Two Pathways for Cyclooxygenase-2 Protein Degradation in Vivo*

COX-2, formally known as prostaglandin endoperoxide H synthase-2 (PGHS-2), catalyzes the committed step in prostaglandin biosynthesis. COX-2 is induced during inflammation and is overexpressed in colon cancer. In vitro, an 18-amino acid segment, residues 595–612, immediately upstream of the C-terminal endoplasmic reticulum targeting sequence is required for N-glycosylation of Asn594, which permits COX-2 protein to enter the endoplasmic reticulum-associated protein degradation system. To determine the importance of this COX-2 degradation pathway in vivo, we engineered a del595–612 PGHS-2 (Δ18 COX-2) knock-in mouse lacking this 18-amino acid segment. Δ18 COX-2 knock-in mice do not exhibit the renal or reproductive abnormalities of COX-2 null mice. Δ18 COX-2 mice do have elevated urinary prostaglandin E2 metabolite levels and display a more pronounced and prolonged bacterial endotoxin-induced febrile response than wild type (WT) mice. Normal brain tissue, cultured resident peritoneal macrophages, and cultured skin fibroblasts from Δ18 COX-2 mice overexpress Δ18 COX-2 relative to WT COX-2 expression in control mice. These results indicate that COX-2 can be degraded via the endoplasmic reticulum-associated protein degradation pathway in vivo. Treatment of cultured cells from WT or Δ18 COX-2 mice with flurbiprofen, which blocks substrate-dependent degradation, attenuates COX-2 degradation, and treatment of normal mice with ibuprofen increases the levels of COX-2 in brain tissue. Thus, substrate turnover-dependent COX-2 degradation appears to contribute to COX-2 degradation in vivo. Curiously, WT and Δ18 COX-2 protein levels are similar in kidneys and spleens from WT and Δ18 COX-2 mice. There must be compensatory mechanisms to maintain constant COX-2 levels in these tissues.

There are two prostaglandin endoperoxide H synthases (PGHSs) 2 called PGHS-1 and PGHS-2. The enzymes are commonly referred to as COX-1 and COX-2, and we use this latter terminology here. COXs catalyze the committed step in prostanoid biosynthesis, which is the conversion of one molecule of arachidonic acid, two O 2 molecules, and two electrons to prostaglandin endoperoxide H 2 (1)(2)(3)(4). PGH 2 is the immediate precursor of what are considered to be the major bioactive prostaglandins (PGs), including PGD 2 , PGE 2 , PGF 2␣ , PGI 2 , and thromboxane A 2 . A different gene encodes each COX isoform (5)(6)(7).
In general, COX-1 is expressed constitutively, and COX-2 is expressed inducibly in response to various stimuli (6,7). However, COX-2 is expressed constitutively in kidney (8) and brain (9). Each COX isoform subserves a distinct set of biologies (10 -12). Both the enzymatic properties and the patterns of expression of the isoforms contribute to their unique functions (12). In part because of the association of COX-2 expression with pathologies such as inflammation and colon cancer (13)(14)(15), many studies have been performed to elucidate the mechanisms of transcriptional regulation of COX-2 as reviewed recently (6,7). It is now also clear that COX-2 is regulated post-transcriptionally at the level of both mRNA degradation (16) and protein degradation (17,18).
In vitro studies with cultured murine NIH 3T3 fibroblasts and with HEK293 cells heterologously expressing various forms of COX-2 have suggested that there are two independent pathways of COX-2 protein degradation (17,18) dubbed "endoplasmic reticulum-associated degradation" (ERAD) and "substratedependent degradation" pathways. N-Glycosylation of Asn 594 , which appears to occur post-translationally, leads to entry of COX-2 into the ERAD pathway. Sequences both upstream and downstream of Asn 594 modulate the rate of Asn 594 glycosylation and collectively include a 27-amino acid instability motif  involving residues Glu 586 to Lys 612 (18). The second COX-2 degradation process is substrate turnover-dependent, * This work was supported, in whole or in part, by National Institutes of correlates with COX-2 suicide inactivation, and is independent of N-glycosylation at Asn 594 . In vitro, this pathway can be attenuated or blocked by various nonspecific and nonsteroidal anti-inflammatory drugs and COX-2 inhibitors (3,18). The biochemical mechanism underlying the disappearance of immunoreactive COX-2 associated with substrate-dependent inactivation is unresolved.
Although there is considerable evidence that COX-2 can be degraded by both ERAD and suicide pathways in vitro, the relative importance of these processes in COX-2 protein homeostasis in vivo has not been determined. To address this issue, we developed a knock-in del595-612 PGHS-2 (⌬18 COX-2) mouse in which residues 595-612 are deleted. Absent key residues important in Asn 594 glycosylation, ⌬18 COX-2 cannot undergo degradation via the ERAD pathway. However, ⌬18 COX-2 is kinetically and pharmacologically indistinguishable from native COX-2 (17), and like native COX-2 it undergoes a substrate-dependent disappearance (18). Here, we provide evidence that both inactivation processes function in vivo in normal and endotoxin-treated mice but that the contribution of each process to COX-2 protein degradation is different in different tissues.

