Loss of Long Noncoding RNA NXTAR in Prostate Cancer Augments Androgen Receptor Expression and Enzalutamide Resistance

Prostate cancer is the second most reported cancer in American men (1). The limited efficacy of androgen deprivation therapies often results in progression of the disease to lethal castration-resistant prostate cancer (CRPC) in 1 to 2 years, with reversion of androgen receptor (AR) activity (2–4). To combat CRPCs, abiraterone, an androgen biosynthesis inhibitor, and enzalutamide, a drug that inhibits AR nuclear translocation, have been widely used; however, the overall disease-free survival for these two AR antagonists as single agents has been modest (5, 6), and their combination has provided little advantage (7). Robust amplification of the AR enhancer region in almost 81% of enzalutamide-treated patients reinforces the notion that AR signaling is the epicenter of drug resistance (8–10). The bypassing of AR signaling by increased expression of glucocorticoid receptor is thought to be one of the early mechanisms of enzalutamide resistance (11). A pluripotency transcription factor, SOX2, was later shown to be required for the lineage plasticity and enzalutamide resistance through the loss of TP53 and RB1 (12). AR regulating its own expression by epigenetic modification (phospho-Tyr88-H4) of its enhancer in an enzalutamide-rich environment has emerged as another mechanism to maintain high AR levels and thus confer resistance (13). In addition, many patients with CRPC develop resistance by expressing an AR splice variant called AR-V7 (14) that lacks the ligand-binding domain, nullifying the effects of these two AR antagonists. Collectively, these data have firmly established that AR is not only important in the pre-CRPC scenario, but also retains a dominant role in the post-CRPC stage, including in the drug-resistant state.

The identification of a large number of noncoding RNAs (ncRNA) has firmly established the much awaited paradigm that the human genome is pervasively transcribed, regardless of the protein coding abilities of the resultant RNAs (15), with ncRNAs comprising almost 90% of the transcribed genome (16). Long ncRNAs (lncRNA), typically >200 nt long, with mostly no evident open reading frames but sometimes with a polyadenylated tail, form a large portion of these ncRNAs (17, 18). The surge in studying these lncRNAs has revealed their important contribution in myriad biological processes, including cell proliferation, drug resistance, metabolism, apoptosis, and senescence (19–23). lncRNAs often participate in these processes by orchestrating enzymatic protein regulation or degradation or by fine tuning different chromatin states, indicating that deregulation of certain lncRNAs could lead to the development of a disease.

Because AR activity is paramount to all stages of prostate cancer, the importance of prostate cancer–specific or prostate cancer–regulating lncRNAs has long been speculated (24). Studies have revealed that successive binding of AR to two overexpressed lncRNAs in prostate cancer, PRNCR1 and PCGEM1, enhances its binding to androgen response elements so as to induce AR-mediated gene expression in androgen-dependent or CRPC tumors (25). ARLNC1 (AR-regulated long noncoding RNA 1) was not only induced by the AR protein, but it also stabilized the AR transcript via RNA–RNA interaction (26). Overall, several studies have revealed the significance of lncRNA in AR signaling (27); however, direct negative regulation of AR and AR-V7 transcription by lncRNA has not yet been reported. Serendipitously, we identified a novel tumor suppressor lncRNA, NXTAR, that not only suppressed AR and AR-V7 transcription, but also when NXTAR expression was reinstated, it inhibited the proliferation of enzalutamide-resistant prostate cancer cells and tumor development in prostate models. Here, we present key data to demonstrate the importance of the NXTAR lncRNA and its potential therapeutic utility using a therapeutic oligonucleotide derived from this lncRNA.

RWPE-1 (normal human prostate cell line; CRL-11609; RRID:CVCL_3791), 22Rv1 (CRL-2505; RRID:CVCL_1045), LNCaP (CRL-1740; RRID:CVCL_1379), VCaP (CRL-2876; RRID:CVCL_2235), and PC3 (CRL-1435; RRID:CVCL_0035) cells were obtained from the ATCC. LAPC4 and C4-2B cells were grown as described earlier (13). Cells were grown in 5% charcoal-stripped FBS (Gibco, catalog no. 12676029) containing medium (CSM), mimicking androgen deprivation, or serum-free medium for experimental purposes. All the cell lines were routinely tested for Mycoplasma (ATCC, catalog no. 30-1012K). siRNAs for AR and ACK1 are as described previously (13). NCOR1 siRNA was purchased from Horizon. pBABE-puro retroviral expression plasmids (kindly provided by Mani Sendurai) were used for NXTAR expression.

Patients provided written informed consent and the study was conducted under Washington University in St. Louis (St. Louis, MO) Institutional Review Board–approved genitourinary banking protocol (HRPO#:201411135). Human prostate tumor and normal samples were obtained from the written consented patients with prostate cancer following radical prostatectomy at the Washington University School of Medicine in St. Louis, Urologic surgery. The specimens were deidentified before processing in the laboratory. The normal (far from the tumor site) and tumor (center core of cancerous lesion) tissues were dissected by a board-certified genitourinary pathologist based on MRI and pathology report of the biopsies from each patient. About 2 mm of the collected tissue was hematoxylin and eosin–stained and reviewed again by pathologist, who assigned a Gleason score. All the tumor samples had Gleason score of 6–9. The specimen were washed in PBS and RNA was prepared (Qiagen).

