TR3 enhances AR Variant Production and Transactivation, Increasing Androgen Independency of AR Signaling in Prostate Cancer Cells

Increased expression of constitutively active androgen receptor variants (AR-Vs) is associated with the development of advanced castration-resistant prostate cancer. The pro-oncogenic function of TR3, an orphan nuclear receptor, has been reported in various cancers including prostate cancer. However, the roles of TR3 in androgen receptor (AR) expression and signaling in prostate cancer cells are poorly understood. Western blotting and quantitative RT-PCR were used to evaluate AR and AR-V expression levels affected by TR3 expression level. RNA-seq analysis, coimmunoprecipitation, cross-linked RNA-immunoprecipitation, and single-strand RNA protection and pull-down assays were conducted to elucidate the molecular mechanisms by which TR3 affected AR-V production. blot pull-down assay using lysates extracted AdCtrl. Total proteins from each cell lysate used in analysis (A). CLIP analysis using anti-hnRNP A2B1 antibody showed an enrichment of hnRNP A2B1 at Reg4 and Reg9 within AR pre-mRNA in TR3-overexpressing CWR22rv cells (B). C-D, TR3 interacts with AR pre-mRNA. ssRNA protection assays performed using puried TR3 and splicing factors (hnRNP A2B1 and HUB/HUR) showed protection of a 30mer ssRNA oligo of AR-Reg4 labelled with biotin at the 5’ end (ssRNA oligo) (C). ssRNA pull-down assays performed using the same puried protein fractions show the binding of TR3 and splicing factors to ssRNA oligo (D). Protein levels were analyzed via western blot analysis using anti-TR3, anti-hnRNP A2B1, and anti-HUR antibodies.

variants (AR-Vs), which lack LBD, is associated with the development of advanced castration-resistant prostate cancer (CRPC), leading to the failure of hormone therapy [2]. Among the multiple AR-Vs reported, AR-V7 (AR3) is the most common in prostate cancer cell lines and tumors and is generated by splicing between exon 3 and cryptic exon 3 (CE3) as well as exon 3 duplication [3]. Mutations and genomic rearrangements are the two major causes of multiple drug resistance during prostate cancer progression, especially during the development of advanced CRPCs [4,5]. These mutations and genomic rearrangements result in the generation of alternative promoters and cryptic splice sites that drive an alternative splicing process.
Several splicing factors have been found to be associated with alternative splicing events in the AR precursor mRNA (pre-mRNA) in prostate cancer cells. Among these, heterogeneous nuclear ribonucleoprotein (hnRNP) A2B1, hnRNP E1/E2, and HUR (ELAV-like protein 1, ELAVL1) are RNA-binding proteins (RBPs) that bind to AR pre-mRNA and regulate the expression of AR-Vs in advanced CRPCs [6,7]. hnRNP E1/E2 contains KH domains and HUR contains RRM domains, and both of these domain types bind to the UC-rich region in AR pre-mRNA [6]. In contrast, hnRNP A/B proteins, especially hnRNP A1 and A2B1, which contain RRM domains, bind to AR pre-mRNA splice sites, and enhance the expression of some AR-Vs, particularly AR-V7, AR-V1, AR-V4, and AR-V5, in CRPCs [7]. Enrichment of hnRNP A1/A2B1 at splice sites involved in AR-V production was further increased in enzalutamide-resistant prostate cancer cells [7]. Additionally, NOVA-2 is an alternative splicing factor, whose expression is upregulated in some cancers [8], although its role in prostate cancer is poorly understood. It is a KH domain-containing RBP, which binds to the YCAY sequence in RNA and regulates the stabilization and transport of intron-excised RNA.
Human orphan nuclear receptor TR3 (also known as NR4A1) is an immediate early response gene induced by a diverse range of signals, including stressors, cytokines, growth factors, and small molecular compounds. TR3 expression is associated with many physiological and pathological processes, such as cell survival and death, in ammation, and cancer development [9]. TR3 is upregulated in many cancers, including lung, colorectal, breast, and prostate cancers [10]. In addition, TR3 signaling activation enhances cancer cell proliferation and tumor progression [10], while loss of TR3 function by retinoid and its derivative compounds induces apoptosis [11]. TR3 overexpression increases prostate cancer cell viability [10]. However, the roles of TR3, especially in AR expression and signaling, are poorly understood.
