Metabolic stress regulates ERK activity by controlling KSR‐RAF heterodimerization

Abstract

Altered cell metabolism is a hallmark of cancer, and targeting specific metabolic nodes is considered an attractive strategy for cancer therapy. In this study, we evaluate the effects of metabolic stressors on the deregulated ERK pathway in melanoma cells bearing activating mutations of the NRAS or BRAF oncogenes. We report that metabolic stressors promote the dimerization of KSR proteins with CRAF in NRAS‐mutant cells, and with oncogenic BRAF in BRAF^V600E‐mutant cells, thereby enhancing ERK pathway activation. Despite this similarity, the two genomic subtypes react differently when a higher level of metabolic stress is induced. In NRAS‐mutant cells, the ERK pathway is even more stimulated, while it is strongly downregulated in BRAF^V600E‐mutant cells. We demonstrate that this is caused by the dissociation of mutant BRAF from KSR and is mediated by activated AMPK. Both types of ERK regulation nevertheless lead to cell cycle arrest. Besides studying the effects of the metabolic stressors on ERK pathway activity, we also present data suggesting that for efficient therapies of both genomic melanoma subtypes, specific metabolic targeting is necessary.

Introduction

The MAPK (mitogen‐activated protein kinase)/ERK (extracellular signal‐regulated protein kinase) signaling pathway coordinates diverse physiological processes and plays a significant role in the development of human cancers, including melanoma 1, 2. Somatic mutations in BRAF and NRAS genes are common in melanoma. They are found in about 45 and 17% of all cases, respectively (www.sanger.ac.uk/genetics/CGP/cosmic/). A glutamic acid substitution for a valine at position 600 (BRAF^V600E) accounts for more than 90% of the mutations in BRAF in cancer and results in a constitutive activation of BRAF kinase activity and downstream MEK‐ERK signaling 3. In NRAS‐mutated melanomas, point mutations impair GTPase activity of RAS, locking the protein into its active conformation, or render RAS insensitive to inactivation by the RAS GTPase‐activating proteins 4, 5. Even though both BRAF^V600E and mutant NRAS stimulate the ERK pathway, the routes to ERK activation differ. In contrast to the RAS‐independent BRAF^V600E signaling, mutant NRAS activates MEK kinases through CRAF 6. To ensure proper signal transduction, components of the ERK module are precisely and timely assembled into multiprotein complexes through dedicated interfaces and scaffold proteins 7, 8. One of these scaffolds is the kinase suppressor of Ras (KSR1 and KSR2 in mammals). KSR proteins constitutively bind MEK kinases and present them to RAF for phosphorylation and activation 9, 10. In addition to their scaffolding property, KSR proteins allosterically stimulate RAF catalytic activity independently of RAS 11, 12, 13.

Oncogenic NRAS and BRAF signaling alter metabolic pathways in cancer cells to support unrestricted neoplastic proliferation and cell survival 14. These metabolic alterations promote the production of biosynthetic precursors and macromolecules through the glycolysis pathway in the cytosol and the tricarboxylic acid cycle (TCA) in mitochondria 15. In proliferating cells, the TCA cycle functions predominantly as a biosynthetic pathway rather than a bioenergetic one, using glutamine as a main carbon source to produce biomass 16. Glucose and glutamine are major contributors to energy production also in cancer cells, but unlike normal cells, they rely less on the oxidative phosphorylation (OXPHOS) which takes place in mitochondria 15. The metabolic disparities between normal and cancer cells are viewed as promising targets for cancer therapy, and several drugs modifying cell metabolism were already approved as cancer treatments with many more currently undergoing clinical testing 17.

While there is a clear role for activated ERK signaling in promoting the metabolic switch supporting cancer cell proliferation 17, 18, 19, surprisingly little is known about the potential impact of metabolic perturbations on the activity of ERK signaling in cancers bearing oncogenic BRAF or NRAS mutations. In this study, we analyzed the response of the dysregulated ERK signaling to metabolic perturbations in these two most common genomic melanoma subtypes. We report that metabolic stressors promote RAF association with KSR proteins, enhancing ERK pathway activity. Nevertheless, there are significant differences in the response of the two genomic subtypes when a higher level of metabolic stress is induced, activating AMPK. The ERK pathway is even more stimulated in NRAS‐mutant cells, while its activity is abrogated in BRAF^V600E‐mutant cells. ERK pathway downregulation correlated with BRAF^V600E dissociation from KSR, an event mediated by activated AMPK. These two opposite types of ERK regulation nevertheless result in cell cycle arrest. Therefore, we evaluated the potential of metabolic targeting as a possible therapeutic strategy for melanoma. Our data indicated that NRAS‐mutant cells could be better candidates for metabolic targeting compared to BRAF^V600E‐mutant cells.

Results

Metabolic stressors promote RAS‐independent hyperactivation of the ERK pathway in NRAS‐mutant melanoma cells

We selected three metabolic stressors for our study that are targeting critical steps in glycolysis and OXPHOS, the two major cellular pathways of energy metabolism. To inhibit glycolysis, we used the hexokinase inhibitor 2‐deoxy‐D‐glucose (2DG), which is currently undergoing clinical testing in phase I/II trials for the treatment of several cancer types. To inhibit OXPHOS, we chose two inhibitors of the mitochondrial complex I, rotenone (Rot) and metformin (Met). Rotenone was selected for its specificity and well‐characterized mechanism of action, while metformin was selected because it is widely used in the clinic as a treatment for type 2 diabetes mellitus 17, 20, 21. We tested the response of MelJuso, NRAS‐mutant melanoma cells, to increasing concentrations of these drugs and observed that all three induced hyperactivation of the ERK pathway after 14 h of treatment (Fig 1A). ERK pathway activation was detectable already after 1 h of treatment (Fig EV1A), but the effect was stronger with more prolonged treatment. We also detected hyperactivation of the ERK pathway in MelJuso cells in response to other metabolic stressors such as 5‐thio‐D‐glucose (5TG, glycolysis inhibitor), 6‐aminonicotinamide (6AN, 6‐phosphogluconate dehydrogenase inhibitor), oligomycin A (ATP synthase inhibitor), antimycin A (inhibitor of electron transfer at complex III), and piericidin A (NADH dehydrogenase inhibitor) (Fig 1B).

Figure 1. Metabolic stressors promote CRAF kinase activity and activation of the MEK‐ERK pathway in NRAS‐mutant melanoma cells.

1. MelJuso, NRAS‐mutant cells were treated with 2DG, rotenone, and metformin at the indicated concentrations for 14 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2), total ERK1/2, phospho‐MEK1/2 (pMEK1/2), and total MEK1/2.
2. MelJuso cells were treated with 5TG, 6AN, oligomycin A, antimycin A, and piericidin A at the indicated concentrations for 14 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2) and total ERK1/2.
3. IPC298 and SKMel30, NRAS‐mutant cell lines were treated with 2DG, rotenone, and metformin at the indicated concentrations for 14 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2), total ERK1/2, phospho‐MEK1/2 (pMEK1/2), and total MEK1/2.
4. MelJuso cells were treated with 2DG (11 mM), rotenone (5 μM), and metformin (10 mM) for 4 h. Endogenous CRAF was immunoprecipitated and subjected to kinase assay in the presence of recombinant kinase‐dead MEK1 (K97M) (500 ng) and ATP (20 μM). The kinase reaction was Western‐blotted for endogenous CRAF, pMEK1, and total MEK1. CRAF activity (%) is relative to the untreated sample. Bars show mean % ± SEM (n = 6). Differences between untreated and the treated samples were examined with unpaired t‐test (2DG, *P = 0.0347). Below: MelJuso cells were treated with the metabolic stressors for 4 h and Western‐blotted for phospho‐CRAF (pCRAF) S338. CRAF levels in the lysate are shown as a loading control.

Source data are available online for this figure.

Figure EV1. The MEK‐ERK pathway activity is enhanced shortly after metabolic stressors administration and independently of RAS .

1. MelJuso cells were treated with 2DG, rotenone, and metformin at the indicated concentrations for 1 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2), total ERK1/2, phospho‐MEK1/2 (pMEK1/2), and total MEK1/2.
2. MelJuso cells were treated with 2DG (11 mM), rotenone (Rot; 5 μM), metformin (Met; 10 mM), and/or forskolin (Fo; 10 μM) for 4 h. Cell extracts were Western‐blotted for phospho‐MEK2 (pMEK2), total MEK2, and α‐tubulin.

Source data are available online for this figure.

