Melanocytic nevi and melanoma: unraveling a complex relationship

Abstract

Approximately 33% of melanomas are derived directly from benign, melanocytic nevi. Despite this, the vast majority of melanocytic nevi, which typically form as a result of BRAF^V600E-activating mutations, will never progress to melanoma. Herein, we synthesize basic scientific insights and data from mouse models with common observations from clinical practice to comprehensively review melanocytic nevus biology. In particular, we focus on the mechanisms by which growth arrest is established after BRAF^V600E mutation. Means by which growth arrest can be overcome and how melanocytic nevi relate to melanoma are also considered. Finally, we present a new conceptual paradigm for understanding the growth arrest of melanocytic nevi in vivo termed stable clonal expansion. This review builds upon the canonical hypothesis of oncogene-induced senescence in growth arrest and tumor suppression in melanocytic nevi and melanoma.

INTRODUCTION

Growth arrest after activation of individual oncogenes can prevent cancer formation. Melanocytic nevi are neoplasms resulting from the proliferation of melanocytes, the normal pigment-producing cells in the skin. Nevi are growth arrested, clonal neoplasms of melanocytes initiated by well-defined oncogenic mutations in the mitogen-activated protein kinase (MAPK) pathway, most commonly by BRAF^V600E-activating mutation. In addition, they are pigmented in nature and located in skin, making nevi readily identifiable by visual examination and allowing for monitoring in real time. Given their well-defined genetics and accessibility, nevi have been used as a model by which to study the growth arrest of lesions after oncogene mutation.

In this review, the fundamental mechanisms regulating growth arrest of nevi will be discussed in the context of clinical features commonly observed in nevi and in light of new observations in mouse models and human tissue. In addition, although nevus growth arrest is very robust and the vast majority of nevi will remain benign over time, a small proportion will progress to melanoma. Mechanisms by which growth arrest of nevi can be overcome and lead to melanoma formation will also be considered. These observations will be integrated into an updated model of growth arrest of melanocytic nevi after oncogene activation, a process we call stable clonal expansion. Stable clonal expansion in nevi is akin to the subclinical clonal expansion observed in other cell types (including in skin) and is also discussed below.

HISTORICAL CONTEXT AND ONCOGENE-INDUCED SENESCENCE

The most well-known hypothesis explaining how individual critical oncogenes can be activated, yet not give rise to cancer, is termed oncogene-induced senescence (OIS). The concept of OIS is based on the phenomenon of replicative senescence (RS), a process during which cultured cells cease proliferation after a finite number of passages in vitro.^1,2 During RS, cells lose the ability to re-enter the cell cycle, even in the presence of mitogenic stimuli.³ In culture, senescent cells exhibit a distinct cellular morphology; they become large, flattened, dendritic and often multinucleated. Senescent cells express characteristic markers such as senescence-associated beta-galactosidase (SA-β-gal) and upregulate tumor suppressors including p16^INK4A and p21^CIP1.⁴ RS is thought to result from progressive shortening of telomeres and is in part driven by the activation of a DNA damage response that occurs when telomeres reach a critically shortened length.^5,6 RS can be overcome by expression of telomerase, which can restore and maintain telomeric DNA.⁷

The OIS hypothesis dates back to the 1980s, when an interesting phenotype was noted after the introduction of individual oncogenes into non-immortalized-cultured cells. Rather than transforming the cells, oncogene expression instead induced a senescence-like phenotype.^8,9 These observations led to the early hypothesis that these senescence-like responses have a tumor-suppressive role in cancer.^10,11 More formal support for the OIS hypothesis came in 1997, when Serrano et al.¹² showed that expression of oncogenic HRAS^G12V in cultured primary cells paradoxically induced a permanent G1 cell cycle arrest; with growth arrested cells exhibiting a morphologic phenotype similar to the cells that had undergone RS.

Since this time, it has been shown that cells that have undergone OIS express similar markers to cells that have undergone RS including: SA-β-gal, H3K9Me3, γ-H2AX and p16^INK4A, among others (discussed further below).⁴ However, contrary to RS, in OIS critical shortening of telomeres does not necessarily occur. Accordingly, expression of telomerase is insufficient to bypass OIS in culture.¹³ These observations suggest that despite morphologic and biomarker similarities, OIS and RS may be fundamentally different processes.

Although OIS is well-defined in vitro, it has been more difficult to identify and study in vivo, where its exact role is debated.¹⁴ In tissue, cells with oncogenic changes exhibit some markers of senescence, but do not appear to rigidly adhere to the OIS phenotype as defined in vitro. Further, it has recently been noted that oncogenic mutations are very common in vivo in phenotypically normal tissue such as skin, and result in ‘invisible’ expansion of a quilt work of numerous, overlapping clonal lesions.¹⁵ Despite these oncogenic mutations resulting in finite clonal outgrowth, the mutant cells appear to largely maintain their ability to proliferate, differentiate and perform their normal functions. In the following sections, the evidence for and against acquisition of senescence-like features after oncogene activation in melanocytic nevi will be considered.

MELANOCYTIC NEVI

Natural history

Melanocytes are pigment-producing cells in the skin and typically reside within the epidermis, at the dermoepidermal junction and within hair follicles. Several benign neoplasms are derived from melanocytes and are typically the result of individual oncogenic mutations.¹⁶ This review will focus on the most common of these lesions, benign-acquired melanocytic nevi (referred to as nevi from here on).

Many adults have nevi, but their abundance varies tremendously from individual to individual, ranging from just a few nevi up to hundreds of lesions per person. Nevi are rarely present at birth and when they are, are known as congenital nevi. Rather, most nevi form later on in life, typically during the first and second decades.^17,18 Total nevus number in any given individual is thought to peak during the third decade of life.¹⁹ This peak is due to reduced formation of new nevi (which becomes less common after 30 years of age) combined with the clinical regression of some existing nevi. Clinical regression of nevi is a poorly understood process during which nevi involute and can disappear entirely. The frequency of nevus regression increases with advancing age.^20,21

Compared with other clinically apparent, benign, but potentially precancerous lesions, melanocytic nevi are unique as they arise relatively early in life. In contrast, for example, actinic keratosis, which can be a precursor of cutaneous squamous cell carcinoma, are uncommon prior to the age of 40 and become much more prevalent with advancing age, even into the 80s and 90s.²² The reason(s) why nevi arise primarily during the first two decades of life and less so with advancing age is unclear. The reason why some individuals get only a few nevi, whereas others get hundreds are also not well understood. In terms of abundance, a combination of inherited causes and ultraviolet radiation and other environmental mutagens, are likely at play.^23,24 Germline mutations such as in CDKN2A, which affects both nevus size and total nevus counts, underlie this phenotype in a small subset of patients.^25–28 Inherited variation in nevus and melanoma risk will be discussed in more detail below.

Clinical and histopathologic features

Nevi are most often 2–6 mm in size and have a uniform color and symmetric architecture clinically. Nevi are grouped into one of three major categories: junctional (melanocytes confined to the epidermis only), intradermal (confined to the dermis only) and compound (both an epidermal and a dermal component). The relationship among these three different types of nevi and what factor(s) result in the formation of one type versus another are not well understood. BRAF^V600E mutations, which are found in the majority of nevi, appear to occur with relatively similar frequencies in all three types, but may be slightly more common in nevi with a dermal component.^29–31 Despite the heterogeneity in clinical and histologic appearance of these types of acquired nevi, all are thought to share a relatively similar natural history and relationship to melanoma. For the purposes of this review, all three types will be considered together. It should be noted that additional types of benign melanocytic nevi such as: blue nevi,³² Spitz nevi³³ and deep penetrating nevi³⁴ exist, however, are relatively less common and will not be discussed in detail. Dysplastic nevi will be considered separately below.

Microscopically, nevi are well circumscribed, symmetric and are composed of melanocytes with a monotonous, banal cytology. Two cardinal histopathological features of nevi are nesting and maturation. Nesting refers to the tendency of nevus melanocytes to form small clusters of cells within tissue (Figure 1). Maturation is a feature of nevi with a dermal component and refers to a gradual and progressive change (from superficial to deep) in nest architecture and melanocyte cytology. As one goes deeper into the lesion, nest size decreases, cell and nuclear volume decreases, pigment production decreases and changes in cell shape occur³⁵ (Figure 1).

Figure 1.

Schematic of melanocytic nevus architecture. (a) Low power image of an intradermal melanocytic nevus stained with hematoxylin and eosin (H&E). The nevus shows features of maturation. (b) Junctional nevi are confined to the epidermis and appear as pigmented macules. Compound nevi have both an intra-epidermal and dermal component. Intradermal nevi are entirely confined to the dermis. Type A, B and C melanocytes are morphologically distinct and found at different depths within the skin. With increasing depth, nevi are less pigmented, smaller, have smaller nuclei, fewer mitoses, increased number of apoptotic cells and increased neural features. Nest size decreases with maturation. (c) High power images of type A melanocytes in the most superficial portion of the nevus. H&E stained section. (d) Type C melanocytes in the deepest portion of the nevus showing neural (Schwannian) differentiation. H&E stained section.

Cytologic features of maturation have been used to divide the melanocytes in individual nevi into three groups, types A, B and C. Type A melanocytes are most similar in morphology to normal epidermal melanocytes and are found in nests in the most superficial portions of nevi, including the epidermis and superficial dermis. Type B melanocytes are found in the mid dermis in relatively smaller nests and are also relatively smaller in size and rounder in shape. Type C melanocytes are found primarily as individual cells in lower portions of the dermis and have a more spindled/fusiform morphology. The complex architecture observed in nevi suggests that both cell intrinsic and extrinsic factors act in concert to shape nevus formation, prevent uncontrolled growth and maintain homeostasis. In melanoma, organized nesting and maturation tend to be lost. It is possible that nesting and/or maturation reflect poorly understood tumor-suppressive interactions within the tissue microenvironment, however, there is currently no data to support what (if any) active role these processes have in constraining nevus growth.

BRAF-ACTIVATING MUTATIONS CAUSE NEVUS FORMATION

Genetics of human nevi

The MAPK pathway is a central activator of cellular proliferation (Figure 2). RAF proteins are serine/threonine kinases, which when activated either by upstream mitogenic signals or via activating mutations, drive signaling through this pathway. In 2002, it was noted that BRAF-activating mutations, which render its kinase function constitutively active, commonly occurred in human cancers, including melanoma.³⁶ BRAF-activating mutations (most commonly V600E) are present in about 50% of melanomas.³⁷ The central role of BRAF as a melanoma oncogene is supported by the marked (albeit typically temporary) responses observed when BRAF inhibitors are used to treat BRAF-mutant melanomas.³⁷ Constitutive MAPK pathway activation is likely a shared feature of most melanomas and is achieved through a variety of mechanisms including mutation of other components of the MAPK pathway, such as NRAS and NF1.^16,37

Figure 2.

