Evidence for a Common Step in Three Different Processes for Modulating the Kinetic Properties of Glucocorticoid Receptor-induced Gene Transcription*

The dose-response curve of steroid hormones and the associated EC50 value are critical parameters both in the development of new pharmacologically active compounds and in the endocrine therapy of various disease states. We have recently described three different variables that can reposition the dose-response curve of agonist-bound glucocorticoid receptors (GRs): a 21-base pair sequence of the rat tyrosine aminotransferase gene called a glucocorticoid modulatory element (GME), GR concentration, and coactivator concentration. At the same time, each of these three components was found to influence the partial agonist activity of antiglucocorticoids. In an effort to determine whether these three processes proceed via independent pathways or a common intermediate, we have examined several mechanistic details. The effects of increasing concentrations of both GR and the coactivator TIF2 are found to be saturable. Furthermore, saturating levels of either GR or TIF2 inhibit the ability of each protein, and the GME, to affect further changes in the dose-response curve or partial agonist activity of antisteroids. This competitive inhibition suggests that all three modulators proceed through a common step involving a titratable factor. Support for this hypothesis comes from the observation that a fragment of the coactivator TIF2 retaining intrinsic transactivation activity is a dominant negative inhibitor of each component (GME, GR, and coactivator). This inhibition was not due to nonspecific effects on the general transcription machinery as the VP16 transactivation domain was inactive. The viral protein E1A also prevented the action of each of the three components in a manner that was independent of E1A's ability to block the histone acetyltransferase activity of CBP. Collectively, these results suggest that three different inputs (GME, GR, and coactivator) for perturbing the dose-response curve, and partial agonist activity, of GR-steroid complexes act by converging at a single step that involves a limiting factor prior to transcription initiation.

The dose-response curve of steroid hormones and the associated EC 50 value are critical parameters both in the development of new pharmacologically active compounds and in the endocrine therapy of various disease states. We have recently described three different variables that can reposition the dose-response curve of agonist-bound glucocorticoid receptors (GRs): a 21-base pair sequence of the rat tyrosine aminotransferase gene called a glucocorticoid modulatory element (GME), GR concentration, and coactivator concentration. At the same time, each of these three components was found to influence the partial agonist activity of antiglucocorticoids. In an effort to determine whether these three processes proceed via independent pathways or a common intermediate, we have examined several mechanistic details. The effects of increasing concentrations of both GR and the coactivator TIF2 are found to be saturable. Furthermore, saturating levels of either GR or TIF2 inhibit the ability of each protein, and the GME, to affect further changes in the dose-response curve or partial agonist activity of antisteroids. This competitive inhibition suggests that all three modulators proceed through a common step involving a titratable factor. Support for this hypothesis comes from the observation that a fragment of the coactivator TIF2 retaining intrinsic transactivation activity is a dominant negative inhibitor of each component (GME, GR, and coactivator). This inhibition was not due to nonspecific effects on the general transcription machinery as the VP16 transactivation domain was inactive. The viral protein E1A also prevented the action of each of the three components in a manner that was independent of E1A's ability to block the histone acetyltransferase activity of CBP. Collectively, these results suggest that three different inputs (GME, GR, and coactivator) for perturbing the doseresponse curve, and partial agonist activity, of GR-steroid complexes act by converging at a single step that involves a limiting factor prior to transcription initiation.
Steroid receptors are best known for their ability to modify the rates of transcription of target genes. In the most fre-quently studied situations, these rates are increased following the binding of agonist steroids to their cognate receptors. Antisteroids block the action of agonist steroids but generally retain some residual agonist activity. For both classes of receptor-steroid complexes, the binding of steroid to its cognate receptor is thought to be the rate-limiting step in steroid hormone action (1)(2)(3). Ligand binding to receptors is followed by an irreversible step called activation, binding to the hormone response elements of the regulated genes, recruitment of coactivators or corepressors, and interaction with the transcriptional machinery. The recent discovery of histone acetyltransferase (HAT) 1 activity in most coactivators (4 -7), and of histone deacetylase activity with corepressors (8 -10), has permitted models of histone acetylation-induced modifications of nucleosome structure (11)(12)(13) to be incorporated into those for steroid regulated gene expression (14 -16).
