Down-regulation of CYLD expression by Snail promotes tumor progression in malignant melanoma

To get a first insight into the role of CYLD in malignant melanoma, we evaluated-CYLD expression in six different human melanoma cell lines, as well as in freshly isolated human melanoma cells. We found strong reduction of CYLD mRNA (Fig. 1 A) and protein (Fig. 1 B) levels when compared with normal human epidermal melanocytes (NHEMs). Also, in situ, primary melanomas and melanoma metastases displayed strongly down-regulated or absent CYLD mRNA (Fig. 1 C) and protein (Fig. 1 D). In contrast, melanocytes in normal epidermis displayed high CYLD expression, as confirmed by costaining for tyrosinase and CYLD and quantification of costaining (Fig. 1, E and F). Because mutations and promoter methylation of the CYLD gene could be excluded (unpublished data), we concluded that down-regulation of CYLD likely occurs at the transcriptional level.

Analysis of the CYLD promoter revealed three potential binding sites for the transcriptional repressor Snail1 (Fig. S1, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1). We found that Snail1 messenger RNA (mRNA) was highly increased in melanoma cell lines compared with NHEM, an expression pattern that correlated inversely with CYLD (Fig. S2 A vs. Fig. 1 A). In contrast, expression of Snail2 (Snai2), a Snail1 homologue that has been shown to affect expression of several genes important for melanoma progression (21), in melanoma was similar or lower than in NHEM (Fig. S2 A).

Chromatin immunoprecipitation (ChIP) assays demonstrated Snail1 binding to the CYLD promoter in primary melanoma cells and the melanoma cell lines Mel Im and Mel Juso, but not in NHEM (Fig. 2 A). Furthermore, neither Snail2 nor Twist (another E-box binding factor associated with poor prognosis in melanoma [22]) bound to the CYLD promoter in ChIP assays (Fig. 2 A). Snail1 recruitment to the CYLD promoter was observed to the first E-box at position −19 to −14, which resembles the consensus Snail1 binding sequence (CAGGTG; Fig. S1), but not to E-box II (-382 to -377) or III (−429 to −424; unpublished data).

Reporter gene assays using the CYLD promoter (−550 to −1) revealed inducible activity in melanoma cells after transient transfection with an antisense Snail1 construct (as-Snail). Conversely, transient transfection of a Snail1 expression construct completely repressed CYLD promoter activity (Fig. 2 B). Mutation of the first E-box of the CYLD promoter at position −19 resulted in strong promoter activity in Mel Im cells, which was not affected by transient transfection with asSnail or Snail1 (Fig. 2 B). In contrast, mutations in E-box II or III of the CYLD promoter did not change the reporter activity (unpublished data).

These data indicate that recruitment of Snail1 to the first E-box binding site located in the CYLD promoter suppresses CYLD promoter activity in malignant melanoma.

To further assess the effect of Snail1 on CYLD expression in melanoma cells, we either stably transfected Mel Im cells with asSnail expression plasmids or depleted Snail1 with specific siRNAs, and found a strong up-regulation of CYLD protein (Fig. 2 C). Furthermore, transient expression of as-Snail expression plasmids was sufficient to induce CYLD expression in melanoma cell lines and primary human melanoma cells (Fig. 2 D and Fig. S2 B). Accordingly, transient transfection of NHEM with a Snail1 expression plasmid strongly inhibited CYLD expression (Fig. 2 E).

Snail1 transcription depends on ERK signaling in squamous cell carcinoma and epithelial cells (15, 23), and melanoma cells exhibit high ERK activity compared with NHEMs (24). Interestingly, we found that pharmacological inhibition of the ERK/mitogen-activated protein kinase (MAPK) cascade induced an almost complete loss of Snail1 mRNA expression and significantly increased CYLD mRNA levels in melanoma cells (Fig. 2 F). In accordance, treatment of NHEM with ERK inhibitors to further reduce basal levels of ERK activity also increased expression of CYLD (Fig. 2 G).

