A novel direct activator of AMPK inhibits prostate cancer growth by blocking lipogenesis

1Metabolism in cancer cells is reprogrammed to facilitate the incorporation of nucleotides, amino acids, and lipids into the biomass needed to produce a new cell (Vander Heiden et al, ). Increased de novo fatty acid (FA) and cholesterol synthesis is a hallmark of prostate cancer (PCa) and correlates with tumor progression and poorer prognosis (reviewed in Pelton et al,; Zadra et al, ). Increased protein synthesis due to alterations in phosphatase and tensin homologue/phosphatidylinositol 3-kinase/Akt/mammalian target of rapamycin (PTEN/PI3K/Akt/mTOR) pathway is also a common feature of both primary and metastatic PCas (Majumder & Sellers, ), and is associated with increased lipogenesis (Van de Sande et al, ). However, despite promising preclinical results, targeting mTOR complex 1 (mTORC1) with rapamycin and its analogues failed to show clinical efficacy in PCa (Amato et al, ) due to the emergence of survival feedback loops (O'Reilly et al,; Carracedo et al,; Rodrik-Outmezguine et al, ) or concomitant development of a tumor-driven lipogenic phenotype (Menendez & Lupu, ). The side effects of existing FA synthesis inhibitors have thus far precluded their clinical use in cancer (Flavin et al, ). Thus, alternative strategies to target these pathways in PCa patients are in great need.

25′ AMP-activated kinase (AMPK) is an energy sensor serine/threonine kinase that stands at the crossroads linking cell metabolism and oncogenesis. It is a heterotrimeric complex (composed of a catalytic α subunit and regulatory β and γ subunits) that responds to an increased AMP/ATP ratio by turning on ATP-generating pathways, while switching off ATP-consuming ones (Hardie & Carling, ). AMPK directs the switch from an anabolic to a catabolic state both by phosphorylating key metabolic targets, and by regulating their gene expression (Hardie, ). AMPK activation requires phosphorylation of Thr172 within the α subunit by an upstream Ser/Thr protein kinase, mainly the tumor suppressor liver kinase beta 1 (LKB1) and the calcium/calmodulin-dependent protein kinase kinase-β (CaMKKβ) (Woods et al,; Hawley et al, ). Germline inactivating mutations of LKB1 are responsible for the Peutz-Jeghers hereditary cancer syndrome. Somatic mutations of LKB1 are also found in a significant fraction of non-small cell lung carcinomas (NSCLCs) and cervical tumors, thus suggesting that LKB1/AMPK axis may act as a tumor suppressor (Shackelford & Shaw, ). When pharmacologically activated, AMPK exerts pleitropic effects resulting in the suppression of tumorigenesis and tumor progression, including inhibition of mTORC1 signaling members tuberous sclerosis complex 2 (TSC2) and Raptor, FA and cholesterol biosynthesis, cell cycle progression, as well as induction of autophagy and apoptosis (Shaw,; Fogarty & Hardie, ). In contrast, AMPK loss fosters tumor progression (Faubert et al, ). Epidemiologic and experimental evidence using metformin and AICAR suggest that AMPK activation is a valid anti-cancer strategy (Xiang et al,; Evans et al,; Bowker et al,; Zakikhani et al, ). However, these activators are known to have additional molecular targets and AMPK-independent effects, leaving unanswered the question of whether AMPK activation per se is necessary and sufficient to affect tumor growth (Ben Sahra et al,,; Kalender et al,; Santidrian et al, ). Direct targeting of AMPK with A-769662 (Abbott Laboratories) has initially shown promising results in the treatment of metabolic disorders and cancer (Cool et al,; Huang et al, ). However, AMPK-independent toxic effects and reduced efficacy in cancer cells have been recently reported (Moreno et al,; Garcia-Garcia et al, ). Thus, development of novel AMPK direct activators is the focus of intense efforts (Pang et al,; Lee et al, ).

3Androgens are the major drivers of PCa carcinogenesis/progression and androgen deprivation therapy (ADT) is an effective first-line therapy for advanced PCa. Despite good initial responses, most patients relapse within several years with a more aggressive castration-resistant PCa (CRPC), where androgen receptor (AR) reactivation occurs through a variety of mechanisms (Yuan et al, ). Hence targeting the AR remains a critical component of the current CRPC therapies, including the AR antagonist MDV3100, and the inhibitor of androgen synthesis abiraterone (Higano & Crawford, ). However, emergence of treatment resistance to these new drugs represents a major limitation (Yuan et al, ). Previous groups have shown AR-induced de novo lipogenesis, which is enhanced during the emergence of androgen independence to contribute to the survival/growth of CRPC cells (Swinnen et al,; Ettinger et al, ). Since AMPK activation negatively regulates the expression/activity of lipogenic transcriptional factors (SREBPs, Sterol Regulatory Element-Binding Proteins) and the major lipogenesis and cholesterol synthesis enzymes, we reasoned that combinatorial regimens of AMPK activators and AR pathway inhibitors might result in a therapeutic benefit and overcome resistance to ADT.

