Forkhead Homologue in Rhabdomyosarcoma Functions as a Bifunctional Nuclear Receptor-interacting Protein with Both Coactivator and Corepressor Functions*

, In a search for novel transcriptional intermediary factors for the estrogen receptor (ER), we used the ligand-binding domain and hinge region of ER as bait in a yeast two-hybrid screen of a cDNA library derived from tamox-ifen-resistant MCF-7 human breast tumors from an in vivo athymic nude mouse model. Here we report the isolation and characterization of the forkhead homologue in rhabdomyosarcoma (FKHR), a recently described member of the hepatocyte nuclear factor 3/forkhead homeotic gene family, as a nuclear hormone receptor (NR) intermediary protein. FKHR interacts with both steroid and nonsteroid NRs, although the effect of ligand on this interaction varies by receptor type. The interaction of FKHR with ER is enhanced by estrogen, whereas its interaction with thyroid hormone receptor and retinoic acid receptor is li-gand-independent. In addition, FKHR differentially regulates the

The nuclear hormone receptors (NRs) 1 play an important role in a variety of physiological functions such as cell growth, development, differentiation, and homeostasis (1,2). The NR superfamily is often divided into steroid and nonsteroid receptor subfamilies, which show different features in DNA binding and dimerization and a different effect on the basal transcriptional activity of the target (2,3). The estrogen receptor (ER), a member of the steroid receptor family, is critical for the development and progression of breast cancer, and it is a useful diagnostic and therapeutic target (4 -8). Like other NRs, ER contains two distinct transactivation function domains (AFs): the ligand-independent (AF-1) and ligand-dependent (AF-2) activation domains (4,8). A large number of ER-interacting proteins have been identified that modify ER activity. Several coactivators have been characterized recently including SRC-1, Grip1/TIF2, RIP140, Trip1, CBP/P300, SPA/L7, and AIB1/ ACTR/RAC3/p/CIP (9 -12). In addition, several corepressors have also been identified including N-CoR and SMRT (13). The relative expression and/or activity of coactivators and corepressors in a particular environment may modulate the agonistic/ antagonistic activities of the partial ER antagonist, tamoxifen (Tam) (14 -17). Most recently, two bifunctional NR intermediary proteins, TIF1 and NSD1, have been described that can regulate transcription either positively or negatively, depending on both the promoter context and the cell type (18,19).
To identify novel transcriptional intermediary factors for ER that might contribute to estrogen-dependent cell proliferation, we used the AF-2 and hinge region of ER as bait in a yeast two-hybrid screen of a cDNA library derived from Tam-resistant breast tumor tissues from an MCF-7 athymic nude mouse model (20,21). Here we report the isolation and characterization of FKHR (forkhead homologue in rhabdomyosarcoma), a previously described member of the hepatocyte nuclear factor 3/forkhead homeotic gene family (HNF3/FKH), as a novel bifunctional NR-interacting protein that displays corepressor activity on steroid receptors and coactivator activity on nonsteroid receptors (22,23). Consistent with these observations, overexpression of FKHR in MCF-7 cells, an estrogen-dependent human breast cancer cell line, dramatically inhibits their proliferation. FKHR has been been shown recently to be an important player in several signal transduction pathways regulated by the AKT protein kinase (24 -32).

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Screen-The plasmid pAS2-1/ER (DEF), coding for a fusion protein containing the GAL4 DBD and the AF-2 and hinge regions of hER␣, was constructed from the vector pAS2-1 (CLON-TECH) and used as bait. The cDNA library was prepared using Tamresistant breast tumor tissues from an MCF-7 athymic nude mouse model constructed in pGAD10 by CLONTECH (20,21). The screen was performed in the presence of estrogen (E 2 ). The positive clones were tested further for ligand-dependent interactions with ER by ␤-gal filter lift assay after cotransformation of DBD-ER (DEF) and AD-FKHR (amino acids 402-629) in the absence and presence of ligand.
GST Pull-down Assay-The constructs GST-TR and GST-RAR were provided kindly by Michael G. Rosenfeld (33). GST-ER (DEF) was generated by inserting the hinge and AF-2 regions of hER␣ into the BamHI/ EcoRI sites of the pGEX-2kt vector (Amersham Pharmacia Biotech). The plasmid pcDNA3.1/AIB1 was provided kindly by Paul S. Melzer (9), and pcDNA3/FKHR was constructed by subcloning the full-length FKHR into pcDNA3 at the Klenow-filled BamHI and XhoI sites. The GST pull-down assay was performed as described previously (34).
