Transcriptional regulation by GATA-4, GATA-5, and GATA-6 in intestine and liver was explored using a transgene constructed from the proximal promoter of the rat liver fatty acid binding protein gene (Fabpl). An immunohistochemical survey detected GATA-4 and GATA-6 in enterocytes, GATA-6 in hepatocytes, and GATA-5 in neither cell type in adult animals. In cell transfection assays, GATA-4 or GATA-5 but not GATA-6 activated the Fabpl transgene solely through the most proximal of three GATA binding sites in the Fabpl promoter. However, all three factors activated transgenes constructed from each Fabpl site upstream of a minimal viral promoter. GATA factors interact with hepatic nuclear factor (HNF)-1α, and the proximal Fabpl GATA site adjoins an HNF-1 site. GATA-4, GATA-5, or GATA-6 bounded to HNF-1α in solution, and all cooperated with HNF-1α to activate the Fabpl transgene. Mutagenizing all Fabpl GATA sites abrogated transgene activation by GATA factors, but GATA-4 activated the mutagenized transgene in the presence of HNF-1α. These in vitro results suggested GATA/HNF-1α interactions function in Fabpl regulation, and in vivo relevance was determined with subsequent experiments. In mice, the Fabpl transgene was active in enterocytes and hepatocytes, a transgene with mutagenized HNF-1 site was silent, and a transgene with mutagenized GATA sites had identical expression as the native transgene. Mice mosaic for biallelic Gata4 inactivation lost intestinal but not hepatic Fabpl expression in Gata4-deficient cells but not wild-type cells. These results demonstrate GATA-4 is critical for intestinal gene expression in vivo and suggest a specific GATA-4/HNF-1α physical and functional interaction in Fabpl activation.
- intestinal epithelium
the six members of the GATA transcription factor family share conserved zinc finger DNA-binding domains that recognize the DNA sequence WGATAR (84). GATA factors bind to DNA as monomers, and only one GATA factor occupies the cognate DNA binding site at a time (33). Transcripts for GATA-4, GATA-5, and GATA-6 have been localized to the developing mammalian liver and gastrointestinal tract. Transfection assays in the intestinal cultured cell line Caco-2 have demonstrated that these factors activate numerous intestinal genes: GATA-6 binds to a site in the promoter of the lactase phlorizin hydrolase gene (30), and GATA-4, GATA-5, and GATA-6 transactivate a transgene made from the proximal lactase promoter (29, 40). A transgene containing the proximal sucrase isomaltase gene promoter is activated by GATA-5 but not GATA-4 or GATA-6 (11, 40). The sodium-hydrogen exchanger isoform 3 gene is highly active in differentiated enterocytes, and a transgene constructed from the proximal promoter of this gene is activated by GATA-5 and, to a lesser extent, GATA-4 and GATA-6 (38). GATA-4 activates a transgene made from the rat liver fatty acid binding protein gene through interaction with a site in the proximal promoter (25). These in vitro studies suggest a potentially important role for GATA factors in intestinal gene regulation.
In vivo evidence has also been obtained for a role of GATA factors in intestinal gene regulation using a transgenic approach. A duodenum-specific enhancer in the adenosine deaminase gene contains consensus GATA binding sites that bind GATA factors in duodenal nuclear extracts (28). A transgene constructed from the adenosine deaminase regulatory sequences including the enhancer had high expression levels in murine duodenum, but a homologous transgene with selective mutation of the three enhancer GATA binding sites had markedly reduced duodenal expression and alterations in the proximal to distal and cell-specific expression patterns (28). A transgene constructed from the proximal promoter of the sucrase-isomaltase gene was active in murine enterocytes (79). A GATA-binding site was identified in the proximal promoter that overlapped with a binding site for the Cux/CDP transcriptional repressor (10). GATA factors did not activate the transgene through this site in cultured cells, but Cux/CDP repressed activity. Mutagenesis of this site resulted in transgene activation in the colon of mice but no loss of expression in the small intestinal epithelium, suggesting that the repressor had dominant function (10). Mice with targeted null mutations for Gata4, Gata5, and Gata6 have been created, but these animals have not been informative for the role of these factors in regulating gene expression in the intestine and liver. Gata4-deficient mice die around embryonic day 9 with severe ventral developmental defects (41, 49). In chimeric mice produced from a high-percentage Gata4 null embryonic stem cells and a low percentage of wild-type embryonic stem cells, Gata4+/+ cells were observed only in visceral endoderm and gut endoderm at embryonic day 9 in embryos that were otherwise entirely derived from Gata4−/− cells (56). GATA-4 is required for gastric epithelial differentiation during development (36), and gastric genes are activated by GATA factors in cultured cells (2, 62, 72). Gata5-deficient mice develop normally except for a defect in female genitourinary development and are healthy and fertile as adults (50), although an analysis of hepatic or gastrointestinal gene expression in these animals has not been reported. Loss of Gata6 results in lethality shortly after implantation (39).
Knowledge of the precise cellular expression patterns of GATA-4, GATA-5, and GATA-6 in mammalian liver and intestine would aid determination of their in vivo function. GATA-4 expression was defined in the murine intestinal epithelium, and expression was detected only in differentiated intestinal enterocytes (11). In the Xenopus laevis intestinal epithelium, GATA-4, GATA-5, and GATA-6 are expressed in distinct but overlapping patterns (31). Electromobility shift assays conducted with intestinal extracts revealed that GATA-4 was the only major binding activity with putative GATA binding sites, despite the presence of Gata5 and Gata6 transcripts (10, 28). These differences have been attributed to posttranscriptional regulation of the GATA factors (14) or posttranslational modification of binding activity (28). Multiple GATA factors are present in cells of other tissues and were found to interact with target genes in distinct ways: GATA-6 but not GATA-4 activates Dab2 in visceral endoderm (55), and GATA-4 is used preferentially over GATA-6 in the regulation of myosin α- and β-chains (17). This specificity has been attributed to preferential binding of individual GATA factors to their cognate site in the promoters of the myosin genes (17), interleukin-5 gene promoter (85), and sucrase-isomaltase gene promoter (10), perhaps through interaction with sequences flanking the core GATA binding sequence. Interactions with cofactors or other transcription factors that bind near a particular site may mediate preferential binding or activity of individual GATA factors (17, 26, 43, 55, 57, 70, 75, 80). In particular, GATA factors have been reported to interact physically and functionally with hepatic nuclear factor (HNF)-1α in activating intestinal gene expression in cultured cells (11, 40, 80).
