Transducin-like enhancer of split proteins, the human homologs of Drosophila groucho, interact with hepatic nuclear factor 3beta.

Members of the hepatic nuclear factor 3 (HNF3) family, including HNF3alpha, HNF3beta, and HNF3gamma, play important roles in embryonic development, the establishment of tissue-specific gene expression, and the regulation of gene expression in differentiated tissues. We found, using the glutathione S-transferase pull-down method, that the transducin-like Enhancer of split (TLE) proteins, which are the human homologs of Drosophila Groucho, directly associate with HNF3beta. Conserved region II of HNF3beta (amino acids 361-388) is responsible for the interaction with TLE1. A mammalian two-hybrid assay was used to confirm that this interaction occurs in vivo. Overexpression of TLE1 in HepG2 and HeLa cells decreases transactivation mediated through the C-terminal domain of HNF3beta, and Grg5, a naturally occurring dominant negative form of Groucho/TLE, also increases the transcriptional activity of this region of HNF3. These results lead us to suggest that TLE proteins could influence the expression of mammalian genes regulated by HNF3.

The HNF3 proteins are pleiotropic transcription factors, and the isoforms have selective effects. These proteins are expressed in tissues other than the liver, such as the pancreas and lung, where they are involved in establishing tissue-specific gene transcription (11)(12)(13). For example, mice homozygous for a null mutation in HNF3␣ have reduced islet glucagon gene expression, but the genes involved in hepatic gluconeogenesis are not affected, suggesting a specific role of HNF3␣ in the pancreas (6). HNF3 family members also play an important role in development. HNF3␣ and HNF3␤ are required for normal mesoderm and neural axis formation (15). Mice lacking HNF3␤ do not have a properly formed notochord, which results in defects in neural tube formation, somite organization, and gut endoderm invagination (16). Furthermore, in transgenic mice in which HNF3␤ is ectopically expressed, the hindbrain region is converted to the floorplate (17).
The three HNF3 family members share four distinct conserved regions (CRI-CRIV) (see Fig. 1). There is more than 90% homology between the family members at CRI, which functions as the DNA binding domain ( Fig. 1) (18). CRII and CRIII are located in the C terminus, whereas CRIV is located in the N terminus ( Fig. 1). CRII, CRIII, and CRIV are required for the transcriptional activity of HNF3␤ (19). CRII and CRIII are part of a position-independent transactivation domain, and deletions of either region result in decreased transactivation activity of HNF3␤ (19). Deletion of CRIV also reduces transcriptional activity of HNF3␤; however, CRIV cannot function independently, as its activity requires the participation of the C-terminal transactivation domain (CRII and CRIII) (19). Interestingly, this C-terminal transactivation domain is required for the accessory factor activity of HNF3␤ in the stimulation of PEPCK gene transcription by glucocorticoids (20), and CRIV is not involved. By contrast, CRIV plays a crucial role in the activation of apoAI gene transcription in conjunction with HNF4 (21). Thus, these domains of HNF3 are capable of distinct functional interactions with other transcription factors.
We show here that members of the transducin-like Enhancer of split (TLE) transcriptional corepressor family interact with the CRII domain of HNF3␤. Overexpression of TLE1, a member of the TLE family, specifically represses the transcriptional activity of a chimeric protein that consists of the GAL4 DBD and the HNF3␤ C-terminal transactivation domain. This inter-action could play an important role in determining the physiologic actions of HNF3␤.

