Disruption of Meox or Gli Activity Ablates Skeletal Myogenesis in P19 Cells*

Gli2 and Meox1 are transcription factors that are expressed in the developing somite and play roles in the commitment of cells to the skeletal muscle lineage. To further define their roles in regulating myogenesis, the function of wild type and dominant-negative forms of Gli2 and Meox1 were examined in the context of differentiating P19 stem cells. We found that Gli2 overexpression up-regulated transcript levels of Meox1 and, con-versely, Meox1 overexpression resulted in the up-regulation of Gli2 transcripts. Furthermore, dominant-negative forms of either Meox1 or Gli2 disrupted the ability of P19 cells to commit to the muscle lineage and to properly express either Gli2 or Meox1, respectively. Finally, Pax3 transcripts were induced by Gli2 overexpression and lost in the presence of either mutants Meox1 or Gli2. Taken together, these results support the existence of a regulatory loop between Gli2, Meox1, and Pax3 that is essential for specification of mesodermal cells into the muscle lineage. Skeletal muscle development occurs in the somite of the developing embryo. Somites give rise to the sclerotome, dermomyotome, and myotome. Differentiation of epaxial muscle in the myotome is controlled by the myogenic regulatory factor family of muscle-specific transcription factors, including MyoD, Myf-5, myogenin, and MRF4 (1,

Gli2 and Meox1 are transcription factors that are expressed in the developing somite and play roles in the commitment of cells to the skeletal muscle lineage. To further define their roles in regulating myogenesis, the function of wild type and dominant-negative forms of Gli2 and Meox1 were examined in the context of differentiating P19 stem cells. We found that Gli2 overexpression up-regulated transcript levels of Meox1 and, conversely, Meox1 overexpression resulted in the upregulation of Gli2 transcripts. Furthermore, dominantnegative forms of either Meox1 or Gli2 disrupted the ability of P19 cells to commit to the muscle lineage and to properly express either Gli2 or Meox1, respectively. Finally, Pax3 transcripts were induced by Gli2 overexpression and lost in the presence of either mutants Meox1 or Gli2. Taken together, these results support the existence of a regulatory loop between Gli2, Meox1, and Pax3 that is essential for specification of mesodermal cells into the muscle lineage.
Skeletal muscle development occurs in the somite of the developing embryo. Somites give rise to the sclerotome, dermomyotome, and myotome. Differentiation of epaxial muscle in the myotome is controlled by the myogenic regulatory factor family of muscle-specific transcription factors, including MyoD, Myf-5, myogenin, and MRF4 (1,2). Signaling factors from tissues surrounding the somite, including Wnts and Sonic hedgehog (Shh), 1 direct the processes of somite patterning and subsequent skeletal muscle development in the myotome (3)(4)(5)(6). Several transcription factors appear to mediate the response to signals from surrounding tissues and regulate the commitment of cells into the muscle lineage. These include Gli2 (7,8), Pax3 (9 -11), and Meox1/2 (12,13).
Wnt and Shh cooperate in the activation of the Gli genes in avian somite formation (8). The Drosophila homolog of the Gli family of zinc finger transcription factors, cubitus interruptus, is a downstream effector of Shh signaling in the fly (14). The Gli family, including Gli1, Gli2, and Gli3, are all expressed in the developing somite (15). Further analysis in avian embryos indicate that Gli1 is expressed predominantly in the ventromedial somite, whereas Gli2 and Gli3 become restricted to the dorsomedial somite, which gives rise to the epaxial myotome (8). Their expression pattern suggests that the Gli factors may play a role in the commitment of cells to the skeletal muscle pathway. Indeed, the Gli factors regulate the maintenance of Myf-5 expression in the early somite, although initial activation is independent of Shh function and an intact Gli binding site (7,16,17). The Glis also play a role in MRF expression and skeletal muscle development in Zebrafish (18,19) and Xenopus (20). Together, these studies suggest a conserved role for the Gli factors in regulating the expression of the myogenic regulatory factors, leading to skeletal muscle development.
