Targeted alternative splicing of TAF4: a new strategy for cell reprogramming

Reprogramming of somatic cells has become a versatile tool for biomedical research and for regenerative medicine. In the current study, we show that manipulating alternative splicing (AS) is a highly potent strategy to produce cells for therapeutic applications. We demonstrate that silencing of hTAF4-TAFH activity of TAF4 converts human facial dermal fibroblasts to melanocyte-like (iMel) cells. iMel cells produce melanin and express microphthalmia-associated transcription factor (MITF) and its target genes at levels comparable to normal melanocytes. Reprogramming of melanoma cells by manipulation with hTAF4-TAFH activity upon TAFH RNAi enforces cell differentiation towards chondrogenic pathway, whereas ectoptic expression of TAF4 results in enhanced multipotency and neural crest-like features in melanoma cells. In both cell states, iMels and cancer cells, hTAF4-TAFH activity controls migration by supporting E- to N-cadherin switches. From our data, we conclude that targeted splicing of hTAF4-TAFH coordinates AS of other TFIID subunits, underscoring the role of TAF4 in synchronised changes of Pol II complex composition essential for efficient cellular reprogramming. Taken together, targeted AS of TAF4 provides a unique strategy for generation of iMels and recapitulating stages of melanoma progression.

that splicing events in the region encoding the co-activator hTAF4-TAFH domain control the differentiation of human neural progenitors (NHNPs) 35,36 and human adipose-derived mesenchymal stem cells (hMSCs) 34 . Targeted proteolysis of Taf4 was demonstrated to be necessary for differentiation of mouse F9 embryonic carcinoma cells 37 and myogenic differentiation of myoblasts 38 , whereas enforced expression of TAFH domain blocked differentiation of F9 cells towards early endodermal lineages 39 . Moreover, inactivation of Taf4 in mouse epidermis resulted in hyperplasia and development of aggressive melanomas in the dermis compartment 40,41 . In keratinocytes, the absence of Taf4 led to ectopic expression of melanocyte-specific melan-A and melanoma-associated antigen 9 (MAGEA9) and development of invasive melanomas 40 . Interestingly, an upstream enhancer containing binding sites for pluripotency transcription factors Oct4 and Nanog was identified in mouse Taf4 gene 42 . Most recent findings described TAF4 as one of the critical components in converting human fibroblasts into iPSCs 43 . Thus, with important functions in maintenance of TFIID stability and integrity, TAF4 represents a unique tool for manipulating the whole TFIID composition and in promoting specific cellular programs.
Here, we provide a new concept for cell reprogramming, where instead of changing the transcription regulatory networks by forced expression of lineage-specific transcription factors or use of different miRNAs, we advocate for targeted AS of the core components of RNA Pol II transcription machinery. As an example, we targeted the activity of TAF4 by TAFH-specific RNAi to examine the potential of this approach for reprogramming of human dermal fibroblasts. Data presented here allow us to conclude that targeting AS of TAF4 affects the entire TFIID complex, providing a unique model system to induce iMels and reprogram tumour cells to less or more aggressive cancer phenotypes.

Results
Differential activity patterns of hTAF4-TAFH are characteristic to dermal fibroblasts, melanocytes and melanoma cells. Previously, we have demonstrated that exons encoding the hTAF4-TAFH domain are subjected to extensive AS in hMSCs and NHNP cells 34,35 . To assess whether AS of TAF4 exons V, VI and VII encoding the hTAF4-TAFH (Fig. 1a) is prevailing in cells of neural crest (NC) origin, we examined the expression of TAF4 ASVs in facial dermal fibroblasts, melanocytes and melanomas.
Using transcript-specific RT-PCR approach, we exclusively examined the splicing of TAF4 exons encoding hTAF4-TAFH domain. TAFH_v1 mRNAs encoding intact hTAF4-TAFH were detected in all analysed cells, while melanocytes, primary melanoma tissues and cultured melanoma cells next to TAFH_v1 mRNAs expressed TAFH_v2 mRNAs encoding proteins with no hTAF4-TAFH domain at comparable to TAFH_v1 levels (Fig. 1b). In primary melanoma tumours and cultured cells, TAFH_v2 was expressed along with TAFH_v4 and TAFH_v6, which all result in isoforms with no hTAF4-TAFH activity (Fig. 1a,b).
