Vaccinia virus protein C4 inhibits NF-κB activation and promotes virus virulence

Vaccinia virus (VACV) strain Western Reserve protein C4 has been characterized and its function and contribution to virus virulence assessed. Bioinformatic analysis showed that C4 is conserved in six orthopoxvirus species and shares 43 % amino acid identity with VACV protein C16, a known virulence factor. A recombinant VACV expressing a C-terminally tagged version of C4 showed that, like C16, this 37 kDa protein is expressed early during infection and localizes to both the cytoplasm and the nucleus. Functional assays using a firefly luciferase reporter plasmid under the control of a nuclear factor kappa B (NF-κB)-dependent promoter demonstrated that C4 inhibits NF-κB activation at, or downstream of, the inhibitor of kappa kinase (IKK) complex. Consistent with this, C4 inhibited interleukin-1β-induced translocation of p65 into the nucleus. A VACV lacking the C4L gene (vΔC4) showed no significant differences from wild-type virus in growth kinetics or spread in cell culture, but had reduced virulence in a murine intranasal model of infection. vΔC4-infected mice exhibited fewer symptoms, lost less weight and recovered 7 days earlier than animals infected with control viruses expressing C4. Furthermore, bronchoalveolar lavage fluid from vΔC4-infected mice had increased cell numbers at day 5 post-infection, which correlated with reduced lung virus titres from this time onward. C4 represents the ninth VACV protein to inhibit NF-κB activation and remarkably, in every case examined, loss of each protein individually caused an alteration in virus virulence, despite the presence of other NF-κB inhibitors.


INTRODUCTION
Vaccinia virus (VACV) is the prototypical member of the genus Orthopoxvirus (OPV) of the Poxviridae, a family of large, complex viruses with dsDNA genomes of 135 kb or more (Moss, 2007). Like other poxviruses, VACV replicates in cytoplasmic factories and encodes many proteins needed for virus transcription and DNA replication (Moss, 2007). VACV is the live vaccine used to immunize against smallpox, an eradicated human disease caused by the antigenically related OPV variola virus (Fenner et al., 1988). After smallpox eradication, interest in VACV continued due to the development of recombinant VACVs as candidate vaccines for other micro-organisms, for instance hepatitis B virus (Smith et al., 1983;Moss et al., 1984;Paoletti et al., 1984), and because VACV is a useful expression vector (Moss, 1996), a tool for immunologists (Bennink et al., 1984: Yewdell et al., 1985McMichael et al., 1986) and a good model system for studying virus-host interactions (Cudmore et al., 1995;Alcamí & Smith, 1996;Doceul et al., 2010).
The VACV genome contains approximately 200 ORFs (Goebel et al., 1990) with a highly conserved central region encoding proteins required for VACV transcription, replication and assembly (Upton et al., 2003;Gubser et al., 2004). In contrast, the more variable terminal regions encode proteins that are non-essential for virus replication in cell culture, but which affect virus host range, virulence and immunomodulation in vivo. The latter group includes proteins that are secreted from infected cells to bind cytokines, chemokines, interferons (IFNs) or complement factors, and intracellular proteins that inhibit apoptosis, synthesize steroid hormones or block signalling cascades leading to activation of transcription factors that promote expression of IFNs and pro-inflammatory molecules (Smith, 1994;Alcami, 2003;Seet et al., 2003).
A subset of these VACV immunomodulatory proteins inhibit activation of nuclear factor kappa B (NF-kB), a transcription factor that is retained in the cytosol of resting cells bound to the inhibitor of kappa B alpha (IkBa). When upstream signalling pathways are activated by, for instance, engagement of the interleukin (IL)-1 receptor with IL-1b, tumour necrosis factor (TNF) receptor with TNF-a, or Toll-like receptors (TLRs) with their respective ligands, IkBa is phosphorylated, ubiquitinated and degraded. Consequently, NF-kB is released and translocates to the nucleus, where it activates NF-kB-responsive genes (Hayden & Ghosh, 2008). NF-kB is an important transcription factor for the expression of many pro-inflammatory molecules and therefore it is not surprising that viruses encode proteins that inhibit activation of NF-kB. VACV is no exception and encodes many proteins that target this pathway, including A46, A52, N1, B14, M2, K1, K7 and E3.
