Nongenetic Mechanisms of Drug Resistance in Melanoma

Resistance to targeted and immune-based therapies limits cures in patients with metastatic melanoma. A growing number of reports have identified nongenetic primary resistance mechanisms including intrinsic microenvironment- and lineage plasticity–mediated processes serving critical functions in the persistence of disease throughout therapy. There is a temporally shifting spectrum of cellular identities fluidly occupied by therapy-persisting melanoma cells responsible for driving therapeutic resistance and metastasis. The key epigenetic, metabolic, and phenotypic reprogramming events requisite for the manifestation and maintenance of so-called persister melanoma populations remain poorly understood and underscore the need to comprehensively investigate actionable vulnerabilities. Here we attempt to integrate the field’s observations on nongenetic mechanisms of drug resistance in melanoma. We postulate that the future design of therapeutic strategies specifically addressing therapy-persisting subpopulations of melanoma will improve the curative potential of therapy for patients with metastatic disease.


INTRODUCTION
Stage IV (metastatic) melanoma is the deadliest skin malignancy, with a five-year survival rate of 22.5% (Am. Cancer Soc. 2019). Approximately 192,310 new cases of melanoma will be diagnosed and an estimated 7,230 melanoma patients will die in the United States in 2019. Although the incidence of melanoma continues to rise disproportionally relative to other cancers, mortality has steadily decreased due to major advancements in targeted and immune-based therapeutic modalities, with 13 new FDA (US Food and Drug Administration)-approved therapies since 2011 , Postow et al. 2015, Queirolo et al. 2015, Sharma et al. 2011. The BRAF inhibitor (BRAFi) and MEK inhibitor (MEKi) combination for patients whose melanomas harbor activating BRAF V600E/K mutations possesses a ∼76% response rate and represents the best targeted therapy strategy in the field (Figure 1) (Dummer et al. 2018. Immune checkpoint inhibition with anti-CTLA-4 or anti-PD-1 blockade has a response rate of ∼20% or ∼40%, respectively, is effective independent of tumor genotype (i.e., BRAF mutation or NRAS mutation), and represents the best immunotherapy strategy in the field (Carlino et al. 2018, Schachter et al. 2017. Despite the encouraging efficacy of targeted and immune-based therapies, most patients (∼80%) with advanced melanoma do not experience cures. The shortfall of long-term therapeutic efficacy is largely attributed to the manifestation of resistance to targeted therapy and the predominate lack of response (due to innate resistance or toxicity) to immunotherapy (Mehta et al. 2018;Rebecca et al. 2014a,b,c;Zaretsky et al. 2016).
The emerging consensus across the melanoma field accepts that resistance to therapy arises via two hallmark mechanisms: (a) genetic mechanisms of acquired resistance driven by secondary mutations that render melanoma cells insensitive to therapy (Botton et al. 2013;Teh et al. 2018;Vido et al. 2018;Villanueva et al. 2010Villanueva et al. , 2013 and (b) nongenetic mechanisms of innate or early adaptive resistance that precipitate the acquisition of secondary mutations (Fischer et al. 2019, Ma et al. 2014, Nazarian et al. 2010, Ojha et al. 2019, Rebecca et al. 2017, Smith et al. 2016, Sun et al. 2014. Genetic mechanisms of resistance have been extensively reviewed in the literature and are not discussed here; however, for reference we created a brief synopsis in Supplemental Figure 1. In contrast, this review looks at cases where no clear genetic cause for resistance is detectable, raising the possibility of nongenetic mechanisms of resistance. Melanoma cells display tremendous intrinsic transcriptional, translational, and posttranslational plasticity, which underlies melanoma's phenotypic versatility and heterogeneity in therapy sensitivity and researchers' inability to eliminate all tumor cells preclinically and clinically (Falletta et al. 2017;Han et al. 2018;Roesch et al. 2010Roesch et al. , 2013Saei & Eichhorn 2018;Shaffer et al. 2017).
Within the melanoma cell population there is a diverse array of dynamic cell states in constant flux depending on stochastic and deterministic variables (Long et al. 2019, Rambow et al. 2018, Tirosh et al. 2016. Cell-to-cell physical and paracrine interactions with adjacent melanoma cells, as well as other cellular compartments (i.e., fibroblasts, endothelial cells, T cells, and B cells) in the surrounding tumor microenvironment (TME) constitute a complex microecosystem that further stimulates a fluid spectrum of cellular identities transiently inhabited by melanoma cells attempting to survive a temporally shifting multitude of environmental stresses (i.e., lack of oxygen supply, starvation due to insufficient nutrient availability, and therapy stress) (Almeida et al. 2019, Atay et al. 2019, Flach et al. 2011, Moellering et al. 2008. Accumulating evidence suggests the existence of multiple melanoma cell states with varying levels of differentiation/dedifferentiation, cell cycle activity, metabolic activity, mobility, invasiveness, metastatic potential, and sensitivity to therapy (Perego et al. 2018;Rambow et al. 2018;Roesch et al. 2010Roesch et al. , 2013Shaffer et al. 2017;Tirosh et al. 2016;Zhang et al. 2016). Multiple models attempt to precisely define the biological mechanisms underlying the manifestation and maintenance of these cell states, including the microphthalmia-associated transcription factor (MITF)-rheostat proliferative/invasive phenotype-switching model (Hoek et al. 2006(Hoek et al. , 2008, the interferon gamma (IFN-γ) proinflammatory cytokine model (Bai et al. 2019, Cho et al. 2011, Ribas & Wolchok 2018, and the cancer stem cell (CSC) model (Schatton et al. 2008;Zabierowski & Herlyn 2008a,b). Interestingly, the epithelial-mesenchymal transition (EMT) cassette in other advanced cancers (i.e., breast and pancreatic) largely overlaps in protein markers and transcriptional programs observed in melanoma cell states implicated in therapy resistance and metastasis (Caramel et al. 2013, Perotti et al. 2019. Insights gained in melanoma may translate to other cancer types and vice versa. A common set of mechanistic threads likely exists amid the multitude of reports on plasticity and adaptive therapeutic resistance across the melanoma field; however, a clear and effective strategy to eliminate these persisting minor populations of melanoma cells following targeted and immune therapy remains to be found in most patients. In this review, we survey the latest findings in mechanisms of primary resistance to therapy, with a special emphasis on melanoma lineage plasticity and the molecular properties/drivers thereof.

