ROS production induced by BRAF inhibitor treatment rewires metabolic processes affecting cell growth of melanoma cells

Selective BRAF inhibitors PLX4032 (Vemurafenib) and GSK2118436 (Dabrafenib) were purchased from Selleckchem. BRAF inhibitors were dissolved in DMSO according to the manufacturer’s instructions and stored at −80 °C. Working aliquots were diluted in 100% ethanol at a concentration of 1 mM for PLX4032 and 100 μM for GSK2118436 and stored at −20 °C. The MEK inhibitor GSK1120212 (Trametinib) was purchased from Selleckchem and was dissolved in DMSO at a concentration of 10 mM and stored at −20 °C. Dichloroacetic acid (DCA) was purchased from Sigma-Aldrich and dissolved in distilled water before use. AZD7545 was purchased from Selleckchem and was dissolved in 100% ethanol according to the manufacturer’s instructions. 10 mM working aliquots were stored at −20 °C. Mitoquinone mesylate (MitoQ) was purchased from Medkoo Biosciences, dissolved in DMSO according to the manufacturer’s instructions and stored at −20 °C. The following antibodies were used for western blot detection: anti phospho-PDH E1α (Merck Millipore); anti PDK1 (Enzo Life Sciences); anti phospho-ERK1/2 (Cell Signaling); anti ERK1/2 (Santa Cruz); anti cleaved-PARP (Cell Signaling); anti PARP (Cell Signaling); anti α-tubulin (Santa Cruz); anti HIF1α (Abcam); anti phospho- AMPK (Cell Signaling).

All melanoma cells were purchased from ATCC while 501Mel was obtained from Dr. Ruth Halaban (Dermatology department, Yale School of Medicine, USA). All cell lines were cultured in RPMI 1640 medium containing ultraglutamine (Lonza BioWhittaker), supplemented with 10% FCS (Foetal Calf Serum, GIBCO) and 1% PS (10′000 U/ml Penicillin and 10′000 U/ml Streptomycin, Lonza BioWhittaker) and grown at 37 °C in a humidified atmosphere at 5% CO2. Cells were regularly tested to be mycoplasma free. Drug-resistant cells were generated by culturing parental A375 cells in presence of 1 μΜ PLX4032 for 4–6 weeks. Twenty surviving clones were picked and then combined in equal proportions to obtain an heterogeneous pool of resistant cells.

The plasmid encoding the cytosolic (cyto)-human glutaredoxin 1 (Grx1)-roGFP2 was obtained by amplifying the cDNA Grx1-roGFP2 sequence without the Peroxisomal Targeting Sequence 1 (PTS1) from the plasmid encoding the peroxisomal (po)-Grx1-roGFP2 vector [19] by PCR (primers; pCyto-EcoRI-5′ 5′-gga gga gga tca gga gga gaa ttc gtg agc aag ggc gag gag-3′ (forward) and pCyto-XbaI-3′ 5′-ctc gac tta tct aga tta ctt gta cag ctc gtc-3′ (reverse)) and subcloned into PCR2.1-TOPO (Invitrogen). The resulting plasmid was digested with EcoRI and XbaI, prior to subcloning into the EcoRI/XbaI digested pcDNA3.1-Grx1-roGFP2-PTS1 vector to obtain the p-cyto-Grx1-roGFP2. The plasmid was verified by DNA sequencing.

The ERK1/2 siRNAs were obtained from GE Dharmacon (siGenome Human). siRNA transfections were performed using 3 μL HiPerfect transfection reagent (Qiagen) per reaction according to the manufacturer’s instructions. The final concentration of siRNA was 50 nM for each ERK1/2 and 100 nM for scrambled control. ERK1 and ERK2 siRNA transfections were performed 48 h prior to the 24 h incubation with PLX4032 (1 μM) and/or Trametinib (3 nM).

The PDK 1–4 siRNAs were obtained from GE Dharmacon (ON-TARGETplus Human). siRNA transfections were performed using 1.5 μL Lipofectamine RNAiMAX (Invitrogen) per reaction according to the manufacturer’s instructions. The final concentration of siRNA was 25 nM for each PDK (PDK1–4) and 100 nM for scrambled control. PDK1–4 siRNA transfections were performed 24 h prior to 24 h incubation with PLX4032 (1 μM).

