Structural Evidence for Ligand Specificity in the Binding Domain of the Human Androgen Receptor IMPLICATIONS FOR PATHOGENIC GENE MUTATIONS*

The crystal structures of the human androgen receptor (hAR) and human progesterone receptor ligand-binding domains in complex with the same ligand metribolone (R1881) have been determined. Both three-dimensional structures show the typical nuclear receptor fold. The change of two residues in the ligand-binding pocket between the human progesterone receptor and hAR is most likely the source for the specificity of R1881 to the hAR. The structural implications of the 14 known mutations in the ligand-binding pocket of the hAR ligand-binding domains associated with either prostate cancer or the partial or complete androgen receptor insensitivity syndrome were analyzed. The effects of most of these mutants could be explained on the basis of the crystal structure.

The crystal structures of the human androgen receptor (hAR) and human progesterone receptor ligandbinding domains in complex with the same ligand metribolone (R1881) have been determined. Both threedimensional structures show the typical nuclear receptor fold. The change of two residues in the ligandbinding pocket between the human progesterone receptor and hAR is most likely the source for the specificity of R1881 to the hAR. The structural implications of the 14 known mutations in the ligand-binding pocket of the hAR ligand-binding domains associated with either prostate cancer or the partial or complete androgen receptor insensitivity syndrome were analyzed. The effects of most of these mutants could be explained on the basis of the crystal structure.
Androgen (AR) 1 and progesterone receptors (PR) are members of the superfamily of nuclear receptors that includes the steroid receptors, among others, as well as the vitamin D, thyroid, retinoic acid receptors, and the so-called orphan receptors. In addition, AR and PR are members of a group of four closely related steroid receptors including the mineralocorticoid receptor and the glucocorticoid receptor recognizing the same hormone response element. In general, steroid receptors are comprised of five to six domains and act as ligand-activated transcription factors that control the expression of specific genes. To date, no experimentally determined three-dimensional structure is available for a complete receptor. During the past few years, x-ray structures have been published for two of the domains, the DNA-binding domain as well as for a number of ligand-binding domains (LBD) including LBD⅐ligand complexes of the estrogen receptor ␣ and ␤, the PR, the vitamin D receptor, the retinoic acid receptors (X,RXR; acid, RAR), the thyroid hormone receptor, and the peroxisome proliferatoractivated receptors (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)(13). Despite the low sequence homology of as low as 20% between the LBDs of different nuclear receptor families, all these proteins share a similar fold. They are comprised of up to 12 helices and a small ␤-sheet arranged in a so-called ␣-helical sandwich, a kind of fold that up to now has only been observed for the LBDs of nuclear receptors. Depending on the nature of the bound ligand, agonist, or antagonist, the carboxyl-terminal helix H12 is found in either one of two orientations. In the agonist-bound conformation, helix H12 serves as a "lid" to close the ligand-binding pocket (LBP), whereas in the antagonist-bound conformation helix H12 is positioned in a different orientation thus opening the entrance to the LBP.
Androgens and their receptors play an important role in male physiology and pathology. AR binds the male sex steroids, dihydrotestosterone (DHT) and testosterone (14), and regulates genes for male differentiation and development. Therefore, mutations in the androgen receptor gene may lead to several disease states like prostate cancer (PC) or the androgen insensitivity syndrome (AIS). In males, defects in the AR gene result in a spectrum of developmental abnormalities ranging from a phenotypic female to varying degrees of incomplete genital phenotype. These mutations are well documented in the Androgen Receptor Gene Mutations Data Base of the Lady Davis Institute for Medical Research (15).
