Functional Antagonism between Msx2 and CCAAT/Enhancer-binding Protein a in Regulating the Mouse Amelogenin Gene Expression Is Mediated by Protein-Protein Interaction*

Ameloblast-specific amelogenin gene expression is spatiotemporally regulated during tooth development. In a previous study, the CCAAT/enhancer-binding protein a (C/EBP a ) was identified as a transcriptional activator of the mouse amelogenin gene in a cell type-specific manner. Here, Msx2 is shown to repress the promoter activity of amelogenin-promoter reporter constructs independent of its intrinsic DNA binding activity. In transient cotransfection assays, Msx2 and C/EBP a antagonize each other in regulating the expression of the mouse amelogenin gene. Electrophoresis mobility shift assays demonstrate that Msx2 interferes with the binding of C/EBP a to its cognate site in the mouse amelogenin minimal promoter, although Msx2 itself does not bind to the same promoter fragment. Protein-protein interaction between Msx2 and C/EBP a is identified with co-immunoprecipitation analyses. Functional antagonism between Msx2 and C/EBP a is also observed on the stably transfected 2.2-kilobase mouse amelogenin promoter in ameloblast-like LS8 cells. Furthermore, the carboxyl-terminal residues 183–267 of Msx2 are required for protein-protein interaction, whereas the amino-ter-minal residues 2–97 of Msx2 play a less critical role. Among three family members tested (C/EBP a , - b , and - g ),

Enamel is the only epithelially derived calcified tissue in vertebrates. Amelogenin, the major organic component of enamel matrix, is essential to the proper regulation of enamel mineralization. Amelogenin proteins comprise ϳ90% of the enamel matrix proteins. Several mutations in the human Xchromosomal amelogenin gene have been identified from patients with the inherited enamel defect X-linked Amelogenesis imperfecta (1)(2)(3). Disruption of amelogenin synthesis during tooth development with either antisense oligonucleotides or ribozymes results in disorganized enamel (4,5). Amelogenin expression is ameloblast specific and developmentally regu-lated at the temporal and spatial level (6 -13). A 2263-nucleotide proximal promoter element from the mouse X-chromosomal amelogenin gene has been demonstrated by transgenic mouse analysis to recapitulate the spatiotemporal expression pattern of the endogenous amelogenin gene (10). Extensive homologies (70% identity) in the 300-nucleotide region upstream of the transcription initiation site exist between the murine, bovine, and human X-chromosomal amelogenin gene, suggesting that this region is likely involved in the transcriptional regulation of tissue-specific amelogenin gene expression. In a previous study, the minimal promoter of the mouse amelogenin gene (Ϫ70/ϩ52) was identified, which contains a CCAAT/enhancer-binding protein (C/EBP) 1 consensus binding site, and C/EBP␣ activated amelogenin transcription in a celltype specific manner through binding to its cognate site (14).
The C/EBPs consist of a family of related basic region leucine zipper transcription factors that are critical regulators of cellular differentiation and function in multiple tissues. Six different members of the family (C/EBP␣, -␤, -␥, -␦, -⑀, and -) have been isolated and characterized. The expression of C/EBPs is tissue-and stage-specific during development. C/EBPs have been shown to play a key role in regulating cellular differentiation, terminal function, and response to inflammatory insults (15)(16)(17)(18)(19).
The Msx2 gene family (20) is the mammalian counterpart of the Drosophila msh (muscle segment homeobox) gene. Three unlinked members, Msx1 (21,22), Msx2 (23), and Msx3 (24), have homeobox sequences very similar to each other and to the Drosophila msh gene. The murine Msx3 is expressed only in the dorsal neural tube (25)(26)(27), which appears to exclude the possibility of functional redundancy of Msx3 on the role of Msx1 and Msx2 in tooth development. During odontogenesis, Msx1 is expressed at all stages in dental mesenchymal cells but not in epithelial cells (28,29). The expression pattern of Msx2 changes with the differentiation of different germ layers. Msx2 is strongly expressed in undifferentiated inner enamel epithelia in which amelogenin expression is barely detectable; but Msx2 is absent in differentiated ameloblasts in which robust expression of amelogenin is detected. On the other hand, Msx2 is weakly expressed/absent in undifferentiated dental papilla mesenchyme, whereas it is strongly expressed in odontoblasts and differentiated dental papilla cells (30). Msx2 has been shown to function as a transcriptional repressor independent of its intrinsic DNA binding activity through the homeodomain. Instead, the repression is mediated by protein-protein interac-tion with either components of basal transcription machinery or other transcription factors (31)(32)(33)(34)(35).
