Cellular transcriptional profiling in influenza A virus-infected lung epithelial cells: The role of the nonstructural NS1 protein in the evasion of the host innate defense and its potential contribution to pandemic influenza

Wild-type influenza A/PR/8/34, delNS1, and NS1 (1–126) viruses were generated, propagated in 7-day-old eggs, and titrated by plaque assay as described (9, 13, 14). Influenza A/WSN/33 virus and 1918 NS WSN recombinant virus were propagated and maintained as described (15). The lung epithelial cell line A549 was purchased from American Type Culture Collection (ATCC, Manassas, VA) and maintained in DMEM containing 10% FBS, 2 mM glutamine, 100 units/ml penicillin, and 100 units/ml streptomycin at 37°C.

Infections with wt and recombinant influenza A viruses were performed in A549 cells grown in 10-cm² dishes or T-150 flasks. In all cases, the multiplicity of infection (moi) was adjusted before each independent experiment to ensure that at least 80% of the cells were infected 8 h postinfection as determined by immunofluorescence using a viral nucleoprotein (NP)-specific antibody. Total RNA extraction and mRNA selection procedure is described (ref. 16; http://ra.microslu.washington.edu/Website/protocol/archive_protocol.html).

EMSAs were done to determine the activation of NF-κB as described (9), by using nuclear extracts from infected A549 cells and a DNA probe containing a mouse H2 κB binding site. Northern blot analysis of total RNA isolated from A549-infected cells using specific [³²P]ATP-labeled probes was performed as previously described (9).

Microarrays were constructed by the University of Washington's Center for Expression Array Technology with PCR products generated from the 15,000 set of sequence-verified IMAGE consortium clones (http://ra.microslu.washington.edu/Website/genelist/genelist.html). For microarray conditions, see supplemental text, which is published as supporting information on the PNAS web site, www.pnas.org. In accordance with proposed MIAME (minimum information about a microarray experiment) standards (17), raw data, including sample information, intensity measurements, error analysis, microarray content, and slide hybridization conditions will be made available in the public domain (http://expression.microslu.washington.edu).

Recent evidence in cell lines and in mice suggests that the influenza A NS1 protein antagonizes the cellular IFN response (13) in part by blocking NF-κB, IRF3, and IRF7 activation (9–11). We examined whether the same was true in A549 cells, a human lung epithelial cell line that may more accurately represent the infected cells found at the physiological sites of influenza virus infection. Both NF-κB activity (Fig. 1A) and IFN-β mRNA levels (Fig. 1B) increased dramatically by 9 h postinfection with a recombinant influenza A virus lacking the NS1 gene (delNS1 virus). In contrast, the activation of NF-κB and production of IFN-β mRNA were considerably lower in cells infected with wt PR8 virus. We confirmed that expression of IFN and NF-κB regulated genes were induced in delNS1 virus-infected cells by examining a selected set of IFN and/or NF-κB stimulated genes by Northern blot (Fig. 1B). The results demonstrate that the NS1 protein of influenza A virus down-regulates the IFN response in a human lung epithelial cell system and support the hypothesis that this viral protein has an inhibitory effect on IFN and NF-κB pathways.

We next examined global cellular gene expression levels in cells infected with viruses containing mutations in the NS1 gene and compared them with those in cells infected with the parental wt PR8 virus. Two recombinant viruses were tested, delNS1 virus (13), and NS1 (1–126) virus lacking the C-terminal 104 aa of the NS1 protein (9). A549 lung epithelial cells in monolayers were infected at an moi resulting in approximately 80% of cell infection (Fig. 6, which is published as supporting information on the PNAS web site, www.pnas.org.). We chose this moi to infect most cells with 1–2 virus particles. At 8 h postinfection, total RNA was extracted to be analyzed by microarray. Differentially expressed genes were selected based on ratio and statistical criteria from combined replica experiments. Because approximately 20% of cells are not infected, down-regulated genes in the array represent substantially down-regulated genes in infected cells or genes that are also down-regulated in noninfected neighboring cells. PR8 wt virus infection perturbed the expression of the fewest number of cellular genes (84 genes) whereas delNS1 virus infection resulted in the largest regulation of cellular genes (115 genes). The NS1 (1–126) virus regulated an intermediate number of cellular genes (93 genes). Thus, influenza A virus infection resulted in ≈0.5–1% of all of the cellular genes on the array being differentially regulated. A summary of the number of genes that are significantly regulated by each virus and the degree of overlap between gene sets is shown [Fig. 2; and see Table 1 (which is published as supporting information on the PNAS web site) for data on individual experiments]. Genes that were regulated by each virus can be classified in three primary groups: genes whose expression was similar between wt and mutant NS1 virus infections (Fig. 2, solid black and diagonal hatched), genes whose expression was similar between mutant NS1 viruses but not wt virus infections (Fig. 2, horizontal hatched), and genes whose expression is more significantly regulated by the individual viruses alone (Fig. 2, solid gray).

