Regulation of Murine TGFβ2 by Pax3 during Early Embryonic Development*

Previously our laboratory identified TGFβ2 as a potential downstream target of Pax3 by utilizing microarray analysis and promoter data base mining (Mayanil, C. S. K., George, D., Freilich, L., Miljan, E. J., Mania-Farnell, B. J., McLone, D. G., and Bremer, E. G. (2001) J. Biol. Chem. 276, 49299-49309). Here we report that Pax3 directly regulates TGFβ2 transcription by binding to cis-regulatory elements within its promoter. Chromatin immunoprecipitation revealed that Pax3 bound to the cis-regulatory elements on the TGFβ2 promoter (GenBank™ accession number AF118263). Both TGFβ2 promoter-luciferase activity measurements in transient cotransfection experiments and electromobility shift assays supported the idea that Pax3 regulates TGFβ2 by directly binding to its cis-regulatory regions. Additionally, by using a combination of co-immunoprecipitation and chromatin immunoprecipitation, we show that the TGFβ2 cis-regulatory elements between bp 741-940 and bp 1012-1212 bind acetylated Pax3 and are associated with p300/CBP and histone deacetylases. The cis-regulatory elements between bp 741 and 940 in addition to associating with acetylated Pax3 and HDAC1 also associated with SIRT1. Whole mount in situ hybridization and quantitative real time reverse transcription-PCR showed diminished levels of TGFβ2 transcripts in Pax3-/- mouse embryos (whose phenotype is characterized by neural tube defects) as compared with Pax3+/+ littermates (embryonic day 10.0; 30 somite stage), suggesting that Pax3 regulation of TGFβ2 may play a pivotal role during early embryonic development.

Pax3 encodes a paired homeobox containing transcription factor that is expressed in neuroepithelium, neural crest, and presomitic mesoderm (1). Homozygous mouse embryos carrying a loss-of-function Pax3 allele (Pax3 Ϫ/Ϫ ) develop open neural tube defects, such as exencephaly or spina bifida (2), and die around embryonic day 14 (E14) 4 as a consequence of heart defects (3). More recently, Pax3 has been shown to function at the nodal point in melanocyte stem cell differentiation (4). Heterozygous embryos (Pax3 ϩ/Ϫ ) are viable but exhibit white patches on their bellies caused by defective development of neural crest-derived melanocytes. This suggests that disruption of the Pax3-dependent developmental program may cause defects in the development of neural crest-derived structures.
Two main Pax3-binding sites have been identified and are found in most Pax3 target elements as follows: (i) a binding site derived from the Drosophila Paired (ATTA N5 GTTCC), and (ii) a Pax3 paired domain binding site (CGTCAC(G/A)(C/ G)TT) identified by Epstein et al. (5) by CASTing (DNA-binding site selection assays) in the c-Met promoter region. In addition several paired domains, such as GTTCC, CAGTGT, GTTAT, GTGTGA, and CAAGG (6), as well as the homeodomain ATTA, have been suggested to be "putative Pax3-binding motifs" (7,8). More recently, Corey and Underhill (9) demonstrated that Pax3 can regulate target genes through alternative modes of DNA recognition. They observed that although the microphthalmia-associated transcription factor element is characterized by suboptimal recognition motifs for the paired domain and homeodomain, it sustains a higher level of Pax3 binding than TRP-1, which contains a canonical paired domain site. The basis for this difference involved a context-dependent cooperative binding event requiring both the paired and homeodomain, whereas the paired domain alone was sufficient for TRP-1 recognition.
Since Pax3 is important in diverse cellular functions during development, we wanted to identify additional genes regulated by Pax3. To accomplish this we utilized oligonucleotide arrays and RNA isolated from Pax3-transfected cell lines and promoter-based data mining (6). Based on the putative Pax3-binding motif score (motif score is defined as the total number of times a paired or a homeodomain appears in the promoter sequence of a given gene) in the promoter regions of the genes whose expression was changed by Pax3 transfection, we postulated 17 new genes that may be direct Pax3 downstream targets and another 17 genes (whose promoter sequences are not known) as putative Pax3 downstream targets (6). The expression of one of these genes, human TGF␤2, was increased in two of the three Pax3 transfectants studied, and this gene showed the highest putative Pax3-binding motif score, making it a likely gene regulated by Pax3.
