Targeting ASCT2‐mediated glutamine uptake blocks prostate cancer growth and tumour development

Patient specimens

Prostate cancer specimens from radical prostatectomy (n = 194) were obtained from the Vancouver Prostate Centre Tissue Bank (http://www.prostatecentre.com/our‐research/core‐facilities/biorepository). This project was approved by the institutional review boards at the University of British Columbia (Vancouver, Canada) and the CHUQ Research Centre (Québec, Canada). Written informed consent was obtained from all participants. The haematoxylin and eosin (H&E)‐stained slides were reviewed and desired areas were identified on paraffin blocks. Tissue microarrays (TMAs) were manually constructed (Beecher Instruments, MD, USA) by punching duplicate cores of 1 mm for each sample, with quantitative analysis calculated from individual cores (individual Gleason scores) or the average of duplicate cores (NHT TMA analysis).

TMA immunohistochemistry

TMA staining was conducted using a Ventana auto‐stainer, model Discover XT™, with an enzyme‐labelled biotin–streptavidin system and solvent‐resistant DAB Map kit, using a 1:4000 dilution of rabbit anti‐SLC1A5 (HPA035240). Staining was graded by a pathologist (LF), using a semi‐quantitative, four‐point scale, in which 0 represents no staining, 1 faint or focal staining, 2 moderate intensity in at least a quarter of neoplastic cells, and 3 intense staining in the majority of neoplastic cells. Examples of the staining are shown in Figure S1B (see supplementary material).

Cell culture and uptake assay

Human prostate cancer cell lines LNCaP‐FGC, PC‐3 and DU145 were purchased from ATCC (Rockville, MD, USA). LNCaP cells had been passaged directly from original low‐passage stocks (2009), and we confirmed PC‐3, LNCaP and DU145 cell identity by STR profiling in 2010 and 2014 (Cellbank, Sydney). Cells were cultured in RPMI 1640 medium containing 10% v/v fetal bovine serum (FBS), penicillin–streptomycin solution and 1 mm sodium pyruvate. Cells were maintained at 37 °C in an atmosphere containing 5% CO₂. Chemicals were diluted as follows, with control wells treated with the appropriate vehicle controls: H‐Ser(Bzl)‐OH (BenSer; Bachem; diluted in H₂O); l‐γ‐glutamyl‐p‐nitroanilide (GPNA; MP Biochemicals; diluted in H₂O); bicalutamide (Astra Zeneca; diluted in DMSO). The [³H]‐l‐leucine and [³H]‐l‐glutamine uptake assays were performed as detailed previously 18, 32.

Knockdown of ASCT2

ASCT2 shRNA knockdown was performed as previously described 18. Two different shRNAs for ASCT2 were used in this study (Sigma): shASCT2#1 (CCGGGCTGCTTATCCGCTTCTTCAACTCGAGTTGAAGAAGCGGATAAGCAGCTTTTTG) and shASCT2#2 (CCGGCTGGATTATGAGGAATGGATACTCGA GTATCCATTCCTCATAATCCAGTTTTTG).

Gene expression analysis

PC‐3 cells were incubated in the presence or absence of GPNA or BenSer for 48 h, the cells were harvested and total RNA was extracted from them using Trizol. RNA quality was confirmed (RIN values 9.8–10), using RNA 6000 Nano Chips on an Agilent 2100 Bioanalyser (Agilent Technologies). Libraries were prepared using TruSeq Stranded Total RNA kit (Illumina, San Diego, CA, USA) and paired‐end sequencing was performed on an Illumina HiSeq 2500 at the Kinghorn Cancer Centre (Sydney). Paired‐end RNA‐sequencing reads were trimmed and mapped to annotated human genome (hg19/GRCh37.p13), using Tophat2 with default settings. For each gene, the number of reads were counted and normalized to the library size. The threshold for expression was set at five reads in at least one experimental group. Exact tests were applied to compare differences in the means of read‐counts from four replicates between the treated and untreated groups. The gene expression levels were estimated counts/million reads (CPM). The National Center for Biotechnology Information Gene Expression Omnibus number for mRNA sequencing datasets described in this study is GSE65112.

