The Seven Amino Acids (547–553) of Rat Glucocorticoid Receptor Required for Steroid and Hsp90 Binding Contain a Functionally Independent LXXLL Motif That Is Critical for Steroid Binding*

Hsp90 association with glucocorticoid receptors (GRs) is required for steroid binding. We recently reported that seven amino acids (547–553) overlapping the amino-terminal end of the rat GR ligand-binding domain are necessary for hsp90 binding, and consequently steroid binding. The role of a LXXLL motif at the COOH terminus of this sequence has now been analyzed by determining the properties of Leu to Ser mutations in full-length GR and glutathioneS-transferase chimeras. Surprisingly, these mutations decreased steroid binding capacity without altering receptor levels, steroid binding affinity, or hsp90 binding. Single mutations in the context of the full-length receptor did not affect the transcriptional activity but the double mutant (L550S/L553S) was virtually inactive. This biological inactivity was found to be due to an increased rate of steroid dissociation from the activated mutant complex. These results, coupled with those from trypsin digestion studies, suggest a model in which the GR ligand-binding domain is viewed as having a “hinged pocket,” with the hinge being in the region of the trypsin digestion site at Arg651. The pocket would normally be kept shut via the intramolecular interactions of the LXXLL motif at amino acids 550–554 acting as a hydrophobic clasp.

Steroid hormone receptors are members of a superfamily of proteins with a common zinc finger motif for DNA binding. While the initially discovered members of this superfamily bind steroid and thyroid hormones, the majority of the family members still have no defined ligand (1,2). Those receptors that bind hormone are ligand activated transcription factors and are thus of interest both as models for the control of gene transcription and as mediators of hormone action.
Steroid binding to a receptor requires both a properly folded protein and an intact ligand-binding domain (LBD). 1 In general, the carboxyl-terminal ϳ250 amino acids of receptors encodes the LBD, although the precise boundaries have rarely been defined. For rat glucocorticoid receptors (GR), the last 246 amino acids (positions 550 -795) encompass the LBD (3). Some receptors, such as the estrogen receptor, contain additional residues after the LBD, commonly called the F domain. However, these added amino acids do not appear to be required for ligand binding (4). It was initially thought that any mutation in the LBD was detrimental for steroid binding. Subsequent research, although, has revealed that many mutations are without effect (5) and some increase the affinity and/or selectivity of receptors (6,7).
Defining the determinants of protein folding that are required for steroid binding has proven to be a difficult task. As with many protein domains, LBDs are transferable and retain their activity when attached to other proteins (8,9). This led to the view that receptor LBDs were independent entities containing all of the information necessary for ligand binding. It was, therefore, surprising to find that the GR LBD was not stable as an isolated unit but required other, upstream sequences, which did not have to be those of GR, in order to obtain a stable protein (3).
Another crucial component for hormone binding to GRs is hsp90. In fact, the GR cannot bind hormone unless the receptor is already complexed with hsp90 (10,11). Hsp90 is an ubiquitous cellular protein that is part of a multiprotein complex with chaperone activity (12). When hsp90 dissociates from GRs, steroid binding is lost but can be regenerated by incubating immobilized GRs with reticulocyte lysate or with a five-protein, minimal chaperone system consisting of hsp90, hsp70, Hop, hsp40, and p23 (11)(12)(13). Several sequences of the GR have been implicated in the binding of hsp90, but most are near the middle of the GR LBD (amino acids 568 -671 of the rat GR (14 -16)). We recently reported (3) that a more amino-terminal sequence of 547-553, which lies at the border of the GR LBD, was essential for the binding of hsp90 (17). This seven-amino sequence appeared to be specifically required, as a polyalanine spacer was incapable of maintaining hsp90 binding, and consequently steroid binding activity.
No common sequence required for hsp90 binding has yet been discerned among the numerous transcription factors and proteins kinases with which hsp90 forms complexes (reviewed in Refs. 12 and 18). Indeed, three different regions of both the human GR (15) and the chicken progesterone receptor (PR) (19) were sufficient to confer hsp90 binding. It is generally thought that, like other chaperones, hsp90 initially binds to exposed hydrophobic residues of partially denatured proteins and that this binding assists the protein in folding to its final, native conformation (20 -23). However, for those proteins that undergo ATP-dependent assembly into stable heterocomplexes with hsp90, such as steroid receptors, this simple model is not sufficient. In addition to creating steroid-binding sites, hsp90 complexation with GRs also opens up the LBD to permit both the attack of thiol residues by a thiol-derivatizing agent (24) and the proteolytic cleavage at basic amino acids by trypsin (25,26). These data, coupled with those of GR mutants (17), support the notion (27) that the hsp90-based chaperone machinery directs the partial unfolding of the LBD, thus opening the hydrophobic steroid binding cavity for access by steroid.
