The Slow Cycling Phenotype: A Growing Problem for Treatment Resistance in Melanoma

Targeted therapy of metastatic melanoma carrying BRAF oncogenic mutations is heralded as a breakthrough into an era of personalized medicine (1, 2). Inhibition of the hyperactive MAPK pathway in BRAF-mutant melanomas induces rapid tumor regression that significantly improves patient survival. However, clinical management of melanoma using targeted therapy remains beset by short-lived treatment responses (2, 3). This is largely due to subpopulations of tumor cells that survive drug cytotoxicity, and quickly acquire new mechanisms of resistance to enable tumor regrowth. Even upon tumor remission, where no trace of melanoma is detectable by clinical examinations or radiographic imaging, relapse can subsequently occur, which suggests that a small residual number of melanoma cells remains undamaged throughout the treatment (4).

Tumor heterogeneity is regarded as one of the main reasons for mixed responses to treatment. A frequently observed pattern in melanoma patients is tumor regression followed by treatment resistance. Furthermore, different tumors from the same patient may display a wide variety of treatment responses, despite sharing the same oncogenic mutations, illustrating the complexity of melanoma heterogeneity that extends beyond genomics (5).

Accumulating evidence suggests that distinct tumor cell phenotypes underlie heterogeneous treatment responses in patients, independent of oncogenic mutations. Slow-cycling or growth-arrested melanoma cells, which appear senescent, are increasingly described as a major determinant for treatment resistance (6–8). The majority of treatment strategies, including chemotherapy and targeted therapies, target the bulk of the tumor with a proliferative phenotype, whereas slow-cycling cancer cells may be left undamaged (8, 9). Consequently, proliferating melanoma cells undergo rapid apoptosis, while those that either retain or switch to a slow-cycling phenotype escape therapeutic stress. Indeed, studies demonstrate that BRAF-mutant melanoma cells exhibit increased features of senescence upon exposure to targeted therapies such as vemurafenib (which targets the BRAFV600E mutation; ref. 10) and chemotherapy (11).

Although it has long been thought that inhibition of tumor cell growth is a favorable outcome, and is analogous to tumor cell killing, more recent studies indicate that tumor cells with a slow-cycling phenotype may be metabolically active and highly aggressive, with increased potential to form new tumors and metastasize (8, 12–14). This suggests an urgent need to therapeutically target the slow-cycling phenotype in tumor cells, along with the bulk of the tumor that is proliferative, to increase tumor cell death and reduce the risk of developing treatment resistance. Recently, there have been several studies that may offer insight into the molecular mechanisms underlying slow-cycling cells. This review briefly overviews molecular changes associated with the slow-cycling phenotype in melanoma, and potential therapeutic implications. We discuss several key features of molecular reprogramming in melanoma, accompanied by the rewiring of signaling cascades mediating dedifferentiation, chromatin remodeling, and a shift in the metabolic state of oxidative phosphorylation (12).

Microphthalmia-associated transcription factor (MITF) is a master regulator of development and differentiation in the melanocytic lineage (15). In melanoma, MITF also plays a critical role in cell phenotype and tightly controls the transcriptional activity and repression of a large array of genes. Indeed, low and high MITF levels, respectively, have been associated with distinct global gene expression patterns (16, 17). Low levels of MITF have been shown to drive dedifferentiation in melanoma cells as demonstrated by the effect of depletion of MITF leading to increased expression of stem cell markers, while decreasing expression of melanocytic differentiation markers (6, 18). In parallel to a dedifferentiated phenotype, depleting MITF has been shown to reduce cell growth and induce senescence and/or apoptosis (6, 11, 18–21). MITF silencing in human melanoma cell lines leads to increased expression of cell-cycle inhibitors including p21 (CDKN1a) and p27 (CDKN1b; refs. 6, 11, 21). Moreover, silencing MITF triggers a DNA damage response (DDR) pathway that upregulates p53, which then mediates cell-cycle arrest (11). The characteristics of dedifferentiation and a slow-cycling phenotype are key features of malignant melanoma stem or initiating cells (MMIC), which have crucial roles in tumor initiation, progression, metastasis, and treatment resistance. While MITF is an important regulator of the differentiation of melanocytes from a neural crest precursor cell lineage, a full discussion of MMICs in the context of melanoma is beyond the scope of this article and has been reviewed by others (22–24).

