Phosphoproteome Analysis of Cells Infected with Adapted and Nonadapted Influenza A Virus Reveals Novel Pro- and Antiviral Signaling Networks

Infection with IAVs leads to the induction of complex signaling cascades, which apparently serve two opposing functions. On the one hand, the virus highjacks cellular signaling cascades in order to support its propagation; on the other hand, the host cell triggers antiviral signaling networks. Here we focused on IAV-triggered phosphorylation events in a systematic fashion by deep sequencing of the phosphoproteomes. This study revealed a plethora of newly phosphorylated proteins. We also identified 37 protein kinases and a range of phosphatases that are activated or inactivated following IAV infection. Moreover, we identified new phosphorylation sites on IAV-encoded proteins. Some of these phosphorylations support the enzymatic function of viral components, while other phosphorylations are inhibitory, as exemplified by PB1 Thr223 modification. Our global characterization of IAV-triggered patterns of phospho-proteins provides a rich resource to further understand host responses to infection at the level of phosphorylation-dependent signaling networks.

ABSTRACT

Influenza A viruses (IAVs) quickly adapt to new environments and are well known to cross species barriers. To reveal a molecular basis for these phenomena, we compared the Ser/Thr and Tyr phosphoproteomes of murine lung epithelial cells early and late after infection with mouse-adapted SC35M virus or its nonadapted SC35 counterpart. With this analysis we identified a large set of upregulated Ser/Thr phosphorylations common to both viral genotypes, while Tyr phosphorylations showed little overlap. Most of the proteins undergoing massive changes of phosphorylation in response to both viruses regulate chromatin structure, RNA metabolism, and cell adhesion, including a focal adhesion kinase (FAK)-regulated network mediating the regulation of actin dynamics. IAV also affected phosphorylation of activation loops of 37 protein kinases, including FAK and several phosphatases, many of which were not previously implicated in influenza virus infection. Inhibition of FAK proved its contribution to IAV infection. Novel phosphorylation sites were found on IAV-encoded proteins, and the functional analysis of selected phosphorylation sites showed that they either support (NA Ser178) or inhibit (PB1 Thr223) virus propagation. Together, these data allow novel insights into IAV-triggered regulatory phosphorylation circuits and signaling networks.

IMPORTANCE Infection with IAVs leads to the induction of complex signaling cascades, which apparently serve two opposing functions. On the one hand, the virus highjacks cellular signaling cascades in order to support its propagation; on the other hand, the host cell triggers antiviral signaling networks. Here we focused on IAV-triggered phosphorylation events in a systematic fashion by deep sequencing of the phosphoproteomes. This study revealed a plethora of newly phosphorylated proteins. We also identified 37 protein kinases and a range of phosphatases that are activated or inactivated following IAV infection. Moreover, we identified new phosphorylation sites on IAV-encoded proteins. Some of these phosphorylations support the enzymatic function of viral components, while other phosphorylations are inhibitory, as exemplified by PB1 Thr223 modification. Our global characterization of IAV-triggered patterns of phospho-proteins provides a rich resource to further understand host responses to infection at the level of phosphorylation-dependent signaling networks.

INTRODUCTION

Influenza A viruses (IAVs) are still a major threat to human and animal health since they can cause seasonal epidemics or occasional pandemics, which can lead to severe and sometimes even lethal pneumonia (1, 2). Wild aquatic birds comprise the natural reservoir for IAVs, which have an RNA genome of eight segments with negative polarity that encode at least 10 viral proteins. The viral RNA (vRNA), together with the three subunits of the heterotrimeric, RNA-dependent RNA polymerase (RdRp) complex (PB1, PB2, and PA) and the nucleoprotein (NP), form active ribonucleoprotein complexes (vRNPs) that mediate viral replication and transcription (1, 3). After infection, the vRNPs enter the nucleus, where viral genome replication and transcription take place. The vRNA is used as a template for transcription to generate viral mRNAs and complementary RNAs (cRNAs), which serve as the templates for vRNA synthesis mediated by the RdRp (4). After completion of viral replication, newly synthesized vRNPs are exported from the nucleus to the cytoplasm, where they reach the viral assembly sites at the plasma membrane and bud to the extracellular lumen (5, 6). Various mechanisms, including the lack of proofreading function of the viral RdRp, cause constant accumulation of mutations. Further genetic diversity can be generated during infection of a cell with different IAVs and via reassortment of the segmented viral genomes (7–9). Therefore, the genetic flexibility of IAVs allows them to not only escape selection pressures such as the immune response but also gain the capacity to infect new host species (10). One well-studied model system for host-specific adaptation of an avian IAV in a mammalian host is the H7N7-type chicken cell-adapted IAV (SC35), which was originally derived from the seal isolate A/Seal/Mass/1/80 (H7N7) (11). This virus was then adapted to mice (SC35M) due to changes at only nine amino acid positions in six viral proteins (12, 13).

IAV infection triggers massive changes in chromatin compaction, gene expression, protein synthesis, vesicle trafficking, and organization of the cytoskeleton (14–16). Many of the virus-induced cell responses serve to facilitate or antagonize virus propagation. Virus-supportive cell functions include translation, posttranslational modification and maturation of viral proteins, including the viral glycoprotein hemagglutinin (HA). Accordingly, various genome-wide RNA interference (RNAi) screens have enabled the identification of host factors supporting IAV replication (17–19), although with a surprisingly low incidence of overlap (20). These screens showed that the cellular proteins supporting IAV replication participate in all basic cellular functions, including signal transduction, nucleic acid metabolism and transport, vesicle trafficking, and all steps of gene expression from chromatin organization to transcription and translation (20). On the other hand, infected host cells initiate signaling pathways that counteract viral infection, as exemplified by the interferon (IFN) system. The production and release of these antiviral cytokines is initiated by a signaling cascade triggered by the RNA-sensing RIG-I protein. The activated RIG-I can then bind to its downstream effectors to trigger the activation of protein kinases that ultimately lead to the activation of downstream transcription factors of the IRF and NF-κB families (21–23). These signaling events heavily rely on the reversible modification of proteins by posttranslational modifications (PTMs) such as phosphorylation, ubiquitination, and SUMOylation (24–26).

Phosphorylations occur at Ser/Thr and Tyr residues and function as a typical starting point for modification cascades, while degradative ubiquitination is an irreversible endpoint. Ultradeep phosphoproteome analysis revealed the distinct regulatory nature of Tyr and Ser/Thr-based signaling. Phosphotyrosines account for less than 1% of the identified phosphorylation sites, and phosphate groups at this amino acid usually have a short half-life owing to the high activity of phospho-Tyr phosphatases. Moreover, phosphotyrosines are enriched on high-abundance proteins, which led to the suggestion to consider phospho-Tyr as a functionally separate PTM of eukaryotic proteomes (27). Regulated phosphorylation events function as reversible molecular switchboards that control many features such as the formation of multiprotein assemblies, as well as protein stabilities and enzymatic activities (28). Therefore, there is considerable interest in studying IAV-triggered phosphorylation patterns in a systematic fashion, as has already been done in order to study the dynamic changes of protein SUMOylation in IAV-infected cells (24). This can be comprehensively achieved by mass spectrometry-based approaches, as demonstrated in some recent studies of cells infected with different RNA viruses (26, 29–32) and DNA viruses (33, 34).

In this study, we aimed to investigate the phosphoproteomic changes of IAV-infected mouse lung epithelial cells at early and late phases of infection. Since the genetic variability of viruses is of great (patho)physiological relevance in IAV infections, we conducted this study using mouse-adapted and highly infectious SC35M and chicken-adapted SC35 viruses. We identified thousands of regulated modification sites that were overrepresented in several pathways regulating chromatin organization and cell adhesion, including a focal adhesion kinase (FAK)-regulated network regulating actin dynamics. Inhibition of FAK interfered with virus replication. This study also identified new phosphorylation sites on virus-encoded proteins that serve to assist or antagonize viral replication and maturation.

RESULTS

Identification of IAV-regulated phosphorylations.

