Increased antiandrogen enzalutamide sensitivity via altering prostate cancer stem cell traits through modulating the androgen receptor-mediated CDR1/circCDR1-AS/miR-1290/BMP4 signaling

Abstract

While the recent FDA-approved antiandrogen enzalutamide (Enz) might prolong the survival of castration-resistant prostate cancer (CRPC) patients by an additional 4.8 months, most patients eventually might still develop Enz resistance within 6–12 months. Although few genes have been linked to Enz resistance in prostate cancer (PCa), the detailed mechanism(s) are still underinvestigated. Here, we found that Enz might function by altering androgen receptor(AR)-mediated CDR1/circCDR1-AS/miR-1290/BMP4 signaling to modulate PCa stem cells (CSCs) to increase Enz resistance. Mechanistic analysis revealed that Enz/AR signaling can transcriptionally regulate CDR1 expression by reducing binding to androgen response elements (AREs) on the CDR1 5’ promoter to alter circCDR1-AS expression. Enz/AR/CDR1/circCDR1-AS signaling might then increase BMP4 expression by altering miR-1290 expression, which involves direct binding to the 3’ UTR of BMP4 mRNA. Preclinical studies using a CWR22Rv1 xenograft mouse model and integrative analysis of GEO cohort data further demonstrated that targeting this newly identified Enz/AR/CDR1/circCDR1-AS/miR-1290/BMP4 signaling pathway with miR-1290, circCDR1-AS-shRNA, or BMP4-shRNA may help develop novel therapies to combat Enz resistance at the later stage of CRPC.

Introduction

Prostate cancer (PCa) is the most commonly diagnosed cancer and the second leading cause of cancer-related death among men in the United States [1]. Targeting androgen receptor (AR) signaling via androgen deprivation therapy (ADT) with various antiandrogens, including the recent FDA-approved powerful antiandrogen enzalutamide (Enz), has become the standard therapy to treat castration-resistant prostate cancer (CRPC) [2, 3]. However, even though Enz can prolong patient survival by an additional 4.8 months, most patients may ultimately fail to respond to ADT therapy because of the development of Enz resistance after 1–2 years [4]. The mechanisms that contribute to Enz resistance are still unclear. Early studies indicated that the induction of the AR splicing variant ARv7 might contribute to the development of Enz resistance [2, 5, 6]. However, other studies have also indicated that non-ARv7 mechanisms, including the induction of other AR mutants or long noncoding RNAs (lncRNAs), could also alter Enz sensitivity [7–9].

Increasing evidence has demonstrated that prostate cancer stem cells (CSCs) may play important roles in the development of ADT resistance in CRPC [10]. Early studies have indicated that alterations in CSCs detected using sphere formation and/or CSC surface markers, including CD44, CD133, SOX2, NANOG, BMP4, OCT4 and integrins [11–13], might contribute to tumor progression and could be targeted to suppress PCa progression [14]. For example, ADT with Enz may alter prostate CSCs to influence therapeutic efficacy [15], and altered CSCs may also contribute to the development of resistance to chemotherapy [16].

Circular RNAs (circRNAs) are a new class of noncoding RNAs that are generated from the noncanonical splicing of linear pre-mRNAs [17, 18]. CircRNAs may regulate gene expression at the post-transcriptional or transcriptional level by acting as microRNA (miRNA) sponges to reduce their availability within cells [19–23]. Interestingly, early studies indicated that miRNAs can also alter CSC traits via multiple mechanisms [24, 25], suggesting that some circRNAs may modulate CSC traits by altering miRNAs.

Here, we assumed that Enz might function by altering circRNAs through AR signaling to control a subset of miRNAs to alter CSC traits and reduce Enz resistance in CRPC.

Materials and methods

Reagents and materials

Anti-AR (1:1000; Santa Cruz Biotechnology) and anti-GAPDH (1:1000; Santa Cruz Biotechnology) antibodies were purchased from Santa Cruz Biotechnology, and the anti-BMP4 antibody was purchased from GeneTech. The anti-mouse and anti-rabbit antibodies used for Western blot were obtained from Invitrogen, and the normal rabbit IgG was also obtained from Santa Cruz Biotechnology.

Generation of Enz-resistant cell lines

EnzR1 was generated by treating C4-2 cells with continuously increasing Enz concentrations from 10 µM to 30 µM for over 1 year. For EnzR2, C4-2 cells were incubated with 10 µM Enz for 6 months. After generation, both EnzR1 and EnzR2 were maintained in media supplemented with 10 µM Enz. CWR22Rv1 cells are naturally resistant to Enz and were referred to as EnzR3 [26]. For EnzR4, C4-2B cells were incubated with 5 µM Enz at 40 µM for 1 year and maintained in medium supplemented with 20 μM Enz [27].

In vitro cell culture/maintenance

C4-2, C4-2B and CWR22Rv1 cell lines were obtained from American Type Culture Collection (ATCC) and maintained in RPMI 1640 media supplemented with 10% fetal bovine serum (FBS), antibiotics (100 units/ml penicillin, 100 µg/ml streptomycin), and 2 mM glutamine (Invitrogen) in a humidified 5% CO₂ environment at 37 °C. All the cell lines were confirmed to be mycoplasma-free or bacteria-free following the instructions of the ATCC three months prior to the experiments.

cDNA library construction and sequencing

EnzR1 cells and their parental Enz-sensitive C4-2 cells were processed with the AllPrep DNA/RNA Mini Kit (Qiagen) to extract RNA. For cDNA library construction and sequencing, mRNA was first isolated from total RNA treated with DNase I using magnetic oligo (dT) beads and then fragmented. Then, the double-stranded cDNA was synthesized with random hexamer primers and further subjected to end-repair and adapter ligation using T4 DNA ligase. The products of the ligation reaction were purified on a 2% agarose gel, and cDNA fragments (~200 bp) were recovered. PCR was carried out to enrich the purified cDNA template. Finally, the cDNA library was constructed. After validation on a Quit and Bioanalyzer, the library was sequenced using an Illumina HiSeq 2500 according to the manufacturer’s instructions.

