The Proline-rich Homeodomain Protein Recruits Members of the Groucho/Transducin-like Enhancer of Split Protein Family to Co-repress Transcription in Hematopoietic Cells*

The proline-rich homeodomain protein (PRH/Hex) is important in the control of cell proliferation and differentiation and in the regulation of multiple processes in embryonic development. We have shown previously that PRH contains two domains that can independently bring about transcriptional repression. The PRH homeodomain represses transcription by binding to TATA box sequences, whereas the proline-rich N-terminal domain of PRH can repress transcription when attached to a heterologous DNA-binding domain. The Groucho/ transducin-like enhancer of split (TLE) family of proteins are transcriptional co-repressors that interact with a number of DNA-bound transcription factors and play multiple roles in development. Here we demonstrate that the proline-rich N-terminal domain of PRH binds to TLE1 in vitro and in yeast two-hybrid assays. We show that PRH and TLE proteins are co-expressed in hematopoietic cells and interact in co-immunoprecipi-tation assays. We demonstrate that TLE1 increases repression by PRH in transient transfection assays and that titration of endogenous TLE proteins by co-expres-sion of Grg5, a natural trans-dominant negative protein, alleviates transcriptional repression by PRH. Finally, we show that a mutation

The proline-rich homeodomain (PRH) 1 protein, also known as Hex (hematopoietically expressed), is an orphan homeodomain protein that functions as an important regulator of hematopoiesis (1). PRH was first identified in human and avian hematopoietic cells (1). PRH is strongly expressed in pluripotent hematopoietic progenitors, in erythromyeloid and B-cell progenitors but not in T-cell lineages (2)(3)(4)(5) and is down-regulated in most hematopoietic lineages during differentiation (3,6). PRH interacts with the growth control protein PML (7) and has been shown to regulate cell growth or differentiation in a number of different tissues (8 -10). Up-regulation of PRH expression is linked with several lymphoid leukemias (11,12). In mice transplanted with bone marrow transduced with a PRHexpressing retrovirus, PRH can act as an oncogene and cause T-cell-derived lymphomas (11,12). However, PRH also acts as a tumor suppressor in some hematopoietic lineages. PRH can inhibit oncogenic transformation by the translation initiation factor eIF4E by disruption of the mRNA transport activity of eIF4E through a direct interaction with eIF4E in the nucleus (13). Disruption of the nuclear localization of PRH is associated with the loss of this key regulatory function in a subset of myeloid leukemias (14). Thus, the effects of PRH on growth and differentiation are both dose-dependent and context-dependent. PRH also has an important role in the regulation of early embryonic patterning (4,15,16). Indeed, it plays a central role in the formation of the vertebrate head and the formation of many endoderm-derived organs such as liver and thyroid (5).
PRH functions as a transcriptional repressor in hematopoietic, liver, thyroid, and embryonic stem cells (15,(17)(18)(19). However, it has also been reported to activate transcription of its own gene in thyroid cells (20). PRH represses transcription in hematopoietic cells by binding to TATA box sequences and the TATA box binding protein using the PRH homeodomain. In addition, however, the proline-rich PRH N-terminal repression domain can repress transcription independently of the DNA binding activity of the homeodomain by as yet unidentified mechanisms (19). Although the PRH N-terminal repression domain is known to bind to PML, eIF4E, and the proteosome subunit C8, these interactions have not been shown to influence transcriptional repression in vivo (7,13,22).
The TLE proteins are members of a family of co-repressor proteins that includes the murine Grg proteins and the Drosophila Groucho protein. Groucho/TLE family proteins are involved in many developmental decisions, including neuronal and epithelial differentiation, segmentation and sex determination, and differentiation of hematopoietic, osteoblast, and pituitary cells (23)(24)(25)(26)(27). Members of the Groucho/TLE family do not have DNA binding activity but are instead recruited to DNA by interactions with DNA-binding proteins. In general, these proteins use a C-terminal region known as the WD repeat domain to interact with DNA-binding proteins (28,29) and an N-terminal glutamine-rich Q domain for tetramerization (30 -32). Some members of the TLE family of proteins, such as Grg5, lack the WD repeat regions but retain the oligomerization Q domain. Overexpression of Grg5 can relieve TLE-mediated corepression, presumably because this protein retains the ability to tetramerize with full-length TLE proteins (32)(33)(34). Once recruited to a promoter, the Groucho/TLE proteins can bring about transcriptional repression by recruiting histone deacetylases (35)(36)(37) or by directly interacting with histones (38) or with the basal transcriptional machinery (39,40).
Here we show that the PRH N-terminal repression domain can interact with TLE1 in vitro and in vivo using a short sequence of amino acids that corresponds to an Engrailed homology (Eh-1) motif. Moreover, we demonstrate that the interaction between TLE1 and PRH is required for enhanced repression of transcription by PRH. Furthermore, we show that titration of endogenous TLE proteins by Grg5 decreases repression by PRH. Thus, TLE proteins co-repress transcription with PRH.

EXPERIMENTAL PROCEDURES
Bacterial Expression Plasmids-A GST-tagged avian PRH N terminus (GST-PRH N1-141 ) expression vector has been described previously (19). PCR fragments encoding amino acids 1-125 or 61-141 of avian PRH and flanked by 5Ј SalI and 3Ј SpeI restriction sites were cloned into pGEX20T (Amersham Biosciences) downstream of the GST moiety to create pGEX-PRH N61-141 and pGEX-PRH N1-125 . The GST-tagged human PRH N terminus expression vector pGEX-MycPRH N1-132 was created as follows. Briefly, a DNA sequence encoding the Myc-tagged human PRH N terminus (amino acids 1-132) was cloned as a BamHI-StuI fragment from pMUG1-MycPRH (see below) into pGEX20T (Amersham Biosciences). The PRH fragment was inserted between the unique BamHI site and the XbaI site in the vector by first modifying the XbaI restriction site with Klenow enzyme to blunt the XbaI restriction site. The pGEX-MycPRH F32E N1-132 plasmid that expresses the PRH N terminus carrying the F32E mutation was created exactly as described above except that the PRH N-terminal fragment was obtained from pMUG1-MycPRH F32E (see below). The DNA sequence of these plasmids and the plasmids described below were verified by DNA sequencing.
