The present study examined the effects of treatment with a COX-1 or COX-2 selective inhibitor on the host response to influenza A viral infection in mice. Treatment with the COX-1 selective inhibitor SC-560 was associated with greater mortality and greater infection-induced changes in body temperature and body weight compared to treatment with the COX-2 selective inhibitor celecoxib or no inhibitor (control). Numbers of inflammatory cells in the BAL fluid were increased in the celecoxib treated group compared to the SC-560 and control groups. Inhibition of either COX enzyme led to decreases in BAL fluid levels of TNF-α and G-CSF. In contrast only inhibition of COX-1 led to a decrease in BAL fluid levels of KC and only inhibition of COX-2 led to a decrease in levels of IL-6. Viral titres were similar between the treatment groups.
We previously demonstrated a biphasic temperature response to influenza viral infection in mice — an initial hyperthermic response followed by a progressive hypothermic response 
. Mice and other small rodents tend to develop hypothermia rather than fever in response to infectious stimuli 
. In the present study, inhibition of either COX-1 or COX-2 blocked the initial hyperthermic response. Subsequently, inhibition of COX-1 led to profound hypothermia whereas inhibition of COX-2 led to a normal hypothermic response. These results contrast with observations in COX-deficient mice in that COX-1 deficiency led to a greater hyperthermic response and COX-2 deficiency abrogated the development of hypothermia 
. The present findings are similar to the prior study with knockout mice in that COX-2 deficiency abolished the hyperthemic response and COX-1 deficiency worsened the degree of hypothermia. Considerable evidence supports the role of COX enzymes in thermoregulation. COX-2 has predominantly been implicated in modulating body temperature changes in response to infection. However, some studies have also implicated COX-1. Studies in rats have shown that LPS-induced hypothermia is blocked by the COX-1 inhibitors SC-560 
and valeryl salicylate 
, but enhanced by the COX-2 inhibitor SC-236 
. In contrast, an earlier study, also in rats, found that the COX-2 inhibitor SC-236 blocked LPS-induced hyperthermia but that the COX-1 inhibitor SC-560 resulted in profound hypothermia in response to LPS 
. The results of the latter study are consistent with the observations of the current study.
A critical challenge for the immune system is balancing the immune response to control infection while minimizing damage to the host. It is thought that much of the morbidity and mortality associated with influenza infection can be attributed to an over exuberant immune response leading to excessive production of cytokines and excessive inflammation at the site of infection 
. Indeed, when we examined the response to influenza A virus in the COX deficient mice, clinical signs of infection correlated with the inflammatory response 
. Inflammation was reduced in COX-2 deficient mice and this was consistent with less morbidity. In contrast, inflammation in COX-1 deficient mice was enhanced and this was associated with a poorer clinical outcome. Surprisingly, in the present study clinical signs of infection in the COX inhibitor treated mice did not correlate with inflammation. The celecoxib treated group had mildly elevated levels of inflammatory cells in the BAL fluid and several pro-inflammatory cytokines remained unchanged whereas others were reduced compared to the control group. There was little difference between the celecoxib treated and control groups with respect to clinical signs. In contrast, SC-560 treatment had no significant effect of BAL fluid inflammatory cell numbers, although there was a trend for decreased neutrophils which may be related to the depressed levels of the neutrophil chemokine KC on day 4 of infection. Otherwise, SC-560 treatment produced a very similar BAL fluid cytokine profile to celecoxib treatment. Yet, the SC-560 group exhibited more severe clinical signs of illness and 100% mortality. These results suggest that inflammation can be dissociated from clinical outcome in the COX inhibitor treated animals.
In addition to immune system activation, inflammatory or infectious stimuli induce a highly coordinated central nervous system response which modulates body temperature changes. The COX-1 inhibitor induced profound hypothermia in influenza infected mice suggesting that COX-1 is required for the suppression of hypothermia following infection with influenza in mice. The degree of hypothermia in mice can predict mortality: when body temperature drops below a certain point in various infection models, death is almost inevitable 
. It is difficult to say definitively whether the hypothermia was the cause of excessive mortality in the SC-560 or a consequence/marker of some other process.
