Absence of Myeloid COX-2 Attenuates Acute Inflammation But Does Not Influence Development of Atherosclerosis in Apolipoprotein E Null Mice

Abstract

Objective

The role of myeloid cell cyclooxygenase-2 (COX-2) in the progression of atherosclerosis has not been clearly defined.

Methods and Results

We investigated the role of COX-2 expressed in the myeloid lineage in the development of atherosclerosis using a myeloid-specific COX-2^−/− (COX-2^−M/−M) mouse on a hyperlipidemic apoE^−/− background (COX-2^−M/−M/apoE^−/−). Myeloid COX-2 depletion resulted in significant attenuation of acute inflammation corresponding with decreased PGE₂ levels in an air pouch model. COX-2 depletion in myeloid cells did not influence development of atherosclerosis in COX-2^−M/−M/apoE^−/− when compared to apoE^−/− littermates on either chow or western diets. The unanticipated lack of contribution of myeloid COX-2 to the development atherosclerosis is not due to altered maintenance, differentiation, or mobilization of myeloid and lymphoid populations. Moreover, myeloid COX-2 depletion resulted in unaltered serum prostanoid levels and cellular composition of atherosclerotic lesions of COX-2^−M/−M/apoE^−/− mice.

Conclusions

Our results suggest that COX-2 expression in myeloid cells, including macrophages, does not influence the development of atherosclerosis in mice.

Atherosclerosis is the primary mechanism underlying the development of coronary heart disease (CHD).¹ Atherosclerosis is a chronic, inflammatory disease that is often correlated with increased low-density lipoprotein (LDL) and decreased high-density lipoprotein (HDL) levels.² During the progression of the disease, LDL is oxidized and becomes trapped in the aortic subendothelial space, resulting in increased inflammation and leukocyte recruitment. The lesions that develop in the aortic root and proximal aorta are characterized by infiltration of macrophages and other immune cells, including polymorphonuclear neutrophils (PMN) and lymphocytes.³

Macrophages are one of the major cell types found in atherosclerotic lesions. In addition to their role in the inflammatory response, macrophages play an important anti-atherogenic role by preventing lipid accumulation through lipid efflux to acceptors such as HDL and apolipoprotein A-I (apoA-I).⁴ An imbalance between lipid uptake and efflux often results in increased lipid accumulation in macrophages, resulting in a pro-atherogenic formation of foam cells and early fatty streaks.⁵

While macrophages are the most well-studied inflammatory cells in atherosclerotic plaques, other myeloid cells including PMN impact the progression of atherosclerosis.⁶ PMN release myeloperoxidases at the lesion site which may participate in oxidative modification of LDL, accelerating the development of the atheroma.⁷

Cyclooxygenase-2 (COX-2) is the inducible form of the COX enzyme and is responsible for the rate-limiting step in the conversion of arachidonic acid to prostanoids (prostacyclin, prostaglandins, and thromboxanes).⁸ COX-2 is highly expressed in macrophages, PMN, and other leukocytes in both acute and chronic inflammation and its expression is well-characterized in chronic inflammatory disorders such as atherosclerosis.⁹ COX-2 is induced in human¹⁰ and mouse¹¹ atherosclerotic lesions and is expressed in macrophages, endothelial cells, and smooth muscle cells.¹⁰

Clinical data have implied a potential anti-atherogenic role for COX-2 since prolonged use of COX-2-specific inhibitors (coxibs) results in an increase in cardiovascular events, such as heart attacks and strokes.^12–14 While these data clearly implicate COX-2 in cardiovascular disease, no mechanism has been proposed to fully explain the role of COX-2 in atherosclerosis, the underlying cause of CHD. Although there have been a large number of studies performed to attempt to elucidate the mechanism, there is no consensus in the literature. COX-2 inhibition has been shown to either decrease^15–17, increase¹⁸, or have no effect^19–21 on the progression of atherosclerosis in mice, depending on the mode and duration of COX-2 inhibition as well as the types of mouse models used. Similarly, the role of macrophage COX-2 in atherosclerosis has been explored, but has not been conclusively elucidated.^{15, 16, 22}

In addition to their role in lipid metabolism at the site of the atherosclerotic lesions described above, macrophages, as well as other myeloid cells, play a large role in the regulation of cytokine production and other inflammatory mediators, including COX-2-derived prostanoids. In this paper, we used a myeloid cell-specific COX-2^−/− (COX-2^−M/−M) mouse to determine whether COX-2 in myeloid cells, including macrophages, plays a role in the development of atherosclerotic lesions in mice.