To identify IAV-dependent phosphorylation events. we interrogated the cellular phosphorylation responses at early (1-h) and late (8-h) time points postinfection (p.i.). In this experiment, we also aimed to compare the responses elicited in widely used murine MLE-15 cells (representing the distal bronchiolar and alveolar epithelium) by mouse-adapted SC35M virus and the nonadapted SC35 strain. A cell extract was prepared in a denaturing buffer to preserve all phosphorylations and inhibit all enzymes (kinases or phosphatases) that could affect the result after lysis. Samples from these lysates obtained from the control and virus-infected cells were validated via Western blot experiments by testing the occurrence of several well-documented phosphorylations in the activation loops of JNK, ERK, and AKT kinases (35) (Fig. 1A). These extracts were then further processed as schematically shown in Fig. 1B. Proteins were digested with trypsin and phosphorylated peptides were enriched by two different strategies. First, peptides phosphorylated at Ser, Thr, or Tyr were enriched via immobilized metal affinity chromatography (IMAC) columns. Since phospho-Tyr does not exceed 5% of all intracellular phosphorylation events (27), we also enriched the Tyr-phosphorylated peptides with specific anti-phospho-Tyr antibodies (pY) from the same extracts. The label-free-based PTMScan method was used for identification and quantification of peptides (36). Immunopurified phosphopeptides were identified by replicate liquid chromatography-tandem mass spectrometry (LC-MS/MS) runs of each sample. The resulting mean intensity values of peptides derived from nanoHPLC profiles were used for further quantification. To normalize global differences between samples, a fixed value normalization strategy was applied separately for IMAC- and pY-enriched peptides.

FIG 1.

Phosphoproteome analysis of IAV-infected MLE-15 mouse lung cells. (A) Cells were infected with SC35M or SC35 for 1 or 8 h as shown. Cells were lysed in a denaturing buffer, and equal amounts of protein contained in the lysates were analyzed by immunoblotting for the occurrence and phosphorylation of the indicated proteins with specific antibodies. (B) Schematic display of the workflow employed for the phosphoproteomic screen. The cell extracts were analyzed using the quantitative label-free PTMScan approach (Cell Signaling Technology). Extracted proteins were digested with trypsin and separated from nonpeptide material by solid-phase extraction with Sep-Pak C₁₈ cartridges. One fraction of the extract was used to enrich peptides phosphorylated at Ser/Thr or Tyr by immobilized metal affinity chromatography (IMAC). Another fraction was taken to enrich Tyr-phosphorylated peptides with a specific antibody. Enriched phosphopeptides were analyzed twice in separate LC-MS/MS runs, and spectra were assigned by using SORCERER 2 v.4.0. Relative quantification was based on the chromatographic peak apex intensity or integrated peak area in the MS1 channel. Validated target lists were then assembled by score filtering and then searched for intended targets or for peptides homologous to intended targets. (C) Total number of peptides and the assigned unique proteins. (D) Distribution of the number of phosphorylation sites per protein.

This normalization was done by multiplying the mean intensity values of each peptide with a fixed constant derived by subtracting the differences between individual log₂ transformed medians and the overall sample median. Medians were thereby adjusted, which allowed the identification and quantification of 6,935 nonredundant, unique peptides enriched by IMAC and 1,679 nonredundant peptides enriched by pY-specific antibodies. Altogether, this data set represents more than 32,000 phosphorylation sites (Fig. 1C and Table 1). These modification sites were then assigned to their corresponding genes and proteins, as shown in Table 1 and in Tables S1 and S2 in the supplemental material. The gene identifiers were used for all further analyses eliminating the (intrinsic) problem that peptide sequences rarely allow discrimination between different splice variants of a gene or between products of different genes with homologous or even identical stretches of amino acid sequences. However, for consistency we decided to refer to the protein names of the corresponding genes throughout this study. Most of the phosphopeptides purified with IMAC columns contained one or two phosphorylated serines. Likewise, the vast majority of peptides enriched by the phospho-Tyr antibody contained at least one phospho-Tyr (data not shown), indicating that the purifications worked properly. The number of IAV-regulated phosphorylation sites obtained for individual proteins varied over a wide range (Fig. 1D). IMAC-enriched phosphorylations typically occurred on several residues, and 28.8% of proteins were phosphorylated at more than five different residues (Fig. 1D, left). In contrast, the majority of Tyr phosphorylations were found only once per protein (Fig. 1D, right).

TABLE 1.

Statistical analysis of the phosphoproteome screen^a

^aCells were infected with SC35M or SC35 influenza viruses for 1 or 8 h. Cell extracts were prepared and pooled, followed by trypsin digestion and enrichment of phosphopeptides either by IMAC or with phospho-Tyr specific antibodies (pY). Data from two replicate LC-MS/MS runs per sample were filtered using customized scripts in R statistical language for nonredundant phosphorylation sites occurring on Ser, Thr, or Tyr residues. All entries for modified residues were merged to the corresponding gene identifier and its corresponding protein names, eliminating multiple entries or ambiguities resulting from differential splicing or protein processing. Upregulated and diminished phosphorylations are shown in red and blue, respectively.

Figure 2 summarizes the overall number of proteins undergoing regulated phosphorylation events, as assessed by dividing normalized peptide intensities from IAV-infected samples by those from uninfected samples. IMAC-enriched phosphorylations (>1.5-fold regulation) in cells infected with SC35 or SC35M were highly abundant and comparable at early or late time points of infection, as was the direction of their regulation (Fig. 2A and B). A further increase in the threshold to >2-fold and >3-fold regulated proteins decreased the number of phosphorylated proteins (Fig. 2A and B). Similarly, the number of SC35/SC35M-induced Tyr phosphorylation sites p.i. also dropped from 1,020 (1 h) or 1,061 (8 h) (>1.5-fold) to 380 (1 h) or 440 (8 h) when the threshold was increased to >3-fold regulation (Fig. 2C and D). Importantly, nearly half of the IMAC-enriched phosphorylations were affected by both virus variants, pointing to a core set of host cell responses (Fig. 2A and B, sum of white stacks). In contrast, both viruses showed more differences in the regulation of Tyr phosphorylation (Fig. 2C and D). The analysis of our data also showed that not all phosphorylation sites within a given protein were uniformly up- or downregulated. Rather, we found the frequent occurrence of oppositely regulated phosphorylations within the same protein, as illustrated by the Venn diagrams in Fig. 2E.

FIG 2.

Summary of IAV-regulated protein phosphorylation events. (A) The IMAC-purified phospho-peptides were mapped to proteins. The proteins experiencing elevated or reduced phosphorylation 1 h after infection with SC35M or SC35 are grouped according to the threshold of the fold regulation. Upregulation is shown in red; downregulation is shown in blue. (B) The analysis was done as in panel A, with the difference that only regulated phosphorylations occurring after 8 h of infection are shown. (C) Proteins harboring Tyr phosphorylation sites identified by enrichment with pY-specific antibodies were grouped according to the threshold of the fold regulation. The numbers of up- and downregulated events after 1 h of infection with SC35 or SC35M are shown. (D) The analysis was done as in panel C, with the difference that only regulated phosphorylations occurring after 8 h of infection are shown. (E) (Upper part) Venn diagrams show a comparison of proteins containing >3-fold regulated phosphopeptides enriched by IMAC. The left side shows SC35-regulated events and displays common and distinct phosphorylations according to time and direction (up or downregulation). The right side shows SC35M-regulated modifications. (Lower part) Venn diagrams displaying results from the same analysis performed for pY-enriched phospho-peptides. Corresponding protein/gene lists and further details are provided in Table S4.

We then extended these analyses to differences and commonalities between virus strain-mediated regulated phosphorylations, including the kinetic information from the 1- and 8-h time points after infection. Totals of 29.6% (SC35) and 23.9% (SC35M) of upregulated and IMAC-enriched phospho-proteins were seen at early and also late times of infection with SC35 (Fig. 3A), while only a small part of the Tyr phosphorylations was seen at both time points (Fig. 3B). In contrast, only 18.8% (SC35) and 10.6% (SC35M) of the proteins showing reduced phosphorylation were shared between early and late infection times (Fig. 3A and B). The same analysis for SC35M-infected cells showed comparable results with the greatest degree of overlap for IMAC-enriched and upregulated phosphorylations (Fig. 3A and B). Since the proteins of SC35 and SC35M viruses only differ at nine amino acid positions, it was interesting to study the impact of these changes on the phosphoproteomes by filtering the data sets for overlapping sets of regulated phospho-proteins. The analysis for IMAC-purified peptides showed that >50% of the upregulated phospho-proteins, and yet only one-third of the proteins with reduced phosphorylations were commonly affected by both viruses (Fig. 3C). Common Tyr phosphorylations were only observed for approximately one-third of the proteins (Fig. 3D). In summary, these data reveal a profound and complex change of the host cell phosphoproteome upon IAV infection by showing that a large proportion of induced Ser/Thr modifications are affected by both viruses at early and late time points. However, this overlap is significantly lower for downregulated Ser/Thr phosphorylations and for Tyr modifications.