The purpose of this study was to determine the role of amino acids 547-553 of rat GR in the binding of hsp90 and steroid. We noted that this seven-amino acid sequence contains an LXXLL motif, which has been found to mediate some protein-protein interactions (28 -30). Several point mutations of this motif were, therefore, prepared to assess its function in the context of glutathione S-transferase (GST) chimeras and full-length GRs. Both the kinetics and thermal stabilities of steroid and hsp90 binding were monitored along with the susceptibility of the receptors to protease digestion. These studies demonstrated that the seven amino acids 547-553 are important for two separable activities, hsp90 binding and steroid binding. A molecular model for some of the determinants of these two activities is presented that is based on the x-ray structure of the progesterone LBD (31).

MATERIALS AND METHODS
Unless otherwise indicated, all operations were performed at 0°C. Chemicals- [1,2,4, H]Dexamethasone (Dex; 81 Ci/mmol) and the Thermo Sequenase radiolabeled terminator cycle sequencing kit were obtained from Amersham Pharmacia Biotech. [6, H]Triamcinolone acetonide (TA; 38 Ci/mmol) and 125 I-conjugated goat anti-mouse and anti-rabbit IgGs were from NEN Life Sciences Products. Nonimmune IgG and the monoclonal anti-GST antibody (clone GST-2) were from Sigma. The AC88 monoclonal IgG against hsp90 was from Stressgen (Victoria, British Columbia), the BuGR2 monoclonal IgG against the GR was from Affinity Bioreagents (Golden, CO), and the anti-GR antibodies aP1 (32) and hGR␣ (33) were gifts from Dr. Bernd Groner (Georg-Speyer-Haus, Frankfurt am Main, Germany) and Dr. George Chrousos (NICHD, National Institutes of Health), respectively. Vent DNA polymerase and the restriction enzymes were obtained from New England Biolabs. The GeneEditor in vitro Site-Directed Mutagenesis System and the Dual-Luciferase Reporter Assay System were from Promega. LipofectAMINE Reagent was obtained Life Technologies, Inc. Enzymatic manipulations were performed according to the manufacturer's specifications. Hydrofluor scintillation mixture was from National Diagnostics. TPCK-treated trypsin was obtained from Sigma and soybean trypsin inhibitor was from Calbiochem.
Colony selection was carried out on LB plates containing 100 mg/ml ampicillin (Diagene Diagnostics, Inc.) and then grown in LB broth (Quality Biologicals, Inc.). Plasmid DNAs were extracted and purified by the JETstar purification system (GENOMED GmbH, Bad Oeynhausen, Germany).
Cell Growth-Monolayer cultures of COS-7 and CV-1 cells were grown at 37°C with 5% CO 2 in Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 5 and 10% of heat-inactivated fetal bovine serum, respectively.
Assay for Steroid Binding-Transient transfection of COS-7 cells with 10 g/100-mm dish of cDNA plasmids for wild type and mutant pSVLGR and GST fusion proteins was used to overexpress the proteins, as described previously (35). Cytosols of transfected cells containing the steroid-free receptors were obtained by the lysis of cells at Ϫ80°C and centrifugation at 15,000 ϫ g as described previously (36). [  Overexpressed receptor proteins were visualized following Western blot analysis (see below) using enhanced chemiluminescence. Gel shift bands were quantitated on a Macintosh Power G3 computer using the public domain program NIH Image 1.6 (developed at the U.S. National Institutes of Health and available on the Internet). An equal size region of the film without any bands was used to calculate the background. The intensity of the specific protein band was then determined by subtracting the background value from that of the receptor protein. To normalize the steroid binding per unit receptor protein, the specific disintegrations/min of [ 3 H]steroid bound to each receptor were divided by the above calculated amount of receptor protein.
Assay for Biological Activity-Forty ng of wild type and mutant pSVLGR constructs, 1 g of a GREtkLUC reporter, 200 ng of a Renilla pRLtkLUC reporter (Promega), and enough pBSKϩ vector to bring total DNA concentration to 3 g/60-mm dish, were transiently transfected into CV-1 cells. As recommended by the supplier, 10 l of Lipo-fectAMINE was combined with the 3 g of DNA in serum-free Dulbecco's modified Eagle's medium and added to the cells for 3 h. At the end of the transfection period, cells were washed, replenished with complete media, and incubated overnight at 37°C. The cells were then treated with EtOH Ϯ Dex or Dex 21-mesylate (Dex-Mes) for an additional 24 h. At the end of the incubation period, the cells were harvested in 1ϫ Passive Lysis Buffer (0.5 ml/dish, Promega). Luciferase activity was assayed using luciferin reagents in the Promega Dual Luciferase Reporter Assay kit.
Proteolytic Digestion of Receptors-The proteolytic fragments were generated by incubating 30% cytosol containing steroid-free receptor with 0 to 25 g/ml trypsin (TPCK treated) for 1 h at 0°C. A 10-fold excess of soybean trypsin inhibitor was then added to prevent further proteolysis, and the samples were frozen at Ϫ80°C.
