Relationships of the antiproliferative proteins BTG1 and BTG2 with CAF1, the human homolog of a component of the yeast CCR4 transcriptional complex: involvement in estrogen receptor alpha signaling pathway.

We have reported previously the physical interaction of B-cell translocation gene proteins (BTG)1 and BTG2 with the mouse protein CAF1 (CCR4-associated factor 1) and suggested that these proteins may participate, through their association with CAF1, in transcription regulation. Here we describe the in vitro and in vivo association of these proteins with hPOP2, the human paralog of hCAF1. The physical and functional relationships between the BTG proteins and their partners hCAF1 and hPOP2 were investigated to find out how these interactions affect cellular processes, and in particular transcription regulation. We defined their interaction regions and examined their expression in various human tissues. We also show functional data indicating their involvement in estrogen receptor alpha (ERalpha)-mediated transcription regulation. We found that BTG1 and BTG2, probably through their interaction with CAF1 via a CCR4-like complex, can play both positive or negative roles in regulating the ERalpha function. In addition, our results indicate that two LXXLL motifs, referred to as nuclear receptor boxes, present in both BTG1 and BTG2, are involved in the regulation of ERalpha-mediated activation.

The pathways that inhibit cell proliferation allow normal cycling cells to exit from the cell cycle in response to changes in environmental conditions (e.g. nutrient deprivation, growthinhibiting factors, or high cell density). The BTG 1 family, whose founding member is BTG1 (B-cell translocation gene 1) (1,2), is a family of functionally related genes involved in the negative control of the cell cycle. In vertebrate, this family comprises at least nine distinct members: BTG1, BTG2/ TIS21/PC3, BTG3/ANA, TOB, TOB2, B9.10, PC3K, PC3B, and B9. 15. Two short conserved domains (Box A and Box B) define the signature of this family (3). BTG family proteins have been reported to be involved in some aspects of cell growth, differentiation, and survival (4 -9). For example BTG1, BTG2/PC3, TOB, TOB2, and ANA were reported to display antiproliferative properties (2, 10 -14). Furthermore, BTG2 expression is regulated by p53 and has been found to be involved in DNA damage-induced G 2 /M cell cycle arrest (8). Rat BTG2, known as PC3 (for pheochromocytoma cell-3) was recently shown to inhibit S-phase entry in an Rb-dependent fashion, correlated with the inhibition of cyclin D1 expression (15). Therefore the authors suggest that PC3 may act as a transcriptional regulator of cyclin D1. These results support our previous work, indicating that BTG1 and BTG2 may play a role in transcription regulation. We have shown that these proteins associate physically and functionally with the homeoprotein HOXB9 and enhance HOXB9-mediated transcription (16). In addition, we have demonstrated that both BTG1 and BTG2 interact with mCAF1 (17), whose yeast homolog is a component of the CCR4⅐NOT transcriptional complex, which can affect transcription either positively or negatively (18).
The association of BTG1 with hCAF1, the human homolog of mCAF1, has been confirmed by a study (19) that showed that hCAF1 overexpression in different cell lines leads to a proliferation block, demonstrating its involvement in growth suppression. Furthermore, a search in GenBank with the hCAF1 cDNA sequence revealed sequence identity with a human cosmid (DDBJ/EMBL/GenBank accession number AB020860), localized in the short arm of chromosome 8 at 8p21.3-p22, a region frequently deleted in numerous human tumors (20 -22). CAF1 was subsequently shown to interact with two other members of the BTG family, TOB and TOB2 (13). Recently, the paralog of the hCAF1 gene, hPOP2, has been identified (23) (EMBL/GenBank accession number AF053318). The authors mapped this gene on chromosome 5q31-q33 and suggested that hPOP2 might be the tumor suppressor gene associated with the development of the myelodysplastic (5q-) syndrome. The human POP2 protein was later described (under the name of CALIF) as interacting with hNOT2 and hNOT3, the human homologs of the yeast proteins involved in the formation of the CCR4⅐NOT complex (24).
In other words, CAF1 and POP2 seem to be involved in transcriptional regulation, and both are localized on chromosome regions frequently deleted in human tumors. The structural and functional characterization of these genes should help to establish their role in transcription regulation and in tumorogenesis. In the present study, we demonstrate that POP2 protein, like CAF1, interacts with both BTG1 and BTG2, and we define their interaction regions. We also examine the expression of CAF1 and POP2, together with BTG1 and BTG2, in different human tissues. Finally, we present functional results indicating the involvement of these proteins in estrogen receptor ␣ (ER␣)-mediated transcription regulation.
