crystal structure and possible catalytic mechanism of microsomal prostaglandin e synthase type 2...

doi:10.1016/j.jmb.2005.03.035 J. Mol. Biol. (2005) 348, 1163–1176

Crystal Structure and Possible Catalytic Mechanism ofMicrosomal Prostaglandin E Synthase Type 2(mPGES-2)

Taro Yamada1, Junichi Komoto1, Kikuko Watanabe2, Yoshihiro Ohmiya3

and Fusao Takusagawa1*

1Department of MolecularBiosciences, University ofKansas, 1200 Sunnyside AveLawrence, KS 66045-7534USA

2Division of Applied LifeScience, Graduate School ofIntegrated Science and ArtUniversity of East Asia, 2-1Ichinomiya-gakuenchoShimonoseki, Yamaguchi751-0807, Japan

3National Institute of AISTThe Special Division for HumanLife Technology, Cell DynamicsResearch Group, MidorigaokaIkeda, Osaka 563-8577, Japan

0022-2836/$ - see front matter q 2005 E

Abbreviations used: GSTase, gluthPGDS, hematopoietic prostaglandindomethacin; mPGES-2, microsomsynthase type 2; PG, prostaglandin;D synthase; PGES, prostaglandin Eprostaglandin F synthase; PGFS-1, PE-mail address of the correspond

[email protected]

Prostaglandin (PG) H2 (PGH2), formed from arachidonic acid, is anunstable intermediate and is converted efficiently into more stablearachidonate metabolites (PGD2, PGE2, and PGF2) by the action of threegroups of enzymes. Prostaglandin E synthase catalyzes an isomerizationreaction, PGH2 to PGE2. Microsomal prostaglandin E synthase type-2(mPGES-2) has been crystallized with an anti-inflammatory drug indo-methacin (IMN), and the complex structure has been determined at 2.6 Aresolution. mPGES-2 forms a dimer and is attached to lipid membrane byanchoring the N-terminal section. Two hydrophobic pockets connected toform a V shape are located in the bottom of a large cavity. IMN bindsdeeply in the cavity by placing the OMe-indole and chlorophenyl moietiesinto the V-shaped pockets, respectively, and the carboxyl group interactswith Sg of C110 by forming a H-bond. A characteristic H-bond chainformation (N–H/Sg–H/Sg/H–N) is seen through Y107–C113–C110–F112, which apparently decreases the pKa of Sg of C110. The geometrysuggests that the Sg of C110 is most likely the catalytic site of mPGES-2.A search of the RCSB Protein Data Bank suggests that IMN can fit into thePGH2 binding site in various proteins. On the basis of the crystal structureand mutation data, a PGH2-bound model structure was built. PGH2 fitswell into the IMN binding site by placing the a and u-chains in the V-shaped pockets, and the endoperoxide moiety interacts with Sg of C110. Apossible catalytic mechanism is proposed on the basis of the crystal andmodel structures, and an alternative catalytic mechanism is described. Thefold of mPGES-2 is quite similar to those of GSH-dependent hematopoieticprostaglandin D synthase, except for the two large loop sections.

q 2005 Elsevier Ltd. All rights reserved.

Keywords: prostaglandin E synthase; mPGES-2; prostaglandin E2; indo-methacin; catalytic mechanism
*Corresponding author
Introduction

Prostaglandins (PGs) have numerous and diversebiological effects on a variety of physiological andpathological events, such as contraction of smoothmuscle, inflammation, and blood clotting. In

lsevier Ltd. All rights reserve

athione-S-transferase;in D synthase; IMN,al prostaglandin EPGDS, prostaglandinsynthase; PGFS,GFS type-1.ing author:

humans, the most important PG precursor isarachidonic acid, a C20 polyunsaturated fatty acidwith non-conjugated double bonds. PGs syn-thesized from arachidonic acid have the subscript2 (the series 2 PGs), such as PGD2, PGE2, PGF2, andPGH2, which is an unstable intermediate formedfrom arachidonic acid by the action of PGHsynthase in the arachidonate cascade. In mam-malian systems, PGH2 is converted efficiently intomore stable arachidonate metabolites, PGD2, PGE2,and PGF2, by the action of the respective synthasesfor these products.1 The crystal structures of PGHsynthase (COX-1 and COX-2), PGD synthase(hPGDS), and PGF synthase (PGFS-1) have beendetermined.2–4 In particular, the structure and

d.

1164 Microsomal Prostaglandin E Synthase-2

function of COX-1 and COX-2 have been studiedintensively, and more than 30 different coordinatesets have been deposited in the RCSB Protein DataBank (PDB).5

PGE2 was first discovered in sheep seminalvesicles. PGE2 is distributed widely in variousorgans, and exerts control over various biologicalactivities, such as relaxation/contraction of smoothmuscle,6 excretion of NaC,7 body temperature,8,9

and the physiological sleep/wake cycle.10 Thebiosynthesis of PGE2 requires three sequentialenzymatic steps: the release of arachidonic acidfrom membrane phospholipids by phospholipaseA2 (PLA2),11,12 conversion of arachidonic acid intoPGH2 by COX-1 or COX-2,13 and transformation ofPGH2 into PGE2 by PGES. Following inflammatorystimuli, PGE2 biosynthesis occurs in kineticallydistinct phases: an immediate phase, occurring inseconds to minutes; and a delayed phase, occurringover hours.14 Stimuli that increase the concentrationof cytoplasmic Ca2C rapidly can elicit the immedi-ate increase in PGE2 synthesis via functionalcoupling of pre-existing PG-biosynthetic enzymes.The second delayed increase in PGE2 biosynthesisis accompanied by the coordinated induction ofCOX-2 and several inducible PLA2 enzymes. It isgenerally thought that inflammation-associatedPGE2 production is a result of this coupledinduction of PLA2 and COX-2.11,14,15

PGES (EC 5.3.99.3) catalyzes the isomerization ofPGH2 to PGE2. Over the last 30 years, severalgroups have purified this enzyme.16–19 At leastthree different types of mammalian PGESs havebeen isolated. Ogorochi et al. and Meyer et al.independently purified the enzyme from thecytosol of human brain18 and Ascaridia galli,19

respectively. This cytosolic enzyme requires gluta-thione (GSH), belongs to the GSH-S-transferase(GSTase) family, and is named cPGES. The enzymehas been expressed in Escherichia coli.20 The mem-brane-associated PGES (mPGES-1) was partiallypurified from microsomal fractions of bovine andsheep vesicular glands, and was shown to requireGSH.16,17,21 Jakobsson et al. expressed human GSH-specific, mPGES-1 in E. coli.22 cPGES and mPGES-1in many organs are GSH-dependent enzymes.