To investigate the role of Msx2 in the regulation of amelogenin gene expression, various amelogenin-promoter reporter constructs were transiently transfected into ameloblast-like LS8 cells with a Msx2 expression plasmid. The functional relationship between Msx2 and C/EBP␣, a transcriptional activator of amelogenin, was examined with cotransfection assays. The potential of Msx2 to interfere with the binding of C/EBP␣ to its cognate site on the amelogenin minimal promoter as well as the ability of Msx2 itself to bind to the promoter was assessed with electrophoresis mobility shift assays (EMSA). Whether Msx2 is able to interact with C/EBP␣ in LS8 cells was further determined with co-immunoprecipitation analyses. The functional antagonism between Msx2 and C/EBP␣ was tested on a stably transfected amelogenin-promoter reporter construct. Finally, the ability of Msx2 to interact with two other C/EBP family members, C/EBP␤ and C/EBP␥, was examined with cotransfection assays.

MATERIALS AND METHODS
Cell Culture and Plasmids-A mouse ameloblast cell line (LS8) established by immortalizing primary cultures of enamel organ epithelium with SV40 large T antigen was maintained in Dulbecco's Modified Eagle's Medium (Sigma) supplemented with fetal bovine serum (10%), penicillin (100 units/ml), and streptomycin (100 g/ml) (36). The mouse amelogenin-promoter reporter constructs were described in a previous study (14). The expression plasmids of Msx2FL, Msx⌬N, and Msx2⌬C (31) were generously provided by Dr. Dwight Towler (Washington University).
Transfection and Luciferase Assay-Transient transfection and luciferase assays were performed as described previously (14). Stable cell line LS8/p2207 was established by transfecting LS8 cells with amelogenin-promoter reporter construct p2207. After selection with hygromycin (750 ng/ml), the resistant colonies were isolated, expanded, and screened with luciferase assays for reporter gene activity. Seven clones with luciferase activity 1,000 -10,000-fold of that in parental LS8 cells were selected for further studies. One of the selected clones with the highest basal luciferase activity was designated LS8/p2207 and maintained in the presence of hygromycin (750 ng/ml).
EMSA-Double-stranded oligonucleotide probes were generated by annealing antisense strand to a 10-fold excess of sense strand and filling in with [␣-32 P]dATP (NEN Life Science Products) and Klenow (exo Ϫ ). The GelShift buffer kit (Stratagene) was used for the binding reaction, which was performed as recommended by the manufacturer. The reaction mixtures were resolved in a 6% nondenaturing polyacrylamide gel provided in the kit. The gel was dried, and the bands were visualized by autoradiography. The sequences of the oligonucleotides were: amel antisense strand 5Ј-GAACAGCCAATCAGGTTTCTGAATG-AA-3Ј, amel sense strand 5Ј-TTTTTCATTCAGAAACCTGA TTGGCTG-TTC-3Ј (nucleotide 1510 -1539, GenBank TM accession number AF083091); msx2 antisense strand 5Ј-ACTTTGAACAGCCAATTAGT TTCTGAATGAA-3Ј, msx2 sense strand 5Ј-TTCATTCAGAAACTAATT-GGCTGTTCA-3Ј. Nuclear extracts were prepared from LS8 cells transfected with a C/EBP␣ expression plasmid, as described previously (14).
Immunoprecipitation and Western Blot Analysis-LS8 cells, 80 -90% confluent in a 100-mm cell culture plate, were collected in ice-cold radioimmune precipitation buffer (1ϫ phosphate-buffered saline, 1% Nonidet P-40, 0.5% sodium deoxycholate, 0.1% SDS, 0.1 mg/ml phenylmethylsulfonyl fluoride, 30 l/ml aprotinin (Sigma), and 1 mM sodium orthovanadate) and lysed by passing through a 22-gauge needle six times at 4°C. After centrifugation at 3,000 rpm for 15 min at 4°C, the protein concentration of the supernatant was determined using a Bio-Rad protein assay kit (Bio-Rad) with bovine serum albumin standards. A primary antibody (2 g) was added to 1 ml of cell lysate (500 g of total cellular protein) and incubated overnight at 4°C with rotation. Twenty-five microliters of protein G PLUS-Agarose (Santa Cruz Biotechnology) was added and incubated for 6 h at 4°C with rotation. Immunoprecipitates were collected by centrifugation at 2,500 rpm for 5 min at 4°C. Pellets were washed three times with 1.0 ml of radioimmune precipitation buffer and once with 1.0 ml of phosphate-buffered saline (pH 7.4) at 4°C. After final wash, pellets were resuspended in an equal volume of 2ϫ SDS loading buffer, boiled for 3 min, and stored at Ϫ80°C. Samples (10 l) were analyzed by Western blot as described previously (14).
