Genetic-deletion of Cyclooxygenase-2 Downstream Prostacyclin Synthase Suppresses Inflammatory Reactions but Facilitates Carcinogenesis, unlike Deletion of Microsomal Prostaglandin E Synthase-1

Prostacyclin synthase (PGIS) and microsomal prostaglandin E synthase-1 (mPGES-1) are prostaglandin (PG) terminal synthases that function downstream of inducible cyclooxygenase (COX)-2 in the PGI2 and PGE2 biosynthetic pathways, respectively. mPGES-1 has been shown to be involved in various COX-2-related diseases such as inflammatory diseases and cancers, but it is not yet known how PGIS is involved in these COX-2-related diseases. Here, to clarify the pathophysiological role of PGIS, we investigated the phenotypes of PGIS and mPGES-1 individual knockout (KO) or double KO (DKO) mice. The results indicate that a thioglycollate-induced exudation of leukocytes into the peritoneal cavity was suppressed by the genetic-deletion of PGIS. In the PGIS KO mice, lipopolysaccharide-primed pain nociception (as assessed by the acetic acid-induced writhing reaction) was also reduced. Both of these reactions were suppressed more effectively in the PGIS/mPGES-1 DKO mice than in the PGIS KO mice. On the other hand, unlike mPGES-1 deficiency (which suppressed azoxymethane-induced colon carcinogenesis), PGIS deficiency up-regulated both aberrant crypt foci formation at the early stage of carcinogenesis and polyp formation at the late stage. These results indicate that PGIS and mPGES-1 cooperatively exacerbate inflammatory reactions but have opposing effects on carcinogenesis, and that PGIS-derived PGI2 has anti-carcinogenic effects.


Results
For the determination of whether PGIS and mPGES-1 were indeed knocked out in the null mice, the protein levels in thioglycollate-induced peritoneal macrophages (MΦ s) prepared from the four genotypes of mice were analyzed by Western blotting (Fig. 1A). In the control and PGIS-deficient MΦ s, mPGES-1 protein was detected in normal culture conditions as was COX-2, and mPGES-1 was up-regulated by the lipopolysaccharide (LPS) treatment. On the other hand, PGIS protein was detected in the control and mPGES-1-deficient MΦ s. Neither mPGES-1 nor PGIS was detected in MΦ s prepared from PGIS/ mPGES-1 DKO mice. Since it has been reported that LPS-stimulated PGE 2 production was markedly suppressed and the production of other prostanoids was increased in MΦ s derived from mPGES-1 KO mice relative to those derived from wild-type (WT) mice 9,16,25,26 , we further measured prostanoids in culture medium from four genotypes of MΦ s. As reported previously, mPGES-1 deficiency decreased the PGE 2 levels but conversely increased the levels of PGs other than PGE 2 including 6-ketoPGF 1α , a PGI 2 metabolite, in culture medium (Fig. 1B), indicating that COX-2-derived PGH 2 is metabolically shunted into the other prostanoid synthetic pathway in mPGES-1-deficient MΦ s.
In the DKO MΦ s, a more marked shunting reaction into the other prostanoids was observed. These shunting phenomena were not observed in PGIS-deficient MΦ s, in whose culture medium the levels of prostanoids including PGE 2 were similar to those in the control MΦ s. These results suggested that the intracellular shunting of COX-2-derived PGH 2 might have a single direction in mouse peritoneal MΦ s. In MΦ s, mPGES-1 might be not able to metabolize PGH 2 , which should be supplied to PGIS from COX-2.
It has been shown that PGIS and mPGES-1 are expressed in mouse kidney and lung even under basal conditions 9,16,18,19,27 . We next measured the prostanoid levels in kidneys and lungs prepared from the four genotypes of mice and found that these two organs also showed a shunting reaction similar to that of the MΦ s ( Fig. 2A). Metabolic shunting of PGH 2 was observed in the mPGES-1-deficient mice but not in PGIS-deficient mice. We further found that the DKO mice developed renal disorders with arterial sclerosis and hypertrophy of vessels (Fig. 2B), as did the PGIS single-KO mice 28 . These results suggested that renal disorders observed in PGIS-deficient mice might be induced by a breakdown of PGI 2 levels but not by shunting into prostanoids other than PGI 2 .
