ELOVL5 Is a Critical and Targetable Fatty Acid Elongase in Prostate Cancer

Prostate cancer is the most common cancer diagnosed in men from the developed world and is a major cause of mortality (1). Androgens acting through the androgen receptor (AR) transcription factor drive prostate cancer initiation and progression. Accordingly, androgen deprivation therapy (ADT) has remained the first-line treatment strategy for advanced prostate cancer since the 1940s. Although initially effective, ADT fails to achieve an enduring remission, and patients ultimately develop castration-resistant prostate cancer, which has a poor prognosis despite the clinical implementation of new AR-directed agents (2). Underpinning the development of more effective therapeutic strategies is a more detailed understanding of the critical cellular processes through which androgens regulate prostate tumor biology and promote cancer cell survival.

It is well established that androgens have a major stimulatory effect on de novo lipogenesis in prostate cancer cells (3). This synthesis of fatty acids is activated in the majority of prostate cancers, and is associated with poor prognosis (4, 5). Approximately 80% of de novo synthesized fatty acids are utilized for the production of phospholipids, the dominant component of cellular membranes, and is linked to the tumor cell's increased demand for cellular membranes and their protection from oxidative stress (6). This biology has elicited major efforts to target lipid synthesis therapeutically via inhibition of the key lipogenesis enzyme, fatty acid synthase (FASN). FASN is directly and indirectly regulated by the AR, and although FASN inhibitors have proved promising in the preclinical setting, off-target effects, poor solubility, and pharmacokinetics have prevented their clinical development (7).

It is known that androgens stimulate expression of over 20 enzymes involved in lipid synthesis, binding, uptake, metabolism and transport (8, 9); however, the precise regulatory mechanisms and influence of these factors on lipid composition and metabolism in prostate tumor cells remain largely undefined. The aim of this study was to gain insight into regulation of the prostate tumor lipidome by androgens in prostate cancer cells, and ultimately identify novel metabolic effectors of androgen action for therapeutic intervention.

Human prostate cancer cell lines LNCaP (clone FGC), 22Rv1, and LAPC4 were from the ATCC. All cell lines were authenticated by short tandem repeat profiling at Cell Bank Australia (NSW, Australia) in July 2016 and again in July 2020. Cell lines were cultured in RPMI-1640 medium (Life Technologies) containing 10% fetal bovine serum (FBS) and 2 mmol/L L-glutamine (Life Technologies). For androgen treatment, cells were initially seeded in medium containing 5% dextran charcoal coated FBS (dcc-FBS), and after 24 hours, 1 nmol/L mibolerone or DHT added and cells cultured for an additional 72 hours. For antiandrogen treatment, LNCaP cells were cultured in growth medium containing 1 μmol/L enzalutamide for 72 hours.

Patient-derived explant (PDE) experiments were conducted according to Declaration of Helsinki principles. Prostate tumor specimens were obtained with written informed consent through the Australian Prostate Cancer Bioresource from men undergoing robotic radical prostatectomy at St. Andrew's Hospital (Adelaide, Australia). Tumors were cultured as PDEs for 48 hours in the presence or absence of 10 μmol/L enzalutamide, as previously described (10). Half was formalin-fixed and paraffin-embedded for histology and half was snap frozen for lipidomics.

Cells or tissues were homogenized in ice-cold 800 μL phosphate-buffered saline (PBS). Cell/explant lysate (100 μL) was set aside to quantify DNA for normalization then lipids extracted using a modified Bligh-Dyer protocol by adding 800 μL CHCl₃, 900 μL 1N HCl:CH₃OH 1:8 (v/v), and 500 μg of the antioxidant 2,6-di-tert-butyl-4-methylphenol (Sigma). Following centrifugal phase separation, the lower organic fraction was collected, evaporated, and the lipid pellet stored under argon gas at −20°C. Lipid pellets were reconstituted in diluent (CH₃OH:CHCl₃:NH₄OH; 90:10:1.25, v/v/v) according to DNA quantification (1 μL diluent/1 μg DNA). Phospholipids were analyzed by electrospray ionization tandem mass spectrometry (ESI-MS/MS) on a hybrid quadrupole linear ion trap mass spectrometer (4000 QTRAP system, AB SCIEX) equipped with a TriVersa NanoMate robotic nanosource (Advion Biosciences) as described previously (6).

RNA was extracted using the PureLink total RNA purification system (Life Technologies) according to the manufacturer's instructions, and concentration assessed using a NanoDrop DM-1000 spectrophotometer (NanoDrop Technologies). RNA was reverse transcribed using Superscript II RT and random hexamer primers according to the manufacturer's instructions (Invitrogen). RT-qPCR was performed on a 7500 Fast Real-Time system (Applied Biosystems) using primer sequences detailed in Supplementary Table S1. Expression levels were normalized against 18S, GUSB, or L19 as reference genes.

Gene set enrichment analysis (GSEA) was implemented using the Broad Institute's public GenePattern server essentially as described previously (11). Briefly, the correlation between expression levels of ELOVL2, ELOVL5, ELOVL7, and 26,447 genes was calculated using data from a cohort of 131 patients from Memorial Sloan-Kettering Cancer Center (MSKCC; ref. 12). Genes were subsequently ranked according to Pearson correlation coefficient (r) value, and this ranked gene list was used for GSEA pre-ranked analysis.

AR chromatin immunoprecipitation sequencing (ChIP-seq) data from clinical specimens (7 normal prostate and 13 primary tumors; ref. 13) were obtained from GEO (GSE56288). Mean signals from normal and tumor samples were generated using the Integrated Genome Browser (14).

