Purification, reconstitution, and steady-state kinetics of the trans-membrane 17 beta-hydroxysteroid dehydrogenase 2.

Human membrane 17 beta-hydroxysteroid dehydrogenase 2 is an enzyme essential in the conversion of the highly active 17beta-hydroxysteroids into their inactive keto forms in a variety of tissues. 17 beta-hydroxysteroid dehydrogenase 2 with 6 consecutive histidines at its N terminus was expressed in Sf9 insect cells. This recombinant protein retained its biological activity and facilitated the enzyme purification and provided the most suitable form in our studies. Dodecyl-beta-D-maltoside was found to be the best detergent for the solubilization, purification, and reconstitution of this enzyme. The overexpressed integral membrane protein was purified with a high catalytic activity and a purity of more than 90% by nickel-chelated chromatography. For reconstitution, the purified protein was incorporated into dodecyl-beta-D-maltoside-destabilized liposomes prepared from l-alpha-phosphatidylcholine. The detergent was removed by adsorption onto polystyrene beads. The reconstituted enzyme had much higher stability and catalytic activity (2.6 micromol/min/mg of enzyme protein with estradiol) than the detergent-solubilized and purified protein (0.9 micromol/min/mg of enzyme protein with estradiol). The purified and reconstituted protein (with a 2-kDa His tag) was proved to be a homodimer, and its functional molecular mass was calculated to be 90.4 +/- 1.2 kDa based on glycerol gradient analytical ultracentrifugation and chemical cross-linking study. The kinetic studies demonstrated that 17 beta-hydroxysteroid dehydrogenase 2 was an NAD-preferring dehydrogenase with the K(m) of NAD being 110 +/- 10 microM and that of NADP 9600 +/- 100 microM using estradiol as substrate. The kinetic constants using estradiol, testosterone, dihydrotestosterone, and 20 alpha-dihydroprogesterone as substrates were also determined.

The members of the 17␤-hydroxysteroid dehydrogenase (17␤-HSD) 1 family are crucial in the biosynthesis and metabolism of active steroid hormones in a variety of tissues. Estrogens and androgens in turn control a variety of important physiological functions such as growth, reproduction, and differentiation. Using NAD as cofactor, 17␤-HSD2, with its pre-dominantly oxidative activity, primarily converts the highly active 17␤-hydroxysteroids such as estradiol, testosterone, and dihydrotestosterone into their inactive keto forms. Furthermore, studies carried out in vitro indicate that 17␤-HSD2 is able to use C 20 -steroids as substrates, namely to catalyze the oxidation of 20␣-dihydroprogesterone to progesterone. The expression of the mRNA of human 17␤-HSD2 has been detected in a large variety of tissues. Its 1.5-kb mRNA is highly expressed in the endometrium, placenta, liver, and small intestine and also in smaller amounts in the pancreas, colon, kidney, and prostate (1)(2)(3)(4)(5). Human 17␤-HSD2 mRNA has also been found to be present in human breast, endometrial, and prostate cancer cell lines (3). In addition, both rodent and human 17␤-HSD2 enzymes are widely distributed in the gastrointestinal and urinary tracts, in the liver, as well as in the adrenals of adults and developing fetuses (2)(3)(4)(5). Recently, the correlation between 17␤-HSD2 and colonic cancer was reported (6). The broad tissue distribution, together with the predominant oxidative activity of 17␤-HSD2, suggests that the enzyme plays an essential role in the inactivation of highly active 17␤-hydroxysteroids. It may have a protective role by lowering the active steroid concentrations and reducing excessive sex hormone action in target tissues.
17␤-HSD2 is a trans-membrane protein, which is demonstrated by its subcellular distribution in the endoplasmic reticulum (7). 17␤-HSD2 cDNA encodes a predicted protein of 387 amino acids with a molecular mass of 42,782 daltons. The primary structure shows that it belongs to the type II signal anchor membrane protein, which is characterized by possessing a cluster of positively charged amino acids and followed by a hydrophobic core of about 33 nonpolar amino acids close to the N terminus of the protein (1,8,9). The carboxyl terminus has a luminal carboxyl-terminal endoplasmic reticulum retention motif (KKK) (1). Based on the trans-membrane helices prediction using a hidden Markov model (10), there are two proposed trans-membrane helices close to the N terminus of 17␤-HSD2, the first one situated in amino acids 5-27 and the second one in 34 -56. The latter is much more hydrophobic than the former. The enzyme is thus suggested to be an integral membrane protein (7).