Mammalian Expression and Reporter Plasmids-The pTK-PRH and pTK-GAL reporter plasmids have been described previously (19). pSV-␤-galactosidase control vector (pSV-lacZ) was obtained from Promega. pBSK II-HPRH is a vector carrying the full-length human PRH cDNA and was a gift from Dr. G. Manfioletti (University of Trieste). The mammalian expression plasmid pFLAG-Grg5 (pCMV-Tag1-AES) expresses the Grg5 protein tagged with the pFLAG epitope and has been described previously (41) and was a gift from Dr. T. Okamoto (Nagoya University). The mammalian expression plasmid pCMV2-FLAGTLE1 contains the TLE1 coding sequence in frame with the FLAG epitope. The FLAGTLE1 construct was generated by first digesting a pBluescript-TLE1 plasmid with BanII, followed by removal of protruding ends with T4 DNA polymerase and recovery of a 1.6-kb fragment encoding the N-terminal region of TLE1. This fragment was subcloned into pCMV2-FLAG digested with EcoRV. A second pBluescript-TLE1 restriction fragment obtained after digestion with SmaI and encoding the remaining half of TLE1 was then ligated in frame into the first ligation product to generate pCMV2-FLAGTLE1 expressing full-length TLE1.
The mammalian expression plasmid pMUG1-MycPRH expresses full-length human PRH and was created as follows. pUHD15-1 (42) was modified by replacing the sequence between the unique BamHI and EcoRI sites with a linker that destroys these two restriction sites and contains a multiple cloning sequence (MCS). The sequence of the linker is as follows: 5Ј-A ATT GGA TCC ATG GGA ATT CGA GGT CGA CAG TGA-3Ј. The linker contains a translational start signal (boldface type) and BamHI, NcoI, EcoRI, and SalI restriction sites. The resulting pMUG1 plasmid contains the CMV promoter with an MCS downstream. A BamHI-SmaI double-stranded oligonucleotide encoding a Myc tag (Myc 9E10 epitope) (5Ј-GATCCATGGAACAAAAACTCATCT-CAGAAGAGGATCTG-3Ј) and a SmaI-EcoRI fragment from pBSK-HPRH carrying the human PRH coding sequence from amino acid 7 was inserted between the BamHI and EcoRI sites in pMUG1. This results in an expression construct where the PRH coding sequence was placed in frame with the Myc tag and the ATG in the MCS.
Yeast Plasmids-pACT2 contains the GAL4 activation domain upstream of a MCS (Clontech). To create pACT2-HPRH, a SmaI-EcoRI fragment from pBSK-HPRH carrying the PRH coding sequence from amino acid 7 was ligated into pACT2 between the BamHI and EcoRI restriction sites. The BamHI site in pACT2 was filled in using Klenow enzyme to allow blunt end ligation with the PRH SmaI site. pGBT9-TLE1 contains the TLE1 cDNA in frame with the GAL4 DNA binding domain and has been described previously (31,43). pAS2-1 contains the GAL4 DNA binding domain upstream of a MCS (Clontech). To create pAS2-1-PRH N1-132 , an EcoRI-StuI fragment from pBSK-HPRH encoding the N-terminal 132 amino acids of PRH was ligated between the EcoRI and SmaI sites of pAS2-1. An oligonucleotide was inserted between the EcoRI site in pAS2-1 and the internal SmaI site in HPRH to achieve the correct reading frame (5Ј-CATGCAGTACCCGCACCCC-3Ј). To create the deletion mutant pAS2-1-PRH N1-98 , the pAS2-1-PRH N1-132 construct was digested with BamHI and partially digested with ApaI. An ApaI-BamHI oligonucleotide (5Ј-CGCCGCGCCCACG-3Ј) was then ligated between the ApaI site located at amino acid 98 within the PRH amino acid sequence and the unique BamHI site in the vector pAS2-1. pGAD424-TLE1 contains the full-length TLE1 coding sequence and expresses a GAL4-TLE1 fusion protein. pGAD424-TLE1Q-SP and pGAD424-TLE1WD contain the Q (amino acids 1-135), Q-SP (amino acids 1-435), and WD (amino acids 444 -770) domains of TLE1, respectively. These plasmids have all been described previously (31,43).
Yeast Two-hybrid Assay-The two-hybrid assay (44) was carried out in yeast strain MaV203 (Clontech). The two-hybrid assay and ␤-galactosidase assays in liquid culture were performed as described by Fields and Song (44) and the Clontech manual. Colony lift assays were performed by transferring yeast colonies onto 3MM filter paper. The yeast were then lysed by rapidly freeze thawing in liquid nitrogen and then placed on filter paper saturated with Z-buffer (60 mM Na 2 HPO 4 , 40 mM NaH 2 PO 4 , 10 mM KCl, 1 mM MgCl 2 , 50 mM ␤-mercaptoethanol) containing X-gal indicator (0.32 mg/ml). The blue color for each transformant was compared after incubation of the filter at 37°C for 3 h.
Mutagenesis-The QuikChange kit (Stratagene) was used for the mutagenesis of pMUG1-MycPRH to PRH F32E and was used according to the manufacturer's instructions. The resulting mutant was fully sequenced to confirm the sequence change.
Expression and Purification of Tagged PRH Proteins-The human and avian GST-PRH fusion proteins were expressed in BL21 pLysS cells (Novagen). Fusion protein expression was induced with 1 mM isopropyl-1-thio-␤-D-galactopyranoside. Cells were harvested and lysed by incubation with 100 l of lysozyme (1 mg/ml) for 20 min followed by sonication in PBS plus 1% Triton X-100. GST-PRH N fusion proteins were purified over glutathione-Sepharose 4B beads (Sigma) according to the manufacturer's instructions and snap frozen in liquid nitrogen. Aliquots of these proteins were eluted with 10 mM glutathione and assayed for purity by SDS-PAGE followed by staining with Coomassie Blue. Proteins were quantified using the Bio-Rad phosphoric acid protein assay.
