Activation of MAPK Signaling by CXCR7 Leads to Enzalutamide Resistance in Prostate Cancer Free

The androgen receptor (AR) is a major driver of the growth of prostate cancer (1, 2) and aberrant AR signaling remains to play major roles in final-stage castration-resistant prostate cancer (CRPC) that have developed resistance to the second-generation AR antagonist enzalutamide, or the androgen biosynthesis inhibitor abiraterone acetate (3–7). Furthermore, elevated expression of the glucocorticoid receptor can bypass dependence on the AR by promiscuously activating AR target genes (8) or through lineage plasticity driven by SOX2 (9). Clearly, there are multiple escape pathways to enzalutamide resistance, and a comprehensive understanding of these underlying mechanisms is critical for identifying novel therapeutic targets to improve CRPC outcomes.

The chemokines and chemokine receptors play important roles in chemotaxis, inflammation, and cancer dissemination. CXCR7 is an atypical chemokine receptor (ACKR3), as it functions through G-protein–independent mechanisms. Activated CXCR7 interacts with β-arrestin 2 (ARRB2) and the complex internalizes into clathrin-coated endosomes within the cells and recruits MAPK proteins for the phosphorylation of ERK1/2 (10). Although able to bind CXCL12 (also called SDF-1α), CXCR7 has been shown to be constitutively active in some cells including breast cancer cells, demonstrating ligand-independent activation (11). CXCR7 expression was detected in prostate cancer (12), and emerging evidence suggests important roles of CXCR7 in prostate tumorigenesis (13–15). Here we identify CXCR7 as a top upregulated gene in enzalutamide-resistant prostate cancer and demonstrate important roles of CXCR7/MAPK/ERK signaling in driving prostate cancer resistance to enzalutamide.

LNCaP, VCaP, and HEK293T cell lines were obtained from the ATCC. To authenticate the cells, LNCaP and VCaP cells were subjected to short tandem repeat DNA profiling (Genetica DNA Laboratories) in February 2017, and subsequently the amplified DNA sequences were compared with reference cell database. Only percent match >80% was regarded as authenticated. All the cell lines were tested negative of Mycoplasma contamination by using MycoAlert Detection Kit (Lonza). The number of passages of cells used in all described experiments is within 20 passages. For androgen stimulation cells were first androgen-starved in phenol-red–free RPMI supplement with 5% charcoal-stripped FBS for 3 days, then incubated with either ethanol or 1 nmol/L R1881 for 24 hours for gene expression analysis or 10 nmol/L R1881 for 16 hours for chromatin immunoprecipitation (ChIP). To generate enzalutamide-resistant LNCaP and C4-2B, cells were continuously selected under 1 μmol/L or 10 μmol/L enzalutamide (Selleckchem), respectively, for at least 3 months. C4-2B was a kind gift from A.M. Chinnaiyan. CWR-R1 enzalutamide-resistant cell line was a gift from Dr. Donald Vander Griend (University of Chicago, Chicago, IL). C4-2B abiraterone-resistant cell line was a gift from Dr. Allen Gao (University of California, Davis, Davis, CA). To study MAPK pathway, cells were treated with either 1 μmol/L trametinib (MEK1/2 inhibitor) or SCH772984 (ERK1/2 inhibitor; both Selleckchem) for indicated time points. Enzalutamide, trametinib, and SCH772984 were dissolved in DMSO (Sigma). Recombinant human CXCL12 (SDF1) and EGF were purchased from R&D and Sigma-Aldrich correspondently. Erlotinib was purchased from Selleckchem. All PCR primers used in this study are listed in the Supplementary Table S1. qPCR was performed using SYBR Green by StepOne Plus.