Methods

Animals

COX-2 floxed “knock-in” mice, in which exons 4 and 5 of mouse COX-2 are flanked by loxP sites, have been described previously.²³ Lysozyme M (LysM) Cre mice²⁴ on a C57BL6/J background were generously provided by Dr. Yibin Wang. ApoE^−/− mice on C57BL6/J background were obtained from Jackson Labs (Bar Harbor, ME). To obtain COX-2^−M/−M mice on an apoE^−/− background (COX-2^−M/−M/apoE^−/−), COX-2 floxed mice were crossed to LysM Cre mice to obtain COX-2^−M/−M (COX-2^flox/flox, LysM^Cre/+) mice and wild type (COX-2^flox/flox, LysM^+/+) littermates. COX-2^−M/−M mice were further crossed to apoE^−/− mice to obtain COX-2^−M/−M/ apoE^−/− (COX-2^flox/flox, LysM^Cre/+, apoE^−/−) mice. ApoE^−/− controls used in experiments were COX-2^flox/flox, LysM^+/+, apoE^−/− littermates. All mice were maintained on a standard chow (6% fat) diet. In atherosclerosis studies, female mice were maintained on chow diet for 14 weeks or “western diet” (42% fat, 0.15% cholesterol, Harlan Teklad, Madison, WI) for 18 weeks starting at the age of 6 weeks. Serum samples were isolated from overnight fasted mice, cryopreserved in 10% sucrose and kept at −80°C until use.

Primary Macrophages and Treatments

Primary macrophages were obtained by lavage 3 days after intra-peritoneal injection of 3% thioglycollate into mice. Cells were plated at a concentration of 750,000–1,000,000 cells/mL, and macrophages were allowed to adhere for 2 hours at 37°C and 5% CO₂. Unattached cells were washed out and the adherent macrophages were grown in DMEM medium (GIBCO-BRL, Grand Island, NY) with 10% FBS for 2 days at 37°C and 5% CO₂.

Cells were used for experiments 2 days after adhering. Prior to addition of tumor necrosis factor α (TNFα) (Invitrogen, Carlsbad, CA) cells were incubated with serum-free DMEM with 0.2% fatty-acid free bovine serum albumin (BSA) (Sigma-Aldrich, St. Louis, MO) overnight. The next day, medium was changed to fresh serum-free DMEM with 0.2% BSA and TNFα was added for the indicated time periods. To obtain protein, cells were lysed with Passive Lysis Buffer (Promega, Madison, WI) plus protease inhibitor cocktail (Roche, Indianapolis, IN).

Western Blotting

Protein samples were fractionated on 8% SDS-polyacrylamide gels and electroblotted onto nitrocellulose membranes (Bio-Rad, Hercules, CA) using a wet transfer apparatus (Bio-Rad, Hercules, CA). Membranes were blocked with phosphate-buffered saline (PBS), 0.1% Tween, and 5% nonfat dried milk for 1 h; and then incubated overnight at 4°C with rabbit anti-mouse COX-2 polyclonal antibody (Cayman Chemical, Ann Arbor, Michigan) diluted 1:1000 in PBS, 0.1% Tween, or monoclonal β-actin antibody (Sigma) diluted 1:5000 in PBS, 0.1% Tween. Membranes were washed 3 times (10 min/wash) with PBS, 0.1% Tween, incubated with horseradish peroxidase-conjugated anti-rabbit IgG or anti-mouse IgG (Amersham, Pittsburgh, PA) at 1:10,000 dilution in PBS, 0.1% Tween, and 2% nonfat dried milk for an additional 1 hour. Membranes were washed again and proteins detected using chemiluminescence (Millipore, Billerica, MA).

ELISA

ELISA for PGE₂, TNFα, IL-1β, IL-6, 6-keto-PGF_1α, and TXB₂ were done according to manufacturer’s instructions (Assay Designs, Ann Arbor, MI) on cryopreserved serum and/or exudate samples.

