Differential interaction of nuclear receptors with the putative human transcriptional coactivator hTIF1.

Hormonal regulation of gene activity is mediated by nuclear receptors acting as ligand-activated transcription factors. Intermediary factors interacting with their activation functions are required to mediate transcriptional stimulation. In search of such receptor interacting proteins, we have screened a human cDNA expression library and isolated a human protein that interacts in vitro with transcriptionally active estrogen receptors (ER). Sequence analysis reveals that this protein is the human homolog of mouse TIF1 (transcription intermediary factor 1) shown to enhance nuclear receptor ligand-dependent activation function 2 (AF2) in yeast. We have characterized the nuclear receptor binding site on hTIF1 and shown that a region of 26 residues is sufficient for hormone-dependent binding to the estrogen receptor. As shown by point mutagenesis, the AF2 activation domain of ER is required for the binding of hTIF1 but not sufficient, since a short region encompassing the conserved amphipathic alpha-helix corresponding to this domain fails to precipitate hTIF1. We also demonstrate that hTIF1 association with DNA-bound ER requires the presence of estradiol. Finally, we show that the interaction of hTIF1 with receptors is selective since strong in vitro hormone-dependent binding is only observed with some members of the nuclear receptor superfamily.

Hormonal regulation of gene activity is mediated by nuclear receptors acting as ligand-activated transcription factors. Intermediary factors interacting with their activation functions are required to mediate transcriptional stimulation. In search of such receptor interacting proteins, we have screened a human cDNA expression library and isolated a human protein that interacts in vitro with transcriptionally active estrogen receptors (ER). Sequence analysis reveals that this protein is the human homolog of mouse TIF1 (transcription intermediary factor 1) shown to enhance nuclear receptor liganddependent activation function 2 (AF2) in yeast. We have characterized the nuclear receptor binding site on hTIF1 and shown that a region of 26 residues is sufficient for hormone-dependent binding to the estrogen receptor. As shown by point mutagenesis, the AF2 activation domain of ER is required for the binding of hTIF1 but not sufficient, since a short region encompassing the conserved amphipathic ␣-helix corresponding to this domain fails to precipitate hTIF1. We also demonstrate that hTIF1 association with DNA-bound ER requires the presence of estradiol. Finally, we show that the interaction of hTIF1 with receptors is selective since strong in vitro hormone-dependent binding is only observed with some members of the nuclear receptor superfamily.
The estrogen receptor (ER) 1 belongs to a superfamily of nuclear receptors that function as ligand-dependent transcription factors (1,2). Transcription is mediated by means of two activation regions, AF1 located in the NH 2 -terminal domain and AF2 located in the hormone binding domain, whose activities vary depending on the responsive promoter and cell type (3)(4)(5). These two activation regions appear to function independently on certain promoters, but, in some cases, both are required for full transcriptional activity (4). Recently, a ligand-dependent functional interaction between the two AFs has been demonstrated using the two-hybrid system in mammalian cells (6).
One essential element required for the activity of AF2 is a COOH-terminal region, which is highly conserved in the nuclear receptor family and has been shown to be essential for transcriptional activation by estrogen and glucocorticoid receptors (7), thyroid hormone receptors (8,9), and retinoid receptors (10). This sequence (⌽⌽XE⌽⌽) has the characteristics of an amphipathic ␣-helix (7). Determination of the crystal structure of RAR-␥, RXR-␣, and TR-␣ ligand-binding domains has revealed that this helix is repositioned upon ligand-induced conformational transition and functions as a lid for the ligand pocket (11)(12)(13).
The mechanism whereby transcriptional stimulation is achieved remains unknown, but it is believed that direct or indirect interactions with target proteins within the transcriptional apparatus could be involved to allow the assembly of a preinitiation complex (14,15). A number of these receptors have been shown to bind directly to the TATA box-binding protein (TBP) and TFIIB in vitro (16 -20). One of the TBPassociated factors, hTAFII30, also interacts selectively with the hormone binding domain of the estrogen receptor and appears to contribute to transcriptional activity in vitro (21). However, the interaction was unaffected by the binding of either 17␤estradiol or antiestrogens such as 4-hydroxytamoxifen and was mapped to a region that is inactive in mammalian cells (22).
