Cholesterol Regulates the Tumor Adaptive Resistance to MAPK Pathway Inhibition

Abstract

Although targeted MAPK pathway inhibition has achieved remarkable patient responses in many cancers, the development of resistance has remained a critical challenge. Adaptive tumor response underlies the drug resistance. Furthermore, such bypass mechanisms often lead to the activation of many pro-survival kinases, which complicates the rational design of combination therapies. Here we performed global tyrosine phosphoproteomic (pTyr) analyses and demonstrated that targeted MAPK signaling inhibition in melanoma leads to a profound remodeling of the pTyr proteome. Intriguingly, altered cholesterol metabolism might drive, in a coordinated fashion, the activation of these kinases. Indeed, we found an accumulation of intracellular cholesterol in melanoma cells (with BRAF^V600E mutations) and non-small cell lung cancer cells (with KRAS^G12C mutations) treated with MAPK and KRAS^G12C inhibitors, respectively. Importantly, depletion of cholesterol not only prevents the feedback activation of pTyr signaling but also enhances the cytotoxic effects of MAPK pathway inhibitors, both in vitro and in vivo. Together, our findings suggest that cholesterol contributes to the tumor adaptive response upon targeted MAPK pathway inhibitors. These results also suggest that MAPK pathway inhibitors could be combined with cholesterol-lowering agents to achieve a more complete and durable response in tumors with hyperactive MAPK signaling.

PRIDE identifier: PXD021877.

Graphical Abstract

Introduction

Aberrant activation of the Ras-BRAF-MEK-ERK pathway underlies the pathogenesis of many human malignancies. Genetic mutation events (e.g., BRAF/Nras mutations in melanoma and Kras mutations in non-small cell lung cancer) result in the constitutive activation of this mitogen-activated protein kinase (MAPK) signaling pathway, which leads to uncontrolled cell proliferation, evasion of cell death, and eventually, oncogenic transformation. The identification of these genetic alterations provides the strong rationale to target these unique tumor-acquired vulnerabilities, with multiple MAPK pathway inhibitors recently approved by the FDA (e.g., BRAF and MEK inhibitors for the treatment of melanoma). Furthermore, MAPK pathway inhibitors (e.g., KRAS^G12C inhibitors) are evaluated, either alone as single agents or in combination with chemo- and radiation-therapies, against a wide variety of other solid tumors.

Although many cancer patients with these actionable mutations initially respond to the targeted therapies, single agent MAPK pathway inhibitors rarely achieve a durable response, and therapeutic resistance almost invariably occurs. Although the underlying mechanisms are complex, a universal theme of the resistant phenotype is the sustained pro-survival signaling in the cells that evade MAPK pathway inhibition ^1,2. This is exemplified by the identification of the “acquired resistance” mechanisms, which include the development of activating MEK mutations (e.g., MEK1^{C121S 3} and MEK2^Q60P4), or the expression of alternative splicing isoforms of BRAF (e.g., p61BRAF^V600E)⁵ in relapsed melanoma patients. In addition to genomic resistance mechanisms, an emerging resistance mechanism to MAPK pathway inhibitors involves the reshaping of the signaling network in tumor cells that allows these cells to adapt to the inhibition of this key survival pathway (termed as “adaptive response”) ⁶. Indeed, increased activation of receptor tyrosine kinases (RTKs), including the platelet-derived growth factor receptor (PDGFR)-β⁷ and insulin-like growth factor-I receptor (IGF-IR)⁸ is frequently observed in post-treatment biopsy samples obtained from melanoma patients. Enhanced phosphotyrosine (p-Tyr) signaling then results in the (re)activation of downstream pathways (e.g., MAPK and the phosphatidylinositol-3-OH kinase (PI(3)K)-AKT), and thereby provides an alternative means to sustain tumor growth under the MAPK-inhibited conditions. Besides in the context of MAPK-inhibited melanoma, the activation of “bypass” signaling pathways is a nearly universal paradigm for targeted therapies in multiple cancer types ⁹, including the activation of EGFR in KRAS^G12C inhibitor-treated non-small cell lung cancer (NSCLC) cells ¹⁰, IGF1R in mTORC1 inhibitor-treated breast cancer cells ¹¹ and FGFR in Met inhibitor-treated leukemia cells ¹²

More recent studies have shown that a common feature of the adaptive tumor response to targeted therapies is often not the activation of one or two kinases, but rather a systematic remodeling of the receptor tyrosine kinome ^13,14. Furthermore, different sets of kinases can be activated, in a context-specific manner, in various tumor cells following pharmacological perturbations ¹⁵. Because of the simultaneous engagement of multiple, and often redundant survival signals, combination therapies involving a targeted therapeutic agent together with an additional kinase inhibitor are less likely to succeed.

A key question is then, upon the treatment of targeted therapies, whether there exists a regulator that controls the concerted activation of multiple RTKs? We envision that a thorough understanding of the molecular nature of such a factor will greatly facilitate the design and implementation of improved therapeutic strategies to achieve more complete and sustained responses in a broad spectrum of solid tumors. Here, using high sensitivity mass spectrometry, we performed an unbiased assessment of the global p-Tyr proteome in melanoma cells treated with various targeted MAPK pathway inhibitors. The results showed that MEK/BRAF inhibitor treatment results in significant changes in the pTyr status of many proteins, leading to adaptive alterations in the pTyr proteome. Intriguingly, subsequent bioinformatic interrogation of our quantitative proteomic dataset pointed to a common cholesterol-binding motif among the activated Tyr kinases, suggesting that altered cholesterol metabolism might be linked to the activation of “bypass” p-Tyr signaling. Indeed, increased cellular cholesterol level was observed in BRAF^V600E melanoma and KRAS^G12C lung cancer cells treated with BRAF and KRAS^G12C inhibitors, respectively. Depletion of cholesterol completely abrogated the activation of p-Tyr signaling in melanoma cells treated with MAPK pathway inhibitors. Finally, we showed that a combination therapy involving MAPK pathway inhibitors and the cholesterol biosynthesis blocker overcame adaptive resistance, leading to improved cytotoxicity in both in vitro and in vivo models of melanoma and NSCLC. Our findings thus point to cholesterol as a contributing factor of the feedback signaling program, which warrants the evaluation of the proposed combinatorial therapeutic approach in future clinical studies.