Mammalian Reporter Plasmid-The pP1-CAT reporter plasmid contains the P1 promoter (nucleotides Ϫ900ϩ13) of the human ER␣ gene fused to the CAT gene (25). The pERE-Luc, which contains three copies of the ER␣ consensus elements (ERE) upstream from the TATA box fused to the luciferase gene, was provided by V. Laudet (ENSL, Lyon, France). The pG4-TK-CAT reporter plasmid contains six GAL4 consensus elements upstream from the thymidine kinase (TK) promoter region, fused to the CAT gene.
Mammalian Expression Constructs-All mammalian expression constructs used are derivatives of the SV40 promoter-driven expression vector pSG5 (Stratagene). The plasmid pSG5Flag was derived from pSG5 by insertion between the EcoRI and BamHI sites of an oligonucleotide containing the Flag peptide sequence (IBI Flag system; Eastman Kodak Co.) and a polylinker. The full-length BTG1 cDNA and the fragments coding for the regions containing amino acids 1-96, 1-117, 1-126, and 38 -171, obtained by PCR, were cloned in-frame with the Flag epitope to generate pSG5FlagBTG1, pSG5FlagBTG1/1-96, pSG5FlagBTG1/1-117, pSG5FlagBTG1/1-126, and pSG5FlagBTG1/ 38 -171. pSG5FlagBTG1S159A and pSG5FlagBTG1⌬BoxB were obtained from pSG5FlagBTG1 by directed mutagenesis (USE mutagenesis kit, Amersham Pharmacia Biotech), respectively, replacing the AGC codon (coding for serine 159) by a GCG (coding for alanine), and deleting bases 292-351, corresponding to Box B. pSG5FlagBTG1ML2, M3L2, ML1L2, and M3L1 were obtained from pSG5FlagBTG1 by the same technique, replacing, at one or both LXXLL sites (bases 130 -144 and 280 -294), one or three CTN codons (coding for leucine) by a GCG (coding for alanine). pSG5FlagCAF1 has been described (17); for pSG5FlagPOP2, the full-length hPOP2 coding sequence, amplified by reverse transcriptase-PCR as described above, has been cloned in pSG5Flag vector. pSG5HEO has been described (26). All of the GAL4 fusion plasmids were obtained by subcloning the different coding regions from pSG5Flag vectors into the Gal4PolyII plasmid (27) in-frame with the yeast GAL4 DNA binding domain coding sequence, as described (17). The fragment corresponding to BTG1 amino acids 76 -171, used to generate pGALBTG1/76 -171, was obtained by digesting the BTG1 cDNA with BamHI. pVPBTG1, pVPBTG2, and pVPCAF constructs were obtained by cloning the respective coding regions in-frame with the VP16 transactivation domain coding sequence into the pSG5FNV vector, as already described (17). The fragments coding for the regions containing amino acids 1-267 and 1-226 of CAF1 were obtained by digesting the 3Ј-end of CAF1 cDNA with the Bal31 exonuclease (Roche Molecular Biochemicals) and then used to construct pVPCAF/1-267 and pVPCAF/1-226. The fragments coding for the regions containing amino acids 1-52 and 53-285 of CAF1 were obtained by digesting the CAF1 cDNA with EcoRI and then used to construct pVPCAF/1-52 and pVPCAF/53-285. pVPCAF⌬229 -247 and pVPCAF⌬11-31 were obtained from pSG5FlagCAF by directed mu-tagenesis (USE mutagenesis kit), deleting, respectively, bases 724 -780 (corresponding to amino acids 229 -247) and 70 -132 (corresponding to amino acids .
The cloned products were verified by DNA sequencing, and the correct expression of all the proteins was checked.
Transfections-The plasmids used for transfection were prepared by the alkaline lysis method and purified by polyethylene glycol/LiCl. HeLa cells were grown in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 10% (v/v) fetal calf serum and seeded at 2.5 ϫ 10 5 cells/well in six-well microtiter plates 8 h prior to transfection. The transfected DNA included various amounts of reporter and expression vectors, as detailed in the figure legends, with 50 ng of the pCMV-␤GAL plasmid as an internal control, in 5 l of Lipo-fectAMINE (Life Technologies, Inc.), and 1 ml of Opti-MEM (Life Technologies, Inc.). The amount of transfected SV40 promoter was kept constant by the addition of pSG5 to the transfection mixture. After 24 h, the cells were washed and treated, where necessary, with a medium containing 10 nM 17␤-estradiol for 24 h. The transfected cells were washed and collected 48 h after transfection.
