Uncoupling of the LKB1-AMPKα Energy Sensor Pathway by Growth Factors and Oncogenic BRAFV600E

Background Understanding the biochemical mechanisms contributing to melanoma development and progression is critical for therapeutical intervention. LKB1 is a multi-task Ser/Thr kinase that phosphorylates AMPK controlling cell growth and apoptosis under metabolic stress conditions. Additionally, LKB1Ser428 becomes phosphorylated in a RAS-Erk1/2-p90RSK pathway dependent manner. However, the connection between the RAS pathway and LKB1 is mostly unknown. Methodology/Principal Findings Using the UV induced HGF transgenic mouse melanoma model to investigate the interplay among HGF signaling, RAS pathway and PI3K pathway in melanoma, we identified LKB1 as a protein directly modified by HGF induced signaling. A variety of molecular techniques and tissue culture revealed that LKB1Ser428 (Ser431 in the mouse) is constitutively phosphorylated in BRAFV600E mutant melanoma cell lines and spontaneous mouse tumors with high RAS pathway activity. Interestingly, BRAFV600E mutant melanoma cells showed a very limited response to metabolic stress mediated by the LKB1-AMPK-mTOR pathway. Here we show for the first time that RAS pathway activation including BRAFV600E mutation promotes the uncoupling of AMPK from LKB1 by a mechanism that appears to be independent of LKB1Ser428 phosphorylation. Notably, the inhibition of the RAS pathway in BRAFV600E mutant melanoma cells recovered the complex formation and rescued the LKB1-AMPKα metabolic stress-induced response, increasing apoptosis in cooperation with the pro-apoptotic proteins Bad and Bim, and the down-regulation of Mcl-1. Conclusions/Significance These data demonstrate that growth factor treatment and in particular oncogenic BRAFV600E induces the uncoupling of LKB1-AMPKα complexes providing at the same time a possible mechanism in cell proliferation that engages cell growth and cell division in response to mitogenic stimuli and resistance to low energy conditions in tumor cells. Importantly, this mechanism reveals a new level for therapeutical intervention particularly relevant in tumors harboring a deregulated RAS-Erk1/2 pathway.


Introduction
Melanoma is the most lethal human skin cancer and its incidence is rapidly rising world-wide [1]. The development of effective therapeutics designed to target melanoma requires a comprehensive understanding of the underlying biochemical and genetic processes contributing to melanocytic neoplasic transformation and the subsequent progression to an advanced melanoma disease stage. Therefore, dissecting the aberrant signaling pathways that are critical to melanomagenesis and understanding the mechanisms by which these pathways interact with each other has become the recent focus of research directed at melanoma therapeutic intervention.
Dysfunctional receptor tyrosine kinase (RTK) signaling, in particular through the hepatocyte growth factor (HGF) tyrosine kinase receptor c-Met signaling pathway, is one important hallmark of melanoma. HGF signaling activates Ras-Erk1/2 and PI3K-AKT pathways, and Ras pathway activation has been shown to play a role in melanoma development and maintenance [2]. Notably, BRAF, a downstream activator in the RAS pathway is mutated in nearly 70% of human melanoma (BRAF V600E activating mutation) while NRAS activating mutations occurs in 30% of melanomas (NRAS Q61L activating mutation) [3]. In addition, support for PI3K-AKT pathway signaling dysfunction in melanomagenesis has been demonstrated by the documented loss of the tumor suppressor PTEN-containing chromosomal region in 5-20% of melanomas as well as the over expression of AKT3 in the advanced stages of this disease [4,5]. Strikingly, however, single mutations within these two pathways are not sufficient to promote melanoma development suggesting that a complex interplay of these aberrant signalling pathways, under poorly understood circumstances, promote melanomagenesis [2].
We chose to investigate the potential interplay among the HGF RTK signaling, the RAS Ras-Erk1/2 and the PI3K-AKT pathways using the HGF transgenic mouse model in which HGF is over-expressed and which develops melanoma in response to neonatal ultraviolet (UV) radiation. This model is unique in that it develops melanocytic neoplasms in stages that are highly reminiscent of the human cutaneous malignant melanoma with respect to biological, genetic and etiologic criteria [6,7].
To begin the analysis, we first searched for possible molecular candidates with potential to mediate the HGF complex signaling and identified the multitasking serine/threonine kinase, LKB1 [8] as one candidate. LKB1 is involved in cell cycle control [9,10], cellular energy metabolism [11] and cell polarity [12]. The cellular localization and activity of LKB1 is controlled through its interaction with the STE20-related adaptor (STRAD) and the armadillo repeat-containing mouse protein 25 (Mo25); [13,14]. These finding led to the discovery that LKB1 is the upstream kinase to AMP-activated protein kinase (AMPK) and is linked to mTOR through the AMPK-TSC1/TSC2 cascade [15,16,17]. LKB1 is phosphorylated on at least 8 residues, and evidence suggests that LKB1 auto-phosphorylates itself on at least four of these, whereas the other four are phosphorylated by upstream kinases [8]. Recent studies show that human LKB1 Ser428 (the equivalent mouse residue is LKB Ser431 ) is phosphorylated in response to mitogenic signals including EGF, TPA, elevated levels of cAMP as well as by PKCf [18,19,20,21,22], where the EGF-mediated phosphorylation of LKB1 Ser428 is dependent on the activation of p90 RSK . Although experiments conducted in G361 melanoma cells indicate that this residue is involved in LKB1-mediated cell growth inhibition [8,23], and several other investigations implicate LKB1 Ser431 residue in the activation of AMPK and BRK1/BRSK2 kinases (SAD-B/SAD-A) [21,24,25], a recent publication stating that LKB1 phosphorylation in the C-terminal is not required for regulation of AMPK BRSK1/2 and cell cycle arrest contradicts the previous findings [26].
