AMPK promotes survival of c‐Myc‐positive melanoma cells by suppressing oxidative stress

Abstract

Although c‐Myc is essential for melanocyte development, its role in cutaneous melanoma, the most aggressive skin cancer, is only partly understood. Here we used the Nras ^Q61K INK4a ^−/− mouse melanoma model to show that c‐Myc is essential for tumor initiation, maintenance, and metastasis. c‐Myc‐expressing melanoma cells were preferentially found at metastatic sites, correlated with increased tumor aggressiveness and high tumor initiation potential. Abrogation of c‐Myc caused apoptosis in primary murine and human melanoma cells. Mechanistically, c‐Myc‐positive melanoma cells activated and became dependent on the metabolic energy sensor AMP‐activated protein kinase (AMPK), a metabolic checkpoint kinase that plays an important role in energy and redox homeostasis under stress conditions. AMPK pathway inhibition caused apoptosis of c‐Myc‐expressing melanoma cells, while AMPK activation protected against cell death of c‐Myc‐depleted melanoma cells through suppression of oxidative stress. Furthermore, TCGA database analysis of early‐stage human melanoma samples revealed an inverse correlation between C‐MYC and patient survival, suggesting that C‐MYC expression levels could serve as a prognostic marker for early‐stage disease.

Introduction

Malignant melanoma is an aggressive cancer of melanocytes and frequently originates in the skin. Cutaneous melanoma accounts for 80% of skin cancer‐related deaths (Siegel et al, 2013). Disease incidence and associated mortality rates increased continuously over the years (Gray‐Schopfer et al, 2007; Purdue et al, 2008). Neoplastic transformation of melanocytes results in mutations of proto‐oncogenes, which often cause hyperactivation of the MAP–kinase pathway (Chin et al, 2006). BRAF ^(V600E) and NRAS ^(Q61R) mutations are amongst the most frequent genetic aberrations found in patients with cutaneous melanoma. Mutation frequencies are 50–60% and 20%, respectively (Maldonado et al, 2003; Curtin et al, 2005; Xia et al, 2014). Other genetic lesions, which predispose to human melanoma formation, are loss‐of‐function mutations of the INK4A/ARF locus (Chin, 2003). More recently, next‐generation sequencing has been used to characterize the mutational landscape of human melanomas. These studies confirmed the presence of frequent driver mutations found in BRAF and NRAS but additionally identified novel genes whose mutations were associated with either NRAS‐ or BRAF‐driven melanomas (Berger et al, 2012; Xia et al, 2014). Interestingly, approximately one‐third of melanoma patients do not have any mutations in NRAS or BRAF indicating that other mutations, which do not necessarily lead to the hyperactivation of the MAP–kinase pathway, can also cause and/or influence progression of disease (Xia et al, 2014).

The bHLH zipper Myc proto‐oncogenic transcription factor family is comprised of three members: c‐, L‐, and N‐Myc. Myc expression in adult tissues is generally low and restricted to stem or progenitor cells (Laurenti et al, 2009). Myc function has been linked to many cellular processes including cell proliferation, growth, energy metabolism, and various biosynthetic pathways (Eilers & Eisenman, 2008; Kress et al, 2015). Although Myc genes are not frequently mutated in tumors, they are recurrently overexpressed in a plethora of cancers. The reason being that Myc is a downstream effector of many signaling pathways that are involved themselves in oncogenic processes. Subsequently, Myc is upregulated during disease progression. Consistently, activating mutations in MYC genes have not been identified in human melanoma, but C‐MYC has been found to be overexpressed in melanoma metastases as well as in tumor‐derived melanoma cell lines (Kraehn et al, 2001; Zhuang et al, 2008; Pouryazdanparast et al, 2012a). Overexpression of C‐MYC in metastases has been linked to copy number gain (Gerami et al, 2011; Pouryazdanparast et al, 2012a). Functionally, c‐Myc overexpression seems to be necessary to counteract oncogene‐induced senescence in NRAS^Q61R‐ or BRAF^V600E‐driven melanoma (Zhuang et al, 2008). Additional roles that c‐Myc overexpression may play during metastasis of cutaneous melanoma are currently unknown.

While specific roles of c‐Myc overexpression in melanoma require further analysis, the physiological function of Myc in melanocytes has been well established during development (Pshenichnaya et al, 2012). c‐Myc loss of function (LoF) in melanocyte precursors resulted in reduced numbers of melanoblasts and mice revealed a hair graying phenotype. Interestingly, c‐Myc‐deficient melanocyte progenitors upregulated N‐Myc. Combined LoF of c‐Myc and N‐Myc resulted in a complete loss of pigmentation indicating that (i) N‐Myc partially compensates for loss of c‐Myc and (ii) Myc is essential for the melanocytic lineage.

The present study employs a metastasizing Nras ^Q61K INK4a ^−/− mouse melanoma model (Ackermann et al, 2005; Pshenichnaya et al, 2012) to address the function of c‐Myc and downstream target signaling for development, maintenance, and progression of disease. We assessed the functional role of c‐Myc either by inactivating c‐Myc or interfering with downstream target molecules. Results were compared and correlated to human melanoma for prognostic and predictive value of the disease.

Results

c‐Myc is essential for initiation of Nras‐driven INK4a‐deficient melanoma

To investigate the role of c‐Myc for melanoma development, we used a genetic LoF approach. We intercrossed mice carrying conditional alleles of c‐Myc (c‐Myc ^lox/lox; Trumpp et al, 2001) with transgenic melanoma‐prone mice (Tyr::Nras ^Q61K INK4a ^−/−). The Nras ^Q61K oncogene is expressed under the control of the tyrosinase promoter in combination with loss of the tumor suppressor INK4a (Ackermann et al, 2005). In addition, tyrosinase‐driven Cre‐recombinase ensures efficient inactivation of conditional c‐Myc alleles within the melanocytic lineage (Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox and Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox Tyr:Cre mice hereafter referred to as Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^Δ/Δ ) (Delmas et al, 2003; Ackermann et al, 2005; Pshenichnaya et al, 2012).

In agreement with previous reports (Ackermann et al, 2005), Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox mice developed primary naevi at age of 2 months that progressed with time to melanotic melanoma invading the reticular dermis and subcutis. At 6–7 months, 100% of the mice have developed melanoma and more than 30% showed metastases in lymph nodes (LN), lung, and other organs (Figs 1 and EV1A–L). In contrast, Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^Δ/Δ mice did not develop melanoma within the investigated time frame, but a hair graying phenotype with normal skin morphology (Fig 1A and C). To test whether the incapacity of developing melanoma in c‐Myc‐mutant mice is simply a consequence of melanocyte deficiency, we performed Fontana‐Masson staining for melanin and also quantified cellular melanin content (Pshenichnaya et al, 2012) of the skin from Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox , Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^Δ/Δ, C57BL/6, and NSG mice as controls. Positive staining confirmed the presence of residual melanocytes in the skin of Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^Δ/Δ mice (Fig 1A). The melanin content of Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^Δ/Δ mice was 15.9‐fold reduced compared to Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox but comparable to C57BL/6 mice (Fig 1B). This is in agreement with a previous report showing that loss of c‐Myc in the melanocytic lineage results in reduced although detectable numbers of melanocyte precursors causing a hair graying phenotype in Tyr::Cre c‐Myc ^Δ/Δ mice (Pshenichnaya et al, 2012). The result indicates the inability to develop melanoma is not simply due to a complete loss of melanocytes during development.

Figure 1. c‐Myc is essential for melanoma initiation in Tyr::Nras ^Q61K INK4a ^−/− mice.

1. Representative gross morphology of a Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox melanoma bearing mouse (4 months) and an age‐matched tumor free Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^∆/∆ mouse (top row). Histological analysis (Fontana‐Masson stain) of skin sections derived either from a Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox mouse or from a Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^∆/∆ mouse showing normal skin architecture (bottom row). Scale bars on images represent 200 μm (40× magnification).
2. Bar graphs represent melanin concentration in the skin of indicated genotypes and are shown as mean ± standard deviation (s.d.). A significant decrease (15.9‐fold) in melanin concentration was observed in skin samples collected from Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^∆/∆ animals (n = 7, dark gray bar) compared to Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox (n = 6, black bar). Skin samples from C57BL6/J (n = 5, light gray bar, WT) and NSG (n = 5, white bar) were included as base line and negative controls, respectively. Skin samples from 4‐month‐old animals were analyzed. Each sample was measured at least twice. **P < 0.03, ***P < 0.01; Student's t‐test.
3. Kaplan–Meier graph of melanoma incidence of Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox (n = 100) and Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^∆/∆ (n = 100) mice, for melanoma incidence scoring refer to materials and methods.

