Novel Transactivation Domain in Erythroid Kruppel-like Factor (EKLF)*

Erythroid Kruppel-like Factor (EKLF) is an erythroid-specific transcription factor that plays a critical role in γ- to β-globin gene switching during development. To identify essential domains required for EKLF transactivation function, we cotransfected a human erythroleukemia cell line (K562) with a locus control region γ/Luc-β/Cat reporter and an EKLF expression vector. In this assay EKLF mediates a 500-fold induction of β/CAT expression compared with controls. To map essential transactivation domains, progressive NH2-terminal and internal deletion mutants of EKLF were constructed. All EKLF mutants were expressed at wild-type levels, localized to the nucleus, and bound DNA. When mutant EKLF proteins were tested for β/CAT activation, a novel transactivation domain was identified. This novel domain, encompassing amino acids (aa) 140–358, is sufficient for maximal β/CAT activation. An 85-amino acid subdomain within this region (aa 140–225) is essential for its activity. Interestingly, this central transactivation subdomain is functionally redundant with the amino-terminal domain (aa 1–139). Thus, EKLF possesses at least two potent transactivation domains that appear to function in a redundant manner.

The genes encoding ␤-like subunits of human hemoglobin are expressed in a tissue-and developmental stage-specific pattern of expression (1). Expression is exclusive to erythroid tissues, and ⑀-, ␥-, and ␤-globin gene expression is predominately restricted to the yolk sac, fetal liver, and bone marrow, respectively. The genes are present within a 100-kb 1 locus that contains a potent 22-kb enhancer termed the locus control region (LCR) (2)(3)(4)(5). Several groups have proposed that precise tissue and developmental regulation is accomplished by complex protein-protein interactions between factors that bind the LCR and those that bind the promoters of individual genes (6,7). The competition model suggests that a number of erythroidspecific and ubiquitous factors bind the LCR and enable this region to function as a potent enhancer. Downstream genes then compete for productive interactions with the LCR (2)(3)(4)(5). Presumably, yolk sac-, fetal liver-, and bone marrow-specific factors bind to globin gene promoters and proximal enhancers and provide individual genes with a competitive advantage for interaction with the LCR at the appropriate developmental stage.
Erythroid Kruppel-like Factor (EKLF) is one factor that is critical for the developmental stage-specific switch from the ␥to ␤-globin expression (reviewed in Ref. 8). This erythroidspecific transcription factor binds the ␤-globin promoter and activates high level expression (9,10). EKLF is a 358-amino acid protein containing an amino-terminal, proline-rich transactivation region (aa 1-275) and a COOH-terminal DNA binding domain (aa 276 -358) (10). The DNA binding domain consists of three C2H2 Kruppel-like zinc fingers that bind specifically to the CCACACCCT motif at Ϫ90 of the ␤-globin promoter (11). Although EKLF also binds to the CACCC box in the ␥ promoter, the binding affinity to ␥ CACCC is 8-fold lower than the binding affinity for ␤ CACCC and EKLF preferentially activates the ␤ gene in transient transfection assays (12). Targeted deletion of EKLF in mice results in a drastic reduction in the ␤-globin gene expression, but ␥ gene expression remains unaffected (13,14). Finally, persistence of ␥ gene expression during development is observed in EKLF knockout mice (15,16). These observations are consistent with the view that EKLF plays a central role in ␥to ␤-globin gene switching by binding specifically to the ␤ promoter and providing a competitive advantage for interactions with the LCR in adult erythroid tissue. Recent studies also suggest that EKLF exerts important functions at sites other than the ␤-globin promoter (17,18).
