NIH Public Access
`Author Manuscript
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`Published in final edited form as:
`Cell. 2013 December 5; 155(6): 1309–1322. doi:10.1016/j.cell.2013.11.012.
`
`Glucocorticoid Receptor Confers Resistance to Anti-Androgens
`
`by Bypassing Androgen Receptor Blockade
`
`Vivek K. Arora1,2, Emily Schenkein1, Rajmohan Murali1,3, Sumit K. Subudhi2, John
`Wongvipat1, Minna D. Balbas1,4, Neel Shah1,4, Ling Cai1, Eleni Efstathiou5, Chris
`Logothetis5, Deyou Zheng6, and Charles L. Sawyers1,7
`1Human Oncology and Pathogenesis Program, Memorial Sloan-Kettering Cancer Center, New
`York, NY 10065.
`
`2Department of Medicine, Memorial Sloan-Kettering Cancer Center, New York, NY 10065
`
`3Department of Pathology Memorial Sloan-Kettering Cancer Center, New York, NY 10065
`
`4Louis V. Gerstner, Jr. Graduate School of Biomedical Sciences, Memorial Sloan-Kettering
`Cancer Center, New York, NY 10065.
`
`5Department of Genitourinary Medical Oncology, The University of Texas MD Anderson Cancer
`Center Houston, TX 77030
`
`6Departments of Neurology, Genetics and Neuroscience, Albert Einstein College of Medicine,
`Bronx, NY 10461
`
`7Howard Hughes Medical Institute, Chevy Chase, MD 20815
`
`Summary
`
`The treatment of advanced prostate cancer has been transformed by novel antiandrogen therapies
`such as enzalutamide. Here we identify induction of glucocorticoid receptor (GR) expression as a
`common feature of drug resistant tumors in a credentialed preclinical model, a finding also
`confirmed in patient samples. GR substituted for the androgen receptor (AR) to activate a similar
`but distinguishable set of target genes and was necessary for maintenance of the resistant
`phenotype. The GR agonist dexamethasone was sufficient to confer enzalutamide resistance
`whereas a GR antagonist restored sensitivity. Acute AR inhibition resulted in GR upregulation in a
`subset of prostate cancer cells due to relief of AR-mediated feedback repression of GR expression.
`These findings establish a novel mechanism of escape from AR blockade through expansion of
`cells primed to drive AR target genes via an alternative nuclear receptor upon drug exposure.
`
`Introduction
`
`Recently approved drugs that target androgen receptor (AR) signaling such as abiraterone
`and enzalutamide have rapidly become standard therapies for advanced stage prostate cancer
`(Scher et al., 2012b) (de Bono et al., 2011). Despite their success, sustained response with
`these agents is limited by acquired resistance which typically develops within ~6-12 months.
`Clinical success of kinase inhibitors in other tumors such as melanoma, lung cancer,
`
`© 2013 Elsevier Inc. All rights reserved.
`
`Correspondence: sawyersc@mskcc.org.
`
`Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
`customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
`the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
`discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`Amerigen Exhibit 1137
`Amerigen v. Janssen IPR2016-00286
`
`

`
`Arora et al.
`
`Page 2
`
`leukemia and sarcoma is similarly transient (Sawyers et al., 2002) (Chapman et al., 2011)
`(Demetri et al., 2002) (Maemondo et al., 2010), resulting in numerous efforts to define
`mechanisms of acquired resistance. One strategy that has proven particularly useful is
`prolonged treatment of drug-sensitive preclinical models to derive drug-resistant sublines,
`followed by genome-wide profiling studies to ascertain differences that may play a causal
`role in conferring drug resistance. A common mechanism that has emerged from these
`kinase inhibitor studies is reactivation of the signaling pathway targeted by the drug, directly
`by mutation of the kinase target or indirectly by bypassing pathway inhibitor blockade
`through amplification of an alternative kinase (Glickman and Sawyers, 2012). Both
`scenarios have been validated in clinical specimens and are guiding efforts to discover next
`generation inhibitors and to develop rational drug combinations.
`
`Clinically relevant mechanisms of resistance to hormone therapy in prostate cancer have
`also been elucidated using preclinical models. Hormone therapy, through the use of drugs
`that lower serum testosterone or competitively block the binding of androgens to AR, has
`been the mainstay of treatment for metastatic prostate cancer for decades but is not curative.