Other factors acting as bridging proteins are probably involved in transcriptional activation by nuclear receptors. Using in vitro protein interaction assays or yeast two-hybrid systems with different nuclear receptor hormone-binding domains as bait, several factors that associate with activated receptors have been characterized. These include hRIP140 (23)(24)(25)(26), mTIF1 (27), hTRIP1/mSUG1 (28,29), several isoforms of SRC-1 (30 -33), and CBP/p300 (34 -37). Two other factors, namely N-CoR and SMRT, that interact with unliganded thyroid hormone and retinoic acid receptors, have been isolated and shown to act as transcriptional corepressors, which are released upon ligand binding (38,39). In fact, it seems likely that these intermediary factors (activators or repressors) form a large family of molecules with differential binding sites or affinities for various nuclear receptors.
We report here the cloning of a human factor, which interacts in vitro with transcriptionally active estrogen receptors and appears to be the homolog of mouse TIF1 (27). We have characterized the interaction of hTIF1 with nuclear receptors and shown that a short region is sufficient for hormone-dependent binding to various members of the superfamily. However, it appears that this interaction is selective, since in vitro hormonedependent interaction with hTIF1 is not observed with receptors belonging to the glucocorticoid receptor subclass.

EXPERIMENTAL PROCEDURES
Isolation of hTIF1 cDNA-A random-primed human fetal liver gt11 expression library (CLONTECH) was screened using the in vitro 32 Plabeled GST-AF2 probe, as described previously (24). A single positive clone (clone 8) containing a 675-base pair insert was isolated, screened again and purified. This insert was then randomly labeled with [ 32 P]dCTP (Megaprime DNA labeling system, Amersham) and used to screen an oligo(dT)-primed ZR75-1 human breast cancer cells cDNA library constructed in ZAPII (24). Several positive clones were isolated and pBluescript phagemids containing the inserts of interest were excised in vivo by coinfection with the R408 helper phage. Inserts were then sequenced by the dideoxy chain termination method using a Sequenase version 2.0 DNA sequencing kit (U. S. Biochemical Corp.). The 5Ј end of the cDNA was obtained by reverse transcriptase-PCR using total RNA from ZR75-1 cells and specific primers generated from the mTIF1 cDNA sequence (27).
Recombinant Vectors-The recombinant vectors allowing expression in Escherichia coli of GST-TFIIB, GST-AF2 (wild-type and mutants), or the plasmid used for in vitro labeling of hRIP140 (pBRIP140) were described previously (24). GST-AF2-AD was constructed by inserting a PCR-generated 96-nucleotide fragment encoding residues 525-556 of mouse ER into the EcoRI site of pGEX-2TK (Pharmacia Biotech Inc.). The nine GST-hTIF1 constructs (wild-type and mut1-mut8) encoding different domains of hTIF1 receptor binding site fused to glutathione S-transferase were constructed by inserting the corresponding DNA fragments produced by PCR from the original clone 8 into the EcoRI site of pGEX-2TK. GST-hRIP140, which encoded a fusion between GST and residues 752-1158 of hRIP140, was generated by inserting a BamHI/ BglII fragment from pBRIP3 (24) into the BamHI site of pGEX2TK.
GST Pull-down Assay-In vitro binding assays were performed essentially as described (24). Briefly, 35 S-labeled proteins (receptors or interacting proteins) were cell-free-synthesized using the TNT lysate system (Promega) and incubated overnight at 4°C with bacterially expressed and purified GST, GST-AF2, GST-AF2-AD, GST-hRIP140, or GST-hTIF1 fusion proteins in the presence of the cognate ligands at micromolar concentrations. Protein interactions were analyzed by SDS-PAGE followed by fluorography (Amplify; Amersham) and quantified using a PhosphorImager (Fujix BAS1000). In some cases, the gel was stained with Coomassie Brilliant Blue (Bio-Rad) prior to fluorography, to visualize the GST fusion proteins present in each track.
Gel Shift Assay-The double-stranded oligonucleotide corresponding to the vitellogenin A 2 estrogen response element (ERE) (40) was 32 Plabeled using Klenow enzyme. Binding reactions (20 l) were performed in the presence or absence of ligands (1 M) on ice for 1 h using 4 l of ER-primed reticulocyte lysate and 30 g of purified GST or GST fusion proteins in 10 mM Tris, pH 7.5, 75 mM KCl, 5% glycerol, 0.5 mM EDTA, 0.5 mM dithiothreitol, 0.1 g/l poly(dI-dC), plus protease inhibitors. The labeled ERE (50 fmol) was then added, incubated for 20 min at room temperature, and analyzed on a 5% polyacrylamide gel.

