|Home | About | Journals | Submit | Contact Us | Français|
The influenza A virus PB1-F2 protein has been implicated as a virulence factor, but the mechanism by which it enhances pathogenicity is not understood. The PB1 gene segment of the H1N1 swine-origin influenza virus pandemic strain codes for a truncated PB1-F2 protein which terminates after 11 amino acids but could acquire the full-length form by mutation or reassortment. It is therefore important to understand the function and impact of this protein. We systematically assessed the effect that PB1-F2 expression has on viral polymerase activity, accumulation and localization of PB1, and replication in vitro and in mice. We used both the laboratory strain PR8 and a set of viruses engineered to study clinically relevant PB1-F2 proteins. PB1-F2 expression had modest effects on polymerase activity, PB1 accumulation, and replication that were cell type and virus strain dependent. Disruption of the PB1-F2 reading frame in a recent, seasonal H3N2 influenza virus strain did not affect these parameters, suggesting that this is not a universal function of the protein. Disruption of PB1-F2 expression in several backgrounds or expression of PB1-F2 from the 1918 pandemic strain or a 1956 H1N1 strain had no effect on viral lung loads in mice. Alternate mechanisms besides alterations to replication are likely responsible for the enhanced virulence in mammalian hosts attributed to PB1-F2 in previous studies.
Seasonal influenza is responsible for significant morbidity and mortality worldwide. In the 1990s, it was estimated to kill 36,000 persons annually in the United States alone and 250,000 to 500,000 persons in the developed world, although hospitalization rates and mortality figures varied considerably from season to season based on the circulating strains (19, 20). Influenza A viruses also have the capability to cause a pandemic if they are sufficiently novel. Strains may emerge whole or in part from animal reservoirs and establish long-term (years to decades) zoonotic lineages in humans (23). The most striking example of this phenomenon occurred in 1918, when an avian virus of the H1N1 subtype crossed the species barrier and established related lineages in two mammalian hosts, swine and humans (16). This pandemic is thought to have killed more than 40 million persons worldwide. In 2009, a novel H1N1 influenza virus of swine origin (H1N1 S-OIV) emerged and is now causing the first pandemic the world has seen in more than 40 years (14). Because of the history of pandemic influenza and the current circulation of a novel pandemic strain, there is intense interest and urgency in understanding viral factors that allow expression of disease in humans.
One such virulence factor is the influenza A virus protein PB1-F2 (8). This small (87 to 90 amino acids), 11th gene product was discovered in 2001 in a search for CD8+ epitopes in alternative reading frames of influenza A virus genes (2). It is encoded in the +1 reading frame of the PB1 gene segment and is translated from an AUG codon downstream of the PB1 start site, probably accessed through leaky ribosomal scanning. It has been shown to contribute to virulence both directly and indirectly, through modulation of responses to bacteria (3, 11). The exact mechanism(s) through which virulence is increased by PB1-F2 expression, however, is not yet understood. Three effects of PB1-F2 expression have been suggested so far. It has been demonstrated to cause cell death in some cell types (2, 5), it has been shown to induce inflammation by recruitment of inflammatory cells in mice (11), and it has been determined to bind PB1 and to increase the activity of the influenza virus polymerase in vitro (10).
The function of the PB1-F2 protein in the life cycle of influenza virus is as unclear as its precise role in virulence. Given that almost all avian influenza virus strains express a full-length PB1-F2 protein (27), it is likely to contribute to survival or transmission in the natural avian host. After introduction of viruses into mammalian hosts such as humans or swine, however, the protein often becomes truncated during adaptation, implying that any effects it might induce are not necessary for virus viability and transmission in these hosts. The 1918 H1N1 virus had a full-length PB1-F2 protein, which has been demonstrated to contribute to virulence in mice (3, 11). During the evolution of H1N1 viruses in humans over time, a stop codon at position 58 in the PB1-F2 amino acid sequence appeared around 1950 and has been retained in the human H1N1 lineage since its reemergence in 1977. Similarly, multiple swine lineages of influenza A virus have had truncations appear at different positions, including position 58, such that 25% of swine PB1-F2 sequences in GenBank lack the C-terminal portion of the protein (27). The H3N2 lineage of viruses in humans has retained a full-length PB1-F2 protein since the introduction of a new PB1 gene segment during the 1968 pandemic, although considerable variation in sequence has occurred during evolution since that time. It is tempting to map these differences in PB1-F2 expression onto patterns of human excess mortality over time, since higher mortality was associated with H1N1 epidemics in the 1930s and 1940s than has been seen since and more excess mortality occurred in recent years with H3N2 viruses than with either H1N1 or influenza B viruses (reviewed in reference 12). Differences in primary virulence or the association with bacteria mediated by PB1-F2 expression could be at least partly responsible for these observed epidemiologic trends.
