TRIM59 is Suppressed by Androgen Receptor and Acts to Promote Lineage Plasticity and Neuroendocrine Differentiation in Prostate Cancer

Background: The incidence of treatment-induced neuroendocrine prostate cancer (t-NEPC) has been greatly increasing after the usage of second-generation androgen receptor (AR) pathway inhibitors (ARPIs). Neuroendocrine differentiation (NED) is closely associated with ARPI treatment failure and poor prognosis in prostate cancer (PCa) patients. However, the molecular mechanisms of NED are not fully understood. Methods: TRIM59 expression was evaluated in PCa samples from patients at rst diagnosis or at relapse stage post ARPI treatment by immunohistochemistry; in vitro effects of TRIM59 were determined by cell proliferation, sphere formation and cell migration assays; while in vivo analysis was performed using subcutaneous tumor model. Western blot, qPCR assay, dual luciferase assessment, chromatin immunoprecipitation and RNA sequencing were applied for mechanistic exploration. Results: Here we report that upregulation of TRIM59, a TRIM family protein, is strongly correlated with ARPI treatment mediated NED and shorter patient survival in PCas. AR binds to TRIM59 promoter and represses its transcription. ARPI treatment leads to a reversal of repressive epigenetic modications on TRIM59 gene and the transcriptional restraint on TRIM59 by AR. Upregulated TRIM59 then drives the NED of PCa by enhancing the degradation of RB1 and P53 and upregulating downstream lineage plasticity-promoting transcription factor SOX2. Conclusion: Altogether, TRIM59 is negatively regulated by AR and acts as a key driver for NED in PCas. Our study provides a novel prognostic marker for PCas and shed new light on the molecular pathogenesis of t-NEPC, a deadly variant of PCa. detect the neuroendocrine markers The diagnosis of NEPC based on both morphology and elevated neuroendocrine markers androgen-dependent PCas at the rst into NEPCs after a of treatment. from 10 pairs of sequential pre- and post-ARPI treatment for AR, TRIM59, CgA and SYP. All the other for TRIM59. Cell TRIM59-Flag vector and 2µg P53 or RB1 expression vector, while 10 6 LNCaP cells was transfected with 5µg TRIM59-Flag or empty vector. The samples were lysed at 24 hours, 48 hours and 60 hours after transfection. For MG132 treatment, LNCaP transfected with 10µg TRIM59-Flag or empty vector were cultured in medium with 10µg/ml cycloheximide (Sigma-Aldrich) at 36 hours post-transfection. genetic alterations. In this study, we nd that TRIM59 promotes the degradation of RB1 and P53 in PCa cells. Moreover, the NED-inducing effect of TRIM59 overexpression is dependent on RB1 and P53. We further demonstrate that knockdown of the reprograming factor SOX2 signicantly overrides the NED phenotype in TRIM59 overexpressed PCa cells. Therefore, TRIM59 drives the NED of prostate cancer by enhancing the degradation of P53 and RB1 and subsequently upregulating downstream lineage plasticity promoting transcription factor SOX2. This provides an explanation of downregulation in RB1 and P53 proteins in NEPCs without RB1/TP53 genetic alterations progression