EXPERIMENTAL PROCEDURES
Plasmid Construction-Oligonucleotides that were used in this report were synthesized by an Expedite 8909 oligonucleotide synthesizer (Perceptive Biosystems, Framingham, MA). The plasmids that express various GST⅐HNF3␤ chimeric proteins in Escherichia coli were generated in two steps. First, the nucleotide sequences of rat HNF3␤ corresponding to amino acids 361-458 (CRII/CRIII), 361-442 (CRII), and 388 -458 (CRIII) were generated by polymerase chain reaction using the primer pairs H3/N/361 (GCGCGGATCCGTGAGGCCCACCTGAA-GCC) and H3/C/458/E (CGGGAATTCTTAGGTCGAGTTCATAATAG), H3/N/361 and H3/C/442/E (CGGGAATTCTTAGTCTGCAGCCAGGGG-C), and H3/N/388 (GCGCGGATCCGTCATCATCATCACAGCCACCAC) and H3/C/458/E, respectively. These polymerase chain reaction fragments were then digested with EcoRI and BamHI and subcloned into the pGEX3X plasmid (Amersham Pharmacia Biotech). The construction of plasmids that encode various GAL4⅐HNF3␤ fusion proteins has been described previously (20). To construct a mammalian expression plasmid for the TLE1 protein, the cDNA of TLE1 was isolated by digesting pBluescript TLE1, which is used for the in vitro translation of TLE1, with HindIII and XbaI. This DNA fragment was then subcloned into the Rous sarcoma virus-promoter driven plasmid, pGM4 (22). The plasmid TLE1-VP16 was generated by the subcloning of a polymerase chain reaction fragment that contains full-length TLE1 cDNA (using primers CGGGAATTCATGTTCCCGCAGAGCCG and GCGCGGATCCTCAGT-AGATGACTTCATAGA) into the EcoRI and BamHI sites of pVP16 plasmid (CLONTECH). The construction of the (GAL4) 5 E1bLuc reporter plasmid has been described previously (23).
GST Pull-down Assay and Protein Purification-BL21 cells containing each of the GST⅐HNF3␤ expression plasmids were grown in LB media, and expression of the fusion proteins was induced by treating the cells with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 4 h at 37°C. Bacteria were then collected, resuspended with lysis buffer (50 mM Tris Cl, pH 8.0, 150 mM NaCl, 100 mM EDTA, 1% Triton X-100, 0.2 units/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride, 10 mM dithiothreitol), and sonicated at 4°C. After centrifugation, an aliquot of the supernatant was incubated with 50% glutathione-agarose overnight at 4°C. The agarose beads bound with various GST⅐HNF3␤ fusion proteins were washed with lysis buffer three times and incubated for 1 h at 4°C with either [ 35 S]methionine-labeled (0.1 mCi/ml) H4IIE whole cell lysate, HeLa cell nuclear extracts, or TLE1 protein that was translated in vitro using the Promega TNT system. After washing three times with lysis buffer, 2ϫ SDS sample buffer was added to the agarose beads, the mixture was boiled for 5 min, and the eluent was analyzed by SDS-PAGE. After electrophoresis, the gels were enhanced, dried, and exposed to Kodak XAR5 film, and proteins were visualized by autoradiography. H4IIE whole cell lysates were prepared by adding lysis buffer to phosphate-buffered saline-washed cells for 15 min at 4°C, and then the lysates were clarified by centrifugation. The preparation of the nuclear extracts is described elsewhere (24).
The procedure used to purify proteins that interact with GST⅐HNF3␤ fusion proteins was basically a scaled-up version of the protocol described above with some modifications. A HeLa cell nuclear extract (provided by Dr. P. A. Weil, Vanderbilt University, Nashville, TN) containing 75 mg of protein was first incubated with the agarose beads bound to GST protein to exclude proteins that interact with either GST or the agarose beads. The HeLa cell nuclear extract was then incubated with agarose beads bound to the C-terminal transactivation domain of HNF3 (GST⅐HNF3␤(CRII/CRIII)). About 2.8 mg of both the GST and GST⅐HNF3␤ proteins were used for this purpose. Proteins that specifically associated with GST⅐HNF3␤ (CRII/CRIII) were analyzed on a preparative SDS-PAGE gel. The 100 -110-kDa band was excised and sent to the Harvard Microchemistry Facility for sequencing.
Cell Culture, Transient Transfection, and Luciferase Assays-The transient transfection procedure and the maintenance of HepG2 and HeLa cells are described elsewhere (23,25). Luciferase activity was measured by the dual luciferase kit (Promega).