Up to now, most of the information obtained for 17␤-HSD2 is about genes and mRNA studies. Although an N-29 amino acid truncated form, in which the first proposed transmembrane helix was deleted, retained about 60% of its catalytic activity as compared with wild type in the intact cells and was purified using a detergent ␤-octyl glucoside, knowledge about the purification of the enzyme is still limited (7). In order to elucidate the structure and function of the protein, we carried out the overproduction, purification, reconstitution, and characterization of N-terminal His 6 -tagged full-length 17␤-HSD2, which are reported here.
Construction and Production of Recombinant Baculovirus-The fulllength human 17␤-HSD2 cDNA coding sequence was obtained by PCR amplification from pCMV/17␤-HSD2 (11). A nucleotide sequence coding for 6 histidine residues followed by a Factor Xa cleavage site was added at the 5Ј terminus of the 17␤-HSD2 cDNA. The forward primer contained a BamHI site (underlined), 5Ј-CGGGATCCATGGGCAGCCATCA-TCATCATCATCATCATAGCAGCATCGAAGGCCGTGGCGGCATGAG-CACTTTCTTCTCGGACACAGCATGG-3Ј, and the reverse primer contained an EcoRI site (underlined), 5Ј-GGAATTCTTACTAGGTGGC-CTTTTTCTTGTA-3Ј. The 1.2-kb amplified products were digested with the appropriate enzymes and subcloned into the corresponding sites of the pBlue Bac4.5 vector. Using this method, we also constructed N-terminal His 10 -tagged as well as C-terminal His 6 -tagged 17␤-HSD2 and the enzyme lacking the first 38, 52, and 61 amino acids of the N terminus (N-38-, N-52-, and N-61-deleted 17␤-HSD2). The recombinant vectors were identified using dideoxynucleotide sequencing (Big Dye TM Terminator Cycle Sequencing Ready Reaction Kit, PerkinElmer Applied Biosystems, 373 sequencer with XL Upgrade). Linearized AcMNPV DNA (Bac-N-Blue DNA) (0.5 g) was used to co-transfect monolayers of Sf9 cells in the presence of InsectinPlus liposomes, according to the manufacturer's instructions. Five days after co-transfection, the media were collected, and the recombinant baculoviruses were purified using three rounds of plaque assay in the presence of Bluo-Gal.
Cell Culture and Virus Infection-Sf9 cells were grown as monolayers in flasks containing Grace's insect cell culture medium with 5% fetal bovine serum and maintained at 27°C. The wild type baculovirus and the recombinant virus carrying 17␤-HSD2 were used to infect the 90% confluent cells at a multiplicity of infection of 0.1-0.5 for virus amplification and an multiplicity of infection of 5-10 for protein overproduction. The infected cells were harvested 72 h postinfection, washed with cold phosphate-buffered saline, pelleted, and stored at Ϫ80°C for later use.
Purification of N-terminal His 6 -tagged 17␤-HSD2-The cell pellets from 6 -8 ϫ 175-cm 2 flasks (which represents about 1.6 -2 ϫ 10 8 cells) were suspended in 50 ml of buffer A. The remaining procedures were carried out at 4°C or on ice unless otherwise specified. The cells were lysed by sonication. The suspension was incubated for 15 min and centrifuged for 30 min at 180,000 ϫ g. The pellets were then solubilized in 100 ml of buffer B (40 mM Tris, pH 8.0, 150 mM NaCl, 10% glycerol, 8 mM imidazole, 20 mM NAD, 0.4 mM phenylmethylsulfonyl fluoride, 0.5% ␤-DDM, 1 g/ml each of the protease inhibitors) and incubated by rotating for 1 h. The supernatants were collected after centrifugation for 45 min at 180,000 ϫ g, adjusted to 300 mM NaCl, mixed with 3 ml of nickel-chelated resin pre-equilibrated with buffer B (containing 300 mM NaCl), and incubated by rotating for 1 h. The mixture was loaded onto the column. The column was washed with 10 column volumes of buffer C (buffer B containing 300 mM NaCl, 0.3% ␤-DDM, 15 mM imidazole, and 15% glycerol) and 10 column volumes of buffer D (buffer B containing 45 mM imidazole, 200 mM NaCl, 0.3% ␤-DDM, and 20% glycerol). Bound proteins were eluted with buffer E (40 mM Tris, pH 7.5, 150 mM NaCl, 20% glycerol, 0.2% ␤-DDM, 250 mM imidazole, 40 M NAD, 0.4 M phenylmethylsulfonyl fluoride, and 1 g/ml protease inhibitors). The fractions with high 17␤-HSD2 activity were collected, frozen in liquid nitrogen, and stored at Ϫ80°C. The purified enzyme without adding cofactor NAD in the purification procedures was used to detect cofactor kinetic constants.
