Human 3α-HSD3 (3α-hydroxysteroid dehydrogenase type 3) plays an essential role in the inactivation of the most potent androgen 5α-DHT (5α-dihydrotestosterone). The present study attempts to obtain the important structure of 3α-HSD3 in complex with 5α-DHT and to investigate the role of 3α-HSD3 in breast cancer cells. We report the crystal structure of human 3α-HSD3·NADP+·A-dione (5α-androstane-3,17-dione)/epi-ADT (epiandrosterone) complex, which was obtained by co-crystallization with 5α-DHT in the presence of NADP+. Although 5α-DHT was introduced during the crystallization, oxidoreduction of 5α-DHT occurred. The locations of A-dione and epi-ADT were identified in the steroid-binding sites of two 3α-HSD3 molecules per crystal asymmetric unit. An overlay showed that A-dione and epi-ADT were oriented upside-down and flipped relative to each other, providing structural clues for 5α-DHT reverse binding in the enzyme with the generation of different products. Moreover, we report the crystal structure of the 3α-HSD3·NADP+·4-dione (4-androstene-3,17-dione) complex. When a specific siRNA (100 nM) was used to suppress 3α-HSD3 expression without interfering with 3α-HSD4, which shares a highly homologous active site, the 5α-DHT concentration increased, whereas MCF7 cell growth was suppressed. The present study provides structural clues for 5α-DHT reverse binding within 3α-HSD3, and demonstrates for the first time that down-regulation of 3α-HSD3 decreases MCF7 breast cancer cell growth.
Most breast cancers are sex-hormone-related [1,2]. According to statistics from the American and Canadian Cancer Society, breast cancer is the second leading cause of death from cancer in women in North America. E2 (17β-oestradiol), the most potent oestrogen, stimulates the proliferation of breast cancer cells, whereas the androgen 5α-DHT (5α-dihydrotestosterone) can reduce breast cancer cell growth [3–7]. 5α-DHT is accepted as the most potent androgen, and is intimately involved in the development of prostate cancer [8–11]. In humans, 3α-HSD3 (3α-hydroxysteroid dehydrogenase type 3) plays a major role in 5α-DHT reduction to produce the inactive steroid 3α-diol (5α-androstane-3α,17β-diol) in the presence of NADPH [12–14], indicating that 3α-HSD3 acts as a pre-receptor regulator of AR (androgen receptor) in prostate cancer [15–18]. However, limited information exists concerning the role of 3α-HSD3 in breast cancer cells.
Four isoenzymes (types 1, 2, 3 and 4) have been identified in the human 3α-HSD family, which are also named AKR (aldo-keto reductase) 1C4, 1C3, 1C2 and 1C1 in the AKR superfamily [19,20]. These isoenzymes have various degrees of 3-, 17- and 20-ketosteroid reductase activities and are involved in the inactivation of androgens, progestins and bile acid precursors [18,21,22]. They are highly homologous enzymes and show differential tissue distributions [13,22,23]. Human 3α-HSD1 (AKR1C4) is a liver-specific enzyme that acts on steroid metabolism in the liver [22,23]. Human 3α-HSD2 (AKR1C3, also named 17β-HSD5) is highly expressed in prostate and mammary glands, and shows a marked ability to biosynthesize testosterone from 4-dione (4-androstene-3,17-dione) [24,25]. Human 3α-HSD3 (AKR1C2) is expressed in several tissues including liver, lung, brain, prostate, testis, mammary gland and the adrenals [13,22]. In vitro, 3α-HSD3 can oxidize 3α-diol to 5α-DHT in the presence of NAD+. However, it practically acts as a reductase to inactivate 5α-DHT in the cellular environment [13,14,26]. Human 3α-HSD3 is also recognized as a bile-acid-binding protein and can be inhibited by bile acids such as ursodeoxycholate [27,28]. In the pathogenesis of female hirsutism, a reduced expression of 3α-HSD3 along with an elevated 5α-DHT level is found in tissue of hirsute women . Human 3α-HSD4 (AKR1C1, also named 20α-HSD) is found in the mammary gland, ovary, uterus and placenta, and predominantly exerts its 20-ketosteroid reductase activity to inactivate progesterone [22,30,31]. It is noteworthy that 3α-HSD3 and 3α-HSD4 share 97.8% sequence identity, with seven residues difference between the two enzymes, of which only one residue is different in their active sites. However, they show distinctive specificity for steroidal substrates.