CAT-ELISA and Luciferase Assay-CAT-ELISA and luciferase assay were performed using the CAT-ELISA kit (Roche Molecular Biochemicals), and the Luc Kit (Promega) following the manufacturers' instructions. The transfected cells were lysed in 0.150 ml of lysis buffer. The supernatants were assayed for CAT and luciferase protein production. All transfection data were normalized by ␤-galactosidase assays, quantified by o-nitrophenly ␤-D-galactopyranoside assay using a standard linear curve. Reporter activity was expressed as the ratio of fold induction to the activity of the reporter vector alone or as pg of CAT protein/ml of cell lysate. Each set of experiments was repeated at least three times, and similar results were obtained in each case.
Immunoblot Analysis-For protein expression assays, 50 g of transfected HeLa cell lysate was subjected to electrophoresis on a 10% polyacrylamide-SDS gel. The separated proteins were transferred to a polyvinylidene difluoride membrane (Millipore) by electroblotting. Equal amounts of protein were loaded onto each lane, as measured by Bradford assay and confirmed by Red Ponceau staining of the transferred membrane. ER␣ was detected with an anti-human monoclonal antibody (HC-20; Santa Cruz Biotechnology, Inc.). The membranes were then incubated with horseradish peroxidase-conjugated rabbit anti-mouse immunoglobulins. The protein was visualized using the enhanced chemiluminescence kit (Amersham Pharmacia Biotech), following the manufacturer's instructions.
Glutathione S-Transferase (GST) "Pull-down" Experiments-GST and GST fusion proteins were expressed in Escherichia coli DH5␣, purified on glutathione-Sepharose beads (Amersham Pharmacia Biotech), and quantified using the Bio-Rad method. SDS-PAGE and Coomassie staining were used to confirm the integrity of the full-length fusion proteins. For in vitro protein-protein interaction assays, 10 g of GST and GST fusion proteins was incubated for 1 h at room temperature with 50 l of glutathione-Sepharose beads in the binding buffer (20 mM Tris-HCl, pH 7.4, 100 mM NaCl, 1 mM EDTA, 1 mM dithiothreitol, 1% milk, 10% glycerol) and packed in minicolumns. 15 l of [ 35 S]methionine-labeled proteins synthesized in vitro (TNT T7 Quick Kit, Promega) were suspended in the binding buffer and passed through GST-, GST-BTG1-, GST-BTG2-, or GST-CAF1-glutathione-Sepharose minicolumns. After washing, the retained proteins were eluted with 10 mM glutathione in 50 mM Tris-HCl, pH 8, analyzed on a 12% SDS polyacrylamide gel, and visualized by autoradiography.
Far Western Analysis-For in vitro protein-protein interaction assays, 5 g of GST, GST-CAF1, or GST-FLRG purified proteins were subjected to 10% SDS-PAGE and transferred to membrane. After denaturation in 6 M and renaturation in 0.187 M guanidine-HCl in HB buffer (25 mM Hepes, pH 7.2, 5 mM NaCl, 5 mM MgCl 2 , 1 mM dithiothreitol), the blots were saturated at 4°C in buffer H (20 mM Hepes, pH 7.7, 7.75 mM KCl, 0.1 mM EDTA, 25 mM MgCl 2 , 1 mM dithiothreitol, 0.05% Nonidet P-40, 1% milk), then incubated for 2 h at 4°C with 50 l of [ 35 S]methionine-labeled in vitro translated ER␣, synthesized in a reticulocyte lysate-coupled transcription/translation system (Promega) in the presence or absence of 100 nM 17␤-estradiol. After washing in buffer H for 1 h at 4°C, the filters were dried and autoradiographed.
Northern Blot Analysis-The pre-made multi-tissue Northern blots (CLONTECH) were hybridized to the indicated 32 P-probes labeled by random priming (Rediprime kit, Amersham Pharmacia Biotech).