This study reports a characterization of VACV strain Western Reserve (WR) protein C4, encoded by gene C4L. Protein C4 has a predicted size of 37.2 kDa and is conserved within other VACV strains and OPVs (http://www.poxvirus. org), suggesting an important function. Protein C4 also shows 43 % amino acid identity to VACV protein C16, which is an immunomodulator and affects virus virulence (Fahy et al., 2008). However, C4 and C16 are functionally distinct because viruses lacking C16 exhibit an attenuated phenotype despite the presence of C4. Data presented here show that C4 is expressed early, is present in both the cytoplasm and the nucleus and inhibits activation of NF-kB at, or downstream of, the IKK complex. Despite being the ninth VACV protein assigned this function, deletion of C4 from the VACV genome caused attenuation of virus virulence. Possible explanations for this observation are discussed.

Construction of C4L deletion, revertant and TAP VACVs
Several recombinant VACVs (strain WR) were constructed (Methods) to study the C4 protein within VACV-infected cells. These included a plaque-purified wild-type virus (vC4), a virus lacking the C4L gene (vDC4) and a revertant virus in which the C4L gene was reinserted at its natural Vaccinia protein C4 blocks NF-kB and is a virulence factor locus (vC4-Rev). To characterize the C4 protein during VACV infection in the absence of a C4 antibody, a virus expressing C4 from its natural promoter and TAP-tagged at the C terminus (vC4-TAP) was constructed. PCR utilizing C4L primers confirmed the presence of C4L in vC4, vC4-Rev and vC4-TAP and its absence in vDC4 (Fig.  S1, available in JGV Online). Analysis of genomic DNA by restriction endonuclease digestion showed that the only discernible difference between these viruses was at the C4L locus (data not shown).

Analysis of C4 expression during VACV infection
To determine when C4 is expressed, BSC-1 cells were infected with vC4-TAP in the presence or absence of cytosine arabinoside (AraC), an inhibitor of viral DNA replication and late protein expression, and extracts of cells were analysed by immunoblotting at different times post-infection (p.i.) (Fig. 1). C4-TAP was detected as a 37 kDa protein, consistent with its predicted size. Like protein C16 (Fahy et al., 2008), the C4 protein was detected from 2 h p.i. onwards and in the presence of AraC, indicating early expression. In contrast, AraC inhibited expression of D8, a virion structural protein expressed late during infection (Niles & Seto, 1988). Immunoblotting with anti-a-tubulin confirmed equivalent loading of samples.
The subcellular localization of C4 was investigated by infecting HeLa cells with vC4-TAP and then either examining fixed cells by immunofluorescence (Fig. 2a) or performing immunoblotting of biochemically fractionated cellular lysates (Fig. 2b). Consistent with temporal expression data, immunofluorescence detected the C4 protein from 2 h p.i. in the cytoplasm, but C4 subsequently became increasingly nuclear. By 6 h p.i., C4 was detectable within both cytoplasm and nucleus, but by 24 h it was predominantly nuclear. Immunoblotting of fractionated cells corroborated data obtained by immunofluorescence and showed that C16 had a distribution similar to that of C4 (Fig. 2b). Successful separation of cytoplasmic and nuclear compartments was confirmed by the presence of atubulin and lamin A/C in only the cytoplasmic and nuclear fractions, respectively.