Data information: See also Fig EV1.

Figure EV1. Tyr::Nras ^{Q61 K}INK 4a ^−/− mouse model undergoes development of metastasis.

• A–L
  Representative macroscopic pictures and H&E staining of sections taken from skin, lung, and lymph nodes (inguinal) of a metastatic melanoma bearing Tyr::Nras ^Q61K INK4a ^−/− mouse compared to an age‐matched wild‐type (WT) control animal. Picture panels: (A, C and E) represent wild‐type (WT) skin, lung, and lymph node, respectively. (G, I and K) depict metastatic Tyr::Nras ^Q61K INK4a ^−/− skin, lung, and lymph node, correspondingly. H&E staining sections shown in panels (B, D, and F) represent WT skin, lung, and lymph node. Panels (H, J, and L) show metastatic Tyr::Nras ^Q61K INK4a ^−/− skin, lung, and lymph node, respectively. Scale bar for panels (A, C, E, G, I, and K) = 5 mm; scale bar for panels (B,D,F, H, J, and L) = 500 μm.

Increased c‐Myc expression correlates with high tumor initiation potential and is preferentially confined to metastatic melanoma

We investigated next whether c‐Myc protein expression may change during disease progression analyzing primary versus metastatic sites in our (Tyr::Nras ^Q61K INK4a ^−/− ) melanoma animals. Thus, we made use of GFP‐c‐Myc knock‐in reporter mice (c‐Myc ^G/G), which express a functional GFP‐c‐Myc fusion protein (Huang et al, 2008). c‐Myc ^G/G were intercrossed with Tyr::Nras ^Q61K INK4a ^−/− mice to generate c‐Myc ^G/G Tyr::Nras ^Q61K INK4a ^−/− reporter animals (hereafter c‐Myc ^rep mice) (Fig 2A). c‐Myc protein expression in primary and metastatic tumors in c‐Myc ^rep mice was analyzed at 7 months of age. Interestingly, CD45⁻CD31⁻ melanoma cells revealed an increase in both relative numbers and expression levels of GFP‐c‐Myc‐positive cells (hereafter c‐Myc^hi) at metastatic sites compared to primary tumor. At metastatic sites (LN, spleen, and lung), the percentage of c‐Myc^hi cells ranged from 36 to 85% compared to only approximately 4% at the primary tumor site (Fig 2B). Next, tumor initiation capacity was assessed comparing c‐Myc^hi melanoma cells versus c‐Myc^lo cells. Thus, one thousand CD45⁻CD31⁻ Myc^hi or ^lo cells were FACS sorted from primary tumors and transplanted in Matrigel™ subcutaneously (s.c.) into NSG mice. c‐Myc^hi cells initiated tumor growth within 25 days post–transplantation, while tumor growth of c‐Myc^lo cells was detectable only 90 days post‐transplantation (Fig 2C). No metastases were observed. Ninety‐five percent of tumor cells derived from Myc^hi cells retained c‐Myc expression at experimental end‐stage analysis. Interestingly, 40% of melanoma cells derived from c‐Myc^lo cells were c‐Myc^hi 100 days post‐transplantation indicating that c‐Myc^lo cells can give rise to c‐Myc^hi tumors (Fig 2C).

Figure 2. c‐Myc is preferentially expressed in metastatic melanoma and correlates with high tumor initiation potential.

1. Schematic depiction of the experimental strategy to generate a c‐Myc reporter melanoma mouse model intercrossing Tyr::Nras ^Q61K INK4a ^−/− with c‐Myc ^G/G reporter mice bearing a GFP cDNA targeted in frame into the second exon of the endogenous c‐Myc locus generating Tyr::Nras ^Q61K INK4a ^−/− c‐myc ^G/G animals.
2. Histograms show representative flow cytometric analysis of GFP‐c‐Myc expression in melanoma cells (CD45⁻CD31⁻DAPI⁻) isolated from primary (skin) and metastatic sites (lymph node, lung, spleen) of 7‐month‐old Tyr::Nras ^Q61K INK4a ^−/− c‐myc ^G/G mice (gray filled histograms). Numbers indicate percentage of GFP⁺c‐Myc⁺ cells. Melanoma cells from age‐matched Tyr::Nras ^Q61K INK4a ^−/− mice (open histograms) were used as control (n = 3).
3. Graph shows tumor initiation and growth of GFP‐c‐Myc (“hi”, black filled circles; GFP‐c‐Myc^hi) and (“lo”, gray filled triangle; GFP‐c‐Myc^lo) cells derived from metastatic lung melanoma of a Tyr::Nras ^Q61K INK4a ^−/− c‐myc ^G/G, which were s.c. injected in Matrigel™ into NSG mice (n = 5 mice per group). Data points are presented as mean ± standard deviation (s.d.). Histograms show representative flow cytometric analysis of GFP‐c‐Myc expression after tumor resection at endpoint analysis as indicated (n = 3). GFP expression was analyzed on CD45⁻CD31⁻DAPI⁻ cells. Pictures show gross morphology of s.c. grown tumors at endpoint analysis as indicated.

Data information: See also Fig EV2.

Previous reports identified the neural crest stem cell (SC) marker CD271 as a potential cancer SC marker in human melanoma (Boiko et al, 2010; Civenni et al, 2011). Whether CD271 could also be a potential cancer SC marker in our murine melanoma model, its expression was analyzed in c‐Myc ^rep mice in primary and metastatic tumors. Although CD271 expression was mostly confined to c‐Myc^hi melanoma cells, transplantation experiments of CD271⁻c‐Myc^hi, CD271⁺c‐Myc^hi and CD271⁺c‐Myc^lo, CD271⁻c‐Myc^lo failed to confirm such cancer SC role for CD271 in our model. Differential CD271 expression of c‐Myc^hi cells did not significantly alter tumor initiation potential (Fig EV2).

Figure EV2. Differential CD271 expression in Tyr::Nras ^{Q61 K}INK 4a ^−/−‐driven melanoma does not alter tumor‐initiating potential.

1. Flow cytometric analysis shows representative FACS plots of GFP‐c‐Myc and CD271‐expressing (CD45⁻CD31⁻DAPI⁻) melanoma cells isolated from primary (skin) or metastic sites (LN) of a 7‐month‐old Tyr::Nras ^Q61K INK4a ^−/− c‐myc ^G/G animal.
2. Graph shows tumor initiation and growth of sorted GFP‐c‐Myc^hi versus GFP‐c‐Myc^lo and/or CD271 (+) and (−) melanoma cells in Matrigel™ (1,000 s.c. cells per mouse) from Tyr::Nras ^Q61K INK4a ^−/− c‐myc ^G/G mice injected into Rag2γ_c ^−/− recipients. Groups were as follows: black filled diamond, CD271⁺GFP^hi; black squares, CD271⁻GFP^hi; gray triangles, CD271⁺GFP^lo; gray squares, CD271⁻GFP^lo (n = 5 mice per group). For each time point, data are presented as mean ± standard deviation (s.d.).

c‐Myc is essential for survival of Tyr::Nras ^Q61K INK4a ^−/−‐driven melanoma

Although c‐Myc is essential for the initiation of Tyr::Nras ^Q61K INK4a ^−/−‐driven melanoma and correlates with high tumor initiation potential, its function in established melanoma is unclear. Thus, we generated two c‐Myc‐positive melanoma cell lines, mM1 and mM2, derived from Tyr::Nras ^Q61K INK4a ^−/− and Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox mice, respectively. We stably introduced a tamoxifen‐inducible Cre‐ERT transgene followed by subsequent treatment with either vehicle (ETOH) or 4‐OH‐tamoxifen (4‐OHT) to excise the conditional c‐Myc alleles. Cell growth of mM2 in vitro was monitored using the alamarBlue^® assay up to 72 h post‐treatment. Vehicle‐treated melanoma cells showed continuous cell growth (Fig 3A, left panel). In contrast, efficient c‐Myc inactivation as determined by Western blot analysis resulted in severely impaired growth and caused apoptosis of Myc‐deficient cells (Fig 3B, left panel). Cre‐ERT‐infected melanoma mM1 cells grew normal in the presence of both ETOH and 4‐OHT confirming that apoptosis is indeed a consequence of loss of c‐Myc function and not due to 4‐OHT treatment or aberrant Cre‐ERT activity (Fig EV3A and B). Genetic c‐Myc LoF results were further validated using the BET bromodomain protein inhibitor JQ1, which was previously shown to inhibit BRD4 at superenhancers, some of which are implicated in regulating c‐Myc transcription (Delmore et al, 2011; Loven et al, 2013). BET inhibition by (+)JQ1 but not the (−)JQ1 enantiomer resulted in reduced c‐Myc expression levels in mM1 and mM2 cells. This correlated with reduced cell growth and induction of apoptosis (Figs 3A and B, and EV3C). Reducing Myc protein levels in the frequently used B16F10 murine cell line revealed similar results (Fig EV3D).