Next, we evaluated ERK pathway activation after the treatment with our initial set of metabolic drugs in two other NRAS‐mutant melanoma cell lines, IPC298 and SKMel30. As shown in Fig 1C, MEK1/2 and ERK1/2 were hyperactivated in both cell lines as well. Since constitutively active RAS and CRAF are the upstream components activating MEK kinases in RAS‐mutant cells, we measured CRAF kinase activity after a 4‐h treatment with the metabolic stressors. Endogenous CRAF was immunoprecipitated from MelJuso cells and used in an in vitro kinase assay in the presence of recombinant kinase‐dead MEK1 (K97M) as a substrate and ATP. As shown in Fig 1D, CRAF kinase activity increased upon treatment with the metabolic stressors. The phosphorylation of the N‐region serine residue S338 that is required for CRAF activation 22 was also enhanced (Fig 1D). Next, we analyzed the binding of CRAF to mutant NRAS and, surprisingly, we observed that CRAF interacts less with NRAS, while MEK was more activated after 2 h of treatment with the metabolic stressors, especially with rotenone (Fig 2A). CRAF is known to be subject to negative regulation by protein kinase A (PKA), preventing its binding to RAS‐GTP 23. The adenylyl cyclase and PKA activator forskolin (Fo) was therefore used as a positive control in this experiment, efficiently disrupting the interaction between CRAF and mutant NRAS (Fig 2A). The reduced interaction of CRAF with NRAS after the treatment with the metabolic stressors was also observed 14 h post‐treatment, and the effect was more pronounced compared to the 2‐h time point for 2DG and metformin (Fig 2A). To confirm that the activation of the MEK‐ERK pathway after the treatment with the metabolic stressors was independent of NRAS, we tested the effect of the metabolic stress in MelJuso cells depleted of NRAS by RNA interference. While the downregulation of NRAS markedly attenuated basal MEK2 kinase activation, the NRAS‐depleted cells treated with 2DG still exhibited increased MEK2 kinase activity compared to NRAS‐depleted cells (Fig 2B). Additionally, the CRAF mutant (CRAF^R89L) that cannot bind to RAS 24 had an enhanced kinase activity in cells treated with 2DG compared to the untreated sample (Fig 2C). Lastly, the inhibitory action of forskolin on MEK activity through the disruption of CRAF binding with RAS‐GTP was overcome when the metabolic stressors were co‐added (Fig EV1B). Together, these data suggested that CRAF binding to NRAS is not necessary for the enhanced CRAF kinase activity by the metabolic stressors.

Figure 2. Metabolic stressors promote the dissociation of the 14‐3‐3 proteins from the N‐terminus of CRAF and induce a RAS‐independent activation of CRAF in NRAS‐mutant cells.

1. MelJuso cells were treated with 2DG (11 mM), rotenone (Rot; 5 μM), metformin (Met; 10 mM), and forskolin (Fo; 10 μM) for 2 h or with 2DG (5.5 mM), rotenone (Rot; 5 μM), metformin (Met; 5 mM), and forskolin (Fo; 10 μM) for 14 h. Endogenous NRAS was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous NRAS and CRAF. Endogenous NRAS, CRAF, phospho‐MEK1/2 (pMEK1/2), and total MEK2 levels in the cell lysates are also shown.
2. MelJuso cells were transfected with two different siRNAs against NRAS (si#1, si#2) and their combination (siPool) or with a non‐targeted siRNA sequence (NTsiRNA). A double siRNA transfection (48 and 72 h) was performed to have an optimal knockdown. Cells were treated with 2DG (11 mM) for 4 h. Cell lysates were Western‐blotted for NRAS, phospho‐MEK2 (pMEK2), total MEK2, and α‐tubulin.
3. Myc‐epitope‐tagged CRAF^WT or CRAF^R89L were transfected into MelJuso cells. After 24 h, cells were treated with 2DG (11 mM) for 4 h. Myc‐tagged CRAF^WT and myc‐tagged CRAF^R89L were immunoprecipitated and subjected to kinase assay in the presence of recombinant kinase‐dead MEK1 (K97M) (500 ng) as a substrate and ATP (20 μM). CRAF^R89L activity (%) is relative to the untreated CRAF^R89L sample. Bars show mean % ± SEM (n = 3). Difference between untreated and 2DG‐treated samples was examined with unpaired t‐test (**P = 0.0076).
4. MelJuso cells were treated with 2DG (11 mM), rotenone (Rot; 5 μM), metformin (Met; 10 mM), and forskolin (Fo; 10 μM) for 2, 4, and 8 h. Cell extracts were Western‐blotted for phospho‐CRAF (pCRAF) S259. CRAF levels in the lysate are shown as a loading control.
5. MelJuso cells were treated with 2DG (11 mM) or rotenone (Rot; 5 μM) for 4 h. Endogenous 14‐3‐3 proteins were immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous 14‐3‐3 and CRAF. Endogenous 14‐3‐3 and CRAF levels in the cell lysates are also shown.
6. FLAG‐epitope‐tagged CRAF^WT and FLAG‐epitope‐tagged CRAF^S259A were transfected into MelJuso cells. Cells were treated with 2DG (11 mM) for 4 h. FLAG‐tagged CRAF was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous 14‐3‐3 and FLAG‐CRAF. Endogenous 14‐3‐3 and FLAG‐CRAF levels in the cell lysates are also shown.

Source data are available online for this figure.

The commonly accepted view is that the interaction with RAS‐GTP serves to displace dimeric 14‐3‐3 proteins from phosphorylated CRAF^S259, causing conformational changes in CRAF that are necessary for its stable activation 25, 26. As we found that metabolic stressors promote CRAF activation independently of RAS, next we investigated the phosphorylation status of CRAF^S259 and the binding of the 14‐3‐3 proteins to the kinase. Dephosphorylation of CRAF^S259 was detectable after 2 h of treatment with 2DG and rotenone and augmented over the time (Fig 2D), while an 8‐h treatment with metformin was necessary to observe a decrease in phosphorylation of this site (Fig 2D). These differences between the two inhibitors might come from the fact that rotenone is an irreversible and much more potent mitochondrial complex I inhibitor than metformin and thus cellular effects are seen earlier. As 2DG and rotenone induced dephosphorylation of CRAF^S259 at shorter time points than metformin, in the next step we analyzed the binding of the 14‐3‐3 proteins to CRAF after a 4‐h treatment with these two inhibitors. As shown in Fig 2E, 2DG and rotenone diminished the binding of 14‐3‐3 proteins to CRAF. The CRAF mutant (CRAF^S259A), which cannot bind 14‐3‐3 in the CR2 domain 27, did not display any differences in 14‐3‐3 binding to CRAF between control and 2DG‐treated samples (Fig 2F). These data indicated that more CRAF molecules were in a pro‐active conformation after the treatment with the metabolic stressors compared to the non‐treated cells. However, it was unclear how the dissociation of the 14‐3‐3 proteins from the CR2 domain of CRAF was accomplished as the depletion of NRAS did not prevent enhanced MEK2 activation after 2DG (Fig 2B) and the CRAF^R89L mutant was activated in response to 2DG (Fig 2C). Thus, these data suggested that further regulatory events were at play.

Metabolic stressors promote CRAF interaction with KSR proteins in NRAS‐mutant melanoma cells

Considering the RAS‐independent nature of CRAF activation under the conditions of metabolic stress, we searched the literature for RAS‐independent mechanisms of RAF activation. The kinase suppressor of Ras (KSR), one of the scaffold proteins of the RAF pathway, has been described to activate RAF in a RAS‐independent manner 12. KSR proteins bind to MEK constitutively, while their binding to RAF is transient 13. First, we tested by immunoprecipitations whether 2DG and rotenone modulate the interaction between CRAF and the KSR proteins (KSR1 and KSR2 in mammals). We overexpressed V5‐epitope‐tagged KSR1^WT and KSR2^WT in MelJuso cells, immunoprecipitated KSRs using anti‐V5 antibody, and found that 2DG and rotenone promoted the interaction of CRAF with the KSR proteins (Fig 3A). We also tested the possibility that metabolic stressors might support dimerization of CRAF with BRAF, as CRAF‐BRAF heterodimers could also contribute to enhanced MEK activation. However, we did not observe any interaction of these two kinases in response to 2DG (Fig EV2A). The multi‐kinase inhibitor sorafenib was used as a positive control, as it can induce dimerization of CRAF with BRAF in NRAS‐mutant cells 28. To form dimers, KSR proteins must partially dissociate from the 14‐3‐3 proteins and localize to the plasma membrane 29, 30, 31. We immunoprecipitated endogenous 14‐3‐3 and observed a significant dissociation of these proteins from KSR1 upon 2DG treatment (Fig EV2B), suggesting that KSR proteins might be more capable of binding CRAF in response to metabolic perturbations.

Figure 3. KSR dimerizes with CRAF and enhances its kinase activity under metabolic stress conditions.

