The X-ray Structure of RU486 Bound to the Progesterone Receptor in a Destabilized Agonistic Conformation

Abstract

Here we describe the 1.95 Å structure of the clinically used antiprogestin RU486 (mifepristone) in complex with the progesterone receptor (PR). The structure was obtained by taking a crystal of the PR ligand binding domain containing the agonist norethindrone and soaking it in a solution containing the antagonist RU486 for extended times. Clear ligand exchange could be observed in one copy of the PR ligand binding domain dimer in the crystal. RU486 binds while PR is in an agonistic conformation without displacing helix 12. Although this is probably because of the constraints of the crystal lattice, it demonstrates that helix 12 displacement is not a prerequisite for RU486 binding. Interestingly, B-factor analysis clearly shows that helix 12 becomes more flexible after RU486 binding, suggesting that RU486, being a model antagonist, does not induce one fixed conformation of helix 12 but changes its positional equilibrium. This conclusion is confirmed by comparing the structures of RU486 bound to PR and RU486 bound to the glucocorticoid receptor.

The drug RU486, also known as mifepristone, is the only clinically approved antiprogestin (trade name Mifegyne® or Mifeprex®). It is applied to terminate pregnancy and has been clinically tested in many more indications (1, 2). Recently, it was shown that RU486 can prevent mammary tumorigenesis in Brca1/p53-deficient mice, implying a use for RU486 in breast cancer therapy (3).

RU486 exerts its clinical effect by binding to the ligand binding domain of the progesterone receptor, although RU486 can also bind to the glucocorticoid receptor (GR)² and weakly to the androgen receptor (4). All these nuclear receptors are close sequence homologs (5). Because the anti-GR activity of RU486 might be problematic in chronic administration (1), past research has focused on finding RU486 variants with more selectivity (4, 6–10) (Table 1).

TABLE 1.

Structure and activities of progestins and antiprogestins

ago, agonistic activity; ant, antagonistic activity, efficacy is measured by curve height at maximal effect and expressed as percentage of reference compound effect (see “Experimental Procedures”); nr, no response; nd, not determined. The indicated error margins are S.D. calculated over all measurements (minimum of four data points). PR activity was measured on the B isoform. The top left panel shows the nomenclature of the steroid ring for clarity.

Despite the clinical importance of RU486, there is currently no three-dimensional structure of it bound to PR, its principal target. However, other complexes have been informative, such as that between the PR ligand binding domain and asoprisnil, which is biochemically a full antagonist and is chemically related to RU486 (Table 1 and Ref. 11). Also informative is the crystal structure of RU486 bound to the GR LBD (12). In all these structures the antagonists bind to a receptor conformation in which the C-terminal helix (called helix 12) is displaced compared with structures of bound agonists. This so-called helix 12 displacement was first seen in the structure of raloxifene bound to the estrogen receptor α, and it is commonly thought to be a general nuclear receptor mechanism (5, 13).

The indications from x-ray structures that in PR, RU486 can induce displacement of helix 12 are supported by biochemical data. For instance the truncation of the PR C terminus induces RU486 to act as agonist (14). Also, the C terminus of PR becomes prone to proteolysis when RU486 binds (15). Finally, from modeling RU486 into the structure of bound progesterone, it has been concluded that the 11β substitution of RU486 (Table 1) is sterically incompatible with the agonistic conformation of helix 12 (16).

Despite the evidence that RU486 can induce displacement of helix 12, it is still unclear if RU486 obligately dissociates helix 12 through steric repulsion or if RU486 allows multiple positions of helix 12 but changes their dynamic equilibrium. In this latter theory, called the dynamic model, RU486 would also be able to bind when helix 12 is in an agonist conformation. Indeed, under rare conditions RU486 can function as an agonist (7), and the compound asoprisnil, in vivo, is known as a partial agonist (11). The use of corepressor peptides in crystal complexes, such as in the PR-asoprisnil complex (11), might lock helix 12 into its final antagonist position, whereas the compound alone would have more subtle effects. The dynamic model controversy has also arisen with other nuclear receptors, indicating its wider scope (13, 17).