Figure 3. c‐Myc is essential for survival of Tyr::Nras ^Q61K INK4a ^−/−‐driven melanoma in vitro .

• A, B
  Proliferation using the alamarBlue^® staining assay (A) and analysis of apoptosis using AnnexinV/7AAD staining (B) was analyzed on Tyr::NRAS ^Q61K INK4a ^−/− c‐Myc ^lox/lox melanoma cells (mM2) stably expressing Cre‐ERT. These cells were treated with either ethanol (ETOH) or 4‐OH‐tamoxifen (4‐OHT) to inactivate the conditional c‐Myc gene (left panels) or treated with the bromodomain inhibitor (+)JQ1 to lower c‐Myc expression (right panels). The inactive (−)JQ1 enantiomer was used as vehicle control for (+)JQ1. Assays were performed at three independent time points (24, 48, and 72 h). Cre‐ERT‐ or JQ1‐mediated gene inactivation and expression of c‐Myc was controlled by Western blot analysis 48 h after treatment. Data are presented as mean ± standard deviation (s.d.) of one representative experiment. Two independent experiments were performed, and in each individual experiment, all data points were done in triplicates (**P < 0.03, ***P < 0.01; Student's t‐test).

Data information: See also Fig EV3.

Figure EV3. JQ1‐mediated downregulation of c‐Myc in Tyr::Nras ^{Q61 K}INK 4a ^−/− melanoma cells interferes with cell growth and survival.

• A, B
  Proliferation assay using alamarBlue^® staining on Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^+/+ mM1 melanoma cells (A) or mM1 melanoma cells stably transduced with a Cre‐ERT construct (B) treated with either 4‐OH‐tamoxifen (4‐OHT) to or ethanol (ETOH) as vehicle control. Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates.
• C
  Proliferation analysis using alamarBlue^® staining (left panel) and quantification of apoptosis by AnnexinV/7AAD staining (left panel) of Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^+/+ mM1 cells treated with the bromodomain inhibitor (+)JQ1 to lower c‐Myc expression levels. Inactive (−)JQ1 enantiomer was used as control. Data are presented as mean ± s.d. of one representative experiment. Two independent experiments were performed. In each experiment, all samples were done in triplicates. c‐Myc protein levels were assessed by Western blot analysis.
• D
  Quantification of apoptotic cells using AnnexinV/7AAD staining of B16F10 melanoma cells 48 or 72 h post siRNA‐mediated knockdown of c‐Myc (dark gray bars) versus control siRNA (light gray bars). Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates.
• E
  Graph shows tumor growth of Tyr::NRAS ^Q61K INK4a ^−/− c‐Myc ^+/+ (mM3) melanoma cells injected s.c. (1 × 10⁵ mM3 cells per mouse) in Matrigel™ into Rag2γ_c ^−/− mice. At day 5 post‐transplantation, animals with a tumor volume of 50 mm³ were randomized into two groups (n = 6/group), injected with either (−)JQ1 (gray filled circles; 50 mg/kg/day) or (+)JQ1 (black filled circles; 50 mg/kg/day) starting on day 5. At experimental endpoint, tumors were harvested for gross morphology (bottom panel).

Data information: ***P < 0.01; Student's t‐test.Source data are available online for this figure.

To investigate in vivo whether tumor maintenance is c‐Myc dependent, 1 × 10⁵ mM1 melanoma cells in Matrigel™ were injected s.c. into Rag2γ_c ^−/− mice. Seven days post‐transplantation, animals with a tumor volume of 100 mm³ were randomized into two groups (n = 11/group) and treated either with (+) or (−)JQ1. Tumor growth of the (+)JQ1 treated cohort was significantly retarded compared to the (−)JQ1 group (Fig 4A). A similar experiment using another Tyr::Nras ^Q61K INK4a ^−/−‐derived cell line (mM3) was performed to exclude the possibility that results are specific to mM1 cells (Fig EV3E). Furthermore, tumors were harvested at 5 and 9 days post JQ treatment for histological analysis of c‐Myc, Ki67, and cleaved caspase‐3 (Fig 4B). (+)JQ1‐treated animals show reduced staining for c‐Myc and Ki67, compared to tumors of the (−)JQ1‐treated cohort. Quantification of Ki67‐stained tumors indicates a reduction of 57% in (+)JQ1‐treated tumors compared to (−)JQ1. Cleaved caspase‐3 staining was comparable between the two cohorts indicating that in vivo (+)JQ1‐mediated tumor growth retardation might mostly be due to inhibition of proliferation. Stainings were similar for both time points investigated. These results strongly suggest that c‐Myc is essential for maintenance of transplanted melanoma cells in vivo.

Figure 4. c‐Myc is essential for survival of Tyr::Nras ^Q61K INK4a ^−/−‐driven melanoma in vivo .

1. Graph (left panel) shows tumor growth of Tyr::NRAS ^Q61K INK4a ^−/− c‐Myc ^+/+ (mM1) melanoma cells injected s.c. (1 × 10⁵ mM1 cells per mouse) in Matrigel™ into Rag2γ_c ^−/− mice. At day 7 post‐transplantation, animals with a tumor volume of 100 mm³ were randomized into two groups (n = 11/group), injected with either (−)JQ1 (gray filled circles; 50 mg/kg/day) or (+)JQ1 (black filled circles; 50 mg/kg/day) starting on day 7. For each time point, data are presented as mean ± standard deviation (s.d.). At experimental endpoint, tumors were harvested for gross morphology (right panel) (**P < 0.03; Student's t‐test).
2. Histological analysis of tumors excised 9 days post JQ treatment. Depicted are representative H&E, nuclear c‐Myc, Ki67, and cleaved caspase‐3 stainings on tumor sections taken from either (−) or (+)JQ‐treated animals. Bar graph depicts quantification of Ki67‐positive cells represented as mean ± standard deviation (s.d.) (n = 5 tumors for (−)JQ1 and 7 for (+)JQ1; ***P < 0.01, Student's t‐test). Scale bars on images represent 200 μm and 100 μm in insets.

c‐Myc alters gene expression of metabolic signaling pathways

We next wanted to elucidate potential mechanisms by which c‐Myc exerts its essential function in Nras/INK4a ^−/−‐driven melanoma. We sorted GFP‐c‐Myc^hi and GFP‐c‐Myc^lo cells from primary and metastatic tumor sites of c‐Myc ^rep mice and performed gene expression analysis. In GFP‐c‐Myc^hi compared to GFP‐c‐Myc^lo melanoma cells, 13,993 genes were found to be significantly up‐ and 3,856 downregulated (Fig 5A). GO analysis indicated that metabolic processes followed by regulation of gene expression, transcription, RNA biosynthesis, and cell‐to‐cell communication were amongst the most significantly changed processes in c‐Myc^hi melanoma cells (Fig 5B). Expectedly, many genes regulating RNA processing and transcription, cell cycle and growth, and migration and invasion were significantly upregulated in melanoma cells expressing c‐Myc (Appendix Fig S1). KEGG pathway analysis pointed again to metabolic pathways including the 5′AMP‐activated protein kinase (AMPK), mTOR, and PI3K–AKT pathways as the most significantly deregulated pathways (Fig 5C and D). We therefore focused on changes in metabolic gene signatures and tested the relevance of two important metabolic regulators 3‐phosphoinositide‐dependent protein kinase‐1 (PDK1) and the AMPK signaling pathway. PDK1 is a master serine/threonine kinase important for the activation of AKT as well as other AGC kinases (Pearce et al, 2010) and was previously linked to melanoma (Scortegagna et al, 2014). Our reasoning to investigate the AMPK pathway closer was driven by the observation of increased expression of two isoforms of the β regulatory subunit of AMPK (Prkab1 and Prkab2, which encode AMPKβ1 and β2, respectively; Fig 5D) in c‐Myc‐positive melanoma cells. Additionally, AMPK is a central energy sensor, which plays critical roles in regulating metabolism and growth (Mihaylova & Shaw, 2011; Hardie et al, 2012). We knocked down Pdpk1 (which encodes PDK1) and Prkab2 in mM1 and mM2 cells. Knockdown of Pdpk1 revealed a minor but significant effect on survival in mM2 melanoma cells. In contrast, knockdown of Prkab2 induced apoptosis to a high percentage in both cell lines indicating that the AMPK pathway is functionally important for the survival of c‐Myc‐driven melanoma cells (Fig 5E). To further uncover a mechanistic function of c‐Myc in melanoma, we focused more specifically on the AMPK signaling pathway.