Watanabe et al. reported that the GSH-non-specific PGES activity was distributed widely inthe microsomal fractions of rat and sheep organs.23,24 The enzyme activity in heart, spleen, and uterinemicrosomes did not specifically require GSH for itscatalytic activity, although the catalytic rate isincreased two- to fourfold in the presence of DTT,dihydrolipoic acid, GSH or other thiol com-pounds.25 This GSH-non-specific, membrane-associated PGES is named mPGES-2. A smallamount of the N-terminal truncated enzyme wasfound in the microsomal fraction of bovine heart.The N-terminal sequence of the truncated enzymestarts from Glu88 of the protein derived from thehuman/monkey cDNA.24 The intact and the N-terminal truncated mPGES-2 show similar catalyticactivity. The intact and the N-terminally truncated

proteins expressed in E. coli have the same enzy-matic properties as the enzyme purified frombovine heart microsomes.25 The amino acidsequence of mPGES-2 contains 110C-x-x-C113,which is seen in the active sites of glutaredoxinand thioredoxin.26 The C110S mutation abolishesthe catalytic activity but the C113S mutated enzymeretains the catalytic activity, suggesting that C110 isinvolved in the catalytic reaction.26

Indomethacin (IMN) is a widely used non-steroidal anti-inflammatory drug and is generallyknown to exhibit its multiple biological functions byinhibiting COX.27 In spite of the therapeutic utility,however, the drug has significant adverse effects, acircumstance that limits its use. Of these, gastro-intestinal and renal toxicities are of majorconcern.28,29 Recent reports suggest that IMN andother non-steroidal anti-inflammatory drugs bindto prostaglandin F synthase (PGFS-1) and protectagainst the progression of gastrointestinaltumors.30–32 There is increasing evidence thatthese drugs may also protect against a variety ofother cancers, including prostate carcinoma and,most recently, leukemia.33–39

Here, we present the crystal structure ofN-terminal truncated mPGES-2 complexed withthe non-steroidal anti-inflammatory drug IMN. Thecrystal structure indicates that IMN inhibits bothPGH2 synthesis and PGE2 synthesis. On the basis ofthis structure and a PGH2-bound model structure,we propose a catalytic mechanism of the isomeriza-tion reaction from PGH2 to PGE2 by mPGES-2.

Results

Overall structure

The enzyme used in this study is a recombinantprotein whose N-terminal amino acid residues 1–87were truncated and an extended His-tag (36 aminoacid residues) was attached to residue 88. Thecrystal structure of N-terminally modified mPGES-2complexed with the non-steroidal anti-inflam-matory drug IMN is determined. The crystallo-graphic refinement parameters (Table 1), final(2FoKFc) and (FoKFc) maps and conformationalanalysis by PROCHECK40 indicate that the struc-ture of mPGES-2 has been determined successfully.A crystallographic asymmetric unit contains foursubunits related by a non-crystallographic 222symmetry. The two subunits interact strongly andform a dimer, while the dimer–dimer interaction isapparently weak. Chloride ions were locatedbetween the dimers. For simplicity, the followingdescription refers mainly to subunit A.

The structure of residues K36 to K1 (extendedHis-tag) and P8 to 99 were not determined due todisorder. The N-terminal truncated in PGES-2(residues 100 to 373) is composed of three domains(Figure 1). The N-terminal domain (residues 100–179) has a four b-strand open a/b structure, thecentral domain (residues 180–222) is composed of a

Table 1. Crystallographic statistics

A. Data collection and phasingData Native SeMet enzymeAnomalous sites of Se – Edge Peak RemoteWavelength (A) 0.99108 0.97943 0.97927 0.96441Resolution (A) 2.6 2.6 2.6 2.6No. total reflections 1,072,662 880,864 881,673 874,951No. unique reflections 47,060 40,750 40,636 40,903Completeness 99.2 97.0 97.0 97.1Rmerge

a(%) 4.8 7.4 8.1 6.8Phasing power – 1.993 1.738 1.226FOM – 0.5396 (before DM), 0.9474 (after DM)B. RefinementNo. independent protein non-H atoms 8860 (four independent subunits)No. independent non-solvent molecules 4 IMN, 4ACT, 4ClNo. independent solvent molecules 73Resolution (A) 20–2.6No. independent reflections 47,060No. reflections used in Rcryst

b 42,337No. reflections used in Rfree 4723Rcrystal

c (outer shell) 0.217 (0.301)Rfree (outer shell) 0.255 (0.321)rmsd from idealityBond distances (A) 0.009Bond angles (deg.) 1.4Torsion angles (deg.) 12.2Ramachandran plotMost favored region (%) 88.0Additionally allowed region (%) 12.0

The space group is C2; the unit cell dimensions are: aZ128.24 A, bZ122.83 A, cZ111.53 A, bZ110.68; the molecular mass of the subunitis 37,137 Da; there are 16 subunits in the unit cell; VM Z2.77 A3; and the solvent content is 55% (v/v).

a RmergeZP

h

Pi jIhiK hIhij=

Ph

Pi jIhij.

b RcrystZSjFoKFcj/SjFoj.c Outer shellZ 2.6–2.7 A resolution.

Microsomal Prostaglandin E Synthase-2 1165

large anti-parallel loop with a short helix at bothends, and the C-terminal domain (residues 223–373)is the largest and has an a-structure composed ofeight helices connected with loops. The IMNbinding site is located between the N and C-terminal domains. To our knowledge, mPGES-2has a unique main-chain fold.

Two subunits related by a non-crystallographic2-fold axis form a dimer. The central domains ofthe two subunits are involved mainly in thedimerization. Intact mPGES-2 is a dimericenzyme and attaches to the microsomal mem-brane. The direction of the non-crystallographic2-fold axis and the position of the N terminus(residue 100) suggest that the dimeric mPGES-2

sits on the lipid bilayer and the truncated N-terminal section apparently enters into the lipidbilayer. A secondary structure prediction by thePSA server41 suggests that residues 1–99 formfive a-helices. A hydropathic index plot indicatesthat two helices (residues 1–20, and 60–72) arequite hydrophobic in character, suggesting thatthese helices are deep in the membrane and theother three helices are located on the surface ofmembrane.

Catalytic site in the cavity

As shown in Figure 2(b), mPGES-2 has a largecavity in each subunit, where the IMN binding site

Figure 1. Topology diagram. Thecircles and triangles representa-helix and b-strand, respectively.The 6 and 7 indicate the upwardand downward direction of the b-strand, respectively. The aminoacid residues in the second-ary structure are: b1(101–106)-a1(111–122)-b2(126–132)-a2(138–140)-b3(150–155)-b4(158–162)-a3(165–178)-a4(182–188)-b5(190–196)-b6(200–206)-a5(214–221)-a6(224–240)-a7(243–250)-a8(254–267)-a9(272–296)-a10(303–318)-a11(332–342)-a12(348–355)-a13-(360–370).