The present study is the rst to demonstrate that TR3 affects AR splicing process as well as its expression and further enhances androgen-independent AR activity in prostate cancer cells. These results, together with the stimulation of tumorigenesis by TR3 overexpression, suggest that TR3 is a pivotal controller of AR signaling in prostate cancer cells and plays a crucial role in cancer progression and the maintenance of advanced CRPCs.
Next-generation sequencing (NGS) high throughput RNA-Seq analysis CWR22rv cells were infected with AdTR3 or AdCtrl, maintained for 24 h, and then harvested for RNA isolation followed by whole transcriptome analysis. RNA was converted to cDNA using a Clontech SMARTer Stranded RNA-Seq Kit with the proprietary SMART stranded N6 primer and SMARTScribeTM reverse transcriptase. The postcapture sequencing libraries were pooled and sequenced on the Illumina HiSeq 2500 platform using the 2 × 100 bp settings. Reads were mapped to the human genome reference HG19 using MapSplice. Gene expression was quanti ed using EdgeR and quantile normalized. Gene set enrichment analysis was performed using GSEA program and DAVID analysis on a preranked list of differentially expressed (> 1.5 fold) genes between AdTR3-and Ad-Ctrl-infected CWR22rv cells. Genes were clustered, and the difference in gene expression was classi ed using the MeV program. Novel splice variants were detected by aligning the novel splices against whole transcriptomes. Alterations in the splice junction rates between exons in full-length AR gene (AR8) were analyzed through Integrative Genomic Viewer (IGV) analysis.

Cross-linked RNA-immunoprecipitation (CLIP)
CWR22rv cells were infected with AdTR3 or AdCtrl, maintained for 24 h, and cross-linked with 1% formaldehyde (Sigma-Aldrich, St Louis, MO, USA) for 10 min. The xed cells were then subjected to crosslinked RNA-immunoprecipitation (CLIP) analysis as previously described [12] with minor modi cations. The protein-bound primary RNA fragments were enriched through immunoprecipitation with anti-hnRNP A2B1 antibodies (Supplementary Table S2). DNA contaminants were removed by treating with RNase-free DNase (ThermoFisher Scienti c, Rockford, IL, USA) at 37°C for 30 min, following which RNA fragments were obtained using chloroform extraction, and cDNA was synthesized using random hexamers (Enzynomics) and M-MLV Reverse transcriptase kit (Promega, Republic of Korea). Enrichment of AR pre-RNA fragments was analyzed via RT-PCR using the CLIP primers listed in Supplementary Table S1.
Single-strand RNA (ssRNA) protection assays Thirty-mer ssRNA labeled with biotin at the 5' end (ssRNA oligo) was designed for RNA-protein interaction assays. ssRNA oligo sequence was identical to that found within Reg4, which contains one hnRNP A2B1 binding site overlapping with one HUB/HUR and two other HUB binding sites but has no hnRNP E1/E2 binding site. ssRNA oligo was incubated with or without the puri ed protein (FLAG-TR3 or splicing factors hnRNP A2B1 and HUB/HUR) in a binding buffer [20 mM Tris (pH 8.0), 0.5% sucrose, 1 mM DTT, 1 mM MgCl 2 , 1 mM EDTA, and 2% glycerol in DEPC water] overnight at 4°C. RNA and protein degradation was minimized by adding RNase (RNasin) and protease inhibitors, respectively. Reactions were analyzed on a 3% horizontal agarose gel in TBE buffer containing 10 µg/ml EtBr. Gels were cooled and run in cooled TBE buffer. Free RNA and RNA-protein complex signals were detected using a UV transilluminator. ssRNA sequence is listed in Supplementary Table S1. ssRNA oligo pull-down assays The interactions between ssRNA oligo and proteins (hnRNP A2B1, hnRNP E1/E2, HUB/HUR, and TR3) were examined using ssRNA oligo pull-down assays as previously described [3], but with minor modi cations. ssRNA oligo was immobilized onto AccuNanoBead™ Streptavidin Magnetic Nanobeads (400 nm, 5 mg/ml; Bioneer, Korea) for 1 h at 4°C in the presence of RNasin, brewel yeast tRNA (Roche, Basel, Switzerland), and salmon sperm-sheared DNA (ssDNA). Puri ed TR3 and splicing factor proteins were added into each interaction reaction in the presence of protease inhibitors and BSA. The interaction reactions were incubated overnight at 4°C. ssRNA oligo-bound proteins were analyzed using western blot analysis. RNA pull-down assay was also performed using whole cell lysates extracted from CWR22rv cells infected with AdTR3 or AdCtrl. Total proteins from each cell lysate were used in each binding reaction.