As BRAF activity acts upstream of ERK, and V600E mutations of BRAF are common in melanoma development (in accordance to our finding that BRAF mutations are present in all melanoma cell lines used in this study [unpublished data]), we tested whether mutated BRAF can modulate CYLD regulation in NHEM. Transfection of NHEM with V600E BRAF resulted in up-regulation of Snail1 and down-regulation of CYLD expression levels (Fig. 2 H).

Collectively, these data indicate that high Snail1 expression modulated by BRAF-mediated ERK activity is involved in the transcriptional down-regulation of CYLD in melanoma.

Snail1 has been shown to play an important role in progression of melanoma. To study whether Snail1 facilitates tumorigenicity of melanoma cells via repression of CYLD, the two melanoma asSnail clones (Fig. 2 D) were stably transduced with viral vectors encoding siRNA against CYLD, resulting in complete suppression of CYLD (Fig. 3 A).

Suppression of Snail1 in Mel Im cells resulted in significantly impaired proliferation (Fig. 3 B, clone 1; Fig. S3 A, clone 2, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1) and migration rates (Fig. 3 C, clone 1; Fig. S3 B, clone 2) compared with control cells. siRNA-mediated depletion of CYLD almost completely rescued the proliferative and migratory potential of Snail1-depleted melanoma cells (Fig. 3, B and C, and Fig. S3, A and B).

To assess the relevance of Snail1-mediated tumor progression in vivo, we used a xenograft tumor model. Mel Im asSnail clones formed significantly smaller tumors than Mel Im control cells when injected subcutaneously into nude mice. In contrast, CYLD siRNA reconstituted the diminished growth of Mel Im antisense Snail1 clones (Fig. 3 D). Histological analysis of tumors formed by Mel Im asSnail clones in nude mice revealed nodular growth (Fig. 3, E, I). In contrast, Mel Im asSnail clones transduced with siRNA against CYLD showed more diffuse infiltration of the surrounding tissue (Fig. 3, E, II).

These findings indicate that CYLD plays an important role in Snail1-mediated progression of melanoma.

We have shown in a previous study that BCL-3 associates with the NF-κB p50 or p52 subunits to enhance cell proliferation by activating the Cyclin D1 promoter in mouse keratinocytes (17). CYLD prevents nuclear accumulation of BCL-3 and hence reduces Cyclin D1 expression and proliferation of keratinocytes.

Next, we examined whether a similar mechanism operates in human melanoma cells. In contrast to NHEM, BCL-3 was primarily found in the nucleus of the melanoma cell lines (Fig. S4 A, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1). Transduction with lentiviral vectors expressing CYLD prevented nuclear translocation of BCL-3 (Figs. S4, B and C), whereas a catalytically inactive mutant of CYLD (CYLD C/S) still binding to BCL-3 was unable to prevent nuclear translocation (Fig. S4, C and D).

To analyze whether CYLD is also regulating Cyclin D1 expression in melanoma cell lines, Mel Im and Mel Juso cells were transduced with viral vectors carrying CYLD. Clones expressing CYLD at similar levels as NHEMs were used as demonstrated by Western blot analysis (Fig. 4 A). In agreement with the mechanism in keratinocytes (17), CYLD markedly reduced expression of Cyclin D1 (Fig. 4 B), cyclin D1 promoter activity (Fig. 4 C), and BCL-3 recruitment to the cyclin D1 promoter in a complex with p50 or p52 (Fig. 4 D, left) as compared with cells transduced with viral vectors carrying a catalytically inactive mutant of CYLD (C/S-CYLD), a GFP expression cassette or noninfected cells.

Because melanoma cells with BRAF V600E mutation exhibit an increased production of the immunosuppressive IL-10 (25), and the production of IL-10 is reduced in BCL-3–null cells (26), we further analyzed BCL-3 binding to the IL-10 promoter. CYLD, but not C/S-CYLD, abrogated BCL-3 binding of IL-10 promoter, confirming the important role of CYLD in the regulation of BCL-3 (Fig. S4 E).

