Antitumor activity and expression profiles of genes induced by sulforaphane in human melanoma cells

DL-SFN, dimethyl sulfoxide (DMSO), propidium iodide (PI), Ponceau S and collagenase were purchased from Sigma-Aldrich. Trypan blue, Dulbecco’s modified Eagle medium (DMEM), Roswell Park Memorial Institute (RPMI) 1640, l-glutamine 100×, foetal bovine serum (FBS), penicillin–streptomycin, 0.25% trypsin–EDTA, Medium 254, human melanocyte growth supplement-2 PMA-Free (HMGS-2), RNAse-A, DEPC–water, collagen I bovine, Hank’s balanced salt solution (HBSS), trypsin neutralizer solution and 0.025% trypsin/EDTA solution were obtained from Gibco Life Technologies. Other chemical supplements and compounds were purchased from several companies as follows: CellTiter-Glo luminescent cell viability assay (Promega), turbo DNA free (Ambion, Life Technologies), FITC Annexin V apoptosis detection kit I (BD Pharmingen), TRIzol reagent (Ambion, Life Technologies), methylene blue (Merck), BCA protein assay kit and NP-40 (Pierce, Thermo scientific), ECL Prime Western blotting detection reagents (Amersham), stripping buffer stripAblot and Tween 20 (EuroClone), complete protease inhibitor cocktail (Roche), PageRule prestained protein ladder (Fermentas, part of Thermo Fisher Scientific), superscript III Reverse Transcriptase (Life Technologies) and iQ™ SYBR^® green supermix (Bio-Rad).

A375 is a primary human malignant melanoma cell line that was obtained from American Type Culture Collection (ATCC). The cells were maintained in DMEM medium containing 10% FBS, 1% l-glutamine and 1% penicillin–streptomycin (10,000 units/ml). The 501MEL cell line was obtained from surgically removed tumours from patients with melanoma at the National Cancer Institute in Milan, Italy, and they were maintained in RPMI 1640 medium supplemented with 10% FBS, 1% l-glutamine (200 mM) and 1% penicillin–streptomycin (10,000 units/ml). Human epidermal melanocyte (HEMa) is a cell line isolated from lightly pigmented (LP) adult skin purchased from Invitrogen (Life Technologies). The cells were propagated in Medium 254 supplemented with HMGS-2 and incubated at 37 °C with 5% CO₂ until they reached approximately 70% confluence. They were then treated with different concentrations of SFN dissolved in dimethyl sulfoxide (DMSO). The same DMSO concentration used to dilute the SFN was utilized as a negative control (untreated).

Cells were seeded at a density of 3 × 10⁵ cells/well onto six-well plates, and after 24 h treated with SFN at final concentrations of 1, 2 and 5 μg/ml. The control cells were treated with the same concentration of DMSO (0.001%). The cell lines were photographed under a phase-contrast microscope at 10× magnification (Nikon Eclipse) to observe morphological changes. For the viability assay, A375 and 501MEL cells were plated at a density of 3 × 10⁵ cells/well and treated with SFN for 24 and 48 h. Dead cells were stained with 0.4% trypan blue dye. In addition, cell viability was evaluated using the Cell Titer-Glo assay. The cells were plated at a density of 5 × 10³ cells/well in 100 μL of medium on 96-well white plates (Cellstar, Greiner) and treated with different concentrations of SFN for 24 and 48 h. The cells were subsequently lysed using Cell Titer-Glo luminescent reagent following the manufacturer’s protocol, and the results are expressed as a percentage based on the ratio of the absorbance of treated cells to that of the control cells (100%).

Cells were plated at a density of 3 × 10⁵ cells/well on six-well plates and treated with SFN for 24 and 48 h. After incubation, the cells were trypsinized, washed with ice-cold phosphate-buffered saline (PBS) and fixed in 70% ethanol overnight. Next, the cells were re-suspended in PBS containing RNAse-A (30 mg/mL) and incubated at 4 °C for 5 min. After FACS buffer (PBS + 2% FBS) and PI (1 mg/mL) were added, the cells were incubated at 4 °C for 30 min and then immediately analysed for DNA content using the FACSCalibur flow cytometer with CellQuest software (Becton–Dickinson). In total, 20,000 events per sample were recorded, and experiments were performed in triplicate. Apoptotic cells with hypodiploid DNA content were measured by quantifying the sub-G₁ peak in the cell cycle pattern.