LNCaP cells were seeded at 3×10⁶ cells/plate in 15 cm plates in RPMI-1640 containing 10% dcc-FBS for 3 days and then treated for 4 hours with 10 nmol/L DHT or vehicle (ethanol). AR ChIP was performed as described previously (15).

ON-TARGET plus human small interfering RNAs (siRNA) and a control siRNA (D-001810-01-20 ON-TARGET plus Nontargeting siRNA #1) were from Dharmacon. The most effective siRNA from a pool of four was selected for targeting ELOVL2 (J-009531-10; J-009531-11), ELOVL5 (J-009260-10; J-009260-12), ELOVL7 (J-028493-09; J-028493-11), and SREBP1 (J-006891-07) and reverse transfected at a concentration of 5 nmol/L using Lipofectamine RNAiMAX transfection reagent (Invitrogen) according to the manufacturer's instructions.

Five tissue microarrays (TMA) comprised of 1-mm triplicate tumor cores, and single/duplicate benign tissue cores were used for evaluation of ELOVL5 expression. TMAs were constructed from a cohort of 249 radical prostatectomy prostate cancer patients who had archival tissue available through the Gelb Center at Dana-Farber Cancer Institute (Boston, MA). Human studies were conducted according to Declaration of Helsinki principles under institutional review board. A written informed consent was received from participants prior to inclusion in the study.

A multiplex tyramine signal amplification (TSA) method was performed on 4-μm TMA sections for the detection of ELOVL5 protein, with pan-Cytokeratin and p63 for masking the tumor and benign epithelium, respectively. Staining was performed on a Leica Bond Rx automated immunostainer, using a protocol for sequential TSA-amplified immunofluorescence labels, with DAPI as the nuclear counterstain. Slides were incubated for 30 minutes with antibodies against ELOVL5 (HPA047752, Sigma, 1:2,000), pan-Cytokeratin CAM5.2 (452M, Cell Marque, 1:500), and an HMWCK-P63 cocktail comprised of p63 (4892, Cell Signaling Technology, 1:100) and cytokeratin 5/4 (API3025AA, Biocare, ready-to-use), followed by 15-minute incubation with TSA-conjugated Opal dye (Opal 690 for ELOVL5, Opal 520 for pan-Cytokeratin, Opal 570 for HMWCK and p63). A tissue and cell segmentation algorithm was developed in PerkinElmer InForm software and applied to the TMAs. All images were reviewed by a pathologist and manually edited for accuracy before the final segmentation data were generated. Cores with major folding or staining artifacts were excluded. Signal was expressed as fluorescent count normalized to exposure time. Per-cell ELOVL5 cytoplasmic signal was normalized to nuclear DAPI signal before comparison between tumor and benign glands. Unpaired t test was used when variances were equal; otherwise, nonparametric Mann–Whitney was applied.

Alexa Fluor-488 labeled PAMAM(G4)-porous silicon nanoparticles (pSiNP) were prepared as described previously (16), and loaded with either siCTRL or siELOVL5 overnight prior to experimental day. On the day of PDE treatment, loaded NPs were washed 2× in RNase-free water and added to explant culture media at a concentration of 200 μg/mL. Explants were prepared in triplicate as described above and cultured in a 24-well plate containing 500 μL NPsiCTRL or NPsiELOVL5 containing growth medium or growth medium without any NPs as a control. After 24 hours, an additional 50 μL of NP containing media was added, and explants were cultured for an additional 48 hours, making a total culture time of 72 hours. At the end of the experiment, explants were formalin-fixed and paraffin-embedded prior to immunostaining with antibodies for ELOVL5 (HPA047752, Sigma, 1: 2,000) or Ki67 (M7240, DAKO, 1:200), as described previously (10). Slides were scanned with an Axio Z.1 Scanner (Carl Zeiss Pty. Ltd.) using 40× objective, and images captured using Zen Blue 2.6 Software (Carl Zeiss Pty. Ltd.). Ki67 positivity was manually counted, as described previously (10). ELOVL5 staining intensity was measured using ImageJ software.

For shELOVL5, LNCaP cells were transduced with universal negative control shRNA lentivirus [pLenti-H1-shRNA(Neg-Control)-Rsv(Puro)] or ELOVL5 shRNA lentivirus [pLenti-H1-shRNA(hELOVL5)-Rsv(Puro)], while for hELOVL5, cells were transduced with universal null control hRNA lentivirus [pLenti-TetCMV(hRNA(Null-Control))-Rsv(GFP-Puro)] or ELOVL5 hRNA lentivirus [pLenti-TetCMV(hRNA(ELOVL5))-Rsv(GFP-Puro)]. All lentiviruses were designed by GenTarget Inc., according to the manufacturer's instructions. Briefly, 1–5×10⁵ cells were seeded in 24-well plates overnight and then treated with hexadimethrine bromide (8 μg/mL) and transduced at MOI of 5. After 3 days, cell selection was undertaken using 4 μg/mL puromycin dihydrochloride (Gibco). ELOVL5 downregulation or upregulation was confirmed by Western blot as described previously (17).

LNCaP or 22Rv1 cells were transfected with siRNA overnight in 24-well plates at a density of 2×10⁴ cells/well, then treated as indicated for 6 days and counted as previously described (17). Fatty acids, arachidonic acid (AA), docosahexaenoic acid (DHA), or cis-vaccenic acid (cVA) were supplemented as indicated for rescue studies.

LNCaP cells transfected with siRNA in 6-well plates were collected after 48 hours and suspended at 7.5×10⁴ cells/mL in RPMI-1640 containing 10% FBS. Cell suspensions (1,500 cells in 20 μL) were pipetted inside the lid of a petri dish and 10–15 mL PBS added to the base. Dishes were reassembled and incubated at 37°C. After 5 days, the formed spheres were imaged, and sphere volume was determined using the equation V = (L × W2)/2.

