Regulation of Hex gene expression by a Smads-dependent signaling pathway.

The homeobox gene Hex is expressed in multiple cell types during embryogenesis and is required for liver and monocyte development. Hex is expressed in the foregut region of late gastrula avian and mammalian embryos in a pattern that overlaps with expression of bone morphogenetic proteins (BMPs). Here we investigate the relationship between BMP signaling and Hex gene expression. We find that Hex expression in avian anterior lateral endoderm is regulated by autocrine BMP signaling. Characterization of the mouse Hex gene promoter identified a 71-nucleotide BMP-responsive element (BRE) that is required for up-regulation of Hex by an activated BMP signaling pathway. The Hex BRE binds Smad4 and Smad1-Smad4 complexes in vitro, and in transfection assays, it is responsive to Smad1 and Smad4 but not to Smad2 and Smad4 or Smad3 and Smad4. The BRE contains two copies of a GCCGnCGC-like motif that in Drosophila is the binding site for Mad and Madea followed by two CAGAG boxes that are similar to sequences required for transforming growth factor-beta/activin responsiveness of several vertebrate genes. Mutation of the GC elements, but not the two CAGAG boxes, abolishes Smads responsiveness in the intact Hex promoter, whereas mutations in both the GC elements and CAGAG boxes show that they act cooperatively to confer Smads responsiveness to the Hex promoter. The Hex BRE can confer Smads responsiveness to a heterologous promoter, and in this context, both the GC-rich elements and the CAGAG boxes are required for Smads-dependent promoter activity. An element almost identical to the Hex BRE is present within the BMP-responsive Nkx2-5 gene promoter, suggesting that the Hex BRE represents a common response element for genes regulated by BMP signaling in the foregut region of the embryo.

The transforming growth factor (TGF␤) 1 superfamily of signaling molecules, including bone morphogenetic proteins (BMPs), TGF␤s, activins, inhibins, and mullerian-inhibiting substance, regulate a wide range of biological processes during embryonic development and in the adult (1). BMPs were first identified due to their ability to induce cartilage and bone formation (2) and have subsequently been shown to regulate many aspects of embryogenesis, including early patterning of the embryo (3)(4)(5).
Like other TGF␤ family members, BMPs exert their biological effects by binding to heteromeric complexes of type I and type II serine/kinase receptors (6). The type II receptor kinase activates the type I receptor kinase, which in turn phosphorylates members of the Smad protein family that are responsible for transducing receptor-mediated signals to the nucleus (7). Smad proteins fall generally into three classes: receptor-regulated Smads (R-Smads); the co-Smad, Smad4; and the inhibitory Smads (I-Smads), Smad6 and Smad7. The R-Smads Smad2 and Smad3 are preferentially phosphorylated by TGF␤ and type I activin receptors, whereas BMPs preferentially induce phosphorylation of Smad1, Smad5, and Smad8. Phosphorylated R-Smads dissociate from the type I receptor, bind to Smad4 in the cytoplasm, and translocate to the nucleus. Smad complexes, in conjunction with various co-factors, then bind to DNA sequences in gene regulatory regions.
Promoter analyses have defined cis elements to which Smads can directly bind and have identified other DNA binding partners with which Smads cooperate to regulate gene transcription. In Drosophila, the vestigial and tinman genes are responsive to the BMP-related factor Dpp and bind the Drosophila R-Smad, Mad, at a GCCGnCGC motif in their promoters (8,9). This motif can also bind phosphorylated Smad1 in the presence of Smad4, and heterologous reporter genes containing concatamers of this sequence respond to BMP but not TGF␤/ activin signaling (10). A similar motif is responsible for BMP responsiveness of the mouse Smad6 and Id1 gene promoters (11)(12)(13). Other Smad binding elements include the sequence CAGAC, multimers of which are found in TGF␤-responsive promoter elements and are required for TGF␤ responsiveness (14 -16). Recent studies have shown that a CAGAC element can also mediate BMP responsiveness of the Xvent2 and Xvent2-B gene promoters (6,17). Binding site selection studies defined the palindromic sequence GTCTAGAC as the optimal binding site for Smad3 and Smad4 (16). Analysis of the crystal structure of the Smad3 MH1 domain bound to this palindrome identified the sequence GNCT, or its reverse complement AGNC, as crucial for Smad binding (18). This sequence, which is embedded within the CAGAC element, can bind Smad1, Smad3, and Smad4, although with somewhat varying affinities probably related to flanking sequences. An emerging view from these studies is that several related motifs are capable of binding multiple Smads with specificity arising through a combination of flanking sequences adjacent to the core binding sites and to the availability of co-factors (9). R-Smad-Co-Smad complexes in fact bind DNA poorly, although following interaction with other DNAbinding proteins, they associate with high affinity to promoter elements to regulate gene transcription (7, 9, 17, 19 -23).
