Introduction
In recent years, the incidence and mortality rate of melanoma have considerably increased (1), whereby the need to discover new therapeutic strategies has become increasingly urgent (2–6). E-cadherin has been shown to play a key role in cell adhesion, growth and differentiation (7), frequently overexpressed in melanoma. Several studies have revealed that E-cadherin is expressed in the normal epidermis and lost in invasive and metastatic melanoma cells (8). Depletion of functional E-cadherin induces the escape of melanocytes from keratinocyte-mediated growth and phenotypic control, thus allowing invasion and migration (9).
Although reduced expression of E-cadherin is known to be caused by promoter hypermethylation in several types of tumors (10), such an epigenetic event has been only partially investigated in melanoma. Moreover, the existence of a relationship between epigenetic changes, E-cadherin expression and clinicopathological features has not yet been evaluated in melanoma.
In the present study, we aimed to investigate whether E-cadherin expression is epigenetically regulated in a cohort of patients affected by cutaneous, mucosal and uveal melanoma and is associated with specific clinicopathological parameters. Findings from the present study revealed a strong association between E-cadherin promoter methylation and several significant pathological characteristics, whereby they may provide insights into the biology of this cancer.
Materials and methods
Cell cultures
Normal human epidermal melanocyte (NHEM) cells were grown in Melanocyte Medium plus Bullet kit (both from Lonza, Walkersville, MD, USA). Cutaneous (G361, WM-115 and WM-266-4), mucosal (MMel-1 and MMel-2) and uveal (OCM-1, OCM-3 and 92.1) melanoma cells were grown as previously described (11,12).
Treatment with the DNA demethylating agent 5-aza-dC
Cells were treated with 5-aza-2′-deoxycytidine (5-aza-dC) (Sigma Chemical Co.) by addition of fresh medium containing 5-aza-dC (10 µmol/l) every day for three consecutive days.
Tissue specimens
Formalin-fixed paraffin-embedded (FFPE) tissue sections of 130 cutaneous, 82 mucosal and 64 uveal melanomas, and 65 normal skin specimens were collected from the Department of Human Pathology, University of Messina, Messina, Italy. The data regarding patients are shown in Table I. The investigation adhered to the Declaration of Helsinki and was approved by the Ethics Committee of the University Hospital of Messina. Informed consent was provided by the patients.
Determination of mitotic index
Mitotic index was determined by counting the number of mitoses in 10 consecutive non-overlapping high power fields (HPFs) with commencement in an area of high mitotic activity using a magnification of ×400.
DNA and RNA extraction
Total RNA and DNA extraction from cells was performed using TRIzol reagent (Invitrogen), and Recover All Total Nucleic Acid Isolation kit (Ambion Inc., Austin, TX, USA) was used for extractin from FFPE samples.
Reverse transcription and qPCR
Total RNA was reverse-transcribed with IMProm-II™ Reverse Transcriptase kit (Promega, Milan, Italy). qPCR was performed using the ABI Prism 7500 Real-Time PCR system (Applied Biosystems, Milan, Italy). Primers and probes were previously described (11). The mRNA levels of E-cadherin were normalized to endogenous β-actin (Applied Biosystems). The control was represented by NHEM cells (described in the paragraph 'Cell culture'). Their expression was considered as00201.
Bisulfite modification and MSP
Bisulfite-modified DNA obtained using the EpiTect Bisulfite kit (Qiagen, Milan, Italy) was amplified using the following primers as previously reported (13): methylated DNA-specific primers: forward primer, 5′-TTAGGTTAGAGGGTTATCGCGT-3′ and reverse primer, 5′-TAACTAAAAATTCACCTACCGAC-3′ (115 bp; genomic position relative to TSS, -176/-61); unmethylated DNA-specific primers: forward primer, 5′-TAATTTTAGGTT AGAGGGTTATTGT-3′ and reverse primer, 5′-CACAACCAA TCAACAACACA-3′ (97 bp; genomic position relative to TSS, -181/-84). PCR products were separated by 2% agarose gel containing ethidium bromide.
