Physical Interaction between GATA-5 and Hepatocyte Nuclear Factor-1α Results in Synergistic Activation of the Human Lactase-Phlorizin Hydrolase Promoter*

GATA-4, -5, and -6 zinc finger and hepatocyte nuclear factor-1α (HNF-1α) homeodomain transcription factors are expressed in the intestinal epithelium and synergistically activate the promoter of intestinal genes. Here, we demonstrate that GATA-5 and HNF-1α physically associate both in vivo andin vitro and that this interaction is necessary for cooperative activation of the lactase-phlorizin hydrolase promoter. Furthermore, physical association is mediated by the C-terminal zinc finger of GATA factors and the homeodomain of HNF-1α. Deletion of HNF-1α activation domains or interruption of HNF-1-binding sites in the lactase-phlorizin hydrolase promoter resulted in a complete loss of cooperativity, whereas deletion of GATA-5 activation domains or interruption of GATA-binding sites resulted in a reduction, but not an elimination, of cooperativity. We hypothesize that GATA/HNF-1α cooperativity is mediated by HNF-1α through its activation domains, which are oriented for high levels of activation through binding to DNA and physical association with GATA factors. These data suggest a paradigm whereby intestine-specific gene expression is regulated by unique interactions among tissue-restricted transcription factors coexpressed in the intestine. Parallel mechanisms in other tissues as well as in Drosophila suggest that zinc finger/homeodomain interactions are an efficient pathway of cooperative activation of gene transcription that has been conserved throughout evolution.

GATA-4, -5, and -6 zinc finger and hepatocyte nuclear factor-1␣ (HNF-1␣) homeodomain transcription factors are expressed in the intestinal epithelium and synergistically activate the promoter of intestinal genes. Here, we demonstrate that GATA-5 and HNF-1␣ physically associate both in vivo and in vitro and that this interaction is necessary for cooperative activation of the lactasephlorizin hydrolase promoter. Furthermore, physical association is mediated by the C-terminal zinc finger of GATA factors and the homeodomain of HNF-1␣. Deletion of HNF-1␣ activation domains or interruption of HNF-1binding sites in the lactase-phlorizin hydrolase promoter resulted in a complete loss of cooperativity, whereas deletion of GATA-5 activation domains or interruption of GATA-binding sites resulted in a reduction, but not an elimination, of cooperativity. We hypothesize that GATA/HNF-1␣ cooperativity is mediated by HNF-1␣ through its activation domains, which are oriented for high levels of activation through binding to DNA and physical association with GATA factors. These data suggest a paradigm whereby intestine-specific gene expression is regulated by unique interactions among tissuerestricted transcription factors coexpressed in the intestine. Parallel mechanisms in other tissues as well as in Drosophila suggest that zinc finger/homeodomain interactions are an efficient pathway of cooperative activation of gene transcription that has been conserved throughout evolution.
The intestinal epithelium is a dynamic structure that undergoes a highly regulated process of cell division, migration, cell fate determination, and differentiation (1)(2)(3). During intestinal development, interactions between visceral endoderm and mes-oderm at E8 1 in mice result in the formation of a primitive foregut that rapidly undergoes cytodifferentiation, so that by E19, an epithelial monolayer overlies nascent villi. During the first 2 weeks of postnatal life, a proliferating compartment develops into the crypts of Lieberkü hn. Stem cells located near the base of crypts rapidly divide and give rise to four terminally differentiated cell types, which migrate both basally and apically. Cells migrating to the base of crypts become Paneth's cells, whereas those migrating up the crypt toward villi become absorptive enterocytes, goblet cells, and enteroendocrine cells. At the crypt-villus junction, the proliferative phase ends, and cells acquire a differentiated phenotype characterized by the synthesis of functionally relevant proteins. The cells continue to migrate up the villi, enter an apoptotic cycle, and are shed into the intestinal lumen ϳ3 days after their initial appearance on villi. The molecular mechanisms underlying the dynamic processes of intestine-specific gene expression and cellular differentiation during development are poorly understood.