The importance of TGF␤2 in the process of neural tube closure was first demonstrated by Sanford et al. (10) who showed that TGF␤2 knock-out mice had spina bifida occulta and unfused neural arches at the midline of neural tube along with other developmental abnormalities. In this paper, we demonstrate the following: (i) Pax3 regulates TGF␤2 by binding to cis-regulatory elements of the promoter; (ii) acetylated Pax3 is associated with p300/CBP, HDAC1, and SIRT1; and (iii) TGF␤2 transcript levels are diminished in the Pax3 Ϫ/Ϫ embryos. Thus our microarray data (6), along with the present findings in this paper in combination with the results of Sanford et al. (10), indicate that TGF␤2 is regulated by Pax3 and could be involved in early embryonic development.
Pax3-GST Fusion Protein Expression and Isolation-The Pax3-GST fusion cDNA was provided by Dr. Mary R. Loeken (Harvard Medical School, Boston). The Escherichia coli was transformed with Pax3-GST fusion protein construct in pcDNA3, and the cells were grown overnight in LB medium. Pax3-GST fusion protein expression and isolation were performed as described previously (6).
Co-immunoprecipitation-Pax3 transfectants, B9 (6), were treated with or without 2 M trichostatin A in serumfree DMEM for 4 h at 37°C and then washed three times with cold PBS. The washed cells were collected in cold PBS and then lysed on ice for 30 min in RIPA buffer containing 50 mM Tris-HCl (pH 8.0), 150 mM NaCl, 5 mM EDTA, 1% Nonidet P-40, 0.1% SDS, and 0.5% sodium deoxycholate. Protease inhibitors (Calbiochem) were added just before lysis. Immunoprecipitations were carried our by using polyclonal HDAC1 (Abcam) and acetylated protein antibody (Abcam) as per the manufacturer's instructions. The immunoprecipitates were collected by 50 l of 50% slurry of protein G-Sepharose (Sigma). The precipitates were washed three times in lysis buffer, and boiled in 2ϫ SDS sample buffer for 5 min and centrifuged for 1 min. The proteins in the supernatant were resolved in 8% SDS-PAGE and immunoblotted with Pax3 monoclonal antibody (Developmental Studies Hybridoma Bank, University of Iowa) and with p300/CBP (Santa Cruz Biotechnology, Inc.).

Whole Mount in Situ Hybridization-
Whole mount in situ hybridization was done according to the method described previously (14). Washes were performed at 65°C. The plasmid pmTGF␤2 containing mouse TGF␤2 (442 bp) was kindly provided by Dr. M. Azhar (University of Cincinnati, Department of Molecular Genetics, Cincinnati, OH). The antisense digoxigenin labeled TGF␤2 riboprobe was synthesized by linearizing the PCR products with EcoRI and using SP6 polymerase. The sense digoxigenin-labeled TGF␤2 riboprobe was synthesized by linearizing the PCR products with XhoI and using T7 polymerase. Embryos were stained and photographed with a Spot Camera (Diagnostic Instruments Inc.).