The following antibodies were used: anti-CXCR7 (GTX100027; Genetex); anti-EGFR (A300-387) from Bethyl; anti-EEA1 (610456) from BD Biosciences; Anti-HDAC1 (ab11966), anti-H3 (ab1791), and anti-GAPDH (ab9385) from Abcam; anti-p53 (SC126) and anti-α-tubulin (DM1A) from Santa Cruz Biotechnology; monoclonal anti-Flag (M2) (F1804) and polyclonal anti-Flag (F7425) from Sigma; anti-Ki-67 (9027), anti-phospho-MEK1/2 (Ser217/221) (9154), anti-Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (4370), anti-p44/42 MAPK (Erk1/2) (4695), anti-PSA (2475), and anti-ARRB2 (3857) were purchased from Cell Signaling Technology; and Anti-CXCR7 (MAB42273) and anti-CXCR4 (MAB170SP) are purchased from R&D Systems. pHAGE-Flag-CXCR7 construct was made through subcloning CXCR7 ORF into pHAGE-Flag vector. ARRB2 was first cloned into gateway entry vector pCR8, and then transferred into destination vector pLentiSFB with LR Clonase (Invitrogen). CXCR7p and CXCR7e guide RNA were subcloned into Lenti-CRISPRV2 vector, CXCR7 shRNAs (TRCN0000014509 and TRCN0000363205), ARRB2 (TRCN0000165387 and TRCN0000280686), and CXCR4 shRNA (TRCN0000256866) in pLKO vector and were purchased from Sigma. Packed lentiviral particles were collected from 293T cultures 48 hours post transfection with lentiviral constructs, pSPAX2 and pMD2G. Viral supernatants were passed through a 0.45-mm filter, supplemented with 4 μg/mL polybrene, and used to infect prostate cancer cells.

Briefly, total RNA was isolated from frozen CRPC metastases with RNA STAT-60 (Tel-Test). The purity and yield of the RNA were determined on a NanoDrop by using the sample absorbance at 260 nm and 280 nm, both normalized to the background 340 nm. RNA integrity was assessed on an Agilent 2100 Bioanalyzer, which provided RNA integrity number values for the samples.

To study subcellular protein localization, cells were trypsinized and spun at 500 × g. Compact cell pellets were lysed in subcellular protein fractionation kit for cultured cells (Thermo Fisher Scientific) according to the product instruction. To determine the interaction between CXCR7 and ARRB2, We performed S-beads protein pull-down as described previously.

Cell proliferation assay was carried out using WST-1 Kit according to the manufacturer's instruction (Clontech). To determine synergy of the two drugs, combination index (CI) values were calculated following the manufacturer's instruction (CompuSyn). For the colony formation assay, 4 × 10³ cells were seeded into 6-well plates and monitored for colony formation for 2 weeks. Then, colonies were stained in 0.25% crystal violet in 20% methanol, photographed, and counted. For cell invasion assay, 2 × 10⁴ cells were plated into the top chamber of Matrigel-coated transwell insert (BD Biosciences) in media containing 1% FBS, while 40% FBS condition media, harvested from 3T3, were added to the bottom chamber. After 48 hours, the transmigrated cells were stained with 0.25% crystal violet in 20% methanol, imaged, and quantified. All assays were triplicated.

Cells grown on poly-lysine precoated coverslips were fixed in 4% paraformaldehyde in PBS at room temperature for 15 minutes. After fixation, the cells were permeabilized with a 0.5% Triton X-100 for 15 minutes and blocked with 3% BSA for 30 minutes. Cells were incubated with anti-Flag (1:2,000), anti-EEA1 (1:50) antibodies in blocking buffer at room temperature for 2 hours, then washed extensively with PBS and incubated with Alexa Fluor-488 or -594 secondary antibodies (Invitrogen) at room temperature for 1 hour. Nuclei were counterstained with DAPI (Invitrogen) and mounted using Prolong Gold Antifade Reagent (Invitrogen). Images were captured with Axiovert 200 Fluorescence Microscope (Carl Zeiss) equipped with a Plan Fluor 40× objective lens (1.3 oil) and AxioCam HRc camera. Images were modified by Photoshop CS4 (Adobe).

Standard IHC was performed with antibodies against CXCR7 and p-ERK1/2 on four tissue microarrays representing clinically localized prostate cancer and metastatic prostate cancer using the Bond Polymer Refine Detection Horseradish Peroxidase (HRP; Leica Biosystems, DS9800) method. Four-micron thick sections were cut and dried in oven at 60°C for 1 hour, then the deparaffinization and antigen retrieval (pH6; 20 minutes) was done on-line using Leica Bond-Max (Leica Biosystems). Antibody was stained as such, the sections washed with bond dewax solution (Leica Biosystems, AR9222) at 72°C then washed with 100% ethanol, followed by washing with bond wash solution, after that sections target antigen were unmasked by incubation in Bond ER 1 Solution (pH6) for 20 minutes at 100°C (Leica Biosystems, AR9961), subsequently, slides were washed with bond wash solution. Immunostaining for the primary antibodies was done by incubation of primary antibodies for 15 minutes at room temperature. Following primary antibody incubation, sections were incubated with nonconjugated secondary rabbit anti-mouse IgG for 15 minutes at room temperature. Subsequently, following washing sections were incubated with HRP-conjugated polymer anti-rabbit poly-HRP-IgG for 15 minutes at room temperature. Then the endogenous peroxidase activity was blocked with hydrogen peroxide for 5 minutes. Immunoreactivity was visualized with 3, 3-diaminobEnzidine (DAB) chromogen for 10 minutes at room temperature. Finally, sections were counterstained with Mayer Hematoxylin for 5 minutes and dehydrated by graded alcohol to xylene using Leica Autostainer XL and mounted by Leica Cover Slip (Leica CV5030). Product score of staining percentage and intensity was used as an overall measure. Intensity was scored as negative (score = 0), weak (score = 1), moderate (score = 2), or strong (score = 3), which was multiplied by staining percentage to produce the product score for each core. In total, 30 localized prostate cancer and 106 metastatic prostate cancer tissue cores were evaluable out of 4 tissue microarrays each for CXCR7 and p-ERK1/2 expression.

