The present study used eight melanoma cell lines (listed in Table SI). Their mutational status (BRAF and NRAS mutations) is shown in Table SII. The origin of cells has been described previously (31,32). Cells were cultivated in RPMI medium (MilliporeSigma) supplemented with 10% FCS (Thermo Fisher Scientific, Inc.), glutamine and antibiotics (MilliporeSigma) at 37°C and 5% CO₂ in 100% humidity. All cell lines were authenticated and tested for mycoplasma using a mycoplasma detection kit (MP0035; MilliporeSigma). Cells were passaged every 72–96 h using a trypsin-EDTA solution. When plated, cells (5×10⁵) were seeded from a stable culture into 12-well plates and incubated for 24 h at 37°C prior to inhibitor treatment or transfection, unless specified otherwise. Normal human melanocytes were purchased from Cascade Biologics Inc. and cultivated according to the manufacturer´s instructions. The generation of inducible melanoma cell lines in which melanoma-associated transcription factor (MITF) protein can be downregulated by the addition of doxycycline (Tet-on system) has been previously described (32).

GANT61 (stock prepared in DMSO) and cyclopamine (stock prepared by dissolving in ethanol) were purchased from Selleck Chemicals LLC. The chemicals were applied to cells as indicated in the appropriate figures for 20 h before cell harvesting if not stated otherwise. The addition of 20 µM GANT61 for 20 h was used after the optimization of both the concentration and time to follow the changes in expression of GLI-dependent genes, when no signs of apoptosis had been yet detected in cells.

To obtain whole-cell extracts for immunoblotting analysis, cells were lysed in RIPA buffer (1% NP-40, 150 mM NaCl, 5 mM EDTA, 0.5% sodium deoxycholate, 50 mM Tris-HCl pH 7.5 and 0.1% SDS) with the addition of the protease and phosphatase inhibitors aprotinin, pepstatin and leupeptin at 1 mg/ml each. cOmplete (Roche Diagnostics) was added as recommended by the supplier. Then, 1 mM phenylmethylsulfonyl fluoride (MilliporeSigma) and phosphatase inhibitor PhosSTOP (Roche Diagnostics) were added. Equal amounts of protein (30 µg; concentration determined by Bradford's assay) were loaded on 10–12% SDS-polyacrylamide gels, separated by electrophoresis and transferred onto PVDF membranes. Blots were blocked in 5% non-fat dry milk Blotto (cat. no. sc-2325, Santa Cruz Biotechnology, Inc.), and incubated with 1:1,000 diluted primary antibodies for 2 h, washed, and then incubated at room temperature for 1 h with 1:4,000 diluted secondary anti-mouse or anti-rabbit horseradish peroxidase-conjugated antibodies (cat. no. sc-2055 or cat. no. sc-2030; Santa Cruz Biotechnology, Inc.). Chemiluminescent detection was used. For western blotting shown in Fig. 5A and B, cells were lysed in PLB buffer (Promega Corporation), used in dual luciferase measurements and these extracts were then directly utilized. Primary antibodies for western blotting were: GLI1 (cat. no. 3538; Cell Signaling Technology, Inc.), GLI2 (cat. no. TX46056; GeneTex, Inc.), and GLI3 (cat. no. AF3690; R&D Systems, Inc.). Anti-Slug (cat. no. sc-166476) was purchased from Santa Cruz Biotechnology, Inc. Klf4 (cat. no. LS-C415468) and ALDH1A1 (cat. no. LS-B10149) from LSBio. Anti-FLAG (M2; cat. no. F1804) and anti-β-actin (A5316) antibodies were purchased from MilliporeSigma. Anti-livin antibody (cat. no. sc-30161) was from Santa Cruz Biotechnology, Inc. and MITF (D5) antibody (cat. no. NBP2451590) was from Thermo Fisher Scientific, Inc. Brn-2 (cat. no. sc-514295) was from Santa Cruz Biotechnology, Inc. E- and N-cadherins, vimentin, Zeb1, and Zeb2 antibodies were from Cell Signaling Technology (cat. no. 9782). Western blot bands were quantified using ImageJ (v. 1.52j) software (National Institutes of Health; data not shown).

