Molecular Characterization of a First Human 3(α→β)-Hydroxysteroid Epimerase*

In this report, we describe the isolation and characterization of a cDNA encoding an enzyme that exhibits catalytic characteristics of a 3(α→β)-hydroxysteroid epimerase (3(α→β)-HSE). The enzyme overexpressed in human 293 embryonic kidney cells transforms androsterone into epi-androsterone in two steps: the oxidation of androsterone to 5α-androstane-3,17-dione, followed by the reduction of the latter to epi-androsterone. The reverse reaction, 3(β→α)-hydroxysteroid epimeration, is approximately 10-fold weaker. These results are confirmed by V max/K m determination, which shows that the enzyme catalyzes the oxidation of androsterone to 5α-androstane-3,17-dione and the reduction of 5α-androstane-3,17-dione to epi-androsterone more efficiently than the reverse reactions. The selective catalysis of the reaction following the 3(α→β) direction is also observed in intact transfected cells in culture, which better reflect physiological conditions. In vitro assays reveal that the recombinant enzyme prefers NAD+ and NADH as cofactors and could recognize both C-19 and C-21 3α-hydroxysteroids as substrates. DNA sequence analysis predicts a protein of 317 amino acids. Tissue distribution analysis using RT-PCR reveals that the mRNA of the enzyme is expressed in various tissues, including liver, brain, prostate, adrenal, and uterus, with the most abundant expression in the liver. Because active hydroxysteroids generally exert their effect in a stereo-specific manner, 3(α→β)-HSE could thus potentially play an important role in regulating the biological activities of various steroids.

In this report, we describe the isolation and characterization of a cDNA encoding an enzyme that exhibits catalytic characteristics of a 3(␣3␤)-hydroxysteroid epimerase (3(␣3␤)-HSE). The enzyme overexpressed in human 293 embryonic kidney cells transforms androsterone into epi-androsterone in two steps: the oxidation of androsterone to 5␣-androstane-3,17-dione, followed by the reduction of the latter to epi-androsterone. The reverse reaction, 3(␤3␣)-hydroxysteroid epimeration, is approximately 10-fold weaker. These results are confirmed by V max /K m determination, which shows that the enzyme catalyzes the oxidation of androsterone to 5␣androstane-3,17-dione and the reduction of 5␣-androstane-3,17-dione to epi-androsterone more efficiently than the reverse reactions. The selective catalysis of the reaction following the 3(␣3␤) direction is also observed in intact transfected cells in culture, which better reflect physiological conditions. In vitro assays reveal that the recombinant enzyme prefers NAD ؉ and NADH as cofactors and could recognize both C-19 and C-21 3␣hydroxysteroids as substrates. DNA sequence analysis predicts a protein of 317 amino acids. Tissue distribution analysis using RT-PCR reveals that the mRNA of the enzyme is expressed in various tissues, including liver, brain, prostate, adrenal, and uterus, with the most abundant expression in the liver. Because active hydroxysteroids generally exert their effect in a stereospecific manner, 3(␣3␤)-HSE could thus potentially play an important role in regulating the biological activities of various steroids.
Epimeration reactions have been shown to play important roles in both bacterial and mammalian systems (1,2). However, the most studied epimerases are nucleotide-sugar epimerases of bacterial or fungal origin (2)(3)(4). Only a few human epimerases have been characterized at the molecular level. One of the best known human epimerases is the human UDPglucose 4-epimerase (EC 5.1.3.2), which catalyzes the conversion of UDP-glucose into UDP-galactose and has been found to be associated with the disease called galactosemia (5)(6)(7)(8).