C4 does not affect virus replication in cell culture
The fact that not all VACV strains contain gene C4L indicated that it was non-essential for virus replication, and this was confirmed for VACV strain WR by the isolation of the C4 deletion mutant, vDC4. To ascertain whether C4 affected virus replication or spread, the size of plaque formed by vDC4 was compared with that of control viruses in RK-13 and BSC-1 cells; no significant differences were observed (Fig. 3a). Next, the replication of vDC4 in BSC-1 cells was investigated after infection at low (0.01) or high (10) m.o.i. and viruses in the intra-and extracellular fractions were titrated by plaque assay. Again, no differences were observed between vDC4 and control viruses (Figs 3b, c and S2). Collectively, these data indicate that C4 is nonessential for virus replication and spread.

C4 inhibits NF-kB activation
Given that C4 was intracellular, its proposed possible function as an extracellular IL-1ra-like protein seemed improbable. Therefore, we investigated whether C4 inhibited intracellular signalling pathways, using a reporter plasmid with the IFN-b promoter driving expression of firefly luciferase. This was transfected into HEK293T cells that were stimulated subsequently by transfection with poly(dA : dT), a ligand for intracellular DNA sensors, or poly(I : C), a ligand of retinoic acid-inducible gene (RIG)-Ilike receptors. These stimuli each induced luciferase activity, which was inhibited by C4 but not by a GFP control ( Fig. 4a, b). Inhibition was also achieved by VACV protein B14, which inhibits NF-kB activation , and by C6, which inhibits interferon regulatory factor (IRF) 3 activation (Unterholzner et al., 2011), because the IFN-b promoter contains binding sites for both IRF3 and NF-kB.
To dissect the mechanism of action of C4, its ability to inhibit activation of IRF3, NF-kB and activator protein 1 (AP-1) was investigated with reporter plasmids specific for these transcription factors. To measure IRF3 activity, an ISG56.1-dependent promoter was stimulated by the transfection of poly(dA : dT) and poly(I : C). ISG56.1 promoter activity was inhibited by C6, a recently characterized antagonist of IRF3 activation (Unterholzner et al., 2011), but not by C4 or B14 (Fig. 4c, d). Next, the effect of C4 on NF-kB-specific promoter activity was assessed using an NF-kB-dependent promoter (Fig. 4e, f). Stimulation of HEK293T cells with IL-1b and TNF-a resulted in a 70-and 18-fold induction of NF-kB promoter activity, respectively, and both were inhibited by C4. As expected B14, but not C6, was able to inhibit NF-kB activation Unterholzner et al., 2011). Satisfactory stimulation of an AP-1 reporter plasmid was not achieved under several conditions tried and so the effect of C4 on this transcription factor could not be determined. Collectively, these data indicate that C4 inhibits IFN-b promoter activity by inhibiting NF-kB, but not IRF3, activation.
The mechanism by which C4 inhibited NF-kB activation was investigated by overexpression of proteins acting at different stages in the signalling cascade. Expression of TRAF2, which acts downstream of the TNF receptor, or TRAF6, which acts downstream of the IL-1 receptor, both stimulated NF-kB expression and both pathways were inhibited by C4 and B14 (Fig. 5a, b). C4 was therefore inhibiting both IL-1-and TNF-mediated signalling and so was likely to act at, or downstream of, the position where these pathways converge. To test this hypothesis, NF-kB was activated by overexpression of IKKb, part of the IKK complex (Fig. 5c); once again both C4 and B14, but not C6, inhibited this activation. Taken together, these data demonstrate that C4 inhibits the activation of NF-kB at, or downstream of, the IKK complex. This was confirmed by measuring the translocation of p65 into the nucleus in response to IL-1b stimulation (Fig. 6). In resting cells p65 was cytoplasmic but, in response to IL-1b, it became nuclear and this was inhibited by C4 expression (Fig. 6a, b).  Fig. 4. C4 inhibits NF-kB activation. HEK293T cells were transfected with a firefly luciferase reporter plasmid under the control of an (a, b) IFN-b-, (c, d) ISG56.1-or (e, f) NF-kB-dependent promoter, a Renilla luciferase transfection control, and a C4, B14, C6 or GFP expression plasmid or empty vector control (EV). Cells were (a-d) transfected 24 h later with (a, c) poly(dA : dT) or (b, d) poly(I : C) at 800 ng per well for 24 h or (e, f) stimulated directly with (e) IL-1b or (f) TNF-a at 40 ng ml "1 for 8 h and then harvested. Firefly luciferase activity was normalized to Renilla luciferase activity. Data are from one experiment representative of at least three, each performed in triplicate and presented as means±SD. *P,0.05; **P,0.005; ***P,0.0005, compared with EV.