Absorptive enterocytes compose ϳ95% of epithelial cells on villi and are the cells responsible for the terminal digestion and absorption of nutrients. Lactase-phlorizin hydrolase (LPH), the enzyme critical for the digestion of milk lactose, is an absorptive enterocyte-specific protein that serves as a marker for intestine-specific gene expression and intestinal differentiation (4,5). In rats, LPH mRNA is detected as early as E18 in the proximal intestine when primitive villi are formed (6). LPH expression is highest at birth and continues to be highly expressed throughout the suckling period. After weaning, LPH expression per enterocyte is reduced and is also restricted to the jejunum and proximal ileum (4,5). This developmental decline also occurs in humans at around age 5, although a subset of the human population continues to synthesize high levels of LPH throughout adulthood (7,8). The close correlation between the lactase activity and its mRNA in rats (4,9) and humans (10,11) and transcription rate experiments (4) indicate that LPH is regulated mainly by gene transcription.
Transgenic studies indicate that information for enterocytespecific LPH gene expression in vivo is contained in the 5Ј-flanking region (12)(13)(14). Identification of specific binding sites within the first 100 bp of 5Ј-flanking region in the LPH genes of several species has led to the demonstration that GATA, hepatocyte nuclear factor-1␣ (HNF-1␣), and Cdx-2 transcription factors are activators of the LPH promoter (15)(16)(17)(18)(19)(20)(21). Due to the close proximity of GATA and HNF-1 binding sites on the LPH promoter and the nearly exclusive coexpression of these two families of transcription factors in the intestinal epithelium (22)(23)(24)(25)(26)(27)(28)(29), we hypothesized that members of the GATA and HNF-1 families of transcription factors interact to modulate LPH gene expression.
The GATA family of transcription factors has been implicated in cell lineage differentiation during vertebrate development. Defined by two evolutionarily conserved zinc fingers of the motif Cys-X 2 -Cys-X 17 -Cys-X 2 -Cys that mediate binding to the consensus DNA sequence WGATAR (where W ϭ A or T, and R ϭ A or G), the GATA family is generally categorized into two classes based on expression patterns and amino acid homologies. GATA-1, -2, and -3 are expressed in developing bone marrow cells and are critical for hematopoiesis (30), whereas GATA-4, -5, and -6 have a more diverse pattern of expression that includes the small intestine, heart, liver, lungs, and gonads (22)(23)(24)(31)(32)(33). The GATA-4, -5, and -6 subfamily has been shown to modulate promoter function of intestinal genes, including the rat and human LPH (16,17,21), human sucraseisomaltase (21), and Xenopus intestinal fatty acid-binding protein (25) genes. Although the wide-ranging expression patterns of GATA-4, -5, and -6 argue against these proteins as being master regulators of tissue or cell type-specific gene expression, there is increasing evidence that this subfamily might be critical in regulating cell-specific gene expression through unique interactions with other semirestricted transcription factors and cofactors (34).
HNF-1␣ is a member of a transcription factor family that contains a modified homeodomain and that binds as a dimer to the consensus sequence GTTAATNATTAAC (35,36). Originally thought to be liver-specific, HNF-1␣ is expressed in the intestinal epithelium (26 -29, 37) and has been shown to modulate the promoter of many genes expressed in the intestine (18, 20, 21, 38 -45). We have recently shown that GATA subfamily members and HNF-1␣ synergistically activate the LPH and sucrase-isomaltase promoters (21), suggesting that members of these two transcription factor families interact to produce high levels of enterocyte-specific gene expression.
The goal of these experiments was to characterize the role that GATA and HNF-1␣ transcription factors play in regulating intestine-specific gene expression by defining the mechanism by which these two transcription factors function to synergistically activate the human LPH promoter. In these studies, the importance of critical structures in GATA factors and HNF-1␣, including domains responsible for protein/protein interaction, DNA-binding domains, and activation domains, for cooperative activation was determined. Using GATA-5 as a model, the results of this study reveal that physical association between GATA-5 and HNF-1␣ is required for the synergistic activation of the human LPH promoter and that this interaction occurs through the C-terminal zinc finger and basic regions of GATA-5 and the homeodomain of HNF-1␣. Identification of parallel mechanisms of protein/protein interaction between zinc finger and homeodomain proteins in other vertebrate tissue (46 -48) as well as in Drosophila (49,50) suggests that zinc finger/homeodomain interactions are an efficient mechanism for synergistic activation of gene transcription that has been conserved throughout evolution.