A recent paper from Wise et al. has shown that a 12th influenza A virus gene product, N40, is also expressed from the PB1 gene segment (24). A delicate balance between PB1, PB1-F2, and N40 appears to be in place. Polymerase activity measured by an in vitro assay was affected by changes in this balance, suggesting a potential importance for replication. If these differences translate to differences in replication, then this could be a key factor in virulence in the host. However, to this point, most studies have utilized a single laboratory variant of influenza A virus, A/Puerto Rico/8/34 (H1N1; PR8), in a limited set of cell types, in assays performed in vitro. We undertook this study to determine the relevance of potential changes in replication mediated by PB1-F2 expression, utilizing several different epidemiologically important virus strains. We found that the effects on polymerase activity and in vitro replication efficiency were virus and cell type specific and did not mediate changes in viral lung load in animals.
A set of plasmids were generated on the pHW2000 backbone as described previously (7), carrying the PB1 gene segment of PR8, A/Vietnam/1203/04 (H5N1), or A/Wuhan/359/95 (H3N2). In each of these backgrounds, the open reading frame for PB1-F2 was disrupted by altering the start codon (T120C mutation by PB1 numbering) so translation will not initiate and inserting a stop codon after 11 residues (C153G) to ensure a complete knockout (11, 26). The PR8 PB1-F2 sequence was further altered by QuikChange site-directed mutagenesis (Stratagene), as described previously (11), so that the protein expressed was identical to that of either the A/Brevig Mission/1/18 (H1N1; 1918 PB1-F2) or A/Beijing/11/56 (H1N1; Beij PB1-F2) virus. In no case did these mutations in the PB1-F2 reading frame cause nonsynonymous mutations in the PB1 reading frame. The N40 start codon (24) was intact in all plasmids. These eight PB1 plasmids were then incorporated into a corresponding set of eight viruses on a seven-gene PR8 backbone (7:1) by reverse genetics as described previously (6). Resulting viruses were rescued by one passage in MDCK cells and then propagated a single time in eggs for stocks to be used in these studies. All viruses were fully sequenced to ensure that no inadvertent mutations occurred during virus rescue and propagation and then were characterized in tissue culture and eggs as previously described (21).
Subconfluent monolayers of 293T, HEK293, and A549 cells (~7.5 × 105 cells/ml in each well of 12-well tissue culture dishes) were transfected (Mirus Bio) with 2 μg firefly luciferase reporter plasmid (enhanced green fluorescent protein open reading frame in pHW72-EGFP replaced with the luciferase gene), 1 μg of the control pRL-TK vector, which expresses the Renilla luciferase, and 1 μg each of PB1, PB2, PA, and NP plasmids. The PB2, PA, and NP plasmids for PR8 and A/Wuhan/359/95 (H3N2) were the kind gift of Robert Webster (St. Jude Children's Research Hospital) (9, 17), while the PB1 plasmids were those constructed for this study. At 24 h posttransfection, cell extracts were prepared in 500 μl lysis buffer, and luciferase activity was measured using a dual-luciferase assay system (Promega) and a BD Monolight 300 luminometer (BD Biosciences). A Renilla luciferase construct was used as an internal control to normalize relative luciferase activity, and results were expressed as the relative activity compared to that in the reaction containing the respective wild-type (WT) PB1 plasmid for each experiment. The identities of the HEK293 and 293T cells were confirmed by selective sequencing and identification of the large T antigen in the 293T cells but not the HEK293 cells.