`The late stage of disease, which is refractory to hormone therapy, is termed castration
`resistant prostate cancer (CRPC). We previously examined the molecular basis of
`progression to CRPC in mouse models and discovered that increased AR expression was the
`primary mechanism (Chen et al., 2004). We then used this observation to screen for novel
`anti-androgens that restore AR inhibition in the setting of increased AR levels. These efforts
`yielded three second-generation anti-androgens: enzalutamide, ARN-509, and RD162 (Tran
`et al., 2009) (Clegg et al., 2012). Enzalutamide and ARN-509 were further developed for
`clinical use, culminating in FDA approval of enzalutamide in 2012 based on increased
`survival (Scher et al., 2012b).
`
`Now with widespread use, resistance to enzalutamide is a major clinical problem. We and
`others have recently identified an AR point mutation as one resistance mechanism by
`derivation of drug-resistant sublines following prolonged exposure to enzalutamide or
`ARN-509 (Balbas et al., 2013) (Joseph et al., 2013) (Korpal et al., 2013). This AR mutation
`has also been recovered from patients with resistance to ARN-509 but only in a minority of
`cases (Joseph et al., 2013). Here we define a novel and potentially more prevalent
`mechanism of resistance by which tumors bypass AR blockade through upregulation of the
`glucocorticoid receptor (GR).
`
`Results
`
`GR is expressed in antiandrogen-resistant tumors
`
`We previously showed that LNCaP/AR xenograft tumors regress during the first 28 days of
`treatment with ARN-509 (Clegg et al., 2012), enzalutamide or RD162 (Tran et al., 2009). In
`a pilot study to explore mechanisms of acquired resistance to these drugs, we treated mice
`continually and harvested tumors after progression (mean 163 days, Supplemental Table
`1A). Tissue from fifteen resistant tumors obtained from long term antiandrogen treated mice
`(n=6 ARN-509, n=9 RD162) and from three control tumors from vehicle treated mice were
`analyzed by expression array. Aggregated data from resistant and control tumors in this pilot
`cohort were compared to identify expression changes commonly associated with resistance
`(Figure 1A). Among the most up-regulated genes in the resistant tumors was the
`glucocorticoid receptor (GR, gene symbol NR3C1) which shares overlapping target
`specificity with AR (Mangelsdorf et al., 1995). Of note, several of the most differentially
`expressed genes were known androgen regulated genes (confirmed by transcriptome
`analysis of short term DHT treated LnCaP/AR cells, in vitro (Supplemental Table 1B)), but
`they were altered in directions that did not reflect restored AR signaling. On the one hand,
`SGK1 (Serum Glucocorticoid Induced Kinase 1), a known AR and GR-induced target gene,
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 3
`
`was among the most up-regulated genes, but several other androgen-induced genes
`(PMEPA1, SNAI2, KCNN2, LONRF1, SPOCK1) were among the most repressed.
`Conversely, several androgen-repressed genes (UGT2B15, PMP22, CAMK2N1, UGT2B17)
`were among the most up-regulated (Figure 1A). These findings indicated that resistance in
`this model system is unlikely to be mediated by simple restoration of AR activity and raised
`the possibility that GR may play a role.
`
`To explore this question further, we generated an independent set of drug-resistant tumors
`(the validation cohort), focusing on the two second generation antiandrogens in clinical use,
`enzalutamide and ARN-509 (Figure 1B). GR mRNA levels in 10 control, 8 short term
`treated (4 day) and 16 resistant tumors were substantially higher in resistant tissues
`compared to control (median 26.9-fold increase) or 4 day treated tumors (Figure 1C). Of the
`tissues analyzed by RT-qPCR, most were also analyzed for GR expression by western blot,
`based on availability of protein lysates (control n=6, 4 day n=5, resistant n=13). No GR was
`detected in control samples, minimal expression was noted in 4 day treated samples, and
`substantial expression was found in most resistant tumors in a pattern that tended to
`correlate with GR mRNA levels (Figure 1D). There was no correlation between GR
`expression and the specific antiandrogen treatment used (Supplemental Table 1C). In
`contrast to GR, AR RNA or proteins levels were not consistently different across the
`treatment groups (Figure 1C,1D).