Both the responses to endogenous steroids and the pharmacology of steroid hormone-based endocrine therapies depend critically on the dose-response curve, which relates steroid concentration to biological activity. Depending on where the curve is positioned, more or less steroid will be required to achieve half of the maximal desired response, or the EC 50 . A lower EC 50 , which is synonymous with a left-shifted doseresponse curve, is highly desirable in clinical situations as it leads to fewer adverse side effects. This is because the responses of other regulated genes with higher EC 50 values would be greatly reduced at the lower steroid concentrations. Similarly, antisteroids that are selective receptor modulators, which are antagonists for designated genes but agonists for all other responsive genes, would be most useful in antisteroid therapies. Over the last several years, numerous examples have been reported of genes with different EC 50 values for steroidal agonists and partial agonist activity for antisteroids (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32)(33)(34)(35). The current models for steroid hormone action do not help in understanding these phenomena as they predict that the same receptor-agonist complex will evoke identical doseresponse curves (and EC 50 values) for all responsive genes. Likewise, a specific antisteroid should always produce the same amount of partial agonist activity. These invariant predictions suggest that additional components or processes may be needed to account for the above diverse transcriptional properties of individual steroid receptor-steroid complexes.
Three components/processes have recently been identified that can modulate both the dose-response curves of agonists and the partial agonist activity of antagonists bound to glu-cocorticoid receptors (GRs). The first to be described was the glucocorticoid modulatory element (GME), which was identified as being 1 kilobase pair upstream of the glucocorticoid response elements (GREs) of the rat tyrosine aminotransferase gene (28, 36 -38). The action of the GME appears to involve the binding of two recently cloned novel proteins called GMEB-1 and -2 (28, 35, 39 -41). The second component was found to be changing concentrations of GR (22,23). This effect was observed both in HeLa cells, where receptor is not a limiting factor, and in CV-1 cells, where GR is limiting. The third component was reported to be varying concentrations of coactivators and corepressors (23). Thus, elevated levels of the coactivators TIF2/GRIP1 (42,43), SRC-1 (44), AIB1 (45), and CBP (46) all caused a shift in the dose-response curve of the glucocorticoid dexamethasone to lower steroid concentrations. At the same time, these cofactors each increased the partial agonist activity of antiglucocorticoids, such as dexamethasone 21-mesylate, without increasing the concentration of GR (23). Conversely, added corepressor SMRT antagonized the actions of added coactivator.
The transcriptional responses to the three components (GME, GR, or coactivator) in our previous studies were qualitatively identical (22, 23, 28, 36 -38). In every case, the induction properties of all agonists and antagonists examined were similarly modified. We also never observed any dissociation of the effects on the dose-response curve from those on the partial agonist activity of antagonists. Furthermore, the changes in these induction parameters were independent of changes in total transactivation or -fold induction (22,23,36,47). This coincidence of phenotypical properties suggested that each process might modify the GR induction properties via a common step that could involve a titratable factor.
Among the various components that we have examined, TIF2 and the GME binding proteins (GMEB-1 and -2) were particularly interesting because they were all devoid of HAT activity (35,48). Therefore, to the extent that HAT activity is involved in the actions of these components, it would have to be present in some other protein in the transactivation complex. A ubiquitous recruited component of steroid hormone action is CBP, which does possess HAT activity (4). CBP is known to bind TIF2 (48) and interacts with both GMEBs in mammalian two-hybrid assays (35). For these reasons, a specific inhibitor of CBP HAT activity would be a useful mechanistic probe. The adenovirus protein E1A, which occurs as both 12 S and 13 S isoforms, appears to be ideal for several reasons. E1A interacts with numerous molecules, including CBP and p300, with diverse consequences (49). E1A-13S can both increase and repress the activation of various genes. However, repression correlates with the ability of E1A to bind CBP/p300 (49). E1A-13S has also been reported to inhibit estrogen receptor activation (50), possibly by inhibiting the HAT activity of CBP (51). Hormone-induced transactivation and histone acetylation by estrogen receptor are reported to be highly dependent on the HAT function of CBP/p300 (52), whereas E1A-12S (full length ϭ 1-242) inhibits acetylation of histones by CBP (53). In contrast, Li et al. (54) described that an N-terminal deletion of E1A has no direct effect on the induction of transcriptional activation by retinoid X receptor/TR complexes but rather stimulates transcription at a subsequent step by acetylating proteins associated with the Xenopus TR␤A promoter. Thus, E1A should be an interesting probe of the mechanism by which TIF2, GME, and increased GR modulate GR induction properties.
