Influenza virus is a major human pathogen that causes epidemics and pandemics with increased morbidity and, especially in the elderly and those with pre-existing medical conditions, increased mortality. Influenza is characterised by respiratory symptoms and constitutional symptoms. Whilst knowledge of the mechanisms underlying host and tissue specificity has advanced considerably of late we still know relatively little about other aspects of influenza virus virulence. In this review, we will explore what is known about the role of apoptosis in respiratory epithelial cell damage and the role of cytokines in inflammation and constitutional symptoms with particular emphasis on the link between apoptosis, inflammation, fever and cytokine production.
Influenza virus is a major human pathogen that causes epidemics and pandemics of significant morbidity. In the elderly and those with pre-existing medical conditions, it can cause increased mortality. Influenza is characterised by respiratory symptoms (nasal obstruction and discharge, sore throat and cough) and constitutional symptoms (fever, chills, myalgia, anorexia, headache and sleepiness) that vary in severity for different strains. For example, H1N1 outbreaks since 1977 were less severe than those caused by H3N2 viruses [1–3] whilst the 1997 and 2003/2004 H5N1 Asian outbreaks produced much more severe disease [4, 5]. This review will attempt to draw together current knowledge of the mechanisms involved in symptom expression.
Mechanism of production of constitutional symptoms
An intriguing question was how does an infection that is essentially localised to the respiratory tract produce constitutional symptoms and what factors control their severity? Studies in infected ferrets showed that inflammatory cells, which flowed into the upper respiratory tract to combat the infection, released “endogenous pyrogen” (EP), that induced fever when inoculated into ferrets or rabbits [6, 7]. The amount of EP released by these phagocytes correlated with the severity of fever in the ferrets from which the phagocytes were lavaged . Human EP produced all the constitutional symptoms associated with influenza when inoculated into human volunteers .
Nature of endogenous pyrogen
Cytokines and chemokines, produced by mononuclear phagocytes, are known to cause fever and flu-like symptoms when inoculated into humans or animals and thus are EPs.
Cytokines are local hormones that activate cells, particularly of the immune system. Some, such as the tumour necrosis factor (TNF) and interleukin (IL)-1 families, are pro-inflammatory and induce fever (see below) while others, such as IL-2, IL-12 and IL-15, upregulate cells of the innate immune response [10, 11].
Chemokines are a subset of cytokines that act as chemoattractants for cells of the immune system. There are four broad groups of chemokines classified according to cysteine residue arrangement for the first two cysteines in the protein: CC, CXC (α chemokines), XC (γ chemokines) and CX3C (δ chemokines) (for review, see . CXC chemokines may also contain an ELR motif (glutamine, leucine and arginine) and generally act on neutrophils, whilst CC chemokines activate monocytes and, together with endothelial monocytic attractant molecules such as ICAM-1, promote their extravasation. Chemokines promote the stable binding of leukocytes to vascular endothelial cells and direct their migration along a gradient of increasing chemokine concentration towards the site of infection.
Cytokines and fever
The most important cytokines involved in the pathogenesis of fever are IL-1α/β, TNFα/β IL-6 [15, 16] and interferon (IFN) α/γ however several others are known to act as endogenous pyrogens such as IL-8  and macrophage inflammatory protein (MIP)-1α (CCL3) . The classical model for fever induction (Fig. 1) is that these cytokines, produced locally or systemically following interaction of exogenous pyrogen (influenza virus) with phagocytes, reach the central nervous system (CNS). Here, they interact in the organum vasculosum of the lamina terminalis (OVLT) of the hypothalamus to produce prostaglandins (PG), particularly PGE2. This increases the thermostatic set point and triggers complex thermoregulatory mechanisms to increase body temperature . However, dissecting the role of particular cytokines in fever has proved complicated because of their pleiotropy and cross-talk amongst signalling pathways.
Studies in humans experimentally or naturally infected with influenza virus have shown that nasopharyngeal washes contain IL-6, TNF-α, IFN-α, IFN-γ, IL-10, monocyte chemotactic protein (MCP)-1 and MIP-1α/β[21–24]. Similarly, nasal phagocytes from influenza virus-infected ferrets released IL-1, IL-6 and TNF-α and human phagocytes incubated in vitro with influenza viruses produced TNF-α, IL-1β, IL-6, IL-8, MIP-1α and IFN-γ. However, in none of these studies has any one cytokine correlated with disease severity.
