Issues affecting molecular staging in the management of patients with melanoma.

Prediction of metastatic potential remains one of the main goals to be pursued in order to better assess the risk subgroups of patients with melanoma. Detection of occult melanoma cells in peripheral blood (circulating metastatic cells [CMC]) or in sentinel lymph nodes (sentinel node metastatic cells [SNMC]), could significantly contribute to better predict survival in melanoma patients. An overview of the numerous published studies indicate the existence of several drawbacks about either the reliability of the approaches for identification of occult melanoma cells or the clinical value of CMC and SNMC as prognostic factors among melanoma patients. In this sense, characterization of the molecular mechanisms involved in development and progression of melanoma (referred to as melanomagenesis) could contribute to better classify the different subsets of melanoma patients. Increasing evidence suggest that melanoma develops as a result of accumulated abnormalities in genetic pathways within the melanocytic lineage. The different molecular mechanisms may have separate roles or cooperate during all evolutionary phases of melanocytic tumourigenesis, generating different subsets of melanoma patients with distinct aggressiveness, clinical behaviour, and response to therapy. All these features associated with either the dissemination of occult metastatic cells or the melanomagenesis might be useful to adequately manage the melanoma patients with different prognosis as well as to better address the different melanoma subsets toward more appropriate therapeutic approaches.

Introduction

The incidence of melanoma is steadily increasing within the Caucasian population [1]. A light phototype (dark-skinned populations show a lower melanoma incidence than fair-skinned populations exposed to similar levels of incident sunlight [2-5]), a large number of acquired common nevi [5-6], the presence of atypical nevi [5-6], or the recurrence of the disease into the family (one-tenth of melanoma patients presents a history of at least one additional affected family member [7-8]) have been associated with an increased risk of melanoma.

Melanoma is usually characterized by a high tendency to develop metastasis; considering the small size of most primary lesions, the metastatic potential of melanoma is considerably greater than that of other solid tumours [9]. The most important factor for reducing the melanoma mortality is early diagnosis, allowing treatment to be undertaken at a stage when cure is readily achievable. Dermoscopy has been demonstrated to be a reliable tool for the differential diagnosis of cutaneous pigmented lesions and, therefore, the early diagnosis of cutaneous melanoma [10]. Actually, Breslow tumour thickness [11], tumour ulceration and metastatic involvement of the regional lymph nodes contribute to define the stage of disease, according to the recent American Joint Committee on Cancer (AJCC) guidelines [12]. Although other prognostic factors may also have a role (including level of invasion, mitotic rate, presence of regression, sex and age of the patient, anatomic site of the primary tumour), stage of disease remains the overriding prognostic factor at present and the survival for melanoma patients is progressively worsening as Breslow thickness of primary tumours increases [13]. Despite an earlier onset of the disease and a significantly increased risk for melanoma in kindreds with recurrent melanoma, a positive family history does not act as an independent prognostic factor and does not seem to influence the overall survival or tumour specific survival [14].

Hence, there is a generalized request for improved methods to predict the clinical outcome in melanoma patients. The hypothesis that a more appropriate evaluation of both the biological behaviour and the molecular mechanisms underlying the tumourigene-sis could improve prognostic prediction and clinical management of melanoma patients is based on at least two issues.(1) The detection of melanoma cells in peripheral blood (circulating metastatic cells [CMC]) [15-19] or in sentinel lymph nodes (sentinel node metastatic cells [SNMC]) [20-24], which represent the first nodes receiving migrating cells from the primary tumour, has been proposed as a potential tool to select patients with higher risk of relapse at the time of the diagnosis.(2) The characterization of the various molecular mechanisms involved in development and progression of melanoma (referred to as melanomagenesis) [6, 25] could represent the second potential tool to identify the molecular profiles underlying aggressiveness, clinical behaviour or response to therapy as well as to better classify the subsets of melanoma patients with different prognosis.