1. V5‐epitope‐tagged KSR1^WT and V5‐epitope‐tagged KSR2^WT were transfected into MelJuso cells. After 24 h, cells were treated with 2DG (11 mM) or rotenone (Rot; 5 μM) for 4 h. V5‐tagged KSR1^WT and KSR2^WT were immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for V5 and endogenous CRAF. V5‐tagged KSR1^WT, KSR2^WT, endogenous CRAF, phospho‐ERK1/2 (pERK1/2), and total ERK1/2 levels in the cell lysates are also shown.
2. FLAG‐epitope‐tagged CRAF^WT and V5‐epitope‐tagged KSR1^WT or KSR2^WT were transfected into HEK293 cells. After 24 h, cells were treated with 2DG (11 mM) or rotenone (Rot; 5 μM) for 4 h. FLAG‐tagged CRAF was immunoprecipitated (IP) and the immunocomplexes were Western‐blotted for V5 and FLAG. V5‐tagged KSR1^WT, KSR2^WT, FLAG‐tagged CRAF, phospho‐ERK1/2 (pERK1/2), and total ERK1/2 levels in the cell lysates are also shown.
3. FLAG‐epitope‐tagged CRAF^WT was transfected into HEK293 cells. After 24 h, the cells were treated with 2DG (11 mM) and/or PD184352 (1 μM) for 4 h. FLAG‐tagged CRAF was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous KSR1 (left panel), endogenous KSR2 (right panel), and FLAG. Endogenous KSR1 (left panel), endogenous KSR2 (right panel), and FLAG levels in the cell lysates are also shown.
4. V5‐epitope‐tagged KSR1^WT, KSR1^R665H, KSR2^WT, or KSR2^R718H were transfected into MelJuso cells. After 24 h, the cells were treated with 2DG (11 mM) for 4 h. V5‐tagged KSR1^WT, KSR2^WT, and their respective mutants were immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for V5, phospho‐MEK1/2 (pMEK1/2), and MEK2.
5. V5‐epitope‐tagged KSR1^WT, KSR1^R665H, KSR2^WT, KSR2^R718H, and FLAG‐epitope‐tagged CRAF^WT or CRAF^K375M were transfected into HEK293 cells. After 24 h, cells were treated with rotenone (Rot; 5 μM) for 4 h. Cell extracts were Western‐blotted for V5‐tagged KSR1^WT, KSR2^WT, FLAG, phospho‐ERK1/2 (pERK1/2), and total ERK1/2.

Source data are available online for this figure.

Figure EV2. 2DG does not promote CRAF association with BRAF and induces the dissociation of the 14‐3‐3 proteins from KSR1.

1. MelJuso cells were treated with 2DG (11 mM) or sorafenib (10 μM) for 4 h. Endogenous CRAF was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous CRAF and BRAF. Endogenous CRAF, BRAF, phospho‐MEK1/2 (pMEK1/2), and MEK2 levels in the cell lysates are also shown.
2. MelJuso cells were treated with 2DG (11 mM) for 4 h. Endogenous 14‐3‐3 was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous 14‐3‐3 and endogenous KSR1. Endogenous 14‐3‐3, endogenous KSR1, phospho‐MEK2 (pMEK2), and MEK2 levels in the cell lysates are also shown.

Source data are available online for this figure.

To determine whether the enhanced interaction of CRAF with KSR proteins in NRAS‐mutant cells is a general mechanism that can be extended to other cell types, FLAG‐epitope‐tagged CRAF^WT and V5‐epitope‐tagged KSR1^WT and KSR2^WT were overexpressed in HEK293 cells, and FLAG‐CRAF^WT was immunoprecipitated after treatment with 2DG or rotenone. 2DG did not promote CRAF binding to KSR proteins in HEK293 cells while rotenone did (Fig 3B). However, the binding of CRAF to the KSR proteins after 2DG in HEK293 could be induced when the MEK inhibitor PD184352 was added (Fig 3C). PD184352 on its own only slightly induced CRAF binding to KSR proteins, demonstrating that the enhanced heterodimer formation after 2DG and PD184352 treatments was due to 2DG action (Fig 3C). Enhanced ERK pathway activation by 2DG probably increased negative feedback from ERK on RAF‐KSR dimer formation, and the inhibition of this feedback using a MEK inhibitor stabilized 2DG‐induced dimerization. These data indicated that 2DG and rotenone were able to induce the interaction between CRAF and KSR proteins in different cellular contexts.

The dimerization of RAF with KSR is mediated by their kinase domains and is critical for allosteric activation of the RAF kinase activity 12. To confirm that the dimerization of CRAF with the KSR proteins was required for the increased CRAF kinase activity in response to metabolic drugs, we engineered two KSR mutants with impaired transactivation potential: KSR1^R665H and KSR2^R718H. The critical arginine residue is located at the center of the dimer interface between RAF and KSR 12, 32. WT and transactivation‐impaired forms of KSR were expressed in MelJuso cells and immunoprecipitated. As KSR and MEK proteins form constitutive complexes 9, 33, MEK kinase phosphorylation was evaluated in the immunoprecipitates. As depicted in Fig 3D, MEK phosphorylation was enhanced in 2DG‐treated KSR1^WT‐ and KSR2^WT‐transfected cells, but not in 2DG‐treated KSR1^R665H‐ and KSR2^R718H‐transfected cells. These data suggested that KSR proteins stimulate CRAF kinase activity. Moreover, the co‐expression of CRAF^WT with WT or mutant KSR constructs in HEK293 treated with rotenone induced hyperactivation of the ERK pathway in CRAF^WT plus KSR1^WT‐ or KSR2^WT‐transfected cells, but not in CRAF^WT plus KSR1^R665H‐ or KSR2^R718H‐transfected cells (Fig 3E, left panel). We performed a similar experiment by replacing CRAF^WT with its kinase‐dead (KD) version, CRAF^K375M, and observed that the pathway was not activated when cells were treated with rotenone (Fig 3E, right panel). In fact, rotenone even decreased MEK phosphorylation when CRAF^K375M was expressed, presumably because the overexpressed KD CRAF sequesters KSR from endogenous CRAF. These data indicated that the dimerization and transactivation of CRAF by KSR cause the hyperactivation of the ERK pathway after metabolic stress.

NRAS‐ and BRAF^V600E‐mutant melanoma cells are differentially regulated by metabolic stressors

The above‐described experiments showed that metabolic perturbations enhance the activity of the ERK pathway by promoting the interactions between KSR and CRAF proteins. Next, we were interested in comparing the response to metabolic stressors of NRAS‐mutant cells to that of the more common genomic melanoma subtype bearing the oncogenic BRAF^V600E mutation. In BRAF^V600E‐mutant A375 melanoma cells, ERK activity was enhanced upon the treatment with 2DG, rotenone, and metformin albeit to a lesser extent than in RAS‐mutant cells (Figs 4A and EV3A). The enhanced ERK activity after 2DG was dose‐dependent as the pathway was downregulated at the highest concentration tested (11 mM) and this response was similar in RKO, another BRAF^V600E‐mutant cell line (Figs 4A and EV3A). Next, we examined the response of A375 to other metabolic stressors (Fig 4B), as we did in MelJuso cells (Fig 1B). An increased activity of the ERK pathway was observed in response to 5TG, 6AN, oligomycin A, and piericidin A but not to antimycin A (Fig 4B). However, similarly to 2DG, a decreased ERK pathway activity was detectable at the highest doses used (Fig 4B).

Figure 4. Under metabolic stress, KSR dimerizes with oncogenic BRAF stimulating ERK pathway activation, yet the pathway is downregulated when the stress is higher.