Previous crystal structures of 3α-HSD3 show that this enzyme possesses an (α/β)8-barrel motif, and its steroid-binding pocket possesses considerable flexibility, which is formed by five loops including loop A (residues 117–143), loop B (residues 217–238), loop C (residues 299–323) and two short loops (residues 23–32 and 51–57) [32–34]. Although these structures provide some detailed information, the crucial structure of 3α-HSD3 in complex with 5α-DHT is still not available.
In the present paper, we report the high-resolution crystal structure of the 3α-HSD3·NADP+·A-dione (5α-androstane-3,17-dione)/epi-ADT (epiandrosterone) complex, which was obtained by co-crystallization with 5α-DHT. It was surprising to observe that oxidoreduction of 5α-DHT occurred during the crystallization process to produce a mixture of steroids. Location of A-dione and epi-ADT was identified in the enzyme cavities. Moreover, we report the crystal structure of the 3α-HSD3·NADP+·4-dione complex. Using a crystallization condition without the acetate ion, the steroid-binding conformation was more reasonable and less distorted. In addition, the use of a specific siRNA to suppress 3α-HSD3 expression in breast cancer cells permitted further investigation of the impact of 3α-HSD3 expression on the concentration of 5α-DHT in breast cancer cells and the growth of such cells.
MATERIALS AND METHODS
MCF7 cells and the MTT cell proliferation assay kit were purchased from the A.T.C.C. (Manassas, VA, U.S.A.). FBS (qualified, U.S. origin) and Lipofectamine™ 2000 transfection reagent were purchased from Life Technologies Corporation. Low-glucose DMEM (Dulbecco's modified Eagle's medium) was from Sigma–Aldrich. siRNA was from Shanghai GenePharma. The 5α-DHT ELISA kit was from Alpha Diagnostic International. 5α-DHT and testosterone were from Steraloids. Diethyl ether, acetone, toluene and dichloromethane were purchased from Fisher Scientific. TLC plates were purchased from EMD Millipore. 14C-labelled 5α-DHT (53.5 mCi/mmol) was purchased from PerkinElmer Life Sciences. 14C-labelled testosterone (55 mCi/mmol) was purchased from American Radiolabeled Chemicals. NADPH, NADP+ and other chemicals were purchased from Sigma–Aldrich.
Protein purification and crystallization
Human 3α-HSD3 was purified following a previously reported procedure : in brief, the recombinant enzyme with a GST tag was initially purified with a glutathione–agarose column and the GST-fusion tag was then cut by thrombin. The second step of purification used a Q-Sepharose column. The enzyme was concentrated to approximately 30 mg/ml in a 10 mM KH2PO4/K2HPO4 buffer (pH 7.0) containing 0.5 mM DTT, 0.06% β-octyl glucoside, 1 mM EDTA, 1 mM NADP+ and 40 μM 5α-DHT or 40 μM testosterone. Crystals were obtained in the sitting drops by using the vapour-diffusion method at room temperature. Before crystallization, either 5α-DHT or testosterone was added into the 0.5 ml well solution to reach a final concentration of 1 mM. The crystallization condition included 100 mM sodium cacodylate (pH 6.0), 200 mM ammonium sulfate and 24–26% (w/v) PEG 3350. Crystals appeared after 3 days and reached their full size within 2 weeks. 5α-DHT can undergo oxidation to A-dione and then reduction to epi-ADT by the enzyme during the crystallization period. The well solution containing 15% glycerol was selected as the cryoprotective solution.
Data collection and structure determination
The X-ray diffraction data were collected at beamline 31-ID-D (λ=0.979 Å; 1 Å=0.1 nm) of the Advanced Photon Source (APS) at Argonne National Laboratory. Datasets were indexed and integrated by iMOSFLM . Using the reported 3α-HSD3 structure (PDB code 1J96) , the two crystal structures were solved by molecular replacement with the program MOLREP in CCP4 [37,38]. The programs REFMAC in CCP4 and COOT were used for refinement and model building respectively [39,40]. After the amino acid residues of the enzyme and water molecules were refined, NADP+ and steroids were fitted into the 2Fo−Fc omit electron density and refined. Finally, the quality of the models was verified with PROCHECK .