Role of the RT-PCR in prognostic prediction

Metastatic melanoma cells are not found in either circulation or tissue sections of normal individuals. Consequently, detection of melanoma cells in samples from patients at early-stage disease could indicate a dissemination of the tumour cells and, thus, a high risk of development of distant metastases. However, it is to underline that mobilization of cells from the site of the primary lesion is necessary, but not sufficient, to produce distant metastases [26]. Indeed, physical invasion of blood stream by tumour cells is among the earliest events in the tumour progression cascade and many other steps are required for metastatic colonization of distant parenchymas [26]. Detection of CMC can be thus considered as a surrogate marker of the initial events toward the establishment of distant metastases. In this sense, identification of melanoma cells in histologically negative regional lymph nodes could probably represent a more useful marker for staging melanoma patients (as previously suggested by our group [21]). As here pointed out, several controversial data have been reported about the role of the detection of CMC or SNMC among melanoma patients. To date, the main question to be answered is whether detection of CMC or SNMC could really contribute to predict survival in melanoma patients, providing clinicians with new tools for a more accurate staging, more appropriate follow-up schedules and/or more effective adjuvant therapies.

Melanoma micrometastases in peripheral blood

Reverse transcriptase-polymerase chain reaction (RT-PCR) has been demonstrated to detect a single specific messenger RNA (mRNA) in a mixed cell population, becoming a sensitive method for identification of circulating tumour cells [27-30]. As the use of a unique marker could be of limited value in the management of melanoma patients, multi-marker assays including tyrosinase (an enzyme that is involved in the melanin biosynthesis pathway [31]), p97 [32], MUC18 [33], MelanA/MART1 [34], MAGE3 [35] and/or gp100/pMel-17 [36] markers have been proposed, in order to improve sensitivity and specificity of the procedure [37]. Our group has previously reported a positive association between clinical stage and detection of tumour-associated mRNAs in peripheral blood of melanoma patients using a multi-marker RT-PCR assay [38-40]. However, it is strongly debatable whether the use of poorly represented mRNAs may add significant information to tyrosinase as marker highly and specifically expressed in cells of the melanocytic lineage. On the other hand, we cannot exclude that if both the number of positive markers and the probability of developing distant metastases were a function of the amount of circulating cells, a higher number of markers could be more sensitive in identifying the heterogeneous population of metastatic cells. In addition to tyrosinase, previous reports indicated the MelanA/MART1 mRNA as the marker whose RT-PCR amplification may indeed increase sensitivity in CMC detection [40-42].

Many studies have dealt with the possible prognostic value of the presence of a minimal residual disease in melanoma patients who have undergone apparently curative surgery. For CMC, conflicting findings have been reported world-wide (Table 1). About two-fifths of studies were negative, excluding a role for the PCR-based detection of CMC as predictive factor for clinical outcome (though blood detection of mRNAs corresponding to melanoma-associated molecular markers was significantly correlated with the stage of disease in all analysed series) [19, 40, 42–46]. The majority (about 60%) of published studies presented positive conclusions, supporting the prognostic role of such a procedure [37, 41, 47–54]. In other words, the presence of CMC appeared to be at least a surrogate marker for clinical staging (in negative studies) or a putative predictive factor for disease-specific survival (in positive studies).

1

Published studies on prognostic value of RT-PCR assays in peripheral blood

Author, year [ref.]	No.of patients	RT-PCR markers	Disease stages	Median follow-up (months)	Multivariate analysis
Positive studies
Hoon, 1995 [37]	119	Tyr, MAGE3, MUC18, p97	I–IV	6	No
Battayani, 1995 [47]	93	Tyr	I–IV	8	No
Kunter, 1996 [48]	64	Tyr	I–IV	20	No
Mellado, 1996 [49]	91	Tyr	I–IV	18	Yes
Curry, 1998 [50]	123	Tyr, MART1	I–III	18	No
Curry, 1999 [51]	186	Tyr, MART1	I–III	24	Yes
Mellado, 1999 [52]	57	Tyr	I–III	27	No
Schittek, 1999 [41]	225	Tyr, MART1	I–IV	4	No
Hoon, 2000 [53]	46	Tyr, MAGE3, MUC18, p97	I–IV	48	Yes
Schrader, 2000 [54]	31	Tyr, MART1, MAGE3	IV	11	No
Negative studies
Hanekom, 1999 [43]	181	Tyr	I–IV	12	No
Aubin, 2000 [44]	39	Tyr	I–III	10	No
Waldmann, 2001 [45]	20	Tyr	IV	20	No
Strohal, 2001 [46]	76	Tyr, MART1	I–IV	11	No
Brownbridge, 2001 [42]	299	Tyr, MART1	I–IV	12	No
Palmieri, 2003 [40]	200	Tyr, MART1, p97	I–IV	44	Yes
Scoggins, 2006 [19]	1,446	Tyr, MART1, MAGE3, gp100	I–IV	30	Yes