PGIS is involved in inflammatory reactions. Studies using KO mice have revealed that IP and PGE 2 receptor subtypes EP2 and EP4 are involved in inflammatory reactions including swelling and pain [21][22][23][24] . Thus, to examine the effects of PGIS and mPGES-1 deficiency on inflammatory reaction, we next counted the number of exudate leukocytes and measured the levels of prostanoids in the peritoneal fluids prepared from the four genotypes of thioglycollate-treated mice. As shown in Fig. 3, in the control mice, the peritoneal leukocyte number was increased steadily from day 2 to 4 after the injection of thioglycollate, accompanied by an increase in 6-ketoPGF 1α levels. PGIS deficiency did not significantly affect the number of exudate leukocytes on day 2, but significantly decreased their number on day 4 compared to the control mice. Our morphological analysis revealed that among leukocytes, the exudation of MΦ s was especially suppressed by PGIS deficiency (Supplementary Table 1). In the PGIS KO mice, 6-ketoPGF 1α was not detected in the peritoneal fluid. Unlike the PGIS-deficient MΦ s, PGE 2 and the other prostanoid levels in the PGIS KO mice were higher than those in the control mice. These results indicated that PGIS-derived PGI 2 might play an important role in the exudation of MΦ s at the late stage of inflammatory reaction.
Moreover, the migration of MΦ s into the peritoneal fluid was not affected by mPGES-1 deficiency but was suppressed more effectively in the PGIS/mPGES-1 DKO mice than in the PGIS KO mice. mPGES-1-derived PGE 2 might contribute to the exudation of MΦ s, but its contribution might be smaller than that of PGIS-derived PGI 2 .
We further investigated the effect of PGIS deficiency on inflammatory pain hypersensitivity, as assessed by the LPS-primed acetic acid-induced writhing reaction. The LPS pretreatment induced the expressions of COX-2 and mPGES-1 and then enhanced the writhing reaction. As we previously reported 9 , the injection of acetic acid into the peritoneum of mice induced a stretching behavior, that peaked at 5-10 min and then declined gradually over 30 min. The writhing reaction was reduced in both the PGIS KO and mPGES-1 KO mice, and it was suppressed more effectively in the PGIS/mPGES-1 DKO mice compared to the PGIS KO and mPGES-1 KO mice (Fig. 4A). In the PGIS KO mice, 6-ketoPGF 1α was not observed but the PGE 2 levels were markedly increased compared to the control mice, although a similar shunting phenomenon was not observed in the mPGES-1 KO mice (Fig. 4B). In the PGIS/mPGES-1 DKO mice, the levels of both 6-ketoPGF 1α and PGE 2 were reduced. These results indicate that PGIS-derived PGI 2 facilitates inflammatory pain hypersensitivity in a coordinated manner with mPGES-1-derived PGE 2 .
PGIS is involved in chemically-induced carcinogenesis. We next injected azoxymethane (AOM) intraperitoneally into these four genotypes of mice once a week for 6 weeks to induce colon carcinogenesis. To examine the involvement of PGIS in an early phase of carcinogenesis, we first killed animals 6 weeks after the last injection, and evaluated the preneoplastic aberrant crypt foci (ACF) formation (Fig. 5A). The results indicated that PGIS deficiency did not induce spontaneous colon carcinogenesis but significantly increased the number of ACF, whereas mPGES-1 deficiency decreased the ACF number, as described previously 11 . In the PGIS/mPGES-1 DKO mice, the ACF number was similar to that in the control mice. These results indicate that the genetic-deletion of PGIS facilitates tumor propagation even though it is not sufficient for tumor initiation.
At 20 weeks after the last injection, the AOM administration induced the development of multiple tumors in the control colons, whereas mPGES-1 deficiency suppressed the colon carcinogenesis as described previously 11 . In contrast to mPGES-1 deficiency, PGIS deficiency tended to increase the polyp numbers (Fig. 5B). In addition, the number of large polyps was significantly increased in the PGIS KO mice. As shown in Fig. 5C, histologically, adenocarcinomas were observed in the control, PGIS KO and DKO mice, but only adenomas were observed in the mPGES-1 KO mice. Among them, the adenocarcinomas in the PGIS KO mice were considerably larger than those in the control mice. Balb/c background mice were used for these analyses, because it was shown that Balb/c mice are more sensitive to AOM-induced colon carcinogenesis than other strain mice 29 . We also found that the number of colon polyps (and the number of large polyps in particular) was increased in the PGIS KO mice compared to the wild type (WT) mice even in the C57BL/6 mouse strain, which has been well characterized as being highly resistant to colon tumor induction by AOM 29 (Fig. 5D). These results indicated that the genetic-deletion of PGIS exacerbates chemically induced colon carcinogenesis.   We next analyzed the levels of prostanoids in these colon polyp tissues (Fig. 6A). As expected, in polyp tissues, mPGES-1 deficiency decreased the PGE 2 levels, and 6-ketoPGF 1α was not detected in the PGIS KO mice or DKO mice. The levels of PGs other than PGE 2 including 6-ketoPGF 1α in the polyp tissues of the mPGES-1 KO mice were higher than those in the control mice, but a similar shunting phenomenon was not observed in the colon polyp tissues of the PGIS KO mice. In the PGIS KO mice, the level of PGE 2 was similar to that in the control mice. We analyzed the expression levels of COX-2, PGIS and mPGES-1 in colon tissues by quantitative RT-PCR. As shown in Fig. 6B, the levels of both COX-2 and mPGES-1 mRNA in polyp tissues were substantially higher than those in normal tissues of the colon. On the other hand, the PGIS mRNA level in the polyps was similar to that in the normal tissues. In our immunohistochemical analysis, the positive immunostaining signal of PGIS was observed only in blood vessels, not in tumor cells or tumor stromal cells (Fig. 6C). These results suggested that PGIS in host-derived vascular cells might be involved in carcinogenesis and that the breakdown of PGI 2 in host-derived vascular cells might lead to an exacerbation of colon tumors in PGIS KO mice.