MAPK pathway alterations in melanoma. RAS (usually NRAS-activating mutation), BRAF-activating mutations and NF1-inactivating mutations are common drivers of constitutive MAPK pathway activation in melanoma. Activation of the MAPK pathway in isolation provides a strong proliferative signal, but ultimately negative feedback loops result in growth arrest. *Small proportion of HRAS, KRAS mutations. **KIT, GNAQ and GNA11 mutations. Mutation data from The Cancer Genome Atlas.³⁷

In 2003, it was noted that BRAF-activating mutations were also present in many nevi.³⁸ Most studies have shown that BRAF is mutated in ∼80% of nevi.^39–41 NRAS mutations have been found in about 5.9–18.2% of nevi.⁴⁰ MAPK-activating mutations appear to be a shared characteristic of most benign cutaneous melanocytic neoplasms. For example, NRAS mutations are common in congenital nevi,⁴² HRAS mutations and copy number gains are found in in Spitz nevi⁴³ and GNAQ/GNA11 mutations are present in blue nevi.⁴⁴ As an important aside, MAPK pathway mutations tend to occur in a mutually exclusive fashion in melanocytic (and other) neoplasms,³⁷ suggesting a functional redundancy with no added selective advantage of having multiple mutations in this pathway. In fact, having two different MAPK pathway-activating mutations may confer a proliferative disadvantage.⁴⁵ A variety of approaches have been used to show that nevi are clonal,^16,46–49 suggesting that the formation of nevi in humans occurs as a result of a single MAPK pathway-activating mutation in an individual melanocyte.

The etiology of BRAF^V600E mutations is debated. Ultraviolet (UV) light is thought to have a positive role in melanoma pathogenesis (especially in melanomas arising on chronically sun damaged skin), and as a group, melanoma genomes carry a tremendous burden of UV damage.³⁷ Interestingly, however, the T-to-A transversion that underlies the V600E mutation is not a classic direct UV signature mutation (C-to-T or CC-to-TT), and the distribution of nevi clinically does not match the areas of skin with the highest exposure to UV light. Furthermore, xeroderma pigmentosum (XP) patients, who are deficient in nucleotide excision repair needed for optimal repair of UV-induced DNA changes, have a markedly elevated rate of melanoma formation, yet only 11% of XP melanomas contain BRAF^V600 mutations.⁵⁰ However, some authors still implicate UV light, arguing that T-to-A transversions are a rare, but direct byproduct of damage from UV.⁵¹ Other authors have suggested nevus formation could be stochastic and due to occasional mistakes during DNA replication, which are then highly selected for and lead to nevus formation.⁵² Other, unidentified environmental mutagens have been proposed to have a role. For example, papillary thyroid carcinoma also commonly carries BRAF^V600E mutations, however, in these neoplasms UV would not be predicted to have a pathogenic role. Interestingly, it has been noted that certain geographic areas have increased rates of both papillary thyroid cancer and melanoma, relative to surrounding areas, suggesting another unknown environmental mutagen may have a role in the formation of BRAF^V600 mutations.⁵³ Overall, this issue remains to be resolved.

Overall, the important hypothesis generated from these findings is that although individual MAPK pathway mutations may initiate inappropriate proliferation resulting in nevus formation, they are not sufficient for melanoma formation in isolation. This line of reasoning provided a conceptual link between the formation of nevi and the OIS hypothesis, as in both cases oncogene activation leads ultimately to a growth arrest phenotype rather than cancer formation. A series of important studies examining the effect of the BRAF^V600E mutation in the melanocytic lineage both in vitro and in mice followed and will be discussed in the following sections.

Functional evaluation of BRAF-activating mutations

In 2005, Michaloglou et al.⁵⁴ reported that expression of BRAF^V600E in cultured melanocytes resulted in a rapid proliferative arrest. Interestingly, there was no clear initial period of proliferative advantage provided by BRAF mutation, as presumably occurs in vivo and leads to nevus formation (the timing of growth arrest is discussed further below). BRAF-mutant melanocytes were found to exhibit cytologic features and expressed markers of OIS (p16^INK4A and SA-β-gal) in this model.⁵⁴ In vivo correlates of these findings included a panel of congenital melanocytic nevi, which were also shown to stain with OIS markers p16^INK4A and SA-β-gal. As would be predicted based on the OIS hypothesis, the melanocytes in this panel of nevi exhibited preserved telomere length.⁵⁴

In 2006, Gray-Schopfer et al.⁵⁵ expanded these findings to common acquired nevi, which were also found to stain positively for p16^INK4A and SA-β-gal. These findings showed that melanocytes in nevi share some features with melanocytes that have undergone OIS in culture. However, despite these similarities, the cytologic changes exhibited during OIS in vitro (large, flat, dendritic and multinucleate) do not tend to be reflected in nevus melanocytes found in tissue. In contrast, nevus melanocytes tend to be small, mononucleate and relatively less dendritic than normal melanocytes. Altogether, these observations raise the possibility that despite the similarity in marker expression, melanocytes in nevi may be distinct from cultured melanocytes that have undergone OIS.

The first functional evaluation of BRAF-activating mutations in the melanocytic lineage in vivo was performed in 2005 in a zebrafish model. In this model, melanocyte-specific BRAF^V600E expression induced the formation of benign melanocytic proliferations called ‘fish nevi’.⁵⁶ This study provided support for the hypothesis that BRAF activation is sufficient to drive nevus formation, but does not in itself result melanoma formation in vivo. In 2009, multiple groups, observed that melanocyte-specific expression of Braf^600E in mice also resulted in the formation of benign melanocytic lesions akin to human nevi.^57–59 The melanocytes composing mouse nevi also expressed senescence markers such as SA-β-gal, but similarly to human nevi did not assume the morphologic features of OIS melanocytes in culture. Subsequent work (discussed in more detail below) has shown that although Braf^V600E-induced mouse nevi remain in a stable growth-arrested state over time, a small subset will later give rise to melanoma.⁶⁰ Variability in the penetrance of melanoma with Braf^600E in mouse models has been observed and will be discussed further below.^58–60

It is now generally accepted that BRAF activation in vitro leads to OIS and in vivo results in the formation of nevi. On the basis of the above data, there is undeniably phenotypic overlap between these two states, however, there are also clear differences. In addition, the observation that nevi serve as precursor lesions in about 25% of melanomas suggests that nevi are not inextricably terminally growth arrested in vivo. Given these differences, it is unclear if at a functional level these two processes are mediated by the same, similar, or different mechanism(s). In the following sections, the relationship between OIS, nevi and melanoma will be discussed.

RELATIONSHIP BETWEEN NEVI AND MELANOMA

The Clark model of melanoma pathogenesis posits that a series of steps occur during progression from normal melanocyte to melanoma.⁶¹ These steps include formation of banal nevi, then dysplastic nevi, then melanoma in situ, and ultimately invasive melanoma; a path thought to be driven by the progressive accumulation of pathogenic genetic/epigenetic changes.⁶² Although linear, step-wise progression may characterize the natural history of a subset of melanomas, significant evidence suggests that in most melanomas, progression is more complex and includes many distinct paths which may be dictated in part by distinct oncogenic hits (Figure 3).⁶³ Interesting new data from Bastian and colleagues regarding the sequence of different mutations in melanocytic neoplasms is discussed below.

Figure 3.

Natural history of melanocytic lesions. Traditionally progression from normal melanocyte to melanoma has been depicted in a linear fashion (linear progression), however, in individual lesions, certain stages may be skipped or never occur at all (non-linear progression pathways). Linear progression through all stages in any individual lesion is probably fairly uncommon. Melanocytes that acquire a BRAF^V600Emutation give rise to banal melanocytic nevi. Melanocytes that acquire NRAS and BRAF^non-V600E mutations may more commonly form de novo dysplastic nevi. Approximately 2/3 of melanomas arise without a known benign precursor lesion, possibly as a result of late acquisition of a MAPK pathway mutation in already sensitized melanocyte(s) with other oncogenic changes such at PTEN and/or CDKN2A inactivation. The vast majority of nevi will never progress to melanoma, many will remain clinically stable over a lifetime, whereas others will regress (dead end pathways). The most common natural history for nevi is highlighted in yellow. *Some banal nevi may later give rise to dysplastic nevi, but this is probably fairly uncommon. **It is not clear that dysplastic nevi progress to melanoma more commonly than banal nevi.

Approximately 25–33% of cutaneous melanomas arise from nevi.^64,65 Nevi which do not arise from melanoma are considered further below. In high-risk patients, such as those with numerous nevi, this number may be as high as 50%.⁶⁶ Dysplastic nevi are also discussed separately below. Transformation of nevi to melanoma has been shown to occur most commonly in non-chronically sun damaged (non-CSD) skin (intermittently sun exposed areas such as the trunk and proximal extremities) in relatively younger patients. Superficial spreading melanoma is the most common histologic subtype in these lesions.^51,67 One study suggested that junctional and compound nevi may be relatively more likely to give rise to melanoma than intradermal nevi, but this has not been definitively shown.⁶⁷

In contrast, melanomas that develop in CSD skin (such as the head and neck) are only rarely associated with nevi.^51,67 Some melanomas arising in CSD skin show a pattern termed lentigo maligna melanoma. Bastian and colleagues have proposed that melanomas arising in CSD skin and non-CSD skin are indeed fundamentally different based on divergent genetics. Non-CSD melanomas have more BRAF^V600E and PTEN mutations, whereas CSD melanomas have more NF1 and TP53 mutations.⁵¹ CSD and non-CSD melanomas likely have distinct natural histories; the subsequent discussion will be more relevant to non-CSD melanoma, given the current data available.

In melanomas that arise from pre-existing nevi, remnants of the original nevus are often evident histologically. Genetic analyses of such histologically contiguous benign nevus-melanoma pairs support the hypothesis that the melanoma cells were derived directly from the nevus cells.^31,68–72 Although cases of driver mutation (that is, RAF/RAS) discordance between paired nevus and melanoma have been reported^70,73 and are interesting mechanistically, these cases appear to be much less common and some may represent coincidental collision lesions between unrelated nevi and melanomas.

Although a small proportion of nevi will ultimately give rise to melanoma, the vast majority never will. It has been estimated that the annual transformation rate of any single nevus ranges from ∼ 1 in 200 000 in individuals under 40 years old to about 1 in 33 000 in men over 60 years old.⁷⁴ Extended over a lifetime, the risk of progression of any individual nevus to melanoma is about 1 in 3000 for men and 1 in 11 000 for women.⁷⁴ For this reason, prophylactic removal of nevi is not part of clinical practice; however, screening for progression of nevi and de novo melanoma development with serial skin examination may result in identification and treatment of melanomas at earlier stages.

Nevi are also an independent marker of overall melanoma risk. There is a well-established, positive, dose-dependent relationship between the total number of melanocytic nevi and the risk of developing melanoma.⁷⁵ This increased risk is distinct from the risk of progression of any individual nevus to melanoma and applies to (and is additive for) both banal as well as histologically dysplastic nevi.⁷⁵ The exact explanation for this observation is not completely understood, but likely relates to shared genetic and environmental factors predisposing to melanocyte neoplasia.

The overall low rate of nevus progression to melanoma suggests that robust tumor-suppressive mechanisms are enacted following BRAF and other mutations. Understanding how and why individual nevi progress to melanoma and why individuals with many nevi are at a higher risk for melanoma formation will be considered further below.

