Microphthalmia Transcription Factor Induces Both Retinal Pigmented Epithelium and Neural Crest Melanocytes from Neuroretina Cells*

Mitf encodes a basic helix-loop-helix transcription factor that plays an essential role in the differentiation of the retinal pigmented epithelium (RPE) and neural crest-derived melanocytes. As cells containing melanogenic enzymes (TRP2) are found in Mitf mouse mutants, it is not clear whether Mitf is a downstream factor or a master regulator of melanocyte differentiation. To fur-ther study the role of Mitf in committing cells to the melanocyte lineage, we express Mitf in the cultured quail neuroretina cells. This leads to the induction of two types of pigmented cells: neural crest-derived melanocytes, according to their dendritic morphology, phys-iology, and gene expression pattern are observed together with pigmented epithelial RPE-like cells. The expression of Mitf is lower in pigmented epithelial RPE-like cells than in neural crest-derived melanocytes. Ac-cordingly, overexpression of Mitf in cultured quail RPE causes cells to develop into neural crest-like pigmented cells. Thus, Mitf is sufficient for the proper differentiation of crest-like pigmented cells from retinal cells and its expression level may determine the type of pigment cell induced. In vertebrates,

In vertebrates, melanin is synthesized in a specialized lysosome-related organelle, the melanosome, by transmembrane glycoproteins of the tyrosinase family (1,2). These organelles are found in two types of cell: the neural crest-derived melanocytes (NC-M) 1 and the retinal pigmented epithelium (RPE). The RPE is a monolayer of cuboidal pigmented cells, involved in the maintenance and function of the vertebrate eye. The RPE and neural retina (NR) are generated from common precursors derived from the proximal part of the budding neural epithelium that retains the capacity to transdifferentiate into each others cell type both in vivo and in vitro (3)(4)(5)(6).
The neural crest is a transitory structure that arises from cells of the neuroepithelium as it folds to form the neural tube. Neural crest cells migrate from the neural tube to characteristic peripheral sites, giving rise to diverse cell types. These include melanocytes of the integument, iris, and inner ear (7). Neural crest-derived cells do not overtly differentiate until after they disperse. Therefore, it is unclear whether the appearance of differentiated cells in a precise location is the result of the differential localization of developmentally distinct precursors or whether factors in these locations instruct developmentally multipotent precursors to adopt specific fates. It has been suggested that at least some neural crest cells become committed to these phenotypes via a series of developmental restrictions, which progressively limit their potential fates (8 -10). Clonal analysis of randomly labeled neural crest cells indicated that c-Kit-expressing cells are fate-restricted melanocyte precursors (9). Indeed, NC-M migrate along the dorsallateral pathway, through the space between the ectoderm and dermomyotome of the somite to give rise to the skin melanocytes. These c-Kit-expressing cells respond to the MGF (also known as Kit ligand or stem cell factor), which is required for normal migration and differentiation of melanocyte precursors (11)(12)(13). In mammals, these cells also express the endothelin receptor B (ETRB). Spontaneous or experimental inactivation of the ETRB gene, as that expressing the endothelin 3 (EDN3) ligand, causes defects in body pigmentation and posterior gut innervations (14,15). In the avian embryo, ETRB starts to be expressed by premigratory neural crest cells. Cells traveling along the dorsal-lateral pathway down-regulate the ETRB gene and switch to express another subtype, EDNRB2 (16). EDN3, a potent mitogen for cultured quail neural crest cells, promotes their differentiation into melanocytes and stimulates EDNRB2 expression, whereas increasing the survival and proliferation of committed melanocytes and glial-melanocytic precursors (17). Other factors from the Wnt and BMP families are also involved in the lineage segregation of NC-M in avian embryos (18).
RPE and neural crest-derived melanocytes differ in several important ways. Neural crest-derived melanocytes, but not cuboidal RPE cells, show a dendritic morphology and play an essential role in the delivery of melanosomes to the surrounding keratinocytes (19). The trafficking of melanosomes in the RPE and neural crest-derived melanocytes involves myosin VIIa and myosin Va, respectively (20,21). None of the growth factors required for NC-M development or their receptors are required for RPE differentiation, as shown by the normal eye phenotype of mutant mice (22)(23)(24). Additional NC-M-specific antigens have also been characterized. These include MelEM (25), which encodes the ␣ subunit of the glutathione S-transferase expressed in melanoblasts committed to produce black pigment (26). However, MelEM is not found in the RPE (25).
