Expression of Microphthalmia-associated Transcription Factor (MITF), Which Is Critical for Melanoma Progression, Is Inhibited by Both Transcription Factor GLI2 and Transforming Growth Factor-β*

Background: Microphthalmia-associated transcription factor (MITF) levels are critical for melanoma progression. GLI2, a TGF-β target, antagonizes MITF to promote melanoma cell invasion. Results: TGF-β, via inhibition of the cAMP pathway, and GLI2 inhibit MITF transcription via separate cis-elements. Conclusion: TGF-β and GLI2 down-regulation of MITF transcription involves distinct mechanisms. Significance: These data provide novel insights into MITF regulation as it relates to melanoma progression. The melanocyte-specific transcription factor M-MITF is involved in numerous aspects of melanoblast lineage biology including pigmentation, survival, and migration. It plays complex roles at all stages of melanoma progression and metastasis. We established previously that GLI2, a Kruppel-like transcription factor that acts downstream of Hedgehog signaling, is a direct transcriptional target of the TGF-β/SMAD pathway and contributes to melanoma progression, exerting antagonistic activities against M-MITF to control melanoma cell invasiveness. Herein, we dissected the molecular mechanisms underlying both TGF-β and GLI2-driven M-MITF gene repression. Using transient cell transfection experiments with M-MITF promoter constructs, chromatin immunoprecipitation, site-directed mutagenesis, and electrophoretic mobility shift assays, we identified a GLI2 binding site within the −334/-296 region of the M-MITF promoter, critical for GLI2-driven transcriptional repression. This region is, however, not needed for inhibition of M-MITF promoter activity by TGF-β. We determined that TGF-β rapidly repressed protein kinase A activity, thus reducing both phospho-cAMP-response element-binding protein (CREB) levels and CREB-dependent transcription of the M-MITF promoter. Increased GLI2 binding to its cognate cis-element, associated with reduced CREB-dependent transcription, allowed maximal inhibition of the M-MITF promoter via two distinct mechanisms.

TGF-␤ signals via serine-threonine kinase receptors. Liganddependent receptor activation results in phosphorylation and nuclear accumulation of proteins of the SMAD family of transcription factors, leading to target gene regulation (1,2). Often considered a tumor suppressor pathway as it negatively controls the cell cycle in normal and premalignant cells, TGF-␤/ SMAD signaling also promotes cancer cell invasion and metastasis through autocrine and paracrine mechanisms. This occurs most notably via its capacity to induce an epithelial-mesenchymal transition, a phenomenon whereby epithelial tumor cells acquire a migratory and invasive, mesenchymal-like, phenotype (3). Also, TGF-␤ acts in a paracrine manner to favor peritumoral angiogenesis and suppress anti-tumoral immunity (4). In melanoma, constitutive TGF-␤ signaling through the SMAD pathway promotes invasiveness and metastatic potential of melanoma cells (5).
GLI2, a member of the Kruppel family of transcription factors, was recently identified as a direct TGF-␤/SMAD target ubiquitously induced in normal and transformed cells (6,7). Induction occurs at the transcriptional level via mechanisms that include the recruitment of both SMAD3 and ␤-catenin to its promoter (6). We found that GLI2 controls melanoma cell invasiveness and metastatic potential via acquisition of a mesenchymal phenotype characterized by loss of CDH1 expression and enhanced MMP2 and MMP9 secretion (8). Similar changes in gene expression were found in human melanoma tumors, suggesting that GLI2 may also represent a marker of poor prognosis for patients with melanoma (8). Recently, we determined that GLI2 expression in melanoma cells is inversely correlated with that of the melanocyte-specific isoform of the microphthalmia transcription factor (M-MITF) 5 and its associated tran-scriptional program leading to pigmentation (9). In view of these findings, GLI2 appears as a critical transcription factor antagonizing M-MITF function to promote melanoma cell phenotypic plasticity and invasive behavior. A number of recent studies point to a broader and critical role for GLI2 in tumor progression and metastasis in various cancer types (for review, see Ref. 10).