Oxidoreduction assay of steroids and GC–MS analysis
5α-DHT oxidoreduction was performed at 37.0±0.3°C in a volume of 1 ml of 25 mM Tris/HCl buffer (pH 7.4) containing 10% (v/v) glycerol, 0.1 mM 5α-DHT (with 0.1 μM 14C-labelled 5α-DHT), 1 mM NADP+ and 0.01 mg/ml BSA. The reaction was initiated by adding the enzyme to the reaction mixture, and the final concentration of the enzyme was 0.2 mg/ml. The reaction was terminated by adding 3 ml of diethyl ether after overnight incubation. The reactions were performed in duplicate. Steroids were separated by a standard TLC protocol . The TLC plate was exposed and quantified by a Storm imaging system (Molecular Dynamics). Similarly, testosterone oxidation was performed using the same experimental conditions as for 5α-DHT oxidoreduction except that 5α-DHT and 14C-labelled 5α-DHT were replaced by testosterone and 14C-labelled testosterone respectively. In addition, the reaction mixture for 5α-DHT oxidoreduction (in the absence of 14C-labelled 5α-DHT) was analysed by GC–MS. The GC–MS analysis was carried out by the bioanalytical platform [Centre Hospitalier Universitaire (CHU) de Québec Research Center (CHUL), Université Laval, Québec, Canada].
MCF7 cells were cultured in Phenol Red-free low-glucose DMEM in the presence of 10% (v/v) FBS. Cells were maintained in a humidified incubator at 37°C supplied with 5% CO2. When indicated, FBS was treated with 2% (w/v) dextran-coated charcoal overnight at 4°C to remove the endogenous steroids from the serum.
Synthesis and transfection of siRNA
Human 3α-HSD3 siRNA (sense, 5′-AAGCUCUAGAGGC-CGUCAAAUTT-3′; antisense, 5′-AUUUGACGGCCUCUA-GAGCUUTT-3′) was designed according to a previous study  and the specificity of the siRNA was checked by BLAST against the human genome . Duplex 3α-HSD3 siRNA and NC (negative control) siRNA were synthesized and purified with HPLC by Shanghai GenePharma. siRNA was transfected using Lipofectamine™ 2000 following the manufacturer's protocol.
qRT-PCR (quantitative real-time reverse transcription–PCR)
MCF7 cells were plated into a six-well plate at 4×105 cells/well with 2 ml of medium and were cultured for 1 day before siRNA transfection. MCF7 cells were transfected with 100 nM 3α-HSD3-specific siRNA or NC siRNA for 24 h and then cells were treated with TRIzol® reagent. The treated samples were sent to the qRT-PCR platform [Centre Hospitalier Universitaire (CHU) de Québec Research Center (CHUL), Université Laval, Québec, Canada] for quantification of 3α-HSD3 and 3α-HSD4 mRNAs and the Hprt1 (hypoxanthine phosphoribosyltransferase 1) housekeeping gene was used for the qRT-PCR data normalization. The design of primer sequences for 3α-HSD3 (5′-CCTAAAAGTAAAGCTCTAGAGGCCGT-3′ and 5′-CAACTCTGGTCGATGGGAATTGCT-3′) and 3α-HSD4 (5′-AAAGTAAAGCTTTAGAGGCCACC-3′ and 5′-AAATGAATAAGGTAGAGGTCAACATAA-3′) was based on previous studies [26,45].
Determination of 5α-DHT levels by ELISA
MCF7 cells were seeded into 24-well plates at 5×104 cells/well with 0.5 ml of medium treated with dextran-coated charcoal. After culture for 24 h, cells were transfected with 100 nM 3α-HSD3-specific siRNA or NC siRNA and grown for another 48 h. Then, the cell culture medium was replaced with fresh charcoal-treated medium containing 0.7 nM 5α-DHT. Cell culture medium supernatants were collected after 48 h. 5α-DHT levels in supernatants were determined using a commercial 5α-DHT ELISA kit (Alpha Diagnostic International) following the manufacturer's protocol.
MCF7 cells (5×104 cells/well) were seeded into 24-well plates to evaluate the conversion of 5α-DHT into 3α-diol/3β-diol (5α-androstane-3β,17β-diol). The medium used in the activity assay was treated with dextran-coated charcoal. After transfection with 100 nM 3α-HSD3-specific siRNA or NC siRNA for 48 h, 100, 200 or 300 nM 14C-labelled 5α-DHT was added to the medium. Each activity assay was carried out for three incubation times (4, 8 or 12 h) before the reaction was stopped. Steroids were separated using a standard TLC protocol for quantification .
Cell proliferation assay
Cell proliferation was determined by the MTT assay. MCF7 cells were seeded into 24-well plates in a similar manner to that used in the ELISA and cultured for 24 h. Next, cells were transfected with 100 nM 3α-HSD3-specific siRNA or NC siRNA. After incubation for another 24 h, the culture medium was replaced with fresh charcoal-treated medium containing different concentrations of 5α-DHT (0.1 nM or 1.0 nM). Cell growth study was performed for 96 h and the medium was replaced every 48 h. The MTT assay was carried out as described previously . Absorbance was read at 570 nm with a reference at 650 nm using a plate reader (Spectra Max 340PC, Molecular Devices).
