RNA-binding protein PSF functions as an epigenetic modifier by interacting with long noncoding RNAs and the corepressor complex. PSF also promotes RNA splicing events to enhance oncogenic signals. In this study, we conducted an in vitro chemical array screen and identified multiple small molecules that interact with PSF. Several molecules inhibited RNA binding by PSF and decreased prostate cancer cell viability. Among these molecules and its derivatives was a promising molecule, No. 10–3 [7,8-dihydroxy-4-(4-methoxyphenyl)chromen-2-one], that was the most effective at blocking PSF RNA-binding ability and suppressing treatment-resistant prostate and breast cancer cell proliferation. Exposure to No. 10–3 inhibited PSF target gene expression at the mRNA level. Treatment with No. 10–3 reversed epigenetically repressed PSF downstream targets, such as cell-cycle inhibitors, at the transcriptional level. Chromatin immunoprecipitation sequencing in prostate cancer cells revealed that No. 10–3 enhances histone acetylation to induce expression of apoptosis as well as cell-cycle inhibitors. Furthermore, No. 10–3 exhibited antitumor efficacy in a hormone therapy–resistant prostate cancer xenograft mouse model, suppressing treatment-resistant tumor growth. Taken together, this study highlights the feasibility of targeting PSF-mediated epigenetic and RNA-splicing activities for the treatment of aggressive cancers.
This study identifies small molecules that target PSF–RNA interactions and suppress hormone therapy–refractory cancer growth, suggesting the potential of targeting PSF-mediated gene regulation for cancer treatment.
RNA-binding proteins (RBP) primarily form ribonucleoprotein complexes in cells to control RNA processing via splicing, polyadenylation, modification, nuclear export, decay of RNA transcripts, and protein translation (1). RBPs play an important role in the development and physiologic function of organs (2). The expression and protein localization of RBPs are dysregulated in human diseases such as dementia and cancer (3–5). A representative RBP, polypyrimidine tract-binding protein (PTB)-associated splicing factor (PSF), or splicing factor, proline and glutamine-rich (SFPQ), is a component of the large spliceosomal complex, consisting of the DNA-binding domain, the RNA recognition motifs (RRM), the NonA/ParaSpeckle (NOPS) domain, and the coiled-coil domain (6–8). In the nucleus, PSF is implicated in both transcription and RNA processing (8). PSF is physiologically critical in neuron and brain development (9–11), regulating comprehensive gene expression by targeting microtubule-associated protein gene, TAU, amyloid precursor protein (APP) gene, and other neuron-associated genes at the RNA level (12, 13). PSF is known to be a component of nuclear speckles, which are composed of long noncoding RNA, nuclear paraspeckle assembly transcript 1 (NEAT1), or another RBP protein, non-POU domain containing octamer binding (NONO) (6, 14). Importantly, PSF suppresses its target gene transcription epigenetically by recruiting histone deacetylase (HDAC) enzymes to specific genomic-binding regions (15, 16).
The dysregulation of RBP expression and function via genomic mutation or transcriptional regulation facilitates cancer progression (17). We have previously shown that PSF coordinates the expression and complex formation of various spliceosome factors in prostate cancer (18). The dysregulation of the spliceosome complex enhances the spliceosome activity for androgen receptor (AR) and its variant production, which are the central signals in prostate cancer, to lead to a more aggressive, hormone therapy–refractory, castration-resistant prostate cancer (CRPC; refs. 19–22). Importantly, PSF also modifies the epigenetic status by binding to specific gene regulatory regions by interacting with AR-induced long noncoding RNA (lncRNA), C-terminal binding protein 1 antisense (CTBP1-AS) gene, and the corepressor complex, including HDAC (15). Moreover, aggressive types of tumors in breast cancer, which is developed due to estrogen receptor α (ERα)-mediated signals, show resistance to hormone therapy using an ERα antagonist, such as tamoxifen (23–26). In an ERα-positive breast cancer, PSF regulates mRNA expression of oncogenic signals, such as of ERα (ESR1), Sec1 family domain containing 2 (SCFD2), transformer-2 protein homolog beta (TRA2B), and abnormal spindle-like microcephaly-associated protein (ASPM) at the posttranscriptional level (27). Furthermore, PSF confers 4-hydroxytamoxifen (OHT) resistance to breast cancer tumor growth. It was also reported that in colon cancer cells, PSF physically associates with metastasis associated lung adenocarcinoma transcript 1 (MALAT1) to increase tumor proliferation (28). In ovarian cancer, PSF associates with a spliceosome component, serine and arginine rich splicing factor 2 (SRSF2), to confer chemoresistance (29). Overall, these findings revealed that PSF and its related pathways promote the progression of treatment-resistant cancers.
Small molecules that modulate the activity of RBPs have been investigated to analyze their pathophysiologic roles in cells. In addition, RBP inhibitors are assumed to be useful as therapeutic agents for a treatment-resistant cancer (17). In this study, we performed a high-throughput screening (HTS; refs. 30, 31) to identify candidate agents for PSF inhibitors. By investigating PSF-associating molecules using HTS of small-molecule libraries by using chemical arrays, we identified compounds that inhibit PSF activity to modulate its target genes. We also screened more structure-related compounds to enhance its activity. We performed in silico docking analysis, RNA pulldown in vitro, and cell growth assays. Then, we identified a compound, No. 10–3 (7,8-dihydroxy-4-(4-methoxyphenyl)chromen-2-one), that inhibited the PSF-mediated RNA splicing and the epigenetic pathway that promotes apoptosis and inhibits cell cycle in both breast and prostate cancer cells. In a xenograft mouse model of treatment-resistant prostate cancer, pharmacologic inhibition of PSF function impaired tumor growth in vivo. Thus, our findings highlight the feasibility of targeting PSF-mediated epigenetic and RNA splicing activities to overcome treatment resistance in prostate and breast cancers.