The “low MITF” slow-cycling state has important therapeutic implications in melanoma. Low MITF activity confers intrinsic and acquired resistance to MAPK pathway inhibitors in BRAF-mutant melanoma cell lines and is associated with vemurafenib-resistance in patient biopsies (25–28).

The therapeutic resistance associated with the low MITF dedifferentiated state suggests that mechanisms promoting survival in response to MAPK pathway inhibitors have been activated. Current clinical strategies involving targeted treatment generally target melanoma cell populations that are “addicted” to the hyperactive MAPK pathway for survival and rapid cell division. However, increasing evidence suggests that melanomas in a low MITF–mediated dedifferentiated state have “rewired” signaling cascades that redistribute oncogenic signaling toward multiple pathways, including non-MAPK pathways, in addition to the MAPK pathway (25, 26, 28). In other words, instead of total dependence on a single growth-regulatory pathway, dedifferentiation allows usage of multiple pathways to confer drug resistance (12).

Upon MAPK pathway inhibition, a low MITF state in small subpopulations of slow-cycling cells within tumors allows them to utilize alternative mechanisms for survival. This correlates with various signaling activators and confers treatment resistance, and therefore permits time for slow-cycling cells to acquire additional mutations and/or epigenetic aberrations that subsequently can allow tumor regrowth and tumor relapse (29). In the following, we discuss several of the key signaling activators, including WNT5A and EGFR, which are associated with low MITF expression and with the slow-cycling phenotype.

WNT5A signaling occurs via a noncanonical WNT pathway and elicits dedifferentiation by suppressing MITF levels (30), promoting a senescence-like cellular phenotype (14). Melanoma cells with high WNT5A expression were shown to respond to therapeutic stress induced by vemurafenib treatment by further increasing WNT5A signaling, which then led to cell-cycle arrest via upregulation of the cell-cycle inhibitor, p21. Increased levels of conventional senescence markers including senescence-associated β-galactosidase (SA-β-gal), senescence-associated heterochromatic foci (SAHF), H3K9Me chromatin markers, and PML bodies accompanied the growth arrest (14). However, although cell division was suppressed, cells expressing high levels of WNT5A were able to reenter the cell cycle, indicating they were not truly senescent (thus termed senescence-like).

High WNT5A expression levels in melanoma cell lines and tumors from patients prior to treatment, were associated with significantly poorer response to vemurafenib compared with tumors and cell lines with low WNT5A expression levels, which suggests that WNT5A signaling may be functionally linked to innate resistance (31, 32). In addition, following BRAF inhibitor treatment, melanoma cell lines and tumors generally had increased expression of WNT5A (31, 32). Potential BRAF inhibitor resistance mechanisms associated with WNT5A signaling include reactivation of the MAPK pathway; indeed, not only does WNT5A signaling lead to resistance to BRAF inhibitors, but elevated WNT5A expression also increases ERK activity in melanoma cell lines (32). Mechanisms that result in activation of the MAPK pathway lead to reduced dependency on the activity of mutant BRAF, including switching to an alternative RAF isoform such as CRAF (33, 34), and activation of the FGF receptor 3/RAS–signaling axis (35). WNT5A has also been shown to activate the PI3K–AKT pathway via the WNT receptors FZD7 and RYK (31).

WNT5A-expressing growth-arrested cells were found to possess enhanced invasiveness upon increasing therapeutic dosage (14). Given that WNT5A suppresses MITF, the increased metastatic potential of WNT5A-expressing cells is consistent with reports that low MITF also confers increased invasiveness (16, 21, 32, 36). This indicates that targeted therapy may rapidly kill proliferative cells, but that it may simultaneously promote invasiveness and metastasis in a subpopulation of slow-cycling WNT5A-high cells that are able to adaptively resist therapeutic stress. Indeed, WNT5A-high cells were capable of reentering the cell cycle and forming new tumor colonies at distant body sites in in vivo experiments (14). There are a number of reports supporting the notion that BRAF inhibitors promote metastasis. For example, proliferative melanoma cell lines adopt invasive characteristics upon MAPK pathway inhibition in in vitro experiments (27), and resistance to vemurafenib selects for a highly malignant phenotype with increase metastatic potential to the lungs in in vivo mouse experiments (37). Consistent with a dedifferentiated state, the metastatic cells in the lungs of these mice were shown to express higher levels of cancer stem cell markers such as CD271, JARD1B, and fibronectin. Also in the clinical setting, targeted therapy has been reported to promote the development of new metastases during treatment relapse (38).

