Testing the divergent pathway hypothesis for melanoma: recent findings and future challenges

Melanomas are common cancers of the pigment cells of the skin for which the incidence continues to rise in most fair-skinned populations [1–4]. Mortality from melanoma has also increased in many populations in recent decades [5], although for patients presenting with thin lesions, gains in survival have been observed, which have been attributed to improvements in early detection and better surgical treatment [6]. Notwithstanding the encouraging trends in survival, an understanding of the causes of melanoma remains fundamental to reducing the burden of this cancer. During the past several decades, an avalanche of research across disciplines as diverse as epidemiology, genetics, molecular biology, histopathology and dermatology has identified an intriguing interplay of environmental and genetic factors that cause melanomas of the skin.

From the environmental perspective, sunlight is the key causal factor. While experimental evidence is scarce, the epidemiologic data are persuasive and consistent: melanoma incidence is markedly higher among fair-skinned than dark-skinned peoples [7]; the incidence of melanoma in countries spanning many degrees of latitude (e.g., Australia [8], New Zealand [9] and the USA [10]) generally rises with proximity to the equator, and fair-skinned migrants from low- to high-solar environments develop melanoma at rates many-fold higher than their country of origin [11]. Yet despite this compelling descriptive evidence, several observations indicate that the association between sunlight and melanoma does not accord with a simple model in which risk increases directly with exposure. For example, the findings that melanomas occur more commonly in indoor than outdoor workers [12,13], that most melanomas develop on body sites habitually covered by clothing [14] and that the distribution of melanomas across body sites varies consistently by age [9,15] all suggest that melanoma development depends upon other factors in addition to sunlight.

We recently proposed a model for melanoma that explicitly accommodates variations in melanoma development by incorporating interactions between host characteristics, anatomical site and sun exposure [16]. The premise (as yet untested experimentally) is that melanomas are a consequence of unchecked melanocytic proliferation, and that the degree of proliferation is a trait that is variably distributed in the population. The model predicts that only modest levels of sun exposure are required to initiate melanoma in people with inherently high proliferative capacity (putatively characterized by large numbers of nevi on the skin), and that these melanomas will be numerically most common at body sites with large melanocyte populations, such as the trunk. By contrast, people with lower proliferative capacity (i.e., with few nevi) require larger amounts of sun exposure to drive melanoma development, and their tumors will tend to arise on habitually sun-exposed sites, such as the face, head and neck.

A number of epidemiologic observations support this notion. Descriptive epidemiologic analyses of melanoma notifications to population-based cancer registries around the world have highlighted striking differences in the age-distribution of melanoma by anatomic site [17]. Thus, whereas the incidence of melanomas occurring on the trunk peaks in middle-age (younger than for most other cancers), head and neck melanomas become increasingly common with advancing age. A recent analysis of USA cancer registry data used advanced mathematical models and found strong evidence for divergent age-dependent pathways for melanoma development [18]. One interpretation is that younger patients with melanoma represent a distinct subgroup of the population susceptible to melanocytic proliferation. Inferences from registry data are limited by the absence of sun-exposure information, but other epidemiologic studies have addressed this limitation by collecting detailed sun-exposure histories from melanoma patients. Case–control studies have reported that patients with truncal melanomas report significantly lower levels of accumulated sun exposure than patients with head and neck melanomas [19]. The concordant findings from descriptive and analytical studies are consistent with melanomas at different body sites arising through different pathways, but do not shed light on the mechanisms of pathogenesis. To address this issue, investigators have systematically appraised histological features and noted that melanomas arising from a benign melanocytic nevus differ from those arising de novo, being more frequent on the trunk than the head and neck, and being associated with young age and high nevus counts [20–23]. These congruent findings point to one group of melanomas arising through a high-nevus/low-sunlight pathway (predominantly on the trunk) and another group of mainly head and neck melanomas arising through a low-nevus/high-sunlight pathway. The key questions which arise are: is there a genetic basis underlying these ‘divergent pathways’, and if so, which genes are involved?