EXPERIMENTAL PROCEDURES
Materials-129X1/SvJ genomic DNA (stock number 000691) was obtained from The Jackson Laboratory. The Expand Long Template PCR system, Complete EDTA-free protease inhibitor mixture, and endoglycosidase H were purchased from Roche Applied Science. QuikChange TM site-directed mutagenesis kit was obtained from Stratagene. Ex Taq TM DNA polymerase (TAK_RR001) was purchased from Takara Mirus Bio. Dulbecco's modified Eagle's medium (DMEM), minimum essential medium (MEM), RPMI 1640 medium, heat-inactivated fetal bovine serum (HFBS), penicillin/streptomycin, Lipofectamine 2000 reagent, Opti-MEM medium, tetracycline, collagenase type II, and NuPAGE were obtained from Invitrogen. Collagenase type A, Escherichia coli lipopolysaccharide (LPS), and protease inhibitors were purchased from Sigma. Cycloheximide (CHX) and MG132 were obtained from Calbiochem. Kifunensine (KIF) was purchased from Toronto Research Chemicals. S-Flurbiprofen and arachidonic acid were obtained from Cayman Chemical Co. DNeasy blood and tissue kit and RNeasy kit were obtained from Qiagen. [1-14 C]Arachidonic acid (1.85 GBq/mmol) was purchased from American Radiolabeled Chemical. Precoated silica gel 60 glass plates for TLC were obtained from Merck. Mouse anti-actin antibody was purchased from MP Biomedicals. Fluorescein isothiocyanate-labeled goat anti-rabbit IgG and goat anti-mouse IgG were obtained from Bio-Rad.
SDS-PAGE and Western Blotting-Frozen cell pellets or solubilized microsomes prepared from mouse tissue were used as the source of protein for Western blotting. Frozen cell pellets were lysed in 20 mM Tris-HCl, pH 7.4, containing 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 50 mM NaF, 10% glycerol, and a mixture of protease inhibitors (Roche Applied Science) for 10 min on ice. Insoluble material was removed by centrifugation at 16,000 ϫ g at 4°C for 20 min. Protein concentrations were determined using a Pierce BCA protein assay kit. Proteins were separated by electrophoresis on 4 -12% polyacrylamide gradient gels or 7% Tris acetate polyacrylamide gels (Invitrogen). For immunoblotting, the proteins were electroblotted onto a polyvinylidene fluoride membrane with a semi-dry blotter (Bio-Rad). The membranes were washed (blocked) overnight in 25 mM Tris-HCl, pH 7.4, containing 0.8% NaCl, 0.02% KCl, 0.1% Tween 20 (TBS-T), and 5% skim milk. The membranes were then incubated with appropriate antibodies against COX-1, COX-2, or actin for 2 h. After the membrane had been rinsed three times for 10 min with TBS-T containing 1% skim milk, it was incubated with goat anti-rabbit or goat anti-mouse horseradish peroxidase-conjugated secondary antibodies (Bio-Rad) for 1 h. After washing four times for 10 min with TBS-T, immunodetection was performed using a Pierce SuperSignal West Pico chemiluminescent substrate kit followed by exposure to x-ray film.
Engineering the ⌬18 muCOX-2 Knock-in Mouse- Fig. 1A illustrates the overall process used to generate the targeting vector. The entire Ptgs2 (COX-2) gene, including exons 1-10, was amplified by PCR using murine 129/SvJ genomic DNA (The Jackson Laboratories, stock number 000691) as a template and P1 (5Ј-TTGTTTTGAGCAGGGGTCTT-3Ј) and P2 (5Ј-AGCAAGAGCAGAACCATTTCA-3Ј) as primers. The primers were designed on the basis of sequence data from the Celera sequence data base and from the description of the original COX-2 null mouse (20). PCR was performed using the Expand long template PCR system (Roche Applied Science). The PCR product was isolated by electrophoresis on a 0.6% agarose gel and subcloned into pCR2.1 (Invitrogen) (Fig. 1A).
Following subcloning, the DNA fragment was verified as muCOX-2 by DNA sequencing of exons 1-10 and the adjoining intron/exon boundaries. The sequences were all identical to the COX-2 sequence in the Celera data base. The sequence encoding amino acids 595-612 of muCOX-2 was deleted using a QuikChange TM site-directed mutagenesis kit (Stratagene) and the mutagenesis oligomer primers 5Ј-CAGCCACCATCAAT-AGGCGTTCAAC-3Ј and 5Ј-CAGCTCAGTTGAACGCCT-ATTGATG-3Ј. PCR was performed under the following conditions: denaturation at 95°C for 1 min, 18 cycles with steps at 95°C for 30 s, 55°C for 1 min, 68°C for 30 min, and 37°C for 30 min. To insert a NotI site in the intron between exons 9 and 10, PCR was performed under the latter conditions again using a QuikChange TM site-directed mutagenesis kit with 5Ј-CAGCTAGTGAGCGGCCGCAAGCCTATGCT-CAAC-3Ј and 5Ј-GTTGAGCATAGGCTTGCGGCCGCTCA-CTAGCTG-3Ј as primers.
