The LIM-only Protein DRAL/FHL2 Interacts with and Is a Corepressor for the Promyelocytic Leukemia Zinc Finger Protein*

Members of the four-and-a-half-LIM domain (FHL) protein family, which are expressed in a tissue- and stage-specific manner, have been reported previously to function as transcriptional coactivators. One of these is the p53-inducible protein DRAL/FHL2 (where DRAL is down-regulated in rhabdomyosarcoma LIM domain protein). In this work, we identified potential binding partners for DRAL/FHL2 using an inducible yeast two-hybrid system. We present evidence of a functional interaction between the promyelocytic leukemia zinc finger protein (PLZF) and DRAL/FHL2. PLZF is a sequence-specific transcriptional repressor whose function relies on recruitment of corepressors that form part of the histone deacetylase complex involved in chromatin remodeling. DRAL/FHL2 interacts specifically with PLZFin vitro and in vivo and augments transcriptional repression mediated by PLZF. This is the first reported incidence of a bona fide FHL protein-mediated corepression and supports the notion of these proteins having a role as coregulators of tissue-specific gene expression.

Many cancers occur in association with chromosomal translocations that frequently involve genes coding for transcription factors. This structural interference results in a downstream disruption of crucial regulatory pathways such as may be involved in the control of growth, differentiation, and survival of normal cells. Regulation of such transcription factors often occurs via chromatin remodeling in that these proteins can themselves interact with components of the transcription complex such as the corepressors (NCoR/SMRT) 1 or with histone deacetylases (HDACs). In a variant form of acute promyelocytic leukemia, the promyelocytic leukemia zinc finger protein (PLZF) is fused to the retinoic acid receptor-␣ as a result of a chromosomal translocation, t(11,17)(q23;q21); this form is re-fractory to retinoic acid treatment as a result of the association of PLZF with HDAC and the corepressors NCoR/SMRT (1,2). PLZF is a sequence-specific DNA-binding transcriptional repressor, bearing nine Krü ppel-like C 2 H 2 zinc fingers at the C terminus. The N terminus contains a POZ domain, which allows self-association as well as heterodimerization with other proteins. A second repression domain, RD2, is located downstream of the POZ domain (3,4). PLZF is differentially regulated during embryogenesis and in adult tissues (5), where it is thought to function as a growth suppressor, through transcriptional repression of target genes.
LIM domains represent a type of protein interaction domain and are cysteine-and histidine-rich motifs of ϳ50 amino acids that form two specialized zinc fingers. To date there is no evidence that LIM domains can interact with nucleic acids despite their characteristic structure, but rather it is patently clear that they mediate protein-protein interactions (6 -8). LIM domain proteins can homodimerize or heterodimerize (9, 10) as well as command an ability to interact with other protein motifs such as PDZ domains (11), ankyrin repeats (12), and helix-loop-helix domains (13,14). LIM-containing proteins have variously been localized to the nucleus and cytoplasm, with many being found in association with the cytoskeleton (12,15,16). Several subclasses of LIM domain proteins are recognized and designated according to the presence or absence of additional motifs such as homeodomains or kinase domains (6,7). One of these subgroups is the LIM-only family, which possess only LIM domains. A number of LIM-only proteins are clearly implicated in transcriptional regulation: LMO2, which interacts with GATA-1 and Tal1 and plays a definitive role in angiogenesis (17,18); and MLP, which interacts with the myogenic transcription factor MyoD and is required for the correct development of cardiac cytoarchitecture (14,19,20).
Within this subgroup is a cohesive family of LIM-only proteins, which are highly homologous, are characterized by a specific arrangement of domains, bearing four complete and one N-terminal half LIM domain (FHL family), and are expressed tissue-specifically and in distinct cellular compartments. Recently, members of the FHL family were also shown to behave as transcriptional coactivators. ACT functions as a coactivator of CREM (21,22), and DRAL/FHL2 has been reported to enhance the transcriptional activity of the androgen receptor (23). Additionally, an isoform of FHL1 known as KyoT2 interacts with RBP-J, a DNA-binding transcription factor, thus negatively regulating transcription (24). Accordingly, it is tenable that this family of FHL proteins may perform a function in transcriptional modulation.
