Conversion of Human Steroid 5β-Reductase (AKR1D1) into 3β-Hydroxysteroid Dehydrogenase by Single Point Mutation E120H

Background: The aldo-keto reductase (AKR) subfamily AKR1D comprises steroid 5β-reductases, whereas the AKR1C subfamily comprises hydroxysteroid dehydrogenases. Results: A single E120H mutation in AKR1D1 converts human steroid 5β-reductase into a 3β-hydroxysteroid dehydrogenase. Conclusion: Glu120 controls the positional specificity of hydride transfer and facilitates double bond reduction. Significance: This is an example of a perfect change-of-function that can be achieved by a single point mutation. Human aldo-keto reductase 1D1 (AKR1D1) and AKR1C enzymes are essential for bile acid biosynthesis and steroid hormone metabolism. AKR1D1 catalyzes the 5β-reduction of Δ4-3-ketosteroids, whereas AKR1C enzymes are hydroxysteroid dehydrogenases (HSDs). These enzymes share high sequence identity and catalyze 4-pro-(R)-hydride transfer from NADPH to an electrophilic carbon but differ in that one residue in the conserved AKR catalytic tetrad, His120 (AKR1D1 numbering), is substituted by a glutamate in AKR1D1. We find that the AKR1D1 E120H mutant abolishes 5β-reductase activity and introduces HSD activity. However, the E120H mutant unexpectedly favors dihydrosteroids with the 5α-configuration and, unlike most of the AKR1C enzymes, shows a dominant stereochemical preference to act as a 3β-HSD as opposed to a 3α-HSD. The catalytic efficiency achieved for 3β-HSD activity is higher than that observed for any AKR to date. High resolution crystal structures of the E120H mutant in complex with epiandrosterone, 5β-dihydrotestosterone, and Δ4-androstene-3,17-dione elucidated the structural basis for this functional change. The glutamate-histidine substitution prevents a 3-ketosteroid from penetrating the active site so that hydride transfer is directed toward the C3 carbonyl group rather than the Δ4-double bond and confers 3β-HSD activity on the 5β-reductase. Structures indicate that stereospecificity of HSD activity is achieved because the steroid flips over to present its α-face to the A-face of NADPH. This is in contrast to the AKR1C enzymes, which can invert stereochemistry when the steroid swings across the binding pocket. These studies show how a single point mutation in AKR1D1 can introduce HSD activity with unexpected configurational and stereochemical preference.

Two subfamilies of NAD(P)(H)-dependent steroid transforming aldo-keto reductase (AKR) 2 exist in humans: 5␤-reductase (AKR1D1) and hydroxysteroid dehydrogenases (HSDs; also known as AKR1C1-1C4). AKR1D1 catalyzes the irreversible reduction of ⌬ 4 -3-ketosteroids to 5␤-dihydrosteroids to introduce a 90°bend at the steroid A/B ring junction (1). This cis-configuration is critical for bile acids to properly emulsify dietary fats and cholesterol (2). 5␤-Reduced steroid hormones may be involved in parturition (3), porphyrinogenesis (4,5), and xenoprotection (6,7). By contrast, AKR1C enzymes catalyze the interconversion between 3-, 17-, and 20-ketosteroids and 3␣/␤-, 17␤-, and 20␣-hydroxysteroids (Scheme 1) (8). By functioning as HSDs, AKR1C enzymes regulate the ratio of active and inactive androgens, estrogens, and progestogens that can bind to nuclear receptors in target tissues (9). AKR1D1 and AKR1C enzymes are highly homologous enzymes, sharing over 50% sequence identity and a common (␣/␤) 8 barrel fold (10 -12). Both subfamilies of enzymes bind the steroid perpendicularly to the NAD(P)(H) cofactor in the active site so that either the steroid A or D ring faces the nicotinamide ring. During reduction, a hydride from the 4-pro-(R)position on the nicotinamide ring is stereospecifically transferred to the recipient double bond/carbonyl group on the steroid substrate (13). AKR1C enzymes perform the hydride acid group of Glu 120 is presumed to adopt the fully protonated anti-oriented conformation in order to donate a hydrogen bond directly to the steroid carbonyl oxygen. This arrangement together with the general acid Tyr 58 creates a superacidic environment that facilitates hydride transfer by activating the ⌬ 4 -double bond and stabilizing the enolate intermediate (Scheme 2, right) (10).
Despite the homology between AKR1D1 and the AKR1C enzymes, no mixed activity has been observed between the two subfamilies. The glutamate-histidine substitution appears to be the key mutation that distinguishes 5␤-reduction from HSD activity. To examine the role of Glu 120 further, we introduced the E120H mutation in AKR1D1. Our kinetic characterization of the mutant demonstrates that the E120H mutation completely abolishes the 5␤-reductase activity and converts AKR1D1 into a highly efficient 3␤-HSD that exhibits opposite stereopreference to the majority of AKR1C enzymes, which show preference for 3␣-HSD activity. Analysis of the crystal structures of the E120H mutant in complex with epiandrosterone (epiA; 3␤-hydroxy-5␣-androstan-17-one), 5␤-dihydrotestosterone (5␤-DHT), and ⌬ 4 -androstene-3,17-dione (⌬ 4 -Adione) reveals that the bulky histidine imidazole ring introduced by the E120H substitution prevents penetration of the steroid into the active site so that hydride transfer is now only possible to the C3-position for ketosteroid reduction. The binding conformations of different steroid isomers also illustrate the mechanism by which HSD stereospecificity is controlled.
