PA from an H5N1 highly pathogenic avian influenza virus activates viral transcription and replication and induces apoptosis and interferon expression at an early stage of infection

Background Although gene exchange is not likely to occur freely, reassortment between the H5N1 highly pathogenic avian influenza virus (HPAIV) and currently circulating human viruses is a serious concern. The PA polymerase subunit of H5N1 HPAIV was recently reported to activate the influenza replicon activity. Methods The replicon activities of PR8 and WSN strains (H1N1) of influenza containing PA from HPAIV A/Cambodia/P0322095/2005 (H5N1) and the activity of the chimeric RNA polymerase were analyzed. A reassortant WSN virus containing the H5N1 Cambodia PA (C-PA) was then reconstituted and its growth in cells and pathogenicity in mice examined. The interferon promoter, TUNEL, and caspase 3, 8, and 9 activities of C-PA-infected cells were compared with those of WSN-infected cells. Results The activity of the chimeric RNA polymerase was slightly higher than that of WSN, and C-PA replicated better than WSN in cells. However, the multi-step growth of C-PA and its pathogenicity in mice were lower than those of WSN. The interferon promoter, TUNEL, and caspase 3, 8, and 9 activities were strongly induced in early infection in C-PA-infected cells but not in WSN-infected cells. Conclusions Apoptosis and interferon were strongly induced early in C-PA infection, which protected the uninfected cells from expansion of viral infection. In this case, these classical host-virus interactions contributed to the attenuation of this strongly replicating virus.


Background
Influenza A viruses cause disease in humans, pigs, other mammals, and birds [1]. The genomes of influenza A viruses are composed of 8 negative-sense single-stranded RNA segments; this segmented genome allows gene reassortment between viruses in co-infected cells to produce new viruses. Reassortment of influenza A virus genes caused the deadly H2N2 "Asian flu" and the H3N2 "Hong Kong flu" pandemics in 1957 and 1968, respectively. During these pandemics, the avian virus PB1, HA and NA, or PB1 and HA genes, respectively, were introduced into circulating human viruses [2,3]. The last pandemic strain, the novel swine-origin influenza virus A/H1N1 (S-OIV), carries PB2 and PA genes of avian origin [4]. In addition to the S-OIV pandemic flu, H5N1 highly pathogenic avian influenza viruses (HPAIV) have caused severe or fatal disease in humans in Asia, the Middle East, and Africa since their emergence in Hong Kong in 1997 (WHO, http://www.who.int/csr/disease/ avian_influenza/en/). The H5N1 influenza virus ribonucleoprotein complex (RNP) contributes to viral pathogenesis in chickens [5,6]. Influenza viruses with high polymerase activity have also shown high pathogenicity [7,8]. These lines of evidence suggest that the influenza RNA-dependent RNA polymerase (RdRp) contributes to its pathogenesis.
In this paper, we describe the activation of the polymerase activity of A/Puerto Rico/8/1934 (PR8, H1N1) and A/WSN/1933 (WSN, H1N1) RNPs by the H5N1 HPAIV PA of A/Cambodia/P0322095/2005, which was isolated from a Cambodian victim [40], and the reconstitution of the chimeric virus to analyze the effect of this H5N1 PA in the background of WSN, a well-studied mouse influenza infection model. We found a discrepancy between the viral polymerase activity and proliferation efficiency in cells and its pathogenesis in mice. We then analyzed the mechanism of the attenuation and the low pathogenicity of WSN carrying H5N1 PA.