DRAL/FHL2, the first FHL protein described, was isolated in our laboratory by virtue of its being down-regulated in rhabdomyosarcoma cells as compared with their non-malignant equivalent, normal human myoblasts (25). We have since shown that DRAL/FHL2 expression is regulated by p53 since mRNA levels are augmented by transient expression of functional p53 in rhabdomyosarcoma cells, as well as by endogenous p53 stimulated by ionizing radiation treatment. Moreover, overexpression of DRAL/FHL2 in both normal and tumorderived cell lines efficiently induces an apoptotic program (26). Therefore, it is conceivable that DRAL/FHL2 has a role in tumor development.
In this work we present the results of a yeast two-hybrid screen with DRAL/FHL2, performed to further delineate the molecular purpose of this protein. It is noteworthy that DRAL/ FHL2 was isolated in several two-hybrid screens and is an interaction partner for the androgen receptor (23), presenilin-2 (27), CDC47 (28), several ␣and ␤-integrin cytoplasmic tails (29), hNP220 (30), and IGFBP5 (31). Here we demonstrate a physical and functional interaction between DRAL/FHL2 and the promyelocytic leukemia zinc finger protein (PLZF). DRAL/ FHL2 acts as a corepressor for PLZF-mediated transcriptional repression, supporting the notion that DRAL/FHL2 is a tissuespecific transcriptional modulator.

EXPERIMENTAL PROCEDURES
DNA Constructs-The full-length cDNA for human DRAL/FHL2 was cloned in-frame into the inducible yeast expression vector pGILDA (LexA Matchmaker Two-hybrid System, CLONTECH), as a fusion to the LexA DNA-binding domain (DNA-BD) by PCR using primers with a 5Ј BamHI site and a 3Ј NotI site. The following DRAL/FHL2 deletion constructs were generated by PCR also with a 5Ј BamHI site and 3Ј NotI site, as well as containing a Kozak consensus start and a Cterminal FLAG tag, and cloned into pcDNA3.1 (Invitrogen): LIM (1)(2)(3)(4) eliminating the N-terminal half LIM domain; LIM(0.5-2) the two and a half N-terminal LIM domains; LIM(3-4) the two C-terminal LIM domains; LIM1, LIM2, LIM3, and LIM4, each complete single LIM domain numbered from the N to C terminus. A similar construct was designed using the full-length DRAL/FHL2 cDNA. The primers used to manufacture these constructs were designed such that they were compatible with the mammalian two-hybrid system expression vector pBIND (Checkmate, Promega Corp., Madison, WI) and so could be used to generate in-frame fusions of DRAL and deletion constructs thereof, with the GAL4 DNA-BD of pBIND. The full-length PLZF cDNA was amplified from pSG5-PLZF (32) using BamHI 5Ј-and NotI 3Ј-modified PCR primers and cloned into the mammalian two-hybrid system VP16activation domain (AD) expression plasmid pACT (Promega Checkmate). pcDNA3.1 myc/his-PLZF and deletion constructs of PLZF, del-POZ (encoding amino acids 121-673), delRD2 (lacking amino acids 199 -313), and delPOZ/delRD2 (encoding amino acids 122-199 and 314 -673) were described previously (4). pGEX-3X-DRAL was described previously (25) as were the IL3R-tk-luc and tk-luc reporters (33). IMAGE clones corresponding to human FHL1, FHL3, ACT, CRP2, and PINCH were obtained from the UK HGMP Resource Centre and transferred into pcDNA3. The integrity of all constructs was confirmed before use. Specific details of cloning and PCR primers are available upon request.
Yeast Two-hybrid Assay-A human adult cardiac cDNA library in the LexA system-inducible B42-activation domain (pB42-AD) yeast expression vector was acquired from CLONTECH and amplified according to the manufacturer's directives. The library was transformed into yeast strain EGY48 to screen for interaction partners with pGILDA-DRAL, as per the manufacturer's protocols. Yeast plasmids were prepared from colonies sustaining growth on leucine-deficient medium and that were positive for ␤-galactosidase activity. By using primers designed to target the pB42-AD plasmid inserts, relevant inserts were amplified and partially sequenced. Approximately 200 putative positive clones were selected and sequenced. Escherichia coli KC8 was then transformed with chosen yeast plasmids, and the library plasmid was nutritionally selected. Isolation of the correct plasmid clone was confirmed by re-sequencing. To confirm specific two-hybrid interactions, small scale transformations were performed and assayed as per instruc-tions. Liquid ␤-galactosidase assays were performed using O-nitrophenyl ␤-D-galactopyranoside (Sigma) as substrate, as described by the manufacturer.
