cDNA Cloning, Expression, and Mutagenesis Study of Liver-type Prostaglandin F Synthase*

Prostaglandin (PG) F synthase catalyzes the reduction of PGD2 to 9α,11β-PGF2 and that of PGH2 to PGF2α on the same molecule. PGF synthase has at least two isoforms, the lung-type enzyme (K m value of 120 μm for PGD2 (Watanabe, K., Yoshida, R., Shimizu, T., and Hayaishi, O. (1985) J. Biol. Chem. 260, 7035–7041) and the liver-type one (K m value of 10 μm for PGD2 (Chen, L. -Y., Watanabe, K., and Hayaishi, O. (1992)Arch. Biochem. Biophys. 296, 17–26)). The liver-type enzyme was presently found to consist of a 969-base pair open reading frame coding for a 323-amino acid polypeptide with aM r of 36,742. Sequence analysis indicated that the bovine liver PGF synthase had 87, 79, 77, and 76% identity with the bovine lung PGF synthase and human liver dihydrodiol dehydrogenase (DD) isozymes DD1, DD2, and DD4, respectively. Moreover, the amino acid sequence of the liver-type PGF synthase was identical with that of bovine liver DD3. The liver-type PGF synthase was expressed in COS-7 cells, and its recombinant enzyme had almost the same properties as the native enzyme. Furthermore, to investigate the nature of catalysis and/or substrate binding of PGF synthase, we constructed and characterized various mutant enzymes as follows: R27E, R91Q, H170C, R223L, K225S, S301R, and N306Y. Although the reductase activities toward PGH2 and phenanthrenequinone (PQ) of almost all mutants were not inactivated, the K m values of R27E, R91Q, H170C, R223L, and N306Y for PGD2 were increased from 15 to 110, 145, 75, 180, and 100 μm, respectively, indicating that Arg27, Arg91, His170, Arg223, and Asn306 are essential to give a low K m value for PGD2 of the liver-type PGF synthase and that these amino acid residues serve in the binding of PGD2. Moreover, the R223L mutant among these seven mutants especially has a profound effect on k cat for PGD2 reduction. TheK m values of R223L, K225S, and S301R for PQ were about 2–10-fold lower than the wild-type value, indicating that the amino acid residues at 223, 225 and 301 serve in the binding of PQ to the enzyme. On the other hand, the K m value of H170C for PGH2 was 8-fold lower than that of the wild type, indicating that the amino acid residue at 170 is related to the binding of PGH2 to the enzyme and that Cys170 confer high affinity for PGH2. Additionally, the 5-fold increase in k cat/K m value of the N306Y mutant for PGH2 compared with the wild-type value suggests that the amino acid at 306 plays an important role in catalytic efficiency for PGH2.

F series prostaglandins (PG) 1 are widely distributed in various organs of mammals and exhibit a variety of biological activities including constriction of pulmonary arteries (1,2). PGF 2 is synthesized from PGE 2 , PGD 2 , or PGH 2 by PGE 9-ketoreductase, PGD 11-ketoreductase, or PGH 9,11-endoperoxide reductase, respectively. PGF synthase (EC 1.1.1.188) was purified from bovine lung by Watanabe et al. (3). It forms 9␣,11␤-PGF 2 (4) from PGD 2 (PGD 2 11-ketoreductase activity) and PGF 2␣ from PGH 2 (PGH 2 9,11-endoperoxide reductase activity) on the same molecule in the presence of NADPH (3,4). This enzyme catalyzes the reduction of other carbonyl compounds including 9,10-phenanthrenequinone (PQ) as well as that of PGD 2 and PGH 2 but does not catalyze the reduction of PGE 2 . Although PGD 2 11-ketoreductase activity was competitively inhibited by PQ, PGH 2 9,11-endoperoxide reductase activity was not inhibited by PGD 2 or PQ (3). PGF synthase belongs to the aldo-keto reductase family. The bovine lung PGF synthase is a monomeric protein with a M r of 36,666 consisting of 323 amino acids, and its amino acid sequence shows high homology compared with that of other aldo-keto reductase family members (5). PGF synthase has two isozymes, one in the lung (3) and the other one in the liver (6), with different K m values for PGD 2 (120 and 10 M, respectively). The regulation by metals, the sensitivity to chloride ions, the inhibition by CuSO 4 and HgCl 2 , and the profile of immuno-precipitation with anti-bovine lung PGF synthase antibody are different between the two isozymes (6). Although Kuchinke et al. (7) isolated a clone (PGFS II) of PGF synthase from bovine liver and determined its amino acid sequence, the 99% similarity with the amino acid sequence of lung PGF synthase and the high K m value for PGD 2 of this recombinant PGFS II indicated that its cDNA was that of the lung-type enzyme even though it had been isolated from liver. Until now, the primary structure of the liver-type PGF synthase and the amino acids related to the affinity for PGD 2 have not yet been defined.