Materials and Methods
Chemicals and chemical array screening
The chemical library containing 30,707 compounds used for screening was provided by RIKEN Natural Product Depository (NPDepo). A total of 12 chemical arrays (NPDepo Array ver.2) was used for this screening assay. Our chemical assays spotted and immobilized small-molecule compounds on glass slides to discover ligands of mCherry-tagged proteins. mCherry protein, mCherry-tagged PSF-RNA binding domain (RBD) and ΔN-terminal PSF (ΔN) proteins were produced by overexpressing in 293T cells. Protein molecules binding to immobilized compounds are detected by evaluating fluorescence intensity. Details for Chemical screening were described in Supplementary Materials and Methods section and previously (31).
The cell lines used in the present study were obtained from ATCC. MCF7 and OHTR cells were provided from Saitama Medical University. Identities of the cells were confirmed by short tandem repeat (STR) analyses in 2019 (BEX co. Ltd.). All cell lines were grown at 37C in a 5% CO2 atmosphere. Also, we routinely checked for Mycoplasma contamination using a PCR-based kit, Mycoplasma Detection Kit (Jena Bioscience). Additional viral-infection was checked using PCR method (ICR monitoring center, Kanagawa, Japan). We maintained stocks of low-passage cells and restarted our cell culture with a fresh vial at least once a month. VCaP, 293T and TIG3 cells were cultured in DMEM supplemented with 10% FBS, 50 U/mL penicillin, and 50 μg/mL streptomycin. 22Rv1 and LNCaP cells were cultured in RPMI medium supplemented with 10% FBS, 50 U/mL penicillin, and 50 μg/mL streptomycin. Long-term androgen deprivation (LTAD) cells were established from LNCaP cells cultured in phenol-red free RPMI medium supplemented with 10% charcoal-dextran stripped FBS, 50 U/mL penicillin, and 50 μg/mL streptomycin. 4-hydroxytamoxifen (OHT)-resistant OHTR cells were established from MCF7 cells cultured in DMEM supplemented with 10% FBS, 1 μmol/L OHT, 50 U/mL penicillin, and 50 μg/mL streptomycin (27). OHT was purchased from Sigma. Dihydrotestosterone (DHT) and 17b-estradiol (E2) were purchased from Wako. Transfection of siRNAs targeting PSF (siPSF #1 and #2) and p53 (#1 and #2) were described before (15, 18).
RNA was isolated using the ISOGEN II (Nippon Gene) in accordance with manufacturer's instructions. Complementary DNA was synthesized from equivalent concentrations of total RNA using Prime Script (TAKARA bio) according to manufacturer's protocols. Amplification was performed in a StepOne PCR System (Thermo Fisher Scientific) using KAPA SYBR Green (Sigma). Fold changes for experimental groups relative to the loading control (GAPDH, MB) were calculated by ΔΔCt method. Primer sequences were described before (15, 18, 27) and Supplementary Table S1.
RNA pulldown assay
RNA pulldown was performed as described previously (15) with some modifications. Biotin-labeled CTBP1-AS probe was prepared using Biotin RNA Labeling Mix (Roche) and T7 RNA polymerase as described. Biotinylated RNAs were treated with RNase-free DNase (Qiagen), and 10 pmol biotinylated RNA was heated to 60°C for 10 minutes and slowly cooled to 4°C. The RNA was mixed with 100 μg of precleared nuclear extract in RNA immunoprecipitation (RIP) buffer (15) supplemented with tRNA (0.1 μg/μL) and incubated at 4°C for 8 hours. A total of 60 μL of washed Streptavidin Agarose beads (Thermo Fisher Scientific) was added and incubated for an additional 1 hour at 4°C. The beads were washed 5 times with RIP buffer and boiled in SDS buffer, and the retrieved proteins were analyzed by Western blot analysis. To analyze the effect of small compounds on the interaction, chemicals were added to nuclear lysates 5 hours prior to mixing with RNA probes.
We used EZ-magna RIP RNA-binding protein immunoprecipitation kit (Millipore) according to the manufacturer's protocol. Briefly, the cell lysates were incubated with PSF antibody coupled beads at 4°C overnight. The RNA/antibody complex was washed six times. The RNA was then extracted using ISOGEN, and subjected to qRT-PCR.
RNA samples were validated to be high quality (RNA integrity score > 9.0) by using RNA bioanalyzer (Agilent). For gene expression microarrays, the Human Affymetrix Human Gene 1.0 ST Array (Affymetrix) was used in accordance with the manufacturer's protocol. Data analysis was performed using the Affymetrix Microarray Suite software.