MDCK cells were grown on glass coverslips and infected with virus as indicated. The cells were then washed with warm phosphate-buffered saline (PBS) at the indicated time points postinfection and fixed with 4% formaldehyde in PBS at 37°C for 30 min in a humid environment. After being washed, cells were permeabilized in 0.1 to 0.2% Triton X-100 for 10 min and washed again, 200 μl Image-iT FX signal enhancer (Invitrogen) was added, and the cells were further incubated at 37°C for 30 min. Cells were then washed thrice with warm PBS and relevant primary antibody diluted 1:1,000 in staining buffer (0.1% NaN3 plus 1% bovine serum albumin in PBS) for 1 h at 37°C. After being washed, cells were incubated with the relevant secondary antibody at a 1:100 dilution in staining buffer at 37°C for 1 h. After a final wash step with warmed PBS, cells were mounted with ProLong Gold antifade reagent (Invitrogen) containing DAPI (4′,6-diamino-2-phenylindole) for nuclear staining. Fluorescence was visualized with a Nikon C1Si confocal lens on a TE2000 microscope. The PB1 protein in infected cells was detected by anti-PB1 goat polyclonal antibody against the N terminus (clone VK-20; Santa Cruz Biotechnology) and by Alexa Fluor 555-coupled donkey anti-goat antibody (Invitrogen). PB1-F2 in infected cells was detected by an anti-PB1-F2 rabbit polyclonal antibody which recognizes the N termini of H1N1 virus-derived PB1-F2 proteins (a gift from Jon Yewdell, NIH) and by Alexa Fluor 488-conjugated donkey anti-rabbit antibody. Semiquantitative analysis was conducted utilizing NIS Elements software (v3.0). The location and intensity of PB1 staining were identified through masking on DAPI-positive (DAPI+; nuclei) events and DAPI− (outside nuclei) areas to define total versus cytoplasmic accumulation. Colocalization of PB1 and PB1-F2 was determined by quantitation of both PB1+ and PB1-F2+ fluorescence intensities above the threshold of the negative control background noise levels and determined to intersect over given areas. Analyses were conducted on at least five randomly selected fields per sample.
Viral replication was assayed in vitro in confluent MDCK cells by the method of Reed and Muench (15) to determine the 50% tissue culture infective dose (TCID50) as described previously (13). Plaque area was determined using at least 50 plaques per virus on MDCK cells overlaid with soft agar as described previously (11). For in vivo experiments, 8-week-old female BALB/c mice were infected intranasally under 2.5% isoflurane anesthesia with 100 TCID50 of virus in the BL2 area of the Animal Resources Center of St. Jude Children's Research Hospital under an Institutional Animal Care and Use Committee-approved protocol. Viral lung load is expressed as the TCID50 per ml of supernatant after lungs were harvested, resuspended in 500 μl of sterile PBS, and homogenized.
Comparisons of viral lung titer, plaque size, polymerase activity, colocalization, and PB1 expression between groups were done using analysis of variance (ANOVA) for multiple comparisons and Student's t test for matched, single comparisons. P values of <0.05 were considered significant for these comparisons. SigmaStat for Windows (v 3.11; SysStat Software, Inc.) was utilized for all statistical analyses.
Previous studies suggested that PB1-F2 expression altered polymerase activity in an in vitro minigenome system using the PR8 strain and 293 epithelial cells (10, 24). We sought to confirm these observations and to extend them to other viruses and cell types. In a four-plasmid system based on PR8 PB2, PB1, PA, and NP and using luciferase activity as a readout for polymerase activity, we studied the contributions of differences in PB1-F2 to polymerase activity. We first altered the PR8 PB1 so it would not express PB1-F2 (ΔPB1-F2) or would instead express PB1-F2 of the 1918 pandemic strain (1918 PB1-F2) or of a 1956 H1N1 strain (Beij PB1-F2) that is truncated at 57 amino acids. We also removed the capacity to express PB1-F2 from two other PB1 plasmids, one derived from a highly pathogenic avian influenza virus of the H5N1 subtype (H5N1 ΔPB1-F2) and the other from a human H3N2 strain (H3N2 ΔPB1-F2). We compared the effects of these changes to those of the WT parent plasmids in a constant background.