Gene expression quantification

To quantify gene expression, the sequencing data were cleaned with fastp55 (v.0.23.2) and aligned to the GRCh38 reference using STAR56 (v.2.7.10a). Read counts were generated by subread57 (v.2.0.1) with default parameters, and the transcripts per million (TPM) values of genes were quantified using RSEM58 (v.1.3.1). Gene annotation was performed using Gencode release 33.

GEO data

Prostate cancer patient gene expression data and corresponding clinical information were obtained from the GEO database (https://www.ncbi.nlm.nih.gov/geo/). There were 94 PCa samples from GSE70769 and 76 PCa samples from GSE74367. The GSE70769 dataset included 92 patients whose survival information was available. The GSE74367 dataset consists of data from 11 primary prostate cancer patients, 20 CRPC patients, and 45 metastatic castration-resistant prostate cancer (MCRPC) patients. In this study, we used the primary PCa and CRPC data for subsequent analysis.

circRNA data

The CIRCMedia v2 database (http://yang-laboratory.com/circpedia/) includes circRNA annotations from more than 180 RNA-seq datasets spanning six different species: humans, rats, mice, fruit flies, zebrafish, and nematodes. The CSCD database (http://gb.whu.edu.cn/CSCD/) contains tumor-specific circRNAs identified through bioinformatics analysis of circRNA data from 87 tumor samples, with a focus on those expressed exclusively in cancer patients. By comparing the circRNAs from both databases, the shared circRNAs were identified as human circRNAs.

Identification of differential circRNAs

Differential expression analysis was performed on EnzR1 cells and their parental Enz-sensitive C4-2 cells using the limma package in R. To identify resistance-related differentially expressed genes, thresholds of P < 0.001 and |log2FC|> 5 were set. Human circRNAs were then mapped to resistance-related differentially expressed genes to identify circRNAs associated with resistance.

Lentiviral expression plasmid construction and virus production

According to Addgene’s protocol, 20 µg of the lentiviral package (the PLKO-shAR, PWPI-AR, PLVTHM-shcircCDR1-AS, PWPI-circCDR1-AS, PLVTHM-miR-1290, and PLVTHM-shBMP4), 10 µg of the PAX2 packaging plasmid, and 10 µg of the PMD2G envelope plasmid were transfected into HEK293T cells using the standard calcium chloride transfection method for 48 or 72 h to obtain the lentiviruses. The lentiviruses were collected and concentrated for immediate use or frozen at −80 °C for later use.

CircCDR1-AS overexpression plasmid construction

PWPI-circCDR1-AS was constructed via PCR amplification of the circRNA locus, including 1 kb upstream and 200 bp downstream of the nonlinear splice sites, using circCDR1-AS forward/reverse primers. The PCR fragment was inserted into HindIII- and NotI-digested PWPI (Invitrogen). An 800 bp DNA stretch upstream of the splice acceptor was amplified using circCDR1-AS forward/reverse primers and inserted downstream in the reverse orientation in an XhoI-digested PWPI-circCDR1-AS-ir, thus generating PWPI-circCDR1-AS. For the generation of PWPI-circCDR1-AS-fs, PCR amplification of the circCDR1-AS exon using circCDR1-AS-fs forward/reverse primers was inserted into PWPI using HindIII and XhoI. The lentiviruses were collected and concentrated for immediate use or frozen at −80 °C for later use.

RNA extraction and quantitative real-time PCR analysis

Total RNAs were isolated using TRIzol reagent (Invitrogen) following the manufacturer’s instructions, and then, 2 µg of RNA was reverse transcribed using Superscript III transcriptase (Invitrogen). Quantitative real-time PCR (qRT‒PCR) was performed using a Bio-Rad CFX96 system and SYBR Green Master Mix to determine the mRNA expression level of a gene of interest. The qRT‒PCR protocol was as follows: 50 °C for 2 min, 95 °C for 8 min 30 s, followed by 45 cycles at 95 °C for 15 s, and 60 °C for 1 min. The extension was 95 °C for 1 min, 55 °C for 1 min, and 55 °C for 10 s. GAPDH was used as the control for normalization.

The miRNAs were extracted using a PureLink® miRNA kit. Briefly, 2 µg of RNA was used for poly(A) polymerase at 37 °C for 20 min, and then, we conducted reverse transcriptase by annealing at 65 °C for 5 min and at 4 °C for 2 min after adding 50 µm RT anchor primer. The last step involved cDNA synthesis at 42 °C for 60 min, after which 2 μl of 10 mM dNTPs, 2 μl of 5x RT buffer, 1 μl of reverse transcriptase and ddH₂O were added to a total volume of 20 μl. The qRT‒PCR protocol was as follows: 95° for 2 min, followed by 45 cycles at 95 °C for 15 s and 60 °C for 45 s. U6 and/or 5S were used as controls for normalization.

Western blot analysis

Cells were lysed in lysis buffer on ice, and the proteins (50 µg) were separated on 10–12% SDS–PAGE gels and then transferred onto PVDF membranes (Millipore, Billerica, MA). The membranes were blocked with 5% bovine serum albumin (Sigma‒Aldrich, St. Louis, MO) for 1 h at room temperature and then diluted with primary antibodies overnight at 4 °C. The following primary antibodies were used: anti-GAPDH (1:1000; Santa Cruz Biotechnology), anti-AR (1:1000; Santa Cruz Biotechnology), and anti-BMP4 (1:1000; GeneTex Biotechnology). The next day, anti-mouse or anti-rabbit IgG secondary antibodies were used for 1 h at a concentration of 1:5000 at room temperature, after which the samples were rinsed 3 times with TBST for 10 min each. The bands were visualized using an enhanced chemiluminescence (ECL) detection system (Thermo Fisher Scientific, Rochester, NY).