Bisulfite genomic sequencing
Bisulfite-treated DNA was amplified by PCR with primers that were specific for modified DNA but did not contain any CpG sites in their sequence. The primers were: S1 (TTT AGT AAT TTT AGG TTA GAG GGT T, upstream, nt 836–861; GeneBank accession no. L34545) and S2 (CTA ATT AAC TAA AAA TTC ACC TAC C, downstream, sequence position nt 965–940) (14). The PCR conditions were 94°C for 2 min; 35 cycles of 94°C for 20 sec, 48°C for 20 sec and 72°C for 30 sec; and a final extension at 72°C for 5 min. The PCR product was extracted from the gel with the QIAquick Gel Extraction kit (Qiagen). The purified DNA samples were sequenced with the CEQ DTCS Quick Start kit, and with an automated DNA sequencer (Beckman Coulter CEQ 2000 analysis system) (both from Beckman Coulter, S.p.A.).
Quantification of methylation levels
A single cytosine signal at the corresponding CpG site was considered 100% methylation, a single thymine signal was considered no methylation, and overlapping cytosine plus thymine signals were considered partial methylation. In the latter condition, the percentage of methylation was expressed as the ratio of the peak values of the cytosine to cytosine plus thymine signals.
Transient transfections
For transient knockdown of DNMT1, DNMT3a and DNMT3b, cells were transfected with specific targeting small interfering RNA (siRNA) or non-targeting control siRNA (Invitrogen, Milan, Italy) at a final concentration of 100 nM 24 h after plating using siPORT Lipid Transfection Agent (Ambion, Milan, Italy).
Invasion assay
The anti-invasive activity of 5-aza-dC was assessed using the Cultrex® BME Cell Invasion assay (Trevigen, Gaithersburg, MD, USA).
Statistical analysis
The Pearson's correlation test was used to assess the association of E-cadherin expression with promoter methylation. Differences in E-cadherin expression levels and clinical characteristics were evaluated by χ2 test between patient subgroups. Kaplan-Meier method was used in the evaluation of the overall and disease-free survival time. The log-rank test was used in comparing the differences between the periods of survival among the examined patients. A p-value of <0.05 was considered to indicate a statistically significant result.
Results
E-cadherin expression correlates with promoter methylation in cutaneous, uveal and mucosal melanoma
E-cadherin mRNA values were firstly measured in 65 normal skin samples from healthy donors and in NHEM cells in order to determine the cut-off point for abnormal E-cadherin expression. E-cadherin levels (as defined by the ratio between the values measured in skin samples over those of NHEM cells) ranged between 1.52 and 2.1 (mean, 1.84±0.133). Values equal or below 1.44 (determined as the mean minus 3 SD) were considered to represent under-expression of E-cadherin. As shown in Fig. 1, a strong reduction or loss of expression of E-cadherin was detected in 55/130 cutaneous (42.3%), 36/64 uveal (56.2%) and 49/82 mucosal (59.7%) melanoma samples. In order to investigate whether promoter hypermethylation correlated with E-cadherin expression, we performed MSP analysis in the series of melanoma specimens examined. Of the 55 cutaneous melanomas down-expressing E-cadherin, 51 samples were hypermethylated (92.7%) and 4 samples were unmethylated (7.3%). All of the 75 cutaneous melanoma sections which expressed amounts of E-cadherin mRNA above the cut-off point were unmethylated. Of the 36 uveal melanoma samples exhibiting levels of mRNA under the cut-off point, 33 (91.7%) were methylated and 3 (8.3%) were unmethylated, whereas only 6 out of 28 (21.4%) specimens exhibiting consistent amounts of E-cadherin expression were methylated. Among the 49 mucosal melanomas that expressed lower mRNA levels of E-cadherin, 46 specimens (93.9%) exhibited promoter methylation and 3 specimens (6.1%) were unmethylated. Six out of 33 (18.2%) mucosal melanoma samples with high levels of E-cadherin were found to be methylated while the rest were unmethylated. Pearson correlation analysis revealed that E-cadherin mRNA levels were inversely correlated with promoter methylation (for cutaneous melanoma, correlation coefficient, −0.899, p<0.001; for uveal melanoma, correlation coefficient, −0.729, p<0.001; for mucosal melanoma, correlation coefficient, −0.754, p<0.001).
