Dual E627K and D701N mutations in the PB2 protein of A(H7N9) influenza virus increased its virulence in mammalian models

The ongoing avian H7N9 influenza outbreaks in China have caused significant human fatal cases and the virus is becoming established in poultry. Mutations with potential to increase mammalian adaptation have occurred in the polymerase basic protein 2 (PB2) and other viral genes. Here we found that dual 627K and 701N mutations could readily occur during transmission of the virus among ferrets via direct physical contact, and these mutations conferred higher polymerase activity and improved viral replication in mammalian cells, and enhanced virulence in mice. Special attention needs to be paid to patients with such mutations, as these may serve as an indicator of higher virus replication and increased pathogenicity.

Since the initial outbreak in Eastern China in 2013 1 , the novel avian influenza A (H7N9) virus has caused more than 600 human infections with a mortality rate of over 35% 2 . Infected chickens at live-poultry markets were identified as the source of human infections and the H7N9 virus now appears to be becoming established in China 3,4 . The occurrence of repeated outbreak waves has posed an ongoing potential pandemic threat to the world.
With the development of the H7N9 virus in the field, some amino acid mutations have occurred which are believed to be associated with host preference and specificity in mammalian species 3,4 . As residues at positions 627 and 701 of the polymerase basic protein 2 (PB2) are considered critical for the mammalian adaptation of avian influenza viruses [5][6][7][8] , several studies have independently shown that single E627K (glutamic acid to lysine) or D701N (aspartic acid to asparagine) mutations could increase polymerase activity and viral replication in mammalian cells and the pathogenicity of H7N9 viruses in the BALB/c mouse model [9][10][11] . Yet it is still unknown whether spontaneous emergence of dual D701N and E627K mutations can occur readily in infected mammals and whether these will synergistically enhance the virulence of H7N9 viruses.
Here we used a ferret transmission model and deep-sequencing of the viruses shed by ferrets experimentally exposed to the prototype H7N9 virus, A/Anhui/1/2013 (AH1, with 627K and 701D in PB2) 1 and an early precursor-like virus, A/Shanghai/05/2013 (SH5, with 627E and 701D in PB2) 12 . Our findings show that both the E627K and D701N mutations occurred in ferrets that had direct contact with infected animals within a few days post-exposure. Dual mutations of 627K and 701N, introduced by

Results
Transmission of H7N9 viruses in ferrets. All ferrets directly inoculated with a 10 6 median tissue culture infective dose (TCID 50 ) of the wild type AH1 or SH5 viruses shed viruses through the nasal cavity for 6 days post-inoculation (dpi) with peak titers, which occurred at 2 dpi, ranged from 6.0 ∼ 6.3 log TCID 50 /ml for AH1 and from 5.3 ∼ 6.3 log TCID 50 /ml for SH5 (Table 1). In the AH1 transmission group, all physical and airborne exposed ferrets seroconverted, and viruses were detected in all physical contact animals and two out of three airborne exposed ferrets (Table 1). Although seroconversions were observed in all three physical contact ferrets exposed to SH5, virus was only detected in the nasal washes of one animal (SH5-PC3) and no airborne exposed ferrets seroconverted or shed virus. The SH5-PC3 ferret started to shed virus 24 hours post-exposure (hpe) and the virus shedding titers peaked at 4 days post-exposure (dpe, Table 1). Relatively high shedding titers (> 3.5 to 5.8 log TCID 50 /ml) were maintained in SH5-PC3 till 8 dpe (Table 1).
PB2-627K and -701N mutations occurred in ferrets exposed to H7N9 viruses. All ferrets that shed virus after exposure to the wild type AH1 virus retained the 627K residue in the PB2 gene but two of the three physical contact animals carried a mixed population at the 701 position, with the D701N mutation occurring at a frequency of 94.2% and 86.8% respectively on their peak shedding day (3 and 4 dpe respectively, Table 1). Neither of the two AH1 airborne exposed ferrets that shed virus possessed the D701N mutation even though they were paired with the contact exposed animals that did. In the SH5 physical contact ferret (SH5-PC3), all viruses shed at 4 dpe retained 627E and 701D, but at 8 dpe 30% of the viral sequences contained 627K and 1.1% had 701N (Table 1) [8]: E627K(30%), D701N(1.1%) 80 Table 1. Virus shed by ferrets inoculated with or exposed to the human H7N9 influenza viruses. -: virus or HI antibody not detected. dpi/dpe: days post-inoculation or post-exposure. HI titer: heamagglutination inhibition titer. In-#: ferret intranasally inoculated with the virus; PC-#: ferret exposed to the infected animals via physical contact; AE-#: ferret exposed to the infected animals via airborne exposure.  Fig. 1). At all temperatures tested, PB2 627E mutants (ED and EN) reduced activity (0.27 -0.92 fold) below that of the wild type virus (KD) while 701N (KN) increased activity by 0.43-1.15 fold (Fig. 1). Polymerase activity at 33 °C was generally lower than that at higher temperatures (p < 0.01), except for the mutant with PB2-627E+ 701D where the level of activity was not affected by temperature (Fig. 1). For mutants with 627K (KD or KN), polymerase activity at 35 °C was also significantly lower than that at the higher temperatures (p < 0.01).  loss at 8 dpi for a dose of 10 2 TCID 50 was less than 10% for the rgAH1-KN viruses and weight loss was not observed in other cases ( Fig. 3a,b). At a dose of 10 3 TCID50, two mice inoculated with rgAH1-KN died at 9 dpi, but no rgAH1-KD and rgAH1-EN inoculated mice died and no body weight loss was recorded in the rgAH1-EN inoculated mice (Figs 3c and 4a). At inoculation doses of 10 4 TCID 50 or above, significant body weight loss occurred in all three groups of mice, with the rgAH1-EN inoculated mice consistently showing the least weight changes and numbers of fatalities (Figs 3d-f and 4b-d). The MLD 50 of the rgAH1-EN, rgAH-KD (wild type) and rgAH1-KN viruses were ≥ 10 6.3 , 10 4.5 , and 10 3.3 TCID 50 (Fig. 4), while the MID50 were 10 2.5 , 10 1.1 , and 10 0.9 TCID 50 , respectively (Table 2). Overall, the virulence of the rgAH1-KN virus was the highest, followed in order by the rgAH1-KD (wild type), and rgAH1-EN viruses (

