Friend of GATA 2 physically interacts with chicken ovalbumin upstream promoter-TF2 (COUP-TF2) and COUP-TF3 and represses COUP-TF2-dependent activation of the atrial natriuretic factor promoter.

Friend of GATA (FOG)-2 is a multi-zinc finger transcriptional corepressor protein that binds specifically to GATA4. Gene targeting studies have demonstrated that FOG-2 is required for normal cardiac morphogenesis, including the development of the coronary vasculature, left ventricular compact zone, and heart valves. To better understand the molecular mechanisms by which FOG-2 regulates these cardiac developmental programs, we screened a mouse day 11 embryo library using a yeast two-hybrid interaction trap with the fifth and sixth zinc fingers of FOG-2 as bait. Using this approach, we isolated clones encoding the orphan nuclear receptors chicken ovalbumin upstream promoter-transcription factor (COUP-TF) 2 and COUP-TF3. COUP-TF2-null embryos die during embryonic development with defective angiogenesis and cardiac defects, a pattern that partly resembles the FOG-2-null phenotype. The interaction between COUP-TF2 and FOG-2 in mammalian cells was confirmed by co-immunoprecipitation of these proteins from transfected COS-7 cells. The sites of binding interaction between COUP-TF2 and FOG-2 were mapped to zinc fingers 5 and 6 and fingers 7 and 8 of FOG-2 and to the carboxyl terminus of the COUP-TF proteins. Binding to COUP-TF2 was specific because FOG-2 did not interact with the ligand-binding domains of retinoid X receptor alpha, glucocorticoid receptor, and peroxisome proliferating antigen receptor gamma, which are related to the COUP-TF proteins. Full-length FOG-2 markedly enhanced transcriptional repression by GAL4-COUP-TF2(117-414), but not by a COUP-TF2 repression domain mutant. Moreover, FOG-2 repressed COUP-TF2dependent synergistic activation of the atrial natriuretic factor promoter by both GATA4 and the FOG-2-independent mutant GATA4-E215K. Taken together, these findings suggest that FOG-2 functions as a corepressor for both GATA and COUP-TF proteins.

Transcriptional cofactors that bind multiple DNA-binding proteins can broadly affect the transcription of and developmental programs underlying mammalian organogenesis (1). The nuclear proteins Friend of GATA (FOG) 1 -1 and FOG-2 are multi-zinc finger transcriptional corepressors that bind specifically to members of the GATA family of transcription factors. FOG-1 is required for normal erythroid and megakaryocyte development (2)(3)(4), whereas FOG-2 is required for cardiac development (5,6). Mammalian FOG-1 and FOG-2 originate from a single ancestral gene represented by the Drosophila gene Ush. FOG-1/FOG-2/Ush have both CCHH zinc fingers, commonly regarded as RNA and DNA binding structures, and a distinct subclass of zinc fingers identified by the CCHC motif that bind to protein (7). The location and the tertiary structure of the CCHC fingers within the FOG-1/FOG-2/Ush molecules are conserved, suggesting that the multiple zinc fingers have distinct functions that have been retained during evolution (8).
The FOG-1/FOG-2 proteins were first identified by their interaction with the GATA family of transcription factors, which have two CCCC zinc fingers (9 -11). The GATA carboxyl zinc finger binds the WGATAR DNA element and transcription factors including the cardiac homeobox protein Nkx2.5 (12,13). The amino zinc finger stabilizes DNA binding and also binds the FOG-1/FOG-2 CCHC zinc fingers (9). Among the FOG-2 CCHC zinc fingers, the GATA proteins bind zinc fingers 1 and 6 strongly and fingers 5 and 8 weakly but do not bind to finger 7 (8,14). Conserved amino acids required for binding have been identified within the first zinc finger of FOG-1 and the amino zinc finger of GATA (14). GATA proteins do not bind the FOG-2 CCHH fingers.