The 12×GLI-TK-Luc plasmid was obtained from Professor R Toftgard, Karolinska Institute, Sweden. pGL3-PTCH1 was kindly donated by Professor Aberger, University of Salzburg, Austria. The PATCHED gene promoter contains one active GLI-binding site responsible for its activity (data not shown). The pGL3-slug-prom-full-length promoter (−5216+112) and pGL3-slug-prom-Δmiddle (−5216-4635, −2092+112) were cloned as XhoI-HindIII (New England BioLabs, Inc.) inserts in the pGL3-basic plasmid (Promega Corporation). pGL3-slug-prom-Δproximal (−5216-2092) was cloned as the XhoI-HindIII insert. The-4635+112 promoter was cloned as the NheI-HindIII insert. pGL3-slug-prom-middle (−4635-2092) was cloned as the NheI-NheI (New England BioLabs, Inc.) insert and the pGL3-slug-proximal promoter (−2092+112) was cloned as the NheI-HindIII insert. Cloning of all GLI expression vectors has been described previously (31). Briefly, original GLI1 (GL1; cat. no. 16419), GLI2 (pCS2-MT GLI2 FL; cat. no. 17648), ΔNGLI2 (pCS2-MT GLI2 delta N; cat. no. #17649) and GLI3 (GLI3 bs-2; cat. no. 16420) were acquired from the nonprofit plasmid repository Addgene, Inc. Respective coding sequences were amplified by PCR and subsequently cloned into the pcDNA3 expression vector or into the pFLAG-CMV-4 plasmid (MilliporeSigma) background to obtain FLAG-tagged GLI proteins. The types of GLI expression plasmids were similarly effective in promoter activation. The melastatin promoter plasmid with the three MITF-binding sites was constructed as previously described (33). The construction of the short hairpin plasmid directed to MITF has been previously described (32). All final plasmid inserts were verified by sequencing on both strands (GATC Biotech).

Transient cell transfections of the promoter reporters were performed using 12-well plates at 37°C, seeded and incubated for 24 h before transfections, and the transfection reagent Mirus TransIT-2020 (Mirus Bio LLC) following the manufacturer's instructions and harvested 48 h after transfection. For detection, a dual luciferase system (Promega Corporation) was used according to the manufacturer's instructions. The pGL3-basic vector was used as a negative control. Expression vectors or controls (pCDNA3 or pFLAG-CMV-4) were cotransfected together with the reporter plasmids. Cell lysates were used for dual luciferase assays performed as recommended by the manufacturer's instructions. The luciferase values were acquired on a Turner Designs 20/20 luminometer (Promega Corporation). Data were normalized to Renilla luciferase activity (internal control) as arbitrary units. Statistical analysis of luciferase assay results was performed using a two-tailed unpaired Student's t test and SigmaPlot software V10.0 (Systat Software Inc.). Where indicated, cells were treated with GANT61 or cyclopamine 20 h at 37°C before harvesting. Tomatidine, a compound inactive in the HH pathway but structurally similar to cyclopamine, was tested as a negative control and revealed similar results as vehicle (data not shown).

Total RNA was isolated using TRIzol^® (Invitrogen; Thermo Fisher Scientific, Inc.) according to the supplier´s instructions (3×10⁵ cells per 30 mm well). Then 2 µg of RNA was reverse transcribed using SuperScript IV reverse transcriptase (Thermo Fisher Scientific, Inc.), and qPCR was performed using a TaqMan QuantiTect Probe PCR kit (Qiagen GmbH) on a ViiA7 Real-Time PCR system (Thermo Fisher Scientific, Inc.) following the manufacturer's instructions (cycling: 30 sec at 94°C and 1 min at 60°C). Data analysis was performed by QuantStudio 6 Software (Thermo Fisher Scientific, Inc.). Concurrent results were obtained in three independent experiments with the following PCR primers and probe for SLUG: Forward, 5′-AGAACTCACACGGGGGAGAAG-3′; reverse, 5′-CTCAGATTTGACCTGTCTGCAAA-3′; probe, 5′-6-FAM-TTTTTCTTGCCCTCACTGCAACAGAGC-TAMRA-3′. β-Actin was utilized as an internal standard control: Forward, 5′-ATTGCCGACAGGATGCAGAA, reverse, 5′-GCTGATCCACATCTGCTGGAA; probe, 6-FAM-CAAGATCATTGCTCCTCCTGAGCGCA-TAMRA. Fluorescent probes were purchased from Thermo Fisher Scientific, Inc.