LNCaP/22Rv1 cells transfected with siRNA were seeded into the upper chamber of 24-well polycarbonate transwell inserts (3422, Sigma-Aldrich) at a density of 3.5 × 10⁴/2 × 10⁵ cells/well in RMPI-1640 containing 1% bovine serum albumin. The bottom chamber was filled with 650 μL of RPMI-1640 containing 5% FBS and cells incubated at 37°C for 48 hours. Transwell membranes were then collected, washed with PBS, and nonmigrated cells were scraped from the upper face. The membranes were fixed in 4% paraformaldehyde for 30 minutes, stained with 1% crystal violet for 1 hour and mounted using permount mounting medium. The number of migrated cells was counted microscopically. Results are reported as the number of migrated cells normalized to the number of viable cells counted in a viability assay at the same time point. For rescue experiments, cells were seeded in serum-free RPMI-1640 containing fatty acids as indicated, and data are presented as number of migrated cells.

LNCaP cells transfected with siRNA were fixed after 72 hours in 1.25% glutaraldehyde, 4% sucrose, 4% paraformaldehyde in PBS (pH 7.2) for 1 hour at room temperature, post-fixed in 1% osmium tetraoxide in water, dehydrated in a graded ethanol series, and embedded in epoxy resin Embed812 (Electron Microscopy Science). Ultrathin sections (70 nm) were cut on a Leica UCT6 ultramicrotome, collected on formvar-coated copper-iridium slot grids, counterstained with 2% uranylacetate and Reynold's lead citrate, and imaged in an FEI Tecnai G2 Spirit transmission electron microscope at 80 kV.

LNCaP cells were transfected with siRNA for 6 days in 24-well plates. For negative controls, cells were treated with 500 μmol/L Trolox (Sapphire Bioscience, NSW, Australia) 24 hours prior to collection, followed by treatment with 200 μmol/L tert-butyl hydroperoxide (TBHP; Thermo Fisher Scientific) for 1 hour prior to collection. For positive controls, cells were treated with 200 μmol/L TBHP for 1 hour prior to collection. Cells were collected into fluorescence-activated sorting tubes and stained with 500 nmol/L CellROX Orange Reagent (Thermo Fisher Scientific) for 1 hour at 37°C. Cells were centrifuged, washed with PBS, and then stained with 1 μmol/L CellROX Orange Reagent (Thermo Fisher Scientific) for 15 minutes on ice. The samples were then read on a BD LSRFortessa X20 flow cytometer.

ELOVL5 stable downregulated (shELOVL5), overexpressing (hELOVL5), or negative control cells (shCTRL) were plated in 6-well plates (3–500 cells/well) and were incubated for up to 3 weeks at 37°C, with medium replenished every 4–7 days. Cells were then washed with PBS, fixed with 4% paraformaldehyde, and stained with 1% crystal violet for 30 minutes. Colonies with >50 cells were counted manually in a blinded manner.

Subcutaneous xenograft studies were approved by the Austin Health Animal Ethics Committee (approval number A2015/05311). LNCaP cells (5 × 10⁶ cells in 50 μL RPMI-1640 containing 10% FBS) were coinjected subcutaneously with 50 μL Matrigel in 5- to 6-week-old NOD/SCID Gamma male mice (Bioresource Facility, Austin Health). When tumors reached ∼200 mm³, mice were randomized to three treatment groups; untreated (n = 5), vehicle control (10% DMSO/PBS; n = 5), and enzalutamide (10 mg/kg enzalutamide in 10% DMSO/PBS, n = 5). Five of the ten castrated mice were treated daily with enzalutamide (10 mg/kg MDV3100 in 10% DMSO/PBS) by oral gavage for 7 days, beginning 5 days after surgery. After the animals were sacrificed, half the tumor was formalin-fixed and paraffin-embedded, and half was analyzed for lipidomics.

Subcutaneous and intraprostatic xenograft studies were approved by the University of Adelaide Animal Ethics Committee (approval number M-2016-160). Eight-week-old male NOD/SCID mice (Animal Resources Centre, Western Australia) received intraprostatic injections (1 × 10⁶ shCTRL or shELOVL5 LNCaP cells in 10 μL of PBS) or subcutaneous flank injections (2 × 10⁶ shCTRL or shELOVL5 LNCaP cells in 50 μL of PBS mixed with 50 μL Matrigel-HC). Each study used 7 animals per group. Noninvasive, whole-body imaging to monitor luciferase-expressing cells was performed at the time of injection and once weekly using the IVIS Spectrum In Vivo Imaging System (PerkinElmer) as described previously (18). Caliper measurements were taken regularly throughout the study. After the animals were sacrificed, internal organs and prostate glands were excised for ex vivo imaging using the IVIS system.

Hematoxylin and eosin (H&E) staining was performed on 2-μm sections that were stained with hematoxylin (Australian Biostain), blued with 1% acid alcohol, and counterstained with eosin (Australian Biostain). IHC was performed on 2-μm sections that were blocked for endogenous peroxidase before being subjected to heat-induced epitope retrieval. Sections were blocked in 10% normal goat serum (Sigma) and incubated with AR primary antibody (1:200, Abcam ab108341) overnight, biotinylated anti-rabbit secondary antibody (1:400, DAKO) for 1 hour, and streptavidin-horseradish peroxidase complex (1:500, DAKO) for 1 hour. Positive cells were visualized with DAB and counterstained with hematoxylin (Australian Biostain). Images were captured on a NanoZoomer digital slide scanner (Hamamatsu).