Pax3 and Pax7 are functionally similar members of a family of paired box transcription factors (21). Pax3 was found to be initially expressed throughout the somite before becoming restricted to the dermomyotome and the cells that migrate to the limb bud (22,23). The splotch mouse carries a mutant allele of the Pax3 gene (24). Splotch mutant embryos were devoid of muscle precursors that migrate from the dermomyotome to populate the limb bud, whereas epaxial muscle appeared unaffected (22,25). Because Pax7 expression was shown to be expanded into regions that normally express only Pax3 in Splotch mutant embryos (26), it is likely that Pax7 compensated for the loss of Pax3 in epaxial muscle.
A cross between the Splotch mouse and the Myf-5-null mouse implied a role for Pax3 in epaxial muscle, because both epaxial and hypaxial muscle precursors were lost in double mutant embryos (11). The lack of MyoD in the double mutant indicated that MyoD is regulated by both Pax3 and Myf-5. Overexpression of Pax3 in avian paraxial mesoderm induced MyoD, Myf-5, and myogenin expression (10). Furthermore, overexpression of a dominant-negative Pax3 in P19 embryonal carcinoma cells resulted in a loss of MyoD and myogenin expression and subsequent myogenesis (9). Together these studies indicate a crucial role for Pax3 in activating the expression of the myogenic regulatory factors in the myotome.
Meox1 and Meox2 are closely related homeobox proteins expressed in the early developing somite (27). At later stages, Meox1 was found in the dermomyotome, whereas Meox2 was found in the developing limb bud (27,28) (12), consistent with a role for Meox2 in hypaxial muscle development. Interestingly, Meox factors have been shown to interact with members of the Pax family (29), although the effect of this interaction on activity of Meox and Pax during muscle development is not yet known.
We have shown previously that Wnt signaling via ␤-catenin is essential and sufficient for the expression of Pax3, Meox1, and Gli2, and the subsequent induction of skeletal myogenesis in aggregated P19 cells (13). Furthermore, Pax3 expression is essential and sufficient for the expression of the transcription factor Six1 and the induction of skeletal myogenesis (9). Interestingly, Pax3 did not activate the expression of Gli2 and Gli2 expression was not lost in cell lines expressing a dominantnegative Pax3 mutant. Conversely, Meox1 expression was induced by Pax3 overexpression and down-regulated in the absence of functional Pax3. These results suggest a complex relationship between Gli2, Pax3, and Meox1.
To study the roles of Gli2 and Meox1 in skeletal myogenesis, we examined the function of wild type and dominant-negative forms of Meox1 and Gli2 in the context of aggregated P19 cells. Gli2 was shown to up-regulate the expression of Meox1, Pax3, and the MRFs leading to low levels of skeletal muscle development. Dominant-negative Gli2, termed Gli/EnR, inhibited skeletal myogenesis and resulted in the down-regulation of Meox1, Pax3, and MRF expression. Meox1 activated the expression of Gli2 but could not induce Pax3 expression or skeletal myogenesis. However, dominant-negative Meox1, termed Meox/EnR, down-regulated Pax3 and Gli2 expression and inhibited skeletal myogenesis. Together, these data indicate that overexpression of Gli2, but not Meox1, is sufficient to induce low levels of skeletal myogenesis, whereas either Meox or Gli factors, or genes with Meox or Gli binding sites, are essential for skeletal myogenesis in P19 cells.

MATERIALS AND METHODS
Plasmid Constructs-Gli1 and Gli2 cDNA, both driven by the CMV promoter, were kindly provided by H. Sasaki (30). To construct the Gli/EnR chimera, the C-terminal domain of Gli2, which includes the activation domain, was removed by utilizing ApaI. This deletion corresponds to the ⌬C4 mutant that could not activate a multimerized Gli-site promoter in transfection assays (30). The remaining N terminus of Gli2, containing the zinc finger domain, was fused to the 198-amino acid N-terminal repression domain of the mouse EN-2 protein. The engrailed repression domain was amplified by PCR utilizing oligonucleotides 5Ј-AAGGGCCCGGGAGGAGAAGGATTCCAAGCCC and 3Ј-AAGCGGCCGCCTAGCCCAGAGTGGCGCTGGCTT. The 5Ј and 3Ј oligonucleotides contained ApaI and NotI sites, respectively (bold), to facilitate cloning and the 3Ј-oligonucleotide contained a stop codon (underlined). Gli/EnR was subcloned into the PGK vector, which contains the phosphoglycerate kinase (pgk-1) promoter (31). The PGK-Puro and PGK-LacZ constructs were described previously (32).
