1 INTRODUCTION

Melanocytes are melanin pigment-producing cells of neural crest embryonic origin that are normally found in the skin (integument) and inner ear, the eye and the internal organs.1 Cutaneous melanocytes are normally present scattered along the basal part of the epidermis. Non-cutaneous melanocytes are present in multiple sites of the body including the posterior and anterior regions of the uvea, the conjunctiva,2 mucosal surfaces such as the nasal and oral cavities, vulva/vagina, anus/rectum and glans penis,3 the prostate and urethra,4-6 the biliary tract,7 the cochlea and vestibular labyrinth of the inner ear, the valves and septa of the heart1 and the meninges.8 Epidemiological, clinical and morphological features as well as genomic characteristics have been used to define nine melanoma subtypes9, 10 that can be broadly separated into those that are epithelium-associated, which includes mucosal-associated, and non-epithelium-associated melanomas, which include the uveal tract and internal organs.9, 10

Large sequencing efforts have resulted in the development of computational approaches that distinguish putative driver gene mutations—mutations that directly or indirectly confer a selective growth advantage to the cell in which they occur11—from passenger mutations.12 The ability of these methods to detect driver mutations is limited by sample size and the background mutational rate of individual tumor types.13 As such, discovery of genetic drivers in non-cutaneous melanomas is directly correlated with the incidence of each disease subtype and the type and extent of genomic instability they have undergone through their evolution.14 In the USA, 91.2% of melanomas are cutaneous (incidence rate of 153.5 cases per million/year), 5.3% are ocular (6 cases per million/year), and 1.3% are mucosal (2.2 cases per million/year).15, 16 Within ocular melanomas, 85% involve the uveal tract—mostly from choroid (4.3 cases per million/year)—and 4.8% are conjunctival (0.4 cases per million/year) (Figure 1A). Within mucosal melanomas, 55.4% occur in head and neck mucosal sites (mostly from nasal cavity, oral cavity and accessory sinuses at incidence rates of 0.3, 0.2 and 0.2 cases per million/year, respectively, Figure 1B), 23.8% are anorectal (0.4 cases per million/year, Figure 1C), and 18% are female genital (1.6 cases per million/year in females, Figure 1D). Other non-cutaneous melanomas have been reported to involve a wide variety of primary sites such as biliary tract, oesophagus, meninges and adrenal glands but are extremely rare.17 Overall, the low incidence rates of non-cutaneous compared to cutaneous melanomas hinder the discovery of genetic drivers within these specific subtypes. In the case of mucosal melanomas, detection of early driver mutations is aggravated by its advanced stage at diagnosis, with 61%, 23% and 21% of anorectal, female genital tract and head and neck mucosal melanoma cases, respectively, presenting with involved lymph nodes at diagnosis.15 Furthermore, in vivo validation of computationally-identified driver gene mutations is necessary to demonstrate their ability to promote disease and to understand their functional impact in a cell- and tissue-specific manner.

Anatomical locations of non-cutaneous melanomas. (A) Non-cutaneous melanomas of the eye include those in the uveal tract—choroid, ciliary body and iris—and the conjunctiva. (B) Non-cutaneous melanomas of the head and neck include those in the oral cavity, nasal cavity, sinuses, nasopharynx and oesophagus. (C-D) Non-cutaneous melanomas in the anus, rectum, urinary tract, male and female genitalia also include those in the urethra, glans penis, vulva, vagina and cervix. Primary melanomas in internal organs are rare and have been reported in the lung, bladder and prostate. (E) Evolutionary timeline of genetic driver mutation acquisition in melanomas of the uveal tract. Short and long chromosomal arms are annotated as p and q. Copy number gains and losses are annotated as ‘+’ and ‘−’, with increasing number of ‘+’ indicating higher number copy number gains.