Preparation of Liposomes and Protein Reconstitution-The reconstitution method basically depended on the strategies described by Rigaud (12). Three kinds of phospholipid (PC, phosphatidylethanolamine, and L-␣-phosphatidylinositol) were tested in the reconstitution system. They were mixed with different ratios, which were around the lipid compositions of the human liver. The mixtures were dissolved in chloroform and dried under a stream of nitrogen gas to minimize the lipid oxidation. The remaining trace of solvent was removed under vacuum for at least 2 h. The dried films were suspended in buffer F (20 mM Tris, pH 7.4, 150 mM NaCl, 1 mM EDTA, and 0.5 mM DTT) at a concentration of 20 mg of lipids/ml. The liposomes were obtained by sonication (1-s burst, 5-s interval for 45 min at output 2.0) in an ice bath under nitrogen gas. The suspensions were frozen in liquid nitrogen and thawed at room temperature three times. The liposomes were extruded through 400-nm polycarbonate membranes (Nuclepore Track-Etch Membrane; Corning) three times, frozen in liquid nitrogen, and stored at Ϫ80°C. To determine the physical state of the detergent-solubilized liposomes, the liposomes were aliquoted and mixed with different amounts of ␤-DDM in a final concentration of 4 mg of lipids/ml using buffer F. The samples were incubated at room temperature for 3 h under constant agitation. The turbidity of the different phospholipiddetergent suspensions was measured at 540 nm with a spectrophotometer (Backman DU 7400).
To reconstitute 17␤-HSD2, the liposomes were diluted to 4 mg of lipids/ml, saturated with ␤-DDM, and equilibrated under constant agitation for 3 h at room temperature. The purified N-terminal His 6tagged 17␤-HSD2 was adjusted to 0.15 M NaCl and the same concentration of detergent as that in the saturated liposomes, mixed with the saturated liposomes in a ratio of liposomes to protein of 14:1 (w/w), and incubated for 2 h at 4°C under rotating. The detergent was removed by three successive extractions with 80 mg/ml, wet weight, polystyrene beads and rotated at 4°C. The first, second, and third extractions lasted for 2, 2, and 4 h, respectively. The beads were removed by filtration over glass wool. Then the mixture was adjusted to 6% glycerol using buffer F. The proteoliposomes were harvested by centrifugation at 180,000 ϫ g for 45 min, resuspended in a buffer (40 mM Tris, pH 7.4, 20% glycerol, 150 mM NaCl, 1 mM EDTA, and 0.5 mM DTT), and stored at Ϫ80°C.
Enzyme Assay in the Purification-The activity of 17␤-HSD2 was monitored by the spectrophotometric assay. It was initiated by the addition of 17␤-HSD2 in 0.5 ml of reaction mixture (50 mM sodium carbonate, pH 9.2, 1 mM NAD, and 25 M testosterone). The reactions were monitored by spectrophotometric measurement of the reduction of NAD at 340 nm. A reaction mixture containing no cofactor or substrate was used as control. One unit of enzyme activity is defined as 1 mol of product formed in 1 min.
Enzyme Activity in Cultured Cells-Enzyme activities in cultured cells were measured by plating the cells in the six-well plates at a density of 1.2 ϫ 10 6 /well. The cells were set for 1 h to attach followed by the infection of virus at a multiplicity of infection of 10. A mock infection was set as a background control. After a 50-h incubation at 27°C, the medium was removed, and 2 ml of serum-free TNM-FH medium with 10 M of each 14 C-labeled estrone, estradiol, testosterone, androstenedione (4-dione), and dihydrotestosterone was added to each well. The reaction was set at room temperature, at different time intervals (3, 10, 20, 40, and 60 min), and aliquots of the media were moved to the tubes containing cold diethyl ether. The steroids were extracted and quantified as mentioned under "Steady-state Kinetics." Steady-state Kinetics-The kinetic constants of 17␤-HSD2 were determined using purified and reconstituted N-terminal His 6  constants of NADH. The kinetic constants of NADP were determined with a 10 M constant concentration of estradiol and various concentrations of cofactor NADP (0.6 -10 mM). The initial velocity was measured with less than 5% substrate conversion. The reactions were carried out at 37°C and stopped by removing 0.5 ml of reaction mixture to the cold diethyl ether at four different time intervals (0, 20, 40, and 60 s). The steroids were extracted with ethanol in dry ice and dried by evaporation. They were then dissolved in dichloromethane, applied onto thin layer chromatograms (TLC), separated by toluene/acetone (4:1, v/v), and quantified by Storm 860 Laser Scanner (Molecular Dynamics, Inc., Sunnyvale, CA; ImageQuant software). At least three independent experiments were carried out for each kinetic constant. The kinetic results were fitted for the Michaelis-Menten equation and calculated using a Lineweaver-Burk plot. The values of the catalytic constant, k cat , were calculated from the V max values with the homodimer molecular mass of 90 kDa (k cat is the turnover number, i.e. the number of moles of substrate transformed per second per mole of enzyme).