Sphere formation assay

The sphere formation assay was performed as described earlier [28, 29]. Briefly, 1 × 10³ cells were seeded in 6-well ultra-low cluster plates (Corning Inc., Corning, NY) and cultured under non-adherent conditions on defined media (RPMI-1640 with B-27(1:50, Gibco), 20 ng/ml EGF, 20 ng/ml bFGF and 4 µg/ml insulin) for up to 7 days, passaged 2-5 times for experimental use, and revived every 2-3 months. Mix the cell suspension (1 × 10³, in 50 µl culture medium) with 50 µl Matrigel (BD) and place three experiments along the edge of a 24 well plate. Place the plate in a 37 °C incubator for 10 minutes to solidify the mixture, then add 500 µl of culture medium. After 7–14 days, count the number of balls under an optical microscope and observe the differences in ball size.

Chromatin immunoprecipitation (ChIP) assay

Cells were crosslinked with 4% formaldehyde for 10 min, followed by cell collection and sonication with predetermined power to yield 300–1000 bp long genomic DNA fragments. Lysates were precleared sequentially with normal rabbit IgG (sc-2027, Santa Cruz Biotechnology) and protein A-agarose. Then, 2 µg of anti-AR antibody was added to the cell lysates, which were incubated at 4 °C overnight. For the negative control, IgG was used in the reaction. Specific primer sets were designed to amplify a target sequence within the human CDR1 gene promoter. The PCR products were analyzed by agarose gel electrophoresis.

Luciferase reporter assays

The human promoter region of the CDR1 5’ promoter was inserted into the pGL3-basic vector (Promega, Madison, WI). Site-directed mutagenesis of the AR binding site in the CDR1 5’ promoter was achieved via quick change mutagenesis. Then, 1100 bp fragments of the BMP4 3’ UTR containing wild-type or mutant miRNA-responsive elements were cloned and inserted into the psiCHECK2 vector construct (Promega) downstream of the Renilla luciferase ORF. C4-2 and Enz-R1 cells were plated in 24-well plates, and the cDNA was transfected with Lipofectamine 3000 transfection reagent (Invitrogen) according to the manufacturer’s instructions. PRL-TK was used as an internal control that served as the baseline control response. Luciferase activity was measured 36–48 h after transfection via a dual-luciferase assay (Promega) according to the manufacturer’s manual.

RNA pull-down assay

The cells were collected with 1 ml of cell lysis buffer after receiving the designated treatments for 72 h. qRT‒PCR was used to test the concentration of GAPDH to ensure that the input of each group was equal for the following steps. The quantitated cells were rotated overnight at 4 °C after adding 1.5 µl of RNase inhibitor and 500 pM antisense oligos. The cells were rotated for 2 h at 4 °C after the addition of 10 µl of streptavidin agarose beads. The mixture was subsequently centrifuged at 3000 rpm for 2 min, after which the beads were washed with cell lysis buffer 5 times. Total RNA was extracted with TRIzol (Invitrogen) according to the manufacturer’s protocol and subjected to qRT‒PCR analysis.

Cell invasion assay

Cells were seeded in 6-well plates and incubated for 72 h after the designated treatments. The upper chambers of 8 µm pore size polycarbonate membrane inserts (Corning Incorporated, Corning NY) were coated with diluted Matrigel (1:20) (BD Corning) 2–4 h before the cells were plated. Then, the cells were collected with serum-free media and plated into the upper chambers at 1 × 10⁵/ml, and 750 µl of 10% FBS media was added to the lower chambers for incubation at 37 °C in a 5% (v/v) CO₂ incubator for 6‒8 h. The noninvading cells and Matrigel in the upper chambers were removed gently with cotton swabs, and the cells that migrated to the bottom of the membranes were permeabilized with methanol and stained with 0.1% (w/v) crystal violet. The invading cells were counted in five randomly chosen microscopic fields (100×) in each experiment and averaged for quantification. Each sample was run in triplicate, and multiple experiments were performed.

In vivo mouse studies

Male BALB/c mice that were 6–8 weeks old were purchased from Beijing Vital River Laboratory Animal Technology Co. Ltd., China, and divided into 4 groups (8 mice/group) for the injection of EnzR3 cells transduced with luciferase, shcircCDR1-AS, oemiR-1290 or shBMP4. The cells (1 × 10⁶) infected with different lentiviral constructs were mixed with Matrigel (1:1) and orthotopically implanted into the APs of nude mice. After 2 weeks of tumor growth, all the mice were i.p. injected with Enz at 30 mg/kg every other day. Tumor development was monitored by a noninvasive fluorescence imager (IVIS Spectrum, Caliper Life Sciences, Hopkinton, MA) once a week. Six to seven weeks after implantation, tumor growth was monitored again via IVIS analysis, the mice were sacrificed, and the tumors were removed for studies.

Immunohistochemical (IHC) staining

Tissues were fixed in 10% (v/v) formaldehyde in PBS, embedded in paraffin, cut into 5-μm-thick sections and subjected to IHC staining with specific primary antibodies against AR (Abcam, 14C8), BMP4 (GeneTex, N1N2) or CD133 (GeneTex, C1C2). To enhance antigen exposure, the slides were treated with 1 × EDTA at 98 °C for 10 min for antigen retrieval. The slides were incubated with endogenous peroxidase blocking solution and then incubated with the primary antibody at 4 °C overnight. After being rinsed with Tris-buffered saline, the slides were incubated for 45 min with biotin-conjugated secondary antibody, washed, and then incubated with enzyme-conjugated horseradish peroxidase (HRP)-streptavidin. Freshly prepared DAB (Zymed, South San Francisco, CA) was used as a substrate to detect HRP. Finally, the slides were counterstained with hematoxylin and mounted with aqueous mounting media. The number of positive cells was calculated as the number of immunopositive cells × 100% divided by the total number of cells/field in 10 random fields at 400× magnification.