FIG 3.

Analysis of overlapping phosphorylation events. (A) Venn diagrams show the numbers of proteins containing SC35 and SC35M-dependent and >3-fold regulated IMAC-enriched phosphorylations at early (1-h) or late (8-h) time points postinfection. (B) The analysis was done as in panel A, with the difference that pY-enriched phosphorylations were analyzed. (C and D) Overlap of proteins showing a >3-fold regulation of phosphorylation during at least one time point triggered by SC35, SC35M, or both viruses. This analysis was performed separately for IMAC-enriched (C) or pY-enriched phosphorylations (D). Corresponding protein/gene lists and further details are provided in Table S5.

Coregulated dynamic phosphorylation patterns define distinct phospho-protein interaction networks regulating genome organization, mRNA biogenesis, and processing.

The data were then analyzed for changes in the extent of phosphorylation for a given protein dependent on the time or virus strain. Stronger changes in phosphorylation not only show a more dynamic regulation but may also indicate a biologically more important modification (37, 38). To differentiate the highly dynamically regulated IMAC-enriched modifications systematically, we filtered the results to display strong changes for up- or downregulated phosphorylations at one time point (using a threshold of >8-fold) and less regulation (defined by a threshold of <3-fold) for the second time point (Fig. 4A). Heat map visualizations of individual protein lists extracted by this strategy reveal highly overlapping but also clearly distinct sets of regulated proteins for each time point and virus strain (Fig. 4A). Thus, early SC35-affected protein modifications were rarely seen at later time points or after infection with SC35M (Fig. 4A, first panel). In contrast, late SC35-regulated phosphorylations were largely identical to late SC35M-regulated modifications (Fig. 4A, second panel). Similarly, the pattern of highly dynamic early (Fig. 4A, third panel) and late (Fig. 4A, fourth panel) SC35M-regulated phosphorylation events were largely mirrored in SC35-infected cells. These data suggest that particularly strongly regulated modifications are often coregulated by both viruses and are therefore likely to be of general relevance for the host response. Moreover, these dynamically modified protein sets appear to be unique for each time point since many components could be assigned to four distinct protein interaction networks (Fig. 4B), based on information deposited in the STRING database (39). These analyses further suggest that these dynamic modifications can be separated into two principal groups. One group distinguishes SC35 from SC35M-induced patterns and is mainly represented by the phosphorylations triggered during early SC35 infection. The other (larger) group reflects commonalities between SC35 and SC35M and is represented by changes at the late time points after infection. This conclusion is corroborated by the identification of 102 proteins regulated uniformly >3-fold at all time points by both virus strains (see the center of the Venn diagram in Fig. 4C). The heat map of individual ratio values shows that >80% of these phosphorylations are induced, suggesting that they support viral replication and/or are part of the (antiviral) host response (Fig. 4D). In support, STRING database analysis showed extensive interactions between these predominantly nuclear proteins, which mainly function to regulate genome and chromatin organization, transcription and signaling (Fig. 4E). Most of these phosphorylation sites have been described and are deposited in the PhosphoSitePlus database, and our study now reports a biological situation where they are strongly regulated. Furthermore, 32% of the identified sites are novel. We then performed the same systematic analysis for the Tyr modified proteins. In contrast to the IMAC-enriched peptides, there was little overlap between the various conditions (Fig. 5A), but some of these proteins occur in complexes (Fig. 5B). We only found 19 pY-enriched proteins that are coregulated under all experimental conditions (Fig. 6A and B). This set of phospho-proteins included JAK1 and STAT5a/b, whose phosphorylation was decreased consistent with a report on IAV-mediated regulation of JAK-STAT proteins (40). The JAK-STAT complex was the only protein complex that emerged from STRING analysis of the 19 coregulated pY proteins (Fig. 6C).

FIG 4.

Identification of virus strain- and time-dependent coregulated networks of IMAC-enriched phospho-proteins. (A) Proteins containing virus-regulated IMAC-enriched phosphorylations were filtered for >8-fold changes at one time point and little regulation (<3-fold) at the second time point. Ratio values for the other two conditions are included for comparison: induced (red) and decreased (blue) phosphorylations are color coded. Coefficient of variation (CV) values for technical replicates below 50% are indicated in green. (B) Lists of the four groups of regulated phospho-proteins shown in panel A were loaded into the STRING database. Protein-protein interaction network models were built based on experimentally documented interactions only with a medium confidence score of >0.4 and were visualized in Cytoscape. The network statistics for nodes and their interactions (=edges) are indicated below each graph. The edge thickness represents the protein-protein interaction probability. (C) Venn diagram showing the overlap of all >3-fold regulated phospho-proteins. A protein list of 102 commonly regulated phospho-proteins (central region of the Venn diagram) is displayed as a heat map in panel D. (E) The 102 commonly regulated phospho-proteins were analyzed via STRING, and a network model was built based on a confidence score of >0.4. As indicated by the P value, the number of interactions was higher than expected by chance. The color code shows the biological functions of the proteins. The table at the right side indicates already known (red) and new phosphorylation sites (black) according to the PhosphoSitePlus database.

FIG 5.

Identification of virus strain- and time-dependent coregulated networks of pY-enriched phospho-proteins. (A) Proteins containing virus-dependent and regulated pY-enriched phosphorylations were filtered for >8-fold changes at one time point and little regulation (<3-fold) at the second time point. Ratio values for the other proteins are included for comparison: induced (red) and decreased (blue) phosphorylations are indicated by the indicated colors, and CV values for technical replicates below 50% are indicated in green. (B) Lists of regulated phospho-proteins shown in panel A were loaded into the STRING database. Only experimentally documented interactions with a medium confidence score of >0.4 are shown, and networks were visualized in Cytoscape. The edge thickness represents the interaction probability. The network statistics are shown below each graph.

FIG 6.

Analysis of coregulated networks of pY-enriched phospho-proteins. (A) Venn diagram showing the overlap of all >3-fold regulated pY-enriched phospho-proteins. The 19 commonly regulated phospho-proteins (central region of the Venn diagram) are displayed as a heat map in panel B. The protein-protein interactions of these components were analyzed by STRING as described in the legends for Fig. 4 and 5 and are displayed in panel C.

In summary, these results provide strong evidence for a prevailing coregulation of proteins phosphorylated at Ser/Thr but not Tyr residues in the host response upon infection with SC35M or SC35 and define a core set of a canonical IAV-induced phosphoproteome. We therefore focused subsequent analyses on this part of the host response to both viruses in order to identify virus genotype-independent common principles in host cell signaling.

Identification of novel IAV-regulated protein kinases.

