Glutamate Decarboxylase 65 Signals through the Androgen Receptor to Promote Castration Resistance in Prostate Cancer

Prostate cancer is the most common cancer and the second leading cause of cancer-related mortality in men (1). Androgen receptor (AR) signaling has been shown to promote prostate cancer progression. Accordingly, androgen-deprivation therapy (ADT) is a major component in the treatment of recurrent and metastatic prostate cancer. However, a considerable fraction of patients receiving ADT ultimately develop castration-resistance, mostly due to acquired mutations in the AR signaling pathway (2). These mutations typically occur after metastatic spread (3). As a result, the capability to survive ADT independent of AR pathway mutations has been hypothesized to be an early step in the evolution of some aggressive, AR-dependent castration-resistant prostate cancer (CRPC).

Metabolic reprogramming exemplified by the Warburg effect is a one of the hallmarks of cancer cells, and contributes to process of carcinogenesis and cancer progression. Identification of metabolic vulnerability of human cancers has led to the development of novel treatment strategies (4, 5). For prostate cancer, metabolic change has been an area of active inquiry with potential application in metabolic imaging and biomarker diagnostics as well as for targeted therapies (6, 7). Despite of significant genetic and clinical heterogeneity among prostate cancers, common metabolic reprogramming properties have been recognized. In comparison to normal prostate cells, citrate secretion is decreased and the Krebs cycle is reactivated in prostate cancer cells, with subsequent changes in cellular redox state and metabolic flux through interdigitated pathways (8). Another early and common metabolic alteration in prostate cancers is increased de novo lipogenesis (9, 10). Increasing evidence suggests that metabolic reprogramming could promote tumor cell growth and survival in diverse microenvironmental conditions. For example, cancer cells adapt to hypoxia by utilizing physiologic adaptation pathways that promote a switch from oxidative to glycolytic metabolism (11). Metabolic reprogramming is also critical for cancer cell survival upon distant metastasis in new environment. Specifically, upon metastasis into the liver, colorectal cancer cells undergo metabolic changes that allow using fructose (more abundant in the liver) as a major energy source (12). In this study, we characterized metabolic reprogramming that allows prostate cancer cells to survive upon androgen deprivation.

The PC-AD and PC-CR xenografts were mechanically dissociated and cultured in RPMI1640 (Invitrogen) medium supplemented with 10% FBS (for PC-AD cells) or 10% charcoal-stripped FBS (for PC-CR cells). After several serial passages, the serum was reduced to 3%, and 100 ng/mL IGF1 (SAFC Biosciences) was added to enhance the growth of epithelial cells and suppress the proliferation of stromal cells. Prostate cancer cell lines LNCaP, abl, CWR22Rv1, and HEK293T cell line were obtained from ATCC between 2010 and 2015 and maintained in RPMI1640 supplemented with 10% FBS (for LNCaP cells and HEK293T cells) or 10% charcoal-stripped FBS (for abl and CWR22Rv1 cells). The identity of the cell lines was verified by short tandem repeat (STR) profiling (Shanghai Biowing Applied Biotechnology), absence of Mycoplasma infection was verified by conventional PCR with the PromoKine PCR Mycoplasma Test Kit I/C. All stable cell lines were aliquoted and frozen after derivation and expansion, experiments were performed within the first 25 passages following thawing.

Immunofluorescence, immunoprecipitation, and Western blotting were performed as described previously (13).

FRAP assay was performed in live cell chambers (Corning) at 37°C using a Zeiss LSM510 microscopy. The cells expressing GFP-AR were bleached with argon laser at 488 nm wavelength. About 90% of the GFP signal was bleached in both the nuclei and cytoplasm. Images were then taken with 35 frames at 25 seconds intervals. Average intensities in regions of interest were measured using MetaMorph software (Molecular Devices). The intensity in the prebleach image was set to 100%, and the first postbleach image was set as time point 0. The recovery curves represent the average of at least 6 cells.