Glycerol Gradient-The apparent functional molecular mass of Nterminal His 6 -tagged 17␤-HSD2 was estimated by cosedimentation with protein standards on 8 -30% glycerol gradients. Glycerol gradients (13 ml; Beckman SW 40Ti rotor) were prepared by using a gradient maker with equal volumes of 8 and 30% glycerol buffer containing 20 mM Tris, pH 7.4, 0.15 M NaCl, 40 M NAD, 1 mM EDTA, 0.5 mM DTT, and 0.1% Triton X-100. The glycerol gradients were equilibrated at 4°C for about 8 h before loading the samples. The samples contained 20 g of purified and reconstituted 17␤-HSD2, 100 g of each protein standard (rabbit skeletal muscle aldolase, bovine serum albumin, and chicken ovalbumin (13)), and the same buffer as in the gradient, but the glycerol concentration was less than 8%. The samples were equilibrated at 4°C for 1 h and then layered on top of the glycerol gradients and centrifuged at 40,000 rpm for 40 h at 4°C. The gradients were fractionated from the bottom into 0.3-ml fractions. 17␤-HSD2 fractions were verified by the enzyme activity assay and Western blot. The positions of the standard markers were determined by SDS-PAGE.
Chemical Cross-linking of 17␤-HSD2-This was performed according to the method described by Knoller (14) with modifications. 60 l of reaction buffer contains 50 mM potassium phosphate, pH 7.4, 20% glycerol, 1 mM EDTA, 0.5 mM DTT, 3 g of purified and reconstituted N-terminal His 6 -tagged 17␤-HSD2, and cross-linking reagent BS with concentrations at 0, 0.25, 1, and 3 mM, respectively. The reaction proceeded for 30 min at 25°C and stopped by adding glycine to a final concentration of 30 mM. The samples were analyzed by 5-15% gradient SDS-PAGE followed by Western blot.
SDS-PAGE and Western Blot Analysis-SDS-PAGE was performed according to the method of Laemmli (15) using 12% polyacrylamide gel or 5-15% gradient SDS-PAGE. The samples in reducing loading buffer were incubated at 40°C for 30 min instead of boiling before loading to prevent the aggregation of the membrane protein (16). The gel after migration was stained with Coomassie Blue. For Western blot analysis, blots were probed with rabbit polyclonal antibody raised against human 17␤-HSD2 as the first antibody and horseradish peroxidase (HRP)conjugated donkey anti-rabbit polyclonal antibody (Amersham Biosciences) as the second antibody. The immunoreactive blots were detected with ECL reagents (PerkinElmer Life Sciences) and exposed to x-ray film (Eastman Kodak Co.).
Protein Concentration Determination-Protein concentrations without detergent were determined using the Bradford reagent (Bio-Rad). The concentrations of proteins with detergents or with phospholipid were determined by the method of microgram quantities of protein determination (17) to prevent the alteration of detergents and phospholipids in the protein concentration by conventional methods.

RESULTS
Overproduction of Various Recombinant 17␤-HSD2-Human 17␤-HSD2 cDNA with a 6-histidine coding sequence and a Factor Xa cleavage site at its 5Ј terminus was subcloned into the baculovirus transfer vector pBlueBac 4.5. The incorporation of the Factor Xa cleavage site allowed the removal of the His tag after purification of the recombinant protein, leaving only two additional glycines at the N terminus of 17␤-HSD2. Sf9 cells were co-transfected with Bac-N-Blue DNA and the above transfer vector of 17␤-HSD2 to produce the recombinant baculovirus. Protein expression was optimized by evaluating the expression levels of the infection at different time intervals. The activity was first detected 24 h postinfection and reached a maximum between 60 and 72 h postinfection (Fig. 1, A and B), whereas no activity could be detected in wild-type AcMNPV virus-infected cells. Thus, the protein expression conditions were set as follows: infection of the cells at a multiplicity of infection from 5 to 10 and harvest in 72 h postinfection. Under these conditions, the overexpressed 17␤-HSD2 constitutes about 3% of the total protein in the insect cell lysate, with a specific activity of 0.012 units/mg in the cell homogenate.