1. A375, BRAF^V600E‐mutant melanoma cells were treated with 2DG, rotenone, and metformin at the indicated concentrations for 14 h. RKO, BRAF^V600E‐mutant colorectal cells were treated with 2DG at the indicated concentrations for 14 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2) and total ERK1/2.
2. A375 cells were treated with 5TG, 6AN, oligomycin A, antimycin A, and piericidin A at the indicated concentrations for 14 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2) and total ERK1/2.
3. A375 cells were treated with 2DG (5.5 mM) or rotenone (Rot; 5 μM) for 4 h. Endogenous BRAF^V600E was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous BRAF^V600E and endogenous KSR1 and KSR2. Endogenous BRAF^V600E, KSR1, and KSR2 levels in the cell lysates are also shown.
4. HA‐epitope‐tagged BRAF^V600E was transfected into HEK293 cells. After 24 h, cells were treated with 2DG (11 mM) for 4 h. HA‐tagged BRAF^V600E was immunoprecipitated (IP) with HA antibody, and the immunocomplexes were Western‐blotted for HA and endogenous KSR1 and KSR2. HA and endogenous KSR1 and KSR2 levels in the cell lysates are also shown.
5. A375 cells were treated with 2DG (5.5 and 11 mM) for 4 h. Endogenous BRAF^V600E was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous BRAF^V600E and KSR2. Endogenous BRAF^V600E, KSR2, phospho‐ERK1/2 (pERK1/2), total ERK1/2, phospho‐AMPKα T172 (pAMPKα), and AMPKα levels in the cell lysates are also shown.
6. MelJuso, IPC298, and SKMel30, NRAS‐mutant melanoma cells and A375, RVH421, RKO, BRAF^V600E‐mutant cells were treated with 2DG (5.5 mM) and/or rotenone (Rot; 5 μM) and/or metformin (Met; 5 mM) for 14 h. Cell extracts were Western‐blotted for phospho‐MEK1/2 (pMEK1/2) and total MEK2.
7. MelJuso and A375 cells were treated with 2DG (5.5 mM), rotenone (Rot; 5 μM), and metformin (Met; 5 mM) for 14 h. Cell extracts were Western‐blotted for phospho‐AMPKα T172 (pAMPKα), total AMPKα, and α‐tubulin.
8. ATP levels in MelJuso and A375 lysates were measured using a luciferase‐based assay after 14 h of treatment with the indicated stressors. Bars show mean % ± SEM (n = 4). Differences between control and experimental groups were evaluated by Student's t‐test: MelJuso cells [2DG (***P < 0.0001); rotenone (***P = 0.0009); 2DG + rotenone (***P < 0.0001); metformin (P = 0.4480); 2DG + metformin (**P = 0.0078)]. A375 cells [2DG (P = 0.0805); rotenone (P = 0.9770); 2DG + rotenone (**P = 0.0041); metformin (P = 0.9749); 2DG + metformin (**P = 0.0014)].

Source data are available online for this figure.

Figure EV3. Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) changes in response to glucose, oligomycin, and 2DG in A375 and MelJuso cells pre‐treated with the metabolic stressors.

1. A375 and RKO, BRAF^V600E‐mutant melanoma cells and MelJuso and IPC298, NRAS‐mutant melanoma cells were treated with 2DG at the indicated concentrations for 14 h. Cell extracts were Western‐blotted for phospho‐ERK1/2 (pERK1/2) and total ERK1/2.
2. ECAR and OCR measurements performed in A375 cells pre‐treated for 4 h with 2DG (11 mM) and/or rotenone (5 μM) and/or metformin (10 mM). Points represent means ± SEM (n = 3).
3. ECAR and OCR measurements performed in MelJuso cells pre‐treated for 4 h with 2DG (11 mM) and/or rotenone (5 μM) and/or metformin (10 mM). Points represent means ± SEM (n = 3).

Source data are available online for this figure.

As we did in the case of CRAF in NRAS‐mutant cells, we tested whether oncogenic BRAF and KSR dimerize upon the treatment with our initial set of metabolic stressors in BRAF^V600E‐mutant cells. Interestingly, oncogenic BRAF associated with endogenous KSR1 and especially KSR2 in response to 2DG and rotenone in A375 cells (Fig 4C). HA‐epitope‐tagged BRAF^V600E transiently transfected in HEK293 cells also dimerized with the KSR proteins after the treatment with 2DG (Fig 4D). HA‐BRAF^V600E bound more strongly to endogenous KSR1 than KSR2 in response to metabolic stress in HEK293 cells (Fig 4D), while the opposite was seen in A375 cells (Fig 4C). However, we do not know how cells regulate the distinct dimer formation between the two KSR isoforms and mutant BRAF. It is of note that 2DG could induce the dimerization when BRAF^V600E was transfected into HEK293 cells but not when wild‐type CRAF was overexpressed in the same cell line without the addition of PD184352 (Fig 3B and C). We attributed this difference to the fact that oncogenic BRAF, but not wild‐type CRAF, can bypass the inhibitory effect of the negative feedback regulation by ERK 34.

As we observed downregulation of the ERK pathway at the highest used concentrations of 2DG (11 mM) (Fig 4A) and other inhibitors (Fig 4B) in A375 cells, we also investigated the association of BRAF^V600E with KSR2 when the cells were treated with increasing concentrations of 2DG. As shown in Fig 4E, 2DG at 5.5 mM promoted KSR2 binding to oncogenic BRAF and enhanced ERK pathway activation, while a higher dose (11 mM) did not induce heterodimerization of the two proteins. Thus, it appeared that low and high levels of metabolic stress differently impact the behavior of the ERK pathway in BRAF^V600E‐mutant cells while a uniform hyperactivation of the pathway by metabolic stress was observed in NRAS‐mutant cells (Fig 1A–C).

In the next set of experiments, we investigated possible reasons why BRAF^V600E‐mutant cells exhibit a reduced ERK pathway activity when exposed to high metabolic stress. Apart from using high doses of 2DG and other metabolic drugs, we noticed that we could induce high metabolic stress by combining metabolic stressors hitting different pathways of cell energy metabolism. As shown in Fig 4F, the combination of 2DG with a mitochondrial complex I inhibitor, rotenone or metformin, downregulated MEK activity in three different BRAF^V600E‐mutant cell lines while it further enhanced it in NRAS‐mutant cells. To ensure that the metabolic stress induced by the combinations is higher than the stress caused by individual drugs, we measured the extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) of the cells after a 4‐h treatment with the metabolic stressors. The drug combination decreased the basal levels of ECAR and OCR in both A375 and MelJuso cells (Fig EV3B and C). Furthermore, cells treated with the combinations of metabolic stressors responded only slightly to glucose, oligomycin, and 2DG when challenged using the Seahorse Glycolytic Stress Test Kit, suggesting that the metabolic capacity of these cells was low (Fig EV3B and C). We also evaluated the activity of the central energy sensor, AMP‐activated protein kinase (AMPK) in both cell types. We found that the kinase was activated after the treatment with the metabolic stressors alone or in combination in NRAS‐mutant cells but not in BRAF^V600E‐mutant cells. Here, AMPK was activated solely in response to a high concentration of 2DG (Fig 4E) or the combination of metabolic stressors (Fig 4G), leading to a significant ATP depletion (Fig 4H).

The downregulation of the BRAF signaling correlated with AMPK activation in the cells. We, therefore, investigated whether AMPK could play a role in the reduced ERK signaling in BRAF^V600E‐mutant cells. Two‐hour treatment of A375 cells with 2DG plus rotenone activated AMPK and induced the dissociation of KSR1 from BRAF^V600E (Fig 5A). The combination of 2DG and metformin did not activate AMPK, most probably because the 2‐h treatment with the drugs was not long enough to activate the kinase and we did not observe the dissociation of KSR1 from BRAF^V600E (Fig 5A). Interestingly, in 2DG plus metformin‐treated sample, AMPK was found to strongly bind to BRAF^V600E, although the kinase was not activated (Fig 5A). A more prolonged treatment of the cells with 2DG and metformin led to AMPK activation and a partial unbinding of KSR1 from BRAF^V600E (Fig 5B). Under these conditions, we could hardly detect any AMPK in the immunoprecipitated fraction of the 2DG and metformin‐treated sample (Fig 5B). Next, we wanted to determine whether AMPK plays a role in the dissociation of KSR1 from mutant BRAF. HEK293 cells were transfected with HA‐BRAF^V600E, depleted of AMPKα by siRNAs, and the mutant BRAF protein was immunoprecipitated and assessed for KSR1 binding. As shown in Fig 5C, the binding of KSR1 with BRAF^V600E after 2DG plus rotenone could be rescued, even strengthened, when AMPKα was depleted by RNAi. Furthermore, the activity of the ERK pathway was downregulated in 2DG plus rotenone‐treated sample but not in 2DG plus rotenone‐treated cells with depleted AMPKα levels (Fig 5C), suggesting that AMPK plays a critical role in regulating the ERK pathway activity by modulating the dimerization of oncogenic BRAF with KSR1 upon metabolic stress. Importantly, we confirmed these data in A375 cells where the depletion of AMPKα in metabolically highly stressed cells also preserved the dimer formation between endogenous KSR2 and endogenous BRAF^V600E (Fig 5D).

Figure 5. AMPK mediates the dissociation of KSR from mutant BRAF but not CRAF in high metabolic stress conditions.

