Ringing the alarm bells: signalling and apoptosis in influenza virus infected cells

Authors


*E-mail ludwigs@uni-muenster.de; Tel. (+49) 251 83 57791; Fax (+49) 251 83 57793.

Summary

Small RNA viruses such as influenza viruses extensively manipulate host-cell functions to support their replication. At the same time the infected cell induces an array of defence mechanisms to fight the invader. These processes are mediated by a variety of intracellular signalling cascades. Here we will review the current knowledge of functional kinase signalling and apoptotic events in influenza virus infected cells and how these viruses have learned to misuse these cellular responses for efficient replication.

Introduction

Influenza A viruses are highly contagious pathogens for humans and several animal species. These viruses belong to the family of Orthomyxoviridae and possess a segmented negative-stranded RNA genome of roughly 13 kB, coding for at least 10 viral proteins (Lamb and Krug, 2001). The viral genome is replicated and transcribed in the nucleus, a feature that requires bidirectional transport through the nuclear membrane. Due to the limited coding capacity of the genome these viruses extensively employ functions of their host-cell for efficient replication (Ludwig et al., 1999).

Cell fate decisions in response to extracellular agents, including pathogenic invaders are commonly mediated by intracellular signalling cascades that transduce signals into stimulus specific actions, e.g. changes in gene expression patterns, alterations in the metabolic state of the cell or induction of programmed cell death (apoptosis). Thus, these signalling molecules are at the bottleneck of the control of cellular responses. Many DNA- and retroviruses are known to induce cellular signalling mainly to drive cells into a proliferative state. The reason is quite obvious because these pathogens partly employ the DNA synthesis machinery for their replication. The consequences of signalling induced by RNA viruses, including influenza viruses were less clear because this area was not a focus of research for a long time. However, in the last couple of years reports on functional signal transduction processes induced by RNA viruses rapidly accumulated. As detailed below, a majority of the more recent reports address the signalling events concomitant with the innate immune response. Nevertheless, there are also findings suggesting that RNA viruses misuse cellular signals to support their replication. Here we will discuss recent advances in the analysis of influenza virus-induced signalling pathways and the insights these studies provide for our understanding of the viral replication process.

Influenza viruses and MAP kinase cascades

Mitogen-activated protein kinase (MAPK) cascades are important signalling pathways that convert a variety of extracellular signals into a multitude of cellular responses (reviewed in Pearson et al., 2001). These signalling cascades regulate cellular decision processes as diverse as proliferation and differentiation, but also cell activation and immune responses (Dong et al., 2002). Four different prototype members of the MAPK family that are organized in separate cascades have been identified so far: ERK (extracellular signal-regulated kinase), JNK (Jun-N-terminal kinase), p38 and BMK-1/ERK5 (Big MAP kinase) (Pearson et al., 2001) (Fig. 1). For each MAP kinase different isoforms are known. All these enzymes have in common that they are activated by a dual phosphorylation event on threonine and tyrosine mediated by upstream MAP kinase kinases (MEKs or MKKs) (Fig. 1). The MAP kinase ERK1 and ERK2 are activated by the dual-specific kinases MEK1 and 2 that are controlled by the serine threonine kinase Raf. Raf, MEK and ERK form the prototype module of a MAP kinase pathway. This three-kinase module is also known as the classical mitogenic cascade. The MAP kinases p38 and JNK are activated by MKK3/6 and MKK4/7, respectively, and are predominantly activated by proinflammatory cytokines and certain environmental stress conditions. ERK5, also known as big-MAP kinase (BMK-1) is activated by MEK5 (Fig. 1). This kinase module is special because it is both activated by mitogens and certain stress inducers.

Figure 1.

Schematic representation of intracellular signalling pathways that are activated upon influenza virus infection or treatment with viral components and the proposed function of these cascades in the infected cell. For clarity, the major pathways of the antiviral IRF activation and induction of the IFN response are shown in a separate figure (Fig. 2).

Influenza virus infection of cultured cells results in the activation of all four known MAPK family members (Kujime et al., 2000; Ludwig et al., 2001; Pleschka et al., 2001) (Fig. 1). Activation of p38 MAPK after influenza virus infection has been linked to expression of RANTES and IL-8, chemokines involved in the attraction of eosinophils and neutrophils respectively (Kujime et al., 2000; Guillot et al., 2005). Furthermore, inhibition of p38 and other MAPKs results in decreased prostaglandin E2 release (Mizumura et al., 2003), indicating that p38 MAPK activation controls the onset of an inflammatory response. This is also supported by the recent finding that hyperinduction of TNF-α upon infection with avian H5N1 influenza viruses occurs in a p38-dependent manner (Lee et al., 2005).