In the present study, we describe the primary structure of the bovine liver-type PGF synthase and the enzymatic properties of the recombinant enzyme in COS-7 cells. Based on the comparison of the amino acid sequences among the liver-type and the lung-type PGF synthases and human liver DDs, several mutants were constructed, and their enzymatic properties were examined. The results of mutagenesis indicated the amino acid residues related to the binding sites of PGD 2 , PGH 2 , and PQ.
Internal Amino Acid Sequences of the Liver-type PGF Synthase-The bovine liver PGF synthase was purified to apparent homogeneity as described previously (6). S-Carboxymethylation and lysyl-endopeptidase digestion of the purified enzyme and the purification of fragments digested by lysyl-endopeptidase were done as described previously (5). The purified peptide fragments were sequenced by automated Edman degradation with an Applied Biosystems Inc. sequencer (Perkin-Elmer).
Degenerative reverse transcriptase-polymerase chain reaction (PCR) using mixed oligonucleotide primers was performed to obtain a partial cDNA fragment for screening of a bovine liver cDNA library. Mixed oligonucleotide primers were designed according to the amino acid sequences of peptides I, II, III, and IV shown in Fig. 1 and to the nucleotide sequence of the lung-type PGF synthase (5). Each primer was synthesized with an Applied Biosystems Inc. DNA synthesizer (Perkin-Elmer), and two sets of the sequences of sense and antisense primers were used as follows: (i) 5Ј-TTCCGCCATAT(CTA)GACAGT-GCT-3Ј (corresponding to peptide I, 21-mers) named P1 and 5Ј-GT-CAAACACCTG(TGA)ATATTCTC-3Ј (corresponding to peptide III, 21mers) named P2; (ii) 5Ј-GTGTCCAACTTCAACCACAAG-3Ј (corresponding to peptide II, 21-mers) named P3 and 5Ј-ATATTCTTC(AGC-T)ACAAATGGGTA-3Ј (corresponding to peptide IV, 21-mers) named P4. The reverse transcriptase-PCR was conducted with mRNA used as a template and primers P1/P2 for one PCR and P3/P4 for the other one under the conditions of denaturation at 98°C for 15 s, annealing at 65°C for 30 s, and elongation at 74°C for 30 s by KOD polymerase (Toyobo, Japan). After 20 cycles of PCR, the products were ethanolprecipitated and separated on 1% agarose gel. Two bands were recovered from the gel by use of a QIAEX II Gel Extraction Kit (Qiagen, Netherlands). Each band was ligated to a BlueScript SK II (ϩ) vector (Toyobo, Japan) by use of a ligation kit (Takara, Japan), and the resulting constructs were transfected into Escherichia coli DH10B competent cells (Life Technologies, Inc.). Plasmids were purified with a Qiagen plasmid purification kit and sequenced with an Applied Biosystems Inc. automated DNA sequencer 373A (Perkin-Elmer). Two bands, one of 720 base pairs (P1/P2) and one of 477 base pairs (P3/P4), encoded the internal cDNA and were used as two different probes for screening of the library.
