The potent AMPK inhibitor BAY-3827 shows strong efficacy in androgen-dependent prostate cancer models

Abstract

Purpose

5′ adenosine monophosphate-activated kinase (AMPK) is an essential regulator of cellular energy homeostasis and has been associated with different pathologies, including cancer. Precisely defining the biological role of AMPK necessitates the availability of a potent and selective inhibitor.

Methods

High-throughput screening and chemical optimization were performed to identify a novel AMPK inhibitor. Cell proliferation and mechanistic assays, as well as gene expression analysis and chromatin immunoprecipitation were used to investigate the cellular impact as well as the crosstalk between lipid metabolism and androgen signaling in prostate cancer models. Also, fatty acid turnover was determined by examining lipid droplet formation.

Results

We identified BAY-3827 as a novel and potent AMPK inhibitor with additional activity against ribosomal 6 kinase (RSK) family members. It displays strong anti-proliferative effects in androgen-dependent prostate cancer cell lines. Analysis of genes involved in AMPK signaling revealed that the expression of those encoding 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMGCR), fatty acid synthase (FASN) and 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 2 (PFKFB2), all of which are involved in lipid metabolism, was strongly upregulated by androgen in responsive models. Chromatin immunoprecipitation DNA-sequencing (ChIP-seq) analysis identified several androgen receptor (AR) binding peaks in the HMGCR and PFKFB2 genes. BAY-3827 strongly down-regulated the expression of lipase E (LIPE), cAMP-dependent protein kinase type II-beta regulatory subunit (PRKAR2B) and serine-threonine kinase AKT3 in responsive prostate cancer cell lines. Also, the expression of members of the carnitine palmitoyl-transferase 1 (CPT1) family was inhibited by BAY-3827, and this was paralleled by impaired lipid flux.

Conclusions

The availability of the potent inhibitor BAY-3827 will contribute to a better understanding of the role of AMPK signaling in cancer, especially in prostate cancer.

Supplementary Information

The online version contains supplementary material available at 10.1007/s13402-020-00584-8.

Introduction

5′ adenosine monophosphate-activated kinase (AMPK) is a heterotrimeric serine/threonine kinase composed of a catalytic α1 or α2 subunit, a scaffolding β1 or β2 subunit and a regulatory γ1, γ2 or γ3 subunit [1]. It is involved in cellular energy homeostasis and responds to low AMP cellular concentrations by rerouting metabolic processes in order to reduce consumption and increase generation of ATP [1–3]. Additional activation mechanisms such as glucose sensing have also been described, indicating a role beyond coordination of energy homeostasis [4]. The activity of AMPK is critically dependent on the phosphorylation of threonine T172 located in the activation loop of the α subunit, which is controlled by calcium-calmodulin-dependent kinase 2 (CAMKK2) and liver kinase B1 (LKB1) [5, 6]. The function of AMPK is tightly regulated by competitive binding of adenine nucleotides to dedicated AMPK γ subunit sites, which regulate access of different protein phosphatases that can reverse T172 phosphorylation. The downstream targets phosphorylated by AMPK are involved in various metabolic processes. One of the best characterized ones is acetyl-CoA carboxylase 1 (ACC1), a biotin carboxylase and carboxyltransferase involved in fatty acid biosynthesis [7]. A number of other targets, many of which are not involved in cellular metabolism, are phosphorylated by AMPK at a consensus recognition motif [5]. More recently, a role of AMPK in phosphorylating the methylcytosine dioxygenase ten-eleven translocation 2 (TET2) dependent on extracellular glucose levels, thus leading to epigenetic reprogramming, has been reported [8].

In view of the essential role of AMPK in different cellular processes, it is not surprising that several pathologies including cancer, metabolic dysfunction, inflammatory disorders and neurodegeneration have been associated with this kinase [3, 9–11]. Concerning cancer, pro-tumorigenic and tumor suppressor functions have been reported for AMPK, depending on parameters such as subcellular location, complex members and upstream modulators [12]. A link to MYC oncogenic activity and an impact on anabolic pathways ultimately leading to sustained cell proliferation has been proposed, but details remain to be elucidated [10]. In prostate cancer, important metabolic reprogramming takes place and the expression of several main players in this process, including CAMKK2, is controlled by androgens [13]. The role of CAMKK2-AMPK signaling in prostate cancer is gradually being unraveled [14, 15]. This pathway is stimulated during late-stage prostate cancer and may represent an essential survival mechanism [16–19]. Activated AMPK induces autophagy, promotes mitochondrial biogenesis and is involved in apoptosis [18, 20, 21], thereby fostering prostate cancer growth and survival, whereas silencing AMPK expression reduces prostate cancer cell proliferation and migration [13, 22]. However, other studies report that AMPK has a tumor suppressor role in prostate cancer [19, 23–25], so that potent and selective chemical probes [26] will be essential to understand the enigmatic function of AMPK in cancer and other pathologies [2, 3, 9, 12].

AMPK activators mainly bind to the allosteric drug and metabolite site located at the α/β subunit interface and have different isoform selectivities [11]. An inhibitory impact of these compounds on the proliferation of prostate cancer models has been reported, but in several cases promiscuous or weak activators were used, which makes interpretation difficult [5, 11, 12]. Data with more selective compounds such as MT 63–78 or 991 also suggest a role of AMPK in impairing prostate cancer cell proliferation via reduction of lipogenesis and independent of CAMMK2 [24, 25]. A phase 1 clinical study focusing on advanced tumors was initiated with the AMPK activator ASP4132 already several years ago, but only few preclinical data have been reported for this compound [27].