Materials and Methods

Cell culture

Human melanoma cell lines A375, Mel-juso and non-small cell lung cancer cell lines H2122, SW1573 and A549 were obtained from the American Type Culture Collection (ATCC, Manassas, VA, USA). All cell lines have been DNA fingerprinted using the PowerPlex 1.2 kit (Promega) and were found to be mycoplasma free using the e-Myco kit (Boca Scientific). Cells were cultured as a monolayer in RPMI-1640 medium supplemented with 10% of fetal bovine serum (Invitrogen). Metabolic Labeling/SILAC cell culture was performed as described previously¹¹. Cells were maintained in an incubator with a humidified atmosphere of 5% CO₂ at 37°C.

Antibodies and reagents

Phospho-p44/42 MAPK (Erk1/2) (Thr202/Tyr204) (#9101), ERK, Phospho-Stat3 (Tyr705) (#9145), STAT3, Phospho-p90RSK, Phospho-S6 Ribosomal Protein (Ser235/236), p-MEK, Phospho-IRS-1 (Ser302), IRS1, GAPDH antibodies and the bead-conjugated rabbit anti-phosphotyrosine (P-Tyr-1000) were purchased from Cell signaling technology. AZD6244, PLX4032 were purchased from Selleck. AMG-510, MRTX-1257 were purchased from MedChemExpress (MCE). Atorvastatin (calcium salt) (Item No. 10493) was from Cayman chemical. Insulin, MβCD and FITC-CTXb were from Sigma.

SILAC cell culture

A375 and Mel-Juso cells were grown in light ([¹²C₆¹⁴N₂]Lys, [¹²C₆¹⁴N₄]Arg) and heavy ([¹³C₆¹⁵N₂]Lys, [¹³C₆¹⁵N₄]Arg) RPMI (Cambridge Isotope Labs). Both light and heavy RPMI were supplemented with 10% dialyzed FBS (Invitrogen) to avoid the introduction of light amino acids that are naturally present in the serum. Cells were grown in the corresponding media for 8 passages, at which point an incorporation check was performed. Specifically, heavy cells were isolated, lysed and digested overnight with sequencing-grade trypsin (Promega) at a 1:100 (enzyme/substrate) ratio. Peptides were desalted using home packed SepPak C18 StageTips and were subsequently analysed by LC–MS/MS experiments on an LTQ Velos Pro Orbitrap mass spectrometer (Thermo). The incorporation rate of heavy amino acids ([¹³C₆¹⁵N₂]Lys, [¹³C₆¹⁵N₄]Arg) was found to be around 97% under these conditions. For drug treatment, cells cultured in heavy media were treated with 1 μM PLX4032 or 1 μM AZD6244 for 48 hours, while the light cells were treated with DMSO for the same durations.

Sample preparation for mass spectrometric analysis

The heavy and light cells were lysed in urea buffer (8M urea, 20 mM HEPES pH 7.0, 75 mM β-glycerolphosphate, 1 mM sodium vanadate, 1 mM DTT and 1.5 mM EGTA) and the protein concentration was determined by the Pierce™ BCA Protein Assay Kit (Thermo). Then the lysates were combined at a 1:1 ratio (normalized by cell lysates). Proteins were extracted by methanol–chloroform precipitation, and were then solubilized in 8 M urea. Cysteines were reduced by adding dithiothreitol to a final concentration of 3 mM, followed by incubation at room temperature for 20 min. Then cysteines were alkylated by adding iodoacetamide to a final concentration of 50 mM, followed by incubation in the dark for 20 min. The lysates were diluted to a final concentration of 2 M urea by addition of 100 mM ammonium bicarbonate (pH 7.8) and were digested overnight with sequencing-grade trypsin (Promega) at a 1:100 (enzyme:substrate) ratio. Digestion was quenched by addition of trifluoroacetic acid to a final concentration of 0.1% and precipitates were removed by centrifugation at 4,000 rpm for 30 min. Peptides were desalted on SepPak C18 columns (Waters) according to manufacturer’s instructions. Phospho-tyrosine peptides were enriched by PTMScan® Phospho-Tyrosine Rabbit mAb (P-Tyr-1000) Kit (CST). Briefly, lyophilized peptides were resuspended in 1.4 mL Immunoaffinity Purification (IAP) buffer and incubated with the antibody-bead slurry for 2 hr at 4°C. Then the beads were washed with 1mM IAP buffer for two times, and 1mL chilled HPLC water for three times. After elution with 0.15% TFA, the p-Tyr modified peptides were desalted/concentrated with home packed C18 StageTips. In the LC-MS/MS analysis, peptides were separated on a 75 μm × 15 cm in-house packed RP column (Maccel 200-3-C18AQ, 3 μm, 200 Å) using a gradient developed over 120 min ranging from 100% Solvent A (100% H2O, 0.1% formic acid) to 32% buffer B (80% acetonitrile, 0.1% formic acid). Peptides were directly introduced into the mass spectrometer using a PicoFrit® nanospray emitter (Tip ID = 15 μm, New Objective).

Mass spectrometry analysis and data processing

The p-Tyr modified peptide samples were analyzed by LC-MS/MS on an Orbitrap Velos Pro mass spectrometer (Thermo, San Jose, CA) with an ionization voltage of 1.8 KV using a top-20 CID (collision-induced dissociation) method ¹⁶. During the chromatographic separation, the LTQ Orbitrap Velos was operated in a data-dependent mode and under direct control of the Xcalibur software (Thermo Scientific). The MS data were acquired using the following parameters: 20 data-dependent collisional-induced-dissociation (CID) MS/MS scans per full scan; full scans were acquired in Orbitrap at resolution 60,000, with 35% normalized collision energy (NCE) in CID ±2.0 Da isolation window, 0.250 activation Q and 10 ms activation time. MS/MS spectra were searched against a composite database of the human IPI protein database (Version 3.60) and its reversed complement using the Sequest algorithm (Ver28) embedded in an in-house-developed software suite. Search parameters allowed for a static modification of 57.02146 Da for Cys and a dynamic modification of phosphorylation (79.96633 Da) on Ser, Thr and Tyr, oxidation (15.99491 Da) on Met, stable isotope (10.00827 Da) and (8.01420 Da) on Arg and Lys, respectively. Search results were filtered to include <1% matches to the reverse data base by the linear discriminator function using parameters including Max. missed cleavages = 2, charge state (exclude 1+ peptides), mass accuracy, all heavy or light Lys and Arg, peptide length and fraction of ions matched to MS/MS spectra. Phosphorylation site localization was assessed by the Ascore algorithm ¹⁷ based on the observation of phosphorylation-specific fragment ions and peptide quantification was performed by using the CoreQuant algorithm ^17,18.

