Modulation of Induction Properties of Glucocorticoid Receptor-Agonist and -Antagonist Complexes by Coactivators Involves Binding to Receptors but Is Independent of Ability of Coactivators to Augment Transactivation*

Coactivators such as TIF2 and SRC-1 modulate the positioning of the dose-response curve for agonist-bound glucocorticoid receptors (GRs) and the partial agonist activity of antiglucocorticoid complexes. These properties of coactivators differ from their initially defined activities of binding to, and increasing the total levels of transactivation by, agonist-bound steroid receptors. We now report that constructs of TIF2 and SRC-1 lacking the two activation domains (AD1 and AD2) have significantly less ability to increase transactivation but retain most of the activity for modulating the dose-response curve and partial agonist activity. Mammalian two-hybrid experiments show that the minimum TIF2 segment with modulatory activity (TIF2.4) does not interact with p300, CREB-binding protein, or PCAF, which also modulates GR activities. DRIP150 and DRIP205 have been implicated in coactivator actions but are unable to modulate GR activities. The absence of synergism by PCAF or DRIP150 with SRC-1 or TIF2, respectively, further suggests that these other factors are not involved. The ability of a TIF2.4 fragment (i.e. TIF2.37), which is not known to interact with proteins, to block the actions of TIF2.4 suggests that an unidentified binder mediates the modulatory activity of TIF2. Pull-down experiments with GST/TIF2.4 demonstrate a direct interaction of TIF2 with GR in a hormone-dependent fashion that requires the receptor interaction domains of TIF2 and is equally robust with agonists and most antiglucocorticoids. These observations, which are confirmed in mammalian two-hybrid assays, suggest that the capacity of coactivators such as TIF2 to modulate the partial agonist activity of antisteroids is mediated by the binding of coactivators to GR-antagonist complexes. In conclusion, the modulatory activity of coactivators with GR-agonist and -antagonist complexes is mechanistically distinct from the ability of coactivators to augment the total levels of transactivation and appears to involve the binding to both GR-steroid complexes and an unidentified TIF2-associated factor(s).

Coactivators are molecules that have recently been found to be important components of the transactivation activity of ste-roid hormone receptors (1)(2)(3). Among the first discovered, and the most widely studied, are the p160 coactivators of SRC-1 (4), TIF2/GRIP1 (5,6), and AIB1 (7). The coactivators were originally defined as factors that increase the total amount of induced gene product with saturating, or pharmacological, concentrations of hormone (4 -7). For the p160 coactivators this action is thought to involve domains with intrinsic transactivation activity (8 -10), called activation domains (ADs). 1 For many of the coactivators, their intrinsic transactivation activity may be related to their histone acetyltransferase (HAT) activity (1,3,11). In addition, and for those coactivators without HAT activity like TIF2/GRIP1 (8), coactivator recruitment of comodulators with HAT activity, such as CBP (8,12), p300 (12,13), and PCAF (14), is considered important. In general, the comodulators are thought to bind only to the p160 coactivators (13,15). However, PCAF, which was initially identified as a binder of p300/CBP (16), has also been reported to bind to glucocorticoid receptors in pull-down assays in a ligand-independent fashion (14). Furthermore, the ligand-dependent association of PCAF with retinoic acid receptors is different from its association with SRC-1 or p/CIP (ϭAIB1) (17). The multiprotein complexes called DRIP/TRAP/Mediator are thought to be recruited by the above coactivator-comodulator complexes to relay the original signal from the steroid receptors (18,19). These complexes are proposed then to modify other components of the transcriptional complex to give the final transcriptional response.
We have recently shown that another property of the p160 coactivators is to modulate both the dose-response curve of agonists, i.e. the concentration of steroid required for halfmaximal induction (EC 50 ), and the partial agonist activity of antagonists. This behavior has been seen for glucocorticoid receptors (GR) (20 -22) and progesterone receptors (23). This modulation of the dose-response curve (or EC 50 ) and of the partial agonist activity of antisteroids has important theoretical and clinical consequences. The EC 50 for induction of most genes is near the physiological levels of hormone. Therefore, changes in the EC 50 of one steroid-responsive gene versus another will both cause different levels of induction from the two genes by endogenous steroids and provide a sensitive mechanism for the differential control of expression of these two genes during development, differentiation, and homeostasis. Antagonists are extremely useful in endocrine therapies to inhibit the actions of endogenous hormones in situations such as inflammation (24), conception (25), and hormone-dependent cancers (26,27) but often produce undesirable side effects. If one could increase the partial agonist activity for genes other than the one to be suppressed, this would greatly reduce the number of side effects that result from the indiscriminate repression of all responsive genes.
Interestingly, the overexpression of corepressors (SMRT and NCoR) (28,29) often produces a response opposite to that of increased coactivators. Thus, corepressors can cause both a right shift in the dose-response curve to higher EC 50 values and a decrease in the partial agonist activity of antisteroids for GR (20,30) and progesterone receptors (23, 30). Furthermore, the ability of added SMRT to antagonize the actions of overexpressed TIF2 suggests that the final responses are determined by an equilibrium reaction with receptors that depends upon the ratio of coactivators to corepressors (20).