Discussion
We here established mice that are doubly deficient for PGIS and mPGES-1, both of which are preferentially coupled with COX-2 as their upstream enzymes, and we then investigated the phenotypes of these mice. PGIS/mPGES-1 DKO mice were born. However as well as COX-2 KO mice 30 , when heterozygous KO mice were intercrossed, the number of DKO mice was slightly less than the expected Mendelian The numbers of total polyps and large (> 2 mm) polyps per mouse in the colon tissues are shown. Results are mean ± SEM (n = 6-9). *P < 0.05 and **P < 0.01 vs. control. ratio. We also observed the development of renal disorders in the DKO mice (Fig. 2B) as well as in the COX-2 KO 30 and PGIS KO mice 28 . These results suggested that renal disorders might be induced by a breakdown of COX-2/PGIS-derived PGI 2 levels.
The inflammatory reactions and inflammatory pain hypersensitivity were significantly suppressed in the PGIS KO mice, and they tended to be suppressed more effectively in the PGIS/mPGES-1 DKO mice than in the PGIS KO mice (Figs 3A and 4A). It has been shown that IP as well as EP receptors and mPGES-1 are involved in inflammatory reactions including swelling and pain [21][22][23][24] . The present results indicated that PGIS-derived PGI 2 acts on IP at inflamed sites and exacerbates inflammation together with mPGES-1-derived PGE 2 . It is noteworthy that PGIS deficiency increased the PGE 2 levels in inflammatory exudates but mPGES-1 deficiency did not affect the production of PGs other than PGE 2 in inflammatory exudates (Figs 3B and 4B).
The shunting pattern of PGH 2 observed in the exudates from KO mice was different from that in KO MΦ s (Fig. 1B). At inflamed sites, as well as MΦ s, several types of cells (including other leukocytes, vascular cells and stromal cells) are able to produce PGs. In the present case, MΦ s might contribute less to PG production. Otherwise, the intracellular transport of PGH 2 from PGIS-deficient MΦ s to the other inflammatory cells might occur. For the development of inhibitors specific for each PG terminal synthase as anti-inflammatory drugs, it is necessary to determine which types of cells are involved in the process of the target illnesses. Here, we found that mPGES-1 deficiency suppressed the acetic acid-induced pain response but did not affect the thioglycollate-induced leukocyte exudation (Figs 3A and 4A). Unlike COX-2, mPGES-1 might have only limited involvement in certain types of inflammatory reactions.
As shown in Fig. 5, the deletions of PGIS and mPGES-1 showed opposite effects on the chemically-induced colon carcinogenesis. mPGES-1 deficiency suppressed the AOM-induced ACF and polyp formation, but PGIS deficiency enhanced both of them. Keith et al. reported that in a smoke-exposure model, pulmonary-specific PGIS-overexpressing mice were chemoprotected from developing lung tumors 31 . Together these findings indicate that PGIS-derived PGI 2 functions as an anti-carcinogenic agent in several types of cancers. In colon tumor tissues, PGIS was expressed in blood vessels (Fig. 6C) and mPGES-1 was expressed in tumor cells and tumor stromal cells 11 . When tumor and host-associated COX-2/mPGES-1-derived PGE 2 increase and then exceed the anti-carcinogenic actions of vascular COX-2/PGIS-derived PGI 2 , a tumor might begin to progress. As described previously 11 , the ablation of mPGES-1 resulted not only in the suppression of carcinogenic PGE 2 production but also in the enhancement of anti-carcinogenic PGI 2 production. An mPGES-1-specific inhibitor is expected to be a more effective anti-carcinogenic agent than a COX-2-specific inhibitor.
