Overexpression of Meloe Gene in Melanomas Is Controlled Both by Specific Transcription Factors and Hypomethylation

The melanoma antigens MELOE-1 and MELOE-2 are encoded by a messenger, called meloe, overexpressed in melanomas compared with other tumour cell types and healthy tissues. They are both able to elicit melanoma-specific T cell responses in melanoma patients, and MELOE-1-specific CD8 T cells have been involved in melanoma immunosurveillance. With the aim to develop immunotherapies targeting this antigen, we investigated the transcriptional mechanisms leading to the preferential expression of meloe messenger in the melanocytic lineage. We defined the minimal promoter region of meloe gene and identified binding motifs for a set of transcription factors. Using mutagenesis, co-transfection experiments and chromatin immunoprecipitation, we showed that transcription factors involved in meloe promoter activity in melanomas were the melanocytic specific SOX9 and SOX10 proteins together with the activated P-CREB protein. Furthermore, we showed that meloe promoter was hypomethylated in melanomas and melanocytes, and hypermethylated in colon cancer cell lines and mesotheliomas, thus explaining the absence of P-CREB binding in these cell lines. This was a second key to explain the overerexpression of meloe messenger in the melanocytic lineage. To our knowledge, such a dual transcriptional control conferring tissue-specificity has never been described for the expression of tumour antigens.


Introduction
We previously identified a new transcript meloe (melanoma overexpressed antigen) over-expressed in human melanomas and under-expressed in other tumour cell types or healthy tissues. This transcript appears specifically expressed in humans, since in silico analyses did not reveal any homologous transcript in other mammalian species. This unconventional mRNA codes for at least two antigens namely MELOE-1 and MELOE-2 recognized by melanoma specific T cells in the HLA-A2 context [1,2]. We also showed that the adjuvant infusion of MELOE-1 specific CD8 T cells prolonged relapse-free survival of melanoma patients treated by adoptive transfer of tumour infiltrating lymphocytes [1]. The immunological interest of MELOE-1 protein was further strengthened by the discovery of a vast CD8 T cell repertoire specific for this antigen, in all HLA-A2 patients [3] and by the characterization of multiple class II helper epitopes from this antigen [4,5]. These properties make this antigen an attractive target for immunotherapy protocols of melanoma, and thus we sought to investigate the transcriptional mechanisms leading to the overexpression of meloe in the melanocytic lineage.
Aside from mutated antigens, tumour antigens can be divided into different groups based on their expression profile in healthy and malignant tissues. Classical overexpressed antigens, such as P53 [6] and Telomerase [7], are highly expressed in a variety of tumour cells and at a lower level in their normal cell counterparts. Genes coding cancer germline antigens (such as MAGE genes), are in turn expressed in many different tumours but are silent in normal cells, except in male germline cells [8]. Finally, tissue differentiation antigens, such as melanocytic antigens [9], are specifically expressed in a cell lineage, including tumour and healthy tissues. None of these expression profiles corresponds exactly to that of meloe messenger, overexpressed in melanomas but weakly expressed in other tumour cell types and healthy tissues [1], suggesting a regulation specific of melanomas.
Two main mechanisms have been involved in the regulation of tumour antigen expression: regulation by tissue specific transcription factors (TFs), responsible for the expression of differentiation antigens and hypomethylation of the gene promoters in tumour tissues, as reported for cancer germline antigens, such as MAGE, BAGE, GAGE and NY-ESO-1 antigens [10,11,12]. Such epigenetic alteration in the cancer cell genome leads to a shared expression of these tumour antigens between different types of cancers, that does not fit with meloe expression profile. Differentiation antigens, in turn, are exclusively expressed in the melanocytic lineage, such as Melan-A, gp100, Tyrosinase or TYRP1. Their tissue specificity is conferred by melanocytic specific TFs, such as micropthalmia associated TF (MITF) [13,14,15,16]. Unlike these antigens, meloe expression was not formally restricted to the melanocytic lineage as a residual expression can be detected in other cancer cell types, however, at a level too low to induce the activation of MELOE-1 or MELOE-2 specific T lymphocytes [1,2]. Thus, in this study we investigated the mechanisms responsible for the unusual expression profile of meloe messenger, first by defining its minimal functional promoter and then by looking for TFs and regulation mechanisms involved in its overexpression in melanomas.