Crystal growth of human 3α-HSD3 in complex with A-dione/epi-ADT and 4-dione
5α-DHT is the principal steroid transformed by human 3α-HSD3. However, its complex structure with the enzyme is still not available. At the beginning, crystal growth of human 3α-HSD3 in complex with 5α-DHT was performed using the reported crystallization condition, which included 100 mM sodium citrate (pH 5.6), 260 mM ammonium acetate, 1 mM 5α-DHT and 26–27% (w/v) PEG 4000 . After obtaining the crystal, the solved crystal structure showed that acetate and citrate molecules were deeply bound in the steroid-binding site, whereas 5α-DHT could not be observed. The entire structure resembled a more recently reported human 3α-HSD3·NADP+·citrate/acetate complex . To solve this problem, the crystallization condition was adjusted to contain 100 mM sodium cacodylate (pH 6.0), 200 mM ammonium sulfate, 1 mM 5α-DHT and 24–26% (w/v) PEG 3350. By doing so, the crystal was obtained and the electron density of the steroid was shown in the steroid-binding site. Surprisingly, the solved structure shows that the bound steroids can be fitted with A-dione and epi-ADT, but not with 5α-DHT. Oxidoreduction of 5α-DHT occurred during the crystallization and a mixture of steroids including A-dione and epi-ADT was generated (Supplementary Figure S1); this phenomenon was also observed for human steroid 5β-reductase co-crystallization with 5α-DHT . Furthermore, in the 3α-HSD3·NADP+·testosterone/acetate complex reported previously , the acetate molecule from the crystallization condition with a high concentration (260 mM) saturated the active site of the enzyme and was able to distort the actual steroid-binding conformation. Therefore human 3α-HSD3 was co-crystallized with testosterone using an amended crystallization condition that did not contain acetate molecules, resulting in the acquisition of the 3α-HSD3·NADP+·4-dione complex structure.
Overall structures of human 3α-HSD3 in complex with A-dione/epi-ADT and 4-dione
The crystal structures of the 3α-HSD3·NADP+·A-dione/epi-ADT complex and the 3α-HSD3·NADP+·4-dione complex both possess the monoclinic space group of P21. There are two 3α-HSD3 molecules (named molecule A and molecule B) contained in an asymmetric unit for each crystal structure (Figure 1A). The 3α-HSD3·NADP+·A-dione/epi-ADT complex was refined at 1.20 Å with a crystallographic R-factor of 15.5% and an Rfree-factor of 17.4%. The 3α-HSD3·NADP+·4-dione complex was refined at 1.75 Å with an R-factor of 16.4% and an Rfree-factor of 20.5%. The protein backbones of the two complexes are quite similar to each other and the RMSD of the backbone Cα atoms in both molecules A is 0.42 Å. However, loop A (residues 117–143) of both complexes undergoes a considerable conformational variation and the RMSD value for 27 Cα atoms of loop A in both molecules A is 1.32 Å. It is noteworthy that molecule A has a lower thermal B-factor than that of molecule B in both complexes. The cofactor is tightly bound in the enzyme with an extended conformation and its binding mode is almost unchanged in different complexes [32–34]. The number of hydrogen bonds, salt bridges and Van der Waals interactions contribute to the stabilization of NADP+ in the enzyme as described previously [32,33]. The Ramachandran plot shows that more than 93% of residues of the two complexes are located in the most favoured regions. Data collection and refinement statistics are summarized in Supplementary Table S1.
Structure of the 3α-HSD3·NADP+·A-dione/epi-ADT complex
Human 3α-HSD3 in complex with A-dione/epi-ADT
The human 3α-HSD3·NADP+·A-dione/epi-ADT complex was obtained by co-crystallization in the presence of 5α-DHT. The difference Fourier map defines that two different steroids exist in the cavities of molecules A and B. In molecule A, A-dione well depicts the majority of the electron density map with the lower average B-factors. In molecule B, the presence of epi-ADT agrees with the electron density map with the lower average B-factors. However, partially weak electron density can also be found to accommodate epi-ADT in molecule A, and, correspondingly, a similar phenomenon can be observed in molecule B. Nevertheless, A-dione and epi-ADT represent the major steroid-binding modes in molecules A and B respectively, and were determined in the structure. It was surprising that, although 5α-DHT was introduced during the crystallization, the observed electron density defined that A-dione or epi-ADT bound in the cavity. Thus reactions simulating the crystallization conditions were performed. The results show that 5α-DHT is oxidized to A-dione in the presence of 1 mM NADP+, and then 5α-DHT and A-dione are reduced to 3α-diol/3β-diol and ADT/epi-ADT using the reduced cofactor respectively (Supplementary Figure S1). Following overnight incubation, A-dione and epi-ADT comprise 26% and 17% of the steroid mixture respectively. Moreover, the GC–MS analysis indicates further that the output ratios of 3α-diol/3β-diol and ADT/epi-ADT are approximately 2.2:1 and 2.1:1 respectively. It should be noted that product profiles of 5α-DHT oxidoreduction mimicking the crystallization conditions may not be directly compared with that of the previous in vitro and in vivo studies due to the different experimental conditions .