The purpose of this study was to determine whether the effects on GR transactivation properties of the GME and increased concentrations of GR or coactivators, such as TIF2, proceed through distinct or common pathways. The ability of these components to modulate GR transcriptional properties was examined by the kinetic method of determining if one or more of the various components could competitively inhibit the actions of each other. Additional evidence for action by a common intermediate was sought by investigating whether inhibitors of one process would be equally effective in blocking the other processes. One of the inhibitors used, for the reasons outlined above, was the 13 S form of E1A, which blocks the HAT activity of CBP (51). Our studies suggest that multiple pathways can converge to modify GR induction properties in a manner that does not require the HAT activity of CBP.
Cell Culture and Transient Transfection-CV-1 cells were grown in Dulbecco's modified Eagle's medium (DMEM; Life Technologies, Inc.) supplemented with 10% fetal calf serum (FCS; Biofluids Inc., Rockville, MD) at 37°C in a humidified incubator (5% CO 2 ). The transient transfections were performed using calcium phosphate in presence of DMEM plus 10% FCS followed by a 16-h incubation in DMEM plus 10% FCS before being induced with the appropriate steroid for 24 h (22). Transfected cells were lysed and assayed for reporter gene activity using luciferase assay reagent according to manufacturer's instruction (Promega, Madison, WI.). Luciferase activity was measured in an EG&G Berthhold luminometer (Microlumat LB 96 P).
Data Analysis-All statistical analyses were, unless specified otherwise, by the two-tailed Student's t test using the program InStat 2.03 for Macintosh (GraphPad Software, San Diego, CA).

Effects of Increased GR Are Saturable-
The presence of increasing concentrations of GR in transiently transfected cells causes an increase in the -fold induction at each concentration of steroid (Fig. 1A). The effect of added GR on the dose-response curve is most easily appreciated when all of the data are replotted as percentage of maximal induction by a saturating concentration of dexamethasone (1 M). When this is done, it is clear that there is a marked shift in the dose-response curve (and EC 50 ) to lower steroid concentrations (Fig. 1B). We have previously reported similar effects of added GR (22), GME (36), and TIF2 (23) without changing the shape of the curve. To facilitate similar types of measurements under numerous conditions, we commonly use an abbreviated assay in which the activity of a constant, subsaturating concentration of the agonist dexamethasone is expressed as percent of the maximal activity displayed by saturating concentrations of dexamethasone (1 M). In this assay, an increase in the percentage of maximal agonist activity seen with a subsaturating concentration of dexamethasone (such as 0.4 nM, indicated by dashed vertical line in Fig. 1B) is diagnostic of a shift in the doseresponse curve to the left (22,23,35,36,47). At the same time, this abbreviated assay includes a single point with a saturating concentration of antisteroid so that we can simultaneously determine the effect of the specific reaction conditions on the partial agonist activity of an antiglucocorticoid.
Typical transient transfections with CV-1 cells utilize 0.04 -0.4 g of GR plasmid ( Fig. 1) (22,23,56). Concentrations of the plasmid pSVLGR as high as 4 g can be used without signs of squelching, as determined by both the total amount of transactivation and the -fold induction by 1 M dexamethasone (data not shown). However, the ability of increasing GR plasmid to cause a progressively greater percentage of maximal activity with subsaturating concentrations of dexamethasone (synonymous with further left shifts in the dexamethasone dose-response curve), and higher amounts of partial agonist activity of progesterone, reaches a plateau by 2 g of pSVLGR ( Fig. 2A). This suggests that the effects of added GR have a finite limit, consistent with the titration of an intermediate factor or saturation of a step in the transcriptional activation process.
Actions of GME and High GR Concentrations Display Competitive Inhibition-The classical method for establishing that two processes share a common intermediate is to show that they exhibit competitive inhibition, wherein saturating concentrations of one component prevent the action of a second component (58). We therefore asked whether the GME can augment the action of both low and plateau concentrations of GR. If the presence of the GME is able to cause increased activity at low but not high concentrations of GR, one can conclude that the two processes compete at a common reaction step that is already maximally stimulated at high levels of GR. The sensitivity of such assays is greatly diminished near saturation of the dose-response curve, where increasing amounts of steroid have progressively smaller effects (see Fig. 1). Therefore, we concentrated on the changes in the percent of maximal induction at low levels of dexamethasone so that we would still be in the sensitive part of the dose-response curve (Յ50% of maximal induction) with plateau concentrations of GR. As expected from previous results, the presence of the GME increases the various relative activities seen with low concentrations of GR ( Fig. 2A; p Յ 0.0022 for treatments with 0.1 ng of GR Ϯ GME) (28, 36 -38). However, the GME is completely inactive at high concentrations of GR. We, therefore, conclude that the actions of the GME and GR competitively inhibit each other and proceed through a shared step.
