Activation of myelin genes during transdifferentiation from melanoma to glial cell phenotype.

Induction of myelin genes occurs around birth in the last stage of Schwann cells differentiation and is reactivated in case of nerve injury. Previous studies showed that activation of the gp130 receptor system, using as ligand interleukin-6 fused to its soluble receptor (IL6RIL6), causes induction of myelin genes such as myelin basic protein (MBP) and myelin protein zero (Po) in embryonic dorsal root ganglia Schwann cells. We also reported that in murine melanoma B16/F10.9 cells, IL6RIL6 causes a shut-off of melanogenesis mediated by a down-regulation of the paired-homeodomain factor Pax3. The present work demonstrates that these IL6RIL6-treated F10.9 cells undergo transdifferentiation to a myelinating glial phenotype characterized by induction of the transcriptional activities of both Po and MBP promoters and accumulation of myelin gene products. For both Po and MBP promoters, a repression by Pax3 and stimulation by Sox10 can be demonstrated. Because after IL6RIL6-treatment, Pax3 disappears from the F10.9 cells (as it does in mature myelinating Schwann cells) whereas the level of Sox10 rather increases, we modulated the relative level of these factors and show their involvement in the induction of myelin gene expression by IL6RIL6. In addition, however, we show that a C/G-rich CACC box in the Po promoter is required for activation by IL6RIL6, as well as by ectopic Sox10, and identify a Kruppel-type zinc finger factor acting through this CACC box, which stimulates Po promoter activity.

Induction of myelin genes occurs around birth in the last stage of Schwann cells differentiation and is reactivated in case of nerve injury. Previous studies showed that activation of the gp130 receptor system, using as ligand interleukin-6 fused to its soluble receptor (IL6RIL6), causes induction of myelin genes such as myelin basic protein (MBP) and myelin protein zero (Po) in embryonic dorsal root ganglia Schwann cells. We also reported that in murine melanoma B16/F10.9 cells, IL6RIL6 causes a shut-off of melanogenesis mediated by a down-regulation of the paired-homeodomain factor Pax3. The present work demonstrates that these IL6RIL6-treated F10. 9

cells undergo transdifferentiation to a myelinating glial phenotype characterized by induction of the transcriptional activities of both Po and MBP promoters and accumulation of myelin gene products. For both Po and MBP promoters, a repression by Pax3 and stimulation by Sox10 can be demonstrated.
Because after IL6RIL6-treatment, Pax3 disappears from the F10.9 cells (as it does in mature myelinating Schwann cells) whereas the level of Sox10 rather increases, we modulated the relative level of these factors and show their involvement in the induction of myelin gene expression by IL6RIL6. In addition, however, we show that a C/G-rich CACC box in the Po promoter is required for activation by IL6RIL6, as well as by ectopic Sox10, and identify a Kruppel-type zinc finger factor acting through this CACC box, which stimulates Po promoter activity.
Axonal myelination is a function of specialized glial cells, oligodendrocytes in the brain and myelinating Schwann cells (SC) 1 in peripheral nerves, whose impairment in demyelinating diseases affects nerve function and integrity. Maturation of cells producing myelin occurs around birth and needs to be reactivated for repair of nerve injury (1). The developmental stages from neural crest-derived precursors to mature myelinating SC were defined by specific gene markers (see Refs. 2-5 for reviews). At early stages, embryonic SC express low affinity nerve growth factor receptor, neural cell adhesion molecule (N-CAM), glial fibrillary acidic protein (GFAP), the transcription factors paired homeodomain Pax3, and later POU domain Octamer-6 (Oct-6)/suppressed cAMP-inducible POU (SCIP). This phenotype also characterizes the adult non-myelinating SC and reappears after nerve injury. Maturation of myelinating SC is marked by disappearance of these early markers, including Pax3, and the induction of myelin genes such as myelin basic protein (MBP) and protein zero (Po or myelin protein zero). Promoters driving glial-specific expression of Po (6 -8) and of MBP (9 -14) were identified along with regulatory transcription factors (see Ref. 15 for review). Thus, the high mobility group (HMG) domain Sox10, already present in early neural crest cells, activates the Po promoter (16). Krox20/Egr2, a late marker of myelinating SC differentiation required for myelination (17), stimulates expression of several myelin genes (18,19). Conversely, Pax3 represses expression of MBP (20), and so does SCIP (21), in line with the need for of Pax3 and SCIP to decrease during the terminal differentiation of myelinating SC.
Among extracellular factors acting on SC, only a few could be shown to activate myelin gene expression. The main SC growth factors, neuregulins nerve-derived factor (NDF)/glial growth factor (GGF) (22), stimulate proliferation at early SC differentiation stages but have negative effects on the induction of MBP and Po genes (23). Similarly, fibroblast growth factor or transforming growth factor-␤ inhibit Po gene expression (24). Axonal contacts are thought to provide positive stimuli for myelin gene expression in SC cultures (25) and in transected nerves (26). The molecules mediating these axonal cues are not well known. Intracellular cyclic AMP elevation (e.g. by forskolin) induces Po and MBP (25,27) whereas it represses Pax3 (20). Activating effects of forskolin were seen on the promoters of the genes encoding myelin proteins Po (6), MBP (28,29), and peripheral myelin protein-22 (PMP-22) (30). Hormones, such as progesterone and glucocorticosteroids, also stimulate Po and PMP-22 promoter activities (31,32). A third group of factors are IL-6 family cytokines that activate the gp130 receptor system. As a prototype of this large family (see Ref. 33 for review), we have used a recombinant protein IL6RIL6 resulting from the fusion of IL-6 to its soluble IL-6 receptor (sIL-6R) that activates the gp130 signaling pathway in many cell types and has a high affinity for gp130 (34,35). We showed that this gp130 activator potently stimulates the expression of MBP and Po mRNAs and proteins in cultures of mouse E14 embryonic dorsal root ganglia (DRG), as well as in derived SC (36,37). IL6RIL6 strongly down-regulated the expression of Pax3 in these embryonic cells. Moreover, IL6RIL6 enhanced in vivo the number of myelinated fibers in regenerating sciatic nerve (37). Others observed that IL6RIL6 induces MBP RNA in brain * This work was supported by Interpharm (Weizmann Industrial Park, Israel) and Ares Serono Group (Geneva, Switzerland). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We show here another system in which the gp130 activator IL6RIL6 switches on the expression of myelin genes and causes transdifferentiation of melanoma cells toward a glial phenotype. The murine melanoma B16/F10.9 cells undergo terminal growth arrest when exposed to the combination of IL-6 with its agonistic soluble IL-6 receptor (39) or to the IL6RIL6 chimera (34). There is a silencing of the melanogenic pathway because of a profound reduction in Pax3, which then down-regulates the microphtalmia-associated transcription factor (MITF) gene and, in turn, the tyrosinase activity (40). The present work demonstrates that IL6RIL6 subsequently induces expression of the myelin Po and MBP mRNAs through promoter-mediated transcriptional activation. This melanoma provides a new model to study the transcription factors involved in the regulation of myelin genes expression and their induction by IL-6 family cytokines.