Tumor cell lines
Melanoma cell lines were established in the GMP Unit of Cellular Therapy and in our laboratory from lymph node metastases and belong to the Biocollection PCU892-NL (CHU Nantes). Written consents were obtained from all patients. This biocollection was approved by the local ethic committee of Nantes hospital (GNDES), and registered under the CNIL number «1278197». Human Mesothelioma cell lines, Meso45, Meso61, Meso85 and Meso163, previously characterized [17] and belonging to the Biocollection PCU892-MG were gifts from M. Grégoire (INSERM U892, Nantes, France). Breast cancer cell line MCF-7, lung carcinoma cell lines A549 and H69, colorectal carcinoma cell lines HCT116, SW707 and SW480, renal carcinoma cell line A498 and neuroblastoma cell line SH-SY5Y were obtained from the ATCC. Melanocytes were gifts from M. Regnier (L'Oréal Laboratory, Paris, France). Cell lines were cultured in RPMI-1640 medium, or DMEM medium for the neuroblastoma cell line, supplemented with 10% heat-inactivated fetal calf serum (PAA), 2 nM L-glutamine, 100 IU/mL penicillin and 0.1 mg/mL streptomycin (Gibco). qPCR RNA samples, extracted from tumour cell lines and melanocytes by RNA purification system NucleoSpin RNA II (Macherey-Nagel, Hoerdt-FRANCE) exhibited an RNA integrity number .9. Retrotranscription was performed using 1 mg of total RNA, oligodT, and SuperScript III reverse transcriptase (Invitrogen-Life-Technologies, Saint-Aubin-FRANCE). Relative quantification of meloe and house keeping genes (HKG) RPLPO and cyclophilin-A expression was performed using Brilliant SYBR Green qPCR (Stratagene-Agilent Technologies, Les-Ulis-FRANCE). cDNA samples (20 ng) were added to SYBR Green Master Mix with specific primers at 200 nM. RPLPO: 59-GTGATGTGCAGCTGATCAAGACT-39 and 59-GATGAC-CAGCCCAAAGGAGA-39; cyclophilin-A: 5 9-CCACCGTGTTCTTCGACAT-39 and 59-CCAGTGCTCA-GAGCACGAAA-39; meloe: 59-GTCCTCCCCAGCACCA-GAGT-39 and 59-AGCCTGCCATCTGCAATCCT-39. For the three genes, thermal cycling was 95uC for 109, 40 cycles at 95uC for 3099, 60uC for 19, and 72uC for 19. The efficiency of PCR reaction was validated with duplicate series of 10-fold-diluted cDNA from the melanoma cell line M113, performed in parallel to plot the standard curves for the three genes. Mean threshold cycle (CT) values from duplicate PCR reactions were normalized to mean CT values for the two HKG from the same cDNA preparations. The relative expression ratio of a target gene was calculated based on the PCR efficiency (E) and the CT deviation between a given cell line (x) and a calibrator (eight melanocyte cultures), expressed in comparison with the mean of the HKG: ratio = (E target) DCT target (calibrator -x) /mean ((E HKG) DCT HKG (calibrator -x) ).
The PCR product was ligated into the pSC-B-amp/kan cloning vector (Stratagene-Agilent Technologies). The cloned 59RACE constructs were sequenced by the DNA Sequencing Facility of the SFR Sante.

Construction of truncated and mutated promoters
All constructs using Firefly luciferase reporter gene were performed in the promoterless pGL4.10 vector (Promega, Charbonnières-FRANCE). Various truncated meloe promoters were generated by PCR with a shared reverse primer: 59ccatggtggcgaagggatgttcacactgccttgg-39. Forward primers were designed with restriction sites for subcloning in pGL4.10, at various positions starting from 21565 bp. The promoter fragment of the melan-A gene, spanning 230 nucleotides, was used as a positive control [20]. Mutated

Plasmid transfections and Dual Luciferase Reporter assay
Cells were transfected with a mixture of 300ng of the pGL4.10 Firefly luciferase vector under meloe promoter control and 20 ng of Renilla luciferase vector pRL-CMV (Promega), with Lipofectamine (0.5 mL, Invitrogen) in serum free medium. For co-transfection experiments, 50 to 150 ng of co-transfected or empty vector were added to the transfection mixture. After 2 h30, complete medium was added to the transfection mixture. After 48 h, luciferase activities of cell extracts were measured using the Dual-Glo Luciferase assay system (Promega). For each experiment, the average ratio of the Firefly luciferase to the Renilla luciferase activity of three independent transfections was calculated.