Figure 5. c‐Myc expression in Tyr::Nras ^Q61K INK4a ^−/− melanoma cells alters gene expression profiles of metabolic signaling pathways.

1. Schematic depiction of the experimental design for RNA‐seq analysis on isolated melanoma cells derived from Tyr::Nras ^Q61K INK4a ^−/− c‐myc ^G/G mice. Sorted GFP‐c‐Myc^hi and GFP‐c‐Myc^lo cells were gated on CD45⁻CD31⁻DAPI⁻ cells, and RNA was isolated from GFP‐c‐Myc^hi melanoma cells (two samples were derived from skin and one metastatic lymph node), and the GFP‐c‐Myc^lo counterpart populations were used for RNA‐seq (left panel). Right panel, pie chart indicates absolute and relative numbers of differentially expressed transcripts (upregulated transcripts in red and downregulated in blue) between GFP‐c‐Myc^hi and GFP‐c‐Myc^lo cells.
2. Gene ontology analysis on upregulated genes comparing GFP‐c‐Myc^hi and GFP‐c‐Myc^lo cells.
3. KEGG pathway analysis on upregulated genes comparing GFP‐c‐Myc^hi and GFP‐c‐Myc^lo cells.
4. Heatmap of RNA expression patterns comparing changes between GFP‐c‐Myc^hi and GFP‐c‐Myc^lo cells derived from skin (primary site) and LN (metastatic site) indicating major alterations of metabolic genes. Green indicates low and red high gene expression.
5. Quantification of apoptotic cells by AnnexinV/7AAD staining performed on two mouse melanoma cell lines mM1 and mM2 upon siRNA‐mediated knockdown of Pdpk1 (red bars) or Prkab2 (green bars). Scrambled siRNA was used as control (siCtrl, blue bars). Data are presented as mean ± s.d. of n = 3 independent experiments. In each experiment, all samples were done in triplicates. Knockdown efficiencies of Pdpk1 and Prkab2 were controlled at the protein level by Western blot analysis (two right panels) 72 h post siRNA‐mediated knockdown assessing AMPKβ2 (siPrkab2) and PDK1 (siPdpk1) protein levels (*P < 0.05, **P < 0.03, ***P < 0.01; Student's t‐test).

Data information: See Appendix Fig S1.Source data are available online for this figure.

AMPK is important for survival of murine c‐Myc‐positive Tyr::Nras ^Q61K INK4a ^−/−‐driven melanoma cells

AMP‐activated protein kinase is a heterotrimer consisting of a catalytic subunit (α) and two regulatory subunits (β and γ). In mammalian cells, there are two genes encoding the AMPKα catalytic subunit (α1, α2), two β genes (β1, β2), and three γ subunits (γ1, γ2, γ3) (Hardie et al, 2012). As expression of some of these isoforms can be tissue specific (Bultot et al, 2016), we characterized the expression of the different AMPK isoforms in mM1 and mM2 cells with and without genetic inactivation or pharmacological inhibition of c‐Myc. Protein extracts from mouse liver and skeletal muscle tissues were used as positive controls to account for different cell type‐specific AMPK isoform expression. In mM1 and mM2 cells, AMPKα1, β1/β2, and γ1 are the predominantly expressed isoforms. Cre‐ERT‐mediated inactivation or pharmacological inhibition of c‐Myc in mM2 and mM1 cells resulted in a pronounced reduction in AMPKα1 and AMPKβ2, while β1 and γ1 were unaffected (Fig 6A, left panels). To investigate whether reduced expression of AMPKα1 and β2 had affected downstream targets of AMPK, we monitored phosphorylation of various AMPK substrates in c‐Myc sufficient versus c‐Myc‐depleted mM2 cells. mM2 c‐Myc‐depleted cells showed a robust reduction in phosphorylation of all tested AMPK substrates compared to control indicating that depletion of c‐Myc protein results in reduced cellular AMPK activity (Fig 6A, right panel). Whether inhibition of AMPK is sufficient to cause cell death of c‐Myc‐positive melanoma cells, mM1 and mM2 cells were treated with either vehicle or the small molecule AMPK inhibitor dorsomorphin for 48 or 72 h, respectively. Dorsomorphin treatment resulted in AMPK inhibition as indicated by reduced pRaptor phosphorylation and correlated with increased numbers of apoptotic cells (Figs 6B and EV4A). Results for mM1 and mM2 cells were comparable. Although dorsomorphin is frequently used as an AMPK inhibitor, the same compound inhibits additional signaling pathways such as BMP or ALK (Vogt et al, 2011). To confirm that dorsomorphin‐induced cell death of c‐Myc‐positive melanoma cells is indeed mediated through AMPK inhibition, we knocked down Pkrab2 and Prkaa1, which encode for the AMPKβ2 and α1 subunit in mM1, mM2, and B16F10 cells. Knockdown of Pkrab2 as well as of Prkaa1 resulted in increased cell death of c‐Myc‐positive melanoma cells (Figs 6C and D, and EV4B–D) confirming that AMPK is essential for survival of c‐Myc‐positive melanoma cells.

Figure 6. c‐Myc ensures survival of Tyr::Nras ^Q61K INK4a ^−/− melanoma cells via AMPK pathway.

• A
  Protein expression profiling of the different AMPK isoforms in mM2 mouse melanoma cells in steady state and upon c‐Myc inactivation assessed by Western blot analysis 48 h after indicated treatments. Protein extracts from mouse liver and skeletal muscle tissues were used as positive controls for different AMPK isoforms (left and middle panel). Right panel shows phosphorylation status of known AMPK targets in the mM2 mouse melanoma cell line in steady state and upon c‐Myc inactivation.
• B
  Quantification of apoptotic cells using AnnexinV/7AAD staining of mM2 melanoma cells 48 or 72 h after treatment with either the AMPK inhibitor dorsomorphin (black bars) or vehicle control (light gray bars). Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates. Dorsomorphin inhibition efficiency on the AMPK pathway was monitored by Western blot analysis assessing phosphorylation of Raptor (pS792Raptor).
• C, D
  Quantification of apoptotic cells using AnnexinV/7AAD staining of mM2 melanoma cells 48 or 72 h post siRNA‐mediated knockdown of AMPKβ2 or AMPKα1 (black bars) versus control siRNA (light gray bars). Data are presented as mean ± s.d. of one representative out of three independent experiments. In each experiment, all samples were done in triplicates. Knockdown efficiency of AMPKβ2 or AMPKα1 was confirmed by Western blot analysis at 48 and 72 h post‐treatment.
• E
  Quantification of apoptotic cells by AnnexinV/7AAD staining of mM2 melanoma cells 24 and 48 h post Cre‐ERT‐mediated inactivation of c‐Myc in presence or absence of the AMPK activator 991. Ethanol (ETOH) and DMSO were used as vehicle controls. Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates. Knockdown efficiency of c‐Myc was confirmed by Western blot analysis 48 h after treatment.

Data information: **P < 0.03, ***P < 0.01; Student's t‐test.Source data are available online for this figure.

Figure EV4. siRNA‐mediated downregulation of AMPK in mM1 and B16F10 melanoma cells induces cell death.

