Basic Krüppel‐like factor (BKLF) is a zinc finger protein that recognizes CACCC elements in DNA. It is expressed highly in erythroid tissues, the brain and other selected cell types. We have studied the activity of BKLF and found that it is capable of repressing transcription, and have mapped its repression domain to the N‐terminus. We carried out a two‐hybrid screen against BKLF and isolated a novel clone encoding murine C‐terminal‐binding protein 2 (mCtBP2). mCtBP2 is related to human CtBP, a cellular protein which binds to a Pro‐X‐Asp‐Leu‐Ser motif in the C‐terminus of the adenoviral oncoprotein, E1a. We show that mCtBP2 recognizes a related motif in the minimal repression domain of BKLF, and the integrity of this motif is required for repression activity. Moreover, when tethered to a promoter by a heterologous DNA‐binding domain, mCtBP2 functions as a potent repressor. Finally, we demonstrate that mCtBP2 also interacts with the mammalian transcripition factors Evi‐1, AREB6, ZEB and FOG. These results establish a new member of the CtBP family, mCtBP2, as a mammalian co‐repressor targeting diverse transcriptional regulators.
Basic Krüppel‐like factor (BKLF) (Crossley et al., 1996) is a zinc finger protein that belongs to the subfamily of Krüppel‐like proteins, which includes erythroid Krüppel‐like factor (EKLF) (Miller and Bieker, 1993), lung Krüppel‐like factor (LKLF) (Anderson et al., 1995) and gut‐enriched Krüppel‐like factor (Shields et al., 1996) [also known as endothelial zinc finger protein, EZF (Garrett‐Sinha et al., 1996; Yet et al., 1998)]. Members of this subfamily contain three characteristic Cys–Cys: His–His Krüppel‐like zinc fingers and recognize CACCC motifs in the promoters and enhancers of various genes. The founding member of the family, EKLF, is expressed in erythroid and mast cells and specifically recognizes the CACCC‐box in the β‐globin promoter (Miller and Bieker, 1993). Naturally occurring mutations which inhibit the binding of EKLF to this element are associated with clinical β‐thalassaemia (Feng et al., 1994), and knockout studies in mice have demonstrated that EKLF is required for transcription of the β‐globin gene (Nuez et al., 1995; Perkins et al., 1995).
Specific target genes regulated by other members of the family have not been as clearly defined. LKLF was identified originally as a CACCC‐box protein that was expressed abundantly in the lung, as well as in a number of other tissues (Anderson et al., 1995). Targeted mutation of the LKLF gene in mice has revealed important roles for this protein in T cell activation (Kuo et al., 1997b) and in vascular endothelial cells (Kuo et al., 1997a). GKLF/EZF is also expressed in vascular endothelial cells (Yet et al., 1998), as well as in the gut (Shields et al., 1996) and in the epidermal layer of skin (Garrett‐Sinha et al., 1996). It is believed to be instrumental in controlling the proliferation of epithelial tissues, and there is some evidence that it can act as either an activator or a repressor of transcription (Yet et al., 1998).
Like EKLF, BKLF is expressed at high levels in erythroid cells, but it is found additionally in other tissues, most notably the brain (Crossley et al., 1996). In vitro binding studies have demonstrated that like EKLF, it binds to the CACCC motifs found in the β‐globin promoter as well as to CACCC‐boxes in the globin locus control regions, and several of the heme biosynthetic genes. The exact target genes regulated by BKLF, however, have not been defined. BKLF knockout mice have been generated, and initial analysis indicates that they suffer from a myeloproliferative disorder, and thus it seems likely that BKLF plays a role in hematopoiesis (Perkins et al., 1997). It has been demonstrated previously that BKLF can activate transcription from a minimal promoter containing a single BKLF‐binding site, although activation was significantly weaker than achieved by other Krüppel family proteins and was only observed with very high levels of BKLF (Crossley et al., 1996). In this study, we have extended our analysis of the functional activity of BKLF and demonstrate that it can also act as a potent repressor. We show that the minimal repression domain maps to a 74 amino acid region within the N‐terminus, and that this domain retains its activity when fused to a heterologous DNA‐binding domain. In an effort to understand the mechanism by which BKLF represses transcription, we carried out a yeast two‐hybrid screen against the repression domain of BKLF, and identified a novel cofactor protein which we refer to as murine C‐terminal‐binding protein 2 (mCtBP2).
mCtBP2 is a new member of the CtBP family of related proteins. The first member of the family, human CtBP (now hCtBP1), was identified initially as a cellular protein that bound to the C‐terminal region of the adenovirus E1a oncoprotein (Boyd et al., 1993; Schaeper et al., 1995). It was shown that deletion of this region dramatically enhanced the tumorigenecity of E1a, suggesting that the binding of hCtBP1 significantly modulated the effect of E1a in vivo. Subsequently, it was demonstrated that an equivalent deletion to a gal4–E1a fusion protein dramatically enhanced its ability to activate transcription (Sollerbrant et al., 1996). This result suggested that hCtBP1 might be involved in the negative regulation of transcription. Recently, a Drosophila homologue of human CtBP1 has been identified as a partner protein of the transcriptional repressors Hairy, Snail, Knirps and Enhancer of split [E(spl)] mδ (Nibu et al., 1998; Poortinga et al., 1998). It has also been suggested that dCtBP binds the repressor Krüppel (Nibu et al., 1998). The in vivo studies in Drosophila indicate that dCtBP is a genuine co‐repressor protein that plays an important role in repressing gene expression during development.