`
`To explore AR and GR signaling in more detail, we established cells lines from control and
`drug-resistant tumors by adaptation to growth in vitro. LREX’ (LnCaP/AR Resistant to
`Enzalutamide Xenograft derived) was derived from an enzalutamide-resistant tumor with
`high GR expression, and CS1 was derived from a vehicle treated tumor. We also developed
`a flow cytometry-based assay to measure GR expression on a cell-by-cell basis. In both
`LNCaP/AR and CS1, most cells showed no evidence of GR expression, with the exception
`of a small subpopulation (black arrow, discussed later) (Figure 1E). In contrast, essentially
`all LREX’ cells expressed GR. Intracellular AR staining confirmed that AR levels in LREX’
`did not notably differ from control cells (Figure S1A).
`
`LREX’ tumors are dependent on GR for enzalutamide-resistant growth
`
`Having established the LREX’ model as representative of high GR expression, we next
`confirmed that these cells maintain a resistant phenotype in vivo. LREX’ or control cells
`were injected into castrated mice that were then immediately initiated on antiandrogen
`treatment. LREX’ showed robust growth whereas LNCaP/AR or CS1 lines were unable to
`establish tumors in the presence of antiandrogen (Figure 2A,2B). Strong expression of GR
`was confirmed in multiple LREX’ xenograft tumors by western blot and by IHC (Figure
`S1B, 2C). As expected, untreated LNCaP/AR tumors were negative for GR expression with
`the exception of rare GR-positive cells (Figure 2C). Although many of these GR-positive
`cells had morphologic features of stromal or endothelial cells (blue arrows), some appeared
`epithelial (black arrow), consistent the with flow cytometry analysis (Figure 1E, black
`arrows).
`
`To determine whether GR expression is required to maintain the drug-resistant phenotype,
`LREX’ cells were infected with a shRNA targeting GR (shGR) and stable knockdown of
`GR protein was confirmed (Figure 2F). Tumor growth of shGR infected LREX’ cells was
`significantly delayed relative to shNT (non targeted)-infected cells in castrated mice treated
`with enzalutamide (Figure 2D). In contrast, shGR had no impact on the growth of GR-
`negative CS1 xenografts, diminishing the possibility of an off-target effect (Figure 2E). Of
`note, shGR LREX’ xenografts harvested on day 49 showed decreased GR protein
`knockdown compared to the pre-implantation levels, indicative of selective pressure against
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 4
`
`GR silencing in the setting of enzalutamide treatment (Figure 2F). These findings provide
`direct evidence that GR drives enzalutamide resistance in vivo.
`
`GR expression is associated with clinical resistance to enzalutamide
`
`To determine whether GR expression is a feature of clinical antiandrogen resistance, we
`evaluated GR expression in bone metastases from patients receiving enzalutamide. Bone
`marrow samples were obtained prior to enzalutamide treatment (baseline) and again after 8
`weeks of treatment, as previously reported in a cohort of abiraterone-treated patients
`(Efstathiou et al., 2012). Using a GR IHC assay optimized for use in bone marrow samples,
`we quantified the percentage of GR-positive tumor cells and dichotomized the data based on
`clinical response. Patients who continued to benefit from therapy for greater than 6 months
`were defined as good responders, while those in whom therapy was discontinued earlier than
`6 months due to a lack of clinical benefit were classified as poor responders (Figure 3A).
`Consistent with the designation of good versus poor clinical response based on treatment
`status at 6 months, 11 of 13 good responders but only 1 of 14 poor responders had a
`maximal PSA decline greater than 50% (Figure 3C). Akin to the findings in the preclinical
`model, GR positivity at baseline was low: 3% of tumor cells in good responders and 8% in
`poor responders. Of note, 3 of 22 tumors had evidence of high GR expression at baseline (≥
`20% of tumor cells) and all three had a poor clinical response (Figure 3C,D). At 8 weeks,
`the mean percentage of GR positive cells was higher than baseline levels in both response
`groups but was more significantly elevated in poor responders (29% vs 8%, p=.009). In
`addition, the percentage of GR-positive cells at 8 weeks was significantly higher in poor
`compared to good responders (29% versus 10%, p=.02) (Figure 3C,D), and similar results
`were obtained when the analysis was limited to patients from whom matched baseline and 8
`week samples were available for analysis (Figure 3E). Furthermore, when GR IHC data was
`dichotomized based on PSA decline instead of clinical response, GR induction was also
`associated with a limited PSA decline (Figure S2). These findings establish a correlation
`between GR expression and clinical response to enzalutamide and raise the possibility that
`AR inhibition may induce GR expression in some patients. The fact that PSA levels also
`correlate with GR expression raises the question of whether transcriptional regulation of a
`canonical AR target gene may be regulated by GR.