Results
Effect of H5N1 Cambodia PA on the PR8 and WSN replicons and in vitro RdRp activity We first examined the replicon activity in 293 T cells of a chimeric PR8 RNP containing H5N1 Cambodia PA ( Figure 1A). Influenza replicon activity was measured as previously described [41,42]. The replicon activity was about 200.0 ± 8.2% that of the PR8 RNP (Student's t; p < 0.005), while the Cambodia RNP showed 43.8 ± 2.9% of the replicon activity of the PR8 RNP (p < 0.005). The expression of RdRp and NP in the transfected cells was confirmed by western blotting (Additional file: Figure S1).
As we found 2-fold activation of the PR8 replicon by H5N1 Cambodia PA, we purified both PR8 RdRp and a chimeric RdRp (PR8PB2-PR8PB1-H5N1 Cambodia PA) in order to investigate the effect of the H5N1 Cambodia PA on RdRp activity in vitro. Sodium dodecyl sulfatepolyacrylamide gel electrophoresis (SDS-PAGE) showed that H5N1 Cambodia PA migrated to a higher position than did PR8PA; the Cambodia PA band almost overlapped with that of 18 × HisPR8PB2 ( Figure 1B). The identity of H5N1 Cambodia PA was confirmed by western blotting against PA ( Figure 1C).
We next compared the activities of the purified RdRps in vitro. We tested the transcription activity using v84 and globin mRNA primers ( Figure 1D and E, mRNA), and replication activity using v84 and c84 with and without the dinucleotide primer ApG ( Figure 1D, E, F, and G). ApG-primed and de novo initiation of v84 transcription produced 84-and 83-nt products, respectively, while globin mRNA-primed transcription produced 96nt products. Because virion RdRp uses 10-15 nucleotide primers to initiate from the C at the 2nd position from the 3′ end of the genome [18,22], the 96-mer products were assigned to the transcripts from the 13th G next to the 12th U of the cap-1 structure (m 7 GmACACUUG-CUUUU) of rabbit β-globin mRNA (GenBank; M10843). The mean and standard deviation (error bar in the graph) of the polymerase activity relative to that of PR8 RdRp were calculated from 2 independent measurements of 3 different RdRp preparations. The relative ApG-primed replication activity of the chimeric RdRp was 95.9 ± 4.5% of that of PR8 RdRp, while its de novo replication activity was 126 ± 40% of that of PR8 RdRp. The chimeric RdRp produced 171 ± 31% (p < 0.05) of the amount of 96-nt transcription products. ApGprimed initiation of c84 produced 84-and 81-nt products, while de novo initiation produced 81-nt products.
The relative replication activity of ApG-primed initiation of c84 by the chimeric RdRp was 116 ± 16% of that by PR8 RdRp, while its de novo initiation was 156 ± 29% of that by PR8 RdRp (p < 0.05). We thus confirmed that H5N1 Cambodia PA enhanced both the transcription and replication activities of PR8 RdRp. Before reconstituting WSN carrying H5N1 Cambodia PA, we tested the effect of H5N1 Cambodia PA on the WSN replicon ( Figure 1A). WSN replicon activity (91.3 ± 3.2%) was similar to that of PR8. The replicon activity of WSN containing the H5N1 Cambodia PA (which was 133.9 ± 5.8% of that of the PR8 replicon) was about 1.5-fold higher than that of the WSN replicon (p < 0.005). Therefore, we also observed an activation effect of H5N1 Cambodia PA on the WSN replicon.

Effect of H5N1 Cambodia PA on virus growth in MDCK cells, chicken embryo fibroblasts (CEF), and Vero cells
Next, we tested whether the RdRp activation effect of H5N1 Cambodia PA affected virus growth. As the RdRp subunits of PR8 and WSN are highly homologous, with 96, 97, and 98% amino acid identity between the PB2, PB1, and PA genes, respectively, and as the activation effect of H5N1 Cambodia PA was confirmed in both the WSN and PR8 replicons, we used a WSN reconstitution system to analyze the effect of H5N1 Cambodia PA in H1N1 virus. The WSN virus carrying H5N1 Cambodia PA (C-PA) was reconstituted successfully. First, we compared the multi-step growth of C-PA with that of WSN in MDCK, CEF, and Vero cells.
In MDCK cells, the C-PA titer was always lower than that of WSN and plateaued (4.7 ± 0.2 × 10 4 PFU/mL) 16 hr post-infection (pi), while the WSN titer was 2.2 ± 0.03 × 10 5 PFU/mL 16 hr pi and plateaued (5.6 ± 1.6 × 10 6 PFU/mL) 36 hr pi ( Figure 2A). The C-PA titer in MDCK cells 60 hr pi was significantly higher than that of WSN (p < 0.001). The titers of C-PA and WSN in CEF did not plateau even 60 hr pi. However, the C-PA titer 60 hr pi (7.1 ± 1.4 × 10 5 PFU/mL) was about half of that of WSN (1.3 ± 0.4 × 10 6 PFU/mL), a statistically significant difference (p < 0.001, Figure 2B). We therefore observed a discrepancy between the effects of H5N1 Cambodia PA on RdRp activity and on virus growth in MDCK and CEF cells. However, the C-PA titer in Vero cells between 24 hr and 60 hr pi was significantly higher than that of WSN (p < 0.05, Figure 2C). Both titers plateaued at 5.6 ± 0.1 × 10 5 PFU/mL 70 hr pi. No cytopathic effect was apparent in Vero cells (data not shown). We next compared single-step growth of C-PA and WSN in MDCK and Vero cells (Figure 3). The C-PA titers in both MDCK and Vero cells between 4 and 12 hr pi were higher than those of WSN. The C-PA titer in MDCK cells 12 hr pi (1.6 ± 0.9 × 10 6 PFU/mL) was significantly higher than that of WSN (1.3 ± 0.7 × 10 5 PFU/mL) (p < 0.005, Figure 3A). The C-PA titer in Vero cells 12 hr pi (1.6 ± 0.9 × 10 5 PFU/mL) was significantly higher than that of WSN (2.9 ± 0.2 × 10 4 PFU/mL) (p < 0.05, Figure 3B). The growth of C-PA in both MDCK and Vero cells plateaued 6 hr pi. WSN growth in Vero cells plateaued 8 hr pi, while its titer in MDCK cells continued to increase up to 12 hr pi.