In 293T cells, inclusion of the ΔPB1-F2 construct led to significantly lower polymerase activity than that of WT PR8, as did substitution of the truncated Beij PB1-F2 (Fig. (Fig.1A).1A). However, expression of the 1918 PB1-F2, which is linked to increased virulence of the virus (11), also decreased polymerase activity. Knockout of PB1-F2 in an H5N1 PB1 background significantly decreased activity, but removal of PB1-F2 expression from an H3N2 PB1 did not. To ensure that the latter effect was not due to a mismatch of the H3N2 PB1 to PR8 PB2, PA, or NP, we repeated the assay with all four plasmids derived from the parent H3N2 virus (H3N2 PB1/H3N2) and achieved the same results (Fig. (Fig.1A).1A). We next studied the same set of plasmids in a potentially more relevant cell type, the human lung cell line A549. The finding of decreased activity when expression of PB1-F2 was prevented or the truncated form was introduced did not hold true in this cell type; activity actually increased with the ΔPB1-F2 plasmid present (Fig. (Fig.1B).1B). The phenotype seen in 293T cells was preserved using the H5 ΔPB1-F2 plasmid, but as before, no differences were seen on the H3N2 background. Since previous reports had used different 293-derived cells (10, 24), we also tested the plasmids in HEK293 cells and found no differences in polymerase activity for any comparisons using different PB1 plasmids (data not shown).
Mazur et al. postulated that the increased polymerase activity derived from PB1-F2 expression in vitro was due to an interaction of the C terminus of the protein with PB1 (10). To determine whether this explained the PB1-F2-related differences in polymerase activity in the minigenome assay, we created viruses on a seven-gene-segment PR8 background, expressing the various PB1/PB1-F2 combinations described earlier, and infected MDCK cells. After 24 h, the PB1-F2 from PR8 was shown to be located predominantly in the cytoplasm of most cells, with nuclear localization in a minority. Many cells showed a distinct pattern suggesting mitochondrial localization (Fig. (Fig.2C).2C). The 1918 PB1-F2 was distributed in a similar pattern, but the Beij PB1-F2, which lacks the C-terminal mitochondrial targeting sequence, was found diffusely throughout the nucleus and cytoplasm. When both PB1 and PB1-F2 were visualized together, less than 10% of PB1-F2 was found to colocalize with PB1 (Fig. 2A and B). Colocalization was seen only with PR8 and the 1918 PB1-F2 and was seen to occur throughout the cell and not to be confined to the nuclear phase. The Beij PB1-F2, which lacks the C terminus required for PB1 binding (10), was not found to colocalize. These data support and extend the concept that PB1 colocalization with PB1-F2 occurs during infection and requires expression of full-length PB1-F2, which implies that a binding interaction occurs.
To determine whether the potential for PB1 binding, as documented by Mazur et al. (10) and suggested by our colocalization experiments, had an effect on the transition of PB1 from the nucleus to the cytoplasm, a time course study of PB1 expression and localization was done. Nuclear and total cellular PB1 peaked at 6 h postinfection when WT PR8 PB1-F2 was expressed, and then it decreased (Fig. 3A and B). In cells infected with the ΔPB1-F2/PR8 virus, PB1 expression followed a very similar time course early but peaked later, at 8 h postinfection. With both the 1918 PB1-F2/PR8 and Beij PB1-F2/PR8 viruses, nuclear and total cellular PB1 proteins were delayed relative to those in PR8 and were significantly lower at 8 h. Time points between 8 and 24 h were not assessed, so it is not clear whether PB1 expression from these viruses ever reached the maximum WT levels within this period.
