|Home | About | Journals | Submit | Contact Us | Français|
Prostaglandin E2 (PGE2) plays an important role in the normal physiology of many organ systems. Increased levels of this lipid mediator are associated with many disease states, and it potently regulates inflammatory responses. Three enzymes capable of in vitro synthesis of PGE2 from the cyclooxygenase metabolite PGH2 have been described. Here, we examine the contribution of one of these enzymes to PGE2 production, mPges-2, which encodes microsomal prostaglandin synthase-2 (mPGES-2), by generating mice homozygous for the null allele of this gene. Loss of mPges-2 expression did not result in a measurable decrease in PGE2 levels in any tissue or cell type examined from healthy mice. Taken together, analysis of the mPGES-2 deficient mouse lines does not substantiate the contention that mPGES-2 is a PGE2 synthase.
Prostaglandins are lipid mediators produced by all cell types and regulate multiple biological processes. The E type prostaglandins are associated with the regulation of a wide range of normal physiological functions including reproduction, smooth muscle contraction/dilation, induction of pain, sodium homeostasis and immune modulation. In addition, prostaglandin E2 (PGE2) is implicated in many pathological conditions such as inflammation  and certain epithelial cancers [2–4]. Prostaglandin E2 biosynthesis begins with the conversion of arachidonic acid (AA) released from phospholipids to prostaglandin endoperoxide (PGH2) by one of two cyclooxygenase enzymes, COX-1 or COX-2. PGH2 is next isomerized to PGE2 by terminal PGE2 synthase enzymes. To date this activity has been ascribed to three different genes: mPGES-1 and mPGES-2, reportedly encoding microsomal prostaglandin E2 synthases, and cPGES/p23, that was reported to encode a cytolsolic prostaglandin E2 synthase .
cPGES/p23 was purified from rat brain extracts, and similar to mPGES-1, the assignment of cPGES/p23 synthase activity and PGE2 biosynthesis was attributed principally on in vitro studies . More recently, the generation of cPGES/p23 deficient mice has raised questions regarding the direct contribution of this protein to PGE2 biosynthesis . For example, analysis of cPGES/p23 in vivo suggests that this protein has PGE2 independent functions. Recent data demonstrates that cPGES/p23 is essential for glucocorticoid receptor function [5, 7].
mPGES-1 is a member of the membrane-associated proteins involved in eicosanoid and glutathione metabolism (MAPEG) family of proteins, and its structure and regulation have been well characterized . mPGES-1 expression is induced by various pro-inflammatory stimuli in cell populations and organs involved in immune responsiveness such as macrophages [9, 10] and the spleen and lung [11–13]. However, constitutive expression is also detected in some organs such as the stomach, spleen and kidney . The role of mPGES-1 in both basal and inducible PGE2 biosynthesis by partnering with either COX-1 or COX-2 in vivo has been extensively documented by recent studies using mPGES-1 deficient mice [9–11, 14].
mPGES-2 was first isolated as a microsomal protein with PGE2 synthase activity from the bovine heart, and cDNAs encoding human and monkey homologs were subsequently identified . mPGES-2 is a 41 kDa protein, which is structurally distinct from mPGES-1 and, unlike mPGES-1, does not exclusively depend on glutathione (GSH) for its catalytic activity . mPGES-2 has a conserved glutaredoxin and thioredoxin like domain, although it is not similar to the GSH-S transferase family. Human mPGES-2 expression was detected in a number of cell lines and a wide spectrum of tissues. The highest mRNA levels are detected in the heart, various regions of the brain, skeletal muscle and liver . Interestingly, high levels of mPGES-2 were detected in organs such as the intestine, heart and brain, which normally express very low levels of mPGES-1 , suggesting that mPGES-2 may be a major contributor to normal PGE2 biosynthesis in these tissues. In contrast to mPGES-1, expression of mPGES-2 was not inducible by LPS treatment in organs such as the brain, heart and liver , again supporting a more constitutive function for mPGES-2. Also, unlike mPGES-1, whose expression was markedly induced in pathological human tissues, such as the heart after acute myocardial infarction, cirrhotic livers, joints with rheumatoid arthritis  and gastric ulcers , mPGES-2 appeared to be expressed equally in both the normal and pathological samples . These findings together supported a more “housekeeping” type of role for mPGES-2 in vivo.