A set of five DNA oligonucleotide probes complementary to NXTAR RNA was synthesized, with biotin labels at their 3′ end (IDT). Nonspecific oligonucleotide recognizing lacZ RNA was synthesized as controls. The sequences are shown in Supplementary Table S1. The biotinylated oligonucleotides were incubated with lysates (2 hours), followed by incubation with streptavidin magnetic beads overnight. Beads attached to RNAs were divided equally and processed for either RNA isolation or Western blotting. For RNA isolation, beads were washed three times and then separated using proteinase K treatment. RNA was then isolated and purified, and DNA was digested in the process. The real-time quantitative RT-PCR was performed on isolated RNA to check the efficiency of pulldown of NXTAR RNA. The remaining beads were heated at 80°C for 5 minutes in loading buffer, followed by Western blotting. The blots were probed with EZH2 (Cell Signaling Technology, 5246S; RRID:AB_10694683) antibody and Actin (SCBT, sc-47778; RRID:AB_2714189) was used as the loading control.

Biotin-labeled NXTAR-N5 (complementary to the AR promoter) and globin oligos were synthesized (IDT; Supplementary Table S1). Chromatin isolation by RNA purification (ChIRP) were performed using lysate from formaldehyde fixed VCaP cells and oligos that were immobilized to streptavidin beads, as described earlier (28). The amount of the chromatin DNA was determined by real-time PCR. For detection of NXTAR-N5/EZH2 binding, biotin-labeled NXTAR-N5 oligo and control globin oligo were incubated with VCaP cell lysates and were subjected to chromatin pulldown, followed by immunoblotting with EZH2 antibodies.

NXTAR-N5 oligonucleotide was synthesized by IDT, and the oligonucleotide sequence is provided in Supplementary Table S1 and Fig. 7C (* represents phosphorothioate bond modifications to avoid degradation by exonucleases). Equal numbers of VCaP and 22Rv1 cells were transfected with NXTAR-N5 or globin-derived oligonucleotide using X-tremeGENE360 (Sigma-Aldrich, catalog no. 08724121001), and the number of live cells were counted 96 hours post-transfection using trypan blue exclusion method (Sigma, catalog no. T8154). RNA was prepared from these cells, and qRT-PCR [SYBR PremixEx Taq II (Tli RNase H Plus) ROX, Clontech Takara, catalog no. RR82LR] for AR and AR-V7 was performed as described earlier (13).

Chromatin immunoprecipitation (ChIP), quantitative RT-PCR, and ChIP-qPCR were performed as described earlier (13, 29, 30). Briefly, cells were either transfected, infected, or treated with various inhibitors (R)-9b; CPTH2 (Sigma, catalog no. C9873)], fixed in formaldehyde, lysed and ChIP was performed with antibodies recognizing AR (SCBT, sc-7305; RRID:AB_626671), GCN5 (SCBT, SC-365321; RRID:AB_10846182), H3K14ac (Cell Signaling Technology, 7627S; RRID:AB_10839410), H3K27me3 (Cell Signaling Technology, 9733S; RRID:AB_2616029), NCOR1 (Bethyl Laboratories, catalog no. A301-145A; RRID:AB_873085), EZH2 (Cell Signaling Technology, 5246S; RRID:AB_10694683), or IgG (Abcam, ab2410; RRID:AB_303052), immobilized on Protein A magnetic beads (Bio-Rad, catalog no. 161-4013). The complexes were washed with ChIP buffers, eluted with elution buffer (Active Motif, catalog no. 53008), and subjected to reverse crosslinking, followed by RNase and proteinase K treatments. A part of soluble chromatin was processed separately at the same time and designated as input DNA. Treated ChIP DNA and input DNAs were purified using PCR-DNA purification columns (Qiagen, catalog no. 28106). The amount of immunoprecipitated DNA was determined by real-time PCR.

Cells were transfected, infected, or treated [(R)-9b; EPZ6438, TargetMol, catalog no. T1788] accordingly. RNA was used for cDNA preparation using High-Capacity cDNA Reverse Transcription Kit (Thermo Fisher Scientific, catalog no. 4368814). Resulting cDNA was then analyzed by qPCR using an Applied Biosystems 7900 Real-Time PCR System (Thermo Fisher Scientific) and SYBR Green PCR Master Mix [SYBR PremixEx Taq II (Tli RNase H Plus) ROX, Clontech Takara, catalog no. RR82LR] according to the manufacturers' instructions. Dissociation curves were generated for each plate to verify the integrity of the primers. The relative expression of RNAs was calculated using the comparative C_t or standard curve method. 18S rRNA or Actin were used as internal controls. The primers used are shown in Supplementary Table S1.