The first report of a hydroxysteroid epimerase activity in animal systems was made several decades ago: a 16-epimerase acting on estriol and 16-epi-estriol in the human placenta was documented in 1968 (9). The activity that transforms 5␣-androstane-3␣,17␤-diol (3␣-diol) into its 3␤-epimer in the rat ovary was also reported, although it was speculated that the conversion was due to the combined actions of 3␣-HSD and 3␤-HSD 1 (10). An ecdysone 3-epimerase from the midgut of Manduca sexta (L.) was partially purified and was shown to act on position 3 to yield the 3-epimer of ecdysone (11). In addition, epimeric conversion of corticosteroids has also been documented for both human and hamster liver preparations (12). Despite the fact that the past few decades have seen rapid progress in the area of steroidogenesis at the molecular level, few reports can be found in recent literature on the cloning and characterization of steroid epimerases. The biological activity of steroid hormones and their metabolites is often associated with the stereo conformation of the molecule. A well documented example of this is the neuroactive steroid 5␣-pregnane-3␣-ol-20-one (allo-pregnanolone), one of the most abundant naturally occurring neuroactive steroids. It has been reported that this neurosteroid can play a role in the modulation of the reproductive function by suppressing the release of hypothalamic gonadotropin-releasing hormone in female rats (13). However, a recent report has demonstrated that its 3␤-epimer, 5␣-pregnane-3␤-ol-20-one (iso-pregnanolone), is ineffective in regulating the hypothalamic activities (14). Recent findings have also suggested that this molecule might be, at least in part, responsible for the premenstrual syndrome (15). Another example of inactivation of steroids via carbon-3 (C-3) epimeration is that of the secosteroid hormone 1␣,25dihydroxyvitamin D 3 , which, when converted to 1␣,25-epi-dihydroxyvitamin D 3 , becomes biologically much less active (16,17). These examples illustrate well the importance of C-3 epimeration in regulating the biological activities of steroid hormones.
In the present study, we report the cloning and characterization of a human cDNA encoding a 3(␣3␤)-hydroxysteroid epimerase (3(␣3␤)-HSE), which catalyzes the transformation of C-19 and C-21 steroids. To our knowledge, this is the first molecular characterization of a mammalian 3(␣3␤)-HSE reported to date.

EXPERIMENTAL PROCEDURES
Isolation of Human 3(␣3␤)-HSE-A cDNA fragment of 174 bp, corresponding to the conserved region situated near the N-terminal of the enzymes belonging to the RoDH family, was obtained by reverse transcription from human liver poly(A) ϩ (CLONTECH Inc., Palo Alto, CA) followed by 30 cycles of amplification using PCR and specific oligoprimer pairs (5Ј-GCC-GAA-TTC-GTG-GTG-AGC-CAT-CTA-CAT-3Ј and 5Ј-CGC-GGA-TCC-CAC-TGT-CTC-CAG-CCT-GTC-3Ј). The restriction sites EcoRI and BamHI were added to the oligoprimers for subsequent subcloning. The cDNA fragment thus obtained was used for the subsequent screening of a human liver gt11 cDNA library (CLON-TECH). The positive recombinant plaques were purified, and phage DNA was isolated by centrifugation for 90 min at 105,000 ϫ g, followed by phenol extraction. DNA inserts from positive phages were digested with EcoRI and ligated into pBluescript II SKϩ for sequencing with the T 7 sequencing kit (Amersham Pharmacia Biotech). The longest inserts containing the ATG codon and the poly(A) ϩ tail were cloned into a pCMV-neo expression vector for activity studies. Plasmid DNA was prepared using a Mega kit (Qiagen, Chatsworth, CA).
Site Epimerase Activity Assays-The determination of the enzyme activity in intact cells was done following the method previously described (19). Briefly, 0.1 M of the radiolabeled steroid was added to a freshly changed culture medium in a 24-well culture plate. When cell homogenates or 100,000 ϫ g fractions were used, the reaction was performed in FIG. 1. Nucleotide and putative amino acid sequences of the human 3(␣3␤)-HSE (AF223225). Numbers above the sequences correspond to the nucleotide sequence relative to the translation start site; numbers below the sequences correspond to the amino acid sequence. The AATAAA polyadenylation signal is underlined.
a 50 mM sodium phosphate buffer, pH 7.4, containing 20% glycerol, 1 mM EDTA, 0.1 M radioactive steroids, and 1 mM of the required cofactor. Incubations were performed at 37°C for 2 h. After the incubation, the steroids were extracted twice with 1 ml of ether. The organic phases were pooled and evaporated to dryness. The steroids were solubilized in 50 ml of dichloromethane, applied to a Silica Gel 60 thin layer chromatography (TLC) plate, then separated by migration in the toluene-acetone (4:1) solvent system. Metabolites were identified by comparison with reference steroids, revealed by autoradiography, and quantified using the PhosphorImager system (Molecular Dynamics, Sunnyvale, CA).