Immunoblot analysis.

For immunoblot analysis, the cells were extracted in lysis buffer (1% SDS, 10 mM HEPES, pH 7.0, 2 mM MgCl₂, universal nuclease 20 U/ml), and extracts were mixed with the 5 × reducing buffer (60 mM Tris-HCl, pH 6.8, 25% glycerol, 2% SDS, 14.4 mM 2-mercaptoethanol, 0.1% bromophenol blue). Samples were boiled for 5 min and subject to electrophoresis using the standard SDS–PAGE method. Proteins were then transferred to a nitrocellulose membrane (Whatman). The membranes were blocked with a TBST buffer (25 mM Tris-HCl, pH 7.5, 150 mM NaCl, 0.05% Tween 20) containing 3% non-fat dried milk, and probed overnight with primary antibodies (1:1,000 dilution) at 4 °C and for 1 h at room temperature with peroxidase-conjugated secondary antibodies. Blots were developed using enhanced chemiluminescence, exposed on autoradiograph film and developed using standard methods as described¹⁹.

Immunofluorescence staining

A375 cells were cultured in 35 mm glass bottom dishes (MatTek), incubated with fluorescein-conjugated cholera toxin B subunit (FITC-CTXb; Sigma) at a concentration of 10 μg/ml in D-PBS for 30 min at 37 °C. Cells were then washed with D-PBS and treated with 1 μM AZD6244 in complete RPMI-1640 medium for 24 h.

For confocal fluorescence microscopy analysis, cells were fixed with 4 % PFA, permeabilized with 0.25 % Triton X-100, then stained with appropriate primary antibodies followed by Alexa Fluor antibodies (Life Technologies) as secondary antibodies. Counter staining of cell nuclei was performed using DAPI (Santa Cruz Biotechnology). Immunofluorescence images were acquired with Zeiss LSM880 Airyscan confocal laser scanning microscope (Zeiss) with ×63 glycerol-immersion objective and scanning resolution of 512 × 512 pixels, zoom factor 6.4 for a subset of images. Immunofluorescence intensity were analyzed using ImageJ Software (NIH,version 1.52).

For Total internal reflection fluorescence structured illumination microscopy (TIRF-SIM) analysis, cells were fixed with 4 % PFA, and mounted in ProLong™ Gold Antifade Mountant (Thermo Fisher). Immunofluorescence images were acquired with DeltaVision OMX SR imaging system (Cytiva).

Cell viability measurement

Cell viability was measured using the CellTiter-Glo assay kit (Promega). Briefly, after drug treatment, room temperature CellTiter-Glo reagent was added 1:1 to each well and the plates were incubated at room temperature for 2 min. Luminescence was measured with the Synergy HT Multi-Detection Microplate Reader (BioTek) and was normalized against control cells treated with DMSO.

Cholesterol extraction and measurement

Three to five million cells were homogenized into 200 μL chloroform-methanol (v/v = 2:1), centrifuged for 10 min at 12,000 g at 4 °C. The organic phase was transferred to a clean tube and vacuum dried. The abundance of cholesterol was determined by the Cholesterol Fluorometric Assay Kit (Cayman). Briefly, the dried organic cell extraction was reconstituted in 200 μL assay buffer. In a typical experiment, 50 μL of the sample and a series dilution of the standard were added into each well of a 96 well plate. Fifty microliters of the assay cocktail (4.745 mL assay buffer, 150 μL cholesterol detector, 50 μL HRP, 50 μL cholesterol oxidase, and 5 μL cholesterol esterase) was added in the well and mixed. After 30 min incubation at 37 °C in the dark, the fluorescence signal was determined by a plate reader using excitation wavelengths between 530–540 nm and emission wavelengths between 585–595 nm. The relative abundance of cholesterol was normalized to the DMSO treatment group.

In vivo drug treatment experiments

PLX4032, AMG-510, and MRTX-1257 were dissolved in 2% DMSO + 30% PEG 300 + 5% Tween 80 + ddH2O. Tumors were engrafted in NSG (NOD-SCID) mice (The Jackson Laboratory) by subcutaneous injection of 3 × 10⁵ A375 or H2122 cells in RPMI-1640 medium supplemented with 50% Matrigel (BD Biosciences, cat. no. 354234). Seven days after the injection, animals were assigned randomly to control and various treatment groups (n = 5 for each group). Tumor bearing mice were administered by oral garvage: (1) Vehicle, 2% DMSO + 30% PEG 300 + 5% Tween 80 + ddH2O; (2) PLX4032, 1 mg/kg body weight/day; (3) Atorvastatin, 1.2 mg/kg body weight/day; (4) AMG-510, 5 mg/kg body weight/day; (5) MRTX-1257, 3 mg/kg body weight/day. The mice were treated every other day. Tumors were measured with an external caliper, and the volume was calculated as (4π/3) × (width/2)² × (length/2).

Statistical analysis

Quantifications were analyzed by software GraphPad prism version 8 and presented as median with SEM. Statistical significance was determined by t-test, one-way or two-way analysis of variance (ANOVA) with Tukey’s multiple comparison post hoc test for comparisons involving more than two groups. In many experiments, the mean values of the control groups were set to 1, and all other values were expressed as fold changes compared with the respective controls. P < 0.05 was considered significant, and significance was represented as per the GraphPad prism software. For in-silico analyses, the cutoffs, thresholds, and other statistical measurements are mentioned in the individual sections.

In-silico analyses

Gene Ontology analysis was performed with DAVID and the selection of Gene Ontology Terms for visualization was curated manually. The most relevant pathways were identified by Ingenuity pathway analysis. Benchmarking substrate-based kinase activity inference was analyzed in gene set enrichment analysis (GSEA). The correlation coefficient analyses, motif enrichment analyses were performed in R 3.6.1. Synergy scoring model analysis was built with the synergyfinder package (2.2.4).in R 3.6.1.

Data availability

The mass spectrometry data have been deposited to the ProteomeXchange Consortium (https://www.ebi.ac.uk/pride/archive/) via the PRIDE partner repository with the dataset identifiers: PXD021877. Computer code and all the other data supporting the findings of this study are available from the corresponding author upon request. The tyrosine peptide enrichment statistic summary was provided as Table S4.

