Cellular Prostaglandin E2 Production by Membrane-bound Prostaglandin E Synthase-2 via Both Cyclooxygenases-1 and -2*

Current evidence suggests that two forms of prostaglandin (PG) E synthase (PGES), cytosolic PGES and membrane-bound PGES (mPGES) -1, preferentially lie downstream of cyclooxygenase (COX) -1 and -2, respectively, in the PGE2 biosynthetic pathway. In this study, we examined the expression and functional aspects of the third PGES enzyme, mPGES-2, in mammalian cells and tissues. mPGES-2 was synthesized as a Golgi membrane-associated protein, and spontaneous cleavage of the N-terminal hydrophobic domain led to the formation of a truncated mature protein that was distributed in the cytosol with a trend to be enriched in the perinuclear region. In several cell lines, mPGES-2 promoted PGE2 production via both COX-1 and COX-2 in the immediate and delayed responses with modest COX-2 preference. In contrast to the marked inducibility of mPGES-1, mPGES-2 was constitutively expressed in various cells and tissues and was not increased appreciably during tissue inflammation or damage. Interestingly, a considerable elevation of mPGES-2 expression was observed in human colorectal cancer. Collectively, mPGES-2 is a unique PGES that can be coupled with both COXs and may play a role in the production of the PGE2 involved in both tissue homeostasis and disease.

duction elicited by agonists promptly mobilizing intracellular Ca 2ϩ , a situation in which a burst release of AA occurs (1)(2)(3)(4)(5). The inducible COX-2 is essential for delayed PG generation induced by proinflammatory stimuli, during which AA is gradually supplied over long periods, and also promotes immediate PG production if it already exists in cells primed by particular stimuli (1)(2)(3)(4)(5). Current studies employing isozyme-specific inhibitors and knockout mice have revealed that the two COXs play distinct roles in vivo (6 -10), and segregated utilization of these enzymes at the cellular level has been explained not only by their distinct expression profiles but also by subtle differences in their AA requirement, hydroperoxide sensitivity, and subcellular localization (4,(11)(12)(13). In addition, selective coupling with various terminal PG synthases has also been shown to influence crucially the utilization of the two COX isoforms during the different phases of cell activation (14 -16).
Recently, a third PGES, which shows broad specificity of thiol requirement, was purified from the microsomal fraction of bovine heart (27), and the whole sequences of the human and monkey enzymes were determined (28). This enzyme, designated mPGES-2, has an N-terminal hydrophobic domain followed by a glutaredoxin or thioredoxin homology region. The N-terminal 22 amino acid sequence of the purified bovine enzyme was identical to residues 88 -109 of the full-length * This work was supported by grants-in-aid for scientific research from the Ministry of Education, Science, Culture, Sports and Technology of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
In this paper, we report the cellular PGE 2 biosynthetic function and tissue expression of mPGES-2 in comparison with those of mPGES-1. We provide evidence that mPGES-2 undergoes N-terminal cleavage during protein maturation in the Golgi membrane, is coupled with both COX-1 and COX-2 leading to PGE 2 production, and is expressed relatively constitutively, rather than inducibly, in various cells and tissues. Moreover, immunohistochemical analyses of mPGES-1 and mPGES-2 in human tissues with various diseases allowed us to speculate on their general roles in human pathophysiology.
Preparation of Antibody against mPGES-2-cDNA for N-terminally truncated mPGES-2 (mPGES-2-del-N, which lacked the first 87 amino acids) subcloned into pTrc-His A (28) was transformed into the competent cell BL21-D3 (Stratagene). After culture for an appropriate period with 0.5 mM isopropyl-␤-D-(Ϫ)-thiogalactopyranoside, cells were spun down, freeze-thawed, and suspended in phosphate-buffered saline (PBS) containing 3 g/ml leupeptin, 3 g/ml antipain, 1 mM phenylmethylsulfonyl fluoride, and 1 mM dithiothreitol. After sonication and centrifugation for 10 min at 10,000 ϫ g, the resulting supernatant was dialyzed against 20 mM Tris-HCl (pH 7.4) containing 150 mM NaCl (TBS) and 1 mM EDTA overnight. Then the dialyzed sample was applied to a nickel-nitrilotriacetic acid-agarose column (Novagen), and the bound proteins were eluted with 40 -80 mM imidazole at a flow rate of 10 ml/h. Fractions containing pure His 6 -tagged mPGES-2 protein were collected and dialyzed against PBS. The purified protein gave a single band with a predicted size on SDS-PAGE followed by staining with Coomassie Brilliant Blue.
New Zealand White rabbits (male, 1 kg; Saitama Animal Center) were immunized subcutaneously with the purified mPGES-2-del-N protein (0.3 mg each) mixed with Freund's complete adjuvant (Difco). After several booster immunizations with Freund's incomplete adjuvant (Difco) at 2-week intervals, blood was collected, and the serum titer was assayed by enzyme-linked immunosorbent assay and Western blotting with recombinant mPGES-2-del-N protein as purified above. The specific binding of the antibody to mPGES-2 was verified by immunoblotting using mPGES-2-transfected mammalian cell lines (see below) in comparison with parental cells.