The absence of any increase in the percent of maximal activity at high concentrations of GR Ϯ GME for low concentrations of dexamethasone does not appear to be due to a saturation of the transcriptional machinery because the total amount of transactivation was elevated by added GR (see legend for Fig.  2A). Support for this conclusion came from studies with the C656G mutant rat GR. This mutation affords a "super" receptor with an approximately 10-fold increased affinity for dexamethasone and a greater specificity among various steroids (56,57). Although the relative activities of subsaturating concentrations of dexamethasone for induction of GREtkLuc reach a plateau value with 4 g of transfected wild type GR ( Fig. 2A), both the relative and total activities can be increased even more in the presence of 4 g of C655G super receptor (Fig. 2B). At the same time that the C656G receptor causes a further left shift in the dexamethasone dose-response curve, the partial agonist activity with dexamethasone 21-mesylate increased by 1.35 Ϯ 0.25-fold (range, n ϭ 2). These results thus suggest that the plateau values in percentage of maximal activity that are seen with the GME and higher concentrations of GR are not due to an inherent limitation of the transcriptional machinery that somehow restricts the maximal agonist activities of the assay.
Effects of Coactivators Are Competitive for Those of the GR-We recently documented that added coactivators, such as TIF2/GRIP1 and SRC-1, can also reposition the GR dose-response curve to the left and increase the partial agonist activity of antiglucocorticoids. The behavior of TIF2 mimics that of added GR but is manifested in a manner that does not involve coactivator-induced increases in GR concentration (23). In the few studies that describe the effects of higher levels of coactivator, it appears that the total transactivational activity increases in parallel (59,60). We therefore asked if added coac- tivator would further modify the EC 50 and partial agonist activity of antisteroids in the presence of the high levels of GR that afford plateau values in the EC 50 and partial agonist activity. With low GR (40 ng of plasmid), 0.2 g of TIF2 induces the usual and highly significant increases in activity of both subsaturating concentrations of dexamethasone (p Ͻ 0.0001, n ϭ 3) (Fig. 3A) and saturating concentrations of dexamethasone 21-mesylate (p ϭ 0.0015, n ϭ 3) (Fig. 3B). However, with plateau level concentrations of GR, there was no additional increase with added TIF2 (p ϭ 0.35 to 0.93). It should be noted that 0.2 g of TIF2 causes a similar, 2.2-3.0-fold increase in the total transactivation by 1 M dexamethasone at all levels of GR (data not shown). Therefore, the inability of added TIF2 to augment the EC 50 and partial agonist activity values seen with high concentrations of GR is not due to an inactivation of TIF2. Rather the ability of TIF2 to enhance the levels of total transactivation is dissociated from the effects of TIF2 on the levels of EC 50 and partial agonist activity. Furthermore, we interpret this ability of TIF2 to modify the EC 50 and partial agonist activity at low, but not high, GR concentrations as GR and TIF2 competitively inhibiting the actions of each other. This, in turn, indicates that the actions of added GR and TIF2 involve a common mechanistic intermediate.
Similarly, the effects the coactivator SRC-1 and high concentrations of GR were found not to be additive. With low amounts of GR (100 ng), additional transfected SRC-1 (1.8 g) results in higher activities, as a percentage of maximal activity, of subsaturating concentrations of dexamethasone and saturating concentrations of dexamethasone 21-mesylate (Fig. 3C) (23). However, the cotransfected SRC-1 is without any effect at plateau concentrations of GR (2 g; see Fig. 2). Again, SRC-1 was still active as a coactivator under these conditions because the total amount of transactivation with 1 M dexamethasone and 2 g of GR was increased by 40 -600% by the additional SRC-1 (data not shown). Therefore, we conclude that GR and SRC-1 competitively inhibit the actions of each other at the level of EC 50 and partial agonist activity but not at the level of total transactivation.