For the detailed analysis of hTAF4-TAFH function, we elaborated RNAi approach with TAFH-specific siRNAs targeting exons V or VI of TAF4 to affect the expression of different ASVs (Fig. 1a). Upon TAFH RNAi, silencing of TAFH_v1 mRNAs both in fibroblasts and melanoma SkMel28 and WM 266-4 cells promoted the expression of transcripts encoding isoforms without hTAF4-TAFH (Fig. 1c) that was also verified by changed patterns of TAF4 isoform expression (Fig. 2e). Expression analysis data along with TAFH RNAi intrigued us to study the effects of hTAF4-TAFH activity in NC-derived cells in more detail.
Silencing of hTAF4-TAFH converts facial dermal fibroblasts to iMels. We have established that abrogation of hTAF4-TAFH activity promotes differentiation of hMSCs and NHNPs 34,35 . In the current work, hTAF4-TAFH silencing in fibroblasts by RNAi resulted in highly obvious visual colour change of cell pellets from pale to dark brown in a siRNA dose-dependent manner (Fig. 2a). This finding was in a good agreement with RT-PCR analysis data establishing that expression of TAF4 ASVs in TAFH RNAi-treated fibroblasts resembled that of melanocytes. Namely, TAFH_v1 was expressed at low and TAFH_v2 at high levels ( Fig. 1c). Upon TAFH RNAi, the induction of melanogenesis in dermal fibroblasts was further confirmed by the results of melanin content assay (Fig. 2b) and expression analysis of MITF and genes of melanin biosynthesis pathway (Fig. 2c,d). Specifically, treatment of fibroblasts with TAFH-specific siRNAs induced the concentration-dependent increase in the content of melanin relative to the control siRNA-treated cells (Fig. 2b). Conclusively, RT-qPCR analysis of TAFH RNAi-treated fibroblasts compared to the control siRNA-treated cells showed the following (Fig. 2c): a) induced expression of SNAI2 and PAX3, early transcription factors of melanocyte development; b) stimulation of expression of total MITF mRNA and melanocyte-specific alternative form MITF-M mRNA; c) increased expression of MITF target genes TYR, TYRP1 and DCT. When RT-qPCR data of TAFH RNAi-treated fibroblasts were normalised to the levels of gene expression in primary fibroblasts (arbitrarily referenced as 0%) or melanocytes (taken for 100%), it became apparent that the expression of melanocyte-specific genes in induced melanocytes (iMels) was highly similar to that in melanocytes (Fig. 2c). While immunofluorescence (IF) analysis further confirmed the nuclear expression of MITF in TAFH RNAi-treated fibroblasts similar to melanoma cells (Fig. 2d). Altogether, these data suggested that hTAF4-TAFH activity perturbs the spontaneous conversion of dermal fibroblasts to iMels.
Silencing of hTAF4-TAFH activity in fibroblasts induced the expression of TP53 mRNAs and accumulation of phospho-p53 (Fig. 2e) similar to the hMSCs 34 , indicating that the function of p53 in the process of hTAF4-TAFH-associated reprogramming is conserved across different cell types.
Altogether, these data argue that TAFH RNAi is an efficient strategy to generate iMel cells from facial dermal fibroblasts.
hTAF4-TAFH activity restricts differentiation potential of melanoma cells. Based on the data from iMel cells, we expected hTAF4-TAFH activity to control the differentiation of melanoma cells. To examine this, we compared TAFH RNAi effects in two metastatic SkMel28 and WM 266-4 melanoma cells. In both cells, TAFH siRNA treatment suppressed cell proliferation by day 5 (Fig. 3a). Moreover, RT-qPCR analysis revealed that hTAF4-TAFH activity was essential for NC-and stem cell-specific gene expression in melanoma cells (Fig. 3b).