Transfection Studies-Transfection of cDNAs into HEK293 cells was performed by lipofection as described previously (4,14). Briefly, 1 g of plasmid (the N-terminally FLAG-tagged full-length mPGES-2 (mPGES-2-FL) in pCMV-Tag2 (Stratagene), FLAG-tagged, N-termi-nally truncated mPGES-2 (mPGES-2-del-N) in pCDNA3.1/hyg (ϩ), and COX-1 or -2 in pCDNA3.1/neo (ϩ)) was mixed with 2 l of Lipofect-AMINE 2000 in 100 l of Opti-MEM medium for 30 min and then added to cells that had attained 40 -60% confluence in 12-well plates (Iwaki Glass) containing 0.5 ml of Opti-MEM. After incubation for 6 h, the medium was replaced with 1 ml of fresh culture medium. After overnight culture, the medium was replaced with 1 ml of fresh medium, and culture was continued at 37°C in an incubator flushed with 5% CO 2 in humidified air. To obtain stable transfectants, the cells were cloned by limiting dilution in 96-well plates in culture medium containing appropriate antibiotics (5 g/ml hygromycin or 1 mg/ml geneticin). After culture for 3-4 weeks, wells containing a single colony were chosen, and the expression of each protein was monitored by RNA blotting and/or Western blotting. The established clones were expanded and used for the experiments described below. To establish BEAS-2B cells stably expressing COX-1 or COX-2, their cDNAs in pCDNA3.1/neo (ϩ) were transfected with LipofectAMINE 2000, and the geneticin-resistant clones were selected in a similar way.
C-terminally FLAG-tagged mPGES-1 and N-terminally FLAGtagged mPGES-2-del-N cDNAs were transfected into BEAS-2B and 3Y1 cells with the ViraPower lentiviral expression system according to the manufacturer's instructions. Briefly, the FLAG-tagged mPGES-1 or mPGES-2-del-N cDNA insert was amplified by PCR with the Advantage cDNA polymerase mix (Clontech) and was subcloned into the pLenti6/V4-D-TOPO vector (Invitrogen). The resulting plasmid was transfected into 293FT cells (Invitrogen) with LipofectAMINE 2000, and an aliquot of the supernatant harvested 3 days after transfection was then added to BEAS-2B and 3Y1 cells. The cells were cultured in the presence of 30 g/ml blastcidin, and the surviving cells that expressed appropriate levels of mPGES-1 or mPGES-2-del-N proteins were used in subsequent studies.
Construction of mPGES-2 Mutant-To construct the catalytically inactive mutant mPGES-2-CS, in which cysteine residues in the thioredoxin/glutaredoxin motif have been replaced by serine residues (31), mismatched PCR was conducted with the Advantage cDNA polymerase mix, as reported previously (14). The primers used for construction of the mutant were as follows. FLAG-del-N primer, 5Ј-CACCATGGACT-ACAAGGACGACGATGACAAGGAGCGCTCAGCAGTGCAGCTC-3Ј (FLAG sequence underlined), m2-3Ј primer, 5Ј-TCAGTGCGCTGGGG-AGGCCTCG-3Ј; CS-S primer, 5Ј-TACAAGACGAGTCCCTTCAGCAGC-AAGG-3Ј; and CS-AS primer, 5Ј-CCTTGCTGCTGAAGGGACTCGTCT-TGTA-3Ј. The first PCR was performed with a set of FLAG-del-N and CS-AS primers and a set of CS-S and m2-3Ј primers with mPGES-2 cDNA as a template. The resulting fragments were mixed, annealed, and then subjected to a second PCR with a set of FLAG-del-N and m2-3Ј primers. The PCR conditions were 25 cycles of 94°C for 30 s, 55°C for 30 s, and 72°C for 30 s. The resulting fragment was purified by the Wizard SV gel and PCR clean-up system (Promega), subcloned into the pLenti6/V4-D-TOPO vector, and sequenced with an autofluorometric DNA sequencer 310 Genetic Analyzer (Applied Biosystems) to confirm the mutation. The plasmid was then transfected into COX-2-expressing HEK293 cells with the ViraPower lentiviral expression system, and virus-infected cells were selected by culturing with blastcidin, as noted above.
RNA Blotting-Approximately equal amounts (5-10 g as required for experiments) of total RNA obtained from the cells were applied to separate lanes of 1.2% (w/v) formaldehyde-agarose gels, electrophoresed, and transferred to Immobilon-N membranes (Millipore). The resulting blots were then probed with the respective cDNA probes that had been labeled with [ 32 P]dCTP (Amersham Biosciences) by random priming (Takara Biomedicals). All hybridizations were carried out as described previously (4,14).
Cell Activation-Cells (2-5 ϫ 10 4 cells/ml) were seeded into 48-well plates and cultured for 3-4 days to confluence. To assess immediate PGE 2 production, cells were stimulated for 30 min with 10 M A23187 in serum-free medium. To assess delayed PGE 2 production, HEK293 cells were incubated for 4 h with 1 ng/ml human IL-1␤ (4,14), 3Y1 cells for 24 h with mouse IL-1␤ (30), and BEAS-2B cells for 4 h with 10 3 units/ml TNF␣ in culture medium. Aliquots of the supernatants were taken for the PGE 2 enzyme immunoassay.