Xenograft animal model
Healthy and microbiologically monitored 4-week-old male NOD.CB17-Prkdc SCID/J mice, obtained from Korea Research Institute of Bioscience and Biotechnology (Daejeon, Korea), were gently injected with 100 µl of injectable anesthetics [Zoletil 50: Rompun: saline buffer (20:10:270)]. Mice were then warmed on a veterinary warming system. Inducible TR3-overexpressing (TR3; subline #2) or control (EV) CWR22rv cell clones (2×10 6 cells/site mixed 1:1 with Matrigel) were subcutaneously implanted into the shoulder of each mouse. The following day, mice were provided distilled water with or without 2 µg/ml doxycycline. Tumor size and volume were monitored and measured three times a week. Two days after the last tumor measurement, the animals were subjected to CO 2 -induced euthanasia and sacri ced, and the tumors were extracted and weighed. Animals were maintained inside a cleaned bench in an animal room with a 12-h light/dark cycle and controlled temperature during all processes. The sterilized mouse cages, water, food, and bedding used for the animal study were replaced once a week. Ultra-ne II short needles (U-100 INSULIN 30 gauge 5/16'' (8 mm) needle) were used for all injections to minimize any pain and lesions. Tumor volumes were calculated using the following formula: length × width × height × 0.5236 (V = π × L × W × H/6 = L × W × H × 0.5236) [13]. Statistical signi cance was calculated using two-tailed t-test analysis. p-value < 0.05 was considered statistically signi cant. All animal procedures were approved by the Institutional Animal Care and Use Committee (IACUC) of Chonnam National University (permit number: 2012-44). Animal experiments have been performed in accordance with the ARRIVE/NC3R guidelines.

Quanti cation and statistical analysis
Animal experiments were performed using six mice in each group. Data, which were obtained from more than three independent experiments, are presented as the mean ± SEM. Statistical signi cance was calculated using one-way ANOVA with Tukey's post hoc test and two-tailed t-test analysis. p-value < 0.05 was considered statistically signi cant.

TR3 regulates the expression of AR and AR-Vs in prostate cancer
To investigate the oncogenic function of TR3 in prostate cancer, we analyzed its expression pro le in prostate tumors (Gene Expression Omnibus database) and evaluated the correlation between TR3 and AR expression levels in prostate tumor patients (The Cancer Genome Atlas database). We found that TR3 expression is upregulated in primary prostate tumors (Fig. 1A) and that its expression level is positively correlated with that of AR (Fig. 1B). Intriguingly, equally high levels of TR3 expression were observed in the loci adjacent to primary tumors. This is probably because immune de ciency and alteration of the tumor microenvironment by TR3 affects the physiology of tumor-adjacent cells, facilitating cancer progression [14]. In addition, TR3 expression was detected in all tested AR-positive prostate cancer cell lines, but not in AR-negative prostate cancer cell lines and was induced by 5α-dihydrotestosterone (DHT) (Fig. S1A), indicating TR3 as an androgen-responsive gene.
