The 11th influenza A virus protein PB1-F2 was previously shown to enhance apoptosis in response to cytotoxic stimuli. The 87 amino acid protein that is encoded by an alternative reading frame of the PB1 polymerase gene was described to localize to mitochondria consistent with its proapoptotic function. However, PB1-F2 is also found diffusely distributed in the cytoplasm and in the nucleus suggesting additional functions of the protein. Here we show that PB1-F2 colocalizes and directly interacts with the viral PB1 polymerase protein. Lack of PB1-F2 during infection resulted in an altered localization of PB1 and decreased viral polymerase activity. Consequently, mutant viruses devoid of a functional PB1-F2 reading frame exhibited a small plaque phenotype. Thus, we have identified a novel function of PB1-F2 as an indirect regulator of the influenza virus polymerase activity via its interaction with PB1.
Influenza A virus (IAV) is a pathogen of humans and animals with pandemic potential. IAV has a segmented negative sensed RNA genome that encodes for up to 11 viral proteins. An important part of the IAV replication cycle occurs in the nucleus of infected host cells. The viral RNA (vRNA) is synthesized by the viral polymerase complex that consists of the polymerase subunits PB1, PB2, PA and the RNA binding nucleoprotein (NP). The newly formed viral ribonucleoprotein complexes (vRNPs) have to be transported out of the nucleus involving cellular and viral proteins (Boulo et al., 2006) and active as well as passive regulatory processes of the host cell (Ludwig et al., 2006).
One attribute of IAV infection is the induction of apoptosis in the infected cells with the typical markers as chromatin condensation, DNA fragmentation, cell shrinkage and caspase activation (Takizawa et al., 1993; Hinshaw et al., 1994; Mori et al., 1995; Balachandran et al., 1998; Lin et al., 2002; Brydon et al., 2003). Apoptosis induction is primarily an antiviral response but also appears to be beneficial for the virus in later stages of the replication cycle (Wurzer et al., 2003; 2004; Mazur et al., 2007). The viral components that have been reported to induce apoptosis include the viral neuraminidase (NA) (Schultz-Cherry and Hinshaw, 1996; Morris et al., 1999), viral single or double stranded RNA intermediates (Balachandran et al., 1998; 2000) and the non-structural protein NS1 (Schultz-Cherry et al., 2001). Nevertheless, the role of NS1 in virus-induced apoptosis is still under debate as NS1 deficient mutant viruses are also strong apoptosis inducers (Zhirnov et al., 2002). A novel player in the IAV induced apoptosis was reported a few years ago (Chen et al., 2001). This 11th IAV gene product named PB1-F2 is an 87 amino acid protein (dependent on the virus isolate) encoded by an alternative open reading frame of the PB1 gene segment (Chen et al., 2001) of most human and avian virus isolates (Zell et al., 2006). Unlike initially reported, the PB1-F2 protein does not induce apoptosis by itself, but rather seems to sensitize cells in response to apoptotic stimuli (Yamada et al., 2004; Zamarin et al., 2005). PB1-F2 was identified to interact with the mitochondrial permeability transition pore complex components ANT3 (adenine nucleotide translocator 3) and VDAC1 (voltage dependent anion channel 1) and may thus play a role in the induction of mitochondria-mediated apoptosis (Zamarin et al., 2005). Knockout of PB1-F2 in laboratory isolates of the H1N1 subtype showed no adverse effect on viral replication in tissue culture but a re-assortant attenuated H1N1 virus mutant that lacks PB1-F2 exhibited diminished pathogenicity in mice (Zamarin et al., 2006). More recently, PB1-F2 has been identified as an important pathogenicity factor owing to a role of the protein PB1-F2 in the establishment of fatal secondary bacterial pneumonia (McAuley et al., 2007). Furthermore, it has been demonstrated that a single mutation in the PB1-F2 of highly pathogenic H5N1 avian influenza viruses and the Spanish influenza virus isolate from 1918 contributes to increased virulence (Conenello et al., 2007). However, the molecular basis of this role of PB1-F2 in pathogenicity is still elusive (Conenello and Palese, 2007). PB1-F2 localizes to mitochondria, but is also found in the cytoplasm and in the nucleus of infected cells (Chen et al., 2001; Gibbs et al., 2003; Yamada et al., 2004). The different localization patterns suggest that the protein might have different functions during the IAV replication cycle.