Although a large amount of information suggests a critical role for EKLF in globin gene switching, the mechanisms underlying EKLF transactivation functions are not clearly understood. EKLF is expressed, and is functional at all developmental stages, yet its effects are apparent only during definitive hematopoiesis (19 -21). The molecular basis of this specificity is unknown. The dissection of EKLF functional domains is an important step in elucidating these mechanisms. Previous studies have localized a transactivation domain at the aminoterminal region (aa 1-104) (22). In this report, we describe the identification of a new transactivation domain. This domain, consisting of amino acids 140 -358, is sufficient for maximal transactivation of the ␤-globin promoter. Furthermore, an essential region within this domain (aa 140 -225) is completely redundant with the amino-terminal transactivation domain (aa 1-139).

EXPERIMENTAL PROCEDURES
Plasmid Constructions-The HA-EKLF plasmid was constructed by subcloning a 63-base pair double-stranded oligomer (top: 5Ј-CATGGG-TAGCAGCTACCCTTACGACGTGCCCGACTACGCCAGCCTGGGCG-GCCCTAGCAGAGG-3Ј; bottom: 5Ј-CATGCCCCTGCTAGGGCCGCC-CAGGCTGGCGTAGTCGGGCACGTCGTAAGGGTAGCTGCTACC-3Ј) encompassing the HA epitope at the NcoI site of the EKLF expression plasmid pSG5/EKLF (10). This HA-EKLF plasmid was used as the * This work was supported by a grant from the NHLBI. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. ‡ Current address: National Institutes of Health, NICHD, Laboratory of Molecular Embryology, Bethesda, MD 20892-5431.
Cell Culture and Transfections-K562 cells were grown and electroporated as described previously (12). COS cells were grown in Dulbecco's modified Eagle's medium, 10% fetal bovine serum and transfected with different mutants along with CMV-Lac-Z using the calcium chloride coprecipitation method. Forty-eight hours after transfection, nuclear extracts were prepared (23), and Western blots were performed using ϳ50 g of nuclear extracts that were normalized by ␤-gal activity. The anti-HA antibody 12CA5-HRP (Roche Molecular Biochemicals) was used at a 1:2500 dilution, and gels were developed using the ECL system (Amersham Pharmacia Biotech). Transactivation assays were performed as described previously (12). Each construct was tested in at least three independent experiments. Within each experiment, each construct was electroporated in duplicate. CAT assays were performed in duplicate on each electroporated sample, and ␤-gal assays were performed in triplicate on each electroporated sample. The CAT assay values were normalized by ␤-gal values.
Indirect Immunofluorescence-Indirect immunofluorescence was conducted as described previously (24), except that the cells were blocked with 10% fetal bovine serum/phosphate-buffered saline for 10 min after the permeabilization step. Coverslips were mounted in media containing propidium iodide (Vector Laboratories). 12CA5 anti-HA antibody (generous gift of Dr. Susan Ruppert) was used (1:500 dilution) as the primary antibody, and a FITC anti-mouse antibody (Santa Cruz) was used as the secondary antibody (1:500 dilution). Images were captured with a Hamamatsu 3CDC camera mounted on a Nikon eclipse E800 microscope.
Gel Shift Assays-COS cell nuclear extracts that were normalized by ␤-gal activity (ϳ20 g) as described above were used in gel shift assays using a double-stranded oligomer encompassing the mammalian ␤ promoter CACCC box (8). The binding reactions were carried out in a buffer containing 5 mM Tris, 0.5 mM dithiothreitol, 0.5 mM EDTA, 25 mM NaCl, and 1% Ficoll. Before loading the sample on the gel, a 0.1 volume of 20% Ficoll was added to each reaction. For supershift/ablation assays, HA antibody was included in the binding reaction.