Importantly, sporadic mutations in the lkb1 gene have been documented in cancers of the breast, pancreas, lung, prostate, cervical and ovary as well as in Peutz-Jeghers syndrome, a rare disorder characterized by the appearance of intestinal polyps and mucocutaneous melanocytic macules [27,28,29,30,31]. In Peutz-Jegher patients, LKB1 may function as a tumor suppressor and is associated with loss of heterozygosity or somatic mutation at the lkb1 locus (for review [8]). More importantly, lkb1 mutations have been described in melanoma [32] and based on this information we determined if LKB1 could function as a potential link between an activated RAS pathway and dysfunctional c-Met signaling and play a role in melanoma development and progression.
In this study, we identify the mouse LKB1 Ser431 residue as a phosphorylation target, not only for EGF, but also for HGF signaling and demonstrate that this LKB1 phosphorylation is executed in an Erk1/2-p90 Rsk -dependent manner, as previously described in response to EGF stimulation [18,19]. We demonstrate that LKB1 Ser428 residue is constitutively phosphorylated in cells harboring BRAF V600E activating mutations, and is found frequently phosphorylated in mouse tumor samples with an increased receptor tyrosine kinase activity suggesting a functional connection between BRAF oncogenic pathway and LKB1. Interestingly, BRAF V600E mutant cells show a very limited response to metabolic stress that appears to be mediated by mechanism that involves the uncoupling of the energy stress sensor pathway LKB1-AMPK-mTOR. Importantly, inhibition of RAS-Erk1/2 pathway in BRAF V600E mutant melanoma cell lines restores the LKB1-AMPK-mTOR pathway response to metabolic stress promoting apoptosis in coordination with the BH3-family proteins Bad and Bim and the Bcl-2 family member Mcl-1.

HGF induces LKB1 Ser428 phosphorylation in a RAS-Erk1/2-p90 RSK pathway-dependent manner
We proposed to identify novel molecules involved in melanoma development and progression analyzing the HGF specific signaling in the UV induced HGF transgenic melanoma mouse model. To understand the HGF specific signaling contributions we used 37-31E-mouse melanoma cell line isolated from neoplasic lesions raised in the HGF transgenic-UV irradiated mice and performed a proteomic screening of the phospho-protein complexes induced after the growth factor treatment (data not shown). As a result, we identified LKB1 as a kinase that becomes phosphorylated in response to HGF. Since previous studies implicates RAS pathway in the modification of this residue we decided to use different cell lines harboring either, BRAF wild type (37-31E, 37-31T, B16F1 and MeWo) or BRAF V600E mutant cell lines (UACC903, A375 and SKMel28). As shown in Figure 1A, in isolated phosphoprotein complexes from 37-31E cells, LKB1 Ser431 was specifically phosphorylated in response to HGF treatment since this phosphorylation was totally prevented by the pretreatment with the specific c-Met inhibitor PHA. LKB1 Ser428 (Ser431 in mouse) is phosphorylated in response to EGF through RAS-Erk1/2-p90 RSK pathway [18,20]. Since HGF triggering activates RAS and PI3K pathway [33,34,35], we used specific Mek1/2 (U0126) and PI3K (LY294002) inhibitors to determine which pathway was involved in the LKB1 Ser431 phosphorylation. Figure 1B shows that in mouse and human melanoma cells, HGF-induced phosphorylation of LKB1 Ser431 was totally abolished by the specific Mek1/2 inhibitor U0126, whereas the PI3K inhibitor LY294002 had no effect on HGF-induced phosphorylation of LKB1 Ser431 . Furthermore, time course experiments showed that p90 RSK became phosphorylated in response to HGF and its phosphorylation profile correlated with LKB1 Ser431 phosphorylation (Fig. 1C). Additionally, analysis of B16F1 melanoma cells showed that the phosphorylation of the LKB1 Ser431 was p-Erk1/2 dependent (Fig. 1C) and, inhibition of Mek1/2 after HGF treatment in 37-31E cells totally abolished the phosphorylation of Erk1/2, p90 RSK and LKB1 Ser431 (Fig. 1C). To confirm the p90 RSK participation we used the p90 RSK specific inhibitor BI-D1870. Treatment of 37-31E cells with BI-D1870 abolished the HGF-mediated phosphorylation of LKB1 Ser431 (Fig. 1D). The observed activation upon BI-D1870 treatment of Erk1/2 and the increased levels of p-CREB Ser133 is in agreement with the suggested p90 RSK negative-feedback loop that regulates Erk1/2 described by other authors [20].