`
`GR expressing drug-resistant tumors show uneven restoration of AR target genes
`
`Having implicated GR as a potential mediator of antiandrogen resistance, we next asked if
`restored AR pathway activity also plays a role by comparing the mRNA transcript levels of
`74 direct AR target genes in control, 4 day, and resistant tumors from the validation cohort
`(Figures S3) as well as eight LREX’ tumors (Figure 4A) (see experimental procedures and
`Supplementary Table 2 for details on gene selection). Consistent with the data generated in
`the pilot cohort (Figure 1A), some AR target genes in resistant tissues showed elevated
`levels relative to control (SGK1, STK39) while other genes (NDRG1, TIPARP, PMEPA1)
`showed no evidence of restored expression.
`
`To examine restoration of AR signaling across the entire set of 74 target genes, we
`calculated a fractional restoration value using log 2 transformed expression values and the
`equation (Resistant – 4 day) / (Control – 4 day). With this approach, a gene whose
`expression in resistant tissue equals the expression in control tumors calculates as 1, while a
`gene whose expression in resistance equals its expression after 4 days of antiandrogen
`treatment equals 0. (Values greater than one indicate hyper-restoration in resistance relative
`to control and values below zero suggest further inhibition as compared to acute treatment.)
`These data confirmed that the pattern of restoration varied gene by gene, but this pattern was
`consistent in LREX’ xenografts and in the validation cohort tumors (Pearson r .64, p = 7.54
`× 10−10, Figure 4B). This finding is most consistent with a model in which AR remains
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 5
`
`inhibited in drug-resistant tumors but expression of certain AR target genes is restored by an
`alternative transcription factor, possibly GR. The fact that AR restoration values were
`somewhat higher in the LREX’ analysis correlates with higher GR expression in these
`tumors (Figure 4C).
`
`GR drives expression of AR target genes in resistant tissues
`
`To determine if GR can drive expression of this subset of AR target genes, we compared, in
`vitro, DHT-induced (AR) and dexamethasone (Dex)-induced (GR) expression of 7 AR
`targets that represent the spectrum of restoration noted in the in vivo analysis, as well as PSA
`(Figure 4D). All 8 genes were regulated by DHT as expected, and this regulation was
`blocked by enzalutamide. Thus, AR signaling remains intact and can be inhibited by
`antiandrogens in these drug-resistant cells, making an AR-dependent mechanism of drug
`resistance less likely.
`
`In contrast to DHT, the effect of Dex on these same target genes was variable but closely
`matched the pattern observed in drug resistant xenografts. For example, Dex strongly
`induced SGK1 and STK39 but did not induce TIPARP, NDRG1, and PMEPA1. Of note,
`KLK3 (PSA) was comparably induced by either DHT or Dex, providing evidence that
`persistent PSA expression in patients responding poorly to enzalutamide could be driven by
`GR. As expected, enzalutamide did not notably affect Dex activity. To confirm that this
`pattern of GR-dependent gene expression is not unique to LREX’ cells, we introduced a GR
`expressing retrovirus into parental LNCaP/AR cells and observed a similar pattern of DHT-
`versus Dex-induced gene expression (Figure S4A, S4B). To be sure that the effects of Dex
`in these models are mediated through GR, we co-treated cells with a previously described
`competitive GR antagonist that lacks AR binding called compound 15 (Wang et al., 2006).
`Compound 15 significantly decreased expression of Dex-induced genes, confirming that
`Dex activity in the LREX’ model is GR-dependent (Figure S4C). Lastly, siRNA
`experiments targeting AR confirmed that AR is not necessary for Dex-mediated gene
`activation (Figure S4D). Collectively these experiments demonstrate that GR is able to drive
`expression of certain AR target genes independent of AR.