For in vivo protein/protein interaction studies, an expression vector for FLAG-tagged GATA-5 (G5-FLAG) was constructed by fusing the GATA-5 coding region to that of FLAG in the pFLAG-CMV-2 expression vector (Sigma). This was done using site-directed mutagenesis (52) by inserting a second EcoRI site at the 5Ј-side of the GATA-5 coding region (mutagenic oligonucleotide, 5Ј-CTTTGGTACATGGAATTCGAGAGCT-CCCAAC-3Ј), resulting in EcoRI sites flanking the open reading frame. The GATA-5 coding region was then subcloned in-frame into the EcoRI site of pFLAG-CMV-2, oriented by restriction digests, and confirmed by sequencing.
For transfection studies, the human LPH promoter containing 118 bp of 5Ј-flanking region was fused 5Ј to the human growth hormone reporter (called h118wt) as previously described (21). This region contains two GATA sites and an HNF-1 site as previously described (21). Transcription factor/DNA interactions were characterized, in part, using human promoter-reporter constructs containing mutations in the two GATA sites (h118mG1G2), in the HNF-1 site (h118mH), and in all three sites together (h118mG1G2H). Mutations introduced into these sites have been previously shown to disrupt specific protein/DNA interactions (21). pRC-CMV (Invitrogen) served as a negative control expression vector for all cotransfection experiments.
Immunoprecipitation of FLAG-tagged Protein-In vivo GATA-5/ HNF-1␣ interactions were determined by supershift electrophoretic mobility shift assay (EMSA) detection of HNF-1␣ in proteins from COS-7 cells cotransfected with G5-FLAG and H1␣-CMV and immunoprecipitated with anti-FLAG antibodies. COS-7 cells were cotransfected at 70% confluence with G5-FLAG and H1␣-CMV using Effectene TM reagent (QIAGEN Inc.) according to the manufacturer's protocol. After 2 days, confluent 100-mm plates were washed using 10 ml of cold phosphate-buffered saline and scraped in 1 ml of phosphate-buffered saline. Cells were then pelleted in a microcentrifuge at 4°C, and supernatants were discarded. The pelleted cells were resuspended in 150 l of lysis buffer (10 mM HEPES (pH 7.0), 1 mM EDTA, 60 mM KCl, 0.5% Nonidet P-40, 1 mM dithiothreitol, and 1% proteinase inhibitor (Sigma)), incubated on ice for 5 min, and centrifuged for 5 min at 4°C. The supernatants were discarded, and the pelleted nuclei were resuspended in 650 l of Buffer A (150 mM NaCl, 40 mM Tris (pH 8), 10% glycerol, 0.3% Nonidet P-40, and 1% proteinase inhibitor) and incubated on ice for 20 min. After centrifugation for 10 min at 4°C, the supernatants were transferred to a new tube and incubated for 1 h with protein A/G Plus-agarose beads (Santa Cruz Biotechnology) to clear nonspecific proteins. Beads were pelleted, and supernatants were transferred to a fresh tube. Anti-FLAG antibody M2 affinity gel (Sigma), which is a purified murine anti-FLAG IgG 1 monoclonal antibody attached to agarose, was gently mixed with nuclear extracts for 5 h at 4°C. The agarose was then washed three times with Buffer A, followed by three washes with 1ϫ EMSA buffer (20 mM HEPES (pH 7.5), 60 mM KCl, 1 mM MgCl 2 , 0.5 mM dithiothreitol, 0.1 mM EDTA, and 12.5% glycerol). G5-FLAG⅐HNF-1␣ complexes were than released by incubating the agarose in 40 l of 0.8% deoxycholic acid (Sigma) on ice for 15 min. After adding 10% Nonidet P-40 to ϳ1% (final concentration), the agarose was pelleted, and 12 l were analyzed by EMSAs.
EMSAs-EMSAs were carried out as previously described (21) using previously characterized GATA-and HNF-1-binding sites as probes and competitors. These sites included the GATA site from the Xenopus intestinal fatty acid-binding protein gene (25) and the HNF-1 site in the rat ␤-fibrinogen gene (53). Probes were made by annealing singlestranded oligonucleotides and extending with [ 32 P]dATP (PerkinElmer Life Sciences) using the large fragment of DNA polymerase I (Klenow, Invitrogen) as previously described (21). The specific activity of all probes exceeded 10 6 cpm/pmol. Proteins were incubated with 10,000 cpm of probe for 20 min at room temperature and separated on 8% nondenaturing polyacrylamide gels. For competition or supershift EMSAs, competitors (200-fold molar excess) or antibodies (1 l, undiluted), respectively, were preincubated with the nuclear extract for 10 min prior to addition of the probe. All antibodies were purchased from Santa Cruz Biotechnology.