The JNK subgroup of MAPKs came into focus in the context of an influenza virus infection because a very early activation of activator-protein 1 (AP-1) transcription factors (Karin et al., 1997) was observed in productively infected cells (Ludwig et al., 2001). AP-1 factors include c-Jun and ATF-2 that are phosphorylated by JNKs to potentiate their transcriptional activity (Karin et al., 1997). Accordingly, activation of JNK was observed with different virus strains in a variety of permissive cell lines (Kujime et al., 2000; Ludwig et al., 2001). JNK activation required productive replication and was induced by accumulating RNA produced by the viral polymerase.

Together with NF-κB and IRF-3 the JNK effectors c-Jun and ATF-2 are critical to regulate the expression of IFNβ, one of the most potent antiviral cytokines (Samuel, 2001) (Figs 1 and 2). Accordingly, inhibition of the cascade by dominant-negative mutants of c-Jun, JNK or the JNK activator MKK7 in the infected cell resulted in impaired transcription from the IFNβ promoter and an enhanced virus production. Thus, the JNK pathway appears to be a crucial mediator of the antiviral response to an influenza virus infection by coregulating IFNβ expression (Ludwig et al., 2001).

Figure 2.

Induction pathways of IFN-α/β genes during virus infection.
Left. Productive virus infection induces the appearance of dsRNA, that is recognized by the RNA helicase RIG-I. This leads, via the adapter protein IPS-1, to the phosphorylation and activation of the key transcription factors IRF3 and/or IRF7 by the protein kinases TBK1 and/or IKK-ɛ. Phosphorylated IRF-3/-7 dimerizes and translocates to the nucleus, where it becomes engaged in the activation of IFN-α/β genes. The influenza virus NS1 protein is known to inhibit activation of IRF-3/7 indicating a blockade of dsRNA-dependent upstream signals at a currently unidentified step.
Right. Uptake of influenza virus into plasmacytoid dendritic cells results in the exposure of virion single-stranded (ss) RNAs to Toll-like receptor (TLR)-7 in an endosomal compartment. Agonist recognition of the TLR leads through the adapter MyD88 and TRAF6 to recruitment of an unknown kinase that activates IRF-7. DsRNA released from infected cells can activate TLR3 either in the endosome or on the cell surface, which leads through to the recruitment of the adapter protein TRIF and TBK1 to the activation of IRF-3.

In contrast to the JNK pathway the Raf/MEK/ERK cascade appears to serve as a module that is beneficial for the virus (Pleschka et al., 2001). Blockade of the pathway by specific inhibitors of MEK, or dominant-negative mutants of ERK or Raf resulted in a strongly impaired growth of both influenza A and B viruses (Pleschka et al., 2001; Ludwig et al., 2004). Conversely, virus titres were enhanced in cells expressing active mutants of Raf or MEK (Ludwig et al., 2004; Olschlager et al., 2004). This has not only been demonstrated in cell culture but also in vivo in infected mice expressing a constitutively active form of the Raf kinase in the alveolar epithelial cells of the lung (Olschlager et al., 2004). In the wild-type situation influenza viruses primarily infected bronchiolar epithelial cells, while in the alveolar layer replication occurs most exclusively in cells carrying the transgene. As a consequence this resulted in an earlier death of the transgenic animals (Olschlager et al., 2004).

Strikingly, inhibition of the pathway did not affect viral RNA or protein synthesis (Pleschka et al., 2001). The pathway rather appears to control the active nuclear export of the viral RNP complexes. RNPs are readily retained in the nucleus upon blockade of the signalling pathway. This is most likely due to an impaired activity of the viral nuclear export protein NEP (Pleschka et al., 2001). Thus, active RNP export appears to be at least in part an inducible event, a hypothesis supported by a late activation of ERK in the viral life cycle. However, the detailed mechanism of how ERK regulates export of the RNPs remains to be elucidated.