It is probable that cross-talk between cytokines is the explanation and the nature of the cytokines involved may differ between influenza viruses or individuals. Inoculation of specific antiserum has implicated IFN-α and IL-1α in the fever response during influenza of mice . Interestingly, a consequence of anti-IFN-α treatment was a reduction in IL-1 levels supporting the view that a cytokine network may be in operation. Studies in animals with gene knock-outs showed that IL-1β-induced fever is mediated through IL-6 . There may be further regulatory control in that there are anti-pyretics such as IL-10, TNF-α, arginine vasopressin and glucocorticoids .
Nature of the influenza virus exogenous pyrogen
There have been few attempts to identify the virion components that stimulate the production of pyrogenic cytokines. As double stranded (ds) RNA is a good inducer of IFN and other cytokines it has been proposed that fever is induced by viral dsRNA made as a by-product of viral replication . Total lung RNA isolated from virus infected, but not sham inoculated mice induced fever when injected into rabbit brain  and synthetic viral dsRNA (108 or 661 bp) derived from the N-terminal sequence of segment 3 of A/Puerto Rico/8/34 influenza virus was pyrogenic when injected intracerebroventrically into rabbits . Such dsRNA is released from infected cells when they die and thus may stimulate cytokine production in uninfected cells . However, virion components are also pyrogenic as virosomes containing no RNA but including viral lipid, haemagglutinin and neuraminidase induced fever, although less so than virions killed by limited heating or UV-inactivation, which destroys infectivity without affecting pyrogenicity [34–37]. Individual components were not pyrogenic explaining why whole virus vaccines can produce influenza-like symptoms while subunit vaccines do not.
Production of respiratory symptoms
Replication in the respiratory tract leads to cell damage and liberation of cytokines and chemokines, which lead to inflammation and respiratory symptoms. The mechanisms involved in cell death and cytokine production are now becoming better understood and appear to be intrinsically linked.
Apoptosis or programmed cell death is a series of defined cellular events that culminates in the efficient removal of the cell and its contents . The original definition of apoptosis states that this form of cell death does not inaugurate an inflammatory response . However, recently this classic definition has been revised as apoptosis in certain situations, such as pathogen invasion, can induce an inflammatory response, which may promote the activation of an immune response .
Apoptosis is characterized by several morphological changes, including cytoskeleton disruption, condensation of the cytoplasm and chromatin, membrane blebbing, loss of mitochondrial function, fragmentation of DNA into 180 base pair oligomers and ultimately the formation of small membrane bound particles known as apoptotic bodies, which in vivo are rapidly cleared by phagocytic cells such as macrophages . Apoptosis can be triggered via one of several pathways, dependent on the initial stimulus. Many of these pathways involve receptor stimulation, activation of the protein kinase/phosphatase cascade [including tyrosine kinases, serine/threonine kinases, mitogen activated kinases (MAP) and protein kinase R (PKR)] and the release of secondary messengers, which act as positive or negative transcription factors for specific genes (see Fig. 2). Another common event in most of these pathways is the activation of a set of cysteine proteases (caspases). Caspases can be separated into two groups, the initiator caspases (caspases-8, -9, -10 and -12) and the effector caspases (caspases-1, -3, -6 and -7). The initiator caspases are recruited to death receptors or other cytoplasmic adaptor molecules and become activated by proteolytic cleavage. These active caspases initiate a caspase cascade resulting in the activation of the effector caspases. The effector caspases cleave a whole range of proteases and nucleases that mediate the morphological transition seen in apoptotic cells . For comprehensive reviews of apoptosis signalling pathways, see [41, 42]
Mechanisms of influenza virus-induced apoptosis
Influenza virus induces apoptosis in a variety of cell types both in vitro and in vivo [43–48]. However, the precise mechanism of virus-induced apoptosis is unclear. One of the major problems elucidating the pathways involved is the large number of cell types being studied, as the apoptotic response can depend on cell type and activation stimulus. Furthermore, many of the pathways activated are not mutually exclusive; the initiation of one pathway can induce multiple signal transduction cascades through feedback loops . However, regardless of the pathways triggered, the final outcome of the apoptotic response appears to be universal (Fig. 2).