Sodium bisulfite DNA sequencing was carried out to assess the promoter methylation pattern for 10 specific CpG sites (genomic positions nt 863, 865, 873, 879, 887, 892, 901, 918, 920 and 940; GeneBank accession no. L34545), as represented in Fig. 2A. Fig. 2B shows the percentage of methylation of E-cadherin in the normal skin and melanoma samples. Cancer samples showed significantly higher methylation levels compared with the normal tissues (p<0.001). Notably, the CpG sites nt 892 and 940 were 90–100% methylated in all the melanoma types examined and the others were partially methylated (range, 53–86%). The methylation rate of the E-cadherin gene was low in the normal tissues (range, 5–24%).
To test whether promoter methylation was responsible for repression of E-cadherin transcription, cutaneous (G361, WM-266-4 and WM-155) and uveal (OCM-1, OCM-3 and 92.1) melanoma cell lines previously found to be E-cadherin-methylated (11) as well as primary mucosal melanoma cells (M-Mel-1 and M-Mel-2) that are methylated for this gene (data not shown), were treated with the demethylating drug 5-aza-dC or transfected with siRNAs targeting DNMT1, DNMT3a or DNMT3b. As shown in Fig. 3, E-cadherin mRNA levels were significantly increased following 5-aza-dC treatment or DNMT1 silencing. In contrast, DNMT3a or DNMT3b knockdown did not restore E-cadherin expression.
Correlation between E-cadherin expression and clinical parameters in cutaneous, mucosal and uveal melanoma
We next examined the correlation between E-cadherin expression and different clinicopathological features. Low E-cadherin expression was observed in 78.5% of the head and neck cutaneous melanomas as compared to 12.8% of melanomas occurring in other regions (p<0.0001; χ2 test). Ulcerated lesions exhibited reduced levels of E-cadherin in 95.7% of the cases relative to 12.1% observed in the non-ulcerated tumors (p<0.0001; χ2 test). A high percentage of melanomas with elevated mitotic index (84.2% for 11–15 mitoses/10 HPFs; 92.3% for 15–20 mitoses/10 HPFs) or with metastasis (68.75%) showed decreased amounts of E-cadherin as compared with melanomas with a low number of mitoses (3.5% for 1–5 mitoses/10 HPFs; 4.7% for 6–10 mitoses/10 HPFs) (p<0.0001; χ2 test), or no metastatic melanomas (12.1%; p<0.0001; χ2 test), respectively. No statistical differences were found between E-cadherin expression and clinical stage or Breslow thickness (Fig. 4A). Mucosal melanoma presented a similar pattern, since a very significant correlation was found between decreased E-cadherin expression and head/neck localization (p<0.0001; χ2 test), ulceration (p<0.0001; χ2 test), mitotic index (p<0.0001; χ2 test) and metastasis (p<0.0001; χ2 test) (Fig. 4B). Data relative to uveal melanoma showed that a high frequency of reduced E-cadherin levels occurred in melanomas of the choroid (p<0.0001; χ2 test) as well as in those with elevated mitotic index (p=0.002; χ2 test) or metastasis (p<0.0001; χ2 test) (Fig. 4C).
E-cadherin expression discriminates site-specific melanoma metastasis
Analysis of the expression of E-cadherin in metastatic tumor according to the metastasis site revealed that the reduction in E-cadherin expression was significantly associated with lymph node metastatic localization (Fig. 5).
Restoration of E-cadherin negatively correlates with the invasive potential of melanoma cells
Since a relationship between E-cadherin silencing and melanoma metastasis was noted (Fig. 4), we aimed at evaluating the effect of 5-aza-dC-induced E-cadherin reactivation on melanoma cell invasiveness. As shown in Fig. 6, demethylation markedly inhibited the invasion of all the examined melanoma cell lines.
