A Fatty Acid Oxidation-Dependent Metabolic Shift Regulates the Adaptation of BRAF-Mutated Melanoma to MAPK Inhibitors

Author Manuscript Published OnlineFirst on August 2, 2019; DOI: 10.1158/1078-0432.CCR-19-0253 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

A fatty acid oxidation-dependent metabolic shift regulates the adaptation of BRAF-

mutated melanoma to MAPK inhibitors

Andrea Aloia1,*, Daniela Müllhaupt1, Christophe D. Chabbert1, £, Tanja Eberhart1, Stefanie Flückiger-

Mangual1, Ana Vukolic1, Ossia Eichhoff2, Anja Irmisch2, Leila T. Alexander3, §, Ernesto Scibona4,

Dennie T. Frederick5, Benchun Miao5, Tian Tian6, Chaoran Cheng6, Lawrence N. Kwong7, Zhi Wei6,

Ryan J. Sullivan5, Genevieve M. Boland8, Meenhard Herlyn9, Keith T. Flaherty5, Nicola Zamboni3,

Reinhard Dummer2, Gao Zhang9, #, Mitchell P. Levesque2, $, Wilhelm Krek1, †, Werner J. Kovacs1, 10, $, *

1ETH Zurich, Institute of Molecular Health Sciences, 8093 Zurich, Switzerland

2University Hospital Zurich, Department of Dermatology, 8091 Zurich, Switzerland

3ETH Zurich, Institute of Molecular System Biology, 8093 Zurich, Switzerland

4ETH Zurich, Institute of Chemical and Bioengineering, 8093 Zurich, Switzerland

5Massachusetts General Hospital Cancer Center, Boston, MA 02114, USA

6Department of Computer Science, New Jersey Institute of Technology, Newark, NJ 07102, USA

7Department of Translational Molecular Pathology, The University of Texas MD Anderson Cancer

Center, Houston, TX 770303, USA

8Department of Surgery, Massachusetts General Hospital Cancer Center, Boston, MA 02114, USA

9Molecular and Cellular Oncogenesis Program and Melanoma Research Center, The Wistar Institute,

Philadelphia, PA 19104, USA

10Lead Contact

$Co-senior authors

£Current address: Roche Innovation Center Zurich, 8952 Schlieren, Switzerland

§Current address: SIB Swiss Institute of Bioinformatics, Personalized Health Informatics, 4056 Basel,

Switzerland

#Current address: Department of Neurosurgery, The Preston Robert Tisch Brain Tumor Center, and

Department of Pathology, Duke University Medical Center, Durham, NC 27710, USA

†Deceased 29. August 2018

*Correspondence to: Werner Kovacs, Institute of Molecular Health Sciences, ETH Zurich, Otto-Stern-

Weg 7, HPL H16.1, CH-8093 Zurich, Switzerland. Tel.: +41 44 633 3084. E-mail: [email protected]

*Correspondence to: Andrea Aloia, Institute of Molecular Health Sciences, ETH Zurich, Otto-Stern-

Weg 7, HPL H23.2, CH-8093 Zurich, Switzerland. Tel.: +41 44 633 3360. E-mail: [email protected]

Running title

MAPKi induce fatty acid oxidation in BRAFV600E melanomas

Keywords

BRAF-mutated melanoma, adaptive drug resistance, fatty acid oxidation, CD36, MAPK inhibitors

Abstract

Purpose: Treatment of BRAFV600E-mutant melanomas with mitogen-activated protein kinase inhibitors

(MAPKi) results in significant tumor regression, but acquired resistance is pervasive. To understand non-mutational mechanisms underlying the adaptation to MAPKi and to identify novel vulnerabilities of melanomas treated with MAPKi, we focused on the initial response phase during treatment with

MAPKi.

Experimental Design: By screening proteins expressed on the cell surface of melanoma cells, we identified the fatty acid transporter CD36 as the most consistently upregulated protein upon short-term treatment with MAPKi. We further investigated the effects of MAPKi on fatty acid metabolism using in vitro and in vivo models and analyzing patients’ pre- and on-treatment tumor specimens.

Results: Melanoma cells treated with MAPKi displayed increased levels of CD36 and of peroxisome proliferator-activated receptor  (PPAR)-mediated and carnitine palmitoyltransferase 1A (CPT1A)- dependent fatty acid oxidation (FAO). While CD36 is a useful marker of melanoma cells during adaptation and drug-tolerant phases, the upregulation of CD36 is not functionally involved in FAO changes that characterize MAPKi-treated cells. Increased FAO is required for BRAFV600E-mutant melanoma cells to survive under the MAPKi-induced metabolic stress prior to acquiring drug resistance. The upfront and concomitant inhibition of FAO, glycolysis and MAPK synergistically inhibits tumor cell growth in vitro and in vivo.

Conclusions: Thus, we identified a clinically relevant therapeutic approach that has the potential to improve initial responses and to delay acquired drug resistance of BRAFV600E-mutant melanoma.

Translational Relevance

Transiently resistant cells adapting to mitogen-activated protein kinase inhibitors (MAPKi) are responsible for acquired drug resistance. We identified CD36 as a marker of transiently resistant and

MAPKi-tolerant BRAFV600E melanoma cells. In addition, we describe an early metabolic reprogramming mechanism induced by MAPKi through increased FAO which is required to survive

MAPKi-induced metabolic stress prior to acquiring drug resistance. FAO inhibitors increase the glycolytic flux in untreated and MAPKi-treated melanoma cells as a compensatory mechanism to FAO inhibition. To exploit melanoma cells’ metabolic plasticity for therapeutic intervention, we propose a triple combination of MAPK, FAO and glycolytic inhibitors as a novel treatment option to prevent drug resistance in BRAF-mutated melanomas.

DECLARATION OF INTEREST

CDC is a full-time employee of Roche AG and a shareholder in AstraZeneca.

Other authors declare no competing interests.

Introduction

Malignant melanoma is an aggressive skin tumor with poor prognosis due to metastasis and therapeutic resistance (1). Approximately 50% of melanoma patients carry gain-of-function mutations in the BRAF gene (2). The most frequent mutation is the substitution of valine at position 600 by glutamic acid (BRAFV600E) resulting in constitutive activation of the mitogen-activated protein kinase

(MAPK) signal transduction pathway (3). BRAF plays a pivotal role in the MAPK pathway by promoting cell division, proliferation, and survival (4). Since BRAF is a critical mediator of melanomagenesis, BRAFV600E inhibitors (BRAFi) and MEK/ERK kinase inhibitors (MEKi) were developed to target key components of the MAPK signaling pathway and induce cell death (5,6).

BRAFi alone or in combination with MEKi have shown clinical efficacy in BRAF-mutated metastatic melanoma patients (7,8). However, drug resistance inevitably develops and most patients relapse within a few months with limited benefits. Drug resistance represents a common complication of targeted therapies hampering long-term treatment success (9). To date, numerous mechanisms of acquired drug resistance have already been identified in metastatic melanoma patients (10).

Mechanisms of therapy resistance are quite heterogeneous because they can be different among patients (inter-patient heterogeneity), co-exist in tumors of the same patient (intra-patient heterogeneity), and be identified within a single tumor (intra-tumor heterogeneity) (11). Resistance can be classified as pre-existing in cells that are insensitive to treatments (primary or intrinsic resistance), as acquired when progressive disease occurs after clinical benefits (secondary or acquired resistance) or as initial attenuation of therapeutic interventions preceding acquired resistance (adaptive resistance). When MAPKi-treated melanoma cells develop adaptive resistance, they become transiently resistant by modifying their molecular phenotype through non-genetic mechanisms such as phenotype switching or metabolic reprogramming (12). Continuous drug treatment drives transiently resistant cells into a stable drug-tolerant state (13). Phenotype switching is the transition from a proliferative (epithelial-like) state to an invasive (mesenchymal-like) one which has been described as a characteristic feature of drug-tolerant melanoma cells and linked to increased resistance to MAPKi (14). Metabolic reprogramming is an essential aspect in the emergence of the drug-tolerant state, in particular as MAPK inhibition reduces glycolysis (15), induces oxidative phosphorylation (OXPHOS) (16,17), and triggers the expression of a mitochondrial biogenesis gene signature (18). Mitochondria are the location of essential energy producing

pathways including OXPHOS, tricarboxylic acid cycle (TCA) and fatty acid oxidation (FAO).

However, the role of lipids in cancer metabolism has been less studied and whether lipid metabolism plays a role during the adaptation to MAPKi remains to be elucidated.

To identify markers of transiently resistant cells and to elucidate mechanisms by which

MAPKi initiate adaptive processes in BRAF-mutated melanoma cells, we performed a targeted screening of cell surface markers in cells treated short-term with the BRAF inhibitor PLX4720. We identified the membrane fatty acid transporter cluster of differentiation 36 (CD36) as the most consistently upregulated cell surface protein accompanying the adaptation of BRAF-mutated melanomas to MAPKi. Because inhibition of oncogenic BRAF inhibits glycolysis, we examined whether fatty acids act as energy sources during the adaptation to MAPKi. We characterized metabolic adaptations and identified a dependency of BRAF-mutated melanoma cells on FAO in the adaptation phase of MAPKi treatment. The simultaneous inhibition of FAO and glycolysis in MAPKi- treated melanoma cells synergistically inhibits tumor cell proliferation and growth in vitro and in vivo, offering new therapeutic insights in overcoming therapy resistance in this disease setting.

Materials and Methods

Cell culture

Melanoma cell lines A375P, UACC62 and SK-Mel28 were obtained from ATCC and cultured in

DMEM (Life technology) containing 4.5 g/l glucose supplemented with 2 mM L-glutamine and 10%

Fetal Calf Serum (FCS) (BioConcept). All patient-derived melanoma cells were obtained from the

University Hospital of Zurich where tumor material was obtained after surgical removal of melanomas from patients after written informed consent (19). Patient-derived melanoma cells were cultured in RPMI-1640 supplemented with GlutaMAXTM (Life technology) and 10% FCS. All cell lines were regularly tested for mycoplasma contamination with a PCR-based assay (#A8994; AppliChem).

Patient material

Freshly isolated tumor biopsies from metastatic stage IV melanoma patients paired pre- and early-on treatment (patients consented to DF/HCC protocol 11-181) from the Massachusetts General Hospital were immediately snap-frozen in liquid nitrogen. Paired biopsies were formalin-fixed and stained with

H&E to estimate tumor and stroma percentage by pathologists. RNA was extracted using the Qiagen

RNeasy Mini kit and 250 ng of RNA per sample was used to arrange RNA libraries with Illumina protocols. RNA-seq was performed at the Broad Institute (Illumina HiSeq2000) and the Wistar

Institute (Illumina NextSeq 500). Library preparation was performed from RNA samples ribo-zero treated using Epicentre’s ScriptSeq Complete Gold kit. Using the High Sensitivity DNA kit, quality check was performed on the Bioanalyzer and quantification was performed using KAPA

Quantification kit. Raw RNA-seq data (BAM files) read counts were summarized by featureCounts

(20) with parameters that only paired-ended, not chimeric and well mapped (mapping quality ≥20) reads were counted. Then normalization was applied to eliminate bias from sequencing depths and gene lengths by edgeR (21), thus RPKMs (Reads Per Kilobase of transcript per Million mapped reads).

Cell surface marker screening of BRAFi-treated A375P cells

Cell surface proteins were analyzed using the LEGENDScreenTM Human PE Kit from Biolegend

(#700001) containing 332 PE-conjugated monoclonal antibodies (AB) specific for human cell surface

markers, 10 mouse, rat or hamster Ig isotype controls and 4 unstained controls (Table S1), arrayed on four 96-well plates with one AB per well. Before staining, lyophilized ABs were reconstituted with

80 μl/well deionized water. Cells were cultured in DMSO or 1 M PLX4720 (PLX) for 36 or 60 h and harvested using Accutase (# A1110501; Life Technology). 24x106 cells from each treatment group were used for the subsequent analysis. To analyze the four different conditions contemporaneously and reduce technical biases, cells from the treatment groups 36 h DMSO, 36 h PLX, 60 h DMSO, and 60 h PLX were labelled using CellTrace Violet according to the manufacturer's instructions at concentrations of 0, 0.1, 1 and 5 μM, respectively. The four groups were mixed and aliquoted in each well of the kit to allow the staining of each PE-conjugated antibody. Each cell population was discriminated by gating on VioBlue-A versus FSC-A plot. SYTOX Green (#S7020; Life technology) was used as viability dye at a concentration of 2.5 μM. Flow cytometry was performed on a BD

LSRFortessa instrument (BD Biosciences) and data were analysed using FlowJo software (Tree

Star, Inc.). All markers with MFI similar to unstained controls were excluded from the analysis.