Confocal Laser Microscopy-Cells grown in subconfluency on glass bottom dishes (Matsunami) pre-coated with 5 g/ml fibronectin (Sigma) were fixed with 3% paraformaldehyde for 30 min in PBS. After three washes with PBS, the fixed cells were sequentially treated with 1% (w/v) bovine serum albumin and 1% (w/v) saponin in PBS (blocking solution) for 1 h, with first antibodies (1:500 dilution for FLAG, 1:200 dilution for COX-1 and mPGES-2, and 1:100 dilution for COX-2) for 2 h in blocking solution, and then with species-matched FITC-conjugated second antibodies (1:200 dilution) for 2 h in blocking solution. For double immunostaining with organelle markers of the endoplasmic reticulum (ER) and the Golgi apparatus, anti-GRP78 and anti-GM130 antibodies, respectively, from the Organelle Sampler Kit (Transduction Laboratories) were used as first antibodies (1:250 dilution for both), followed by incubation with Cy3-conjugated second antibody (1:100 dilution). After six washes with PBS, specific immunofluorescent signals were visualized with a laser scanning confocal microscope (IX70; Olympus), as described previously (14).
Reverse Transcriptase (RT)-PCR-Lipopolysaccharide (LPS) (5 mg/ kg) or vehicle (PBS) was administered intraperitoneally into 7-week-old female C57BL/6 mice (Saitama Animal Center). After 24 h, these mice were sacrificed by bleeding; their organs were removed, and total RNA was extracted by homogenization in TRIzol reagent by 10 strokes of a Potter homogenizer at 1,000 rpm. Then aliquots (1 g) of RNA were subjected to RT reaction using RNA PCR Kit (Takara Biomedicals). The primer sets used for PCR were as follows: for mouse mPGES-2, 5Ј-GT-ACAAGACATCTCCCTTCTGC-3Ј and 5Ј-CACCATGGCCCAGGCGGC-GCGCCTTTC-3Ј; and for mouse mPGES-1, 5Ј-GAGCCCACCGCAACG-ACATG-3Ј and 5Ј-CAGATGGTGGGCCACCTCCC-3Ј. The PCR was conducted with ExTaq polymerase (Takara Biomedicals) with 30 cycles of 94°C for 30 s, 56°C for 30 s, and 72°C for 30 s. The reaction products were electrophoresed on 1% agarose gels, and the bands were visualized with ethidium bromide staining, ligated into pCR3.1 (Invitrogen), and transfected into Top10FЈ supercompetent cells (Invitrogen). The plasmids were isolated and sequenced with a Taq cycle sequencing kit (Takara Biomedicals) and an autofluorometric DNA sequencer 310 Genetic Analyzer (Applied Biosystems) to confirm the sequences.

FIG. 1. Processing of mPGES-2.
A, schematic structures of mPGES-2-FL and mPGES-2-del-N, which are tagged with the FLAG epitope at their N termini. An N-terminal hydrophobic region and a thioredoxin/glutaredoxin-like domain are noted. B, HEK293 cells were transfected with FLAG-mPGES-2-FL or FLAG-mPGES-2-del-N by lipofection using LipofectAMINE 2000. After 2 days of transfection, lysates of the transfectants were subjected to immunoblotting with anti-FLAG and anti-mPGES-2 antibodies. C, the transfectants were cultured for 1 month in the presence of geneticin, and their lysates were subjected to immunoblotting with anti-FLAG and anti-mPGES-2 antibodies. D, after 2 days of transfection, cell lysates were centrifuged for 1 h at 100,000 ϫ g, and equal portions of the resultant supernatant (Sup) and membrane fractions as well as the lysates were taken for immunoblotting with anti-FLAG antibody. E, the same membrane in the upper panel of D was reprobed with anti-mPGES-2 antibody. F, lysates of parental HEK293 cells and stable FLAG-mPGES-2-FL transfectants were each separated into the 100,000 ϫ g supernatant and pellet and were subjected to immunoblotting with anti-mPGES-2 antibody. Representative results of three independent experiments are shown. In the blots with anti-mPGES-2 antibody, exposure of the ECLreacted membranes to x-ray films was stopped before endogenous mPGES-2 became visible (B, C, and E). F, endogenous mPGES-2 protein was visualized by extending the exposure time.

mPGES-2 and PGE 2 Production
Immunohistochemistry-Immunohistochemical analyses on human tissues were performed as described previously (22). Briefly, the tissue sections were incubated with Target Retrieval Solution (Dako) as required, incubated for 10 min with 3% (v/v) H 2 O 2 , washed 3 times with TBS for 5 min each, incubated with 5% (v/v) skim milk for 30 min, washed 3 times with TBS/Tween for 5 min each, and incubated for 2 h with anti-mPGES-1 or anti-mPGES-2 antibody (1:100 dilution in TBS). Then the sections were treated with the CSA system staining kit (for mPGES-1; Dako).
Other Procedures-Protein concentrations were determined by BCA protein assay kit (Pierce). Data were analyzed by Student's t test.
Results are expressed as the mean Ϯ S.E., with p ϭ 0.05 as the limit of significance.

