The N Termini of Friend of GATA (FOG) Proteins Define a Novel Transcriptional Repression Motif and a Superfamily of Transcriptional Repressors*

Members of the Friend of GATA (FOG) family of transcriptional co-factors are required for the development of both the cardiovascular and hematopoietic systems. FOG proteins physically interact with members of the GATA family of transcriptional activators and modulate their activity. We have previously shown that FOG-2 can bind to the N-terminal zinc finger of GATA4 and, via this interaction, repress GATA4-mediated transcriptional activation of various cardiac promoters. In this report we further characterize the domain of FOG-2 necessary for repression of GATA4 transcriptional activity. We show that FOG-2-mediated repression is not blocked by the histone deacetylase inhibitor tricostatin A, suggesting that FOG-2 repression of GATA4 occurs via a histone deacetylase independent mechanism. N-terminal deletion mutants of FOG-2 revealed that the first 12 amino acids of FOG-2 are necessary for FOG-2-mediated repression. Fusion of these 12 amino acids to the DNA binding domain of GAL4 demonstrated that this region is sufficient to mediate transcriptional repression even when recruited to a heterologous promoter. Single amino acid substitutions within this N-terminal domain of FOG-2 defined the critical amino acid sequence as RRKQxxPxxI. Interestingly, a search of the NCBI protein data base identified several other partially characterized zinc finger transcriptional repressors from various vertebrate species that contained this motif at their N terminus. Taken together, these observations define a novel transcriptional repression motif and a superfamily of zinc finger transcriptional repressors.

It is becoming apparent that transcriptional repression plays an important role in the development of organ systems in many different model organisms. In Drosophila, the repressors knirps and even-skipped have been shown to be critical for the initial formation of body segments within the developing embryo (1,2). In mammals, the transcriptional repressor N-CoR is critical for normal thymocyte, erythrocyte, and central nervous system development (3), whereas chicken ovalbumin upstream promoter transcription factor II (coup-TFII) is required for the development of the heart and vascular system (4). Transcriptional repressors can be broadly categorized into three classes; those that inhibit the basal transcriptional machinery, those that block the function of activators, and those that modulate chromatin structure (5). Inhibitors of the basal transcriptional machinery include even-skipped, which is thought to block transcription by binding directly to TATA-binding protein (TBP) 1 and preventing TBP interaction with the TATA box (6). An example of a repressor that functions through blocking a transcriptional activator is the repressor Id, which inhibits transcription by binding to other basic helix-loop-helix-containing transcriptional activators and blocking their ability to bind DNA (7). Inhibitors of activator activity may also function by increasing activator degradation or altering the subcellular localization of the activator (5). Finally, repressors that alter chromatin structure are often dependent on histone deacetylases (HDACs) and include repressors such as retinoblastoma tumor suppressor protein (Rb) and C-terminal-binding protein (CtBP) (8 -10). Repressors such as these are thought to function by recruiting various HDACs to the local chromatin, resulting in the deacetylation of the N-terminal tail of the histones within that chromatin region, leading to a more condensed chromatin structure and inhibiting access of other transcriptional activators to the target gene. Other mechanisms of chromatin remodeling include the inhibition of histone acetyltransferases or the recruitment and activation of histone methyltransferases (5). It is likely that we are just beginning to realize the complexity of repressive mechanisms involved in the regulation of development.