GST fusion proteins were expressed in BL21(DE3) bacteria, lysed by using a French press, isolated using glutathione beads, eluted in 50 mmol/L Tris-Cl (pH 8.0) containing 10 mmol/L glutathione, and dialyzed into PBS. His-tagged proteins were expressed and purified using a Talon metal affinity resin (Clontech). Recombinant proteins were incubated in Tris-buffered saline Nonidet P-40 buffer (20 mmol/L Tris, pH 8.0, 150 mmol/L NaCl, 0.5% NP-40, 5 mmol/L EGTA, 1.5 mmol/L EDTA, 0.5 mmol/L Na₃VO₄, 20 mmol/L p-nitrophenyl phosphate) supplemented with a cocktail of protease inhibitors and 250 μmol/L GABA at 4°C for 4 hours. Glutathione-agarose beads (GST pull-down) or Ni-NTA-agarose beads (His pull-down) were then added in the incubation for additional 2 hours. The bound proteins were resolved by SDS-PAGE and subjected to Western blotting.

Patients provided their written informed consent for the use of samples from their tumors for future investigations, and experiments were conducted in accordance with the Declaration of Helsinki and approved by the Institutional Review Board of Second Military Medical University.

Four-week-old male NOG mice were used as transplant recipients. All animal procedures were performed in compliance with the institutional ethical requirements and approved by the Institutional Animal Care Use Committee of Second Military Medical University.

Statistical analysis was carried out with SPSS version 19.0 (SPSS Inc.). Experimental data are expressed as means ± SD and analyzed by 2-tailed unpaired t test. Association between Sirt5 expression and disease recurrence was analyzed by the Kaplan–Meier method, and curves were compared by the log-rank test. P < 0.05 was considered to indicate statistical significance throughout the study.

Biopsy specimens were obtained from 7 patients with locally advanced or metastatic prostate cancer, and were immediately implanted subcutaneously into NOG mice in the presence of Matrigel. After a minimum of 3 generations of serial passaging, we were able to establish transplantable patient-derived xenograft (PDX) lines in 5 of the 7 patients (PC-AD1, PC-AD2, PC-AD3, PC-AD4, and PC-AD7). The transplantable tumor lines retained major histopathologic characteristics of their matched patient tumor (Supplementary Table S1). Host castration resulted in a dramatic decrease in tumor volume in all established lines; however, a vestige of the regressed xenograft remained at the inoculation site. Within 7 months of continuing observation, 3 out of the 5 regressed xenografts relapsed (PC-AD1, PC-AD3, and PC-AD4; Supplementary Fig. S1A), indicating progression to an androgen-independent state. We successfully established the stable subcutaneous xenograft model from the first relapsed xenograft after 5 subcutaneous passages in castrated NOG mice (named PC-CR1, PC-CR3, and PC-CR4, respectively). We next compared the requirement for glucose or glutamine of hormone-dependent (AD) with paired castration-resistant (CR) lines. Similar to that in most proliferating cells, both PC-AD and PC-CR lines proliferated at a high rate when glucose and glutamine were abundant (Supplementary Fig. S1B), with high levels of glucose and glutamine consumption and lactate production (Supplementary Fig. S1C). Also, no apparent difference was observed in fatty acid uptake between the paired PC-AD and PC-CR xenograft cells (Supplementary Fig. S1D). TCA activity, as reflected by oxygen consumption rate (OCR), was similar between PC-AD and PC-CR cells (Supplementary Fig. S1E). Notably, steady-state level of succinyl-CoA, but not other TCA cycle metabolites (αKG, succinate, or malate), was reproducibly diminished in all PC-CR lines tested as compared with their androgen-dependent counterparts (Fig. 1A). In PC-AD cells, virtually all intracellular αKG, succinyl-CoA, and succinate were rapidly labeled after addition of [U-¹³C]glucose or [U-¹³C]glutamine. In contrast, a substantial fraction of succinyl-CoA failed to become labeled with [U-¹³C]glucose or [U-¹³C]glutamine in PC-CR cells (Fig. 1B). Quantification of metabolite fluxes revealed significantly lower glucose and glutamine fluxes through succinyl-CoA in PC-CR cells (Fig. 1C) despite of similar flux of glucose- or glutamine-derived carbons through αKG in PC-CR versus PC-AD (Supplementary Fig. S1F).

