Human embryonic kidney 293T cells (HEK293T) and HeLa cells were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% fetal bovine serum, streptomycin, and penicillin at 37°C with 5% CO₂.

The expression plasmids [backbone: pFLAG-CMV2 (Sigma–Aldrich) for FLAG-tagged PB1-F2 (PR8) and PB-F2 (HK156) were provided by S.-R. Shih (Chang Gung University, Taiwan) (Chen et al., 2010). Plasmid constructions for FLAG-tagged chimeric PB1-F2 (HP1, HP3, and HP3) were done by overlapping PCR extension cloning. For each construct, a pair of flanking primers and either a pair (HP1 and HP3) or two pairs (HP2) of internal primers were used to make the final product containing chimeric PB1-F2 (Tables 1, 2). Coding region of each chimeric PB1-F2 was cloned into pFLAG-CMV2 at 5′-HindIII and 3′-EcoRI sites. For making expression plasmids for EGFP-PB1-F2, coding region of each PB1-F2 was amplified by PCR from pFLAG-CMV2-based plasmids, and then inserted into pLNFG (a gift of P.-Y. Wu, Academia Sinica, Taiwan) (Pan et al., 2015) at 5′-XhoI and 3′-ClaI. Site-directed mutagenesis of amino acids 68–71 was carried out by primer extension. Briefly, segments of PB1-F2 coding region with amino acids 68–71 mutated were amplified from the pLNFG-PB1-F2 (wild type) using flanking primers and internal primers containing desired mutation. PCR products were cloned into pLNFG at 5′-XhoI and 3′-ClaI sites. Fragments containing coding region of PB1-F2 (PR8-TQDS) and PB1-F2 (HK156-ILVF) were PCR amplified and subcloned from pLNFG-based constructs into pFLAG-CMV2 at 5′-HindIII and 3′-EcoRI sites. Primer design and their uses, as well as the templates are listed in Tables 1, 2. All constructs were sequenced for accuracy.

All plasmids for transfection were purified by using the Plasmid Midi kit (QIAGEN). Transfection of HEK293T and HeLa was performed using Lipofectamine 2000 reagent (Invitrogen) or jetPRIME transfection reagent (Polyplus-transfection) according to the manufacturer’s instruction.

Total cells were lysed in sample buffer (1% sodium dodecyl sulfate, 1.5% β-mercaptoethanol, 4% glycerol, 0.01% bromophenol blue in 50 mM Tris–HCl), resolved by SDS–polyacrylamide gel, and electro-transferred onto polyvinylidene difluoride membranes. After blocked in Tris–buffered saline (50 mM Tris–HCl [pH 7.4], 0.2 M NaCl and 0.1% Tween 20) containing 5% non-fat milk, the membranes were incubated with various antibodies. Membranes were then washed with TBS-T and incubated with horseradish peroxidase-conjugated secondary antibody. Signals were detected with ECL Western blotting detection reagent (GE Healthcare). Commercial primary antibodies used were anti-FLAG (Sigma), anti-GFP (GeneTex), anti-GAPDH (Novus), anti-β-actin (GeneTex) anti-PB1 (GeneTex), anti-NA (GeneTex), and anti-NS1 (Santa Cruz) antibodies. The house-made rabbit polyclonal anti-PB1-F2 antibody was raised against amino acid residues 2–16 (GQEQDTPWILSTGHI) of PB1-F2 (PR8).

HEK293T cells were transfected plasmids expressing FLAG-tagged PB1-F2 for 16 h before cycloheximide treatment (50 μg/ml). At 0, 1, 2, 4, and 8 h post-cycloheximide treatment, cells were harvested and lysed in SDS sample buffer. An additional 0.5 h time point was added in Figure 1. Protein levels of FLAG-PB1-F2 were examined by Western analysis using anti-FLAG antibody. Chemiluminescence intensity for bands specific to PB1-F2 at each time point were captured and measured by a biomolecular imager (ImageQuant LAS 4000). Signals were plotted and analyzed after normalized with β-actin as percentage to time zero using Multi Gauge software (Fujifilm). Band intensity at time zero was set to 100%. The protein half-life (t1/2) for PB1-F2 was calculated from regression formulation taken from three independent experiments. For proteasome inhibition experiments, HEK293T cells were transfected with pFLAG-PB1-F2 or pEGPF-PB1-F2 plasmids. After being transfected for 6–8 h, cells were treated with of MG132 (5 μM) for 16 h to block proteasome function. Chemiluminescence intensity for bands specific to EGFP-PB1-F2 and GAPDH were captured and measured by a biomolecular imager (ImageQuant LAS 4000). Signals with or without MG132 treatment were compared after normalized with GAPDH. Intensity of EGFP-PB1-F2 (MG132 (-)) after normalized with GAPDH was set as 1x.

