Structural and Mutational Analysis of Trypanosoma brucei Prostaglandin H 2 Reductase Provides Insight into the Catalytic Mechanism of Aldo-ketoreductases*

Trypanosoma F 2 synthase is an aldo-ketoreductase that catalyzes the reduction of prostaglandin H 2 to PGF 2 (cid:1) in addition to that of 9,10-phenan- threnequinone. We report the crystal structure of TbPGFS (cid:1) NADP (cid:2) (cid:1) citrate at 2.1 Å resolution. TbPGFS adopts a parallel ( (cid:1) / (cid:3) ) 8 -barrel fold lacking the protrudent loops and possesses a hydrophobic core active site that contains a catalytic tetrad of tyrosine, lysine, histidine, and aspartate, which is highly conserved among AKRs. Site-directed mutagenesis of the catalytic tetrad residues revealed that a dyad of Lys 77 and His 110 , and a triad of Tyr 52 , Lys 77 , and His 110 are essential for the reduction of PGH 2 and 9,10-PQ, respectively. Structural and kinetic analysis revealed that His 110 , acts as the general acid catalyst for PGH 2 reduction and that Lys 77 facilitates His 110 protonation through a water molecule, while exerting an electrostatic repulsion against His 110 that maintains the spatial arrangement which allows the formation of a hydrogen bond between His 110 and C 11 carbonyl of PGH 2 . We also show that Tyr 52 acts as the general acid catalyst for 9,10-PQ reduction, and thus we not only elucidate the catalytic mechanism of a PGH 2 reductase but also provide an insight into the catalytic specificity of AKRs. at 37 °C. Blanks without enzyme or without substrate were also in- cluded. k cat and K m values were calculated from initial velocity studies over a wide range of pH values using the triple buffer system containing 50 m M sodium phosphate, 50 m M sodium pyrophosphate, and 50 m M 3-[(1,1-dimethyl-2-hydroxyethyl)amino]-2-hydroxypropanesulfonicacid.Thereactionwasinitiatedbytheadditionofasmallaliquotofenzymeinphosphate-bufferedsaline,andnosignificantchangeinthepHofthereactionmixturewasobservedbeforeoraftertheadditionofenzyme. UV and CD Spectroscopy— CD spectra of the wild-type and mutant enzymes were recorded in Tris/Cl buffer (pH 8.0) on a J-600 spectropo-larimeter (Jasco International Co.) equipped with a 1-mm path length cuvette. The temperature in the cell holder was maintained at 10 °C by a circulating water thermostat. The reported spectra were the mean of 10 scans at 50 nm (cid:1) min (cid:2) 1 . Fluorescence Quenching Assay— To determine the binding of NADPH to wild-type and mutant TbPGFS, we mixed various concentrations of NADPH with TbPGFS in a 200 (cid:3) l reaction mixture that contained 50 m M sodium phosphate (pH 7.0). After incubation at 37 °C for 2 min, the intrinsic tryptophan fluorescence was measured by an FP-750 spectroflu- orometer (Jasco International Co.) at an excitation wavelength of 282 nm and an emission wavelength of 338 nm (37). The quenching of tryptophan fluorescence caused by nonspecific interactions with each ligand was corrected with 1.5 (cid:3) M N -acetyl- L -tryptophanamide. The K d value for NADPH binding to TbPGFS was calculated as described by Levine (38).

Trypanosoma brucei prostaglandin F 2␣ synthase is an aldo-ketoreductase that catalyzes the reduction of prostaglandin H 2 to PGF 2␣ in addition to that of 9,10-phenanthrenequinone. We report the crystal structure of TbPGFS⅐NADP ؉ ⅐citrate at 2.1 Å resolution. TbPGFS adopts a parallel (␣/␤) 8 -barrel fold lacking the protrudent loops and possesses a hydrophobic core active site that contains a catalytic tetrad of tyrosine, lysine, histidine, and aspartate, which is highly conserved among AKRs. Site-directed mutagenesis of the catalytic tetrad residues revealed that a dyad of Lys 77 and His 110 , and a triad of Tyr 52 , Lys 77 , and His 110 are essential for the reduction of PGH 2 and 9,10-PQ, respectively. Structural and kinetic analysis revealed that His 110 , acts as the general acid catalyst for PGH 2 reduction and that Lys 77 facilitates His 110 protonation through a water molecule, while exerting an electrostatic repulsion against His 110 that maintains the spatial arrangement which allows the formation of a hydrogen bond between His 110 and C 11 carbonyl of PGH 2 . We also show that Tyr 52 acts as the general acid catalyst for 9,10-PQ reduction, and thus we not only elucidate the catalytic mechanism of a PGH 2 reductase but also provide an insight into the catalytic specificity of AKRs.
Earlier structural and functional studies on AKRs have elucidated the catalytic mechanism of the oxidation/reduction reaction for physiological substrates such as monosaccharides (8 -10), steroid hormones (4), and aldehydes (1,11). These studies revealed that bacterial and mammalian AKRs catalyze the oxidation/reduction through a catalytic mechanism that involves a catalytic tetrad of aspartate, tyrosine, lysine, and histidine. Although the roles of these residues on catalysis of the oxidation/reduction reaction for a number of substrates have been elucidated, little is known about the way they affect the oxidation/reduction of prostaglandins. Nonetheless, the positional and evolutionary conservation of the residues of this catalytic tetrad in AKRs from different biological kingdoms would suggest that this catalytic mechanism is probably common to all AKRs. PGF 2␣ is one of the earliest discovered prostanoids for which the synthetic pathways and biological functions have been investigated extensively. PGF 2␣ is synthesized by NADPH-dependent reduction catalyzed by AKRs of either the 9,11-endoperoxide moiety of PGH 2 or the 9-keto group of PGE 2 (12,13). PGF 2␣ synthase (EC 1.1.1.188) was first isolated from mammals (14), and the mammalian enzymes, which belong to the 1C subfamily of the AKRs (see Refs. 15 and 16 for AKR superfamily classification), catalyze the reduction of PGH 2 to PGF 2␣ and PGD 2 to 9␣,11␤-PGF 2 (a stereoisomer of PGF 2␣ ) (17), in addition to the oxidation of 9␣,11␤-PGF 2 to PGD 2 (18,19). In mammals, PGF 2␣ is a potent mediator of various physiological processes (20 -22) including regulation of vascular tone, constriction of uterine muscle (23) and pulmonary arter-ies (24,25), and induction of luteolysis during the estrous cycle and prior to parturition (26,27). During pathological processes in mammals, PGF 2␣ overproduction causes ovarian dysfunction and miscarriage (28 -30).