The γ-aminobutyric acid (GABA) shunt bypasses the αKG dehydrogenase complex (KGDHC) and succinyl-coenzyme A ligase (SUCL) for the conversion of αKG into succinate, and thus decreases the formation of succinyl-CoA (14). PC-CR cells had substantial glutamine or glucose-dependent GABA and succinic semialdehyde (SSA) production (Fig. 1D). Also, these cells preserved larger intracellular pools of GABA and SSA than their parental hormone-naive cells (Fig. 1E). Glutamic acid decarboxylase (GAD), GABA-transaminase (GABA-T), and succinic semialdehyde dehydrogenase (SSADH) are key enzymes in the GABA shunt. To assess the role of the GABA shunt in CRPC growth, we impaired the enzyme activity individually using RNA interference (Supplementary Fig. S1G). The growth of CRPC was markedly decreased by silencing the GAD65 isoform (Supplementary Fig. S1H–S1I), but not by knockdown of GAD67, GABA-T, or SSADH (Supplementary Fig. S1J). Consistent with this finding, ¹³C-tracing experiments showed increased glutamate (and total glutamate) and decreased GABA (and total GABA) upon GAD65 knockdown in PC-CR4 cells (Fig. 1F), suggesting an increased GABA shunt activity in CRPC.

Silencing GAD65 produced more profound inhibition on the growth in PC-CR4 cells than in matched PC-AD4 cells (Fig. 2A; Supplementary Figs. S1H–S1I and S1K). More robust inhibition by GAD65 silencing was also observed in the androgen-dependent prostate cancer cell line LNCaP than in its androgen-independent derivative LNCaP-abl (abl; Fig. 2B; Supplementary Fig. S1L). In an in vivo mouse xenograft CRPC model using CWR22Rv1 cells, tumors derived from GAD65 knockdown cells demonstrated decreased growth rate and masses with reduced cell proliferation rate assessed by IHC staining of Ki-67, compared with those derived from control cells (Fig. 2C, top). However, the growth of xenografts derived from PC-AD4 and PC-AD4 GAD65 knockdown cells did not differ (Fig. 2C, bottom). We generated “rescue” CRPC cells (PC-CR4 and abl) with stable knockdown of endogenous GAD65 and rescue expression of wild-type (WT) or an enzyme-dead K396R mutant (15) of a shRNA-resistant, flag-tagged human GAD65 form. Rescue expression of Flag-GAD65^WT, but not the K396R mutant, prevented GAD65 shRNA from inhibiting the proliferation of PC-CR4 and abl cells (Fig. 2D). Moreover, rescue CWR22Rv1 cells expressing GAD65^WT but not the K396R mutant demonstrated restored potential of tumor formation and growth (Fig. 2E). These data together suggested that GAD65 is necessary for CRPC proliferation.

To clarify whether activation of the GABA shunt in CRPC is due to increased GAD65 enzyme activity, we expressed flag-tagged GAD65 in PC-AD and PC-CR cells (Supplementary Fig. S2A). GDA65 activity after immunoprecipitation was significantly higher in PC-CR cells than in PC-AD cells (Fig. 3A). Phosphorylation status plays an important role in regulation of GAD activity (16). Four putative phosphorylation sites (serine 3, serine 6, threonine 95, and serine 417) have been identified in GAD65 by mass spectrometry (17). We mutated each of the 4 putative phosphorylation sites individually to alanine (A), and examined their phosphorylation. GAD65 phosphorylation was decreased by mutation at T95, and more so with S6 (by approximately 90%), but not at S3 and S417 (Fig. 3B), indicating that S6 is the major phosphorylation site in GAD65. To test the effect of S6 phosphorylation, we also generated the GAD65^S6E mutant, in which S6 was substituted with glutamate to mimic constitutive phosphorylation, and overexpressed flag-tagged mutant or WT GAD65 in HEK293T cells (Supplementary Fig. S2B). GAD65 enzyme activity was significantly increased by substitution with glutamate (GAD65^S6E) but not with alanine (GAD65^S6A; Fig. 3C). Consistent with an stimulative effect of phosphorylation on GAD65 activity, inhibition of phosphatase by okadaic acid (OA) and vanadate (Vd) treatment significantly increased GAD65 activity in cells overexpressing WT GAD65 (Fig. 3D), but not in cells expressing either GAD65^S6A or GAD65^S6E mutant (Fig. 3D). Taken together, these data suggest that phosphorylation at S6 activates GAD65.

