Plasticity of Drug-Naïve and Vemurafenib- or Trametinib-Resistant Melanoma Cells in Execution of Differentiation/Pigmentation Program

Melanoma plasticity creates a plethora of opportunities for cancer cells to escape treatment. Thus, therapies must target all cancer cell subpopulations bearing the potential to contribute to disease. The role of the differentiation/pigmentation program in intrinsic and acquired drug resistance is largely uncharacterized. MITF level and expression of MITF-dependent pigmentation-related genes, MLANA, PMEL, TYR, and DCT, in drug-naïve and vemurafenib- or trametinib-treated patient-derived melanoma cell lines and their drug-resistant counterparts were analysed and referred to genomic alterations. Variability in execution of pigmentation/differentiation program was detected in patient-derived melanoma cell lines. Acute treatment with vemurafenib or trametinib enhanced expression of pigmentation-related genes in MITF-Mhigh melanoma cells, partially as the consequence of transcriptional reprograming. During development of resistance, changes in pigmentation program were not unidirectional, but also not universal as expression of different pigmentation-related genes was diversely affected. In selected resistant cell lines, differentiation/pigmentation was promoted and might be considered as one of drug-tolerant phenotypes. In other resistant lines, dedifferentiation was induced. Upon drug withdrawal (“drug holiday”), the dedifferentiation process in resistant cells either was enhanced but reversed by drug reexposure suggesting involvement of epigenetic mechanisms or was irreversible. The irreversible dedifferentiation might be connected with homozygous loss-of-function mutation in MC1R, as MC1RR151C  +/+ variant was found exclusively in drug-naïve MITF-Mlow dedifferentiated cells and drug-resistant cells derived from MITFhigh/MC1RWT cells undergoing irreversible dedifferentiation. MC1RR151C  +/+ variant might be further investigated as a parameter potentially impacting melanoma patient stratification and aiding in treatment decision.


Introduction
Targeted therapies brought hope for melanoma patients; however, the initial clinical response is not achieved in every patient and the development of drug resistance is observed in the majority of responders within one year. Heterogeneity and plasticity of melanoma cells are well recognized as causative factors of resistance [1][2][3][4]. The possible switch between diverse phenotypic states creates a plethora of opportunities for melanoma cells to escape the treatment [4]. In this respect, the role of the differentiation program in intrinsic and acquired resistance to targeted drugs is not sufficiently elaborated. 2 Journal of Oncology suggesting that targeted therapies enhance expression of MITF-M and melanosomal genes [9,10], which lead to increased pigmentation [11,12]. MITF-M is the major transcription factor that regulates the phenotype of both melanocytes and melanoma cells [13][14][15], and it plays an important prosurvival role in melanoma cells [16]. The role of MITF-M in the development of resistance is controversial. It has been demonstrated that downregulation of MITF enhances effects of targeted therapeutics and reduces the acquisition of resistance [17][18][19][20], but other reports have shown that low MITF predicts early resistance to targeted drugs [21], and the acquisition of resistance is accompanied with dedifferentiation and markedly reduced MITF level [7]. These discrepancies can be partially explained by high intratumour heterogeneity and coexistence of MITF high melanoma cells and MITF low melanoma cells expressing the AXL kinase at a high level [22][23][24]. Since these two subpopulations are present in the tumour in different proportions as the result of genetic and epigenetic mechanisms, but also therapeutic insult or microenvironmental stimuli, either MITF high differentiated phenotype or AXL high invasive phenotype might dominate and be detected at the bulk tumour level.
Using our preclinical model of patient-derived melanoma cells cultured in stem cell medium (SCM), we investigated effects of targeted drugs, vemurafenib and trametinib, on MITF level and expression of MITF-dependent pigmentation/differentiation genes. The study included MLANA and PMEL encoding transmembrane proteins, Melan-A/MART-1 (melanoma antigen recognized by T cells 1) and PMEL17 (premelanosome protein 17/gp100; HMB45), both proteins functioning in stage I/II of melanosomal differentiation, and two genes, TYR and DCT encoding enzymes active in stage III/IV of melanin synthesis, tyrosinase, and DOPAchrome tautomerase/TYRP2, respectively. Choosing SCM as the microenvironment for melanoma cells was crucial, as we have shown previously with transcriptomic analysis that serum present in the medium drastically reduces expression of MITF-M and 74 MITF-dependent genes, including TYR, DCT, and MLANA [21]. Moreover, SCM better preserves the original melanoma cell characteristics than serumcontaining medium [25][26][27][28].