HNF3␤ Interacts with Proteins from H4IIE and HeLa Cells-
Affinity chromatography was used to identify proteins that interact with the C-terminal transactivation domain of HNF3␤. The C-terminal transactivation domain of HNF3␤ fused to glutathione S-transferase, GST⅐HNF3␤(CRII/CRIII), is highly expressed in E. coli (data not shown). GST or GST⅐HNF3␤(CRII/CRIII) was bound to agarose beads and incubated with a whole cell lysate obtained from [ 35 S]methionine-labeled H4IIE hepatoma cells. The beads were washed with either 150 or 500 mM NaCl, and the proteins that remained on the beads were eluted and analyzed by SDS-PAGE. Two major bands were associated with the beads containing GST⅐HNF3␤(CRII/CRIII) but not with those that contained GST ( Fig. 2A). One protein had a molecular mass of approximately 100 -110 kDa, and the other had a mass of about 84 kDa as determined by migration on SDS-PAGE gels. Washing the beads with a high salt concentration (500 mM) did not influence the interaction between GST⅐HNF3␤(CRII/CRIII) and these two proteins ( Fig. 2A). Although a nonspecific 84-kDa protein appears to bind to GST, a band of stronger intensity was bound to GST⅐HNF3␤(CRII/CRIII). Thus, we conclude that this 84-kDa protein specifically associates with GST⅐HNF3␤(CRII/ CRIII) and comigrates with a GST-interacting protein. These two proteins were also identified when the nuclear extracts prepared from [ 35 S]methionine-labeled H4IIE cells were used to perform the GST⅐HNF3␤(CRII/CRIII) pull-down experiments (data not shown). We also determined that these GST⅐HNF3␤(CRII/CRIII)-interacting proteins were present in HeLa S3 cells, which can be grown in suspension for production of large quantities of nuclear extracts for further protein purification (26,27) (Fig. 2B).
Previous studies have shown that deletion of either CRII or CRIII markedly reduces the activity of the C-terminal transactivation domain of HNF3␤ (19). We speculated that co-regulatory proteins that are important for transmitting the activity of the C-terminal transactivation domain would not interact with mutants that lack either of these domains. Two GST fusion proteins with deletions of either CRII (GST⅐HNF3␤-(CRIII)) or CRIII (GST⅐HNF3␤(CRII)) were prepared. These two proteins, along with GST or GST⅐HNF3␤(CRII/CRIII), were bound to agarose beads and incubated with nuclear extracts isolated from [ 35 S]methionine-labeled HeLa cells. GST, GST⅐HNF3␤(CRII/CRIII), GST⅐HNF3␤(CRIII), and GST⅐HNF3␤-(CRII) were all expressed at equal amounts in E. coli (data not shown). One of the two previously identified bands, the 100 -110-kDa band, interacts with both GST⅐HNF3␤(CRII/CRIII) (lane 1) and GST⅐HNF3␤(CRII) (lane 2) but not with GST⅐ HNF3␤(CRIII) (lane 3) (Fig. 2B). This suggests the presence of a nuclear protein that associates with the CRII of HNF3␤. The 84-kDa band interacts with all of the GST⅐HNF3␤ proteins but does not interact with GST alone (Fig. 2B). Thus, the interac-tion of this 84-kDa protein with the C-terminal transactivation domain of GST⅐HNF3␤ does not require either CRII or CRIII but does require amino acids 388 -442.
TLE Proteins Interact with HNF3␤-This report focuses on the 100 -110-kDa nuclear protein(s) because it appears to interact with CRII, which is a crucial component of the C-terminal transactivation domain of HNF3␤. The GST pull-down protocol used in the experiments described above was employed to purify the 100 -110-kDa protein from HeLa cell nuclear extracts. This extract was preincubated with GST-bound agarose beads to remove nonspecific binding proteins. The residual nuclear extract was then incubated with agarose beads bound with GST⅐HNF3␤(CRII/CRIII). After extensive washing, the proteins that remained on the agarose beads were eluted and subjected to a preparative SDS-PAGE gel. The 100 -110-kDa band was excised and sent to the Harvard Microchemistry Facility for sequencing. Three peptide sequences were obtained. Using the NCBI blast search protocol, two peptides had a sequence identical to that of TLE3, a member of the TLE family of proteins (28). The sequence of the other peptide is very similar to that of the other TLE proteins, TLE1, TLE2, and TLE4 (Fig. 3). Because all four members of the TLE family have a similar molecular weight and a very similar structural organization, we may have isolated more than one TLE protein in the purification process.