DYNAMICS OF GROWTH ARREST

The timing of growth arrest after BRAF activation is different in vitro and in vivo. BRAF activation in vitro leads to nearly immediate growth arrest (within days) with no clear period of initially increased proliferation.⁵⁴ In contrast, BRAF activation in vivo leads to an initial period of enhanced proliferation leading to nevus formation, but is ultimately followed by clinical growth arrest as a mature nevus.^59,60 A similar phenotype is observed after RAS activation, with near immediate induction of a growth arrest in vitro,¹² but an initial period of proliferation in vivo followed by growth arrest.⁷⁶ This same pattern has been noted in other cell types. The reason for this discrepancy in the timing of growth arrest is unknown, but is explored in the following section. In mouse models, the proliferation induced by Braf^V600E lasts for 14–21 days, after which lesion expansion ceases and a mature nevus is formed.⁶⁰ Braf^V600E induced nevi remain stable in size over time as mice age.⁶⁰ When Braf is activated in the context of melanocyte-specific Cdkn2a inactivation (Braf/Cdkn2a model) (Figure 4), nevus area is increased about threefold, with a final area of ∼ 0.75 mm². A similar model based on Nras^Q61K and Cdk4^R24C-activating mutations also results in nevi of roughly the same size.^77,78 On the basis of these estimates, mouse nevi, which are thought to be derived from an individual parental melanocyte, are composed of ∼1500 to 3000 melanocytes.⁷⁷

Figure 4.

Mouse models of melanocytic nevi and melanoma. Braf^V600E mutation in isolation results in the formation of small, growth-arrested nevi. When Pten in inactivated in the setting of Braf activation (Braf/Pten) no growth arrest is observed; rapid progression to metastatic melanoma ensues. When Cdkn2a is inactivated in the setting of Braf activation larger melanocytic nevi form, but are still stably growth arrested. With increasing age, a small proportion of nevi progress to melanoma at rates similar to human nevi. When Lkb1 is inactivated (constitutive mTORC1 activation) in the setting of Braf activation, growth arrest of nevi is abrogated, but melanoma never forms. When Dnmt3b is inactivated in the Braf/Pten model, most melanocytic lesions growth arrest, with rare progression to melanoma with advancing age.

In humans, it is less clear over what period of time BRAF and NRAS induced nevus formation occurs, but it perhaps can be estimated indirectly based on certain clinical observations. For example, serial nevus photography in children and adolescents (when rates of new nevus formation are highest) has shown that most enlarging nevi grow over a period of months up to a year and then stop.^79,80 In a different scenario, eruptive nevi, numerous new nevi develop within a short period of time in individual patients. Studies looking at eruptive nevi suggest that formation of the nevi typically occurs over one to several months.^81–83 However, how closely eruptive nevus formation mimics sporadic nevus formation, and the precise mechanism(s) underlying this phenomenon are unclear. Taking these observations together, it could be roughly estimated that nevus formation in humans occurs as quickly as within 1–2 months, but may take a year or more.

Most nevi in adults range in size from 2 to 6 mm and have been estimated to be composed of several tens of thousands up to hundreds of thousands of melanocytes depending on the size and type of nevus.^16,84 On the basis of this size estimate, roughly 13–16 rounds of cell division would be required to generate a nevus from a single precursor melanocyte if clonal expansion occurred equally among all daughter cells without any loss of progeny. More rounds of division are likely required as in reality proliferation is probably not perfectly exponential. Nonetheless, this estimated number of divisions is importantly significantly lower than the 60–80 rounds of cell division that would be required to result in critical telomere shortening,⁸⁵ consistent with the observation that telomere length is preserved in nevi and they do not appear to undergo RS.^54,86 If similar logic is applied to murine melanocytes, 10–11 rounds of division would be required to form a mouse nevus. In addition, telomeres are much longer in laboratory mouse strains (50–150 kb) than typically seen in humans (5–6 and 10–12 kb, adults and newborn humans, respectively)⁸⁷ and significant telomere erosion in mouse nevi is unlikely to occur.

Altogether these data strongly support the hypothesis that at least in a subset of melanocytes that have acquired BRAF^V600E mutations, growth arrest has significant latency in vivo and does not occur as quickly as it does in vitro. The reason for this discrepancy is not entirely clear. One hypothesis is that the mechanism of growth arrest in vitro is different from that occurring in vivo. If true, this may be related to the non-physiologic conditions of cell culture, where cells are already constitutively proliferating and are in the presence of favorable concentrations of growth factors and nutrients. This hypothesis is supported by the differences in time frame of growth arrest and differences in cytology between growth-arrested cells in vitro and in vivo.

An alternative hypothesis is that an immediate growth arrest phenotype analogous to that observed in vitro does occur in vivo, but is not routinely appreciated because no clinically apparent lesion develops. In this scenario in order for the formation of a visible nevus to occur, immediate senescence programs would need to be either ineffective or somehow rapidly bypassed. Along these lines, some authors have hypothesized that patients with relatively fewer nevi are more effectively able to enact an immediate senescence response after BRAF mutation in melanocytes. In these hypothetical patients, although activating BRAF mutations occur, they rarely result in formation of clinically visible nevi due to the robust and rapid onset of growth arrest programs. In contrast, patients with less robust immediate growth arrest programs would develop more clinically visible nevi, as they would rely more on secondary mechanisms of growth arrest that act with longer latency.¹⁶ However, which tumor-suppressive mechanisms potentially act immediately versus those that are secondary are not well-defined and there is no direct experimental evidence in support of this hypothesis.

The hypothesis that nearly immediate melanocytic growth arrest after BRAF activation occurs in vivo predicts that individual melanocytes or subclinical melanocytic proliferations harboring BRAF^V600E mutations should be detectable in skin. Indeed, subclinical melanocytic proliferations are encountered as chance findings in skin excisions for other cutaneous neoplasms.⁵² However, the frequency with which these lesions occur is not well characterized and it is not known if they contain BRAF mutations. Some authors have suggested that many nevi in humans may never grow larger than 1 mm in size and thus have been largely overlooked by most studies of nevi in humans.⁸⁸

In another example, eruptive nevi, in which numerous new nevi synchronously appear, BRAF^V600E mutations are present in most lesions. This observation suggests that subclinical BRAF-mutant melanocytes may have been present and then triggered to grow. Alternatively, but less likely, new BRAF mutations could be induced in multiple cutaneous locations over a relatively short time period. Last, many BRAF-mutant melanomas do not develop from preexisting nevi, suggesting either that clinically silent BRAF-mutant melanocytes preceded the melanoma, or alternatively that BRAF mutation was instead acquired relatively late in melanomagenesis, leading to de novo melanoma formation (Figure 3). This issue is considered more below.

MECHANISMS OF GROWTH ARREST

Several specific mediators of growth arrest after activation of critical oncogenes have been proposed and are based on both in vitro and in vivo experimental evidence. Although these experiments are numerous and have been performed in many cell types, the following sections will focus primarily on experiments performed in the melanocytic lineage. Where possible in vitro and in vivo data will be discussed together.

Negative feedback within the MAPK pathway

Although BRAF mutation and activation of the MAPK pathway is important in nevogenesis, MAPK pathway activation do not appear to be sustained at high levels in nevi after growth arrest. Time-course studies performed by our group in the Braf/Cdkn2a mouse model of nevus formation show that the MAPK pathway is activated only transiently after Braf mutation and corresponds to the phase of active melanocyte proliferation during nevus formation.⁶⁰ MAPK activity is significantly lower during stable growth arrest in this model. As might be predicted, the MAPK pathway is re-activated at high levels in melanoma.⁶⁰ Analysis of human nevi shows a similar pattern, with relatively low levels of MAPK pathway activation in nevi relative to melanoma.^89–91 Retention of low levels of pathway activity may be supported by the observation that treatment of patients with BRAF inhibitors results in changes in the appearance of pre-existing BRAF-mutant nevi.^92,93 BRAF-mutant melanomas show high levels of MAPK pathway activation and are clearly dependent on BRAF-induced MAPK pathway activation given the efficacy of BRAF and MEK inhibitors.⁹⁴

The mechanisms by which MAPK signaling in nevi is attenuated during growth arrest are not well characterized. Negative feedback loops involving dual specificity MAPK phosphatases (MKPs or DUSPs) or Sprouty proteins are defined inhibitors of the MAPK pathway generally, but have not been carefully studied in nevi.^95–97 Progression to melanoma appears to rely in part upon reactivation of MAPK signaling,⁶⁰ which may be facilitated by copy number gains and upregulation of mutant BRAF, but ultimately is likely also related to disruption of negative feedback loops.^91,98 Concomitant dysregulation of additional pathways (such as PI3K/ AKT and mTOR) appears to facilitate sustained MAPK pathway activation in melanoma^59,60 and will be discussed further below.

CDKN2A

The CDKN2A locus encodes two distinct proteins, p16^INK4A and p14^ARF, both of which are considered bonafide tumor suppressors in melanoma. p16^INK4A opposes Cyclin D-Cdk4/6 mediated cell cycle progression through the G₁/S restriction point via phosphorylation of pRB.⁹⁹ More recently, Cyclin D-Cdk4 has been implicated in regulating cellular glucose metabolism independently of cell cycle progression.¹⁰⁰ p14^ARF (p19^Arf in mice) inhibits MDM2-mediated degradation of p53 and may also function as a tumor suppressor by opposing ribosome production.¹⁰¹ Metabolic implications of CDKN2A loss will be discussed in more detail below.

CDKN2A is the prototypic familial melanoma susceptibility locus and accounts for ∼40% of familial melanoma. Multiple different germline mutations resulting in loss-of-function of one copy of p16^INK4A and/or p14^ARF have been reported in melanoma kindreds.^102,103 The clinical phenotype in many patients with germline CDKN2A mutations is characterized by an increased abundance and larger size of nevi^25,26 and a significantly increased risk of developing cutaneous melanoma.^104,105 However, a subset of these patients do not have elevated nevus counts, yet are still at an increased risk for developing melanoma. The observed alteration in nevus size and number may argue that CDKN2A gene products have a role in rapid induction of growth arrest, and when impaired nevus melanocytes must rely on other mechanisms that act with longer latency. In the nevi in these patients, one normal copy of the CDKN2A locus is still thought to be expressed.¹⁰⁶ CDKN2A mutations are also very common in sporadic melanomas. Inactivation of one copy of CDKN2A is also common in melanoma in situ; inactivation of both copies is more commonly found in advanced melanomas.^37,51,98 Altogether, these observations suggest that in humans there is a complex and potentially dose-dependent effect of inactivating mutations in CDKN2A in nevus and melanoma biology.

Both p16^INK4A and p14^ARF are canonical tumor suppressors and have also been implicated in growth arrest after oncogene inactivation at a functional level. p16^INK4A in particular is highly upregulated in cells that have undergone both OIS and RS, and is one of the most commonly used markers of these states.^107,108 In human melanocytic lesions, p16^INK4A staining is typically higher in nevi than in melanomas, where expression tends to be reduced or lost entirely.^109–112 Interestingly, at a functional level, neither p16^INK4A nor p14^ARF appear to be required for induction of OIS phenotypes in melanocytes in vitro.¹¹³ Similarly, they do not appear to be required for growth arrest in vivo. For example, simultaneous inactivation of both p16^Ink4a and p19^Arf in the Braf/ Cdkn2a mouse model does not abrogate Braf^V600E-induced growth arrest (Figure 4).⁶⁰ However, similarly to patients with germline CDKN2A mutations, Braf-induced mouse nevi are both more numerous and larger when Cdkn2a is disrupted.⁶⁰

The in vivo data from mice and humans suggest that CDKN2A inactivation results in a temporary disruption, but not abrogation of growth arrest, and that in the absence of CDKN2A, other growth arrest programs can still control growth of the lesions. However, in the Braf/Cdkn2a mouse model, the additional loss of Cdkn2a (compared with Braf activation alone) has the important effect of increasing melanoma penetrance from near 0 to 100%.^59,60 Interestingly, this increase in melanoma penetrance appears to be independent of growth arrest of nevi in the Braf/Cdkn2a model. Although robust growth arrest of nevi is observed, a small subset of nevi will progress to melanoma as mice age. Interestingly, the progression rate in this model is similar to that observed in other mouse nevus models and in human nevi.^60,77 Dhomen and colleagues also noted that in Braf-mutant melanocytes, that p16^INK4a loss was not required for senescence in vivo, but its loss did increase tumor penetrance and decrease tumor latency.⁵⁸ These models and factors regulating progression of nevi to melanoma will be discussed further below.

