Human PCa LNCaP, DU145 and 22rv1 cell lines were purchased from the American-Type Culture Collection (ATCC, ON, Canada). In particular, we have selected LNCaP cells, which express a mutated AR gene and grow in castrate testosterone concentrations [24] and in DU145 cells, which are AR^- and a model of anaplastic CRPC [25,26]. LNCaP cells were cultured in RPMI-1640 ATCC modification from GIBCO (cat# A1049101) and DU145 and 22rv1 cells (DU145-NC, DU145-OE, 22rv1-NC and 22rv1-OE; see below and in Supplementary Figure 4) in RPMI-1640 from GIBCO (cat# 21875034), supplemented with 10% of fetal bovine serum (FBS) and 1% antibiotics (penicillin and streptomycin). All the cell lines were cultured according to the protocols available from the ATCC website. Cells were cultured at 37°C in a 5% CO₂-humidified incubator.

DU145 cells that express HORAS5 at undetectable levels were stably transfected with a lentivirus-derived particle that induced HORAS5 overexpression (Genecopoeia, MD, USA, Cat# LPP-GS266B-Lv105-050). 7.0 × 10⁴ AR^- Du145 and AR⁺ 22rv1 cells were first seeded in 24-well plates and were incubated overnight. The following day, old media was replaced by media supplemented with polybrene (aka Hexadimethrine Bromide, Sigma Aldrich, Gillingham, UK, Cat# H9268) at a final concentration of 8 μg/ml to increase transduction efficiency [27]. After that, 5 μl of the purified human linc00161 (HORAS5) lentiviral particles (Titer: 1.37 × 108 TU/ml where 1 TU = 100 copies of viral genomic RNA) were added to the wells and incubated overnight. The following day, the wells were washed with RPMI-1640 containing 10% FBS three-times and were subsequently left for 2 days to reach confluence. 48- to 72 h post-transduction, cells were split one-in-two into six-well plates and were allowed to adhere for 5–6 h prior to antibiotic selection using puromycin (Gibco, Loughborough, UK, Cat# A1113803). Selection lasted for 2 weeks with media change every 3–4 days. To achieve high copy number, 3 μg/ml of puromycin was selected. These cells were called Du145-OE and 22rv1-OE, respectively, and were passaged and frozen for long-term storage in liquid nitrogen. All overexpression experiments were normalized to cells transduced with the empty vector (DU145-NC and 22rv1-NC, respectively) as lentiviral vectors did not induce phenotypic changes.

Total RNA was extracted using the RNeasy plus mini Kit (Qiagen, Manchester, UK) from cultured cells according to the manufacturer’s protocol.

Nuclear- and cytoplasmic-fractionated RNA extraction was performed on cultured cells using the PARIS™ kit (Ambion, Loughborough, UK) following the manufacturer’s protocol. DNAse digestion of the fractionated RNA was performed using the TURBO DNA-free™ Kit (Ambion, Cat# AM1907).

Upon extraction, 1 μg of total RNA was reverse‐transcribed using high capacity cDNA reverse transcription kit (Applied Biosystems, Loughborough, UK) following the kit instructions. The cDNA obtained was diluted ten-times prior the quantitative PCR (qPCR). TaqMan assays (Applied Biosystems) were used for the qPCR to assess gene expression as per the manufacturer’s protocol. The TaqMan assays used were LINC00161 (Hs00863167_g1) and BCL2A1 (Hs00187845_m1). HPRT1 (Hs02800695_m1) was used as housekeeping control in all the reverse transcription (RT)-qPCR experiments. For subcellular localization of RT-qPCR experiments, the probes MALAT1 (Hs00273907_s1) and GAPDH (Hs02786624_g1) were also used as nuclear and cytoplasmic control respectively.