Transfections, Luciferase, and Growth Inhibition Assays-The reporter genes 2xERE-Tk-Luc (35), PRE/GRE-TATA-Luc (36), pC3-Tk-Luc (37), TRE-Tk-Luc, and RARE-Tk-Luc (38) were described previously. Monkey kidney-derived COS-1 (ATCC) and human hepatocyte carcinoma HepG 2 cells (ATCC) maintained in improved Eagle's medium supplemented with 10% fetal bovine serum (Life Technologies, Inc.) were seeded in phenol red-free improved Eagle's medium (Life Technologies, Inc.) containing minimal essential components and transfected 24 h later with the DNAs indicated in the figure legends using Fugene 6 (Roche Molecular Biochemicals). The total amount of DNA was kept constant in all transfections by the addition of empty vector DNA as carrier. Twelve hours later, the cells were treated with different ligands for 24 h and harvested for ␤-gal and luciferase activities. The luciferase activities were normalized to the ␤-gal activities. The data are presented as the average Ϯ S.E. of triplicates and are representative of at least three independent experiments. The single cell proliferation assay (39) and colony reduction assay (40) were performed as described previously.

RESULTS
Using ER as bait, the yeast two-hybrid screen of the cDNA library derived from Tam-resistant breast tumors was utilized to identify novel ER-interacting proteins. A 0.8-kilobase open reading frame DNA fragment obtained from this screen showed 100% identity with the C-terminal one-third of FKHR (amino acids 402-629), a member of the HNF3/FKH transcription factor family (22,23). The sequence of FKHR protein exhibits certain interesting features (22). As shown in Fig. 1A, the protein contains at its N terminus a forkhead domain that is highly homologous among the HNF3/FKH family and is necessary for DNA binding, and at its C terminus a proline-rich and acidic serine/threonine-rich transactivation domain (AD). Other motifs in FKHR include an SH3 binding site and an alanine-rich region that has been associated with transcriptional repression in other proteins. FKHR and two other closely related members, FKHRL1 and AFX (41,42), are relatively divergent from other HNF3/FKH family members both within and outside the forkhead region. As shown in Fig. 1B, their forkhead domain lacks the N-terminal KPPY motif common to most HNF3/FKH family genes and contains a novel 5-amino acid insert (DKGDS) instead. Interestingly, at their C termini an NR-interacting domain or LXXLL motif (Fig. 1A, NID) (43-45) is highly conserved among FKHR, FKHRL1, and AFX but not other HNF3/FKH members (Fig. 1C).
The interaction between FKHR and ER was defined further using the yeast two-hybrid and GST pull-down assays. In the yeast two-hybrid assay ( Fig. 2A), in the presence of E 2 , coexpression of two chimeric proteins (GAL4 AD-FKHR (amino acids 402-629) and GAL4 DBD-hER␣) resulted in good cell growth on selection medium, and the colonies turned blue within 1 h in the ␤-gal filter lift assay ( Fig. 2A, E 2 ), whereas a minimal growth was seen either in the absence of ligand ( Fig.  2A, No ligand) or the presence of anti-estrogen ( Fig. 2A, Tam). Surviving colonies in the absence of ligand or with Tam did not turn blue within 16 h in a ␤-gal filter lift assay. In the GST pull-down assay, FKHR only weakly interacted with ER both in the absence of hormone (Fig. 2B, lane 2) and the presence of Tam (Fig. 2B, lane 4). E 2 treatment enhanced the interaction by 2-3-fold (Fig. 2B, lane 3). GST alone did not interact with FKHR even in the presence of E 2 (Fig. 2B, lane 5), indicating the specific interaction between ER and FKHR. As reported previously (9 -11), the known ER-interacting protein AIB1 interacted with ER only in the presence of E 2 (Fig. 2B, lane 8). Taken together these results suggest that FKHR is an ERinteracting protein that preferentially associates with E 2bound ER.