As TR3 and AR expression levels were positively correlated in prostate tumors, we investigated whether TR3 regulates AR expression. Overexpression of TR3 enhanced the protein levels of AR in androgenindependent CWR22rv cells and other AR-positive prostate cancer cells (LNCaP and C4-2) treated with or without androgen ( Fig. 1C and Fig. S1B). Intriguingly, we observed other AR-Vs in TR3-overexpressing cells, which were absent in the controls (Fig. 1C). In addition, the mRNA levels of AR and AR-V7 were enhanced in these TR3-overexpressing cells treated with or without androgen, when examined using primers speci cally covering exon 3/exon 4 (AR) and exon 3/CE3 (AR-V7) ( Fig. 1D and Fig. S1C). Furthermore, TR3 overexpression increased the mRNA level of AR-V7 signi cantly more than that of the exon 1/exon 2-containing AR in CWR22rv cells (Fig. 1E), which was quanti ed through qPCR, pointing to the effects of TR3 on AR pre-mRNA splicing process as well as AR transcription. As expected, silencing TR3 strongly decreased the protein and mRNA levels of AR and AR-Vs in AR-positive prostate cancer cells ( Fig. 1F-G and Fig. S1D). In addition, treatment with the TR3 antagonist, DIM-C-pPhOH, strongly decreased the protein level of AR and AR-Vs in the presence or absence of DHT (Fig. S1E). However, in TR3-knockdown CWR22rv cells, the decreased AR and AR-V protein levels by TR3 silencing were not restored following treatment with MG-132 and chloroquine, which are blockers of protein degradation pathways (Fig. S1F), suggesting that TR3 regulates AR expression at mRNA level.
TR3 regulates the expression of AR splicing variants, altering splice junction rates between exons The coupling of transcription and splicing processes is well-known [15][16][17]. Therefore, to further investigate the effects of TR3 on AR splicing process and expression, we rst analyzed the potential binding sites of TR3 (TRE: AGGTCA) of TR3 ~ 5 kb upstream of the transcription start site and in the region covering the whole AR gene using Annhyb, a bioinformatics-based tool for sequence analysis. Four potential binding sites (P, A, B, and C) were found; site P was located ~ 3.4 kb upstream of the transcription start site, while the other three sites (A-C) were located within AR gene introns. Site A was located near the 3' splicing site (3'ss) of exon 3, site B near the 5' splicing site (5'ss) of CE3, and site C near the 5'ss of exon 4 ( Fig. 2A). ChIP assays showed that TR3 might bind to all four sites; however, TR3 overexpression preferably enhanced TR3 binding to the regions containing sites P and C, but not sites A and B (Fig. 2B). These results suggest that TR3 binds to TREs in the AR gene and regulates its expression. However, we cannot rule out the possibility that TR3 binding to intron regions, especially to site C, affects AR pre-mRNA splicing process.
To investigate the alteration of AR splicing events in TR3-overexpressing prostate cancer cells, we performed next-generation sequencing (NGS) high-throughput RNA-Seq analysis of RNA obtained from TR3-overexpressing and control CWR22rv cells. The INDEL coupling with Integrative Genomic Viewer (IGV) analysis of RNA-Seq data of AR transcripts revealed that splice junction rates between exons within the pre-mRNA of full-length AR gene (AR8) were generally decreased (up to 50%) when TR3 was overexpressed in CWR22rv cells, while the splice junction rate between exon 1b y and exon 2 increased (~ 30%) (Fig. 2C). The splicing between exon 1b y and exon 2 has been identi ed in AR transcript variant 2 (also known as AR45), a short transcript that lacks exon 1, which is produced by utilizing an alternative transcription start site located between − 3 and + 5 of exon 1b y [18]. Overexpression of AR45 may either repress or stimulate AR transactivation, depending on the cellular context.
We then validated the expression of AR splicing variants by performing RT-PCR analysis using speci c primer pairs covering exon 1b y , exon 2, exon 3, cryptic exon CE1, cryptic exon CE3, and exon 4. The results showed that the increase in the level of RNA containing splice junctions between exon 1b y and exon 2 (found in AR45) and exon 3 and CE3 (found in AR-V7) was higher than the general increase in total AR mRNA containing the exon 3-exon 4 junction, which was upregulated by TR3 activation of the AR promoter (Fig. 2D). Overexpression of TR3 also markedly increased the levels of RNA containing exon 1b y -exon 2 and exon 3-CE1 in C4-2 cells, whereas AR-Vs containing exon 3-CE1 or exon 3-CE3 were hardly detectable in LNCaP cells (Fig. S2). These results suggest that TR3 affects the AR splicing process, especially in advanced CRPC cells, such as CWR22rv and C4-2 cells.
