Prerequisites for the acquisition of mammalian pathogenicity by influenza A virus with a prototypic avian PB2 gene

The polymerase of avian influenza A virus (AIV) is a heterotrimer composed of PB2, PB1, and PA. PB2 plays a role in overcoming the host barrier; however, the genetic prerequisites for avian PB2 to acquire mammalian pathogenic mutations have not been well elucidated. Previously, we identified a prototypic avian PB2 that conferred non-replicative and non-pathogenic traits to a PR8-derived recombinant virus when it was used to infect mice. Here, we demonstrated that key amino acid mutations (I66M, I109V, and I133V, collectively referred to as MVV) of this prototypic avian PB2 increase the replication efficiency of recombinant PR8 virus carrying the mutated PB2 in both avian and mammalian hosts. The MVV mutations caused no weight loss in mice, but they did allow replication in infected lungs, and the viruses acquired fatal mammalian pathogenic mutations such as Q591R/K, E627K, or D701N in the infected lungs. The MVV mutations are located at the interfaces of the trimer and are predicted to increase the strength of this structure. Thus, gaining MVV mutations might be the first step for AIV to acquire mammalian pathogenicity. These results provide new insights into the evolution of AIV in birds and mammals.

We previously identified avian polymerase genes with different degrees of mammalian pathogenicity. A prototypic PB2 gene of an H9N2 low-pathogenic AIV (LPAIV) strain, A/chicken/Korea/01310/2001 (H9N2) (01310), was not replicative and non-pathogenic (no body weight loss) in BALB/c mice after inoculation of 7 + 1 PR8-derived recombinant virus. Furthermore, the PB2 gene of the H9N2 LPAIV strain A/Korea/KBNP-0028/2000 (H9N2) (0028) was replicative but non-pathogenic in BALB/c mice, and we identified candidate amino acids related to the replication of 0028 PB2 in mouse lungs by comparing the amino acid sequences of PB2 proteins 22 . Importantly, the identification of key amino acids in 0028 PB2 that confer replicative ability, but not pathogenicity, may improve understanding of the first-step mutations that occur in prototypic 01310 PB2 to facilitate the acquisition of fatal mammalian pathogenicity. Here, we identified key mutations (I66M, I109V, and I133V, collectively referred to as "MVV") that increase replication efficiency in both avian and mammalian hosts and are predicted to increase the structural integrity of the trimeric polymerase. These MVV mutations are essential prerequisites for the subsequent acquisition of fatal mammalian pathogenic mutations. Collectively, these results provide important insights into the first evolutionary step taken by AIV to acquire pathogenicity in mammals.
SCientiFiC RePoRtS | 7: 10205 | DOI: 10.1038/s41598-017-09560-z To assess viral growth efficiency in MDCK cells and viral pathogenicity in mice, we generated PR8-derived recombinant viruses containing single amino acid mutations in 01310 PB2. Recombinant viruses with the I133V, L373I, or E627K mutations produced higher titres than the parent strain rPB2(01310) (P < 0.05) (Fig. 1b). In murine pathogenicity experiments, most of the mutant viruses, except the rPR8 virus, did not cause body weight loss or mortality during the observation period. Only rPB2(01310)-E627K caused slight body weight loss of up to 4% over 7-10 days post-infection (dpi) (P < 0.05) (Fig. 2a,b). During mouse infectivity screening, however, we observed detectable growth of mutant viruses containing the I66M, I109V, I133V, E627K, or A674T mutations ( Table 1). The L373I mutation, which increases viral replication efficiency in MDCK cells, was excluded from further analysis because viruses carrying this mutation did not proliferate sufficiently in mouse lung. The E627K mutation has a well-characterized role related to mammalian pathogenicity, whereas the A674T mutation, which is conserved among most human influenza viruses, does not appear to be related to mammalian pathogenicity 23,24 . 66M, 109V, and 133V are novel mutations that were first characterized in the present study.

Frequency of key amino acid mutations.