Studies across the evolutionary spectrum have shown that BMP signaling plays important roles in many processes during early embryogenesis. In avian embryos, BMP2 is expressed in late gastrula anterior lateral (AL) endoderm and is required for up-regulation of several cardiogenic genes in the adjacent mesoderm (24 -34). AL endoderm itself gives rise to the liver, a process that requires signals from the cardiogenic mesoderm, including fibroblast growth factors (35)(36)(37)(38). Heart development in the mesoderm and liver development in the adjacent AL endoderm is therefore interrelated through reciprocal growth factor-mediated inductive interactions.
Recent studies have also shown that BMP signaling plays an important role during liver development. In mouse, BMP4 expressed in the septum transversum mesenchyme is required for expression of liver-specific markers, although the direct downstream targets of BMP signaling in the liver development pathway remain unknown. One potential target of BMP signaling is the homeobox containing the gene Hex, which is expressed in multiple cell types during embryonic development in patterns suggesting roles in anterior posterior axis formation, in vasculogenesis, and in the development of several endodermally derived organs including the liver and thyroid (39 -42). Hex is expressed in the foregut region of late gastrula chick and mouse embryos and is the earliest known marker for the liver primordium (42). Mutant mice lacking Hex expression die at around embryonic day 10.5 due to impaired liver development (41,43). In the late gastrula chick embryo, Hex (also known as Prh; 54) expression coincides with the domain of BMP2 expression in AL endoderm.
Here we have investigated the relationship between BMP2 signaling and Hex gene expression. We show that Hex expression in chick AL endoderm requires autocrine BMP2 signaling, and we identify and characterize a BMP-responsive element (BRE) within the mouse Hex gene promoter. The BRE contains two GC-rich elements and two CAGAG boxes that act cooperatively to confer BMP responsiveness the Hex promoter. A similar BRE element is present within the BMP-responsive Nkx2-5 gene promoter (52,55), suggesting that the Hex BRE represents a common response element for genes regulated by BMP signaling in the foregut region of the embryo.

EXPERIMENTAL PROCEDURES
Isolation of the Mouse Hex Promoter-The 5Ј region of the mouse Hex gene was isolated by screening one million recombinants from a mouse 129SVJ genomic library (Stratagene) using a PCR product containing exon 1 of the mouse Hex gene. A single positive clone, mHex-14, containing a 15.8-kb insert, including 8.7 kb upstream of transcription initiation, was selected. The plasmid pBSHex-14 was generated by this insert into the NotI site of pBluescript KS (ϩ) (Stratagene).
Cell Culture and DNA Transfection-C3H10T1/2 and P19 cells were maintained in Dulbecco's modified Eagle's medium with 10% fetal bovine serum. For transient transfection, 80% confluent cells were transfected in 24-well plates using LipofectAMINE (Invitrogen) according to the manufacturer's instructions. For transfection in C3H10T1/2 cells, Hex promoter-luciferase reporter constructs were transfected alone or with various combinations of expression plasmids for human Smad1, Smad4 (gift of R. Derynck), Smad2, Smad3 (gift of X. Feng), and the constitutively activated BMP type I receptor ALK3 (Q233D; gift of J. L. Wrana). Luciferase activity was assayed 24 h after transfection using the Dual-Light luciferase assay kit (Tropix). Values were normalized with ␤-galactosidase activity expressed from the pCMV ␤-galactosidase plasmid (Clontech). P19 cells were transfected alone or with various combinations of expression plasmids for human Smads 1-4. Following transfection, cells were maintained in low serum medium and were treated with BMP2 and TGF␤ as described (56). Luciferase activity assays were performed 18 h later.