RNA isolation and quantitative RT-PCR (RT-qPCR)

RNA was isolated with the NucleoSpin RNA II kit (Macherey & Nagel) according to the manufacturer’s protocol. 0.5 to 2 μg of total RNA were reverse transcribed using the High-Capacity

RNA-to-cDNA™ Kit (#4368813; Applied Biosystems) according to the manufacturer’s instruction. qPCR was performed on a Roche LightCycler LC480 instrument. The amplification mixture consisted of 2x KAPA SYBR® FAST qPCR Mastermix (#KK4601; KAPA Biosystems), 7.5 pmol forward and reverse primers and ~20 ng of cDNA template. Thermal cycling was carried out with a 5 min denaturation step at 95 °C, followed by 45 three-step cycles: 10 sec at 95 °C, 10 sec at 60 °C, and

10 sec at 72 °C. Melt curve analysis was carried out to confirm the specific amplification of a target gene and absence of primer dimers. All reactions were run in duplicate. Relative mRNA amount was calculated using the comparative threshold cycle (CT) method (22). Cyclophilin and TATA-box binding protein (TBP) were used as the invariant control. Primer sequences are listed in Table S2.

Flow cytometry-based CD36 analysis

Cells were trypsinized and collected using PBS with 2.5% FCS (FACS buffer) containing either

DMSO or MAPKi, stained on ice with PE anti-human CD36 antibody (1:70) and resuspended in

FACS buffer containing 2.5 M SYTOX Green for exclusion of dead cells. Unstained cells were used as control to set gates. Flow cytometry was performed on a BD LSRFortessa and data were analyzed using FlowJo software.

Fluorescence-activated cell sorting (FACS)

Cells were staining according to the protocol described above for CD36 analysis. For each sorting session unsorted, CD36+ and CD36- fractions were collected for each cell line using the aBD FACS

Aria IIIu instrument. Sorted cells were reanalyzed to evaluate sorting efficiency and reseeded either in DMSO- or MAPKi-containing medium supplemented with 2% penicillin/streptomycin.

FAO assay

Radioactively labelled palmitic acid ([9,10-3H(N)]palmitic acid, 32Ci/mmol, NET043001MC) was purchased from Perkin Elmer. FAO of [9,10-3H(N)]palmitic acid was assessed by the production and release of tritiated water according to a previously described protocol (23).

Fatty acid uptake assay

500 μM stock solution of [9,10-3H(N)]palmitic acid was prepared as described for the FAO assay.

For the fatty acid uptake assay, cells were either grown in medium supplemented with DMSO

(1:5’000) or 1 μM PLX and 0.5 μM AZD for 16 h. For DMSO and MAPKi treatment 50’000 and

100'000 cells per well were seeded into a 48-well plate, respectively. The uptake was initiated by replacing the culture medium with 200 μl of the palmitate/BSA stock solution diluted 4-fold with serum-free medium. The culture plate was incubated for 15 min at 37 °C and subsequently the medium was removed, the plate washed with PBS, and cells were collected in 200 μl RIPA buffer [20 mM TRIS (pH 7.5), 150 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1% (v/v) NP-40, 1% (v/v) sodium deoxycholate]. 100 μl of lysates were transferred into scintillation vials and 3 ml scintillation fluid were added.

Xenograft of human melanoma cells

Two million A375P cells were resuspended in 100 l PBS and injected subcutaneously into the posterior flanks of 6-8-week-old immunodeficient BALB/cAnNRj-Foxn1nu/nu female mice (Janvier

Laboratories). Short-term MAPKi treatment started when tumors reached 200 mm3 and mice were divided into 2 groups (5 mice per group). One group received by oral gavage sodium carboxymethyl cellulose (CMC solution) as vehicle control and the second group received a combination of vemurafenib (30 mg/kg) and selumetinib (15 mg/kg) diluted in CMC solution every 12 h for 36 or 60 h. For combinational treatment studies, two million A375P cells were resuspended in PBS, mixed with matrigel (60% PBS, 40% matrigel) and injected subcutaneously into the posterior flanks of

BALB/cAnNRj-Foxn1nu/nu mice as described above. Mice were divided in groups of 8 when tumors reached 60 mm3. Four groups received by oral gavage either vehicle control alone or with 100 mg/kg dichloroacetate (DCA) or 32 mg/kg ETO or the combination of DCA and ETO. Four groups received by oral gavage either a combination of vemurafenib (24 mg/kg) and selumetinib (12 mg/kg) alone or with 100 mg/kg DCA or with 32 mg/kg ETO or the combination of DCA and ETO. Tumor formation was monitored every 3 days and tumor volume was calculated by the ellipsoidal formula (tumor volume = ½ * (width2 * length). All protocols for animal use and experiments were approved by the

Veterinary Office of Zurich (Switzerland).

Glycolysis stress test and lactate measurements

For the glycolysis stress test, melanoma cells treated with DMSO or the indicated inhibitors for 48 h were seeded in quintuplicates in a Seahorse XF Microplate. Cells were incubated overnight in a humidified 37 °C incubator with 5% CO2. ECAR measurements were performed using the XF24

Extracellular Flux analyzer (Seahorse Bioscience). Prior to performing an assay, growth medium was exchanged with the appropriate unbuffered assay medium (Krebs-Henseleit buffer). 450 µl of the assay medium containing the corresponding inhibitors were added to each well and the plate was incubated for 1 h at 37 °C in a non-CO2 incubator. Each measurement cycle consisted of a mixing time of 3 minutes, a waiting time of 2 minutes and a data acquisition period of 2 minutes. ECAR data points refer to the average rates during the measurement cycles. All compounds were prepared at appropriate concentrations in assay medium and adjusted to pH 7.4. In a typical experiment, 3 baseline measurements were taken prior to the addition of 15 mM glucose, 3 measurements were taken prior the addition of 1 μM oligomycin, 3 measurements were taken prior and after the addition of 200 mM 2-deoxy-D-glucose. ECAR was normalized to cell number in each experiment.

Lactate concentrations were determined using the Cedex Bio Analyzer instrument (Roche

Diagnostics) in cell-free supernatants of cells treated with MAPKi for 24 h. Values were normalized to integral of viable cells.

Proliferation assay

To determine the effects of ETO as single agent or in combination with DCA on cell proliferation,

10000 cells were seeded in 96-well plates and after attachment treated with DMSO, 1 M PLX alone or in combination with 0.5 M AZD. ETO, DCA or their combination were added at the indicated concentrations after 24 h. After 72 h cells were washed and incubated for 30 min at 37 °C with

PrestoBlue® Cell Viability Reagent (#A13262; ThermoFisher Scientific). The converted fluorescent dye was measured using the Infinite® M1000 PRO microplate reader (Tecan). Values were normalized to DMSO control. Compound synergy score was determined based on the BLISS model using Combenefit (24).

CRISPR/Cas9 gene editing

A375P cells were genetically engineered to generate CD36 knockout (KO) cells using the lentiCRISPRv2 plasmid (#52961; Addgene). Three sgRNAs for CD36 were designed using the

ATUM gRNA Designer (https://www.atum.bio/eCommerce/cas9/input). CD36 sgRNA sequences used are: CD36 KO1: 5’-(GTCTCTTTCCTGCAGCCCAA)NGG-3’; CD36 KO2: 5’-

(GGAGGTATTCTAATGCCAGT)NGG-3’; CD36 KO3: 5’-(ACTTTGAGAGAACTGTTATG)NGG-3’.

Lentiviral particles were prepared by transiently transfecting HEK293T cells with lentiviral vectors together with packaging vectors (pMD2 and psPAX2) using the polyethylenimine transfection protocol. Supernatants were collected 48 h posttransfection, passed through a 0.45 m filter (BD

Biosciences), and stored at -80 °C (25).

For lentiviral transduction cells were seeded in 6 well-plates and infected with 0.5 ml lentiviral particles in 1.5 ml of complete medium. LentiCRISPRv2 without sgRNA was used as control. Infected cells were selected with 2 g/ml puromycin. CD36 KO was confirmed by FACS.

CD36 protein levels were significantly reduced using CD36 KO1 and KO3 sgRNAs, whereas CD36

KO2 sgRNA did not decrease CD36 and was excluded from further experiments. Single cell sorting was performed with CD36 KO1 and KO3 pools, and after expansion CD36 expression in several

clones was analyzed by FACS. 2 clones from each sgRNA with a complete deletion of CD36 expression were used for fatty acid uptake and FAO assays. Three clones were isolated from cells transduced with lentiviral particles expressing the empty LentiCRISPRv2 and used as controls.

siRNA cell transfection

Cells were grown to 50% confluence in 24-well plates and transfected using Lipofectamine

RNAiMAX (Life Technologies) according to the manufacturer’s protocol with two siRNAs against

PPARA (siPPARA#1: 5’-(GAUCUAGAGAGCCCGUUAUCU)dTdT-3’; siPPARA#2: 5’-

(GGAGCAUUGAACAUCGAAU)dTdT-3’) or two siRNAs against CPT1A (siCPT1A #1: 5’-(

UGUGCUGGGCUGGAAAGAA)dTdT-3’; siCPT1A #2: 5’-(GGGUAAACUUUUGUUUUGU)dTdT-3’).

AllStars negative control siRNA (Qiagen) was used as scramble control. 24 h after transfection cells were treated with DMSO, PLX alone or in combination with AZD. Cells were analyzed after treatment for 72 h.

RNA sequencing and data processing

A375P cells were treated with 0.01% (v/v) DMSO as control or 1 M PLX for 36 h. RNA was isolated with the NucleoSpin RNA II kit according to the manufacturer’s protocol. RNA integrity was checked using the Bioanalyzer system (RIN above 8.70) (Agilent Technologies) and the concentration was measured using Quant-IT RiboGreen RNA Assays (Life Technologies). Libraries for Illumina sequencing have been prepared with 200 ng total RNA input material using the TruSeq Stranded

Total RNA Library Prep kit Ribozero (Illumina) and quality checked using the Fragment Analyzer standard sensitivity NGS kit (AATI). SR126 sequencing was performed on an Illumina HiSeq 2500 system (HiSeq SBS kit v4). The RNA-seq data are deposited in ArrayExpress (E-MTAB-7453).

A detailed description of the computational methods used to analyze the RNA-seq dataset is available in the supplemental information.

Statistical test

All statistical tests were performed using GraphPad Prism version 7.0. For in vitro experiments, unless differently specified, data were shown as the mean ± SD of 3 independent experiments.

Statistical differences between conditions were determined using ordinary one-way Anova or

Student’s t-test. Xenograft tumors were analyzed using 2-way Anova to assess differences in average tumor growth. Changes in gene expression of human tissues were correlated using

Spearman’s rho rank correlation.

Results

Identification of expression change patterns of cell surface proteins during MAPK inhibition

The majority of therapeutics target membrane proteins, accessible on the cell surface, to alter cellular signaling and metabolism (26). To identify cell surface markers altered by MAPKi, the BRAF- mutated melanoma cell line A375P was treated with the BRAF inhibitor PLX at 1 M for 36 or 60 h.