To further elucidate the mechanism of RU486, we have determined the three-dimensional structure of RU486 bound to PR. For this, we developed a novel protocol in which ligands are exchanged in an existing PR crystal in which norethindrone is bound. The crystal lattice restricts the position of helix 12 to the agonist position, but RU486 is still able to bind, proving that it is sterically compatible with an agonist position of helix 12 and suggesting that RU486 works through changing the dynamic equilibrium of helix 12. Comparing our structure with the asoprisnil complex gives insight into the mechanism of helix 12 destabilization, which is confirmed by comparing our structure to that of RU486 bound to GR.

EXPERIMENTAL PROCEDURES

Expression and Purification of PR-LBD

The PR LBD, comprising residues 678–933, was cloned in pET15b (Novagen). Expression was performed in Escherichia coli BL21(DE3) star (Invitrogen) by overnight induction at 20 °C in the presence of 10 μm norethindrone. Bacteria were lysed in buffer A (50 mm Tris, pH 7.8, 250 mm NaCl, 10% glycerol) with 0.4 mm pefabloc (Roche Applied Science) and 50 μm norethindrone and purified on nickel-nitrilotriacetic acid. Fractions were eluted with buffer A to which we added 10 mm β-mercaptoethanol, 10 μm norethindrone, and 100 mm imidazole. Elution fractions of the first three column volumes were discarded, and the others were collected and treated with 2.5 wt/wt % thrombin (Kordia) overnight at 4 °C to remove the N-terminal His tag. Thrombin was removed by adding benzamidine-Sepharose (GE Healthcare), centrifuging for 10 min at 5000 × g, and harvesting the supernatant. To make the final crystallization sample, the protein was dialyzed to buffer A to which we added 1 mm EDTA, 1 mm dithiothreitol, and 10 μm norethindrone and subsequently concentrated in a stirring cell to about 2 mg/ml as measured by its absorption at 280 nm. The sample was stored at 4 °C.

Ligand Replacement

Crystals of the PR LBD in complex with norethindrone were grown at room temperature from 2-μl drops hanging over a mother liquor of 12–25% polyethylene glycol 4000, 0.1 m Hepes, pH 6.5, and 100 mm Li₂SO₄. Drops consisted of 1 μl of protein sample and 1 μl of mother liquor. Crystals usually appeared after about 3 days and were kite-shaped.

For reference, one crystal was transferred to a cryoprotectant solution of 80% mother liquor and 20% glycerol, dipped in liquid nitrogen, and shipped for data collection to 1.6 Å at 100 K. Another crystal was transferred to mother liquor to which 10 μm RU486 was added. The crystal was stored in a sitting drop at room temperature. The solution surrounding the crystal was replaced by fresh solution 10 times over a period of 10 days. After this period, the crystal was frozen as described above and shipped for data collection to 1.95 Å at 100 K.

All data were collected by the MXpress service of the European Synchrotron Radiation Facility, Grenoble, France. Structures were solved and refined using the ccp4i interface of the CCP4 software suite (18). PDB entry 1A28, stripped of solvent, ligands and cofactors, was used for molecular replacement. The structure was refined with refmac5 (19) using two TLS (translation, liberation, and screw) domains, one for each protein chain and ligand. Corina (20) was used to generate geometry restraints for the ligands. For final data statistics, see Table 2. For electron density, see Fig. 1 and supplemental Fig. 1. The structure and structure factor of the RU486 complex were submitted to the protein data bank (PDB identifier 2W8Y).

TABLE 2.

Final crystallographic data and refinement statistics

Data in parentheses indicate the last resolution shell. r.m.s.d., root mean square deviation.

