Catalytic mechanism of the primary human prostaglandin F synthase, aldo-keto reductase 1B1 – prostaglandin D2 synthase activity in the absence of NADP(H)

Authors

Errata

This article is corrected by:

  1. Errata: Corrigendum Volume 278, Issue 22, 4450–4451, Article first published online: 28 October 2011

Y. Urade, Department of Molecular Behavioral Biology, Osaka Bioscience Institute, 6-2-4 Furuedai, Suita, Osaka 565-0874, Japan
Fax: +81 6 6872 2841
Tel: +81 6 6872 4851
E-mail: uradey@obi.or.jp

Abstract

Aldo-keto reductase 1B1 and 1B3 (AKR1B1 and AKR1B3) are the primary human and mouse prostaglandin F (PGF) synthases, respectively, which catalyze the NADPH-dependent reduction of PGH2, a common intermediate of various prostanoids, to form PGF. In this study, we found that AKR1B1 and AKR1B3, but not AKR1B7 and AKR1C3, also catalyzed the isomerization of PGH2 to PGD2 in the absence of NADPH or NADP+. Both PGD2 and PGF synthase activities of AKR1B1 and AKR1B3 completely disappeared in the presence of NADP+ or after heat treatment of these enzymes at 100 °C for 5 min. The Km, Vmax, pK and optimum pH values of the PGD2 synthase activities of AKR1B1 and AKR1B3 were 23 and 18 μm, 151 and 57 nmol·min−1·(mg protein)−1, 7.9 and 7.6, and pH 8.5 for both AKRs, respectively, and those of PGF synthase activity were 29 and 33 μm, 169 and 240 nmol·min−1·(mg protein)−1, 6.2 and 5.4, and pH 5.5 and pH 5.0, respectively, in the presence of 0.5 mm NADPH. Site-directed mutagenesis of the catalytic tetrad of AKR1B1, composed of Tyr, Lys, His and Asp, revealed that the triad of Asp43, Lys77 and His110, but not Tyr48, acts as a proton donor in most AKR activities, and is crucial for PGD2 and PGF synthase activities. These results, together with molecular docking simulation of PGH2 to the crystallographic structure of AKR1B1, indicate that His110 acts as a base in concert with Asp43 and Lys77 and as an acid to generate PGD2 and PGF in the absence of NADPH or NADP+ and in the presence of NADPH, respectively.

Abbreviations
AKR

aldo-keto reductase

PG

prostaglandin

PGDS

prostaglandin D synthase

PGFS

prostaglandin F synthase

TLC

thin-layer chromatography

Introduction

Aldo-keto reductases (AKRs) are soluble monomeric proteins with molecular masses of 37 kDa with NADPH-dependent oxidoreductase activity. AKR proteins are widely distributed in prokaryotes and eukaryotes, fall into 15 families [1] and metabolize a number of substrates, including aldehydes, monosaccharides, steroids, polycyclic hydrocarbons, isoflavonoids and prostaglandins (PGs) in the presence of NADPH [2]. Aldose reductase (EC 1.1.1.21), named AKR1B1 in human and AKR1B3 in mouse, is considered to be the prototypical enzyme of the AKR superfamily. In addition to these conical aldose reductases, a second group, named aldose reductase-like proteins, has been characterized on the basis of sequence homology (at least 60–70% identity with aldose reductase). AKR1B7, initially characterized as a mouse vas deferens androgen-dependent protein, belongs to the aldose reductase-like proteins. X-Ray crystallographic structures of members of the AKR superfamily have shown these enzymes to have a common three-dimensional fold, known as the (α/β)8-barrel fold [3–6]. The nucleotide cofactor binds in an extended conformation at the top of the α/β-barrel, with the nicotinamide ring projecting down into the center of the barrel and pyrophosphate straddling the barrel lip [7]. Kubiseski et al. [8] have established that the enzyme follows a sequential ordered mechanism in which NADPH binds before the aldehyde substrate and NADP+ is released after the alcohol product is formed. When, in 1992, the first complete crystal structure of human AKR1B1 was solved, the conserved Tyr48 was shown to fulfill the role of a catalytic acid for NADPH-dependent reduction [5].