In Vitro Binding Assays-The plasmid used for the in vitro transcription and translation of TLE1, pcDNA3-TLE1, has been described previously (27,43). Transcription and translation was carried out using a TNT kit (Promega) according to the manufacturer's protocol. Approximately 20 g of each of the GST fusion proteins or an equimolar amount of GST protein as judged by Coomassie staining was bound to 50 l of glutathione-Sepharose 4B beads (Sigma) according to the manufacturer's instructions. To assay for specific interactions, 10 l of [ 35 S]methionine-labeled in vitro translated protein was added and incubated in binding buffer (20 mM Hepes pH 7.8, 200 mM KCl, 5 mM MgCl 2 , 0.5 mM dithiothreitol, 0.5% Nonidet P-40, 50 ng/l bovine serum albumin) with gentle agitation for 60 min at 4°C. The beads were then washed six times with 1 ml of binding buffer. Bound proteins were eluted by boiling the beads in protein sample buffer containing 1% SDS and analyzed by SDS-PAGE and fluorography.
Pull-downs and Western Blotting-Whole cell extract from 2 ϫ 10 8 K562 cells was made as follows. The cell pellet was washed in PBS twice and then resuspended in 1 ml of high salt lysis buffer (500 mM NaCl, 50 mM Tris, pH 7.5, 0.1% SDS, 0.1% Nonidet P-40). The cell suspension was drawn up and down six times with a 3ϫ Monojet needle (1.1 ϫ 50 mm, 19 gauge 2 inches) and then incubated on ice for 5 min. The extract was centrifuged at maximum speed for 5 min at 4°C in an Eppendorf microcentrifuge. Whole cell extract was added to ϳ10 g of GST-HPRH N protein or 10 g of GST protein bound to glutathione resin. After 2 h at 4°C with tumbling, the resin was collected by centrifugation in an Eppendorf microcentrifuge, washed three times in 1 ml of wash buffer (150 mM NaCl, 50 mM Tris, pH 7.5, 0.1% SDS, 0.1% Nonidet P-40), and then resuspended in 50 l of 2ϫ SDS loading buffer. All operations were carried out at 4°C and in the presence of protease inhibitors. After SDS-PAGE, the proteins were immunoblotted onto Immobilon-P membrane (Sigma). TLE proteins were detected using a rat monoclonal pan TLE antibody raised against the conserved carboxyl-terminal WD40 domain of Groucho/TLEs (25,45) and an ECL kit (Amersham Biosciences). HC8 was detected with a mouse monoclonal antibody (Affiniti).
Immunofluorescence-K562 cells were grown to a density of 1 ϫ 10 6 cells/ml. 10 ml of cells were collected by centrifugation and resuspended in 1 ml of PBS. Glass slides were coated with poly-L-lysine (Sigma) for 10 min at 20°C and then washed with water. The washed K562 cells were incubated on the coated slide for 10 min at 20°C and then washed in PBS. The cells were fixed in paraformaldehyde for 30 min and then rinsed twice in PBS and permeabilized by incubation in 0.1% Triton X-100 in PBS for 10 min. The cells were then rinsed with PBS and incubated with PBSA (3% bovine serum albumin in PBS) for 20 min to block nonspecific antibody binding. After rinsing in PBS, antibody staining was performed with a 1:10 dilution of a rat monoclonal pan-TLE antibody and a 1:10 dilution of a mouse polyclonal anti-PRH antibody for 1 h at 20°C. The cells were rinsed in PBS twice and in PBSA twice and then incubated with secondary antibodies for 1 h at 20°C. PRH was detected with a TRITC donkey anti-mouse secondary antibody (Stratec) that had been preadsorbed for immunoreactivity against rat antibodies. MycPRH was detected with a monoclonal antibody (Santa Cruz Biotechnology, Inc., Santa Cruz, CA). TLE was detected with a 1:100 dilution of a biotinylated rabbit anti-rat secondary antibody (Vector Laboratories Inc.) and a 1:100 dilution of a fluoresceinlabeled goat anti-Biotin tertiary antibody (Vector Laboratories). Immunostained cells were viewed on a Leica DM IRBE confocal microscope and imaging performed using Leica confocal software version 2.00.
Cell Culture and Transient Transfection Assays-K562 cells were grown in glutamine-added Dulbecco's modified Eagle's medium (Sigma) supplemented with 10% fetal calf serum to a density of ϳ1 ϫ 10 6 cells/ml. The cells were then collected by centrifugation and then resuspended in medium plus 10% fetal calf serum to a density of 5 ϫ 10 7 cells/ml. 5 g of the luciferase reporter plasmid with 5 g of the ␤-galactosidase reporter plasmid and the amount of expression plasmids indicated under "Results" were electroporated into 1 ϫ 10 7 cells using a Bio-Rad Genepulser (200 V, 975 microfarads). The cells were rested for 10 min and then incubated overnight in 10 ml of supplemented medium. After 24 h the cells were harvested, and luciferase activity was assayed using the Promega luciferase assay system according to the manufacturer's instructions. ␤-Galactosidase assays were performed as an internal control for transfection efficiency. After subtraction of the background, the luciferase counts were normalized against the ␤-galactosidase value.
Co-immunoprecipitation Assays-K562 cells (2 ϫ 10 7 ) were co-transfected with 2 g of pMUG1-MycPRH and either 2 g of pCMV2-FLAGTLE1 or 2 g of pCMV-FLAG-Grg5 as described above. Five transfections were pooled (1 ϫ 10 8 cells) to make the nuclear extract for each co-immunoprecipitation. Nuclear extracts were made by the method of Dignam et al. (46) with modifications described by Dorn et al. (47). Nuclear extracts were incubated with 4 l of a monoclonal anti-Myc9E10 antibody (Santa Cruz Biotechnology) for 30 min at 4°C. 100 l of a 50% slurry of Protein G beads (Sigma) was then incubated with the nuclear extracts for a further 90 min at 4°C. After this time, the resin was collected by centrifugation in an Eppendorf microcentrifuge (13,000 rpm for 1 min), washed three times in 1 ml of wash buffer B (150 mM NaCl, 50 mM Tris, pH 7.5, 0.2% Nonidet P-40), and then resuspended in 50 l of 2ϫ SDS loading buffer. All operations were carried out at 4°C and in the presence of protease inhibitors. After SDS-PAGE, the proteins were immunoblotted onto Immobilon-P membrane, and TLE and Grg5 proteins were detected using anti-FLAG antibodies (Sigma). Approximately 10% of the nuclear extract (200 g of protein) was used for Western blotting experiments as described above.