Reconstitution of GST/GR Hsp90 Heterocomplexes-Aliquots (200 l) of undiluted cytosol from transfected COS-7 cells were immunoadsorbed with 15% anti-GST antibody or nonimmune IgG and subsequently incubated with 8-l pellets of protein A-Sepharose. Immunoadsorbed GST fusion proteins were stripped of associated hsp90 by incubating the immune pellets an additional 2 h at 4°C with 0.5 M NaCl, followed by one wash with 1 ml of TEG buffer (10 mM TES, pH 7.6, 50 mM NaCl, 4 mM EDTA, 10% glycerol) and a second wash with 1 ml of Hepes buffer (10 mM Hepes, pH 7.4). Immune pellets containing stripped fusion proteins were suspended in 50 l of rabbit reticulocyte lysate. Dithiothreitol (1 l) was added to each incubation to a final concentration of 5 mM, and 5 l of an ATP-regenerating system (50 mM ATP, 250 mM creatine phosphate, 2 mM MgOAc, and 100 units/ml creatine phosphokinase) were added to all assays to yield a final assay volume of 56 l. The assay mixtures were incubated for 20 min at 30°C with suspension of the pellets by shaking the tubes every 5 min. At the end of the incubation, the pellets were washed twice with 1 ml of ice-cold TEGM buffer (TEG buffer with 20 mM sodium molybdate), and GST fusions and hsp90 were assayed in each sample by Western blotting.
A portion of each immune pellet was assayed for steroid binding by incubation overnight in 100 l of TEGM buffer plus 4 mM dithiothreitol and 50 nM [ 3 H]triamcinolone acetonide (TA). Samples were then washed twice with 1 ml of TEGM and counted by liquid scintillation spectrometry as described previously (13). The steroid binding is expressed as counts/minute of [ 3 H]TA bound/anti-GST immune pellet prepared from 100 l of cytosol.
Assay of Steroid Dissociation-Prior to immunoadsorption, the BuGR2 antibody was prebound to protein A-Sepharose pellets by incubating 65 l of a 20% slurry of Protein A-Sepharose for 2 h at 4°C with 45 l of antibody at a concentration of 100 g/ml and 200 l of 10 mM Hepes, pH 7.4, followed by centrifugation and washing with Hepes buffer. The transfected wild type GR and L550S/L553S mutant GR that had been prebound with [ 3 H]Dex Ϯ 500-fold excess [ 1 H]Dex were allowed to dissociate for various lengths of time and then adjusted to contain 20 mM molybdate to block further dissociation of hsp90. The receptors were then immunoadsorbed from 600 l of diluted cytosol by rotation for 2 h at 4°C with 13 l of protein A-Sepharose prebound with BuGR2 antibody, followed by washing the immunopellets three times with 1 ml of TEGM.
Gel Electrophoresis and Western Blotting-For assay of GST fusion proteins and associated hsp90, immune pellets were boiled in SDS sample buffer with 10% ␤mercaptoethanol, and proteins were resolved on 7% SDS-polyacrylamide gels. Proteins were then transferred to Immobilon-P membranes and probed with 0.01% aP1 for fusion proteins, 1 g/ml BUGR-2 for full-length GR, and 1 g/ml AC88 for hsp90. The immunoblots were then incubated a second time with the appropriate 125 I-conjugated counter antibody to visualize the immunoreactive bands. For quantitative Western blotting, the GST/GR HBD fusion proteins, or hsp90, were excised from the immunoblots and counted for associated 125 I-conjugated goat anti-rabbit (for GR) or anti-mouse (for hsp90) radioactivity. The determined ratio of hsp90 per unit of GR chimera was then expressed as a percent of that found for the GST520C control or, in the case of full-length GR, as a percent of the zero time control.
Alternatively, full-length or chimeric GR that were overexpressed in COS-7 cells were usually separated on 13% SDS-polyacrylamide gels and visualized by enhanced chemiluminescence as previously reported (26). Trypsinized receptors were separated on 15% SDS-polyacrylamide gels using 30% cytosol samples as described previously in detail (26). The wild type and mutant pSVLGR receptors were detected using an anti-GR antibody, aP1 (1:10,000), whereas the wild type and mutant GST-GR fusion receptors were detected using an anti-GST antibody (1:2,000) or hGR␣ (1:5,000).

