Open Full Page

Published OnlineFirst February 12, 2018; DOI: 10.1158/0008-5472.CAN-17-2215

Cancer Metabolism and Chemical Biology Research

RSK Regulates PFK-2 Activity to Promote Metabolic Rewiring in Melanoma Thibault Houles1, Simon-Pierre Gravel2,Genevieve Lavoie1, Sejeong Shin3, Mathilde Savall1, Antoine Meant 1, Benoit Grondin1, Louis Gaboury1,4, Sang-Oh Yoon3, Julie St-Pierre2, and Philippe P. Roux1,4

Abstract

Metabolic reprogramming is a hallmark of cancer that includes glycolytic flux in melanoma cells, suggesting an important role for increased glucose uptake and accelerated aerobic glycolysis. This RSK in BRAF-mediated metabolic rewiring. Consistent with this, phenotypeisrequiredtofulfill anabolic demands associated with expression of a phosphorylation-deficient mutant of PFKFB2 aberrant cell proliferation and is often mediated by oncogenic decreased aerobic glycolysis and reduced the growth of melanoma drivers such as activated BRAF. In this study, we show that the in mice. Together, these results indicate that RSK-mediated phos- MAPK-activated p90 ribosomal S6 kinase (RSK) is necessary to phorylation of PFKFB2 plays a key role in the metabolism and maintain glycolytic metabolism in BRAF-mutated melanoma growth of BRAF-mutated melanomas. cells. RSK directly phosphorylated the regulatory domain of Significance: RSK promotes glycolytic metabolism and the 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase 2 (PFKFB2), growth of BRAF-mutated melanoma by driving phosphory- an enzyme that catalyzes the synthesis of fructose-2,6-bisphosphate lation of an important glycolytic enzyme. Cancer Res; 78(9); during glycolysis. Inhibition of RSK reduced PFKFB2 activity and 2191–204. 2018 AACR.

Introduction but recently developed therapies that target components of the MAPK pathway have demonstrated survival advantage in pati- Melanoma is the most aggressive form of skin cancer and arises ents with BRAF-mutated tumors (7). Drugs that inhibit mutated from melanocytes, which reside in the skin, eye, mucosal epithe- BRAF (vemurafenib, dabrafenib) lead to tumor regression and lia, and meninges of the brain and spinal cord (1). In cutaneous improved overall patient survival (8–10). Despite the promise of melanoma, the RAS/MAPK pathway is frequently activated due BRAF inhibitors in the clinic, intrinsic and acquired resistances to mutations in BRAF (52%), NRAS (28%), and NF1 (14%) were found to limit the therapeutic beneﬁt of these drugs (11). genes (2). More than 90% of BRAF mutations encode a protein The RAS/MAPK signaling cascade regulates a wide range of harboring the V600E mutation, which constitutively activates ERK cellular functions, including cell growth, proliferation, survival, 1/2 signaling and drives cell proliferation (3–5). Malignant mel- and metabolism (12, 13). Upon activation by upstream RAS anoma is highly resistant to conventional chemotherapy (6), isoforms or in response to mutational activation, RAF phosphorylates and activates MEK1/2, which themselves phosphorylate and activate ERK1/2 (14). Once activated, ERK1/2 phosphorylate 1Institute for Research in Immunology and Cancer (IRIC), Universite de Montreal, many substrates, including members of the p90 ribosomal S6 2 Montreal, Quebec, Canada. Department of Biochemistry, Goodman Cancer kinase (RSK) family of proteins (15). The RSK family comprises of Research Centre, McGill University, Montreal, Quebec, Canada. 3Department of four Ser/Thr kinases (RSK1–4) that share 73% to 80% sequence Cancer Biology & Pharmacology, University of Illinois College of Medicine, Peoria, Illinois. 4Department of Pathology and Cell Biology, Faculty of Medicine, identity and belong to the AGC family of basophilic protein Universite de Montreal, Montreal, Quebec, Canada. kinases (16). RSK has been shown to be hyperactivated in BRAF-mutated melanoma (17), and accordingly, inhibition of Note: Supplementary data for this article are available at Cancer Research Online (http://cancerres.aacrjournals.org/). RSK activity abrogates melanoma growth in mice (18). Although RSK regulates a number of substrates involved in cell growth and Current address for S.-P. Gravel: Faculty of Pharmacy, Universite de Montreal, Montreal, Quebec, Canada; current address for M. Savall: Institut Cochin, Uni- proliferation, a potential role for RSK signaling in cell metabolism versite Paris Descartes (Paris V); Centre National de la Recherche Scientiﬁque remains unknown. (CNRS), UMR8104, Paris 75014, France; Institut National de la santeetdela One of the hallmarks of cancer is metabolic reprogramming, recherche medicale (INSERM), U1016, Paris 75014, France; and current address which facilitates the uptake and incorporation of nutrients needed for J. St-Pierre: Ottawa Institute of Systems Biology (OISB), Department of to support the continuous growth and proliferation of cancer cells Biochemistry, Microbiology and Immunology, University of Ottawa, Ottawa, (19, 20). In particular, cancer cells have accelerated aerobic Ontario, Canada. glycolysis, which converts glucose to lactate regardless of oxygen Corresponding Author: Philippe P. Roux, UniversitedeMontr eal, Montreal, availability (21). Although much is known about the altered Quebec H3C 3J7, Canada. Phone: 514-343-6399; Fax: 514-343-5839; E-mail: glucose metabolism of cancer cells, more recent work has shown [email protected] that many of these changes are directly driven by oncogenes and doi: 10.1158/0008-5472.CAN-17-2215 tumor-suppressor genes (22). In melanoma, oncogenic BRAF is a 2018 American Association for Cancer Research. key player in metabolic reprogramming as it affects glucose

www.aacrjournals.org 2191

Houles et al.

metabolism in several ways (23). Activated BRAF downregu- after thawing. A375 stable cell lines were generated using lates expression of peroxisome proliferator–activated receptor g pBabe-puro–derived retroviral particles, and expressing cells were coactivator 1a (PGC-1a; refs. 24, 25), a transcriptional coactiv- selected using puromycin (2 mg/mL). HEK293 cells were trans- ator that promotes mitochondrial biogenesis and enhances oxi- fected by calcium phosphate precipitation as previously described dative phosphorylation (26), and stimulates glycolysis in part (37). Cells were grown for 24 hours after transfection and serum- by upregulating hypoxia-inducible factor 1 (HIF1) target genes starved overnight using serum-free DMEM where indicated. involved in glucose uptake and subsequent utilization along Starved cells were pretreated with PD184352 (10 mmol/L), the pathway (27, 28). One particular example is the upregulation LJH685 (10 mmol/L), BI-D1870 (10 mmol/L), PI-103 (1 mmol/L), of 6-phosphofructo-1-kinase (PFK-1; ref. 29), which is a key rapamycin (25 nmol/L), or Ku-0063794 (10 mmol/L; Biomol), regulatory enzyme that controls glycolytic ﬂux (30). One of the where indicated, and stimulated with phorbol 12-myristate most potent activators of PFK-1 is fructose-2,6-bisphosphate 13-acetate (PMA; 100 ng/mL) or EGF (25 ng/mL) before being (Fru-2,6-P2), whose levels are controlled by the bifunctional harvested. Unless indicated otherwise, all drugs and growth 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase (PFK-2) factors were purchased from Invitrogen. family of enzymes (PFKFB1–4; ref. 31). Although the expression of some PFK-2 family members has been shown to depend on RNA interference and viral infections HIF1 activity (32), a potential role for BRAF in the direct regu- Short hairpin RNA (shRNA)–mediated knockdown was lation of PFK-2 activity remains unknown. achieved using lentiviruses produced with vectors from the Mis- Recent phosphoproteomic studies performed in BRAF- sion TRC shRNA library (PFKFB2, TRCN37959, TRCN37960, mutated melanoma cells revealed PFKFB2 as a potential phos- TRCN37962; RSK1, TRCN470; RSK2, TRCN537). Cells were phorylation substrate of RSK (33, 34). Using various agonists infected in the presence of 4 mg/mL polybrene, and 3 days after and inhibitors of RAS/MAPK signaling, we found that PFKFB2 viral infection, A375 and WM164 cells were treated and selected is directly phosphorylated by RSK on sites that regulate its with 2 mg/mL puromycin. Retroviruses were produced in the function. RSK was found to promote glycolytic ﬂux in mela- Phoenix cell line using the pLPC-puro vector to overexpress noma cells, which supports melanoma growth. ectopic PFKFB2 WT, S466/483A, or S466/483D mutants. Two days after infection, cells were selected with 2 mg/mL puromycin.