Recently, we have reported that human AKR1B1, mouse AKR1B3 and mouse AKR1B7 are associated with PGF synthase (PGFS; EC 1.1.1.188) activity, which catalyzes the reduction of PGH2, a common intermediate of various prostanoids of the two series, to PGF [9]. PGF plays a variety of physiological roles in the body, such as the contraction of uterus, bronchial, vascular and arterial smooth muscles [10], regulation of pressure in the eye [11], renal filtration [12], stimulation of hair growth [13] and regulation of the ovarian cycle through the induction of luteolysis [14]. More recently, human AKR1B1 and mouse AKR1B3 were identified to be the primary PGFS [15,16]. Three different pathways have been reported for PGF production [10], i.e. 9,11-endoperoxide reduction of PGH2, 9-ketoreduction of PGE2 and 11-ketoreduction of PGD2, although the latter results in the production of a PGF stereoisomer, 9α,11β-PGF2, not PGF [17]. PGFS was first isolated from mammals as an enzyme that catalyzes the reduction of PGH2 to PGF, and of PGD2 to 9α,11β-PGF2 [18]. The first identified mammalian PGFS belongs to the AKR1C family [19,20], and protozoan PGFS to the AKR5A subfamily [21,22]. PGF ethanolamide synthase, which belongs to the thioredoxin-like superfamily, has also recently been found to convert PGH2 to PGF [23].

In this study, we introduced site-directed mutagenesis into the catalytic tetrad of AKR1B1, and found that His110, not Tyr48, was crucial for PGFS activity in the presence of NADPH. Furthermore, we found that AKR1B1 and AKR1B3, but not AKR1B7 and AKR1C3, also catalyzed the isomerization of PGH2 to PGD2 in the absence of NADPH or NADP+. In combination with the mutagenesis analyses and pH titration studies, we found that His110 acted as a base to generate PGD2 in the absence of NADPH or NADP+ and as an acid to generate PGF in the presence of NADPH. Thus, this is the first report demonstrating the proton donor/acceptor function of His110 during the conversion of PGH2 catalyzed by AKR1B1.

Results

Formation of PGF and PGD2 from PGH2 by AKR1B1

Recombinant human AKR1B1, mouse AKR1B3, mouse AKR1B7 and human AKR1C3 were expressed in Escherichia coli and purified to be a single band as judged by SDS/PAGE. We incubated these purified AKR proteins with 5 μm [1-14C]PGH2 in the presence or absence of 0.5 mm NADPH or NADP+ and analyzed the reaction products by thin-layer chromatography (TLC). AKR1B1 catalyzed the reduction of the 9,11-endoperoxide group of PGH2 to produce PGF in the presence of NADPH, which was defined as the PGFS activity, and the isomerization of PGH2 to PGD2 in the absence of NADPH or NADP+, which was defined as the PGD2 synthase (PGDS) activity (Fig. 1A). Both PGDS and PGFS activities were not found in the presence of NADP+ at all and were completely inactivated by heat treatment of AKR1B1 at 100 °C for 5 min. The PGFS and PGDS activities catalyzed by AKR1B1 were calculated to be 2.4 and 3.7 nmol·min−1·(mg protein)−1, respectively (Fig. 1B). AKR1B3 with 85.8% identity of the amino acid sequence of AKR1B1 also catalyzed both PGFS activity [3.6 nmol·min−1·(mg protein)−1] in the presence of NADPH and PGDS activity [3.3 nmol·min−1·(mg protein)−1] in the absence of NADPH or NADP+. However, AKR1B7 (71.2% and 69.6% identity with AKR1B1 and AKR1B3, respectively) and AKR1C3 (47.4% and 47.1% identity with AKR1B1 and AKR1B3, respectively) did not catalyze PGDS activity, although these AKRs showed PGFS activity [3.9 and 0.9 nmol·min−1·(mg protein)−1, respectively] in the presence of NADPH. These results suggest that PGDS activity is selective to AKR1B1 and AKR1B3 among these mammalian AKR proteins.

Figure 1.

 PGDS and PGFS activities of recombinant AKR1B1 protein. (A) Autoradiogram of TLC after incubation of AKR1B1 (each 10 μg protein) with 5 μm 1-[14C]PGH2 in the presence or absence of cofactor at 37 °C for 2 min at pH 7.0 with or without heat treatment at 100 °C for 5 min. (B) Enzyme activities of wild-type AKR1B1, AKR1B3, AKR1B7 and AKR1C3 obtained from the respective PGDS and PGFS assays at pH 7.0. Data are presented as the mean ± SD from four to six independent experiments.