cDNA Expression Vectors-RT-PCR with F10.9 total RNA was used as described (40) to amplify the entire coding sequences of Pax3c cDNA (NM_008781; nt 255-1759) and Sox10 cDNA (AF047043; nt 39 -1339). The cDNAs were cloned in the pcDNA3 expression vector (Invitrogen). The plasmid pAdRSV Krox20G (a gift of Drs. P. Charnay and P. Topilko, Paris, France) contains 5.2 kbp of Krox20 mouse genomic DNA sequence (including the complete protein coding sequence) in 3Ј of Rous sarcoma virus long terminal repeat and in 5Ј of the pIX gene polyadenylation domain.
We prepared permanently transfected clones of F10.9 cells where Sox10 or Pax3 gene expression can be increased following addition of tetracycline (or doxycyclin) to the growth medium (Tet-on system). We first selected F10.9 cell clones expressing the recombinant reverse tetracycline repressor rtTA isolated from the pTet-On regulator plas-mid (Clontech) cloned in pEFIRES-puro, using puromycin as before (40). The clones were screened for rtTA activity, wild-type growth rate, and response to IL6RIL6 by morphological change. In the selected clone cells, we introduced the vector pSV2Hygro, together with pBI vectors containing a bidirectional Tet-regulated element (Clontech). In the pBI vectors, downstream to the Tet-regulated element, we cloned the cDNA for the green fluorescent protein, enhanced green fluorescent protein (Clontech), either alone (control) or with Sox10 or Pax3 cDNA (in the opposite direction). F10.9-rtTA cell clones growing in the presence of hygromycin and becoming fluorescent after treatment with doxycycline (200 ng/ml) were selected. Expression of the ectopic mRNAs was verified by RT-PCR, using the vector reverse primer 5Ј ACTCACCCT-GAAGTTCTCAG and forward primers Sox10 (AF047043; nt 924 -948) or Pax3 (NM_008781; nt 850 -880).
Promoter Reporter Gene Constructs-The MBP reporter plasmid pBG1b (9) (a gift of Dr. C. Kioussi, San Diego, CA) was excised with BglII and BspeI and blunt-ended to obtain the 5Ј flanking MBP gene sequence Ϫ1320/ϩ33. The fragment was cloned in front of the luciferase coding sequence in the pGL3 basic plasmid (Promega) cut by XhoI and HindIII, blunt-ended, to create the plasmid pGL3MBP/Ϫ1.3. The latter was cut with BglII (in 5Ј) and PstI (in the MBP sequence) and closed by ligation to create pGL3MBP/Ϫ0.65 or was cut with Ecl136 -1 and PvuII to create pGL3MBP/Ϫ105. The rat Po promoter (6) was generated by genomic PCR, using the following primers: F, 5Ј-GACATTATCCCTCC-CATCCCCTTATTTCCC-3Ј; and R, 5Ј-GCCCAGAGCGTCTGT-GGGGT-GGAGAGAGCG-3Ј. After end polishing, the amplified fragment (genomic sequence Ϫ912/ϩ45 relative to the start site) was cloned upstream of luciferase in pGL3 basic cut with SmaI to create pGL3Po-912. For pGL3Po-500, Po gene sequence Ϫ500/ϩ45 was made by PCR with pGL3Po-912 as template and primers F, 5Ј-GGGGACGCGTCCAGGA-TGCAGGGAGATG-3Ј, with a MluI site (underlined); and R, 5Ј-GGG-GAAGCTTGCCCAGAGCGTCTGTGGG-3Ј, with a HindIII site. The MluI/HindIII PCR fragment was inserted in pGL3 basic cut by MluI/ HindIII. The other 5Ј deletion plasmids were similarly prepared. For 3Ј deletions of the Po promoter, the pGL3Po-300 plasmid was used as template for PCR amplification between primer F, 5Ј-GGGGGCTAGC-TCTATCCCTCAGAGAAGT-3Ј, with a NheI site; and R, 5Ј-GGGGAG-GCCTCCCCTGGATCCCCAGCAT-3Ј or 5Ј-GGGGAGGCCTGGGGCAT-TGTATACTCTG-3Ј, with a StuI site. The amplified fragments contain sequences Ϫ299/Ϫ30 or Ϫ299/Ϫ137 of the Po promoter. All 3Ј deletions were made with the same method and were introduced in front of the minimal herpes simian virus (HSV)-TK promoter (Ϫ73/ϩ57) in the PGL3-TK plasmid cut by NheI/StuI. PGL3-TK was made by using as template a construct containing the Ϫ155/ϩ57 sequence of HSV-TK in front of the luciferase gene in PGL3 basic. We amplified by PCR a large fragment of this plasmid by using primer F, 5Ј-GGGGGCTAGCAGGC-CTAACACGCAGATGCAGTCGG containing in 5Ј NheI (bold) and StuI (underlined) sites in front of the TK sequence from Ϫ73. The reverse primer 5Ј-TCTGGCATGCGAGAATCTCACGC-3Ј overlapped the unique SphI site of the vector. The PCR fragment cut by NheI/SphI was cloned in PGL3 basic vector cut by NheI/SphI.