RESULTS
PRH and TLE1 Interact in Yeast-PRH contains two motifs that could be involved in an interaction with TLE proteins. Amino acids 30 -39 (TPFYIEDILG) contain a sequence that matches the Eh-1 motif (FXIXXIL) identified originally in the Drosophila homeodomain proteins Engrailed and Goosecoid and shown to mediate the interaction of both these proteins with Groucho (48,49). The second is a putative "Runt/Hairy" motif (LLWSPF amino acids 124 -129 in PRH), which loosely resembles the tryptophan containing motifs WRPY and WRPW that are used to recruit the Groucho protein to the Runt and Hairy transcription factors, respectively (50, 51). To investigate whether PRH might interact with TLE proteins, we made use of the yeast two-hybrid assay (Fig. 1). The human PRH cDNA was placed in frame with GAL4 activation domain in the vector pACT2 to create pACT2-PRH (see "Experimental Procedures"). pGBT9-TLE1 contains the full-length TLE1 cDNA (amino acids 1-770) in frame with the GAL4 DNA binding domain (31,43). pACT2-PRH and pGBT9-TLE1 were co-transformed into yeast strain MaV203. This strain contains integrated copies of the LacZ and His reporter genes under the control of GAL4-dependent promoters. A functional interaction between the two hybrids in this strain would be expected to produce ␤-galactosidase activity and histidine prototrophs. Although expression of pGBT9-TLE1 with a vector containing the GAL4 activation domain or expression of pACT2-PRH with a vector containing the GAL4 DNA binding domain did not result in the production of ␤-galactosidase activity (Fig. 1B, columns 1 and 2), significant ␤-galactosidase activity was detected when pGBT9-TLE1 and pACT2-PRH were co-expressed (Fig. 1B, column 3). These data suggest that a functional transcription factor is produced only when both proteins are present and therefore that PRH interacts with TLE1 in this assay. However, the co-expression of PRH and TLE1 results in very poor growth of the transformed yeast strain when histidine is added to the growth medium and no growth in the absence of histidine (not shown). This suggests that the interaction of the full-length PRH and TLE proteins is toxic for this yeast strain.
To confirm that the PRH-TLE1 interaction occurs and to establish whether the N-terminal domain of PRH is responsible for the interaction, we repeated the yeast two-hybrid assay with a truncated PRH construct (Fig. 1C). In this case, fulllength TLE1-(1-770) was fused to the GAL4 activation domain in pGAD424 to create pGAD424-TLE1 (31,43). A fragment of the human PRH cDNA encoding the first 98 amino acids of PRH was cloned into the GAL4 DNA binding domain vector pAS2-1 to create pAS2-1-PRH N1-98 (see "Experimental Procedures"). This fragment of PRH lacks the Runt/Hairy-like motif (LLWSPF) described earlier but contains the putative Eh-1 motif. These constructs were then transformed into yeast. To inhibit leaky expression of the histidine gene, transformants were assayed for growth on dropout medium plates containing 50 mM 3-aminotriazol (3AT) and subsequently assayed for lacZ expression. Yeast co-transformed with PRH N1-98 and TLE1 grew well on ϪHLT (ϩ50 mM 3AT) dropout medium, whereas yeast transformed with only one of the partners and the corresponding empty vector either did not grow or grew very poorly on this medium (Fig. 1C). Moreover, only yeast co-transformed with PRH N1-98 and TLE1 resulted in colonies that produced ␤-galactosidase activity in X-gal colony lift assays (Fig. 1C, compare sectors 7 and 2 on X-gal). Thus, it appears that the N-terminal 98 amino acids of PRH are sufficient for the interaction of PRH with TLE1 and that the Runt/Hairy-like sequence found in PRH is not required for the interaction.
The TLE1 protein is 770 amino acids long and has several functional domains (Fig. 1A) (45). The N-terminal region of the protein (amino acids 1-135) is a glutamine-rich tetramerization domain known as the Q domain. The middle of the protein (amino acids 135-450) contains a glycine/proline-rich G/P domain, a nuclear localization and CKII phosphorylation region (CcN), and a Ser/Thr/Pro-rich SP domain. The C terminus of the protein contains a WD repeat domain (amino acids 450 -770) that can mediate protein-protein interactions with transcription factors (28,29,31,52). To determine which regions of TLE1 are involved in the interaction between PRH and TLE1, a series of TLE1 deletion mutants were fused to the GAL4 activation domain in pGAD and co-transformed into yeast with pAS2-1-PRH N1-98 . The TLE deletion mutants pGAD-TLEQ-(1-135), pGAD-TLEQ/SP-(1-435), and pGAD-TLEWD-(444 -770) have been described previously (31). Interestingly, when cotransformed with pAS2-1-PRH N1-98 , all of the TLE deletion mutants allowed growth of the corresponding yeast transformants on ϪHLT (ϩ50 mM 3AT) medium and also resulted in ␤-galactosidase activity in X-gal filter lift assays (Fig. 1C, sectors 7-10) and in liquid ␤-galactosidase assays (not shown). These data suggest that PRH N1-98 is able to interact with both the Q domain at the N terminus of TLE1 and also with the WD repeat domain at the C terminus of TLE1.
To confirm the growth phenotypes of the yeast co-transformed with these constructs, we made use of the Ura Ϫ (ϩ5fluoroorotic acid) counterselection assay (53). Yeast strain MaV203 is Ura Ϫ and contains an integrated copy of the Ura3 gene under the control of a Gal4-dependent upstream activating sequence. Only yeast carrying interacting proteins become Ura ϩ and can convert 5-fluoroorotic acid (5-FOA) to the highly toxic compound 5-fluorouracil. As can be seen from the data in Fig. 1C, yeast containing any of the TLE1 proteins and PRH N1-98 die on medium containing 5-FOA (Fig. 1C, sectors 7-10, on FOA). In contrast, yeast transformed with the two empty vectors pAS2-1 and pGAD424 (sector 1) or each of the TLE1 proteins with pAS2-1 (sectors 2-5), or PRH N1-98 with pGAD424 (sector 6) show very limited growth on ϪHLT (ϩ3AT) medium, no ␤-galactosidase activity, and growth on medium containing 5-FOA. Taken together, these data confirm that PRH and TLE1 can interact in yeast. Furthermore, these data suggest that both the Q domain and the WD domains of TLE1 are involved in the interaction with the first 98 amino acids of PRH.
