Potential role of microsomal prostaglandin E synthase-1 in tumorigenesis.

Microsomal prostaglandin E2 synthase-1 (mPGES-1) is a stimulus-inducible enzyme that functions downstream of cyclooxygenase (COX)-2 in the PGE2-biosynthetic pathway. Given the accumulating evidence that COX-2-derived PGE2 participates in the development of various tumors, including colorectal cancer, we herein examined the potential involvement of mPGES-1 in tumorigenesis. Immunohistochemical analyses demonstrated the expression of both COX-2 and mPGES-1 in human colon cancer tissues. HCA-7, a human colorectal adenocarcinoma cell line that displays COX-2- and PGE2-dependent proliferation, expressed both COX-2 and mPGES-1 constitutively. Treatment of HCA-7 cells with an mPGES-1 inhibitor or antisense oligonucleotide attenuated, whereas overexpression of mPGES-1 accelerated, PGE2 production and cell proliferation. Moreover, cotransfection of COX-2 and mPGES-1 into HEK293 cells resulted in cellular transformation manifested by colony formation in soft agar culture and tumor formation when implanted subcutaneously into nude mice. cDNA array analyses revealed that this mPGES-1-directed cellular transformation was accompanied by changes in the expression of a variety of genes related to proliferation, morphology, adhesion, and the cell cycle. These results collectively suggest that aberrant expression of mPGES-1 in combination with COX-2 can contribute to tumorigenesis.

Clinical, genetic, and biochemical evidence has suggested that prostaglandin (PG) 1 E 2 produced via the cyclooxygenase (COX)-2-dependent pathway plays a crucial role in the development of colorectal cancer and possibly other cancers (1). Non-steroidal anti-inflammatory drugs, which inhibit COX-2, reduce the incidence of colorectal cancer (2)(3)(4). The major prostanoid produced by several types of cancer is PGE 2 , which is produced by three biosynthetic reactions involving phospho-lipase A 2 (PLA 2 ), COX, and terminal PGE 2 synthase (PGES). PGE 2 promotes survival and motility of colon cancer cells in vitro and promotes tumorigenesis and angiogenesis in vivo (5)(6)(7). High levels of constitutive expression of COX-2 have been found in various cancer cells and tissues (8,9), and studies employing overexpression, antisense suppression, and specific inhibitors of COX-2 have demonstrated that COX-2 contributes to the progression of several types of cancer (10 -12).
More direct evidence for the role of COX-2 and its product PGE 2 in colorectal tumorigenesis has been provided by gene targeting studies. Gene disruption of either COX-2 (13) or the PGE receptor EP2 (14) results in reduction of the number and size of intestinal polyps in Apc mutant mice, a model for human familial adenomatous polyposis. In another model, disruption of the genes for the PGE receptors EP1 (15) or EP4 (16) suppresses the development of colorectal cancer induced by carcinogen. Moreover, gene knockout of cytosolic PLA 2 ␣ (cPLA 2 ␣), which supplies the substrate arachidonic acid to COX-2, also leads to reduced polyposis in Apc mutant mice (17,18).
PGES catalyzes the conversion of PGH 2 , which is produced from arachidonic acid by COX-1 or COX-2, to PGE 2 . Recent advances in this field have led to identification of at least three PGES enzymes, including cytosolic PGES (cPGES) (19), microsomal PGES (mPGES) -1 (20 -22), and mPGES-2 (23). Among them, microsomal PGES-1 (mPGES-1) has received much attention, as this enzyme is induced by proinflammatory stimuli, down-regulated by anti-inflammatory glucocorticoids, and functionally coupled with COX-2 in marked preference to COX-1 (20 -22). In comparison, cPGES (the heat shock proteinassociated protein p23) is constitutively and ubiquitously expressed and is selectively coupled with COX-1 (19). mPGES-2 does not show homology with mPGES-1 and has a unique N-terminal hydrophobic domain and a glutaredoxin-like domain (23), although its cellular function has not yet been addressed.
mPGES-1 is a member of the MAPEG (for membrane-associated proteins involved in eicosanoid and glutathione metabolism) superfamily, to which other proteins involved in arachidonic acid metabolism, such as 5-lipoxygenase-activating protein (FLAP) and leukotriene C 4 synthase, also belong (20 -22). Induced expression of mPGES-1 has been postulated to be associated with various pathophysiological events in which COX-2-derived PGE 2 has been implicated, such as rheumatoid arthritis (24), febrile response (25), reproduction (26,27), bone metabolism (21), and Alzheimer's disease (28). A recent gene targeting study of mPGES-1 has shown that PGE 2 production by lipopolysaccharide-stimulated peritoneal macrophages depends almost entirely on this enzyme (29). Induced expression of mPGES-1 is regulated by the NF-IL-6 pathway (29) or the mitogen-activated protein kinase pathway (30), the latter of * This work was supported by a Showa University grant-in-aid for Innovative Collaborative Research Projects and a Special Research grant-in-aid for Development of Characteristic Education from the Japanese Ministry of Education, Culture, Sports, Science 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.