TR3 physically interacts with some splicing factors involved in AR splicing process Because splicing factors and spliceosomal complexes play a crucial role in the regulation of the splicing process, we rst examined whether TR3 overexpression altered the expression of some splicing factors, particularly NOVA-2, HUB (ELAV-like protein 2, ELAVL2), and other splicing factors involved in alternative splicing of AR pre-mRNA, such as hnRNP A2B1, hnRNP E1/E2, and HUR. RNA-Seq analysis revealed ~ 6.4fold induction of NOVA-2 and ~ 1.7-fold induction of HUB in TR3-overexpressing cells compared with the control (Fig. 3A). The increase in NOVA-2 and HUB mRNA levels was con rmed through RT-PCR analysis (Fig. 3B), although western blot analysis results revealed that NOVA-2 protein level was only marginally increased (Fig. S3). Further, other splicing factors, such as hnRNP A2B1 and hnRNP E1/E2, were expressed at high basal levels in CWR22rv cells and showed no signi cant changes in their mRNA and protein levels upon TR3 overexpression.
hnRNP A2B1 plays a critical role in the control of the AR splicing process [7], and it likely interacts with TR3, based on the analysis of the protein interaction network [19]. Because the protein-protein interaction between TR3 and splicing factors might alter spliceosomal component recruitment and complex formation to control the AR splicing process, we investigated the physical interactions between TR3 and splicing factors (hnRNP A2B1, HUB/HUR, and hnRNP E1/E2) in TR3-overexpressing CWR22rv cells. Coimmunoprecipitation assays revealed that TR3 interacted with splicing factors hnRNP E1/E2 and hnRNP A2B1 (Fig. 3C). These results suggest that TR3 may be associated with spliceosomal complexes on AR pre-mRNA and may alter the splicing process.
TR3 overexpression alters the recruitment of some splicing factors near splicing sites, interacting with AR pre-mRNA We further investigated whether TR3 altered the recruitment of splicing factors near splice sites. We chose 9 regions, designated Reg1-9, located ~ 1 kb upstream and/or downstream of the exon/intron junction of AR pre-mRNA and covering exons 1-4 (Fig. 4A). This is because spliceosomal components, such as snRNPs (U2/U2AF snRNP and U1 snRNP) and other hnRNPs, mostly bind to 3'ss and 5'ss of introns and branch sites [20], and the region covering exons 1-4 contains cryptic exons, which are frequently found in AR-Vs. Putative binding sites of some RBPs (hnRNP A2B1, hnRNP E1/E2, HUB/HUR, and NOVA2) in the 1.17-Mb AR pre-mRNA were investigated and found within Reg1-9, using SpliceAid program (a database of strictly experimentally assessed target RNA sequences in humans). Additionally, cross-linked RNA immunoprecipitation (CLIP) assays using anti-hnRNP A2B1 antibody and TR3overexpressing or control CWR22rv cells revealed strong enrichment of hnRNP A2B1 at Reg4, which contains several putative hnRNP A2B1 binding sites, upon TR3 overexpression (Fig. 4B). We also detected the enrichment of hnRNP A2B1 at Reg9, which contains no putative binding site for hnRNP A2B1, but it has hnRNP E1/E2 and HUB/HUR binding sites. It is possible that hnRNP A2B1 enrichment at Reg9 was due to protein-protein interactions between hnRNP A2B1 and other spliceosomal complex components.
To examine whether TR3 is associated with the spliceosomal complex at Reg4, we performed in vitro RNA-protein interaction assays involving single-strand RNA (ssRNA) protection and ssRNA oligo pulldown assays. A 30-mer ssRNA labeled with biotin at the 5' end (ssRNA oligo) was designed to include the sequence found within Reg4, which contains one hnRNP A2B1 binding site overlapping with one HUB/HUR and two other HUB binding sites, but it has no hnRNP E1/E2 binding site. ssRNA protection assays, which were performed using the ssRNA oligo and either puri ed TR3 or splicing factors (hnRNP A2B1 and HUB/HUR), showed that both TR3 and the splicing factors bound to and protected this ssRNA oligo from degradation (Fig. 4C). The ssRNA oligo migrated on agarose gel based on the formation of RNA-TR3 or RNA-splicing factor complex. Furthermore, ssRNA oligo pull-down assays performed using the same puri ed TR3 and splicing factors revealed that TR3 and the splicing factors, hnRNP A2B1 and HUB/HUR, bound to the ssRNA oligo (Fig. 4D). Together, these results suggest that TR3 binds to AR pre-mRNA and alters a recruitment of some splicing factors near splicing sites, which is a novel function of TR3.