DNA damage response and p53

The role of DNA damage responses (DDR) have been intensively studied in cells that have undergoing OIS in culture. In 2006, it was shown that introduction of MAPK-activating mutations such as activated HRAS into cultured cells induced a DNA hyper-replication phenotype causing replication stress and resulting in double stranded DNA breaks. In these models, double strand breaks triggered DDR programs, which were themselves required for effective enforcement of OIS.^114,115 More recently, multiple groups have shown that natural depletion of cellular nucleotide pools after oncogene-induced hyper-replication may also lead to replication stress and contribute to activation of DDR during OIS.^116–118

The functional role of DDR programs in the growth arrest of nevi, however, is less clear. An initial evaluation of dysplatic nevi and melanomas showed that markers of DDR (such as, γ-H2AX and CHK2) were present in both types of lesions, but not normal skin.¹¹⁹ Subsequent analyses show that banal nevi also express γ-H2AX, at levels that appear to be higher than normal melanocytes, but lower than melanoma.^120,121 These data can be interpreted in different ways. For example, one could argue that a DDR, which was initially effective in enforcing growth arrest, has become ineffective (but is sustained) in melanoma, which would explain the higher levels in melanoma relative to nevi. Alternatively, is also possible that a stronger, but transient DDR occurs during growth arrest, but is not sustained during homeostatic conditions after growth arrest.

p53 is a potent tumor suppressor and is a central regulator of DDRs.^122,123 In melanoma, the TP53 gene is mutated at relatively low rates compared with most other solid malignancies.^37,98 TP53 mutations are enriched in melanomas arising on CSD skin and associated with thicker invasive melanomas, but tend to be relatively less common in non-CSD melanomas.^37,51,98 Immunohistochemical analysis of histologically contiguous human nevus-melanoma pairs has shown that the melanoma portion of the lesions tend to have higher p53 expression, whereas p53 expression is relatively lower in the nevus portion of the lesion.¹¹⁰ However, these data are difficult to interpret in the absence of knowing the TP53 mutational status.

Studies testing the functional role of p53 loss on nevus formation in mice have been performed. In the Braf/p53 model developed by our group, where p53 is simultaneously inactivated in Braf^V600E -mutant melanocytes, stable growth arrest of nevi still occurs despite the absence of functional p53 and an impaired p53-dependent DDR.⁶⁰ However, similarly to the Cdkn2a/Braf model, inactivation of p53 results in an increased total number of nevi, larger nevi, but nevi that still growth arrest (Figure 4). However, as in the Cdkn2a/Braf model, 1–4 melanomas typically arise within 100 days of life in these mice.⁶⁰ Viros et al.¹²⁴ also found that inactivation of p53 in the setting of Braf activation leads to increased melanoma formation in mice. Altogether, these observations suggest that p53 and DDR do not have an obligate role in growth arrest of nevi, but do alter the phenotype of nevi slightly and regulate the rare, stochastic progression to melanoma. This is perhaps not surprising as there are likely multiple levels of protection from transformation after BRAF activation.

Epigenetics

Epigenetics broadly refers to chromosomal alterations that affect processes such as gene expression, but do not change the actual DNA sequence. DNA methylation and histone modifications are common examples of epigenetic alterations. Epigenetics and epigenetic regulators are known to be dysregulated in cancer including melanoma and in many instances contribute to cancer formation and progression.¹²⁵ In this section, the role of epigenetics in the formation nevi and subsequent progression to melanoma will be discussed.

In nevi, ultrastructural studies using electron microscopy have shown that heterochromatin (more tightly packed, less transcriptionally active) predominates over euchromatin (relatively less condensed, more transcriptionally active). Not surprisingly, in melanoma, euchromatin predominates.^126,127 This pattern is common when compared between benign and malignant lesions in other tissues. In fact, one of the most commonly used markers of senescence, senescence-associated heterochromatic foci (SAHF), reflects an epigenetic modification, which leads to heterochromatin formation. SAHF were initially described in vitro and functionally are thought to promote senescence by silencing of E2F target genes by affecting chromatin structure. E2F target genes are critical for cell cycle progression from G₁ to S phase.¹²⁸ SAHF are detected using antibodies specific for trimethylation of lysine-9 of histone H3 (H3K9me3); however, this marker can be difficult to quantitate.

Heterochromatin formation in nevi has been proposed to be mediated by specific factors, including histone deacetylase 1 (HDAC1), the activity of which can be partially inferred by H3K9me3 staining.¹²⁹ Although initial evidence suggested HDAC1 was upregulated in nevi, subsequent analyses found H3K9me3 staining to be essentially equivalent in nevi and melanomas.^120,129 Other studies have implicated the expression of histone variant macroH2A in heterochromatin formation in nevi, with expression of macroH2A tending to be lost with progression to melanoma.¹³⁰ MacroH2A may promote the senescence-associated secretory phenotype¹³¹ (discussed below).

DNA methylation is globally dysregulated in melanoma. For example, tumor suppressor genes are commonly silenced by hypermethylation of GpG islands at promoter sites.¹³² In fact aberrant DNA methylation may be the most common genomic alteration in melanoma, with certain loci being methylated in >95% of melanomas.¹³³ Several groups have characterized the differences in DNA methylation patterns between nevi and melanoma.^134–137 Detection of certain epigenetic marks may be usefully clinically and are being developed as serum biomarkers for melanoma.¹³⁸

In 2012, Lian et al.¹³⁹ showed that the specific epigenetic modification, hydroxymethylation at the 5 position of cytosine (5-hmC), was common in nevi, but was nearly universally lost in melanoma.¹³⁹ Follow-up studies have confirmed this pattern.¹⁴⁰ The mechanism by which this epigenetic change is regulated and the functional significance in melanocytic lesions is not completely clear. Isocitrate dehydrogenase 2 (IDH2) and ten-eleven translocation (TET) proteins, such as TET2 may have a role in induction of 5-hmC in nevi.^139,141 Functional evaluation of the role of this modification in nevus and melanoma formation in vivo is likely to be very informative. The latency with which growth arrest in nevi occurs after BRAF mutation may also argue that epigenetic modifications (which may take time to take effect), have an important role in constraining growth.

Epigenetic modifications can also have a permissive role in melanomagenesis. For example, DNMT3B is a DNA methyltransferase responsible for de novo DNA methylation. In a study from our group, we found that inactivation of Dnmt3b in the highly penetrant and rapidly lethal Braf/Pten mouse model (Figure 4) markedly impaired melanoma formation and rather, resulted in the formation of nevus-like growths through a mechanism discussed below.¹⁴² This observation provides strong evidence to support the hypothesis that epigenetic modifications (specifically de novo methylation of DNA) are required for melanocytic proliferations to grow beyond a nevus-like size, even in the presence of Braf and Pten mutations that would otherwise lead to melanoma formation. DNA methylation may regulate feedback loops that might otherwise limit MAPK and PI3K signaling, which are thought to be required for melanoma growth.

Other epigenetic regulators have been proposed to have a role in melanomagenesis, but will only be mentioned briefly. JARID1B (KDM5B) is expressed at higher levels in melanoma than nevi,¹⁴³ and has been shown to be required for continuous growth of melanoma in experimental models.^144,145 SETDB1 is a methyltransferase responsible for H3K9me3 methylation (as seen in SAHF) that interestingly has actually been shown to be recurrently amplified in melanoma.¹⁴⁶ Germline SETDB1 sequence variants have been shown to confer increased susceptibility to melanoma formation.¹⁴⁷ The SWI/SNF (switch/sucrose nonfermentable) complex regulates chromatin remodeling via nucleosome sliding; components of this complex have been shown to be recurrently mutated in melanoma.^37,148,149 Last, EZH2, a histone methyltransferase is mutated in a small proportion of melanomas,^37,149 and has been noted to be upregulated in melanomas compared with nevi;¹⁵⁰ however, the functional role of this protein in melanocytic lesions is not well understood yet.

Modification of gene expression by non-coding RNAs is often considered along with epigenetics. Gene product regulation by microRNAs (miRNAs) in particular is likely to have an important role in both establishment and escape from growth arrest. For example, using the Braf/Cdkn2a model, we found that miR-99a, miR-99b and miR-100 are upregulated after Braf activation and likely help to enforce growth arrest via downregulation of mTOR signaling⁶⁰ (mTOR is discussed in detail below). miR99/100 are expressed at high levels in human nevi relative to melanoma, consistent our observations in mice.^151,152 In addition, using the Dnmt3b/Braf/Pten model discussed above, we identified miR-196B as an important suppressor of mTORC2 activation after Pten loss, by targeting mTORC2 component Rictor.¹⁴² Other microRNAs also appear to be involved in melanoma pathogenesis at various stages and have been recently reviewed.¹⁵³ Long non-coding RNAs such as MIR31HG and have been reported to have a role in BRAF-induced OIS in vitro,^154,155 but have yet to be studied in nevi.

Cellular metabolism

In recent years, study of metabolic alterations in cancer cells has regained focus. Metabolic reprogramming is central to cancer formation and progression, and is classically referred to as the Warburg effect. In normal cells, under conditions of normoxia, glucose if fully oxidized to carbon dioxide via the citric acid cycle and mitochondrial oxidative phosphorylation. This pathway is very efficient in terms of ATP production. Glycolysis, the alternative pathway of glucose metabolism, is less energetically efficient and most normal cells is only used under conditions of hypoxia. Cancer cells, however, preferentially metabolize glucose via glycolysis regardless of oxygen abundance.¹⁵⁶ Although less efficient in terms of ATP production, glycolytic pathways generate molecules useful in nucleotide, amino acid and lipid synthesis, and facilitate generation of biomass.¹⁵⁷ Rapid glucose uptake by cancer cells is so conserved that a clinical imaging modality commonly used in cancer patients (fludeoxyglucose positron emission topography or FDG-PET) specifically measures this aberration to localize cancer within the body. Specific mediators of metabolic reprogramming in cancer cells are beginning to be understood and their role in nevus and melanoma formation will be considered in this section.

Early studies in senescence and OIS showed that although senescent cells permanently exit from the cell cycle, they maintain metabolic activity.^12,158,159 Since this time, several groups have shown that oxidative phosphorylation favors development and maintenance of OIS, possibly by generating redox stress.^160–163 Consistent with this hypothesis, introduction of BRAF^V600E into cultured fibroblasts promotes oxidative phosphorylation by inhibiting pyruvate dehydrogenase kinase 1 (PDK1).¹⁶⁰ Pyruvate kinase is a second regulator of glycolysis and has also been implicated in metabolic reprogramming of cancer cells. The M2 splice isoform of pyruvate kinase (PKM2) has been shown to be preferentially upregulated in cancers and may specifically induce a Warburg metabolism.^164,165 As a side-note, despite this observation, a subset of melanomas appear to maintain and tolerate high levels of oxidative phosphorylation by upregulating reactive oxygen species (ROS) detoxification capacity.¹⁶⁶

PDK1 levels tend to be higher in human nevi than in melanoma,¹⁶⁷ consistent with the hypothesis that oxidative phosphorylation is the predominate means by which glucose is metabolized in nevi. PKM2 levels have not been specifically compared between nevi and melanomas. Functional analysis of nevi in vivo at a microscopic level in the Cdkn2a/Braf mouse model using an imaging modality analogous to FDG-PET (2-NBD glucose uptake) showed that nevi do not take up glucose at high levels (whereas the melanomas that develop in this model do).⁶⁰ In our experience in clinical practice, human nevi, including nevi larger than 1 cm² are also not FDG-PET positive. On the basis of these observations, it is reasonable to hypothesize that restriction of Warburg metabolism is likely an important factor that limits nevus growth after BRAF mutation; however, how specifically this occurs in nevi remains unclear.