Concerning inhibitors, the pyrazolopyrimidine derivative compound C is a widely used ATP-competitive AMPK inhibitor which however lacks target selectivity, being additionally a strong blocker of different anaplastic lymphoma kinases and also a hinderer of cellular entry of the AMP analog AICAR, an activator of AMPK [28, 29]. MT 47–100 is both an inhibitor and activator of different AMPK complexes [30]. SU6656 is a SRC family kinase inhibitor also acting as an ATP-competitive AMPK antagonist [31]. SBI-0206965 was originally identified as an Unc-51 like autophagy activating kinase 1 (ULK1) inhibitor [32, 33]. It is a mixed-type inhibitor with IC₅₀ values around 330–400 nM for AMPK and 1 μM for ULK1, and blocks downstream cellular signaling [32]. It may however also inhibit additional kinases, as the selectivity panel used to characterize it only covers 11% of the human kinome [32]. Sunitinib is a multi-kinase inhibitor which also blocks AMPK, and derivatives with enhanced AMPK selectivity have very recently been identified and tested in a leukemia cell line [34].

Here we describe BAY-3827, a highly potent, low nanomolar inhibitor of AMPK. Profiling in a panel comprising 331 kinases showed only limited off-target activity. Strong cellular activity was demonstrated by determining ACC1 phosphorylation. Proliferation assays revealed that only androgen-dependent prostate cancer cells, and to a lesser extent multiple myeloma cells, strongly responded to this compound. Analysis focusing on genes involved in AMPK signaling showed the expression of many of them to be highly androgen-dependent. BAY-3827 repressed the expression of several genes involved in lipid metabolism. Consistent with this, the compound strongly impacted lipid turnover in treated prostate cancer cells.

Materials and methods

Compounds and cell lines

BAY-3827 and BAY-974 were synthesized in-house [35] and are available from the Structural Genomics Consortium as donated chemical probes. The androgen R1881 (methyltrienolone) was also synthesized in-house. Enzalutamide was purchased from Selleck Chemicals (Houston, TX, USA). Cell lines were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA) or the Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ, Braunschweig, Germany), and were routinely grown in the recommended media at 37 °C and in a 5% CO₂ atmosphere. Authentication was performed by the DSMZ using short tandem repeat DNA typing analysis. Cell lines were confirmed to be free of mycoplasma using the MycoAlert Mycoplasma Detection Assay (Lonza, Basel, Switzerland).

Kinase activity assays

A time-resolved fluorescence resonance  energy transfer (TR-FRET)-based AMPK activity inhibition assay was established. Full-length human AMPKα2 N-terminally fused to glutathione-S-transferase (GST) was coexpressed with GST-PRKAB1 and PRKAG1 using a baculovirus expression system (#102–114, lot 10CBS-1133D; Carna Biosciences, Kobe, Japan). Purification of the GST-AMPKα2/ß1/γ1 complex was performed using glutathione sepharose chromatography. The biotinylated peptide biotin-Ahx-HMRSAMSFAEPG (C-terminus in amide form) was used as substrate for the kinase reaction (Biosyntan, Berlin, Germany).

For the kinase assay, 50 nl of a 100-fold concentrated solution of the test compound in DMSO was pipetted into either a black low-volume 384-well microtiter plate or a black 1536-well microtiter plate (Greiner Bio-One, Frickenhausen, Germany). Then, 2 μl of a solution of GST-AMPKα2/ß1/γ1 in aqueous assay buffer (50 mM Hepes pH 7.5, 10 mM MgCl₂, 5 mM β-glycerophosphate, 2.5 mM dithiothreitol, 0.5 mM EGTA, 0.01% (w/v) bovine γ-globulin, 0.01% (v/v) Triton X-100) was added and the mixture was incubated for 15 min at 22 °C to allow pre-binding of the test compounds to the enzyme. For the kinase reaction with the low adenosine-triphosphate (ATP) condition, ATP was added at a final concentration of 10 μM, adenosine-monophosphate (AMP) at a final concentration of 2 μM and substrate at a final concentration of 0.5 μM in assay buffer, and the resulting mixture was incubated for 90 min at 22 °C. For the high ATP condition, ATP was added at a final concentration of 1 mM, AMP at a final concentration of 2 μM and substrate at a final concentration of 0.5 μM in assay buffer, and the resulting mixture was incubated for 30 min at 22 °C. The concentration of GST-AMPKα2/ß1/γ1 was adjusted depending on the activity of the preparation batch and was chosen to have the assay in the linear range, a typical concentration being 0.05 nM. The reaction was stopped by addition of 3 μl of a solution of TR-FRET detection reagents (0.2 μM streptavidine-XL665 [Cisbio Bioassays, Codolet, France], 3.33 nM anti-phospho-serine antibody STK [Merck Millipore, Burlington, MA, USA] and 3.33 nM anti-mouse IgG-Tb cryptate, a Terbium-cryptate labeled anti-mouse IgG antibody [Cisbio Bioassays]) in an aqueous EDTA solution (166.7 mM EDTA, 0.06% (w/v) bovine serum albumin in 50 mM HEPES pH 7.5). The resulting mixture was incubated for 1 h at 22 °C to allow the formation of a complex between the phosphorylated biotinylated peptide and the detection reagents. Subsequently, the amount of phosphorylated substrate was evaluated by measurement of resonance energy transfer from the Tb-cryptate to the streptavidine-XL665. For that, fluorescence emissions at 620 nm and 665 nm after excitation at 350 nm were measured in a TR-FRET reader (Pherastar FS, BMG Labtechnologies, Ortenberg, Germany or ViewLux, PerkinElmer, Waltham, MA, USA). The ratio of the emissions was taken as the measure of the amount of phosphorylated substrate. The data were normalized (enzyme reaction without inhibitor = 0% inhibition, all other assay components but no enzyme = 100% inhibition). As a rule, the test compounds were analyzed on the same microtiter plate in 11 different concentrations ranging from 20 μM to 0.07 nM in duplicate values. IC₅₀ values were calculated using the Genedata Screener™ software (Basel, Switzerland).