In Drosophila, Ush is co-expressed in the cardiogenic mesoderm with the GATA homologue pannier. Loss of Ush results in overproduction of cardiac and precardiac cells, whereas Ush overexpression inhibits cardioblast formation (23). Drosophila cardioblasts harboring hypomorphic and null Ush alleles have impaired cell migration, whereas embryos transheterozygous for Ush showed excessive migration. FOG-2-null mice die during embryonic development with complex cardiac defects including tricuspid atresia, pulmonic stenosis, atrial septal defect, ventricular septal defect, hypoplasia of the left ventricular compact zone, and absent coronary vessel formation (5,6). The complex valvular and septal abnormalities suggest impaired migration of nonproliferating myocardial cells into the superior and inferior atrioventricular cushions. In response to inductive signals from the myocardium, subpopulations of proepicardial and epicardial cells undergo an epithelial-mesenchymal transformation and invade the myocardium to form the coronary vessels. The absent coronary vasculature and tricuspid valve in FOG-2-null embryos likely reflect the loss of FOG-2-dependent mesenchymal signaling and migration (5,6).
The molecular mechanism(s) underlying the congenital heart defects seen in FOG-2-deficient mice is poorly understood. Although these abnormalities may reflect dysregulation of a purely GATA-dependent transcriptional program, we hypothesized that the complex phenotype might alternatively reflect a loss of FOG-2 interaction with additional families of transcription factors. To address this possibility, we screened a mouse embryonic day 11 library using a FOG-2 bait in a yeast two-hybrid interaction trap. Using this approach, we isolated clones encoding the orphan nuclear receptors COUP-TF2 and COUP-TF3 (24 -26). COUP-TF2 is required for development of the sinus venosus and cardiac atrium beyond a primitive tube, as well as for remodeling of the primitive capillary plexus (27). Like FOG-2, the cardiovascular abnormalities in COUP-TF2null embryos demonstrate an important role of COUP-TF2 in mesenchymal-endothelial and -epithelial interactions (27). This report details our studies on the molecular interaction of FOG-2 with COUP-TF2, in which we demonstrate that FOG-2 can serve as a corepressor protein for COUP-TF2 in addition to GATA4.
Co-immunoprecipitation-COS-7 cells transfected with expression plasmids using the Superfect transfection reagent (Qiagen) were harvested 24 -48 h after transfection in radioimmune precipitation buffer (28) supplemented with the Complete mixture of protease inhibitors (Roche Molecular Biochemicals). Preimmune or immune rabbit antisera was mixed with clarified cell extract and radioimmune precipitation buffer supplemented with 1 mg/ml bovine serum albumin for at least 1 h, followed by precipitation with protein A-agarose beads (Roche Molecular Biochemicals). After washing in radioimmune precipitation buffer three times on ice, proteins were eluted with sample buffer, subjected to 8% SDS-polyacrylamide gel electrophoresis, and blotted COUP-TF2 Interacts with FOG-2 onto nitrocellulose. The membrane was then divided, and the upper half was probed with dilute FOG-2 antisera, and the lower half was probed with anti-X-press (Invitrogen) followed by anti-rabbit or anti-mouse IgG-horseradish peroxidase. The chemiluminescent image was developed on Kodak BioMax MR film.
Transfection-NIH 3T3, COS-7, and CV-1 cells were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum (Summit, Fort Collins, CO), 100 units/ml penicillin, and 100 g/ml streptomycin in a humidified atmosphere at 37°C with 5% CO 2 . 293 cells were cultured in the same medium supplemented with 20 mM HEPES and 1 mM nonessential amino acids. Transient transfection assays were performed in 12-or 24-well dishes with 50 ng of reporter plasmid, and a total of 225 ng of expression plasmid and 25 ng of cytomegalovirus ␤-galactosidase served as a control for transfectional efficiency. At 24 -48 h after transfection, cellular extracts were harvested in Reporter Lysis Buffer (Promega), and luciferase activity was measured on a luminometer (Autolumat 953; EG&G, Gaithersburg, MD) using the Promega luciferase system; ␤-galactosidase activity was measured by conversion of o-nitrophenyl ␤-D-galactopyranoside. Each sample was performed in triplicate, and each experiment was repeated at least three times.