We have identified a unique PGF 2␣ synthase from the protozoan parasite Trypanosoma brucei (31), the etiological agent of African trypanosomiasis in humans and animals. T. brucei PGF 2␣ synthase belongs to 5A subfamily of the AKRs (16,31). The uniqueness of this synthase stems from the discovery that it exhibits a high specificity toward PGH 2 and shows no enzymatic activity toward substrates such as PGD 2 , PGE 2 , or PGF 2␣ . TbPGFS catalyzes the NADPH-dependent reduction of PGH 2 to PGF 2␣ in addition to the reduction of 9,10-PQ (a non-physiological substrate) and has the lowest K m value for PGH 2 among the various AKRs. Although it has been shown that PGF 2␣ elicits vacuole formation and thus may play the signal coupling role during phagocytosis in the protozoan parasite Amoeba proteus (32), little is known as to why T. brucei produces PGF 2␣ . However, because African trypanosomiasis is characterized by miscarriage due to PGF 2␣ overproduction correlated with parasitemia peaks (29), findings that T. brucei possesses a TbPGFS suggest that this enzyme may well play a role in the pathogenesis of trypanosomiasis. Hence, the elucidation of TbPGFS structure and mechanism of catalysis is of importance because it may lead to the design of specific inhibitors needed for investigating the physiological role(s) of the enzyme in this organism.
Here, we describe the x-ray crystallographic structure of the TbPGFS⅐NADP ϩ ⅐citrate ternary complex at 2.1 Å resolution. Tb-PGFS structure showed the positional and evolutionary conservation of Asp 47 , Tyr 52 , Lys 77 , and His 110 in the catalytic pocket. Although the tyrosine residue has been essential for catalysis in most AKRs, our mutagenesis experiments on the catalytic tetrad revealed that TbPGFS required Lys 77 and His 110 but not Tyr 52 for the biological reduction of PGH 2 to PGF 2␣ , whereas it needed Tyr 52 , Lys 77 , and His 110 for the reduction of 9,10-PQ. Thus, we propose a catalytic mechanism by which a catalytic dyad of Lys 77 and His 110 in TbPGFS is involved in proton transfer to the 9,11-endoperoxide moiety of PGH 2 .

MATERIALS AND METHODS
Structure Determination and Refinement-Initially, we obtained some crystals with ammonium sulfate as a precipitant (33). After observing that citrate inhibited TbPGFS activity, we then crystallized TbPGFS in the presence of 1.4 M sodium citrate (used as a precipitant) at pH 7.0. These cryocooled crystals were obtained in the presence of trehalose, and the x-ray diffraction data were obtained up to 2.1 Å resolution at SPring-8 beam-line 12B2.
The structure of TbPGFS was determined by a combination of molecular replacement methods using the AMoRe program (34). The structure of Sus scrofa aldose reductase (Protein Data Bank code 1AH3) was used as a search probe (41% identity). After a rotational search at 3.0 Å resolution, we obtained two solutions corresponding to the two molecules in the asymmetric unit that had a correlation coefficient of 52.1% and R-value of 46.4%. Subsequent cycles of the structure refinement were done with the TURBO-FRODO program (35) by alternating the model building after the rigid body minimization that allowed a complete polypeptide chain tracing. The quality of the final model was assessed from Ramachandran plots and by the analysis of the model geometry. The plot indicated that 87.7% of the residues were in the favorable regions, and 11.1% of the residues were in the allowed regions. Crystallographic refinement was carried out with the CNS program (36). The refinement procedure included simulated annealing, positional refinement, restrained temperature factor refinement, and maximum likelihood algorithms as provided by the CNS program. Water molecules were inserted manually and then checked by inspecting the F o Ϫ F c map. NADP ϩ and the citrate anion could be inserted into the clearly defined electron density model only after the first cycles of refinement. All residues in the asymmetric unit were similar and exhibited a root mean square deviation of 0.26 Å upon structural alignment.
Site-directed Mutagenesis-For point mutations of TbPGFS, highly conserved amino acid residues to be mutated were selected based on the detailed structural information of the TbPGFS⅐NADP ϩ ⅐citrate ternary complex. Polymerase chain reaction mutagenesis was carried out to generate the appropriate mutated cDNAs. A series of sense (F) and antisense (R) primers containing mismatched codons for each mutant were designed for PCR mutagenesis. Site-directed mutagenesis was used to produce single (D47N, Y52F, K77L, K77R, H110A, and H110F), double (D47N/H110A, Y52F/K77L, Y52F/H110A, and K77L/H110A), and triple (Y52F/K77L/H110A) mutants. The single mutant enzymes were obtained by using the following respective oligonucleotide primer pairs: ( . Wild-type TbPGFS open reading frame from pGEX4T-1 was subcloned into pUC18, and the resulting Wt-pUC18 plasmid was used as template for PCR amplification. The first PCR, amplified with Pyro-Best polymerase (Takara Shuzo), was carried out with gene-specific sense F-primer for each mutant and antisense M13 R-primer, as well as with sense M13 F-primer and gene-specific antisense R-primer for each mutant. The resultant products from the first PCR of each mutant were mixed and used as a template for the second PCR, which used genespecific sense F-primer of each mutant and antisense M13 R-primer, as well as gene-specific antisense R-primer for each mutant and sense M13 F-primer. Products from the second PCR of each mutant were cloned into pUC18 and sequenced. Subsequently, the open reading frames that contained mismatched codons were cloned in pGEX4T-1, and their sequences were checked.