Statistical analysis was carried out using GraphPad Prism software v7.02 (2016, GraphPad Software), and data are representative of at least two independent experiments. Bar graphs represent the mean ± SEM of at least three biological replicates unless otherwise indicated. Dot plots represent the mean ± SEM of individually counted cells, spheroids, or lipid droplets from at least three biological replicates unless otherwise indicated. Box plots represent the mean ± interquartiles, and whiskers represent minimum and maximum values. For cell line and animal studies, significance was measured by two-tailed unpaired t test or one-way ANOVA with Dunnett multiple comparison test as indicated. For clinical data, significance was measured using Mann–Whitney U test. Significance is expressed as *, P < 0.05; **, P < 0.01; ***, P < 0.001. Heat maps, hierarchical clustering, and Pearson correlations were generated using GENE-E (www.broadinstitute.org/cancer/software/GENE-E/). All heat maps shown are row-normalized for presentation purposes. Pearson correlation with average linkage was used for clustering.

To investigate the influence of androgens on the prostate cancer lipidome, lipids extracted from androgen-treated AR-positive LNCaP prostate cancer cells were subjected to quantitative lipidomic profiling by tandem mass spectrometry (Fig. 1A). In cells treated with low (1 nmol/L) and high (10 nmol/L) physiologic doses of the synthetic androgen mibolerone, total phospholipid abundance increased dose dependently compared with vehicle-treated cells (Fig. 1B). Analysis of the four major phospholipid classes, phosphatyidyl-choline (PC), -ethanolamine (PE), -inositol (PI), and -serine (PS), revealed an increase in PC abundance with the low dose of 1 nmol/L mibolerone, and marked increases in abundance across all four classes with the high dose of 10 nmol/L mibolerone compared with vehicle (Fig. 1C). Androgen treatment did not alter saturated fatty acid (SFA) levels but significantly increased the abundance of monounsaturated fatty acids (MUFA; defined as phospholipids containing 1–2 unsaturations, i.e., one for each fatty acid tail) and polyunsaturated fatty acids (PUFA; defined as phospholipids containing 3 or more unsaturations) at the high dose of 10 nmol/L mibolerone compared with vehicle (Fig. 1D).

The most prominent effect of androgens on the lipidome was elongation of MUFAs and PUFAs across all four phospholipid classes analyzed. Elongation was evident as decreased abundance of short chain phospholipid species and a concomitant increase in long-chain phospholipids with a given number of double bonds, and was observed in LNCaP cells treated with either 1 nmol/L mibolerone compared with vehicle (Supplementary Fig. S1) or 1 nmol/L of the native androgen 5α-dihydrotestosterone (DHT; Fig. 1E; Supplementary Fig. S1). The opposite effect on phospholipid chain length was observed in LNCaP cells treated with the clinical AR antagonist enzalutamide (Fig. 1E; Supplementary Fig. S1). Androgen-induced elongation was also observed in AR-positive LAPC4 and 22Rv1 prostate cancer cells, but not in AR-negative PC-3 prostate cancer cells (Supplementary Fig. S2A), which is consistent with the low level of ELOVL5 expression observed in PC-3 cells compared with AR-positive cell lines (Supplementary Fig. S2B). The clinical relevance of this androgenic effect was addressed by undertaking phospholipidomic profiling in matched prostate cancer PDEs treated with enzalutamide or vehicle (n = 19; Fig. 1F). The major trend observed for the majority (17/19; 89.4%) of enzalutamide-treated PDEs was inhibition of fatty acid chain elongation in MUFAs and PUFAs with up to 4 unsaturations (Fig. 1G; Supplementary Fig. S3). Taken together, we have demonstrated that androgens have a significant effect on phospholipid composition in prostate cancer cells and tumors, whereby androgens induce and enzalutamide inhibits fatty acid chain elongation (Fig. 1H).

Consistent with the effects of androgens on fatty acyl chain length, we observed androgenic induction of three ELOVL fatty acid elongases (ELOVL) in LNCaP, LAPC4, and 22Rv1 prostate cancer cells, as well as induction of the known AR-regulated gene FASN (Fig. 2A). The most marked androgen-induced increases were observed for ELOVL2, ELOVL5, and ELOLV7, which was also evident in an independent, published LNCaP RNA-seq data set (Fig. 2B; ref. 19). Expression of these three ELOVLs was strongly associated with androgen-induced prostate cancer gene sets (Nelson, Fig. 2C; Wang, Supplementary Fig. S4A; refs. 20, 21). As expected, the AR antagonist enzalutamide significantly decreased ELOVL2, ELOVL5, and ELOLV7 expression compared with vehicle in LNCaP cell lines (Fig. 2D) and LNCaP xenograft tumors (Fig. 2E). Expression of ELOVL2 was not detected in the LNCaP xenografts. Importantly, ELOVL2, ELOVL5, and ELOVL7 were repressed by androgen deprivation in prostatic biopsies from patients with locally advanced or metastatic prostate cancer after short-term neoadjuvant ADT (n = 7; Fig. 2F; ref. 22), and in radical prostatectomy samples from neoadjuvant intense ADT patients (n = 18; Supplementary Fig. S4B; ref. 23). Previous studies suggested that androgens may mediate ELOVL expression indirectly through activation of the master lipogenic transcription factor, sterol regulatory element-binding protein (SREBP1; ref. 24), or directly via AR transcriptional activity (25). To explore regulation through SREBP1, we knocked down SREBP1 in LNCP cells in the presence and absence of androgens (Supplementary Fig. S4C; 84% decrease in gene expression with siSREBP1 compared with siCTRL). As expected, siSREBP1 inhibited androgen induction of FASN expression, but was unable to suppress androgen induction of ELOVL2, ELOVL5, or ELOVL7 (Supplementary Fig. S4C), which indicates that SREBP1 is not the primary mediator of androgen-induced ELOVL expression. To explore direct AR regulation, we examined AR ChIP-seq data from clinical samples (13). Multiple AR binding sites proximal to the ELOVL2, ELOVL5, and ELOVL7 genes were evident in human prostatic tissues, many of which were enriched in malignant compared with normal tissues (Fig. 2G; Supplementary Fig. S5). Supporting their relevance, these AR binding sites were coincident with histone H3 lysine 27 acetylation (H3K27ac), a mark of active promoters and enhancers. We provide evidence of these AR binding sites in many published ChIP-seq data sets using ChIP-Atlas (Supplementary Fig. S6A–S6C), and validated AR binding at a subset of these sites by ChIP-qPCR in LNCaP cells (Fig. 2H). Collectively, these data indicate that AR binds to multiple sites proximal to the transcriptional start sites of the ELOVL2, ELOVL5, and ELOVL7 genes, supporting a direct mode of AR-mediated transcriptional regulation.