Scientific RepoRts | 6:30852 | DOI: 10.1038/srep30852 Upon TAFH RNAi, expression of pluripotency-associated NANOG and NC-specific MSX2, PAX7, SOX10, and SNAI1 mRNAs was significantly reduced, whereas the levels of stem cell factors KLF4 and OCT4 remained steady (Fig. 3b). These results suggested that depletion of hTAF4-TAFH activity was needed to initiate differentiation of cancer cells. We tested this hypothesis by using culture conditions supporting lineage-specific differentiation. As expected, silencing of hTAF4-TAFH activity in melanoma cells promoted the initiation of mixed routes of differentiation towards chondrogenic, adipogenic and neurogenic lineages. In differentiation-supported conditions, TAFH RNAi-treated melanoma cells started to express chondrogenic-specific SOX9, NKX3.2 and RUNX2; adipogenic PPARG; and neurogenic NTRK2, NF-M and SYP mRNAs at 48 h post-stimulation of differentiation (Fig. 3b).
Our recent findings established that silencing of hTAF4-TAFH specifically supported the differentiation of hMSCs along chondrogenic lineages 34 . Considering this and given that in general the cartilaginous differentiation of melanomas is regarded to be a rare event 44,45 , we decided to analyse TAFH siRNA-treatment effects on melanoma differentiation in chondrogenesis-supporting media conditions in more detail. Immunoanalysis data revealed that at 7 days of differentiation, TAFH RNAi-treated melanomas showed high levels of expression of chondrocyte-specific SOX9 proteins (Fig. 3c) and prominent staining with Alcian blue that evidenced the synthesis of cartilage-specific ECM-proteins (Fig. 3d). All these data further argued that hTAF4-TAFH rendered TAF4 to control chondrogenic differentiation in melanoma cells.
In sum, TAFH RNAi strategy was beneficial for reprogramming of melanoma cells to recapitulate different stages of cancer progression and unmask the differentiation potential of melanomas along lineages supported by the environmental surrounding.

hTAF4-TAFH controls migration of iMels and invasiveness of melanoma cells. Given that dif-
ferentiation is accompanied by enhanced migratory activity of lineage-committed cells, we analysed the effects of TAFH RNAi on cell motility. For that, migration and invasion potential of iMels and melanoma cells was evaluated by measuring their migration activity towards 10% FBS across polycarbonate membrane. For invasion experiments, membranes were coated with ECM proteins derived from primary fibroblasts that were initially cultivated on the same transwell membranes. TAFH RNAi induced a robust (7-fold) rise in migration of iMels and also supported the invasion of melanoma cells (Fig. 4a). Consistent with these findings, both, iMels and cancer cells, showed high expression of CDH2 and matrix metallopeptidase 3 (MMP3) mRNAs and low expression of CDH1 and keratin 14 (KRT14) mRNAs at 72 h after TAFH RNAi (Fig. 4b). Western blot analysis further confirmed the induction of expression of N-cadherin (CDH2) and inhibition of E-cadherin (CDH1) upon effective hTAF4-TAFH silencing in both cell types (Fig. 4c).
The results from the migration and invasion assays established that suppression of hTAF4-TAFH activity was critical for high motility of iMels and invasiveness of melanoma cells.

Enforced expression of TAF4 supports cancer stem cell-like features in melanoma cells.
To further assess the role of hTAF4-TAFH activity in cell reprogramming, we investigated whether melanoma cells were amenable to reprogramming towards induced cancer stem cell (iCSC)-like cells by using forced expression of TAF4. Overexpression of TAF4 resulted as expected in the heightened levels of TAF4 protein (Fig. 5a) and significantly increased TAF4_v1 mRNA expression (Fig. 5b), and brought along phenotypic changes of SkMel28 melanoma towards cells with drastically reduced invasive potential (Fig. 5c). These cells had lost their ability to migrate because of the high hTAF4-TAFH activity that switched cells to predominant expression of E-cadherin (Fig. 5a). These changes were also observed at mRNA levels ( Fig. 5b). Consistently, CDH2 and MMP3 levels were reduced, and CDH1 and KRT14 mRNA levels increased in these cells (Fig. 5b). On the other hand, enforced expression of TAF4 promoted expression of pluripotency-associated KLF4, OCT4 and NANOG, and NC-related MSX2, PAX7, SOX10 and SNAI1 (Fig. 5b), reaffirming the critical role of hTAF4-TAFH activity in the maintaining of multipotency in melanoma cells.