Materials and Methods
. . Drug. Vemurafenib and trametinib were purchased from Selleck Chemicals LLC (Houston, TX, USA).

. . Ethical Approval, Melanoma Cell Line Generation, and
Culture. The study was approved by Ethical Commission of Medical University of Lodz. Each patient signed an informed consent before tissue acquisition. All research was performed in accordance with relevant guidelines and regulations. Melanoma cell populations from drug-naïve patients were investigated. Cell lines were named DMBC11, DMBC12, DMBC17, DMBC21, DMBC28, DMBC29, and DMBC33 (Department of Molecular Biology of Cancer, DMBC). Tumour tissues were processed immediately after surgical procurement and suspensions of melanoma cells for culturing were generated within 2 h. After several washes, tumour fragments were minced with scissors and incubated in HBSS (Sigma Aldrich, St Louis, MO, USA) supplemented with 3 mM calcium chloride and 1 mg/mL collagenase IV for 2-3 h at 37 ∘ C. DNase I (10 g/mL) was added and cells were filtered through a 70 m pore size filter. Cells were cultured in complete medium (RPMI-1640 with 10% FBS) for 1 day to remove dead and nonadherent cells. They were maintained in serum-free stem cell medium (SCM), consisting of DMEM/F12 low osmolality medium (Lonza, Basel, Switzerland), B-27 supplement (Gibco, Paisley, UK), insulin (10 g/mL), heparin (1 ng/ml), 10 ng/mL bFGF (basic fibroblast growth factor), 20 ng/mL EGF (epidermal growth factor) (BD Biosciences, San Jose, CA, USA), and antibiotics (100 IU/mL penicillin, 100 g/mL streptomycin) as described previously [29].
For experiments, cells were treated with 5 M vemurafenib or 50 nM trametinib and then collected for RNA isolation (after 22 h), protein lysates (after 24/48 h), and immunophenotype analysis (after 44 h).
To generate cells resistant to vemurafenib or trametinib, melanoma cells were cultured for 4-5 months with increasing concentrations of drugs, from 1 M to 10 M and from 1 nM to 50 nM, respectively. For "drug holiday" experiments, drugs were removed for 10 days.
. . Acid Phosphatase Activity (APA) Assay. Cells were plated at a density of 3.2-4 × 10 3 /well in 96-well plates and cultured in 100 l of culture medium containing 5 M vemurafenib or 50 nM trametinib. To assess the number of viable cells, the activity of acid phosphatase was measured. After indicated time intervals, the medium was replaced with 100 l of buffer containing 0.1 M sodium acetate (pH = 5), 0.1% Triton X-100, and 5 mM p-nitrophenyl phosphate, and plates were incubated for 2 hours at 37 ∘ C. The reaction was stopped by adding 10 l of 1 M NaOH and the absorbance values were measured at 405 nm using a microplate reader (Tecan Group Ltd., Salzburg, Austria). . . Cell Lysate Preparation and Western Blotting. Melanoma cells were lysed for 30 min at 4 ∘ C in RIPA buffer consisting of 50 mM Tris-HCl pH 8.0, 150 mM NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS, and freshly added MS-SAFE protease and phosphatase inhibitor cocktail (Sigma-Aldrich, St. Louis, MO, USA). Protein concentration was determined by Bradford assay (BioRad, Hercules, CA, USA). The lysates were diluted in 2x Laemmli sample buffer (125 mM Tris-HCl pH 6.8, 0.004% bromophenol blue, 20% glycerol, 4% SDS, and 10% -mercaptoethanol). Samples (15 g of proteins) were loaded on 7% SDS-polyacrylamide gel followed by electrophoresis at constant voltage 25 V/cm. GAPDH or -actin was used as loading control. The proteins were transferred onto Immobilon-P PVDF membrane (Millipore, Billerica, MA, USA) using BioRad transfer system. The membrane was blocked either in 5% non-fat milk or in phosphoBLOCKER (Cell Biolabs, San Diego, CA, USA) in PBS containing 0.05% Tween-20 (Sigma-Aldrich) for 1 hour. Primary antibodies detecting PARP, SOX10, DCT, GAPDH (Santa Cruz Biotechnology, Santa Cruz, CA, USA), phospho-ERK1/2 (Thr 202 /Tyr 204 ), ERK1/2, MITF (Cell Signaling, Danvers, MA, USA), or -actin (Sigma-Aldrich) were used followed by binding of the secondary HRP-conjugated antimouse or anti-rabbit antibodies (Santa Cruz Biotechnology). The membrane was incubated with Pierce5 ECL Western Blotting Substrate (Pierce, Rockford, IL, USA) for 1 min and the proteins were visualized on a medical X-ray film (Foton-Bis, Bydgoszcz, Poland) or by using ChemiDoc Imaging System (Biorad).