In this study we present the crystal structure of the human hAR LBD in complex with the ligand metribolone (R1881) in comparison with the crystal structure of the human hPR LBD in complex with the same ligand. AR and PR belong to the same steroid receptor subfamily and share a 54% LBD sequence identity (Fig. 1). A number of different ligands bind with similar binding affinities to both receptors (14). The x-ray structure analysis of both receptors in complex with the same ligand (R1881, Fig. 2) should increase our understanding of ligand specificities. Furthermore, the analysis of published mutant data on the basis of the hAR LBD crystal structure might give us a deeper insight into AR-related diseases.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-The cDNAs coding for hAR and hPR were obtained from the groups of A. Cato (Forschungszentrum Karlsruhe, Germany) and P. Chambon (Institut de Génétique et de Biologie Moléculaire  the hAR (amino acid residues 663-919) and the hPR (amino acid residues 677-933) were amplified by the polymerase chain reaction technology using the appropriate primers and cloned into a pGEX-KG vector (16). The resulting fusion proteins consisted of a glutathione S-transferase, containing a carboxyl-terminal thrombin cleavage site, optimized by a glycine-rich "kinker" region followed by the corresponding LBD. The constructs were then transformed into the Escherichia coli strain BL21 (DE3).
Protein Expression and Purification-Fermentation using the corresponding rec E. coli strains expressing hAR LBD was carried out in 2X YT medium in the presence of ampicillin (200 g/ml) supplemented with 10 M R1881. Expression was induced with 30 M isopropyl-␤-Dthiogalactoside, and the fermentation (10 liters) was continued at 15°C for 14 -16 h. Cells were harvested by centrifugation and disrupted twice in a continuous high pressure homogenizer (9000PSI) in a buffer containing 50 mM Tris-HCl, pH 8, 150 mM NaCl, 5 mM EDTA, 10% glycerol, 100 M R1881, 100 M phenylmethylsulfonyl fluoride, and 10 mM DTT. All buffers were purged with nitrogen before adding DTT. The supernatants from ultracentrifugation were loaded onto a glutathione-Sepharose column, washed with 50 mM Tris-HCl, pH 8, 150 mM NaCl, 5 mM EDTA, 10% glycerol, 10 M R1881, 0.1% n-octyl-␤-glucoside, and 1 mM DTT, and the fusion protein was eluted using the same buffer supplemented with 15 mM reduced glutathione. The eluate was diluted with 100 mM HEPES, pH 7.2, 150 mM NaCl, 0.5 mM EDTA, 10% glycerol, 10 M R1881, 1 mM DTT, and 0.1% n-octyl-␤-glucoside up to a fused protein concentration of 1 mg/ml. A thrombin cleavage (2 NIH units/mg fusion protein) was performed overnight at 4°C. The protein mixture was further diluted 3-fold with 10 mM HEPES, pH 7.2, 10% glycerol, 10 nM R1881, 10 mM DTT, and 0.1% n-octyl-␤-glucoside and loaded onto a Fractogel SO 3 Ϫ column and eluted with a gradient of 50 -500 mM NaCl in a 10 mM HEPES buffer, pH 7.2, 10% glycerol supplemented with 10 nM R1881, 10 mM DTT, and 0.1% n-octyl-␤glucoside. Approximately 2.4 mg of purified hAR LBD were recovered from 1 liter of E. coli cell cultures. Protein concentration was determined with Bio-Rad Protein Assay. Fermentation and purification of the hPR LBD was performed identically, but a HEPES pH 7.3 buffer was used from the beginning.
Comparative Modeling-A model of the hAR LBD was built based on the coordinates of the hPR LBD-progesterone complex (molecule A) (9). Amino acid substitutions were made based on the sequence alignment in Fig. 1 using the Insight 98.0 software (MSI Inc., San Diego). Soaking of the initial model and the energy minimization protocols applied are described in detail elsewhere (17).
Crystallization, Data Collection, and Structure Determination-Both proteins were dialyzed after purification with buffer containing 50 mM HEPES pH 7.2 for hAR LBD or 10 mM HEPES pH 7.2 for hPR LBD, respectively, 10% glycerol, 10 mM DTT, 0.1% n-octyl-␤-glucoside, 10 mM R1881, and 150 mM Li 2 SO 4 and were concentrated up to 3 mg/ml for the hPR LBD-R1881 and up to 4.4 mg/ml for the hAR LBD-R1881, respectively. Vapor diffusion method was used at 20°C for the hAR LBD complex and at 4°C for hPR LBD complex. Crystallization experiments had to be set up immediately after concentration. For the hAR LBD⅐R1881 complex, the reservoir solution contained 0. at beam line BM14 at the ESRF (Grenoble, France) to a resolution of 2.8 Å. Before data collection was complete the crystal decomposed in the x-ray beam.