Human CtBP1 recognizes signature Pro‐X‐Asp‐Leu‐Ser motifs in the C‐terminus of E1a (Boyd et al., 1993; Schaeper et al., 1995; Sollerbrant et al., 1996). Drosophila CtBP has been shown to bind similar motifs found within the C‐terminus of Hairy and E(spl)mδ (Poortinga et al., 1998), and within the body of the Snail and Knirps proteins (Nibu et al., 1998). The mammalian BKLF protein contains a similar motif within its repression domain. We show that mCtBP2 associates with BKLF through this element and that it is instrumental in mediating the repressor activity of BKLF.
We searched the sequences of previously reported mammalian repressor proteins and found that potential CtBP recognition motifs occurred in several of these proteins, most notably in the zinc finger proteins Evi‐1 (Morishita et al., 1988), AREB6 (Watanabe et al., 1993) and ZEB (Genetta et al., 1994). The signature motif also occurs in Friend of GATA (FOG) (Tsang et al., 1997), and a related Drosophila repressor protein, U‐shaped (Haenlin et al., 1997). We used the two‐hybrid system to demonstrate that mCtBP2 is able to associate specifically with the mammalian proteins AREB6, ZEB, Evi‐1 and FOG. Our results indicate that mCtBP2 is a co‐repressor protein that can associate with a number of mammalian transcription factors.
BKLF can act as a potent repressor of transcription
In order to assess the function of BKLF, we carried out co‐transfection assays in Drosophila Schneider line 2 (SL2) cells. These cells are used conventionally for the study of CACCC‐box factors since they are devoid of ubiquitous CACCC proteins (such as Sp1) which complicate interpretations of experiments in mammalian cell lines. It previously has been reported that CACCC‐box factors can synergize with the glucocorticoid receptor (GR) on appropriately configured composite CACCC‐glucorticoid response elements (GREs) (Schüle et al., 1988; Strähle et al., 1988). With a view to assessing whether BKLF would co‐operate with or quench the activity of GRs, we prepared a test promoter containing three copies of a composite CACCC‐GRE site [which had a spacing that previously had been reported to give maximal synergy (Strähle et al., 1988)] (Figure 1). First we tested the activity of BKLF alone, and in the absence of the glucocorticoid dexamethasone, and found it reduced the expression of this reporter gene slightly (Figure 1A, columns 1–4). In contrast, EKLF potently activated transcription (columns 5–7). This result suggested that BKLF was capable of repressing transcription. We next co‐expressed increasing amounts of BKLF together with a constant amount of EKLF and found that BKLF could silence the activation mediated by EKLF (Figure 1B). We noted that this inhibition occurred with relatively low levels of BKLF (gel retardation data not shown), suggesting that the effect was not due entirely to competition between BKLF and EKLF for the three CACCC‐boxes and that it might be due to an active repression domain within BKLF. We therefore tested a deletion derivative of BKLF, in which the entire N‐terminus of the protein had been removed, leaving only the zinc finger DNA‐binding domain (Figure 1C). This truncated protein was able to compete with and inhibit EKLF‐mediated activation to some extent, but was significantly less effective than full‐length BKLF. These results raised the possibility that BKLF contained a repression domain, capable of repressing activated transcription.
We then proceeded to assess whether BKLF could repress activation mediated by the GR in the presence of dexamethasone. We found that BKLF potently repressed GR‐activated transcription (Figure 1D, columns 1–4). In contrast, EKLF (and Sp1) (Figure 1D, columns 5–7, and data not shown) efficiently synergized with the GR on our promoter. We also tested the deletion derivative of BKLF, that only contains the zinc fingers, and found that it synergized weakly with the GR (Figure 1E). We conclude that BKLF contains a repression domain in its N‐terminus that can silence both EKLF‐ and GR‐activated transcription. In order to map this repression domain more precisely, we prepared a number of additional deletion constructs. We first used gel shift assays to verify that these mutants were all expressed at equivalent levels (data not shown), and then tested their ability to silence GR‐activated transcription (Figure 1E). Deletion of the first 74 amino acids dramatically reduced repression activity, suggesting that the relevant domain included this region.