These results show that, LKB1 Ser428 is phosphorylated in response to HGF treatment in an Erk1/2-p90 RSK dependent manner. LKB1 Ser428 is phosphorylated in response to different growth factors and is highly phosphorylated in melanoma cells harboring BRAF V600E mutations and in tumor samples Next, we investigated if this post-translational modification was broader in scope. We tested several growth factors related to cancer development in human and mouse melanoma cell lines to check whether or not these ligands were able to induce LKB1 Ser428 phosphorylation. As suggested by previous investigations [18], all ligands tested, including HGF, EGF, basic fibroblast growth factor (FGF2), insulin like growth factor 1 (IGF-1), platelet derived growth factor (PDGF), tumor necrosis factor-alpha (TNF-a), herregulin, insulin and phorbol-ester tumor promoter TPA were able to induced LKB1 Ser428 phosphorylation in a cell type dependent manner ( Fig. 2A). In all cases, the ligands that activated Erk1/2 and p90 RSK kinases led to the phosphorylation of LKB1 Ser428 (Fig. 2A).
The aberrant regulation of the Ras-Erk1/2 pathway represents one of the hallmarks in cancer. Considering that 70% of human melanomas harbor BRAF activating mutations [3], we determined the phosphorylation status of LKB1 Ser428 in different human melanoma cell lines harboring BRAF V600E activating mutations in serum free and complete medium conditions. As expected, SKMel28, A375, and UACC903 human melanoma cell lines harboring the BRAF V600E mutation demonstrated a constitutively phosphorylated LKB1 Ser428 residue whereas the MeWo human melanoma cell line that harbors the wild type alleles did not (Fig. 2B).
Based on the above results, tumor samples with a deregulated tyrosine kinase pathway and/or with enhanced RAS-mediated and p-90 RSK Thr359/Ser363 after HGF triggering (40 ng/ml) under serum starvation conditions. LKB1 total protein is shown as a loading control. On the right, LKB1 Ser431 is phosphorylated in response to HGF in an Erk1/2-p90 RSK dependent manner. Time course shows the phosphorylation of Erk1/2 Thr202/Tyr204 and LKB1 Ser428 in B16F1 cells. Down below, 37-31E melanoma cells were serum starved and triggered with HGF (40 ng/ml) for 5 minutes. Then, cells were treated with the Mek1/2 specific inhibitor U0126 (10 mM) for the indicated increasing times. Fifty mg of total lysates were resolved by SDS-PAGE and membrane was probed with the indicated antibodies. (D) 37-31E cells were treated for 10 min in serum starvation with HGF (40 ng/ml) in the presence or absence of U0126 (10 mM) or BI-D1870 (10 mM). Western-blots show the levels of p-LKB1 Ser431 , p-Erk1/2 Thr202/Tyr204 and p-CREB Ser133 . doi:10.1371/journal.pone.0004771.g001 mitogenic activity would be expected to exhibit elevated LKB1 Ser428 phosphorylation. In this matter, the phosphorylation state of LKB1 Ser431 in spontaneous tumor samples raised in UVirradiated HGF transgenic mice and in xenographed tumors from the 37-31E-melanoma cells correlated with elevated levels of p-Erk1/2 (Fig. 2C).
All together, these data indicated the existence of a RAS pathway and LKB1 crosstalk suggesting that LKB1 might be involved in some of the RAS-Erk1/2 induced-responses, and more importantly would be contributing to BRAF oncogenic signaling.

BRAF mutant melanoma cells have a dysfunctional LKB1-AMPK energy stress-induced pathway response
Melanoma cells are especially resistant to different types of stress. LKB1 is the AMPKa upstream kinase that becomes phosphorylated in response to metabolic stress controlling protein synthesis through mTOR pathway. Considering the LKB1 Ser428 phosphorylation as a read out of the RAS and LKB1 pathways interaction, the constitutive phosphorylation of LKB1 Ser428 in BRAF mutant cells suggested a possible interplay between LKB1-AMPK pathway and BRAF oncogenic signaling. Therefore, we investigated the LKB1-AMPK pathway activation in three different BRAF V600E mutant melanoma cell lines under low energy conditions and the contribution of BRAF V600E signaling to the energy sensor pathway. To test this hypothesis we starved BRAF mutant melanoma cells under serum free and low glucose conditions in the presence or absence of the Mek1/2 inhibitor U0126, and investigated the activation of LKB1-AMPK-mTOR pathway. UACC903, SKMel28 and A375 cells showed a very limited response to energy withdrawal as measured by the induction of