Effects of TIF2 Are Saturable and Competitive for Those of the GME-The cotransfection of small amounts of TIF2 markedly increases the percent of maximal activities of both subsaturating concentrations of dexamethasone, due to a left shift in the dose-response curve (23), and saturating concentrations of the antisteroid dexamethasone 21-mesylate (Fig. 4). At low concentrations of transfected TIF2 plasmid (15 ng), the GME further increases both activities (p Ͻ 0.0001 for 1 nM dexamethasone, p ϭ 0.013 for 1 M dexamethasone 21-mesylate; n ϭ 4). However, the presence of the GME is unable to cause any significant increase in conjunction with higher amounts of TIF2 (p ϭ 0.85 to 0.27, n ϭ 4). This competitive inhibition of GME activity by TIF2 argues that these two components exert their effects via a common intermediate. Therefore, not only do TIF2 and GME compete with each other (Fig. 4), but competitive inhibition is also seen between TIF2 and GR (Fig. 3, A and  B) as well as between GR and GME ( Fig. 2A). These results indicate that the actions of all three components proceed through a shared step. Once that step is fully induced, further input from other directions is unable to cause any additional increase.
Specific Inhibition of GR, TIF2, and GME Activities by the Dominant Negative Suppressor TIF2.7-To test our conclusion that GR, TIF2, and GME all influence GR transactivational properties by pathways that converge at a common pretranscriptional step, we asked whether inhibitors of one component might similarly prevent the action of the other two components.
Overexpressed transcription factors often block the activity of the same or other transcription factors, which is commonly thought to result from squelching, or the non-productive titration of limiting factors (22,(61)(62)(63). For example, TIF2.7 is a fragment of TIF2 corresponding to amino acids 870 -1179, which contains the more central activation domain of TIF2 (called AD1) and possesses intrinsic transactivation activity (48). In our assays, the cotransfection of 200 ng of TIF2.7 plasmid routinely decreases the total transactivation, and -fold induction by 1 M dexamethasone, by 25-75% (n ϭ 15). At the same time, TIF2.7 suppresses the percent of maximal activity of 1 nM dexamethasone and 1 M dexamethasone 21-mesylate (Fig. 5A). This influence of TIF2.7 on GR induction properties is reversible because the effect of TIF2.7 is counterbalanced by higher amounts of GR (compare 200 versus 40 ng in Fig. 5A). Likewise, added TIF2 antagonizes the inhibitory effects of TIF2.7 with 40 ng of GR (Fig. 5A). Interestingly, there is very little effect of TIF2.7 on the -fold induction (or total transactivation) by 1 M dexamethasone in the presence of TIF2 (data not shown). This appears to be yet another example of how separate parameters can modify GR -fold induction independently of changes in the dose-response curve (22,23,36,47). TIF2.7 is even more efficient in blocking the effects of the GME (Fig. 5B). Whereas TIF2.7 is unable to completely reverse the increased activities caused by higher levels of GR or TIF2 (Fig. 5A), added TIF2.7 prevents any of the increased activities that result from the presence of the GME element in the GREt-kLuc reporter (Fig. 5B). In summary, the observation that TIF2.7 inhibits the activities of GR, TIF2, and GME is consistent with our proposal that a single step is shared by GR, TIF2, and GME and that this step can be inhibited by TIF2.7.
To determine whether the repressive activity of TIF2.7 is specific, as opposed to a nonspecific effect on transcription in general, we asked whether similar decreased GR induction properties would be detected with overexpression of another potent but unrelated transactivator, the VP16 activation domain. In these experiments, a chimera of the GAL4 DNA binding domain with the VP16 transactivation domain was used so that the presence of functionally active GAL-VP16 fusion protein could be independently assessed by its ability to induce a luciferase reporter regulated by the 5Ј-upstream GAL4-responsive elements (UAS). As shown in Fig. 6, 20 ng of GAL/VP16 is sufficient for nearly maximal transactivation of the UAS 5 LUC reporter but has no effect on the activity of subsaturating concentrations of dexamethasone or saturating concentrations of dexamethasone 21-mesylate. Therefore, the inhibitory effects of TIF2.7 in Fig. 5 are relatively specific and are not simply the result of nonspecific interactions with any overexpressed transactivation domain.
Actions of GR, TIF2, and GME Do Not Require HAT Activity of CBP-Considerable experimental support exists for the model that one role of coactivators in the mechanism of steroid hormone action is to increase the acetylation of histones and possibly other proteins (14 -16). Although most coactivators possess intrinsic transactivation activity (5)(6)(7)9), TIF2 does not (48). However, all of the coactivators interact with CBP (5, 48, 50, 64 -67), which does have HAT activity (4). The adenovirus 13 S protein, E1A, has been described to be a nontoxic inhibitor of CBP HAT activity. This activity has been localized to the CR2 domain of E1A-13S (Fig. 7A) (51). To test whether the HAT activity of CBP is required for the modulation of GR induction properties by GME or increased concentrations of either GR or TIF2, we investigated the ability of E1A-13S to prevent these effects in a manner that depends upon the CR2 domain of E1A.