The HNF-1 homeodomain transcription factors are also critical regulators of intestinal and hepatic gene expression and cellular differentiation (78). HNF-1α and HNF-1β are the two family members that share highly homologous homeodomain DNA-binding motifs that recognize the same DNA sequence (48, 60) and bind to a single cognate DNA site as heterodimers or homodimers. Cooperative gene activation has been identified between HNF-1α and GATA-4 or GATA-5 (11, 80) in cultured cells. A physical interaction between GATA-5 and HNF-1α has been reported (80), in which the two factors bind to each other in solution in the absence of DNA. In addition, GATA-4, HNF-1α, and Cdx-2 cooperate together to activate a sucrase-isomaltase transgene in cultured cells, and all three factors form a complex in solution (11). However, in vivo evidence for the significance of interactions between GATA and HNF-1 family factors has not yet been obtained.
We have used the rat liver fatty acid binding protein gene (Fabpl) as an experimental model to study gene regulation in the intestinal epithelium and liver in vivo (67). Rat Fabpl is highly expressed in hepatocytes and enterocytes, and expression is primarily regulated at the transcriptional level (5). A transgene constructed from Fabpl nucleotides −596 to +21 relative to the start site of transcription is active in murine hepatocytes, all small intestinal epithelial cells, all proximal colonic epithelial cells, and some gastric epithelial cells (71). We identified functional binding sites for GATA factors, HNF-1 factors, and four additional endodermal transcription factor families in the proximal Fabpl promoter (25). Furthermore, cooperative activation was observed between GATA-4 and HNF-1α or HNF-1β in activating the Fabpl transgene in cultured cells (25). The importance of HNF-1α in Fabpl transcriptional regulation is demonstrated by the complete loss of hepatic Fabpl expression in mice with a targeted disruption of HNF-1α (1). The role of GATA factors in Fabpl regulation in vivo remains unknown. We now report that GATA family factors are localized in the same cells as those reported to express HNF-1α in the adult murine intestine and liver (11) and that GATA-4 is critical for intestinal Fabpl expression. We demonstrate that cooperative functional interactions occur between HNF-1α and GATA-4, GATA-5, or GATA-6 to activate an Fabpl transgene in cultured cells. We show that HNF-1α binds to GATA-6 as well as GATA-4 and GATA-5 in solution and provide evidence that HNF-1α interacts specifically with GATA-4 to regulate Fabpl expression.
MATERIALS AND METHODS
Transcription factor binding sites in the Fabpl promoter (GenBank accession no. M13501) were identified with transcription element search system (TESS) (63) and by direct examination using binding site matrices from TRANSFAC (84).
An Fabpl transgene was constructed from Fabpl nucleotides −596 to +21, relative to the start site of transcription, linked to the entire human growth hormone (hGH) gene lacking regulatory sequences as previously described (25). The glucocorticoid receptor binding site in the first intron of hGH was inactivated by mutagenesis in this construct. Site-directed mutagenesis of the Fabpl transgene was performed with a commercial kit (Stratagene, Quikchange). The sequence of sense strand oligonucleotides used to induce mutations shown in Fig. 2 and not previously described (25) was as follows: GATA-128: 5′-GCCCATTCTGATTTTTAGTGTTGACCATTGCTCTCAGGA-3′; GATA-130: 5′-CTTCTGCCTTGCCCATTCTACTTTTTATCGTTGACCATTGC-3′. The underlined bases are those changed. Targeted mutations were confirmed by sequencing the entire Fabpl sequence, and the presence of a functional hGH reporter was verified by protein production in cultured cells (see below). Promoters with multiple mutations were created by sequential rounds of mutagenesis. Plasmids with Fabpl transgenes containing mutagenized sequences were termed pTS173 (GATA-128 site mutated), pTS181 (GATA-130 site mutated), pTS184 (GATA-229 and -557 sites mutated), pTS266 (GATA-128, -130, and -557 sites mutated).
Transcription factor expression plasmids for HNF-1α, HNF-1β, and GATA-4 were produced by inserting the transcription factor coding sequences into pSG5 (Stratagene) (25). An expression plasmid for GATA-5 (pTS218) was created by amplifying the GATA-5 open reading frame from a murine heart cDNA library using PCR and vent polymerase (New England Biolabs). Outside primers were 5′-AGTGATCCTGCCCTAGACGTCTGC-3′ and 5′-TCCTGTTCCTGGGATGTACTGTGG-3′, and inner primers were 5′-GCTATGAATTCTTCTCTGCAGGTCAAGCTCG-3′ and 5′-GCTATTGATCACCTAGGCCAAGCCAGAGCA-3′. The 5′- and 3′-primers contained EcoRI and BclI sites, respectively. The GATA-5 coding sequence was digested with these enzymes and ligated into pSG5 cut with EcoRI and BglII. Sequencing was used to verfy the absence of PCR-induced errors. An expression plasmid for GATA-6 (pTS236) was constructed by digesting pcDNA3-GATA-6 (53) (a kind gift from Edward Morrisey) with XhoI and blunting with T4 DNA polymerase. The GATA-6 coding sequence was then released with a subsequent digest with BamHI and ligated into pSG5 digested with BglII and BamHI, where the BglII site had been blunted.
Synthetic “short” promoter constructs were produced to place GATA binding sites 5′ to an SV40 minimal promoter. The promoter was amplified from pGL3-promoter (Promega) with a 5′-primer that contained the GATA site of interest or control sequence. Primers used for sites shown in Fig. 2 were as follows: GATA-128, 5′-cgatctcgagcattctactttttaTC-gttgactgcatctcaattagtcagcaaccatagtc-3′;GATA-130, site 5′-cgatctcgagcattctgatttttagtgttgactgcatctcaattagtcagcaaccatagtc-3′; GATA-228, 5′-cgatCTCGAGatgagcggtgataagacaccaatgcatctcaattagtcagcaaccatagtc-3′; GATA-556, 5′-cgatCTCGAGttagggactgataaaatatatgtgcatctcaatt-agtcagcaaccatagtc-3′; J6 GATA site (7), 5′-cgatCTCGAGaggcaatggagatagggagggatgcatctcaattag-tcagcaaccatagtc-3′; and control (no GATA site), 5′-CGATCTCGAGTGCATCTCAATTAGTCAGCAACCATAGTC-3′. The same reverse primer 5′-GAACggatccAAGCTTTTTGCAAAAGCCTAGGCCTC-3′ was used for all reactions and contains a BamHI recognition sequence. The amplimers containing the SV40 promoter and GATA site or control sequence were digested with XhoI and BamHI, then ligated into pTS154 (25) digested with the same enzymes. These constructs contained the viral promoter with or without added GATA sites driving expression of hGH as a reporter. Sequencing was used to verify the absence of PCR-induced errors.