The requirement of Raf/MEK/ERK activation for efficient influenza virus replication suggests that this pathway may be a cellular target for antiviral approaches. Besides the antiviral action against both type A and B viruses (Ludwig et al., 2004), MEK inhibitors meet two further criteria which are a prerequisite for a potential clinical use. Although targeting an important signalling pathway in the cell, the inhibitors showed surprisingly little toxicity (i) in cell culture (Planz et al., 2001; Pleschka et al., 2001; Ludwig et al., 2004); (ii) in an in vivo mouse model (Sebolt-Leopold et al., 1999); and (iii) in clinical trials for the use as anticancer agent (Cohen, 2002). In the light of the latter finding it was hypothesized that the mitogenic pathway may only be of major importance during early development of an organism and may be dispensable in adult tissues (Cohen, 2002). Another very important feature of MEK inhibitors is that they showed no tendency to induce formation of resistant virus variants (Ludwig et al., 2004). Although targeting of a cellular factor may still raise the concern of potential side-effects of a drug, it appears likely that local administration of an agent such as a MEK inhibitor to the primary site of influenza virus infection, the respiratory tract, is well tolerated. Here the drug primarily affects differentiated epithelial cells, for which a proliferative signalling cascade such as the Raf/MEK/ERK cascade may be dispensable. Following this approach it was recently demonstrated that the MEK inhibitor U0126 is effective in reducing virus titres in the lung of infected mice after local administration (Ludwig et al., 2003; Klumpp, 2004).

Protein Kinase C (PKC) as a regulator of viral entry

Activation of the kinase Raf within the Raf/MEK/ERK module is complex and involves phosphorylation by several other kinases including members of the PKC family (Kolch et al., 1993; Cai et al., 1997).

The PKC superfamily consists of at least 12 different PKC isoforms that carry out diverse regulatory roles in cellular processes by linking into several downstream signalling pathways (Toker, 1998). It has been noticed quite some time ago that influenza virus infection or treatment of cells with purified viral haemagglutinin results in rapid activation of PKCs upon binding to host-cell surface receptors (Rott et al., 1995; Arora and Gasse, 1998; Kunzelmann et al., 2000) (Fig. 1). However, the functional consequences of this activation remained elusive. Given the variety of downstream effectors of PKCs (Toker, 1998) it appears likely that beside a regulation of the Raf/MEK/ERK cascade and other downstream pathways, PKCs may have additional functions during viral replication. This assumption is supported by the finding that the viral M1 protein gets phosphorylated by PKC in vitro and binds to the cellular receptor of activated C kinase RACK1 (Reinhardt and Wolff, 2000). A role of PKCs in the process of entry of several enveloped viruses has been proposed based on the action of protein kinase inhibitors H7 and staurosporine (Constantinescu et al., 1991). In more recent studies it was shown that the pan-PKC inhibitor bisindolylmaleimide I prevented influenza virus entry and subsequent infection in a dose-dependent and reversible manner (Root et al., 2000). Using a dominant-negative mutant approach this function was assigned to the PKCβII isoform: Overexpression of a phosphorylation-deficient mutant of PKCβII revealed that the kinase is a regulator of late endosomal sorting. Accordingly, expression of the PKCβII mutant resulted in a block of virus entry at the level of late endosomes (Sieczkarski and Whittaker, 2002; Sieczkarski et al., 2003). This identifies PKCβII as a specific regulator of influenza virus entry (Fig. 1).

Influenza Virus and the classical pathway of NF-κB activation

Another important signalling pathway which is commonly activated upon virus infection is the IκB kinase (IKK)/NF-κB signalling module (Hiscott et al., 2001). The NF-κB/IκB family of transcription factors promote the expression of well over 150 different genes, such as cytokine or chemokine genes, or genes encoding for adhesion molecules or anti- and pro-apoptotic proteins (Pahl, 1999). The classical mechanism of NF-κB activation includes activation of IKK that phosphorylates the inhibitor of NF-κB, IκB and targets the protein for subsequent degradation (Karin and Ben-Neriah, 2000) (Figs 1 and 3). This leads to the release and migration of the transcriptionally active NF-κB factors, such as p65 or p50 to the nucleus (Ghosh, 1999; Karin and Ben-Neriah, 2000). The IKK complex consists of at least three components, namely the enzymatically active IKK1/IKKα and IKK2/IKKβ and the scaffold protein NEMO/IKKγ. The most important isozyme for NF-κB activation via the degradation of IκB is IKK2 while IKK1 seems to primarily phosphorylate other factors of the NF-κB/IκB family namely p100/p52 (reviewed in Bonizzi and Karin, 2004). The large IKK complex (Karin and Delhase, 2000) appears to contain still other kinases such as MEKK1 (MAPK kinase kinase 1), TAK1 (TGFβ-activated kinase), MLK-3 (Mixed-lineage kinase 3), NIK (NF-κB inducing kinase), and the double-stranded (ds) RNA-activated protein kinase PKR (reviewed in Hiscott et al., 2001).