Kinetic analysis of the PGFS and PGDS activities of AKR1B1

Figure 2A shows the pH–rate profiles of AKR1B1 for PGFS and PGDS activities. The PGFS activity decreased with increasing pH, with an optimum of pH 5.5. The pKb value of PGFS activity was calculated to be 6.19 ± 0.05 by nonlinear fitting of Eqn (1) (see Materials and methods section) to the pH–rate profile data. However, the PGDS activity of AKR1B1 increased with increasing pH, with an optimum of pH 8.5. The pKa value of PGDS activity was calculated by Eqn (1) to be 7.94 ± 0.07. Nonenzymatic autodegradation of PGH2 was almost constant in a range from pH 4 to pH 9 and increased at alkaline pH values, especially at pH > 11 [24], suggesting that the pKa value of C11 may be higher than pH 9. As PGH2 does not ionize in the pH range examined, the pH profiles of the reaction velocity reflect the pH-dependent ionization of the catalytic residue for the PGFS and PGDS activities of AKR1B1. AKR1B3 also showed similar pH–rate profiles to AKR1B1 for PGFS and PGDS activities, although the PGDS activity of AKR1B3 was 24% of the PGFS activity (Fig. S1A). The optimum pH values were found to be pH 5.0 for PGFS activity and pH 8.5 for PGDS activity of AKR1B3. The pKb value of PGFS activity and the pKa value of PGDS activity were calculated by Eqn (1) to be 5.39 ± 0.09 and 7.57 ± 0.04, respectively, by nonlinear fitting of Eqn (1) to the pH–rate profile data.

Figure 2.

 Kinetic analysis of PGFS and PGDS activities of wild-type AKR1B1. (A) V value versus pH for PGH2 conversion catalyzed by wild-type AKR1B1. Wild-type AKR1B1 was incubated with 5 μm 1-[14C]PGH2 in the presence of 0.5 mm NADPH for PGFS activity (○) or in the absence of NADPH or NADP+ for PGDS activity (•) at 37 °C for 1 min. PGH2 conversion to PGD2 or PGF was analyzed by TLC and autoradiography. The corresponding values of the spot densities were plotted. (B) Michaelis–Menten plots of PGFS (○) and PGDS (•) activities for AKR1B1 on PGH2 at their optimum pH. Wild-type AKR1B1 protein was incubated with various concentrations of PGH2 in the presence of 0.5 mm NADPH for PGFS activity or in the absence of NADPH or NADP+ for PGDS activity. For PGDS activity, the enzymes were incubated at 37 °C for 1 min with 5 μm 1-[14C]PGH2 in the absence of NADPH or NADP+ at pH 8.5, and, for PGFS activity, in the presence of 0.5 mm NADPH at pH 5.5.

PGDS and PGFS activities of AKR1B1 were characterized at their optimum pH values (pH 5.5 for PGFS and pH 8.5 for PGDS) by Michaelis–Menten kinetics (Fig. 2B). AKR1B1 exhibited a Km value for PGH2 of 29 μm and a Vmax value of 169 nmol·min−1·(mg protein)−1 for PGFS activity in the presence of NADPH at pH 5.5, and values of 23 μm and 151 nmol·min−1·(mg protein)−1, respectively, for PGDS activity in the absence of NADPH or NADP+ at pH 8.5. However, AKR1B3 showed a Km value of 33 μm and Vmax value of 240 nmol·min−1·(mg protein)−1 for PGFS activity at pH 5.0, and 18 μm and 57 nmol·min−1·(mg protein)−1, respectively, for PGDS activity at pH 8.5 (Fig. S1B). Similar affinities for the substrate PGH2 and Vmax values of PGFS and PGDS activities of AKR1B1 and AKR1B3 suggest that the substrate is bound in a similar fashion.

Mutagenesis analyses of the effect of the AKR tetrad in AKR1B1 on PGDS and PGFS activities