Air Pouch Model of Acute Inflammation

Air pouch model for acute inflammation was performed as described previously.²⁵ Briefly, 3 mL of sterile air was injected into backs of mice subcutaneously and pouches were allowed to form for 4 days. Two mL sterile saline or 2% zymosan (in sterile saline) was injected into the air pouch. Air pouches were flushed with 2 mL DMEM with 10% FBS 24 hours later. An aliquot of the exudates was taken for cell counting using a Hemavet cell counter (Drew Scientific, Waterbury, CT). The remaining cells were collected by centrifuging the exudates, plating out cells, allowing them to adhere for 2 hours, and collecting adherent and non-adherent cells as a myeloid and a non-myeloid population, respectively. The remaining exudate (collected after centrifugation) was immediately frozen at −80°C for further analysis of cytokines and prostanoids.

Aortic Root Lesions

Atherosclerotic lesion area in the aortic root was determined as described previously. ²⁶ Briefly, heart and proximal aorta from mice were obtained and embedded in OCT compound. Serial 10 µm-thick cryosections from the middle portion of the ventricle to the aortic arch were collected, mounted on pre-coated slides, and stained with Oil Red O and hematoxylin. The lipid-containing area on each section, centered around the aortic valves, was determined in a blinded fashion, using an ocular with a 20 × 20 µm² grid on a light microscope. The average lesion area per aorta, calculated from 5–10 sections of each aorta, was scored.

Serum Lipids

Serum lipids were determined as described previously.²⁷

Immunohistochemistry

Fresh-frozen aortic root sections were stained for CD68, α-actin, and COX-2. Briefly, after fixation in ice-cold acetone, sections were blocked in 4% BSA plus 10% goat serum for 3 hours at room temperature. Rat anti-mouse CD68 (1:100, Serotec, Oxford, UK), rabbit anti-mouse α-actin (1:100, Spring Bioscience, Pleasanton, CA), and rabbit anti-mouse COX-2 (1:200, Cayman Chemical) were used with an overnight incubation at 4°C. Goat anti-rat and goat anti-rabbit alkaline phosphatase secondary antibodies (Jackson ImmunoResearch, West Grove, PA) were used at 1:200 with a 1 hour incubation at room temperature. Immunostaining was visualized using Vector Red substrate plus levamisole to inhibit endogenous alkaline phosphatase activity (Vector Laboratories, Burlingame, CA). Immunostaining was quantified in a blinded fashion using an ocular with a 20 × 20 µm² grid on a light microscope.

Flow Cytometry

Single cell suspensions were prepared from each tissue prior to performing flow cytometry. Spleens and periaortic lymph nodes were gently dissociated between two glass slides for single cell isolation. Peripheral blood was isolated directly into BD Vacutainer K2 EDTA tubes (BD Biosciences, San Jose, CA) at room temperature and red blood cells (RBC) were lysed (Sigma-Aldrich, Saint Louis, MO). Mouse bone marrow cells were harvested by flushing the tibias and femurs of the mice followed by RBC lysis. Single cell suspensions were filtered and incubated for 30 min on ice with the following purified and FITC-, PE-, APC-, PerCP-Cy5.5-, PE-Cy7- and APC-Cy7 conjugated antibodies: CD45, CD11b, Gr-1, and F4/80 were purchased from eBioscience (San Diego, CA); Ly6C, CD4, and 7-AAD were purchased from BD Biosciences). Cells were washed twice prior to analysis on the BD LSR-II flow cytometer (BD Biosciences). Data were analyzed with FlowJo software (TreeStar, Ashland, OR).

For sorting experiments, single cells suspensions of bone marrow were either untreated or treated with LPS (2 µg/mL) (Sigma) in DMEM 10% FBS at 37°C for 4 hours. Cell suspensions were then incubated with CD11b, Ly6C, and Gr-1 antibodies as described above. Cells were sorted through a BD FACSAriaII cell sorter (BD Biosciences) and genomic DNA (Qiagen, Valencia, CA) and protein were harvested immediately.

Genomic PCR

Cre-mediated recombination of genomic DNA was detected using the following primers flanking the COX-2 floxed region and designed for COX-2 introns 3 and 5, respectively: 5´- AAAGTTCAGCCATTGTACAGCAGGG-3´ and 5´- GAATCTCCTAGAACTGACTGG-3´. PCR conditions used were 95°C 3 minutes, 30 cycles of 95°C 30 sec, 60°C 1 min and 72°C 2 min. The larger band (2067bp) corresponds to the non-recombined allele and the smaller band (651bp) corresponds to a recombined allele.