Figure 3.

Model of the chain of events that links virus-induced NF-κB activation to caspase induction and enhanced RNP export. During productive virus infection the proapoptotic factors TRAIL, Fas and FasL are expressed in an NF-κB-dependent manner. These factors induce caspase activation in an auto- and paracrine fashion. Active caspases allow an enhanced release of RNP complexes from the nucleus, presumably due to caspase-mediated disruption of the active nuclear pore complex (see text for further details). Alternatively, caspase activation can also be achieved by expression of the viral PB1-F2 protein although this appears not to be linked to RNP export. It should be stressed that the mechanism shown may not be the only event regulating the NF-κB- and caspase-mediated outcome on viral propagation.

Influenza virus infection results in an activation of NF-κB (reviewed in Ludwig et al., 1999; 2003; Julkunen et al., 2000) although the level of activation is kept in a certain limit due to the action of the viral NS1 protein (Wang et al., 2000), as will be further discussed below. Nevertheless, the induced activity is sufficient to control expression of a variety of genes (Wurzer et al., 2004; Bernasconi et al., 2005). Viral induction of the transcription factor involves activation of IKK2 (Flory et al., 2000; Wurzer et al., 2004; Bernasconi et al., 2005) and is also achieved with isolated influenza virus components. This includes single-stranded (ss) RNA (Heil et al., 2004; Kawai et al., 2004) or double-stranded (ds) RNA (Chu et al., 1999) as well as overexpression of the viral HA, nucleoprotein (NP) or M1 proteins (Pahl and Baeuerle, 1995; Flory et al., 2000) (Fig. 1). As gene expression of many proinflammatory or antiviral cytokines, such as IFNβ or TNF-α, is controlled by NF-κB (Pahl, 1999) the concept emerged that IKK and NF-κB are essential components in the innate immune response to virus infections (Chu et al., 1999). Accordingly, influenza virus-induced IFNβ promoter activity is strongly impaired in cells expressing transdominant negative mutants of IKK2 or IκBα (Wang et al., 2000; Wurzer et al., 2004).

Nevertheless, IKK and NF-κB might not only have antiviral functions. Two recent studies demonstrate that influenza viruses exhibit higher levels of replication in cells where NF-κB is preactivated (Nimmerjahn et al., 2004; Wurzer et al., 2004). Conversely, a dramatic reduction of influenza virus titres could be observed in cells were NF-κB signalling was impaired (Nimmerjahn et al., 2004; Wurzer et al., 2004). This is different from the situation with other RNA viruses, e.g. Borna disease virus (BDV) where constitutive activation of NF-κB clearly leads to a drop in virus titres (Bourteele et al., 2005). Thus, in the context of an influenza virus infection NF-κB appears to have a supportive function for viral replication that is dominant over the antiviral activity induced by the transcription factor. On a molecular basis this was shown to be at least in part due to the NF-κB-dependent expression of proapoptotic factors, such as TNF-related apoptosis inducing ligand (TRAIL) or FasL (Wurzer et al., 2004). Inhibition of virus induced expression of these factors results in strongly impaired viral growth. These findings link the pro-influenza action of NF-κB to the induction of apoptosis, a process that will be further discussed below (Fig. 3). Finally, viral need for NF-κB activity suggests that this pathway may be suitable as a target for antiviral intervention. To this end it has been shown recently that pharmacological inhibitors of NF-κB act antiviral in vivo without toxic side-effects or the tendency to induce resistant virus variants (Ludwig et al., 2003).

Mounting the antiviral state – signalling events inducing the type I interferon response

A large proportion of the signalling events in infected epithelial cells at the primary viral replication sites is devoted to the induction of cellular responses that aim to prevent the spread of the invading virus in the tissue and the establishment of a persistent infection. A major part of the antiviral response is the expression and secretion of the interferons (IFN)-α and -β. These antiviral cytokines bind to the IFN-α/β receptor, which by signalling through the JAK-STAT pathway leads to the formation of the trimeric transcription factor ISGF3 that in turn activates a multitude of latent gene products, many of which have strong antiviral activities such as the Mx, p56 and 2′-5′-oligoadenylate synthetases (Samuel, 2001). The potency of the IFN-α/β system is illustrated by the exquisite sensitivity of STAT1–/– mice to viral infections (Durbin et al., 1996). However, during coevolution with their hosts probably most viruses have evolved gene products that interfere with the IFN-α/β system at the induction or effector level as a prerequisite for efficient replication (Garcia-Sastre, 2004). Thus, depending on the particular virus and the target(s) of its antagonistic gene product one may observe high or low contributions of distinct cellular components to the establishment of an antiviral state (Leib et al., 2000).