In contrast to regulation of a slow-cycling phenotype by WNT5A, there are also reports that WNT5A is associated with proliferation (31). In two melanoma cell lines, WNT5A overexpression increased proliferation, whereas knockdown of WNT5A reduced cell division (31), highlighting the complexity of the regulation of cell division by WNT5A.

WNT5A is a potential therapeutic target for targeting slow-cycling cells as a first-line combination therapy. Knockdown of WNT5A inhibits invasive potential and ablates the formation of in vivo metastases of primary human melanoma cell lines injected into the tail vein of athymic nude mice (14, 30).

EGFR signaling and enhanced expression of EGFR induce a slow-cycling phenotype, and elevated levels of cell-cycle inhibitors including p21^Cip1 (CDKN1A), p27^Kip1 (CDKN1B), as well as activated hypophosphorylated retinoblastoma protein (7). Activation of various receptor tyrosine kinase (RTK) signaling pathways has been demonstrated to confer resistance to MAPK pathway inhibitors (7, 26, 28). However, EGFR signaling is reported as being crucial for the development of resistance to MAPK pathway inhibitors in melanoma (7, 28, 39).

When EGFR signaling was assessed in post-treatment patient biopsies compared with pre-treatment tumors, 6 of 16 patients had increased EGFR activation as detected by IHC (7). Treatment resistance attributed to EGFR signaling was also shown in melanoma cell lines (28). EGFR signal transduction pathways have been shown to be part of the same transcriptional program as WNT5A, promoting resistance to vemurafenib in whole-transcriptome microarray analysis (31). In colorectal cancer (CRC), resistance to vemurafenib treatment of BRAF-mutant CRC is mediated by EGFR expression, which is a consequence of the epithelial lineage of CRC. Intrinsically activated EGFR signaling pathways are found to promote innate resistance to BRAF inhibitors in CRC (40, 41). It has been demonstrated that upon BRAF inhibition, EGFR signaling enables feedback re-activation of the MAPK pathway (42) and activation of the PI3K–AKT pathway (41). Increased EGFR transcription results following the demethylation of DNA enhancer elements in the EGFR promoter (39). While the EGFR signaling pathway is normally inactive in the melanocyte neural crest lineage (43), dedifferentiation in melanoma as a result of low MITF levels leads to increased activity of the EGFR signaling pathway (28). Indeed, EGFR-mediated resistance in melanoma is associated with profound alterations in gene expression patterns, which correlate with dedifferentiation, as defined by a reduction in MITF levels together with reduction in downstream targets of MITF and other important melanocytic regulators, such as SOX10, PAX3, and LEF1. Knockdown of MITF leads to the activation of an autocrine drug resistance loop that consists of increased secretion of EGF ligand and increases EGFR expression (28). Forced expression of MITF in melanoma inhibits EGFR signaling and increases sensitivity to BRAF and MEK inhibitors (28). Interestingly, increased expression of MITF in CRC cells was also found to increase sensitivity to BRAF inhibitors (28).

In one study, slow-cycling populations of melanoma cells with high EGFR expression were associated with low levels of the transcription factor SOX10, which is known to both upregulate, and interact with MITF to regulate melanocyte lineage development (7). Like low MITF, low expression of SOX10 promotes TGFβ signaling, which in turn activates numerous receptor tyrosine kinases, including EGFR, PDGFR-B, ERBB3, and AXL (7, 26). Cells expressing high EGFR and low SOX10 were resistant to BRAF and MEK inhibitors and exhibited slow cycling. However, upon treatment with MAPK inhibitors, cells with high EGFR and low SOX10 were selectively enriched, while cell division was increased. Interestingly, upon discontinuation of treatment with MAPK inhibitors, there was a depletion of cells with high EGFR and low SOX10. These data suggest that the high EGFR and SOX10 axis is only beneficial in the presence of MAPK inhibitors and without MAPK inhibitors, the cells return to a drug-sensitive state (7).