Gene knockdown (KD) experiments were performed using the reverse transfection method [28]. Cells were seeded in six‐well or 96‐well plates with the lipid:siRNA mixture prepared using the RNAiMAX reagent (Invitrogen, Loughborough, UK) according to the manufacturer’s protocol. Final siRNA treatment dosages were 2 nM concentrated. All duplexes were purchased from Integrated DNA Technologies (IDT) (Leuven, Belgium): anti‐HORAS5 (aka Linc00161) DsiRNA hs.Ri.LINC00161.13.2, anti-BCL2A1 DsiRNAs hs.Ri.BCL2A1.13.1 and hs.Ri.BCL2A1.13.2 and nontargeting negative control (scramble) DS NC1. After 48 or 72 h post‐transfection, treated cells were harvested for extracting total RNA and/or total protein.

5 mg of cabazitaxel (Jevtana, Selleckchem, Ely, UK, cat#S3022) was resuspended in dimethyl sulfoxide (DMSO) in order to obtain a final stock solution of 5 mM. 5 mg of enzalutamide (MDV3100, Selleckchem, cat#S1250) was resuspended in DMSO in order to obtain a final stock solution of 1 mM. Both drugs were stored at –80°C. Carboplatin was resuspended in water to obtain a stock solution of 10 mg/ml and stored at –20°C. All drugs were thawed at room temperature and diluted in cell culture media to treat the cells at different concentrations, according to the experiments:

Cabazitaxel was used to treat LNCaP and DU145 cells for 24–48–72 h at (5–50) nM in gene expression experiments, caspase assays, RNA sequencing and at (000005–0.0005–0.05–0.5–5–50–100) nM in the Trypan blue-based cell counting and IC₅₀ calculation experiments. 22rv1 were treated for 48–72 h with 5–50 nM of cabazitaxel in gene expression experiments and Trypan blue-based cell counting;

All the drugs concentrations have been selected based on the IC₅₀ and concentrations used in the literature for CRPC cells, following a log₁₀ criteria, in order to select a wide range of concentrations for our analyses [29–32].

Cell proliferation was assessed via Trypan blue-based cell counting; in this contest cell metabolic assays can give altered results due to taxane interference with mitochondria metabolism [33].

2 × 10⁵ DU145-NC and -OE (and 22rv1-NC and -OE, Supplementary Figure 4E) cells were seeded in a six-well plate and treated with DMSO/cabazitaxel in the concentrations specified above (see drug treatments).

2.5 × 10⁵ LNCaP cells and 5 × 10⁵ DU145-OE cells were seeded in a six-well plate and reverse transfected with 2 nM of control siRNA and either anti-HORAS5 siRNA or anti-BCL2A1 siRNAs, respectively. At day two post-transfection, the cells were treated with DMSO/cabazitaxel in the concentrations specified above (see drug treatments).

For all the cells, at day 3 (LNCaP) and 2 (DU145-NC/-OE and 22rv1-NC/-OE) after the drug treatment, trypan blue-based cell counting was performed and the IC₅₀ was calculated by nonlinear regression analysis (variable-slope inhibitor fitting), after normalization to untreated (DMSO) cells.

10⁴ DU145-NC and -OE cells were seeded in a white, flat-bottom 96-well plate and treated with DMSO/5 nM of cabazitaxel.

1.25 × 10⁴ LNCaP cells and 2.5 × 10⁴ DU145-OE cells were seeded in a white, flat-bottom 96-well plate and reverse transfected with 2 nm of control siRNA and either anti-HORAS5 siRNA or anti-BCL2A1 siRNAs, respectively. At day 2 post-transfection, the cells were treated with DMSO/5 nM/50 nM of cabazitaxel.

At day 3 (LNCaP) and 2 (DU145-NC and -OE) after the drug treatment, Caspase-Glo reagent (Promega, Southampton, UK) was added to the cells and total luminescence was quantified, following the manufacturer’s protocol. Results were normalized to the cell count at the respective treatment concentration and time of treatment.