Western blot analysis
Protein concentration was determined by performing the bicinchoninic acid (BCA) assay (Pierce). Obtained lysates were loaded on SDS-polyacrylamide gels, separated using electrophoresis, and subsequently electrotransferred. Membranes were incubated with the specific primary antibodies at 4°C overnight, and then incubated with secondary antibodies. Antibody–antigen complexes were detected using Western Blotting Detection Reagents (Pierce). For immunoprecipitation, extracts were incubated with the indicated antibodies overnight at 4°C. Following 2 hours of incubation with protein G agarose (GE Healthcare), beads were washed three times with lysis buffer, resuspended in 1x Sample buffer (Nacalai) and boiled for 5 minutes.
Cell proliferation assay
Cells (3 × 103) were plated and cultured in 96-well dishes. CellTiter 96 Aqueous Kit (Promega), an [3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium, inner salt] (MTS)-based assay was used to measure cell vitality and cell growth rate.
Flow cytometry analysis
22Rv1 and OHTR cells were seeded in 6-cm culture plates containing a cell culture medium with various concentrations of compounds or vehicle. The apoptotic cells were quantified using an Annexin V-fluorescein isothiocyanate (FITC)/propidium iodide (PI) Apoptosis Detection Kit (Thermo Fisher Scientific). The stained cells were analyzed with LSRFortessa (BD). Cell cycle was evaluated by measuring cellular DNA content after staining with PI. Data were analyzed using Modifit and FlowJo software (BD Biosciences).
Formalin-fixed tissues were embedded in paraffin and sectioned. A Histofine kit (Nichirei), which employs the streptavidin–biotin amplification method, was used for Antigen retrieval was performed by heating the slides in an autoclave at 120°C for 5 minutes in citric acid buffer (2 mmol/L citric acid and 9 mmol/L trisodium citrate dehydrate, pH 6.0). The antigen–antibody complex was visualized with a 3,3′-diaminobenzidine solution (1 mmol/L 3,3′-diaminobenzidine, 50 mmol/L Tris-HCl buffer, pH 7.6, and 0.006% H2O2). During IHC analysis, immunoreactivity was evaluated in more than 1,000 carcinoma cells in each case and the percentage of immunoreactivity (labeling index; LI) was determined by pathologists.
CRPC xenograft model
The ethics committee of animal experiments at the Tokyo Metropolitan Institute of Gerontology approved our study protocol. Prostate cancer cells, 22Rv1 and DU145, suspended in medium were mixed with Matrigel (BD Biosciences) and subcutaneously injected into one side of twenty 5-week-old male BALB/c nude mice (CLEA Japan). After the tumor volume reached 100 mm3, we performed castration in the mice harboring 22Rv1 tumors and randomly divided into two groups. Then we started intrapleural injection of 1 (22Rv1) or 5 mg/kg (DU145) No. 10–3 or vehicle (DMSO) diluted by PBS daily. Tumor dimensions were monitored using a caliper. Tumor volume was determined according to the formula 1/2 × a × b2 (a and b represent the minimal and maximal diameter of tumor, respectively). When mice were sacrificed, blood samples for serum biochemical analysis (Oriental Kobo) were collected from hearts. Tumors were homogenized in and RIPA buffer with protease inhibitor cocktail (Nacalai) for Western blot analyses. For IHC, tumors were formalin fixed and embedded in paraffin. To evaluate the toxicity, body weights for mice were recorded every week (18, 32).
Chromatin immunoprecipitation and chromatin immunoprecipitation sequencing
Chromatin immunoprecipitation (ChIP) and quantitative PCR (qPCR) were performed as described previously (15, 18). The fold enrichment relative to input was measured by performing qPCR, using KAPA SYBR Green PCR master mix (Sigma Genosys) and the ABI StepOne system (Thermo Fisher Scientific). We performed AcH3 ChIP-seq in LTAD/LNCaP cells using an Illumina HiSeq 2500 (Illumina). Details for ChIP-seq are as described previously (18, 32). Signal scores of transcription factor bindings were calculated using model-based analysis of ChIP-seq (MACS) and the threshold for the binding sites was set as P < 1e-4 and 1e-5.
We performed all experiments at least twice and confirmed similar results. Data are expressed as the mean ± SD. In most experiments using cell lines, we used the two-sided Student t test to determine the statistical significance between groups. Significance was defined as P < 0.05. qPCR analyses in cell lines were performed in technical replicates. Cell growth assay were performed in biological replicates. Other statistical tests are described in Figure legends. Microsoft Excel (Microsoft) or GraphPad Prism software ver. 6.0 was used for the statistical analysis.
The Gene expression omnibus (GEO) accession numbers for Microarray, RIP-seq and ChIP-seq data used in this study are GSE155803, GSE169139, and GSE155804.