Figure 2 (legend next page)


Figure 2. (a) A dimeric mPGES-2 sits on the lipid bilayer. Two subunits are shown by aquamarine and light-pink,respectively. The arrow indicates a non-crystallographic 2-fold axis. Parts of the truncatedN-terminal section (residues 1–87)and disorder section (residues 88-99) shown bywhite are built based on a secondary structure prediction and a hydropathicindex analysis, and are apparently associatedwith the lipid bilayer. (b)Aviewof the subunit showing the anti-inflammatorydrug indomethacin (IMN) (magenta bonds) and an acetate ion (yellow bonds) that bind in the large cavity. Every tenthresidue is shown from100 to370.TheN-terminal, the central, and theC-terminaldomains are colored light-pink, yellow, andcyan, respectively. (c) An (FoKFc) map showing the residual electron density peak of the bound IMN. The map wascalculated after 50 cycles of positional refinement of proteins and solvents. Thefinalmodel of IMN is superimposed, and thecontour is drawnat the 2.0s level. (d)An IMNbinding site view showing interactions between the bound IMNandmPGES-2. (e) Aview showingH-bond networks around Sg of C110, which is surrounded by four rings (Y107, F112, P111, and P149).The geometry suggests that the possible binding site of the endoperoxidemoiety or the C9 of PGH2 is limited to near the O1

binding site of the bound IMN. A characteristic H-bond chain (N–H/S–H/S/H–N) through Y107–C113–C110–F112apparently decreases the pKa value of –SH of C110. Possible H-bonds are indicated by thin lines.


is. There are two hydrophobic pockets connected toform a V shape in the bottom of the cavity(Figure 2(b)). I264 is wedged between the twodeep pockets (pocket 1 and pocket 2). The anti-inflammatory drug IMN binds in the cavity(Figure 2(c)), and the OMe-indole and chlorophenylmoieties enter deep into the hydrophobic pocket 1and pocket 2, respectively (Figure 2(d)). Thecarboxyl group participates in a H-bond with Sg

of the essential C110.26 The –SH group of C110participates in a characteristic H-bond chain,N–H/S–H/S/H–N seen through Y107/C113/C110/F112 (Figure 2(e)). An acetate ion is stackedon the indole ring of IMN as if it fills a spacebetween IMN and the protein.

Discussion

The fold of mPGES-2 is similar to those ofhPGDS and GSTase

The crystal structure of the functionally similarhematopoietic PGD synthase (hPGDS) has beendetermined.3,42,43 hPGDS belongs to the GSTasefamily, and the fold of hPGDS is quite similar tothose of known GSTases (a class,44 m class,45 pclass,46 and s class47). Although mPGES-2 does notbelong to the GSTase family, the fold of mPGES-2 issimilar to those of hPGDS and GSTase. By using theprogram SARF2,48 the structure of mPGES-2 wascompared with those of hPGDS (PDB code 1PD2)


and human GSTase (PDB code 19GS). The totalnumber of amino acid residues in hPGDS andGSTase are 199 and 207, respectively. In total, 159 Ca

(79%) of hPGDS and 141 Ca (68%) of GSTase aresuperimposable on the corresponding Ca positionsof mPGES-2 with rmsd values of 2.34 A and 2.30 A,respectively. The amounts of identical residues inthe superimposable section are 12% and 14% forhPGDS and GSTase, respectively. As shown inFigure 3(a), the main-chain of mPGES-2 is nearlysuperimposable on those of hPGDS and GSTase,except for the central domain region (residues178–215) and the helix-loop-helix region (residues249–301). The helix-loop-helix region (residues 249–301) of mPGES-2 participates in forming the IMNbinding site, whereas such a helix-loop-helix regionis not present in the hPGDS structure. Interestingly,the essential –SH groups of C110 and the bound co-factor GSH are located in nearly the same position

Figure 3. (a) A superimposed view of three structure(aquamarine). The bound IMN, GSH, and GSH analogue adiagram showing the relative positions of IMN (magenta) and

in a superimposed hPGDS-mPGES-2 structure(Figure 3(b)).

IMN can fit into the PGH2 (PGG2) binding site

A PDB search shows that five crystal structurescontaining IMN have been reported, with COX-1,49

COX-2,50 PGFS-1,51 phospholipase A2 (PLA2; PDBCode: 1TI0). Except for PLA2, PGH2 (or PGG2) iseither the substrate or product of these enzymes,suggesting that IMN can fit into the PGH2 bindingsite although IMN and PGH2 have quite differentstructures (Figure 4(a)). IMN molecules in thecomplex structures of COX-1 and COX-2 are mostlyburied in the proteins, and thus, IMN stronglyinhibits COX-1 (IC50 0.08 mM)50 and COX-2 (IC50

0.96 mM).50 In the structure of PGFS-1, an IMNmoiety binds deep in the relatively large active sitepointing the carbonyl O3 to the oxyanion hole. IMN

s mPGES-2 (magenta), hPGDS (yellow), and GSTasere included. (b) A magnified view of the superimposedC110 (light-pink) in mPGES-2, and GSH (cyan) in hPGDS.

Figure 4. (a) Chemical structures of IMN and PGH2. (b) A model structure of mPGES-2 complexed with PGH2

(magenta bonds) and a water molecule (w1). Possible H-bonds are indicated by thin lines. (c) A superimposed view ofthe bound IMN (aquamarine) and the modeled PGH2 (magenta).


inhibits PGFS-1 moderately with IC50 4.1 mM.52 Inthe mPGES-2 structure, although the chlorophenylmoiety of IMN has extensive interaction with theprotein, the other portions have less interactionwith the protein. The OMe-indole moiety stackswith an acetate ion, and the carboxylate group ishydrated. The structure of mPGES-2:IMN suggeststhat the protein–IMN interaction is much weaker

than those seen in the COX-1, COX-2, and PGFS-1complexes. Indeed, IMN inhibits mPGES-2 muchmore weakly (IC50 w1 mM).As shown in Table 2, the bound IMN molecule

has various conformations, indicating that IMN isable to have different conformations in order to fitinto the active site of several enzymes. There is norelation between the conformation and the

Table 2. Torsion angles of the backbone of IMN found in the complex structures of mPGES-2, PGFS-1, COX-2, and PLA2

Name C18–C17a C17–C7 C7–C8 C8–N N–C9 C9–C10

a

MPGES-2 52 K124 180 180 K157 67PGFS-1 67 K110 172 181 K141 50COX-2 8 K140 183 174 127 K17PLA2 K20 18 181 167 161 K14

The atom numbering is shown in Figure 4(a). The values of IMN in the COX-1 complex are not listed because of low resolution (4.5 A).a The torsion angles of C18–C17 and C9–C10 can have 1808 added because of a different selection of carboxyl oxygen or phenyl ring

carbon atoms.


inhibitory activity. The IMN molecules found inmPGES-2 and PGFS-1 have quite similar backboneconformations, while the IMN molecules found inCOX-2 and PLA2 have some similarities. Thecarbonyl group (C]O) of the IMN moleculefound in mPGES-2 and PGFS-1 are pointed upwardwith respect to the indole plane, whereas thosefound in COX-2 and PLA2 are oriented downward.

PGSF-1 is an NADPH-dependent aldo-ketoreductase, and catalyzes the PGD2/9a,11b-PGF2reduction reaction, and the PGH2/PGF2aisomerization–reduction reaction.53 AlthoughPGD2 and PGH2 are structurally analogous, thebinding sites for PGD2 and PGH2 in this enzyme arepredicted to be different.53 The crystal structures ofPGFS-1:PGD2

4 and PGFS-1:IMN51 have beenreported. Interestingly, there is little overlapbetween the bound PGD2 and IMN in the struc-tures, indicating that PGD2 and IMN bind to PGFS-1 quite differently. Since IMN tends to bind thePGH2 binding site, PGH2 might bind to the IMNbinding site found in the PGFS-1:IMN complexstructure.51

Putative PGH2 binding scheme

In the crystal structure, the –SH group of C110forms H-bonds with O1 of the bound IMN(S–H/O1) and with Sg of C113 (S/H–Sg). TheC110S mutation abolishes catalytic activity but theC113S mutated enzyme retains catalytic activity.26

This mutagenesis study and the crystal structure ofmPGES-2 indicate that the catalytic process doesnot require disulfide bond formation between C110and C113. C110 is most likely involved in thecatalytic reaction, whereas C113 participates inlowering the pKa of the –SH group of C110 byforming a H-bond because the C113S mutatedenzyme has nearly full catalytic activity.