Cell surface epitope expression was analysed by flow cytometry using a cell surface marker screening kit containing Phycoerythrin-conjugated (PE-conjugated) antibodies against 332 cell surface epitopes (Table S1). The expression was assessed in 4 different conditions, including control 36 h (36 h DMSO), PLX-treated for 36 h (36 h PLX), control 60 h (60 h DMSO) and PLX- treated for 60 h (60 h PLX). To discriminate each condition by fluorescence-activated cell sorting

(FACS), A375P cells were labelled with different concentrations of CellTrace Violet dye and then stained for the antibody panel (Figure S1A and S1B). Mean fluorescence intensities (MFI) for each

PE-conjugated antibody were measured and overlaid for the 4 different conditions (Figure S1C-E).

With a threshold >1.6 and <0.6 fold change (FC) for at least one time point, 284 cell surface epitopes

(85.5%) showed no differential expression after 36 or 60 h of BRAF inhibition (Figure S1C, Table

S3). In contrast, the expression of 48 cell surface proteins (14.5%) was changed upon treatment. In particular, 24 membrane proteins decreased expression and 24 membrane proteins increased expression in response to PLX treatment for 36 and/or 60 h (Figure 1A, S1D and S1E, Table S3).

CD36 mRNA and protein levels are increased in melanoma cell lines, xenograft and human tumors treated with MAPKi

In order to identify general surface proteins that were adaptively increased by MAPKi, the hits identified in our screen were further investigated by qRT-PCR in 2 melanoma cell lines (A375P and

UACC62) and 4 early passage melanoma cultures (M980513, M080326, M130820, and M130309).

Changes in mRNA levels were consistent with changes in cell surface protein levels in A375P cells upon BRAF inhibition. Overall, 4 hits out of the 24 upregulated proteins showed a consistent increase at the mRNA level after 36 h of BRAF inhibition. In particular, mRNA levels of CD24, CD36 and CD166 increased significantly in all PLX-treated cell lines, while CD271 expression increased in

5 out of 6 tested cell lines (Figure 1B). Among the 24 downregulated proteins, the mRNA levels of

CD61 and CD96 were most consistently downregulated in PLX-treated cell lines (Figure 1B). In order to assess the clinical relevance of our finding, we evaluated the expression changes of our hits by analyzing RNA sequencing (RNA-seq) data derived from paired pre- and early on-treatment tumor biopsies that were available in two independent cohorts of patients with BRAF-mutated melanoma. Among the short-listed candidates, CD36 showed the most consistent upregulation in on- treatment melanoma biopsies when compared with the pre-treatment biopsies (Figure 1C and D). In particular, in a cohort (#1) from the Massachusetts General Hospital (MGH) 7 out of 11 biopsies from patients treated for 14-16 days with BRAFi alone or in combination with MEKi showed an upregulation of CD36 (Figure 1C). Data were further validated in a second cohort (#2) of patients from

MGH and CD36 expression was increased in 7 out of 8 on-treatment biopsies when compared with paired pre-treated tumors (Figure 1D). In addition, CD24 and CD61 showed expression patterns similar to the one observed in vitro, whereas CD166, CD271 and CD96 revealed inconsistent expression changes (Figure S2A). In order to examine if increased CD36 mRNA levels lead to elevated protein levels and/or an increase of a CD36-positive (CD36+) cell population, A375P,

UACC62 and 5 early passage melanoma cell cultures were treated with PLX or the MEKi AZD6244

(AZD) for 36 h. In all treated cell lines compared to untreated controls, the percentage of CD36+ cells increased ranging from 15.9% to 63.7% and from 25.2% to 92.4% after PLX or AZD treatment, respectively (Figure 1E). Conversely, the percentage of CD36+ cells did not increase after PLX treatment in the non-BRAF-mutant melanoma cell line MeWo and in the BRAFV600E/NRASQ61R double-mutated early culture M121224 which is resistant to BRAF inhibition (Figure S2B). However, the percentage of CD36+ cells in MeWo and M121224 increased upon MAPK pathway inhibition by

MEKi, indicating a strong dependence of CD36 upregulation on the inhibition of MAPK signaling

(Figure S2B and Table S4). Interestingly, the sensitive melanoma cell lines SK-Mel28 and M140124 which intrinsically express high levels of CD36 also upregulate CD36 MFI after PLX or AZD treatments (Figure S2C). Finally, we injected A375P cells into immunodeficient nude mice and examined if MAPK inhibition increases CD36 expression in subcutaneous A375P tumors. After the tumors reached ~200 mm3, mice were treated with vemurafenib (PLX4032) combined with AZD or with DMSO as solvent control for 36 and 60 h. After tumor collection and dissociation into single cells, the proportion of CD36+ cells was measured by flow cytometry. The fraction of CD36+ cells increased in tumors treated with MAPKi compared to control tumors both at 36 and 60 h (Figure

S2D). Taken together, these data show that the percentage of CD36+ cells increases during the adaptation phase to MAPKi and suggest CD36 as a candidate marker for identifying transiently resistant cells in BRAF-mutated melanomas.

MAPK inhibition induces and maintains the CD36+ population in melanoma

Since the percentage of CD36+ cells increased after MAPKi treatment, we examined whether the

CD36+ subpopulation was generated or selected during MAPK inhibition. Untreated and 36 h MAPKi- treated (PLX+AZD) UACC62 and M130820 cells were FACS sorted into CD36+ and CD36- subpopulations while the bulk population was used as control. After sorting, DMSO- or MAPKi- treated cells were reseeded into DMSO- or MAPKi-containing medium, respectively, and CD36 expression was determined after 3 days (Figure S3A). CD36+ and CD36- subpopulations from

DMSO-treated cells did not maintain their phenotype and emerged as the mixed population that characterized unsorted cells in DMSO-containing medium (Figure S3B). In contrast, CD36+ subpopulations from MAPKi-treated cells maintained their phenotype after 3 days of continuous

MAPK inhibition and a substantial amount of previously CD36- cells became positive (Figure S3C).

These data indicate that CD36+ cells are adaptively induced rather than selected by MAPKi and that continuous MAPK inhibition is necessary to sustain high CD36 expression levels. Furthermore,

CD36+ and CD36- UACC62 cells were sorted after 36 h of MAPKi treatment and re-plated into

DMSO- or MAPKi-containing medium to determine CD36 levels at day 2, 5, 8 and 12 (Figure S3D).

In control medium, the CD36+ fraction of UACC62 cells gradually decreased, whereas in MAPKi- containing medium the CD36+ subpopulation maintained its CD36 expression over 12 days. In addition, the CD36- subpopulation increased its CD36 expression in MAPKi-containing medium and reached comparable levels to the CD36+ subpopulation after 5 days (Figure S3D).

To assess whether the trend observed in vitro could be identified in vivo, we analyzed a publicly available single cell RNA-seq dataset (27) generated from a patient-derived xenograft model

(MEL006) treated with the combination of BRAFi and MEKi. Data showed increased CD36 expression at the single cell level during MAPK inhibition. Individual melanoma cells treated with

MAPKi showed higher CD36 expression during the adaptation and the drug-tolerant phases compared with the pre-treatment stage (Figure 1F). Interestingly, CD36 expression decreased at the acquired resistance phase when MAPK pathway is re-activated or compensatory pathways are

activated (Figure 1F, S3E). In conclusion, these data indicate that MAPK inhibition induces and maintains CD36 expression in BRAF-mutated melanoma cells during both adaptation and drug- tolerant states and suggest increased CD36 expression as a candidate marker for adaptation and drug-tolerant phases of MAPK inhibition.

Short-term MAPK inhibition induces expression of genes involved in FAO

As existing data show a role for CD36 in the process of fatty acid uptake and FAO (28,29), expression profiles of genes involved in lipid metabolism were investigated in BRAF-mutated melanoma cells after short-term MAPKi treatment. The analysis of RNA-seq data from A375P cells treated with PLX for 36 h showed a significant downregulation of genes involved in glycolysis and upregulation of genes involved in fatty acid transport and mitochondrial and peroxisomal lipid catabolism including acyl-CoA synthetases, acyl-CoA dehydrogenases, and transcription factors involved in the regulation of lipid metabolism (Figure 2A, S4A). RNA-seq data were validated by qPCR in A375P and a panel of BRAF-mutated melanoma cells (UACC62, M980513, M130820 and

M140122) that were treated with PLX or AZD. A set of genes involved in FAO was consistently upregulated after 36 h of MAPK inhibition (Figure 2B, 2C, S4B). In contrast, the expression of glycolytic genes as well as extracellular signal-regulated kinase (ERK) target genes was decreased after MAPKi treatment (Figure 2B, 2C, S4B and S4C). The expression of genes involved in FAO

(CD36, PPARA, CPT1A, ACAD10 and ACSBG1), in glycolysis (HK2, GLUT3, LDHA), and the ERK target gene DUSP6 was similarly affected in A375P xenograft tumors from mice treated with MAPKi for 36 and 60 h (Figure 2D, 2E). In contrast, in A375P xenograft tumors treated with MAPKi for 24 days which acquired resistance, the mRNA levels of FAO genes as well as glycolytic and ERK target genes showed similar levels compared to the control group (Figure S4E, S4F). Interestingly, CPT1A expression was upregulated in 7 out of 19 biopsies from on-treatment patients compared with paired pre-treatment stage (Figure S4D). Taken together, these data suggest that metabolic reprogramming occurs in the early stage of MAPK inhibition, namely decreasing glycolysis and increasing FAO.

MAPK inhibition decreases glycolytic flux in BRAFV600E melanoma cells

To test whether the expression changes drive a metabolic program in the adaptive stage of MAPK inhibition, we determined key cellular metabolic parameters in DMSO- and MAPKi-treated melanoma cell lines. The glycolytic flux was determined by measuring the extracellular acidification rate (ECAR) in melanoma cells treated with PLX alone or in combination with AZD for 72 h. In accordance with the gene expression profiles, the rate of glycolysis and glycolytic capacity were significantly reduced in melanoma cell lines after treatment with MAPKi (Figure 3A, S5A and S5B). Accordingly, MAPKi- treated cells showed lower levels of lactate production (Figure 3B). When the capacity to utilize glycolysis under oxidative stress (glycolytic reserve) was evaluated, MAPKi-treated M130820 cells showed a significantly reduced glycolytic reserve but no significant differences were observed in

A375P and SK-Mel28 cells (Figure 3A, S5A and S5B). Metabolomic analysis of A375P cells revealed that levels of glycolytic intermediates were lower in DMSO-treated compared to cells treated with PLX for 72 h, confirming a decreased glycolytic flux in PLX-treated cells (Figure S5C and Table S5). Accordingly, untreated A375P were extremely vulnerable to 72 h of glucose starvation compared to cells treated with MAPKi due to a more pronounced glycolytic state (Figure

3C). Overall, these data suggest that MAPK inhibition reduces glycolytic flux in BRAF-mutated melanoma cells.

MAPK inhibition increases FAO in BRAFV600E melanoma cells

The analysis of gene expression data suggested a metabolic switch towards an increased rate of

FAO in BRAFV600E melanoma cells upon MAPK inhibition (Figure 2, S4A, S4B, S4D). To test whether the increased expression levels of genes associated with FAO translated into functionally

3 elevated levels of FAO, radioactively labelled palmitic acid was added to the medium and H2O was measured as readout for the rate of FAO. Strikingly, we found significantly higher levels of FAO in 6

BRAF-mutated melanoma cell lines after treatment with PLX as a single agent or in combination with

AZD (Figure 3D). To validate if the metabolic changes are exclusively induced by MAPK inhibition, the BRAFV600E/NRASQ61R cell line M121224 and the BRAF wild-type cell line MeWo were treated with

PLX or with a combination of PLX and AZD. M121224 and MeWo are resistant to PLX but sensitive to the combinational treatment. Accordingly, PLX did not affect FAO, whereas MEKi increased FAO in MeWo cells (Figure 3E). Only the BRAF-mutated, PLX-sensitive cell line M080326 and the double-mutated PLX-resistant but AZD-sensitive M121224 did not show increased FAO in response

to MAPK inhibition (Figure 3F). Strikingly, these 2 cell lines that failed to upregulate FAO exhibited the lowest absolute FAO levels and the highest sensitivity to combined MAPKi treatment after 5 days

(Figure 3G, S5D). In summary, our data suggest that MAPK inhibition triggers a metabolic reprogramming in melanoma cells and that upregulation of FAO allows MAPKi-tolerant cells to survive under the metabolic stress induced by MAPKi.