Mutagenesis, Expression, and Purification of Recombinant AKR1D1 E120H Mutant-The expression plasmid for the AKR1D1 E120H mutant was made using the pET28a-AKR1D1 construct (10) as a template by conducting site-directed mutagenesis using the QuikChange method (Stratagene) following the manufacturer's protocol . The forward primer was  5Ј-dGGA TCT TTA CAT CAT TCA CGT ACC AAT GGC  CTT TAA GCC-3Ј, and the reverse primer was 5Ј-dGGC TTA  AAG GCC ATT GGT ACG TGA ATG ATG TAA AGA TCC-3Ј. The mutation was verified by dideoxysequencing. The mutant enzyme was overexpressed in Escherichia coli C41(DE3) cells in LB-Miller medium (BD Diagnostics) by the addition of isopropyl 1-thio-␤-D-galactopyranoside (1 mM final concentration) at inoculation. The cells were grown for 20 h at 37°C, harvested by centrifugation, and resuspended in buffer A (20 mM Tris (pH 8.6)) supplemented with 1 mg/ml lysozyme and 1 g/ml DNase I (Roche Applied Science). The cells were lysed by sonication, and the protein in the soluble fraction was purified using a DEAE-Sepharose column with a linear salt gradient using buffer A and buffer B (20 mM Tris (pH 8.6), 250 mM NaCl). The mutant enzyme eluted at around 60 mM NaCl. Fractions containing the mutant were identified by SDS-PAGE analysis and pooled, and the final imidazole concentration was adjusted to 20 mM using buffer C (20 mM Tris (pH 7.9), 400 mM imidazole, 500 mM NaCl). The pooled fractions were then applied to a HisTrap FF column (5 ml), and the enzyme was purified by stepwise elution using 50% buffer C and 50% buffer D (20 mM Tris (pH 7.9), 20 mM imidazole, 500 mM NaCl). The purity of the protein was assessed by SDS-PAGE, and the protein concentration was determined by a Bradford protein assay using bovine serum albumin as a standard. The reductase activity of the pooled fractions after each chromatographic step was measured by the fluorimetric assay described under "Kinetic Analysis" using 1 M 5␣-DHT and 7.5 M NADPH as substrates. The mutant was obtained in a 65% yield. C]progesterone (55 mCi/ mmol), and 180 g of AKR1C1 in a final volume of 0.5 ml. The reaction was incubated at 37°C for 2 h with ϳ75% conversion of progesterone. [ 14 C]20␣-OHP required further purification from the unreacted progesterone by chromatography (silica 60, CH 2 Cl 2 /ether, 60:5 (v/v)). After incubation, all reactions were quenched and extracted with ethyl acetate. The purity of the synthesized steroids was assessed by TLC as described below. The specific radioactivity of the synthesized substrates was based on the value for the starting material.
Product Identification by TLC-The 5␤-reductase activity of the AKR1D1 E120H mutant was examined using testosterone, progesterone, and ⌬ 4 -Adione as substrates with NADPH as cofactor. The ketosteroid reductase activity of the E120H mutant was examined using 5␣-DHT, 5␤-DHT, androsterone, ⌬ 4 -Adione, and progesterone as substrates and NADPH as cofactor. The reactions were performed in 200-l systems containing 100 mM potassium phosphate buffer (pH 6.0), 4% acetonitrile, 200 M NADPH, 2.8 g of AKR1D1 E120H, and one of the following substrates: 5 M The hydroxysteroid oxidase activity of the E120H mutant was evaluated using androsterone, testosterone, 5␣-DHT, 5␤-DHT, 17␤-estradiol, and 20␣-OHP as substrates, using NAD(P) ϩ as cofactor. The reactions were performed in 200-l systems containing 100 mM potassium phosphate buffer (pH 7.0), 4% acetonitrile, 2. All reactions were incubated at 37°C and quenched by 1 ml of ethyl acetate at different time intervals. Control reactions that contained the same components as the reaction mixture less enzyme were performed for each substrate using a 60-min incubation. The aqueous phase was extracted with ethyl acetate, and the extracts were applied to an LK6D Silica 60 TLC plate (Whatman), and the radioactive peaks were detected by a TLC-linear analyzer (Bioscan Imaging Scanner system 200-IBM with AutoChanger 3000; Bioscan). Products of the reactions were identified by co-chromatography with authentic synthetic standards. The chromatograms for reactions containing 5␣-DHT, ⌬ 4 -androstene-3,17-dione, androsterone, or 17␤estradiol were developed three times in CH 2 Cl 2 /ether (110:10, v/v). The chromatograms for reactions containing testosterone or 5␤-DHT were developed twice in toluene/acetone (80:20, v/v). The chromatograms for reactions containing progesterone or 20␣-OHP were developed once in CH 2 Cl 2 /ethyl acetate (80:20, v/v).
