Inhibition of Human Steroid 5β-Reductase (AKR1D1) by Finasteride and Structure of the Enzyme-Inhibitor Complex*

The Δ4-3-ketosteroid functionality is present in nearly all steroid hormones apart from estrogens. The first step in functionalization of the A-ring is mediated in humans by steroid 5α- or 5β-reductase. Finasteride is a mechanism-based inactivator of 5α-reductase type 2 with subnanomolar affinity and is widely used as a therapeutic for the treatment of benign prostatic hyperplasia. It is also used for androgen deprivation in hormone-de pend ent prostate carcinoma, and it has been examined as a chemopreventive agent in prostate cancer. The effect of finasteride on steroid 5β-reductase (AKR1D1) has not been previously reported. We show that finasteride competitively inhibits AKR1D1 with low micromolar affinity but does not act as a mechanism-based inactivator. The structure of the AKR1D1·NADP+·finasteride complex determined at 1.7 Å resolution shows that it is not possible for NADPH to reduce the Δ1-2-ene of finasteride because the cofactor and steroid are not proximal to each other. The C3-ketone of finasteride accepts hydrogen bonds from the catalytic residues Tyr-58 and Glu-120 in the active site of AKR1D1, providing an explanation for the competitive inhibition observed. This is the first reported structure of finasteride bound to an enzyme involved in steroid hormone metabolism.

The ⌬ 4 -3-ketosteroid functionality is present in many important steroid hormones, e.g. testosterone, cortisone, and progesterone. An initial step in steroid hormone metabolism is the reduction of the ⌬ 4 -ene, which in humans is mediated by steroid 5␣-reductases (SRD5A1, SRD5A2) or steroid 5␤-reductase (AKR1D1) 3 to yield the corresponding 5␣-or 5␤-dihydrosteroids, respectively (1,2). The products of these reactions are not always inactive. 5␣-Reductase is responsible for the conversion of testosterone to 5␣-dihydrotestosterone (5␣-DHT), which is the most potent natural ligand for the androgen receptor. By contrast, in addition to being involved in bile acid biosynthesis, 5␤-reductase is responsible for generating 5␤-pregnanes, which are natural ligands for the pregnane-X receptor (PXR) in the liver (3,4). PXR is involved in the induction of CYP3A4, which is responsible for the metabolism of a large proportion of drugs (5,6). Thus both 5␣-reductase and 5␤-reductase are involved in the formation of potent ligands for nuclear receptors.
Finasteride is a selective 5␣-reductase type 2 inhibitor that reduces plasma 5␣-dihydrotestosterone levels and shrinks the size of the prostate (7). It is a widely used therapeutic agent in the treatment of benign prostatic hyperplasia (8,9), it is used in androgen deprivation therapy to treat prostate cancer (10), and it has been examined as a chemopreventive agent for hormonedependent prostate cancer (11). Finasteride was originally thought to act as a competitive inhibitor with nanomolar affinity for 5␣-reductase type 2 (12). More recently, it was found that finasteride acts as a mechanism-based inactivator of this enzyme (13). Subsequent to inhibitor binding, there is hydride transfer from the NADPH cofactor to the ⌬ 1-2 -ene double bond of finasteride. The intermediate enolate tautomerizes at the enzyme active site to form a bisubstrate analogue in which dihydrofinasteride is covalently bound to NADP ϩ (13). The bisubstrate analogue has subnanomolar affinity for 5␣-reductase type 2 (Fig. 1). No structural information exists for 5␣-reductase type 1 or type 2; therefore, it is not possible to determine how finasteride would bind to the active site of a human steroid double bond reductase in the absence of an experimentally determined crystal structure.
As part of an ongoing inhibitor screen of AKR1D1, we now report that finasteride acts as a competitive inhibitor with low * This work was supported, in whole or in part, by National Institutes of Health Grants R01-DK47015 and P30-ES013508 (to T. M. P.) and R01-GM56838 (to D. W. C. Expression of Recombinant AKR1D1-Previously we reported the expression of AKR1D1 using the prokaryotic expression vectors pET16b and pET28a (15). Recombinant AKR1D1 was purified to homogeneity as described previously (15). Wild-type AKR1D1 was obtained in 56% yield and had a final specific activity of 80 nmol of testosterone reduced per minute per mg of purified enzyme under standard radiometric assay conditions.
Standard Radiometric Assay and Product Verification-The standard assay contained 2 M [4-14 C]testosterone (40,000 dpm), 8 M unlabeled testosterone, 5% acetonitrile, and 100 mM phosphate buffer (pH 6.0). Reactions were initiated by the addition of 200 M NADPH and performed at 37°C. The substrate and product of the quenched reaction were separated by TLC and quantitated by scintillation counting. The 5␤-dihydrotestosterone (5␤-DHT) product was identified by co-chromatography with the authentic standard.