Site-directed mutagenesis was done by using an Expand high fidelity PCR system (Roche Molecular Biochemicals). Two primers in opposite orientation, each one carrying half of the mutated site in its 5Ј, were used to amplify the mutated plasmid in a single PCR reaction. Typical conditions for the PCR reaction were according to the manufacturer's protocol. Amplification conditions were 95°C (3 min) for one cycle; 94°C (15 s (sec), 58°C (30 s), and 68°C (4 min) for 10 cycles; 94°C (15 s), 58°C (30 s), and 68°C (4 min) for 15 cycles; and 72°C (7 min) for 1 cycle. The PCR product was precipitated with ethanol and then phosphorylated by T4 kinase. The phosphorylated fragment was self-ligated by T4 DNA ligase and digested with restriction enzyme DpnI to eliminate the non-mutated template. The mutated plasmid was cloned and amplified in Escherichia coli (DH5-␣ strain) competent cells.
Cell Transfections-For transfections, growing F10.9 cells were seeded in 6-cm Nunc plates. After 24 h, each well received 2 ml of a mixture containing a constant amount of 3.3 g of DNA composed of 0.5 g of the Po or MBP promoter-luciferase pGL3 plasmids, 100 ng of pSV40-Renilla luciferase plasmid (Promega) and completed with empty pCDNA3, alone or with Sox10 or Pax3 expression vectors, plus 20 l of LipofectAMINE (Invitrogen), all in medium without antibiotics. After culture for 12 h at 37°C, each transfected plate was split into 10 wells of 6-well plates, and half the wells were treated with IL6RIL6, and half were left untreated. After further culture for 72 h, the dual luciferase assay system kit (Promega) was used to measure luciferase activities according to the manufacturer's protocol. Results were calculated from quadruplicate wells.
Electrophoretic Mobility Shift Assays-Nuclear extracts were pre-2 M. Pizzi, personal communication. pared from F10.9 cells grown in 9-cm dishes for different times in the presence or absence of IL6RIL6 (300 ng/ml). The cell monolayer, washed with PBS (minus calcium, magnesium), was scraped with a rubber policeman in 2 ml of phosphate-buffered saline into Eppendorf tubes. Pellets recovered by centrifugation at 3000 rpm were frozen and stored in liquid nitrogen until use. Pellets were thawed on ice and homogenized in hypotonic Buffer A (4 volumes per volume pellet) by mixing five times with pipette, left 10 min on ice, and centrifuged for 10 min at 4°C at 5000 rpm. The pellet compacted by a 10-s spin at 14000 rpm was resuspended in 2.5 volumes (relative to original cell pellet) of Buffer B by mixing five times with pipette, left on ice for 10 min, and centrifuged 14000 rpm for 10 min at 4°C. The supernatants (nuclear extracts) were kept at Ϫ70°C. Buffer A contained 10 mM Hepes, pH 7.9, 10 mM NaCl, 5% glycerol, 2 mM EDTA, 2 mM EGTA, 50 mM NaF and was completed at the last moment with 1 mM dithiothreitol, 10 mM sodium molybdate, 0.1 mM sodium orthovanadate, and a mix of protease inhibitors (Calbiochem) diluted 1/50 in Buffer A. Buffer B is the same as Buffer A with 400 mM NaCl. Complementary oligonucleotides of Po promoter 176/Ϫ151 segment (see Fig. 6D) were 5Ј-labeled with [␥-32 P]ATP (10 4 cpm/fmol) and polynucleotide kinase. After annealing, the double-stranded oligonucleotide was purified on a non-denaturing 8% polyacrylamide gel. About 20,000 cpm of the oligonucleotide probe (20 fmol) was incubated with 2 l of nuclear extracts for 20 min on ice in a final volume of 20 l. The incubation buffer final composition was 20 mM Hepes, pH 7.9, 60 mM NaCl, 1 mM dithiothreitol, 5% glycerol, 5 mM MgCl 2 , 0.1 mM ZnCl 2 , and 100 g/ml bovine serum albumin, with, respectively, 2 and 0.1 g of poly(dG)-poly(dC) and poly(dI)-poly(dC) alternate copolymers per assay (Roche Molecular Biochemicals). For competition, 2 pmol of cold wild-type Ϫ176/Ϫ151 probe or oligonucleotides with mutated CACC site were used as follows: 5Ј-TGTGTCCCT- One-hybrid System-The MATCHMAKER one-hybrid system from Clontech was used according to the manufacturer's protocols. Complementary synthetic oligonucleotides containing the CACC box sequence Ϫ174/Ϫ156 of the myelin Po promoter cTGTCCCCCCACCCCCCTACa were placed in four tandem repeats upstream of the pHis-1 and pLAcZ1 plasmids. Target reporter strains of Saccharomyces cerevisiae YM4271 were obtained after transformation with these plasmids and tested for minimal background growth in minimal medium lacking histidine and containing calibrated concentrations of 3-amino-1,2,4-triazole (3-AT). A cDNA library (minimal length 400 bp) was prepared from RNA extracted from F10.9 melanoma cells that had been treated for 48 h with IL6RIL6. The amplified and tailed cDNAs were used for recombinationmediated cloning in yeast with the SmaI-linearized pGADT7-Rec plasmid to form fusion products with the Gal4 activation domain upon transformation into the target reporter yeast strain. Large positive yeast colonies growing in the minimal synthetic dropout (SD) medium lacking histidine and leucine and supplemented with 30 mM 3-AT were selected. Plasmids isolated from the yeast colonies by cloning into E. coli DH5␣ were tested individually in the yeast target reporter strain. Screening was done in the histidine-and leucine-free selective medium for colonies that grew as efficiently in the presence of at least 30 mM 3-AT as in the absence of 3-AT, thereby eliminating false-positive clones. Candidate plasmids were isolated and sequenced from 5Ј and 3Ј ends of the insert. Specific primers were prepared for RT-PCR with RNA from F10.9 cells treated or not with IL6RIL6 (for ZBP-99, forward, 5Ј-GAGGACACATAGTGGAGAAAAGCC-3Ј; reverse, 5Ј-TTTCTACT-GAATAACTATGCATGT-3Ј). Constructs with full-length open reading frames were made in pcDNA3 expression vector and used for transfections as above.