TR3 enhances androgen-independent and androgen-dependent transactivation of ARs
We then examined the effect of TR3 on the transcriptional activity of ARs (full-length AR, AR-FL; AR Nterminal domain, AR-NTD; and AR-V7) based on luciferase reporter assays. The AR-NTD lacking LBD was used to represent the ligand-independent nature of AR, and AR-V7 was used to mimic AR-V behavior in CWR22rv cells. Overexpression of TR3 markedly enhanced both androgen-dependent and -independent transactivation of exogenous AR-FL, AR-NTD, and AR-V7 overexpressed in PPC1 cells in a dose-dependent manner (Fig. 5A-C and Fig. S4A-C). In addition, overexpression of TR3 signi cantly enhanced the androgen-induced transactivation of endogenous AR in androgen-dependent LNCaP cells (Fig. 5D).
The subcellular localization of ARs and recruitment of coactivators to AR target genes are important for AR transactivation. TR3 overexpression promoted the nuclear translocation of ARs even in the absence of androgen (Fig. 5G). In addition, co-overexpression of TR3 seemed to facilitate the recruitment of AR coactivators, such as SRC-2, to the AR response element, synergistically enhancing the transactivation of AR (Fig. 5H). Furthermore, the physical interaction between TR3 and androgen-independent AR-NTD (Fig.   S4D) suggests a molecular mechanism through which TR3 increases the activity of ARs in both an androgen-dependent and -independent manner. Collectively, these results show the profound effect of TR3 on the transactivation of ARs in prostate cancer, both in an androgen-dependent and -independent manner.
TR3 overexpression promotes prostate cancer cell proliferation and mobility, and enhances in vivo tumorigenesis of androgen-independent prostate cancer cells RNA-Seq analysis showed the upregulation of genes that are involved in cell proliferation, migration, and invasion as well as the PI3K/AKT/MAPK signaling pathway when TR3 is overexpressed (Fig. 6A-B). PI3K/AKT/MAPK signaling is ampli ed as prostate cancer progresses into CRPC [21,22]. Therefore, we further explored the function of TR3 in cell proliferation and mobility, which are the downstream cell behaviors of AR signaling. Overexpression of TR3 enhanced the proliferation of androgen-independent CWR22rv cells (Fig. 6C, left) as well as androgen-sensitive LNCaP cells under androgen-, outlaw IL6-, and forskolin-stimulated conditions (Fig. S5A). In contrast, TR3 silencing signi cantly inhibited the proliferation of these prostate cancer cells (CWR22rv and LNCaP) (Fig. 6C, right and Fig. S5B), which was con rmed using viability assays of TR3-knockdown CWR22rv cells (Fig. 6D). TR3 also affected cell mobility, with TR3 overexpression markedly increasing CWR22rv cell migration and invasion (Fig. 6E, left  and 6F, top), and TR3 silencing resulting in a decrease in migration and invasion (Fig. 6E, right and 6F,  bottom). These results suggest that TR3 increases cell proliferation and mobility of prostate cancer cells.
The effect of TR3 on prostate tumorigenesis was investigated using the inducible TR3-overexpressing (TR3) or control (EV) CWR22rv xenograft mouse models (Fig. S5C-D). We observed that the average volume of TR3 tumors was approximately 10-fold larger than that of EV tumors after ~ 7-week treatment with doxycycline ( Fig. 6G-H). The average weight of tumors from TR3 mice was also approximately 10fold greater than that of tumors from EV mice, although there were no signi cant differences in body weight (Fig. 6I-J). These results suggest the pivotal role of TR3 in in vivo tumorigenesis of advanced CRPC cells.