RESULTS
The Two Members of the CAF/POP Family Are Widely Expressed in Human Tissues: Relationship with BTG1 and BTG2-hCAF1 and hPOP2 proteins exhibit 76% sequence identity (see Fig. 1A), but their corresponding cDNA are divergent in the 5Ј-and 3Ј-nontranslated regions. To determine the tissue-specific patterns of CAF1 and POP2 gene expression, Northern blots containing human RNA from various tissues were performed using specific probes corresponding to the 3Јnontranslated regions of each transcript.
Transcripts of hCAF1 and hPOP2 were observed in a wide variety of tissues, with the highest levels for hCAF1, in the skeletal muscle, heart, and pancreas and for hPOP2, the heart and pancreas. A 2.4-kb mRNA species was detected with both the CAF1 and POP2 probes (see Fig. 1B). Besides the 2.4-kb transcript, the hCAF1 probe also detected significant amount of a 4.3-kb transcript in the skeletal muscle and a 1.35-kb transcript, which was the most abundant in the testis and was not detected in the lung or brain (Fig. 1B). Both forms were absent in the stomach, small intestine, and thymus. The nature and function of these variant transcripts are still unknown. At the same time we monitored the expression of BTG1 and BTG2 (see Fig. 1B): BTG1 transcript was found, though at barely detectable levels, in most tissues assayed other than the pancreas, heart, and lung, unlike BTG2 expression, which was abundant in the majority of the tissues analyzed, less so in the brain, and absent from the liver and testis.
Interaction of hPOP2 with BTG1 and BTG2-Given that hPOP2 and hCAF1 exhibit 76% amino acid sequence identity, we next investigated whether hPOP2 interacted with BTG1 and BTG2. We also tested the possibility that hPOP2 could interact with CAF1.The hCAF1 and mCAF1 proteins have only one amino acid sequence difference (Fig. 1A), so we used CAF1 to indicate both human and mouse proteins. The cDNA encompassing the entire hPOP2 open reading frame was cloned by reverse transcriptase-PCR, as described under "Experimental Procedures." The interaction assay was performed in the mammalian two-hybrid system, as already described (17). As shown in Fig. 2A both BTG1 and BTG2 interacted with hPOP2 in mammalian cells, but hPOP2 did not associate with CAF1. To verify that hPOP2 can interact directly with BTG1 and BTG2, we performed in vitro association assays with purified recombinant GST fusion proteins. GST-BTG1, GST-BTG2, and, as a control, GST alone, were coupled to glutathione-Sepharose beads and incubated with [ 35 S]methionine-labeled hPOP2. As shown in Fig. 2B, the specific retention of hPOP2 was observed with the GST-BTG1 and GST-BTG2 beads, but not with the control GST beads. The incubation of GST-BTG1 and GST-BTG2 with [ 35 S]methionine-labeled luciferase, used as a control, failed to show any specific interaction (Fig. 2B). These results point to a direct physical interaction of hPOP2 with both BTG1 and BTG2 and indicate that POP2 does not interact with CAF1 and that it does not homodimerize (data not shown).
Mapping of the Protein Domains Required for BTG-CAF1 Interaction-Our previous studies indicated that BTG1 and BTG2 were able to interact with mCAF1 in yeast and in mammalian cells and that Box B was necessary to this interaction (17). More recent work has confirmed that BTG1 protein interacts with hCAF1 (19), but the authors indicate that this inter- action is mediated by the phosphorylation of BTG1 Ser-159. To define further the BTG1 regions involved in the interaction with CAF1, we used the mammalian two-hybrid system, given that proteins are more likely to be appropriately modified posttranscriptionally and that the results are therefore more likely to represent biologically significant interactions. Several deletion mutants of BTG1, described in Fig. 3A, were fused to the DNA binding domain of the GAL4 protein and assayed for possible interaction with CAF1 fused to the VP16 transactivation domain (VP16CAF1) in HeLa cells (see Fig. 3B). As expected, no interaction was found with the GALBTG1⌬BoxB chimera, lacking Box B (amino acids 98 -117), whereas GALBTG1S159A, in which the BTG1 Ser-159 