501mel cell cultures were transfected using Mirus TransIT-2020 (Mirus Bio LLC) with the pcDNA3-GLI1, pcDNA3-GLI2 or pcDNA3-GLI3 expression plasmids in 90 mm plates. After 24 h of treatment, fresh medium was applied. After another 24 h, cells were crosslinked with 1% formaldehyde for 10 min at room temperature and incubated with glycine solution, and chromatin immunoprecipitation was performed by using the ChIP-IT High Sensitivity kit (Active Motif, Inc.) in accordance with the manufacturer's protocols. Anti-acetylated histone H3 antibody (MilliporeSigma) was used as the positive control, and rabbit nonimmune IgG (MilliporeSigma) was used as a negative control. To detect GLI1 bound to the promoter, a rabbit anti-GLI1 antibody (cat. no. sc-20687; Santa Cruz Biotechnology, Inc.) was used. Rabbit anti-GLI2 (cat. no. ab26056; Abcam) was used to precipitate GLI2. For GLI3, a rabbit anti-GLI3 antibody (cat. no. 3538; Cell Signaling Technology, Inc.) was used and 2 µg of each antibody was added in each reaction. To assess the amount of ChIP-generated DNA, PCR was performed using Phusion High-Fidelity DNA polymerase (Thermo Fisher Scientific, Inc.). The amplification of the regions of the proximal SLUG promoter was performed with four alternative primer pairs: Region A (−2108-1766): sense, 5′-GCATACGTGTTACTCGCTAGCAC-3′, antisense, 5′-TCCTTGTTTCACTCTACACAGTCTATTCAC-3′; region B (−1769-1163): sense, 5′-AGGAAATAATAGCCATGGCGATA-3′, antisense, 5′-CATCTCTGTCCATTGCAGAC-3′; region C (−1182-490): sense, 5′-GTCTGCAATGGACAGAGATGC-3′, antisense, 5′-GGGAAGCGGGAAGACAAAG-3′; and region D: (−509+112): sense, 5′-CCTTTGTCTTCCCGCTTCCCCCTTCC-3′, antisense, 5′-ACACGGCGGTCCCTACAGCATCG-3′. PCR products were resolved on 1% agarose gels. The intensity of the final PCR bands was quantified using ImageJ (v. 1.52j) software (National Institutes of Health).

For immunofluorescence, four cell lines (501mel, Hbl, SK-MEL-5 and SK-MEL-28) were seeded into 8-well Lab-Tek II Chamber Slides (Thermo Fisher Scientific, Inc.). After 48 h, 20 µM GANT61 was added for 20 h at 37°C. Vehicle (DMSO) alone was added to the controls. The cells were fixed with 3% paraformaldehyde at room temperature for 10 min, washed, permeabilized with 0.1% Triton X-100, and blocked with 5% goat serum. Slides were then stained with Slug primary antibody (1:1,000; cat. no. sc-166476; Santa Cruz Biotechnology, Inc.) at room temperature for 2 h followed by 1:1,000 secondary anti-mouse fluorescein-coupled antibody (cat. no. FI-2000-1.5, Vector Laboratories, Inc.) and mounted in medium with DAPI to stain nuclei. The visualization was performed using an Olympus IX70 microscope with cellSens software V2.2 (Olympus Corporation).