The isomerization reaction from PGH2 to PGE2

requires at least two processes (i.e. O9–O11 bondcleavage by reduction and C9 oxidation). Therefore,it would be reasonable to assume that Sg of C110interacts with either the endoperoxide moiety orC9 of PGH2 when PGH2 binds to the cavity ofmPGES-2. As shown in Figure 2(e), the essential Sg

of C110 is located in the bottom of bowl-shapedenvironment surrounded by the rings of Y107, P111,F112, and P149, and participates in two H-bonds(Sg/H–N[F112] and Sg/H–Sg [C113]). Therefore,

the binding site of the endoperoxide moiety or theC9 of PGH2 would be limited to the open space ofthe bowl, which is occupied by the O1 of the boundIMN.

The following assumptions were used to build aPGH2 bound mPGES-2 model.

(1)
The active-site geometry of mPGES-2 is notaltered drastically upon PGH2 or IMN binding.
(2)
The –SH group of C110 is involved directly ineither the O9–O11 bond cleavage or the C9
oxidation.
(3) Either the O9–O11 moiety or the C9 of PGH2 is
located near the O1 site of the bound IMNmolecule.

(4)
The bound PGH2 molecule has good polar andnon-polar interactions with the protein.
Initially, two models were built. One (C9 fixedmodel) was built by placing the C9 of PGH2 on theO1 position of the bound IMN, and the other (O9

fixed model) was built by placing the O9 on the O1

position. The PGH2 model structure was rotated atC9 (or O9) and the conformation of a and u-chainswere varied in order to fit into the active site ofmPGES-2. However, rigidity of the cyclopentane-endoperoxide moiety excluded the C9 fixed modelbecause the chains had too many short contactswith the protein. On the other hand, in the O9 fixedmodel, the two chains fit relatively well into the twohydrophobic pockets connected to form a V shapein the bottom of cavity. A similar model was built byfixing O11 on O1 of bound IMN. The initial O9/O11

fixed model suggested six possible H-bondsbetween PGH2 and the protein. These are: thecarboxylate group of the a-chain forms a pair ofH-bonds with the amide groups of G268 and K269,the hydroxyl group of the u-chain forms twoH-bonds with Og of S247 and Oh of Y251, O9 orO11 forms a H-bond with Sg of C110, and C9 forms aC–H/O H-bond with a bound water molecule(w1). The position and conformation of PGH2 wererefined by using X-PLOR.54 On the basis of thefollowing conditions, 100 cycles of positionalrefinement by X-PLOR were carried out:

(1)
The bound acetate was removed from themodelstructure.
(2)
The atoms of the protein, water molecules, andchloride ion were fixed on the original X-raypositions.


(3)
The atomic positions of PGH2 were varied onlywith the constraint of cis-C6]C7, trans-C13]C14
and the correct chiralities of C8, C12, and C15.
(4) The weight for X-ray part (wa) was set to zero. (5) The six possible H-bonds described above were
restrained to 3.0 A.

Although the conformations of the protein werenot altered during the model building, the a andu-chain fit well into pocket 1 and pocket 2 ofmPGES-2, respectively (Figure 4(b)). The modeledPGH2 molecule has a relatively large number ofhydrophobic interactions, but does not have anyshort contact with the protein. The modeled PGH2

molecule overlaps the bound IMN significantly(Figure 4(c)). For example, the a and u-chainsapproximately overlap the OMe-indole and thechlorophenyl moieties of IMN, respectively. In themodel structure, the carboxyl group of the a-chaincan form a pair of H-bonds with the amido groupsof G268 and K269, while the hydroxyl group of theu-chain can form two H-bonds with Og of S247 andOh of Y251, and O9 and O11 interact with Sg of theessential C110. A water molecule (w1) is locatedbetween Oh of Y107 and O2 of IMN in the crystalstructure. In the model structure, this watermolecule is now able to participate in a H-bondchain between Oh of Y107 and C9 of PGH2 (i.e.[Y107]Oh–H/O/H–C9[PGH2]). There is a largespace above the H-bond chain, where various R–SHmolecules can bind and place their –SH groups inthe water binding site. Therefore, the –SH group ofR–SH could form a similar H-bond chain([Y107]Oh–H/S/H–C9[PGH2]).

It should be noted that there is no crystalstructure of any of the enzymes containing PGH2,because PGH2 is a relatively unstable compound.The PGH2 binding scheme in COX-1 and COX-2 hasbeen proposed on the basis of the crystal structurescontaining the substrate arachidonic acid andinhibitors.49,50,55 For example, COX-2 structurescontaining arachidonic acid and COX-2 inhibitorshave been determined by X-ray analysis.50,55 On thebasis of the crystal structures, the initial PGH2

binding scheme was deduced. However, in order tobe consistent with data from mutagenesis studies,the proposed PGH2 binding site is shifted signifi-cantly from the arachidonate binding site.55 Themodeled PGH2 molecule overlaps approximatelywith the bound IMN in the COX-2 structure, andthe cyclopentane-endoperoxide moiety, a-chain,and u-chain are located approximately in thebinding sites of the chlorophenyl, the carboxyl,and OMe-indole moieties of the bound IMN,respectively.

A possible catalytic mechanism of mPGES-2

mPGES-2 catalyzes the PGH2 to PGE2 isomeriza-tion reaction without the presence of an R–SHcompound. However, the catalytic rate is increasedtwo- to fourfold in the presence of an R–SHreagent.25 Double SH reagents (DTT and

dihydrolipoic acid) enhance the catalytic activitymore than single SH reagents (GSH and2-mercaptoethanol).25 Since double SH reagentstend to form an intramolecular S–H/S H-bondbetween the two SH groups, the pKa value of Sb islowered and a deprotonated thiolic anion (–SK) isreadily generated. Therefore, the enhancement ofthe catalytic rate by a double SH reagent might bedue to the low pKa of one of the –SH groups. Asdescribed above, mPGES-2 is active in the absenceof an R–SH reagent, but the catalytic activity isincreased by the presence of an R–SH reagent,suggesting that a water molecule and the –SHgroup of an R–SH bind the same site and participatein the same catalytic role.hPGDS catalyzes the isomerization of PGH2 to