Background
Androgen receptor (AR) pathway inhibitors (ARPIs), such as abiraterone, enzalutamide, etc, have become the rst-line treatment of metastatic prostate cancers (PCas) [1,2]. However, with the application of ARPIs, especially the second-generation ARPIs, the prevalence of a highly aggressive castration resistant prostate cancer (CRPC) subtype named neuroendocrine PCa (NEPC) with small cell carcinoma-like morphology, including indistinct cell borders, high nuclei-cytoplasm ratio, hyperchromatic nuclei and elevated neuroendocrine-related biomarker expression, has been steadily increasing. NEPC now accounts for approximately 20% in PCa patients who develop resistance to ARPIs, which stands in great contrast to only 1% in untreated newly diagnosed PCa patients [3]. NEPC patients are now commonly treated with platinum-based chemotherapy, but the disease prognosis is highly unfavorable [4]. Our incomplete understanding of the molecular drivers and regulators in neuroendocrine differentiation (NED) hinders the development of better treatment approaches.
Most of PCas at rst diagnosis are histologically adenocarcinoma with a luminal-cell phenotype [5,6]. Although the mechanisms are still elusive, it is suggested that lineage plasticity and luminal-to-neuroendocrine transdifferentiation contribute to the emergence of treatmentinduced NEPC [7,8]. Several signaling molecules have been shown to participate in this process. These include concurrent loss or mutation of RB1 and TP53 [3,[9][10][11], increased lineage plasticity associated with the reprogramming factor SOX2 [12], upregulation of neuronal transcription factor SOX11 [13], BRN2 [14], oncogene AURKA [15] and , splicing factor SRRM4 [21], and epigenetic modi er EZH2[18] and etc. It is believed that the use of highly potent ARPIs and the loss of AR signaling are initial triggers of NED [3,9]. However,how AR activity loss modulates the molecular events leading to NED remains to be further explored.
TRIM59, a member of the tripartite motif (TRIM)-containing protein family, is comprised of one Ring nger region, one B-box region, two crimped helices and one transmembrane domain [22]. Due to its Ring nger domain, TRIM59 is mainly considered to be a putative E3 ubiquitin ligase [22,23]. Our previous study has shown that TRIM59 played an oncogenic role in gastric carcinogenesis [24]. Moreover, TRIM59 upregulation has been shown to be a driver of prostatic tumorigenesis in a genetically engineered murine model, in which Trim59 overexpression was controlled by a prostate speci c PSP94 gene promoter [25]. Intriguingly, the prostate tumor tissue from PSP94-Trim59 mice displayed features of NEPC and expressed the neuroendocrine marker gene Chga [25]. Therefore, TRIM59 may act as an essential gene in PCa progression and NED. However, the role and molecular mechanism of TRIM59 in castration resistance of human PCa and NED has not been elucidated.
Here we nd a prominent elevation of TRIM59 protein levels in human PCa samples after the ARPI treatment. TRIM59 upregulation is strongly correlated with NED in PCas. We further demonstrate that TRIM59 is transcriptionally repressed by AR. TRIM59 promotes degradation of RB1 and P53 and then upregulates downstream SOX2, therefore potentiates the lineage plasticity and drives the NED of PCa.
Patients Patient sample and medical information collection protocols used in this study is approved by the Medical Ethical Committee at Ren Ji Hospital, School of Medicine, Shanghai Jiao Tong University. Patients diagnosed with NEPC (n = 21) or CRPC-adenocarcinoma (n = 39) from January, 2010 to December, 2019 were enrolled in the NEPC/CRPC-adenocarcinoma cohort. For pathological classi cation, fresh samples were collected and para n embedded. 5µm thickness sections were prepared for H&E staining to observe cell morphology. When characteristic NEPC morphology as previously described in the manuscript was identi ed, immunohistochemical staining was further carried out to detect the neuroendocrine markers expression. The diagnosis of NEPC was based on both tumor morphology and elevated neuroendocrine markers by certi ed pathologists in Department of Pathology, Ren Ji Hospital. In the Renji t-NEPC cohort, 10 androgendependent PCas at the rst diagnosis eventually develop into NEPCs after a series of ARPI treatment. Para n slices from 10 pairs of sequential pre-and post-ARPI treatment samples were stained for AR, TRIM59, CgA and SYP. All the other samples were stained for TRIM59. IHC were conducted as described previously [26]. Antibodies used for IHC were listed as follows: TRIM59 (ab166793, Abcam), AR (5153, Cell Signaling Technologies), SYP (36406, Cell Signaling Technologies), CgA (ab15160, Abcam).
In vitro sphere forming and migration assay The single cell suspension at 10 4 cells/ml were seeded in 6-well low-attachment plates with DMEM-F12 containing 10µM enzalutamide. Spheroid (> 100µm) were numerated under a light microscope (Zeiss).
For cell migration assay, 2x10 4 cells were plated in the upper chamber with serum-free medium (DMEM for PC3 and DU145, RPMI-1640 for LNCaP), separated from FBS-containing medium in the lower chamber. After 24-hour incubation at 37°C in 5% CO2, non-migrated cells were removed. The number of migrated cells were stained with crystal violet (Beyotime, C0121) and then counted using microscope in 5 randomly chosen elds.