A possible linkage of mPGES-1 with tumorigenesis has been provided by a recent observation that mPGES-1 is constitutively expressed in several cancers, most of which also express COX-2 constitutively (32)(33)(34). In this study, we have used colon cancer cell lines and mPGES-1-transfected cells to examine the expression of mPGES-1 in colorectal cancer tissues and cells and evaluate its potential role in tumorigenesis.  (21), human COX-1 and COX-2 (35), and human cPGES (19) were described previously. cDNA probes for human rhoA and human c-myc were donated by Dr. M. Shibanuma (Showa University). HEK293 cells stably expressing COX-2 and/or mPGES-1 were described previously (21). The enzyme immunoassay kits for PGE 2 and PGF 2␣ and the COX-2 inhibitor NS-398 were purchased from Cayman Chemicals. The MAPEG inhibitor MK-886 was from Toronto Research Chemicals. The rabbit anti-human cPLA 2 ␣, goat anti-human COX-1 and goat anti-human COX-2, goat anti-human RhoA and goat anti-human ezrin antibodies were purchased from Santa Cruz. Mouse monoclonal antibody against human ErbB3 and mouse monoclonal anti-FLAG antibodies were from Sigma. Rabbit antibody against human Egr-1 was provided by Dr. H. Naraba (National Cardiovascular Center Research Institute, Japan). Fluorescein isothiocyanate-conjugated anti-goat IgG, Cy3-conjugated anti-rabbit IgG, and horseradish peroxidase-conjugated anti-goat, -mouse, and -rabbit IgGs were purchased from Zymed Laboratories Inc.. Rabbit antiserum for human cPGES was prepared as described previously (19). Human interleukin (IL)-1␤ was purchased from Genzyme. LipofectAMINE 2000, oligofectamine, Opti-MEM medium, TRIzol reagent, geneticin, hygromycin, zeocin, and the pcDNA3.1 series of mammalian expression vectors were obtained from Invitrogen.
New Zealand White rabbits (male, 1 kg; Saitama Animal Center) were immunized subcutaneously with the purified mPGES-1 protein (0.3 mg/head) mixed with Freund's complete adjuvant (Difco). After several booster immunizations with Freund's incomplete adjuvant (Difco) at 2-week intervals, the blood was collected and the serum titer was assayed by enzyme-linked immunosorbent assay and Western blotting with recombinant mPGES-1 protein. The antiserum obtained was further purified on an immunoaffinity Hi-Trap NHS-activated column (Amersham Biosciences) that had been conjugated with mPGES-1 protein. The purified antibody was used in subsequent studies.
Transfection Studies-Transfection of cDNAs into HEK293 cells was performed by lipofection as described previously (21,35). Briefly, 1 g of plasmid (mPGES-1 in pCDNA3.1/hyg and COX-1 or -2 in pCDNA3.1/neo) was mixed with 2 l of LipofectAMINE 2000 in 100 l of Opti-MEM 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. The cells were cloned by limiting dilution in 96-well plates in culture medium contain-ing appropriate antibiotics (10 g/ml hygromycin or 1 mg/ml G418). After culture for 3-4 weeks, wells containing a single colony were chosen, and the expression of each protein was assessed by RNA blotting. The established clones were expanded and used for the experiments described below.
The C-terminal FLAG-tagged mPGES-1 cDNA was transfected into HCA-7 cells by the ViraPower lentiviral expression system (Invitrogen) according to the manufacturer's instructions. Briefly, the FLAG-tagged mPGES-1 cDNA insert was amplified by polymerase chain reaction with the Advantage cDNA polymerase mixture (Clontech) and was subcloned into the pLenyi6/V4-D-TOPO vector (Invitrogen). The resulting plasmid was transfected into 293FT cells (Invitrogen) with Lipo-fectAMINE 2000, and an aliquot of the supernatant harvested 3 days after transfection was then added to HCA-7 cells. The cells were cultured in the presence of 40 g/ml blastcidine (Invitrogen), and the antibiotic-resistant cells were used in subsequent studies.
Antisense Experiments-HCA-7 cells (6 ϫ 10 4 cells) were seeded into 6-well plates and cultured for 2 days. Then the mPGES-1 antisense S-oligonucleotide (0.2 nmol) 5Ј-GAGGAAGACCAGGAAGTGCAT-3Ј was transfected into HCA-7 cells with oligofectamine reagent. After 48 h, cell numbers and the PGE 2 released into the supernatants during culture were quantified. The cell lysates were subjected to Western blotting to verify mPGES-1 expression.