Pax3 Binds to cis-Regulatory Regions of Murine TGF␤2
Promoter-An unbiased survey of the entire murine TGF␤2 promoter (GenBank TM accession number AF118263) to ascertain putative Pax3-binding elements was done by employing chromatin immunoprecipitation on E10.0 (30 somite stage) Pax3 ϩ/ϩ embryos using rabbit polyclonal Pax3 antibody. The forward and reverse primer sets A-F amplified 200-bp fragments from the input sample (Fig. 1, lane 1) and immunoprecipitates (Fig. 1, lane 3). No amplification was observed when ␤-actin forward and reverse primers were used. Immunoprecipitation reactions using IgG from normal rabbit serum did not show any amplification of the TGF␤2 promoter in the immunoprecipitated DNA (Fig. 1,  lane 2). All ChIP samples were tested for false-positive PCR amplification by sequencing the 200-bp amplified product to ascertain the specificity of the Pax3 binding to the cis-regulatory elements. Additionally, false-positive PCR amplification was ruled out by amplifying the sequence from murine ␤-actin gene, which failed to give any PCR product in the immunoprecipitates. With the exception of primer set A, all other primer sets (B-F) amplified a 200-bp fragment from the DNA extracted from the immunoprecipitates (Fig. 1, lane 3). It is important to note that even though the primers to probe specific regions are spaced at 200-bp intervals, the level of resolution is determined by the fragment DNA length following sonication. This is generally on the order of 200 -600 bp, depending on the conditions of the sonication.
Functional Analysis of Murine TGF␤2 Promoter Activity-To determine the functional interaction of Pax3 with murine TGF␤2 promoter and the effect on the promoter activity, we employed transient co-transfection experiments using luciferase-reporter plasmids containing different murine TGF␤2 promoter sequences as described by Wilder et al. (11) with Pax3 expression plasmids. Murine TGF␤2 promoter plasmids were subcloned into the pPGB11 plasmid and labeled as TG1 (bp 1226 -1290), TG2 (bp 1191-1290), TG3 (bp 1143-1290), TG4 (bp 950 -1290), TG6 (bp 734 -1290), and TG7 (bp 1-1290) (11). A full-length promoter sequence and each construct are shown in Fig. 2. The promoterless luciferase gene in pPGB11 (13) and the plasmid constructs containing murine TGF␤2 promoter sequences were assayed for promoter activity by transient co-transfection of Pax3-pcDNA3 into wild type DAOY cells (Fig. 3). Luciferase activity was normalized to Renilla luciferase activity by cotransfecting an internal control plasmid, pRL-null, carrying cDNA encoding the Renilla luciferase gene. Our results showed that the promoter constructs TG4 and TG7 displayedmaximumluciferaseactivity with TG4 showing slightly higher activity than TG7 construct. These observations suggest that these constructs contain transcriptional activating elements that may be responsive to Pax3 binding. The promoter constructs TG1, TG2, TG3, and TG6, on the other hand, displayed no increase in luciferase activity. It is intriguing to know that TG6 promoter construct did not show any increase in promoter activity, although it encompassed the TG4 construct promoter region. This obser-   Fig. 2) or pcDNA3 and murine TGF␤2 promoter constructs and pPGB11 (promoter-less luciferase gene) were co-transfected into wild type DAOY cells. The luciferase activity was assayed 48 h after transfection as described under "Experimental Procedures." Values of pPGB11 were subtracted from the TGF␤2 promoter-luciferase values. Experiments were performed in quadruplicate with each data point in duplicate. Note the region of (gt) 32 and the region between bp 1-734 and bp 950 -1143 that may have contributed to increased TG4 and TG7 construct activity. AUGUST 25, 2006 • VOLUME 281 • NUMBER 34 vation suggests that the TG6 promoter construct may contain silencing elements 5Ј to the boundary of the TG4 promoter construct that significantly decreases promoter activity. One plausible explanation for this latter observation could be due to the presence of a dinucleotide (GT) 32 repeat region in this promoter sequence (Fig. 2). Previous studies by Yoshikawa and co-workers (15) have identified a functional and polymorphic (GT) n repeat in the 5Ј-regulatory region of N-methyl-D-aspartate receptor 2A subunit (GRIN2A) gene that resulted in the length-dependent inhibition of transcriptional activity. Similarly, Yamada et al. (16) suggested that the larger the size of a (GT) n repeat in the hemoxygenase-1 gene promoter, the greater the transcriptional inhibition. In sum, the observations from our promoter-luciferase study concluded that TG7 and TG4 promoter constructs may be having Pax3 responsive regions between bp 1-734 and bp 950 -1143, respectively.

