Conjunctival malignant melanoma in Denmark: epidemiology, treatment and prognosis with special emphasis on tumorigenesis and genetic profile

Anatomy and function of the conjunctiva

The conjunctiva is derived from ectoderm during embryogenesis, and it can be distinguished at about the tenth week of gestation (Seregard et al. 2015). It is a thin, clear mucous membrane that covers the anterior surface of the eye (the bulbar conjunctiva) except the cornea (Fig. 1) and covers the inside of the eyelids (the palpebral conjunctiva), and forms the superior and inferior fornices. A loose fold of bulbar conjunctiva (the plica semilunaris), which is a remnant of the posterior surface of the nictitating membrane in animals, is situated in the nasal canthus of the eye. The lacrimal caruncle is situated nasally to the plica semilunaris. The caruncle is a remnant of the anterior surface of the nictitating membrane, is covered by modified skin lined with conjunctival epithelium, and contains hair follicles, sebaceous glands, sweat glands and accessory lacrimal tissue (Seregard 1998; Levy et al. 2009; American Academy of Ophthalmology and Levine 2015; Seregard et al. 2015). The function of the conjunctiva is to provide a smooth, frictionless surface that allows free eye movement. It protects the eye from infection and contributes substantially to tear film stability and corneal transparency (Knop & Knop 2000; Seregard et al. 2015). The conjunctiva is innervated by the ophthalmic division of the trigeminal nerve (V1). The arterial supply is provided by the anterior ciliary arteries, and the peripheral and marginal arcades of the eyelids. Venous drainage of the conjunctiva occurs through anterior ciliary, superior and inferior ophthalmic veins. The conjunctiva drains to preauricular/parotid, submandibular and deeper cervical lymph nodes (Seregard 1998; American Academy of Ophthalmology and Levine 2015).

Microscopically, the conjunctiva is composed of a conjunctival epithelium and a stroma. The epithelium consists of two or more layers of non-keratinized stratified columnar or squamous epithelium (Fig. 2), interspersed with mucin-producing goblet cells. The conjunctival stroma is divided into a superficial lymphoid layer and a deeper, richly vascularized fibrous layer (Folberg et al. 1989; Seregard 1998; Seregard et al. 2015). Conjunctival melanocytes reside in the basal layers of the epithelium and in the substantia propria (stroma; Folberg et al. 1989; Seregard 1998).

Epidemiology

The most common location of melanoma is the skin, accounting for approximately 90% of cases. Less frequently, melanoma develops in the ocular region or in mucosal membranes (Chang et al. 1998). Ocular melanoma accounts for 5% of cases and is the second most common site of malignant melanoma. Uveal melanoma accounts for the majority of these cases, whereas 5% of ocular melanomas are of conjunctival origin (Østerlind 1987; Chang et al. 1998; Seregard 1998; Brownstein 2004). Mucosal melanomas represent 1.3% of all melanomas and most frequently involve the head and neck region (Chang et al. 1998). The incidence of CM is 0.2–0.8 per million in Caucasian populations (Seregard & Kock 1992; Nørregaard et al. 1996; Tuomaala et al. 2002; Yu et al. 2003; Isager et al. 2005; Missotten et al. 2005; Triay et al. 2009). Taking into account the small surface area of the conjunctiva, the prevalence of melanoma is relatively high compared to the skin and other mucosal membranes. Whereas the incidence rate of CM in Denmark during the period 1943–1997 was reported to be stable (Isager et al. 2005), an increase in incidence analogous to cutaneous melanoma has been reported in newer studies from Sweden and Finland (Tuomaala et al. 2002; Triay et al. 2009).

Conjunctival melanoma occurs mainly in Caucasian populations, and is rarely seen in blacks or Asians (Hu et al. 2008). No clear gender predilection has been established, but in some studies males have tended to be younger than females at the time of diagnosis (Nørregaard et al. 1996; Tuomaala et al. 2002; Missotten et al. 2005; Triay et al. 2009; Shields et al. 2011). Conjunctival melanoma is primarily a disease of the middle-aged, with the mean age ranging from 55 to 65 years, and it is rarely reported in children (Nørregaard et al. 1996; Tuomaala et al. 2002; Missotten et al. 2005; Taban & Traboulsi 2007; Hu et al. 2008; Triay et al. 2009; Shields et al. 2011).

Aetiology and pathogenesis

Lesions predisposing to the development of CM have been reported with variable frequency. Primary acquired melanosis (Fig. 3) is the most common precursor lesion and is reported in 42–75% of cases, mainly affecting middle-aged Caucasian patients, but may occur regardless of skin pigmentation (Jakobiec et al. 1989b; Anastassiou et al. 2002; Tuomaala et al. 2002; Missotten et al. 2005; Shields et al. 2008; Shields & Shields 2009). The clinical appearance of PAM is usually described as unilateral, flat and variably brown, with patches of pigmentation occurring in any part of the conjunctiva with or without extension to eyelid skin or cornea (Jakobiec et al. 1989b; Seregard 1998; Shields & Shields 2009). Primary acquired melanosis is caused by a neoplastic proliferation of epithelial melanocytes, and by using histological criteria PAM may be designated with or without atypia (PAM+ or PAM−, respectively). As opposed to PAM−, which rarely progresses into melanoma, PAM+ appears to progress into melanoma much more frequently, particularly if there is vertical invasion of epithelium by atypical melanocytes, pagetoid spread or epithelioid cytology (PAM with severe atypia; Folberg et al. 1985b; Jakobiec et al. 1989b; Seregard 1998; Shields et al. 2008; Shields & Shields 2009). The grading of PAM has been a topic of much debate, since the term melanosis includes both benign and premalignant melanocytic lesions. As a consequence, the more precise histological terms: ‘conjunctival melanocytic intraepithelial neoplasia’ (C-MIN) and ‘hypermelanosis’ have been proposed together with a new, improved grading system (Damato & Coupland 2008). Furthermore, the term PAM has been proposed to be reserved for the clinical description of the CM.

Conjunctival nevi (Fig. 3) rarely give rise to CM (Shields et al. 2004), but 2–40% of CMs have been described to arise in a nevus, and 5–8% arise in a combination of a nevus and PAM (Folberg et al. 1985a, 1989; Anastassiou et al. 2002; Tuomaala et al. 2002; Missotten et al. 2005; Shields et al. 2011). Conjunctival nevi are typically formed in the first decade of life and most commonly consist of nests of nevus cells residing in the junctional zone (junctional nevi). The nevus cells subsequently descend into the substantia propria (compound and subepithelial nevi), dragging with them surface epithelium and goblet cells, which may account for the cyst formation commonly seen in these lesions (Folberg et al. 1989). Conjunctival melanoma has been described to mainly arise from junctional or compound nevi, both of which have epithelial components (Folberg et al. 1989; Shields et al. 2004).

In 16–26% of cases, the CM arises without any apparent predisposing lesion (de novo), that is the CM presents with no signs of either associated PAM or nevi by histopathological examination (Seregard 1998; Anastassiou et al. 2002; Missotten et al. 2005; Shields et al. 2011).

The conjunctiva is the only mucosal membrane in the body that is naturally exposed to sunlight. This characteristic, together with the close anatomical relationship with both skin and sinonasal mucosa, suggests that CM may share pathogenic risk factors with these subtypes of melanoma. However, the exogenous risk factors in CM are mostly unknown. In cutaneous melanoma, intermittent exposure to sun (as opposed to chronic sun exposure) and sunburn history have been determined as risk factors for melanoma development (Gandini et al. 2005). There is some evidence that sun exposure may also be a risk factor in CM–mainly, due to an increased incidence of CM in sun-exposed conjunctival locations (Triay et al. 2009) and identification of BRAF and NRAS mutations, which are associated with sun exposure in cutaneous melanoma (El-Shabrawi et al. 1999; Maldonado et al. 2003; Gear et al. 2004; Spendlove et al. 2004; Goldenberg-Cohen et al. 2005; Beadling et al. 2008; Populo et al. 2010; Lake et al. 2011; Griewank et al. 2013a; Sheng et al. 2015).

Clinical characteristics

Conjunctival melanoma patients usually present with a growing pigmented nodular lesion of the perilimbal conjunctiva, lacking mobility in relation to the sclera, or with a subtle thickening of a PAM, but CM may occur in all conjunctival locations with or without PAM (Jakobiec et al. 1989b; Paridaens et al. 1994b; Seregard 1998; Anastassiou et al. 2002; Tuomaala et al. 2002; Werschnik & Lommatzsch 2002; Brownstein 2004; Missotten et al. 2005; Shields & Shields 2009; Triay et al. 2009; Fig. 4). In the majority of cases (56–79%), the CM is confined to the epibulbar conjunctiva (the corneal limbus or bulbus). Less frequently, the CM involves extrabulbar conjunctiva (the fornix and palpebral conjunctiva; in 9–29% of cases), or the caruncle (in 1–7% of cases; Paridaens et al. 1994b; Seregard 1998; Shields 2000; Anastassiou et al. 2002; Werschnik & Lommatzsch 2002; Missotten et al. 2005; Tuomaala et al. 2007; Triay et al. 2009). Benign melanocytic lesions in fornical or palpebral conjunctiva are rare, and malignancy should always be suspected in these cases (Buckman et al. 1988; Jakobiec et al. 1989b; Seregard 1998). Conjunctival melanoma grows in a nodular pattern in most cases (48–63%), and the tumour is most commonly pigmented rather than having a mixed or nonpigmented appearance (Jakobiec et al. 1989b; Seregard 1998; Shields 2000; Anastassiou et al. 2002; Shields et al. 2011). Local invasion to the cornea, adjacent skin, the nasolacrimal system, the sinuses, or – rarely – to the orbit may occur (Paridaens et al. 1994b; Seregard 1998; Shields 2000; Anastassiou et al. 2002; Werschnik & Lommatzsch 2002; Shields et al. 2011). Multiple lesions have been reported, and these are usually associated with PAM (Anastassiou et al. 2002; Shields et al. 2011).