HeLa cells were grown on coverslips before being transfected with individual EGFP-tagged PB1-F2 expression plasmid. At 16 h post-transfection, cells were stained with 250 nM reduced rosamine MitoTracker Orange (M7511, Invitrogen) in HBSS (Gibco), fixed with 4% formaldehyde and then permeabilized with 0.5% Triton X-100. Nuclei were stained with DAPI (Sigma). Confocal microscopy was performed with a Zeiss LSM700 microscope. The images were acquired using Plan-Apochromat 1.4 Oil DIC M27 objectives with power of 63× and 100× with appropriate filter combinations. Ten fields were randomly selected to calculate the percentage of the EGFP-PB1-F2 that colocalized with MitoTracker signals.

3 × 10⁵ HEK293T cells were co-transfected with individual EGFP-tagged PB1-F2 expressing plasmid and p125-luc IFN-β luciferase reporter (a gift of L.-H. Hwang, National Yang-Ming University, Taiwan) in 12-well plates. 24 h post-transfection, cells were infected with Sendai virus (SeV) H4 (50 HAU) to induce IFN-β signaling pathway. Cells were harvested 24 h post-infection for luciferase activity using Luciferase Assay System (Promega). Two independent experiments were performed in triplicate for each sample.

Recombinant A/Puerto Rico/8/1934 carrying PB1-F2 (68T, 69Q, 70E, 71S) was made using an eight-plasmids reverse genetics system (Hoffmann et al., 2000). Plasmids were kindly provided by Dr. M.-S. Ho (Academia Sinica, Taiwan). Coding region of amino acid residues 68–71 was mutated from ILVF to TQES without affecting the coding of PB1 in plasmid pHW192 using overlapping PCR extension cloning. Primer design and their uses, as well as the templates are listed in Tables 1, 2. Resulting plasmid pHW192 or pHW192-(PB1-F2 TQES) was transfected with seven other pHW19x plasmids into HEK293 cells using jetPRIME transfection reagent to generate PR8 wild-type virus or PR8 (PB1-F2-TQES) mutant virus. At 16 h post-transfection, medium was replaced with DMEM without serum. Additional 32 h later, the supernatant was collected and used to inoculate A549 cells. A549 cells were harvested 24 h post-infection for Western analysis using antibodies as indicated.

To compare the stability of PB1-F2 variants, we measured protein half-lives of PB1-F2 from A/Hong Kong/156/1997 (H5N1) (herein PB1-F2 (HK156)) and A/Puerto Rico/8/1934 (H1N1) (herein PB1-F2 (PR8)) by cycloheximide chase assay. The plasmid expressing either FLAG-PB1-F2 (HK156) or FLAG-PB1-F2 (PR8) was transfected to HEK293 cells for 16 h before cycloheximide was added to stop translation. Cells were harvested at 0, 0.5, 1, 2, 4, and 8 h post-cycloheximide treatment for Western analysis. Intensity for bands specific to PB1-F2 at each time point was measured by densitometer and plotted. A representative Western analysis (Figure 1A) and a graph of the mean band intensities and regression functions from three sets of experiments were shown in Figure 1B. We found that the estimated half-life of PB1-F2 (PR8) calculated by regression analysis is 6.30 h, while that of PB1-F2 (HK156) is 2.14 h, suggesting that PB1-F2 (HK156) has a more rapid turnover rate. The longer protein half-life of PB1-F2 (PR8) may contribute to its higher protein production at zero time point since protein was allowed to accumulate for 16 h before cycloheximide was added. However, we cannot rule out the possibility that transcript of FLAG-PB1-F2 (PR8) can be translated more efficiently than that of FLAG-PB1-F2 (HK156).