Standard Spectrofluorometric Assay-Spectrofluorometric analysis of the reduction of testosterone to 5␤-DHT was performed by monitoring the decrease in NADPH emission on a fluorescence spectrophotometer F-4500 (Hitachi America, Ltd., New York, NY). Assays contained 10 M testosterone, 15 M NADPH, 4% acetonitrile in 100 mM potassium phosphate buffer (pH 6.0) in a final volume of 1 ml. Reactions were initiated by the addition of enzyme and monitored at 37°C. NADPH emission was monitored by using an excitation wavelength of 340 nm (slit width 5 nm) and an emission wavelength set at 460 nm (slit width 10 nm). A standard curve was performed daily to convert the arbitrary fluorescence units measured at 460 nm to NADPH concentration. The curve was used to calculate the amount of NADPH oxidized per minute in the assays. The specific activity observed was in agreement with the specific activity measured radiometrically and gave a value of 80 nmol of testosterone reduced per minute per mg of purified enzyme.
Enzyme Inhibition Studies-Enzyme inhibition studies were performed using the standard radiometric assay to monitor the reduction of 2.0 M [ 14 C]testosterone in the absence or presence of increasing finasteride concentrations at 37°C. The K i value was determined by producing a family of lines using different fixed substrate concentrations and by varying the finasteride concentration. The family of lines was fit using GraFit, and the pattern was examined to determine the mode of inhibition, i.e. competitive, noncompetitive, and uncompetitive. Time-dependent incubation studies were performed with finasteride in which AKR1D1 (7.8 M) was preincubated with finasteride (31.0 M) in the presence and absence of NADPH (144 M) in 100 mM buffer (pH 6.0) at 37°C. Aliquots were removed from the incubation over time and directly diluted (100-fold dilution) into the standard spectrofluorometric assay, and the amount of enzyme activity remaining was determined.
Crystallography-The AKR1D1⅐NADP ϩ ⅐finasteride complex was crystallized by the hanging drop vapor diffusion method at 4°C. Typically, a drop containing 3.0 l of enzyme solution (5.0 mg/ml AKR1D1, 2.0 mM NADP ϩ , 0.5 mM finasteride, 10.0 mM Tris (pH 7.4)) and 4.0 l of precipitant buffer (0.1 M Tris-HCl (pH 7.0), 10 -20% (w/v) polyethylene glycol 4000, 10% isopropyl alcohol) was equilibrated against a 1-ml reservoir of precipitant buffer. Crystals of the AKR1D1⅐ NADP ϩ ⅐finasteride complex appeared within 1 week and were then soaked for 24 h in the same mother liquor solution augmented with 2.0 mM NADP ϩ , 2.0 mM finasteride, and 30% isopropyl alcohol. Following transfer to a 32% Jeffamine solution and flash-cooling, these crystals yielded diffraction data to 1.70 Å resolution at the National Synchrotron Light Source (beamline X6A, ϭ 1.00 Å, 100 K). Diffraction intensities measured from these crystals indicated that they belonged to the space group P2 1 2 1 2 1 with unit cell parameters similar to those previously measured from crystals of AKR1D1 complexed with different steroids (15). Two monomers occupy the asymmetric unit in this crystal form. Data reduction was achieved using the programs HKL2000 and SCALEPACK (19). Data collection and reduction statistics are reported in Table 1.
The structure of the AKR1D1⅐NADP ϩ ⅐finasteride complex was solved by the difference Fourier method using the model of the AKR1D1⅐NADP ϩ ⅐cortisone complex less ligand atoms and water molecules (PDB: 3CMF) as a starting model. The programs CNS (20) and PHENIX.refine (21) were used to refine the model, and the graphics program Coot (22) was used for map fitting. In the final stage of refinement, the atomic coordinates of finasteride were retrieved from entry WOLXEA (23) in the  Table 1.

RESULTS
Inhibition Studies-Finasteride, which is a mechanism-based inhibitor of 5␣-reductase with subnanomolar affinity, is a competitive inhibitor of recombinant AKR1D1 with a K i value of 2.1 M when AKR1D1 is saturated with NADPH and testosterone is used as a varied substrate (Fig. 2). Knowing that finasteride is a mechanism-based inactivator of 5␣-reductase type 2 and that the mechanism of inactivation has an obligatory requirement for NADPH, we conducted preincubation studies of AKR1D1 with finasteride in the presence and absence of added NADPH over 140 min. Upon dilution of aliquots into a standard AKR1D1 activity assay, we obtained no evidence for time-dependent inactivation of the enzyme.
Structure of the AKR1D1⅐NADP ϩ ⅐Finasteride Complex-Finasteride is bound with full occupancy to monomer B in the crystal structure of the AKR1D1⅐NADP ϩ ⅐finasteride complex.
The structure of this complex is very similar to the crystal structures of the ternary AKR1D1 complexes containing NADP ϩ and either cortisone or progesterone, with root mean square deviations of ϳ0.20 Å for 324 C␣ atoms (15). The electron density envelope of finasteride is clear and unambiguous (Fig.  3a), showing that the steroid binds perpendicular to the NADP ϩ cofactor.