	Norethindrone	RU486
Space group	P21	P21
Unit cell (Å3)	57.71 × 64.26 × 70.12	58.16 × 63.90 × 70.05
β-Angle (°)	95.65	95.57
Wavelength (Å)	0.939	0.934
Resolution (Å)	69.84-1.55 (1.59-1.55)	69.67-1.95 (2.06-1.95)
Completeness	98.5% (85.9%)	99.8% (99.3%)
Rpima	0.058 (0.378)	0.047 (0.284)
Mn(I/σ)	11.4 (1.9)	13.9 (3.1)
R factor/Rfreeb	0.156/0.187	0.174/0.208
Protein atoms	4111	4113
Ligand atoms	44	54
Water molecules	573	199
Other molecules e.g. glycerol (SO4)2−	35	21
r.m.s.d. bonds, Å	0.014	0.008
r.m.s.d. angles (°)	1.4	1.096
B-factors (average Å2)
Main chain	16.2	24.2
Side chain	18.2	24.9
Water molecules	30.3	27.2
Ligand A-chain	16.5	25.6
Ligand B-chain	17.1	26.4
Wilson B-factor (Å2)	16.4	21.2
PDB identifier	Identical to 1SQN	2W8Y

^a R_pim = Σ_h([1/(N − 1)]1/2Σ_i|Ii(h) − 〈I(h)〉|)/Σ_hΣ_iI_i(h), where I is the observed intensity, and 〈I〉 is the average intensity of multiple observations from symmetry-related reflections. It is an indicator of the precision of the final merged and averaged dataset.

^b R_free was determined using 5% of the data.

FIGURE 1.

F_o − F_c OMIT electron density around the ligands in the PR LBD after soaking. The map is contoured at three standard deviations (3σ). A, the PR copy in the asymmetric unit, which shows norethindrone bound. Residue Met-909 assumes a double conformation. B, the other PR copy, which shows electron density for RU486 and displacement of Met-909. The map was calculated after taking out the ligands from the final PDB file followed by additional refinement with refmac (19) to remove remaining phase bias. The structure shown is the final PDB file. In supplemental Fig. 1, a 2F_o − F_c OMIT map, calculated according to Bhat's protocol (35), is shown.

Activity Measurements

Agonistic and antagonistic profiles of compounds on PR, GR, and androgen receptor were determined in cellular assays using protocols similar to those published earlier for PR (21–23). For agonistic profiles we measured nuclear receptor-induced activation of a murine mammary tumor virus-luciferase reporter in Chinese hamster ovary cells (22). These Chinese hamster ovary cells were stably transfected with the reporter gene and the receptor of interest. For antagonistic profiles, we measured the decrease of activation, induced by a reference compound, by competition with a compound of interest (21). Reference compounds were dihydrotestosterone for androgen receptor, Org 2058 for PR (23), and dexamethasone for GR. Assays were run in a standardized fashion. The results are the average of a duplicate test in two different experiments (Table 1).

Binding Data

Binding Data were taken from the literature (Tables 3 and 4). Usually RU486 data are reported as EC₅₀ values or relative affinities in radioactive competition assays. Because different reference ligands are used, these data are hard to compare. To alleviate this problem we converted EC_{50 RU486} to K_i,RU486 with the Cheng-Prusoff equation (24) as follows: K_i,RU486 = EC_{50 RU486}/(1 + [ref]/K_d,ref). Here [ref] is the reference ligand concentration, and K_d,ref is the reference ligand dissociation constant. Wherever possible, all input values were taken from the original assay data source (Table 4).

TABLE 3.

Binding constants of RU486 and asoprisnil for PR

In the LBD, rabbit and human PR have identical sequences. Data are from Ref. 8.

	RU486	Asoprisnil
	nm	nm
Rabbit uterine PR (Ki)	0.82 ± 0.01	0.85 ± 0.01

TABLE 4.

Binding constants of RU486 for PR and GR

Data were taken from literature. Unless otherwise indicated, all data in a row originate from the reference in the first column. ruPR, rabbit uterine PR; hPR, human PR; rtGR, rabbit thymic GR; hGR, human GR. PR-A and PR-B are isoforms, prog, progesterone; dex, dexamethasone; nr, not reported.