The phosphoglycerate kinase neo poly(A) cassette flanked with two loxP sites was inserted at the NotI site in the reverse orientation. The targeting vector was sequenced to ensure that no mutation had been introduced and then linearized by XhoI digestion and electroporated into R1 ES cells (129/SvJ strain). Homologous recombinant clones were obtained from G418resistant colonies. To screen for ES cell clones that had undergone homologous recombination correctly with the targeting vector, mouse genomic DNA was isolated from ES cells clones and screened using PCR with P7 (5Ј-CCACTTGTGTAGCGC-CAAGT-3Ј) and P8 (5Ј-ATGTCTCTGTAGTACATAGCAC-3Ј) as primers (supplemental Fig. S2A). PCR was carried out using Ex Taq TM DNA polymerase (Takara Biochemical). Primers P7 and P8 amplified a 3.3-kbp fragment and were used to sequence the deletion site that is located in exon 10 (data not shown).
Genomic DNA (10 g) from ES cells providing a positive response in the PCR screen was digested with BglII and separated by electrophoresis on 0.6% agarose gels. DNA was transferred to Hybond-N ϩ (Amersham Biosciences) and hybridized with 32 P-labeled 5Ј and 3Ј probes (supplemental Fig. S2B). Two positive clones (WS2D2 and WS2G5) were injected into C57BL/6J blastocysts. Male chimeras were obtained, and appropriate chimeras were mated with C57BL/6J females to obtain heterozygous mice. To identify the heterozygous mice, genomic DNA was extracted from mouse tail and used for screening by PCR with P3 (5Ј-CAGT-GACATCGATGTCATGG-3Ј) and P4 (5Ј-ACAAACACCGG-AGGGAATCT-3Ј) as primers. PCR was carried out using Ex Taq TM DNA polymerase under the following conditions: denaturation at 94°C for 2 min, 30 cycles with steps at 94°C for 30 s, 55°C for 1 min, and 68°C for 1 min, and 68°C 10 min. Primers P3 and P4 amplify a 524-bp native allele fragment and a 470-bp mutant allele fragment (Fig. 1C).
Breeding was performed to remove the neomycin gene and to put the animals into a C57BL/6J genetic background. The conditional heterozygous mice were mated with EIIa Cre mice having a C57BL/6J background to remove the loxP phosphoglycerate kinase neo poly(A) cassette. Genomic DNA extracted from the tails of the offspring were screened by PCR using primers P3 and P4 as described above and by sequence of intron 9 in genomic DNA to identify the removal of the loxP phosphoglycerate kinase neo poly(A) cassette. Homozygous ⌬18 COX-2 knock-in mice used in the studies reported here were offspring that were derived from the breeding of heterozygous mice that had been backcrossed a minimum of six times with C57BL/6J mice. Expression of ⌬18 muCOX-2 protein in mouse tissue was determined by Western blotting using antibodies to residues 595-612 of muCOX-2 and antibodies reactive with residues 583-592 of muCOX-2 as detailed above. All animal studies reported here were performed under protocols approved by University of Michigan Committee on the Use and Care of Animals.
Isolation of Mouse Dermal Fibroblasts-Mouse skin was cut from the backs of euthanized newborn mice and washed three times with ice-cold PBS containing penicillin/streptomycin. The skin was incubated in MEM containing 10% HFBS, 0.05% collagenase type A, and penicillin/streptomycin for 48 h at 4°C.
The skin was moved into a fresh 35-mm dish, broken up using a cell scraper and repeated pipetting, and incubated at 37°C in humidified 5% CO 2 until the fibroblasts grew to 80% confluence.
Cyclooxygenase Assays-Mouse dermal fibroblast cells were cultured in DMEM medium, including 0.2% FBS for 24 h, and then treated with 1 mM aspirin for 30 min to eliminate COX-1 and residual COX-2 activities. The serum-starved cells were then washed three times with DMEM, including 0.2% FBS, to remove the aspirin and then were serum-stimulated by incubation in DMEM, including 20% FBS, to induce muCOX-2 protein. After 3 h, the medium was removed and replaced with 2 ml of MEM containing 10% FBS and 10 nmol of [1-14 C]arachidonic acid (ϳ300,000 cpm) and incubated for 10 min at room temperature. Subsequently, 1 N citric acid and 10% butylated hydroxytoluene were added to stop the reaction (21), and prostaglandins were extracted from cell suspension by adding 8 ml of hexane/ethyl acetate (1:1, v/v) and centrifuged to separate the phases for 10 min at 4°C. The organic phase was transferred into a clean glass tube maintained on ice and the extraction step repeated, and two organic phases were pooled. The organic phases were dried under N 2 ; the samples were reconstituted in 300 ml of ice-cold ethyl ether and applied directly to TLC plates at 4°C. TLC plates were developed at room temperature with ethyl acetate/2,2,4-trimethylpentane/acetic acid/water (110:50:20:100, v/v). Radioactivity on the TLC plate was detected by exposure to x-ray film.