RESULTS
Processing of mPGES-2 Protein-N-terminally FLAG epitope-tagged mPGES-2-FL and mPGES-2-del-N (structures illustrated in Fig. 1A) were each transfected into HEK293 cells, and their protein expression was assessed by Western blotting with anti-FLAG and anti-mPGES-2 antibodies. After 2 days of transfection, both antibodies recognized a single 34-kDa band in cells transfected with FLAG-mPGES-2-del-N (slightly larger than mPGES-2-del-N (33 kDa) because of the presence of the FLAG epitope) (Fig. 1B). In mPGES-2-FL-transfected cells, anti-FLAG antibody recognized a single 43-kDa band, which corresponded to the size of mPGES-2-FL (42 kDa) plus the FLAG epitope, whereas two bands (44 and 33 kDa) were detected by anti-mPGES-2 antibody (Fig. 1B). Furthermore, after longer culture of mPGES-2-FL-transfected cells, only the lower band was detected by anti-mPGES-2 antibody, whereas no protein band was visualized by anti-FLAG antibody (Fig. 1C). A single 34-kDa band was again detected in mPGES-2-del-Ntransfected cells by both antibodies. A similar molecular shift from 44-kDa mPGES-2-FL to a 33-kDa smaller species, which is likely to correspond to mPGES-2-del-N, occurred when mPGES-2-FL was transfected into several other cell lines (data not shown). Moreover, mPGES-2 endogenously expressed in these cells, which was visualized by anti-mPGES-2 antibody, was detected predominantly as the smaller species (see below). These results suggest that mPGES-2-FL is spontaneously processed during cell culture by proteolysis to a mature, N-terminally truncated form.
Cell homogenates transiently transfected with FLAG-tagged mPGES-2-FL or mPGES-2-del-N were centrifuged at 100,000 ϫ g to separate into the supernatant and membrane fractions. When these samples were subjected to Western blotting with anti-FLAG antibody, mPGES-2-FL was predominantly detected in the membrane fraction, whereas mPGES-2del-N was distributed preferentially in the supernatant fraction (Fig. 1D). This result suggests that the N-terminal hydrophobic region, which is subsequently removed by proteolytic cleavage, is required for tight association of mPGES-2-FL with membrane, and the mature, N-terminally truncated enzyme behaves as a soluble protein that retains a weak membrane affinity. Interestingly, when the same samples from the mPGES-2-FL-transfected cells were blotted with anti-mPGES-2 antibody, the truncated fragment produced from the full-length protein was mainly detected in the membrane (Fig.  1E). However, if the mPGES-2-FL-transfected cells were cultured for a longer period until the full-length protein was invisible, the truncated form (an only species recognized by anti-mPGES-2 antibody; see Fig. 1C) was preferentially distributed in the supernatant fraction (Fig. 1F), as predicted from the blot with anti-FLAG antibody (Fig. 1D, lower panel). Moreover, endogenous mPGES-2 protein, which appeared as the same size as the truncated form of the overexpressed enzyme, was also preferentially distributed in the supernatant fraction (Fig. 1F). Thus, it appears that the truncated mPGES-2 protein remains membrane-associated soon after the initial cleavage of the full-length protein and is then shifted to a soluble form in accordance with the disappearance of the full-length protein.
Subcellular Localization of mPGES-2-In keeping with the above results, we performed confocal microscopic analyses to determine the subcellular localization of mPGES-2. To this end, HEK293 and BEAS-2B cells transfected with FLAGtagged mPGES-2-FL or mPGES-2-del-N were subjected to immunocytostaining with anti-FLAG antibody. In HEK293 cells, a strong immunoreactive signal for mPGES-2-FL was detected in one side of the perinuclear region ( Fig. 2A). This signal was colocalized with the Golgi marker GM130 (Fig. 2B) but not with the ER marker GRP78 (data not shown), indicating that mPGES-2-FL is associated with the Golgi membrane. Similarly, mPGES-2-FL signal was overlapped with that for GM130 in BEAS-2B cells in the perinuclear region (Fig. 2C). No detectable signals for mPGES-2-FL were found in the perinuclear envelope and cytosol. In contrast, the signal for mPGES-2del-N was distributed throughout the cytoplasm with a trend that the perinuclear area yielded more intense signal in both HEK293 (Fig. 2D) and BEAS-2B (Fig. 2E) cells. Immunocytostaining with anti-mPGES-2 antibody provided a similar staining pattern (data not shown). These results suggest that mPGES-2 is initially synthesized as a Golgi membrane-associated protein, and after complete proteolytic removal of the N-terminal hydrophobic domain, the truncated enzyme is dispersed into the cytoplasm. Nonetheless, a significant fraction of mPGES-2-del-N was still colocalized with the Golgi marker GM130, especially in BEAS-2B cells (Fig. 2F).
COX Coupling-To assess whether mPGES-2 promotes cellular PGE 2 biosynthesis and, if so, to compare directly the COX coupling selectivity of mPGES-2 with that of mPGES-1, mPGES-2 (in this case mPGES-2-del-N, a putative mature form (see above)) or mPGES-1 was transfected into HEK293 cells in combination with COX-1 or COX-2. After verifying the appropriate expression levels of these enzymes in the respective transfectants by Western blotting (Fig. 3A), the PGE 2 biosynthetic capacities of these cells from exogenous or endogenous AA were investigated.
Transfection of COX-1 or COX-2 alone into HEK293 cells resulted in dose-dependent increases in production of PGE 2 from exogenous AA (Fig. 3B), as reported previously (4). Although transfection of mPGES-2 alone did not increase PGE 2 significantly, cells cotransfected with mPGES-2 in combination with either COX-1 or COX-2 produced more PGE 2 than did cells transfected with the respective COX alone in an AA concentration-dependent manner (Fig. 3B). Augmentation of COX-1-mediated PGE 2 production by mPGES-2 became evident at more than 2-5 M AA, whereas that of COX-2-mediated production increased almost linearly in proportion to incremental concentrations of AA. In comparison, mPGES-1 enhanced COX-2-directed conversion of AA to or COX-2 and control cells were incubated for 30 min with the indicated concentrations of AA, and PGE 2 released into the supernatants was quantified. D and E, production of PGE 2 from endogenous AA. Cells expressing mPGES-2 (D) or mPGES-1 (E) alone or in combination with COX-1 or COX-2 and control cells were stimulated for 30 min with A23187 or for 4 h with IL-1␤, and PGE 2 released into the supernatants was quantified.