. . Flow Cytometry. To exclude dead cells from the analysis, LIVE/DEAD5 Fixable Violet Dead Cell Stain Kit (Life Technologies, Eugene, OR, USA) was used. Cells were fixed with 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 and stained with Anti-Melan-A primary antibody (Dako, Glostrup, Denmark) and FITC-conjugated secondary antibodies (BD Biosciences, San Jose, CA, USA), and then Alexa Fluor 647-conjugated Ki-67 antibodies (BD Biosciences, San Jose, CA, USA). For MITF staining, Alexa488-conjugated antibody (Abcam, Cambridge, Great Britain) was used. Appropriate isotype controls were included in each experiment. Flow cytometric data were acquired with FACSVerse (BD Biosciences, San Jose, CA, USA) and analysed using BD FACSuite.
. . DNA Extraction, Whole-Exome Sequencing (WES), and WES Data Analysis. DNA was isolated from 10 6 melanoma cells using a DNeasy Blood & Tissue kit (Qiagen, Hilden, Germany). Further steps were performed by Macrogen (Geumcheon-gu, Seoul, Korea). In brief, DNA samples were quantified using Picogreen (Invitrogen Thermo Fisher Scientific, Carlsbad, CA, USA) and resolved by 1% agarose gel electrophoresis (30 min, 100 V) to confirm the presence of high molecular weight fragments. DNA samples were then prepared according to an Agilent Sure-Select Human All Exome V6 kit (Agilent Technologies) which is a solution-based system utilizing ultra-long 120mer biotinylated cRNA baits to capture regions of interest. Targeted regions were selected using magnetic streptavidin beads, amplified, and loaded on the sequencer. The libraries were sequenced with Illumina HiSeq 4000 System (Illumina). Bcl files were converted to FastQ data immediately after the run. Raw data are publicly available under the accession numbers E-MTAB-6978 (drug-naïve melanomas) and E-MTAB-7248 (drug-resistant melanomas) at ArrayExpress. Data were mapped to the reference genome GRCh37/hg19 using BWA package (version bwa-0.7.12). VCF files were generated to identify somatic single nucleotide variants and short insertions or deletions (indels). Functional effects of single nucleotide polymorphisms were predicted in silico by the Polyphen-2 software available online (genetics.bwh.harvard.edu/pph2/index.shtml). Polyphen-2based predictions were classified as benign (scores 0.000-0.449), possibly damaging (scores 0.450-0.959) or probably damaging (scores 0.960-1.000).
. . Statistical Analysis. Graphs represent mean ± SD of three biological replicates, unless otherwise noted. Figure 2(b) shows mean results of three technical repeats from one typical experiment. Student's t-test was used to determine significant differences between the mean values of the tested parameters. The difference was considered significant if p < 0.05.
Expression of MITF-M was previously compared between all V600E BRAF patient-derived cell lines at the transcript and protein levels [5]. Both MITF-M high (DMBC21, DMBC28, DMBC29, and DMBC33) and MITF-M low (DMBC11 and DMBC12) cell lines were identified. In the present study, expression of four genes, MLANA, PMEL, TYR, and DCT, encoding structural/enzymatic proteins crucial for melanosomal differentiation was assessed by qRT-PCR (Figure 1(c)). Two melanoma cell lines, DMBC17 and DMBC33, expressed these genes at high levels relative to the median expression estimated for all seven cell lines. In two cell lines, DMBC21 and DMBC29 transcript levels were close to the median values, in DMBC28 markedly lower, whereas, in MITF low cell lines, DMBC11 and DMBC12 were almost undetectable. As MITF cannot induce the expression of several pigmentation-related genes in the absence of SOX10, we checked its expression. SOX10 protein level was even higher in DMBC11 and DMBC12 cell lines than in lines with enhanced expression of pigmentation-related genes (Figure 1(a)).