4In this study, we use a novel small molecule AMPK activator to investigate whether AMPK per se plays an anti-cancer role in PCa, and whether this can be monitored in vivo by PET imaging. We dissect the molecular underpinning mechanisms of AMPK mediated-growth inhibition and we seek to determine the effect of AMPK activation in CRPC and its combination with AR signaling inhibitors.

5The small molecule MT 63–78 (now Debio 0930) was identified in a targeted screening using purified human recombinant AMPK α1β1γ1 (Fig 1A). Oral bioavailability and pharmacokinetic characterization of the compound is provided in Supplementary Fig 1A. Using a cell-free assay, we demonstrated that MT 63–78 allosterically activates recombinant AMPK in a dose-dependent manner (Fig 1B). In addition, MT 63–78, like AMP, inhibits AMPK dephosphorylation on Thr172 by protein phosphatase 2C alpha (Fig 1C). We tested whether MT 63–78 was able to activate AMPK in PCa cells, using androgen-dependent LNCaP (PTEN null) and androgen-independent PC3 (PTEN and p53 null) cell lines as models. We observed a dose-dependent phosphorylation of the two major AMPK targets Acetyl-CoA Carboxylase (ACC) on Ser79 and of Raptor on Ser792, after 30 min of treatment. A corresponding increase in Thr172 phosphorylation on the AMPK α subunit was also observed (Fig 2A). AMPK activity induced by MT 63–78 was significantly increased in both cell lines in a dose-dependent manner (EC50 = 25 μM) (Fig 2B). This increase was significantly stronger compared to treatment with the current available AMPK activator A-769662 (Abbott Laboratories) and AICAR (Supplementary Fig 1B). In contrast to metformin, 2-deoxyglucose, and oligomycin, addition of the compound did not cause any changes in intracellular ATP, ADP levels in LNCaP and PC3 cells, demonstrating that AMPK activation by MT 63–78 is not an indirect effect of increased energy stress (Fig 2C and D). We also confirmed these data in HepG2 cells by measuring ATP, ADP, and AMP levels using high-performance liquid chromatography (HPLC). Reduction in ATP levels and increased ADP, AMP levels were only observed at 200 μM of the compound, which is far beyond the concentrations used in this study (Supplementary Fig 2).

6To evaluate the specificity of MT 63–78, AMPK α1−/− and α2−/− and wild-type (wt) mouse embryonic fibroblasts (MEFs) were incubated with MT 63–78 for 30 min. In contrast to wt MEFs, no phosphorylation of ACC and Raptor was observed in AMPK α1−/− and α2−/− MEFs (Fig 2E). We also screened MT 63–78 at 5 and 25 μM in cell-free assays against a panel of 93 protein kinases other than AMPK (Kinase Profiler, Millipore). Only 2 kinases (IGF-1R, Flt3) were slighltly activated at the highest concentration, whereas the majority of the kinases were not stimulated by either of the two concentrations, including the rat homolog of AMPK and the human members of AMPK-related kinase family (SIK, ARK5, MARK1). At 25 μM, six kinases were marginally inhibited, but only in one case (MST1) the inhibition was more than 50% (Supplementary Fig 3).

7The compound exerts its effect by binding to the regulatory β subunit. To discriminate between the β1 and β2 subunits, we tested the in vitro activity of four different AMPK heterotrimeric complexes (α1β1γ1, α2β1γ1, α1β2γ1, α2β2γ1), following incubation with MT 63–78, A-769662, and AMP. We observed that both MT 63–78 and A-769662 show maximal AMPK activation in β1 subunit-containing heterotrimers (α1β1γ1, α2β1γ1) and a stronger effect for A-769662. At high doses MT 63–78 was also able to activate β2 subunit-containing heterotrimers, whereas A-769662 showed no activation (α1β2γ1) or low activation (α2β2γ1), suggesting a low-affinity binding of MT 63–78 to the β2 subunit, especially when β2 is complexed with the α2 subunit (Supplementary Fig 4A). However, using PCa cell models, we observed that β1 subunit knock down significantly reduces the basal activity of AMPK (80 and 92% for LNCaP and PC3, respectively) and totally abolishes its activation by MT 63–78 (Fig 2F). In contrast, in β2 subunit-silenced LNCaP and PC3 cells AMPK activity and ACC phosphorylation were still increased (although at lower extent) when cells were treated with MT 63–78 (Supplementary Fig 4B and C). This suggests that MT 63–78 activates AMPK mainly by binding the β1 subunit in a cell-based context.