PGD2 in the presence of GSH. hPGDS andmPGES-2have similar main-chain folding patterns, and in asuperimposed structure, the essential –SH groupsof the bound cofactor GSH in hPGDS and of theessential C110 in mPGES-2 were located in the samesite (Figure 3(b)). Therefore, it is reasonable toassume that hPGDS and mPGES-2 have similarcatalytic mechanisms. A catalytic mechanism ofhPGDS has been proposed on the basis of thehPGDS:GSH complex structure.3 In the proposedmechanism, the –SH group of the bound GSH isdeprotonated by forming a H-bond with a tyrosineresidue (Y8). The thiolic anion of the deprotonatedGSH attacks O11 of PGH2, breaking the O9–O11

bond, and forming a covalent O11–S bond betweenPGH2 and GSH.When a second GSH (deprotonatedGSH) abstracts the hydrogen atom attached to C11,the O11–S bond is broken, and the C11]O11 carbonylgroup is formed. A third GSH (protonated GSH)donates a proton to the hydroxyl anion (O9) tocomplete the isomerization.On the basis of the crystal structure of mPGES-2

and the PGH2-bound model structure, and inconsideration of the proposed catalytic mechanismof hPGDS, we have postulated a mechanism forthe mPGES-2 reaction. A characteristic H-bondchain N–H/S–H/S/H–N is seen throughY107/C113/C110/F112 in the catalytic site(Figure 2(e)). In this environment, Sg of C110 issurrounded by two positive character protons, andthus, the pKa of the S

g of C110 is decreased. The pKa

value would be decreased to w7.0 when thehydrogen atoms are brought close to the Sg bymolecular vibrations. When PGH2 binds to thecavity in mPGES-2 and the endoperoxide moiety islocated in the catalytic site as described in themodeling, the isomerization reaction is initiated bya proton transfer from Sg of C110 to O11 of PGH2.O11 becomes positive and the deprotonated Sg ofC110 is produced. The thiolic anion of thedeprotonated Sg of C110 attacks the O9 of PGH2,since the positively charged O11 withdrawselectrons from O9. The nucleophilic attack resultsin forming a covalent O9–S

g bond between PGH2

and C110 and breaking the O9–O11 bond.Awater molecule (w1) or –SH group of the R–SH

reagent located between Oh of Y107 and C9 of PGH2


forms a H-bond chain ([Y107]Oh–H/O/H–C9[PGH2] or [Y107]Oh–H/S/H–C9[PGH2]). Thewater or –SH group of R–SH is polarized byforming a H-bond with Y107 as seen inGSTases,56–58 and acts as a base to abstract thehydrogen atom attached to C9. The abstraction ofhydrogen from C9 is an energetically unfavorablereaction, but it occurs by coupling with theenergetically favorable O9–S

g bond breakage. Theisomerization is completed by breaking the O9–S

g

bond and forming the C9]O9 carbonyl group(Figure 5).

Figure 5. A proposed catalytic mechanism of mPGES-2 onmodel structure of mPGES-2:(PGH2CRSH/H2O). (a) PGH2 bpocket 1 and pocket 2, respectively. The H-bond chain, N–H/the pKa of S

g of C110. A molecular vibration enforces the H atoand the proton on Sg is transferred to O11 of PGH2. (b) The depto break the O9–O11 bond. (c) A R–SH (or a water) bound betwH-bond with Y107, and consequently, the –SH group of R–SHthiolic anion of R–SH (or hydroxyl anion) removes the hydrogand C110 is cleaved to form the C9]O9 carbonyl group in a coof mPGES-2 and reagent R–SH.

Alternative catalytic mechanism

In our modeling of PGH2 binding, we assumedthat the PGH2 molecule makes good polar and non-polar interactions with the protein, and conse-quently the C9 fixed model was excluded. However,if the two chains are allowed to be in the solventregion with little interaction with the protein, the C9

fixed model is possible. In this case, the deproto-nated Sg of C110 abstracts the hydrogen atomattached to C9, and the O9–O11 bond is cleaved byacid catalysis with a water or R–SH molecule

the basis of the crystal structure of mPGES-2:IMN and ainds to the active site by placing the a and u-chains intoS–H/S/H–N seen in Y107–C113–C110–F112 decreasesms close to Sg and thus, the pKa of S

g is decreased further,rotonated Sg of C110 attacks O9 to form a Sg–O9 bond andeen Oh of Y107 and C9 of PGH2 is polarized by forming a(or water) is deprotonated at neutral pH. The resulting

en atom attached to C9, and the O9–Sg bond between PGH2

ncerted manner. (d) Formation of PGE2, and regeneration

Figure 6. An alternative catalyticmechanism. A PGH2 moleculebinds in the cavity of mPGES-2 bylocating C9 in the C110 catalyticbowl, while the carboxyl group ofthe a-chain interacts with a posi-tively charged region where H241,H244, R292, and R296 are located.The deprotonated Sg of C110abstracts the hydrogen atomattached to C9, and the O9–O11

bond is cleaved by acid catalysisby a water molecule or R–SH.


(Figure 6). This alternative catalytic mechanism issupported by the following facts.

The catalytic efficiency (kcat/KM) of the PGH2 toPGE2 reaction by mPGES-2 is 6.5!104 sK1 MK1

(kcat Z1.82 sK1 and KMZ28 mM),24 which is amoderate rate, suggesting that tight binding ofPGH2 to the protein might not be required.Furthermore, PGH2 is a relatively unstable com-pound and is decomposed non-enzymatically toPGD2 and PGE2.

59 The endoperoxide moiety ofPGH2 is quite susceptible to acid catalysis thatopens the O9–O11 bond. Oxidation of C11 or C9 inPGH2 produces PGD2 or PGE2, respectively. There-fore, a minimum requirement of the enzyme is toprovide a C9 oxidation-favored geometry. There is apositively charged region on the surface of thecavity, where H241, H244, R292, and R296 arelocated. The carboxyl group of the a-chain ofPGH2 might interact with this positively chargedregion, while the cyclopentane-endoperoxidemoiety of PGH2 enters the C110 catalytic bowl,whose shape might prohibit a C11 approachbut allow a C9 oxidation-favored geometry(Figure 2(e)).

Materials and Methods

Purification and crystallization

The N-terminal truncated (residues 1–87) monkeymPGES-2 gene was cloned into the pTrc-HisA vectorand transformed in E. coli BL21,25 which were grown at37 8C in 1 l of LB medium containing 50 mg of ampicillinand 100 mg of Fe(NO3)3. IPTG was added to a finalconcentration of 1 mM after 1.5 hours of culturing, andincubation was continued for an additional 15 hours.Cells were harvested by centrifugation and suspended in60 ml of 50 mM Tris–HCl (pH 7.5), 0.5 mM EDTA. Celllysis was carried out by treatment with egg-whitelysozyme (1 mg/ml of the suspension at 0 8C for onehour), followed by freezing and thawing. The mixturewas subjected to brief sonication. The centrifuged

supernatant was treated with 300 g/l of ammoniumsulfate, and the precipitated protein was recovered bycentrifugation. Ammonium sulfate, EDTA, and Tris–HClwere removed by dialysis in buffer A (30 mM potassiumphosphate (pH 7.2), 0.2% (v/v) Tween-20). The proteinwas loaded onto a column of DE52 (2.4 cm!10 cm)equilibrated with buffer A. The enzyme was eluted by alinear gradient between 100 ml each of 30 mM and200 mM potassium phosphate (pH 7.2) containing 0.2%Tween-20. Fractions having a reddish color were pooled,and imidazole was added to a final concentration of10 mM. The solution was loaded onto an affinity column(1.0 cm!5.0 cm) containing nickel-chelating resin (Ni-CAMe HC Resin (Sigma)) equilibrated with 30 mMpotassium phosphate buffer (pH 7.2) containing 10 mMimidazole, 0.01% (w/v) n-dodecyl-beta-D-maltopyrano-side. The column was washed with the same buffer untilthe absorption at 280 nm of the eluate reached that of thewashing buffer. The enzyme was eluted by a lineargradient between 50 ml each of 10 mM and 200 mMimidazole in 30 mM potassium phosphate buffer (pH 7.2)containing 0.01% n-dodecyl-beta-D-maltopyranoside. Thereddish fractions were pooled and concentrated to20 mg/ml solution using an Amicon concentrator witha 30 kDa cut-off membrane. The purity of enzyme waschecked by SDS-PAGE, and the enzyme activity wasexamined as described.25