Cells were plated at a density of 3 × 10⁵ cells/well on six-well plates, and 24 h later treated with SFN for 24 and 48 h. Floating cells were collected, combined with the trypsinized adherent cells and stained with Annexin V and PI according to the manufacturer’s recommended protocol. The samples were analysed by FACS within 1 h. For this assay, 20,000 events were counted. The analyses were performed in triplicate.

Cells were washed with PBS and then lysed with lysis buffer [1 M Tris, 2.5 M NaCl, 10% glycerol, 0.5 M glycerophosphate, 1% Tween 20, 0.5% NP-40 and a complete protease inhibitor tablet (Roche)] for 20 min on ice. The cell lysates were separated by 10 or 12% denaturing SDS-PAGE. The proteins were then transferred to nitrocellulose membranes (Whatman), blocked with PBS containing 5% milk and 0.05% Tween and incubated with specific primary antibodies overnight. After being washed with PBS containing 0.05% Tween, the blots were incubated with peroxidase-conjugated secondary antibodies labelled with horseradish peroxidase (HRP) (Sigma, Italy) and developed using ECL according to the manufacturer’s instructions. Rabbit polyclonal anti-GAPDH (ab 9485) and anti-procaspase-9 (ab 32068) were purchased from Abcam (Italy) and diluted 1:500. Rabbit anti-PARP (9542) and anti-p53 (Ser15-9286) were purchased from Cell Signaling Technology (Italy) and diluted 1:500. Anti-cleaved caspase-3 (9661), also purchased from Cell Signaling Technology, was diluted 1:1000. Rabbit anti-Bcl2 (ab 59348) and rabbit polyclonal anti-cyclin B1 (4138), purchased from Santa Cruz Biotechnology (Italy), were diluted 1:500. Rabbit anti-procaspase-8 (ab 49853) was purchased from Sigma and diluted 1:1000.

A375 cells were seeded in six-well plates at a concentration of 3 × 10⁵ cells/well in complete medium and incubated for 24 h. After 24 h, the cells were treated with mitomycin (15 µg/ml, Sigma) to prevent cell proliferation for 45 min at 37 °C. Next, the medium was removed, and a vertical scratch was created in the centre of the well using a sterile tip. The cells were subsequently washed with PBS and treated with 2 µg/ml SFN. The groove was monitored and photographed both immediately and 24 h after the scratch was created using a phase-contrast microscope (Evos, Zeiss) with 4× magnification. The migration was evaluated as the residual area of the groove, and three different fields were counted for each condition. The wound area was calculated by tracing a line along the border of the wound using ImageJ software, and the percentage of wound closure was calculated using the following equation: [Wound area (0 h) − Wound area (X h)] × 100/Wound area (0 h) = %Wound closure.

The collagen matrix was generated from bovine type I collagen at a final concentration of 1 mg/ml according to the manufacturer’s protocol. Collagen was plated immediately onto 24-well plates and then incubated at 37 °C with 95% humidity for 30 min. After collagen polymerization, cells were seeded at a density of 3 × 10⁴ cells/well and treated with 2 µg/ml SFN. After treatment, viability and invasion were evaluated by analysing and counting the cells in the supernatant, the adherent cells harvested using PBS/EDTA (5 min at 37 °C) from the upper collagen surface and the cells remaining in the collagen matrix after the adherent cells were removed. Collagen was fixed with paraformaldehyde (4%), and migrated cell nuclei were stained with blue methylene (1:10). The samples were analysed using a microscope (Olympus BH-2) to count cell numbers (n = 6 independent fields for each condition), and cell images were captured with an Olympus U-PMTVC camera. The analyses were performed in triplicate for statistical evaluation.

cDNA was synthesized using Superscript III Reverse Transcriptase (Sigma) according to the manufacturer’s instructions. qPCR was performed in 96-well plates using iQ™ SYBR Green Supermix (Bio-Rad) and an I-cycler iQ Real-Time PCR instrument (Bio-Rad). Measurements were performed in triplicate with a variability <0.5 Ct. mRNA expression was analysed by normalizing to that of the housekeeping gene GADPH. Primers were designed with PERL primer software using NCBI EntrezGene reference sequences as templates and synthesized by Sigma. The primers used are listed in Supplementary Table S4.