All animal experiments were performed using the standards for humane care in accordance with the NIH Guide for the Care and Use of Laboratory Animals. Mice breeding and colony maintenance were performed according to Institutional Animal Care and Use Committee protocols approved by Washington University in St. Louis. A total of 1.5 × 10⁶ VCaP cells were suspended in 200 μL of PBS with 50% Matrigel (Corning, catalog no. 354428) and implanted subcutaneously into the dorsal flank of 6-week-old male SCID mice (Charles River, Strain code:236). Once the tumors reached approximately 100 mm³, mice were randomized and injected subcutaneously with (R)-9b mesylate salt suspended in 6% Captisol (Cydex Pharmaceuticals). Tumor volumes were measured twice weekly using calipers. At the end of the study, all mice were humanely euthanized and tumors were extracted and weighed. To assess the role of NXTAR in tumor growth, 1.5 × 10⁶ VCaP cells that were either infected with pBABE vector or NXTAR and selected with puromycin (Gibco, catalog no. A1113803), implanted and growth of tumors was recorded. The personnel taking tumor volume measurement were blinded for the treatment groups.

The 22Rv1, VCaP, or LNCaP cells were retrovirally infected with control or NXTAR vectors and selected in puromycin medium containing 5% charcoal-stripped serum. Equal numbers of selected cells were seeded in 6-well plates followed by treatment with either (R)-9b, enzalutamide (Selleck Chemicals, catalog no. S1250), or abiraterone (Selleck Chemicals, catalog no. S2246), and the number of viable cells was counted by trypan blue exclusion assay.

Cells were harvested, lysed by sonication in receptor lysis buffer (31) and 20–50 μg of protein lysates were fractionated by SDS-PAGE, and transferred onto a polyvinylidene difluoride membrane (Immobilon, catalog no. IPVH00010). After blocking in 5% nonfat dry milk (RPI, catalog no. M17200; or 3% BSA, Goldbio, catalog no. A-420-500), membranes were incubated with the following mAbs: AR (SCBT, catalog no. sc-7305; RRID:AB_626671), pY267-AR (29), ACK1 (SCBT, sc-28336; RRID:AB_626629), pACK1 (Upstate, 09-142; RRID:AB_612088), EZH2, and Actin. The signals visualized by using the enhanced chemiluminescence (ECL; Pierce, 32106) system.

A total of 1 mg of lysate prepared from HEK293 cells transfected with 6XHis-tagged-EZH2 plasmid was incubated with Ni-NTA agarose beads (Qiagene) overnight at 4°C. Beads were washed three times using wash buffer (20 mmol/L Tris/HCl, 200 mmol/L NaCl, 5 mmol/L imidazole; pH 7.5) and purified EZH2 was eluted using elution buffer (20 mmol/L Tris/HCl, 200 mmol/L NaCl, 200 mmol/L imidazole; pH 7.5).

lncRNA-Protein binding study was performed as described earlier (32). Briefly, 10 pmol biotinylated NXTAR-N5 oligo or biotinylated globin oligos were diluted in 40 μL RNA structure buffer, heated at 90°C for 2 minutes, immediately transferred onto ice for 2 minutes and then kept at room temperature for further 20 minutes. A total of 10 μmol/L of purified EZH2 protein was added to these biotinylated oligos with 500 μL of NETN buffer and the tubes were placed at 4°C for 2 hours. Post washing with cold NETN buffer, streptavidin magnetic beads (60 μL) were added into prepared oligo-protein mix and incubated at 4°C overnight. Beads were washed three times with NETN buffer, followed by Western blot analysis. For NXTAR-EZH2 binding studies, 10 pmol biotinylated oligos complimentary to NXTAR or lacZ oligos were incubated with 1 μmol/L RNA prepared from NXTAR-overexpressing VCaP cells in hybridization buffer (50 mmol/L Tris-HCl pH 7.0, 750 mmol/L NaCl, 1 mmol/L EDTA, 1% SDS, 15% Formamide), along with prewashed streptavidin magnetic beads at room temperature for 4 hours. A total of 10 μmol/L of purified EZH2 protein was added and after incubation for 2 hours magnetic beads were washed three times with NETN buffer, followed by Western blot analysis.

Data were summarized by descriptive statistics. Data between two groups were analyzed with unpaired Student t tests using Graphpad Prism (Graphpad Software Inc., RRID:SCR_002798). The qRT-PCR gene expression in tumor normalized to matched samples of a gene was tested against one (indicating equality to matched normal samples) by one sample Wilcoxon test. All tests were two sided and a P value of ≤0.05 was considered significant.

The data generated in this study are available upon request from the corresponding author.

We identified a convergent lncRNA (LOC105373241) positioned next to the AR gene, which we named NXTAR (Fig. 1A; Supplementary Fig. S1A). The gene is 40,180 nt long (67,784,890 to 67,744,711 nt position), encoding a 1,476-nt transcript with five exons of 81, 94, 39, 103, and 1,159 nt, respectively (Fig. 1A). NXTAR exhibited overall homology with rhesus monkey, mouse, dog, and elephant sequences, but not with chicken, Xenopus frog, zebrafish, or lamprey (Supplementary Fig. S1A). Although exon 5 showed a small stretch of homologous sequence with Xenopus, NXTAR appears to be a mammalian lncRNA, and its regions of conservation with other mammals is shown in the Supplementary Table S2.