Statistics

Statistical significance was determined by Student’s T-test. Statistical significance was defined as p-value < 0.05.

Results

To characterize the COX-2^−M/−M mouse model, peritoneal macrophages were isolated from wild type and COX-2^−M/−M mice and COX-2 protein expression and PGE₂ levels were determined. Since COX-2 is induced by inflammatory stimuli, macrophages were treated with 50 ng/mL TNFα for 4 hours to determine COX-2 protein expression in cell lysates and for 18 hours to determine accumulation of PGE₂ in the culture supernatants. Macrophages from COX-2^−M/−M mice did not express COX-2 protein or produce PGE₂ when stimulated with TNFα (Figure 1), suggesting that the COX-2 gene was successfully knocked out in macrophages of COX-2^−M/−M mice.

Figure 1.

Macrophages from COX-2^−M/−M mice do not express COX-2 or produce PGE₂. Peritoneal macrophages from wild type (WT) and COX-2^−M/−M mice were either untreated (NT) or treated with TNFα (50 ng/mL) for 4 hours and COX-2 expression was determined by western blot (A) as described in Methods. Peritoneal macrophages from WT and COX-2^−M/−M mice were either untreated (NT) or treated with TNFα (50 ng/mL) for 18 hours and PGE₂ in the culture supernatant was determined by ELISA (B) as described in Methods. Data are represented as the average concentration plus one standard deviation. * = p-value < 0.05, † = p-value < 0.05 compared to NT.

COX-2^−M/−M mice were crossed to hyperlipidemic apoE^−/− mice to obtain COX-2^−M/−M/apoE^−/− mice. All of the subsequent experiments were carried out using COX-2^−M/−M/apoE^−/− mice and apoE^−/− littermates.

Macrophages and other myeloid cells regulate a variety of acute inflammatory processes primarily through the production of PGE₂.^{28, 29} To determine if myeloid COX-2 affects acute inflammation in vivo, the air pouch model of acute inflammation was studied in COX-2^−M/−M/apoE^−/− mice and apoE^−/− littermates. Air pouches made on the backs of either COX-2^−M/−M/apoE^−/− mice or apoE^−/− controls were injected with either zymosan or saline, and inflammatory exudates were collected 24 hours later. Exudates from saline treated air pouches contained only negligible numbers of infiltrating cells and concentrations of inflammatory mediators (Figure 2). Injection of zymosan into air pouches resulted in a large infiltration of leukocytes; however, the total number of leukocytes in zymosan-injected air pouches in COX-2^−M/−M/apoE^−/− mice was no different from those in air pouches of apoE^−/− controls (Figure 2A). Both adherent (mostly myeloid) cells and non-adherent (mostly non-myeloid) cells obtained from the air pouch exudates of apoE^−/− mice injected with zymosan express COX-2 protein. In contrast and as expected, COX-2 protein expression was only detectable in non-myeloid cells from COX-2^−M/−M/apoE^−/− mice while COX-2 protein was depleted in the infiltrating myeloid cells (Figure 2B). Moreover, PGE₂ in COX-2^−M/−M/apoE^−/− exudates is decreased compared to exudates from control mice, suggesting a reduction in overall COX-2 activity, potentially related to specific COX-2 knockdown in myeloid cells (Figure 2C). Furthermore, the acute inflammatory mediator TNFα, was significantly decreased in exudates obtained from COX-2^−M/−M/apoE^−/− mice (Figure 2D), demonstrating an attenuation of acute inflammation correlated with loss of COX-2 expression and activity in myeloid cells. Unlike TNFα, IL-1β and IL-6 levels were not affected in COX-2^−M/−M/apoE^−/− mice in the air pouch model of acute inflammation (data not shown).

Figure 2.

Lack of COX-2 in myeloid cells results in a decrease in acute inflammation. Air pouches were created in COX-2^−M/−M/apoE^−/− mice and apoE^−/− littermates (n=4). Exudates from air pouches were collected 24 hours after injection of 2% zymosan or sterile saline and total cell counts were determined (A). Adherent and non-adherent cells were isolated as described in Methods and COX-2 and β-actin expression were determined by western blot (B). Air pouch exudates were assayed for PGE₂ (C) and TNFα (D). Data are represented as the average concentration of samples from 4 mice plus one standard deviation. * = p-value < 0.05, # = p-value = 0.07.