Expression and Purification of Mutant Enzymes-The coding region of wild-type and mutant TbPGFS cDNAs that carried EcoRI and SalI restriction sites at their respective 5Ј-end were cloned into the corresponding sites of the pGEX-4T-1 expression vector (Amersham Biosciences) to prepare TbPGFS-glutathione S-transferase fusion proteins. The resultant recombinant vectors were used for transformation of Escherichia coli BL21(DE3). Transformed cells were cultured for 8 -10 h in the presence of 0.5 mM isopropyl-␤-D-thiogalactopyranoside at 30°C. E. coli BL21(DE3) transformants were harvested by centrifugation; washed with phosphate-buffered saline containing a mixture of reversible and irreversible inhibitors (1 tablet in 25 ml) of pancreas extract, Pronase, thermolysin, chemotrypsin, trypsin, and papain (Complete TM ; Roche Diagnostics); suspended in the same buffer; and disrupted by sonication. After removal of debris by centrifugation, the recombinant proteins in the supernatant were purified by affinity chromatography on reduced glutathione-Sepharose 4B resin (New England Biolabs) according to the manufacturer's protocol. Wild-type and mutant TbPGFS fusion proteins bound to the resin via their glutathione S-transferase were cleaved with thrombin, and then TbPGFS was eluted with phosphate-buffered saline. The resulting recombinant proteins were dialyzed against 20 mM Tris/Cl (pH 8.0) buffer and applied to a DEAE anion-exchange column that had been equilibrated with the same buffer. The proteins were eluted with an increase in linear gradient of 0 -500 mM NaCl in the same buffer.
Protein concentrations were determined by use of bicinchoninic acid reagent (Pierce) with bovine serum albumin as a standard according to the manufacturer's protocol. The purity of the proteins was assessed by SDS-PAGE on 14% (w/v) gels, and the gels were stained with Sypro Orange (Bio-Rad) or Coomassie Brilliant Blue (Daiichi Pure Chemicals).
Enzyme Assay and Kinetic Measurements-PGH 2 reductase activity was measured by incubating at 37°C for 2 min a reaction mixture containing an appropriate enzyme aliquot, 5 M [1-14 C]PGH 2 (final concentration), 100 l of 100 mM sodium phosphate (pH 7.0), and 500 M NADPH (31). After the reaction had been terminated by the addition of 250 l of stop solution (30:4:1 (v/v/v) diethyl ether/methanol/2 M citric acid), aliquots were analyzed by thin-layer chromatography as described previously (31). For the reduction of 9,10-PQ, the reaction mixture consisted of 100 mM sodium phosphate (pH 7.0), the enzyme, 100 M NADPH, and 40 M 9,10-PQ, in a total volume of 500 l. The reaction was initiated by the addition of the substrate, and then the decrease in absorbance at 340 nm (⑀ ϭ 6270 M Ϫ1 cm Ϫ1 ) was monitored at 37°C. Blanks without enzyme or without substrate were also included. k cat and K m values were calculated from initial velocity studies over a wide range of pH values using the triple buffer system containing 50 mM sodium phosphate, 50 mM sodium pyrophosphate, and 50 mM 3-[(1,1-dimethyl-2-hydroxyethyl)amino]-2-hydroxypropanesulfonic acid. The reaction was initiated by the addition of a small aliquot of enzyme in phosphate-buffered saline, and no significant change in the pH of the reaction mixture was observed before or after the addition of enzyme.
UV and CD Spectroscopy-CD spectra of the wild-type and mutant enzymes were recorded in Tris/Cl buffer (pH 8.0) on a J-600 spectropolarimeter (Jasco International Co.) equipped with a 1-mm path length cuvette. The temperature in the cell holder was maintained at 10°C by a circulating water thermostat. The reported spectra were the mean of 10 scans at 50 nm⅐min Ϫ1 .
Fluorescence Quenching Assay-To determine the binding of NADPH to wild-type and mutant TbPGFS, we mixed various concentrations of NADPH with TbPGFS in a 200 l reaction mixture that contained 50 mM sodium phosphate (pH 7.0). After incubation at 37°C for 2 min, the intrinsic tryptophan fluorescence was measured by an FP-750 spectrofluorometer (Jasco International Co.) at an excitation wavelength of 282 nm and an emission wavelength of 338 nm (37). The quenching of tryptophan fluorescence caused by nonspecific interactions with each ligand was corrected with 1.5 M N-acetyl-L-tryptophanamide. The K d value for NADPH binding to TbPGFS was calculated as described by Levine (38).

RESULTS AND DISCUSSION
We determined the structure of the TbPGFS⅐NADP ϩ ⅐citrate ternary complex by molecular replacement at 2.1 Å resolution with R cryst and R free values of 23.1 and 26.8%, respectively. Other crystallographic properties and refinement statistics of the structure are shown in Table I.
TbPGFS is composed of a cylindrical core of eight parallel ␤-strands surrounded by 8 ␣-helices running antiparallel to the ␤-sheets (Fig. 1, A and B). This (␣/␤) 8 structure is interrupted by two extra ␣-helices (H1 and H2) situated toward the COOHterminal side of the molecule. The helix H1 lies in between ␤7 and ␣7, whereas H2 follows ␣8. The bottom of the barrel is sealed by a ␤ hairpin-like structure made up of two additional anti-parallel ␤-strands, B1 and B2, corresponding to residues 7-9 and 14 -16, joined by a tight turn. These extra structures, i.e. H1, H2, B1, and B2, have also been described in other members of the AKR superfamily (2,10,39). The COOH-terminal 19 amino acid residues sketch out the upper part of the substrate-binding pocket by constructing a lip-like structure around the edge of the COOH-terminal face of the central ␤-barrel. The overall structure of TbPGFS is homologous to that of Corynebacterium 2,5-diketo-D-gluconic acid reductase A (2,5DKG) (8,16), both of which belong to the fifth subfamily of the AKR superfamily (Fig. 1B). It is also essentially similar to the structures of members of the first subfamily of AKRs, including human aldose reductase (ADR) holoenzyme (2), porcine ADR (10), and rat liver 3␣-hydroxysteroid/dihydrodiol dehydrogenase (3␣-HSD) (4). However, one of the major structural differences between TbPGFS and members of the first subfamily of AKRs is the presence of shorter loops in the structure of TbPGFS. A superimposition, with a root mean square deviation of 1.5 Å from the corresponding C ␣ atoms, of the TbPGFS model on the rat 3␣-HSD (Fig. 1C) showed deletions in loops ␤1-␣1, ␤4-␣4, and ␤7-H1 in TbPGFS in contrast to the presence of significantly longer loops in enzymes of the first subfamily of AKRs. As a result, TbPGFS possesses shallow cofactor-and substrate-binding sites, whereas rat 3␣-HSD shows a deep corresponding site created by the longer loops. The presence of short loops in TbPGFS and Corynebacterium 2,5DKG may thus be a common characteristic of the members of the fifth subfamily of AKRs.