In molecule A, the O17 atom of A-dione is oriented towards the nicotinamide ring of NADP+ (Figure 1B). The α-face of A-dione interacts with the side chains of Val54 and Val128. Its β-face mainly interacts with the indole ring of Trp227, and the steroid C18 and C19 methyl groups point towards Trp227. It is worth mentioning that a water molecule is located at the catalytic core and forms hydrogen bonds with the side chains of Tyr55, His117 and the O17 atom of A-dione. In molecule B, the O3 atom of epi-ADT points towards the nicotinamide ring (Figure 1C). The α-face of epi-ADT is stacked with the indole ring of Trp227, and the steroid C18 and C19 methyl groups point towards Val54 and Tyr55. The O3 atom of epi-ADT forms a hydrogen bond with the His222 Nε atom. When molecules A and B are superimposed, the O3 and O17 atoms of both steroids are oriented upside-down relative to each other, and the planes of both steroids are flipped approximately 145° around the long axis of C3–C17 (Figure 1D). After superimposition, the RMSD value is 0.43 Å for their backbone Cα atoms. In molecules A and B, loop A and a short loop (residues 23–33) possess considerable fluctuation and the RMSD values for their Cα atoms are 1.06 and 1.05 Å respectively. The residues of Val54, Tyr55, His117, His222 and Leu308, which belong to the steroid-binding pocket, are relatively static, whereas the side chains of Val128, Ile129 and Leu306 display apparent movement to consort with steroids. In particular, the side chain of Leu306 contains two different conformations in molecules A and B, which may correspond to the two different orientations of steroids.
Human 3α-HSD3 in complex with 4-dione
The human 3α-HSD3·NADP+·4-dione complex was obtained by co-crystallization with testosterone in the presence of NADP+. During the crystallization process, the oxidation of testosterone occurred and 4-dione was generated (Supplementary Figure S2). The difference Fourier map clearly shows that 4-dione is located in the steroid-binding site of molecules A or B (Figure 2A). When molecules A and B are superimposed, the orientation and conformation of 4-dione are quite similar to each other and only a small shift of the steroid plane towards the indole ring of Trp227 can be observed (Figure 2B). The backbones of molecules A and B overlap each other satisfactorily, and the RMSD value for the backbone Cα atoms is 0.36 Å. Loop A still displays its flexibility and the RMSD value for its 27 Cα atoms is 0.96 Å. The side chains of Val54, Val128, Ile129 and Leu308 in the steroid-binding pocket also exhibit subtle adjustments in molecules A and B. The C3 atom and the A-ring of 4-dione point towards the nicotinamide ring. The α-face of 4-dione interacts with the side chain of Trp227, and the C18 and C19 methyl groups of 4-dione point towards the side chain of His117. The O3 atom of 4-dione forms a hydrogen bond with the His222 Nε atom.
Structure of the 3α-HSD3·NADP+·4-dione complex
Compared with the orientation of testosterone in the reported 3α-HSD3·NADP+·testosterone/acetate complex , it can be seen clearly that the plane of 4-dione in the 3α-HSD3·NADP+·4-dione complex generates an approximately 45° rotation around its C3–C17 long axis (Figure 2B). Meanwhile, the steroid penetrates deeper into the cavity of the enzyme, and the distance between the C3 atoms of steroids and the C4 atoms of the nicotinamide rings is changed to 4.3 Å from the former 6.2 Å. The acetate molecule in the testosterone/acetate complex, which originates from the crystallization conditions, is situated between the steroid and the cofactor, and occupies the catalytic site of the enzyme. Obviously, the repulsion between the O3 atom of testosterone and the ketone group of acetate produces a remarkable twist in the steroid orientation. However, the acetate molecule is replaced by a water molecule in the 4-dione complex, and the latter forms hydrogen bonds with the side chains of Tyr55 and His117. Moreover, the side chains of Val54, Ile129, Trp227, Leu306, Leu308 and the nicotinamide ring make slight adjustments to accommodate the more suitable orientation of the steroid.