Induction of Myelin Gene Products following Loss of Melanogenesis in Melanoma
Cells-As previously reported (40), treatment of the melanoma F10.9 cells by the gp130 activator IL6RIL6 induces a transition in morphology from epitheliallike cells to elongated cells with extended processes that align to form cell tracks. Concomitantly to the cell shape changes, the IL6RIL6-treated cells stopped releasing melanin pigment in the medium, and the rate-limiting melanogenic enzyme tyrosinase decreased. We have demonstrated (40) that the loss of tyrosinase upon IL6RIL6 treatment results from the decrease in the transcription factor MITF, decrease that is itself mainly because of the loss of Pax3 from the cells. Fig. 1A shows that following the Pax3 RNA down-regulation, there was a strong induction of MBP transcripts in the IL6RIL6-treated melanoma cells. A progressive accumulation of MBP mRNA was observed from 12 to 48 h (Fig. 1B). Transcripts for myelin Po, which is the most abundant component specifically found in the peripheral nerve myelin made by Schwann cells, were strongly induced by IL6RIL6 in the F10.9 cells starting from 24 h (Fig.  2C). The CNPase protein, another constant component of myelin (43), was similarly induced by IL6RIL6 in the F10.9 melanoma cells (Fig. 2D). There was also an induction of PMP-22 and of galactocerebroside (GalC; not shown). The F10.9 melanoma cell response to IL6RIL6 can be defined as a transdifferentiation, because the cells transit from a melanocytic phenotype, which is enhanced if the cells are treated by forskolin (40), to a phenotype characteristic of myelinating Schwann cells.
Transcriptional Activation of Myelin Po and MBP Gene Promoters in IL6RIL6-treated F10.9 Melanoma Cells: Roles of Pax3 and Sox10 -To determine whether the induction of the MBP and Po mRNAs following IL6RIL6 treatment results from transcriptional gene activation, we cloned the 5Ј flanking Ϫ1320/ϩ33 region of the murine MBP proximal promoter (9) in front of the luciferase reporter gene. This region of the promoter confers tissue-specific expression in cell lines and in myelinating oligodendrocytes in vivo (44). Transfection of the F10.9 cells with this reporter gene demonstrated that the MBP gene undergoes a transcriptional activation of about 5-fold in response to the IL6RIL6 stimulus ( Fig. 2A). For the myelin Po gene, the 5Ј flanking sequences Ϫ912 to ϩ45 of the rat gene, which confer specific expression in Schwann cells (6,7), were cloned in the luciferase reporter gene and transfected into F10.9 cells. IL6RIL6 induced 10 -15-fold the activity of this myelin Po promoter construct (Fig. 2B). Because one of the major effect of IL6RIL6 in the transdifferentiating F10.9 cells is the decrease of Pax3 protein (Fig. 2D) (40), we investigated if Pax3 may be involved in the regulation of the myelin Po gene promoter activity, as reported for the MBP gene in primary Schwann cells (20). The F10.9 cells were transfected with a Pax3 expression vector in addition to the MBP and Po promoter reporter genes. The IL6RIL6-dependent induction of the promoter activities for both MBP ( Fig. 2A) and Po (Fig. 2B) was inhibited upon expression of Pax3. Increasing doses of Pax3 cDNA produced parallel reductions in the effect of IL6RIL6 on the myelin Po and MBP gene promoters (Fig. 2C). Thus the repressor effect of Pax3 on the MBP gene is operating in the melanoma F10.9 cells and can now be extended to the peripheral myelin Po gene, as well. Up to 80% inhibition by Pax3 was seen with the Ϫ912/ϩ45 Po promoter construct (Fig. 2B), and a dose-dependent decrease could also be observed with a Ϫ300/ ϩ45 Po promoter construct (data not shown). Inspection of the rat Po promoter sequence does not indicate the presence of Pax3 binding sites, leaving the possibility that Pax3 acts indirectly. In any event, this repression effect on the myelin Po and the MBP promoters strongly suggests that the down-regulation of Pax3 by IL6RIL6 contributes significantly to the induction of the myelin gene expression.
The B16/F10.9 cells express endogenous Sox10, an HMG factor found to be important in both melanocytic (45)(46)(47)(48) and SC differentiation (15,16,49). An increase in Sox10 RNA was observed by RT-PCR with RNA extracted 72 h after IL6RIL6 addition to the cells (Fig. 2E). This induction was ascertained by real-time PCR (Light Cycler), which indicated an increase in Sox10 transcripts of 2-and 5.5-fold, respectively, at 48 and 72 h post-IL6RIL6 addition. The amount of Sox10 protein measured by immunoblots in nuclear extracts of F10.9 cells showed also an increase, which contrasts with the decrease in Pax3 protein (Fig. 2D). We reported before (40) that this increase in Sox10 is not seen during the first day after IL6RIL6 treatment, where Sox10 was even lower than in untreated cultures, but becomes apparent at days 2 and 3 amounting to a 50% increase in the level of Sox10 protein in the treated cell nuclei.