In general, several other factors have been proposed to drive metabolic reprogramming in cancer and include C-MYC and HIF-1α.^168–170 In melanocytes, over expression of C-MYC has been shown to suppress OIS in vitro,¹⁷¹ though whether or not this effect was related to changes in cellular metabolism was not studied. C-MYC transcriptional activity is thought to be higher in melanomas than in nevi, consistent with this hypothesis.^171,172 In terms of HIF-1α, one study found higher levels of HIF-1α in melanomas than in nevi.¹⁷³ In the Braf/Pten model, inactivation of HIF-1α and HIF2α does not affect primary tumor formation, but does decrease metastasis.¹⁷⁴

Interestingly, although in isolation BRAF^V600E mutation promotes oxidative phosphorylation, in a fully transformed state, such as melanoma, mutant BRAF alternatively appears to promote glycolysis and support the Warburg effect.¹⁷⁵ For example, in patients with BRAF-mutant melanomas treated with BRAF inhibitors, a rapid and marked reduction in glucose uptake (including by FDG-PET) is observed; this change has been shown to correspond to a decrease in volume of melanoma cells.^176,177 The differential role of BRAF in these two contexts is likely a reflection of whether BRAF activation occurs in relative isolation (as in nevi) or rather occurs in the context of other cooperative driver mutations, which likely cooperate to coordinately dysregulate cellular metabolism promoting the Warburg effect. For example, dysregulation of the PI3K/AKT and mTOR pathways in melanoma appear to have a central role in the metabolic reprogramming of melanoma cells and allowing outgrowth of BRAF-mutant melanocytes as melanoma (the role of these pathways will be discussed in detail below).

Autophagy and endoplasmic reticulum stress

Several studies have suggested that autophagy has an important role in OIS. Autophagy is a process by which cellular proteins and organelles can be degraded under unfavorable conditions to generate both energy and macromolecular building blocks. When autophagy is activated, cellular substrates are encircled by autophagic vesicles and delivered to lysosomes for bulk degredatation.¹⁷⁸ A variety of cellular stressors can activate autophagy, including oncogene activation.^178,179

Activation of autophagy has been proposed to promote OIS in part by facilitating the senescence-associated secretory phenotype (SASP) in senescent cells.¹⁷⁹ SASP is discussed in more detail below. Interestingly, SA-β-gal which is commonly used as a marker of senescent cells, labels lysosomes and may reflect increased levels of autophagy.^180,181 Complicating interpretation of the role of autophagy in melanomagenesis is the observation that autophagy can alternatively promote or repress tumorigenesis in different contexts. For example, autophagy promotes tumor cell survival in the setting of anti-cancer therapy, including during treatment with BRAF-mutant melanomas with BRAF inhibitors.^182,183

These seemingly disparate roles for autophagy are perhaps reconcilable if one considers the context in which they occur. For example, a study in mice showed that inactivation of autophagy had opposite effects based on whether or not functional p53 was present. In this study, when p53 is intact (that is, early in tumorigenesis) autophagy has a tumor-suppressive function, however, when p53 is lost (that is, as later in tumor progression), autophagy alternatively promotes tumor progression.¹⁸⁴ The role of autophagy in cancer more broadly was recently reviewed.¹⁶²

Analysis of human melanocytic lesions supports the hypothesis that autophagy has a context-dependent role. Immunohistochemical analyses have shown that relative to early melanomas, nevi show increased staining for markers of autophagy including LC3B, Beclin1 and ATG5.^185–187 However, when comparing levels of autophagy in primary versus metastatic melanoma, metastatic lesions appeared to have higher levels of autophagy based on LC3B staining.^188,189 These findings appear to support the hypothesis that autophagy correlates with growth arrest in nevi, but may also promote progression of melanoma once it becomes invasive.

Ultrastructural analysis of nevi has shown that the number and size of most cytoplasmic organelles decrease from superficial dermal cells to deeper dermal cells, which correlates with a marked reduction in cell size/volume.¹⁹⁰ It could be hypothesized that levels of autophagy increase as a function of depth within the dermis, possibly explaining the decrease in cell size and organelle content. Maturation in nevi is discussed above and summarized in Figure 1. In this scenario, autophagy might be induced in melanocytes as they leave the epidermis/superficial dermis and venture further into the potentially less favorable microenvironmental conditions in the mid and deep dermis. Interestingly, Ivanov et al.¹⁹¹ have shown that increased autophagy-mediated degradation of histones occurs along with maturation and increases in deeper portions of the nevi. However, previous ultrastructural analyses were not able to detect changes in the number of autophagosomes as a function of nevus maturation.¹⁹⁰ Further study will be required to more clearly delineate any possible relationship between autophagy and maturation and the potential functional relevance of either process to growth arrest of nevi.

Endoplasmic reticulum (ER) stress occurs in the setting of very high levels of protein translation when misfolded and unfolded proteins accumulate in the ER, leading to an unfolded protein response (UPR).¹⁹² Activation of ER stress/UPR promotes cell survival under such adverse conditions by decreasing rates of translation and promoting degradation of misfolded proteins.¹⁸² The high levels of protein translation that occur after oncogene activation is one setting in which ER stress can occur.¹⁹² ER stress can also activate autophagy, a process which has been implicated in resistance of BRAF-mutant melanomas to BRAF inhibitors.^182,192

In 2006, Denoyelle et al.¹⁹³ reported that HRAS, but not BRAF or NRAS activation triggered ER stress in the melanocytic lineage.¹⁹³ HRAS mutations are more common in Spitz nevi, but relatively uncommon in acquired nevi. In this study, although evidence of sustained ER stress was noted in Spitz nevi, it was not evident in more typical acquired nevi. Subsequent analysis of melanocytic lesions using GRP78, a marker of ER stress, showed low levels of ER stress in nevi, but much higher levels in melanoma.¹⁹⁴ In summary, the role of ER stress in the growth arrest of nevi, if any, remains unclear.

Microenvironmental mediators

A major difference between in vivo and in vitro systems is the presence or absence of a physiologic microenvironment. In vivo, the tissue microenvironment consists of multiple cellular and non-cellular entities, which directly and indirectly interact with melanocytes and undoubtedly influence their behavior. It can be hypothesized that features of nevi observed in tissue, but not in growth-arrested BRAF-mutant melanocytes in culture, such as nesting and maturation, may be a reflection of interactions among nevus melanocytes and with the tissue microenvironment (Figure 1). During maturation, melanocytes become smaller, less pigmented and change their shape. Further, within nests themselves, melanocytes at the edges of the nest tend to be smaller, whereas those centrally tend to be larger. These patterns suggest that the phenotype of an individual melanocyte is influenced by its position within the nevus and within in the skin.

The specific factors regulating nesting and maturation are difficult to study and not well understood. In terms of maturation, a study by Perez et al.¹⁹⁵ showed that levels of the matrix metalloproteinase MT1-MMP, an extracellular matrix degradation enzyme, differ as a function of nevus maturation. However, it is unclear if or how MT1-MMP is functionally related to maturation or nesting. Extracellular matrix composition is thought to vary significantly between nevi and melanomas, however the specific factors which influence nevus nesting and maturation are not known.^196,197

The likely importance of nesting and maturation in tumor suppression is underscored by the observation that in melanoma these features tend to be disrupted. In fact, patterns of nesting and maturation are key histologic features used by dermatopathologists to distinguish nevi from melanoma in biopsy specimens. In melanoma, nest morphology is altered with nests tending to be larger, irregularly sized and/or more tightly packed, whereas in some melanomas the nesting phenotype is lost almost entirely. Similarly, maturation is lost in melanoma. Although the factors that mediate these processes are poorly understood, they are likely to have an important role in nevus formation, and at least in part, reflect a regulatory influence of the tissue microenvironment.

In skin biopsies, both normal individual melanocytes, as well as nevus melanocytes appear to prefer to be in close association with keratinocytes. Melanocytes in tissue tend to be concentrated in areas adjacent to both basal interfollicular and follicular keratinocytes. This observation also appears to be true in vitro also. Cultured melanocytes prefer to be associated with and grow better in association with keratinocytes. For this, reason a keratinocyte feeder layer is often used in the culture of melanocytes.¹⁹⁸ The specific signals that underlie this phenomenon and whether or not proximity to keratinocytes in the epidermis has a role in the process of maturation in nevi is not known.

Nevus melanocytes likely interact with other cells in their microenvironment actively through secreted molecules. For example, senescent cells including those that have undergone OIS are highly secretory, a characteristic termed the SASP.¹⁶² SASP has been shown to have a functional role in growth arrest through propagation of this phenotype in an autocrine/paracrine manner. SASP has even been proposed to activate immune surveillance of lesions in tissue.^199,200

In the setting of BRAF^V600E mutation, specific secreted factors including both inflammatory (IL-1, IL-6 and type I interferons)^201,202 and non-inflammatory (IGFBP7)²⁰³ factors have been reported to influence growth arrest phenotypes. For example, secretion of IGFBP7 was shown to drive BRAF^V600E-induced OIS in melanocytic neoplasms in 2008 by Wajapeyee et al.,²⁰³ however, these findings have been debated in the literature.^204,205 Interestingly, IGFBP7 can inhibit signaling through the IGF-1 receptor (IGF1R).²⁰⁶ We have found that Igf1r activation and in turn activation of PI3K/AKT signaling is associated with progression of nevi to melanoma in the Cdkn2a/Braf mouse model and hypothesize this is an important oncogenic driver in PTEN wild-type melanomas, by providing an alternate way to activate PI3K/AKT signaling.⁶⁰ In human specimens, nevi tend to have lower levels of IGF1R expression than melanoma,²⁰⁷ consistent with this hypothesis. This link between IFGBP7 and IGF1R in nevi and melanoma is only hypothetical.

In terms of secreted inflammatory mediators, upregulation of IL-6 and IL-8 have been shown to occur after BRAF activation in vitro and are thought to reinforce senescence in a cell autonomous manner.²⁰¹ Other inflammatory mediators such as IL-1 have been shown to regulate paracrine senescence in other models.²⁰⁰ Type I interferons have recently been shown to have an important non-cell autonomous role in growth arrest after BRAF activation²⁰² and will be discussed further in the following section.

In human tissue, both IL-1 and IL-6 are upregulated in benign nevi relative to dysplastic nevi and melanoma.²⁰⁸ Altogether, these observations are consistent with a growth suppressive role for these interleukins, however, it remains unclear to what degree in vivo these and other inflammatory mediators act at by inhibiting melanocyte growth directly versus activating immune surveillance. The role of the immune system in suppression of melanocytic neoplasia will be the focus of the following section.