To address PGE 2 production from endogenous AA, cells were stimulated with A23187 for 30 min or with IL-1␤ for 4 h. Cotransfection of mPGES-2 with COX-1 or COX-2 resulted in marked augmentation of PGE 2 production as compared with transfection of either COX alone in both A23187-and IL-1␤stimulated responses (Fig. 3D). In comparison, when mPGES-1 was cotransfected with either COX enzyme, augmented PGE 2 production occurred via COX-2 in marked preference to COX-1 following both stimuli (Fig. 3E). Thus, irrespective of the sources of AA, the COX coupling profiles of the two mPGES enzymes are apparently distinct in that mPGES-2 is COX non-selective and mPGES-1 exhibits a marked COX-2 preference. It is notable that more PGE 2 was produced via mPGES-2 after stimulation with A23187 than with IL-1␤, whereas mPGES-1-mediated PGE 2 production was greater after stimulation with IL-1␤ than with A23187 in this cell system. There was no significant increase in PGE 2 production after transfection of mPGES-2-CS, a catalytically inactive mPGES-2 mutant (31), into COX-2-expressing HEK293 cells (Fig. 3F), implying that the catalytic activity of mPGES-2 is required for its cellular function.
To ensure that the coupling between the two COXs and the two mPGESs observed above was not a peculiarity of HEK293

FIG. 4. Coupling of COXs and mPGESs in human lung epithelial BEAS-2B cells.
A, expression of mPGES-2 (del-N), mPGES-1, COX-1, and COX-2 proteins in BEAS-2B transfectants obtained by lentivirus-mediated gene transfer, as assessed by immunoblotting with anti-mPGES-2, -mPGES-1, -COX-1, and -COX-2 antibodies, respectively. B and C, subcellular localization of COX-1, COX-2, and mPGES-1 in the transfectants. Cells grown on fibronectin-coated glass dishes were fixed with paraformaldehyde and treated sequentially with anti-FLAG antibody and FITC-conjugated second antibody, and fluorescent signals (green) were detected by confocal laser microscopy. C, the cells were also incubated with anti-GRP78 antibody and Cy3-conjugated second antibody to visualize the ER and associated perinuclear envelope (red). Colocalization of COX-1 or mPGES-1 with GRP78 is shown by yellow (merge). Staining of control cells is shown in B. D, these cells were stimulated for 30 min with A23187 or for 4 h with TNF␣ to assess PGE 2 production. Representative results of 3 independent lentivirus-infection studies are shown.

FIG. 5. Coupling of COXs and mPGESs in rat fibroblastic 3Y1 cells. A, expression of mPGES-2 (del-N) and mPGES-1 proteins in 3Y1
transfectants obtained by lentivirus-mediated gene transfer, as assessed by immunoblotting with anti-mPGES-2 and -mPGES-1 antibodies, respectively. B and C, PGE 2 production by 3Y1 cells transfected with mPGES-2 (B) or mPGES-1 (C) and control cells after stimulation for 30 min with A23187 or for 24 h with IL-1␤. Representative results of 3 independent lentivirus-infection studies are shown. cells, we next introduced these enzymes, alone or in combination, into human lung epithelial BEAS-2B cells by lentivirusmediated gene transfer. Appropriate expression of these proteins in the transfectants, as assessed by immunoblotting with their respective antibodies, is shown in Fig. 4A. In this cell line, expressions of endogenous COX-1, COX-2, and mPGES-1 were below detection, whereas endogenous mPGES-2 was weakly detected at a size similar to that of mPGES-2-del-N ( Fig. 4A; also see below). Immunocytostaining of BEAS-2B transfectants with antibodies against COX-1, COX-2, and mPGES-1 indicated their locations in the ER membrane and the perinuclear envelope (Fig. 4B), as shown by colocalization with the ER marker GRP78 (Fig. 4C) (see Fig. 2 for localization of mPGES-2 in BEAS-2B cells). COX-1 appeared to be distributed rather uniformly throughout the ER membrane, whereas COX-2 was enriched in the perinuclear region. Similar locations of these enzymes were seen in HEK293 cells, as described previously (14,16).
Expression of Endogenous mPGES-1 and mPGES-2-Expression of endogenous mPGES-1 and mPGES-2 in several cell lines (Fig. 6, A-C) and mouse tissues (Fig. 6D), before and after proinflammatory stimuli, was next investigated. Endogenous mPGES-2 protein (assessed by immunoblotting) (Fig. 6A) and mRNA (by Northern blotting) (Fig. 6B) were expressed constitutively in HEK293, BEAS-2B, 3Y1, and HCA-7 cells, in which its expression was minimally altered by IL-1␤ (for HEK293, 3Y1, and HCA-7) or TNF␣ (for BEAS-2B) over 24 h. The majority of endogenous mPGES-2 protein detected by immunoblotting in these cells appeared to correspond to the del-N form in size (Fig. 6A). RT-PCR was performed to detect endogenous mPGES-1 because its expression levels in these cells (except HCA-7 cells; see below) were relatively low. As shown in Fig.  6C, endogenous mPGES-1 expression was induced significantly in HEK293 and 3Y1 cells after IL-1␤ stimulation and was constitutive in BEAS-2B and HCA-7 cells. Although mPGES-1 mRNA expression in 3Y1 cells and HCA-7 cells was detectable in an agarose gel by staining with ethidium bromide (Fig. 6C, upper panel), expression in HEK293 and BEAS-2B cells was evident only by RT-PCR followed by high sensitivity Southern blotting (Fig. 6C, lower panel). Notably, the expression levels of both mPGES-1 and mPGES-2 in HCA-7 cells, a human colorectal adenocarcinoma cell line, were rather higher than those in the other three cell lines.