Despite the major quantitative effects on steroid receptor gene transactivation that are caused by varying the intracellular levels of coactivators, only general features of the mechanism by which coactivators modulate steroid receptor induction properties are known. The ability of coactivators to augment the total amount of induced gene transcripts requires coactivator binding to the agonist-bound receptors and may result from more efficient binding/recruitment of transcription complex core factors (31,32). In contrast, the conventional model of steroid hormone action is silent on how coactivators might change either the EC 50 , and the position of the doseresponse curve, or the partial agonist activity of antisteroids. In fact, the current thinking is that coactivators do not interact with antagonist-bound receptors (1,3,11), in which case the modulation of the partial agonist activity of antisteroid complexes by coactivators would have to involve indirect interactions.
Whereas the ability of coactivators to increase the total levels of transactivation with receptor-agonist complexes proceeds via binding to receptors, it is not known if similar interactions are required for the modulation of the EC 50 , and the dose-response curve, of receptor-agonist complexes. One indication that separate pathways are involved comes from studies with TIF2, which does not contain the intrinsic HAT activity that is associated with many coactivators (1,3,11). TIF2 binds to CBP (8), although, which does have HAT activity (12). However, the ability of TIF2 to modulate GR activities was found to be independent of the HAT activity of CBP (21), which can itself modulate GR transactivation properties (20). A second indication of separate pathways is the observation that several factors can alter the dose-response curve and partial agonist activity of GR complexes in a manner that is independent of their effects on total transcriptional activity of GR (20,30,(33)(34)(35). Finally, a recent report indicates that TIF2 acts via a ratelimiting intermediate or step that does not regulate the total levels of receptor-mediated gene induction (21).
The purpose of this study, therefore, was to examine in greater detail whether the intrinsic ability of a coactivator to increase the total transactivation of GR is required for its ability to modulate the EC 50 , or dose-response curve, of GRagonist complexes and the partial agonist activity of GR-antagonist complexes. We took the approach of asking if the minimal domains of coactivators that are required for GR modulation are the same or different from those needed for conventional coactivator activity. At the same time, we inquired if the effects of this minimal modulatory domain could be replaced by any other proteins that are thought to interact with the p160 coactivators. The ability of receptor-antagonist complexes to interact with coactivators was also determined. From these studies, we conclude that the modulation of GR transactivation activi-ties involves direct interactions of coactivators with both GRagonist and -antagonist complexes in a manner that is mediated by parameters that are unrelated to both the conventional coactivator domains and known coactivator-associated proteins.
Antibodies-The monoclonal mouse anti-GAL DBD antibody (sc-510) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA).
Cell Culture and Transient Transfection-Monolayer cultures of CV-1 cells were grown in Dulbecco's modified Eagle's medium (Invitrogen) supplemented with 10% fetal calf serum (Biofluids Inc., Rockville, MD) at 37°C in a humidified incubator (5% CO 2 ). Transient transfections were achieved using calcium phosphate precipitation (39) or 1.0 l of LipofectAMINE Reagent (Invitrogen) or 0.5 l of FuGENE 6 (Roche Molecular Diagnostics) in 0.3 ml of Opti-MEM I (Invitrogen) without serum in 24-well plates (10-fold more in 60-mm dishes). As a control for the transfected cofactors, equimolar amounts of the empty plasmid vector were added. The total transfected DNA was adjusted to 300 ng/well of a 24-well plate (or 3 g/60-mm dish) with pBluescriptII SK ϩ (Stratagene). Renilla TK or TS (5-10 ng/well of a 24-well plate) was included as an internal control. Cells were incubated with plasmid DNA, Opti-MEM I, and LipofectAMINE or FuGENE 6 reagent for 24 h, after which this mixture was replaced by the normal media (10% fetal calf serum, Dulbecco's modified Eagle's medium). The cells were induced with steroids for 20 -24 h and then harvested. The cells were lysed and assayed for reporter gene activity using the Dual Luciferase Assay reagents according to manufacturer's instruction (Promega, Madison, WI). Luciferase activity was measured by an EG&G Berth-hold's luminometer (Microlumat LB 96P). The data were normalized either for total protein or Renilla null luciferase activity.
Mammalian Two-hybrid Assays-The recommended procedure for the Mammalian Matchmaker two-hybrid assay kit (Clontech) was modified slightly by changing from a chloramphenicol acetyltransferase reporter to a Luciferase Reporter pFRLuc (Stratagene), which is under the control of five repeats of the upstream activating sequence for the binding of GAL4.
Bacterial Expression of Proteins-pGEX series of plasmids were transformed into Escherichia coli (BL21; Amersham Biosciences) according to the manufacturer's procedure. A single colony was picked and inoculated into 3 ml of LB broth with 100 ng/ml ampicillin. After overnight culture, 1 ml of bacterial culture was diluted into 50 ml of LB broth containing 100 ng/ml ampicillin, shaken at 37°C for 2 h, adjusted to 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside, and shaken at 37°C for another 3 h. The cells were harvested by centrifugation, washed once with phosphate-buffered saline, resuspended in 10 ml of phosphate-buffered saline, and sonicated for 30 cycles at 30% of maximum power. The supernatant was collected for use after centrifugation (5000 ϫ g for 20 min).
