to human β-actin RNA. F, TMPRSS2:ERG (VDR target gene) (22) RNA expression (*, P < .05, significant effect of vehicle vs EB1089). G and H, PLAT and c-Myc (ERG target genes) (19, 20) RNA expression (*, P < .05; **, P < .01, significant effect of vehicle vs EB1089). I, E2F1 (c-Myc target gene) RNA expression (*, P < .05, significant effect of vehicle vs EB1089). J, CYP24A1 (VDR target gene) RNA expression. K, TRPV6 (VDR target gene) RNA expression where individual values for tumors were plotted (*, P < .05, significant effect of vehicle vs EB1089). L–M, VCaP cells plated at 400 000 cells/well were treated with the indicated amounts of EB1089 for 72 hours, RNA was isolated and purified to measure gene expression. L, TMPRSS2; M, CYP24A1; N, % maximum induction. For each gene, the basal (ethanol) levels of gene expression was subtracted from the level in each treated sample, and the level at the highest dose of EB1089 was set as 100%, with the remaining samples compared with this level (**, P < .01; ***, P < .001).

Effects of DHT on VDR-mediated gene expression

There are reports of effects of AR activation and antagonist bound AR on VDR function in LNCaP cells, although some of these responses are cell line specific (22, 34, 35). DHT reduces VDR-mediated induction of CYP24A1 RNA expression in AR positive LNCaP cells but not in AR negative PC3 and DU145 cells (36, 37). The mice in this study were intact and thus had normal circulating levels of androgens. Therefore, we determined the effect of DHT or EB1089 singly and in combination on VDR-mediated gene expression in VCaP cells. EB1089 induced TMPRSS2:ERG RNA expression to a similar extent as DHT alone, and the combination of EB1089 and DHT minimally increased levels beyond that observed with DHT or EB1089 (Figure 7A). DHT alone had no effect on CYP24A1 expression, but it reduced EB1089-mediated induction of CYP24A1 to about 25% of EB1089 alone (Figure 7B). However, fold induction in the presence of both hormones was still substantial (20- to 30-fold). In contrast, DHT had much less effect on EB1089-mediated induction of another VDR target gene, TRPV6 (Figure 7C). To determine whether either receptor altered expression of the other receptor, we treated VCaP cells with DHT, EB1089, or the combination and assessed receptor expression by Western blotting (Figure 7D). There was a modest reduction in VDR when cells were treated with both hormones, suggesting that this may contribute to the modest reductions in TRPV6 expression, but this is unlikely to be the sole reason for the large reduction in CYP24A1 activity or the failure to detect additive or synergistic induction of TMPRSS2:ERG. VDR and AR are in the same family of transcription factors and likely play similar roles in inducing this gene. Thus, if either receptor is optimally activated, the other may provide little additional transcriptional activation.

Figure 7.

Differential effects of DHT on VDR-mediated induction of target genes in VCaP cells. VCaP cells were treated with the indicated amounts of DHT and/or EB1089 (EB) for 24 hours, RNA was extracted, and target gene expression was measured by qRT-PCR and normalized to 18S RNA. A, TMPRSS2:ERG RNA expression. B, CYP24A1 RNA expression. C, TRPV6 RNA expression (*, P < .05; **, P < .01; ***, P < .001). D, VCaP cells were treated with vehicle (EtOH), 5nM EB, 10nM DHT, or 10nM EHT + 5nM EB (DHT + EB) for 24 hours, lysates prepared, 25-μg protein separated on an sodium dodecyl sulfate (SDS) gel, and AR, VDR, and tubulin protein levels were detected by Western blotting.

Discussion

Whether VDR signaling reduces risk for PCa or can be used to aid in treatment is unresolved. Our finding that 1,25D induces expression of TMPRSS2 and of TMPRSS2:ERG fusions (22) suggests that responses to VDR activation may differ in fusion positive tumors compared with fusion negative tumors. An earlier study showing that some, but not all, ETS transcription factors can cooperate with VDR to induce rat CYP24A1 raised the question of whether the remarkable overexpression of ETS transcription factors caused by the genomic translocations that produce the fusions (>1000-fold in some cases) (15) cooperates with VDR to increase CYP24A1 expression. Our ERG depletion and overexpression studies show that ERG synergizes with VDR to induce CYP24A1 but not other VDR target genes. Remarkably, 10nM 1,25D treatment in ERG-expressing cells induced CYP24A1 expression to a level higher than 1μM 1,25D treatment in non-ERG-expressing cells (Figure 5B). This is clinically significant in TMPRSS2:ERG positive PCa, because it suggests that the normal feedback mechanism is altered in tumor cells; in the presence of ERG overexpression arising from TMPRSS2:ERG fusion, 1,25D-mediated hyperinduction of CYP24A1 could limit the local amount of 1,25D as suggested by our studies of residual 1,25D in VCaP-conditioned medium (Figure 2C). This would result in induction of the highly sensitive TMPRSS2:ERG but not other growth inhibitory actions of 1,25D that may require higher levels of 1,25D. In our studies, we achieved only a partial depletion of ERG in the VCaP cells. Moreover, the expression of the type VI+72 ERG isoforms in LNCaP cells was much less than that of endogenous ERG in VCaP PCa cells, suggesting that the alterations in CYP24A1 expression may be an underestimate of what occurs in TMPRSS2:ERG fusion-containing tumors. A recent publication showed that CYP24A1 expression typically is higher in malignant human prostate tissues compared with benign tissues (11). Although the authors did not measure TMPRSS2:ERG, the high frequency of this fusion (>50%) suggests that it may be a major factor in the reported elevation of CYP24A1. Interestingly, a recent study showed that PCa patients receiving 40 000-IU vitamin D/d for 3–8 weeks before prostatectomy had variable levels of 1,25D in prostate tissue (<20–100 pmol/kg) (38). Those in the highest quartile had reduced Ki67 (human nuclear antigen defined by the monoclonal antibody Ki-67) staining, suggesting that the local levels of 1,25D regulate prostate proliferation (38). This supports the concept that differential metabolism of vitamin D can play a role in the level of prostate cell/tumor proliferation.