A375 cells were seeded at a density of 3 × 10⁵ cells/well on six-well plates and treated with SFN for 2, 6, 24 and 48 h. Total RNA was isolated using a standard TRIzol (Life Technologies) extraction protocol. Contaminating DNA was removed by Turbo DNase treatment (Life Technologies), and RNA was purified using the RNeasy Mini Kit (Qiagen). RNA quality was measured by the RNA 6000 Bioanalyzer Nano Kit (Agilent Technologies) according to the manufacturer’s instructions. Only samples with a minimum RIN score of 9.8 were used for further analyses. Double-stranded cDNA libraries were generated from 1 μg of total RNA using a TruSeq RNA Sample Preparation Kit (Illumina) following the manufacturer’s instructions. Libraries were validated using an Agilent High-Sensitivity DNA Kit on an Agilent 2100 Bioanalyzer (Agilent Technologies) and quantified by qPCR using a KAPA Library Quantification Kit (Kapa Biosystems) on a StepOne Real-Time PCR system (Applied Biosystem) at 95 °C (5 min), followed by 35 cycles of 95 °C (30 s) and 60 °C (45 s). After being quantified, the libraries were normalized to a final concentration of 10 nM. The libraries (2 nM) were pooled at final concentrations of 9 and 7 pM and loaded into a HiSeq SR flow cell v3 (Illumina) using a cBot (Illumina) according to the manufacturer’s instructions. The loaded flow cell was sequenced on a HiSeq 1500 platform (Illumina, by PoloGGB) to perform a 100 bp single-read run following the manufacturer’s instructions. The images were processed using the Illumina RTA base calling pipeline, converted in FASTQ using Casava 1.8 software and evaluated for quality using FastQC (v 0.10.1). The reads were trimmed with Trimmomatic (v 0.32) and all reads with >Q20 were selected and aligned with STAR (v 2.3.0) to hg19/GRCh37 in the UCSC Genome Browser. Bam and GTF files were used to create a matrix associating all mapped read numbers to each sample and gene using Htseq. The obtained matrix was processed with the DeSeq R package to select for all differential expressed genes (DEGs) that had a false discovery rate (FDR) <10%. Expression differences were considered significant if their p value was <0.05 and their induction (or repression) ratio was ≥1.5. All graphs were produced with R software (v 3.0.0). Functional clustering was performed with DAVID 6.7 and the Web Gene Set Analysis Toolkit (WebGestalt) for enrichment analysis of the differentially expressed genes. DEGs were screened using enrichment analysis based on the hypergeometric distribution WebGestalt algorithm.

Using the online database Search Tool for the Retrieval of Interacting Genes (STRING) v. 9.1 ([25]; http://​string-db.​org), interactions between the DEGs were predicted. The interactions include direct (physical) and indirect (functional) associations derived from four sources: genomic context, high-throughput, co-expression and prior knowledge.

All experiments were performed at least three times independently. The data are expressed as the mean ± standard deviation (SD). Data were analysed by the paired two-tailed Student’s t test [p < 0.05 (*), p < 0.01 (**) and p < 0.001 (***)]. The Dunnett’s test was used to assess the significance of the viability assay results. All statistical analyses were performed with Prism 5 Software.