Cloning of hTIF1 and Interaction with GST-AF2 Probes-To
isolate cDNAs encoding putative mediators of the ligand-dependent activation function of nuclear receptors, we have applied the strategy previously used to clone hRIP140 cDNA (24). We have screened a human fetal liver random-primed cDNA expression library with the 32 P-labeled probe containing the mouse ER hormone binding domain fused to glutathione Stransferase (GST-AF2 probe). We have isolated a single positive clone (clone 8) encoding a 225-amino acid peptide, which interacted directly with the wild-type GST-AF2 probe in an estrogen-dependent manner (Fig. 1A). The binding characteristics of the protein encoded by this clone were similar to those previously observed for hRIP140 and hRIP160 (23), i.e. lack of interaction (i) in the absence of ligand (ii) in the presence of the antiestrogen 4-hydroxytamoxifen (iii) with transcriptionally defective mutants in the conserved amphipathic ␣-helix (Fig. 1A).
The cDNA clone 8 contained a 675-base pair insert, which did not cross-hybridize with hRIP140. When used as a probe in Northern blot analysis of poly(A) ϩ RNA isolated from MCF7 breast cancer cells, this fragment detected a major band of about 4.5 kilobases (data not shown). Sequence analysis confirmed that this cDNA was different from hRIP140 and from sequences present in the data bases at the time of the cloning. We then isolated longer cDNA clones from an oligo(dT)-primed ZR75-1 human breast cancer cell cDNA library and sequenced the 5Ј end of the longest clone (clone 16). This revealed strong homology with the sequence of mouse TIF1, which was present in the data base as the NH 2 -terminal moiety of the mouse T18 oncoprotein (41), and indicated that we have isolated the human TIF1 cDNA. As shown in Fig. 2A, the overall conservation of TIF1 amino acid sequence between mouse (27) and human species appears relatively high (more than 92%; see also Fig. 2B for sequence comparison of the nuclear receptor binding site). Two clusters of amino acids in the NH 2 terminus of the molecule (residues 14 -25 and 107-111 in hTIF1) are, however, highly divergent. By contrast, we found in the ZR75-1 cDNA library a clone coding for an hTIF1 isoform with the same insertion of 34 amino acids already described for the mouse protein (27) suggesting that this variant form, perfectly conserved between the two species, could be of physiological importance.
Using GST pull-down experiments with a 90-kDa peptide in vitro expressed from clone 16, we confirmed that hTIF1 interaction with wild-type GST-AF2 was strongly increased (ϳ14fold) by estradiol (Fig. 1, panel B, lanes 3 and 4, and panel C) but not by the antiestrogens 4-hydroxytamoxifen or ICI164,384 (lanes 5 and 6). This association was also dramatically reduced by various mutations in the conserved amphipathic ␣-helix that abolish AF2 transcriptional activity (7). When either of two pairs of hydrophobic residues (547/548 or 543/544) or the negatively charged glutamic acid residue at position 546 were mutated, the effect of estradiol on hTIF1 binding was reduced to less than 2-fold (Fig. 1C). In these GST pull-down assays, hTIF1 binding to wild-type GST-AF2 in the presence of estradiol (57% of input; 14-fold increase versus control; Fig. 1C) was very similar to that of hRIP140 (56% of input; 12-fold increase versus control; data not shown).
Binding Sites Characterization-The initial hTIF1 cDNA clone 8 isolated from the library encoded a 225-amino acid peptide (Fig. 2B). This fragment overlapped the region between residues 539 and 750 in the mouse protein, which was shown to be the minimal domain sufficient to interact with RXR␣ in the yeast two-hybrid system (27). To define more accurately the border of the nuclear receptor binding site (NRBS) on the hTIF1 molecule, we generated by PCR a GST-hTIF1 vector and several deletion mutants (mut1-mut8; Fig. 2C). We then tested their ability to interact with 35 S-labeled in vitro translated ER in GST pull-down experiments. As shown in Fig. 2C, strong estradiol-dependent interaction (19-fold increase over control) was observed with GST-hTIF1. Deletion of 102 residues from the COOH terminus (mut1) affected only moderately this interaction (10-fold induction in the presence of estradiol). By contrast, when the hTIF1 fragment was reduced to 96 (mut2) or 73 amino acids (mut3), hormone-dependent binding was totally abolished, indicating that an important motif was localized in the last 27 residues of mut1. Starting from mut1, we then generated five other mutants (mut4 -mut8). In GST pull-down assays, hormone-inducible binding of ER was retained with mut4 -mut7 but not with mut8. Together, these results indicate that the minimal NRBS on hTIF1 is localized between residues 716 and 741 and that the 11 amino acids at the NH 2 terminus of mut7 are necessary for ER binding. As shown in Fig. 2D, this sequence exhibits some homologies (conservation of Ser and Leu residues) with sequences in the two NRBS of RIP140 (25).