A plasmid containing the rat LFABP protein coding sequence in pZERO 2.0 for generation of in situ probes was constructed (pTS252). PCR primers 5′-GACGggatccctcattgccaccatgaactt-3′ and 5′-gacgggatccctaaattctcttgctgactc-3′ contain BamHI sites and were used to amplify the coding sequence from pMONLFABP (20). pZERO 2.0 (InVitrogen) and the amplimer were digested with BamHI and ligated together. Orientation and absence of sequence errors were confirmed by sequencing.
A bacterial expression plasmid for production of the glutathione-S-transferase (GST)/HNF-1α fusion protein was produced by introducing the coding sequence for HNF-1α into pGEX-2T (Amersham). The murine HNF-1α sequence was amplified from pTS158 (25) with primers incorporating a BamHI site at the 5′ end (5′-CAGCggatccatggtttctaagctgaGCCAGC-3′) and an EcoRI site at the 3′ end (5′-CACGGAATTCTTACTGGGAAGAGGAGGCCATC-3′). The vector and PCR fragment were digested with both enzymes and ligated together to create plasmid pTS358. The absence of errors in the HNF-1α sequence was verified by sequencing.
Cell culture and transfections.
CaCo-2, HepG2, and LLC-PK1 cells were purchased from American Type Culture Collection and maintained as recommended. Transient transfections were performed by the calcium phosphate precipitation method (25). Briefly, cells were transfected at a density of 40–60% in six-well plates and were just confluent when harvested. The transfection mixtures contained an Fabpl reporter plasmid, transcription factor expression plasmids, and pGL3-control plasmid (Promega), included to normalize for differences in expression efficiency. The total amount of plasmid DNA (5–9 μg/well) was kept constant in a given experiment by addition of pSG5 plasmid. Assays were performed 48 h after transfection. HGH was detected in the media using a specific radioimmunoassay (Nichols Institute), whereas luciferase production from pGL3-control was detected with a commercial kit (Promega). Values were calculated as the average of the three replicate wells for each DNA solution, and error was in standard deviations or propagated error for calculated values. Values are reported as fold activation over the activity of the native Fabpl reporter with no added transcription factor expression plasmids. Significance is calculated by Student's two-tailed t-test. All experiments were repeated at least twice with similar results.
Genetically altered mice.
Approval was obtained from the Washington University Institutional Animal Care and Use Committee for all experiments involving animals. Transgenes were injected into a pronucleus in fertilized eggs derived from inbred FVB/N mice by standard procedures (58). All founder animals were killed, and tissue was fixed for immunohistochemical analysis as previously described (66). Chimeric mice were produced as previously described (36).
Immunohistochemistry, histological staining, and in situ analysis.
Immunohistochemical localization of GATA factors was performed with sections from paraffin-embedded tissues of Naval Medical Research Institute mice and mice of C57Bl/6J and 129/Ola mixed background that had been fixed in 4% paraformaldehyde. Primary antibodies were polyclonal goat anti-mouse GATA-4 IgG, polyclonal goat anti-mouse GATA-5 IgG, and polyclonal rabbit anti-mouse GATA-6 IgG (all from Santa Cruz Biotechnology) used in dilutions of 1:200, 1:50, and 1:50 for GATA-4, GATA-5, and GATA-6, respectively. Nonimmune IgG was used as the primary antibody for negative controls. A commercially available avidin-biotin immunoperoxidase system was used to visualize bound antibody (Vectastain Elite ABC Kit, Vector Laboratories) with 3,3′-diaminobenzidine (Sigma) as substrate. Staining for each gastrointestinal tract region and liver sample was repeated two or more times. Immunohistochemistry for hGH was performed as previously described (66). β-galactosidase color development and in situ hybridization were performed as previously reported (36). The antisense probe for Fabpl was derived from the entire coding sequence of rat Fabpl (pTS252), which is highly homologous to the mouse sequence.
Solution binding assays.
The GST-HNF-1α fusion protein was produced in BL21 (DE3) Escherichia coli and purified with glutathione resin according to the protocol suggested by the manufacturer (Amersham). Before use, the fusion protein was dialyzed against a buffer containing 50 mM TRIS, pH 7.5, 150 mM NaCl, 100 μM ZnCl2, 0.3% Nonidet P-40, and 1 mM dithiothreitol. 35S-labeled target proteins were synthesized from mammalian expression plasmids pTS186 (GATA-4), pTS218 (GATA-5), and pTS236 (GATA-6) using the TNT T7 coupled reticulocyte lysates kit (Promega) and conditions recommended by the manufacturer. Binding assays were conducted by mixing 4 μg of GST/HNF-1α fusion protein with 20 μl of each target synthesis reaction in a total volume of 400 ul binding buffer [in mM: 20 HEPES (pH 7.5), 100 KCl, 5 EDTA, and 5 EGTA (80)] with 0.5% BSA added. Binding was allowed to occur for 2 h at 4°C, then 20 ul glutathione Sepharose beads equilibrated in binding buffer were added, and the reaction continued overnight. The beads were then washed once in binding buffer with BSA and twice with binding buffer without BSA. Proteins were eluted from the beads by boiling in SDS sample buffer and separated by SDS-PAGE. Radiolabeled proteins were visualized with a phosphor capture system.
Cellular localization of GATA-4, GATA-5, and GATA-6 in the adult murine gastrointestinal tract and liver.