PKR is a key regulatory component in many apoptotic pathways  and is induced by IFN and activated by dsRNA. Activated PKR has been detected in a number of different cell types infected with influenza virus [51–54]. Activation of PKR leads to several sequential downstream events including phosphorylation of eukaryotic initiation factor (eIF)-2, activation of Nuclear Factor (NF)-κB and transcriptional induction of numerous pro-apoptotic genes including those encoding Fas, p53 and Bax [52, 55]. Active PKR, by an unknown mechanism, also results in the recruitment of caspase-8 by the cytoplasmic protein Fas associated death domain (FADD). This results in the initiation of a caspase cascade, which includes the activation of caspase-9 [55, 56]. Influenza virus-induced apoptosis in murine 3T3 cells and MDCK cells correlates with upregulated expression of Fas [57, 58]. In addition, the initiator caspase, caspase-8 and effector caspase, caspase-3, are activated during infection of these cells . The anti-apoptotic mitochondrial protein, Bcl-2, has been shown to inhibit virus-induced apoptosis in MDCK cells  indicating mitochondrial dysfunction may occur after activation of caspase-8, thus augmenting the caspase cascade. Together these data suggest that virus-induced apoptosis is, at least in part, mediated via the formation of a FADD/caspase-8 complex by PKR, which initiates a caspase cascade, independently of Fas . However, treating infected cells with antibodies directed towards Fas Ligand (FasL) partially inhibits the induction of apoptosis  indicating Fas may also be involved in viral induced apoptosis, perhaps by a secondary mechanism. The level of FasL expressed on the cell surface is also upregulated in these cells , and may play an additional role in the induction of apoptosis.
Studies using equine influenza virus have shown that influenza virus infection of MDCK cells induces apoptosis via a stress-activated pathway . Transforming growth factor (TGF)-β, upregulated in infected cells, initiates either a c-Jun N-terminal kinase (JNK) or stress activated protein kinase (SAPK) cascade, which modulates the activity of the apoptosis-promoting regulatory factor c-Jun/AP-1 [62, 63]. Viral neuraminidase activates latent TGF-β and influenza virus-induced apoptosis of MDCK cells is partially inhibited by TGF-β-specific antibodies  providing further evidence for the role of TGF-β in virus-induced apoptosis. These latter studies were performed using avian and human influenza viruses indicating that the TGF-β pathway is not dependent on strain. This pathway is not limited to MDCK cells; TGF-β activation has been implicated in the apoptosis of lymphocytes and thus lymphopenia observed during acute infection .
Recent studies have shown that caspase-8 activation is the predominant apoptotic pathway in human bronchiolar cells . There is no evidence for the role of the mitochondria in this pathway, as caspase-9 is not activated. However, when cells are treated with the caspase-8 inhibitor Z-IETD-fmk complete inhibition of apoptosis is not achieved  indicating the involvement of a second pathway during viral induced apoptosis of bronchiolar cells. This caspase-8 independent pathway is mediated via the activation of apoptosis signal-regulating kinase (ASK)-1, which triggers a phosphorylation cascade . Endoplasmic reticulum (ER) stress, due to the over expression of viral glycoproteins, induces this signal transduction pathway through the formation of an IFN response element-1(IRE-1)–TNF receptor-associated factor 2 (TRAF2)–ASK1 complex . The activation of PKR may also potentiate this system, by direct activation of ASK-1 .