The ligand-dependent interaction between FKHR and ER suggests a potential role for FKHR as an ER coregulator. Therefore, we tested whether FKHR can directly affect ER transactivation by cotransfection of FKHR and ER along with reporters driven by either an artificial ERE promoter (Fig. 2C) or a natural ERE promoter (Fig. 2D) into mammalian cells. In both cases, co-expression of ER and FKHR into HepG 2 cells repressed 2-3-fold the transcriptional activity of the reporter in the presence of E 2 in a dose-dependent manner. A similar result was obtained in COS-1 cells (data not shown). A slight repression of reporter activity was also observed in the absence of ligand and the presence of Tam, which could be either caused by residual E 2 in the medium or the weak interaction of ER and FKHR under those conditions. Alternatively, FKHR could act as a general transcriptional repressor. However, the transcriptional activity of the CMV-␤-gal (used as an internal control) was not significantly affected (data not shown), and the repres-sion was observed only when the ER was cotransfected (Fig.  2C), indicating that the repression is receptor-dependent. To exclude the possibility that FKHR may down-regulate ER, we measured ER protein levels in cells transfected with various amounts of FKHR. No difference in ER protein levels was found (data not shown), indicating that the repressive effect of FKHR on ER-mediated transactivation is not caused by ER down-regulation.
To test whether FKHR can counteract the activator activity mediated by other known coactivators, HepG 2 cells were cotransfected with a constant amount of the AIB1 expression plasmid along with increasing amounts of FKHR (Fig. 3A) or vice versa (Fig. 3B). As reported previously (9), the transient transfection of AIB1 resulted in a 2-fold enhancement of ERmediated transactivation, confirming its ER coactivator activity. Cotransfection of increasing amounts of FKHR along with AIB1 resulted in a dose-dependent repression of AIB1-enhanced ER-mediated transactivation (Fig. 3A). Conversely, cotransfection of increasing amounts of AIB1 along with a constant amount of FKHR gradually reversed the repressive effect of FKHR on ER-mediated transactivation (Fig. 3B). These data demonstrate that FKHR and AIB1 antagonize the effects of each other on ER-mediated transactivation.
To investigate whether the repressor activity of FKHR is limited to ER, we examined the effect of FKHR on other NRs. Similar to ER, cotransfection of FKHR resulted in a gradual repression of PR-and GR-mediated transactivation (Fig. 4, A  and B), and this effect was solely observed in cells cotransfected with the receptor and treated with their cognate ligands. In addition, FKHR interacted with two nonsteroid NRs: GST-RAR and GST-TR (Fig. 4C). However, unlike ER, FKHR bound to TR and RAR both in the absence and presence of their cognate ligands, indicating the ligand independence of the interaction. In contrast to steroid receptors, cotransfection of FKHR resulted in 2-3-fold stimulation, rather than repression, of RARand TR-mediated transactivation both in the absence and presence of their cognate ligands (Fig. 4, D and E).
To determine the physiologic relevance of the interaction of FKHR with ER, we tested the effect of FKHR on the growth of an estrogen-dependent human breast cancer cell line, MCF-7. Using a single colony reduction assay (Fig. 5A), we observed that empty vector-transfected cells had a significantly different distribution in the cell number per colony or cluster than FHKR-transfected cells (p Ͻ 0.0001). Clusters containing single cells or doublets were much more common in cells transfected with FKHR, whereas clusters containing more than 10 cells were common in control dishes (Fig. 5A). This difference in colony size is more apparent when comparing the median cell number per colony for pcDNA-transfected cells (seven cells/ colony) with the FKHR-transfected cells (two cells/colony). In the colony reduction assay (Fig. 5B), in comparison with vector alone, overexpression of FKHR greatly reduced colony formation (50% reduction). Similarly, overexpression of p21, a known negative regulator for cell growth, resulted in a 70% reduction of colony formation (40).
Finally, Western blot analysis revealed the differential expression pattern of FKHR in different human tissues (Fig. 6). FKHR is expressed in most of the tissues tested with higher levels in ovaries and testes and intermediate levels in brain, heart, kidneys, liver, and skeletal muscle. There is very low FKHR expression in lungs and no detectable levels in placenta and spleen. Interestingly, there is a doublet band in muscle (heart and skeletal muscle) but not in other tissues tested, although we do not know currently the nature of this doublet. The tissue differences in FKHR protein levels may play a role in tissue specificity of NR-mediated responses to various hormones or antihormones. DISCUSSION We have shown that FKHR interacts with several members of the NR superfamily including ER, RAR, and TR, although the effect of ligand on this interaction varies by receptor type. Its interaction with ER is enhanced by estrogen, whereas its interaction with TR and RAR is ligand-independent. The characteristic features of the interaction of FKHR with ER compared with TR and RAR are similar to those described for the previously described NR intermediary protein NSD1 (19). The two distinct NR-interacting domains identified in NSD1 could be also present in FKHR. Alternatively, the different binding features of FKHR could also reside within the structure of the NRs themselves. As mentioned, steroid and nonsteroid receptors show distinct features in their DNA binding and dimerization and in their effects on the basal transcriptional activity of target genes. First, steroid receptors form homodimers in their active state, whereas nonsteroid receptors heterodimerize with RXR upon the addition of ligand. Conceivably, FKHR could interact differently with homodimer versus heterodimer partners. Second, unliganded steroid receptors are complexed with chaperone proteins and remain in an inactive state, whereas unliganded nonsteroid receptors are bound to DNA and are complexed with corepressors, resulting in the repression of basal transcription of target genes. FKHR could bind with different affinities and/or mechanisms to DNA-bound versus free NRs. Because there is no DNA or promoter present in our in vitro assay, the interaction seen in vitro may not necessarily reflect the in vivo situation.