In Vitro Expression of Protein with TNT-coupled Wheat Germ Extract System-The in vitro transcription-coupled translation reactions were performed using a TNT T7-coupled wheat germ extract system, according to the manufacturer's recommendation (Promega). Briefly, 25 l of TNT wheat germ extract, 2 l of TNT reaction buffer, 1 l of TNT RNA polymerase, 0.5 l of amino acid mixture (minus leucine, 1 mM), 0.5 l of amino acid mixture (minus methionine, 1 mM), 1 l of RNasin ribonuclease inhibitor (40u/l), 2 l of template DNA (0.5 mg/ml), and 18 l of nuclease-free H 2 O were added into a 0.6-ml centrifuge tube. The translation mixture was incubated at 30°C for 90 min and then stored at Ϫ80°C. The [ 35 S]methionine was included in the reactions in a parallel experiment to determine the translation efficiency of individual protein.

Msx2 Was a Transcriptional Repressor of Mouse Amelogenin
Promoter-To determine whether Msx2 could regulate the promoter activity of the mouse amelogenin gene, a Msx2 expression plasmid was cotransfected into ameloblast-like LS8 cells with a series of 5Ј-deletion amelogenin-promoter reporter constructs. In a previous study, we demonstrated that the Ϫ70/ ϩ51 region of the mouse amelogenin promoter functions as a minimal promoter in LS8 cells (14). Exogenous Msx2 was able to down-regulate the promoter activity of each reporter construct so long as that construct contained the mouse amelogenin minimal promoter (p2207, p454, p349, p194, and p70 all contained the region Ϫ70/ϩ51), whereas no inhibitory effect was observed on either the reporter construct p51 or the promoter-less construct pGL3-Basic (Fig. 1A). The reporter construct p51 by itself exhibited very little promoter activity in LS8 cells (Fig. 1A (14)). The effect of Msx2 on the full-length promoter (p2207) and the minimal promoter (p70) was further tested in LS8 cells with transient cotransfection assays. Transfection of different amounts of a Msx2 expression plasmid resulted in a roughly linear increase in the expression level of the exogenous Msx2 protein (Fig. 1B, middle panel, Msx2FL). Compared with that of the exogenous Msx2 protein (Msx2FL), the expression level of endogenous Msx2 protein (Msx2) in LS8 cells was very low (Fig. 1B, lower panel). In response to an increasing amount of exogenous Msx2 protein, the reporter gene activity was decreased up to 15-fold for p2207 and up to 7-fold for p70 (Fig. 1B), respectively. Furthermore, overexpression of Msx2 in LS8 cells had no effect on both the simian virus 40 (SV40) and the human cytomegalovirus immediate-early gene (CMV) promoter (data not shown). Taken together, these data indicated that Msx2 repressed the mouse amelogenin promoter in a dose-dependent manner in ameloblast-like LS8 cells.
C/EBP␣ and Msx2 Antagonized Each Other in Regulating Mouse Amelogenin Promoter Activity-C/EBP␣ has been demonstrated to function as a transcriptional activator of the mouse amelogenin gene through its cognate binding site in the Ϫ70/Ϫ52 region of the mouse amelogenin promoter (14). To investigate the functional relationship between C/EBP␣ and Msx2 in regulating the mouse amelogenin promoter activity, C/EBP␣ and Msx2 expression plasmids were cotransfected into LS8 cells with either amelogenin-promoter reporter construct p2207 or p70. Three different Msx2 expression plasmids were used in the study to generate amino-terminal FLAG epitope-tagged Msx2 proteins. Msx2FL was a full-length protein, including murine Msx2 residues 2-267. Msx2⌬N was an aminoterminal deletion containing residues 98 -267, whereas Msx2⌬C was a carboxyl-terminal deletion containing residues 2-183.
Msx2 Interfered with the Binding of C/EBP␣ to Its Cognate Site on Mouse Amelogenin Promoter-As a first step to understand the mechanism underlying the functional antagonism between Msx2 and C/EBP␣, the effect of Msx2 on the binding of C/EBP␣ to the mouse amelogenin promoter was assessed with an EMSA. Various forms of Msx2 protein were generated using an in vitro transcription-coupled translation system (TNT-coupled wheat germ extract system, Promega). A functional C/EBP cognate site has been identified in the Ϫ70/Ϫ52 region of the mouse amelogenin promoter (14). A 32 P-labeled doublestranded oligonucleotide containing the C/EBP cognate site (Fig. 4C, amel probe) was able to form a C/EBP␣-containing protein-DNA complex using nuclear extracts prepared from LS8 cells overexpressing C/EBP␣ protein (Fig. 4A, lane 12). Only modest changes in the intensity of the C/EBP␣-probe complex were observed when different amounts of either shamtreated TNT extracts (Fig. 4A, lanes 2 and 3) or TNT-expressed luciferase protein (Fig. 4A, lanes 10 and 11) were included in the EMSA reactions. On the contrary, TNT-expressed Msx2FL diminished the intensity of the EMSA complex to 20% (Fig. 4A,  lanes 4 and 5). Reduction in the intensity of the EMSA complex was also observed for the two deleted forms of Msx2, down to 25% for Msx2⌬N (Fig. 4A, lanes 6 and 7) and 37% for Msx2⌬C (Fig. 4A, lanes 8 and 9). Furthermore, increasing amounts of purified full-length Msx2 protein was capable of reducing the were cotransfected into LS8 cells, respectively. Similar parameters for transfection experiments were used for Msx2⌬N and Msx2⌬C, respectively. pCMV-lacZ was used as an internal control for transfection efficiency. The relative luciferase activity is the normalization of luciferase activity with ␤-galactosidase activity. The mean Ϯ S.D. from at least three independent experiments is represented, and the basal level of p2207 (A) and p70 (B) was set as 1, respectively.