In addition to uncontrolled proliferation, malignant melanoma is characterized by early invasiveness and metastasis. One of the hallmarks of the increased migratory and invasive potential of malignant melanoma cells is the expression of N-cadherin (2–4). Interestingly, the N-cadherin gene contains NF-κB binding sites in its promoter region (27). ChIP analysis revealed DNA-bound BCL-3 in a complex with p50 or p52 in control cells, and CYLD C/S– or GFP-expressing cells, whereas CYLD-expressing cells showed no BCL-3 recruitment to the NF-κB binding sites of the N-cadherin promoter (Fig. 4 D, right).

In line with these findings, expression of CYLD in melanoma cells strongly reduced N-cadherin promoter activity (Fig. 4 E) and protein levels (Fig. 4 F), whereas no effects on Snail1, Snail2, Twist, or E-cadherin expression (Fig. 4 F) were found compared with GFP- or CYLD C/S–transduced controls, respectively. In agreement, down-regulation of CYLD in asSnail clones 1 and 2 resulted in up-regulation of Cyclin D1 and N-cadherin (Fig. 3 A).

Depletion of BCL-3 expression in melanoma cells using siRNA resulted in reduced expression of both N-cadherin and Cyclin D1 without changes in Snail1 expression levels (Fig. 4 G). Furthermore, siRNA against Snail1, which up-regulated CYLD expression (Fig. 4 H), reduced nuclear BCL-3 levels, confirming that reexpression of CYLD blocks nuclear translocation of BCL-3 in melanoma (Fig. 4 H).

These results suggest that BCL-3–induced cyclin D1 and N-cadherin expression is blocked by the expression of CYLD.

To investigate the functional role of the suppressive effect of CYLD on Cyclin D1 and N-cadherin, we used the two different melanoma cell lines (Mel Im and Mel Juso) stably expressing CYLD (Fig. 4 A). CYLD expression caused diminished proliferation in Mel Im cells (Fig. 5 A) and Mel Juso cells (Fig. S5 A, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1), as well as a decrease in colony growth in soft agar in both cell lines (Fig. 5 B and Fig. S5 B) compared with control, GFP, or CYLD C/S–infected cells.

Analysis of the motile and invasive phenotype of CYLD-expressing melanoma cells revealed that CYLD expression resulted in diminished migration and invasion rates as assessed in time-lapse scratch assays (Fig. 5 C and Fig. S5 C), spheroid migration assays (Fig. 5 D and Fig. S5 D), and Boyden chamber assays (Fig. 5 E and Fig. S5 E).

In contrast, forced expression of N-cadherin reversed the CYLD-mediated reduction in melanoma cell migration (Fig. 5 F and Fig. S5 F), and depletion of BCL-3 expression resulted in diminished migration rates (Fig. 5 G and Fig. S5 G).

These data indicate that CYLD regulates proliferation, as well as migration and invasion, of melanoma by controlling N-cadherin expression.

To test the effect of CYLD on tumor growth in vivo, we used a murine xenograft model. Mel Im cells stably expressing CYLD were injected subcutaneously into nude mice, revealing significantly impaired growth compared with cells stably expressing GFP or noninfected cells (Fig. 6 A). Similar results were obtained with Mel Juso cells (Fig. S6 A, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1). 14 d after injection, none of the mice from the CYLD group, but 6 out of 10 animals from the GFP and noninfected control groups, developed tumors. After 4 wk, 4 mice from the CYLD group also developed tumors; however, these tumors were significantly smaller (4.3 ± 2.8 vs. 28.6 ± 14.7 mm³ [GFP] and 28.5 ± 10.7 mm³ [noninfected]; Fig. 6 A).