A variable-number tandem repeat polymorphism was detected in the promoter region of human PGIS gene that is associated with promoter activity 32 . Poole et al. reported that the PGIS promoter polymorphism might affect the risk of the development of colorectal polyps 33 . They showed that having fewer than six repeats on both PGIS alleles was associated with an increased risk of adenoma development compared with the WT genotype, although having more than six repeats reduced the risk. The PGIS promoter polymorphism may be a predictive marker for the risk of colorectal polyps. It is also noteworthy that IP deficiency did not affect the AOM-induced colonic ACF formation 34 . These results suggested that PGI 2 might suppress colon carcinogenesis through an unknown receptor other than IP.
A candidate target for the anti-carcinogenic action of PGI 2 is peroxisome proliferator-activated receptor δ (PPARδ ). Gupta et al. showed that endogenously synthesized PGI 2 could serve as a ligand for PPARδ 35 . It was also reported that PPARδ deficiency enhanced AOM-induced polyp formation 36 . PGIS-derived PGI 2 might act on PPARδ and then suppress carcinogenesis. On the other hand, Zuo et al. reported that PPARδ deficiency conversely inhibited colon tumorigenesis in mice 37 . Further studies are needed to clarify the involvement of PGI 2 /PPARδ in colon carcinogenesis.
In conclusion, our present findings demonstrate that PGIS promotes inflammatory reactions cooperatively with mPGES-1 but attenuates carcinogenesis, in contrast to mPGES-1. Both PGIS and mPGES-1 functionally couple with COX-2 as their upstream enzymes 5,15 . Toward the development of mPGES-1-specific inhibitors as novel NSAIDs without adverse side effects, the involvement of PGIS in target illnesses should be fully elucidated.

Methods
Animals. All animal experiments were performed in accordance with protocols approved by the Institutional Animal Care and Use Committees of Showa University, in accordance with the Standards Relating to the Care and Management of Experimental Animals in Japan. Balb/c and C57BL/6 mice were purchased from Saitama Experimental Animals Supply Co. (Saitama, Japan). mPGES-1 KO mice and PGIS KO mice on a C57BL/6 × 129/SvJ background were described previously 9,28 . In a colon carcinogenesis model, we used these KO mice, backcrossed 3 times onto Balb/c background 11  . These mice were housed in microisolator cages in a pathogen-free barrier facility. All mice were 6-10 weeks old when used for the described experiments.

Induction of peritonitis and preparation and activation of peritoneal MΦs.
Thioglycollate medium (Becton Dickinson, Sparks, MD) (1 mL/20 g of body weight) was intraperitoneally injected into mice, and peritoneal exudate cells and fluids were collected on day 2 and 4 by washing the cavity with 8 mL of PBS as described previously 9 . Cell number was determined by Trypan Blue exclusion. Cytocentrifuge preparations were Giemsa-stained, and cell subsets were identified and counted. For preparation of MΦ s, the peritoneal cells were washed, counted and then seeded into 12-well plates (Iwaki Glass, Tokyo, Japan) at a cell density of 10 6 cells/mL in 1 mL of RPMI medium (Nissui, Tokyo, Japan) supplemented with 10% (v/v) fetal calf serum. After incubation for 2 hours in a CO 2 incubator, the supernatants and non-adherent cells were removed. More than 90% of adherent cells were peritoneal MΦ s. The cells were then incubated with or without 10 μ g/mL LPS from Escherichia coli O111:B4 (Sigma, St. Louis, MO) in medium containing 2% serum for 24 hours. The culture media were taken for measurements of prostanoids.
Acetic acid writhing reaction. The acetic acid writhing reaction was induced in mice by an intraperitoneal injection of 0.9% (v/v) acetic acid solution into mice at a dose of 5 mL/kg, as described previously 9,21 . For the induction of COX-2, LPS (10 μ g/0.1 mL of saline/mouse) was given intraperitoneally 18 hours before the injection of acetic acid solution. The number of writhing responses was counted every 5 min. For the measurement of prostanoids, mice were sacrificed 10 min after the acetic acid injection, and their peritoneal cavities were washed with 5 mL of PBS.