Cytoplasmic PB1 accumulation directly mirrored total PB1 expression (Fig. (Fig.3C);3C); there was no apparent delay in cytoplasmic localization caused by expression of WT PR8 PB1-F2 (Fig. (Fig.3D).3D). In fact, a delayed appearance of PB1 in the cytoplasm was seen when either the 1918 PB1-F2 or Beij PB1-F2 was expressed (Fig. 3C and D), arguing that the kinetics of PB1 synthesis, accumulation, and migration are not due to binding of PB1-F2 to PB1 after both proteins have been translated. Expression of N40 is not predicted to differ between the WT PR8, 1918 PB1-F2/PR8, and Beij PB1-F2/PR8 viruses (24), so posttranslational interactions with this gene product are an unlikely explanation. It is more likely that transcriptional or translational differences affecting total cellular PB1 synthesis are responsible for the disparate outcomes. The H5N1 PB1 is synthesized more rapidly than other PB1 proteins assayed, but no significant differences in total PB1 accumulation at 8 h or in the transition to the cytoplasm were seen by disruption of the PB1-F2 reading frame in either the H5N1 PB1 or H3N2 PB1 background (data not shown).
The PB1-F2 protein has been shown to contribute to virulence (3, 11, 26). If the PB1-F2-PB1 interaction and differences in polymerase activity make a meaningful contribution, this would be expected to manifest itself as differences in replication and viral titer in the host. We tested the multistep replication efficiency of this panel of viruses in MDCK cells at a multiplicity of infection (MOI) of 0.001. In the H1N1 PR8 background, an advantage in replication could be seen for the two viruses expressing a full-length PB1-F2, WT PR8 and 1918 PB1-F2/PR8, compared to the two viruses lacking a full-length PB1-F2, either a virus that does not express PB1-F2 (PR8 ΔPB1-F2) or one truncated after 56 amino acids (Beij PB1-F2/PR8) (Fig. (Fig.4A).4A). Of note, this advantage was at early time points only, since by 48 h titers were indistinguishable, did not correlate with the differences in polymerase activity established earlier (Fig. (Fig.1A),1A), and were dependent on infection at a low MOI. In earlier studies at a higher MOI (11; data not shown), no differences in replication could be appreciated. In the H5N1 and H3N2 PB1 backgrounds, disruption of PB1-F2 expression did not alter in vitro replication (Fig. 4B and C), indicating that this effect is virus strain specific.
Plaque size did not follow the same pattern as in vitro replication, indicating that the mechanism for expansion of plaques during soft agar overlay differs from that responsible for total viral titer in liquid medium. Plaque size was larger with expression of both 1918 PB1-F2 and Beij PB1-F2 but smaller after deletion of the H5N1 PB1-F2 (Fig. (Fig.55).
In mice, no differences in viral lung load could be appreciated in any background at any time point with disruption or substitution of PB1-F2 (Fig. (Fig.6).6). From these data, we conclude that the modest and variable effects on polymerase activity and replication in vitro that can be attributed to PB1-F2 do not translate to differences in viral load in vivo and are unlikely to contribute in a meaningful way to virulence.
A few times every century, a novel influenza A virus strain emerges from the animal reservoirs of the world and causes a pandemic. Understanding the basis of virulence of these strains is an important and as yet unrealized goal. In order to prioritize resources, we need to know which specific virulence factors are critical and understand how they contribute to pathogenicity, which will tell us which strains are important and deserving of monitoring and further study. Few molecular signatures are known with any specificity that can be utilized in this manner. PB1-F2 may be one such virulence factor, but the mechanism by which it might enhance disease is not currently understood. It is clear that it was important in the pandemic of 1918 (11) and likely contributed to the virulence of other pandemic strains which expressed full-length PB1-F2 proteins. We demonstrate here that the PB1-F2 protein's potential effects on virulence are unlikely to be mediated by enhanced replication. Although modest effects are seen in some in vitro assays with some strains under some conditions, these do not impact total virus load in an animal model and do not map to previously determined differences in pathogenicity (3, 11).
All influenza A viruses which infect humans are zoonoses derived ultimately from the bird reservoir. Full-length PB1-F2 proteins are expressed by nearly all avian influenza virus strains (27) but often become truncated during adaptation in mammalian hosts. This pattern implies that there is some evolutionary utility for the protein in birds but not in mammals. In the three pandemics in the 20th century, the only two gene segments taken from the avian reservoir in each case were the hemagglutinin and the PB1 segments (18), and each PB1 gene segment encoded a full-length PB1-F2. The H1N1-lineage PB1-F2 became truncated around 1948 to 1950, losing the C-terminal region that contains the mitochondrial targeting sequence (2), can mediate PB1 binding (10), and promotes inflammation (11). The currently circulating H3N2-lineage PB1-F2 proteins are still full length but may have undergone evolutionary changes during adaptation. The hypothesis that unadapted PB1-F2 proteins derived from the avian reservoir were important for virulence of those pandemic strains is compelling.