Materials and Methods Immunoprecipitations and immunoblotting DNA constructs and recombinant proteins All cell lysates were prepared as previously described (38). The original plasmid encoding human PFKFB2 (pCR4-TOPO- Briefly, cells were washed 3 times with ice-cold PBS and lysed in PFKFB2) was obtained from the PlasmID Repository at Harvard BLB [10 mmol/L K3PO4, 1 mmol/L EDTA, 5 mmol/L EGTA, Medical School (Clone MGC:138310). This DNA construct was 10 mmol/L MgCl2, 50 mmol/L b-glycerophosphate, 0.5% used as template for generating 6Myc-tagged PFKFB2 in the Nonidet P-40, 0.1% Brij 35, 0.1% deoxycholic acid, 1 mmol/L fi pcDNA3 and pLPC backbones, as well as for the ampli cation sodium orthovanadate (Na3VO4), 1 mmol/L phenylmethylsul- of a C-terminal fragment of PFKFB2 (aa 431–505) that was fonyl fluoride, and a cOmplete protease inhibitor cocktail tablet inserted in pGEX-2T for bacterial expression. All PFKFB2 mutants (Roche)]. For immunoprecipitations, cell lysates were incubated (S466A, S483A, S466/483A) were generated using the Quik- with the indicated antibodies for 2 hours, followed by a 1-hour Change methodology (Stratagene). The vectors encoding for incubation with protein A-Sepharose CL-4B beads (GE Health- constitutively active MEK1 (MEK-DD) and RSK1 (WT and KD) care). Immunoprecipitates were washed 3 times in lysis buffer, were described previously (18, 35). The plasmids encoding b, s, g, and beads were eluted and boiled in 2 reducing sample buffer and z 14-3-3 isoforms were obtained from Addgene and described (5 buffer is 60 mmol/L Tris-HCl, pH 6.8, 25% glycerol, 2% SDS, previously (33, 36). 14.4 mmol/L 2-mercaptoethanol, 0.1% bromophenol blue). Eluates and total cell lysates were subjected to 8% to 12% Antibodies SDS-PAGE, and resolved proteins were transferred onto poly- Antibodies targeted against the Arg-X-X-pSer/Thr motif, vinylidene difluoride membranes for immunoblotting. The PFKFB2 and phospho-PFKFB2 (S483), Akt and phospho-Akt data presented are representative of at least three independent (S473), ERK1/2 and phospho-ERK1/2 (T202/Y204), S6K1 and experiments. phospho-S6K1 (T389), and S6 and phospho-S6 (S235/236) were purchased from Cell Signaling Technology. The phospho- Protein phosphotransferase assays PFKFB2 (S466) antibody was purchased from EMD Millipore. For RSK kinase assays, recombinant-activated RSK1 purchased Anti-Myc, anti-HA, anti-Flag, and anti-tubulin monoclonal from SignalChem was used with bacterially purified recombinant antibodies were purchased from Sigma-Aldrich. All secondary GST-PFKFB2 (aa 431–505) as substrate (WT and S466/483A), horseradish peroxidase–conjugated antibodies used for immu- under linear assay conditions. Assays were performed for 10 to 60 noblotting were purchased from Chemicon. minutes at 30C in kinase buffer [25 mmol/L Tris-HCl (pH 7.4), 10 mmol/L MgCl2, and 5 mmol/L b-glycerophosphate] supple- Cell culture and transfection mented with 5 mCi of [g-32P]ATP. All samples were subjected to HEK293, HeLa, A375, and WM164 cells were obtained from SDS-PAGE followed by immunoblotting, and incorporation of the American Type Culture Collection without further authenti- radioactive 32P label was determined by autoradiography using a cation. Cells were maintained at 37C in DMEM with 4.5 g/L Fuji PhosphorImager with ImageQuant software. The data pre- glucose supplemented with 5% (v/v) FBS, 100 IU/mL penicillin, sented are representative of at least three independent experi- and 100 mg/mL streptomycin. Cells were regularly PCR tested to ments. For PFKFB2 kinase assays, PFKFB2 was immunoprecipi- exclude mycoplasma contamination and used within 3 months tated with anti-Myc antibodies and washed thrice with ice-cold

2192 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFKFB2 and Glycolysis

lysis buffer and twice with ice-cold PFKFB2 kinase buffer Gas chromatography-mass spectrometry analyses (100 mmol/L Tris-HCl, pH 7.4, 500 mmol/L DTT, 100 mmol/L Cells grown in 6-well plates at 70% confluence and serum- EDTA, 2.5 mmol/L KH2PO4, 5 mmol/L MgCl2). Assays were starved overnight were treated with vehicle control (DMSO performed for 10 minutes at 30C in PFKFB2 kinase buffer 0.1% v/v) or with specific MEK/RSK inhibitors for 6 hours. supplement with 500 mmol/L of ultra-pure ATP and 10 mmol/ Cells were rinsed twice with ice-cold saline solution (9 g/L L of Fru-6-P. Kinase activity was quantified using ADP-Glo from NaCl) and quenched with 600 mL dry ice-cold 80% (v/v) Promega. Briefly, detection was carried out by incubating 10 mLof methanol/distilled and deionized water. Samples were sonicat- kinase reaction, 10 mL of ADP-Glo Reagent for 40 minutes, ed for 10 minutes using the Bioruptor (Diagenode Inc.) kept in followed by adding 20 mL of Kinase Detection Reagent for 60 a cold room and set at "high" setting with 30-second ON and minutes. Luminescence was detected using a LumiStar lumin- 30-second OFF cycles. Samples were cleared by centrifugation at ometers from BMG LabTech. The data presented are representative 13,000 g for 10 minutes at 4C, and 800 ng of the internal of at least three independent experiments. standard myristic-acid-D27 dissolved in pyridine was added to each tube. Samples were dried overnight in an orbital trap Cell proliferation assays device (Labconco). Dry pellets were dissolved in methoxya- Cell proliferation in vitro was assessed by cell counting. Expo- mine-HCl (10 mg/mL; Sigma-Aldrich) in pyridine, vortexed nentially growing A375 and WM164 cells were grown in 12-well for 30 seconds, sonicated for 30 seconds, and centrifuged for plates in DMEM supplemented with 5% FBS, and the relative 10 minutes at 13,000 g. Pellets were discarded, and samples number of viable cells was measured at 24, 48, 72, and 96 hours. were placed in capped amber gas chromatography-mass spec- The results displayed represent the mean of triplicates SEM. trometry (GC-MS) injections vials and incubated for 30 minutes at 70C (methoximation), followed by derivatization with N-tert-Butyldimethylsilyl-N-methyltrifluoroacetamide with 1% Xenotransplantation in nude mice tert-Butyldimethylchlorosilane (MTBSTFA þ 1% t-BDMCS, Sig- 5 A375 cells (2 10 ) stably expressing PFKFB2 WT, or the S466/ ma-Aldrich) for an additional 60 minutes at 70C. Note that 483A and S466/483D mutants were washed once in 200 mLof 1 mL was injected into an Agilent 5975C GC/MS configured for fl PBS and injected subcutaneously in the anks of nude mice scan acquisition and detailed elsewhere (39). Data analyses 6 (8 per groups). Note that 3 10 WM164 shNT or shRSK1/2 were performed using the Chemstation software (Agilent). All fl were injected subcutaneously in the anks of nude mice (5 per metabolites shown were validated using authentic standards groups). Mice were regularly monitored, and tumors were (Sigma-Aldrich). Derivatized metabolites were identified accord- measured every 2 to 3 days using a caliper. Mice were euthanized fi 3 ing to their speci c retention time and fragmentation pattern once tumors reached 1,500 mm . Animals were handled in strict (presence of quantifier and 1 to 3 qualifier ions with expected fi accordance with good animal practice as de ned by the relevant relative intensities) that were previously listed (39, 40). Meta- local animal welfare bodies, and all experiments were approved bolites amounts obtained from mþ0 intensities were corrected by the Canadian Council on Animal Care. for the internal standard myristic-acid-D27 ion integration and cell counts obtained with a TC10-automated cell counting L-lactate and D-glucose measurements device (Bio-Rad). Note that 10 mL of cell growth media were collected from cell plates after 6- to 24-hour incubations on serum-starved con- Fructose 1,6-bisphosphate measurements by LC-MS dition, spun down to remove cells, and diluted in cold 80% For targeted metabolite analysis and semi-quantitative con- (v/v) methanol/distilled and deionized water. Samples were centration determination of fructose 1,6 bis phosphate, 5 mLof stored at 80C until further manipulations. Collected sam- each sample resuspended in 50 mL of water was injected onto an ples were thawed on ice and spun down at 21,000 x g for 10 Agilent 6430 Triple Quadrupole (QQQ)-LC-MS/MS. Chromato- minutes at 4C to remove residues. Cleared samples were dried graphy was achieved using a 1290 Infinity ultraperformance LC in an orbital trap device (Labconco) for 6 to 16 hours. Pellets system (Agilent Technologies) consisting of vacuum degasser, were solubilized in 100 mL deionized water, vortexed, and autosampler, and a binary pump. Mass spectrometer was equip- sonicated to ensure solubilization. Each sample (50 mL) was ped with an electrospray ionization source, and samples were assessed for L-lactate content using the L-lactate colorimetric analyzed in negative mode. Multiple reaction monitoring transi- assay Kit I (Eton Biosciences Inc.), and following the manu- tions were optimized on standards for each metabolite quanti- facturer's instructions with the little modifications, standard tated. For fructose 1,6-bisphosphate, the following quantifier curve for L-lactate (#71718, Sigma-Aldrich) was prepared in and qualifier transitions were monitored respectively: 339!97 growth media (0.3–30 mmol/L), diluted in 80% (v/v) meth- and 339!79. Gas temperature and flow were set at 350C and anol/distilled and deionized water, dried, and reconstituted 10 L/min, respectively, nebulizer pressure was set at 50 psi, and similarly to media collected from cell cultures. Assay was capillary voltage was set at 4,000 V. Chromatographic separa- performed in 96-well plates with each experimental condition tion was performed on a Scherzo SM-C18 column 3 mm, 3.0 assessed in triplicates. Absorbance at 490 nm was monitored 150 mm (Imtakt Corp) maintained at 10C. The chromatographic with a FLUOstar Omega microplate reader (BMG Labtech gradient started at 100% mobile phase A (100 mmol/L formic GmbH). For D-glucose measurements, 2 mL of cell growth acid in H2O) with a 2-minute hold followed with a 6-minute media were collected from cell plates after 6- to 24-hour gradient to 80% B (200 mmol/L ammonium formate in 30% incubations, spun down to remove cells, and analyzed with acetonitrile, pH 8) at a flow rate of 0.4 mL/min. This was followed the NOVA Bioprofile 400 (Nova Biomedial Corporation). by a 5-minute hold time at 100% mobile phase B and a sub- D-glucose uptake was corrected for cell counts (million cells) sequent re-equilibration time (6 minutes) before next injec- measured with a TC10-automated counting device (Bio-Rad). tion. Relative concentrations were determined from external