C4 affects virus virulence and cell recruitment
As C4 inhibited NF-kB activation, it was possible that C4 might affect virus virulence, and this was tested utilizing intradermal and intranasal murine models of VACV infection (Williamson et al., 1990;Tscharke & Smith, 1999). In the intradermal model, there was no difference in lesion size in mice infected with vDC4 compared with control viruses (Fig. S3). However in the intranasal model, mice infected with vDC4 lost less weight, exhibited fewer signs of illness and recovered faster than mice infected with control viruses (Fig. 7a, b). These differences were statistically significant (P,0.05) from days 6 and 5 p.i., respectively. Measurement of virus titres in lungs at day 2 p.i. showed that all viruses had replicated to equivalent levels, demonstrating that, as in vitro, the loss of C4 did not impair virus replication in vivo. However, by days 5 and 8 p.i., vDC4-infected mice had significantly lower virus titres than controls, suggesting more rapid clearance of virus in the absence of C4 (Fig. 8a). To investigate the basis for this phenotype, the number of viable cells in bronchoalveolar lavage (BAL) fluids was investigated on days 2, 5 and 8 p.i. (Fig. 8b). This revealed that, as early as day 2 p.i., there was an increase in the number of cells in BALs from vDC4-infected mice, although this difference was not statistically significant. However, by day 5 a larger and statistically significant difference was seen. By day 8 p.i., there were fewer cells in the BAL fluid of mice  infected with vDC4 compared with controls. The recruitment of cells into BALs therefore showed quantitative and kinetic differences following infection by vDC4, with earlier recruitment and higher numbers of cells, and more rapid resolution. Note that, at day 8 p.i., the number of cells present in BALs of mice infected with control viruses was equivalent to that at day 5, whereas in vDC4-infected animals, the cell number had already declined from that at day 5, indicating recovery from infection. The more rapid and greater recruitment of cells into BAL fluids of vDC4infected mice correlated well with the reduced virus titres at days 5 and 8 p.i. C4 is therefore both an immunomodulator and a virulence factor.

DISCUSSION
Characterization of the C4 protein from VACV strain WR is reported. Data presented show that C4 is an intracellular, 37 kDa protein that is expressed early after infection and is non-essential for virus replication. Functional assays demonstrated C4 to inhibit NF-kB activation, acting at or downstream of the IKK complex, and C4 was shown to be a virulence factor in a murine intranasal model of infection.
The C4L gene is conserved (95-99 % amino acid identity) in six of eight OPV species and seven of 15 sequenced VACV strains. It is also related closely to VACV protein C16 and a family of OPV proteins (Fahy et al., 2008). It was noteworthy that C4 is encoded by the three OPVs that do not encode C16 (CMLV, TATV and MPXV). The close relatedness of C4 and C16 (43 % amino acid identity) and the fact that C4 is present in those OPVs lacking C16 suggest that these proteins have an important role in the virus life cycle and may have partially overlapping functions. However, it is clear that C4 and C16 are not redundant because removal of either gene causes an alteration in virus virulence, as shown in this study and by Fahy et al. (2008).