All in all, data of melanoma cell reprogramming by enforced expression of TAF4 evidenced the fast induction of multipotent features in tumours just by altering the hTAF4-TAFH activity levels in cells.

TAFH RNAi results in changed splicing of transcripts encoding TFIID complex subunits. Finally,
we examined whether TAF4-associated cell reprogramming might affect the expression of other members of TAF complex. Firstly, we analyzed the expression of TAF4b, TAF4 paralog, in response to TAF4-TAFH siRNA treatments (Supplementary Figure S1). Western blot (Fig. S1a) and RT-PCR (Fig. S1b) data established that levels of TAF4b remained unaffected upon TAFH silencing. Next, RT-PCR analysis revealed that human fibroblasts and melanoma cells, in general, expressed a wide variation of ASVs encoding TFIID complex subunit proteins ( Fig. 6A and data not shown). Transcript-specific RT-PCR analysis showed that many of these had retained ORFs and thus the potential to code different isoforms ( Fig. 6A and Supplementary Figure S2). In fibroblasts, silencing of hTAF4-TAFH activity led to the reduced expression of distinct ORF-retaining TAF2, TAF6 and TAF12 ASVs. In melanomas, TAFH RNAi induced the differential expression of TBP, TAF1, TAF2, TAF6, TAF10, and TAF12 ASVs. More specifically, TAFH RNAi in melanoma cells led to the high levels of TAF6_v3 encoding the protein isoform with altered histone fold domains (HFD) similar to the apoptosis-dependent form of TAF6, namely TAF6δ 46 ( Fig. 6a and Supplementary Figure S2). Furthermore, in SkMel28 melanoma cells, TAFH RNAi promoted the expression of TBP_v2, which encodes protein isoforms with no poly-glutamine Armadillo-like helical region that is functionally important for hetero-oligomeric interactions 47 (Fig. 6a and Supplementary Figure S2). Splicing of TAF1, the largest and multifunctional TFIID subunit, was also affected by perturbations in TAFH structure, introducing alterations into its protein kinase and histone acetyltransferase domains ( Fig. 6a and Supplementary Figure S2). Altogether, these data argued for the tight control of TFIID composition and function at the level of AS of TAF4. However, yet more data is needed to draw more definite conclusions.
In whole, our data demonstrated for the first time that subunits of TFIID complex were targets of extensive cell-specific AS. Changes in hTAF4-TAFH activity affected the expression of TFIID complex components as a part of the molecular control of cellular reprogramming and cancer development.

Discussion
This study concludes that splicing patterns retaining high levels of hTAF4-TAFH activity lead to cell phenotypes with high multipotency but low capacity to migrate, while splicing events resulting in low hTAF4-TAFH activity enforce cells to migrate and differentiate (Fig. 6B). Simultaneous presence of TAF4 isoforms with high and low hTAF4-TAFH activities in normal cells and in cancer suggests that dominance of any of these forms is likely to have dramatic effects upon development and in tumour progression.