INTRINSIC RESISTANCE
The relatively short progression-free survival (PFS) experienced by most patients treated with either targeted therapy [the 3-year PFS of combinatorial dabrafenib and trametinib therapy is 23% (Schadendorf et al. 2017) and the 5-year PFS is 13%  or immunotherapy [pembrolizumab 5-year PFS is 21% (Hamid et al. 2019)] suggests that the acquisition of resistance occurs in most patients (Figure 1). Although activating BRAF V600E/K mutations predict sensitivity to BRAF inhibition and BRAF/MEK inhibition, a large number of melanoma cell lines and patient-derived xenograft (PDX) models display intrinsic therapy resistance (Ndoye et al. 2017, Paraiso et al. 2011, Roesch et al. 2016, Song et al. 2017). In addition, approximately 24% of patients whose melanomas harbor BRAF V600E/K mutations do not meet the RECIST (Response Evaluation Criteria in Solid Tumors) threshold for a response to BRAF/MEK inhibition . Intrinsic resistance to targeted therapy occurs through a large number of nongenetic mechanisms including the hyperactivation of compensatory signaling pathways [i.e., FAK (Hirata et al. 2015), SRC (Girotti et al. 2013), and STAT3 (Vultur et al. 2014  extensive intratumoral metabolic heterogeneity exists whereby differing dependencies on aerobic glycolysis versus oxidative phosphorylation can be observed among melanoma subpopulations. Exploiting this metabolic vulnerability with the use of mitochondrial HSP90 inhibitors could increase therapeutic efficacy by targeting metabolic heterogeneity (Zhang et al. 2016). Although the mechanisms described are diverse, resistance to BRAF/MEK inhibition primarily channels through the hyperactivation of parallel PI3K/mTOR signaling (Caporali et al. 2016, Deng et al. 2012, Sweetlove et al. 2015 or the reactivation of the mitogen-activated protein kinase (MAPK) pathway (Ojha et al. 2019, Welsh et al. 2016. Unfortunately, the effective use of existing PI3K and mTOR inhibitors in combination with BRAF/MEK inhibition remains to be seen due potential overlapping toxicities (Lemech et al. 2013, Welsh et al. 2016.
Further contributing to the difficulty of clinically addressing intrinsic resistance to therapy is the immense intratumoral transcriptomic and proteomic heterogeneity observed within a b c High Low