Site-directed Mutagenesis-Site-directed mutagenesis was carried out using the method of Kunkel (52). To delete or disrupt specific structures within GATA-5 and HNF-1␣, mutations were introduced into G5-pGEM, which contains the GATA-5 coding region oriented for efficient transcription from the SP6 RNA polymerase site in pGEM-7Zf(Ϫ) (21), and H1␣-CMV. The G5-pGEM vector was used because mutagenesis with G5-CMV was not achieved. DNA-containing GATA-5 mutants synthesized from the G5-pGEM uracil template were rein-serted into the pcDNA3 expression vector for transient transfection studies using a BamHI/EcoRI subcloning strategy. Fig. 1 shows the mutagenic oligonucleotides used in this study. The oligonucleotides were phosphorylated using T4 polynucleotide kinase and ATP and annealed to the uracil templates by incubation for 2 min at 70°C, followed by slow cooling to 40°C. Uracil templates for both G5-pGEM and H1␣-CMV required the use of reverse strand oligonucleotides. After annealing, a double-stranded hybrid was synthesized using T7 DNA polymerase and T4 DNA ligase and used to transform competent E. coli cells. After an overnight incubation at 37°C, plasmids were isolated using a miniprep kit (QIAGEN Inc.) and tested by a BamHI or EcoRI digest. The identities of the mutant clones were confirmed by DNA sequencing.
In Vitro Transcription and Translation-Unlabeled and labeled wild-type and mutant GATA-5 and HNF-1␣ proteins were synthesized using the TNT TM transcription/translation kit (Promega) according to the manufacturer's instructions. G5-pGEM and H1␣-CMV utilized SP6 and T7 RNA polymerases, respectively. Labeled proteins were synthesized using [ 35 S]methionine (Redivue, Amersham Biosciences).
GST Pull-down Assays-To characterize in vitro protein/protein interactions, GST pull-down assays were carried out. GST fusion proteins were expressed in E. coli DH5␣ and purified using GSH-Sepharose beads (Amersham Biosciences) according to standard protocols (54). Beads coated with glutathione were incubated in 160 l of binding buffer (20 mM HEPES (pH 7.5), 100 mM KCl, 5 mM EDTA, 5 mM EGTA, and 0.5% bovine serum albumin) and mixed for 2 h at 4°C. The beads were then washed five times with phosphate-buffered saline, and bound proteins were released by boiling in SDS sample buffer and resolved by SDS-PAGE. The gels (10% polyacrylamide) were run at a constant voltage (150 V) for 40 min, dried, and exposed to film.
Cell Culture and Transient Cotransfection Assays-HeLa cells were used for transient cotransfection assays because GATA-specific and GATA/HNF-1␣ cooperative activation can be distinguished in cells that do not synthesize endogenous HNF-1␣. Transient cotransfection assays were carried out by electroporation as previously described (21). The amount of human growth hormone reporter secreted into the medium over 24 h was measured using an 125 I radioimmunoassay kit (Allegro hGH, Nichols Institute). To control for transfection efficiency, all transcriptional activities were expressed relative to pXGH5, a constitutively FIG. 2. GATA-5 and HNF-1␣ physically associate in vivo. EMSAs using the rat ␤-fibrinogen HNF-1 site as a probe (53) were carried out on nuclear extracts from COS-7 cells that were cotransfected with G5-FLAG and H1␣-CMV and immunoprecipitated using beads coated with anti-FLAG antibodies. A major protein⅐DNA complex (arrowhead) was detected (lane 2) that formed a supershift complex (SC) using an anti-HNF-1␣ antibody (lane 3). Nuclear extracts from COS-7 cells cotransfected with FLAG-CMV and H1␣-CMV were used as a negative control (lane 4).
active metallothionein I promoter fused to the human growth hormone gene.
Statistics-The t test or one-way analysis of variance was employed in all statistical analyses using InStat software (GraphPAD Software, Inc.). Multiple comparisons were carried out by the Dunnett multiple comparison test.