Meloe is overexpressed in melanomas and melanocytes
To strengthen our previous results concerning the expression profile of meloe in the melanocytic lineage compared to cancer cell lines of different origins [1], we expanded our sampling of melanocytes and melanoma cell lines, for extensive qPCR analysis. We also tested a number of house-keeping genes (HKG), and choose two genes whose expression remained constant between all the tested tumour cell lines and the melanocytes. As shown in Figure 1, the overexpression of meloe was confirmed in melanomas and more surprisingly in melanocytes, compared with other tumour cell types, such as colon, renal, lung, breast, mesotheliomas and neuroblastoma cell lines. In our previous study, we found that meloe was expressed to a lower extent in two of these same melanocytes (01MO8 and 98MO9) [1]. This discrepancy can be explained by the HKG used in our previous study. Indeed, one of them (ß2-microglobulin) was expressed at a higher level in melanocytes than in melanoma cell lines, thus artificially decreasing meloe relative expression. The very low expression of meloe in other healthy tissues was confirmed using the new HKG (data not shown), thus we can assert that this antigen is a tumour antigen overexpressed in the melanocytic lineage.
Minimal promoter definition and validation of transcription factor binding sites meloe messenger comprises multiple short ORFs (Open Reading Frame). The most proximal ORF shown to be translated is MELOE-2 [2]. We first defined the transcription initiation site by 59 RACE PCR analysis starting from MELOE-2 proximal 59 region. The longest amplified sequence places the start codon 544 bp upstream of MELOE-2 ATG, and thus presented 259 additional bases compared to the public sequence of meloe transcript [NR_026664].
In silico analysis of meloe proximal promoter region (from 22000 to +1 nucleotide), located in the third intron of the HDAC-4 gene (NG_009235.1) [1], conducted with Jaspar database (Jaspar http://jaspar.cgb.ki.se/), highlighted the putative TATA element located 30bp upstream of the +1 transcription start and a number of motifs able to bind TFs, some of them illustrated on Figure 2A. We focused on TFs involved in melanocytic differentiation or melanoma progression. We found two E-boxes (CA [T/C]GTG) consensus sequences (2659 and +495) able to bind MITF [22], along with ETS (2409), AP1 (2354), SOX (293), and CREB (280) consensus sequences. We also identified a potential PAX3 consensus sequence (2635) sharing features with common patterns validated in a previous study [23].
In order to define the minimal promoter region and validate each of these consensus sequences, we conducted deletion and mutational approaches with a reporter plasmid, replacing MELOE-2 sequence by the Firefly luciferase gene. Each construct was transfected in the melanoma cell line M113, together with a control plasmid coding Renilla luciferase. The promoter of the melanocytic differentiation antigen Melan-A was used as a positive control [20]. Results first show that meloe promoter (from P-1565 to P-415) is as active as Melan-A promoter in melanoma cells ( Figure 2B). Promoter activity decreased when the region between 2415 and 2268 (containing ETS and AP1 consensus sequences) was deleted (p,0.01), and was completely abrogated after deletion of the region containing SOX and CREB consensus sequences (P-76) (p,0.001). To formally validate these putative binding sites, we tested the activity of the promoter region P-644, containing mutated consensus sequences ( Figure 2C). A significant decrease in luciferase activity was observed with promoters mutated on ETS or AP1 binding sites (p,0.01), in line with the decrease of promoter activity observed with P-268 truncated plasmid. Mutation of SOX and CREB binding sites entailed a drastic loss of promoter activity (P,0.001). Thus, SOX and CREB seem to be key factors involved in meloe transcription. As expected from deletion experiments, the mutation of the putative PAX3 binding site does not affect promoter activity. Finally, mutation of the two potential MITF binding sites (2659 and +495), performed on longer and shorter promoter regions (P-1376 and P-415), did not affect promoter activity (data not shown).
Validation of the activator role of SOX9 and SOX10 on meloe promoter activity Among the SOX factor family, SOX9 and SOX10 belong to the SOX-E factors implicated in melanogenesis regulation. They are both expressed in melanocytes and melanomas and involved in the transcription of various genes [24,25]. We assessed the expression of these two factors, together with the active P-CREB in melanoma cell lines, compared with two other tumour cell types that poorly or did not express meloe gene (colon carcinoma cell lines and mesotheliomas). As shown in Figure 3A, SOX9 is strongly expressed in melanoma cells, but weakly in colon carcinoma cells or mesotheliomas. SOX10, in turn, is expressed exclusively in melanoma cells. As expected, P-CREB is detected in each cell lysate.
In order to validate the implication of SOX9 and/or SOX10 factors in meloe promoter activation, we co-transfected expression vectors coding these factors, together with the P-644 plasmid in a colon carcinoma cell line (SW480) and a mesothelioma cell line (Meso163). The presence of each of these factors restores meloe promoter activity in these cell lines, strongly suggesting their involvement in meloe transcription in melanoma cells ( Figure 3B).
A complementary approach was conducted using a dominantnegative SOX10 factor. This mutated form (E189X) contains the high-mobility-group binding domain but is defective for the activation of gene transcription [21]. The lack of activity of this truncated form, compared to the wild type form, was first confirmed upon co-transfection into SW480 cell line, with the P-644 meloe promoter plasmid ( Figure 3C, left panel). Conversely, its co-transfection into M113 cell line with the P-644 plasmid strongly decreases promoter activity in a dose-dependent manner, by competing with endogenous SOX9 and SOX10 proteins. As expected, co-transfection of increasing doses of SOX10 wild type form has no effect in these cells spontaneously expressing high levels of SOX10 and SOX9 proteins ( Figure 3C, right panel).