TbPGFS does not possess the Rossmann fold for binding NADPH. The NADP ϩ pyridine nucleotide is bound in an extended conformation within a crevice that extends from the outer edge of the barrel to the inner core. The nicotinamide ring is positioned deep in the central region of the ␤-barrel. The NADP ϩ adenine and pyrophosphate groups are wedged on the surface of the depression between the ␣7 and ␣8 helices (Fig.  1B). The electron density map of the NADP ϩ molecule bound to TbPGFS ( Fig. 2A) showed that the enzyme⅐NADP ϩ complex was stabilized by 16 hydrogen and 3 ionic bonds and 1 aromatic stacking interaction (Fig. 2B). The number of interactions is lower than that for AKRs of the first subfamily, such as human ADR holoenzyme and rat 3␣-HSD, both of which have 19 hydrogen bonds, 3 salt bridges, and 1 aromatic stacking (2,4). Fluorescence quenching assay (data not shown) revealed that the K d value of TbPGFS for NADPH (5.3 M) was 1 order of magnitude higher than that of rat 3␣-HSD, i.e. 143 nM (40), indicating a relatively loose binding of NADPH in TbPGFS as compared with the tight binding in AKRs of the first subfamily. The weaker affinity for NADPH may be responsible, in part, for the 3-fold higher NADPH K m value of 5.7 M (31) for TbPGFS compared with 2 M for human ADR holoenzyme (10). In Tb-PGFS, the pyrophosphate group of NADP ϩ is bound in a crevice on the surface of the barrel. On one side, the crevice lines Ser 188 , Leu 190 , and Gln 192 , all provided by loop ␤7-H1. On the other side, it has only Lys 230 , held by loop ␤8-␣8. The pyrophosphate group is held in place by hydrogen bonding interactions with Ser 188 , Leu 190 , Gln 192 , and one molecule of water in one side and Lys 230 in the other side (Fig. 2, A and B). In contrast, in the AKRs of the first subfamily, the extended polypeptide region of loop ␤7-H1 folds over the pyrophosphate group of NADPH and builds a short tunnel, known as the "safety belt," through which the pyrophosphate moiety passes. In human ADR holoenzyme, salt bridges between Asp 216 , Lys 21 , and Lys 262 stabilize this safety belt (2). However, the safety belt is absent in TbPGFS, and Asp 216 has been deleted from loop ␤7-H1. Lys 24 (similar to Lys 21 in human ADR holoenzyme) is located too far, i.e. ϳ6.3 Å, from the pyrophosphate group. Although the salt linkage between the pyrophosphate group and Lys 230 has been conserved in the structure of TbPGFS, the lack of the safety belt may account for the high K d value of NADPH observed for TbPGFS. In addition, AKR members have conserved most of the residues involved in NADP ϩ pyridine nucleotide binding. However, the involvement of Met 22 , Trp 187 , Gln 192 , and Gly 232 in NADP ϩ binding appears to be of the data, which were used during the course of the refinement. c R free ϭ ⌺ʈF o ͉ Ϫ ͉F c ʈ/⌺ ͉F o ͉, calculated from 5% of the data, which were obtained during the course of the refinement.
FIG . 1. A, structural alignment of TbPGFS (AKR5A2) with several members of the AKR superfamily. Alignments were calculated with the COMPARER program (51) using Protein Data Bank entries 1A80, 2ACQ, 1AH3, and 1AFS for Corynebacterium 2,5DKG (AKR5C), human ADR holoenzyme (AKR1B1), porcine ADR (AKR1B6), and rat 3␣-HSD (AKR1C9), respectively. The secondary structure of TbPGFS is indicated above the unique for TbPGFS. We also found that in TbPGFS structure, the amido atom of 3-carboxyamide side chain of the NADP ϩ ring hydrogen bonds Ser 139 and Gln 161 , both of which are conserved among the AKR members (Fig. 1A). Trp 187 indole ring is involved in ainteraction with the NADP ϩ ring. In contrast, in human ADR holoenzyme, porcine ADR, and rat 3␣-HSD, Tyr that has substituted Trp 187 maintains theinteraction with the NADP ϩ ring. On the other hand, phosphate group of 2Ј-AMP is hydrogen-bonded to Lys 230 and Arg 236 of TbPGFS, two residues that are conserved among AKRs (with the exception of rat 3␣-HSD, in which Lys 230 has been substituted by Arg 230 ). Moreover, with regard to the 2Ј-AMP phosphate group binding, a unique feature of TbPGFS is that the amido nitrogen atom of Gly 232 (a residue that is not conserved among AKRs) also hydrogen bonds the phosphate group. However, because the amido nitrogen atom of the main chain forms this interaction, the phosphate group bonds to either Val 232 in Corynebacterium 2,5DKG, human ADR holoenzyme, and porcine ADR or Phe 232 in rat 3␣-HSD.