Using the same method, we also overproduced N-38-, N-52-, and N-61-deleted 17␤-HSD2, as well as N-terminal His 10tagged and C-terminal His 6 -tagged 17␤-HSD2. The truncated N-38 form was expressed at about 2% of the total protein in the insect cell lysate, with a specific activity of 0.005 units/mg in the cell homogenate. Although this form could be solubilized to a higher level with detergent from membrane vesicles (solubilized 45% of the enzyme in the presence of 0.4% ␤-DDM) than that of N-terminal His 6 -tagged 17␤-HSD2 (solubilized 30.5% of the enzyme in the presence of 0.4% ␤-DDM), it was unstable in the solubilized state with detergents even in intact cells and showed a very strong tendency to degrade and aggregate in the cell homogenate. The truncated N-52 and N-61 forms were expressed in fairly low amounts in Sf9 insect cells. The N-52 form was even more unstable than the N-38 form. We found that the N-52 form in fresh cultured cells still retained a little activity, but it completely lost activity in several hours at 4°C. Moreover, the N-61 form was totally inactive in fresh cultured cells. The N-terminal His 10 -tagged 17␤-HSD2 was expressed to a high level (about 5% of the total protein); however, it retained a quite lower specific activity (about 0.001 units/mg) than that of the N-terminal His 6 -tagged form (about 0.012 units/mg). The C-terminal His 6 -tagged 17␤-HSD2 was expressed at about 2% of the total protein in the insect cell lysate, with a specific activity of 0.009 units/mg in the cell homogenate. This recombinant was solubilized to a lower level from membrane vesicles (solubilized 21% of the enzyme in the presence of 0.4% ␤-DDM) than that of N-terminal His 6 -tagged 17␤-HSD2 and exhibited a very high tendency to aggregate, as seen in SDS gel and Western blot analysis. Indeed, the majority of the protein presented as a polymer staying in the sample-loading place or as a dimer (data not shown). These findings suggest that there is a stronger membrane interaction in C-terminal His 6 -tagged 17␤-HSD2 than in N-terminal His 6 -tagged 17␤-HSD2. Furthermore, several purification tests using various detergents demonstrated that the C-terminal His 6 tag of this form was not able to be effectively bound to nickel-chelated affinity matrix. Finally, N-terminal His 6 -tagged 17␤-HSD2, although still highly labile, was found to be able to retain full biological activity, to be expressed in a fairly good amount in the baculovirus expression system, and to facilitate its purification. Therefore, this form was chosen in our study.
Effects of Detergents on the Solubility and Activity of 17␤-HSD2-N-terminal His 6 -tagged 17␤-HSD2 is still a very labile protein with a strong tendency to aggregate and degrade. The choice of an optimal detergent is the crucial step in the purification. Much effort was devoted to finding a suitable detergent to solubilize this recombinant from the cell membranes. Several commonly used detergents were tested for the protein stability, solubility, and binding capacity with nickel matrix. The results are summarized in Table I. The pH during solubilization was kept at 8.0, since the protein was subsequently used to bind to the Ni 2ϩ -agarose matrix. Sodium cholate showed no significant inhibition of 17␤-HSD2 activity, but the protein solubility was very low with this detergent. ␤-Octyl glucoside had low ability in protein solubilization and strong inhibition to the enzyme activity at increasing concentrations. C 12 E 8 and decyl-␤-D-maltoside displayed medium ability both in solubilizing and in maintaining the enzyme activity. Triton X-100 showed high protein solubility, but it significantly inhibited the enzyme activity. Although none of the detergent could assist the enzyme to get more than 50% solubility, ␤-DDM gave the best results both in solubilizing and in maintaining the enzyme activity among those detergents tested. Typically, we used 0.5% ␤-DDM in the solubilization of His 6 -tagged 17␤-HSD2, considering that higher concentrations of ␤-DDM did not further improve the solubility notably but rather inhibited the enzyme activity and solubilized more contaminants.