The bovine liver cDNA library was constructed from 5 g of poly(A) ϩ RNA with SuperScript Plasmid System (Life Technologies, Inc.) for cDNA synthesis and a Plasmid Cloning Kit (Life Technologies, Inc.) according to the manufacturer's manual. The resulting constructs were transformed into ElectroMax, DH12S competent cells (Life Technologies, Inc.) by the electroporation method using a Gene-Pulser (Bio-Rad). The library yielded 2.0 ϫ 10 5 independent clones. Full-length cDNA clones were obtained by the colony hybridization method. All clones were spread on 20 sheets of nylon filters for the master filter, and then two filters were replicated from each master filter. The replicate filters were alkaline-denatured and fixed by baking at 80°C for 2 h. Two reverse transcriptase-PCR products, 720 and 477 base pairs described above, were randomly labeled by [␣-32 P]dCTP (111 TBq/mmol, Amersham Pharmacia Biotech, UK) with a Megaprime random primer labeling kit (Amersham Pharmacia Biotech, UK) and used as two probes for hybridization. After hybridization in 5ϫ SSCP, 0.1% SDS, 100 g salmon sperm DNA, and 10ϫ Denhardt's solution at 65°C for overnight, each filter was washed extensively twice in 2ϫ SSC, 0.1% SDS at room temperature for 5 min and twice in 0.5ϫ SSC, 0.1% SDS at 60°C for 30 min. Thirty two double-positive clones against the two different probes were obtained from 2.0 ϫ 10 5 independent clones. Six clones were picked up and sequenced with an Applied Biosystems Inc. automated DNA sequencer 373A. All clones coded for full-length cDNAs of bovine liver PGF synthase. One of these six clones was named pSPORT-BLiFS27.
Northern Blot Analysis-For Northern blot analysis, total RNA (20 g), which was isolated from bovine liver with a total RNA purification kit (Amersham Pharmacia Biotech, UK) according to the manufacturer's manual, was separated on 1.0% agarose gel and transferred to nylon membranes. The 477-base pair fragment (P3/P4 described above) was labeled with a BcaBEST Labeling Kit (Takara, Japan) according to the manufacturer's manual using [␣-32 P]dCTP and was used as a probe. The membranes were prehybridized in Rapid Hybridization buffer (Amersham Pharmacia Biotech, UK) at 65°C for 1 h and then hybridized for 2.5 h after addition of the radiolabeled probe. The membranes were washed for 5 min once at room temperature in 2ϫ SSC, 0.1% SDS, twice for 30 min at 65°C in 0.5ϫ SSC, 0.1% SDS, and twice for 30 min at 65°C in 0.2ϫ SSC, 0.1% SDS. Autoradiograms were examined with a BAS-2000 system analyzer (Fuji Film, Japan).
Expression of Bovine Liver PGF Synthase in COS-7 Cells and Purification of Its Expressed Protein-The bovine liver cDNA insert was removed from pSPORT-BLiFS27 by digestion with EcoRI and AflII, and its ends were blunted. The blunt-ended insert containing the complete coding region was subcloned into the blunt-ended BstXI sites of the pEF-BOS mammalian expression vector (20). Monkey COS-7 cells were cultured in Dulbecco's modified Eagle's medium (Nissui Co., Tokyo) containing 10% fetal calf serum. COS-7 cells (5 ϫ 10 6 cells) were transfected with 20 g of plasmid DNA by the electroporation method using a Gene Pulser (Bio-Rad). These cells were incubated in a 5% CO 2 incubator at 37°C for 72 h. The transfected COS-7 cells were sonicated in 3 volumes of 10 mM potassium phosphate buffer (KPB) (pH 7.0). The recombinant enzyme was purified by the method of Chen et al. (6) with a minor modification. The cytosol fraction of the homogenated cells, which was centrifuged at 100,000 ϫ g, was subjected to ammonium sulfate fractionation between 40 and 75% saturation. The precipitate formed was suspended in 500 l of 10 mM KPB (pH 7.0) and desalted by passage through a NAP-10 column. The desalted sample was applied to a Red Sepharose column, and the enzyme was eluted with 10 mM KPB (pH 7.0) containing 1 M KCl and 1 mM NADP. About 2.8-fold purification of the recombinant protein was achieved, and the apparent homogeneity was concluded following SDS-polyacrylamide gel electrophoresis (PAGE) and staining with Two-dimensional Silver Stain⅐II "Daiichi" (Dai-ichi Pure Chemicals Co., Ltd., Japan). A polyclonal antibody against PGF synthase was raised in a rabbit by the same procedure as described previously (3), with the enzyme purified from bovine liver used as the immunogen (6). For Western blot analysis, the purified enzyme was subjected to SDS-PAGE and electrophoretically transferred to a polyvinylidene difluoride membrane (Amersham Pharmacia Biotech, UK). Protein bands were immunostained with the anti-bovine liver PGF synthase antibody and reagents from a Vectastain ABC kit (Vector Laboratories) and visualized with an Enhanced Chemiluminescence Kit (ECL, Amersham Pharmacia Biotech, UK).