The FOG family of proteins are transcriptional co-factors that have been shown to be critical for development of the hematopoietic as well as cardiovascular system in flies, frogs, and mice (11)(12)(13)(14)(15). Indeed, mice deficient in FOG-1 fail to form mature erythrocytes or megakaryocytes and die at embryonic day (E) 11.5 due to severe anemia (13). FOG-1 also plays a role in heart development as mice with a FOG-1 gene disruption in the developing cardiac and vascular endothelium die at E14.5 secondary to a double outlet right ventricle and a common atrioventricular valve (16). Mice deficient in FOG-2 die in utero at E13.5 of cardiac malformations that include a double outlet right ventricle, pulmonic stenosis, a common atrioventricular valve, ventricular wall hypoplasia, and lack of coronary arteries (11,12). Taken together, these results demonstrate the importance of FOG proteins for the development of the hematopoietic and cardiovascular system. All FOG proteins contain multiple zinc finger domains and physically interact with members of the GATA family of transcriptional activators (17)(18)(19)(20)(21). FOG proteins specifically bind to the N-terminal zinc finger of GATA factors, and this interaction can be disrupted by point mutations within this zinc finger (18,22). Mice homozygous for such mutations in GATA4 die in utero of congenital heart defects that are similar to those seen in FOG-2-deficient mice (23). Furthermore, mice with mutations in the N-terminal zinc fingers of GATA1 and GATA2 recapitulate the phenotype seen in FOG-1deficient mice (24). These results support the notion that interaction with GATA factors is required for function of FOG proteins in development.
The FOG-GATA interaction can result in either repression or synergistic activation of target promoters (18,19,(25)(26)(27)(28)(29)(30)(31). Recent work suggests that the presence of a binding site for the Ets-family transcription factor Fli-1 in close proximity to a GATA site can convert FOG-1 from a co-repressor to a coactivator (26). However, in most cell and promoter contexts examined to date, FOG proteins behave as transcriptional corepressors, decreasing the ability of GATA factors to activate transcription. One potential mechanism for this repression would be the recruitment of other transcriptional repressors to the promoter complex. One such repressor is the chicken ovalbumin upstream promoter transcription factor II, which has been shown to interact with a domain of FOG-2 that includes zinc fingers 5 and 6 (32). However, a truncated form of FOG-2 lacking this domain is still able to efficiently mediate repression in vitro (28). FOG proteins have also been shown to interact with the CtBP co-repressor family. Both FOG-1 and FOG-2 have consensus binding sites for CtBP proteins and have been shown to interact with them in vitro (15,20,28,33). However, disruption of this interaction by mutagenesis does not abolish the ability of FOG-2 to repress transcription in vitro (20,28). In addition, targeted disruption of the CtBPinteracting domain of FOG-1 in mice results in no obvious hematopoietic phenotype, suggesting that the FOG-CtBP interaction is not required for erythroid or megakaryocyte development (33). Thus, the mechanism by which FOG proteins mediate transcriptional repression is complex and incompletely understood.
In our previous work we demonstrated that a transcriptional repression domain was also present in the N terminus of FOG-1 and FOG-2 (28). In this report we further characterize this repression domain and refine it to the first 12 amino acids within FOG proteins. We show this domain to be both necessary and sufficient to mediate transcriptional repression when linked to a heterologous DNA binding domain. We demonstrate that this repression is mediated in an HDAC-independent fashion and define the critical amino acid residues within this domain that are required for repression. Furthermore, we demonstrate that these amino acids are conserved across vertebrate species in the FOG proteins and have identified this repression motif in other partially characterized zinc finger containing transcriptional repressors. Taken together, these observations define a novel transcriptional repression family and a repression motif that may be critical in a number of developmental processes.

EXPERIMENTAL PROCEDURES
Plasmid Construction-The constructs p-638 ANF GH, pSV40Gal4-Luc, pVR␤Gal, pcDNA GATA4, and pcDNA FOG-2 have been described previously (28). Expression constructs for mutant FOG-2 proteins were generated using PCR and the following primers: 5Ј-TTCAGGTGCATT- The resulting fragments were digested with BamH1 and XbaI and inserted into pcDNA3. Subsequently, a 3.5-kilobase XbaI fragment encoding the 3Ј end of the coding region of FOG-2 (1208 -4770 bp) from pcDNA FOG-2 was inserted into the XbaI site to generate the final construct.