Enzalutamide-resistant LNCaP and DMSO LNCaP cell cultures were trypsinized and washed in 1× PBS. Cells were resuspended, and aliquoted in 5 × 10⁵ cells in 100 μL per staining condition. Cell samples were fixed in 0.5× IC Fixation Buffer (eBioscience, Invitrogen) and stained with primary anti-CXCR7 antibodies, clone 11G8 (MAB42273) from R&D, supplemented or not with 1× Permeabilization Buffer (eBioscience, Invitrogen). The secondary staining was done with goat anti-rabbit antibodies conjugated with Alexa Fluor-488 antibodies (Invitrogen). Cell nuclei were stained with DAPI (Invitrogen) solution. Sample acquisition was done on BD LSRFortessa analyzer, data were analyzed on FlowJo V10.0.8r1 (FlowJo, LLC).

Briefly, cells were cross-linked with 1% formaldehyde for 10 minutes and the reaction was quenched by 0.125 mol/L glycine for 5 minutes at room temperature. Cells were then rinsed with cold 1× PBS twice, incubated with cell lysis buffer and subsequently nuclear lysis buffer. Chromatin was sonicated and fragmented to a size of 200–500 bp, precleared with agarose/protein A or G beads (Upstate), and incubated with 3–5 μg of anti-AR antibodies (N-20) overnight. Protein–DNA complexes were precipitated, washed, and eluted. Microarray expression profiling was performed using HumanHT-12 v 4.0 Expression BeadChip (Illumina). Bead-level data were preprocessed and normalized by GenomeStudio. Differentially expressed genes were identified by Bioconductor limma package (cutoff P < 0.005). Heatmap view of differentially expressed genes was created by Cluster and Java TreeView. Gene Ontology (GO) term enrichment was analyzed using DAVID and plot was drawn by R package ggplot2. Gene set enrichment analysis (GSEA) was performed according to the manufacturer's instruction (Broad Institute). The Gene Expression Omnibus accession number for the data reported in this study is GSE104935.

Human prostate cancer specimens were obtained as part of the University of Washington Medical Center Prostate Cancer Donor Program, which is approved by the University of Washington Institutional Review Board. All specimens for IHC were formalin-fixed (decalcified in formic acid for bone specimens), paraffin-embedded and examined histologically for the presence of nonnecrotic tumor. Tissue microarrays (TMA) were constructed with 1-mm diameter duplicate cores (number of cores = 212) from CRPC patient tissues (number of patients = 34) consisting of visceral metastases and bone metastases (number of sites = 106) from patients within 8 hours of death. TMAs of primary prostate cancer (number of patients = 30, number of sites = 30) were generated at the Northwestern University Pathology Core through the prostate SPORE program, approved by Northwestern University Institutional Review Board.

All procedures were approved by the Northwestern University Institutional Animal Use and Care Committee. Five-week-old male nude athymic BALB/c nu/nu mice (Charles River Laboratory) were purchased for the experiment. To study the function of CXCR7 in enzalutamide-resistant prostate cancer growth, 6 × 10⁶ enzalutamide-resistant C4-2B stably expressing pLKO nontarget control or shCXCR7 (1:1 to Matrigel; BD Biosciences) were inoculated into the right dorsal flank of mice by subcutaneous injection. To evaluate the therapeutic effect of MEKi treatment on enzalutamide-resistant CRPC, 6 × 10⁶ C4-2B cells or enzalutamide-resistant VCaP were inoculated into the right dorsal flank of mice by subcutaneous injection. Once the tumors reached approximately 200–300 mm³, tumor bearing mice were castrated and awaited castration-resistant growth for 2 weeks. Then, mice were randomized into 2 groups: (i) treated with enzalutamide (10 mg/kg/day) alone and (ii) enzalutamide (10 mg/kg/day) + trametinib (1 mg/kg/day) for C4-2B or (3 mg/kg/day) for VCaP orally for 5 days once a day and total of 4 weeks. Tumor size was measured twice a week and calculated by the formula [length (mm) × width² (mm²) × 0.5]. At the endpoint mice were euthanized and tumor tissues were excised, weighted, and extracted for protein lysates.