Pax3
Regulates Murine TGF␤2 Promoter-The above study prompted us to look for the paired or the homeodomain binding sequences within bp 1-734 (TG7) and bp 950 -1143 (TG4) that contributed to the maximum activity. Therefore, we made several promoter-luciferase constructs on the TG7 backbone with point mutations, either in a paired domain binding site (PDBS) or homeodomain-binding site (HDBS) between bp 1-734 and bp 950 -1143, representing the regions in TG7 and TG4 that showed highest luciferase activity in Fig. 3. The results (Fig. 4) show that point mutations T514 (Tgm6) in the PDBS "gttat" and T664 (Tgm10) in the PDBS "gttcc" showed a decrease in promoter activity with T1056 (Tgm14) in the PDBS "ccttg" showing a marginal decrease in promoter activity. The Tgm6 and Tgm10 mutants gave the maximum reduction in promoter activity followed by Tgm14. By contrast, a point mutation T526 (Tgm7) in the PDBS "gtgtga" did not show any decrease in promoter activity. Except for point mutation in the HDBS A178 (Tgm4), point mutations in HDBS represented by A569 (Tgm8) and A766 (Tgm11) showed no appreciable decrease in promoter activity. These results suggest that not all paired domains or homeodomain-binding sites contribute equally to TGF␤2 promoter activity by Pax3. Thus, our data show that the PDBS "gttat" (bp 512-516) and ccttg (bp 1053-1057) and HDBS "atta" (bp 176 -179) contribute maximally to promoter activity.
The above results as well as our chromatin immunoprecipitation data suggested that there may be more than one Pax3response elements in the TG7 promoter construct. To evaluate how different regions may combine to influence promoter  activity, we introduced several deletion as well as point mutations into the TG7 promoter construct (as outlined in Fig. 5). Deletion and point mutation locations were based on the mutation studies mentioned above. These studies defined maximal promoter activation located in the PDBS gttat (bp 512-516) and ccttg (bp 1053-1057). We also chose another PDBS gtgtga (bp 523-528) adjacent to gttat. This PDBS was described by Epstein and co-workers (16,17) as binding to Pax3 and mediating activation of the c-RET enhancer. We introduced a deletion mutation to verify whether this region was important for transcriptional activity of the TGF␤2 promoter. Our results (Fig. 5) showed that deletion of the paired domain binding motif, gttat (TG7⌬2, bp 512-516), in the TG7 construct decreased promoter activity marginally by 42%, whereas deletion of another paired domain binding motif, gtgtga (TG7⌬3, bp 523-528), did not reduce promoter activity. These results suggested that the PDBS gttat (bp 512-516) may be part of a Pax3-response element in the murine TGF␤2 promoter.
Identification of Pax3-binding Sequences in Murine TGF␤2 Promoter-To identify Pax3-binding sequences in the murine TGF␤2 promoter, we used EMSA with 32 P-labeled 30-mer oligonucleotides and purified Pax3-GST fusion protein. In our initial screen for the choice of oligonucleotides, we found that oligonucleotide containing gtgtga weakly bound Pax3-GST in EMSA (data not shown). These data corroborated those of Lang et al. (17). Because the ON-c (bp 504 -533) oligonucleotide used in our EMSA studies also contains a gtgtga region, we wanted to confirm the specificity of the Pax3 binding by introducing a point mutation (T526) within the gtgtga region of ON-c. However, even after introducing a point mutation (T526), Pax3-GST still bound the oligonucleotide (data not shown). This clearly suggested an additional region within the oligonucleotide ON-c-bound Pax3-GST. Our promoter luciferase data suggested that the Pax3-GST binding region in ON-c might be gttat (bp 512-516) (Fig. 6). Deleting gtgtga (bp 523-528) as in TG7⌬3 (Fig. 6) and a point mutation (T526) as in the mutant promoter construct Tgm7 (Fig. 5) did not change the activity substantially when compared with the native TG7 promoter construct. These data suggested that gtgtga (bp 523-528), while binding to Pax3-GST, was not as transcriptionally significant as gttat (bp 512-516) in the context of our experiments using the TGF␤2 promoter. On the contrary, the sequence gtgtga was reported to be significant for the activation of the c-RET enhancer, where Pax3 binding is modified by the Sox10-binding site adjacent to the Pax3-binding site (17,18). We therefore chose to introduce two point mutations, one being T514 within the gttat region and the other being T526 within the gtgtga region in the ON-c oligonucleotide for EMSA and not in the luciferase assays. Our results showed that introducing two point mutations in ON-c abolished binding to purified Pax3-GST.