Chemical library screening to identify inhibitors of PSF–RNA interaction
We first prepared four mCherry-tagged PSF deletion proteins overexpressed in 293T cells and applied them to characterize their interaction with their target lncRNAs (Fig. 1A). An RNA pull-down assay was performed using biotin-labeled RNA probes of CTBP1-AS (15), an AR-regulated lncRNA that was demonstrated to interact with PSF in prostate cancer cells (Fig. 1B). We found that the deletion of the RNA-binding domain (RBD) (ΔR) impaired its interaction with the RNA probe. Among the PSF deletion mutant proteins, we observed that N-terminal deleted PSF protein (PSF ΔN) and only the RBD of PSF protein (PSF RBD) interact with the lncRNA and are produced more efficiently in comparison with full-length one. Therefore, in this study, we first applied these lysates including proteins with RBD for chemical screening in vitro to identify small molecules modulating RBD function. Then, we carried out chemical array screening using the RIKEN Natural Products Depository (NPDepo) chemical library containing 30,707 compounds (Fig. 1C; refs. 30, 31). The binding of small molecules to PSF proteins tagged using mCherry on a chemical array was measured by detecting the mCherry signals. By comparing these signals with those on the array to which mCherry only protein was applied, we identified small molecules that were specifically bound to mCherry-tagged PSF proteins. Then, we observed that three compounds could interact with PSF RBD; meanwhile, 15 compounds were found to interact with PSF ΔN (Supplementary Fig. S1A).
To explore whether these candidate compounds function to inhibit PSF protein from interacting with CTBP1-AS, we further screened using an RNA pull-down assay, as well as an MTS-based cell proliferation assay. Regarding the RNA pull down assay, we observed that six compounds (No. 8, 10, 11, 12, 13, and 14) inhibited the interaction of the CTBP1-AS lncRNA probe with the PSF proteins (Fig. 1D). In the MTS assay, these compounds exerted repression of the proliferation of two prostate cancer cell lines at a concentration of 30 μmol/L (Supplementary Fig. S1B). Next, we investigated whether these compounds have an impact on hormone therapy–resistant prostate cancer using two CRPC models, created with two cell lines, 22Rv1 (18) and long-term androgen deprivation (LTAD) cells (15). The MTS assay results demonstrated that, among these compounds, No. 10 and No. 14 significantly repressed the proliferation of these cell lines (Supplementary Fig. S2A). Moreover, PSF is responsible for androgen-mediated repression of genes, such as CTBP1, p53, and SMAD3 by interacting with AR-induced CTBP1-AS in AR-dependent prostate cancer cells (15). Then, we screened whether these compounds abrogated transcriptional changes mediated by androgen and PSF in LNCaP cells. We observed that No. 10 and No. 14 significantly reversed the androgen-mediated CTBP1 repression, showing the inhibitory effect of these compounds on PSF function in prostate cancer cells (Supplementary Fig. S2B). Notably, we observed no significant toxic effect on cell proliferation of human normal fibroblasts, TIG3 cells, at this concentration (Supplementary Fig. S2C). Therefore, we focused on No. 10 (7-hydroxy-4-(4-methoxyphenyl)-8-(piperidylmethyl)chromen-2-one) and No. 14 (3,3-dimethylindoline) compounds to analyze their impact on the function of PSF (Fig. 1E).
No. 10–3 effectively inhibits PSF–lncRNA interaction
Then, we investigated the compounds with a similar structure to that of compounds No. 10 and No. 14 to identify more effective small compounds that inhibit PSF action. A total of five compounds similar to No. 10 (Supplementary Fig. S3A) and seven compounds similar to No. 14 (Supplementary Fig. S3B) were used for further analyses. We found three compounds [No. 14–5 (1-Methylindole-2-carboxylic acid), No. 10–1 (4-(4-methoxyphenyl)-9-benzyl-8H,10H-chromeno[8,7-e]1,3-oxazaperhydroin-2-one), No. 10–3 (7,8-dihydroxy-4-(4-methoxyphenyl)chromen-2-one)] that repressed prostate cancer cell proliferation more effectively than the original ones (Supplementary Fig. S3C and S3D). Therefore, we measured the inhibitory concentrations (IC50s) of these compounds using an RNA pull-down assay. Strikingly, among them, we found that No. 10–3 has the lowest IC50 (2.2 pmol/L) in inhibiting the interaction of PSF with CTBP1-AS probe (Fig. 2A–C). Meanwhile, we determined IC50 in the 0.5 - 2 μmol/L range for the other molecules (Fig. 2D). To obtain mechanical insights on the action of No. 10–3, we further investigated the binding mode of No. 10–3 that inhibits PSF action in silico. In the molecular modeling analysis, we used registered PSF structures in the Protein Data Bank. A PSF protein structural model was prepared for docking simulation based on its X-ray structure (PDB ID: 4WII) with high resolution (2.1 Å) (8). To identify the most reasonable model, we tried to find models bound tightly to No. 10–3 with high scores in the docking simulation. Thus, we obtained two representative models of No. 10–3 inhibiting PSF using different interaction patterns. One model is based on the interaction of No. 10–3 with the groove between RBD and the coiled coil domain and the other uses the pocket between RBD and DNA-binding domain (DBD). To confirm which model is compatible with our experimental data, we further performed an RNA pull-down assay by using deletion mutants of PSF (Supplementary Fig. S4A–S4D). We observed that the interaction of RBD-2-, RBD-4-, and ΔN-PSF with lncRNA probes was blocked by adding No. 10–3. However, no inhibitory effect was revealed regarding the interaction of RBD- and RBD-3-PSF with lncRNA, suggesting the importance of the region between RBD and the coiled-coil (488–524 a.a) for the inhibitory function of No. 10–3 (Fig. 2E; Supplementary Fig. S4A–S4D). More specifically, by introducing mutations to three amino acids (Y490, K516, and D517) expected to be associated with No. 10–3, the activity of No. 10–3 to inhibit the interaction was alleviated (Fig. 2F and G). Thus, these data supported the predicted bound structure of No. 10–3 with the groove formed between coiled-coil and RBD.