GR Mutations and
Truncations-Examination of the sevenamino acid sequence previously found to be required for hsp90 binding to GR (amino acids 547-553) (17) revealed the majority of an LXXLL sequence at the carboxyl end. Such sequences have recently been shown to be important for protein-protein interactions between steroid receptors and coactivators (28 -30). Furthermore, a double mutation of this motif (L553G/ L554G) has been reported to eliminate the cell-free affinity labeling of GR by Dex-Mes and the biological activity of Dex in whole cells (37). Thus, we suspected that this same LXXLL sequence might be a critical component for hsp90 binding to GR. To investigate this hypothesis, we prepared the Leu 553 and/or Leu 550 point mutants of full-length GR (from pSVLGR) and of GST chimeras containing the GR LBD plus an additional amino-terminal 3 (547-795 ϭ 547C) or 30 (520 -795 ϭ 520C) amino acids (Fig. 1). The GST547C construct was selected in order to examine the effect of mutations in the context of a chimera with the minimum GR segment that both bound steroid and contained the sequence of 547-553 (17). However, trypsin digestions of steroid-free GST547C would not be expected to yield the characteristic 16-kDa fragment (3). In order to be able to look at the effect of these mutations on the tertiary structure of GR, the GST520C chimera was also selected, as proteolysis of this ligand-free chimera does afford the 16-kDa species (38). This construct would further reveal any contributions of the hinge region in hsp90 binding to GR. In order to examine the biological consequences of these mutations, fulllength GR mutants were prepared.
Steroid Binding Activity and Expression of Mutant Proteins-Western blots indicated that each mutant was expressed at about same level as the wild type construct (Fig. 2,  insets). Surprisingly, though, the level of [ 3 H]steroid binding was not similarly constant. The single mutations decreased the amount of binding of the chimeras by about 75% while the double mutant displayed almost no binding (Fig. 2, A and B). In all cases, the effect of each mutation was muted in the context of the full-length receptor (Fig. 2C). However, Scatchard analysis revealed that each mutation did not alter the affinity of those receptors that did bind steroid (Table I). This decrease in binding with maintenance of wild type affinity could be explained by a heterogeneity among the receptors, some of which were unable to bind steroid. Alternatively, the kinetics of binding could be very different among the mutant GRs. However, [ 3 H]Dex binding to each of the GST520C chimeras was found to reach and maintain plateau values over 2-17 h at 0°C (data not shown). Therefore, the decreased binding of the mutants GRs in Fig. 2 did not simply result from an insufficient time to reach maximal binding but appeared to reflect receptor heterogeneity.
Effect of Mutations on Interactions with Hsp90 -We had previously found that GST547C, overexpressed in COS-7 cells, was associated with hsp90 (17). One model for the decreased steroid binding capacity without altered affinity of the L550S and/or L553S mutants is that hsp90 binding to GR has become more labile and only a fraction of the receptors remain associated with hsp90. Thus, under cell-free conditions at 0°C that do not permit hsp90 reassociation with GR (11), only those receptors that retained hsp90 would bind with an affinity that might be indistinguishable from wild type receptors. This hy- pothesis was addressed by heating ligand-free GRs at 20°C to accelerate the dissociation of hsp90 and the loss of steroid binding. For GST547C Ϯ L550S chimeras, the loss of steroid binding activity was found to be identical (t1 ⁄2 ϭ 60 min; data not shown). These results indicated that the decreased binding of GST547C/L550S was not a consequence of weaker hsp90 binding that reduced the fraction of hsp90-bound GR and, thus, the steroid binding capacity of the total GR.
Another possibility was that the mutations altered the folding of the GR proteins such that there was a partitioning of the receptors between GR that could and could not bind hsp90. Thus, those molecules that reached a conformation capable of binding hps90 might then bind steroid with the same affinity as wild type receptors. To assess this possibility, we quantitated the steroid and hsp90 binding of the GST520C mutants. These chimeras were used because they seemed to have the same tertiary structure as the wild type receptors, as determined by protease digestion. A 16-kDa tryptic fragment, corresponding to the sequence 652-795 (25,38), is obtained after trypsin digestion of steroid-free, full-length receptors and chimeras containing sequences upstream of amino acid 547 (such as 537C, 520C, and 494C) but not from constructs starting at 547C (3, 38).

B-E, lane 2).
Even the L550S/L553S double mutant with little residual binding contained large amounts of hsp90 (Fig. 3E,  lane 2). In all cases, the amount of associated hsp90 was dramatically reduced in the stripped pellets (lane 3 of Figs. 3, B-E) and restored after incubation with reticulocyte lysate (lane 4 of Fig. 3, B-E). Most surprising, however, was that the ratio of hsp90 to GR was relatively constant for all constructs (Fig. 3F). Thus, the quantity of hsp90 associated with the wild type 520C and each mutant was about the same even though there were major differences in total steroid binding activity. This demonstrates that, while hsp90 has been found to be necessary for steroid binding (10,11), hsp90 binding to GR is not sufficient for steroid binding. These data also establish that the decreased binding of the L550S, L553S, and L550S/L553S mutations of GST520C occur without any major disruption of hsp90 binding to GR. Therefore, the amino acid sequence of 547-553 is involved in the expression of two independent activities of GR, steroid binding and hsp90 binding.