To ascertain that both TNT-expressed and -purified Msx2 proteins were capable of binding to DNA, a 32 P-labeled doublestranded oligonucleotide containing a Msx2 cognate site (Fig.  4C, msx2 probe) was used in gel shift analyses. The TNTexpressed full-length (Msx2FL) and amino-terminal deleted (Msx2⌬N) Msx2 protein efficiently bound to the msx2 probe (Fig. 4C, lanes 7, 8, and 12), whereas the TNT-expressed carboxyl-terminal deleted Msx2 protein (Msx2⌬C) and luciferase protein failed to form an EMSA complex, respectively (Fig. 4C,  lanes 9 and 10). An EMSA complex with the msx2 probe was also observed for the purified Msx2 protein (Fig. 4C, lane 12). Neither of the various Msx2 proteins nor the luciferase control protein could bind to the amel probe that contained a C/EBP binding site instead of a Msx2 cognate site (Fig. 4C, lanes 2-5  and 11). The EMSA analyses indicated that Msx2 protein was able to interfere with the binding of C/EBP␣ to the mouse amelogenin promoter in a dose-dependent manner. Given the fact that Msx2 protein itself did not bind to the mouse amelogenin promoter, the observed interference appeared to result from protein-protein interaction instead of competition for binding to overlapping cognate sites on the promoter. Deletion of either amino-or carboxyl-terminal domain of Msx2 only had a modest effect on the ability of Msx2 to interfere with the binding of C/EBP␣ to the amelogenin promoter in the EMSA analyses (Fig. 4A). However, the same deletions resulted in a dramatic decrease in the antagonistic potency of Msx2 on C/EBP␣ in the transfection studies (Fig. 2).
Msx2 Interacted with C/EBP␣ in LS8 Cells-To investigate whether Msx2 could interact with C/EBP␣, a co-immunoprecipitation analysis was performed in LS8 cells. A C/EBP␣ expression plasmid was cotransfected into LS8 cells with an empty vector, Msx2FL, Msx2⌬N, and Msx2⌬C expression plasmid, respectively. Comparable amounts of C/EBP␣ protein were expressed in all four transfected cell populations (Fig. 5, lanes 1-4, panels III and IV). The C/EBP␣ protein was coimmunoprecipitated efficiently with Msx2FL (Fig. 5, lane 2, panel I) but to a lesser extent with Msx2⌬N (Fig. 5, lane 3, panel I). However, C/EBP␣ protein was barely detected in the Msx2⌬C-containing immunocomplex (Fig. 5, lane 4

, panel I).