www.aacrjournals.org Cancer Res; 78(9) May 1, 2018 2193

Houles et al.

calibration curves prepared in water. Data were analyzed using pho-S483 antibody (Cell Signaling Technology; #13064) for 15 MassHunter Quant (Agilent Technologies). LC-MS grade water, minutes at room temperature. Detection of specific signal was acetonitrile, formic acid, and ammonium formate were purchased acquired by using Bond Polymer DAB Refine kit (#DS9800; Leica from Fisher, and authentic metabolite standards were purchase Biosystems) according to the provider's recommendations. Slides from Sigma-Aldrich Co. were counterstained automatically with hematoxylin included in the Polymer DAB Kit. Stained slides were scanned using the Stable isotope tracer analyses Hamamatsu's NanoZoomer Digital Pathology system 2HT. Vir- For assessment of cellular glucose metabolism activity, cells tual slides were then imported in Visiopharm Integrator System growing in 6-well plates at 70% confluence and serum-starved (3.4.1.0). Scanned tissue microarray was processed by using the overnight were treated with vehicle control (DMSO 0.1% v/v) or array imager module. Then each core was classified using Visio- with specific inhibitors for 6 hours. Cells were then equilibrated morph's K-Means clustering method. CMN biomarker scoring for 1 to 2 hours in tracing media without tracer (with treat- algorithm was applied. VS ¼ (MI Area)LOW þ (MI Area)MED þ ments), followed by media change with tracing media containing (MI Area)STRONG/Total Area of ROI. VS (Global visiomorph 13 25 mmol/L uniformly labeled C6-D-glucose ( 99 atom %, score) is range between 0 and þ255, MI (average intensity). #389374, Sigma-Aldrich) for the indicated times (with treatments). Cellular metabolite extracts were prepared and analyzed as de- Statistical analysis scribed in the GC-MS analyses section above, with the exception Statistical analyses were performed using a two-sample that single ion monitoring was performed to ensure optimal unequal-variance Student t test. Data are presented as the mean fi þ ... þ quanti cation of isotopomers (m 1, ,m n), where m is the SEM, and P values < 0.05 were considered to be statistically m/z ratio of analyzed metabolite fragments, and n is the maximal significant. Data are representative of results from at least three number of labeled carbons associated with these fragments. In independent experiments. order to remove abundances of naturally occurring isotopes (2H, 13C, 14C, 15N, etc.) and to determine the specific contribution 13 of C6-D-glucose to the isotopomer pools, mass distributions Results vectors were multiplied to in-house–prepared correction matrices, RSK regulates glucose metabolism in BRAF-mutated 13 as fully described previously (41). Typical C6-D-glucose tracing melanoma will generate enrichments in mþ3 lactate (glycolysis end-product), RSK was previously shown to be constitutively activated in and mþ2 forward citric acid cycle intermediates (due to oxidative BRAF-mutated melanoma (17), and to be required for melanoma decarboxylation of mþ3 pyruvate, generating mþ2acetyl-CoA cell proliferation (18). In order to dissect its contributions on the þ and m 1CO2). As these analyses only provide information on regulation of glycolysis, we examined the effects of specific the proportional labeling of metabolites pools, specificisotopomer enzyme inhibitors on the metabolism in BRAF-mutated melano- 13 enrichments were further multiplied by metabolite steady-state ma cell. For this, we performed metabolic flux analysis using C6- values obtained from separate unlabeled cell plates, thus revealing glucose as tracer in response to pharmacologic inhibition of RSK 13 – the relative amounts of C6-D-glucose labeled isotopomers, as (Fig. 1A). Notably, we found that RSK inhibition induced a 13 described previously (40). decrease in glucose flux as assayed through C6-glucose incorporation into lactate [m þ 3] metabolite pool levels (Fig. 1B). 13 RNA extraction and quantitative real-time PCR Interestingly, we also found decreases in C6-glucose incorpo- Total RNA from A375, WM164, was extracted by using the ration into tricarboxylic acid (TCA) cycle metabolites, including RNeasy mini Kit (Qiagen). Total RNA was reverse transcribed by citrate [m þ 2], a-ketoglutarate [m þ 2], and malate [m þ 2] using a cDNA reverse transcription kit (Applied Biosystems) (Fig. 1B). Assessment of amino acid levels under similar condi- according to the manufacturer's instructions. Gene expression tions did not reveal significant changes in response to RSK levels of the endogenous controls HPRT and ACTB were deter- inhibition (Supplementary Fig. S1). Using GC-MS, we also per- mined by using prevalidated TaqMan gene expression assays formed steady-state quantifications of metabolites in both A375 (Applied Biosystems). Expression levels of queried and control and WM164 cells subjected to RSK inhibitors. These experiments genes were determined by using assays designed with the Uni- revealed significant decreases in lactate, citrate, a-ketoglutarate, versal Probe Library from Roche. PCRs were performed on an ABI succinate, and malate in both cell types (Fig. 1C and D). In Real Time 7900HT cycler and analyzed with SDS 2.2 software. parallel, we performed differential analysis of cell media from Quantitative real-time PCR (qRT-PCR) was performed with Taq- A375 and WM164 cell cultures subjected to RSK inhibitors. Man gene expression master mix (Applied Biosystems), and all Lactate release was decreased by RSK inhibition over 6 hours in samples were tested in duplicate. mRNA levels of queried genes both BRAF-mutated melanoma cell lines (Fig. 1E). To confirm the were normalized to the averaged levels of both HPRT and ACTB. role of RSK, we depleted RSK1 and RSK2 in melanoma cells (A375 and WM164) using lentiviral shRNA constructs and found similar Tissue microarray reductions in lactate release (Fig. 1F). We also assessed glucose Immunohistochemical staining against PFKFB2 phospho-S483 consumption in melanoma cells treated with RSK inhibitors and was carried out on paraffin-embedded formalin-fixed samples found that it was reduced upon inhibition of RSK for 6 hours (Fig. using the automated Bond RX staining platform from Leica 1G). We also measured glycolytic metabolites and found that RSK Biosystems. Sections were deparaffinized inside immunostainer. inhibition decreased intracellular fructose-1,6-bisphosphate (Fig. Antigen recovery was conducted using heat retrieval (Heat- 1H). Given that fructose-1,6-bisphosphate and fructose-2,6- Induced Epitope Retrieval) with ER1 (Leica Biosystems proprie- bisphosphate cannot be easily differentiated using our LC-MS tary Epitope Retrieval using a low pH buffer) for 20 minutes. settings, we cannot exclude the possibility that some fructose-2,6- Sections were then incubated with 150 mL of anti-PFKFB2 phos- bisphosphate might contribute to the observed abundance of