The N‐terminal domain of BKLF interacts with the co‐repressor mCtBP2
In order to identify possible cofactors which might bind to this domain and mediate BKLF's ability to repress transcription, we carried out a yeast two‐hybrid screen using a murine erythroleukaemia (MEL) cell library (Tsang et al., 1997) and a gal4 DNA‐binding domain (DBD)–BKLF(1–268) fusion bait protein. This bait protein contains the repression domain of BKLF and adjoining sequences, but does not contain the BKLF zinc fingers. The analysis of ∼1.2×106 colonies led to the isolation of 110 that were His+, and 35 that were both His+ and β‐galactosidase positive. Sequencing of these isolates revealed a full‐length clone that encodes a member of the CtBP family of proteins. This protein is most highly related to a sequence generated through the assembly of database expressed sequence tags (ESTs) termed mCtBP2 (Katsanis and Fisher, 1998); however, it differs at four amino acids, and contains an additional 25 amino acids at the N‐terminus. The sequences of recognized CtBP family proteins are shown in Figure 2.
Finally we showed that as well as interacting with gal4DBD–BKLF(1–268), mCtBP2 interacted with gal4DBD–full‐length BKLF (Figure 3), but did not interact with three control baits, p53(72–390) (Iwabuchi et al., 1993), the N‐finger of GATA‐1 (Tsang et al., 1997) or human lamin C(66–230) (Bartel et al., 1993). We also exchanged the bait and prey, cloning BKLF(1–268) in‐frame with the gal4 activation domain (AD) and mCtBP2 in‐frame with the gal4DBD. A strong interaction was also observed in this experiment (Figure 3).
The repression domain and mCtBP2‐binding domain of BKLF co‐localize
We next used the two‐hybrid assay to determine whether mCtBP2 bound to the region of BKLF previously implicated in transcriptional repression. We constructed a panel of gal4DBD–BKLF deletion constructs and used the two‐hybrid system to test which of these interacted with mCtBP2 (Figure 3). Significantly, deletion of the N‐terminal repression domain abolished the interaction with mCtBP2, but not with other BKLF‐interacting proteins isolated from the library screen (data not shown). This result suggests that mCtBP2 binds within the BKLF repression domain.
Within this N‐terminal repression domain of BKLF, we noted the motif Pro‐Val‐Asp‐Leu‐Thr, which is closely related to a recognized hCtBP1‐binding site in E1a (Pro‐Val‐Asp‐Leu‐Ser) (Schaeper et al., 1995). It previously has been reported that mutation of the core Asp–Leu residues to Ala–Ser abolished the interaction between E1a and hCtBP1 (Schaeper et al., 1995). We therefore used the two‐hybrid system to test whether the analogous mutation in either the original BKLF(1–268) bait or in the full‐length BKLF bait affected their interaction with mCtBP2. In both cases, this mutation significantly impaired the interaction (Figure 3), suggesting that mCtBP2 directly targets this motif in BKLF. Furthermore, exchange of the bait and prey (as above) also demonstrated that this mutation impaired the interaction between BKLF and mCtBP2 (Figure 3).
mCtBP2 binds BKLF in vitro
We next sought to determine whether the interaction between mCtBP2 and BKLF could be detected in vitro. We prepared GST–BKLF fusion proteins immobilized on agarose beads and tested their ability to retain in vitro translated 35S‐radiolabelled mCtBP2. GST–full‐length BKLF and GST–BKLF(1–268) fusions were able to bind mCtBP2 efficiently, whereas proteins containing the core Ala–Ser mutation, the zinc fingers of BKLF alone or GST alone could not retain mCtBP2 (Figure 4A). Thus, the interaction of mCtBP2 with BKLF in vitro requires the N‐terminal repression domain, and is dependent on the integrity of the Pro‐Val‐Asp‐Leu‐Thr motif.
BKLF, mCtBP2 and CACCC‐box DNA form a ternary complex
We also investigated whether mCtBP2 could associate with full‐length BKLF bound to a CACCC sequence in DNA. We expressed full‐length BKLF, or full‐length BKLF containing the core Ala–Ser mutation, in SL2 cells and used the nuclear extracts in gel mobility shift experiments with the β‐globin CACCC site as a probe (Figure 4B). Retarded complexes corresponding to BKLF and mutant BKLF can readily be observed (Figure 4B, lanes 2 and 7). However, when increasing amounts of mCtBP2 are co‐expressed, the intensity of the BKLF–DNA complex diminishes, whereas the intensity of the mutant BKLF–DNA complex is unaffected (Figure 4B, compare lanes 2–6 and 7–11). The reduction in the observable BKLF–DNA complex suggests that either mCtBP2 is interfering with BKLF's ability to bind DNA, or that a higher molecular weight complex is formed but is obscured by another band. Additional experiments revealed the presence of a high molecular weight complex which co‐migrates with a fainter endogenous band (Figure 4B, lanes 13 and 15). This complex is only observed when intact BKLF is co‐expressed with mCtBP2, and addition of anti‐BKLF antisera leads to the disruption of this complex and the formation of a new supershifted complex (Figure 4B, lanes 14 and 16). This latter complex is of only slightly lower mobility than the BKLF–mCtBP2–DNA complex, and co‐migrates with the BKLF–Ab–DNA complex (data not shown), suggesting that the antibody displaces mCtBP2. These results indicate that BKLF and mCtBP2 can form a complex on a CACCC‐box oligonucleotide.