Statistics

All the experiments were performed in triplicate. The data are expressed as the Mean ± SEM from at least 3 independent experiments. All the statistical analyses were performed with GraphPad Prism software 10.1.2 and R software (version 4.4.0; https://www.R-project.org). For comparisons between two groups, the Wilcoxon rank-sum test was used. Differential expression analysis of circRNAs was performed using the limma package in R, with a significance threshold of P < 0.001 and |log2FC|>5. Survival analysis was conducted using the Kaplan‒Meier method, and significance was assessed with the log-rank test. A P value < 0.05 was considered statistically significant.

Results

Enz increases circCDR1-AS expression to alter Enz sensitivity

We first generated Enz-resistant (EnzR) CRPC cells via chronic culture of C4-2 cells in media with increasing Enz concentrations from 10 to 30 µM over 1 year (named EnzR1) and then cultured the CRPC C4-2 cells in media supplemented with 10 µM Enz for 6 months (named EnzR2). CWR22Rv1 cells (named EnzR3) are naturally resistant to Enz, and CRPC C4-2B cells were cultured in media with Enz concentrations ranging from 5 to 40 µM for 1 year (named EnzR4) (Supplementary Fig. S1A–D).

We then applied the RNA-seq approach to compare these EnzR1 cells with their parental Enz-sensitive (EnzS) C4-2 cells (treated with DMSO for 1 year) to identify the differentially expressed circRNAs, with thresholds of P < 0.001 and |log2FC|>5. The results revealed that the expression of CDR1 antisense RNA (CDR1-AS or circCDR1-AS), which is a widely studied circRNA in tumors, was significantly increased (Fig. 1A). Consistent results from qRT‒PCR also revealed that circCDR1-AS expression was significantly increased in 3 pairs of EnzR cell lines compared with their parental Enz-sensitive lines (Fig. 1B).

Fig. 1. Enz treatment increases circCDR1-AS expression, which is linked to the development of Enz resistance.

A Volcano plots of the most significantly changed circRNAs in EnzR cells. The differential expression of the EnzR and parental cells was examined with RNA-seq analysis, and those with P < 0.001 and |log2FC|>5 were significantly up- or downregulated. B qRT‒PCR analysis of circCDR1-AS mRNA levels in C4-2 vs. EnzR1 cells (upper panel), C4-2 vs. EnzR2 cells (middle panel), and C4-2B vs. EnzR4 cells (lower panel). C circCDR1-AS was knocked down in EnzR1 cells. EnzR1-PLKO and EnzR1-shcircCDR1-AS cells were treated with Enz, and cell viability was analyzed via the MTT assay. The left panel shows the results of qRT‒PCR analysis of circCDR1-AS levels in labeled cells, and the right panel shows the results of the MTT assay. D C4-2-PWPI and C4-2-oecircCDR1-AS cells were treated with or without Enz, and then, the cell viability was analyzed via an MTT assay. The left panel shows the results of qRT‒PCR analysis of circCDR1-AS levels in C4-2-PWPI and C4-2-oecircCDR1-AS cells, and the right panel shows the results of the MTT assay. E circCDR1-AS was knocked down in EnzR4 cells. EnzR4-PLKO and EnzR4-shcircCDR1-AS cells were treated with Enz, and cell viability was analyzed via the MTT assay. The left panel shows the results of qRT‒PCR analysis of circCDR1-AS levels in labeled cells, and the right panel shows the results of the MTT assay. F C4-2B-PWPI and C4-2B-oecircCDR1-AS cells were treated with or without Enz, and then, the cell viability was analyzed via an MTT assay. The left panel shows the results of qRT‒PCR analysis of circCDR1-AS levels in C4-2B-PWPI and C4-2B-oecircCDR1-AS cells, and the right panel shows the results of the MTT assay. The data are presented as the Mean ± SEM, **P < 0.01, ***P < 0.001,****P < 0.0001, ns not significant.

To further confirm the potential impacts of circCDR1-AS on Enz resistance, we knocked down circCDR1-AS in EnzR cells, and the results of the MTT growth assay revealed that suppressing circCDR1-AS led to significantly increased Enz sensitivity, further suppressing EnzR1,EnzR4 cells proliferation (Fig. 1C, E). The addition of circCDR1-AS resulted in decreased Enz sensitivity in C4-2,C4-2B cells (Fig. 1D, F).

Together, the results shown in Fig. 1A–F from multiple EnzR CRPC cell lines demonstrated that a high level of circCDR1-AS in EnzR cells contributes to Enz resistance and that lowering circCDR1-AS expression could increase Enz sensitivity to further suppress EnzR cell proliferation.

Enz/AR signaling can alter circCDR1-AS expression via transcriptional regulation of its host gene CDR1

To dissect the mechanisms by which Enz can alter circCDR1-AS expression via AR signaling, we focused on CDR1, the host gene of circCDR1-AS [30], since the expression of many circRNAs is controlled by their host gene transcription and splicing [31]. The qRT‒PCR results revealed that suppressing AR signaling led to increased CDR1 and circCDR1-AS expression in C4-2 cells and C4-2B cells (Fig. 2A).

Fig. 2. Mechanistic analysis of how Enz treatment regulates circCDR1-AS expression.

A qRT‒PCR analysis of CDR1 and circCDR1-AS mRNA levels after AR was knocked down in C4-2 (left) and C4-2B (right) PLKO and shAR cells. B Schematic depiction of the cloning of the 2.5 kb CDR1 promoter into the pGL3 basic luciferase reporter vector (pGL3). Site-directed mutagenesis of the ARE by changing part of the ARE through to the BamH1 cutting site (-GGATCC-). C ChIP assay to identify AR binding to the CDR1 promoter region. The PCR products were detected by agarose gel electrophoresis, and the results revealed that AR could bind to the potential ARE on the CDR1 promoter in C4-2 and EnzR1 cells. D Cotransfection of wild-type or mutant circCDR1-AS promoter pGL3-luciferase constructs with/without shAR or oeAR into C4-2 cells. The reporter activity was analyzed by a luciferase assay. The data are presented as the Mean ± SEM, *P < 0.05, **P < 0.01, ***P < 0.001,****P < 0.0001, ns not significant.