RESULTS
We performed a mutational analysis of EKLF using a previously established transient transfection assay (12) to map essential transactivation domains. The transactivation assay utilizes the human erythroid cell line K562, which expresses little endogenous EKLF (12), and an LCR ␥-␤ reporter (HS2-␥/Luc-␤/Cat) that contains wild-type human ␥and ␤-globin promoters (Fig. 1). In this assay, full-length EKLF activates ␤/CAT expression 500-fold compared with vector alone; therefore, the system provides a sensitive, quantitative assay for transactivation domains. We first tagged EKLF at the NH 2 terminus with the influenza HA epitope (HA-EKLF) to allow detection of the protein in transfected cells. HA-EKLF transactivated the ␤-globin promoter in the HS2-␥/Luc-␤/Cat reporter at the same level (500-fold) as wild-type EKLF. Using this parental plas-mid, we constructed three progressive amino-terminal truncations of HA-EKLF: HA-⌬4 -139 that deletes amino acids 4 through 139, HA-⌬4 -225 that deletes amino acids 4 through 225, and HA-⌬4 -254 that deletes amino acids 4 through 254 ( Fig. 2A). The same methionine start site was retained for all mutants to avoid differences in the site of translation initiation. As an added control, untagged versions of each mutant were also produced and tested in the same assay; no differences between HA-tagged and untagged proteins were observed (data not shown).
Each deletion construct was cotransfected with the LCR ␥-␤ reporter and Lac-Z control and tested for the ability to transactivate the ␤-globin promoter. The results are depicted in Fig.  2B. In these experiments full-length HA-EKLF (WT) activated the ␤-globin promoter 500-fold relative to the vector alone and is depicted as 100%. The activities of different mutants are expressed as a percentage of WT. Surprisingly, HA-⌬4 -139 activated the ␤ promoter at wild-type levels (99 Ϯ 21%) even though this deletion removed the transactivation domain previously identified by Chen and Bieker (amino acids 1-104) (22). A further 85 amino acid deletion (HA-⌬4 -225) drastically reduced ␤/CAT activation to 2.9 Ϯ 2%. Deletion to amino acid 255 (HA-⌬4 -254) resulted in very low levels of ␤/CAT activation (1.0 Ϯ 0.6%). These data demonstrate that amino acids 140 -358 encompass a potent transactivation domain that is sufficient for maximal transactivation of the ␤-globin promoter. Furthermore, the results suggest that an 85-amino acid subdomain (amino acids 140 -225) is critical for the function of this domain.
To ensure proper integrity of the mutants, we first assessed EKLF protein levels after transfection into COS cells. Mutant constructs were cotransfected with a Lac-Z plasmid to control for transfection efficiency, and Western blots were conducted on normalized nuclear extracts using an anti-HA antibody. These results are illustrated in Fig. 3. All mutant EKLFs were detected at near wild-type levels, suggesting that the proteins were stable.
We next assessed the subcellular localization of the mutants to determine whether the inactivity of HA-⌬4 -225 and HA-⌬4 -254 could be explained by protein mislocalization. Mutant constructs were transfected into COS cells, and HA-EKLFs were detected by indirect immunofluorescence with a FITC-labeled (green) anti-HA antibody (Fig. 4, A-E). Propidium iodide staining (red) was utilized to define the nucleus (Fig. 4, H-L), and a two-color merge (Fig. 4, O-S) was used to assess EKLF nuclear FIG. 1. Constructs used in the transactivation assay. The reporter plasmid HS2 ␥/Luc-␤/Cat has been described previously (12). This construct contains a human 1.5-kb KpnI-BglII HS2 fragment linked to a human ␥-globin gene promoter (Ϫ299 to ϩ37) driving the firefly luciferase (Luc) gene and a human ␤-globin promoter (Ϫ265 to ϩ48) driving the chloramphenicol acetyltransferase (Cat) gene. The activator plasmid (pSG5/HA-EKLF) contains an SV40 promoter driving HA-EKLF (wild-type or mutant). A CMV/Lac-Z gene was cotransfected with the activator and the reporter into K562 cells to serve as a control for transfection efficiency.
localization. As illustrated in the figure, all mutant EKLFs localized to the nucleus. These data suggest that the transcriptional inactivity of HA-⌬4 -225 and HA-⌬4 -254 does not result from protein mislocalization.