Both data sets were integrated and reduced using DENZO and SCALEPACK (18). Statistics of x-ray data collection and processing are summarized in Table I.
Contrary to the hPR LBD⅐progesterone complex that crystallizes with one homodimer in the monoclinic space group P2 1 the hAR LBD crystallizes with one monomer in the orthorhombic space group P2 1 2 1 2 1 . The hPR LBD⅐R1881 complex crystallizes in the same monoclinic space group P2 1 and with similar cell constants as the hPR LBD⅐progesterone complex. The structure determination for the hAR LBD⅐R1881 and hPR LBD⅐R1881 complexes were carried out using the molecular replacement method in AMoRe (19) with the coordinates of the hPR LBD⅐progesterone complex (Protein Data Bank entry, 1A28 (9)). Clear solutions were obtained for both structures using data between 15.0 and 3.5 Å for the hAR LBD and 12.0 and 3.5 Å for the hPR LBD, respectively.
Refinement of hAR LBD⅐R1881 Complex-The molecular replacement solution obtained was refined using X-PLOR (20). In all refinements and map calculations with X-PLOR a bulk solvent correction was used, and all low resolution data were included. Prior to the refinement calculations, a random 5% sample of the reflection data was flagged for R-free calculations (21). All model interactive visualization and editing was carried out using TURBO (22). Refinement started using data up to 3.5 Å, and resolution was gradually extended to 2.4 Å. The model was edited according to the known hAR LBD sequence (23) using 2 F o Ϫ F c and F o Ϫ F c maps calculated at 3.2 Å resolution and simulated annealed omit maps. The fast wARP (24, 25) molecular replacement protocol was also applied after each XPLOR refinement to improve further the 2 F o Ϫ F c electron density map. Prior to its inclusion in the model, the electron density for the R1881 ligand was clearly visible in all maps. A model for the ligand was obtained from the Cambridge Structural Data Base entry HMESTR (26,27). The XPLOR topology and parameter dictionaries were built using program XPLO2D (28). In the final refinement at 2.4 Å, 26 water molecules were included in the model, and individual restrained B-factors were refined for all nonhydrogen atoms. The final values of R and R-free were 21.0 and 29.7%, respectively. The R-free/R ratio is only slightly smaller than expected (29) for the number of atoms and reflections used in the refinement. The final refinement results and statistics are shown in Table II.
Refinement of hPR LBD⅐R1881 Complex-The molecular replacement solution obtained was refined using REFMAC (30) using the maximum likelihood approach. Bulk solvent scaling of F o and F c was applied based on Tronrud's solvent correction, and all available data with no cutoffs were used. All map calculations were done including calculated F values for missing reflections. To avoid model bias, calculated maps using only F o were checked. After the first refinement step the SigmaA-weighted calculated 2 F o Ϫ F c and F o Ϫ F c maps were inspected using the program O (31), and electron density of the ligand was clearly observed. The ligand was built up in SYBYL 6.5 (Tripos Inc.) and was included in further refinement steps. A dictionary file for distance restraints for the R1881 molecule was prepared using MAKE-DICT (32). Towards the end of the refinement, only one water molecule in the LBP of molecule A was added due to the low resolution and missing data. The final model comprises 4027 protein atoms, 42 ligand atoms, and 1 water molecule with final R values of r ϭ 21.7% and R-free ϭ 34.3%, respectively. A summary of the refinement and model statistics is included in Table II.
Data Deposition-Coordinates have been deposited with the Protein Data Bank (33,34). Protein Data Bank accession codes are 1e3g for hAR LBD-R1881 and 1e3k for hPR LBD-R1881.

RESULTS AND DISCUSSION
Protein Expression and Purification-Glutathione S-transferase fusion proteins can be expressed to very high levels in the E. coli strain BL21 (DE3) (16). This system was used successfully for the production of the hPR LBD (9) and hAR LBD. The expression of soluble fusion proteins strongly depends on the presence of ligand in the cells during fermentation (data not shown). During cell disruption, purification, and concentration, any protein oxidation was avoided by purging all buffers with nitrogen and by using DTT as an antioxidant. Fusion proteins were purified by the use of glutathione-Sepharose and subsequently cleaved with thrombin. Cation exchange chromatography yielded purified LBDs. Concentration was performed with a nitrogen pressure diafiltration system.