8Taken together, these results show that the small molecule MT 63–78 is a specific and potent tool for studying the effect of a direct activation of AMPK in PCa.

9Activation of AMPK requires phosphorylation of Thr172 mainly by LKB1 and CaMKKβ (Woods et al,; Hawley et al, ). Both upstream kinases are expressed in normal prostate cells and in PCa cells, except for DU145 cells, which are LKB1 null (Fig 3A). We treated DU145 and HeLa cells (the model commonly used for LKB1 deficiency) with MT 63–78 and we observed a dose-dependent activation of AMPK in both cell lines (Fig 3B). HeLa as well as wt and LKB1−/− MEFs were then exposed to the CaMKKβ inhibitor STO-609, prior to incubation with the compound. STO-609 almost completely abolished AMPK activation by MT 63–78 as well as the phosphorylation of AMPK downstream targets ACC and raptor in HeLa and LKB1−/− MEFs, whereas STO-609 did not have any significant effect on wt MEFs (Fig 3C and D).

10These results suggest that activation of AMPK in response to MT 63–78 is independent of LKB1 when CaMKKβ is functional, opening new perspectives for the treatment of LKB1-null tumors.

11Growth of AR + and AR− LNCaP and PC3 cells was analyzed over 4 days, following the addition of MT 63–78. A dose-dependent decrease in cell number, concomitant to the activation of AMPK signaling was observed (Fig 4A). The growth inhibitory effect of MT 63–78 was also evaluated in several AR+ CRPC cell models (CL1, C4-2, and 22Rv1), as well as in the C4-2B subline derived from bone metastasis. Forty-eight hours incubation with MT 63–78 significantly reduced cell growth in all CRPC cell models (Fig 4B). High concentrations of the compound (50 μM) exerted a more moderate effect in immortalized non-transformed iPrEC and RWPE-1 cells, and low concentrations (5 μM) do not affect RWPE-1 cells, while still inhibiting PCa cell growth. These data suggest that the compound mostly affects cancer cell growth (Supplementary Fig 5A and B). We compared the efficacy of MT 63–78 and A-769662 in inhibiting AR sensitive and CRPC PCa cell growth and showed that MT 63–78 has 16–40 times higher potency than A-769662 (Supplementary Fig 5C). In addition, the growth inhibitory effect of MT 63–78 was not limited to PCa cells, and included LKB1-null A549 and BRAF mutated (V600E) KTC-1 cells (Supplementary Fig 6).

12Using AMPK α1−/− and α2−/− MEFs, we confirmed that MT 63–78-mediated inhibition of cell growth is strictly dependent on the functional activity of AMPK (Fig 4C) in contrast to AICAR and metformin, as previously shown by other groups (Ben Sahra et al,; Park et al,; Gonzalez-Girones et al, ). At concentrations that do not affect cell number (1 μM), MT 63–78 still inhibits anchorage-independent growth (Fig 4D and E).

13Cell cycle analysis using propidium iodide (PI) showed that 24-h treatment with MT 63–78 induces a significant enrichement in the G2/M population in both androgen sensitive and CRPC cell models (Fig 5A). G2-M arrest was particularly dramatic in the p53 null PC3 cells, suggesting a p53-independent effect. In contrast, in LNCaP cells, G2/M arrest was more evident after 48-h treatment (Supplementary Fig 7A). These finding were also confirmed in other tumor models (Supplementary Fig 7A). In depth morphological and molecular characterization indicated that MT 63–78 induces arrest in M phase. We observed accumulation of rounded and detached cells, and chromosome condensation in pro-metaphase (Fig 5B). Western blot analysis showed increased phosphorylation of histone H3 (marker of mitosis) in treated cells and accumulation of the mitotic regulator cyclin B1 (Fig 5C), suggesting that arrest takes place before anaphase when most of cyclin B1 is degradated (Clute & Pines,; Chang et al, ), most likely during pro-metaphase. In contrast to nocodazole treatment, no significant changes in cdc2 expression were observed (Supplementary Fig 7B). Moreover, the viable cells recovered from the conditioned media of MT 63–78 treated dishes showed high levels of cyclin B1 and positivity for phospho-histone H3, in keeping with the known looser attachment of cells to the substrate during mitotic rounding (Fig 5C). Consistent with the accumulation of cyclin B1, we observed accumulation and/or increased activation of key mitotic kinases such as Aurora kinases A, B, and Polo-like kinase 1 (PLK1) (Fig 5C). Evidence is now accumulating that a fine-tuned spatial and temporal activation of AMPK at the mitotic apparatus is required for correct spinde assembly/orientation, chromosal segregation, and proper cell division (Lee et al,; Vazquez-Martin et al,,; Thaiparambil et al, ). Thus, our results suggest that MT 63–78-induced sustained activation of AMPK alters these dynamics and does not allow cells to undergo a correct cytockinesis, as previously shown with metformin (Vazquez-Martin et al, ). Recently, it has been demonstrated that drug-induced prolonged M phase arrest leads to telomere deprotection and activation of DNA damage response with ATM activation and increased phosphorylation of histone H2AX (γ-H2AX), which can also be found in mitotic arrested cells undergoing apoptosis (Rogakou et al,; Hayashi et al, ). Markedly increased phosphorylation of histone H2AX was indeed observed upon mitotic block induced by MT 63–78 (Fig 5D). Mitotic arrested PCa cells undergo cell death via activation of the intrinsic apoptotic pathways. We observed p53 activation, caspase-3, parp cleavage, and outer membrane permeabilization in all the models assessed, including the p53 null PC3 cells. The extent of apoptosis in PC3 cells was however reduced with respect to p53 wt cells (Fig 6B). To identify the underlying molecular mechanisms, we evaluated the expression of the most critical Bcl-2 family members. MT 63–78 treatment induces reduction of anti-apoptotic Mcl-1 in concert with accumulation of the pro-apoptotic BH3-only protein Puma (Fig 6A) in all PCa cells. In contrast, Noxa accumulation was only observed in 22Rv1 cells. Bcl-2 levels were moderately reduced in C4-2, C4-2B, CL1, and 22Rv1 cells. Pro-survival XIAP protein was also found reduced in C4-2 and C4-2B cells (Supplementary Fig 8).