1. HA‐epitope‐tagged BRAF^V600E was transfected into HEK293 cells. After 24 h, cells were treated with 2DG (11 mM) and/or rotenone (Rot; 5 μM) and/or metformin (Met; 10 mM) for 2 h. HA‐tagged BRAF^V600E was immunoprecipitated (IP) with HA antibody, and the immunocomplexes were Western‐blotted for HA, endogenous KSR1, and endogenous AMPKα. HA, endogenous KSR1, phospho‐AMPKα T172 (pAMPKα), and total AMPKα levels in the cell lysates are also shown.
2. HA‐epitope‐tagged BRAF^V600E was transfected into HEK293 cells. After 24 h, cells were treated with 2DG (11 mM) and/or metformin (Met; 10 mM) for 3 h. HA‐tagged BRAF^V600E was immunoprecipitated (IP) with HA antibody, and the immunocomplexes were Western‐blotted for HA, endogenous KSR1, and endogenous AMPKα. HA, endogenous KSR1, phospho‐AMPKα T172 (pAMPKα), and total AMPKα levels in the cell lysates are also shown.
3. HA‐epitope‐tagged BRAF^V600E and V5‐epitope‐tagged KSR1^WT were transfected into HEK293 cells; 6 h later, one of the transfected dishes was transfected with an AMPKα siRNA. After 48 h, cells were treated with 2DG (11 mM) and rotenone (Rot; 5 μM) for 2 h. HA‐tagged BRAF^V600E was immunoprecipitated (IP) with HA antibody, and the immunocomplexes were Western‐blotted for HA and V5. HA, V5, phospho‐ERK1/2 (pERK1/2), and total ERK1/2 levels in the cell lysates are also shown. The efficiency of the AMPKα knockdown is shown in the lower panel.
4. One of the dishes of A375 cells was transfected with an AMPKα siRNA. After 48 h, cells were treated with 2DG (5.5 mM) and rotenone (Rot; 5 μM) for 2 h. Endogenous BRAF^V600E was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous BRAF^V600E and endogenous KSR2. Endogenous BRAF^V600E and KSR2 levels in the cell lysates are also shown. The efficiency of the AMPKα knockdown is presented in the lower panel.
5. MelJuso cells were treated with 2DG (11 mM) and/or rotenone (Rot; 5 μM) and/or metformin (Met; 10 mM) for 2 h. Endogenous CRAF was immunoprecipitated (IP), and the immunocomplexes were Western‐blotted for endogenous CRAF, KSR2, and AMPKα. Endogenous CRAF, KSR2, phospho‐AMPKα T172 (pAMPKα), total AMPKα, phospho‐ERK1/2 (pERK1/2), and total ERK1/2 levels in the cell lysates are also shown.

Source data are available online for this figure.

Considering the role of AMPK in controlling the dimerization of oncogenic BRAF with KSR, we also investigated whether AMPK could be found at the CRAF‐KSR complexes under high metabolic stress conditions in NRAS‐mutant cells. Surprisingly, AMPK localized at the complexes in these conditions (Fig 5E), and in contrast to BRAF^V600E‐mutant cells, we did not observe any dissociation of KSR2 from CRAF (Fig 5E). In fact, increased stress promoted AMPK activation, AMPK binding to CRAF, and CRAF association with KSR2 and hence also the ERK pathway hyperactivation (Fig 5E).

NRAS‐ and BRAF^V600E‐mutant cells regulate the ERK pathway to promote cell cycle arrest in response to metabolic stress

The above‐presented data demonstrated that NRAS‐ and BRAF^V600E‐mutant cells differently regulate the ERK pathway when facing high metabolic stress; while the ERK pathway was strongly hyperactivated in NRAS‐mutant cells, it was inhibited in BRAF^V600E‐mutant cells. Thus, our next set of experiments was to investigate the possible reasons for this differential regulation.

The RAF‐ERK pathway is one of the key signal transduction pathways that participate in cell cycle control. We first analyzed the progression through the S phase of the cell cycle in MelJuso and A375 cells. Importantly, we confirmed that 2DG induced a comparable inhibition of the glycolysis in the two cell lines and that rotenone and metformin promoted a similar decrease in the oxygen consumption at the concentrations used (Fig 6B). 2DG as a single agent significantly diminished the number of cells in S phase in MelJuso but not in A375 cells (Fig 6A). To better understand this disparity between the two cell lines in response to 2DG, we analyzed the percentage of cells in each cell cycle phase using propidium iodide DNA staining. As shown in Fig 6C, MelJuso cells treated with 2DG massively accumulated in G0/G1 phase (65% vs. 43% control). It has been demonstrated that a sustained activation of the ERK pathway is necessary for G1 to S phase progression 35, 36. However, the duration and the magnitude of the activation both have to be controlled as not only the inhibition but also too strong activation of this pathway can lead to a reversible or permanent cell cycle arrest 37, 38. In this study, we showed that 2DG promoted a stronger activation of the ERK pathway in MelJuso than in A375 cells (Figs 1A and 4A, and EV3A) and the difference became even more prominent when the level of metabolic stress was higher (Fig 4F). These data suggested that in MelJuso cells, the enhanced ERK pathway activation after 2DG could be high enough to induce cell cycle arrest in G0/G1 phase, but not in A375 where the activation was weaker. Indeed, it did not seem to be sufficient to promote cell cycle arrest or a significant increase in cell proliferation (Fig 6A and C). High intensity or long‐term ERK activation can cause cell cycle arrest due to the induction and accumulation of the cell cycle inhibitor p21 39, 40, promoted by an ERK‐driven transcriptional induction and increased protein stability 41, 42. As shown in Fig 6D, p21 expression is highly induced by 2DG in MelJuso but not in A375 cells. Together, these data indicated that NRAS‐mutant cells hyperactivate the ERK pathway in response to metabolic stress to efficiently block cell cycle progression. This strategy was not adopted in BRAF^V600E‐mutant cells that still proliferated when subjected to mild metabolic stress (Fig 6A and C). However, when the metabolic stressors were combined, and the levels of stress were higher (Fig 4G and H), the ERK pathway was downregulated (Fig 4F) and the cell cycle progression arrested in BRAF^V600E‐mutated cells (Fig 6A).

Figure 6. Metabolic targeting might not be an efficient therapeutic strategy for BRAF^V600E‐mutant cells.

1. A375 and MelJuso cells were treated with 2DG (5.5 mM) and/or rotenone (5 μM) and/or metformin (5 mM) for 24 h followed by incubation with Edu nucleotide analog for 30 min. EdU‐positive cells were visualized using the Click‐iT EdU Alexa Fluor 488 Imaging Kit and quantified by flow cytometry. Bars show mean % ± SEM of EdU‐positive cells (n = 4). Differences between control and experimental groups were evaluated by Student's t‐test: A375 cells (2DG (P = 0.1108); rotenone (P = 0.3295); 2DG + rotenone (***P = 0.0002); metformin (P = 0.4669); 2DG + metformin (***P = 0.0003). MelJuso cells (2DG (**P = 0.0046); rotenone (P = 0.0537); 2DG + rotenone (**P = 0.0087); metformin (P = 0.7328); 2DG + metformin (**P = 0.0032).
2. ECAR measurements representing the glycolytic flux in control and 2DG‐treated cells and OCR measurements depicting the basal respiration in control, rotenone, and metformin‐treated cells (n = 3). Cells were pre‐treated for 4 h with 2DG (11 mM) and/or rotenone (5 μM) and/or metformin (10 mM). Bars show means ± SEM (n = 3). Differences between the groups were evaluated by Student's t‐test: ECAR (A375 control vs. MelJuso control (*P = 0.0174)); (A375 2DG vs. MelJuso 2DG (P = 0.1225)) OCR (A375 control vs. MelJuso control (*P = 0.0308)); (A375 rotenone vs. MelJuso rotenone (P = 0.9602)); (A375 metformin vs. MelJuso metformin (P = 0.0648)).
3. A375 and MelJuso cells were treated with 2DG (5.5 mM) for 24 h and processed as described in the Materials and Methods section. Bars show mean % ± SEM of number of cells in each cell cycle phases (G0/G1, S, and G2/M) (n = 3). Differences between control and 2DG were evaluated by Student's t‐test: A375 (G0/G1: P = 0.1658; S: P = 0.4641; G2/M: P = 0.1880), MelJuso (G0/G1: ***P < 0.0001; S: **P = 0.0084; G2/M: P = 0.1558).
4. A375 and MelJuso cells were treated with 2DG (5.5 mM) for 24 h. Cell extracts were Western‐blotted for p21^Cip1 and α‐tubulin.
5. Percentage of apoptotic A375 cells after 48 h of treatment with 2DG (5.5 mM) and/or rotenone (5 μM) and/or metformin (5 mM). Bars show means ± SEM of PI‐positive cells (n = 3). Student's t‐test: control vs. rotenone (***P = 0.0002); rotenone vs. 2DG + rotenone (***P = 0.0003).
6. A375 cells were treated with 2DG (5.5 mM) and/or rotenone (Rot; 5 μM) for 14 h. Cell extracts were Western‐blotted for cleaved PARP, PARP, cleaved caspase 3, caspase 3, cleaved caspase 9, and caspase 9. Below: Amount of cytochrome c released in the cytosolic fraction of A375 cells.
7. Quantification of intracellular glutamine. Before metabolite extraction, cells were treated with 2DG (5.5 mM), rotenone (5 μM), and/or metformin (5 mM) for 14 h. Bars show mean % ± SEM (n = 3). Differences between control and each group were evaluated by Student's t‐test: 2DG (P = 0.0778); 2DG + rotenone (***P < 0.0001); 2DG + metformin (P = 0.0792).
8. Percentage of apoptotic A375 cells after 48 h of treatment with 2DG (5.5 mM), rotenone (5 μM), and/or metformin (5 mM) ± BPTES (10 μM). Bars show means ± SEM of PI‐positive cells (n = 3). Differences between the groups were evaluated by Student's t‐test: 2DG + rotenone vs. 2DG + rotenone + BPTES (***P < 0.0001); 2DG + metformin vs. 2DG + metformin + BPTES (**P = 0.0011).

