We first used Western analysis to evaluate SKI and SnoN protein levels in a panel of human melanoma cell lines as compared to normal melanocytes. As shown in Figure 1A, SKI and SnoN protein levels were barely detectable in normal melanocytes. On the other hand, all melanoma cell lines tested (WM793, 1205Lu, WM852, WM983B and SK28) expressed high levels of SKI and SnoN protein (Figure 1A). The non-tumorigenic MNT1 cell line expressed relatively similar levels of SKI protein, after correction for β-actin content, as compared to other melanoma cell lines with tumorigenic potential. Additional cell lines (Dauv-1, Fo-1, WM239A, WM1341D, SK-mel501, SK-mel888) exhibited similar high SKI protein content (not shown). These data are consistent with previous report on the subject 17 . P-SMAD3, a marker of constitutive TGF-β receptor activity, was detected in all melanoma cell lines that we examined, not in normal melanocytes, consistent with our initial observations of autocrine SMAD signaling in various human melanoma cell lines in culture 35 . SKI mRNA levels, as measured using quantitative RT-PCR (Figure 1B) were highly variable across melanoma cell lines, not higher than in normal melanocytes, and did not correlate with SKI protein levels, suggesting uncoupling of gene transcription and protein expression. Similar results were found for SnoN mRNA levels (not shown). Together, these data are consistent with the literature that describes SKI and SnoN proteins as targets for proteasomal degradation in response to TGF-β 22 23 24 25 26 27 28 29 .

We next examined the expression of the ubiquitin ligases Arkadia and Smurf2, as these proteins are essential for proteasome-mediated degradation of SKI and SnoN proteins. As shown in Figure 1C, all melanoma cell lines exhibited elevated and rather similar levels of Arkadia and variable levels of Smurf2. Arkadia was hardly detectable in normal melanocytes, in which no expression of Smurf2 was found. Remarkably, treatment of normal melanocytes with the proteasome inhibitor MG132 allowed for a dramatic recovery of SKI protein levels (Figure 1D). MG132 treatment of 1205Lu melanoma cells treated resulted in increased SKI protein content, consistent with a role of the proteasome in controlling SKI protein levels, both in normal and malignant melanocytes.

Given our extensive phenotypic characterization of various melanoma cell lines using Matrigel™ invasion in vitro as well as subcutaneous tumor growth and bone metastasis in nude mice 36 42 , we thought to determine whether basal SKI protein levels in culture may be predictive of a given invasive, tumorigenic, or metastatic behavior of melanoma cells. As shown in Table 1, SKI protein levels did not correlate with the capacity of melanoma cells to invade Matrigel™. Neither did they correlate with their capacity to form subcutaneous tumors in nude mice or with the incidence of bone metastasis following intracardiac inoculation of tumor cells into nude mice. Remarkably, all of these cellular activities are efficiently altered upon TGF-β inhibition by either SMAD7 overexpression or pharmacologic inhibitors of TβRI kinase activity in vitro or in vivo 36 37 38 , attesting for pro-tumorigenic and pro-metastatic activities of autocrine TGF-β signaling despite high SKI and SnoN protein levels.

We next investigated whether high SKI levels in melanoma cells are associated with an absence of transcriptional responses to TGF-β. Incubation of 1205Lu melanoma cells with increasing concentrations of TGF-β for 30 min lead to a dose-dependent decrease in SKI protein content (Figure 2A), accompanied with an inversely correlated increase in P-SMAD3 levels. Parallel transient cell transfection experiments with SMAD3/4-specific (CAGA)₉-MLP-luc reporter construct indicated dose-dependent transcriptional activation in response to TGF-β (Figure 2B).

To determine the kinetics of SKI degradation in response to TGF-β, three distinct human melanoma cell lines that exhibit high SKI protein levels in basal cell culture conditions were incubated with TGF-β; SKI protein content was monitored over time by Western blotting. Results shown in Figure 2C indicate a rapid, time-dependent, degradation of the SKI protein in all cell lines, which was abolished when cells were incubated with the TGF-β receptor type I (TβRI/ALK5) kinase inhibitor SB431542 1 h prior to TGF-β addition (Figure 2D).

In view of these experiments, it appears that despite high expression of the SKI protein, melanoma cells exhibit a strong transcriptional response to exogenous TGF-β. Rapid degradation of SKI occurs within minutes following TGF-β challenge and is accompanied with strong SMAD-dependent transcriptional activity.