14The induction of autophagic response upon AMPK activation has been previously reported (Hoyer-Hansen & Jaattela,; Egan et al, ). However, only a modest effect of MT 63–78 on authophagy flux was observed in LNCaP and PC3 cells following 24-and 48-h treatment, respectively (Fig 6C).

15We next turned our attention on the two AMPK downstream pathways known to play a dominant role in the control of PCa cell growth/proliferation and survival. We, therefore, assessed the relative contribution of mTORC1 and lipogenesis in mediating PCa growth inhibition, following AMPK direct activation. AMPK inhibits the mTORC1 pathway through phosphorylation of the TSC2 complex and Raptor, whereas it affects de novo FA and cholesterol synthesis through phosphorylation of the rate limiting enzymes ACC and 3-hydroxy-3-methyl-glutaryl-CoA reductase (HMGCR) and of the transcriptional regulators SREBPs, as well as via transcriptional repression of lipogenic genes (Clarke & Hardie,; Zhou et al, ). Both the mTORC1 (assessed by phosphorylation of Raptor, S6RP, and 4EBP-1) and the lipogenic pathways were inhibited by MT 63–78. Increased phosphorylation of ACC was observed in LNCaP and PC3 cells while expression of ACC and FA synthase (FASN) was also significantly downregulated (Fig 7A). Levels of the mature form of SREBP-1, responsible for the transcription of ACC and FASN, were significantly reduced (Fig 7B). FASN activity was repressed as well (Supplementary Fig 9A). However, when we compared MT 63–78 to the mTORC1 inhibitor rapamycin, we observed that the former caused a greater inhibition of PCa cell growth despite a more modest reduction of S6RP phosphorylation (Fig 7C and D). Moreover, re-activation of the mTORC1 pathway, by transient transfection of constitutively active and rapamycin insensitive S6 Kinase 1, was not sufficient to rescue MT 63–78 mediated-cell growth inhibition (LNCaP, P = 0.9950; PC3, P = 0.9740, one-way anova followed by Tukey's post hoc test) (Fig 7E).

16Taken together, these results suggest that the blockade of the mTORC1 pathway is not the principal mechanism responsible for growth inhibition mediated by direct AMPK activation. In contrast to rapamycin treatment, no increase in Akt and MAPK phosphorylation was detected in LNCaP and PC3 cells treated with MT 63–78, suggesting that direct AMPK activation does not induce undesired mitogenic feedback loops (Fig 7F).

17To corroborate the inhibitory effect of MT 63–78 on lipogenesis, we measured 14C-acetate incorporation into lipids extracted from LNCaP cells and observed a dose-dependent reduction in 14C-acetate incorporation at all the time points analyzed (Fig 8A). Comparable results were observed in PC3 cells (Supplementary Fig 9B). As expected, most of the radioactivity was incorporated into the phospholipid fraction, with lower amounts in triacylglycerols and other neutral lipids. AMPK activation affects primarily de novo production of phospholipids but also significantly decreases the production of neutral lipids (triacylglycerides, cholesterol, cholestererol esters, and diacylglyclerols) (Fig 8B). We then determined whether the repression of lipogenesis accounts for MT 63–78-mediated cell growth inhibition. Addition of exogenous mevalonate and palmitate, the final products of HMGCR and FASN, rescued partially but significantly LNCaP cell growth. Surprisingly, no additive effect was observed with palmitate and mevalonate together (Fig 8C). Of note, neither palmitate nor mevalonate showed an effect on growth of untreated LNCaP cells (Supplementary Fig 9C). Moreover, the addition of the ACC inhibitor TOFA and HMGCR inhibitor simvastatin, alone or combined, inhibited cell growth similarly to MT 63–78 (Fig 8D). These results suggest that the growth inhibitory effect of direct AMPK activation in PCa cells is mediated in large part through inhibition of FA and cholesterol synthesis (Fig 8E).