A similar amount of FLAG-tagged Msx2 protein was present in each immunoprecipitated complex (Fig. 5, lanes 2-4, panel II), whereas no protein was detected in empty vector-transfected cells (Fig. 5, lane 1, panel II) using the same monoclonal anti-FLAG antibody (M2Ab) as that in the immunoprecipitation process. The reciprocal experiment was also performed, in which immunoprecipitation with a C/EBP␣-specific antibody was followed by Western blot analysis using the anti-FLAG M2Ab (data not shown). The Msx2FL protein was readily detected in the C/EBP␣-containing immunocomplex and so was Msx2⌬N to a less extent. However, the migration rate of the C/EBP␣-specific antibody light chain was very close to that of Msx2⌬C in SDS-polyacrylamide gel electrophoresis, thereby compromising the detection of Msx2⌬C band in Western blot analyses. These data indicated that Msx2 interacted with C/EBP␣ at the protein level in LS8 cells. Moreover, the carboxyl-terminal domain (residues 183-267) of Msx2 was required, whereas the amino-terminal domain (residues 2-97) of Msx2 contributed somehow to the interaction between Msx2 and C/EBP␣ protein. Taken together with the transfection data in Figs. 2 and 3, the carboxyl-terminal domain of Msx2 was indispensable for the repressive effect as well as the antagonism between Msx2 and C/EBP␣ on the promoter activity of the mouse amelogenin gene, whereas the amino-terminal domain of Msx2 played a less critical role. Complexes were separated by electrophoresis and visualized by autoradiography. The intensity of the EMSA complex was quantified by densitometry and plotted. The density level of the EMSA complex in the absence of TNT extract was set as "100." B, the same nuclear extracts (NE/␣) and probe as described above were utilized. His-tagged Msx2 proteins (His-Msx2) purified by nickel-nitrilotriacetic acid resin (Promega) were included in the EMSA reactions. Four different doses of His-Msx2 (50, 100, 150, and 200 ng) or 200 ng of bovine serum albumin were used. Complexes were separated by electrophoresis and visualized by autoradiography. The intensity of the EMSA complex was quantified by densitometry and plotted. The density level of the EMSA complex in the absence of His-Msx2 or bovine serum albumin was set as 100. C, a 32 P-labeled double-stranded oligonucleotide containing a consensus Msx2 binding site (msx2 probe) was used in addition to the probe described above (amel probe). Purified Msx2 protein (His-Msx2) and TNT extract expressing Msx2FL (TNT/FL), Msx2⌬N (TNT/⌬N), Msx2⌬C (TNT/⌬C), or luciferase protein (TNT/luc) was incubated with the two probes, respectively. Complexes were separated by electrophoresis and visualized by autoradiography. The sequences of the two probes were shown, with C/EBP-and Msx2 binding site in bold, respectively. The amelogenin probe (amel probe) was derived from the mouse amelogenin promoter (GenBank TM accession number AF083091).  Msx2⌬C). Whole cell lysates were prepared 24 h after transfection. Comparable amounts of cell lysate from four cell populations were electrophoresesed, transferred to Immobilon-P membrane, and immunoblotted with a C/EBP␣-specific antibody (Santa Cruz Biotechnology; panel III). Equal loading was demonstrated using an anonymous invariant band (panel IV). Equal amounts of protein from the four cell lysates were subject to immunoprecipitation with a monoclonal anti-FLAG antibody (M2Ab, Sigma). The immunoprecipitates were then electrophoresesed, transferred to Immobilon-P membrane, and immunoblotted with a C/EBP␣-specific antibody (panel I). After stripping, the same membrane was reprobed with the M2Ab (panel II). IP, immunoprecipitate; WB, Western blot. established by stably transfecting the amelogenin-promoter reporter construct p2207 into LS8 cells. The basal level of luciferase activity in these stable transfected cells was 1,000-to 10,000-fold of that in the parental LS8 cells (data not shown), indicating that the 2.2-kilobase mouse amelogenin promoter in the context of chromosome was very efficient in directing the expression of the reporter gene luciferase. LS8/p2207, one of the cell lines with the highest basal luciferase activity, was selected for further studies on the Msx2-and C/EBP␣-mediated regulation of the mouse amelogenin promoter. Transient transfection of a C/EBP␣ expression plasmid into LS8/p2207 cells resulted in a 3-4-fold increase in the reporter gene activity (Fig. 6A, lane 2), whereas little effect on the basal promoter activity was observed for each of the Msx2 constructs alone (Fig. 6A, lanes 3-5). Furthermore, cotransfection of Msx2FL with C/EBP␣ was able to decrease the C/EBP␣-mediated transactivation from 3.3-fold to 2-fold of the basal activity, whereas either Msx2⌬N or Msx2⌬C interfered little with C/EBP␣ function (Fig. 6A, lanes 6 -8).

Functional Antagonism between Msx2 and C/EBP␣ Was Observed on the 2.2-Kilobase Mouse Amelogenin Promoter in the
In our transient transfection experiments, 20 -30% transfection efficiency was consistently achieved. In other words, 70 -80% of the LS8/p2207 cells assayed for luciferase activity were not transfected with an expression plasmid, but these nontransfected cells still contributed to the background signal. To circumvent this problem, FACS analyses were performed. A ␤-galactosidase expression plasmid pCMV-lacZ was included in each transfection. After incubation with a fluorescent ␤-galactosidase substrate, di-␤-D-galactopyranoside, cells containing the plasmid pCMV-lacZ were sorted with flow cytometry and assayed for luciferase activity. Given the fact that pCMV-lacZ only comprised 5% of the total amount of the plasmid DNA in each transfection, the ␤-galactosidase positive cells should also contain the plasmid of interest (Msx2 or C/EBP␣). A 6 -7-fold increase and 75% decrease in reporter gene activity were observed for C/EBP␣ and Msx2FL, respectively. Furthermore, cotransfection of equal amounts of C/EBP␣ and Msx2FL expression plasmids gave rise to a 3-fold increase in reporter gene activity (Fig. 6B). Therefore, the 2.2-kilobase mouse amelogenin promoter in the chromosome context was responsive not only to the C/EBP␣-mediated activation and Msx2-mediated repression but also to the functional antagonism between C/EBP␣ and Msx2.