Nuclear receptor(cell line, reference)	RU486 data EC50	Reference ligand(identity, nm)	Reference liganda(Kd, nm)	RU486b (Km, nm)
PR
ruPR (8)	nr	R5020 (5)	2.0	0.82 ± 0.01
hPR-A (9)	10.6 nm ± 1.3	prog (6.8)	1.2	1.6 ± 0.4
hPR-B (9)	9.5 nm ± 0.9	prog (6.8)	0.9	1.1 ± 0.2
ruPR (9)	11.5 nm ± 0.9	prog (6.8)	4.8	4.8 ± 0.5
ruPR (10)	3.0 nmc	R5020 (0.4)	2.0d	2.5c
hPR (T47D, Ref. (10))	1.0 nmc	R5020 (0.4)	2.0d	0.83c
hPR LBD (36)	nr	prog analog (nr)	10	10.0 ± 0.3
hPR LBD (37)	307 nmc	prog analog (nr)	10	40c
GR
rtGR (9)	9.1 nm ± 0.8	dex (6)	2.2	2.4 ± 3.2
hGR (IM9, Ref. (10)	2.0 nmc	dex (0.4)	2.2e	1.7c
hGR LBD (37)	100 nmc	dex analog (1)	0.18	15c

^a As reported by a literature source or calculated from the EC₅₀ of a reference ligand in the same assay through K_d,ref = EC_{50 ref} − [ref] (24).

^b Converted from EC₅₀ using K_i,RU486 = EC_{50 RU486}/(1 + [ref]/K_d,ref). The average K_i of RU486 for full-length PR (not LBD only) is 1.9 nm ± 1.4.

^c No errors reported.

^d Taken from Pullen et al. (8), as Palmer et al. (10) report 0.05 nm, which seems very low.

^e Taken from Attardi et al. (9).

Conformer Calculation

16 RU486 conformers were generated with the package “Catalyst” (Accelrys©). These were minimized at the Hartree-Fock level with basis set 3–21G with the program Gaussian03 (Gaussian Inc., 2004). The resulting five different conformations were subjected to a rigid body fit into the electron density of the GR and PR RU486 complexes.

RESULTS AND DISCUSSION

Structure of the Norethindrone-bound Complex

To compare structures properly, it is very important that they are determined using the same crystallographic protocols. Therefore, we first re-determined the structure of the PR LBD coexpressed with the agonist norethindrone to 1.55 Å. This structure is essentially identical to the published PR-norethindrone complex 1SQN (25), with a root mean square deviation (C_α) of 0.118 Å. The asymmetric unit contains two copies of the PR-LBD, and in both copies norethindrone is present. However, the protein conformation of both copies is not identical (25), as differences occur in the regions comprising residues 703–712, 785–808, and 893–923, which are located N-terminal from helices 3, 6, and 12, respectively (for their location in the LBD, see Fig. 2A).

FIGURE 2.

Binding of RU486 (purple) in the PR LBD (blue/red). A, ribbon view of the PR LBD copy in the dimer that binds RU486. The coloring represents B-factor changes compared with the equiconformational LBD copy in the PR norethindrone complex. Changes range from −4 (blue) to +44 (red) Ų and are predominantly restricted to the loop 785–808 and helix 12. The conformation of helix 12 is agonistic, closely packed against the LBD core. B, superposition of the two PR LBD copies in the dimer, one containing RU486 (coloring as in panel A) and one containing norethindrone (gray). Important residues and loops are marked.

The PR Ligand Exchange Protocol Allows Visualization of New Compound Classes

Using RU486 in the same manner as norethindrone in coexpression and purification did not generate crystals but led to precipitation of the receptor. Therefore, we resorted to ligand replacement by soaking norethindrone-containing crystals.

In general, soaking protocols are a good way to obtain atomic information on compounds that are scarce or destabilize the protein, such as RU486. However, they have not been widely applied to nuclear receptors because the presence of high affinity ligands is generally required for LBD stability during purification (26), and such ligands exchange only slowly. Promising experiments for mutant estrogen receptor crystals have recently been described (27), but the most advanced protocol with PR thus far involves exchange of ligands during purification (26), which has not worked for RU486 so far.

To circumvent problems with slow exchange, we soaked our crystals in RU486 for a long time (10 days). This resulted in clear replacement of norethindrone by RU486 in one copy of the crystal dimer. This is evidenced by the appearance of electron density for the 11β substitution (Fig. 1).