2194 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFKFB2 and Glycolysis

Figure 1. RSK regulates glycolysis in BRAF-mutated melanoma cells. A, Schematic representation of metabolites involved in glucose metabolism (glycolysis, TCA cycle). The black circles represent the carbons within metabolites that are measured during metabolic flux experiments (12C, black isotope; 13C, red 13 isotope from C6-glucose). B, A375 cells were treated with LJH685 (10 mmol/L) for 6 hours, and glucose catabolism was assessed by quantification of 13 13 C atoms incorporated from C6-glucose to lactate [m þ 3], citrate [m þ 2], a-ketoglutarate [m þ 2], and malate [m þ 2] metabolites. A375 (C)and WM164 (D) cells were treated with LJH685 (10 mmol/L), and quantification of metabolites was assessed by GC-MS after 6 hours. E, A375 and WM164 cells were treated with LJH685 (10 mmol/L) for 6 hours, and lactate release was quantified by colorimetric assay. F, Lactate release was quantified by colorimetric assay in A375 and WM164 cell lines expressing a control shRNA (shNT) or shRNAs against RSK1/2. G, A375 and WM164 cells were treated with LJH685 (10 mmol/L) for 6 hours, and glucose uptake was quantified using NOVA BioProfile analyzer. H, A375 cells were treated with LJH685 (10 mmol/L) for 6 hours, and fructose1,6-biphosphate was quantified using LC-MS. Statistically significant changes compared with the control condition are indicated by asterisks (#, P < 0.1; , P < 0.05; , P < 0.01; , P < 0.001 by unpaired Student t test).

www.aacrjournals.org Cancer Res; 78(9) May 1, 2018 2195

Houles et al.

fructose-1,6-bisphosphate in cellular extracts. We confirmed that ylation targets of RSK involved in metabolism. Recent phos- RSK inhibition does not reduce the expression of the main phoproteomic screens have identified the glycolytic enzyme glycolytic genes (Supplementary Fig. S2), suggesting that RSK PFKFB2 as being potentially regulated by the RAS/MAPK path- may be directly regulating a glycolytic enzymes. Together, these way (33, 34). Two phosphorylation sites in PFKFB2 were results indicate that RSK is necessary to maintain glycolytic activity shown to be modulated by MEK1/2 and ERK1/2 inhibitors in BRAF-mutated melanoma cells, indicating its potential partic- (Ser466 and Ser483), which are conserved among vertebrate ipation in metabolic rewiring induced by oncogenic BRAF. species (Fig. 2A). To specifically verify this, we used phospho- specific antibodies directed against Ser466 and Ser483 of Identification of PFKFB2 as a substrate of the RAS/MAPK PFKFB2, as well as phospho-motif antibodies targeted against pathway the sequence conserved around both residues (Arg/Lys-X- To understand the precise mechanisms by which RSK reg- X-pSer/Thr, where X is any aa). HEK293 cells transfected with ulates glucose anabolism, we investigated potential phosphor- Myc-tagged human PFKFB2 were serum-starved overnight, and

Figure 2. The RAS/MAPK pathway promotes the phosphorylation of PFKFB2 on Ser466 and Ser483. A, Amino acid alignment of PFKFB2 from several vertebrate species showing the relative conservation of two phosphorylation sites within its regulatory domain (RD). B, HEK293 cells were transfected with Myc-tagged PFKFB2 WT or unphosphorylatable mutants (S466A, S483A, or S466/483A), serum-starved overnight, and stimulated with PMA (100 ng/mL) for 30 minutes. Immunoprecipitated Myc-PFKFB2 was assayed for phosphorylation by immunoblotting with a phospho-motif antibody that recognizes the RXXpS/T consensus motifs, as well as two anti-PFKFB2 phospho-speciﬁc antibodies targeted against the Ser466 and Ser483. C, Same as B,exceptthatcells were treated with PD184352 (10 mmol/L), PI-103 (1 mmol/L), Ku-0063794 (10 mmol/L), or rapamycin (25 nmol/L) for 30 minutes prior to PMA stimulation. D, Immunoblotting of HEK293 cells transfected with increasing concentrations of MEK-DD, an activated form of MEK1, and serum-starved prior to being harvested for SDS-PAGE and immunoblotting.

2196 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFKFB2 and Glycolysis

Figure 3. RSK phosphorylates PFKFB2 in vitro on Ser466 and Ser483. A, Recombinant-activated RSK1 was incubated with recombinant PFKFB2 WT or the unphosphorylatable mutant S466/483A (SA/SA) in a kinase reaction with [g-32P]ATP. The resulting samples were subjected to SDS–PAGE, and the dried Coomassie-stained gel was autoradiographed. B, Same as in A,exceptthatkinaseassayswereperformedinthepresenceofBI-D1870(10mmol/L) or LJH685 (10 mmol/L). C, Same as in B, except that samples were immunoblotted with speciﬁc PFKFB2 phospho-Ser466, phospho-Ser483, and GST antibodies after 0, 15, and 60 minutes of kinase reaction.

stimulated with PMA to activate the RAS/MAPK pathway, as RSK directly phosphorylates PFKFB2 at Ser466 and Ser483 shown by ERK1/2 phosphorylation(Fig.2B).Immunoprecipi- RSK is a basophilic protein kinase that operates down- tated PFKFB2 was then analyzed for phosphorylation by immu- stream of the RAS/MAPK pathway (16). To test its potential noblotting using the anti-pS466, anti-pS483, and anti-RXXpS/T involvement in the regulation of PFKFB2 phosphorylation, we antibodies. With this approach, we found that activation of performed in vitro kinase assays with purified proteins and the RAS/MAPK pathway correlated with the phosphorylation g[32P]ATP. Recombinant active RSK1 purified from Sf9 insect of PFKFB2 on Ser466 and Ser483 (Fig. 2B). The use of PFKFB2 cells was incubated in a kinase reaction buffer with bacterially mutants at either Ser466 (S466A) or Ser483 (S483A) confirmed purified GST, GST-tagged PFKFB2 wild-type (WT), or the dou- the specificity of the antibodies. The phospho-motif anti- ble S466/483A mutant (SA/SA). We found that activated body revealed that all PMA-induced immunoreactivity was lost RSK1 robustly increased [32P] label incorporation in purified when both Ser466 and Ser483 were simultaneously mutated GST-PFKFB2 WT, but not in GST alone or the GST-PFKFB2 (S466/483A), indicating that these residues are the predomi- S466/483A mutant (Fig. 3A). The phosphotransferase activity nant RXXpS/T phosphorylation sites in PFKFB2. of RSK1 was found to be necessary for this effect, as two The involvement of the RAS/MAPK pathway was further ATP-competitive RSK inhibitors (BI-D1870, LJH685) comp- confirmed using pharmacologic inhibitors, and by monitoring letely abrogated PFKFB2 phosphorylation (Fig. 3B). We also endogenous PFKFB2 phosphorylation using the anti-pS483 performed nonradioactive assays using phospho-specificanti- antibody. Using both HEK293 and HeLa cells, we found that bodies to measure RSK-mediated phosphorylation of PFKFB2. MEK1/2 inhibition using PD184352 decreased Ser483 phos- These experiments confirmed that RSK1 directly targets both phorylation induced by PMA stimulation (Fig. 2C), suggesting Ser466 and Ser483 for phosphorylation (Fig. 3C). Together, that a MEK1/2-dependent protein kinase regulates PFKFB2 these data are consistent with the idea that RSK is the likely phosphorylation. Notably, we found that mTOR inhibitors MEK1/2-dependent protein kinase phosphorylating PFKFB2 at (Ku-0063794, rapamycin) or a dual PI3K/mTOR inhibitor Ser466 and Ser483. (PI-103) did not affect PFKFB2 phosphorylation, indicating that To test the potential involvement of RSK in cells, we used both the PI3K/mTOR pathway is not involved in PFKFB2 phosphor- gain- and loss-of-function approaches. Using pharmacologic ylation. The requirement for MEK1/2 signaling was further inhibitors that are selective for RSK (BI-D1870, LJH685), we confirmed by expressing a constitutively activated form of MEK1 confirmed that RSK activity is required for PFKFB2 phosphor- (MEK-DD; S212/218D), which was found to be sufficient to ylation at Ser466 and Ser483 in response to PMA stimulation promote PFKFB2 phosphorylation at Ser483 in serum-starved (Fig.4A).ThereductioninPFKFB2 phosphorylation was simi- cells (Fig. 2D). Expression of activated MEK1/2 did not mod- lar to that observed with the MEK1/2 inhibitor (PD184352), ulate endogenous PFKFB2 protein levels, suggesting that the suggesting that RSK is the predominant PFKFB2-kinase operat- RAS/MAPK pathway does not affect PFKFB2 expression and/or ing downstream of MEK1/2. Notably, we also found that degradation. Together, these findings demonstrate that agonists inhibition of MEK1/2 and RSK reduced PFKFB2 phosphoryla- of the RAS/MAPK pathway promote MEK1/2-dependent phos- tion at RXXpS/T consensus phosphorylation sites (Fig. 4A), phorylation of PFKFB2 on Ser466 and Ser483. indicating that Ser466 and Ser483 are the predominant RSK

www.aacrjournals.org Cancer Res; 78(9) May 1, 2018 2197

Houles et al.