Combination treatment of melanoma cells with gefitinib, an EGFR inhibitor, and vemurafenib was unable to inhibit proliferation of the drug-resistant cells, indicating that other routes of drug resistance were also present (7). As RTKs, such as PDGFR-B and ERBB3, are expressed in melanoma cells with low SOX10, and given that many RTKs stimulate both MAPK and PI3K–AKT downstream pathways, the combined inhibition of MAPK and PI3K–AKT pathways using BRAF and PI3K inhibitors was found to restore inhibition of proliferation in the drug-resistant melanoma cells (7).

A range of other signal activators was found to be associated with a low MITF dedifferentiated state and treatment resistance in melanoma. The expression of AXL tyrosine receptor kinase expression was significantly increased by low MITF levels, and conferred both intrinsic and acquired resistance in melanoma cell lines (25, 26). Expression and activation of AXL, as determined by phosphorylation, as well as other RTKs was further increased upon exposure to vemurafenib. AXL was able to maintain MAPK pathway activity and activate the PI3K–AKT pathway when exposed to BRAF or MEK inhibition (25). The high AXL/low MITF expression profile was also observed in TCGA data, which includes transcriptional profiles for 356 patients, as well as in tumor samples obtained from patient-derived xenografts (26). Other RTKs that increased with high AXL included EGFR and PDGFR. It is thought that MITF suppresses these RTKs either indirectly or directly, and thus reduced MITF levels lead to the activation of these various RTKs (26). Inhibiting AXL did not result in significant tumor cell death in drug-resistant cell lines suggesting that, although AXL contributes to the drug-resistant phenotype, other RTK activity may also need to be cotargeted (25).

The activation of NFκB was shown to be an important player in promoting treatment resistance in MITF-low cells (25). In treatment-naïve melanoma cell lines, cells with high NFκB and low MITF expression were intrinsically resistant to both single-agent inhibition at multiple points in the MAPK pathway, and to combination BRAF/MEK inhibition (25). Furthermore, NFκB signaling was sufficient to promote a switch to a low MITF state that enabled resistance to MAPK pathway inhibition (25).

As previously discussed, the depletion of MITF mediates senescence via dedifferentiation in melanoma (21). However, MITF activity has been shown to be a central regulator of distinct cellular phenotypes, highlighting its complex role. For instance, MITF plays an essential role in cell division by suppressing the cell-cycle inhibitor p27^Kip1 (CDKN1B; refs. 20, 21, 44, 45). However, other studies have reported that high MITF levels suppress cell division by activating p16^Ink4a (CDKN2A; ref. 46) and p21^Cip1 (CDKN1A; refs. 46, 47). To explain the paradoxical function of MITF in regulating both proliferation and cell-cycle arrest, the rheostat model was proposed (21). This model proposes that different MITF levels drive distinct cellular phenotypes. Low levels of MITF induce dedifferentiation resulting in a senescence-like phenotype, intermediate MITF levels favor cell proliferation, and higher levels of MITF result in differentiation-mediated senescence (ref. 21; Fig. 1).

Increasing numbers of studies have expanded the rheostat model with therapeutic implications. High MITF expression and increased expression of its melanocyte-specific target genes were reported to be associated with treatment resistance to various single-agent MAPK pathway inhibitors, as well as to combined inhibition of BRAF and MEK (26, 48–52). Ectopic MITF overexpression was reported to promote resistance to BRAF, MEK, or ERK inhibitors (25, 48). Moreover, genomic amplification of MITF is found in 15.2% of metastatic melanoma patients and is associated with poorer prognosis and resistance to chemotherapeutic agents (44).

In contrast, MITF expression has also been shown to be necessary in BRAF-mutant melanomas to facilitate the response to MAPK pathway inhibition (25–27). BRAF-mutant melanoma cell lines were responsive to MAPK pathway inhibition only in those cell lines that exhibited MITF activity (25, 26). The association of MITF with both treatment resistance and susceptibility is reminiscent of the dual role of MITF in proliferation and differentiation-mediated senescence. According to the rheostat model, intermediate MITF levels regulate the proliferative phenotype and confer susceptibility to MAPK pathway inhibition, whereas higher MITF expression levels mediate differentiation-mediated senescence and confer drug resistance (Fig. 1).