The role of mPGES-2 as a functional terminal PGE2 synthase has been primarily characterized in vitro to date . Over expression of the cDNAs encoding mPGES-2 and either COX-1 or COX-2 in human embryonic kidney cells revealed that mPGES-2 can partner with either COX-1 or COX-2 for increased PGE2 production upon exposure of cells to AA. Similarly, in cells stimulated by Ca++ ionophore A23187 or IL1-β to release endogenous stores of AA, mPGES-2 coupled both with exogenously expressed COX-1 and COX-2 for increased PGE2 biosynthesis. In contrast to mPGES-1, which preferentially augmented PGE2 synthesis when co-expressed with COX-2, mPGES-2 did not display a COX-1 or COX-2 coupling selectivity. Thus, mPGES-2 was shown to participate in both acute (Ca++ ionophore stimulated) and delayed (IL1-β stimulated) PGE2 biosynthesis. Similar findings in lung epithelial and rat fibroblast cell lines co-expressing mPGES-2 with COX-1 and COX-2 following stimulation with either Ca++ ionophore or TNF-α supported the idea that mPGES-2 non-selectively couples with COX-1 and COX-2 in multiple cell types . Since mPGES-2 is co-expressed with both COX-1 and COX-2 in many organs, the findings of Murakami et al. further strengthened the hypothesis that mPGES-2 is non-selective in its COX isoform coupling in vivo.
mPGES-2 was also independently isolated from an IFN-γ stimulated macrophage cDNA library as a novel protein, GATE Binding Factor 1 (GBF-1), that bound and activated transcription from the IFN-γ responsive element GATE . Sequence analysis of GBF-1 revealed that it was identical to mPGES-2, and Gbf-1 expression was similarly detected in most tissues with highest levels in the heart, liver, kidney and brain. Hu et al. demonstrated that Gbf-1/ mPges-2 expression was induced by IFN-γ in multiple organs and that GBF-1 was localized to both the nuclear and cytoplasmic compartments in MEF and HeLa cell lines. Additionally, GBF-1 was shown to bind the GATE element and stimulate gene expression of reporter constructs that were driven specifically by GATE sequences. GBF-1 over-expression in primary MEFs, upon addition of IFN-γ, also increased expression of endogenous IRF-9, an IFN-γ regulated gene with a naturally occurring GATE element in its promoter [18, 19]. Thus, Hu et al. proposed that GBF-1/mPGES-2 is a novel transcription factor capable of modulating IFN-γ target gene expression and speculated that the glutaredoxin domain may allow this transcription factor to be regulated via redox status.
To better define and clarify the proposed roles of mPGES-2 in vivo as a terminal PGE2 synthase and/or a novel IFN-γ stimulated transcription factor, we generated mice deficient in mPGES-2 by gene-targeting. Analysis of these mice fails to support the hypothesis that mPGES-2 contributes to production of PGE2 in vivo, nor does it provide evidence for a role in regulating IFN-γ gene expression.
All animal colonies were maintained according to standard guidelines as defined by the NIH Guide for the Care and Use of Laboratory Animals and the Institutional Animal Care and Use Committee of UNC-Chapel Hill. The generation of mPGES-1 and COX-1 deficient mouse lines was described previously [9, 20]. mPGES-1 deficient mice were generated on a DBA/1lacJ genetic background and COX-1 deficient mice were bred on a C57BL/6 genetic background.
mPges-2 cDNA sequence from Macaca fascicularis was used to identify the corresponding mouse cDNA and genomic sequences by homology search. The targeting construct was designed to disrupt exon 4 based on the conserved protein sequence. A neomycin resistance marker in the plasmid pJNS2 replaced 8 base pairs of exon 4 in the endogenous mPtges-2 gene. The targeting plasmid was introduced into 129/SvEv derived ES cells by electroporation. Colonies resistant to both neomycin and gancyclovir were screened for the desired recombination event by Southern blot analysis. A 432 base pair probe was prepared from DNA downstream of the targeted locus, used to identify cells that had undergone homologous recombination, and later to genotype the mice. A targeted ES cell clone was injected into 3.5-day old C57BL/6 blastocysts and implanted in pseudopregnant females. Chimeras were crossed directly to 129/SvEv (Tatonic) to maintain the mutation on the 129 background, and thus generate co-isogenic mPges-2 mutant lines for studies. In addition, the mutation was moved to the C57BL/6 genetic background, by 12 consecutive crosses to this laboratory strain.