Figure 4. RSK promotes the phosphorylation of PFKFB2 in cells. A, HEK293 cells were transfected with Myc-PFKFB2, serum- starved overnight, and treated with PD184352 (10 mmol/L), BI-D1870 (10 mmol/L), or LJH685 (10 mmol/L) for 30 minutes before PMA (50 ng/mL) stimulation. PFKFB2 was immunoprecipitated with anti-Myc antibodies, and immunoblotting was performed on immunoprecipitates using indicated antibodies. B, HEK293 and HeLa cells were serum-starved and treated with BI-D1870 (10 mmol/L) or LJH685 (10 mmol/L) for 30 minutes before PMA stimulation (50 ng/mL). Endogenous PFKFB2 phosphorylation was assessed by immunoblotting with a PFKFB2 phospho-specific antibody targeted against the Ser483. C, HEK293 cells were transfected with RSK1 WT or kinase inactive (KD), serum-starved, and stimulated for 30 minutes with PMA (50 ng/mL) or 10 minutes with EGF (25 ng/mL). Endogenous PFKFB2 phosphorylation was assessed by immunoblotting with PFKFB2 phospho-specific antibodies against the Ser483. D, HEK293 cells were transfected with MEK-DD, serum-starved, and treated with BI- D1870 or LJH685 for 30 minutes. Endogenous PFKFB2 phosphorylation was assessed by immunoblotting with PFKFB2 phospho-specific antibodies against Ser483.

sites in PFKFB2. These results were confirmed with endo- consensus score, suggesting that RSK-mediated PFKFB2 phosphor- genous PFKFB2 in both HEK293 and HeLa cells, where RSK ylation promotes 14-3-3 binding to its C-terminal regulatory do- inhibitors (BI-D1870, LJH685) were found to reduce Ser483 main. To investigate this, we transfected HEK293 cells with WT phosphorylation (Fig. 4B). We also overexpressed WT or kinase PFKFB2 or the S466/483A mutant (SA/SA), as well as four different inactive (KD) forms of RSK1 to determine the role of RSK 14-3-3 isoforms (b, g, a,andz). Using coimmunoprecipitation activity in PFKFB2 phosphorylation. As expected, we found that assays, we found that WT PFKFB2 interacts with all four 14-3-3 RSK1 WT, but not RSK1 KD, increased PFKFB2 phosphoryla- isoforms, and that mutation of Ser466 and Ser483 almost comp- tion in response to PMA or EGF stimulation (Fig. 4C). Finally, letely prevented the interaction with the 14-3-3b, g,andz isoforms we found that RSK inhibitors prevented PFKFB2 phosphory- (Fig. 5A). We then assessed the relative role of Ser466 and Ser483 lation induced by constitutively activated MEK1 (MEK-DD; in 14-3-3 binding and found that both phosphorylation sites S212/218D; Fig. 4D), indicating that RSK is both required and contribute to 14-3-3b binding (Fig. 5B). To determine if RSK sufficient for PFKFB2 phosphorylation in response to agonists activity was required for these interactions, we treated cells with of the RAS/MAPK pathway. MEK1/2 (PD184532) or RSK (LJH685) inhibitors and found that both cell treatments disrupted 14-3-3b binding to PFKFB2 RSK promotes 14-3-3 binding to PFKFB2 and increases its (Fig. 5C). These data show that RSK promotes 14-3-3-binding to kinase activity PFKFB2 via the phosphorylation of Ser466 and Ser483. PFKFB2 was previously shown to interact with 14-3-3, which PFKFB2 phosphorylation was previously shown to promote correlated with increased PFKFB2 activity (42, 43). Using 14-3- the production of Fru-2,6-P2 (45, 46). To determine if RSK 3-Pred (44), we analyzed the amino acid sequence of human could similarly regulate PFKFB2 activity, we developed an in vitro PFKFB2 for potential 14-3-3 binding sites. Notably, we found that PFKFB2 kinase assay based on the release of ADP upon phos- Ser466 and Ser483 are the only two residues with a significant phorylation of fructose-6-phosphate (Fru-6-P) by PFKFB2

2198 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFKFB2 and Glycolysis

Figure 5. RSK regulates PFKFB2 kinase activity and binding to 14-3-3. A, HEK293 cells were cotransfected with Myc-PFKFB2 WT or S466/483A (AA) with 14-3-3b, g, s,orz isoforms. Cells were then serum- starved overnight and stimulated for 30 minutes with PMA (50 ng/mL). Cell lysates were immunoprecipitated with anti-Myc antibody and immunoblotting was performed. SE, short exposure; LE, long exposure. B, Same as in A, except that 14-3-3b was cotransfected with Myc-PFKFB2 WT, S466A, S483A, or S466/483A (SA/SA). C, HEK293 cells were cotransfected with Myc-PFKFB2 WT and 14-3-3b, serum-starved, and treated with PD184352 (10 mmol/L) or LJH685 (10 mmol/L) for 30 and 60 minutes before PMA stimulation. Cell lysates were immunoprecipitated with anti-Myc antibodies and immunoblotting was performed. D, HEK293 cells were transfected with Myc-PFKFB2 WT, serum-starved, and treated with LJH685 (10 mmol/L) or BI-D1870 (10 mmol/L) for 90 minutes before PMA stimulation. Cell lysates were immunoprecipitated with anti-Myc antibodies, and quantiﬁcation of PFKFB2 kinase activity toward Fru-6-phosphate was assessed by quantifying ADP release. Statistically signiﬁcant changes compared with the untreated condition (no PMA) are indicated by asterisks (, P < 0.05 by unpaired Student t test).

(Fig. 5D). Using this approach, we evaluated the kinase activity of HeLa cells, we found very high basal levels of endogenous PFKFB2 immunoprecipitated PFKFB2 isolated from HEK293 cells treated phosphorylation at Ser483 in serum-starved A375 and WM164 with PMA and/or RSK inhibitors (BI-D1870, LJH685). Our results cells (Fig. 6C). These cells also displayed high basal ERK1/2 show that PMA signiﬁcantly stimulated PFKFB2 kinase activity phosphorylation, consistent with the idea that the RAS/MAPK toward Fru-6-P, which was prevented by the inhibition of RSK pathway, and more particularly RSK, regulates PFKFB2 phos- activity (Fig. 5D). Taken together, these results indicate that RSK phorylation. Accordingly, we found that treatment of A375 and promotes phosphorylation-dependent binding of 14-3-3 to WM164 cells with MEK1/2 (PD184352) or RSK (BI-D1870, PFKFB2, which correlates with increased PFKFB2 kinase activity LJH685) inhibitors abrogated PFKFB2 phosphorylation at in response to agonists of the RAS/MAPK pathway. Ser483 (Fig. 6C and D), indicating that RSK is the predominant kinase regulating PFKFB2 in BRAF-mutant melanoma cells. As PFKFB2 is required for melanoma cell proliferation and is Akt was previously shown to phosphorylate PFKFB2 in response constitutively phosphorylated by RSK to amino acids in HeLa cell and rat cardiomyocytes (46), we used To determine the importance of PFKFB2 in melanoma cells Akt inhibitors to address its potential role in melanoma cells. and discriminate its effect from other PFK-2 isoforms, we used a Consistent with a predominant role for RSK, we found that knockdown approach with lentiviral shRNAs targeted against PFKFB2 phosphorylation was not affected by the Akt inhibitor three different target sequences in the PFKFB2 mRNA. Compared MK-2206 (Fig. 6E) or by PI3K/Akt pathway inhibition using with a control shRNA (shNT), we found that all three shRNAs (Ku-0063794, PI-103; Fig. 6C). Using endogenous PFKFB2 strongly reduced PFKFB2 expression in two different melanoma immunoprecipitations, we conﬁrmed that RSK promotes the cell lines harboring the BRAF V600E mutation (A375 and phosphorylation of both Ser466 and Ser483 (Fig. 6F). In addition, WM164; Fig. 6A and B). In both cell lines, depletion of PFKFB2 we analyzed PFKFB2 activity in melanoma cells treated with correlated with a severe reduction in cell proliferation (Fig. 6A RSK inhibitors and found that RSK activity is required for opti- and B), indicating that PFKFB2 plays an important and non- mal endogenous PFKFB2 activity in both A375 and WM164 redundant function in melanoma cells. melanoma cells (Fig. 6G and H). Together, these results indicate Next, we characterized PFKFB2 phosphorylation in melanoma that RSK constitutively phosphorylates PFKFB2 in melanoma cells using pharmacologic inhibitors. Unlike in HEK293 and cells and regulates its activity.

www.aacrjournals.org Cancer Res; 78(9) May 1, 2018 2199

Houles et al.

Figure 6. RSK phosphorylates and regulates PFKFB2 in BRAF-mutated melanoma cells. Proliferation assays were performed in A375 (A) and WM164 (B) cell lines knocked down for PFKFB2 using three different shRNA constructs. shRNA efficiency was assessed by immunoblotting with the anti-PFKFB2 antibody. C, A375 and WM164 cells were serum-starved overnight and treated with PD184352 (10 mmol/L), PI-103 (1 mmol/L), Ku-0063794 (10 mmol/L), or rapamycin (25 nmol/L) for 30 minutes, and endogenous PFKFB2 phosphorylation was assessed by immunoblotting. D, Same as in C, except that A375 and WM164 cells were treated with BI-D1870 (10 mmol/L) or LJH685 (10 mmol/L) for 30 minutes. E, Same as in D, except that A375 and WM164 cells were treated with BI-D1870 (10 mmol/L) and/or MK-2206 (1 mmol/L). F, WM164 cells were serum-starved overnight and treated with PD184352 or LJH685 for 30 minutes. Endogenous PFKFB2 was immunoprecipitated, and immunoblotting was performed on immunoprecipitates using indicated antibodies. A375 (G) and WM164 (H) cells were serum-starved overnight and treated with LJH685 for 90 minutes. Endogenous PFKFB2 was immunoprecipitated, and quantification of PFKFB2 kinase activity toward Fru-6-phosphate was assessed by quantifying ADP release. Endogenous PFKFB2 phosphorylation was assessed by immunoblotting. Statistically significant changes compared with control are indicated by asterisks (, P < 0.05; , P < 0.01 by unpaired Student t test).

