Structural basis of the catalytic mechanism operating in open-closed conformers of lipocalin type prostaglandin D synthase.

Lipocalin type prostaglandin D synthase (L-PGDS) is a multifunctional protein acting as a somnogen (PGD2)-producing enzyme, an extracellular transporter of various lipophilic ligands, and an amyloid-beta chaperone in human cerebrospinal fluid. In this study, we determined the crystal structures of two different conformers of mouse L-PGDS, one with an open cavity of the beta-barrel and the other with a closed cavity due to the movement of the flexible E-F loop. The upper compartment of the central large cavity contains the catalytically essential Cys65 residue and its network of hydrogen bonds with the polar residues Ser45, Thr67, and Ser81, whereas the lower compartment is composed of hydrophobic amino acid residues that are highly conserved among other lipocalins. SH titration analysis combined with site-directed mutagenesis revealed that the Cys65 residue is activated by its interaction with Ser45 and Thr67 and that the S45A/T67A/S81A mutant showed less than 10% of the L-PGDS activity. The conformational change between the open and closed states of the cavity indicates that the mobile calyx contributes to the multiligand binding ability of L-PGDS.

Prostaglandin (PG) D synthase (PGDS; PGH2 d-isomerase (EC 5.3.99.2)) (1, 2) produces PGD2, having 9α-hydroxy and 11-keto groups, from PGH2, which bears the chemically labile 9,11-endoperoxide group and is produced as a common intermediate of all prostanoids by the action of cyclooxygenase (PGH2 synthase). Two distinct types of PGDS have evolved from phylogenetically distinct protein families (2, 3). One is hematopoietic PGDS (H-PGDS), which belongs to the σ class of GSH S-transferases (4, 5), and the other is lipocalin type PGDS (L-PGDS), a member of the lipocalin family (6, 7). L-PGDS is the only enzyme in the lipocalin family and is identical to β-trace, a major protein in human cerebrospinal fluid (8, 9). Although H-PGDS and L-PGDS catalyze the same reaction, their amino acid sequences and tertiary structures are quite different from each other, indicating that these enzymes are a new example of functional convergence (2, 3).

L-PGDS is expressed in the heart, central nervous system, and male genital organs of various mammals and is involved in various physiological and pathological functions (reviewed in Refs. 6 and 7). In the brain, L-PGDS produces PGD2, which is involved in the regulation of pain and non-rapid eye movement sleep, as was shown in studies using gene knock-out mice (10, 11) and human enzyme transgenic mice (12). L-PGDS is regulated by SOX9 and is involved in the differentiation of male genital organs (13–15). This enzyme is also expressed in adipocytes (16), vascular smooth muscle cells (17), and myocardial cells (18, 19) and is involved in adipocyte differentiation, the progression of arteriosclerosis (20), and the protection against hypoxemia (18) or ischemia/reperfusion injury (19). L-PGDS binds various lipophilic compounds, such as retinoids (21), bilirubin, biliverdin (22), gangliosides (23), and amyloid-β peptides (24, 25), with high affinity, acting as an extracellular transporter of these compounds and serving as an endogenous amyloid-β chaperone to prevent amyloid deposition in vivo (24).

Although many biochemical and physiological studies suggest important roles of PGD2 and L-PGDS/β-trace in the regulation of sleep and other biological functions, the crystal structure of L-PGDS has not been resolved. In this study, we determined the crystal structures of two different forms of the Δ1–24-C65A mutant of mouse L-PGDS in both open and closed conformations. L-PGDS was shown to possess a typical lipocalin fold, the β-barrel, which is a unique structural component specific to L-PGDS and comprises a mobile E-F loop and a large central cavity with two compartments. By performing site-directed mutagenesis of Δ1–24-L-PGDS and the Δ1–24-C65A mutant, we found that the Cys65 surrounded by the hydroxyl side chains of Ser45, Thr67, and Ser81 was activated to contribute to the catalysis by L-PGDS.