Formalin-fixed paraffin-embedded tissues (skin, nevus and melanoma) were retrieved from the archive of the Department of Pathology and Molecular Medicine, Second Faculty of Medicine, Charles University, University Hospital Motol, Prague. At least four samples of each tissue were processed and similar results were obtained. Deparaffinized, rehydrated in descending ethanol series, and blocked (with 3% H₂O₂) sections were stained with 1:1,000 primary antibodies. Immunohistochemistry for MITF was performed with the primary antibody MITF (cat. no. D5; cat. no. NBP2451590, Thermo Fisher Scientific, Inc.). GLI2 was stained with antibody cat. no. GTX46056 (GeneTex, Inc.) and Slug with anti-Slug (A-7) sc-166476 antibody (Santa Cruz Biotechnology, Inc.). Detection was performed using an EnVision+ avidin-biotin detection system (Dako; Agilent Technologies, Inc.). Each tissue was examined by two pathologists. Tissues were scored on a scale of 0 to 4 based on the combined extent and intensity of staining. The final score represented the predominant staining pattern of both combined parameters. Section fields were imaged using a BX51 microscope (Olympus Corporation) equipped with a PROMICAM 3-3CP 3.1 camera and QuickPHOTO CAMERA 3.2 software (Olympus Corporation).

Statistical significance (P-values) was assessed using a two-tailed Student's t test and Mann-Whitney test. Standard error of the mean values are depicted in graphs as bars within each column in the reporters and RT-qPCR. Data not significant (P>0.05) were not labeled, values of 0.01 <P<0.05 are marked by an asterisk, and values with P<0.01 are marked by two asterisks. SigmaPlot software V10.0 (Systat Software Inc.) and GraphPad Prism v.8.4.3 software (Dotmatics) were used to perform statistical analysis. Statistical analysis was verified using one-way ANOVA followed by post hoc tests as specified in figure legends. P<0.05 was considered to indicate a statistically significant difference.

Snail1 and Snail2 (Slug) are among the main players in the tumorigenic program of EMT (19). SLUG gene deletions have been found in Waardenburg syndrome and piebaldism in humans (34,35), indicating the involvement of Slug in melanin pigmentation. The SLUG gene has long been thought to be under the transcriptional control of MITF (34). To further explore the regulation of Slug expression in melanoma, the present study examined the SLUG promoter and found 85 GLI-binding sites in the long SLUG promoter (−5216 +112; the start of translation is +1) and 31 sites in the proximal promoter (−2092 +112; Fig. 1A and Table SIII). Although none of the sites was a full consensus, GLI proteins, which are the final executive factors of the Hedgehog signaling pathway, bind to and are also active from sites with two or three mismatched nucleotides. This implies that SLUG gene expression could be controlled by the Hedgehog pathway.

To further investigate this hypothesis, Slug protein expression was examined in melanoma cell lines treated with GANT61, a potent and specific pan-GLI transcriptional inhibitor. It was found that GANT61 decreased Slug protein in all eight melanoma cell lines investigated as well as in normal melanocytes (Figs. 1B and SI). BRAF- or NRAS-mutated or wild-type cells for both oncogenes were present among the analyzed melanoma cell lines (Table SII). The levels of other proteins known to be involved in EMT, such as E- and N-cadherins, vimentin, Zeb1 and Zeb2, were not appreciably changed after GANT61 treatment in six melanoma cell lines. Likewise, the CSC markers ALDH1A1 and Klf4 were generally only slightly affected (Fig. 1B). A mild increase in Klf4 was noted in the cell lines 501mel, MeWo, SK-MEL-5 and SK-MEL-28, while ALDH1A1 levels were somewhat increased in 501mel cells and slightly diminished in SK-MEL-28 cells after GANT61 treatment (Fig. 1B). Only Brn2 (N-Oct-3, POU3F2) protein, a known repressor of MITF (36), noticeably decreased in five of six cell lines, suggesting that its transcription may also be regulated by the HH pathway in most melanomas. MITF was slightly downregulated in three cell lines (Fig. 1B). Immunofluorescence staining confirmed blunting of Slug expression after the treatment of melanoma cells with GANT61 (Fig. S2).