To understand the frequency of each identified mutation (I66M, I109V, I133V, E627K, and A674T), we examined IAVs collected from birds, pigs and humans ( Table 2). The human IAVs were divided into two groups: the first contained the H5, H6, H7, H9, and H10 IAVs (birds to humans), and the second contained the H1, H2, and H3 IAVs (humans), excluding the 2009 pandemic H1N1 virus. Interestingly, most of the IAVs already possessed the 66M, 109V, or 133V mutation, regardless of the host Figure 2. Comparison of mouse pathogenicity of 01310 PB2 variants. The virulence of 01310 PB2 single amino acid variants (a,b) and multi-amino acid variants (c,d) was observed by body weight loss (a,c) and mortality (b,d) of infected mice. Five 6-week-old BALB/c mice were challenged with 1.0 × 10 6 EID 50 of each virus or PBS (mock). Mortality and weight loss were observed for 14 days. The average weight loss ± s.d. was measured by comparing to the initial weight of each mouse. Statistical significance was analysed using Student's t-test (compared to rPB2(01310), *P < 0.05). species they were isolated from. However, viruses bearing the 627K and 674T mutations were significantly more common in 'birds to humans' and human IAVs than bird and pig IAVs. The well-known mammalian pathogenic factor, D701N mutation, occurs less frequently than the E627K mutation in the PB2 genes of "human" and "bird to human" IAVs (Table 2) 17,25 . Thus, we concentrated on the effect of E627K and its synergistic effects with the MVV mutations with regard to mammalian pathogenicity.
Effects of multiple amino acid mutations. Based on the high frequencies of the MVV and MIV mutations among the examined AIVs, we introduced multiple mutations combined with E627K and A674T into 01310 PB2 (MIV, MVV, MVVT, MIVKT, MVVK, and MVVKT) and performed mini-genome assays. The combination of 66M and 133V (MIV), as well as that of 66M, 109V, and 133V (MVV), significantly increased polymerase activity compared to the individual amino acid mutations at 33 °C and 37 °C ( Fig. 1a and S1). The combination of MVV with 627K (MVVK) significantly increased polymerase activity compared to the single 627 K mutation at 33 °C and 37 °C ( Fig. 1a and S1) (P < 0.05). The 674T mutation did not influence polymerase activity; however, in combination with MVV (MVVT), it increased polymerase activity significantly compared to MVV alone.
To further evaluate the effects of the MVV mutations, we constructed loss-of-function (LOF) genes using the mouse pathogenic PR8 PB2 gene (Fig. 3a). The polymerase activities of the LOF mutants possessing the Virus Positive rate* (Virus titre (logEID 50 /0.1 ml)) 1 st experiment 2 nd experiment 3 dpi 3 dpi 6 dpi We next generated LOF mutant viruses possessing the above-described mutated PR8 PB2 genes and compared their replication efficiency in MDCK cells. The virus titre of rPB2(PR8)-III at 3 dpi was slightly lower than that of rPR8, but it recovered at 5 and 7 dpi (Fig. 3b). Furthermore, rPB2(PR8)-IIIE and rPB2(PR8)-IIIEA replicated less efficiently than rPR8 and rPB2(PR8)-K627E.

Acquisition of fatal mammalian mutations.
To investigate the effect of the MVV mutations on the acquisition of fatal mammalian adaptive mutations, we performed quasi-species analysis using the lungs of mice infected with rPB2(01310)-MVV at 6 dpi (Table S1). Cloned amplicons encoding amino acid residues 590-701 were sequenced, and well-known mammalian pathogenic mutations such as Q591R/K, E627K, and D701N were identified in four out of five lungs. Surprisingly, at least half of the quasi-species of three mice possessed mammalian pathogenic mutations (Q591K, E627K, or D701N), and double mutations [E627K (10/10) and Q591R (1/10)] were observed in one mouse.

Structure-function relationship.
To predict the structure-function relationship of I66M, I109V, and I133V, we located each amino acid residue in a 3D structure of the PB1, PB2, and PA trimer from a bat IAV 3 . Interestingly, all the residues at the 66th, 109th, and 133rd positions were predicted to be located at the interfaces of the polymerase complex (Fig. 5). The 66th and 109th residues were located close to 628N, 629N, and 630P as well as 613W and 621R of the PB1 protein. The 133rd residue was located close to 429P, 430I, and 433I of PA. We compared the neighbouring amino acid residues of bat IAV PB1 and PA with those of 01310 and PR8 PB1 and PA and found that most (except 621 and 628 of PB1) were conserved. Residues 621 and 628 of 01310 and PR8 are glutamine and leucine, respectively. By analysing the frequencies of the neighbouring amino acid residues in PB1 and PA, it was demonstrated that the 613W, 629N, 628L/M, and 630P residues of most IAVs are located in PB1 regardless of host species, but position 621 in PB1 carries different amino acids depending on the host species (Table S2). Residues 429P, 430I, and 433I in PA were conserved among most avian and mammalian IAVs (434P, 435I, and 438I), but the precise locations differed due to a 5-amino-acid deletion in bat PA. Considering that the neighbouring amino acids are conserved, except for residue 621 in PB1, these residues may form key interactions . At 3 dpi, the allantoic fluid of the embryonated eggs and the cell culture supernatants of the MDCK cells were collected, and the PB2 gene was amplified by RT-PCR and sequenced. The codon for 66I is ATA, 109I is ATC, 133I is ATT, and 627E is GAG; for the PB2 variants, the codon for 66M is ATG, 109V is GTC, 133V is GTT, and 627K is AAG. The DNA sequence chromatograms shown correspond to one sample from each host.