The TGF␤2 cis-Regulatory Elements Bind Acetylated Pax3 and Are Associated with p300/CBP and Histone Deacetylases-
The above studies demonstrated that Pax3 not only bound the TGF␤2 promoter but also participated in its transcriptional regulation. To understand the molecular mechanisms that underlie the regulation of the TGF␤2 promoter by Pax3, we used a combination of co-immunoprecipitation assays coupled with chromatin immunoprecipitations to map the trans-acting factors that might modulate Pax3 function. To demonstrate that Pax3 gets acetylated in cells, we used stable Pax3 transfec-tants B9 and incubated with or without trichostatin A. The cells were harvested and subjected to immunoprecipitation with acetylated protein antibody and then immunoblotted with Pax3 antibody. The results (Fig. 7) showed that Pax3 gets acetylated in stable Pax3 transfectant B9. To ascertain whether HDAC1 is associated with acetylated Pax3, we performed immunoprecipitation assays using HDAC1 antibody and immunoblotted with Pax3 antibody. The presence of Pax3 in the immunoprecipitates showed that HDAC1 is associated with acetylated Pax3. To ascertain whether p300/CBP also forms the part of the transcriptional regulatory complex, we performed immunoblotting of HDAC1 immunoprecipitates with p300/CBP. The presence of p300/CBP in the immunoprecipitates showed that p300/CBP forms the part of transcriptional regulatory complex. Thus our co-immunoprecipitation data showed the following: (i) Pax3 gets acetylated, and (ii) Pax3 associates with p300/CBP and HDAC1.
Next we wanted to know whether chromatin immunoprecipitation can identify HDAC1, SIRT1, or acetylated protein interaction domains in the murine TGF␤2 promoter, which was earlier defined by Pax3. By using the same set of TGF␤2 promoter forward and reverse primers as used in Fig. 1, which defined the Pax3 recognition domains, we were able to map the HDAC1 interaction domains to the 200-bp PCR-amplified product, with the exception of primer set A (Fig. 8). However, when SIRT1 antibody was used for chromatin immunoprecipitation, only primer sets C, D, and F amplified the 200-bp fragment from the immunoprecipitates. When acetylated protein antibody was used for immunoprecipitation, only primer sets D and F amplified the 200-bp fragment from the immunoprecipitates. These observations suggest that not all regions of the TGF␤2 promoter behave equally even though Pax3 might bind to at least seven oligonucleotides ( Fig. 6) spanning five cisregulatory regions of the TGF␤2 promoter (Fig. 1). The presence of deacetylases such as HDAC1 and SIRT1 might also mean association of other acetylated transcription factors such as p53 (19 -21) within the 200-bp cis-regulatory region of TGF␤2 promoter. The 200-bp fragment amplified by primer set D (between bp 741 and 940) is noteworthy because this region of the promoter as defined by promoter construct TG6 (Fig. 3) has the (gt) 32 repeats that showed diminished promoter activity in our promoterluciferase reporter assays. p53 is a known substrate for SIRT1 (21,22) involved in transcriptional repression of promoters. On the other hand, the 200-bp fragment amplified by primer set E (between bp 1012 and 1212) does not seem to  have SIRT1 associated to it. Interestingly, the TGF␤2 promoter construct TG4 (Fig. 3), which showed an increase in the promoter-luciferase activity, houses this 200-bp fragment. Further evaluation of which other trans-acting factors modulate the TGF␤2 promoter activity along with Pax3 during early embryonic development is currently underway in our laboratory. TGF␤2 mRNA Expression Is Reduced in Pax3 Ϫ/Ϫ Mice-Based on the above findings it was tempting to ask whether Pax3 Ϫ/Ϫ embryos, whose phenotype is characterized by neural tube defects, show diminished levels of TGF␤2 expression. To test the hypothesis that Pax3 regulates the expression of TGF␤2 