M-MITF, a basic helix-loop-helix leucine zipper transcription factor, is central to a variety of cellular processes in the melanocyte lineage (migration, morphology, proliferation, survival), including terminal differentiation of melanocytes leading to pigment production as well as malignant transformation and melanoma progression to metastasis (11,12). In patients with melanoma, M-MITF may either be amplified or lost, the outcome of M-MITF function depending both on its expression levels and on the presence or absence of concomitantly occurring genetic alterations such as activating B-RAF and N-RAS mutations (11,12). It is thought that by switching back and forth between low and high M-MITF phenotypes, melanoma is driven to progress (13,14).
We recently identified TGF-␤ as a potent repressor of M-MITF expression and as an inducer of GLI2 expression in melanoma cells and found that GLI2 overexpression inhibits M-MITF promoter activity (9). In this report we have elucidated the precise molecular mechanisms underlying M-MITF transcriptional control by TGF-␤ and GLI2. We demonstrate that inhibition of M-MITF expression by TGF-␤ occurs via repression of protein kinase activity leading to reduced CREBdependent transcription, whereas GLI2-driven transcriptional repression of M-MITF is an independent phenomenon. Distinct cis-elements within the M-MITF promoter, namely a novel Kruppel-like transcription factor (KLF)/GLI2 binding region and a proximal CREB binding site, drive M-MITF silencing by TGF-␤ and GLI2, respectively.
Biochemical Methods-Protein extraction and Western blotting were performed as previously described (8,9). Mouse anti-M-MITF and rabbit anti-GLI2 antibodies were purchased from Neomarkers (Thermo Scientific, Kalamazoo, MI) and Cell Signaling (Ozyme, St-Quentin en Yvelines, France), respectively. Mouse monoclonal anti-actin and goat anti-ubiquitous kinesin heavy chain as well as secondary donkey anti-rabbit and antimouse HRP-conjugated antibodies were from Santa-Cruz Biotechnology (Santa-Cruz, CA). Protein kinase A (PKA) activity was measured using a commercial kit from Promega (nonradioactive PKA Assay) utilizing a fluorescent PKA-specific peptide substrate. In brief, 5 ϫ 10 6 cells were incubated in medium supplemented with 1% FCS in the absence or presence of TGF-␤ (5 ng/ml). Cells were harvested, and samples were processed according to the manufacturer's protocol. Phosphorylated and unphosphorylated peptides were separated by agarose gel electrophoresis and visualized under UV light. Quantitation of fluorescence was measured at 540 nm after gel dissolution into 16 M acetic acid (Sigma).
Cell Transfections-For reporter assays, melanoma cells were seeded in 24-well plates and transfected at ϳ70 -80% confluency in fresh medium containing 1% FCS with either the polycationic compound FuGENE TM (Roche Diagnostics) or JetPEI (Polyplus Transfection Inc., New York, NY). After incubation, luciferase activities were determined with a Dual-Glo TM luciferase assay kit (Promega) using a Fluoroskan Ascent FL (Thermo Labsystems). For stable expression of GLI2⌬N, melanoma cells were transfected with JetPEI at ϳ70 -80% confluency with 10 g of either empty pcDNA3 expression vector or the same vector carrying constitutively active GLI2⌬N (18) per 100-mm diameter culture dish. Three days later, G418 (Sigma, 0.7 mg/ml) was added to the culture medium. Selection of stably transfected cell populations occurred within a 3-week period. RT-PCR and measurement of GLI-dependent transcription were used to verify expression and activity of transfected GLI2⌬N, respectively. All experiments were performed at least three times independently using triplicate dishes.
Gene Silencing in Human Melanoma Cells-Infection of melanoma cells with lentiviral particles expressing either control, non-targeting shRNA (shCtrl, Sigma SHC002V), or shRNA targeting GLI2 (Sigma SHVRS clone ID TRCN0000033329 and TRCN0000033330) has been described previously (8,9). Transduced cell populations were selected with puromycin (2 g/ml). Efficacy of GLI2 knockdown over time was verified by real time RT-PCR and Western blotting after each cell passage.