Co-transfections with Sox10 expression vectors showed that Sox10 is positively involved in the activation of the MBP promoter in the F10.9 melanoma cells non-treated with IL6RIL6 (Fig. 3A). In the same experiment, the Pax3 vector again strongly repressed the MBP promoter. Induction by Sox10 was observed with the three MBP constructs tested, including the shortest construct containing only the proximal Ϫ105/ϩ33 fragment of the MBP promoter (Fig. 3A). A further shortened construct, Ϫ70/ϩ33, failed, however, to respond to either Sox10 or IL6RIL6 (not shown). These experiments show that MBP transcription in F10.9 cells is dependent on Sox10, as found very recently (50) in oligodendrocytes by in vivo and in vitro experiments. Likewise, transfection of Sox10 into the F10.9 melanoma stimulated the Po promoter transcriptional activation in a dose-dependent manner in the absence of IL6RIL6 (Fig. 3B). At high ratios of Sox10 over the Po reporter the activation was close to that produced by IL6RIL6. A truncated form of Sox10 (E188X) was not comparably active (not shown). In the cells treated by IL6RIL6, Sox10 did not increase the Po promoter activity (Fig. 3B) and did not increase the MBP promoter activity (not shown). Hence, in these cells, Sox10 appears to be a limiting factor in the activity of the MBP and Po gene, unless the cells have been treated by the gp130 activator.
Modulation of the Concentrations of Pax3 and Sox10 in Stably Transfected F10.9 Cells Regulates the Level of Cellular Myelin Po mRNA-To further evaluate the contributions of Pax3 and Sox10 on the induction of Po mRNA, the F10.9 cells were permanently transformed with vectors containing Pax3 and Sox10 cDNAs placed under the control of a tetracyclineregulated element. The levels of Pax3 and Sox10 transcripts were measured by semiquantitative RT-PCR in the cells treated by doxycycline or IL6RIL6 or both (Fig. 4A), and in parallel we measured the amount of myelin Po mRNA. In cells that overexpressed Pax3 when treated with doxycycline, the level of Po RNA induced by IL6RIL6 was decreased by about 50% as compared with the level induced in the absence of doxycyclin (Fig. 4B). The basal level of Po RNA was also very  low in the Pax3 expressers. Cells harboring a Tet-regulated enhanced green fluorescent protein control vector had the same endogenous Po mRNA induction with or without doxycycline (noted as 100% on the graph in Fig. 4B). These results confirm that the induction of the Po mRNA by IL6RIL6 is impaired if Pax3 is artificially re-increased in the IL6RIL6-treated cells, in line with the hypothesis that the down-regulation of Pax3 by the gp130 activator accounts for a significant part of the accumulation of the myelin gene product.
Conversely, with a clone that overexpressed ectopic Sox10 in response to doxycycline (Fig. 4A), we observed an increase of Po mRNA (Fig. 4B). However, under conditions where the increase in total Sox10 RNA caused by doxycycline was comparable with that in IL6RIL6-treated cells (Fig. 4A), the increase in Po was less than the one induced by IL6RIL6 treatment (Fig. 4B). Thus, the increase in Sox10 caused by IL6RIL6 could account for part, but probably not all, of the induction of the Po mRNA by IL6RIL6 treatment. This conclusion was confirmed when we next examined the Po promoter sequences required for the transcriptional activation by IL6RIL6.
Sequences Required for the Activation of the Po Promoter by IL6RIL6: Involvement of Sox10 Sites and of a CACC Box-The rat Po promoter sequence mediating tissue-specific expression of Po contains a number of binding sites for known transcription factors (8,16), some of which are indicated in Fig. 5A. Po promoter constructs with sequences from Ϫ912, Ϫ500 or Ϫ300 bp to ϩ45 all gave 10 -15-fold activation by IL6RIL6 (Fig. 5B). Further 5Ј deletions from Ϫ300 to Ϫ136 resulted in a progres-sive loss of induction by IL6RIL6, most of the decrease occurring between Ϫ222 and Ϫ137 (Fig. 5C). By further deletion from Ϫ136 to Ϫ63, the basal activity of the promoter decreased steeply probably because of the loss of the two CAAT boxes (Fig. 5, A and C). Several 3Ј deletions were made and inserted upstream of a minimal TK promoter. As shown in Table I, a TK construct containing the Po Ϫ300 to Ϫ137 sequence was highly inducible by IL6RIL6, although its overall activity was lower than the Ϫ300/Ϫ30 construct. The Ϫ222/Ϫ137 segment of the Po promoter (Fig. 5A, dotted lines) contains the Sox10 sites C (Ϫ204/Ϫ198), CЈ (Ϫ193/Ϫ187), and B (Ϫ147/Ϫ140). These sites, as characterized by Peirano et al. (16), bind bacterially produced Sox10 and mediate the rat Po promoter activation in cells stimulated to express Sox10 by doxycycline. We therefore examined the effect of mutating these sites as part of a series of linker scanning mutations in the Ϫ300 to Ϫ137 region (relevant mutants indicated in Fig. 5D). Mutation of the Sox10 site CЈ by itself caused a 77% reduction in the IL6RIL6-induced activity of the promoter and markedly reduced the -fold induction (Table I). Mutation of the Sox10 C and B sites also reduced the induction by IL6RIL6, the reduction being greater when both sites were mutated (Table I). However, we found that additional mutations located between the Sox10 CЈ and B sites also affected the activity of the Po promoter and its inducibility by IL6RIL6. In particular, mutating the C/G-rich Sp1-like site or CACC box at Ϫ168/Ϫ161 (mutation of CCCACCCC; see Fig.  5D) consistently reduced the activity and response to IL6RIL6 in the Po Ϫ300/ϩ45, Po Ϫ222/ϩ45, and Po TK Ϫ300/Ϫ137 constructs (Table I). The sequences at and around Ϫ168/Ϫ161 do not resemble a Sox10 site, indicating that the induction by IL6RIL6 is not only because of an effect on Sox10.