MAPK activity
High Low

MAPK activity
High Low

PI3K/mTOR activity
High Low SRC activity Cell 1

Cell 2
Low activity for all pathways

Figure 3
Single-cell signaling heterogeneity. (a) Two neighboring melanoma cells can possess different activity kinetics of the MAPK pathway at any given moment in time. (b) These different signaling kinetics likely occur across all major signaling pathways, producing (c) heterogeneous tumors containing melanoma with a spectrum of signaling pathway intensities shifting over time. melanoma patients' tumor samples , Tirosh et al. 2016). Melanoma subpopulations with heterogeneous cell cycle and therapy resistance transcriptional programs can be observed within the same melanoma patient's tumor and vary from patient to patient (Tirosh et al. 2016). At the single-cell level, the activity and expression of proteins in a given signaling pathway oscillate dynamically in response to idiosyncratic intracellular and extracellular cues around a preset homeostatic level unique to each cell's epigenetic landscape, gene regulatory network, and intratumoral localization (Figure 3a) (Buszczak et al. 2014, Huang 2013. For example, intracellular signaling kinetics of the MAPK pathway within individual melanoma cells in vivo are highly heterogeneous and influenced by spatial dynamics [i.e., oxygen/nutrient availability and proximity to elements of the TME (Hirata et al. 2015)]. Class switching between RAF isoforms has also been observed in melanoma cells and likely occurs to varying extents within a population of melanoma cells (Dumaz et al. 2006).
The activity of every relevant signaling pathway within melanoma cells is heterogeneous, with transiently distinct cell populations possessing, for example, high activation of MAPK but low PI3K/AKT and SRC pathway activity (Figure 3b) (Buszczak et al. 2014). In contrast, there are likely melanoma cell subpopulations with high activation or low activation of all three pathways. These different, temporally shifting signaling kinetics result in the complete gamut of melanoma cellular states, including a spectrum of metabolic (anabolic and catabolic) activity, cell cycle progression rate, metastatic potential, immune mimicry, and sensitivity to therapy (Figure 3c). The reality of this plasticity should trigger caution in the interpretation of static data points and may underlie the difficulty in the development of biomarkers that can reliably predict whether a patient will respond to a given therapeutic, which is highly desired for targeted and immune therapy-based modalities.

