Dehydroepiandrosterone and Dihydrotestosterone Recognition by Human Estrogenic 17β-Hydroxysteroid Dehydrogenase

Steroid hormones share a very similar structure, but they behave distinctly. We present structures of human estrogenic 17β-hydroxysteroid dehydrogenase (17β-HSD1) complexes with dehydroepiandrosterone (DHEA) and dihydrotestosterone (DHT), providing the first pictures to date of DHEA and DHT bound to a protein. Comparisons of these structures with that of the enzyme complexed with the most potent estrogen, estradiol, revealed the structural basis and general model for sex hormone recognition and discrimination. Although the binding cavity is almost entirely composed of hydrophobic residues that can make only nonspecific interactions, the arrangement of residues is highly complementary to that of the estrogenic substrate. Relatively small changes in the shape of the steroid hormone can significantly affect the binding affinity and specificity. TheK m of estrone is more than 1000-fold lower than that of DHEA and the K m of estradiol is about 10 times lower than that of DHT. The structures suggest that Leu-149 is the primary contributor to the discrimination of C-19 steroids and estrogens by 17β-HSD1. The critical role of Leu-149 has been well confirmed by site-directed mutagenesis experiments, as the Leu-149 → Val variant showed a significantly decreased K m for C-19 steroids while losing discrimination between estrogens and C-19 steroids. The electron density of DHEA also revealed a distortion of its 17-ketone toward a β-oriented form, which approaches the transition-state conformation for DHEA reduction.

structure itself (3). Steroid hormones share a very similar structure that has as its core the cyclo-pentenophenanthrene ring. The addition of an 18th carbon atom results in the estrogen ring (C-18), the addition of a 19th carbon atom results in the C-19 ring (including androgens such as DHT 1 and testosterone), and the addition of a two-carbon lateral chain produces the pregnane ring (C-21). Despite their very similar structures, steroid hormones have significantly different physiological activities. DHT, the most potent androgen, estradiol (E 2 ), the most potent estrogen, and DHEA, the most abundant hormone precursor in humans and other mammals, exert numerous and diverse actions on their target tissues ( Fig. 1) (4 -6). The detailed mechanism of steroid discrimination has not been well elucidated up to now.
The enzymes of the 17␤-hydroxysteroid dehydrogenase (17␤-HSD) family are responsible for the last step in the formation of all androgens and estrogens. The reduction of the 17-ketone to a 17␤-hydroxyl increases the affinity of the steroids to their cognate receptors (7). The 17␤-reduction activity of these enzymes is thus required for the synthesis of all active androgens and estrogens. Molecular cloning has revealed the presence of seven isozymes of 17␤-HSD, six of which are members of the short chain dehydrogenase/reductase superfamily (8,9). 17␤-HSD1 primarily catalyzes the interconversion of estrone (E 1 ) and estradiol, but it also has some catalytic activity for the interconversion between DHEA and 5-androstene-3,17-diol, 4-androgene-3,17-dione and testosterone, as well as between A-dione and DHT (9,10), thus constituting a good model to study C-18/C-19 steroid or estrogen/androgen discrimination. We have previously proved that 17␤-HSD1 is a dimer consisting of two identical subunits (11). Immunochemical analyses have confirmed the presence of this enzyme in human placenta, breast, and ovary granulosa cells (12,13). Since it is well known that estradiol stimulates the proliferation of mammary tumor cells, this enzyme is an important target for breast cancer therapy (14,15).
The structure of 17␤-HSD1 from human placenta was first determined at 2.2 Å resolution (16). More recently, the structure of the complex of 17␤-HSD1 with E 2 , as well as the enzyme structure in the presence of E 2 and NADP, have been reported (17,18). Nevertheless, it is still unclear why a small structural modification, like the addition of a C-19 methyl group, can result in a very different catalytic activity, and how steroids are recognized by different isozymes and receptors. To answer these questions, we have determined and compared the structures of 17␤-HSD1 in complex with different steroids. Here we report the crystal structures of 17␤-HSD1 complexed with ei-ther DHEA or DHT. The comparison of the structures of DHEA, DHT, and E 2 complexes with 17␤-HSD1 gives an informative picture of important steroid hormone recognition and discrimination by the enzyme.