Inhibition of autocrine TGF-β signaling by stable overexpression of SMAD7 in the 1205Lu cell line did not significantly alter SKI protein content, yet dramatically inhibited Matrigel™ invasion, and almost entirely blocked subcutaneous tumor growth and the appearance of experimental bone metastases in mice (Table 2), Together, these results suggest uncoupling of the pro-invasive and pro-metastatic activities of TGF-β with SKI protein levels in melanoma cells, or at least indicate that SKI function is relatively marginal as compared to the tumor promoter activities of TGF-β

As expected from the literature, the proteasome inhibitor MG132 efficiently abolished TGF-β-dependent SKI degradation (Figure 3A). Also, a 1 h pre-treatment of 1205Lu and Dauv-1 melanoma cells with the proteasome inhibitors MG132 and ALLN strongly inhibited SMAD3/4-specific transcriptional response induced by TGF-β in transient cell transfection experiments with (CAGA)₉-MLP-luc (Figures 3B and 3C, respectively). Likewise, a 1-h pre-treatment with MG132 attenuated TGF-β-induced IL-11 and PTHrP expression in 1205Lu cells (Figures 3D and 3E, respectively), two known SMAD genes targets implicated in melanoma and breast cancer metastasis to bone 36 43 44 45 . Thus, although SKI has little influence on TGF-β response because of its rapid degradation, it is likely that prevention of SKI degradation, as achieved by MG132 or ALLN pre-treatment of the cells, contributes to the attenuation of TGF-β-dependent transcriptional responses. This experimental approach does not however exclude that other proteasome-mediated events, independent from SKI, may also be implicated in the attenuation of TGF-β responses.

To better understand the contribution of endogenous SKI levels to melanoma cell behavior, SKI expression was knocked down by stable expression in 1205Lu melanoma cells of a specific shRNA. Despite a 90% reduction in SKI protein content (Figure 4A), there was no significant alteration of SMAD3/4-specific transcriptional responses to TGF-β, as estimated in transient cell transfection experiments with (CAGA)₉-MLP-luc (Figure 4B). Likewise, induction of IL-11 and PTHrP expression in response to TGF-β was not significantly altered in SKI-knockdown cells as compared to mock-transfected cells (Figure 4C). These data were further validated by means of SKI-specific siRNA transfection experiments in 1205Lu, WM852 and 888mel cells (not shown). Also, SKI knockdown did not alter the capacity of 1205Lu and WM852 (not shown) melanoma cells to invade Matrigel™ (Figure 4D).

These observations are consistent with the notion that the high levels of SKI are effectively degraded by TGF-β in these melanoma cells and therefore do not play a critical role in antagonizing, or preventing, TGF-β responses. Accordingly, we previously provided direct evidence that the invasive capacity of melanoma cells is highly dependent upon autocrine TGF-β signaling 36 , further suggesting that SKI levels do not strongly influence or attenuate TGF-β effects.

It has been suggested that high SKI expression in melanoma cells is responsible for the lack of growth inhibitory activity of TGF-β, by blocking TGF-β-driven p21 expression 17 39 . Given the ample evidence for efficient TGF-β signaling and associated transcriptional responses in all melanoma cell lines tested thus far in our laboratory (see above and refs 35 36 37 ), we tried to reproduce these data in the 1205Lu melanoma cell line, which is both highly invasive, strongly resistant to TGF-β growth inhibitory activity, capable of a strong SMAD3/4-specific transcriptional response to exogenous TGF-β stimulation, yet expresses high levels of SKI and SnoN proteins (see Figure 1A). Firstly, parallel transient cell transfections with either a 2.4-kb p21/WAF1 promoter luciferase construct or (CAGA)₉-MLP-luc were performed in 1205Lu cells. TGF-β had no effect on p21 promoter activity despite efficient SMAD3/4-specific gene transcription, as measured using the highly sensitive (CAGA)₉-MLP-luc construct (Figure 5A). As expected, p21 promoter transactivation in response to TGF-β was readily observed in HaCaT keratinocytes. These data confirm our initial observations that melanoma cells efficiently respond to TGF-β by a strong SMAD-specific transcriptional response 35 36 37 , and that the lack of induction of p21 is highly gene-specific and is probably not due to a general inhibition of TGF-β signaling by SKI or SnoN, as SMAD3/4-specific transcription and induction of other TGF-β target genes, such as IL-11 or PTHrP, is intact. Remarkably, both the proliferative rate and the weak growth inhibition exerted by TGF-β (approx. 10% after 72 h) were virtually identical in both mock- and shSKI-transduced 1205Lu cells (Figure 5B). Also, SKI knockdown did not restore p21 promoter transactivation in response to TGF-β (Figure 5C). Likewise, oligonucleotide siRNA-mediated SKI knockdown in transient cell transfection experiments using 1025Lu, WM852 and 888mel cells did not allow p21 expression or promoter transactivation in response to TGF-β in any of those cell lines (not shown). These results are fully consistent with our previous work and with the observations provided herein that indicate that high SKI levels in melanoma cells do not antagonize the pro-tumorigenic activities exerted by TGF-β. Neither do they interfere with TGF-β-driven gene responses. It should be noted that lack of p21 induction by TGF-β in 1205Lu cells is specific, as we previously demonstrated that JNK inhibition efficiently activates p21 expression and promoter transactivation in this cell line 46 .