CD36+ cells have higher rates of FAO

Next, we assessed whether there is a functional relationship between the MAPKi-mediated increase of CD36 levels and rates of FAO. Untreated and MAPKi-treated UACC62, M130820 and M140122 cells were sorted into CD36+ and CD36- subpopulations after 36 h of treatment. An unsorted population was taken as control (Figure 4A and S6B). Sorted cells were reseeded into DMSO- or

MAPKi-containing medium and the rate of FAO was determined on the following day (Figure 4A).

Unsorted MAPKi-treated cells had higher levels of FAO compared to control cells, and the FAO rate was significantly higher in MAPKi-treated CD36+ compared to CD36- subpopulations (Figure 4B).

Interestingly, CD36+ M130820 and M140122 cells cultured in DMSO-containing medium showed higher levels of FAO compared to CD36- cells. Furthermore, changes in expression levels of CD36 and CPT1A, the rate limiting enzyme of FAO, correlated significantly in a cohort of 20 paired pre- and post-treatment tumor tissues (Figure 4C).

To assess the functional role of CD36 in FAO, we knocked out CD36 in A375P cells using

CRISPR/Cas9. We chose A375P cells because both CD36 protein and mRNA levels as well as the

FAO rate were highly increased in response to MAPKi treatment. We used two sgRNAs targeting

CD36 and selected two knockout (KO) single cell clones from each sgRNA for further analyses.

While three MAPKi-treated control clones showed high CD36 expression, the expression in the four

CD36 KO clones was undetectable (Figure S6A). To address the possible role of CD36 as fatty acid transporter in MAPKi-treated melanoma cells, fatty acid uptake was measured in control and CD36

KO cells after 16 hours of MAPKi treatment. No differences in fatty acid uptake were found both in vehicle- and MAPKi-treated as well as control and CD36 KO cells (Figure S6C). These results suggest that CD36 does not function as a fatty acid transporter in melanoma cells. However, no differences may be explained by compensatory mechanisms that can occur upon genetic deletion of

CD36. To evaluate potential compensatory mechanisms, expression of fatty acid transport proteins

FATP1, FATP2, FATP3, FATP4, FABP1 and FABP4 was measured in CD36 WT and KO A375P cells. FATP2, FABP1 and FABP4 mRNA levels were undetectable in untreated and MAPKi-treated conditions, and FATP1, FATP3 and FATP4 expression was similar in WT and KO cells (Figure

S6D).

Next, we asked whether CD36 affects FAO in MAPKi-treated melanoma cells. The FAO rate was similar in control and CD36 KO cells, whereas MAPKi treatment similarly increased FAO in control and KO cells (Figure S6E). In addition, the FAO rate was also comparable in vehicle- and MAPKi- treated control and CD36 knockdown (KD) cells using 5 different lentiviral-mediated shRNAs (Figure

S6F). Finally, we used the FATP inhibitor Lipofermata to assess the role of fatty acid transporters in

FAO. Lipofermata decreased the rate of FAO in untreated and MAPKi-treated A375P cells (Figure

4D), but CD36 KO did not show an additive effect upon Lipofermata treatment (Figure S6G), indicating a nonfunctional role of CD36 in MAPKi-induced FAO.

FAO upregulation is mediated by the transcription factor PPAR in melanoma cells

Peroxisome proliferator-activated receptor alpha (PPAR) is a key transcriptional regulator of genes involved in FAO. We determined the expression level of PPARA to examine whether PPAR is involved in the FAO upregulation in MAPKi-treated BRAF-mutated melanoma cells. PPARA expression was upregulated in melanoma cells and A375P xenograft tumors in the early stage of

MAPK inhibition (Figure 2, S4A). To investigate if PPARA KD suppresses the induction of FAO in response to MAPKi treatment, PPARA expression was efficiently decreased with two siRNAs

(Figure S7A) in vehicle- and MAPKi-treated A375P, UACC62 and M980513 cells. PPARA KD significantly reduced FAO levels as well as expression of CPT1A in vehicle- and MAPKi-treated cells

(Figure 5A and 5B) without affecting CD36 expression (Figure S7B).

To confirm these findings, we tested whether pharmacological inhibition of PPAR decreases FAO levels and expression of CPT1A. The PPAR antagonist GW6471 significantly decreased FAO levels in vehicle- and MAPKi-treated A375P, UACC62 and M980513 cells (Figure

5C). GW6471 also reduced the expression of CPT1A in MAPKi-treated A375P cells (Figure 5D).

Conversely, the PPAR agonist GW7647 increased FAO and CPT1A expression in MAPKi-treated

A375P and M980513 cells (Figure 5E, 5F). However, the effect of GW7647 on CD36 expression was inconsistent as it increased CD36 expression in A375P cells but no effect could be observed in

M980513 and M130820 cells (Figure S7C). Taken together, these data reveal a role of PPAR in the upregulation of CPT1A and mitochondrial FAO in BRAF-mutated melanoma cells in response to

MAPK inhibition.

Pharmacological inhibition of FAO induces glycolysis in MAPKi-treated melanoma cells

Etomoxir (ETO) is an irreversible pharmacological inhibitor of CPT1A and thereby of mitochondrial

FAO. We examined the effects of ETO on cell proliferation and viability to further investigate the role of FAO during the adaptation to MAPKi. Incubation with ETO for 4 and 72 h significantly reduced the

FAO rate of vehicle- and MAPKi-treated A375P cells (Figure S8A). However, treatment with ETO for

72 h resulted in only a minor reduction of cell viability both in vehicle- and MAPKi-treated cells

(Figure S8B). These results prompted us to investigate how glucose is utilized in ETO-treated melanoma cells. Interestingly, FAO inhibition significantly increased the glycolytic flux both in vehicle- and MAPKi-treated A375P, SK-MEL-28, and M130820 melanoma cells (Figure 6A, S8C and S8D); however, the mRNA levels of glycolytic genes were unchanged (data not shown). In accordance with an increased glycolytic flux induced by FAO inhibition, CPT1A KD as well as the PPAR antagonist

GW6471 increased ECAR both in untreated and PLX-treated A375P cells, hence showing similar effects on glycolysis as etomoxir (Figure 6B and S8D). Taken together, these data suggest an increased glycolytic flux as a compensatory mechanism to overcome the metabolic stress induced by FAO inhibition.

Inhibitors of FAO and glycolysis act synergistically to maximize the efficacy of MAPKi

To confirm the reactivation of glycolysis as a compensatory mechanism during FAO inhibition, vehicle- and MAPKi-treated cells were treated with ETO in glucose-free medium. ETO increased cell death of MAPKi-treated cells in glucose-free medium (Figure 6C). To further investigate whether

MAPKi-treated BRAF-mutated melanoma cells show vulnerability to concomitant inhibition of FAO and glycolysis, 5 cell lines were treated with the glycolytic inhibitor DCA and ETO in a matrix combination design. Cell viability and apoptosis were assessed during MAPK inhibition. While

MAPKi in combination with ETO for 5 days did not significantly affect cell viability and apoptosis in melanoma cells, the BLISS interaction analysis revealed that the combination of ETO and DCA synergistically reduced cell viability in MAPKi-treated A375P, UACC62, M980513, M130820 and

M140122 cells (Figure 6D and S8E). Accordingly, annexin V/propidium iodide staining showed increased apoptosis when ETO and DCA were used in combination with MAPKi on A375P and

M980513 cells (Figure S8F). To explore the mechanism by which DCA and ETO cooperate to enhance the efficacy of MAPKi, we tested the effects of DCA, ETO and the combination of DCA and

ETO on ECAR and FAO. DCA inhibited glycolysis (Figure S8G), but it increased FAO in DMSO- and

MAPKi-treated A375P cells (Figure 6E). In contrast, the FAO inhibitor ETO decreased FAO (Figure

6E and S8A) but increased glycolysis (Figure 6A and S8G). Only the double combination of ETO and DCA efficiently decreased glycolysis and FAO in MAPKi-treated BRAF-mutated melanoma cells, because ETO cannot overturn the glycolytic inhibition caused by DCA and DCA cannot overturn

FAO inhibition caused by ETO (Figure 6E and S8G). In summary, these results suggest that concurrently targeting both FAO and glycolysis might be a promising strategy to overcome drug resistance to MAPKi.

FAO inhibition enhances tumor growth but delays tumor relapse in combination with glycolytic inhibitors in MAPKi-treated A375P xenografts

MAPK inhibitors suppress glycolysis and provide a strong clinical benefit in BRAF-mutated melanoma, and reactivation of glycolysis characterizes the relapse phase of acquired resistance.

Simultaneous inhibition of BRAF and glycolysis induces cell death in BRAF inhibitor-resistant melanoma cells (15). Since our data show that upregulation of FAO plays a role in the adaptive response to MAPK inhibition, and FAO inhibition reactivates glycolysis in MAPKi-treated melanoma cells, we tested the in vivo efficacy of the combination of ETO and DCA with MAPKi in a xenograft model of A375P cells. When tumors reached 60 mm3, mice were treated with 32 mg/kg ETO, 100 mg/kg DCA or the combination of both with or without MAPKi (24 mg/kg PLX4032 and 12 mg/kg

AZD). As expected, the combination of BRAF and MEK inhibitors impaired tumor growth compared to tumors grown in vehicle-treated mice (Figure S9A). Surprisingly, treatment with ETO alone significantly increased tumor growth compared to tumors from mice treated with vehicle, DCA or a combination of DCA with ETO without affecting mouse body weight (Figure 6F, S9B and S9C). ETO administered in combination with MAPKi significantly increased tumor growth compared to MAPKi alone (Figure 6G, S9B). In contrast, inhibition of glycolysis by DCA in combination with MAPKi significantly impaired tumor growth and delayed the onset of tumor resistance compared to treatment

with MAPKi alone (Figure 6H). Finally, the combination of MAPKi, DCA, and ETO resulted in a significantly greater effect when compared to MAPKi alone or in combination with DCA without affecting mouse body weight (Figure 6H, S9B and S9D). In summary, in vivo and in vitro data show that concomitant inhibition of FAO and glycolysis may induce apoptosis in MAPKi-treated melanoma cells and indicate that patients may benefit from the concomitant addition of FAO and glycolysis inhibitors to their MAPKi therapies (Figure 6I).

DISCUSSION

Short-term treatment with MAPK inhibition causes a transient molecular reprogramming in BRAF- mutated melanoma cells. Drug-tolerant cells survive treatments, in a quiescent state, before acquiring drug resistance-conferring secondary mutations or activating compensatory pathways that drive resistance (30). We identified a metabolic reprogramming of BRAF-mutated melanomas that occurs during the initial response phase of treatment and may be therapeutically exploited to enhance the efficacy of MAPKi. We performed a FACS-based screen to identify cell surface proteins that are dysregulated by short-term treatment with MAPKi and validated them in various pre-clinical melanoma models and patients’ tumor specimens treated with MAPKi. Our data show that CD36 is the most consistently upregulated cell surface protein upon short-term treatment with MAPKi, suggesting that CD36 may be a reliable marker of transiently resistant and MAPKi-tolerant melanoma cells. In agreement with recent studies (31), the stem cell marker CD271 was also consistently upregulated upon MAPK inhibition.

Several studies have linked CD36 to patient prognosis, metastatic progression, and drug resistance in human cancers (28,32). We show that MAPK inhibition induced and maintained CD36 expression in BRAF-mutated melanoma during the adaptation and drug-tolerant phases, while CD36 expression returned to pre-treatment baseline levels after acquired resistance developed.