In addition to its different binding properties with steroid and nonsteroid NRs, FKHR differentially regulates the transactivation mediated by different NRs. Co-expression of FKHR in mammalian cells (HepG 2 and COS-1) dramatically represses transactivation mediated by ER, PR, and GR. In contrast, FKHR stimulates rather than represses transactivation mediated by RAR and TR. The differential effects of FKHR on the transactivation of different NRs might be explained by the presence of both coactivation and corepression domains in the FKHR molecule, as described for NSD1 (19). Sequence analysis of the FKHR gene has shown a transactivation domain at its C terminus as well as a repression region at its N terminus (22). Binding to different receptor types could result in conformational changes, exposing either the transactivation or repression domains of FKHR. In addition, homodimers and heterodimers of NRs might recruit different sets of transcriptional components that might then be differentially regulated by FKHR. Finally, because its regulatory function depends on the nature of the receptor, it is possible that FKHR activates transcription functions of TR and RAR by either sequestering TRand RAR-specific corepressors or blocking the histone deacetylase activity associated with these corepressors. Further experiments are required to address these possibilities.
The HNF3/FKH transcription factor family has been implicated in diverse biological functions varying from embryonic development to adult tissue-specific gene expression (46 -48). In addition, variants of several genes of this family (especially the FKHR subfamily) have shown oncogenic potential (46,49,50). FKHR was originally cloned from a rhabdomyosarcoma because of its aberrant fusion with another transcription factor, PAX3, resulting from a unique chromosomal translocation t(2;13) (22,23). The resulting fusion protein PAX3-FKHR is a hallmark of these tumors (51)(52)(53)(54) and is thought to play a crucial role in muscle cell transformation and evolution to rhabodmyosarcoma. Little is known about the underlying mechanism of this transformation process (55)(56)(57). Characterization of FKHR as an NR transcriptional intermediary protein should provide clues about the biological function of FKHR and possibly the oncogenic mechanism of PAX3-FKHR. It is possible that the chromosomal translocation in rhabdomyosarcoma results in not only the activation of PAX3 but also disruption of functional FKHR, which may be essential for RAR-dependent muscle cell differentiation. This loss of a differentiation function of FKHR, rather than a gain of function by the PAX3-FKHR fusion, could conceivably contribute to the development of rhabdomyosarcoma. In addition, FKHR has been shown recently to play a role in several signal transduction pathways (24 -32). Further studies on the mechanism of the regulatory function of FKHR and its biological relevance, especially its effect on hormone-dependent cell proliferation and differentiation, may help reveal the role of FKHR in the development and progression of cancers such as breast cancer, leukemia, and rhabdomyosarcoma.

FIG. 5. FKHR inhibits the growth of MCF-7 breast cancer cells.
A, single cell proliferation assay. MCF-7 cells maintained in improved Eagle's medium plus 10% fetal bovine serum were cotransfected with 5 g of pcDNA3 (n ϭ 100) or pcDNA 3/FKHR (n ϭ 102) plus 0.5 g of pCMV-␤-gal. After 3-4 doublings, the cells were fixed and stained for ␤-gal in situ. Colonies containing blue cells were scored for the number of blue cells per colony and subjected to biostatistical analysis. B, colony formation assay. MCF-7 cells maintained in improved Eagle's medium plus 10% fetal bovine serum were transfected with 1 g of pcDNA3, pcDNA3/FKHR, pcDNA/p21, or mocktransfected and subjected to G418 selection 48 h after transfection. The surviving colonies were stained and counted after 14 days of selection. The number of colonies on each plate was counted and graphed as averages Ϯ S.E. from triplicates.