Since the integrity of the AF2 activation domain core plays a crucial role in the binding of hTIF1 to ER (Fig. 1, A and C), we have examined whether the region of the receptor encompassing this domain was sufficient to support in vitro interaction with hTIF1. The region from amino acid 525 to 556 was fused to GST and used to precipitate labeled hTIF1. As shown in Fig.  3 (lane 5), no specific binding was obtained, even though the level of GST-AF2-AD was slightly higher than those of GST-AF2 wild-type or mut. This suggests that other regions of the ER ligand binding domain are probably required to generate the complete binding site.
Since in vitro binding characteristics of hTIF1 (i.e. the requirement for hormone and for an intact AF2 activating domain) were similar to those obtained for hRIP140 (23)(24)(25), it was possible that these two potential transcriptional coactivators contact the same region of the hormone binding domain. To investigate this possibility, we performed pull-down experiments to test the ability of ER to contact simultaneously hTIF1 and hRIP140. We used GST-hTIF1 or GST-hRIP140 to precipitate, respectively, labeled hRIP140 or hTIF1 bound to ER. As shown in Fig. 4A, labeled ER efficiently bound (almost 50% of input) to GST-hTIF1 in an hormone dependent manner (com- pare lanes 3 and 4). We then preincubated increasing amounts of labeled hRIP140 with ER before precipitation with GST-hTIF1 (lanes 5-9) and looked for the formation of a ternary complex GST-hTIF1⅐ER⅐hRIP140. Whereas binding of ER was still clearly observable, specific precipitation of hRIP140 in these conditions was almost undetectable since for the highest amount of hRIP140 used (lane 9), less than 2% of input material was bound. Moreover, most of this binding was not ER-mediated since a similar amount of hRIP140 was precipitated by GST-hTIF1 in the absence of ER (lane 10). Identical results were obtained in the converse experiment using GST-hRIP140 to pull-down hTIF1 (Fig. 4B). Together these results indicated that, in our pull-down conditions, simultaneous binding of hTIF1 and hRIP140 on ER is not possible.
Interaction of hTIF1 with DNA-bound ER-The electromobility shift assay was used to examine the association of GST-hTIF1 with ER bound to DNA. In vitro expressed ER was incubated with purified GST or GST-hTIF1 before addition of 32 P-labeled ERE. As shown in Fig. 5, the retardation profile obtained in the presence of GST was in agreement with previously published results (42). Binding of ER to DNA was not dependent on the presence of hormone, but the mobility of the complexes varied according to the ligand. In the presence of estradiol, the ER⅐ERE complexes migrated faster than in the absence of ligand or in the presence of 4-hydroxytamoxifen (lanes 1-3), probably reflecting a different conformation of the receptor. Addition of purified GST-hTIF1 did not modify the position or intensity of the retarded bands in the absence of ligand or in the presence of 4-hydroxytamoxifen (compare lanes  1 and 3 to lanes 4 and 6). By contrast, in the presence of estradiol, we observed a shift in the position of the ER⅐ERE complex and a slight decrease in the intensity of the band (compare lanes 2 and 5). This shift was reproducibly obtained even when ER was produced in COS-1 cells or in baculovirusinfected insect cells (data not shown), and it probably corresponds to the formation of an heteromeric complex of GST-hTIF1 and ER on DNA. We have checked that this altered migration was not due to a modification in the estradiol binding capacity of ER or in the dissociation kinetics of hormone after binding of GST-hTIF1 (data not shown). Moreover, similar results were obtained when we used GST-hRIP140 fusion protein (Fig. 5, lane 8), thus confirming previous results showing that, in DNA-dependent assays for protein-protein interaction, the binding of hRIP140 to ER (25) or to RAR/RXR heterodimers (43) was also ligand-dependent. By contrast, no retardation was observed with GST-TFIIB (Fig. 5, lane 10), emphasizing the specificity of this effect.