`
`AR and GR have overlapping transcriptomes and cistromes
`
`To explore AR and GR transcriptomes in an unbiased fashion, we performed expression
`profiling after short-term treatment of LREX’ cells with DHT or Dex in the presence or
`absence of enzalutamide. AR and GR signatures were respectively defined as all genes with
`absolute expression change greater than 1.6 fold (FDR<.05) after 1 nM DHT or 100 nM Dex
`treatment (Supplementary Table 3). Of the 105 AR signature genes and 121 GR signature
`genes, 52 were common to both lists (Figure 5A). An even larger proportion of AR or GR
`signature genes (>80%) showed evidence of regulation by the reciprocal receptor using
`different thresholds for expression differences (Supplementary Table 3). Heatmap analysis
`of these genes confirmed significant overlap in DHT- versus Dex-induced gene expression
`and showed that Dex-induced gene expression is not impacted by enzalutamide treatment
`(Figure 5B). These findings support the hypothesis that GR activity can bypass
`enzalutamide-mediated AR inhibition by regulating a distinct but significantly overlapping
`transcriptome.
`
`We next addressed the question of whether transcriptomes of enzalutamide-resistant tumors
`are more likely to be explained by AR- or GR-driven gene expression using gene set
`enrichment analysis (GSEA). To define gene sets that distinguish AR and GR activity,
`expression of AR and GR signature genes was first evaluated by GSEA in the DHT- and
`Dex-treated samples from which they were derived. As expected, GR signature genes were
`enriched in the Dex-treated samples and AR signature genes were enriched with DHT
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 6
`
`treatment (Figure 5C). Because several of the genes did not distinguish AR and GR status
`due to their overlapping transcriptional activities, we refined the lists into AR selective
`genes (defined as the AR induced signature genes that were also more highly expressed in
`DHT treated samples relative to Dex treated samples, n=39) and GR selective genes (defined
`as the converse, n=67) (Supplementary Table 3). GSEA analysis of these selective gene lists
`revealed that GR selective genes were strongly enriched in the enzalutamide-resistant
`LREX’ tumors whereas AR selective genes were strongly enriched in the control tumors
`(Figure 5D). These data provide compelling, unbiased evidence that drug resistance is
`associated with a transition from AR- to GR-driven transcriptional activity.
`
`One prediction of this model is that GR should occupy a substantial portion of AR binding
`sites in drug resistant cells. To address this question, we conducted ChIP-seq experiments to
`define AR and GR DNA binding sites in LREX’ cells after DHT and Dex treatment
`respectively. Of note, 52% of the AR binding sites identified after DHT treatment were
`bound by GR after Dex treatment (Figure 5E). We examined the remaining 48% of AR
`peaks more closely to be sure that these peaks were not scored as GR negative simply
`because they fell just below the threshold set by our peak calling parameters. When we
`plotted the average AR and GR signal as a measure of the relative strength of AR and GR
`peaks, we found little evidence of GR binding at the AR unique sites (Figure S5A),
`confirming that these peaks were indeed unique to AR. Next we conducted motif analysis to
`explore potential differences between AR/GR overlap versus AR unique sites. The core
`ARE/GRE consensus sequence was present in both groups (66% and 68% of peaks) but AR/
`GR overlap peaks were relatively enriched for the FoxA1 motif (64% versus 45% of peaks,
`p=2.2×10-16) (Figure 5E). Similar analysis of the GR cistrome defined GR unique and AR/
`GR overlap peaks and revealed that a higher proportion of GR binding sites were unique to
`GR. Interestingly, GR unique peaks were highly enriched for the FoxA motif (Figure 5F),
`while the classic ARE/GRE was not reported by the motif discovery algorithm (MEME) and
`was found only 25% of the time.
`
`Although these cistrome studies provide evidence of substantial overlap between AR and
`GR binding sites in enzaluamide-resistant cells, several lines of evidence indicate that the
`transcriptional differences in DHT- versus Dex-induced gene expression cannot be
`explained solely by DNA binding. For example, ChIP RT-qPCR experiments showed
`significant AR and GR DNA binding at genes induced by both receptors (SGK1, FKBP5,
`PSA) but also at genes such as NDRG1 that are transcriptionally activated by DHT but not
`Dex (Figure S5B). Integrative ChIP-seq and transcriptome analysis provided further
`evidence that DNA binding is not sufficient to determine transcriptional competence. Of the
`56 AR signature genes found to have an AR binding peak, 49 showed at least some
`transcriptional regulation by GR (1.2 fold expression change, p<.05). 38 of these 49 GR
`regulated genes (78%) had an overlapping AR/GR binding peak, confirming substantial
`overlap at co-regulated genes. But GR peaks were also found in 3 of the 7 AR targets genes
`(43%) with no apparent GR transcriptional regulation (Figure S4C). Others have reported
`evidence of allosteric regulation of hormone receptor complexes by specific DNA sequences
`independent of binding affinity (Meijsing et al., 2009), a phenomenon that may also be
`relevant here.