RNA Extraction and Gene Expression Analysis-RNA extraction procedure and reverse transcription-polymerase chain reaction (RT-PCR) methodologies have been described previously together with primer sequences (8). The latter are available upon request.
Electrophoretic Mobility Shift Assays (EMSAs)-Oligonucleotides spanning the regions Ϫ334/Ϫ304 and Ϫ334/Ϫ296 were radiolabeled with [ 32 P]dCTP using Klenow polymerase and used as probes to detect DNA-protein interactions within the GLI2-responsive region of the M-MITF promoter. Nuclear extracts from 1205Lu melanoma cells were isolated using a small scale preparation (22). For competition experiments, a 100-fold excess of unlabeled consensus GLI2 (23) or Sp1 (24) oligonucleotide was used. For antibody interference experiments, nuclear extracts (15 g) were incubated for 1 h at 4°C with 1 g of either anti-GLI2 (generous gift from Rüne Toftgard, Karolinska Institute, Stockholm, Sweden) or anti-TCF3 antibodies (Santa Cruz Biotechnology) before the addition of the radiolabeled probes (5 ϫ 10 4 cpm/lane). Binding mixtures were separated electrophoretically on native 4% acrylamide gels and visualized by autoradiography after gel drying.
Amplification of GAPDH sequences was used as an internal control with primers provided in the Active Motif kit. Amplimers were visualized with ethidium bromide after agarose gel electrophoresis.
In Silico Promoter Analysis-For bioinformatics analyses, the Genomatix software package (Genomatix Software, München, Germany) was used. Proximal promoter sequences (2 kb upstream and 0.2 kb downstream of the transcriptional start site) were extracted with help of the ElDorado module (release 4.6). Identification of transcription factor binding sites in the M-MITF promoter was performed using the MatInspector module of the Genomatix data base together with Matrix Family Library version 8.3 (October 2010) (25). All analyses were conducted with high threshold values (core similarity 1.0, matrix similarity 0.9). In a second set of experiments, SKMel28 melanoma cells, which express detectable levels of both M-MITF and GLI2 in the basal state, were transduced with either non-targeting or GLI2-specific shRNA lentiviral vectors. Quantitative RT-PCR was used to evaluate the efficacy of GLI2 knockdown; both constitutive and TGF-␤-induced GLI2 expression was reduced in Next, GLI2 expression was down-regulated in 888mel cells by means of transiently transfected GLI2-specific siRNA oligonucleotides. Unlike the partial effect of lentiviral knockdown (above), this transient siRNA approach fully abolished GLI2 induction by TGF-␤ (supplemental Fig. S1B, left panel). Yet it did not oppose the inhibitory effect of TGF-␤ on M-MITF expression (supplemental Fig. S1B, right panel) as seen above after stable GLI2 knockdown. These data are consistent with the hypothesis that mechanisms other than GLI2 induction are involved to mediate M-MITF inhibition by TGF-␤. Similar results were obtained in 501mel melanoma cells transfected with GLI2 siRNA (supplemental Fig. S1C).

Identification of Two Distinct Regulatory Regions Implicated in TGF-␤ Down-regulation of M-MITF Promoter-We used the
Genomatix software MatInspector module for in silico analysis of putative transcription factor binding sites present within the Ϫ334/ϩ1 region of the human M-MITF promoter ( Fig. 2A). Results are summarized in Fig. 2A. Noteworthy, a previously undescribed putative Kruppel-like transcription factor binding site was identified at position Ϫ316.