The detection of C4 early during infection is consistent with the results of transcriptional analysis which reported that C4 mRNA was detectable within 30 min p.i. (Assarsson et al., 2008). The expression of C4 also parallels that of C16, which is also made early during infection  (Fahy et al., 2008), and this timing is consistent with the immunomodulatory function of these proteins. VACV proteins that inhibit the innate immune response to infection are generally expressed rapidly after infection to prevent activation of signalling cascades leading to production of IFNs and pro-inflammatory cytokines and chemokines. One exception to this principle is the case of the VACV IL-1b-binding protein encoded by VACV strain WR gene B15R, which is expressed late during infection and controls the body temperature of infected animals (Smith & Chan, 1991;Alcamí & Smith, 1992, 1996. The localization of C4 during infection is unusual and shows a predominant cytoplasmic expression early (2 h) p.i., but this changes with time and by 6 h it is in both nuclear and cytoplasmic fractions, and by 24 h is predominantly nuclear (Fig. 2). Similarly, the related protein C16 is also predominantly cytoplasmic early after infection and accumulates in the nucleus as infection progresses (Fahy et al., 2008). C4 does not contain an identifiable nuclearlocalization sequence, suggesting that it might bind to another viral or cellular protein to mediate this translocation, or that it can move between cytoplasmic and nuclear compartments by passive diffusion (which is possible for proteins ¡50 kDa). However, the strong nuclear predominance later during infection argues against simple diffusion for this localization. Proteins C6 and E3 are other examples of VACV immunomodulatory proteins that have a nucleocytoplasmic localization during the virus life cycle (Yuwen et al., 1993;Unterholzner et al., 2011).
The absence of C4 in several VACV strains and in ECTV indicated that the protein was non-essential for OPV replication and this was consistent with isolation of a VACV strain WR deletion mutant vDC4. Despite this, a minor reduction in plaque size was noted, which, although not statistically significant with the sample size studied, was observed consistently in multiple cell types. Similarly, vDC16 also produced a smaller plaque, but in that instance, the difference was large enough to be statistically significant (Fahy et al., 2008). In the future, it will be interesting to investigate the plaque size of a virus lacking both genes. An analysis of virus replication kinetics in vitro and in vivo showed no difference between vDC4 and control viruses.
Functional screens revealed C4 to be an inhibitor of IFN-b promoter activation in response to poly(dA : dT) and poly(I : C). Mechanistically, this was attributable to inhibition of activation of NF-kB rather than IRF3. C4 is therefore distinct from VACV protein C6, which inhibits IRF3 activation by binding to the adaptor proteins TANK, SINTBAD and NAP1 (Unterholzner et al., 2011), but similar to B14, which inhibits NF-kB activation by binding to IKKb Graham et al., 2008;Benfield et al., 2011). The position in the pathway leading to NF-kB activation at which C4 functions was deduced to be at, or downstream of, the IKK complex, because C4 inhibited activation mediated by overexpression of TRAF2, TRAF6 and IKKb. This was consistent with its ability to block NF-kB activation by both IL-1b and TNF-a. C4 is therefore comparable to B14, which binds to IKKb and prevents the phosphorylation of the IKKb activation loop McCoy et al., 2010;Benfield et al., 2011).
Given the similar function of C4 and B14 and the fact that C4 represents the ninth VACV protein to inhibit NF-kB (see Introduction), it was surprising that loss of C4 caused an attenuated phenotype (Fig. 7). Furthermore, in all cases where the loss of an individual NF-kB inhibitor has been studied in vivo -A46 (Stack et al., 2005), A52 (Harte et al., 2003), B14 , N1 (Bartlett et al., 2002) and C4 (this paper) -there is also an attenuated phenotype, despite the presence of other inhibitors. These observations indicate that these proteins are non-redundant and that this non-redundancy might be explained in different ways. First, the proteins may have multiple functions, as has been demonstrated for N1, which is an inhibitor of both apoptosis and NF-kB activation (DiPerna et al., 2004;Cooray et al., 2007;Maluquer de Motes et al., 2011). Second, the position in the NF-kB activation pathway at which an inhibitor functions could affect outcome. For instance, if an inhibitor blocked only IL-1b or TNF-a-induced NF-kB activation, the influence in vivo could be different from that of an inhibitor that acts on both pathways. Furthermore, crosstalk between NF-kB and other signalling pathways could also allow an inhibitor to influence outcomes in vivo differently, depending on where it acts.