Correlation between E-cadherin levels and patient survival
The overall and disease-free survival of melanoma patients were also explored and the association with E-cadherin expression was analyzed. As shown in Fig. 7, a shorter survival time for patients with reduced levels of E-cadherin was documented (p<0.0001). Compared to the overall survival time for patients with preserved levels of E-cadherin (from 2 to 60 months, average 32.3 months, median 36 months in cutaneous melanoma; from 2 months to 60 months, average 33.2 months, median 36 months in mucosal melanoma; from 2 to 60 months, average 25.5 months, median 24 months in uveal melanoma), a shorter overall survival time in patients with reduced expression of E-cadherin was observed (from 2 to 52 months, average 19.4 months, median 20 months in cutaneous melanoma; from 2 to 44 months, average 19.2 months, median 24 months in mucosal melanoma; from 2 to 28 months, average 15.2 months, median 14 months in uveal melanoma). Disease-free survival time was also shorter in patients with loss of expression of E-cadherin, compared with the other patients. In the presence of E-cadherin expression it was: 2 months, 60 months, average 32.7 months, median 28 months in patients with cutaneous melanoma; 2 months, 60 months, average 31.6 months, median 28 months in patients with mucosal melanoma; 2 months 60 months, average 31.9 months, median 32 months in patients with uveal melanoma. In the rest of the patients it was: 2 months, 40 months, average 21.4 months, median 20 months in patients with cutaneous melanoma; 2 months, 56 months, average 21.5 months, median 20 months in patients with mucosal melanoma; 2 months, 36 months, average 16.9 months, median 16 months in patients with uveal melanoma.
Discussion
The role of E-cadherin as a 'suppressor of invasion' (15) has been well established in the context of melanoma, since its re-expression was accompanied by suppression of tumor invasiveness and metastasis (16). Moreover, the significant role played by E-cadherin in melanoma was very recently evidenced by observations that the resistance to BRAF inhibitors displayed by melanomas harboring mutations in BRAF was strictly associated with a significant decrease in E-cadherin expression (17). However, although reduced E-cadherin expression has been well documented in malignant melanoma and is associated with specific features of the tumor (17–19), to the best of our knowledge, the underlying mechanisms have not been extensively determined. In fact, in only two cellular melanoma models the relationship between E-cadherin silencing and promoter methylation was evidenced (19,20). In the present study, we showed that in cutaneous, uveal, and mucosal melanomas the methylation at specific CpG sites in the E-cadherin promoter was responsible for reduced gene expression (Figs. 1 and 2) and several clinicopathological features. Indeed, E-cadherin downregulation was assumed to be an important step in the progression from a melanoma in situ to an invasive and metastasizing lesion and a determinant of tumor depth (8,21). Thus, the demonstration that E-cadherin expression is epigenetically regulated through promoter methylation and that it can be reverted by demethylating tools, such as 5-aza-dC treatment or DNMT1 silencing it may be of clinical significance (Fig. 3). This latter data accounts for the prominent role of DNMT1 in repressing E-cadherin transcription, as already shown in another cancer model (22), and is in agreement with studies reporting the near selective activity of 5-aza-dC in inhibiting DNMT1 and neglecting DNMT3a and DNMT3b enzymes (23,24). The identification of a likely mechanism for E-cadherin downregulation in melanoma may pave the way to alternative therapeutic strategies. This possibility is further suggested by additional results of the present study, which indicate that methylation and reduced expression of E-cadherin are closely associated with clinicopathological markers, either specific or common to each type of melanoma. This may be of some interest, since up-to-date conflicting data emerge from the literature. Initially, E-cadherin expression was shown to be rare in both nevocellular nevi and early-stage melanoma, but preserved in advanced-stage melanoma and melanoma metastases (25). More recently, however, E-cadherin expression was found to be altered in different stages of melanoma progression (19). Moreover, information with regard to the actual meaning of E-cadherin epigenetic silencing in melanoma was lacking. In the present study, we showed that in cutaneous melanoma reduced E-cadherin expression via promoter aberrant methylation was strongly associated with several poor prognostic factors, namely ulceration, head/neck localization, mitotic count and metastasis as well as with reduced overall/disease-free survival (Figs. 4 and 7). No statistical association was found with clinical stage or Breslow thickness. Our findings allow us to hypothesize that reduced E-cadherin expression through promoter