Pathogenicity of the C-PA virus in mice
We also examined the pathogenicity of C-PA in mice. The LD 50 for mouse nasal infection was calculated from the survival rate of the infected mice by the method of Reed and Münch ( Figure 4) [43]. The LD 50 values of C-PA and WSN were 5 × 10 5 PFU and 1 × 10 4 PFU, respectively. The pathogenicity of C-PA is therefore lower than that of WSN, despite its higher RdRp activity, both in cell culture (as reflected by the multi-step growth) and in mice.
We thus confirmed the discrepancy between the RdRp activity and pathogenicity both in cells (virus titer) and in mice. The genome sequence of the C-PA stock was confirmed to be identical to that of the genome reconstitution plasmids.

Interferon induction
Because the multi-step growth activity of C-PA was lower than that of WSN in MDCK and CEF cells but better in Vero cells, and because the single-step growth activity of C-PA was better than that of WSN in both MDCK and Vero cells, we examined the effect of C-PA  the induction of apoptosis ( Figure 6B). The activities of caspases 3 (14.7 ± 9.8 × 10 6 RFU/mg protein, p < 0.0005), 8 (3.6 ± 0.2 × 10 6 RFU/mg protein, p < 0.005), and 9 (1.5 ± 0.5 × 10 6 RFU/mg protein, p < 0.005) were higher in C-PA-infected cells than in WSN-infected cells 10 hr pi (comparisons evaluated by Student's t test). The caspase activities in WSN-infected cells were similar to those in mock-infected cells, indicating that WSN infection did not induce apoptosis 10 hr pi. The activation of caspase 9 in C-PA-infected cells indicates that C-PA infection induces apoptosis through the mitochondrial pathway [47][48][49], and the apoptosis induction by C-PA began at an early time after infection at which no apoptosis was induced by WSN infection.

Histopathology and TUNEL assay of infected mice
Finally, we compared the pathological changes in the lung between C-PA-and WSN-infected mice. The histopathological appearances were similar (Figure 7). The major difference between C-PA-and WSN-infected lungs is that pulmonary edema around blood vessels was present in the C-PA-infected lungs from day 1 pi, although it was also present in WSN-infected lungs on days 3 and 4 pi. On the first day of infection, a moderate amount of lymphocyte infiltration was observed around the bronchioles of C-PA-infected mouse lungs, and this inflammation decreased on days 3 and 4. More lymphocyte infiltration around the bronchioles was observed in WSN-infected mouse lungs on days 1-3, with the most severe inflammation on day 2, and only mild inflammation was observed on day 4. No pathological change was observed in the mock-infected or non-infected mouse lungs.
We simultaneously analyzed these samples by TUNEL assay. A few TUNEL-positive cells were observed in both C-PA-and WSN-infected mouse lungs on days 1-4 pi, Caspase activity is expressed as the fluorescence intensity/total cellular protein (mg) calculated from 3 independent measurements. Mock: mock-infected MDCK cells. Statistical significance was evaluated with Student's t-test. *p < 0.05, ***p < 0.005, ****p < 0.0005. but the numbers did not clearly differ between the 2 viruses. The histopathological findings indicated recovery of the C-PA-infected mouse lungs, which is consistent with the lower pathogenicity of C-PA.