`
`Activation of GR by dexamethasone is sufficient to confer enzalutamide resistance
`
`Whereas LNCaP/AR cells acquire GR expression after prolonged exposure to enzalutamide,
`some prostate cancer cell lines derived from CRPC patients (DU145, PC3, VCaP) express
`endogenous GR (Figure 6A). DU145 and PC3 cells are AR-negative and hence resistant to
`enzalutamide but VCaP cells are enzalutamide-sensitive in vitro (Tran et al., 2009). IHC
`analysis showed diffuse, primarily cytoplasmic GR expression under standard culture
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 7
`
`conditions that lack glucocorticoid supplementation (Figure S6A). To test if GR activation
`by addition of glucocorticoids impacts antiandrogen sensitivity, we treated VCaP cells with
`enzalutamide in the presence or absence of Dex. Enzalutamide inhibited growth as expected,
`but co-treatment with Dex reversed this growth inhibition (Figure 6B). Additional studies
`with the GR antagonist, compound 15, or with GR shRNA restored enzalutamide sensitivity,
`provided pharmacologic and genetic evidence that GR confers resistance (Figure 6C, 6D,
`6E). Of note, GR knockdown (which inhibits GR more completely than compound 15,
`which has mixed agonist/antagonist properties(Wang et al., 2006)) augmented the activity of
`enzalutamide even in the absence of Dex (Figure 6D,F), suggesting that even the weak basal
`GR activity seen under our standard cultures conditions can confer relative resistance to
`enzalutamide. This result also suggests that a pure GR antagonist could enhance the activity
`of enzalutamide in prostate cancers co-expressing GR and AR.
`
`To determine if Dex activates a subset of AR target genes in VCaP (as we observed in the
`LREX’ model), we derived a list of AR target genes in VCaP cells exposed to DHT and
`asked whether Dex could modulate these same AR target genes in the presence of
`enzalutamide. Dex restored expression of some targets (KLK2, FKBP5, HOMER2,
`SLC45A3) but not others (DHCR24, SLC2A3, TRPM8, TMEM79), analogous to the uneven
`restoration we observed in the LNCaP/AR model (Figure 6G). Dex also induced expression
`of the clinical biomarker PSA in these cells, further supporting the hypothesis that GR can
`drive PSA progression in enzalutamide-resistant patients (Figures S6B, C). To confirm that
`Dex activated genes via the glucocorticoid receptor, we evaluated the effect of compound 15
`on Dex induced transcriptional activity. As expected, compound 15 reduced Dex induction
`of the GR targets KLK2 and FKBP5 (Figure 6H). Similarly, GR knock-down prevented
`Dex-mediated induction of target genes (Figure S6C). As in the LREX’ system
`(Supplementary Table 3), the vast majority of genes robustly regulated by GR activation in
`VCaP cells were also regulated by AR activation with DHT (Supplementary Table 4).
`Others have recently shown substantial overlap in the AR and GR cistromes in VCAP as
`well(Sahu et al., 2013). These findings extend our hypothesis that GR promotes
`enzalutamide resistance largely by replacing AR activity at a subset of genes to a second
`model system.
`
`A subset of prostate cancer are primed for GR induction in the setting of AR inhibition
`
`In considering potential mechanisms for increased GR expression in drug-resistant tumors,
`we noted several observations that suggested two distinct models. First, flow cytometry
`analysis of LNCaP/AR and CS1 cells revealed GR expression in a rare subset of cells (Fig
`1E), raising the possibility that these cells clonally expand under the selective pressure of
`antiandrogen therapy. Consistent with this model, we observed rare GR-positive cells in a
`tissue microarray analysis of 59 untreated primary prostate cancers (Supplementary Table
`5). However, we also observed a modest (~2 fold) but significant increase in GR mRNA
`levels in LNCaP/AR xenografts after only 4 days of antiandrogen treatment, reminiscent of
`an older report of increased GR expression in normal ventral rat prostate after castration
`(Davies and Rushmere, 1990). These findings suggest a second model of adaptive resistance
`whereby AR inhibition causes an increase in GR levels due to loss of AR-mediated negative
`feedback.