The induction of Po promoter activity by ectopic Sox10 was, as expected, abrogated by mutations of the Sox10 sites B and C or mutation of site CЈ (Fig. 6A). Interestingly, the Ϫ168/Ϫ161 CACC box mutation decreased the induction by ectopic Sox10 as it did for the IL6RIL6 inducer (Fig. 6A, right panel). Changing just the central CAC (Ϫ165/Ϫ163) of this element into GGG also reduced the activity of the Po promoter in response to IL6RIL6 and Sox10 (Table I), as did a mutation of Ϫ176/Ϫ169, which shortens the C/G-rich arm (Fig. 6A). These results strongly suggest a model in which Sox10 acts with one or more cooperative factor binding to the CACC box and/or its C/G-rich arms.
A Kruppel-type Factor Is Involved in the Function of the Ϫ168/Ϫ161 CACC Box in the Po Promoter-Electrophoretic mobility shift assays with a Ϫ175/Ϫ151 probe demonstrated binding of two slow migrating complexes (A and B) and one fast migrating complex (C), which were specifically competed by the unlabeled probe but not when that probe had the Ϫ168/Ϫ161 mutation destroying the CACC box (Fig. 6B, lanes 3 versus 4). Using probes with the Ϫ165/Ϫ163 CAC mutated to GGG or TTT also affected the formation of these complexes (not shown). Although the complexes were seen with nuclear extracts of untreated cells, IL6RIL6 produced a significant change in the pattern of complex formation, with a consistent increase in complex C but a decrease in complex A (Fig. 6B, lanes 1 versus 2).
In search of a factor that could act through this element of the Po promoter, we used a one-hybrid selection approach in yeast. A cDNA library prepared from IL6RIL6-treated F10.9 cells was screened for clones encoding a protein that binds a target reporter construct containing the Ϫ174/Ϫ156 sequence of the Po promoter. Screening of about 0.5 ϫ 10 6 clones, followed by re-screening under stringent conditions of yeast growth (see "Materials and Methods"), identified a clone with a 3.6-kb insert. This clone was found by sequencing to contain the entire open reading frame of a cDNA designated ZBP-99 (accession number XM_129403), and RT-PCR indicated that its corresponding mRNA was significantly increased following treatment of the F10.9 cells by IL6RIL6 (Fig. 7A).
ZBP-99 (797 amino acids) has been described (51) as a member of a class of transcription factors that contain a characteristic array of four Kruppel-type zinc fingers and bind to CACC boxes, such as that in the gastrin gene promoter whose sequence and length (CCCCCCACCCCGCCCC) resembles that of the Ϫ171/Ϫ161 element in the Po promoter (see Fig. 5D). Although ZBP-99 was described as an inhibitor of transcription (51), we found that it stimulates the myelin Po promoter activ-ity (Fig. 7B). When expressed in the F10.9 melanoma cells, the full-length ZBP-99 stimulated by itself the transcriptional activity of the Po-300 promoter, whereas a truncated form of ZBP-99 had no effect (Fig. 7B). The combination of Sox10 with ZBP-99 produced a 10.2-fold stimulation of the Po promoter, above the additional effect of each factor alone, and similar to that seen following IL6RIL6 treatment of the cells (Fig. 7B). The truncated ZBP-99, lacking its C-terminal domain, was inactive or slightly inhibitory even with Sox10 or IL6RIL6.
FIG. 6. The CACC box (؊168/؊161) affects IL6RIL6-and Sox10induced Po promoter activity. Sequence-specific factor attachment to this CACC box is modulated by IL6RIL6. A, F10.9 cells co-transfected with the Po promoter Ϫ300/ϩ45 luciferase construct or different Po promoter mutants (0.5 g), with the expression vector pcDNA3-Sox10 (left panel, Sox10/Po-luciferase plasmids ratio 5:1; right panel, 2:1). The luciferase activity in cells treated by IL6RIL6 (normalized as in Fig. 2A) was taken as 100 (black bars). Empty bars, no IL6RIL6; point-filled bars, with Sox10; oblique stripes, with Sox10 and IL6RIL6 treatment. B, binding of sequence-specific factors to the CACC box in F10.9 extracts, demonstrated by electrophoretic mobility shift assays. Oligonucleotide representing sequence Ϫ176/Ϫ151 of the Po promoter, 32 P-end labeled, was incubated with nuclear extracts of F10.9 cells treated with IL6RIL6 (300 ng/ml) or non-treated for 48 h, as indicated. The first two lanes show induction of complex C and reduction of complex A with IL6RIL6 treatment. The next two lanes show specific binding to the CACC box, complexes A, B, and C disappearing upon competition with a 100-fold excess of cold probe but persisting when CACC box sequence Ϫ168/Ϫ161 in the cold competitor is mutated (mutation indicated in Fig. 5D).  Fig. 5D.
Mutation of the Ϫ168/Ϫ161 CACC box in the Po Ϫ300/ϩ45 promoter (designated as P1 mutation) strongly reduced the activity of ZBP-99, indicating the functional role of this CACC box for ZBP-99 action, as seen for the IL6RIL6-dependent response (Fig. 7C). In addition, the effect of ZBP-99 was also abolished when the Sox10 B and C sites in the Po promoter were mutated, supporting a cooperative action between the CACC box factor and Sox10 in the activation of the Po promoter (Fig. 7C).
Krox20 Also Synergizes with Sox10 to Activate the Myelin Po Promoter but Does Not Act through the Ϫ168/Ϫ161 Element-In addition to ZBP-99, which is positively involved in the activity of the Po promoter and increases after IL6RIL6 treatment, other proteins appear to bind to the Ϫ175/Ϫ171 probe and contribute to the three DNA-protein complexes observed. DNA-protein UV cross-linking experiments (not shown) indicated that other proteins are present in complexes A and B than in complex C, which increases after IL6RIL6. Because Krox-10/Egr-2 has been shown to be required for myelination in SC (17) and binds to C/G-rich Sp1-like sequences (52), we examined whether it may be one of the proteins acting on the Ϫ168/Ϫ161 element, possibly through the C/G-rich arms of this CACC box.