A cDNA encoding authentic, full-length TLE1 (28) was used to test for an interaction between this protein and the transactivation domain of HNF3␤. TLE1 was translated in vitro in the presence of [ 35 S]methionine and incubated with agarose beads bound to various GST⅐HNF3␤ fusion proteins. In vitro translated TLE1 interacted with GST⅐HNF3␤(CRII/CRIII) and with GST⅐HNF3␤(CRII) (Fig. 4, lanes 3 and 4) but not with GST⅐HNF3␤(CRIII) (Fig. 4, lane 5). These results provide additional confirmation that the proteins identified are members of the TLE family.
A mammalian two-hybrid assay was used to confirm an in vivo interaction between TLE1 and HNF3␤. The basal transcription activity of the various GAL4⅐HNF3␤ fusion proteins The sequence of TLE3 is identical to peptides 1 and 2 and shares 92% identity (12 of 13 amino acids) with peptide 3. In contrast, the sequence of TLE1 is identical to peptide 3 but shares 80% identity (12 of 15 amino acids) with the sequence of peptide 1 and 91% identity (10 of 11 amino acids) with the sequence of peptide 2. The amino acids that do not match are shown in boldface.
was monitored by cotransfecting their cognate expression vectors with a luciferase reporter plasmid that contains five copies of a GAL4 binding site positioned upstream of a E1B TATA box ((GAL4) 5 E1bLuc). Overexpression of a chimeric protein that consists of TLE1 and the VP16 activation domain (TLE1⅐VP16) in HeLa cells does not influence the transcriptional activity of the GAL4 DBD (Fig. 5). However, overexpression of TLE1⅐VP16 potentiates the transcriptional activities of both GAL4⅐HNF3␤(CRII/CRIII) and GAL4⅐HNF3␤(CRII) but not GAL4⅐HNF3␤(CRIII), when compared with the activity provided by VP16 alone (Fig. 5). These results are consistent with the in vitro GST pull-down experiments described above and suggest that the HNF3␤ CRII directly associates with TLE1 in vivo.
TLE Regulates the Activity of the Transactivation Domain of HNF3␤-TLE proteins share a high degree of homology with the Drosophila Groucho protein, which functions as a corepressor in conjunction with several transcription factors (29). We tested whether TLE proteins can modulate the activity of the C-terminal transactivation domain of HNF3␤ in order to begin to understand the physiological significance of the TLE-HNF3␤ interaction. The expression vector that encodes the TLE1 protein was cotransfected with plasmids that encode various GAL4⅐HNF3␤ chimeric proteins and the (GAL4) 5 E1bLuc reporter gene. In HepG2 hepatoma cells, the transcriptional activities of GAL4⅐HNF3␤(CRII/CRIII), GAL4⅐HNF3␤(CRIII), and GAL4⅐HNF3␤(CRII) are 7-fold, 4-fold, and 1-fold that of the GAL4 DBD, respectively (Fig. 6A) (8). Overexpression of TLE1 significantly reduced the activity of GAL4⅐HNF3␤(CRII/ CRIII) and GAL4⅐HNF3␤(CRII), but the activity of GAL4⅐ HNF3␤(CRIII), which does not interact with TLE1, was not affected (Fig. 6A and see above). As in HepG2 cells, the overexpression of TLE1 inhibits the activity of GAL4⅐HNF3␤(CRII/ CRIII) and GAL4⅐HNF3␤(CRII) in HeLa cells but has no effect on GAL4⅐HNF3␤(CRIII) in these cells (Fig. 6A). Thus, in both HepG2 and HeLa cells, TLE1 co-represses transcription through a specific interaction with the CRII of HNF3␤.
All of the constructs are more active in HeLa cells, an observation that may be related to transfection efficiency. Some qualitative differences in the responses are noteworthy, however. Deletion of CRII resulted in decreased activity in HepG2 cells and a 5-fold increase in activity in HeLa cells; thus, CRII could act in a cell-specific fashion. It appears to be an activation domain in HepG2 cells and a repression domain in HeLa cells. This observation could be explained if HeLa cells have a higher concentration of TLE, which, in the absence of CRII, is unable to function. An immunoblot analysis was performed to test this possibility. As shown in Fig. 6B, HeLa cells express approximately five times more TLE than HepG2 cells.