In Vitro Transcription and Translation Assays-pRBAL-GR (1 l), 40 l of Promega TNT (SP6) quick coupled transcription/translation system master mixture, and 2 l of [ 35 S]methionine (Amersham Biosciences) were brought up to a total volume of 50 l with H 2 O and incubated at 30°C for 90 min according to the manufacturer's (Promega) recommendations. Different steroid hormones (Dex, Dex-Mes, Dex-Ox, progesterone, and RU486) were added into the transcriptiontranslation reaction during the 30°C incubation to both increase the stability of the synthesized GR and to cause activation of the GR-steroid complex.
Pull-down Assays-Sonicated bacterial lysate (0.5 ml) containing overexpressed GST, GAL/TIF2.4, or GAL/TIF2.4m123 was incubated with 20 l of glutathione-Sepharose 4B beads for 1 h at 0°C. The mixture was centrifuged (12,000 ϫ g), the supernatant was discarded, and the pellet was washed with phosphate-buffered saline (3 ϫ 1 ml). Each 20-l sample of immobilized GST or GST-chimera was then incubated overnight with 10 l of hormone prebound, activated, 35 S-labeled GR translation product. The matrix was washed with phosphate-buffered saline (4ϫ 1 ml). The immobilized proteins were removed from the beads by heating at 90°C for 5 min in 20 l of 2ϫ SDS loading buffer. The proteins were then separated on 10% SDS-PAGE gels and the bound GR was located by autoradiography.
Statistics-Unless otherwise noted, all experiments were performed in triplicate several times. The error bars in graphs of individual experiments correspond to the S.D. of the triplicate values. KaleidaGraph 3.5 (Synergy Software, Reading, PA) was used to determine a leastsquares best fit (R 2 was almost always Ն0.95) of the experimental data to the theoretical dose-response curve, which is given by the equation derived from Michaelis-Menten kinetics of y ϭ [free steroid]/([free steroid] ϩ K d ) (where the concentration of total steroid is approximately equal to the concentration of free steroid because only a small portion is bound), to yield a single EC 50 value. The values of n independent experiments are then analyzed for statistical significance by the twotailed Student's t test using the program "InStat 2.03" for Macintosh (GraphPad Software, San Diego, CA). In paired analyses, all comparisons in a given figure are performed at the same time with the same seeding of cells and then analyzed for n different experiments.

Minimum Domain of Coactivator TIF2/GRIP1 for Modulation of GR Properties-TIF2/GRIP1
is one of the three p160 coactivators that increase the total induced levels of gene product with steroid receptors (5,6). Of the various domains of TIF2/GRIP1 that have been identified (Fig. 1A), the receptor interaction domains (RIDs) and the two activation domains (AD1 and AD2) have been found to be required for full coactivator activity (8). Amino acids 1057-1109 of TIF2/GRIP1 are within the AD1 domain and are required for CBP/p300 binding (8,40,41). This binding of CBP/p300 is thought to be particularly important for the coactivator activity of TIF2/GRIP1 because TIF2/GRIP1 does not possess any intrinsic HAT activity (8) but rather relies on the binding of CBP/p300 to provide HAT activity.
We previously demonstrated that TIF2 causes a left shift in the dose-response curve for agonist complexes, and increased partial agonist activity of antagonist complexes, of both GR and progesterone receptors (20,21,23). To identify the domain(s) of TIF2/GRIP1 that are required for this modulation, we now use an abbreviated assay in which increases in the activity of a subsaturating concentration of steroid, expressed as percent of maximal induction by saturating concentrations of agonist, have been shown to be diagnostic of a left shift in the doseresponse curve to lower EC 50 values (20,21,42,43). The main advantage of this abbreviated assay is that it requires many fewer samples to reach the same conclusion. The changes in partial agonist activity of the antisteroid were assessed by the normal method of expressing the activity of 1 M antagonist as percent of maximal induction by saturating concentrations of agonist (20,21,23,30,33).
Using this abbreviated assay, we assessed the modulatory activity of GRIP1 mutants lacking the activation domains AD2 and/or segments of AD1 (Fig. 1B). GRIP⌬1095-1106 is missing a portion of AD1 that is involved in binding to p300 (40). However, the ability of this mutation to cause a left shift of the GR dose-response curve, as determined from the increased activity of 1 nM Dex relative to that of the samples without added GRIP, is indistinguishable from the full-length GRIP. This mutation also displays an undiminished capacity to increase the partial agonist activity of the antiglucocorticoid Dex-Mes. Similarly, GRIP⌬1057-1109, which lacks almost all of the AD1 domain (8,40,41), and GRIP(5-1121)⌬AD1, which is missing both AD1 and AD2 domains, display the same amount of modulatory activity as does the intact GRIP1. We, therefore, conclude that none of these deleted residues are required for the modulatory activity of GRIP1.