To further dissect the molecular mechanism at the transcriptional level, we first searched for potential androgen response elements (AREs) and found one potential ARE in the 2.5 kb 5’ promoter region of CDR1 (Fig. 2B). The results from the chromatin immunoprecipitation (ChIP) in vitro binding assay revealed that AR could bind to this potential ARE on the CDR1 promoter region in C4-2 cells. However, in EnzR1 cells, such binding was attenuated significantly, suggesting that in EnzR cells, AR loses its ability to bind to the CDR1 promoter (Fig. 2C).

The results from the luciferase reporter assay also confirmed that downregulating AR via the addition of AR-shRNA led to increased luciferase reporter activity in the wild-type cells but not the mutant cells. In contrast, upregulating AR via the addition of AR-cDNA (oeAR) led to a decrease in luciferase reporter activity in wild-type but not mutant C4-2 cells (Fig. 2D).

Together, the results shown in Fig. 2A–D suggest that Enz/AR signaling can transcriptionally regulate CDR1 gene expression by binding to the ARE on the 5’ promoter of the CDR1 gene to modulate circCDR1-AS expression.

Enz/AR/CDR1/circCDR1-AS signaling can alter Enz resistance by altering CSC traits

Next, to study the mechanisms by which Enz/AR/CDR1/circCDR1-AS signaling can alter Enz resistance, we tested and found that adding circCDR1-AS led to increased expression of several biomarkers of CSCs in C4-2 cells (Fig. 3A, B), suggesting that circCDR1-AS might function by altering CSC traits to impact Enz sensitivity. This finding is in agreement with recent reports showing that altered CSCs might be linked to inherent resistance to therapy-induced cytotoxicity [32–34]. Furthermore, a recent study also indicated that PCa CSC alterations might play key roles in resistance acquisition after ADT [15, 16].

Fig. 3. Enz/AR-mediated circCDR1-AS upregulation can alter Enz resistance via the modulation of CSC traits.

A The mRNA levels of CSC markers were analyzed by qRT‒PCR in C4-2 PWPI and oecircCDR1-AS cells. B The protein levels of CSC markers were analyzed by Western blot analysis in C4-2 PWPI and oecircCDR1-AS cells. C Sphere formation assays were performed to evaluate the CSC population in C4-2/EnzR1 cells and C4-2B/EnzR4 cells. The quantitative data are on the right. D qRT‒PCR analysis to examine the expression of different CSC markers in C4-2 and EnzR1 cells. E Western blot analysis was used to detect the expression of different CSC markers in C4-2 and EnzR1 cells. F circCDR1-AS was knocked down in EnzR1 and EnzR4 cells, and a sphere formation assay was performed to evaluate the CSC population. The quantitative data are on the right. G The mRNA levels of CSC markers were analyzed via qRT‒PCR in EnzR1 PLKO and shcircCDR1-AS cells. H Western blot analysis of the protein levels of CSC markers in EnzR1 PLKO and shcircCDR1-AS cells. I circCDR1-AS was overexpressed in C4-2 and C4-2B cells, and a sphere formation assay was performed to evaluate the CSC population. The quantitative data are on the right. The data are presented as the Mean ± SEM, *P < 0.05, **P < 0.01, ***P < 0.001,****P < 0.0001.

We then applied a sphere formation assay, another standard assay used to monitor CSCs [35], to confirm the biomarker expression results, and the results revealed that the number of CSCs in EnzR1 cells was greater than that in C4-2 cells; similar results were also obtained when we replaced EnzR1/C4-2 cells with EnzR4/C4-2B cells (Fig. 3C). The addition of circCDR1-AS also resulted in increased sphere-forming capacity of EnzR1 and EnzR4 cells (Supplementary Fig. S1E, F). Importantly, results from measuring the expression of other key CSC biomarkers, including CD44, CD133, NANOG, SOX2, OCT4, BMP4, and integrin β1, also revealed greater expression of these CSC biomarkers in EnzR1 cells than in C4-2 cells (Fig. 3D, E).

Finally, suppressing circCDR1-AS also led to a decrease in the sphere-forming ability of both EnzR1 and EnzR4 cells (Fig. 3F) and the expression of CSC biomarkers in EnzR1 cells (Fig. 3G, H), and adding circCDR1-AS to C4-2 and C4-2B cells also resulted in increased sphere formation (Fig. 3I).

Together, the results shown in Fig. 3A–I suggest that Enz/AR/CDR1/circCDR1-AS signaling may alter CSC traits to increase Enz resistance in CRPC cells.

Enz/AR/CDR1/circCDR1-AS signaling alters CSC traits by altering miR-1290 expression

Next, to dissect the mechanism by which Enz/AR/CDR1/circCDR1-AS signaling can alter CSC traits, we focused on miRNAs since circRNAs may function by regulating selected miRNAs to alter their target genes [36]. Using two public bioinformatic prediction databases (CircNet [37] and StarBase v2.0 [38]), we found that 5 potential miRNAs (miR-7-5p, miR-1290, miR-1299, miR-6782 and miR-876-5P) might interact with circCDR1-AS in PCa cells (Fig. 4A). We then compared the expression of these miRNAs in EnzR1 and C4-2 cells and found that miR-1290 expression was lower in EnzR1 cells than in C4-2 cells, but miR-1299 and miR-6782 expression was increased (Fig. 4B). In contrast, suppressing circCDR1-AS led to increased expression of miR-1290 in EnzR1 cells and decreased expression of miR-6782 (Fig. 4C).