Mice that bear severe mutations in the Mitf gene are completely unpigmented and deaf, because they lack neural crestderived melanocytes (27). This shows that Mitf function is required very early in the development of the melanocyte lineage (28). These mice are also microphthalmic because of the abnormal development of the RPE, which appears unpigmented and often pluristratified (23,29). Mitf encodes the microphthalmia-associated transcription factor of the basic helix-loop-helix and leucine zipper family. Mitf is expressed as multiple isoforms termed Mitf-M, Mitf-A, Mitf-D, Mitf-C, and Mitf-H (30). Mitf-M expression is restricted to neural crestderived melanocytes, whereas Mitf-A is strongly expressed in the RPE (31). In line with these observations, Mitf activates the promoter regions of genes involved in terminal differentiation of pigmented cells, including the genes encoding the melanosome glycoprotein QNR71, the melanogenic enzyme tyrosinase (Tyr), and the tyrosinase-related proteins TRP-1 and TRP-2, by specifically binding to the hexameric motif CATGTG (M-box) present in all these promoters (reviewed in Ref. 31). Mitf is one of the targets of the signaling cascade initiated by the c-Kit tyrosine kinase (31). The c-Kit promoter has been shown to be directly regulated by Mitf in mast cells (32). Mutations in transcription factors, including positive regulators of Mitf such as Pax3 (33) and Mitf itself (31,34), can result in pigmentation defects (Waardenburg syndromes 1 to 4). Mitf is therefore one of the most important transcription factors governing the development and differentiation of pigment cells in vertebrates. However, its role in the commitment to the melanocyte lineage is unclear because melanoblasts containing Mitf mRNA can appear and survive for 48 h in the absence of a functional Mitf protein (35,28).
We previously showed that ectopic expression of Mitf led to the appearance of pigmented foci from neuroretina cells (6). This study showed that both RPE-like epithelial foci and NC-M-like dendritic foci were produced. NC-M-like cells were MelM-, Steel-, Kit-, Pax3-, and EDNRB2-positive. Their proliferation and pigmentation increased following the addition of EDN3 to the cultures. RPE-like foci expressed lower levels of Mitf protein than NC-M-like foci. To demonstrate that a high level of Mitf could induce an NC-M-like phenotype in RPE, we overexpressed Mitf-M in cultured RPE cells. We observed that Mitf expression, a dendritic morphology, increased pigmentation and MelM staining. We propose that Mitf is probably involved in a common developmental event that precedes cell fate specification and can induce the genesis of RPE-or NC-Mlike pigmented cells depending on the level of Mitf protein synthesized in neuroepithelial cells.

MATERIALS AND METHODS
Cell Culture and Transfection Experiments-Dissociated neuroretina cells obtained from 6-day-old quail embryos were plated in Dulbecco's modified Eagle's, F-12 medium supplemented with 10% fetal calf se-rum, 1% minimal essential medium vitamins 100 times, and 10 g/ml conalbumin (6). The pigmented retinal epithelium was dissected from 7-day-old quail embryos. The transfected cells were passaged one time on gelatin-coated dishes (100 mm diameter) and grown for 1 month in medium containing G418 (300 g/ml) before testing. The pigmentation usually became apparent 10 days after selection. The LE-Mitf-M vector was described previously (36). The Mitf-A and Mift-M isoforms were inserted into the pVNC3 vector for comparison. The LE and pVNC3 vectors were similarly efficient.
Construction of EGFP Fusion Proteins-To visualize Mitf proteins, enhanced green fluorescent protein (EGFP, Clontech) was fused to Mitf. The following plasmids were constructed: a BglII-NotI fragment of EGFP was inserted into pVNC3 at the BglII and Bsp120I sites to produce pVNC3EGFP. A 1299-nucleotide XbaI-SalI fragment of Mitf-M was inserted into the XbaI-SalI sites of pVNC3EGFP to give pVNC3Mitf-M-EGFP, with the GFP in-frame with Mitf. To quantify the GFP signal in the Mitf-M-EGFP-expressing cells, fluorescent signals from dendritic or epithelial cell nuclei were captured with a 5-MHz Micromax 1300Y interline CCD (charge-coupled device) camera (Roper Instruments, France) as 16-bit images using identical parameters. The fluorescence content of each nucleus was calculated with the help of the integrated morphometry analysis option of the Metamorph software (Universal Imaging) and plotted for each focus.