Upregulation of CD36 pointed toward a role of fatty acid metabolism in the adaptation of BRAF- mutated melanoma to MAPKi (28,29). Therefore, we focused on fatty acid metabolism during the initial phase of MAPK inhibition. Several studies associated the MAPK pathway and melanoma with changes in lipid metabolism (33-35), but its role in allowing melanoma cells to survive and acquire resistance to MAPK inhibition has not yet been investigated. We showed that MAPK inhibition induced the expression of genes involved in FAO. In particular, mRNA levels of CPT1A, a rate- limiting mitochondrial enzyme of FAO that mediates the transport of fatty acids into the mitochondria

(36), were consistently increased. Accordingly, the rate of FAO increased in BRAF-mutated MAPKi- treated cells. Since MAPK inhibition decreased glycolytic flux, our data suggest that oxidized fatty acids are required for energy production and might serve as an alternative carbon source during the initial response phase of MAPKi treatment.

To examine whether CD36 played a role in the regulation of FAO in response to MAPK inhibition, we determined FAO levels in CD36+ and CD36- subpopulations. CD36+ cells showed

higher rates of FAO compared to CD36-. However, in contrast to existing data suggesting a role of

CD36 in fatty acid uptake and metabolism (37-39), neither knockdown nor knockout of CD36 impaired fatty acid uptake or FAO in MAPKi-treated BRAF-mutated melanoma. Even though inhibition of fatty acid transporters using lipofermata inhibited FAO in vehicle- and MAPKi-treated cells, CD36 KO did not show synergistic effect with lipofermata. Furthermore, our data excluded the involvement of compensatory mechanisms of other fatty acid transporters in CD36-depleted BRAF- mutated melanoma cells. Since CD36 is a multi-ligand scavenger receptor that functions in a context-dependent manner (40), we cannot exclude that other ligands or lipids bind CD36 in MAPKi- treated melanoma cells.

To explore the molecular mechanism involved in the upregulation of FAO upon MAPK inhibition, we examined the expression of genes involved in the regulation of lipid metabolism.

Expression of PPARA was significantly increased in melanoma cells treated with BRAFi or MEKi.

PPAR is a member of the nuclear receptor superfamily of transcription factors that control nutrient sensing and transcriptional regulation of metabolic pathways, especially fatty acid transport and FAO

(41). Pharmacological inhibition or downregulation of PPAR significantly decreased the expression of CPT1A and subsequently the rate of FAO. Interestingly, the expression of other genes involved in

FAO as well as CD36 was not affected by PPAR KD or inhibition. PPAR agonists increased both

CPT1A mRNA and FAO levels, but expression of other genes involved in fatty acid catabolism remained unchanged. Taken together, our data indicated that PPAR, by inducing CPT1A expression, is the major regulator of FAO upon MAPK inhibition.

BRAF inhibition suppresses glycolysis through controlling a network of glycolytic regulators

(15) and increases oxidative phosphorylation (16). We showed that short-term treatment with MAPKi decreased glycolytic flux and concomitantly upregulated FAO (Figure 6I). Fatty acids are the major source of energy and metabolized in mitochondria by FAO. FAO fuels the TCA cycle with acetyl-CoA as well as reducing equivalents for OXPHOS (42). Our data suggest that the dependency on FAO might represent a metabolic vulnerability of MAPKi-tolerant cells that could be exploited as a novel treatment option. Therefore, we examined if pharmacological inhibition of FAO with ETO inhibits cell viability. ETO strongly decreased energy metabolism and reduced tumor growth in triple negative breast cancer with high levels of fatty acid catabolism (43) and re-sensitized nasopharyngeal carcinoma to radiation therapy (44). However, FAO inhibition did not lead to enhanced cell death or

reduced proliferation of MAPKi-treated BRAF-mutated melanoma cells due to a compensatory activation of glycolysis (Figure 6I). ETO and the PPAR antagonist GW6471 significantly reduced

FAO levels but concomitantly increased glycolytic flux in MAPKi-treated cells. Therefore, ETO showed a strong effect on cell viability only in glucose-free medium. Accordingly, increased glycolytic flux in response to FAO inhibition with ETO has already been described (45). High proliferation rates are often associated with high levels of aerobic glycolysis (46) and accordingly increased glycolysis in response to treatments with ETO alone or in combination with MAPKi paradoxically enhanced the growth of A375P xenograft tumors. However, the combination of MAPK, FAO and glycolysis inhibitors significantly reduced both glycolysis and FAO showing a synergistic effect on tumor growth when compared to treatment with MAPKi alone or in combination with DCA or ETO (Figure 6I). The concomitant inhibition of glycolysis and OXPHOS has previously been proposed as therapeutic treatment option for ovarian cancer. DCA in combination with metformin, an OXPHOS inhibitor, synergistically suppressed tumor growth (47), while metformin and starvation reduced cancer progression in colon and breast cancer (48).

In summary, BRAF-mutated melanoma cells are characterized by high levels of PPAR- mediated and CPT1A-dependent FAO in the initial phase of MAPKi treatment. Increased FAO is essential for melanoma cells to survive under the MAPKi-induced metabolic stress during the adaptation stage that precedes the occurrence of acquired resistance. Our studies identified drug- tolerant melanoma cells marked by CD36. As an extracellular, accessible surface marker, CD36 could be exploited as a therapeutic target for antibody drug conjugates, bispecific T cell engager or chimeric antigen receptors in a combination treatment with MAPKi. Such combinations were not investigated in this study and need to be further evaluated. Targeting FAO induced higher glycolytic flux, demonstrating the extraordinary metabolic plasticity of melanoma cells. Due to the promising therapeutic advances of cancer immunotherapy in melanoma and the role of metabolism in the regulation of tumor microenvironment (49,50), the effects of FAO activators or inhibitors, alone or in combination with MAPKi, could be tested in immunocompetent syngeneic and genetically modified mouse models. A triple combination using MAPK, FAO and glycolytic inhibitors should be considered as a novel treatment option to prevent the development of acquired drug resistance in BRAF- mutated melanomas (Figure 6I).

ACKNOWLEDGEMENTS

We thank Dr. Malgorzata Kisielow and Anette Schütz (ETH Zurich flow cytometry facility) for their technical support. We thank Drs. Michael Prummer and Nora Toussaint (NEXUS Personalized

Health Technologies, ETH Zurich) and Dr. Christian Beisel, Katja Eschbach, Elodie Burcklen, and

Manuel Kohler (Genomics Facility, D-BSSE) for their support in RNA-seq processing. We thank the

University Research Priority Project (URRP) in cancer research biobank of the University Hospital

Zurich for the early passage melanoma cultures. This work was supported in part by the Swiss

Cancer Research Grant KFS-3651-02-2015 and a HMZ (Hochschulmedizin Zurich) Seed Project

Grant to W.K, the Swiss National Science Foundation (SNSF) Grant 31003A_166245 to W.J.K., in part by NIH grants P01 CA114046, P50 CA174523, U54 CA224070, DoD - PRCRP WX1XWH-16-1-

0119 [CA150619] and the Dr. Miriam and Sheldon G. Adelson Medical Research Foundation to M.H.

AUTHOR CONTRIBUTIONS

Conceptualization, A.A., S.F., W.K., W.J.K.; Methodology, A.A., W.J.K.; Validation A.A, D.M., C.D.C,

W.K., W.J.K.; Formal analysis, A.A., C.D.C.; Investigation, A.A., D.M, T.E., S.F, O.E., E.S., A.V.,

L.T.A.; Resources, A.I., G.Z., D.T.F, B.M., R.J.S., G.M.B., T.T., C.C., L.N.K., K.T.F., Z.W., M.H.,

R.D., M.L.; Writing – Original Draft, A.A., C.D.C., T.E., M.L., W.J.K.; Writing – Review and Editing,

A.A., C.D.C., T.E., S.F., G.Z., M.L., W.J.K.; Visualization, A.A., C.D.C., N.Z., W.J.K.; Supervision,

A.A., W.K., W.J.K.; Project administration, A.A., W.K., W.J.K.; Funding acquisition, N.Z., M.H., M.L.,

W.K., W.J.K.

REFERENCES

1. Balch CM, Gershenwald JE, Soong SJ, Thompson JF, Atkins MB, Byrd DR, et al. Final version of 2009 AJCC melanoma staging and classification. J Clin Oncol 2009;27(36):6199-206 doi 10.1200/JCO.2009.23.4799. 2. Davies H, Bignell GR, Cox C, Stephens P, Edkins S, Clegg S, et al. Mutations of the BRAF gene in human cancer. Nature 2002;417(6892):949-54 doi 10.1038/nature00766. 3. Caenepeel S, Cooke K, Wadsworth S, Huang G, Robert L, Moreno BH, et al. MAPK pathway inhibition induces MET and GAB1 levels, priming BRAF mutant melanoma for rescue by hepatocyte growth factor. Oncotarget 2017;8(11):17795-809 doi 10.18632/oncotarget.14855. 4. Wellbrock C, Karasarides M, Marais R. The RAF proteins take centre stage. Nat Rev Mol Cell Biol 2004;5(11):875-85 doi 10.1038/nrm1498. 5. Tsai J, Lee JT, Wang W, Zhang J, Cho H, Mamo S, et al. Discovery of a selective inhibitor of oncogenic B-Raf kinase with potent antimelanoma activity. Proc Natl Acad Sci U S A 2008;105(8):3041-6 doi 10.1073/pnas.0711741105. 6. Gopal YN, Deng W, Woodman SE, Komurov K, Ram P, Smith PD, et al. Basal and treatment-induced activation of AKT mediates resistance to cell death by AZD6244 (ARRY-142886) in Braf-mutant human cutaneous melanoma cells. Cancer Res 2010;70(21):8736-47 doi 10.1158/0008-5472.CAN-10-0902. 7. Chapman PB, Hauschild A, Robert C, Haanen JB, Ascierto P, Larkin J, et al. Improved survival with vemurafenib in melanoma with BRAF V600E mutation. N Engl J Med 2011;364(26):2507-16 doi 10.1056/NEJMoa1103782. 8. Larkin J, Ascierto PA, Dreno B, Atkinson V, Liszkay G, Maio M, et al. Combined vemurafenib and cobimetinib in BRAF-mutated melanoma. N Engl J Med 2014;371(20):1867-76 doi 10.1056/NEJMoa1408868. 9. Chan MM, Haydu LE, Menzies AM, Azer MW, Klein O, Lyle M, et al. The nature and management of metastatic melanoma after progression on BRAF inhibitors: effects of extended BRAF inhibition. Cancer 2014;120(20):3142-53 doi 10.1002/cncr.28851. 10. Lo RS, Shi H. Detecting mechanisms of acquired BRAF inhibitor resistance in melanoma. Methods Mol Biol 2014;1102:163-74 doi 10.1007/978-1-62703-727-3_10. 11. Shi H, Hugo W, Kong X, Hong A, Koya RC, Moriceau G, et al. Acquired resistance and clonal evolution in melanoma during BRAF inhibitor therapy. Cancer Discov 2014;4(1):80-93 doi 10.1158/2159-8290.CD-13-0642. 12. Hugo W, Shi H, Sun L, Piva M, Song C, Kong X, et al. Non-genomic and Immune Evolution of Melanoma Acquiring MAPKi Resistance. Cell 2015;162(6):1271-85 doi 10.1016/j.cell.2015.07.061. 13. Shaffer SM, Dunagin MC, Torborg SR, Torre EA, Emert B, Krepler C, et al. Rare cell variability and drug-induced reprogramming as a mode of cancer drug resistance. Nature 2017;546(7658):431-5 doi 10.1038/nature22794. 14. Konieczkowski DJ, Johannessen CM, Abudayyeh O, Kim JW, Cooper ZA, Piris A, et al. A melanoma cell state distinction influences sensitivity to MAPK pathway inhibitors. Cancer Discov 2014;4(7):816-27 doi 10.1158/2159-8290.CD-13-0424. 15. Parmenter TJ, Kleinschmidt M, Kinross KM, Bond ST, Li J, Kaadige MR, et al. Response of BRAF-mutant melanoma to BRAF inhibition is mediated by a network of transcriptional regulators of glycolysis. Cancer Discov 2014;4(4):423-33 doi 10.1158/2159-8290.CD-13-0440.