Histopathological characteristics

The diagnosis of CM is made after histopathological evaluation of the lesion. Conjunctival melanoma is characterized by invasion by melanoma cells from the epithelium into the substantia propria (Folberg et al. 1985a; Jakobiec et al. 1989b). It has been found that 75% of CMs are associated with PAM and 20% are associated with a preexisting nevus with or without PAM (Jakobiec et al. 1989b; Seregard 1998). The melanoma cells may be small polyhedral cells, epithelioid cells, spindle-shaped cells, or balloon cells, usually in a variable mixture (Folberg et al. 1985a; Jakobiec et al. 1989b; Fig. 5). The cells may grow in nests or in a sheet-like fashion (Jakobiec et al. 1989b; Seregard 1998).

Benign conjunctival nevi may be difficult to differentiate from CM (Folberg et al. 1985a; Jakobiec et al. 1989b; Seregard 1998), in particular, a junctional nevus may be indistinguishable from PAM+ (Folberg et al. 1989). Patient age >40 years, evidence of an intraepithelial component that extends horizontally, pagetoid spread (upward-spreading of melanocytes into the epithelium) and mitotic activity are features associated with CM and can be used to differentiate a CM from a benign conjunctival nevus (Jakobiec et al. 1989b; Seregard 1998). Immunohistochemical studies detecting melan-A, S-100 and HMB-45 are also helpful in the diagnosis of CM, and may help to distinguish amelanotic CM from conjunctival nevi or squamous cell carcinoma (Jakobiec et al. 1989b, 2010; Seregard 1998; Heegaard et al. 2000). Uveal melanomas presenting with extraocular extension or cutaneous melanoma metastasis to the conjunctiva are rare differential diagnoses that are important to exclude (Jakobiec et al. 1989a; Seregard 1998).

Treatment

Because of the rarity of this disease, there is no consensus regarding the optimal management of CM, so treatment has mainly evolved from physician experience and the findings from case series (Shields et al. 1997; Seregard 1998; Shields 2000; Damato & Coupland 2009b; Shields & Shields 2009). Previously, many patients were treated with orbital exenteration (Paridaens et al. 1994a). Exenteration is currently reserved for extensive disease with intraocular or orbital invasion, but systemic metastases may already have occurred in these patients (Paridaens et al. 1994a). Incisional biopsies are discouraged, since these appear to be associated with a higher frequency of local recurrence (Shields 2000; Damato & Coupland 2009a). Current treatment modalities include surgical excision with a 3- to 5-mm free conjunctival margin, alcohol corneal epithelialectomy in cases where the tumour extends to the corneal limbus, and a ‘no touch’ technique in which swabs are not reused, touching of the tumour is avoided and instruments are changed (Shields et al. 1997; Seregard 1998; Shields 2000; Damato & Coupland 2009b; Shields & Shields 2009; Shildkrot & Wilson 2010; Kenawy et al. 2013). Excision in combination with cryotherapy, adjunctive brachytherapy (Fig. 6) and/or topical chemotherapy (mytomycin C) or interferon alfa-2b appears to be effective in eradicating most lesions (De Potter et al. 1993; Shields 2000; Anastassiou et al. 2002; Shields et al. 2002; Tuomaala et al. 2002; Missotten et al. 2005; Pe'er & Frucht-Pery 2005; Finger et al. 2008; Damato & Coupland 2009b; Shildkrot & Wilson 2010; Kenawy et al. 2013; Sheng et al. 2015). Excisions without adjuvant therapy are discouraged, since correlations with local recurrence and CM-related mortality have been reported (Shields 2000; Werschnik & Lommatzsch 2002; Missotten et al. 2005). Sentinel lymph node biopsies (SLNBs) are currently recommended in CM cases with a high risk of regional metastasis, that is tumours that are more than 1–2 mm thick, that have palpebral or non-limbal location, and that have histological ulceration (Tuomaala & Kivelä 2004, 2008; Tuomaala et al. 2007; Esmaeli 2008; Damato & Coupland 2009b; Savar et al. 2011; Esmaeli et al. 2012; Maalouf et al. 2012). The findings of a positive SLNB and consequently regional lymphadenectomy may have a favourable effect on prognosis (Cohen et al. 2013).

Prognosis and prognostic features

Local recurrence is very common in CM, occurring in approximately 50% of cases, and it is associated with a poor prognosis (Shields 2000; Anastassiou et al. 2002; Tuomaala et al. 2002; Missotten et al. 2005; Shields et al. 2011). Unlike uveal melanomas, which mainly spread haematogenously to the liver, CM has metastatic potential, both locally and systemically, and metastasis occurs in 16–32% of cases (De Potter et al. 1993; Seregard 1998; Shields 2000; Esmaeli et al. 2001; Anastassiou et al. 2002; Tuomaala et al. 2002; Werschnik & Lommatzsch 2002; Tuomaala & Kivelä 2004; Missotten et al. 2005). Conjunctival melanoma may spread initially via the lymphatics to ipsilateral lymph nodes (preauricular, submandibular and parotid or cervical) and eventually develop distant metastases involving the skin, adrenals, brain, lungs, heart, peritoneum, bowels, pancreas, kidneys, bones and spleen (De Potter et al. 1993; Seregard 1998; Shields 2000; Esmaeli et al. 2001; Werschnik & Lommatzsch 2002; Missotten et al. 2005). However, distant metastases without previous lymph node metastasis are not uncommon (Seregard 1998; Esmaeli et al. 2001; Werschnik & Lommatzsch 2002; Tuomaala & Kivelä 2004). In studies with long-term follow-up, the melanoma-related mortality of CM has been reported to be approximately 25–30% at 10 years (Folberg et al. 1985a; Fuchs et al. 1989; Paridaens et al. 1994b; Nørregaard et al. 1996; Seregard 1998; Shields 2000; Anastassiou et al. 2002; Tuomaala et al. 2002; Werschnik & Lommatzsch 2002; Isager et al. 2005).

Only a few multicentre and population-based prognostic studies of CM have been performed. Most published data are based on case reports or series of referred patients, the results may have been skewed by unrecognized bias, and there have been few multivariate analyses. In light of these shortcomings, clinicopathological features with a prognostic association are outlined here.

Prognostic features in CM have mainly concerned tumour location and extension. An association with local tumour recurrence, metastasis and melanoma-related death has been reported in CMs with an ‘unfavourable location’, described as ‘non-epibulbar tumours’, ‘non-limbal location’ or ‘location not touching the limbus’ – that is, tumours located in the palpebral conjunctiva, the fornix, the caruncle and/or plica semilunaris, the eyelid margin, or those involving adjacent tissue structures such as eyelid skin (Fuchs et al. 1989; Paridaens et al. 1994b; Shields 2000; Anastassiou et al. 2002; Tuomaala et al. 2002, 2007; Werschnik & Lommatzsch 2002; Missotten et al. 2005; Shields et al. 2011). Medial tumour location has also been associated with a higher risk of local recurrence (Damato & Coupland 2009a). Caruncular location appears to show contradictory results, regarding prognosis, when other unfavourable locations are excluded (Tuomaala et al. 2002; Damato & Coupland 2009a). Conjunctival melanomas extending more than one quadrant, with a diameter exceeding 10 mm, and particularly with increasing tumour thickness, have been associated with regional and distant metastasis and CM-related death (Folberg et al. 1985a; Fuchs et al. 1989; Lommatzsch et al. 1990; Seregard & Kock 1992; Paridaens et al. 1994b; Tuomaala et al. 2002; Missotten et al. 2005; Savar et al. 2011; Esmaeli et al. 2012). Whereas tumour thickness of >0.8 mm has been associated with the risk of metastasis, tumour thickness exceeding 2 mm also appears to be associated with higher mortality (Folberg et al. 1985a; Fuchs et al. 1989; Lommatzsch et al. 1990; Tuomaala et al. 2002). Other poor prognostic features include: multifocal tumour presentation (particularly if the CM is located in a favourable location; Fuchs et al. 1989; De Potter et al. 1993; Paridaens et al. 1994b), nodular appearance rather than superficial appearance of the tumour (Anastassiou et al. 2002; Shields et al. 2011), patient age <55 years (Werschnik & Lommatzsch 2002) and local tumour recurrence (De Potter et al. 1993; Shields 2000; Tuomaala et al. 2002, 2007). Histopathologically, tumour excisions without clear margins have been reported to be associated with local recurrence, CM metastasis and CM-related death (Shields 2000; Anastassiou et al. 2002; Shields et al. 2011). A high degree of epithelioid and mixed cell types, especially in CMs occurring in favourable locations (corneo-limbal and bulbar conjunctiva) may correlate with an increasing risk of mortality (Paridaens et al. 1994b). A high mitotic count/index and presence of histological ulceration also appear to correlate with metastasis and CM-related death (Folberg et al. 1985a; Seregard & Kock 1992; Paridaens et al. 1994b; Anastassiou et al. 2002; Tuomaala et al. 2007; Savar et al. 2011; Esmaeli et al. 2012). An increasing number of tumour-infiltrating lymphocytes may also be associated with a higher mortality (Paridaens et al. 1994b; Tuomaala et al. 2007; Esmaeli et al. 2012). Different prognostic associations of CM origin have been reported over the years, and may be a consequence of changing classifications and nomenclature (Liesegang & Campbell 1980; Folberg et al. 1985a; Anastassiou et al. 2002; Shields et al. 2011). However, a large study of 382 cases found that de novo origin was associated with CM metastasis and CM-related death (Shields et al. 2011).