Meox1 cDNA was kindly provided by B. Mankoo. The Meox1 cDNA was subcloned into the PGK vector using SacI/KpnI. To construct the Meox/EnR chimera, Meox1 cDNA was amplified utilizing oligonucleotides 5Ј-AAGAGCTCCAGCAGATGGATCCAGTG and 3Ј-AAGGTACCG-ATATCCTCTGAACTTGGAGAAGCTGC and subcloned into pGEM® T-Easy (Promega) vector (the start codon is underlined). The Meox1 cDNA was cut out using SacI/KpnI and subcloned into the PGK-vector by utilizing SacI/KpnI sites. The engrailed repression domain was amplified by PCR utilizing oligonucleotides 5Ј-AAGATATCGAGGAG-AAGGATTCCAAGCCC and 3Ј-AAGATATCCTAGCCCAGAGTGGCG-CTGGCTT. The 5Ј-and 3Ј-oligonucleotides contained EcoRV sites (bold) to facilitate cloning and the 3Ј-oligonucleotide contained a stop codon (underlined). The 198-amino acid N-terminal repression domain of the mouse EN-2 protein was then fused to the C terminus of the Meox1 cDNA by subcloning the amplified engrailed repression domain into the PGK-Meox1 vector by utilizing EcoRV. The activation domain of Meox1 is currently unknown and thus was not removed prior to construction of the Meox/EnR chimera.
P19(control), P19(Gli2), and P19(Meox1) cells were differentiated in the absence of Me 2 SO and P19(control), P19(Gli/EnR), and P19(Meox/ EnR) cells were differentiated in the presence of 0.8% Me 2 SO, as described (9, 32, 34 -37, 39). Differentiation involved aggregation of cells for 4 days in Petri dishes (in the absence or presence of Me 2 SO) followed by plating into tissue culture dishes on day 4. Cells were harvested for total RNA on day 9. At least three clonal populations, which behaved similarly, were isolated for each cell line.
Immunofluorescence-Cells were plated onto gelatin-coated coverslips on day 4 and fixed on day 9 in Ϫ20°C methanol. Cells were rehydrated in phosphate-buffered saline and myosin heavy chain expression (MHC) was detected utilizing monoclonal MF20 antibody supernatant (40) and nuclei were detected by Hoechst stain as described previously (9). In brief, 40 l of MF20 supernatant in 40 l of phosphate-buffered saline was incubated on coverslips for 1 h at room temperature. Coverslips were then washed 3ϫ 5 min in phosphatebuffered saline, and incubated with 80 l of 1:100 dilution of goat anti-mouse Cy3-linked antibody (Jackson ImmunoResearch) for 1 h at room temperature. After phosphate-buffered saline washes, coverslips were mounted and immunofluorescence was visualized with a Zeiss Axioskop microscope. Images were captured on a Sony 3CCD camera and processed utilizing Axiovision, Adobe Photoshop 7, and Canvas 8 software.
Northern Blot Analysis-Total RNA was isolated from each differentiation utilizing the LiCl method, as described (32). Twelve g of total RNA were separated on a 1% agarose, formaldehyde gel, transferred onto Hybond-N (Amersham Biosciences) by capillary action and crosslinked by UV irradiation. Blots were then hybridized to DNA probes labeled with [␣-32 P]dNTP by multiprime labeling (Amersham Biosciences) for 16 h at 42°C. Probes for Pax3, Meox1, Six1, MyoD, myogenin, and 18s (for standarization of loading) have been described previously (9). The Gli2 probe consisted of an EcoRI-ApaI N-terminal fragment of mouse Gli2 cDNA, provided kindly by H. Sasaki (30). Northern blots were then washed 5ϫ 5 min in low stringency wash (2ϫ SSC, 0.2% SDS) at 42°C and for 15 min in high stringency wash (0.1ϫ SSC, 0.2% SDS) at 65°C. Blots were visualized by autoradiography. Figures shown were spliced from the same autoradiogram.