The kinases mediating protein phosphorylations can be predicted from the identification and database comparisons of common substrate phosphorylation motifs. Here, we used the motif-x-tool to identify de novo motifs and compared them to known protein kinase consensus motifs (41). IMAC-purified peptides after 1 h and 8 h of IAV infection showed many peptides containing the RXXpS motif (Fig. 7A). This motif analysis also identified frequent modification of serines directly flanked by prolines or in the vicinity of acidic amino acids. These sequences correspond to the preferred substrates of different kinases, including AKT, calmodulin-dependent protein kinase II, MAPKAPK1, and PKA (RXXpS motif), as well as GSK-3, MAPKs, and CDK2 (pSP motif) (Fig. 7A). The analysis of IMAC-purified peptides after 1 and 8 h of IAV infection showed many peptides containing the RXXpS motif (Fig. 7A). The analysis of substrate motifs for phospho-Tyr showed the prevalence of very short motifs where the phosphorylated amino acids are directly flanked by Ser, Gly, or Ala or alternatively harbor a Pro in the +3 position (Fig. 7B). However, the absence of typical substrate phosphorylation motifs for Tyr kinases precluded the identification of candidate kinases. The IAV-mediated regulation of protein kinases can also be deduced from the phosphorylation state of key amino acids present in the activation loop, which is contained in most kinase domains. Phosphorylation of these activation loop residues is typically required for full kinase activity (42), and thus the phosphorylation state of the kinase activation loop is often used as proxy to analyze the activation state of the respective kinase (27). All protein kinases undergoing a >2-fold change in the phosphorylation of residues contained in the activation loop during at least one time point after infection with at least one virus are displayed in Fig. 7C and Table S3. This analysis allowed the identification of 37 regulated protein kinases, with the number of activated kinases (12) being comparable to the number of downregulated kinases (14). The remaining kinases showed high volatility in their activation loop phosphorylation due to differential phosphorylation in response to the different viruses, infection time points, or phosphorylation at different amino acids. This activation loop analysis confirmed the regulation of kinases already known to be involved in IAV infection (e.g., JNKs, ERKs, p38, and JAK1) and also revealed 20 regulated kinases that have not been implicated in IAV infection so far (boldface names in Fig. 7C).

FIG 7.

Identification of phosphosite motifs and IAV-regulated kinases and phosphatases. (A) IMAC enriched phosphopeptides regulated by >3-fold after 1 or 8 h of IAV infection were collected. A total of 324 (1 h IAV) or 579 (8 h IAV) nonredundant 15-mer protein sequences −7, +7 amino acid residues around the phosphorylated site were extracted. The phosphosite motif analysis was performed by using the motif-x tool. Sequence frequency graphs of significantly enriched phosphosite de novo motifs for centered Ser are shown; the number of different peptides used for the generation of the logos is indicated at the right. Candidate kinases potentially modifying these sites were identified using the motif-x tool. (B) The phosphosite motif analysis of pY-enriched peptides was performed as in panel A, with the exception that phosphorylation motifs for centered Tyr were identified. Totals of 533 (1 h IAV) or 705 (8 h IAV) nonredundant 15-mer protein sequences −7, +7 amino acid residues around the phosphorylated site were analyzed. (C) Regulation of key residues in the activation loop of kinases regulated by at least one virus. Only kinases where SC35 or SC35M caused >2-fold changes in activation loop phosphorylation during at least one time point are displayed. Arrows pointing up or down indicate increased or decreased phosphorylations, respectively. Kinases in boldface have not been implicated in IAV signaling thus far. Peptide lists identified for the phosphorylated kinases and further details are provided in Table S3. (D) Human proteins associated with the process of dephosphorylation were extracted from the DEPOD database (43), and the corresponding mouse orthologues were identified. These proteins were queried for the occurrence of IAV-regulated phosphorylations. The heat maps display proteins undergoing >2-fold regulation for IMAC-enriched and phospho-Tyr-enriched peptides; the CV values for technical replicates below 50% are indicated in green.

Since the phosphorylation status is a result of the balance between phosphorylation and antagonizing dephosphorylation, it was interesting to investigate whether also phosphatases undergo IAV-regulated phosphorylations. From the 238 catalytic and regulatory subunits of phosphatases extracted from the dephosphorylation DEPOD database (43), 237 mouse orthologues were retrieved. From this group, 25 IMAC-enriched and 10 phospho-Tyr enriched proteins undergo dynamic phosphorylation (>2-fold) during at least one time point in response to at least one virus (Fig. 7D). These data are in agreement with the concept that the dynamic increases and decreases of phosphorylation in response to IAV infection are due to differential activities of kinases and their antagonizing phosphatases.

Identification of host cell pathways enriched in IAV-dependent phosphorylation events.

We then performed an enrichment analysis of functional annotations using KEGG (Kyoto Encyclopedia of Genes and Genomes) and gene ontology (GO) databases and overrepresentation (ORA) analysis. All proteins carrying at least one phosphorylation site that was regulated >3-fold were compared to the entire genome (proteome). The GO analysis for cellular components showed enrichment for annotations related to many different subcellular structures including cell anchoring, adherence junctions, and vesicle trafficking (Fig. 8A), which is consistent with the viral life cycle involving many intracellular compartments. Also, the GO analysis for molecular function showed the involvement of diverse functions, including signaling events, transcription factors, and their regulators, as well as mRNA metabolism (Fig. 8B). The GO analysis for biological processes highlighted several processes, including adhesion and locomotion, as well as the organization of actin and the cytoskeleton (Fig. 8C). These key pathways were also revealed in an independent analysis using the KEGG annotation tool (data not shown). Direct visualization of the modified proteins in the KEGG pathways shows the overrepresentation of phosphorylated proteins in the actin cytoskeleton pathway (Fig. 9A) and focal adhesion pathway (Fig. 9B).

FIG 8.

Enrichment of IAV-regulated host cell functions and signaling networks. (A to C) All proteins showing regulated phosphorylation by >3-fold were analyzed for overrepresentation of the GO terms cellular components (A), molecular functions (B), and biological processes (C). Significant of enrichment is indicated by adjusted P values and dot colors. Dot sizes show the relative number of proteins in the experimental set compared to all proteins belonging to the respective GO term.

FIG 9.

Network diagram displaying the IAV-dependent phosphorylation changes in the actin cytoskeleton and focal adhesion KEGG pathways. (A) Proteins of the actin cytoskeleton pathway are indicated by rectangles. Phosphorylation events measured either after IMAC or phospho-Tyr enrichment at 1 h or 8 h after infection are displayed. Changes in phosphorylation by >2-fold are marked by colors using the R/Bioconductor package Pathview. (B) The same network diagram set up was used to project phosphorylation events onto the focal adhesion KEGG pathway.

FAK-dependent signaling is required for efficient IAV replication.

The focal adhesion pathway and actin reorganization are regulated by the cytoplasmic tyrosine kinase FAK, which is also regulated upon IAV infection (Fig. 7C). Since this kinase can be specifically inhibited by Defactinib (VS-6063, PF-04554878), a small molecule inhibitor already used in several phase II clinical studies for the treatment of cancer (44), we treated MLE-15 cells with Defactinib before and after infection with SC35M and measured the impact on virus propagation. Defactinib treatment interfered with virus propagation, while the inhibition of virus replication was less prominent when the inhibitor was administered 1.5 h p.i. (Fig. 10A), suggesting a role for FAK in the early events during IAV infection. Since FAK controls several downstream pathways, including phosphatidylinositol 3-kinase and MAPK signaling (45), we measured the effects of Defactinib on IAV-triggered signaling cascades. Preincubation of cells with Defactinib slightly inhibited IAV-induced activation of TBK1 and also interfered with the activation of NF-κB signaling, as judged by absent IκBα phosphorylation in the presence of the inhibitor (Fig. 10B). FAK inhibition resulted in slightly increased virus-induced c-Jun phosphorylation despite diminished JNK activity and had no impact on the phosphorylation of AKT or p38. The same experiment also revealed that FAK inhibition also caused delayed and diminished synthesis of the viral NS1 protein (Fig. 10B), in line with the inhibitory activity of Defactinib on IAV replication. The administration of Defactinib to cells 1.5 h p.i. only marginally affected the major signaling pathways (data not shown), suggesting that FAK is important for the entry steps of IAV early during infection.

FIG 10.

Effects of Defactinib on IAV replication and signaling in MLE-15 cells. (A) MLE-15 cells were incubated with 10 μM Defactinib or DMSO 1 h before or 1.5 h after infection with SC35M (MOI = 1). Focus assays were performed at 24 h p.i. on MDCK II cells. The virus titers (FFU/ml) are indicated; error bars show the standard errors of the mean (SEM) obtained from three independent experiments performed in triplicates. (B) MLE-15 cells were incubated for 1 h with 10 μM Defactinib, followed by infection with SC35M (MOI = 1) for 1 h. Cells were lysed at the times (h) p.i., as shown. Equal amounts of protein were analyzed by Western blotting for the occurrence and phosphorylation of the indicated proteins; the results from one of three representative experiments are shown.

Identification of novel phosphorylation sites at IAV-encoded proteins.