• A
  Quantification of apoptotic cells using AnnexinV/7AAD staining of mM1 melanoma cells 48 or 72 h after treatment with either dorsomorphin (black bars) or vehicle control (light gray bars). Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates.
• B, C
  Quantification of apoptotic cells using AnnexinV/7AAD staining of mM1 melanoma cells 48 or 72 h post siRNA‐mediated knockdown of AMPKβ2 (B) or AMPKα1 (black bars, C) versus control siRNA (light gray bars). Data are presented as mean ± s.d. of one representative out of three independent experiments. In each experiment, all samples were done in triplicates. Knockdown efficiency of AMPKβ2 or AMPKα1 was confirmed by Western blot analysis at 48 and 72 h post‐treatment.
• D
  Quantification of apoptotic cells using AnnexinV/7AAD staining of B16F10 melanoma cells 48 or 72 h post siRNA‐mediated knockdown of AMPKβ2 (Prkab2, dark gray bars) or AMPKα1 (Prkaa1, black bars) versus control siRNA (light gray bars). Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates. Knockdown efficiency of AMPKβ2 or AMPKα1 was confirmed by Western blot analysis 48 h post‐treatment.
• E
  Quantification of apoptotic cells by AnnexinV/7AAD staining of mM2 melanoma cells 24 and 48 h post Cre‐ERT‐mediated inactivation of c‐Myc in the presence or absence of a constitutive active form of AMPKα (AMPKα CA). Ethanol (ETOH) and DMSO were used as vehicle controls. Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates.
• F
  p‐AMPK immunostaining was performed on representative sections from s.c. transplanted tumors excised either 5 or 9 days post initial JQ treatment [(−)JQ or (+)JQ at 50 mg/kg at day 7 post‐transplantation]. Scale bars on images represent 200 μm.

Data information: ***P < 0.01, Student's t‐test.Source data are available online for this figure.

We then investigated whether AMPK activation would be sufficient to counteract at least partially the apoptotic phenotype in c‐Myc‐mutant melanoma cells. mM2 cells were treated with either vehicle or 4‐OHT to activate Cre‐ERT in the presence of DMSO or the highly specific allosteric AMPK activator, 991 (Xiao et al, 2013). The ability of 991 to activate AMPK was verified by analyzing phosphorylation of bona fide AMPK substrates by Western blot analysis in MEFs and mM2 cells (Fig 6A and E right panel). As expected, lowering c‐Myc levels genetically resulted in pronounced cell death of mM2 cells. In contrast, genetic depletion of c‐Myc in the presence of 991 or the expression of a constitutive active form of AMPKα1 (Crute et al, 1998) significantly prevented the induction of apoptosis in mM2 c‐Myc‐depleted cells (Figs 6E and EV4E). We further confirmed in vivo that (+)JQ1‐treated melanoma bearing animals compared to (−)JQ1 treated animals revealed reduced p‐AMPK staining in tumor sections (Fig EV4F). These results suggest that activation of the AMPK pathway downstream of c‐Myc is important for survival of c‐Myc‐positive melanoma cells.

AMPK suppresses ROS in c‐Myc‐positive melanoma cells

Unraveling mechanisms of cell death in c‐Myc‐depleted melanoma cells will elucidate how AMPK downstream of c‐Myc ensures survival of tumor cells. Suppression of oxidative stress has been previously linked to the pro‐tumorigenic properties of AMPK (Liu et al, 2012; Saito et al, 2015). We thus investigated whether c‐Myc‐depleted mM2 cells undergo apoptosis due to increased oxidative stress. We found that melanoma cells with reduced c‐Myc significantly upregulated ROS (Fig 7A), which correlated with increased apoptosis (Fig 3B). To test whether increase in ROS is causative for induction of cell death, we treated c‐Myc‐depleted melanoma cells with the antioxidant/ROS scavenging agent α‐tocopherol. α‐tocopherol prevented ROS accumulation and rescued to a large extent c‐Myc‐depleted melanoma cells from cell death (Fig 7B). Since 991‐driven AMPK activation significantly prevents apoptosis in c‐Myc‐depleted melanoma cells, we assessed whether this is mediated via suppression of oxidative stress. Activation of 991‐mediated AMPK in c‐Myc‐depleted cells suppressed accumulation of ROS 48 h post‐treatment (Fig 7C), which correlated with pronounced survival compared to vehicle‐treated cells (Fig 6E). Therefore, depletion of c‐Myc and subsequent downregulation of AMPK signaling compromise survival partially by inducing oxidative stress‐mediated cell death. It appears that AMPK functions to counterbalance for deregulated c‐Myc‐induced biosynthesis, and if so, the need for AMPK could be alleviated by attenuating c‐Myc‐induced biosynthesis. Thus, mM1 and mM2 cells were treated with the AMPK inhibitor dorsomorphin in presence or absence of a RNA Pol III inhibitor, which slows down c‐Myc‐induced t‐RNA and ribosome biogenesis. Dorsomorphin‐induced apoptosis was significantly reduced in both mM1 and mM2 cells treated with the Pol III inhibitor (Fig 7D) supporting the hypothesis that AMPK provides a negative feedback loop for Myc‐induced biosynthesis. Moreover, we monitored cellular energy status by measuring adenine nucleotide levels (ATP, ADP, and AMP) in control and Myc‐depleted mM2 cells using ultra‐high performance liquid chromatography mass spectrometry (Pluskal et al, 2010; Fig EV5). Genetic c‐Myc depletion caused a robust and significant decrease in ATP and ADP levels, while AMP levels were not significantly altered, indicating that c‐Myc is an important driving force for ATP production. Consistent with the results of c‐Myc depletion, knockdown of AMPKα1/2 resulted also in significantly decreased ATP and ADP levels (Fig EV5). The reduction in ATP and ADP levels in AMPK knockdown cells was less pronounced compared to c‐Myc‐depleted cells, which is consistent with the possibility that c‐Myc exerts its important function in ATP biosynthesis not exclusively through AMPK.

Figure 7. AMPK suppresses ROS in c‐Myc‐positive Tyr::NRAS ^Q61K INK4a ^−/− melanoma cells.

1. Representative flow cytometric histograms showing ROS production measured by the H2DCF‐DA probe in mM2 cells 48 h post c‐Myc inactivation (blue line, 4‐OHT). Ethanol (red line, ETOH) was used as vehicle control. Gray filled histograms represent untreated cells (n = 3). Deletion of c‐Myc was assessed by Western blot analysis.
2. Representative flow cytometric histograms (top panels) showing ROS production measured by the H2DCF‐DA probe in mM2 cells 48 h post c‐Myc inactivation (blue line, 4‐OHT) in the absence (vehicle; left panel) or presence (right panel) of the antioxidant vitamin E (α‐tocopherol, 100 μM). Ethanol (red line, ETOH) was used as vehicle control. Gray filled histograms represent vehicle alone or α‐tocopherol alone treated cells. Bar diagram represents quantification of apoptotic cells by AnnexinV/7AAD staining of mM2 melanoma cells 48 h post‐inactivation of c‐Myc in the presence or absence of the antioxidant vitamin E (α‐tocopherol, 100 μM). Ethanol (ETOH) or 4‐OH‐tamoxifen (4‐OHT) was used as vehicle control (left bottom panel). Deletion efficiency of c‐Myc was controlled by Western blot analysis (right bottom panel). **P < 0.03, ***P < 0.01, Student's t‐test.
3. Representative flow cytometric histograms showing ROS production measured by the H2DCF‐DA probe in mM2 cells 48 h post c‐Myc inactivation in the presence or absence of the AMPK activator 991. Ethanol (ETOH) and DMSO were used as vehicle controls. Two independent experiments were performed.
4. Bar graph depicts quantification of apoptotic cells using AnnexinV/7AAD staining of mM1 and mM2 melanoma cells 48 h post‐treatment: DMSO (white bar), DMSO + RNA Pol III inhibitor (light gray bar, 50 μM), dorsomorphin (dark gray bar, 10 μM), and RNA Pol III inhibitor + dorsomorphin (black bar, 50 μM and 10 μM, respectively).

Data information: Data are presented as mean ± s.d. of one representative out of three independent experiments. In each experiment, all samples were done in triplicates. **P < 0.03, ***P < 0.01, Student's t‐test.Source data are available online for this figure.

Figure EV5. c‐Myc is a driving force for ATP production.