EXPERIMENTAL PROCEDURES

Preparation and Crystallization

The Δ1–24-C65A mutant of mouse L-PGDS was expressed in Escherichia coli, purified to homogeneity, and crystallized, as reported previously (26). The selenomethionyl (Se-Met) protein was also expressed in E. coli cultured in LeMaster broth with Se-Met and purified by using the same method (26). Both native and Se-Met proteins were crystallized in a hanging drop vapor diffusion chamber at 22.5 °C, as described earlier (26). The mixed drop consisted of 2 μl of the 10 mg/ml Se-Met protein solution and an equal volume of mother liquor containing 1.25 m sodium citrate, 10% (v/v) 1,4-dioxane, 2% (v/v) Triton X-405, and 10 μm retinoic acid in 0.1 m Tris-HCl (pH 9.5). The bar-shaped crystals grew to their maximum dimensions of 0.1 × 0.1 × 0.4 mm within 3 weeks. Crystals of the Se-Met protein belonged to the orthorhombic space group C2221 with the cell dimensions of a = 45.7, b = 66.8, and c = 104.5 Å and contained 1 protein molecule/asymmetric unit. The Δ1–24-C65A crystals with cell dimensions of a = 46.3, b = 67.1, and c = 104.6 Å were obtained under the same conditions but without Triton X-405. Other crystals of Δ1–24-C65A were obtained from 2 m malonate in 0.1 m Tris-HCl (pH 8.0) and showed a different space group, P212121, with similar unit cell dimensions of a = 46.2, b = 66.8, and c = 105.3 Å and contained 2 molecules/asymmetric unit.

Data Collection and Structure Determination

Diffraction data sets of both Se-Met and native crystals were collected at the RIKEN beamline I (BL45XU-PX) at SPring-8 (27, 28). The crystals were washed in the mother liquor containing 0.5 m trehalose for cryoprotection and transferred into a liquid N2 gas stream at 100 K. Collected multiwavelength anomalous dispersion data sets were processed by using the programs DENZO and SCALEPACK (29). The CCP4 suite was used for successive crystallographic calculations (30). The sites of the three selenium atoms were found from selenium Bijvoet and dispersive difference Patterson maps. Phase refinement and density modification were performed with the SHARP (31) and SOLOMON programs (32). The programs O (33) and XtalView (34) were used for model building. Crystallographic refinement was performed by alternating cycles of CNS (35) refinement and manual rebuilding of the model. The crystallographic R and Rfree values were 0.23 and 0.28, respectively, calculated from diffraction data up to 2.5 Å resolution. The partially refined model of the Se-Met C65A crystal was the search model for successive analysis of the native crystals done by using the molecular replacement method. The models of the molecular replacement phasing were refined by REFMAC5 in the CCP4 suite (30). The final crystallographic R and Rfree values were 0.24 and 0.28, respectively, for both forms of crystals. There were no residues in the disallowed region of the Ramachandran plots in either of the structural coordinates. The crystallographic statistics are summarized in Table 1.

TABLE 1

Data collection and structure determination statistics

Parameter	Value
Data collection statistics
Space group	P212121	C2221	C2221
Data collection	Native	Native	Se-Met
Unit cell dimension (Å)	a = 46.2, b = 66.8, c = 105.3	a = 46.3, b = 67.1, c = 104.6	a = 45.7, b = 66.8,c = 104.5
			Edge	Peak	Remote
Wavelength (Å)	1.0100	1.0100	0.9803	0.9797	1.1000
Resolution (Å)	56.8-2.10	10.0-2.0	2.50	2.50	2.50
No. of total reflections	61,094	32,068	31,453	38,570	29,691
No. of independent reflections	16,942	8,773	5,801	5,846	5,673
Completeness (%)	91.6	77.2	98.4	99.3	98.5
Rmerge (%)	4.7	5.2	4.8	7.4	5.3
Refinement statistics
Resolution (Å)	56.8-2.10	10.0-2.00	20.0-2.50
No. of reflections used for Ra	16,942 (90.9)	7,821 (77.06)	5,419 (93.2)
No. of reflections used for Rfreea	922 (5.2)	850 (9.8)
R (%)b	24.3 (31.5)	24.2 (26.6)
Rfree (%)b	27.9 (32.8)	27.8 (31.4)
No. of atoms (protein/solvent)	2,425/102	1,219/68
Root mean square deviation in bond length (Å)/angles (degrees)	0.006/1.48	0.012/1.39

Completeness of reflections against expected reflections/proportion of total reflections in parentheses (%).

The highest shell R/Rfree value (%) is shown in parentheses.

Structural Comparison and Modeling

Amino acid sequences of L-PGDS from various mammals were aligned by using T-COFFEE (36). The electrostatic potential calculations and comparisons of L-PGDS and ligands with other lipocalins were modeled by using Quanta/CHARMm (Accelrys) and PyMOL (DeLano Scientific, LLC). All figures were prepared by using PyMOL and Grasp (37).

Site-directed Mutagenesis

All mutants of mouse Δ1–24- and Δ1–24-C65A L-PGDS were obtained using a QuikChange® site-directed mutagenesis kit (Stratagene). All mutants were overexpressed and purified as described previously (26), except that the proteins were expressed at 16 °C.