To assess the relative contributions of ELOVL2, ELOVL5, and ELOVL7 to the androgen-induced elongation phenotype, siRNA-mediated knockdown of all three ELOVLs in combination or alone was performed in LNCaP cells. Two independent siRNAs were tested per ELOVL gene in LNCaP cells, each achieving at least 80% knockdown within 48 hours (Supplementary Fig. S7A). We observed no evidence of ELOVL2, ELOVL5, or ELOVL7 knockdown alone causing compensatory increases in expression of the other ELOVLs (Fig. 3A). When compared with siCTRL, complete reversal of the androgen-mediated increase in longer fatty acid chains was observed in all four phospholipid classes with simultaneous knockdown of ELOVL2/5/7 in LNCaP cells (Fig. 3B; Supplementary Fig. S7B). Notably, almost identical changes to the combined ELOVL2/5/7 knockdown were achieved by ELOVL5 knockdown alone (Fig. 3B; Supplementary Fig. S7B). ELOVL2 and ELOVL7 knockdown showed no discernible effect, implicating ELOVL5 as the primary contributor to androgen regulation of phospholipid fatty acyl chain elongation in LNCaP cells. Supporting ELOVL5 as the major elongase in prostate cancer, ELOVL5 was the most highly expressed of all seven ELOVL genes in two independent transcriptomic data sets, derived from 419 localized primary prostate tumors generated by The Cancer Genome Atlas (TCGA) Research Network (http://cancergenome.nih.gov/; Fig. 3C) and 118 metastatic castration-resistant prostate tumors from the Stand Up to Cancer (SU2C) data set (Fig. 3D). Given the high level of ELOVL5 expression in prostate cancer, we investigated whether this is aberrant expression compared with normal prostate tissue. Data from three clinical cohorts were evaluated, including transcriptomic data sets from MSKCC (Fig. 3E), TCGA (Fig. 3F), and a proteomic data set from Tampere University Hospital (Supplementary Fig. S8A; ref. 26). In all three cohorts, ELOVL5 expression was significantly higher in prostate cancer compared with normal/benign prostatic hyperplasia. Further supporting these observations, ELOVL5 protein levels were analyzed by IHC in five tissue microarrays (TMA) comprised of 249 primary prostate cancer samples. DAPI-normalized ELOVL5 intensity was significantly increased in tumor compared with adjacent benign prostate epithelial cells in the entire cohort (Fig. 3G and H), or when each TMA was independently evaluated (Supplementary Fig. S8B).

The significant role of ELOVL5 in modifying the cellular lipidome suggests that it may contribute to cellular proliferation and energetic processes. Indeed, transient knockdown of ELOVL5 by two independent siRNAs in LNCaP (Supplementary Fig. S7A) and 22Rv1 cells (Supplementary Fig. S8C) significantly attenuated viability of both prostate cancer cell lines (Fig. 4A and B), but did not induce death (Fig. 4A and B) or apoptosis (Supplementary Fig. S8D). In contrast, constitutive overexpression of ELOVL5 (hELOVL5⁺) in LNCaP cells significantly increased viable cell number compared with control cells (hCTRL; Fig. 4C). LNCaP colony formation was affected in a similar manner, where stable knockdown of ELOVL5 with short hairpin RNA (shELOVL5⁻) significantly reduced colony formation (Fig. 4D) while stable overexpression of ELOVL5 increased colony formation (Fig. 4E). Given we observed no impact of ELOVL5 depletion on cell death, we evaluated proliferation as the driver of ELOVL5-mediated increase in cell viability. Transient knockdown of ELOVL5 with siRNA significantly attenuated LNCaP and 22Rv1 cell proliferation compared with control siRNA, as determined by a Cyquant assay (Fig. 4F), which was associated with cell-cycle arrest at the G₁–S phase (Fig. 4G). Encouraged by the antiproliferative effects of ELOVL5 knockdown in prostate cancer cell lines, we again used our PDE model to determine efficacy of ELOVL5 targeting in clinical prostate tumors. Significant ELOVL5 depletion (44.7%) was achieved in PDE tissues using siRNA-loaded nanoparticles (Fig. 4H; Supplementary Fig. S8E), and this resulted in a marked decrease in expression of the proliferative marker ki67 from 52.9% in siCTRL PDE tissues to 21.06% in siELOVL5 PDE tissues (Fig. 4H; Supplementary Fig. S8E). Finally, reduced spheroid growth was observed in siELOVL5-depleted LNCaP cells compared with siCTRL (Fig. 4I), and constitutively expressing shELOVL5 LNCaP cells showed significantly reduced capacity for tumor growth compared with shCTRL cells when grown subcutaneously in NOD/SCID mice (n = 7 mice/treatment; Fig. 4J).