2 THE ROLE OF ULTRAVIOLET-INDUCED MUTATIONAL PROCESS IN AETIOLOGY OF NON-CUTANEOUS MELANOMAS

The occurrence of most non-cutaneous melanomas in sun-protected anatomical locations suggests a lack of a major pathogenic role for ultraviolet (UV) radiation9 and correlates with a lower tumor mutational burden and absence of mutational signatures associated with UV light exposure.18 Posterior uveal melanomas of the choroid and ciliary body have a much lower mutation burden than cutaneous melanomas19-24 (0.5–1.1 versus 49.2 mutations/Mb). Johansson et al23 reported the dominant presence of mutational signatures of unknown aetiology in posterior uveal melanoma. In contrast, non-cutaneous melanomas occurring in areas exposed to sunlight such as conjunctival melanomas25, 26 and anterior uveal melanomas of the iris23, 27 frequently harbour a UV signature and have a significantly higher tumor mutational burden compared to their posterior uveal tract counterparts (5 mutations/Mb in iris melanomas23 and 1.1–15.6 mutations/Mb in conjunctival melanomas25, 28). The wide anatomical distribution of mucosal melanomas and corresponding varying levels of sun exposure corresponds with the variable presence of UV-induced damage. While mucosal melanomas of lower body regions such as anorectal tract and urogenital region show no signatures of UV-induced damage, this signature has been detected infrequently in upper body mucosal melanomas of east-Asian ancestry and at a lower extent compared to cutaneous melanomas.29 Instead, a signature of spontaneous deamination of 5-methylcytosine has been found to be prevalent in mucosal melanomas, particularly in those of lower body regions,24, 29, 30 with a mutational load that resembles that of acral melanomas,31 also of minimal sunlight exposure (1.5–4.53 mutations/Mb in mucosal melanomas29, 32, 33 versus 2.6 mutations/Mb in acral melanomas24).

A recent study using a cohort of 10 patients with presumed primary pulmonary melanomas34—an extremely rare melanoma recognized as an entity by the thoracic WHO classification35—revealed a high mutational load with dominant UV signature and frequent mutations of genetic drivers characteristic of cutaneous melanomas, suggesting that the presumed primary pulmonary melanomas were probably metastases from occult or regressed primary cutaneous melanomas. It is also possible that some primary melanomas in other internal organs actually also represent metastases from unrecognized primary cutaneous melanomas.

3 GENETIC DRIVERS OF POSTERIOR UVEAL MELANOMAS

Posterior uveal melanomas are more common than other non-cutaneous melanomas, and this relatively high incidence together with its high rate of metastasis (50%) and poor survival of advanced stage patients has prompted concerted efforts to elucidate the genetic drivers and subtypes of clinical significance in this disease.

3.1 Recurrent chromosomal alterations and differential expression define Class 1 and Class 2 groups in posterior uveal melanoma

The relatively simple mutational background of uveal melanomas, characterized by mutual exclusivity of recurrent mutations and copy number alterations (CNAs) enabled identification of recurrent events. Cytogenetic analyses of small cohorts of posterior uveal melanomas36-38 revealed that a single copy (monosomy) of chromosome 3, gain of 8, loss of the long arm of chromosome 6 (the q arm) and isochromosome formation in the short arm of chromosome 6 (the p arm) and chromosome 8q were recurrent events, with monosomy of chromosome 3 and gain of 8q co-occurring frequently. Karyotype and comparative genomic hybridization (CGH) characterization of chromosome 3 status in a retrospective cohort of 54 patients showed monosomy of chromosome 3 to be strongly associated with shorter relapse-free and overall survival in posterior uveal melanomas independent of clinical and histological variables.39 Monosomy of chromosome 3 and gain of chromosome 8 in the same tumor was associated with worse survival compared to monosomy of chromosome 3 alone,40, 41 with increasing dosage of chromosome 8q associated with an even poorer survival.42 Abnormalities of chromosome 6 appear associated with better survival, even within samples with altered chromosome 3 and 8.40 Tumors with only partial loss of chromosome 3 show no association with worse disease-specific survival.41, 43 Clustering of gene expression changes44, 45 results in two classes of uveal melanoma expression profiles—denominated class 1 and class 2—the latter class including monosomy of chromosome 3 and amplification of chromosome 8 associated with down- and upregulation of genes in those chromosomes, respectively, as well as worse survival.45

3.2 Oncogenic mutations in Gαq signalling genes are prevalent in class 1 and class 2 posterior uveal melanomas

Based on a previous finding that hypermorphic germline mutations in Gnaq and Gna11 confer dermal hyperpigmentation in mice,46 van Raamsdonk et al47 identified G protein subunit alpha Q—GNAQ (9q21.2)—as an oncogene activated via somatically acquired mutations exclusively in codon Gln209 in 46% of uveal melanomas. Mutually exclusive Gln209 and Arg183 mutations on GNAQ and the paralogous gene G protein subunit alpha 11—GNA11 (19p13.3)—were subsequently identified in 83% of uveal melanomas.48 G proteins are attached to the cell surface plasma membrane and transduce signals from G protein-coupled receptors (GPCRs). The alpha subunit of G proteins (Gα) is divided into families based on sequence similarity, with GNAQ and GNA11 members of the q family (Gαq).49 G protein activity is regulated cyclically by the association of GTP with the α subunit, with a GTP-bound state linked with activation.50 Unlike class 1 and class 2 groups, Gαq mutational status and distinct driver mutations within this pathway are not associated with differences in disease-free survival.51, 52 Lack of association with clinical and pathological features, aneuploidy events and class 2 expression profile suggested that GNAQ mutations are an early oncogenic event in uveal melanomas.53