Study of Tissue Distribution by RT-PCR-Human poly(A) ϩ RNAs from different tissues, namely, liver, brain, prostate, adrenal, spleen, testis, uterus, placenta, and mammary gland (CLONTECH), were analyzed for the expression of the epimerase mRNA by semiquantitative RT-PCR. First strand cDNA was reverse-transcribed using poly(T) oligoprimer and a Superscript II kit manufactured by Life Technologies, Inc. and mRNA from the types of tissues mentioned above. One-tenth of the synthesized cDNA was used for PCR amplification. The conditions for PCR amplifications were: denaturation at 94°C for 1 min; annealing at 55°C for 1 min, and then elongation at 72°C for 1 min. Various numbers of cycles were tried for each tissue in an attempt to find the linear range of PCR amplification. The amplified products were analyzed on a 1.5% agarose gel and visualized by staining with ethidium bromide.

RESULTS
Cloning of 3(␣3␤)-HSE-Using a 174-bp probe, corresponding to the conserved region of the retinol dehydrogenase family to screen a human liver cDNA library, we have obtained several positive clones. Sequencing of these inserts identified one of them as a full-length cDNA encoding a putative protein of 317 amino acids (Fig. 1). The predicted protein sequence exhibits all the characteristic motifs found in the short chain alcohol dehydrogenase family (SDR), including the cofactor binding site Gly 36 -(Xaa) 2 -Ser-Gly-Xaa-Gly, the catalytic site Tyr 176 -(Xaa) 3 -Lys, and the residues of unknown function Leu 109 -Xaa-Asn-Asn-Ala-Gly (20). Moreover, five of the six conserved cysteine residues are found in the sequence, with a sixth nonconserved one at Cys 122 . Comparison of the deduced amino acid sequence with that of other SDR members (Fig. 2) revealed that human 3(␣3␤)-HSE shares 94.7% amino acid sequence identity with the human RoDH-like enzyme having an oxida-tive 3␣-HSD activity (21); 71.4% identity with the rat type 6 17␤-HSD (21); 67.2% and 67.5% identity, respectively, with the two newly reported human sterol/retinol dehydrogenases (22,23); 66 -67% identity with the three rat RoDHs (24 -26); and 50% sequence identity with human 11-cis-RoDH or 9-cis-RoDH (27,28).
Determination of Enzymatic Activity of the Recombinant Enzyme-Because the clone we obtained shows 94.7% identity with one of the previously reported enzymes, we wanted to determine whether our clone also encodes a protein exhibiting the same activity. To our surprise, when we incubated the microsomal fraction of HEK-293 cells stably overexpressing the enzyme with [ 3 H]ADT in the presence of NAD ϩ , we obtained two metabolites. One was 5␣-dione, the expected oxidative metabolite of ADT, whereas the second was identified as epi-ADT (Fig. 3A). The identity of this second metabolite was further confirmed by high performance liquid chromatography (data not shown). The results thus strongly suggest that this single enzyme possesses stereo-selectivity for the oxidative and reductive reactions. To further confirm that the reductive reaction preferentially produces the 3␤-OH isomer, we incubated the enzyme with [ 14 C]5␣-dione in the presence of NADH, as illustrated in Fig. 3B. The major end product was indeed 3␤diol (two-thirds of the products), as expected, whereas approximately one-third of the products was found to be ADT.
Determination of the Preferred Reaction: 3(␣3␤) or 3(␤3␣)-Because the 3(␣3␤)-HSE possesses differential stereo-selectivity, it is expected that the enzyme shows directional selectivity: 3(␣3␤) versus 3(␤3␣). To determine which direction is preferably catalyzed by the enzyme, we incubated the microsomal fraction of HEK-293 cells overexpressing 3(␣3␤)-HSE with 0.1 M ADT and epi-ADT as substrates in the presence of 1 mM of both oxidative (NAD ϩ ) and reductive (NADH) cofactors. As illustrated in Fig. 4, the enzyme catalyzes the conversion of ADT into epi-ADT approximately four times more than the transformation of epi-ADT into ADT, thus suggesting that the preferred orientation is 3(␣3␤). The kinetic data shown in Table I  of 13 and 1.17 for ADT and epi-ADT as substrates, respectively), whereas in the reductive reaction the 3␤-isomers are the major end products (V max /K m values of 12.5 and 3.5 for the reduction of 5␣-dione to epi-ADT and ADT, respectively).