To search for potential drivers of prostate cancer resistance to enzalutamide, we performed expression profiling of DMSO-treated and enzalutamide-resistant LNCaP cells (Supplementary Fig. S1A and S1B). GSEA demonstrated that AR-induced genes were downregulated in enzalutamide-resistant cells, while AR-repressed genes upregulated, confirming the on-target molecular effects of enzalutamide even in resistant cells (Supplementary Fig. S1C and S1D). Analysis of triplicate experiments identified 284 and 294 genes (>2-fold; FDR adjusted P < 0.001) that were induced and repressed, respectively, in enzalutamide-resistant LNCaP cells (Fig. 1A). GO analysis revealed that these genes were involved in previously reported pathways such as steroid hormone biosynthesis but also in novel concepts such as signal peptides and membrane proteins, suggesting increased receptor signaling (Fig. 1B). Interestingly, CXCR7 was among the most upregulated genes in enzalutamide-resistant cells as compared with the DMSO-treated cells.

To extend this to CRPC models, global expression profiling of enzalutamide-resistant C4-2B cells (Supplementary Fig. S1E and S1F) revealed many upregulated genes ranked top, among which was CXCR7 (Fig. 1A). qRT-PCR confirmed that CXCR7 was increased by 8 and 5 fold in enzalutamide-resistant LNCaP and C4-2B cells, respectively (Fig. 1C). Time-course analysis revealed that CXCR7 mRNA began to increase at 3 days after enzalutamide treatment and continued to increase over time (Fig. 1D), while Western blotting confirmed CXCR7 protein increase in four enzalutamide-resistant prostate cancer lines as compared with their DMSO-treated control (Fig. 1E and F), indicating a common mechanism. To determine whether CXCR7 drives enzalutamide resistance, control and CXCR7-expressing LNCaP cells were treated with increasing doses of enzalutamide (Fig. 1G). Cell growth assays demonstrated that CXCR7 overexpression greatly reduced the sensitivity of LNCaP cells to enzalutamide (Fig. 1G). On the contrary, knockdown of CXCR7 in enzalutamide-resistant LNCaP cells markedly resensitized them to enzalutamide (Fig. 1H), supporting CXCR7 as a driver of prostate cancer enzalutamide resistance.

We have recently shown that AR can function as a transcriptional repressor to directly inhibit target genes such as CCN3 (16, 17). Interestingly, we found that both CXCR7 mRNA and protein levels were decreased by synthetic androgen R1881 in a dose- and time-dependent manner (Fig. 2A and B; Supplementary Fig. S2A and S2B). Concordantly, AR knockdown restored CXCR7 expression (Fig. 2C and D). Examining previously published AR ChIP-seq data (18), we found that R1881 treatment, which promotes AR nuclear translocation, leads to a strong AR binding event at 110 kb downstream of the transcription start site of the CXCR7 gene and a weaker peak at the promoter (Fig. 2E). ChIP-qPCR confirmed abundant AR occupancy at the CXCR7 enhancer and weaker but significant AR enrichment at the promoter (Fig. 2F). Similar results were also observed in another prostate cancer cell line (Supplementary Fig. S2C and S2D).

To examine the importance of the AR-binding peaks in regulating CXCR7 gene expression, we first performed motif analysis and identified a well-centered androgen-response element (ARE) at both CXCR7 promoter and enhancer sites (Supplementary Fig. S2E and S2F). Next, we designed guide RNA pairs flanking these ARE elements, which were cotransfected to LNCaP cells with CRISPR-Cas9 lentivirus. CRISPR-Cas9–mediated deletion in a pooled population was validated through PCR analysis of genomic DNA revealing a wild-type and a shorter, ARE-deleted PCR product (Fig. 2G). ChIP-qPCR showed substantially decreased AR recruitment to the CXCR7 enhancer and to a lesser degree the CXCR7 promoter (Fig. 2H). Concordantly, qRT-PCR analysis demonstrated that deletion of the CXCR7 enhancer led to markedly restored expression of CXCR7, whereas deletion of CXCR7 promoter only slightly increased CXCR7 expression (Fig. 2I). Therefore, the downstream enhancer is essential for CXCR7 repression by AR.