Current cell separation and cultivation methods have been limiting the use of melanocytes in clinics 48 . Generation of functional melanocytes from a highly accessible cell source such as dermal fibroblasts by cell reprogramming method as described herein would establish an easily scalable supply of autologous melanocytes, which could be beneficial for developing cell-based treatments of congenital pigment disorders, or in the studies of the etiopathogenesis of melanomas. Previous data have substantiated that reprogramming of dermal fibroblasts into functional melanocytes implies the use of certain combinations of exogenously added factors including transcription factors MITF, SOX10 and PAX3 16 . Instead of changing the transcriptional networks by enforced expression of specific transcription factors or microRNAs, this study provides a new concept for manipulating cell fate by targeted AS of general transcription machinery components. Here we show that targeted splicing of exons encoding hTAF4-TAFH domain alters TAF4 activity and converts fibroblasts directly into iMel cells. This is the first study to demonstrate that TAFH RNAi approach is an effective tool for generating the iMel cells with characteristics of normal melanocytes. Moreover, our findings establish a link between hTAF4-TAFH activity and activation of p53 to accompany with effective reprogramming of fibroblasts into iMels. Involvement of p53 in tanning response 49 and hyperpigmentation of epidermis has been described 50 . Simultaneous activity of MITF and p53 in pigmented skin has suggested a model of crosstalk between these two cellular pathways as necessary for protection of skin against UV harms 49 . Thus, TAF4-p53 interactions may be crucial for effective reprogramming of cells. However, this hypothesis remains yet to be tested. (b) RT-qPCR analysis of pluripotency, NC (neural crest) and migration-associated genes upon enforced expression of TAF4 in SkMel28 cells showed that high levels of TAF4 expression support multipotency and inhibit invasion. Data are normalised to GAPDH expression relatively to control plasmid-treated cells and represented in LN scale. At least three independent experiments were performed for each analysis and represented as mean± SD with * * * p < 0.001, and * * p < 0.01. (c) Cell invasion assays showed that high levels of TAF4 suppress the invasion potential of SkMel28 cells. The number of invaded cells was determined by counting from five independent microscopic fields, normalised to cell numbers that were transfected with control vector (Contr) and represented as mean± SD with * * * p< 0.001 determined by Student t-test. Recent studies along with ours have demonstrated that differentiated cells do not express canonical TAF4 protein at high levels 30,38,51 . Herein we show that malignantly transformed cells, melanoma in particular, represent a unique model where both, canonical and alternative TAF4 isoforms, are expressed at comparable levels providing an attractive context for manipulating cancer cells in vitro. Similar to hMSCs 34 , neural progenitor cells 35 , and facial dermal fibroblasts, silencing of hTAF4-TAFH activity in melanoma promotes these cells to differentiate. Melanoma is a highly divergent neoplasia, with prominent genetic and clonal heterogeneity 52,53 . The phenomenon of reversal differentiation is well described for NC-derived lineages 54,55 . In the current study, we have demonstrated the critical role of hTAF4-TAFH in driving the fibroblast-like (mesenchymal) features in melanoma cells, as silencing of hTAF4-TAFH activity promoted melanoma cells to differentiate along mesenchymal lineages. In addition, enforced expression of hTAF4-TAFH revealed that melanoma cells avidly acquire phenotypes of highly plastic NC stem cell-like cells 56 and characteristics of iCSCs 57,58 , supporting the expression of pluripotency KLF4, NANOG and OCT4 and genes associated with multipotent NC phenotype. It is highly conceivable that similar changes in AS of TAF4 with consequences to reprogramming of cancer cells can spontaneously occur in vivo. In sum, our study is the first to pinpoint to the role of hTAF4-TAFH activity of TAF4 as one of the mechanisms in driving heterogeneity in melanoma, where cells with low hTAF4-TAFH activity differentiate along different lineages and cells with high activity of hTAF4-TAFH retain multipotency.
Our current data suggest that integrity of hTAF4-TAFH domain is critical for cell migration. We established that changes in the activity of hTAF4-TAFH can spontaneously convert dermal fibroblasts into highly motile iMels and support invasion of melanoma cells. Low levels of hTAF4-TAFH activity induce E-to N-cadherin switches in these cells. Changes in the expression of these ECM proteins are a characteristic of epithelial-to-mesenchymal transition (EMT), which is a crucial process in development and in cancer progression 59 . Current findings provide a new understanding of how one of the core processes of EMT is under the control of hTAF4-TAFH activity.