Discussion
To date, various amino acid changes in avian IAVs related to mammalian replication and pathogenicity have been reported, and PB2 is one of the most important genes for overcoming host barrier defences 36,38 . In the present study, we identified the novel amino acid mutations I66M, I109V, and I133V, which increased polymerase activity and replication efficiency in mammalian and avian hosts as well as pathogenicity in mice both independently and in combination. The E627K mutation was previously reported to play a key role in the mammalian pathogenicity of avian IAVs 10, 14, 15, 37 . The single E627K mutation in the prototypic 01310 PB2 increased polymerase activity and replication efficiency more effectively than the MVV mutations, but it did not cause apparent body weight loss or mortality in mice (Figs 1a,b and 2a,b). However, the combination of E627K with the MVV mutations was sufficient to cause severe body weight loss and mortality (Fig. 2c,d). Although the MDT of the LOF mutant rPB2(PR8)-III was prolonged, the mutant retained sufficient pathogenicity to cause 100% mortality in mice. However, the K627E mutation eliminated the pathogenicity of the LOF mutants rPB2(PR8)-K627E and rPB2(PR8)-IIIE in mice (Fig. 3c,d). Thus, the E627K mutation has more important implications than the MVV mutations, but the MVV mutations are present at a much higher frequency than E627K in avian, swine, and human IAVs and therefore may represent the first-step mutations acquired by the prototypic PB2. Based on the growth competition between MVV and III recombinant viruses, the MVV mutations may be required for efficient replication in avian as well as mammalian hosts. Therefore, the very low frequencies of the intermediate mutants MIV, MVI, IVV, MII, IVI, and IIV and the prototype III in IAVs may imply that the evolution of polymerase activity in birds is approaching 'evolutionary stasis' , raising a question about the presence of other hosts of ancient IAVs and their transmission to modern birds and mammals (Table 2) 2 .
The significantly lower frequencies of MVV and MIV with additional mammalian pathogenicity mutations in birds compared to pigs and humans may reflect negative selection in birds, but the significantly higher frequencies of MVV and MIV with additional mammalian pathogenicity mutations in pigs and humans may reflect positive selection ( Table 2). The different selectivity of the 627E and 627K mutations in avian and mammalian hosts, respectively, has been a matter of conflict [39][40][41][42][43][44] . In addition, the 627K mutation increased the mammalian pathogenicity of most H5N1 IAVs, but it did not increase the pathogenicity of other IAVs, including equine and some swine IAVs, some H5N1 viruses, and pandemic H1N1 viruses 10,[45][46][47] . The growth competition experiment clearly demonstrated that viruses harbouring 627E grew more rapidly than those harbouring 627K in avian hosts and that those harbouring 627K grew more rapidly than those harbouring 627E in mammalian cells (Fig. 4). Thus, the conflicting results of previous reports may be caused by different backgrounds of additional mammalian pathogenicity mutations 48 . Therefore, the prototypic 01310 PB2 may be useful for comparing the effects of mammalian pathogenicity mutations on viral growth efficiency in different hosts and pathogenicity in mammals.
Our quasi-species study revealed that the MVV mutations were the minimum requirement for the acquisition of additional mammalian pathogenicity mutations. Acquisition of additional E627K, D701N, or Q591K/R mutations in 4 out of 5 mice during the first mammalian infection was unexpected, but it also supports the fact that the MVV mutations are the minimum essential predisposing mutations to acquire mutations for mammalian pathogenicity. When we compared the mammalian pathogenicity mutation patterns in PB2 from pig and human IAVs, it was found that multiple mutations in different combinations had been accumulated in PB2 (data not shown). These different accumulation patterns may be the result of growth competition in natural hosts. Therefore, tracking of accumulated mutations in PB2 using the prototypic 01310 PB2 may provide information on the effect of single or accumulated mutations on mammalian pathogenicity, providing some insight into the evolutionary steps taken by PB2 during adaptation to mammals under competition between IAVs.