Multi-variate analysis assessing the interaction between the disease stage and the other candidate prognostic factors has not been performed in majority of these studies [37, 41–48, 50, 52, 54]. Although identification of circulating melanoma cells at the time of diagnosis may be indeed the first step toward the subsequent development of metastatic disease, it may also simply reflect the stage of disease at diagnosis. In the five studies that used multi-variate analysis (two [19, 40] negative and the remaining ones [49, 51, 53] positive for association with prognosis), the CMC detection did not act as an independent prognostic factor (its strongest correlation was again with stage at diagnosis). However, conclusions of these studies cannot be considered definitive due to the limited number of analysed cases (less then 100 patients [49, 53]), the short time of observation (median follow-up less than 3 years [19, 51]), or the low rates of events for statistical analysis (due to preponderance of patients presenting a AJCC Stage I disease [40]). Moreover, majority of the published studies was focused on progression-free survival, a surrogate end-point that, although reliable, cannot completely substitute the value of overall survival. Nevertheless, all such studies are substantially retrospective, being based on the analysis of patients who underwent CMC assay; again, selection biases cannot be definitively excluded.

The complex biology of metastasis formation could provide possible explanations for all these controversial data. To become invasive, tumour cells need to change their adhesive properties, to loose contact with other cells in the primary tumour, and make new contacts with the extra-cellular matrix of host cells they encounter as they invade [26, 55]. They also need to be able to penetrate into the surrounding host tissue, where the modulation of prodtease activity in the vicinity of the tumour cells plays a critical role. To migrate away from the primary location, tumour cells also need to gain motility functions. These same properties are also thought to be important when circulating tumour cells exit the circulatory system and start metastatic colonization in secondary organs [55]. Overall, colonization of distant tissues and development of metastasis represent the result of a multi-step cascade of events occurring to cancer cells during tumour dissemination (i.e. viabili-ty in circulation, capability of exiting blood stream and starting tissue invasion, presence of adequate growth potential for metastasis formation) [26, 55].

In this light, a prolonged presence of melanoma cells in blood stream may indeed contribute to select viable cancer cells with better capacity to begin colonization process at the distant site. All the above-mentioned studies were based on RT-PCR analyses of baseline blood samples obtained from melanoma patients at time of diagnosis (generally, early-stage patients have been enrolled if no more than 4 weeks had elapsed from the surgical treatment, whereas those with advanced disease had baseline blood sample collected before systemic therapy). One could speculate that, after an initial peak of circulating cancer cells (strictly correlated with the tumour burden) at the time of surgical excision of the primary melanoma, progressive disappearance of CMC from peripheral blood (i.e. due to lack of their viability in circulation) could be related to the favourable outcome in patients classified as PCR-positive at baseline. Conversely, appearance of CMC in peripheral blood during follow-up (i.e. due to the existence of previously quiescent melanoma cells, which might have acquired the capability of entering the blood stream) could justify the observation of relapses in PCR-negative patients at baseline (Fig. 1).

1

Comparison between circulating metastatic cells (CMC) detection and prognosis. The different hypotheses to explain the controversial data about prognostic value of RT-PCR assays on peripheral blood of melanoma patients at the time of diagnosis (baseline) are provided. On the left, representation of the cancer cells entering (top), surviving (middle) and exiting (bottom) the blood stream.

Supporting this hypothesis, recent studies suggest that metastatic cancer cells might be constantly present in blood circulation in a subset of recurrent melanoma patients, before the establishment of distant metastases (as assessed by detection of frequently positive PCR-based assays in peripheral blood samples serially obtained during follow-up visits) [57-58]. In such studies, both disease-free survival and overall survival were significantly higher for patients always showing negative RT-PCR results in comparison with those who tested positive in more than one RT-PCR assay during a prolonged follow-up observation (about 6 years for each of the two series) [57-58]. Similarly, another study defined the existence of changes in CMCs during interferon (IFN) therapy among melanoma patients with more advanced disease (AJCC stages II–IV) [59]. Using multi-variate analysis, these authors have shown that patients who became CMC-negative during IFN therapy were significantly associated with better disease-free survival than those who remained or became positive during therapy [59]. Dynamic studies, based on routinely repeated evaluations of the presence of circulating malignant cells by RT-PCR assay during the follow-up period or after the treatment administration among patients with melanoma, could therefore represent a way toward the assessment of the threshold above which sensitivity for clinically significant micrometastases can be optimized and false-positives can be minimized.