Sequences between Amino Acids 520 and 546 Are Not Required for Hsp90 Binding-We have previously demonstrated that GST547C contains sufficient amino acid residues for the association of receptors with hsp90 in intact cells. However, the possibility remained that residues between 520 and 546 could contribute to hsp90 binding and were responsible for the unaltered binding of hsp90 to the GST520C mutants in Fig. 3. The same experiments were therefore conducted in the context of the GST547C chimera to see if a similar insensitivity of hsp90 binding existed in the presence of mutations that reduced steroid binding. As shown in Fig. 4, the ratio of hsp90 to GR was not appreciably different for GST520C and GST547C. Most significantly, this ratio was virtually identical between GST520C and GST547C/L550S, which retained only 25% of the steroid binding activity of the wild type GST547C (see Fig. 2A). Therefore, as for the GST520C chimeras, mutations in the LXXLL sequence of 547-553 can dramatically reduce steroid binding activity with little or no effect on hsp90 binding. Furthermore, the sequence of 547-553 is sufficient to convey steroid and hsp90 binding to the rest of the GR LBD (554 -795) as  residues between 520 and 546 do not make significant contributions. Whole Cell Biological Activity of Full-length GR Mutants-Dex inducibility of a transiently transfected reporter construct by full-length receptors containing a single mutation was uniformly higher (1.3-2.6-fold) than that by wild type GR (Fig. 5). In contrast, the activity of the double mutant was decreased 5-fold (0.19 Ϯ 0.06-fold; S.D., n ϭ 3). These changes in biological activity could not be accounted for by the observed alter-  C, D, and E) replicate native GST520C-hsp90 heterocomplexes were immunoadsorbed and samples were assayed for GR-associated hsp90 by SDS-PAGE and immunoblotting. To prepare the graph, the GST/GR HBD fusion proteins and co-immunoadsorbed monkey hsp90 on the immunoblot were excised from a separate gel and counted for 125 I-conjugated goat anti-rabbit or anti-mouse IgG radioactivity. After correction for the relative amount of hsp90, the values were expressed as a percent of the construct B control. The data represent the means from two experiments. The immune pellets were washed, and the GST/GR chimeras and hsp90 were resolved by SDS-PAGE and Western blotting. The bar graph presents the mean hsp90/GR ratio (Ϯ range) determined from duplicate samples as described for Fig. 3. Similar results were observed in a second experiment.

FIG. 5. Biological activity of full-length wild type and mutant
GRs. Triplicate dishes of CV-1 cells were transiently transfected with reporter (GREtkLUC), control plasmid (Renilla pRLtkLUC), and the indicated full-length GR plasmid. Cells were treated with EtOH Ϯ 1 M Dex or Dex-Mes (DM) and then harvested for quantitation of the luciferase activity per unit of Renilla activity as describe under "Materials and Methods." For each GR, the average fold induction Ϯ S.D. of luciferase activity by the various treatments was expressed relative to that for EtOH, which was defined as 1. Similar results were obtained in four additional experiments. ations in steroid binding capacity (Fig. 2C) or affinity (Table I), suggesting an effect of the mutations on processes other than those investigated so far at 0°C. Of further interest was the marked reduction in agonist activity of the antiglucocorticoid Dex-Mes in the mutant receptors relative to the wild type receptor.
Steroid Binding Properties of Full-length L550S/L553S Mutant GR-The discrepancy between the effect of LXXLL motif mutations on steroid binding versus biological activity of the full-length GR (Figs. 2C versus 5) could be due to the different temperatures of the two assays (0 versus 37°C). In fact, when steroid-free receptors were heated at 20°C, the full-length double mutant (WT/L550S/L553S) lost steroid binding activity much more rapidly than did the wild type GR (Fig. 6). This could result from either a more rapid dissociation of hsp90 from the double mutant or weakened intramolecular interactions in the double mutant, thus permitting a more facile opening of the LBD. When receptors pre-bound with steroid were heated, conditions that cause the activation/transformation of receptorsteroid complexes (39), the loss of steroid was again more rapid from the double mutant (Fig. 7). However, this more rapid loss of steroid from the activated mutant complexes occurred without any preferential loss of hsp90 (circle data points). Thus, as seen above in Figs. 3 and 4, steroid binding capacity can be altered without affecting the binding of hsp90.