It has been known for decades that dsRNA is a molecular pattern commonly associated with viral infections and that the intracellular appearance of dsRNA mediates a strong induction of IFN-α/β genes in many cell types (summarized in Majde, 2000). However, although it was established a while ago that IFN induction is controlled by the transcription factors IRF3/-7, NF-κB and ATF-2/c-Jun (Wathelet et al., 1998), the molecular events facilitating this antiviral reaction had remained elusive for a long time. Lately, several breakthrough studies have elegantly identified molecular sensors and at least some of the mediators that ultimately lead to the activation of IRF-3/-7 and hence, IFN-α/β genes. Yoneyama and colleagues recently showed that intracellular dsRNA, a frequent by-product of viral replication, is detected by the dsRNA helicases RIG-I (Yoneyama et al., 2004; Kato et al., 2005) and its homologue mda-5 (Andrejeva et al., 2004), that both contain two caspase recruitment domains (CARD) (Fig. 2). RIG-I mediates the activation of the transcription factors IRF3 and NF-κB and expression of IFN-α/β after virus infection by interaction with another CARD protein, termed interferon-β promoter stimulator 1 (IPS-1; also identified as MAVS, VISA, Cardif) (Kawai et al., 2005; Meylan et al., 2005; Seth et al., 2005; Xu et al., 2005). This protein was shown to interact with overexpressed TBK-1 and IKK-ɛ (Meylan et al., 2005; Xu et al., 2005) that are two members of the IKK family known to activate the latent IRF-3 and IRF-7 transcription factors by phosphorylation (Fitzgerald et al., 2003; Sharma et al., 2003) (Fig. 2). Furthermore, IPS-1 facilitated the recruitment of IRF-3 to RIG-I (Xu et al., 2005) and cells with abolished or ‘knocked-down’ levels of RIG-I or IPS-1, respectively, were severely impaired in IFN secretion and produced highly increased virus titres (Kato et al., 2005; Kawai et al., 2005; Seth et al., 2005). Collectively, these findings span the framework for a model, in which RIG-I and IPS play central roles in initiating the antiviral response (Fig. 2).

In our consideration of how RNA viruses are detected in body cells we like to distinguish between the processes described above for cells that support productive virus replication from the uptake of viruses in professional antigen presenting cells such as dendritic cells (DCs), which mainly serves to initiate an adaptive immune response (Fig. 2). Immature DCs express a variety of pattern recognition receptors belonging to the Toll-like receptor (TLR) family, that recognize various microbial components such as lipopolysaccharide (LPS), peptidoglycan or CpG motif-containing DNA (Akira and Takeda, 2004). Agonist binding to TLRs triggers via the adapter molecules MyD88 and TRIF the induction of pro-inflammatory and antiviral cytokine genes including IFN-α and/or IFN-β, which results in DC maturation (Takeda and Akira, 2005) (Fig. 2). As the TLR family members 3 (dsRNA sensor) (Alexopoulou et al., 2001), 7 and 8 (ssRNA sensors) (Heil et al., 2004; Kawai et al., 2004) are known to be activated by genetic materials also found in RNA viruses, it appeared possible that TLRs might mediate an RIG-I-independent detection of influenza virus infections also in non-haematopoetic cells (Guillot et al., 2005). However, the unaltered course and pathogenesis of influenza- and other negative strand RNA virus infections in TLR3–/– and MyD88–/– mice (Edelmann et al., 2004; Lopez et al., 2004; Barchet et al., 2005) argues against such a generalized role for TLRs as antiviral effectors in the early phase of infection.