X-Ray crystallographic and biochemical analyses of AKR1B1 revealed that this protein contains a catalytic tetrad composed of Asp43, Tyr48, Lys77 and His110, which is highly conserved among members of the AKR family and constructs the common active site with a hydrophobic core in this family [3–6]. To identify the catalytic residues involved in the PGDS and PGFS activity catalyzed by AKR1B1, we introduced site-directed mutagenesis into the tetrad, generating the D43N, D43E, Y48F, K77L, K77R, H110F and H110A mutants, and assessed their PGDS and PGFS activities with 5 μm [1-14C]PGH2 at the optimum pH 8.5 for PGDS activity and pH 5.5 for PGFS activity in the absence and presence of 0.5 mm NADPH, respectively. The typical autoradiograms of TLC used for PGDS and PGFS assays are shown in Fig. 3A,B, respectively. Under these conditions, the Y48F mutant changed slightly both PGDS and PGFS activities from wild-type AKR1B1 (138% and 69% of wild-type AKR1B1, respectively), although this mutant decreased the p-nitrobenzaldehyde reductase activity to 0.2% (Fig. 3C), indicating that the catalytic Y48 residue is essential for AKR activity but not necessary for either PGDS or PGFS activity. However, all other mutants of the tetrad, including H110, showed some trace activity on both PGDS and PGFS activities (Fig. 3C), suggesting that the triad of Asp43, Lys77 and His110 residues is essential for these activities.

Figure 3.

 Mutagenesis analyses of the catalytic tetrad of AKR1B1. Typical autoradiogram of TLC used for the PGDS assay at optimum pH 8.5 (A) and the PGFS assay at the optimum pH 5.5 (B) for AKR1B1. (C) Summarized enzyme activities obtained from the respective PGDS and PGFS assays at the optimum pH and the NADPH-dependent p-nitrobenzaldehyde reductase (PNBR) activity at pH 7.0 by the wild-type and mutants. (D) Typical fluorescence quenching curves of wild-type AKR1B1 and its mutants. The binding of NADP+ to these proteins was determined by measuring the decrease in fluorescence emission at 338 nm (excitation at 282 nm). Data are presented as the mean ± SD from three to six independent experiments.

When site-specific mutagenesis was introduced to Y48 and H110 of AKR1B3 (Fig. S2C), the Y48F mutant changed slightly both PGDS and PGFS activities (133% and 286% of wild-type AKR1B3, respectively), and the H110F mutant decreased the PGDS and PGFS activities to 37% and 1%, respectively. The double mutant Y48F/H110F completely lost PGDS activity and showed a weak PGFS activity (2.9%). These results suggest that both Tyr48 and His110 residues are essential for PGDS activity in the case of AKR1B3 (Fig. S2A–C).

The p-nitrobenzaldehyde reductase activity of AKR1B1 [291 nmol·min−1·(mg protein)−1 at pH 7.0, Fig. 3C] was decreased remarkably to less than 1% in the Y48F, K77L and H110F mutants, to 4% in the D43N mutant and to 12% in the H110A mutant. The AKR activity of the D43N and K77L mutants was partly restored in the charge-recovered (D43E and K77R) mutants to 41% and 20%, respectively (Fig. 3C). However, the p-nitrobenzaldehyde reductase activity of AKR1B3 [542 nmol·min−1·(mg protein)−1 at pH 7.0] was also decreased to less than 1% in the Y48F and H110F mutants (Fig. S2C). These results are consistent with previous reports that the p-nitrobenzaldehyde reductase activity of AKR1B1 is catalyzed by the triad composed of Tyr48, Lys77 and His110 and assisted by ionic interaction with Asp43 [25,26], and suggest that Tyr48 is also crucial for the p-nitrobenzaldehyde reductase activity of AKR1B3 (Fig. S2C).

Furthermore, all these mutants of AKR1B1 and AKR1B3 showed fluorescence quenching of intrinsic Trp residues after incubation with NADP+ in a concentration-dependent manner. The Kd values of NADP+ for AKR1B1 and AKR1B3 are summarized in Tables 1 and S1, respectively, and the typical fluorescence quenching curves of AKR1B1 and its mutants are shown in Fig. 3D. The Kd value of the K77L mutant of AKR1B1 was similar to that of the wild-type enzyme (0.3 μm) and the values of the D43N, D43E, Y48F, K77R, H110F and H110A mutants were 12–360 times higher than that of the wild-type AKR1B1. However, the Kd value of the Y48F/H110F double mutant of AKR1B3 was similar to that of the wild-type enzyme (5.7 μm) and the values of the Y48F and H110F mutants were 12 times higher than that of the wild-type AKR1B3. These results confirm that these mutations do not affect significantly the overall three-dimensional structure of the cofactor-binding site within the catalytic pocket.