Total RNA was extracted from frozen kidney, lung and proximal small intestine samples using RNABee (Tel-Test). The RNA was fractionated on a 1% agarose/formaldehyde gel and subsequently transferred to a nitrocellulose membrane (Millipore). A 733 bp cDNA probe for mPges-2 was obtained from an EST (Invitrogen). The fragment encodes for a portion of the cDNA from the middle of exon 5 to the end of exon 7, including the stop codon and the 3' UTR. The probe was labeled with 32P[dCTP] (Amersham) by random priming (Roche). The blots were stripped and re-probed for mPges-1 and β-actin expression using the corresponding mouse cDNA probes.
Total RNA was extracted from tissue samples as described above. RNA was purified with Qiaprep RNeasy Kit (Qiagen) and the quality of RNA was evaluated with an Agilent 2100 Bioanalyzer. 5–10 µg RNA was reverse transcribed using the High-Capacity Reverse Transcription Kit (Applied Biosystems) according to manufacturer’s instructions. cDNA was amplified with Taqman PCR Universal Master Mix (Applied Biosystems) using the Applied Biosystems 7900 HT Fast Real-Time PCR System. All samples were run in quadruplicate. Genes of interest were normalized to a housekeeping gene (Gapdh or Hprt1). All Taqman primer/probe sets were purchased from Applied Biosystems. Data was analyzed using the comparative CT method (ΔΔCT).
Hearts were harvested from 4 mPges-2 −/− mice and their wild-type controls. Total RNA was extracted and the integrity was assessed as described above. RNA was synthesized and amplified using the Agilent Low RNA Input Fluorescent Linear Amplification Kit according to the manufacturer’s specifications. The RNAs were labeled with either cyanine 3-CTP (Cy-3) or cyanine 5-CTP (Cy-5) (Perkin Elmer). 750 ng of Cy-3 labeled reference and Cy-5 labeled sample cRNA were hybridized to Agilent Whole Mouse Genome microarrays overnight and scanned using an Agilent DNA microarray scanner G2565B and Feature Extraction software. Aliquots containing equal amounts of RNA were combined from 4 wild-type heart samples to serve as the reference pool. All samples, both mPges-2 +/+ and −/−, were compared to the same reference pool. The Rosetta Resolver Gene Expression Data Analysis system was used to normalize, transform and combine the fluor-reversed duplicates. A KO-specific gene set was established based on a statistically significant up- or down-regulation only in the knock-out mice. First, the Resolver error-model was used to assign measures of confidence to each of the two experimental conditions . For each probe, an error-weighted measure of the relative mRNA abundance for each of two comparisons (WT vs. WT and KO vs. WT) was calculated. Evidence of differential expression was only considered if the combined absolute fold change was greater than 1.25 and if the combined p-value was less than 0.05 only in the KO vs. WT group. 868 sequences were identified in this analysis. Second, an ANOVA analysis was applied to compare the 4 WT vs. WT and the 4 KO vs. WT experiments. 1,646 sequences with ANOVA p-value less than 0.05 were selected.
The final KO-specific gene set of 171 sequences was obtained by taking the common sequences from these two analyses. These 171 sequences were further analyzed by using the Ingenuity pathway analysis tool .
Peritoneal macrophages were harvested from mice three days after an intraperitoneal (IP) injection of 0.1% solution of Zymosan A (Sigma) in PBS. The peritoneum was lavaged with 4 mL DMEM (Gibco) containing 20 mM HEPES buffer (Gibco). Cells were cultured and incubated at 37°C in a humidified incubator containing 5% CO2 for 45 minutes, after which the media was aspirated, replaced with warmed media, and incubated for an additional 30 minutes. The supernatant was collected and replaced with media containing 5 µg/mL Ca++ ionophore (A23187) (Sigma) or 10−5M arachidonic acid (AA) (Sigma). After 30 minutes of stimulation, the media was collected. Macrophages stimulated with lipopolysaccharide (LPS) (Sigma) were cultured in DMEM/10% FBS for 16 hours. The cell count was verified using a β-hexosaminidase assay as previously described , and PGE2 levels of pre- and post-stimulated samples were determined by ELISA (Amersham).