Purification and Reconstitution-The purification was car-ried out in a single affinity chromatography step using ␤-DDM as detergent. The results are summarized in Table II and are presented in Fig. 2, A and B. Most contaminants were removed by washing the column with 45 mM imidazole, and N-terminal His 6 -tagged 17␤-HSD2 was purified with a purity of more than 90% based on Coomassie Blue staining and densitometric analysis. There were two bands with molecular masses of about 44 and 90 kDa on the SDS gel, which were confirmed to be a monomer form and a dimer form, respectively, by Western blot analysis. The yield of the purification was typically about 1 mg of homogeneous protein with a specific activity of about 0.9 unit/mg using estradiol as substrate from 2 ϫ 10 8 cells. We found that the enzyme activity in both the homogenate and the supernatant using ␤-DDM as detergent could be kept for several days at 4°C without significant loss (data not shown). However, the ␤-DDM-solubilized and purified 17␤-HSD2 had so strong a tendency to denature that the protein stored at 4°C for 3 and 24 h would lose half and total activity, respectively (data not shown). The fractions of the elution had to be frozen by liquid nitrogen as quickly as possible and stored at Ϫ80°C or immediately used for reconstitution. Liposomes from the mixtures of PC, phosphatidylethanolamine, and L-␣-phosphatidylinositol and from single PC were tested to compare their ability to reconstitute 17␤-HSD2 activity. The proteoliposomes formed from single PC demonstrated the highest activity and were thus chosen for the protein reconstitution. To follow the physical state of the liposomes, the absorbance at 540 nm was measured at various concentrations of ␤-DDM. Purified 17␤-HSD2 was tested for the incorporation into four different stages of liposomes with ␤-DDM (i.e. the liposomes before saturation, saturated (onset solubilization), halfway through the breakdown of the liposomes, and fully solubilized (micellar state)). The concentrations of ␤-DDM at these four points corresponded to 0.2, 0.4, 0.5, and 0.9%, respectively. The physical states of the liposomes with ␤-DDM and the activities of the reconstituted 17␤-HSD2 are shown in Fig. 3. The highest activity was obtained when the liposomes saturated with ␤-DDM. Typically the liposomes with a slight oversaturation (0.42% ␤-DDM) were used in the reconstitution system.
The activities in the proteoliposomes depended not only on the physical state of the liposomes at the beginning of the reconstitution but also on the liposome/protein ratios, the concentration of glycerol, and the ionic strength. Different liposome/protein ratios were tested, and the optimal ratio was around 14:1 (w/w). Using a lower liposome/protein ratio in the reconstitution system, the enzyme was not able to incorporate into the liposomes totally, and using a higher ratio, the harvested proteoliposomes were less active or totally inactive. The presence of glycerol at less than 15% concentration in protein incorporation and detergent removal led to significant loss of enzyme activity. At the last step of reconstitution, before the proteoliposomes were harvested by centrifugation, the glycerol was diluted to 6% to facilitate the proteoliposomes pellet formation, whereas dilution to an extremely lower concentration of glycerol led the harvested proteoliposomes to be totally inactive. It was also found that using fairly high ionic strength (0.15-0.2 M NaCl) in the reconstitution procedure could produce a higher level of active proteoliposomes than that in a lower ionic strength (data not shown). The reconstituted 17␤-HSD2 using our optimal conditions showed higher specific activity (2.6 units/mg with estradiol) and much higher stability than before reconstitution. The proteoliposomes were kept at 4°C for 2 months without significant loss of enzyme activity. The reconstituted protein was also purer than before due to the  (1). The apparent functional molecular mass of this protein was calculated based on three independent glycerol gradient experiment results with 0.1% Triton X-100, which showed a molecular mass of 90.4 Ϯ 1.2 kDa (Fig. 4, A-C). The Western blot results confirmed that the protein was present exclusively in the fractions corresponding to the peak of 17␤-HSD2 activity. No 17␤-HSD2 bands were detected in the rest of the fractions. To further clarify the association of the subunits, we performed a cross-linking experiment using BS as reagent. There was one major band at monomer and one minor band at dimer positions in the control sample (Fig. 5). In the samples with 0.25, 1, and 3 mM of BS, the rather defused dimer band intensity increased with increasing BS concentration while the monomer intensity decreased. There were no clear bands of trimer and tetramer. The polymer form of the protein presented in every sample loading position. With the BS concentration increasing, the polymer bands became stronger. These results demonstrate that 17␤-HSD2 presents as a homodimer in the natural state.
The kinetic constants for different substrates (testosterone, estradiol, dihydrotestosterone, and 20␣-dihydroxyprogesterone) and those for cofactors NAD(H) and NADP are summarized in Tables III and IV. The K m app values for steroid substrates were close to those published results measured in cell homogenates (1) and in purified N-29-deleted 17␤-HSD2 (7). We also measured the kinetic constants for diphosphate cofactors NAD(H) with saturated concentrations of steroids (estradiol, testosterone, estrone, and 4-dione) and triphosphate cofactor NADP with saturated concentration of estradiol. Similar to the K m app values for substrates, the K m app values for NAD between two oxidative substrates (estradiol and testosterone) and the K m app values for NADH between two reductive substrates (estrone and 4-dione) were also very close. Although the K m app values for NAD were 20 -30-fold higher and the apparent V max values were also 10 -15-fold higher than those for NADH, the enzyme had almost the same apparent catalytic specificity for both oxidation and reduction. However, it was almost unidirectional in favor of oxidative reaction in intact Sf9 cells (Fig.  6), and the same results were reported in cultured HEK293 cells (7,11). The K m app for NADP with the estradiol as substrate at a saturating level, however, reached 9600 M, more than 80-fold higher than that for NAD with the same substrate. This finding suggests that the cofactor concentration in the cells is a key factor to decide the reaction direction. It is well known that the reductase uses NADPH as cofactor and the dehydrogenase uses NAD as cofactor in vivo. As a result of the enzyme's kinetic property and since the intracellular concentration of NAD is remarkably higher than that of NADH, with a ratio of about 1000 (18), the reaction will be in the oxidation direction by using NAD as cofactor for this enzyme in vivo.