Kinetic Analysis-Reductase activity was monitored fluorimetrically by measuring the disappearance of the NADPH signal ( ex ϭ 340 nm, em ϭ 460 nm) in 1-ml systems containing 100 mM potassium phosphate buffer (pH 6.0), 4% acetonitrile, 7.2 M NADPH, and varied concentrations of steroid (0.2-60 M) at 37°C. The oxidase activity was monitored fluorimetrically by measuring the appearance of NADH signal ( ex ϭ 340 nm, em ϭ 460 nm) in 1-ml systems containing 100 mM potassium phosphate buffer (pH 7.0), 4% acetonitrile, 2.3 mM NAD ϩ , and varied concentrations of steroid (0.2-60 M) at 37°C. Reactions were initiated by the addition of enzyme and were corrected for the non-enzymatic rate. Kinetic constants were calculated by fitting the initial velocity data to either the Michaelis-Menten equation or the modified Michaelis-Menten equation with a term for substrate inhibition (Equation 1) using GraFit (Erithacus Software Ltd.). The iterative fits gave estimates of the steady-state kinetic parameters, mean Crystallography-Four different complexes were obtained for the E120H mutant (E120H⅐NADP ϩ , E120H⅐NADP ϩ ⅐epiA, E120H⅐NADP ϩ ⅐⌬ 4 -Adione, and E120H⅐NADP ϩ ⅐5␤-DHT) using the hanging drop vapor diffusion method at 4°C. Besides the steroid, drops typically contained 3.0 l of protein solution (10 mg/ml AKR1D1 E120H mutant, 2.0 mM NADP ϩ , and 10 mM Tris-HCl (pH 7.4)) and 3.0 l of well solution (0.1 M Tris-HCl (pH 7.0 -7.7), 14 -20% (w/v) polyethylene glycol 4000, and 10% isopropyl alcohol). Steroid solutions of 100 mM were prepared in isopropyl alcohol and added to the protein solution in co-crystallization trials or to the soaking solution to obtain the desired steroid concentration without adjustment of the final isopropyl alcohol concentration to a constant value. Crystals appeared and grew to a suitable size for diffraction in approximately a week. The ternary complex of E120H⅐NADP ϩ ⅐epiA was obtained by co-crystallization in the presence of 2 mM 5␣-DHT, and the E120H⅐NADP ϩ ⅐⌬ 4 -Adione complex was obtained in the presence of 2 mM testosterone. The steroids in the structures were different from those used in co-crystallization due to substrate turnover (see "Results"). The E120H⅐NADP ϩ ⅐5␤-DHT complex was obtained by soaking the binary complex E120H⅐NADP ϩ in the mother liquor augmented with 5 mM NADP ϩ and 5 mM 5␤-DHT for 6 h. Prior to flash-cooling, crystals were soaked for 5 min in a 16% (w/v) Jeffamine ED-2001 cryoprotective solution containing 2 mM NADP ϩ and 2 mM steroid to maintain full occupancy of the steroid ligand. Diffraction data of the E120H⅐NADP ϩ , E120H⅐NADP ϩ ⅐⌬ 4 -Adione, and the E120H⅐NADP ϩ ⅐5␤-DHT complexes were collected at beamline 24-ID-C ( ϭ 0.979 Å) of the Advanced Photon Source at Argonne National Laboratory (Argonne, IL). The data of the E120H⅐NADP ϩ ⅐epiA complex were collected at beamline X29 ( ϭ 1.075 Å) of the National Synchrotron Light Source at Brookhaven National Laboratory (Upton, NY). All crystals are of the P2 1 2 1 2 1 space group with unit cell parameters a ϭ 50.03, b ϭ 109.71, and c ϭ 129.34 Å, similar to those of the WT AKR1D1 (10). The asymmetric unit of the unit cell contains two monomers of the E120H mutant. Data were integrated and scaled with HKL2000 and Scalepack (18). Data collection and reduction statistics are reported in Table 1 for all complexes.
The structure of the E120H⅐NADP ϩ complex was solved by molecular replacement performed with PHASER (19) from the CCP4 suite using the coordinates of the WT AKR1D1⅐NADP ϩ binary complex (PDB code 3BV7) (10) less ligand and solvent molecules as a search model. The structures of ternary complexes E120H⅐NADP ϩ ⅐epiA, E120H⅐NADP ϩ ⅐⌬ 4 -Adione, and E120H⅐NADP ϩ ⅐5␤-DHT were solved by the difference Fourier method using the E120H⅐NADP ϩ complex as a starting model.
The programs CNS (20), PHENIX (21), and COOT (22) were used for refinement and model fitting. NADP ϩ and steroids were built into the electron density map at the final stage of refinement. The quality of the models was verified with Molprobity and PROCHECK. PROCHECK identified residue Thr 224 in both monomers in all of the AKR1D1 E120H complexes in the disallowed region. The refinement statistics are reported in Table 1.

RESULTS
Loss of 5␤-Reductase Activity-The E120H mutation effectively abolished the 5␤-reductase activity found in WT AKR1D1. No 5␤-reduced product was observed with the AKR1D1 E120H mutant when [ 14 C]testosterone, [ 14 C]progesterone, or [ 14 C]⌬ 4 -Adione was employed as the substrate in the presence of NADPH after a 1-h-long incubation at 37°C. The limit of detection of the assay was 0.2 pmol/min, and we estimated that if any residual activity remained, it was less than 0.8% of wild-type enzyme.