Next, we examined whether No. 10–3 blocks the RNA-binding activities of PSF or protein-binding of CTBP1-AS. Interestingly, RNA pull-down assay of CTBP1-AS indicates that CTBP1-AS could interact with several RNA-binding proteins and splicing factors such as U2AF2. This result is consistent with our past report that lncRNAs upregulated in CRPC tissues regulate splicing factor activity (33). Although the interaction of CTBP1-AS with PSF or PSF-centered RBP complex such as NONO (6) was repressed by No. 10–3 addition, the interaction with U2AF2 or other RNA-binding proteins (hnRNPU and DDX23) were preserved compared with PSF (Supplementary Fig. S4E). This result indicated that binding of CTBP1-AS with other protein complex is independent of PSF and not affected by No. 10–3. Moreover, PSF overexpression could increase the amount of interacted CTBP1-AS with PSF in the presence of No. 10–3 (Supplementary Fig. S4F and S4G), suggesting that PSF is the target of No. 10–3 to block the interaction. Therefore, we performed comprehensive study, RIP sequencing (RIP-seq) analysis, to investigate the effect of No. 10–3 on RNA binding ability of PSF. We found that bindings of some PSF-binding partners including CTBP1-AS to PSF were abrogated by No. 10–3 (Supplementary Fig. S4H). Importantly, although the bindings of most of PSF binding RNAs were observed to decrease by No. 10–3 treatment, a subset of target genes (17%) were severely affected (fold < 0.6; Supplementary Fig. S4I and S4J; Supplementary Table S2), suggesting the specificity of No. 10–3 on the regulation of binding genes at RNA level.
A secondary assay, an MTS assay, was performed to ensure that the compounds inhibited PSF activity, using several treatment-resistant cancer cells. The repression of prostate cancer cell proliferation by No. 10–3 was more evident in a CRPC model using 22Rv1 cells (IC50 = 1.2 μmol/L) compared with LNCaP cells (IC50 = 23 μmol/L; Fig. 3A), in line with the increased expression level and activity of PSF in CRPC models (18). We observed that the high expression of PSF and CTBP1-AS is correlated with a poor prognosis of breast cancer patients, as well as prostate cancer (Supplementary Fig. S5A), in line with the past reports (15, 27). Then, we examined the effect of No. 10–3 on treatment-resistant breast cancer cells, and observed that No. 10–3 treatment inhibits OHT-resistant (OHTR) breast cancer cells, compared with parental hormone-sensitive MCF7 cells (Fig. 3B). Moreover, the effect of No. 10–3 and No. 10–1 on the induction of apoptosis was measured. We observed that Annexin V/PI–positive apoptosis cells and PARP1 cleavage were increased by adding No. 10–3 to 22Rv1 (Fig. 3C and 3D) and OHTR cells (Fig. 3D and E) in line with PSF knockdown, demonstrating that No. 10–3 effectively induces apoptosis in these treatment-resistant cancer cells.
The epigenetic and splicing activity of PSF is modified by no. 10–3 treatment
In our past study, PSF epigenetically regulated the expression of cell-cycle regulators by binding to a lncRNA CTBP1-AS and HDAC complex (15). In addition, PSF is responsible for the gene regulation by directly binding to pre-mRNA to facilitate RNA processing and increase the important signals for cancer progression (18, 27, 29). We first analyzed the effect of No. 10–3 on such gene regulations at the transcriptional level using epigenetic control. We treated AR-positive LNCaP cells with 10 nmol/L dihydrotestosterone (DHT), a representative androgen, for 24 hours, and observed an androgen-dependent transcriptional repression of CTBP1, p53, and SMAD3, which are PSF target genes (15). qRT-PCR analysis revealed that the addition of No. 10, No. 10–1, and No. 10–3 reversed the repression induced by DHT treatment (Fig. 4A). The repression of PSF and HDAC complex recruitment to the enhancer and promoter regions of these genes were also demonstrated using a ChIP assay (Fig. 4B). Moreover, the expression of PSF-binding genes, such as AR (34), AR-V7 (35–37), SchLAP1 (18, 38), and AR-regulated genes (FKBP5 and ACSL3), at the RNA level (18) is positively maintained by PSF. qRT-PCR analysis showed that these target genes were repressed in 22Rv1 cells using No. 10–3 treatment (Fig. 4C). Mechanistically, the RIP assay demonstrated that the binding of PSF to these genes at the RNA level was abrogated (Fig. 4D; Supplementary Fig. S5B). In OHTR breast cancer cells, we observed the repression of PSF-target genes at the RNA level (ERα and SCFD2) by using No. 10, No. 10–1, and No. 10–3 treatment (Fig. 4E). Consistently, the interaction of PSF with these target pre-mRNAs was inhibited by using No. 10–3 (Supplementary Fig. S5C). Consistently, estrogen dependent induction of a representative ERα target gene, GREB1, was also repressed by No. 10–3, in line with the decreased expression of ERα by using No. 10–3 treatment (Supplementary Fig. S5D). By measuring the expression level of pre-mRNA of these PSF-binding genes, we observed that the expression level of pre-mRNA is not reduced, rather increased by No. 10–3 treatment, indicating that No. 10–3 blocks the process of transcribed RNA to be spliced into mRNA (Supplementary Fig. S5E and S5F).