Published phenotypic studies of ELOVL5 knockout mice demonstrate that hepatic steatosis induced by ELOVL5 knockout can be rescued by supplementation with the PUFAs AA (C20:4, n-6) and docosahexaenoic acid (DHA; C22:6, n-3; ref. 27), as these PUFAs are reliant on ELOVL5 for synthesis (Fig. 5A). We found that although AA and DHA levels were reduced in ELOVL5-depleted LNCaP cells (Supplementary Fig. S9A), supplementation with AA or DHA did not alter ELOVL5 knockdown-mediated changes in cell viability (Fig. 5B). Also distinct from the mouse model, ELOVL5 knockdown caused no induction of key lipid metabolism enzymes (SREBP1, FASN, ACACA, and ATGL; Supplementary Fig. S9B), nor any alteration to cellular lipid content, as determined by the number of lipid droplets observed per cell by electron microscopy (Supplementary Fig. S9C). These findings indicated that the role of ELOVL5 in prostate cells differs substantially from those in the liver. In considering other known products of ELOVL5, we tested an omega-7 MUFA, cis-vaccenic acid (cVA; C18:1, n-7), which has previously been implicated in ELOVL5 action (Fig. 5A; ref. 28). Remarkably, supplementation with cVA completely reversed the inhibitory effect of ELOVL5 knockdown on cell viability (Fig. 5C), indicating MUFA elongation as a key ELOVL5 function in prostate cancer.

Electron microscopy examination of ELOVL5-depleted LNCaP cells showed a marked increase in the size and length of the cellular mitochondria compared with control cells, as well as darker and denser cristae (Fig. 6A), indicative of cellular stress. Further investigation of the mitochondria was undertaken using the Seahorse mitochondrial stress test assay. Decreased basal and maximal respiration rates were observed in ELOVL5-depleted LNCaP cells (Fig. 6B), whereas ELOVL5 overexpression increased maximal mitochondrial respiration (Fig. 6C). A decrease in mitochondrial respiration, as observed upon ELOVL5 knockdown, is a strong indicator of mitochondrial dysfunction and can result in aberrant generation of reactive oxygen species (ROS; ref. 29). The fluorogenic dye CellROX Orange was used to investigate whether ELOVL5 targeting influences ROS production. Markedly increased ROS levels were observed upon ELOVL5 knockdown in LNCaP cells, which was completely reversed by the potent antioxidant/ROS scavenger Trolox (Fig. 6D). Trolox also rescued the inhibitory effects of ELOVL5 depletion on LNCaP cell viability (Fig. 6E), revealing ROS production as a key antiproliferative mechanism for ELOVL5 targeting. Consistent with these results, ELOVL5-mediated changes in ROS production were also rescued by supplementation with cVA (Fig. 6E). Confirming that ROS induction was not part of a generalized stress response, stable overexpression of ELOVL5 markedly reduced ROS levels in LNCaP cells (Fig. 6F).

Having established a critical role for ELOVL5 in prostate cancer cell proliferation, we next investigated its influence on metastatic tumor features in vitro and in vivo. ELOVL5 knockdown significantly reduced transwell migration through Matrigel in both LNCaP (Fig. 7A) and 22Rv1 cells (Fig. 7B), whereas ELOVL5 overexpression promoted LNCaP migration (Fig. 7C). Importantly, supplementation of ELOVL5-depleted LNCaP cells with cVA reversed the migratory phenotype (Fig. 7D), reinforcing a critical role for this MUFA in prostate cancer. To determine whether ELOVL5 influences spontaneous metastasis, we used an orthotopic assay in which LNCaP cells were injected directly into the prostate (i.e., intraprostatic xenografts). Weekly monitoring of the mice revealed that tumor incidence was 100% for both shELOVL5 and shCTRL cells, but the shELOVL5 tumors grew significantly more slowly and measured almost 40% smaller than shCTRL tumors after 50 days (Fig. 7E). Ex vivo bioluminescent imaging of the major organs following intraprostatic tumor growth revealed that shCTRL cells spread to major visceral organs, most notably the lungs, in 4 of 7 mice (Fig. 7F). Histologic evidence of LNCaP tumor metastases to the lung was demonstrated by staining for AR and H&E (Fig. 7F). Significantly, no overt metastasis by shELOVL5 cells was detected in any mice (Fig. 7F).

The advent of high-throughput “omics” approaches has revolutionized the molecular characterization of cancer, but in contrast to the genome, transcriptome and proteome, the lipidome remains largely underexplored. We utilized tandem mass spectrometric analysis to profile the androgen-regulated lipidome of prostate cancer cell lines and PDEs and uncovered ubiquitous increases in fatty acyl chain elongation mediated by the fatty acid elongase, ELOVL5. Molecular targeting of ELOVL5 reversed fatty acyl chain elongation in prostate cancer cells, which through induction of mitochondrial oxidative stress, significantly reduced prostate tumorigenesis and metastatic potential in vitro and in vivo. Moreover, we demonstrate that the monounsaturated fatty acid product of ELOVL5 activity, cVA, is a critical modulator of ELOVL5-induced oxidative stress and survival in prostate cancer. Our findings provide new insights into the downstream metabolic actions of AR in prostate cancer cells, beyond the previous almost exclusive focus on de novo lipogenesis, and identify fatty acid elongation as a critical target of androgen action for therapeutic intervention.