Next, we generated an antibody specifically recognizing the S6-phosphorylated GAD65. The specificity of the anti-phospho-GAD65^S6 antibody was verified by recognition of the S6-phosphorylated peptide but not the unphosphorylated control peptide (Supplementary Fig. S2C). Phosphorylation of GAD65 could readily be detected by this antibody; the signal was diminished by GAD65 knockdown (Supplementary Fig. S2D) and completely blocked by preincubation with the antigen peptide (Supplementary Fig. S2E), confirming the specificity of the anti-phospho-GAD65^S6 antibody. GAD65^S6 phosphorylation was negligible or absent in hormone-sensitive PC-AD and LNCaP cells but present in the paired castration-resistant PC-CR and abl cells (Fig. 3E). Consistent with the key role of PKCϵ in GAD65 phosphorylation (16), we found higher levels of the active PKCϵ in castration-resistant cells (Supplementary Fig. S2F). The activity of PI3K, an upstream activator of PKCϵ (18) and an essential regulator of CRPC (19), was also increased in castration-resistant cells (Supplementary Fig. S2G), suggesting major contribution of the PI3K/PKCϵ signaling to GAD65^S6 phosphorylation in CRPC. Furthermore, we evaluated the kinetics of GAD65^S6 phosphorylation in serial passages of PC-CR4 tumors in castrated mice. Percentage of the cells with GAD65^S6 phosphorylation increased with passaging of the PC-CR4 line, from ∼37% after the first passage to >80% by passage 5 (Fig. 3F). Such an increase was accompanied with slight increase in AR expression, moderate increases in PSA/TMPRSS2 expression (Supplementary Fig. S2H), and decreased tumor doubling time (9 days in P5 vs. 15 days in P1), suggesting a growth advantage of phosphorylated over unphosphorylated GAD65^S6 cells in castrated mice.

To determine whether GAD65^S6 phosphorylation can be detected in clinical CRPC, the anti-phospho-GAD65^S6 antibody was characterized by its suitability for IHC. We found that this antibody could detect strong signals that were specifically blocked by the phospho-S6 antigen peptide in paraffin-embedded normal colon tissues (Supplementary Fig. S2I). In soft-tissue and bone metastases obtained from 16 men who died from CRPC, majority (12/16) exhibited high level of GAD65^S6 phosphorylation (>30% phosphorylated S6-positive cells; Fig. 3G and H). We also stained tissue microarrays containing samples from individuals with benign prostatic hyperplasia (BPH), hormone-naive prostate cancer, prostate cancer treated with neoadjuvant hormone ablation and CRPC. We detected high GAD65^S6 phosphorylation in 1.5%, 12.3%, 56.7%, and 83.1% of these samples, respectively. The mean percentage of cells positive for GAD65^S6 phosphorylation among all samples increased from 2.7% in benign prostatic hyperplasia to 5.3% in hormone-naive disease, 40.5% in men treated with neoadjuvant ADT and 76.2% in CRPC (P < 0.01; Fig. 3I and J). These data demonstrate that GAD65^S6 phosphorylation is rare in untreated androgen-dependent prostate cancer, increases with androgen deprivation, and is highest in CRPC.

To decipher the role of GAD65 in bioenergetics and anabolic biosynthesis of CRPC, we performed a set of metabolic assays using CRPC cells with GAD65 knockdown. In comparison with the empty vector control, GAD65 knockdown did not affect intracellular ATP level, glucose uptake, glycolytic rate, lactate production, oxygen consumption rate, glutamine uptake, and glutaminolysis rate (Supplementary Fig. S3A–S3G). GAD65 knockdown also did not affect oxidative pentose phosphate pathway (PPP) flux (Supplementary Fig. S3H) and the overall lipid and RNA biosynthesis (Supplementary Fig. S3I–S3J).

Because GAD65 knockdown reduced cell proliferation and tumor growth, we hypothesized that other mechanisms beyond metabolic defects might be responsible for the proliferative disadvantage in cells conferred by GAD65 knockdown. Further analyses showed that knockdown of GAD65 selectively inhibited transcriptional activity of AR and reduced the mRNA levels of AR target genes (PSA and TMPRSS2) in CRPC cells (Fig. 4A and B). Such effects were reversed by ectopic expression of shRNA-resistant GAD65^WT or the GAD65^S6E mutant, but not the GAD65^S6A mutant (Fig. 4C and D). GAD65 knockdown did not significantly change the levels of AR (Supplementary Fig. S4A), but substantially reduced the levels of nuclear AR in abl and PC-CR3 cells (Fig. 4E), which was confirmed by immunofluorescence (Fig. 4F). No similar phenomena were detected in the matched androgen-sensitive cells (Fig. 4E and F).