Fig. 4. The AR/CDR1/circCDR1-AS axis promotes CSC traits by altering miR-1290 expression.

A Five potential miRNAs (miR-7-5p, miR-1290, miR-1299, miR-6782, and miR-876-5P) might interact with circCDR1-AS. The expression levels of miRNAs (miR-7-5p, miR-1290, miR-1299, miR-6782 and miR-876-5P) were analyzed via qRT‒PCR in (B) C4-2 and EnzR1 cells and in (C) EnzR1 PLKO and shcircCDR1-AS cells. D RNA pull-down assays were performed to test whether miR-1290, miR-1299, or miR-6872 can specifically bind to the circCDR1-AS or ASS1 probe in EnzR1 cells. E The expression levels of miR-1290 were analyzed by qRT‒PCR in EnzR1 and EnzR4 cells overexpressing miR-1290. A sphere formation assay was performed to evaluate the CSC population via the overexpression of miR-1290 in EnzR1 (F) and EnzR4 (G) cells. H Sphere formation assay to evaluate the CSC population in C4-2, C4-2B cells in which the expression of circCDR1-AS and miR-1290 was manipulated. I The protein and mRNA levels of CD133 were analyzed by Western blot analysis and qRT‒PCR in C4-2 cells after overexpressing circCDR1-AS and miR-1290. For (F, G), the quantitative data are on the right. The data are presented as the Mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001,****P < 0.0001, ns not significant.

The results from the RNA pull-down assay with a biotinylated oligo (ASS1) probe (5’-GGATAATGCACATGTGTATATCT-3’) to target the circular junction of circCDR1-AS also revealed that miR-1290, but not the miRNAs (miR-1299 and miR-6782), was enriched in the pull-down product, suggesting that miR-1290 can bind directly to circCDR1-AS (Fig. 4D).

Furthermore, the addition of miR-1290 (Fig. 4E) decreased the number of CSCs in EnzR1, EnzR4 (Fig. 4F, G), whereas the addition of miR-1299 and miR-6782 to EnzR1,EnzR3 and EnzR4 cells had little effect on the number of CSCs, suggesting that only miR-1290, but not other miRNAs, can regulate the CSC population in EnzR cells (Supplementary Fig. S1G–I).

Importantly, the results from interruption approaches also indicated that adding miR-1290 could abrogate the changes in CSC traits induced by circCDR1-AS (Fig. 4H) as well as downregulate CSC biomarker (CD133) expression (Fig. 4I) in C4-2 cells.

Together, the results shown in Fig. 4A–I suggest that Enz/AR/CDR1/circCDR1-AS signaling may function by binding to miR-1290 and altering its expression to regulate CSC traits.

Enz/AR/CDR1/circCDR1-AS/miR-1290 signaling can promote CSC traits by altering BMP4 expression

To dissect the mechanism by which Enz/AR/CDR1/circCDR1-AS/miR-1290 signaling can promote CSC traits, we focused on relevant CSC-related genes in bioinformatics databases (circBase and CircNet) that may have direct regulatory effects on circCDR1-AS or miR-1290 and identified 3 CSC-related genes, BMP4, SOX2, and FOXC1, for further study (Fig. 5A). The qRT‒PCR results revealed that the addition of miR-1290 significantly decreased the expression of BMP4 but not the other two CSC-related genes at the mRNA level in EnzR1 cells (Fig. 5B). We then further studied the relationship between BMP4 and circCDR1-AS and between BMP4 and miR-1290.

Fig. 5. The AR/CDR1/circCDR1-AS/miR-1290 axis regulates CSC traits by altering BMP4 signaling in PCa.

A Diagram showing that BMP4, FOXC1 and SOX2 may be regulated by circCDR1-AS and miR-1290 in EnzR cells. B qRT‒PCR assay for screening CSC-related genes in EnzR1 cells with oemiR-1290 compared with the vector PLVTHM. C The protein level of BMP4 in EnzR1 PLVTHM and oemiR-1290 cells was analyzed via Western blot. The mRNA (right) and protein (left) levels of BMP4 were analyzed via qRT‒PCR and Western blot, respectively, in (D) C4-2 and EnzR1 cells, in (E) C4-2 PLKO and shAR cells, and in (F) EnzR1 PLKO and shcircCDR1-AS cells. The mRNA and protein levels of BMP4 were analyzed via qRT‒PCR and Western blot in EnzR1 (G) cells transfected with circCDR1-AS shRNA or treated with a miR-1290 inhibitor or in C4-2 (H) cells transfected with oecircCDR1-AS or oemiR-1290. I The protein levels of BMP4 in EnzR1 and EnzR4 cells were analyzed via Western blot after transfection with shBMP4. J C4-2 cells were transfected with oecircCDR1-AS and shBMP4, and then, the sphere formation ability of the cells was tested, as shown in the lower panel. K Predicted sequence alignment of the BMP4 3’ UTR with the wild-type (WT) or mutant potential miR-1290 target site. Luciferase reporter activity after the transfection of wild-type or mutant BMP4 3’ UTR reporter constructs in EnzR1 (L) and C4-2 (M) cells compared with that in control cells. The data are presented as the Mean ± SEM; *P < 0.05, **P < 0.01, ***P < 0.001,****P < 0.0001, ns not significant.

As expected, the addition of miR-1290 decreased the expression of BMP4 at the protein level in EnzR1 cells (Fig. 5C), and higher expression of BMP4 was detected in EnzR1 cells than in C4-2 cells (Fig. 5D). Suppressing AR could then increase the expression of BMP4 in C4-2 cells (Fig. 5E), and suppressing circCDR1-AS led to decreased BMP4 expression in EnzR1 cells (Fig. 5F).