The influenza virus NS1 protein interferes with the antiviral response

Although IFN-α/β were originally discovered during the study of influenza virus (Lindenmann et al., 1957), it is ironic to note that reverse genetic studies revealed that these viruses actually possess the capacity to suppress the expression of these cytokines. In comparison to wild type, influenza viruses with deleted NS1 genes proofed to be much stronger inducers of IFN-α/β genes and their growths were highly attenuated in hosts with an intact IFN system (Garcia-Sastre et al., 1998; Dauber et al., 2004). A second known activity of the 26 kDa NS1 protein is to inhibit activation of the antiviral dsRNA-dependent protein kinase R (PKR), which is probably due to NS1's capacity to bind to dsRNA produced during virus replication (Lu et al., 1994; Bergmann et al., 2000). PKR can phosphorylate the translation factor eIF2α, which leads to a sustained arrest in cellular translation and, hence, to an inhibition of viral gene expression (Samuel, 2001). The strong IFN induction during influenza virus infection in the absence of NS1 expression correlated directly with a prominent activation of the transcription factors IRF-3/-7, NF-κB and ATF-2/c-jun (Ludwig et al., 2003). Thus, the NS1 protein is capable of inhibiting the virus-induced dsRNA signals leading to activation of antiviral cytokine genes (Fig. 2). Whether NS1 functions merely by intracellular sequestration of dsRNA molecules or by direct targeting of any of the recently described factors such as RIG-I or IPS-1 that transduce these signals is currently under investigation. Preliminary data showed that recombinant influenza viruses expressing NS1 proteins with inactivated RNA binding can still effectively block viral IFN induction suggesting that dsRNA sequestration plays only a minor role in IFN antagonization (B. Dauber and T. W., unpubl. obs.). The NS1 protein has also been described to inhibit the maturation of cellular pre-mRNAs raising the possibility that this activity additionally reduces secretion of IFN-α/β from infected cells (reviewed in Krug et al., 2003). The strong induction of IFN-α in plasmacytoid DCs by wild-type influenza viruses (Diebold et al., 2004) suggests that the viral NS1 protein does not interfere significantly with TLR-dependent signals, but rather specifically targets the RIG-I-dependent pathway of IFN induction.

Influenza virus-induced caspase activation and apoptosis – submerging a suicide program

An important cellular signalling response commonly observed upon virus infections is the induction of the apoptotic cascade. Apoptosis is a morphologically and biochemically defined form of cell death (Kerr et al., 1972) and has been demonstrated to play a role in a variety of diseases including infections by viruses (Razvi and Welsh, 1995; Fischer and Schulze-Osthoff, 2005). Apoptosis is mainly regarded to be a host cell defence because many viruses express antiapoptotic proteins to prevent this cellular response. The central component of the apoptotic machinery is a proteolytic system consisting of a family of cysteinyl proteases, termed caspases (for review see Cohen, 1997; Thornberry and Lazebnik, 1998). Two groups of caspases can be distinguished: upstream initiator caspases such as caspase 8 or caspase 9 which cleave and activate other caspases and downstream effector caspases, including caspases 3, 6 and 7, cleaving a variety of other cellular substrates, thereby disassembling cellular structures or inactivating enzymes (Thornberry and Lazebnik, 1998). Caspase 3 is the most intensively studied effector caspase. Work on MCF-7 breast carcinoma cells which are deficient in caspase 3 due to a deletion in the casp3 gene has revealed the existence of a crucial caspase 3 driven feedback loop which mediates the apoptotic process (Janicke et al., 1998; Slee et al., 1999). Thus, caspase 3 is a central player in apoptosis regulation and the level of procaspase 3 in the cell determines the impact of a given apoptotic stimulus.

It is long known that infection with A and B type influenza viruses results in the induction of apoptosis both in permissive and non-permissive cultured cells as well as in vivo (Takizawa et al., 1993; Fesq et al., 1994; Hinshaw et al., 1994; Mori et al., 1995; Ito et al., 2002). Interestingly, viral activation of MAPKs or upstream kinases has been linked to the onset of apoptosis. In a mouse model for a neurovirulent influenza infection, JNK but not p38 activity correlated with apoptosis induction in the infected brain (Mori et al., 2003). In embryonic fibroblasts deficient for the MAPK kinase kinase ASK-1 (Fig. 1) virus-induced p38 and JNK activation was blunted concomitant with an inhibition of caspase 3 activation and virus-induced apoptosis (Maruoka et al., 2003). As an extrinsic mechanism of viral apoptosis induction it has been noted quite early on that the Fas/FasL apoptosis inducing receptor/ligand system (Takizawa et al., 1993; 1995; Wada et al., 1995; Fujimoto et al., 1998) is expressed in a PKR-dependent manner in infected cells (Takizawa et al., 1996). This most likely contributes to virus-induced cell death via the receptor mediated FADD/caspase 8-dependent pathway (Balachandran et al., 2000). Another mode of viral apoptosis induction might occur via activation of TGF-β, a known apoptosis inducer that is converted from its latent form by the viral neuraminidase (Schultz-Cherry and Hinshaw, 1996). Within the infected cell the apoptotic program is mediated by activation of caspases (Takizawa et al., 1999; Zhirnov et al., 1999; Lin et al., 2002) with a most crucial role of caspase 3 (Wurzer et al., 2003).