Discussion
Constitutive transcriptional activity of AR-Vs is important for prostate cancer progression. However, the molecular mechanisms underlying the production of AR-Vs remain unclear. In the present study, for the rst time, we found a novel function of TR3, which controls AR splicing events in prostate cancer cells. Overexpression of TR3 upregulated the expression of mRNAs and proteins of smaller AR-Vs. Consistent with this nding, TR3 overexpression altered the splicing junction rates of AR pre-mRNA, resulting in the enhancement of mRNA levels of AR-Vs, such as AR45, AR-V7, and AR-V1, in advanced CRPC cells (CWR22rv and C4-2). We also demonstrated that TR3 overexpression alters the expression of certain splicing factors and that TR3 interacts with several known AR splicing factors (hnRNP A2B1 and hnRNP E1/E2) as well as the intron regions in AR pre-mRNA. Interestingly, TR3 functions in RNA splicing process while also regulating gene expression as a transcription factor. The association of TR3 with splicing factors and AR pre-mRNA might control the recruitment of spliceosomal complex components or identi cation of intron/exon boundaries.
Many splicing factors, including hnRNP A family (A0, A1/A1P10, and A2/B1) [7], hnRNP E1/E2 and HUR [6], proline-rich splicing factor PSF/SFPQ [23,24], and U2AF1/U2AF65 [3], are involved in the splicing of AR-V7. Among them, PSF regulates AR splicing events by interacting with spliceosomal complex components as an integrator [23,24]. It binds to the intronic region of AR transcripts and coordinates the complex formation of the spliceosome to promote the production of AR-V7 variant. Similar to PSF, TR3 directly or indirectly binds to the intronic region of AR transcripts at Reg4 and Reg9 and interacts with spliceosomal complex components, such as hnRNP A2B1 and hnRNP E1/E2. Many known RBPs contain the RNA recognition motif (RRM), KH, serine-rich, or poly-proline domains. Although the TR3 amino acid sequence does not contain RRM or KH domains, it has a putative poly-proline rich motif that shows high homology with that of PSF. These similarities suggest that TR3 may control AR splicing events as PSF does in prostate cancer cells.
The coupling of transcription and splicing processes is well-documented [15][16][17]. Transcription factors bound to the promoter or enhancers are likely involved in the recruitment of splicing factors to the site of transcription [15]. In addition, the C-terminal domain of RNA polymerase II (RNA pol II CTD) binds to some processing factors in the spliceosomal complexes and supports splicing factor recruitment to splice sites as shown with PSF and U2AF65-Prp19 complex [3,[23][24][25]. Interestingly, TR3 also interacts with RNA pol II CTD [19]. Therefore, TR3 overexpression may lead to the formation of alternative spliceosomal complexes by recruiting splicing factors to unusual splice sites through its interaction with RNA pol II CTD in AR pre-mRNAs, thus promoting the production of AR-Vs. In addition, TR3 bound to the intronic regions of the AR gene may directly recruit some splicing factors, such as hnRNP A2B1 and hnRNP E1/E2, through protein-protein interactions. Otherwise, it may indirectly change the local chromatin structure, thus inhibiting the splicing between exon 3 and exon 4 of AR-FL, which eventually promotes the production of small transcripts, such as AR-V7 [26].
TR3 overexpression causes immune de ciency and activates immune tolerance, which may create a potent immune suppressive tumor microenvironment during CRPC progression [27]. In the present study, RNA-Seq analysis revealed that TR3 overexpression signi cantly enhanced the expression of immune tolerance genes (Fig. S6A) while decreasing the expression of immune defense genes, particularly in ammatory genes ( Fig. S6B-C), to allow tumor cell survival and escape from immune attack. Additionally, mature myeloid-derived suppressor cells (PMN-MDSCs) are important in CRPC progression, and the IL23-ROR -STAT3 axis signaling is activated when the proportions of circulating PMN-MDSCs are increased during CRPC progression [27]. Notably, the IL23-ROR -STAT3 signaling axis was upregulated when TR3 was overexpressed in CWR22rv cells (Fig. S6D-F). Collectively, these results suggest that TR3 promotes the susceptibility of prostate tumors to escape from immune checkpoints.

Conclusions
TR3 controls AR expression, splicing events, and androgen-independent function, and it likely alters immune responses to facilitate prostate tumorigenesis, suggesting that TR3 is a pivotal factor in prostate cancer progression. Therefore, TR3 could be an alternative therapeutic target for the treatment of advanced prostate cancer.