Importantly, suppressing miR-1290 via treatment with the miR-1290 inhibitor led to increased BMP4 expression and partly abrogated the circCDR1-AS-shRNA-mediated suppression of BMP4 expression in EnzR1 cells (Fig. 5G). In contrast, increasing miR-1290 via the addition of the miR-1290 mimic led to decreased BMP4 expression and partly abrogated the circCDR1-AS-induced increase in BMP4 expression in C4-2 cells (Fig. 5H).

To further confirm the suppression of BMP4 specificity and effect, we also constructed a BMP4-shRNA and validated the effects of BMP4 knockdown in EnzR1 and EnzR4 cells (Fig. 5I). Finally, the results of the sphere formation assays revealed that suppressing BMP4 with shBMP4 also partially abrogated the circCDR1-AS-induced increase in CSC traits in C4-2 cells (Fig. 5J).

Together, the results shown in Fig. 5A–J suggest that Enz/AR/CDR1/circCDR1-AS/miR-1290 signaling can promote CSC traits by altering BMP4 expression.

Enz/AR/CDR1/circCDR1-AS/miR-1290 signaling can increase BMP4 expression via direct binding to the 3’ UTR of BMP4 mRNA

Finally, to dissect the mechanism by which miR-1290 can alter BMP4 expression, we first searched for and identified potential binding sites located on the 3’ UTR of BMP4 mRNA (http://www.targetscan.org). We then constructed psiCHECK2 vectors carrying the wild-type or mutant BMP4 miRNA target sites for the luciferase reporter assay (Fig. 5K), and the results revealed that miR-1290 could decrease the luciferase reporter activity of the wild-type 3’ UTR of BMP4 but had little effect on the mutant 3’ UTR of BMP4 mRNA in EnzR1 cells (Fig. 5L). As expected, treatment with the miR-1290 inhibitor also increased the luciferase reporter activity of the wild-type 3’ UTR of BMP4 but not the mutant 3’ UTR of BMP4 in C4-2 cells (Fig. 5M).We then test the results above by qRT‒PCR and Western blot analysis. The results demonstrate that miR-1290 could decrease relative BMP4 mRNA expression of wild-type of BMP4 but had little effect on the mutant of BMP4 mRNA in EnzR1 cells. Also, treatment with the miR-1290 inhibitor also increased relative BMP4 mRNA expression of the wild-type of BMP4 but not the mutant of BMP4 in C4-2 cells (Supplementary Fig. S2A, B). Western blot analysis shows same results (Supplementary Fig. S2C, D).

Together, the results shown in Fig. 5K–M suggest that Enz/AR/CDR1/circCDR1-AS/miR-1290 signaling can alter BMP4 expression via direct binding to the 3’ UTR of BMP4 mRNA.

A preclinical study using a mouse model and integrative analysis of GEO cohort data confirmed that Enz may promote CSC traits by altering Enz/AR/CDR1/circCDR1-AS/miR-1290/BMP4 signaling

To further confirm the above in vitro data from the in vivo mouse model, we established orthotopic xenografts of EnzR3-luc cells expressing firefly luciferase. We then transfected cells with circCDR1-AS-shRNA, oemiR-1290, or BMP4-shRNA at 8 mice/group (1: EnzR3-luc control; 2: EnzR3-luc+circCDR1-AS-shRNA; 3: EnzR3-luc+oemiR-1290; and 4: EnzR3-luc+BMP4-shRNA). After 2 weeks of tumor growth, the mice were injected with Enz at 30 mg/kg every other day, and an in vivo imaging system (IVIS) was used for 6‒7 weeks to monitor tumor growth.

As shown in Fig. 6A, C, targeting this newly identified AR/CDR1/circCDR1-AS/miR-1290/BMP4 signaling pathway (circCDR1-AS-shRNA, oemiR-1290, and BMP4-shRNA) significantly suppressed CRPC growth, as did the expression of the CSC biomarkers CD133 and BMP4, which is consistent with our results from multiple in vitro cell lines showing that circCDR1-AS and miR-1290 can regulate BMP4 expression to alter CSC traits (Fig. 6B, D, E).

Fig. 6. A preclinical study using a mouse model and integrative analysis of GEO cohort data confirmed that Enz may promote CSC traits by altering Enz/AR/CDR1/circCDR1-AS/miR-1290/BMP4 signaling.

A IVIS imaging was used to detect the tumor volume in mice bearing orthotopic PCa xenografts. Two representative mouse IVIS bioluminescent images for each group are shown. B Representative images and quantitative analysis of the IHC results for BMP4 and CD133 in each group. C Quantification of IVIS imaging in the four groups. D, E Quantification of the IHC results for BMP4 and CD133 in each group. F Overall survival based on GEO database indicated that higher levels of BMP4 were associated with a worse survival. G The expression of BMP4 in primary PCa and CRPC. The data are presented as the Mean ± SEM; **P < 0.01, ***P < 0.001,****P < 0.0001.

Finally, to substantiate all the above in vitro/in vivo results using human clinical data, we obtained a dataset (GSE70769) from the GEO (https://www.ncbi.nlm.nih.gov/geo/) database [39], which included the RNA expression profiles and detailed clinical information of 94 PCa patients and performed a survival analysis. The results revealed that PCa patients with higher BMP4 gene expression had significantly shorter overall survival (P = 0.0022, Fig. 6F), suggesting that BMP4 may play a positive role in promoting PCa progression. The data of 11 patients with primary PCa and 20 patients with CRPC were selected from GSE74367 [40] for analysis of the differences in BMP4 expression levels between primary PCa and CRPC patients. The results revealed that the expression of the BMP4 gene in CRPC patients was significantly greater than that in primary PCa patients (Fig. 6G) and that high expression of BMP4 may facilitate the transition from primary PCa to CRPC.

Together, these results (Fig. 6A–G) demonstrate that Enz/AR/CDR1/circCDR1-AS/miR-1290/BMP4 signaling may function by altering CSC traits to alter Enz resistance and that targeting this newly identified signaling pathway with small molecules may be an effective strategy for inhibiting EnzR CRPC cell growth.