Epimerase Activity in Intact Transfected Cells in Culture-Previously, we have shown that intact transfected cells in culture are more suitable for determining the preferred reac-tion of oxidoreductases and for characterizing their physiological activities (19,29,30). Thus, to further determine whether the enzyme could catalyze epimeration without the addition of exogenous cofactors and under conditions more similar to physiological conditions, we incubated HEK-293 cells stably overexpressing 3(␣3␤)-HSE in a culture medium with 0.1 M ADT and epi-ADT for various time periods. As illustrated in Fig. 5, the enzyme efficiently catalyzes the transformation of ADT into epi-ADT, whereas the transformation of epi-ADT into ADT is done at a much lower rate. The activity catalyzed by nontransfected HEK-293 under the same conditions was not significant.

Substitution of Amino Acid Residues in 3(␣3␤)-HSE with the Corresponding Ones in the RoDH-like Enzyme-Because
the reported human RoDH-like enzyme having oxidative 3␣-HSD activity shows only a few differences in amino acid residues with the 3(␣3␤)-HSE, we examined whether and how these changes in amino acid residues could affect the 3(␣3␤)-HSE activity. We thus performed site-directed mutagenesis to change the Glu at position 63 into Asp (E63D) and the Gly at position 105 into Arg (G105R) separately and in combination with the insertion and deletion of the nucleotide G to introduce the changes in 17 amino acid residues from 157 to 174 in the amino acid sequence. Expression vectors bearing these mutant sequences were analyzed for their activities. As illustrated in Table II, although the E63D substitution does not significantly affect the 3(␣3␤)-HSE activity, the G105R substitution and the changes in the 17 amino acid residues both completely inactivate the 3(␣3␤)-HSE activity of the recombinant enzyme.

FIG. 5. Determination of 3(␣3␤)-HSE activities in intact transfected HEK-293 cells in culture.
Cells stably expressing human 3(␣3␤)-HSE were seeded into 6-well plates at a density of approximately 500,000 cells/well. 0.1 M 3 H-labeled ADT or epi-ADT was added to freshly changed medium for testing the activities. After incubation for the indicated periods, the media were collected and extraction was performed using ethyl ether. The substrates and products were separated on TLC and quantified as described under "Experimental Procedures."   6 and 7) with the microsomal fraction of HEK-293 cells overexpressing 3(␣3␤)-HSE and mock-transfected cells, respectively. The reactions were performed at 37°C for 2 h in 50 mM phosphate buffer (pH 7.4) containing 1 mM both NAD ϩ and NADH. The reactions were stopped by adding 1 ml of ethyl ether. The procedures were as described under "Experimental Procedures." Histograms showing the percentage of the products in the corresponding reactions appear below the autoradiograph.
Tissue Distribution of the 3(␣3␤)-HSE mRNA-The tissue distribution of the epimerase mRNA was assessed by semiquantitative RT-PCR using poly(A) ϩ RNA from CLONTECH. As illustrated in Fig. 6, we detected the expression of epimerase transcript in various tissues, namely, adrenal, brain, liver, lung, mammary gland, placenta, prostate, testis, and uterus. Relative levels of the transcript were determined by varying the quantity of poly(A) ϩ RNAs for the reverse transcriptase reaction and the number of PCR amplification cycles to determine the linear range of PCR amplification for each tissue. It was determined that the highest level of expression is in the liver. In the other tissues, the order of the epimerase transcript levels is as follows: spleen Ͼ prostate Ͼ adrenal Ͼ brain Ͼ uterus Ͼ mammary gland Ͼ placenta (data not shown). No amplification product was detected from the testis tissue, even when the PCR product was probed by an epimerase-specific oligonucleotide with a longer exposure time.