PGDS Assay

PGDS activity was measured as described previously (1). Briefly, L-PGDS was incubated at 25 °C for 1 min with [1-14C]PGH2 in 50 μl of 0.1 m Tris-HCl (pH 8.0) containing 1 mm dithiothreitol or 2 mm GSH as the reducing agent, and the use of either dithiothreitol or GSH made no difference. The products were separated by thin layer chromatography. The conversion rate from [1-14C]PGH2 to [1-14C]PGD2 was measured by using an imaging plate system (Fuji Film, Tokyo, Japan). For determination of the K value of either the U-44069 or U-46619 inhibitor (Cayman Chemical), the PGDS activity was determined with 2 mm GSH and 5 PGH2 concentrations (2.5–40 μm) in the absence or the presence (40, 80, or 160 μm) of each inhibitor (Fig. 2B). The specific activity of Δ1–24-L-PGDS (Vmax = 5.9 μmol/min/mg protein) was comparable with that of L-PGDS purified from rat brain (1) and higher than that of the enzyme from human cerebrospinal fluid (9).

FIGURE 2.

Kinetics and binding of PGH A, reaction equation of PGDS from PGH2 to PGD2 and chemical structures of PGH2 analogues U-44069 and U-46619. B, Lineweaver-Burk plots of the Δ1–24-l-pgds activity inhibited by U-44069 and U-46619. Shown are sensograms (C) and the dose-response curves (D) of total (dashed lines), nonspecific (dotted lines), and specific (solid lines) binding of U-44069 (white circle) and U-46619 (black circle) to Δ1–24-C65A L-PGDS, as assessed by the surface plasmon resonance analysis.

Assays for Binding of U-44069, U-46619, and Retinoic Acid

The binding affinities of U-44069 and U-46619 (Cayman Chemical) for L-PGDS and its various mutants were determined by using the surface plasmon resonance assay on a Biacore 2000 (GE Healthcare), as reported previously (24). The binding affinity of the L-PGDS mutants for all-trans-retinoic acid was determined by monitoring fluorescence quenching of intrinsic tryptophan residues after incubation with various concentrations of all-trans-retinoic acid, as described previously (21).

Assays for Activated SH Group

The activation of the Cys65 residue was measured spectrophotometrically with a V560 spectrophotometer (JASCO, Tokyo, Japan) at 340 nm after incubation with a 50 μm concentration of 2,2′-dithiodipyridine (Dojindo Chemical, Kumamoto, Japan), a thiol-specific reacting agent, and 5 μm enzyme, under a low reactivity condition in 0.1 m sodium citrate (pH 4.0) containing 1 mm EDTA (38, 39).

Data Analysis

All of the analyses, including analysis of PGDS activity, were calculated with non-linear fitting for each target function by using S.D. values. SigmaPlot was used for the data analyses (Systat Software).

RESULTS

Overall Structure

We crystallized mouse Δ1–24-C65A L-PGDS as two different crystal forms in the presence of retinoic acid as a crucial additive for crystallization (26). These crystals were grown from citrate or malonate as precipitants with space groups of C2221 or P212121, respectively. The molecular packing forms of these crystals were principally the same regardless of space group differences. Two molecules in a local 2-fold symmetry in the P212121 crystal corresponded to those in a crystallographic 2-fold symmetry in the C2221 crystal. The first two amino acid residues from the linker sequence and the following nine N-terminal residues of L-PGDS (GS-25QGHDTVQPN33) in both crystals and Gln88 in the C2221 crystal were not included in the models due to poor electron density (Fig. 1, A–C). The conformational differences between the C2221 and P212121 crystals were observed at the top and bottom ends of the β-barrel structure of L-PGDS. In an asymmetric unit of the C2221 crystal, the L-PGDS molecule was in the closed conformation. In that of the P212121 crystal, one of the two L-PGDS molecules was in the open conformation, but we could not determine the conformation of the other molecule due to the poor electron density of the E-F loop.

FIGURE 1.