1. Bar graph (left panel) depicts adenine nucleotide (ATP, ADP, and AMP) levels in mM2 cells either left untreated (Ctrl), treated with ethanol (ETOH) or genetic depletion of c‐Myc (4‐OHT) 48 h post‐treatment, and normalized to total amino acid content. Right panel shows phosphorylation status of known AMPK targets in the mM2 mouse melanoma cell line in steady state and upon c‐Myc inactivation 48 h post indicated treatments (untreated, Ctrl; vehicle, ETOH; c‐Myc inactivation, 4‐OHT).
2. Bar graph (left panel) depicts ATP, ADP, and AMP nucleotide measurements in mM2 cells using siRNA‐mediated knockdown of AMPKα1 (siPrkaa1, black bars) versus control siRNA (siCtrl, light gray bars) 48 h post‐treatment and normalized to either protein content or total amino acid content. Right panel shows phosphorylation status of known AMPK targets in the mM2 mouse melanoma cell line in steady state and upon c‐Myc inactivation 48 h post indicated treatments (control siRNA, siCtrl; AMPKα1 depletion, siPrkaa with two (1,2) independent siRNAs).

Data information: Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done at least in triplicates. **P < 0.03, ***P < 0.01, Student's t‐test.Source data are available online for this figure.

Elevated C‐MYC expression correlates with reduced median survival of melanoma patients

Our data indicate that elevated c‐Myc protein expression is preferentially confined to metastatic sites in our mouse melanoma model. To investigate whether this is also the case in human melanoma, we tested C‐MYC protein expression in 15 patient‐derived human melanoma cell lines (Caballero et al, 2010; Nikolaev et al, 2011). They were either derived from primary tumors (5 cell lines) or from metastatic sites (10 cell lines). Three out of five melanoma cell lines from the primary tumor site showed low but detectable protein C‐MYC expression (LAU‐T126, LAU‐T921A, and LAU‐Me36), whereas in the two remaining cell lines (LAU‐Me300 and LAU‐Me 300A) C‐MYC expression was below detection limit. In contrast, 7 (LAU‐Me 252, LAU‐Me275, LAU‐T618A, LAU‐T1194, LAU‐T12, LAU‐T38713, and LAU‐T33) out of 10 melanoma cell lines derived from metastatic melanoma patient samples showed high C‐MYC expression levels (Fig 8A). There was no correlation between C‐MYC expression levels and NRAS or BRAF‐mutated patient samples, as C‐MYC expression was detected in either patient samples. To investigate the MYC‐AMPK‐ROS connection in human melanoma samples, we first analyzed the protein composition of the AMPK subunits in tumor cell lines of primary (LAU‐T921A) and metastatic (LAU‐Me 252) origin with low C‐MYC expression and two of metastatic origin (LAU‐Me275, LAU‐T333A) with high C‐MYC expression. All AMPK subunits were expressed though at different levels in the cell lines investigated, with the exception of AMPKα2 in LAU‐T333, which was not detected (Fig 8B). To test whether AMPK signaling in these four cell lines is important for survival, we knocked down PRKAA1/2 (AMPKα1/2). Interestingly, knockdown of AMPKα1/2 induced apoptosis and ROS production significantly in LAU‐T921A and LAU‐Me252 cells with high C‐MYC levels, while survival and ROS production was not significantly changed in LAU‐T921A‐ and LAU‐Me 252‐expressing low levels of C‐MYC (Fig 8C and D).

Figure 8. Elevated C‐MYC expression in human metastatic cell lines.

1. Protein expression of C‐MYC in 15 human melanoma cell lines (5 primary = 1° skin tumors and 15 from metastatic origin) was assessed by Western blot analysis. Protein extracts from human melanocytes were included as negative control.
2. Protein expression profiling of human AMPK isoforms in melanoma cell lines (one primary and three metastatic) was performed by Western blot analysis.
3. Quantification of apoptotic cells using AnnexinV/7AAD staining of human melanoma lines 48 h post siRNA‐mediated knockdown of AMPKα1 (black bars) or control siRNA (light gray bars). Data are presented as mean ± s.d. of one representative out of two independent experiments. In each experiment, all samples were done in triplicates (***P < 0.01; Student's t‐test). Knockdown efficiency of AMPKα1 was confirmed by Western blot analysis 48 h post‐treatment.
4. Representative flow cytometric histograms showing ROS production measured by the H2DCF‐DA probe in human melanoma cell lines (one primary and three metastatic) 48 h post siRNA‐mediated knockdown of AMPKα1 (solid blue line) or control siRNA (solid red line). Three independent experiments with comparable results were performed.

Source data are available online for this figure.

Next, we performed immunohistochemical analysis for C‐MYC and p‐AMPK on paraffin sections of unpaired patient‐derived tissue samples (eight patients—four primary tumors stage III‐IV and four metastatic tumors). No C‐MYC or p‐AMPK expression was detected in the four primary tumors, whereas all samples from metastatic sites (axillary region or liver) stained positive for C‐MYC and p‐AMPK (Fig 9A). We further analyzed paired tumor samples from 10 patients by immunostaining for C‐MYC and p‐AMPK to obtain a better understanding of C‐MYC and p‐AMPK expression during melanoma progression. In five cases (patients 2 and 7–10), C‐MYC expression was enhanced in metastatic compared to primary tumors, whereas in four patient samples (3–6) C‐MYC immunostaining was comparable between primary tumors and metastasis (Fig 9B). Furthermore, five out of six patients (5, 6, 8–10) with low C‐MYC expression in the primary tumor were also low for p‐AMPK expression. Moderate to high p‐AMPK staining was found in 5 out of 10 metastatic patient samples (2, 4, 6, 7, and 10) three of which correlated with high C‐MYC expression (patients 4, 7, and 10) indicating that high p‐AMPK‐expressing patients represent a subgroup of the high C‐MYC‐expressing population. Interestingly, patients whose primary tumor samples were weakly to moderately positive (below 50%) for C‐MYC had a longer time interval until occurrence of metastasis compared to patients in which the primary tumor sample had already high (over 50%) to very high (all cells) C‐MYC levels (Fig 9B).

Figure 9. High C‐MYC expression in melanoma patients correlates with reduced median survival.

1. C‐MYC (set of two left panels) as well as p‐AMPK (set of two right panels) immunostaining was performed on sections from primary (skin; top row) and metastatic (axillary region, middle row; liver, bottom row) of melanoma patient biopsies. A 5× gross morphology panel set and a 40× high magnification are depicted. Scale bars represent 2.5 μm (5×) and 100 μm (40×).
2. Table shows semi‐quantitative evaluation of C‐MYC and p‐AMPK expression in primary and metastatic melanomas of 10 paired biopsies (skin, Clark/Breslow stages III and IV) and the time interval between the examined samples. Grading is indicated in the footnote.
3. Kaplan–Meier survival graph plots C‐MYC RNA expression levels of 221 melanoma patients derived from TCGA database, divided into two groups C‐MYC ^lo (blue line) versus C‐MYC ^hi (pink line), and correlated with overall survival. Solid black line indicates 50% median survival. p‐value was calculated using log‐rank statistical test.
4. Kaplan–Meier survival graph depicts 74 metastasis‐free melanoma patients from the TCGA database divided into four groups according to tumor (T) stage at diagnosis combined with C‐MYC protein expression levels (low/T(0‐2), solid blue line; high/T(0‐2), solid pink line; low/T(4b), stippled blue line; high/T(4b), stippled pink line). Solid line black indicates 50% median survival. P‐values were calculated using log‐rank statistical test.
5. Kaplan–Meier survival graph plots AMPKα protein expression levels of 127 melanoma patients derived from TCGA database, divided into two groups AMPKα^lo (blue line) versus AMPKα^hi (pink line), and correlated with overall survival. Solid black line indicates 50% median survival. P‐value was calculated using log‐rank statistical test.

To test whether C‐MYC expression levels might correlate with poor prognosis, we analyzed TCGA database case sets and correlated survival of melanoma patients based on C‐MYC RNA expression of 352 patients, which were classified into high‐ and low‐expressing cohorts. The high expression cohort corresponds to the 33% (n = 110) of patients expressing highest C‐MYC RNA levels, whereas the low cohort includes the 33% (n = 111) of patients with the lowest expression. The Kaplan–Meier plot demonstrates that the patient cohort with high C‐MYC RNA expression levels has a significant reduction in median survival time (MST) of 3.6 years compared to patients with low C‐MYC expression (Fig 9C). The difference is even more significant if analysis of melanoma patients is based on C‐MYC protein expression correlated with tumor T stage at diagnosis. The analysis was performed on a cohort of 197 patients for which C‐MYC protein expression levels and corresponding tumor pathology information were available. Melanoma patients were divided into four groups: C‐MYC^lo/T0‐2; C‐MYC^hi/T0‐2; C‐MYC^lo/T4b; and C‐MYC^hi/T4b. Differential C‐MYC protein expression did not correlate with MST in T4b patients. In contrast, high C‐MYC protein expression in T0‐2 patients correlated significantly with reduced MST. The difference in reduced MST between the C‐MYC high and low patient cohorts was > 4.9 years (Fig 9D). These data support the hypothesis that elevated C‐MYC expression at early‐stage melanoma correlates well with reduced patient MST. Whether AMPK expression levels might correlate with poor prognosis, TCGA database case sets were analyzed and correlated with survival of melanoma patients based on AMPKα protein expression of 192 patients. They were classified into the 33^rd percentile of patients with highest and lowest expression. Although not significant, patients that survive beyond 3 years of diagnosis show a clear trend with reduced MST of 5.8 years between AMPKα high and low patient cohorts (Fig 9E).