Chick Embryo Culture-Embryos were removed from fertile chick eggs (Gallus domesticus, Rosemary Farms, Santa Maria, CA or SPAFAS, Inc., Preston, CT), incubated until they reached Hamburger-Hamilton stage 5, and then placed in New culture (53). Aggregates of control or noggin-expressing CHO cells (gift of R. Harland) were prepared by placing 2,000 -5,000 cells in hanging drop culture for 3-5 days in Dulbecco's modified Eagle's medium plus 10% fetal bovine serum. Resulting aggregates were implanted into the ventral side of the embryo through a small slit cut into the endoderm. Embryos were incubated for 6 -8 h and then processed for whole mount in situ hybridization as described (42). Digoxigenin-labeled antisense cHex probe was generated according to the manufacturer's instructions (Roche Molecular Biochemicals) from full-length chicken Hex (Prh) cDNA (gift of G. H. Goodwin, Institute of Cancer Research, London, UK).
Electrophoretic Mobility Shift Assays-The Ϫ493/Ϫ345 Hex promoter fragment was amplified by PCR and labeled with ␥-32 P using T4 polynucleotide kinase. Binding reactions were performed as described previously with minor modifications (4). Fusion constructs for GST-Smad1 and GST-Smad4 and the isolation of bacterially expressed Smads have been described (15). Briefly, 20-l reactions containing 25 mM Tris-HCl, pH 8.0, 50 mM KCl, 5 M MgCl 2 , 1 mM dithiothreitol, 10% glycerol, and 400 ng of poly(dI⅐dC) were prepared. 200 ng of purified GST-Smad1 or GST-Smad4 was added separately or together to each binding reaction, and reactions were incubated at 4°C for 60 min. For super gel shift, 1 l of anti-Smad1 or Smad4 antibody (Santa Cruz Biotechnology) was added subsequently to addition of labeled probe and incubated at 4°C for 60 min. Complexes were resolved on a 4% polyacrylamide gel and analyzed by autoradiography

Hex Gene Expression in Avian Anterior Lateral Endoderm
Requires BMP Signaling-The homeobox gene Hex is expressed in the AL endoderm of late gastrula avian embryos in a pattern that overlaps with BMP2 gene expression (32,33,42). To investigate the relationship between BMP2 signaling and Hex gene expression in AL endoderm, aggregates of CHO cells expressing the BMP antagonist noggin were implanted into the anterior region of chick embryos at the late gastrula stage (stage 5). Embryos were then allowed to develop for 6 h. Whole mount in situ hybridization showed that Hex transcripts were absent from a region within several hundred micrometers of noggin-expressing CHO cells, whereas control CHO cells had no effect on Hex gene expression (Fig. 1). Since BMP2 is expressed in the foregut region endoderm, these results demonstrate that Hex expression in AL endoderm requires autocrine BMP signaling.
Responsiveness of the Hex Gene Promoter to Smads-To investigate transcriptional regulation of Hex by BMP signaling, a 8.7-kb region of the mouse Hex gene upstream of a previously identified transcription initiation site (46) was isolated from a phage genomic library and subcloned into the pGL3-Basic luciferase reporter plasmid (Hp8.7-Luc). To determine whether this Hex promoter fragment contained BMP-responsive elements, co-transfection assays were performed in P19 cells. Although treatment of Hp8.7-Luc-transfected P19 cells with BMP2 stimulated promoter activity only slightly above control levels ( Fig. 2A), co-transfection of Hp8.7-Luc plus Smad1 and Smad4 expression vectors, followed by addition of BMP2 to the culture medium, increased luciferase levels ϳ2.8-fold. TGF␤ signaling preferentially utilizes Smads 2, 3, and 4 (9). Cells transfected with Hp8.7-Luc plus Smad2 and Smad4 expression vectors and exposed to TGF␤ showed only a modest (1.2-fold) increase in promoter activity.