DISCUSSION
As described in the Introduction, there are two proposed trans-membrane helices located at the N terminus of 17␤-HSD2, but the more hydrophobic domain centers on the second proposed trans-membrane helix. We have devoted considerable effort to find a truncated form, which was able to retain enzyme activity as well as having better solubility. We found that the truncated N-38 form retained about 40% of its catalytic activity as compared with the full-length enzyme, but this form was extremely unstable either in the cell homogenate or in the detergent-solubilized state. For the other recombinants, the N-52 form only retained some activity in fresh cultured cells, and the N-61 form was totally inactive. It was reported that the truncated N-29 17␤-HSD2 (in which the first proposed transmembrane helix was deleted) retained about 60% of its cata-  lytic activity as compared with the wild type enzyme (7). However, the truncated N-80 form was totally inactive (7). All of these results suggest that the first proposed trans-membrane helix is less important and that the second one is crucial in maintaining the enzyme functions.
We found that N-terminal His 6 -tagged 17␤-HSD2 was the most suitable form to study among our three overexpressed His-tagged 17␤-HSD2 recombinants. Our experiments demonstrated that C-terminal His 6 -tagged 17␤-HSD2 was much more hydrophobic than the N-terminal His 6 -tagged form. This suggests that the soluble His tag on the N terminus not only facilitates its purification but also weakens the hydrophobicity centered on the N-terminal region of the enzyme. Therefore, the N-terminal His 6 -tagged protein should be less hydrophobic than the full-length one without His tag protein. However, the N-terminal His 10 -tagged protein retained only about 10% of the wild type catalytic activity, and this indicates that the longer His tag may perturb the enzyme structure. We also found that the overproduction level of the N-terminal His 6 -tagged form was higher than the C-terminal His 6 -tagged form and lower than the N-terminal His 10 -tagged form. This suggests that less hydrophobicity close to the N terminus of the protein will overproduce a higher level of the protein.
According to the primary structure of human 17␤-HSD2, there is a strong hydrophobic core possessing 33 nonpolar amino acids close to its N terminus and a quite hydrophilic motif in the other region (1), which indicates that 17␤-HSD2 has a high tendency to aggregate especially in the detergentsolubilized state. Based on this consideration, several precautions had been taken in the purification and reconstitution procedures. First, an appropriate concentration of glycerol was required when treating 17␤-HSD2. We found that glycerol played an important role in stabilizing the protein, but high concentration of glycerol reduced the binding capacity of the His 6 -tagged 17␤-HSD2 with Ni 2ϩ matrix. Thus, 10% glycerol was used in the sample buffer when the protein bound with Ni 2ϩ matrix, and 20% glycerol was used in the other steps. Second, we found that the proper ionic strength (0.15-0.2 M NaCl) could strengthen the detergent acting on the hydrophobic region of the protein, so that it could stabilize the protein.
High ionic strength could increase the specific binding of the protein with Ni 2ϩ resin, but excessive high ionic strength (Ͼ0.5 M NaCl) caused serious enzyme aggregation and degradation. A concentration of 0.3 M NaCl was thus used when the protein bound with Ni 2ϩ matrix, and a lower concentration of NaCl was used in the other steps. Third, the proper amount of cultured cells and Ni 2ϩ resin was required. We tested different amounts of the cultured cells and Ni 2ϩ resin for purification. Using a large scale of cultured cells and large amount of Ni 2ϩ resin or using a pressure on the column always led to the protein aggregation and to its denaturation. Finally, we found that cofactor NAD had an effect on stabilizing the enzyme. Using NAD in the purification procedure helped to obtain an enzyme preparation with higher activity.