Comparison of Model and Crystal hAR LBD Structures-The model and the crystal structures of the hAR LBD are very similar with respect to their overall structure, the LBP and the ligand orientation. The root mean square (r.m.s.) deviation between 149 equivalent C ␣ atoms in helices between the model and crystal structure of the hAR LBD is 1.09 Å. It is comparable to the r.m.s. deviation of 0.84 Å between the crystal structures of the hAR LBD and the hPR LBD⅐progesterone complex. The most striking difference between the model and the crystal structure was found in helix H6, where an ␣-helix was identified in the crystal structure in this region, whereas in the hPR LBD⅐progesterone complex (molecule A) no ␣-helix is observed. The ligand orientation in both the hAR LBD-R1881 model and crystal structure is very similar, and the same hydrogen bonds are found.   (35), and their stereochemical quality parameters were within their respective confidence intervals. In the Ramachandran (36) , plot, 87.7% for the hAR LBD-R1881 and 85% for the hPR LBD-R1881 structures, respectively, lie within the most favored regions. For the hAR LBD⅐R1881 complex no residue is outside the normally allowed regions, whereas in the hPR LBD⅐R1881 complex two residues are located in disallowed regions (Asn 705 and Ser 793 in molecule A), and three residues (Thr 796 in molecule A and Asn 705 and Ser 793 in molecule B) are located in generously allowed regions. These residues are not involved in ligand binding and are located in loop regions that are most probably not involved in ligand recognition. In the hAR LBD-R1881 structure there is only one close contact (2.6 Å) between Met 895 and Ala 896 carbonyl oxygens. In the hPR LBD-R1881 structure some close contacts were observed, but due to the resolution and completeness of the data this is not surprising. The overall fold of the hAR and hPR LBD-R1881 structures is very similar and also with that of hPR LBD complexed with progesterone (9). On the basis of the secondary structure calculated with PROCHECK (35) according to Kabsch and Sander (37), the hAR LBD-R1881 structure contains 9 ␣-helices, two 3 10 helices, and four short ␤-strands associated in two anti-parallel ␤-sheets. The helices are arranged in the typical "helical sandwich" pattern as in hPR LBD⅐progesterone complex (9), and helices H4, H5, H10, and H11 are contiguous. In hAR LBD-R1881 helix H12 seems to be split into two shorter helical segments, with nine and five residues each, respectively. This observation was not seen in the hPR LBD-R1881 structure, although a bending of helix H12 is also seen here. Fig. 1 shows a comparison between the amino acid sequences of hAR LBD and hPR LBD. A ribbon diagram of the hAR LBD-R1881 structure is shown in Fig. 3 along with a superimposed C ␣ trace of the hAR LBD-R1881 and hPR LBD-R1881 molecules.
The crystal structure coordinates of hAR LBD-R1881 were superimposed with those of hPR LBD-R1881 (molecule A) and hPR LBD-progesterone (molecule A) using LSQKAB (38). For the superposition the main chain atoms except three aminoterminal (Cys 669 -Pro 671 ) and one carboxyl-terminal (Thr 918 ) residues were used. The r.m.s. coordinate deviations were 1.16 and 1.22 Å, respectively, again an indication of the similarity of the overall fold of these three molecules. In hAR LBD-R1881, Cys 669 and Cys 844 are very close, and a disulfide bridge be-tween them was modeled, based on the electron density. However there is no supporting biochemical evidence so far, and it should be noted that the temperature factors of both cysteine residues and the adjacent residues are very high. A cis peptide bond is found at position Pro 849 in hAR LBD-R1881.
Comparative Modeling-The model of the hAR LBD which is based on the hPR LBD⅐progesterone complex is very similar to the hAR LBD crystal structure with respect to the overall fold and ligand orientation but shows a stronger bending of helices H10 and H11. Our model structure differs from other published models (39) with respect to the secondary structure alignment. The secondary structure assignment by Yong et al. (39) as compared with the hAR LBD crystal structure is similar between helices H3 and H10 but differs most for helices H11, H12, and the additional helix at the carboxyl-terminal end.