In addition, melanoma cells were injected i.v. into nude mice to assess in vivo metastasis. MIA serum levels are an established marker to monitor melanoma metastasis (28), and therefore, an ELISA selectively detecting human MIA was applied to monitor the serum content of MIA in this xenograft model. 4 wk after injection, animals receiving CYLD-expressing melanoma cells had significantly lower MIA serum levels (1.1 ± 0.4 ng/ml) than animals injected with GFP-transduced (3.8 ± 0.8 ng/ml) or nontransduced (3.8 ± 1.6 ng/ml) cells (Fig. 6 B and Fig. S6 B).

We performed immunohistochemical analysis sections using MART 1 and MIA, two tumor-associated antigens, to detect lung metastasis. Mice receiving Mel Im cells stably transduced with CYLD showed almost no MART 1 (Fig. 6 C) or MIA (not depicted) staining compared with mice injected with control cells, indicating that they had significantly less metastatic deposits (Fig. 6 D and Fig. S6 C).

In summary, these data show that reduced CYLD expression induces tumorigenicity of melanoma, and that reexpression of CYLD reduces their tumor growth and metastasis in vivo.

To analyze the clinical relevance of reduced CYLD expression in melanoma, we performed immunohistochemical analysis of a tissue microarray consisting of 88 primary human melanomas (Fig. 7 E) and found a significant correlation between loss of CYLD expression and Clark level/tumor invasiveness and tumor thickness (Table I). Moreover, loss of CYLD expression inversely correlated with overall (Fig. 7 A) and progression-free survival (Fig. 7 B). In contrast, patients with primary tumors expressing CYLD developed no tractable metastasis in the time period analyzed.

Notably, and in accordance with the Snail1-CYLD connection described here, expression of Snail1 was inversely correlated to CYLD expression in primary tumors (P < 0.001; Fig. 7, E and F). In compliance, Snail1 expression also correlated with tumor thickness (Table I) and overall (Fig. 7 C) and progression-free survival (Fig. 7 D).

These clinical findings support a link between the known tumor promoter Snail1 and its transcriptional target CYLD, indicating an important role for Snail1 to suppress CYLD during progression of melanoma and for CYLD to suppress melanoma cell proliferation and invasion.

The currently available adjuvant therapies for human melanoma have only limited effects on overall survival and can have severe side effects (for review see reference [29]). This highlights the strong need for rational, mechanism-based therapies. One approach with great potential for achieving this goal is the identification and targeting of tumor-activated signaling cascades.

Our study uncovers a new key signaling pathway that is responsible for progression of malignant melanoma and is related to poor prognosis in melanoma patients. This novel molecular mechanism is based on the suppression of CYLD expression by the transcriptional regulator, Snail1, driving melanoma cells to an aggressive phenotype (depicted as a model in Fig. 8).

Notably, our findings reveal a thus far unknown link between Snail1, which plays an important role in EMT and favors tumor progression (9, 11), and CYLD. CYLD was originally identified as a tumor suppressor that is mutated in familial cylindromatosis (16). In melanoma, we did not find mutations of the CYLD gene (unpublished data), but we identified direct recruitment of the transcriptional repressor Snail1 to the CYLD promoter as a molecular mechanism leading to strongly reduced or even absent CYLD expression in all melanoma tissues and cell lines tested. In accordance, melanoma cells in which Snail1 expression was abrogated showed a strong up-regulation of CYLD and reduced growth in vitro and in vivo.

Interestingly, we found that ERK activation is the major upstream signaling mechanism responsible for high constitutive Snail1 expression in malignant melanoma cells, as well as for induced Snail1 expression in normal melanocytes. In melanoma cells, the Ras–Raf–MEK–ERK (MAPK) signaling pathway is constitutively active and exerts key functions in melanoma development and progression (30). Activating mutations in BRAF or N-ras have been identified in most cutaneous melanomas (31), corresponding to our finding that BRAF mutations were present in all melanoma cell lines tested, but not in primary human melanocytes.