Relatively few studies have examined the expression of SKI in melanocytic lesions in humans. We thus used immunohistochemistry to detect SKI protein in a panel of 12 nevi, 37 primary melanomas at various clinical and pathological stages of disease progression, 17 cutaneous and 10 lymph node metastases (Table 3). SKI was detected in 8 (66%) nevi, 8 (21.6%) primary melanomas, and 8 (21.7%) metastases (Table 4). Representative results for SKI staining are shown in Figure 6. We found no evidence for a link between SKI expression and histological or pathological staging within each melanoma group of samples. These data are remarkably similar to those recently reported in a larger cohort of 120 patients treated for cutaneous melanoma 47 .

We further analyzed the activation of TGF-β signaling in tissues by means of P-SMAD3C immunohistochemistry in a subset of melanomas and metastases. Nuclear expression of P-SMAD3C was observed in all melanocytic lesions, albeit at varying intensity (Figure 7). Intriguingly, staining intensity of SKI and phospho-SMAD3C on consecutive sections appeared to be inversely correlated (Figure 7). Although these immunohistochemical analyses do not allow quantification of protein expression, they support our observation that high TGF-β signaling can drive SKI degradation.

Taken together, the results presented herein unambiguously demonstrate (a), that SKI levels in melanoma cells are not predictive of their tumorigenic, invasive or metastatic propensity, (b), that TGF-β signals lead to rapid degradation of SKI proteins in a proteasome-dependent manner, and (c), that TGF-β induces a efficient SMAD3/4-dependent transcriptional response in melanoma cells despite high expression of c-SKI and SnoN in these cells. Furthermore, our results support the notion that there is no correlation between SKI expression and tumor progression or histogenetic subtype of human cutaneous melanomas.

Melanoma cell lines (1205Lu, WM852, WM983B, SK28, MNT1, Dauv-1, Fo-1, WM239A, WM1341D, SK-mel501, SK-mel888) have been described previously 35 42 69 70 71 . NHEM (Normal Human Epidermal Melanocytes) were purchased from Promocell (Heidelberg, Germany) and cultured in ready-to-use medium, also supplied by Promocell. All cells were grown at 37°C in a humidified atmosphere of 5% CO2. The reporter plasmids (CAGA)₉-MLP-luc 72 and 2.4 kb p21/WAF1 promoter luciferase reporter construct 73 were gifts from Drs. Sylviane Dennler (INSERM U697, Paris, France) and Bert Vogelstein (Johns Hopkins University, Baltimore, MD), respectively. The pRL-TK vector was from Promega (Madison, WI). pSuper vector expressing SKI shRNA has been described previously 22 . Human recombinant TGF-β1, purchased from R&D System Inc. (Minneapolis, MN) is referred to as TGF-β throughout the text. The ALK5/TβRI inhibitor SB431542, and the proteasome inhibitor leu-leu-leu-al (MG132) were both from Sigma-Aldrich (St. Louis, MO). Acetyl-leu-leu-norleu-al (ALLN) was purchased from CalBiochem (San Diego, CA).

Protein extraction and Western blotting were performed as previously described 74 . Anti-SMAD3 and anti-Actin antibodies were purchased from Zymed (San Francisco, CA) and Sigma-Aldrich, respectively. The rabbit anti-phospho-SMAD3 antibody 75 was a generous gift from Dr. Edward Leof (Mayo Clinic College of Medicine, Rochester, MN). Anti-c-SKI, anti-SnoN, anti-SMURF2, anti-HSP60 and secondary anti-mouse and anti-rabbit horseradish peroxidase-conjugated antibodies were from Santa Cruz Biotechnology Inc. (Santa Cruz, CA). Anti-Arkadia was obtained from Abnova Gmbh (Heidelberg, Germany).

Melanoma cells were seeded in 24-well plates and transfected at approximately 80% confluency with the polycationic compound Fugene (Roche Diagnostics, Indianapolis, IN) in fresh medium containing 1% FCS. TGF-β and/or inhibitors were added 4 h after transfections. Following a 16-h incubation, cells were rinsed twice with PBS and lysed in passive lysis buffer (Promega). pRL-TK (Promega) was co-transfected to assess transfection efficiency. Luciferase activities were determined with a Dual-Glo luciferase assay kit according to the manufacturer's protocol (Promega). Each experiment was repeated at least three times using triplicate dishes in each of them for each experimental condition.