is mutated into Ala, did not prevent BTG1 interaction with CAF1. The results of this assay indicate that the phosphorylation of serine 159, unlike Box B, is not indispensable for BTG1-CAF1 interaction. In contrast to the results obtained with yeast (17), the deletion mutant GALBTG1/1-117 was unable to interact with CAF1, showing that Box B is necessary but not sufficient for this interaction in HeLa cells. As the deletion mutant GALBTG1/ 1-126 was still able to interact with CAF1, we conclude that the interaction in question does not require the C terminus of BTG1 but is strictly dependent on the short sequence flanking Box B (amino acids 118 -126), which is highly conserved between BTG1 and BTG2. The deletion mutant GALBTG1/38 -171 did not interact with CAF1, showing that the N terminus of BTG1 is required for the interaction. This region (amino acids  is also involved in the interaction of both BTG1 and BTG2 with HOXB9 protein in yeast (16) and contains a short motif (EIAAAV) that is conserved in all the members of the BTG family. It is possible that this motif has some functional significance, at least in protein-protein interactions. All of the results obtained with the GALBTG1 chimeric mutants were confirmed in vitro, using a GST pull-down method with purified GST-CAF1 and 35 S-radiolabeled mutant BTG1 proteins (Fig. 3D). We conclude that the interactions observed were direct and not modified by the presence of the GAL4 domain in the fusion proteins. Analysis of the interactions of the VP16POP2 chimeric protein with the GALBTG1 fusion mutants gave similar results (data not shown). Taken together, these results demonstrate that the direct interaction of CAF1 and POP2 with BTG1 involve two regions of BTG1 (amino acids 1-38 and 98 -126). As these two regions are highly conserved between BTG1 and BTG2, it is probable that they are also necessary for the interactions of CAF1 and POP2 with BTG2.
Seeking then to delineate the regions of CAF1 which are important for its association with BTG1 and BTG2, we made a series of deletion constructs of CAF1 (Fig. 4A). Using the mammalian two-hybrid assay (Fig. 4B), we found that two regions, corresponding to residues 11-31 and 229 -247, are important for the interaction between CAF1 and both BTG2 (Fig. 4B) and BTG1 (data not shown). None of the hybrid proteins activated expression of CAT reporter gene on its own. In fact, as described in our previous studies (17), CAF1 and BTG proteins do not seem to be capable of stimulating transcription when tethered to multimerized DNA sites through a GAL4 binding domain in HeLa cells. We confirmed all of these interactions by in vitro interaction using a GST pull-down assay, showing that the interactions observed in HeLa cells are direct (Fig. 4D). As these regions are conserved between CAF1 and POP2, they are probably also involved in BTG-POP2 interactions.
Modulation of ER␣-mediated Transcriptional Activation by   FIG. 2. hPOP2 interacting with BTG1 and BTG2 in vitro and in vivo. Panel A, hPOP2-BTG interaction in vivo using the mammalian two-hybrid system. 200 ng of GAL4 and VP16 expression plasmids corresponding to indicated fusion proteins were transiently cotransfected into HeLa cells with 0.5 g of a reporter gene containing six GAL4 binding sites upstream from a minimal thymidine kinase promoter fused to the CAT gene. Cells were transfected as described under "Experimental Procedures." Total DNA was kept constant at 1 g. Reporter activity, which is expressed as pg of CAT protein/ml of lysate, was normalized with ␤-galactosidase activity. Bars indicate standard deviations from the mean of at least three independent transfections. Panel B, physical interaction of hPOP2 with BTG1 and BTG2 in vitro. 35 S-Labeled full-length in vitro translated hPOP2 and luciferase were suspended in binding buffer and passed through GST, GST-BTG1, and GST-BTG2-glutathione-Sepharose minicolumns. The beads were washed and then eluted with 10 mM glutathione. The eluted proteins and one-third of input radiolabeled proteins were analyzed by 12% SDS-PAGE and visualized by autoradiography. Molecular size markers are given in kDa.