Subsequent refinement of the ligand with full occupancy and loose B-factor restraints results in an average B-factor for the 11β-substituent of only 3.5 Ų (2.5 S.D.) above that of the average of the 17-atom steroid scaffold. This is comparable with the B-factor of the ketone at the 3-position, and therefore, this refinement test indicates a high level of exchange. Thus, ligands can be replaced in PR crystals if long soaking times are used.

A drawback of ligand exchange is that the crystal lattice constrains the conformation of the protein. When a compound lacks affinity for the restrained receptor conformation, ligand exchange does not take place, or the crystals deteriorate because of induced fit. Both are frequent results in our hands (data not shown). However, when we do see exchange, as for RU486, the ligand has affinity for the restrained protein conformation, and therefore, the observed binding mode has thermodynamic relevance.

The Flexible Loop 785–808 Could Be a Ligand Entry Route

Interestingly, compound exchange of RU486 takes place in only 1 copy of the dimeric PR LBD. In the other copy, norethindrone remains bound. This selective affinity is also observed when soaking with other compounds (data not shown). However, sometimes exchange is observed in both copies, indicating that the second PR LBD copy is accessible in the crystal and that it must have lower steroid affinity or a slower rate of ligand entry. This difference must arise from conformational differences in the PR LBDs, which are most prominent around the loop 785–808. This loop is also a region of substantial conformational variability (∼1.0 Å) when our structure is compared with the PR-norethindrone, PR-asoprisnil, and the GR-RU486 complexes (see below). A C798V mutation in this loop abrogates all progesterone binding activity of PR (28) even though Cys-798 is not in direct contact with the ligand. Interestingly, the same region was recently suggested to allow ligand entry in estrogen receptor α (27). Therefore, we propose that also in PR, the flexible loop 785–808 is involved in ligand entry.

The Binding Mode of RU486 Explains Anti-progestin Resistance

The interactions of the steroid rings of RU486 are similar to those of other steroids such as norethindrone (Ref. 11 and Fig. 2) and in line with mutagenesis results (28). The ketone of the A-ring anchors to Arg-766 and Gln-725. The 17β-hydroxyl on the D-ring forms a water-mediated hydrogen bond to Asn-719 and is oriented toward Thr-894 at 4.5 Å distance. The 17α-propynyl of RU486 is accommodated in a cavity formed by Leu-715, Leu-718, Phe-794, Leu-797, Met-801, and Tyr-890 (Ref. 11 and Figs. 2 and 3).

FIGURE 3.

Superposition of the PR LBDs bound to asoprisnil and to RU486. The asoprisnil complex (copper) was obtained from the PDB (code 2OVH, Ref. 11). The RU486 complex is shown in blue. The corepressor SMRT in this asoprisnil complex is colored green. A, overview of the total LBD; B and C, detailed views of the ligand binding pocket.

Additional interactions are formed by the 11β substitution of RU486. This binds into the hydrophobic pocket lined by Gly-722, Trp-755, Met-759, and Met-909 (Figs. 2 and 3), which is also the 11β-substituent pocket in other nuclear receptors (5, 11).

Interestingly, the terminal dimethylamine is not coplanar with the phenyl moiety of the 11β substitution. This strained conformation is supported by the electron density (Fig. 1). To see if it also occurs in small molecule crystal structures, we analyzed 1347 N,N-dimethylanilines without ortho substituents in the Cambridge Structural Data base (29). These show a median angle of 8.8° between the plane through the phenyl ring and the plane through the dimethylamino group, revealing that the skew is a property of dimethylanilines.

The terminal dimethylamine engages in a π-π interaction with the main chain carbonyl group of Gly-722 (3.9 Å). This Gly-722 is a close neighbor of Gln-725 that anchors the steroid A-ring, and together they form a loop that clasps one side of the antiprogestin.

In the past a G722C variation in the PR of chicken and hamsters was found to cause resistance to RU486 (30). From modeling, it was concluded that a Cys-722 side chain could potentially block RU486 binding (16). This is confirmed in our structure. If modeled, the Cβ of Cys-722 would be at 2.0 Å distance from the phenyl moiety in RU486, leading to incompatible steric clashes.