The overexpressed enzyme whose N-terminal residues1–87 were truncated and an “extended” His-tag(MGGSHHHHHHGMASMTGGQQMGRDLYDDDDKDRWGS) was attached to residue 88 was purified and usedfor crystallization. The enzyme was crystallized with theanti-inflammatory drug, IMN. The crystallization con-ditions were: 1.7 M ammonium sulfate, 100 mM sodiumacetate/HCl buffer (pH 5.5), 1.5 mM IMN, and 10 mg/mlof enzyme using the sitting-drop, vapor-diffusionmethod. Reddish plate crystals (w0.3 mm!0.2 mm!0.1 mm) were grown in three days at 22 8C.For the SeMet enzyme, the cloned pTrc-HisA

vector was transformed into the Met auxotrophic E. colistrain B834. The transformed bacteria were grown inM9 medium supplemented with 0.06 g/l of SeMet.The SeMet enzyme was purified and crystallizedusing the same procedure as that used for the nativeenzyme.


Data measurement

Diffraction data were collected from cryo-cooledcrystals at beamline 19BM at the Advanced PhotonSource in the Argonne National Laboratory. The crystals(w0.3 mm!0.2 mm!0.1 mm) of the native and SeMetenzymes were scooped with a nylon loop and dipped intoa cryoprotectant solution forw5 seconds before theywerefrozen in liquid nitrogen. The cryoprotectant solution wascomposed of the original mother liquor containing 25%(v/v) ethylene glycol. The crystals diffracted up to 2.6 A.Multiwavelength anomalous diffraction (MAD) datawere collected from a single crystal at 100 K to 2.6 Aresolution, and four data sets of the native enzyme werecollected at 2.6 A resolution. The data were processedwith the program DENZO/SCALEPACK.60 The fournative data sets were merged to one data set. Datastatistics are given in Table 1.

Structure determination

The space group (C2) and the unit cell dimensions (aZ129.0 A, bZ122.7 A, cZ112.2 A, and bZ110.58) indicatethat four independent molecules are in an asymmetricunit and the water content is 55% (v/v) (VM 2.77 A3). Thestructure determination was carried out by the MADmethod using 2.6 A resolution data. The 24 Se positionswere determined by using Shake-and-Bakemethod.61 TheSe positions were refined by using CNS,62 and the initialMAD map was obtained. Residues 100–373 were built inthe MAD map using XstalView.63 The model was refinedby CNS. The (FoKFc) maps did not give any significantelectron density for the amino acid residues K36 to K1(extended His-tag) and 88–99 (N-terminal section),indicating that these residues were heavily disordered.The (FoKFc) maps showed two large significant residualelectron density peaks in region of the active site. SincemPGES-2 was crystallized in the presence of an excess ofIMN (1.5 mM) and acetate (100 mM), IMN and acetatemolecules were fit into the electron density peaks(Figure 2(c)). Another large residual electron densitypeak was seen between the two subunits. This peak issurrounded by R137, R146, and Y287 of the other subunit.Judging from the environment and peak height, a chlorideion was assigned to this peak because mPGES-2 wascrystallized in the presence of acetate/HCl (100 mM).Other well-defined residual electron density peaks indifference maps were assigned to water molecules ifpeaks were able to bind to the protein molecules withhydrogen bonds. Although the protein was purified fromE. coli grown in LB medium containing 0.4 mM Fe(NO3)3and had a reddish color, the (FoKFc) maps did not showany significant residual electron density peak for Fe,suggesting that Fe ions bound to the disordered extendedHis-tag section. During the refinement, the four subunitsrelated by a non-crystallographic 222 symmetry wererestrained tightly to have the same structure in order toincrease the accuracy of coordinates. The structure wasrefined with all reflections (no s-cut off) from 20 A to2.6 A resolution.

Protein Data Bank accession code

The atomic coordinates and structure factors have beendeposited with the Brookhaven Protein Data Bank (entryname: 1Z9H).

Acknowledgements

We express our thanks to the 19BM beamline staffat APS for assistance to Dr Hiroyuki Kojina formeasurment of the enzyme activity and to ProfessorRichard H. Himes for a critical reading of themanuscript and very valuable comments. The workhas been supported by grant GM37233 (to F.T.) fromthe National Institutes of Health. Use of theArgonne National Laboratory Structural BiologyCenter beamline at the Advanced Photon Sourcewas supported by the US Department of Energy,Office of Energy Research, under contract no. W-31-109-ENG-38.

References

1. Johnson, M., Carey, F. & McMillan, R. M. (1983).Alternative pathways of arachidonate metabolism:prostaglandins, thromboxane and leukotrienes.Essays Biochem. 19, 40–141.

2. Picot, D., Loll, P. J. & Garavito, R. M. (1994). The X-raycrystal structure of the membrane protein prosta-glandin H2 synthase-1. Nature, 367, 243–249.

3. Kanaoka, Y., Ago, H., Inagaki, E., Nanayama, T.,Miyano, M., Kikuno, R. et al. (1997). Cloning andcrystal structure of hematopoietic prostaglandin Dsynthase. Cell, 90, 1085–1095.

4. Komoto, J., Yamada, T., Watanabe, K. & Takusagawa,F. (2004). Crystal structure of human prostaglandin Fsynthase (AKR1C3). Biochemistry, 43, 2188–2198.

5. Garavito, R. M. & Mulichak, A. M. (2003). Thestructure of mammalian cyclooygenases. Annu. Rev.Biophys. Biomol. Struct. 32, 183–206.

6. Smith, W. L., Marnett, L. J. & DeWitt, D. L. (1991).Prostaglandin and thromboxane biosynthesis.Pharmacol. Ther. 49, 153–179.

7. Vander, A. J. (1968). Direct effects of prostaglandin onrenal function and renin release in anesthetized dog.Am. J. Physiol. 214, 218–221.

8. Milton, A. S. & Wendlandt, S. (1971). Effects on bodytemperature of prostaglandins of the A, E, and Fseries on injection into the third ventricle ofunanaesthtized cats and rabbits. J. Physiol. 218,325–336.

9. Matsumura, H., Goh, Y., Ueno, R., Sakai, T. &Hayaishi, O. (1988). Awaking effects of PGE2 micro-injected into the preoptic area of rats. Brain Res. 444,265–272.

10. Hayaishi, O. (1991). Molecular mechanisms of sleep–wake regulation: roles of prostaglandins D2 and E2.FASEB J. 5, 2575–2581.