We further analyzed the impact of No. 10–3 on prostate cancer cell-cycle conditions, because PSF promotes cell-cycle progression by epigenetically inhibiting cell-cycle regulators (15). Consistently, No. 10–3 treatment inhibits cell-cycle progression at the G2–M phase (Fig. 5A). Using Western blot analysis, we observed an enhanced expression of cell-cycle inhibitors, p53, p21 and p27, in 22Rv1 cells treated with No. 10–3 (Fig. 5B; Supplementary Fig. S6A). Notably, the formation of a PSF, HDAC, and Sin3A complex to epigenetically repress the gene expression was attenuated by adding No. 10–3 to cells in a dose-dependent manner, suggesting that the formation of such complex is also targeted by No. 10–3 (Fig. 5C). Indeed, PSF and HDAC complex recruitment to the promoter regions of these cell-cycle regulators was suppressed (Fig. 5D), suggesting the induction of cell-cycle regulators by blocking the histone deacetylation of the promoter. In contrast, we showed that histone acetylation of these regions was enhanced by using No. 10–3 treatment (Fig. 5D). We observed that p53 knockdown reversed the growth inhibition as well as p21 induction in 22Rv1 cells, suggesting the functional role of p53 (Supplementary Fig. S6B and S6C). In contrast to 22Rv1, p21 induction was not observed in AR-negative prostate cancer cells, DU145 cells, which possess mutated p53 (39). However, p27 induction (Supplementary Fig. S6A and S6B) and growth inhibition by No. 10–3 treatment was observed in DU145 cells independent of p53 (Supplementary Fig. S6C and S6D), suggesting that No. 10–3 can suppress cancer cell growth by modulating other pathways independent of p53. We also observed the repression of S-phase in No. 10–3 treated hormone therapy–refractory breast cancer OHTR cells (Fig. 5E). Using Western blot analysis, we found that cell-cycle inhibitors (p21, p27, and p53) were induced, although ERα expression at the protein level was repressed using No, 10–3 treatment (Fig. 5F). In line with the result obtained for prostate cancer cells, the repression of PSF/HDAC complex recruitment by No. 10–3 was demonstrated in this cell line (Fig. 5G).
Global transcriptional regulation by inhibiting PSF epigenetic function
Furthermore, to explore the comprehensive effect of the small molecule No. 10–3 on PSF-mediated signals, we analyzed the genome-wide histone acetylation status using chromatin immunoprecipitation (ChIP) and massive sequencing (ChIP-seq). Both LNCaP and LTAD cells were treated with No. 10–3 or vehicle, and their histone acetylated DNA regions were revealed by ChIP-seq analysis. To analyze the effect on androgen signaling, we also treated androgen-sensitive LNCaP cells with DHT or the vehicle. In total, the number of significant histone H3 acetylation (AcH3) was decreased in the presence of DHT and recovered using an No. 10–3 treatment, which is compatible with the PSF action of androgen-dependently repressing histone acetylation (Fig. 6A; Supplementary Fig. S7A). We also observed the upregulation of AcH3 modification using No. 10–3 treatment in LTAD cells. Venn diagrams showed the AcH3 sites newly induced using the No. 10–3 treatment, suggesting the distinct epigenetic activation induced by No. 10–3 to modify the global transcriptional signals (Fig. 6B; Supplementary Fig. S7A and S7B). We also confirmed these results by conventional ChIP assay or additional study in LTAD cells (Supplementary Fig. S7C–S7E). Then, we performed microarray analysis and identified specific transcriptional signals induced by No. 10–3 treatment, with enhanced AcH3 sites (Fig. 6C). Gene ontology (GO)-term analysis revealed that No. 10–3 treatment was strongly enriched in some processes, such as apoptosis or cell-cycle regulation, including regulators of cyclin-dependent protein kinase activity (Fig. 6D). Moreover, using gene set enrichment analysis (GSEA) of a set of differentially expressed genes by using No. 10–3 treatment, we identified an enrichment of apoptotic signaling in response to endoplasmic reticulum stress pathway (Fig. 6E). In addition, the induction of cyclin-dependent kinase (CDK) inhibitors was also observed using No. 10–3 treatment in microarray analysis, showing the epigenetic role of PSF in repressing these cell-cycle regulators (Fig. 6F). Meanwhile, we revealed that other cell-cycle regulators, particularly mitotic nuclear division-associated signals, were severely downregulated using GO-term and GSEA analyses (Fig. 6G). Thus, collectively, our analysis indicates that cell-cycle inhibition or apoptosis is promoted by global epigenetic and transcriptional regulation due to PSF-inhibition with the treatment by No. 10–3.