We further explored the molecular mechanism by which GAD65 regulates AR nuclear localization. GAD65 did not form a protein complex with AR. We speculated that GAD65 may indirectly regulate AR nuclear localization by controlling intracellular levels of GABA shunt intermediates. Addition of GABA up to 250 μmol/L in the culture media restored intracellular GABA levels in GAD65 knockdown cells (Supplementary Fig. S4B). Interestingly, the decrease in AR nuclear localization caused by GAD65 knockdown in CRPC cells was also reversed by GABA treatment, but not by succinate or SSA treatment (Fig. 4G and H). We also demonstrated the ability of GABA to restore AR transcriptional activity in GAD65 knockdown cells by quantifying ARE reporter activity (Fig. 4I), as well as the expression of AR target genes (Fig. 4J). Consequently, addition of GABA, but not succinate or SSA, significantly reversed the reduced cell proliferation due to GAD65 knockdown (Fig. 4K). Based on these results, we hypothesize that GAD65-mediated conversion of glutamate into GABA functions in a pathway that is used to promote AR nuclear accumulation in CRPC.

To distinguish whether GAD65 facilitates AR nuclear accumulation through either nuclear import or nuclear retention, we analyzed the nuclear localization of AR in GAD65 knockdown abl cells after treatment with leptomycin B (LMB), an inhibitor of protein nuclear export by CRM1/exportin 1 (20). As shown in Fig. 5A, AR was primarily localized in the nucleus in the presence of LMB, strongly suggesting that the cytoplasmic localization of AR after GAD65 knockdown is not due to defects in nuclear import but due to a combination of deficient nuclear retention and effective nuclear export. Coimmunoprecipitation of AR, followed by a mass spectrometry analysis suggested binding of AR to zinc finger HIT domain-containing protein 3 (ZNHIT3) in PC-CR4 cells (Supplementary Table. S2), as well as in PC-CR3 and abl cells (Fig. 5B). Such an interaction was significantly disrupted by treating cells with a GAD65 shRNA (Fig. 5C). Next, we performed cell-free pull-down assays using purified recombinant AR incubated with purified ZNHIT3, in the presence of increasing amounts of GABA. Although ZNHIT3 displayed negligible interaction with AR, GABA (50–250 μmol/L) increased AR-bound ZNHIT3 in a concentration-dependent manner (Fig. 5D). More importantly, we found that purified AR pretreated with increasing concentrations of GABA displayed increasing binding ability to ZNHIT3, whereas GABA pretreated ZNHIT3 had no effect (Fig. 5E), suggesting that GABA directly affects the AR–ZNHIT3 complex. In line with this, a radiometric metabolite–protein interaction analysis using ¹⁴C-labeled metabolites revealed that labeled GABA but not succinate, or SSA, was retained on purified recombinant AR (Supplementary Fig. S4C). To determine the selectivity of GABA binding to AR, we generated an AR mutant with substitutions at L704, Q711, M745, R752, and F764. These residues were predicted to be critical for GABA binding to AR by molecular docking (Fig. 5F). The results demonstrated that the L704A/Q711A/M745A/R752A/F764A (5A) mutant was resistant to GABA binding (Supplementary Fig. S4D); the association between the 5A mutant and ZNHIT3 was no longer induced by GABA (Fig. 5G). To clarify whether ZNHIT3 facilitates AR nuclear retention, we performed fluorescence recovery after photobleaching (FRAP) GFP-AR in abl cells expressing scramble shRNA or ZNHIT3 shRNA (Fig. 5H). After photobleaching of GFP-AR in the nucleus, no significant differences were observed in the recovery of nuclear GFP-AR among control and ZNHIT3 knockdown cells. In contrast, after photobleaching of cytoplasmic AR, reduction of nuclear GFP-AR was significantly less in control cells compared with ZNHIT3 shRNA-expressing cells, thus indicating an important role of ZNHIT3 in AR nuclear retention.