Following administration of LPS to mice, the expression of mPGES-1 was markedly elevated over its basal, low level of expression in all tissues examined (brain, heart, liver, lung, and colon) (Fig. 6D), in agreement with previous reports (14). In contrast, mPGES-2 was expressed rather constitutively in these tissues, with no increase (brain, heart, and lung) or only a modest increase (liver and colon) after LPS treatment (Fig.  6D). These results suggest that the expression of mPGES-2 is relatively ubiquitous and constitutive, whereas that of mPGES-1 is varied among different cell types and is often stimulus-inducible.
Immunohistochemistry-Finally, we performed immunohistochemistry of the two mPGES enzymes in human tissues with inflammation, damages, and cancer.
Rheumatoid Arthritis (RA)-PGE 2 is a critical mediator of the pathogenesis of RA, in which COX-2 plays a pivotal role (32)(33)(34). Immunohistochemical staining of paraffin-embedded sections of RA joints from several patients revealed that mPGES-2 immunoreactivity was distributed in synovial lining and sublining cells (representative results from tissue sections of two patients are shown in Fig. 7, A-C). The staining pattern was similar among different sections from individual patients with different pathological states (A and B from a patient with active RA and C from a patient with inactive RA in Fig. 7). In contrast, mPGES-1 immunoreactivity was restricted to synovial lining cells in RA tissues in the active state (Fig. 7, D and  E), whereas it was barely detectable in tissues in the resting state (Fig. 7F). Consistently, immunoblot analysis showed that mPGES-1 expression was remarkably high in sample b, from the patient who exhibited the most severe symptoms diagnostically, as compared with samples a and c, from the patients with milder RA (Fig. 7G). mPGES-2 expression was similar among three different RA samples a-c (even though sample b appeared to contain slightly more mPGES-2 than samples a and c) (Fig. 7G). In support of these observations, the expression of COX-2 and group IIA secretory PLA 2 , which has been

mPGES-2 and PGE 2 Production
shown to be correlated with the severity of RA (35,36), was highest in sample b (Fig. 7G). Expression of cytosolic PLA 2 was also most abundant in sample b, whereas that of COX-1 was similar among the three samples, as was that of ␣-tubulin used as a control.
Acute Myocardial Infarction-mPGES-2 immunoreactivity was detected in the whole area of normal myocardium (Fig.  8A), in which mPGES-1 was undetectable (Fig. 8B). In infarct lesions, mPGES-1 expression was particularly detected in damaged but surviving myocytes (Fig. 8D), whereas the intensity of mPGES-2 staining in these damaged cells was similar to that in undamaged cells (Fig. 8C). Interstitial fibrotic foci and coronary arterial walls were negative for both mPGES-1 and mPGES-2.
Viral Hepatitis-Hepatocytes from the intact liver were positively stained for mPGES-2 ( Fig. 9A) but not for mPGES-1 (Fig. 9B). Hepatitis C virus-infected hepatocytes from a patient with progressive liver cirrhosis were strongly positive for mPGES-1, with persistent presence of mPGES-2 at the perilobular area (Fig. 9, C and D). Staining at the portal area, into which lymphocytes were infiltrated, was minimal for both enzymes.
Colorectal Cancer-Expression of mPGES-1 was barely detectable in normal large intestinal gland and submucosa, whereas it was intensely positive in colorectal adenocarcinomatous lesions (Fig. 10, A and B). mPGES-2 was also markedly expressed in adenocarcinoma per se, whereas staining of normal gland was weak (Fig. 10, C and D). In addition, a scattered signal for mPGES-2 was found in the submucosa (Fig. 10C). DISCUSSION mPGES-2 was originally purified as an N-terminally truncated form from the microsomal fraction of bovine heart (27), and recombinant human and monkey enzymes (both fulllength and N-terminally truncated forms) exhibit significant and comparable PGES activities that can be activated by glutathione, lipoic acid, and other reducing agents (28,31). The k cat /K m value of recombinant mPGES-2 is about one-fifth that of recombinant mPGES-1 under their optimal assay conditions (28, 37). The gene for mPGES-2 is located on human chromosome 9q33-34, a locus where the genes for COX-1 (9q32-33), mPGES-1 (9q34.3), and lipocalin-type PGD synthase (9q34.2-34.3) are clustered, suggesting that this chromosome locus is related to the metabolisms of PGs. To better understand the regulatory functions of mPGES-2 in mammalian cells, we herein examined the cellular PGE 2 biosynthetic function and expression of mPGES-2 in various cells and tissues.