Measurement of PGES Activity-PGES activity was measured by assessment of conversion of PGH 2 to PGE 2 as previously reported (21). Briefly, cells were harvested from culture dishes with a cell scraper and disrupted by sonication with a Branson sonifier (10 s, 3 times, 50% duty) in 10 mM Tris-HCl (pH 8.0) containing 150 mM NaCl. After centrifugation of the sonicates at 100,000 ϫ g for 1 h at 4°C, the membrane fractions were used as an enzyme source. An aliquot (10 g of protein equivalents) was incubated with 0.5 g of PGH 2 for 30 s at 24°C in 0.1 ml of 0.1 M Tris-HCl (pH 8.0) containing 1 mM glutathione and 5 g of indomethacin. After stopping the reaction by the addition of 100 mM FeCl 2 , PGE 2 contents in the reactions were quantified by use of the enzyme immunoassay kit. Protein concentrations were determined by the bicinchoninic acid protein assay kit (Pierce) with bovine serum albumin as a standard.
RNA Blotting-Approximately equal amounts (ϳ5 g) 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 (35).
Semisoft Agar Assay-Cells (10 4 cells/ml) were suspended in cell culture medium containing 1% (w/v) low-melt agarose and plated on 60-mm culture dishes. After culture for 10 days at 37°C in a CO 2 incubator, colony numbers in each plate were counted. Relative colony size was determined by measuring 10 random colonies in each plate, and the mean for each treatment set was calculated and compared with that of controls.
Experiments with Nude Mice-Cells (5 ϫ 10 6 cells) were suspended in 100 l of phosphate-buffered saline and injected subcutaneously into BALB/c-nu/nu mice (6-week-old males) (Crea Japan). After 3 months, solid tumors were removed surgically and fixed in 10% (v/v) formalin. After embedding in paraffin, thin sections (4ϳ6 m thickness) of tumor tissues were prepared on glass slides.
Immunohistochemistry-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 the first antibodies in TBS (1:50 and 1:200 dilutions for anti-mPGES-1 and anti-COX-2 antibodies, respectively). Then the sections were treated with the LSAB2 staining kit (for COX-2; DAKO) or the CSA system staining kit (for mPGES-1; DAKO).

mPGES-1 and Tumorigenesis
cDNA Array Analysis-mRNAs isolated from COX-2-expressing and COX-2/mPGES-1-coexpressing HEK293 cells (10 7 cells for each) were reverse-transcribed into cDNA and 32 P-labeled with Atlas cDNA Expression Arrays kit (Clontech). Hybridization was performed on the Atlas Human 1.2 Array (Clontech). After exposure to an image plate (BAS III; Fuji Photo Film), signals were analyzed by AtlasImage 1.0 Software (Clontech).
Other Procedures-Confocal laser microscopy was performed as described previously (21). 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.

mPGES-1 Is Expressed in Human Colorectal
Cancer Tissues-Surgically resected human colorectal adenocarcinoma (Fig. 1, A and B) and adenoma ( Fig. 1, C and D) tissues were fixed with formaldehyde, embedded in paraffin, and serial sections were immunostained with antibodies against mPGES-1 ( Fig. 1, A, C, and E) and COX-2 ( Fig. 1, B, D, and F). In the sections shown in Fig. 1, virtually all adenocarcinoma cells (panels A and B) and adenoma cells (panels C and D) were positively stained with anti-mPGES-1 and anti-COX-2 antibodies. Signal for mPGES-1 was distributed throughout the cytoplasm with a punctate pattern, whereas that of COX-2 was enriched in the luminal side of the nuclei. Considering that mPGES-1 is an integral membrane protein, it is likely that the observed mPGES-1 immunostaining reflects its distribution in the endoplasmic reticulum membrane. Immunoreactive signals for mPGES-1 and COX-2 were barely seen in normal colon sections (Fig. 1, E and F). Of several colon cancer tissue sections examined, 9 of 9 adenocarcinoma tissues and 5 of 7 adenoma tissues were positive for mPGES-1 immunoreactivity. Furthermore, 5 of 9 mPGES-1-positive adenocarcinoma tissues and 4 of 5 mPGES-1-positive adenoma tissues were also COX-2-positive.
To examine the role of PGE 2 produced by the COX-2/ mPGES-1 pathway in the growth of HCA-7 cells, we first examined the effects of NS-398, a well known COX-2 inhibitor (38), and MK-886, an mPGES-1 inhibitor that also inhibits other MAPEG proteins such as FLAP and LTC 4 synthase (22), on cell growth and PGE 2 production. Treatment of HCA-7 cells with NS-398 almost completely abolished PGE 2 production, accompanied by ϳ40% reduction of cell growth (Fig. 2D). When HCA-7 cells were treated with MK-886 at a concentration completely inhibiting mPGES-1 enzymatic activity in vitro (data not shown), there was a ϳ40% reduction in cell growth (as in the case of NS-398 treatment), even though the inhibition of PGE 2 production was only partial (ϳ60%) (Fig. 2E).