E1A-13S unexpectedly causes both a right shift in the dexamethasone dose-response curve and decreased amounts of agonist activity for the antiglucocorticoid dexamethasone 21-mesylate (Fig. 7B). The magnitude of these effects depends on the concentration of E1A-13S, as seen by the progressively lower values with 0.4 g of GR (Fig. 7C). The inhibitory activity of E1A-13S was also reversible, as demonstrated by the ability of added GR (0.4 versus 0.04 g) to attenuate the effects of 0.2 g of E1A-13S (Fig. 7C). Most interestingly, the repressive activity of E1A-13S was maintained with the truncated protein, E1A/ N108 (Fig. 7A), which lacks the CR2 domain that is required for the inhibition of CBP HAT activity (51). Therefore, CBP HAT activity does not seem to be required for expression of the increases in the percentage of maximal activity that are seen with higher GR concentrations. Similar experiments revealed that E1A-13S also reverses both the effects of TIF2 (Fig. 7D) and GME (Fig. 7E). In each case, the truncated E1A/N108 is fully active. Thus, E1A-13S is equally effective in inhibiting the effects of added GR, TIF2, or GME on GR induction parameters. Furthermore, in each case, E1A-13S appears to be acting via a mechanism that is independent of its capacity to block CBP HAT activity. DISCUSSION Two parameters of GR induction of responsive genes that are important for the differential control of gene expression are the positioning of the steroid dose-response curve, or EC 50 , and the amount of partial agonist activity displayed by antiglucocorticoids. Although previous models of steroid hormone action predict that these parameters will be constant for a given receptorsteroid complex within a given cell, the variety of reports to the contrary (17-35) indicate a need for modifications of these models. We have previously described three different components that can modulate these two parameters: the cis-acting element called the GME (28, 36 -38, 47), varying concentrations of GR (22,23,56), and varying concentrations of coactivators (23). We now propose that all three components share a common mechanistic step.
Four independent experimental approaches support our current hypothesis. First, the increase in activity, as a percentage of maximal induction, both of subsaturating concentrations of the agonist dexamethasone (indicative of a left shift in the dose-response curve to lower EC 50 values), and of saturating concentrations of the antagonist dexamethasone 21-mesylate (due to increased partial agonist activity) is not unlimited but reaches plateau values with high concentrations of transfected GR or TIF2 (Figs. 2A, 3, and 4). This indicates that some process is being saturated, possibly by titrating out a limiting factor. Second, under these plateau conditions of excess GR or TIF2, none of the components, including the GME, are able to effect any further changes in EC 50 or partial agonist activity (Figs. 2-4). This mutually competitive inhibition by each of the three components is diagnostic kinetic evidence for a shared mechanistic step. The plateau is not an artifact of some intrinsic limitation of the transcriptional machinery, as further increases can be achieved with the C656G mutant GR (Fig. 2B). The plateau is also not due to an inactivation of GR or TIF2, as both proteins still increase the levels of total transactivation. Third, overexpression of a TIF2 fragment (TIF2.7) containing intrinsic transactivation activity prevents the higher activities (as percentage of maximal induction by dexamethasone) of subsaturating concentrations of dexamethasone, and saturating concentrations of dexamethasone 21-mesylate, that are seen not only with the parent molecule, TIF2, but also with each of the other components, i.e. GR and GME (Fig. 5). This inhibition by TIF2.7 is reversible with added GR or TIF2 (Fig.  5). TIF2.7 inhibition is reasonably specific because a similar repression is not observed in the presence of another overexpressed transcription domain, i.e. the VP16 activation domain (Fig. 6). Finally, added E1A-13S blocks the actions of GR, TIF2, and GME in a reversible and concentration-dependent manner (Fig. 7). Both full-length E1A-13S and an N-terminal fragment were equally active with each of the components. Collectively, these data strongly support the conclusion that the three different components that have been described to modulate the EC 50 of GR agonists and the partial agonist activity of GR antagonists do so through a common step (Fig. 8). It should be realized that the magnitude of variation in the EC 50 and the partial agonist activity values will probably vary between tissues, and between normal and malignant cells, in view of the differences in coactivator (45,68,69) and receptor levels.