To determine the functional roles of CXCR7, we performed CXCR7 knockdown and observed significantly reduced cell proliferation, colony formation, and cell invasion in both enzalutamide-resistant LNCaP and C4-2B cells, with the amplitude of inhibition proportional to the degree of knockdown (Fig. 3A–D; Supplementary Fig. S3A–S3D). Next, to examine whether CXCR7 regulates xenograft prostate cancer growth in vivo, we inoculated pLKO nontarget vector or shCXCR7-transduced enzalutamide-resistant C4-2B cells subcutaneously into the right dorsal flanks of nude mice. Xenograft tumor growth was measured biweekly. Our data revealed that CXCR7 depletion significantly reduced enzalutamide-resistant prostate cancer tumor growth (Fig. 3E). Accordingly, xenograft tumor weight at the endpoint was also much smaller for the grafts with CXCR7-knockdown cells (Fig. 3F and G).

CXCR7 has been shown to activate the MARK pathway, which is pivotal for cell survival and proliferation (19). Indeed, we observed that pERK1/2 was significantly elevated in enzalutamide-resistant LNCaP, C4-2B, VCaP, and CWR-R1 cells as compared with their DMSO-treated controls (Fig. 4A and B). Furthermore, GSEA demonstrated that a MAPK downstream target gene set, as defined previously (20), was enriched for upregulation in enzalutamide -resistant prostate cancer cells (Fig. 4C). To address whether this resistance mechanism is specific to enzalutamide or more broadly to androgen inhibition, we also examined the pERK1/2 level in abiraterone-resistant cell lines. We demonstrated that CXCR7 is upregulated in abiraterone-resistant C4-2B cells with concordance elevation of p-ERK (Supplementary Fig. S4A). To demonstrate that increased CXCR7 level is responsible for MAPK pathway activation in enzalutamide-resistant prostate cancer, we silenced CXCR7 in enzalutamide-resistant LNCaP and observed substantial reduction of pERK1/2 levels, with the amplitude of decrease proportional to the degrees of CXCR7 knockdown (Fig. 4D).

To determine how CXCR7 is activated, we first analyzed CXCL12 expression by ELISA but failed to detect any autocrine production of CXCL12 in enzalutamide-resistant LNCaP or C4-2B cells. To determine whether exogenous CXCL12 can induce pERK1/2 in enzalutamide-resistant cells, we serum-starved the cells for 48 hours followed by treatment with 100 ng/mL of CXCL12 for up to 5 hours. We found that serum starvation does not affect pERK1/2 levels, suggesting CXCL12-independent, constitutive activation of CXCL12 (Fig. 4E). Exogenous CXCL12 further increased pERK1/2 levels after 2 hours of treatment, a delayed response typical of CXCR7-mediated signaling (21). Interestingly, exogenous CXCL12 also induced pERK1/2 in the control cells, although to a lesser degree, likely due to less amount of CXCR7 in these cells. Recent studies have suggested that EGF-stimulated EGFR, which results in EGFR phosphorylation (pEGFR) and activation, may interact with CXCR7 and stimulate pERK1/2 (13, 14). However, our data showed that EGFR expression was abolished to an undetectable level, rather than increased, in enzalutamide-resistant cells as compared with control and pEGFR could not be detected, precluding its role in this context (Supplementary Fig. S4B and S4C).

We next attempted to examine whether CXCR4 is required for CXCR7 activation in prostate cancer. First, we found that CXCR4 was expressed at low levels in most prostate cancer lines; CXCR7 level was several magnitudes higher (Fig. 4F). As controls, CXCR4 level was high in 293T cells, wherein CXCR7 was not expressed. In expression profiling data, we did not observe CXCR4 upregulation in enzalutamide-resistant prostate cancer cells. Moreover, CXCR4 knockdown in enzalutamide-resistant cells (Supplementary Fig. S4D) did not affect pERK1/2 level, whereas CXCR7 knockdown abolished pERK1/2 (Fig. 4G). In aggregate, we demonstrate that MAPK pathway in enzalutamide-resistant prostate cancer cells is activated by CXCR7 in a CXCR4- and CXCL12-independent manner but may be further enhanced by CXCL12, if present.