Crystal structures of L-PGDS with open and closed cavities. A, ribbon diagrams of the l-pgds structure. β-Strands and α-helices are labeled A-I and 1–3, respectively. The open and closed l-pgds structures of the P212121 and C2221 crystals were superimposed on an orthogonal view of the barrel. The conformational differences between open and closed forms are colored in sky blue and pink, respectively, in the E-F loop, Phe34 of an N-terminal tail (N) and Gln88 of the C-D loop, in which Phe34 (F34), Trp54 (W54), Gln88 (Q88), Pro110 (P110), and His111 (H111) are shown in a stick model form. The Phe34 residue was defined in the closed form but disordered in the open form, unplugging the bottom of the β-barrel entry, whereas Gln88 of the C-D loop was identified in the open form (Q88) but not in the closed form. C65A residue (C65A), a disulfide bond between Cys89 and Cys186 (C89/186), and two putative N-linked glycosylation sites, Asn51 (N51) and Asn78 (N78) are also indicated as stick models (green, carbon; red, oxygen; blue, nitrogen). B, l-pgds structures in another view are shown after about 90° rotation of A from the vertical. C, another view of the open-closed l-pgds structures is represented after horizontal rotation of B. D, a view of the superimposed mobile E-F loops from the closed (pink) and open (sky blue) structures. Trp54 and His111 at the top of the barrel are shown as stick models. The Pro110 residue in the up and down conformations of pyrrolidine ring puckerings (upP110o in sky blue and downP110c in pink) is shown in the respective open and closed forms of the l-pgds structures. E and F, architectures of the mobile E-F loop in the open (E) and closed (F) forms. The non-bonding interactions are indicated by dashed lines. In F, Trp54 in helix 2 of the Ω loop is also shown. G, open and closed lids of the l-pgds cavity are represented as a silver molecular surface. The open and closed E-F loops are colored in sky blue and pink, respectively (left and right panels, respectively). H, the upper hydrophilic compartment with several polar residues, including the C65A (C65A), Ser45 (S45), Thr67 (T67), Ser81 (S81), His116 (H116), Ser133 (S133), Thr147 (T147), and Tyr149 (Y149) shown in stick-model form (green, carbon; red, oxygen; blue, nitrogen).

The crystal structure of L-PGDS in both the C2221 and P212121 crystals showed a typical lipocalin fold (i.e. the β-barrel, which consisted of a single domain that comprised an eight-stranded anti-parallel β-barrel (strands A–H), three α-helical regions (regions 1–3), and the C-terminal β-strand (I) (Fig. 1, A–C, and supplemental Fig. S1). The barrel was 40 × 30 × 35 Å and linked with the conserved disulfide bridge of Cys89–Cys186 at the outside of the barrel. The outer molecular surface of l-pgds was mostly composed of charged or polar amino acid residues. The molecular surface of the putative substrate entry point of the enzyme was electrostatically positive to attract the negatively charged PGH2 (supplemental Fig. S2). The catalytic residue Cys65, which was replaced by Ala in the crystal structure, was located at the N-terminal edge of β-strand B (Fig. 1A). The N-linked glycosylation sites Asn51 and Asn78 (40) were located at the base of the Ω-loop (A-B loop containing short α-helix 2) and at the N terminus of β-strand C, respectively, at the antipodes of the l-pgds molecule (Fig. 1B).

Open-Closed Conformers

In the crystal structure with the space group of P212121, both the upper entrance (Fig. 1, D and left panel of G) and the bottom of the large central cavity were open to the exterior. On the other hand, in the C2221 structure, the large cavity was isolated from the protein exterior by closure due to an aromatic bridge between Trp54 of the Ω-loop and His111 of the E-F loop at the top of the barrel (Fig. 1, D and right panel of G) and by plugging of the bottom of the barrel by Phe34 (Fig. 1, A and B).

In the open conformation, β-strands E and F were stretched (Fig. 1, D and E). The short E-F loop was stabilized by the conformationally strained residue Pro110 synergizing with the backbone and side chain hydrogen bonds among Ser109, Ser112, and Ser114 (Fig. 1E). On the other hand, in the closed conformation, β-strands E and F were shortened (Fig. 1F), and the E-F loop was extended with a β-sheet-like bent structure formed by hydrogen bonding between the hydroxyl group of Ser109 and the backbone imino group of Ser114 (Fig. 1F). The E-F loop was bent toward Trp54 in helix 2 to make an aromatic bridge with His111 of the E-F loop in the closed conformer. The basal region of the expanded E-F loop may have been stabilized by the interaction between the aromatic ring of Tyr107 on strand E and that of His116 on strand F, regardless of the conformer type. The open-closed structures indicated that the loops of the calyx are mobile and flexible and are consistent with the fact that the crystallographic B-factors are higher for this loop region than those for the envelope core structure.