Peritoneal macrophages stimulated with IFN-γ were collected and cultured as described above. Macrophages were stimulated with RPMI/10%FBS containing 100 ng/mL recombinant murine IFN-γ (Sigma) for 8 and 16 hours, respectively. After IFN-γ treatment, cells were rinsed twice with PBS and lysed with RNABee for total RNA extraction.
Proximal small intestine, kidney, lung, brain and heart were harvested from mPges-2 −/−, mPges-1 −/− and Cox-1 −/− mice and their respective wild-type controls. All tissues were flash frozen in liquid nitrogen and stored at −80°C. The tissues were homogenized in 1X PBS/1 mM EDTA containing 10 µM indomethacin, on ice. Prostanoids were precipitated with 75% ethanol and glacial acetic acid. The supernatant was run over a C18 column (Amersham), prostaglandins were eluted in ethyl acetate, evaporated with N2 and resuspended in PGE2 EIA assay buffer. The quantity of PGE2 was determined using the Prostaglandin E2 Biotrak Enzymeimmunoassay System (Amersham) or the Correlate-EIA Prostaglandin E2 Enzyme Immunoassay Kit (Assay Designs) according to manufacturer's specifications. Each sample was run in duplicate. The detection ranges of the kits are 50 – 6400 pg/mL and 39.1 – 2500 pg/mL, respectively; however, their cross-reactivity with isomeric PGE2 analogs, such as 8-iso-PGE2 has not been determined.
Measurement of PGE2 by LC/MS was performed as described previously . The lower limit of detection for PGE2 by LC/MS is 0.002 ng/mg protein. Isomeric analogs of PGE2 were not separated in LC/MS or ELISA, thus the amount of PGE2 measured potentially includes the levels of these compounds.
Statistical analysis was performed using JMP Statistical Software package v5.1 (SAS) and Prism 4 (GraphPad Software). Comparisons of the mean were made by Student’s t-test or ANOVA followed by Tukey-Kramer’s HSD post hoc test as necessary unless otherwise noted. Data are shown as mean ± SEM. Differences with p<0.05 were considered statistically significant.
A mutation was introduced into the mouse mPges-2 gene, which encodes for microsomal prostaglandin E2 synthases (mPGES-2), in ES cells by homologous recombination, using the strategy outlined in Figure 1A. The recombination event results in the replacement of 7 base pairs of exon 4, which encode for amino acids 194 to 196, with the neomycin phosphotransferase gene. ES cells in which the recombination event occurred were used to generate mice carrying the mutant mPges-2 allele. Intercross of these heterozygous animals yielded offspring homozygous for the mPges-2 mutation at expected frequencies. These mice are healthy, fertile, have normal growth rates and cannot be distinguished from littermates by observation alone. Morphological and histological analysis of the major organ systems failed to identify a role for mPGES-2 in normal development. Examination of thymus, spleen and lymph nodes showed normal development of T cells, B cells and macrophages (data not shown).
To verify that the mutation introduced into the mPges-2 gene generated a null allele, RNA was prepared from tissues of mPges-2 −/− mice and littermate controls. Three tissues reported to express high levels of mPGES-2, the kidney, lung and intestinal tract were included in the study. For comparison, RNA was also prepared from the same tissues of mPges-1 −/− mice, which encodes for microsomal prostaglandin E2 synthases (mPGES-1). As expected, high levels of mPges-2 were observed in the kidney, lung and intestinal tract of all wild type mice, as well as mice lacking mPges-1. In contrast, no mPges-2 expression was detected in any of the tissues prepared from the mPges-2 −/− animals (Fig. 1B). Similar to mPges-2, mPges-1 expression is easily detected in the lung and kidney. However, in contrast to the high levels of mPges-2 in the small intestine, mPges-1 expression is barely detectable in this tissue. No obvious compensatory change in the expression of mPges-1 was observed in the mPges-2 −/− organs, and conversely, loss of mPges-1 did not result in an easily detectable increase in mPges-2 RNA levels.