By analysing the expression of ASVs of different subunits of TFIID in response to TAFH RNAi, we found that cell-specific AS is highly common for components of TFIID complex. The majority of TAF subunits are expressed in tissue-and development-specific manner. The functions of TAF4, TAF8 and TAF10 have been found to be critical for maintaining pluripotency 43 , whereas TAF2, TAF7 and TAF12 have come across as ESC determinants 60 . Silencing of TAF3 and TAF5, on the other hand, results in loss of OCT4 function 61,62 . Differential expression of TAF subunits drives lineage-specific differentiation 29 . For example, TAF3 is predominantly expressed in myocytes 38 ; expression of TAF1, TAF3, TAF4 and TAF9 is absent in liver 63 ; TAF8 is essential for adipogenic but not for myogenic differentiation 64 . Our data extend these findings by showing that TAF isoforms generated by cell-specific AS control the composition and activity of cellular TFIID complexes through dynamic regulation of their tissue-and development-specific functions. Flexible arrangement of TFIID complexes in different cellular contexts offers an efficient means to control cell fate by targeted AS of a single component of TFIID, such as established by us TAF4 subunit. Previous studies have shown that TAF4b can compensate for the loss of TAF4 at early stages of embryogenesis 65 , consistent with the earlier findings that both proteins retain specific and non-redundant functions 66 . Thereby, although TAF4 and TAF4b have at some extent overlapping roles, part of their functions is specific. As TAF4-TAFH-dependent mechanism of cell reprogramming is clearly not associated with the compensatory functions of TAF4b, this suggests that alternative splicing of TAF4 leads to TAFH-less protein isoforms that lack the ability to substitute for canonical TAF4 activity in RNA PolII-dependent transcription, but behave as dynamic fine-tuners of specific cell differentiation programs dictated by the surrounding.
In conclusion, a more detailed understanding is required to clarify the role of TAF4 in the sequence of events driving the generation of iMels and cancer progression. However, it is quite clear from our studies that targeted AS of TAF4 could be used as a new way for cell reprogramming with advantages in cell-based therapies.

Materials and Methods
Ethics Statement. Use of human biological materials for the study was pre-approved by the local ethical committee at the National Institute for Health Development, Tallinn, Estonia (Approval No 2234 from Dec 09, 2010). All experiments were performed in accordance with relevant guidelines and regulations. Written informed consent was obtained from all participants prior to the study.
Cell culture. Human primary dermal fibroblasts were freshly isolated from the facial eye-lid skin of women aged between 30 to 50 years 67 , expressed typical for fibroblasts specific cell-surface markers CD73, CD90, CD105, and were negative of CD45 by analysis of flow cytometry as previously described 68 . All fibroblasts expressed NC markers including SLUG/SNAIL, SOX9, SOX2, PAX3, PAX7 and MSX1 (Supplementary Table S3). The cells were propagated in the medium of low glucose DMEM with glutamine (DMDM-LG) (PAA Laboratories, Austria) supplemented with 1% penicillin/streptomycin (PAA Laboratories, Austria) and 10% heat-inactivated foetal bovine serum (FBS) (PAA Laboratories, Austria) in a humidified atmosphere at 37 °C and 5% CO 2 . Throughout the study, fibroblasts from low (< 7) number passages were used in the functional assays. WM266-4 and SkMel28 human melanoma cell lines were obtained from ATCC and cultured in DMEM high glucose medium with glutamine (DMEM-HG)(PAA Laboratories, Austria) supplemented with 10% FBS and 1% penicillin/streptomycin.

RNA isolation, RT-PCR and Real-Time PCR.
Human primary melanoma and normal tissue RNA panels were purchased from BioChain Institute (USA). For cultured cells, total RNA was purified using TRIzol ® reagent (Invitrogen, Life Technologies, UK) following manufacturer's recommendations. cDNA was synthesised from DNase-treated (Ambion, UK) RNA using Superscript III RT (Invitrogen, UK) and a mix of oligodT and random hexamers, according to manufacturer's recommendations. RT-PCR was carried out using HOT FIREpol ® Master Mix (Solis Biodyne, Estonia). RT-qPCR was performed in triplicates using LightCycler ® SYBR Green I Master Mix (Roche Applied Science) and the LightCycler ® 480 Real-Time PCR System (Roche Applied Science). The fold of change was calculated relatively to control siRNA treatments after normalization to GAPDH expression, using 2 −ΔΔCt method, where Δ Ct is (gene of interest Ct) − (GAPDH Ct), and Δ Δ Ct is (Δ Ct treated)-(Δ Ct control). Primer sequences are listed in Supplementary Table S4. Immunofluorescence and immunoblot analyses. TAFH siRNA-treated and untreated human dermal primary fibroblasts and melanoma cells were grown on 22-mm 2 glass slides to about 70% confluence. Anti-MITF (ab59232, Abcam) as primary, and Alexa Fluor 546 (Molecular Probes, Invitrogen, Eugene, OR, USA) as secondary antibodies were used for immunofluorescence analysis. Images were visualised using Nikon Eclipse 80i fluorescence microscope (Nikon Instruments Inc., USA).