Results

Quantitative Tyr phosphoproteomic analyses of the adaptive response in melanoma cells

As a model system, we used two melanoma cell lines, A375 and Mel-Juso, to investigate the feedback regulation of p-Tyr signaling by the MAPK pathway inhibitors. These two cell lines are driven by activated mutant BRAF (BRAF^V600E in A375) and NRAS (NRAS^Q61L in Mel-Juso) respectively. They are therefore characterized by their hyperactive MAPK signaling (Figure 1A). Immunoblotting analyses confirmed that the acute treatment of A375 cells with a BRAF inhibitor PLX4032 (Vemurafenib) for 2 hours abrogated the phosphorylation of its key downstream target proteins, including MEK, ERK, p90 ribosomal S6 kinase (RSK) and S6 (Figure 1B). Whereas prolonged PLX4032 treatment for 24–48 hours led to adaptive response, indicated by the observed reactivation of the proteins downstream of BRAF (Figure 1B) and the recovered cell proliferation (Figure 1D). Similar findings were obtained when these cells were treated with a MEK inhibitor (AZD6244, Selumetinib) (Figure 1C, D). In Mel-Juso cells (BRAF^WT/NRAS^Q61L), AZD6244 caused the adaptive resistance whereas PLX4032 did not produce the adaptation (Figure 1E–G). This is because the BRAF^V600E-specific inhibitor induced transactivation of RAF dimers, leading to the rapid and paradoxical activation of MAPK signaling in the context of BRAF^{WT 20}. Here we included the treatment of PLX4032 in Mel-Juso cells as a control group in the experimental model, to study the common features of the adaptive responses found in the other three treatment conditions. It is important to note that the developed “adaptive resistance” was not induced by the elimination of the inhibitors with 48 hours, because the conditioned medium of the cell culture treated with the inhibitors for 24–48 hr could still completely block the p-ERK in A375 cells (Figure S1).

Figure 1. Inhibition of MAPK pathway induces adaptive response in melanoma cells.

(A) The schematic representation of the RAS-RAF-MEK-ERK signaling pathway. The structures of PLX4032 (a BRAF V600E inhibitor) and AZD6244 (a MEK inhibitor) are shown.

(B-C) Immunoblots of cell lysates from A375 cells treated with 1 μM PLX4032 (B) or 1 μM AZD6244 (C) for the indicated times. GADPH served as a loading control.

(D) The cell growth data of A375 cells treated with 1 μM PLX4032 or 1 μM AZD6244 for the indicated times. The normalized growth rate was defined as the fold change of the cell amount in 24 hours.

(E-F) Immunoblots of cell lysates from Mel-Juso cells treated with 1 μM PLX4032 (E) or 1 μM AZD6244 (F) for the indicated times. GADPH served as a loading control.

(G) The cell growth data of A375 cells treated with 1 μM PLX4032 or 1 μM AZD6244 for the indicated times. Data are shown as mean ± SEM. Data are representative of three independent experiments. n.s, not significant; **, p < 0.05; ***, p < 0.001, Two-Way ANOVA post-hoc analysis.

To further characterize the feedback regulation of p-Tyr signaling, we carried out a stable isotope labeling by amino acids in cell culture (SILAC)-based approach to generate quantitative pTyr proteomic data to evaluate the effects of 48 hr treatment with either the PLX4032 BRAF^V600E inhibitor, or the AZD6244 MEK inhibitor to block ERK signaling downstream of BRAF or NRAS, compared to DMSO control-treated cells (Figure 2A). In the first experiment (A375_PLX4032), we treated the light and heavy A375 cells with DMSO and PLX4032, respectively, for 48 hr. Cells were harvested, and the lysates were combined at a 1:1 ratio. Proteins were digested, with the resulting peptides subject to p-Tyr peptide enrichment and quantitative LC-MS/MS analyses (see Materials and Methods text for the detailed description of the experiment). Similar experiments were also performed on A375 cells treated with DMSO and AZD6244 (A375_AZD6244), Mel-Juso cells treated with DMSO and AZD6244 (Mel-Juso_AZD6244), or DMSO and PLX4032 (Mel-Juso_PLX4032).

Figure 2. Quantitative Tyr phosphoproteomic analyses of the adaptive response in melanoma cells.

(A) The general schematic of the quantitative Tyrosine phosphoproteomics workflow. Briefly the SILAC labeled A375 and Mel-Juso cells were treated with AZD6244 or PLX4032 for 48 hours as indicated, followed by rapid lysis, protein extraction, and digestion into tryptic peptides. Then the peptides were enriched by the anti-Phospho-Tyr mAb beads. The resulting phosphopeptides were analyzed by LC-MS/MS experiments.

(B) Venn diagrams showing the overlap between the phosphorylated proteins identified in A375_AZD6244, A375_PLX4032, Mel-Juso_AZD6244 and Mel-Juso_PLX4032.

(C) Venn diagrams showing the overlap between the phosphorylated peptides identified in A375_AZD6244, A375_PLX4032, Mel-Juso_AZD6244 and Mel-Juso_PLX4032.

(D) Immunoblot analysis of A375 cells that were treated with PLX4032 or AZD6244 of indicated doses for 48 hours. GADPH served as a loading control. Data are representative of three independent experiments.

From these four sets of SILAC experiments, we were able to identify a total of 316 (A375_AZD6244), 232 (A375_PLX4032), 698 (Mel-Juso_AZD6244) and 705 (Mel-Juso_PLX4032) unique phosphopeptides from 164, 160, 373 and 386 proteins, respectively (Figure 2B–C, Table S1). We performed Gene Ontology Functional Annotation analyses and found that the identified proteins were enriched for cytoskeleton, cell-cell adherens junction and focal adhesion, all of which are known to be linked to p-Tyr signaling (Figure S2). These proteins were also enriched for biological processes (BP) including transmembrane RTK activity, ephrin receptor binding, non−membrane spanning tyrosine kinase activity and molecular functions (MF) including VEGFR signaling pathway and ephrin receptor signaling pathway (Figure S2).

Next we performed quantitative assessment on the changes of the phosphorylation level of specific Tyr residues. From the A375_AZD6244 dataset, we identified 15 and 16 p-Tyr proteins that were up- and down-regulated (by more than 2-fold, coefficient of variation (CV) < 25, Table S2) after AZD6244 treatment, respectively. The up-regulated phosphorylated proteins included proteins involved in the transmembrane RTK signaling pathway (e.g., IRS2, STAT3, STAT5A) (Figure S3A). We performed similar analyses on the upregulated phosphoproteins from the PLX4032-treated A375 cells and found that many of these proteins were also linked to RTK signaling (e.g., IRS2, IL6ST, JAK1, STAT3) (Figure S3B). We observed STAT3 was one of the upregulated p-Tyr proteins in both A375_PLX4032 and A375_AZD6244 datasets. Accordingly, we observed marked induction of STAT3 Y705 phosphorylation in A375 cells after the treatment with AZD6244 or PLX4032 (Figure S3C, S3D). Furthermore, pSTAT3_Y705 changes in the presence and absence of AZD6244 or PLX4032 were validated by immunoblotting experiments (Figure 2D).