3␤-Hydroxysteroid Dehydrogenase Activity of AKR1D1 E120H Mutant-The ability of the E120H mutant to act as an HSD was examined for the reduction of 3-, 17-, and 20-ketosteroids and the oxidation of the corresponding hydroxysteroids (Scheme 3). Enzyme activity was monitored by measuring the change of fluorescence emission of the cofactor to generate steady-state kinetic parameters. Concurrently, reactions were performed with radiolabeled steroid substrates so that product profiles of the reactions could be ascertained. Product identity was achieved by TLC by measuring co-migration of the radiolabeled peaks with authentic synthetic standards.
The E120H mutant was expected to function as a 3␣-HSD like the AKR1C enzymes but to maintain a preference for 3-ketosteroids with a 5␤-configuration because these are produced by WT AKR1D1. To our surprise, the E120H mutant acted predominantly as a 3␤-HSD that preferred steroids with a 5␣-configuration over those with a 5␤-configuration. In ketosteroid reduction, the mutant stereospecifically reduced 5␣-DHT to 5␣-androstane-3␤,17␤-diol ( Fig. 1) with a catalytic efficiency of 2.1 ϫ 10 5 M Ϫ1 s Ϫ1 , which was 10-fold higher than that for the reduction of 5␤-DHT, which has the opposite configuration (Table 2). For hydroxysteroid oxidation, the mutant oxidized the 3␤-hydroxysteroid, epiA, to 5␣-androstane-3,17dione with a catalytic efficiency of 1.3 ϫ 10 5 M Ϫ1 s Ϫ1 , which was at least 15-fold higher than the catalytic efficiency of the other tested hydroxysteroids regardless of the configuration of the steroid A/B ring junction and the position of catalysis (3-, 17-, or 20-position) ( Table 3).
Comparison of AKR1D1 E120H Mutant with AKR1C Enzymes-AKR1C enzymes perform ketosteroid reduction and exhibit similar catalytic efficiencies for the reduction of 5␣and 5␤-DHT and preferably form the 3␣-hydroxysteroid over the 3␤-hydroxysteroid (23,24). The exception is AKR1C1, which reduces 5␣-DHT to a mixture of 3␣-and 3␤-isomers in a 1:3 ratio. The AKR1D1 E120H mutant exhibited superior specificity and efficiency for the reduction of 5␣-DHT to yield 5␣-androstane-3␤,17␤-diol than observed with AKR1C1 ( Table 2). The E120H mutant produced only the 3␤-isomer and had a catalytic efficiency for 5␣-DHT that was 400-fold greater than that observed for the same reduction reaction catalyzed by AKR1C1. In addition, the k cat /K m value for the reduction of 5␣-DHT catalyzed by the E120H mutant was about 15-fold greater than that observed for the reduction of this steroid to the opposite stereoisomer, 5␣-androstane-3␣,17␤-diol, catalyzed by AKR1C2. The values obtained with the mutant were similar to that observed for the reduction of 5␣-DHT to 5␣-androstane-3␣,17␤-diol catalyzed by AKR1C4, the most efficient human AKR1C enzyme (23).
In the oxidation direction, the E120H mutant slowly converted 3-hydroxysteroids with a 3␣,5␣-configuration (supplemental Fig. S3) or a 3␣,5␤-or 3␤,5␤-configuration to the corresponding 3-ketosteroids with similar catalytic efficiencies. Their catalytic efficiencies were much lower than that of the oxidation of epiA, which contains the preferred 3␤,5␣configuration ( Table 3). The E120H mutant also oxidized 17␤hydroxysteroids and converted 5␣-DHT to 5␣-androstane-3,17-dione, 17␤-estradiol to estrone, testosterone to ⌬ 4 -Adione, and 5␤-DHT to 5␤-androstane-3,17-dione (supplemental Fig. S4 and Table 3). Oxidation of 5␣-DHT at the 17␤-position exhibited slightly higher catalytic efficiency than that observed for the oxidation of the 3-hydroxysteroids except for epiA. The oxidation of the 17␤-position of 5␣-DHT was also preferred over the oxidation of the 17␤-position of 5␤-DHT and gave a catalytic efficiency that was 50-fold higher. The E120H mutant also catalyzed a very slow oxidation of 20␣-hydroxyprogesterone to progesterone, but accurate kinetic parameters were not obtained due to the slow reaction rate and substrate inhibition (supplemental Fig. S5 and Table 3). The catalytic efficiencies of these minor oxidative activities on the 3␣/␤-, 17␤-, and 20␣-hydroxysteroids were 20 -800-fold lower than that obtained for the oxidation of epiA catalyzed by the 3␤-HSD activity of the E120H mutant.
Overview of Structure of E120H Mutant-We determined the crystal structures of the E120H mutant in complex with NADP ϩ , NADP ϩ and epiA, NADP ϩ and ⌬ 4 -Adione, and NADP ϩ and 5␤-DHT to develop a structural explanation for the observed change of function. Each complex contained two monomers of the E120H mutant in the asymmetric unit.   (42). Thr 224 in both monomers of all the AKR1D1 E120H complexes was found in the disallowed region.