Total RNA was isolated using an RNeasy™ kit (Qiagen GmbH, Hilden Germany). Genomic DNA contaminations were eliminated by DNAse I treatment. One microgram of RNA from each sample was reverse transcribed using the Thermoscript kit (InVitrogen) following the manufacturer's protocol. The resulting cDNAs were then processed for either semi-quantitative or real-time PCR using SYBR Green technology. In the latter case, reactions were carried out in a 7300 Real-time PCR System (Applied Biosystem) for 40 cycles (95°C for 15 sec/60°C for 1 min) after an initial 10-min incubation at 95°C, using the following primer sets: SKI (sense, 5'-gagaaattcgactatggcaacaag-3'; antisense, 5'-gtcatctgttttgggtcttatgga-3'; IL-11 (sense, 5'-actgctgctgctgaagactc-3'; antisense, 5'-ccacccctgctcctgaaata-3'); PTHrP (sense, 5'-tttacggcgacgattcttcc-3'; antisense, 5'-ttcttcccaggtgtcttgag-3'); GAPDH (sense, 5'-gctcctcctgttcgacagtca-3'; antisense, 5'-accttccccatggtgtctga-3'). Data were analyzed using Applied Biosystems Sequence Detection Software (version 1.2.1).

Tissue culture Transwell^® inserts (8-μm pore size, Falcon, Franklin Lakes, NJ) were coated for 3 h with 10 μg of growth factor-reduced Matrigel™ (Biocoat, BD Biosciences, San Jose, CA) in 100 μl of PBS at 37°C. After air-drying the chambers for 16 h, the Matrigel™ barrier was reconstituted with 100 μl DMEM for 24 h at 37°C. The chambers were then placed into 24-well dishes containing 750 μl of W489 medium supplemented with 0,1% FCS. Cells (5 × 10⁴) were added to the upper well of each chamber in 500 μl of serum-free W489 medium. After a 24 h-incubation period, the number of invading cells was counted by bright-field microscopy at ×200 in six random fields. Additional details of the procedure may be found in 37 .

Melanoma cells were plated in 12-well plates at an initial density of 5000 cells/well. Cell growth was measured after 72 h in 1% FCS, with or without TGF-β (10 ng/ml), by counting the cells after trypsinization using a Malassez cell. For each experimental condition, duplicate dishes were counted. Experiments were performed at least three times with similar results.

Formaldehyde-fixed and paraffin-embedded naevi (n = 12), primary cutaneous melanomas (n = 37), cutaneous and lymph node metastases (n = 17 and 10, respectively) from adult patients were obtained from the pathology archives of the Radboud University Nijmegen Medical Centre, and re-evaluated by an expert pathologist (see corresponding Results section for details). Tissues were obtained according to local ethical guidelines and approved by the local regulatory committee.

Paraffin-embedded 4 μm sections on superfrost slides (Menzel-Glaser, Braunscheig, Germany) were de-waxed in xylene, rehydrated through graded alcohol baths, then rinsed with PBS. After quenching of endogenous peroxidases, an antigen retrieval step (sodium citrate 10 mM, pH 6.0 for 10 min. at 95°C) was performed. Tissue sections were subsequently pre-incubated with 20% normal goat serum in PBS, followed by an overnight incubation with rabbit polyclonal SKI antibody (H-329, 1 μg/ml, Santa Cruz Biotechnology Inc.) or affinity-purified rabbit polyclonal anti-phospho-Smad3C (P-SMAD3C) antibody 75 in PBS containing 1% bovine serum albumin overnight at 4°C. For detection of SKI, the Powervision system (Immunovision Labs, Brisbane, CA) was used as a secondary reagent with 3-amino-9-thylcarbazole served as a chromogen. For detection of P-SMAD3C, a biotin-avidin peroxidase complex was generated according to standard procedures (Vector, Burlingame, CA) and developed with 3,3'-diaminobenzidine. Counterstaining was performed with hematoxylin. Samples with nuclear SKI appear purple and were scored positive in case positivity was detected in at least 10% of melanocytic cells.

Data were entered in a computerized database and analyzed using SPSS software (version 16.0.1 for Windows). The binomial test was used to analyze frequency of SKI expression in nevi. The Mann-Whitney U test was used to correlate SKI expression and tumor thickness. The correlation between SKI expression and the level of invasion was determined by the Pearson Chi-square test. Fisher exact test for small sample numbers was used to determine the correlation between SKI expression in cutaneous and nodal metastases.