RNA-seq and RT-PCR
Total RNAs were extracted from cells using TRIZOL following manufacturer's instruction. Then mRNA was reverse-transcribed to barcoded cDNA fragments using oligo-dT primer with adapter attached. Barcoded cDNA libraries were sequenced using the Illumina HiSeq 4000 PE150 platform. RNA-seq reads were aligned to the reference genome (GRCh37/hg19) using HISAT2 after quality assessment. StringTie was used to assemble and quantitate the abundance of transcripts. DESeq2 was used to perform differential gene expression analysis to the normalized data.
For RT-PCR, reverse transcription was performed using HiScript III RT SuperMix for qPCR (R323-01, Vazyme) according to the manufacturer's instructions. Real-time PCR was conducted in quadruplicates using ChamQ Universal SYBR qPCR Master Mix (Q711-02, Vazyme). Primers used in this study was in the supplementary Table 3.

Immunoblotting
Whole cell lysates were prepared in Cell lysis Buffer (C500035, Sangon). About 10 7 cells were lysed with 1ml lysis buffer containing protease inhibitor cocktail (MCE, HY-K0010). SDS-PAGE and immunoblotting experiment was conducted following conventional methods. Antibodies

Lentiviral transfection
The lentiviral particle was produced in HEK293T cells after co-transfection with lentiviral plasmid and packaging vectors (psPAX2 and PMD2.G) at a 6:1:7 ratio using Polyetherimide in 10-cm dishes. The lentivirus was harvested 36-60 hours later after replacement of media at 12 hours post-transfection. To generate stable transgenic lines, LNCaP and LAPC4 were infected at a 50% con uency with the lentiviral particle and 8ug/ml polybrene (Sigma-Aldrich). The positively transfected cells were selected and enriched by applying puromycin (5µg/ml) in the culture medium for 2 weeks.
Luciferase assay TRIM59-promoter-luciferase lentivirus were produced and used for transfecting LNCaP cells. Stably transfected LNCaP cells were seeded in 12-well plate and treated with vehicle (DMSO), 2nM R1881 or 20µM Enzalutamide. Dual-luciferase Reporter Assay System (Promega, E2920) was used to detect re y luciferase activity 24hrs post treatment according to the manufacturer's protocol. Fire y luciferase activity was normalized to cell numbers.

Tumor xenograft experiment
All the mice were kept in the pathogen-free facility at Ren Ji Hospital. 5x10 6 cells were suspended in 100µl 50% Matrigel and injected into the right ank of nude mice. When testing the capacity of in-vivo castration resistance, nude mice were castrated 2 weeks prior to subcutaneous tumor cell implantation. Tumor were harvested, imaged and weighed after sacri ce of the experimental mice.
Protein stability assay LNCaP/AR or 293T cells were used for performed protein stablility assays as preciously described [24]. For dose dependent protein stability assay, LNCaP cells was transfected with 1µg, 5µg, 10µg TRIM59-Flag or empty vector. All the samples were lysed to detect RB1 and P53 protein levels at 48 hours after transfection. For time dependent protein stability assay, 10 6 293T cells were co-transfected with 2µg TRIM59-Flag vector and 2µg P53 or RB1 expression vector, while 10 6 LNCaP cells was transfected with 5µg TRIM59-Flag or empty vector. The samples were lysed at 24 hours, 48 hours and 60 hours after transfection. For MG132 treatment, LNCaP transfected with 10µg TRIM59-Flag or empty vector were cultured in medium with 10µg/ml cycloheximide (Sigma-Aldrich) at 36 hours post-transfection.