Cell Culture and Transfections-Murine NIH 3T3 fibroblasts were cultured at 37°C, 5% CO 2 in Dulbecco's modified Eagle's medium, 10% fetal bovine serum, 1% glutamine, and 1% penicillin/streptomycin (In- Reporter Assays-Forty-eight hours after transfection, medium was harvested. Cells were washed twice with phosphate-buffered saline, and then 0.3 ml of 1ϫ reporter lysis buffer (Promega, Madison, WI) was added to the cells. After a 15-min incubation at room temperature, cells were scraped from the plate, lysis buffer was removed, and cell debris was pelleted. Five l of this lysate was used in a Bradford protein assay to determine protein concentration (Bio-Rad). ␤-Galactosidase activity was measured in each cell lysate by mixing 15 l of lysate with 85 l of 1ϫ reporter lysis buffer and 100 l of 2ϫ ␤-galactosidase assay buffer (200 mM sodium phosphate, pH 7.3, 2 mM MgCl 2 , 100 mM ␤-mercaptoethanol, and 1.33 mg/ml o-nitrophenyl ␤-D-galactoside) and incubating at 37°C in a Molecular Devices 96-well microtiter plate reader, with absorbance at 420 nM, read every 60 s. Relative ␤-galactosidase activity was taken as the change in absorbance over time and normalized to the total protein concentration in the lysate. Luciferase activity was determined using 20 l of cell lysate and 100 l of luciferin (Promega) and measured in a TD-20/20 luminometer. Human growth hormone (hGH) was measured using 1-5 l of cell media and a commercially available enzyme-linked immunosorbent assay (Roche Applied Science). Relative promoter activity was calculated as the hGH concentration divided by the relative ␤-galactosidase activity.

FOG-2-mediated Repression of GATA4
Activity Is HDACindependent-As a first step in characterizing the nature of the repression domain of FOG-2, we used the HDAC inhibitor, tricostatin A (TSA), to determine the dependence of FOG-2 on HDAC activity. Tricostatin A has been previously shown to inhibit both class I and class II HDACs and relieve transcriptional repression mediated through HDAC-dependent pathways (35,36). As we have shown previously, FOG-2 is an effective repressor of GATA4-mediated transactivation of the ANF promoter in cultured fibroblasts (Fig. 1, first and second  bars). The addition of TSA from 50 to 200 nM had no effect on the ability of FOG-2 to repress GATA4 transactivation of the ANF promoter (Fig. 1, third through fifth bars). These results support the notion that FOG-2 represses transcription though an HDAC-independent pathway in murine fibroblasts.
Conservation of the N-terminal Amino Acids of FOG Proteins-Our previous work had demonstrated that both murine FOG-1 and FOG-2 were capable of repressing GATA4 transac-tivation of the ANF promoter and that this repression required the N termini of each protein (28). Furthermore, fusion protein analysis between the DNA binding domain of GAL4 and the N terminus of FOG-2 demonstrated that the N-terminal 45 amino acids were sufficient to mediate transcriptional repression in a heterologous system (28). To further refine the repression domain within the first 45 amino acids of FOG proteins, we reasoned that the amino acids within this domain critical for repression might be conserved in sequence across species. To identify such residues, we first searched the NCBI genomic and EST data base for the sequences of FOG proteins from other species. We identified genomic or cDNA sequence encoding the N-terminal amino acids of FOG-2 from both human and cow sequences. For FOG-1, we identified sequences from human, cow, chicken, frog, and zebra finch. Fig. 2 shows the alignment of amino acid sequences obtained from conceptual translation of these genomic and cDNA fragments. As can be seen, 14 of the first 16 residues are identical in all of these proteins, suggesting that the amino acids critical for repression lie in the first 16 amino acids of the FOG proteins.
Fine Mapping of the Critical Residues in the N Terminus of FOG-2-To test the importance of the conserved region of the N terminus of FOG-2 for transcriptional repression, we generated a mammalian expression construct encoding a murine FOG-2 protein that lacked amino acids 2-18 and then tested the ability of this protein to repress GATA4-dependent transactivation in a transient transfection assay as described previously (18,28). As can be seen in Fig. 3B, the elimination of amino acids 2-18 completely abrogated the ability of FOG-2 to repress transcription (compare the second and third bars, p Ͻ 0.001). This result demonstrates that the first 18 amino acids are required for FOG-2 mediated repression.