The conventional definition of a coactivator is a molecule that increases the total level of steroid receptor-induced gene product in the presence of saturating concentrations of an agonist (1,3,11). We, therefore, asked whether the coactivator activity segregates with the modulatory activity of the above GRIP1 mutants. As shown in Fig. 1C (average data of n ϭ 4), deletion of the AD1 domain reduces the coactivator activity of GRIP1 (paired p ϭ 0.020 for ⌬1057 versus GRIP1). Interestingly, deletion of both AD1 and AD2 to give GRIP-(5-1121)⌬AD1 did not further decrease the coactivator activity. Nevertheless, when compared with Fig. 1B, it appears that there is no correlation between the modulatory activity and coactivator activity of GRIP1 with the various deletions.
These questions were examined in greater detail using a larger set of deletions in TIF2, which is the human ortholog of the mouse GRIP1. TIF2 is 94% identical to GRIP1 (44) and has no known different properties from GRIP1. All of the TIF2 modulatory activity was found to be contained in TIF2.1 (624 -1287) and TIF2.3 (624 -1179) ( Fig. 2A, top and middle panels, respectively). These results suggest that neither the aminoterminal half of TIF2, nor the AD2 domain, are required for the activity of TIF2 in our assays. TIF2m123 contains a LXXLL to LXXAA mutation in each of the three RIDs (Fig. 1A) and has been described to no longer bind to steroid receptors (8,45). Thus, the inactivity of TIF2m123 in our assay (top panel) indicates that TIF2 binding to GR is required for the modulation of GR activities. Not surprisingly, the TIF2 AD2 domain by itself (TIF2.2) is inactive (top panel). Further removal of AD1 and the auxiliary RID domain (41) from TIF2.3 gives TIF2.4, which still retains almost all of the activity of the wild type TIF2 (bottom panel). In experiments examining multipoint dose-response curves directly, the average left shift is 2.41 Ϯ 0.17-fold (S.E., n ϭ 16, p Ͻ 0.0001) for TIF2 and 1.80 Ϯ 0.09-fold (S.E., n ϭ 13, p Ͻ 0.0001) for TIF2.4 (data not shown). For comparison, the -fold left shift in the dose-response curve with TIF2m123, which is unable to interact with GR, is 1.09 Ϯ 0.08 (S.E., n ϭ 6). Thus, TIF2.4 is the smallest fragment of TIF2 that we have found with nearly the full modulatory activity of the intact TIF2.
We next asked how much coactivator activity is preserved in TIF2.3 and TIF2.4. As shown in Fig. 2B, TIF2.3 is just as effective as the full-length TIF2 in augmenting the total level of gene induction by GR with saturating concentrations of the agonist Dex. In contrast, TIF2.4 has Ͻ40% of the activity of TIF2 ( Fig. 2B; average -fold increase is 1.61 Ϯ 0.16 for TIF2.4 versus 2.69 Ϯ 0.22 for TIF2 (ϮS.E., n ϭ 11, paired p Ͻ 0.0001)). Thus, whereas TIF2.4 retains most of the modulatory activity of TIF2, it has lost more than half of the coactivator activity. Collectively, the data with TIF2 and GRIP1 mutants suggest that the modulation of GR properties is independent of the coactivator activity of TIF2/GRIP1. This implies that the two activities of TIF2, coactivator and modulator, are expressed by different mechanisms.
The active mutant TIF2.3 has been divided in half to give TIF2.5 (amino acids 624 -869) and TIF2.7 (amino acids 870 -1179) (8). TIF2.7 (870 -1179) functions as a dominant suppressor in the coactivator and modulator assays (21), whereas TIF2.5 has weak activity in both assays (data not shown). When equal molar amounts of TIF2.5 and TIF2.7 plasmids are cotransfected, the suppressor activity of TIF2.7 is again dominant (data not shown). Thus, a simple mixture of the two halves of the TIF2.3 protein is not sufficient to reconstitute the activity of the intact TIF2.3.
Region of Coactivator SRC-1 Retaining Modulatory Activity of GR Properties-SRC-1 is another p160 coactivator (4) that modulates GR activities (20). Whereas generally similar in organization to TIF2/GRIP1, notable differences of SRC-1 include an activation domain at the amino-terminal end (AD1), a fourth RID in the region of amino acids 1139 -1441 (9), and a repressive or negative domain in the COOH-terminal 56 amino acids (46) of SRC-1 (Fig. 3A). Interestingly, the two activation domains (AD1 and AD2), the three central RIDs, and the Q-rich domain all make minimal contributions the bulk of the modulatory activity of SRC-1 (Fig. 3A). A more detailed analysis of SRC-1 versus SRC-1/1139 -1441 (Fig. 3B) confirms that this activity of SRC-1 is largely contained in the COOH-terminal 300 amino acids. This region is known to contain a RID that shows very strong binding to GRs (47).