Lung tissue injury, following influenza virus infection, has been associated with cellular oxidative stress and reactive oxygen species (ROS) generation . There is also evidence for the induction of nitric oxide synthase (NOS)-2 in human airway epithelium, by dsRNA dependent-PKR . This leads to the formation of toxic reactive nitrogen intermediates. Oxidative stress of infected cells leads to the activation of the transcriptional regulatory proteins, activating protein (AP)-1, C/EBP, and NF-κB , all of which are potential pro-apoptotic and/or pro-inflammatory regulators. However, anti-oxidants had little effect on the level of apoptosis induced in the bronchiolar cell line NCI-H292 . In contrast, we have found that ROS generation is the predominant pathway in human nasal cells (E. Brydon, unpublished observation). ROS and free radicals usually act by causing mitochondria membrane dysfunction, resulting in the release of cytochrome c and consequently caspase-9 activation . However, caspase-9 is not activated in influenza virus-infected nasal cells (E. Brydon, unpublished observation). Studies using the human embryonic kidney cell line HEK293 have shown over expression of M, HA or NP generates ROS, which leads to IκB kinase (IKK) expression and ultimately NF-κB activation . These results indicate nasal cell apoptosis may involve transcriptional activation of cellular oxidative stress genes and a direct attack on DNA integrity by ROS. However, the role of oxygen radicals in activation of NF-κB has been questioned; they may be cell and/or stimulus specific, acting as facilitators rather than causatively . A caspase-8 signalling cascade may also be involved in virus-induced apoptosis of nasal cells, as low levels of caspase-8 were detected in infected cells. However, the overall contribution of caspase-8 to influenza virus-induced apoptosis in this cell type is unclear (E. Brydon, unpublished observation).
Many of the apoptotic pathways described above result in the activation of transcription factors such as AP-1 and NF-κB, which are known to regulate the expression of several cytokines and chemokines. Thus, the activation of these transcription factors provides a direct link between apoptosis, cytokine expression and inflammation.
Apoptosis and virus replication
It has been postulated that the induction of apoptosis is a host defence mechanism, stopping the replication and spread of virus. However, inhibiting influenza virus-induced apoptosis by Bcl-2 expression reduces virus yield, spread and HA glycosylation . Furthermore, caspase-3 activation is essential for efficient virus replication as is the induction of the pro-apoptotic factors TRAIL and FasL by NF-κB [74, 75]. In contrast, it has been suggested that influenza virus may avoid the possible anti-viral properties of apoptosis through rapid multiplication as apoptosis was only observed after virus had replicated . This may be virus strain or cell dependent, as caspase inhibitors did not affect virus replication in human bronchiolar cells (E. Brydon, unpublished observations). In addition, some viruses that replicate well in tissue culture cells induce only low levels of caspase 3 and apoptosis in MDCK cells [77; S. Morris, unpublished observations].
Nature of the influenza virus component involved in induction/inhibition of apoptosis
The downstream pathways of molecular signalling after influenza virus infection are subject to intense research at present (see above). However, the upstream signalling events and what it is specifically that initiates these responses is less clear. It is likely that apoptosis induction is multifactorial and highly regulated, as influenza virus strains have been shown to differ in their ability to induce apoptosis [78, 79]. In addition, the multiple pathways discussed above, many of which appear to be cell type specific, may be initiated by different viral components. Several proteins have been reported to play a role in influenza virus-induced apoptosis. However, many of these proteins have been identified using a single protein expression approach that may not reflect events during infection.
The viral NA was the first influenza virus protein shown to have a role in the induction of apoptosis [64, 78, 80]. It can activate latent TGF- β at the cell surface by facilitating cleavage of TGF-β into its active form. However, NA is not the sole contributor to apoptosis as UV-irradiated virus, which retains 100% NA activity, induced only low levels of apoptosis . Similarly, ammonium chloride, used to prevent virus entry, and amantadine, which inhibits virus uncoating, reduced virus-induced apoptosis . The requirement for endosome acidification in apoptosis induction is interesting in the light of recent observations that ssRNA and influenza virus may interact with Toll-like receptor 7 to initiate IFN responses and this may be blocked by neutralisation of endosome acidification . As described above dsRNA may also induce apoptosis.
The M1 matrix protein has been shown to have some amino acid similarity to protease inhibitors, specifically a motif between residues 18-22 that bore a structural resemblance to the active site of aprotinin. Subsequently a protease-binding domain in the N-terminal region of M1 was defined  suggesting that it might bind caspases. Indeed, M1 binds to caspase-8, and weakly to caspase-7. How it functions in this complex is not yet known but it may be an influenza virus-encoded inhibitor of caspase-8. However, caspase-8 activation does occur in virus-infected MDCK epithelial cells and 3T3 fibroblasts [59, 83] so the ability of M1 to bind and inhibit caspase-8 activation may vary between strains or be regulated by another, as yet unidentified, factor. In contrast, both M1 and M2 expressed individually from plasmids induced apoptosis .