GST Pulldown-Expression of full-length DRAL/FHL2 as a GST fusion protein from pGEX-3X (Amersham Biosciences) was performed as described previously (25). 35 S-Labeled proteins were produced using Promega's TNT coupled in vitro transcription-translation (IVT) system, using the following constructs all in pcDNA3.1-myc/his: full-length PLZF, delPOZ, delRD2, and delPOZ/delRD2. 5 l of the IVT reactions were incubated with 1 g of GST-DRAL bound to glutathione-Sepharose in NET-80 buffer (80 mM NaCl, 20 mM Tris, 1 mM EDTA) and mixed for 3 h at 4°C. Beads were then washed extensively in NET-80, resuspended in 30 l of SDS-PAGE gel loading buffer, and resolved on a 10% polyacrylamide gel followed by autoradiography.
A U937 monocytic cell line expressing PLZF under the control of the tetracycline repressor (34) was grown in the presence or absence of tetracycline. These cells were harvested with phosphate-buffered saline at 4°C and exposed to lysis buffer (150 mM NaCl, 20 mM Tris-Cl, pH 7.4, Tween 20, plus protease inhibitors) on ice for 15 min. This suspension was centrifuged at 6000 rpm for 10 min at 4°C. The supernatant was pre-cleared by exposure to beads bound to a nonspecific rabbit IgG (Zymed Laboratories Inc., San Francisco) and mixed for approximately 1 h at 4°C. The beads were then pelleted and exposed to PLZF antibodies (IgG 2a isotype) (35) covalently bound to protein A-Sepharose and mixed on a rotator overnight. The pellets were washed in fresh cold lysis buffer six times, the last three times using Nonidet P-40 instead of Tween 20 in the lysis buffer. The beads were then placed in Laemmli buffer and the precipitated proteins released by boiling. This was followed by electrophoresis through a 12% SDS-polyacrylamide gel and transfer to an Immobilon-P membrane (Millipore, Bedford, MA). The blot was incubated with a 1:1000 dilution of rabbit anti-DRAL/FHL2 antibody followed by a 1:3000 dilution of horseradish peroxidase-conjugated anti-rabbit secondary antibody (Roche Molecular Biochemicals). Autoradiography was performed using the ECL chemiluminescence kit (Amersham Biosciences).
Northern Analysis-Northern blotting and hybridization were carried out as described previously (26) on total RNA isolated from normal human tissues from autopsy material. cDNAs of DRAL/FHL2, PLZF, and ␤-actin were used as probes.
Mammalian Two-hybrid Assays-The full-length cDNA for PLZF was cloned in-frame into pACT, the mammalian two-hybrid system VP16 activation domain expression vector (Promega Checkmate System). 293T cells were plated at a density of 3 ϫ 10 5 cells/60-mm dish. The following day 1 g each of pBIND-DRAL and pACT-PLZF were transfected, together with 1 g of the GAL4-responsive reporter pG5luc. pRSV-␤-Gal was included in all transfections at a concentration of 100 ng/dish to correct for transfection efficiency. After 48 h of incu-bation, cells were washed twice with phosphate-buffered saline, harvested in reporter lysis buffer (Promega), and assayed for ␤-galactosidase activity (Promega ␤-Galactosidase Enzyme Assay System) and luciferase activity (Promega Luciferase Assay System). Similar experiments were performed using the DRAL/FHL2 and PLZF deletion constructs as described under "Results." Results are presented as fold activation of normalized luciferase activity in the presence of DRAL/ FHL2 and/or PLZF compared with normalized activity in the presence of empty vectors and are the average of at least four independent experiments. Western blotting of cell extracts was performed using anti-FLAG M2 (Sigma) at 1:1500 dilution, followed by alkaline phosphatase-conjugated goat anti-mouse IgG (Sigma), and developed using Tropix Western Light kit (Applied Biosystems).