Enzyme Assay-The PGD 2 11-ketoreductase, PGH 2 9,11-endoperoxide reductase, and PQ reductase activities of the recombinant protein were measured as described previously (3). The standard assay mixture for PGD 2 11-ketoreductase contained 0.1 M KPB (pH 6.5), 0.5 mM NADP, 5 mM glucose 6-phosphate, glucose-6-phosphate dehydrogenase (1 unit), 1.5 mM [ 3 H]PGD 2 (3.7 KBq), and enzyme in a total volume of 50 l. Incubation was carried out at 37°C for 30 min. The PGH 2 9,11endoperoxide reductase activity was assayed under the same conditions as those of the PGD 2 11-ketoreductase activity except that 40 M [1-14 C]PGH 2 (4 MBq) was used as a substrate in place of 1.5 mM [ 3 H]PGD 2 and that the incubation time was 2 min. The PQ reductase activity was measured spectrophotometrically at 37°C by following a decrease in absorbance at 340 nm in the assay mixture consisting of 0.1 M KPB (pH6.5), 80 M NADPH, 10 M PQ, and enzyme in a total volume of 0.5 ml. One unit of enzyme activity was defined as the amount that produced 1 mol of PGF 2 per min at 37°C. Specific activity was expressed as the number of units/mg of protein. Protein was determined according to the method of Lowry et al. (21).
Site-directed Mutagenesis-A mutagenesis study was performed by the method of Jones et al. (22). The mutagenesis primers for the mutant R27E were designed as follows: LiFSPN, 5Ј-CAAACAATGGATCC-3Ј; LuFSP1, 5Ј-TTGCACCTGAGGAGGTTCC-3Ј; LuFSP2, 5Ј-CCTCCT-CAGGTGCAAAGGTT-3Ј; LiFSPC, 5Ј-GAATGCACGTGTACAGCT-3Ј. The bold letters of LuFSP1 and LuFSP2 were the sites of mutagenesis. As shown in Fig. 1, one PCR between LiFSPN and LuFSP2 and the other one between LuFSP1 and LiFSPC were conducted with pSPORT-BLiFS27 harboring the full-length cDNA for bovine liver PGF synthase as a template. The conditions of PCR were described above. After 20 cycles of PCR, the products were purified by 1% agarose gel electrophoresis, and the purified products were treated with Klenow fragment of DNA polymerase I after recovery. These products were mixed at the ratio of 1 to 1, and the second PCR was conducted using LiFSPN and LiFSPC as described above. The product of the second PCR was bluntended and then was ligated to the blunted BstXI sites of the pEF-BOS expression vector. Consequently, the mutant of R27E was formed. The other mutant enzymes were formed by the same procedure as used for the R27E mutant.

RESULTS
cDNA Cloning of Bovine Liver PGF Synthase-Screening of 2.0 ϫ 10 5 clones with the two probes (P1/P2 and P3/P4 products described under "Experimental Procedures") gave 32 doublepositive clones, and six independent clones of these 32 clones were picked up. DNA sequencing confirmed that these clones coded for full-length cDNAs of bovine liver PGF synthase. The deduced amino acid sequences of bovine liver PGF synthase contained all the amino acid sequences of the nine peptide fragments obtained from the native bovine liver enzyme (Fig.  1). The bovine liver PGF synthase cDNA clone BLiFS27 contained a polyadenylation signal after the stop codon ( Fig. 1), showing that it coded for a full-length PGF synthase. BLiFS27 contained an open reading frame of 969 base pairs coding for 323 amino acids. The calculated M r of the bovine liver enzyme was 36,742, a value similar to that of the native enzyme, which is about 36 kDa (6). The identity between the liver and lung enzymes was 87% at the amino acid level and 90% at the nucleotide level. As shown in Fig. 2, this enzyme showed a high identity in amino acid sequence with not only bovine lung PGF synthase (87%) but also human liver DD1 (79%), DD2 (77%), and DD4 (76%). Moreover, its sequence was identical with that of bovine liver DD3.
Identification and Size Determination of Bovine Liver PGF Synthase mRNA by Blot Hybridization Analysis- Fig. 3A shows the result of the Northern blot analysis of bovine liver mRNA with the P3/P4 PCR product (477 base pairs) used as a probe. From its migration in a denaturing gel system, the sequence of PGF synthase mRNA from bovine liver was estimated to be 1400 nucleotides. Therefore, assuming a length for the poly(A) tail of 100 -150 nucleotides, the insert cDNA sequence of 1139 nucleotides extended nearly the full length of the mRNA.