2200 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFKFB2 and Glycolysis

RSK-mediated PFKFB2 phosphorylation is required for Fru-6-P, an important glycolytic metabolite (Fig. 5). Using BRAF- melanoma growth mutated melanoma cell lines, we found that RSK mediates the To determine if RSK is required for melanoma growth and constitutive phosphorylation of PFKFB2, which is a PFK-2 iso- proliferation, we depleted RSK1/2 in two melanoma cell lines form required for melanoma cell proliferation (Fig. 6). Our results and assessed their proliferation in vitro and growth as xenografts also show that an unphosphorylatable mutant of PFKFB2 in nude mice. Our results demonstrate that RSK1/2 are indeed decreases glycolytic activity in melanoma cells as well as tumor required for efficient melanoma cell proliferation in vitro growth in immunocompromised mice (Fig. 7), suggesting that (Fig. 7A and B). We also found that RSK1/2 depletion strongly RSK-mediated PFKFB2 phosphorylation plays a key role in BRAF- reduced the growth of WM164 xenografts when injected (3 mutated melanoma growth. 106 cells) subcutaneously in the flanks of nude mice (5 per The RSK family comprises four related isoforms (RSK1–4) groups;Fig.7C),indicatingthatRSKisrequiredformela- that have divergent functions (16). Although RSK1 and RSK2 noma growth. To investigate if RSK-mediated phosphorylation promote cancer cell growth, survival, and proliferation, RSK3 of PFKFB2 at Ser466 and Ser483 provides an advantage to and RSK4 were shown to possess functions akin to tumor BRAF-mutated melanoma cells, we generated A375 cell lines suppressors (47). This may explain the fact that RSK3 and RSK4 stably expressing PFKFB2 WT, or the S466/483A (AA) and are expressed to very low levels in most cells and tissues, S466/483D (DD) mutants. All three alleles were found to be whereas RSK1 and RSK2 are abundantly expressed in most expressed at similar levels, and we also confirmed that PFKFB2 tissues and cell lines, including melanoma cells (17, 18). Our WT was the only allele showing specific phosphorylation at results indicate that RSK promotes PFKFB2 phosphorylation Ser483 (Fig. 7D). Although we did not observe significant and thereby regulates glucose metabolism. These results are in differences in the proliferation rates of the different cell lines line with previous findings showing that RSK phosphorylates under standard tissue culture conditions (Fig. 7D), we found and regulates AS160 (Akt substrate of 160 kDa; ref. 48), a that cells expressing the S466/483A mutant of PFKFB2 had protein involved in GLUT4 (glucose transporter 4) transloca- reduced lactate release compared with the other stable cell lines tion and glucose uptake (49). AS160 phosphorylation was also (Fig. 7E). These results suggested that RSK-mediated PFKFB2 shown to promote 14-3-3 binding (48), suggesting that RSK and phosphorylation does not affect melanoma cell proliferation 14-3-3 could collaborate in the regulation of glucose uptake and under nutrient replete conditions, but that these phosphory- metabolism via AS160 and PFKFB2, respectively. Interestingly, lation events participate in enhancing glycolytic activity. RSK was also shown to negatively regulate IRS-1 (insulin To investigate the role of PFKFB2 phosphorylation in melano- receptor substrate-1) by promoting its phosphorylation on ma tumor growth, we used a xenotransplantation assay in nude Ser1101 (50). These results suggest that RSK is a negative mice. For this, A375 cells stably expressing PFKFB2 WT, or the regulator of insulin signaling and a potential mediator of different phosphorylation mutants (S466/483A, S466/483D), insulin resistance. Together, these data indicate that RSK plays were injected (2 106 cells) subcutaneously in the flanks of nude unexpected but important roles in glucose metabolism, and mice (10 per groups). We found that all cell lines rapidly formed future work using genetically engineered mouse models should tumors, reaching 0.5 cm3 within approximately 30 to 50 days of help reveal its exact contribution in both normal and patho- injection, with the exception of cells expressing the S466/483A physiologic settings. mutant of PFKFB2, which yielded significantly smaller tumors PFKFB2 was originally found to be expressed in the heart (Fig. 7F). Indeed, mice expressing the unphosphorylatable (45, 51), but evidence now shows that it is present in other mutant of PFKFB2 (AA) survived longer than any other groups tissues and cultured cells, including ovarian (52) and prostate (P < 0.01; Fig. 7G), consistent with the idea that phosphorylation (53) cancer cells. Unlike other PFK-2 isoforms, PFKFB2 was of Ser466 and Ser483 promotes tumor growth. We also investi- shown to display nearly equal kinase and phosphatase activities gated whether PFKFB2 is phosphorylated in human melanomas (31). This feature suggests that its simple upregulation is not and found by immunohistochemistry that Ser483 phosphoryla- sufficient to modulate Fru-2,6-P2 levels, and that additional tion was increased in melanomas compared with adjacent normal inputs are necessary to regulate its function (31). Certain skin (Fig. 7H). Together, these data indicate that RSK-mediated stimuli have been shown to increase PFKFB2 kinase activity phosphorylation of PFKFB2 promotes glycolysis and is necessary by promoting the phosphorylation of residues located in its for melanoma growth in vivo (Fig. 7I). regulatory domain, such as androgen (53) and amino acids (46). Phosphorylation of Ser466 and Ser483 at the C-terminal Discussion end of PFKFB2 was shown to activate its kinase activity by decreasing the Km for Fru-6-P and by increasing the Vmax Oncogene-driven metabolic rewiring is an adaptation to low without affecting its phosphatase activity (31). Mutagenesis nutrient and oxygen conditions in the tumor microenvironment. of Ser466 and Ser483 to glutamic acid, to mimic the effects of In melanoma, the RAS/MAPK pathway mediates part of these phosphorylation, indicated that Ser466 phosphorylation is fi metabolic changes through speci c genetic and epigenetic responsible for the increase in Vmax, whereas both Ser466 and changes (23). In this study, we show that the MAPK-activated Ser483 phosphorylations are necessary to decrease the Km for protein kinase RSK regulates glucose metabolism in melanoma Fru-6-P (54). These results indicate that the function of the cell lines with a BRAF V600E mutation (Fig. 1). We found that the C-terminal regulatory domain of PFKFB2 may be to inhibit its RAS/MAPK pathway (Fig. 2), and more precisely that RSK, pro- kinase activity, and thereby shift the balance toward its phos- motes PFKFB2 phosphorylation at Ser466 and Ser483 (Fig. 3) in phatase activity, which can be relieved through phosphoryla- multiple cell types (Fig. 4). Interestingly, phosphorylation of tion of Ser466 and Ser483. Our results show that RSK-mediated Ser466/Ser483, located in the regulatory domain of PFKFB2, phosphorylation of Ser466 and Ser483 promotes the interac- promotes 14-3-3 binding and increases its kinase activity toward tion of PFKFB2 with different 14-3-3 isoforms. Although the

www.aacrjournals.org Cancer Res; 78(9) May 1, 2018 2201

Houles et al.

Figure 7. RSK-mediated PFKFB2 phosphorylation promotes BRAF-mutated melanoma growth. Proliferation assays were performed in WM164 (A) and A375 (B) cells stably expressing a control shRNA (shNT) or shRNAs against RSK1/2. shRNA efficiency was assessed by immunoblotting with the anti-RSK1 and anti-RSK2 antibody. C, WM164 cells stably expressing a shNT or shRNAs against RSK1/2 were injected as subcutaneous xenografts in nude mice. Mice were monitored for tumor development, and graphs represent the growth rate of subcutaneous tumors. Values represent the average volume SEM of 10 tumors (from 5 mice). Endogenous RSK1 and RSK2 protein levels in injected cells were monitored by immunoblotting. D, Stable cell lines expressing PFKFB2 WT, PFKFB2 AA (S466/483A), or PFKFB2 DD (S466/483D) were generated in A375 parental cells. Proliferation assays were performed in A375 stable cell lines over 96 hours. The expression of Myc-PFKFB2 and its phosphorylation at Ser483 in each cell line was assessed by immunoblotting. E, Lactate release was quantified by colorimetric assay in A375 stable cell lines, 6 hours after media change. F, PFKFB2 stable cell lines were injected as subcutaneous xenografts in nude mice. Tumor volumes were measured after 29 days of growth (8 mice). G, Survival curves representing the proportion of surviving mice after subcutaneous injections of PFKFB2 stable cell lines. H, IHC analysis of PFKFB2 phospho-Ser483 in human melanoma tissue microarrays. A representative image of normal skin and melanoma tissues label showing increased labeling with phospho-S483 antibodies (right) and hematoxylin/phloxin safran (HPS, left). PFKFB2 phospho-Ser483 was quantified in normal skin (black) and lesion tissues (red) in the same IHC for each case. Values represent the quantification of PFKFB2 phospho-Ser483 (9 patients). I, Schematic representation of our working model. Oncogenic BRAF V600E activates RSK to phosphorylate and activate PFKFB2, which binds to 14-3-3 and promotes glycolysis. Statistically significant changes compared with the control condition are indicated by asterisks (, P < 0.01; , P < 0.0001 by unpaired Student t test).