The phosphoproteome screen also allowed the detection of phosphorylated residues in viral proteins. We have confirmed seven of the already known phosphorylation sites (46–48) and identified 13 new modified amino acids in several viral proteins (Fig. 11A and B). All modifications were detected for SC35 and also SC35M-encoded proteins. Some of the identified phosphorylation sites occur at sites conserved between various IAV strains with a potential functional relevance. For example, we identified an NS1 modification at Thr49, which is immediately adjacent to the previously identified phosphorylation site at Ser48 in a region that is directly involved in RNA binding and therefore functionally relevant (47, 49, 50). In order to obtain insight into the structural consequences of Ser48/Thr49 phosphorylation, published NS1 structures (51, 52) were used as the templates to model the structure of SC35M NS1 using the Swiss-Model server (Fig. 11C). Both phosphorylation sites are at the end of a α-helix that interacts with dsRNA. The in silico model also suggests direct contact between phosphorylated Thr49 and the RNA ribose. These NS1 phosphorylations were validated by an independent experimental approach using phospho-specific antibodies that were raised against this epitope. Western blotting confirmed the specificity of the phospho-specific antibodies (Fig. 11D) and the occurrence of NS1 phosphorylation (Fig. 11E).

FIG 11.

Phosphorylation of viral proteins. (A) The upper Venn diagram displays the numbers of identified phosphorylation sites on viral proteins, as detected after enrichment of peptides via IMAC columns or pY-specific antibodies. The lower Venn diagram shows the overlap with known phosphorylation sites. (B) List of identified phosphorylated residues in SC35M virus proteins; newly discovered phosphorylation sites are indicated in red. (C) The published NS1 structures were used as templates in order to calculate the structure of SC35M NS1 using the Swiss-Model server. The phosphorylations at Ser48 and Thr49 are shown in red, and the possible contact to dsRNA bases is visualized. (D) Recombinant SC35M viruses expressing either a wild-type NS1 gene or mutants thereof with both Ser48 and Thr49 changed either to Glu or Ala were rescued. These different viruses were used to infect MLE-15 cells (MOI = 1), followed by cell lysis after 8 h and immunoblotting to detect the occurrence and phosphorylation of NS1, as shown. (E) MLE-15 cells were infected with SC35 or SC35M for the indicated periods. Cells were harvested, and cell extracts were analyzed by immunoblotting for NS1 phosphorylation using phospho-specific antibodies.

The functional interactions of virus protein phosphorylations are influenced by the genetic background of the virus.

To investigate the functional relevance of some newly discovered and already documented phosphorylations, we selected conserved residues within functionally important regions and a known structure. These selected residues were changed to create phosphorylation-defective (STY/AF) and phospho-mimicking (STY/E) mutations. A reverse genetics system was used to create mutated SC35M viruses, as visualized in the upper part of Fig. 12A. This system allows to generate recombinant viruses upon transfection of eight plasmids encoding the viral RNAs into a coculture of 293T/MDCK II cells, followed by collection of virions and determination of virus titers. Phosphorylation-defective and phospho-mimicking mutations were produced for SC35 and SC35M-encoded NS1 proteins and also for the polymerase subunit PB1 and neuraminidase (NA). We rescued SC35M viruses for all mutations except for the mutation of PB1 Thr223 to Glu, which did not yield infectious virus particles in four independent experiments (Fig. 12A, middle). Interference with PB1 phosphorylation at Tyr657 by mutation to Phe (PB1 Y657F) allowed virus replication (Fig. 12A), suggesting that this phosphorylation is not required. However, we could not test the effects of constitutive phosphorylation at this site, since a bulky side chain is required at this position and already mutation to alanine abrogated its function (data not shown). In other cases, virus growth was prohibited only by the simultaneous mutation of several phosphorylation sites on different viral proteins. For example, mutation of NS1 at Ser48/Thr49 to Glu (NS1 ST48,49EE) did not impair virus production (Fig. 12A). However, mutation of NS1 at Ser48/Thr49 to Glu, together with the mutation of NA Ser178 either to Glu or to Ala (NA S178A/E), failed to yield infectious particles (Fig. 12A). These data suggest a functional interaction between the different phosphorylation sites. It was also interesting to study the potential impact of the genetic background of the viral proteins on the function of the phosphorylation. To address this question, we mutated the Ser178 phosphorylation site also in NA encoded by SC35. While the SC35M-encoded NA Ser178/Glu mutant resulted in virus progeny, the SC35-encoded NA Ser178/Glu mutant did not yield infectious IAV particles (Fig. 12A). Intriguingly, the SC35M- and SC35-derived NA proteins differ only at amino acid 340 (12), showing that the genetic background of the virus is an important codeterminant for the functional outcome of IAV protein phosphorylation. To further explore the relative contribution of the viral genetic background on the effect of viral protein phosphorylation, we generated recombinant SC35 expressing mutated NA and NS1 of SC35M. In this setting the phospho-mimicking versions of SC35M NA and NS1 (which yielded infectious virus particles in the SC35M background) failed to generate virus progeny (Fig. 12B), reinforcing the notion that the functional effects of virus protein phosphorylation depend on the genetic background of the virus.

FIG 12.

Functional analysis of virus protein phosphorylation. (A) The top panel shows a schematic display of the workflow for the production of recombinant IAV viruses. In the the bottom panel, 293T/MDCK II cells were transfected with plasmids encoding SC35M genome segments expressing wild-type proteins, together with the plasmids expressing genome segments encoding either the indicated SC35M (black) or SC35 (blue) proteins in their wild-type or mutated forms in an SC35M background. These viruses were used to infect MLE-15 cells (MOI = 0.001), and the virus titers were determined at 24 h p.i. Errors bars indicate the standard deviations from three independent experiments performed in triplicates. (B) The experiment was performed as in panel A in a SC35 background using the indicated plasmids. In all experiments, successful transfection of 293T/MDCK II cells was ensured by qPCR detecting viral RNAs (not shown).

Phosphorylation of IAV proteins supports or antagonizes their function.

Next, we tested the function of virus protein phosphorylation by infecting MLE-15 lung cells with recombinant SC35M viruses, as schematically shown at the top of Fig. 13A. All viruses were able to propagate in MLE-15 cells and most of the changes in virus titers where within one log₁₀ scale (Fig. 13A, lower). Significantly elevated infectivity was seen in a double mutant expressing a phosphorylation-deficient NS1 mutant together with a phospho-mimicking NA Ser178Glu variant, while the lowest virus titers were seen in viruses expressing the NA Ser178Ala mutant (Fig. 13A, lower). These results raise the possibility that inhibition of NA Ser178 phosphorylation interferes with its enzymatic activity. To test this hypothesis, we determined the NA activity of the SC35M virus and its derivatives by expressing NA phosphorylation-deficient and phospho-mimicking versions. Various dilutions of IAVs were used to determine NA activity using an in vitro assay. The phosphorylation-deficient NA Ser178Ala mutant consistently showed impaired NA activity (Fig. 13B), suggesting either that NA Ser178 phosphorylation assists in its enzymatic activity or that the Ser as such is important. Molecular modeling showed that Ser178 is close to the active center, which was visualized by displaying the location of the competitive inhibitor Laninamivir (Fig. 13C, upper panel). A molecular model predicting the structural changes triggered by NA Ser178 phosphorylation showed that the phosphate group will result in the movement of the Glu226 side chain and might possibly make contact with Thr224. The phosphorylation-induced structural changes might facilitate substrate docking while leaving the enzymatic center intact (Fig. 13C, lower panel). While NA phosphorylation is an example of a stimulatory phosphorylation, it was also interesting to investigate inhibitory phosphorylations. This is exemplified by PB1 Thr223 phosphorylation, which does not allow IAV generation (Fig. 13D, left). To test whether this modification affects polymerase activity, we used a plasmid-based minireplicon system (53). A polymerase I (Pol I)-driven plasmid encoding the luciferase reporter gene was cotransfected with plasmids encoding the viral PB2, PA, NP, and PB1 proteins and its mutated derivatives into 293T cells. The analysis of luciferase activity showed that the phospho-mimetic PB1 Thr223Glu variant had completely lost its enzymatic activity (Fig. 13D, right). In silico modeling of PB1 suggests that phosphorylation of Thr223 in the polymerase fingertip might allow interaction with the side chain of R350, thus negatively affecting RNA template binding and NTP channeling (Fig. 13E).