EXPERIMENTAL PROCEDURES
Materials-DHEA, DHT, 17␤-estradiol, and estrone were purchased from Aldrich. NAD, NADH, ␤-octylglucoside, dithiothreitol, polyethylene glycol 4000, and glycerol were from Sigma. We packed the Q-Sepharose ion exchange and the Blue-Sepharose CL-6B columns ourselves with media and columns from Amersham Pharmacia Biotech (Montreal, Canada). Phenyl-Superose HR 10/10 was purchased directly from the same company. All reagents were of the best grade available. Spodoptera frugiperda cells (Sf9 cells), Bac-N-Blue TM transfection kit, and transfer vector pBlueBac4.5 were purchased from Invitrogen Corp.; Grace's insect cell culture medium, yeastolate, and lactalbumin hydrolysate were from Life Technologies, Inc.; cell culture grade fetal bovine serum was from HyClone.
Crystallization, Data Collection, and Refinement-17␤-HSD1 was purified from human placenta with a previously described rapid purification procedure (11). The crystallization was carried out with the vapor-diffusion technique in hanging drops. Crystals were obtained in several days with 30% polyethylene glycol 4000, 0.06% (w/v) ␤-octylglucoside, 0.12 M MgCl 2 , 0.1 M Hepes, pH 7.5, and in the presence of saturating DHEA or DHT (1 mM). In the absence of glycerol, 17␤-HSD1 gets inactivated readily at low temperature (at less than 10 -15°C), and it gets reactivated by warming at 30°C where the enzyme is stable (29). The enzyme stability is necessary for obtaining high quality crystals with a high occupancy of substrate. We thus attempted the crystallization at 27°C, which yielded good results. 2 The data for the DHEA complex was collected on an R-axis IIC image plate area detector on a Rigaku RU-H2R rotating anode generator equipped with a graphite monochromator and a 0.5-mm collimator at room temperature. Data for the DHT complex were collected at Ϫ150°C at beamline X12-C at Brookhaven National Laboratory. The two data sets were processed using DENZO and SCALEPACK (30,36). Both data sets were isomorphous with the native and E 2 complex structures and were refined with X-Plor (31) using the protein portion of the structure of the E 2 complex as a starting point. Model building was done using the program O (32). The refinement was begun with data between 10 and 3 Å resolution and subsequently was extended to the high resolution limit of each data set.
Both structures were refined with good geometry ( Table I). The electron densities for DHEA and DHT are well defined, and they have been modeled and refined with full occupancies. The electron densities for the 43 C-terminal residues and for the flexible segment between residues 192 and 207 are stronger than those found in the first enzyme structure and in the estradiol complex (17), but they are still not clear enough to define the atomic location of those residues (15% of the protein by molecular mass). These missing portions of the model may contribute to the somewhat high free R factor for the structure of the DHT complex.
Steady-state Kinetic Study-Kinetic data for the wild type 17␤-HSD1 with DHEA and DHT and for the mutant 17␤-HSD1 with E 1 , E 2 , DHEA, and DHT were obtained by spectrophotometric measurement of NADP(H) reduction and/or oxidation at 37 Ϯ 1°C by measuring the absorbance change at 340 nm. The reaction mixture contained 50 mM phosphate buffer, pH 7.5, 100 M NADP(H), and 10 -60 M steroid (with smaller upper concentrations for estrogens: 10 -30 M for estradiol and 1-10 M for estrone). The cell subfractions obtained from the wild type, and from the variants, were sonicated and centrifuged at 100,000 ϫ g for 0.5 h to separate the mitochondrial and microsomal fractions resulting in samples ready for assaying the enzyme activity for various steroids. Lineweaver-Burk plots were used to determine the K m and V max values. A blank value lacking steroids was obtained under the same condition and subtracted. Protein concentrations were determined by the method of Bradford (26). Protein content was rapidly assayed using a Coomassie Blue dye-binding procedure combined with scanning densitometry (27). The enzymatic activity was calculated as described previously (11).
Site-directed Mutagenesis-The human 17␤-HSD1 cDNA coding region was subcloned from the vector pVL/17␤-HSD (28) by polymerase chain reaction to reconstruct the vector pBlueBac4.5/17␤-HSD1 for site-directed mutagenesis. The mutation to obtain the variant L149V, was constructed by using the QuickChange site-directed mutagenesis kit (Stratagene) following the manufacturer's instructions. The mutation used two primers: 5Ј-GGAGGATTGATGGGGGTGCCTTTCAAT-GACG-3Ј and 5Ј-CGTCATTGAAAGGCACCCCCATCAATCCTCC-3Ј. After verifying the mutation by dideoxynucleotide sequencing, the mutated transfer vector was cotransfected in Sf9 cells with linearized pBlueBac DNA following the protocol described by the manufacturer (35).