Transcriptional Repression Assays-A reporter construct containing four high affinity PLZF-binding sites found in the IL3R␣ chain promoter was used to measure transcriptional activity (IL3R␣-tk-luc), and a tk-luc construct without the specific PLZF-binding sites was used as negative control. 3 ϫ 10 5 293T cells were seeded in 60-mm dishes and transfected the following day. Luciferase activity was measured after 48 h. Transfection efficiency was normalized using ␤-galactosidase activity produced by pRSV-␤-gal included in all transfections. Total DNA transfected was the same for all samples using empty vector DNA. Results are presented as fold repression of normalized luciferase activity in the presence of PLZF and/or effectors compared with normalized activity in the presence of empty expression vectors. Results are the average of four individual experiments. For inhibition of histone deacetylase, cells were transfected as described in the presence of 6 mM sodium butyrate. For repression assays in NIH3T3 cells, 6 ϫ 10 5 cells were seeded in 60-mm dishes in Dulbecco's modified Eagle's medium, 0.5% FCS and transfected 20 h later with pcDNA3-DRAL; post-transfection cells were maintained in 0.5% FCS. 24 h later pSG5-PLZF and the IL3R␣-tk-luc reporter were transfected. Following this second transfection cells were given either 10 or 0.5% FCS for 3 h as indicated in the figure legend and were thereafter assayed for luciferase and ␤-galactosidase activity. Western blotting of cell extracts was performed using anti-FLAG M2 (Sigma) at 1:1500 dilution, followed by alkaline phosphatase-conjugated goat anti-mouse IgG (Sigma), and developed using Tropix Western Light kit (Applied Biosystems).

Identification of Putative Protein Interaction Partners for
DRAL/FHL2-In order to identify potential binding partners for DRAL/FHL2 as well as to further delineate its molecular role, we have exploited the yeast two-hybrid system. A human adult cardiac cDNA library was selected for screening since we and others have shown previously (23,25,26,(37)(38)(39)(40) that expression of DRAL/FHL2 is highest in heart muscle, and specifically the protein localizes to diverse structures such as the nucleus, focal contacts, and the Z-discs and M-bands of cardiac myofibrils (26). An inducible LexA-based two-hybrid system was used, and ϳ1 ϫ 10 7 independent clones were screened yielding a large number of putative positive clones upon selection for interacting transformants. Based on its described growth-suppressive properties, one of these was selected for further characterization, namely the promyelocytic leukemia zinc finger protein, PLZF (41,42). To confirm the DRAL/FHL2-PLZF interaction, LexA-DRAL/FHL2 and B42-PLZF yeast plasmids were tested for their in vivo interaction in a liquid ␤-galactosidase assay (Table I). PLZF fused to the B42-activation domain interacts with DRAL/FHL2 fused to the LexA DNA-binding domain, thus increasing ␤-galactosidase reporter activity to a level comparable with that of the positive control LexA DNA-BD-p53 and B42 AD-SV40 large T antigen, whereas either plasmid alone or in combination with SV40 large T antigen or p53, respectively, did not result in ␤-galactosidase activity ( Table I).
The association of DRAL/FHL2 with PLZF was further corroborated using a monoblastoid cell line, U937, expressing PLZF in a tetracycline-repressible manner (34); only when PLZF expression is induced can DRAL/FHL2 also be precipitated by the anti-PLZF antibody (Fig. 1, IP Ϫtet), suggesting an interaction between DRAL/FHL2 and PLZF both in the yeast two-hybrid system as well as in mammalian cells.
Specific Domains of PLZF Are Required for the DRAL/FHL2-PLZF Interaction-At this juncture, having demonstrated an interaction between the two proteins, it was instructive to attempt to define the domains required for the interaction. Therefore we first performed an in vitro protein binding assay. 35 S-Labeled in vitro translated PLZF or deletion constructs thereof ( Fig. 2A) were incubated with GST-DRAL/FHL2 recombinant protein (25) immobilized on glutathione-Sepharose beads. Bound complexes were washed extensively, prior to elution and resolution by SDS-PAGE. This GST-pulldown assay provided further credence to the interaction between the full-length proteins (Fig. 2B, lane 5). Deleting the POZ domain of PLZF does not rescind the interaction with GST-DRAL/ FHL2 (Fig. 2B, lane 7), and this is also true when the RD2 domain is deleted (Fig. 2B, lane 8); however, deletion of both of these domains drastically reduces the ability of the protein to complex with GST-DRAL on the glutathione beads (Fig. 2B,  lane 6). This would imply that either the POZ or RD2 domains of PLZF interact with DRAL/FHL2, and deletion of both abrogates this interaction.