Discussion

Recent studies have indicated that several mechanisms might contribute to the development of Enz resistance in CRPC. For example, AR splicing results in the generation of mutant ARv7 [41] or mutant ARF876L [7, 42, 43], altering glucocorticoid receptor signals [44], activating NFkB2/p52 [45], and activating the IL6/Stat3/AR axis [46]. However, in view of the complexity and diversity of the mechanism of CRPC resistance to Enz, this mechanism does not fully explain the phenomenon of drug resistance in all CRPC patients. Here, we identified another new mechanism showing that Enz could also function by modulating circRNA/miRNA signaling to alter CSC traits to influence Enz resistance (Fig. 7).

Fig. 7. Schematic model showing the AR/CDR1/circCDR1-AS/miR-1290/BMP4 axis regulating the resistance of prostate cancer to Enz.

The AREs on the 5’ promoter of CDR1 bind with AR to inhibit the expression of circCDR1-AS. Enz treatment can inhibit the entry of AR into the nucleus and reduce the interaction between AR and CDR1 to promote the expression of circCDR1-AS, thus inhibiting the expression of miR-1290 to increase BMP4 expression, which leads to the promotion of CSC traits and the development of Enz resistance.

CircRNAs are widely detected in various human cells, and their expression is much greater than that of their linear isomers [47]. CircRNAs have highly conserved sequences with better stability, which might allow them to become novel biomarkers for the diagnosis of various types of cancer [48, 49]. Functional studies have indicated that circRNAs might act as sponges of miRNAs to alter their function [48–50]. For example, circCDR1-AS can sponge miR-7 in embryonic zebrafish [51]. In our study, we found that Enz can increase circCDR1-AS expression by inhibiting the binding of AR to the promoter of its host gene CDR1 and that circCDR1-AS can also serve as a sponge of miRNAs to reduce and inhibit miRNA function.

Importantly, early studies also indicated that circCDR1-AS could function as a post-transcriptional regulator by binding to miR-7 [51], that circCDR1-AS could also compete with endogenous miR-7 [52] to alter breast CSCs [53], and that the restoration of miR-7 effectively inhibited CSC stemness [54]. Other studies also indicated that adding miR-34a to PCa cells suppressed prostate CSC traits by inhibiting CD44 expression [55], and adding miR-200 suppressed prostate CSC stemness by indirectly inhibiting ZEB1/2 and SNAIL2 expression [56].

In this study, we found that the expression of circCDR1-AS in EnzR cells was significantly greater than that in EnzS C4-2 cells. Additionally, suppressing circCDR1-AS led to increased expression of miR-1290 in EnzR1 cells. Moreover, circCDR1-AS markedly increased sphere-forming efficiency and notably elevated the expression of the stemness markers CD133 and BMP4 in prostate cancer cells. BMP4 belongs to the TGF-β superfamily and may play important roles in tumor cell proliferation, angiogenesis, and metastasis [57, 58]. Interestingly, recent studies have also indicated that BMP4 could play pivotal roles in promoting cancer cell stemness [12, 59, 60]. Here, we confirmed that miR-1290 could directly target the 3’ UTR of BMP4 mRNA to inhibit CSC traits in PCa. In addition to its role in CSC regulation, circCDR1-AS may also contribute to Enz resistance by promoting tumor cell invasion. Interestingly, in EnzR1, EnzR4 cells, knocking down circCDR1-AS led to suppressed cell invasion, suggesting that circCDR1-AS could facilitate tumor cell invasion in EnzR cells. These findings align with those of previous studies indicating that circRNAs can regulate various aspects of cancer progression (Supplementary Fig. S2E).

In summary, our preclinical studies using multiple in vitro cell lines and an in vivo mouse model demonstrated that Enz could promote prostate CSC traits by altering Enz/AR/CDR1/circCDR1-AS/miR-1290/BMP4 signaling and that targeting this newly identified signaling pathway with small molecules may be an effective strategy for treating CRPC at the later stage, when Enz resistance is likely to occur.

The limitation of this study is that the patient cohort analyzed was derived from public databases rather than from newly collected clinical samples. While these publicly available datasets provide valuable insights, they have inherent limitations, particularly in the relatively low coverage of miRNA and circRNA data, which may affect the comprehensiveness of our analysis. Currently, we are actively collecting patient cases to address these gaps and strengthen our findings. However, due to the extended timeframe required for comprehensive case collection, these data have not yet been incorporated into the present study. Future investigations will focus on integrating newly acquired clinical samples with more complete miRNA and circRNA profiling to increase the robustness and clinical applicability of our conclusions.

Author contributions

KW, HY, and XC designed and supervised the study; DH, MS, KY, XF, HL, DL, ZC, JL, RX, XS, YD, XM, XW, and PL did the experiments; DH, MS, KY and XF analyzed the data; DH, MS, and KY wrote the paper; DH, MS, KY, XC, KW and HY scientifically revised the manuscript. All authors contributed to and approved the manuscript.

Data availability

The data that support the findings of this study are available from the corresponding author upon reasonable request.

Code availability

The computational analyses in this study were performed using R (version 4.4.0) with the limma package for differential expression analysis. All custom scripts used for data processing and analysis have been provided in the supplementary materials for full transparency and reproducibility.

Competing interests

The authors declare no competing interests.

Ethics approval

All animal care and use procedures were performed according to the Principles of Animal Care provided by the National Society for Medical Research and the Guide for the Care and Use of Laboratory Animals (Institute of Laboratory Animal Resources, NIH). All the animal experiments were approved by the ethics committee of Fourth Affiliated Hospital of Harbin Medical University(approval number: 2024-DWSYLLCZ-29). All methods were performed in accordance with the relevant guidelines and regulations.

Supplementary information

The online version contains supplementary material available at 10.1038/s41388-025-03482-1.