Additional transfection analyses in C3H10T1/2 mesenchymal cells further illustrated the selective specificity of Hp8.7 Luc to different Smad proteins. Transfection of Hp8.7-Luc plus the BMP type I ALK3 receptor (Q233D), alone or in combination with Smad1 or Smad4, resulted in a consistent slight reduction of luciferase activity versus Hp8.7 alone (Fig. 2B). When Hp-8.7-Luc was co-transfected with all three expression plasmids (Smad1, Smad4, and ALK3), however, luciferase activity levels were enhanced more than 40-fold. As has been observed with other Smads-responsive promoters (45), an equivalent enhancement of luciferase activity was observed in the absence of the ALK3 (Q233D) receptor. Co-transfection of Hp8.7Luc with Smad2 and Smad4 or Smad3 and Smad4 failed to increase luciferase activity above background levels, indicating that the ability of Smads to up-regulate Hex promoter activity was limited to Smad1 and Smad4. Taken together, these promoter analyses indicate that cis elements within 8.7 kb upstream of transcription initiation are preferentially responsive to a Smad1-Smad4-dependent BMP signaling pathway.
To further define Smads-responsive elements in the Hex promoter, various promoter segments were cloned upstream of the TK minimal promoter (Fig. 3A) and then co-transfected into 10T1/2 cells alone or with Smad1, Smad4, and ALK3 (Q233D) expression plasmids. Although constructs containing promoter fragments between Ϫ8736 and Ϫ2094 failed to confer Smads responsiveness to the TK promoter (Fig. 3B), when a promoter fragment between Ϫ2099 and Ϫ12 linked to the TK promoter was co-transfected with Smad1, Smad4, and ALK3, luciferase activity was increased more than 30-fold. A construct in which the Hex promoter fragment was deleted to Ϫ833 retained more than 26-fold induction, indicating that one or more BMPresponsive elements are located within this proximal upstream promoter region.
Deletion analysis of the intact Hex promoter confirmed these findings and further defined the location of BREs. Although the heterologous promoter assays above demonstrated that sequence upstream of Ϫ2099 contained no identifiable Smads-responsive regions, progressive deletion of the Hex promoter showed that upstream regions were nevertheless capable of modulating activity of more proximally located BRE(s) (Fig. 4). Progressive truncation of the proximal promoter region further localized BRE(s) to within 500 nucleotides of transcription initiation. A promoter fragment truncated at Ϫ493 retained full responsiveness, whereas deletion to Ϫ406 reduced responsiveness by ϳ80%, and deletion to Ϫ321 completely abolished responsiveness to the BMP signaling pathway.
Analysis of Hex promoter sequence between Ϫ493 and Ϫ321 revealed several motifs showing homology to previously identified Smads binding elements (Fig. 5A). Two GC-rich elements, located at Ϫ448 (GC1) and Ϫ428 (GC2), show homology with the sequence GCCGnCGC that has been shown to bind Drosophila Mad and Madea and mammalian Smad1 and Smad4 (9,11). GC1 contains two nucleotide differences, and GC2 contains a single nucleotide difference, from the Drosophila consensus sequence. GC2 is also identical to an element that is responsible for BMP responsiveness of the mouse Smad6 promoter (13). Two CAGAG boxes (CA1 and CA2) lo- Aggregates of control or noggin-expressing CHO cells were inserted into the anterior lateral regions of late gastrula (stage 5) embryos. Embryos were incubated for 6 -8 h and then processed for whole mount in situ hybridization to localize Hex mRNAs. A, the rostral portion of a control embryo showing typical Hex mRNA localization to AL endoderm. B, an embryo containing control (arrowhead) or noggin-expressing (asterisks) CHO cell aggregates. Note that Hex mRNAs are absent on the side of the embryo containing the noggin-expressing cells. Aggregates of control CHO cells had no effect on Hex expression (arrowhead). cated at Ϫ411 and Ϫ385 are similar to Smad binding elements found in human and Xenopus BMP-responsive gene promoters (7,15,16).