The expression and siRNA suppression of 3α-HSD3 in MCF7 cells
The transcription of human 3α-HSD3 and 3α-HSD4 mRNA in MCF7 cells was evaluated by qRT-PCR. The results show that the mRNA level of 3α-HSD3 is approximately 2.5-fold that of 3α-HSD4 in MCF7 cells (Figure 3A). As a result of the high similarity between 3α-HSD3 and 3α-HSD4 at the nucleotide level, selective suppression of 3α-HSD3 expression was performed using a specific siRNA to avoid interference with 3α-HSD4. The qRT-PCR results show that, compared with control siRNA-treated cells, 3α-HSD3-specific siRNA significantly suppresses 3α-HSD3 expression, with negligible effect on the expression of 3α-HSD4 (Figure 3B).
The expression and siRNA suppression of 3α-HSD3 in MCF7 cells
Suppression of 3α-HSD3 expression increases the 5α-DHT concentration in MCF7 cells
To investigate the impact of 3α-HSD3 expression on 5α-DHT concentration in MCF7 cells, ELISAs were carried out using supernatants from the cell culture medium (Figure 4A). After siRNA treatment, MCF7 cells were cultured for 48 h with charcoal-treated medium containing 0.7 nM (203 pg/ml) 5α-DHT. Compared with the control siRNA-treated groups, the 3α-HSD3-specific siRNA treated group showed a significantly higher 5α-DHT concentration (∼154 pg/ml), which was approximately 2.3-fold the value of the control group (∼68 pg/ml).
Suppression of 3α-HSD3 in MCF7 cells enhanced 5α-DHT-mediated inhibition of cell growth
Suppression of 3α-HSD3 expression decreases the conversion of 5α-DHT into 3α-diol/3β-diol in MCF7 cells
The contribution of 3α-HSD3 to the conversion of 5α-DHT into 3α-diol/3β-diol was studied using the specific siRNA. After siRNA treatment, MCF cells were cultured with the addition of 5α-DHT (100, 200 or 300 nM) in the culture medium and each assay was carried out for 4, 8 or 12 h. The comparison of the 3α-diol/3β-diol formed after specific suppression of 3α-HSD3 expression is shown in Supplementary Table S2. Compared with the control group, the 3α-diol/3β-diol formed was decreased to 47.7±0.9%, 50±2.3% and 49.6±1.3% with the addition of 5α-DHT (100, 200 and 300 nM) respectively.
Suppression of 3α-HSD3 expression decreases MCF7 cell proliferation in the presence of 5α-DHT
To assess the potential role of 3α-HSD3 in MCF7 cell proliferation, siRNA-specific suppression of 3α-HSD3 expression in MCF7 cells was performed. After siRNA treatment, MCF7 cells were cultured for 96 h with charcoal-treated medium containing 0.1 or 1.0 nM 5α-DHT. MCF7 cells transfected with 3α-HSD3-specific siRNA exhibited a significantly decreased growth rate compared with MCF7 cells transfected with control siRNA, with approximately 20% or 25% reduction under these culture conditions containing 1.0 or 0.1 nM 5α-DHT respectively (Figure 4B).
In order to determine the molecular characteristics of 5α-DHT binding to human 3α-HSD3, a high-resolution crystal was obtained by co-crystallization in the presence of 5α-DHT and NADP+. However, during the crystallization process, 5α-DHT initiated a series of oxidoreductive reactions and generated a mixture of steroids. Therefore A-dione and epi-ADT were identified and located in two 3α-HSD3 molecules (molecules A and B) in a crystal asymmetric unit respectively. In molecule A, the O17 atom of A-dione points into the catalytic site of the enzyme, which means that 5α-DHT has been oxidized on its C17 hydroxy group to produce A-dione; the generated A-dione resides in the cavity of the enzyme. In molecule B, the O3 atom of epi-ADT is oriented towards the catalytic site of the enzyme, which implies that A-dione has been reduced on its C3 carbonyl group to produce ADT and epi-ADT. Only epi-ADT is retained in the cavity of the enzyme. Although an assay of 5α-DHT oxidoreduction mimicking the crystallization process showed that 5α-DHT was reduced to 3α-diol and 3β-diol, these two steroids and ADT were released from the cavity after the reactions. Nevertheless, the different orientations of A-dione and epi-ADT provide structural clues for 5α-DHT reverse binding in the cavity of the enzyme to generate different steroids. In molecules A and B, loop A is located in different neighbouring environments from the nearby symmetric molecules and displays an apparent flexibility. Furthermore, in the 3α-HSD3·NADP+·4-dione complex, testosterone was introduced before crystallization, whereas the observed electron density in the cavity agreed with 4-dione. This means that the C17 hydroxy group of testosterone first enters into the catalytic site of 3α-HSD3, and then testosterone is oxidized to 4-dione in the presence of the cofactor. Then, 4-dione is released from the catalytic site and re-enters with its C3 carbonyl group pointing inside. Of interest is that this process is reversed in the 3α-HSD3·NADP+·testosterone/acetate complex reported previously . Therefore this evidence lends further support to the different binding modes of steroids in 3α-HSD3.