Ligand-binding Pocket Interactions-There are a total of 18 amino acid residues in hAR LBD and hPR LBD that interact with the bound ligand (either R1881 or progesterone). These residues are highlighted in Fig. 1 and included in Fig. 4. Most of these residues are hydrophobic and interact mainly with the steroid scaffold, whereas a few are polar and may form hydrogen bonds to the polar atoms in the ligand.
The hydrogen-bonding scheme to O-3 of R1881 and progesterone is similar but not identical, as shown in Fig. 4. In the hAR LBD-R1881 crystal structure, this oxygen atom forms a hydrogen bond to Arg 752 (Arg 766 in hPR LBD), but in contrast with the hPR LBD⅐progesterone complex the distance of 3.9 Å to Gln 711 (Gln 725 in hPR LBD) does not allow a hydrogen bond. There is a water molecule near O-3 that is hydrogen-bonded to three other residues with a nearly triangular geometry (Arg 752 N 1 , Met 745 O, and Gln 711 O ⑀1 in hAR LBD; Arg 766 N 1 , Met 759 O, and Gln 725 O ⑀1 in hPR LBD-progesterone). Two of these residues are acceptors; therefore, a third acceptor atom (O-3 in either progesterone or R1881) in a direction perpendicular to the plane of the triangle is unlikely, also due to unfavorable geometry. The water molecule hydrogen-bonded to Gln 711 N ⑀2 in hAR LBD (Gln 725 in hPR LBD) has hydrogen bonds to two other residues (Val 685 O and Phe 764 O in hAR LBD, Ile 699 O and Phe 777 O in hPR LBD), and in hAR LBD it is hydrogen-bonded to a further water molecule, the overall hydrogen bond geometry being distorted tetrahedral. In the hPR LBD-R1881 structure, the ligands in molecules A and B possess slightly different hydrogen bond patterns. In molecule A, O-3 of R1881 forms two hydrogen bonds (3.2 Å to Gln 725 N ⑀2 and 2.9 Å to Arg 766 N 2 ). One water molecule was located in the F o Ϫ F c electron density with the same tetrahedral geometry as observed in the hAR LBD-R1881 structure. In molecule B, the ligand is in a slightly different position, and the hydrogen bond pattern differs from that observed in molecule A. The O-3 of R1881 forms again one hydrogen bond to Arg 766 N 2 with a distance of 2.9 Å, whereas the distance to Gln 725 N ⑀2 is now 3.7 Å, outside the acceptable range for a hydrogen bond.
The 17␤ hydroxyl group of R1881 forms different hydrogen bonds, when bound to hAR LBD or hPR LBD (Fig. 4). In hAR LBD, the 17␤ hydroxyl group is hydrogen-bonded to Asn 705 O ␦1 (2.8 Å) and Thr 877 O ␥ (2.9 Å). The same pattern is observed in molecule B of the hPR LBD⅐R1881 complex where the 17␤ hydroxyl group of R1881 also forms a strong interaction to Asn 719 O ␦1 (2.8 Å), whereas in molecule A the corresponding distance of 3.5 Å is only in the range of a weak interaction. In contrast to the hAR LBD, in both hPR LBD monomers Cys 891 (Thr 877 in hAR LBD) shows only a weak interaction with the 17␤ hydroxyl group of R1881 (3.7 Å in molecule A and 4.0 Å in molecule B). However, the relative orientation of the Cys 891 side chain with regard to the hydroxyl group does suggest that this interaction is relevant to the binding of the ligand.