SRC-1/1139 -1441 has no ability to increase the total activity of GR-Dex complexes (data not shown). As previously reported, SRC-1 displays very little conventional coactivator activity with GR (4,20). Therefore, we cannot say whether SRC-1/ 1139 -1441 has less coactivator activity than SRC-1. However, it is clear that the modulatory activity of SRC-1 can be restricted to a domain (amino acids 1139 -1441) that is devoid of conventional coactivator activity. In this respect, the modulatory activity of SRC-1, like that of TIF2, is independent of any changes by SRC-1 in the total activity of saturating concentrations of agonist.
PCAF Modulates GR Properties but Does Not Synergize with SRC-1-Amino acids 1139 -1250 of SRC-1 have been reported to interact with PCAF (48). Because these sequences are contained within the COOH-terminal modulatory fragment of SRC-1 (Fig. 3), PCAF is a potential downstream target for SRC-1 in its modulation of GR activity. PCAF is also considered to be a target of CBP and other coactivators, such as AIB1, although there is no report of interactions of PCAF with TIF2/ GRIP1 (14,16,17). CBP has recently been reported to have little to no direct binding to receptors (15,49) but does bind to PCAF, which directly interacts with various steroid receptors including GR (14). Collectively, these data suggest that the modulatory effects of changing concentrations of GR, and of coactivators and corepressors, could be mediated by PCAF. This hypothesis was tested directly by looking at the effects on GR transactivation properties of cotransfected PCAF. As seen in Fig. 4A, PCAF causes an ϳ2.8-fold left shift in the GR dose-response curve and an increase in the partial agonist activity of DM. At the same time, the added PCAF produces a slight increase in total activity (1.35 Ϯ 0.29-fold increase) and a greater increase in -fold induction (2.57 Ϯ 0.77-fold increase, Ϯ S.E., n ϭ 3).
The ability of PCAF to alter GR transactivation properties, coupled with its binding to the COOH-terminal active region of SRC-1 (SRC-1/1139 -1441) suggests that PCAF could be a necessary component of SRC-1 modulatory activity. To test this hypothesis, we asked if the combination of PCAF and SRC-1/ 1139 -1441 afforded any additional, or synergistic, response over the sum of the individual components. To maximize our ability to observe any combined response, we used submaximal amounts of each cofactor. The addition of 0.5 units of SRC-1/ 1139 -1441 (Fig. 4B) and PCAF (data not shown) causes less shift in the dose-response curve, and change in the partial agonist activity of Dex-Mes, than does 1 unit of each factor. Importantly, the combination of 0.5 units of SRC-1/1139 -1441 and PCAF is not significantly more effective than 0.5 units of SRC-1/1139 -1441 alone (Fig. 4B). Because PCAF does not increase, either synergistically or even additively, the effects of SRC-1/1139 -1441 on the GR transactivation properties, we conclude that PCAF does not participate in the modulatory activity of SRC-1/1139 -1441.
Modulatory Activity of TIF2.4 Does Not Involve Binding to CBP, p300, or PCAF-While the abilities of SRC-1 and TIF2 to modulate GR properties appear to be identical (20), the mechanistic details could be different. Therefore, we inquired whether PCAF might contribute to the actions of TIF2. There is no report of PCAF binding to TIF2/GRIP1 (14,16,17). This question was pursued by asking if PCAF can interact in a mammalian two-hybrid assay with those fragments of TIF2 that possess modulatory activity (TIF2.3 and TIF2.4 of Fig.  2A). Under conditions where a chimera of PCAF fused to the GAL4 DNA-binding domain (GAL/) interacts with a VP16/GR chimera bound with Dex, there was no interaction of GAL/ PCAF with either TIF2.3 or TIF2.4 fused to VP16 (Fig. 5A). Therefore, we conclude that PCAF does not interact with TIF2.4, which is the modulatory domain of TIF2, and is not involved in TIF2 modifications of GR properties.
The transactivation activity of TIF2 is thought to require the activation domains, AD1 and AD2 (Fig. 1A), with CBP and p300 binding to TIF2 being localized to the AD1 domain (8,40,41). The observation in Fig. 2A that TIF2.4, which lacks both AD1 and AD2, still augments the total activity of GR-Dex complexes suggests that there might be some weak binding of CBP or p300 to TIF2.4. Using the same mammalian two-hybrid assay as in Fig. 5A, we investigated the interactions of CBP, and p300, with TIF2 fragments containing (TIF2.3) and lacking (TIF2.4) the AD1 domain. Strong interactions are seen between CBP or p300 and TIF2.3 (Fig. 5B). However, no interaction of CBP or p300 with TIF2.4 is observed. Therefore, we conclude that the modulatory activity of TIF2.4 does not involve PCAF, CBP, or p300.