Association of the two proteins in vivo was further supported by coimmunoprecipitation experiments. Human kidney carcinoma cells were transfected with full-length PLZF and DRAL/ FHL2 expression constructs; the cell lysates were then immunoprecipitated with a monoclonal PLZF antibody and immunoblotted with a polyclonal DRAL/FHL2 antibody. DRAL/FHL2 was immunoprecipitated by the full-length PLZF protein only when both specific antibodies were used (Fig. 2C, lane 1) but not with preimmune serum (Fig. 2C, lane 2). We then performed additional coimmunoprecipitation experiments  with lysates from 293T cells transfected with the different PLZF constructs. As shown in Fig. 2C, DRAL/FHL2 can also be immunoprecipitated by the PLZF constructs del-POZ and del-RD2 (Fig. 2C, lanes 3 and 5, respectively). The band shown in Fig. 2C, lane 3, is faint but clearly visible upon longer exposure (data not shown). In concurrence with the GST pulldown assay, deleting both the POZ and RD2 domains from PLZF abrogates binding by DRAL/FHL2 (Fig. 2C, lane 7).
DRAL/FHL2 Interaction with PLZF in Vivo Requires an Intact Protein-In order to probe further the integrity of the putative DRAL/FHL2-PLZF interaction in vivo, a mammalian two-hybrid system was utilized. 293T cells were transfected with the full-length DRAL/FHL2 bait construct as a fusion to the GAL4 DNA binding domain, and this activated the reporter up to 9.25-fold when in the presence of the full-length PLZF prey construct, fused to the VP16 activation domain (Fig. 3B,  lane 4), whereas the positive control consisting of Id and MyoD, exploiting two known interacting transcription factors, showed approximately a 16-fold increase of reporter activity (Fig. 3B,  lane 1). The autoactivation control for DRAL/FHL2 shows a negligible level of reporter activation at 1.4-fold (Fig. 3B, lane  3), whereas PLZF does not show any intrinsic activity (Fig. 3B,  lane 2). The same results were also obtained in a different cellular background with HT1080 cells suggesting that the interaction observed in this assay is independent of other tissue-specific cofactors (data not shown).
In order to reveal the functional domains of DRAL/FHL2 that mediate the interaction with PLZF, the same mammalian two-hybrid assay was used with various deletion constructs of DRAL/FHL2 (Fig. 3A). As shown in Fig. 3C, it is apparent that removing the N-terminal half LIM domain of DRAL/FHL2 has minimal effect on the avidity of the binding between DRAL/ FHL2 and PLZF, reducing the activity from 14-to 9.7-fold (Fig.  3C, lanes 2 versus 3). However, the constructs LIM(0.5-2) and LIM (3)(4) show a very reduced activity when cotransfected with full-length PLZF, namely only 3.6-and 1.1-fold, respectively (Fig. 3C, lanes 4 and 5), which is not related to differences in expression levels (Fig. 3E, upper panel). This would suggest that either the relevant LIM domain or domains required to mediate the interaction are internal or, alternatively, that any manipulation of the protein in any permutation eliminates the possibility of a full interaction. To examine these hypotheses, we explored the effect of single LIM domain constructs transfected together with the full-length PLZF fusion protein. No activity is observed with these single LIM constructs (Fig. 3D, lanes 3-6), despite their being expressed at similar levels (Fig. 3E, lower panel), suggesting that no single LIM domain can mediate the interaction, but instead the integrity of the complete protein seems to be required.
PLZF-mediated Transcriptional Repression Is Enhanced by DRAL/FHL2-Next we wanted to examine the possible functional significance of the observed interaction. As PLZF is a known transcriptional repressor (3,33,42), and since other proteins that interact with PLZF have proven to function in the manner of a corepressor (4), we determined if DRAL/FHL2 might have an analogous function. A reporter system composed of four high affinity PLZF-binding sites from the IL3R␣ chain promoter upstream of a minimal promoter and a luciferase cassette has been described previously (4) (see Fig. 4A); when this construct is expressed in 293T cells, a particular level of luciferase activity is observed, which can be transcriptionally repressed by inclusion of a PLZF expression construct in the system, as shown in Fig. 4B, lane 3 (ϳ2.9-fold repression). Addition of DRAL/FHL2 expressed from a cytomegalovirus promoter results in an enhancement of this repression up to 5.6fold (Fig. 4B, lane 4), whereas use of DRAL/FHL2 alone has no effect on reporter activity (data not shown). This potentiation of repression was not observed when a reporter lacking the PLZFspecific binding sites was used (Fig. 4C), indicating that the effect is specific and dependent on PLZF binding. However, titration experiments revealed that a higher concentration of  1 and 2), the deletion constructs delPOZ (lanes 3 and 4), delRD2 (lanes 5 and 6), or delPOZ/delRD2 (lanes 7 and 8) were performed in 293T cells as described under "Experimental Procedures." Immunoprecipitation (IP) was done with anti-PLZF antibody and immunoblotting (IB) with anti-DRAL/FHL2-specific antibody (lanes 1, 3, 5, and 7) or preimmune serum (lanes 2, 4, 6, and 8). Lanes 9 -12 represent the input lysate control of lanes 1, 3, 5, and 7. DRAL/FHL2 with a constant amount of PLZF results in a lower level of repression, with a peak repression level of 5.7-fold using 400 ng of DRAL/FHL2 DNA (Fig. 4D, lanes 4 -7). Expression of DRAL/FHL2 in the absence of PLZF has no effect on reporter activity (Fig. 4D, lane 10).