Until recently, it was thought that influenza A viruses encoded only 10 proteins from 8 RNA segments. During experiments characterising the viral proteins recognised by mouse CD8+ T cells, Chen and colleagues  serendipitously discovered a small protein encoded by an alternative +1 reading frame in the PB1 protein. The protein localises to mitochondria where it permeabilises and destabilises the mitochondrial membrane leading to leakage of cytochrome c. Monocytes, but not fibroblast or epithelial cell lines, expressing PB1-F2 underwent apoptosis. The suggestion is that this protein targets leucocytes to undergo apoptosis, possibly resulting in the lymphopoenia often seen in influenza .
The non-structural 1 protein (NS1), like M1, induces apoptosis when expressed from a plasmid in the absence of the other viral proteins . In addition, the NS1 protein shows approximately 50% of structural homology to the pro-apoptotic Fas receptor . However, these results may be aberrant of the situation during viral infection, as the NS1 protein has recently been shown to downregulate influenza virus-induced apoptosis as apoptosis increased in cells infected with an NS1 deletion mutant . This is possibly through its capacity to inhibit type 1 IFN production , which potentiates virus-induced apoptosis , or through competing with Fas due to their structural homology. NS1 also inhibits activation of transcription factors such as NF-κB  and JNK/AP-1 . Recently we have shown, using reverse genetics, that the NS RNA segment 8 derived from a poor apoptosis-inducing virus, converted a virus that induced high levels of apoptosis into a poor apoptosis-inducing virus. Furthermore, RNA segment 8 from the good inducer enhanced apoptosis induced by the poor inducer (S. Morris, unpublished observations). Whether this is due to the NS1 or NS2 protein is presently unknown but illustrates the need to use different strains in such investigations.
The inflammatory response during influenza
During influenza virus infection, it is neutrophils that dominate the early infiltrate, with other inflammatory cells increasing in numbers later on in infection . Neutrophils are attracted to the site of infection and activated by the CXC chemokine IL-8 (CXCL8) (Fig. 3). However, other chemokines such as MIP-1α/β, MCP-1 and RANTES (Regulated on Activation, Normal T cell Expressed and Secreted) may also be important in inflammation induced by influenza virus [92, 93]. They promote the activation and extravasation of macrophages by an increasing concentration gradient towards the site of infection and increasing their interaction with ICAM-1 on the surface of vascular endothelial cells via their receptor LFA-1.
Cytokines and chemokines are produced usually only after a cell has received an induction signal. Which cytokines are produced is determined by the signal of induction, the cell type, its location and developmental age, and the specific gene spectrum induced by transcription and translation factors within the cell. Their expression and release is tightly controlled at almost all levels including transcription, translation and protein processing.
Control of cytokine gene expression
The expression of a cytokine mRNA depends on the transcription factor binding site(s) within its promoter. Thus, most cytokine expression is due to the transduction of signals from the plasma membrane through adaptor molecules and kinases to the activation of transcription factors such as NF-κB and NF-IL6, IFN regulatory factors (IRFs), signal transducers and activators of transcription (STATs), AP-1 and others. Many cytokines and chemokines contain NF-κB binding sites in their promoters as well as others; IL-6 and IL-8 expression requires the binding of both NF-κB and NF-IL6 to the response elements in their promoters . This implies that cytokine expression is a transient event initiated in response to a stimulus. However, IL-6 expression is constitutive in tumour cells  as also IL-8 may be . For a recent review of cytokine transcription factor induction by influenza virus, see .