Table 1.   NADP+-binding affinities of wild-type (WT) and mutants of AKR1B1.
 Kdm)
AKR1B1WT0.31 ± 0.06
D43N88 ± 39
D43E110 ± 49
Y48F22 ± 4
K77L0.69 ± 0.16
K77R45 ± 9
H110F27 ± 9
H110A5.2 ± 1.0

Discussion

Catalytic mechanism of PGDS and PGFS activities of AKR1B1

Mutational analysis of the catalytic tetrad of AKR1B1 and pH titration analysis revealed that the His110 residue functioned as a proton acceptor and donor during the conversion of PGH2 to PGD2 and PGF, respectively. pH titration analysis of PGDS and PGFS activities demonstrated that PGD2 formation required a deprotonated group with a pKa value of 7.9 for AKR1B1, and PGF formation required a protonated group with a pKb value of 6.2 (Fig. 2A). In the light of the expected acidity, His110 was deduced to act as the proton acceptor and donor for PGH2 to produce PGD2 and PGF, respectively, at physiologic pH, because the imidazolium side chain of His has a pKa value in the range 6–7, whereas the value for the hydroxyl group of Tyr is about 10, and those of Asp and Lys are about 3.6 and 10.5, respectively.

Molecular docking simulation of PGH2 to the crystallographic structure of AKR1B1 (PDB code, 2qxw; resolution, 0.8 Å) demonstrated that the substrate PGH2 was bound to the substrate-binding cavity in an extended conformation at the top of the (α/β)8-barrel (Fig. 4A,B). The docking calculation, including molecular dynamics, revealed that the 11-endoperoxide oxygen atom of PGH2 was accessible to His110 within the AKR tetrad at a distance of 2.9 Å, and the substrate PGH2 was stabilized by both hydrophobic and hydrophilic interactions with Trp20, Val47, Trp79, Trp111, Phe122, Pro218, Trp219 and Leu300 (Fig. 4C). In the presence of NADP+, when H atoms were added to the protein crystal structure of AKR1B1 by the myPresto/tplgene program [27] and used for the construction of the energy minimization model by the cosgene molecular dynamics simulation program, the distance between the H atom of the OH group (O34) of NADP+ and the carboxyl O atom of Asp43 was 2.61 Å, within a hydrogen-bonding distance.

Figure 4.

 Overall views of the ternary complex of human AKR1B1–NADPH–PGH2. The three-dimensional structure was calculated by the program sievgene/myPresto [http://medals.jp/myPresto/index.html; http://presto.protein.osaka-u.ac.jp/myPresto4/]. (A) Schematic drawing of the ternary structure using the program PyMOL version 1.0 [39] showing the TIM barrel structure of AKR1B1 (PDB code, 2QXW; resolution, 0.8 Å; green), NADP+ (yellow), the substrate, PGH2 (orange), Tyr48 (purple), Asp43 (green), Lys77 (cyan) and His110 (magenta). (B) Stereoview of the active site model with the NADPH (yellow) and PGH2 (orange) molecules (stick models) bound in the active-site cleft consisting of Asp43 (green), Tyr48 (purple), Lys77 (cyan) and His110 (magenta). (C) Schematic representation of the interactions between PGH2 (orange) and human AKR1B1. PGH2 is predicted to be stabilized by both hydrophobic (blue) and hydrophilic (red) interactions with Trp20, Val47, Trp79, Trp111, Phe122, Pro218, Trp219 and Leu300.

These results suggest that the His110 residue is the catalytic residue of PGDS and PGFS activity. The role of Lys77 could be to deprotonate the protonated His110, but it might just form a stable hydrogen bond, or a hydrogen bond network around the active site, to assist acid–base catalysis. The Asp43 residue is also important for the hydrogen bond network. Furthermore, the observation that Km for NADP+ is significantly altered in these mutants also suggests that Lys77 and Asp43 may have roles in NADPH binding as well as catalysis. The hypothetical catalytic mechanisms of PGDS and PGFS activities of AKR1B1 are shown schematically in Fig. 5. In the absence of NADPH, the concerted reaction of Asp43, Lys77 and His110 increases the basicity of His110 and extracts the proton C11 of PGH2. Another proton is provided to the O9 atom of PGH2 from an unidentified proton donor (EnzA-H) to produce PGD2. However, in the presence of NADPH, the hydride ion is transferred from NADPH to the O9 atom of the peroxide oxygen of PGH2, and a proton is provided from His110 to O11 to produce PGF. In the presence of NADP+, Asp43 forms a hydrogen bond with NADP+ and disrupts the catalytic triad, which is essential for the production of PGD2. However, the function of Tyr48 is not clear at present.