Total RNA samples were isolated from DU145-NC/OE cells untreated (DMSO) versus treated for 48 h with 5 nM of cabazitaxel. Next-generation sequencing (NGS) based on the Ion Torrent Semiconductor technology (Thermo Fisher Scientific, Loughborough, UK) and bioinformatics analysis were carried out by the Institute of Pathology of the University Hospital Basel, Switzerland. The resulting dataset was further analyzed to determine the protein-coding genes upregulated when HORAS5 is overexpressed (DU145-OE) versus negative control (DU145-NC) upon cabazitaxel treatment (cabazitaxel-driven genes in Figure 4A). The expression threshold was set as log2 fold-change >2 and p < 0.01 for the cells overexpressing HORAS5. The 87 genes were then filtered for DU145-NC p-value and sorted in descending order (top 25 genes in Figure 4A). The final shortlist was obtained by ranking the top three genes based on literature evidence on cancer and taxane resistance.

Cell lysates were obtained using 15–50–100 μl of radioimmunoprecipitation assay (RIPA) buffer (Tris pH 8.0 [Sigma Aldrich]; NaCl [Sigma Aldrich]; EDTA [Sigma Aldrich]; Igepal [Sigma Aldrich]; SDS [Sigma Aldrich], NaF [Sigma Aldrich]; NaVO3 [Sigma Aldrich]) according to the number of cells used. Proteins were quantified with the Peierce BCA assay (Thermo Fisher Scientific) as per the manufacturer’s protocol. 15 mg of proteins were resolved via gel electrophoresis on reducing SDS-polyacrylamide gels (Tricine 10–20%, Thermo Fisher Scientific) run at 110 V for 2 h. Protein transfer was performed using nitrocellulose membrane at 300 mA for 2.5 h. The membranes were blocked in 8% skimmed milk dissolved in tris-buffered saline (TBS; Sigma Aldrich) containing 0.1% tween-20 (TBS-T) at room temperature for 1 h. After 1 h, the blots were incubated overnight at 4°C with protein-specific primary antibodies dissolved in 5% BSA diluted in TBS-T for anti-BCL2A1 (Cell Signalling Technology, Leiden, The Netherlands, A1/Bfl-1 (D1A1C) Rabbit mAb, Cat# 14093) and 5% milk diluted in TBS-T for anti-GAPDH (Sigma Aldrich, cat# G9545). After the overnight incubation, protein blots were washed three-times with TBS-T for 10 min. Blots were then incubated with horseradish peroxidase (HRP)-conjugated anti-rabbit secondary antibody (Thermo Fisher Scientific, cat#31460) dissolved in 8% milk diluted in TBS-T at room temperature for 1 h. After the incubation, blots were washed four-times in TBS‐T for 10 min each.

After washing, enhanced chemiluminescence (ECL) western blotting substrate kit was used (Millipore, Watford, UK) to visualize blot chemiluminescence, using Syngene Gbox with GeneTools software (Syngene, Bangalore, India).

To assess the clinical relevance of HORAS5 expression in PCa samples, CBioPortal (www.cbioportal.org) was queried using a publicly available Agilent microarray dataset [34]. This dataset consisted of a single study with 63 patients, of which 15 patients did not undergo chemotherapy treatment and ten were treated with taxane only. We compared none versus taxane only treatment.

1.5 × 10⁵ LNCaP and 1.5 × 10⁵ DU145-NC cells were seeded in six-well plates and after 24 h, the cells were transfected with 2 nM of a negative control antisense oligonucleotides (ASOs) (Eurofins Genomics, Wolverhampton, UK, ASO-NC: C*C*T *T*C*C *C*T*G *A*A*G *G*T*T *C*C*T *C*C) and HORAS5-ASO3 (Eurofins Genomics, HORAS5 V3*: G*G*C *T*G*C *T*G*C *A*T*G *T*C*T *A*C*A *G*T) preselected as the most effective of eight tested ASO sequences for HORAS5 KD (Figure 6A), using the RNAiMAX reagent (Invitrogen, Loughborough, UK), according to the manufacturer’s protocol. At day 2 post ASO treatment, the cells were treated with DMSO/cabazitaxel at concentrations specified above (see drug treatments). At day 3 (LNCap) or day 2 (DU145-NC) postdrug treatment, the cells were counted using the tripan blue-based method.