18Enhanced lipogenesis is associated with the emergence of CRPC phenotype. The current standard of care for advanced CRPC is based on the administration of AR antagonists/inhibitors of androgen synthesis. Recently, MDV3100 and abiraterone have been FDA approved for the treatment of naïve or post-docetaxel advanced CRPC. However, emergence of treatment resistance to these new drugs represents a major limitation. Thus, we tested whether a combination treatment of MT 63–78 with AR antagonists/androgen synthesis inhibitors could result in increased therapeutic response. MT 63–78 significantly enhances the growth inhibitory effect of AR antagonists (bicalutamide and MDV3100) and abiraterone in both LNCaP and CRPC cells (Fig 9A–C, Supplementary Fig 10). MT 63–78 reduces AR levels and combinatorial treatments result in further reduction of AR and PSA expression levels when compared to single treatments (Fig 9A–C).

19We next sought to determine if the direct activation of AMPK could inhibit tumor growth in an in vivo model. Thirty-two nude mice bearing LNCaP tumors measuring 50–100 mm3 were treated intraperitoneally (i.p.) with MT 63–78 at 30 mg/kg daily (n = 17) or with 5% hydroxypropyl beta-cyclodextrine as control (n = 15) for 14 days. MT 63–78 led to a 33% inhibition of tumor growth (P = 0.049, Fig 10A). Consistent with our in vitro findings, MT 63–78 promotes AMPK activation in xenograft tumors and inhibits the lipogenesis and mTORC1 pathways, as assessed by increased ACC and Raptor phosphorylation, respectively (Fig 10B and C). Importantly, the highest levels of ACC phosphorylation were found in the two mice showing the most significant tumor volume reduction (Fig. 10C, black circles). Despite the fact that MT 63–78 reduced the glycemia and improved glucose tolerance in animal models of obesity and insulin resistance (Supplementary Fig 11A and B), no effect on weight, serum glucose and triglycerides levels were observed in non-obese nude mice maintained on normal chow diet, following MT 63–78 treatment (Supplementary Fig 11C and D). Prolonged treatment (21 days) with higher dose of MT 63–78 (60 mg/kg) resulted in a more robust inhibition of LNCaP xenograft growth (P = 0.004, end of treatment) (Fig 10D) as well as increased pharmacodinamic markers p-ACC and p-Raptor in treated tumors compared to controls (Fig 10E). Slight increase of apoptotic marker cleaved-parp was observed in the treated tumors (Supplementary Fig 11E). Serum PSA levels were measured at the end of the treatment. Mean PSA level in treated group was lower than in control, although the difference did not reach statistical significance (Fig 10F).

20Since our results demonstrated that AMPK direct activation inhibits PCa tumor growth, we hypothesized that the loss of AMPK catalytic alleles and, as a result, activity, would favor PCa development. Thus, we crossed prostate specific FASN transgenic (FASN-Tg) mice, which develop age-dependent prostate epithelial neoplasia (PIN) (Migita et al, ) with AMPK α2−/− mice. At the age of 13–16 months, 34 FASN-Tg/AMPK α2+/+, 28 FASN-Tg/AMPK α2+/−, 7 FASN-Tg/AMPK α2−/− mice were sacrificed and the three prostate lobes (anterior, ventral, dorso-lateral) were collected for histological examination. Despite the expression of AMPK α1 catalytic isoform in the murine prostate, we observed increased incidence of PIN in FASN-Tg/AMPK α2−/−, suggesting that loss of α2 subunit is sufficient to enhance FASN-induced tumorigenesis (Fig 10G). The odds of PIN occurrence increase with the number of knockout alleles (logistic regression P = 0.077, with post hoc calculated power of 0.44).