mCtBP2 is a co‐repressor
To determine whether BKLF was repressing transcription primarily by recruiting mCtBP2 (or a related family member) to the promoter, we tested whether the core Ala–Ser mutation [which reduces the binding of mCtBP2 to BKLF (Figures 3 and 4)] would impair BKLF's ability to repress transcription in the SL2 cell assay (Figure 5A). In this system, full‐length BKLF repressed GR‐mediated activation in a dose‐dependent manner, to >35‐fold (Figure 5A, columns 2–4). In contrast, the mutant version of BKLF (which contains the core Ala–Ser mutation) caused only a 5‐fold repression, which was not dose dependent (Figure 5A, columns 5–7). This result strongly suggests that BKLF is repressing transcription by recruiting CtBP to the reporter gene promoter. In this case, we conclude that BKLF is associating directly with endogenous dCtBP.
We carried out an additional experiment to demonstrate that mCtBP2 itself was capable of directly silencing gene expression. We constructed a chimaeric gene in which the coding sequence of mCtBP2 was fused directly to the zinc finger region of BKLF (Figure 5B, columns 1–6). In this case, the entire BKLF repression domain is removed and replaced by mCtBP2. mCtBP2 can therefore be targeted to the CACCC‐box promoter, not by piggy‐backing onto BKLF, but directly by means of the linked zinc finger domain. We tested the ability of this mCtBP2‐finger chimaeric protein to repress GR‐mediated transcriptional activation in SL2 cells. As shown in Figure 5B (columns 1–6), it repressed transcription efficiently in a dose‐dependent manner. This result indicates that mCtBP2 is capable of repressing transcription, at least in Drosophila SL2 cells.
Additional experiments in mammalian cells provide further evidence that BKLF and mCtBP2 associate to mediate repression. We constructed a gal4DBD–BKLF(1–268) fusion protein and tested its ability to repress a gal4‐dependent promoter driving growth hormone expression in NIH 3T3 cells (Figure 5C). We chose these cells since they express both endogenous BKLF and mCtBP2 (unpublished results). As shown in Figure 5C (columns 1–4), this fusion protein efficiently silenced gene expression in a dose‐dependent manner. This result is consistent with our previous conclusion that the N‐terminus of BKLF contains a repression domain. In order to determine whether the abrogation of mCtBP2 binding influenced BKLF's ability to repress transcription, we tested whether a mutant gal4DBD–BKLF(1–268) fusion protein, containing the core Ala–Ser substitution, could also mediate repression (Figure 5C, columns 5–7). In contrast to the result in SL2 cells, in NIH 3T3 cells this mutation had no discernible effect on BKLF's repression activity.
Reasoning that BKLF might possess an additional repression domain that functions in NIH 3T3 cells and compensates for the loss of CtBP‐mediated repression, we constructed a minimal gal4DBD–BKLF(1–75) fusion protein that contained only the previously defined minimal repression domain of BKLF (see Figure 1). This protein also repressed reporter gene expression efficiently in a dose‐dependent manner (Figure 5D, columns 1–4). In the case of the minimal construct, however, we found that mutation of the core Asp–Leu residues, within the mCtBP2‐binding site, severely impaired its ability to repress transcription (Figure 5D, columns 5–7). This result suggests that the gal4DBD–BKLF(1–75) fusion represses transcription by recruiting mCtBP2 (or another member of the CtBP family expressed in NIH 3T3 cells).
Finally, in order to determine whether mCtBP2 itself was capable of repressing transcription in mammalian cells, we constructed a gal4DBD–mCtBP2 fusion and tested its activity against the gal4 site‐dependent promoter. As shown in Figure 5E (columns 1–6), the gal4DBD–mCtBP2 chimaeric protein efficiently repressed the expression of the reporter gene in a dose‐dependent manner.
Taken together, these results demonstrate that mCtBP2 is a true co‐repressor in that it is capable of repressing transcription when delivered to a target promoter, either by binding to BKLF or when provided with its own BKLF zinc finger or gal4DBD. While mutant versions of BKLF, unable to bind CtBP, are unable to repress transcription in SL2 cells, it is interesting that the long form of the mutant gal4DBD–BKLF fusion could still repress gene expression in NIH 3T3 cells. This result suggests that additional cofactors may exist in mammalian cells that compensate for the loss of direct CtBP binding. Thus, our results indicate that while BKLF and CtBP co‐operate to repress transcription, additional proteins are also likely to be involved (see Discussion).
The putative dehydrogenase activity of mCtBP2 is not required for transcriptional repression
It has been noted that CtBP family proteins have significant homology to the family of d‐isomer‐specific 2‐hydroxy acid dehydrogenases (Schaeper et al., 1995), and it has been suggested that dCtBP may repress transcription by means of this dehydrogenase activity (Nibu et al., 1998). Attempts to demonstrate dehydrogenase activity of hCtBP1, however, have been unsuccessful, and an alternative hypothesis that the homology indicates structural similarity only and may reflect a conserved dimerization domain has been proposed (Schaeper et al., 1995; Poortinga et al., 1998). The dehydrogenases most similar to CtBP family proteins function as homodimers in vivo (Goldberg et al., 1994), and it has also been shown that dCtBP can dimerize (Poortinga et al., 1998). We tested mCtBP2's ability to dimerize in both the yeast two‐hybrid assay and in in vitro GST pull‐down experiments, and found that mCtBP2 was also able to homodimerize efficiently (data not shown). Thus it is clear that like the dehydrogenase enzymes, CtBP family proteins can dimerize, and it is possible that the similarity to dehydrogenases arises primarily from conservation of the dimerization surfaces.