Immunostaining of PRH and TLE Proteins in K562 Cells-To investigate further the biological significance of the interaction between PRH and TLE proteins, we next examined the distribution of these proteins in K562 hematopoietic cells. PRH is known to be strongly expressed in this cell line, and other PRH-interacting proteins have been identified from yeast two-hybrid screens with cDNAs obtained from this cell line (6,22). The K562 cell line was originally obtained from a patient with chronic myeloid leukemia in blast crisis, and in culture  (2), and pACT2-PRH and pGBT9-TLE1 (3). The experiment was performed several times, and the results of one experiment performed in triplicate are shown. The data are presented as the mean and S.E. C, top, a schematic of the fusion proteins GAL4 DNA binding domain-PRH N and GAL4 activation domain-TLE1 at the GAL4 upstream activating sequence. Below, growth phenotypes on ϪLT medium, ϪHLT (ϩ50 mM 3AT) medium, ϪLT (ϩ0.2% 5FOA) medium, and ␤-galactosidase activity of yeast transformants in a colony lift assay (X-gal). The sectors contain pAS2-1 and pGAD424 (1), pAS2-1 and pGADTLE1 (2), pAS2-1 and pGADTLE1 WD (3), pAS2-1 and pGADTLE1 Q (4), pAS2-1 and pGADTLE1 Q-SP (5) pAS2-1-PRH N1-98 and pGAD424 (6), pAS2-1-PRH N1-98 and pGADTLE1 (7), pAS2-1-PRH N1-98 and pGADTLE1 WD (8), pAS2-1-PRH N1-98 and pGADTLE1 Q (9), and pAS2-1-PRH N1-98 and pGADTLE1 Q-SP (10). these cells spontaneously give rise to multiple cell types, larger myeloblasts, and smaller myeloid cells at various stages of differentiation (55). To determine the intracellular localization of the endogenous PRH and TLE proteins in the K562 cell line, we used confocal laser microscopy and immunofluoresence. Fig.  2A shows endogenous TLE staining of K562 cells with a fluorescein-labeled pan-TLE antibody (green signal), and Fig. 2B shows DNA staining with DAPI (blue signal). TLE is present in both the nucleus and the cytoplasm of K562 blasts and smaller myeloid cells. However, the staining pattern of the cells is not uniform. For example, cells 1 and 2 both show a predominant cytoplasmic staining for TLE, and cells 3 and 4 show predominant nuclear staining for TLE (compare DAPI and fluorescein staining). The differences in subcellular localization of TLE proteins in these cells may reflect the differentiation status of these cells. Certainly, TLE proteins are known to become more strongly associated with the nuclear compartment during the neural differentiation of P19 embryonic carcinoma cells (56).   2C shows endogenous PRH staining of the same cells with a TRITC-labeled anti-PRH mouse polyclonal antibody (red signal). This antibody stains both the nucleus and the cytoplasm of K562 cells. In all cases when cells were stained without the primary antibody, there was little, if any, detectable immunofluoresence (data not shown). As expected, co-immunofluoresence experiments using TRITC-labeled anti-PRH mouse polyclonal antibodies and the same fluorescein labeled pan-TLE rat monoclonal antibodies show that a significant proportion of, but not all, K562 cells contain both PRH and TLE proteins in the nucleus (Fig. 2G). Fig. 2 also shows a high magnification image of a K562 cell, where endogenous PRH is present in the nucleus and cytoplasm (Fig. 2G, TRITC staining) and endogenous TLE proteins are strongly nuclear (Fig. 2F, FITC staining). Co-immunofluoresence of this cell shows that both endogenous PRH and TLE proteins are present in the nucleus (Fig. 2H).
Since both TLE and PRH proteins can be found in both cytoplasmic and nuclear compartments, we wanted to determine whether overexpressed TLE1 and PRH proteins are predominantly cytoplasmic or nuclear. Transfection of K562 cells with either pMUG1-MycPRH or pFLAGTLE1 followed by immunostaining with a Myc monoclonal antibody or the rat pan-TLE antibody, respectively, results in around 10% of cells expressing the transfected proteins. However, in both cases, immunofluoresence experiments show that the transfected proteins are located in the nucleus (data not shown). Similarly, when K562 cells are co-transfected with both PRH and TLE1 expression plasmids, both MycPRH and FLAGTLE1 are strongly localized to the nucleus. Fig. 2I shows a high magnification transmitted light image of a transfected K562 cell. Fig.  2, J and K, shows TLE1 (FITC) and PRH (TRITC). Fig. 2L shows co-localization of TLE1 with PRH in the nucleus of a transfected K562 cell.
PRH and TLE Proteins Interact in Vitro and in K562 Hematopoietic Cells-To provide biochemical evidence for the interaction between PRH and TLE1, we carried out in vitro binding studies and pull-down experiments. The equivalent N-terminal regions of human PRH (amino acids 1-132) and avian PRH (amino acids 1-141) were expressed in bacteria as GST fusion proteins and partially purified. In vitro transcription and translation was used to produce labeled TLE1. Glutathione beads carrying GST or the GST-PRH N proteins were incubated with labeled TLE1 and then washed extensively. Bound protein was eluted by boiling in SDS-PAGE loading buffer and analyzed by SDS-PAGE. Fig. 3A shows that labeled TLE1 binds to the human GST-PRH N1-132 (lane 3) and avian GST-PRH N1-141 (lane 4) fusion proteins but binds only very weakly, if at all, to GST alone (lane 2). Thus, both the human and the avian PRH N-terminal domains bind to TLE1 in vitro. To extend our in vitro binding studies, we determined whether the purified human PRH N terminus is able to interact with endogenous TLE proteins in K562 whole cell extracts. Human GST-PRH N1-132 or GST alone were immobilized on glutathione-Sepharose beads and incubated with K562 whole cell extracts. After extensive washing, bound proteins were eluted by boiling in SDS-PAGE loading buffer and separated by SDS-PAGE. The bound proteins were then probed with the pan-TLE rat monoclonal antibody in a Western blot. The GST-PRH N1-132 protein is clearly able to retain the endogenous TLE proteins present in the whole cell extracts (Fig. 3B, lane 3). However, an equivalent amount of GST (as judged by Coomassie staining) does not retain TLE (Fig. 3B, lane 2). We conclude that the human PRH N-terminal protein purified from bacterial cells is able to interact with the endogenous TLE proteins present in K562 cell extracts.