To assess the effects of SFN on cell viability, primary human melanoma A375, metastatic human melanoma 501MEL and HEMa cells were treated with increasing concentrations of SFN (1–5 µg/ml) for 24 and 48 h. SFN treatment significantly reduced the viability of both A375 and 501MEL cells in a dose- and time-dependent manner (Fig. 1a), though A375 cells were more sensitive to SFN treatment than 501MEL cells. In fact, SFN began to significantly reduce the viability of 501MEL cells only at the concentration of 2 µg/ml. After 48 h, 2 µg/ml SFN decreased the viability of the A375 cells by 64% compared to only 46% for the 501MEL cells. As a control, HEMa cells were treated with the same concentrations of SFN, and only the higher dose (5 µg/ml) significantly decreased their viability. These results were also confirmed by cell proliferation analysis using the trypan blue exclusion assay (Supplement Fig. S1a) and phosphorylated AKT (p-AKT) expression analysis (Supplement Fig. S1b). Trypan blue cell counting showed significantly reduced proliferation of both A375 and 501MEL cells at 24 and 48 h post-treatment. Moreover, high levels of p-AKT in both melanoma cell lines were significantly decreased after 24 h of SFN treatment.

To determine cell death due to reduced viability, cell morphology was observed in the confluent monolayer after treatment with 2 and 5 µg/ml SFN. Specifically, the A375 cells displayed increased size, irregular shape and membrane blebbing after 48 h (Fig. 1b). Morphological alterations were also observed in the 501MEL treated cells, as they retracted into a spherical shape and formed suspended clusters. In contrast to melanoma cells, HEMa cells did not exhibit any significant morphological alterations at the 2 µg/ml dose of SFN, and only high concentrations of SFN induced rounded melanocytes, irregular morphology and membrane blebbing. Because 2 µg/ml SFN had no inhibitory effect on HEMa cells but was extremely effective in both melanoma cells lines, this concentration was used for further analysis.

To further investigate the inhibitory effects of SFN on cell viability, we analysed cell cycle progression and apoptosis by both flow cytometry and Western blot. SFN exposure changed the cell cycle phase distribution in both melanoma cell lines, and, in agreement with the cell viability data, no changes were observed in the HEMa cells (Fig. 2a). A375 and 501MEL cells treated with SFN significantly accumulated in the G₂/M phase, as up to 55 and 50% of cells were observed in this phase at 24 h post-treatment, respectively. These numbers shifted down to 40 and 45%, respectively, after 48 h (Fig. 2a). Conversely, the proportion of cells in the G₀/G₁ phase was markedly decreased in both cell lines, while the percentage of cells in the S phase remained stable.

To evaluate whether SFN induces apoptosis in human melanoma cell lines, we used several experimental approaches. The assessment of cells in the sub-G₁ phase using cell cycle analysis showed a time-dependent significant increase in the percentage of cell death after SFN exposure at 48 h in both melanoma cell lines, as up to 29 and 20% of A375 and 501MEL cells were dead, respectively (p = 0.0339 and p = 0.0023, respectively). In contrast, no significant effects were observed in the HEMa cells (Fig. 2b). To validate the pro-apoptotic effects of SFN, both the A375 and the 501MEL cells were analysed by flow cytometry using Annexin V–FITC. The percentage of total (early + late) apoptotic cells increased in a time-dependent manner. At 48 h post-treatment, the number of apoptotic cells reached 38 and 22% in the A375 and 501MEL cells, respectively (Fig. 2c). As previously shown, SFN exhibited a stronger apoptotic effect on the A375 cells than on the 501MEL cells. No effects were observed in the HEMa cells.