Isolation of Fibroblasts from Mouse Tails-Approximately 2-cm lengths of mouse tails were used to isolate fibroblasts using a protocol described by Salmon et al. (22). The mouse tail was incubated in DMEM containing 10% HFBS, 100 units/ml penicillin, and 100 g/ml streptomycin for 1 h on ice. The tail was washed in ice-cold PBS containing penicillin/streptomycin, washed in 70% ethanol, and washed again in ice-cold PBS containing penicillin/streptomycin. The tail was diced into pieces of less than 0.5 mm 3 and digested overnight with collagenase type II (400 units/ml, 1600 units total per tail) dissolved in DMEM supplemented with 10% HFBS and penicillin/streptomycin at 37°C in a humidified incubator with 5% CO 2 . After collagenase treatment, cells were dislodged from digested tissue by repeated pipetting and filtration using a 70-l nylon cell strainer (Falcon). The samples were centrifuged, and the collagenase solution was aspirated from the cell pellet. Cells were resuspended in DMEM with 10% HFBS and penicillin/streptomycin, split into a 60-mm dish, and incubated until the cells were confluent. Cells (0.7-0.8 ϫ 10 6 cells/dish) were seeded into 100-mm dishes with DMEM containing 10% HFBS and penicillin/streptomycin for subsequent passage. All experiments were done using cells in passages 1-6.
Expression and Degradation of COX-2 Protein in Primary Fibroblasts-Primary cells (0.35-0.40 ϫ 10 6 cells/dish) were seeded in 100-mm dishes. For serum induction of WT and ⌬18 COX-2 proteins, mouse fibroblast cells were grown to about 80% confluence in 100-mm tissue culture dishes in DMEM containing 10% FBS and then serum-starved for 24 h in 8 ml of DMEM containing 0.2% FBS. The medium was changed to DMEM containing 20% FBS with or without an appropriate inhibitor such as KIF or flurbiprofen, and the cells were incu-bated. The cells were subsequently harvested at different times, and protein levels were quantified by Western blot analyses.
To measure the rate of degradation of COX-2 protein via the ERAD pathway in primary fibroblasts, the serum-starved cells were stimulated with 20% serum containing 100 M S-FBP for 4 h, and the medium was then changed to DMEM containing 20% serum, 100 M S-FBP, and 50 M cycloheximide (CHX) with or without 25 M KIF. After a 1-h incubation, the cells were harvested at the indicated times, and cell lysate protein (15 g/lane) was subjected to Western blotting with anti-COX-2 antibodies (␣Gln 583 -Asn 594 ).
Preparation of Solubilized Microsomes from Mouse Organs for Western Blotting-Mouse organs were homogenized on ice with a Polytron homogenizer in ice-cold PBS, including 2 mM EDTA, a mixture of protease inhibitors, and 20 M indomethacin, and then sonicated three times for 5 s. The homogenates were centrifuged at 10,000 ϫ g at 4°C for 20 min. The supernatants were centrifuged at 100,000 ϫ g for 60 min to prepare a microsome fraction. Microsomes was suspended in 20 mM Tris-HCl, pH 7.4, containing 150 mM NaCl, 2 mM EDTA, 1% Nonidet P-40, 50 mM NaF, 10% glycerol, and a mixture of protease inhibitors using a glass-Teflon homogenizer and then incubated for 20 min on ice. The suspension was centrifuged at 100,000 ϫ g at 4°C for 60 min. A fraction from this supernatant (15 g of protein) was used for Western blotting.
Isolation of Peritoneal Residential Macrophage for Western Blotting-After mice were euthanized, the peritoneal cavities were lavaged with a total of 10 ml of ice-cold PBS. Cells were harvested by centrifugation and suspended in RPMI 1640 medium with 100 units/ml penicillin and 0.10 mg/ml streptomycin, and the cells were counted. The cell were collected by centrifugation, and the cell pellet was suspended in RPMI 1640 medium supplemented with 10% heat-inactivated FBS and 100 units/ml penicillin and 0.10 mg/ml streptomycin. Cells (4.5-5.0 ϫ 10 6 ) were plated on 35-mm tissue culture dishes and incubated in 2 ml of medium for 2 h. The media were changed to 2 ml of RPMI 1640 medium containing 10% HFBS, penicillin/ streptomycin, 1 g/ml LPS with or without COX inhibitor, and/or KIF for 24 h.
Isolation of Resident Peritoneal Macrophages and Assay of PGs-After the mice were euthanized, resident peritoneal macrophages were collected from untreated male mice (ϳ12 weeks of age) by peritoneal lavage with 5 ml of ice-cold PBS (23). Cells were harvested by centrifugation and suspended in RPMI 1640 medium with 100 units/ml penicillin and 0.10 mg/ml streptomycin to count the cells. Then 2.0 ϫ 10 5 cells were plated in 96-well plates and incubated for 30 -60 min in a CO 2 incubator at which time the media were removed and fresh RPMI 1640 medium supplemented with 10% HFBS, penicillin, and streptomycin was added. After 24 h, the media were changed to 100 l of RPMI 1640 medium containing penicillin and streptomycin with or without inhibitors, and the cells were incubated for 30 min. Then an additional 100 l of RPMI 1640 medium containing penicillin, streptomycin, and 200 ng/ml LPS was added to each well, and the cells were incubated for different times, and the supernatants were harvested and kept at Ϫ80°C until analyzed. PG levels in the culture medium were determined using enzyme-linked immunosorbent assay kits from Assay Designs (Ann Arbor, MI).
Measurement of Urinary PGs-Three male or female mice, either WT or ⌬18 COX-2 knock-in mice, were placed in metabolic cages for 24-h urine collections. Samples were collected from a total of nine 12-week-old mice in each of the four groups. To collect the urine, three mice were placed in each metabolic cage (i.e. three metabolic cages were used for each group). Urine was collected each day for 3 days, and then randomly five samples from each group were analyzed; thus, two samples of the five were from the same cage but were collected on different days. Urine was centrifuged to remove any precipitated matter. The supernatants were kept Ϫ80°C until analyses. The urinary metabolite of prostaglandin E 2 was measured using a liquid chromatography/tandem mass spectrometric method described previously (24).