Nevertheless, it is noteworthy that all reported CtBP family members contain a conserved histidine residue that appears to correspond to the active site histidine in the enzyme d‐lactate dehydrogenase (H296 in d‐lactate dehydrogenase, H321 in mCtBP2 and H314 in hCtBP1) (Taguchi and Ohta, 1993; Schaeper et al., 1995) (Figure 2). Thus it remains possible that CtBP proteins possess undetected dehydrogenase activity. It has been shown that mutation of H296 in d‐lactate dehydrogenase severely impaired its enzymatic activity (Taguchi and Ohta, 1993). We therefore constructed a mutation of H321 in mCtBP2 and carried out experiments to determine whether this mutation influenced the ability of mCtBP2 to repress gene expression. We tested both mutant mCtBP2–BKLF zinc finger and gal4DBD–mCtBP2 fusion proteins, each containing an H321 to A substitution (Figure 5B and E respectively). The substitution had no significant effect on the repression activity of either the mCtBP2–BKLF finger chimaera in SL2 cells (Figure 5B, columns 7–11) or the gal4DBD–mCtBP2 fusion in NIH 3T3 cells (Figure 5E, columns 7–11). These results suggest that even if mCtBP2 does possess undetected dehydrogenase activity, this activity is not essential for its role in the silencing of gene expression.
mCtBP2 binds a number of mammalian transcription factors
Noting that CtBP family members recognize the consensus motif Pro‐X‐Asp/Asn‐Leu‐Ser/Thr, we carried out database searches to determine whether this motif was found in other mammalian repressors. As shown in Table I, this motif (or closely related sequences) occurs in a number of proteins and most notably within the previously defined small repression domains of the zinc finger proteins Evi‐1 and AREB6 (Bartholomew et al., 1997; Ikeda et al., 1998) (Figure 6). The motif also occurs in FOG (Tsang et al., 1997) and in the AREB6‐related protein ZEB (Genetta et al., 1994). Evi‐1 is an oncogene, whose expression is associated with myeloid leukaemias in humans and mice (Nucifora, 1997). Although Evi‐1 appears to be capable of activating transcription, there is now good evidence that it can also act as a transcriptional repressor (Bartholomew et al., 1997). AREB6 is a large zinc finger homeodomain protein, which has also been implicated in both the repression and activation of transcription (Watanabe et al., 1993).
We used the yeast two‐hybrid system to test whether a gal4DBD–mCtBP2 bait could interact with gal4AD hybrids containing portions of these mammalian transcription factors (Figure 6). We first tested the repression domain of Evi‐1 (amino acids 514–726; Bartholomew et al., 1997). This region contains the sequences Pro‐Phe‐Asp‐Leu‐Thr and Pro‐Leu‐Asp‐Leu‐Ser, and was strongly positive for interaction with mCtBP2. We also tested a 32 amino acid region within the repression domain of AREB6, encompassing a Pro‐Leu‐Asn‐Leu‐Thr motif. This domain also interacted efficiently with mCtBP2. We then studied a slightly larger domain from the protein ZEB (amino acids 697–795). ZEB is highly related to AREB6 but differs in a small region between amino acids 784 and 802, and therefore lacks the Pro‐Leu‐Asn‐Leu‐Thr motif of AREB6. However, N‐terminal to this region two additional Pro‐X‐Asp‐Leu‐Thr‐like motifs occur (Table I). We found that mCtBP2 interacted strongly with this subdomain of ZEB. Finally, mCtBP2 was found to bind to a region within FOG (amino acids 563–859) which contains the motif Pro‐Ile‐Asp‐Leu‐Ser. Our results therefore suggest that mCtBP2 can interact with recognition motifs found in several different mammalian transcription factors.
Sequence‐specific DNA‐binding proteins play a primary role in the regulation of gene expression. The DNA recognition domains of these proteins tend to form distinct structures, such as zinc fingers or homeodomains, and are usually easily recognizable. The activation and repression domains of transcription factors have been much more difficult to define. Activation domains have been classified in general terms as acidic, glutamine or proline rich (Mitchell and Tjian, 1989), while repression domains are often basic, alanine or proline rich (Hanna‐Rose and Hansen, 1996). It now appears that many of these domains function by recruiting accessory proteins that mediate the activation or repression of transcription, and that in some cases the crucial protein–protein interactions involve only very small sequence motifs within the larger functional domains (Fisher et al., 1996; Heery et al., 1997). For example, it recently has been reported that the interaction between certain steroid receptor proteins and their co‐activators occurs through short Leu‐X‐X‐Leu‐Leu motifs in the co‐activators (Heery et al., 1997). We have shown here that Pro‐X‐Asp/Asn‐Leu‐Ser/Thr motifs within the repression domains of mammalian DNA‐binding proteins mediate their interaction with the co‐repressor mCtBP2, while others have demonstrated the necessity of this motif for interaction of dCtBP with the repression domains of Drosophila transcription factors (Nibu et al., 1998; Poortinga et al., 1998).