Interaction of hTIF1 with Other Nuclear Receptors-To study the specificity of hTIF1 interaction with nuclear receptors, we have tested whether other members of the superfamily were able to interact with GST-hTIF1 in the presence of their cognate ligands. As shown in Table I, vitamin D 3 (VDR), thyroid hormone (TR␣1), retinoid X (RXR␣), and retinoic acid (RAR␣) receptors bound to GST-hTIF1 in the presence of their cognate ligand (more than 12% of input). Specific binding (which represents less than 2.5% of input in the absence of ligand) is increased by more than 8-fold in the presence of hormone. In addition, as observed for ER in the presence of antiestrogens, the interaction of RAR␣ was not significantly increased in the presence of the antagonist Ro 41-5253 (44). Moreover, we have checked that chimeric proteins found in acute promyelocytic leukemia and resulting from the fusion of RAR␣ to promyelocytic leukemia protein (PML) in t(15;17) or promyelocytic leukemia zinc finger protein (PLZF) in t(11;17) (45) also retained a strong retinoic acid-dependent interaction with GST-hTIF1 (data not shown).
By contrast, when we evaluated the interaction of the other steroid receptors, namely glucocorticoid (GR), progesterone (PR form A and B), androgen (AR), and mineralocorticoid (MR) receptors, we observed a weak interaction (less than 5% of input) with GST-hTIF1 even in the presence of the cognate hormone. A slight increase in binding to GST-hTIF1 was, however, detectable with MR in the presence of aldosterone. In addition, we analyzed the association of GST-hTIF1 with two orphan receptors that bind DNA as homodimers and behave as transcriptional repressor (COUP-TFII) or activator (hepatocyte nuclear factor 4; HNF4) and observed very low specific interaction representing less than 3% of input (data not shown).
Other transcription factors such as c-Fos, c-Jun, CREB, E1A, or MyoD, shown to bind CBP (46), also failed to interact specifically with hTIF1 in vitro (data not shown). Together these results indicate that, in our experimental conditions, hTIF1 selectively interacts in a ligand-dependent fashion only with specific members of the nuclear receptor superfamily.

DISCUSSION
Using a protein-protein interaction assay, we have searched for human transcription intermediary factors able to bind the hormone-dependent activation domain of the estrogen receptor. We have isolated a single cDNA clone, which appears to be the human counterpart of mTIF1, a murine factor isolated using a yeast genetic screen based on the capacity to stimulate RXR␥ activity (27). This factor, initially characterized as a fusion protein with the equivalent of the human proto-oncogene B-RAF (41), exhibits interesting structural features. Several domains (a RING finger, B box fingers, a coiled-coil domain, a plant homeodomain finger, and a bromodomain) potentially involved in protein-DNA or protein-protein interactions (47)(48)(49) have been identified. Most of the consensus residues in these domains are conserved between mouse and human, and the overall conservation between the two species is around 93% (for comparison, another potential coactivator for nuclear receptor, hTRIP1/mSUG1 is 99.3% conserved) (29). The region of hTIF1 located in the middle of the molecule is one of the most divergent from mTIF1, and it also contains the potential alternative splicing site generating an isoform with a 34-amino acid insertion perfectly conserved between the two species.
We have shown that a short sequence of 26 residues is sufficient for the in vitro binding of hTIF1 to the estrogen receptor in the presence of estradiol. This domain presents significant homologies with the two receptor binding sites characterized on hRIP140 (25), suggesting that they could contact the same interface on nuclear receptors. On the other hand, our results also indicate that the ␣-helix of ER, which contains the AF2 activation domain, is required but not sufficient for the interaction between the receptor and hTIF1. This suggests that other parts of the hormone binding domain participate in the formation of the binding interface and is apparently not in accord with several reports, which have shown that the amphipathic ␣-helix from mouse RXR␣ or ␤ (10, 50) and chicken TR␣ (9) can autonomously transactivate when fused to a DNAbinding domain. However, it should be noted that, in all three cases, transcriptional activation was lower than that obtained with the full-length hormone binding domain, suggesting that only suboptimal configurations of the activation domain are generated in such chimeric proteins and thus possibly explaining the undetectable in vitro association with hTIF1. Moreover, it has been shown in the case of mSUG1 that the AF2-AD, while essential, was also not sufficient for an efficient interaction in yeast two hybrid experiments (29).
