Differential COX-2 induction by viral and bacterial PAMPs: Consequences for cytokine and interferon responses and implications for anti-viral COX-2 directed therapies☆

Highlights • We report interactions of Toll-like receptors (TLRs) with COX enzymes in vivo.• COX-2 was broadly induced by LPS (TLR4) but more locally by poly(I:C) (TLR3).• COX-1/2 deletion modified the response to TLR activation in a TLR-specific manner.• COX-2 deletion enhanced interferon responses to viral-type TLR3/7/9 ligands.• COX-2 inhibition could provide a novel anti-viral therapeutic strategy.


Introduction
Cyclooxygenase (COX) enzymes catalyze the two-step conversion of arachidonic acid to the unstable prostaglandin (PG) intermediate PGH 2 , which is then further converted to a range of prostanoid mediators that include PGE 2 , prostacyclin (PGI 2 ) and thromboxane (TXA 2 ). Two COX isoforms, which catalyze identical reactions, exist. COX-1 is constitutively expressed in many tissues, and generally plays a role in homeostatic function [1,2]. In contrast, COX-2 is generally not expressed in most healthy tissues but is rapidly induced in response to mitogens [3] and cytokines [4] and is often present in elevated levels at sites of inflammation. COX-2 produces prostanoids that contribute to vasodilation, increased vascular permeability, leukocyte chemotaxis, fever and that potentiate nociception. Indeed, COX-2 is the target of both traditional non-steroidal anti-inflammatory drugs (NSAIDs) such as aspirin and ibuprofen, which inhibit both COX-1 and COX-2, and of newer COX-2 selective inhibitors such as celecoxib. These agents are widely used for the symptomatic control of pain and inflammation, particularly in patients with arthritis, reflecting COX-2 induction in the inflamed arthritic joint.
In addition to its expression in sterile inflammation, it is clear that COX-2 is induced by pathogens. For example, gram-negative bacteria, or lipopolysaccharide (LPS), which activates the prototypical bacterial pattern recognition receptor Toll-like receptor (TLR)4, can induce COX-2 expression both in isolated cells [5,6] and in vivo [5,7]. Viruses or viral mimetics such as poly(I:C) can also induce COX-2 in cultured cells [8][9][10]. However, no systematic comparison of the ability of bacterial and viral pathogen-associated molecular patterns (PAMPs) to induce COX-2 in different tissues has been described. Moreover, whilst prostaglandins are recognized as important modulators of the immune system [11], there is no study that compares the roles of COX-1 versus COX-2 in inflammatory responses induced by viral and bacterial PAMPs. Here, we use a ''knock-in'' Cox2 luciferase reporter mouse [7], in which luciferase activity expressed from the endogenous Cox2 promoter reflects Cox2 expression, to both visualize and quantify the tissue distribution of LPS-and poly(I:C)-induced Cox2 gene expression. We then explore the role of COX-1-and COX-2-derived prostanoids in modulating the inflammatory response to specific PAMPs, using mice deficient in either COX isoform.

Animals
Cox1 À/À [1], Cox2 À/À [12] and Cox2 fLuc/+ mice [7] were backcrossed onto a C57Bl/6J background and identified by genomic PCR. Experiments were performed on 10-12 week old male and female mice. Animal procedures were conducted in accordance with Animals (Scientific Procedures) Act 1986 and after local ethical review by the Imperial College Ethical Review Panel or the UCLA Animal Research Committee, as appropriate for individual experiments.

Bioluminescent imaging
Cox2 fLuc/+ mice were injected with poly(I:C) (8 mg/kg; i.p.; Sigma, UK), LPS (from Escherichia coli serotype 055:B5; 0.1 or 10 mg/kg; i.p.; Sigma, UK) or vehicle (saline) under brief isoflurane anesthesia. 4 h later, D-luciferin (125 mg/kg; i.p.; Xenogen, USA) was administered. After a further 15 min, the dorsal skin was shaved and bioluminescent emission from this area was recorded over 3 min, using an IVIS imaging system (Xenogen, USA). Animals were then euthanized by isoflurane overdose and tissues were rapidly dissected and arranged on cell culture dishes, following which luciferase activity was imaged ex vivo. In both cases, collected photon number and images were analyzed using Living Image software (Perkin Elmer, USA) and quantified as the peak photon release/pixel detected from each tissue.

Luciferase activity assays
After bioluminescent imaging, tissues were snap frozen for biochemical measurement of luciferase activity. Tissues were dissociated using a Precellys24 bead homogenizer and the supernatants loaded into white 96 well microtitre plates. The luminescence was then read after injection of Luciferase Assay Reagent (Promega, UK). Protein concentration of homogenates was determined using the bicinchoninic acid method (Perbio, UK) and luciferase activity was normalized for protein concentration.