Aurora A kinase activity was determined using N-terminally His-tagged recombinant protein (Merck Millipore, Burlington, MA, USA) and the biotinylated FMRLRRLSTKYRT peptide (Jerini AG, Berlin, Germany). An anti-phospho-Akt antibody was used for detection, and europium cryptate-labeled protein-A and streptavidin-Xlent! for signal measurement (Cisbio Bioassays). FLT3 kinase activity was determined with N-terminally GST-tagged recombinant protein (Merck Millipore) and the biotinylated GGEEEEYFELVKKKK peptide (Biosyntan). Signals were determined following treatment with PT66-K and streptavidin-Xlent! (Cisbio Bioassays). c-Met kinase activity was determined using an in-house produced protein and PolyGT-Biotin (Cisbio Bioassays) as substrate. Signals were measured using Eu-W1024 PT66 (PerkinElmer) and streptavidin-Xlent! (Cisbio Bioassays). Ribosomal 6 kinase (RSK) 4 activity was determined using 6His-tagged recombinant protein (Eurofins, Brussels, Belgium) and the biotinylated KKLNRTLSFAEPG peptide (Biosyntan). Detection was carried out using an anti-phosphoserine antibody (Amersham Bioscience), and signals were measured after treatment with Eu-W1024 PT66 (PerkinElmer) and Dy648-conjugated streptavidin (Dyomics GmbH, Jena, Germany). A low ATP concentration (10 μM) was used in these assays.

Selectivity profiling was performed at Eurofins by testing 331 different wild-type human kinases (Table S1) and IC₅₀ values were determined for selected kinases (Table S2).

Cellular mechanistic assay

A phospho-ACC1 (Ser79) HTRF assay (Cisbio Bioassays) was used to determine the levels of ACC1 phosphorylation in cell lysates. Detection was performed through a sandwich TR-FRET assay using an anti-total ACC1 antibody labeled with the acceptor d2 and an anti-phospho-ACC1 antibody labeled with the donor europium cryptate. On day 1, the cells were seeded in a 384-well Small Volume™ plate (Greiner) in the appropriate culture medium. Next, they were treated with the inhibitor or DMSO and incubated for 1 h at 37 °C. Following incubation, the cells were lysed in 4 μl lysis buffer for 1 h on ice. Then 4 μl antibody solution containing equal amounts of total ACC1 and phospho-ACC1 were added and the samples were incubated overnight at 4 °C. The plate was measured the next day using a PHERAstar FS reader (BMG Labtech, Ortenberg, Germany). IC₅₀ values were calculated using the DRC Master Spreadsheet (in-house Bella software) and setting DMSO-treated cells as the minimum inhibition (C0) and staurosporine-treated cells (1 μM of staurosporine) as the maximum inhibition (Ci).

Cell proliferation assays

LNCaP, VCaP, 22Rv1, C4-2B, PC-3 and DU-145 prostate cancer cells were seeded at concentrations of respectively 600, 2400, 1200, 600, 600 and 1000 per well in 384-well white plates in RPMI-1640 medium without phenol red supplemented with 10% charcoal-stripped FBS, except in the case of DU-145 cells where DMEM/Ham’s F12 medium supplemented with stable glutamine and 10% FBS was used. Sister wells were seeded in a separate plate for time zero determination. All plates were incubated overnight at 37 °C. On the next day, the androgen R1881 was added at a final concentration of 1 nM (LNCaP), 0.1 nM (VCaP), 0.1 nM (22Rv1), 1 nM (C4-2B) and 0.1 nM (PC-3), whereas no R1881 was added in the case of DU-145 cells. This was followed by addition of the inhibitor in serial dilutions. The plates were incubated at 37 °C for 6 days, except for PC-3 and DU-145 cells which were incubated for 4 and 3 days, respectively. Cell viability was determined using a CellTiter-Glo® assay (Promega, Madison, WI, USA). Luminescence was measured in a Victor X3 reader. Background values measured in wells containing only medium were subtracted from all other values. Control wells containing cells with culture medium, DMSO and R1881 were used to determine control cell growth at the end of the experiment compared to cells treated only with DMSO.