Statistical analysis
All statistical analysis was performed using the GraphPad 7.0 software. Statistical analysis method used in this study was provided in corresponding gure legend. P < 0.05 is considered signi cant.

TRIM59 is highly expressed in castration resistant NEPCs
To determine the signi cance of TRIM59 in NEPC, we rst analyzed a published NEPC-enriched CRPC dataset (the Beltran-2016 cohort) [9] containing 15 castration resistant NEPC samples and 34 CRPC with adenocarcinoma phenotype (CRPC-AD) samples (Fig. 1A). The NEPCs showed signi cantly higher mRNA levels of TRIM59 compared to the CRPC-AD subgroup (Fig. 1A, 1B). In addition, the high expression level of TRIM59 was signi cantly associated with worse survival in PCa patients based on data analysis of The Human Protein Atlas prostate cancer dataset (Fig. 1C). To further examine the TRIM59 protein expression in clinical samples, we examined our in-house cohort (the Renji cohort, n = 60) of CRPC specimens including 39 CRPCs with adenocarcinoma phenotype (CRPC-AD) and 21 NEPCs (Fig. 1D). Statistical analysis of the IHC staining result showed that TRIM59 was signi cantly upregulated in NEPCs than CRPC-AD (Fig. 1E). Consistent with the previous results in The Human Protein Atlas prostate cancer dataset (Fig. 1C), we observed a signi cantly shorter overall survival (Fig. 1F) and faster progression to abiraterone resistance in high TRIM59 expression group than the low expression group in our cohort (Fig. 1G). Interestingly, in addition to the elevated TRIM59 expression in NEPC, we also found a signi cant increase of the TRIM59 mRNA level in small cell lung cancers, which share pathohistological similarities with NEPCs, from data analysis at The Cancer Genome Atlas (TCGA) and the SCLC cohort (supplementary Fig. 1A-1C) [29]. Those data suggested a strong association of TRIM59 with neuroendocrine cancers.
TRIM59 is negatively correlated with the AR signaling in human CRPCs Since NED often occurs after potent ARPI treatment, we focused on the relationship between TRIM59 and AR. We rst analyzed the AR and TRIM59 mRNA level in the Beltran-2016 dataset and found an inverse correlation between AR and TRIM59 transcripts ( Fig. 2A, 2B). Furthermore, we uncovered a signi cantly negative correlation between expression of AR signature genes (referred as AR score here) in the Beltran-2016 cohort (Fig. 2C). In another published PCa RNA-seq dataset (TCGA rehose), the TRIM59 transcript was also negatively correlated with the expression level of classic AR downstream genes including KLK3, KLK2 and PART1 (supplementary Fig. 2). To further con rm this nding, we performed IHC analysis of TRIM59 and AR on human CRPC samples (n = 44). In concordance with the Beltran-2016 dataset analysis, our IHC experiments revealed that AR-positive CRPC samples exhibited low or negative staining for TRIM59, while ARnegative CRPC samples showed high or medium TRIM59 staining (Fig. 2D, 2E). In addition, we assessed TRIM59, AR and SYP protein levels by IHC staining in PCa samples from 10 treatment-induced NEPC (t-NEPC) patients, all of which included hormone-naïve PCa samples before ARPI treatment and t-NEPC samples after ARPI treatment from the same patient. Among them, 8 (80%) patients showed a marked upregulation of TRIM59 expression when the disease progressed to t-NEPC after ARPI treatment compared to treatment-naive primary PCa (Fig. 2G, 2H). One patient showed no change and one patient sample exhibited a slight decline of TRIM59 expression along the t-NEPC development. Collectively, our ndings suggested a robust increase in the TRIM59 expression upon prolonged ARPI treatment in PCa patients.