MICROENVIRONMENT-INDUCED RESISTANCE
Outside of what is possible within a population of melanoma cells lies an ever more complex interactome with other cell compartments and aspects of the local and distant TME (Almeida et al. 2019). The TME has long been recognized to exert homeostatic control over melanoma cells and contribute to their escape from therapy and metastatic potential (Villanueva & Herlyn 2008). In 1889 Paget first postulated the so-called seed and soil theory that some in vivo microenvironments were more receptive to the establishment of tumor metastases than others (Fidler 2003). A tour de force study tested the potential of the microenvironment to confer resistance to therapy in melanoma cells. In all, the study leveraged 23 stromal cell types and 45 cancer cell lines alongside 35 anticancer drugs and concluded that stromal-mediated drug resistance is common, especially to targeted therapy (Straussman et al. 2012). Stromal secretion of hepatocyte growth factor (HGF) was found to play the most significant role in conferring melanoma resistance to BRAF inhibition through activation of the cognate receptor c-MET. Melanoma cells dynamically interact with and reprogram adjacent host cells in the TME via the paracrine secretion of bFGF (basic fibroblast growth factor) (Berking et al. 2001a;Shih & Herlyn 1993, PDGF (platelet-derived growth factor) (Anderberg et al. 2009, Elias et al. 2010, Willenberg et al. 2012, and TGF-β (transforming growth factor beta) (Li et al. 2014, Menter et al. 1995, which activate stroma to a cancerassociated fibroblast phenotype and promote endothelial cell-dependent angiogenesis. Melanoma cells dynamically communicate with stromal cells through the secretion of PDGF, which stimulates the stromal secretion of IGF-1 capable of increasing melanoma proliferation and therapy resistance (Figure 4a) (Ruiter et al. 2002). Stromal cells can also increase melanoma aggressiveness through the secretion of endothelins and bFGF (Ruiter et al. 2002). Melanoma-derived TGF-β also promotes the deposition of the extracellular matrix (ECM; collagen and tenascin),  (a) Melanoma-mediated resistance mechanisms to BRAFi. Melanoma cells remodel local and distant microenvironments through the secretion of TGF-β, FGF-2, and VEGF, which activate fibroblasts to a cancer-associated fibroblast state, promote the secretion of tumor-promoting IGF-1 by B cells, and stimulate de novo vasculature formation by endothelial cells, respectively. Melanoma cells also secrete several growth factors that promote autocrine activation of key survival pathways. (b) There also exists therapy-mediated reprogramming mechanisms of the tumor microenvironment to a protumorigenic state. BRAFi activate the MAPK pathway in macrophages, triggering the production of VEGF to promote melanoma resistance, and in CD4 + and CD8 + T cells, which was correlated with elevated T cell activation and T cell antitumor activity. BRAFi also activate melanoma-associated fibroblasts and promote ECM production and remodeling, leading to therapy escape. Abbreviations: BRAFi, BRAF inhibitors; ECM, extracellular matrix.
immunosuppression, and phenotype plasticity (Berking et al. 2001b). Through integrin β1, melanoma cells can dynamically interact with the ECM in the TME, triggering downstream FAK and SRC survival signaling (Colo et al. 2012, Frame & Serrels 2015, Hirata et al. 2015. In addition, the secretion by melanoma cells of VEGF (vascular endothelial growth factor) and PlGF (placental growth factor) in response to environmental stresses including hypoxia, acidic pH, and hypoglycemia triggers the formation of de novo vasculature to restore optimal oxygen, pH, and glucose levels (Levati et al. 2011, Pagani et al. 2016. Cells in the immune compartment also play a role in the escape of melanoma cells from therapy. Recently, human tumor-infiltrating B cells were observed to contribute to acquired drug resistance through a paracrine mechanism (Figure 4a) (Somasundaram et al. 2017). The paracrine secretion of FGF-2 (fibroblast growth factor 2) from melanoma cells activated tumor-infiltrating B cells and induced their secretion of IGF-1. B cell-derived IGF-1 was critical in promoting heterogeneous subpopulations of melanoma cells with hyperactivation of FGFR-3. A pilot trial utilizing a CD20 antibody to deplete B cells showed encouraging antitumor activity, highlighting the potential to overcome nongenetic, microenvironment-mediated mechanisms of therapy resistance.
In addition to the ability of melanoma cells to activate surrounding host cells to a protumorigenic secretory phenotype, targeted therapy itself has been demonstrated to reprogram host cells to contribute to melanoma therapy resistance (Figure 4b). BRAF inhibition paradoxically activates the MAPK pathway in macrophages, triggering production of VEGF to promote melanoma resistance (Wang et al. 2015a,b). BRAF inhibition also activates melanoma-associated fibroblasts and promotes their production and remodeling of ECM, leading to increased integrin β1/FAK/SRC signaling and therapy escape in melanoma cells (Hirata et al. 2015). Notably, BRAF inhibition has also been observed to activate MAPK activity in CD4 + and CD8 + T cells, which was correlated with elevated T cell activation and increased T cell antitumor activity (Callahan et al. 2014), underscoring the complexity of this system and the need to consider all potential consequences a given therapy can have on the tumor and host cells.
Interestingly, in addition to the ability of melanoma cells and therapy to reprogram tumor host cells to a protumorigenic state, the age of the patient has also been observed to be a contributing factor for the aggressiveness of melanoma. Stromal fibroblasts from aged patients (older than 55 years of age) confer significant aggressiveness and resistance to therapy to melanoma cells relative to fibroblasts from younger patients (less than 35 years of age) (Kaur et al. 2016). Melanoma cells xenografted into aged or young mice demonstrated that an aged microenvironment can increase the metastatic potential of melanoma to the lungs and decrease the sensitivity to targeted therapy. These results reflect clinical observations whereby older patients experience lower response rates to targeted therapy relative to their younger counterparts. Interestingly, aged patients respond better to immunotherapy than younger patients, underscoring the need to comprehensively understand the role the TME serves in response to therapy (Kugel et al. 2018).