Recent evidence indicates that Drosophila CtBP interacts with the helix–loop–helix repressors Hairy and E(spl)mδ (Poortinga et al., 1998), the zinc finger protein Snail and the nuclear receptor protein Knirps (Nibu et al., 1998). Here we show that the related murine protein, mCtBP2, interacts with the zinc finger transcription factors BKLF, AREB6, ZEB, Evi‐1 and FOG. As we have demonstrated, BKLF can act as a potent repressor of transcription, though it has also been shown to activate transcription, albeit weakly (Crossley et al., 1996). It has been proposed that BKLF may be involved in the activation of the SHP‐1 phosphatase gene (Perkins et al., 1997), and it is possible that BKLF acts as either an activator or repressor of transcription, depending on promoter or cellular contexts. Whilst ZEB is well recognized as a transcriptional repressor, AREB6 and Evi‐1 have been reported to be involved in either the repression or activation of transcription, depending on contexts (Watanabe et al., 1993; Genetta et al., 1994; Morishita et al., 1995; Tanaka et al., 1995; Bartholomew et al., 1997; Ikeda et al., 1998). Likewise, the GATA‐1‐binding protein FOG appears to be able to activate or repress transcription depending on context (Tsang et al., 1997, and unpublished data). Interestingly, a CtBP recognition motif also occurs in the Drosophila repressor protein U‐shaped which is similar to FOG (Figure 6) in that it also contains nine zinc fingers and specifically binds to a GATA factor, the Drosophila factor Pannier (Haenlin et al., 1997). It will be interesting to determine whether the motif in U‐shaped is required for the protein's repression activity during Drosophila development. The motif also occurs in the apparent Drosophila counterpart of AREB6, Zfh1, a transcription factor required for germ cell migration and gonadal mesoderm development (Broihier et al., 1998). It has also been suggested that dCtBP may bind the founding member of the Krüppel family, the Drosophila Krüppel protein (Nibu et al., 1998). We have demonstrated that mCtBP2 interacts with a small region of Krüppel (amino acids 356–502), encompassing the previously defined C‐terminal repression domain (Hanna‐Rose and Hansen, 1996) (Figure 6) and two potential CtBP recognition motifs (Table I) (unpublished results).
We have presented evidence that mCtBP2 is involved directly in the repression activity of BKLF. mCtBP2 binds to the minimal repression domain of BKLF, and mutations which abrogate its binding significantly impair BKLF's ability to repress transcription. Our results also suggest that CtBP proteins play a direct role in mediating repression by Evi‐1 and AREB6. In the case of Evi‐1, it is significant that two CtBP‐binding motifs map within the repression domain of this protein. In the case of AREB6, there is even stronger evidence that CtBP is involved. We have demonstrated that mCtBP2 binds a region of AREB6 that contains a single CtBP recognition motif (Pro‐Leu‐Asn‐Leu‐Thr). Moreover, it has been reported previously that an Asn to Gln substitution within this motif abolished repression, though the explanation for this loss of activity was not elucidated (Ikeda et al., 1998). Our results strongly suggest that CtBP is involved in facilitating repression by AREB6. This example, however, is complicated by the fact that AREB6 also contains two additional potential CtBP‐binding motifs slightly N‐terminal to the critical motif. These additional motifs are conserved in the related protein ZEB, and we have demonstrated that these motifs can also bind mCtBP2 in two‐hybrid assays. It may be that AREB6 contains three CtBP‐binding sites and, since the inactivation of one of these compromises repression, it is likely that all three are required for full repressor activity. Interestingly, Snail and Krüppel also appear to have more than one CtBP‐binding site (Nibu et al., 1998), as does Evi‐1 (Figure 6 and Table I). Thus, it seems that like the Leu‐X‐X‐Leu‐Leu motifs that are repeated in the co‐activators of nuclear receptor proteins (Heery et al., 1997), Pro‐X‐Asp‐Leu‐Ser motifs also occur in clusters within a single protein.
It is notable that several of the proteins targeted by mCtBP2 can act as either activators or repressors. It will be important to determine whether the activity of CtBP family members in particular cell lines correlates with the ability of these factors to repress transcription. Existing evidence from Northern blot analysis and EST studies suggests that hCtBP1, mCtBP2 and dCtBP mRNAs are widely expressed (Katsanis and Fisher, 1998; Poortinga et al., 1998) but, clearly, the presence of CtBP mRNA may not indicate that the protein is active. It has been reported that hCtBP1 is modified differentially by phosphorylation during the cell cycle (Boyd et al., 1993) and, moreover, repression by CtBP is likely to require additional cofactors (Nibu et al., 1998; Poortinga et al., 1998), so the availability of these proteins must also be taken into account.