Given its genomic position and convergent location (Fig. 1A), we reasoned that NXTAR could influence AR expression or might be affected by it (or both). Prostate cancer cell lines were deprived of androgen, the ligand for AR, by growing them in CSM, followed by assessment of NXTAR RNA expression by qRT-PCR. LNCaP and 22Rv1 cells exhibited NXTAR upregulation upon androgen deprivation (Supplementary Fig. S1B and S1C). This increase in NXTAR was further enhanced upon removal of growth factors by growing cells in serum-free conditions (Supplementary Fig. S1B and S1C), suggesting that loss of AR activity could augment NXTAR levels.

Methylation of the fourth amino acid residue from the N-terminus of histone H3, H3K4me3, is a histone modification associated with the promoters and transcription start sites (TSS) of actively transcribed genes (33). Although the AR promoter is enriched for H3K4me3 in most cancer samples, including CRPCs (34) and AR-positive LNCaP cells (Supplementary Fig. S2A), H3K4me3 deposition on the NXTAR promoter was found to be lacking in LNCaP cells (Gene Expression Omnibus:GSM945240; Supplementary Fig. S2B). These data open the possibility of an inverse correlation between AR and NXTAR expression in prostate cancer.

To test this hypothesis, we designed a primer pair that amplified the 177-bp region spanning exons IV and V of NXTAR and performed qRT-PCR of 35 human prostate cancer and paired normal prostate derived RNAs, which revealed two distinct scenarios. Type I: Twenty-four of 35 patients with prostate cancer (68%) exhibited a significant decrease in NXTAR expression in tumors, compared with normal tissues (Fig. 1B and C). Among these samples, the PSA/KLK3 expression was significantly upregulated relative to normal (Fig. 1D and E). Three of these 10 patients (#376, 422, and 460) were treated with androgen deprivation therapy in past and were considered as CRPCs. Type II: In 11 of 35 patients with prostate cancer, NXTAR levels were either higher than normal or not significantly altered (Fig. 1F and G). The KLK3 levels in these patients are also shown (Fig. 1H and I). Data representing mean of relative NXTAR and KLK3 expression between normal and tumor samples are shown (Fig. 1J and K). In addition, we also assessed NXTAR levels in various grades of prostate cancer and observed that in low-grade cancer (Gleason grade 6) there was no statistically significant change; however, NXTAR levels were significantly downregulated in Gleason grade 7 or medium-grade cancers (Fig. 1L).

Furthermore, NXTAR expression was under-expressed in various AR-positive prostate cancer cell lines, including 22Rv1, LNCaP, and LAPC4. In contrast, “normal” prostate cells (RWPE) exhibited detectable NXTAR transcript levels (Fig. 1M). The highest relative expression of NXTAR was observed in PC3 cells that have no detectable AR expression (Fig. 1M). A semiquantitative PCR using the same set of primers showed a pattern of NXTAR expression similar to that seen in qRT-PCR of different cell lines, with the highest expression in PC3 cells (Supplementary Fig. S1D).

To validate the role of AR, NXTAR expression levels were measured after depletion of AR levels in LNCaP, C4-2B, 22Rv1, and VCaP cell lines using AR siRNA (Supplementary Fig. S3A). A significant increase in NXTAR levels were seen in all prostate cancer cell lines after AR depletion (Fig. 2A–D). Because withdrawal of growth factors also enhanced NXTAR upregulation, we reasoned that growth factor–regulated tyrosine kinase signaling could suppress NXTAR expression. To determine the role of a growth factor–regulated tyrosine kinase in suppressing NXTAR expression, we focused on ACK1 because of its regulation by multiple receptor tyrosine kinases such as insulin receptor, HER2, PDGFR, MERTK, and EGFR (13, 29, 35–37). ACK1, also known as TNK2, is a non-receptor tyrosine kinase whose protein levels (by SIAH-mediated ubiquitinylation) and activation are precisely regulated; a significant increase in its activation occurs as hormone-sensitive prostate cancer progresses to CRPC (31, 38–41). Kinase screen revealed that prostate cancer stem-like cells depends on ACK1 for their survival (42). Recently, we demonstrated that ACK1 regulates AR expression by epigenetically modifying the AR enhancer with a novel epigenetic mark: histone H4 Tyr88 phosphorylation (13). Recognizing the significance of ACK1/AR signaling, we developed the ACK1 small molecule inhibitor (R)-9b, which not only inhibited ACK1 kinase activity but also suppressed AR expression (13, 43). LNCaP and 22Rv1 cells treated with (R)-9b exhibited robust increases in NXTAR levels in a dose-dependent manner (Fig. 2E and F). Efficient knockdown of ACK1 was obtained using two distinct sets of siRNAs (Fig. 2G). Knocking down ACK1 also led to significant induction of NXTAR levels in a time-dependent manner in LNCaP (Fig. 2H) and 22Rv1 (Fig. 2I) cells.