hypermethylation may have a major impact on the proliferation rate of melanoma. The inverse association between E-cadherin expression and the mitotic index found by us is in line with various studies, which reveal that downregulation of E-cadherin prevents the control of melanocyte proliferation by keratinocytes and is involved in melanocyte transformation (16,26). Moreover, other studies have reported that expression of E-cadherin results in inhibition of cell motility and local invasion in melanoma cells (16,27), as well as in other cell systems (28,29). The higher frequency of E-cadherin methylation and silencing in cutaneous melanoma occurring in head and neck regions, as detected in the present study, deserves to be considered separately. The incidence of melanoma of the head and neck has increased in proportion to sun exposure, and it is often linked to poor prognosis and reduced survival (30). Intense vascularization and abundance of lymphatic vessels in the region may account for these malignancy characteristics, but although it is generally believed that molecular mechanisms different from those involved in melanoma occurring in sun-protected areas are engaged (31,32), little information about distinctive molecular markers of head/neck melanoma are available. The positive correlation between the methylation of E-cadherin and the location of melanoma in the head and neck region highlighted here may explain the apparent discrepancies with other studies reporting a low frequency of methylation of the gene in this tumor. Indeed, in case studies by Liu et al, melanomas of the head and neck were not included (33). The silencing of E-cadherin through aberrant promoter methylation may contribute to add new insights into the mechanisms accounting for the higher malignancy of head/neck melanoma. A similar hypothesis may be advanced for head/neck mucosal melanomas, which, although rare (34), represent half of all mucosal melanomas and exhibit a more aggressive behavior as compared to mucosal melanoma of other regions (35). Indeed, it is quite intriguing that, among the various types of mucosal melanoma, methylation-mediated downregulation of E-cadherin is almost entirely restricted to the head and neck area (Fig. 4). Likewise, we showed that the presence of ulcerated lesions, high mitotic index, metastasis and reduced overall/disease-free survival are conditions frequently associated with epigenetic downregulation of E-cadherin in mucosal melanomas (Figs. 4 and 7). These data allow us to hypothesize that cutaneous and mucosal melanomas not only use the same mechanism for regulating E-cadherin expression, but also that both adopt it in the same circumstances of transformation. Regarding uveal melanoma, it is worth mentioning that the expression of E-cadherin, along with other biomarkers, is considered to be the most accurate predictor of metastasis (36). Therefore, our findings that the downregulation of E-cadherin correlates inversely with the mitotic index and directly with the choroidal site, metastasis and reduced overall/disease-free survival (Figs. 4 and 7) are not surprising, albeit indicative from a clinicopathological point of view. In fact, it is well recognized that high expression of E-cadherin is strictly associated to the epithelial phenotype, while loss of E-cadherin is indicative of mesenchymal transition and acquirement of invasiveness characteristics (37,38). In addition, it is even more interesting that all the melanoma types share the association of methylation-induced E-cadherin reduction with the metastatic phenotype (Fig. 4), the lymph nodal metastasis localization (Fig. 5), the invasive character (Fig. 6), and the reduced overall/disease-free survival (Fig. 7). Such data, streaming from the well known role exerted by E-cadherin as cell-cell adhesion molecule and central hub of inhibitory signals of cell motility (39,40), contribute to clarify why melanoma preferentially metastasizes to distant sites and lymph nodes. Taken together, the results reported in the present study highlight the crucial role of E-cadherin in the events affecting melanoma progression and support the proposal that E-cadherin itself may be considered as a prognostic factor for melanoma.
In conclusion, in the light of the above considerations, the findings reported in the present study, indicate that promoter hypermethylation plays a significant role in switching off E-cadherin gene expression in malignant melanoma and suggest that loss of E-cadherin may contribute to increased cell proliferation, invasiveness, ulcerative process, metastasis, prevalently at lymph nodes, head and neck or choroidal localization, and reduced disease-free/overall survival. Taken together, the findings reported in the present study suggest that E-cadherin downregulation may itself be considered a poor prognostic factor in melanoma and raise the possibility that reactivation of E-cadherin through promoter demethylation may represent a promising therapeutic strategy for the treatment of melanoma.