To date, the clinical role of CMC detection in melanoma patients remains controversial. Although it seems reasonable to wait for more consistent and definitive results (by performing serial RT-PCR assays in large series of patients, with a longer follow-up in order to observe a higher number of melanoma-related deaths), such procedure should be limited to clinical trials and should not enter clinical practice nor affect treatment-decision making.

Melanoma micrometastases in lymph nodes

Development of regional node metastasis is able not only to change tumour staging but also to significantly affect patient survival, becoming a negative prognostic factor in melanoma [12-13]. Early identification of occult metastasis, before the development of clinical disease, is thought to improve survival in melanoma patients. In comparison with the CMC detection, identification of occult melanoma metastasis in regional lymph nodes could be more effective for the assessment of the minimal residual disease after surgical excision of the primary melanoma.

In past years, selective lymphadenectomy has been introduced for the treatment of melanoma by using a lymphatic mapping technique (initially based on intradermic injection of vital blue dye, and then improved by a radio-guided methodology) [60-62]. Intraoperative lymphatic mapping and sentinel node biopsy has become the standard approach for staging the regional lymph nodes in early-stage melanomas [63]. Recently, it has been reported that sentinel-node biopsy may identify subsets of melanoma patients with nodal metastases, whose survival can be prolonged by immediate lym-phadenectomy [64]. Moreover, staging of melanoma patients has been revised with the introduction of a distinction between macroscopic (clinical) and microscopic lymph node involvement [12]. To increase the sensitivity of the morphological analysis based on haematoxylin-eosin staining and improve the detection of occult metastases in sentinel nodes, conventional immunohistochemistry using specific melanocytic markers such as HMB-45, S-100, and/or MelanA/MART1 [39, 63–64] should be associated.

The sentinel node metastatic cells (SNMC) could be however identified by application of highly sensitive molecular approaches. Amplification of tissue-specific mRNA by RT-PCR assays on sections of sentinel lymph nodes may be considered as a further attempt to improve the sensitivity in detecting occult melanoma cells [21, 65–66]. The highest sensitivity of the RT-PCR assay has been reported using fresh or frozen sentinel nodes [65-66]. Since this type of tissue samples are not usually available in majority of cases, our group previously defined protocols to amplify total RNA from paraffin-embedded sentinel nodes of melanoma patients (AJCC stages I-III) [21]. Despite a good consistency in detection of RT-PCR products among frozen and paraffin melanoma sections, the multi-marker assay used in our study (including tyrosinase and MelanA/MART1 mRNAs as markers) failed to identify a quite high fraction (about one third) of sentinel nodes containing melanoma cells as assessed by immunohistochemical analysis [21]. This strongly suggests that sensitivity is deeply lowered when RT-PCR is performed on paraffin-embedded specimens. Also taking into account this weakness of the methodology, our study showed that increasing number of PCR-positive markers in histologically proven tumour-free sentinel nodes from melanoma patients was significantly associated with higher rates of relapses after a median follow-up of about 3 years [21]. Several studies have been performed in order to identify subgroups of melanoma patients with RT-PCR-positive sentinel nodes and high risk of disease recurrence (Table 2). A poorer disease-specific survival has been registered in patients with histopathologically negative sentinel nodes that expressed at least one melanoma-associated mRNA marker [65-70]. When an extended follow-up of patients with histologically negative sentinel nodes has been evaluated, detection of occult melanoma cells by RT-PCR assays seems not to select patients with an increased probability of disease recurrence [71-72].