Implantation of Telemeters and Monitoring Body Temperature-Radiofrequency telemeters (model ETA10-F20, Data Sciences International, St. Paul, MN) were surgically implanted in the peritoneal cavity of male mice (16 -20 weeks old) under anesthesia, as described previously (25,26). At least 21-28 days were allowed for recovery from surgery before the experiments began. Temperature data were acquired every 10 min as detailed previously (26). Circadian temperature variation was assessed by monitoring the body temperature of undisturbed animals at 10-min intervals over two consecutive 24-h periods. These 48-h data were averaged and represented a single 24-h period. Base-line body temperature was defined as the average body temperature for the 60 min before any treatment. Mice were subsequently injected intraperitoneally with pyrogen-free saline, and body temperature was recorded for 24 h. The same mice were then injected intraperitoneally with 10 g of LPS (E. coli O111:B4, Sigma) dissolved in pyrogen-free saline, and body temperature was measured for another 24 h. Each mouse administered pyrogen-free saline served as its own control for subsequent LPS exposure. Treatments were conducted at 9:00 a.m. to reduce the effects of circadian variation on the results.
For either WT (n ϭ 6) or ⌬18 COX-2 mice (n ϭ 5), statistical comparisons between the mean hourly body temperature data following saline and LPS treatment were conducted using a two-way analysis of variance with Bonferroni post hoc test. For comparisons between WT and ⌬18 COX-2 mice treated with LPS, differences between pyrogen-free saline-treated body temperature and matched LPS-treated body temperature measurements were calculated (denoted ⌬T b ), and then sequential 10 min ⌬T b values were averaged hourly to calculate mean hourly ⌬T b . Matched hourly ⌬T b data were compared by Student's t test. p Յ 0.05 was considered statistically significant.
Cardiovascular Parameters of ⌬18 COX-2 Mice-Cardiovascular parameters were measured by the University of Michigan Medical School Integrative Genomics using nine, 12-week-old mice in each group (27). Ejection fractions were determined with echocardiograms (27,28). Values are reported as mean Ϯ S.D. NOVEMBER

RESULTS
Preparation of ⌬18 COX-2 Mouse-Previous studies have indicated that human (hu) COX-2 expressed heterologously in HEK293 cells undergoes N-glycosylation of Asn 594 as a prelude to entry into the ERAD pathway. Additional studies with mutants of murine (mu) COX-2 also expressed in HEK293 cells corroborate these results (supplemental Fig. S1).
To determine the importance in COX-2 biology of the 27amino acid instability motif (27-IM) located near the C terminus of the protein, we prepared a knock-in (⌬18 COX-2) mouse lacking 18 residues (595-612) of the 27-IM. The ⌬18 COX-2 protein is catalytically indistinguishable from native COX-2, but unlike wild type (WT) COX-2, it is unable to undergo N-glycosylation of Asn 594 and degradation via the ERAD pathway (17). Fig. 1A illustrates how the targeting vector was constructed. To generate the ⌬18 COX-2 knock-in mice, homologous recombination in ES cells was used to create the mutant allele (Fig. 1B), and selected recombinant ES cells (supplemen-tal Fig. S2, A and B) were used for the generation of the ⌬18 COX-2 knock-in mice. To remove the Lox P phosphoglycerate kinase neo poly(A) cassette, the knock-in mice were mated with Cre recombinase expressing EIIa Cre transgenic mice, and the genotypes of the mice were identified by PCR (Fig. 1C). PCR fragments of appropriate sizes (524 and/or 470 bp) were observed for DNA isolated from mice expressing only WT COX-2, only ⌬18 COX-2, or both WT and ⌬18 COX-2.
To determine the level of COX-2 protein expression, we cultured dermal fibroblasts from newborn WT, ⌬18 COX-2, and heterozygous mice. The cultured cells were subjected to serum starvation, which typically reduces the levels of native COX-2 in normal fibroblasts to near zero and then, as shown in Fig. 1D, the cells were stimulated with serum to induce COX-2 gene transcription. The ␣Ser 598 -Lys 612 antibody cannot recognize ⌬18 COX-2 protein, whereas the ␣Gln 583 -Asn 594 antibody can recognize both WT COX-2 and ⌬18 COX-2 protein. As expected, there is no COX-2 immunoreactivity in cell  extracts from ⌬18 COX-2 mice in Western blots performed with the ␣Ser 598 -Lys 612 antibody. Western blots performed with the ␣Gln 583 -Asn 594 antibody indicate that ⌬18 COX-2 levels are higher than WT COX-2 levels in the serum-stimulated fibroblasts.
As shown in Fig. 1E, serum-stimulated dermal fibroblasts from WT COX-2, ⌬18 COX-2, and heterozygous mice retain COX-2 catalytic activity in approximate proportion to the amount of immunoreactive COX-2 protein and form the same products from arachidonic acid (e.g. PGE 2 ). These results are consistent with previous in vitro data showing that WT COX-2 and ⌬18 COX-2 are kinetically indistinguishable (17).