16. Haq R, Shoag J, Andreu-Perez P, Yokoyama S, Edelman H, Rowe GC, et al. Oncogenic BRAF regulates oxidative metabolism via PGC1alpha and MITF. Cancer Cell 2013;23(3):302-15 doi 10.1016/j.ccr.2013.02.003. 17. Vazquez F, Lim JH, Chim H, Bhalla K, Girnun G, Pierce K, et al. PGC1alpha expression defines a subset of human melanoma tumors with increased mitochondrial capacity and resistance to oxidative stress. Cancer Cell 2013;23(3):287-301 doi 10.1016/j.ccr.2012.11.020. 18. Zhang G, Frederick DT, Wu L, Wei Z, Krepler C, Srinivasan S, et al. Targeting mitochondrial biogenesis to overcome drug resistance to MAPK inhibitors. J Clin Invest 2016;126(5):1834-56 doi 10.1172/JCI82661. 19. Raaijmakers MI, Widmer DS, Maudrich M, Koch T, Langer A, Flace A, et al. A new live-cell biobank workflow efficiently recovers heterogeneous melanoma cells from native biopsies. Exp Dermatol 2015;24(5):377-80 doi 10.1111/exd.12683. 20. Liao Y, Smyth GK, Shi W. featureCounts: an efficient general purpose program for assigning sequence reads to genomic features. Bioinformatics 2014;30(7):923-30 doi 10.1093/bioinformatics/btt656. 21. Robinson MD, McCarthy DJ, Smyth GK. edgeR: a Bioconductor package for differential expression analysis of digital gene expression data. Bioinformatics 2010;26(1):139-40 doi 10.1093/bioinformatics/btp616. 22. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods 2001;25(4):402-8 doi 10.1006/meth.2001.1262. 23. Knobloch M, Pilz GA, Ghesquiere B, Kovacs WJ, Wegleiter T, Moore DL, et al. A Fatty Acid Oxidation-Dependent Metabolic Shift Regulates Adult Neural Stem Cell Activity. Cell Rep 2017;20(9):2144-55 doi 10.1016/j.celrep.2017.08.029. 24. Di Veroli GY, Fornari C, Wang D, Mollard S, Bramhall JL, Richards FM, et al. Combenefit: an interactive platform for the analysis and visualization of drug combinations. Bioinformatics 2016;32(18):2866-8 doi 10.1093/bioinformatics/btw230. 25. Kuroda H, Kutner RH, Bazan NG, Reiser J. Simplified lentivirus vector production in protein-free media using polyethylenimine-mediated transfection. J Virol Methods 2009;157(2):113-21 doi 10.1016/j.jviromet.2008.11.021. 26. Yin H, Flynn AD. Drugging Membrane Protein Interactions. Annu Rev Biomed Eng 2016;18:51-76 doi 10.1146/annurev-bioeng-092115-025322. 27. Rambow F, Rogiers A, Marin-Bejar O, Aibar S, Femel J, Dewaele M, et al. Toward Minimal Residual Disease-Directed Therapy in Melanoma. Cell 2018;174(4):843-55 e19 doi 10.1016/j.cell.2018.06.025. 28. Pascual G, Avgustinova A, Mejetta S, Martin M, Castellanos A, Attolini CS, et al. Targeting metastasis-initiating cells through the fatty acid receptor CD36. Nature 2017;541(7635):41-5 doi 10.1038/nature20791. 29. Glatz JFC, Luiken J. Dynamic role of the transmembrane glycoprotein CD36 (SR-B2) in cellular fatty acid uptake and utilization. J Lipid Res 2018;59(7):1084-93 doi 10.1194/jlr.R082933. 30. Hangauer MJ, Viswanathan VS, Ryan MJ, Bole D, Eaton JK, Matov A, et al. Drug- tolerant persister cancer cells are vulnerable to GPX4 inhibition. Nature 2017;551(7679):247-50 doi 10.1038/nature24297. 31. Boiko AD, Razorenova OV, van de Rijn M, Swetter SM, Johnson DL, Ly DP, et al. Human melanoma-initiating cells express neural crest nerve growth factor receptor CD271. Nature 2010;466(7302):133-7 doi 10.1038/nature09161.

32. Ye H, Adane B, Khan N, Sullivan T, Minhajuddin M, Gasparetto M, et al. Leukemic Stem Cells Evade Chemotherapy by Metabolic Adaptation to an Adipose Tissue Niche. Cell Stem Cell 2016;19(1):23-37 doi 10.1016/j.stem.2016.06.001. 33. Banks AS, McAllister FE, Camporez JP, Zushin PJ, Jurczak MJ, Laznik-Bogoslavski D, et al. An ERK/Cdk5 axis controls the diabetogenic actions of PPARgamma. Nature 2015;517(7534):391-5 doi 10.1038/nature13887. 34. Bost F, Aouadi M, Caron L, Even P, Belmonte N, Prot M, et al. The extracellular signal-regulated kinase isoform ERK1 is specifically required for in vitro and in vivo adipogenesis. Diabetes 2005;54(2):402-11. 35. Zhang M, Di Martino JS, Bowman RL, Campbell NR, Baksh SC, Simon-Vermot T, et al. Adipocyte-Derived Lipids Mediate Melanoma Progression via FATP Proteins. Cancer Discov 2018;8(8):1006-25 doi 10.1158/2159-8290.CD-17-1371. 36. Houten SM, Wanders RJ. A general introduction to the biochemistry of mitochondrial fatty acid beta-oxidation. J Inherit Metab Dis 2010;33(5):469-77 doi 10.1007/s10545- 010-9061-2. 37. Ibrahimi A, Sfeir Z, Magharaie H, Amri EZ, Grimaldi P, Abumrad NA. Expression of the CD36 homolog (FAT) in fibroblast cells: effects on fatty acid transport. Proc Natl Acad Sci U S A 1996;93(7):2646-51. 38. Rozovski U, Harris DM, Li P, Liu Z, Jain P, Ferrajoli A, et al. STAT3-activated CD36 facilitates fatty acid uptake in chronic lymphocytic leukemia cells. Oncotarget 2018;9(30):21268-80 doi 10.18632/oncotarget.25066. 39. Xu S, Jay A, Brunaldi K, Huang N, Hamilton JA. CD36 enhances fatty acid uptake by increasing the rate of intracellular esterification but not transport across the plasma membrane. Biochemistry 2013;52(41):7254-61 doi 10.1021/bi400914c. 40. Son NH, Basu D, Samovski D, Pietka TA, Peche VS, Willecke F, et al. Endothelial cell CD36 optimizes tissue fatty acid uptake. J Clin Invest 2018;128(10):4329-42 doi 10.1172/JCI99315. 41. Murray AJ, Montgomery HE, Feelisch M, Grocott MPW, Martin DS. Metabolic adjustment to high-altitude hypoxia: from genetic signals to physiological implications. Biochem Soc Trans 2018;46(3):599-607 doi 10.1042/BST20170502. 42. Nsiah-Sefaa A, McKenzie M. Combined defects in oxidative phosphorylation and fatty acid beta-oxidation in mitochondrial disease. Biosci Rep 2016;36(2) doi 10.1042/BSR20150295. 43. Camarda R, Zhou AY, Kohnz RA, Balakrishnan S, Mahieu C, Anderton B, et al. Inhibition of fatty acid oxidation as a therapy for MYC-overexpressing triple-negative breast cancer. Nat Med 2016;22(4):427-32 doi 10.1038/nm.4055. 44. Tan Z, Xiao L, Tang M, Bai F, Li J, Li L, et al. Targeting CPT1A-mediated fatty acid oxidation sensitizes nasopharyngeal carcinoma to radiation therapy. Theranostics 2018;8(9):2329-47 doi 10.7150/thno.21451. 45. Schlaepfer IR, Glode LM, Hitz CA, Pac CT, Boyle KE, Maroni P, et al. Inhibition of Lipid Oxidation Increases Glucose Metabolism and Enhances 2-Deoxy-2- [(18)F]Fluoro-D-Glucose Uptake in Prostate Cancer Mouse Xenografts. Mol Imaging Biol 2015;17(4):529-38 doi 10.1007/s11307-014-0814-4. 46. Ratnikov BI, Scott DA, Osterman AL, Smith JW, Ronai ZA. Metabolic rewiring in melanoma. Oncogene 2017;36(2):147-57 doi 10.1038/onc.2016.198. 47. Li B, Li X, Ni Z, Zhang Y, Zeng Y, Yan X, et al. Dichloroacetate and metformin synergistically suppress the growth of ovarian cancer cells. Oncotarget 2016;7(37):59458-70 doi 10.18632/oncotarget.10694. 48. Marini C, Bianchi G, Buschiazzo A, Ravera S, Martella R, Bottoni G, et al. Divergent targets of glycolysis and oxidative phosphorylation result in additive effects of

metformin and starvation in colon and breast cancer. Sci Rep 2016;6:19569 doi 10.1038/srep19569. 49. Ho PC, Bihuniak JD, Macintyre AN, Staron M, Liu X, Amezquita R, et al. Phosphoenolpyruvate Is a Metabolic Checkpoint of Anti-tumor T Cell Responses. Cell 2015;162(6):1217-28 doi 10.1016/j.cell.2015.08.012. 50. Zhang Y, Kurupati R, Liu L, Zhou XY, Zhang G, Hudaihed A, et al. Enhancing CD8(+) T Cell Fatty Acid Catabolism within a Metabolically Challenging Tumor Microenvironment Increases the Efficacy of Melanoma Immunotherapy. Cancer Cell 2017;32(3):377-91 e9 doi 10.1016/j.ccell.2017.08.004.

FIGURE LEGENDS

Figure 1. Cell surface marker screening and validation of differentially expressed cell surface protein during MAPK inhibition. A, Differentially expressed cell surface proteins in A375P cells after 36 and

60 h of PLX treatment. B, Gene expression of cell surface proteins that are consistently up- or downregulated in 2 melanoma cell lines (A375P, UACC62) and 4 early passage melanoma cultures

(M980513, M080326, M130820, M130309) after PLX treatment for 36 h. Each value represents the amount of mRNA relative to that in cell lines incubated with DMSO, which was arbitrarily defined as

1. Data are mean  SD (n = 3). TBP or cyclophilin were used as the invariant control. Statistical analysis was performed using Student’s t test. *, p < 0.05; **, p < 0.01; ***, p < 0.001. C, Expression of CD36 measured by RNA-seq in 11 BRAF-mutated melanoma patients from the Massachusetts

General Hospital (MGH). Each value represents the amount of mRNA in tumor biopsies of patients treated for 14-16 days relative to that prior treatment, which was arbitrarily defined as 1. D,

Expression of CD36 measured by RNA-seq in 8 BRAF-mutated melanoma patients from a second

MGH cohort. Each value represents the amount of mRNA in tumor biopsies of patients on treatment for different periods of time relative to that prior treatment, which was arbitrarily defined as 1. E, Plots depict representative images of FSC-A versus CD36-PE of 2 melanoma cell lines (A375P, UACC62) and 5 early passage melanoma cultures (M980513, M080326, M130309, M130820 and M140122) treated for 36 h either with 1 M of the BRAF inhibitor PLX or 0.5 M of the MEK inhibitor AZD.

DMSO was used as solvent control. Percentage of CD36+ cells is indicated in each dot plot. For each cell line the experiment has been performed at least 3 times. F, Violin plot of single cell RNA- seq data highlighting the distribution of CD36 expression at single cell level in different phases of

MAPKi treatments.

Figure 2. MAPKi induce the expression of genes involved in lipid catabolism in BRAFV600E melanomas. A, Gene expression values obtained from RNA-seq data of A375P cells that were treated with 1 M PLX for 36 h. B-C, mRNA levels of genes involved in fatty acid -oxidation or glycolysis measured using qRT-PCR in A375P (B) and 4 additional melanoma cell lines (C) treated with 1 M PLX or 0.5 M AZD for 36 h. Each value represents the amount of mRNA relative to that in cell lines incubated with DMSO, which was arbitrarily defined as Log2 of 1. TBP and cyclophilin

were used as the invariant control. Data are mean  SD (n = 3). Statistical analysis was performed using one-way Anova test. D-E, Gene expression data measured using qRT-PCR in A375P xenograft tumors from mice that were treated either with DMSO as solvent control or a combination of 30 mg/kg PLX4032 and 15 mg/kg AZD6244 for (D) 36 and (E) 60 h. TBP and cyclophilin were used as the invariant control. Statistical analysis was performed using Student’s t test.