Chromatin remodeling provides a major mechanism to control gene transcription. The chemical modification of chromatin opens up DNA regions, which are otherwise condensed, for access to transcription factors and RNA polymerase (53). Cancer cells are capable of epigenetically reprogramming their transcriptional processes through chromatin-remodeling factors to maintain an oncogenic phenotype (54). In melanoma, aberrant chromatin remodeling through histone demethylation has been shown to play an important role in switching to a slow-cycling state.

In melanoma cell lines and in tumors of patients with advanced melanoma, the chromatin-remodeling factor, JARID1B (KDM5B), was highly expressed in small subpopulations of cells (8, 55, 56). JARID1B is a member of the highly conserved family of jumonji/ARID1 and it functions to remove methyl groups from histone 3 lysine 4 (H3K4) residues, which in turn alters the chromatin structure and transcriptional expression patterns. JARID1B-high cells exhibit slow-cycling properties, with a cell division doubling rate of more than 4 weeks (13). Moreover, JARID1B-high cells lack protein expression of proliferation marker Ki-67 (13) and have increased activity of retinoblastoma protein, which downregulates E2F target genes and induces cellular senescence (57).

Slow-cycling JARID1B-positive cells display intrinsic resistance to various chemotherapeutic drugs (cisplatin, bortezomib, temozolomide), and to vemurafenib (8). This resistance mechanism results in enrichment of JARID1B-positive cells upon vemurafenib exposure in melanoma cell lines and in in vivo human melanoma xenografts. Moreover, melanoma biopsies obtained before and after relapse from the same patients, who were treated with vemurafenib, displayed increased JARID1B expression post-treatment (8). Further investigations revealed that the primary mechanism of enrichment of JARID1B-high cells following drug exposure was through the selection of pre-existing JARID1B-positive slow-cycling cells. In addition, induction via an adaptive mechanism also occurred, although to a lesser degree (8). Pathway analysis revealed increased PI3K–AKT pathway signaling in JARID1B-high cells, which supports the notion that slow-cycling cells have altered their dependency on growth-stimulatory pathways to allow them to escape programmed cell death or senescence when the MAPK pathway is inhibited (8). A quiescent cellular state with drug resistance has also been associated with histone demethylase JARID1A (KDM5A) in numerous cancer types, including melanoma (58). This is associated with increased IGFR1 signaling, which is again indicative of activation of an alternative growth-stimulatory pathway, and escape from therapeutic stress resulting from MAPK pathway inhibition.

As alluded to above, adaptive mechanisms may occur during induction of drug resistance in contrast to the enrichment of pre-existing drug-resistant cells in an initial cell population (8). For example, in one study where vemurafenib-induced drug-tolerant cells (IDTC) were generated, global chromatin modifiers were also induced, including upregulation of JARID1A, JARID1B, and the activity of various other histone demethylases (12). In the IDTCs, the transcription of genes involved in cell proliferation and DNA replication was reduced, as well as cell growth rates, which is characteristic of a quiescent phenotypic state. Dedifferentiation was demonstrated by the increased expression of putative melanoma stem cell markers, such as CD271, and the loss of melanocyte-specific antigens including tyrosinase and melan-A (12). Moreover, multiple signaling pathways were upregulated including the MAPK pathway, the PI3K–AKT pathway, and the mTOR pathway.

Cancer cell proliferation is associated with a shift in cell metabolism toward glycolysis, a phenomenon commonly referred to as the Warburg effect (59, 60). This process is also termed aerobic glycolysis given that glycolysis is maintained, despite sufficient oxygen supply and an intact mitochondrial respiration system (61, 62). Glycolysis permits fast production of ATP and the molecular building blocks required to support rapid cell division and biomass accumulation (61). Indeed, BRAF-mutant malignant melanomas are dependent on glycolysis to permit continuous cell division, as demonstrated by reduced oxygen consumption rate, and increased levels of lactate and glycolytic enzymes (63, 64).