According to computational structure-function analysis, the 66th, 109th, and 133rd residues were predicted to be located at the interfaces of the polymerase complex of a bat IAV (Fig. 5) 3 . The conserved amino acids in PB1 and PA neighbouring the residues at positions 66, 109, and 133 of PB2 supports the importance of the I66M, I109V, and I133V mutations. The position 66 isoleucine and methionine have hydrophobic side chains, but methionine may interact more strongly with neighbouring 628L and 630P. Because the side chain of isoleucine is bulkier than that of valine, 109V and 133V may fit better into the interface of PB1 and PA, which may affect the structural integrity of the polymerase complex and thereby increase polymerase activity. Therefore, the MVV mutations are not host-specific but rather universal mutations for avian and mammalian host adaptation. The question of why the polymerase evolved to obtain greater structural integrity remains to be answered. However, a recent report indicated that oviraptorids, ancient relatives of birds in the Cretaceous period, had a lower body temperature than modern birds and mammals, which could have necessitated a more structurally stable complex 49 .
In conclusion, the MVV mutations in PB2 have an important role in IAV replication. These mutations not only affect the structural integrity of the polymerase complex but are also an essential prerequisite for the subsequent acquisition of mammalian pathogenic mutations. Our results raise questions surrounding how the prototypic PB2 was transmitted from ancient hosts, with lower body temperatures, to modern birds, with higher body temperature. In addition to improving understanding of the molecular steps taken by IAV to acquire mammalian pathogenicity, the prototypic PB2 may be useful as a template for the grafting of certain mutations to compare their effects on mammalian pathogenicity.

Materials and Methods
Viruses, eggs, and cells. This study used A/PR/8/34 (H1N1) (PR8) virus and A/chicken/Korea/01310/2001 (H9N2) (01310), a strain used for an inactivated oil emulsion vaccine in Korea. To synthesize recombinant PR8 virus (rPR8), a Hoffmann vector system was used as described previously 50,51 . Recombinant PR8 viruses were generated and passaged three times in 10-day-old SPF embryonated chicken eggs (ECEs) (Charles River Laboratories, North Franklin, USA) and then used in experiments. Additionally, 293T, MDCK, and A549 cells were purchased from the Korean Collection for Type Cultures (KCTC, Daejeon, Korea), and PK-15 cell was acquired from American Type Culture Collection (ATCC, VA, USA). 293T, MDCK, and PK-15 cells were maintained in DMEM supplemented with 10% FBS (Life Technologies Co., CA, USA), and A549 cell was maintained in DMEM/F12 supplemented with 10% FBS.
Cloning of the PB2 gene and site-directed mutagenesis of 01310 PB2 genes. Each viral segment was cloned into Hoffmann's bi-directional transcription vector pHW2000 50 . The insert sequence was confirmed by sequencing with primers cmv-SF (5′-TAAGCAGAGCTCTCTGGCTA-3′) and bGH-SR (5′-TGGTGGCGTTTTTGGGGACA-3′). Site-directed mutagenesis of specific amino acid substitutions in PB2 genes from the 01310 virus was implemented using a Muta-direct Site Directed Mutagenesis Kit (iNtRON, Korea) as per the manufacturer's protocol.
Mini-genome assay. To evaluate the polymerase activity of each mutated virus, we constructed pHW-NP-Luc plasmids, which have pHW2000 backbones containing the untranslated region of the PR8 NP gene inserted in the antisense direction between the RNA polymerase I promoter and the terminator of the vector. The firefly luciferase gene from the pGL-3 vector was inserted between the NP gene 5′ and 3′ non-coding regions. 293T cells in 12-well plates were co-transfected with 0.1 μg each of pHW-NP-Luc and mutated 01310 PB2 and PR8 PB1, PA and NP genes. Additionally, 0.1 μg of the Renilla luciferase plasmid pRL-TK (Promega, USA) was also co-transfected, which served as an internal control to normalize variations in transfection efficiency and sample processing. Then, 24 hours after transfection, luminescence was assessed using a Dual-Glo Luciferase Assay System (Promega, USA) in accordance with the manufacturer's instructions on a TECAN Infinite200 pro machine (Tecan Benelux bv, Giessen, Netherlands). All results shown are the average from triplicate experiments, and the standard deviation was calculated.