Melanin content assay. Melanin content was determined according to the modified method of Hosoi et al. 69 .
Human dermal fibroblasts were cultured and subjected to mock or siRNA treatments. Melanoma SkMel28 cells were treated with 1 mM dibutyryl-cAMP for 6 days as a control for melanin synthesis. After 24 h, cells were washed by PBS, harvested by trypsinisation and counted. The cell pellet was solubilised in 200 μ l of 0.2 M NaOH containing 10% DMSO per 1 × 10 6 cells at 80 °C for 1 h. The absorbance of each well was measured at 405 nm using SPECTRAmax 340 PC spectrophotometer (Molecular Devises, USA) and the values were normalised to the total amount of protein.
Proliferation assay. WM266-4 and SkMel28 melanoma cells were grown in 96-well flat bottom tissue culture plates up to 80% confluence and treated with control or TAFH siRNAs. WST-1 analysis (Roche Applied Science) was performed according to manufacturer's instructions. Absorption at 450 nm was measured using SPECTRAmax 340 PC Microplate Reader (Molecular Devises LLC, USA). The data were analysed using Softmax Pro 3.12 software and normalised to the values at the day of plating. Cell viability was evaluated every 24 h, up to 5 days post-treatment.

Differentiation.
For melanoma differentiation studies, 24 h after siRNA transfection, cells were cultured in conditions supporting adipogenic, chondrogenic or neuronal differentiation as previously described 34,35 . The extent of chondrogenic, adipogenic and neuronal differentiation was analysed by RT-qPCR of lineage-specific marker genes. To confirm effective chondrogenic differentiation, immunofluorescence detection by chondrocyte-specific SOX9 (AB5535, Millipore) was performed 7 days after culturing of siRNA-treated melanoma cells in chondrogenic differentiation-supporting conditions. Alcian blue staining (Sigma) was performed 2 weeks after differentiation stimulation. Cells were fixed in 4% PFA, washed with distilled water, stained with Alcian blue and analysed under 10× magnification for glycosaminoglycan expression.
Cell migration and invasion assays. Cell migration and invasion were analysed by in vitro migration assays, using 8 μ m-pore polycarbonate transwell filters (QCM TM Colorimetric Cell Migration Assay, Millipore, USA). For melanoma invasion assay, transwell filters were coated with extracellular matrix components (ECM) extracted from human fibroblasts 70 . For that, fibroblasts were plated onto transwell filters, cultivated until the confluence, and ECM were extracted by using cell lysis buffer (10 mM TRIS-HCl, 1 mM EDTA, at pH7.4) for 16 h. For migration assay, transwell filters were left uncoated. For migration/invasion studies, fibroblasts or melanoma cells were treated with TAFH siRNAs for 72 h and thereafter starved in serum-free medium for 16 h. Cells were seeded at density of 1 × 10 5 cells per insert in DMEM-LG medium without the addition of FBS, whereas the bottom chamber was filled with DMEM-LG supplemented with 10% FBS. The total number of cells that migrated to the lower chamber was quantified after 24 h of incubation of cells at 37 °C with 5% CO 2 . Migrated cells on the lower surface of the filter membrane were fixed and stained by cell stain solution with 0.1% crystal violet before counting. The number of migrated and invaded cells per well was counted in five randomly selected microscope fields in three independent experiments using Nikon Eclipse 80i fluorescence microscope (Nikon Instruments Inc., USA).
Scientific RepoRts | 6:30852 | DOI: 10.1038/srep30852 Statistical analysis. At least three independent replicates were performed for each experiment. Statistical analysis was performed using an unpaired Student's t-test with a 2-tailed p value. Differences were considered significant when the p < 0.05.