Figure 3. MAPK inhibition decreases glycolysis and increases the rate of fatty acid oxidation in

BRAF-mutated melanoma cells. A, Extracellular acidification rate (ECAR) was determined using a

Seahorse XF24 analyzer to evaluate the glycolytic flux in A375P cells treated with MAPKi for 72 h.

Glycolysis, glycolytic capacity and glycolytic reserve were determined by the sequential addition of

15 mM glucose, 1 M oligomycin and 200 mM 2-D-glucose. Values represent mean ± SD of 4 experiments performed in quintuplicates. Values were normalized to cell number. B, Effect of MAPK inhibition on lactate production in A375P cells. Lactate levels were normalized to integral viable cell number. Data are mean ± SD (n = 2) C, Percentage of dead cells after treatment of A375P cells with

MAPKi in complete or glucose-free medium for 72 h. Data are mean  SD (n = 3). Statistical analysis was performed using one-way Anova test. D, FAO levels measured using radioactive 3H-labeled palmitic acid in BRAF-mutated melanoma cell lines treated with MAPKi for 72 h. Each value represents the rate of FAO relative to that in cell lines incubated with DMSO, which was arbitrarily defined as 1. Data are mean ± SD (n = 3) and were normalized to total protein content. E, FAO rate in the BRAF wild-type melanoma cell line MeWo. F, FAO rate measured in the BRAFV600E melanoma cell line M080326 and in the BRAFV600E/NRASQ61R cell line M121224. nd = not detectable G,

Quantification of viable cells using AnnexinV/PI staining in melanoma cell lines treated with 1 M

PLX and 0.5 M AZD for 5 days. Statistical significance was analysed by one-way Anova test.

Figure 4. CD36+ melanoma cells have a higher rate of FAO. A, DMSO- and MAPKi-treated (1 M

PLX and 0.5 M AZD) cells were sorted into CD36+ and CD36- subpopulations after 36 h. An unsorted fraction was taken as control. The 3 fractions were replated in DMSO- or MAPKi-containing media and FAO was measured the next day. B, Radioactive FAO measurements using 3H-labeled palmitic acid. Each value represents the rate of FAO relative to that in cell lines incubated with

DMSO, which was arbitrarily defined as 1. Data are mean ± SD (n = 3). C, Correlation between

changes in gene expression in patients on-treatment with MAPKi compared to pre-treatment. The expression changes of CPT1A and CD36 were computed using reads per kilobase of transcript per million mapped reads (RPKM) estimates for each gene before and after treatment. CD36 fold changes were plotted against CPT1A fold changes and correlation was evaluated using Pearson’s rho rank correlation. D, FAO levels measured in A375P cells treated with MAPKi alone or in combination with 3 M lipofermata for 72 h. Each value represents the rate of FAO relative to that in cells incubated with DMSO, which was arbitrarily defined as 1. Data are mean ± SD (n = 2) and were normalized to total protein content.

Figure 5. MAPKi increase FAO through PPAR-mediated CPT1A induction. A, FAO measurements via radioactively labelled 3H palmitic acid in A375P, UACC62 and M980513 cells with PPARA knockdown (KD) and treated with the indicated MAPKi for 72 h. PPARA was knocked down using 2 different siRNAs transfected 24 h before treatment. Each value represents the rate of FAO relative to that in cell lines incubated with DMSO and transfected with scrambled siRNA, which was arbitrarily defined as 1. Data are mean ± SD (n = 3). B, CPT1A expression in A375P, UACC62 and M980513 cells with PPARA KD and treated with the indicated MAPKi for 36 h. Each value represents the amount of mRNA relative to that in cell lines incubated with DMSO and transfected with scrambled siRNA 24 h before the begin of the treatment. Data are mean ± SEM (n = 3). C, FAO measurements in A375P, UACC62 and M980513 cells treated with the indicated MAPKi and the PPAR antagonist

GW6471 for 72 h. Each value represents the rate of FAO relative to that in cell lines incubated with

DMSO, which was arbitrarily defined as 1. Data are mean ± SD (n = 2). D, CPT1A expression in

A375P cells treated with the indicated MAPKi and PPAR antagonist GW6471 (50 M) for 36 h.

Each value represents the amount of mRNA relative to that in cell lines incubated with DMSO. Data are mean ± SD (n = 3). Statistical analysis was performed using one-way Anova test. E, FAO measurements in A375P and M980513 cells treated with the indicated MAPKi and PPAR agonist

GW7647 (100 nM) for 72 h. Each value represents the rate of FAO relative to that in cell lines incubated with DMSO, which was arbitrarily defined as 1. Data are mean ± SD (n = 2). F, CPT1A expression in A375P and M980513 cells treated with the indicated MAPKi and PPAR agonist

GW7647 for 36 h. Each value represents the amount of mRNA relative to that in cell lines incubated

with DMSO. Data are mean ± SD (n = 3). Statistical analysis was performed using one-way Anova test.

Figure 6. Role of glycolysis and FAO inhibitors in proliferation and resistance to MAPKi. A, Glycolytic flux, indicated as ECAR of A375P cells untreated or treated with indicated MAPKi and with or without

50 M ETO, was measured by a Seahorse XF24 Analyzer. Data are mean ± SD (n = 3). B, Glycolytic flux of untreated or PLX-treated A375P WT and CPT1A KD cells. Data are mean ± SD (n = 2). C,

Percentage of dead cells after treatment of A375P cells with the indicated MAPKi alone or in combination with 50 M ETO in glucose-free medium. Data are mean  SD (n = 2). Statistical analysis was performed using one-way Anova test. D, BLISS interaction analyses of glycolysis, FAO, and MAPK inhibitors in BRAFV600E cell lines (n ≥ 2 for each cell line). E, FAO measurements in A375P cells treated with the indicated MAPKi alone or in combination with 50 M ETO, 40 mM DCA or a combination of ETO and DCA for 72 h. Data are mean ± SD of 3 biological replicates. F, Tumor volume of A375P xenografts treated with the vehicle control or 100 mg/kg DCA or 32 mg/kg ETO or a combination of DCA+ETO. Mice were xenografted in both flanks with A375P cells. Data are mean ±

SEM of at least 8 tumors. G, Tumor volume of A375P xenografts treated with 24 mg/kg PLX4032 and

12 mg/kg AZD6244 (MAPKi) alone or in combination with 32 mg/kg ETO. H, Tumor volume of A375P xenografts treated with 24 mg/kg PLX4032 and 12 mg/kg AZD6244 (MAPKi), either alone or in combination with 100 mg/kg DCA or in combination with 100 mg/kg DCA and 32 mg/kg ETO.

Statistical analysis was performed using one-way Anova test. I, Summary of the metabolic reprogramming in the initial response phase of MAPKi treatment. MAPKi-treated cells have a decreased glycolytic flux and display increased levels of CD36 and PPAR-mediated and CPT1A- dependent FAO. The concomitant treatment with MAPK and FAO inhibitors is not sufficient to impair tumor growth due to re-activation of glycolysis. The triple combination of MAPK, FAO and glycolytic inhibitors induces apoptosis and reduces tumor growth in vivo.

Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2019 American Association for Cancer Research. Author Manuscript Published OnlineFirst on August 2, 2019; DOI: 10.1158/1078-0432.CCR-19-0253 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Figure 1 CD24 CD36 A B 12 900 *** Down-regulated Up-regulated DMSO 600 ** * CD28 CD9 pre ss ion 8 50 PLX *** CD41 CD24 ** *** 4 ** 10 CD49c CD29 ex gene 8 * *** *** CD49e CD32 4 0 0 CD49f CD36 tive Rela tive

CD55 CD49a A375P A375P UACC62M980513M080326M130820M130309 UACC62M980513M080326M130820M130309 CD57 CD49d CD166 CD271 CD61 CD59 45 * 25 30 15 CD73 CD74 *** pre ss ion * * * CD96 CD107a 15 10 10 ** 3 CD108 CD132 ** ex gene *** 2 CD137L CD138 5 * *** 1 CD156c CD143 0 0

CD201 CD146 Rela tive A375P A375P CD203a CD162 UACC62M980513M080326M130820M130309 UACC62M980513M080326M130820M130309 CD221 CD166 CD96 CD61 CD252 CD245 1.0 1.2 CD257 CD253 * 0.8 pre ss ion ** CD276 CD271 *** 0.5 *** *** *** EGFR CD294 *** *** 0.4 ex gene *** MSC CD340 MICA/B CD344 0.0 0.0

Notch4 DLL1 Rela tive A375P A375P NPC HLA DQ-DR UACC62M980513M080326M130820M130309 UACC62M980513M080326M130820M130309

C D F Pre-treatment Adaptation phase 7 15 Drug-tolerant phase Acquired resistance phase 6 10 pre ss ion 5 6

expres sion * 4 *** 4 3 12 *** ex mean mean 2 2

CD36 1 CD36 8 0 0 2 6 7 9

38 35 Nor mExp 12 13 19 24 34 10 16 43 BI1 200 409 PPP IPIPD 4 tive Rela tive tive Rela tive MGH Cohort #1 patient number MGH Cohort #2 patient number

Pre-treatment Early-on treatment 0 CD36 single cell expression E (Rambow et al., 2018) A375P UACC62 M980513 M080326 M130309 M130820 M140122

5 5 5 5 5 5 5 10 0.16% 10 21.8% 10 15.1% 10 24.0% 10 11.0% 10 37.6% 10 18.1%

4 4 4 4 4 4 4 10 10 10 10 10 10 10

3 3 3 3 3 3 3 10 10 10 10 10 10 10 DMSO 0 0 DMSO 0 0 0 0 0

3 3 3 3 3 3 3 -10 -10 -10 -10 -10 -10 -10

0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K

5 5 5 5 5 5 5 10 63.9% 10 46.4% 10 40.3% 10 70.1% 10 26.9% 10 53.4% 10 65.4%

4 4 4 4 4 4 4 10 10 10 10 10 10 10

EP-63DC 3 3 3 3 3 3 3 10 10 10 10 10 10 PLX 10 0 0 PLX 0 0 0 0 0

3 3 3 3 3 3 3 -10 -10 -10 -10 -10 -10 -10

0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K

5 5 5 5 5 5 5 10 92.6% 10 70.8% 10 40.3% 10 79.9% 10 37.6% 10 70.4% 10 75.8%

4 4 4 4 4 4 4 10 10 10 10 10 10 10

3 3 3 3 3 3 3 10 10 10 10 10 10 10

50K 100K 150K 200K 250K

AZD 0 0 0 0 0 0 0

3 3 3 0 3 3 3 3 -10 -10 -10 -10 -10 -10 -10

0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K 0 50K 100K 150K 200K 250K

Downloaded from clincancerres.aacrjournals.orgFSC-A on September 28, 2021. © 2019 American Association for Cancer Research. Author Manuscript Published OnlineFirst on August 2, 2019; DOI: 10.1158/1078-0432.CCR-19-0253 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited. Figure 2 A DMSO PLX B

CD36 CPT1A PPARA A375P 10 11.5 9.0

11.4 ) 8 2 14 PLX 11.3 8.8 12 AZD 6 11.2 8.6 10

Log2 (Counts) 11.1 8

4 (Log re ssion 11.0 8.4 6 * * ** DMSO PLX DMSO PLX DMSO PLX 4 ** *** *** ** ** ** 2 *

EHHADH PGC1A ACADS *** *** *** *** *** *** 0 5.2 8 -2 8.25 7

4.8 exp gene Relative -4 8.00 6 4.4 HK2 CD36 ECH1 LPIN1 LDHA 5 CPT1APPARAPGC1AACADS ACAT1 GLUT3 7.75 ACAD10EHHADH 4.0 4 Log2 (Counts) 7.50 Lipid metabolism Glycolysis