Immunocytochemistry-Transfected cells were fixed for 20 min in 4% paraformaldehyde in phosphate-buffered saline, and incubated with a polyclonal anti-MITF antiserum (6) or with anti-bromodeoxyuridine (BrdUrd). Primary antibodies were detected with Cy3-labeled goat antirabbit immunoglobulin secondary reagent (Jackson ImmunoResearch). c-Kit-expressing cells were detected as described in Ref. 9 using a monoclonal antibody kindly provided by B. Wehrle-Haller.
Electron Microscopy-Transfected cells were fixed overnight at 4°C in 2.5% glutaraldehyde in 0.1 M cacodylate buffer. Cells were post-fixed with 2% OsO 4 for 60 min, dehydrated in increasing concentrations of ethanol, and embedded in Epon. Ultrathin sections were counterstained with uranyl acetate and then observed under a Philips CM120 Electron Microscope (FEI Company, Eindoven, The Netherlands).
In Situ Hybridization-Riboprobes were prepared by in vitro transcription of appropriate plasmids using either T3 or T7 polymerase and incorporation of digoxygenin-labeled UTP. Nonradioactive in situ hybridization was performed as described in Ref. 33. The EDNRB2 probe was kindly provided by L. Lecoin.
Reverse Transcriptase-PCR Assays-Total RNAs were extracted using the guanidinium isothiocyanate-cesium chloride method and then incubated with RQ1 DNase. Total RNA (2-4 g) was incubated with random hexamer oligonucleotides (PerkinElmer Life Sciences) or oligo(dT) for 10 min at 70°C. Murine Moloney leukemia virus-reverse transcriptase was added in the appropriate buffer to a final volume of 20 l, and the mixture was incubated for 60 min at 42°C. After dilution to 100 l, a 2-l aliquot was mixed with 0.5 units of Taq polymerase Goldstar (Eurogentec), 2 mM MgCl 2 , 1 M of each oligonucleotide, and 0.25 mM dNTP in the appropriate buffer and submitted to PCR amplification in a PerkinElmer Life Sciences thermocycler with the oligonucleotides described in Table I.

Expression of Mitf in Quail Neuroretina Cells Induces Two
Types of Pigmented Foci-Overexpression of the melanocytic isoform of Mitf (Mitf-M) in neuroretina cells is sufficient to induce a pigmented phenotype, suggesting that the protein encoded can activate the genetic network that triggers RPE transdifferentiation alone (6,37). To determine whether Mitf could also induce a neural crest-derived melanocyte phenotype, we stably transfected dissociated cells derived from E6 quail   1A). Immunocytochemistry detected Mitf in cell nuclei from pigmented foci with a dendritic appearance (Fig. 1D), but detected only a faint signal in the nuclei of epithelial RPE-like cells (Fig. 1C). All the foci that expressed Mitf were pigmented. Half of the foci in the Mitf-M-transfected cells and three foci in the Mitf-A-transfected cells were of dendritic appearance. Analysis of ultrathin sections of these cells embedded in Epon revealed that epithelial cells bore a homogeneous population of round melanosomes (Fig. 1E), whereas the dendritic cells exhibited heterogeneous melanosomes with a characteristic elliptical morphology at distinct stages of maturation (Fig. 1F), which were significantly smaller than those from epithelial cells. Dendritic Pigmented Foci Express Neural Crest-derived Melanocyte Markers-We then asked whether the induction of the dendritic appearance of the pigmented foci was accompanied by the expression of neural crest-specific markers in these cells. We first tested the expression of the MelEM antigen in these cells. As previously reported, the RPE cells were negative for MelEM staining (25), whereas the choroidal melanocytes, present in this culture of quail RPE cells and easily recognized by their dendritic morphology, were strongly labeled (Fig. 2, A and B). Dendritic foci of transfected QNR cells (Fig. 2C) expressing high levels of Mitf (Fig. 2E) were immunostained with anti-MelEM (Fig. 2G). The pigmented foci with an RPE-like morphology (Fig. 2D) were negative for MelEM (Fig. 2H). We subsequently tested for several neural crest melanocyte-specific markers: growth factors (Steel), growth factor receptors (c-Kit and EDNRB2), and the transcription factor Pax3. Reverse transcriptase-PCR experiments performed on RNA extracted from the G418-selected Mitf-M-transfected NR cells revealed a specific band with all of the primers used (data not shown). To demonstrate that these genes were specifically expressed in the pigmented foci, we also performed in situ hybridization with digoxygenin-labeled RNA probes onto photobleached foci. This allowed clear detection of the hybridization signal in the cells. The probes derived from the fragments cloned from the reverse transcriptase-PCR experiments detected Steel/MGF RNA (Fig. 3A), EDNRB2 RNA (Fig. 3B), and Pax3 RNA (Fig. 3, C and D, and the Mitf Pax3 plate for a low magnification view of the transfected plate) in the cells of pigmented foci. Non-pigmented G418-resistant foci were negative for these RNAs (data not shown and control plate for Pax3 probe). Panels E and F of Fig. 3 show the Mitf signals in dendritic and epithelial foci, respectively.