We further investigated the mechanisms underlying cell death induced by SFN by assessing proteins that play a crucial role in the cell cycle and apoptosis regulation. Immunoblotting on protein cell extracts revealed that SFN completely inhibited the expression of cyclin B1, a G₂/M phase marker, in A375 cells at 48 h post-treatment, which agrees with the accumulation of cells in the G₂/M phase. In the 501MEL cells, there was only a partial reduction (Fig. 2d, e). The levels of total p53 and phosphorylated p53 [p-p53 (Ser-15)] were strongly increased in a time-dependent manner in both melanoma cell lines. The levels of Bcl-2, an anti-apoptotic protein and a transcriptional target of p53, were also analysed. In accordance with p53 expression, Bcl-2 expression was decreased in both melanoma cell lines after SFN exposure (Fig. 2d, e). Moreover, to determine whether SFN-induced apoptosis in melanoma cells is caspase-dependent, caspase-8, caspase-9 and caspase-3 expression was assessed. Procaspase-9 was strongly decreased in a time-dependent manner in A375 cells after SFN treatment, suggesting a major role of the intrinsic apoptotic pathway. In contrast, no differences were detected in the 501MEL at any time point. Procaspase-8 was used to assess the extrinsic apoptotic pathway, and its levels were slightly reduced in 501MEL cells at 48 h post-treatment and slightly reduced in A375 cells at both 24 and 48 h post-treatment. Because of caspase-8 and caspase-9 activation, caspase-3 was completely cleaved in A375 cells at 48 h after SFN treatment, while caspase-3 cleavage started later and it was only partially cleaved after 48 h in 501MEL cells (Fig. 2d, e). Because caspase-3 is the most efficient processing enzyme for poly (ADP-ribose) polymerase (PARP), we also analysed its activation. SFN activated PARP cleavage in both melanoma cell lines. PARP was remarkably cleaved at 24 h in A375 cells and persisted until at least 48 h after treatment, while PARP cleavage was only detected in 501MEL cells at 48 h (Fig. 2d, e). These findings highlight the differences in the apoptotic mechanism of SFN between primary and metastatic human melanoma cells.

To evaluate the effects of SFN on melanoma cell migration, we performed a scratch assay. The data demonstrated that SFN treatment strongly reduced wound closure in A375 cells at 24 h. Only 18% of the tumour cells migrated compared to 89% for the untreated cells (p = 0.0022; Fig. 3a). The difference in the migration ability between the treated and untreated cells cannot be ascribed to a difference in their proliferation rates because the cells were treated with mitomycin prior to SFN treatment. To measure the impact of SFN on tumour cell invasion, we counted the number of cells that could penetrate type I collagen matrix. The data showed a significant reduction in the number of tumour cells invading the collagen compared to that of untreated cells (108 vs. 35.5, p = 0.0057) (Fig. 3b). All together, these results clearly suggest that SFN affects the migration and invasion ability of melanoma cells.

To decipher the antitumour molecular mechanisms driven by SFN, we investigated the whole transcriptome of A375 cells at four different time points (2, 6, 24 and 48 h after SFN treatment) using Illumina RNA-Seq technology. The results from two independent experiments revealed a total of 329 differently expressed genes (DEGs). Among these genes, 219 were affected by SFN treatment (Table 1), while 110 genes were intrinsically modulated during cell growth in culture (without compound treatment) (Supplementary Table S1). Out of the 219 genes, 74 were downregulated (34%) and 145 were upregulated (66%) by SFN at least at one time point (Fig. 4a, b; Table 1). Only two genes were upregulated at 2 h post-treatment, heat shock 70 kDa protein 1A (HSPA1A) and heme oxygenase (decycling) 1 (HMOX1). The number of DEGs modulated by SFN was much higher at 6 and 24 h after exposure compared to that at 48 h post-treatment (Fig. 4a, b; Table 1). Most genes were specifically affected at a unique time point, and only a few genes shared differential expression over several hours post-SFN treatment, specifically between 6 and 24 h (Fig. 4c).

Enrichment cluster analysis of gene functions and biological processes was carried out using the WEB-based GEne SeT AnaLysis Toolkit (WebGestalt). Gene ontology analysis based on biological processes, molecular functions and cellular components are presented in Table 2, and the list of genes in each category is presented in Supplement Table S2. Clustering based on the biological process classification revealed that SFN-regulated genes are predominantly involved in response to stress (86 genes, p = 2.43e−12), the apoptotic process/cell death (60 genes, p = 3.44e−11), response to topologically incorrect proteins (16 genes, p = 9.47e−09), response to different stimuli (129 genes, p = 9.47e−08) and positive regulation of metabolic processes (50 genes, p = 0.0001). Molecular function clustering revealed that SFN regulates genes generally involved in protein binding (135 genes, p = 6.97e−08). According to the cellular component classification, SFN induces the expression of genes involved in the intracellular ferritin pathway (2 genes, p = 0.012) and the glutamate–cysteine ligase complex (2 genes, p = 0.012).