PB1-F2 shows variable localization patterns in infected versus transfected cells
In cells infected with the influenza virus strain A/Puerto-Rico/8/34 (PR8), PB1-F2 is expressed in parallel to the viral NP or the PB1 polymerase with a slightly delayed kinetic (Fig. 1A). While NP and PB1 expression is constantly high after several replication cycles, we observed that PB1-F2 expression significantly decreases at later time points in A549 cells (Fig. 1A) as well as in other cell types (data not shown). If PB1-F2 is expressed from plasmids the protein is almost exclusively found in mitochondrial patterns (Fig. 1B, upper panel). This is different from its localization in infected cells. While the protein is still visible at mitochondrial structures, there is also a diffuse cytoplasmic and weak nuclear staining (Fig. 1B, middle panel). Furthermore, the expression levels are very variable and in a small percentage of infected cells PB1-F2 is exclusively localized to the nucleus (Fig. 1B, lower panel). This confirms earlier observations (Chen et al., 2001; Gibbs et al., 2003; Yamada et al., 2004) and suggests that PB1-F2 might have additional functions besides its apoptotic role at the mitochondria.
PB1-F2 knockout viruses show impaired spread in cell monolayers
For functional investigations of PB1-F2 we generated recombinant PR8 viruses lacking the PB1-F2 reading frame. To ensure complete abolition of PB1-F2 protein expression, we not only created a recombinant virus with a mutated start codon (T96→C96, F2C) but also generated a second mutant virus with an additional stop codon at amino acid 12 of PB1-F2 (C129→G129, 2xΔF2) (Fig. 2A) according to a strategy described by Zamarin et al. (2006) to avoid aberrant expression from an upstream start site. While the selected substitutions did not result in amino acid changes in the PB1 protein sequence and did not affect PB1 or NP expression, both virus mutants completely lost their ability to express PB1-F2 (Fig. 2B).
Consistent with previous publications (Chen et al., 2001; Zamarin et al., 2005; 2006) we did not observe significant differences of progeny virus titres between the recombinant wild type (wt) PR8 and the PB1-F2 knockout mutants upon infection of A549, MDCK, MDBK, HeLa or A301 cell lines (data not shown).
However, a striking difference in plaque size between wt and mutant viruses was observed in MDCK monolayers (Fig. 2C). Infection with the PB1-F2 deficient PR8 mutants resulted in significantly smaller plaques. As PB1-F2 was shown by others to interfere with apoptosis induction we wondered whether the smaller plaque phenotype may be due to reduced cell lysis and death. To address this issue we investigated formation of plaques at a microfoci level in immunofluorescence studies (Fig. 2D). While already at this early stage a reduced viral spread in the cell layer was observed upon infection with the PB1-F2 mutant viruses no significant cell lysis was detected in wt or mutant infected cells (Fig. 2D). This argues against a prominent role of apoptosis in the observed phenomenon. To further analyse the apoptosis-inducing features of wt and mutant viruses we performed additional assays. Neither the overall apoptosis rate, as measured by staining of dead cells (data not shown), nor the level of caspase activation, indicated by the cleavage of the caspase substrate poly ADP-ribose polymerase (PARP) differed upon infection with wt and mutant viruses (Fig. 2B). Thus, we concluded that the lack of PB1-F2 results in a reduced replication and spread of the virus independent of the proapoptotic features of PB1-F2.