2

Published studies on prognostic value of RT-PCR assays in hystopathologically negative sentinel lymph nodes

Author, year[ref.]	No. of patients	RT-PCR markers	Median follow-up (months)	Multi-variate analysis
Positive studies
Bostick, 1999 [65]	55	Tyr, MAGE3, MART1	12	Yes
Blaheta, 2000 [66]	101	Tyr	19	Yes
Palmieri, 2001 [21]	61	Tyr, MART1	36	Yes
Denninghoff, 2004 [67]	42	Tyr	37	No
Ulrich, 2004 [68]	288	Tyr	37	Yes
Gradilone, 2004 [69]	129	Tyr	24	Yes
Romanini, 2005 [70]	101	Tyr, MART1	30	No
Negative studies
Kammula, 2004 [71]	97	Tyr	67	Yes
Mangas, 2006 [72]	142	Tyr	45	Yes

Again, knowledge of the biological behaviour of the melanocytic cells could provide some possible explanation for these conflicting results. It is documented that nevus cells can be found within the fibrous capsule and trabeculae of lymph nodes, with tendency to form melanocytic aggregates into the lymph node parenchyma [73-75]. The increased frequency of nodal nevi in sentinel nodes might be due to a mechanical transport of nevus cells [75] or to an active migration of melanocytic cells [76]. Therefore, nevi inclusion occurring in lymph nodes of melanoma patients are not uncommon and may mimic metastases [74-76]. Morphological analysis using haematoxylin-eosin staining, for the assessment of the architectural and phenotypic features of the cell clusters into the lymph node, and immunohistochemistry using melanoma-associated markers seem to be able to differentiate melanocytic aggregates from melanoma metastases in large majority of cases [24, 64]. For such a purpose, the tumour suppressor p16CDKN2A protein has been demonstrated to be a useful immunohistochemical marker [77]. When screening of lymph node sections from melanoma patients is instead performed by RT-PCR assay (which detects one specific cell in 106–107 background cells), using genes specifying for antigens of melanocytic differentiation (mainly, tyrosinase and MelanA/MART1) as markers, it is not possible to determine whether amplified mRNAs come from normal or tumour cells.

In conclusion, RT-PCR analysis has been found to be more sensitive than haematoxylin-eosin staining and immunohistochemistry in detecting cells of the melanocytic lineage within sentinel lymph nodes. Although such a molecular approach could indeed support conventional histopathological analysis (which is retained to underestimate the real incidence of minimal disease) and improve the identification of occult metastases, lack of specificity and limits in the availability of fresh-frozen tissue specimens make this technique impractical for routine use. Therefore, RT-PCR assays on sentinel nodes remain investigational and should not be used to direct adjuvant therapy in melanoma patients at this time. However, studies should be addressed to the enhancement of the PCR-based strategies for detecting occult metastases through either the isolation of more specific molecular targets or a better comprehension of the biological mechanisms underlying the metastasis formation in regional lymph nodes. To date, the Multi-centre Lymphadenectomy Trial II has been designed in order to test the clinical significance of tumour-positive sentinel lymph nodes as inferred by haematoxylin-eosin staining, immuno-histochemistry or RT-PCR assay [24]. Results from this study will provide additional clues about the role of regional lymph nodes into the metastatic process of melanoma patients as well as the clinical value, in terms of costs and benefit, of detecting occult node metastases using the PCR-based technique.

Molecular complexity of melanomagenesis

During the recent past years, an increasing number of evidence has pointed out to the importance of correlating the molecular pathways involved into the de-regulation of melanocyte growth with the clinical and pathological aspects of melanoma [6, 78]. The hypothesis that several genetic and molecular abnormalities co-operating in melanomagenesis do generate different subgroups of patients may have important implications in predicting the clinical outcome as well as planning effective therapeutic strategies in each of these subgroups (especially, in the light of a steadily increasing use of new gene-targeted anticancer molecules in combination with conventional drugs for melanoma treatment).

Melanocytic transformation is thought to occur by sequential accumulation of genetic and molecular alterations [6, 25, 79–80]. Although the pathogenetic mechanisms underlying melanoma development are still largely unknown, several genes and metabolic pathways have been shown to carry molecular alterations in melanoma.

Re-arrangements or deletions of the short arm of chromosome 9 represent a common genetic alteration detected in tumour tissues from patients with melanoma [7, 81]. Molecular and cytogenetic investigations have indicated the CDKN genetic locus at the chromosome 9p21 as a candidate region involved in melanoma pathogenesis [81-82]. The CDKN2A gene maps at the CDKN locus and encodes two proteins: the cyclin-dependent kinase inhibitor p16CDKN2A, which is a component of the CyclinD1-RB pathway, and the tumour suppressor p14CDKN2A, which has been functionally linked to the MDM2-p53 pathway [83] (Fig. 2). Alterations of the CDKN2A gene have been widely reported as the most common cause of inherited susceptibility to melanoma [84-87]. In melanocytic cells, the p16CDKN2A protein acts as a proliferation inhibitor by binding the CDK4/6 kinases and blockingphospho-rylation of the RB protein, which lead cells to cycle arrest [6, 78]. Dysfunction of the proteins involved into the p16CDKN2A pathway has been demonstrated to promote a de-regulation of the cell cycle with an uncontrolled cell growth, which may induce cell proliferation and increase aggressiveness of transformed melanocytic cells [88]. Therefore, genetic (deletion or mutation) [6, 78, 84–87] or epigenetic (gene silencing) [89-90] inactivation of the CDKN2A gene may be required for tumour progression and metastasis formation (melanoma cells tend to inactivate both alleles of such a tumour suppressor gene and increase their aggressiveness and malignancy).