To further characterize the N-terminal repression domain, we generated expression constructs encoding alanine substitutions in amino acids 2-6, 7-12, or 13-18 of murine FOG-2. These proteins were also tested for their ability to repress GATA4 transactivation as shown in Fig. 3B. Alanine substitution of amino acids 2-6 or 7-12 resulted in complete loss of repression by FOG-2 (Fig. 3B, fourth and fifth bars), whereas substitution of amino acids 13-18 had no statistically significant effect in FOG-2 mediated repression (Fig. 3B, second and  sixth bars, p ϭ 0.41). These results further narrowed the critical region to the first 12 amino acids. To test each residue individually, we generated expression constructs encoding single alanine substitutions in residues 2-12 (Fig. 4). Substitution of amino acid 3 (R3A), 4 (R4A), or 5 (K5A) resulted in a complete loss of repression by FOG-2, whereas substitution of amino acids 2, 7, 8, 10, or 11 had no significant effect on repressive activity (p Ͼ 0.05). Mutation of amino acids 6, 9, and 12 resulted in a partial loss of repression that was statistically significant (p ϭ 0.0002, 0.0001, and 0.0001, respectively). Given the potency of FOG-2 as a transcriptional repressor, very little FOG-2 expression was necessary to achieve complete repression. As a consequence, expression of these constructs could not be reliably detected by Western analysis of transfected cell lysate. However, all constructs were shown to program expression of FOG-2 in 3T3 fibroblasts when transfected at high levels (data not shown). These results suggest that amino acids 3-6, 9, and 12 are required for FOG-2 mediated repression. Importantly, these amino acids are all completely conserved in FOG proteins across species as shown in Fig. 2.
The N-terminal 12 Amino Acids Are Sufficient to Mediate Transcriptional Repression-To determine whether the critical repression motif of FOG proteins identified above is sufficient to mediate transcriptional repression, we generated mammalian expression vectors encoding the first 132, 18, or 12 amino acids of FOG-2 fused to the DNA binding domain of the yeast transcription factor GAL4 (Fig. 5A). These constructs were transiently transfected into NIH 3T3 fibroblasts along with a reporter plasmid containing the luciferase gene under the control of a modified SV40 promoter with four copies of the GAL4 DNA binding elements. Consistent with our past observations, when the first 132 amino acids of FOG-2 were fused to GAL4, the resulting fusion protein was an effective transcriptional repressor, inhibiting SV40 promoter activity by 78 Ϯ 9% (Fig.  5B, first and second bars). Similar results were obtained using just the first 18 amino acids of FOG-2 (Fig. 5B, third bar) or the first 12 amino acids (Fig. 5, fourth bar). Furthermore, a fusion protein containing the first 12 amino acids of FOG-2 with an alanine substitution at position 5 (1-12 K5A) is unable to significantly repress the SV40 promoter (Fig. 5B, fifth bar) despite similar levels of protein expression to the wild-type construct (data not shown). Taken together, these results define the critical repression motif in the N terminus of FOG proteins as RRKQxxPxxI.
Identification of the FOG Repression Motif in Other Transcriptional Repressors-With the critical residues of the FOG repression motif defined, we searched the NCBI data base for any other proteins that might contain this motif. The results are shown in Fig. 6. We identified 14 proteins from multiple vertebrate species that contain this motif. Of note, this motif was not found in proteins from the Drosophila or yeast genomes. Interestingly, all of the proteins identified in this search contain zinc finger domains, and many have been previously characterized as transcriptional repressors (37)(38)(39). In addition, the FOG repression motif in all of these proteins is at the very N terminus of each protein. Finally, all of the critical residues identified in Fig. 5 with the exception of the isoleucine at residue 12 were conserved in all the proteins. At position 12, several of the identified proteins had a conservative substitution of leucine for isoleucine or other hydrophobic amino acids such as valine or phenylalanine. Taken together, these results demonstrate that the FOG repression motif is found in other zinc finger-containing transcriptional repressors and, thus, define a novel vertebrate superfamily of zinc finger transcriptional repressors that would include members of both the FOG and Spalt families. DISCUSSION In this report, we have further characterized the N-terminal repression domain of FOG-2 and have uncovered a novel repression motif found in both FOG-1 and FOG-2 that is conserved across vertebrate species. In addition, we have identified this motif in other transcriptional repressors, suggesting that this motif may be utilized as a general mechanism of transcriptional repression. This pathway appears to work through an HDAC-independent mechanism, as tricostatin A, a general HDAC inhibitor, was unable to block repression mediated by FOG-2. It is still possible that the FOG repression motif may function in chromatin remodeling by the blocking of histone acetyltransferases or the activation of histone methyltransferases. This motif may also function either by interacting directly with the basal transcriptional machinery or RNA polymerase itself or recruiting a currently unidentified co-repressor to the promoter complex. Experiments are currently under way to address this question.