The different binding modes of steroids in enzymes have also been found in other hydroxysteroid dehydrogenases such as human 3α-HSD2 (also named 17β-HSD5 or AKR1C3). In 3α-HSD2, the crystal structure shows that two reverse orientations of testosterone can be observed in the cavity of the enzyme . Moreover, for human steroid 5β-reductase (also named AKR1D1), the enzyme E120H mutant was co-crystallized with 5α-DHT . Oxidoreduction of 5α-DHT occurred during the crystallization process and produced a mixture of steroids. 3β-diol and epi-ADT were found to occupy the cavities of molecules A and B in the crystal asymmetric unit, which indicated that the different binding modes of steroids in the enzyme led to the multispecificity of the enzyme. Taken together, the different binding modes of steroids in the enzyme originate from the pseudo-self-symmetry property of steroids and the flexibility of the steroid-binding pocket.
Recently, the reported human 3α-HSD3·NADP+·progesterone complex showed that, although the steroid D-ring points towards the enzyme cavity, progesterone possesses two different orientations in the cavity . When superimposed with the 3α-HSD3·NADP+·A-dione/epi-ADT complex, the orientations of A-dione and progesterone resemble each other in both molecules A, and only the D-rings of the two steroids show limited detachment (Figure 5A). The C18 and C19 methyl groups of the two steroids point towards Trp227. The side chains of Val54, His222, Leu306 and Leu308 are flipped approximately 180°, which reflects the subtle flexibility of the steroid-binding site. Furthermore, although the A, B, C and D rings of epi-ADT and progesterone are located in opposition to each other in both molecules B, the orientations of the two steroid planes resemble each other, and both α-faces of the steroids interact with the side chain of Trp227 (Figure 5B). In addition, in the reported 3α-HSD3·NADP+·ursodeoxycholate complex, an overlay shows that, although the facial orientation of ursodeoxycholate is similar to that of epi-ADT in the cavity, their C3–C17 long axes are reversed to each other. There is a small swing between two ligands around the vertical axis of the steroid plane (Figure 5C). Therefore two major steroid orientations can be found in the cavity of 3α-HSD3, which implies further that the steroid-binding site of 3α-HSD3 possesses considerable flexibility.
Superimposition of the steroid-binding sites of the related 3α-HSD3 structures
The overall conformation of the enzyme in the 3α-HSD3·NADP+·4-dione complex is identical with molecule A of the 3α-HSD3·NADP+·progesterone complex , and the RMSD value is only 0.14 Å between the backbone Cα atoms of both molecules A. Although an overlay shows that the two steroid molecules are oriented back-to-back with their C18 and C19 methyl groups pointing towards opposite directions, the orientations of the two steroids resemble each other (Figure 5D). This indicates that the steroid-binding pocket of 3α-HSD3 has sufficient space and flexibility to allow the steroid planes to flip within.