We found that mPGES-2 predominantly exists as a shorter form than its predicted full-length size in several cell lines with different tissue and animal origins. Moreover, mPGES-2-FL overexpressed in several cells was found to be spontaneously converted to an N-terminally truncated form similar in size to mPGES-2-del-N (lacking the N-terminal 87 amino acids), which corresponds to the enzyme originally purified from the heart (27). It is thus likely that mPGES-2 undergoes proteolytic processing between amino acid residues 87 and 88 ((Ala 87 / Glu 88 )-Arg-Ser) or another site close to this position by a certain protease to give rise to the mature enzyme in cells. Fractionation (Fig. 1) and immunocytostaining (Fig. 2) studies revealed that mPGES-2-FL is associated with the Golgi membrane and is dispersed in the cytoplasm after removal of the N-terminal region. Nonetheless, mPGES-del-N still possesses a weak membrane affinity, which may contribute to its partial enrichment in the perinuclear area. Alignment of primate and mouse mPGES-2 proteins demonstrates that the homology within the N-terminal first 48 amino acids is low, followed by a highly homologous, alanine-and leucine-rich hydrophobic region (residues 49 -87), which is likely to be responsible for the tight association of the full-length enzyme with the Golgi membrane. At present, however, which sequence motif in this region allows targeting of the enzyme into the Golgi membrane and which type of protease cleaves the enzyme to be released into the cytoplasm remain unclear. Of note, the truncated form produced from mPGES-2-FL is first retained on the membrane and is then released into the cytoplasm in accordance with the disappearance of the full-length protein (Fig. 1E). This observation provides some impact on the mechanism for the proteolytic processing of mPGES-2. Conceivably, the cleavage of mPGES-2-FL may occur sequentially at two proximal sites, where the initial, intermediate truncated product still retains the high membrane-binding capacity and the second cleavage gives rise to a mature, cytosolic form. Alternatively, the truncated mPGES-2 may be oligomerized with the full-length enzyme on the Golgi membrane and the completion of the proteolytic processing eventually leads to the release of the mature enzyme into the cytoplasm.
Functional assessment of mPGES-2 (in comparison with mPGES-1) by transfection studies has revealed that the mature mPGES-2-del-N indeed exhibits PGE 2 biosynthetic action in cells. Although the sensitivity of different cells to distinct stimuli appears varied, an emerging picture is that mPGES-2 is an enzyme capable of coupling with both COX isozymes in both the immediate and delayed phases, whereas mPGES-1 plays more profound role in COX-2-dependent PGE 2 production. The COX coupling profiles of the two mPGES enzymes agree with their distinct expression patterns in that mPGES-2 is constitutively expressed, whereas mPGES-1 is stimulus-inducible (in many, if not all, cases), in cultured cells stimulated with proinflammatory cytokines (Fig. 6C), in tissues of LPStreated mice (Fig. 6D), and in human tissues with various disorders (Figs. 7-10).
Current evidence has suggested the importance of the compartmentalization of COXs and various terminal PG synthases for their efficient functional coupling (14 -16, 38). In our previous cotransfection studies, the perinuclear terminal PG synthases (e.g. mPGES-1, thromboxane synthase, and prostacyclin synthase) show a marked coupling preference for COX-2 over COX-1, whereas cytosolic terminal PG synthases (e.g. cPGES and hematopoietic PGD synthase) are capable of being coupled with COX-1 distributed throughout the ER membrane (14 -16). The ability of the mature mPGES-2, which is dispersed in the cytosol, to be coupled with COX-1 appears to be in line with this scenario. Because mPGES-2 and COX-1 are constitutively expressed, their functional coupling argues for a general role of mPGES-2 in the production of PGE 2 crucial for tissue homeostasis. Considering that there are two COX-1-coupled PGES enzymes in cells, mPGES-2 and cPGES, a key question is how these enzymes are functionally segregated. Because the cellular PGE 2 biosynthetic action of cPGES requires the supply of high concentrations of AA (15) and regulatory factors (26), its PGES function may be manifested only when the cPGES activation machinery is operative.
Notably, mPGES-2 represents the second PGES that can be coupled with COX-2. This coupling seems reasonable because mPGES-2 displays some perinuclear-enriched location and because COX-2 is intrinsically favored over COX-1 at lower AA or hydroperoxide levels (4,12,13). The existence of a COX-2coupled PGES other than mPGES-1 has been anticipated by several knockout mouse studies. COX-2-deficient mice show numerous female reproductive abnormalities with defects in ovulation, fertilization, implantation, and decidualization (39), and the PGE receptor EP2 knockout mice also exhibit a defect in the ovulation process (40). These observations imply a crucial role for COX-2-derived PGE 2 in the female reproductive process. Even though induction of mPGES-1 expression has been observed during the period of pregnancy (21), no abnormality has so far been found in the reproductive system of mPGES-1-deficient mice (23). Moreover, although mice deficient in both COX-1 and COX-2 (41) or the PGE receptor EP4 (42) frequently undergo neonatal death due to failure of closure of the ductus arteriosus, mPGES-1-null mice are all viable after birth (23), indicative of no gross abnormality in this developmental process. From this standpoint, our present finding suggests that mPGES-2 may compensate for mPGES-1 in these physiological events. Nevertheless, mPGES-2 transcript is more abundantly distributed in the brain, heart, skeletal muscle, kidney, and liver than in other tissues (28), whereas mPGES-1 transcript is primarily detected in organs related to the immune response, such as lung and spleen, as well as in male genital organs (14,17), suggesting that these two enzymes are not always redundant but rather exhibit tissuespecific functions. Thus, in cells and tissues in which mPGES-1 induction is limited, mPGES-2 could be a major player acting downstream of COX-2 for PGE 2 biosynthesis.