In addition to activating mutations in GNAQ/GNA11, mutations in the same signalling pathway are found in a mutually exclusive manner in codon Leu129 of cysteinyl leukotriene receptor 2—CYSLTR2 (13q14.2)20, 54 and in codon Asp630 of phospholipase C beta 4—PLCB4 (20p12) in 4% and 2.5% of tumors, respectively.21 Activation of Gαq signalling has been shown to result in downstream activation of MAPK, JNK/p38, YAP and β-catenin pathways47, 48, 55-59; the distinct role of these pathways in uveal melanoma pathogenesis is still being elucidated. Interestingly, mutations in these genes have also been characteristic of blue naevi, a subgroup of melanocytic tumors that share many morphological similarities with uveal melanomas,60-62 as well as primary leptomeningeal melanocytic tumors.63

3.3 Class 2 posterior uveal melanomas carry BAP1 inactivation

Whole-exome sequencing and targeted Sanger sequencing of 31 class 2 uveal primary melanomas64 revealed that 84% of the patients in this group and ~40% of uveal melanomas overall carried deleterious mutations in the BRCA1-associated protein-1—BAP1—a gene located in chromosome 3p21.1. Concordant with a two-hit inactivation mechanism of tumorigenesis and typical of tumor suppressor genes,65 deleterious mutations in BAP1 co-occurred with chromosome 3 monosomy in all class 2 patients with cytogenetic data available.64 In one patient, a frameshift mutation in this gene was found in the germline DNA; BAP1 germline mutations predispose to a number of tumors including benign melanocytic tumors, mesothelioma and uveal melanoma, in a condition called the BAP1 cancer syndrome.66-68 BAP1 encodes a deubiquitinating enzyme of tumor-suppressing activity.69, 70 BAP1 plays several cancer-protective roles in nuclear and cytoplasmic processes.68 Among those roles, BAP1 ensures genome stability via recruitment of the INO80 chromatin remodelling complex to stalled forks during replication stress71 and participates in DNA double-strand break repair by homologous recombination.72 BAP1 has been shown to repress expression of Solute Carrier Family 7 Member 11—SLC7A11—resulting in inhibition of cysteine uptake leading to lipid peroxidation accumulation and ferroptosis,73 a metabolic stress-induced non-apoptotic form of cell death.74 Additionally, BAP1 localizes to the endoplasmic reticulum and modulates calcium release to the cytosol and mitochondria via binding, deubiquitylation and stabilization of type 3 inositol-1,4,5-trisphosphate receptor—IP3R3—promoting apoptosis.75

3.4 Class 1 posterior uveal melanomas carry recurrent mutations in SF3B1 and EIF1AX

Patients with class 1 uveal melanomas lack inactivation of BAP1, present disomy of chromosome 3 and have better survival than class 2 patients; gain of chromosome 6p is associated with class 1.45 Within class 1, splicing factor 3b subunit 1—SF3B1 (2q33.1)—and eukaryotic translation initiation factor 1A X-linked—EIF1AX (Xp22.12)—are mutually exclusive and recurrently mutated genes. Mutations in SF3B1 were first reported in a cohort of 102 primary uveal melanomas,76 being exclusively somatic and occurring predominantly at the Arg625 codon in the HEAT repeat domain. SF3B1 mutations were present in 27% of class 1, 7% of class 2 uveal melanomas and overall in 19% of uveal melanomas. A subsequent study77 based on whole-exome sequencing of 22 uveal melanomas and targeted sequencing of an additional 66 tumors confirmed the mutational status of SF3B1 and revealed EIF1AX mutations in 48% of tumors with disomy of chromosome 3. In agreement with the good prognosis observed in patients with partial monosomy of chromosome 3,41, 43, 77 54% and 7% of tumors with this specific alteration had mutations in SF3B1 and EIF1AX, respectively.