Staging

Conjunctival melanoma may be staged according to the seventh edition of the American Joint Committee on Cancer (AJCC) Tumour, Lymph node and Metastasis (TNM) classification (Edge et al. 2009). This classification covers the location (bulbar, non-bulbar or caruncular) and extension (quadrants, local invasion) of the primary tumour (T), the presence of regional lymph node metastases (N) and the presence of distant metastases (M; Table 1). Recent studies have shown a correlation between the latest TNM classification and survival – concerning local recurrence in particular (Shields et al. 2012; Yousef & Finger 2012; Kivelä & Kujala 2013).

The histopathological classification (Table 2) is based on tumour thickness and invasion of the substantia propria. Furthermore, the designation ‘conjunctival melanoma in situ’ is included. No other histopathological or genetic prognostic features have been reported with any consistency, and they are therefore not included in the current edition (Kivelä & Kujala 2013). Further studies are needed to relate clinicopathological features and TNM staging to long-term follow-up data.

Oncogenic mutations in melanoma

Cancer is a result of multiple genetic aberrations mainly occurring through activation of proto-oncogenes and/or inactivation of tumour suppressor genes, which in normal cells control functions such as transcriptional regulation, signal transduction and growth (Weinberg 1994).

One of the main regulatory pathways involved in melanoma development is the mitogen-activated protein kinase (Ras-Raf-MEK-ERK or MAPK) pathway (Spendlove et al. 2004; Dahl & Guldberg 2007). Activation of the MAPK pathway (Fig. 7) in melanoma most commonly occurs through activating mutations in the BRAF-, NRAS- or KIT gene (Dahl & Guldberg 2007). The BRAF gene encodes a serine-threonine-specific protein kinase which activates MEK. Mutations in BRAF are common in cutaneous melanoma, with a reported frequency of approximately 40% (Platz et al. 2008). In contrast, BRAF and NRAS mutations are very rare in uveal melanoma, and oncogenic mutations of GNAQ appear to represent an alternative route to MAPK activation (Cohen et al. 2003; Cruz et al. 2003; Rimoldi et al. 2003; Onken et al. 2008; Van Raamsdonk et al. 2009). The majority of BRAF mutations in cutaneous melanoma (80%) are caused by a substitution of thymine with adenine at nucleotide 1799, leading to a substitution of valine (V) with glutamic acid (E) at residue 600 (V600E; Dahl & Guldberg 2007). BRAF V600K (valine to lysine) mutations account for approximately 20% of cases of BRAF mutations, whereas other types of BRAF mutations are rare in cutaneous melanoma (Menzies et al. 2012).

BRAF V600E mutations have been identified in up to 80% of cutaneous nevi, suggesting that BRAF mutations are one of the first of many genetic events in the formation and progression of cutaneous melanoma (Pollock et al. 2003; Dahl & Guldberg 2007). BRAF-mutated cutaneous melanoma appears to form a biologically distinct subtype in which patients are younger at the time of diagnosis, the melanoma is located in intermittently sun-exposed (non-chronic) sites, it is more often of the superficially spreading kind, and it is more frequently associated with a pre-existing nevus (Edlundh-Rose et al. 2006; Liu et al. 2007; Platz et al. 2008; Bauer et al. 2011; Menzies et al. 2012; Kim et al. 2015). The RAS gene family (HRAS, KRAS and NRAS) encodes small guanine-nucleotide binding proteins that are activated by receptor tyrosine kinases (RTKs), consequently leading to activation of the MAPK pathway (Fig. 5). NRAS mutations may activate the phosphatidylinositol 3-kinase-AKT (PI3K-AKT) pathway, which is involved in cell proliferation and survival mechanisms (Dahl & Guldberg 2007). NRAS mutations are present in approximately 20% of cutaneous melanoma, and they most often involve a substitution of glutamine at residue 61, causing constitutive activation of the protein (Dahl & Guldberg 2007; Platz et al. 2008). These mutations appear to occur more frequently in cutaneous melanoma developed in chronically sun-exposed areas of the skin (Platz et al. 2008). The KIT gene encodes a RTK that functions in mediating extracellular stimulations from stem cell factors, causing activation of several signal transduction pathways, including the MAPK and PI3-AKT pathways. KIT mutations are relatively common in cutaneous melanoma on chronically sun-damaged skin, but they are very rare in cutaneous melanoma arising on intermittently sun-exposed skin (Cohen et al. 2004; Curtin et al. 2006; Dahl & Guldberg 2007).

Whereas BRAF mutations are rare in MM in general, NRAS and KIT mutations occur at a higher frequency (Cohen et al. 2004; Curtin et al. 2006; Beadling et al. 2008). KIT mutation rates in MM appear to vary with anatomical site, that is there is a higher frequency of KIT mutations in anorectal, vulvar, or vaginal MM than in oral and nasal MM (Beadling et al. 2008; Omholt et al. 2011; Colombino et al. 2013; Turri-Zanoni et al. 2013; Zebary et al. 2013).

Based on identification of oncogenic mutations in the different melanoma subtypes, significant progress has been made in the development and use of targeted treatment in advanced melanoma disease. Targeted treatment with BRAF- and MEK inhibitors has substantially improved the survival of patients with stage-IV cutaneous melanoma (Long et al. 2011; Dossett et al. 2015). In addition, for patients with KIT- mutated melanoma, a phase-II trial showed promising effects of targeted treatment with tyrosine-kinase inhibitors (Imatinib; Carvajal et al. 2011).

The prognostic implications of BRAF mutations in cutaneous melanoma have been widely debated. A meta-analysis found a significant increase in the risk of mortality in melanoma patients with BRAF mutations (Safaee et al. 2012). Also, recent results have shown associations involving poor melanoma-specific survival in patients with BRAF-mutated, stage-I to stage-III cutaneous melanoma (Mar et al. 2015). This emphasizes the need for early detection of tumours and mutations and for studies evaluating the impact of adjuvant-targeted treatment in cutaneous melanoma. For this purpose, immunohistochemical staining using the antibody VE1 has proven to be a rapid and efficient method for detection of BRAF V600E mutations in cutaneous melanoma (Lade-Keller et al. 2013; Long et al. 2013).

Oncogenic mutations in conjunctival melanoma

Only limited cytogenetic studies of oncogenic aberrations in CM had been performed when the present PhD study began (El-Shabrawi et al. 1999; Gear et al. 2004; Spendlove et al. 2004; Goldenberg-Cohen et al. 2005; Beadling et al. 2008; Populo et al. 2010; Lake et al. 2011; Wallander et al. 2011). During the course of the study, additional information on BRAF, NRAS and KIT mutations in CM has been published and has contributed to our understanding of CM tumorigenesis (Alessandrini et al. 2013; Griewank et al. 2013a; Sheng et al. 2015). Recently, telomerase reverse transcriptase (TERT) promotor mutations leading to increased TERT expression were identified in 32–41% of CMs and in 8% of PAM+ cases, but they were absent in conjunctival nevi (Griewank et al. 2013b; Koopmans et al. 2014). However, the significance of these abnormalities in CM remains to be established.

BRAF mutations have been identified in up to 50% of CMs, much resembling the frequency reported in cutaneous melanoma (Gear et al. 2004; Spendlove et al. 2004; Goldenberg-Cohen et al. 2005; Beadling et al. 2008; Platz et al. 2008; Populo et al. 2010; Lake et al. 2011; Griewank et al. 2013a; Sheng et al. 2015). KIT mutations appear to occur only rarely in CM (Beadling et al. 2008; Wallander et al. 2011; Alessandrini et al. 2013; Griewank et al. 2013a). However, a recent study of 53 Chinese CM patients revealed KIT mutations in 11% of CMs and a very low prevalence of BRAF mutations (8%), suggesting that there may be different pathways in tumour development in different ethnic groups. NRAS mutations have been reported in up to 18% of CMs, closely resembling the proportions reported for NRAS mutations in both cutaneous and sinonasal MM (El-Shabrawi et al. 1999; Beadling et al. 2008; Platz et al. 2008; Populo et al. 2010; Griewank et al. 2013a; Turri-Zanoni et al. 2013; Zebary et al. 2013). The distributions and comparison of BRAF, NRAS and KIT mutations in conjunctival, cutaneous and sinonasal MM are summarized in Table 3, with comparisons. The inconsistency in the reported frequencies of BRAF mutations in CM could be due to the use of different detection methods, different sample types (formalin-fixed or fresh frozen), small sample sizes, different sample selection regarding location and origin of the CM, and patient characteristics such as ethnicity. Consequently, clinicopathological and prognostic associations regarding BRAF mutations in CM have not yet been explored thoroughly, and large population-based studies are needed to establish these features. Little is known about the genetic mutations that drive CM development. BRAF mutations have been suggested to be early events, by analogy with cutaneous melanoma, since these have been identified in 14 of 28 conjunctival nevi, but not in PAM+ cases (Goldenberg-Cohen et al. 2005). However, no analyses have been performed on consecutive premalignant lesions and CM. An increase in the incidence of CM analogous to that in cutaneous melanoma has been reported, particularly in melanomas located in the sun-exposed conjunctiva (Tuomaala et al. 2002; Yu et al. 2003; Triay et al. 2009). Whether or not this increase in incidence is associated with BRAF mutations has not been determined. The new advances in targeted therapy underscore the importance of characterizing mutations in CM. Immunohistochemical staining with the antibody VE1 is a rapid method for detection of BRAF V600E mutations in cutaneous melanoma (Capper et al. 2011; Lade-Keller et al. 2013). This antibody has not been investigated previously in CM.

