Pathogenic influenza B virus in the ferret model establishes lower respiratory tract infection

Influenza B viruses have become increasingly more prominent during influenza seasons. Influenza B infection is typically considered a mild disease and receives less attention than influenza A, but has been causing 20 to 50 % of the total influenza incidence in several regions around the world. Although there is increasing evidence of mid to lower respiratory tract diseases such as bronchitis and pneumonia in influenza B patients, little is known about the pathogenesis of recent influenza B viruses. Here we investigated the clinical and pathological profiles of infection with strains representing the two current co-circulating B lineages (B/Yamagata and B/Victoria) in the ferret model. Specifically, we studied two B/Victoria (B/Brisbane/60/2008 and B/Bolivia/1526/2010) and two B/Yamagata (B/Florida/04/2006 and B/Wisconsin/01/2010) strain infections in ferrets and observed strain-specific but not lineage-specific pathogenicity. We found B/Brisbane/60/2008 caused the most severe clinical illness and B/Brisbane/60/2008 and the B/Yamagata strains instigated pathology in the middle to lower respiratory tract. Importantly, B/Brisbane/60/2008 established efficient lower respiratory tract infection with high viral burden. Our phylogenetic analyses demonstrate profound reassortment among recent influenza B viruses, which indicates the genetic make-up of B/Brisbane/60/2008 differs from the other strains. This may explain the pathogenicity difference post-infection in ferrets.


INTRODUCTION
The influenza virus is a constant health concern. Influenza B infections average between 20-30 % of total influenza incidence but can be 50 % or greater depending on the season (Paul Glezen et al., 2013). Two influenza B lineages, B/Yamagata/16/88 (B/Yam) and B/Victoria/2/87 (B/Vic), have been co-circulating since the 2009H1N1 pandemic (WHO, 2014. There are also significant differences in the evolution and epidemiology between influenza A and B viruses. Influenza A infects a broad range of natural vertebrate hosts which gives rise to various subtypes whereas humans are the primary host of influenza B viruses (Ohishi et al., 2002;Osterhaus et al., 2000). For influenza B viruses, a narrow host range is thought to limit subtype diversity and gene reassortment; furthermore, the rate of antigenic drift has been estimated to be slower than for A viruses (Chen & Holmes, 2008;Lindstrom et al., 1999;McCullers et al., 1999). As a result, influenza B viruses are thought to evolve slower than influenza A viruses; there has been no evidence of influenza pandemic caused by a B virus (Cheng et al., 2012;Hay et al., 2001;Paul Glezen et al., 2013).
Influenza B viruses are often considered to cause mild clinical disease predominantly in paediatric and geriatric patients (Chan et al., 2013;Chi et al., 2008;Li et al., 2008;Wang et al., 2008). Despite this, several lines of evidence exist suggesting these viruses are also capable of causing significant human disease. Comparative human clinical reports suggest that influenza B infection may induce severe clinical and inflammatory responses comparable to influenza A (Kaji et al., 2003;Sakudo et al., 2012). Influenza B virus dominant seasons have also been documented such as the 2002-2003 season when the prevalence of influenza B infection was higher than that of influenza A infection in most countries around the world.
Ferrets are a suitable model to study influenza infection as they manifest human-like flu symptoms and immune responses (Banner & Kelvin, 2012;Hamelin et al., 2010;Hause et al., 2013;Huang et al., 2013;Leó n et al., 2013). Previously, we found distinct clinical and immunological patterns between ferrets infected with B/Vic and B/Yam lineages, where B/Vic appeared to be the more virulent lineage (Huang et al., 2011). Here we investigated influenza B virus infection in ferrets utilizing four unique strains, two B/Yam (B/Florida/04/2006 and B/Wisconsin/01/2010) and two B/Vic viruses (B/Brisbane/60/2008and B/Bolivia/1526, that are representative of circulating B viruses. We then analysed clinical characteristics, viral kinetics and histopathology throughout the respiratory tract. Phylogenetic analysis of representative influenza B strains from the past 30 years was employed to understand the genetic relationship of these viruses. These findings will help to characterize the clinical severity of influenza B and establish the ferret influenza B model. This work provides insight into the development of appropriate influenza therapeutic strategies.