RESULTS
GATA-5 and HNF-1␣ synergistically activate the human LPH promoter (21), suggesting that these two transcription factors physically associate. To test the hypothesis that GATA-5 and HNF-1␣ physically associate in vivo, COS-7 cells were cotransfected with G5-FLAG and H1␣-CMV, and FLAGassociated proteins were immunoprecipitated from nuclear extracts using beads coated with anti-FLAG antibodies. Detection of HNF-1␣ in the G5-FLAG immunoprecipitate indicates GATA-5/HNF-1␣ interactions. As shown by EMSAs using an HNF-1-binding site as a probe (Fig. 2), a protein⅐DNA complex was identified that supershifted with an anti-HNF-1␣ antibody, demonstrating that HNF-1␣ was present in the anti-FLAG immunoprecipitate. HNF-1␣ was not detected in nuclear extracts from COS-7 cells cotransfected with FLAG-CMV and H1␣-CMV, indicating that HNF-1␣ does not bind to FLAG alone. These data demonstrate that HNF-1␣ physically associates with GATA-5 in vivo.
To characterize critical domains in GATA-5 necessary for physical association and functional cooperativity with HNF-1␣, mutations were introduced into GATA-5 to delete or disrupt specific structures (Fig. 3A). The capacity of wild-type and mutant GATA-5 proteins to associate physically with HNF-1␣ in vitro was then tested by GST pull-down assays using GST fused to HNF-1␣. As indicated by the presence of a labeled protein on SDS-polyacrylamide gels (Fig. 3B, upper panel), GATA-5 proteins that have intact C-terminal zinc finger and basic regions (wild-type GATA-5 and GATA-5 mut1, mut3, and  (24). The C-terminal domain is deleted in mut1, whereas the C-terminal domain and basic region are deleted in mut2. Both activation domains are deleted in mut3, whereas both activation domains and the N-terminal zinc finger are deleted in mut4. GATA-5 mut5 is a cysteine-to-serine substitution in the C-terminal zinc finger at amino acid 270. B, the C-terminal zinc finger and basic regions of GATA-5 are necessary for physical association with HNF-1␣. GST pull-down assays (upper panel) were carried out using GST-H1␣ incubated with labeled, in vitro transcribed and translated wild-type (WT) (lane 1) and mutant (lanes 2-6) GATA-5. Wild-type GATA-5 and GATA-5 mut1, mut3, and mut4 were pulled down by GST-H1␣, whereas GATA-5 mut2 and mut5 were not. GATA-5 was not pulled down by GST alone (lane 7). All proteins used in the GST pull-down assays were synthesized as shown by direct loading of TNT products (lower panel). C, the C-terminal zinc finger and basic regions are critical for binding to DNA. EMSAs were carried out using the Xenopus intestinal fatty acid-binding protein GATA site as a probe (25) and wild-type and mutant GATA-5 proteins as indicated (lanes 2-9). Wild-type GATA-5 formed a complex (lane 2) that was competed with a specific oligonucleotide (S; lane 3), but not with a nonspecific oligonucleotide (N; lane 4). Wild-type GATA-5 and GATA-5 mut1, mut3, and mut4 all bound to DNA, whereas GATA-5 mut2 and mut5 did not. mut4) were pulled down by GST-H1␣. In contrast, GATA-5 mut2, in which the basic region and C-terminal domain are deleted, and GATA-5 mut5, in which the structure of the Cterminal zinc finger is disrupted, were not pulled down by GST-H1␣. GST alone did not pull down wild-type GATA-5, indicating that GST does not interact with GATA-5. Direct loading of labeled TNT products (Fig. 3B, lower panel) indicates that proteins of predicted sizes were synthesized. These data demonstrate that the C-terminal zinc finger and basic regions of GATA-5 are required for physical association with HNF-1␣.
To map DNA-binding domains in GATA-5, EMSAs were carried out using a GATA-binding site as a probe and in vitro synthesized wild-type and mutant GATA-5 proteins (Fig. 3C). A protein⅐DNA complex was formed with wild-type GATA-5 that was competed away by a specific oligonucleotide, but not by a nonspecific oligonucleotide, indicating that GATA-5 binds the DNA specifically. GATA-5 mut1, mut3, and mut4 bound DNA, but GATA-5 mut2 and mut5 did not. These data are parallel to the GST pull-down experiments demonstrating that the domains critical for GATA-5/DNA interaction also map to the C-terminal zinc finger and basic regions. Thus, regions in GATA-5 that mediate protein/protein interaction with HNF-1␣ and DNA binding are co-localized to the C-terminal zinc finger and basic regions.