To confirm these results, co-immunoprecipitation studies were carried out using full-length PRH tagged with the Myc9E10 epitope (pMUG1-MycPRH) and TLE1 tagged with a FLAG epitope (pCMV2-FLAGTLE1) (Fig. 3C). K562 cells were transiently co-transfected with MycPRH and FLAGTLE1 plasmids, and nuclear extracts were made from the transfected cells. Proteins in these nuclear extracts were separated on SDS-polyacrylamide gels and probed with an antibody raised against the FLAG epitope. FLAG-tagged TLE1 proteins with a molecular mass of ϳ90 kDa were detected with this antibody in Western blotting experiments (Fig. 3C, lane 1). Shorter exposures show the presence of a doublet of TLE1 proteins (data not shown). The same nuclear extract was incubated with either Protein G-Sepharose beads alone or with an anti-Myc mouse monoclonal antibody bound to Protein G-Sepharose beads. After extensive washing, the bound proteins were separated on SDS-polyacrylamide gels and probed with the anti-FLAG antibody. In the presence of the Myc antibody, a doublet of the same molecular weight as that of the TLE1 proteins is coimmunoprecipitated (Fig. 3C, lane 3). This doublet is not present when Protein G-Sepharose beads alone are incubated with nuclear extracts made from the transfected cells (Fig. 3C, lane  2). Thus, FLAGTLE1 can interact with MycPRH in hematopoietic cells.
Some apparent protein-protein interactions occur indirectly and are observed because two proteins both independently bind to contaminating DNA present in the nuclear extract. For example, Oct-2 and the nonspecific DNA binding subunits of the Ku protein co-immunoprecipitate only in the presence of DNA, whereas the interaction between Rb and E1A is not dependent on the presence of DNA (57). EtBr intercalates into DNA and can abolish protein-DNA interactions without disrupting protein-protein interactions. To determine whether the PRH-TLE1 interaction requires the presence of DNA, we carried out the co-immunoprecipitation experiment with MycPRH and FLAGTLE1 described above in the presence of different concentrations of EtBr. The direct protein-protein interactions between Rb and E1A are insensitive to EtBr concentrations of 0.2 mg/ml (57). Fig. 3C shows that TLE co-immunoprecipitates with MycPRH even in the presence of 0.4 mg/ml EtBr (Fig. 3C,  compare lanes 3 and 4). Since EtBr does not appear to significantly affect the co-immunoprecipitation of TLE1 and PRH, it is likely that this is a protein-protein interaction that does not require promoter DNA.
To establish whether transfected MycPRH interacts with endogenous TLE proteins, nuclear extracts were made from K562 cells alone or K562 cells transfected with MycPRH. Proteins in these nuclear extracts were separated on SDS-polyacrylamide gels and probed with the rat pan-TLE antibody raised against the C terminus of TLE proteins. In the presence of transfected MycPRH, nuclear TLE proteins with a molecular mass of ϳ90 kDa were detected in Western blotting experiments (Fig. 3D, lane 1). Nuclear extracts expressing transfected MycPRH were incubated with an anti-Myc mouse monoclonal antibody bound to Protein G-Sepharose beads (Fig. 3D,  lane 4) or incubated with Protein G-Sepharose beads alone (Fig. 3D, lane 3). As a further control, untransfected K562 nuclear extracts were incubated with an anti-Myc mouse monoclonal antibody bound to Protein G-Sepharose beads (Fig. 3D,  lane 2). After washing, the bound proteins were separated on SDS-PAGE gels and probed with the pan-TLE antibody. Small amounts of TLE proteins nonspecifically bound to Protein G-Sepharose beads are detected by the pan-TLE antibody even in the absence of the Myc antibody (Fig. 3C, lane 3) or in the absence of transfected MycPRH (Fig. 3C, lane 2). However, in the presence of transfected MycPRH, a doublet of the same molecular weight as that of the endogenous TLE proteins is strongly co-immunoprecipitated by the Myc antibody (Fig. 3D,  lane 4). We conclude that TLE proteins can associate with PRH both in vitro and in hematopoietic cells.
TLE Proteins Can Co-repress Transcription with PRH-To determine the biological significance of the interaction between PRH and TLE1, we investigated whether TLE1 could function as a co-repressor of transcription with PRH. We have shown previously that the reporter plasmid pTK-PRH is repressed by chicken PRH in avian hematopoietic BM2 cells (19) and by human PRH in K562 cells (22). The pTK-PRH reporter and the control reporter plasmid pSV-␤-gal were transiently transfected into K562 cells with 100 ng of pMUG1-MycPRH . As can be seen in Fig. 4A, this amount of pMUG1-MycPRH decreases pTK-PRH promoter activity to ϳ60% of its unrepressed activity (compare columns 1 and 2). Co-transfection with 200 ng of the TLE1 expression plasmid pCDNA3-TLE1 results in increased repression by PRH (Fig. 4A, compare columns 2 and 3). This effect, although small, is highly reproducible. Transfection of pCDNA3-TLE1 and the pTK-PRH reporter alone does not decrease reporter activity, and if anything there is an increase in promoter activity (Fig. 4A, compare columns 1 and 4). These data suggest that TLE1 can co-repress transcription, albeit weakly, with PRH in transient transfection assays.
There are a number of possible reasons why we might not observe very strong co-repression between PRH and TLE1 in transient transfection assays. K562 cells strongly express several TLE family members, 2 and one possibility is that trans-fection of the TLE1 plasmid into K562 cells does not strongly increase the amount of TLE1 proteins in the cells available to interact with PRH. A complementary approach to examine co-repression between PRH and TLE proteins in these cells is to remove endogenous TLE proteins and measure any effects on PRH activity. Several studies have shown that Grg5, a naturally occurring protein that is related to the first 200 amino acids of Groucho/TLE, can tetramerize with TLE proteins and function as a dominant negative regulator (32)(33)(34). We therefore examined whether we could relieve repression by PRH by using Grg5 to titrate out endogenous TLE proteins. Transfection of 100 ng of pMUG1-MycPRH decreases pTK-PRH promoter activity to ϳ60% of its unrepressed activity (Fig.  4B, compare columns 1 and 2). Co-transfection of Grg5 with PRH relieves repression by PRH in a dose-dependent fashion (Fig. 4B, lanes 3-5). In the presence of 5 g of pFLAG-Grg5, reporter activity returns to ϳ75% of its unrepressed level (Fig.  4B, lane 5). The presence of pFLAG-Grg5 did not similarly increase reporter activity in the absence of PRH (Fig. 4B, lanes  6 -8). In fact, 5 g of pFLAG-Grg5 decreases reporter activity in the absence of PRH to ϳ65-70% of its normal level (Fig. 4B,  lane 8). However, FLAG-Grg5 does not appear to significantly 2 T. E. Swingler and P.-S. Jayaraman, unpublished observations.