The data in Fig. 1 indicate that engineered ⌬18 COX-2 mice exhibit homozygous expression of a functional ⌬18 COX-2 of the appropriate size. ⌬18 COX-2 knock-in mice were backcrossed at least six generations into the C57BL/6J mouse strain prior to performing the experiments described below.
Cardiovascular parameters for WT and ⌬18 COX-2 male and female mice are summarized in Table 1. No statistically significant differences were observed in heart rates, blood pressure, or cardiac output (i.e. ejection fraction) with age-and sexmatched mice. We did observe that the mutant mice showed smaller variations in blood pressure. We were somewhat surprised by the lack of any obvious cardiovascular phenotype because we had expected ⌬18 COX-2 mice to overexpress COX-2 and to overproduce vascular PGI 2 (29), which, if uncompensated, would have led to changes in cardiovascular performance (e.g. decreased blood pressure).
The levels of the urinary metabolite of PGE 2 were significantly higher (ϳ50%) in 12-week-old male ⌬18 knock-in mice than in male WT control mice (Fig. 2). Approximately half of urinary PGE 2 metabolites are formed via COX-2 in WT mice (30), so a 50% increase in the total PGE 2 metabolites in the knock-in mice may represent an approximate doubling of COX-2-derived PGE 2 formation. This would be consistent with overexpression of COX-2. It is not known why the levels of urinary PG metabolites are so much higher in the male than the female mice in C57BL/6J mice. This is clearly dependent on the genetic background of the mice because females have higher levels of PGE 2 urinary metabolites than the males in the DBA/1 mouse background (31).
COX-2 Expression in Tail Fibroblasts from WT Versus ⌬18 COX-2 Knock-in Mice-Primary fibroblasts prepared from the tails of WT and ⌬18 COX-2 mice were used to evaluate the importance of residues 595-612 of COX-2 in the degradation of the protein via the ERAD pathway (Fig. 3, A-C). In these experiments serum-starved cells were treated with 20% serum to induce COX-2 protein expression. The experiments depicted in Fig. 3, A and B, were performed in the presence or absence of the following: (a) kifunensine (KIF), which blocks processing of N-glycosylated carbohydrate and retards entry of N-glycosylated protein into the ERAD pathway, and/or (b) the nonspecific COX inhibitor S-flurbiprofen, which prevents substrate turnover. No inhibitor of translation was included in the experiments shown in Fig. 3, A and B. In the absence of KIF or flurbiprofen, WT and ⌬18 COX-2 protein undergo degradation in a temporally similar manner, although the level of COX-2 expression is much higher with ⌬18 COX-2. The disappearance of immunoreactive COX-2 is completely blocked by flurbiprofen in fibroblasts from ⌬18 COX-2 but not WT COX-2 mice; flurbiprofen slows but does not prevent the disappearance of WT COX-2. In the presence of KIF alone or KIF plus S-flurbiprofen, two immunoreactive COX-2 bands are present in fibroblasts from WT but not ⌬18 COX-2 mice. The upper band results primarily from N-glycosylation of Asn 594 in WT COX-2. ⌬18 COX-2 is missing the consensus sequence for N-glycosylation of Asn 594 and cannot be glycosylated on this residue. KIF by itself does not stabilize ⌬18 COX-2.
The experiment shown in Fig. 3C was performed in the presence of both an inhibitor of protein synthesis cycloheximide FIGURE 2. Urinary PGE 2 metabolites from adult male WT or ⌬18 COX-2 mice. Analyses of urinary PGE 2 metabolites (PGE-M) were performed as described in detail under "Experimental Procedures" using urine collected from 12-week-old mice. Data are means Ϯ S.E. * indicates a difference from the value for the male WT; ** indicates a difference from the value for male WT but no difference between the values for the WT and ⌬18 female mice (p Ͻ 0.05).

TABLE 1 Cardiovascular parameters of WT and ⌬18 COX-2 mice
Cardiovascular parameters were measured using nine, 12-week-old mice in each group except for the ⌬18 COX-2 female group where there were eight animals. Details are provided under "Experimental Procedures." The abbreviations used are as follows: bpm, beats/min; bp, blood pressure; EF%, ejection fraction. Mean Ϯ S.D.

Animal
Body weight Heart rate Systolic bp EF (CHX) and FBP, which inhibits degradation via the suicide pathway. It can be seen that WT COX-2 is degraded, and degradation is inhibited by KIF. However, ⌬18 COX-2 is not degraded under these conditions. The data in Fig. 3, A-C, indicate that ⌬18 COX-2 protein can be degraded via the substratedependent degradation pathway but not the ERAD pathway. The lack of a functional 27-IM caused by deletion of residues 595-612 stabilizes ⌬18 COX-2 in cells from the knock-in mouse where ⌬18 COX-2 is under the control of the WT muCOX-2 promoter. These results obtained with mouse tail fibroblasts are consistent with those obtained in vitro in HEK293 cells heterologously expressing WT or mutant forms of human COX-2 (18). COX-2 Expression in Brain, Kidney, Spleen, and Heart-Western transfer blotting was performed on extracts of several tissues from WT and ⌬18 COX-2 mice to assess the relative importance of the ERAD pathway in COX-2 degradation (supplemental Fig. S3 and Fig. 4). The basal levels of immunoreactive COX-2 are lower in brain tissue from WT COX-2 mice than ⌬18 COX-2 mice suggesting that the ERAD pathway is important in COX-2 degradation in brain. In contrast, the basal levels of immunoreactive COX-2 are similar in heart (supplemental Fig. S4), kidney, and spleen from WT COX-2 and ⌬18 COX-2 mice.