Since the protein half-life of PB1-F2 (HK156) is three-times shorter as that of PB1-F2 (PR8), we asked if there are particular amino acid residues responsible for this result. Alignment between two proteins reveals 57.8% identity and no apparent domains may contribute to the difference. We therefore constructed plasmids expressing chimeric proteins and measured their half-lives. Used plasmid expressing PB1-F2 (HK156) as the base, coding sequence for the first, second or last third of the protein was replaced with that of PB1-F2 (PR8), resulting pFLAG-HP1, pFLAG-HP2 and pFLAG-HP3, respectively. All three plasmids expressing chimeric FLAG-PB1-F2 were transfected into HEK293 cells. Cycloheximide chasing assay was performed as shown in Figure 1. We found the expression patterns of PB1-F2 (HP1) and PB1-F2 (HP2) are similar to that of PB1-F2 (HK156), while that of PB1-F2 (HP3) resembled the expression pattern of PB1-F2 (PR8) (Figure 2A). The results indicated the last 1/3 of protein containing residues determining protein half-life of PB1-F2. Within this region, sequence of amino acid 68–71 is completely different between PB1-F2 (HK156) and PB1-F2 (PR8). We therefore constructed two plasmids to produce PB1-F2 with these four amino acids swapped. Resulting plasmids pFLAG-HK156-ILVF and pFLAG-PR8-TQDS were transfected into HEK293 cells. Cycloheximide chasing assay was performed as in Figure 2A. We found that PB1-F2 (HK156-ILVF) now had an expression pattern similar to that of PB-F2 (PR8), while PB1-F2 (PR8-TQDS)’s resembled that of PB1-F2 (HK156), suggesting that amino acids 68–71 are the key residues affecting protein half-life of PB1-F2 (Figure 2B). A sequence alignment between the last third of PB1-F2 (HK156) and PB1-F2 (PR8), as well as the sequence of amino acids 68–71 in different constructs are shown in Figure 2C. It worth to note that PB1-F2 (HP2) may have a faster degradation rate than PB1-F2 (HK156), suggesting the middle third or PB1-F2 may contain residues contributing to its protein stability.

Because one factor affecting protein turnover is how fast the protein is degraded, so we were interested in if variants and mutants of PB1-F2 are subjected to different degradation pathway. Others have shown that treatment of lactacystin, a proteasome inhibitor, increased the protein level of a PB1-F2 mutant but not its wild-type counter part (Schmolke et al., 2011), therefore we examined if various PB1-F2 proteins have equal sensitivities to proteasome-mediated degradation. Using the plasmids expressing FLAG-PB1-F2 of wild-type (PR8 and HK156) or chimeric (HP1 and HP3) proteins, we transfected HEK293 for 6 h before MG132, a proteasome inhibitor, was added. Cells were harvested and analyzed for protein levels of PB1-F2 in the presence or absence of MG132 by Western analysis. We found that MG132 treatment only increased levels of PB1-F2 of HK156 and HP1, but not those PR8 and HP3, indicating that the last 1/3 of PB1-F2 regulated proteasome-mediated PB1-F2 degradation (Figure 3A). Similarly we examined if amino acids 68–71 are responsible for this observation. Plasmids expressing EGFP-PB1-F2 were used to transfect HEK293 cells and experiments were carried out as in Figure 3A. We found that swapping of amino acids 68–71 changed its sensitivity to MG132. The increasing level by MG132 treatment for wild-type PB1-F2 (HK156) dropped from 2.56× to 1.12× when it came the ILVF mutant. On the other hand, the increasing level by MG132 treatment for wild-type PB1-F2 (PR8) increased from 0.99× to 2.91× when it came the TQDS mutant indicating that T68, Q69, D70 and S71 prompted PB1-F2 to proteasome degradation (Figure 3B). These results suggested that higher expression level of PR8-F2 (PR8) may be due to its resistance to proteasome degradation, and amino acids 68–71 (ILVF) is important for this resistance.