Reverse Transcription-Polymerase Chain Reaction-Total RNA was harvested on day 9 utilizing TRIzol reagent, as per the manufacturer's protocol (Invitrogen), or the LiCl method, followed by purification with RNeasy® Mini Kit (Qiagen, Mississauga, Ontario, Canada). Approximately 1 g of DNase I-treated RNA was used to synthesize first strand DNA, utilizing the SuperScript First-strand Synthesis kit (Invitrogen). Random hexamers (50 ng) were utilized as per the manufacturer's protocol. Two l of first-strand DNA was then used for PCR. Oligonucleotides to amplify Myf-5 were 5Ј-GAGCTGCTGAGGGAACAGGTG-GAGA and 3Ј-GTTCTTTCGGGACCAGACAGGGCTG at an annealing temperature of 68°C. The oligonucleotides and PCR conditions for myogenin and tubulin have been described previously (37). First-strand reactions were tested for linearity with each set of oligonucleotides, and negative controls included RT experiments in the absence of RNA or reverse transcriptase enzyme and PCR in the absence of first strand reaction. Southern blot analysis was carried out to analyze RT-PCR products utilizing probes for myogenin and tubulin as described previously (37) and a full-length cDNA fragment of mouse Myf-5.
Myogenic Conversion Assay-C3H 10T1/2 fibroblasts were grown in 10% 1:1 cosmic calf-fetal calf bovine serum in ␣-minimum Eagle's medium. 10T1/2 cells were transfected on gelatin-coated coverslips utilizing the FuGENE 6 transfection reagent. Myogenic conversion was induced by 1.4 g of PGK-MyoD. To determine whether Gli2, Meox1, or Gli2 could induce myogenic conversion, 10T1/2 cells were transfected with 1.4 -3 g of CMV-Gli2, PGK-Meox1, or PGK-Pax3. All transfections included 0.5 g of pEGFP-N1 (Clontech Laboratories, Inc., Palo Alto, CA) for transfection efficiency. Total DNA in each transfection was brought up to 4 g with PGK vector plasmid. Twenty-four hours after transfection, cells were transferred to differentiation media containing 2% horse serum. Transfection efficiency was calculated by counting green fluorescent protein-positive cells 2-3 days after transfection. Six days after transfection, cells were fixed and stained for the presence of myosin heavy chain by immunofluorescence with MF20. Myogenic conversion was quantitated by counting cells expressing myosin heavy chain, normalized to transfection efficiency.

Gli2 and Meox1 Activate the Expression of Each Other in
Aggregated P19 Cells-To examine the ability of Gli2 to induce skeletal myogenesis, cell lines stably expressing Gli2 (termed P19(Gli2) cells) were isolated. P19(Gli2) cells and control cells were aggregated in the absence of Me 2 SO. Under these conditions, control cells do not differentiate into skeletal muscle, as shown by the lack of MHC-expressing cells in these cultures on day 9 (Fig. 1B). In contrast, P19(Gli2) cells did differentiate into skeletal muscle, as shown by the presence of bipolar MHCexpressing myocytes (Fig. 1D), representing Ͻ4% of these cultures. Therefore, Gli2 can induce low levels of skeletal myogenesis in aggregated P19 cells.
To identify the factors activated by Gli2 expression, total RNA was harvested on days 0 and 9 of differentiation. By Northern blot analysis, Gli2 transcript levels were high in P19(Gli2) cell lines, but not in control cell lines (Fig. 2, panel I,  A). Transcription factors expressed in the dermomyotome of the developing somite, such as Pax3 and Meox1, were activated by day 9 of differentiation in P19(Gli2) cell lines but not control cell lines (Fig. 2, panel I, B and C). Interestingly, Meox1 transcripts were also present in monolayer cultures, suggesting that Gli2 can activate Meox1 expression in the absence of additional factors induced by aggregation (Fig. 2, Panel I, B). Furthermore, expression of Gli1 in aggregated P19 cells induced skeletal myogenesis as well as Meox1 and Pax3 transcript levels (data not shown), similar to the Gli2-induced skel-etal muscle program. Therefore, Gli factors can induce the expression of Pax3 and Meox1, which are factors thought to be involved in the specification of mesodermal precursors to the myogenic lineage.