Discussion

c‐Myc is one of the best‐studied oncogenes promoting tumorigenesis in many tissues (Kress et al, 2015; Dejure & Eilers, 2017; Kalkat et al, 2017). It is frequently over expressed in cancer, correlates with tumor aggressiveness and poor clinical outcome (Nesbit et al, 1999; Beroukhim et al, 2010). Although gene‐targeting experiments for c‐Myc have revealed an essential role during melanocyte development (Pshenichnaya et al, 2012), its function in cutaneous malignant melanoma is not well known. Here, we have used a tyrosinase Nras ^Q61K‐driven INK4a‐mutant melanoma mouse model (Ackermann et al, 2005) to demonstrate that c‐Myc is essential during initiation, progression, and maintenance of disease, in part through activating AMPK signaling to suppress oxidative stress.

Ablation of c‐Myc during tumor initiation never led to melanoma formation, but instead, similar to wild‐type mice with melanocyte‐specific c‐Myc inactivation, exhibited a hair graying phenotype. The hair graying phenotype is the consequence of reduced numbers of melanoblasts, which still developed into functional melanocytes (Pshenichnaya et al, 2012). Interestingly and in agreement with this study, residual melanocytes were also detected in our c‐Myc‐mutant melanoma mouse model suggesting that the inability to develop melanoma is not simply a consequence of lack of melanocytes. Why remaining melanocytes do not develop into melanoma is currently unknown. However, it is possible that c‐Myc is essential for proliferation and thereby for the expansion of tumor‐initiating cells. The fact that none of the c‐Myc‐mutant melanoma mice developed disease strongly suggests that there is no functional redundancy with the other Myc family members.

Our studies using c‐Myc ^rep mice to monitor c‐Myc protein expression during melanoma progression show that only a minor fraction of melanoma cells are c‐Myc positive at the primary tumor site. Nonetheless, protein expression levels and frequency of c‐Myc‐positive cells are increased at metastatic sites. A similar trend was observed in analysis of patient‐derived cell lines or tumor samples with low or no C‐MYC expression in melanoma from the primary tumor site and high expression at the site of metastasis. These findings are in agreement with previous reports showing that elevated C‐MYC expression is preferentially observed in metastatic human melanoma patient samples correlating with and caused by copy number gains in C‐MYC at 8q24 (Kraehn et al, 2001; Gerami et al, 2011; Pouryazdanparast et al, 2012a,b).

Moreover, melanoma patients with copy number gains at 8q24 often reveal aggressive clinical progression (Pouryazdanparast et al, 2012b). Our TCGA analysis of C‐MYC expression in melanoma patients is in line with these reports and further extends our current knowledge. C‐MYC protein expression, in particular at early tumor stage (T0‐2), seems to be a significant predictor of reduced MST. Thus, assessing early lesions of melanoma patients for C‐MYC expression would be important and could bear prognostic value.

We do not know whether elevated c‐Myc expression at metastatic sites in our mouse melanoma model is also caused by gene amplification. Nevertheless, our transplantation experiments of c‐Myc^hi versus c‐Myc^lo melanoma cells demonstrated that c‐Myc‐positive melanoma cells are more aggressive and have a higher tumor initiation potential compared to c‐Myc‐negative melanoma cells. Whether melanoma is hierarchically organized in phenotypically distinct subpopulations of tumorigenic versus non‐tumorigenic is controversially discussed (Quintana et al, 2008, 2010; Schatton et al, 2008; Boiko et al, 2010; Civenni et al, 2011).

c‐Myc expression in murine melanoma causes a constitutive growth program upregulating associated gene expression signatures including RNA biosynthesis, RNA processing, transcription, and metabolism being in agreement with the role of Myc reported in other cancers (Eilers & Eisenman, 2008; Dang et al, 2009). All cellular processes, including metabolism, have to adapt to manage increased energy consumption and production of macromolecules in order to cope with proliferative stress to ensure survival of tumor cells. In this context, we focused on AMPK, as it is a central energy and redox sensor. AMPK is activated by metabolic stresses, which lower the cellular energy status by decreasing catabolic generation of ATP or accelerating ATP consumption. Upon activation, it functions to restore cellular energy homeostasis by switching off anabolic pathways and other ATP‐consuming processes while switching on ATP‐producing catabolic pathways (Hardie et al, 2012). The role of AMPK in cancer is complex and context dependent. Historically, AMPK was first associated with tumor suppressive functions based on the discovery that the tumor suppressor serine–threonine kinase liver kinase B1 is a direct activator of AMPK (Woods et al, 2003; Shaw et al, 2004). The ability of AMPK to activate the tuberous sclerosis complex, a negative regulator of mTORC1 and frequently activated in cancer through upstream activators such as PI3K and AKT, further suggested that AMPK plays a tumor suppressive role (Lizcano et al, 2004; Inoki et al, 2006). Genetic loss of studies showing that inactivation of AMPK accelerates tumor growth in an experimental model of lymphomagenesis is also in agreement with a tumor suppressive function of AMPK (Faubert et al, 2013). Although AMPK has been associated with tumor suppressive functions, there is accumulating evidence that activation of AMPK can act pro‐tumorigenic in a context‐dependent manner (Hardie, 2015). AMPK has been shown to be important in oncogene‐induced tumorigenesis as in Hras^V12 ± PTEN‐induced astrocytic tumor cell proliferation (Rios et al, 2013) and kinase suppressor of Ras 2‐induced anoikis resistance of MEFs and several cancer cell lines (Fernandez et al, 2012). Interestingly, AMPK‐related kinase 5 (ARK5) was identified in a synthetic lethal RNAi screen for kinases that are selectively required to support tumor cells when Myc is overexpressed (Liu et al, 2012). Depletion of ARK5 prolonged survival in Myc‐driven hepatocellular carcinoma mouse models. Inhibition of AMPK signaling in Myc‐expressing cells resulted in cellular ATP collapse and caused increased levels of reactive oxygen species causing apoptosis (Liu et al, 2012). A similar role has been suggested for AMPK regulating NADPH homeostasis to promote tumor cell survival under circumstances of energy stress (Jeon et al, 2012). In a murine myeloid leukemia model, AMPK protected leukemia‐initiating cells from metabolic stress via suppression of ROS (Saito et al, 2015). Our results are in line with these reports showing that survival of c‐Myc^hi Nras ^Q61K‐driven INK4a‐mutant melanoma cells activates and is dependent on AMPK signaling to maintain ATP homeostasis and to suppress oxidative stress. Interestingly, specific AMPK subunits/isoforms (AMPKα1 and AMPKβ2) are amplified in various human tumors and coincide with activation and/or amplifications of various oncogenes including KRAS, BRAF, AKT, and C‐MYC across several cancer types such as breast, bladder cancer, and melanoma (Monteverde et al, 2015). These reports combined with our results support a pro‐tumorigenic role of AMPK signaling. Therefore, it is likely that its role is multifaceted and context dependent.

In conclusion, our results reveal an important AMPK‐dependent oncogenic mechanism by which c‐Myc‐positive melanomas ensure survival coping with tumor gene‐induced proliferative stress including suppression of oxidative stress. Moreover, we show that C‐MYC protein expression at an early stage of human melanoma has prognostic value for melanoma patients.