AR signaling suppresses the transcription of TRIM59 in prostate cancer cells
We further analyzed a published dataset (GSE71797), wherein LNCaP and VCaP cells were stimulated with R1881. Consistently, we found a reduced expression of TRIM59 when AR was activated by R1881 (Fig. 3A, 3B). The other downregulated genes included previously known AR repressed genes, such as DDC, OPRK1, NOV and SERPIN1 (Fig. 3A, 3B). We then examined the role of AR signaling in the regulation of TRIM59 using PCa cell lines. As shown in Fig. 3C, stimulation of LNCaP cells with synthetic androgen R1881 (2nM) led to an activation of AR signaling evidenced by the upregulation of AR target genes KLK3 and NKX3.1. AR activation resulted in a signi cant decrease in the expression of TRIM59 both at mRNA and protein levels (Fig. 3C). Treatment with the ARPI, Enzalutamide (20µM), remarkably increased the TRIM59 transcription and protein levels, accompanied with reduced expression of androgen-driven genes KLK3 and NKX3.1 (Fig. 3C). Moreover, we knocked down or overexpressed AR in LNCaP and LAPC4 cells and examined changes in TRIM59 expression. Consistent with the AR antagonist and agonist treatment experiment results, LNCaP and LAPC4 cells exhibited a signi cant increase in the expression of TRIM59 after AR knockdown, and a robust decrease in the TRIM59 expression when AR was overexpressed (Fig. 3D-F). To test whether the elevated TRIM59 expression by AR signaling blockage was a transient phenomenon, LNCaP cells were treated with 10µM enzalutamide for 3 months to establish a long-term enzalutamide-treated LNCaP cell line (Fig. 3G). We observed a persistent higher expression of TRIM59 in the enzalutamide-resistant LNCaP cells compared to control cells (Fig. 3H, 3I).
To explore the mechanism underlying the negative regulation of TRIM59 by AR signaling, we cloned the promoter region of TRIM59 (-1462bp + 1032bp) and utilized a promoter driven luciferase assay to examine the impact of AR signaling on transcription of TRIM59. As shown in Fig. 3J, a signi cant decrease in the luciferase activity was observed in R1881-stimulated LNCaP cells. We then analyzed the promoter region of TRIM59 in search of transcriptional factor binding sites using JASPAR, an online transcription factor binding pro le database. Interestingly, 3 putative androgen response elements (AREs) which were located in -861bp~-845bp, -589bp~-573bp, and + 550bp ~ + 566bp were identi ed in the TRIM59 promoter (Fig. 3K). To validate those predicted ARE sites, we performed ChIP-qPCR for AR on LNCaP cells treated with 2nM R1881 or 20µM enzalutamide. A signi cant AR enrichment at the three binding sites (ARE-1, ARE-2 and ARE-3) in TRIM59 promoter was observed in R1881-stimulated LNCaP cells with respect to DMSO control (Fig. 3L). AR recruitment at the three AREs was reversed in the presence of enzalutamide (Fig. 3L). To further investigate how the transcription of TRIM59 was repressed by AR, we performed ChIP-qPCR for gene-expression-related epigenetic markers including H3K4me3, H3K27ac and H3K9me3. Among all the epigenetic modi cations we examined, the most differentially changed epigenetic marker along with AR activation or suppression on TRIM59 promoter AREs was H3K9me3, a repressive histone modi cation (supplementary Fig. 3A). Additionally, knockdown of EHMT1/EHMT2, histone methyltransferases to speci cally methylate H3K9, resulted in a signi cant upregulation in TRIM59 transcription (supplementary Fig. 3B). Together these ndings support the notion that AR functioned as a direct transcriptional repressor of TRIM59. Attenuation of the AR signaling via potent ARPIs, such as enzalutamide, relieved the transcriptional repression of TRIM59 by AR.