SF3B1 encodes a subunit of the splicing factor SF3B complex, required for binding the U2 small nuclear ribonucleoprotein complex (U2 snRNP) to the branch point upstream of the 3′ splice site; hotspot mutations in the C-terminal HEAT domains of this protein78 cause aberrant 3′ splice selection and aberrant splicing. Mutated SF3B1 has been linked to aberrant splicing and repression of the Bromodomain Containing 9—BRD9—a component of the non-canonical BAF chromatin remodelling complex and of demonstrated tumor-suppressing activity in uveal melanoma79—and to stabilize MYC and impair apoptosis via promoting decay of transcripts encoding the protein phosphatase 2A (PP2A) subunit PPP2R5A.80 Mutations in additional splicing factors have been detected in class 1 tumors in a mutually exclusive manner with SF3B1, suggesting alternative mechanisms of spliceosome disruption.22

The protein encoded by EIF1AX—eIF1A—is a component of the preassembled 43S preinitiation complex, a ribonucleprotein complex recruited to the 5′ end of mRNAs to initiate translation via scanning in a 5′ to 3′ direction for correct positioning of a start codon in the P site of the 40S ribosomal subunit.81 The C-terminal and N-terminal tails of eIF1A regulate start codon selection in opposing manners: while the C-terminal tail promotes continued scanning at non-AUG codons, the N-terminal tail (NTT) arrests scanning and promotes eIF1 release at AUG codons.82, 83 Mutations in EIF1AX occur in the first 15 amino acids within the NTT and might result in reduced rates of bulk translation and altered start site recognition.77

3.5 Evolution of posterior uveal tract melanomas is characterized by early acquisition of Gαq signalling mutations and subsequent bi-allelic BAP1 inactivation, SF3B1/EIF1AX mutations and gain of 8q.

The studies that unveiled main driver mutations and CNAs in posterior uveal melanoma formed the basis for subsequent efforts aimed at discovering additional genetic drivers and those seeking to determine the order of genetic event acquisition within primary tumors and from primary to metastatic disease (Figure 1E).21, 22, 84, 85 In the uveal melanoma The Cancer Genome Atlas (TCGA) cohort,21 BAP1 somatic mutations were found to have an overall lower cancer cell fraction than the co-occurring loss of heterozygosity (LOH) event in chromosome 3, indicative of the LOH event preceding mutation of BAP1 during progression.21 The study by Field et al22 collated 92 primary uveal melanomas and identified mutations in Gαq genes to be predominantly clonal, indicating early occurrence in tumor evolution. BAP1/SF3B1/EIF1AX mutations as well as monosomy of chromosome 3, gain of 6p and 8q occurred in 100% of the cells within most tumors, supporting a scenario of punctuated emergence of these events followed by selectively neutral mutations, as identified in a subset of 12 tumors with available whole-genome sequencing. In a minority of cases, these mutations and alterations were found in subclones, including BAP1 mutations and most alternative spliceosome mutations to those in SF3B1.

Assessment of matched primary-metastasis tumors has further helped elucidate the cascade of events in posterior uveal melanomas by identifying those ubiquitous across lesions versus events shared and private to specific lesions. Shain et al85 pursued multiregional sampling in a cohort of predominantly class 2 primary tumors with matched liver metastases followed by targeted sequencing of a 500-gene panel at high coverage across 35 patients. In this study, in addition to the canonical Gαq and BAP1/SF3B1/EIF1AX mutations occurring early during tumor progression, additional subclonal oncogenic mutations impacting CDKN2A, MED12 and genes of the PI3 K pathway were identified—particularly in tumors with SF3B1 and EIF1AX mutations—as well as mutations in chromatin remodelling factors PBRM1 and EZH2 occurring after acquisition of BAP1 mutations in disease progression. While commonly acquired early during disease progression, inactivating BAP1 mutations and chromosome 8q gains were observed after metastatic dissemination in some cases. Based on the extent of the presence of truncal versus branchial events across evolutionary trees, Shain et al concluded that Gαq pathway mutations occur earliest, followed by BAP1/SF3B1/EIF1AX events and gain of chromosome 8q. In contrast, loss of 6q, gain of 1q and higher gains of 8q were significantly enriched in metastases. Other CNAs occurred at intermediate points through disease progression. Rodrigues et al84 sequenced 91 primary and metastatic uveal melanomas from 25 patients, with 15 patients having matched primary-metastasis. Early occurrence of Gαq and BAP1/SF3B1/EIF1AX mutations as well as recurrent loss of 6q, gain of 1q and higher gains of 8q towards metastasis were also observed in this study. They showed that two patients with hypermutated tumors had inactivation of methyl-CpG binding domain 4, DNA glycosylase—MBD4 (3q21.3)—a gene responsible for integrity of methyl-CpG sites.86 This hypermutated phenotype resulted in a mutational burden similar to that of cutaneous melanomas and was associated with higher heterogeneity and putative additional driver mutations in CDKN2A, GNAS, TP53 and SMARCA484 acquired through disease progression. These studies provide evidence for ongoing acquisition of driver mutations in advanced disease.