Figure 5.

 Schematic drawings of the PGDS (top) and PGFS (bottom) activities in the absence and presence of NADPH, respectively (see Discussion for detailed description).

Comparison of AKR1B1- and AKR1B3-catalyzed PGDS and PGFS activities with other AKR-mediated reactions

In the p-nitrobenzaldehyde reductase activity of AKR1B1 and AKR1B3 in the presence of NADPH, Tyr48 acts as the proton donor, consistent with previous reports from the mutational analysis of AKR1B1 [25,26] and various other members of the AKR superfamily [28,29], in which all AKRs have been shown to retain the same active site, and the conserved Tyr residue in the catalytic tetrad has been identified to play a crucial role in the catalysis of NADPH-dependent reduction. Alternatively, we propose a mechanism in the PGDS and PGFS reactions catalyzed by AKR1B1 in which His110 acts as a base in concert with Asp43 and Lys77 to generate PGD2 in the absence of NADPH or NADP+, and as an acid to generate PGF in the presence of NADPH.

However, the H110F mutant of AKR1B3 retained more than 25% of the wild-type PGDS activity, so that we generated a Y48F/H110F double mutant of AKR1B3. This double mutant completely lost PGDS activity and showed only 2.9% of PGFS activity (Fig. S2C). These results suggest that both Tyr48 and His110 residues are essential for PGDS activity in the case of AKR1B3, different from AKR1B1. The determination of the X-ray crystallographic structure of AKR1B3 is needed to elucidate the catalytic mechanism of PGDS and PGFS activities of AKR1B3.

We have reported previously that Trypanosoma brucei PGFS (AKR5A2 with 40.1% amino acid sequence identity with human AKR1B1) utilizes His110, but not Tyr48, as the catalytic residue for the reduction of PGH2 to PGF in the presence of NADPH. Therefore, the catalytic mechanism of PGFS activity of AKR1B1, AKR1B3 and T. brucei PGFS may be considered to be identical. However, T. brucei PGFS did not catalyze PGDS activity in the absence of NADPH or NADP+ (N. Nagata & Y. Urade, unpublished results). These results indicate that PGDS activity is selective to AKR1B1 and AKR1B3, but not AKR1B7, AKR1C3 and AKR5A2, and suggest that the tertiary structure of the catalytic pocket, especially the PGH2-binding site, of AKR1B1 and AKR1B3 is very similar and different from that of other members of the AKR family.

Human AKR1B1 has been reported recently to function as PGFS in the endometrium and is a potential target for the treatment of menstrual disorders [15], and mouse AKR1B3 has been reported to be involved in the suppression of adipogenesis through FP receptors [16]. Further characterization of the in vivo function of AKR1B1 in the endometrium and AKR1B3 in adipocytes as PGFS is essential to understand the development of menstrual disorders and metabolic disorders, such as diabetes and obesity, respectively. However, the catalytic mechanisms of PGFS catalyzed by several isozymes of mammalian AKRs are not clearly understood, because the X-ray crystallographic structures of AKR1B3 and AKR1B7 have not yet been determined. Our findings are useful for the design of inhibitors selective to AKR1B1, which can be employed for the evaluation of its contribution to the biosynthesis of PGF in various systems.