PB1-F2 knockout mutants show decreased viral polymerase activity
Virus replication and spread is primarily determined by the activity of the viral polymerase complex. To selectively investigate the viral polymerase activity in the presence or absence of PB1-F2 we used a recombinant polymerase system driving a mini-genome harbouring the firefly luciferase as reporter gene (described in Pleschka et al., 1996). Briefly, we transfected Pol II-driven expression plasmids for PB2, PA, NP and either PB1 wt or the respective mutants lacking the PB1-F2 reading frame (F2C and 2xΔF2) into HEK293. To provide a virus like RNA template we cotransfected a plasmid that drives expression of a reporter gene RNA in negative sense orientation flanked by Pol I promoter and terminator sites. This arrangement allows the transcription of negative orientated virus-like RNA (mini-genome), which can be transcribed into mRNA by the viral polymerase components expressed from the Pol II-driven constructs. Thus, the activity of the luciferase reporter gene is a measure of viral polymerase activity in the presence or absence of PB1-F2. These experiments revealed that in the absence of PB1-F2 the viral polymerase complex displayed a significantly decreased activity (Fig. 3A). The same results were obtained upon infection of mini-genome-harbouring cells with wt or PB1-F2 deficient recombinant PR8 viruses (Fig. 3C). This indicates that PB1-F2 either directly or indirectly interferes with the function of the viral polymerase.
PB1-F2 colocalizes and interacts with PB1
Our results obtained so far suggest a possible role of PB1-F2 in the transcription of the viral genome. This may be achieved by interaction of PB1-F2 with components of the viral polymerase complex. In that respect it was interesting to note, that PB1-F2 partially colocalizes with the PB1 protein but not with other components of the RNP complexes, such as the NP or the PB2 protein (Fig. 4). The colocalization of PB1-F2 with the PB1 subunit suggests a direct interaction or complex formation of both proteins. To investigate this hypothesis we performed interaction studies in yeast two-hybrid assays. For this purpose we fused both, PB1-F2 and PB1 (with knocked-out PB1-F2 reading frame to prevent the expression of non-fused PB1-F2) to the binding domain (BD) or activation domain (AD) of the Gal4 transcription factor (see Experimental procedures). AD- and BD-tagged proteins were coexpressed in yeast and we obtained a strong interaction in both combinations of the alternately fused proteins, indicated by the staining owing to expression of the β-galactosidase reporter gene (Fig. 5A and D). We also observed the interaction between PB1-F2 BD and PB1-F2 AD (Fig. 5G) consistent with the previous findings that PB1-F2 forms dimers and trimers (Chen et al., 2001; Henklein et al., 2005; Bruns et al., 2007). A similar interaction with PB1 was observed with the PB1-F2 protein of another influenza virus strain, A/WSN/33 (WSN) (Fig. 6A). In line with this finding, a knockout of PB1-F2 in the WSN background also results in reduced polymerase activity (Fig. 6B) and in a smaller plaque-phenotype of recombinant viruses (Fig. 6C), although not as pronounced as observed for the PR8 virus.
The interaction between PB1 and PB1-F2 was confirmed in co-immunoprecipitation experiments from virus-infected cells (Fig. 7). Remarkably, PB1-F2 coprecipitated with PB1, but neither with the PA or PB2 polymerase subunit, nor with the viral NP or M1 (Fig. 7, left panel). Conversely, NP, which is the major component of the RNP complexes coprecipitated with PB1 and PA but not with PB1-F2 (Fig. 7, right panel). This indicates that PB1-F2 interacts with the free PB1 protein only, but not with PB1 assembled in RNP complexes.
Lack of PB1-F2 expression leads to a cytoplasmic localization of the viral PB1 protein
The remaining question was, how the exclusive binding of PB1-F2 to PB1 but not to the polymerase complex may lead to an altered polymerase activity. To this end we compared localization of the NP and the PB1 protein in infected cells in the presence or absence of PB1-F2. While the NP, which is the main constituent of the RNP complexes, is still localized in the nucleus 5 h post infection (p.i.) (Fig. 8A), the RNPs are readily exported to the cytoplasm at 7 h p.i. (Fig. 8B). This localization pattern is not altered in the absence of PB1-F2. While there is also no differences in nuclear PB1 localization 5 h p.i. with either wt or PB1-F2 knockout viruses the situation changes dramatically at 7 h p.i. At this time point the PB1 in wt virus infected cells was still predominantly located in the nucleus, while lack of PB1-F2 resulted in a cytoplasmic localization of the polymerase protein (Fig. 8C). This indicates that in the presence of PB1-F2 the PB1 protein is partially retained in the nucleus of infected cells at time points when a major portion of the RNPs are already transported to the cytoplasm. Thus, PB1-F2 may prevent premature transport of PB1 to the cytoplasm and thereby ensures prolonged polymerase activity in the nucleus of the cell.