All data were obtained from at least two or three independent experiments and analyzed using GraphPad Prism 7 software. Values are presented as mean ± standard deviation (SD). Significant differences between the groups were calculated using linear trend test, Student’s t-test, one-way ANOVA with Tukey’s multiple comparison post-test and two-way ANOVA with Sidak’s multiple comparison post-test and nonlinear fit (log inhibitor vs normalized response-variable slope) IC₅₀ analysis. A p < 0.05 was set as threshold for statistical significance. Outlier test was carried out to remove the extreme experimental replicate for the IC₅₀ calculation.

Based on previously published data on HORAS5 expression in a panel of PCa cells [22], we selected LNCaP (AR⁺) as cell line expressing this lncRNA and DU145 (AR^-) as cell line with undetectable levels of HORAS5 [22]. To investigate the effects of HORAS5 in AR^- PCa cells, HORAS5 expression was artificially induced via DU145 cells’ stable transduction with lentiviral particles containing HORAS5 gene, expressed under the strong CMV promoter (Figure 1A). This transduction induced elevated HORAS5 expression, which was stable for at least 20 passages (Figure 1B). We called these cells DU145-OE and the corresponding cells transfected with the empty vector were called DU145-NC. Additionally, we have confirmed HORAS5 overexpression in AR⁺ 22RV1 cells, which express undetectable endogenous levels of HORAS5 (Supplementary Figure 4A). Our previous studies showed that HORAS5 is located in the cytoplasm of LNCaP cells [22]. Our results in DU15-OE cells indicate that the artificial expression of this lncRNA preserved its cytoplasmic location (Figure 1C). Our analyses in 22rv1-OE also confirmed HORAS5 cytoplasmic location (Supplementary Figure 4B). Additionally, we observed no morphological changes induced by HORAS5 overexpression in DU145-OE versus DU145-NC and DU145-WT (Figure 1D) and HORAS5 overexpression did not influence cell proliferation in untreated cells (Figure 1E).

Since our previous studies have shown HORAS5 upregulation in CRPC versus hormone sensitive PDXs [22] and other studies suggested that HORAS5 modulates drug response [13,21], we sought to investigate whether exposure to clinically employed drugs could affect the expression of HORAS5. To this aim, we selected three drugs, which are representative of the most commonly used treatments in CRPC patients: the AR inhibitor enzalutamide, which is used in AR⁺ CRPC [5,35,36]; the microtubule inhibitor cabazitaxel, which is active against both AR⁺ and AR^- CRPCs [37,38]; the platinum agent carboplatin, which displayed some activity in AR^- CRPCs (Table 1) [39,40]. In line with clinical indications, enzalutamide was used in LNCaP and 22rv1-OE cells, carboplatin in DU145-OE cells only and cabazitaxel in all the cell lines with doses ranging around the IC₅₀ and concentrations found in the literature for CRPC cells, on a log₁₀ basis [29–32]. Based on our criteria (p < 0.05 and R² >0.5) enzalutamide did not determine a significant change in HORAS5 expression in LNCaP (p = 0.2451, R² = 0.06073) (Figure 2A) and 22rv1-OE cells (p = 0.0029, R² = 0.3195) (Supplementary Figure 4C) at the clinically achievable concentrations used [41]. Similarly, no effect was observed for carboplatin (p = 0.0061, R² = 0.1657) (Figure 2B) [42,43]. Cabazitaxel induced a concentration-dependent increase of HORAS5 expression in LNCaP (p < 0.0001, R² = 0.6513) (Figure 2C), DU145-OE (p < 0.0001, R² = 0.7563) (Figure 2D) and 22rv1-OE cells (p < 0.0001, R² = 0.7185) (Supplementary Figure 4D). All the concentrations of cabazitaxel used in these experiments are clinically achievable [44]. Overall, just cabazitaxel treatment determined a linear concentration-dependent increase in the expression of HORAS5 (R² >0.5 linear trend test). A time-course experiment revealed that cabazitaxel-induced HORAS5 upregulation is time-dependent: we did not observe any significant transcript upregulation up to 24 h after treatment. However, HORAS5 was significantly upregulated in DU145 (48 h, 72 h) and in LNCaP (72 h) cells (Figure 2E & F). For this reason, we chose 72 h for LNCaP and 48 h for DU145-OE as time points for further experiments.