Although the four mammalian proteins we have studied are all zinc finger proteins that have established roles in haematopoiesis, we believe that mammalian CtBP proteins will target a much wider group of factors. The results from Drosophila indicate that dCtBP interacts with a variety of different DNA‐binding proteins and, likewise, our analysis of databases suggests that CtBP recognition motifs occur in a number of other transcriptional repressors and, interestingly, also in co‐repressor proteins (data not shown). The observation that we and others have made, that both mCtBP2 and dCtBP (Poortinga et al., 1998) can dimerize, raises the possibility that CtBP dimers may function to bring together repressors and co‐repressors containing Pro‐X‐Asp‐Leu‐Ser motifs. CtBP may thus be an adaptor protein that facilitates the assembly of a larger repressor complex. Such a role is in keeping with the suggestion that dCtBP acts as a member of a large complex (Nibu et al., 1998; Poortinga et al., 1998), and with our result that dehydrogenase activity is not required for its ability to repress gene expression.
Although our results fit with the suggestion from work on dCtBP that CtBP acts as part of a repressor complex, the mechanism by which this complex represses gene expression is not clear. It may be significant that dCtBP is believed to target short‐range repressors (Nibu et al., 1998). These repressors are classified as proteins that quench the action of activators bound within 100 bp or directly inhibit the core transcriptional machinery. Here we show that mCtBP2 quenches the activity of GR bound to an adjacent element, but we cannot exclude the possibility that it is working primarily on TATA‐box‐binding protein (TBP) or an associated factor critical to the activity of the basal transcriptional machinery. The previously described co‐repressors NCoR, SMRT and Rb have been shown to mediate repression by recruiting histone deacetylase proteins (Nagy et al., 1997; Brehm et al., 1998; Lin et al., 1998; Magnaghi‐Jaulin et al., 1998). It may be that mCtBP2 also recruits a histone deacetylase, but other mechanisms are also possible, and currently are being investigated.
Although CtBP proteins associate with DNA‐binding proteins, it should be remembered that hCtBP1 was first isolated as a protein that interacted with the adenovirus protein E1a (Boyd et al., 1993). E1a is generally regarded as a virally encoded transcriptional cofactor, and it is interesting that as well as hCtBP1, other transcriptional regulatory proteins (such as p300, Rb, TBP and certain TBP‐associated factors) (Lee et al., 1991; Zamanian and La Thangue, 1992; Geisberg et al., 1995; Mymryk, 1996; Yang et al., 1996) are also targeted by E1a. Unlike these proteins, hCtBP1 binds the C‐terminus of E1a. This region is not required for oncogenic transformation but it has been shown to play a significant role in modulating the transforming activity of E1a (Boyd et al., 1993). Thorough in vivo studies have demonstrated that the deletion of the CtBP‐binding site significantly increases E1a's transforming and tumorigenic activities, while reducing its ability to inhibit metastases (Boyd et al., 1993). Although the molecular mechanisms underlying these activities are not known, it seems likely that by binding CtBP, E1a alters normal patterns of gene expression and in this way alters the proliferative capacity of cells. It is possible that E1a is involved in delivering CtBP to additional target genes or, alternatively, by sequestering CtBP, it may deprive other transcription factors of this protein. It will be particularly interesting to determine whether the sequestering of CtBP alters the expression of genes regulated by oncogenic transcription factors such as Evi‐1.
If mCtBP2 is proven to be exclusively a short‐range repressor, it may be useful as a tool for specifically silencing particular genes, without affecting the activity of closely linked neighbouring genes. It appears that CtBP normally is delivered to target genes by piggy‐backing onto DNA‐bound transcription factors. In our study, we fused the zinc fingers from BKLF, which recognize CACCC‐box sequences, to mCtBP2 and showed that the resulting protein could silence a CACCC‐dependent promoter. Thus we demonstrated that if CtBP is provided with its own zinc finger DNA‐binding domain it can silence gene expression directly. It is likely that zinc fingers with different specificities could also be used and, in this way, mCtBP2 could be targeted to other promoters. Indeed much progress has been made recently in engineering novel zinc fingers with a wide range of desired sequence specificities (Choo et al., 1994, 1997; Kim and Pabo, 1997, 1998; Kim et al., 1997). It thus seems possible that artificial zinc finger–mCtBP2 fusions with the required sequence specificity could be directed against various chosen target promoters. It should be remembered, however, that our experiments only demonstrated silencing of reporter plasmids in transient transfection assays. Nevertheless, there is reason to believe that the silencing of endogenous genes may be successful. Firstly, it should be noted that the aim would be to target CtBP to fully active promoters, likely to be contained in accessible chromatin. Secondly, results from Drosophila have already indicated that a gal4DBD–dCtBP fusion is able to repress the transcription of an integrated gal4‐dependent promoter in vivo (Nibu et al., 1998). In this case, however, repression was weaker than expected, suggesting the requirement for additional cofactors. We have evidence of a novel mCtBP2‐interacting protein, which also associates with BKLF (unpublished results). This latter protein binds to a distinct domain within BKLF and may account for the residual repressor activity observed in assays with the mutant BKLF proteins that are unable to bind CtBP directly (Figure 5A and C). We currently are investigating this protein, in the expectation that it is another component of a larger repressor complex. Ultimately, it may be possible to deliver the entire complex to chosen promoters, and in this way to control their activity artificially.