FIG 13.

Pro- and antiviral effects of IAV protein phosphorylation. (A) Schematic display of the workflow. Seven plasmids encoding SC35M were combined with the indicated plasmids from SC35M (black) to produce viruses mutated only in the respective SC35M proteins in an SC35M background. Alternatively, plasmids encoding SC35 proteins (blue) were mutated and combined with seven plasmids from SC35M to produce viruses mutated only in SC35 proteins within the SC35M genetic background. These reassortant viruses were produced in 293T/MDCK II cells and then titrated to infect MLE-15 cells (MOI = 0.001). Virus titers were determined at 24 h p.i. In order to facilitate comparison, the titer of the SC35M virus was set as 100%. The relative virus titers and SEM derived from three independent experiments performed in triplicates are shown on a log₁₀ scale. (B) Reassortant viruses expressing wild-type NA or phosphorylation-deficient or phospho-mimicking versions were produced, and 60 HA units were used in various dilutions to measure HA activity with an ELISA. Relative fluorescence units (RFUs) measuring the conversion of the fluorescent MUNANA substrate are shown; error bars indicate the standard deviations. (C) The upper part shows a molecular model of NA from SC35M based on the high homology to the structure for PDB 4QN7 showing neuraminidase subtype N7 complexed with oseltamivir. The electrostatic potential of the molecular surfaces of NA are color coded; the position of the catalytic center is visualized by the binding of laninamivir, and the position of Ser178 at the surface is shown. The lower part shows ribbon models for NA from SC35M in the unphosphorylated (blue) and phosphorylated (green) state. (D) On the left, reassortant SC35M viruses harboring the wild-type or point-mutated forms of SC35M PB1 were generated in 293T/MDCK II cells, and the titer was determined. Error bars show SEM obtained from three independent experiments performed in triplicates. On the right, 293T cells were transfected to express a Pol I-driven firefly luciferase reporter gene, together with a control Renilla luciferase gene and expression vectors encoding PB2, PA, NP, and both wild-type and mutated forms of PB1. After 1 day, cells were harvested and split into two aliquots. The first aliquot was analyzed for proper and comparable PB1 expression by qPCR (data not shown), while the second aliquot was used for the determination of reporter firefly and Renilla luciferase activities. The values show corrected firefly luciferase activities as the fold induction relative to the empty vector control; bars indicate the SEM from two independent experiments performed in triplicates. (E) The published PB1 structure was used to model the impact of PB1 Thr223 phosphorylation on template binding. The impact of Thr223 phosphorylation on the contact to uracil 15 in the RNA is shown in the enlarged section.

Functional and evolutionary conservation of proteins undergoing IAV-dependent phosphorylation changes.

It was then interesting to study the possible role of host cell phosphorylations in proteins with a known importance in the IAV life cycle. Such candidate proteins were revealed in a number of different whole-genome shRNA screens (17, 18, 54, 55). Of the 128 human proteins that affected influenza virus replication in at least two independent screens (56), 126 proteins are also present in mice. One-third of these proteins with a putative role in the IAV life cycle were dynamically regulated by IAV-dependent phosphorylations (Fig. 14A), showing a strong overrepresentation of phosphorylation events for these proteins.

FIG 14.

Meta-analysis of IAV-mediated protein phosphorylation. (A) The human genome encodes at least 128 host cell proteins with a functional relevance for IAV replication, as revealed by their identification in at least two independent loss-of-function screens (56). The mouse genome encodes 126 orthologues, and our study reveals that 42 of them undergo dynamic IAV-dependent changes of phosphorylation, as summarized by the table. The proteins are grouped according to the strength of their regulated phosphorylation. (B) The total set of 1,675 human phospho-proteins identified in human macrophages before and after IAV infection (26) was used to assign 929 murine orthologues. As indicated by the Venn diagram, 887 phospho-proteins overlapped with our study. (C) A heat map visualizes all phospho-proteins of this group with a >4-fold regulation in at least one infection condition. The orthology analysis and the nonredundant peptides visualized in panels B and C are shown in Table S6 in the supplemental material. (D) The known and novel phosphorylation sites of viral proteins identified in this study were combined, and the percentage of phosphorylated residues was calculated based on SC35M proteins.

Since IAVs infect many different species ranging from birds to horses, pigs, and aquatic mammals, further phosphoproteomic studies are needed in order to allow identification of protein residues that are also frequently phosphorylated in other host species in response to IAV infection. One further phosphoproteomic study investigated the host cell response after IAV infection of human macrophages (26). To investigate whether different influenza viruses infecting different cell types from different species regulate the phosphorylation of an overlapping set of proteins, we compared the data sets from the study of Söderholm et al. with our study. The Söderholm study identified phosphorylation of 1,675 unique proteins (corresponding to 1,670 unique gene names). One half (887) of these proteins were also dynamically phosphorylated in MLE-15-infected cells (Fig. 14B). Of these, 266 were regulated by >4-fold, revealing a surprisingly high overlap of proteins undergoing regulated phosphorylation despite differences in host cell species, cell types, and IAV strains. The heat map displayed in Fig. 14C visualizes the strongly regulated phosphorylations (>4-fold) for the group of overlapping phospho-proteins.

DISCUSSION

We identified here thousands of IAV-regulated phosphorylation events in mouse lung cells. This dynamic adjustment of the phosphoproteome either will be a direct consequence of virus infection or is the result of secondary cell stress events, especially during late infection times. The comparative analysis of mouse-adapted SC35M and chicken-adapted SC35-mediated phosphorylations showed a stronger overlap between the two viruses for inducible phosphorylations, suggesting that in this setting the upregulated phosphorylations are of broader relevance. While approximately half of the regulated Ser/Thr phosphorylations were common to SC35 and SC35M, this overlap was significantly lower for Tyr phosphorylations. This finding supports the previous notion of a distinct regulatory nature of Tyr and Ser/Thr-based signaling (27).

At this stage, we cannot formally differentiate between changes in phosphorylation for a given peptide and changes in the level of the protein that are reflected in the phosphopeptide abundance. Thus, further validation of candidates will be necessary to differentiate between these two possibilities. This study provides a valuable data set to start from, but follow-up studies are needed to validate the candidate sites. The systematic description of the dynamic phosphoproteome provides a rich resource for further functional studies. Rather than focusing on the role of individual phosphorylations, it will be important to focus these studies on the function of IAV-regulated kinases, since they act as upstream regulators affecting the modification of many downstream phosphorylation events. This screen not only confirmed the involvement of known kinases such as ERK1/2 and JNK signaling in IAV infection (35) but also revealed new candidate kinases as identified by regulated activation loop phosphorylation and substrate motif analysis. It will be interesting to test the contribution of these new kinases to IAV infection in order to reveal new potential targets for antiviral drugs. The number of IAV-activated kinases roughly corresponds to the number of downregulated kinases, which mirrors the finding of a comparable up- and downregulation of substrate protein phosphorylations. The activation loops of some kinases, including FAK, show the simultaneous occurrence of increased and decreased phosphorylations at specific residues. This variation might reflect truly opposing regulation within the same protein, the existence of different pools of up- or downregulated kinase species within the same cell or simply cell-to-cell variability. It will therefore be important to determine the activation status of IAV-regulated kinases with the help of phospho-specific antibodies at the single cell level. We discovered that FAK inhibition by the preclinically used FAK inhibitor Defactinib interferes more efficient with virus replication upon preincubation prior to infection, consistent with a proposed role of FAK for the early steps of virus infection (57). FAK controls multiple downstream pathways and its inhibition will not only affect JNK activity as shown in this study but most probably also affect additional effector pathways. At this stage it cannot be excluded that the inhibitory effects of Defactinib on IAV replication might depend on both suppression of FAK activity and inhibition of other kinases, since this clinically used inhibitor might cotarget further enzymes (58). FAK and its downstream effectors cooperate to mediate the massive cytoskeletal reorganizations that facilitate virus entry and budding (15) but probably also control intercellular spreading of the virus (59).