Compartmentalized Central Cavity

The large central cavity of l-pgds was separated into two compartments by five large hydrophobic amino acid residues: Leu79, Leu96, Val118, Leu131, and Tyr149 (supplemental Fig. S3). The upper compartment contained eight polar residues (C65A, Ser81, Thr67, Ser45, Tyr149, Thr147, Ser133, and His116) surrounded by hydrophobic amino acids (Fig. 1H). The catalytically essential Cys65 residue was surrounded by a cluster of the hydroxyl side chains of Ser45, Thr67, and Ser81 within the range of hydrogen bonding, suggesting that these residues form a hydrogen bond network to contribute to the activation of the thiol group of Cys65 and to stabilize the thiolate anion as the catalytic species, as described below.

The lower compartment was composed of abundant hydrophobic amino acid residues, similar to those of other lipocalins, such as epididymal retinoic acid-binding protein (41) and neutrophil gelatinase-associated protein (lipocalin 2 or siderocalin) (42), suggesting the presence of evolutionary pressure to maintain the same interior environment among these lipocalins. There were only two distinct corresponding amino acid residues between l-pgds and epididymal retinoic acid-binding protein (i.e. Gln36 and Ile8, respectively) at the bottom of the β-barrel and Leu96 and Ala67 as a compartmentalized residue, respectively, in the middle of the central cavity (supplemental Fig. S4). The bulky side chain of Leu96 in l-pgds narrowed the tunnel between the upper and lower compartments of the central cavity as compared with the wide tunnel of epididymal retinoic acid-binding protein.

Mutational Analysis

Based on the crystal structures, we used site-directed mutagenesis to change several residues in the putative catalytic pocket of the Δ1–24-l-pgds and its C65A mutant and measured the enzyme activity and the binding affinities for U-44069, U-46619, and all-trans-retinoic acid. U-44069 and U-46619 are the stabilized regioisomer analogues of the substrate PGH2 at C9 and C11, respectively, whose oxygen atom of the endoperoxide group is substituted by a carbon atom in either analogue (Fig. 2A). These compounds inhibited the PGDS activity in a competitive manner with a K of 30 and 36 μm, respectively (Fig. 2B). The results of the mutational analysis are summarized in Table 2.

TABLE 2

Kinetic and binding parameters of L-PGDS

Shown are K and Vmax for PGH2 in PGDS activity and K for U-44069, U-46619, and all-trans-retinoic acid of the mutants.

	PGH2		U-44069 Kd	U-46619 Kd	All-trans-retinoic acid Kd
	Km	Vmax
	μm	μmol/min/mg protein	μm	μm	μm
Δ1–24-C65A		NDa	20.2 ± 2.5	22.1 ± 1.5	233 ± 148
Δ1–24-C65A + W54A		NDa	49.2 ± 3.7	42.4 ± 3.4
Δ1–24-C65A + P110A		NDa	22.4 ± 1.4	23.3 ± 2.0
Δ1–24-C65A + H111A		NDa	38.1 ± 2.4	37.2 ± 3.5
Δ1–24-L-PGDS	0.8 ± 0.1	5.4 ± 0.2	17.5 ± 1.2	22.8 ± 1.6	292 ± 195
Δ1–24-L-PGDS + W54A	1.3 ± 0.5	3.6 ± 0.2	20.3 ± 1.9	44.0 ± 3.9
Δ1–24-L-PGDS + P110A	1.6 ± 0.5	4.8 ± 0.3	13.5 ± 3.1	26.3 ± 3.1
Δ1–24-L-PGDS + H111A	1.4 ± 0.4	5.1 ± 0.2	20.6 ± 2.3	20.3 ± 1.3
Δ1–24-L-PGDS + H116A	1.4 ± 0.1	4.4 ± 0.1	27.3 ± 2.5	48.5 ± 4.4
Δ1–24-L-PGDS + W54A/H111A	0.8 ± 0.1	4.9 ± 0.1	20.6 ± 1.4	48.5 ± 4.4	301 ± 242
Δ1–24-L-PGDS + S45A	0.5 ± 0.2	2.3 ± 0.1	15.2 ± 2.4	16.2 ± 1.4
Δ1–24-L-PGDS + T67A	2.8 ± 0.9	2.8 ± 0.9	12.3 ± 3.0	26.9 ± 5.3
Δ1–24-L-PGDS + S81A	1.5 ± 0.4	4.7 ± 0.2	14.6 ± 2.4	17.0 ± 1.7
Δ1–24-L-PGDS + S45A/T67A	2.1 ± 0.6	1.6 ± 0.1	7.2 ± 1.6	19.1 ± 2.5
Δ1–24-L-PGDS + S45A/S81A	1.2 ± 0.3	1.5 ± 0.1	12.0 ± 2.7	19.1 ± 2.6
Δ1–24-L-PGDS + T67A/S81A	2.1 ± 0.2	1.3 ± 0.0	13.5 ± 2.3	19.6 ± 2.3
Δ1–24-L-PGDS + S45A/T67A/S81A	1.2 ± 0.1	0.5 ± 0.0	12.9 ± 2.4	22.0 ± 2.0	418 ± 124

ND, not detected (<0.01 μmol/min/mg protein).