Recently, there has been a growing body of evidence supporting the notion that cancers use multiple strategies of metabolism (59). For instance, oxidative phosphorylation underlies a subpopulation of slow-cycling cells in melanoma (8). Subpopulations of slow-cycling cells with high levels of JARID1B chromatin-remodeling factor were shown to feature a high level of oxidative phosphorylation. Proteomic profiling revealed the JARID1B-overexpressing cells had elevated expression of mitochondrial enzymes of electron transport chain complexes, including ATP-synthase and NADPH dehydrogenase (complex I). Accordingly, the JARID1B-high cells also had higher mitochondrial energy production, consumed more oxygen, and generated more peroxide compared with control cells (8).

Increasing evidence suggests that oxidative phosphorylation mediates treatment resistance in various cancers including melanoma (65). For example, treatment of BRAF-mutant melanoma cell lines with vemurafenib enriched for resistant cells that had increased oxidative phosphorylation (8, 66, 67). Oxidative phosphorylation was demonstrated with upregulation of numerous mitochondrial proteins, increased oxygen consumption rates (8, 67), increased mitochondria density (66), and reduced levels of lactate (66, 67). Consistently, BRAF-mutant melanoma samples excised from patients who have developed treatment resistance were found to exhibit increased expression of oxidative phosphorylation gene expression (66). Overall, this suggests that metabolic heterogeneity in melanoma may underlie different cellular phenotypes as well as distinct responses to therapeutic drugs.

While proliferative melanoma cells utilizing glycolytic pathways are susceptible to therapeutic drugs, oxidative phosphorylation in a subpopulation of slow-cycling melanomas allows the energy usage to be diverted from cell division toward more efficient energy production, favoring cell survival (8, 65, 66). Consequently, slow-cycling cells with an enhanced potential for oxidative phosphorylation, and suppression of proliferative potential, are selected upon therapeutic stress. Slow-cycling cancer cells that utilize oxidative phosphorylation for drug resistance have also been observed in pancreatic tumors (68, 69), glioblastoma (70), and acute myelogenous leukemia (71). In pancreatic tumors, subpopulations of slow-cycling tumor cells with cancer stem cell–like features were identified that depended on mitochondrial respiration to survive the toxic effects of therapeutic drugs, such as doxycycline (68). These drug-resistant cells characteristically expressed cancer stem cell markers, and displayed increased mitochondrial respiration (68). Drug-resistant tumor cells can remain dormant for months in a slow-cycling state until tumor relapse.

Like JARID1B, PGC1α has a role in oxidative phosphorylation and resistance to therapy. PGC1α is one of the key drivers mediating oxidative phosphorylation and treatment resistance in melanoma (66, 72). PGC1α is a transcriptional coactivator known for its role in the regulation of mitochondrial biogenesis. It has a central role in upregulating oxidative phosphorylation in various tissue types (73). BRAF-mutant melanoma cell lines treated with vemurafenib exhibit significantly increased levels of PGC1α-mediated oxidative phosphorylation. In addition, an upregulated gene expression program of mitochondrial oxidative phosphorylation, an increased oxygen consumption rate, and a reduction of lactate secretion, glycolysis, and glucose uptake are observed in vemurafenib-treated melanoma cells (66, 72). PGC1α also has a role in the detoxification of reactive oxygen species (72).

Even though JARD1B and PGC1α both govern oxidative phosphorylation, these two factors do not cooperate in the same pathway, and current data suggest they represent two distinct mechanistic pathways. More than one mechanism controls oxidative metabolism during treatment resistance in melanoma. Melanoma cell lines treated with vemurafenib upregulate oxidative metabolism regardless of PGC1α expression (67). In addition, JARID1B and PGC1α appear to represent two distinct states of differentiation (Fig. 1). JARID1B is reported to be a putative stem cell marker with roles in stem cell biology and in inducing pluropotency (74, 75). High JARID1B–expressing cells were shown to have increased tumorigenic potential with increased ability to sustain continuous tumor growth (13). Throughout serial xenograft transplantation where human melanoma is grown through serial passage in mice, JARID1B-negative cells eventually reach a replicative ceiling and become exhausted, whereas JARID1B-positive cells can be continuously passaged indefinitely. In contrast, in a subpopulation of melanoma cells, a high MITF melanocyte-specific differentiation program drives PGC1α-mediated oxidative phosphorylation, as well as the downstream targets of MITF, such as TYR, MLANA, and DCT (66). MITF was found to directly stimuate PGC1α expression by binding to its promoter (66, 72).