RU486 Binds to an Agonist-like Conformation of the LBD

When the RU486-bound PR LBD is superimposed on a norethindrone-bound PR LBD (we used the equivalent copy in the unsoaked structure), this shows that helix 12 has a similar orientation (Fig. 2). Therefore, RU486 binds to an agonistic conformation of the PR LBD. Clearly, this is caused by the restraints of the crystal contacts, which were formed when the crystal was grown in presence of norethindrone.

It is surprising that RU486 binds to the “agonistic” PR-LBD conformation at all, as in the past it was predicted that the dimethylaniline would clash with Met-909 in helix 12 and Trp-755 (16). In our structure we only observe a more subtle steric conflict. In comparison to the norethindrone complex, the side chain of Met-909 becomes disordered, and the neighboring Ile-913 assumes a different rotamer (Fig. 1). There is only incomplete electron density for the side chain of Met-909, and any rotamer we model remains sterically uncomfortably distanced from the terminal dimethylamine. In reality we probably observe the average of an ensemble of subtly different dimethylaniline and Met-909 orientations, each sterically allowed.

Other differences were observed compared with the norethindrone complex in the two flexible loops 785–808 and 703–712. After RU486 binding, these have shifted 0.5 and 0.3 Å, respectively, inducing a (partially occupied) peptide bond flip of Ser-792. These shifts widen the hormone binding pocket. It appears as if RU486 wrenches itself in the LBD, between Phe-794 in the loop 785–909 and Met-909 in helix 12 (Fig. 2). Thus, although RU486 is bound with helix 12 in the agonist position, the LBD is clearly changed.

RU486 Binding Induces Higher B-factors in Helix 12

The structures of the PR-LBD bound to norethindrone and RU486 were determined using the same crystallographic protocols, and because they have the same crystal contacts, we can attribute any differences in main-chain B-factors to ligand binding. B-factor differences between the individual structures were calculated as B_i,RU486 − B_{i,norethindrone} for every atom i for the equivalent PR copies. When plotted, these differences indicate substantial effects, particularly in the region of helix 12 (Fig. 2A and supplemental Fig. 1). Thus, even though helix 12 has an “agonist-like” orientation, RU486 clearly increases its flexibility.

The Mechanism of Helix 12 Destabilization

It is interesting how the single steric clash between Met-909 and RU486 can lead to destabilization of the entire helix 12 as measured by B-factors. This can be explained because Met-909 on the helix interface protrudes deeply into the LBD core and represents a hydrophobic hook that can be dissociated by RU486. In addition, the main chain N atom of Met-909 is located at the N-terminal side of helix 12, where it is capped by Glu-723 (Fig. 2B), which functions as the residue that stabilizes the helix 12 dipole (31). A small shift of Met-909 will, therefore, lead to disruption of this dipole system, leading to destabilization of the entire helix.

Differences in the Binding Mode to PR of RU486 and Asoprisnil

Recently, the three-dimensional structure of the PR LBD was reported, bound to the antagonist asoprisnil, and a SMRT peptide (PDB code 2OVH, Ref. 11) (Fig. 3). For unknown reasons, asoprisnil and SMRT stabilize the complex sufficiently to allow cocrystallization experiments, whereas in our hands, RU486, even in combination with peptides, yields an unstable complex. Nevertheless, RU486 and asoprisnil bind to PR with similar affinities (Table 3).

In the asoprisnil structure helix 12 is rearranged (Ref. 11 and Fig. 3) to make room for binding of the SMRT peptide. Thereby, the asoprisnil complex represents a PR structure in full antagonistic mode (5, 11).

Apart from the helix 12 shift, our PR-RU486 and the PR-asoprisnil structures overlap well, although again the loops 703–712 and 785–808 show conformational flexibility. RU486 and asoprisnil have a similar binding orientation, but asoprisnil is positioned 0.8 Å “upwards” toward Met-909 in helix 12, which is now out of the way. At the bottom of the pocket, Phe-794 (1.1 Å) and Leu-797 (0.8 Å) are upwardly shifted to match the 17α-substituent of asoprisnil. The Phe-794 and Leu-797 side chains of the asoprisnil structure would need to shift downward to accommodate the 17α-substituent of RU486. Therefore, compared with RU486, asoprisnil appears squeezed out of the LBD, toward helix 12 (Fig. 3). This demonstrates that the loop 785–808 and the 17α-substituent also contribute to helix 12 displacement and, thereby, to antiprogestin activity (Fig. 4).