Not applicable
Available data and materials The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Competing interests
The authors declare that they have no competing interests.  AdCtrl-infected CWR22rv cells, which were treated with 10 nM DHT, 10 µM MDV-3100 or vehicle only. GFP was used as a control for the e ciency of adenovirus infection. qPCR analysis showed alterations in the level of AR mRNAs, which contain exon 1-exon 2 (AR) and exon 3-CE3 (AR-V7) in CWR22rv cells transfected with FLAG-TR3 or empty vector (EV) (E). F-G, Silencing of TR3 expression or inhibiting of TR3 function signi cantly decreases the expression levels of AR and AR-Vs. Western blot analysis showing AR and AR-V protein levels in CWR22rv cells silenced with siTR3 (F). qPCR analysis showing AR, AR-Vs and TR3 mRNA levels in CWR22rv cells transfected with siTR3 or siCtrl (G). Data are shown as mean ± SEM. *, p<0.05; **, p<0.01; ***, p<0.001; one-way ANOVA analysis with Tukey's post hoc test.

Figure 2
TR3 regulates the expression of AR splicing variants, altering splice junction rates between exons. A-B, TR3 binds to the promoter and intron regions of AR gene. Schematic presentation of putative TR3 binding sites (P, A, B, and C) in AR gene (A). Recruitment of TR3 protein to putative TR3 binding sites within the promoter and intron regions of AR gene was determined via ChIP assays using anti-TR3 antibody (B). CWR22rv cells were infected with AdTR3 or AdCtrl. Changes in TR3 enrichment at putative TR3 binding sites was examined using PCR. The loading control (Ctrl) was β-actin. C-D, TR3 overexpression alters the splice junction rate between exons of AR pre-mRNA. CWR22rv cells were infected with AdTR3 or AdCtrl, and the samples were analyzed using Next-Generation Sequencing (NGS) high-throughput RNA-Seq analysis. Splice junction rate between exons within the pre-mRNA of AR-FL (AR8) was analyzed using Integrative Genomic Viewer (IGV) analysis (C). RT-PCR analysis showed alterations of the splice junction rate between AR-V exons, particularly exon 1by/exon 2 and exon 3/CE3 (D). C, TR3 physically interacts with splicing factors involved in the splicing of AR pre-mRNA.
Schematic presentation of intronic regions marked Reg1-9 in AR pre-mRNA, which were selected for CLIP Figure 5 TR3 positively regulates the transactivation of ARs in both androgen-dependent and androgenindependent manners. TR3 enhances the androgen-dependent and -independent transactivation of ARs.
A-D, TR3 overexpression signi cantly enhances the transactivation of ARs. Both PPC1 cells overexpressed with AR-FL (A), AR-NTD (B), or AR-V7 (C), and LNCaP cells (D) were transiently transfected with FLAG-TR3 or empty vector (EV) along with pARE2-TATA-luc and treated with 1 nM DHT or vehicle. E-F, TR3-mediated enhancement of the transactivation of ARs was impaired by treatment with TR3 antagonist (DIM-C-pPhOH). PPC1 cells were co-transfected with AR-FL (E) or AR-V7 (F) together with FLAG-TR3 or EV along with pARE2-TATA-luc and treated with 20 µM or vehicle in the presence or absence of 10 nM DHT. G, TR3 increases androgen-independent nuclear translocation of ARs. HEK 293T cells were transfected with GFP-AR-FL or GFP-AR-NTD together with TR3 or EV in the absence of androgen. Subcellular localizations of ARs and TR3 were detected as green uorescent protein (GFP) and red Alexa Fluor568 signal, respectively. Nuclei were stained blue with TOPRO-3. Arrowheads indicate strong signals of nuclear-localized AR proteins in TR3-coexpressing cells. Images were acquired using a confocal microscope. Scale bars, 25 µm. H. TR3 overexpression enhances coactivator recruitment to AR. PPC1 cells were co-transfected with AR, SRC-2, and FLAG-TR3 or empty vector along with pARE2-TATA-luc and incubated with or without 1 nM DHT. Data are shown as mean ± SEM. *, p<0.001; ns, not signi cant; oneway ANOVA with Tukey's post hoc test.