2202 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFKFB2 and Glycolysis

exact molecular function of 14-3-3 remains unknown, one Authors' Contributions possibility is that its interaction with PFKFB2 relieves the Conception and design: T. Houles, J. St-Pierre, P.P. Roux inhibitory activity of the regulatory domain through steric Development of methodology: T. Houles, S.-P. Gravel, G. Lavoie hindrance and/or conformational changes. Although future Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): T. Houles, S.-P. Gravel, G. Lavoie, S. Shin, M. Savall, work will be aimed at studying the molecular role of 14-3-3 A. Meant, B. Grondin, L. Gaboury, S.-O. Yoon, J. St-Pierre in the regulation of PFKFB2 activity, this is likely to be chal- Analysis and interpretation of data (e.g., statistical analysis, biostatistics, lenged by the lack of speciﬁc and permeable 14-3-3 inhibitors, computational analysis): T. Houles, S.-P. Gravel, G. Lavoie, S. Shin, L. Gaboury, as well as the fact that different 14-3-3 isoforms regulate S.-O. Yoon, J. St-Pierre, P.P. Roux additional metabolic enzymes (55, 56). Writing, review, and/or revision of the manuscript: T. Houles, S.-P. Gravel, Metabolic reprogramming in melanoma occurs in response L. Gaboury, P.P. Roux Study supervision: J. St-Pierre, P.P. Roux to oncogenic stimuli and as an adaptation to stress associated with the tumor microenvironment. Although hypoxia reduces ﬂ Acknowledgments glucose ux into the TCA cycle, it also stimulates glycolysis by We thank all members of the laboratory for their insightful discussions and upregulating HIF1 target genes (27, 28). Our results show comments. We also thank Julie Hinsinger for technical support and important that, under normal culture conditions, expression of a phos- advice at the Histology platform of IRIC. phorylation mutant of PFKFB2 does not impair melanoma This work was supported by grants from the Canadian Institutes for cell proliferation. However, we found that these cells show Health Research (P.P. Roux and J. St-Pierre) and the Natural Sciences reduced lactate release and tumor growth in nude mice, and Engineering Research Council of Canada (P.P. Roux). P.P. Roux and J. St-Pierre are scholars of the Fonds de la recherche du Quebec - Sante suggesting that stresses associated with the tumor microenvi- (FRQS). Metabolite measurements and tracer analyses were performed at the ronment increase the dependence on PFKFB2 phosphoryla- Rosalind and Morris Goodman Cancer Research Centre Metabolomics tion and activation. These results raise the possibility that Core Facility supported by the Canada Foundation for Innovation, The melanoma cells may be more dependent on RSK in response Dr.JohnR.andClaraM.FraserMemorial Trust, the Terry Fox Foundation to certain types of stresses affecting glucose metabolism, but (TFF Oncometabolism Team Grant 1048 in partnership with the Foundation this remains to be formally tested. Nonetheless, our results du Cancer du Sein du Quebec), and McGill University. suggest that RSK may be a suitable target for the treatment The costs of publication of this article were defrayed in part by the payment of of melanoma, as well as other cancer types characterized by page charges. This article must therefore be hereby marked advertisement in similar metabolic rewiring. accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

Disclosure of Potential Conflicts of Interest Received July 21, 2017; revised August 3, 2017; accepted February 6, 2018; No potential conflicts of interest were disclosed. published first February 12, 2018.

References 1. Lo JA, Fisher DE. The melanoma revolution: from UV carcinogenesis to a 15. Cargnello M, Roux PP. Activation and function of the MAPKs and their new era in therapeutics. Science 2014;346:945–9. substrates, the MAPK-activated protein kinases. Microbiol Mol Biol Rev 2. Wan PT, Garnett MJ, Roe SM, Lee S, Niculescu-Duvaz D, Good VM, et al. 2011;75:50–83. Mechanism of activation of the RAF-ERK signaling pathway by oncogenic 16. Romeo Y, Zhang X, Roux PP. Regulation and function of the RSK family mutations of B-RAF. Cell 2004;116:855–67. of protein kinases. Biochem J 2012;441:553–69. 3. Davies H, Bignell GR, Cox C, Stephens P, Edkins S, Clegg S, et al. Mutations 17. Old WM, Shabb JB, Houel S, Wang H, Couts KL, Yen CY, et al. Functional of the BRAF gene in human cancer. Nature 2002;417:949–54. proteomics identifies targets of phosphorylation by B-Raf signaling in 4. Dhomen N, Marais R. New insight into BRAF mutations in cancer. melanoma. Mol Cell 2009;34:115–31. Curr Opin Genet Dev 2007;17:31–9. 18. Romeo Y, Moreau J, Zindy P-J, Saba-El-Leil M, Lavoie G, Dandachi F, et al. 5. Gray-Schopfer VC, da Rocha Dias S, Marais R. The role of B-RAF in RSK regulates activated BRAF signalling to mTORC1 and promotes melanoma. Cancer Metastasis Rev 2005;24:165–83. melanoma growth. Oncogene 2013;32:2917–26. 6. Miller AJ, Mihm MC Jr. Melanoma. N Engl J Med 2006;355:51–65. 19. Hay N. Reprogramming glucose metabolism in cancer: can it be exploited 7. Scheier B, Amaria R, Lewis K, Gonzalez R. Novel therapies in melanoma. for cancer therapy? Nat Rev Cancer 2016;16:635–49. Immunotherapy 2011;3:1461–9. 20. Hanahan D, Weinberg RA. Hallmarks of cancer: the next generation. Cell 8. Flaherty KT, Puzanov I, Kim KB, Ribas A, McArthur GA, Sosman JA, et al. 2011;144:646–74. Inhibition of mutated, activated BRAF in metastatic melanoma. N Engl J 21. Vander Heiden MG, Cantley LC, Thompson CB. Understanding the Med 2010;363:809–19. Warburg effect: the metabolic requirements of cell proliferation. Science 9. Chapman PB, Hauschild A, Robert C, Haanen JB, Ascierto P, Larkin J, et al. 2009;324:1029–33. Improved survival with vemurafenib in melanoma with BRAF V600E 22. Levine AJ, Puzio-Kuter AM. The control of the metabolic switch in cancers mutation. N Engl J Med 2011;364:2507–16. by oncogenes and tumor suppressor genes. Science 2010;330:1340–4. 10. Sosman JA, Kim KB, Schuchter L, Gonzalez R, Pavlick AC, Weber JS, et al. 23. Ratnikov BI, Scott DA, Osterman AL, Smith JW, Ronai ZA. Metabolic Survival in BRAF V600-mutant advanced melanoma treated with vemur- rewiring in melanoma. Oncogene 2017;36:147–57. afenib. N Engl J Med 2012;366:707–14. 24. Haq R, Shoag J, Andreu-Perez P, Yokoyama S, Edelman H, Rowe GC, et al. 11. Lito P, Rosen N, Solit DB. Tumor adaptation and resistance to RAF Oncogenic BRAF regulates oxidative metabolism via PGC1alpha and inhibitors. Nat Med 2013;19:1401–9. MITF. Cancer Cell 2013;23:302–15. 12. Murphy LO, Blenis J. MAPK signal specificity: the right place at the right 25. Vazquez F, Lim JH, Chim H, Bhalla K, Girnun G, Pierce K, et al. PGC1alpha time. Trends Biochem Sci 2006;31:268–75. expression defines a subset of human melanoma tumors with increased 13. Ballif BA, Blenis J. Molecular mechanisms mediating mammalian mitogen- mitochondrial capacity and resistance to oxidative stress. Cancer Cell activated protein kinase (MAPK) kinase (MEK)-MAPK cell survival signals. 2013;23:287–301. Cell Growth Differ 2001;12:397–408. 26. St-Pierre J, Lin J, Krauss S, Tarr PT, Yang R, Newgard CB, et al. Bioenergetic 14. Roux PP, Blenis J. ERK and p38 MAPK-activated protein kinases: a family analysis of peroxisome proliferator-activated receptor gamma coactivators of protein kinases with diverse biological functions. Microbiol Mol Biol 1alpha and 1beta (PGC-1alpha and PGC-1beta) in muscle cells. J Biol Rev 2004;68:320–44. Chem 2003;278:26597–603.

www.aacrjournals.org Cancer Res; 78(9) May 1, 2018 2203

Houles et al.