Our kinetic experiments cannot determine whether the various inhibitors (TIF2.7, E1A-13S, and E1A/N108) prevent a reaction before or after the step containing our proposed molecule X (Fig. 8). However, we suspect that the site of repression is distal to X simply because it is easier to envisage each inhibitor blocking one step after X than three different steps prior to X. Whether TIF2.7 and the E1A molecules would prevent a common step after X is unknown.
All of the components examined here (GR, TIF2, SRC-1, and GME) cause a left shift in the dexamethasone dose-response curve and increased agonist activity of antiglucocorticoids. Similarly, the coactivator AIB1 (45) and the comodulator CBP (46) have the same effects on GR induction properties (23). Although we have not further examined the properties of these species, we predict that they too will be found to competitively inhibit the actions of GR, TIF2, and GME, and thus proceed through the common intermediate "X" that is proposed in Fig.  8. It should be noted that this intermediate is not a target of all proteins with intrinsic transactivation activity as added, Nterminal truncated GRs (22), progesterone receptors (22), and a GAL/VP16 chimera (Fig. 6) are each inactive.
The nature of the key molecule X of Fig. 8 is not yet known. Since the ability to modulate the kinetic parameters of GR transactivation has the physiologically beneficial consequence of allowing differential control of gene transcription, the greatest regulatory diversity would be achieved if the various mechanisms converged as some common point close to the final step. As some of the key intermediary factors for the expression of GME activity are different from those that are utilized by GR (47), attractive candidates for X are components of the macromolecular complexes that have recently been implicated in the transcriptional process, such as TRAP (70), DRIP (71), ARC (72), and CRSP (73), but may not be directly involved in steroid-induced transactivation. Given the importance of differential control of gene expression in development, differentiation, and homeostasis, it would also not be surprising if yet additional pathways exist that are capable of modifying the GR induction parameters. In fact, we have recently discovered another process that does not act through the intermediate X identified in this study. 2 This, plus the observation that the activation domain of the general transactivator VP16 does not influence either the EC 50 or the partial agonist activity of GR complexes, suggests that factor X is involved in a step that is common to several but not all transactivation pathways.
One intriguing, recent mechanistic proposal for the modulation of GR induction properties is that added coactivators may bind to the receptors and change the affinity of steroid for receptor by decreasing the rate of steroid dissociation (74). We believe that this attractive hypothesis is not operative in our proposed model (Fig. 8) for several reasons. First, GR does not bind to DNA until after steroid binds to the receptor and the resulting complex is activated and translocated into the nucleus. Therefore, any effect of GME-containing DNA sequences on the dose-response curve by modifying the steroid binding affinity of GR would have to be manifested several steps after the initial equilibrium binding of steroid to receptor. Second, simply increasing the amount of GR will increase the total amount of steroid that can specifically bind. However, thermodynamic considerations argue that increased GR concentrations do not alter the affinity of steroid for GR. Third, as the affinity of antisteroid binding to GR is independent of the amount of partial agonist activity (75), it is not clear how an increase in affinity could result in higher amounts of partial agonist activity (23). Fourth, TIF2.7 is capable of reversing the effects of added coactivator TIF2, presumably by titrating out some limiting factor. Because TIF2.7 does not bind GR (76), it is likely that TIF2 is titrating out an additional, non-GR molecule to modify the GR dose-response curve. Finally, as the GME, increased GR, and increased coactivator each compete at an apparently common step to cause the left shift in the agonist dose-response curve (Fig. 8), we conclude that this behavior does not result from a possible increase in steroid binding affinity to GR with additional coactivator. It should be noted that changes in steroid dissociation rates can be accompanied by compensating differences in the rate of ligand binding so that the steroid binding affinity does not change (77,78).