We used immunohistochemistry to determine the cellular expression pattern of GATA-4, GATA-5, and GATA-6 in the adult murine stomach, intestine, and liver (Fig. 1). Abundant gastric GATA-4 expression was detected in the neck region (Fig. 1A). Positive mucous, parietal, and chief cells were found in the cardiac and corpus glands as well as in the surface epithelial cells of the cardia. Nuclear GATA-6 expression was most abundant in the distal regions of the stomach (Fig. 1I). Approximately one-half of the surface epithelial cells were immunoreactive for GATA-6, and stronger staining was also detected in the nuclei of glandular parietal and duct cells. Cytoplasmic immunoreactive GATA-6 was also detected in most cells where nuclear GATA-6 immunostaining was evident, and this staining is likely nonspecific, because similar cytoplasmic but not nuclear staining was observed in sections of brain where GATA-6 is not expressed (Fig. 1M; Brn). Previous work (3) has shown the specificity of the GATA-6 immunostaining procedure in the female reproductive tract. Considerably fewer cells immunoreactive for GATA-5 were observed in the gastric mucosa compared with GATA-4 and GATA-6 (Fig. 1E), and these were primarily located in the transitional region between the cardia and corpus (Fig. 1E, inset) along with some surface epithelial cells.
GATA-5 was not detected in the adult murine small intestine (Fig. 1F). GATA-4 was abundantly expressed in villus enterocytes as previously reported (11) as well as Brunner's glands (Fig. 1B). Nuclear GATA-6 was detected in the majority of villus and crypt epithelial cells, Brunner's gland cells, and rare cells of the lamina propria (Fig. 1J). Neither GATA-4 nor GATA-5 were detected in the colon (Fig. 1, C and G), where GATA-6 was abundant in epithelial cell nuclei but not in the nuclei of lamina propria cells (Fig. 1K). The expression pattern of GATA-4, GATA-5, and GATA-6 immunoreactive protein was identical to that observed with in situ analysis of transcript localization for these three factors in the small and large intestines (data not shown). Within the liver, GATA-4 was not present in hepatocytes but was abundantly expressed in other liver cell types, presumably sinusoid endothelial cells (Fig. 1D). GATA-5 was not detected in the liver (Fig. 1H), but GATA-6 was abundant in hepatocytes as well as bile duct epithelial cells (Fig. 1L). Nonimmune IgG as the primary antibody gave only background staining in all tissues (data not shown).
GATA-4, GATA-5, and GATA-6 directly activate the Fabpl transgene through interaction with cognate sites in the proximal promoter.
GATA-4 activates a transgene constructed from rat Fabpl nucleotides −596 to +21 in cultured cells (25). This transgene is active in mice in all small intestinal epithelial cells, in all proximal colonic epithelial cells, in hepatocytes, in renal proximal tubular epithelial cells, and in a subpopulation of gastric epithelial cells (61, 66, 71, 74). Because GATA factors are present in many of these cell populations (Fig. 1), the potential of GATA factors to regulate the transgene was explored. The 624-bp Fabpl proximal promoter was examined for the core “GATA” binding sequence using the TESS (63) as well as direct sequence evaluation. Four potential GATA binding sites were identified (Fig. 2). These sites were designated by most the proximal base of the consensus binding site relative to the start site of transcription. Two of the GATA sites are distinct, whereas the −128/−130 GATA sites overlap substantially (Fig. 2). All the sites were conserved between the orthologous human and rat Fabpl promoter sequences except for the −130 site. The −130 site was also not an ideal match with the consensus but might function as a GATA binding site if the −128 site were absent as discussed below. In addition to these GATA sites, other functional transcription factor binding sites in Fabpl have been identified, including an HNF-1 site at −95 (1, 25).
Transient transfection assays were performed in three cell lines to determine the potential of GATA-4, GATA-5, or GATA-6 to activate the Fabpl transgene. Caco-2 cells, HepG2 cells, and LLC-PK1 cells were chosen to represent the three tissues in which the transgene is active in mice: intestinal epithelium, hepatocyte, and proximal tubular epithelium (71). Caco-2 cells endogenously express GATA-4, GATA-5, and GATA-6, with GATA-6 being the predominant form (30), HepG2 cells express GATA-4 (83), and the endogenous expression of GATA factors has not been reported in LLC-PK1 cells. A plasmid containing the Fabpl transgene produced the transgene reporter (hGH) when transfected into all three cell lines (Fig. 3). The addition of expression constructs for GATA-4 or GATA-5 activated the transgene in all three cell lines (P < 0.001 for all cases; Fig. 3), whereas the addition of a GATA-6 expression plasmid produced only a small increase in transgene activity in Caco-2 (1.4-fold; P = 0.026) and HepG2 cells (3.8-fold; P = 0.019; Fig. 3). GATA-6 activation of the Fabpl transgene was increased at higher doses of expression plasmid in these two cell lines (data not shown). A transgene was constructed with all four potential GATA sites mutagenized to abrogate GATA factor binding (Fig. 2) as a tool to determine whether GATA activation of Fabpl was directly due to binding of these factors to the promoter. The transgene with no functional GATA was not activated by GATA-4, GATA-5, or GATA-6 (P > 0.2 for all cases; Fig. 3), indicating these factors activate Fabpl through direct interaction with one or more of the mutagenized sites.
GATA factors activate the Fabpl transgene through preferential interaction with the most proximal of three GATA binding sites.
Transgenes were created where mutagenesis was used to destroy individual Fabpl GATA sites or various combinations of Fabpl GATA sites. These transgenes provided a tool to determine which sites mediated Fabpl transgene activation by GATA-4 (Fig. 4, top row) or GATA-5 (Fig. 4, bottom row). In all three cell lines, the Fabpl transgene lacking a functional −128 site had the greatest drop in GATA-stimulated activity of any promoter with a single site mutagenized. Mutagenesis of both the −128 and −130 overlapping sites resulted in a loss of activation with either GATA-4 or GATA-5 comparable with that achieved with inactivation of all four sites (Caco-2 GATA-4 1.11 vs. 0.82, P = 0.037; HepG2 GATA-4 2.27 vs. 1.62, P = 0.009; LLC-PK1 GATA-4 1.08 vs. 0.57, P = 0.003; Caco-2 GATA-5 2.94 vs. 2.05, P = 0.005; HepG2 GATA-5 3.19 vs. 2.64, P = 0.15; LLC-PK1 1.25 vs. 0.6). GATA factors could presumably only occupy one of the two overlapping −128/−130 sites at any one time, and the −130 site may only be used when the −128 site is lost. In some cells, transgenes lacking a functional −229 site also exhibited reduced activity, indicating that this site may also be used to a minor extent (LLC-PK1 GATA-4 12.6 vs. 6.68, P = 0.0002; HepG2 GATA-5 42.23 vs. 24.03, P = 0.012; LLC-PK1 GATA-5 13.30 vs. 4.66, P = 0.001).