It has been established that PLZF mediates transcriptional repression through its ability to recruit corepressors such as the NCoR, which in turn can then recruit the HDAC complex; furthermore, PLZF itself can interact with HDAC1 and -2 (42,43). To test for the specificity of the observed repression, we included an HDAC inhibitor, namely sodium butyrate, throughout the transfection period. As expected, this eliminated PLZF-mediated repression regardless of the addition of DRAL/FHL2 (Fig. 4D, lanes 8 and 9), suggesting that augmented repression mediated by DRAL/FHL2 is also depending on the formation of a HDAC complex. To exclude any potential effects of cell background, all results have been duplicated in human fibrosarcoma cells, HT1080 (data not shown).
To investigate domains of DRAL/FHL2 required to mediate the observed corepression activity, deletion constructs were coexpressed with PLZF in the repression assay. This revealed that deleting the N-terminal LIM domain has no effect on the corepressor function of DRAL/FHL2 (Fig. 4E, lanes 2 and 3). However, the two constructs LIM(0.5-2) and LIM (3)(4) do not significantly enhance PLZF repressor activity (Fig. 4E, lanes 4  and 5), despite showing comparable expression levels (Fig. 4F,  lanes 3 and 4). Hence, corepressor activity of DRAL/FHL2 requires an intact protein, supporting the conclusion from the interaction assay. In both experimental settings, however, the N-terminal half LIM domain is dispensable for activity.
DRAL/FHL2 can be localized to various structures in the cytoplasm such as focal adhesions as well as the nucleus (26). Recently it has been shown that nuclear accumulation can be enhanced through serum stimulation (47). Therefore, we wanted to investigate whether transcriptional corepression by DRAL/FHL2 can be further enhanced when DRAL/FHL2 is predominantly localized in the nucleus. Accordingly, we transfected NIH3T3 cells with the IL3R␣-tk-luc reporter, full-length PLZF, and full-length DRAL. After transfection and serum starvation, cells were treated for 3 h with serum and transcriptional repression analyzed. While no repression was observed without PLZF (Fig. 5A, lanes 1 and 2), serum stimulation also had no effect on PLZF repression (Fig. 5A, lanes 3 and 4). In contrast, a marked increase in corepression by DRAL/FHL2 (Fig. 5A, lanes 5 and 6; 7.7-versus 3.3-fold) was observed which was indeed paralleled by nuclear accumulation of DRAL/FHL2 (Fig. 5B). These experiments further support the corepressor activity of nuclear DRAL/FHL2. individual LIM domains (22). Thus we examined other FHL proteins, namely FHL1, FHL3, and ACT, for their ability to act as cofactors in PLZF-mediated repression. Cotransfection of FHL3 with PLZF also stimulates repression of the reporter to a level comparable with that obtained with DRAL/FHL2 (Fig.  6A, lane 3 versus 2), increasing repression from 1.7-fold with PLZF alone to 2.6-fold with the combination. No corepression was seen when ACT was cotransfected with PLZF (Fig. 6A,  lane 4) nor when FHL1 was included in the assay (Fig. 6A, lane  5). Consequently, this suggests that corepression is specific to some members of the FHL family, namely DRAL/FHL2 and FHL3.

Specificity of DRAL/FHL2 as a Cofactor for PLZF-mediated
Having established that only FHL3 may substitute for the corepressor function of the DRAL/FHL2, we wanted to test other LIM-only proteins in order to further demarcate specificity. Three LMOs were chosen: MLP, which is involved in myogenesis (14,19,20); CRP2, which is related to MLP but expressed in smooth muscle (45); and PINCH, which interacts with integrin-linked kinase (12). When each of these were cotransfected with PLZF, no additional repression of the reporter was detected (Fig. 6B, lanes 3-5, respectively). Accordingly, we concluded that significant functional repression of PLZF targets is restricted to DRAL/FHL2 and FHL3 only.