Validation of the activator role of SOX9/SOX10 and P-CREB by ChIP
To demonstrate that SOX9, SOX10 and P-CREB are actually recruited on meloe promoter in vivo, a ChIP experiment focusing on a restricted promoter region bearing SOX and CREB binding sites ( Figure 4A) was conducted on three tumour cell lines: M113 cell line strongly expressing meloe, and two non melanoma cell lines poorly expressing meloe (Meso163 and SW480). The results show that these three factors are recruited on meloe promoter in melanoma cells. None of these factors is recruited on meloe promoter in mesothelioma cells, consistent with the very low expression of meloe in these cells. Finally, only SOX9 seemed to be fixed on meloe promoter in the colon carcinoma cell line, to a lesser extend compared with melanoma cells ( Figure 4B).

Meloe proximal promoter methylation status
In order to explain the absence of binding of P-CREB on meloe promoter in non-melanoma cell lines, we evaluated the methylation status of meloe proximal promoter. Indeed, it is widely documented that transcription efficiency depends on the methylation status of CpGs on promoter' genes, allowing or not the binding of TFs. The CREB binding site contains a CpG dinucleotide, whose methylation could impair CREB binding. Thus, a bisulfite conversion of genomic DNA from adult melanocytes, melanomas, mesotheliomas and colon carcinoma cell lines, was performed. Methylation status of the 26 CpGs included in the meloe region spanning from 2270bp to +544 bp was analyzed ( Figure 5). Results show that this region is almost totally unmethylated in melanocytes and melanoma cells. On the contrary, most of the CpG motifs of this region, including that of the CREB binding site, are hypermethylated in mesothelioma cells (100% in Meso45 and 77% in Meso163), and in colon carcinoma cells (64% in SW480 and 85% in SW707). The hypermethylation of the promoter region is consistent with the underexpression of meloe gene in non-melanocytic cells.