Expression of Bovine Liver PGF Synthase in COS-7 Cells and Purification of Its Expressed Protein-pBOS-BLiFS27 carrying the full-length bovine liver PGF synthase in the mammalian expression vector pEF-BOS was prepared by use of the strategy described under "Experimental Procedures." COS-7 cells were transfected with pBOS-BLiFS27. The recombinant protein was expressed transiently in COS-7 cells and was located in the cytosol fraction. The recombinant protein was purified to an apparent homogeneity by the following purification steps: ammonium sulfate fractionation between 40 and 75% saturation and Red Sepharose column chromatography, as described under "Experimental Procedures." About 2.8-fold purification of the PGD 2 11-ketoreductase activity was achieved from the cytosol of COS-7 cells with a yield of 44% (Table I). A sample of each of the purification steps was subjected to SDS-PAGE. Silver staining of the gel indicated that an approximately 36.7-kDa protein was produced in the cells harboring pBOS-BLiFS27, and this protein was purified to an apparent homogeneity (Fig. 3B). Western blot analysis of each sample revealed that the 36.7-kDa protein was recognized by antibovine liver PGF synthase antibody (Fig. 3C). The molecular weight of the expressed enzyme was the same as that of the native enzyme, as shown in Fig. 3, B and C. No protein from the control COS-7 cells bearing pEF-BOS without the insert DNA interacted with this antibody (data not shown).
The purified recombinant protein exhibited enzymatic properties similar to those of the native enzyme. The K m values for PGD 2 , PGH 2 , and PQ were 15, 25, and 1.1 M, respectively (Table II). The low K m value for PGD 2 was essentially identical to that of the native liver-type enzyme and was different from the high K m value (120 M) of the lung-type one. The specific activities of the recombinant protein for PGD 2 , PGH 2 , and PQ were 23, 9, and 186 milliunits/mg of protein, respectively. However, PGE 2 was not reduced. Moreover, the IC 50 value for inhibition of PGD 2 11-ketoreductase activity of the expressed enzyme by CuSO 4 was 0.5 mM (data not shown), which was almost the same as that for inhibition of the native liver-type enzyme (0.4 mM) but unlike that for the lung-type enzyme (0.003 mM) (6). These results indicate that the cloned cDNA coded for the liver-type PGF synthase and not for the lung-type synthase.
Site-directed Mutagenesis-The comparison of amino acid sequences among PGF synthases of bovine liver and lung, and human DD1, DD2, and DD4 is shown in Fig. 2. The K m values of the lung-type (3) and the liver-type (6) PGF synthases and the human liver DD1 and DD2 for PGD 2 (16) were 120, 10, 12, and 79 M, respectively. DD4 does not catalyze the reduction of PGD 2 (16). To determine which amino acid residues were related to PGF synthase activity, especially to PGD 2 11-ketoreductase activity, we conducted a site-directed mutagenesis study. Arg 27 , Arg 91 , Arg 223 , Lys 225 , Ser 301 , or Asn 306 of the liver-type PGF synthase with a low K m value for PGD 2 was changed to Glu, Gln, Leu, Ser, Arg, or Tyr, respectively, the latter of which are the residues of the lung-type PGF synthase with a high K m value for PGD 2 . In addition to these mutations, His 170 was changed to the Cys of DD4 (Fig. 2), which has no PGD 2 11-ketoreductase activity. The mutant enzyme, R27E, R91Q, H170C, R223L, K225S, S301R, or N306Y, was expressed in COS-7 cells and was purified to an apparent homogeneity (Fig. 4) as described under "Experimental Procedures." The results of their final purification step are shown in Table I. The k cat , K m , and k cat /K m values for three representative substrates, i.e. PGD 2 , PGH 2 , and PQ, of the purified mutant enzymes are shown in Table III.