Leu 302 -Tyr 326 ), although no significant conformational differences were evident between the two monomers. This finding is consistent with the reported structure of WT AKR1D1 (10). The E120H mutation did not perturb the overall structure of the protein. The root mean square deviations between the AKR1D1⅐NADP ϩ ⅐cortisone structure (PDB code 3CMF) and the structures of the four complexes of the E120H mutant were within 0.18 Å for the 325 residues in monomer B and were similar to the root mean square deviations when the individual mutant complexes were compared (0.09 -0.12 Å). When the AKR1D1⅐NADP ϩ ⅐HEPES complex structure is compared with the AKR1D1 E120H⅐NADP ϩ complex structure, the similarity in the structures is apparent (Fig.  2A). The His 120 residue in the E120H mutant adopts the same conformation as the Glu 120 residue in WT AKR1D1. The two residues were essentially superimposed with each other at their backbones and side chains up to the C-␥ atom (Fig. 2B). Cofactor binding was undisturbed by the mutation, as expected, due to the lack of direct interaction between the cofactor and Glu 120 . NADP ϩ maintained the same hydrogen bond network as well as the salt link between Arg 279 and the 2Ј-phosphate of AMP moiety that is responsible for the tight binding of NADPH (10,25). The positioning of the nicotinamide ring of the cofactor in the central active site cavity retained the same location as found in WT AKR1D1.
E120H⅐NADP ϩ Complex-The E120H⅐NADP ϩ complex was obtained by co-crystallization with NADP ϩ . In the binary complex, a water molecule occupied the oxyanion site through hydrogen-bonding to Tyr 58 and His 120 (Fig. 2B). In the absence of a steroid ligand, Trp 230 swung into the steroid binding channel, triggering a slight movement of Loop B.
E120H⅐NADP ϩ ⅐epiA Complex-The E120H⅐NADP ϩ ⅐epiA complex was obtained by co-crystallization with 5␣-DHT. Simulated annealing omit maps indicated that two androgens distinct from 5␣-DHT were bound in the two monomers of the complex. EpiA agreed well with the electron density map in monomer B with lower average B-factors (Fig. 3A), whereas 5␣-androstane-3␤,17␤-diol best depicted the map in monomer

Engineering 3␤-HSD into Steroid 5␤-Reductase
A (supplemental Fig. S6). An assay mimicking the crystallization condition was performed and showed that 5␣-DHT oxidoreduction occurred during the crystallization process and produced a mixture of products that included epiA, 5␣-androstane-3␤,17␤-diol, and 5␣-androstane-3,17-dione (supplemental Fig. S4). The electron density map of the ligand identified the steroid that occupied the active site most frequently. Despite the difference in functional groups at the C17-position, the two 3␤,5␣-tetrahydrosteroids occupied identical positions in the steroid binding channel. The structure of monomer B with epiA was selected to represent the complex. EpiA was oriented with the A-ring toward the active site center, placing the 3␤-hydroxy group in the oxyanion site through interactions with Tyr 58 and His 120 . Trp 230 receded from the steroid channel, packing against the ␣-face of the steroid. The distance of 3.2 Å between the C3-position of the steroid and the C4 atom on the nicotinamide ring was within a suitable distance for hydride transfer. The steroid ␣-face was directed toward the re-face of the cofactor, which allowed oxidoreduction of the 3␤,5␣-reduced hydroxysteroid, consistent with the observed stereopreference of the E120H mutant. Compared with WT AKR1D1, the E120H mutant bound steroids with an opposite facial orientation, and the oxyanion site in the mutant was shifted "upward" by 1.5 Å (Fig. 3B). This movement placed the C3-position of the steroid in the E120H mutant at a position that could be superimposed on the oxyanion site in the AKR1C enzymes, illus-trated by the position of the 3-ketone group in testosterone in the AKR1C9⅐NADP ϩ ⅐testosterone complex (PDB code 1ASF) (Fig. 3C).
E120H⅐NADP ϩ ⅐⌬ 4 -Adione Complex-The E120H⅐NADP ϩ ⅐⌬ 4 -Adione was obtained by co-crystallization with testosterone. Akin to the phenomenon observed with the E120H⅐ NADP ϩ ⅐epiA complex, oxidation of the 17␤-hydroxy group of testosterone was observed. ⌬ 4 -Adione was found in monomer B of the complex with lower average B-factors (Fig. 4A), whereas testosterone was found in monomer A (supplemental Fig. S7). Both ⌬ 4 -3-ketosteroids occupied identical positions in the steroid binding channel. ⌬ 4 -Adione bound to the steroid binding channel with its A-ring, rather than the D-ring, pointing toward the active site, placing the 3-ketone group at the oxyanion site in a nonproductive orientation. Despite the structural resemblance between ⌬ 4 -Adione and epiA, the binding orientation of ⌬ 4 -Adione resembles 5␤-DHT by orientating its ␤-face toward the cofactor (i.e. the steroid face has been flipped). In the WT AKR1D1⅐NADP ϩ ⅐testosterone structure, testosterone was bound perpendicular to the steroid binding channel in an alternative binding pocket, which is responsible for the reported substrate inhibition of WT AKR1D1 (Fig. 4B) (26). In the E120H mutant structure, testosterone was bound in the steroid binding channel, although the mutation did not disrupt the alternative binding pocket. Substrate inhibition observed for testosterone oxidation catalyzed by the mutant could be elicited either by steroid fitting into the alternative binding pocket as in the wild type AKR1D1 or binding nonproductively in the steroid channel so that its A-ring is in proximity with the cofactor.