All three Fabpl GATA sites confer equal activation by GATA-4, GATA-5, and GATA-6 to a minimal viral promoter.
Three GATA binding sites in the Fabpl promoter contain canonical GATA factor binding sequences (Fig. 2). The most proximal site might be preferentially used due to preferred binding to sequences outside the core consensus recognition sequence (85). Synthetic promoter transgenes were created to test the interactions of GATA-4, GATA-5, and GATA-6 with the Fabpl binding sites. These transgenes contained a promoter consisting of a single GATA binding site located just 5′ to a minimal SV40 promoter. These synthetic promoters were linked to the entire hGH gene as a reporter. GATA binding sites used were the Fabpl sites or the GATA site from the murine J6 gene that mediates GATA-4 activation (7). A 22-nt sequence was used for each site, centered on the GATA sequence. The activities of the proximal tandem site were differentiated by using sequences with one site or the other selectively mutagenized as indicated in Fig. 2. The activities of GATA family factors in transactivating these synthetic promoters were determined with transient transfections in HepG2 cells. Each of the Fabpl GATA binding sites conferred similar activation by GATA-4, GATA-5, or GATA-6 to the viral promoter (Fig. 5). Activation by GATA-4 was approximately twofold (P < 0.002 for all sites) and 1.5-fold for GATA-5 and GATA-6 (P < 0.009 for all cases). This activation was comparable with that of a characterized GATA-4 binding site in the J6 gene promoter (7). Furthermore, GATA-6 activated the viral promoter transgenes to the same extent as GATA-5, in contrast to activation mediated by the same sites in the Fabpl promoter, where GATA-4 and GATA-5 but not GATA-6 activated. These results suggest that there is no significant difference in GATA factor interaction with the three sites and that an alternative mechanism produces selective site and factor utilization.
Cooperative activation of the Fabpl transgene is observed between GATA family factors and HNF-1 family factors in cultured cells.
Selective site and factor specificity might be determined by interactions of the GATA factors with other transcription factors that bind to the promoter. HNF-1α functionally and physically interacts with GATA-4 (11) and GATA-5 (80), and an HNF-1 binding site is adjacent to the most proximal GATA binding site (Fig. 2 and Ref. 25). Furthermore, GATA-4 and HNF-1α or HNF-1β functions cooperatively to activate the Fabpl transgene (25). This cooperativity was manifest as activation of the Fabpl transgene by HNF-1α or HNF-1β and GATA-4 together to more than twice the sum of the activations of the two factors separately. Cooperativity between GATA-4, GATA-5, or GATA-6 and HNF-1α or HNF-1β in activating the Fabpl transgene was determined using transient transfections in the three cell lines (Fig. 6). HNF-1α is expressed endogenously in Caco-2 cells (25) and HepG2 cells (46) but not in LLC-PK1 cells (68). GATA-4 exhibited greater than twofold cooperative activation with HNF-1α in all three cell lines in activating the Fabpl transgene and ∼1.5-fold cooperative interaction with HNF-1β in activating the Fabpl transgene in all three cell lines; compare activation of both factors together (dark gray bars) with the sum of the activities of each factor separately (arrows). At least twofold cooperative activation was also observed between GATA-5 or GATA-6 and HNF-1α in Caco-2 and HepG2 cells, despite a lack of Fabpl transgene activation by GATA-6 alone (Fig. 3). These observations suggest that interaction between the multiple GATA factors found in intestine and liver with HNF-1 family factors may participate in Fabpl transgene activation.
The HNF-1 binding site but not GATA binding sites is critical for Fabpl transgene expression in vivo.
The potential significance of GATA family factors or HNF-1 family factors to Fabpl transgene expression in vivo was determined by examining the expression pattern of the mutagenized transgenes in mice. Transgenic animals were created by pronuclear injection of the native Fabpl transgene or by injection of a transgene with all GATA binding sites mutagenized or by injection of a transgene with the HNF-1 binding site mutagenized. These transgenes were identical to those used in the cell transfection studies. Six independent founder animals were generated for the native transgene: six for the transgene with mutagenized GATA sites and eight for the transgene with the mutagenized HNF-1 site. Serum hGH levels were measured in these animals, and levels above 1 ng/ml serum were detected in four of the founder animals with the native transgene, two of the founder animals with the transgene containing no GATA binding sites, and none of the founder animals without a functional HNF-1 site. Immunohistochemistry was used to define the precise cellular transgene expression pattern in all founder animals regardless of serum hGH levels. Transgene reporter immunoreactivity was only detected in those animals with detectable serum hGH. The expression pattern of the native transgene was identical to that previously described (67), with expression in hepatocytes, all cells in the small intestinal epithelium, proximal colonic epithelial cells, renal proximal tubular epithelial cells, and some cells in the gastric epithelium (Fig. 7A and data not shown). The transgene with mutagenized GATA sites had an identical expression pattern to that of the native transgene in all tissues (Fig. 7B and data not shown). No hGH was detected in mice carrying the transgene with a mutagenized HNF-1 site by radioimmunoassay or immunohistochemical analyses (Fig. 7C and data not shown). These results establish that the HNF-1 binding site is required for Fabpl transgene expression, consistent with a requirement for HNF-1α for activity of the endogenous gene (1). GATA factors either do not play an essential role in regulating Fabpl transgene expression or GATA factors regulate Fabpl expression independently of the presence of their cognate DNA binding site(s) through binding to other factors as has been suggested (11)
GATA-4 is required for murine intestinal Fabpl expression.