To assess the contribution of mPGES-1 to cell growth and PGE 2 production in a more comprehensive way, we used an antisense oligonucleotide for mPGES-1 to reduce its expression. Treating the cells with an mPGES-1-specific antisense oligonucleotide resulted in marked reduction of mPGES-1 protein expression with no appreciable change in COX-2 expression (Fig. 3A). In this setting, both cell growth (Fig. 3B) and PGE 2 production (Fig. 3C) were reduced partially, as observed in the experiments with MK-886 (Fig. 2E). The production of PGF 2␣ was unaffected by antisense treatment (Fig. 3D), verifying that the antisense acted specifically on mPGES-1 but not on upstream enzymes or other terminal enzymes. Control oligonucleotide did not affect mPGES-1 expression, PGE 2 production, or cell growth (not shown). These results suggest that PGE 2 produced via the COX-2/mPGES-1 pathway is partially involved in the proliferation of HCA-7 cells.
We next examined if, conversely, overexpression of mPGES-1 would facilitate the growth of HCA-7 cells. To this end, C-terminal FLAG-tagged mPGES-1 cDNA was transfected into HCA-7 cells by lentivirus-mediated gene transfer. As shown in Fig. 4A, the virus-infected cells expressed FLAGtagged mPGES-1 protein just above the position of endogenous mPGES-1, whereas the constitutive expression of the upstream enzymes cPLA 2␣ and COX-2 was unaltered. PGES activity in vitro in the membrane fraction of cell lysates (Fig. 4B) and PGE 2 release into medium during culture (Fig. 4C) were markedly increased in mPGES-1-transfected cells relative to control cells. Furthermore, the cell growth rate of mPGES-1transfected cells was significantly faster than that of control cells (Fig. 4D).

FIG. 1. Immunohistochemical detection of mPGES-1 and COX-2 in human colorectal tumor tissues.
Paraffin-embedded sections of human colorectal adenocarcinoma (A and B) and adenoma (C and D) tissues as well as normal tissues (E and F) were subjected to immunostaining with anti-mPGES-1 (A, C, and E) and anti-COX-2 (B, D, and F) antibodies. After immunostaining, the sections were counterstained with hematoxylin and eosin. A, B and E, ϫ400; C, D, and F, ϫ200.

mPGES-1 and Tumorigenesis
Transformation of HEK293 Cells by Overexpression of mPGES-1-As shown in Fig. 5A, HEK293 cells cotransfected with COX-2 and mPGES-1 grew more rapidly than parental cells over the entire culture periods. COX-2/mPGES-1 cotransfection into HEK293 cells led to cell aggregation, rounding and piling up, and both enzymes were colocalized in the perinuclear region in the aggregated cells (Fig. 5B). These morphological changes were less pronounced in replicate COX-2/mPGES-1cotransfected cells cultured in the continued presence of NS-398, added immediately after transfection, or in cells cotransfected with COX-2 and mPGES-1-R110S, which has very weak enzyme activity (21) (Fig. 5B). These results suggest that the catalytic functions of COX-2 and mPGES-1 are both required for triggering cellular transformation. However, addition of NS-398 to already transformed COX-2/mPGES-1-expressing clones failed to reverse their growth and aggregated morphology (not shown), indicating that transformation of HEK293 cells by COX-2/mPGES-1 cotransfection is an irreversible event.
As anchorage-independent growth is considered to be an in vitro test for tumorigenesis, we examined the growth of COX-2/mPGES-1-cotransfected HEK293 cells in a semisoft agar medium. As demonstrated in Fig. 6A, the COX-2/mPGES-1-coexpressing cells exhibited marked anchorage-independent growth, as manifested by the appearance of a number of large colonies. On the other hand, cells expressing COX-2 alone or cells coexpressing COX-1 and mPGES-1 formed fewer and smaller colonies, and parental cells did not grow appreciably in soft agar (Fig. 6A). Quantification of the numbers and sizes of colonies formed in this colony assay is summarized in Fig. 6B.