E120H⅐NADP ϩ ⅐5␤-DHT Complex-The E120H⅐NADP ϩ ⅐ 5␤-DHT complex was obtained by soaking 5␤-DHT into the E120H⅐NADP ϩ complex. Although 5␤-DHT is a substrate for the E120H mutant, no turnover was observed due to its low catalytic efficiency and relatively short incubation time with the enzyme. The steroid nucleus of 5␤-DHT occupied a similar position in the steroid binding channel as epiA with the A-ring approaching the cofactor (Fig. 5A). However, 5␤-DHT bound distally from the catalytic site and instead a water molecule was anchored to the oxyanion site. This represents a non-productive binding mode witnessed in monomer A of the AKR1D1⅐ NADP ϩ ⅐5␤-DHP complex (PDB code 3CAV) (25), which resembles a conformation that precedes steroid product release from the WT enzyme. The ␤-face of 5␤-DHT was directed toward the cofactor, retaining the same orientation utilized for steroid 5␤-reduction and coincides with the stereopreference of the mutant to produce the 3␣,5␤-reduced steroid product from 5␤-DHT. Slight movements of Trp 230 and Tyr 132 were observed to accommodate the flip of the C18 and C19 angular methyl groups. In the WT AKR1D1⅐NADP ϩ ⅐5␤-DHT structure (PDB code 3DOP) (27), the steroid was also bound through a water molecule; however, the 17␤-hydroxy group of the steroid was directed toward the oxyanion site instead of the 3-ketone group (Fig. 5B).

DISCUSSION
Glutamate-Histidine Substitution Controls 5␤-Reductase and HSD Activity-Engineered enzymes usually exhibit crippled catalytic efficiency and partially retain their original activ-

Engineering 3␤-HSD into Steroid 5␤-Reductase
ity. Here we demonstrate that a single glutamate-to-histidine mutation in AKR1D1 eliminates its 5␤-reductase activity and converts the enzyme into a proficient 3␤-HSD. The stereopreference of the E120H mutant is different from most of the AKR1C enzymes except AKR1C1 (23, 28). The introduced 3␤-HSD activity is much higher than observed with AKR1C1 and is comparable with and in some cases even higher than native 3␤-HSDs in the short-chain dehydrogenase/reductase family (29). We have previously engineered 5␤-reductase activity into rat 3␣-HSD AKR1C9 by introducing the corresponding

Engineering 3␤-HSD into Steroid 5␤-Reductase
H120E mutation (numbering according to AKR1D1) (30). Characterization and kinetic analyses on the AKR1C9 H120E mutant showed that it exhibited moderate k cat values ϳ0.005 s Ϫ1 and K m values of ϳ20 M toward four different ⌬ 4 -3-ketosteroids, with catalytic efficiencies approximately one-thirtieth of the corresponding values of purified rat liver 5␤-reductase AKR1D2 (31). The two studies complement each other and prove that the functional difference between AKR1D1 and AKR1C enzymes originates from a single glutamate-histidine substitution in the catalytic tetrad residue at position 120. Crystal structures of the AKR1D1 E120H mutant were determined to elucidate the structural basis for the change-of-function. Glu 120 probably plays a dual role in determining the activity of AKR1D1 and the AKR1C enzymes; this residue controls the positional specificity of hydride transfer and facilitates double bond reduction. Glu 120 Controls Positional Specificity of Hydride Transfer-The E120H mutation does not disturb the overall structure of AKR1D1 or the position of the cofactor (Fig. 2). The histidine backbone in the E120H mutant is placed at essentially the same position as the glutamate residue in WT AKR1D1 (Fig. 3B). However, compared with the glutamate, which "swings" its side chain away from the steroid, the bulky imidazole ring of the histidine residue in the E120H mutant protrudes toward the catalytic center, adopting an orientation that closely resembles the conformation exhibited by His 120 in the AKR1C enzymes (Fig. 3C). As a result, the steroid can no longer penetrate deep into the active site, and the oxyanion site in the mutant is shifted upward away from the cofactor to a location almost identical to the one it occupies in the AKR1C enzymes. This shift changes the recipient group for hydride transfer, from the C5-position to the C3-position of the steroid, and results in the loss of 5␤-reductase activity and the gain of 3␤-HSD activity. This phenomenon is reminiscent of what is seen in the short-chain dehydrogenase/ reductase family, in which plant Digitalis lanata progester- Atoms are colored-coded as follows. Yellow, protein carbon; black, steroid carbon; blue, nitrogen; red, oxygen; orange, phosphorus. Hydrogen bonds are indicated by red dashed lines. The steroid A-rings are shown in ball-and-stick representations. B, superposition of AKR1D1 E120H⅐NADP ϩ ⅐epiA (yellow) and WT AKR1D1⅐NADP ϩ ⅐cortisone complex (PDB code 3CMF; blue). The AKR1D1 E120H mutation allows the C3 carbonyl to reside ϳ1 Å closer to the 4-pro-(R)-hydride on the nicotinamide ring. The trajectory of hydride transfer is indicated by a green dashed line. C, superposition of AKR1D1 E120H⅐NADP ϩ ⅐ (yellow), AKR1C2⅐NADP ϩ ⅐ursodeoxycholate (PDB code 1IHI; gray), and AKR1C9⅐NADP ϩ ⅐testosterone complex (PDB code 1ASF; purple). The testosterone 3-ketone group in AKR1C9 superimposed well with the 3-hydroxy group of epiA in the E120H mutant. Residues are numbered as in AKR1D1.