To investigate the potential role of these elements in conferring BMP responsiveness to the Hex promoter, mutations in Hp-493 were made within each element, alone or in combination. Mutations within GC1 reduced Smads-dependent promoter activity ϳ40% (Fig. 5B), whereas mutations within GC2 reduced promoter activity by ϳ60%. Combined mutations within GC1 and GC2 reduced promoter activity to 18% of wild type. Mutations within CA1 and CA2, individually or in combination, had no effect on promoter activity. Combining CA box mutations with a mutation in GC2, however, reduced promoter activity ϳ50% below levels observed with the GC2 mutation alone. These findings show that GC1 and GC2 are important for Smads-dependent promoter activity and that in the presence of Smads, they can act cooperatively with the CAGAG boxes to increase transcriptional activity.
To determine whether the promoter region containing these motifs can confer Smads responsiveness to a heterologous promoter, concatamers containing three copies of the promoter sequence from Ϫ454 to Ϫ403, containing GC1, GC2, and CA1, or sequence from Ϫ437 to Ϫ369, containing GC2, CA1, and CA2, were cloned upstream of the TK minimal promoter. Although the TK promoter alone (pTKGL3) showed no up-regulation following co-transfection with the Smads expression vectors, 3WTTK, containing three wild type concatamers of the sequence from Ϫ454 to Ϫ403, was up-regulated 22-fold (Fig.  6A). Mutation of GC1 (3GC1m) reduced responsiveness ϳ65%, whereas mutation of GC2 (3GC2m) reduced responsiveness to less than 20% of the wild type concatamer. Combined mutations within GCm1 and GCm2 abolished Smads responsiveness. In contrast to results obtained with the intact promoter, mutations of CA1 alone also abolished Smads responsiveness.
Similar results were obtained with concatamers of Ϫ437 to Ϫ369 (Fig. 6B). Whereas the wild type concatamer was upregulated 8-fold when co-transfected with Smads expression vectors, mutation of GC2 or CA1 and CA2 abolished Smads responsiveness. The BRE located between Ϫ437 and Ϫ369 can therefore confer Smads responsiveness to a heterologous promoter, and in the context of this heterologous promoter construct, both the GC elements and the two CAGAG boxes are required for Smads-dependent transcriptional activity.
Smads Binding to the Hex BRE-We next examined the ability of Smad1 and Smad4 to bind directly to a promoter fragment containing the BRE in an electrophoretic mobility shift assay. When a Hex promoter fragment from Ϫ493 to Ϫ345 was combined with GST-Smad1 protein, no shifted bands were observed (Fig. 7). Combining GST-Smad-4 with the probe, however, produced two shifted bands, indicating that the Hex BRE binds purified Smad4 but not Smad1. When both GST-Smad fusion proteins were combined with probe, an additional shifted complex was observed, and all three Smads-specific bands were supershifted following addition of Smad4 antibody. Although mutation of either GC2 or the two CAGAG boxes within the Ϫ493 to Ϫ345 promoter fragment had no effect on Smads binding, when these mutations were combined, Smadsspecific binding was reduced. Finally, mutations within both GC1 and GC2 abolished Smads-dependent binding to the Hex BRE. DISCUSSION We have shown that the homeobox-containing gene Hex requires autocrine BMP signaling for expression in chick AL (foregut) endoderm. In transfection assays, the Hex promoter is FIG. 3. Heterologous promoter assays localize Smads-responsive elements to within 833 nucleotides of transcription initiation. As shown in A, Hex promoter fragments from Ϫ8736 to Ϫ4232, Ϫ8736 to Ϫ6680, Ϫ6679 to Ϫ4232, Ϫ4237 to Ϫ2094, Ϫ2099 to Ϫ12, or Ϫ833 to Ϫ12 were cloned immediately upstream of the mouse TK minimal promoter driving luciferase. B, luciferase activities of various Hex TK promoter constructs following transfection alone (Control) or with Smad1, Smad4, and ALK3 expression plasmids into C3H10T1/2 cells. Cell extracts were assayed for luciferase activity 24 h following transfection.

FIG. 4. Deletion analysis of the Hex gene promoter.