To determine whether COX-2 depletion in myeloid cells modulates chronic inflammatory conditions such as atherosclerosis, COX-2^−M/−M/apoE^−/− and apoE^−/− control females were fed either a standard chow diet or the western diet, and atherosclerotic lesion development was determined in all groups (please see Methods for details). Both groups of mice, with and without western diet, developed typical atherosclerotic lesions as determined by Oil Red O staining (Figure 3A–B, D–E). However, there were no significant differences in atherosclerotic lesion development between COX-2^−M/−M/apoE^−/− mice and apoE^−/− control mice on either the standard chow diet or the western diet (Figure 3C, F). Moreover, lipid levels in circulation, including HDL and LDL levels, were unaffected by myeloid COX-2 depletion, regardless of diet (supplemental Table I, please see http://atvb.ahajournals.org).

Figure 3.

Lack of COX-2 in myeloid cells does not influence atherosclerotic lesion development in mice. Female COX-2^−M/−M/apoE^−/− mice and apoE^−/− littermates were fed chow for 14 weeks (n=13–15) or western diet for 18 weeks (starting at 6 weeks of age) (n=12–14). Representative aortic root sections stained with Oil Red O from apoE^−/− (A) and COX-2^−M/−M/apoE^−/− (B) mice on chow. Individual lesion areas in aortic roots from apoE^−/− and COX-2^−M/−M/apoE^−/− mice on chow are shown in (C). Representative aortic root sections stained with Oil Red O from apoE^−/− (D) and COX-2^−M/−M/apoE^−/− (E) mice fed western diet for 18 weeks. Individual lesion areas in aortic roots from apoE^−/− and COX-2^−M/−M/apoE^−/− mice on western diet for 18 weeks are shown in (F). Open circles are averaged lesion areas from individual mice and black bars represent average lesion area of group.

Lesion composition, in addition to lesion size, is another indicator of susceptibility to atherosclerosis.³⁰ In general, an increase in macrophage number and decreased smooth muscle cell proliferation, as well as the presence of thin fibrous cap encasing the lesion suggest a more unstable, rupture-prone plaque.³¹ To determine if lesions in COX-2^−M/−M/apoE^−/− mice contained differences in phenotype, compositions of macrophages and smooth muscle cells were quantified in atherosclerotic lesions by immunohistochemistry using antibodies to CD68 and α-actin, respectively (Figure 4). Figure 4, A–D (CD68) and G–H (α-actin) show representative abundance of macrophages and smooth muscle cells, respectively, from both apoE^−/− and COX-2^−M/−M/apoE^−/− mice. Lesions from COX-2^−M/−M/apoE^−/− showed no significant differences in macrophage or smooth muscle cell expression regardless of diet (Figure 4E–F, I), suggesting that lesions from these mice, in addition to being similar in size to those from apoE^−/− mice, also are similar in composition and phenotype.

Figure 4.

Lack of COX-2 in myeloid cells does not affect lesion phenotype. Aortic root sections from female COX-2^−M/−M/apoE^−/− mice and apoE^−/− littermates fed chow or western diet for 18 weeks were immunostained for CD68 (A–F) or α-actin (G–I) (red staining). Representative aortic root sections stained with CD68 from apoE^−/− on chow (A), COX-2^−M/−M/apoE^−/− on chow (B), apoE^−/− on western diet (C), and COX-2^−M/−M/apoE^−/− on western diet (D) are shown. Individual areas of CD68 staining in aortic roots from apoE^−/− and COX-2^−M/−M/apoE^−/− on chow (n=8–9) (E) and western diet (n=6) (F) were determined as described in Methods. Representative aortic root sections stained with α-actin from apoE^−/− on western diet (G) and COX-2^−M/−M/apoE^−/− on western diet (H) are shown. Individual areas of α-actin staining in aortic roots from apoE^−/− and COX-2^−M/−M/apoE^−/− on western diet (n=6) (I) were determined as described in Methods. Open circles are averaged staining areas from individual mice and black bars represent average area stained per group.

Figure 5.