To determine whether Gli2 overexpression can activate MRF expression in P19 cells, transcripts were analyzed on day 9 of differentiation in P19(Gli2) and control cell lines. MyoD and myogenin transcripts were not observed by Northern blot analysis, indicating that MRFs were not induced to high levels in P19(Gli2) cells (data not shown). By RT-PCR analysis, both Myf-5 and myogenin transcripts were detected in P19(Gli2) cell lines, confirming that commitment into the skeletal muscle lineage occurred in these cultures and not control cells (Fig. 2,  panel II, E and F). Very low levels of MRF transcripts were present in the absence of muscle in the control cultures on days 0 (data not shown) and 9, in agreement with studies in other systems (41,42). Although exogenous Gli2 was expressed at high levels on day 0 in P19(Gli2) cells, Myf-5 was not activated on day 0 (data not shown). Together, these data indicate that Gli2 can up-regulate the expression of the myogenic factors, and that the activation is dependent on cellular aggregation in P19 cells.
In contrast to the results in P19 cells, Gli2 was not able to convert 10T1/2 fibroblasts into skeletal muscle using a myogenic conversion assay (data not shown). This is similar to findings with Pax3, which can induce myogenesis in P19 cells (9) as well as paraxial and lateral plate mesoderm explants (10), but not in 10T1/2 fibroblasts (10). It is likely that Gli2 function requires factors present in aggregated P19 cells but not in fibroblasts. Because Meox1 transcripts were up-regulated by Gli2 expression and during Me 2 SO-induced myogenesis (9), we examined the ability of Meox1 to induce skeletal myogenesis. Stable cell lines overexpressing Meox1, termed P19(Meox1) cells were created and aggregated in the absence of Me 2 SO, along with control cell lines. Cultures were examined by immunofluorescence (data not shown), Northern blot analysis, and RT-PCR (Fig. 3). Bipolar skeletal myocytes were not observed after staining with an anti-MHC antibody (data not shown). Therefore, in contrast to results with Gli2, and previous results with activated ␤-catenin (13) and Pax3 (9), Meox1 overexpression was not sufficient to induce skeletal myogenesis in aggregated P19 cells.
By Northern blot analysis, Meox1 transcripts were present at high levels in P19(Meox1) cells on day 9 and lower levels on day 0 (Fig. 3, panel I, A). This variability in PGK-driven transcript expression is because of the enhancer/silencer effects at the site of insertion and is seen in many P19 cell lines (35,37). Interestingly, Gli2 transcripts were up-regulated in P19(Meox1) cells, compared with control cells, whereas Pax3 transcripts were not significantly increased over background (Fig. 3, panel I, B and C). As a positive control, both Gli2 and Pax3 transcripts were detected in control cells treated with Me 2 SO to induce the endogenous skeletal muscle program (Fig.  3, panel I, B and C, lane 1). Therefore, Meox1 can activate Gli2 but not Pax3 expression in aggregated P19 cells.
MyoD and myogenin transcripts were not detectable by Northern blot analysis (data not shown). By RT-PCR analysis, very low levels of myogenin mRNA were detected in some cell lines (Fig. 3, panel II). However, Myf-5 transcripts were not observed to be significantly over background. Therefore, while Meox1 expression can up-regulate Gli2 expression, it cannot up-regulate Pax3 or MRF expression, leading to skeletal myogenesis. Taken together, it appears that Meox1 and Gli2 can activate the expression of each other, implying the presence of a regulatory loop.
Gli/Engrailed Expression Inhibits Muscle Specification in P19 Cells-Because we have shown that Gli2 is sufficient to induce the skeletal muscle program in P19 cells, we tested if Gli2 activity was also necessary for myogenesis by construction of a Gli/engrailed chimera (termed Gli/EnR). The activation domain of Gli2 was replaced with the engrailed repression domain that actively blocks transcription by interacting with transcriptional repressor proteins (43,44). P19(Gli/ EnR) and control cells were aggregated in the presence of 0.8% Me 2 SO and examined on day 9 of differentiation. Under these conditions, bipolar skeletal myocytes were detected by immunofluorescence in control cells but not P19(Gli/EnR) cells (data not shown). P19 cells expressing the engrailed repressor domain alone differentiated efficiently into skeletal muscle (data not shown). Analysis by RT-PCR showed a lack of Myf-5 mRNA in P19(Gli/EnR) cells when compared with P19(control) cells (Fig. 4, panel I), indicating that Gli factors, or genes containing Gli binding sites, are essential for Myf-5 expression and skeletal myogenesis.