When parental and COX-2/mPGES-1-expressing HEK293 cells, as well as HCA-7 cells used as a positive control, were injected subcutaneously into athymic nude mice, COX-2/ mPGES-1-expressing HEK293 cells and HCA-7 cells, but not parental HEK293 cells, formed large solid tumors after 3 months (Fig. 7A). Histopathologic examination of a fraction of whole tumor tissues from COX-2/mPGES-1-expressing HEK293 cell xenografts is shown in Fig. 7B. At the site of implantation, a whitish nodular tumor was formed in the subcutaneous tissue and exhibited a well circumscribed mass (Fig.  7B, panel a). The tumor cells had swollen nuclei of round or polygonal shape with sporadic mitosis, and their cytoplasm was scare and chromophobic (Fig. 7B, panel b). These characteristic features of the tumor suggest its malignancy. The tumor cells

mPGES-1 and Tumorigenesis
were proliferating with scanty interstitium that was mainly composed of capillaries (Fig. 7B, panel c, arrow). These capillaries consisted of swollen endothelial cells, and were likely to be newly formed vessels. Immunohistochemistry for mPGES-1 (Fig. 7B, panel d) and COX-2 (Fig. 7B, panel e) revealed the location of both enzymes around the nuclei in tumor cells.
The HCA-7 xenograft tumor appeared as a nodular mass and was well demarcated (Fig. 7C, panel a). Tumor with a thin fibrous capsule was visible in the subcutaneous region (Fig. 7C,  panel b, arrow). The tumor cells had scare cytoplasm and hyperchromatic nuclei of various sizes with sporadic mitosis, and were partially proliferating with tubular formation. These features indicated adenocarcinoma differentiation (Fig. 7C, panel c).

FIG. 4. Effects of mPGES-1 overexpression on HCA-7 cells.
A, expression of mPGES-1 and other enzymes related to PGE 2 biosynthesis (COX-1 and -2, cPGES and cPLA 2 ␣) in HCA-7 cells with (mPGES-1) or without (parent) lentivirusmediated transfer of FLAG-mPGES-1, as assessed by immunoblotting using their respective antibodies. The protein band for overexpressed FLAG-mPGES-1 was detected just above the band for endogenous mPGES-1. Equal protein (20 g protein equivalent)) from the cell lysates were subjected to Western blotting at the same time with the corresponding antibody to allow for a direct comparison. B-D, effects of mPGES-1 overexpression on PGES activity in the membrane fraction of cell lysates (B), PGE 2 production (C), and cell growth (D) after culturing for 4 days (n ϭ 3, *, p Ͻ 0.05).

FIG. 5. Effects of co-overexpression of COX-2 and mPGES-1 on cell growth and morphology of HEK293 cells.
A, growth acceleration of HEK293 cells by COX-2 and mPGES-1 co-overexpression. HEK293 cells stably expressing COX-2 and mPGES-1 and parental cells were seeded at 10 4 cells/ml and cultured for the indicated periods. Cell numbers were counted at each time point. B, confocal microscopic analysis of subcellular localization of COX-2 (green) and mPGES-1 (red) in the COX-2/mPGES-1cotransfected cell aggregates. Yellow color (merge) indicates the co-localization of both enzymes. Fluorescent signals were overlaid on phase-contrast views. C, morphology of HEK293 cells transfected with COX-2 alone, cotransfected with COX-2 and mPGES-1, cotransfected with COX-2 and mPGES-1-R110S mutant, and cotransfected with COX-2 and mPGES-1 and cultured in the presence of 100 ng/ml NS-398. (Fig. 7C, panel e) were detected in the tumor regions.

mPGES-1 and Tumorigenesis
Identification of a Panel of Genes Altered by Overexpression of mPGES-1-To further elucidate the mechanisms of cellular transformation by co-overexpression of COX-2/mPGES-1, we surveyed a panel of genes, the expression of which was significantly altered in COX-2/mPGES-1-cotransfected HEK293 cells relative to cells expressing COX-2 alone, by cDNA array analysis. As summarized in Table I, increased genes included those required for cytoskeletal regulation (the small G protein RhoA), cell growth (the receptor tyrosine kinases ErbB3 and Flt1, the cyclin-dependent kinase CDK5, and the tumor necrosis factor signaling molecule TRAF-1), gene transcription (c-Myc, GATA4, and YL-1), protein synthesis (the ribosomal proteins S19 and S3A), and so on. Conversely, decreased genes included the ␣ subunit of protein phosphatase 1, cytoskeleton regulators (tubulin ␤ and ezrin), cell adhesion regulators (several integrins and ␣ 1 -catenin), the transcription factor Egr-1, and thymosins (Table I).