PC‐3–luc xenografts and bioluminescence imaging

Athymic (nu/nu) male nude mice (Animal Resource Centre, Perth, Australia), aged 6–8 weeks, were housed in a specific pathogen‐free facility, in accordance with the University of Sydney animal ethics committee guidelines. Mice were anaesthetized via 2% isoflurane inhalation and received subcutaneous (s.c.) injections of 1 × 10⁶ PC‐3–luc cells resuspended in 100 ml Hanks' balanced salt solution (HBSS). Xenografts were transplanted in both the right and left ventral flanks of mice, as detailed previously 30. Tumour growth was monitored via bioluminescence imaging performed 48 h following cell implantation and biweekly thereafter for 32 days. During the experiments, two shASCT2 mice were dead 4 days after injection, due to fighting; they were not included in the analyses. Anaesthetized mice received intraperitoneal (i.p.) injections of d‐luciferin substrate (150 mg/kg in DPBS; Gold Biotechnology) and images were acquired after a 15 min interval, using the Xenogen in vivo imaging system (IVIS) Lumina II (Caliper Life Science, MA, USA). Regions of interest were determined using Living Image software (Caliper Life Science) and quantified in photons/s (p/s). After 32 days, the animals were sacrificed following the final imaging time point. Livers and lungs were removed for IVIS‐Lumina II analysis to detect spontaneous metastases. After being imaged and weighed, tumours were collected in either Trizol for RNA analysis or lysis buffer for western blotting analysis, or fixed in 10% v/v neutral buffered formalin for sectioning and immunostaining.

Statistical analysis

Data are expressed as mean ± SEM. All experiments were performed with at least three replicate experiments and analysed using a Mann–Whitney U‐test or one‐way ANOVA in GraphPad Prism v. 6. MTT assay and tumour growth curve analysis used the two‐way ANOVA test. Fisher's exact test was used in tumour metastasis analysis. All statistical tests were two‐sided.

ASCT2 expression is increased in prostate cancer

To determine whether ASCT2 expression is increased in prostate cancer compared to normal tissue, we compared patient‐matched TCGA mRNA expression data for normal prostate versus prostate cancer (Figure 1A). There was a significant increase in ASCT2 expression in tumour samples compared to normal (pairwise t‐test, p = 0.025), with 24/36 patients showing increased tumour ASCT2 levels compared to patient‐matched normal prostate tissue (χ² test, p = 0.046). Analysis of Oncomine datasets showed significant increases in ASCT2 expression (p < 0.05) in seven prostate cancer datasets (Table 1). To determine ASCT2 protein levels, we performed immunohistochemical staining on a human prostate cancer tissue microarray (TMA; Figure 1B). At all stages of prostate cancer, ASCT2 protein was detected in the membrane of the prostate epithelial cells, but not in the cytoplasm (see supplementary material, Figure S1A). There were no significant differences in expression between Gleason grades in either the TMA or the TCGA dataset (see supplementary material, Figure S1B, C). Comparing untreated with patients undergoing neoadjuvant hormone therapy (NHT), a significant decrease in ASCT2 protein expression was observed during 1–6 and 7–12 months of treatment (p = 0.04; p < 0.001; Figure 1C). This supports previous data showing that ASCT2 expression is regulated by the androgen receptor 23. Interestingly, ASCT2 expression significantly increased in patients with recurrent prostate cancer (p < 0.001; Figure 1C), indicating either re‐activation of androgen receptor signalling or that other signalling pathways might up‐regulate ASCT2 expression upon disease recurrence. Analysis of an LNCaP xenograft model 24 at various stages precastration (intact), postcastration during regression, at the nadir of PSA, and in recurrent and castration‐resistant prostate cancer (CRPC) further confirmed this dynamic regulation of ASCT2 expression. ASCT2 expression decreased significantly post‐castration (regression and nadir), before increasing in recurrent and CRPC, where ASCT2 expression was no longer significantly different from intact ASCT2 levels (Figure 1D). These human and xenograft data suggest that ASCT2 levels may be important, both in the primary tumour and during tumour recurrence and CRPC.