Coexpression of DRAL and PLZF-Specific expression of DRAL/FHL2 has been assigned previously to cardiac muscle in terms of both mRNA and protein (23,26,39), and PLZF also presents expression in mouse heart (5). To corroborate these findings as well as demonstrate endogenous coexpression of DRAL/FHL2 and PLZF, we analyzed expression of these proteins in a small panel of human tissues with a Northern blot (Fig. 7A) and immunostaining of adult rat cardiomyocytes (Fig. 7B). Northern blotting indeed shows expression of both DRAL and PLZF in both skeletal muscle and heart (Fig. 7A). Immunohistochemical analysis of adult rat cardiomyocytes shows nuclear expression of PLZF (Fig. 7B, panel a), whereas DRAL is found both in the same nuclei, as well as at the Z-discs and M-bands of the cardiomyofibrils (26) (Fig. 7B,  panel b), suggesting that coexpression does in fact occur, at least in cardiomyocytes. DISCUSSION While many of the primary components involved in chromatin remodeling have been recognized, there remain additional cofactors that must work in concert with transcription factors to achieve spatio-temporal variations in and control of gene expression. Here we present evidence of a functional interaction between DRAL/FHL2 and PLZF, a transcriptional repressor with a role in the control of cellular proliferation and Hox gene regulation (41,42,46). The interaction has been validated by a number of in vitro and in vivo assays, including yeast two-hybrid, mammalian two-hybrid, in vitro protein binding, and coimmunoprecipitation. More importantly, we can demonstrate that DRAL/FHL2 functionally enhances PLZF-mediated repression, thus propounding the notion that it may represent a novel class of tissue-specific corepressor.
From the yeast two-hybrid screen using DRAL/FHL2 as bait, one clone was selected that upon partial sequencing was found to correspond to PLZF. Analysis of the incomplete cDNA clone revealed that it began at amino acid 100, which eliminates the POZ/BTB domain of this protein. More detailed analyses suggest that either the POZ and/or RD2 domain of PLZF is important for this interaction, since deletion of either domain alone does not affect the interaction with DRAL/FHL2 (Fig. 2, B and C), whereas deletion of both domains simultaneously abrogates binding of recombinant DRAL/FHL2 in vitro (Fig. 2B) as well as preventing precipitation of DRAL/FHL2 by the PLZF antibody in vivo (Fig. 2C). On the other hand, it appears that the full complement of LIM domains of DRAL/FHL2 is necessary for an interaction, as implied by the mammalian two-hybrid experiments (Fig. 3). The only small domain that could be deleted without affecting the interaction was the N-terminal half LIM domain. This domain is also not required for the assayed DRAL/FHL2 function and therefore remains uncharacterized. It has, however, recently been shown that the Nterminal half LIM domain is involved in Rho signaling (47).
PLZF is a known transcriptional repressor by reason of its ability to recruit corepressors such as NCoR/SMRT, Sin3A, and histone deacetylases that together form part of the Sin3A/B complex involved in chromatin rearrangement. It binds to target gene promoters, such as that of the IL3R␣ chain (33) and cyclin A (42) by means of its C-terminal zinc fingers. Here we have shown that DRAL/FHL2 can enhance transcriptional repression mediated by PLZF (Fig. 4) and thus can also be considered as a corepressor.
Recently, it was shown that transcriptional activation by DRAL/FHL2 occurs concomitantly with its translocation to the nucleus as a result of stimulation of the Rho signaling pathway (47). Serum stimulation of NIH3T3 cells results in an augmentation of DRAL/FHL2-enhanced corepression of PLZF target genes (Fig. 5A), which is associated with nuclear translocation (Fig. 5B), and thus may also involve a similar signaling mechanism.