phospho-AMPKa (Fig. 3A). Under these conditions all cells retained considerable mTOR activity as indicated by the phosphorylation levels of ribosomal protein S6 (Fig. 3A). However, the addition of U0126 (10 mM) recovered AMPKa pathway (100 ng/ml), Herregulin (50 ng/ml), IGF-1 (50 ng/ml), PDGF (50 ng/ml), TNF-a (100 ng/ml) Insulin (100 nM) and TPA (200 nM). Fifty mg of total lysates were separated by SDS-PAGE and same membranes were incubated against the indicated antibodies. (B) MeWo (BRAF wild type), A375 (BRAF V600E ), SKMel28 (BRAF V600E ) and UACC903 (BRAF V600E ) human melanoma cells were growth in complete medium (CM) or serum starvation (SF) conditions as indicated. Fifty mg of total lysates were analyzed by SDS-PAGE. The phosphorylation status of LKB1 Ser428 , p-Erk1/2 Thr202/Tyr204 and p-90 RSK Thr359/Ser363 is shown. Total Erk1/2 is used as a loading control. Cell genotypes are showed. (C) p-LKB1 Ser431 and p-Erk1/2 Thr202/Tyr204 levels in mouse melanoma tumor samples. Samples 1-7 primary tumors raised in HGF-UV irradiated transgenic mice. Samples 8 and 9 show xenograph tumors generated from 37-31E cells in FVB mice with high and low p-Erk1/2 levels, respectively. As a control fifty micrograms of protein from 37-31E melanoma cell line treated with HGF (40 ng/ml) for 10 minutes was added (Total lysates, T.L.). Same membrane was blotted against the indicated antibodies. Quantifications of phospho-proteins normalized against total protein are showed in the graphs below. doi:10.1371/journal.pone.0004771.g002 activation in response to low energy conditions resulting in the complete abrogation of mTOR activity as indicated by phospho-S6 ribosomal protein (Fig. 3A). In contrast, 37-31E melanoma cells harboring wild type BRAF did not show this effect by the addition of U0126 under low energy conditions (Fig. 3A). Interestingly, the re-activation of the AMPKa pathway in BRAF mutant cells correlated with the total inactivation of Erk1/2 and the unphosphorylated LKB1 S428 (Fig. 3A). Importantly, the addition of U0126 up to 10 mM in serum free high glucose conditions did not induce the activation of AMPKa [36] (Fig. 3B). To confirm the reconnection of the AMPK pathway after inhibition of oncogenic BRAF signaling we used AICAR (5-Aminoimidazole-4-carboxyamide ribonucleoside) instead of low glucose in order to stimulate the activation of the AMPK pathway. As expected, the addition of AICAR, which increases AMPK phosphorylation levels by a mechanism that appears to be due to the inhibition of AMPK dephosphorylation [37,38,39], slightly increased the p-AMPK levels in serum starvation. The addition of U0126 inhibitor resulted in a clear increment the p-AMPKa levels (Fig. 3C). Since this effect was observed in BRAF V600E mutant cells, we repeated the experiments using the BRAF inhibitor sorafenib. Notably, inhibition of BRAF V600E signaling with sorafenib recovered the activation of AMPK pathway in response to metabolic stress. Interestingly, sorafenib treatment under low glucose condition reduced Erk2 protein levels by a currently unknown mechanism (Fig. 3D). Importantly, experiments knocking-down BRAF V600E performed in serum free and low glucose medium also resulted in an increased of p-AMPKa levels ( Fig 3D).
In addition, we tested whether p90 RSK signaling was mediating the observed effect using the p90RSK specific inhibitor BI-D1870. Experiments were done in the presence of EGF to assured the activation of RAS pathway. Treatment of cells with the p90 RSK inhibitor BI-D1830 did not recover the cells response to low energy conditions as indicated by the p-AMPK levels ( Fig. 3E) suggesting, that the oncogenic BRAF-mediated LKB1 Ser428 phosphorylation was not sufficient to account for the observed response. As previously shown, BI-D1870 treatment induced the Erk1/2 and CREB Ser133 phosphorylation mediated by the suggested p90 RSK negative-feedback loop that regulates Erk1/2 [20].
Altogether these data show evidences that indicate that melanoma cells harboring oncogenic BRAF have a diminished response to metabolic stress. Importantly, the inhibition of the oncogenic BRAF signaling, that connects RAS pathway to LKB1, restored the AMPK-mediated energy stress sensor pathway. However the results also indicate that the inhibition of the LKB1 Ser428 phosphorylation is not enough to recover the pathway LKB1-AMPK-mTOR response, suggesting the existence of additional mechanisms.