Northern blot analysis was performed on a time course of differentiation to determine at what stage the myogenic program was disrupted (Fig. 4, panel II). The P19(Gli/EnR) cell line expressed high levels of Gli/EnR transcripts, as shown by hybridization to a cDNA fragment of the engrailed repressor domain (Fig. 4, panel II, A). The lower band provides the expected size for the fusion transcript and the upper band is not seen in all clonal populations isolated. It may occur due to the use of alternative polyadenylation signals at the site of insertion. Brachyury T, a protein expressed in the primitive streak during gastrulation, was expressed in both P19(control) and P19(Gli/EnR) cell lines, indicating that mesoderm induction still occurred (Fig. 4, panel II, B).
Factors found in the differentiating somite, including Gli2, Meox1, Pax3, and Six1 were all expressed during Me 2 SO-induced differentiation of control cell lines by days 3-5 (Fig. 4,  panel II, C-F). Expression of the Gli/EnR chimera did not down-regulate the transcript levels of endogenous Gli2 compared with control cells (Fig. 4, panel II, C). However, Meox1, Pax3, and Six1 expression was down-regulated in P19(Gli/ EnR) cell lines (Fig. 4, panel II, D-F). Meox1 transcripts were first detected on day 5 in P19(Gli/EnR) cells, 2 days after expression was initiated in the control cell lines. Subsequently,  (panel I, lanes 1 and 2, and panel II, lanes 1-10) and P19(Meox/EnR) cells (panel I, lanes 3-6, and panel II, lanes 11-20) were differentiated in the presence of 0.8% Me 2 SO. Northern blots with 12 g of RNA were probed with the factors indicated on the left. Lanes are indicated at the bottom. Panel I, total RNA was harvested on days 0 and 9 of differentiation. Panel II, total RNA was harvested on each day 9-day time course.
Meox1 transcripts were lost on days 6 -9 in P19(Gli/EnR) cells (Fig. 4, panel II, D). Similar to Meox1, Pax3 transcripts were initially detectable at very low levels on day 5 in P19(Gli/EnR) cells, 1 day after expression was initiated in control cell lines. Then, Pax3 transcripts were absent on days 6 -9 in P19(Gli/ EnR) cells (Fig. 4, panel II, E). Finally, Six1 transcript levels were not present from days 6 to 9 in P19(Gli/EnR) cells compared with control cell lines (Fig. 4, panel II, F). Therefore, the earliest disruption in the myogenic program in P19(Gli/EnR) cultures was the down-regulation of Pax3 and Meox1 expression on days 3-4 of differentiation.
MyoD transcripts were present on day 7 in the control cells but not in P19(Gli/EnR) cultures (Fig. 4, panel II, G). Myf-5 expression was also initiated on day 7 in control cells but not in P19(Gli/EnR) cultures (data not shown). Therefore, the disruption of wild type Gli2 activity resulted in the loss of the expression of somite patterning factors, such as Meox1 and Pax3, leading to the absence of MRF expression and commitment into the muscle lineage.
Meox/Engrailed Expression Inhibits Muscle Specification in P19 Cells-Although Meox1 expression was not sufficient to induce skeletal myogenesis, it was still important to test whether it is essential. A dominant-negative Meox1 chimera was constructed by fusing the engrailed repressor domain to the C terminus of Meox1. Stable cell lines were created and termed P19(Meox/EnR) cells. P19(control) and P19(Meox/EnR) cells were differentiated in the presence of Me 2 SO and examined by immunofluorescence and Northern blot analysis. Immunofluorescence with an anti-MHC antibody indicated a loss of detectable bipolar skeletal myocytes in P19(Meox/EnR) cells, compared with control cells (data not shown).
Analysis of transcript levels at day 9 of differentiation showed overexpression of Meox/EnR transcripts in the P19(Meox/EnR) cells, compared with control cells (Fig. 5, panel  I, A). The expression of the MRFs, MyoD, and myogenin, as well as the factors found in the developing somite, such as Pax3, Pax7, Six1, Eya2, and Gli2 was down-regulated in P19(Meox/EnR) cells, compared with control cells (Fig. 5, panel  I, B-H).