21Since acetate is a precursor for FA and cholesterol synthesis, we performed Positron Emission Tomography (PET) to investigate whether inhibition of lipogenesis by MT 63–78 was reflected by a reduction of 11C-acetate uptake by tumor xenografts. Three mice treated with MT 63–78 (30 mg/kg) or vehicle, 1 mouse treated with AICAR (400 mg/kg), and 1 mouse treated with FASN inhibitor C75 (30 mg/kg) were imaged with 11C-acetate PET before and after 24-h treatment with compounds. The FASN inhibitor C75 caused a 20% reduction in the uptake, whereas AICAR induced a slight increase (16%). No difference was observed with MT 63–78 (Fig 11A). These results are consistent with the role of activated AMPK in promoting nutrients uptake, their catabolism for ATP production, and mitochondrial biogenesis (Steinberg & Kemp, ). To investigate whether reduction in 11C-acetate uptake for lipids synthesis may be compensated for by increased uptake for its oxidation, we incubated LNCaP cells, previously treated with MT 63–78 and AICAR for 4 h, with 14C-acetate (2uCi) for 2 h. At the end of the treatment, the production of 14C-CO2 was measured. A significant increase in 14C-CO2 was observed, confirming that AMPK activation promotes acetate oxidation in the TCA (Fig 11B). No increase in 14C-CO2 production was observed after 30 min of AMPK activation, consistent with our previous data showing no difference in ATP levels at that time point (Fig 2C).

22The master metabolic regulator AMPK has recently emerged as a potential therapeutic target for cancer treatment. AMPK activators such as AICAR and metformin have been shown to inhibit PCa growth in culture, in xenografts, and in genetically engineered mouse models (GEMMs) (Xiang et al,; Huang et al,; Zakikhani et al, ). Moreover, epidemiologic data including biguanides, physical activity, caloric restriction, and increased adiponectin levels have been reported to both activate AMPK and lower cancer risk, including PCa (Jiang et al,; McTiernan,; Li et al, ). However, AMPK-independent effects of drugs tested to date (Ben Sahra et al,,; Moreno et al,; Garcia-Garcia et al,; Kalender et al,; Santidrian et al, ) prevented investigators to unequivocally determine whether AMPK activation was the necessary and sufficient mechanism of inhibiting carcinogenesis. In search for an adequate tool to address this unresolved question, we characterized and utilized in our experiments the novel AMPK direct activator MT 63–78. This new tool allowed us demonstrating that AMPK activation itself is necessary and sufficient for the suppression of PCa cell growth, in both androgen-dependent and CRPC models. Importantly, the growth inhibitory response to MT 63–78 was not affected by the status of the upstream AMPK-activating kinase LKB1. Since 30% of somatic NSCLCs harbor LKB1 inactivating mutations, often associated with KRAS mutations (Mahoney et al,; Shackelford & Shaw, ), our results open new therapeutic opportunities for LKB1 deficient and LKB1 deficient/KRAS activated tumors.

23The use of a complementary approach with a mouse model of genetic ablation of α2 subunit in the context of FASN overexpression underscored the crucial role of AMPK pathway in PCa progression, previously demonstrated for other tumor models (Huang et al,; Faubert et al, ) and further supports the potential of targeting AMPK axis for PCa treatment.

24Cell cycle analysis shows that MT 63–78 induces mitotic arrest with accumulation/activation of cyclin B1, Aurora kinases A and B, and PLK1, involved in mitotic spindle assembly, chromosome segregation and cytokinesis. These results are in line with literature observations highlighting the role of AMPK as cell cycle regulator, beyond its metabolic activity (Lee et al,; Vazquez-Martin et al,,b,c; Banko et al, ). From these studies it has emerged that a fine-tuned biphasic activation of AMPK is required for proper mitotic progression, whereas any alteration of AMPK expression or function (obtained by pharmacological sustained activation or AMPK depletion) alters its spatial and temporal regulation, resulting in microtubule misalignement, spindle misorientation, abnormal chromosome segregation followed by mitotic catastrophe and polyploidy (Vazquez-Martin et al,; Banko et al,; Thaiparambil et al, ). Our data are also supported by recent experiments done in our laboratory, using synchronized HeLa cells as model, which show that increased de novo fatty acid synthesis concomitant to reduced AMPK activity and phosphorylation of ACC is required for cytokinesis initiation (Natalia Scaglia, unpublished results). In this view, AMPK orchestrates the completion of mitosis by coordinately regulating mitotic spindle assembly, chromosomal segregation, and cytoskeleton rearrangement with membrane synthesis. Thus, MT 63–78 induced mitotic arrest could be ascribed to both a persitent AMPK-mediated inhibition of de novo FA synthesis (metabolic role) as well as to mitotic spindle assembly/chromosome segregation abnormalities (non-metabolic role). Moreover, we observed induction of DNA damage signaling following prolonged mitotic arrest, which potentially serves as a mitotic duration checkpoint, responsible for eliminating cells that fail to progress through mitosis properly. Indeed, activation of the intrinsic apoptotic pathway was observed in our PCa models concurrent to Puma accumulation and reduction of Mcl-1 expression. This is consistent with the finding that AMPK activation by AICAR or upon block of glycolysis results in mTORC1 inhibition-dependent decrease of Mcl-1 through regulation of its translation (Pradelli et al, ). Interestingly, we also observed cell cycle arrest and apoptosis in p53-deficient context, suggesting that MT 63–78 might also be suitable for the treatment of p53 null/mutated PCas not responsive to biguanides (Ben Sahra et al, ).