Finally, we examined all mutant EKLFs for the ability to bind DNA. Gel shift assays were performed with nuclear ex-tracts obtained from COS cells following transfection of the HA constructs described above. The binding sequence was a labeled oligomer (20 base pairs) encompassing the mammalian ␤-globin CACCC site (Fig. 5). Control extracts from cells transfected with the vector alone are shown in lanes 2 and 3. The background bands result from endogenous proteins that are known to bind the CACCC motif (8); as expected, HA antibody does not ablate or supershift these bands. Lanes 4 and 5 represent extracts from cells transfected with full-length HA-EKLF. Unfortunately, the EKLF gel shift band is obscured by a background band and, therefore, cannot be detected. However, unique gel shift bands that are distinct from endogenous bands are detected for all deletion mutants. Lanes 6, 8, and 10 represent extracts from cells transfected with EKLF mutants HA-⌬4 -139, HA-⌬4 -225, and HA-⌬4 -254, respectively. Specific bands that migrate at the predicted size are observed in each of these lanes. These data demonstrate that all mutant EKLFs bind DNA. To confirm that unique bands are EKLFspecific gel shifts, HA antibody was included in the binding reactions. Lanes 7, 9, and 11 demonstrate that HA antibody specifically ablates HA-⌬4 -139, HA-⌬4 -225, and HA-⌬4 -254  After 48 h, indirect immunofluorescence was performed using 12CA5 anti-HA primary antibody and FITC anti-mouse secondary antibody. Mounting medium contained propidium iodide, and cells were viewed through a fluorescence microscope using different color filters. The FITC filter depicts the nuclei expressing the HA tagged protein in green (panels A-G), Texas red filter shows propidium iodide stained nuclei in red (panels H-N), and twocolor merge demonstrates colocalization (yellow, panels O-U). Representative nuclei for each construct are shown. Over 100 transfected cells for each construct were analyzed, and greater than 99% of these cells showed similar results. gel shifts. We conclude that all EKLF mutants bind DNA and, therefore, that the transcriptional inactivity of HA-⌬4 -225 and HA-⌬4 -254 results from deletion of a novel transactivation domain.
The results described above demonstrate that amino acids 4 to 139 of EKLF are dispensable for maximal activation of the ␤-globin promoter, although this region encompasses the minimal activation domain described by Chen and Bieker (22). One possible explanation for this apparent discrepancy is that different systems were utilized to define transactivation domains. Chen and Bieker (22) used a GAL4 DNA binding domain fused to EKLF and tested the ability of fusion proteins to activate a heterologous target promoter (GAL4 binding sites-TATA-CAT) in 32DEpo1 cells. On the other hand, we utilized native EKLF and the natural ␤-globin promoter in K562 cells. To determine whether the amino-terminal domain of EKLF would transactivate in our assay, we constructed a new mutant, HA-⌬140 -226 ( Fig. 2A), that essentially fuses the amino-terminal domain (amino acids 1-139) to the transcriptionally inactive mutant HA-⌬4 -225. We then tested the ability of this protein to transactivate the ␤-globin promoter in our HS2-␥/Luc-␤/Cat reporter (Fig. 2B). Interestingly, this mutant activated the ␤ promoter at WT levels (95 Ϯ 23%). This result confirms that the aminoterminal region of EKLF contains a potent transactivation domain as originally described by Chen and Bieker (22). Fur-thermore, our data demonstrate that this amino-terminal domain is functionally redundant with the domain at aa 140 -225. As expected, the HA-⌬140 -226 EKLF protein is stable (Fig. 3), localizes to the nucleus (Fig. 4 panels F, M, and T), and binds DNA (Fig. 5, lanes 12 and 13). DISCUSSION The results described above identify a novel transactivation domain in the erythroid-specific transcription factor EKLF. This domain, which is contained within amino acids 140 -358 of the protein, is sufficient for maximal transactivation of the ␤-globin promoter (see HA-⌬4 -139, Fig. 2B). Furthermore, an 85-amino acid subdomain of this region (aa 140 -225) is essential for activity; deletion of this subdomain dramatically reduces transactivation (HA-⌬4 -225, Fig. 2B). Interestingly, this subdomain can be functionally replaced by amino acids 1-139 (HA-⌬140 -226, Fig. 2B), which contains a potent transactivation domain described by Chen and Bieker (22). These studies demonstrate that EKLF possesses at least two transactivation domains that function in a redundant manner. In this respect, EKLF is similar to another Kruppel-like family member, Sp1, which also contains two potent transactivation domains that are functionally redundant (25). Mutant constructs designed to delimit subregions within the 85-amino acid central domain (aa 140 -225) produced unstable proteins (data not shown), precluding further definition of this domain.