To assess whether restoration of NXTAR levels affected prostate tumor growth, enzalutamide-resistant VCaP cells were implanted subcutaneously in male SCID mice. Once tumors became palpable, the mice were injected with either the vehicle (6% Captisol in PBS) or (R)-9b (12 or 20 mg/kg of body weight), five times a week. (R)-9b treatment significantly decreased tumor growth (Fig. 3A–C). Further, no significant decrease in body weight was observed in (R)-9b–treated mice (Fig. 3D) and no significant decreases in AR levels were noted in mice prostates or brain (Fig. 3E and F). These data suggest that (R)-9b treatment does not have off-target effects in normal tissues and may not pass the blood–brain barrier. However, the xenograft tumors exhibited a significant reduction in AR and in mRNA levels of its target genes, KLK3 and TMPRSS2 (Fig. 3G–I). In contrast, we observed a significant increase in NXTAR levels in (R)-9b–treated tumors (Fig. 3J). The immunoblotting of tumor lysates further confirmed loss of ACK1, phospho-ACK1, AR, and phospho-AR protein levels upon (R)-9b treatment (Fig. 3K). Together, these data suggest that one possible mechanism underlying the significant loss of tumor growth upon overcoming ACK1/AR signaling is the restoration of the tumor suppressor lncRNA NXTAR.

H3K14ac is a hallmark of gene activation and it is also present on bivalent promoters that are usually inactive but primed for activation if stimulus is presented. Levels of H3K14ac correlate with the CpG content of the promoters (44). Several families of histone acetyltransferase have been reported to mediate acetylation of Lys14 of H3, including GCN5 (45). To determine whether an increase in NXTAR upon ACK1 inhibition was due to an increase of H3K14ac deposition at its putative promoter, 22Rv1 and LNCaP cells were treated with (R)-9b alone or in combination with CPTH2, a selective GCN5 inhibitor. The significant increase in NXTAR levels observed following (R)-9b treatment was severely compromised by GCN5 inhibition in 22Rv1 and LNCaP cells (Fig. 4A–C). We checked whether this increase in NXTAR coincided with an increase in H3K14ac status in the NXTAR promoter region by designing primers for this region (designated Ppr3 and Ppr4; Fig. 4A). ChIP revealed a significant increase in H3K14ac mark deposition at these sites in (R)-9b–treated 22Rv1 and LNCaP cells, an increase that was abrogated when (R)-9b was combined with CPTH2 (Fig. 4D and E; Supplementary Fig. S3B and S3C; see Supplementary Fig. S9A–S9D). To further validate the role of GCN5 in deposition of H3K14ac marks, ChIP was performed using GCN5 antibodies, which revealed significant binding of GCN5 to the putative NXTAR promoter in (R)-9b–treated samples (Fig. 4F and G; Supplementary Fig. S3D and S3E).

Androgen deprivation or suppression of AR by siRNA or (R)-9b resulted in robust transcriptional activation of NXTAR (Figs. 1 and 2). Together with the significant GCN5 binding in (R)-9b–treated samples, we reasoned that AR and GCN5 could be competing for binding the same location in the NXTAR promoter. The binding of AR at Ppr3 and Ppr4 was compromised in cells that were deprived of the growth factors or subjected to (R)-9b treatment (Fig. 4H and I; Supplementary Fig. S3F and S3G; see also Supplementary Fig. S9E abd S9F). These data indicate that a decrease in AR levels or activity could promote GCN5 occupation of the NXTAR promoter, upregulating its expression by deposition of H3K14ac marks.

A distal intragenic enhancer was identified within the first intron of the NXTAR gene (chrX: 67779247–67781971), using the ENCODE (Fig. 4A, blue box; ref. 46). Multiple transcription factor binding sites were noticed in this region, including nuclear receptor corepressor1 (NCOR1), which functions as a corepressor for AR. NCOR contains a pair of SW13/ADA2/NCoR/TFIIB (SANT) domains (47). The N-terminal SANT is a critical component of a deacetylase activation domain that binds and activates HDAC3, promoting histone deacetylation and transcriptional repression (48). We reasoned that because inhibition of ACK1 leading to loss of AR levels reinstates NXTAR expression (Fig. 2D–I), AR might recruit corepressor such as NCOR1 to accomplish NXTAR suppression. To test this possibility, 22Rv1 cells were treated with (R)-9b, followed by ChIP with NCOR1 and GCN5 antibodies. qPCR with IRF8 binding site primers revealed increased GCN5 binding upon (R)-9b treatment (Fig. 4J). In contrast, qPCR with NXTAR enhancer region specific primers revealed NCOR1 binding to this region, which was significantly decreased upon (R)-9b treatment (Fig. 4K). Furthermore, transfection with NCOR1 siRNA2, 3 and pooled siRNAs exhibited significant downregulation of NCOR1 expression (Fig. 4L), which was reflected in increased NXTAR expression (Fig. 4M). Taken together with Fig. 4F–I; Supplementary Fig. S3A–S3D, these findings uncover the molecular dynamics at NXTAR locus wherein NCOR1/AR keep NXTAR levels in check by assembling at the promoter/enhancer region, however, upon suppression of AR allows GCN5 binding to reinitiate transcriptional activity.