DRIP150 and DRIP205 Do Not Modulate GR Transcriptional Properties-Gene induction by steroid/nuclear receptors has been proposed to proceed via sequential complexes, one of which is the DRIP (19) or TRAP (18) complex. Two components of the DRIP/TRAP complex that are of particular interest are DRIP205 and DRIP150. DRIP205 may have an effect on the dose-response curve of some vitamin D analogs (50) and interacts, with the help of DRIP150, with GR (51). Furthermore, DRIP150 has been shown to interact with the AF-1 domain of GR (51). Finally, DRIP150 is a component of mediator-like complexes. Another component of mediator complexes is hSur2 (52), which we have recently found is also capable of modulating GR activities (22). Despite these many interesting associations, the cotransfection of neither DRIP150 nor DRIP205 has any modulatory activity on the transactivation properties of GR (data not shown). Even when high concentrations of GR are used, thereby allowing us to see the involvement of species downstream of the previously identified rate-limiting step X (21) that are "kinetically invisible" at low concentrations of GR (35), there is no effect of DRIP150 or DRIP205 (data not shown). Thus, DRIP150 and DRIP205 are not limiting factors for the modulation of GR transcription properties under our conditions in CV-1 cells.
Whereas the DRIPs are inactive by themselves, they might synergize with TIF2. To address this possibility, we asked if the co-addition of DRIP150 could significantly increase the effects of TIF2. We found that the addition of 70 ng of DRIP150 to 10 ng of TIF2 was no more effective in changing the GR doseresponse curve and amount of partial agonist activity than was 20 ng of TIF2 (average Ϯ S.E. (n ϭ 3) values for EC 50 and percent agonist activity are for control (13.2 Ϯ 1.7 nM; 21.7 Ϯ 0.4%), 10 ng of TIF2 (8.5 Ϯ 1.2 nM; 31.8 Ϯ 1.9%), 10 ng of TIF2 ϩ 70 ng of DRIP150 (6.7 Ϯ 1.7 nM, 41.8 Ϯ 2.5%), and 20 ng of TIF2 (6.9 Ϯ 1.8 nM; 38.3 Ϯ 2.5%)). Therefore, we conclude that DRIP150 does not synergize with TIF2 to modify GR transcriptional properties.
TIF2 Modulatory Activity Requires the Binding of Additional Factors-The modulatory activity of TIF2.4, and TIF2, is presumed to involve TIF2 as a physical intermediate that connects GR to some other factor. This connection is broken with TIF2m123, which is inactive ( Fig. 2A). However, the above experiments suggest that none of the potential binders of TIF2.4 are involved because they are not able to augment the activity of endogenous TIF2 and other coactivators (53). In fact, no proteins are known to bind to TIF2.4 other than GR and the other steroid/nuclear receptors. To obtain more direct evidence for a possible novel binder to TIF2.4, we asked if TIF2.4 fragments lacking the three RIDs ( Fig. 2A), and thus do not bind GR, can competitively inhibit the actions of TIF2.4 by titrating out our putative TIF2-binding factor. The fragments that we prepared are TIF2.37, which is TIF2.4 minus the RIDs, and TIF2.38, which is a 68-amino acid extension of TIF2.37 and contains the auxiliary domain but not the AD1 transactivation domain (Fig. 6A) (41). As shown in Fig. 6B, added TIF2.37 and TIF2.38 each block the modulatory activity of TIF2.4. Thus, TIF2.37 and TIF2.38 prevent the average 1.90 Ϯ 0.26-fold left shift (S.E., n ϭ 9, p ϭ 0.0080) in the GR dose-response curve by TIF2.4. Similarly, both TIF2.37 and TIF2.38 reverse the 1.59 Ϯ 0.08-fold (S.E., n ϭ 9, p ϭ 0.0001) increase in Dex-Mes partial agonist activity by TIF2.4. TIF2.37 and TIF2.38 by themselves are unable to shift the dose-response curve to the left, or to increase the partial agonist activity of Dex-Mes, and may even reduce the activities of GR, consistent with TIF2.37 and TIF2.38 antagonizing endogenous coactivators (data not shown). We, therefore, conclude that TIF2.37 and TIF2.38 competitively bind a factor(s) that is required for TIF2.4 to modulate GR transactivation properties.
TIF2 Modulatory Activity Proceeds via Binding of TIF2 to Agonist-and Antagonist-bound GRs-A distinguishing characteristic of the modulation of GR activities by coactivators is that the properties of both GR-agonist and GR-antagonist complexes are affected (20,21). In general, coactivators are not thought to interact with antagonist-bound receptors (1, 3, 11). However, this question has not yet been examined directly with GR. We, therefore, used a mammalian two-hybrid assay to assess the ability of TIF2 to associate in intact cells with GR complexes of the agonist Dex and various antiglucocorticoids, i.e. Dex-Mes, progesterone, Dex-Ox, and RU486. Each of these antiglucocorticoids are effective antisteroids and display low amounts of partial agonist activity with the GREtkLUC reporter in CV-1 cells (20 -22, 33, 38). A large, Dex-dependent increase in luciferase activity is seen with the chimeras GAL/ GRIP1 and VP16/GR that is not observed with GAL plus VP16/ GR, and thus is specific for the presence of GRIP1 (Fig. 7A). A similar, strong GRIP1-specific increase in reporter activity is obtained for GR bound by the antiglucocorticoids Dex-Mes, Dex-Ox, and progesterone. The response with RU486 is usually somewhat weaker.