Although the k cat /K m values of almost all mutants for PGH 2 and PQ were retained above 50% of the wild-type value, these values of all mutants for PGD 2 decreased. The k cat /K m values of R27E, R91Q, H170C, and N306Y for PGD 2 were only about 10% that of the wild type, and the K m values of R27E, R91Q, H170C, R223L, and N306Y for PGD 2 were 110, 145, 75, 180, and 100 M, respectively. These K m values were 5-10-fold higher than the wild-type value and were almost the same as that of the lung-type PGF synthase. Considering the amino acid residues of these mutants were changed from the livertype PGF synthase to the lung-type synthase for R27E, R91Q, R223L, and N306Y or to DD4 for H170C, these results suggest that Arg 27 , Arg 91 , His 170 , Arg 223 , and Asn 306 are essential to give a low K m value for PGD 2 and that these amino acid residues play an important role on the binding for PGD 2 to PGF synthase. In addition, the R223L mutant increased k cat for PGD 2 5-fold, indicating that the amino acid residue at 223 has a profound effect on k cat for PGD 2 reduction. The k cat /K m values of R27E, R91Q, and N306Y mutants for PQ were 50 -80% of that value of the wild type, indicating that Arg 27 , Arg 91 , and Asn 306 have little effect on the catalytic efficiency for PQ, less than that of these residues on the catalytic efficiency for PGD 2 . Moreover, the K m values of R223L, K225S, and S301R for PQ were about 2-10-fold lower than the wild-type value, and the k cat /K m values of these mutants for PQ were 3-15-fold higher than the value of the wild type. These results suggest that the amino acid residues at 223, 225, and 301 are related to the binding for PQ to the enzyme and that the binding to the carbonyl group is different between PG and quinone compounds. On the other hand, the K m value of H170C for PGH 2 was 8-fold lower than that of the wild type, and the k cat /K m values of H170C and N306Y for PGH 2 were 5-6-fold higher than that of the wild type. These results indicate that the amino acid residue at 170 is related to the binding for PGH 2 and that Cys 170 seems to confer greater affinity for PGH 2 than His. Moreover, the amino acid residue at 306 plays an important role in catalytic efficiency for PGH 2 . DISCUSSION We isolated a clone of the liver-type PGF synthase with a low K m value for PGD 2 from the cDNA library of bovine liver, expressed the enzyme in COS-7 cells, and constructed seven mutants. Moreover, we examined the enzymatic properties of the wild-type enzyme and of the mutant enzymes, and we investigated the amino acid residue(s) related to the affinity of PGF synthase for the substrates.
The amino acid sequence of the liver-type PGF synthase consisted of 323 amino acid residues with a M r of 36,742 (Fig.  1) and showed 87% identity with that of the lung-type synthase (Fig. 2). When the liver-type enzyme was expressed in COS-7 cells (Fig. 3), the recombinant purified protein (Fig. 3, B and C, and Table I) was essentially identical to the native liver-type enzyme and not to the lung-type enzyme, based on the enzymatic properties (Table II) including the low K m value (15 M) for PGD 2 and the results of Western blot analysis (Fig. 3).
Recently, Jez et al. (23) reported on a structure/function analysis of the aldo-keto reductase superfamily. They reported that five amino acid residues, i.e. Asp 50 , Asn 167 , Gln 190 , Ser 271 , and Arg 276 , and three residues, i.e. Asp 50 , Tyr 55 , and Lys 84 , function in cofactor binding and in the active site, respectively. Based on the locations of the cofactor-binding pocket and the active site, a putative substrate-binding site was also proposed. However, the binding site for PGs has not yet been reported. The amino acid sequence of the liver-type PGF synthase was highly homologous with those sequences of DD1, DD2, and DD4 of human liver (Fig. 2), and DD1 and DD2 exhibited PGD 2 11-ketoreductase activity (16). Based on the comparison among the amino acid sequences of human liver DDs (15,16) and the lung-type and the liver-type PGF synthases, we studied the site-directed mutagenesis to change seven amino acid residues as follows: R27E, R91Q, H170C, R223L, K225S, S301R, and N306Y. All mutants expressed the proteins in COS-7 cells to almost the same extent as the wild-type protein, suggesting that the tertiary structures of these mutants were not drastically changed. Moreover, all mutants retained the activity with above 50% of k cat /K m values of the wild type for PGH 2 and PQ, indicating that the structures of all mutants were conserved. Therefore, the change in the enzymatic properties of the mutants reflected the mutation of the amino acid residue and not a change in the structure. Judging from the results of this study, Arg 27 , Arg 91 , His 170 , Arg 223 , and Asn 306 are essential to give a low K m value for PGD 2 and play an important role in the binding of PGD 2 ; and the amino acid residue at 223 has a significant effect on k cat for PGD 2 reduction. Moreover, the results of K m of the R223L, K225S, and S301R mutants for PQ and PGD 2 show that R223L, K225S, and S301R mutations acquired high binding for PQ and low for PGD 2 . Leu 223 , Ser 225 , and Arg 301 of the lung-type PGF synthase have a positive effect  R27E, R91Q, H170C, R223L, K225S, S301R, and N306Y mutants All activity measurements were performed under standard assay condition for the PGD 2 11-ketoreductase activity as described under "Experimental Procedures." One milliunit of enzyme activity was determined as the amount that produced 1 nmol of PGF 2 /min at 37°C. and native lung PGF synthase Enzyme assays of expressed liver PGF synthase with PGD 2 , PGH 2 , and phenanthrenequinone were conducted under standard assay conditions including the enzyme (3.4 g for PGD 2 , 17 g for PGH 2 , and 3.4 g for PQ), and measured by the radioisotope method for PGD 2 and PGH 2 and by the spectrophotometric method for phenantherenequinone at 37°C as described under "Experimental Procedures." One milliunit of enzyme activity was determined as the amount that produced 1 nmol of PGF 2 /min at 37°C. Enzymes purified to apparent homogeneity were used as enzyme sources.  on the binding of PQ. In terms of the binding site for PGH 2 , the amino acid residue at 170 seems to be related to the binding of this PG and that at 306 plays an important role in k cat /K m of this PG.