25PCa is characterized by alterations in the PI3k/Akt/mTOR pathway and by a unique lipogenic reprogramming, active both at tumor initiation and in advanced stages (Suburu & Chen, ). Since multiple components of the PI3k/Akt/mTOR pathway are deregulated in a wide variety of solid tumors, considerable attention has been focused on the biochemical and genetic characterization of this pathway in cancer and on developing new drugs to inhibit it (Wong et al,; Laplante & Sabatini, ). However, activation of de novo lipogenesis has recently been recognized, not only as an adaptation to fulfill the metabolic requirements of highly proliferating cancer cells, but as a hallmark of tumorigenesis, tumor progression, chemoresistance, and development of androgen-resistance (Ettinger et al,; Santos & Schulze, ). Newly synthesized lipids affect a number of cellular processes such as membrane biogenesis, signal transduction, intracellular trafficking, and lipid-based protein modification to promote prostate cancer cell proliferation and survival (Zadra et al, ). Indeed, overexpression and increased activity of FASN, a key enzyme responsible for the terminal step in FA synthesis, represents one of the most frequent phenotypic alterations in cancer cells (Brusselmans & Swinnen,; reviewed in Flavin et al, ). Our data support the essential role of lipogenesis in PCa maintenance and progression and point out that AMPK-mediated repression of lipogenesis plays a dominant role in tumor growth inhibition.

26Androgens are the major drivers of PCa carcinogenesis and progression and AR signaling is still active in castration resistant disease. Thus, targeting AR pathway remains the gold standard for both androgen sensitive and CRPCs. The AR antagonist MDV3100 and the CYP17A1 inibitor abiraterone have been FDA approved for the treatment of CRPC. However, despite encouraging results, emergence of treatment resistance represents a major limitation (Yuan et al, ). Previous studies have underscored the tight connection between AR signaling and increased de novo lipogenesis, expecially in the context of emergence of androgen independence (Swinnen et al,; Ettinger et al, ), suggesting that targeting lipogenesis will be particularly beneficial in CRPC. Our data showing MT 63–78-mediated growth inhibitory effect in CRPC models and the discovery of lipogenesis inhibition as the effector arm of AMPK-mediated suppression of PCa growth, prompted us to evaluate whether combination of AR signaling inhibitors and direct AMPK activation would result in a potentiated growth inihibitory effect. Low concentration of MT 63–78 was sufficient to significantly enhance the rate of growth inhibition induced by AR antagonists bicalutamide, MDV3100 or by the steroidogenesis inhibitor abiraterone (recently also discovered as AR antagonist) (Richards et al, ). This enhanced anti growth effect also correlates with reduction in AR expression and activity (PSA production). These results support previous data using combination of bicalutamide and the indirect activator metformin (Colquhoun et al, ) and point to the crucial role of AMPK activation rather than the inhibition of the insulin pathway in mediating the additive effect of the combinatorial treatment. Results from the phase II clinical trial NCT01677897, which evaluates the impact of the addition of metformin to abiraterone in metastatic PCa patients, will also inform on the benefit of adjuct AMPK activators to current standard treatments.

27PET is the most advanced technique for metabolic imaging and one of the most accurate tools for tumor staging in the pre-treatment and follow-up phases. Vavere et al showed a reduced tumor uptake of 11C-acetate in mice treated with FASN inhibitor C75, providing validation for development of 11C-acetate PET as measure of FA synthesis (Vavere et al, ). Thus, we investigated whether 11C-acetate could serve as in vivo biomarker of response to AMPK activation therapy. While our data with C75 fully confirm their findings, we report no decrease in 11C-acetate uptake following MT 63–78 treatment, showing that this is compatible with AMPK-mediated induction of 11C-acetate uptake for oxidation in the TCA cycle, concurrent with inhibition of FA synthesis. Our results thus discourage the use of 11C-acetate to monitor treatment with AMPK activators in the clinical setting, while the proliferation marker fluoro-L-thymidine (18F-FLT) might be a promising alternative (Habibollahi et al, ).

28In conclusion, this study provides the biochemical and molecular underpinning for direct targeting of AMPK in PCa therapy, with the downregulation of lipogenesis representing the major downstream effect of this therapeutic approach. This study set also the groundwork for testing the combination of AMPK activators and AR signaling inhibitors in CRPC pre-clinical models.

29A detailed list of the reagents used in the study is provided in the Supplementary materials and methods.