FIG. 4. PRH and TLE proteins co-repress transcription.
A, K562 cells were transiently transfected with 5 g of a luciferase reporter plasmid (pTK-PRH) containing five PRH binding sites upstream of the minimal TK promoter (columns 1-4). pMug1-MycPRH (100 ng) was co-transfected into the same cells either alone or together with pCDNA3-TLE1 (200 ng) (column 2 and 3, respectively). Luciferase activity was normalized for transfection efficiency using a co-transfected plasmid expressing ␤-galactosidase. The data are presented as promoter activity relative to the reporter alone, and the values represent the mean and S.E. of at least three experiments. B, K562 cells were transiently transfected with 5 g of pTK-PRH (columns 1-8). 100 ng of a PRH expression vector (pMug1-MycPRH) was co-transfected into the same cells (columns 2-5) with 0.5 g (column 3), 1 g (column 4), or 5 g (column 5) of the Grg5 expression vector pFLAG-Grg5. As a control, the same amounts of pFLAG-Grg5 were transfected into K562 cells in the absence of PRH (columns 6 -8). Luciferase activity is normalized and presented exactly as in A.
affect the activity of the viral CMV promoter that is used to express both MycPRH and FLAG-Grg5 (data not shown). Thus, repression of the TK-PRH promoter by PRH can be relieved by co-expression of Grg5. To examine whether Grg5 can itself interact with PRH in K562 cells, we carried out co-immunoprecipitation studies using pMycPRH and pFLAG-Grg5. We were unable to demonstrate co-immunoprecipitation of Grg5 and PRH, although in a parallel experiment, TLE1 and PRH were co-immunoprecipitated under wash conditions of the same stringency (data not shown). Thus, whereas we cannot rule out the possibility that Grg5 might bind directly to PRH and inhibit PRH-dependent repression, our experiments indicate that, most likely, the titration of TLE proteins by Grg5 results in decreased repression by PRH. We conclude that endogenous TLE proteins do co-repress with PRH and that overexpressed TLE1 can co-repress with PRH.
Mapping the TLE Interaction Motif in PRH-The human and avian PRH proteins are strongly conserved (1), and both proteins contain an Eh-1 motif within the first 60 amino acids of the PRH N terminus (Fig. 5A). To determine whether this motif is required for the interaction of TLE proteins with PRH, we made use of three avian GST-PRH N fusion proteins in in vitro binding assays. The N terminus of avian PRH (amino acids 1-141) or N-terminal fragments of avian PRH (amino acids 1-125 and 61-141) were expressed in bacteria as GST fusion proteins and purified. In vitro transcription and translation was used to produce labeled TLE1, and binding assays were carried out as described above. Fig. 5B shows that labeled TLE1 binds to the GST-PRH N1-141 (lane 3) and GST-PRH N1-125 (lane 5) fusion proteins but does not bind to GST-PRH N61-141 (lane 4) or to GST alone (lane 2). These data confirm the interaction data in yeast, which showed that the Runt/Hairy motif is not required for the PRH-TLE interaction. Furthermore, these data strongly suggest that the Eh-1 domain located within the first 60 amino acids of PRH is required for the interaction. To confirm these in vitro binding studies, we determined whether the purified GST-PRH N deletion proteins are able to bind to endogenous TLE proteins in pull-down experiments. Bound proteins were probed with pan-TLE rat monoclonal antibody or with a mouse monoclonal antibody raised against the C8 subunit of the proteosome. Thus, whereas all of the fusion proteins appear to be functional for protein-protein interactions with C8, the GST-PRH N fusion protein lacking the first 60 amino acids of PRH is unable to interact with endogenous TLE proteins.
To establish whether the Eh-1 motif in PRH is responsible for the interaction of PRH with TLE, we mutated phenylalanine 32 to glutamic acid (F32E) in the Myc-tagged human PRH mammalian expression plasmid pMUG1-MycPRH to create pMUG1-MycPRH F32E . This amino acid is the most conserved within the Eh-1 motif, and changing this amino acid to glutamic acid has previously been shown to abolish the interaction of the homeodomain proteins Goosecoid and Engrailed with Groucho (49,58). The wild type human PRH N terminus (amino acids 1-132) and the mutated PRH N terminus were transferred to a GST expression vector, and GST-PRHN 1-132 and GST-PRHN 1-132 F32E fusion proteins were expressed in bacteria and purified. In vitro binding assays were carried out with these fusion proteins and in vitro transcribed and trans-lated TLE1 as described above. Fig. 5D shows that labeled TLE1 binds to the human GST-PRH N1-132 (lane 3) but binds only very weakly, if at all, to GST-PRH N1-132 F32E (lane 4) or to GST alone (lane 2). Pull-down experiments were also carried out as described above with these fusion proteins and K562 nuclear extracts. Fig. 5E shows that endogenous TLE proteins bind to the human GST-PRH N1-132 (lane 2) but bind only very weakly, if at all, to GST-PRH N1-132 F32E (lane 3) or to GST alone (lane 1). In contrast, Fig. 5E shows that proteosome subunit C8 binds to GST-PRH N1-132 (lane 2) and GST-PRH N1-132 F32E (lane 3) but binds only very weakly, if at all, to GST alone (lane 1). Thus, we conclude that the Eh-1 domain in PRH mediates the interaction between PRH and TLE proteins present in K562 cells.  1-4). pMug1-MycPRH F32E (100 ng) was co-transfected into the same cells either alone or together with pCDNA3-TLE1 (200 ng) (lanes 2 and 3, respectively). Luciferase activity is normalized and presented exactly as in Fig. 4.