Elongation is a critical step in the production of fatty acids with a chain length of more than 16 carbons, and is a stepwise process whereby carbon units are added to the carboxyl terminal of fatty acid chains (30). The reaction is catalyzed by enzymes called elongases, encoded by a family of seven ELOVL genes (ELOVL1–7) that display distinct tissue distribution and substrate preference (31). ELOVL5 is most highly expressed in the adrenal glands, testis, and prostate (32). Increasing evidence suggests that ELOVL expression is frequently altered in cancer. ELOVL6 is the most commonly studied, with reported overexpression in hepatocellular cancer (33), lung cancer (34), and breast cancer (35). In prostate cancer, ELOVL7 has been reported as overexpressed in hormone-naïve prostate cancer (36, 37), with influences on cell viability, xenograft tumor growth, and potentially de novo steroidogenesis (36, 38); however, we found its abundance in clinical prostate tumors to be very low. When directly comparing all seven ELOVL family members at both the RNA and protein levels, ELOVL5 was the most highly expressed in both primary and castration-resistant clinical prostate tumors, and was overexpressed in malignant compared with nonmalignant prostate. In published overexpression studies in primary rat hepatocytes, adenoviral overexpression of ELOVL5 induced elongation of C20 PUFAs and gave rise to altered hepatic fatty acid content (39). The reverse was observed in ELOVL5 knockout mice where inhibition of C20 and C22 PUFA elongation and concomitant increase in C18 PUFAs was observed, and resulted in development of hepatic steatosis (27). Unfortunately, the effect of ELOVL5 ablation on the mouse prostate was not reported, but we likewise observed inhibition of fatty acid elongation in prostate cancer cells with ELOVL5 knockdown. Phenotypically, targeting ELOVL5 had profound effects on prostate cancer cell proliferation, colony formation, spheroid growth, and invasion through Matrigel, in preclinical models that range from typical cell lines to complex 3D cultures and PDEs. The latter provides the most physiologically and clinically relevant preclinical model currently available, as the critical tumor microenvironment remains intact (10). The therapeutic efficacy of ELOVL5 depletion was also significant in vivo where primary tumor growth was suppressed, and spontaneous metastasis to distant sites was restricted by the knockdown.

Demonstrated in this study is a previously unknown relationship between ELOVL5-driven lipid elongation in prostate cancer, mitochondrial homeostasis, and intracellular ROS levels. ELOVL5′s dual effects on mitochondrial respiratory activity and intracellular ROS levels are consistent, given mitochondria are a major source of intracellular ROS (40). Importantly, our data indicate that ROS induction is not a nonspecific cellular stress response to ELOVL5 targeting, as ectopic ELOVL5 overexpression markedly suppresses ROS production. It is well established that ROS production is elevated in cancer cells, as a consequence of increased metabolic demand, mutation rate, and hypoxia, as a means of promoting cancer cell proliferation and tumor progression (41). Although these effects are beneficial for cancer cell growth and survival, the increase in baseline ROS also represents a vulnerability to further oxidative stress, as high levels of ROS are cytotoxic (42). Unlike their benign counterparts with low baseline ROS levels, small increases in ROS production can push cancer cells over the “cytotoxic ROS threshold” (41) to suppress cell survival and/or induce cell death. The role of ELOVL5 in prostate cancer appears to be, at least in part, as a regulator of cellular redox homeostasis, to ensure that prostate cancer cells do not exceed this “cytotoxic ROS threshold,” whilst maintaining appropriate ROS production to drive tumorigenesis. Although an increasing number of cancer treatments target ROS as a therapeutic strategy, including both chemotherapies and molecular targeted agents (43), specific targeting of lipid metabolism through ELOVL5 now provides a new avenue for ROS induction as a treatment for prostate cancer.

One of the most significant findings of this study was rescue of ELOVL5-mediated proliferation, metastasis, and ROS induction by supplementation with the monounsaturated product of ELOVL5, cis-vaccenic acid, but not by supplementation with long-chain polyunsaturated fatty acids. This was an unexpected, because ELOVL5 was previously considered to be mainly involved in production of PUFAs rather than MUFAs, and the main hepatic phenotype of the ELOVL5 knockout mouse was fully rescued by PUFA supplementation. In light of recent reports of the critical importance of MUFAs in maintaining redox homeostasis in cancer cells (44), our data suggest that cis-vaccenic acid production underpins this aspect of ELOVL5 activity, and may serve to reduce oxidation of ELOVL5-derived PUFAs. Although it is the least studied of ELOVL5′s products, cis-vaccenic acid has been reported to be linked to coronary heart disease risk when incorporated into red blood cell membranes (45), has displayed antilipogenic effects in bovine adipocytes (46) and its ratio with its precursor has been suggested as an indicator of biotin deficiency (47). A study undertaken in HepG2 (kidney) cells suggested that cis-vaccenic was the mediator of ELOVL5 regulation of the mTORC2–Akt–FOXO1 pathway (28). Importantly, its role in metabolism is largely unknown (46) and there is little to no published data on its role in cancer, highlighting this as an important focus for future investigations.

Important for the potential of ELOVL5 as a therapeutic target, the murine ELOVL5 knockout was not lethal. This is in contrast to the master fatty acid synthesis enzyme FASN, which is lethal when knocked out in mice (48), and suggests that targeted inhibition of ELOVL5 could have a higher therapeutic window than inhibition of FASN—currently the most widely studied metabolic target in cancer. Moreover, loss of ELOVL5 did not elicit changes to the expression of other ELOVL family members, key lipid synthetic enzymes, or to cellular fatty acid content. These findings further highlight the potential of ELOVL5 as a therapeutic target, as compensatory adaptive responses that can mitigate efficacy have frequently been observed with investigative metabolic drugs and limit their clinical application. For example, FASN inhibitors target de novo lipid synthesis but can result in compensatory increases in uptake of exogenous dietary lipids or expression of genes in the fatty acid synthesis pathway including ACACA, ACLY, as well as FASN itself, to satisfy the cell's demand for fatty acids (49, 50).