This question has been addressed recently using two approaches: constitutional genotyping to identify those genes conferring melanoma susceptibility and genotyping of cancer cells to identify somatic mutations likely to reflect the causal origins of melanoma. At the level of constitutional genotyping, genome-wide scans have largely confirmed the associations of a clutch of promising candidate genes for melanoma susceptibility in the general population [24,25]. Genes currently accepted as being positively associated with melanoma at high-frequency (i.e., not familial melanoma genes) include MC1R, ASIP, MTAP, OCA and TYR1. Most of these genes had been identified previously as determinants of phenotypes associated with melanoma from epidemiological studies, including nevus counts, hair color and freckling. It remains to be determined whether these genes are associated with melanomas at younger or older ages, as one might predict assuming the ‘divergent pathway’ hypothesis to be true.

At the level of somatic mutations, an exciting new era of gene discovery is imminent. Mutations in three oncogenes (BRAF, NRAS and KIT) and three tumor-suppressor genes (TP53, CDKN2A and PTEN) have been consistently identified in melanomas [26]. It is generally agreed that these mutations are causal for at least some melanomas, and evidence is accumulating that the origins of the cancer might be established by observing the patterns and types of mutations. Early studies examined, for example, the determinants of p53 expression in melanoma and found that associations with key risk factors differed depending upon expression status [27,28]. Since that time, the 1799 T>A transversion in BRAF (V600E) has proven to be the most common somatic mutation in melanoma, accounting for half of all reported mutations [26]. Importantly, the prevalence of BRAF mutations has been shown to differ depending upon the anatomical site of the melanoma, host characteristics (such as nevus density and age) [29–31], levels of sun exposure in early life [32,33] and the presence or absence of contiguous neval remnants [34,35]. These observations are consistent with the divergent pathway hypothesis, but the mechanisms through which BRAF mutations arise remain unclear. Sun exposure in early life is suspected of causing BRAF mutations in melanocytes [36], but direct evidence is lacking. There is little doubt that new causal somatic mutations for melanoma will be discovered as a consequence of ‘next-generation’ sequencing. To date, only one melanoma genome has been sequenced in its entirety, with genetic material sourced from a cell-line established from a metastatic lesion, not a primary lesion [38]. Even so, the analysis provided new insights into the way in which UV radiation-specific mutations are distributed in the melanoma genome, and demonstrated the potential of the new genotyping and sequencing technologies for understanding the interplay of genes and sunlight for melanoma.

While research into the molecular origins is of intrinsic interest, arguably of greater relevance to patients is whether effective therapies can be developed by specifically targeting the somatic mutations identified. Some success has been established using nonspecific receptor tyrosine kinase inhibitors against melanomas harboring KIT mutations [37]; agents directed against mutant BRAF (including sorafenib, RAF-265 and PLX4032) are undergoing clinical trials.

Given recent advances in technology and renewed interest in the conceptual understanding of melanoma, the prospects for important and useful discoveries are encouraging. To maximize returns on research investment however, investigators will need to keep abreast of developments across all disciplines, and will need to incorporate tests of new hypotheses into their study designs. For example, stratification by the anatomical site and host nevus count would be advisable in any future genome-wide scans, since the genes determining host susceptibility to melanoma are highly likely to differ according to these characteristics. Similarly, the interpretation of somatic mutation patterns in melanomas will be most meaningful when tumors are analyzed separately according to anatomic site, histology and host phenotype.

Finally, notwithstanding the dominance of genetics in recent melanoma research, it cannot be overemphasized that much melanoma biology remains poorly understood. Little is known about the mechanisms underlying the proliferative responses of melanocytes to sunlight for example, yet this phenomenon is likely to yield extremely valuable insights for the development of melanoma. The challenge for future research will be to integrate research findings across all disciplines into a coherent causal model for melanoma.

Financial & competing interests disclosure

The author has no relevant affiliations or financial involvement with any organization or entity with a financial interest in or financial conflict with the subject matter or materials discussed in the manuscript. This includes employment, consultancies, honoraria, stock ownership or options, expert testimony, grants or patents received or pending, or royalties.

No writing assistance was utilized in the production of this manuscript.