If Steel autocrine regulation was involved in the formation of neural crest-like pigmented foci in response to Mitf, we should be able to detect the c-Kit receptor in these cells. We found that a significant number of pigmented cells were labeled by a monoclonal antibody raised against a peptide within the extracellular domain of c-Kit (9) (Fig. 3, G and H). A few nonpigmented cells, close to the pigmented foci, were also labeled, but no labeled cells were found in the RPE-like pigmented foci (data not shown).
Biological Activity of EDN3-As EDNRB2 was expressed in the dendritic foci of pigmented cells, we asked whether these cells could respond to the addition of EDN3 to the culture medium, as shown for cultured skin-derived pigment cells (17). The addition of 100 nM EDN3 to the culture medium of the G418-resistant foci resulted in a clear increase in the size and the pigment content of the selected foci, but not in the number of pigmented foci (compare Fig. 4A, Mitf with and without EDN3). EDN3 did not induce any pigmentation in control cells transfected with the NeoR-expressing control vector (Fig. 4A,  control). To demonstrate that EDN3 increased pigment cell proliferation, we performed a BrdUrd incorporation assay in Mitf-M-transfected cells, with and without EDN3 in the culture medium. In the cells cultured without EDN3, very few nuclei were detected (see arrow, Fig. 4B) and double labeling showed that these cells contained Mitf (Fig. 4D, F, is a merge of B and  D). In contrast, in the presence of EDN3, most, if not all, of the Mitf-positive nuclei (Fig. 4E) incorporated BrdUrd (Fig. 4C, G is a merge of C and E). We concluded that the expression of EDNRB2 in Mitf-transfected neuroretina cells is connected to a functional downstream transduction pathway, leading to melanocyte-like proliferation.
Quantification of MITF in Pigmented Foci of Quail Neuro- This fusion protein activated the QNR-71 promoter as efficiently as did the wild-type form (data not shown) and induced a similar number of pigmented foci in the transfected QNR cells following G418 selection, suggesting that the chimeric protein is fully active (Fig. 5A). Therefore, the GFP signal present in the nucleus, which directly reflects the amount of Mitf protein, could be quantified by use of the integrated morphometry analysis option of the Metamorph software. We recovered the GFP fluorescence of each cell making distinct dendritic or epithelial foci and plotted the fluorescence intensity of each individual nucleus in each pigmented foci. The GFP signal was on average 2-fold (up to 10-fold) higher in the cells making dendritic foci than in the cells of epithelial morphology (Fig.  5B). Representative pigmented foci with dendritic and epithelial morphology are shown in Fig. 5C (panels a and c, respectively), together with the corresponding GFP signal (panels b and d, respectively).
Conversion of RPE Cells into NC-M by Mitf-Following the hypothesis that the amount of Mitf determines the type of pigmented cell, we reasoned that the expression of Mitf in RPE epithelial cells would convert these cells into a neural crest-like phenotype. We transfected cells with a Mitf-M-encoding vector, and then selected transfected cells. Cultured RPE cells became depigmented and did not exhibit any Mitf staining. In contrast, pigmented cells containing the Mitf protein (Fig. 6, B and F) were apparent in the transfected culture (Fig. 6, A and E) and were positive for MelEM immunostaining (Fig. 6, D and H). G418-selected cells transfected with an empty control vector did not exhibit such pigmented foci. However, when these control cells were maintained at confluency for several days, they recovered both pigmentation and epithelial morphology. Therefore, expression of Mitf in a differentiated RPE cell is sufficient to trigger NC-M differentiation. DISCUSSION Genetic and molecular analyses in different species have shown that Mitf has a key function in the development of melanin-producing cells, including the RPE (31). Mitf is therefore one of the most important transcription factors governing the development and differentiation of the pigment cells in vertebrates. However, its role in the commitment to the melanocyte lineage is unclear because melanoblasts containing Mitf mRNA can appear and survive for 48 h in the absence of a functional MITF protein, as shown in the mi mutant (28,35). Several other factors required for pigment cell development have been also described based on the phenotypes of mutants. All these factors, which are important at distinct steps of melanocyte development, form a biochemical cascade that ultimately modulates Mitf function (31). Gain of function studies in avian NR cultures have shown that Mitf-M is sufficient to induce a pigmented phenotype in transfected cells (6). Morphological examination revealed two types