BTG Proteins-The fact that BTG1 and BTG2 interact with CAF1 and POP2, which are homologs of the yeast transcription factor yCAF1/POP2 and act as cofactors for HOXB9-mediated transcription (16), supports the hypothesis that these proteins play a role in transcription regulation. In addition, BTG1 and BTG2 contain two copies of an LXXLL motif known as the NR (nuclear receptor) box (Fig. 6A), which was identified as being essential for the interaction of a number of coactivators with nuclear receptors (29). One motif (referred to below as L1) is located in the N-terminal part of the two proteins; the other motif (referred to below as L2) is located within the middle part, at the beginning of Box B, one of the two conserved domains that constitute the BTG family signature. These observations incited us to study the possible role of the BTG proteins in the transcriptional regulation of the nuclear receptors. We first focused on ER␣ because both BTG and ER␣ are involved in the regulation of cell proliferation: hormone binding to ER␣ induces conformational changes leading to the recruit- ment of transcriptional auxiliary factors, binding of ER␣ to EREs in gene promoters, and regulation of transcriptional activity of genes involved in proliferation, development, and differentiation (for review, see Refs. 30 -33). HeLa cells, which lack endogenous ER␣, were transfected with a vector expressing ER␣, pSG5HEO, and a luciferase reporter linked to a multimer palindromic ERE sequence, pERE-Luc, along with either a control plasmid or vectors expressing BTG1 and BTG2, in the presence of 17␤-estradiol. The ER␣ expressed by pSG5HEO contained a substitution of a valine for a glycine (point mutation in codon 400) which made its binding to ERE strictly dependent on estradiol, at least in vitro (26). Both BTG1 and BTG2 significantly enhanced the ER␣-mediated activation of the luciferase reporter gene, as shown in Fig. 5A. No effect of BTG1 or BTG2 on reporter gene activity was observed in the absence of ER␣, and the cotransfection of pSG5FlagFOLL, which encodes an unrelated protein that was used as a control, showed no effect on reporter gene activation (Fig. 5A).
We performed similar transfection assays using a CAT reporter gene driven by the P1 promoter (nucleotides Ϫ900ϩ13) of the human ER␣ gene (pP1-CAT), which is responsive to estradiol through three half-EREs dispersed in the promoter (25). HeLa cells were transfected with the pSG5HEO vector, expressing ER␣, and pP1-CAT, with vectors expressing BTG1 and BTG2, in the presence of 17␤-estradiol. In marked contrast to the previous results, both proteins significantly inhibited ER␣-mediated activation of the CAT reporter as shown in Fig. 5B. The control reporter construct pBLCAT3, which lacks the P1 promoter, was not activated when cotransfected with the ER␣-expressing vector (data not shown), and no effect of BTG1 or BTG2 on reporter gene activity was observed in the absence of ER␣. Cotransfection of pSG5FlagFOLL used as a control, showed no effect on reporter gene activation (Fig. 5B). Similar results were obtained in COS-7 and MCF7 cells (data not shown).
Taken together, the results of these transfection studies suggest that BTG1 and BTG2 can function either as coactivator or corepressor of ER␣ (Fig. 5, A and B), depending on the promoter context, which in turn suggest that the BTG proteins can act as effectors of the ER␣ signaling pathway.
Role of NR Motifs in Transcriptional Regulation by BTG1-To determine the relative importance of the two LXXLL motifs, which are common to both BTG1 and BTG2, we constructed a series of full-length BTG1 derivatives bearing individual leucine to alanine substitutions as illustrated in Fig. 6A. These constructs were expressed in HeLa cells and assayed for ER␣dependent transcriptional activation. The results of four independent experiments (Fig. 6, B and C) demonstrated that the mutation of LXXLL to LXXAL in both NR boxes (see Fig. 6B; ML1L2) were fully responsive to ER␣. In contrast, mutations that converted the three hydrophobic leucine to alanine (M3L1 and M3L2) in the L1 or in the L2 motif, prevented BTG1 for having an effect on ER␣ transcriptional activity, suggesting that both motifs participate in the observed regulation. As expected the mutant BTG1/1-96, lacking Box B and L2, had no effect on ER␣-mediated transcriptional activation. Thus, the activity of BTG1 and BTG2 on ER␣ appears to depend on the presence of two functional NR motifs.
Mechanisms for BTG Regulation of ER␣-mediated Activation-To investigate whether the ER␣-BTG functional interaction takes place directly, we carried out GST pull-down experiments. In vitro translated ER␣ did not appear to interact directly with either GST-BTG1 or GST-BTG2 used as baits, either in the presence or the absence of the ligand (data not shown). But although the BTG proteins seem to not interact directly with ER␣, they possibly interact with other components of regulatory complexes involved in ER␣-dependent transactivation.