MicroRNA expression

The role of epigenetic modulation in cancer has been investigated intensively in recent years, and in addition to DNA methylation and chromatin remodelling, miRNAs are important epigenetic regulators (Esquela-Kerscher & Slack 2006; Finnegan & Pasquinelli 2013). MicroRNAs are small, non-coding RNA molecules that mainly cause posttranscriptional silencing by binding to the 3′UTR of messenger RNAs, causing either de-stabilization or translational block (Finnegan & Pasquinelli 2013). The biogenesis of miRNA involves several steps (Fig. 8; Finnegan & Pasquinelli 2013). One miRNA molecule can influence the expression of several protein-encoding genes, so small changes in miRNA are likely to have a great impact on tumorigenesis (Esquela-Kerscher & Slack 2006).

Several miRNAs have been shown to function as oncogenes and tumour suppressors, and are involved in the pathogenesis of almost all types of cancer (Esquela-Kerscher & Slack 2006; Zhang et al. 2007; Kunz 2013). MicroRNAs can be reliably detected in formalin-fixed, paraffin-embedded (FFPE) melanoma samples (Glud et al. 2009; Larsen et al. 2014), making retrospective studies possible. Whereas several miRNAs have been reported to be differentially expressed in the progression of cutaneous melanoma (Chen et al. 2011; Penna et al. 2011; Xu et al. 2012; Kunz 2013), the miRNA expression profile in CM has not been explored.

Patients and tumour samples

This PhD project was based on tissue samples identified by searching pathology reports at the Eye Pathology Institute and the Danish Registry of Pathology using SNOMED (Systematized Nomenclature of Medicine) codes for primary, recurrent and metastatic malignant melanoma, together with topographical codes for conjunctiva, eyelid skin and caruncle, from 1 January 1960 to 31 December 2012.

Patients with CMs that had been biopsied or removed surgically were included in the study. Archived FFPE tissue blocks, fresh frozen tissue samples, histological slides, pathology reports and clinical patient records were obtained and histopathological diagnoses were re-evaluated. Cases with CM in situ were excluded. If there was invasion to adjacent eyelid skin, each case was evaluated and primary cutaneous melanomas were excluded. For each CM patient, the records at the Eye Pathology Institute were also searched for previous premalignant lesions, that is nevi or PAM in the same eye and conjunctival location, and these were defined as paired lesions.

Altogether, 139 CM cases were identified from the period 1960 to 2012. In the first genetic study investigating BRAF mutations in CM (study I), patients were included if diagnosed in the time period 1994–2012, if sufficient tumour material was obtained from the Eye Pathology Institute, and if clinical data were available. This meant that 56 CM cases with FFPE tissue available were included. In our population-based study of BRAF mutations in CM (study II), patients were included if diagnosed between 1960 and 2012, and if tumour material and clinical data were available. Furthermore, pathology reports were traced back to identify archived paired lesions from which tissue samples could be obtained. This gave 119 CM cases and 25 paired lesions with FFPE tissue available.

For miRNA microarray profiling experiments (study III), patients diagnosed with CM were identified by searching the archives at the Eye Pathology Institute and the Danish Registry of Pathology during the years 2000 to 2012, and archived and fresh frozen tissue samples were collected. Archived tissue samples from normal conjunctiva were collected from the Eye Pathology Institute during the years 1999–2000. Fresh frozen tissue samples from nasal and oral MM patients were identified and obtained from the Danish Registry of Pathology and the Danish Cancer Biobank during the years 2000–2014. Sufficient archived tumour material was obtainable from 40 of 55 CM samples and from seven normal conjunctiva samples. Fresh frozen tumour samples were obtained from six CMs, three sinonasal melanomas and one laryngeal melanoma.

Clinicopathological data and follow-up

For the 139 CM patients, data on gender, age and date of diagnosis were collected in all but one case (where age at diagnosis could not be obtained). Clinical patient records, pathology requisitions, pathology reports and death certificates were collected and reviewed, and the Danish Registry of Pathology, the Danish Registry of Causes of Death and the National Patient Administrative System were searched in order to obtain data on tumour location (bulbar-, corneo-limbal- and/or palpebral conjunctiva; fornix; plica semilunaris; caruncle and/or eyelid margin), tumour extension (1–4 quadrants), invasion to adjacent tissue structures, colour of the tumour (pigmented, non-pigmented, mixed), and configuration (nodular, flat, mixed), presence or absence of clinical PAM, largest basal diameter, presence of regional or distant metastatic disease, type of treatment and status.

Tumour location was characterized as being epibulbar (where the CM is confined to the bulbar and corneo-limbal conjunctiva), extrabulbar (where the CM is either confined to or extends to the fornix, palpebral conjunctiva or eyelid margin) or caruncular. We characterized sun-exposed lesions as being epibulbar and/or caruncular, and non-sun-exposed lesions as being extrabulbar. Tumour, Lymph node and Metastasis staging was based on available clinical information and was performed according to the American Joint Commission on Cancer TNM classification, seventh edition (Edge et al. 2009).

Follow-up data were collected and included: presence of and time from diagnosis to development of local recurrence (recorded if recurrence was reported in the same eye at any site and with a histopathological diagnosis of CM), regional and/or distant metastatic disease (any metastasis), distant metastatic disease (distant metastasis), death from melanoma-related causes (patients included reported dead due to tumour progression, based on death certificates and clinical patient records), death from all causes (including patients who died from an unknown cause).

Haematoxylin- and eosin-(HE) stained slides were reviewed by the author and an eye pathologist (JUP) in order to determine the predominant cell type (spindle-shaped, epithelioid or mixed) and the tumour thickness (measured from the epithelial surface to the deepest tumour cell), and divided into three categories (≤0.5, >0.5 to 1.5, and >1.5 mm) in line with the current TNM classification, and additionally into two categories (≤2 and >2 mm). The origin (PAM+, PAM+ and nevus, nevus or de novo) was established by reviewing HE-stained slides, stainings against melan-A and cytokeratin and clinical information. Haematoxylin- and eosin-stained slides or FFPE tissue were not obtained in eight CM patients. In these cases, available information on cell type, tumour thickness and origin was obtained from pathology reports. The proliferative index (mitotic rate) was evaluated in slides stained with anti-Ki-67 (Dako Denmark, Glostrup, Denmark) and the percentage of Ki-67-positive cells was assessed, allowing division of CMs into two groups (≤15% and >15%).

Discussion

Patients included in the present study were identified by searching two population-based registries covering the years 1960–2012. We searched the archives at the Eye Pathology Institute, which provides a nationwide service free of charge to private ophthalmologists and departments of ophthalmology in Denmark. Since some ophthalmology departments may send their specimens to local pathology departments, the Danish Registry of Pathology was also searched. From the 1960s, the major specialized hospitals in Denmark registered their diagnoses in the Danish Registry of Pathology, and by the 1980s all the pathology departments in Denmark were using this registry. Since the present study period began in 1960, some cases at the beginning of this time period may have been missed. However, since the hospitals that registered diagnoses from the 1960s are the same as the ones that treated CM patients, there must have been very few missing cases and the incidence of CM can be considered reliable. Conjunctival melanoma is a rare disease, and consequently most published data are case reports or based on series of referred patients from specialized departments. Such data may be biased; furthermore, loss to follow-up may skew the results. Since 1968, each resident of Denmark has been given a unique identification number from the Central Population Registry (CPR). These CPR numbers enable accurate linkage of patient data across various registries and patient records. Thus, only a few patients were lost to follow-up in the present study, and the loss of clinical data was minimal. Thus, the present study was not prone to selection or referral bias.

Droplet digital PCR analysis

DNA was extracted from FFPE tumour samples using the QIAamp^® Deparaffinization Solution and the DNA FFPE Tissue Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. The concentration of DNA was measured spectrophotometrically using a NanoDrop ND-1000 (Thermo Scientific, Wilmington, DE, USA). Samples with a DNA concentration of ≥10 ng/μl were included (study I). No lower limit criterion for DNA concentration was set in study II. Detection and quantification of BRAF mutations (c.1799T>A, p.V600E and c.1798_1799GT>AA, p.V600K) were performed by ddPCR analysis using the QX200 system (Bio-Rad Laboratories, Hercules, CA, USA) and hydrolysis probe-based assays (PrimePCR ddPCR Mutation Detection Assays; Bio-Rad). The PCR mixture contained 11 μl of ddPCR supermix for probes (no dUTPs), 1.1 μl of mutation primer/probe mix (FAM), 1.1 μl of wildtype primer/probe mix (HEX) and 5 μl of DNA (11.6–508.4 ng/μl in study I) in a total volume of 22 μl. Twenty microlitres of this mixture and 70 μl of droplet generation oil were transferred to different wells of a droplet generation cartridge. After emulsification of the samples using the droplet generator, samples were transferred to a 96-well PCR plate and amplified for 40 cycles of 94°C for 30 seconds and 55°C for 60 seconds. Droplets were analysed on the droplet reader, and quantasoft software (version 1.4.0.99, BioRad Laboratories, Hercules, CA, USA) was used for analysis of DNA concentrations. Cut-off settings were determined using mutation-positive and mutation-negative control DNA samples.

BRAF V600E immunohistochemistry

For immunohistochemical detection of BRAF V600E oncoprotein (VE1) detection, CM and paired premalignant specimens were placed on slides with positive tissue controls and examined using the antibody VE1 (dilution 1:40; Spring Bioscience, Pleasanton, CA, USA), which targets the BRAF V600E oncoprotein. Slides were pretreated with heat-induced epitope retrieval (pH 8.5, CC1 buffer) using the Ventana Benchmark XT CC1S program (Roche A/S, Hvidovre, Denmark). For visualization, the UltraView Universal Alkaline Phosphatase Red Detection kit (Roche A/S) was used. The intensity of the oncoprotein stain was blindly and independently evaluated by one observer (JLK) and scored as absent (score = 0), uncertain (very weak, almost absent staining; score = 1), weak to moderate (score = 2) or strong (score = 3), and defined as negative (intensity score ≤ 1) or positive (intensity score ≥ 2).