MELANOMA LINEAGE PLASTICITY-MEDIATED RESISTANCE
Treatment with combination BRAF/MEK inhibition eliminates the bulk of tumor cells with the exception of rare subpopulations of melanoma cells that persist, metastasize, and ultimately regrow with acquired resistance (Rambow et al. 2018, Shaffer et al. 2017). This resistance is driven in part by the ability of melanoma cells to adjust their phenotype by adopting a MAPK-independent lineage identity to survive and thrive. Perhaps the most widely accepted model of lineage plasticity in melanoma rests upon the expression of MITF, a master regulator of melanocyte function and differentiation from neural crest stem cell (NCSC) progenitor cells (Hemesath et al. 1998, Hoek & Goding 2010, Yasumoto et al. 1994  The relationship between MITF, melanoma proliferation/invasion, and therapy resistance. (a) Overlaid expression of MITF and proliferative capacity from literature findings that MITF levels were directly correlated with melanoma proliferation and inversely correlated with invasiveness/motility in vitro and proliferation in vivo (Carreira et al. 2006;Hoek et al. 2006Hoek et al. , 2008Kaur et al. 2016;Vervaillie et al. 2016 to stratify melanoma cells into two phenotypes: (a) a MITF high proliferative (noninvasive) phenotype and (b) a MITF low invasive (nonproliferative) phenotype. This model is based on seminal papers in 2006 and 2008 by several independent laboratories including our own reporting that MITF levels were correlated with the proliferative and invasive capacity of melanoma cells in culture, as well as with the proliferative capacity of melanoma cells in xenograft mouse models (Figure 5a) (Carreira et al. 2006;Hoek et al. 2006Hoek et al. , 2008. However, melanoma cells expressing inducibly regulated short hairpin RNAs specific for MITF do not robustly exhibit reduction in proliferation or an appreciable change in stem cell markers upon downregulation of MITF (Vlckova et al. 2018), suggesting that baseline MITF expression may serve as a biomarker and not a functionally important molecule for this phenotype switch. Several additional genes/proteins have been reported to correlate with MITF levels and participate in the execution of the two phenotypes, including the lysosomal acid ceramidase ASAH1 (Leclerc et al. 2019), the EMT inducer ZEB1 transcription factor (Caramel et al. 2013, Richard et al. 2016, the RTK AXL (Konieczkowski et al. 2014, Muller et al. 2014, the SRY (sex-determining region Y) box 10 (SOX10) gene (Sun et al. 2014), and BRAF-mediated regulation of MITF via a PAX3/BRN2 rheostat (Smith et al. 2019).
The potential therapeutic significance of the MITF-rheostat model lies in observations that therapy-naïve MITF high melanoma cells are initially sensitive to therapy. It is postulated that MITF high tumors transition to an MITF low invasive state in the context of targeted therapy, hypoxia, and the aged TME (Figure 5b). Recent reports have observed that MITF expression and transcriptional signatures increase in vivo following BRAF/MEK inhibition (Rambow et al. 2018, Smith et al. 2016, suggesting a complex role of MITF in therapy sensitivity (Figure 5c). Careful examination of tumor tissues from paired pretreatment, on-treatment, and relapse has not been extensively done due to poor availability. However, there is evidence of MITF high and MITF low subpopulations in nearly all tumors (regardless of resistance state) analyzed by single-cell resolution techniques such as scRNA-seq (single-cell RNA sequencing) and RNA FISH (fluorescence in situ hybridization), which suggests that the potential role of MITF in vivo is immensely complex and heterogeneous (Rambow et al. 2018, Tirosh et al. 2016. Furthermore, although the phenotype-switching model has been demonstrated at the tumor population level, it remains to be seen whether an individual proliferative, noninvasive melanoma cell initially expressing high MITF levels can decrease its MITF levels and transition to a nonproliferative and highly invasive state (Figure 5d). Advances in lineage analysis techniques with the use of unique DNA barcodes may provide opportunities to determine whether individual cells can undergo a phenotype switch and whether MITF is the master regulator.