The following additional cancer cell lines were tested by seeding the indicated numbers in 384-well black plates: Oci-Ly-7 (1000 cells), Ramos (1000 cells), REC-1 (5000 cells), RAJI (4000 cells), SU-DHL-10 (1000 cells), L-363 (4000 cells), JJN-3 (1000 cells), AMO-1 (2000 cells), SK-N-F1 (1600 cells), LP-1 (4000 cells), OPM-2 (4000 cells), SW780 (2000), IMR32 (1600 cells), U251 (2500 cells), IMR-5/75 (800 cells), MDA-MB-231 (800 cells), MDA-MB-453 (4000 cells), HeLa (800 cells), Colo201 (800 cells), Colo320 (800 cells), LS-174 T (800 cells), SW480 (800 cells), A549 (300 cells), NCI-H292 (500 cells), NCI-H460 (1000 cells), NCI-H520 (2000 cells), A2780 (800 cells), BxPC3 (2500 cells), PANC-1 (900 cells), A375 (4000 cells), Snu16 (1000 cells). Alternatively, 96-well plates were used for Daudi (20,000 cells), U2-OS (3000 cells) and Snu398 (3000 cells) cell proliferation assays. Culture conditions recommended by the provider were used. Sister wells were seeded on a separate plate for time zero determination. All plates were incubated overnight at 37 °C. On the next day, the inhibitor was added in serial dilutions and the plates were incubated at 37 °C for 72 to 144 h, as detailed (Table S3). The time zero plate was measured by adding 25 μl/well of CellTiter-Fluor solution (Promega), followed by incubation for 30 min at 37 °C and measurement of fluorescence on a PHERAStar reader. After the required incubation time, the plates were measured as described above. Background values measured in wells containing only medium were subtracted from all other values. Control wells containing cells with culture medium and DMSO were used to determine the control cell growth, compared to the initial number of cells (time zero value). To distinguish between cell growth inhibition and cell killing, the fluorescence values were corrected for the mean fluorescence observed for the time zero wells at the day of drug treatment start (time zero value). IC₅₀ values were calculated as above.

Western blotting

LNCaP cells were grown in 10-cm dishes and treated with 1 nM R1881 for the indicated timepoints. Protein extracts (40 μg) were separated on a NuPAGE 3–8% tris acetate gel (ThermoFisher Scientific, Waltham, MA, USA) and blotted onto a polyvinylidene difluoride membrane (Invitrogen, Carlsbad, CA, USA). Protein expression analyses were performed using monoclonal antibodies diluted as indicated and directed against ACC1 (1/1000; #3676), pSer79-ACC1 (1/1000; #3661), AMPKα (1/500; #2532), pThr172-AMPKα (1/500; #2535) or HSP90 (1/2000; #4874) (Cell Signaling Technology, Danvers, MA, USA), or against CAMKK2 (1/500; H00010645-M01; Abnova, Taipei City, Taiwan). Detection was performed using an Odyssey Imaging System (LI-COR Biosciences, Lincoln, NE, USA). ImageJ software (National Institute of Health, Bethesda, MD, USA) was used for quantification.

RNA expression analysis

LNCaP and VCaP cells were treated with 1 nM R1881 or BAY-3827, as indicated for 24 or 48 h. RNA was extracted using an RNeasy Plus Mini kit (Qiagen, Hilden, Germany). Synthesis of cDNA was performed using a SuperScript® III First Strand Synthesis SuperMix for qRT-PCR (ThermoFisher). Expression analysis was performed using the RT² Profiler™ PCR Array Human AMPK Signaling 330,231 tool (Qiagen). RNA expression levels were additionally obtained from several untreated cell lines by extracting data from the Cancer Cell Line Encyclopedia [36].

Chromatin immunoprecipitation DNA-sequencing (ChIP-seq) and bioinformatics analysis

VCaP cells were treated with 1 nM R1881 and 2 μM enzalutamide, as indicated, for 22 h. Cells were frozen and sent to Diagenode (Liège, Belgium) for ChIP reactions, as reported before [37]. Approximately 300 base pair-long DNA fragments were prepared using Bioruptor Pico with sonication beads (Diagenode). Immunoprecipitation was performed using antibodies specific for the AR (06–680, Merck Millipore). Sequencing reads were mapped to human genome hg19 using the Burrows-Wheeler alignment tool with default settings [38]. Additional steps were performed as reported before [37].

Lipid droplet determination by confocal microscopy and image analysis

Lipid droplet formation was determined using a fluorescent synthetic fatty acid precursor (Red fluorescent BODIPY 558/568 C12, D3835, ThermoFisher) [39]. Briefly, LNCaP cells were cultured in 96-well plates in RPMI-1640 supplemented with 10% charcoal stripped cFBS and treated with the indicated compounds for 2 or 4 days. Following treatment, they were first incubated for 24 h with 0.1 μM BODIPY 558/568 C12, then washed with medium and treated for 30 min with 4 μM Hoechst 33342 (ThermoFisher) to stain the nuclei and, finally, fixed with 1% formalin. In the case of VCaP cells, DMEM supplemented with 10% cFBS was used for culture and 1 μM C12 as precursor. This was followed by fixation and washing with 1% cFBS before staining with Hoechst 33342. Imaging was performed with a laser scanning microscope (LSM 700, Carl Zeiss Microscopy, Jena, Germany) using the predefined BTMR (C12 signal) or H342 (Hoechst 33342 signal) channels. Data were analyzed using ImageJ software and the ratio of lipid droplet to nuclei determined for each experimental condition.