Sf9 cells were grown as monolayers at 27°C in Grace's insect cell culture medium containing 10% fetal calf serum and 10 g/ml gentamycin. Cells were infected with virus at a multiplicity of infection of 0.1-0.5 for virus amplification, and at a multiplicity of infection of 5-10 for protein expression. Cells were harvested 72 h after infection, washed with phosphate-buffered saline, pelleted in aliquots (1 ϫ 10 7 ), and stored at Ϫ80°C for later use.

RESULTS AND DISCUSSION
Binding Pocket-In these complexes, the core of the protein maintains the same structure as that of the enzyme (16)   The binding pocket is approximately 7.5 Å by 9 Å in crosssection and 16 Å deep. Hydrophobic atoms constitute 79% of the total 335 Å 2 surface of the pocket. It also contains the Tyr-X-X-X-Lys sequence that is conserved in all members of the short chain dehydrogenase/reductase superfamily and essential for the enzymatic activity.

DHEA and DHT Recognition and Sex Hormone Discrimination-
The O-17 atom of DHEA (and that of DHT) forms two hydrogen bonds with the protein. These hydrogen bonds involve Tyr-155 and Ser-142 in the active site, and form a triangular-shaped network similar to that observed between the steroid and the catalytic residues in the 17␤-HSD-E 2 complex. The O-3 atom in DHEA forms two hydrogen bonds with N-⑀2 of His-221 and O-⑀2 of Glu-282. The hydrogen bond between Glu-282 and O-3 is weak, as shown by the flexibility of the side chain of Glu-282, which has a high B-factor and a weak electron density. The hydrogen bond formed with His-221 is strong and important for orientating the bound substrate. The latter hydrogen bond is also found in the 17␤-HSD1-DHT complex, and a similar hydrogen bond is found in the complex of rat liver 3␣-hydroxysteroid dehydrogenase with testosterone (19).
When the structures of the three substrate complexes of 17␤-HSD1 are superimposed, it is found that the core of DHEA rotates by nearly 20°about an axis along the length of the steroid binding pocket, and shifts 1.5 Å at the O-3 end and 0.6 Å at the O-17 end as compared with the position of E 2 in the E 2 complex (Figs. 3 and 4). DHT shifts 1.4 Å at the O-3 end and 0.6 Å at the O-17 end, but the core of DHT rotates only slightly.
These results are surprising, since it was thought that the binding pocket fitted the steroid tightly (16). In all structures of DHEA, DHT, and E 2 , the C-10 is in a precise alignment with the fork-like side chain of Leu-149. In the E 2 structure, the distance between E 2 and the C␦ of Leu-149 is 4.6 Å. Because there is a ␤-methyl C-19 attached to C-10 of DHEA and DHT, these steroids obviously cannot remain at the same position due to steric hindrance. The position of Leu-149 is stabilized by two main chain hydrogen bonds with Pro-150 O and Gly-148 O. The electron density of Leu-149 is well defined, and the Bfactors of its C␦1 and C␦2 are half the average B-factor for the whole protein. The interaction of Leu-149 with both DHEA and DHT is sufficiently strong to shift C␦1 and C␦2 of Leu-149 by 0.4 Å from their original position in the presence of E 2 . This rigidity of part of the active site forces the shift in position of DHEA and DHT. Compared with E 2 binding, the different binding locations of DHT and DHEA result in reduced interactions of these steroids with several important non-polar residues in the binding pocket, although they can make new interactions with the more polar Ser-222 and Pro-187. The E 2 complex structure and our modeling study 3 have revealed that Phe-259 plays an important part in E 2 binding. Its side chain forms five van der Waals contacts with the A-ring of E 2 . Because the interactions with DHEA and DHT are decreased, we observed that the side chain of Phe-259 in the DHEA and DHT complexes has shifted a little from its binding position in the E 2 complex and that its B-factor has increased. Due to the hydrophobic nature of steroids, van der Waals contacts between 17␤-HSD1 and the non-polar atoms of the steroid provide major contributions to the binding energy.  (Fig. 3). The flexibility of helix ␣GЈ, which contains several residues in the binding pocket, may be important for steroid binding.
Evidence for the relative binding affinities of various substrates can be obtained from the steroid electron density in the different complex structures. The position of E 2 in the binding pocket was proven by its high level of electron density. In contrast, the electron densities of DHEA and DHT are weaker than that of E 2 in the cognate complex, although they were crystallized in the presence of higher concentrations of DHEA and DHT (about 1 mM) than the one that was used in the cocrystallization of E 2 (500 M) (17). The slightly higher B factor values of the DHEA and DHT complexes also show that their positions are less well defined than was E 2 in the binding pocket.