In the next experiment, GST-fusion constructs of different AR truncation mutants lacking the NTD, hinge, or DBD domains (Fig. 6A) were immobilized on glutathione resin; in vitro binding assays were performed to examine their ability of binding to ZNHIT3 in the presence of GABA. As illustrated in Fig. 6B, GABA induced ZNHIT3 binding only to AR constructs that contain the hinge region, and not any other AR domains. To locate the segment critical for ZNHIT3 association, we generated and characterized several mutants bearing smaller deletions within the AR hinge region (Fig. 6C). The AR (Δ625-634) and AR (Δ653-660) mutants had full binding affinity for ZNHIT3 upon GABA treatment, whereas AR (Δ625-660) completely lost binding affinity for ZNHIT3 (Fig. 6D), suggesting amino acids 635-652 is absolutely required for GABA-induced AR binding to ZNHIT3. Next, we transfected HEK293 cells with the GFP-AR or GFP-ARΔ635-652 construct. Transfected AR constructs underwent nuclear import within 4 hours after DHT induction. In the absence of DHT, GFP-AR, but not GFP-ARΔ635-652, exhibited prolonged nuclear retention in the presence of GABA (Fig. 6E). These results indicate that GABA-induced ZNHIT3 binding in the hinge region retains AR in the nucleus to promote its nuclear accumulation. To further explore how GABA promotes ZNHIT3 binding, we generated an antibody to a specified hinge segment (amino acids 633–653) including amino acids 635 to 652 for the purpose of testing whether the GABA binding alters hinge accessibility. Immunoblotting showed that this antibody binds to hinge, hinge lacking the amino acids 655 to 668, but does not recognize NTD, LBD, and DBD (Fig. 6F). In the next experiment, this antihinge antibody was used to probe for hinge accessibility in the GST-hinge-LBD (amino acids 624–920) by ELISA. The GST-hinge-LBD protein was adsorbed to microtiter wells, based on detection with an anti-GST antibody. The addition of GABA increased the antihinge antibody binding to GST-hinge-LBD protein by approximately 3 folds (Fig. 6G), indicating that GABA binding promotes AR hinge accessibility to ZNHIT3.

A retroviral vector, in which GAD65 shRNA was expressed under the control of the doxycycline (Dox)-inducible activator rtTA (LV-tet-on-shGAD65), was constructed and used to infect PC-CR3, abl, and CWR22Rv1 cells. Dox treatment induced knockdown of endogenous GAD65 in these cells (Supplementary Fig. S5A). After palpable tumor formation, castrated mice bearing infected PC-CR3, abl, or CWR22Rv1 xenografts were treated twice weekly with PBS or Dox for 5 weeks. Dox administration significantly slowed the growth of all 3 tumor models, with the most prominent effect in CWR22Rv1 xenograft (Fig. 7A). Dox-treated tumors were pale, nonadherent to underlying muscle, and noninvasive histologically, whereas control tumors grossly invaded underlying muscle (Fig. 7B). Metastasis to lymph nodes was less frequent in Dox-treated mice (1 of 5 mice) versus the control (5 of 5; Fig. 7C). Prolonged administration of Dox led to long-term growth suppression (Supplementary Fig. S5B) and increased the duration of survival by >100% in mice bearing abl tumors. Dox-treated tumors had large areas of cell loss, reduced proliferation (Ki-67 staining), and fewer blood vessels (CD31 staining) compared with the PBS control (Fig. 7D).

In the last set of experiments, androgen-sensitive PC-AD3 cells were transfected with the Dox-inducible vector for GAD65 shRNA expression, and then implanted subcutaneously into castrated mice. Treatment of the mice with Dox significantly delayed time to castration resistance (Fig. 7E). At week 6 after implantation, progression to castration-resistance (defined as serum PSA at >8 ng/mL) was apparent in 8 of 10 mice and none of the 10 Dox-treated mice (Fig. 7F). Of note, AR immunostaining in xenografts treated with Dox was predominantly localized to the cytoplasm of tumor cells, whereas the PBS control tend to have higher level of AR in the nucleus (Supplementary Fig. S5C). These findings suggest that GAD65 inhibition maintain the sensitivity of AR in androgen-sensitive tumors in castrated hosts and inhibit the development of castration resistance.