In order to examine the specificity of DRAL/FHL2-governed corepression of PLZF target genes, we employed other members of the FHL family of LIM-only proteins in the IL3R␣ functional assay. To our surprise, corepression was not entirely specific to DRAL/FHL2 but was in fact shared by FHL3, which displayed a similar level of activity (Fig. 6A). Other FHL proteins assayed, ACT and FHL1, did not manifest any corepressor activity (Fig. 6A) nor did other LIM-only proteins such as MLP, CRP2, and PINCH (Fig. 6B). Thus, the ability to corepress transcription is exclusive to a tissue-specifically expressed subset of FHL proteins. Interestingly, FHL3, apart from being structurally related to DRAL/FHL2, also shares a similar pattern of expression, being expressed predominantly in skeletal muscle and in heart (48), tissues which both also express PLZF (Fig. 7). Functional redundancy might therefore explain the lack of an obvious phenotype in DRAL/FHL2 knockout mice (49), a notion that is supported by the fact that DRAL/FHL2 and FHL3 are able to interact with each other (40); DRAL/FHL2 can also interact with ACT at lower affinity (22). However, cotransfection of combinations of FHL3, ACT, and DRAL/FHL2 together with PLZF did not result in a cooperative augmentation of repression (data not shown).
Previous work suggests that FHL proteins can act as activators or coactivators of transcription: ACT, which is expressed specifically in testis (21), serves as a coactivator of CREM, and DRAL/FHL2 is a known coactivator of the androgen receptor (23). Here we observed an additional repressive function of DRAL/FHL2, suggesting that these proteins can have a dual role depending on the promoter context. There are other examples of transcription factors possessing a dual activator-repressor function, like PML which associates with CREB-binding protein to activate transcription (50) and also interacts with several corepressors and HDAC1 (51). Recently FKHR, a member of the hepatocyte nuclear factor 3/forkhead homeotic gene family (HNF3/FKH), has been described to interact with nuclear receptors, exhibiting corepressor activity on steroid receptors and coactivator activity on non-steroid receptors (52). Moreover, a splicing isoform of FHL1 bearing only the first two and a half LIM domains can interact with and negatively regulate the activity of RBP-J, a transcription factor involved in the Notch signaling pathway (24). One possible explanation for the dual role of FHL proteins in transcription might be that they act to stabilize the transcriptional complexes as a type of bridging factor. This interpretation is supported by the fact that, in addition to PLZF, we found an interaction between DRAL/FHL2 and the corepressor NCoR in our two-hybrid screen (data not shown). Three clones were isolated, all of which corresponded to the region upstream of the second receptor interaction domain (IDII) of NCoR. The POZ domain of PLZF is known to be the interface mediating the interaction with NCoR (53,54), whereas the putative DRAL/FHL2 interacting domain would appear to be downstream of this domain (Fig. 2, B and C).
It has been reported that DRAL/FHL2 might be an interaction partner of myocyte nuclear factor, a winged-helix/forkhead protein that forms a corepressor complex with Sin3B (49,55) which in turn can also recruit NCoR (56). Hence, it remains to be seen if DRAL/FHL2 might also take part in this transcriptional complex, making it likely that additional repressing or activating complexes containing FHL proteins will be identified in the future.
Expression of both DRAL/FHL2 and PLZF has been described previously (5,23,25,26,(37)(38)(39) in cardiac tissue. This has again been substantiated by examination of mRNA expression from various tissues (Fig. 7A, lane 4). Furthermore, immunohistochemical analysis of adult rat cardiomyocytes finds both proteins endogenously coexpressed, PLZF presenting its typical nuclear-speckled pattern (Fig. 7B, panel a), corresponding to nuclear bodies where it colocalizes with LAZ3/BCL6 (57), and DRAL/FHL2 with expression both in nucleus and along the Z-discs and M-bands of the myofibrils (Fig. 7B, panel b) (26), suggesting endogenous colocalization of these proteins in the nucleus.
Up to now, a possible implication for PLZF in heart function has not been discussed. However, PLZF is also known to effect an increase in growth arrest and/or apoptosis upon overexpression in myeloid cells (41). Because another putative interaction partner of DRAL/FHL2 in myocyte nuclei, myocyte nuclear factor, can also negatively regulate cell growth (55) and suppress oncogenic transformation, one might speculate that DRAL/FHL2 participates in cellular growth control. This is supported by the initial observation that DRAL/FHL2 is downregulated in myogenic tumor cells compared with their normal counterparts (25) and that its expression can be stimulated by p53 (26). However, definitive assessment of such a function will require analysis of mutant animals that lack expression of all FHL proteins within a given expression domain.
FHL proteins appear to participate in transcriptional complexes where they can modulate tissue-specific activity of activators or repressors. It remains to be seen whether their presence is also required for the regulation of specific downstream target genes.