Growth factor treatment and oncogenic BRAF V600E induce LKB1-AMPK disassembly The LKB1 tumor suppressor kinase activity is not related to its phosphorylation state [14]. Thus, its contribution to the different biological processes is likely to be mediated by its interaction with other proteins and/or its cellular localization. LKB1 controls protein synthesis and cell growth through the AMPK-TSC1/ TSC2 cascade [15,16,17]. Mitogenic responses coordinate simultaneously cell growth with cell division. Since the activation of AMPK-TSC1/TSC2 pathway by LKB1 controls energy metabolism and protein synthesis, we examined whether the growth factor treatment leads to LKB1-AMPK dissociation, providing a mechanism that would assure cell growth upon a mitogenic stimulus and resistance to energy stress conditions. To investigate the underlying mechanism, we transfected 293T cells with Flagtagged LKB1 and GST-AMPKa and then treated the cells with or without EGF in order to activate RAS pathway. After immunoprecipitation of the Flag-LKB1 complexes we checked for the presence of GST-AMPKa in the immunocomplexes. The data indicated that some fraction of AMPKa is constitutively bound to LKB1. Interestingly, treatment of cells with the growth factor induced dissociation of the LKB1-AMPKa complexes (Fig. 4A). Moreover, this effect was totally independent of LKB1 kinase activity since Flag-tagged kinase dead LKB1 (LKB1 KD ) reproduced exactly the same result. Since growth factor stimulation promoted LKB1-AMPKa disassembly and this effect correlated with the phosphorylation of LKB1 Ser428 upon growth factor stimulation, we examined the role of the Ser428 (Ser431 in mouse) residue on this effect. We transfected LKB1 WT wild type, LKB1 S431A mutant, LKB1 S431D phospho-mimetic mutant and the LKB1 KD constructs together with GST-AMPKa in 293T cells and repeated the previous experiment. Again when LKB1 WT and LKB1 KD were transfected the EGF treatment promoted the disassembly of the LKB1-AMPKa complex. However, GST-AMPKa did not form a complex with either the LKB1 S431A or LKB1 S431D mutants, suggesting that, in response to growth factors, the Ser428 residue would be involved in the binding or stability of the LKB1-AMPKa complexes (Fig. 4B).
The above data suggested that the limited response to metabolic stress of BRAF mutant melanoma cells could be caused by the dissociation of LKB1-AMPKa complexes. Thus, the inhibition of BRAF signaling by U0126 inhibitor would permit the reconnection of the pathway. To confirm that, we performed an immunoprecipitation of the endogenous LKB1 in the BRAF mutant melanoma cells and examined the AMPKa association under low energy conditions with or without U0126 inhibitor. UACC903 and A375 melanoma cells showed an increase in the number of AMPKa molecules associated to LKB1 when BRAF signaling pathway was blocked (Fig. 4C). This re-assembly was also associated with an increase in AMPKa T172 phosphorylation levels. Similar results were obtained when endogenous p-AMPKa T172 was immunoprecipitated from SKMel28 melanoma cells under same conditions, confirming the suggested mechanism (Fig. 4C). Additionally, we reconstitute the system in Hela cells that do not express endogenous LKB1. Hela cells were transfected with Flag-LKB1 and GST-AMPKa in the presence or absence of oncogenic myc-BRAF V600E under low glucose conditions. The expression of oncogenic BRAF V600E induced the complex dissociation that was totally rescued by the addition of U0126 inhibitor (Fig. 4D).
These data showed evidences that support the growth factor treatment and RAS pathway activation mediated disassembly of the LKB1-AMPK complexes. Importantly the inhibition of the RAS pathway in cells harboring BRAF V600E mutation restores the LKB1-AMPKa pathway by permitting the re-association of the LKB1-AMPKa complexes. Furthermore, these data also suggested that although growth factor stimulation induces LKB1 Ser428 phosphorylation, additional mechanisms should be involved promoting the RAS pathway-dependent LKB1-AMPK disassembly.

Restoration of the LKB1-AMPKa pathway in BRAF V600E melanoma cells under energy stress conditions induces apoptosis in coordination with Bad, Bim and Mcl-1
Next, investigated the cell survival response of BRAF V600E mutant melanoma cell lines with a restored LKB1-AMPKa pathway under stress energy conditions. According to the accepted mechanism, elevation of intracellular AMP levels will activate LKB1 that in turn activates AMPK and regulates apoptosis in response to energy stress [16]. Additionally, it has been shown that blocking BRAF signaling in BRAF V600E mutant cell lines for long periods of time (24-48 hours) promotes apoptosis through the regulation of BH3-family proteins [40,41]. We subjected UACC903 and A375 melanoma cells to metabolic stress conditions for a maximum of 12 hours in the presence or absence of the Mek1/2 inhibitor. Then, we measured cell viability and apoptosis by nuclear staining exclusion (Guava ViaCount) and Annexin V and propidium iodide (PI) double staining. UACC903 and A375 melanoma cell lines showed some spontaneous apoptosis under normal growing conditions: 5.71% and 3.27% respectively. The addition of Mek1/2 inhibitor for 12 hours in high glucose medium did not promote any increment in the apoptosis rate (UACC903 5.38% and A375 2.70%). Glucose starvation for 12 hours resulted in slight increase in the number of double positive Annexin V-PI cells respect normal growing conditions: 1.27 fold for UACC903 cells and 2.35 fold for A375 cells. However, the restoration of the LKB1-AMPKa pathway by inhibition of Mek1/2 under low glucose conditions resulted in a considerable number of apoptotic cells (5.7 and 6.4 fold increase respectively); (Fig 5A). Similar results were observed when viable cells were analyzed by PI exclusion (Fig. 5A). The results were correlated with the molecular status of the pathways implicated in a time course fashion. As shown before, the inhibition of Erk1/2 phosphorylation resulted in the decrease of phopho-LKB1 Ser428 levels and the restoration of the LKB1-AMPKa pathway. In turn, the LKB1-AMPKa pathway was able to sense the low energy conditions as soon as 4 hours after glucose starvation, indicated by the levels of p-AMPKa T172 (Fig. 5B). Interestingly, the increase in the number of apoptotic cells did not correlate with p53 stabilization. On the contrary, p53 levels were down-regulated under these conditions in all cell lines, suggesting the participation of BRAF signaling in the stabilization of p53 and a p53independent apoptotic mechanism (Fig. 5B).