MicroRNA expression analysis

Archived CM samples were evaluated using HE- and melan-A-stained slides to establish the amount of tumour tissue, and 8–10 sections (each of 20 μm thickness) were cut. In cases where the amount of tumour tissue was evaluated to be <80–90%, the sections were transferred to glass slides and tumour tissue was macrodissected by hand with a scalpel.

For automated purification of total RNA from formalin-fixed, paraffin-embedded tissue, the QIAsymphony RNA Kit (Qiagen, Valencia, CA, USA) was used according to the manufacturer's instructions. The AllPrep DNA/RNA Micro Kit (Qiagen) was used according to the manufacturer's instructions for purification of total RNA, including small RNAs, from fresh frozen tumour samples. The degree of RNA degradation was evaluated prior to microarray analysis using the Agilent RNA 6000 Nano kit on an Agilent 2100 Bioanalyser. Total RNA (130 μg) was prepared from each sample and hybridized on a GeneChip 4.0 Array (Affymetrix, Santa Clara, CA, USA) according to the manufacturer's instructions. Briefly, total RNA was labelled with FlashTag™ Biotin HSR through poly (A) tailing and subsequent FlashTag™ Biotin HSR ligation, and incubated at 99°C for 5 min and 45°C for 5 min with Hybridization Master Mix. Arrays were hybridized at 48°C and 60 rpm for 16–18 hr in an Affymetrix^® hybridization oven 640 (Affymetrix) and washed and stained with phycoerythrin-conjugated streptavidin (SAPE) in the Affymetrix Fluidics Station 450. Arrays were scanned in an Affymetrix^® GeneChip Scanner 3000 to generate fluorescent images, as described in the Affymetrix Gene Chip^® protocol. Cell intensity files were generated in the GeneChip^® Command Console^® Software (AGCC; Affymetrix). Cell intensity files were RMA normalized using the Affymetrix Expression Console and 2578 probe sets representing all human mature miRNAs were selected for further analysis. The expression matrix was imported into the r environment (R Core Team 2015) and differential expression between samples was defined using ebayes modelling, as implemented in the r limma package (Smyth 2004). MicroRNAs were defined to show differential expression if fold changes were above 2 and Benjamini-Hochberg adjusted p-values (false discovery rates) were below 0.1. Differentially expressed miRNAs were visualized using hierarchical clustering based on average linkage and Euclidian distance, as implemented in Qlucore Omics Explorer bioinformatics software (www.qlucore.com).

For the analysis of miRNA expression in CMs from archived tissue, three samples were excluded because they showed outlier miRNA signals. This resulted in the inclusion of 37 CM samples for further analysis. For the identification of tumour-specific miRNAs, all pooled tumour samples were compared to normal conjunctiva and for the identification of prognostic miRNAs, TNM stage T1, T2, and T3 tumours were compared to normal conjunctiva. Analysis of fresh frozen tumour samples was performed as described for archived tissue, and the expression of tumour-specific miRNAs was compared in CM and in nasal and oral MM.

Statistical evaluation

For the calculation of incidence, the time period was divided into five groups of 10 years each. Age was categorized in three groups (<50, 50–65 and >65 years). Crude incidence rates (overall, time period-specific, age-specific, gender-specific, sun exposure-specific, location-specific and BRAF-mutation-specific) were calculated as the number of cases divided by the number of person-years in the corresponding category, and expressed as cases per 1 000 000 person-years. A single patient was excluded from incidence calculations because of unknown age and unknown clinical presentation at diagnosis. The total number of people in Denmark in each age group was obtained from the Statistics Denmark website (www.dst.dk) for each year of the study period. Confidence intervals were calculated by assuming a Poisson distribution. A multivariable Poisson regression model, adjusting for age and gender, was used to determine the independent effects of time period, age, gender, sun exposure and location on the incidence. These effects are expressed as rate ratios (RRs) relative to the first time period (1960–1969), age <50 years, male gender, non-sun-exposed location, bulbar location or BRAF-wildtype mutation status. The significance of the variables was tested with likelihood ratio tests (II).

Associations between primary outcomes and patient characteristics, clinical features and treatment were assessed in Cox regression models (I). Analyses were adjusted for possible confounding by also including gender, age, location and local invasion as covariates in multivariable Cox regression models. Fisher's exact test (with binary variables), Chi-square test (with more than two variables), Mann–Whitney U-test (with continuous variables) and the log-rank test were used where appropriate to compare BRAF-mutation status with patient characteristics, clinicopathological features and outcome (I and II). Chi-square and Kruskal–Wallis test were used to compare demographic characteristics and tumour features with clusters of low, intermediate and high miRNA expression.

Statistical calculations were performed using spss (version 20; IBM SPSS, Ardmonk, NY, USA) and sas (version 9.4; SAS Institute Inc., Cary, NC, USA).

Discussion

We used ddPCR analysis and microarray miRNA expression analysis, which are well-known and well-established detection methods routinely performed at the Danish Cancer Society and at the Microarray Centre, Rigshospitalet. The analyses were mainly performed on FFPE tissue, which were more than 50 years old in some cases. Overall, precautions were taken to maximize the quality of FFPE sections. This included sectioning of the material shortly before the analysis and storing the sections under cold conditions. Formalin fixation has been known to generate DNA-protein cross-linkage and cause DNA fragmentation, and the duration of formalin fixation in particular appears to be a determining factor (Bass et al. 2014). Long-term storage of FFPE tissue appears less likely to have an influence on the DNA quality, but the amplifiability of some gene fragments may decrease over time (Bass et al. 2014). This might explain the variation in DNA quality observed, leading to the exclusion of mainly old CM samples. Droplet digital polymerase chain reaction requires very small amounts of DNA and it can detect mutations with high sensitivity, even in cases where samples contain only a small fraction of tumour cells (Nadauld et al. 2012). For each sample, DNA is fractionated in up to 20 000 droplets. Polymerase chain reactions are then run on each individual droplet and the mutation status is then based on the evaluation of the droplets as a whole. Because of the potentially poor quality of mainly old CM samples, our first analysis included newer CM samples with a concentration of ≥10 ng/μl. In a subset of these CM cases, fresh frozen tissue was also obtained. These cases were analysed for BRAF mutations, and revealed a similar BRAF-mutation status to what was observed in our analysis of the corresponding FFPE tissue (data not shown) – thus validating our results. We did not observe a time-dependent decrease in DNA concentration, so we proceeded by including all the samples that were available from 1960 to 2012.

Formalin-fixed, paraffin-embedded tissue can be stored for several years for IHC, provided that sections are freshly cut (Bass et al. 2014). We therefore included all the samples available from 1960 to 2012. The use of FFPE blocks >25 years old is, however, not generally recommended, so some discrepancy was expected with very old CM samples (Bass et al. 2014). BRAF V600E IHC analysis was performed in collaboration with the Department of Histopathology, Aarhus University Hospital, where this analysis is routinely performed. To verify that the antibody staining observed was specific, positive tissue controls were used on each slide.

In the process of formalin fixation, RNA molecules may be degraded and modified (Bass et al. 2014). Whereas fresh frozen tissue is optimal for assessment of differentially expressed miRNAs, miRNAs are relatively unaffected and well preserved in FFPE tissues, and FFPE specimens are therefore also suitable for studies of miRNA expression (Glud et al. 2009). We were able to retrieve and analyse fresh frozen CM and MM samples as an additional part of the study. To evaluate prognostic features in CM, we adjusted for the possible confounding effects of tumour location and local invasion in our statistical analysis. These features were chosen based on their inclusion and important role in the current TNM classification (Edge et al. 2009).

Ethics

This study adhered to the tenets of the Declaration of Helsinki and it was approved by the local Scientific Ethics Committee (entries no. H-2-2011-131 and H-6-2014-060) and the Danish Data Protection Agency (entry no. 2011-41-6580).

Epidemiology

During the study period, 139 patients with CM were identified in Denmark. The median age at diagnosis (obtained from 138 patients) was 67 years, with a range from 14 to 100 years.

The incidence increased over the study period, from 0.36 cases/1 000 000/year (1960–1969) to 0.87 cases/1 000 000/year (2000–2009; Table 4). When adjusting for age and gender, the increase in incidence involved an 11-fold increase in CM in patients >65 years old (RR = 10.88, p < 0.0001) and a more than twofold increase in CM that had developed in sun-exposed conjunctiva (RR = 2.46, p < 0.0001). In particular, epibulbar CM showed higher incidence than caruncular and extrabulbar CM (p < 0.0001; II).

Clinical presentation and prognosis

Follow-up data were available for 132 of 139 patients (Table 5; I). Patients were staged according to the seventh edition of the AJCC TNM classification (Edge et al. 2009) and were classified as T1 in 86/130 (66%), as T2 in 27/130 (21%) and as T3 in 17/130 (13%).