in vivo during early embryonic development, we used Pax3 ϩ/ϩ and Pax3 Ϫ/Ϫ mouse embryos (E10.0; 30 somites). The neural tube in these genotypes is closed in the thoracicolumbar region, whereas the caudal region is open in Pax3 Ϫ/Ϫ mouse embryos. We performed whole mount in situ hybridization of E10.0 (30 somites) of Pax3 ϩ/ϩ and stage-matched Pax3 Ϫ/Ϫ littermates with digoxigenin-labeled murine TGF␤2 riboprobe. We observed a significant reduction of TGF␤2-positive staining in the migratory neural crest cells in Pax3 Ϫ/Ϫ embryos (Fig. 9, C and D) as compared with the age-matched Pax3 ϩ/ϩ embryos (Fig. 9, A and B). In sum, our data suggest that Pax3 is involved in the regulation of TGF␤2 expression in the migrating neural crest cells. Taken together, our data suggest that Pax3 is involved in the regulation of TGF␤2 expression during early embryonic development.

DISCUSSION
Implications for TGF␤2 Regulation by Pax3-Early embryonic development involves neural crest induction, epithelialto-mesenchymal transition neural crest cell migration, and subsequent neural tube formation (23). Failure of any of these processes results in a neural tube defect, characterized by exencephaly and anencephaly in the rostral part and as spina bifida in the caudal part of the neural tube (24). Pax3 is expressed in the developing neural crest (25). Neural crest cells emerge from the dorsal neural tube and migrate throughout the developing embryo, where they give rise to a range of cell types, including neurons and glial cells of the peripheral nervous system and melanocytes. In Pax3 Ϫ/Ϫ embryos, neural crest cells fail to migrate correctly, resulting in either a severe reduction or a complete absence of neural crest derivatives (26). Studies by Serbedzija and McMahon (27) indicate that the defect in the Splotch (Pax3 ϩ/Ϫ ) mutation is not intrinsic to the neural crest cells themselves but appears to reflect inappropriate cell interactions either within the neural tube or between the neural tube and the somite. When caudal neural tubes from Splotch (Pax3 ϩ/Ϫ ) embryos were grafted in a nonsplotch host, normal neural crest migration was observed, suggesting that these interactions are defective in Splotch (Pax3 ϩ/Ϫ ) mice and that neural crest generation requires interaction with neighboring tissue (27). These experiments imply that a soluble mediator or the extracellular matrix is involved in neural crest cell migration. TGF␤2 may be a candidate for this soluble mediator. Several studies discussed below support this possibility.
TGF␤s have been shown to modify neural crest cell differentiation and migration (28). TGF␤s regulate key aspects of embryonic development and major human diseases (29).
TGF␤2 mediate effects on cell migration by regulating genes responsible for remodeling the cell-extracellular matrix as well as genes related to adhesion molecules/receptors or the cytoskeleton. In support of this view, Col2A(I) has been shown to be a downstream target of TGF␤2 signaling (30). Similarly, an increase in versican, a chondroitin sulfate proteoglycan, and associated proteins in response to growth factors, such as platelet-derived growth factor and TGF␤1, have been shown to cause an increase in the pericellular matrix of the cells and expansion of the extracellular matrix (31). Previously, we showed that Pax3 overexpression causes an up-regulation of the V2 splice variant of versican and down-regulation of the V3 splice variant (6). It is quite plausible that Pax3 may affect versican regulation via TGF␤2 signaling. Zavadil et al. (32) have shown increased gene expression of a host of extracellular matrix molecules, such as VEGF, tenascin, lamininB1, lami-ninA3, fibronectin1, thrombospondin, integrin␣2, following TGF␤ treatment. Similarly, genes important for cytoskeletal remodeling have been shown to be up-regulated following TGF␤ treatment (32).