Materials and methods

Expression and purification of recombinant AKR enzymes

Open reading frames of the wild-type enzymes of AKR1B1, AKR1B3, AKR1B7 and AKR1C3, and their mutants, were inserted between NdeI and BamH1/EcoRI sites of the expression vector pET-28a, as described previously [30,31], and used for the transformation of E. coli BL21DE3 (Invitrogen, Carlsbad, CA, USA). The outside primers used for PCR amplifications of the inserts were as follows: 5′-1B1 NdeI primer (5′-CGGCAGCCATATGGCAAGCCGTC-3′) and 3′-1B1 EcoRI primer (5′-CGGAATTCGGGCTTCAAAACTCTTCATGG-3′); 5′-1B3 NdeI primer (5′-CGGCAGCCATATGGCCAGCCATC-3′) and 3′-1B3 EcoRI primer (5′-CACGAATTCCAGAGAGACACAGGACACTTGC-3′); 5′-1B7 NdeI primer (5′-CGGCAGCCATATGGCCACCTTCGT-3′) and 3′-1B7 BamHI primer (5′-CGGGATCCCGTCAGTATTCCTCGTGG-3′); and 5′-1C3 NdeI primer (5′-GGAATTCCATATGGATTCCAAACACCAGTG-3′) and 3′-1C3 EcoRI primer (5′-CGGAATTCTTAATATTCATCTGAATATG-3′). Site-directed mutagenesis was performed using a QuikChange® site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA, USA). The D43N-, D43E-, Y48F-, K77L-, K77R-, H110A- and H110F-substituted recombinant enzymes for AKR1B1 and the Y48F- and H110F-substituted recombinant enzymes for AKR1B3 were obtained using the following respective oligonucleotide primer pairs: AKR1B1 D43N forward (F) (5′-GTACCGCCACATCAACTGTGCCCATGTG-3′) and reverse (R) (5′-CACATGGGCACAGTTGATGTGGCGGTACC-3′); AKR1B1 D43E (F) (5′-GGGTACCGCCACATCGAATGTGCCCATGTG-3′) and (R) (5′-CACATGGGCACATTCGATGTGGCGGTACCC-3′); AKR1B1 Y48F (F) (5′-CTGTGCCCATGTGTTCCAGAATGAGAATG-3′) and (R) (5′-CATTCTCATTCTGGAACACATGGGCACAG-3′); AKR1B1 K77L (F) (5′-CTTCATCGTCAGCCTGCTGTGGTGCACG-3′) and (R) (5′-CGTGCACCACAGCAGGCTGACGATGAAG-3′); AKR1B1 K77R (F) (5′-CTCTTCATCGTCAGCAGGCTGTGGTGCACG-3′) and (R) (5′-CGTGCACCACAGCCTGCTGACGATGAAGAG-3′); AKR1B1 H110F (F) (5′-CCTCTACCTTATTTTCTGGCCGACTGGC-3′) and (R) (5′-GCCAGTCGGCCAGAAAATAAGGTAGAGG-3′); AKR1B1 H110A (F) (5′-CCTCTACCTTATTGCCTGGCCGACTGGC-3′) and (R) (5′-GCCAGTCGGCCAGGCAATAAGGTAGAGG-3′); AKR1B3 Y48F (F) (5′-GACTGCGCCCAGGTGTTCCAGAATGAGAAG-3′) and (R) (5′-CTTCTCATTCTGGAACACCTGGGCGCAGTC-3′); AKR1B3 H110F (F) (5′-GATCTCTACCTTATTTTCTGGCCAACGGGG-3′) and (R) (5′-CCCCGTTGGCCAGAAAATAAGGTAGAGATC-3′) (the italic codons indicate the sites of mutations). Transformed cells were precultured overnight at 30 °C. Induction was started by the addition of 1 mm isopropyl thio-β-d-galactoside (final concentration, 1 mm) when the absorbance (A) at 600 nm of the culture had reached 0.5–0.6, and further cultivation was carried out for 6 h at 30 °C. The recombinant protein was purified by chromatography with nickel nitrilotriacetate His-Bind resin (Merck, Darmstadt, Germany) according to the manufacturer’s protocol, followed by digestion with thrombin to remove the 6× His tag. The recombinant protein was further purified by gel filtration chromatography with Hiload 16/60 Superdex 75 pg (GE Healthcare, Amersham, Buckinghamshire, UK) in Dulbecco’s phosphate-buffered saline. Protein purity was assessed by SDS/PAGE on 10–20% gradient gels after staining with Coomassie Brilliant Blue. Protein concentrations were measured using a BCA Protein Assay Kit (Pierce Biotechnology, Rockford, IL, USA).

Enzyme activity assays

The PGFS and PGDS activities of AKR proteins were determined as described previously [32]. In brief, the purified recombinant enzymes were incubated at 37 °C for 2 min with 5 μm 1-[14C]PGH2 as a substrate in the presence or absence of 0.5 mm NADPH in 50 mm sodium phosphate, pH 7.0. The reaction was terminated by the addition of 300 μL of diethyl ether–methanol–2 m citric acid (30 : 4 : 1, v/v/v). The reaction products recovered into the organic phase were separated by TLC. The conversion rate from 14C-labeled substrate to 14C-labeled product was calculated using an imaging plate system (Fuji Film, Tokyo, Japan). The kinetic constants were determined from Lineweaver–Burk plots prepared with sigmaplot software (version 10.0 for Windows; Systat Software, Inc., San Jose, CA, USA).