FIGURE 4.

The mechanism of antagonistic action in PR. The circle outlines the PR LBD pocket, capped by helix 12 (rectangle). Top left, when progesterone is bound, the LBD is in a tightened agonist conformation (model after PDB structure 1A28). Top right, when an antagonist such as RU486 displaces progesterone, the LBD pocket widens by pressure on Phe-794 and Met-909 (model after our structure). Low right, this pressure can be released by helix 12 displacement followed by return of Phe-794 and the loop 785–808 to a more relaxed conformation. This conformation is further stabilized by corepressor binding (model after the PDB structure 2OVH).

Our RU486 Complex Is an Intermediate in Antagonist Action

Using the above analyses, we can place our structure in the route of antiprogestin action (Fig. 4). Normally, before treatment, progesterone is bound to the PR-LBD. This is represented by the structure with PDB code 1A28. Upon treatment, progesterone is replaced by RU486, and helix 12 is destabilized. Its new position is a dynamic equilibrium that includes the agonist position (as in our structure) and the antagonist position (as in the asoprisnil complex). When corepressor proteins bind, the equilibrium position of helix 12 is further driven toward the antagonist position (Fig. 4). In short, the RU486 complex represents an intermediary stage in antagonist action and provides a unique view on the mechanism of antihormones, implicating new residues throughout the LBD.

RU486 Binds with Similar Affinity to PR and GR

To compare the binding affinities of RU486 for PR and GR, we converted historic binding EC₅₀ values to K_i (Table 4). This indicates that RU486 binds to native PR with an average K_i of 1.9 nm and that the K_i for GR is similar at about 2 nm. The K_i for the separate LBDs (15–40 nm) is higher than for full-length receptors but still substantial (Table 4).

Next we superimposed our PR-RU486 complex on the published GR-RU486 complex (PDB code 1NHZ, Ref. 12). This latter structure again shows a shifted helix 12 conformation, although in the GR-RU486 structure, the space freed up by the helix 12 shift is filled up by helix 12 from another GR monomer, and thus, care should be taken in interpreting this structure as a full antagonistic conformation (12). Nevertheless, the comparison still yields meaningful insights.

RU486 Binds with a Different Conformation to PR and GR

Interestingly, the steroid A ring of RU486 has a different puckering conformation in both complexes (Fig. 5). To confirm this, we calculated the experimental electron densities belonging to both complexes and show that the RU486 conformation in the 1NHZ structure does not fit our density and vice versa (Fig. 5).

FIGURE 5.

2F_o − F_c electron density contour (1.2σ) for our PR-RU486 complex (2W8Y, panel A) and the GR-RU486 complex (1NHZ, panel B, diffraction data obtained from the PDB). Ligands modeled in the original x-ray structures are in gray. The cyan conformation is the low energy inverted half-chair conformation (1β, 2α) that fits the GR density. The dark blue conformation shows the normal half-chair conformation (1α, 2β) that fits the PR density and which has somewhat higher energy in vacuo. The ab initio ligands (cyan/blue) were rigid body fitted into the density using the program Coot.

To check the physical reality of the different conformations, we calculated RU486 conformers ab initio. The lowest energy conformation is similar to that observed in the GR electron density (12). A higher energy conformation, which matches Cambridge Structural Data entry FAFGOI (32) (+1.7 kcal/mol in vacuo), fits in the electron density of our PR structure (Fig. 5). Similar disorder in the A ring of 3-keto-Δ⁴ steroids has been observed before in the x-ray structures of 3-keto-desogestrel (33) and nandrolone (34). We conclude that in PR, RU486 binds in a somewhat strained conformation. Because this conformation is also seen in the PR-asoprisnil complex, it is probably not induced by the agonistic helix 12 position but by differences in the binding pockets of GR and PR.