Steady-state kinetic studies of these steroids with 17␤-HSD1 were carried out in parallel, to compare with the structural information from various complexes. These kinetic studies show that the K m for DHEA is more than 1000-fold greater than that for estrone, and the K m value for DHT is about 10-fold greater than that for estradiol, in agreement with the structural information (Table II).
The catalytic mechanism of 17␤-HSD1 for estrone reduction is thought to involve the direct transfer of a hydride ion from the C-4 position of the nicotinamide nucleotide to the acceptor carbonyl C-17 of estrone to produce a 17␤-hydroxyl. The NADP ϩ molecule binds in an extended conformation, with the nicotinamide moiety in a syn conformation pointing toward the substrate binding site of 17␤-HSD1 (17,20). The shift of DHT and DHEA could increase the distance between C-17 of the steroid and C-4 of the nicotinamide. A modeling study using the model from Azzi et al. (17) and the NADP ϩ model from Breton et al. (18) indicated that the distance is 3.6 Å for E 2 , 4.0 Å for DHT, and 3.9 Å for DHEA. If these longer distances are maintained when 17␤-HSD1 binds both the cofactor and these substrates, then we would expect a lower rate of hydride transfer for the C-19 steroids. Thus, we propose that the weak affinity of DHEA and DHT for 17␤-HSD1 is consistent with their increased K m values and that the longer distance between C-17 and C-4 of NADP ϩ is in accord with their lower k cat values when compared with E 2 . As a result of the K m and k cat modification, the specificity difference between the estrogens and C-19 steroids could be between 250-and 50,000-fold. The discrimination between these steroids is thus guaranteed. Of all the steroid hormones, only the estrogens E 1 and E 2 lack the C-19 methyl group. 17␤-HSD1 is well designed for its high specificity toward estrogen hormones. Although the binding cavity is almost entirely composed of hydrophobic residues having nonspecific interactions, the arrangement of residues yields highly specific interactions allowing only one orientation for the steroid, corresponding to the strongest binding. Thus, relatively small changes in the shape of the hormones can significantly affect the binding position and affinity resulting in critical hormone recognition and substrate selectivity. The interactions between C-19 and the hydrophobic side chains around Leu-149 appear to be the most important for discriminating C-19 steroids and in preventing them from binding in an ideal position. Therefore, the shape of the binding pocket plays a critical role in steroid hormone discrimination. 1 In parallel with the crystallographic study, site-directed mutagenesis was carried out to produce the variant by replacing the Leu-149 by the very similar but smaller valine. As expected from our knowledge of the structure, the catalytic efficiency and substrate discrimination were significantly modified in the variant, which, unlike the native enzyme, demonstrates similar affinities for E 2 and DHEA. The K m of variant enzyme for DHEA has decreased from 33 M to 4.3 M, while the K m for E 1 has increased from 0.03 M to 0.3 M (Table II). Similarly, the variant enzyme showed a lower K m value for DHT as compared with the wild type, though the effect is not so significant as in the case of DHEA (Table II). These changes appear to be due to some loss of favorable interactions with estrogens and to the removal of unfavorable interactions with C-19 steroids and thus are in good agreement with the above structural analysis.
In Table II, we can see that the specificity values of the Leu-149 3 Val variant for estrogens may decrease, while these values significantly increased for C-19 steroids, resulting in a loss of C-18/C-19 steroid discrimination. In fact, the specificity value for C-19 steroids and estrogens became similar (Table II) (21). Therefore, the conformation of the D-ring was adjusted to fit the density (Fig. 5). O-17, C-17, and C-16 were moved 1.2, 0.6, and 0.7 Å, respectively, toward the ␤-surface. The O-17-C-17 bond was shifted by 24°compared with its orientation in the energy-minimized and small molecular models. In this new conformation, the 17-ketone oxygen can form hydrogen bonds with Tyr-155 (3.6 Å) and Ser-142 (3.3 Å). This suggests that the hydroxyl groups of Tyr-155 and Ser-142 can induce the new position of the 17ketone oxygen of DHEA, thereby allowing Tyr-155 to donate its proton and to form a hydrogen bond. These hydrogen bonds could stabilize this high energy state, thus facilitating the proton transfer from Tyr-155 to O-17, and the hydride transfer from nicotinamide ring to C-17 of the steroid. The presence of the positively charged side chain of Lys-159, in proximity to the Tyr-155 hydroxyl oxygen, could also facilitate the stabilization of the ␤-oriented ketone and proton transfer. By comparison, in the structure of the E 2 complex, the 17-hydroxyl and D-ring of estradiol are ␤-oriented, forming strong hydrogen bonds with Tyr-155 and Ser-142. Thus, the ␤-oriented ketone shows the characteristics of the intermediate of 17␤-HSD reduction, and this is evidence for enzyme-induced strain of the substrate. The stabilization of a ␤-oriented ketone is an obligatory step in the formation of the 17␤-hydroxyl, because it decreases the energy barrier for the conversion of 17-ketone to 17␤-hydroxyl. Therefore, the present results may indicate a new catalytic mechanism that includes a conformational modification during the reduction of DHEA, and this may also apply to the reduction of other 17-keto-steroid hormones (Fig. 6).