Targeting PSF interaction with lncRNA exhibits antitumor action in treatment-resistant prostate cancer
Notably, RNA-binding properties have essential roles in aggressive cancer states. However, it has not been fully clarified whether targeting the interaction of PSF with RNA could be a potent therapeutic target in treatment-resistant cancer. Thus, we investigated the potency of pharmacologically targeting PSF-RNA associations using No. 10–3. In xenograft models of AR-positive 22Rv1 cells, we performed castration to inhibit androgen action to mimic hormone therapy. A marked inhibition of castration-resistant tumor growth was observed using No. 10–3 treatment (1 mg/kg, five times per week; Fig. 7A). Meanwhile, no significant toxic effect was observed by measuring the body weight and performing biochemical testing in this experiment (Fig. 7B; Supplementary Fig. S8A). IHC analysis showed a decrease in the cell proliferation marker Ki67 and AR protein expression using No. 10–3 treatment (Fig. 7C and D). In line with the IHC results, Western blot analysis dramatically showed a repressed AR protein level and the induction of p27 and p21 proteins, along with an enhanced histone acetylation (Fig. 7E). In qRT-PCR analysis, we observed a decreased mRNA expression level of AR and PSF-interacting lncRNA in 22Rv1 cells, SchLAP1 (Fig. 7F; ref. 18). Furthermore, in xenograft models of AR-negative and p53-mutated DU145 cells, we tested the efficacy of No. 10–3 for tumor growth inhibition. We observed inhibition of tumor growth by administrating No. 10–3 treatment (5 mg/kg, five times per week) without obvious toxic effect on mice (Supplementary Fig. S8B and S8C). By Western blot analysis, we also observed remarkably enhanced histone acetylation and cleaved PARP, suggesting the apoptosis induction by regulating histone modification in tumors (Supplementary Fig. S8D). We also noticed p27 induction and repression of SchLAP1 expression level, in line with the blockade of PSF action (Supplementary Fig. S8E). Together, these results demonstrated the efficacy of targeting RBP interaction with RNA to overcome treatment-resistant tumor growth.
Physiologically, PSF is a critical factor in maintaining the transcriptional elongation of long genes in neurons (9, 40). In cancer, PSF regulates cell-cycle–associated gene expression by regulating the epigenetic conditions (15). Moreover, PSF binds to its target mRNAs to promote RNA processing (6, 11, 12, 41, 42). In this study, using high-throughput chemical screening, we identified PSF inhibitors that demonstrated antitumor activity. Notably, No. 10–3 inhibits the in vitro RNA interaction with PSF at the picomolar level of IC50 value. Then, we explored the molecular mode of No. 10–3 action in silico using docking analysis. We showed the decreased effect for the repressive action of No. 10–3 on the interaction with CTBP1-AS by mutating PSF proteins at specific amino acids predicted to interact with No. 10–3, supporting the notion that No. 10–3 targeted the groove produced by RBD and the coiled-coil domain. Coiled-coil domain is important for the action of PSF on nucleic acids in the nucleus (8). Consistently, further analysis of the effect of No. 10–3 on the PSF-binding genes at RNA level by RIP and comprehensive RIP-seq in hormone therapy–refractory cancer cells revealed that No. 10–3 inhibits the interaction of PSF with most of binding genes at RNA level, severely on a subset of the target genes, although further analysis to determine the specificity of No. 10–3 on its target genes is necessary in the future study. One important function of PSF is the processing of mRNA posttranscriptionally to regulate AR, ESR1, spliceosome genes, and SCFD2 mRNA expression in hormone therapy-refractory cancer cells (18, 27). We demonstrated that the interaction of PSF with these target RNAs was inhibited via No. 10 and No. 10–3 treatment. Subsequently, qRT-PCR analysis revealed that the expression of these target genes was also repressed. Thus, these results demonstrated that No. 10–3 would inhibit the RNA recognition by PSF in hormone therapy-refractory cancer cells.
Cellular and xenograft studies revealed that the inhibition of PSF action leads to the activation of apoptosis, epigenetic modification, and the modulation of hormone receptor activity. These inhibitors may be useful as a tool for the investigation of the cellular function of PSF. Androgen-induced lncRNA CTBP1-AS mediates androgen-dependent PSF action in prostate cancer (15). Previously we have shown that the epigenetic control via PSF binding to genomic regions through DBD accelerates tumor growth by repressing cell cycle regulators, such as p53, SMAD3, and CDK inhibitors (15, 18). We have also shown that CTBP1-AS interacts with PSF in the nucleus and promotes AR signaling by repressing AR-associated corepressors, such as CTBP1 (15). In this study, we observed that the inhibition of PSF-lncRNA action by small molecules suppressed the binding of PSF to the promoter and enhancer of cell-cycle regulators and activates histone acetylation. Moreover, No. 10–3 also inhibits epigenetic regulation by PSF through inhibiting genomic binding and HDAC1 interaction with PSF. Previous study demonstrated that coiled-coil domain is important for genomic action of PSF (8). In addition, modulation of RBD at the protein level is required for interaction with HDAC1, promoting deacetylase activity (43). Because No. 10–3 targets the groove made by RBD and coiled-coil domain according to our predicted model, we speculate that No. 10–3 mediated blockade of PSF also affects the interaction of PSF with these deacetylase complex and the activity of coiled-coil domain for genomic binding. Furthermore, we integrated microarray and ChIP-seq analyses, and then investigated the genome-wide changes induced by the inhibitors. We found that apoptosis-related pathway and cell cycle–related genes, such as CDK inhibitors, were the main targets of histone acetylation activation via No. 10–3 treatment, indicating that the epigenetic action via PSF recruitment was blocked to induce apoptosis. Therefore, we can assume that the inhibition of PSF results in the alleviation of the epigenetic action, subsequently exerting an antitumor action.