Discussion
By qPCR analysis, we showed that meloe messenger was overexpressed only in the melanocytic lineage, although a weak expression can be found in other cancer cell types, such as colon carcinoma cell lines (Figure 1), and in healthy tissues ( [1], and data not shown). This expression profile does not fit with that of a differentiation antigen, expressed in a specific lineage and totally absent in other tissues, nor with that of a cancer germline antigen normally expressed in germ cells and trophoblast tissues and aberrantly expressed in a variety of human malignancies [26]. In order to characterize mechanisms involved in meloe transcription, we first defined the minimal promoter region active in melanoma cells and look for TF binding sites that could be relevant in the melanocytic lineage. As shown on Figure 2C, CREB and SOX binding sites appeared essential for promoter activity whereas AP-1 and ETS binding sites seemed only involved in optimal promoter activation. Conversely, the putative PAX3 binding site (2635) did not seem involved in promoter activity, as well as the two proximal and distal putative MITF binding sites, that were actually E-boxes (CA [T/C]GTG) known to bind factors belonging to the basic helix-loop-helix (bHLH) and bHLH leucine zipper (bHLH-LZ) families. MITF, a bHLH-LZ TF critical for the regulation of melanocyte functions, recognizes an «AGTCA [T/C]GTG» DNA motif termed ''M-box'', identified on gene promoters regulated by this factor, such as Tyrosinase or TRP-1 [27,28]. It has to be stressed that the two E-boxes found on the analyzed sequence were not in the context of a M-box, thus not promoting MITF binding.
In conclusion, these putative binding sites are not involved in promoter activity, but this does not exclude a possible fixation of PAX3 and/or MITF at more distant sites, and their possible role in meloe transcription regulation.
The AP1 family of bHLH-LZ TFs includes JUN and FOS that heterodimerize to form DNA-binding complexes and stimulate transcription of genes containing the AP1 consensus DNA-binding site TGA(C/G)TCA [29]. In melanomas, it has been demonstrated that c-JUN protein, together with CEBP/ß TF, was involved in transcriptional regulation of a specific melanoma differentiation associated gene (mda-7) [30]. The family of ETS factors regulate a number of genes implicated in tumour progression such as those coding metalloproteinases [31]. In melanomas, ETS-1 has been recently identified as a key factor in upregulation of Mcl-1 gene, upon endoplasmic reticulum stress, and thus in the resistance of melanoma cells to apoptosis [32]. Members of these two TF families could thus contribute to meloe overexpression in melanoma cells and melanocytes, without being absolutely necessary for minimal promoter activity.
Conversely, SOX and CREB proteins appeared essential for meloe expression as mutations of their specific binding sites completely abrogate promoter activity ( Figure 2C). For this reason, we focused this first study on these factors, absolutely needed for meloe transcription. Furthermore, we also checked the absence of single nucleotide polymorphisms in these two binding sites, by in silico analysis [33], in order to exclude any interindividual variation.
CREB is an ubiquitous bHLH-LZ TF that regulates the expression of numerous genes suppressing apoptosis, inducing cell proliferation, mediating inflammation and tumour metastases [34]. It has been involved in the tumorigenicity and metastatic potential of melanomas, through the regulation of expression of metalloproteinases and adhesion molecules [35,36]. SOX TFs are grouped into nine classes. Among them, those participating in melanocyte differentiation and melanogenesis belong to the SOX-E group, including SOX9 and SOX10 [37], both expressed in established melanoma cell lines. SOX10 factor is crucial for the melanocytic development process, because of its regulation of MITF gene [38], but after the establishment of melanocyte precursor, SOX9 plays a similar role in this differentiation process in adults [39]. SOX9 and SOX10, overexpressed in melanomas  [40], are implicated in the regulation of genes involved in melanogenesis such as MITF and tyrosinase [25].
We therefore endeavoured to formally demonstrate the implication of CREB, SOX9 and SOX10 factors in meloe promoter activity, first by evaluating their expression in melanoma cells. As expected, the activated form of CREB protein (P-CREB) was present in melanoma, colon cancer and mesothelioma cell lysates ( Figure 3A). SOX9 protein was strongly present in melanoma cells, and weakly detected in colon cancer and mesothelioma cell lines whereas SOX10 expression was restricted to melanoma cell lines. These results are consistent with known SOX9 and SOX10 expression profiles [41]. The involvement of these two factors in meloe promoter activity has been further validated either upon co-transfection of wild-type SOX9 and SOX10 factors into non-melanoma cell lines ( Figure 3B), or using a SOX10 dominant-negative form co-transfected into melanoma cells ( Figure 3C). ChIP experiments further formally demonstrated the in vivo binding of SOX9, SOX10 and P-CREB on meloe promoter region in melanomas (Figure 4). Considering the proximity of SOX and CREB binding sites (13bp), we can hypothesise that they could act synergistically, as previously shown for MITF promoter [42].
Considering the ChiP experiments, none of these three factors (not even the ubiquitous CREB protein) was bound to meloe promoter in mesotheliomas, and only SOX9 was poorly bound to this region in colon carcinoma cells. Since the CREB binding is dependent of CpG dinucleotide demethylation included in the CRE motif, we further investigated the methylation status of proximal meloe promoter in these three tumour cell types and in melanocytes. The region from 2270 to +544 bp of meloe was almost unmethylated in melanoma cell lines and in melanocytes, and hypermethylated in colon carcinoma cell lines and even more in mesotheliomas ( Figure 5). The CpG dinucleotide of CREB binding site is thus hypermethylated in non-melanoma cell lines and likely impairs P-CREB binding in these cells, contributing to the under-expression of meloe gene. In cancer cells, epigenetic modifications are very frequent, regulating positively or negatively gene expression. Many tumours show hypomethylation of their genome, but the hypomethylation of the region of 800bp surrounding meloe transcription start site concerns both melanomas and their normal counterparts and thus appears specific to the melanocytic lineage. In the same line of thought, a melanoma specific hypomethylation has been described for the first intron of FOXP3 gene, demethylated in regulatory T cell clones and in melanoma cells, compared to other cancer cells such as lung and colorectal carcinoma cells [43].
To our knowledge, such a dual transcriptional control conferring tissue specificity for a gene coding tumor antigens has never been described, and the mechanisms leading to the melanoma specific hypomethylation of meloe gene will be explored in depth.