The N terminus of FOG-1 has been previously shown to be required for megakaryocyte development (40). Using an in vitro hematopoietic differentiation assay with FOG-1-deficient mouse embryonic stem cells, Cantor et al. (40) demonstrated that the N-terminal 144 amino acids of FOG-1 is not required for rescue of erythropoiesis in the differentiation of FOG-deficient ES cells but is critical for megakaryocyte production from these cells. This result in combination with the work presented in this report suggests that the FOG repression motif may play a role in the regulation of megakaryocyte differentiation. This observation also suggests that the FOG repression motif may not be utilized in all developmental processes (e.g. erythropoiesis) for which FOG proteins are required.
In Drosophila there is only one FOG family protein, Ushaped (14,(41)(42)(43). This protein has been demonstrated to act as a genetic repressor of the GATA factor pannier and be critical for normal heart development in the fly. As in other FOG proteins, U-shaped contains a consensus binding site for CtBP and can physically interact with CtBP in vitro. However, U-shaped does not have the FOG repression motif identified in this report. Consistent with this observation, transient transfection assays suggest that U-shaped only weakly represses GATA4-mediated transactivation in NIH 3T3 fibroblasts. 2 Moreover, we were unable to identify any protein in the Drosophila genome that contained this motif. In contrast, this motif was detected in proteins from frog, chicken, mouse, rat, cow, and human proteomes. Taken together, these observations suggest that the mechanism of repression utilized by the FOG repression motif may be restricted to vertebrates.
Many of the proteins that contain the FOG repression motif (Fig. 6) have been previously identified to be transcriptional repressors. For example, zinc finger protein 423 (early B-cell zinc finger transcription factor or Roaz) is a multi-zinc finger protein that binds DNA directly and represses the activation of olfactory-specific genes by O/E-1 (39). BCL11b (also known as chicken ovalbumin upstream promoter transcription factor IIinteracting protein 1) is also a zinc finger-containing protein that functions as a transcriptional repressor (37). It has been shown to contain two repression domains, one of which is localized to the N-terminal 171 amino acids. The FOG repression motif is encoded by the N-terminal 12 amino acids in this protein, so it is possible that this motif is responsible for the repression activity described for this domain. Consistent with our observations with FOG-2, the repression mediated by the N terminus of BCL11b is TSA insensitive.
Finally, the transcriptional repressor SALL1 is another example of a zinc finger transcriptional repressor that contains the FOG repression motif. SALL1 is critical for the development of the kidney, and mutations in SALL1 cause Townes-Brocks syndrome in humans (44 -46). Previous work on this protein demonstrated that the repression domain was localized to the N-terminal 130 amino acids (46). As with BCL11b, the FOG repression motif of SALL1 is localized to the very N terminus of the protein. In contrast to FOG-2 and BCL11b, the N-terminal 130 amino acids of SALL1 were shown to interact with HDAC1 and 2, but the repression mediated by this domain was only partially blocked by the HDAC inhibitor TSA (38,46). Therefore, it was postulated that this domain might repress transcription by two distinct mechanisms, one that is HDAC-dependent and the other by an HDAC-independent mechanism. Given the finding of a FOG repression motif within this domain, the HDAC-independent mechanism likely functions through this motif. As the genomes of other species are sequenced and characterized, it is possible that other proteins containing the FOG repression motif will be identified.