Results

Identification of BAY-3827, a potent small-molecule AMPK inhibitor

A high-throughput biochemical screening was performed to identify AMPK inhibitors. Approximately 4 million compounds were tested and a hit from the dihydropyridine-dicarbonitrile series was selected. Optimization towards potency and selectivity led to the design of BAY-3827 (Fig. 1a). This compound inhibited AMPK kinase activity with IC₅₀ values of 1.4 nM at a low, 10 μM ATP concentration and 15 nM at a high, 2 mM ATP concentration (Table 1). Potential off-target activity was evaluated for Aurora A, FLT3, c-Met and RSK4, which showed IC₅₀ values ranging from 36 to 1324 nM, at least 25 times higher than that measured for AMPK (Table 1). Furthermore, a large panel of human, wild-type kinases was screened in the presence of 1 μM BAY-3827 at 10 μM ATP concentration. Inhibition leading to enzymatic activity below 20% of control levels was recorded for only a few kinases (Table S1). IC₅₀ values were subsequently determined to be 9 nM (RSK1), 52 nM (RSK2), 24 nM (RSK3), 43 nM (MSK1) and 94 nM (MST3) (Table S2). Altogether the data showed BAY-3827 to be a potent AMPK inhibitor with over 500-fold selectivity for most of the 331 kinases tested, and RSK family members being the closest off-targets identified. In order to have a proper negative control we generated BAY-974 (Fig. 1b), which had no inhibitory activity on AMPK or on FLT3, and only micromolar activity on c-Met (Table 1).

Fig. 1.

Androgen-dependent prostate cancer cell lines are sensitive to AMPK inhibition. a Chemical structure of the AMPK inhibitor BAY-3827. b Chemical structure of BAY-974, a structurally related inactive compound. c Cellular AMPK signaling measured by assessing ACC1 Ser79 phosphorylation with a TR-FRET assay following treatment with a dose-range of BAY-3827 overnight. Data represent mean IC₅₀ ± SD. d Prostate cancer cell viability determined by CellTiter-Glo® after treatment with a dose-range of BAY-3827 for 6 days. Data represent mean IC₅₀ ± SD

Table 1.

Kinase inhibitory activity of BAY-3827 and BAY-974

IC50 (nM)
	AMPK (low)	AMPK (high)	Aurora A	FLT3	c-Met	RSK4
BAY-3827	1.4 ± 0.6	15 ±10	1324 ± 272	124 ±140	788 ±120	36 ± 8
BAY-974	> 20,000	> 20,000	n.d.	> 20,000	8350 ±71	n.d.

Mean IC₅₀ ± SD values are given from at least 6 biological replicates for BAY-3827, and for 2 biological replicates for BAY-974. n.d.: not determined

BAY-3827 inhibits cellular AMPK signaling

ACC1 Ser79 phosphorylation was determined by TR-FRET to assess the cellular activity of BAY-3827 (Fig. 1c). A strong reduction of phosphorylation was observed in LNCaP and VCaP cells, and to a lesser extent in IMR-32 and especially in Colo320 cells, following compound treatment. The negative control BAY-974 had only weak effects on ACC1 Ser79 phosphorylation with IC₅₀ values of 4.2 and 6.4 μM in LNCaP and VCaP cells, respectively (not shown).

BAY-3827 shows preferential anti-proliferative activity in androgen-dependent prostate cancer models

The anti-proliferative effects of BAY-3827 were determined in a panel of cancer cell lines. Cell viability was measured after 6 treatment days and strong inhibitory effects were observed for LNCaP and VCaP cells, two prostate cancer cell lines that are responsive to androgen stimulation, whereas no activity was observed in andogen receptor (AR)-negative prostate cancer cell lines (Fig. 1d). A comparison of AMPK subunit expression showed no striking difference between responsive and non-responsive prostate cancer cell lines, but there was a higher expression of CAMKK2 in AR-positive models (Table S4). The RSK family members were expressed as well, but not FLT3. BAY-3827 also showed low micromolar efficacy in multiple myeloma cell lines (Table S3). These cell lines do not express the AR, whereas the AMPK subunit levels are comparable to those of the prostate cancer models we examined (not shown). Little impact of BAY-3827 was observed on a variety of models originating from different hematological and solid tumors (Table S3). The negative control compound BAY-974 had no anti-proliferative activity on prostate cancer cell lines (Table 2).

Table 2.

Anti-proliferative activity of BAY-3827 and BAY-974 in prostate cancer cell lines

IC50 (μM)
	LNCaP	VCaP	22Rv1	C4-2B	PC-3	DU-145
BAY-3827	0.28 ± 0.11	1.71 ± 0.28	5.55 ± 1.77	> 10	> 10	> 10
BAY-974	> 10	> 10	> 10	> 10	> 10	> 10

Mean IC₅₀ ± SD values from at least 2 biological replicates are given

Several AMPK pathway genes are strongly androgen-regulated

In view of the strong impact of BAY-3827 on prostate cancer models, we further explored the crosstalk between AMPK and AR signaling in LNCaP cells. Following treatment with different concentrations of the synthetic androgen R1881 for 24, 72 or 144 h, a dose-dependent phosphorylation increase of ACC1 was observed mainly at the latest timepoint (Fig. 2a, b). The effects were less pronounced for AMPK phosphorylation and CAMKK2 protein levels (Fig. 2a, b), but are in agreement with previously published data [13, 22].