To confirm that induction of NF-κB and IFN-β regulated genes could be detected by microarray, we examined the mRNA levels for the cellular genes tested by Northern blot. The NFKBIA, IL-8, TNFRSF6, and HLA-C genes were all induced at 8 h postinfection with delNS1 virus but not by wt PR8 virus (Fig. 1B and Fig. 7, which is published as supporting information on the PNAS web site). In contrast, the levels of actin-α were unaffected (actin-β not present). It is important to stress that the Cy3/Cy5 ratio obtained by repeated microarray experiments underestimates the relative change in mRNA expression. Thus, a 2-fold change in expression by microarray is most likely a conservative cutoff for a differentially regulated gene, because the levels of the same mRNA measured by Northern blot appear to be more dramatically changed. The discrepancy between Northern and microarray results has previously been observed by using the same array platform (18).

There were 24 genes that were induced more than 2-fold with greater than 99% confidence (P ≤ 0.01) during infection with both wt and mutant PR8 viruses (Fig. 2, solid black, and Fig. 3A). There were 10 more genes whose expression was coregulated by all 3 viruses but with reduced confidence levels (P value > 0.01) in one set of experiments (Fig. 2, diagonal hatched, and Fig. 3A). Sixteen genes in this group were previously known to be stimulated by IFN and 4 more are potential antiviral genes. The IFN-stimulated genes (ISGs) up-regulated by viral infection include the influenza virus inhibitory IFN-induced gene MX1/MXA, in addition to IFI75, IFI41, IRF1, ISGF3-γ/IRF9 and others (Fig. 3A, asterisk). Note that, even though these genes were differentially regulated during infection with both wt and mutant viruses, some genes (e.g., GBP and IRF1) were induced to higher levels in delNS1 virus-infected cells (Table 2, which is published as supporting information on the PNAS web site). Other potential antiviral genes of interest that were consistently induced by PR8 virus infection include TNF-related apoptosis inducing ligand (TRAIL/TNFSF10), complement C1S component, and MYD88. The latter is an adapter molecule that participates in Toll-like receptor signaling (19).

To determine whether the sequence of differentially regulated expressed sequence tags (ESTs) aligned with any known or predicted gene, a blast search of the human genome database was performed on the available sequences for each uncharacterized clone. Three ESTs aligned with regions containing known or predicted genes with potential antiviral activity (Fig. 9, which is published as supporting information on the PNAS web site). These ESTs align with an exon of cig5 similar to inflammatory response protein 6 (IMAGE ID 120600), an exon of LOC129607 which shares similarity with thymidylate kinase (IMAGE ID 207838), and the 3′ noncoding region of LOC143274 that is similar to IFIT2/ISG54 (IMAGE ID 289496).

We would predict from biochemical analysis in cell culture (Fig. 1B and refs. 9 and 10) and phenotypic analysis in mice (13, 14) that the mutant NS1 viruses would induce a more pronounced antiviral response involving genes regulated by IFN and/or NF-κB. Indeed, delNS1 and NS1 (1–126) viruses induce a generally higher increase in ISGs and in NF-κB mediated gene expression than wt virus [Fig. 3A, Fig. 10 (which is published as supporting information on the PNAS web site), and Table 2]. In addition, both mutant viruses induce the expression of an additional 24 genes (Fig. 3B), 5 of which were known to be associated with the IFN pathway (including 2 ESTs revealed by blast; see Fig. 9) and 6 were potential antiviral genes based on available annotation. Furthermore, if one also considers the ISGs that are not exclusively coregulated by both mutants but enhanced during either NS1 (1–126) (1 ISG and 7 genes with potential antiviral activity, Fig. 3C) or delNS1 virus (3 ISGs and 5 genes with potential antiviral activity, Fig. 3D) infection, it is evident that the antiviral response, including IFN and NF-κB regulated genes, is more robust in the absence of the NS1 protein (Fig. 3 and Table 2). Interestingly, two genes that are induced to high levels in delNS1 virus-infected cells (PMAIP1 and GBP1) also appear to be induced by overexpression of a constitutively active form of IRF3 (20). The higher activation of IRF3 observed during infection with delNS1 virus as compared with wt PR8 virus (10) may explain these results.