The actions of the viral protein E1A-13S in our system are particularly interesting. E1A interacts with three regions of CBP (79): amino acids 1-1450, which bind the LBDs of steroid receptors (65,80), the HAT activity domain encompassing residues 1805-1891, and the coactivator binding domain at positions 2058 -2163. Both E1A-13S and E1A/N108 block the activities of GR, GME, and TIF2 (Fig. 6, C-E). However, these properties appear to be independent of effects on CBP HAT activity because amino acids 121-138 of E1A (domain CR2) are reported to be required to block histone acetylation by CBP, whereas the N108 fragment of E1A-13S is almost inactive (51). The inhibitory properties of E1A-13S and E1A/N108 are unlikely to stem from alterations of the N-and C-terminal domains of CBP, which are involved in CBP binding to receptor LBDs or coactivators respectively, because an N-terminal mutation of E1A affects only the binding to the HAT domain in the middle of CBP (79). Therefore, we suspect that the region of E1A-13S that is active in blocking the effect of GR and TIF2 (amino acids 1 ϭ 108) is different from that which is required for histone acetylation. This, in turn, suggests that the effects of GR and TIF2 on the dose-response curve of agonists and the 2 S. Chen and S. S. Simons, Jr., manuscript in preparation.
indicated concentrations of GR without or with either E1A-13S or E1A/N108. After treating the samples with 1 nM dexamethasone or 1 M dexamethasone 21-mesylate, the induced luciferase activities above basal levels (EtOH) were determined and expressed as percentage of the maximal activity with 1 M dexamethasone, as described under "Materials and Methods." The error bars indicate the S.D. of triplicate values. Similar results were obtained in two additional experiments. D, relative activities of E1A-13S and E1A/N108 with TIF2. Triplicate samples of CV-1 cells were transiently transfected with 1.5 g of GREtkLuc and the indicated concentrations of GR Ϯ TIF2 without or with either E1A-13S or E1A/N108. After treating the samples with 1 nM dexamethasone or 1 M dexamethasone 21-mesylate, the induced luciferase activities above basal levels (EtOH) were determined and expressed as percentage of the maximal activity with 1 M dexamethasone, as described under "Materials and Methods." The error bars indicate the S.D. of triplicate values. Similar results were obtained in two additional experiments. E, relative activities of E1A-13S and E1A/N108 with GME. Triplicate samples of CV-1 cells were transiently transfected with 1.5 g of GREtkLuc or GMEGREtkLuc, and the indicated concentrations of either E1A-13S or E1A/N108. After treating the samples with 1 nM dexamethasone or 1 M dexamethasone 21-mesylate, the induced luciferase activities above basal levels (EtOH) were determined and expressed as percentage of the maximal activity with 1 M dexamethasone, as described under "Materials and Methods." The error bars indicate the S.D. of triplicate values. Similar results were obtained in a second experiment. In all cases, the molar amount of E1A vector was constant.
FIG. 8. Proposed model for convergent pathways by which GME, added GR, and increasing concentrations of coactivators modify the dose-response curve of GR-agonist complexes and the partial agonist activity of GR-antagonist complexes. The mutual competitive inhibition of each of the three components, combined with the common inhibition by TIF2.7, E1A, and E1A/N108, suggest that a common intermediate (X) is active in the expression of the final biological responses. The precise step that is inhibited by TIF2.7, E1A, and E1A/N108 is shown to be the same only for ease of display and may be different for TIF2.7 and the E1A proteins. See text for further details. partial agonist activity of antagonists are exerted by mechanisms other than modifying histone acetylation. This conclusion is consistent with our findings that the total levels of transactivation, which are directly affected by cofactors that are thought to act through histone acetylation (4 -7, 9, 10), can be dissociated from the modulation of other GR transcriptional properties, such as the EC 50 and the partial agonist activity of antisteroids (22,23,36,47). Similarly, others have reported that the stimulatory effects of SRC-1 and CBP on the cell-free transcriptional activation by progesterone receptors from DNA templates are independent of histone acetylation (81,82). Experiments with transfected templates in Xenopus oocytes argue that E1A has no direct effect on the induction of transcriptional activation by retinoid X receptor/TR complexes but rather stimulates transcription at a subsequent step by acetylating proteins associated with the TR␤A promoter (54). Therefore, our proposed intermediate target, X, may be component of the promoter complex, which would be consistent with our recent evidence that the GMEBs affect GR transactivation properties by contacting some proteins that are different from those directly interacting with GR (47).
In view of the extensive homology between glucocorticoid receptors and other steroid receptors, a natural question is whether the above processes similarly affect any other steroid receptor. We have recently found that the induction parameters for progesterone receptors are also affected by increasing progesterone receptor concentration and to a lesser extent by some coactivators and corepressors. 3 However, there are significant differences, as witnessed by the fact that added GR, but not progesterone receptor, modified the properties of GR induction (22). Thus, it is possible that the induction properties of other steroid receptors may be modified by pathways similar to those proposed in our model (Fig. 8) but utilizing different specific factors.