We have previously revealed that a high expression of PSF in cancer tissues predicts a worse prognosis of patients with prostate and breast cancer (18, 27). PSF expression is higher in endocrine therapy—resistant breast and prostate cancer models (LTAD, 22Rv1, and OHTR cells), compared with the parental hormone-sensitive cells. Therefore, using a CRPC xenograft model, we assessed the in vivo efficacy of No. 10–3 in treatment-resistant cancer tumor growth. In prostate cancer, whereas androgen deprivation therapy is effective initially, many of these patients develop a more aggressive lethal therapy-resistant CRPC (19, 20). It is well known that an increased AR signaling has a critical role in the transition to CRPC (19). Ligand-independent splice variants, such as AR-V7 (36), are also induced under castrate conditions through spliceosome dysregulation in CRPC (18). Thus, targeting epigenetic or chromatin modeling factors involved in AR activation is promising targets to be therapeutically effective against CRPC. We demonstrated that castration-resistant tumor growth of 22Rv1 CRPC model cells was severely impaired via No. 10–3 administration. Notably, AR expression was suppressed, suggesting the blockage of PSF-mediated AR mRNA processing via No. 10–3. Consistently, the expression of cell-cycle inhibitors, such as p21 and p27 (44), was increased, along with an enhanced histone acetylation. These results indicate that epigenetic regulation via No. 10–3 induced these cell cycle inhibitors to alleviate tumor growth. Moreover, we also showed the in vivo antitumor action of No. 10–3 on AR-negative DU145 cells with mutated p53, suggesting that modulation of other signals independent of p53 such as p27 and SchLAP1 would be effective for inhibiting tumor growth. We also found that No. 10–3 could inhibit OHT-resistant breast cancer cell proliferation. Therefore, this study raises the possibility that targeting RBP action by blocking the interaction with RNA is a relevant tool for the improvement of hormone therapy to overcome the treatment resistance of endocrine-related cancers.
Recently inhibition of CDK4/6 by small-molecule inhibitors attracted much attention because it has remarkably improved the outcome of patients with ER-positive breast cancer in combination with endocrine therapy (45). Mechanistically, blockade of CDK4/6 repressed phosphorylation of retinoblastoma (Rb) protein, that accelerate cell cycle. It is also shown that CDK4/6 inhibitors also repressed CRPC tumor growth without significant attenuation of androgen signaling (46, 47). However, the clinical efficacy of CDK4/6 inhibitors in CRPC would be limited because disruption of Rb is frequently observed in the late phase of CRPC (48). Thus, the activity of CDK4/6 inhibitors was dependent on the integrity of Rb. Moreover, CDK4/6 inhibition is limited by an inability to induce complete and durable cell-cycle arrest due to early adaptation mediated by persistent G1–S-phase cyclin expression via CDK2 (49). Similarly, we demonstrated that No. 10–3 repressed cell-cycle progression by inducing cell-cycle inhibitors such as p53, CDK1A/p21, and CDK1B/p27. In contrast to CDK4/6 inhibitors, these CDK inhibitors works at multiple steps of cell-cycle checkpoints including CDK4/6 and CDK2 (44, 50), suggesting that No. 10–3 might be effective for prostate tumors with Rb loss. Therefore, it is postulated that No. 10–3-mediated cell-cycle inhibition would be beneficial for patients with CRPC and hormone therapy-refractory breast cancer with resistance to CDK4/6 inhibitors. We also observed that genome-wide epigenetic modification by histone acetylation induce apoptosis, suggesting that No. 10–3 regulates pathways to induce apoptosis in cancer cells. More importantly, blockade of PSF-mediated mRNA production by No. 10–3 repressed SchLAP1, AR, SCFD2, and ERα expression level in hormone therapy–refractory cancer cells. Thus, it will be of interest to examine the combination of No. 10–3 with other drugs targeting hormone signals or chemotherapy in hormone therapy–refractory cancers.
Y. Kondoh reports grants from AMED P-CREATE during the conduct of the study. Y. Suzuki reports grants from University of Tokyo during the conduct of the study and grants from University of Tokyo outside the submitted work. M. Yoshida reports grants from AMED P-CREATE during the conduct of the study and grants from Meiji Co., Ltd. outside the submitted work. No disclosures were reported by the other authors.
K. Takayama: Conceptualization, data curation, formal analysis, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. T. Honma: Funding acquisition, investigation, visualization, writing–original draft. T. Suzuki: Investigation, visualization. Y. Kondoh: Resources, data curation, funding acquisition, investigation, methodology, writing–review and editing. H. Osada: Resources, supervision. Y. Suzuki: Investigation. M. Yoshida: Resources, supervision, funding acquisition. S. Inoue: Supervision, funding acquisition, writing–original draft.
This work was supported by grants of AMED under grant number JP18ck0106194 (to S. Inoue) and JP19cm0106002 (to M. Yoshida, T. Honma, Y. Kondoh); by grants from the JSPS [number 20K21667 (to S. Inoue) and number 17H04334 (to K. Takayama); by grants from Takeda Science Foundation (to S. Inoue, K. Takayama), Japan; by grants from Sagawa Foundation for Promotion of Cancer Research (to K. Takayama), Japan; by grants from The Mochida Memorial Foundation for Medical and Pharmaceutical Research (to K. Takayama), Japan.
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.