Discussion
Influenza A viruses can readily acquire mutations through virus replication in host cells, owing to the low fidelity of their RNA dependent polymerase 13,14 . At the beginning of the first outbreak wave of H7N9 in China, several potential mammalian adaptation markers had already appeared in H7N9 viruses that had infected humans, including PB2 627K and 701N 1,3,4,12 .
As of 4 th December 2014, all available H7N9 virus sequences from the poultry and environmental samples were found to exclusively carry PB2-627E while the majority (69.0%, 89/129) of the human isolates had a lysine mutation at this position (E627K), and a tree sparrow virus identified in Shanghai possessed both E and K 15 . An Aspartic acid to Asparagine mutation, relative to avian or environmental samples, was observed at position 701 in PB2 (D701N) in 8.5% (11/129) of human isolates. In January 2014, an H7N9 virus (A/Shanghai/PD-02/2014) carrying both the E627K and D701N mutations in the PB2 protein (GenBank accession number KJ549801) was isolated from a fatal human case, indicating that such dual mutations can occur naturally.
In this study, by using a ferret transmission model, we demonstrated that both the E627K and D701N mutations could occur within one passage in physical contact ferrets, and these two mutations could co-emerge in the same virus particle. The dual 627K and 701N mutations in the PB2 protein of AH1 viruses was associated with increased polymerase activity, better replication in human cells and enhanced virulence in mice. However, as the E627K and D701N mutations were not found in airborne exposed ferrets, this suggests they may not increase the airborne transmissibility of the viruses. Structural studies of the C-terminal domain of PB2 show that residue 701 is located in the nuclear localization signal (NLS) and importin binding domain, and it might disrupt the salt bridge between 701D and 753R, thereby facilitating the unfolding of the NLS and its exposure to and binding of the importin molecules [16][17][18][19] . An increase in the nuclear accumulation of PB2 in mammalian cells could enhance the polymerase activity of an influenza virus carrying a 701N residue. PB2 proteins containing 627K or 627E possess nearly identical structures, but the electrostatic surface potential is more positively charged with 627K 16,17,20 . The large basic surface on the PB2 protein was assumed to be essential for efficient polymerase activity in human cells and may be involved in vRNP assembly or interactions with viral or host proteins and/or RNAs 17,19,20 .
The avian H7N9 influenza virus that emerged in 2013 has caused three outbreak waves in humans but relatively few viral sequences are available, given the large numbers of affected birds and humans. This paucity of information on the possible adaptations of H7N9 viruses within infected birds or patients makes the prediction of clinical outcomes difficult. Our study, together with previous reports [9][10][11]    Full-length viral genes were cloned into the plasmid pHW2000 21,22 . Mutations were introduced into the pHW2000-AH1-PB2 plasmid using the QuikChange ® II Site-Directed Mutagenesis Kit (Stratagene) to generate the K627E, D701N and the double mutations (627E plus 701N). Plasmid pRL-TK (Promega) was renamed as pRluc in this study, which uses thymidine kinase (TK) promoter to drive the constitutive expression of Renilla luciferase (Rluc). Reporter plasmid pFluc was constructed by cloning the firefly luciferase (Fluc) gene from the pGL3 vector (Promega) into the RNA polymerase I promoter/terminator cassette. The untranslated regions (UTRs) of the influenza A/WSN/33 NP segment was introduced into the flanking regions of the Fluc gene to produce artificial influenza NP-like RNA segment. All plasmids were confirmed to have the exact sequences as designated.