The more rapid rate of dissociation of steroid from the double mutant (Fig. 7), and the lack of biological activity (Fig. 5), could be due either to a mutation-induced blockage of activation, thus preventing the activation/transformation of the mutant receptor-steroid complex to the more slowly dissociating form (40 -42), or to more rapid dissociation of steroid from the activated form of mutant GR. To distinguish between these two possibilities, we made use of the fact that sodium molybdate blocks the conversion of the rapidly dissociating, unactivated receptorsteroid complex to the more stable, activated complex for both GR (41) and estrogen receptors (43). Therefore, the rates of steroid dissociation at 20°C from unactivated (in the presence of molybdate) and activated (in the absence of molybdate) double mutant GR-[ 3 H]Dex complexes were determined. As shown in Fig. 8, the receptor-steroid complex was quite stable in the presence of molybdate. The slight increase in steroid binding H]Dex and then allowed to dissociate at 20°C. At the indicated times, the samples were placed in ice to stop the dissociation. All samples were treated with 20 mM molybdate at the end of the assay (120 min) to block further dissociation of hsp90. Receptor with its associated hsp90 was then immunoadsorbed using either a nonimmune antibody or anti-GR antibody (BUGR2). The graph displays the bound [ 3 H]Dex at each time as percent of initial binding for full-length, wild type GR (pSVLGR, f), and the L550S/L553S double mutant (Ⅺ) in addition to the hsp90/GR ratio for full-length wild type GR (q) and the L550S/L553S double mutant (E). The autoradiogram (inset) shows the Western blots of immunoadsorbed GR and coadsorbed hsp90. These bands were incubated with 125 I-labeled secondary antibody, excised, and counted to generate the hsp90/GR ratios in the graph. Similar results were obtained in a second experiment. with time at 20°C (triangles) was most likely due to the fact that the initial incubation period of 2 h was not sufficient to bind all of the receptors. However, in the presence of excess non-radioactive steroid to prevent the rebinding of [ 3 H]Dex, the rate of dissociation of steroid in the absence of molybdate was much faster (t1 ⁄2 Ϸ 20 min) than in the presence of molybdate (t1 ⁄2 Ϸ 80 min). Thus, a temperature-induced modification of the full-length L550S/L553S mutant GR-steroid complex has occurred which causes a more rapid dissociation of steroid. Furthermore, this temperature-induced change is inhibited by sodium molybdate. We interpret this as indicating that the L550S/L553S mutation of the full-length GR does not block activation but rather accelerates the dissociation of bound steroid once the activated complex has been formed. This rapid loss of steroid from activated complexes would further explain why the full-length double mutant displays good steroid binding ability at 0°C (Fig. 2C) but is biologically inactive (Fig. 5).
Trypsin Digestion of Steroid-free Mutant Receptors to Yield the 16-kDa Fragment-The atypical accelerated dissociation of steroid from activated versus unactivated mutant complexes in Fig. 8 is most readily explained as resulting from changes in GR tertiary structure due to mutations of the LXXLL motif. We, therefore, looked for additional evidence of conformational differences in the mutant GRs. We previously documented that steroid-free GR is uniquely vulnerable to trypsin digestion to yield a 16-kDa fragment (25) corresponding to the sequence of amino acids 652-795 (38). The formation of this 16-kDa tryptic digest product is inhibited by ligand binding, due to a steroidinduced conformational change, thus indicating that the trypsin-sensitive residue (Arg 651 ) is in a conformationally mobile portion of the LBD (25,44). In this same trypsin digestion assay with the GST520C chimeras, the wild type construct yielded the 16-kDa fragment but much less 16-kDa species was obtained from the mutant chimeras (Fig. 9A). Furthermore, there was a general correlation between the reduced steroid binding capacity (Fig. 2B) and the decreased yield of tryptic 16 kDa for the various GST520C chimeras. To confirm these observations, the trypsin digestion analysis was repeated with the full-length receptors (Fig. 9B). Again, there was a good correlation between reduction in steroid binding capacity (Fig.  2C) and yield of the 16-kDa fragment. This was most clearly seen with the digests of the full-length GR exposed to 14 and 25 g of trypsin.

DISCUSSION
The seven-amino acid sequence of 547-553 of rGR is established here to participate in two distinct activities. This sequence was previously reported to be essential for hsp90 binding to GR (17). The absence of steroid binding in constructs lacking this sequence was thought to result from the loss of hsp90 binding. However, mutations within this sequence now illustrate that, while hsp90 binding to GR is required for steroid binding (10,11), mutations of Leu 550 and Leu 553 to serine dramatically reduced steroid binding (Fig. 2) and biological activity (Fig. 5) without altering the binding of hsp90 (Figs. 3, 4, and 7). Therefore, the sequence of 547-553 contributes to the expression of two independent and separable GR activities, steroid binding and hsp90 binding. The steroid binding activity depends on a LXXLL motif at amino acids 550 -554.
LXXLL motifs have recently been described to mediate the protein-protein interactions of transcriptional cofactors with steroid/nuclear receptors (28 -30). The presence of such a motif at the COOH terminus of amino acids 547-553 of rGR suggested that it might similarly be crucial for the protein-protein interactions required in hsp90 binding and steroid binding. Convincing evidence for the role of the LXXLL sequence in steroid binding per se came from the rates of steroid dissociation in the presence and absence of molybdate from full-length GR complexes containing the L550S/L553S double mutation (Fig. 8). The rate of steroid dissociation from the wild type GR decreases after being exposed to activating conditions, such as elevated temperatures (40 -43). In contrast, the L550S/L553S mutations resulted in an accelerated dissociation of bound steroid after activation. Furthermore, discrepant rates of steroid dissociation from GR complexes were observed under conditions causing no difference in the amount of hsp90 retained by the wild type or mutant GR (Fig. 7).