Distinct clinical features among influenza B infections in ferrets
A clear clinical and pathological picture of recent B/Yam and B/Vic lineage infections has not yet, to our knowledge, been described. Here we studied the pathogenicity of B viruses to enhance the understanding of disease progression and strain variation. We infected healthy male naive adult ferrets with either B/Florida/04/2006 (B/Fla), B/ Wisconsin/01/2010 (B/Wisc), B/Brisbane/60/2008 (B/Bris) or B/Bolivia/1526/2010 (B/Bol). We monitored daily temperature and weight changes, nasal symptoms, sneezing and inactivity level for 7 days post-infection (pi).
Generally, influenza B infected animals experienced fever of a 2 day duration peaking on day 2 pi (except B/Wisc) (Fig.  1a). No significant difference was found between the temperature peak values. Conversely, kinetics of weight change were more distinct among infections (Fig. 1b). B/Wisc, B/Fla and B/Bol infected animals experienced weight loss (peak values: 4.5, 3.1 and 2.1 %, respectively) for 1 or 2 days pi then began recovery. In contrast, B/Bris infected animals had a biphasic weight loss curve and began to lose weight 1 day pi, peaking on day 2 pi (6 %), then experienced a secondary weight loss on day 4 pi (4 %). Nasal discharge was prominent among all infections, with the highest amount found in B/Bris infected ferrets (6/8 animals) (Fig. 1c). Sneezing was not readily observed and the activity level was not affected. B/Bris infected animals displayed minimal inactivity (not significantly different from the other viruses). The time-courses pi of nasal discharge, sneezing and inactivity level are included in Fig.  S1 (available in the online Supplementary Material). Nasal discharge peaked 2 to 3 days pi, which followed the fever kinetics (1 to 3 days pi). Impact on the activity of the B/Bris infected ferrets was observed on day 3 pi, which followed fever and preceded the second peak weight loss. No prominent sneezing was observed in this study. In summary, we found B/Bris virus (B/Vic) gave the highest morbidity (most weight loss and longest duration) while B/Bol (B/Vic) induced the mildest clinical illness, suggesting strains from the same lineages do not necessarily impose a similar host clinical impact.

B/Bris virus replicated in the ferret lower respiratory tract
Severe influenza disease has been associated with the development of lower respiratory tract infection (Taubenberger & Morens, 2008). Since ferrets and humans have similar distribution of influenza receptors in their respiratory tracts (Jayaraman et al., 2012;Maher & DeStefano, 2004;Suzuki et al., 2000;van Riel et al., 2007), ferrets are a suitable model to evaluate human influenza replication in vivo. Here, we measured viral titres in nasal wash, trachea and lung to represent viral burden in the upper, middle and lower respiratory tract during influenza B virus infection.
From the titre kinetics in nasal washes, all influenza B viruses readily replicated in the nose, although analyses of the middle and lower respiratory tract found that only B/Bris virus was able to significantly infect the lower reaches of the respiratory tract (Fig. 2). Besides replication location, there was also a difference in virus levels. Nasal wash viral titres of B/Bris were significantly higher than the titres of B/Wisc and B/Bol on day 3 (40-and 70-fold higher than B/Wisc and B/Bol, respectively) and day 5 pi (80-and 10-fold higher than B/Wisc and B/Bol, respectively) (Fig.  2a). The virus levels between B/Bris and B/Fla were not significantly different. Importantly, B/Bris trachea and lung virus levels were markedly high and the other influenza B infections were not detectable on either day 3 or 7 pi (Fig. 2b,  c). A similar trend was also observed when using real-time PCR assays, where B/Bris had the highest viral RNA (vRNA) copy number 3 days pi, although slight vRNA levels were also detected in B/Fla and B/Wisc (Fig. S2). vRNA was not detected in the B/Bol samples. These results suggested that while influenza B viruses were mainly found to be an upper respiratory tract pathogen in ferrets, there were differences in replication efficiency in the upper respiratory tract, and B/Bris replicated efficiently in the lower respiratory tract.