Cell lysis was performed at 4 °C using ice cold buffers. Cells were lysed on the dish with lysis buffer containing 30 mM Tris/HCl pH 6.7, 5% glycerol, 2.5% mercaptoethanol, 1% SDS. Protein extracts were further analysed by SDS-PAGE and Western blotting. ECL signals were detected as described before [20]. Before re-probing, blots were stripped as described before [21]. All experiments were performed in three biological replicates and one representative replicate is shown. Western blot quantification of ECL signals was performed using both the Image Lab 4.0.1 software from Bio-RAD and Bio1D analysis package (Vilber).

Total RNA was extracted using the Quick-RNA™ miniprep kit (Zymo Research) according to the manufacturer’s instructions and the concentration and quality was determined using a NanoDrop Spectrophotometer. 250 or 500 ng of total RNA was reverse-transcribed with the miScript II RT kit (Qiagen) in a volume of 10 μL, according to the manufacturer’s instructions. Quantitative real time PCR (qPCR) was carried out on a CFX96 Detection System (BioRad) in a total volume of 10 μL (10 pmol of each primer and containing cDNA corresponding to 50 ng RNA template). The housekeeping genes PPIA, HPRT and the target genes were assayed in parallel for each sample. Melting curve analysis was performed to guarantee the specificity of the qPCR primers as previously described [22]. All samples were run in biological triplicates each consisting of three technical replicates.

Gene-specific qPCR primers for PDK1, PDK3, PDK4 and HIF1a were purchased from Eurogentec (Belgium). For the detection of PDK2 a RT2 qPCR primer assays from Qiagen were used. The geometric mean of two housekeeping genes was calculated and a normalization factor for each sample was generated using geNorm (VBA add-in for Microsoft Excel). The normalization factor was used to calculate the relative amount of each target mRNA in each sample. Each sample was normalized to the untreated control.

Cells were transfected with a plasmid coding for Grx1-roGFP2 or mitochondrial (mito)-roGFP2 (pMF1762) as previously described [23], using Lipofectamine LTX2000 (Invitrogen). Two days post transfection the cells were either subjected to a 1 h pretreatment with mitoQ (150 nM), a treatment with PLX4032 (3 h for A375; 10 h for 501 Mel or IGR37) or a combination thereof. roGFP2 is a genetically engineered GFP, which contains two cysteine residues that can be oxidized. Depending on the oxidation state, roGFP changes its excitation spectrum (reduced: 488/530 nm; oxidized: 400/530 nm). By measuring the ratio (400/488 nm) of the fluorescence emitted by the two excitation states of the roGFP2, the redox status of a subcellular compartment can be monitored [24, 25]. The higher this ratio, the higher is the redox state. The analysis was performed using the ImageJ software on pictures taken using an Andor Revolution W1 spinning disc confocal microscope, mounted on a Nikon Ti microscope (60× oil objective). Between 10 and 20 ROIs (regions of interest) were analysed per cell with a minimum of 12–20 cells analysed for each condition. The excitation record time was set to 250 msec or 300 msec for the 400 nm channel and 50 msec or 60 msec for the 488 nm channel for the Grx1-roGFP2 or mito-roGFP2 respectively. Two independent experiments were performed.

50 × 10³ cells/well of 5 melanoma cell lines were seeded in 12-well plates and 24 h later stimulated with 10 μM of AZD7545. Cellular growth was monitored in the IncuCyte ZOOM live cell microscope (Essen BioScience) and images were taken in phase contrast every 3 h for a total of 90 h. Proliferation assay were carried out for three biological replicates and for each figure one representative replicate is shown.