Mice chimeric with cells containing targeted inactivation of both Gata4 alleles and with wild-type cells were used to determine whether GATA-4 is required for endogenous Fabpl expression in vivo. These chimeric mice were produced by injection of embryonic stem cells with targeted inactivation of both Gata4 alleles into blastocysts harvested from ROSA26 mice (36). ROSA26 mice express bacterial β-galactosidase in most cells, including those of the intestinal epithelium and hepatocytes. In situ analysis for Fabpl transcript expression was performed on sections from a mouse at day 18 of gestation (Fig. 8). Some intestinal segments visible in these sections were derived entirely from wild-type cells and displayed β-galactosidase staining (arrow in Fig. 8A), uniform Fabpl expression in villus epithelial cells (arrow in Fig. 8B), and uniform Gata4 and Gata6 expression (arrows in Fig. 8, D and F, respectively). Some segments were derived from both wild-type and Gata4-deficient cells (arrowhead in Fig. 8A). Those segments lack both Fabpl and Gata4 expression in identical regions (arrowheads in Fig. 8, B and D, respectively). Cellular morphology was identical between the Gata4-deficient and wild-type intestinal epithelial cells (Fig. 8A), and Gata6 expression was maintained in those cells lacking Gata4 activity (arrow in Fig. 8F), indicating that there is no global defect in tissue differentiation in Gata4-deficient cells. These results suggest that GATA-4 is required to sustain Fabpl transcription in the intestinal epithelium, in contrast with the absence of a phenotype when Fabpl transgene GATA sites are lost. Fabpl activity was also determined in the livers of chimeric animals, where equal expression was observed in Gata4-deficient and wild-type hepatocytes, suggesting different regulatory mechanisms between hepatocytes and enterocytes.
GATA factors bind directly to HNF-1α and activate the Fabpl transgene in the absence of GATA binding sites if HNF-1α is present.
The possibility of GATA factors interacting with HNF-1 family factors in the absence of a GATA binding site was explored, because loss of GATA binding sites did not affect transgene expression in vivo, but loss of Gata4 resulted in endogenous Fabpl silencing. Previously published studies (11, 80) have established that HNF-1α binds directly to GATA-4 or GATA-5 in solution, and HNF-1α functionally cooperated with GATA-4, GATA-5, or GATA-6 to activate the Fabpl transgene (Fig. 6). Solution binding assays were conducted to determine whether GATA-6 could also bind to HNF-1α. A GST-HNF-1α fusion protein was used to pull down radiolabeled GATA-4, GATA-5, or GATA-6. This assay revealed that HNF-1α could bind to all three GATA factors in solution (Fig. 9).
We sought evidence that this interaction between HNF-1 and GATA factors could occur in the cell nucleus in the context of the Fabpl promoter. Activation of the Fabpl transgene by GATA-4 or GATA-5 or GATA-6 plus either HNF-1α or HNF-1β was performed in Caco-2 cells as shown in Fig. 6. Activation by these combinations of factors was compared for a transgene constructed from the native promoter, a transgene constructed from a promoter lacking all GATA sites, a transgene constructed from a promoter lacking the HNF-1 site, or a transgene constructed from a promoter lacking both GATA and HNF-1 sites. The results were examined for evidence of Fabpl transgene activation by a GATA factor in the presence of an HNF-1 factor but in the absence of a GATA binding site. This evidence was found for HNF-1α and GATA-4 but not any other combination (Fig. 10 and data not shown). As previously observed (Fig. 6), both GATA-4 and HNF-1α were able to activate a transgene constructed from the native Fabpl promoter, and the two factors together exhibited significant cooperativity in transgene activation (Fig. 10). HNF-1α did not activate a transgene constructed from the Fabpl promoter with a mutagenized HNF-1 binding site. In addition, GATA-4 activation of this transgene lacking a functional HNF-1 binding site was also decreased by approximately one-half. This decrease could be due to loss of cooperative interaction with endogenous HNF-1α, which is expressed at low levels in Caco-2 cells under the conditions used (25). GATA-4 was unable to activate the transgene constructed from an Fabpl promoter with mutagenized GATA binding sites, whereas HNF-1α activation of this transgene was unaffected. However, HNF-1α activation of the transgene constructed from an Fabpl promoter with no GATA binding sites was significantly increased by addition of GATA-4, although GATA-4 alone could not activate this transgene (Fig. 10; bracket). Neither factor alone nor both factors together activated a transgene constructed from an Fabpl promoter lacking both GATA and HNF-1 binding sites. These findings suggest an asymmetrical interaction among GATA-4, HNF-1α, and the Fabpl promoter, where GATA-4 is tethered to HNF-1α in the absence of DNA binding by GATA-4 but not the reverse (Fig. 11). No such interaction was found between HNF-1α and GATA-5, which are also known to physically and functionally interact (Figs. 6 and 9 and Refs. 40 and 80), or GATA-6 and HNF-1α, which we have now shown to functionally and physically interact (Figs. 6 and 9).
GATA expression patterns.
GATA-4, GATA-5, and GATA-6 are localized in the developing or mature intestine, stomach, and liver (4, 24, 42, 53, 69), are critical for development of these tissues (36), and have been implicated in gene regulation in the adult liver (9, 22, 34), intestine (2, 29–31, 34, 38, 80), and stomach (62). We conducted a detailed analysis of the GATA protein cellular expression patterns in these tissues of the adult mouse, because differences between mRNA and protein expression patterns have been identified (14). Notably, hepatic GATA-4 was only present in cells other than hepatocytes, although GATA-4 has been implicated in hepatocyte gene expression based on nuclear extracts prepared from the entire liver (22, 34). Nuclear GATA-6 was detected in hepatocytes, consistent with a report that this was the predominant GATA family member in the liver based on mRNA abundance (69). GATA-4 and GATA-6 have distinct but overlapping expression patterns in the intestinal epithelium. Nuclear GATA-6 is expressed in both crypt and villus epithelial cells, whereas GATA-4 is limited to the villus epithelium. GATA-6 was also detected in the nuclei of lamina propria cells of the murine small intestine, and expression in smooth muscle cells has previously been reported (52). In addition, nuclear GATA-6 expression extends into the proximal colon. In situ analysis of expression of GATA-4, GATA-5, and GATA-6 in X. laevis revealed a similar expression pattern for xGATA-4 and xGATA-6 to that of the mouse, but expression of the xGATA-5 transcript was abundant in differentiated villus cells of the adult small intestinal epithelium. We did not detect GATA-5 in adult murine tissues, although the transcript has been detected in murine fetal intestine and liver (54) and adult distal small intestine (28). This discrepancy may reflect a difference in transcript vs. protein expression as characterized for GATA-6 (14) or a difference in assay sensitivity between Northern blot analysis and immunohistochemistry. We also detected cytoplasmic GATA-6 expression in most cells where nuclear GATA-6 was evident but not in cells without nuclear staining. GATA-6 has been identified in the cytoplasmic compartment as well as the nucleus of ovarian tumor cells (15), and it is tempting to speculate that the cytoplasmic staining may be specific in those cells. However, no definitive conclusion can be made, because cytoplasmic staining was also detected in the brain cells, which are known not to express GATA-6.