Materials and Methods

Mice

Tyr::Nras ^Q61K INK4a ^−/− mice were previously described (Ackermann et al, 2005) and crossed to c‐Myc ^lox/lox (Trumpp et al, 2001) and c‐Myc ^G/G (Huang et al, 2008) to generate Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox and c‐Myc ^G/G Tyr::Nras ^Q61K INK4a ^−/− (c‐Myc ^rep ), respectively. Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^lox/lox were crossed to a melanocyte‐specific Cre line—Tyr::Cre (Delmas et al, 2003) to generate Tyr::Nras ^Q61K INK4a ^−/− c‐Myc ^∆/∆. Rag2γ_c ^−/− (B6.Rag2^tm1FwaII2g^tmWjl) and NSG (NOD.Cg‐Prkdc ^scid Il2rγ^tm1Wjl/SzJ) were purchased from TACONIC and the Jackson Laboratories (United States).

Cell lines

Human melanoma cell lines have been described (Caballero et al, 2010; Nikolaev et al, 2011) and were cultured in RPMI1640 media supplemented with 10% FCS and 50 μg/ml gentamycin (RPMI10). Primary mouse melanoma cell lines (mM1, mM2, mM3) were cultured in RPMI10 supplemented with 5 μg/ml insulin (Sigma), 100 ng/ml α‐MSH (Sigma), 25 ng/ml FGF (Invitrogen), 200 ng/ml TPA (Sigma), 30 ng/ml endothelin3 (Sigma), and 5 μM dexamethasone (AppliChem). GFP‐IRES‐Cre‐ERT stable cell lines (mM1 and mM2) were generated transducing a GFP‐IRES‐Cre‐ERT construct. GFP⁺ cells were FACS sorted and expanded.

Single cell preparation of primary, metastatic, and s.c. melanomas

Tumors were minced into small pieces and digested using dispase (4 mg/ml, Gibco) and collagenase (1 mg/ml, Gibco) for 2 h at 37°C. Cell clumps were removed using a quick spin. Supernatant containing the single cell suspension was collected and filtered (40 μm cell strainer). Washing steps were repeated 3×. Two additional washes with 1× HBSS (without Mg²⁺ and Ca⁺⁺) were performed prior to FACS staining or s.c. injections.

Immunohistochemistry

Mouse organs and primary patient samples (obtained from the Department of Pathology of CHUV, Lausanne) were fixed with 4% paraformaldehyde (4°C) and embedded in paraffin; 3‐μm sections were stained with H&E using standard methods. Anti‐c‐Myc (Y69, Abcam) staining was performed according to manufacturer's instructions.

Flow cytometry

Single cell suspensions of c‐Myc ^rep mice (5 × 10⁵–1 × 10⁶ cells) were resuspended in staining media (SM; 1× HBSS, 2% FCS), stained with α‐CD31 and α‐CD45 using standard flow cytometric staining procedures, and analyzed on a Dako CyAn Flow Cytometer for GFP expression; dead cells were excluded using DAPI (CD31⁻CD45⁻DAPI⁻). Data were processed using FlowJo^® software. For subsequent RNA‐seq and s.c. transplantation assays, cells were sorted using a BC MoFlo Astrios^EQ cell sorter. Purity of GFP⁺ and GFP⁻ live cell populations (CD31⁻CD45⁻DAPI⁻) was >90%.

s.c. transplantation in vivo

In vivo transplantation assays were performed using sorted primary cell populations from c‐Myc ^rep mice (1,000 cells per recipient NSG mouse of either GFP^lo or GFP^hi (CD31⁻CD45⁻DAPI⁻) fraction) or using the mM1 or mM3 cell lines transplanting 1 × 10⁵ cells s.c. into Rag2γ_c ^−/− mice. Cells were injected into the flank of recipient mice in 100 μl of Matrigel^™. Once tumor volumes reached 50–100 mm³ in the cell line transplantation assays, c‐Myc expression was inhibited using (−), (+)JQ1 at 50 mg/kg/day injecting mice i.p.

RNA‐seq

Total RNA was prepared from GFP^hi or GFP^lo FACS‐sorted cells and collected in triplicates using RNAqueous‐Micro kit (Ambion). RNA‐seq was performed by the Genomic Technologies Facility platform (University of Lausanne, Switzerland). An RNA amplification step was done using Ovation^® RNA Amplification System V2 (Nugen). The Ovation^® Ultralow Library System V2 (Nugen) was used for library preparation. Library sequencing was done using an Illumina HiSeq 2500 sequencer. RNA‐seq mapping and analysis are described in the Appendix Supplementary Methods.

Antibodies and Western blotting

Western analysis was performed under standard methods. All antibodies used in this study are listed in the Appendix Table S1.

Cell proliferation and apoptosis detection

Cell proliferation was analyzed using the alamarBlue^® cell viability reagent (Invitrogen) according to manufacturer's instructions. Melanoma cells were plated (2,500 cells/well). Cells were treated 24 h post‐seeding with different compounds and harvested at different time points as indicated. Each condition was done in triplicates. Total apoptotic cells were assessed by AnnexinV‐Cy5/7AAD staining using the AnnexinV‐Cy5 Apoptosis Detection Kit (BD Biosciences) according to manufacturer's instructions. Samples were analyzed on a Dako CyAn Flow Cytometer and data processed using FlowJo^® software.

siRNA

siRNA transfections were performed using Lipofectamine RNAiMAX (Invitrogen) according to manufacturers manual. Different siRNAs were purchased as smartPools (FlexiTube GeneSolution, GS17869) from Qiagen. Allstar negative control (Qiagen, SI03650318) was used as negative control.

In vitro assays

ROS staining

ROS analysis was detected using H2DCFDA (Sigma) staining. Twenty‐five thousand cells were stained with H2DCFDA at a final concentration of 50 μM. Cells were incubated for 45–60 min under standard culture conditions (37°C, 5% CO₂), and then, H2DCFDA fluorescence was analyzed on a Dako CyAn Flow Cytometer. Data were processed using FlowJo^® software. Each condition was analyzed in triplicates in three independent experiments.

c‐Myc inhibition

c‐Myc depletion in c‐Myc ^lox/lox Cre‐ERT‐expressing cells was induced treating cells with 4‐hydroxy‐tamoxifen (4‐OHT) (Sigma) at a final concentration of 1 μM in ethanol (ETOH). ETOH was used as vehicle control. c‐Myc expression was downregulated treating cells with the BRD4 inhibitor (+)JQ1 (Selleckchem) at a final concentration of 1 μM. The (−)JQ1 (Sigma, 1 μM) inactive enantiomer was used as control.

AMPK pathway modulation

AMP‐activated protein kinase pathway signaling was inhibited treating the cells with dorsomorphin (Sigma, 10 μM). AMPK pathway activation was induced treating cells with 991 (5‐{[6‐chloro‐5‐(1‐methylindol‐5‐yl)‐1H‐benzimidazol‐2‐yl]oxy}‐2‐methyl‐benzoic acid (Bultot et al, 2016); compound at a final concentration of 10 μM.

Antioxidant treatment

α‐tocopherol (Sigma; 100 μM) was used as antioxidant treatment to counteract ROS‐induced apoptosis. Cells were pretreated with α‐tocopherol for 24 h and subsequently maintained continuously in the media for the whole experiment.

Kaplan–Meier and statistical analysis

Protein and RNA‐seq expression levels from melanoma patients were downloaded from the TCGA website. Kaplan–Meier survival curves were created with survfit function from the survival R package and statistical comparisons with the function comp from the survMisc R package. For comparison of pooled data between two groups, unpaired t‐tests were used to determine significance (*P < 0.05, **P < 0.03, ***P < 0.01).

Adenine nucleotide measurements

Sample preparation for adenine nucleotide analysis was based on Gheldof et al (2017), see Appendix Supplementary Methods.

Ethics statement

All animal work was conducted according to Swiss national guidelines. All mice were kept in the animal facility under EPFL animal care regulations. This study has been reviewed and approved by the cantonal veterinary office.

Author contributions

AK, MA, CC, JS‐D, MPG, and SC performed experiments and analyzed data. ML performed statistical and data base analysis. LL analyzed pathological slides from human melanoma patients. UK, AT, SM, KS, and FB analyzed data and provided conceptual and experimental guidance throughout the study and helped writing the manuscript. FR conceived the study, analyzed data, and wrote the manuscript.

Conflict of interest

C. Collodet, M.P. Giner, S. Christen, S. Moco, and K. Sakamoto are employees of the Nestlé Institute of Health Sciences SA (Switzerland). The remaining authors declare that they have no conflict of interest.

Supporting information

Appendix

Expanded View Figures PDF

Source Data for Expanded View

Review Process File

Source Data for Figure 5

Source Data for Figure 6

Source Data for Figure 7

Source Data for Figure 8

The EMBO Journal (2018) 37: e97673