A battery of deletion constructs of the human M-MITF promoter was assessed for transcriptional regulation by either exogenous TGF-␤ stimulation or GLI2⌬N overexpression in transient cell transfection experiments with 888mel cells. As shown in Fig. 2B, efficient transcriptional repression (35-70%) of all 5Ј-end deletion constructs from Ϫ2135/ϩ136 to Ϫ177/ ϩ136 was observed in response to exogenous TGF-␤ (Fig. 2B,  central panel). Interestingly, and in contrast to the results obtained with TGF-␤, overexpression of GLI2⌬N only inhibited the activity of constructs extending upstream of nucleotide Ϫ334, whereas further 5Ј-end deletions of the promoter led to complete loss of responsiveness to GLI2 overexpression (Fig. 2B, right panel). These experiments gave matching results when repeated in 501mel cells (supplemental Fig. S2). Together, they identify the region Ϫ334/Ϫ296 of the M-MITF promoter as a GLI2-responsive region, whereas another regulatory element essential for transcriptional repression by TGF-␤ lies 3Ј of position Ϫ177. Noteworthy, transcriptional activity of the proximal region of the M-MITF promoter downstream of Ϫ177 has been previously shown to largely depend upon the functionality of a CREB binding site at position Ϫ140 (20).
To further establish the respective implication of these two promoter regions in the context of GLI2 and/or TGF-␤-mediated repression, 888mel human melanoma cells were transfected in parallel with Ϫ334/ϩ136MMITFluc and Ϫ177/ ϩ136MMITFluc reporter constructs together with either empty pcDNA or GLI2⌬N expression vector before TGF-␤ stimulation. Although the activity of both constructs was dramatically repressed by TGF-␤, only the longer one (Ϫ334/ ϩ136MMITFluc) was repressed by GLI2⌬N overexpression (Fig. 2C), in line with the data presented above. As expected, GLI2⌬N and TGF-␤ exerted an additive inhibitory activity on Ϫ334/ϩ136MMITFluc. Also, the extent of TGF-␤ repression on Ϫ177/ϩ136MMITFluc was similar in the absence or presence of GLI2⌬N overexpression. Together, these data demonstrate that GLI2 does not mediate TGF-␤ effect on M-MITF promoter inhibition.
Chromatin Immunoprecipitation Reveals GLI2 Binding to M-MITF Promoter-Our previous work identified melanoma cell lines with distinct levels of GLI2, directly associated with their invasive and metastatic capacity (8). Also, we determined that melanoma cell lines with high GLI2 expression have low M-MITF levels, whereas pigmented cell lines expressing M-MITF exhibit low GLI2 mRNA and protein levels (9). The latter is consistent with the above demonstration of transcriptional repression of M-MITF by GLI2. To study the possible recruitment of GLI2 to the GLI2-responsive region of the M-MITF promoter identified by 5Ј-end deletion studies between nucleotides Ϫ334 and Ϫ296 (see above), we performed a first set of chromatin immunoprecipitation (ChIP) experiments in three distinct melanoma cell lines previously shown to express high (1205Lu), intermediate (SKMel28), or low (888mel) levels of GLI2 protein (8). As shown in Fig. 3A, strong constitutive GLI2 recruitment to the Ϫ393/Ϫ206 region of the M-MITF promoter was observed in 1205Lu cells, although less and less detectable in SKMel28 and 888mel cells, thus following a pattern consistent with the respective steadystate expression levels of GLI2 in these cell lines.
As shown in Fig. 3B, left panel, GLI2 recruitment to the M-MITF promoter in SKMel28 cells was dramatically reduced after stable GLI2 knockdown, the latter also resulting in an upregulation of M-MITF mRNA steady-state levels (right panel).
On the other hand, in WM983A melanoma cells, in which we previously identified that TGF-␤ induces a solid SMAD-dependent transcriptional response (16,17) as well as strong GLI2 induction (8), GLI2 recruitment to the M-MITF promoter was enhanced in response to TGF-␤ (Fig. 3C, left panel), accompanied with reduced M-MITF expression (right panel). From these experiments, we conclude that levels of M-MITF pro-moter occupation by GLI2 in region Ϫ387/Ϫ296 are correlated with GLI2 expression levels.