Role of the immune surveillance

The immune system likely has a role in controlling growth of nevi and preventing progression to melanoma. The ability of the immune system to interact with melanocytes in a functionally relevant manner is supported by several clinical observations. For example, vitiligo is a condition characterized by complete loss of melanocytes in affected areas of skin leading to the formation of depigmented patches. Although vitiligo pathogenesis is complex, melanocyte depletion is thought to be at least in part mediated by targeted destruction by CD8+ cytotoxic T cells.^209,210 In a similar example, halo nevi, nevus melanocytes are targeted for destruction by the immune system. In halo nevi, nevus melanocytes are recognized and destroyed by CD8+ T cells leading to formation of a depigemented patch of skin around a pre-existing nevus and sometimes disappearance of the nevus altogether.^211,212

Melanoma is associated with a relatively high mutation burden and is considered an immunogenic cancer.^37,213 It has long been hypothesized that rare cases of spontaneous regression of metastatic melanoma are related to immune-mediated tumor recognition and destruction.²¹⁴ In the 1970s, it was noted that a subset of patients responded to early immune-based therapies such as bacillus calmette-guerin.^215–217 Since this time, we have learned that systemic immune stimulatory therapies such as high dose IL-2 can induce durable tumor remission in a small subset of patients with metastatic melanoma.²¹⁸ Most recently, blockade of inhibitory immune checkpoints using inhibitors of CTLA-4 and PD-1 have been shown to induce durable anti-tumor responses in a subset of melanoma patients.²¹⁹ One case of CTLA-4 inhibitor-induced regression of benign nevi has been reported to date, suggesting checkpoint inhibitors can also stimulate recognition and destruction of nevus melanocytes.²²⁰ Vitiligo-like depigmentation can also occur in patients with melanoma treated with checkpoint inhibitors (or spontaneously) and is considered a good prognostic sign.²²¹

An additional informative clinical observation from patients relates to immunosuppression. Immunosuppressed patients such as solid organ transplant recipients and patients with chronic lymphocytic leukemia have an approximately twofold increased incidence of invasive melanoma compared with non-immunosuppressed individuals.^222–224 This observation suggests that adaptive immunity has a role in suppressing melanoma formation and/or progression. Importantly, though, the increased risk of melanoma in immunosuppressed patients is relatively modest compared with some other malignancies. For example, solid organ transplant recipeints have a 65–100-fold increased risk of developing cutaneous squamous cell carcinoma.²²⁵

Matin et al.²²⁶ have suggested that the proportion of melanomas developing from nevi may be slightly higher in transplant recipients based on results from a larger study. However, the relative proportion of nevi developing from nevi versus de novo in transplant recipients has not been specifically studied and is already known to be highly variable between different studies.

Although it is clear that the adaptive immune system can recognize and eliminate melanocytes under a variety of conditions, it is unclear to what degree lymphocytes and other immune cells interact with nevi under homeostatic conditions and whether or not these interactions constrain growth and/or prevent transformation to melanoma. Some murine models have shown that in certain circumstances, cells with senescence phenotypes can be recognized and eliminated by both innate and adaptive arms of the immune system,^227–229 however, this has not specfically been studied in nevi. In tissue, banal acquired nevi tend to be relatively pauci-inflammatory in contrast to melanomas, which in general show more robust lymphocytic infiltration.^230,231 CD8+ T cell infiltration and histologic evidence of cytotoxic responses are not usually observed in nevi.⁵²

Regression is a histologic phenomenon observed in some early melanomas and is characterized by focal areas of apparent tumor cell loss and replacement by fibrosis and inflammation.²³² When observed in histologic specimens, regression is usually partial, rather than complete. Regression is typically not observed in nevi other than the outer perimeter of halo nevi. The generally accepted view is that regression reflects prior immune-mediated destruction of a portion of the melanoma, however, this is based mainly on inference from the clinical and histological appearance of regressed melanomas. Interestingly, the T cells found in areas of regression actually differ from the T cells in conditions such as vitiligo and halo nevi.^233,234 In regression, primarily CD4+, not CD8 + T cells are present.²³³ CD8+ T cells predominate in vitiligo and halo nevi. The significance of this observation is unclear. An alternative hypothesis to explain regression is that tumor cell loss is instead driven by genomic crisis occuring in incipient melanomas and leading to apoptosis^235–237 (this hypothesis will be discussed further below).

Recently, type I interferon signaling was also shown to have a tumor-suppressive role in the Braf^V600E mouse model.²⁰² In this study, inactivation of type 1 interferon receptor, Ifnar1, resulted in impaired growth arrest of Braf^V600E-mutant melanocytes and increased melanoma penetrance in vivo. The tumor-suppressive effect of interferon signaling in this model appeared to be partially melanocyte autonomous and partially melanocyte non-autonomous, suggesting one possibility is that interferon could stimulate immune surveillance of nevus melanocytes, however, this was not specifically studied. Interestingly, previous work has shown that secretion of type I interferon by senescent cells is mediated by activation of the DNA damage response.²³⁸ Therapeutic interferon α (IFN-α) has been used as an adjuvant agent in melanoma, however, its efficacy has been debated.²³⁹

The complex interplay between the immune system and neoplastic cells is underscored by the observation that chronic inflammation, can alternatively promote tumorigenesis over time.²⁴⁰ Other cell types including some myeloid-derived cells and macrophages can support the formation and progression of cancer by multiple mechanisms.²⁴¹ For example, Gr-1+ myeloid cells have been shown to oppose senescence in a murine model of prostate cancer.²⁴² In melanoma, tumor-associated macrophages have been shown to facilitate melanoma progression by various mechanisms.^243,244 The role of chronic inflammation, macrophages and other myeloid-derived cell populations have not been closely studied with respect to nevus biology.

The role of telomeres

Telomeres are protective structures at the ends of chromosomes formed by a repetitive DNA sequence and an associated protein complex (shelterin). The DNA portion of telomeres becomes progressively shorter after each round of cell division and upon reaching a critically shortened length triggers RS (as discussed above). The number of cell divisions required for induction of this process is called the ‘Hayflick limit’ and has been estimated to require ∼60 to 80 population doublings.⁸⁵ The observations that growth arrest after oncogene activation in vitro occurs rapidly,⁵⁴ that telomerase expression cannot overcome this growth arrest,¹³ an estimated 13–16 rounds of cell division are required for formation of a nevus, and that telomere length tends to be preserved in nevi,^54,86,245 all support the hypothesis that RS should not contribute to nevus growth arrest.

In 2003, it was reported that the promoter of the telomere reverse transcriptase (TERT) is mutated at very high frequencies in melanomas, but not in nevi.^37,246,247 Telomerase can extend shortened telomeres and is often aberrantly re-activated in cancers allowing proliferation beyond the Hayflick limit. Nevi have been shown to have absent or relatively low telomerase expression, but in melanoma telomerase is commonly expressed at relatively high levels, especially in advanced lesions.^248–250

TERT promoter mutations are thought to result in increased TERT gene expression by creating binding motifs for ETS/TCF transcription factors.²⁵¹ Given that these transcription factors are activated downstream of oncogenic pathways such as MAPK and WNT, it is possible that in the presence of TERT promoter mutations, activation of these pathways drives expression of TERT. TERT promoter mutations have been shown to be associated with increased telomerase expression in melanoma.²⁵²

In the context of these observations, one might predict that TERT promoter mutations provide a selective advantage in advanced melanomas, but might have relatively less important role in nevi and in situ melanomas. However, recent findings by Shain et al.⁹⁸ have provided evidence somewhat contrary to this hypothesis.⁹⁸ They show that TERT promoter mutations are found in combination with BRAF mutations in indeterminate melanocytic lesions (dysplastic-nevus spectrum) and melanoma in situ,⁹⁸ suggesting TERT promoter mutations provide an early selective advantage.

One hypothesis to reconcile these seemingly disparate observations is that although telomeres are not critically shortened in nevi, they become so during the transition to melanoma. Along these lines, Bastian and colleagues have proposed that histologic regression observed in early melanomas (as discussed above) reflects the aftermath of a catastrophic genetic event that is initiated by critical telomere shortening. In this hypothetical model, melanocytic subpopulations of incipient melanomas that are not able to overcome genetic stress induced by critical telomere shortening undergo apoptosis, resulting in the loss of areas of neoplastic melanocytes.^235,236 In continuing with this line of reasoning, this group has also proposed that telomere shortening actually does occur in nevi as patients age and explains the eventual regression of nevi in older patients. In this model, melanocytes in nevi would be predicted to slowly replicate overtime leading to critical telomere shortening that drives disappearance of nevi after RS. Although intriguing, to date, there is little experimental evidence to directly support this hypothesis.

An alternative hypothesis is that TERT expression provides a telomere-independent function that is important in early stages of melanomagenesis. For example, TERT has been shown to promote C-MYC stabilization;²⁵³ C-MYC is known to suppress growth arrest phenotypes in melanoma.¹⁷¹ Telomerase has other telomere-independent functions, which may also be important and have been recently reviewed.²⁵⁴ Last, oncogene activation has been shown to cause telomere dysfunction, which can induce growth arrest even in the presence of non-critically shortened telomeres through induction of DNA damage responses.^255–257 It is possible that TERT expression may also relate to this observation in some way, however, telomere dysfunction has not been well documented in nevi.

The importance of telomere biology in melanocytic neoplasia is underscored by the observation that inherited mutations conferring an increased risk of melanoma cluster on genes that encode components of the telomere shelterin complex, in addition to TERT itself. These genes include POT1, ACD and TERF2IP.^258–261 It has curiously even been reported that inherited variation in telomere length correlates with both total nevus number and nevus size.¹⁹ An improved understanding of the role that telomeres and TERT promoter mutations have in melanocytic neoplasia has the potential to significantly advance our understanding of factors regulating early stages of melanoma formation.

The role of PTEN, PI3K and AKT

PTEN is a tumor suppressor in the PI3K/AKT pathway. Functional data from murine models and observations in human lesions strongly implicate PTEN in restricting BRAF^V600E-induced melanomagenesis in vivo; perhaps more convincingly that any other proposed mechanism of growth arrest after BRAF activation in melanocytes.

The PI3K/AKT and mTOR signaling pathways are central regulators of cell growth. These pathways are highly conserved and are typically activated downstream of receptor tyrosine kinases, but are also regulated by sensors of intracellular conditions.²⁶² Activation of PI3K signaling results in activation of PDK1, which in turn activates the critical downstream kinase, AKT (Figure 5). PI3K and AKT signaling is also highly integrated with mTOR signaling (discussed in more detail below). PTEN, through its lipid phosphatase activity is a critical inhibitor of the PI3K/AKT pathway and considered a canonical cancer tumor suppressor.²⁶³ Although the discussion below will primarily focus on the role of PI3K signaling through AKT, PDK1 also has targets other than AKT, which are likely also important in melanoma pathogenesis.^264,265

Figure 5.

Overview of the PI3K/AKT/mTOR pathway. When activated via mutation, this pathway provides a constitutive cellular growth signal. PTEN is a central tumor suppressor upstream of PDK1/AKT. LKB1 can inhibit mTORC1 via AMPK and TSC signaling. mTORC2 activates AKT by phosphorylating the S473 residue, whereas PDK1 activates AKT by phosphorylating T308. Activation of both mTORC1 and mTORC2 are required for progression of nevi to melanoma. Canonical outputs of mTORC1 include S6K and 4E–BP1. Tumor suppressors: red, oncogenic effect: blue. RTK, receptor tyrosine kinase.