Distinct expression patterns were evident in the gastric epithelium for GATA-4, GATA-5, and GATA-6. Neither the rat nor mouse Fabpl is active in the adult stomach (35, 61), but the Fabpl transgene is expressed in surface and pit cells (61). The native Fabpl transgene expression pattern did not bear any obvious relationship to that of the GATA factors and was not altered by mutagenesis of the GATA binding sites. Expression of the Fabpl transgene is likely the result of interactions among the many transcription factors that interact with the included Fabpl regulatory sequences (25). For example, we found dominant regulation by GATA-4 only in the intestine where HNF-1α is present, and it is possible that GATA factors may play a less important role in other tissues and/or interact with different factors.
GATA site and GATA factor specificity.
GATA-4 and GATA-5 activate the Fabpl promoter transgene primarily through the −128 site, yet all Fabpl GATA sites confer similar transactivation potential to a minimal viral promoter by these factors. These results are consistent with binding of all GATA factors to each of the Fabpl sites with similar avidity but in contrast to data demonstrating differences in binding affinity of GATA factors to DNA sequences that differ outside the core GATA nucleotides (10, 40, 85). The Fabpl GATA binding sites all share a perfect GATA core sequence but different flanking sequences. Although all three GATA factors similarly activate the viral promoter constructs, GATA-4 and GATA-5 activate the Fabpl transgene significantly better than GATA-6. Activation of the viral constructs by GATA factors are less than activation of the Fabpl transgene. These results are consistent with selective interactions between the GATA family factors and other transcription factors that bind to the Fabpl proximal promoter, including HNF-1α. Interactions with cofactors or other transcription factors that bind near a GATA site have been hypothesized to mediate preferential binding or activity of individual GATA factors (23, 27). The most proximal GATA site at −128 is located close to an HNF-1 binding site at −95 (Fig. 1), and nearby sites for C/EBP, Cdx, HNF-4, Fox, and HNF-6 family factors (see Ref. 25 and data not shown). No other factors binding to sequences surrounding the more distal GATA sites have yet been identified.
Murine expression of Fabpl and the Fabpl transgenes.
The in vivo results indicate a critical role for HNF-1α in Fabpl expression. Mutagenesis of the Fabpl transgene HNF-1 site silenced the transgene, and mice with targeted inactivation of both HNF-1α gene alleles have no endogenous Fabpl expression in hepatocytes (1). A critical role for HNF-1α activation of the sucrase-isomaltase gene was similarly identified. A transgene constructed from the proximal promoter of the sucrase-isomaltase gene was active in Caco-2 cells and murine enterocytes but silenced by mutagenesis of an HNF-1 binding site in the proximal promoter (13). These results also support the proposed central role for HNF-1α in gene expression in the intestinal epithelium (73) and liver (59).
Removal of the GATA sites in the Fabpl transgene had no effect on transgene expression, indicating an asymmetrical significance to the binding of GATA factors compared with HNF-1 factors. The observation that Fabpl expression was dependent on GATA-4 in the embryonic intestinal epithelium demonstrates that GATA-4 plays a critical role in Fabpl expression in vivo and fits with the strong Fabpl transgene activation in cultured cells by GATA-4. One explanation for the difference in phenotype between loss of binding site and loss of factor could be that a GATA site required for Fabpl expression exists outside the sequences included in the transgene, but this possibility is unlikely because the transgene is active in enterocytes and hepatocytes where the endogenous gene is expressed. Alternatively, GATA factors could be tethered to the proximal promoter without binding DNA directly through binding to HNF-1α or other factors as has been hypothesized (Fig. 11) (11).
Removal of the GATA binding sites had no effect on transgene cellular expression patterns, opening the question of their function in the Fabpl regulatory sequences. One possibility is that the GATA site provides a redundant function although GATA factors can bind to the promoter in the absence of the site. Fabpl is expressed at very high levels in intestinal epithelial cells and hepatocytes (5), so a particularly robust activator region may also be required. Alternatively, the transgene may not recapitulate some aspect of endogenous Fabpl regulation.
Experimental evidence was obtained for an interaction between GATA-4 and HNF-1α at the Fabpl promoter, where HNF-1α tethered GATA-4 but not the reverse in the cell transfection assay (Figs. 10 and 11). This asymmetrical interaction is consistent with the transgene expression data. We did not observe an interaction between GATA-5 or GATA-6 and HNF-1α or HNF-1β in this assay, despite identifying functional and physical interactions between HNF-1α and GATA-4, GATA-5, or GATA-6. These results may indicate that interaction of GATA-5 or GATA-6 with HNF-1α in the solution binding assay does not correspond to a relevant in vivo interaction or to the affinity of the factors in the cell transfection assay. Cell transfection assays performed with a sucrase-isomaltase transgene in Caco-2 cells revealed that GATA-5 could interact with HNF-1α in the absence of GATA binding sites (40) and also that GATA-4 could cooperate simultaneously with both Cdx-2 and HNF-1α in the absence of GATA binding sites, leading to the suggestion that other factors could tether GATA-4 to the promoter (11). This binding may explain the lack of a defect in small intestinal sucrase-isomaltase transgene expression when the GATA binding site is mutagenized (12). A transgene constructed with an enhancer from the adenosine deaminase gene loses murine duodenal expression when GATA sites in the enhancer are mutated (28). In this case, PDX-1 but not HNF-1 binding sites are reported to be present in the enhancer. Binding in each case may be dependent on the presence of other factors in the complex as discussed below and may differ between genes. We observed that endogenous Fabpl was silent in Gata4-deficient enterocytes despite the presence of abundant GATA-6 (Fig. 8). This finding is consistent with a lack of interaction between GATA-6 and HNF-1α in vivo or with a unique role for GATA-4 in activating Fabpl in enterocytes. Support for a unique role for GATA-4 in Fabpl expression is obtained by comparing Fabpl expression in the liver and intestine of the chimeric Gata4-deficient mice. Endogenous Fabpl is expressed only in differentiated enterocytes and hepatocytes. GATA-4 and GATA-6 as well as HNF-1α are expressed in both these cell types at embryonic day 18 when the chimeric mouse was analyzed (8, 16), yet loss of GATA-4 in the chimeric mice resulted in loss of Fabpl expression only in enterocytes. In the adult animal, GATA-6 was found in both enterocytes and hepatocytes, whereas GATA-4 was only detected in enterocytes. These results are consistent with a critical role specifically for GATA-4 in enterocytic Fabpl expression.