Extrabulbar location was associated with increased melanoma-related and all-cause mortality, with hazard ratios (HRs) of 2.72 (95% CI, 1.12–6.59; p = 0.03) and 2.21 (95% CI, 1.25–3.90; p = 0.006) respectively. Conjunctival melanoma patients with local invasion to adjacent tissue structures had a higher risk of distant metastasis and any metastasis with HR of 3.38 (95% CI, 1.21–9.45; p = 0.02) and 3.04 (95% CI, 1.13–8.18; p = 0.03) respectively (Table 5; I). There was no significant association between TNM tumour stage (T) and local recurrence or regional metastasis (p = 0.99 and p = 0.20, respectively, log-rank test), but lower rates of distant metastasis were identified in T1- and T2-stage CM than in T3-stage CM (p = 0.002), and a reduced proportion of melanoma-related deaths were identified in T1 compared to T2 and T3 CM (p = 0.004; Fig. 8). Local recurrence was significantly associated with regional metastasis, distant metastasis and melanoma-related death (p = 0.003, p = 0.003, and p = 0.05, respectively, log-rank test).

Histopathological features and prognosis

The histopathological origin was PAM+ in the majority of cases (80/129; 62%). The median thickness measured in 125 CM was 2 mm (range, 0.1–20 mm). All CMs invaded the substantia propria, and 73% of CMs involving the extrabulbar conjunctiva (27 of 37) exceeded 1.5 mm in thickness (I). The predominant cell type was mixed in 59% of cases (n = 69), epithelioid in 36% (n = 42) and spindle-like in 6% (n = 7). No significant prognostic association of cell type was identified in log-rank analyses. In 102 cases, the median Ki-67 proliferative index was 10% (range, 0–50%) and the majority of CMs (60%, 61/102) had a Ki-67 proliferative index of ≤15%. A tendency of increased risk of death from all causes was observed in melanoma developed de novo (HR = 1.94; 95% CI, 0.98–3.84; p = 0.06). Additionally, melanoma developed in a nevus compared to melanoma developed in PAM+ tended to be associated with a lower risk of local recurrence (HR = 0.41; 95% CI, 0.16–1.07; p = 0.07) and any metastasis (HR = 0.51; 95% CI, 0.25–1.02; p = 0.06; I). A division based on two categories of thickness (≤2 and >2 mm) revealed a significantly reduced proportion of melanoma-related deaths in patients with CM <2 mm in thickness (p = 0.034, log-rank test).

Treatment and prognosis

Data on primary treatment were obtained for 129 patients. Incisional or diagnostic biopsy was performed before other treatment in 27 patients. Most patients were treated with excision of the tumour combined with adjuvant therapy (Table 5). Sentinel lymph node biopsies was performed in three patients, all initially seen with eyelid skin involvement (I). Patients who initially had an incisional biopsy performed had a higher risk of distant metastatic disease (HR = 2.46; 95% CI, 1.08–5.58; p = 0.03). Patients treated with excision without adjuvant treatment more frequently developed local recurrence (HR = 1.97; 95% CI, 0.11–3.48; p = 0.02) and distant metastasis (HR = 2.51; 95% CI, 1.07–5.91; p = 0.03) and had increased risks of melanoma-related (HR = 2.27; 95% CI, 0.97–5.33; p = 0.06) and all-cause mortality (HR = 1.80; 95% CI, 1.05–3.08; p = 0.03; I).

BRAF mutations in conjunctival melanoma

BRAF mutation results were obtained from 47/56 CMs in study I and 111/119 CMs in study II. The proportion of BRAF mutations was slightly higher in CMs diagnosed from 1994 to 2012 (19/47, 40%) than in CMs diagnosed from 1960 to 2012 (39/111, 35%; I and II).

The majority of BRAF mutations were V600E-type (82%) whereas V600K-type mutations accounted for 18%. The distribution (Fig. 9) and adjusted RRs of BRAF-mutated CM, relative to BRAF-wildtype CM, did not change over time (p = 0.64; II).

Clinicopathological features associated with BRAF mutations are presented in Table 6. In our study II investigating BRAF mutations in 111 CM patients, the results were similar to those reported in study I. Overall, a stronger statistical association was observed – except for tumour colour. Male patients more often had a BRAF-mutated tumour (24 males and 15 females; p = 0.016, Fisher's exact test). Patients with a BRAF-mutated CM (mean age, 56.8 years) were significantly younger than patients with BRAF-wildtype tumours (mean age, 69.8 years; p = 0.001, Mann–Whitney U-test). The BRAF-mutated CMs occurred more frequently in sun-exposed sites (p = 0.01). Patients with BRAF-mutated CMs rarely presented with clinically described PAM (p < 0.001), and more often presented with a mixed pigmented or non-pigmented lesion (p = 0.023). BRAF mutations were more frequent in CMs originating in a nevus (16/27; 57%) than in CMs originating from PAM+ (22/69; 32%) or de novo (1/12; 8%; p = 0.005). The BRAF-mutated CMs were more frequently classified as stage-T1 and -pT1 tumours (p = 0.007). A tendency of more frequent epithelioid cell type was also identified in BRAF-mutated CM (p = 0.06, chi-square test; II).

Distant metastasis occurred in seven of 47 CM patients (15%) diagnosed from 1994 to 2012 (I). The majority of these patients (6/7; 86%) had a BRAF-mutated CM, and a reduced distant metastasis-free survival was observed in BRAF-mutated CM in univariate analysis (Fig. 10). In Cox regression models adjusting for gender, age, location and local invasion, higher risks of local recurrence, metastasis or mortality were not observed in BRAF-mutated CM than in BRAF-wildtype CM (I and II).

BRAF mutations in paired premalignant lesions

BRAF mutation results were obtained in 20/25 of the paired lesions. BRAF mutations were identified in 11/20 (55%) of the paired lesions, the majority of these being nevi. The mutation status of the premalignant lesions corresponded with that of the paired primary CM in 19/20 of the cases (II).

BRAF V600E immunohistochemistry

Data on IHC BRAF V600E were obtained from 102 primary CMs. In 26/30 (87%) of the cases, both IHC and mutational analysis detected a BRAF V600E mutation. In 60/67 (90%) of the CM cases, both types of test were negative. All of the BRAF V600K-mutated CM cases (5/5) corresponded to a negative IHC (II). The non-corresponding cases were based on slides cut from 9- to 54 year-old FFPE tissue blocks (median age, 34.5 years). The BRAF V600E IHC stain detected mutations with a sensitivity of 0.94 and a specificity of 1.00 (Fig. 11) when patients from more recent years were analysed (2000–2012, n = 47; II).

MicroRNA expression in conjunctival melanoma

We identified 25 significantly differentially expressed miRNAs (24 up-regulated and one down-regulated) in CM compared to normal conjunctiva (Table 7). A supervised hierarchical cluster analysis of the 25 differentially expressed miRNAs differentiated normal conjunctiva samples from melanoma. The analysis showed a gradient of increased expression from melanomas resembling normal tissue to a highly up-regulated expression pattern (III).

Differentially expressed microRNAs associated with tumour stage

In addition to six repeated miRNAs, a single miRNA (miR-30d-5p) was identified as up-regulated in both stage-T1 and -T2 CM relative to normal conjunctiva (Table 7, Fig. 12A). No differentially expressed miRNAs were identified in stage-T3 CM. A two-way cluster analysis based on these seven differentially expressed miRNAs divided the samples in clusters of low expression, intermediate expression and high expression (Fig. 12B). Conjunctival melanomas with a high expression pattern were significantly thicker than CMs with intermediate or low expression (p = 0.007, Kruskal–Wallis test; III).

MicroRNA expression in conjunctival melanoma compared to mucosal melanoma

Comparing fresh frozen tissue samples from six CM patients and four MM patients, variations in – but no significant change in – expression of the tumour-specific miRNAs was observed (III).

Epidemiology

The overall incidence of primary CM in Denmark between 1960 and 2012 was 0.5/1 000 000/year, and a more than twofold increase in incidence of CM was observed over the study period. We found a similar overall incidence of CM to that previously reported in Denmark, estimated as 0.45 per million between the years 1960 and 1980 (Nørregaard et al. 1996). Since the increase in incidence was observed after the 1990s, it is less likely to have been caused by missing registrations of CM. Furthermore, the increase in incidence corresponds well with the results of other population-based studies of CM in Scandinavia (Tuomaala et al. 2002; Triay et al. 2009). In addition, our reports of a more than 11-fold increase in incidence in CM patients aged >65 years and a more than twofold increase in incidence in CMs confined to the epibulbar conjunctiva are similar to results from Sweden (Triay et al. 2009). It is possible that the observed rise in incidence reflects an increase in detection of CM cases. Additionally, the increase may also be driven in part by an increase in the proportion of older adults in the Danish population. However, due to the striking and statistically significant rise in the incidence of CM in older patients and at sun-exposed sites, our results suggest that accumulated exposure to sunlight may play a role in CM development.

Clinicopathological features, treatment and prognosis

The age, gender and clinical presentation of CM patients in the present study was mostly similar to those in other large studies (Seregard & Kock 1992; De Potter et al. 1993; Paridaens et al. 1994b; Nørregaard et al. 1996; Shields 2000; Anastassiou et al. 2002; Tuomaala et al. 2002, 2007; Werschnik & Lommatzsch 2002; Missotten et al. 2005; Triay et al. 2009; Shields et al. 2011). Accordingly, the gender ratio was not statistically significantly different from unity, and the mean age (64 years) was similar to that in a previous population-based study performed in Denmark (Nørregaard et al. 1996).

The rate of local recurrence was 48% in the present study, and it was confirmed that local recurrence was associated with more frequent development of metastases and melanoma-related death. A higher risk of local recurrence, distant metastasis and melanoma-related death were identified in patients who were treated with excision alone. These findings highlight the importance of supplementing excision with adjuvant therapy such as cryotherapy, brachytherapy or topical chemotherapy when treating CM patients. The origin of the tumour may explain the poor outcome observed in patients treated with excision alone – since there tended to be more frequent local recurrence in CMs originating from a PAM+ rather than a nevus, and since practices regarding treatment have changed over the considerably long follow-up time. The result therefore also reflects the change in practice over time, particularly concerning the treatment of CM developed from PAM+. In this work, we did not have enough data to specifically identify a preferred type of adjuvant treatment, but recent studies have shown improvement in local recurrence rates when brachytherapy is used alone or in combination with local chemotherapy (Damato & Coupland 2009a,b; Kenawy et al. 2013). This treatment is currently the preferred choice at our institution.