Although it is now well established that influenza virus infection induces caspases and subsequent apoptosis, the consequence of this activation for virus replication or host cell defence is still under a heavy debate (reviewed in Schultz-Cherry et al., 1998; Ludwig et al., 1999; Lowy, 2003). Early studies demonstrated that overexpression of the antiapoptotic protein Bcl-2 results in impaired virus production correlating with a misglycosylation of the viral surface protein haemagglutinin (Hinshaw et al., 1994; Olsen et al., 1996). Furthermore, it has been shown that the viral NS1 protein has pro-apoptotic features and induces apoptosis when ectopically expressed (Schultz-Cherry et al., 2001). These data have been challenged by the finding that a recombinant influenza virus lacking the NS is a stronger apoptosis inducer than the wild type suggesting an antiapoptotic function of NS1 (Zhirnov et al., 2002a). The findings link viral apoptosis induction to the antiviral type I interferon (IFNα/β) response, because the NS1 protein was shown to be an efficient IFN-α/β antagonist (Garcia-Sastre, 2004) and type I interferons are believed, besides the Fas/FasL system, to be main inducers of influenza virus induced apoptosis (Balachandran et al., 2000). Another finding in favour of an antiviral role of apoptosis is caspase-mediated cleavage of the NP of human influenza virus strains (Zhirnov et al., 1999). The truncated form of the NP is not packaged into viral particles, an observation that has led to the suggestion that caspases may act to limit amounts of virus protein for proper assembly. Furthermore it has been demonstrated in an in vitro binding assay that the viral M1 protein specifically binds to caspase 8 and weakly to caspase 7, suggesting interference of M1 with a caspase 8 mediated apoptosis pathway (Zhirnov et al., 2002b).

The PB1-F2 protein has recently been characterized as a new pro-apoptotic influenza A virus protein that is expressed from a + 1 reading frame of the PB1 polymerase gene segment (Chen et al., 2001). PB1-F2 induces apoptosis via the mitochondrial pathway when added to cells and infection with recombinant viruses lacking this protein results in reduced apoptotic rates of lymphocytes (Chen et al., 2001) (Fig. 3). The protein contains a mitochondrial target sequence (Gibbs et al., 2003; Yamada et al., 2004) and interacts with the inner mitochondrial membrane adenine nucleotide translocator 3 (ANT3) and the outer mitochondrial membrane voltage-dependent anion channel 1 (VDAC), both of which are implicated in the mitochondrial permeability transition during apoptosis (Zamarin et al., 2005).

These results have let to the assumption that apoptosis induction by PB1-F2 may be required for the specific depletion of lymphocytes during an influenza virus infection, a process which is observed in infected animals (Van Campen et al., 1989; Tumpey et al., 2000). Others have suggested that apoptosis may serve also to boost the induction of cytotoxic T cell responses, because apoptotic cells or materials are efficiently taken up by macrophages or DCs by phagocytosis (Albert et al., 1998; Watanabe et al., 2002). Furthermore, it was reported that viral induction of the apoptotic process limits the release of proinflammatory cytokines and thereby may reduce the severity of the inflammatory response to infection (Brydon et al., 2003). However, no direct proof for each of the suggested functions has been given so far.

A recent study adds a new aspect to the open discussion by the surprising observation that influenza virus propagation was strongly impaired in the presence of caspase inhibitors (Wurzer et al., 2003). This dependence on caspase activity was most obvious in cells where caspase 3 was partially knocked-down by siRNA (Wurzer et al., 2003). Consistent with these findings, poor replication efficiencies of influenza A viruses in cells deficient for caspase 3 could be boosted 30-fold by ectopic expression of the protein. Mechanistically, the block in virus propagation appeared to be due to the retention of viral RNP complexes in the nucleus preventing formation of progeny virus particles (Wurzer et al., 2003) (Fig. 3). As influenza viral apoptosis induction has been linked to the pro-viral activity of NF-κB due to a virus-supportive effect of NF-κB-dependent proapoptotic factor (see above Wurzer et al., 2004) a chain of events from NF-κB induction to caspase-mediated regulation of RNP export unravels (Fig. 3).