To determine the earliest stage that Meox/EnR expression disrupts myogenesis, time courses of differentiation experiments were examined by Northern blot analysis. Endogenous Meox1 and MyoD transcripts were completely lost throughout the differentiation time course in P19(Meox/EnR) cells compared with control cells, in which Meox1 transcripts are detected on day 4 and MyoD by day 7 (Fig. 5, panel II, A and  B). In contrast, Brachyury T transcripts were present in both P19(control) and P19(Meox/EnR) cells, indicating that mesoderm induction still occurred in P19(Meox/EnR) cells (Fig. 5, panel II, C). Pax3 expression was severely down-regulated in P19(Meox/EnR) cell lines from days 4 to 9, compared with control cell lines (Fig. 5, panel II, D). Gli2 expression was down-regulated but not lost in P19(Meox/EnR) cells from days 4 to 9, compared with control cell lines (Fig. 5, panel II,  E). Therefore, disruption of myogenesis occurred as early as day 4 of differentiation, with the down-regulation of Meox1 and Pax3 expression. Taken together, Meox factors, or genes having Meox binding sites, are essential for the proper expression of factors involved in somitogenesis and for subsequent skeletal myogenesis. DISCUSSION We have shown that Gli2 and Meox1 can activate the expression of each other in aggregated P19 cells. In addition, Gli2 can up-regulate Pax3 and MRF transcript levels, leading to skeletal myogenesis. Finally, overexpression of either dominantnegative Gli2 or Meox1 inhibited the specification of mesoder-mal precursors into the muscle lineage, resulting in the loss of MRF expression. In P19(Gli/EnR) cells, Meox1 levels were down-regulated and in P19(Meox/EnR) cells Gli2 levels were down-regulated. Therefore, these findings have led to a model in which Gli and Meox activate the expression of each other and are essential for the specification of cells into the skeletal muscle lineage (Fig. 6).
Expression of either Meox/EnR or Gli/EnR fusion proteins resulted in the down-regulation of Pax3 transcript levels, indicating that Meox or Gli factors, or genes containing Meox or Gli binding sites, are necessary for Pax3 expression (Fig. 6). Gli2 but not Meox1 overexpression was sufficient to induce transcription of Pax3. Previous results have shown that Pax3 overexpression could initiate transcription of Meox1 and that a Pax3/EnR fusion could down-regulate Meox1 transcript levels (9). Taken together, these results support the presence of a regulatory loop between Gli2, Meox1, and Pax3 that is important for the maintenance of their expression and subsequent commitment to the skeletal muscle lineage (Fig. 6).
The observation that Gli/EnR expression does not abolish the endogenous transcript levels of Gli2, indicates that Gli2 does not regulate its own expression. Conversely, Meox/EnR expression resulted in the loss of endogenous Meox1 transcripts and Pax3 was shown to be self-regulating in P19 cells by expression of a Pax3/engrailed chimera (9). Meox factors have been shown to bind Pax factors (29), although the affect of this association on the function of these factors remains unknown. Taken together, Meox1 and Pax3 can regulate their own expression, either directly or indirectly, and this self-regulation likely contributes to the stabilization of the Gli2-Meox1-Pax3 regulatory loop essential for proper muscle specification (Fig. 6).
Previous results have shown that ␤-catenin was essential and sufficient for the initiation of Gli2, Meox1, and Pax3 expression in aggregated P19 cells (13), implying an upstream role for Wnt signaling via ␤-catenin (Fig. 6). In addition, P19 cells grown in monolayer cultures expressing an activated form of ␤-catenin showed an up-regulation in Gli2 and Pax3 transcript levels, but not Meox1 (13). Consequently, it is possible to detect Pax3 transcripts in the absence of Meox1 transcripts, indicating that Meox1 is likely not the sole contributor regulating Pax3 expression. Meox1 was expressed strongly on day 5 of differentiation in P19(Gli/EnR) cultures (Fig. 4, panel II, D,  lane 16). This indicates that a portion of Meox1 expression is controlled in a Gli-independent fashion. Gli2 expression was down-regulated but not lost in P19(Meox1/EnR) cultures, indicating that part of Gli2 expression is independent of Meox function. The extent to which ␤-catenin regulates Gli2, Meox1, and Pax3, and whether this regulation is direct or indirect remains to be determined.