Little Functional Interaction with Msx2 Was Observed for Two Other C/EBP Family Members-To determine whether Msx2 is also able to interact with other C/EBP family members in regulating the mouse amelogenin promoter activity, C/EBP␤ and C/EBP␥ expression plasmid was transfected into LS8 cells together with a Msx2 expression plasmid, respectively. The amelogenin-promoter construct p2207 or p70 was used as a reporter in this study. C/EBP␣ not only potently activated the basal promoter activity but also efficiently overcame the Msx2mediated transcriptional repression of p2207 (Fig. 7A, lanes  1-6) or p70 (Fig. 7B, lanes 1-6). C/EBP␤ by itself weakly activated the basal promoter activity of the reporter construct, and an increasing amount of C/EBP␤ modestly antagonized Msx2 repression of p2207 (Fig. 7A, lanes 7-12) or p70 (Fig. 7B,  lanes 7-12). C/EBP␥ had little effect on either the basal promoter activity or the Msx2-mediated transcriptional repression of p2207 (Fig. 7A, lanes 13-18) or p70 (Fig. 7B, lanes 13-18). These cotransfection data indicated that C/EBP family members functioned differentially in the regulation of the mouse amelogenin promoter. Furthermore, only C/EBP␣, but not C/EBP␤ or C/EBP␥, was able to efficiently antagonize the repressive effect of Msx2 on the amelogenin gene in the functional analyses.

DISCUSSION
The Msx2-mediated Transcriptional Repression of the Mouse Amelogenin Gene Is DNA Binding Independent-We tested the hypothesis that Msx2 is involved in the regulation of mouse amelogenin gene expression. Amelogenin expression is ameloblast specific and spatiotemporally regulated during tooth development. The transcription of amelogenin is restricted to inner enamel epithelial cells that undergo terminal differentiation to the ameloblast phenotype (6 -8). In transgenic mouse analyses, the same 2.2-kilobase mouse amelogenin promoter as that used in the p2207 construct is able to recapitulate the spatiotemporal expression pattern of the endogenous amelogenin gene (10). Furthermore, the minimal promoter of the mouse FIG. 6. Functional antagonism between Msx2 and C/EBP␣ on the mouse amelogenin promoter in a stable cell line LS8/p2207. The reporter construct p2207 was transfected into LS8 cells to establish a stably transfected cell line LS8/p2207, in which the reporter gene luciferase is constitutively expressed at high level under the control of the 2.2-kilobase mouse amelogenin promoter. A, LS8/p2207 cells in 12-well plates were transfected, respectively, with equal amounts (500 ng) of empty vector (vector), C/EBP␣ expression plasmid (C/EBP␣) and Msx2 expression plasmids (Msx2FL, Msx2⌬N, and Msx2⌬C), or cotransfected with C/EBP␣ and Msx2 expression plasmid together (C/ EBP␣ϩMsx2FL, C/EBP␣ϩMsx2⌬N, and C/EBP␣ϩMsx2⌬C). pCMV-lacZ was used as an internal control for transfection efficiency. The relative luciferase activity is the normalization of luciferase activity with ␤-galactosidase activity. B, LS8/p2207 cells in 60-mm dishes were cotransfected with pCMV-lacZ (0.2 g) and expression plasmids for C/EBP␣, Msx2FL, or C/EBP␣ plus Msx2FL as described under "Materials and Methods." The ␤-galactosidase positive cells were sorted by FACS analysis and subjected to luciferase assay. The relative luciferase activity is the normalization of luciferase activity with protein concentration of the cell lysate. The mean Ϯ S.D. from three independent experiments is represented, and the level of luciferase activity in the presence of empty vector was set as 1.
amelogenin gene has been identified, in which a 19-nucleotide stretch (Ϫ71/Ϫ52 region) is required for the basal promoter activity (14).
The expression patterns of Msx2 during initiation and development of the murine teeth have been identified with in situ hybridization (30). In molar teeth, the expression pattern of Msx2 changes with the differentiation of each germ layer derivative, being strongly expressed in undifferentiated inner enamel epithelia but absent in differentiated ameloblasts and weakly expressed/absent in undifferentiated dental papilla mesenchyme but strongly expressed in odontoblasts and differentiated dental papilla cells. At the late bell stage (E18), moderate levels of Msx2 are expressed in the stratum intermedium and stellate reticulum (30). The reciprocal expression pattern of Msx2 and amelogenin in ameloblast cell lineage suggests that Msx2 may repress the expression of the mouse amelogenin gene.