The mechanisms regulating hypothermia are not fully understood but cytokines such as TNF-α and interleukins have been shown to induce or modulate the hypothermic response 
. Studies have shown that the degree of hypothermia correlates with levels of certain pro-inflammatory cytokines 
. We did not observe any elevation in cytokine levels in the SC-560 treated group relative to control that might explain the excessive hypothermia; cytokines examined were either unchanged or blunted in relation to control.
The differential effects of SC-560 on the temperature response may in fact be mediated further upstream in the process (i.e. centrally in the brain). Several studies support a role of COX products in the central nervous system response to infectious stimuli. For example, microinjection of a COX inhibitor into the preoptic area of rat brain, a region believed to be responsible for thermoregulation, decreased the fever response to LPS suggesting that prostaglandin biosynthesis in that region of the brain is necessary to modulate the thermoregulatory response to infection 
produces fever when injected intracerebroventricularly 
and mice lacking the EP3 receptor also lacked an appropriate febrile response to PGE2 
. Oka and colleagues found that the EP3 receptor is necessary to produce fever and also necessary to prevent profound hypothermia in response to LPS 
. The effects of COX inhibitors on the brain circuitry that is activated as part of the thermoregulatory response to infection in mice have not been thoroughly investigated and would be an interesting area for future study.
Treatment of influenza A virus infected mice with selective COX inhibitors did not recapitulate the phenotypes observed using the same model in COX knockout mice. This is not a unique observation: we have previously observed discordance between studies with COX knockout mice and COX inhibitor treated mice in an allergic airway disease model 
. There are several potential explanations for the discordance. First, in inhibitor studies, the COX enzymes are inhibited at the time of study, whereas COX knockout mice are genetically deficient in the enzyme from the point of conception. Second, the COX knockout mice have a total absence of the respective COX activity, a condition likely not attainable, even with higher doses of COX inhibitors 
. Third, the COX enzymes are known to play differential roles in immune development 
and thus immune phenotypes of the COX knockout mice may be a consequence of developmental effects of COX deficiency rather than inhibition of the enzyme. Finally, it is also possible that some of the discordance between the knockout and inhibitor studies could be due to COX-independent effects of the inhibitors 
An interesting observation was that treatment with either COX inhibitor led to depressed levels of G-CSF in BAL fluid on day 4 of infection and a similar trend was observed in serum. G-CSF expression is often induced during infection and is thought to play an important role in the regulation of the systemic and local neutrophil response to the infection 
, a process known as stress or emergency granulopoiesis. However, the depressed levels of G-CSF in the COX inhibitor treated groups appear not to have significantly affected BAL fluid neutrophil levels; in fact, neutrophils were slightly elevated on day 4 in the celecoxib treated group. It is possible that the importance of G-CSF in regulating stress granulopoiesis is pathogen/route dependent: in G-CSF null mice, neutrophilia is normal in response to intravenous Candida albicans
or intraperitoneal Listeria monocytogenes
but blunted in response to intravenous Listeria monocytogenes 
. Nevertheless, the observation that COX inhibitors can depress G-CSF levels during stress granulopoiesis is important and may have significant consequences in other infectious models/states.
The COX enzymes are a major pharmaceutical target. Because of their analgesic effects and their potent anti-inflammatory and anti-pyretic properties, NSAIDs are amongst the most widely prescribed drugs in the western world. The classical NSAIDs inhibit both COX-1 and COX-2 but tend to be more selective towards COX-1 
. While one cannot extrapolate from animal studies directly to human populations, the profound effect of the COX-1 inhibitor SC-560 on the thermoregulatory response to influenza A virus is noteworthy and deserves more detailed study. A key focus of future studies will be to examine how the thermoregulatory response in the brain of influenza A virus infected mice is modulated by SC-560 and other COX inhibitors.
In the present study the COX inhibitors were administered orally and in chow given ad libitum. As we observed, the model was characterized by weight loss and so it is possible that mice were not receiving a consistent dose of drug each day. Nevertheless, the fact that the COX-1 inhibitor had such a profound effect with the greatest weight loss suggests the COX-1 is critical to host response to influenza viral infection and can still have a potent, long lasting effect, even at lower doses.
In summary, we have shown that treatment with a COX-1 inhibitor during influenza A viral infection is detrimental to the host, whereas treatment with a COX-2 selective inhibitor does not significantly modulate disease severity. Furthermore, our studies point to a critical role for COX-1 in controlling the thermoregulatory response to influenza infection in mice.