Immunohistochemical analyses of several human tissue sections also support the distinct expression (and therefore distinct roles) of the two mPGES enzymes. RA is a complex autoimmune syndrome involving a number of proinflammatory factors, one of which is PGE 2 (34,(43)(44)(45). The crucial role of COX-2 in producing proinflammatory PGE 2 in RA has been confirmed by clinical trials of COX-2 inhibitors (33,46,47) and by studies using COX-2 knockout mice (32). Recently, it has been reported that induction of mPGES-1 occurs in cultured rheumatoid synoviocytes stimulated with IL-1␤ or TNF␣ (19) and that mPGES-1 is markedly up-regulated, whereas mPGES-2 is constitutively expressed in rat adjuvant arthritis (48). Consistent with these studies, we found here that mPGES-1 is mainly located in the synovial lining cells of RA in the active state, whereas its expression in the quiescent state is below the limit of detection. On the other hand, mPGES-2 is detected in both active and resting RA tissues, where it was located in synovial cells and subsynovial tissues. A previous immunohistochemical study has demonstrated the localization of COX-2 in synovial lining cells as well as in blood vessel endothelium, chondrocytes, and sublining fibroblasts in correlation with the severity of RA, whereas COX-1 is expressed in synovial cells with no disease-related differences (35). These observations support the view that mPGES-1, downstream of COX-2, contributes to augmented PGE 2 production by synovial lining cells, whereas mPGES-2 expressed rather constitutively in synovial lining, and sublining tissues may play a role in the basal production of PGE 2 unrelated to the progress of the disease.
COX-derived PGs have beneficial effects on myocardiocytes in the ischemic heart, and functional loss of COXs is associated with enhanced ischemia/reperfusion injury (49 -51). It is therefore likely that the induced expression of mPGES-1 and attendant augmented production of PGE 2 in myocardiocytes after infarction may contribute to protection of myocardiocytes from injury and reduction of cardiac cell death, whereas PGE 2 produced via the constitutive mPGES-2 may play a role in the maintenance of cardiac homeostasis. Likewise, COX-derived PGs, particularly PGE 2 , play a role in liver regeneration and protection of hepatic tissues from damage (52)(53)(54)(55)(56). Hepatocyte proliferation after partial hepatectomy is significantly reduced when COX-1-deficient mice are treated with a COX-2 inhibitor (55), suggesting the involvement of both COX isoforms in liver regeneration. Expression of mPGES-1 in the hepatitis and associated cirrhosis caused by infection with hepatitis C virus suggests its role in augmented production of PGE 2 that protects hepatocytes from disease-associated damage, whereas mPGES-2 constitutively expressed in both normal and pathologic hepatocytes may play a housekeeping role. Recently, mPGES-2 was found to be identical to GBF-1, a novel factor that is modestly increased in IFN-␥-stimulated RAW cells (57). Considering that mPGES-2 expression is elevated modestly in the liver of LPS-treated mice (Fig. 6B), its expression could be up-regulated in particular cell types (such as liver cells) under particular conditions. Because IFN-␥ is a critical cytokine for the pathogenesis of viral infection, some increase in the immunoreactive signal for mPGES-2 in the restricted area of viral hepatitis tissues may reflect the local response to IFN-␥.
A marked elevation of mPGES-2, as well as mPGES-1, in human colorectal adenocarcinoma tissues and cell lines (HCA-7 cells) is intriguing, because PGE 2 produced via the COX-2 pathway has been implicated in the development of colorectal cancer and possibly other cancers (58 -68). Recent evidence suggests that mPGES-1 expression is elevated in colorectal and lung cancers and that mPGES-1 can promote tumorigenesis in cooperation with COX-2 (22,69,70). Elevated expression of mPGES-1 is accompanied by increased expression of the receptors for epidermal growth factor and vascular endothelial cell growth factor (22), consistent with the notions that PGE 2 transactivates the epidermal growth factor receptor, thereby triggering mitogenic signaling (71), and that PGE 2 promotes angiogenesis (65,66,72) in cancers. The present observation that mPGES-2, as well as mPGES-1, is highly expressed in colorectal adenocarcinoma cells implies that this enzyme also has the potential to contribute to tumorigenesis, a possibility that is now under investigation.
In summary, the present study has revealed cell biological, functional, and immunohistochemical aspects of mPGES-2. We have shown that mPGES-2 undergoes proteolytic processing, is functionally coupled with both COX-1 and COX-2, and is expressed in cells and tissues relatively constitutively with the exception of marked elevation in colorectal cancer. Thus, mammalian cells contain at least three PGES enzymes, a COX-1selective cPGES, a COX-2-preferring mPGES-1, and mPGES-2 that is non-selective for both COXs. The remaining questions that need to be addressed in future studies are as follows: (i) the mechanism of the Golgi sorting and proteolytic processing of mPGES-2, including the identification of the protease involved in this process; (ii) the function of the full-length mPGES-2 that appears transiently on the Golgi membrane; (iii) post-translational modification of mPGES-2 (such as phosphorylation) if it exists; (iv) the relative contribution of the three PGES enzymes in PGE 2 biosynthesis in different cell types; and (v) the physiological and pathological roles of these enzymes in vivo. As mPGES-1 knockout mice are now available (23), the in vivo functions of mPGES-1 will be solved within the next few years. Together with the forthcoming gene targeting of cPGES and mPGES-2, these studies will give us comprehensive answers to these critical questions and provide the basis for future studies that will evaluate if these enzymes are bona fide therapeutic targets.