11. Bonventre, J. V., Huang, Z., Taheri, M. R., O’Leary, E.,Li, E., Moskowitz, M. A. & Sapirstein, A. (1997).Reduced fertility and postischaemic brain injury inmice deficient in cytosolic phospholipase A2. Nature,390, 622–625.

12. Murakami, M., Kambe, T., Shimbara, S., Higashino,K., Hanasaki, K., Arita, H. et al. (1999). Differentfunctional aspects of the group II subfamily (TypesIIA and V) and type X secretory phospholipase A(2)sin regulating arachidonic acid release and prosta-glandin generation. Implications of cyclooxygenase-2induction and phospholipid scramblase-mediatedcellular membrane perturbation. J. Biol. Chem. 274,31435–31444.


13. Smith, W. L. & Langenbach, R. (2001). Why there aretwo cyclooxygenase isozymes. J. Clin. Invest. 107,1491–1495.

14. Murakami, M., Nakatani, Y., Tanioka, T. & Kudo, I.(2002). Prostaglandin E synthase. Prostaglandins LipidMediat. 68–69, 383–399.

15. Scott, K. F., Bryant, K. J. & Bidgood, M. J. (1999).Functional coupling and differential regulation of thephospholipase A2-cyclooxygenase pathways ininflammation. J. Leukoc. Biol. 66, 535–541.

16. Ogino, N., Miyamoto, T., Yamamoto, S. &Hayaishi, O.(1977). Prostaglandin endoperoxide E isomerase frombovine vesicular gland microsomes, a glutathione-requiring enzyme. J. Biol. Chem. 252, 890–895.

17. Moonen, P., Buytenhek, M. & Nugteren, D. H. (1982).Purification of PGH-PGE isomerase from sheepvesicular glands. Methods Enzymol. 86, 84–91.

18. Ogorochi, T., Ujihara, M. & Narumiya, S. (1987).Purification and properties of prostaglandin H–Eisomerase from the cytosol of human brain: identifi-cation as anionic forms of glutathione S-transferase.J. Neurochem. 48, 900–909.

19. Meyer, D. J., Muimo, R., Thomas, M., Coates, D. &Isaac, R. E. (1996). Purification and characterization ofprostaglandin-H E-isomerase, a sigma-class gluta-thione S-transferase, from Ascaridia galli. Biochem. J.313, 223–227.

20. Tanioka, T., Nakatani, Y., Semmyo, N., Murakami, M.& Kudo, I. (2000). Molecular identification of cytosolicprostaglandin E2 synthase that is functionally coupledwith cyclooxygenase-1 in immediate prostaglandin E2

biosynthesis. J. Biol. Chem. 275, 32775–32782.21. Tanaka, Y., Ward, S. L. & Smith, W. L. (1987).

Immunochemical and kinetic evidence for twodifferent prostaglandin H–prostaglandin Eisomerases in sheep vesicular gland microsomes.J. Biol. Chem. 262, 1374–1381.

22. Jakobsson, P. J., Thoren, S., Morgenstern, R. &Samelsson, B. (1999). Identification of human prosta-glandin E synthase: a microsomal glutathion-depen-dent, inducible enzyme, constituting a potential noveldrug target. Proc. Natl Acad. Sci. USA, 96, 7220–7225.

23. Watanabe, K., Kurihara, K. &Hayaishi, O. (1997). Twotypes of microsomal prostaglandin E synthase:glutathione-dependent and independent prosta-glandin E synthases. Biochem. Biophys. Res. Commun.235, 148–152.

24. Watanabe, K., Kurihara, K. & Suzuki, T. (1999).Purification and characterization of membrane-bound prostaglandin E synthase from bovine heart.Biochim. Biophys. Acta, 1439, 406–414.

25. Tanikawa, N., Ohmiya, Y., Ohkubo, H., Hashimoto,K., Kangawa, K., Kojima, M. et al. (2002). Identifi-cation and characterization of a novel type ofmembrane-associated prostaglandin E synthase.Biochem. Biophys. Res. Commun. 291, 884–889.

26. Watanabe, K., Ohkubo, H., Niwa, H., Tanikawa, N.,Koda, N., Ito, S. & Ohmiya, Y. (2003). Essential 110Cysin active site of membrane-associated prostaglandin Esynthase-2. Biochem. Biophys. Res. Commun. 306,577–581.

27. Vane, J. R. (1971). Inhibition of prostaglandin syn-thesis as a mechanism of action for aspirin-like drugs.Nature New Biol. 231, 232.

28. Gabriel, S. E., Jaakkimainen, L. & Bombardier, C.(1999). Risk for serious gastrointestinal complicationsrelated to use of nonsteroidal anti-inflammatorydrugs: a meta-analysis. Ann. Intern. Med. 115, 787–796.

29. Epstein, M. (2002). Non-steroidal anti-inflammatorydrugs and the continuum of renal dysfunction.J. Hypertens. 6, S17–S23.

30. Xu, X. C. (2002). COX-2 inhibitors in cancer treatmentand prevention, a recent development. AnticancerDrugs, 13, 127–137.

31. Gupta, R. A. & Dubois, R. N. (2001). Colorectal cancerprevention and treatment by inhibition of cyclo-oxygenase-2. Nature Rev. Cancer, 1, 11–21.

32. Grosch, S., Tegeder, I., Niederberger, E., Brautigam, L.& Geisslinger, G. (2001). COX-2 independent induc-tion of cell cycle arrest and apoptosis in colon cancercells by the selective COX-2 inhibitor celecoxib.FASEB J. 15, 2742–2744.

33. Zhang, X., Morham, S. G., Langenbach, R. & Young,D. A. (1999). Malignant transformation and antineo-plastic actions of nonsteroidal antiinflammatorydrugs (NSAIDs) on cyclooxygenase-null embryofibroblasts. J. Expt. Med. 190, 451–459.

34. Wechter, W. J., Kantoci, D., Murray, E. D., Jr, Quiggle,D. D., Leipold, D. D., Gibson, K. M. & McCracken,J. D. (1997). R-flurbiprofen chemoprevention andtreatment of intestinal adenomas in the APC(Min)/C mouse model: implications for prophylaxis andtreatment of colon cancer. Cancer Res. 57, 4316–4324.

35. Wechter, W. J., Leipold, D. D., Murray, E. D., Jr,Quiggle, D., McCracken, J. D., Barrios, R. S. &Greenberg, N. M. (2000). E-7869 (R-flurbiprofen)inhibits progression of prostate cancer in theTRAMP mouse. Cancer Res. 60, 2203–2208.

36. Kasum, C. M., Blair, C. K., Folsom, A. R. & Ross, J. A.(2003). Non-steroidal anti-inflammatory drug use andrisk of adult leukemia. Cancer Epidemiol. BiomarkersPrev. 12, 534–537.

37. Mills, K. I., Gilkes, A. F., Sweeney, M., Choudhry,M. A., Woodgate, L. J., Bunce, C. M. et al. (1998).Identification of a retinoic acid responsive aldoketo-reductase expressed in HL60 leukaemic cells. FEBSLetters, 440, 158–162.