A375 cells were plated in 12 well plates in complete medium and allowed to adhere overnight. Cells were transduced with lentiviral particles (LV-iRFP-P2A-Puro) containing near-infrared fluorescent protein (iRFP) linked to the puromycin resistance gene (Puro) via a P2A cleavage peptide (Imanis Life Sciences). Virus was prepared in serum-free medium and added to the cells with a multiplicity of infection (MOI) of 10. To increase transduction efficiency, the plate was centrifuged at 300 g for 30 min and placed in the incubator. Four hours after transduction, complete medium was added to the cells. Forty-eight hours after transduction, cells were checked on the Odyssey Infrared imaging system (LI-COR Bioscience) for iRFP expression and selection with puromycin (Invivogen) at 1 μg/ml was initiated. After several rounds of puromycin selection, cells were analysed on the Odyssey Infrared imaging system (LI-COR Bioscience) and on a FACSCanto II flow cytometer using FACSDiva (BD Bioscences) software to confirm that 100% of the cell population is expressing iRFP.

The A375-iRFP cells were used to test the combination of 1 μM PLX4032 and 10 μM AZD7545 versus PLX4032 alone. 10,000 cells were seeded per well (6 well plates) and were treated with the inhibitors for 3 weeks. Medium was changed twice a week. After 3 weeks of treatment, cells were scanned and the intensity of the iRFP signal was measured using the LI-COR Odyssey instrument (LI-COR Biosciences). The iRFP signal was quantified using the Image Studio lite version 4.0 software (LI-COR Biosciences).

A375 cells were plated in round-bottom ultra-low attachment 96-well plates in 90 μl of serum-free DMEM-F12 supplemented with B-27 (1X; Invitrogen), Insulin (4 U/L; Sigma), Heparin (4 μg/ml; Sigma), EGF (20 ng/ml; Biomol); bFGF (20 ng/ml: Miltenyi Biotec), and penicillin/streptomycin (1X, Lonza). After 2 days of sphere formation, drugs were added. Sphere growth was monitored in the IncuCyte ZOOM live cell microscope (Essen BioScience) by measuring four different spheroid diameters after 3 days. Spheroid assays were carried out in three biological replicates.

In the wake of recent publications investigating the role of the BRAF oncogene on metabolic processes, we wanted to explore the effects of BRAF inhibitors on melanoma cells. To this end, time course experiments were performed in a panel of BRAF^V600E and BRAF^WT/NRAS^mut human melanoma cells, which were incubated with two BRAF kinase inhibitors, PLX4032 and GSK2118436, for the indicated time points (Fig. 1). As expected, BRAF inhibition caused the down-regulation of phosphorylated ERK in the BRAF^V600E positive cell lines A375, IGR37 and 501Mel already after 30 min (Fig. 1a). On the contrary and as previously described [26, 27], PLX4032 and GSK2118436 lead to an upregulation of ERK phosphorylation in cells expressing BRAF^WT/NRAS^mut (Fig. 1b). Total PARP and cleaved PARP detections were performed to assess the induction of apoptosis which seems to occur only 48 and 72 h after treatment of the BRAF^V600E positive cells. Interestingly, we found an increased phosphorylation of PDH-E1α on S₃₀₀ from around 7 h post inhibitor treatment in the BRAF^V600E positive cell lines A375, IGR37 and 501 Mel (Fig. 1a and c), increasing at 24 h and generally peaking at 48 h. In the BRAF^WT/NRAS^mut cells, the situation was again different and no change in the phosphorylation of PDH was observed over time. Phosphorylation of PDH-E1α at S₃₀₀ is associated with an inhibition of the activity of the pyruvate dehydrogenase complex (PDH), which converts pyruvate into acetyl-CoA and thus regulates the entry of metabolites from glycolysis into the TCA cycle. Our data suggest that treatment with BRAF inhibitors induces changes in the metabolism of cells harboring the BRAF^V600E mutation in a different way than in cell expressing BRAF^WT.