In 2004, it was noted by Tsao and colleagues that PTEN inactivation tended to occur more frequently in BRAF than NRAS-mutant melanomas.²⁶⁶ In 2004, it was also noted at a functional level that concurrent dysregulation of the MAPK and PI3K/AKT pathways in cultured fibroblasts resulted in G1 to S progression, whereas activation of either pathway in isolation did not.²⁶⁷ In 2006, Courtois-Cox and colleagues noted that constitutive activation of MAPK signaling, which results in OIS in culture, was associated with induction of negative feedback loops that inhibited both MAPK and PI3K/AKT signaling. These data importantly suggested that after activation of MAPK signaling, negative feedback to PI3K/AKT helps constrain proliferation as part of OIS phenotypes in culture.⁹⁵ Along these lines, in 2008, Cheung and colleagues reported that BRAF^V600E can cooperate with AKT3 to drive early melanoma formation in vitro,^268,269 again suggesting coordinated dysregulation of these two pathways is a potent oncogenic driver. Although the exact feedback loops that are important in melanoma need to be more clearly defined, one group has shown that BRAF activation leads to specific negative regulation of AKT signaling by feedback inhibition of the mTORC2 component Rictor (regulation of Akt by mTORC2 is discussed below).²⁷⁰

In 2009, it was noted by our lab that simultaneous Braf^V600E mutation and Pten inactivation showed tremendous synergy in driving melanoma development in a mouse model (Braf/Pten model). As discussed above, Braf^V600E mutation alone induces formation of mouse nevi, but rarely melanoma, even with long latency. However, when Pten is simultaneously inactivated in this model (Braf/Pten model), detectable growth arrest is no longer observed and unabated melanocytic proliferation ensues without delay, leading to synchronous formation of innumerable melanomas and rapid lethality from overwhelming tumor burden within 3–4 weeks (Figure 4).

The phenotype observed in the Braf/Pten model strongly supports the hypothesis that Pten has a central role in restricting melanoma formation after activation of MAPK signaling. Data published by Vredeveld and colleagues supports and extends this hypothesis. They show that that Pten depletion in already established Braf^V600E-induced nevi using an shRNA approach also results in melanoma formation. These data suggest that Pten remains an important tumor suppressor in established nevi and that inactivation of Pten and activation of PI3K/Akt is a mechanism by which nevi may progress to melanoma.²⁷¹ Activation of PI3K/ Akt signaling was also shown to have a similar effect in a mouse model of pancreatic neoplasia based on activated Ras (Ras^G12D).²⁷²

Analysis of human lesions also supports the important role of PTEN in restricting melanoma formation. Immunohistochemical staining for PTEN tends to be strong and uniform in melanocytic nevi.^273,274 In contrast to nevi, PTEN is dysregulated in melanoma. Complete loss of PTEN staining occurs in about ∼ 1/3 of melanomas, with reduced or altered expression found in another ∼1/3.²⁷⁴ PTEN inactivation occurs by mutation in a small subset of melanomas, but is more commonly silenced epigenetically.^98,274–277 As would be predicted based on these observations, levels of AKT activation have been found to be lower in nevi relative to melanomas using phospho-specific antibody staining.^278–280 Immunohistochemical analysis of contiguous nevus-melanoma pairs also shows that although nevus portions of the lesion tend to have high PTEN expression and low levels of AKT activation; the melanoma portions tended to show reduced PTEN expression and increased levels of activated AKT.²⁷¹

As discussed above, PTEN inactivation is enriched in BRAF-mutant melanomas suggesting that there may be a particularly potent synergy between BRAF^V600E and PTEN loss.^37,266,281 These two mutations tend to be enriched in non-CSD melanomas,^16,51 suggesting that these mutations may define a biological subset of melanomas that tend to be less related to chronic ultraviolet exposure, more frequently associated with a nevus precursor, and affect relatively young patients.

The role of mTOR signaling

PI3K/AKT signaling is tightly integrated with mTOR signaling, which occurs as part of two distinct complexes, mTORC1 and mTORC2. mTORC1 integrates signals from growth factors, other pathways, and sensors of cellular nutrient, energy, and redox status to control protein synthesis and other anabolic processes.²⁸² PI3K/AKT is a potent upstream regulator of mTORC1 and provides a strong activation signal to the complex via inhibition of TSC and PRAS40.²⁸² mTORC2, is nutrient insensitive, but is also thought to be activated downstream of growth factor signaling. mTORC2 mediated phosphorylation of AKT on S473 is required for full AKT activation²⁸² (Figure 5).

In the Braf/Pten model, constitutive activation of both mTORC1 and mTORC2 is observed.^59,60 As mTORC1 is a highly conserved output of activated PI3K/Akt signaling, and we found that the mTORC1 inhibitor rapamycin inhibits melanoma growth in the Braf/Pten model,^59,60 we hypothesized that mTORC1 activation was essential to the effect of Pten loss in Braf-mutant melanocytes. To test this hypothesis, we generated mice with Braf activation in the context of either Lkb1 (Braf/Lkb1) or Tsc1 (Braf/Tsc1) inactivation, which result in isolated activation of mTORC1 without directly affecting mTORC2 activity.⁶⁰ Interestingly, constitutive activation of mTORC1 abrogated growth arrest of nevi, however, full transformation to melanoma did not occur as in the Braf/Pten model (Figure 4). In these models, confluent melanocytic neoplasia, rather than formation of discrete nevi was observed; however, although mice were tracked for >1 year, the melanocytic neoplasms in this model never exhibited malignant behavior such as uncontrolled or invasive growth, a Warburg metabolism, or metastasis.

Lack of a fully malignant behavior in the Braf/Lkb1 and Braf/Tsc1 models is likely at least partially due to well-defined negative feedback loops induced by constitutive activation of mTORC1, which oppose activation of mTORC2.²⁸² Interestingly, we found that Cdkn2a inactivation in the Braf/Lkb1 model enabled simultaneous activation of mTORC1 and mTORC2 leading to rapid melanoma formation.⁶⁰ The mechanism by which Cdkn2a inactivation permits mTORC2 activation in this model is unclear, but may be related to novel roles of Ckdn2a in regulating cellular metabolism. For example, Cdk4 was recently shown to control cellular glucose uptake in addition to cell cycle control.¹⁰⁰ Arf has also been shown to regulate cellular metabolism.¹⁰¹

The central importance of mTORC2 in melanoma pathogenesis was also recently illustrated by our lab using the Pten/Braf/Dnmt3b model in which Dnmt3b mediated repression of miR-196b is required to alleviate inhibition of mTORC2 signaling by targeting of the mTORC2 component Rictor.¹⁴² Without this activity of Dnmt3b, full activation of mTORC2 and formation of melanoma does not occur. This work uncovered an unanticipated epigenetic checkpoint that regulates full mTOR activation in the progression of nevi to melanoma. The oncogenic role of Rictor in human melanoma is supported by the observation that the RICTOR gene is frequently amplified in melanoma.¹⁴⁸ Rictor has been shown to be essential for melanoma formation in other models as well.²⁸³

Analysis of human melanocytic lesions supports the hypothesis that activation of both mTORC1 and mTORC2 are important in melanoma, as the activity of both complexes tends to be relatively low in in nevi, but high in melanoma. For example, activation of mTORC1 substrates such as ribosomal protein S6 are only rarely present in nevi, but frequent in human melanomas.²⁸⁴ Similarly, levels of eukaryotic initiation factor (eIF-4E), which is inhibited by mTORC1 substrate eIF-4E binding protein 1 (4E–BP1) are also significantly lower in nevi than in melanoma.²⁸⁵ mTORC2 activity has similarly been shown to be lower in nevi than in melanoma. For example, levels of phospho-AKT-S473, a phosphorylation event catalyzed by mTORC2, are significantly lower in nevi than in melanoma. ^60,278–280

Altogether, these observations support the hypothesis that activation of PI3K/AKT signaling, including mTORC1 and mTORC2, is central to the pathogenesis of BRAF^V600E-mutant melanoma. A requirement for simultaneous activation of mTORC1 and mTORC2 has been proposed in other malignancies and is also likely to be important in NRAS and NF1-mutant melanomas. Although the exact reason why concurrent activation is required for a fully malignant phenotype is not completely clear, it has been hypothesized to relate in part to metabolic reprogramming of tumor cells.^282,286 In this conceptual model, MAPK pathway mutations would provide the initial proliferative signal, whereas dysregulation of PI3K/AKT/mTOR signaling enables sustained growth. In the absence of PI3K/AKT/mTOR dysregulation, BRAF mutation would only induce transient proliferation as seen during nevus formation. In melanomas that do not develop from a nevus precursor, BRAF mutation may occur relatively later in an already sensitized melanocyte with PI3K/AKT/mTOR dysregulation (Figure 3). Pten loss in our murine models does not induce analagous melanocyte proliferation in isolation.

Separate work has suggested that alternatively mTORC1 may become increasingly activated in senescent cells and may help enforce senescence phenotypes through activation of SASP.^287–290 The reason(s) for the discrepancy in these observations with data from mouse models are unclear, but may be related to differences in cell type, in vitro versus in vivo effects, relative levels of mTORC1 activation, and/or difference in subcellular localization of activated mTORC1.

Additional pathways such as Wnt/β-catenin are also important in melanoma pathogenesis. We have shown that stabilization of β-catenin through activating mutation, enhances melanoma metastasis in the Pten/Braf mouse model.^63,291 However, we also showed that β-catenin stabilization results in an altered nevus phenotype with more polymorphous lesion size and increased pigment compared to Braf activation alone. In a lung cancer model, β-catenin has been shown to cooperate with activated Braf to suppress OIS via induction of c-Myc expression.²⁹² Wnt/β-catenin activation is thought to occur in ∼1/3 of human melanomas.²⁹³ The example of β-catenin illustrates that there are many potential combinations of cooperative oncogenic hits in human melanocytic lesions, which give rise to melanocytic proliferations with heterogeneous and often overlapping clinical and histologic features. The possible sequence and differential contribution of various mutations, as recently described by Shain and colleagues, is discussed below.

DYSPLASTIC NEVI

The above discussion has largely focused on banal acquired melanocytic nevi, lesions for which most, if not all, dermatopathologists would agree on the diagnosis based on histologic grounds. Other melanocytic lesions are not as clear cut and can have some features of melanoma and some features of nevi, either clinically, histologically, or both; creating a diagnostic gray zone. Dysplastic and atypical nevi are terms used by clinicians to describe lesions with concerning histologic or clinical features, respectively. As part of clinical practice, histologically dysplastic nevi are often graded based on the degree abnormality into categories of mild, moderate and severe dysplasia; with severe dysplasia bordering on melanoma, but not quite meeting diagnostic criteria. This system of grading histologic dysplasia and its implications are controversial because of a lack of consensus terminology and disagreement over clinical management of these lesions. Further, this grading system implies that progression through different degrees of dysplasia toward melanoma occurs in a linear, progressive fashion.^294,295 However, the natural history and biologic significance of dysplastic nevi is not well characterized. For example, it is not known what proportion of dysplastic nevi develop de novo versus what percentage could represent evolution of previously banal nevi. It should be noted that in the vast majority of cases, definitive histological features of banal nevi and dysplastic nevi are not observed in the same lesion, suggesting progression from one to the other is probably rare.