Statistics and data analysis
Data were analyzed using Prism 5.01 (GraphPad software, USA). Statistical significance was determined by two-way ANOVA with Dunnett's post-hoc test unless otherwise stated, and data sets considered different if p < 0.05. Each n value represents a data point from a separate animal.

Cox2 gene induction by prototypical viral and bacterial PAMPs
Cox2 gene expression is increased globally in Cox2 fLuc/+ animals treated with LPS, a TLR4 ligand [7]. In the current study, we extended our earlier observations and used two doses of LPS to observe a graded, dose-dependent increase in gene expression (Fig. 1A). When surface luminescence values for intact tissues were measured, the following tissues displayed a robust response (P2-fold increase versus vehicle-treated mice) to LPS (0.1 mg/kg): aorta, heart, liver, lung and spleen (Fig. 1A). When a larger LPS dose (10 mg/kg) was administered, all studied tissues measured displayed P2-fold increase in Cox2 gene expression (versus vehicle) with the following rank order spleen ) heart P liver > lung P aorta > skin % kidney % stomach % thymus % gut % brain. In contrast to results obtained with LPS, the viral PAMP/TLR3 ligand, poly(I:C), at a dose (8 mg/kg), close to the maximum tolerated dose [13], produced only a modest induction of Cox2 gene expression; significant induction was only apparent in the spleen (Fig. 1B).
Surface luminescence values do not always reflect the tissue activity in this model, since tissue absorbance of emitted bioluminescence and penetration of substrate in vivo influence measured luminescence. For this reason, we also measured luciferase activity, driven from the Cox2 promoter, in tissue homogenates. These data (Supplementary Fig. 1) confirm those obtained by surface luminescence measurements and also reveal the stomach as a site of modest poly(I:C)-induced Cox2 induction.

Modulatory role of COX-2 on physical features of sepsis induced by LPS in vivo
Wild-type mice treated with LPS displayed observable 'physical condition' responses, including hunching, inactivity and piloerection ( Fig. 2A). The physical condition response to LPS was accompanied by a decrease in body core temperature, reflecting loss of thermoregulation (Fig. 2B), and reduced circulating lymphocyte and platelet counts (versus vehicle treatment), suggesting tissue lymphocyte sequestration and intravascular coagulopathy, respectively ( Supplementary Fig. 2). These LPS-induced behavioral and body temperature changes were COX-2-dependent, but not COX-1 dependent, since they were strongly suppressed in Cox2 À/À , but not Cox1 À/À mice ( Fig. 2A and B).
In contrast to the results for LPS, other PAMPs tested did not affect the gross physical condition or body core temperature of the mice ( Fig. 2A and B). Nonetheless, like LPS, Pam3CSK4, poly(I:C) and R-848 produced lymphocytopenia ( Supplementary Fig. 2), whilst MALP-2, poly(I:C), R-848 and CpG ODN produced an increase in circulating neutrophils, probably reflecting mobilization from the bone marrow ( Supplementary Fig. 2). Cox1 or Cox2 deletion did not affect any of these features.

PAMPs produce specific patterns of cytokine and interferon response in vivo
Each tested PAMP produced a specific pattern of plasma cytokine response, in agreement with reports from previous in vivo studies [14] and reflecting the distinct tissue distribution and signaling pathways of each TLR [15]. The pro-inflammatory cytokines IL-1b (Fig. 2C), TNFa and KC and the anti-inflammatory cytokine IL-10 ( Supplementary Fig. 3) were increased in plasma of mice treated with LPS, but not in mice treated with the other PAMPs. IL-5 was also increased by LPS, as well as Pam3CSK4, whilst plasma IL-4 was elevated in mice treatedwith Pam3CSK4, CpG ODN and R-848 ( Supplementary Fig. 3). Type I IFNs are intrinsically associated with viral infection [15]. Therefore, as expected, plasma IFNa levels were increased in mice treated with the viral PAMPs poly(I:C) and R-848. IFNa levels also tended to be increased in mice treated with the bacterial PAMPs LPS and CpG ODN (Fig. 3A). In wild-type mice, LPS and R-848 produced a significant increase in plasma IFNc levels (Fig. 3B), but only MALP-2 produced a strong increase in the type III interferon, IFNk (Fig. 3C).