2

Molecular pathways involved in melanomagenesis. The proteins have been ordered according to their position into the functional molecular cascade. Straight arrows and barred lines (in blue) indicate induction and inhibition, respectively; bent arrows (in red) indicate interaction between the different pathways.

Detection of 9p21 allelic loss at similar rates in majority of dysplastic nevi, primary tumours and corresponding metastases indicated that such genetic alterations may play a role in melanocytic proliferation and transformation, being maintained in all evolutionary phases of melanocytic tumourigenesis [91-93]. Allelic deletions at CDKN locus and absence of the p16CDKN2A protein expression have been both observed at increased rates moving from early to advanced primary melanomas and to secondary melanoma lesions as well as to melanoma cell lines, being quite completely absent in non-melanoma melanocytic lesions [88, 91–93]. Inactivating epigenetic mechanisms-the 5′ CpG island upstream of p16 gene has been found methylated in several primary tumours and cell lines [89-90]-may be also responsible for gene silencing. Altogether, data support the hypothesis that inactivation of the p16 gene by different mechanism might be selected during the tumour progression and, especially, during the establishment and propagation of melanoma cells in culture [94].

Another crucial level of cell cycle regulation in melanoma involves the pathway starting from the p14 protein, the second product of the CDKN2A gene (this is an example of a gene encoding two different proteins which generate two different cascades of functional events) [96]. As for p16, the p14 protein exerts a tumour suppressor effect by inhibiting the oncogenic actions of the downstream MDM2 protein, whose direct interaction with p53 blocks any p53-mediated activity and targets the p53 protein for rapid degradation [95] (Fig. 2). The MDM2 gene itself has been shown to be amplified in primary tumours [96] and to act as an oncogene in cell cultures [97]. Mutations in p14 gene are much less frequent than those in p16 gene [7–8, 79]; they allow degradation of p53 by releasing its binding partner MDM2 [94-95] (Fig. 2). Analogously, the TP53 gene is mainly inactivated at functional level and rarely mutated in melanoma [83]. Impairment of the p14CDKN2A-MDM2-p53 cascade, whose final effectors are the Bax/Bcl-2 proteins, is implicated in defective apoptotic response to genotox-ic damage and, thus, to anticancer agents (in most cases, melanoma cells present concurrent high expression levels of Bax/Bcl-2 proteins, contributing to further increase their aggressiveness and refractoriness to therapy [98]) [94, 98–99] (Fig. 2).

The MAPK-ERK pathway (including the cascade of NRAS, BRAF, MEK1/2 and ERK1/2 gene products) has been also reported to play a major role in both development and progression of melanoma [100-101]. Constitutive activating mutations in NRAS occur in about 20% of melanoma cell lines [102-103], whereas oncogenic BRAF mutations have been described in 30–60% of primary melanomas [100, 104–106]. The BRAF gene codify for a serine/threonine kinase of the RAS/RAF/MEK/ERK mitogen-activated protein kinase (MAPK) pathway, a major signalling cascade involved in the control of cell growth, proliferation and migration [100] (Fig. 2). The ERK1/2 proteins, which represent the final components of such a signalling kinase cascade, have been found to be constitutively activated in melanomas, mostly as a consequence of mutations in upstream components of the pathway [101]. The increased activity of ERK seems to be implicated in rapid melanoma cell growth, enhanced cell survival and resistance to apoptosis [101]. This high basal level of ERK activity may further induce the metastatic potential of melanoma by increasing the expression of invasion-promoting integrins [107]. Presence of BRAF mutations in benign and dysplastic nevi supports the hypothesis that activation of the RAF/MEK/ERK pathway is an early event in melanoma progression and that additional co-operating genetic events are required to achieve full malignancy [108].