A different situation confronts us now with the emergence of the H1N1 S-OIV in 2009. This strain is a reassortant virus comprised of genes derived from avian, swine, and human viruses, but it does not express a full-length PB1-F2 protein (22). At present, the H1N1 S-OIV is not highly pathogenic like the H5N1 strains recently making incursions into humans in Eurasia and Africa or like the 1918 H1N1 pandemic strain (4). The possibility that the current pandemic strain could reassort with a circulating strain that expresses a full-length PB1-F2 protein, such as the seasonal H3N2 strains, and thus gain virulence makes understanding PB1-F2 critical. In this study, disruption of PB1-F2 expression in the mouse-adapted laboratory strain PR8 decreased polymerase activity in vitro, which correlated with enhanced replication in vitro but not with increased viral lung loads in mice. When the PR8 PB1-F2 was replaced with the PB1-F2 from the 1918 pandemic strain, however, polymerase activity was similarly diminished, but the in vitro replication advantage was retained. Disruption of PB1-F2 expression in either the H5N1 or H3N2 background had no effect on replication in vitro or in vivo. In earlier studies, disruption of PB1-F2 expression in a PR8 background had no effect on replication (at a higher MOI) in multiple cell types, including MDCK, A549, MDBK, HeLa, and A301 cells and eggs (10, 24). Thus, it is unlikely that potential effects of PB1-F2 on polymerase activity are meaningful for virulence.
Lack of the C terminus of the PB1-F2 protein, where the mitochondrial targeting sequence is located, had effects on both mitochondrial localization and PB1 binding. This is likely to affect mechanisms that depend on mitochondrial targeting, such as cell death if mediated by mitochondrial interactions (1, 5), and may affect the immunostimulatory capabilities of the protein, which have also been mapped to the C terminus (11). Since the potentially nonfunctional, truncated PB1-F2 protein is expressed in this scenario, however, the N40 protein is unlikely to be overexpressed with its resulting effects on PB1 (24). We did not see differences in polymerase activity or replication, in cells or animals, potentially attributable to N40 in comparing the ΔPB1-F2 virus to the truncated Beij PB1-F2-carrying virus, but there was a modest delay and decrease in both total and cytoplasmic PB1 accumulation. Both the H5N1 and seasonal H3N2 viruses have full-length PB1-F2 proteins including a mitochondrial targeting sequence, but disruption of PB1-F2 expression did not affect replication in either background.
If PB1-F2 does not alter viral load in mammalian hosts, how does it contribute to virulence in mice or humans? One possibility is that its ability to cause cell death (2) contributes to pathogenesis by causing epithelial cell damage or killing immune effector cells. However, cell death in this context has also been shown to be cell type specific, differs in the context of the whole virus versus expression of PB1-F2 alone, and has been studied only with the laboratory strain PR8. Cell death was engendered in MDBK and HeLa cells through exposure to the PR8 PB1-F2 protein alone (2) but was not seen in Vero, HeLa, or MDCK cells (25). Using whole virus, cell death was observed in U937 and primary human monocytes but was not seen in MDCK, MDBK, A549, or HeLa cells (2). More relevant viruses than the laboratory strain PR8 have not yet been assessed in this manner, and this mechanism has not been studied in vivo. A second possibility is that the reported immunostimulatory properties of the protein are responsible for virulence. Enhanced inflammation in mouse models has been attributed to expression of PB1-F2 proteins derived from several backgrounds, including PR8, H5N1, and the 1918 pandemic strain (3, 11). The mechanism of this contribution to virulence should be worked out in several relevant clinical strains, including the seasonal H3N2 viruses with which the H1N1 S-OIV might reassort.
The work described here was supported by PHS grant AI-66349 and by ALSAC.
We thank Samuel Connell and Jennifer Peters of the Cellular Imaging Shared Resource for technical assistance with confocal microscopy and helpful discussions about the data.
Published ahead of print on 14 October 2009.