Homologues of Groucho/TLE have been isolated from mouse, rat, and human cDNA libraries (28,30). These proteins have an N-terminal Gln (glutamine-rich) domain, a Gly/Pro-rich sequence (Gly/Pro region), and C-terminal Ser/Pro-rich and WD40 motifs. The multimerization of Groucho/TLE, through the Gln domain, is required for its repressive effects (30 -32). The families of Groucho/TLE genes in mouse, rat, and human also each include one gene that encodes a truncated protein that only has the Gln domain and part of the Gly/Pro region (30). For instance, Grg5 is a truncated protein that lacks the C-terminal Ser/Pro and WD40 motifs. Thus, it can dimerize (through the Gln domain) with TLE1/Groucho and block its repressive function (30,33,34). An expression vector that encodes the dimerization domain of the Grg5 protein was cotransfected into HeLa cells with a plasmid that encodes the GAL4⅐HNF3␤(CRII/CRIII) chimeric protein and with the (GAL4) 5 E1bLuc reporter gene. Overexpression of Grg5 increased the transcriptional activity of GAL4⅐HNF3␤(CRII/ CRIII) by about 5-fold (Fig. 7). By contrast, overexpression of Grg5 did not potentiate the transcriptional activity of the GAL4 DBD or of GAL4⅐HNF4(128 -374), neither of which associates with Groucho/TLE (data not shown). These data suggest that Grg5 interferes with the repressive effect of TLEs in this system, probably by dimerizing with TLE1 as shown previously (30,33,34). This experiment provides further evidence that Groucho/ TLE can specifically modulate the transcriptional potential of HNF3␤ through its C-terminal transactivation domain. DISCUSSION Protein-DNA interactions are an essential component of transcription regulatory mechanisms. In recent years, it has become apparent that protein-protein interactions, which can change the magnitude or direction of a response within a given promoter, are equally important. HNF3 binds to a specific DNA element (gAF2) (glucocorticoid accessory factor 2 element) in the PEPCK gene promoter and is an essential component of the glucocorticoid response unit that enhances the transcription of this gene (8). The gAF2 element also serves as an insulin response sequence and is involved in the repression of the PEPCK gene by insulin (35). Protein-protein interactions involving HNF3 were sought in an effort to understand how this protein functions as a regulator of transcription and to help explain its role in the glucocorticoid response unit, and perhaps in the insulin response. We detected an interaction between the TLE1 transcription corepressor protein and the CRII of HNF3␤. The TLE proteins are remarkably similar in functional organization and amino acid sequence (28,36,37), so it is therefore quite possible that these proteins all interact with this domain of HNF3␤. The CRII is conserved between all three mammalian HNF3 proteins (␣, ␤, and ␥) and the Drosophila forkhead protein. Thus, it is likely that other members of the extensive HNF3 family associate with TLE proteins.
The CRII of HNF3␤ does not contain the WRPW motif that is required for the interaction between TLE/Groucho and Hairy/ Enhancer of split-like proteins (29,32,38). The CRII also does not resemble the eh1/GEH motif that is required for the interaction between TLE/Groucho and Engrailed or Goosecoid (39). CRII, however, does have an FNHPF sequence that may serve as a TLE binding site because it has two aromatic residues separated by a proline. In HepG2 cells, GAL4⅐HNF3␤(CRII/ CRIII) is a stronger transcription activator than GAL4⅐HNF3␤(CRIII), which contains CRIII but lacks CRII. This suggests that CRII is required for complete activity of the C-terminal transactivation domain of HNF3␤. However, GAL4⅐HNF3␤(CRII), which contains CRII but not CRIII, does not have transcriptional activity, so maximal activity requires both domains. Interestingly, the CRII functions as a repressor of the C-terminal transactivation domain of HNF3␤ in HeLa cells, and in this cellular context, GAL4⅐HNF3␤(CRIII) is five times more active than GAL4⅐HNF3␤(CRII/CRIII). The function of CRIII is also not completely understood. CRIII appears to independently bind to another protein(s), presumably a coactivator(s), because it can provide strong transcriptional activity in the absence of CRII. The identification of other proteins that associate with the C-terminal transactivation domains of HNF3␤ will be an important step in understanding the regulation of transcription by HNF3␤.