The analyses of upregulated p-Tyr proteins in the Mel-Juso_AZD6244 and Mel-Juso_PLX4032 datasets showed that these two compounds induced different p-Tyr patterns. Similar to the A375_AZD6244 dataset, the up-regulated phosphorylated proteins in the Mel-Juso_AZD6244 dataset were also linked to RTK signaling (e.g., the IL-6-JAK/STAT3 signaling pathway) (Figure. S3E). In contrast, the lack of significant functional enrichment of the p-Tyr proteins in PLX4032-treated Mel-Juso cells could be ascribed to the inability of BRAF inhibitors to block MAPK signaling in these NRAS^mut cells (Figure. S3F).

Cholesterol contributes to the adaptive response to MAPK pathway inhibition

To analyze the common features in the adaptive resistant cells, we performed unsupervised hierarchical clustering based on the change of the phosphorylation level of commonly identified Tyr residues across the four SILAC datasets. The resulting heatmap showed that these site-specific phosphorylation alterations could segregate the four datasets (Figure 3A). Specifically, both PLX4032 and AZD6244 treatment led to significant changes in the pTyr status of many proteins in A375 cells, whereas changes in the pTyr proteome in the Mel-Juso NRAS mutant cells were only noted with AZD6244 treatment. These results were consistent with a model where robust blockade of the MAPK pathway was required for the adaptive activation of p-Tyr-mediated signaling (Figure 1). The aforementioned quantitative mass spectrometric data in two independent cell systems (i.e., A375 and Mel-Juso) showed that MAPK pathway inhibition led to striking adaptive alterations in the p-Tyr proteome in the resistant cells, as indicated by the extensive activation of RTK-mediated signaling, so we hypothesized that there could be some conserved upstream regulators that control the coordinated activation of these RTKs during the adaptive response to MAPK pathway inhibition. Towards this end, we first mapped the identified phosphorylation sites to known kinase-substrate relationships to computationally infer the relevant upstream kinase activities ²¹. To perform the “kinase activation” analysis, lists of kinase-substrate relationships for 348 kinases across a range of kinase families were curated from PhosphoSitePlus ²². Then the phosphorylation sites from the MS data were queried against the kinase-substrate lists using the GSEA algorithm ²³. From the analyses, we found the enriched kinase signatures were similar between A375_AZD6244, A375_PLX4032 and Mel-Juso_AZD6244 (Figure 3B). In the adaptive resistant cells, the activity of many kinases, including AXL, MER, RET, ALK, FER and JAK3, etc., were specifically upregulated (activated kinases). Meanwhile, several AZD6244/PLX4032 downstream kinases (MEK2, ERK1 and MAPKAPK2) showed substrate enrichment for the downregulated phosphorylation peptides (inhibited kinases), indicating the drug treatment was still effective when cells were harvested. The kinome analysis revealed that the activated kinases were enriched for Tyrosine Kinases (TK) family on the KinMap with multiple branches (Figure 3C), indicating AZD6244 or PLX4032 treatment in A375 cells, and AZD6244 treatment in Mel-Juso cells induced the activation of multiple RTKs.

Figure 3. Systematic analyses of the global p-Tyr proteome remodeling in melanoma cells treated with various targeted MAPK pathway inhibitors.

(A) Heatmap showing the unsupervised hierarchical clustering analysis of the phosphorylated peptides commonly quantified across A375_AZD6244, A375_PLX4032, Mel-Juso_AZD6244 and Mel-Juso_PLX4032 experiments. The values represent the z-score for the relative fold change of the ion intensity of the phosphopeptides in each drug treatment condition with respect to DMSO treatment

(B) Kinase activity inference results using the scaled phosphopeptide data set for each experiment. Each dot in the plot represents the normalized enrichment score (NES) for a single kinase.

(C) Phylogenetic kinome tree depicting the protein kinase super families of the enriched kinases.

(D) Protein motifs enriched in the upregulated phosphoproteins and activated kinases in A375_AZD6244, A375_PLX4032 and Mel-Juso_AZD6244.

(E) Schematic representation of the CRAC consensus sequences in AXL, MERTK, ALK and JAK3.

To explore the common features of the activated substrates and tyrosine kinases, we generated a list of 62 proteins (Table S3), which included the MS-identified commonly upregulated phosphorylated proteins as well as the computationally inferred activated kinases from cells with adaptive response after drug treatment (A375_AZD6244, A375_PLX4032 and Mel-Juso_AZD6244). Then we queried the list against the Pfam database which is a comprehensive collection of protein families, clans and domains based on the multiple sequence alignments. Intriguingly, the most represented motifs after the analysis were protein kinase domains (PF07714, PF00069), the cholesterol-binding motif known as the Cholesterol Recognition/interaction Amino acid Consensus sequence (CRAC)²⁴ and the SH3/SH2 domains (PF00017, PF00018) (Figure 3D). Although generally considered as a structural lipid backbone of the cell membrane, cholesterol can interact with proteins via CRAC to trigger protein dimerization, a key step in the activation of RTKs ²⁴. The existence of CRAC in these activated kinases (Figure 3E) suggests that altered cholesterol metabolism could be the underlying mechanism for MAPK pathway inhibition-induced RTK activation, and the subsequent adaptive response in melanoma cells.

Cholesterol leads to increased tyrosine phosphorylation

To directly test this hypothesis, we first used a fluorometric assay to measure the total cholesterol level in A375 cells treated with or without the MEK (AZD6244) and BRAF (PLX4032) inhibitors. We found that both inhibitors (24 hr treatment) resulted in a ~3-fold increase of the cholesterol level (Figure 4A). It has been well established that the assembly of cholesterol, sphingolipid as well as ganglioside can form a stable membrane sub-compartment called lipid rafts. These nanoscale domains function as a key platform to regulate membrane signaling and trafficking ²⁵. We next labeled A375 cells with the fluorescently-tagged cholera toxin B subunit (FITC-CTXb), which specifically binds to the GM1 ganglioside, and thus, visualizes the cholesterol-rich microdomains (lipid rafts) in the plasma membrane ²⁶. Indeed, confocal microscope analyses showed that AZD6244 treatment dramatically increased the fluorescence intensity of cholesterol-rich microdomains (Figure 4B). Quantitative analyses showed that the total fluorescence intensity within an individual cell was about 3-fold higher in A375 cells treated with AZD6244 than that in control cells (Figure S4A, the outlined region was enlarged in Figure 4B). Using a different microscope modality, namely the total internal reflection fluorescence (TIRF), we confirmed that the plasma membrane localization of cholesterol-rich microclusters was also significantly increased in AZD6244-treated A375 cells (Figure S4B).