Engineering 3␤-HSD into Steroid 5␤-Reductase
one 5␤-reductase corresponds to AKR1D1, and human 17␤-HSD type I corresponds to the AKR1C enzymes. In these two short-chain dehydrogenase/reductases, the change from a 5␤-reductase to a 17␤-HSD is achieved by repositioning of a catalytic tyrosine. The shift of this tyrosine alters the depth of the steroid binding pocket so that either the steroid double bond or ketone group is positioned for reduction (32). However, in the short-chain dehydrogenase/reductase family, relocation of the catalytic tyrosine is accompanied by significant sequence and structural changes, whereas in the AKR family, a single glutamate-histidine substitution easily shifts the position of the steroid without rearrangement of the catalytic site.
Catalytic Role of Glu 120 -The AKR1D1 E120A mutation allows steroid penetration in the active site but eradicates both 5␤-reductase and HSD activity. 3 The H120A mutation in AKR1C9 also exhibits diminished 5␤-reductase activity compared with the H120E mutant (30). These findings imply that the glutamate residue plays a role in the chemical mech-anism of 5␤-reduction. Studies on AKR1C9 H120E mutant show that the 5␤-reductase activity is abolished by the Y58F/ H120E double mutation but not by the H120A mutation, indicating that Glu 120 greatly facilitates the reaction but is not obligatory for catalysis (30). The pH-rate dependence of these AKR1C9 mutants confirms that Tyr 58 is the general acid for 5␤-reduction and suggests that Glu 120 lowers the pK b of the general acid by hydrogen bonding to Tyr 58 , resembling the role played by the histidine residue in AKR1C9 (Scheme 2, left) (30). However, the crystal structure of AKR1D1 disagrees with this mechanism. The distance between Glu 120 and Tyr 58 of 4.3 Å is too long for a hydrogen bond to exist between these two residues. Instead, the structure suggests that Glu 120 donates a superacidic hydrogen bond to the steroid C3 carbonyl group, which activates the ␣,␤-unsaturated steroid double bond by stabilizing the enolic intermediate during hydride transfer (Scheme 2, right) (10). The common feature of the two mechanisms is that Glu 120 aids 5␤-reduction by promoting an acidic environment. Currently, there are no kinetic data to support either mechanism directly. Characterization of the other Glu 120 mutations, including E120D, E120Q, and E120N, will probably unveil the catalytic property of this residue. A similar mechanistic problem existed for Pseudomonas ⌬ 5 -3-ketosteroid isomerase, in which Tyr 14 functions as the general acid and Asp 38 acts as the general base, but the role of a neighboring Asp 99 was uncertain. This residue could be either hydrogen-bonded directly to the steroid substrate (33,34) or to Tyr 14 to form a catalytic diad (35,36). The role of Asp 99 was only resolved via NMR studies using a series of catalytic residue single/double mutants, which showed that Asp 99 was directly hydrogen-bonded to the steroid (37).
Substrate Preference for Steroid trans-A/B Ring Fusion-AKR1C enzymes reduce 5␣-DHT and 5␤-DHT with similar catalytic efficiencies (23,24). The preference of the AKR1D1 E120H mutant to reduce a 3-ketosteroid with an A/B trans-ring configuration over a cis-configuration by more than 10-fold was unexpected. This preference mainly results from the difference in substrate binding affinity, as reflected by a 10-fold difference in K m values of 5␣-DHT and 5␤-DHT. The steroid binding channel of the E120H mutant has a cylindrical shape that packs the steroid tightly on three sides, where Tyr 132 and Trp 230 are parallel to the steroid plane, and Tyr 26 is perpendicular to the steroid plane (Fig. 3A). This shape allows easy accommodation of 5␣-DHT but not 5␤-DHT that has a 90°bend at the A/B ring junction. By contrast, using AKR1C2 as an example, Tyr 132 is replaced by a smaller residue, and the indole ring of Trp 230 is placed almost perpendicular to its corresponding position in AKR1D1 (Fig.  3C). These changes forge the steroid binding channel into a curved shape that tolerates the cis-configuration better, resulting in a similar or even slightly favorable K m for 5␤-DHT over 5␣-DHT in 3-ketosteroid reduction ( Table 2).