Complex role of COX-2 in cytokine and interferon responses to viral and bacterial PAMPs
To understand how prostanoids produced by COX-1 and COX-2 modify the cytokine responses to specific viral and bacterial PAMPs, we performed studies using mice deficient in either isoform. In line with what was seen with physical condition and body temperature, in mice-treated with LPS, the IL-1b (Fig. 2C) IL-5 ( Supplementary Fig. 3) and IFNc responses were markedly Wild-type, Cox1 À/À and Cox2 À/À mice were treated with a range of bacterial and viral PAMPs. After 4 h ''animal condition'' was scored (A), and core body temperature change (B) and plasma IL-1b levels measured (C). Only LPS produced an apparent of loss of condition in mice; this phenotype was accompanied by a hypothermic response and increased plasma IL-1b levels. Cox2 gene deletion protected mice against each of these responses. Data are expressed as means ± s.e.m. from n = 6 to 12 individual animals per treatment. ⁄ p < 0.05 vs. wild-type; # p < 0.05 vs. vehicle-treated wild-type.
Cytokine responses to viral PAMPs exhibited a distinctly different pattern of response to Cox1 and Cox2 deletion. The most pronounced interactions was an increase of upto 6-fold in the IFNa, IFNc and IFNk (Fig. 4) and IFNb (Supplementary Fig. 4) responses to poly(I:C) in Cox2 À/À mice. This suggests that COX-2-derived prostanoids act to limit TLR3-mediated IFN release, and contrasts sharply with the pro-inflammatory role of COX-2 in LPS cytokine responses. In agreement, levels of the IFN-associated cytokine and CpG ODN, were better able to stimulate interferon production than their bacterial counterparts. In wild-type mice, poly(I:C) increased IFNa levels and tended to increase IFNc levels, however, in each case the response was markedly increased in Cox2 À/À mice. A similar pattern was true for poly(I:C)-induced IFNk. In contrast, the IFNc response to LPS was suppressed by Cox2 deletion, whilst Cox1 gene deletion limited the IFNa response to R-848 and CpG ODN. Data are expressed as means ± s.e.m. from n = 6 to 12 individual animals per treatment. ⁄ p < 0.05 vs. wild-type; # p < 0.05 vs. vehicle-treated wild-type.
IL-12, and the IFN response cytokine IP-10 were also increased by Cox2 deletion in poly(I:C)-treated mice, as were basal levels of IL-12 in vehicle-treated animals ( Supplementary Fig. 4). Whilst there was a tendency for IFNa to be increased in poly(I:C) treated Cox1 À/ À mice, this effect was less robust than that seen in Cox2 À/À mice and was in opposition to reduced R-848 and CpG ODN IFNa responses in Cox1 À/À mice (Fig. 3). This further illustrates the specificity of the interactions between COX-1/2 activity and PAMP/TLR pathways. These effects of Cox1/2 gene deletion on IFNs and related cytokines were associated with a systemic effect, since no change in IP-10 was noted in whole blood stimulated with PAMPs in culture. This contrasts to the clear increase in KC seen in whole blood stimulated with LPS ( Supplementary Fig. 5).