Basal production of PGE2 by the intestinal tissue of mPges-2 −/− mice and wild-type littermate controls was determined by ELISA and compared to levels measured in similar samples obtained from Cox-1 −/− and mPges-1 −/− mice and appropriate controls. PGE2 can be easily detected in all wild-type samples, albeit at different levels, likely reflecting differences between mouse strains in basal PGE2 production. As expected, the production of PGE2 in healthy intestine is largely dependent on Cox-1 expression (Fig. 2C, p<0.001). In mice lacking the COX-1 enzyme, PGE2 levels are close to the lower limit of detection. Surprisingly, loss of mPges-2 expression did not reduce PGE2 levels in the small intestine (Fig. 2A). In contrast, mPges-1 mRNA levels, as assessed by Northern analysis, are very low in the intestinal tract; however, basal PGE2 production appears to be almost completely dependant on this enzyme, as PGE2 levels are significantly diminished in mPges-1 −/− mice (Fig. 2B, p<0.05). PGE2 levels were also measured in additional tissues from mPges-2 −/− mice and co-isogenic controls. Basal PGE2 levels in lung, kidney, heart and brain of mPges-2 −/− mice did not differ significantly from those of controls (Fig. 3).
In addition to examining PGE2 levels in mPGES-2 deficient mice on a 129/SvEv genetic background, the mPges-2 mutation was also backcrossed onto the C57BL/6 genetic background for twelve consecutive generations. Mice homozygous for the mPges-2 mutation were generated by intercross of these animals, generating congenic C57BL/6 mPges-2 −/− animals. Examination of tissues from these animals also failed to identify a contribution of mPGES-2 to basal levels of PGE2 in these tissues (data not shown).
To further verify these findings we examined levels of PGE2 and other prostanoids in extracts prepared from the proximal small intestine and the brain of 129/SvEv mPges-2 −/− and co-isogenic controls by LC/MS. Both the brain and small intestine express high levels of mPges-2, while mPges-1 expression is barely detectable in these organs in basal conditions (Fig. 1B) . As can be seen in Figure 4A, the level of PGE2 in intestinal tissue is not reduced by loss of mPges-2 expression. In fact, PGE2 levels were slightly elevated in the mPges-2 −/− tissue compared to co-isogenic controls, although this difference did not achieve statistical significance (p=0.0565, n=3). No significant differences were measured in PGE2 levels in either the brain (Fig. 4B) or kidney (data not shown). Taken together, these studies failed to support a role for mPges-2 in basal production of PGE2.
The role of mPGES-1 in the production of PGE2 by macrophages has been established; however, these studies also showed that macrophages can produce some PGE2 independent of expression of this enzyme . To determine if mPGES-2 contributed to PGE2 production by macrophages, peritoneal macrophages were isolated from 129/SvEv mPges-2 −/− mice and their co-isogenic controls. After peritoneal lavage, macrophages were enriched by adherence to plastic, and PGE2 levels in the culture supernatant were determined before and at various times after stimulation of the cells with A23187, AA or LPS. In some cases, parallel cultures were treated with indomethacin, a potent COX-1 and COX-2 inhibitor, prior to stimulation. As expected, both Ca++ ionophore and AA result in a rapid increase in PGE2 production (Fig. 5A). Pre-treatment with indomethacin inhibited PGE2 synthesis to background levels detected in unstimulated cell supernatants (data not shown). However, no difference was observed in the production of PGE2 by the mPges-2 −/− and control macrophages in response to either stimulus (Fig. 5A). As expected, PGE2 synthesis was significantly reduced, but not completely abolished, in mPges-1 −/− peritoneal macrophages in response to both AA and A23187 (Fig. 5B).
Treatment of macrophages with LPS has been demonstrated to induce COX-2 expression and increase production of PGE2 . A functional role for mPGES-1 in LPS stimulated PGE2 biosynthesis has been independently described [9, 10]. In accordance with these data, LPS treatment resulted in high levels of PGE2 in the supernatant of both mPges-2 +/+ and −/− macrophages in a dose dependent manner. However, loss of mPges-2 did not alter the ability of these cells to produce PGE2 (Fig. 5C). These experiments, taken together, do not support a role for mPGES-2 in the production of PGE2 by peritoneal macrophages.