The functional importance of specific structures in GATA-5 for synergistic GATA-5/HNF-1␣ activation of the human LPH promoter was tested by transient cotransfection assays using wild-type or mutant GATA-5 and wild-type HNF-1␣ expression vectors (Fig. 4). To test specific effects of transfected wild-type and mutant GATA-5 and HNF-1␣, a model system utilizing cells that do not synthesize appreciable amounts of these factors was necessary. HeLa cells provided such a model system (15,16,25,55) and were used throughout the remainder of this study. pRC-CMV was used as a negative control. The human LPH promoter (h118wt) was independently activated by HNF-1␣, as previously described in Caco-2 cells (21), whereas GATA-5 alone only minimally activated this promoter. The human LPH promoter was synergistically activated by a combination of GATA-5 plus HNF-1␣, as indicated by a transcriptional activity that was ϳ3-fold greater than the sum of the individual transcriptional activities of GATA-5 and HNF-1␣ alone (indicated by the dashed line). GATA-5 mut1, mut3, and mut4 all demonstrated synergistic activation of the human LPH promoter when cotransfected with HNF-1␣ as indicated by a mean activation that extends to the right of the dashed line. Noteworthy, however, is that synergy, although present, was greatly reduced with mut3 and mut4, which do not contain GATA-5 activation domains. GATA-5 mut2 and mut5, which did not physically associate with HNF-1␣ or bind to DNA, failed to show synergistic activation of the human LPH promoter. These data suggest that the C-terminal zinc finger and basic regions of GATA-5 are required for synergistic activation of the human LPH promoter.
To characterize functional domains in HNF-1␣, mutations that result in C-terminal deletions were introduced into wildtype HNF-1␣ (Fig. 5A). The capacity of these proteins to physically associate with GATA-5 in vitro was assessed by GST pull-down assays using GST linked to GATA-5 (Fig. 5B, upper  panel). As indicated by the presence of labeled protein on SDSpolyacrylamide gels (Fig. 5B, upper panel), wild-type HNF-1␣ and HNF-1␣ mut1, in which the activation domains are de- leted, physically associated with GST-G5. However, HNF-1␣ mut2, in which the activation domains plus a segment of the C-terminal side of the homeodomain are deleted, did not physically associate with GST-G5. All proteins were synthesized as shown by direct loading of the TNT reactions (Fig. 5B, lower  panel).
The role of HNF-1␣ activation domains in the cooperative activation of the human LPH promoter was characterized by transient cotransfection assays using h118wt and wild-type and mutant HNF-1␣ (mut1). For comparison, wild-type GATA-5 and GATA-5 mut3, which contains a deletion of the GATA-5 activation domains, were also used. This allowed comparative analyses of the activation domains of GATA-5 and HNF-1␣ together (Fig. 6). Consistent with previous data (Fig.  4), deletion of GATA-5 activation domains reduced, but did not eliminate, functional synergy. In contrast, deletion of the HNF-1␣ activation domains eliminated independent as well as cooperative activation of the human LPH promoter. Lack of activation by HNF-1␣ mut1 was not due to the inability to associate with GATA-5 or to inefficient DNA binding because these functions remained intact for this mutant (Fig. 5).
Because domains in GATA-5 and HNF-1␣ that confer physical association are co-localized with DNA-binding domains, it is not possible by mutagenesis of GATA-5 or HNF-1␣ to differ-entiate between the importance of GATA/HNF-1␣ interaction and DNA binding. However, it is possible to independently determine the importance of protein⅐DNA interaction by introducing mutations that disrupt DNA interaction into the binding sites in the human LPH promoter. As shown in Fig. 7, transient cotransfection assays were carried out using h118wt and mutant promoter-reporter constructs that have the GATAand HNF-1-binding sites either intact or mutated. As shown by the middle bar, the wild-type human LPH promoter cotransfected with GATA-5 and HNF-1␣ demonstrated synergistic activation. GATA-5 and HNF-1␣ cotransfected with h118mG1G2 demonstrated synergistic activation, although the transcriptional activity was reduced from that of the wild-type promoter. However, constructs containing mutations in the HNF-1-binding site alone (h118mH) or together with the mutated GATA-binding sites (h118mG1G2H) did not show synergistic activation, and their transcriptional activities were significantly lower than that of the wild-type promoter (p Ͻ 0.05). These data demonstrate that HNF-1-binding sites in the promoter are necessary for GATA-5/HNF-1␣ cooperative activation.