Discussion
The activation of replicon activity by H5N1 HPAIV PA observed in this study has been reported several times ( Figure 1) [34,50,51]. Influenza viruses with high polymerase activity have been reported to show high pathogenicity [7,8]. Therefore, reassortment of H5N1 HPAIV PA into human influenza viruses, including the seasonal influenza viruses (H1N1 and H3N2) and the pandemic influenza virus (H1N1), is a serious concern [34], although this event may not occur easily [32]. The major factor for host adaptation in H5N1 HPAIV RNP is PB2, which shows a strong correlation with pathogenicity in mammalian hosts, including humans [30,31,[51][52][53][54][55].
The WSN reconstitution system [56] was used to analyze the effect of H5N1 Cambodia PA in an animal influenza virus (H1N1). Contrary to our expectation, the multi-step growth of C-PA in MDCK and CEF cells was less than that of WSN, and its pathogenicity was low because of this attenuation (Figures 2, 4). We confirmed the absence of mutations in the C-PA genome. As we previously found a similar discrepancy in influenza promoter/origin function due to a difference in activation of the host defense system [42], we examined whether the discrepancy between RdRp activity and pathogenicity was also due to differences in host defense. As the multi-step growth of C-PA in Vero cells was better than that of WSN, and because its single-step growth, which is highly dependent on the polymerase activity, in both MDCK and Vero cells was better than that of WSN (Figure 3), we first examined the activity of the interferon β promoter and found that it was strongly activated ( Figure 5) in CPA-infected cells.
Type I interferon is induced in virus-infected cells by a signal transduction pathway beginning with retinoicacid-inducible gene-I (RIG-I), which recognizes 5′triphosphate-containing influenza RNA [57][58][59][60][61][62]. In C-PA-infected cells, a large amount of vRNA was expressed from early stages of infection, which might more strongly trigger interferon induction (Figures 5 and Additional file: Figure S3) and thereby induce an antiviral state in uninfected cells that protected them from influenza infection (Additional file: Figure S2A) [63][64][65]. The better multi-step growth of C-PA relative to WSN in Vero cells, which are defective in type-I interferon production [66] (Figure 2C), and the better single-step growth of C-PA than of WSN (Figure 3), are consistent with the high in vitro polymerase activity and strong interferon induction in C-PA infected cells.
Apoptosis (programmed cell death) is another mechanism by which cells restrict viral infection including influenza [67]. However, virus-induced apoptosis causes tissue damage, which is one of the mechanisms of influenza pathogenicity [68]. Apoptosis also aids in the release of influenza viruses [69,70]. WSN induced apoptosis in MDCK cells late in infection [71]. However, no apoptosis was induced by expression of any single protein of either WSN or H5N1 Cambodia. Apoptosis was strongly induced in C-PA-infected cells beginning early in infection when no apoptosis was induced in WSNinfected cells (Figure 6). The activation of caspase 9 indicated that this apoptosis was mediated through the mitochondrial pathway [72]. The increased expression of vRNA (Additional file: Figure S3) in C-PA infected cells due to the high replication activity promoted by H5N1 PA (Figure 1) induced both interferon and apoptosis, resulting in attenuation of C-PA proliferation (Figure 2). Neither caspase 3 nor 8 was induced by expression of WSN PA or Cambodia PA alone (data not shown). The strong induction of interferon discussed above also stimulates the induction of apoptosis via RNA-dependent protein kinase (PKR) [73]. Such rapid induction of apoptosis was also observed in duck cells infected with HPAIV [68]. However, in case of C-PA, early induction of apoptosis attenuated the proliferation of the virus and thus decreased its pathogenicity. PB1-F2 is the only influenza virus protein that induces apoptosis in infected monocytes and potentiates apoptosis during infection [74][75][76]. The WSN PB1-F2 proteins of WSN and C-PA are identical. Kinetic analysis of viral RNAs (Additional file: Figure S3) indicates that PB1-F2 is unlikely to contribute to the strong induction of apoptosis in C-PA infected cells.
C-PA infection strongly induced both interferon production and apoptosis early in infection, which attenuated virus proliferation and pathogenicity despite high RdRp activity. This may be another reason, in addition to poor adaptation, for the difficulty of obtaining reassortant viruses carrying H5N1 HPAIV PA [31,32].