As it was observed that PB1-F2 localize to mitochondria, to the cytoplasm and to the nucleus of infected cells, there is a strong indication for a multifunctional role of this viral protein comparable to the influenza virus NS1 protein (Krug et al., 2003; Garcia-Sastre, 2004). Here we show that, besides its reported apoptosis-promoting function, PB1-F2 interacts with the viral PB1 polymerase and determines its nuclear localization late in the infection cycle. As a consequence, a reduced polymerase activity and a small plaque phenotype was observed if cells were infected with virus mutants that lack the PB1-F2 reading frame.
PB1-F2 is translated from an alternate reading frame of the PB1 gene segment. Up to now it was not clear whether initiation at the alternate start codon would need other viral factors or only occurs in the context of virus infection. Here we show that transient transfection of an expression vector for PB1 alone is sufficient to lead to similar amounts of PB1-F2 as expressed in the infected cell, indicating that PB1-F2 expression from the PB1 mRNA exclusively is determined by cellular factors.
Several recent publications show that PB1-F2 is an important virulence factor in animal models; however, the underlying molecular mechanisms have not been addressed in these studies (Zamarin et al., 2006; Conenello et al., 2007; McAuley et al., 2007). In earlier studies PB1-F2 was functionally only analysed with respect to its mitochondrial localization and its capacity to act as an apoptosis enhancer (Chen et al., 2001; Zamarin et al., 2005). Infection with virus mutants that do not express PB1-F2 resulted in a less efficient apoptosis induction; however, this effect was restricted to human immune cells such as monocytes (Chen et al., 2001). As the viral load in mice infected with PB1-F2 knockout viruses was cleared more efficiently than in wt infected animals (Zamarin et al., 2006), it was suggested that PB1-F2 might selectively target immune cells to apoptosis.
Thus, it might appear controversial that we observe functional effects of the protein on viral polymerase activity without any involvement of cell death in epithelial cells. We suggest that PB1-F2 may exert different functions depending on the cell type infected. In immune cells, the proapoptotic effect might be dominant while the polymerase regulating activity may be a modulating event in epithelial cells.
Our data are consistent with other studies reporting that knockout of PB1-F2 does not alter replication characteristics of the mutant viruses in cultured cells. However, it is surprising that we observed a striking small plaque phenotype. While the plaque size reduction is obvious and perfectly correlates with the reduced polymerase activity, these effects may not translate to the overall growth in immortalized cell lines that are optimized for virus replication. It may be hypothesized that the phenotype of the knockout virus may only become apparent under conditions of locally restricted viral spread, while freely released virus may overgrow the inhibitory effects. It is also feasible that polymerase modulation may become more apparent in viruses that carry an N to S mutation at position 66 of the protein that correlates with enhanced pathogenicity (Conenello et al., 2007).
Thus, our data show again that PB1-F2 is not an essential protein but rather appears to exhibit modulating activities on several levels of the viral replication process. In that respect PB1-F2 resembles the non-structural NS1 protein that also has a multitude of different functions (Krug et al., 2003; Garcia-Sastre, 2004). Similar to the NS1, PB1-F2 is only found in the infected cell and not to be incorporated into the virion (U. Schubert, unpubl. obs.). Consistent with this finding, PB1-F2 only was shown to coprecipitate with PB1 that is not associated with PB2, PA or NP. This indicates that PB1-F2 is not a direct cofactor of the RNP complexes.
It has also been observed by others, that PB1 exhibits a prolonged nuclear localization during the late phase of the infection cycle (J. Ortin, Madrid, pers. comm.). It seems that presence of PB1-F2 and binding of the protein to PB1 is a major molecular determinant for this nuclear retention. There are several possible functional consequences of this binding and altered localization. First, PB1-F2 might retain PB1 in the nucleus to ensure efficient RNA replication at late time points when RNP export is already very efficient. This would also explain the enhanced polymerase activity observed in the presence of the protein (Fig. 4). Second, it is also likely that PB1-F2 competes with other factors of the polymerase complex for PB1 binding and may sequester PB1 for other functions in the nucleus. Finally, it is feasible that a strong nuclear localization signal in PB1 helps to redirect PB1-F2 into the nucleus for additional functions.