Transcriptomic studies have suggested for many years that lipid synthesis is enriched by AR activity in prostate cancer but the explicit details of how this contributes to tumorigenicity have remained unclear. Herein we provide a specific new functional role for the AR in maintaining lipid homeostasis via ELOVL5-mediated elongation of fatty acids, and demonstrate the significant impact of elongation on prostate cancer cell malignancy and metastasis. The antimetastatic effects of ELOVL5 knockdown are particularly notable given that metastatic prostate cancer remains a major clinical challenge and is associated with a high rate of death despite multiple new drug approvals in recent years. Based on the findings in this study, targeting specific lipidomic features of prostate tumor cells such as elongation is a promising therapeutic strategy that warrants further investigation.

M.M. Centenera reports grants from Prostate Cancer Foundation of Australia during the conduct of the study. J. Machiels reports a patent for WO2012038525 licensed. I.J. Burvenich reports grants from Movember Foundation and Prostate Cancer Foundation of Australia during the conduct of the study. C.C. Bango reports other employment from Glaxosmithkline outside the submitted work. A.M. Scott reports other funding from AbbVie, EMD Serono, ITM, Medimmune, AVID, Telix, Adalta, Cyclotek, and Theramyc, and personal fees from Life Science Pharmaceuticals and Imagion, outside the submitted work. L.A. Selth reports grants from Cancer Council SA during the conduct of the study. J.V. Swinnen reports grants from Prostate Cancer Foundation of Australia, Research Foundation–Flanders, Stichting tegen Kanker, KU Leuven, and Interreg V-A EMR23 “EURLIPIDS” during the conduct of the study; in addition, J.V. Swinnen had a patent for WO2012038525 issued. L.M. Butler reports grants from Movember Australia, Australian Research Council, Cancer Council South Australia, Prostate Cancer Foundation of Australia, and U.S. Department of Defense during the conduct of the study. No disclosures were reported by the other authors.

M.M. Centenera: Conceptualization, investigation, writing–original draft, writing–review and editing. J.S. Scott: Investigation. J. Machiels: Investigation. Z.D. Nassar: Investigation. D.C. Miller: Investigation. I. Zininos: Investigation. J. Dehairs: Investigation. I.J.G. Burvenich: Investigation. G. Zadra: Investigation. P.M. Chetta: Investigation. C. Bango: Validation and investigation. E. Evergren: Validation and investigation. N.K. Ryan: Investigation. J.L. Gillis: Investigation. C.Y. Mah: Investigation. T. Tieu: Investigation. A.R. Hanson: Investigation. R. Carelli: Investigation. K. Bloch: Investigation. V. Panagopoulos: Investigation. E. Waelkens: Investigation. R. Derua: Investigation. E.D. Williams: Investigation. A. Evdokioou: Investigation. A. Cifuentes-Rius: Writing–review and editing. N.H. Voelcker: Writing–review and editing. I.G. Mills: Writing–review and editing. W.D. Tilley: Writing–review and editing. A.M. Scott: Writing–review and editing. M. Loda: Writing–review and editing. L.A. Selth: Investigation, writing–review and editing. J.V. Swinnen: Conceptualization, supervision, writing–review and editing. L.M. Butler: Conceptualization, supervision, writing–review and editing.

The authors thank the study participants, urologists, nurses, histopathologists, and the South Australian Coordinator of the Australian Prostate Cancer Bioresource; Jessica Savage, who assisted in the recruitment and collection of patient material and information; Swati Irani for assistance with explant tissue culture and analysis; Dr. Ying Huang for immunostaining the tissue microarrays; the University of Adelaide Microscopy facility for technical assistance with electron microscopy. The authors acknowledge financial support from the Movember Foundation and Prostate Cancer Foundation of Australia (MRTA3 to L.M. Butler, J.V. Swinnen, W.D. Tilley, A.M. Scott, L.A. Selth, M.M. Centenera; MRTA1 to E.D. Williams), Prostate Cancer Foundation of Australia (NDDA2711 to L.M. Butler, M.M. Centenera, J.V. Swinnen; NCG1816 to L.M. Butler, A. Cifuentes-Rius, N.H. Voelcker), the Research Foundation–Flanders (FWO G.0841.15 to J.V. Swinnen), the Stichting tegen Kanker (to J.V. Swinnen), KU Leuven (C16/15/073 and C32/17/052 to J.V. Swinnen), Interreg V-A (EMR23 to J.V. Swinnen and E. Waelkens), the National Health and Medical Research Council (1121057 to W.D. Tilley and L.A. Selth), the National Institute of Health (RO1CA131945 to M. Loda), the NCI (P50CA211024 to M. Loda), the Prostate Cancer Foundation (to M. Loda), the United States Department of Defense (PC150263 to G. Zadra; PC160357 to M. Loda; PC180582 to M. Loda, L.M. Butler, J.V. Swinnen), and the Dana-Farber Cancer Institute (Claudia Adams Barr Award in Innovative Basic Cancer Research to G. Zadra). M.M. Centenera was supported by a Prostate Cancer Foundation of Australia Young Investigator Award (YIG0412). Z.D. Nassar was supported by National Health and Medical Research Center Early Career Fellowship (1138648) and Prostate Cancer Foundation of Australia John Mills Young Investigator Award (YI1417). T. Tieu and N.H. Voelcker acknowledge support from the Office of the Chief Executive and CSIRO Manufacturing. T. Tieu acknowledges support of an Australian Government RTP scholarship. A. Cifuentes-Rius was supported by a National Health and Medical Research Council Early Career Fellowship (1112432). L.M. Butler was supported by an ARC Future Fellowship (130101004). L.M. Butler and L.A. Selth are supported by Principal Cancer Research Fellowships awarded by Cancer Council's Beat Cancer project on behalf of its donors, the state government through the Department of Health, and the Australian Government through the Medical Research Future Fund (PRF1117 and PRF2919).

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.