Statistical Analysis-The mean and standard error of the mean were determined for replicate samples. For multiple treatment groups, a factorial analysis of variance was applied followed by Fisher's least significant difference test. A p value of less than 0.05 was considered significant.  1A). Northern analysis confirmed co-expression of FOG-2 and COUP-TF2 in the adult mouse heart, and in situ hybridization of day 13.5 mouse embryos confirmed co-expression in the developing atria and ventricles (data not shown).
COUP-TF2 Interacts with Multiple FOG-2 Zinc Fingers-To assess the relative binding of COUP-TF2 with the zinc fingers of FOG-2, we performed an in vitro binding assay with GST-FOG-2 fusion proteins. The COUP-TF2 in vitro translation product was selectively retained by GST-FOG-2 fusion proteins including the seventh and eight (F7-8) and the fifth and sixth (F5-6) zinc fingers to a greater extent than the sixth zinc finger (F6) or the first four zinc fingers (F1-4). COUP-TF2 was not retained by the fingerless amino terminus of FOG-2 or by GST alone (Fig. 2, top panel). This binding pattern differed from that of GATA4, which interacted equally well with all of the GST-FOG-2 fusion proteins containing zinc fingers (Fig. 2, middle  panel). The most notable difference was the weak interaction of COUP-TF2 with zinc finger 6, whereas GATA4 interacted strongly with zinc finger 6. GST-FOG-2 fusion proteins did not retain [ 35 S]methionine-labeled luciferase (Fig. 2, bottom panel).
Interaction of COUP-TF2 with FOG-2 is Zinc Finger-dependent-Because COUP-TF2 bound GST-FOG-2 fusion proteins containing zinc fingers, we hypothesized that this interaction required intact FOG-2 zinc fingers. In the mammalian twohybrid system, the fifth and sixth zinc fingers of FOG-2(aa 497-831) (F5F6) fused to the GAL4 DNA-binding domain were tested for their ability to interact with COUP-TF2(117-414) fused to the VP16 transcriptional activation domain. Interaction of COUP-TF2(117-414) with F5F6 is demonstrated by an almost 4-fold activation of luciferase activity compared with GAL4 (Fig. 3). Next, plasmids encoding GAL4-FOG-2 cysteineto-serine mutations selectively disrupting the fifth zinc finger  (LBD). B, full-length COUP-TF2 co-immunoprecipitated with FOG-2. Extracts from COS-7 cells transfected with expression plasmids for FOG-2 (left two lanes), COUP-TF2 (middle two lanes), and FOG-2/COUP-TF2 (right two lanes) were subjected to immunoprecipitation (IP) with either preimmune (P) or immune (I) FOG-2 antisera. The proteins were separated by SDS-polyacrylamide gel electrophoresis and blotted on nitrocellulose. The membrane was then cut, and the lower half was probed by Western blotting (WB) with the anti-X-press antibody to detect recombinant COUP-TF2, whereas the upper half was probed with rabbit anti-FOG-2 antisera. A representative of four experiments is shown, demonstrating that COUP-TF2 was isolated only by co-immunoprecipitation with FOG-2.
(M5F6), sixth zinc finger (F5M6), and both the fifth and sixth zinc fingers (M5M6) were tested for interaction with VP16-COUP-TF2(117-414). Western blotting confirmed expression of each GAL4 fusion protein (data not shown). Mutation of the fifth zinc finger reduced the interaction of VP16-COUP-TF2(117-414) with GAL4-FOG-2(aa 497-831) by almost half, whereas mutation of finger 6 had less of an effect (Fig. 3). Mutation of both FOG-2 fingers 5 and 6 abolished the interaction. This finding demonstrates that FOG-2 zinc fingers are necessary for interaction with COUP-TF2.