A further surprise is that losses of C4 or B14 give different in vivo phenotypes, despite blocking the same pathway. With vDB14, the attenuation was evident in the intradermal but not the intranasal model , whereas with vDC4, the opposite is observed. C4 represents another example where a phenotype is seen in only one model and illustrates the value of using both models, as noted previously . Mechanistically, a propensity for an immunomodulator to show a phenotype in only one in vivo model has been attributed to its mechanism of action and the different inflammatory responses elicited by infection via either route . In this case, given that B14 and C4 target the same pathway and at a similar position, the different in vivo phenotype is more surprising and suggests, perhaps, that one or both of these proteins have additional functions.
In summary, VACV protein C4 is an intracellular inhibitor of NF-kB and promotes virus virulence. C4 represents the ninth VACV inhibitor of NF-kB, and the fact that VACV has evolved so many inhibitors of this pathway highlights the importance of NF-kB signalling for the host response to virus infection. A notable feature of these inhibitors is that, in all cases where the virulence of viruses lacking individual proteins has been studied, reduced virulence is observed, despite the presence of the other inhibitors. This indicates additional complexity of these proteins and signalling pathways that is worthy of further investigation and may enhance our understanding of virus pathogenesis and innate immunity.

METHODS
Cell culture. BSC-1, CV-1 and HEK293T cells were grown in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10 % FBS (Harlan Laboratories) that was heated at 56 uC for 1 h to inactivate complement, and penicillin/streptomycin (50 mg ml 21 ; Gibco). HeLa and RK-13 cells were grown in minimum essential medium (Gibco) supplemented as above and with the addition of non-essential amino acids (Sigma). All cell lines were maintained at 37 uC in a 5 % CO 2 atmosphere.
Construction of plasmids. For construction of a VACV lacking the C4L gene, 300-350 bp flanking regions of the C4L gene were amplified by overlapping PCR from VACV WR genomic DNA, and then cloned into the Z11 mammalian expression vector containing Escherichia coli guanine phosphoribosyltransferase (Ecogpt) and enhanced green fluorescent protein (EGFP) genes, generating plasmid Z11DC4, as described previously for another VACV gene (Unterholzner et al., 2011). The 59 DNA fragment was generated with oligonucleotides 59-GCCACGCGTTCCAATTATCTTTACCG-39 (L-FA) containing an MluI restriction site (underlined) and 59-CCCGGGTCCGGAGAG-CTCTAGTATTGGTTAAAAATGAAAATGG-39 containing nucleotides from the 39 fragment (italics) at the 59 end. The 39 DNA fragment was generated with oligonucleotides 59-GAGCTCTCCGGACCCGGGTTT-ATATCACTACGG-39 containing complementary sequence to the 59 fragment (italics) and 59-TAGCGGCCGCGGTGTCCTGTGTTCAGG-39 (L-FB) containing a NotI restriction site (underlined). Left and right amplicons were joined by PCR using the L-FA and L-FB oligonucleotides to generate an FA-FB fragment. For construction of a C4 revertant virus in which the C4L gene was reinserted into its natural locus within the C4 deletion mutant, C4L together with the C4L flanking regions was amplified from VACV WR genomic DNA using oligonucleotides L-FA and L-FB and then cloned into Z11, generating plasmid Z11C4Rev.
The sequence of C4L was codon-optimized for expression in mammalian cells (GeneArt, Invitrogen Life Technologies). C4L was then subcloned into the mammalian expression vector pcDNA4.0/TO (Invitrogen) containing a C-terminal tandem affinity purification (TAP) tag consisting of two FLAG and two StrepII epitopes (Gloeckner et al., 2007), generating plasmid C4-TAP. For construction of the TAPtagged C4 revertant virus (vC4-TAPRev), C4L and the C4L left flanking region were cloned into the pcDNA4.0/TO vector and the C4L right flanking region was then cloned downstream of the TAP tag. The complete C4L locus and TAP tag was subsequently subcloned into the pUC13 mammalian expression vector modified to contain the Ecogpt and EGFP genes (Maluquer de Motes et al., 2011), generating the plasmid pUC13-C4TAPRev. The fidelity of all plasmids constructed using PCR was confirmed by DNA sequencing.