Given the failure to measure a role for mPGES-2 in PGE2 synthesis, it was of interest to determine whether other functions attributed to this gene could be supported by studies of the mPGES-2 deficient mice. mPGES-2 expression was reported to be inducible by IFN-γ and contribute to IFN-γ mediated regulation of IRF-9 gene expression. To assess the role of mPGES-2 as GBF-1, we examined induction of Irf9 expression after treatment of macrophages with IFN-γ. Peritoneal macrophages were harvested by lavage of the peritoneal cavity 72 hours after induction of acute peritonitis with Zymosan A, and enriched by adherence to plastic. Cells were then stimulated with 100 ng/mL IFN-γ for 8 or 16 hours. mPges-2 expression levels were unchanged compared to untreated macrophages at 8 and 16 hours (Fig. 6A). As expected, IFN-γ treatment significantly increases endogenous Irf9 expression after 16 hours of stimulation (Fig. 6B). However, Irf9 expression is not significantly diminished in IFN-γ treated, mPGES-2 deficient macrophages compared to wild-type levels at both time points. Therefore, while Irf9 expression was induced by IFN-γ treatment, this data indicates that mPGES-2 is not essential for IRF-9 induction.
Survey of RNA prepared from various mouse organs by quantitative real-time PCR indicates that moderate levels of mPges-2 are detected in the heart under basal conditions (Fig. 7). The heart is a less complex organ with fewer cell types than the lung and intestine and is also less sensitive to metabolic variations (compared to liver or intestine). Additionally, because low basal levels of PGE2 are produced, despite moderate expression levels of mPges-2 in this organ, it seemed ideal for the identification of functions of mPGES-2 independent of PGE2 synthesis. RNA was prepared from the hearts of four age, sex, and weight matched 129/SvEv mPges-2 −/− mice and their co-isogenic controls for microarray analysis. Analysis revealed surprisingly few differences between the wild-type and mPges-2 −/− hearts. The top genes that increased or decreased in expression in the mPges-2 −/− heart are shown in Table 2 in order of fold change. The expression of only one gene increased, while the expression of two genes decreased significantly with a change greater than 2.0 fold. As anticipated, this analysis showed a significant decrease of mPges-2 mRNA levels in the heart of mPges-2 −/− animals. We chose to verify the change in gene expression for those candidates with commercially available primer/probe sets by quantitative real-time PCR analysis. PCR was carried out both with the samples used for microarray analysis and from independent RNA samples prepared from mPges-2 −/− and control animals. Of these candidates only changes in the expression level of Trim13 were confirmed by quantitative real-time PCR in the heart. To further explore the effect of mPGES-2 on Trim13 expression in mPges-2 −/− mice, kidney and intestinal tract samples were also evaluated for Trim13 mRNA levels by quantitative real-time PCR. Trim13 expression was significantly decreased in mPGES-2 deficient heart, kidney and jejunum to nearly half of wild-type levels (Fig. 8, p<0.005). While, in general, the impact of loss of mPGES-2 was surprisingly modest, these results demonstrate that mPGES-2 loss can directly or indirectly affect the expression of other genes.
Given the observed ability of loss of mPGES-2 to alter Trim13 expression, we considered the possibility that the identification of mPGES-2 as a synthase might reflect its ability to regulate expression of other enzymes involved in either PGE2 synthesis or catabolism, as had been observed for cPGES/p23 . This possibility was heightened by the observation that slightly elevated levels of PGE2 levels were observed occasionally in some tissues lacking mPges-2. To test this hypothesis, we examined the expression of Cox-1, Cox-2 and mPges-1, enzymes critical in PGE2 production in mPges-2 +/+ and −/− small intestine. Expression of the cytosolic prostaglandin E2 synthase, cPges, a gene also purported to convert PGH2 to PGE2 was examined, as well. PGE2 is rapidly metabolized to 15-PGE2 by 15-PGDH (encoded by 15-Pgdh), the primary enzyme involved in PGE2 catabolism. Thus, expression analysis of 15-Pgdh was also incorporated in the study. As seen in Figure 9, there was no change in the expression of Cox-1, Cox-2, mPges-1 and cPges in the mPges-2 −/− small intestine. A significant reduction in expression of 15-Pgdh was initially observed in some samples from the 129 co-isogenic animals (Fig. 9); however, this reduction was not consistently observed in small intestine samples from mPges-2 −/− mice on a C57BL/6 genetic background (data not shown). Taken together, the loss of mPGES-2 expression did not consistently alter either the PGE2 levels or expression of enzymes required for PGE2 synthesis or metabolism.