Regional metastasis occurred in 16 patients (12%), and most of these (13/16; 81%) presented with – or subsequently developed – distant metastases. Most patients who developed distant metastatic disease (17/30; 57%) had no evidence of previous regional metastases. However, because of the considerably long follow-up time in the present study, practices in screening methods for detection of regional metastases have varied. Palpation of regional lymph nodes has mainly been performed, and diagnosis of regional lymph node involvement may therefore have been underdiagnosed. Identification of patients with a higher risk of regional metastasis is important if future patients are to benefit from SLNB. In this work, patients who developed regional and subsequent or concurrent distant metastasis more often presented with local invasion of adjacent tissue structures, mainly eyelid skin. A poor prognosis in patients with involvement of the eyelid margin has been described previously (Robertson et al. 1989; Tahery et al. 1992). These results suggest that CM patients with local invasion in addition to known risk factors, that is tumours more than 1–2 mm thick, palpebral or non-limbal location, and presence of histological ulceration may benefit from SLNB (Tuomaala & Kivelä 2004, 2008; Tuomaala et al. 2007; Esmaeli 2008; Damato & Coupland 2009b; Savar et al. 2011; Esmaeli et al. 2012; Maalouf et al. 2012; Cohen et al. 2013). Due to the fact that SLNB was performed in only three patients, we did not have sufficient data to evaluate the effect of SLNB on prognosis.

Caution against incisional biopsies has been exercised in several studies (Paridaens et al. 1994b; Shields 2000; Werschnik & Lommatzsch 2002). We found a significantly higher risk of distant metastasis in patients in whom an incisional biopsy was performed, regardless of subsequent treatment. This finding highlights the importance of referral to a specialized ophthalmological department and complete excision and adjuvant therapy where possible.

Conjunctival melanoma cases involving extrabulbar conjunctiva were associated with a higher risk of metastasis and melanoma-related mortality. This is in agreement with other studies (Folberg et al. 1985a; Paridaens et al. 1994b; Shields 2000; Esmaeli et al. 2001; Anastassiou et al. 2002; Tuomaala et al. 2002; Werschnik & Lommatzsch 2002; Missotten et al. 2005; Shields et al. 2011). However, an increased mortality rate for CM cases involving the caruncle could not be confirmed (Paridaens et al. 1994b; Anastassiou et al. 2002; Damato & Coupland 2009a). The poor prognosis of CM with caruncular involvement has previously been attributed to the close resemblance to skin (Seregard 1998), and indeed, four of six caruncular CM cases presented with invasion to skin of the eyelid in the present work, but caruncular location alone was not associated with a poor prognosis. Our results are similar to those from a population-based study from Finland in which an adverse prognosis for CM with caruncular involvement could not be demonstrated either (Tuomaala et al. 2002). Tumour thickness has previously been claimed to be the sole sovereign prognosticator in CM (Jakobiec 1980). We assessed tumour thickness of the initial CM, and observed that the thickest tumours were more frequently located in extrabulbar conjunctiva. In addition, tumours exceeding 2 mm were found to be associated with a significantly higher proportion of melanoma-related deaths in univariate analysis. Tumour thickness was evaluated in Cox regression models, correcting for the possible confounding effects of tumour location and invasion to adjacent tissue structures (I). In these analyses, increasing tumour thickness did not show higher risks of local recurrence, metastasis or mortality. Consequently, our results indicate that the poor prognosis with increased tumour thickness may be because tumours have had time to grow unremarked in extrabulbar locations.

The nomenclature and classification of CM have changed considerably over the years. In order to optimally assess tumour origin, we therefore re-evaluated the origin in all the CM cases available. We observed a higher but non-significant risk of all-cause mortality in CM developed de novo compared to CM developed from a PAM+ regardless of age, gender, tumour location and local invasion (HR = 1.94; 95% CI, 0.98–3.84; p = 0.06) of all-cause mortality. Melanomas developed from a nevus rather than a PAM+ appeared to be associated with a lower risk of local recurrence and any metastasis. Conjunctival melanomas originating de novo rather than from a PAM+ have previously been associated with higher risk of metastasis and melanoma-related death (Shields et al. 2011). These findings confirm an unfavourable prognosis when a CM develops de novo, and they highlight the importance of assessing the origin when establishing the initial diagnosis.

Genetic characteristics

We investigated BRAF-mutation status and differentially expressed miRNAs in CM. We observed that BRAF mutations are common in CM, that they associate with a clinicopathological profile that resembles cutaneous melanoma, and that they appear to be early events in CM development. BRAF mutations have been intensively investigated in cutaneous melanoma, and the BRAF-mutated genotype appears to represent a biologically, clinically and prognostically distinct subgroup of cutaneous melanomas (Maldonado et al. 2003; Cho et al. 2005; Edlundh-Rose et al. 2006; Liu et al. 2007; Menzies et al. 2012; Safaee et al. 2012; Mar et al. 2015). Targeted BRAF and MEK inhibitor therapies have been introduced and have substantially improved survival in patients with advanced cutaneous melanoma disease (Long et al. 2011; Dossett et al. 2015).

In contrast, BRAF mutations have only been determined in small cytogenetic studies of CM, and no population-based studies have been performed. BRAF mutations in CM and paired predisposing lesions were investigated in studies I and II. The frequency of BRAF mutations in CM was 40% in study I and 35% in study II, which was higher than reported in a another large study, where 29% of CM cases were BRAF-mutated (Griewank et al. 2013a). Our analysis of CM in study I did not include CMs from other pathology departments (since these could not be obtained at the time of our first analyses) or CMs with insufficient tissue. This could explain the higher frequency of BRAF mutations in study I than in study II. Selection bias, however, was not likely to have been introduced in study I, since the clinicopathological associations regarding BRAF mutations were similar in the two studies.

BRAF mutations have previously been determined in 0–50% of CM cases (Gear et al. 2004; Spendlove et al. 2004; Goldenberg-Cohen et al. 2005; Beadling et al. 2008; Populo et al. 2010; Lake et al. 2011; Griewank et al. 2013a; Sheng et al. 2015). Differences in detection methods, sample types (formalin-fixed or fresh frozen), sample sizes, sample selection (location and origin of the CM, and patient characteristics such as ethnicity) and lack of detection of V600K-type mutations may explain the variations in frequency. The reported frequency of 35% BRAF-mutated CMs in the present study covered a whole nation over 52 years, and to the best of my knowledge it is the first population-based study of BRAF mutations in CM. We observed a distribution of BRAF V600E-type mutations (approximately 80%) and V600K-type mutations (approximately 20%) that is similar to those in reports on cutaneous melanoma (Menzies et al. 2012). Analysis of other genotypes was not performed, so the frequency of mutations in CM may be even higher. However, BRAF mutations apart from V600E-type and V600K-type are rarely reported in cutaneous melanoma (Menzies et al. 2012). The observed frequency of BRAF mutations in CM in this work can therefore be considered to be reliable. We found that BRAF mutations were significantly more frequent in CM diagnosed in younger patients. Similar findings of young age in BRAF V600E-mutated cutaneous melanoma have been described (Edlundh-Rose et al. 2006; Liu et al. 2007). BRAF mutations were predominantly identified in epibulbar and caruncular CMs, and the BRAF-mutated CM was frequently classified as T1 tumours. This suggests that exposure to UVR may be a risk factor in CM, which is similar to what has been described in cutaneous melanoma (Maldonado et al. 2003; Tucker & Goldstein 2003). Whereas an increase in the incidence of CM was observed in Denmark from 1960 to 2012, the relative adjusted rate ratio of BRAF-mutated CM relative to BRAF-wildtype CM did not change over time (II). Thus, the proportion of BRAF mutations in CM remained constant. In light of the increasing incidence of bulbar and caruncular lesions, we expected to find an increasing incidence of BRAF-mutated CM. Surprisingly, this was not the case. Our findings therefore suggest that involvement of different genetic mutations or pathways may account for the increasing incidence observed in CM.

A high prevalence of BRAF mutations has been reported in both conjunctival and cutaneous nevi, in CM with origin in a nevus, and in cutaneous melanoma associated with a preexisting nevus (Pollock et al. 2003; Goldenberg-Cohen et al. 2005; Edlundh-Rose et al. 2006; Griewank et al. 2013a). We can confirm that there was a high prevalence of BRAF-mutated conjunctival nevi and BRAF-mutated CM with origin in a nevus. Also, BRAF-mutated CMs rarely presented with clinical signs of PAM. BRAF mutations were identified in both premalignant paired lesions and CMs. BRAF mutations were, therefore, interpreted to be early events in CM development that mainly occurred in nevoid lesions and persisted if the lesion transformed into a malignant tumour. BRAF mutations have been hypothesized to be early events in the development of cutaneous melanoma, and further malignant progression is probably dependent on several genetic aberrations (Pollock et al. 2003; Dahl & Guldberg 2007; Meyle & Guldberg 2009). Conjunctival melanoma may share these features with cutaneous melanoma, particularly regarding the development of CM within a nevus.