To confirm that the Eh-1 domain mediates the interaction between PRH and TLE1 that occurs during co-repression of the TK promoter, we carried out transient co-transfection assays. The pTK-PRH reporter and the control reporter plasmid pSV-␤-gal were transiently transfected into K562 cells with 100 ng of pMUG1-MycHPRH F32E. Fig. 5F shows that pMUG1-MycHPRH F32E decreases pTK-PRH promoter activity to ϳ60% of its unrepressed activity (compare columns 1 and 2). However, co-transfection of 200 ng of the TLE1 expression plasmid pCDNA3-TLE1 with pMUG1-MycHPRH F32E does not result in increased repression by PRH F32E (Fig. 5F, compare columns 2  and 3). These data suggest that the Eh-1 domain mediates the interaction between PRH and TLE1 in vivo. DISCUSSION We have shown previously that the PRH N-terminal region contains a transcriptional repression domain with potential Groucho/TLE interaction motifs (19). Here we have demonstrated that PRH interacts with TLE proteins in hematopoietic cells, in yeast and in vitro, and that an Eh-1 motif in the PRH N-terminal region mediates the interaction of these proteins. It has been documented previously that Grg5 binds to TLE proteins and functions as a trans-dominant negative (32)(33)(34). We have shown that titration of endogenous TLE proteins by Grg5 results in decreased transcriptional repression by PRH, suggesting that PRH and TLE associate in cells. Furthermore, overexpression of TLE1 increases repression by PRH but does not increase repression by a PRH protein containing a mutation in the Eh-1 motif that eliminates binding to TLE1 in vitro. Taken together, these observations strongly suggest that PRH and TLE proteins interact in K562 hematopoietic cells and that TLE proteins function as co-repressors for PRH. It is of interest to note that the PRH F32E mutation does not significantly decrease the ability of PRH to repress transcription in transient transfection assays. This situation is similar to that reported by Tolkunova et al. (48) for Engrailed (En). En has multiple repression domains and repression mechanisms. En interacts with the Groucho co-repressor and the interaction is essential for full repression in Drosophila embryos but the Eh-1 F to E mutation in En resulted in less than a 10% reduction of En repression activity in standard transient transfection assays (48). Like En, PRH has several repression mechanisms, and therefore the full effect of this mutation might only be clearly seen in genomic DNA in the natural in vivo context. Groucho/TLE-dependent repressors have been found to function as long range repressors (i.e. they block promoter function in a distance-and orientation-independent manner). The ability to recruit TLE proteins to a promoter represents an important mechanism for PRH-dependent transcriptional repression. There are at least two domains in Groucho/TLE proteins that are known to interact with DNA-bound transcription factors: the N-terminal Q domain and the C-terminal WD repeat domain. The Q domain (amino acids 1-135) is a tetramerization domain that is essential for transcriptional repression (23,30) and sufficient for the interaction of TLE with the transcription factors T-cell factor and Blimp-1 (PRD1-BF1) (32). In contrast with the interaction of TLE proteins with T-cell factor and Blimp-1, the WD repeat domain in Groucho is essential and sufficient for direct interactions with the Engrailed and Hairy transcription factors (59). However, the Q and WD repeat domains in TLE1 are both involved in interactions of TLE proteins with the Runt-related transcription factor RUNX2/Cbfa1 and the winged-helix protein BF-1 (60,61). Similarly, we have shown that the PRH N-terminal domain interacts with both the Q domain and the WD repeat domain in TLE1. It has been suggested that the employment of several TLE protein-protein interactions domains allows greater specificity of interaction (62). Alternatively, the use of more than one TLE proteinprotein interaction domain might confer greater stability to a TLE-PRH complex. Certainly, the stable interaction of another homeodomain protein, Pax5, with TLE proteins requires two separate domains in TLE1 and two separate domains in Pax5 (63).
Although the TLE proteins were not originally identified in the hematopoietic compartment, it has become apparent that TLE proteins interact with a number of transcription activator proteins that are found in hematopoietic cells, including the human Runt domain protein, AML1 (also known as RUNX-1/ CEBP-2/PEBP2) (27,64), T-cell factor (also known as lymphoid enhancer factor (LEF-1)) (64), and the B-cell-specific factors PRDI-BF1 and Pax5 (32,63). In each case, the interaction of the transcription activator with TLE proteins results in transcriptional repression. Thus, TLE proteins are likely to be co-repressors for PRH in a number of hematopoietic lineages. In addition, both PRH and TLE are expressed in a variety of tissues in the developing embryo. PRH is expressed in the pregastrulation embryo and is one of the earliest markers for dorsoventral patterning (4). Later, in the early embryo, PRH is expressed in the anterior endoderm (head organizer region) adjacent to tissues expressing the Goosecoid homeodomain protein (15), which co-represses transcription with TLE proteins (49). Interestingly, PRH is essential for forebrain development (16), and TLE proteins have been implicated in neuronal differentiation (65) and dorsoventral patterning of the neural tube (66). Expression of PRH has also been detected in endothelial precursor cells (4,67) and in osteoblasts (3). TLE expression occurs postgastrulation in mesoderm derivatives that go on to produce the endothelial cells of the lining of the heart, muscle, and bone (25). In Drosophila, Groucho is part of the Notch signaling cascade. It is believed that the role of this signaling pathway is to bring about a halt in embryonic neurogenesis so that cells that were committed to become neuroblasts are made competent to enter the epidermal lineage instead (68). Studies with TLE proteins have shown that they play a similar developmental role (25). TLE expression is elevated in undifferentiated or transformed epithelial cells and down-regulated as epithelial cells differentiate. This suggests that TLE proteins are involved in the maintenance of the undifferentiated state (21). Significantly, Notch signaling also plays a fundamental role in hematopoietic development; in general, Notch signaling promotes self-renewal and inhibits differentiation (54). PRH is found in hematopoietic and endothelial progenitors (3, 4) and a variety of tissues derived from endoderm including thyroid and liver (5). However, PRH is absent in terminally differentiated hematopoietic cells and/or endothelial cells (3,4). Furthermore, our studies on myeloid stem cells have shown that PRH expression is down-regulated during differentiation of early myeloid progenitors toward myeloblasts or erythrocytes (6). Thus, PRH expression is associated with the undifferentiated state in hematopoietic and endothelial cells. Here we have shown that undifferentiated hematopoietic blasts contain nuclear PRH and TLE proteins. It is tempting to speculate that the interplay of TLE proteins and PRH in these cells contributes to the control of cell differentiation. Further work will be required to reveal in what other cell types TLE proteins act as co-repressor proteins for PRH and the importance of the TLE1-PRH interaction in differentiation and development.