In contrast to melanoma cells, melanocytes express CYLD at a high level, exhibit low ERK activity, and lack Snail1 expression. Introducing mutated BRAF into primary melanocytes resulted in increased Snail1 expression and reduced CYLD expression. Our findings indicate that after malignant transformation, constitutively high ERK activity and Snail1 expression allow melanoma cells to permanently exploit transcriptional repression of CYLD and activation of BCL-3 to acquire a more aggressive phenotype.

BRAF V600E mutation has been shown to induce NF-κB activity in human melanoma cell lines via regulation of β-Trcp expression (32), and high NF-κB activity resulting in increased tumorigenesis has been reported in malignant melanoma (32–34). Here, we provide a new mechanism for how mutational activation of B-RAF contributes to both constitutive NF-κB and BCL-3 activity via repression of CYLD.

Loss of CYLD promoted BCL-3 recruitment to the Cyclin D1 promoter. This mechanism plays an important role for the growth of melanoma cells because stable expression of CYLD in melanoma cells reduced tumor growth and metastasis in vivo in a xenograft model. More importantly, we showed for the first time that BCL-3 is recruited to the N-cadherin promoter to activate transcription. Thereby, our study provides a novel paradigm of how CYLD affects N-cadherin expression, which has been shown to correlate with tumor cell motility and invasiveness in several malignancies including melanoma (2, 12, 33). Re-expression of CYLD in melanoma cells with constitutively high Snail1 levels inhibited N-cadherin expression, leading to a reduced migratory and invasive potential in vitro and less pulmonary metastasis in a murine in vivo model. Conversely, depletion of CYLD in Snail1-deficient melanoma cells resulted in highly aggressive melanoma cells with strong proliferative and invasive potential. In summary, these findings clearly demonstrate that suppression of CYLD is a central mechanism in how Snail1 facilitates both growth and metastasis in malignant melanoma. Whether this mechanism operates in other types of cancer has to be tested.

Importantly, the clinical findings further support an important role of the Snail1 CYLD pathway in tumor progression and survival in melanoma patients. Snail1 expression shows significant correlation with loss of CYLD and signs of tumor progression like tumor thickness. Even more important, both Snail1 and CYLD expression in primary tumors directly correlate with progression-free survival and overall survival of the patients.

Hence, elevated expression of Snail1 and reduced expression of CYLD in primary melanoma hint to a high metastatic potential in vivo. Although CYLD levels in primary tumor and melanoma might vary, our findings indicate the intriguing possibility that analysis of CYLD and Snail1 protein expression in the primary tumor may be used as a predictive marker for survival of melanoma patients. Such prognostic risk assessment would have major clinical implications for the appropriate selection of treatment modalities and follow up protocols, or as stratification factor for adjuvant therapy.

Finally, our data also suggest a rationale for strategies to rescue CYLD expression in malignant melanoma as therapeutic approaches for this highly aggressive tumor.

The melanoma cell lines Mel Ei, Mel Wei, Mel Ho, and Mel Juso were derived from a primary cutaneous melanoma. Mel Im and Mel Ju were derived from metastases of malignant melanomas (J. Johnson, Ludwig Maximilian University of Munich, Munich, Germany) (35). Furthermore, melanoma cells were freshly isolated from primary human melanomas and used at low passage to confirm key findings obtained with melanoma cell lines. All primary melanoma cells and cell lines used carried the V600E mutation in BRAF. Isolation and culture of primary NHEMs (35) and 5-azacytosine (10 mM) treatment for analysis of promoter methylation were performed as previously described (36).

Melanoma cells and NHEMs were treated with the chemical MAPK inhibitors SP600125 (specific for JNK1, JNK2, and JNK3), PD98059, and UO126 (specific for MEK1 and MEK2; all from Calbiochem) at a concentration of 20 µM for 19 h. The vehicle DMSO alone served as control.