Mouse CAF1 was identified as interacting with the CCR4 complex, a general transcription multisubunit complex that, in yeast, regulates the expression of different genes involved in cell cycle regulation and progression. Our previous data suggested that BTG proteins might participate, through association with CAF1, in transcription regulation. To explore further the functional significance of the interaction between BTG proteins and CAF1, we tested the possibility that the overexpression of CAF1 could affect ER␣-dependent activation. As shown in Fig. 7, A and  B, CAF1, like its partner proteins BTG1 and BTG2, regulates ER␣-mediated transactivation and can function either as coactivator or corepressor depending on the promoter context. To ensure that this effect on ER␣ transcriptional activity was not a result of a modification of ER␣ expression, we performed a West-ern blot analysis. As Fig. 7C illustrates, ER␣ protein is expressed at a comparable level in both the presence and absence of exogenous BTG1, BTG2, and CAF1.
Because BTG1, BTG2, and CAF1 acted as coactivators of ER␣, we tested for additive or synergistic effects on ER␣-dependent activation when both proteins were expressed in the transient transfection assays. The simultaneous expression of CAF1 and BTG did not shown a synergistic effect but produced an enhancement over the activation observed with each protein alone, depending upon ER␣, BTG, and CAF1 expression levels and the cellular context (data not shown).
At this point we looked at the possibility of a direct CAF1-ER␣ interaction performing a far Western blot analysis. Purified GST, GST-CAF1, and GST-FLRG, an unrelated protein used as control, were subjected to SDS-PAGE, transferred from the gel to membrane, and probed with [ 35 S]methionine-labeled ER␣ protein in the presence or absence of 17␤-estradiol. As shown in Fig. 7D, specific hybridization was observed with GST-CAF1, but not with GST and GST-FLRG after incubation with labeled ER␣ in the presence of hormone. [ 35 S]Methioninelabeled ER␣ protein in the absence of hormone failed to show any interaction. This finding suggests that BTG1 and BTG2 may exert their coactivator function on ER␣-mediated transcription by means of CAF1.
We next investigated whether the LXXLL motifs might be involved in the BTG1-CAF1 interaction. As Fig. 8 shows, the BTG1 LXXLL mutant that can modulate ER␣-dependent transcription (BTG1ML1L2) was still able to interact with CAF1. In contrast, the LXXLL mutations that abolish the effect of BTG1 on ER␣ transcription (BTG1M3L1 or BTG1M3L2), also strongly decrease the interaction of BTG1 with CAF1 (Fig. 8). These results were confirmed by mammalian two-hybrid assay (data not shown). Furthermore, the interaction of BTG1 with CAF1 is consistent with its ability to mediate transcription activation and inhibition.

DISCUSSION
In this study we compared structural and functional features of CAF1 and POP2 gene products and their relations with the BTG proteins. The POP2 coding region has a high degree of homology with mouse and human CAF1, resulting in a protein that has 76% amino acid identity with mouse and human CAF1. A noticeable difference between POP2 and CAF1 is found in the C-terminal region, where POP2 contains an extension that is not found in CAF1 (see Fig. 1A). The present study shows that POP2, like CAF1, can interact with BTG1 and BTG2 and that the interactions involve two BTG1 regions (amino acids 1-38 and 98 -126). Because these two regions are highly conserved between BTG1 and BTG2, it is probable that they are necessary for interactions of CAF1 and POP2 with BTG2. Again, two CAF1 regions are involved in the interaction with BTG proteins. It is possible that both regions are necessary for BTG-CAF interaction or that one of them is important for the association, and the other is essential for maintaining the appropriate structure of the proteins.
In an effort to understand how POP2 and CAF1 function in vivo we examined expression of their mRNA in numerous human tissues. Along the 2.4-kb transcript that was detected by both the hCAF1 and hPOP2 probes, the hCAF1 probe also detected two variant transcripts whose structure and function are still unknown. CAF1 and POP2 being partners of BTG proteins, we monitored the expression of BTG1 and BTG2 as well. The high degree of evolutionary conservation of CAF1 and POP2 among eukaryotes indicates that these proteins play an important biological role. Because CAF1 and POP2 were found to be expressed in every cell type tested, it is possible that functional specificity and selectivity could be achieved by in- teractions with different partners. The fact that hPOP2/CALIF but not hCAF1, is able to interact with hNOT2 and hNOT3 (24) suggests that the two proteins are functionally distinct and that they could participate in the formation of different complexes in mammals. This observation is particularly interesting because in yeast the CCR4⅐NOT complex appears to be composed of at least two groups of proteins which are physically and functionally distinct (34).