Here, we demonstrate that Msx2 functions as a transcriptional repressor of the mouse amelogenin gene in a dose-dependent manner (Fig. 1). Two lines of evidence indicate that the primary repressive effect of Msx2 on amelogenin promoter is not DNA binding mediated. First, there is no Msx2 consensus binding site in the mouse amelogenin minimal promoter that is effectively repressed by Msx2. Second, Msx2 itself does not bind to the amelogenin minimal promoter as shown in Fig. 4C. The repression is most pronounced on the full-length promoter (p2207) and decreases gradually with the shortening of the promoter. This is consistent with the observation that the longer promoter constructs have lower basal activity than the minimal promoter does, which may be because of the endogenous Msx2 protein expressed in LS8 cells. Msx2 transcripts are detected in LS8 cells with reverse transcriptase-polymerase chain reaction (data not shown), and very low levels of endogenous Msx2 protein are expressed in LS8 cells as detected by an antipeptide antibody specific for Msx2 (Fig. 1B). During odontogenesis, Msx1 is expressed in all stages in dental mesenchymal cells but not in epithelial cells (28,29). In cotransfection assays, Msx1 can repress amelogenin-promoter reporter constructs only modestly in ameloblast-like LS8 cells (data not shown).
Functional Antagonism between Msx2 and C/EBP␣ Results from Protein-Protein Interaction-Msx2⌬C, the carboxyl-terminal deletion form of Msx2, interacts very poorly with C/EBP␣ in vivo, evidenced by the cotransfection and immunoprecipitation analyses in LS8 cells. However, Msx2⌬C is able to interfere with the binding of C/EBP␣ to its cognate site on the mouse amelogenin minimal promoter in vitro, although less potently than Msx2FL, the full-length protein. Residues 182-193 in murine Msx2 consist of homeodomain helix 3 that is the DNA recognition helix. The nuclear localization signal of Msx2 is overlapped with homeodomain helix 3. Deletion of residues 183-267, as in Msx2⌬C, not only results in the loss of DNA binding activity but also affects nuclear localization (31). Therefore, different subcellular localization most likely accounts for the inability of Msx2⌬C to antagonize C/EBP␣ in cotransfection assays, with Msx2⌬C in cytoplasm and C/EBP␣ in nucleus. Little C/EBP␣ protein is co-immunoprecipitated with Msx2⌬C from a whole cell lysate of LS8 cells in which both Msx2⌬C and C/EBP␣ are overexpressed. The weak interaction between Msx2⌬C and C/EBP␣ cannot withstand the stringent washing condition in co-immunoprecipitation assays; however, the high local concentration of Msx2⌬C in gel shift analyses likely enables Msx2⌬C proteins to interfere with the binding of C/EBP␣ to its cognate site. Therefore, in vivo, the nuclear localization signal in Msx2 is necessary for the functional antagonism between Msx2 and C/EBP␣. However, in co-immunoprecipitation assays, being accessible to each other alone is not sufficient for effective interaction between C/EBP␣ and Msx2 in the absence of Msx2 amino acid residues 183-267. Deletion of the amino-terminal domain (residues 2-97) of Msx2 attenuates, but does not abolish the interaction between Msx2 and C/EBP␣, suggesting that Msx2 amino acid residues 2-97 play a less critical role. In the future, it will be of interest to further delineate the domains(s) responsible for the interaction between Msx2 and C/EBP␣.
Msx2-mediated transcriptional repression has been extensively studied in osteoblasts. Gene repression by Msx2 is independent of the intrinsic DNA binding activity of the Msx homeodomain. Protein-protein interactions are essential to the repressive function of Msx2 (31)(32)(33). Two repressive mechanisms have been proposed, general repression and promoterspecific repression. Msx2 can bind to transcription factor F for RNA polymerase II (TFIIF), a component of the preinitiation complex, thereby repressing basal transcription (31). The DNA binding activity of other interacting partners is required to achieve promoter specificity for Msx2-dependent transcriptional repression (32)(33)(34)(35). Dlx5, another homeodomain transcription factor, up-regulates transcription of the osteoblast- , and 250 ng of pcC/EBP␣ with 250 ng of pcDNA3 were cotransfected into LS8 cells, respectively. Similar parameters for transfection experiments were used for C/EBP␤ and C/EBP␥, respectively. pCMV-lacZ was used as an internal control for transfection efficiency. The relative luciferase activity is the normalization of luciferase activity with ␤-galactosidase activity. The mean Ϯ S.D. from at least three independent experiments is represented, and the basal level of p2207 (A) and p70 (B) was set as 1, respectively. specific osteocalcin gene through binding to its cognate site. Msx2 antagonizes the function of Dlx-5 by forming a Msx2-Dlx5 heterodimer that cannot bind DNA (33). Msx2 also abrogates the induction of the osteocalcin promoter by fibroblast growth factor 2 through inhibiting a DNA binding activity to the fibroblast growth factor 2-response element without Msx2 itself binding to this element (32). However, the identity of this DNA binding activity remains unclear.