mPGES-1 was the first enzyme identified and characterized as capable of catalyzing the conversion of PGH2 to PGE2 . The generation of mice lacking this enzyme verified the role of mPGES-1 in PGE2 synthesis, both in its contribution to basal PGE2 levels and to increased PGE2 production during inflammatory responses [9–11].
A number of observations made during the study of these mice suggested the existence of mPGES-1 independent PGE2 production. The expression levels of mPGES-1 in various tissues did not correspond well with the basal levels of PGE2 produced by these tissues. Furthermore, while mPGES-1 was responsible for the majority of PGE2 biosynthesis, measurable amounts of PGE2 could still be detected in mPGES-1 deficient cells and tissues. Moreover, contrary to expectation, some phenotypes of mice lacking COX-2 and the EP receptors were not evident in mPGES-1 deficient animals. Notably, the patent ductus observed in COX-2 deficient  and EP4 deficient  mice, was not recapitulated in mPGES-1 deficient mice .
A possible explanation for these observations was suggested by the cloning of an enzyme, designated mPGES-2, initially purified from microsomal fractions of bovine heart . Over-expression of mPGES-2 in cell lines resulted in increased PGE2 levels . Several subsequent publications furthered the supposition that mPGES-2 was a PGE2 synthase. Interestingly, mPges-2 is located in a cluster of genes, including Cox-1 and Cox-2, on mouse chromosome 2. To define the in vivo role of mPges-2 in PGE2 biosynthesis, we have generated a mouse carrying a null allele of this gene. No mPges-2 mRNA can be detected in any of the tissues examined to date. However, surprisingly, analysis of PGE2 levels in these mice fails to support the classification and designation of mPges-2 as a prostaglandin E2 synthase. Basal PGE2 levels were not decreased in any of the mPGES-2 deficient tissues examined in this study. Ex vivo studies of macrophages isolated from mPges-2 −/− mice also failed to support a role for mPGES-2 in PGE2 production. mPges-2 −/− macrophages produced amounts of PGE2 similar to wild-type controls in response to AA, Ca++ ionophore and LPS. Under identical conditions, mPGES-1 independent PGE2 production was measured as reported previously [9, 10]. These data suggest that mPGES-2 is not responsible for residual PGE2 production observed in mPGES-1 deficient mice.
These findings are in contrast with studies published by Tanikawa et al. and Murakami et al. [12, 16], which demonstrate that mPGES-2 enzymatically promotes PGE2 biosynthesis in vitro. In studies by Murakami et al., over-expression of mPGES-2 in cell lines, which express mPGES-1 at very low levels, resulted in a measurable increase in PGE2 production. The basis for the apparent discrepancy between these observations and our results is not obvious. It is possible that the ability of transfected cells to increase PGE2 was the indirect consequence of increased levels of this protein and not its function as a synthase. For instance, increases in PGE2 levels may potentially result from facilitating the non-enzymatic conversion of PGH2 to PGE2 or by alterations in the catabolism of PGE2. It is possible that the ability of mPGES-2 to facilitate PGE2 production requires very high and non-physiological levels of expression, which are not observed in most healthy, normal tissues. However, at this time, we also cannot rule out the possibility that populations of cells or tissues not examined in this study can utilize this protein in the production of PGE2.