The present structure and kinetic studies of inhibitors have determined that a 17␤-hydroxyl or 17-ketone in the ligand is essential for tight binding with 17␤-HSD1 (24). This is consistent with the formation of strong hydrogen bonds with Tyr-155 and Ser-142 as seen in the 17␤-HSD1-E 2 complex. Even though the orientations of the core of DHEA and DHT have been changed in the binding pocket, the hydrogen bonds to Tyr-155 have remained. Hydrogen bonds are highly directional and therefore are important in the determination of the binding orientation, while the non-directional van der Waals forces are important for the binding affinity of steroids.
General  regions. The first region recognizes the steroid phenolic A-ring. It contains the conserved His-221 and Glu-282 residues that could form hydrogen bonds with O-3. The second region binds to the central hydrophobic core of the steroid. It contributes to the main thermodynamic force favoring the binding of substrates. The catalytic region surrounds the D-ring, and contains the absolutely conserved catalytic residue Tyr-155 that is located on the ␤-face of the steroid and forms a hydrogen bond with O-17. The ␣-face of the steroid is accessible to the nicotinamide to facilitate hydride transfer. Because both estrogens and androgens (with C-19) may have 3-and 17-hydroxyl or ketone groups, the critical residues that are expected to determine specificity for the estrogenic and androgenic members of the 17␤-HSD family must lie within the hydrophobic binding region. In estrogen-specific enzymes, some residues will sterically hinder the C-19-methyl group of the androgens, resulting in an unfavorable binding position or no binding at all. In androgen-specific enzymes, such as 17␤-HSD3, there may be a hydrophobic cavity that can strongly interact with the C-19. We predict that the hydrophobic region of 17␤-HSD2 will be more tolerant than that of 17␤-HSD1, as it shows similar activity with both C-18 and C-19 hormones.
We propose that other C-19 steroids (androstenedione, androsterone, androst-5-ene-3␤,17␤-diol, and testosterone) must have similar interactions with 17␤-HSD1 as do DHEA and DHT. This prediction is in agreement with our kinetic data, which show that these C-19 steroids have similar K m and K cat values for 17␤-HSD1. 4 The combination of hydrogen bonding and hydrophobic interactions in the construction of the specific steroid binding site in 17␤-HSD1 is expected to be a common feature in other protein-steroid interactions. Hydrogen bonds alone are insufficient for the requisite ligand-binding affinity; van der Waals contacts contribute significantly to binding energy. Similar mechanisms were recently shown in the interactions between the estrogen receptor and 17␤-estradiol (22).
Inhibitor Implication-Forty percent of all cancer cases are sex steroid-sensitive, including breast, prostate, ovarian, and uterine cancers. Steroid-metabolizing enzymes are thus prime candidates for approaches based on control of intracrine activity. Because the 17␤-HSD isozymes have much greater tissue specificity than the receptor, this family of enzymes is an attractive target for the design of potent and selective drugs to combat steroid-related disorders (23,24). The 17␤-HSD1 complex structures suggest several ways to design potent inhibitors. The narrow hydrophobic binding region contributes the main thermodynamic force for binding. For the best access to the restricted pocket, a planar hydrophobic ring structure like that of E 2 should be chosen as the core of an inhibitor. A ␤-oriented electron withdrawing group should be present to form hydrogen bonds with Tyr-155 and Ser-142. In addition, an ␣-oriented hydrophobic group at C-17 or C-16 would help to prevent the nicotinamide nucleotide cofactor from binding. Despite the narrowness of the hydrophobic cleft, there appears to be sufficient space to accommodate substituents at the 7␣position of the steroid. 17␤-Estradiol derivatives with 7␣-alkyl substituents have been synthesized and shown to inhibit 17␤-HSD1 (25). The selectivity of an inhibitor toward androgenic or estrogenic HSDs would be controlled by the shape of its hydrophobic core. A group in the C-19 position of the androgen would greatly decrease its affinity toward 17␤-HSD1 while increasing the affinity toward 17␤-HSD3.