Triggering the cytokine response in influenza
As has been intimated above cytokine and chemokine production during influenza is a complex process likely to involve many pathways, networks of induction and suppression, and cell types. The most widely postulated trigger for cytokine induction on influenza virus infection is dsRNA intermediates produced during replication. The mechanism by which dsRNA triggers production of cytokines is still not well understood. Recent work implicates Toll-like receptor (TLR) 3 as a sensor for dsRNA in the initiation of an innate immune response . Influenza virus produces dsRNA intermediates during replication. However, contrary to this, it has been reported that viral pathogenesis and immune responses were similar in TRL3−/− mice compared to TRL3+/+ mice on infection with lymphocytic choriomeningitis virus, vesicular stomatitis virus (VSV), murine cytomegalovirus or reovirus . In addition, plasmacytoid dendritic cells (pDCs), which do not express TRL3, isolated from the bone marrow of mice and infected in vitro with influenza A virus were found to secrete wild type levels of IFNα and Il-12 . This may mean that TRL3 is redundant amongst the dsRNA sensors, the levels of dsRNA produced during replication are not sufficient to stimulate the receptor, or that the viral dsRNA simply does not enter the compartments where TRL3 is present. In the same study by Lund et al. , influenza A virus, or VSV, infection could not trigger IFNα and IL-12 production in pDCs from TRL7−/− mice. Concurrent studies in other independent laboratories implicate input single-stranded (ss)RNA from virions as the trigger for signalling and activation of transcription factors through the receptors TRL7 and/or TRL8 [97, 98]. Influenza A virus, or human immunodeficiency virus type 1, were used as model ssRNA viruses in these studies. IFNα production in response to VSV infection also requires MyD88-dependent signalling . Further dissection revealed that murine TRL7 and human TLR8 are the ssRNA sensors for each species . They require acidification of the endosome, which occurs after receptor-mediated internalization of the virus, for the recognition of ssRNA. The natural ligands for murine TLR7 and human TRL8 remain to be identified. Studies on TLR signalling, particularly in a viral context, mainly focus on the induction of IFN production through receptor activation and signalling to the IFN enhanceosome. However, the role of TRLs in the induction of inflammation during virus infection has yet to be explored in detail. As dendritic cells express a wide range of TLRs, these findings further indicate their critical role at the crossroads of innate and adaptive immunity. It also implicates TRL7 and DCs as prime players in the potentiation of the innate and proinflammatory responses to influenza virus infection.
The induction of proinflammatory cytokines such as RANTES during influenza virus infection is regulated through the p38 MAP Kinase/JNK pathway  and AP-1 dependent gene expression is also activated by this pathway . As these transcription factors have not been shown to be activated by TLR signalling as yet, it is quite possible that dsRNA activation of PKR is the stimulus for induction of this response to viral infection .
Cytokine production during influenza virus infection
Most in vitro studies of cytokine induction in response to influenza virus infection have been carried out in monocyte/macrophage cell lines. The production of the cytokines IFN-α, TNF- α, IL-1, IL-6 and the mononuclear cell attractant chemokines MIP-1α, MIP-1β, MCP-1, MCP-3, IP-10 and RANTES in cell culture using human monocytes, rat alveolar or murine macrophages in response to influenza virus infection, has been documented [100–103].
However, it is important to note that at the time of infection these immune cells would not be present in great numbers until they have been recruited into the area. In fact, it has been shown that in the normal healthy respiratory epithelium, using human biopsy or brush samples, immune cells typically make up less than 1% (bronchial epithelium) or 2% (nasal epithelium) of intraepithelial cells, the dominant cell type being epithelial . Therefore, it is important not to ignore the role of cytokine production by infected epithelial cells of the respiratory mucosa . In epithelial cells infected with influenza A virus in vitro, chemokine expression and release has now been documented particularly IL-6, IL-8 and RANTES [66, 106–109] and possibly MCP-1 (unpublished results, see reviews [92, 93]). The facts that very few monocytic or dendritic cells are present at the time of infection in an otherwise healthy person and that the neutrophilic chemokine IL-8 is produced by epithelial cells concurs with evidence for the early infiltration of neutrophils. However, even the small numbers of immune cells present could, if infected, produce a large amount of cytokines as well as capture and present antigen. The secretion of MCP-1 and RANTES could act as activators of macrophage function in the few that are present to recruit more to the site and amplify the immune response.
In vivo studies
As with the immune response, in vivo studies of influenza virus infection have centred on the mouse, although ferrets and pigs have also been studied. In several independent studies intranasal infection of mice with A/PR/8/34 has shown that IFN-α TNF-α, IL-1 α and β and IL-6 levels in broncho-alveolar lavage (BAL) fluid or lung homogenates rise soon after infection [27, 110–112]. This is in temporal association with symptom expression and lung pathology. Experiments in which infected mice have been treated with antibodies to cytokine have also been important in investigating the network of cytokine interactions. They implicate IFN-α and IL-1 α, as well as showing a possible cascade of induction; anti-IFN antibody treatment lowers IL-1 levels . In addition, IL-1β knock-out mice (IL-1β−/−) exhibited greater mortalities due to influenza virus infection than wild-type mice , which correlates with work done using IL-1 receptor agonist (IL-1ra) . In addition, it was recently found that influenza virus infection of IL-18 knockout mice (IL-18−/−) exhibited greater mortality than wild-type mice, pathological changes included massive inflammatory cell infiltrate to the lungs . This correlated with a decreased activation of NKT cells in the lung and suggests an important role for the early innate immune response in controlling the level of inflammation as well as viral replication.