Recent studies based on the determination of the crystal structure of different nuclear receptor hormone binding domains (11)(12)(13) have demonstrated that in the unliganded  1 and 4). Electromobility shift assay was then performed as described under "Experimental Procedures" using 32 P-labeled ERE (50 fmol). RXR␣, the amphipathic helix that contains the AF2 activation domain points away from the core of the molecule. In the liganded RAR␥, this helix (H12) is folded back and residues in the core of AF2-AD establish contacts both with the ligand and other part of the HBD such as helix 4. It seems therefore likely that helix 12 alone, while necessary, is insufficient for binding proteins such as hTIF1 and that the interface is rather a tridimensional structure generated upon conformational change of the HBD (involving the swing of H12). In agreement with this hypothesis, it has been shown that a point mutation in H4 of RAR␥ (K264A) or RXR␣ (R307A) decreases the in vitro interaction of mTIF1 (51). The hypothesis that different parts of the HBD participate in the formation of an active binding site for putative transcriptional coactivators could explain our observation showing that only a subgroup of nuclear receptors (ER, TR, RAR, RXR, and VDR) exhibit hormone-dependent interaction with hTIF1. The structure of the AF2-AD core is so well conserved among all members of the superfamily that the binding specificity probably involves other regions of the HBD. It is striking to note for instance that the important position corresponding to the above-mentioned Lys-264 in RAR␥ is occupied by a charged residue only in receptors that bind hTIF1 (51). From a more general point of view, the subfamily of steroid receptors that do not bind hTIF1 (GR, PR, AR, and MR) have strong sequence homologies both in the DNA and hormone binding domains (52,53). A parallel could be made between the discriminatory residues of the P-box in the DNA-binding domain, which select the specific half-site of the response element (54,55), and some amino acids in the HBD yet unidentified, which allow the interaction with hTIF1. The significance of this binding specificity is unknown, but it should be noted that the other potential coactivators (TRIP/SUG1 and hRIP140) that interact with ER also bind TR, RAR, and RXR but not GR (25,28,29). 2 The opposite specificity seems to take place for the effects of PML on nuclear receptors transactivation since PR, MR, AR, and GR are preferentially stimulated by PML cotransfection (56). Similarly, RSP5 potentiates GR and PR but not ER transcriptional activities (57). By contrast, SRC-1/GRIP1/TIF2 (30 -32), CBP (37), and hbrm/hSNF2 (58,59) modulate the activity of receptors from the two subgroups, and RAP46 (60) also binds to receptors without such a discrimination. On the other hand, ARA70 appears to be a specific coactivator for the androgen receptor (61).
Our results and those from Chambon's laboratory (27,29) have provided several lines of evidence supporting the idea that TIF1 can mediate at least in part the nuclear receptor hormone dependent activation function AF2. However, the mechanisms involved remain to be elucidated. We have confirmed that hTIF1 as its mouse homolog is not able to interact in vitro with either TBP or TFIIB (data not shown), and it is possible that TIF1 does not function as a bridging protein between receptors and basal transcription machinery but rather as a mediator of chromatin remodeling. For instance, hTIF1 could facilitate nuclear receptor function by countering transcriptional repression mediated by histones or other chromatin-associated proteins as suggested in the case of the SWI-SNF complex (62).
Several factors have been shown to interact in a liganddependent manner with nuclear receptors hormone binding domains and therefore could play a role of mediator in transcriptional activation as already suggested for SRC-1 (30 -32), RIP140 (25,26), or CBP/p300 (34 -37). It will be now very important to find out what are their respective roles and whether their activities are essential or not, additive or syner-gistic. Concerning hTIF1 and hRIP140, our data suggest that their binding on the receptor is mutually exclusive at least in vitro. However, it is obvious that, in the nucleus, additional factors could allow simultaneous binding of these two proteins. It is also possible that the recruitment of these receptor interacting proteins is sequential, one acting after the other. Depending on the cell type and/or the promoter context, specific binding of only one type of intermediary factor could take place on the receptor, and such a mechanism could be involved, at least partly, in tissue-specific regulation of gene expression. Further studies such as gene inactivation will be necessary to define the exact role of these factors in transcriptional control by nuclear hormone receptors.