FOG-2 Interacts Specifically with COUP-TF1, COUP-TF2, and COUP-TF3-In the absence of ligand, nuclear hormone receptors partner with specific corepressor proteins, which repress transcription (1). We performed in vitro binding assays with GST-FOG-2-F5-6 fusion protein to determine the specificity of interaction between FOG-2 and members of the COUP-TF and nuclear hormone family of proteins. The in vitro translation products of full-length GATA4, COUP-TF1, and COUP-TF3, but not those of RXR␣, were selectively retained by GST-FOG-2-F5-6 compared with GST alone (Fig. 5A). Next, we used a yeast two-hybrid assay to test the interaction of FOG-2 with the ligand-binding domain of several nuclear hormone receptor proteins fused to the GAL4 transcriptional activation domain. A liquid ␤-galactosidase assay provided a After the radiolabeled proteins had been incubated with immobilized GST or GST-FOG-2 fusion proteins, the beads were washed, and the bound input was eluted with sample buffer and subjected to 15% SDS-polyacrylamide gel electrophoresis. COUP-TF2 was selectively retained by GST-FOG-2 F7-8 and F5-6 to a greater extent than F6, F1-4, amino terminus (N-term), or GST. GATA4 was retained by all of the GST-FOG-2 proteins that included zinc fingers. The GST-FOG-2 fusion proteins did not retain luciferase. Coomassie Blue staining of the gel demonstrated strong expression of each GST protein (data not shown). A representative of three experiments is shown.

FIG. 3. The interaction of COUP-TF2 with FOG-2 requires FOG-2 zinc fingers.
Site-directed polymerase chain reaction-based mutagenesis was used to introduce selective mutations of cysteine residues in FOG-2(aa 497-831) required for zinc finger formation. The mutant constructs were cloned into pM vector for expression as fusion proteins with the GAL4 DNA-binding domain. These plasmids were transiently transfected into NIH 3T3 cells with a plasmid encoding VP16-COUP-TF2(117-414) and a GAL4-responsive SV40 promoter luciferase reporter plasmid. The mean Ϯ S.E. fold activation for each pM-FOG-2 construct co-transfected with pVP16-COUP-TF2(117-414) was normalized to the activity of the same plasmid co-transfected with pVP16. GAL4-FOG-2-F5F6 interacted with VP16-COUP-TF2(117-414) to a greater degree than the GAL4-FOG-2 zinc finger 6 mutant (F5M6) and the zinc finger 5 mutant (M5F6). Mutation of both zinc fingers abolished binding, which demonstrates that FOG-2 zinc fingers are required for COUP-TF2 interaction. Results from three experiments performed in triplicate are shown. ‫,ء‬ a statistically significant (p Ͻ 0.05) interaction compared with GAL4.
COUP-TF2 Synergistically Activates the ANF Promoter with GATA4 -COUP-TF2 is required for normal atrial development (27). To further investigate the role of COUP-TF2 in atriumspecific transcription, we examined the effect of COUP-TF2 on a rat ANF (Ϫ638 to ϩ62)-promoter luciferase reporter plasmid (kindly provided by Kenneth R. Chien (34)). We found that GATA4 produced a significant activation of the ANF-luciferase reporter plasmid (mean Ϯ S.E., 21.1 Ϯ 1.4-fold; p ϭ 0.028); in contrast, His-COUP-TF2 alone had little effect on ANF promoter activity (Fig. 7). Co-expression of full-length His-COUP-TF2 with GATA4 produced a synergistic dose-dependent activation of the ANF-luciferase reporter plasmid that was highly significant (up to 300-fold). The carboxyl-terminal deletion mutant His-COUP-TF2(1-335) did not influence GATA-dependent transcription (data not shown).