BLASTN analysis revealed a sequence in exon 5 of NXTAR (the N5 region) with significant homology (in the complementary strand) to two regions of 30 and 43 nt separated by about 1,607 nt in the AR promoter (Supplementary Fig. S4; see also Fig. 7A and B). These data open up the possibility that NXTAR could bind to the AR promoter, regulating its expression reciprocally. To test this hypothesis, we generated a retroviral construct expressing NXTAR and 22Rv1 and VCaP cells were infected. Upon restoration of NXTAR, a significant decrease in AR mRNA level was seen, which was also reflected in significant loss of expression of the AR target genes KLK3 and TMPRSS2 (Supplementary Fig. S5A and S5B). 22Rv1 cells expressed AR-V7, a variant derived from the AR locus, which also was significantly compromised upon reintroduction of NXTAR (Supplementary Fig. S5A). AR protein was decreased upon NXTAR expression in a time-dependent manner in LNCaP and 22Rv1 cells (Supplementary Fig. S5C and S5D, respectively). This regulation of AR protein by NXTAR was further confirmed by its dose-dependent downregulation of AR in 22Rv1 cells (Supplementary Fig. S5F), in contrast, vector-transfected cells had no effect on AR levels (Supplementary Fig. S5E). Further to confirm functionality of LNCaP and 22Rv1 cells, qRT-PCR was performed; increased AR and KLK3/PSA levels post-DHT treatment was significantly compromised upon NXTAR restoration (Supplementary Fig. S6A–S6D).

To examine whether effect of ACK1 inhibitor on AR is mediated through ACK1, a “rescue” experiment was performed, wherein NXTAR overexpression caused significant loss of AR levels. However, when NXTAR expression was accompanied by activated ACK1 (caACK; refs. 49, 50) expression, AR levels were restored (Supplementary Fig. S6E). As an additional evidence for interdependence of AR and NXTAR expression in imparting enzalutamide resistance, AR and NXTAR levels were assessed in Enzalutamide treated VCaP cells. These cells are enzalutamide-resistant and thus exhibited marginal decrease in AR levels, which is also reflected in no change in NXTAR levels (Supplementary Fig. S6F).

To understand precisely how AR transcript levels are suppressed by NXTAR, we took a closer look at the epigenetic landscape of the AR promoter. The histone-methyltransferase enhancer of zeste homolog 2 (EZH2), a subunit of PRC2, catalyzes H3K27 trimethylation (H3K27me3), leading to chromatin compaction and subsequent silencing of genes (51). ChIP with H3K27me3 antibodies followed by qPCR using the primers spanning the region (−0.7 to +3.4 kb) around the TSS of the AR gene (Fig. 5A) revealed a significant increase in deposition of H3K27me3 marks in the AR promoter upon NXTAR restoration in LNCaP, 22Rv1 cells (Fig. 5B; Supplementary Fig. S7A; Supplementary Fig. S9G–S9K and S9L–S9O) and VCaP cells (Supplementary Fig. S7C). To validate that the cause of H3K27me3 deposition is the binding of EZH2 at the AR promoter, ChIP was performed, revealing its significant binding in LNCaP cells overexpressing NXTAR (Fig. 5C; Supplementary Fig. S9P–S9R). Similar observations were made in VCaP (Supplementary Fig. S7B) and 22Rv1 cells (Supplementary Fig. S7F–S7H). Furthermore, (R)-9b treatment too caused a significant binding of EZH2 to the AR promoter in VCaP and 22Rv1 cells (Supplementary Fig. S7D and S7F–S7H), reflecting in increased H3K27me3 levels (Supplementary Fig. S7E).

To explore the potential NXTAR binding to EZH2, RNA pulldown was performed wherein DNA oligonucleotides complementary to NXTAR lncRNA (oligonucleotides complementary to lacZ as a control) were used to perform the pulldown of NXTAR RNA from VCaP cells (which was subjected to retroviral expression of NXTAR). Immunoblotting revealed that EZH2 could bind to NXTAR RNA, but not to the lacZ (Fig. 5D). NXTAR pulldown was confirmed by qPCR (Supplementary Fig. S7I). Furthermore, treatment with EZH2 inhibitor (EPZ6438) restored the AR expression (Fig. 5E). To further assess NXTAR interaction with EZH2, we generated 6XHis-tagged EZH2 construct, which was transfected in 293T cells. EZH2 was purified using nickel affinity chromatography (binding to Ni-NTA beads), followed by elution with imidazole. NXTAR immobilized on streptavidin beads when incubated with purified EZH2, exhibited considerable binding, which was not seen with the LacZ oligo (Fig. 5F). Because, the in vitro RNA/protein binding assay was performed using tagged EZH2 purified from 293T cells, but not from bacteria, we cannot completely rule out the possibility that NXTAR could bind to proteins associated with EZH2 in PRC2 complex, for example, EED, SUZ12, JARID2, AEBP2, RbAp46/48, and PCL (51). Taken together, these data indicate that NXTAR could bind to the AR promoter, recruit PRC2 complex, leading to silencing of AR due to the H3K27me3 marking.