While some of the phosphorylations will be of critical importance for IAV propagation, others will not have a direct physiological consequence. It is assumed that a substantial part of phosphorylation events are nonfunctional phosphorylations that may result from random encounters between protein kinases and degenerated substrate recognition motifs of substrates in the crowded environment of the cell (60, 61). It is tempting to speculate that these nonfunctional phosphorylations may be overrepresented in the group of phosphorylations triggered by only one virus. In the future it will thus be very important to identify the key phosphorylation sites with a trackable functional role in virus replication. Candidates for functionally relevant phosphorylations may be found in some of the cophosphorylated protein complexes identified in this study. These complexes include proteins involved in chromatin compaction and all major steps of gene regulation. Early studies showed the association of IAV proteins and RNAs with nuclear chromatin (14) and more recent work revealed that IAV infection leads to chromatin changes and decreased H3K4 trimethylation (62). Our data also show massively regulated phosphorylation in various chromodomain-helicase-DNA binding (CHD) chromatin remodelers, which are known as regulators of IAV multiplication (62–64). Genome three-dimensional organization during IAV infection is also affected by rampant transcription where readthrough transcription of elongating RNA Pol II disrupts chromatin interactions and leads to reorganization of chromatin loops (65). Also, posttranscriptional regulation is important for IAV propagation, since the nuclear RNA exosome coordinates the initial steps of viral transcription with RNA Pol II at host promoters. The viral polymerase exploits the quality control mechanisms that lead to the degradation of 3′ end of unprotected regulatory RNAs to allow synthesis of host-viral chimeric RNAs, which are important for licensing transcription of antisense genomic viral RNAs (66).

This study also identifies a number of new phosphorylation sites in IAV proteins. The published and new phosphorylations are not equally distributed between the viral proteins, as quantified in Fig. 14D. While no phosphorylations have been found for the PB1-F2 and PA-X proteins, the matrix protein M1 is heavily phosphorylated at multiple residues. An important factor to consider for the functional analysis of IAV protein phosphorylation is the finding that the functional impact of some modifications such as NS Ser48/Thr49 phosphorylation depends on the genetic background of the virus. Furthermore, the effects of some phosphorylations only emerge upon simultaneous modification of several IAV proteins. It will therefore be important to reveal the functions of the phosphorylations occurring at residues that are conserved between different IAV strains. Systematic studies are needed in order to identify residues supporting or antagonizing the functions of viral proteins independent from the genetic background of the virus. This classification will be of potential practical relevance, since the identification of virus-supportive phosphorylations and the responsible kinases could open new avenues for therapeutic interference with IAV replication. On the other hand, phosphorylations inhibiting the function of viral proteins, such as the modification of PB1 at Thr223, provide interesting examples for new antiviral signaling mechanisms.

MATERIALS AND METHODS

Antibodies and plasmids.

Antibodies recognizing phospho-ERK T202/Y204, phospho-JNK T183/Y185, phospho-TBK1 S172, phospho-p38, T180/Y182, phospho-cJun S63, JNK, phospho-IκBα S32,36, phospho-AKT (T308), phospho-AKT (S473), and TBK1 were from Cell Signaling Technology. Antibodies recognizing actin (Abcam) and tubulin (Sigma) were used as loading controls; anti-NS1 antibodies were obtained from S. Ludwig (Münster, Germany), and phospho-NS1 S48/T49 antibodies were produced by Eurogentec. All expression vectors for SC35 and SC35M segments were as described previously (12), and point mutants thereof were produced using the QuikChange mutagenesis kit (Agilent). Luciferase assays were done with the plasmid pHW72-Luci (67), and pSV40-Renilla (68) was used for normalization.

Cell culture and transfections.

Human embryonic kidney (HEK) 293T, MDCK II, and murine MLE-15 lung epithelial cells, were cultured at 37°C and 5% CO₂ in Dulbecco modified Eagle medium containing 10% fetal calf serum, 100 U/ml penicillin, and 100 μg/ml streptomycin. Cells were transfected using Lipofectamine 2000 (Invitrogen) or Rotifect (Roth) according to the instructions of the manufacturer.

Virus amplification and titration.

SC35 was originally derived from the seal isolate A/Seal/Mass/1/80 (H7N7) via serial passages in chicken embryo cells (11), and SC35M was obtained from SC35 by sequential passages in mouse lung (12). SC35M was propagated in MDCK II cells, and SC35 was grown in embryonated chicken eggs. We do not know whether the amplification resulted in the acquisition of additional adaptations. Titers of the virus stocks were determined by titration in MDCK II cells performed in triplicates, as described previously (69).

Generation of recombinant SC35 and SC35M viruses.

The pHW2000 plasmids containing the cloned segments of SC35 and SC35M (12) were transfected into a coculture of 293T/MDCK II cells (ratio 3:1) using Lipofectamine 2000 and further handled as described previously (69).

Cell lysis and Western blotting.

Cells were washed once with phosphate-buffered saline (PBS), harvested by scraping, and collected by centrifugation. The cells were resuspended in 200 μl of 1× sodium dodecyl sulfate (SDS) sample buffer (50 mM Tris-HCl [pH 6.8], 10% [vol/vol] glycerol, 3% [vol/vol] β-mercaptoethanol, 2% [wt/vol] SDS, 0.004% [wt/vol] bromophenol blue). After boiling the samples for 4 min, the samples were sonicated two times for 20 s with a Branson sonifier, and lysates were analyzed by Western blotting. After SDS-PAGE and transfer of the proteins using a semidry blotting apparatus (Bio-Rad), the membrane was incubated with Tris-buffered saline–Tween (TBS-T) containing skimmed milk or bovine serum albumin. The membranes were transferred to primary antibody solutions and incubated at 4°C for several hours. After the membranes were washed and incubated with appropriate secondary antibodies conjugated with horseradish peroxidase, they were washed again with TBS-T. Antigens were revealed by enhanced chemiluminescence as detected with a ChemiDoc Touch instrument (Bio-Rad).

Phosphoproteome sample preparation.

For each sample, MLE-15 cells were grown on 10 dishes (145-mm diameter) to 80% confluence. The cell dishes were infected with SC35 or SC35M (multiplicity of infection [MOI] =1) for various periods, washed twice with PBS (without Ca/Mg), and directly lysed on the plate upon addition of fresh urea lysis buffer (20 mM HEPES [pH 8], 9 M urea, 1 mM activated sodium orthovanadate, 2.5 mM sodium pyrophosphate, 1 mM β-glycerol-phosphate). The cells were carefully resuspended without foaming and extracts were sonified with a Branson device for 4 × 30 s. Different extracts from the 10 dishes per condition were pooled and centrifuged for 15 min at 4°C and 20,000 × g. The supernatant was taken, the protein concentration was determined using a Bradford assay, and an aliquot was used for test Western blotting.

Samples were reduced at 55°C with 4.5 mM dithiothreitol for 30 min, followed by alkylation with iodoacetamide (19 g/liter H₂O) for 15 min at room temperature in the dark. After 1:4 dilution with 20 mM HEPES (pH 8.0), the proteins were digested overnight with 10 μg/ml trypsin-TPCK (tolylsulfonyl phenylalanyl chloromethyl ketone). Digested peptides were acidified with 1% trifluoroacetic acid (TFA) and desalted over 360-mg Sep Pak Classic C₁₈ columns (catalog no. WAT051910; Waters, Milford, MA). Elution of peptides was performed with 40% acetonitrile in 0.1% TFA, followed by drying under vacuum. Enrichment of phosphorylated peptides was done by two strategies in parallel. Five hundred micrograms of protein extract was used to enrich Ser/Thr/Tyr-modified peptides via IMAC columns, while 20 mg of extract was used to enrich phospho-Tyr containing peptides by immunoprecipitation with antibodies.

LC-MS/MS analysis.