. . Mutation Status of Differentiation/Pigmentation-Related
Genes in Patient-Derived Melanoma Cell Lines. We did not find any SNPs and indels in MITF. Extended analysis of mutations in genes encoding transcription factors regulating MITF expression (Suppl . Table S2) did not reveal consistent explanation for variability in MITF level in melanoma cell lines. Focusing on upstream regulators of MITF expression and melanogenesis, several variants of MC R were found (Suppl .  Table S3). Notably, only DMBC11 and DMBC12 cell lines harboured homozygous MC1R R151C alteration (rs1805007), which might partially explain low expression of MITF-M and melanogenesis-related genes.
A probably damaging R419Q substitution in OCA2 (rs1800407), a protein involved in tyrosine transport, was also found exclusively in DMBC11 and DMBC12 cells (Suppl . Table S3). In addition, K198N substitution in EDN1 (endothelin-1; rs5370) was present although genes encoding EDN1 receptors, EDNRA and EDNRB, were not mutated. A homozygous probably damaging variant of TYR (rs1126809) was found in DMBC11, DMBC12, and DMBC17 cells. PLX  -PLX  -PLX  -PLX  -PLX  -PLX  TRA  TRA  TRA  TRA  TRA  TRA   GAPDH   DMBC11  DMBC12  DMBC21  DMBC28  DMBC29  . . e Influence of Vemurafenib and Trametinib on Expression of Pigmentation-Associated Genes. Six melanoma cell lines with diverse differentiation gene expression signature were selected to monitor changes in MITF-M level and expression of pigmentation-associated genes after treatment with vemurafenib or trametinib. To measure drug efficacy, phosphorylation of ERK1/2 and MEK1/2 was assessed. As expected, p-ERK1/2 and p-MEK1/2 levels were substantially reduced or even eradicated after 48 h of treatment (Figure 2(a)). This was accompanied with drug-induced changes in the viable cell number reflected by decreased activity of acid phosphatase relative to its activity in control cells (Figure 2(b)).
MITF-M transcript levels were not significantly changed, except for DMBC17 cells treated with trametinib, whereas protein levels were slightly increased in MITF-M high cells treated with vemurafenib or trametinib (Figures 3(a) and 3(b)). The percentages of MITF-positive cells were not substantial changed by the treatment as assessed by flow cytometry (Figure 3(c)).
Next, we investigated the influence of vemurafenib and trametinib on expression of pigmentation-associated genes. The acute response to vemurafenib and trametinib was not uniform. In a panel of MITF-M high melanoma cell lines, these genes were transcriptionally upregulated, except for DMBC17 cells (BRAF WT ) treated with vemurafenib ( Figure 4(a)).
In MITF-M low cell lines, DMBC11 and DMBC12, transcript levels of pigmentation-associated genes were not substantially enhanced, except for DCT expression. However, the original DCT expression in DMBC11 and DMBC12 cells was almost undetectable (Figure 1(c)); therefore, any alteration in its expression could generate a high fold change. Indeed, when expression levels of pigmentation-associated genes were compared with the median values obtained for all cell lines treated with drugs, transcript levels of DCT but also other pigmentation-related genes were still much lower in DMBC11 and DMBC12 cells than in MITF-M high cells (Figure 4(a), right). Very low and enhanced levels of DCT expression in drug-treated MITF-M low and MITF-M high melanoma cells, respectively, were also confirmed at the protein level by immunoblotting (Figure 4 -PLX  -PLX  -PLX  -PLX  -PLX  -PLX TRA  TRA  TRA  TRA  TRA  TRA   SOX10   DMBC11 DMBC12 DMBC21 DMBC28 DMBC29  unaffected, thus increasing their percentages. However, expression of pigmentation-related genes was already significantly enhanced after 22 h of treatment (Figure 4(a)), when cell viability in drug-treated cultures was similar to that in control cultures (Figure 2(b)). Moreover, when incubation with drugs was prolonged to 44 h causing reduction in viability (Figure 2(b)), enhancement of gene expression was less pronounced (Figure 4(d)). Altogether, it indicates that transcriptional reprograming causing changes in expression of pigmentation-related genes was induced as an early event.