To verify the expression patterns of the DEGs detected by RNA-Seq, 12 genes that are modulated by SFN were selected for quantitative real-time PCR analysis, and GADPH was used as the reference gene. This includes five stress response genes [(HSPA1A, HMOX1, thioredoxin reductase 1 (TXNRD1), glutamate–cysteine ligase, catalytic subunit (GCLC) and modifier subunit (GCLM))], six p53 network genes [early growth response 1 (EGR1), activating transcription factor 3 (ATF3), BCL2-associated X protein (BAX), Fas cell surface death receptor (FAS), growth arrest and DNA damage-inducible beta (GADD45B) and cyclin-dependent kinase inhibitor 1A (CDKN1A))] and integrin beta 4 (ITGB4). The qPCR results showed high concordance with the RNA-Seq data, although the fold change values varied accordingly with the analytical method, suggesting that the RNA-Seq findings are reliable (Fig. 5). However, these data support the observation that SFN regulates the transcription of genes related to the Nrf2-signaling pathway and its involvement in physiological processes in melanoma cells.

The functional partnerships and interactions that occur between proteins corresponding to the revealed DEGs were predicted by STRING network analysis [25]. The downregulated and upregulated genes were linked into the networks according to their physical and functional associations and their involvement in specific cellular pathways (Fig. 6; Supplement Table S3). Early response to the drug (2 h post-treatment) was clearly associated with overexpression of the HSPA1A and HMOX1 genes. We detected upregulation of the HMOX1 gene across all the time points except the last one (48 h), with the maximum expression being reached at 6 h after exposure. At 6 h after SFN exposure, the number of stimulated genes increased greatly (Fig. 4; Table 1). Out of 114 DEGs, 30 upregulated genes could be organized into a network according to their function: (a) response to stress; (b) protein unfolding response and response to temperature; (c) regulation of cell death. Only five interacting proteins were downregulated, all of which were receptors [coagulation factor II (thrombin) receptor (F2R), lysophosphatidic acid receptor 1 and 3 (LPAR1/3), adenosine A1 receptor (ADORA1) and peroxisome proliferator-activated receptor gamma (PPARG)]. Prolonged SFN treatment (24 h) upregulated the expression of 73 genes and downregulated the expression of 36 genes (Fig. 4; Table 1). The upregulated genes mainly included genes related to apoptosis, including p53-related genes (MDM2, BAX, GADD45A, CDKN1A, ATF3, FAS), and genes involved in the growth arrest and proliferation (EGR1, BTG2). The stress response network was still activated, including the TXNRD1 and SOD1 genes and the histone protein cluster genes HIST1H2BD and HIST1H2BK. The downregulated genes include plasminogen activator tissue (PLAT), ITGB2 and ITGB4 grouped into the network, which could be involved in the regulation of cell migration and tissue remodelling (Fig. 6). Transcription factor-specific protein 1 (SP1) has been shown to be upregulated in several type of cancers; it is associated with poor prognosis and could be implicated in prostate cancer chemoprevention [26]. Our analysis demonstrated that SFN-treated A375 cells showed reduced SP1 expression at 24 h, which orchestrated the coordinated overexpression of 16 other proteins in the network (Fig. 6). The transcriptional co-factor transducin beta-like (TBL1X) 1 has been demonstrated to be involved in cellular proliferation and invasiveness in both human and murine pancreatic ductal adenocarcinoma [27]. Our data also indicated a crucial role for this protein, as it is downregulated at 24 h by SFN and probably controls the expression of ten other genes in the network. At 48 h, SFN upregulated a small number of genes related to the final stage of stress response and p53-related genes that play a role in apoptosis, such as TP53-inducible nuclear protein 1 (TP53INP1), sestrin 2 (SESN2), Fas cell surface death receptor (FAS) and glucose-6-phosphate dehydrogenase (G6PD) (Fig. 6; Table 1).