The purified full-length 17␤-HSD2 was unable to keep its activity, and the protein could not be solubilized by 1 M potas-  5. Chemical cross-linking of 17␤-HSD2. Purified and PCreconstituted 17␤-HSD2 were cross-linked with different concentrations of BS. Reaction products were resolved by 5-15% gradient SDS-PAGE and were analyzed by Western blot. Lanes 1-4, protein crosslinked with 0, 0.25, 1.0, and 3.0 mM BS, respectively. sium carbonate and attained less than 20% solubility by 0.1 M sodium hydroxide, a method widely used to discriminate between membrane-associated proteins and integral membrane proteins (19). Based on all of those facts, we consider 17␤-HSD2 to be an integral membrane protein. To obtain active 17␤-HSD2 for biological and structure-function studies, we needed to attempt to reconstitute the enzyme. Finally, we successfully reconstituted the enzyme using the ␤-DDM-mediated method. The purified and lipid PC-reconstituted 17␤-HSD2 was employed to examine the enzyme characteristics. The reconstituted protein has the same physiological properties as previously reported (1). The optimal pH profile with testosterone is similar with the result obtained using HEK 293 cell-expressed 17␤-HSD2 (data not shown) (1). The K m app values with several substrates in the oxidation reaction are similar to the results from HEK 293 cell-expressed 17␤-HSD2 cell homogenate (1) and from purified N-29-truncated protein (7). But the k cat values for several substrates in the oxidation reaction are higher than those reported. All of these demonstrated that the reconstituted protein is fully active.
Our results of cofactor kinetics demonstrate the fact that although the K m app for NAD is higher than that for NADH and both cofactors have the same apparent kinetic specificities, 17␤-HSD2 is undoubtedly a dehydrogenase in vivo (see below). The similar K m app values for NAD and NADH are also reported in its kinetics with cell homogenates (20), which are 70 M for NAD and 14 M for NADH with testosterone and 4-dione as fixed substrates, respectively. Whether the nicotinamide nucleotides for the enzyme reaction were from the intraluminal or the cytoplasmic part of the cell is not clear, but it is well known that intracellular concentration of NAD is remarkably higher than that of NADH, and NADPH concentration is higher than that of NADP (18). It is also well known that the reductases use NADPH as cofactor and the dehydrogenases use NAD as cofactor in vivo (11). This indicates that the cofactor concentration in the cells is a key factor to decide the reaction direction, so that the reaction goes on oxidation when using NAD as cofactor in vivo.
A glycerol gradient in the absence or presence of various concentrations of Triton X-100 was used to estimate the functional molecular mass of 17␤-HSD2. We found that the vesicles of proteoliposomes migrated in the upper part of the gradient in the absence of Triton X-100, and this phenomenon is consistent with the other published gradient ultracentrifugation experiment of reconstituted membrane protein (21). With 0.1-0.5% of Triton X-100 in the gradient, the enzyme activity was reduced from 50% to more than 90%. However, its apparent functional molecular mass was shown to be stable (91-89-kDa area in the gradient) in this Triton X-100 concentration range. Western blot results confirmed that 17␤-HSD2 was present in the fractions corresponding to the peak of 17␤-HSD2 activity whether the activity was high or low. We used 0.1% Triton in the gradient, considering that the enzyme retained lower activity in higher concentrations of the detergent. Our results together with other reports (22,23) demonstrate that Triton X-100 does not interfere with protein functional molecular mass significantly in glycerol gradient. The enzyme with ␤-DDM, although it retained much higher activity than with Triton X-100 under the same detergent concentration, exhibited larger apparent molecular mass (169 -110 kDa, corresponding to 0.04 -0.5% ␤-DDM in the gradient) in the glycerol gradient. Using ␤-DDM concentration higher than 0.5% in the gradient, the enzyme completely lost its activity. This demonstrated that the detergent interfered with the protein functional molecular mass significantly in the glycerol gradient.
In conclusion, this study provides an efficient method to obtain highly labile integral membrane 17␤-HSD2 enzyme. The overexpressed N-terminal His 6 -tagged 17␤-HSD2 was demonstrated to be the most suitable form in our study. ␤-DDM  6. 17␤-HSD2 activities in the intact Sf9 cells. Five different substrates, three (estradiol (E2), testosterone (T), and dihydrotestosterone (DHT)) for the oxidation reaction and two (E1 and 4-dione) for the reduction reaction were used to detect the enzyme's substrate specificity. The results demonstrate that the enzyme is almost unidirectional in favor of the oxidation reaction with the same specific activity among those three substrates.
is the best detergent in both solubilizing the protein and maintaining the enzyme in an active state. 17␤-HSD2 was proved to be a homodimer with a molecular mass of 90.4 Ϯ 1.2 kDa in the presence of a 2-kDa His tag. Our purification and reconstitution procedures provide a new and advanced way to obtain homogeneously and functionally reconstituted 17␤-HSD2. This will permit us to further scale up the cell culture volume and recombinant protein production, thereby yielding sufficient homogeneous protein to approach crystallization and further structure studies. The methods we have introduced here may be applicable for other membrane steroid enzymes.