Gene ablation experiments in the mouse have shown that the Gray arrows indicate the results from previous studies (9,10,13,54,55). loss of functional Meox2, which is expressed in the developing limb bud (27,28), resulted in down-regulation of Pax3 and Myf-5 expression and a decrease in skeletal muscle in the limb (12). Furthermore, mice carrying null mutations for both Meox genes displayed a loss of Pax3, Myf-5, and myogenin expression, leading to severe deficiencies in somitogenesis and skeletal myogenesis (45). Consequently, our experiments in P19 cells, using an EnR fusion approach, have provided similar results to those in mice using gene ablation technology. Furthermore, mice lacking Meox genes display a similar phenotype to mice lacking hedgehog signaling (46), suggesting that Meox may mediate the response to hedgehog signals (45). The results presented here implicate Meox as a mediator of hedgehog signaling by regulating Gli2 expression.
Gli2 was sufficient to induce MRF transcripts and low levels of skeletal myogenesis. However, given that Gli2 expression first induced Meox1 and Pax3 transcripts, it is unclear which MRF was activated by which transcription factor and whether the interaction is direct or indirect. Recent studies have shown that the initial activation of Myf-5 expression is independent of the Gli site and of Shh functioning (16). Furthermore, we (13) and others (47) using Drosophila have shown that the Wnt pathway can regulate Gli/Ci expression and function. Therefore, our model is consistent with these results and suggests a complex combination of factors controlling Myf-5 expression.
The level of conversion into skeletal muscle is lower in P19(Gli2) cells than in other P19 cell lines examined. This may occur for several reasons. First, from our model, Gli2 functions early to commit cells to the myogenic lineage. Furthermore, Gli factors are involved in other pathways, such as cardiac (48) and neural development (49). In fact, P19(Gli2) cells made significant amounts of cardiac muscle. 2 The production of cardiac muscle indicates that some of the cells expressing Gli were routed into a cardiac muscle lineage and thus were not available to make skeletal muscle. Furthermore, Gli/Ci require proteolytic cleavage, nuclear localization, and/or hypophosphorylation for full positive activity (50). This may only happen in a subpopulation of aggregated P19 cells.
Meox1 expression was not sufficient to induce skeletal myogenesis, although Gli2, but not Pax3, transcripts were activated in aggregated P19(Meox1) cells. The finding that skeletal myogenesis did not proceed in the absence of Pax3 expression is consistent with our previous results, showing that a dominantnegative Pax3 disrupted skeletal myogenesis (9). However, given that Gli2 expression can activate Meox1 and Pax3, it is surprising that Meox1 expression activated Gli2 but not Pax3 transcripts. It may be that the levels of Gli2 induced by Meox1 were insufficient to activate Pax3 expression. Alternatively, the complex of factors present during aggregation may favor the cardiomyogenesis pathway, because abundant cardiac muscle was present after aggregation of P19(Meox1) or P19(Gli2) cells. 2 Similar results were found previously for MEF2C, in which MEF2C readily induced cardiomyogenesis in the absence of Me 2 SO but could only induce skeletal myogenesis under particular conditions in the presence of Me 2 SO (35,37). The finding that Meox1 preferentially activates cardiac versus skeletal muscle development implies the presence of the correct combination of factors and conditions to support Meox1induced cardiomyogenesis but not skeletal myogenesis in aggregated P19 cells.
Gli3 appears to act as a negative regulator of transcription (30). Gli factors were shown to act as transcriptional repressors in skeletal muscle differentiation in Xenopus (20). Our data that overexpression of Gli/EnR inhibits myogenesis suggests that Gli factors function as transcriptional activators during skeletal muscle development in P19 cells. Analysis of mice carrying mutations in one or more of the Gli factors suggests that the factors have overlapping functions in somitogenesis (48,(51)(52)(53). Further analysis of the effects of such mutations on myogenesis is needed to characterize the roles of Gli factors during skeletal muscle development in vivo.
In conclusion, our results implicate the presence of a regulatory loop between Gli2, Meox1, and Pax3 that is initiated by Wnt signaling via ␤-catenin and is maintained in part by autoregulation of Meox1 and Pax3. The Gli2-Meox1-Pax3 loop results in the specification of mesodermal cells into the muscle lineage and is essential for the expression of MRFs and thus commitment into muscle.