Generation of recombinant viruses and virus titration. Recombinant viruses rgAH1-KD (with
wild type PB2-627K and 701D), rgAH1-KN (rgAH1-PB2-627K and 701N), rgAH1-ED (rgAH1-PB2-627E and 701D) and rgAH1-EN (rgAH1-PB2-627E and 701N) were generated by co-transfection of the eight reverse-genetically reconstructed plasmids, each carrying a viral gene segment, into 293T/ MDCK co-cultured monolayers 22 . The identity of each propagated mutant virus was ascertained by whole genomic sequencing. Viral titrations were determined using MDCK cells, and the TCID 50 was calculated using the Karber method 23 . Previous reports [9][10][11] have shown that H7N9 virus containing PB2-627E and 701D was less virulent than those carrying either PB2-627K-701D or PB2-627E-701K. Because titers of rgAH1-ED stock prepared in MDCK cells were consistently lower than those of the three other mutants by more than ten fold, and rgAH1-ED did not form clear plaques on MDCK cells, rgAH1-ED was not included in the study of virus growth kinetics on A549 cells and the in vivo mouse experiment.

Virus infection and transmission in a ferret model.
Three six-month-old influenza-free male ferrets (Mustela putorius furo, Wuxi Sangosho Co. Ltd.), held in separate cages, were intranasally inoculated with 10 6 TCID 50 of AH1 or SH5 viruses in 500 μ l of MEM. At 24 hpi, three naïve ferrets (as physical contacts) were introduced to each of the three main cages with the directly inoculated ferrets. Another three were placed in adjacent cages at a 10 cm distance from the main cage, to serve as airborne exposed animals 24 . Nasal washes were collected into 1ml of cold phosphate buffered saline (PBS) on a daily basis and titrated using TCID 50 assays. When viruses shed by the contact ferrets (either physical or airborne contacts) reached their peak titers, RNAs from the nasal swabs were extracted and subjected to deep sequencing on a Mi-Seq desktop sequencer (Illumina), giving a coverage of 6,000 ∼ 25,000× . The PB2 gene sequences at the 627 and 701 positions were analyzed for their genetic heterogeneity. At 14 dpi, sera were collected from each animal for detection of seroconversion and antibody titers by hemagglutination inhibition (HI) tests. Virus infection in the mouse model. Groups of five 8-to 10-week-old specific-pathogen-free (SPF) female C57BL/6 mice (Vital River Laboratories, Beijing) were anesthetized with 0.2 ml of 0.75% pentobarbital sodium and inoculated intranasally with recombinant viruses at the indicated dose (10 1 ∼ 10 6 TCID 50 ) in a 50 μ l volume, or mock inoculated with 50 μ l PBS. Body weights and clinical signs of infections were recorded daily. Mice that lost more than 30% of their original weight were euthanized for humane reason and recorded as a fatal infection. At 14 dpi, blood was collected from each of the surviving mice, and serum was separated for determination of antibody titers using HI tests. Mouse median lethal dose (MLD 50 ) and 50% infectious dose (MID 50 ) were determined using the Karber formula 23 . To determine the virus replication sites, groups of twelve 8-to 10-week old female C57BL/6 mice were intranasally inoculated with 10 3 or 10 5 TCID 50 of the indicated viruses and three mice in each group were euthanized at 1, 3, 5, and 7 dpi. Brain, nasal turbinate, trachea, lung, heart, spleen, kidney, liver, intestinal tract, eyeball, conjunctiva and spinal cord were collected and virus titers were determined by TCID 50 assays.
Histological and immunohistochemical examinations. Lung tissues from the infected mice were perfused with 10% neutral buffered formalin and fixed for over 24 h before processing. The tissues were then embedded in paraffin by standard tissue processing procedures, cut into 3 ∼ 4 μ m sections and affixed on glass slides. Standard hematoxylin and eosin (H&E, Sigma) staining and immunohistochemical staining of NP antigens in the lung tissues were performed by using a mouse anti-NP monoclonal antibody, kindly provided by Professor Ningshao Xia, the National Institute of Diagnostics and Vaccine Development for Infectious Diseases, Xiamen University, and a goat anti-mouse IgG-biotin conjugated secondary antibody (Calbiochem) 24 .
Statistical analysis. Differences between experimental groups were evaluated using the Student's t test. A p-value < 0.05 was considered statistically significant.