In efforts to develop more effective therapies, it is critical to improve the understanding of molecular mechanisms underlying CRPC development. In this study, we approached this question by identifying “metabolic vulnerabilities” specifically required for CRPC development. Distinct metabolic states are required for tumor growth under different environments (11–12, 21). We found increased GABA shunt activity due to GAD65 activation in CRPC. The GAD65 activation was due to S6 phosphorylation, and mediated by the PI3K–PKCϵ signaling axis. This finding is consistent with a previous study showing a critical role of the PI3K–PKCϵ signaling axis in the development and maintenance of CRPC (19). Specifically, we detected GAD65 S6 phosphorylation in multiple CRPC cell lines and in a majority of metastatic and castration-resistant prostate cancer tissues. Interestingly, GAD65 S6 phosphorylation increased with passaging of castration-resistant tumors in the xenograft models, suggesting that phospho-S6-positive cells have a growth advantage over phospho-S6-negative cells. We also provided convincing evidence showing that activated GAD65 promotes CRPC proliferation via facilitating AR nuclear accumulation. It is well recognized that AR is activated by androgen binding and translocates to nucleus to activate transcription of AR-target genes. However, in CRPC cells, AR remains in the nucleus even in the absence of androgens, transactivating androgen-responsive genes and leading to uncontrolled growth (22). Blockade of AR nuclear localization is a putative treatment strategy for CRPC. The results from this study showed that GAD65 selectively promotes AR nuclear retention in CRPC by controlling intracellular levels of its product GABA. GABA is the major inhibitory neurotransmitter in the mammalian brain but recent studies have suggested that GABA is also implicated in many other biological functions (23, 24), as well as in the development of a variety of cancers (25). Uptake of GABA and its subsequent catabolism via the GABA shunt increase NADH levels in tumor microenvironment, which in turn confer a proliferative advantage to tumor cells (26). Also, GABA acting via its receptors ensures a beneficial and important function on tumor cell proliferation and a “GABAergic Mr. Hyde” has been described in different types of tumors where components of the GABAergic signaling are overexpressed (25). Our findings link the GABA shunt to AR nuclear accumulation in CRPC and clarify a signaling function of GABA that is independent of its role in cell metabolism or GABA receptor-mediated effects. GABA serves as a proto-oncometabolite by binding to AR in its ligand binding domain. GABA induces a conformational switch in AR that increases hinge accessibility to ZNHIT3. ZNHIT3 is a nuclear zinc finger protein previously implicated in transcriptional regulation and in small nucleolar ribonucleoprotein particle assembly and thus possibly to preribosomal RNA processing (27, 28). ZNHIT3 inhibits nuclear export and promotes nuclear retention of AR. These results suggest that metabolic intermediates may function as signaling molecules to allow crosstalk between metabolic and cell signaling pathways. For example, we previously reported that TCA cycle intermediate succinate binds to and activates the ROS scavenging enzyme TrxR2 to maintain redox homeostasis (29), whereas others have found that acetoacetate specifically enhances binding of BRAF V600E to MEK1 in V600E-positive cancer cells to promote activation of MEK–ERK signaling (30). On the other hand, increasing evidence suggests that posttranslational modifications, including tyrosine phosphorylation (31, 32) and lysine acylation (33, 34) of metabolic enzymes, are common and important to link cell signaling pathways to metabolic pathways in cancer cells. These findings together represent a realm of crosstalk with “back and forth” signal flows between metabolic and cell signaling networks that “acutely” regulate cell metabolism and proliferation, which, unfortunately, are “hijacked” by cancer cells.

Finally, the findings from this study encourage exploring the GABA shunt, GAD65 in particular, as a molecular target in the treatment of CRPC. Future studies, however, must address the specificity issue due to the critical functional role of GABA in the central nervous system (14).

No potential conflicts of interest were disclosed.

Conception and design: L. Chen, Z.G. Du, Q.Q. Li

Development of methodology: Y. Gao, Z.M. Wu, X.J. Liu, H. Huang

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): Y. Gao, L. Chen, Z.M. Wu, X.J. Liu, Q.Q. Li

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): Y. Gao, L. Chen, Z.G. Du, Q.Q. Li

Writing, review, and/or revision of the manuscript: Q.Q. Li

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): Z.G. Du, W.C. Gao, H. Huang, D.F. Xu

Study supervision: Q.Q. Li

This work was supported by grants from the National Nature Science Foundation of China (81372350, 81572886, and 81874064; all to Q.Q. Li). The authors thank Professor Xiang Wang (Department of Urology, Shanghai General Hospital, Shanghai Jiaotong University, Shanghai, China) for reviewing the English text and helpful discussion.

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.