When PB1-F2 (PR8) was first identified, it has a mitochondrial localization (Chen et al., 2001). Subsequent studies, on the other hand, demonstrated that PB1-F2 (HK156) localized mainly to the cytoplasm and nucleus (Chen et al., 2010). An optimal mitochondrial targeting sequence was identified to region 65–87 (Gibbs et al., 2003). Therefore we asked if swapping of amino acids 68–71 changes subcellular localization of PB1-F2. For this, plasmids expressing EGFP fusion protein with PB1-F2 of PR8, HK156, PR8-TQDS, or HK156-ILVF were constructed and transfected into HeLa cells. Cells were stained with MitoTracker for mitochondria and DAPI for nuclei. The fluorescence images showed that PB1-F2 (PR8) is localized to mitochondria, while PB1-F2 (HK156) is localized to cytosol and nuclei as predicted. However, amino acids 68–71 swapping between two proteins changed their cellular localization drastically. PB1-F2 (PR8) lost its mitochondrial localization and dispersed in cytosol and nucleus when amino acids 68–71 were changed from ILVF to TQDS (Figure 4A, PR8-TQDS); conversely when PB1-F2 (HK156) grained mitochondrial localization when amino acids 68–71 were switched from TQDS to ILVF (Figure 4A, HK156-ILVF). A quantitation of mitochondrial localization was shown in Figure 4B. These results indicated that amino acids 68–71 not only affect PB1-F2 protein stability, but also its cellular localization.

It has been shown that PB1-F2 can modulate type I interferon (IFN) production through interfering RIG-I-mediated signaling at MAVS level (Varga et al., 2011). We asked if swapping of amino acids 68–71 can change this ability. Because NS1 of influenza virus is a potent inhibitor of RIG-I-mediated signaling pathway, we used SeV to induced IFN-β signaling cascade, which could be reflected by IFN-β reporter (p125-luc) activity. Plasmids expressing PB1-F2 was cotransfected with p125-luc reporter into HEK293 cells. 24 h post-transfection, cells were infected with SeV to induce IFN-β response. Infected cells were harvested 24 h post-infection for p125-luc reporter assay. SeV infection induced about 40-fold of p125-luc activity compared to non-infection control. PB1-F2 (PR8) caused an 81% drop of SeV-induced IFN-β promoter activity. On the other hand PB1-F2 (PR8-TQDS) only caused a 33% reduction. This is consistent with the decreased protein stability and reduced mitochondrial localization of this protein. Changing amino acids 68–71 from TQDS to ILVF further enhanced the ability of PB1-F2 (HK156) to inhibit p125-luc from 58 to 31% (Figure 5). This is consistent with the increased protein stability and mitochondrial localization of PB1-F2 (HK156-ILVF). These results indicated that amino acids 68–71 are key residues that affecting protein half-live and the functions of PB1-F2. Although PB1-F2 (HK156) did not show mitochondrial localization, it reduced SeV-mediated p125-luc activity to 58%. This may due to its inhibitory effect on NF-κB (Reis and McCauley, 2013), whose binding site also presents in the p125-luc promoter.

It is important to know if amino acid residues 68–71 effect PB1-F2 production during viral infection. We used PR8 virus as our model virus to address this question. Because PB1-F2 is encoded by the +1 ORF in the PB1 gene (Chen et al., 2001), we mutated the coding sequence from ILVF to TQES instead of TQDS to avoid changing the coding of PB1. Recombinant PR8 carrying PB1-F2 (T68, Q69, E70, S71) was made using an eight-plasmids reverse genetics system resulting the PB8 (PB1-F2-TQES) mutant virus in HEK293 cells. The supernatant was collected and used to inoculate A549 cells. Wild-type PR8 generated by reverse genetics served as a control. 24 h post-infection, cells were harvested for Western analysis. While PB1-F2 can be readily detected in A549 infected with PR8 virus, it was not detected in cells infected with PB8 (PB1-F2-TQES) mutant virus (Figure 6). This discrepancy is unlikely due to unequal inoculation since productions of other viral proteins, PB1, NA, and NS1 were similar in cells infected with different viruses. This result confirms our studies in vitro that amino acid residues 68–71 play an important role in PB1-F2 stability.