Scientific RepoRts | 5:17376 | DOI: 10.1038/srep17376 Induction of colonic tumors by AOM treatment. Mice were intraperitoneally injected with AOM at a dose of 10 mg/kg body weight once a week for 6 weeks as described previously 11 , and then killed 6 or 20 weeks after the last injection of AOM. After laparotomy, the entire colons were dissected and then macroscopically divided into normal-appearing tissues and polyps as normal and tumor tissues, respectively. For the microscopic analysis, the dissected colons were filled with 10% neutral-buffered formalin and then opened longitudinally from the anus to the cecum. For the analysis of ACF formation, mice were killed 6 weeks after the last injection of AOM, and each colon was stained with 0.2% methylene blue in PBS. Colon tissues were scored under a light microscope for the number of ACFs or polyps per colon.
Immunohisotochemical analysis. Immunohistochemistry of the tissue sections was performed as described previously 9 . Briefly, the tissue sections were incubated for 15 min with Target Retrieval Solution (DAKO Japan, Kyoto, Japan) , incubated for 10 min with 3% (v/v) H 2 O 2 , washed three times with TBS for 5 min each, incubated with 5% (w/v) skim milk for 30 min, washed three times with TBS-Tween for 5 min each and incubated for 1 h with anti-PGIS antibody (Genway Biotech, SanDiego, CA) in TBS (1:100 dilution). The sections were then treated with the Envision staining kit (DAKO Japan), followed by counterstaining with hematoxylin.
RT-PCR analysis. Total RNA was extracted from colon mucosa by homogenization in TRIzol reagent. Two micrograms of RNA from each sample were subjected to a reverse transcription (RT) reaction using the High-Capacity cDNA Reverse Transcription Kit (Applied Biosystems, Carlsbad, CA). A SYBR Green-based protocol and real-time PCR detection system (Applied Biosystems) were used to detect mRNA levels. RNA amounts were normalized against the18S rRNA level.

Measurement of prostanoids by liquid chromatography-electrospray ionization-mass spectrometry (LC-ESI-MS).
For the extraction of prostanoids from the peritoneal fluids or culture media, an internal standard (50 pg of PGB 2 ) was added to medium (500 μ g), and then the medium was acidified by the addition of 100 μ L of 0.2% (v/v) formic acid followed by 500 μ L of ethyl acetate. The samples were mixed and centrifuged at 20,000 g for 10 min. The organic layer was retrieved and evaporated to dryness with a vacuum evaporator. Samples were resuspended in 100 μ L of mobile phase A (water/acetonitrile/ formic acid [63:37:0.02, v/v/v]) and injected into a LC-ESI-MS system. For the extraction of prostanoids from the tissues, the snap-frozen tissues were homogenized at 4 °C in SET buffer using a bead crusher (uT-01, TAITEC, Saitama, Japan), and then the homogenates were centrifuged at 3000 g for 5 min at 4 °C. Supernatants were isolated and adjusted to pH 3.0 with 1 M HCl. An internal standard (50 pg of PGB 2 ) was added to the samples, and then the samples were passed through a Sep-Pak C18 cartridge (Waters, Milford, MA). The retained PGs were eluted with 3 mL of ethyl acetate/methanol (9:1 [v/v]). The sample solvents were evaporated, and then the PGs were resuspended in 100 μ L of mobile phase A and injected into the LC-ESI-MS system.
All mass spectrometric analyses were performed using a Prominence HPLC system (Shimadzu, Kyoto, Japan) equipped with a linear ion trap quadrupole mass spectrometer (QTRAP5500, AB Sciex, Framingham, MA, ), as described previously 38 . Briefly, prostanoids were separated by reverse-phase LC on a TSKgel ODS-100 V column (2.0 × 150 mm inner dia, 5-μ m particle, Tohso, Tokyo, Japan) at a flow rate of 300 μ L/min at 30 °C. The column was equilibrated in mobile phase A (water/acetonitrile/formic acid [63:37:0.02, v/v/v]). Samples (10 μ L) were injected using a 50-μ L injection loop and eluted with a linear gradient from 0% to 20% mobile phase B (acetonitrile/isopropanol [1:1, v/v]) between 0 and 6 min. Mobile phase B was increased to 55% from 6 to 6.5 min and held until 10 min. This phase was increased to 100% from 10 to 12 min and held until 16 min. Then, from 16 to 17.5 min, the phase was dropped to 0% and held there until 20 min. Prostanoids were subsequently analyzed using a tandem quadrupole mass spectrometer via multiple-reaction monitoring (MRM) in negative-ion mode. The transitions monitored were m/z: 351/271 for PGE 2 and PGD 2 , 353/193 for PGF 2α , 369/207 for 6-ketoPGF 1α , 369/195 for TXB 2 and 333/235 for PGB 2 . These prostanoids were identified in samples by matching their MRM signal and LC retention time with those of a pure standard.