38. Bunce, C. M., Mountford, J. C., French, P. J., Mole,D. J., Durham, J., Michell, R. H. & Brown, G. (1996).Potentiation of myeloid differentiation by anti-inflam-matory agents, by steroids and by retinoic acidinvolves a single intracellular target, probably anenzyme of the aldoketoreductase family. Biochim.Biophys. Acta, 1311, 189–198.

39. Desmond, J. C., Mountford, J. C., Drayson, M. T.,Walker, E. A., Hewison, M., Ride, J. P. et al. (2003). Thealdo-keto reductase AKR1C3 is a novel suppressor ofcell differentiation that provides a plausible target forthe non-cyclooxygenase-dependent antineoplasticactions of nonsteroidal anti-inflammatory drugs.Cancer Res. 63, 505–512.

40. Laskowski, R. A., MacArthur, M. W., Moss, D. S. &Thornton, J. M. (1993). ROCHECK: a program tocheck the stereochemical quality of protein structures.J. Appl. Crystallog. 26, 283–291.

41. Stultz, C. M., White, J. V. & Smith, T. F. (1993).Structural analysis based on state-space modeling.Protein Sci. 2, 305–314.

42. Inoue, T., Irikura, D., Okazaki, N., Kinugasa, S.,Matsumura, H., Uodome, N. et al. (2003). Mechanismof metal activation of human hematopoietic prosta-glandin D synthase. Nature Struct. Biol. 10, 291–296.

43. Inoue, T., Okano, Y., Kado, Y., Aritake, K., Irikura, D.,Uodome, N. et al. (2004). First determination of theinhibitor complex structure of human hematopoieticprostaglandin D synthase. J. Biochem. (Tokyo), 135,279–283.


44. Sinning, I., Kleywegt, G. J., Cowan, S. W., Reinemer,P., Dirr, H. W., Huber, R. et al. (1993). Structuredetermination and refinement of human a classglutathione transferase A1-1, and a comparison withthe m and p class enzymes. J. Mol. Biol. 232, 192–212.

45. Raghunathan, S., Chandross, R. J., Kretsinger, R. H.,Allison, T. J., Penington, C. J. & Rule, G. S. (1994).Crystal structure of human class m glutathionetransferase GSTM2-2. Effects of lattice packing onconformational heterogeneity. J. Mol. Biol. 238,815–832.

46. Reinemer, P., Dirr, H. W., Ladenstein, R., Schaffer, J.,Gallay, O. & Huber, R. (1991). The three-dimensionalstructure of class p glutathione S-transferase incomplex with glutathione sulfonate at 2.3 Aresolution. EMBO J. 10, 1997–2005.

47. Ji, X., von Rosenvinge, E. C., Johnson, W.W., Tomarev,S. I., Piatigorsky, J., Armstrong, R. N. & Gilliland, G. L.(1995). Three-dimensional structure, catalytic proper-ties, and evolution of a s class glutathione transferasefrom squid, a progenitor of the lens S-crystallins ofcephalopods. Biochemistry, 34, 5317–5328.

48. Alexandrov, N. N. (1996). SARFing the PDB. ProteinEng. 9, 727–732.

49. Loll, P. J., Picot, D., Ekabo, O. & Garavito, R. M. (1996).Synthesis and use of iodinated nonsteroidal anti-inflammatory drug analogs as crystallographicprobes of the prostaglandin H2 synthase cyclo-oxygenase active site. Biochemistry, 35, 7330–7340.

50. Kurumbail, R. G., Stevens, A. M., Gierse, J. K.,McDonald, J. J., Stegeman, R. A., Pak, J. Y. et al.(1996). Structural basis for selective inhibition ofcyclooxygenase-2 by anti-inflammatory agents.Nature, 384, 644–648.

51. Lovering, A. L., Ride, J. P., Bunce, C. M., Desmond,J. C., Cummings, S. M. & White, S. A. (2004). Crystalstructures of prostaglandin D2 11-ketoreductase(AKR1C3) in complex with the nonsteroidal anti-inflammatory drugs flufenamic acid and indo-methacin. Cancer Res. 64, 1802–1810.

52. Matsuura, K., Shiraishi, H., Hara, A., Sato, K.,Deyashiki, Y., Ninomiya, M. & Sakai, S. (1998).Identification of a principle mRNA species forhuman 3a-hydroxysteroid dehydrogenase isoform(AKR1C3) that exhibits high prostaglandin D2 11-ketoreductase activity. J. Biochem. 124, 940–946.

53. Watanabe, K., Yoshida, R., Shimizu, T. & Hayaishi, O.(1985). Enzymatic formation of prostaglandin F2afrom prostaglandin H2 and D2. Purification andproperties of prostaglandin F synthetase from bovinelung. J. Biol. Chem. 260, 7035–7041.

54. Brunger, A. T. (1993). X-PLOR 3.82: A System for X-rayCrystallography and NMR, Yale University Press,London.

55. Kiefer, J. R., Pawlitz, J. L., Moreland, K. T., Stegeman,R. A., Hood, W. F., Gierse, J. K. et al. (2000). Structuralinsights into the stereochemistry of the cyclo-oxygenase reaction. Nature, 405, 97–101.

56. Bjornestedt, R., Stenberg, G., Widersten, M., Board,P. G., Sinning, I., Jones, T. A. & Mannervik, B. (1995).Functional significance of arginine 15 in the active siteof human class alpha glutathione transferase A1-1.J. Mol. Biol. 247, 765–773.

57. Ji, X., Johnson, W.W., Sesay, M. A., Dickert, L., Prasad,S. M., Ammon, H. L. et al. (1994). Structure andfunction of the xenobiotic substrate binding site of aglutathione S-transferase as revealed by X-ray crystal-lographic analysis of product complexes with thediastereomers of 9-(S-glutathionyl)-10-hydroxy-9,10-dihydrophenanthrene. Biochemistry, 33, 1043–1052.

58. Wilce, M. C., Board, P. G., Feil, S. C. & Parker, M. W.(1995). Crystal structure of a theta-class glutathionetransferase. EMBO J. 14, 2133–2143.

59. Andersen, N. H. & Hartzell, C. J. (1984). High-field 1HNMR studies of prostaglandin H2 and its decompo-sition pathways. Biochem. Biophys. Res. Commun. 120,512–519.

60. Otwinowski, Z. & Minor, W. (1997). Processing ofX-ray diffraction data collected in oscillation mode.Methods Enzymol. 276, 307–326.

61. Miller, R., Gallo, S. M., Khalak, H. G. & Weeks, C. M.(1994). SnB: crystal structure determination via Shake-and-Bake. J. Appl. Crystallog. 27, 613–621.

62. Brunger, A. T., Adams, P. D., Clore, G. M., DeLano,W. L., Gros, P., Grosse-Kunstleve, R. W. et al. (1998).Crystallography &NMR system: a new software suitefor macromolecular structure determination. ActaCrystallog. sect. D, 54, 905–921.

63. McRee D. T. (1999). Practical Protein Crystallography.2nd edit., Academic Press, San Diego.

Edited by M. Guss

(Received 29 December 2004; received in revised form 9 March 2005; accepted 14 March 2005)

crystal structure and possible catalytic mechanism of microsomal prostaglandin e synthase type 2...

Documents