. . Pigmentation-Related Program in Vemurafenib-and Trametinib-Resistant Melanoma Cells. We have also used our preclinical model of cultured patient-derived melanoma cells  Journal of Oncology to study drug-induced long-term changes in pigmentationrelated program by recapitulating the clinical scenario of melanoma resistance to targeted therapies. About four to five months was necessary for melanoma cells to develop resistance to lethal drug concentrations. Four cell lines resistant to trametinib 17 TRAR, 21 TRAR, 28 TRAR, and 29 TRAR and three lines resistant to vemurafenib 21 PLXR, 28 PLXR, and 29 PLXR were derived from their drug-naïve counterparts. First, expression of MITF-M and pigmentation-related genes was compared in these isogenically matched pairs of sensitive and resistant cell lines (Figures 5 and 6). The outcome was more complex than expected. Two groups of drug-resistant cell lines could be distinguished. In the first group, 21 TRAR, 28  Analysis of changes in MITF expression using a publicly available microarray data set (Gene Expression Omnibus (GEO)) revealed that the majority of relapsed tumours showed altered expression of MITF ( Figure 5(d)), and the extent of MITF increase or loss was very diverse, similarly as in our study.
To Expression of pigmentation-related genes in resistant cell lines was either reduced or enhanced, which reflected alterations in MITF levels ( Figure 6(a)). This, however, was not the case in 29 PLXR cell line, in which expression of pigmentation-related genes (Figure 6(a)) and percentages of Melan-A-positive cells (Figure 6(b)) were markedly reduced even if MITF-M expression was kept high. Lower DCT transcript level in 29 PLXR (Figure 6(a)), confirmed at the protein level (Figure 6(c)), could be partially explained by acquired heterozygous disruptive in-frame insertion in DCT (Suppl . Table S5).
While differentiation was either enhanced or repressed, the percentages of Ki-67-positive cells in resistant cell lines were similar to those in their original counterparts ( Figure 6(d)), indicating that proliferation of resistant cells became drug-independent.

. . Pigmentation-Related Program in Drug-Resistant Cells a er Drug Discontinuation.
It has been demonstrated that melanoma cells with acquired resistance to targeted therapeutics can develop drug dependency. Unexpectedly, in our experiments drug discontinuation for 10 days did not cause massive cell death (Figure 7(a)) indicating lack of drug addiction.
Next, we asked the question whether a pigmentationrelated program is modified in drug-resistant cells after drug withdrawal, during "drug holiday." Expression of MITF-M at the protein (Figure 7(a)) and transcript levels (Figure 7(b)) was reduced in comparison to that in resistant cells prior drug cessation or was kept undetectable. Expression of pigmentation-related genes was also reduced in the majority of resistant cell lines subjected to drug discontinuation (Figure 7(b)). Interestingly, MITF-M transcript and protein levels in 29 PLXR on-drug-holiday cells were similar/enhanced in comparison to those in resistant cells and drug-naïve cells, whereas MITF-M-dependent expression of MLANA was undetectable in 29 PLXR cells, both resistant and ondrug-holiday (Figure 7(b)). In two resistant cell lines with highly upregulated expression of pigmentation-related genes, 17 TRAR and 29 TRAR, trametinib withdrawal resulted in marked reduction of mRNA levels, almost to the levels assessed in the drug-naïve cells, and as shown for DCT and TYR transcripts even below these levels (Figure 7(b)).
Alterations following drug discontinuation were also diverse at the cell population level (Figures 7(c)-7(d)). Percentages of MITF-positive cells, substantially reduced in resistant cell lines, were kept either low (21 TRAR) or undetectable (28 PLXR) during "drug holiday" or were substantially reduced (28 TRAR, 29 TRAR, 17 TRAR, 21 PLXR) (Figure 7(c)). And again, the percentage of MITF-positive cells in 29 PLXR after vemurafenib withdrawal markedly increased (Figure 7(c)), which was not, however, accompanied with an increase in the percentage of Melan-A-positive cells (Figure 7(d)). Except for 17 TRAR, the percentages of Melan-A-positive cells were very low after drug discontinuation (Figure 7(d)), which is in agreement with reduced expression of MLANA in these conditions (Figure 7(b)).