To test whether NXTAR had any effect on prostate cancer cell proliferation, 22Rv1 and VCaP cells were infected with a pBABE-puro vector or NXTAR-expressing constructs and selected using puromycin. In contrast to vector, severe loss of proliferation was seen upon NXTAR expression in VCaP (Fig. 6A) and 22Rv1 (Fig. 6B) cells. The loss of AR expression upon NXTAR expression was confirmed by immunoblotting (Supplementary Fig. S8A and S8B). We reasoned that a lncRNA that inhibits AR (and AR-V7) could limit the proliferation of AR antagonist-resistant prostate cancer cells. To test this hypothesis, VCaP and 22Rv1 cells expressing vector or NXTAR constructs were treated with enzalutamide, which did not result in a significant decrease in proliferation of either cell line; however, NXTAR expression significantly compromised proliferation upon enzalutamide treatment (Fig. 6C and D). Interestingly, this reduction was comparable to that observed with NXTAR overexpression alone, suggesting that NXTAR expression causing loss of AR makes enzalutamide-mediated AR inactivation irrelevant. In addition to enzalutamide, we assessed another AR antagonist that has a different mode of action: abiraterone, which suppresses androgen synthesis. Although 22Rv1 cells are sensitive to abiraterone when it is used alone, NXTAR expression further sensitized them (Fig. 6E). As a positive control, cells were treated with (R)-9b, which caused cell death in both cell lines (Supplementary Fig. S8C and S8D), suggesting that (R)-9b–mediated NXTAR upregulation is robust and could provide superior benefits compared with retroviral overexpression of NXTAR.

To examine the direct effect of NXTAR restoration on prostate tumor growth, equal numbers of VCaP cells infected with either pBABE vector or NXTAR constructs were injected in the SCID mice. Palpable tumors were formed around week 5 that were measured twice a week thereafter. Compared with mice injected with vector-expressing cells, mice injected with NXTAR overexpressing cells showed a significant reduction in tumor growth (Fig. 6F). The tumor burden in mice injected with NXTAR expressing cells was significantly reduced compared with that of vector-transfected cells (Fig. 6G–I). Furthermore, the NXTAR expressing tumors exhibited significant reduction in AR protein levels (Fig. 6J).

As described earlier, we uncovered two regions of 30 and 43 nt upstream of the AR promoter that have complementary sequences to the N5 region of exon V of NXTAR (Supplementary Fig. S4; Fig. 7A and B). We reasoned that binding of NXTAR upstream of the AR promoter could hinder transcription (see Fig. 7A and Graphical Abstract). To test this hypothesis, we designed an oligonucleotide corresponding to the N5 region of exon V, designated NXTAR-N5 (Fig. 7C). VCaP and 22Rv1 cells transfected with NXTAR-N5 exhibited a significant decrease in AR and AR-V7 transcription, which was also reflected in the compromised proliferation of these cells (Fig. 7D and E). The bindings of NXTAR-N5 to the upstream of AR was validated; briefly, biotinylated-NXTAR-N5 oligo was immobilized onto streptavidin beads, incubated with VCaP lysate, followed by qPCR with the primer AR Up1.1 (Fig. 7A). NXTAR-N5 bound upstream of AR, however, globin binding was significantly reduced (Fig. 7F). In addition, NXTAR-N5 interaction with EZH2 was confirmed by performing pulldown as described above (Fig. 7G). Moreover, NXTAR-N5 interaction with purified EZH2 (as described above) was confirmed by performing pulldown (Fig. 7H). Furthermore, NXTAR-N5 introduction induced histone H3K27 methylation in the AR promoter region (−0.7 kb), which was confirmed by ChIP (Fig. 7I).

Overall, our results demonstrate that a novel tumor suppressor lncRNA NXTAR, located close to the AR gene, suppresses AR expression by recruiting EZH2 methyltransferase and marking the AR promoter with H3K27me3 repressive epigenetic marks. The loss of AR, in turn, allowed GCN5 acetyltransferase to bind to the NXTAR upstream region, leading to deposition of H3K14ac epigenetic marks that enhanced NXTAR expression (Graphical Abstract). To address how GCN5 is recruited, we used MotifMap, a web-based tool for genome-wide maps of regulatory elements to scan the region between enhancer and promoter. It revealed a few distinct transcription factor sites (Supplementary Table S3); the first one was the IRF-8. IFN regulatory factors (IRF) upon binding to a specific DNA element, also bind to coactivators such as HATs-GCN5 and PCAF and become modified (52, 53). ChIP with GCN5 clearly shows its binding to promoter (Ppr3/Ppr4 region; Fig. 4F and G; Supplementary Fig. S3D and S3E) and adjoining IRF-8 binding region (Fig. 4J). Taken together, these findings reveal that NCOR1/AR assembles at enhancer/promoter region of NXTAR to maintain its low levels in prostate cancer; however, suppression of AR allows IRF-8 to recruit GCN5 to this newly vacated region, reinitiating NXTAR transcription. Significantly, this negative feed forward NXTAR-AR circuitry seems to explain low levels of NXTAR in AR-positive prostate tumors or cancer-derived cell lines. Furthermore, interruption of this circuitry by inhibition of AR expression using the ACK1 inhibitor, (R)-9b (13) restored NXTAR levels, establishing the tumor suppressor function of this lncRNA. Moreover, a significant loss of enzalutamide-resistant xenograft tumor growth was observed upon restoring NXTAR expression (Fig. 6). Overall, these data provide new insight into the mechanism by which prostate cancer cells maintain high AR levels continuously, that is, suppression of NXTAR.