Figure 4. Cholesterol leads to increased tyrosine phosphorylation in the resistant cells.

(A) Bar plot of the relative level of cholesterol in A375 cells treated with 1 μM AZD6244 or 1μM PLX4032 for 24 hours. Two-tailed t test. ***, p < 0.001.

(B) Representative confocal immunofluorescent images of FITC-CTXb, AlexaFluro 594-anti–α-tubulin and DAPI staining of A375 cells under DMSO and AZD6244 treatment (1 μM, 24 h) conditions. Scale bar = 10 μm.

(C) Immunoblots of cell lysates from serum starved A375 cells pretreated with DMSO or 5 mM MβCD for 1h at 37°C, then stimulated with 200 nM Insulin or 20 mM Cholesterol for 10 min 37°C. GADPH served as a loading control.

(D) Immunoblots of cell lysates from A375 cells treated with 1 μM AZD6244 or PLX4032, with or without the presence of 3 mM MβCD for 24 hours.

(E) Immunoblots of cell lysates from H2122 and A549 cells treated with 1 μM AMG-510 or MRTX-1257 for 24h at 37°C.

(F) Bar plot of the relative level of cholesterol in H2122, SW1573 and A549 cell lysates treated with 1 μM AZD6244 or 1μM PLX4032 for 24 hours. Two-way ANOVA analysis followed by Tukeýs post-hoc test. ns, not significant; **, p < 0.05; ***, p < 0.001. Data are representative of three independent experiments

Many growth factor receptors have been shown to be enriched in lipid rafts ²⁷, with their dimerization, and subsequent activation further regulated by cellular cholesterol contents ^28,29. Proteins involved in different signaling cascades could be recruited to rafts, which could lead to the interaction or segregation of different signaling cascades, including multiple RTK pathways ³⁰. The dramatic increase in both total cholesterol abundances and lipid raft formation in MAPK pathway-inhibited A375 cells prompted us to investigate whether cholesterol could contribute to control the adaptive remodeling of the p-Tyr proteome. We first confirmed that cholesterol was required for p-Tyr signaling events mediated by various RTKs. As a model system, we found that insulin stimulation in serum-starved A375 cells increased the global p-Tyr level, as well as the phosphorylation of Insulin Receptor Substrate 1 (IRS1), which is a downstream substrate of the Insulin Receptor (IR) (Figure. 4C). Similar results were obtained when these cells were treated with cholesterol (Figure. 4C). Pre-treating the cells with methyl-β-cyclodextrin (MβCD), which is a cholesterol-depleting agent³¹, blocked insulin-induced global p-Tyr, as well as p-IRS1 signals. Pre-complexed MβCD and cholesterol at a 1:2 molar ratio partially rescued p-Tyr inhibition caused by cholesterol depletion (Figure. 4C). We also explored the role of cholesterol in RTK activation induced by MAPK pathway inhibitors. In A375 cells, both AZD6244 and PLX4032 treatment increased global p-Tyr and p-STAT3 levels. However, in the presence of MβCD, the induction of global p-Tyr and p-STAT3 was completely abrogated (Figure. 4D). Collectively, these data indicate that cholesterol is required for the bypass p-Tyr activation caused by various MAPK pathway inhibitors.

We extended our findings by analyzing the correlation between cholesterol/lipid abundances and drug sensitivity in a wider panel of melanoma cell lines. The Cancer Cell Line Encyclopedia (CCLE) project used liquid chromatography–mass spectrometry (LC-MS) and profiled more than 225 metabolites in 928 cell lines from more than 20 cancer types ³². We extracted the data from the 40 melanoma cells that harbor the BRAF V600E mutation. Based on their response to AZD6244 and PLX4720 (the tool compound for PLX4032) (data derived from the Cancer Therapeutics Response Portal), we then classified these cell lines into the “sensitive” (ActArea > 3) and “resistant” (ActArea < 3) groups ³³. Statistical analyses revealed that the resistant cells have significantly higher levels of cholesterol and other lipid species, compared to the sensitive cell lines (Figure. S5 and S6). Taken together, these results suggest that at least for melanoma cells, increased abundances of cholesterol and the related lipid species could also play a role in determining their innate resistance to MAPK pathway inhibitors.

Besides in melanoma, aberrant activation of MAPK signaling is also found in many other human malignancies (e.g., as a result of KRAS mutation in NSCLC) ^34,35. Marked progresses have been reported for the development of covalent inhibitors targeting the KRAS^G12C mutant. Not dissimilar to BRAF/MEK inhibitors for BRAF^mut melanoma, long-term treatment of KRAS^G12C inhibitors also triggers the tumor adaptive response, which is characterized by the reactivation of MEK/ERK under these conditions¹⁰. We then explored the role of cholesterol in regulating the bypass mechanisms in KRAS^G12C tumors. Indeed, treatment of H2122 cells (a KRAS^G12C NSCLC cell line) with two chemically distinct, covalent KRAS^G12C inhibitors (i.e., AMG 510³⁶ and MRTX1257³⁷) resulted in a dramatic increase in the global p-Tyr levels, while no significant changes of p-Tyr signaling were observed in A549 cells (a KRAS^G12S NSCLC cell line) treated with these compounds (Figure. 4E). Accordingly, fluorometric assays showed that there was a two to five-fold increase of the total cholesterol level in H2122 and SW1573 (a KRAS^G12C NSCLC cell line) treated with AMG510 or MRTX1257, while the same treatment in A549 cells did not affect its cholesterol level (Figure. 4F). These findings indicate that aberrant cholesterol metabolism might be a conserved mechanism for the adaptive response in tumors with hyperactive MAPK signaling.