Fig. 2.

Impact of androgen treatment on AMPK signaling in prostate cancer cell lines. a Western blot analysis of ACC1, AMPKα, CAMKK2 protein, and of ACC1 and AMPKα phosphorylation. b  Quantification was performed using ImageJ software. Normalization was versus the respective HSP90 protein levels

We further expanded on these findings by determining the impact of androgen treatment on the expression of genes linked to AMPK signaling. RNA was extracted from LNCaP cells treated with DMSO only or with 1 nM R1881 for 24 h, and analyzed using an AMPK signaling RT² array (Table S5). CAMKK2 was the most strongly upregulated gene with a 3-fold elevated level compared to the DMSO control (Fig. 3a). Expression of the genes coding for 3-hydroxy-3-methyl-glutaryl-coenzyme A reductase (HMGCR), ACC1, fatty acid synthase (FASN), 6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 2 (PFKFB2) and peroxisome proliferator-activated receptor gamma coactivator 1-beta (PPARGC1B) was upregulated over 2-fold (Fig. 3a). The corresponding proteins are all involved in lipid biosynthesis. Analysis of VCaP cells revealed that the expression of PFKFB2, HMGCR and FASN was also strongly upregulated after 1 nM androgen treatment for 24 or 48 h, as well as that of cryptochrome 1 (CRY1), sterol regulatory element-binding transcription factor 1 (SREBF1) and serine/threonine-protein phosphatase 2A catalytic subunit beta isoform (PPP2CB) (Fig. 3a). A few other genes were upregulated over 2-fold in androgen-treated VCaP cells at single time points (Fig. 3a). There was no strong overlap between the AMPK signaling genes regulated by R1881 in LNCaP compared to VCaP cells (Fig. 3c).

Fig. 3.

Impact of androgen treatment and BAY-3827 on AMPK signaling in prostate cancer cell lines. a Expression analysis showing the most strongly up-regulated genes following treatment with the androgen R1881 in the indicated cell lines. b Expression analysis showing the most strongly down-regulated genes following treatment with BAY-3827 and R1881 in the indicated cell lines. c Venn diagram showing the overlap of genes up- or -down regulated over twofold after R1881 or BAY-3827 treatment

We next looked at AR binding to the two most strongly regulated genes in VCaP cells. For HMGCR and PFKFB2 we observed several AR binding peaks along and around the gene body, which were induced upon androgen application and strongly reduced by additional treatment with the AR antagonist enzalutamide (Fig. 4a, b). For CAMKK2, which is not regulated by androgen in VCaP cells, only a single AR peak was observed immediately upstream of the coding region (Fig. 4c). This is in line with a strong androgen regulation of the HMGCR and PFKFB2 genes in these cells.

Fig. 4.

Androgen-dependent AR binding at regulated genes. a HMGCR gene coding region. b PFKFB2 gene coding regions. c CAMKK2 gene coding region. AR ChIP-seq signals are shown for VCaP cells treated for 22 h with DMSO, 1 nM R1881 or 1 nM R1881 plus 2 μM enzalutamide (Enza)

AMPK inhibition down-regulates the expression of genes involved in lipid metabolism

We next looked at the direct transcriptional impact of AMPK inhibition on the regulation of selected genes, as before. LNCaP and VCaP cells grown in the presence of R1881 were treated for 24 or 48 h with BAY-3827. Their RNA was purified and analyzed on an AMPK signaling RT² array (Table S6). A strong, over 10-fold repression of LIPE gene expression was observed in LNCaP cells and 2.5 fold reduction in VCaP cells treated for 24 h (Fig. 3b). In addition, an 8-fold expression reduction of the serine/threonine kinase AKT3 was observed in LNCaP cells (Fig. 3b). Expression of cAMP-dependent protein kinase type II-beta regulatory subunit (PRKAR2B), a regulatory subunit of cAMP-dependent protein kinase type II, was reduced to half in VCaP cells (Fig. 3b) and almost reduced to half in LNCaP cells (not shown). SLC2A4 was the most strongly down-regulated gene in VCaP cells, with a 16-fold inhibition already observed after 24 h treatment with BAY-3827 (Fig. 3b). Expression of several genes from the mitochondrial carnitine palmitoyltransferase (CPT) family, which is involved in acyl carnitine formation, was also blocked by the AMPK inhibitor in VCaP cells (Fig. 3b). Importantly, in LNCaP cells, none of the top androgen-stimulated genes was affected by BAY-3827 treatment (Table S6). In VCaP cells, the situation was slightly different, as PRKAR2B and less so carnitine palmitoyl-transferase 1 (CPT1A) expression was inhibited by BAY-3827 treatment (Table S6). Altogether however, there was no strong overlap or significant correlation between androgen- and AMPK inhibitor-regulated genes, as revealed by Venn diagram and scatter plot analysis, respectively (Fig. 3c and Fig. S1).