The two panels labeled Brain in Fig. 4 represent the same Western blot with different exposure times; the lower Brain panel is included to permit comparisons of the levels of WT and ⌬18 COX-2 in brain, kidney, and spleen. The basal levels of WT COX-2 are low in brain, intermediate in kidney, and relatively high in spleen. LPS treatment of normal mice leads to increased expression of WT COX-2 in brain and spleen but not kidney. Treatment of animals with ibuprofen at analgesic doses (32), which would be expected to interfere with substrate-dependent degradation of COX-2, increased the levels of WT COX-2 in brain from WT animals, and the effect was observed with and without (i.e."0" time point) LPS. In contrast, ibuprofen treatment had no obvious effect on the levels of WT COX-2 in kid-ney or spleen from either LPS-treated or untreated animals. The levels of immunoreactive ⌬18 COX-2 in brain, kidney, and spleen were unchanged by LPS treatment. Overall, the results shown in Fig. 4 suggest that both the ERAD and substrate turnoverdependent pathways are important in COX-2 degradation in brain. It is unclear what pathways are involved in COX-2 protein degradation in kidney and spleen. Compensatory mechanisms involving enhanced expression and/or decreased degradation of COX-2 must exist in kidney and spleen to maintain COX-2 levels constant in these tissues (e.g. when the ERAD pathway is disrupted in the ⌬18 COX-2 mice).  . Expression of COX-2 in tissues from WT and ⌬18 COX-2 mice treated with endotoxin with or without ibuprofen. Microsomal protein was prepared from whole brain, whole kidney, and whole spleen obtained from 12-week-old male WT or ⌬18 COX-2 mice. Animals were subjected to an intraperitoneal injection with LPS (10 g) and were euthanized, and tissues were taken at the indicated times following the LPS injection. Half of the animals received ibuprofen (ϩIBP) in their drinking water (0.2 mg/ml) beginning 2 days prior to the LPS treatment and also received an intraperitoneal injection of ibuprofen (40 mg/kg) in saline 30 min prior to the LPS injection. C, control WT muCOX-2 (0.05 g); the adjacent lane contained molecular weight standards (not visible). The same amounts of microsomal protein were applied in each lane. The two Brain panels represent different exposure times for the same Western blot. The columns in each of the panels represent tissue from the same mouse (i.e. one mouse was used for each time point). Anti-COX-2 (Novus Biologicals) antibody was used for Western blotting.
Prostaglandin Formation by LPS-treated Peritoneal Macrophages-When peritoneal macrophages isolated from WT and ⌬18 COX-2 mice were treated with LPS, cells from the ⌬18 COX-2 mice produced considerably more PGE 2 and PGI 2 (measured as 6-keto-PGF 1␣ ) than cells from WT mice (Fig. 5, A and B). Again, this is consistent with the ERAD pathway playing an important role in COX-2 degradation in macrophages. Macrophages treated with KIF to slow degradation of COX-2 by interfering with ERAD degradation produced slightly higher amounts of PGE 2 in response to the Ca 2ϩ ionophore A23187 (Fig. 5A, lanes  10 and 11).
It can be seen that the levels of immunoreactive COX-2 higher at base line in macrophages from both male and female ⌬18 COX-2 mice are relatively unaffected by LPS, flurbiprofen, and/or KIF (Fig. 5C). This is similar to what is observed with brain, kidney, and spleen from ⌬18 COX-2 mice (Fig. 4). In contrast, COX-2 levels in macrophages from WT mice increase in a predictable manner following treatments with LPS, LPS and flurbiprofen, and LPS plus KIF (18) (Fig. 5C); the lower band of the protein doublet seen when KIF is included is COX-2 with three asparagine residues glycosylated, whereas glycosylation at a fourth asparagine, primarily Asn 594 , results in formation of the band with slower electrophoretic mobility (18).