We next used this assay to examine the properties of TIF2.4, which is the smallest segment of TIF2 to show strong modulatory activity (see Fig. 2A). As a control, we used GAL/ TIF2.4m123, where each of the three LXXLL sequences of the RIDs have been mutated to eliminate GR binding to TIF2 (8,45). Compared with TIF2.4m123, Dex and most of the antiglucocorticoids elicit a strong interaction between GR and TIF2.4. The only antiglucocorticoid that is unable to promote an interaction between GR and TIF2.4 is RU486. The insert shows that the reduced signals with GAL/TIF2.4m123 are not because of lower amounts of expressed chimeric protein.
Cell-free pull-down assays were performed with GST/TIF2.4 and [ 35 S]methionine-labeled, in vitro translated full-length GR in an effort to obtain additional evidence for the binding of TIF2 to the various GR-steroid complexes. Significantly increased binding of GR to GST/TIF2.4 over that seen to GST and GST/ TIF2.4m123 controls is seen for Dex and the antiglucocorticoids Dex-Mes, Dex-Ox, and progesterone (Fig. 7B). No binding to TIF2.4 is seen with RU486-bound GR. These results are very similar to those of the mammalian two-hybrid data (Fig. 7A). Collectively, the above data argue that amino acids 624 -1010 of TIF2 are sufficient to mediate the modulatory activity of TIF2/GRIP1 on GR transcriptional properties via a mechanism that involves the direct binding of TIF2/GRIP1 to both GRagonist and GR-antagonist complexes.

DISCUSSION
The ability of p160 coactivators to modulate both the position of the dose-response curve of GR-agonist complexes, and the partial agonist activity of GR-antagonist complexes, is a relatively new function of p160 coactivators (20,21). The originally described function of coactivators is to increase the total levels of induced gene product by receptor-steroid complexes. We have now used a variety of truncated forms of TIF2/GRIP1 and SRC-1 to show that the modulatory activity of these p160 coactivators is separable from their coactivator activity (Figs.  1-3). For both TIF2/GRIP1 and SRC-1, the modulatory activity can be restricted to a small region of each coactivator that does not include any of the activation domains (AD1 and AD2). Furthermore, the minimum sequence that retains high levels of modulatory activity has a significantly reduced capacity to augment the total amounts of GR-mediated transactivation (Fig. 2B). These results are consistent with our earlier conclusions that various factors can modulate the dose-response curve and partial agonist activity of GR complexes in a manner that is independent from their effects on the total levels of induced gene activity (20,21,(33)(34)(35)54).
The fact that coactivators can alter the partial agonist activity of GR-antagonist complexes is particularly interesting. We had first established a functional consequence of coactivators on the biological properties of GR-antagonist complexes in intact cells (20,21). The current general model proposes that coactivator binding to steroid receptors is seen only in the presence of agonist steroids (1,11). However, this model is based largely on the observations of nuclear receptors, which display significant mechanistic differences with GR. For example, ligand-free receptors are nuclear and DNA-bound for the nuclear receptors versus cytoplasmic for GR, binding of hsp90 is seen for GR and rarely for nuclear receptors (55), only the nuclear receptors heterodimerize with RXR, and the amino acids beyond helix 12 are required for ligand binding to GR (56) but not most nuclear receptors, which lack long amino acid sequences beyond helix 12. Estrogen receptors (ERs) are more like the nuclear receptors in several respects: nuclear localization of ligand-free receptors, diminished importance of hsp90 binding (57), and dispensability of COOH-terminal residues for steroid binding (58,59). Nevertheless, reports exist suggesting an interaction of coactivators with antagonist-bound ERs (60). SRC-1 binding to ER is seen with all forms of ER but is promoted by agonists and is the least with antagonists (60,61). Also, the binding of coactivators to antagonist-bound ER may be responsible for their partial agonist activity as microinjection of anticoactivator antibodies into transfected cells reduces the amount of partial agonist activity (62). We now establish that for GR, the coactivator TIF2/GRIP1 binds to GR complexed with several different antiglucocorticoids. These interactions are seen both in intact cells in a mammalian two-hybrid assay and in vitro in a pull-down assay (Fig. 7). In both assays, the interactions of all antagonists except RU486 were similar in magnitude with that seen with the agonist Dex. RU486bound GR displayed a more robust interaction with TIF2/ GRIP1 when the full-length GR is used, as opposed to a fragment of TIF2/GRIP1. The significance of this result is not yet known. However, the data clearly show that coactivators can bind to antisteroid complexes of GR, even those that normally display negligible amounts of partial agonist activity (33, 38,63). These results also provide biochemical support for our model of an equilibrium interaction of coactivators and corepressors binding to antagonist-bound GR (20) and progesterone receptors (23).
The p160 coactivators contain numerous functional domains. However, the only known functional domains that are involved in receptor modulatory activity are the RIDs, which are required for the binding of receptors. The effects of coactivators on the total levels of steroid receptor-mediated gene induction are mediated by the AD1 domain, which binds CBP/p300 (8,41,46,64). An auxiliary domain for the binding of GR has been identified at residues 1011-1056 of GRIP1 (41). A COOHterminal domain is needed for synergism between the activation sequences of receptors (65). However, none of these domains are present in TIF2.4, which is the minimally active fragment of TIF2.