In order to establish a causal link among the inhibition of BRAF signaling, the AMPK pathway re-activation, and the increased number of dead cells, we knocked-down AMPKa in UACC903 melanoma cells and investigated the response under low glucose conditions to the inhibition of the oncogenic BRAF signaling. As showed in figure 5C, blocking BRAF signal under low energy conditions in control cells resulted in elevated levels of p-AMPKa T172 together with an increment in the number of dead cells. However, AMPKa knock-down cells did not show any increase in the number of dead cells under similar conditions.
As mentioned previously, BRAF suppresses apoptosis, targeting the BH3-family of proteins in BRAF V600E mutant cells [40]. In high glucose medium the addition of the U0126 inhibitor for 12 hours caused a small decrease of p-Bad and the Bcl-2 family member Mcl-1 protein levels together with a slight increase in the amount of Bim EL . Interestingly, when the LKB1-AMPKa pathway was restored under low glucose conditions, the increased number of dead cells correlated with a stronger biochemical response including the complete de-phosphorylation of Bad, the stabilization of the non-phosphorylated Bim EL isoform and the drastic down-regulation of Mcl-1 (Fig. 5D).
Our results indicated that in a BRAF mutant context, the reactivation of LKB1-AMPK-mTOR pathway under low energy conditions together with the inhibition of oncogenic BRAF signaling for short periods of time, promoted a pronounced apoptosis response through the de-phosphorylation of Bad, stabilization of Bim EL and the down-regulation of Mcl-1.

Discussion
The understanding of the molecular and biochemical mechanisms contributing to melanoma development and progression is critical for therapeutical intervention. The UV-induced HGF mouse melanoma model recapitulates chronologically and histopathologically all the stages of human melanoma [6,7]. We investigated the potential interplay among the HGF RTK signaling, the RAS-Erk1/2 and the PI3K-AKT pathways using the HGF mouse model. In neoplasic melanoma cells isolated from spontaneous tumors raised in the mouse model, we identified LKB1 as one of the molecules responsive to HGF triggering. As previously described for EGF [18], we show that HGF and several other growth factors induce the phosphorylation of LKB1 Ser431 through the Ras-Erk1/2-p90 RSK pathway. Interestingly, this residue appears to be constitutively phosphorylated in human melanoma cells harboring BRAF V600E activating mutation as an indicator of the connection between RAS pathway and LKB1. The role of LKB1 in response to growth factors, and its connection to the RAS pathway, is mostly unknown. Our results show that melanoma cells harboring the BRAF V600E oncogenic mutation have a very limited response to metabolic stress. Interestingly, the inhibition of the BRAF signaling restores the ability of the cells to sense the low energy conditions. Notably, growth factor treatment and oncogenic BRAF V600E leads to the uncoupling of LKB1-AMPKa complexes, suggesting a mechanism that disconnects the energy sensor pathway, which is involved in controlling cell growth through the mTOR pathway in response to low energy conditions. Furthermore, inhibition of BRAF oncogenic signaling promotes the association of the LKB1-AMPKa complexes and results in an increase of apoptosis in response to metabolic stress.
Our screening for the discovery of novel molecules involved in HGF signaling in melanoma cells allowed us to identify by DIGE analysis proteins that were directly modified in response to c-Met activation by HGF (data not shown). Previously, it has been described that EGF and forskolin promote the phosphorylation of LKB1 Ser428 in a p90 RSK and PKA dependent manner respectively [18,19]. Our results showed that LKB1 Ser431 phosphorylation occurs in response to HGF and other different growth factors in a RAS-Erk1/2-p90 RSK dependent manner as initially suggested by previous investigations [18,19]. Interestingly, and according to siRNA knock down experiments (data not shown), cells expressing less than five percent of the LKB1 pool still respond to this stimulus, suggesting that LKB1 would have a relevant role in the mitogenic response to growth factors. Virtually all cancers have aberrant signaling of receptor tyrosine kinases (RTKs), growth factors autocrine loops or activating mutations in the RAS pathway (RAS activating mutations or BRAF V600E mutation). We therefore hypothesized that LKB1 would be mediating some of the effects of the RAS and BRAF oncogenes. In agreement with this, LKB1 Ser428 appears to be constitutively phosphorylated in human melanoma cell lines harboring BRAF V600E activating mutations as an indicator of the interplay between RAS pathway and LKB1. Furthermore, LKB1 Ser431 (Ser428 in human) tends to be phosphorylated in mouse tumor samples harboring deregulated tyrosine kinase activities or increased mitogenic signaling, suggesting the direct participation of LKB1 in tumor biology.