Febrile Response-COX-2 is known to be important in thermoregulatory responses to inflammatory insults such as infection (33,34). Therefore, experiments were conducted to investigate the impact of the ⌬18-modified COX-2 on thermoregulatory physiology both at rest and under inflammatory stress, induced by intraperitoneal injection of LPS. As depicted in Fig. 6A, both WT and ⌬18 COX-2 mice exhibited normal basal body temperatures and intact circadian variability. Although both WT and ⌬18 COX-2 mice became febrile following LPS administration (Fig. 6, B and C), significant differences were noted in these responses. In the WT animals, core body temperature was significantly greater after LPS injection compared with saline administration beginning ϳ5 h after the injections and persisting for 3 h (Fig. 6B). In contrast, sig-  WT and ⌬18 COX-2 mice that had had telemeters implanted were treated as detailed under "Experimental Procedures." A, body temperatures of untreated mice demonstrate normal circadian variability in WT (thin line) and ⌬18 COX-2 mice (thick line); B, WT mice were injected with pyrogen-free saline (thin line) or LPS (thick line, 10 g) at time 0. The horizontal line with asterisk denotes statistically significant differences in mean hourly temperature (p Ͻ 0.01). C, ⌬18 COX-2 mice were injected with pyrogen-free saline (thin line) or LPS (thick line, 10 g) at time 0. The horizontal bar with asterisk denotes statistically significant differences in mean hourly temperature (p Ͻ 0.01). Body temperature measurements shown in the figure represent mean and S.E. of values obtained from six WT mice and five ⌬18 COX-2 mice. D, comparison of the mean hourly difference in body temperature (⌬T b ) between pyrogen-free saline and LPS treatment for WT (thin line) and ⌬18 COX-2 mice (thick line). Mean hourly ⌬T b values were calculated as described under "Experimental Procedures." Horizontal black lines with asterisks denote hour periods during which statistically significant differences (p Ͻ 0.05) were revealed between WT and corresponding ⌬18 COX-2 mice. Dark bars on the x axis denote the dark period of the light/dark cycle. nificant differences in core body temperature between LPS and saline administration became evident ϳ4 h after injection in the ⌬18 COX-2 mice and remained statistically significant for 7 h (Fig. 6C). Increases in body temperature induced by LPS over saline alone were significantly more pronounced in the ⌬18 COX-2 mice versus the WT animals for as long as 20 h after the treatments were administered (Fig. 6D). 3

DISCUSSION
In this study, we successfully generated ⌬18 COX-2 knock-in mice and established that the ⌬18 COX-2 protein has COX activity and that ⌬18 COX-2 is more stable than the WT enzyme in at least some tissues (i.e. brain). In combination with other in vitro studies of the mechanism of COX-2 protein degradation (17,18), our results indicate that the ERAD pathway participates in COX-2 protein degradation in vivo. Substratedependent loss of WT COX-2 is inhibited by flurbiprofen in murine 3T3 fibroblasts and in HEK293 cells heterologously expressing human COX-2 (18). Similarly, we find that flurbiprofen leads to accumulation of immunoreactive COX-2 protein in both cultured fibroblasts and macrophages from WT and ⌬18 COX-2 mice. WT COX-2 also accumulates in brain tissue from normal mice treated with ibuprofen, a nonsteroidal anti-inflammatory drug closely related to flurbiprofen. These data suggest that substrate-dependent loss of COX-2 occurs in vivo. Like flurbiprofen, KIF treatment of cultured cells from WT and ⌬18 COX-2 mice also causes the accumulation of COX-2 protein even in the presence of flurbiprofen.
The manner in which the ERAD and substrate turnover-dependent degradation pathways are related is unclear. They may be parallel noninteracting pathways of degradation. However, it is also possible that there is an additional level of complexity such that N-glycosylation at Asn 594 somehow interferes with the ability of the enzyme to encounter substrate and undergo substrate-dependent degradation.
Our most unexpected finding was the lack of accumulation of ⌬18 COX-2 above levels seen with WT COX-2 in several tissues, including kidney, heart, and spleen. This implies that there are previously unappreciated processes involved in maintaining COX-2 protein levels that presumably involve cross-talk between pathways involved in COX-2 synthesis and degradation. These changes could involve decreased rates of transcription and/or translation (6,7,16). We did not measure COX-2 mRNA levels in our samples. There are examples in the literature suggesting that there is cross-talk between PG production and COX-2 gene expression (6,7).
When ⌬18 COX-2 knock-in mice are subjected to a physiological stress, specifically treatment with bacterial endotoxin, the febrile response is enhanced, and ⌬18 COX-2 accumulates in brain. Similarly, resident peritoneal macrophages prepared from ⌬18 COX-2 mice produce more PGs than cells from WT COX-2 mice upon endotoxin treatment. Again, this is consistent with overexpression of ⌬18 COX-2 in cells from knock-in mice.
COX-2-derived PGE 2 is a central mediator in the febrile response to infection and inflammation (33,34), but evidence suggests that eicosanoids do not control normal, nonfebrile thermoregulation (35). Our results support this paradigm, as no differences were observed in either basal body temperature or circadian temperature variation between WT and ⌬18 COX-2 mice. Notably, however, mice harboring the ⌬18 COX-2 variant exhibited aberrant thermoregulation in the face of systemic inflammation. We found a prolonged febrile response in ⌬18 COX-2 mice after intraperitoneal LPS administration compared with WT COX-2 mice, and this correlated with higher levels of ⌬18 COX-2 in the brains of the transgenic animals relative to the levels of normal COX-2 in the WT mouse brains. These data support a role for COX-2 in the febrile response to LPS and suggest that the ERAD pathway for COX-2 degradation plays a role in controlling the amplitude and duration of the febrile response.
The ⌬18 COX-2 knock-in mouse has what appears to be a largely normal phenotype. Nonetheless, there is about a 50% increase in the level of the major PGE 2 urinary metabolite in male ⌬18 COX-2 knock-in mice. This is consistent with an increase in the level of COX-2 protein and activity in ⌬18 COX-2 knock-in mice. Mice 3-6 months of age were used in these studies, and it is not known if the knock-in mice would display any abnormalities with age. It is known that cancer patients and individuals with rheumatoid arthritis have elevated levels of PGE 2 production. ⌬18 COX-2 mice that chronically overproduce prostaglandins via COX-2 may be useful in studying the development of chronic inflammation and colon cancer.