We then performed Northern and Western blot analyses to verify that the expression of these genes identified by cDNA array indeed differ between COX-2-and COX-2/mPGES-1transfected HEK293 cells and between parental and mPGES-1-transfected HCA-7 cells. A representative result of the blotting analyses is shown in Fig. 8. In line with the cDNA array analysis, the expression levels of rhoA, c-myc, and ErbB3 were markedly increased, whereas those of ezrin and Egr-1 were decreased, in COX-2/mPGES-1-cotransfected cells relative to COX-2-transfected cells (Fig. 8A). Moreover, increased expression of rhoA, c-myc, and ErbB3 was also observed in mPGES-1-transfected HCA-7 cells as compared with parental cells (Fig. 8B). DISCUSSION Involvement of COX-2 and its main product, PGE 2 , in cancer development has been well documented by several studies. There is a reduction in the relative risk of colorectal cancer in individuals taking non-steroidal anti-inflammatory drugs (2,3). COX-2 levels are increased in ϳ90% of human colorectal cancers and ϳ50% of premalignant adenomas (1). COX-2 expression in colorectal carcinoma cells provides a growth and survival advantage and increases tumor invasiveness, and PGE 2 increases growth and motility of colorectal carcinoma cells through the EP4 receptor signaling pathway (8 -12). Genetic disruption of cPLA 2 ␣ (17, 18), COX-2 (13), or the EP2 receptor (14) results in decreased number and/or neoplastic growth of colorectal polyps in APC mutant mice. In addition to colorectal cancer, transgenic overexpression of COX-2 in mouse mammary gland results in spontaneous development of mammary tumor (10). Similarly, transgenic mice with keratin 5 promoter-driven COX-2 overexpression in basal epidermal cells exhibit a preneoplastic skin phenotype (39). PGE 2 regulates COX-2-dependent, CD44-and matrix metalloproteinase-2-mediated invasion in non-small cell lung cancer in an autocrine/ paracrine manner via EP4 receptor signaling (40).
Given these observations, the aim of this study was to evaluate the potential contribution of mPGES-1, which lies downstream of the COX-2-dependent PGE 2 -biosynthetic pathway, to tumorigenesis. Elevated expression of mPGES-1 in human cancers has recently been demonstrated in non-small cell lung cancer (32) and endometrial adenocarcinoma (33). In this study, we performed immunohistochemistry of human colon cancer and adenoma tissues with anti-COX-2 and anti-mPGES-1 antibodies and found that both enzymes are coex-

mPGES-1 and Tumorigenesis
pressed in many, even if not all, of malignant and benign colorectal tumor cells (Fig. 1). While this study was underway, Yoshimatsu et al. (34) reported that mPGES-1 is overexpressed in Ͼ80% of human colorectal tumors and adenomas, in line with our present observation.
To assess the role of mPGES-1 in growth of colon cancer cells in cell culture, we took advantage of HCA-7 cells, a human colon cancer cell line that has often been used to investigate COX-2-dependent tumorigenesis (4,36,37). Proliferation of HCA-7 cells is reduced by COX-2 inhibitors, and this is reversed by exogenous PGE 2 (4,36). PGE 2 also prevents COX-2 inhibitor-induced apoptosis by inducing expression of Bcl-2, and increases clonogenicity in HCA-7 cells (37). In this study, we found that both mPGES-1 and COX-2 are constitutively, but not inducibly, expressed in HCA-7 cells (Fig. 2A). Although the transcriptional regulations of COX-2 and mPGES-1 are not entirely identical (24,31), recent observations that the expression of COX-2 in HCA-7 cells is transcriptionally regulated by the NF-IL-6-regulatory element (41) and that mPGES-1 induction in lipopolysaccharide-stimulated mouse macrophages is ablated in NF-IL-6-deficient mice (29) suggest a role for this transcription factor in the expression of both COX-2 and mPGES-1 in HCA-7 cells.
Importantly, studies using MK-886, an inhibitor for MAPEG proteins including mPGES-1 (Fig. 2, C and D), an mPGES-1specific antisense oligonucleotide (Fig. 3) and overexpression of mPGES-1 (Fig. 4) together suggest that mPGES-1 is involved in PGE 2 production and proliferation of HCA-7 cells, even if partially. Somewhat paradoxically, under conditions where the COX-2 inhibitor NS-398 almost completely abolished PGE 2 production, reduction of cell growth by NS-398 was still only partial, whereas reduction of PGE 2 production and cell growth by MK-886 or the mPGES-1 antisense was partial and parallel ( Figs. 2 and 3). A likely explanation for these observations is that the PGE 2 produced via the COX-2/mPGES-1 pathway contributes to cell growth, whereas there is an additional COX-

TABLE I Genes that are increased or decreased by overexpression of mPGES-1 in COX-2-expressing HEK293 cells