The active site of TbPGFS is located at the COOH-terminal end of the ␤-barrel (Fig. 3, aϪe), as in the case of all other ␣/␤-barrel enzymes (39,41). Within the active site pocket, the nicotinamide ring is oriented so that the 4-pro-R hydrogen is directed up toward the pocket opening. The TbPGFS active site pocket has the dimensions of ϳ3 ϫ 5 ϫ 7 Å (length ϫ width ϫ depth) from the nicotinamide ring C 4 position to the upper surface of the TbPGFS molecule (Fig. 3a). These dimensions of the TbPGFS pocket are significantly smaller when compared with 5 ϫ 6 ϫ 9 Å (Fig. 3b) for Corynebacterium 2,5DKG (8) or 7 ϫ 13 ϫ 10 Å (Fig. 3c)  In addition, we used citrate (an inhibitor of the reduction of PGH 2 by TbPGFS) in co-crystallization experiments in order to get insight into its binding to the enzyme. Fig. 4A shows a TbPGFS⅐NADP ϩ ⅐citrate ternary structure with the electron density map of the bound citrate molecule. In this ternary complex structure, citrate was located above the bound NADP ϩ with its carboxyl O 1 penetrating deep into the substrate-binding site. This carboxyl O 1 is hydrogen-bonded to the O ␥ atom of Tyr 52 and the N ⑀ atom of His 110 at a distance of 2.76 and 2.54 Å, respectively (Fig. 4B), whereas Asp 47 and Lys 77 are located far from the citrate, at a distance of 6.61 and 4.98 Å, respectively. Hydrogen bondings similar to those between Tyr 52 -substrate-His 110 (in TbPGFS) were also identified in the recently determined human PGFS (3␣-HSD)⅐PGD 2 ⅐NADP ϩ ternary complex structure (42), in which Tyr 55 and His 117 correspond to Tyr 52 and His 110 of TbPGFS. Furthermore, the carboxyl O 1 of citrate is located at a distance of 3.30 Å to C 4 of NADP ϩ .
A number of highly conserved amino acid residues have been identified in the close vicinity of the NADPH nicotinamide ring in the active site pocket of AKRs. A comparison of the structure of TbPGFS with the structures of Corynebacterium 2,5DKG, human ADR holoenzyme, porcine ADR, and rat 3␣-HSD identified Asp 47 , Ala 49 , Tyr 52 , Lys 77 , Trp 79 , and His 110 as being those highly conserved amino acid residues in TbPGFS, whereas the amino acids at position 51, which is known to determine the specificity of substrates in AKRs, are highly divergent. The crystal structure of Corynebacterium 2,5DKG contains two well-ordered water molecules (Fig. 3b) that play an important role in the orientation of the carbonyl and hydroxyl groups of the substrate within the catalytic pocket (43). However, instead of water, TbPGFS crystal structure contained one citrate molecule (Fig. 3a). A superimposition of NADP ϩ molecules bound to TbPGFS with those bound to human ADR holoenzyme (2), porcine ADR (44), and rat 3␣-HSD (4) with a root mean square deviation of 0.17, 0.20, and 0.24 Å, respectively, revealed that in these enzymes the substratebinding sites corresponded to the site of citrate binding, thus identifying this site as the putative substrate-binding site.
Most AKRs catalyze the oxidation/reduction of both physiological and synthetic (9,10-PQ) substrates by a mechanism thought to be common for all members of the superfamily. To elucidate the mode of action of TbPGFS on PGH 2 and 9,10-PQ reduction, we generated a number of single, double, and triple mutants in the residues of the catalytic tetrad and assessed the effect of each mutation on the reduction of both substrates. Mutants K77L and H110F completely abolished their PGH 2 reductase activity (Fig. 5A, left panel and inset), whereas mutants Y52F, K77L, H110A, and H110F were inactive on 9,10-PQ (Fig. 5B, left panel). Replacement of Leu 77 by a positively charged residue in L77R mutant restored 60% and 52% of the reductase activity toward PGH 2 and 9,10-PQ, respectively. In contrast, H110A mutation produced a mutant that exhibited 2.9% of PGH 2 reductase activity as compared with the wild-type enzyme. However, this same mutation resulted in inactive mutant on 9,10-PQ (Fig. 5, A and B, left panels). Unexpectedly, none of the other single mutations altered the PGH 2 reductase activity, indicating that in TbPGFS, a diad of Lys 77 and His 110 , but not a tetrad of Asp 47 , Tyr 52 , Lys 77 , and His 110 , was essential for the PGH 2 reductase activity. Mutants Y52F/H110A and D47N/H110A retained some of their reductase activity toward PGH 2 , whereas Y52F/K77L, K77L/H110A, and Y52F/K77L/H110A completely lost this activity (data not shown). None of these double and triple mutants exhibited 9,10-PQ reductase activity. These results suggest that TbPGFS uses two independent mechanisms to catalyze the reduction of both PGH 2 and 9,10-PQ. The reduction of PGH 2 to PGF 2␣ appears to be achieved by a catalytically essential diad of Lys 77 and His 110 , but not Tyr 52 and/or Asp 47 , whereas a catalytic triad of Tyr 52 , Lys 77 , and His 110 is needed for the reduction of the non-physiological substrate 9,10-PQ. Previous mutagenesis studies have shown that the Y55F mutant of rat liver 3␣-HSD (an AKRC1C9) abolished the reductase activity toward steroid substrates (45) but retained the reduction of 9,10-PQ (46). In the present study, we show that Y52F mutant of TbPGFS (an AKR5A2) retained the reductase activity toward the physiological substrate (PGH 2 ), while abolishing the reduction of the non-physiological substrate 9,10-PQ. Instead, TbPGFS used a distinct residue (His 110 ) to achieve the reduction of PGH 2 . Although the presence of two different catalytic mechanisms for the reduction of two different substrates within the same alignment. The corresponding amino acid residues are highlighted in yellow for ␤-sheets, blue for ␣-helices, and red for loops. ␤-Sheets and ␣-helices that are not part of the (␣/␤) 8 -barrel structure core are indicated by B1, B2, H1, and H2. Green asterisks denote residues that are positionally conserved among all enzymes investigated, and black boxes indicate the residues that are conserved as catalytic tetrad in AKRs. B, stereoview of the overall structure of TbPGFS with bound NADP ϩ , showing the overall structure viewed down the COOH-terminal end of the central ␤-barrel with the NADP ϩ molecule drawn in ball-and-stick model. The ␣-helices (in blue) and ␤-sheets (in yellow) are separated among them by loops (in orange). One citrate molecule is also depicted as a ball-and-stick model within the putative catalytic cleft. The overall fold of TbPGFS is similar to that of other AKR enzymes. The figures were drawn with MOLSCRIPT (52) and RASTER3D (53). C, side view of the superimposed overall structure of TbPGFS⅐NADP ϩ ⅐citrate ternary complex (in light blue) with rat 3␣-HSD⅐NADPH⅐testosterone ternary complex showing the loops (␤1-␣1, ␤4-␣4, and ␤7-H1, in pink) that are present in 3␣-HSD but absent in TbPGFS and how loop ␤7-H1, a structural feature absent in TbPGFS, impacts NADPH binding in 3␣-HSD. Structural differences of the three loops were drawn using MOLSCRIPT and RASTER3D. active site of AKRs has been reported (45,46), our findings show that at least TbPGFS catalyzes the reduction of PGH 2 through a different mechanism and that the non-physiological substrate 9,10-PQ appears to be reduced by a mechanism similar to that of steroid reduction by rat 3␣-HSD, in which the Tyr residue is essential.