We did not find the downregulation of NXTAR in advanced prostate cancer in publically available databases (Supplementary Fig. S1E). Similar to other studies on lncRNAs, for example, HOTAIR in breast cancer (54), the correlation of expression with the lncRNAs and their target genes is not observed. There are several potential explanations. First, low expression of NXTAR made it hard to differentiate signal from noise in RNA sequencing data. While NXTAR was not differentially expressed in a patient cohort, a small subset of patients shows a negative correlation between AR and NXTAR. Nonetheless, these data are inconclusive due to small sample size and therefore we decided not to include. Second, the AR locus is one of the most complex loci in prostate cancer. AR expression is regulated by numerous genetic and epigenetic mechanisms. In prostate cancer (particularly in advanced patients), the AR locus is constantly altered. For example, the well-known genetic alteration in prostate cancer is amplification of the AR region that results in overexpression of AR and its nearby genes including NXTAR. Although, database analysis of patients with prostate cancer shows a positive correlation of expression between AR and NXTAR, this observation, however, does not negate the negative regulation of AR by NXTAR as this may be one of the many ways AR is regulated. Importantly, we show this negative regulation in our tightly controlled experimental data where the only perturbation was NXTAR expression.

What is the region within NXTAR binding to EZH2 or its interacting proteins in PRC2 complex? The initial analysis suggests that there could be 2–3 contact points between NXTAR and EZH2. The “RNAstructure” MaxExpect results generated a high fidelity NXTAR RNA secondary structure composed of highly probable bps, shown in Supplementary Fig. S10. It shows that (i) 490–1,180 nt, (ii) 1,270–1,450 nt, and (iii) 290–440 nt regions form distinct secondary structures with high probability of base pairing. Interestingly, these regions are also highly conserved in rhesus monkeys (Supplementary Table S2). We believe, by virtue of high probability of base pairing and evolutionary conservation, these region/s might be involved in binding to PRC2 complex proteins.

Although at least 12 tumor suppressor lncRNAs have been reported in androgen-dependent prostate cancer, only three have been shown to be involved in CRPCs: LINC00844, DRAIC, and PCAT29 (27). LINC00844 was identified to be an AR-regulated lncRNA that is downregulated in metastatic prostate cancer (55). A tumor-suppressive locus on human chromosome 15q23 has recently been reported that contains two lncRNAs, DRAIC and PCAT29, both of which are suppressed by androgen-bound AR (56, 57). In contrast, NXTAR is a distinct lncRNA, both functionally and spatially; unlike the others, (i) NXTAR is convergently paired tail-to-tail with the AR gene; (ii) it negatively regulates AR and AR-V7 expression; (iii) it is, itself, negatively regulated by AR; and (iv) its restoration overcomes AR and AR-V7, suppressing enzalutamide-resistant xenograft tumor growth. These unique qualities of NXTAR make it a highly desirable therapeutic target for recurrent prostate cancer, a stage that has become a major cause of prostate cancer–related mortality. Future studies involving restoration of NXTAR by using the small molecule inhibitor (R)-9b or NXTAR-N5 therapeutic oligonucleotide could provide long-term beneficial response in patients with CRPC.

M. Sawant reports a patent for 63/126,916 NXTAR-N5 pending. K. Mahajan reports other support from Technogenesys outside the submitted work. F.Y. Feng reports personal fees from Janssen, Blue Earth Diagnostics, Astellas, Myovant, Roivant, Bayer, Bristol Myers Squibb; other support from SerImmune and BlueStar Genomics; and personal fees and other support from Artera outside the submitted work. N.P. Mahajan reports grants from NIH/NCI, Prostate Cancer Foundation, and Department of Defense during the conduct of the study; other support from TechnoGenesys, Inc., outside the submitted work; in addition, N.P. Mahajan has a patent for 63/126,916 pending. No disclosures were reported by the other authors.

R. Ghildiyal: Data curation, formal analysis, validation, investigation, methodology, writing–original draft. M. Sawant: Data curation, formal analysis, validation, investigation, methodology, writing–review and editing. A. Renganathan: Resources, validation, investigation. K. Mahajan: Formal analysis. E.H. Kim: Resources. J. Luo: Data curation. H.X. Dang: Data curation. C.A. Maher: Data curation. F.Y. Feng: Formal analysis. N.P. Mahajan: Conceptualization, design therapeutic oligos, resources, formal analysis, supervision, funding acquisition, writing–original draft, project administration, writing–review and editing.

N.P. Mahajan is a recipient of NIH/NCI grants (1R01CA208258 and 5R01CA227025), Prostate Cancer Foundation (PCF) grant (17CHAL06) and Department of Defense grant (W81XWH-21-1-0202). F.Y. Feng is supported by NIH/NCI grants (5R01CA227025). K. Mahajan is supported by Department of Defense grant (W81XWH-21-1-0202).

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.