Numbers indicate -fold changes in the expression of genes in COX-2/mPGES-1 cotransfectants relative to cells expressing COX-2 alone. A representative result of three independent experiments is shown.  (16), the PGE 2 produced by COX-2/mPGES-1 in the perinuclear region may be preferentially presented to this G-protein-coupled PGE 2 receptor through an unknown mechanism. Alternatively, the perinuclear PGE 2 (or possibly other unknown metabolites) produced by COX-2/mPGES-1 may act on certain nuclear receptors that in turn promote cell growth. Indeed, several eicosanoids have been shown to stimulate the peroxisome proliferator-activated receptor family of nuclear receptors (43,44). Nonetheless, different sensitivity of PGE 2 production by HCA-7 cells to COX-2 and mPGES-1 inhibitors and mPGES-1 antisense oligonucleotide strongly argues that this cell line may contain an alternative COX-2-dependent PGE 2biosynthetic route that involves other PGES enzyme(s). Although cPGES is abundantly expressed in HCA-7 cells ( Fig.  2A), this enzyme has been reported to be coupled rather specifically with COX-1, not COX-2 (19). mPGES-2, a recently identified enzyme whose transcript is detected in human colon (22), may represent a second PGES that can be coupled with COX-2, a possibility that is now under investigation. In another model using HEK293 cells, cotransfection of COX-2 and mPGES-1 led to cellular transformation, as demonstrated by rapid proliferation, morphological change, piling up, and aggregation in normal culture, large colony formation in soft agar culture, and formation of solid tumors in nude mice . Immunohistochemical examination of the tumor revealed tissue invasion by COX-2/mPGES-1-derived tumor cells as well as angiogenesis (Fig. 7), consistent with the angiogenic effect of COX-2 and PGE 2 (11,45,46). Growth promotion and morphological change of COX-2/mPGES-1-transfected HEK293 cells in culture were less pronounced if NS-398 was added immediately after COX-2/mPGES-1 transfection, or if a catalytically inactive mPGES-1 mutant was transfected in place of native enzyme (Fig. 5C). Once transformed, however, neither NS-398 nor MK-886 reversed the growth and aberrant morphology of COX-2/mPGES-1-cotransfected HEK293 cells. These observations indicate that, even though the COX-2/ mPGES-1 catalytic product triggers cellular transformation, it may not be required for subsequent maintenance of a transformed phenotype in this setting. A similar event has been observed in ECV endothelial cells transfected with COX-1 (not COX-2, which induced apoptosis in these cells), where aggressive growth of COX-1-transfected ECV cells was no longer inhibited by indomethacin (47). Unlike HCA-7 cells (4,36), simple addition of exogenous PGE 2 to HEK293 cells did not induce cellular transformation, as mentioned previously (21). The following possibilities could be considered: (i) continuous production of high levels of PGE 2 around the perinuclear area by COX-2/mPGES-1 is critical for inducing transformation; (ii) some additional components, which act cooperatively with PGE 2 to induce cellular transformation, are induced by COX-2/mPGES-1 overexpression; and (iii) some unknown substances produced by COX-2/mPGES-1 may be involved in transformation.
To gain insights into mPGES-1-promoted cellular transformation, we sought to identify mPGES-1-regulated genes by cDNA array technology (Table I and Fig. 8). The genes identified so far can be categorized into several groups: (i) genes for signaling molecules related to cell proliferation and differentiation; (ii) genes for transcription factors; (iii) genes related to cytoskeletal regulation; (iv) cell adhesion molecules; and (v) genes with unknown functions. Some of the induced genes are proto-oncogenes that have the capacity to promote cellular transformation when transfected alone into cells.
Increased genes related to cell growth and differentiation include those encoding the receptor tyrosine kinases ErbB3 and Flt1, the cyclin-dependent protein kinase CDK5, the ribosomal proteins S3A and S19, the proliferation-associated nucleolar protein NOL1, and TRAF1. ErbB3, a ligand for heregulin, is a member of the epidermal growth factor receptor family and its overexpression has been frequently found in human tumors (48). Induced expression of ErbB3 is in line with a recent report that PGE 2 transactivates epidermal growth factor receptor, thereby switching on the mitogenic signaling pathway in gastric epithelial and colon cancer cell lines (49). Moreover, COX-2 is overexpressed in hereglin 2-positive breast cancer (50). Induction of Flt1, a receptor for vascular endothelial cell growth factor (51,52), may be linked to angiogenesis, which has been associated with COX-2-and PGE 2 -dependent tumor development (11,45,46).
Cyclin-dependent protein kinases generally play crucial roles in cell cycle progression (53). Apart from cell cycle control, CDK5 phosphorylates a diverse list of substrates and regulates a range of cellular processes, including cell adhesion and motility (54,55). Ribosomal proteins are integral components of the basal cellular machinery involved in protein synthesis and have been found to play roles in regulating cell growth and FIG. 8. Altered expression of several genes in COX-2/mPGES-1-cotransfected HEK293 cells and mPGES-1transfected HCA-7 cells. A, comparison between COX-2-expressing and COX-2/ mPGES-1-coexpressing HEK293 cells. ErbB3, Egr-1, and ezrin were detected by Western blotting and c-myc and rhoA by Northern blotting. The Northern membrane was reprobed with a glyceraldehyde-3-phosphate dehydrogenase (GAPDH) probe to confirm equal loading of samples into each lane. B, comparison between parental and mPGES-1-transfected HCA-7 cells. The results of ErbB3 (Western blotting), c-myc and rhoA (Northern blotting) are shown. For Northern blot analysis, ribosomal RNA (rRNA), which was visualized with ethidium bromide, is also shown.