Studies of mPGES-2 deficient mice also failed to support reports that this gene could act as GBF-1, a novel transactivator of IFN-γ dependent gene expression. Hu et al.  demonstrated that mPGES-2/GBF-1 activated transcription of ISGF3g/p48/IRF-9 upon IFN-γ treatment of RAW264.7 murine macrophage cells. While a robust induction of Irf9 was measured, both in wild-type and mPges-2 −/− peritoneal macrophages stimulated with IFN-γ, Irf9 levels measured in mPGES-2 deficient macrophages were not significantly different from the wild-type macrophages (Fig. 6A). Hu et al. also demonstrate that Gbf-1/mPges-2 mRNA expression was induced in RAW cells after 8 hours of IFN-γ stimulation. In contrast, Gbf-1/mPges-2 mRNA levels are not up-regulated in peritoneal macrophages isolated from wild-type mice (Fig. 6B) after 8 or 16 hours of IFN-γ stimulation. In the context of these data, it is not surprising that Irf9 levels were not significantly reduced in mPges-2 −/− peritoneal macrophages. It is possible that peritoneal macrophages, unlike, RAW264.7 murine macrophage cells may not be dependent on GBF-1/mPGES-2, and these cells may have a different transcriptional program resulting from their immortalization.
Meng et al.  recently reported that IFN-γ induces the interaction of GBF-1/mPGES-2 with CEBP-β, and that recruitment of GBF-1 to the GATE/ CEBP-β element is CEBP-β dependent. It is therefore possible that IRF-9 levels are not significantly decreased in the absence of mPGES-2 because IRF9 transcription is predominantly dependent on transcription factors such as CEBP-β in peritoneal macrophages. Alternately, it remains possible that steady state mRNA levels may not accurately represent changes in the rate of transcription of the IRF9 gene.
Because of the report that mPGES-2/GBF-1 may play a role in regulation of gene expression we carried out an additional line of inquiry into this potential function, using genome wide expression analysis of the heart. Surprisingly, few differences were found in the gene expression pattern of hearts lacking mPges-2. Of the genes with statistically significant differences in expression levels, only one of the five candidates tested could be validated by quantitative real-time PCR. We verified that expression levels of Trim13/RFP2/Leu5 (tripartite motif 13, Ret Finger Protein 2), were significantly down regulated in the heart, and interestingly, in the kidney and proximal small intestine, as well.
Trim13/ Leu5/ RFP2, a putative transcription factor, is a member of the tripartite motif family proteins containing RING finger, B-box type and coiled-coil domains . The mouse Trim13/RFP2/Leu5 gene locus located on chromosome 14 is not linked to mPges-2 and potentially encodes for two proteins, RFP2/Leu5 and DLTET, of unknown function [32, 33]. Loss of RFP2/Leu5 appears to be an early event in human B-cell CLL suggesting that RFP2/Leu5 is a tumor suppressor gene . To date, we have not observed any lymphoid tumors in mPges-2 −/− mice that are over 12 months old. Furthermore, FACS analysis of cells isolated from the spleen, thymus and lymph nodes has not revealed an abnormal increase in B-cells or any other lymphoid population. It is unclear at this time if the changes in Trim13 expression are directly caused by loss of mPges-2 dependent transcription or are indirect and potentially compensatory for the loss of mPGES-2 function. Thus, while a role for mPGES-2 in the regulation of IRF-9 is not supported by these studies, the gene expression profiles leave open the possibility that mPGES-2 can act as a transcription factor.
It was therefore of interest to determine if the increased basal levels of PGE2 observed in some tissues lacking mPGES-2 reflected altered expression of genes involved in PGE2 synthesis or metabolism. No differences were observed in expression levels of Cox-1, Cox-2 or mPges-1 mRNA, and while in some cases a decrease in expression of 15-Pgdh was observed, we failed to consistently observe increases in PGE2 levels. Furthermore, the potential increase in PGE2 levels was not observed after backcrossing of the mPges-2 mutation to the C57BL/6 genetic background.
In summary, our data indicates that, similar to cPGES, mPGES-2 does not encode a prostaglandin sythase. Thus mPGES-1 mediated conversion of cyclooxygenase-derived PGH2 to PGE2 may represent the only enzymatic pathway by which PGE2 is generated in vivo, with other non-enzymatic sources of production accounting for the differences in the phenotypes of the prostaglandin receptor-deficient mice and the mice lacking this enzyme.
We thank Anne Latour for assistance in tissue culture. This work was supported by National Institutes of Health Grants DK069896 (B.H.K.), DK38108, and HL068141 (B.H.K.) and the Department of Defense DAMD17-03-1-0453 (S.C.)