Since the transcriptional inhibitor of GLI factors GANT61 downregulated Slug protein expression, the activity of the SLUG promoter was analyzed in reporter assays. Examination of luciferase expression driven by the long (full-length) promoter (−5216+112) and its truncated versions revealed that the middle part of promoter (−4635-2092) was the most active fragment (Fig. 2A). Its deletion greatly decreased the activity of the full-length promoter. The short portion most upstream (−5216-4635) probably performs an inhibitory function because its presence in the contexts of the full-length promoter, middle part and proximal promoter (−2092+112) decreased the luciferase activity (Fig. 2A). As the proximal promoter also showed appreciable activity, it was used in the following experiments.

Next, to test whether the SLUG promoter-reporter is directly activated by cotransfected expression vectors for GLI factors and inhibited by GANT61, it was first investigated whether the proximal promoter activity decreased after the addition of HH pathway inhibitors. Indeed, both GANT61 and cyclopamine lowered the activity in all three melanoma cell lines tested (Fig. 2B). As a subsequent experiment, each of the expression vectors for GLI factors (GLI1-3) were cotransfected with the GLI-responsive promoter containing 12 canonical GLI-binding sites (12×GLI). First, it was ascertained that there was only a minimal difference between the cotransfected control (pcDNA3) and GLI vectors with the pGL3-basic empty control promoter (Fig. 2C, left). By contrast, all three GLIs greatly increased 12×GLI promoter activity compared with the negative control plasmid pcDNA3 (Fig. 2C, right). GLI3 showed the weakest activation of the canonical 12×GLI promoter (but still ~80-fold), whereas the highest activation (800-fold) was achieved by ΔGLI2, a truncated version of GLI2 in which the N-terminal repression domain was removed (37). The activity of the GLI factors was then tested on a known HH-responsive promoter of the PATCHED gene, also a component of the HH pathway. The results were similar to those obtained with the canonical promoter, but the extent of stimulation was substantially lower, ~6-fold in the case of the most active ΔGLI2 (Fig. 2D, first graph). Similar experiments were then conducted with the three types of SLUG promoters, the long (full-length), Dmiddle and proximal promoters (Fig. 2D, second, third and fourth graphs). In all cases, the DGLI2 construct again showed the best activation. In accordance with the results in Fig. 2A, the Dmiddle promoter exhibited the lowest overall activity. When the PATCHED and SLUG promoters were tested, the addition of GANT61 more or less decreased the GLI-stimulated promoter activities (Fig. 2B and D). Western blotting verified that all GLI proteins were similarly expressed (Fig. 2E). These results together indicated that SLUG promoter activity is dependent mainly on GLI2 in reporter assays and that stimulation by GLI factors can be inhibited by GANT61.

To further evaluate the transcriptional regulation of Slug by HH/GLI, the present study examined the effect of GANT61 on Slug RNA levels using real-time PCR. RT-PCR was performed first, followed by qPCR. In all eight melanoma cell lines tested, 20 µM GANT61 significantly (P<0.01) lowered the mRNA level of Slug. This indicates that the positive effect of GLI factors on endogenous SLUG gene expression is mediated through the activation of transcription (Fig. 3).

The proximal SLUG promoter contains 31 GLI-binding sites, of which only two sites harbor two mismatches and all other sites have three mismatches (Fig. 2A and Table SIII). To investigate whether these sites are occupied by GLI proteins in cells, the 501mel cell line was used to chromatin immunoprecipitate DNA fragments bound to GLIs using specific anti-GLI1, GLI2 and GLI3 antibodies (Fig. 4). To obtain improved insight into whether the particular GLI binds to specific regions of the proximal promoter (−2092+112), the immunoprecipitated purified DNA was amplified by four primer pairs (see Materials and methods) that demarcated the four subregions A-D (Fig. 4) of the promoter. Each subregion contained several GLI-binding sequences. The most distal region A and a middle region C were bound by all three GLI factors. The middle region B was remarkably occupied only by the GLI3 factor. The most proximal fragment D was bound by GLI1 and GLI2 but not GLI3 (Fig. 4). Thus, regions B and D showed some preference for GLI occupancy, whereas regions A and C clearly exhibited enrichment by all three GLI factors. Acetylated histone H3, which was precipitated by anti-acetylated H3 antibody as a positive control, showed occupancy at all four subregions. Nonimmune IgG as a negative control showed no binding in any experiment. Taken together, these data indicate that GLI transcription factors are abundantly present at their recognition DNA sites within the SLUG proximal promoter, further confirming their role in the transcriptional regulation of Slug expression through HH signaling. The data quantifying the final DNA amount formed through PCR are in Table SIV.