In addition to the aforementioned studies, several others have further characterized the molecular landscape of posterior uveal tract melanomas. Robertson et al21 performed a comprehensive multi-platform assessment of 80 primary uveal melanomas as part of the TCGA project. In this study, SF3B1 and SRSF2 mutants were associated with differential splicing and expression of several genes, including initiation factors. Distinct global methylation patterns were associated with chromosome 3 status—also reported by Field et al22—with disomy of chromosome 3 presenting distinct CNA patterns depending on SF3B1/SRSF2 versus EIF1AX mutation status and with tumors with monosomy of chromosome 3 presenting distinct subsets both transcriptionally and also in terms of CNAs. Johansson et al23 utilized whole-genome sequencing of 103 uveal melanomas—mostly primaries—of which eight originated from the iris. This study identified centromere protein E—CENPE—TP53, and ribosomal protein L5—RPL5 (1p22.1)—as statistically significantly mutated genes. CENPE is a microtubule-based kinetochore motor protein involved in centromere-microtubule interaction87; intriguingly, tumors with mutated CENPE were found to have a significantly higher extent of CNAs.23 Recurrent mutations of RPL5 and TP53 together with 8q gain—where MYC resides—led the authors to hypothesize an association of these genes via the impaired ribosome biogenesis checkpoint, and accordingly, found that mutations in these genes were associated with gain of 8q. The study by Durante et al88 increased the resolution of the genomic landscape via single-cell sequencing of 8 primary and 3 metastatic uveal melanomas. This study provides further evidence for ongoing subclonal acquisition of CNAs as well as co-occurrence of predominantly class 1 and class 2 tumors with subclones carrying events representative of the alternative class.88

4 GENETIC DRIVERS OF ANTERIOR UVEAL AND CONJUNCTIVAL MELANOMAS

Uveal tract melanomas correspond to 85% of ocular melanomas.15 Iris anterior uveal melanomas represent only 4% of those in the uveal tract,89 whereas conjunctival melanomas—those in the conjunctiva, the mucosal membrane that covers the eye—correspond to only 4.8% of all ocular melanomas.15 The lower incidence of iris and conjunctival melanomas presents a challenge for identification of genetic drivers in these tumor types. The recurrent events identified in posterior uveal melanomas and the suspected role of UV radiation in iris and conjunctival melanomas guided early efforts seeking to identify genetic driver mutations in these tumors.90-96

4.1 Genetic drivers identified in iris melanomas more closely resemble those in posterior uveal tract melanomas than cutaneous melanomas

Identification of BRAF mutations in iris melanomas has been discordant between studies. Henriquez et al identified recurrent Val600 BRAF mutations in 47% of iris melanomas,97 whereas Scholz et al did not identify any mutations in BRAF, NRAS or KIT,98 instead identifying mutually exclusive mutations of GNAQ/GNA11 as well as frequent mutation in EIF1AX in 84% and 42% of iris melanomas98 (Table S1). Van Poppelen et al confirmed this pattern, identifying 77% and 17% of iris melanomas carrying GNAQ/GNA11 and EIF1AX mutations, respectively. They also detected mutations in SF3B1 (3%) and BAP1 (43%), with BAP1 expression not being associated with disease-free survival.99 BRAF, NRAS and KIT mutations were observed mostly co-occurring with oncogenic mutations typical of uveal melanomas and in non-Val600 (for BRAF) and non-Gly12/Gly13/Gln61 (for NRAS) residues, suggesting the possibility of these being passenger mutations.99, 100 Nevertheless, KIT mutations were found in exon 11 in 33% of iris melanomas and 9% of posterior uveal melanomas.101 In iris melanomas, partial or complete loss of chromosome 3 (23–71%), gain of 6p (14%) and 8q (4–35%) and loss of 9p (35%) were observed99, 102-104 (Table S2) as well as expression profiles characteristic of class 1 and class 2 tumors.105