30Cell lines were maintained and cultured as described in the Supplementary materials and methods.

31Western blotting was performed as described in the Supplementary materials and methods.

32MT 63–78 was administered i.p. or orally using the antiacid Mylanta as vehicle in three C57 BL/6 male mice. Compound concentration in the plasma was measured by HPLC/MS, as described in the Supplementary materials and methods.

33AMPK activity in total cell lysates was measured using CycLex AMPK Kinase Assay kit (MBL International Corporation) or alpha screen technology (Perkin Elmer), according to the manufacturer's instructions. In vitro AMPK activity was assessed using recombinant human AMPK (α1β1γ1, α2β1γ1, α1β2γ1, α2β2γ1) and alpha screen technology. Details are provided in the Supplementary materials and methods.

34Kinase profiling was performed using The Upstate® KinaseProfiler service from Millipore.

35AMPK dephosphorylation experiments using recombinant AMPK α1β1γ1 were performed as described in the Supplementary materials and methods.

36Measurment of intracellular ATP, ADP, AMP levels were carried out using commercial kits and HPLC, as described in the Supplementary materials and methods.

37Details regarding siRNA and plasmids are provided in the Supplementary materials and methods.

38Cell growth was assessed by counting number of viable cells using the Vi-cell automatic counter (Beckman). Viability (% of viable cells/total cells) was also measured using the trypan blue exclusion method. Soft agar assays were performed as described in the Supplementary materials and methods.

39Cell cycle and apoptosis studies were performed by flow citometry and western blotting as described in the Supplementary materials and methods.

40Staining with Hoechst 33342 was performed according to standard procedures. Details are provided in the Supplementary materials and methods.

41FASN activity in PCa cell lines was measured as previously described (Kuhajda et al, ). Details are provided in the Supplementary materials and methods.

422-14C-acetate incorporation into lipids and Thin Layer Chromatography (TLC) were performed as previously described (Bligh & Dyer,; Bagnato & Igal, ) Details are provided in the Supplementary materials and methods.

432-14C -acetate oxidation experiments were performed using a standard protocol as described in the Supplementary materials and methods.

44All animal studies were carried out in compliance with National and DFCI guidelines. A detailed description of animal studies (drug treatments, GEMMs, and PET imaging) is provided in the Supplementary materials and methods.

45Collection of murine prostate and histological procedures were performed as described in the Supplementary materials and methods. Hematoxylin and eosin staining was blindly evaluated by an expert pathologist (Sabina Signoretti) for the presence of prostate lesions.

46Serum PSA levels were measured using Total Prostate-Specific Antigen ELISA kit (Alpco). Glucose levels were measured glucometer (Abbott) and QuantiChrom™ Glucose Assay Kit (Bio Assays Systems). Triglycerides were measured by serum triglyceride Determination Kit (Sigma), as described by the manufacturer.

47Statistical analyses were done using two-tailed, unpaired t-test, Wilcoxon Mann-Whitney test, one-way analysis of variance (ANOVA) followed by post-hoc Bonferroni, Tukey's or Dunnett's tests, as required, and two-way ANOVA test. Analyses were performed using the Graphpad software, version 5d, StatPlus: mac 2009, and R statistic packages. P values < 0.05 were considered significant. Logistic regression analysis was performed using R statistics packages.

48While metabolic reprogramming is now recognized as a hallmark of tumorigenesis and tumor progression, the development of metabolic-targeted therapy remains a challenge. Epidemiologic and experimental studies have identified the energy sensor AMPK as an ideal target not only for metabolic diseases but also for cancer. While biguanides such as metformin have cancer preventing/therapeutic properties, their metabolic effects beyond AMPK activation leave unanswered the question of whether AMPK activation per se is necessary and sufficient to affect tumor growth.

49Utilizing a novel, highly specific AMPK direct activator, we show that AMPK activation per se inhibits PCa cell growth in vitro and in vivo, whereas the allelic loss of AMPK catalytic subunits fosters PCa development. In addition, we report that AMPK anti-tumor effects are independent of the status of its upstream activator, the tumor suppressor LKB1. From a mechanistic standpoint, we demonstrate that inhibition of de novo lipogenesis is the dominant downstream effector of AMPK-mediated inhibition over mTORC1. Finally, we demonstrate that co-treatment of the direct AMPK activator with AR signaling inhibitors results in significantly enhanced growth inhibition in both androgen sensitive and CRPC models.

50This study proposes AMPK as a crucial metabolic hub that can be targeted in PCa. Because suppression of lipogenesis is the predominant effector arm of AMPK activation, AMPK activators may be especially effective in lipogenesis-driven tumors. Combining AMPK activators with hormonal therapy results in a synergistic anti-cancer effect. PET imaging can be utilized as a non-invasive biomarker of efficacy of metabolic therapy.