TLE binds to the mammalian transcription activators acute myeloid leukemia-1 and lymphoid enhancer factor-1, interactions that reduce the transactivation potential of these proteins (40,41). Data acquired from the analysis of related proteins suggest that TLE proteins could modulate the transcriptional activity of the HNF3 proteins and do so by several different mechanisms. For example, the Drosophila protein Groucho, a homolog of TLE, interacts with a number of transcription repressors, such as the Hairy, Engrailed, Goosecoid, and Hairy/ Enhancer of split-like proteins. In this context, it serves as a transcriptional corepressor (29,32,38,39,42). In certain promoter contexts, Groucho can also convert an activator into a repressor, as exemplified by its interaction with Drosophila Dorsal (43). In a third mechanism, Groucho binds to the Drosophila T cell factor, and this interaction accounts for the repressive function of Drosophila T cell factor (33,44). However, Drosophila T cell factor acts as a transcriptional activator FIG. 6. TLE1 inhibits the transactivation function of HNF3␤. Expression plasmids (2 g in HepG2, 1 g in HeLa) that encode various GAL4⅐HNF3␤ fusion constructs, with or without an expression plasmid that encodes TLE1 (10 g in HepG2, 2 g in HeLa), were cotransfected with (GAL4) 5 E1bLuc (10 g in HepG2, 5 g in HeLa) into HepG2 or HeLa cells, as displayed in A. Results are presented relative to the activity provided by the GAL4 DBD protein and represent the average Ϯ S.E. of three or more experiments. Immunoblot analysis of HepG2 (lane 1) and HeLa (lane2) cells lysates is illustrated in B. Cell lysates (30 g per lane) were separated by SDS-PAGE and analyzed by immunoblot analysis using a monoclonal antibody that recognizes all of the TLE isoforms. when it associates with Armadillo, a coactivator (33,44). The latter mechanism has precedent in mammalian cells. NcoR, a co-repressor, interacts with and blunts the transactivation function of Pit-1 (45). The activity of Pit-1 is apparently regulated by the balance between a corepressor complex that contains NcoR and a coactivator complex that contains CREBbinding protein/p300 (45). Finally, the ratio of TLE and Grg5 proteins in cells could also influence the transactivation potential of HNF3. Grg5, a naturally occurring dominant-negative form of Groucho/TLE, can also potentiate the transcriptional activities of several Groucho/TLE-associated proteins, such as PRD1-BF/Blimp-1 and Drosophila T cell factor (33,34). Thus, it is possible that the ratio of coactivators and corepressors bound to the C-terminal transactivation domain determines the activity of HNF3.
Although the physiologic significance of the TLE-HNF3 interaction has not been established, certain clues suggest that this could be important. For example, the TLE and HNF3 proteins play critical roles in embryonic development. In mammals, the expression of the TLE1 and TLE3 genes correlates with segmentation, central and peripheral neurogenesis, and epithelial differentiation. This expression pattern is analogous to the role Groucho plays in Drosophila development (46). The absence of HNF3␤ in mice results in major defects in the development of the neural axis, somite organization, and gut endoderm invagination and an early embryonic lethal phenotype (15,16). An interaction between TLE proteins and HNF3␤ may therefore have an important role in these developmental processes.
HNF3 has not been directly linked to the repression of genes, but HNF3 binding sites are required for the induction of gene transcription by glucocorticoids in the PEPCK, tyrosine aminotransferase, and insulin-like growth factor-binding protein-1 gene promoters, and these sites overlap with the insulin response sequences that mediate the repressive effect of insulin on these genes (9,47,48). According to the cis-trans model of gene regulation, there is presumably a protein that binds to the insulin response sequence and mediates the insulin response; however, this protein has not been unequivocally identified. The possibility that HNF3 participates in the insulin response through a protein-protein interaction has not been excluded. Insulin could enhance either the interaction between HNF3 and TLE (as with the Groucho-Dorsal interaction), or the repressive activity of TLE (as with the interactions of Groucho with Hairy, Engrailed, Goosecoid, or Hairy/Enhancer of split-like). Alternatively, because phosphorylation plays an important role in the nuclear localization of TLE (49), the insulin signaling pathway may induce the phosphorylation of TLE and thereby increase the amount of TLE in the nucleus, which would favor its binding to HNF3. In this regard, it is of interest to note that the activity of Groucho is influenced by receptor tyrosine kinase signaling pathways, such as Torso and the epidermal growth factor receptor (14,50). Studies designed to test for a role of TLE in insulinmediated gene repression are under way.