Construction of recombinant VACVs. C4 recombinant viruses were constructed by transfecting plasmid Z11DC4 into VACV-WRinfected CV-1 cells (to make vDC4) or transfecting Z11C4Rev or pUC13-C4TAPRev into vDC4-infected cells (to make vC4-Rev and vC4-TAPRev, respectively). Recombinant VACVs were subsequently isolated by transient dominant selection (Falkner & Moss, 1990) as described for the construction of other VACV deletion mutants Fahy et al., 2008), using three cycles of plaque purification. The genotype of viruses was determined by C4L locus-specific PCR, and a plaque-purified wild-type virus (vC4) and a C4L deletion mutant virus (vDC4) were identified and amplified. A revertant virus expressing C4 and a recombinant virus expressing C4 with a C-terminal TAP tag were isolated from vDC4 by the same methodology. The genomes of all viruses were confirmed to have the predicted structure by PCR and restriction endonuclease digestion of genomic DNA.
Immunoblotting. BSC-1 cells were infected with the indicated viruses and cell lysates prepared as described previously (Bartlett et al., 2002). Antibodies were from the following sources: mouse anti-FLAG (Sigma), mouse anti-a-tubulin (Upstate Biotech) and mouse anti-lamin A/C (Abcam). The rabbit C16 polyclonal antiserum and mouse anti-D8 mAb AB1.1 were described previously (Parkinson & Smith, 1994;Fahy et al., 2008). Secondary antibodies and protein detection were performed as described previously (Bartlett et al., 2002).
Cell fractionation. BSC-1 cells were infected with vC4-TAP at 10 p.f.u. per cell for the indicated times and processed as described previously (Fahy et al., 2008). Proteins were resolved by SDS-PAGE and detected by immunoblotting.
Plaque-size assay. Monolayers of BSC-1 or RK-13 cells were infected in duplicate at 50 p.f.u. per well and incubated for 72 h in medium containing 1.5 % carboxymethylcellulose. The cells were washed once with PBS and then stained with crystal violet. The diameter of 30 plaques per virus was measured using Axiovision 4.6 on a Zeiss Axiovert 200 M microscope (Zeiss).
Virus growth curves. Intracellular and extracellular virus production after high or low m.o.i. was determined as described previously  and the clarified culture supernatant was used for extracellular virus.
Reporter-gene assays. Luciferase reporter-gene assays were performed by seeding HEK293T cells at a density of 10 000 cells per well before transfection with polyethyleneimine (PEI; Polysciences). Transfection reactions contained 10 ng GL3-Renilla control plasmid per well, 60 ng IFN-b-, AP-1-or NF-kB-firefly reporter plasmid per well (gifts from A. Bowie, Trinity College Dublin, Ireland, and G. Sen, Lerner Research Institute, Cleveland, OH, USA) and 70 ng expression vector or pcDNA4.0/TO empty vector control per well. After 24 h incubation, transfected cells were stimulated with 40 ng IL-1b or TNF-a ml 21 (Peprotech) for 8 h, or transfected with 800 ng poly(dA : dT) or poly(I : C) ml 21 (Sigma) for 24 h. The cells were harvested in Passive Lysis Buffer (Promega) and data were analysed using MARS data analysis software on a FLUOstar Omega instrument (BMG Labtech). For data analysis, firefly luciferase activity values were normalized against Renilla luciferase activity values. Experiments were performed in biological triplicate and repeated a minimum of three times. Data are expressed as means±SD.
Analysis of cell populations. Mice were sacrificed and the BAL fluid and lung tissue were prepared and processed as described previously (Clark et al., 2006;Jacobs et al., 2006). Viable cells in BAL fluid were stained with 0.4 % trypan blue (Invitrogen) and counted.