DISCUSSION
In screening for novel interacting partners of FOG-2, the fifth and sixth zinc fingers of FOG-2 were used as bait because they are both CCHC zinc fingers that interact with the amino CCCC zinc finger of GATA4 (14). Our yeast-two hybrid screen isolated cDNA clones encoding the ligand-binding domains of both COUP-TF2 and COUP-TF3, which was unexpected because crystallography has not demonstrated a zinc finger within this region of nuclear receptors (30,33). Because FOG-2 could not bind COUP-TF2 by a finger-finger interaction, we compared which FOG-2 zinc fingers bound GATA4 and COUP-TF2. In agreement with previously published studies (8,14), GATA4 was retained well by GST-FOG-2 proteins encoding zinc finger 6, zinc fingers 7 and 8, and zinc fingers 1-4. COUP-TF2 was retained strongly by fingers 7 and 8, followed by zinc fingers 5 and 6, and COUP-TF2 was retained weakly by finger 6 alone and fingers 1-4 (Fig. 2). If COUP-TF2 interacted with the same zinc fingers as GATA4, with the same affinities, then we would have expected the same pattern of binding by GST-FOG-2 proteins. Our finding of distinct patterns of zinc finger interaction by COUP-TF2 and GATA4 is further supported by our mammalian two-hybrid analysis, which found that mutation of zinc finger 5 had a greater effect on VP16-COUP-TF2(117-414) binding than mutation of finger 6 (Fig. 3). These results demonstrate that GATA4 interacts with zinc finger 6 more strongly than with zinc finger 5, whereas COUP-TF2 clearly favors zinc finger 5 more than zinc finger 6. Whereas both GATA4 and COUP-TF2 can bind both zinc fingers 5 and 6, their relative affinity for different FOG-2 zinc fingers suggests the possibility that FOG-2 could interact with both proteins simultaneously.
The nuclear receptor ligand-binding domain serves as a molecular switch allowing ligand-induced conversion from transcriptional repression to activation (29). Ligand-binding domains are compact structures with a conserved architecture shaped by 10 helices forming a hydrophobic core tethered to the ligand-dependent transactivation domain, AF2, by the tenth helix, H10 (33). H10 is also a protein interaction surface, forming most of the interface for nuclear receptor dimers. Mutation of three leucine residues within the H10 of COUP-TF2 pre-vented FOG-2 binding (Fig. 4). This clearly demonstrates that FOG-2 interacts with the carboxyl terminus of COUP-TF2, a region that is also bound by other corepressor proteins. In addition to nuclear receptor-corepressor (N-Cor) and silencing mediator and thyroid hormone receptor (SMRT) (35), the COUP-TF2 ligand-binding domain binds Alien (36) and the nuclear receptor-corepressor variant RIP13⌬1 (37) corepressor proteins. In the brain, COUP-TF1 is co-expressed with the zinc finger proteins CTIP1 and CTIP2, which bind COUP-TF1 and serve as corepressor proteins (38). We have demonstrated that FOG-2 selectively interacts with the ligand-binding domains of all three COUP-TF proteins, but not with RXR␣, PPAR␥, or GR (Fig. 5). Besides differences in primary sequence (39), the lack of the second helix in the COUP-TF proteins (31) may alter the ligand-binding domain sufficiently to explain their selective interaction with FOG-2. Alien also selectively interacts with some, but not all, nuclear hormone receptors (36). The selective use of tissue-restricted corepressor proteins may be a mechanism by which widely expressed DNA-binding proteins can cause organ-specific transcription. transcription of the ANF promoter. B, FOG-2 represses COUP-TF2-dependent synergistic activation of transcription by GATA4-E215K. 100 ng of pcDNA3.1/His-COUP-TF2(1-414) was co-transfected with each sample. Co-expression of COUP-TF2 with GATA4 produced a 7.1-fold increase in ANF-luciferase activity, and an 11.2-fold increase was produced with GATA4-E215K. FOG-2 significantly repressed the COUP-TF2-dependent synergistic luciferase activity for both GATA4 and GATA4-E215K. COUP-TF proteins repress transcription by binding directly to chromatin as a homodimer or heterodimer (40) (active repression) or after forming a dimer with a different member of the nuclear receptor family (transrepression) (25). Sequestration of coactivators into non-DNA-binding complexes by the COUP-TF proteins has also been proposed, but this mechanism of repression has been disputed (31). Active repression by COUP-TF1 requires the entire ligand-binding domain. Removal of H11, H12, and AF2 results in a loss of COUP-TF1mediated repression, suggesting their role in binding corepressor proteins (35). Using GAL4-COUP-TF2 fusion proteins, we have demonstrated that deletion of the carboxyl terminus of COUP-TF2 abolishes active repression, FOG-2 interaction, and FOG-2-mediated repression. Within the ligand-binding domain, mutations of H1 disrupt binding by nuclear receptorcorepressor and silencing mediator and thyroid hormone receptor; however, H1 mutation likely changes the overall ligand-binding domain structure (33). The crystal structure of nuclear receptors has shown that H10 is an exposed proteinprotein interaction surface (30). Mutation of the leucine residues performed in this study likely altered the conformation of H10 on the ligand-binding domain (33). We have demonstrated that this mutation decreases active repression by GAL4-COUP-TF2, suggesting that this site may also serve as a corepressor binding site. We found that mutation of H10 prevents FOG-2 binding, such that FOG-2 can no longer serve as a corepressor for COUP-TF2. Taken together our results are consistent with a model in which FOG-2 interacts with H10 as well as regions affected by H10 such as AF2, and that the FOG-2 binding site is also used by other corepressor proteins.