The COOH-terminal active segment of SRC-1, corresponding to amino acids 1139 -1441 (Fig. 3), is devoid of activation domains but does contain a HAT activity domain and a RID (9,48). This segment also displays a dominant negative repressor activity (4,46,66), which appears to be due simply to its ability to bind tightly to the receptor LBD, thereby blocking the binding of other coactivators (48). Interestingly, we see little if any dominant negative activity for this SRC-1 fragment with GR (data not shown). Finally, this sequence of SRC-1 does bind PCAF (48). The binding of PCAF to the small active segment of SRC-1, coupled with the ability of PCAF to modulate the doseresponse curve, and partial agonist activity, of GR complexes (Fig. 4), initially made PCAF an attractive mediator of the modulatory activity of coactivators. However, the lack of any additive or synergistic response of PCAF with SRC-1/1139 - Several reports suggest that the coactivator activities of SRC-1 and TIF2 are largely redundant. SRC-1 knockout mice appear to compensate for the lack of SRC-1 by increasing the levels of TIF2 (67) and recent data argue that TIF2 and SRC-1 are not functionally distinct (68). Thus, while we have not established that the modulatory activities of SRC-1 and TIF2 are expressed by identical mechanisms, we currently suspect that they are. It should be noted that there are no reports of PCAF binding to TIF2/GRIP1 (16) and we are unable to detect any interaction of PCAF with the active TIF2 fragments (Fig.  5A). Therefore, as for SRC-1, the available evidence argue against an important role of PCAF in the actions of TIF2.
Several other factors that are known to bind to coactivators are shown not to be involved in the modulation of GR transactivation properties. CBP/p300 does not interact with TIF2.4 (Fig. 5B) whereas the p300 binding site in SRC-1 is reported to be at amino acids 913-977 (13), which is outside of the COOHterminal domain of SRC-1 modulatory activity. The DRIP/ TRAP complex has been found to play an important role in the expression of receptor transactivation activity (69 -71). DRIP150 and DRIP205 bind to receptors and increase the total amounts of receptor transactivation (51, 72) but are inactive in our assays, while DRIP150 does not synergize with TIF2. We cannot presently rule out, however, that some untested combination of factors might synergize with SRC-1 or TIF2. No common motif could be recognized in the short modulatory regions of TIF2 and SRC-1 that might bind some other transcription cofactor. Therefore, we conclude that a currently unidentified factor(s) interacts with the above identified unique domains of TIF2 and SRC-1 to contribute to the modulation of several transactivation properties of GR. This conclusion is supported by the ability of a fragment of TIF2 lacking the RIDs to be able to competitively inhibit the actions of TIF2.4 (Fig. 6), presumably by titrating out the unidentified factor(s).
In summary, we are assembling a model by which the p160 coactivators exert additional, unique effects on the transactivation properties of GRs complexed with both agonist and antagonist steroids. These properties, which include the ability of coactivators to associate with GR-antisteroid complexes, are physically separable from the initially discovered properties of augmenting the total levels of gene induction by receptor-agonist complexes. The domain of TIF2/GRIP1 that is sufficient for this new activity is not known to bind any protein and appears to interact with a currently unidentified factor(s). As the transcriptional properties of other steroid receptors are similarly modulated (23), we anticipate that this putative new factor(s) FIG. 7. Effect of added steroid on interaction of GR with GRIP1/TIF2. A, interaction of agonist-and antagonistbound GR with GRIP1/TIF2 in mammalian two-hybrid assays. Triplicate samples of CV-1 cells were transiently transfected with plasmids for the indicated chimeras, treated with 1 M of the designated steroids, and plotted as described in the legend to Fig. 5A. Similar results were obtained in at least one additional experiment. Inset, Western blot of overexpressed GAL/TIF2.4 and GAL/ TIF2.4m123. Both GAL chimeras were overexpressed in transiently transfected CV-1 cells and then detected by Western blotting as described under "Materials and Methods." B, binding of agonist-and antagonist-bound GR to TIF2 in pulldown assays. Bacterially expressed GST, GST/TIF2.4, and GST/TIF2.4m123 that had been immobilized by matrix-bound anti-GST antibody was treated with in vitro translated GR that had been bound by agonist (Dex), antiglucocorticoid (Dex-Mes, Dex-Ox, Prog, or RU486), or left unbound. GR complexes that were retained on the matrix by binding to the GST chimeras were separated on SDS-PAGE gels and visualized by autoradiography. Specific binding is seen as the GR binding to GST/TIF2.4 that is in excess of the nonspecific binding to GST alone. The binding of GR complexes to GST/TIF2.4 that is in excess of that to GST/TIF2.4m123 represents binding that requires the presence of the RIDs of TIF2. Similar results were obtained in a second experiment.
will similarly affect the dose-response curve of agonists (or EC 50 ), and partial agonist activity of antisteroids, bound to other steroid receptors.