Identification of Functional GLI2 Binding Site within Human M-MITF Promoter-Bioinformatics identified a potential binding site for KLFs; that is, CTCCTCCAAA at position Ϫ316/ Ϫ306 of the M-MITF promoter (see Fig. 2A). Given the data obtained from 5Ј-end promoter deletion (Fig. 2, B and C) and ChIP (Fig. 3) studies, we first hypothesized that this putative KLF binding site may bind GLI2. To address this issue, EMSA experiments were performed to analyze DNA-protein interactions within this M-MITF promoter region. As shown in Fig.  4A, incubation of nuclear extracts from 1205Lu melanoma cells, which strongly express GLI2, with a radiolabeled oligonucleotide representing region Ϫ334/Ϫ296 of the M-MITF promoter and encompassing the putative KLF binding site identified a slow-migrating band (lane 2) that was specifically eliminated by 100-fold excess unlabeled probe corresponding to a consensus GLI2 binding site (lane 3), not by a 100-fold excess consensus Sp1 oligonucleotide (lane 4). Furthermore, incubation of 1205Lu cell nuclear extracts with an antibody against GLI2 before the addition of the radiolabeled probe in the binding mixture abolished DNA-protein complex formation (lane 5), whereas an anti-TCF3 antibody used under the same conditions did not (lane 6). EMSA experiments performed with an additional radiolabeled oligonucleotide overlapping the KLF binding site (Ϫ334/Ϫ304) confirmed these results (supplemental Fig. S3A).
Next, the Ϫ316/Ϫ306 KLF/GLI2 binding site present in the Ϫ334/ϩ136 M-MITF promoter reporter construct was inactivated by either multiple simultaneous point mutations (mut-KLF) or nucleotide deletion (⌬KLF). Remarkably, parallel transient transfection experiments in 888mel cells revealed that both mutant constructs had increased basal activity (2-3-fold) as compared with their wild-type counterpart (Fig. 4B, open  bars). Furthermore, unlike what we observed with the wild-type Ϫ334/ϩ136 construct, neither of these mutants was inhibited by GLI2⌬N overexpression (Fig. 4B, solid bars). In 1205Lu cells, which express high constitutive levels of GLI2, both mutant constructs exhibited an even higher activity than the parent construct (up to 14-fold, supplemental  (20). PKA phosphorylation of CREB at serine 133 in response to cAMP promotes binding of CREB dimers to CREs (a conserved palindromic 8-bp sequence TGACGTCA) and allows their association with the transcriptional coactivators CBP and p300, thereby transactivating cAMP-responsive genes (25). Interestingly, we identified antagonistic interactions between the cAMP and TGF-␤/SMAD pathways (26,27), leading to opposite regulation of M-MITF and GLI2 expression in melanoma cells (9). We thus examined whether the proximal CRE may also play a role in mediating TGF-␤ effect. For this purpose, M-MITF promoter constructs were generated in which the CRE was inactivated by point mutation. Parallel transfection experiments of wild-type and CRE-mutated Ϫ334/ ϩ136 constructs in 888mel melanoma cells confirmed the importance of the CRE for promoter basal activity and revealed that integrity of the CRE is critical for TGF-␤ responsiveness (Fig. 5A, left panel). On the other hand, inhibition of M-MITF promoter activity by GLI2⌬N overexpression was similar in the presence or absence of an intact CRE (Fig. 5A, right panel), consistent with the results presented above implicating the Ϫ316/Ϫ306 KLF binding site for GLI2 responsiveness. Similar results were obtained when the CRE mutation was generated in the context of the longer Ϫ2135/ϩ136 promoter construct (supplemental Fig. S4A). As expected, mutation of the CRE in the Ϫ296/ϩ136 construct (which lacks the KLF binding site) rendered the promoter not only unresponsive to GLI2 overexpression but also to exogenous TGF-␤ (supplemental Fig. S4B). All these experiments were repeated in the 501mel melanoma cell line and gave sub-identical results (supplemental Fig. S4C).