Kinetic analysis of the mutants (Table II) revealed that the catalytic efficiency, k cat /K m (PGH 2 ), of the invariable catalytic tyrosine mutant (Y52F) was similar to that of the wild-type enzyme, proving that neither k cat nor K m (PGH 2 ) was significantly affected by this mutation. H110A mutant showed a 35-fold decrease in k cat and a 5-fold increase in K m (PGH 2 ) that resulted in a 183-fold decrease in the k cat /K m (PGH 2 ) value. On the other hand, D47N and K77R mutants exhibited an increased k cat (almost 1.6-fold) and K m for PGH 2 (10-and 20-fold, respectively), leading to a significant decrease in their k cat /K m values. In contrast, these same mutants (D47N and K77R) showed a decreased k cat (almost 1.4-and 1.9-fold, respectively) FIG. 2. A, NADP ϩ and citrate binding models with 2F o Ϫ F c omit maps colored in blue and orange, respectively. Oxygen, nitrogen, sulfur, and phosphorus atoms are shown in red, blue, yellow, and magenta, respectively. Labeled residues indicate those involved in NADP ϩ and citrate binding. The figures were drawn using the programs MOLSCRIPT and RASTER3D. B, schematic representation of the interactions between NADP ϩ and TbPGFS. In addition to hydrogen bonds and electrostatic interactions, the side chain of Trp 187 is involved in the aromatic stacking interaction with nicotinamide ring of the cofactor. and an increased K m for 9,10-PQ (47-and 16-fold, respectively), which resulted in a very low catalytic efficiency. Because structural analysis revealed hydrogen bonds between Asp 47 -NADP ϩ (Fig. 2, A and B) and Asp 47 -Lys 77 (Fig. 6B), it is most likely that D47N mutation produced a mutant with a weak hydrogen bond between Asn 47 -Lys 77 and low affinity between Asn 47 -NADPH.
This change could well affect substrate binding and the catalytic efficiency for the reduction of both substrates by this mutant. On the other hand, the increment of K m for PGH 2 and 9,10-PQ in K77R may be due to a change in substrate binding properties as a result of Arg 77 rigidity.
The loss of activity observed throughout this study could FIG. 3. The ligand binding pocket of AKRs. a, TbPGFS; b, Corynebacterium 2,5DKG; c, human ADR holoenzyme; d, porcine ADR; e, rat 3␣-HSD. TbPGFS and Corynebacterium 2,5DKG pockets display a citrate anion and two well-ordered water molecules, respectively. Ligand glucose-6-phosphate (G6P), tolrestat (a well-known inhibitor for AKR), and testosterone are shown in red, green, and purple, respectively. They are bound to human ADR holoenzyme, porcine ADR, and rat 3␣-HSD, respectively. NADP(H) molecule is shown as the ball-and-stick model. The figures were drawn using the programs GRASP (54) come from a conformational change rather than from any specific effect of the particular mutation on catalysis. CD spectra of the wild-type and mutant enzymes did not suggest any significant alteration of the ternary structure in the mutant proteins (data not shown), suggesting that these mutations did not result in any significant change in the overall three-dimensional structures. In addition, wild-type TbPGFS and its inactive mutants (Y52F, K77L, H110A, and H110F) did bind NADPH (data not shown) with a binding constant (K d ) that was almost similar to that of the wild-type enzyme (5.3 M), confirming that mutations did not affect the overall threedimensional structures of the resulting mutants. We tested also the effect of PGH 2 analogs (U-44069 and U-46619) on the reduction of PGH 2 and 9,10-PQ by TbPGFS. Chemically, these analogs have structures that are similar to PGH 2 , except that the oxygen atom at C 11 in U-44069 or at C 9 in U-46619 is FIG . 5. A, left panel, specific activities of wild-type and mutant enzymes. Pure recombinant wild-type (Wt) or mutants of TbPGFS (1-2 g) were incubated with 5 M [1-14 C]PGH 2 in the presence of NADPH as described under "Materials and Methods." PGH 2 conversion to PGF 2␣ was qualitatively (inset) and quantitatively (the bar graph) analyzed by thin-layer chromatography and autoradiography as shown in the inset, and the corresponding values of the spot densities were plotted. Right panel, pH dependence of PGH 2 reduction by wild-type and mutant TbPGFS. B, left panel, NADPH-dependent reduction of 9,10-PQ by wild-type and mutant TbPGFS. Experimental details are described under "Materials and Methods." Right panel, pH dependence of 9,10-PQ reduction by wild-type and mutant TbPGFS. C, schematic drawing of PGH 2 and PGH 2 homologs U-44069 and U-46619. D, effect of U-46619 and U-44069 on the reduction of PGH 2 by wild-type TbPGFS. Various concentrations of homologs were pre-incubated with TbPGFS (1-2 g) at 37°C for 2 min, followed by incubation of the enzyme with 5 M [1-14 C]PGH 2 in the presence of NADPH. The residual enzymatic activity was plotted against PGH 2 homolog concentrations. substituted with a CH 2 group (Fig. 5C). Pre-incubation of wildtype TbPGFS with various concentrations of U-44069 had no effect on the reduction of both PGH 2 (Fig. 5D) and 9,10-PQ (data not shown). In contrast, treatment with U-46619 partially inhibited the reduction of PGH 2 by TbPGFS in a dosedependent manner (Fig. 5D), but it had no effect on the reduction of 9,10-PQ (data not shown). These findings provide evidence of the need for free oxygen at the C 11 position of PGH 2 for the substrate to enter the catalytic pocket and bind to TbPGFS.