mPGES-1 and Tumorigenesis
transformation (56,57). NOL1 expression is associated with cell proliferation during G 1 -S phases and represents a biological marker indicative of tumor aggressiveness, particularly the late events of colorectal tumor progression (58,59). TRAF1 is a signal transducer of the tumor necrosis factor receptor family, which has been implicated in cell differentiation (60). Conversely, decreased expression of protein phosphatase 1, a tumor suppressor of which mutations have been found in lymphoma and hepatoma (61,62), may also account for aggressive cell growth of COX-2/mPGES-1 cotransfectants.
Transcription factors whose expressions are altered by COX-2/mPGES-1 include c-Myc, GATA4, and YL-1 (increased) as well as Egr-1 (decreased). Induced expression of c-Myc, a helixloop-helix transcription factor, is of great importance in controlling cell growth and vitality (63), and is commonly amplified in many human tumors (64). Involvement of c-Myc in colon cancer development downstream of the Wnt/APC/␤-catenin signaling pathway has been documented (65,66). Moreover, the activation of the Wnt/APC/␤-catenin signaling pathway results in transcriptional up-regulation of COX-2 in cancer cells (67). GATA-4, a member of the GATA transcription factor family, plays a role in the regulation of cell migration and its overexpression has been found in several types of cancer (68). YL-1, also reported as transcription factor-like 1, has been implicated in anchorage-independent cell growth (69). Egr-1, a COX-2/ mPGES-1-decreased gene, is an inducible transcription factor that binds to GC-rich elements and plays a crucial role in transcriptional activation of the mPGES-1 gene (31). Hence, the reduction of Egr-1 expression in COX-2/mPGES-1 cotransfectants may be a reflection of negative feedback regulation of mPGES-1 expression.
Altered genes encoding proteins for cytoskeletal regulation include RhoA (increased), ezrin, tubulin, and annexins (decreased), which may be responsible for the marked morphological change in COX-2/mPGES-1 cotransfectants. The small G protein RhoA regulates various aspects of actin filament rearrangement and has a key role in growth of tumors (70,71). Ezrin plays structural and regulatory roles in the assembly and stabilization of specialized plasma membrane domains, particularly in surface projections such as microvilli and membrane ruffles where it links the microfilaments to the plasma membrane (72). Annexins interact with cytoskeletal proteins and have been implicated in diverse cellular responses including differentiation and membrane fusion (73). Because annexins are capable of inhibiting cPLA 2 ␣ (74), decreased expression of annexins may lead to increased activation of cPLA 2 ␣, thereby amplifying the COX-2/mPGES-1-dependent PGE 2 biosynthesis. Indeed, cPLA 2 ␣ has often been associated with cellular transformation (75) and cPLA 2 ␣ knockout decreases the incidence of polyposis in Apc mutant mice (17,18). There were decreases in the expression of several cell adhesion molecules, such as integrins and ␣ 1 -catenin, which are critical components for focal adhesion (76,77). Thus, decreased expression of these focal adhesion proteins is consistent with the reduced anchorage dependence of COX-2/mPGES-1 cotransfectants.
The most remarkably decreased genes in COX-2/mPGES-1 cotransfectants are those for thymosins. Prothymosin ␣ is a small highly acidic protein found in the nuclei of virtually all mammalian tissues, and its high conservation in mammals and wide tissue distribution suggest an essential biological role (42). Although the exact mechanism of action of thymosins remains elusive, the one constant has been their relationship with the proliferating state of the cell and its requirement for cellular growth and survival. Overall, although the mechanisms by which overexpression of COX-2/mPGES-1 alters the expression of these genes, and whether these genes, alone or in combination, induce transformation of HEK293 cells, are currently unclear, the present findings shed light on unexplored aspects of the combined action of COX-2 and mPGES-1 in tumorigenesis. Indeed, increases in some of the genes identified in COX-2/mPGES-1-transfected HEK293 cells were also observed in mPGES-1-transfected HCA-7 cells (Fig. 8B), suggesting that our findings could be applicable to at least some types of colorectal cancer. Evaluation of the effect of COX-2/ mPGES in non-transformed intestinal cell lines will give further insight into this critical issue.
In summary, this study demonstrates that mPGES-1, in concert with COX-2, can be associated with cellular transformation and cancer development. Future studies using mPGES-1 knockout mice or mPGES-1-specific inhibitors would open further insights into the role of this critical PGE 2 biosynthetic terminal enzyme in tumorigenesis. Importantly, our observation that mPGES-1 is overexpressed in colorectal tumors provides the basis for future studies that will evaluate whether mPGES-1 is a bona fide therapeutic target.