The SLUG gene has been demonstrated to be transcriptionally regulated by MITF in melanocytes (34) and during Xenopus laevis development (38). However, data relevant to a possible regulation of Slug expression by MITF in melanoma cells are lacking. To test whether MITF overexpression influences the activity of the SLUG gene promoter, we performed promoter-reporter assays in which the proximal promoter was cotransfected with the MITF expression construct into 501mel cells. Additionally, we compared activation by MITF-Vp16, a hyperactive MITF derivative in which the MITF N-terminal activation domain (AD) was replaced by a stronger Vp16 AD (39), with native MITF. While MITF had no effect on promoter activity, MITF-Vp16 increased it ~2-fold (Fig. 5A left). On the other hand, the melastatin promoter, a known MITF target (33), was stimulated by both MITF and MITF-Vp16 ~4-fold and 10-fold, respectively (Fig. 5A middle). Western blotting verified the total MITF protein level (the control sample also showed the MITF protein signal because relatively high endogenous MITF protein is already present in nontransfected 501mel cells; Fig. 5A, right). Only transfected cells with ectopic MITF or MITF-Vp16 were measured for luciferase activity, which explained why promoter activity increased more compared with the overall MITF protein level. The same experiment was repeated with the FLAG-tagged vectors, and the same results were obtained. The expression of proteins expressed from transfected plasmids was verified with the anti-FLAG antibody (Fig. 5B).

To corroborate these results, doxycycline-regulatable melanoma cell lines, in which MITF could be downregulated by inducible expression of shRNA directed at MITF were used (32). Whereas a decrease in endogenous MITF protein was achieved after the addition of doxycycline in all cell lines, Slug expression remained unchanged. By contrast, the level of livin (ML-IAP), a known MITF target (40), mirrored the decrease in MITF protein (Fig. 5C). In agreement with this, overexpression of MITF did not change the endogenous level of Slug protein (data not shown). In a control experiment, inducible control nontargeting shRNA revealed no changes in the expression of all proteins tested. Thus, endogenous Slug seems to be expressed independently of MITF protein levels in human melanoma cells.

To further investigate the relationship between MITF and Slug and GLI2 vs. Slug protein expression in human samples, parallel sections of skin, nevus and melanoma metastasis were compared by immunohistochemical staining (Fig. 6). Positive staining by anti-GLI2 and anti-Slug antibodies was observed in the epidermis of normal skin (scored 2–3), consistent with previous observations (41,42). Only a few strongly MITF-positive (score 4) melanocytes were present in the epidermis. GLI2 and Slug were also stained in nevus cells, albeit somewhat less intensively (score 2, rare cells scored 3 in Slug staining). The nevus showed abundant MITF-positive cells (score 4). Epidermal keratinocytes were, as expected, MITF-negative in the skin and nevus sections (score 0). In metastatic melanoma, both GLI2 and Slug staining was positive in ~half of the cell population (scored 2–4), with Slug staining slightly more positively than GLI2 staining. MITF-specific staining of metastatic melanoma was negative (score 0), with only a small number of positive cells (not shown in the picture). Sections that were negative or very slightly positive for MITF (scored 0–1) showed cells stained for both GLI2 and Slug (scores 2–3; Fig. 6). This is also consistent with the idea that invasive and metastatic cells have low or absent MITF (43). Notably, all three proteins were almost exclusively localized to the nucleus. Thus, Slug staining was associated with GLI2, but not MITF, positive staining in immunohistochemical sections of melanomas (Fig. 6). This result supports the aforementioned findings by demonstrating the positive regulation of Slug expression by GLI2, but not by MITF, in melanoma cells.