All these findings clearly indicate the existence of a complex molecular machinery that provides checks and balances in normal melanocytes. Progression from normal melanocytes to malignant metastatic cell in melanoma patients is the result of a combination of down- or up-regulations of the various effectors acting into the different molecular pathways (Fig. 3). The main interactions underlying the melanomagen-esis could be summarized as follow.

3

Proposed model of melanocytic tumourigenesis. The main molecular alterations underlying each step from melanocytic proliferation to advanced melanoma are indicated.

The 9p21 deletions seem to be set in majority of dysplastic nevi, which have been indicated as pre-neoplastic melanoma lesions, and their incidence does not vary during melanocytic tumouri-genesis [88, 91–93]. This implicates that, although an important step for the initiation of neoplastic transformation, 9p21 allelic loss is insufficient for the development of melanoma. Similarly, BRAF somatic mutations were demonstrated to be present at quite identical rates in nevi, in primary and metastatic melanomas as well as in cultured melanoma cells [100, 104–106, 108]. This strongly suggests that BRAF mutations may co-operate with the 9p21 loss into the initial steps of melanocytic proliferation and melanoma formation.

Activated BRAF may induce formation of melanocyte lesions that rapidly developed into invasive melanomas in transgenic p53-deficient zebrafish[109]. This is a further evidence that activating BRAF mutations have a role in melanocytic proliferations, being a necessary event for starting the cascade of alterations involved into the melanomagenesis. However, BRAF activation alone is not sufficient to induce the malignant process and fully transform proliferating melanocytes, but requires additional, co-operating de-regulative events (such as the inactivation of the p53 pathway) for tumour development.

Activating BRAF mutations have been reported to constitutively induce up-regulation of p16CDKN2A (this phenomenon looks like a sort of protective response to an inappropriate mitogenic signal) [110]. As for the p53 deficiency, a genetic or epi-genetic inactivation of p16CDKN2A gene may strongly contribute to malignant transformation of the BRAF-driven melanocytic proliferation;

The microphthalmia-associated transcription factor (MITF) gene, which is activated by a constitu-tive induction of the MAPK-ERK pathway, has been demonstrated to participate in regulation of cell cycle progression in normal melanocytes [111]. The MITF protein seems to exert this effect by co-operating with either p21CDKN1A, which is a downstream effector of the p14CDKN2A-MDM2-p53 cascade, or pRB, which is the final target of the p16CDKN2A-CyclinD1 cascade [111]. A MITF gene amplification has been reported in melanoma metastases and cell lines, suggesting a role in melanoma progression [112].

Molecular alterations and disease phenotype

The complexity of the sequential accumulation of the above-mentioned molecular alterations, during the development and progression of melanoma, raises the question whether several distinct types of melanoma might exist.

Considering the growth patterns, four histological types of melanoma have been described: superficial spreading, lentigo maligna, nodular and acral melanoma [12]. However, justification for such a distinction is controversial because of histological overlap and lack of prognostic significance [13]. Conversely, different studies have shown that melanomas of the palms and soles (acral melanomas) have distinctive patterns of chromosomal aberrations as compared with those at other sites [113-114]. Comparative genomic hybridization revealed that several genomic regions (mostly, 11q13, 22q11–13, and 5p15) were abnormally amplified in acral melanoma [113-114]; such regions were different from those found altered in superficial spreading or nodular melanoma (mainly, 9p21 and 1p22) [7, 81, 88]. In addition, patients with melanomas of the head and neck differ from patients with melanomas of the trunk in having higher expression levels of p53 protein (TP53-positive melanomas), a higher frequency of associated non-melanoma skin cancers, and lower numbers of melanocytic nevi [25, 80, 94, 115]. By contrast, TP53-negative melanomas were related to high nevus density [78, 115].

Considering the genesis at cutaneous level, most melanomas seem to directly arise from normal melanocytes [6]. However, an increasing number of evidence indicates that (a) melanomas progress from pre-existing melanocytic nevi and (b) dysplastic nevi may be considered as precursors of melanoma [7, 80, 94]. These observations suggest that melanoma may arise from at least two pathways. Evaluation of the features underlying the risk of developing melanoma provide clues that two additional phenotypic expressions may be associated with the disease. For the same level of fairness of the skin, melanomas can arise either in individuals who are prone to freckle with an inability to tan and few melanocytic nevi or in those who present instead an elevated number of such nevi [25, 115]. Again, the involvement of at least two pathogenetic pathways might be hypothesized.