Inhibition of ASCT2 suppresses prostate cancer cell growth

Analysis of ASCT2 protein showed high expression in both LNCaP and PC‐3 cells, with lower levels detected in DU145 cells (Figure 2A). Addition of the ASCT2 competitive inhibitor benzylserine (BenSer; see supplementary material, Figure S2A) 25 significantly reduced both glutamine and leucine uptake in LNCaP, PC‐3 and DU145 cells (Figure 2B, C). Leucine uptake inhibition occurs due to exchange of glutamine for leucine by l‐type amino acid transporter 1 (LAT1; SLC7A5), but may also be through direct LAT1 inhibition 15. Inhibition of androgen receptor signalling by bicalutamide also significantly reduced glutamine uptake in androgen‐sensitive LNCaP cells, but not androgen‐insensitive PC‐3 cells (see supplementary material, Figure S2B, C), confirming that AR signalling regulates ASCT2 levels 23 and ASCT2‐mediated glutamine transport. After 3 days of treatment with BenSer, cell growth was significantly decreased in LNCaP, PC‐3 and DU145 cells (Figure 2D). We have previously shown that LAT1 inhibition affects the cell cycle through E2F‐regulated cell cycle proteins CDK1, CDC20 and UBE2C 23. These three cell cycle‐regulatory proteins are also significantly increased in prostate cancer metastasis 23. CDK1, CDC20 and UBE2C exhibited decreased expression after BenSer treatment (Figure 2E).

Inhibition of ASCT2 suppresses glutamine metabolism in prostate cancer

As leucine and glutamine are involved in the activation of mTORC1 signalling 14, 26, 27, we examined T389 phosphorylation of p70S6K, a downstream target of mTORC1. BenSer reduced the phosphorylation of p70S6K in PC‐3 cells, but not in LNCaP cells (Figure 2F). Utilizing another ASCT2 inhibitor, l‐γ‐glutamyl‐p‐nitroanilide (GPNA; see supplementary material, Figure S2D)28, confirmed that glutamine deprivation inhibits the mTORC1 pathway in PC‐3 cells, but not LNCaP cells (Figure 2F). These data suggest that mechanisms other than mTORC1 must also contribute to regulate cell growth in prostate cancer.

The majority of intracellular glutamine is converted into glutamate, which can be utilized by the TCA cycle for ATP and fatty acid production, or exported from the cell by xCT (SLC7A11) in exchange for cystine, thereby contributing to glutathione production and protection against oxidative stress. Alternatively, glutamine acts as a nitrogen and carbon donor for macromolecular synthesis, including nucleotides and non‐essential amino acids/proteins. We therefore set out to determine how changes in intracellular glutamine affect these downstream pathways. Using the Seahorse flux analyser, BenSer or GPNA treatment resulted in a reduction in basal oxygen consumption rates (OCRs) compared to control in LNCaP and PC‐3 cells (see supplementary material, Figure S2E, F). GPNA, but not BenSer, significantly reduced basal OCR in both LNCaP and PC‐3 cells (Figure 3A, B). However, this reduction in OCR did not appear to result directly from glutamine usage, as there was no significant reduction in oxidized ¹⁴C‐labelled glutamine in either cell line after BenSer or GPNA (Figure 3C, D), despite lower ¹⁴C‐glutamine uptake (see supplementary material, Figure S2G, H). Lipid synthesis from glutamine, however, was significantly decreased in LNCaP cells (Figure 3E), but not in PC‐3 cells (Figure 3F). These data suggest that LNCaP cells directly utilize glutamine for fatty acid synthesis, while PC‐3 cells may utilize glutamine for other pathways, including exchange for leucine (Figure 2C) to activate mTORC1 signalling (Figure 2F), which indirectly affects mitochondrial respiration.