Discussion
COX-2 is an inducible enzyme that regulates the production of prostaglandins in inflammation and infection. It is increasingly recognized that COX-2 has a complex role in immune responses, the extent of which we do not yet fully understand. In the current study we used a mouse model in which expression from the Cox2 gene can be imaged directly, using a luciferase reporter ''knocked in'' to the coding region of the endogenous Cox2 gene, to compare expression after treatment with bacterial and viral PAMPs. We present data showing that bacterial LPS caused a pan-COX-2 induction across all studied tissues. In contrast, the viral PAMP, poly(I:C), induced a much more tissue-specific and limited induction from the Cox2 gene, with increases seen only in the spleen and stomach. These results likely reflect the limited distribution of TLR3 to specific immune cells, compared to broadly expressed TLR4.
In the current study, we show, for the first time, how deletion of Cox1 or Cox2 affects responses to a broad range of PAMPs. These data, summarized qualitatively in Fig. 4, clearly highlight distinct roles of COX-1 and COX-2-derived prostanoids in the modulation of specific PAMP/cytokine responses and suggest a potentially important role for COX-2 in anti-viral interferon responses.
In mouse models, the TLR4 ligand, LPS, induces a systemic inflammatory response that includes pronounced changes in the physical condition of the animal. Typical responses include piloerection, inactivity and loss of thermoregulation. In our study these effects were prevented by Cox2 gene deletion, consistent with previous reports [16], but not by Cox1 deletion. Importantly, these observations reflect what we know about the consequences of COX-2 inhibition in man; NSAIDs, including COX-2 selective inhibitors, are effective anti-pyretic drugs [17].
Whilst the other tested PAMPs did not induce notable changes in physical condition or thermal deregulation, they did produce cytokine changes. These effects were not ubiquitous and each PAMP elicited a relatively specific cytokine response profile (Fig. 4). Cox1 and Cox2 gene deletion studies suggested distinct interactions between COX-1, COX-2 and specific TLR responses. These differences likely reflect both the targeting of unique cell populations by each PAMP and the specific prostanoid pathways associated with these targets. Consistent with the anti-inflammatory effects of COX-2 inhibitors in man and in previous animal studies [16] we noted that Cox2 gene deletion attenuated IL-1b, IL-5 and IFNc responses to LPS. In contrast, the response to other tested bacterial PAMPs were, for the most part, not consistently altered by deletion of either the Cox1 or Cox2 gene.
TLR3, the receptor for poly(I:C) is highly expressed on dendritic cells, which secrete IFNa and IL-12. These ligands, in turn, stimulate IFNc production in natural killer cells [14]. Interestingly, Cox2 gene deletion enhanced poly(I:C)-induced release of type I (IFNa, IFNb), type II (IFNc) interferons and IL-12, consistent with an augmented dendritic cell response, as well as release of type III interferon (IFNk) and the IFN response protein, IP-10. Indeed, elevated basal IL-12 levels in Cox2 À/À mice suggest dendritic cells may be primed for activation in these animals, consistent with previous reports that prostanoids can limit dendritic cell function and survival [18]. TLR7 and TLR9, the receptors for the viral PAMPs R-848 and CpG ODN, respectively, have a broader expression in the immune system than TLR3 and couple to the MyD88 rather than TRIF adapter protein. Unlike TLR3-mediated IFNa responses, TLR7/TLR9 stimulated-release of IFNa was not enhanced by Cox2 gene deletion and was suppressed by Cox1 gene deletion, illustrating a complex interaction between viral PAMPs and the two COX enzymes.
In addition to the predicted anti-inflammatory role of COX-2 deletion, our observations showing enhanced anti-viral interferon responses in Cox2 À/À mice to poly(I:C), suggesting that COX-2 inhibitors could have disease-modifying activity in viral infections. Interactions between COX-2 and viral infection has been suggested in a limited number of studies; COX-2 can be induced by whole virus in isolated cells [8][9][10] and COX-2 protein is increased in biopsy specimens of viral target tissues from patients with active viral infection [19,20]. Moreover, data from isolated cells [8,10] and animal models suggest that COX-derived products may play an important role in the host response. Cox2 deletion reduces mortality in mice infected with influenza A [21]. Moreover, in the same model, Cox1 deletion is associated with worsening of infection, consistent with our data demonstrating that COX-1 can limit TLR7/9 interferon responses. Protective effects of COX-2 inhibition have also been described in rodent models of respiratory syncytial virus [22] and vesicular stomatosis virus (VSV) [23]. Whilst there is little consensus as to the mechanism(s) responsible for this effect, limited previous data point to a role for IFN. For example, COX-2 inhibition in VSV-infected mice increases plasma IFNc and IL-12 levels [23], and treatment of hepatitis B or C patients with the COX-1/COX-2 inhibitor indomethacin increases serum levels of the IFN response product 2 0 5 0 -oligoadenylate synthetase-1 [24]. Our data may, therefore, provide a mechanistic link by which COX-2 regulates viral infection as a result of modulation of innate immune recognition and subsequent interferon response.
Taken together, this study is the first to provide a systematic analysis of the reciprocal interactions between COX enzyme induction and inflammatory responses to a range of bacterial-and virallike stimuli. The ability of Cox2 deletion to suppress physical and cytokine responses to LPS has been previously reported, and contrasts strongly with our new data showing that Cox2 deletion enhances IFN responses to viral PAMPs, particularly poly(I:C). These data suggest a role for COX-2 in limiting the anti-viral cytokine/interferon response to infection, and may provide a plausible explanation for the previously published data showing that Cox2 deletion/COX-2 inhibition is beneficial in animal models of viral infection. If a similar mechanism is present in man, COX-2 inhibitors might be a potential anti-viral therapy, able to boost the endogenous anti-viral response when given soon after infection.

Role of funding sources
This research was supported by a Wellcome Trust program grant (0852551Z108/Z; J.A.M.) and NIH-NCI P50 award CA086306 (H.R.H.). A.K.Z. is the recipient of an American Society of Hematology Scholar Award. No funding source was involved in the study design; data collection, analysis or interpretation; writing of the report or the decision to submit for publication.