There is evidence to suggest that BRAF mutations have a prognostic role in primary and metastatic cutaneous melanoma (Long et al. 2011; Safaee et al. 2012; Mar et al. 2015), but this is regularly debated. Bulbar location in CM is usually associated with a fair prognosis (Shields 2000; Anastassiou et al. 2002; Tuomaala et al. 2002; Missotten et al. 2005; Shields et al. 2011). Previous studies investigating BRAF mutations in CM have not shown significant associations with prognosis (Gear et al. 2004; Spendlove et al. 2004; Goldenberg-Cohen et al. 2005; Beadling et al. 2008; Populo et al. 2010; Lake et al. 2011; Griewank et al. 2013a; Sheng et al. 2015). In a study by Lake et al. (2011), however, BRAF mutations were identified in 50% of CM metastases.

When investigating possible associations between BRAF-mutation status and outcome, we observed a significant association with the development of distant metastasis in univariate analysis (p = 0.02, log-rank test; I). No definitive association, however, was identified in multivariable Cox proportional hazards regression models (I), which could have been due to the small proportion of tumours tested or the possibility that confounding factors rather than BRAF mutations accounted for the association noted in univariate analysis. Using multivariable Cox regression models on the large population-based dataset (1960–2012), we did not observe any significantly increased risks of local recurrence, metastasis or mortality in BRAF-mutated CM (II). The analysis was possibly confounded by the considerably long follow-up time, which may have influenced on outcome due to changes in treatment practices, and missing FFPE samples – particularly from old cases. In conclusion, large studies are needed to determine the true prognostic significance of BRAF mutations, preferably in more recently diagnosed CM.

Overall, the presence or absence of BRAF mutations was determined in 22 patients who subsequently developed distant metastatic disease during the years 1960–2012. Distant metastases occurred in 18% (13/72) of BRAF-wildtype lesions and in 24% (9/29) of BRAF-mutated lesions. Even though no statistical association with prognosis was observed, a division of the T1 stage based on BRAF-mutation status should be considered, given the high proportion of patients with BRAF-mutated T1 tumours who developed distant metastases.

BRAF inhibitors have recently been investigated in BRAF-mutated CM cell lines, showing promising effects of reduced CM proliferation rates (Riechardt et al. 2015). Apart from BRAF inhibitors being potentially useful for the treatment of advanced CM, BRAF inhibition was found to be efficient as an adjuvant treatment in a case of extensive CM without metastasis (Pahlitzsch et al. 2014; Riechardt et al. 2015). Unfortunately, many cutaneous melanoma patients treated with a BRAF inhibitor develop resistance, causing reactivation of the MAPK pathway (Hertzman & Egyhazi 2014). Recent studies, however, have shown improved response rates and better survival in advanced cutaneous melanoma patients treated with combined BRAF and MEK inhibitor therapy (Dossett et al. 2015). Currently, BRAF-mutation status is routinely tested in patients with advanced cutaneous melanoma in Denmark. However, BRAF status in primary CM or even in premalignant lesions may have important implications for future management. Immunohistochemistry BRAF oncoprotein detection was therefore investigated in the present study as a rapid method of detecting BRAF V600E mutations prior to treatment selection. The IHC identification of BRAF V600E oncoprotein corresponded well with the BRAF V600E mutational status in newer CM samples. This has also been reported for cutaneous melanoma (Lade-Keller et al. 2013; Long et al. 2013). The cases that did not correspond were all from old FFPE tissue blocks, suggesting that degradation of epitopes and/or DNA may cause inaccurate results. However, the lack of correspondence could also be caused by intratumour heterogeneity or in rare cases by detection of a BRAF V600E complex mutation by the IHC method. The less prevalent BRAF V600K-type mutations are not detected by this method. Thus, BRAF V600E oncoprotein detection should be supplemented with further detection methods.

The identification of BRAF, NRAS and KIT mutations in CM has shown a striking variation from uveal melanomas and highlights the genetic resemblance to both cutaneous and mucosal melanoma (I and II; Gear et al. 2004; Spendlove et al. 2004; Goldenberg-Cohen et al. 2005; Beadling et al. 2008; Populo et al. 2010; Lake et al. 2011; Griewank et al. 2013a; Larsen et al. 2015; Sheng et al. 2015). The identification of diagnostic, prognostic or therapeutic miRNAs has the potential to improve the prognosis of both cutaneous melanoma patients and uveal melanoma patients (Worley et al. 2008; Kitago et al. 2009; Chen et al. 2010; Philippidou et al. 2010; Gaziel-Sovran et al. 2011; Poliseno et al. 2012; Streicher et al. 2012; Xu et al. 2012; Sand et al. 2013; Forloni et al. 2014; Qi et al. 2014; Saleiban et al. 2014; Fleming et al. 2015; Pinto et al. 2015). However, to the best of my knowledge, no miRNA expression studies have been performed on CM.

The CM patients included in our miRNA study were representative for CM apart from a slightly older age at diagnosis. We identified 24 up-regulated miRNAs and one down-regulated miRNA in CM. Several of the differentially expressed miRNAs have previously been described in cutaneous melanoma; none have been reported previously in uveal melanoma samples (Worley et al. 2008; Radhakrishnan et al. 2009; Larsen et al. 2014). Five of the up-regulated miRNAs in CM have been reported to have key functions in the development and progression of cutaneous melanoma. These miRNAs may be of prognostic or therapeutic relevance in CM.

Up-regulation of miR-20b-5p (miR-20b), which was previously characterized as an oncomiR in cutaneous melanoma, was observed in CM. MiR-20b expression has been reported to be up-regulated in primary and metastatic cutaneous melanoma (Xu et al. 2012). Moreover, up-regulation of miR-20b leads to suppression of phosphatase and tensin homologue (PTEN) expression in both breast cancer and advanced colorectal cancer (Zhou et al. 2014; Zhu et al. 2014). Phosphatase and tensin homologue is a well-known tumour suppressor that antagonizes the oncogenic phosphatidylinositol 3-kinase (PI3K/Akt) pathway by modulating cell proliferation (Wu et al. 2003; Zhou et al. 2014), and a pronounced loss of nuclear PTEN expression has previously been observed in CM (Westekemper et al. 2011). Another study demonstrated a down-regulation of miR-20b in metastatic cutaneous melanoma, associated with activation of proteinase-activated receptor-1 (PAR-1), which is involved in tumour invasion and angiogenesis and consequently tumour metastasis (Melnikova & Bar-Eli 2009; Saleiban et al. 2014). Due to this dual function of miR-20b, further investigations are needed to establish its prognostic and therapeutic potential in CM.

An interesting observation was the up-regulation of miR-146a-5p (miR-146/miR-146a) and miR-146b-5p (miR-146b) in CM. Up-regulation of miR-146a has been observed in primary and metastatic cutaneous melanoma, and it may be specific for metastatic disease (Philippidou et al. 2010; Qi et al. 2014). MiR-146a promotes both initiation and progression of BRAF/NRAS-mutated cutaneous melanoma through increased activation of the NOTCH protein (Forloni et al. 2014). NOTCH1 is required for melanoma formation and it may enhance the metastatic potential of primary melanoma cells through activation of the mitogen-activated protein kinase (MAPK) pathway or the PI3K/Akt pathway. Up-regulation of miR-146b has been reported in cutaneous melanoma, and appears to be associated with melanoma progression (Chen et al. 2010; Philippidou et al. 2010; Sand et al. 2013; Saleiban et al. 2014). Oncogenic roles for miR-146a and miR-146b are therefore evident in cutaneous melanoma, and these miRNAs may have a similar role in CM.

The observed up-regulation of miR-506-3p (miR-506) and miR-509-3p (miR-509), which belong to the miR-506–514 cluster, has also been described in metastatic cutaneous melanoma (Streicher et al. 2012; Saleiban et al. 2014). In a functional characterization of the miR-506–514 cluster, its inhibition led to reduced cell growth and invasion and to increased apoptosis in melanoma cell lines (Streicher et al. 2012). Inhibition of these miRNAs may therefore be a new approach in the treatment of cutaneous and possibly CM.

To identify potentially prognostic miRNAs, we compared the miRNA expression in TNM stages T1, T2 and T3 with miRNA expression in normal conjunctiva. We observed an up-regulation of seven miRNAs, which overlapped between stages T1 and T2. These miRNAs were significantly associated with increased tumour thickness, which is a known feature of poor prognosis in CM (Folberg et al. 1985a; Fuchs et al. 1989; Lommatzsch et al. 1990; Tuomaala et al. 2007). In particular, the up-regulation of miR-30d-5p (miR-30d) was of interest. A correlation has been reported between expression of this miRNA and increased melanoma thickness, melanoma invasion, metastasis and mortality in cutaneous melanoma (Philippidou et al. 2010; Gaziel-Sovran et al. 2011). In addition, miR-30d targets several messenger RNAs involved in melanoma and/or cancer progression (MITF, ITGA5, SERPINE1 and ADAM19; Fleming et al. 2015). Consequently, miR-30d expression may be a prognostic tool in CM.

The associations of tumour thickness with expression patterns may, however, also reflect that the thinner tumours were more difficult to macro-dissect. Consequently, these samples may have contained a higher percentage of normal tissue that interfered with the expression results. Thus, micro-dissection of CM samples or the use of CM samples >2 mm in thickness is recommended in future studies.

Using fresh frozen tissue, we extended the investigation and compared the expression of the 25 tumour-specific microRNAs in MM and in CM. We found that the miRNA expression patterns were similar in MM and CM. This observation may have been due to the use of fresh frozen tissue, which may have a different expression pattern. This is unlikely, however, since a good correlation between miRNA profiles from fresh frozen and archived tissue samples has already been reported (Glud et al. 2009). The subtypes of melanoma investigated all evolve in mucosal membranes, and the miRNA expression may therefore also reflect shared embryological or tissue-specific characteristics.