Ectopic expression of Krox20 in the F10.9 cells induced the Po-300 promoter by itself (Table II). Moreover, a marked synergistic effect was observed when both Krox20 and Sox10 were expressed together in the melanoma cells (Table II). The importance of Sox10 for the response to Krox20 was further seen when Sox10 sites B and C were mutated (Table III). However, when the Ϫ168/Ϫ161 sequence was mutated, the stimulation by Krox20 was still observed, as well as the Krox20-Sox10 synergism, although the effect of Sox10 was reduced (Table II). Thus, the site of Krox20 action must be elsewhere. Indeed, mutation of the upstream CCCCACCCC element (Ϫ247/Ϫ240; see Fig. 5D) reduced the effect of Krox20 more than those of Sox10 or IL6RIL6 (Table III). Another C/G-rich element, GGGGGAGG (Ϫ220/Ϫ213), also affected the response to Krox20 with less effects on the other inducers (Table III). Krox20 may, therefore, work through several distinct upstream sites. The mutation analysis further showed that the effect of IL6RIL6 on Po promoter activity correlates less with the effect of Krox20 than with the response to Sox10. We also did not find that IL6RIL6 treatment significantly changed the level of Krox20 RNA already present in these F10.9 cells (not shown). Instead, the Ϫ168/Ϫ161 CACC box appears to be the main element required, in addition to its surrounding Sox10 binding sites, for the activation of the Po promoter by IL6RIL6, as well as by Sox10 in the F10.9 melanoma cells. DISCUSSION The gp130 activator IL6RIL6 activates myelin gene expression in embryonic Schwann cells (36,37). We show here that IL6RIL6 induces expression of the myelin genes in the murine B16/F10.9 melanoma cells. These cells undergo a transdifferentiation, with loss of their melanocytic phenotype characterized by MITF and tyrosinase expression (40) and acquisition of a myelinating Schwann cell phenotype. Melanoma and nevus cells can exhibit such phenotypic changes in vivo (53,54). The inducible transdifferentiation of the F10.9 cells may recapitulate molecular switches that operate when neural crest-derived progenitors develop into either melanocytic or glial lineages (55,56) and may help identify them. We investigated here mechanisms involved in the transcriptional activation of promoters from two myelin genes, Po and MBP. One of the switches that seems to affect both MITF and myelin gene promoters is a profound down-regulation of Pax3 observed following IL6RIL6 treatment. The Pax3 RNA was decreased at 12 h after IL6RIL6, followed by a decrease in the protein (40). Part of this effect may be on stability of Pax3 RNA, because its half-life decreased from 6 to 3 h, 1 day after IL6RIL6 (not shown). Pax3 is an activator of the MITF gene (57), and we have demonstrated that the loss of Pax3 caused by IL6RIL6 is responsible for the silencing of the MITF promoter, because its activity can be restored by ectopic expression of Pax3 (40). As shown here, expression of Pax3 in the same F10.9 cells conversely represses the ability of IL6RIL6 to activate the promoters of the myelin Po and MBP genes. An inhibition of MBP gene induction by Pax3 had been described in SC and in neuroblastoma (20). This is the first report of a similar repression of the Po gene promoter. It is of interest that the down-regulation of Pax3 by the gp130 activator IL6RIL6 is also observed in cultures of embryonic DRG and derived SC, in which Po and MBP gene products are induced (37). Because the decrease in Pax3 is a constant feature of terminal differentiation of myelinating SC (5), this may be an important mechanism by which gp130 signaling enhances myelin gene expression.
Pax3 acts in synergy with Sox10 to activate the MITF promoter, and mutations in any of these three genes cause some form of Waardenburg syndrome with melanocytic deficiencies (45)(46)(47)(48). Contrasting with the strong decrease in Pax3 caused by IL6RIL6 in F10.9 cells, the Sox10 protein is rather increased after IL6RIL6 treatment for 2-3 days. It may seem paradoxical that IL6RIL6, which inhibits melanocytic differentiation of the cells, causes an increase in Sox10, a factor needed for melanogenesis (45-48). However, we have shown that, in IL6RIL6treated F10.9 cells, Sox10 overexpression does not restore MITF and does not synergize with Pax3 as it does in the FIG. 7. CACC box-specific factor ZBP-99 stimulates Po promoter activity and is increased by IL6RIL6. A, RT-PCR with primers for ZBP-99 on non-saturating amounts (0.25 g/ml) of total cell RNA from F10.9 cells, either untreated (Ϫ) or treated (ϩ) for 24 h with IL6RIL6 (140 ng/ml). G3ЈPDH amplification is shown as control of equal RNA loading. B, F10.9 cells co-transfected with Po promoter Ϫ300/ϩ45 luciferase construct (0.3 g) and pcDNA3-ZBP99, 3 g, fulllength (black bars) or truncated after amino acid 599 (gray bars), and 0.6 g of pcDNA3 carrier. In control group (white bars), cells were transfected with 3.6 g of pcDNA3 only. Where indicated, cells were co-transfected with pcDNA3-Sox10, 0.6 g, and 3 g of pcDNA3 (white bars) or 3 g of pcDNA3-ZBP99 or truncated ZBP99 (black and gray bars, respectively). Cells were either non-treated (NT) or IL6RIL6treated. C, transfections with pcDNA3-ZBP99 comparing Ϫ300/ϩ45 Po promoter constructs with either wild-type sequence, with the Ϫ168/ Ϫ161 CACC box mutation (P1 mut), or with Sox10 site B and site C mutations (Sox10mut). Luciferase activity was expressed as in Fig. 2A. untreated F10.9 cells (40). This suggests that the lowering of Pax3 relative to Sox10, and the deregulation of their synergistic interaction, prevent Sox10 from activating MITF and melanogenesis. It may also free Sox10, making it available for acting on other genes. The repression effect exerted by Pax3 on the Po Ϫ912/ϩ45 promoter occurs without apparent Pax3 DNA binding site and may be indirect, possibly because of proteinprotein interactions between Pax3 and factors (such as Sox10) acting on the myelin gene promoters.