An increased 5α-DHT concentration, as well as a decreased expression of 3α-HSD3 in prostate tumour compared with benign tissue, was previously observed by Stolz and co-workers [16,26]. They speculated that the loss of 3α-HSD3 in prostate cancer cells increased 5α-DHT-dependent cell growth. As generally accepted, excessive 5α-DHT leads to the stimulation of prostate cancer proliferation through binding to AR. 3α-HSD3 eliminates 5α-DHT in the prostate and this can prevent AR activation by excessive androgens. Moreover, these authors reported that, in the presence of progesterone, specific suppression of 3α-HSD4 alone or in combination with 3α-HSD3 in T47D cells led to decreased cell growth . Previous studies indicate that 5α-DHT still plays an important role in anti-proliferation in breast cancer cells, which is mediated by AR [5,6,50,51]. Human 3α-HSD3 is characterized by its marked ability to inactivate 5α-DHT, which prompted us to explore its role in breast cancer cells. Owing to the high sequence identity between 3α-HSD3 and 3α-HSD4, there is a lack of specific antibodies to distinguish these two enzymes . In the present study, 3α-HSD3 expression was suppressed by specific siRNA in MCF7 cells, and a relative elevation of 5α-DHT level was observed using ELISA. Meanwhile, in the presence of 5α-DHT, suppression of 3α-HSD3 expression by specific siRNA resulted in decreased MCF7 cell growth, suggesting that the suppressed expression of 3α-HSD3 strengthened the anti-proliferative effect of 5α-DHT in MCF7 cells. In addition to the outstanding ability of E2 production, human 17β-HSD1 also displays the ability to inactivate 5α-DHT . Suppression of 17β-HSD1 expression was shown to increase the 5α-DHT level in T47D cells, which showed an anti-proliferative effect . However, compared with 5α-DHT inactivation by 17β-HSD1 (Km 32.0 μM, kcat 2.40 min−1 and kcat/Km 0.08 min−1·μM−1) , 3α-HSD3 possesses a significantly higher specificity regarding the reduction of 5α-DHT (Km 1.1 μM, kcat 1.5 min−1 and kcat/Km 1.3 min−1·μM−1) , which implies that 3α-HSD3 may play a more important role than 17β-HSD1 in the inactivation of 5α-DHT. To sum up, the present study demonstrates that specific suppression of 3α-HSD3 expression showed its significant effect in 5α-DHT restoration and suppression of MCF7 cells growth, providing a new dimension of the control of this highly common human cancer. The novel findings of the present study provide a structural basis for designing certain lead compounds to block 5α-DHT depletion and to suppress breast cancer cell growth.
Bo Zhang, Xiao-Qiang Wang and Jean-Francois Theriault carried out the experimental studies. Bo Zhang and Xiao-Jian Hu solved the crystal structures. Dao-Wei Zhu supported the protein crystallization. Sheng-Xiang Lin, Bo Zhang and Peng Shang designed the study. Fernand Labrie contributed to the 3α-HSD3 study and revising the paper critically. Bo Zhang and Sheng-Xiang Lin prepared the paper. All authors read and approved the final paper.
We thank Dr Donald Poirier for providing us with several steroids and helpful discussions. We thank Dr Preyesh Stephen for the revision of the paper. We thank Ms Nathalie Paquet for qRT-PCR services and Mr Patrick Bélanger for GC–MS analysis, from the qRT-PCR platform and the bioanalytical platform [Centre Hospitalier Universitaire (CHU) de Quebec Research Center (CHUL) and Laval University, Québec] respectively. We thank PROTEO (The Quebec Network for Research on Protein Function, Structure, and Engineering) and Dr Albert Berghuis's laboratory at McGill University for providing the X-ray crystallography facilities for the initial X-ray dataset collection. X-ray diffraction datasets were collected at LRL-CAT at the Advanced Photon Source, Argonne National Laboratory, and at the CMCF, Canadian Light Source. Bo Zhang thanks the grant from CIHR (MOP 97917) to Lin et al. and a national scholarship provided by China Scholarship Council (CSC) for supporting his Ph.D. study.
PDB ACCESSION CODES
Crystal structures of human 3α-HSD3 in complex with A-dione/epi-ADT and in complex with 4-dione at 1.2 Å and 1.75 Å respectively have been deposited in the PDB under codes 4XO6 and 4XO7 respectively.
This work was supported by the Canadian Institutes of Health Research (CIHR) [grant number MOP 97917 (to S.-X.L.)]. Use of the Advanced Photon Source, an Office of Science User Facility operated for the U.S. Department of Energy (DOE) Office of Science by Argonne National Laboratory, was supported by the U.S. Department of Energy (DOE) [contract number DE-AC02-06CH11357]. Use of the Lilly Research Laboratories Collaborative Access Team (LRL-CAT) beamline at Sector 31 of the Advanced Photon Source was provided by Eli Lilly Company, which operates the facility. The Canadian Macromolecular Crystallography Facility (CMCF) is supported by the Canadian Foundation for Innovation (CFI), Natural Sciences and Engineering Research Council (NSERC) and Canadian Institutes of Health Research (CIHR). B.Z. was supported by a national scholarship provided by the China Scholarship Council (CSC) and the Canadian Institutes of Health Research (CIHR) grant mentioned above (to S.-X.L. as principal investigator) to support his graduate study. We are grateful to The National Natural Science Foundation of China ([grant number 31011120381] for supporting the international travel expenses of X.-J.H. for his visit to S.-X.L.’s laboratory.
Dulbecco's modified Eagle's medium
quantitative real-time reverse transcription–PCR