So far, we have shown that cabazitaxel induces a dose- and time-dependent increase in HORAS5 expression in both AR⁺ and AR^- CRPC cells. We therefore sought to investigate the functional significance of HORAS5 in CRPC cells exposed to cabazitaxel. We optimized the silencing procedure and obtained a KD of 77%, at day 5 after transfection (Supplementary Figure 1A). This time point results from 2 days of silencing + 3 additional days of cabazitaxel treatment. We then hypothesized that HORAS5 silencing and overexpression can affect CRPC cell proliferation. We analyzed the effect of HORAS5 modulation on cabazitaxel anticancer activity. Our data showed that cabazitaxel induced a dose-dependent growth inhibition (Figure 3A & B). This inhibition decreased in DU145-OE compared with DU145-NC (Figure 3A) with a significant increase of cabazitaxel IC₅₀ of 9.8-times (from 3.11 ± 1.48 nM to 30.55 ± 3.9 nM, Supplementary Figure 1B) in the cells that overexpress HORAS5 (p = 0.0114) (Figure 3C). Vice versa, cabazitaxel-dependent growth inhibition increased in LNCaP cells upon HORAS5 KD (Figure 3B). Our calculation showed that HORAS5 silencing caused dramatic decrease of cabazitaxel IC₅₀ in this cell line (from 20.80 ± 0.74 nM to 2.59 ± 0.77 nM, p = 0.0033) (Figure 3D, Supplementary Figure 1B). These results demonstrate that HORAS5 promotes the survival of both AR⁺ and AR^- CRPC cells exposed to cabazitaxel. In order to investigate the role of the apoptotic pathway in this phenomenon, we measured caspase 3/7 activity in LNCaP and DU145 cells exposed to cabazitaxel. Our data showed that HORAS5 overexpression significantly reduced caspase 3/7 activation in response to cabazitaxel exposure (Figure 3E). In keeping with this result, HORAS5 KD increased caspase 3/7 activation in LNCaP cells exposed to cabazitaxel (Figure 3F). Overall, these findings indicated that HORAS5 promoted cabazitaxel resistance in both AR⁺ and AR^- PCa cells via increase of cell proliferation and inhibition of caspase-mediated apoptosis.

To evaluate the transcriptomic profile of CRPC cells exposed to cabazitaxel and the effects of HORAS5 modulation, we performed an NGS transcriptome analysis (Supplementary Figure 2A–E) using the following samples: DU145-NC, untreated; DU145-NC, exposed to cabazitaxel; DU145-OE, untreated; DU145-OE, exposed to cabazitaxel. Based on this analysis, 87 genes were significantly upregulated (fold change (FC) ≥2, p < 0.01) in DU145-OE treated with cabazitaxel versus untreated cells (Figure 4A). Notably, these genes were not significantly upregulated in DU145-NC exposed to cabazitaxel (Figure 4A). We then ranked these 87 genes according to the DU145-NC p-value and shortlisted the top 25 (Figure 4A, Supplementary Table 1). Three of these 25 genes (SOX9, CCL20, BCL2A1) have been previously implicated in cancer and drug resistance (Figure 4 [45–49]A). Hence, we sought to validate our transcriptome results by measuring the expression of SOX9, CCL20 and BCL2A1 via RT-qPCR. No significant differences in SOX9 expression were found (Figure 4B). CCL20 and BCL2A1 were both significantly upregulated in the HORAS5 expressing cells versus negative control upon cabazitaxel exposure. This differential expression pattern was particularly significant for BCL2A1 (Figure 4C & D), which is also a well described antiapoptotic gene [48,50,51]. For this reason, we decided to investigate the role of BCL2A1 in HORAS5-dependent cabazitaxel resistance. First, we confirmed that BCL2A1 protein expression was induced by cabazitaxel treatment (Figure 4E). According to our results BCL2A1 seemed to be the most consistently upregulated gene in cabazitaxel-treated cells that overexpress HORAS5. Therefore, we investigated whether BCL2A1 KD could rescue the drug-resistant phenotype induced by HORAS5 overexpression.