Collectively, the mutant NS1 viruses appear to have a significant impact on genes in the signal transducer and activator of transcription (STAT)-signaling pathway. For instance, the expression levels of STAT1 (Fig. 3D) and STAT3 (Table 2) were greater in cells infected with NS1 mutant viruses relative to wt A/PR/8/34 virus. In addition, the expression of two members of the suppressor of cytokine signaling (SOCS) family of proteins, STATI2/SOCS2 and SSI-3/SOCS3, whose transcription is regulated by STAT proteins (21–23), was affected in cells infected with delNS1 virus relative to wt (Table 2 and Fig. 4A). These proteins function to modulate cytokine and growth factor signaling in a negative feedback loop. Interestingly, SOCS2/STATI2 is down-regulated in delNS1 virus-infected cells (Fig. 3D), whereas SSI-3/SOCS3 is up-regulated (Fig. 3B), suggesting that the two genes may have opposing functions in modulating cytokine response, at least during influenza virus infection. The expression of another SOCS family member (SOCS1/SSI-1, Fig. 4A) and of the N-Myc interactor (NMI) (Table 2), which has also been shown to interact with STAT proteins (24), was also increased by PR8 virus infection. We examined expression of all SOCS-box-containing proteins, along with its associated cytokines and receptors that were present on the microarray (Fig. 4A). We found that, in addition to the STAT and SOCS genes discussed above, the leukemia inhibitory factor (LIF) and its receptor (LIFR), whose signaling is modulated by SOCS proteins (23, 25), were induced in PR8 virus-infected cells. The SOCS genes modulate cytokine response via different mechanisms. SOCS1/SSI-1 binds directly to Janus kinases (JAKs) and inhibits their catalytic activity (26), SOCS2/SSI-2/CIS binds to the cytoplasmic domain of cytokine receptors (27), and SOCS3/SSI-3 functions via binding to receptors in which JAKs are already bound (28).

Other intriguing evidence that the STAT pathway may be regulated during influenza virus infection is an increase in dual-specificity phosphatases, DUSP4 and DUSP5, mRNAs during delNS1 virus infection (Fig. 3D). DUSP6 mRNA was also increased but its P value was 0.02 (Fig. 11, which is published as supporting information on the PNAS web site). Although the human DUSP genes, which are homologous to vaccinia virus H1 phosphatase (VH1), are known to primarily act on mitogen-activated protein (MAP) kinases [such as extracellular signal-regulated kinase 1 (ERK1)], the viral homolog VH1 has been shown to dephosphorylate and inhibit STAT proteins (29). The mRNA for several members of the DUSP gene family (e.g., DUSP1, -2, and -5) were also found to be induced in other large-scale gene expression studies of both bacterial (30) and viral infections, including influenza virus infection of dendritic cells (31). Another gene critical to IFN signaling that may also be involved in STAT signaling is the IFN-inducible gene PKR. Levels of PKR mRNA increased in cells infected with NS1 mutants relative to wt virus. We presume that increased levels of type I IFN (Fig. 1B) in the absence of NS1 result in increased PKR mRNA levels. Finally, PKR itself has been suggested to be involved in the serine phosphorylation of STAT1 (32) and therefore may also contribute to the overall regulation of STAT signaling.

Proteins containing tripartite motifs are a growing family of known and predicted proteins that have been proposed to function by forming specific subcellular compartments, for example, promyelocytic leukemia (PML) nuclear bodies (33). These genes are predicted to contain, a RING finger (3 zinc binding domains), a B-box (type 1 and/or 2), and a coiled-coil region. Tripartite motif (TRIM) genes regulated during influenza virus infections are shown in Fig. 4B (see also Fig. 12, which is published as supporting information on the PNAS web site). We propose that this set of TRIM genes may represent a new class of antiviral proteins that are coordinately regulated.

To further explore the potential contribution of the NS1 protein in influenza virus virulence, we used a recombinant WSN virus containing the NS gene from the 1918 pandemic strain (15). Because this recombinant virus was generated in the influenza A/WSN/33 virus strain background, it was also necessary to measure the effect of WSN virus infection on host cell gene expression in A549 lung cells so that a direct comparison could be made. Introduction of the 1918 NS sequence into the WSN background resulted in an overall expression pattern that was very similar to the parental WSN virus (Fig. 5). However, differences were observed for specific ISGs. Although both WSN wt and 1918 NS recombinant viruses appeared to induce lower levels of expression of IFN-regulated genes than the PR8 wt virus (Fig. 13, which is published as supporting information on the PNAS web site), the wt WSN virus did induce expression of some ISGs to high levels [MXA/MX1, ISG15, IFITM1, and ISGF3-g (IRF9)]. In contrast, infection with 1918 NS WSN virus failed to significantly induce any ISG (Figs. 5 and 13). We also noted that NS 1918 virus (and to a lesser extent wt WSN virus) had the opposite effect of PR8 virus infection on the expression of SSI-1/SOCS1, STAT1, and leukemia inhibitory factor (Fig. 4A), as well as of some NF-κB-regulated genes (e.g., SAA1 and BIRC2 and -3, Fig. 10). The compare plot in Fig. 5 also reveals that several cellular genes, in addition to ISGs, are differentially expressed between cells infected with wt or 1918 NS WSN viruses. For example, protein kinase inhibitor-β (PKIB) and Toll-like receptor 5 (TLR5) were up- and down-regulated, respectively, by 1918 NS virus but not by wt WSN. Conversely, CSPG6 is down-regulated during infection with wt WSN virus but not by 1918 NS WSN. We also found several ESTs that exhibit different expression levels between the two viruses. These results emphasize the advantage of combining recombinant virus and microarray technology to identify a few candidate genes for future study. Whether these differences in gene expression are partially responsible for the high virulence associated with the 1918 virus remains to be investigated.