In any case, the in vivo phenotype of PB1-F2 knockout viruses and even the apoptotic features described can in part be explained by the polymerase-modulating effects of the protein. A more active polymerase would lead to enhanced accumulation of viral RNA that in turn is a trigger for both, enhanced apoptosis induction as well as an increased cytokine responses.
Taken together, we have identified a novel function of the PB1-F2 protein in epithelial cells. The protein is an interaction partner of PB1, determines localization of the polymerase protein and thereby affects viral polymerase activity.
Cell lines and viral infections
Madin-Darby canine kidney (MDCK) cells were grown in minimal-essential medium (MEM) supplemented with 10% heat-inactivated fetal bovine serum (FBS) and antibiotics. HEK293 cells were grown in D-MEM medium supplemented with 10% heat-inactivated FBS and antibiotics. For infection cells were washed with PBS incubated with virus at the indicated multiplicities of infection diluted in PBS/BA (PBS containing 0.2% BSA, 1 mM MgCl2, 0.9 mM CaCl2, 100 U ml−1 penicillin and 0.1 mg ml−1 streptomycin) for 30 min at 37°C. The inoculum was aspirated and cells were incubated with either MEM or D-MEM containing 0.2% BSA and antibiotics. At the given time points supernatants were collected to assess the number of infectious particles (plaque titres) in the samples. Briefly, MDCK cells grown 90% confluent in six-well dishes were washed with PBS and infected with serial dilutions of the supernatants in PBS/BA for 30 min at 37°C. The inoculum was aspirated and cells were incubated with 2 ml MEM/BA (medium containing 0.2% BSA and antibiotics) supplemented with 0.6% Agar (Oxoid), 0.3% DEAE-Dextran (Pharmacia Biotech) and 1.5% NaHCO3 at 37°C, 5% CO2 for 2–3 days. Virus plaques were visualized by staining with neutral-red or Coomassie blue (0.1% Coomassie brilliant blue G-250 in 40% methanol, 10% acetic acid).
Transfections, immunoprecipitations and Western blots
MDCK cells were transfected with Lipofectamine 2000 (Life Technologies) according to a protocol by Basler et al. (2000). Luciferase-reporter gene assays were carried out as described earlier (Ludwig et al., 2001; 2002). For immunoprecipitations and Western blots, cells were lysed on ice with Triton lysis buffer (20 mM Tris-HCl, pH 7.4; 137 mM NaCl; 10% glycerol; 1% Triton X-100; 2 mM EDTA; 50 mM sodium glycerophosphate, 20 mM sodium pyrophosphate; 5 μg ml−1 aprotinin; 5 μg ml−1 leupeptin; 1 mM sodium vanadate and 5 mM benzamidine) for 30 min. Cell lysates were then centrifuged and protein contents in supernatants were estimated employing a protein dye reagent (BIO-RAD laboratories). Equal amounts of protein were separated by SDS-polyacrylamide gel electrophoresis (PAGE) and blotted on nitrocellulose membranes for Western blot or used for immunoprecipitations with antisera against PB1-F2, PB1 and NP coupled to protein A agarose (Roche).
Generation of recombinant influenza viruses
A set of plasmids allowing the rescue of recombinant influenza virus strains A/Puerto Rico/8/34 (PR8) or A/WSN/33 (WSN) was used for generating PB1-F2 knockout mutants. The reverse genetics system includes eight influenza virus RNA-coding transcription plasmids (pHW2000-PB1, -PB2, -PA, -NP, -HA, -NA, -M and -NS) described by Hoffmann et al. for WSN (Hoffmann et al., 2000) and PR8 (Hoffmann et al., 2002). Mutations in the PB1 gene segment were introduced using the Quickchange mutagenesis kit (Stratagene).