To gain insight into the mechanism(s) by which the CRE is involved in TGF-␤ down-regulation of M-MITF promoter activity, we examined the effects of TGF-␤ on PKA and CREBspecific transcriptional activity. As shown in Fig. 5B, left panel, incubation of 1205Lu or 888mel melanoma cells with TGF-␤ over a 24-h period led to a 45-55% drop in constitutive PKA  3 and 4, respectively). Antibody interference studies were performed with 1 g of either anti-GLI2 (lane 5) or anti-TCF3 (lane 6) antibodies. Similar results were obtained with radiolabeled probes spanning regions Ϫ325 to Ϫ296 (supplemental Fig. S3A). NS, not significant. B, 888mel melanoma cells were transiently transfected with either empty (pc) or constitutively active GLI2 mutant (GLI2⌬N) expression vectors together with either wt, mutKLF, or ⌬KLF Ϫ334/ϩ136 M-MITF promoter constructs. Promoter activity was determined 24 h later. C, 888mel melanoma cells were transiently transfected with either wt, mutKLF, or ⌬KLF Ϫ334/ϩ136 M-MITF promoter constructs, and incubated overnight with TGF-␤, after which promoter activity was determined. Note that inactivation of the KLF binding site prevents GLI2, not TGF-␤, inhibitory effect.
activity. This phenomenon was associated with reduced CREB phosphorylation levels (Fig. 5B, center panel) and CREB-dependent transcription, as measured in transient cell transfection experiments with the CREB-specific pCRE-luc reporter plasmid in both cell lines (Fig. 5B, right panel). Similar results were obtained in the 501mel cell line (supplemental Fig. S4D).
We next determined the kinetics of TGF-␤ effect on both PKA activity and GLI2 expression in 888Mel melanoma cells. A rapid and dramatic (up to 75%) reduction in PKA activity was observed 30 min after TGF-␤ addition (Fig. 5C) followed by a partial and transient recovery at 2 h followed by prolonged secondary inhibition of PKA, around 50% of the initial enzyme activity, from 8 to 48 h. In parallel, maximal GLI2 mRNA induction was found at 8 h (ϳ10-fold) and remained high (at least 6-fold above basal expression levels) over the course of the experiment, as measured by quantitative RT-PCR.
Simultaneous Inactivation of Both KLF Binding Site and CRE of M-MITF Promoter Is Required to Abolish Both TGF-␤ and GLI2 Inhibitory Activities-To definitely establish the functionality of both KLF and CREB binding sites within the M-MITF promoter, simultaneous inactivation of both sites was performed in the Ϫ334/ϩ136 construct, which contains sufficient regulatory sequences to confer inhibition in response to either TGF-␤ or GLI2 (see Fig. 2). Responsiveness of the double mutant (mutKLF/mutCRE) promoter to TGF-␤ or GLI2⌬N overexpression was compared with that of constructs in which only one of the two regulatory elements was nonfunctional (mutCRE or mutKLF). As shown in Fig. 6A, wt and mutKLF constructs were inhibited by TGF-␤, whereas mutCRE and mutKLF/mutCRE were not. On the other hand, wt and mut-FIGURE 5. Inhibition of protein kinase A activity and CREB-dependent transcription by TGF-␤. A, subconfluent 888mel human melanoma cells were transfected with wt and mutCRE Ϫ334/ϩ136 M-MITF promoter constructs before overnight TGF-␤ stimulation (left panel). Alternatively, the constructs were co-transfected with either empty (Ϫ) or GLI2⌬N (ϩ) expression vectors (right panel), and luciferase activity was measured 24 h later. B, subconfluent 1205Lu and 888mel human melanoma cells were incubated in the absence or presence of TGF-␤ for 24 h. Protein kinase A activity was measured using a fluorescent substrate based assay kit (left panel). Alternatively, total protein extracts were subjected to Western blotting to assess CREB and P-CREB levels (center panel). An anti-HSP60 was used as control. Subconfluent 1205Lu and 888mel human melanoma cells were transfected with the CREBspecific reporter construct CRE-luc before stimulation with TGF-␤ (right panel). C, 888mel melanoma cells were incubated for various time points with TGF-␤. PKA activity and GLI2 mRNA steady-state levels were measured in parallel dishes. CRE were inhibited by GLI2⌬N overexpression, not mutKLF or mutKLF/mutCRE (Fig. 6B). Thus, only the double mutant construct was unresponsive to both TGF-␤ and GLI2⌬N overexpression.