Methods
Cell culture 293 T and Vero cells were maintained in Dulbecco's modified Eagle minimal essential medium (DMEM) containing 10% fetal bovine serum (FBS). CEFs were prepared from 11-day-old embryonated chicken eggs by trypsin digestion and stored in liquid nitrogen. CEF and MDCK cells were maintained in DMEM containing 10% or 5% FBS, respectively.

Expression, purification, and in vitro transcription of the influenza virus RdRp
Expression, purification, and in vitro transcription of the influenza virus RdRp were performed as previously described [42,77,78]. Model RNA templates v84 and c84 model RNA templates were prepared as previously reported [11].
Virus growth assay MDCK, CEF, and Vero cells in 3.5-cm dishes (21 dishes of MDCK cells and CEFs, 27 dishes of Vero cells per a virus) were infected with WSN and C-PA at an MOI of 0.01 at 37°C for 1 hr (multi-step growth assay). The cells were washed 3× with phosphate-buffered saline (PBS) and incubated with 2 mL of DMEM containing 2% FBS (without trypsin) at 37°C, as WSN replicates without trypsin [79]. All of the supernatants of 3 dishes of each cell were taken 6, 12, 24, 36, 48, and 60 hr (and also at 72 and 90 hr for Vero cells) after infection and stored at −80°C. Virus growth was also tested in MDCK and Vero cells after infection at an MOI of 5 (single-step growth assay). Supernatants were harvested 2, 4, 6, 8, 10, and 12 hr after infection. The viruses in the supernatants were titered on MDCK cells by a plaque-formation assay.

Mouse infection
Groups of 4 6-week-old female BALB/c mice (Sino-British Laboratory Animal, Shanghai, China) were anesthetized with ether and inoculated with 10 5 , 10 4 , or 10 3 PFU of WSN or 10 6 , 10 5 , 10 4 , or 10 3 PFU of C-PA in a volume of 50 μL by nasal dropping. Four mice were inoculated with 50 μL of PBS as a mock-infection control. The survival rates were monitored daily and the 50% lethal doses (LD 50 s) calculated [41].

Histopathological analysis
Six-week-old female BALB/c mice were anesthetized with ether and inoculated with 10 5 PFU of WSN or C-PA in 50 μL by nasal dropping. Mice were inoculated with 50 μL of PBS as a mock-infection control. A non-treated mouse was used as a non-treated control. The mockinfected and infected mice were sacrificed by cervical dislocation under anesthesia on days 1, 2, 3, and 4 after infection, and their lungs were removed and fixed with 3.5% formalin/PBS at 25°C for 2 days. The lungs were embedded in paraffin blocks, sectioned at 4-μm thickness, and stained with hematoxylin and eosin (HE).

TUNEL assay
Cells were placed on cover slips in 24-well plates and infected with WSN or C-PA at an MOI of 0.01. Eight, 9,10,11,12,13,14,15,16, and 24 hr after infection, cells were fixed with 1% formalin/PBS. The TUNEL assay was performed using the DeadEnd TM Fluorometric TUNEL system according to the company's instructions. The samples were observed using a fluorescence microscope (Leica DM IRB, Leica, Wetzlar, Germany), and the numbers of TUNEL positive cells in 3 random fields were counted.

Caspase activity
MDCK cells were plated in 10-cm-diameter plates and infected with WSN or C-PA at an MOI of 0.01. 10 hr after infection, the cells were harvested and the activities of caspases 3, 8, and 9 measured using the Caspase-3/ CPP32 Fluorometric Assay kit, the Caspase 8/FLICE Fluorometric Assay kit, and the Caspase 9 Fluorometric Assay kit (Biovision, Inc., Milpitas, USA) according to the manufacturer's instructions.

Interferon induction
Interferon induction was analyzed by measuring the luciferase activity of cells transfected with p-125Luc, which was kindly provided by Dr. Fujita [80]. 293 T cells were transfected with p-125Luc (1 μg) and pRL SV40 (100 ng). Four hours post-transfection, the cells were infected with WSN or C-PA at an MOI of 0.01. The interferon promoter activity was measured as the luciferase activity using the Dual-Glo luciferase assay kit and a GloMax 96 Microplate Luminometer (Promega) and normalized as the firefly luciferase/Renilla luciferase activity ratio.

Statistical analysis
The statistical significance levels of the data were evaluated by Student's t-test, with p < 0.05 indicating statistical significance.