The side chain of Trp-230 plays a key role in packing against the ␤-face of finasteride, which is oriented toward the re-face of the nicotinamide ring of NADP ϩ such that the steroid C1-C2 double bond is too far away from the 4-pro-R-hydride of NADPH for double bond reduction to occur (Fig. 3a). The C3 carbonyl oxygen of finasteride accepts hydrogen bonds from the phenolic hydroxyl group of Tyr-58 and the anti-oriented conformer of the carboxylic acid side chain of Glu-120. Additionally, the N4 group and the C20 carbonyl oxygen group of finasteride engage in hydrogen-bond interactions with water molecules (Fig. 3a). Interestingly, although the binding mode of finasteride is similar to that of progesterone, finasteride is rotated by ϳ12°with respect to progesterone (Fig. 3b), and the segment Arg-226 -Val-231 (which contains Trp-230) in loop B moves ϳ1 Å away to accommodate finasteride binding in comparison with progesterone binding.
In monomer A, continuous but uninterpretable electron density is observed, possibly indicating disordered or low occupancy binding of finasteride (supplemental Fig. S1). Accordingly, we have not built finasteride into this density in the final model. Finally, in both monomers, the backbone conformation of Thr-224 adopts a disallowed conformation on the Ramachandran plot. Notably, the electron density corresponding to Thr-224 is clear and continuous (data not shown). Because Thr-224 is adjacent to the substrate and cofactor binding sites (15), it is possible that its unfavorable backbone conformation is accommodated to optimize enzyme function. Herzberg and Moult (25) observe that regions of such steric strain in refined enzyme structures are located overwhelmingly in or near active  sites, concluding that the precision required for substrate binding and catalysis occasionally outweighs the requirement for optimal protein folding. Reduction of ⌬ 1,4 -Dienes-To validate that AKR1D1 is unable to reduce a ⌬ 1-2 -ene double bond, 1,4-androstadien-17␤-ol-3-one was examined as a substrate; co-chromatography with authentic standards was used to identify the product of the reaction. The reaction has three possible product outcomes: testosterone from reduction of the ⌬ 1-2 -ene; 5␤-androst-1-en-17␤-ol-3-one from reduction of the ⌬ 4-5 -ene; or 5␤-DHT from reduction of the ⌬ 1,4 -diene. The product does not co-migrate with either testosterone or 5␤-DHT, suggesting that the enzyme is unable to reduce the ⌬ 1-2 -ene. The k cat and K m for this reaction are 1.8 min Ϫ1 and 3.2 M, respectively.

DISCUSSION
AKR1D1 is inhibited by the 5␣-reductase type 2 selective inhibitor finasteride by acting solely as a reversible competitive inhibitor. The K i value of 2.1 M is considerably higher than that observed for either the competitive inhibition or time-dependent inactivation of 5␣-reductase type 2. By contrast, the K i value of 2.1 M is less than an order of magnitude greater than that observed for the inhibition of 5␣-reductase type 1 (K i value of 300 nM) (26). Both AKR1D1 and 5␣-reductase type 1 play important roles in the hepatic clearance of steroid hormones, suggesting that high dose finasteride may have an adverse effect on hepatic steroid metabolism. Inhibition of AKR1D1 by high dose finasteride would also deprive PXR of its natural 5␤-pregnane ligands, resulting in diminished CYP3A4 induction. This could result in significant drug-drug interactions. Importantly, finasteride itself is metabolized by CYP3A4, suggesting that high dose finasteride might prevent its own metabolism (27).
Finasteride does not act as a mechanism-based inactivator of AKR1D1, in contrast with its mode of inhibition against 5␣-reductase type 2. Although both AKR1D1 and 5␣-reductase type 2 reduce the 4-5 double bond of ⌬ 4 -3-oxosteroids in a similar manner, i.e. hydride transfer to C5 followed by enolization, the structure of the AKR1D1⅐ NADP ϩ ⅐finasteride complex (Fig.  3a) shows that it is not possible to reduce the ⌬ 1-2 -ene of finasteride because this double bond is not proximal to the NADPH cofactor. The similarity in the binding orientations of finasteride and progesterone to AKR1D1, however, account for the competitive inhibition of 5␤-reductase by the drug. It is noteworthy that finasteride forms hydrogen bonds with Tyr-58 and Glu-120, which are part of the catalytic tetrad of AKR1D1.The inability to reduce a ⌬ 1-2 -ene is validated by using 1,4-androstadien-17␤-ol-3-one as a substrate.
No structural information currently exists for 5␣-reductase type 2. However, based on the structure of the AKR1D1⅐ NADP ϩ ⅐finasteride complex, it is anticipated that the cofactor must be aligned for ⌬ 1-2 -ene reduction. It will also be intriguing to see whether there is conservation in catalytic mechanism. We have previously proposed with the aid of site-directed mutagenesis that double bond reduction requires both a catalytic tyrosine (Tyr-58) and a superacidic glutamic acid residue (Glu-120) (15,18). Future x-ray crystallographic studies will determine whether such a mechanism operates for 5␣-reductase type 1 and type 2.