27. Denko NC. Hypoxia, HIF1 and glucose metabolism in the solid tumour. 43. PozueloRubioM,PeggieM,WongBH,MorriceN,MacKintoshC.14- Nat Rev Cancer 2008;8:705–13. 3-3s regulate fructose-2,6-bisphosphate levels by binding to PKB-phos- 28. Kumar SM, Yu H, Edwards R, Chen L, Kazianis S, Brafford P, et al. Mutant phorylated cardiac fructose-2,6-bisphosphate kinase/phosphatase. V600E BRAF increases hypoxia inducible factor-1alpha expression in EMBO J 2003;22:3514–23. melanoma. Cancer Res 2007;67:3177–84. 44. Madeira F, Tinti M, Murugesan G, Berrett E, Stafford M, Toth R, et al. 14-3-3- 29. Semenza GL, Roth PH, Fang HM, Wang GL. Transcriptional regulation of Pred: improved methods to predict 14-3-3-binding phosphopeptides. genes encoding glycolytic enzymes by hypoxia-inducible factor 1. J Biol Bioinformatics 2015;31:2276–83. Chem 1994;269:23757–63. 45.MarsinAS,BertrandL,RiderMH,DeprezJ,BeauloyeC,VincentMF, 30. Mor I, Cheung EC, Vousden KH. Control of glycolysis through regulation et al. Phosphorylation and activation of heart PFK-2 by AMPK has a of PFK1: old friends and recent additions. Cold Spring Harb Symp Quant role in the stimulation of glycolysis during ischaemia. Curr Biol 2000; Biol 2011;76:211–6. 10:1247–55. 31. Rider MH, Bertrand L, Vertommen D, Michels PA, Rousseau GG, Hue L. 46. Novellasdemunt L, Tato I, Navarro-Sabate A, Ruiz-Meana M, Mendez- 6-phosphofructo-2-kinase/fructose-2,6-bisphosphatase: head-to-head Lucas A, Perales JC, et al. Akt-dependent activation of the heart 6-phos- with a bifunctional enzyme that controls glycolysis. Biochem J 2004; phofructo-2-kinase/fructose-2,6-bisphosphatase (PFKFB2) isoenzyme by 381:561–79. amino acids. J Biol Chem 2013;288:10640–51. 32. Minchenko O, Opentanova I, Caro J. Hypoxic regulation of the 6-phos- 47. Houles T, Roux PP. Defining the role of the RSK isoforms in cancer. Semin phofructo-2-kinase/fructose-2,6-bisphosphatase gene family (PFKFB-1-4) Cancer Biol 2017;48:53–61. expression in vivo. FEBS Lett 2003;554:264–70. 48. Geraghty KM, Chen S, Harthill JE, Ibrahim AF, Toth R, Morrice NA, et al. 33. Galan JA, Geraghty KM, Lavoie G, Kanshin E, Tcherkezian J, Calabrese V, Regulation of multisite phosphorylation and 14-3-3 binding of et al. Phosphoproteomic analysis identifies the tumor suppressor AS160 in response to IGF-1, EGF, PMA and AICAR. Biochem J 2007;407: PDCD4 as a RSK substrate negatively regulated by 14-3-3. Proc Natl Acad 231–41. Sci U S A 2014;111:E2918–27. 49. Sano H, Kane S, Sano E, Miinea CP, Asara JM, Lane WS, et al. Insulin- 34. Gnad F, Doll S, Song K, Stokes MP, Moffat J, Liu B, et al. Phosphoproteome stimulated phosphorylation of a Rab GTPase-activating protein regulates analysis of the MAPK pathway reveals previously undetected feedback GLUT4 translocation. J Biol Chem 2003;278:14599–602. mechanisms. Proteomics 2016;16:1998–2004. 50. Smadja-Lamere N, Shum M, Deleris P, Roux PP, Abe J, Marette A. Insulin 35. Carriere A, Cargnello M, Julien LA, Gao H, Bonneil E, Thibault P, et al. activates RSK (p90 ribosomal S6 kinase) to trigger a new negative feedback Oncogenic MAPK signaling stimulates mTORC1 activity by promoting loop that regulates insulin signaling for glucose metabolism. J Biol Chem RSK-mediated raptor phosphorylation. Curr Biol 2008;18:1269–77. 2013;288:31165–76. 36. Yaffe MB, Rittinger K, Volinia S, Caron PR, Aitken A, Leffers H, et al. The 51. Heine-Suner D, Diaz-Guillen MA, Lange AJ, Rodriguez de Cordoba S. structural basis for 14-3-3:phosphopeptide binding specificity. Cell Sequence and structure of the human 6-phosphofructo-2-kinase/fructose- 1997;91:961–71. 2,6-bisphosphatase heart isoform gene (PFKFB2). Eur J Biochem 1998; 37. Roux PP, Ballif BA, Anjum R, Gygi SP, Blenis J. Tumor-promoting phorbol 254:103–10. esters and activated Ras inactivate the tuberous sclerosis tumor sup- 52. Zhao L, Ji G, Le X, Wang C, Xu L, Feng M, et al. Long noncoding RNA pressor complex via p90 ribosomal S6 kinase. Proc Natl Acad Sci U S A LINC00092 acts in cancer-associated fibroblasts to drive glycolysis 2004;101:13489–94. and progression of ovarian cancer. Cancer Res 2017;77:1369–82. 38. Zhang X, Lavoie G, Meant A, Aubert L, Cargnello M, Haman A, et al. 53. Moon JS, Jin WJ, Kwak JH, Kim HJ, Yun MJ, Kim JW, et al. Androgen Extracellular signal-regulated kinases 1 and 2 phosphorylate Gab2 to stimulates glycolysis for de novo lipid synthesis by increasing the promote a negative-feedback loop that attenuates phosphoinositide activities of hexokinase 2 and 6-phosphofructo-2-kinase/fructose- 3-kinase/Akt signaling. Mol Cell Biol 2017;37. 2,6-bisphosphatase 2 in prostate cancer cells. Biochem J 2011;433: 39. Morita M, Gravel SP, Chenard V, Sikstrom K, Zheng L, Alain T, et al. 225–33. mTORC1 controls mitochondrial activity and biogenesis through 4E-BP- 54. Bertrand L, Alessi DR, Deprez J, Deak M, Viaene E, Rider MH, et al. Heart dependent translational regulation. Cell Metab 2013;18:698–711. 6-phosphofructo-2-kinase activation by insulin results from Ser-466 40. Gravel SP, Avizonis D, St-Pierre J. Metabolomics analyses of cancer cells in and Ser-483 phosphorylation and requires 3-phosphoinositide- controlled microenvironments. Methods Mol Biol 2016;1458:273–90. dependent kinase-1, but not protein kinase B. J Biol Chem 1999;274: 41. Nanchen A, Fuhrer T, Sauer U. Determination of metabolic flux ratios from 30927–33. 13C-experiments and gas chromatography-mass spectrometry data: pro- 55. Pozuelo Rubio M, Geraghty KM, Wong BH, Wood NT, Campbell DG, tocol and principles. Methods Mol Biol 2007;358:177–97. Morrice N, et al. 14-3-3-affinity purification of over 200 human phospho- 42. Dubois F, Vandermoere F, Gernez A, Murphy J, Toth R, Chen S, et al. proteins reveals new links to regulation of cellular metabolism, prolifer- Differential 14-3-3 affinity capture reveals new downstream targets of ation and trafficking. Biochem J 2004;379:395–408. phosphatidylinositol 3-kinase signaling. Mol Cell Proteomics 2009;8: 56. Kleppe R, Martinez A, Doskeland SO, Haavik J. The 14-3-3 proteins in 2487–99. regulation of cellular metabolism. Semin Cell Dev Biol 2011;22:713–9.

2204 Cancer Res; 78(9) May 1, 2018 Cancer Research

RSK Regulates PFK-2 Activity to Promote Metabolic Rewiring in Melanoma

Thibault Houles, Simon-Pierre Gravel, Geneviève Lavoie, et al.

Cancer Res 2018;78:2191-2204. Published OnlineFirst February 12, 2018.

Updated version Access the most recent version of this article at: doi:10.1158/0008-5472.CAN-17-2215

Supplementary Access the most recent supplemental material at: Material http://cancerres.aacrjournals.org/content/suppl/2018/02/10/0008-5472.CAN-17-2215.DC1

Cited articles This article cites 55 articles, 23 of which you can access for free at: http://cancerres.aacrjournals.org/content/78/9/2191.full#ref-list-1

Citing articles This article has been cited by 4 HighWire-hosted articles. Access the articles at: http://cancerres.aacrjournals.org/content/78/9/2191.full#related-urls

E-mail alerts Sign up to receive free email-alerts related to this article or journal.

Reprints and To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at Subscriptions [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://cancerres.aacrjournals.org/content/78/9/2191. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.