4.2 Genetic drivers identified in conjunctival melanomas more closely resemble those in cutaneous melanomas than uveal tract melanomas

Driver mutations commonly identified in cutaneous melanoma have been found in their conjunctival counterpart to a greater extent than in iris melanomas. In contrast, driver mutations common in posterior uveal melanomas are absent in conjunctival melanomas, with several studies reporting no presence of GNAQ/GNA11 mutations47, 48, 106, 107 or other mutations characteristic of uveal melanomas26 (Table S1). BRAF mutations have been identified in 8%–60% of conjunctival melanomas26, 122 and their presence associated with sun exposure.113, 117 NRAS mutations have been found in 11%–37.5% of conjunctival melanomas,26, 28, 107, 113, 114, 120 NF1 mutations in 20%–37.5%,26, 28, 107, 122 KIT mutations in 7.7%–11%,111, 116 KRAS mutations in 1.5%107 and TERT promoter mutations in 32%–41%.114, 123 The latter was also found in primary acquired melanosis (a common precursor lesions in conjunctival melanoma) and rarely found in large cohorts of posterior uveal melanomas.114, 123, 124 Recurrent somatic mutations were reported in the epigenetic regulator ATRX in 62.5% of cases in one study,28 also identified in one of five patients by Swaminathan et al.26 Small cohort size in most of the studies reported to date has resulted in large variability in the range of proportions and rates reported for each mutation. For example, Populo et al found no BRAF or NRAS mutations in 6 conjunctival melanomas125 and no mutations were identified in KIT by Wallander et al, Del Prete et al and Mikkelsen et al in 5, 2 and 12 conjunctival melanomas, respectively.101, 122, 126 However, some mutations may require a much larger cohort size for identification. Griewank et al reported one of the largest conjunctival melanoma cohorts published to date (n = 78) and did not identify any KIT hotspot mutations in 42 conjunctival melanomas,113 but identified amplification at the KIT locus in 6 of 30 conjunctival melanomas. Similarly, Scholz et al identified one driver mutation in KRAS in a cohort of 63 conjunctival melanomas.107 Amplification of CCND1 and loss of CDKN2A have also been observed in individual patients.28, 113 In a study of 16 primary and 6 metastatic conjunctival melanomas, frequent amplification of RUNX2 and CDKN1A were observed in primaries—both in 6p21—whereas amplification of MLH1, TIMP2 and deletion of MGMT and ECHS1 were observed in metastases.112

Griewank et al (N = 78)113 and Kenawy et al (N = 59)120 independently identified recurrent amplifications of 1q, 6p, 7, 8q, 12p and 17q and recurrent deletions of 3q, 6q, 8p, 9p, 10, 11q, 12q and 16q (Table S2). Furthermore, Kenawy et al identified amplification of 6p21-25 in 76% of conjunctival melanomas, deletion of ASNS (7q21.3) and amplification of the histone gene cluster-1 (6p22.2) in 76% and 61% of cases, respectively.120

The availability of large cohorts unveiled mutational subtypes107, 113 that show equivalence to those in cutaneous melanomas based on mutually exclusive mutational status of BRAF, NRAS and NF1127 and that are associated with recurrent CNAs. Griewank et al reported increased extent of CNAs in wild-type conjunctival melanomas (as defined by no BRAF or NRAS mutant in that study), gains of 1q, 3p and 17q were less prominent in BRAF mutants, and loss of chromosome 10 and PTEN (10q23) were more common in BRAF mutant tumors.113 Kenawy et al identified regional amplification of 17q associated with NRAS mutants and deletion of 10q11-23 (PTEN) and 10q26 (DMBT1) associated with BRAF mutants120; deletion of 10q24.32-10q26.2—which includes the tumor suppressor genes NEURL1, SUFU, PDCD4 and C10orf90—was associated with lymphatic invasion, increased tumor thickness and reduced metastasis-free survival.120

A few studies have analysed the differential expression of conjunctival melanomas and have identified frequent activation of MAPK and mTOR pathways,119, 125 with decreased PTEN expression compared to uveal melanomas125 and no increase in rate of BRAF mutations compared to nevi.119 Efforts evaluating differential expression of microRNAs in conjunctival melanomas compared to normal conjunctival tissue found upregulation of hsa-miR-181b-5p (chr1q32.1) and hsa-miR-509-3p (chrXq27.3)—the latter also identified as differentially expressed in cutaneous melanoma in two other studies—and associated with Hippo pathway and p53 signaling.118,