Source data are available online for this figure.

Analysis of the therapeutic potential of metabolic targeting in malignant melanoma

Considering the different strategies to block cell cycle progression in the two genomic melanoma subtypes, we decided to determine whether the metabolic stressors, used separately or combined, could constitute a potential therapeutic strategy for the treatment of NRAS‐ or BRAF^V600E‐mutant cells. Most classical small molecule inhibitors of cellular respiration, including rotenone, exhibit significant toxicity and are not suitable for therapeutic use. In contrast, metformin, while being capable of inhibiting complex I activity 43, is commonly used in the clinic as the treatment of type II diabetes 20. Therefore, we were particularly interested in determining the effect of 2DG as a single treatment and in combination with metformin, but we also used the combination of 2DG plus rotenone for comparison. 2DG, metformin, and their combination did not display any acute toxicity toward A375 and MelJuso cells in vitro (Figs 6E and EV4A). Surprisingly, the concomitant treatment of 2DG with rotenone improved A375 and MelJuso viability compared to rotenone‐treated cells (Figs 6E and EV4A). This effect was also observed with 5TG, another glucose analog and glycolysis inhibitor. However, we did not observe a similar protective effect when rotenone was combined with inhibitors of other essential carbon sources, for example, the glutaminase inhibitor BPTES or the inhibitor of fatty acid β‐oxidation etomoxir (Fig EV4B), or when 2DG was replaced by glucose‐free medium (Fig EV4C). The unexpected protective effect of 2DG against rotenone‐induced toxicity was confirmed when we analyzed the activation of caspases, specific proteases that are required for the execution of apoptotic cell death. In A375 cells, rotenone induced the cleavage of procaspase‐9 and procaspase‐3 and of poly(ADP‐ribose) polymerase (PARP), a cellular target of caspase‐3. Activation of caspases and PARP cleavage were reduced when rotenone was applied concomitantly with 2DG (Fig 6F). We also examined the release of cytochrome c into the cytosol, another apoptotic feature, and observed that its release was also reduced when 2DG was used together with rotenone, compared to rotenone alone‐treated cells (Fig 6F).

Figure EV4. Neither glucose starvation nor inhibitions of other carbon sources than glycolysis or autophagy protect rotenone‐treated cells from death as does 2DG .

1. Percentage of apoptotic MelJuso cells after 48 h of treatment with 2DG (5.5 mM) and/or rotenone (5 μM) and/or metformin (5 mM). Bars show means ± SEM of PI‐positive cells (n = 3). Differences between the groups were evaluated by Student's t‐test: control vs. rotenone (***P = 0.0007), rotenone vs. 2DG + rotenone (**P = 0.0065).
2. Percentage of apoptotic A375 cells after 48 h of treatment with rotenone (5 μM), BPTES (10 μM), and/or etomoxir (50 μM). Bars show means ± SEM of PI‐positive cells (n = 3). Differences between the groups were evaluated by Student's t‐test: rotenone vs. rotenone + BPTES (**P = 0.0023), rotenone vs. rotenone + etomoxir (*P = 0.0403).
3. Percentage of apoptotic A375 and MelJuso cells after 24 h of treatment with rotenone (5 μM) or metformin (5 mM) in the presence (+ Glc) or absence (− Glc) of D‐glucose (2 g/l). Bars show means ± SEM of PI‐positive cells (n = 3). Differences between the groups were evaluated by Student's t‐test: A375 control + Glc vs. control − Glc (*P = 0.0366), MelJuso control + Glc vs. control − Glc (***P = 0.0009), A375 control − Glc vs. MelJuso control − Glc (**P = 0.0075).
4. A375 and MelJuso cells were treated with bafilomycin A1 (5 nM) and/or 2DG (5.5 mM) and/or rotenone (Rot; 5 μM) and/or metformin (Met; 5 mM) for 8 h. Cell extracts were Western‐blotted for LC3 I/II, phospho‐AMPKα T172 (pAMPKα), and total AMPKα.
5. Percentage of apoptotic A375 and MelJuso with 2DG (5.5 mM) and/or rotenone (5 μM) and/or metformin (5 mM) ± bafilomycin A1 (Baf A1; 5 nM) for 48 h. Bars show means ± SEM of PI‐positive cells (n = 3). Differences between the groups were evaluated by Student's t‐test: A375 (2DG + rotenone vs. 2DG + rotenone + Baf A1 (P = 0.5045); 2DG + metformin vs. 2DG + metformin + Baf A1 (P = 0.6605)). MelJuso (2DG + rotenone vs. 2DG + rotenone + Baf A1 (P = 0.4085); 2DG + metformin vs. 2DG + metformin + Baf A1 (P = 0.2042)).

Source data are available online for this figure.

Next, we tested the possibility that autophagy, an important catabolic process involved in cell survival and death 44, might play a role in the 2DG‐protective effect of rotenone‐treated cells. For that, we used bafilomycin A1 (Baf A1), an inhibitor of the late phase of autophagy. bafilomycin A1 induced a robust increase in LC3‐II, a specific marker of autophagosome formation, in A375 and MelJuso cells (Fig EV4D), indicating autophagy blockage. Interestingly, cells treated concomitantly with bafilomycin A1 and the combination of metabolic stressors accumulated less LC3‐II than cells treated with bafilomycin A1 alone (Fig EV4D), suggesting that these cells did not turn on autophagy even though AMPK was activated (Fig EV4D). We also evaluated whether autophagy could have an impact on the improved cell viability after the treatment with the combination of metabolic drugs. As shown in Fig EV4E, bafilomycin A1 did not significantly increase cell death in any of the cell types when added to the mix of metabolic stressors. Therefore, we concluded that autophagy did not participate in the 2DG‐induced protective effect of rotenone‐treated cells.

In the next step, we tested whether this pro‐survival effect could be linked to the reduced glycolytic flux and substrate availability for the TCA cycle caused by 2DG. For that, we added the membrane‐permeable pyruvate analog, methyl‐pyruvate (MP) to 2DG and rotenone‐treated cells. To confirm that the externally supplied MP is taken up by the cells, we monitored the levels of alanine, which can be produced from pyruvate, by NMR analysis. As this reaction is coupled to the conversion of glutamate to α‐ketoglutarate (AKG), we also analyzed the level of AKG in the cells. Intracellular alanine increased to about 400% and intracellular AKG to approx. 1,500% of the control levels in MP‐treated A375 cells, suggesting that the exogenously provided MP is indeed metabolized by the cells (Fig EV5A). MP effectively restored cell death, indicating that the glycolytic flux is probably contributing to the decreased viability of melanoma cells treated with rotenone (Fig EV5B). The NMR analysis also revealed that cells treated with 2DG in combination with metformin or rotenone accumulated significantly more (approx. 300%) intracellular glutamine than control cells (Fig 6G). Glucose and glutamine are the two most important carbon sources for cell growth, and glutamine was shown to maintain TCA cycle and cell viability when mitochondrial pyruvate transport is inhibited 45. Reductive carboxylation of glutamine‐derived AKG is stimulated when TCA cycle function is altered 46, 47, 48, 49. To determine whether glutamine metabolism could be contributing to the maintenance of cell viability in 2DG and rotenone‐treated A375 cells, we inhibited glutamine utilization with BPTES. The glutaminase inhibitor reverted the 2DG‐induced protective effect (Fig 6H), suggesting that A375 cells facing metabolic perturbations rely on glutamine to preserve their viability. Importantly, the addition of BPTES strongly potentiated the cytotoxicity of 2DG and metformin combination.

Figure EV5. 2DG pro‐survival effect is linked to the reduced glycolytic flux and substrate availability for the TCA cycle.