To further investigate lipid homeostasis, we measured the levels of cellular phospholipids and neutral lipids. Lipid content analysis of LNCaP cells revealed that GPNA significantly reduced the cellular level of phospholipids (Figure 3G) and BenSer and GPNA reduced the neutral lipids (Figure 3I) as well as lipid droplet number (Figure 3K). Notably, the magnitude of inhibition by BenSer and GPNA was similar to thast of 5‐(tetradecyloxy)‐2‐furoic acid (TOFA), which inhibits the rate‐limiting enzyme of fatty acid synthesis, acetyl‐CoA carboxylase (Figure 3G, I, K). A significant decrease of phospholipids was also observed in PC‐3 cells (Figure 3H); however, there were no significant decreases in neutral lipids (Figure 3J) or lipid droplet numbers (Figure 3L). Taken together, pharmacological inhibition of ASCT2 provoked cell line‐specific effects on the flux of glutamine as a fuel source, yet ultimately led to reduced OCR and cellular lipid levels in both prostate cancer cell lines. However, neither BenSer nor GPNA led to increased levels of reactive oxygen species (ROS; see supplementary material, Figure S2I).

Global effects of BenSer or GPNA inhibition in PC‐3 cells

To determine the global effects of ASCT2 inhibition, we used next‐generation sequencing to determine mRNA expression changes in PC‐3 cells treated with BenSer or GPNA for 48 h. Analysis of genes significantly (p < 0.05) up‐ or down‐regulated by GPNA (see supplementary material, Table S1) compared to control showed a substantial overlap (up‐regulated, 45.1%; down‐regulated, 49.7%) with BenSer gene expression, highlighting common targets of ASCT2 (Figure 4A). BenSer appears to have additional non‐ASCT2‐mediated affects, with more genes showing significant up‐ and down‐regulation (see supplementary material, Table S2) compared to GPNA (Figure 4A). This is not surprising, since BenSer has been shown to inhibit both ASCT2 and LAT1 function in oocytes 18.

Gene set enrichment analysis (GSEA) using Gene Ontology categories showed significant enrichment in control compared to GPNA for RNA and DNA processing and metabolism, as well as cell cycle processes (Figure 4B; see also supplementary material, Table S3). Similar categories were also enriched in control compared to BenSer (Figure 4C; see also supplementary material, Table S4). Analysis of motif enrichment by GSEA showed a significant association with E2F transcription factor motifs for control compared to either GPNA or BenSer (Figure 4D, E; see also supplementary material, Tables S5, S6). We have previously shown that inhibition of leucine uptake in LNCaP cells shows similar E2F‐mediated cell cycle inhibition, with a subset of 122 genes down‐regulated by the leucine uptake inhibitor BCH 23. Analysis of this 122‐gene signature in control versus GPNA and BenSer showed enrichment in the control group (Figure 4F, G), suggesting that there is a common mechanism regulating gene expression after leucine starvation and glutamine starvation in prostate cancer.

Amino acid starvation leads to up‐regulation of the stress‐response transcription factor ATF4, which in turn leads to increased expression of diverse amino acid transporters, including ASCT2 and LAT1 23. ATF4 was the first (p = 1.10E‐11) and sixteenth (p = 4.21E‐21) most significantly up‐regulated gene in GPNA‐ or BenSer‐treated cells, respectively (see supplementary material, Tables S1, S2), which was verified by RT–qPCR (see supplementary material, Figure S3A, B). ATF4 has previously been shown to directly transcriptionally regulate the expression of transporters including ASCT2, ASCT1, xCT and SLC3A2 23, 29. Analysis of gene sets up‐regulated after GPNA inhibition showed significant enrichment for Gene Ontology categories, including amino acid transport (see supplementary material, Figure S3C), with significant up‐regulation of transporters, including SLC1A5 (ASCT2, p = 0.020), SLC1A4 (ASCT1; p = 7.72E‐5), SLC7A11 (xCT; p = 7.02E‐8), SLC38A1 (SNAT1; p = 3.51E‐5), SLC7A8 (LAT2; p = 5.63E‐5), SLC3A2 (4F2hc; p = 0.041) and SLC3A1 (p = 0.005). Another critical gene significantly increased by both GPNA (p = 0.009) and BenSer (p = 0.0002) was EIF2AK3 (PERK), which can phosphorylate eIF2α, leading to global protein synthesis suppression. A set of genes that belong to eukaryotic translation initiation factor are significantly decreased after GPNA treatment, including EIF3B (p = 0.0003), EIF2A (p = 0.003), EIF4H (p = 0.005), EIF5A (p = 0.009), EIF2B3 (p = 0.034), EIF4G1 (p = 0.039), EIF2B4 (p = 0.040) and EIF2B5 (p = 0.048).