However, influenza virus is not a natural pathogen of mice and usually requires several passages through them to become mouse-adapted and cause disease. Moreover, clinical manifestations of the disease differ somewhat from the natural hosts; most mouse strains show a drop in body temperature instead of fever and the infection is usually lethal .
The ferret model is useful as it can be infected with human strains of influenza virus and exhibits the same clinical symptoms as humans during a predominantly upper respiratory tract infection . However, these studies are prohibited by the lack of reagents currently available.
As yet, swine have not been used extensively in influenza research. However, some research has been conducted, involving a gnotobiotic model. Clinically there is lethargy, shivering, anorexia, tachypnoea, laboured respiration and coughing. The pathology includes excessive lung epithelial desquamation and massive infiltration of inflammatory cells into the bronchi and bronchioli, with neutrophils being especially evident . There was a strong temporal association (levels rose and fell together) between virus titres, neutrophil infiltration, disease, and IFN-α, TNF-α, and IL-1 levels in BAL fluids [118, 119].
Human volunteer studies
The response in humans has been studied by infection of human volunteers with wild-type H1N1 or H3N2 strains that produce a characteristic illness with viral shedding and a febrile upper respiratory tract response. Nasal lavage fluids have been shown to contain elevated levels of pro-inflammatory cytokines such as IFN-α, TNF-α, IL-6, IL-8, MIP-1α and β, and MCP-1 [22–24, 120]. The peak of production occurred within 2 or 3 days of inoculation in all studies and correlated directly with nasal virus titres and symptoms (as shown by scoring systems). However, the levels in plasma, denoting systemic cytokine production, were much lower than in the respiratory tract. It can be argued that IFN-α and IL-6 may be responsible for most of the early symptoms [21–23].
Linking apoptosis, inflammation and symptoms in influenza
Apoptosis, as opposed to necrosis, is by definition a non-inflammatory event, phagocytes clearing apoptotic cells before they can release their contents and elicit an inflammatory response. This is certainly true in development but symptoms during influenza infection are an expression of damage. As both apoptosis and inflammation induce damage it is becoming increasingly evident that symptom expression is linked to these processes. There is a good temporal correlation between the induction of the inflammatory response and the onset of constitutional symptoms in ferrets and humans.
From the activation pathways for apoptosis, fever and inflammation discussed above, it would seem that similar cytokines and chemokines are mediators of all three responses. Early cytokines are produced by non-immune cells at the site of infection and are responsible for initiation of the local inflammatory response, as well as some systemic effects . Research has shown that IL-8, a pyrogenic as well as a pro-inflammatory cytokine, is secreted on induction of apoptosis in bronchiolar epithelial cells by Fas ligation . Again, the system is of relevance to influenza due to the virus’ capacity to infect the respiratory tract epithelium and induce Fas-dependent apoptosis [57, 61]. Similarly, IL-1β (pyrogenic and pro-inflammatory) and IL-18 (pro-inflammatory) production is linked with apoptosis in macrophages where they are proteolytically processed by virus-activated ICE/caspase-1 and caspase-3 [122, 123].
IL-1β and IL-8 could work in concert. The former is a powerful stimulator of neutrophil and macrophage function and can also upregulate leukocyte adhesion molecules on the vascular endothelium. This would mediate the first step of neutrophil/macrophage recruitment into the respiratory tract, whilst IL-8 would complete it by promoting stable binding to the vascular endothelium and activation of the neutrophils. This would then lead to a cascade of pro-inflammatory cytokine induction and secretion by activated immune cells to amplify the response. Local release of IL-1β and IL-8 need not necessarily be in large quantities, just large enough to stimulate the local immune cells. This is given further credence when one realises that cells undergoing apoptosis may not be able to upregulate cytokine gene transcription or translation into bioactive protein. This may be compensated for in the case of IL-1β and IL-18. These proteins are synthesized as immature precursors that need to be proteolytically cleaved for activation by the enzyme ICE/caspase-1. If caspase-1 were activated during apoptosis then this event would occur, cleaving the pro-form into mature, active IL-1β and IL-18. The precursor accumulates in the cytosol of the cell as it does not have a secretory signal peptide, thus apoptosis activates and releases it . It has thus been suggested that IL-1β and IL-18 act as emergency proinflammatory cytokines.