To rule out that the PDH phosphorylation is an off target effect of the BRAF inhibitor and since BRAF^V600Eare known to activate ERK1/2 via MEK1, we investigated the involvement of ERK1/2 and MEK1 in mediating PDH-E1α phosphorylation in A375 cells. Using an siRNA approach targeting both ERK1 and ERK2 in combination with PLX4032 we could show that a knock-down of ERK1/2 induces an upregulation of PDH-E1α phosphorylation, comparable to what was achieved with PLX4032 (Fig. 2a). In addition, the same upregulation was obtained when the cells were treated with a MEK1 inhibitor (Trametinib), which also leads to the down-regulation of pERK. Of note, neither of the effects could be increased by combination treatment with both BRAF and MEK inhibitor. To further explore the link between RAS/RAF/MEK/ERK pathway and PDH phosphorylation we used NRAS^mut cells (SKMel30, MelJuso, IPC298) (which did not upregulate pPDH upon BRAF inhibitor treatment) and treated them with Trametinib for the indicated time points. As expected, MEK inhibition caused a down-regulation of ERK phosphorylation already after 30 min (Fig. 2b) also in NRAS^mut cells. Phosphorylation of PDH-E1α on S₃₀₀ was observed in those cells from around 24 h onwards. These results suggest that the phosphorylation of PDH is a specific effect mediated by the BRAF/MEK/ERK pathway and not an off target effect of BRAF inhibitors and that it occurs in all melanoma cells presenting activating mutations in this pathway.

Since PDKs are known to be responsible for PDH phosphorylation [14, 15], we next analysed mRNA levels of the four PDK isoforms in three BRAF^V600E cell lines. HIF1α was included in the study since this transcription factor induces PDK1 expression and has been shown to be upregulated by conditions other than hypoxia [28], notably also by activation of the RAS/MAPK pathway [29]. Furthermore, constitutive HIF1α activity has been described in malignant melanoma [30]. Surprisingly, an upregulation of the mRNA levels of PDK1 and HIF1α was detected in all cell lines at 12 and 24 h after PLX4032 treatment (Fig. 3). However, validation of these results showed that although HIF1α mRNA was upregulated upon treatment, this effect was not detectable at the protein level (Additional file 1: Figure S1). This finding was corroborated by the fact that PDK1 protein levels, which are regulated by HIF1α protein, were also not upregulated upon BRAF inhibitor treatment. Interestingly, the mRNAs of the other PDK family members PDK3 and PDK4 were upregulated upon BRAF inhibitor treatment, while PDK2 was not detected in any cell line. The protein levels of PDK3 and PDK4 could not be investigated since the available antibodies were not specific for these proteins (data not shown). Thus, mRNA levels of the PDKs are upregulated following BRAF inhibitor treatment, although for PDK1 no increase in protein levels was detectable.

The activity of PDKs can also be regulated by alternative post-translational mechanisms [31], and for enzymes the regulation of their activity is often more important than their mere expression level. Therefore, we investigated the importance of PDKs in our inhibitor-mediated PDH-E1α phosphorylation by knocking down all four PDK isoforms using an siRNA approach. The down-regulation of mRNAs for all isoforms ranged between ~10–40% compared to both untreated and scrambled controls (Fig. 4a). As for PDK1 protein levels, a strong reduction following the siRNA treatment could be observed too (Fig. 4b). Following the knock-down of the PDKs, a drastic decrease in the phosphorylation levels of PDH-E1α at serine 300 residue in A375 cells was detected, despite the presence of PLX4032 (Fig. 4b). To further support these results, we used dichloroacetic acid (DCA), a pan-PDK inhibitor, which also resulted in a decreased phosphorylation level of PDH-E1α in A375 (Fig. 4c). In summary, these data indicate that PDK isoforms mediate the PLX4032/GSK118436-induced phosphorylation of PDH-E1α.

AMPK phosphorylation and its subsequent activation upon BRAF inhibitor treatment has previously been described in melanoma and colorectal cancer cells [32, 33]. Since AMPK has indirectly been implicated in mediating PDH phosphorylation [34], we explored the phosphorylation status of AMPK in our setting. In accordance with previous publications [32, 33], we also observed an upregulation of AMPK phosphorylation on T₁₇₂ following inhibition of BRAF^V600E with both PLX4032 and GSK2118436 (Fig. 5). However, the phosphorylation occurred generally at later time points (24 or 48 h after treatment) compared to the onset of PDH phosphorylation. For this reason, we exclude the possibility that the observed PDH phosphorylation following BRAF inhibitor treatment was indirectly induced by AMPK.