1. Quantification of intracellular alanine and AKG. Before metabolite extraction, cells were treated with MP (10 mM) for 14 h. Bars show means ± SEM (n = 3). Student's t‐test (alanine, *P = 0.0117; AKG, ***P < 0.0001).
2. Percentage of apoptotic A375 and MelJuso cells after 48 h of treatment with 2DG (5.5 mM), rotenone (5 μM), and/or MP (10 mM). Bars show means ± SEM of PI‐positive cells (n = 3). Student's t‐test: A375 (2DG + rotenone vs. 2DG + rotenone + MP (*P = 0.0103)), MelJuso (2DG + rotenone vs. 2DG + rotenone + MP (*P = 0.0282)).

Discussion

Fast‐growing tumors are often subjected to periods of metabolic stress caused by factors such as hypoxia or lack of nutrients 50. In our study, we induced metabolic perturbations in the two most common genomic melanoma subtypes, bearing mutated BRAF and NRAS genes, and examined the effects on RAF signaling and cellular functions.

Our data suggest that metabolic stress promotes dimerization of the KSR proteins with RAF kinases; that is, CRAF in NRAS‐mutant cells and oncogenic BRAF in BRAF^V600E‐mutant cells, stimulating ERK signaling. In NRAS‐mutant cells, the formation of dimers positively regulated CRAF kinase activity. Critically, expression of transactivation‐impaired forms of KSR, KSR1^R665H and KSR2^R718H, abrogated the enhancement of CRAF kinase activity induced by metabolic stressors. In BRAF^V600E‐mutant cells, the formation of dimers did not regulate the activity of oncogenic BRAF as the mutant kinase is constitutively activated. However, KSR invariably associates with the RAF substrate MEK 9, 10 and hence, its heterodimerization with BRAF^V600E could bring MEK and oncogenic BRAF into closer proximity. The mutant kinase has been described to signal primarily as a monomer which does not require dimerization for ERK pathway activation. Yet, it has been shown that the capacity of BRAF^V600E to activate ERK is significantly compromised in the absence of KSR1 51, indicating that KSR function is necessary for the oncogenic potential of BRAF^V600E.

Additionally, we found that in NRAS‐mutant cells, the activation of CRAF by metabolic stress was independent of its upstream regulator NRAS. Notably, mutant CRAF (CRAF^R89L) that cannot bind to RAS 24 had an increased kinase activity when cells underwent metabolic stress. This result does not constitute the first example of RAS‐independent activation of RAF. For example, it has been shown that RNA interference‐mediated knockdown of RAS does not affect the ability of KSR to drive RAF activation in Drosophila Schneider S2 cells 12. Leicht et al 52 also demonstrated that MEK1 could activate CRAF in a RAS‐independent manner, an event which was mediated by the interaction of both kinases. The authors suggest that the association might promote a conformational change that enhances CRAF susceptibility to phosphorylation and activation or that the interaction could stabilize the phosphorylated active form of CRAF. Similar mechanisms could be responsible for the positive effect of the enhanced interaction between KSR and CRAF under metabolic stress on CRAF kinase activity.

While the ERK pathway was hyperactivated in both genomic subtypes under metabolic stress, consequences on cell functions were distinct. The glycolysis inhibitor 2DG induced p21 expression and cell cycle arrest in G0/G1 phase in NRAS‐mutant cells but not in BRAF^V600E‐mutant cells. This might be because the ERK pathway was more activated in NRAS‐mutant cells than in BRAF^V600E‐mutant cells in response to 2DG. Too strong activation of the ERK pathway, in the duration or magnitude, can result in a reversible or permanent cell cycle arrest in G1 phase 39, 40. Interestingly, we showed that the cell cycle arrest could also be achieved in BRAF^V600E‐mutant cells, but a higher level of metabolic stress had to be reached. This state was induced by the combination of two metabolic stressors or by a high concentration of one metabolic stressor, both situations leading to the activation of AMPK. In NRAS‐mutant cells, 2DG was already enough to activate AMPK, and therefore, it appears that BRAF^V600E‐mutant cells might be more resistant to metabolic stress than NRAS‐mutant cells. They were also less sensitive to glucose starvation (Fig EV4C). This difference in sensitivity might come from the fact that the AMPK and BRAF^V600E signaling cannot be activated simultaneously 53, but active AMPK and CRAF could co‐exist in NRAS‐mutant cells as we observed AMPK activation concomitantly to ERK pathway hyperactivation. Contrary to NRAS‐mutant cells that strongly hyperactivated the ERK pathway to induce cell cycle arrest, BRAF^V600E‐mutant cells downregulated the ERK pathway when facing high metabolic stress. Both an excessive activation and the inhibition of the ERK signaling can result in cell cycle arrest 37, 38. Thus, even though the two genomic subtypes oppositely control ERK activity, both have a common goal—to block the cell cycle under metabolic insult.

The next question we wanted to answer was why the two cell types used a different strategy to modulate ERK activity when facing high metabolic stress. We showed that in NRAS‐mutant cells, increased stress promoted the association of CRAF with AMPK and additionally with KSR2, enhancing ERK signaling. This scenario was different in BRAF^V600E‐mutant cells where non‐activated AMPK bound to BRAF^V600E and activation of the kinase correlated with the dissociation of oncogenic BRAF from KSR. Critically, RNA interference‐mediated knockdown of AMPKα in metabolically stressed cells rescued the interaction of mutant BRAF with KSR2, demonstrating that AMPK plays a critical role in the dissociation of oncogenic BRAF from KSR. This negatively regulated ERK pathway activity, most probably due to mutant BRAF being diverted from its substrate MEK, which constitutively binds to the KSR proteins. These data suggested that the differential dependence of the two RAF isoforms on AMPK could be the cause of the opposite responses seen in the two genomic melanoma subtypes facing high metabolic stress. We are currently investigating this possibility further. A previous study found that KSR proteins can associate with AMPK and regulate its activity 54. AMPK binds to the CA3 region of KSR and an additional portion located between the CA2 and CA3 regions for KSR2, while RAF and MEK bind to a more carboxy‐terminal region of KSR 54. While these binding sites are distinct, KSR2 mutations affecting key residues involved in the interactions of KSR2 with RAF and MEK often lead to a reduced affinity for AMPK compared to their wild‐type counterpart 55. These data strongly suggest that the association of KSR proteins with AMPK depends on their interaction with RAF kinases.

The observed cell cycle arrest in response to metabolic insult led us to consider the potential of the metabolic stressors as a therapeutic strategy for mutant NRAS‐ and BRAF^V600E‐bearing melanomas. Our data indicated that the combination of 2DG with an OXPHOS inhibitor or a high concentration of 2DG might be necessary to promote AMPK activation, ERK pathway downregulation, and cell cycle arrest in BRAF^V600E‐mutant cells. Additionally, we also showed that in these settings, BRAF^V600E‐mutant cells started to rely on an additional carbon source, glutamine, to preserve their viability. Therefore, we concluded that metabolic targeting using a glycolysis inhibitor such as 2DG might not be a straightforward therapeutic strategy for BRAF^V600E‐bearing melanomas. We demonstrated a role for AMPK in the downregulation of the ERK pathway in BRAF^V600E‐mutant melanoma cells. Therefore, it might be better to consider the use of an AMP mimetic to activate AMPK without the need to expose the cells to high metabolic stress. Furthermore, it is somewhat difficult to precisely define the conditions under which low and high metabolic stress will occur in a physiological context. The tumor microenvironment is spatially and temporally heterogeneous. The accessibility of oxygen and nutrients depends on the proximity of the cells to blood vessels and will determine their basal metabolic state. Thus, the uptake of 2DG will depend on the initial metabolic status of the cancer cell and its avidity for glucose. In contrast, our data suggest that the metabolic drugs might be quite effective in NRAS‐mutant melanomas, where 2DG as a single agent activated AMPK and hyperactivated the ERK pathway, resulting in cell cycle arrest. Furthermore, ERK pathway hyperactivation could already be observed with a relatively low 2DG concentration (0.75 mM). However, further analyses of the activity of metabolic drugs as single agents in NRAS‐mutant cancer cells will be necessary before proceeding to any clinical application, as an insufficient increase of the ERK signaling could promote cancer cell proliferation rather than induce cell cycle arrest.