Discussion
Melanoma cell plasticity is evident and as one of the main causes of low efficacy of treatment and development of drug resistance is in the focus of current research. Several programs that are executed in melanoma cells can be affected by targeted drugs in a different way in different patients. Differentiation/pigmentation program is less extensively studied. Several observations indicate, however, that differentiation status of melanoma cells is highly clinically relevant. Melanoma patients with pigment-producing metastatic lesions have shorter disease-free survival compared with patients with nonpigmented melanomas [30,31].
The reacquisition of proliferating status in metastatic sites is linked to a differentiation program [32]. BRAF inhibition is associated with increased melanoma antigen expression, including Melan-A and TYRP2 (DCT) [9]. Dedifferentiation was recently recognized as a mechanism of resistance to adoptive T-cell transfer therapy to the Melan-A/MART-1 antigen in a patient with metastatic melanoma [33].
We approached the pigmentation/differentiation program at genetic and phenotypic levels. Using only a few patient-derived cell lines, we found a plethora of possibilities how this program can be executed in drug-naïve, drugtreated, and drug-resistant melanoma cells, including those on "drug holiday" (Figure 8).
Differentiation/pigmentation but also proliferation and melanoma cell survival are mediated by MITF-M, an isoform unique for melanocytes and melanoma [14,15,34]. Models linking MITF-M with melanoma phenotype, the rheostat model [8], and phenotype switching model [35] constantly evolve [36]. Most recently, a multistage differentiation model has been presented, which categorizes melanoma differentiation as four distinct stepwise stages [7]. When we aligned patient-derived melanoma cell characteristics shown in this and our previous [5,37]   other cell lines). We have previously shown that short treatment with vemurafenib or trametinib resulted in the enrichment of a small CD271 (NGFR) high /Ki-67 low subpopulation [5]. In this study, the percentages of Melan-A high /Ki-67 low cells increased in response to vemurafenib or trametinib, which altogether indicates that proliferation program executed by melanoma cells can be simultaneously substituted by pigmentation and stem-like cell programs upon acute drug exposure. Moreover, this is not only caused by elimination of the proliferating cell subpopulation as transcript levels of pigmentation-related genes were increased before cell viability was affected. It has been very recently shown that distinct drug-tolerant transcriptional states, pigmented, starvationlike, invasive, and stem-like states, can cooccur in a minimal residual disease established through a nonmutational adaptive process [38].
Upon acquired resistance to vemurafenib or trametinib melanoma cells either progressed to more dedifferentiated subtype or had similar or even enhanced differentiation status in comparison to their original counterparts. These results showing that dedifferentiation can be induced by both drugs but only in selected melanoma cell lines suggest that (1) acquired resistance is not always accompanied by a dedifferentiation process as recently shown [7] and (2) dedifferentiation-associated resistance is rather patientrelated than drug-specific. This raises the question about the relevance of combined treatment (e.g., dabrafenib and trametinib) in patients who already developed dedifferentiationassociated resistance to vemurafenib.
Two cell lines, DMBC11 and DMBC12, expressed MITF-M and pigmentation-related genes at very low levels. This might be caused by homozygous mutation leading to MC1R R151C variant. MC R is a highly polymorphic gene [39,40] that contributes to the diversity of pigmentation [41]. Some variants, including R151C, increase the risk for melanoma [42,43]. MC1R, activated by -MSH, increases cAMP level leading to activation of the signalling cascade involving CREB (cAMP response element-binding protein) and MITF-M, which results in induction of melanogenesisrelated gene expression. -MSH, mainly synthesized by keratinocytes, can be also produced by melanoma cells [44]. All MC1R variants, R151C, V60L, and I155T, found in our study have been already described as having altered activity (Suppl .  Table S6), but only homozygous mutations result in loss of MC1R function in melanocytes [45]. Our study indicates that also in melanoma only the homozygous MC1R R151C variant can be connected with reduced MITF-M expression as DMBC33 cells (MC1R R151C +/− ) showed a high MITF-M level and still efficiently executed the pigmentation program, whereas drug-naïve DMBC11 and DMBC12 cells expressing MC1R R151C +/+ variant and the resistant cells, 21 PLXR and 28 PLXR, that acquired this homozygous mutation expressed MITF-M and pigmentation-related genes at very low levels. MITF-M level and activity are modulated by several mechanisms [14]. To the best of our knowledge, this is the first report that links lack of functional MC1R with very low level of MITF-M in vemurafenib-resistant melanoma cells derived from cells that were originally MITF high /MC1R WT . Interestingly, resistant melanoma cells that acquired a homozygous mutation leading to MC1R R151C +/+ variant and became MITF low cells also acquired several other de novo mutations, the same as originally present in drug-naïve MITF low /MC1R R151C +/+ melanoma cells. Our results support earlier observations that homozygous mutations in MC R can be connected with an elevated mutation burden in melanoma patients [46].