Structural Basis for Ligand Specificity in hAR LBD-The ligand R1881 binds with a relative binding affinity of 290 to the wild-type hAR as compared with a value of 180 for DHT and 100 for testosterone, respectively (14). As for the wild-type hPR, the relative binding affinity of R1881 is 190 with respect to progesterone (relative binding affinity ϭ 100). Overall, R1881 shows comparable good binding affinities to both receptors, which is also reflected in the orientation of the ligand in the LBPs of the hAR LBD and the hPR LBD (Fig. 4). Thr 894 in hPR LBD is replaced by Leu 880 in hAR LBD, and the C ␦2 atom of this leucine makes a van der Waals contact (3.9 Å) with the oxygen atom of the 17␤ hydroxyl group of R1881. This bulkier side chain, along with the substitution of Cys 891 in hPR LBD by Thr 877 in hAR LBD is very likely responsible for the specific recognition of the 17␤ hydroxyl group of R1881 contrary to the 17␤ acetyl group of progesterone. Not only is there an extra polar residue (Thr 877 besides Asn 705 which is conserved in AR) that can form an additional hydrogen bond to the 17␤ hydroxyl oxygen, but the directed decrease in pocket volume caused by the change of Thr 894 to Leu 880 will very likely inhibit the binding of other bulkier ligands such as progesterone. As previously noted (9), there are no strong hydrogen-bonded interactions between the O-20 carbonyl oxygen atom of progesterone and the protein in hPR LBD indicating that the recognition of this group is probably made only through hydrophobic and steric interactions. The hPR LBD can bind R1881 as well as progesterone, and in the crystal structure, the hPR LBD molecule appears to exhibit two different binding modes for R1881, one resembling that of progesterone (O-3 with two hydrogen bonds to the protein and the 17␤ function weakly interacting with the protein) and one similar to that of hAR LBD (O-3 with only one hydrogen bond to the protein and the 17␤ function also hydrogen-bonded to the protein). However, these binding modes do not seem to imply significant changes in ligand position and orientation within the LBP.
Pathogenic Gene Mutations in the hAR LBD Gene-Constitutional mutations in the AR gene can cause the AIS by impairing androgen-dependent male sexual differentiation to various degrees. Complete AIS (CAIS) leads to an unequivocally external female phenotype; PAIS (partial or incomplete AIS) comprises a wide spectrum of clinical phenotypes; and MAIS, a mild form of AIS, is connected to forms of undervirilization (40).
About 50% of the mutated residues reported in the hAR LBD are found to be involved in PC and in AIS (15). These mutations are distributed all over the LBD but seem to be accumulated in helices H4 and H5, a region involved in ligand binding. Comparison of the solvent accessibility of these mutated residues revealed that a nearly even distribution is found between buried, medium, or fully accessible residues. Table III lists all those mutations in or near the AR LBP which are known to be involved in AIS and prostate cancer, their location with respect to secondary structural elements, as well as the potential effect of the mutations. Mutations are reported for 12 of the 18 residues considered to interact with the ligand R1881 within 4.0 Å as discussed above, as well as two additional residues within 5.0 Å of the ligand (Gly 708 and Val 746 in hAR LBD and Gly 722 and Val 760 in hPR LBD). For most of these mutations, a structural effect can be associated with the substitution. The location of these mutations in the three-dimensional structure of hAR LBD-R1881 is shown in Fig. 5, and it can be seen that the mutations involved in PC cluster mainly near the R1881 17␤ hydroxyl group, whereas those involved in AIS are arranged mainly around the other parts of the ligand. One notable exception is Met 749 which has mutations implicated in both PC and CAIS and is located in the vicinity of R1881 O-3, opposite from the other PC-implicated mutations.
Mutations in the LBP Observed in the Prostate Cancer Cell Line LNCaP-The prostate tumor cell line LNCaP contains an AR receptor showing a significantly increased binding affinity for gestagenic and estrogenic steroids but shows identical R1881 binding (41). A single point mutation (T877A) is associated with this abnormal behavior. With an alanine at this position an important hydrogen bond partner for the 17␤ hydroxyl group in R1881, testosterone, or DHT would be missing, but the other hydrogen bond partner, Asn 705 , involved in ligand binding could still orient the ligand in the LBP. Mutagenesis experiments of hPR emphasized the critical role of this asparagine residue in ligand interaction (17). In the crystal structure of the hPR LBD⅐progesterone complex, Cys 891 is found at the position of Thr 877 , but no hydrogen bond of the 17␤ acetyl group of progesterone was observed, although Cys 891 is relatively  C891V) showed a large decrease in relative binding affinity for progesterone, and the purified mutated hPR LBD was completely inactive in binding assays (17).