Hex promoter fragments containing progressive deletions between Ϫ8736 and Ϫ321 were cloned upstream of the luciferase gene and assayed for Smads responsiveness by transfection into C3H10T1/2 cells alone or with Smad1, Smad4, and ALK3 expression plasmids. Cultures were assayed 24 h later for luciferase activity. responsive to Smad1 and Smad4 but not to Smad2 and Smad4 or Smad3 and Smad4. Characterization of the mouse Hex promoter has identified a BRE located between Ϫ448 and Ϫ378 that is required for up-regulation by an activated BMP signaling pathway. The Hex BRE contains two GC-rich elements and two CAGAG boxes that act in concert to bind Smad4 and complexes of Smad1 and Smad4 to activate transcription.
Relatively few BMP-responsive promoters have been characterized in vertebrates (9), and consequently a clear picture of the cis elements involved in transcriptional regulation by Smad1, Smad5, or Smad8 is still emerging. BMP responsiveness of the mouse Smad6 promoter has been localized to a GC-rich motif that is similar to the consensus sequence GC-CGnCGC (13) that binds Drosophila Mad and is required for transcriptional activity of several Dpp-responsive genes (10,11). The sequences GCCGCCCG and GCCGGCGGC within the Hex BRE are similar to the Dpp consensus sequence, and several lines of evidence argue that these sequences are directly involved in conferring BMP responsiveness to the Hex gene. First, deletion and heterologous promoter studies localize BMP responsiveness to a region between Ϫ493 and Ϫ406, which includes these GC elements. Second, mutation of these elements within Hp493 reduces responsiveness by more than 80%, whereas mutations of these elements within a heterologous promoter construct containing three copies of the Hex BRE linked to the TK minimal promoter abolishes BMP re-sponsiveness. Lastly, gel shift studies show that these elements are required for binding of Smad4 and Smad1-Smad4 complexes to the BRE in vitro.
Two copies of the sequence CAGAG (CA1, CA2) present within the Hex BRE are similar to the CAGAC box identified as a Smad consensus binding site (14,16). The sequence CAGAC binds complexes of Smad3 and Smad4 in TGF␤/activin-responsive promoters (15) and contains the reverse complement sequence GTCT identified by binding site selection as the optimal Smad3-Smad4 binding motif (16). The CAGAC motif can also participate in BMP responsiveness as a single CAGAC element is required for BMP responsiveness of the Xenopus Xvent-2 and Xvent-2B promoters (5,7). The CAGAG boxes in the Hex promoter appear to play a subtler role in augmenting BMP responsiveness. Although mutation of CA1 and CA2, alone or in combination, has no effect on Smads responsiveness of the intact Hex promoter, when these mutations are combined with mutations within GC2, responsiveness is reduced an additional 50% below levels observed with the GC2 mutation alone. In contrast, when three copies of the sequence from Ϫ454 to Ϫ403, containing GC1, GC2, and CA1, are placed upstream of the TK minimal promoter, mutation of CA1 abolishes responsiveness, similar to mutations within the two GC elements. Similar results were obtained using a second heterologous promoter construct containing sequence between Ϫ439 and Ϫ369 (GC2, CA1, and CA2); mutation of CA1 and CA2 abolishes Smads responsiveness. One explanation for the differential effects of CA mutations in the intact promoter versus the concatamer constructs is that several CAGA sequences are located proximal to the BRE in the intact promoter (Hp493) that could act redundantly to the CAGAG boxes in the BRE. In contrast, no additional GC boxes have been identified outside of the BRE anywhere within the 8.7-kb promoter fragment, and so mutations within the BRE GC boxes cannot be compensated for.
Electrophoretic mobility shift assays using a promoter fragment from Ϫ493 to Ϫ345 demonstrate that the Hex BRE binds purified Smad4 and complexes of Smad1 and Smad4 but not Smad1 alone. Mutation of either GC2 or the two CAGAG elements has little effect on Smads binding, although mutation of all three sites significantly reduces the intensity of shifted bands, once again suggesting cooperativity between GC boxes and CAGAG elements. Mutations within both GC1 and GC2 abolishes Smads binding. These results suggest either that the elements bind Smad4, whereas Smad1 does not directly contact DNA, or that Smad1 binds to the BRE with low affinity that must be stabilized by binding to Smad4.