An examination of the predicted GR structure, based on the x-ray structure of the closely related human PR LBD (31), suggests a molecular model for the role of the LXXLL motif in steroid binding. This sequence is thought to be ␣-helical and part of helix 1 of the LBD (31,45). Therefore, as adjacent amino acids of an ␣-helix are 3.5 residues apart, the mutations of Leu 550 and/or Leu 553 would affect amino acid substituents that all lie on one side of the ␣-helix. Furthermore, Leu 550 and Leu 553 of rGR are well located for intramolecular van der Waals interactions that would stabilize the tertiary structure of the GR LBD (Fig. 10A). Leu 550 (yellow stick model) would form hydrophobic bonds with residues both in the bend between helices 3 and 4 (Pro 600 of rGR) and in helix 9 (Tyr 711 and Glu 714 of rGR) while Leu 553 (red) would contact Pro 600 (see figure legend for details). Leu 554 (blue) would interact with Trp 595 , Ile 599 , Prp 600 , Leu 682 , Thr 686 , Tyr 711 in helix 3, between helices 3 and 4, and in helix 8 of rGR. It should be noted that almost all of these residues, including Leu 550 , Leu 553 , and Leu 554 , are completely conserved between rGR and the human PR (31). We therefore propose that the LXXLL sequence of FIG. 9. Tryptic digestion patterns of steroid-free GST520C and full-length GR. Cell-free extracts of COS-7 cells that had been transiently transfected with plasmids encoding (A) GST520C or (B) full-length receptors Ϯ point mutations were treated, in the absence of added steroid, with the indicated concentrations of trypsin for 1 h at 0°C. Tryptic fragments, especially the 16-kDa species (indicated by arrow), were separated by SDS-polyacrylamide gel electrophoresis and visualized by Western blotting with anti-GR antibody (aP1) followed by enhanced chemiluminescence. The positions of the molecular weight markers was determined by staining with Ponceau S and marking with fluorescent paint. Comparable results were obtained in 1 and 2 additional experiments for A and B, respectively. 550 -554 is a hydrophobic patch that makes important contributions to the stability of the GR LBD tertiary structure and, consequently, steroid binding activity.
The importance of this LXXLL sequence can be further appreciated from a different view of the LBD, in which the aminoterminal half of the domain (547-651 ϭ helices 1-6) forms one side of the steroid binding pocket (Fig. 10B). The two sides of the pocket are "hinged" in the region of Arg 651 , which appears not to exist as an ␣-helical segment (44), and closed at the other end, in part, by the hydrophobic interactions of Leu 550 , Leu 553 , and Leu 554 .
Additional considerations of the hinged pocket model offer explanations as to why the magnitude of reduced binding capacity at 0°C following various mutations of the LXXLL motif was not uniform in all receptor constructs examined. A single mutation in the GST chimeras of the GR LBD caused a greater loss of steroid binding activity than did the presence of both mutations in the full-length receptor (Fig. 2). Thus, GR sequences upstream of amino acid 520 were able to counteract some of the disruptive effects of the mutations of Leu 550 and Leu 553 . This does not appear to result from a stabilization of hsp90 binding as the rate of loss of steroid binding activity upon heating steroid-free receptors at 20°C was actually faster for the full-length L550S/L553S double mutant (t1 ⁄2 ϭ 20 min, FIG. 10. Predicted rGR residue interactions based on x-ray structure of human PR. A, predicted intramolecular contacts of Leu 550 , Leu 553 , and Leu 554 in LBD of rGR. The backbone structure of the human PR LBD (31) is shown (gold), along with the bound steroid (purple ball and stick model). The orientation of the PR LBD is such that the beginning of helix 1, and the end of helix 3, is closest to the viewer. The letter N designates the amino terminus of the structure (Gln 682 of PR). Unless otherwise indicated, all numbering is for rGR. The predicted locations of GR leucines and the residues which they contact by stabilizing hydrophobic interactions (determined by the CSU software program on the Protein Data Base) are as follow: L550 (yellow stick model ϭ Leu 687 of PR) interacts with CPK space filling residues for Pro 600 (lavender) in bend between helices 3 and 4 and with Tyr 711 (dark green) and Glu 714 (light green) of helix 9, Leu 553 (red stick model ϭ L690 of PR) interacts with Pro 600 (lavender), and Leu 554 (dark blue stick model ϭ L691 of PR) interacts with Pro 600 (lavender) and Tyr 711 (dark green) in addition to the cyan colored residues in helix 3 (Trp 595 ), between helices 3 and 4 (Ile 599 ), and in helix 8 (Leu 682 and Thr 686 ). B, view of PR LBD showing the hinged pocket structure. The letters N and C indicate the amino and carboxyl termini, respectively, of the x-ray structure. The identity of several of the ␣-helices is marked. The two halves of the PR/GR LBD backbone structure have been divided between the residues amino-terminal (light green) and carboxyl-terminal (blue) of R651 (red stick model). The steroid (lavender ball and stick model) resides in the cavity near the hinge, constituted by the amino acids immediately adjacent to and including Arg 651 . The two sides of the pocket are held closed by hydrophobic interactions between, inter alia, Leu 550 , Leu 553 , and Leu 554 (yellow space filling models) and helices 8 and 9. With mutation of Leu 550 and Leu 553 , the two halves of the LBD would be able to dissociate and expose the interior of the LBD. Structures were drawn with RasMac (v2.5). Fig. 6) than for the wild type GR or GST547C/L550S (t1 ⁄2 ϭ 60 min for both ( Fig. 6 and data not shown)). Furthermore, there was no significant difference in the hsp90/GR ratio for wild type GR versus any of the mutant receptors (Figs. 3, 4, and 7). Instead, we suspect that this reflects intramolecular associations with other portions of GR. Definitive data are lacking because the x-ray structures of receptor LBDs do not include more amino-terminal residues (31,45). However, interactions have been reported for LBDs with the amino-terminal regions of androgen (46 -48), estrogen (49,50), and progesterone (51) receptors and implicated for the amino terminus of GR by the requirement of AF1 for steroid regulated transactivation of the murine mammary tumor virus enhancer (52).