Finally, recent findings support the existence of a dual pathway for the development of melanoma: one related to chronic exposure to the sun and the other related to melanocyte instability [78, 115]. It has been reported that melanomas on skin not chronically exposed to sun usually carry either a mutated NRAS or mutated BRAF or concurrently mutated BRAF and PTEN genes [116-117]. Indeed, recent evidence suggest that BRAF and NRAS mutation are mutually exclusive at the single-cell level providing further support to the hypothesis that expression of the two mutations may not occur in the same neoplastic cell [117]. However, activating mutations of both NRAS and BRAF genes may differently segregate into the various cells from the same melanoma; 4etherefore, these mutations can co-exist in the same human melanoma [117]. Since BRAF- and NRAS-mutated clones have been demonstrated to possess different biological properties in vitro [117], co-existence of the two alterations into the same tumour may result in a heterogeneous response to therapy. In the same group of lesions, BRAF mutations have been associated with specific sequence variants of the melanocortin 1 receptor (MC1R) gene [118]. In contrast, melanomas on skin chronically exposed to the sun or on acral skin generally present wild-type BRAF or N-RAS genes with subsequent lack of involvement of the RAS–RAF–ERK pathway [78]. These tumours have instead a genomic instability with an increased number of copies of the proliferation-controlling CyclinD1 or CDK4 genes [78], which belong to the p16CDKN2A-RB cascade [94-95].

All these evidence represent a strong indication that the different molecular pathways associated with the melanomagenesis may correspond to different subsets of melanoma patients, with distinguished biological and clinical behaviour of the disease. Identification of such different patients’ subsets should be introduced in clinical trials by addressing tissue sections from each melanoma patient to molecular analyses: immunohistochemistry using antibodies against the main candidate proteins (p16CDKN2A, p14CDKN2A, pERK1/2, pRB, p10, p53, CyclinD1), in order to assess any alteration of their expression levels, and/or fluorescence in situ hybridization (FISH) analysis, in order to evaluate the existence of pathogenetic gene amplifications (for MITF, CyclinD1 or CDK4 genes).

Conclusions

Molecular classification of melanoma patients could be therefore achieved through the assessment of either the molecular profile of primary tumours (indicating which gene or pathway is affected) or the level of disease dissemination (indicating the presence of melanoma cells into blood stream or regional lymph nodes). This however raises a question:’which is the most appropriate management or the best therapeutic approach for the different subsets of melanoma patients coming from such a molecular staging?’.

Multiple clinical trials on adjuvant therapy have been carried out in melanoma patients using chemotherapy, vaccines, biological drugs, or combinations of these [119]. The only substance yet shown to affect disease behaviour, reproducibly in large randomized controlled clinical trials, is high-dose intravenous IFN-α[119] (low-dose or intermediate-dose IFN offered no overall survival benefit in several trials [120]). However, after several years in which IFN has been utilized in melanoma therapy the real mechanism of action is as yet unknown [121]. None of the trials based on different dosage, route of administration and duration of IFN contained biological end-points aimed to better understand the activity of the molecule [119-121]. Also for this specific purpose, a new patient classification, which would take into account the molecular profile of each melanoma, could be indeed helpful to determine patients who may most benefit from IFN therapy, deeply contributing to shed light on how to overcome IFN resistance or enhance IFN effectiveness [122].

To speak in more general terms, the recent introduction of gene-targeted anticancer molecules in combination with conventional drugs into the clinical practice further support the idea that is time for a more appropriate selection of patients to be addressed to the various innovative therapies [123]. It is unlikely that targeting a single component in the signalling pathway will yield significant antitumour responses. For this purpose, analysis of all known molecular targets could help us to make a prediction, identifying the subsets of patients who would be expected to be more or less likely to respond to specific therapeutic interventions. Nevertheless, several new genes and molecular pathways are being discovered through gene expression profiling based on microarray technology [124-128], making correlations between molecular signatures and clinical outcome [126-128]. These findings along with the advancements of the biotechnologies will do provide even more reliable tools for detailed gene-based analyses, allowing to better characterize molecular biomarkers which may predict prognosis and response to treatment in patients with melanoma.