Twelve amino acid residues among the 42 amino acid residues of the liver-type enzyme that differed from those of the lung-type one were located in the ␣-helix or ␤-sheet, and the other 30 amino acid residues were located in loop structures of the tertiary structure of the liver-type PGF synthase, as inferred from data on human aldose reductase (24,25) and rat 3␣-HSD (26), which showed 46 and 70% identity, respectively, in terms of amino acid sequence with bovine liver-type PGF synthase. As a general rule, an amino acid(s) located in an ␣-helix or a ␤-sheet is involved in supporting the tertiary structure and that in loop structures of aldo-keto reductases is related to define substrate specificity (23,27). Although His 170 is located in an ␣-helix of the inferred tertiary structure of the liver-type PGF synthase, the amino acid residue at this position in the aldo-keto reductase family shows variation and is located near Asn 167 , which makes hydrogen bonds with the carboxyamide moiety of the cofactor (23). His 170 was changed to Cys of DD4, which has no PGD 2 11-ketoreductase activity. The k cat /K m values of this mutant for PGD 2 was 16% that of the wild type, and its 1/K m value decreased to about 20% that of the wild type. On the other hand, the k cat /K m values of this mutant for PGH 2 were about 6-fold that of the wild type, and its 1/K m value of this mutant increased to 8-fold that of the wild type. These results indicate that the amino acid residue at 170 seems to be related to the binding for PGD 2 and to that for PGH 2 . The reverse effects of the H170C mutant on PGD 2 and PGH 2 reductase activities indicate that His and Cys at 170 play an important role in the binding of PGD 2 and PGH 2 .
Arg 27 , Arg 91 , Arg 223 , Lys 225 , Ser 301 , and Asn 306 located in a loop structure are related to the binding site of PGD 2 , PQ, or PGH 2 . Jez et al. (27) reported that Trp 86 and Trp 227 of the 3␣-HSD near the active site may have roles in substrate binding. They proposed that Trp 86 is important in binding to a steroid ligand, whose A-ring lies between this Trp and the cofactor, and that Trp 227 interacts with the C-and/or D-rings of steroid ligands. The effect of R91Q on PGD 2 taken together with the report on Trp 86 of the 3␣-HSD suggests that Arg 91 near Trp 86 is a part of the substrate-binding pocket for PGD 2 . Moreover, although the effects of W227Y on the k cat /K m values for steroids are dramatic, smaller effects of this mutant on those for one-, two-, and three-ring substrates are also shown in this order (27). PGD 2 has one cyclopentane ring and two long tails (-and ␣-chains), and Arg 223 and Lys 225 are also located near Trp 227 . The results of the R223L mutant for PGD 2 suggest that Arg 223 is an essential amino acid residue to give a low K m value for PGD 2 of the liver-type PGF synthase and that the amino acid residue at 223 plays an important role in k cat of PGD 2 reduction. The results of the R223L and K225S mutants for PQ suggest that Leu 223 and Ser 225 have a significant effect on the binding of PQ. PGD 2 and PQ seem to bind to the same apolar pocket of PGF synthase differently, like steroids, nonsteroidal anti-inflammatory drugs, and aldose reductase inhibitors of 3␣-HSD (27). Moreover, the effect of S301R on PQ suggests that Ser 301 also has a role in the binding of PQ to the enzyme. Asn 306 , located in the C-terminal region, may have structural importance for the binding of the substrate. Members of the NADPH-dependent aldo-keto reductase superfamily are in part distinguished by unique C-terminal loops (28 -30). C-terminal loops of aldo-keto reductase are unique for each member and differ drastically in length and amino acid composition; and the C-terminal loop is critical for catalytic efficiency and for substrate and inhibitor specificity (28 -30). In the case of PGF synthases, the C-terminal region may be also critical for the affinity and specificity for the substrate. 2 The effect on the binding of the N306Y mutant for PGD 2 and that on k cat /K m of this mutant for PGH 2 also suggest that Asn 306 is a critical determinant of PGs.