Interestingly, the findings are consistent with a much earlier report showing that upon infection of cells overexpressing the antiapoptotic protein Bcl-2 the viral RNP complexes were retained in the nucleus (Hinshaw et al., 1994) resulting in repressed virus titres (Olsen et al., 1996). The observation of a caspase requirement for RNP nuclear export was quite puzzling because this export process was shown before to be mediated by the active cellular export machinery involving the viral nuclear export protein (NS2/NEP) (O’Neill et al., 1998; Neumann et al., 2000) and the antiapoptotic Raf/MEK/ERK cascade (Pleschka et al., 2001). Caspase activation does not support, but rather inhibit the active nuclear export machinery by cleavage of transport proteins. This suggests the existence of an alternative strategy by which caspases may regulate RNP export, e.g. by directly or indirectly increasing the diffusion limit of nuclear pores (Faleiro and Lazebnik, 2000) to allow passive diffusion of larger proteins. Such a scenario is supported by the finding that isolated NPs or RNP complexes, which are nuclear if ectopically expressed, can partially translocate to the cytoplasm upon stimulation with an apoptosis inducer in a caspase 3-dependent manner (Wurzer et al., 2003). These findings can be merged into a model in which the RNPs are transported via an active export mechanism in intermediate steps of the virus life cycle. Once caspase activity increases in the cells, proteins of the transport machinery get destroyed, however, widening of nuclear pores may allow the viral RNPs to use a second mode of exit from the nucleus (Faleiro and Lazebnik, 2000) (Fig. 3). Such mechanism could further enhance RNP migration to the cytoplasm in late phase of the viral life cycle and thereby support virus replication. The model of a complementary use of both, active Raf/MEK/ERK-dependent and passive caspase-dependent transport mechanisms is supported by the observation that concentrations of MEK and caspase inhibitors, which can not block influenza virus replication completely, impaired virus propagation much more efficiently when used in combination (Wurzer et al., 2003). Thus, while both pathways do not interfere with each other (Wurzer et al., 2003), they appear to synergize to mediate RNP export via different routes.

Taken together, one may conclude that influenza virus has acquired the capability to take advantage of supposedly antiviral host cell responses to support its propagation. This includes early induction of caspase activity, but not necessarily execution of the full apoptotic process that most likely is an antiviral response.

Conclusions

A variety of signalling pathways induced by influenza viruses have been described in the last few years and evidences and suggestions for the activating components and functions in the cell have been provided. Still other pathways are beginning to unravel, such as the Phosphatidylinositol 3 kinase (PI3K)-Akt pathway or the Rac1/p21-activated kinase (PAK) module, that are both activated upon influenza virus infection (Ehrhardt et al., 2004; Guillot et al., 2005) and appear to be required for TBK-1-independent (Sarkar et al., 2004) or TBK-1-dependent IRF-3 activation respectively (Ehrhardt et al., 2004).

The picture emerging implicates, that most of the signalling events are initiated by the infected cell as an alert signal to fight the invader. Thus, virus induced cellular signalling can be considered as an antiviral response. Nevertheless, viruses have not only acquired the capability to suppress these responses but also to misuse the remaining activities to support their replication. A prominent example is the NF-κB pathway. Viral NF-κB activation is partially suppressed by the NS1 protein, presumably to prevent an overshooting expression of IFNβ, but at the same time the virus appears to take advantage of the remaining NF-κB activity for apoptosis related and virus supporting processes. This is an economic way for the virus to control efficient replication without the need for specific viral inducers of cellular activities. Thus, there appears to be no black or white situation. Cellular antiviral responses may not only be partially misused to support viral replication at some point during the life cycle, but even may be turned into a proviral activity. Such a potential bivalent function should be considered when evaluating the impact a given signalling pathway has on virus growth. If a virus-supportive activity is dominant over an antiviral action one may even consider to use these cellular signalling components as targets for an antiviral intervention (Ludwig et al., 2003; Nimmerjahn et al., 2004).

Acknowledgements

This work is dedicated to the 75th birthday of Christoph Scholtissek, a pioneer in influenza virus research. We are thankful for support by different grants from the Deutsche Forschungsgemeinschaft (DFG) and by the Fonds der Chemischen Industrie (FdChI). Furthermore, this work is part of the activities of the VIRGIL European Network of Excellence on Antiviral Drug Resistance supported by a grant (LSHMCT-2004-503359) from the Priority 1 ‘Life Sciences, Genomics and Biotechnology for Health’ programme in the 6th Framework Programme of the EU.

Ancillary