Increasing the levels of Sox10 in F10.9 cells transfected by Tet-regulated expression plasmids led to an induction of Po mRNA. However, this induction was always lower than that produced by IL6RIL6. This is consistent with many lines of evidence that Sox10 does not work alone in the regulation of myelin gene expression. Thus, in embryonic SC precursors, Sox10 is present long before myelin gene products start to be made (49) indicating that other changes are needed. Although Sox10 activates the transcriptional activity of the Po and MBP promoters in the F10.9 melanoma cells, as it does in N2A neuroblastoma cells (16,50), Sox10 fails to activate MBP promoters in HeLa cells (58). This suggests that HeLa cells lack a co-factor for Sox10 or contain an inhibitor. In general, proteins of the Sox family are known to affect DNA topology (DNA bending) and to require partner factors to activate transcription (59). Such partners of Sox10, besides Pax3, are, for example, SCIP/Oct-6 (49) and Sp1 (60). SCIP, like Pax3, is an inhibitor that is down-regulated before SC activate myelin gene expression (5). For Sp1, a binding site at Ϫ55/Ϫ50 in the rat Po promoter was reported as required for the basal Po promoter activity (8), but this site is not in the region that we find essential for the response to IL6RIL6. We, therefore, searched in the Ϫ300/Ϫ137 region of the Po promoter for sequence elements that are positively involved in the response to IL6RIL6 and to Sox10. Our analysis indicates that activation by IL6RIL6 and by Sox10 requires the Sox10 binding sites C (Ϫ204/Ϫ198), CЈ (Ϫ193/Ϫ187), and B (Ϫ147/Ϫ141) defined by Peirano et al. (16) but requires in addition other critical transcriptional elements around these sites.
One such element is a long CACC box, which is functionally defined by the Ϫ168/Ϫ161 mutation inhibiting the Po promoter response to IL6RIL6, as well as to Sox10. Several DNA-protein complexes are formed on a probe containing this CACC box sequence, and some of the complexes are increased following IL6RIL6 treatment whereas other may decrease. We identified a protein that functionally interacts with this CACC box sequence used as cis-acting transcriptional element in a onehybrid yeast selection system, as transcription factor ZBP-99. The mRNA for ZBP-99 is increased in the F10.9 cells following treatment with IL6RIL6. ZBP-99 had been isolated previously as binding a long CACC box from the gastrin gene promoter and shown to belong to a novel family of proteins defined by an array of four Kruppel-type zinc fingers (51). ZBP-99 was reported to inhibit transcription of the Ornithine decarboxylase gene similarly to its homologous factor ZBP-89/BFCOL1 (51,61). However, ZBP-89 has also been shown to activate transcription of the p21/waf1 gene through cooperation with histone acetyltransferase p300 in HT29 cells (51). Transfection of the F10.9 cells with ZBP-99 stimulated the activity of the Po promoter, and its action was potentiated when cotransfected with Sox10. Moreover, ZBP-99 was not active on a Po promoter in which Sox10 sites B and C were mutated. These results demonstrate that a protein, acting through the CACC box, stimulates Po promoter activity and mediates a cooperative action of Sox10 with the CACC box element. The increased expression of the Kruppel-type factor ZBP-99 that is observed in the IL6RIL6-treated melanoma cells may be one of the mechanisms activating the Po promoter in these cells. However, complex interactions with other transcription factors and regulatory proteins are likely to be involved in the function of the CACC box and its interplay with the Sox10 sites.
In addition to Pax3, Sox10 and ZBP-99 we examined the role of Egr-2/Krox20 in the activation of the Po gene in F10.9 melanoma cells. Krox20 is required for the onset of myelination during SC development (17), and it controls a large array of SC genes including Po/myelin protein zero (19). Activation of the Po promoter by Krox20 was seen in SC (18) but not in neuroblastoma (16), and synergism of Krox20 and Sox10 was re- a Relative to non-treated without addition. b Mutant indicated in Fig. 5D. a Type of site mutated. For actual mutated sequences see Fig. 5D. b Activity or -fold induction of mutants in percent of the wild-type Po Ϫ300/ϩ45 luciferase reporter construct, in IL6RIL6-treated cells or without IL6RIL6 but with Sox10 or Krox20 co-transfection. ported on the Connexin-32 promoter (57). In the F10.9 cell system, Krox20 expression activates the Po promoter and exhibits a strong synergism with Sox10. The effect of Krox20 was not related to the Ϫ168/Ϫ161 CACC box, but we identified other G/C-rich sequences that may mediate its action on the Po promoter. The untreated F10.9 cells already contain Krox20 mRNA, and IL6RIL6 did not significantly increase its level, although other types of modulations affecting Krox20 functions are not excluded. In embryonic neuroglial cell cultures from DRG of rat E14 embryos, we did find 3 that IL6RIL6 causes a strong increase in Krox20 mRNA and protein, in correlation with the induction of MBP and Po mRNAs.
There is growing evidence for a major role of the IL-6 cytokine family in the synthesis of myelin. In vivo, the inactivation of the IL-6 family gp130 receptor in newborn mice was shown to impair nerve myelination (38). Using the potent gp130 activator IL6RIL6, we reported in vivo stimulation of myelination in regenerating sciatic nerves (37) and in vitro activation of myelin gene expression in premyelinating Schwann cells from embryonic DRG (36,37). The present melanoma cell system provides a new model to investigate the multiple mechanisms that activate myelin gene transcription in response to the IL-6-type stimulus.