Synergistic effects between cholesterol-lowering agents and MAPK pathway inhibitors

Cholesterol is being increasingly appreciated for its role in regulating the initiation, progression and metastasis of various human cancers ^38,39. Indeed, large-scale meta-analyses have pointed to decreased cancer incidence in patients that use cholesterol-lowering drugs (e.g., statins) ^40,41. Because of the critical role of cholesterol in regulating the tumor adaptive response to MAPK pathway inhibitors, we sought to test whether cholesterol-lowering drugs could enhance the cytotoxic effects of MAPK pathway inhibitors. Using a CellTiter-Glo assay, we performed a four-by-four dose-response validation matrices using Atorvastatin (an FDA-approved, cholesterol-lowering agent, https://www.fda.gov/drugs/information-drug-class/statins), AZD6244 and PLX4032 (Figure. 5A, S7A). The dose-response profiles showed that the combination of Atorvastatin and AZD6244 was highly effective in A375 cells, which was evidenced by the positive Loewe score (Figure. 5B). We also observed a highly combinational effect between Atorvastatin and PLX4032, which was associated with a higher Loewe score (Figure. S7B). The addition of Atorvastatin to either AZD6244 or PLX4032 treatment also greatly enhanced the cytotoxic effects of these MAPK pathway inhibitors in A375 cells (Figure. 5C, S7C). Using similar analyses, we found another HMG CoA reductase inhibitor Simvastatin could also significantly enhance the effect of AZD6244 or PLX4032 treatment in A375 cells (Figure S7D). Collectively our data suggested that HMG-CoA reductase inhibitors (which may limit the cholesterol synthesis) could sensitize the melanoma cells to either AZD6244 or PLX4032 treatment. Next we evaluated the synergistic effects between Atorvastatin and PLX4032 in vivo. In an A375 xenograft model, PLX4032 treatment significantly slowed tumor growth, whereas Atorvastatin alone had no obvious effects. Strikingly, the combination of PLX4032 and Atorvastatin resulted in dramatic regression of the tumors (Figure. 5D).

Figure 5. Synergistic effects between cholesterol-lowering agents and MAPK pathway inhibitors.

(A) The dose-response matrix of A375 cells treated with AZD6244 and Atorvastatin by the indicated concentrations for 48 hours at 37°C. The numbers indicate the inhibition of cell growth (%).

(B) The synergy plots show that the combination of AZD6244 and Atorvastatin is synergistic in A375 cells.

(C) The viability of A375 cells after the treatment of the indicated compounds for 48 hours at 37°C. The concentration for each compound was 1 μM. Data are means ± SEM with three replicates. One-way ANOVA analysis followed by Tukeýs post-hoc test. ***, p < 0.001.

(D) Tumor sizes of an A375 xenograft model (n = 5 per group). Mice were dosed with 1 mg/kg PLX4032, 1.2 mg/kg Atorvastatin and the combination for 2 weeks (after 7 days post injection of the tumor cells). Tumor sizes were measured on the indicated days post injection. Two-way ANOVA analysis followed by Tukeýs post-hoc test.

(E) The dose-response matrix of H2122 cells treated with MRTX1257 and Atorvastatin using the indicated concentrations for 48 hours at 37°C. The numbers indicate the inhibition of cell growth (%).

(F) The synergy plots show that the combination of MRTX1257 and Atorvastatin is synergistic in H2122 cells.

(G) Tumor sizes in an H2122 xenograft model (n = 5 per group). Mice were dosed with 5 mg/kg AMG-510, 3 mg/kg MRTX-1257, 1.2 mg/kg Atorvastatin and indicated combinations for 2 weeks after 7 days post injection of the tumor cells. Tumor sizes were measured on the indicated days post injection. Two-way ANOVA analysis followed by Tukeýs post-hoc test. Data are representative of three independent experiments.

We also explored the potential synergistic effect between cholesterol-lowering agents and KRAS^G12C inhibitors. In this case, MRTX1257 was combined with Atorvastatin in a pairwise, all-versus-all fashion for three NSCLC cell lines. The results showed that the combination of Atorvastatin and MRTX1257 was highly synergistic in both KRAS^G12C lines, i.e., H2122 (Figure. 5E–F) and SW1573 (Figure. S7E–F). However, in the KRAS G12S cell line A549, MRTX1257 had little effect on the proliferation even at 5 μM (Figure. S7G). In addition, MRTX1257 did not show any synergistic effect when combined with Atorvastatin (Figure. S7H). Finally, the xenograft experiments indicated that KRAS^G12C NSCLC tumors were sensitive to AMG510 or MRTX1257 monotherapy, where both KRAS^G12C blockers inhibited tumor growth (Figure. 5G). However, the combination of KRAS^G12C inhibitors with Atorvastatin could achieve a more complete response (Figure. 5G). These findings indicate that besides in melanoma, cholesterol-targeting agents could also overcome the adaptive response in KRAS^G12C-mutated NSCLC.

Discussion

Our understanding of how tumor cells develop resistance to targeted kinases inhibitors has been buoyed by rapid progresses in DNA sequencing technologies and has led to some of the first advances in improved therapeutic strategies with more complete and durable responses. Besides the genetic causes, recent studies have shown that resistance can also arise through non-genetic mechanisms. Indeed, activation of alternate routes of kinase pathways is frequently observed following targeted inhibition of oncogenic kinases. An understanding of the molecular underpinnings of these events will be critical for the rational design of combination approaches to overcome therapeutic resistance. We utilized a quantitative mass spectrometry-based proteomic strategy to comprehensively monitor the dynamic remodeling of the p-Tyr proteome. Indeed, our findings aligned well with the previous studies, with the identification of activated tyrosine kinases (e.g., PDGF, JAK/STAT3, VEGF, c-kit et al.) in melanoma cells treated with various MAPK pathway inhibitors ^8,42–45. Among these pathways, the JAK/STAT3 signaling axis has been well documented as a critical mediator of the adaptive tumor response ^46–49. Activation of JAK/STAT3 signaling is an important adaptive resistance mechanism to EGFR tyrosine kinase inhibitors (TKI) in non–small cell lung cancer (NSCLC) ⁴⁷. Our data suggest that JAK/STAT3 could also be a bypass mechanism that is relevant to MAPK-inhibited melanoma cells. However, because of the extensive reprogramming of the p-Tyr signaling network, combining MAPK pathway inhibitors with compounds targeting one or two of these activated kinases is less likely to be sufficient to overcome the adaptive tumor response.