In addition to acting on genes involved in cell migration, TGF␤2 signaling has also been shown to act on genes involved in neural crest generation and epithelial-mesenchymal transformation. For instance, Slug encoding a zinc finger transcription factor, is a downstream target of TGF␤ signaling (33). Slug is expressed in both pre-migratory and migratory chick neural crest cells (17) and promotes desmosome dissociation (33), a key step in epithelial-to-mesenchymal transition. Recently, Msx1 and Pax3 (34) and Pax3 and Zic1 (35) have been shown to be responsible for Slug induction, a direct TGF␤2 target. TGF␤ also up-regulates Sox4 and Sox9, key regulators in the differentiation of early mesenchymal progenitor cells into cardiac (36) and chondrogenic lineage (37). Thus, Pax3 regulation of TGF␤2 at the time of neural tube closure seems quite plausible in the light of the above findings.
Pax3 is not the only functional regulator of TGF␤2, but one important regulator of neural crest cell migration. Developmental events such as neural tube closure are tightly modulated by gene regulatory networks (23,38,39,41). The relatively small change in the expression of a key regulator gene such as TGF␤2 can either be amplified by a regulatory network of genes to affect the expression of downstream genes or can signal a counter-regulatory loop to compensate for the change. Earlier we had shown that a 2-4-fold difference in the expression of ST8SiaII/STX in Pax3 transfectants had a profound effect on NCAM polysialylation. Furthermore, an increase in PSA-NCAM resulted in a complete phenotypic change by inhibiting NCAM-dependent aggregation and increasing PSA-NCAM heterophilic adhesion to heparin sulfate proteoglycan (13). In summary, our results show the following: (i) Pax3 regulates TGF␤2 gene by binding to the cis-regulatory elements on the promoter, and (ii) the expression of TGF␤2 is reduced in migratory neural crest cells in Pax3 Ϫ/Ϫ mice. Overall, these results support the hypothesis that Pax3 regulates TGF␤2 expression during early embryonic development.

Trans-regulatory Factors Affecting Pax3 Regulation of TGF␤2
Promoter-The question is, "Does TGF␤2 belong to a Pax3-dependent developmental program?" Our data support that Pax3 regulates TGF␤2 but does not suggest that TGF␤2 belongs to a Pax3-dependent developmental program. TGF␤2 regulation by Pax3 may represent an aspect of the Pax3-dependent developmental program, such as migration of neural crest cells, but is not necessarily responsible for all the downstream events. The studies by Pani et al. (42) is particularly noteworthy because they suggested that neural tube closure is a Pax3-independent process, and the sole required function of Pax3 in neural tube closure is to inhibit p53-dependent apoptosis. Similarly, SIRT1 has been shown to inhibit p53-dependent apoptosis (43). Cordenonsi et al. (44) have shown that p53 partners with Smad2 in the activation of multiple TGF␤ target genes. Wilkinson et al. (45) have shown that there is a direct cooperation between p53 and TGF␤ effectors in chromatin modification and transcription repression. Secreted TGF␤2 can activate NF-B, block apoptosis, and is essential for the survival of some tumor cells (40). These observations and our finding that Pax3 regulates TGF␤2 may imply that Pax3 may be performing the same role of blocking apoptosis the way it does to block p53-mediated apoptosis.
In summary, we show the following: (i) Pax3 regulates TGF␤2 directly by binding to cis-regulatory elements on the promoter; (ii) certain regions of the cis-regulatory elements in TGF␤2 promoter bind acetylated Pax3, which is associated with deacetylases and p300/CBP; and (iii) TGF␤2 expression is diminished in migrating neural crest cells in Pax3 Ϫ/Ϫ embryos.