DMSO PLX DMSO PLX DMSO PLX C UACC62 M980513 ) 2 6 6 * ** **

4 4 *** * ** *** *** * * ** *** ** *** * *** * *** *** *** *** *** *** *** *** ** ** ** sion (Log res sion ** ** *** ** 2 2 *** *** * *** ** *** *** *** *** *** 0 0

-2 *** -2 *** *** *** *** ***

Relative gene exp gene Relative -4 -4

HK2 HK2 CD36 ECH1 LPIN1 LDHA CD36 ECH1 LPIN1 LDHA CPT1APPARAPGC1AACADS ACAT1 GLUT3 CPT1APPARAPGC1AACADS ACAT1 GLUT3 ACAD10EHHADH ACAD10EHHADH

) M130820 M140122 2 6 6 ** *

4 ** 4 ** * ** *** * *** * ** *** * * ** *** *** *** ** *** *** ** * * *** ***

2 2 *** *** ** sion (Log res sion ** * * ** ** * * ** ** *** *** * *** *** *** 0 0

-2 -2

Relative gene exp gene Relative -4 -4

HK2 HK2 CD36 ECH1 LPIN1 LDHA CD36 ECH1 LPIN1 LDHA CPT1APPARAPGC1AACADS ACAT1 GLUT3 CPT1APPARAPGC1AACADS ACAT1 GLUT3 ACAD10EHHADH ACAD10EHHADH Lipid metabolism Glycolysis Lipid metabolism Glycolysis D E A375P xenograft 36 h A375P xenograft 60 h 0.4 3 0.010 0.010 ** ** ** * 0.3 2.0 * 2 0.2 0.005 0.2 0.005 1.0 1 0.1 0.000 0.0 0 0.000 0.0 0.0 CD36 PPARA CPT1A CD36 PPARA CPT1A io n

res sion 0.20 8 3 0.10 3 *** * ** * 6 * 2 expres s 4 2 0.10 4 0.05 1 1 2 0 0.00 0 gen e 0 0.00 0 ACAD10 ACSBG1 HK2 ACAD10 ACSBG1 HK2 tive Rela tive Relative gene exp gene Relative 150 15 6 15 10 * * * * 150 ** 100 10 4 100 10 5 50 5 2 50 5 0 0 0 0 0 0 GLUT3 LDHA DUSP6 GLUT3 LDHA DUSP6 DMSO PLX+AZD Downloaded from clincancerres.aacrjournals.org on September 28, 2021. © 2019 American Association for Cancer Research. Author Manuscript Published OnlineFirst on August 2, 2019; DOI: 10.1158/1078-0432.CCR-19-0253 Author manuscripts have been peer reviewed and accepted for publication but have not yet been edited.

Figure 3 A A375P B C Oligomycin A375P A375P A375P Glucose 2-D-Glucose ** 20 25 *** 100 ** ** *** Lactate

10 K c ell s) 20 ** 150 80 15 *** 10 K c ell s) 15 ro l) 60 100 *** con t

ns produ ct io n 40 **

10 f 10 ns s decll aeD )%( ns 50 5 20 o (% 0 5 Lac tate 0 0 ECAR (pMoles/min/ ECAR PLX dium ECAR (pMoles/min/ ECAR 0 DMSO Medium 0 40 80 Glycolysis PLX+AZD ree Me Minutes Complete DMSO PLX PLX+AZD Glycolytic Reserve DMSO PLX PLX+AZD Glycolytic Capacity Glucose-f D BRAFV600E

A375P UACC62 SK-MEL28 M980513 M130820 M140122 4 5.0 3 7.0 6 3 ** * *** *** ** * ** * *** ** ** * FAO ra te 2 4 2 2 2.5 3.5 1 2 1 rm alized

No 0 0 0 0 0 0

PLX PLX PLX PLX PLX PLX DMSO X+AZD DMSO X+AZD DMSO X+AZD DMSO X+AZD DMSO X+AZD DMSO X+AZD PL PL PL PL PL PL

E F G Increased FAO Stable FAO

100 BRAFWT BRAFV600E BRAFV600E PLX-resistant

MeWo M080326 M121224 (%)

3 2 2 ell s ** 50 ns 2 iable c FAO ra te FAO ra te V 1 1 1 rm alized rm alized nd No No 0 0 0 0

PLX PLX PLX DMSO X+AZD DMSO X+AZD DMSO X+AZD PL PL PL A375P A375P UACC62 UACC62 SKMel28 SKMel28 M980513 M130820 M140122 M080326 M121224 M980513 M130820 M140122 M080326 M121224 DMSO PLX+AZD

Figure 4

Overnight Overnight A DMSO 36 h incubation in MAPKi 36 h incubation in DMSO medium MAPKi medium

Cells Cells Unsorted Unsorted

Melanoma Sorting Cells Melanoma Sorting Cells Cells CD36+ FAO assay Cells CD36+ FAO assay

Cells Cells CD36- CD36-

B UACC62 M130820 M140122 4 ** 5 2.0 *** ** *** 4 ** ** 3 1.5 * 3 rate FAO 2 1.0 ns 2 p=0.068 1 0.5 1 Normal ized 0 0 0.0

DMSO CD36- DMSO CD36+DMSO CD36- MAPKi CD36+MAPKi CD36- DMSO CD36+ MAPKi CD36+MAPKi CD36- DMSO CD36+DMSO CD36- MAPKi CD36+MAPKi CD36- DMSO Unsorted MAPKi Unsorted DMSO Unsorted MAPKi Unsorted DMSO Unsorted MAPKi Unsorted

C D A375P Pearson r = 0.407, p = 0.03728, test if r > 0 3 ** 4 Patient IPIPD1001

Patient 6 Patient 200 ra te ** Patient 35 Patient 19 2 Patient 9 Patient 34 Patient 38 2

Patient 409 AO Patient 42 Patient 7 Patient PPP Patient BI1 Patient 13 F ed ns Patient 12 Patient 16 0 Patient 24 1 Patient 43 Patient 10 Normali z −2 Patient 2 0 Log2 (Fold change) CD36 PLX DMSO −4 PLX+AZD −2 −1 0 1 2 Log2 (Fold change) CPT1A PLX+LIPOFERMATA DMSO+LIPOFERMATA PLX+AZD+LIPOFERMATA

Figure 5

A375P UACC62 M980513 A ns 5 ns 6 4 ** * * * 4 ** * ** * ** 3 ** 4 3 2 ed FAO ra te 2 ** 2 ns * ns 1 ** 1 * Normali z 0 0 0 Control Control Control Control Control Control Control Control Control siControl siControl siControl siControl siControl siControl siControl siControl siControl siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#1 siPPARA#2 siPPARA#2 DMSO PLX PLX+AZD DMSO PLX PLX+AZD DMSO PLX PLX+AZD B A375P UACC62 M980513 3 3 4 ns ns ** * * ns * * ** 3 pre ss ion 2 2 *** *** ** ** ns 2 ns 1 *** 1 * *** 1 CPT1Aed ex

0 0 0 Normali z Control Control Control Control Control Control Control Control Control siControl siControl siControl siControl siControl siControl siControl siControl siControl siPPARA#1 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#2 siPPARA#1 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#1 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#2 siPPARA#1 siPPARA#2 siPPARA#2 DMSO PLX PLX+AZD DMSO PLX PLX+AZD DMSO PLX PLX+AZD C A375P UACC62 M980513 5 6 2.0 *** *** *** *** 4 *** ** *** *** *** 1.5 *** *** 4 3 *** ** ***

ed FAO ra te 1.0 2 ns ** 2 ns 1 * 0.5 Normali z 0 0 0.0 Control Control Control Control Control Control Control Control Control 50 µM GW6471 50 µM GW6471 50 µM GW6471 50 µM GW6471 50 µM GW6471 50 µM GW6471 50 µM GW6471 50 µM GW6471 50 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 100 µM GW6471 DMSO PLX PLX+AZD DMSO PLX PLX+AZD DMSO PLX PLX+AZD

D E F A375P A375P M980513 A375P M980513 PPAR antagonist PPAR agonist PPAR agonist PPAR agonist PPAR agonist si on si on 4 6 3 *** 8 *** 6 * ** pre s pre s *** *** *** ** 6 ns 2 ex ex 4 4 A A 2 4 ns * * ns ns 2 1 PT 1 PT 1 2 2 *

0 0 0 0 0 Normalized FAO ra te Control Control Control Control Control Control Control Control Control Control Control Control Control Control Control C Normalized C Normalized GW6471 GW6471 GW6471 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647 GW7647

DMSO PLX PLX+AZD DMSO PLX PLX+AZD DMSO PLX PLX+AZD DMSO PLX PLX+AZD DMSO PLX PLX+AZD

Figure 6 A B C A375P A375P DMSO A375P PLX A375P in glucose-free medium 100 25 *** p=0.0789 30 3 ns 20 * * 80 *** el ls

10 K c ell s) * 15 10 K c ell s) 20 2 60 *** 10 ** * 40 10 1 dead c 5 % 20 (pMole s/ min /

0 (pMole s/ min / 0 0 0

PLX ECAR ECAR

DMSO ECAR PLX+ETOPLX+AZD siControl siControl PLX+ETOPLX+AZD DMSO+ETO siCPT1A #1siCPT1A #2 siCPT1A #1siCPT1A #2 DMSO+ETOPLX Control PLX+AZD+ETO DMSO Control PLX+AZD+ETO

E 10 A375P D 8 *** BLISS ra te **

A375P UACC62 M980513 M130820 M140122 AO 6 Etomoxir [µM] F ed 12 25 50 100 12 25 50 100 12 25 50 100 12 25 50 100 12 25 50 100 4 * 11 8 12 -3 -3 1 15 3 1 -1 6 25 16 15 6 5 2 3 5 13 7 2

10 1 4 8 5 -1 6 2 13 -2 0 6 31 1 0 0 13 6 6 12 8 Normali z 0 20 4 0 5 11 2 0 3 12 -2 -1 12 34 2 0 4 23 -1 -2 -9 4 DCA [mM] ETO 4 7 20 20 7 6 14 19 0 3 8 18 ETO ETO DCA 40 0 1 6 27 5 6 7 10 DCA DCA Control Control Antagonism Synergy Control ETO+DCA ETO+DCA ETO+DCA DMSO PLX PLX+AZD F G H

900 900 600 )

3 ** *

600 600 ** * 400

** ** *

ume (mm ol ume ** 300 300 200 Tumor v Tumor

0 0 0 0 5 10 15 20 25 0 10 20 30 40 0 10 20 30 40 days days days

Control ETO MAPKi MAPKi+ETO MAPKi MAPKi+DCA MAPKi+DCA+ETO DCA DCA+ETO

I Glycolysis inhibitors

V600E V600E BRAF BRAFV600E BRAF MAPK MAPK MAPK inhibitors Glycolysis inhibitors Glycolysis inhibitors Glycolysis MEK MEK + MEK + + PPARα PPARα ERK ERK PPARα ERK + + + + FAO + CPT1A + FAO + CPT1A CPT1A inhibitors + inhibitors FAO CD36 CD36 FAO CD36 FAO

Initial response phase of MAPKi treatment Response to MAPKi and FAOi Cell death

A fatty acid oxidation-dependent metabolic shift regulates the adaptation of BRAF-mutated melanoma to MAPK inhibitors

Andrea Aloia, Daniela Müllhaupt, Christophe D Chabbert, et al.

Clin Cancer Res Published OnlineFirst August 2, 2019.

Updated version Access the most recent version of this article at: doi:10.1158/1078-0432.CCR-19-0253

Supplementary Access the most recent supplemental material at: Material http://clincancerres.aacrjournals.org/content/suppl/2019/08/02/1078-0432.CCR-19-0253.DC1

Author Author manuscripts have been peer reviewed and accepted for publication but have not yet Manuscript been edited.

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 Subscriptions Department at [email protected].

Permissions To request permission to re-use all or part of this article, use this link http://clincancerres.aacrjournals.org/content/early/2019/08/02/1078-0432.CCR-19-0253. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC) Rightslink site.