To generate the recombinant viruses, 1 μg each of the eight plasmids was transfected into HEK293-MDCK cells (ratio 3:1). The cells were grown in Opti-MEM medium without supplements. Twenty-four hours post transfection fresh media was added. After 24 h incubation the supernatant was removed and used for infection of new MDCK cells. After 3 days incubation the supernatant was harvested and the virus titre was determined on MDCK cells by plaque assays.
The PB1 segments of recombinant wt and mutant viruses were sequenced after reverse transcription-PCR amplification from infected cells to verify the presence and propriety of the desired mutations.
Reagents and antibodies
Staurosporine was purchased from Sigma-Aldrich. The anti-PARP monoclonal antibody was purchased from BD Transduction Laboratories. Mouse monoclonal antibodies directed against the viral NP and M1 were purchased from Serotec. The rabbit polyclonal antiserum against PB1-F2 of PR8 used in this study was described in detail previously (Henklein et al., 2005). These earlier studies as well as own interstrain reactivity analysis indicate that this antiserum predominantly binds to the far C-terminus of the PB1-F2 protein. A polyclonal mouse antiserum against the PB1-F2 of PR8 has been generated at the IMV in Münster and was used in immunofluorescence studies shown in Fig. 4 (lower panel). An antibody against PA was kindly provided by J. Ortin, Madrid, Spain. Antisera against PB1 were kindly provided by A. Garcia-Sastre, New York, USA, or purchased from Santa Cruz Biotechnology. A rabbit antiserum against the PB2 protein was previously described (Carr et al., 2006) and was a generous gift of E. Fodor, Oxford, UK. All sera were used at a 1:1000–1:5000 dilution for Western blotting. An antiserum against ERK2 (Santa Cruz Biotechnologies) was used for loading controls.
Indirect immunofluorescence microscopy
MDCK cells were grown on 15 mm glass plates. When 50% confluence was reached, cells were infected with recombinant PR8 viruses. Thirty minutes post infection, the inoculum was aspirated and medium/BA supplemented with DMSO or inhibitors was added. Eight hours post infection, cells were washed twice with PBS, then fixed for 30 min with 3.7% paraformaldehyde (in PBS) at room temperature. After washing, cells were permeabilized with acetone, washed with PBS and blocked with 10% FBS in PBS for 20 min at 37°C. After blocking, cells were incubated with different antisera and antibodies against the viral NP, PB2, PB1 and PB1-F2 in PBS for 30 min. After further washes, cells were incubated with FITC- or Texas-Red labelled secondary antibodies in PBS for 1 h. Finally, cells were washed and mounted with Vectashield mounting medium with DAPI. Fluorescence was visualized using a Zeiss Axiovert 135 fluorescence microscope.
Yeast two-hybrid assays
For the investigation of interaction partners of PB1-F2 the matchmaker yeast two-hybrid system (Fields and Song, 1989; Phizicky and Fields, 1995) from Clontech was used. Y190 cells were transformed with the vector pAS2-1 containing the Gal4 BD and vector pACT2 containing the Gal4 AD, either fused or unfused with PB1-F2 or PB1(ΔF2). The PB1-F2 knockout cDNA of PB1 [PB1(ΔF2)] was used to prevent the expression of the wt protein. The transformed cells were incubated at 30°C on −Leu –Trp −His medium.
We thank Ervin Fodor, Oxford, UK, Adolfo Garcia Sastre, New York, USA, and Juan Ortin, Madrid, Spain, for providing antisera to PB2, PB1 and PA respectively. We are very grateful to Erich Hoffmann, Memphis, TN, USA. for providing the eight plasmid systems to create rcombinant PR8 and WSN viruses. We also thank Nicole Studtruker, Erlangen, for excellent technical assistance and Thorsten Eierhoff, Münster, for help with the immunofluorescence studies. This work was supported by different grants from the Deutsche Forschungsgemeinschaft (DFG), the DFG Graduate Schools GRK1045 and 1409, the BMBF zoonosis network ‘FluResearchNet’ and the IZKF Grant Lud2/032/06 of the University of Muenster Medical School. The work was further supported by a grant from research network FORINGEN, grant IE-S08T06 from the German Human Genome Research Project, and by grants of the DFG SFB 466 and 643 to U.S.