Sequence alignment of GATA-4 and GATA-5 reveals 85% homology in the C-terminal zinc finger and basic regions (24), which are the domains responsible for interaction with HNF-1␣. To test the hypothesis that GATA-4 and HNF-1␣ are also capable of physically associating and cooperatively activating the human LPH promoter, GST pull-down and transient cotransfection assays were carried out. As indicated by the presence of a labeled protein on SDS-polyacrylamide gels (Fig. 8A,  upper panel), GATA-4 was pulled down with GST-H1␣, similar to GATA-5. Immunoprecipitations were carried out using antibodies for GATA-4 and GATA-5, demonstrating that the TNT reactions resulted in the synthesis of authentic GATA-4 and GATA-5 proteins (Fig. 8A, lower panel). These data suggest that GATA-4, like GATA-5, physically associates with HNF-1␣. Transient cotransfection assays carried out in HeLa cells (Fig.  8B) demonstrated that GATA-4 and HNF-1␣ independently and synergistically activated the human LPH promoter. These data suggest that GATA-4 is capable of interacting with HNF-1␣ by a mechanism similar to that described for GATA-5.

DISCUSSION
The GATA zinc finger and HNF-1 homeodomain families have been implicated as regulators of tissue-specific gene expression (34,37). The mRNAs of members of both of these transcription factor families are detected in the foregut as early as E9.5 for GATA factors and E10.5 for HNF-1␣. Both are expressed in the intestinal epithelium throughout adulthood, suggesting a critical role for these two families of transcription factors in intestinal function (29,34,56). This report shows for the first time that physical association between members of each of these transcription factor families, namely GATA-4 or GATA-5 and HNF-1␣, results in the cooperative activation of the promoter of an intestine-specific gene, LPH, suggesting functional convergence of two critical intestinal transcriptional regulatory pathways to maintain high levels of intestine-specific gene expression.
We have previously shown that GATA-5 and HNF-1␣ synergistically activate the human LPH promoter in the Caco-2 intestinal cell line (21). In the present report, we have demonstrated that this synergistic activation requires physical association between GATA-5 and HNF-1␣ and that this interaction is mediated by the C-terminal zinc finger and basic regions of GATA-5 and the homeodomain of HNF-1␣. Our data further demonstrate that HNF-1␣ activation domains are necessary not only for independent HNF-1␣ activation, but also for synergistic activation with GATA-5. Deletion of GATA-5 activation domains results, however, in a reduction, but not an elimination, of synergy, suggesting that the GATA-5 activation do-mains are not necessary, but serve as additional activators for maximal synergy. Mutational analysis of the human LPH promoter further revealed that intact HNF-1-binding sites, but not GATA-binding sites, are necessary for synergistic activation (Fig. 7), suggesting that binding of HNF-1␣ to the LPH promoter is a critical component of synergy. Although GATA-5binding sites may not be directly necessary for cooperative activation, the presence of GATA-binding sites adjacent to the HNF-1 site might recruit GATA factors in close proximity for HNF-1␣ interaction, resulting in maximal levels of synergistic activation.
The requirement for physical association between GATA-5 and HNF-1␣ was demonstrated by mutational analysis of GATA-5 (Figs. 3 and 4) and of the human LPH promoter (Fig.  7). As shown in Figs. 3 and 4, GATA-5 mut5 (C270S) failed to associate with HNF-1␣, bind DNA, and synergistically activate h118wt. However, promoter mutation experiments (Fig. 7) revealed that GATA binding to DNA was not necessary for functional synergy. Thus, failure to synergistically activate the LPH promoter by GATA-5 mut5 must be due to its inability to physically associate with HNF-1␣, rather than its inability to bind DNA. Taken together, these data demonstrate that physical interaction is required for functional synergy. Based on these data, we propose the following model: GATA/HNF-1␣ synergy is mediated by HNF-1␣ through its activation domains, which are oriented for high levels of activation through a combination of binding to DNA and physical association with GATA factors (Fig. 9). GATA/HNF-1␣ interactions might unmask the HNF-1␣ activation domains either by a conformational change or by recruitment of additional proteins that modulate transcriptional activity.