Although high levels of 1,25D or EB1089 inhibit VCaP growth in vitro, the environment in vivo and requirements for tumor growth are likely to be quite different. Moreover, there is a limit to the amount of VDR agonist that can be given without inducing hypercalcemia. In our experiment, there was no difference in size of tumors in response to EB1089 treatment compared with control. That near maximal EB1089 was delivered systemically is shown both by the elevation of renal CYP24A1 RNA expression and of serum calcium. These mice were fed a regular chow diet containing 1500-IU vitamin D/kg chow, which theoretically could have reduced the difference in VDR activity in the 2 groups and thus any difference in growth. However, a subsequent sc xenograft study using a defined chow with reduced levels of vitamin D also showed no difference in tumor size (data not shown). That the tumors responded to some extent to EB1089 is shown by the significant increase in TMPRSS2:ERG fusion RNA. In vitro, VCaP PCa cells express high levels of the TMPRSS2:ERG fusion and induction of the TMPRSS2:ERG fusion by 1,25D resulted in modest induction of CACNA1D RNA expression (22), an ERG target gene. In vivo, we observed induction not only of the TMPRSS2:ERG fusion but of its targets PLAT and c-Myc as well as of the c-Myc target gene, E2F1 (Figure 6). Dose responses for 1,25D-mediated induction of target genes can differ by an order of magnitude. Our in vitro studies (Figure 5) show that TMPRSS2 reaches nearly 50% of maximum by 1nM 1,25D, whereas TRPV6 requires 10nM and CYP24A1 requires more than 10nM 1,25D to reach 50% of maximum. Similar differences in responses were observed in VCaP cells treated with EB1089 (Figure 6, L–N). Gene expression in the tumor (Figure 6) suggests that adequate EB1089 was delivered to induce the TMPRSS2:ERG fusion, that the amount was borderline for TRPV6 induction, with some mice showing enhanced expression and others no response, and that EB1089 levels were insufficient to raise tumor CYP24A1 at least in the presence of circulating androgens. In contrast, growth of LNCaP xenografts is inhibited by EB1089 (9). Serum calcium levels in our previous experiment compared with this experiment suggest similar systemic delivery. To our knowledge, in TMPRSS2:ERG negative tumors, most, if not all, actions of VDR are growth inhibitory. In contrast, in TMPRSS2:ERG positive tumors, VDR may have the added action of stimulating growth through induction of TMPRSS2:ERG. Thus, the growth inhibitory of actions of VDR must more than compensate for this stimulation to inhibit growth. The gene expression data in the VCaP tumors suggest that the level of EB1089 in the tumors was sufficient to induce TMPRSS2:ERG but not to achieve levels of VDR activation sufficient to inhibit the tumor growth (Figure 6). Thus, the VCaP tumors may have taken up less EB1089 and/or it was metabolized or pumped out by a transporter. EB1089 is reported to be CYP24A1 resistant (39, 40), but little is known about its metabolism.

In androgen replete animals and tissues, androgen will blunt the induction of CYP24A1. Under conditions of androgen deprivation, VDR likely plays a greater role in inducing expression of the TMPRSS2:ERG fusion. Moreover, induction of CYP24A1 should be greater, which will potentially limit actions of VDR that require higher levels of hormone. It remains to be determined whether other CYP24A1-resistant VDR agonists with a wider separation between the amount required to activate VDR and the concentration that induces hypercalcemia would be effective in inhibiting VCaP or other TMPRSS2:ERG-expressing tumors. We cannot, however, exclude the possibility that the requirements for growth in vivo differ such that no amount of VDR activation would inhibit TMPRSS2:ERG-expressing tumor growth. This study suggests that clinical trials of 1,25D or its metabolites may show more benefit in patients if trials are limited to patients without fusions.

Funding Statement

This work was supported by Department of Defense Idea Awards W81XWH-09–1-416, W81XWH-13–1-330, and T32 HD07165 and by the National Cancer Institute Cancer Center Support Grant P30CA125123.