DISCUSSION
In this report, we describe cloning and expression of a cDNA encoding an enzyme possessing 3(␣3␤)-HSE activity. Our results clearly indicate that the epimeration reaction is the result of catalysis by a single enzyme possessing both oxidative and reductive activities. As illustrated in Fig. 7, the overall reaction is composed of two steps, the oxidation of a 3␣-hydroxy group to a 3-keto group, followed by the reduction of this 3-keto group into a 3␤-hydroxy group. The 3-ketosteroid conformation is thus an intermediate in the 3(␣3␤)-hydroxysteroid epimeration. To our knowledge, this is also the first observation of an enzyme possessing stereo-selectivity for the oxidative and reductive reactions. Except for the 17 amino acids located from positions 158 -174, and 2 amino acid substitutions at positions 63 and 105 (E63D and G105R) (Fig. 2), the sequence of our clone is identical to that of the RoDH-like human oxidative 3␣-hydroxysteroid dehydrogenase, previously reported by Biswas and Russell (21). It is interesting to observe that the deletion and insertion of a G, respectively, at positions 467 and 520 relative to ATG in the sequence reported by Biswas and Russell (21), will transform these 17 different amino acid residues into the ones reported here.
To further understand the effect of these differences on the activity of the enzyme and to compare the epimerase activity reported here with the RoDH-like oxidative 3␣-HSD activity reported by Biswas and Russell (21), we performed site-directed mutagenesis and examined the activities of each as described in Table II. The E63D substitution, which does not affect 3(␣3␤) activity, is probably due to polymorphism. On the other hand, because the enzyme described by Biswas and Russell (21) is an active one, the G105R substitution and the differing 17 amino acid residues (resulting from the deletion and the insertion of a G at positions 467 and 520, respectively) can probably be explained by sequencing errors. Further evi-dence that the sequence described by Biswas and Russell (21) contains sequencing errors is a sequence submitted by Kedishvilli to GenBank under the accession number AF016509. This sequence is almost identical to ours except for two nucleotides located at positions 18 and 133 downstream from the ATG initiation codon. These changes do not alter the amino acid sequence. In addition, this sequence is probably obtained by PCR amplification, because it does not contain the 5Ј-and 3Ј-untranslated region. Therefore, the possibility that the difference of two nucleotides is due to amplification error is not to be excluded. Under the experimental conditions described by Biswas and Russell (21), when the period of incubation is very short, only one reaction can be detected and the direction of this  The epimerase reaction is a combination of a differential stereo-selective 3␣-HSD activity and of a 3-keto-reductase activity that, respectively, catalyze the transformation of ADT into 5␣-dione and the subsequent conversion of 5␣-dione to epi-ADT. reaction is determined by the added cofactor. Although this is a regular procedure used to analyze enzyme activities in vitro, it is not suitable for this particular enzyme, which catalyzes both the oxidative and the reductive reactions in a stereo-selective manner. This explains why these authors failed to associate both activities into one epimeration reaction. Although an in vitro assay using purified enzyme or preparations of subcellular fractions is a common procedure for performing kinetic or ligand binding studies, we believe that transfected intact cells are more suitable to characterize enzyme activities in living cells, especially for oxidoreductases. This way, the preferred reaction catalyzed by the enzyme, either oxidation or reduction, can be clearly identified (29), and unstable or labile enzymes can be studied with certainty (19). In the present study, using intact transfected cells in culture without the addition of exogenous cofactors, a system which better reflects the actual physiological conditions, we demonstrated clearly that 3(␣3␤)-HSE possesses an epimerase activity capable of converting a 3␣hydroxysteroid into its 3␤-hydroxy epimer (Fig. 5). This is further confirmed by in vitro assays in the presence of both NAD ϩ and NADH as cofactors (Fig. 4).
The physiological significance of the conversion of 3␣-hydroxysteroid by the epimerase has been demonstrated by recent findings regarding the differential biological activities of some stereoisomers of steroids. In fact, 3␣-diol and 3␤-diol have long been considered as breakdown products of androgens. However, recent studies provided evidence that 3␣-diol, but not its 3␤-epimer, might be an active hormone required for parturition in the rat (31). Furthermore, allo-pregnanolone, readily formed from progesterone in the brain, is one of the most potent neuroactive steroids. It can suppress the release of hypothalamic gonadotropin-releasing hormone via the ␥-aminobutyric acid A receptor, whereas its 3␤-epimer does not exert the same effect (14). Allo-pregnanolone is strongly suspected of being involved in the premenstrual syndrome (15). The conversion of 3␣-diol and allo-pregnanolone into their respective 3␤-isomers by 3(␣3␤)-HSE could inactivate these molecules. The detection of expression of the epimerase transcript in the brain provides evidence for the possible involvement of the epimerase in the regulation of allo-pregnanolone levels and, therefore, the possible implication of the epimerase in regulating hormonal changes in the brain.