We constructed a PGH 2 binding model (Fig. 6) based on the electrostatic potential on the surface of TbPGFS, the details of citrate binding to TbPGFS, and the kinetics of PGH 2 analogs binding to the wild-type enzyme. The energy minimization of the PGH 2 binding model was achieved by using the CNS program. In this model, PGH 2 binds above the nicotinamide ring of the cofactor in an extended conformation across the shallow elliptical crevice (Fig. 6A). PGH 2 chains ␣ and extend in opposite directions and straddle the lip of the crevice. This orientation is justified by the analysis of the electrostatic potential surface model of PGH 2 , which revealed that ␣-chain of the substrate would interact with the only basic residue, Lys 24 , and that -chain may interact with the hydrophobic residue Trp 23 . The hydroxyl group in the -chain of PGH 2 is located at a distance of 2.72 Å from the N ⑀ atom of Trp 23 , suggesting that Trp 23 may well recognize the -chain as evidenced in the TbPGFS⅐NADP ϩ ⅐citrate ternary structure, where the carboxyl O 5 of citrate was hydrogen-bonded to the N ⑀ atom of Trp 23 at a distance of 3.44 Å (Fig. 4B). No basic residues were found in the putative binding site around the -chain of PGH 2 . The cyclopentane ring of PGH 2 , with its 9,11-endoperoxide bond, is oriented deep in the cavity in the vicinity of the nicotinamide ring of NADP ϩ , positioning the C 9 oxygen atom of PGH 2 at 3.3 Å from the C 4 of the nicotinamide ring (Fig. 6B). On the other hand, Lys 77 and His 110 are located at 7.69 and 2.62 Å, respectively, from the C 11 oxygen of PGH 2 . These geometrical arrangements indicate that His 110 , but not Lys 77 , is in close vicinity to C 11 oxygen of PGH 2 and therefore may easily be involved in hydrogen bonding of the substrate and proton transfer during the reduction of PGH 2 . More recently, a ternary structure of human PGFS⅐NADP ϩ ⅐PGD 2 was reported (42). The PGH 2 binding model we propose here is in agreement with PGD 2 binding to the human PGFS (3␣-HSD, an enzyme that catalyzes the reduction of PGH 2 to PGF 2␣ , the reduction of PGD 2 to 9␣,11␤-PGF 2 , and the oxidation of 9␣,11␤-PGF 2 to PGD 2 ) because the ternary structure of human 3␣-HSD⅐NADP ϩ ⅐PGD 2 revealed that the cyclopentane ring of PGD 2 (a prostanoid that possesses a cyclopentane ring as PGH 2 ) was oriented deep within the active site cavity, whereas ␣and -chains of PGD 2 remained far from the active site cavity and showed little interaction with the enzyme.
Mutagenesis and kinetic studies of AKRs revealed that bac-terial and mammalian enzymes catalyze the oxidation/reduction reactions through a mechanism that involves a catalytic tetrad of tyrosine, histidine, lysine, and aspartate (1-4, 8, 44, 45, 47, 48). Tyrosine acts as the general acid/base catalyst. Histidine facilitates proton donation during reduction (46) or orientates the substrate carbonyl in the active site (9,43), whereas lysine helps proton removal by tyrosine during oxidation by making a hydrogen bond that lowers the pK a value of the tyrosine. Aspartate forms a salt bridge to stabilize lysine (3,4,40). In order to catalyze the oxidation/reduction reaction according to this mechanism, tyrosine OH has been involved in hydrogen bond networks with histidine and lysine. In the present study, we found that among the residues of the catalytic tetrad of TbPGFS, Lys 77 and His 110 are catalytically important for the reduction of PGH 2 , whereas Tyr 52 , Lys 77 , and His 110 are essential in the reduction of 9,10-PQ. Observations that mutation of Tyr 52 , the putative proton donor in most of AKRs, led to a retention of PGH 2 reductase activity but a loss of 9,10-PQ reductase activity prompted a thorough investigation of the effect of pH on catalysis in an effort to assign the titratable groups involved in the reduction of these two substrates. The pK a and pH-independent values of k cat for the wild-type and mutant enzymes are summarized in Table III. The log k cat value for the reduction of PGH 2 by wild-type TbPGFS (Fig. 5A, right panel) increased with the increase of pH up to pH 6.5 and then decreased with the increase of pH, showing an ionizable group with a pK a value of 7.4 Ϯ 0.08 (Table III) that should be protonated for maximal activity. A comparison of the log k cat versus pH profiles for the tetrad mutants revealed the presence in D47N, Y52F, and K77R mutants of a titratable group with a pK a value of 7.80, 7.30, and 7.40, respectively, that must be protonated for maximal PGH 2 reduction. However, this titratable group is not observed in mutants K77L and H110F. In addition, the log k cat versus pH profiles also show a titratable group with a lower pK a value of 5.8 Ϯ 0.1 for H110A mutant (that exhibited trace of PGH 2 reduction) that must be protonated for maximal PGH 2 reduction. Although Lys 77 and His 110 can each act as a proton donor, structural data revealed that Lys 77 is located too far (6.35 Å) from the C 4 of the nicotinamide ring (Fig. 6B) and thus would probably not act as a proton donor for PGH 2 . Taken together, these data suggest that His 110 (pK a ϭ 5-8 in proteins) but not Lys 77 (pK a ϭ ϳ10) plays a dominant role in catalysis and thus may be the titratable group responsible for PGH 2 reduction. In contrast, the k cat versus pH profiles for the reduction of 9,10-PQ by wild-type TbPGFS (Fig.  5B, right panel) showed a k cat value that increases with increasing pH up to pH 8.0 and then falls with increasing pH, revealing the presence of a titratable group with a pK a value of 8.5 Ϯ 0.05 that must be protonated for maximal activity. The 1.1-pH unit difference of the pK a value between the reduction of PGH 2 and 9,10-PQ suggests that two different ionizable groups might be involved in the reduction of the two different substrates. This finding is consistent with our site-directed mutagenesis data and explains our previous results showing that the reduction of PGH 2 by TbPGFS was not inhibited by 9,10-PQ when both substrates were incubated simultaneously in excess (31). A profile of the k cat versus pH for the tetrad mutants revealed that mutants Y52F, K77L, H110A, and H110F eliminated the titratable group for 9,10-PQ reduction. Residues Tyr 52 , Lys 77 , and His 110 can each act as a proton donor. However, Lys 77 will not be a proton