Taken together, transfection results and gel shift analyses indicate that the Hex BRE is responsible for conferring BMP responsiveness to the Hex promoter and that although the GC-rich elements play a more dominant role in Smads binding and BMP responsiveness, they nevertheless act in a cooperative fashion with CAGAG elements, perhaps both within the BRE and in other regions of the promoter. A recent transgenic analysis of Hex gene regulation in mice identified a promoter region upstream of transcription initiation that is required for Hex expression in the thyroid, in liver, and in endothelial cells (51). The Hex BRE that we have identified falls within this region.
BMP signaling is essential for many aspects of embryogenesis, including development of the heart (24 -26). In amphibians, birds, and mammals, AL endoderm (or endomesoderm) produces BMP ligands that activate cardiogenic genes such as Nkx2-5 in responsive mesoderm (27,29,(31)(32)(33). AL endoderm also forms portions of the gut and gives rise to several endodermally derived organs, including the liver (48). Fibroblast growth factors produced by the cardiogenic mesoderm are required for hepatocyte specification (35,36,38,47,49,50), and Rossi et al. (56) have recently shown that in mouse, BMP4 expressed in the septum transversum mesenchyme is required for expression of differentiation-specific liver genes such as albumin. Here we show an earlier requirement for BMP signaling during hepatogenesis through autocrine regulation of Hex, the earliest known regulatory gene in the liver development pathway. Hex is required for liver development as knockout mice show greatly reduced expression of transcription factors such as Hnf3b, Hnf6, Hnf4␣, and Hnf1 and fail to express liver-specific genes such as ␣-fetoprotein or serum albumin (41,43). Our results showing that Hex expression in foregut endoderm requires autocrine BMP signaling reveals a crucial role for BMPs at the top of a hierarchy of gene expression leading to hepatocyte differentiation. BMP signaling is therefore required for activation of regulatory genes that govern early stages of both heart and liver development. For chick, paracrine BMP signaling activates cardiogenic genes in the mesoderm, whereas autocrine BMP signaling activates Hex in the endoderm. It is interesting to note that the mouse Nkx2-5 promoter contains a Smads consensus regulatory region at Ϫ3059 that is similar to the Hex BRE, containing a GC box 16 and 27 nucleotides upstream of two CAGA boxes (52). This consensus regulatory region is required for initial expression of Nkx 2.5 in the cardiogenic crescent and is responsive to BMP signaling. The Hex and Nkx2.5 Smads consensus regulatory elements may therefore define a common BRE for genes regulated by BMP signaling in the foregut region of the embryo. It will be interesting to determine whether similar elements are present in other genes regulated by BMP signaling during heart and liver development. FIG. 7. Binding of Smads to the Hex BRE. Electrophoretic mobility shift assays were performed using a Hex promoter fragment from Ϫ493 to Ϫ345, either wild type or containing the mutations shown in Fig. 6A in GC2 (GC2m), the two CAGA boxes (CA1m CA2m), or both sets of mutations combined (GC2m CA1m CA2m). A probe containing mutations in both GC1 and CC2 (GC1CG2) was also generated. 32 Plabeled DNA was combined with 200 ng each of GST-Smad1, GST-Smad4, or both proteins, as indicated, incubated at 4°C for 1 h, and resolved on a 4% polyacrylamide gel. Asterisks denote specific shifted complexes. Antibodies specific for Smad4 were added to some reactions.
FIG. 6. Concatamers of the Smads element confer responsiveness to a heterologous promoter. Wild type or mutated concatamers of BRE sequence from Ϫ454 to Ϫ403, containing GC1, GC2 and CA1 (A), or sequence from Ϫ437 to Ϫ369, containing GC2, CA1 and CA2 (B), were linked to the TK minimal promoter driving luciferase. Constructs were transfected into C3H10T1/2 alone (Control) or with ALK3, Smad1, and Smad4 expression plasmids, and cell extracts were assayed 24 h later for luciferase activity.