Tumor cell inoculation and measurement of tumor growth in NMRI (nu/nu) mice was performed as previously described (37). For analysis of in vivo metastasis melanoma, cells were injected i.v. (10⁶ cells/mouse). 4 wk after inoculation, lungs were processed for (immuno-)histology, and intrapulmonary metastases were counted for one representative section per mouse using MART 1 (1:100; Novus Biologicals) and MIA (38) staining to determine metastasis. In addition, MIA serum levels were determined using a MIA-ELISA (Roche). Experiments were approved by the Department of Veterinary Affairs of the local Government of Bavaria, District of Oberpfalz Area.

Tissue samples from primary human melanoma, and melanoma metastasis obtained from patients undergoing surgical treatment, were immediately snap frozen and stored at −80°C. Melanoma cells were selectively retrieved from tumor samples with PALM microlaser technology (PALM) under microscopic control.

The tissue microarray was constructed as previously described (39) and contained primary unifocal malignant melanomas. Clinical follow up data from the Central Tumor Registry were available for all patients. The median follow up for all patients was 54 mo (range 1–135 mo). The University of Regensburg Institutional Review Board granted approval for the project.

Isolation of total cellular RNA from cultured cells and tissues and reverse transcription were performed as previously described (36). To ensure expression of the full-length CYLD mRNA, RT-PCR analysis was performed using two specific primer combinations (Table S1, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1) and the PCR products were resolved on 1.5% agarose gels. Sequences were examined by automated sequencing. Quantitative real time-PCR was performed with specific primers for CYLD, Cyclin D1, N-cadherin, and Snail1 (Table S2) applying LightCycler technology (Roche), as previously described (36).

Protein extraction, analysis, and Western blotting were performed as previously described (36), applying the following primary antibodies: polyclonal anti-CYLD (1:1,000 [17]), anti–Cyclin D1 (1:500; Santa Cruz Biotechnology, Inc.), anti–BCL-3 (1:250; Santa Cruz Biotechnology, Inc.), anti–N-cadherin (1:2,000; BD), anti-ERK1/2 (1:1,000), anti–Laminin A/C (1:1,000) or anti-phospho-ERK1/2 (1:1,000; Cell Signaling Technology), anti-Snail1 (1:500; Abcam), anti-Snail2 (1:500; Abcam), anti-Twist (1:500; Abcam), anti-tubulin (1:1,000; Millipore), or anti–β-actin (1:5,000; Sigma-Aldrich).

Immunohistochemical and immunofluorescence analysis of paraffin-embedded tissues applying anti-CYLD antibody (1:1,500), anti-Snail1 antibody (1:50), or anti-tyrosinase antibody was performed as previously described (17, 36). For negative control, the primary antibody was omitted and IgG isotype control antibodies did not reveal any detectable staining. Immunohistochemical stainings were counterstained with hematoxylin. For analysis of the tissue, microarray positivity for Snail1 was therefore defined as any detectable nuclear staining, and CYLD staining intensity was specified applying a semiquantitative three-step scoring system (score 0, negative; score 1+, weak; and score 2+, strong cytoplasmic staining). Examples for positive and negative CYLD and Snail1 stainings are given in Fig. 7 E. Intracellular BCL-3 localization was visualized by immunofluorescence analysis using an anti–BCL-3-antibody (1:200; Santa Cruz Biotechnology, Inc.).

Cell proliferation was determined by means of BrdU incorporation into cellular DNA using an ELISA-based colorimetric assay (Roche). Colony formation in soft agar, monolayer scratch assays, spheroid migration assays, and migration and invasion assays in Boyden chambers were performed as previously described (40, 41).

ChIP was performed according to the manufacturer's (IMGENEX Corporation) instructions using polyclonal antibodies against Snail1 (clone E-18; Santa Cruz Biotechnology, Inc.), p50, p52, Snail2, Twist (Abcam), or BCL-3 (Santa Cruz Biotechnology, Inc.), and negative control rabbit immunoglobulin (DAKO) as control. The extracted DNA was used for semiquantitative PCR to amplify the Cyclin D1, N-cadherin, IL-10, or CYLD promoter region (Table S3, available at http://www.jem.org/cgi/content/full/jem.20082044/DC1).