AMPK inhibition leads to accumulation of lipid droplets

In view of the strong effects on genes involved in lipid synthesis, we examined the impact of BAY-3827 on fatty acid metabolism. LNCaP and VCaP cells were treated with R1881, and with 1 or 3 μM AMPK inhibitor and 1 μM enzalutamide. Next, BODIPY C12 and Hoechst 33342 were added and the lipid content in relation to cell nucleus number determined (Fig. 5). We found that treatment of androgen-stimulated LNCaP or VCaP cells with the AMPK inhibitor BAY-3827 significantly increased the formation of lipid droplets in comparison to androgen treatment only (Fig. 5a, b). Co-treatment with BAY-3827 and enzalutamide also led to a significant increase of lipid droplet formation in VCaP cells (Fig. 5b).

Fig. 5.

Impact of BAY-3827 on lipid droplet formation. Cells were incubated with the fatty acid precursor BODIPY 558/568 C12 and treated as indicated. Fluorescence was determined by laser scanning microscopy and is shown in relation to the signals measured after nucleus staining. a Ratio of lipid droplets to nuclei in LNCaP cells treated as indicated. *p = 0.0445, **p = 0.0069 compared to R1881-treated group using Dunnett’s multiple comparison test. b Ratio of lipid droplets to nuclei in VCaP cells treated as indicated. *p = 0.0044, **p = 0.0230, ***p = 0.0012 compared to R1881-treated group using Dunnett’s multiple comparison test

Discussion

A role of AMPK as an oncoprotein or a tumor suppressor, depending on subcellular context, complexes formed and upstream regulators, has been reported in several studies [12, 40]. AMPK inhibitors and activators both show anti-proliferative effects, but often the results need to be interpreted with caution, due to the poor efficacy and/or selectivity of many of the compounds evaluated [11, 40]. Here we present the highly potent AMPK inhibitor BAY-3827, which was identified by high-throughput screening followed by compound optimization. It belongs to the dihydropyridine-dicarbonitrile family and showed low nanomolar inhibitory activity for AMPK, both in the presence of low and high ATP concentrations. It displayed over 500-fold selectivity over the majority of human kinases tested, the main off-target activity being for RSK family members. The compound showed cellular activity, as demonstrated by the reduction of pACC1 and of expression of several genes belonging to the AMPK downstream pathway. BAY-3827 exhibited strong anti-proliferative activity only in a subset of tumor models, namely androgen-dependent prostate cancer cell lines and multiple myeloma cell lines.

We focused on prostate cancer models and analyzed the crosstalk between AR signaling and the AMPK pathway in detail. We found that androgen treatment strongly stimulated the expression of several genes belonging to the AMPK pathway, including many involved in lipid biosynthesis, with some differences between the two cell lines tested. This was paralleled by an increase of AR peaks in the gene bodies and upstream of the transcription start site, as exemplified for HMGCR and PFKFB2 in VCaP cells. HMGCR plays an essential role in cholesterol biosynthesis and its overexpression has been linked to enzalutamide resistance [41], possibly due to elevated intracrine androgen synthesis [42]. PFKFB2 promotes glucose uptake and subsequent de novo lipogenesis in prostate cancer cells [43]. CAMKK2 was the most prominently stimulated gene in LNCaP cells. The corresponding protein activates AMPK and reduces downstream lipogenesis, but also directly stimulates lipogenesis in an AMPK-independent way [25].

Conversely, AMPK inhibitor treatment blocked many genes involved in lipogenesis. Expression of LIPE, also named hormone-sensitive lipase, was strongly down-regulated by BAY-3827 in both prostate cancer cell lines analyzed. Few functional data are available on the role of this lipase in prostate, but one study has shown that it is involved in mitochondrial transport of free cholesterol upon interaction with the steroidogenic acute regulatory protein [44]. Co-culture experiments with adipocytes underline the role of LIPE in lipolytic processes associated with breast cancer cell proliferation [45]. The same study additionally linked breast cancer growth to increased levels of CPT1A, an enzyme involved in the transport of carnitine into mitochondria. We found CPT1 family members to be down-regulated by BAY-3827, supporting a similar pathway to be implicated in prostate cancer growth. Recent knock-down and overexpression studies underscore the role of CPT1 and β-oxidation in prostate cancer growth and treatment resistance [46, 47]. Indeed, determination of lipid content in treated cells showed that whereas androgen had little direct impact on lipid droplet formation, treatment with BAY-3827 led to a significant accumulation of intracellular lipids, most likely due to the observed down-regulation of CPT1 family members. Other genes strongly blocked by BAY-3827 have not been directly linked to lipid metabolism in tumor cells, but essential roles of AKT3 and PRKAR2B in prostate cancer proliferation have been reported [48, 49], suggesting that the inhibitory effects observed for BAY-3827 may also be due to the blockade of these genes. SLC2A4 expression was strongly down-regulated following treatment of VCaP cells with BAY-3827. SLC2A4 facilitates the diffusion of glucose into cells where it undergoes phosphorylation before entering glycolysis [50]. Glucose uptake and metabolism are generally associated with tumor growth, but this seems less prominent in prostate cancer [51]. Androgens are known to stimulate SLC2A4 function and glucose uptake in several cell types, but the precise role of this transporter in prostate cancer needs to be thoroughly investigated, especially in view of the overexpression of several SLC2A family members in this tumor type [50].

No significant overlap was observed between genes regulated by androgen and those regulated by BAY-3827, despite the preferential anti-proliferative effects observed in androgen-dependent prostate cancer models. This may be due to our focus on selected genes downstream of AMPK signaling and on the involvement of other regulatory mechanisms that were not analyzed in the present study, such as cross-talk between the AR and other signaling pathways [52], differences between regulation of RNA, protein and activity levels, or epigenetic regulation of AR signaling [53].