Although COUP-TF proteins are generally considered to be repressors of transcription, there are examples of COUP-TFdependent transcriptional activation (41,42). Direct promoter binding by COUP-TF is required for complete induction of phosphoenolpyruvate carboxykinase gene transcription by glucocorticoids (43,44). COUP-TF1 binds Sp1 to activate the NFGI-A gene expression (45). The molecular mechanism of COUP-TF-mediated activation includes binding of steroid receptor activator 1 and p300 coactivator proteins to the carboxyl terminus of COUP-TF1 (45). To our knowledge, we now report for the first time the synergistic activation of the cardiacspecific ANF promoter by GATA4 and COUP-TF2. There are several potential mechanisms that may explain the synergistic activation of the ANF promoter by COUP-TF2 and GATA4, and this is the subject of active work.
FOG-2 repressed COUP-TF2-dependent synergistic activation of the ANF promoter by GATA4. This result left open the possibility that FOG-2 repressed the synergistic activation solely by binding GATA4. To separate the ability of FOG-2 to repress GATA4-dependent transcription from its effects on COUP-TF2, we used the GATA4-E215K mutant, which does not interact with FOG-2 (22). We show that FOG-2 can repress GATA4-dependent but not GATA4-E215K-dependent transcriptional activation of the ANF promoter. However, FOG-2 repressed transcriptional activation by GATA4-E215K in the presence of COUP-TF2, which demonstrates that FOG-2 can serve as a corepressor protein for COUP-TF2 on a cardiacspecific promoter. It is noteworthy that FOG-2 repression of GATA4-E215K/COUP-TF2 transcription was not complete. Crispino et al. (46) created mice harboring the GATA4-V217G mutation (numbering differs from the published report, reflecting mouse GATA4 sequence in GenBank accession number NP 032118), which produces a GATA4 mutant that does not interact with FOG (46). Mice homozygous for this mutation resembled the FOG-2-deficient mice, yet there were differences in eHAND expression and orientation of the outflow tracts. The incomplete phenocopy of the GATA4-V217G knock-in and the FOG-2 knock-out may reflect an interaction of GATA4 with a different FOG protein (46), yet our results raise the possibility that FOG-2 may be interacting with COUP-TF2 in concert with GATA4.
In summary, we have demonstrated the interaction of FOG-2 with a transcriptional regulator other than the GATA proteins. The physiologic importance of this interaction is underscored by the finding that FOG-2-null and COUP-TF2-null mice have defects related to mesenchymal-epithelial and -endothelial interactions, which are important for heart and blood vessel formation. We have demonstrated preferential binding of different zinc fingers by COUP-TF2 and GATA4. COUP-TF2 can utilize FOG-2 as a corepressor protein to down-regulate transcription from both a heterologous promoter and a cardiacspecific promoter. Taken together, these results suggest that FOG-2 interacts with multiple cardiac transcription factors to regulate the complex program of cardiac morphogenesis.