The hinged pocket model can further account for the biological activities of the mutant receptors. Single and double mutations of the full-length GR are tolerated at 0°C (Fig. 2). At the elevated temperatures that cause activation, any destabilization of the complexes resulting from the dissociation of non-receptor proteins like hsp90 may be partially compensated by the steroid-induced compaction of the LBD (45,53). However, the double mutation causes more extensive disruptions of the hydrophobic clasp around Leu 550 . The loss of these important interactions allows the hinged pocket to open more easily at 37°C, with the subsequent dissociation of steroid and the production of a transcriptionally inactive form of GR.
The hinged pocket model of the GR LBD also offers an attractive hypothesis for how mutations around Leu 550 have structural consequences that would affect trypsin digestion at Arg 651 to produce the 16-kDa tryptic fragment, which corresponds to amino acids 652-795 (38) (Fig. 9). Trypsin digestion can be an especially revealing probe of GR structure (25,26) and, in some cases, is a more sensitive probe than steroid binding (3). The GR LBD contains many basic residues but is cleaved by trypsin at only very few sites (25,26,38), presumably due to most of the residues being hidden by the folded tertiary structure and associated proteins. Trypsin digestion of steroid-free GR does not open up the LBD after cutting at Arg 651 in the hinge because the two halves of the LBD (518 -651 and 652-789) remain non-covalently associated (26,38). However, mutation of the hydrophobic clasps at the other end of the hinged pocket (Leu 550 and Leu 553 ) would be expected to reduce the binding of the two tryptic fragments, thereby allowing the dissociation of the two peptides. This would facilitate further trypsin digestion of the LBD, and the 16-kDa fragment, at previously inaccessible residues, just as was observed (Fig.  9). We feel that the observed differences in trypsin digestion reflect a functionally relevant loosening of the receptor tertiary structure as opposed to a nonspecific denaturation because 1) all of the mutants retain nearly equivalent levels of hsp90 binding and 2) the steroid binding at 0°C of each of the fulllength mutant GRs is similar while the biological activities are quite dissimilar.
Four mutant GRs have been described with properties similar to those defined in this study. Mouse GR with the Leu to Gly mutations at positions equivalent to 553 and 554 of rGR was transcriptionally inactive (37). We predict that the steroid binding capacity of this L553G/L554G double mutant will be relatively unchanged at 0°C but that prebound steroid will rapidly dissociate at elevated temperatures following activation and that trypsin digestion will not afford the 16-kDa tryptic fragment, due to inadequate intramolecular contacts of the hinged cover of the LBD. The rGR double mutant K597I/ P600L, which modifies the proline contacted by Leu 550 , Leu 553 , and Leu 554 , lost all Dex binding at 0°C but retained hsp90 binding (54). The effect of the P600L mutation in isolation is not known but we suspect that the substitution by Leu is detrimental not because of an inability to interact with the leucines at positions 550, 553, and 554 but rather because the absence of the proline may prevent the bend between helices 3 and 4. Mutation of the mouse equivalent of Leu 682 , which is predicted to interact with Leu 554 (Fig. 10A), to phenylalanine produced an activation labile phenotype like that of the L550S/ L553S mutant. There was no loss in Dex affinity (cf. Table I) but 200-fold higher Dex concentrations were required for biological activity (55). However, it is not clear what caused the loss of biological activity as the hydrophobic interactions should be maintained by the L682F mutation. Finally, the mutation of human GR equivalent to L771F in rGR results in biologically inactive, activation-labile receptors, in which steroid binding is normal at 0°C but lost once the receptor is transformed to the activated state (56,57). Thus, the loss of steroid binding, especially at elevated temperatures, in the presence of apparently normal hsp90 binding may be a relatively common phenotype.