CXCR7 is known as a membrane protein that, when activated, internalizes into the endosomes forming a complex with ARRB2, which acts as a scaffold protein for MAPK protein assembly and activation (10). We performed flow cytometry on nonpermeabilized prostate cancer cells and indeed found that CXCR7 was present on the plasma membrane of LNCaP and C4-2B cells and was further increased in their enzalutamide-resistant derivatives (Fig. 5A; Supplementary Fig. S5A). Further, fractionation of protein lysates from enzalutamide-resistant LNCaP cells showed CXCR7 localization exclusively to the membranous fraction, which contains membranes of both plasma and cytoplasmic organelle (e.g., endosome; Fig. 5B). We noted that pERK1/2 was present primarily in cytoplasmic and membrane fractions but very weak in nuclear fractions, supporting its activation by CXCR7–ARRB2 complex. Moreover, immunofluorescent costaining validated that CXCR7 localized mainly in cytoplasmic aggregates, colocalizing with early endosome marker EEA1 and ARRB2 (Fig. 5C and D). Finally, coimmunoprecipitation showed that anti-Flag pull down of flag-ARRB2 expressed in enzalutamide-resistant LNCaP cells indeed enriched CXCR7, supporting CXCR7–ARRB2 interaction (Fig. 5E).

To examine whether ARRB2 is involved in CXCR7-mediated MAPK activation in enzalutamide-resistant prostate cancer cells, we knocked down ARRB2 and evaluated pERK1/2 by Western blotting. Depletion of ARRB2 markedly attenuated ERK1/2 phosphorylation (Fig. 5F), echoing the decrease of pERK1/2 upon CXCR7 depletion. Moreover, in a rescue experiment, ARRB2 knockdown abolished pERK1/2 that were induced by ectopic CXCR7 in LNCaP cells, supporting that ARRB2 is required for CXCR7-mediated activation of MAPK (Fig. 5G). Furthermore, analogous to CXCR7 depletion (Fig. 3), ARRB2 knockdown in enzalutamide-resistant cells led to remarkable reduction in cell proliferation, colony formation, and cell invasion (Fig. 5H–J; Supplementary Fig. S5B). We thus conclude that CXCR7 activates MAPK pathway in enzalutamide-resistant cells through cytoplasmic internalization and the recruitment of ARRB2.

To determine the clinical relevance of the CXCR7-MAPK pathway, we analyzed public microarray dataset and observed significant CXCR7 upregulation in CRPC relative to localized prostate cancer in (Fig. 6A; Supplementary Fig. S6A; refs. 22, 23). Furthermore, analysis of RNA-seq data (24) showed substantial CXCR7 upregulation in CRPC compared with primary prostate cancer (Fig. 6B). Furthermore, qRT-PCR of tissue samples confirmed that CXCR7 mRNA level was increased in CRPC relative to localized prostate cancer and was further increased post-abiraterone and/or enzalutamide treatment (Fig. 6C). Moreover, reanalysis of 122 CRPC transcriptomes generated by the International Stand Up To Cancer/Prostate Cancer Foundation (SU2C/PCF) Prostate Cancer Dream Team (25) showed a strong (P = 4 × 10⁻¹¹) correlation of CXCR7 expression and a MEK-functional activation signature (Supplementary Fig. S6B; ref. 26).

Next, we performed IHC assays on several TMA containing localized or metastatic prostate cancer. We found that CXCR7 staining was very high in endothelial cells, which was consistent with previous reports (27), supporting the specificity of the assay. CXCR7 was almost undetectable in localized prostate cancer but expressed strongly in about 30% of CRPC (Fig. 6D and E). Concordantly, pERK1/2 staining was weak in localized prostate cancer and was strongly increased in CRPC (Fig. 6F and G). Importantly, we analyzed a single CRPC available worldwide, that progressed on enzalutamide and was subsequently treated with MEK inhibitor trametinib. After 2 years on trametinib with an ongoing radiographic response characterized by an objective reduction in the size of enlarged retroperitoneal lymph nodes, the patient died of an unrelated cause. Lymph node metastases were collected post enzalutamide but prior to trametinib. IHC staining revealed indeed a concordant increase of CXCR7 and pERK staining compared with localized prostate cancer (Fig. 6H). Taken together, our data confirmed CXCR7 and MAPK pathway upregulation in CRPC and suggested them as promising targets for therapeutic intervention.

To test the effects of MAPK inhibitors, Western blot analysis showed that trametinib at 2 μmol/L is sufficient to abolish pERK1/2 in LNCaP cells induced by enzalutamide (0.5 μmol/L; Fig. 7A). The specificity of trametinib on ERK signaling was validated as various known CXCR7 downstream phospho-proteins such as pAKT, pSTAT3, pJNK, pMEK1/2, and p-p38 remained unaffected upon trametinib treatment (Supplementary Fig. S7A). Concordantly, cell growth assay and colony formation assays showed combinatorial effects of enzalutamide and trametinib in inhibiting proliferation of LNCaP and VCaP cells (Fig. 7B and C; Supplementary Fig. S7B and S7C). In addition, we evaluated the combination effect of enzalutamide and trametinib with the use of CompuSyn software. CI revealed that all the tested doses of enzalutamide + trametinib exhibit synergistic interaction (CI < 1; Fig. 7D; Supplementary Fig. S7D).