Different findings indicated the possibility that BTG1 and BTG2 play a role in transcription regulation: (a) both proteins interact with HOXB9 and modulate its transcription activity (16); (b) both of them interact with CAF1 (17) and POP2 (this report), which are homologs of a yeast transcription factor; (c) BTG2 acts as a transcriptional regulator of cyclin D1 (15); (d) BTG1 and BTG2 contain two copies of an LXXLL motif (see Fig.  6A), which has been identified as being essential for the interaction of a number of coactivators with nuclear receptors (29); (e) they also interact with the protein-arginine N-methyltransferase (PRMT1) (35), and a relationship between the methylation of proteins and the transcription regulation of nuclear receptors has recently been described (36,37). We therefore investigated the possible involvement of the BTG proteins in the transcriptional Shown is a GST pull-down assay with interactions between 35 S-labeled in vitro translated BTG1 proteins mutated in LXXLL boxes, as described in Fig. 6A, and GST-CAF1, as described in Fig. 2B. regulation of the estrogen nuclear receptor ER␣. Our results indicate that BTG1 and BTG2 can function as coactivators and corepressors of ER␣ (Fig. 5) and that the LXXLL sequences are involved in the effect of BTG proteins on ER␣-dependent transcription (Fig. 6). This result along with the fact that (a) the BTG proteins need the LXXLL motifs to modulate ER␣-mediated transcription and to interact with CAF1 and (b) that CAF1 also acted as a modulator in this assay and can bind directly to ER␣ in vitro strongly suggests that BTG proteins modulate ER␣-mediated transcription through their interaction with CAF1 via a CCR4like complex or complexes. As regards yeast, it is thought that a complex of this type may affect transcription either positively or negatively (18). This hypothesis is supported by our recent results showing that in mammalian cells CAF1 and BTG1 bind together in a large multiprotein complex. 2 Alternatively, it may be that the LXXLL sites are important for the correct conformation of BTG proteins and for their biological activity. In fact this motif normally takes on a helical conformation and facilitates protein-protein interaction. BTG proteins may function as bridges between nuclear receptors associated with coactivators or corepressors and component(s) of the transcription machinery. We have also tested the ability of BTG1 and BTG2 to affect transactivation by other members of the nuclear receptor family, and our results indicate that BTG1 and BTG2 also modulate the transcriptional activity of the thyroid hormone receptor ␣ (data not shown).
BTG1 and BTG2 exhibit either corepressor or coactivator characteristics depending on the promoter context. The major difference between the two promoters used for the transcription assay was that the P1 promoter of ER␣ is a natural promoter containing three half-ERE dispersed in the promoter, whereas the ERE-Luc contains multimerized palindromic ERE. In the case of the P1 promoter, ER␣ can bind as monomer, and its regulation can be different from that of a dimeric conformation. In accordance it has been demonstrated that SRC1, an ER␣ coactivator, requires the presence of ER␣ dimer for binding (38). Although the mechanistic basis of the inhibitor effect of BTG1 and BTG2 on ER␣-dependent transcription remains unclear, a number of plausible hypotheses may be put forward: inhibition of DNA binding of ER␣, direct antagonism of cofactors functioning via competition, or transrepression via recruitment of a putative repressor. BTG1 and BTG2 overexpression in vivo may sequester limiting factors required for ER␣ transactivation, thus leading to an imbalance of limiting constituents that generate the functional ER␣ activation complex and leading to transcriptional squelching. Recent studies have shown that PC3, the rat homolog of BTG2, inhibits the expression of cyclin D1, which stimulates ER transactivation (28,39). We can speculate that BTG2 inhibits the activation of the ER␣ P1 promoter by inhibiting cyclin D1 expression. The characterization of the components of the multiprotein complex containing CAF1 and BTG should tell us whether BTG1 and BTG2 modulate ER␣-mediated transcription directly or through the control of ER␣ activator expression.
It is possible that BTG and CAF1 proteins, whose expression inhibits cell proliferation, can function as corepressors for genes whose expression activates cell proliferation and, vice versa, as coactivators in the context of antiproliferative gene promoters. The localization of hCAF1 and hPOP2 on chromosome regions frequently deleted in human tumors suggests a tumor-suppressing role for these proteins. The mechanism by which BTG and CAF1 modulate ER␣-activated transcription remains to be elucidated. Further investigation is also required to understand how BTG-CAF1 interactions mediate transcription regulation in normal physiology and in neoplasia.