LC-MS/MS analysis was done using the LTQ-Orbitrap-Velos instrument. Peptides were loaded directly onto a PicoFrit capillary column (10 cm by 75 μm) packed with Magic C18 AQ reversed-phase resin. The column was developed by using a 150-min linear gradient of acetonitrile in 0.125% formic acid delivered at 280 nl/min. The MS parameter settings were as follows: MS run time, 96 min; MS1 scan range, 300.0 to 1,500.00; and top 20 MS/MS (min signal, 500; isolation width, 2.0; normalized coll. energy, 35.0; activation-Q, 0.250; activation time, 20.0; lock mass, 371.101237; charge state rejection enabled; charge state 1+ rejected; dynamic exclusion enabled; repeat count, 1; repeat duration, 35.0; exclusion list size, 500; exclusion duration, 40.0; exclusion mass width relative to mass; exclusion mass width, 10 ppm). MS/MS spectra were evaluated using SEQUEST (70). The raw mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the data set identifier PXD009634. Each sample was run twice nonsequentially on the instrument both to maximize the number of identifications and also to provide a basis for measurement of analytical reproducibility on the instrument. For each identified peptide, %CV (percent coefficient of variation) values were generated by dividing the standard deviation of the replicate peak area measurements by the average peak area and converting to percentage. For each peptide and fold change reported, the individual %CV values for each sample can be found in Tables S1 and S2, as indicated by the headers.

Some peptides map to more than one protein due to sequence identity. In these cases, to generate the most complete view of the data, we have listed all possible matching proteins separated by semicolon in Tables S1 and S2. No approach was taken to try and resolve to a single protein for these cases since, in the absence of further criteria, it is impossible to unambiguously assign the peptide to one specific protein isoform or paralog. Modification sites are reported as determined by the SEQUEST search. For each peptide the site localization scores (AScore) were calculated. Score values of ≥13 were considered of sufficient confidence to correctly assign a modification (71).

Searches were performed against NCBI Mus musculus version 2011 plus the influenza virus database version 2015 with a mass accuracy of ±50 ppm for precursor ions and 1 Da for product ions; further information is given in the FASTA files provided with the data set PXD009634 on ProteomeXchange. Results were filtered with a mass accuracy of ±5 ppm on precursor ions and the presence of the intended motif. A 5% default false-positive rate was used to filter the Sequest results. Each sample was measured twice in separate MS/MS runs. The results were given as spreadsheets with information on the peptide measurements. There are cases where more than one peptide matches to a LC-MS/MS peak due to ambiguous site localization, and in those cases both possibilities are reported and flagged in the data with red text in the intensity columns (Tables S1 and S2, sheet “non redundant”).

Experimental design and statistical rationale.

Statistical analyses were performed with customized scripts using R statistical language (R Core Team, 2017; https://www.R-project.org/) including packages for R or for R/Bioconductor (72, 73). For Fig. 1C, we used grouping functions of the R package dplyr version 0.7.1 (https://CRAN.R-project.org/package=dplyr). Tables containing color-coded columns and heat maps were created with Microsoft Excel 2010. Fig. 1C and Fig. 2A to D were created with SigmaPlot version 11.2 (Systat Software). The Venn diagram in Fig. 3 was created with Venn Diagram Plotter (v1.5.5228; 25 April 2014, Pacific Northwest National Library [https://omics.pnl.gov/software/venn-diagram-plotter]). The Venn diagrams in Fig. 4C and 6A were created using the Venn diagram web tool from the University of Ghent, Ghent, Belgium (http://bioinformatics.psb.ugent.be/webtools/Venn). The Venn diagrams in Fig. 11 were created by using the R package VennDiagram (74). Sets of regulated phospho-proteins in the kinetic analyses shown in Fig. 4A and D, 5A, and 6B were analyzed with the STRING database, version 10.5 (39), and network models were built based on experimentally documented interactions only with a medium confidence score of 0.4. These were visualized using Cytoscape version 3.5.0 (75, 76) by using the STRINGapp (39).

Networks visualized with Cytoscape show only the protein components with known interactions. The edge thickness additionally represents the interaction probability as deduced from STRING. The phosphosite motif analysis (Fig. 7) was performed by using motif-x version 1.2 10.05.06 (41, 77). The analyses were conducted with the following settings: prealigned foreground sets; 15-mer protein sequences −7, +7 amino acid residues around the phosphorylated residue of all regulated peptides; background set: all mouse proteins in database ‘ipi.MOUSE.fasta’; significance threshold: ‘0.000001’. Sequence frequency graphs were created using Weblogo version 2.8.2 (78). Enrichment analyses for GO terms (Fig. 8A to C) and KEGG pathways (79, 80) were done using the compareCluster function of the R package ClusterProfiler version 3.4.4 (81). These enrichment tests are based on hypergeometric distribution, also taking the false discovery rate of multiple testing into account. The dot charts show at least 10 of the enriched GO/KEGG pathways with P values of ≤0.01 per experimental variant. Redundantly enriched GO terms were removed by applying the simplify procedure (R package GOSemSim version 2.2.0; Wang’s method, p.adjust, >0.7) (82). The KEGG pathways in Fig. 9 were superimposed with protein regulation data by the use of the R/Bioconductor package Pathview version 1.16.1 (83). The strongest peptide regulation was mapped as a color code on the nodes of the networks. The joint overrepresentation analysis of KEGG and GO biological process terms shown in Fig. 8 was done in Cytoscape version 3.5.0 using the app ClueGO version 2.3.4 (84). A total of 1,562 annotated mouse proteins showing >3-fold regulation were submitted as ‘Entrez Gene IDs’ to a two-sided hypergeometric test with Bonferroni step-down correction against a reference set of 21,838 genes selected from the utilized GO/KEGG terms. GO term hits were reduced by using the ‘GO Fusion’ algorithm using a ‘Min GO Level’ = 1, ‘Max GO Level’ = 2, and ‘GO group’ algorithm by using the ‘Kappa statistic’ with a ‘Kappa Score Threshold’ = 0.6 and a ‘Sharing Group Percentage’ = 50.

The data comparison with the study by Söderholm et al. (26) was performed as follows. All gene names of the 1,675 unique phospho-proteins identified in human macrophages infected with IAV (Table S4 from reference 26) were retrieved from the UniProt knowledge database (release 2019_01). The resulting list of 4,343 human gene names mapped to 1,870 orthologous murine genes that were retrieved from the Ensembl database (release 95, January 2019). A total of 929 matching orthologous mouse genes were found in our own study comprising 3,177 total gene names (Table 1). A total of 42 orthologous genes could not be unambiguously assigned between human and mouse genes as they had (i) differing gene names, (ii) a ‘one to many’ relationship in terms of orthology between species, or (iii) were not assigned as ‘gene name (primary)’ on UniProt.

Modeling of protein structures.

The structure of phosphorylated NS1 was modeled according to the published structures of highly similar NS1 proteins (PDB 2ZKO, 2N74, 3M8A, and 2Z0A) using the Swiss-Model automated comparative protein modeling server (85). The templates for the modeling of phosphorylated NA (PDB 4QN7, 1L7F, and 1L7G) and PB1 (PDB 2ZKO, 2N74, 3M8A, and 2Z0A) were published structures. Phosphorylation and subsequent refinement of phosphorylated residues was done with the interactive graphics program Coot (86) and the PHENIX program suite (87). Figure 11E and Fig. 13C and E were created with UCSCF Chimera (88).

Neuraminidase assays.

A coculture of MDCK II and 293T cells was transfected with eight plasmids containing the segments encoding the SC35M genome. After 2 days, the viruses were collected from the supernatant and the samples used for NA assays were treated with 0.2% (vol/vol) Triton X-100 in PBS to inactivate the viruses. Equal titers of the IAVs were serially diluted and NA activity was measured by fluorescence-based assay (NA-Fluor influenza neuraminidase assay kit; Applied Biosystems), and the relative fluorescence was measured by using a FLx800 microplate fluorescence reader with an excitation wavelength of 360 nm and an emission wavelength of 460 nm.

Luciferase assays.

293T cells were seeded in 6-well plates and cotransfected with pHW2000-based plasmids encoding SC35M-derived PB1 or its point-mutated versions, PB2, PA, and NP, together with the Pol I-driven firefly luciferase reporter plasmid pHW72-Luc and a plasmid coding for Renilla luciferase for normalization. One day after Lipofectamine-mediated transfection, the cells were harvested. Cell extracts were prepared in passive lysis buffer from the dual-luciferase reporter assay system (Promega). The activities of firefly and Renilla luciferases were determined by using a Berthold DuoLumat LB 9501 luminometer.