Materials and methods
The EKLF expression vector pPac‐EKLF (Merika and Orkin, 1995) was a gift from M.Merika. pPac‐BKLF was constructed by inserting a blunted EcoRI fragment containing the entire BKLF coding sequence (Crossley et al., 1996) into the blunted BamHI site of pPac. pPac‐GR was constructed by similarly cloning an Asp718–DraI fragment containing the entire coding region of the human GR (Hollenberg et al., 1985) into the blunted BamHI site of pPac. The gal4DBD–mCtBP2 fusion consisted of the mCtBP2 coding sequence fused downstream of the gal4DBD(1–147) all within the mammalian expression vector pcDNA3 (Invitrogen). The mCtBP2–BKLF fingers fusion consisted of mCtBP2 fused upstream of the region encoding the three BKLF zinc fingers (residues 246–344) all cloned into the BamHI site of pPac. SL2 cell reporter plasmids contained three copies of a composite CACCC‐GRE site (tgctAGAACAtccTGTACAgcagagagCCACACCCAtctg; Strähle et al., 1988) upstream of the minimal Adh promoter and the CAT gene in p1970 (Fan and Maniatis, 1989). NIH 3T3 cell reporters contained five gal4‐binding sites upstream of a minimal promoter in pØGH (Nichols Institute).
Transactivation and transrepression assays
SL2 cells and NIH 3T3 cells were cultured in Shields and Sang medium and Dulbecco's modified Eagle's medium (DMEM) respectively, both supplemented with 10% fetal calf serum, penicillin, streptamycin and glutamine. Transfections were carried out by the calcium phosphate precipitation method (Chen and Okayama, 1987; Sambrook et al., 1989) with the amounts of plasmid DNA indicated. A 0.5 μg aliquot of CAT reporter plasmid was used in all SL2 experiments, and 5 μg of growth hormone reporter was used in all NIH 3T3 cell experiments. All data shown represent the sum of several sets of experiments with duplicate preparations of plasmid DNA. Transfections were normalized by reference to co‐transfected lacZ reporter plasmids and/or gel shifts were carried out to verify expression levels of DNA‐binding proteins. In experiments using the human GR, dexamethasone was added to a final concentration of 10−8 M 20 h after the transfection. Cells or media were harvested 48 h after transfection. CAT assays and growth hormone assays were carried out using Amersham Quanti‐Cat and Nichols Inst. Allegro GH assay kits according to the manufacturers' instructions.
Two‐hybrid screening and assays
The Clontech two‐hybrid system was used according to the manufacturer's instructions. A murine erythroleukaemia cell cDNA library in the gal4AD fusion vector, pGAD10, was transfected into the yeast strain HF7c harbouring the gal4DBD–BKLF(1–268) fusion protein expressed from pGBT9. Two full‐length mCtBP2 clones that differed only in the length of their 5′‐untranslated regions were isolated. Surprisingly, intervening stop codons between the gal4AD and the mCtBP2 translational initiation codon suggested that neither clone was being expressed as an in‐frame gal4AD fusion. Interestingly, dCtBP was cloned using a similar strategy and was also present as an out‐of‐frame fusion (Nibu et al., 1998). Interactions between mCtBP2 and Evi‐1, AREB6, ZEB and FOG subdomains were carried out by co‐transfecting a gal4DBD–mCtBP2 bait plasmid together with gal4AD–prey proteins into HF7c. Transformants were selected on Trp/Leu‐deficient plates and patched onto Trp/Leu/His‐deficient media. Growth was scored after 30 h. Controls containing bait and prey alone were also conducted and were negative for growth up to >72 h.
In vitro binding assays
Full‐length mCtBP2 was cloned into the vector pcDNA3 (Invitrogen), and 35S‐labelled in vitro translated protein was generated using T7 polymerase and the TNT system from Promega. GST fusion proteins and binding assays were carried out as previously described (Merika and Orkin, 1995). Gel retardation assays were performed using a 32P‐labelled double‐stranded oligonucleotide (TAGAGCCACACCCTGGTAAG and CTTACCAGGGTGTGGCTCTA) (Crossley et al., 1996).
We are grateful to Gerd Blobel, Archa Fox, Melissa Holmes and Margot Kearns for critical reading of the manuscript, to Louise Rafty for preliminary work on this project, to T.Genetta and T.Kadesch for ZEB plasmids, and to G.Chinnadurai for communicating unpublished results. J.T. is supported by an Australian Postgraduate Award. This work was supported by a grant from the Australian NHMRC to M.C.
- Copyright © 1998 European Molecular Biology Organization