Based on our transcriptomic analysis, we hypothesized that BCL2A1 mediates the drug-resistance phenotype induced by HORAS5. To test this hypothesis, we optimized the KD procedure using two different siRNAs directed against both variants of BCL2A1 mRNA. Our results show that siRNA1 and siRNA2 determined a reduction in BCL2A1 mRNA of 84 and 95% respectively (Figure 5A). Our results also showed that both siRNAs reduced the expression of BCL2A1 protein (Figure 5B). These two siRNAs were used to transfect DU145-OE cells in order to investigate whether BCL2A1 KD could revert the cabazitaxel resistance phenotype induced by HORAS5 overexpression. Our data showed that BCL2A1 KD significantly reduced DU145-OE resistance to cabazitaxel (Figure 5C). We also tested whether BCL2A1 effect on cabazitaxel response was a consequence of its antiapoptotic activity. As we showed in Figure 5D, BCL2A1 KD determined a small increase in apoptosis in untreated cells (Figure 5D). However, BCL2A1 KD caused a highly significant increase in apoptosis when the cells are exposed to both 5 nM (1.98≤FC≤2.21, p < 0.0001) and 50 nM (3.06≤FC≤4.36, p < 0.0001) cabazitaxel. These findings show that BCL2A1, which is upregulated upon HORAS5 expression, enhances cabazitaxel resistance by inhibiting the apoptosis response.

Since HORAS5 modulation via lentiviral-driven overexpression and siRNA-mediated KD affected cabazitaxel resistance, we investigated HORAS5 expression in clinical PCa samples from a published study [34], accessed via cBioPortal. HORAS5 expression was significantly higher (p = 0.0086) in metastatic samples from patients treated with taxanes than in untreated patients (Figure 6A). Given this evidence, we analyzed if HORAS5 could effectively be silenced using ASOs. Since ASOs are currently used in clinical trials [52–54], we thought that this set of experiments could highlight the therapeutic potential of targeting lncRNAs like HORAS5 in the clinical setting. For this set of experiments, we used LNCaP cells, which express endogenous detectable levels of HORAS5. We tested eight different ASOs and found that ASO3 was the most effective inhibitor of HORAS5 expression (78.2%) (Figure 6B, Supplementary Figure 3). Hence, we tested ASO3 in combination with cabazitaxel to analyze if this HORAS5-targeting ASO influenced cabazitaxel IC₅₀. Our results showed that the HORAS5-targeting ASO determined a decrease in the cell count upon cabazitaxel treatment and therefore a significant decrease in the IC₅₀ (FC = 6.55, p = 0.0034) (Figure 6C & D). To rule out off-target effects, we have tested ASO3 in DU145-NC cells, which express undetectable levels of HORAS5. ASO3 does not affect HORAS5 expression (Supplementary Figure 5A) neither cabazitaxel effect on cell count of DU145-NC (Supplementary Figure 5B). This evidence showed that ASO-directed HORAS5 KD decreased cabazitaxel resistance in LNCaP cells and that its role in vivo should be further evaluated. Overall, these findings suggested that HORAS5 could have a clinical relevance and paved the way to HORAS5 targeting in combination with taxanes to increase drug response in CRPC patients.