of pigmented foci. They could be distinguished by the shape of the cells: RPE-like epithelial cells and dendritic cells, reminiscent of the neural crest-derived melanocytes. Very few foci contained both types of cell. Immunodetection of antigens expressed in neural crest melanocytes (c-Kit receptor and MelEM), and in situ hybridization (Steel/MGF, EDNRB2, Pax3) or reverse transcriptase-PCR (not shown) detection of neural crest-specific gene expression, allowed us to demonstrate a neural crest-like pigment cell transdifferentiation in the culture. In addition, we observed that the addition of EDN3 to the culture medium of the Mitftransfected neuroretina increased both pigment cell proliferation and pigmentation, as already shown for neural crest-derived melanocytes (17). Therefore, either multipotent precursors that can differentiate into a neural crest phenotype in response to Mitf exist in the culture, or Mitf is able to trigger the transdifferentiation of neuroretina cells into melanocytic neural crest cells. Multipotent precursors have been found in the ciliary margin of the retina (39), and glial cells can produce neuronal cells in vivo (40), but no differentiation in the neural crest phenotype has been reported so far. Bipotent glial/melanogenic intermediates, cells with latent melanogenic potential, are present in embryonic quail peripheral nerves. Transdifferentiation of these cells into pigment cells could be observed following treatment with the tumor-promoting phorbol ester drug (41) or EDN3 (42). However, in contrast to Dupin et al. (42), we observed no pigmentation induction in control cells grown in the presence of EDN3, and no MelEM (which labels early NC-derived melanoblasts (43))-positive cells were found in the cultured neuroretina prior to transfection (data not shown), suggesting that Mitf is essential for the induction of the pigmented phenotype. As we observed a 10-fold decrease in the number of pigmented foci when the neuroretina was taken from E7 instead of E6 embryos, we favor the hypothesis that Mitf has the potential to induce a transdifferentiated phenotype in undifferentiated neuroblasts.
The pigmentation-inducing effect is not a property shared by many genes, as no pigmentation could be observed in NR cultures transfected with the erbB, ras, src, mil, fos, jun, erbA, ski, or E1A genes. However, we recently observed that Otx2-induced pigmented foci expressed Mitf (37) and that these Mitfexpressing foci exhibited both a dendritic cell morphology and MelEM staining (data not shown). The NR cells positive for the Mitf-A isoform also exhibited this particular phenotype, albeit at a much lower level than Mitf-M (only three MelEM-expressing foci were recovered after Mitf-A transfection). The two Mitf isoforms display different subcellular locations (44); Mitf-A was detected in the nucleus and cytoplasm. Therefore, the different efficiencies of the two Mitf isoforms may be because of the presence of less Mitf-A than Mitf-M in the nucleus. It is noteworthy that considerably less Mitf protein was found in the nuclei of the cells in the RPE-like foci after Mitf-M transfection than in the nuclei of the cells of the dendritic foci obtained in the same plate (as judged by the immunostaining signal intensity, and by measuring the GFP fluorescence fused to Mitf in a biologically active protein). Thus, the genes induced may depend on the amount of Mitf present in the nucleus. It remains possible that low concentrations of Mitf only activate genes involved in pigmentation (high affinity genes), resulting in the appearance of RPE-like cells, but that higher concentrations of Mitf also activate genes involved in neural crest melanocyte differentiation (low affinity genes). It has been demonstrated that E-box flanking sequences play an important role in Mitf recognition and subsequently in promoter activation (45). Therefore, an increase in Mitf concentration in RPE cells may well induce a NC-crest melanocyte phenotype in cells expressing only the high affinity targets. This was indeed observed after the introduction of a Mitf-M-encoding vector into predifferentiated RPE cells. Mitf-induced pigmentation depends on cell type, as shown by in vivo Mitf expression in non-pigmented cells. It has recently been reported that medaka-ESlike cells (46) differentiate into melanocytes in response to this factor. Our data support the hypothesis that Mitf is a master developmental regulator gene, sufficient to initiate and to complete the differentiation of neuroretina cells into melanocytes. They also suggest that the possibility to reprogram neuroretina cells toward distinct differentiation pathways is greater than previously suspected. These findings make this in vitro culture system an attractive model that should provide more insights into this and other differentiation pathways.