Knockdown of ASCT2 suppresses growth in prostate cancer cells

We used lentiviral shRNA‐mediated knockdown of ASCT2 to substantiate the direct ASCT2 effects of BenSer and GPNA in LNCaP and PC‐3 cells. Two different shRNAs decreased ASCT2 expression and glutamine transport in LNCaP and PC‐3 cells compared to a non‐targeting shRNA control (Figure 5A, B; see also supplementary material, Figure S3D). Inhibition using shASCT2#2 also significantly decreased leucine uptake in both LNCaP and PC‐3 cells (Figure 5C), suggesting that ASCT2‐transported glutamine is utilized for leucine uptake.

LNCaP and PC‐3 cells expressing shASCT2#2 showed a significant decrease in cell viability compared to shControl cells (Figure 5D, E). Analysis of BrdU incorporation (Figure 5F) and apoptosis (see supplementary material, Figure S3E) in shASCT2#2‐expressing PC‐3 cells confirmed that ASCT2 knockdown inhibits the cell cycle rather than apoptosis. Similar results for glutamine uptake, BrdU incorporation and apoptosis were also observed in shASCT2#1 PC‐3 cells (see supplementary material, Figure S3D, F, G). Finally, we examined mTORC1 pathway activation by western blotting, showing reduced phosphorylation of p70S6K in PC‐3 cells expressing shASCT2#2 (Figure 5G).

Knockdown of ASCT2 suppresses tumour growth in prostate cancer xenografts

To determine whether ASCT2 function is critical for tumour growth in vivo, PC‐3 cells expressing shControl or shASCT2#2 were transduced with a lentiviral vector co‐expressing eGFP and firefly luciferase 30. Cells were enriched to high purity on a narrow band of high GFP expression, resulting in similar GFP/luciferase expression in each shControl and shASCT2 cell line (see supplementary material, Figure S4A). PC‐3–luc cells were subcutaneously injected into nude mice, with similar luciferase expression in shControl and shASCT2 tumours confirmed after 48 h (Figure 6A). Bioluminescence was analysed twice weekly for 32 days, showing a significant decrease in shASCT2 tumour size by day 25 (Figure 6B).

Mice were euthanized after 32 days, due to the size of the shControl tumours. The tumours were isolated, photographed and weighed, with shControl tumours being significantly larger than shASCT2 tumours (Figure 6C, D; p < 0.001). Western blots of xenograft tumours showed that phosphorylation of p70S6K was decreased in shASCT2 tumours compared to shControl tumours (Figure 6E). Analysis of E2F2, CDK1, CDC20, UBE2C and the proliferation biomarker Ki67 in xenograft sections showed consistently lower levels in the shASCT2 tumours (Figure 6F). There were no changes in cleaved caspase 3 levels (see supplementary material, Figure S4B).

Micrometastases were detected by isolation and ex vivo bioluminescence analysis of organs immediately after euthanasia (see supplementary material, Figure S4C). Analysis of shControl mice showed that nine of 10 had metastases in the liver and four of 10 in the lung, while shASCT2 mice showed that four of nine had metastases in the liver, with no lung metastases (p = 0.004; Figure 6G). Metastatic bioluminescence burden in the liver and lung also showed a significant decrease in shASCT2 xenografted mice (Figure 6H).