Our results also demonstrate that amino acids 255-358 encompass sequences sufficient for nuclear localization; the HA-⌬4 -254 EKLF mutant efficiently localized to the nucleus (Fig. 4, panels E, L, and S), although the protein did not activate transcription. To confirm this result we generated a new construct HA-⌬255-358 that deletes amino acids 255-358. This mutant was expressed at WT levels in whole cell extracts (data not shown). When tested for subcellular compartmentalization, the mutant localized exclusively to the cytoplasm (Fig.  4, panels G, N, and U). This result demonstrates that the first 254 aa of EKLF do not contain a nuclear localization sequence. The best candidate for the nuclear localization sequence is PKRSRR at position 260 -265. This sequence is highly homologous to the nuclear localization sequence (PKRGRR) identified in other Kruppel-like family members (26).
The molecular basis for transactivation through the newly identified, central domain of EKLF is not known. However, some insight can be obtained from studies of the amino-terminal domain. The activity of this domain is regulated by phosphorylation of a conserved threonine residue within the recognition site of casein kinase-II (27). EKLF is heavily phosphorylated on serine and threonine residues, and several consensus sites for phosphorylation are present within the central domain (27); therefore, the activity of this region may also be regulated by phosphorylation. Alternatively, transactivation may be a function of the proline-rich nature of this region. The activation domains of several transcription factors are rich in prolines (28), and previous studies have demonstrated that a proline stretch of 10 residues is sufficient to confer high transactivation capacity when fused to the GAL4 DNA binding domain (29). The entire EKLF protein is rich in prolines, except for the zinc finger domain (aa 275-358); therefore, both the amino-terminal and central domains may function through proline-rich sequences.
The redundancy of amino-terminal and central domains could be achieved in several ways. Both domains may interact independently with the same coactivator. This has been observed for the two redundant Sp1 activation domains, both of which interact with the same coactivator, TAFII110 (a component of the TFIID complex) (30). Alternatively, the two domains may interact with different coactivators that have re-  10 and 11), or HA-⌬140 -226 (lanes 12 and 13) were used in the binding reactions and run on a 5% acrylamide gel. Arrows depict the positions of bands corresponding to the mutants. The band corresponding to WT (HA-EKLF) is predicted to comigrate with one of the two high intensity background bands (the faster migrating band) and is therefore obscured. One l of the anti-HA antibody 12CA5 was used to ablate the gel shift bands in the designated lanes (lanes 3, 5, 7, 9, 11, and 13). Lane 1 shows the probe run alone. dundant activities. Previous studies have demonstrated that EKLF interacts physically with CBP, P300, and P/CAF in vivo, and GATA-1 in vitro (31,32). Furthermore, Armstrong et al. (33) recently purified the SWI/SNF-related complex E-RC 1 based on its ability to interact functionally with EKLF in vitro. Additional experiments will be required to define the coactivators that interact with each EKLF transactivation domain.
As mentioned above, EKLF is phosphorylated extensively, and Zhang and Bieker (31) recently demonstrated that specific lysines are acetylated (Lys 261 , Lys 270 ). These post-translational modifications may promote interactions with specific coactivators and factors of the basal transcription machinery. Differential post-translational modifications of the two transactivation domains during development may provide a mechanism for the temporal-specific transcriptional activity of EKLF on the ␤-globin promoter, and identification of additional EKLF interacting proteins should provide insights into globin gene switching.