LKB1 activity is controlled through its interaction with the STE20-related adaptor (STRAD) and Mo25 [13,14]. LKB1 can be phosphorylated at eight or more different residues, where the modifications at these amino acids have no effect on LKB1 kinase activity [8]. In the last five years a number of publications have reported several critical roles for LKB1 in different biological processes such as: energy metabolism [11,16], cell polarity and division [12,11] and transcriptional regulation [10]. However, most of these studies rely on the presence or absence of LKB1 in these processes. Our data show that growth factor treatment induces both, LKB1 Ser428 phosphorylation and the dissociation of LKB1-AMPKa complexes. The participation of this residue in growth inhibition, cell polarity and the activation of the LKB1 downstream kinases (AMPK, BRSK1/2) have been controversial [8,21,23,25,26]. Indeed, our results regarding the participation of this residue in the growth factor-mediated dissociation of the LKB1-AMPKa complexes are not conclusive and more research is needed in order to elucidate the complete role of this residue. Moreover, we can no exclude the participation of additional residues or proteins in the process. LKB1 is the upstream kinase of AMPKa and is linked to mTOR through the AMPK-TSC1/TSC2 cascade [15,16,17] controlling cell growth under energy stress conditions. One possible interpretation of our results would be that the dissociation of the LKB1-AMPKa complex would provide a mechanism to avoid interruption of protein synthesis through this pathway while cells are responding to a mitogenic stimulus. In this matter, the activation of RAS-Erk1/2 pathway would engage biochemical mechanisms to coordinate cell growth and division to assure cell proliferation. Importantly, the deregulation of the AMPK-mTOR axis by the dissociation of the LKB1-AMPKa complex in cells harboring RAS pathway activating mutations could represent an advantage for proliferation and a significant resistance increased to metabolic stress conditions. Notably, BRAF V600E mutant melanoma cell lines showed a limited sensitivity in response to low energy conditions. Treatment of cells with the Mek1/2 inhibitor U0126 allowed the formation of endogenous LKB1-AMPKa complexes and restored the energy sensor pathway in response to low energy conditions, supporting the proposed mechanism (Fig. 6). Mek1/2 inhibitors (U0126, PD98059) have been reported to activate AMPK at 20 mM concentration, but not at 5-10 mM [36]. These experiments were done in glucose free medium and in the presence of growth factors with high levels of phospho-Erk1/2. According to our data the inhibition of the Erk1/2 pathway under low energy conditions would allow the re-association of LKB1-AMPKa, which in turn would result in an increase the levels of p-AMPKa. Furthermore, our data showed that in high glucose medium the addition of U0126 (10 mM) or sorafenib (data not shown) had no effect on AMPKa phosphorylation in all BRAF mutant cell lines (Fig 5A). Up-to-date the connection between RAS pathway and LKB1 has been limited to the phosphorylation of LKB1 Ser428 residue through p90 RSK . However, our results indicate that this modification is not enough to mediate the observed effect, suggesting the existence of additional biochemical mechanisms mediated by Erk1/2 that will account for the LKB1-AMPKa dissociation mediated by the RAS pathway (Fig. 6).
The activation of AMPKa by LKB1 under energy stress, stimulates glucose uptake and fatty acid oxidation to increase ATP production, inhibits protein synthesis and protects cells from undergoing apoptosis [8]. Surprisingly, the inhibition of Erk1/2 pathway in BRAF mutant melanoma cell lines subjected to metabolic stress resulted in an increase in the number of dead cells (Fig 6). Interestingly, our experiments knocking-down AMPKa suggests a causal link among the inhibition of oncogenic BRAF signaling, the reconstitution of the energy sensor pathway and the resulting cell death. In agreement with the interplay between the oncogenic signaling and AMPK is the recent finding where the activation of AMPK pathway by administration of metformin, phenformin or A-769662 to PTEN(+/2) mice significantly delayed tumor onset, demonstrating that LKB1 is required for activators of AMPK to inhibit mTORC1 signaling as well as cell growth in PTEN-deficient cells [42]. Interestingly, the increased apoptosis rate did not correlate with the stabilization of p53, indicating that the mechanism was apparently p53-independent and that BRAF oncogenic signaling was participating in the stabilization of p53. In fact, it is known that under genotoxic-stress conditions, Erk1/2 signaling mediates p53 stabilization [43,44]. Furthermore, the AMPK-induced p53 activation has been reported to promote cell survival in response to glucose deprivation in MEFs [45], while our data in BRAF mutated melanoma cells clearly showed an increase in apoptosis.
Recent publications have shown that oncogenic BRAF can suppress apoptosis through targeting BH3-only proteins Bad and Bim [40,41]. Our results indicate that the inhibition of oncogenic BRAF signaling at 12 h promotes a slight de-phosphorylation of Bad and the stabilization of Bim EL , most likely, by inhibiting its Erk1/2-dependent phosphorylation and proteasome-mediated degradation [46]. The reactivation of the LKB1-AMPK-mTOR pathway under low energy conditions by the inhibition of BRAF signaling led to a more pronounced effect that included a drastic down-regulation of Mcl-1 (Fig 6). Interestingly, Mcl-1 has been shown to be an important melanoma anti-apoptotic protein [47].