Lack of COX-2 in myeloid cells does not affect maintenance, differentiation, or mobilization of myeloid and lymphoid populations in vivo. Bone marrow, spleen, lymph nodes, and peripheral blood from female COX-2^−M/−M/apoE^−/− and apoE^−/− littermates (n=4) were harvested and cells were analyzed by flow cytometry as described in Methods. Percentages of myeloid (M) and lymphoid (L) cells were determined in the bone marrow (A), spleen (B), lymph nodes (C), and peripheral blood (D) by CD45/CD11b staining and depicted as dot plots. Percentages of leukocyte subsets, monocytes/macrophages (Mono/Mφ), PMN, and CD4+ lymphocytes (CD4), were determined by F4/80/CD11b/Ly6C/Gr1/CD4 staining in bone marrow (E), spleen (F), lymph nodes (G), and peripheral blood (H). Data are represented as the average percentages from 4 mice plus one standard deviation.

COX-2 and its products have been implicated in the differentiation of myeloid cells, including macrophages;^{32, 33} however, it is not clear whether the lack of COX-2 in myeloid cells alters differentiation programs which are most likely mediated by paracrine and/or transcellular mechanisms. We examined if there were differences in maintenance, differentiation, and/or mobilization of myeloid and lymphoid cells in COX-2^−M/−M/apoE^−/− mice. Flow cytometry was performed on bone marrow, spleen, lymph nodes, and peripheral blood harvested from COX-2^−M/−M/apoE^−/− mice and apoE^−/− controls (Figure 5). We observed negligible changes in myeloid and lymphoid populations in any of the tissues tested (Figure 5A–D). Similarly, there were no notable differences in percentages of monocytes/macrophages, PMN, and CD4+ lymphocytes in any of the tissues assayed (Figure 5E–H). These data suggest that depletion of myeloid COX-2 does not influence maintenance of myeloid or lymphoid cells in the bone marrow or percentages of myeloid and some lymphoid cells in peripheral tissues or in circulation.

To further confirm that COX-2^−M/−M/apoE^−/− mice have specific depletion of COX-2 in myeloid cells, bone marrow was isolated from apoE^−/− and COX-2^−M/−M/apoE^−/− mice and Cre-mediated recombination of the COX-2 gene as well as LPS-induced COX-2 protein expression was determined in monocytes/macrophages, PMN, and CD11b negative (non-myeloid) cells (Figure 6). The floxed COX-2 cassette was specifically deleted in monocytes/macrophages and PMN, but not in non-myeloid cells isolated from the bone marrow of COX-2^−M/−M/apoE^−/− mice (Figure 6A). Furthermore, LPS-induced COX-2 expression was noticeably absent in monocytes/macrophages and PMN, but was retained in non-myeloid cells isolated from the bone marrow of COX-2^−M/−M/apoE^−/− mice, confirming the specificity of the COX-2^−M/−M/apoE^−/− mouse model (Figure 6B).

Figure 6.

Myeloid-specific knockout of COX-2 in bone marrow cells. Bone marrow cells from a COX-2^−M/−M/apoE^−/− mouse and apoE^−/− littermate were either untreated (NT) or treated with LPS (2 µg/mL) for 4 hours, incubated with CD11b/Ly6c/Gr1 antibodies, sorted by flow cytometer, and COX-2 expression was determined in monocytes/macrophages (Mono/Mφ), PMN, and CD11b negative (CD11b_neg) cells by genomic PCR (A) and western blot (B) as described in Methods.

To determine patterns of COX-2 expression in the atherosclerotic lesion, immunohistochemistry was performed on lesions from COX-2^−M/−M/apoE^−/− and apoE^−/− controls on western diet (supplemental Figure I, please see http://atvb.ahajournals.org). In apoE^−/− lesions, there is qualitative co-localization between macrophages (CD68) and COX-2 staining, confirming that macrophages at the site of lesions express COX-2. Macrophages and COX-2 expression do not seem to co-localize in COX-2^−M/−M/apoE^−/− lesions. To determine if serum levels of COX products are affected by myeloid depletion of COX-2, levels of PGE₂, 6-keto-PGF_1α (stable metabolite of prostacyclin), and TXB₂ (stable metabolite of TXA₂) were assayed (supplemental Figure I, please see http://atvb.ahajournals.org) in serum samples obtained from COX-2^−M/−M/apoE^−/− and apoE^−/− controls on western diet. There were no differences in serum levels of prostanoids between the two groups of mice (supplemental Figure I, please see http://atvb.ahajournals.org).