The mechanism underlying the Msx2-mediated transcriptional repression on the mouse amelogenin promoter appears to fall into the second category, in which Msx2 inhibits the DNA binding activity of C/EBP␣, a transcriptional activator of amelogenin promoter. By antagonizing C/EBP␣, Msx2 fulfills its role as a promoter-specific transcriptional repressor of the mouse amelogenin gene in ameloblasts.
Various transcription factors in concert with C/EBP␣ have been shown to synergistically activate the responsive promoters (37)(38)(39)(40)(41)(42). Negative regulation of C/EBP-mediated transactivation through protein-protein interaction has also been reported (43)(44)(45)(46). To our knowledge, Msx2, a homeodomain protein, is the first nonbasic region leucine zipper protein identified to date that functions as a transcriptional repressor through its interaction with C/EBP␣ protein. However, the nature of the interaction remains to be delineated. Our data demonstrate that the interaction between C/EBP␣ and Msx2 requires the carboxyl-terminal domain of Msx2, although the interaction is not mediated by the binding of Msx2 to the mouse amelogenin minimal promoter. Notably, this domain contains the third helix of the homeodomain, which is responsible for the binding of Msx2 to its cognate sites.
Transcription factors of the nuclear factor-B families have been reported to have interacted directly with C/EBP␤ via the Rel homology domain of nuclear factor-B and the basic leucine-zipper domain of C/EBP␤ (47,48). The glucocorticoid receptor, transcription factor v-myb or Sp1, interacts directly with C/EBP␤, resulting in synergistic activation of the target genes (49 -51). Among these factors, Sp1 synergizes with C/EBP␤ but not with C/EBP␣, suggesting that the C/EBP family members may interact differentially with Msx2. In the present study, Msx2 preferentially interacts with C/EBP␣ but not with either C/EBP␤ or C/EBP␥ in the regulation of the mouse amelogenin promoter activity.
A previous study has demonstrated an important role for C/EBP␣ in regulating the mouse amelogenin promoter (14). A C/EBP binding site (Ϫ70 to Ϫ61) is located at the minimal promoter of the mouse amelogenin gene and cotransfection of a C/EBP␣ expression plasmid transactivates amelogenin-promoter reporter constructs in a cell type-specific manner. Mutation or deletion of the C/EBP site within the amelogenin promoter not only results in the loss of C/EBP␣-mediated transactivation but also abolishes the basal promoter activity (Fig. 1A (14)). Furthermore, both the endogenous amelogenin gene and the stably transfected amelogenin-promoter reporter construct (p2207) in ameloblast-like LS8 cells are responsive to the transactivation mediated by C/EBP␣ (Fig. 6 (14)). These data indicate that C/EBP␣ is likely to play a critical role in regulating ameloblast-specific expression of the amelogenin gene.
Msx2 has been demonstrated to regulate cellular proliferation and differentiation during development (72)(73)(74)(75). Msx2 prevents differentiation and stimulates proliferation in primary cultured chick calvarial osteoblast (72). In transgenic mouse analyses, enhanced expression of Msx2 transiently inhibits osteoblast differentiation. As a consequence, the increase in osteoblast precursors in growth centers of the developing skull results in augmented bone growth and ultimately craniosynostosis (73). Tissue-specific gene expression during development has been shown to be regulated by Msx2. For example, Msx2 has been suggested to repress the expression of osteocalcin gene in the craniofacial skeleton at stages immediately preceding odontoblast and osteoblast terminal differentiation (76). It is conceivable that Msx2 may function in an analogous way to regulate the amelogenin gene during ameloblast differentiation.
In summary, we demonstrate that the functional antagonism between Msx2 and C/EBP␣ results from the Msx2-mediated interference with the binding of C/EBP␣ to its cognate site on the mouse amelogenin minimal promoter. Protein-protein interaction rather than competition for overlapping binding sites are responsible for the observed antagonism. Furthermore, the carboxyl-terminal residues 183-267 of Msx2 are required for the interaction, whereas the amino-terminal residues 2-97 play a less critical role. These data, together with the identification of C/EBP␣ as a transactivator of amelogenin gene in a previous study (14), support our interpretation that Msx2mediated repression and C/EBP␣-mediated activation operate in concert to regulate the spatiotemporal expression of amelogenin gene during tooth development.