The in vitro data suggest that GATA-5 or GATA-6 could regulate Fabpl expression, but the in vivo data indicate this regulation is not critical in late gestational differentiated enterocytes. It is possible that GATA-6 plays a role in those cell types that support transgene expression in which GATA-6 but not GATA-4 is present: hepatocytes, proximal colonic epithelial cells, and small intestinal crypt epithelial cells. Because removal of the GATA binding sites did not affect transgene expression in any of these cell types, either GATA-6 is not required for transgene expression or GATA-6 can interact with HNF-1α or other factors in the absence of a binding site in a manner analogous to that postulated for GATA-4. Whereas HNF-1α and GATA-6 exhibited a functional and physical interaction, we did not detect evidence of these interactions in the cell transfection assays with mutant promoters. These results may indicate that the transfection assay is not as sensitive as solution binding assays as noted above or that other factors in addition to HNF-1α may be involved in tethering GATA factors to the promoter in vivo as discussed below. Specificity may arise from additional mechanisms. GATA-4 is subject to phosphorylation that can increase binding (18, 45, 51), and binding may be altered by this or other posttranslational modifications. Intestinal extracts showed only GATA-4 but not GATA-6 binding to sequences containing GATA binding sites despite the presence of both factors in this tissue, suggesting regulation of binding (11, 28). Finally, gene expression in vivo is the result of tissue development and cellular differentiation, where factors such as GATA-4 may bind to chromatin in the absence of gene activation to maintain chromatin structure competent to support activation later in development (19). These effects will not be detected by cell transfection or solution binding assays but may contribute to the activity of GATA factors to Fabpl and Fabpl transgene expression in vivo.
GATA factor interactions with other family factors.
We provide evidence for binding and functional cooperation of HNF-1α and GATA-4. It is possible that similar factor-factor interactions occur between GATA proteins and other transcription factors present at the Fabpl proximal promoter. GATA-4 can functionally and physically interact with C/EBPβ (75) and GATA-6 (17). HNF-1α will functionally and physically interact with GATA-5 (40, 80). GATA-4 interacts functionally and physically with Nkx2.5, and GATA-4 and other factors can recruit Nkx2.5 to a cardiomyocyte-specific promoter without binding of the latter factor to DNA (65). A single amino acid substitution in GATA-4 results in heart defects, and this mutation does not interfere with target gene activation by GATA-4 but ablates functional and physical interaction with the transcription factor TBX5 (32). Endodermal GATA factors have also been found to interact with Sp family transcription factors (38), dHAND (21), YY1 (6), SF1 (77), and thyroid transcription factor (47). The GATA family cofactor FOG2 is important in gene activation by endodermal GATA factors (70), which also achieve transcriptional activation, in part, by recruiting histone acetylases such as P300 (37). These numerous functional and physical interactions lend credence to the possibility that the function of GATA factors in intestinal and hepatic gene regulation may be mediated by direct interaction with other factors. For example, GATA-4, Cdx-2, and HNF-1α have been reported to form a single complex in solution, and all three factors interact functionally to activate the sucrase-isomaltase gene (11). A Cdx site is present just proximal to the HNF-1 site in the Fabpl promoter (25), and Cdx factors may participate in formation of a transcriptional initiation complex along with the other factors that bind to nearby sites. These other factors may play a critical role in binding to GATA-4.
Interactions among all the factors that interact at the proximal promoter of Fabpl or other endodermally expressed genes may be significant determinants of transcriptional activation. We previously found that a mixture of six factors that activates the Fabpl transgene exhibits cooperative activation to such an extent that individual activation by any one factor is insignificant (25). GATA-6 by itself displayed very low Fabpl transgene activation in the cell transfection assay. However, GATA-6 could activate the transgene as well as GATA-4 when HNF-1α was present (Fig. 6), and GATA-6 also bound directly to HNF-1α in the absence of DNA. A role for GATA-6 in enterocytic gene regulation has been proposed based on observations that GATA-6 transactivates transgenes constructed from the proximal promoters of intestinal genes in cell transfection assays (2, 29, 30). The evidence from the Fabpl studies indicates that GATA-6 regulation of these genes could be amplified through interaction of GATA-6 with HNF-1α or other factors. Cooperativity among a complex of transcription factors has been described as a general property of promoters where multiple factors bind (44, 82), and loss of this cooperativity has been postulated to explain the high incidence of disease due to haploinsufficiency of transcription factors (64, 81). Two human diseases have been attributed to mutations in GATA-4 that lead to a loss of cooperativity with another transcription factor but not the ability to transactivate a target gene directly. A single amino acid substitution in GATA-4 leads to a loss of interaction with TBX5 and cardiac septal defects (32). A mutation in steroidogenic factor 1 leads to a loss of cooperativity with GATA-4 and sex reversal (76). The direct interactions we observe between GATA factors and HNF-1α at the Fabpl promoter may be part of an interacting complex of factors with other major endodermal transcription factors that also bind to the promoter (25). In addition, our results raise the possibility that interactions with other factors could tether GATA factors to the promoters of genes that contain no GATA recognition sites in their regulatory DNA sequences.
This work was supported in part by funding from the National Institutes of Health Grant DK-5631TCS (to T. C. Simon), research Grant 1-FY99–606 from the March of Dimes Birth Defects Foundation (to T. C. Simon), the Washington University Diabetes Research Training Center (Grant DK-20579), the Washington University Digestive Diseases Research Core Center (Grant DK-52574), and the Sigrid Juselius Foundation (to M. Heikinheimo and D. B. Wilson).
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2004 the American Physiological Society