The Δ1–24-C65A l-pgds did not show any enzymatic activity at all but possessed the same binding affinities for both U-44069 and U-46619 (K = 20 and 22 μm, respectively) and retinoic acid (K = 0.23 μm) as those of Δ1–24-l-pgds (Vmax = 5.9 μmol/min/mg protein and K = 0.8 μm in PGDS activity; K = 18, 23, and 0.30 μm, respectively) (Fig. 2, C and D). These results indicate that the Δ1–24-C65A mutant possessed substantially the same substrate- and retinoic acid-binding pockets as did Δ1–24-l-pgds.

We then substituted the amino acid residues of the aromatic bridge between Trp54 in helix 2 and Pro110 and His111 in the flexible E-F loop (Fig. 1, C and D). All of these mutations resulted in only moderate effects on the PGDS activity (K = 0.8–1.6 μm, Vmax = 3.6–5.1 μmol/min/mg protein) and on the binding of U44069 (K = 13.5–20.6 μm), U46619 (20.3–48.5 μm), and all-trans-retinoic acid (0.30 μm).

The hydroxyl cluster of Ser45, Thr67, and Ser81, which surrounded Cys65 in the catalytic pocket, was then substituted with Ala in Δ1–24-l-pgds (Fig. 1H). Among the single mutants, the PGDS activity decreased to 40–50% for the S45A and T67A mutants (Vmax = 2.3 and 2.8 μmol/min/mg protein, respectively) and to 80% for the S81A mutant (Vmax = 4.7 μmol/min/mg protein; Table 2). Double mutations of these residues decreased the PGDS activity (Vmax = 1.3–1.6 μmol/min/mg protein) more significantly than did the single mutation. The S45A/T67A/S81A triple mutant showed less than 10% of the PGDS activity (Vmax = 0.5 μmol/min/mg protein). However, all of these mutants showed almost identical affinities for PGH2 (K = 0.5–2.8 μm), U-44069 (K = 7.2–15.2 μm), U-46619 (K = 16.2–26.9 μm), and retinoic acid (K = 0.4 μm).

Activation of Cys65 in L-PGDS

In an SH titration assay conducted at acidic pH 4 with 2,2′-dithiodipyridine (Fig. 3A), the chromogenic reaction was 16-fold faster with Δ1–24-l-pgds than with GSH, indicating that the SH group of Cys65 had been activated to become a thiolate anion. Among the various mutants of Δ1–24-l-pgds, the SH activation was unchanged in the S81A mutant and W54A, P110A, and H111A mutants (data not shown for these three mutants) but decreased to 47% in the W54A/H111A mutant, to 9% in the S45A mutant, and to 7% in the T67A mutant. The S45A/T67A/S81A triple mutant showed the same rate of chromogenic reaction as that of GSH. When the Δ1–24-C65A l-pgds was used, no chromogenic reaction was detected. These results indicate that the hydroxyl cluster surrounding Cys65 is significant for activation of the SH group and that the hydrogen bond network of the catalytic Cys65 with the hydroxyl cluster of Ser45, Thr67, and Ser81 synergistically contributes to a decrease in the pK of the thiol of Cys65 and stabilizes the thiolate anion as the reactive group in the PGDS reaction at physiological pH.

FIGURE 3.

Activation of the thiol group of Cys A, thiol titration assay of Δ1–24-l-pgds and its mutants at pH 4 with 2,2′-dithiodipyridine. The chromogenic reaction was monitored spectrophotometrically at 340 nm, as described under “Experimental Procedures.” B, hydrogen bond networks of l-pgds, H-PGDS/the σ class of GSH transferases (σGST), and the θ class of GSH transferases (θGST). C, amino acid residues responsible for thiol activation in l-pgds (pink carbon), H-PGDS (blue), and the θ class of GSH transferases (green) in stereo. (Other color codes are the same as in Fig. 1).