In conclusion, in this report we show that activation of RAS pathway by growth factors and oncogenic BRAF V600E results in the dissociation of LKB1-AMPKa. These results, permit us to speculate that under normal growth conditions, this biochemical mechanism, through the activation of RAS pathway, could be involved in the coordination of two important processes in cell proliferation: cell growth and cell division. Interestingly, BRAF V600E mutant melanoma cells have minimal response to energy stressed conditions due to the constitutive dissociation of the LKB1-AMPKa complexes. However, under metabolic stress conditions the inactivation of BRAF oncogenic signaling restores the LKB1-AMPKa-mTOR pathway-promoting apoptosis in collaboration with BH3-only proteins and Mcl-1 (Fig 6). Importantly, this mechanism reveals a new level for therapeutical intervention triggering apoptosis of tumor cells. This might be particularly relevant in tumors harboring a deregulated RAF-Erk1/2 pathway that survive in energy stress conditions.

Phospho-protein isolation
37-31E cells were serum starved for two hours and then triggered with 40 ng/ml of HGF (R&D) for 10 minutes in the presence or absence of 0.2 mM of the c-Met specific inhibitor PHA (Sugen-Pfizer). Cells were then lysed according to the phosphoprotein purification kit (Qiagen Inc.), and phospho-proteins purified according to manufacturer instructions.

Cell transfection
293T cells were seeded at 60% confluence the day before transfection. Cells were transiently transfected with Lipofectamine Figure 6. A model of the metabolic stress response regulation by oncogenic BRAF in melanoma cells. Resistance to stress conditions is essential for melanoma cells survival. We propose that oncogenic BRAF V600E signaling (left panel) protects to apoptosis by regulating BH3-family members and confers resistance to low energy conditions promoting the uncoupling of LKB1 and AMPK through Erk1/2 and p90 Rsk . Under this condition BRAF mutant cells have a limited response to low energy conditions. On the right panel the inhibition of BRAF signaling allows the formation of the LKB1-AMPK complexes restoring the energy stress pathway and promoting the down-regulation of anti-apoptotic proteins such as Mcl1. The activation of AMPK by metabolic stress conditions and the inhibition of BRAF signaling would have synergistic effects promoting apoptosis. doi:10.1371/journal.pone.0004771.g006 reagent (Invitrogen Inc.) following the manufacturer's protocol. Cells were treated and lysed 36-48 h after transfection.

siRNA transfection experiments
Scramble siRNA, human BRAF siRNA On target-smartpool, and human siRNA AMPKa1 and AMPKa2 On-target-smartpools were purchased from Dharmacon. 100 nM siRNA was transfected into cells using Lipofectamine 2000 (Invitrogen Inc.) following manufacturer protocol. Experiments were performed 72 hours after transfection.

Reagents and Western Blot analysis
PHA c-Met specific inhibitor (Pfizer) was diluted in DMSO and used at the concentrations indicated. Mek1/2 inhibitor U0126 and PI3K inhibitor LY294002 (Cell Signaling) were used at 10 mM concentration. P90RSK inhibitor BI-D1870 was purchased from MSI/WTB University of Dundee and used at 10 mM. Cells were treated with the inhibitors for 2 hours under serum starvation and then treated with HGF (40 ng/ml) for 10 min. Five mg of phosphoproteins or 50 mg of total protein lysates were separated by SDS-PAGE and transferred to a PVDF membrane (Millipore). The membranes were blocked in 5% milk (Santa Cruz) and blotted against different primary antibodies. ERK2 and LKB1 were from Santa Cruz. Anti-DYKDDDDK (Flag), phospho-Erk1/2 (Thr202/ Tyr204), Erk1/2, phospho-ACC (Ser79), p-90 RSK (Thr359/ Ser363), AMPKa, p-AMPK (Thr172), phospho-S6 ribosomal protein (Ser235/236); phosho-Bad (Ser112), anti-Bad and Bim were purchased from Cell Signaling. Additionally p-Bad (Ser112) was purchased from Genscript Co., Anti-GST polyclonal antibody and anti-Flag was purchased from Sigma-Aldrich and Genscript Co. and GAPDH was purchased from Trevigen. Mcl-1 antibody was from DAKO. Anti-Flag resin was obtained from Sigma-Aldrich and glutathione-resin was purchased from Amersham and Genscript Co. Membranes were developed using horseradish linked secondary antibodies (GE Healthcare) and ECL (GE Healthcare).
Immunoprecipitations 36-48 h after transfection cells were treated as needed and lysed in RIPA Buffer containing a protease cocktail II inhibitor (Sigma-Aldrich). 800-1000 mg of total protein was subjected to immunoprecipitation with 30 ml Flag-resin or 30 ml of Glutathion-resin. Then, samples were washed three times with RIPA buffer and SDS-loading sample buffer was added to the samples. Samples were separated by SDS-PAGE.

Cell viability and apoptosis assays
Cell viability and dead cells were counted by using Guava ViaCount reagent (Gevara Technologies) cell counter (ViaCount). Apoptosis was measured using the Annexin V-EGFP apoptosis detection kit (Genscript corporation) following the manufacturer's protocol. Positive cells for Annexin V-EGFP and propidium iodide staining were analyzed and quantified by flow cytometry (FACScalibur).

Image analysis
Bands from tumor samples were quantified using NIH1.6 Image software. Normalization of p-proteins was performed against the normalized amount of the total phosphorylated protein. Other proteins were normalized against GAPDH.