Mutations in the LBP Observed in CAIS-
The three mutations in the hAR LBP described for CAIS are substitutions that considerably change the size of the respective amino acid side chains, N705S (40,42), L707R (43), and M749V (40,44). This change in size alters the LBP such that the local structure and interactions to the ligand are disturbed.
In the AR LBD and PR LBD crystal structures, Asn 705 or Asn 719 , respectively, is one of the hydrogen bond partners to the ligand R1881 but not to progesterone. If this residue is substituted to Val in hPR LBD, only a moderate effect was observed on the binding activity of progesterone, considering the K D and half-life values (17). In the crystal structure of the hPR LBD⅐progesterone complex, Asn 719 is involved in the stabilization of the loop between H11 and H12, via hydrogen bond between its N ␦2 atom and Glu 904 O. In the hAR LBD, an identical stabilization is found. An N705S mutation, observed in a patient suffering from CAIS, would have a 2-fold effect, destabilization of the structure and loss of a hydrogen bond partner for the ligand.
In the described hAR mutant L707R, the structure integrity disturbance is also reflected in the binding constants. Considering a van der Waals cutoff distance of 4.0 Å, the side chain of Leu 707 makes close contacts with the A-ring of R1881 as well as five residues in the protein chain as follows: Val 685 , Ala 687 , Gln 711 , Phe 764 , and Leu 768 . These residues are located in a loop region between H2 and H3, within H-3, and in strands S1 and S2. Clearly, such a variation in the size of the side chain would have a large impact, not only in the LBP but in disrupting the overall protein fold itself. The mutated receptor shows undetectable binding affinity to the ligand R1881 as obtained by Scatchard plot analysis, and no transcriptional activity is found (43).
Mutations in the LBP Observed with PAIS/MAIS-Seven described mutations in the hAR LBP are associated with PAIS/ MAIS, and multiple substitutions were observed for amino acids at position 708 (45) and 742 (46). In the hAR LBD crystal structure, a substitution of Gly 708 to alanine should be tolerated, whereas a valine at this position would interfere with ligand binding. The closest distance of the C ␤ atom of an alanine residue to the ligand would be 3.0 Å; however, the C ␥ atoms of a valine would be too close to the ligand atoms (1.5 Å).
In all steroid receptors, the steroid is stabilized by a hydrogen bond between the A-ring of the ligand and an arginine (Arg 752 in hAR). A smaller amino acid residue at this position (mutation to glutamine in hAR) should have a dramatic impact on ligand binding as the stabilization of the A-ring would be severely hampered due to the lack of an electrostatic interaction (47,48). A similar effect has been reported for the hPR receptor where a mutation (R766H) resulted in a low or even non-detectable binding affinity. The side chain of histidine is too small to serve as a hydrogen bond partner to the O-3 atom in progesterone (17).
In the hAR mutation F764S, R1881 shows a similar binding affinity as the wild-type receptor, but a rapid ligand dissociation is observed (49). In the crystal structure, Phe 764 is involved in the stabilization of the A-ring position. A smaller amino acid like serine would allow binding of the ligand but very likely would not contribute to the tight binding of R1881.
Mutations M742V or M742I both dramatically reduce the binding affinity of R1881 (46). Although Ile and Val fit into the LBP, the changed environment is less tightly packed, and the LBP is enlarged, thus affecting the binding of the ligand.
However, not all mutations can be related to a disturbance of the structure. In case of the M787V mutation in the hAR LBD, it was found by Scatchard analysis that R1881 and DHT binding were undetectable or strongly reduced (50). The lack of androgen binding was thought to be the cause for AIS. In the crystal structure, a methionine to valine substitution could be tolerated. The lack of binding affinity found for R1881 may account for a destabilization in the LBP as the Met 787 side chain is in van der Waals contact with other amino acids like Val 760 and Leu 887 as well as ligand atoms.
Of the 20 amino acid residues involved in ligand interaction as discussed above, only from 6 have no mutations been reported so far. In addition, for many of the published mutated receptors no ligand binding data are available. However, the effects of most of the characterized mutants could be explained on the basis of the crystal structure.  (52) and Raster 3D (53). The view is rotated by about 80°c lockwise about a vertical axis with respect to the orientation shown in Fig. 3a.