In search for a molecular switch that regulates the coordinated activation of multiple RTKs, we identified a cholesterol binding motif that is commonly shared by many of these tyrosine kinases. Biochemical experiments then confirmed the dramatic accumulation of cholesterol in melanoma cells treated with the MAPK pathway inhibitors. Cholesterol is an important component of the “lipid rafts” and the related structures in the plasma membrane of mammalian cells ⁵⁰. Lipid rafts were initially studied in the context of protein tyrosine kinase signaling. When bound to their ligands, certain RTKs move into cholesterol-rich lipid microdomains that contain other kinases, adaptors and scaffolds, which become activated to initiate downstream signaling ^25,51,52. Indeed, we showed that depletion of cholesterol blocks RTK signaling induced by various MAPK pathway inhibitors, and we found that a profound synergistic effect between atorvastatin and BRAF/MEK inhibitors in melanoma, as well as between atorvastatin and a KRASG12C inhibitor in NSCLC, both in vitro and in vivo.

Biosynthesis of cholesterol requires many reaction steps and is under strict regulatory control. HMG-CoA synthase convert acetyl-CoA and acetoacetyl-CoA into 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA), followed by the rate-limiting step: the conversion of HMG-CoA to mevalonate by HMG-CoA reductase⁵³. In addition to atorvastatin, we found that another HMG-CoA reductase inhibitor simvastatin could also sensitize the melanoma cells to either AZD6244 or PLX4032 treatment, suggesting cholesterol biosynthesis could be targeted for the melanoma cells with adaptive resistance. Consistent with our findings, it was reported that HMG-CoA synthase 1 (HMGCS1), the upstream ketogenic enzyme of HMG-CoA, as an additional “synthetic lethal” partner of BRAFV600E ⁵⁴. Also, it has been reported that 3-hydroxy-3-methyl-glutaryl-CoA reductase (HMGCR), a crucial enzyme in the mevalonate pathway for sterol biosynthesis, is elevated in enzalutamide-resistant prostate cancer cell lines⁵⁵. Collectively, these data suggested that elevated cholesterol biosynthesis could be implicated in the resistance phenotype of multiple different cancer types.

Statins have been used by millions of people as hypocholesterolemic agents in cardiovascular and cerebrovascular diseases. Due to their extraordinary safety profiles, statins are among the most commonly prescribed drugs worldwide⁵⁶. Mechanistically, statins are 3-hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase inhibitors, which block the synthesis pathway of the cholesterol precursor mevalonate⁵⁷. Nevertheless, large randomized controlled trials (RCTs) and numerous preclinical studies have shown that, in addition to the hypocholesterolemic activity, statins can also have a protective effect against the development of cancers^38,39, particularly in melanoma ^58–60. Atorvastatin inhibits the survival, invasion and metastasis of human melanoma cells by attenuating RHOC signaling and geranylgeranylation of the Rho family proteins⁶¹. Other independent studies have shown that lovastatin reduced membrane-associated Rho proteins and metastasis in the B16F10 mouse melanoma model⁵⁸. Moreover, statins could be pharmacological inhibitors for melanoma immunotherapy as they favor the innate immune response against tumor cells⁶². The role of statins has also been studied in lung cancer. It was able to decrease the cell proliferation and tumor growth in several cellular models, such as lung cancer (A549, NCI-H292), and in a xenograft mouse model of A549 lung cancer⁶³. Recently Chou et al. have revealed the role of statins in lung cancer in a population-based study⁶⁴. By retrospectively analyzing patients diagnosed with lung cancer between 1998 and 2011, they found a high percentage of patients with p53 mutations observing a reduced 5-years mortality under simvastatin treatment. Then, the impact of simvastatin upon tumor-suppression was validated in both normal and mutant p53 cells by regulating different signaling pathways. Based on these recent advances, we outline important considerations for advancing statins to cancer prevention.

In addition to RTKs, cellular cholesterol is also involved in regulating the activation of other important signaling molecules. For example, it has been shown that, in order to fulfill its signaling role, a fraction of H-Ras needs to be localized to the cholesterol-rich lipid rafts in the plasma membrane. The expression of a dominant-negative mutant of caveolin depletes cell surface cholesterol, and thereby inhibits the membrane partition and the subsequent activation of H-Ras ⁶⁵. Furthermore, lysosomal cholesterol has been shown to activate mTORC1 via the SLC38A9-Niemann-Pick C1 (NPC1) signaling complex ⁶⁶. These results point to the intriguing possibility that MAPK pathway inhibition-induced cholesterol accumulation could also lead to the compensatory activation of H-Ras and mTORC1 (and perhaps many other signaling molecules), and thereby provide yet other forms of pro-survival signals. Taken together, our results demonstrate cholesterol as the central signaling hub that controls the activation of bypass signaling during MAPK pathway inhibition. The potential regulatory mechanisms that lead to cholesterol accumulation under these conditions warrant further investigation.

In summary, we performed a comprehensive characterization of the remodeling of the p-Tyr proteome in melanoma cells treated with various MAPK pathway inhibitors. We showed that the blockade of MAPK signaling leads to the accumulation of cholesterol, which serves as a contributing factor to control the coordinated activation of multiple RTKs. Importantly, the combination of cholesterol-lowering agents and MAPK pathway inhibitors overcomes the adaptive response in melanoma and NSCLC in cell culture and xenograft models. Because both Atorvastatin and several MAPK pathway inhibitors (e.g., PLX4032 and AMG510) have already been approved by the FDA, this provides an actionable combination therapy that can be rapidly translated and tested in clinical studies.

Supplementary Material

supporting information

Figure S1 The inhibitors in the cell culture remain active during the 48 hour treatment

Figure S2 Gene Ontology (GO) analysis

Figure S3 Quantitative analysis of the p-Tyr proteome in response to drug treatment

Figure S4 Images of CTXb labeled membrane patches

Figure S5 The metabolite panel of “sensitive” and “resistant” melanoma cells in response to AZD6244

Figure S6 The metabolite panel of “sensitive” and “resistant” melanoma cells in response to PLX4032

Figure S7 Synergistic effects between statins and MAPK pathway inhibitors

Figure S8 Uncropped Western blot images corresponding to Figure 1

Figure S9 Uncropped Western blot images corresponding to Figure 2

Figure S10 Uncropped Western blot images corresponding to Figure 4

Figure S11 Uncropped Western blot images corresponding to Figure S1

Table S3

The list of activated substrates and tyrosine kinases (xlsx)

Table S2

The highlighted tyrosine peptides and proteins in Figure S3 (xlsx)

Table S4

The tyrosine peptide enrichment statistic summary (xlsx)

Table S1

The identified peptides and proteins (xlsx)