Collectively, the JARID1B and MITF–PGC1α axes could regulate oxidative phosphorylation and drug resistance mechanisms at different ends of the rheostat model. Slow cycling occurs with either low MITF (dedifferentiation) or with high MITF (differentiation). JARID1B mediates oxidative phosphorylation in slow-cycling cells that are dedifferentiated. In contrast, differentiated slow-cycling cells utilize the MITF–PGC1α axis to mediate oxidative phosphorylation and treatment resistance (Fig. 1). Interestingly, the MITF–PGC1α axis correlates with a proliferation expression signature. However, the forced expression of PGC1α did not lead to enhanced cell division, indicating that it does not regulate proliferation (66).

Oxidative phosphorylation serves as a valuable therapeutic target, as slow-cycling cells depend on it for metabolic energy. Mitochondrial inhibitors including oligomycin A, an ATP synthase inhibitor, and other mitochondrial inhibitors were able to suppress the enrichment of a JARID1B-positive subpopulation (8). Tumor cell death was also enhanced in melanoma cells that utilized a MITF–PGC1α axis–mediated treatment resistance pathway, when treated with mitochondrial inhibitors in combination with chemotherapeutic agents, or with vemurafenib (66). However, drugs that inhibit the mitochondrial complexes typically have a narrow therapeutic index, which means that frequently they are too toxic to be used in the clinic. The pro-oxidative drug, elesclomol, has a wider therapeutic index, and demonstrates antitumor activity in vemurafenib-resistant melanoma cells in vitro and in vivo (67). Elesclomol induces apoptosis by increasing oxidative stress (76–78); apoptosis is induced by elesclomol in slow-cycling cells with high JARID1B, preventing the enrichment of treatment-resistant JARID1B-positive cells (79). Cancerous cells with oxidative phosphorylation already have increased levels of endogenous reactive oxygen species (8). Therefore, this generates increased dependence on anti-oxidation for survival, making slow-cycling melanoma cells more sensitive to pro-oxidative drugs compared with normal cells. Elevated levels of reactive oxygen species selectively elicit cell death in tumors with high levels of oxidative phosphorylation. These data demonstrate that multidrug resistance in slow-cycling cells can be overcome by combining elesclomol with therapeutic drugs such as the BRAF inhibitor, vemurafenib (79). Elesclomol has shown clinical efficacy in a phase III trial in a subgroup of metastatic melanoma patients with normal baseline lactate dehydrogenase (LDH; ref. 80). Normal LDH levels reflect slow-cycling cancerous cells that depend on oxidative phosphorylation, and which are susceptible to elesclomol. In contrast, high LDH levels are associated with rapidly proliferating tumor cells, which are largely dependent on glycolysis, have low levels of reactive oxygen species, and are more resistant to elesclomol (79, 81). Therefore, combining targeted therapy with drugs that target oxidative phosphorylation may be a promising therapeutic strategy.

Melanoma cells that are slow cycling are highly resistant to conventional chemotherapy and to current targeted therapy regimes. While keeping cancerous cells in a senescent nonproliferative state has been suggested as an efficient therapeutic option, increasing evidence suggests that these cells could be wrongly interpreted as being senescent, but rather are slow cycling, as they can re-gain proliferative characteristics. Slow-cycling cells can rapidly acquire enhanced levels of proliferation leading to cancer relapse. The characterization of the slow-cycling phenotypes may therefore lead to the identification of vulnerable characteristics of tumor cells that may serve as therapeutic targets to overcome treatment resistance and ultimately improve melanoma therapy.

No potential conflicts of interest were disclosed.

The authors acknowledge financial support from the University of Otago (to A. Ahn), and the New Zealand Institute for Cancer Research Trust (to A. Chatterjee, M.R. Eccles), and the Maurice Wilkins Centre for Molecular Biodiscovery (to A. Chatterjee, M.R. Eccles).