Mitochondrial ROS production after PLX4032 exposure has recently been described [35]. Since we and others have previously demonstrated that the induction of ROS can stimulate the phosphorylation of PDH [34, 36], we investigated a possible connection between ROS and pPDH after PLX4032 treatment in the context of melanoma. In order to measure the redox state in both mitochondria and the cytosol after PLX4032 treatment, A375, IGR37 and 501Mel cells were transfected with a plasmid expressing either the mito-roGFP2 or the Grx1-roGFP2, which can be used to monitor the mitochondrial and the cytosolic redox state, respectively. Both redox states after PLX4032 treatment were compared to untreated controls. In particular, we treated A375 cells with the BRAF inhibitor for 3 h whereas IGR37 and 501Mel were tested for 10 h. The time points were chosen according to the onset of pS₃₀₀PDH after PLX4032 treatment, in particular just before the phosphorylation was becoming prominent. Furthermore, we also used a mitochondria-targeted antioxidant (mitoQ) in order to prevent or decrease oxidative stess induced by PLX4032. Melanoma cells were pre-treated with 150 nM of the antioxidant, mitoquinone mesylate (MitoQ), followed by PLX4032 treatment. For all cell lines, both the mitochondrial and the cytosolic redox status was significantly increased after exposure to PLX4032 and decreased in case of mitoQ pre-treatment (Fig. 6a). We could also observe a decreased phosphorylation of PDH-E1α on S₃₀₀ when the cells were pre-incubated with MitoQ suggesting that ROS are activating PDKs, which in turn leads to inhibition of PDH by phosphorylation (Fig. 6b). As an additional confirmation, cells were treated with 100 μM H₂O₂, which also induced an upregulation of PDH-E1α phosphorylation (Fig. 6c).

Generally a metabolic shift from glycolysis to glucose oxidation has been associated with an increase in the electron transport chain activity and ROS production [37] and oxidative stress in form of increased ROS levels drives cells into apoptosis [38]. Based on this, we hypothesize that the inhibition of PDKs during acute BRAF inhibitor treatment and during early development of resistance might be beneficial for patients as such that sufficient levels of ROS could be generated, which might drive cancer cells into apoptosis and thus delay or prevent resistance. In this context, it has previsouly been shown that the PDK inhibitor DCA negatively affects the growth of melanoma cells, which are sensitive or resistant to BRAF inhibition by suppressing glycolysis [39]. First we tested a PDK inhibitor with low micromolar cellular potency, AZD7545, as single treatment on both BRAF^V600E and BRAF^WT/NRAS^mut human melanoma cells. Interestingly, we detected a 30–40% decrease in cell growth in BRAF^V600E A375 and IGR37 cell lines (Fig. 7a). Interestingly the AZD7545-mediated suppression of growth could also be observed in BRAF/NRAS^mut cells, although, at later time points. This might be attributed to the slower growth of those cell lines (Fig. 7b), which also signal via the MAPK pathway. These data suggest that AZD7545, which we used at 10 μM, might be a valid alternative to Dichloroacetate (DCA), another PDK inhibitor used at millimolar concentrations. To investigate if AZD7545 could delay or influence the occurrence of BRAF inhibitor resistance we tested the combination of the two inhibitors on melanoma cell growth. A375 melanoma cells were transduced with lentiviral particles containing iRFP and were treated for 3 weeks either with 1 μM of PLX4032 or with 1 μM of PLX4032 in combination with 10 μM AZD7545. By measuring the intensity of the iRFP signal, we observed that cells treated with both inhibitors showed a more pronounced reduction in cellular growth (Fig. 8a). This drug combination was also tested on sphere models with similar results (Fig. 8b). It is tempting to speculate that the targeting of PDKs in combination with BRAF inhibitors could be a promising strategy to more efficiently hit melanoma cells or to delay the onset of drug resistance. In addition BRAFi-resistant A375 melanoma cells also showed reduction of growth upon AZD7445 treatment (Fig. 8c).