Expression plasmids of CYLD, N-cadherin, or BCL-3 were used for transient transfection applying the Lipofectamine plus method (Invitrogen) (35, 36). BRAF plasmids (C. Wellbrook, University of Manchester, Manchester, England, UK) and Snail expression constructs (A. Garcia de Herreros, Institute Municipal, Barcelona, Spain) were also used. Stable expression of CYLD, C/S-CYLD, CYLD siRNA, and GFP was achieved by retro- or lentiviral gene transfer (42). In brief, CYLD and GFP were cloned into a retroviral vector (pclMFG). CMV promoter-driven lentiviral constructs were obtained by cloning wt-HA-CYLD, C/S-HA-CYLD, and GFP into a HIV-1–derived lentiviral vector (42).

Transfection with CYLD siRNA (43) or BCL-3 siRNA (17) was performed as previously described using HiPerfect according to the manufacturer's recommendations (Invitrogen), and the transfection efficiency was ∼40%. siRNA oligos were purchased from TAG Copenhagen A/S with the following RNA sequences: Scramble (5′-AGACCCACUCGGAUGUGAAGAGAUA-3′) and Snail1 697 (5′-GUGAAGAGAUACCAGUGUCAGGCCU-3′).

The reporter gene assays with CYLD, Cyclin D1, or N-cadherin promoter-luciferase plasmids were performed as previously described (17, 27, 36).

Values are presented as means ± SEM. Statistical differences were determined using Student's t test. Contingency table analysis and two-sided Fisher's exact tests were used to study the statistical association between categorical clinicopathologic and immunohistochemical variables. Retrospective overall survival and progression-free survival curves comparing patients with and without any of the variables were calculated using the Kaplan-Meier method, with significance evaluated using log-rank statistics. For the progression-free survival analysis, patients were censored at the time of their last tumor-free clinical follow-up appointment. For the overall survival analysis, patients were censored at the time of their last clinical follow-up appointment. Median overall survival time for censored patients was 63.5 mo (range, 3–135 mo). P < 0.05 was considered significant. Statistical analyses were performed using SPSS version 10.0 (SPSS, Inc.).

Fig. S1 shows the CYLD promoter sequence. Fig. S2 shows mRNA and protein expression of Snail and Slug in malignant melanoma cell lines. Fig. S3 includes data on the effect of CYLD analyzed in Mel Im as Snail clone 2. Fig. S4 shows the effect of CYLD on subcellular localization of Bcl-3. In Fig. S5, the effects of CYLD on proliferation and migration of Mel Juso melanoma cells in vitro are presented, and in Fig. S6 the effects of CYLD on in vivo proliferation and migration of Mel Juso melanoma are presented. Table S1 lists the primer used for amplification of the CYLD gene, Table S2 lists primer used for mRNA amplification by RT-PCR, and in Table S3 primer used in ChIP are listed.

We thank Susanne Wallner and Elisabeth Bumes for excellent technical assistance and Drs. J. Johnson (Ludwig Maximilians University of Munich, Germany) for melanoma cell lines, David Goltzman (McGill University, Montreal) for the N-cadherin promoter luciferase vector, Claudia Wellbrock (University of Manchester) for the BRAF constructs, and Antonio Garcia de Herreros (Institute Municipal, Barcelona) for the antisense Snail1 construct.

The work was supported by the German Research Association (to A.K. Bosserhoff, A. Pfeifer, and R. Fassler), the Dr. Mildred Scheel Foundation for Cancer Research, Melanoma Network (to A.K. Bosserhoff), the Swedish Society for Medical Research, Swedish Cancer Foundation, Swedish Medical Research Council, Royal Physiographic Society in Lund, U-MAS Research Foundations (to R. Massoumi), and the Max Planck Society (to R. Fassler).

The authors have no conflicting financial interests.