TRIM59 signi cantly promotes development of castration resistance and NED in PCas
We then examined the functional signi cance of TRIM59 in PCas. As shown in supplementary Fig. 4, overexpression of TRIM59 in hormonaltherapy sensitive PCa cell lines LNCaP and LAPC4 resulted in a signi cant increase in tumor cell proliferation, migration and sphere forming capacity in vitro and an accelerated xenograft tumor growth in vivo. Importantly, we observed development of castration resistance in TRIM59-overexpressing LNCaP or LAPC4 cell lines, manifested as an enhanced cell proliferation rate and sphere forming capacity in ARPIcontaining medium compared to control cells (Fig. 4A-C). To further test the role of TRIM59 in the development of castration resistance of PCas in vivo, we implanted TRIM59-overexpressing or control LAPC4 cells subcutaneously to the castrated nude mice (Fig. 4D). Consistent with the in vitro experimental results, a notably castration independent xenograft growth was found in the TRIM59-overexpressing group, which stood in great contrast to the rare forming of xenograft tumor in control cell implanted group (Fig. 4E). Immuno uorescent staining of LAPC4 xenografts revealed prominent expression of neuroendocrine markers (SYP, NCAM-1 and CgA) in the TRIM59-overexpressing group but not in the control group (Fig. 4F). Knockdown of TRIM59 in CRPC cell lines PC3 and DU145, on the other hand, signi cantly inhibited the proliferation, sphere formation, migration in vitro and in vivo growth of tumor cells (Fig. 5). These data together suggested that TRIM59 functioned as an important factor in promoting castration resistance and NED in PCa.
We next conducted RNA-sequencing to compare the gene expression pro le between LNCaP-TRIM59 and LNCaP control cells. In compliance with the previous results, TRIM59 overexpression resulted in increased transcription of neuroendocrine marker SYP, ENO2, and AURKA, along with a dramatic decline in AR signaling signature genes (Fig. 6A). Those gene expression alterations were further validated at protein levels by immunoblotting (Fig. 6B). Moreover, Gene Set Enrichment Analysis (GSEA) showed that luminal signature genes were markedly downregulated, while the small cell lung cancer (SCLC) gene set were signi cantly enriched in the LNCaP-TRIM59 cell line (Fig. 6C, 6D).
Nervous system development related genes were preferentially expressed in LNCaP-TRIM59 revealed by GO pathway enrichment analysis (Fig. 6E, 6F). Interestingly, we found that AR signaling signature genes were signi cantly downregulated in TRIM59 overexpressing LNCaP cells, which was revealed both by q-PCR and GSEA analysis of RNA-seq data (supplementary Fig. 5), suggesting a negative feedback between TRIM59 and AR signaling in PCa cells. To determine whether TRIM59 was necessary in the NED process, we next constructed TRIM59 knockdown in LNCaP cells and subjected the LNCaP-shTRIM59 or control cells to ARPI-containing medium. shRNA-mediated silencing of TRIM59 prevented the enzalutamide-induced upregulation of NE markers in LNCaP cells (Fig. 6G, 6H). Together these ndings supported an essential role of TRIM59 in PCa cellular plasticity and NED.