`
`To investigate the relationship between AR activity and GR expression, we first asked if the
`high level of GR expression in LREX’ tumors is maintained after discontinuation of
`enzalutamide. Remarkably, GR mRNA levels dropped by ~5 fold 8 days after treatment
`discontinuation (Figure 7A). Because enzalutamide has a prolonged half-life in mice (Tran
`et al., 2009), it is difficult to make definitive conclusions about negative feedback loops
`using in vivo models. Therefore, we conducted similar enzalutamide withdrawal
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 8
`
`experiments in LREX’ cells cultured in vitro. GR mRNA levels dropped as early as 1 day
`after discontinuation and continued to decline throughout the 23 days of the experiment
`(Figure 7B). Additional experiments with LREX’ cells using earlier timepoints in charcoal
`stripped media showed reduced GR mRNA levels after only 8 hours DHT exposure and this
`reduction was reversed by co-treatment with enzalutamide (Figure 7C). This reduction
`correlated precisely with the recruitment of an AR binding peak in an intronic enhancer of
`GR identified by ChIP, suggesting AR directly represses GR expression in these cells
`(Figure 7D).
`
`To determine if the loss of GR expression upon enzalutamide withdrawal occurs across the
`entire cell population or is restricted to a subset of cells, we conducted flow cytometry
`experiments, where a shift in median signal intensity can be used to identify expression
`changes in the bulk cell population. (Expression changes limited to a minority sub-
`population would not affect the median and would instead be identified as a tail population
`by histogram plot.) We observed an exponential decay in median GR protein signal (half-
`life 7.6 days) (Figure 7E,top row, 7F), confirming that the loss in GR expression occurs
`across the entire LREX’ cell population. Extension of this experiment to later time points
`(17 weeks) revealed a plateau in loss of GR expression by 7 weeks (Figure S7A).
`
`Next we conducted the reciprocal experiment of re-exposure of LREX’ cells to
`enzalutamide following GR downregulation after prolonged enzalutamide withdrawal
`(LREX’off). GR expression was regained with induction kinetics essentially reciprocating
`the rate of decay previously seen with removal of drug (doubling time 6.8 days),
`establishing that the resistant line remained poised for GR induction in the setting of AR
`inhibition (Figure 7E,F). Consistent with the time scale, continued drug exposure for 7
`weeks was associated with a clear shift in GR expression in essentially all cells (Figure
`S7A).
`
`We next determined if AR inhibition is sufficient to induce GR expression in LNCaP/AR or
`CS1 cells that had not previously been exposed to enzalutamide. In contrast to LREX’, there
`was no change in median expression intensity in CS1 or LnCaP/AR over the 4 week
`experiment, indicating that most cells do not turn on GR expression simply as a consequence
`of AR inhibition (Figures 7E, 7F, S7C). However, the area under the GR staining population
`did increase. Given the weak antiproliferative effect of enzalutamide in vitro (Figure S7B),
`we conclude that this increase in GR expression is most likely explained by loss of AR-
`mediated negative feedback rather than by clonal expansion. Together, these findings
`support a model in which a subset of prostate cancer cells are “primed” for GR induction in
`the context of AR inhibition through an adaptive resistance mechanism (via AR-mediated
`negative feedback). We postulate that these cells then clonally expand under the selective
`pressure of AR blockade, eventually emerging as drug-resistant tumors whose expression
`profiles may resemble those of AR-driven tumors but are driven by GR (Figure 7G).
`
`Discussion
`
`Following the recent approvals of the next generation AR pathway inhibitors abiraterone
`and enzalutamide, the treatment of metastatic prostate cancer has evolved to a two-stage
`process. Initially patients receive conventional androgen deprivation therapy, typically with
`a gonadotropin-releasing hormone agonist that lowers testosterone (castration), often in
`conjunction with an anti-androgen such as bicalutamide. Preclinical and clinical studies have
`conclusively demonstrated that acquired resistance to conventional androgen deprivation
`therapy is caused by restoration of AR pathway activation, primarily due to increased AR
`expression. These discoveries provided the rationale for the development of next generation
`AR therapies.
`
`Cell. Author manuscript; available in PMC 2014 June 05.
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`NIH-PA Author Manuscript
`
`

`
`Arora et al.
`
`Page 9
`
`Here