Stereospecificity of 3-Ketosteroid/Hydroxysteroid Oxidoreduction-The majority of the AKR1C enzymes prefer the production of 3␣-hydroxysteroids regardless of the steroid A/B-ring configuration. The AKR1D1 E120H mutant exhibits a different stereopreference. It acts as a 3␤-HSD in the presence of 5␣-DHT but produces a mixture of 3␣and 3␤-hydroxysteroids in a 4:1 ratio when 5␤-DHT is the substrate. Two mechanisms have been proposed to control the stereopreference of these HSDs, each through a unique motion of the steroid involving either swinging or flipping. 5␤-DHT adopts a non-productive binding mode in the active site, allowing a water molecule to occupy the oxyanion site. Atoms are color-coded as described in the legend to Fig. 3, except the carbons of the AKR1D1 E120H mutant are colored in blue. Hydrogen bonds are indicated by red dashed lines. The steroid A-rings are shown in ball-and-stick representations. B, superposition of E120H⅐NADP ϩ ⅐5␤-DHT (blue) and WT AKR1D1⅐NADP ϩ ⅐5␤-DHT complex (PDB code 3DOP; red). 5␤-DHT binds with its A-ring approaching the oxyanion site in the AKR1D1 E120H mutant and with its D-ring approaching the oxyanion site in the WT AKR1D1. Water molecules are shown as spheres. Only monomer B is shown. MAY 11, 2012 • VOLUME 287 • NUMBER 20

JOURNAL OF BIOLOGICAL CHEMISTRY 16619
The swinging mechanism was proposed based on docking simulations of 5␣-DHT into AKR1C1 and AKR1C2, where the two enzymes show opposite stereospecificity for 5␣-DHT reduction (23,38). The docking models suggest that the steroid swings across the steroid binding pocket to present different faces for 4-pro-(R)-hydride transfer from the cofactor to ensure different stereochemical outcomes. This swinging mechanism is achieved by the size of the residue at position 57 (AKR1D1 numbering). The longer side chain of Leu 57 in AKR1C1 pushes the steroid toward Trp 230 so that the steroid ␣-face is presented to the cofactor to produce a 3␤-hydroxysteroid, whereas the shorter side chain of Val 57 in AKR1C2 ensures that the steroid ␤-face is presented to the cofactor to produce a 3␣-hydoxysteroid. However, this mechanism cannot be applied to AKR1D1 because Ile 57 is no longer in close proximity to the steroid and is unable to play a central role in steroid orientation. Also, this swinging mechanism may not be generalized to the other AKR1C enzymes because both AKR1C3 and AKR1C4 contain the Leu 57 as in AKR1C1, yet they favor the production of the 3␣-hydroxysteroid as seen in AKR1C2.
Our structures of the E120H mutant support a mechanism in which the faces of the steroid are flipped. The structures of the E120H⅐NADP ϩ ⅐epiA and E120H⅐NADP ϩ ⅐⌬ 4 -Adione complexes show that the steroid may present either its ␣or ␤-face toward the cofactor by flipping 180°along its long molecular axis, using the stationary oxyanion site as the center of rotation (Figs. 3A and 4A). The flipping mechanism complies with the different stereopreference seen with the AKR1D1 E120H mutant and AKR1C9. The competitive inhibitor testosterone in AKR1C9 binds to the active site at almost the same position as epiA in the E120H mutant, except that the two steroids present opposite faces to the cofactor (i.e. the two steroids are flipped) (Fig. 6). Like the E120H mutant, AKR1C9 also packs testosterone on three sides, with Leu 57 and Trp 230 parallel to the steroid plane and Tyr 313 perpendicular to the steroid plane, but Tyr 313 in AKR1C9 resides on the opposite side of the steroid against Tyr 26 in the E120H mutant. The tyrosine residues cause the steroid B-ring to be flipped away from them to avoid clashing, and, as a result, the ␣-face of the steroid is directed toward the cofactor for 3␤-reduction in the E120H mutant, whereas the ␤-face of the steroid is directed toward the cofactor for 3␣-reduction in AKR1C9.
Positional Specificity of Mutant-The E120H mutant exhibits residual activity at the 17-and 20-positions of the steroid. AKR1C enzymes possess this functional plasticity as well (Scheme 1) (8). The functional plasticity derives from the symmetric nature of the steroid molecule and the plasticity of the AKR1C active site. Among the ternary complex structures solved for the human AKR1C enzymes (12, 39 -41), none of the structures depict steroids in a productive binding mode. Only in the AKR1C1⅐NADP ϩ ⅐20␣-OHP and AKR1C2⅐NADP ϩ ⅐ testosterone complexes do the steroids adopt orientations that correspond to observed positional preferences of the enzymes. The AKR1D1 E120H mutant catalyzes the 17␤-hydroxysteroid oxidation of testosterone to yield ⌬ 4 -Adione; however, the structure of the E120H⅐NADP ϩ ⅐⌬ 4 -Adione complex also fails to display a productive binding mode (Fig. 4). Instead, a reverse binding orientation with the 3-ketone group tethered to the oxyanion site is observed. The preference for the 3-ketone over the 17␤-hydroxyl group in the oxyanion site reflects the kinetic data, where the K m values for the 3-ketosteroid reduction for 5␣-and 5␤-DHT are lower than the K m values for the 17␤hydroxysteroid oxidation of the same steroids. In addition, the fact that the E120H mutant allows nonproductive binding of the substrate by reversing the A/D-ring orientation may be the cause of substrate inhibition observed with 5␣-DHT, testosterone, and 20␣-OHP in this mutant.
In conclusion, we have successfully engineered steroid 5␤-reductase into a highly efficient 3␤-HSD by a single glutamate-to-histidine mutation. The bulky histidine alters the shape of the active site, resulting in a shift of the relative position of the steroid to the cofactor, which changes the enzyme function. This E120H mutant elicits a change in stereospecificity of the HSD reaction by flipping the steroid, and it shows an unexpected preference for C19 steroids with a 5␣-configuration over C19 steroids with a 5␤-configuration due to the presence of a cylindrical binding channel. Engineered enzymes rarely exhibit adequate catalytic efficiency for a new activity with complete eradication of the native activity, but here we demonstrate an example of a perfect change of function achieved by a single point mutation.