Interaction between zinc fingers and homeodomains may be  3 and 4, respectively). Loading of GATA-4 and GATA-5 immunoprecipitated with specific antibody (Ab) (lanes 1 and 2) or nonimmune serum (NI) (lanes 3 and 4) indicated that the TNT products were authentic GATA-4 and GATA-5 (lower panel). B, GATA-4 and HNF-1␣ synergistically activate the human LPH promoter. Transient cotransfection assays were carried out in HeLa cells using h118wt and GATA-4 and HNF-1␣ cotransfected separately or together. pRC-CMV was used as a negative control. Transcriptional activities above the dashed line indicate synergistic activation. Data are means Ϯ S.E. (n ϭ 5). a critical biological mechanism for gene regulation. For example, similar to the model presented here is the well characterized model for the synergistic activation of cardiac promoters by GATA-4 and Nkx2.5 (47,48,(57)(58)(59)(60), a homeodomain-containing transcription factor like HNF-1␣. In this model, the C-terminal zinc finger and basic regions of GATA-4 and the homeodomain of Nkx2.5 are required for physical association, which, in turn, is necessary for synergistic activation of several genes (47,48,57). Furthermore, intact Nkx2.5-binding sites on cardiac promoters are required for synergistic activation (47,48,57). HNF-1␣ cooperatively activates the sodium-glucose transporter (SGLT1) with SP-1, a zinc finger-containing protein (44). In Drosophila, interactions between the zinc finger of Ftz-F1, a member of the nuclear receptor superfamily, and the homeodomain of the fushi tarazu Ftz protein result in the cooperative activation of the engrailed gene (49,50). These findings are parallel to the GATA-5/HNF-1␣ model proposed here and suggest an evolutionarily conserved structure/function relationship that preserves a mechanism of cooperative activation of multiple genes in diverse tissues.
HNF-1␣ has been shown to interact through its homeodomain with other transcription factors in the intestine. A report by Sakaguchi et al. (45) suggested that HNF-1␣ is able to enhance claudin-2 promoter activity only in the presence of another member of the homeodomain transcription factor family, Cdx-2. In addition, Mitchelmore et al. (20) demonstrated that HNF-1␣ physically interacts with Cdx-2 to cooperatively activate the pig LPH promoter. Interestingly, similar to the GATA-5/HNF-1␣ model presented in this study, HNF-1␣, but not Cdx-2, must bind to the DNA for cooperative activation. Because HNF-1␣ binds to the DNA as a dimer (35,36), it remains possible that for the specific regulation of the LPH gene in vivo, all three factors form a trimolecular complex whereby the HNF-1␣ dimer binds GATA factors, Cdx-2, or a combination of both. This hypothesis is consistent with our previous studies (21), which show that the human LPH and sucrase-isomaltase promoters demonstrate synergistic activation when cotransfected with GATA-5, HNF-1␣, and Cdx-2 together. A similar mechanism has also been reported for the cardiac ␣-actin promoter, which is up-regulated through combinatorial interactions of at least three cardiac tissue-enriched transcription factors, GATA-4, Nkx2.5 (which also forms dimers, like HNF-1␣), and serum response factor (48). HNF-1␣ may also recruit coactivators such as CBP and p300/CBPassociated factor, forming a trimolecular complex that activates transcription by coupling nucleosome modification with recruitment of proteins for the general transcription machinery (61). Together, these data suggest that HNF-1␣ has the ability to physically associate with diverse proteins and to act as a linker to the DNA to synergistically activate intestinal gene promoters. The magnitude of cooperative activation may be altered by the abundance of different transcription factors and their affinity for HNF-1␣.
Physical association between GATA family members and HNF-1␣ is necessary for the synergistic activation of the human LPH promoter, providing a mechanism by which tissuerestricted transcription factors can interact to attain high levels of tissue-specific gene expression. However, although GATA (22,24) and HNF-1␣ (29) transcription factors are co-localized in the intestinal epithelial, where LPH expression is high (5, 6), they are also coexpressed in primitive intestinal epithelial cells prior to the onset of LPH gene expression (23,24,32,37). We therefore hypothesize that coexpression of GATA factors and HNF-1␣ is necessary, but not sufficient, for high levels of cellspecific LPH gene expression. Interaction with other tissuerestricted transcription factors and cofactors such as SP-1 (44), FOG-2 (62, 63), Cdx-2 (20), and the dimerization cofactor of HNF-1␣ (64) as well as other information farther upstream in the LPH promoter (18, 65) must be considered when characterizing specific expression of the LPH gene in vivo. FIG. 9. Schematic representation of a model for GATA/HNF-1␣ cooperative activation of the human LPH promoter. AD, activation domain; HD, homeodomains; ZnII, C-terminal zinc finger of GATA factors; G, GATA-binding site at the LPH promoter; H, HNF-1-binding site; TATA, TATA box-binding site for TATA box-binding protein.