DISCUSSION

Comparison of the Crystal Structures of L-PGDS with Its NMR Structures and SAXS Models

When we compared the crystallographic structures of Δ1–24-C65A l-pgds with the NMR structures of the Δ1–24-C89A/C186A mutant of mouse l-pgds (Protein Data Bank code 2E4J) (43), the overall structure of the core region of the β-barrel structure was essentially identical (Fig. 4A) and similar to the case of other lipocalin structures. In the Cα carbon atoms of the closed form crystal structure and 15 NMR structures (43), the root mean square deviation of the core structures was 1.11 Å, and that of overall structures was 1.70 Å. The short helix 1 in the N-terminal region, helix 2 in the Ω-loop, and the long helix 3 in the C-terminal region were significantly shifted in the NMR structure from their positions in both of the crystal structures. The unidentified N-terminal sequence and the mobile E-F loop in the crystal structures are consistent with the HN signal broadening of the first eight residues at the N terminus, the E-F loop, and the G-H loop in the NMR structure. The crystal structures were also well superimposed on the SAXS molecular envelope of mouse l-pgds (Fig. 4B), in which binding of lipophilic ligands to l-pgds was demonstrated to induce compact packing of the l-pgds molecule (44). The crystal structures of open-closed conformers, together with those reported previously, indicate that l-pgds is a flexible molecule that can bind diverse ligands, including PGH2, U-44069, U-46619, and retinoic acid.

FIGURE 4.

Superimposed models of the crystal structures of Δ The crystal structures of the open (pink) and closed (sky blue) forms of Δ1–24-C65A l-pgds in schematic models are superimposed on 15 NMR structures of the Δ1–24-C89A/C186A l-pgds in hair ribbon models (PDB code 2E4J) (43) or a transparent surface model of mouse apo-l-pgds by SAXS (44). Phe34, Trp54, Pro110, and Cys89–Cys186 residues are represented by stick models with transparent spheres in A. (Color codes are the same as in Fig. 1A).

Activated Thiol as the Common Reactant between L-PGDS and H-PGDS

In this report, we also demonstrated that the Cys65 residue of l-pgds was activated by the hydrogen bond interaction with the surrounding hydroxyl side chains of Ser45, Thr67, and Ser81 in the catalytic pocket. The changes in activation of this Cys65 residue in Ala-substituted mutants of the hydroxy cluster was well correlated with synergistic decreases in the Vmax values of PGDS activity without a change in the K for PGH2 and K for U-44069 and U-46619, indicating that the thiolate anion of Cys65 plays a pivotal role in the reaction mechanism of l-pgds (Table 2). The relative contribution of those amino acids to activate the Cys65 residue was on the order of Ser45 ≈ Thr67 > Ser81 according to the reactivity with 2,2′-dithiodipyridine at acidic pH (Fig. 3A). In a model of the hydrogen bond cluster based on the crystal structures of Δ1–24-C65A l-pgds (Fig. 3B), Cys65 formed a hydrogen bond with Ser45 and Thr67 and secondarily with Ser81 to facilitate synergistically the thiolate formation of Cys65, being consistent with the thiol modification experiments.

The activation of Cys65 to the thiolate anion by the hydrogen bond network in l-pgds is similar to the case of H-PGDS, which is a member of the σ class of GSH transferases and the θ class of GSH transferases (Fig. 3, B and C). In H-PGDS, the phenolic hydroxyl group of Tyr8 activates the cysteinyl thiol of the bound GSH and decreases the pK from 8.5 to 7.8 (45), similar to many other members of the GSH transferase family. Alternatively, in the θ class of GSH transferase (46), the activated thiol group of the bound GSH is also stabilized by hydrogen bonding with hydroxyl groups of serine residues, as was shown by kinetic analysis.

Reaction Mechanism of L-PGDS

We proposed a reaction mechanism of l-pgds (Fig. 5) based on the similarity of the reaction mechanism between l-pgds and H-PGDS. In the proposed mechanism, PGH2 is bound to the open form of l-pgds and induces the closed conformation of the catalytic pocket with the proper geometry to allow the endoperoxide oxygen atom of C11 of PGH2 to face the sulfur atom of Cys65 for the PGDS reaction (Step 1). The thiolate anion of Cys65 exerts a nucleophilic attack on the endoperoxide oxygen atom at C11 at the base of the closed cavity (Step 2), giving the putative S-O adduct as an unstable reaction intermediate (Step 3). The labile S–O bond breaks autonomously with the proton rearrangement at the hydroxyl of Ser45 and the C11 to form the carbonyl group in a concerted manner (Steps 3 and 4). The product PGD2 is released after the opening of the Trp54–His111 gate together with a conformational change in the E-F loop associated with the change in up- and down-puckerings of the pyrrolidine ring conformations of Pro110 (47). The thiol proton of Cys65 dissociates to form the thiolate anion again as the reactive specimen.