Our study indicates that alterations in expression of differentiation/pigmentation-related genes during development of resistance do not always follow changes in the level of MITF-M. Although MITF-M plays the central role in the pigment formation in melanocytes, several other mechanisms including transcriptional regulation involving p53, LEF-TCF, HNF1 , SOX10, and PAX3 have been described (for review [47,48]) suggesting diverse deregulation possibilities that may occur during melanomagenesis. This might explain the discrepancies between altered expressions of different pigmentation-related genes in acute response to drugs and during the development of resistance. Comparison of changes in percentages of MITF-and Melan-A-positive cells suggests that MLANA was not responsive to MITF-M-dependent transcriptional regulation in the majority of resistant cell lines. Dedifferentiation has been recently shown as a new mechanism that can lead to acquired resistance to cancer immunotherapy [33]. As we demonstrated that MLANA expression can be selectively downregulated in some resistant melanoma cell lines that still exert a high level of MITF and expression of pigmentation-related genes, not all pigmented drug-resistant melanomas might respond to adoptive T-cell transfer therapy targeting the Melan-A/MART-1 antigen.
Growing evidence indicates that the emergence of metastatic and treatment-resistant cells is not exclusively due to mutational mechanisms, but fluctuations in microenvironment/drug-dependent epigenetic states should also be considered [4,38,[49][50][51]. In this study, we have shown that, in the majority of drug-resistant cell lines, MITF-M expression was downregulated in comparison to drug-naïve cell lines and was further reduced upon drug removal ("drug holiday"), but higher percentages of MITF-positive cells returned after a short reexposure to drugs. These results show that epigenetically driven adaptive plasticity is well-preserved in melanomas that become resistant to therapeutics targeting the MAPK pathway. Intriguingly, our results have also shown that resistant cells on "drug holiday" were not drug addicted and did not respond to drug withdrawal with increased lethality as reported previously for four other resistant melanoma cell lines [52]. Therefore, intermittent therapies might not improve efficacy of continuous treatments but this remains to be confirmed in larger studies.

Conclusions
In conclusion, this and our previously published study [5] indicate that acute exposure to vemurafenib or trametinib can lead to simultaneous appearance of more differentiated and more primitive melanoma cells in different proportions, as shown at the transcript and protein bulk levels but also in the composition of cell subpopulations. Interestingly, for the first time we have demonstrated that this balance is much closer to an irreversible dedifferentiation state in those melanoma cell lines, in which loss-of-function mutation in MC R either is originally harboured or was acquired during the development of resistance. In other cell lines, acquired resistance is accompanied with reversible changes in the MITF level. Our results indicate that the development of resistance to targeted therapeutics is not always unidirectional and connected with dedifferentiation of melanoma cells, and therefore differentiated/pigmented state should be also considered as one of drug-tolerant phenotypes of melanoma. If we consider that, depending on the patient, resistance following targeted treatment can be connected with either enhanced differentiation or dedifferentiation process, which in addition could be reversible or irreversible, the need to better characterize melanoma cells in respect to their differentiation status is clear as far as the new combination therapies are worked out. Thus, our results extended by further studies might be diagnostically and therapeutically exploited to limit melanoma drug resistance.

Data Availability
WES data analysis: raw data are publicly available under the accession numbers E-MTAB-6978 (drug-naïve melanomas) and E-MTAB-7248 (drug-resistant melanomas) at ArrayExpress.

Conflicts of Interest
The authors declare that there are no conflicts of interest regarding the publication of this paper.