Together with the data presented in Fig. 5, these experiments suggest that TGF-␤ effect on M-MITF promoter activity occurs primarily via the proximal CRE. The repressed state of the M-MITF promoter may be maintained over time by additional mechanisms involving GLI2 acting through the KLF binding site. The integrity of the latter is not required for M-MITF repression by TGF-␤.

DISCUSSION
Interest in studying the transcriptional control of M-MITF expression is motivated by the critical role played by this transcription factor at all stages of normal and pathological development of the melanocyte lineage (11,12). Most critically, M-MITF is involved in various steps of melanoma progression. In particular, M-MITF is essential for melanoma cell proliferation through the control of cell cycle progression. In this study we have dissected the molecular mechanisms by which TGF-␤ down-regulates M-MITF gene transcription in melanoma cells. We identify two distinct mechanisms involved in TGF-␤ and GLI2 action. Using transient transfection assays and site-directed mutagenesis to inactivate the proximal CREB binding site of the M-MITF promoter, we determined that the latter is absolutely critical for inhibition of M-MITF transactivation by TGF-␤. Inhibition of PKA activity by TGF-␤ results in reduced CREB phosphorylation, an essential step for full transcriptional activity of the CREB transcription factor. Another mechanism is implicated in the down-regulation of M-MITF expression by the GLI2 transcription factor. Using a combination of ChIP, EMSA, site-directed mutagenesis, and transient cell transfection assays, we found that GLI2 exerts a negative regulatory activity on M-MITF transcription via direct binding to a previously uncharacterized KLF/GLI2-specific cis-element located in region Ϫ324/Ϫ294 of the M-MITF promoter. We previously identified GLI2 as a transcriptional target of TGF-␤ signaling whereby gene transactivation occurs in response to TGF-␤-dependent recruitment of both SMAD and ␤-catenin to the GLI2 promoter (6). GLI2, however, was not required for TGF-␤ effect on M-MITF transcription. A schematic representation of our findings is provided in Fig. 7. Yet, GLI2, which can accumulate in melanoma cells after TGF-␤ stimulation and possibly in response to other stimuli that remain to be identified, contributes to the long term maintenance of low M-MITF levels in certain melanoma cell lines. Indeed, we have shown previously that melanoma cells that constitutively express high levels of GLI2 have very low M-MITF expression, whereas on the other hand, cell lines with high M-MITF levels express little GLI2 (9).
Although these findings represent a leap forward in the identification of novel mechanisms of M-MITF silencing in melanoma cells, the transcriptional repressors associated with GLI2 remain to be identified. One possibility may be that the oncoprotein c-Ski or its homolog, SnoN, abundantly expressed in melanoma cells (28,29), may function as co-repressor for GLI2. They have previously been characterized as transcriptional corepressors of the SMAD transcription factors, as they recruit a repressor complex comprising N-CoR SMRT, Sin3A, and HDAC-1 to the target gene promoters, leading to gene silencing (30,31). A two-hybrid screen with the N-terminal fragments of both GLI2 and GLI3 has found that both c-Ski and SnoN interact with GLIs and repress GLI-dependent transcription (32). Work is in progress to further characterize the mechanisms implicated in GLI2-driven M-MITF transcriptional repression and the possible implication of c-SKI/SnoN. The high functional redundancy of c-Ski and SnoN makes it a difficult task, as we have not been able thus far to perform efficient and concomitant knockdown of both proteins in melanoma cells.

CONCLUSION
We previously demonstrated the pro-invasive and pro-metastatic role played by TGF-␤ and GLI2 in various experimental settings to address melanoma cell behavior both in vitro and in vivo (16,29,(33)(34)(35). This work extends our understanding of the transcriptional control of M-MITF expression by TGF-␤ and GLI2. In particular, we provide important new information regarding the fine-tuning of M-MITF promoter regulation by TGF-␤ and GLI2 and unveil the independent contribution of a known proximal CRE and that of a novel KLF/GLI regulatory element to, respectively, allow efficient transcriptional repression of M-MITF by TGF-␤ and GLI2.