TRIM59 drives the NED of prostate cancer by promoting degradation of RB1 and P53
To further elucidate the mechanism of TRIM59 in NED, we carefully analyzed the expression alterations of previously reported neuroendocrine driver genes in LNCaP-TRIM59 RNA-seq data (Fig. 6A). Among them, the lineage plasticity gene SOX2 was the mostly differentially expressed one in LNCaP-TRIM59 cells. Interestingly, GSEA analysis on the LNCaP-TRIM59 RNA-seq data showed that the RB1 and P53 target geneset were signi cantly underexpressed in LNCaP-TRIM59 cells (Fig. 7A), while SOX2 downstream targets were markedly enriched (Fig. 7B). It has been demonstrated that concurrent loss or mutation of RB1 and TP53 were the most common genetic alterations in NEPC [9]. SOX2 was found as a downstream pivotal factor of RB1 and TP53 loss in promoting lineage plasticity and NED [12]. In agreement with our GSEA analysis, we found declines in both RB1 and P53 protein levels in LNCaP and LAPC4 cell lines overexpressing TRIM59 (Fig. 7C). Meanwhile the RB1 and P53 downregulation was accompanied with a rising level of the SOX2 protein. We did not detect overt changes in RB1 and TP53 mRNA levels upon TRIM59 overexpression, suggesting that the downregulation of RB1 and P53 was at post-translational level (Fig. 7D). We then investigated the impact of ectopic expression of TRIM59 on RB1 and P53 protein stability. As shown in Fig. 7E, TRIM59 overexpression led to an accelerated downregulation of protein abundance of RB1 and P53 in 293T cells. Similarly, we found that TRIM59 promoted reduction of endogenous RB1 and P53 levels in LNCaP cells in a time-dependent and dose-dependent manner (Fig. 7F, 7G). We further performed a time-dependent protein stability assay with the addition of 10µg/ml protein synthesis inhibitor CHX with or without 10µM proteasome inhibitor MG132. As shown in Fig. 7H, TRIM59 signi cantly reduced the half-life of RB1 and P53 proteins, which could be reversed by addition of MG132. These data suggested that TRIM59 promoted RB1 and P53 degradation via the proteasome pathway.
To further analyze whether TRIM59 induced NED was acted via downregulating RB1 and P53, we used two cell line models with NED phenotype, that were LNCaP-ENZA-R cells (Fig. 3G) and LNCaP cells stably transfected with RB1 and TP53 shRNA as previously reported [11]. As shown in Fig. 7I, we detected a prominent decline of neuroendocrine markers NSE and SYP, as well as SOX2 in LNCaP-ENZA-R cells upon TRIM59 silencing. In contrast, no signi cant alterations in expression levels of NSE, SYP and SOX2 was found in a shRB1/shTP53 stably transfected LNCaP cells after TRIM59 silencing (Fig. 7J). These data suggested that the neuroendocrine phenotype induced by TRIM59 is dependent on RB1 and P53. To further test the role of SOX2 in TRIM59-induced PCa NED, we conducted shRNA-mediated SOX2 silencing in LNCaP-TRIM59 cells. Importantly, we found that SOX2 knockdown led to a signi cant reduction of NE markers (NCAM-1, SYP, CgA) in LNCaP-TRIM59 cells (Fig. 7K). Therefore, TRIM59 was an AR-suppressed key molecular driver for lineage plasticity and neuroendocrine differentiation in prostate cancer, which acted through enhanced RB1/P53 degradation and upregulation of SOX2 (Fig. 8).