The proposed standard sequence of the cytokine cascade seems to have IFN-α/β, TNF-α, and both IL-1α and β as the earliest mediators of the response, with IL-6 following closely, and then the chemokines. In fact, one model proposes that IFN-α secretion leads to activation and secretion of IL-1 . The others would be activated and secreted as the cascade gained momentum and immune cells became involved either through clearing apoptotic cells, activation by IL-1 or direct infection. This is supported by the fact that type 1 interferons potentiate influenza virus-induced apoptosis  and that the IFN antagonist of influenza virus, NS1, also downregulates virally-induced apoptosis in an IFN-dependent manner .
There is a delicate balance between these responses in limiting the amount of damage they cause whilst eliminating the virus. IL-18 has recently been shown to be important in the resolution of infection and the prevention of an overwhelming inflammatory response in mice . In fact apoptosis as a result of infection may help control an acute inflammatory response by limiting the amount of cytokine produced through the progression of the apoptotic cascade. We have shown that cleavage of the Golgi body during influenza virus-induced apoptosis correlates with a decrease in cytokine release and that caspase inhibitors can increase the release of these cytokines .
Signalling pathways at the crossroads of apoptosis and inflammation
Signalling within the cell after influenza virus infection involves IFN induction pathways, IFN signalling pathways through Jak/STAT if cells are able to overcome NS1-mediated IFN antagonism, apoptotic signalling pathways and other kinase cascades. It is likely there is significant cross-talk between pathways. NF-κB is a key central molecule involved in the transcription of many genes that are responsive to these signalling pathways and is described as a key regulator of life and death of the cell, particularly in lymphoid cells .
Several papers describe the activation of the p38MAPK/JNK pathway leading to AP-1-dependent transcription of cytokines [63, 99]. However, it is also thought that this signalling pathway is activated in stress-induced apoptosis during influenza virus infection  demonstrating the potential for signalling pathways to act upon two different targets. It is not clear if these are temporally spaced responses. The apoptotic module of this signalling pathway seems to be mediated through ASK1 . This stress activated pathway is also the one potentially responsible for NF-κB activation on overexpression of viral glycoproteins [72, 126] and the generation of oxygen radicals has been implicated in both apoptosis  and NF-κB activation [72, 127]. As stated previously the association of oxygen radicals with these signalling events is not yet clear .
It is clear that these signalling pathways converge on multiple targets that affect both the way a cell signals to its neighbours and whether the cell decides to die. Ultimately these pathways have consequences for the protection of the host from viral replication and resolution of the infection. Apoptosis and inflammation in this respect seem to be intrinsically linked, whether one positively or negatively regulates the other may depend on the temporal activation of pathways and the signal that triggers them. In order for the host to resolve the infection, a delicate balance must be maintained between cell life and death, the pro-inflammatory response and the innate antiviral response.
Influenza virus infection of epithelial cells and phagocytes induces apoptosis. This not only leads to direct damage of such cells but it is becoming apparent that such cells produce pro-inflammatory cytokines, some of which require activation by caspases induced during apoptosis. These cytokines and chemokines attract leucocytes to the site of infection. This inflammatory response directly damages tissues inducing respiratory symptoms but it may also indirectly damage the host as the inflammatory cells release pyrogenic cytokines that not only induce fever but all the constitutional signs and symptoms of influenza. This response is primarily protective but a delicate balance needs to be maintained. When upset, as in the overproduction of pro-inflammatory cytokines by H5N1 subtypes isolated in Asia in 1997 and 2003 , the consequences may be extremely serious even fatal.
E.W.A.B. and S.J.M. were supported by BBSRC studentships and S.J.M. by a BBSRC project grant.