Materials and Methods

Cell culture techniques, transfections, and mutagenesis

All cell lines were cultured in RPMI‐1640 containing 2 g/l d‐glucose (Sigma‐Aldrich, R0883) supplemented with 10% fetal bovine serum and 100 units penicillin and 0.1 mg streptomycin per ml. Cells were cultured at 37°C in a humidified atmosphere containing 5% CO₂. For glucose starvation experiment, RPMI‐1640 without glucose was used (Gibco, #11879020). For protein depletion, cells were transfected using X‐tremeGENE siRNA transfection reagent as recommended by the manufacturer (Roche) with the following siRNA: NRAS_siRNA1: 5′‐CACCAUAGAGGAUUCUUAC‐3′; NRAS_siRNA2: 5′‐CUGAGAUACGUCUGUGACU‐3′; AMPKα_siRNA: 5′‐ GAGGAGAGCUAUUUGAUUA ‐3′; non‐targeting (NT) control siRNAs: 5′‐CUGGAGUUGUCCCAAUUCC‐3′; 5′‐AGAAUUGGGACAACUCCAG‐3′. For transient expression studies, cells were transfected using TurboFect transfection reagent as recommended by the manufacturer (Thermo Fisher Scientific). All mutations within CRAF, KSR1, and KSR2 were generated by site‐directed mutagenesis using high‐fidelity PfuTurbo DNA polymerase (Stratagene) and verified by DNA sequencing (Macrogen).

Reagents

Expression vectors for epitope‐tagged CRAF^WT, CRAF^R89L, BRAF^WT, BRAF^V600E, KSR1^WT, and KSR2^WT were kindly provided by WK, NR, and JR. The following antibodies were used for Western blotting: mouse anti‐α‐tubulin (B‐7; sc‐5286), goat anti‐KSR1 (C‐19; sc‐9317), mouse anti‐KSR2 (K75; sc‐100421), rabbit anti‐pMEK1/2 (sc‐7995), goat anti‐MEK1 (C‐18; sc‐219), rabbit anti‐MEK2 (N‐20; sc‐524), rabbit anti‐OctA (D‐8; sc‐807), goat anti‐V5 (G‐14; sc‐83849), mouse anti‐BRAF (F‐7; sc‐5284) (Santa Cruz Technology), rabbit anti‐pERK1/2 T202/Y204 (4370), rabbit anti‐ERK1/2 (9102), rabbit anti‐pCRAF S259 (9421), rabbit anti‐pCRAF S338 (9427), rabbit anti‐pAMPKα T172 (2535), rabbit anti‐AMPK (5831), rabbit anti‐cleaved caspase‐3 (9664), rabbit anti‐caspase‐3 (9665), rabbit anti‐cleaved PARP (5625), rabbit anti‐PARP (9542), rabbit anti‐cleaved caspase‐9 (7237), mouse anti‐caspase‐9 (9508) (Cell Signaling Technology). The following antibodies were used for Western blotting and immunoprecipitation: rabbit anti‐CRAF (C‐20; sc‐227), mouse anti‐CRAF (E‐10; sc‐7267), rabbit anti‐NRAS (C‐20; sc‐519) (Santa Cruz Technology), mouse anti‐14‐3‐3 (CG15) (Thermo Fisher Scientific), rat anti‐HA high‐affinity clone 3F10 (Roche). 2DG, 5TG, 6AN, rotenone, metformin, oligomycin A, antimycin A, bafilomycin A1, BPTES, etomoxir, and methyl‐pyruvate were purchased from Sigma‐Aldrich, piericidin A from Santa Cruz Technology, and forskolin, PD184352, and sorafenib from Selleck Chemicals.

Immunoprecipitation and in vitro kinase assays

Cells were washed once with ice‐cold PBS, scraped from the plates, and lysed as described 56. The lysate was cleared by centrifugation at 13,000 g for 25 min at 4°C. Protein concentrations were determined using the Bradford reagent (Bio‐Rad) with BSA as a standard; 1 μg of an antibody or anti‐FLAG M2 affinity gel (Sigma‐Aldrich) was added to the cell lysates and incubated for 2–3 h at 4°C on a slow rotator. Protein G Sepharose 4 Fast Flow beads (GE Healthcare) were then added and incubated for another hour at 4°C on a slow rotator. Beads–immune complexes were washed three times with ice‐cold lysis buffer, followed by addition of 30 μl of 2× SDS sample buffer. Results were analyzed by Western blotting using standard protocols. Specific bands were detected using home‐made ECL and G:Box imaging system (Syngene). For kinase assays, immunoprecipitated proteins were washed three times with lysis buffer, once with ice‐cold washing buffer and once with ice‐cold reaction buffer as described 56. The kinase reaction was performed in the presence of 20 μM ATP and 500 ng of a recombinant human inactive MEK‐1 substrate (Life Technologies) in a total volume of 20 μl reaction buffer at 30°C for 30 min with gentle agitation. The reaction was ended by adding of 20 μl 2× SDS sample buffer followed by 10‐min boiling. Densitometric quantification of the bands was performed using ImageJ software.

Quantification of intracellular metabolites using NMR spectroscopy

Sample preparation

Acetonitrile extraction 57 was employed to quench cell metabolism and to extract low molecular weight compounds from A375 melanoma cells quantitatively. Following removal of acetonitrile via vacuum concentration, dried extracts were resuspended in 550 μl of D₂O (Sigma‐Aldrich) containing 0.005% sodium 3‐(trimethylsilyl)‐propionate‐2,2,3,3‐d4 (TSP) (Sigma‐Aldrich) used as both chemical shift reference and internal standard for metabolite quantification.

NMR spectroscopy

Information on the concentration of metabolites in individual samples was derived from volumes of corresponding signals in 1D ¹H NMR spectrum. Assignment of signals in NMR spectra of individual samples to a metabolite was achieved via comparison of a sample spectrum with spectra of pure metabolites (Sigma‐Aldrich). The 1D ¹H spectra were measured at 700 MHz using a Bruker Avance III NMR spectrometer equipped with a triple resonance room temperature probe using the zgpr pulse sequence (standard Bruker pulse program library). All spectra were acquired at 20°C and processed using TopSpin 3.2 (Bruker, USA). To make the comparison of metabolite concentration profiles among various samples possible, the signal intensities in individual samples were normalized to total protein concentration.

Propidium iodide (PI) viability assay

Cells were collected 48 h post‐treatment and washed once with ice‐cold PBS. The cell pellets were resuspended in ice‐cold PBS, and 1 μg/ml PI (Sigma‐Aldrich) was added to the suspension and measured with the Attune Acoustic Focusing Cytometer (Applied Biosystems).

Cell cycle analysis

Cells were harvested into ice‐cold PBS and fixed with 70% ethanol for 30 min on ice. After washing with PBS, cells were incubated with RNase A (17 μg/ml) at 37°C for 30 min and then stained with PI (3 μg/ml) for 10 min at room temperature in the dark. The cell cycle profile was measured with the Attune Acoustic Focusing Cytometer. At least 40,000 events were acquired for each sample. FCS Express 6 Plus software was used to analyze the different cell cycle phases and to obtain DNA cycle statistics.

Cell proliferation assay and ATP determination

Cell proliferation was examined by measuring 5‐ethynyl‐2′‐deoxyuridine (EdU) incorporation using the Click‐iT EdU Alexa Fluor 488 Flow cytometry kit (Thermo Fisher Scientific) following the manufacturer's instructions. ATP was quantified with the ATP determination kit from Invitrogen following the manufacturer instructions.

Metabolic assays

Extracellular acidification rate (ECAR) and oxygen consumption rate (OCR) were measured with the Seahorse XFp analyzer. Briefly, 10,000 cells/well were grown overnight, treated for 4 h with the indicated stressors, and incubated for 1 h in a non‐CO₂ incubator with unbuffered DMEM with an adjusted pH of 7.4 according to the manufacturer's instructions. The glycolysis stress test was used to measure ECAR and OCR responses from A375 and MelJuso cells using the following concentrations: 10 mM glucose, 1 μM oligomycin, and 50 mM 2DG. The data were normalized to cell number.

Statistical analysis

Statistics were performed with GraphPad Prism 5 (GraphPad Software, San Diego, CA, USA). Data are represented as mean ± SEM. The number of independent experiments is indicated in the figure legend. Statistical analysis was by Student's t‐test. Value of *P < 0.05, **P < 0.01, and ***P < 0.001 was considered statistically significant.

Author contributions

AV prepared the samples for NMR analysis, performed the mutagenesis, immunoprecipitations, kinase assays, ATP and Seahorse measurements, cell cycle analyses, and viability experiments. MK and LT quantified intracellular metabolites by NMR. WK, NR, and JR provided the epitope‐tagged CRAF^WT, CRAF^R89L, BRAF^V600E, and KSR^WT plasmid constructs. SU performed the EdU cell proliferation assay. ME and JT helped with the Seahorse technology and interpretation of the metabolic data. DP and ZZ aided with the identification of KSR proteins. KS contributed to the evaluation of the metabolic stressors as a potential treatment. AV designed the study, analyzed the data, and wrote the paper with the support from SU and WK.

Conflict of interest

The authors declare that they have no conflict of interest.

Supporting information

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 1A

Source Data for Figure 2A

Source Data for Figure 3A

Source Data for Figure 4A

Source Data for Figure 5A

Source Data for Figure 6D

EMBO Reports (2018) 19: 320–336