Discussion
In the current study, analysis from published DNA and RNA sequencing data of prostate adenocarcinomas and NEPCs, together with IHC staining of tumor tissues from our large in-house CRPC cohort consistently show a marked elevation of TRIM59 expression in t-NEPC. We further demonstrate that upregulation of TRIM59 promotes the lineage transdifferention of PCa cells with a luminal phenotype to the t-NEPC phenotype characterized with underexpression of luminal genes and increased transcription of neuroendocrine maker genes. Notably, this phenotype change is associated with enhanced resistance to the APRI treatment in vitro and in vivo. TRIM59 knockdown signi cantly abrogates NED in PCa cells treated with ARPIs. Altogether, we uncover that TRIM59, an oncogenic TRIM family protein, as an essential molecular driver for the highly lethal t-NEPC formation.
Except for a small proportion of de-novo NEPC, most of NEPCs are induced by potent ARPI therapies. Tracking the change of gene expression characteristics or DNA mutations of the PCa biopsy samples in the same patient before and after ARPI treatment can provide valuable information to elucidate the molecular basis of t-NEPC progression. However, the samples from the same patient are often di cult to obtain.
We have carefully collected 10 pairs of before-and-after samples from prostate adenocarcinoma patients who received ARPI therapy and eventually develop t-NEPC in the past 10 years at Ren Ji Hospital, one of the largest PCa treatment center in China. IHC staining of these sequential samples provides compelling evidence that TRIM59 upregulation is frequently found in ARPI resistant NEPC samples compared to hormone naïve PCa at rst diagnosis. Those data strongly support that the expression of the NED driver gene TRIM59 is induced upon ARPI treatment.
Reports over past few years have highlighted how ARPI treatment [3] or loss of AR signaling promotes the NED of PCa cells. A few genes that have been reported to associate with an NEPC phenotype, such as BRN2 [14] are controlled by AR. Yet how AR activity loss modulates the t-NEPC-inducing molecular events is not fully understood. Our current study demonstrates that AR directly binds to TRIM59 promoter and represses its transcription. ARPI treatment leads to a reversal of the suppressive histone modi cations on TRIM59 gene and a release of transcriptional restraint on TRIM59 by AR. We further uncover that the repressive epigenetic modi cation H3K9me3 in the TRIM59 promoter is dynamically changed along with AR activation or suppression. Knockdown of H3K9me3 methyltransferases releases the transcription repression of TRIM59. Therefore, an androgen-dependent epigenetic program contributes to the regulation of TRIM59 transcription. In addition, we detect a negative feedback between TRIM59 and AR signaling in PCa cells, based on the observation that enforced expression of TRIM59 leads to a suppression of AR target genes. The present work provides insights into the reciprocal negative regulation between AR signaling and TRIM59 in NED.
The concurrent loss or mutations of RB1 and TP53 is reported to be the most frequent genetic alterations in NEPC [3,9,11]. Mechanistic study has shown that RB1 and TP53 loss mediated NED of PCa is dependent on the reprograming transcriptional factor SOX2 [12]. In addition, an independent group also reported that RB1 protein loss was found in 90% of NEPC cases, whereas 23% cases harbored homozygous loss of RB1 [10]. This observation suggests that alternative mechanisms are taken to downregulate RB1 and P53 in NEPCs carrying no RB1/TP53 genetic alterations. In this study, we nd that TRIM59 promotes the degradation of RB1 and P53 in PCa cells. Moreover, the NED-inducing effect of TRIM59 overexpression is dependent on RB1 and P53. We further demonstrate that knockdown of the reprograming factor SOX2 signi cantly overrides the NED phenotype in TRIM59 overexpressed PCa cells. Therefore, TRIM59 drives the NED of prostate cancer by enhancing the degradation of P53 and RB1 and subsequently upregulating downstream lineage plasticity promoting transcription factor SOX2. This provides an explanation of downregulation in RB1 and P53 proteins in NEPCs without RB1/TP53 genetic alterations The emergence of t-NEPC is a complicated pathological process. Understanding how AR activity loss and key molecular drivers trigger this pathological lineage transdifferentiation process is essential to overcome resistance of current ARPI therapies and to develop novel treatment approaches. Our data from human PCa patient sample examinations, in vitro and in vivo functional analysis and mechanistic study collectively demonstrate that TRIM59 is an AR suppressed oncogenic protein and acts as an essential driver in the progression treatmentinduced NEPC.

Conclusion
In summary, we identify TRIM59 as an AR-suppressed molecular driver of neuroendocrine differentiation in prostate cancer. Mechanistically, ARPI treatment leads to a reversal of transcriptional restraint on TRIM59 by AR. Upregulated TRIM59 then drives the NED of PCa by enhancing the degradation of RB1 and P53 and upregulating downstream lineage plasticity-promoting transcription factor SOX2. Our study provides a novel prognostic marker for PCas and shed new light on the molecular pathogenesis of t-NEPC, a deadly variant of PCa.

Consent for publication
All subjects have written informed consent.

Availability of data and material
The RNA-sequencing datasets are available upon request.

Competing interests
The authors declare that they have no competing interests.