Rescue of mutant viruses. PB2 plasmids containing mutations of interesting amino acids and 7 genome segments of PR8 were transfected into 293T cells by transfecting Hoffmann's eight reverse genetics plasmids as described previously with some modifications 50,51 . Briefly, bi-directional pHW2000 vectors containing PB1, PA, HA, NP, M, and NS of the PR8 virus and 01310 virus PB2 containing mutations in amino acids of interest were transfected with 300 ng of each plasmid using Lipofectamine 2000 and Plus reagents (Life Technologies Co., CA, USA). After overnight incubation, 1 ml of Opti-MEM (Life Technologies Co., CA, USA) and 0.5 mg/ ml of L-1-tosylamido-2-phenylethyl chloromethyl ketone (TPCK)-treated trypsin (Sigma-Aldrich, USA) were added. After 24 h, the culture medium was harvested, and 200 µl of the medium was injected into 10-day-old SPF ECEs via the allantoic cavity. Three days after inoculation, the allantoic fluid was harvested and checked for virus growth via HA assay using 1% (v/v) chicken red blood cells (RBCs) according to the WHO Manual on Animal Influenza Diagnosis and Surveillance. All mutant viruses were confirmed by RT-PCR and sequencing.

Titration of viruses.
Each mutant virus was inoculated into five 10-day-old SPF ECEs for virus isolation. To estimate virus titre, each virus was serially diluted from 10 −1 to 10 −9 in 10-fold increments, and each dilution was injected into five 10-day-old SPF ECEs as well as inoculated onto MDCK cells. The 50% chicken embryo infectious dose (EID 50 ) and 50% tissue culture infectious dose (TCID 50 ) were calculated using the Spearman-Karber method.
Animal experiments. Six-week-old female BALB/c mice were purchased from KOATEC (Pyeongtaek, Korea), and a mouse pathogenicity test was carried out by BIoPOA Co (Yongin, Korea) in accordance with national guidelines for the care and use of laboratory animals. To measure the mouse pathogenicity of each mutant virus, five mice were anaesthetized via intraperitoneal injection of 15 mg/kg Zoletil 50 (Virbac, Carros, France) and then intranasally inoculated with 10 6 EID 50 /50 µl of each virus as described previously 22 . Negative control (mock) mice were injected with the same volume of sterilized PBS. Mortality and weight loss were measured for 14 days. Mice that lost more than 30% of their original weight were euthanized and recorded as a death. For the measurement of virus replication in the lungs of infected mice, six mice from each group were injected with PBS (mock) or 10 6 EID 50 /50 µl of mutant virus. The lungs were collected at 3 and 6 dpi and then stored at −70 °C until use. The lungs were ground using a TissueLyzer 2 (Qiagen, Valencia, CA, USA) with 5 mm stainless steel beads and a volume of PBS equal to 10% of the lung weight in suspension. Then, 10 volumes of PBS were mixed with the ground tissues. After centrifugation at 2000 × g for 10 min, the supernatants were used for viral titres, which were measured as described above. To determine whether adaptive mutations were present in the PB2 gene of viruses isolated from infected mouse lung, RNA extraction and RT-PCR were carried out using lung diluents infected with rPB2(01310)-MVV. The PCR products were cloned into a TA cloning vector (RBC, Taiwan). By selecting 10 colonies per specimen and sequencing, the proportions of adaptive mutations were confirmed.
Histopathology. Mice from each group were euthanized on day 6 dpi, and lungs were collected and fixed in 10% phosphate-buffered neutral formalin. One slice from each lung lobe per mouse was pulled, processed routinely, embedded in paraffin, and stained with haematoxylin and eosin for histopathological analyses. Histological changes were evaluated according to the modified methods of McAuley et al. 52 . Each pulmonary lesion was graded on a scale of 0 to 4. The grading system for histological characterization of the lesions was defined as follows: 0, normal lung; 1, mild infiltration of inflammatory cells around airways and vessels; 2, moderate infiltration of inflammatory cells around airways and vessels and mild leukocyte infiltration of alveolar spaces and interstitium; 3, moderate to severe infiltration of inflammatory cells around airways and vessels, moderate leukocyte infiltration of alveolar spaces and interstitium, mild necrosis, and hyperplasia of airway epithelium; 4, severe infiltration of inflammatory cells around airways and vessels, severe leukocyte infiltration of alveolar spaces and interstitium, moderate to severe necrosis, and hyperplasia of airway epithelium.