donor because it is located too far away. Because the pH dependence of k cat (for wild-type) identified an ionizable group with a pK a value of 8.5 Ϯ 0.05 that must be protonated for maximal activity, the ionizable residue responsible for 9,10-PQ reduction appears to be Tyr 52 . Tyrosine usual range of pK a in proteins is 9 -12. In TbPGFS, the pK a value of 8.5 Ϯ 0.05 appears to be lowered by hydrogen bond with Lys 77 . Our mutagenesis and kinetic studies also show that the H110A mutant exhibited 2.9% of the PGH 2 reductase activity relative to wild-type enzyme, whereas it was inactive on 9,10-PQ. Why H110A mutant shows some trace activity on PGH 2 is not clear at present. However, the pH profile of log k cat for the wild-type and H110A mutant shows a decrease in pK a (Table III) for H110A, in addition to the fact that the H110A mutation was the only one to demonstrate a 105-fold reduction in the pH-independent value of k cat . The ⌬pK a of 1.6 pH unit between the wild-type and H110A mutant likely reflects the involvement of a different titratable group in this mutant. In addition, the fact that the k cat versus pH plot is shallow with a slope Ͻ 1.0 may suggest that that the titratable group is not a single ionizable amino acid but rather a group of residues, i.e. the Lys 77 and His 110 dyad. Furthermore, TbPGFS appears to be a rare example of an AKR that uses histidine to transfer proton to a physiological substrate (PGH 2 ) and loses its quinone reductase activity upon mutation of the catalytic tyrosine. Most members of AKR described so far are known to lose their reductase activity toward AKR physiological substrates while they retain quinone reductase activity in the absence of the catalytic tyrosine (45,46,49,50). TbPGFS is also different from the human PGFS because this latter enzyme completely eliminates PGD 2 , PGH 2 , and 9,10-PQ reductase activities in the absence of the catalytic tyrosine (18).
In the crystal structure of TbPGFS⅐NADP ϩ ⅐citrate, we also identified a well-defined water molecule from the water channel between the N ␦1 atom of the imidazole ring of His 110 and the carboxyl group of Lys 77 (Fig. 6B). The presence of this water molecule shows that His 110 is exposed to solvent, and therefore protonation/deprotonation of its imidazole ring will be influenced by the pH of the environment. Under our experimental conditions (aerobic and pH 7.0), His 110 (pK a ϭ 7.4 Ϯ 0.08) should be protonated for maximal activity, whereas the reduction of PGH 2 catalyzed by wild-type TbPGFS would require an obligatory two-electron transfer event from the cofactor (NADPH) to the C 9 carbonyl of PGH 2 . It appears, therefore, that TbPGFS catalyzes hydride transfer to the C 9 carbonyl of PGH 2 by a mechanism that involves His 110 (Fig. 6C). In this mechanism, concomitantly, C 9 carbonyl of PGH 2 (positioned at 2.63 Å from the C 4 of the nicotinamide ring) would accept a hydride ion (H Ϫ , magenta) and then will be protonated, whereas N ⑀2 of the imidazole ring of His 110 (located at 2.62 Å from the C 11 carbonyl) would polarize C 11 carbonyl of the substrate. Thus, His 110 will act as a general acid by donating its proton (H ϩ , red) to C 11 carbonyl (␦ Ϫ ) of the endoperoxide. However, we also found that His 110 can act as a general acid catalyst only when TbPGFS has a positively charged residue at position 77. This positively charged residue seems to create an electrostatic repulsion against His 110 (also a positively charged residue) that maintains the spatial arrangement and allows the formation of a hydrogen bond between His 110 and C 11 carbonyl of the endoperoxide. In addition, through a water molecule bound between its carboxyl group and the N ␦1 atom of the imidazole ring of His 110 , this positively charged residue would facilitate His 110 protonation. This assumption is supported by the fact that a mutation of the positively charged residue Lys 77 to a hydrophobic amino acid, leucine, led to a complete loss of activity by the mutant K77L and that this lost activity was restored by replacement of Leu 77 by arginine (another positively charged residue) in L77R mutant. Our results therefore show the importance of Lys 77 in facilitating the spatial arrangement via an electrostatic repulsion against His 110 and protonation of His 110 through a water molecule, in addition to forming a salt bridge to stabilize Asp 47 .
We have successfully solved the three-dimensional structure of TbPGFS that was co-crystallized in the presence of NADP ϩ and citrate. Our x-ray data provided structural details of NADP ϩ and citrate binding that were used for the modeling of PGH 2 binding to the enzyme. The presence of only two and three catalytic residues for the reduction of PGH 2 and 9,10-PQ instead of four was confirmed by mutagenesis and kinetic studies. Our results provide the first insights into the mechanism of the biological reduction of PGH 2 by a prostaglandin F 2␣ synthase and the critical role of histidine, but not tyrosine, in catalysis. We therefore propose a novel catalytic mechanism for the reduction of PGH 2 to PGF 2␣ that involves a dyad of Lys 77 and His 110 in which His 110 acts as a general acid catalyst.
the PGH 2 binding model is described under "Results and Discussion." The PGH 2 binding model was made with the programs GRASP, MOLSCRIPT, and RASTER3D; colored according to the calculated electrostatic potentials; and contoured from Ϫ15 (intense red) to ϩ15 (intense blue) kT/e. B, stereoview of the PGH 2 binding model showing the putative bonds between PGH 2 and TbPGFS. The catalytic site also shows the highly conserved catalytic tetrad: Asp 47 , Tyr 52 , Lys 77 , and His 110 . The model shows Lys 77 attached to Asp 47 and Thr 48 , whereas His 110 hydrogen bonds with a water molecule (Wat 1). The model is viewed as indicated by the arrow in A (right panel). These figures were drawn using MOLSCRIPT and RASTER3D. C, proposed catalytic mechanism of the enzymatic reduction of PGH 2 . When a substrate binds to the active site of TbPGFS, a NADPH hydride ion attacks the C 9 carbonyl of PGH 2 , whereas a proton from His 110 will be transferred to the negatively charged C 11 carbonyl. Lys 77 appears to have a role of maintaining the position of His 110 and stabilizing Asp 47 . See details under Results and Discussion.