The relationship of dysplastic nevi to melanoma is also incompletely understood. For example, it is unclear if individual dysplastic nevi progress to melanoma at higher rates than banal nevi. In fact, this seemingly straight forward question is difficult to study directly as to establish a diagnosis of dysplasia, the lesion must be biopsied (usually fully removed). Further, clinical aytpia does not necessarily correlate with histologic dysplasia,²⁹⁶ suggesting these lesions cannot reliably be identified clinically and followed. Several studies have indirectly addressed this question by comparing the frequency with which melanomas are associated with the remnants of banal nevi versus the remnants dysplastic nevi.^64,297–301 These analyses have generally shown that dysplastic nevi tend to be associated with melanomas at similar rates as banal nevi, however, this observation may be confounded by the relative abundance of banal nevi relative to dysplastic nevi.^294,295

A recent study by Shain et al.⁹⁸ found that melanocytic lesions in the diagnostic gray zone between nevi and melanoma may be genetically distinct from banal nevi. This group found that whereas banal nevi typically have BRAF^V600E mutation only, ambiguous lesions tended to have NRAS mutations and BRAF^non-V600E mutations, as well as TERT promoter mutations. In this excellent work, the authors also address the sequence in which mutations are likely to have occurred. Other studies have also found relatively lower rates (∼60%) of BRAF^V600E mutation in dysplastic nevi, compared to ∼80% in banal nevi.^{38,40,91,302,303} These data suggest that histologic dysplasia may hold a meaningful, lesion intrinsic biologic significance and also implies that dysplastic nevi are often likely not derived from previously banal nevi. It also suggests that BRAF^V600E-mutant and non-BRAF^V600E-mutant melanomas may show distinct natural histories. Despite these observations by Shain and colleagues, however, it is still not clear that histologically dysplastic lesions have an increased risk of progression to melanoma compared to their banal counterparts, they may just be more morphologically similar to melanoma histologically. Further work will be required to better characterize the relationship among benign nevi, histologically dysplastic nevi and melanoma.

Despite the controversy in this area, a history of a histologically dysplastic nevus is still clinically significant for patients. At a population level, patients with a history of nevi with increasing histologic dysplasia carry a dose-dependent increase in the overall risk of developing melanoma.^304,305 This increased risk appears distinct from the individual lesion actually biopsied and diagnosed as dysplastic. It is unclear if this increased risk is related to exposure to mutagens such as ultraviolet light, an inherent genetic susceptibility to melanocytic neoplasia, or a combination of both.

STABLE CLONAL EXPANSION

OIS is a paradigm that has been used to understand growth arrest after oncogene activation and has significantly advanced our understanding of neoplasia broadly, including the importance of cooperation between multiple oncogenes in cancer. However, this terminology is slightly confusing when applied to melanocytic lesions in tissue, as although cells express some markers of senescence, overall they appear to have relatively few phenotypic features of senescent cells (as discussed above). In fact, in 2012 Tran et al.¹²⁰ found that levels of the most commonly used markers of OIS do not readily distinguish between nevi and melanoma. These markers included p16^INK4A, p53, SA-β-gal, PML, SAHF (H3K9Me, 4′-6-diamidino-2-phenylindole) and DNA damage response (γ-H2AX). Further, authors have also noted that one of the most robust OIS markers, SA-β-gal, can be detected not only in nevi but also in some late stage melanomas, including metastasis.^55,120,306 Although no individual histologic marker is perfect, these observations suggest that thinking about nevi slightly differently could be useful; we propose that stable clonal expansion may be a more useful term in describing this process moving forward than oncogene-induced senescence (Figure 6).

Figure 6.

Mechanisms of growth arrest during stable clonal expansion. After acquisition of individual oncogenic mutations (BRAF^V600E), growth arrest of melanocytic nevi is established and maintained by multiple different, overlapping mechanisms. Progression to melanoma likely requires simultaneous abrogation of multiple growth suppressive pathways.

Several clinical observations suggest that nevi are not static (senescent), even after they reach a seemingly final size. For example, although the majority of nevi do appear to remain relatively stable in size over time, a subset will enlarge. In clinical practice, when enlarging nevi are noted, they are typically biopsied to evaluate for melanoma. But, does enlargement of nevi necessarily mean that growth arrest mechanisms have been bypassed and progression to melanoma has occurred; as would be implied based on the OIS hypothesis? To address this question, Lucas and colleagues carefully tracked individual nevi over time in adults using total body photography and biopsied lesions that had changed appearance (including increased in size). They found that increase in size alone in otherwise non-concerning lesions rarely led to a diagnosis of melanoma.³⁰⁷ Similar studies have also shown that most enlarging nevi, when biopsied, are diagnosed as nevi (both histologically banal and dysplastic) and not melanoma.^79,308,309 These observations are not included to suggest that enlarging nevi should not be biopsied to rule out melanoma in clinical practice, but rather emphasize that from a mechanistic standpoint tumor-suppressive mechanisms constraining progression to melanoma remain intact even during periods of clinical growth. Multiple distinct mechanisms are likely overlapping/redundant in the maintenance of benignity and preventing uncontrolled outgrowth of nevi (Figure 6).

Interestingly, nevi can change appearance in other settings, such as during pregnancy and after exposure to ultraviolet radiation, changes that do not necessarily signify progression to melanoma. For example, in pregnant patients, nevi can change color, dermatoscopic appearance,^310,311 increase in size^312,313 and show increased mitotic rates.³¹⁴ Although melanoma does rarely develop during pregnancy, this is very rare compared to these common changes in benign nevi. Exposure to ultraviolet radiation induces proliferation of nevus cells,^315,316 yet this proliferation does not equate to melanoma formation. In other words, nevi appear to maintain the ability to respond physiologically to various stimuli, including increasing proliferation rates while maintaining their benignity and stability of growth arrest.

Additional observations suggest that, in fact, nevus melanocytes actually retain significant proliferative potential. For example, nevi can regrow in patients when only partially biopsied (incompletely removed) and are known as recurrent nevi. In recurrent nevi, nevi regrow to a similar size within the scar at the biopsy site, but then again stop clinical growth and remain benign.^317,318 Data from in vitro work also suggests that nevus melanocytes retain proliferative capacity. In several reports from the 1980s, it was shown that nevus melanocytes derived from clinical specimens can proliferate in culture, where they actually proliferate faster than normal melanocytes grown under the same conditions.^319–321

In fact, close histologic examination of nevi shows that even common banal nevi have mitotically active melanocytes, which are present at low, but reproducible rates.^322–324 Glatz et al.³²³ estimated that 0.024 mitoses are present per mm² in histologically banal melanocytic nevi. Using Ki67, a marker of actively cycling cells, Soyer et al.⁸⁴ reported that 0.78% of nevus melanocytes, an estimated 2200 cells out of 282 000 cells per mm³ were cycling at the time of biopsy. In nevi, proliferative activity is generally restricted to the most superficial portions of the lesion^322–324 (Figure 1).

If the above estimates are correct, and melanocytes in nevi do slowly divide over time, then melanocyte attrition would be need to be present at a similar rate to maintain a relatively stable size over time. Along these lines, low rates of apoptosis have also been reported in melanocytic nevi.^325,326 It is unclear whether apoptosis in nevi is a melanocyte intrinsic phenomenon or whether apoptosis is induced by some other factor or cell type in the microenvironment; for example, lymphocytes (though satellite cell apoptosis is not routinely observed in nevi). Interestingly, contrary to mitotic activity, which tends to favor the superficial portion of the lesion, apoptotic cells predominate in deeper portions of the dermis and are almost never found in superficial portions of the lesion.³²⁶ Given these observations, it is tempting to hypothesize that in nevi with a dermal component, histologic maturation reflects a process in which melanocytes which are generated in superficial portions of the lesion migrate over time into deeper portions of the lesion where apoptosis occurs (Figure 1). However, to date there is no experimental evidence to support this hypothesis.

Shain et al.⁹⁸ have proposed that continued proliferation of nevus melanocytes over decades results in progressive and ultimately critical shortening of telomeres, resulting in telomere crisis. In this model, eventual telomere crisis, RS, and subsequent clearance of nevus melanocytes might mechanistically underlie the observed clinical regression of nevi with advancing age.⁹⁸ Clinical nevus regression is associated with complete disappearance of the lesion and replacement of melanocytes by fatty and fibrotic tissue.²⁰ However, as discussed above, critical shortening of telomeres in nevi has not been documented to date,^54,86 despite the apparent early selection for TERT promoter mutations during transition to melanoma (as discussed above).

Overall, these observations suggest that although nevi demonstrate tremendous clinical stability, behind the scenes nevus melanocytes are seemingly fairly dynamic. Microscopically low rates of proliferation are balanced by cell attrition. Nevus melanocytes can respond to environmental stimuli and even increase their proliferation without transforming to melanoma. Stable clonal expansion is maintained by multiple, overlapping, nearly fail-safe mechanisms including pathway intrinsic and extrinsic feedback loops, epigenetic reprogramming, microenvironmental effects, metabolic constraints and others (Figure 6).

These dynamic (rather than static/senescent) features of nevus melanocytes are notable in the setting of the recent observation that in clinically normal skin, individual oncogenic keratinocytic clones are actually fairly ubiquitous.¹⁵ In this study, it was shown that potent oncogenic mutations in NOTCH, TP53 and FGFR3 are commonly present in morphologically normal keratinocytes and result in the subclinical expansion of mutant clones, despite the otherwise normal appearance of the skin clinically. These mutant clones provide local selective advantage and ultimately result in formation of a quilt-like pattern of partially overlapping, competing clones. These results are important as they show that despite the presence of potent oncogenic mutations, normal cellular function and tissue viability (including proliferation) is maintained. Constitutive proliferation of keratinocytes is required to continuously turnover skin and maintain epidermal barrier function; a process that seemingly proceeds undisturbed within mutant clones. In keratinocytes at least, a tumor-suppressive response that involved complete and irreversible withdrawal from the cell cycle would be predicted to result in disappearance of the clone over time. Applied to melanocytes and other cell types, these observations suggest that despite oncogenes which result in clonal expansion, typical cellular function can be maintained.

This principle is likely to be important in other tissue types, even those not constantly exposed to a potent mutagen like ultraviolet light. Even in the absence of an outside mutagen, cells are constantly exposed to new oncogenic insults generated during DNA replication. For example, Chandeck and Mooi estimate based on organism wide rates of cell division in humans (5 million every second) and the inherent imperfection in DNA replication which leads to an unrepaired point mutation in 1 out of every 1 billion replicated bases, that by chance activating mutations occurring in any given oncogene (for example, BRAF^V600E) occur approximately every 10 min somewhere in the body.⁵² Although the vast majority of these cells are likely eliminated, never expand, or are shed/lost, a subset likely clonally expand and suggest that the body must deal with a constant barrage of mutant clones, yet maintain tissue function. The importance of this effect is underscored by the more recent observation that differential rates of cancer development in different tissue types correlates with the frequency with which stem cells divide in that tissue. This presumably incidentally leads to enhanced generation of mutant oncogenes, which persist in stem cell populations.³²⁷

Taking these various considerations together, we favor the term stable clonal expansion when referring to melanocytic nevi in tissue, rather than oncogene-induced senescence. To us, this term better reflects nevus phenotypes observed in mouse models and in human lesions. In nevi, despite stable lesion size clinically, the melanocytes in nevi are dynamic and multiple cooperative cell intrinsic and extrinsic factors restrain continuous growth (Figure 6). However, this process can be overcome as part of progression to melanoma, with the acquisition of additional pathogenic mutations and failure of a critical mass of growth suppressive programs.