RU486 Has Different Binding Interactions in PR and GR

The PR-RU486 and the GR-RU486 complexes differ in their conformation of helix 12 and to a lesser extent in the conformations of the loops of residues 785–808 and 703–712 (548–557 and 630–653 in GR) (Fig. 6). Moreover, in GR the 11β-dimethylaniline of RU486 has shifted 1.4 Å and occupies space freed up by helix 12 rearrangement (Fig. 6). In this “GR” orientation, the phenyl group would stereochemically clash at 2.4 Å with Asn-719 in the PR conformation. However, in GR, the equivalent Asn-564 has assumed a different rotamer (made possible by the helix-12 shift), which still binds the ligand 17β-hydroxyl group but now via two water molecules instead of one (Fig. 6).

FIGURE 6.

Superposition of RU486 bound to GR (cyan) and PR (blue). The GR structure was obtained from the PDB (code 1NHZ, Ref. 12). A, overview of the total LBD. B, detail of the ligand binding pocket. The shifted steroid ring orientation and the different puckering in the steroid A-ring can be clearly observed.

In GR, the 17α-propynyl group of RU486 delves 1.6 Å deeper into its subpocket compared with PR. This is because of the presence of the flexible Met-639 in GR instead of the bulky Phe-794 as in PR. This orientation is stabilized in GR by Gln-642 forming a 2.7 Å hydrogen bond to the 17β-hydroxyl of RU486. This hydrogen bond is strengthened by the burial between Met-560, Met-639, and Tyr-735. In PR, Gln-642 is replaced by Leu-797, which is unable to form hydrogen bonds to RU486 (Fig. 6). Therefore, although RU486 binds to both PR and GR with similar affinity, there are clear differences in its binding interactions.

Specificity Differences in RU486-like Compounds

To compare the activities of RU486 and related compounds, we determined their activities in our own set of cellular assays (Table 1). Binding data for many of these compounds were reported earlier (Ref. 4 and Table 4), but here our cellular assays allow discrimination between agonistic and antagonistic activities.

Our results confirm that RU486 has full antagonistic activity in PR and androgen receptor, but interestingly in GR, RU486 is a partial antagonist (with 10% agonistic activity). On the atomic level, this difference might be caused by RU486 binding more deeply into the 17α-propynyl binding pocket in GR. This might ensure fewer clashes between the 11β substitution of RU486 and helix 12, at the other end of the LBD, thereby facilitating binding of helix 12 in its agonist conformation, leading to a more partial compound profile on GR.

Structures Help to Understand Different Profiles of RU486-related Compounds

Our activity data show that asoprisnil and onapristone are more specific for PR than RU486, confirming earlier results (Ref. 7, Table 1). Underlying this increased specificity are the chemical differences between the compounds, which are mostly located around the 11 and 17 positions of the steroid ring (Table 1).

As for the 11β substitution, our structure and biological data (6, 14) suggest that it does not solely disrupt helix 12 but also provides binding energy. Replacing the 11β-terminal dimethylamine by an oxime leads to a 3-fold stronger binding to PR (6). Probably, through its increased coplanarity with the phenyl group, the oxime can stack better to Gly-722. Alternatively, the oxime might fit better in the solvent-exposed environment after helix 12 displacement. However, as binding interactions are essentially similar in PR and GR, it is unlikely that the 11β substituents in Table 1 confer selectivity.

At the 17-position, introduction of an alkyl ether in asoprisnil or alcohol in onapristone reduces GR activity (refs 6, 7, Table 1). Both the 17β-methoxy of asoprisnil and the 17β-hydroxypropyl of onapristone can be accommodated in PR but clash with Gln-642 in GR. In this way differences at the 17-position may play a role in the selectivity of these antiprogestins.

Conclusion

Our PR-RU486 structure resembles an intermediate in antiprogestin action and gives new insight into how RU486 acts on PR. Our structure suggests that RU486 does not induce one particular receptor conformation but changes the equilibrium of the helix 12 position. Importantly, the whole LBD cooperates in helix 12 displacement through the concerted action of various loops and residues. Apart from RU486, many more drugs are 11β-substituted steroids or related antihormones (5, 13). Through this structure, the molecular mechanism of this whole drug class can be better understood.