Discussion

Atherosclerosis, a disease of the large arteries that is the primary cause of CHD and stroke, is a multifactorial, chronic inflammatory disease in which low plasma levels of HDL and high plasma levels of LDL, are a strong predictor of the condition.¹ COX enzymes and their products, prostanoids, (prostacyclin, prostaglandins, and thromboxanes) play a key role in both acute and chronic inflammation, including atherosclerosis.^{8, 34} The inducible form of COX, COX-2, is involved in stimulus-induced prostanoid synthesis and is the target for the class of non-steroidal anti-inflammatory drugs (NSAID) termed coxibs.³⁵ Defining the cardiovascular effects of coxibs has taken an increasing clinical relevance due to an increased risk of myocardial infarction and stroke^12–14, which led to the voluntary withdrawal of two classes of coxibs from the market.³⁶ The issue remains intensely controversial, mainly due to lack of consensus in the literature about the role of COX-2 in the development of atherosclerosis.

Macrophages are the hallmark of inflammatory diseases, including atherosclerosis.⁵ Based on their ability to robustly express COX-2, we reasoned that macrophages present an attractive mechanism to study the role of COX-2 in atherosclerosis. In this study, a mouse model in which COX-2 was deleted in only myeloid cells was developed and the role of macrophages and other myeloid cells in acute inflammation and atherosclerosis was investigated. As expected, macrophage COX-2 contributes to PGE₂ formation during acute inflammation in vitro (Figure 1). We used the air pouch model of inflammation to investigate the role of myeloid COX-2 in acute inflammation in vivo. Our results show that, i) COX-2 is highly expressed in myeloid cells during inflammatory events in vivo, ii) COX-2 in myeloid cells regulates TNFα production at early time points of acute inflammation in vivo, and iii) the attenuation of inflammatory responses in COX-2^−M/−M /apoE^−/− mice correlates with decreased PGE₂ levels (Figure 2). Overall, the results from the air pouch model of inflammation demonstrate that COX-2^−M/−M/apoE^−/− mice responded very similarly to previous reports that utilized total COX-2^−/− mice³⁷ and COX-2 inhibitors in acute inflammatory models.^{29, 38, 39} Although significant differences in inflammatory responses were noted in COX-2^−M/−M/apoE^−/− mice in the air pouch model of acute inflammation when compared to apoE^−/− mice, our results demonstrate that myeloid COX-2 does not play a role in atherosclerotic lesion progression or composition in COX-2^−M/−M /apoE^−/− mice (Figures 3 and 4).

The role of macrophage COX-2 in atherosclerosis has recently been investigated in a fetal liver cell transplant model system.^{15, 16} In these studies, the authors concluded that macrophage COX-2 promotes early atherosclerotic lesion formation in apoE^−/− and LDLR^−/− mice transplanted with fetal liver cells from COX-2^−/− mice. However, it should be noted that the fetal liver cell transplants that were employed in their studies populate all hematopoietic cells, including non-myeloid cells. In our model system, these non-myeloid hematopoietic cells robustly express COX-2 during inflammation (Figures 2 and 6). Therefore, it is possible that the anti-atherogenic effects observed by transplant of COX-2^−/− fetal liver cells are a result of depleted COX-2 activity in non-myeloid leukocytes, including lymphocytes, but not macrophages as suggested.^{15, 16} Our results clearly demonstrate that COX-2 in myeloid cells, including macrophages, is not involved in the progression of atherosclerosis in mice.

There was also no change in lesion phenotype observed in lesions from COX-2^−M/−M/apoE^−/− mice (Figure 4). Lesion phenotype, as it relates to plaque stability, is an important index of risk of atherosclerosis.³⁰ Increased macrophage abundance as well as decreased smooth muscle cell proliferation are indicative of an unstable, rupture-prone plaque.³¹ However, in the current study, COX-2 depletion in myeloid cells did not affect macrophage infiltration, smooth muscle cell proliferation, or characteristics of the fibrotic cap encasing the lesion. Although macrophage COX-2 has been shown to regulate the activity of matrix metalloproteinases, which can contribute to plaque rupture⁴⁰, the current studies do not demonstrate regulation of plaque phenotype by myeloid COX-2.