Next, to test the therapeutic effects of trametinib in vivo, enzalutamide-resistant VCaP or enzalutamide-resistant C4-2B cells were inoculated subcutaneously into the dorsal flanks of nude mice. Upon initial tumor formation, all mice were castrated. Once tumors stabilized after castration, mice were treated with either enzalutamide alone or in combination with trametinib. Tumor size was measured twice per week. Enzalutamide-resistant xenografts treated with enzalutamide alone progressed steadily as expected. Significantly, xenografts treated with the drug combination not only failed to progress but showed remarkable tumor regression (Fig. 7E and F; Supplementary Fig. S7E and S7F). Accordingly, average tumor weight from mice that received combinatorial treatment was significantly lower than those that received enzalutamide monotherapy (Fig. 7G; Supplementary Fig. S7G). Western blotting confirmed decreased pERK1/2 by trametinib (Fig. 7H; Supplementary Fig. S7H). Thus, our data strongly suggest that the MEK inhibitor trametinib may be useful for the treatment of enzalutamide-resistant CRPC.

Recent studies have reported AR as a transcriptional repressor, which is orchestrated by other transcriptional repressors such as EZH2 and LSD1 (16, 28). Here, we identify CXCR7 as a direct target of AR-mediated transcription repression. In contrast to a recent article reporting AR regulation of CXCR7 through the promoter (14), we used ChIP and CRISPR-Cas9 system to establish that CXCR7 is repressed primarily by AR-bound enhancer 110 kb away from the transcription start site. AR binding at the CXCR7 promoter is much weaker and is most likely indirect, due to the AR-bound enhancer forming chromatin loops to the promoter, which is typical for AR-regulated genes (17, 29, 30). Similar to CXCR7, AR-repressed genes such as c-MET, HOTAIR, and SOX2 have previously been shown upregulated upon ADT and act as oncogenes to drive CRPC (30–32).

Since its initial discovery as a cognate receptor of CXCL12 in 2006 (33), CXCR7 has been reported to act through both constitutive- and ligand-dependent manner (11, 19, 34, 35). Concordantly, we found that elevated CXCR7 in enzalutamide-resistant prostate cancer turns on downstream signaling independently of CXCL12, albeit the effect can be further enhanced by CXCL12 stimulation. Although a previous study has suggested the formation of CXCR4-CXCR7 heterodimer upon ectopic expression (36), endogenous complex has not been isolated. Our data showed that CXCR4 is barely expressed in prostate cancer cell lines, wherein CXCR7 was expressed at a level of several magnitudes higher. Furthermore, only CXCR7, but not CXCL12 and CXCR4, was upregulated in enzalutamide-resistant lines and our data showed that CXCR4 is not required for CXCR7-stimulated MAPK activation. We also tested whether CXCR7 becomes constitutively active due to emergence of gene mutation, but no mutation in CXCR7 was detected in our resistance lines, which suggested that CXCR7 activation is less likely due to acquired mutation. A recent study of obese HiMyc mice revealed high levels of CXCL12 in prostate stroma and high staining of CXCR4 and CXCR7 proteins in the epithelial compartment of mice tumors (15). It would be interesting to examine CXCR7 signaling within the tumor microenvironment, which however is beyond the scope of this study.

Unlike CXCR4, activated CXCR7 interacts with ARRB2, rather than G proteins, to stimulate receptor internalization and downstream MAPK/ERK activation (10). Indeed, we found that ARRB2 is required for the roles of CXCR7 in activating MAPK/ERK signaling and promoting enzalutamide-resistant prostate cancer. This is in contrast to a recent study reporting a role of ARRB2 in suppressing CXCR7-mediated EGFR transactivation upon exogenous EGF stimulation (37). We propose that this is due to promiscuous CXCR7 interaction with abundant EGF that was exogenously provided, which may compete with ARRB2. However, whether this interaction is relevant in a physiologic setting is unclear.