FIGURE 5.

Docking model of PGH See “Discussion” for detailed description.

Similarity of the Three-dimensional Structure of the Catalytic Pocket between L-PGDS and H-PGDS

l-pgds and H-PGDS have distinct protein folds as a member of the lipocalin (1, 2) and the σ class of GSH transferase (3, 5, 7) families, respectively (Fig. 6A). However, both enzymes are predicted to possess a similar tertiary structure of the catalytic pocket; because we recently found AT-56, a competitive inhibitor selective for l-pgds (48), from derivatives of HQL-79, which is a competitive inhibitor of H-PGDS (49). We then compared the geometry of the amino acid residues involved in the catalysis and the substrate binding within their catalytic pocket and superimposed to it on the three-dimensional view (Fig. 6B). The active thiol of Cys65 in l-pgds is surrounded by the hydroxyl cluster of Ser45, Thr67, and Ser81, whereas the active thiol of GSH in H-PGDS interacts with the hydroxyl of Tyr8, which overlaps the position of the hydroxyl cluster of l-pgds. The (15S)-hydroxy group-bearing ω-chain of PGH2 is lined with the hydrophilic polar residues Ser133, Thr147, and Tyr149 in l-pgds and Arg14, Tyr152, and Cys156 in H-PGDS. The other tail, the α-chain of PGH2 with a negatively charged carboxyl end, may be considered to be held by a pair of the countercharged groups of Arg85 and Lys92 in l-pgds and of Lys112 and Lys198 in H-PGDS at their protein exterior. In the docking model of PGH2 within the catalytic pocket (Fig. 5A), the cyclopentane head of PGH2 with its endoperoxide group is wrapped by hydrophobic residues, such as Phe83 and Pro110, the side chain conformation of the latter residue being altered in the flexible E-F loop between the open-closed conformers, as described above. In the superimposed model, Pro110 of l-pgds overlaps Trp104 of H-PGDS, which plays a role in opening the catalytic cleft to solvent.

FIGURE 6.

Comparison of the three-dimensional architectures of whole molecules of L-PGDS and H-PGDS ( A, l-pgds is shown on the left and H-PGDS on the right in rainbow-colored schematic diagrams with transparent molecular surfaces. Cys65 in l-pgds and GSH bound to H-PGDS are shown as ball and stick models (black, carbon; red, oxygen; blue, nitrogen; yellow, sulfur). B, superimposed corresponding side chains involved in PGDS catalysis by l-pgds (stick model of pink-colored carbon atoms) and by H-PGDS (sky blue carbon atoms).

This binding mode of PGH2 to l-pgds is different from the previously proposed mode (43), which was predicted from the chemical shift in the NMR spectra after binding of U-44069 and the increase in K and the decrease in Vmax of the PGDS activity by the H116A mutation in Δ1–24-C89A/C186A l-pgds without the Cys89–Cys186 disulfide bridge. However, the H116A mutation in Δ1–24-l-pgds with the S–S bridge had little impact on the PGDS activity (K = 1.4 μm, Vmax = 4.4 μmol/min/mg protein). A similar difference was also observed in the effects of the mutation of Pro110 on the PGDS activity between Δ1–24-l-pgds and its C89A/C186A mutant. The PGDS activity was not changed in the P110A mutant of Δ1–24-l-pgds (Table 2) but was lost completely in the P110A and P110S mutants of C89A/C186A l-pgds. The conserved disulfide bridge plays an important role in the stabilization of the tertiary structure of l-pgds (50), as recently confirmed by NMR analysis of the cysteine residue mutations of rat l-pgds (51). Therefore, both the chemical shift and the changes in the enzyme activity in the C89A/C186A mutants were probably caused by the secondary structural perturbation by introducing further mutation in the enzyme without the disulfide bridge.

The docking studies also indicate that the upper cavity of l-pgds as a reaction field can fully accommodate PGD2 with its bulk and extended structure of its 9α-hydroxyl-11-keto-cyclopentane group after cleavage of the 9α,11α-endoperoxide of PGH2. This indication is consistent with the fact that l-pgds binds PGD2 with a high affinity (K of about 20 nm) and suggests that the release of the reaction product PGD2 from the catalytic pocket of l-pgds may be a rate-limiting step of the reaction, giving the low turnover rate of l-pgds as compared with that of H-PGDS.