NEURAL CREST STEM CELL-LIKE MELANOMA CELLS
Lineage plasticity is emerging as a critical mechanism leveraged by persisting melanoma cells to withstand therapy and metastasis in vivo. The best described lineage transition event occurring in melanoma is the dedifferentiation to a NCSC-like state. Aggressive subpopulations of melanoma that persist after therapy exhibit common features, including (a) their existence prior to therapy, (b) a slow-cycling state of relative dormancy, (c) high metastatic potential, and (d) stem-like molecular and biological properties akin to NCSCs. NCSC-like melanoma cells express NCSC markers [i.e., the H3K4 demethylase JARID1B (Roesch et al. 2010), NGFR (Fallahi-Sichani et al. 2017, EGFR (Shaffer et al. 2017), or AXL (Konieczkowski et al. 2014)] and exhibit NCSC behaviors including high invasiveness, plasticity, and self-renewal capacity (Figure 6a-c). It was initially unclear in the melanoma field whether melanomas follow the CSC model whereby melanoma stem cells function at the top of a tumor differentiation pyramid (Reya et al. 2001), as seen in acute myeloid leukemia (Bonnet & Dick 1997). However, observations that JARID1B and NGFR do not follow a hierarchical CSC model because JARID1B + and NGFR + cells can give rise to JARID1B − and NGFR − cells, respectively, suggest that melanoma lineage plasticity is very dynamic.
The heterogeneity and molecular drivers among NCSC-like melanoma cells are poorly understood. In a report focused on slow-cycling NCSC-like subpopulations that leveraged a label retention method, the secreted protein SerpinE2 was identified as conferring the invasive potential of JARID1B + and JARID1B − slow-cycling melanoma cells (Perego et al. 2018). Although all JARID1B + cells are slow cycling, not all slow-cycling cells are JARID1B + , suggesting that this persisting subpopulation encompasses extensive heterogeneity. Notably, no correlation between MITF and SerpinE2 was observed, demonstrating that MITF may not completely govern melanoma plasticity. Accordingly, a recent report delineated multiple subpopulations of melanoma that persist in minimal residual disease following the tumor debulking phase that occurs in most patients treated with BRAFi/MEKi (Rambow et al. 2018). Here, NCSC-like melanoma cells were again identified, in addition to and distinct from an invasive population, a starved-like melanoma cell population, and a pigmented population. Although four distinct subpopulations were identified by transcriptional signatures, it is unclear whether these four subpopulations represent four clonal populations, one potential clone capable of transiently inhabiting one of four transcriptionally defined states, numerous (greater than four) clones distributed across the four different states, or some combination thereof (Figure 6b). Notably, the MITF signature, predicted to decrease following therapy as cells engage the invasive phenotype, was unexpectedly observed to increase across the Hoek et al. (2006), Verfaillie et al. (2015), and Rambow et al. (2018) MITF signatures following BRAF/MEK inhibition in vivo in PDX models. An explanation behind this conflicting observation may be due to an unappreciated nuance of MITF biology that cannot be addressed by bulk RNA or Western blotting: MITF requires nuclear localization to trigger its transcriptional activity (Fang & Setaluri 1999). The majority of MITF characterization in melanoma cells is based on levels of MITF transcript or protein expression detected in the bulk population of melanoma cells. It is conceivable that there are melanoma cells with high levels of intracellular MITF but low levels of nuclear MITF. The opposite may also be true, where cells have relatively low levels of intracellular MITF but a high level of MITF in the nucleus actively promoting MITF transcriptional programs. In their recent conflicting report, Rambow et al. (2018) performed immunohistochemistry and observed that although the absolute staining of MITF is reduced following BRAF/MEK inhibition, there is an observable increase in cells with high nuclear expression of MITF, which correlates with the increased MITF transcriptional signature detected in melanoma cells treated with BRAFi/MEKi. MITF high versus MITF low perhaps should be carefully redefined as MITF high-nuclear expression versus MITF low-nuclear expression . MITF high and MITF low subpopulations are both clearly present in minimal residual disease following BRAF/MEK inhibition, demonstrating the need to broaden our understanding of what MITF-independent molecules and processes govern these highly plastic subpopulations so that improved therapeutic modalities can be developed to best address resistance.

FUTURE PERSPECTIVES
Although much progress has been made in the development of novel therapeutic modalities that dramatically increase overall survival relative to what was possible just a decade ago by chemotherapy, it is critical that the field address the complex therapeutic resistance mechanisms causing most patients to succumb to disease. It is becoming increasingly clear that the dynamic adaptability of melanoma cells to therapy is a major impediment for a cure. Host cells provide protective sanctuaries to escaping melanoma cells via melanoma-dependent TME reprogramming, dysfunctional secretory phenotypes in aged patients, and therapy-induced reprogramming of the microenvironment to a protumorigenic state. In addition, melanoma cells dynamically transition to different therapy-resistant cellular states allowing access to phenotypic plasticity-resembling melanocyte progenitor cells. Understanding the vulnerabilities of melanoma cells in the context of therapy, the complex microenvironment, and different cellular states will allow for the development of new therapeutic strategies that lead to the elimination of minimal residual disease responsible for driving therapy relapse. The development of improved technologies to more accurately assess the patient tumor landscape and its potential susceptibility to one therapeutic strategy versus another will also be key to the true personalization of medicine and potential curability of therapy tailored to the unique intratumoral heterogeneity of each patient.