For pH–rate profiles, Km values were calculated from initial velocity studies over a wide range of pH values using a triple buffer system containing 50 mm sodium phosphate, 50 mm sodium pyrophosphate and 50 mm 3-[(1,1-dimethyl-2-hydroxyethl)amino-2-hydroxypropanesufonic acid. In analyzing these data, the pKa and pKb values were estimated using the fitting equation

image(1)

prepared with sigmaplot software. C is the pH-independent value of V. The p-nitrobenzaldehyde reductase activity of AKR1B1 was measured with 0.2 mm NADPH and 1 mmp-nitrobenzaldehyde in 100 mm sodium phosphate (pH 7.0). The reaction was initiated by the addition of the substrate, and the decrease in the absorbance at 340 nm was monitored at 25 °C [9].

Fluorescence quenching assay

The binding of NADP+ to wild-type and mutant proteins of AKR1B1 was determined by performing a fluorescence quenching assay, in which various concentrations of coenzyme were incubated with AKR1B1 proteins in 300 μL of 50 mm sodium phosphate (pH 7.0) at 25 °C for 2 min. The intrinsic Trp fluorescence was measured using an FP-6200 spectrofluorometer (JASCO, Tokyo, Japan) operated at an excitation wavelength of 282 nm and an emission wavelength of 338 nm [33]. The Kd values for coenzyme binding to AKR1B1 proteins were calculated from the difference in fluorescence signal observed in the presence and absence of coenzyme, as reported previously [26], with sigmaplot software.

Molecular docking simulation

The docking study was performed by sievgene/myPresto (http://medals.jp/myPresto/index.html; http://presto.protein.osaka-u.ac.jp/myPresto4/) [27]. The prediction accuracies of the sievgene program have already been reported to be 19.2%, 50.78% and 60.0% with rmsd values of less than 1 Å, 2 Å and 3 Å, respectively, in a total of 130 complexes. Among the top 10 docking models, the probabilities increase to 28.5%, 63.1% and 76.9% with rmsd values of less than 1 Å, 2 Å and 3 Å, respectively [27]. The initial three-dimensional coordinates of the small compounds were generated by the Chem3D program (cambridge Software, Cambridge, MA, USA) manually. We used the general AMBER force field [34], and the molecular topology files were generated by tplgeneL/myPresto. The energy optimization of the coordinates of small compounds was performed using cosgene/myPresto [35]. The atomic charges were calculated by the Gasteiger method of Hgene/myPresto [36,37]. The atomic charges of the proteins were the same as the atomic charges of AMBER parm99 [38]. For flexible docking, the sievgene program generated up to 1000 conformers for each compound. We predicted that the C4 atom of nicotinamide reacts with the O atom of PGH2 and that these two atoms should be close to each other. Among the top 10 docking models, two models similar to each other showed C4–O distances of 2.0 and 2.1 Å, consistent with the experimental results, whereas the other eight models showed C4–O distances of more than 6.5 Å. The complex structure depicted in Fig. 4 was given by the energy minimization calculation based on the model with a C4–O distance of 2.0 Å. The position of the ring structure of PG should be 70–80% accurate based on the prediction accuracy of ‘sievgene’. Although it is difficult to predict the position of side chains of PG, the side chains are not important for the discussion of the reaction mechanism.

Acknowledgements

This work was supported in part by the Japan Aero-space Exploration Agency (JAXA), the Program of Basic and Applied Researches for Innovations in Bio-oriented Industry of Japan, Takeda Science Foundation, and Osaka City (to Y.U.) and Grant-in-Aid for Scientific Research (No. 22550152) from the Ministry of Education, Culture, Sports, Science and Technology of Japan (to T.I.). We thank Dr Michele Manin (CNRS UMR6247-GReD, France) for kindly providing the AKR1B1 expression vector; Drs Kenji Mizuguchi and Sukanta Mondal (National Institute of Biomedical Innovation, Ibaraki, Japan) for homology modeling; Dr Toshiyoshi Yamamoto (Department of Molecular Biology and Medicine, Research Center for Advanced Science and Technology, University of Tokyo, Japan) for kinetic analysis; Dr Zakayi Kabututu, Nobuko Uodome and Toshiharu Tsurumura (Osaka Bioscience Institute, Japan) for assistance during the early stage of this research; and Megumi Yamaguchi, Naoko Takahashi and Taeko Nishimoto (Osaka Bioscience Institute, Japan) for secretarial assistance.

Ancillary