Considering the one binding site for NADPH in the inferred tertiary structure of the liver-type PGF synthase, the catalytic site for the various substrates of aldo-keto reductase including PGF synthase seems to be the same. Tyr 55 of aldo-keto reductase is favored as the catalytic acid in the reaction mechanism (23). Tyr 55 of PGF synthase also plays the same role as that of the corresponding position in the other aldo-keto reductases. 2 The Y55F mutation eliminated PGD 2 , PGH 2 , and PQ reductase activities of PGF synthase completely. 2 However, although PGD 2 and PQ reductase activities of Y55Q were eliminated completely, only PGH 2 reductase activity of this mutant was stimulated, 27-fold. 2 These results suggest that Tyr 55 is favored as the catalytic residue, but the reduction mechanism, which involves a hydride transfer from NADPH to the substrate and protonation of the oxygen by a residue of the enzyme acting as a general acid, may be different between PGH 2 and PGD 2 /PQ. The reduction site of PGD 2 is the keto group like other carbonyl compounds but that of PGH 2 is the endoperoxide group. Moreover, the results of the reverse effects of H170C and N306Y on PGD 2 and PGH 2 reductase activities shown in this paper, taken together with the results on the competitive inhibition between PGD 2 and PQ (3) and on the lack of inhibi-  2 , and phenanthrenequinone among expressed liver PGF synthase and mutants The enzyme activities for PGD 2 , PGH 2 , and phenanthrenequinone were measured in the presence of wild-type PGF synthase and seven mutants (0.06 -15 g for PGD 2 , 0.3-19 g for PGH 2 , and 0.3-10 g for PQ) by the methods shown in Table II tion between PGH 2 and PGD 2 /PQ (3,6), indicate that the reduction mechanism for PGD 2 may be the same as that for PQ but not the same as that for PGH 2 . 3␣-HSD, which catalyzes the NADP-dependent reversible oxidation of the 3␣-hydroxy group of various steroids, interacts extensively with bile acids (31); and it and human liver high affinity bile acid-binding protein with minimal 3␣-HSD activity are multifunctional proteins in bile acid transport and xenobiotic metabolism (32). The amino acid sequences of HSD and high affinity bile acid-binding protein are similar to the sequence of bovine liver PGF synthase, and DD2 and DD4 exhibited binding activity for bile acids (33). Therefore, bovine liver PGF synthase may also be expected to exhibit the ability to bind bile acids. In the liver, PGF 2␣ , PGE 2 , and PGD 2 reduce bile flow and bile acid secretion, and especially the effect of PGF 2␣ is more potent than that of PGE 2 or PGD 2 (34). PGF synthase may be multifunctionally involved in the biosynthesis of PGF and in the binding of bile acids, and PGF synthase in the liver may reduce bile flow. Moreover, PGF 2␣ stimulates hepatocyte DNA synthesis and may have a role in promoting hepatocyte proliferation (35,36). Furthermore, PGF 2␣ and PGD 2 are released from primary Ito cell cultures after stimulation by noradrenaline and ATP (37). PGF 2␣ plays an important physiological role in the liver, and the liver-type PGF synthase mainly contributes to the biosynthesis of PGF 2␣ there.
Recently, the sequence of the bovine liver DD3 (17) was reported. The sequence of the liver-type PGF synthase reported in this paper was found to be identical to that of the bovine liver DD3. The liver-type PGF synthase showed the enzyme activity for (S)-(ϩ)-indanol (6.3 mol/min/mg), 3 which is a typical substrate of DD3 (13,17). However, PG(s) is the naturally occurring substrate(s) for this enzyme, as indanole is a xenobiotic compound.