As mentioned above, unlike other tumor types, prostate cancer is not highly dependent on increased glucose uptake and aerobic glycolysis, as evidenced by the poor signals seen in prostate cancer patients when using F-deoxyglucose positron emission tomography [54]. Rather, prostate cancer has a unique dependency on lipid metabolism for energy production, which has been linked to enhanced fatty acid uptake [51], intracrine steroid synthesis and de novo fatty acid synthesis [51, 54]. Despite the fact that the expression of many enzymes involved in synthesis, uptake and oxidation of lipids is under androgen control [55, 56] and often upregulated in prostate cancer patients [57, 58], blockade of androgen signaling will eventually cease to be efficacious in prostate cancer patients, suggesting that novel therapeutic agents that directly address lipid metabolism may represent a novel therapeutic opportunity [59]. Indeed, resistance to androgen deprivation therapy has been associated with increased lipid metabolism, due for example to re-activation of dedicated genes by SREBP1 or the glucocorticoid receptor [60]. Our present data show that blockade of lipid metabolism inside prostate tumor cells using the AMPK inhibitor BAY-3827 has the potential to inhibit proliferation, and future studies are needed to show how this translates into the clinical setting.

Beside AMPK inhibition, BAY-3827 also shows effects on RSK kinases, mainly RSK1 and RSK3. A role of RSK1 in prostate cancer metastasis has been proposed using the PC-3 model [61], but since this model does not express the AR, no conclusion can be drawn on whether inhibition of this kinase is involved in the effects observed. The RSK inhibitor SL0101 affects the proliferation of LNCaP cells, but only at high micromolar concentrations [62, 63]. Comparable inhibitory results have been reported for the AR-negative cell line PC-3 [62]. Also, this compound has later been shown to act independently of RSK [64], so no conclusion can be drawn on the role of RSK in androgen-dependent prostate cancer. Finally, the fact that the RSK family members are expressed at similar levels in both cell lines responsive to BAY-3827 and those that are not, strongly suggests that inhibition of this kinase family is not responsible for the selective effects we observed.

In summary we identified BAY-3827 as a potent inhibitor of AMPK with additional activity against RSK family members. We found that it shows preferential anti-proliferative activity in androgen-dependent prostate cancer models and regulates the expression of several genes involved in lipid metabolism such as LIPE, PRKAR2B, AKT3 and CPT1 family members. The availability of BAY-3827 may help to better understand the role of AMPK in cancer, especially in prostate cancer.

Abbreviations

ACC1

acetyl-CoA carboxylase 1

AMP

adenosine monophosphate

AMPK

5′ adenosine monophosphate-activated kinase

AR

androgen receptor

ATP

adenosine triphosphate

ATCC

American Type Culture Collection

CAMKK2

calcium-calmodulin-dependent kinase 2

ChIP-seq

chromatin immunoprecipitation DNA-sequencing

CPT1

carnitine palmitoyl-transferase 1

CRY1

cryptochrome 1

FASN

fatty acid synthase

DSMZ

Deutsche Sammlung von Mikroorganismen und Zellkulturen

GST

glutathione-S-transferase

HMGCR

3-hydroxy-3-methyl-glutaryl-coenzyme A reductase

HTRF

homogeneous time-resolved fluorescence

LIPE

lipase E

LKB1

liver kinase B1

PFKFB2

6-phosphofructo-2-kinase/fructose-2,6-biphosphatase 2

PPARGC1B

peroxisome proliferator-activated receptor gamma coactivator 1-beta

PPP2CB

serine/threonine-protein phosphatase 2A catalytic subunit beta isoform

PRKAR2B

cAMP-dependent protein kinase type II-beta regulatory subunit

RSK

ribosomal S6 kinase

SREBF1

sterol regulatory element-binding transcription factor 1

TET2

ten-eleven translocation 2

TR-FRET

time-resolved fluorescence resonance energy transfer

ULK1

autophagy activating kinase 1

Author contribution

Clara Lemos, Volker Schulze and Bernard Haendler contributed to the study conception and design. Medicinal chemistry and computational chemistry: Volker Schulze, Tobias Heinrich, Julien Lefranc, Hans Briem, Lara Kuhnke and Clara Christ were in charge of the medicinal chemistry and computational chemistry part. Clara Lemos, Simon Baumgart, Benjamin Bader, Simon Holton, Ulf Bömer, Philip Lienau and Bernard Haendler were involved in the acquisition of pharmacology data. Clara Lemos, Volker Schulze, Simon Baumgart, Ekaterina Nevedomskaya, Benjamin Bader, Clara Christ, Ulf Bömer and Bernard Haendler were involved in the analysis and interpretation of data (i.e., statistical analysis, biostatistics, computational analysis). Franz von Nussbaum, Carl Nising, Marcus Bauser, Andrea Hägebarth and Dominik Mumberg supervised the study. The first draft of the manuscript was written by Bernard Haendler and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Data availability

Complete ChIP-seq data are available at NCBI GEO (https://www.ncbi.nlm.nih.gov/geo/) under GSE148358.

Compliance with ethical standards

Conflict of interest

All authors are/were employees and/or shareholders of Bayer AG.

Code availability

Not applicable.