Lipid levels in circulation were also unchanged (supplemental Table I, please see http://atvb.ahajournals.org), further supporting the lack of impact of myeloid COX-2 on atherosclerosis development. We previously showed absence of COX-2 alters lipoprotein metabolism in mice; however, those results were obtained from total COX-2^−/− mice.⁴¹ Our current studies suggest that the hyperlipidemia observed in the total COX-2^−/− mice is due to cell types from a non-myeloid lineage. It follows that other cell types expressing COX-2 could play a role in lipid regulation in circulation and will be the focus of further studies.

Myeloid cells play a large role in inflammation, but we observed no effect of myeloid COX-2 depletion on important processes in atherosclerotic inflammation, including macrophage recruitment to the lesion site. Since it has been hypothesized that COX-2 is involved in the differentiation of myeloid cells^{32, 33}, we hypothesized that the lack of significant differences in lesion macrophage recruitment may be explained by relative differences in differentiation and/or maintenance of leukocyte populations in COX-2^−M/−M/apoE^−/− mice. However, no differences were found in myeloid and lymphoid cells in COX-2^−M/−M/apoE^−/− mice when compared to apoE^−/− mice (Figure 5). Moreover, a closer inspection of leukocyte populations, including monocytes/macrophages, PMN, and CD4+ T cells also demonstrated a lack of effect of myeloid COX-2 depletion on maintenance of leukocytes in bone marrow and peripheral lymphoid tissues as well as mobilization of leukocytes into circulation (Figure 5).

COX-2 knockdown in the entire myeloid compartment has no effect on lesion formation. An explanation for this unexpected finding could be the prominence of COX-2 staining at the lesion itself, even in COX-2^−M/−M/apoE^−/− animals (supplemental figure I, please see http://atvb.ahajournals.org). There is functional overlap in prostanoid production between various cell types in many biological processes.⁸ In addition, although prostanoids are relatively unstable, they can function in a paracrine fashion in the vasculature and circulation.⁴² Moreover, intermediate COX metabolites including PGH₂ can be processed transcellularly by various downstream prostanoid synthases, resulting in local prostanoid production by distal cells.^{43, 44} Other modes of transcellular prostanoid formation include the release of secretory phospholipase A₂ (sPLA₂) by distal cells that could initiate arachidonic acid metabolism in local cell populations.⁴⁵

In the current studies, COX-2 is expressed robustly not only in macrophages in the lesion, but in endothelial cells, smooth muscle cells, and lymphocytes as well (data not shown). Therefore, in an atherosclerotic lesion, where many cell types exist in close proximity, prostanoids produced by COX-2 in various cell types could compensate for COX-2 depletion in myeloid cells or even act in a paracrine fashion on myeloid cells themselves, maintaining prostanoid-dependent myeloid cell functions. Another possibility is that myeloid cells lacking COX-2 could still process exogenous PGH₂ in a transcellular process, masking any effects of COX-2 depletion. In addition, COX-2 deficient myeloid cells could still produce proinflammatory isoprostanes, generated from non-enzymatic, free-radical catalyzed peroxidation of arachidonic acid⁴⁶, thereby affecting atherosclerosis progression independently of COX activity.

The hypothesis that compensatory mechanisms may help explain the lack effect of myeloid depletion of COX-2 on atherosclerosis is further supported by the finding that serum levels of prostanoids, PGE₂, 6-keto-PGF_1α, and TXB₂ are unchanged in COX-2^−M/−M/apoE^−/− mice (supplemental figure I, please see http://atvb.ahajournals.org). Therefore, the depletion of COX-2 in myeloid cells and the postulated resulting decrease in PGE₂ production by myeloid cells is unable to influence the progression of chronic inflammation in atherosclerosis potentially because of sufficient prostanoid production by other cell types and/or unaltered prostanoid production by myeloid cells via transcellular mechanisms.

The human clinical data suggest that COX-2 activity is anti-atherogenic. This study proves that when modeled in mice, myeloid COX-2, including macrophage COX-2, is not responsible for this anti-atherogenic activity. The use of other conditional COX-2^−/− mice, including those made to deplete COX-2 from endothelial cells or smooth muscle cells, could be fruitful in terms of elucidating an anti-atherogenic role for COX-2 that would help propose a mechanism to explain the clinical data.