Our data suggests CXCR7 and MAPK/ERK pathway as promising targets for therapeutic intervention. Many attempts have been made to generate CXCR7 antagonists with CCX771 being the most prominent. However, recent studies have unfortunately determined that CCX771, despite its ability to bind CXCL12, actually acts as an agonist (38), which raises concerns regarding studies using CCX771 as CXCR7 antagonist (39). Here, we took advantage of the readily available MAPK inhibitors to block CXCR7 downstream pathways. We are aware that MAPK targeting as a single agent may lead to rapid development of resistance (40) and was suggested to be used in combination with other targeted agents to mitigate drug resistance (41). Our data strongly support that trametinib in combination with enzalutamide may be highly effective in the treatment of CRPC.

A. Sharp has received speakers bureau honoraria from Astellas and has provided expert testimony (travel support) for Genetec-Roche and Sanofi. P.S. Nelson is a consultant/advisory board member for Astella and Janssen. F.Y. Feng is a consultant/advisory board member for Medivation/Astellas, Janssen, Sanofi, Bayer, Dendreon, Ferring, Blue Earth, and Celgene, and has provided expert testimony for PFS Genomics. J.S. de Bono is a consultant/advisory board member for Astellas, AstraZeneca, Janssen, Genentech/Roche, CellCentric, Carrick, Eisai, GSK, Pfizer, MSD, Merck Serono, Cekgene, Daiichi, Bayer, and Seattle Genetics. M.B. Rettig reports receiving commercial research grant from Novartis and Johnson & Johnson, has received speakers bureau honoraria from and is a consultant/advisory board member for Johnson & Johnson. No potential conflicts of interest were disclosed by the other authors.

Conception and design: K.-w. Fong, A. Zhang, P.S. Nelson, J.S. de Bono, J. Yu

Development of methodology: K.-w. Fong, G. Gritsina, A. Zhang, J.S. de Bono, J. Yu

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): S. Li, K.-w. Fong, G. Gritsina, J. Kim, W. Yuan, A.M. Chinnaiyan, J.S. de Bono, C. Morrissey, J. Yu

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): K.-w. Fong, G. Gritsina, A. Zhang, J.C. Zhao, J. Kim, A. Sharp, W. Yuan, X.J. Yang, P.S. Nelson, F.Y. Feng, J.S. de Bono, M.B. Rettig, J. Yu

Writing, review, and/or revision of the manuscript: K.-w. Fong, G. Gritsina, J. Kim, A. Sharp, W. Yuan, P.S. Nelson, F.Y. Feng, A.M. Chinnaiyan, J.S. de Bono, C. Morrissey, M.B. Rettig, J. Yu

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): K.-w. Fong, G. Gritsina, C. Aversa, C. Morrissey, J. Yu

Study supervision: J.S. de Bono, J. Yu

We are grateful to Richard J. Miller (Northwestern University, Chicago, IL) for helpful discussions on CXCR4. We also thank David Dolling, Jon Welti, Ines Figueiredo, Ruth Riisnaes, and Daniel NavaRodrigues from the Institute of Cancer Research, United Kingdom for their insights on CXCR7 and pERK in human prostate cancer specimens. We thank the patients and their families who were willing to participate in the Prostate Cancer Donor Program and the investigators Drs. Robert Vessella, Celestia Higano, Bruce Montgomery, Evan Yu, Heather Cheng, Elahe Mostaghel, Paul Lange, and Martine Roudier for their contributions to the University of Washington Medical Center Prostate Cancer Donor Rapid Autopsy Program. This work was supported by RSG-12-085-01 (to J. Yu) and IRG-18-163-24 (to K.-w. Fong) from the American Cancer Society, PC120466 (to J. Yu) and PC160328 (to J. Yu) from the Department of Defense, R01CA172384 (to J. Yu), R50CA211271 (to J.C. Zhao), P50CA180995 (to J. Yu), P50CA097186 (to P.S. Nelson), T32CA09560 (to G. Gristina), and T32DK007169 (to J. Kim) from the NIH, and Prostate Cancer Foundation 2017CHAL2008 (to J. Yu, J.C. Zhao, P.S. Nelson, C. Morrissey, and M.B. Rettig). Work in the de Bono laboratory was supported by funding from the Movember Foundation, Prostate Cancer UK, US Department of Defense, the Prostate Cancer Foundation, and the UK Department of Health through an Experimental Cancer Medicine Centre grant. A. Sharp is supported by the Medical Research Council, the Academy of Medical Sciences, and Prostate Cancer UK. Research supported by an Stand Up To Cancer-Prostate Cancer Dream Team Translational Research Grant (SU2C-AACR-DT0712). Stand Up to Cancer is a division of the Entertainment Industry Foundation. Research grants are administered by the American Association for Cancer Research, the scientific partner of SU2C.

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.