Influenza B infection induced strain-specific lung pathology
Our results from viral titre kinetics demonstrated that influenza B viruses were not confined to the upper respiratory tract. Our previous study with ferrets demonstrated the B/Yam lineage was able to induce pathology in the bronchioles (Huang et al., 2011); here we expanded our histopathological studies.   animals (3 days pi) appeared similar to the mock-infected control. By day 7 pi, epithelial desquamation (red ellipses) and leukocyte infiltration (blue and red arrows) in the bronchioles of B/Wisc and B/Bris infected animals were still observed (Fig. 3b, right panel). Furthermore, epithelial cell sloughing was also noted in B/Fla infection by day 7 pi (red ellipse). Conversely, the bronchioles of the B/Bol infected animals appeared normal and similar to the mockinfected animals.
We also investigated the lowest end of the respiratory tract in ferrets for evidence of pathology. We observed early (3 days pi) small airway histological alteration in B/Fla, B/ Wisc and B/Bris infected animals (Fig. 3c, left panel). Both B/Fla and B/Wisc infection caused mainly multifocal interstitial pneumonia determined by the thickened lining of the alveolar wall (Fig. 3c, green arrows). B/Bris infection, however, not only caused interstitial pneumonia (Fig. 3c, green arrows) by day 3 pi but also induced neutrophil and mononuclear cell infiltration to the alveolar lining and lumens (blue and red arrows respectively) and epithelial shedding (red ellipses). The severity in the alveoli had improved by day 7 pi for all strains except the B/Bris infection ( Fig. 3c, right panel). The alveoli in the B/Fla and B/Wisc infected animals appeared normal and similar to the mock-infected control animals. Nevertheless, interstitial pneumonia (Fig. 3c, green arrows), cell infiltration (blue and red arrows), and epithelial shedding (red ellipses) were still observed in the alveoli of B/Bris infected animals. Alveoli of the lungs of B/Bol infected animals did not incur pathological changes throughout the time-course. Together, the four influenza B strains imposed differential pathology where B/Bris infection gave the most severe and prolonged impact, which included the lower reaches of the respiratory system while B/Bol virus induced the least damage.

Viral segments layout in B/Bris, B/Bol, B/Flo and B/Wis
From above, influenza B infection appeared to trigger a distinct strain-specific disease pattern and severity. Therefore, we hypothesized that each strain was composed of a unique set of gene segments leading to variability in clinical outcomes. To characterize the genetic make-up of the four influenza B strains within the context of the pool of circulating influenza B viruses, the sequences from the genomic segments of B/Bris, B/Bol, B/Flo and B/Wis, together with the sequences of 660 influenza B strains that were isolated between 2006 and 2010, were grouped by maximum-likelihood clustering (Fig. 4) Analyses of the internal protein segments suggested further reassortment events between influenza B viruses of recent years, although all descending from B/Yam lineage, which may explain the in vivo differences (Fig. 5b) (Jonsson et al., 2012). They found the biphasic weight loss might be associated with the migration of viral infection between different lobes of the lung, which may also occur during B/Bris infection when establishing lower respiratory tract infection. In our previous study, we found the kinetics of lung pathology correlated with influenza--specific antibody responses measured by haemagglutination inhibition (HI) assay (Huang et al., 2011). Taking together the HI kinetics from this study (Fig. S3) and the previous studies, we found B/ Bris infection induced higher HI levels on day 14 pi and broader lung pathology by day 7 pi than other influenza B infections. It is possible the persistence of lung pathology from days 3 to 7 pi caused by B/Bris virus stimulated a higher HI response at peak level. Pulmonary abnormality in B/Bol infected ferrets was minimal and correlated with absence of active viral replication in the lower lung and minimal weight loss. As suggested from previous work (Huang et al., 2011(Huang et al., , 2012Marcelin et al., 2011;Wu et al., 2012), it is possible that high viral burden might not be the only attributing feature for severe clinical disease where the resulting destruction may be promoted by increased activated immune responses. However, this study demonstrated a wider distribution of efficient influenza B virus replication for B/Bris infection which reached the lower respiratory system leading to severe disease (Gill et al., 2010;Harms et al., 2010;Mauad et al., 2010;Nakajima et al., 2012;Shieh et al., 2010).
Our phylogenetic analyses demonstrated an on-going complex evolution where antigenic drift of type B viruses and gene reassortment continue to take place, changing the genetic composition of circulating influenza B strains. Our phylogenetic analyses suggested that the studied B/Vic lineage viruses are progeny viruses of reassortant Bris/02 strain and the HA and NA segments are derived from Shangdon/97 (B/Vic strain) and Sichuan/99 (B/Yam strain) respectively. Furthermore, these analyses indicated several internal protein segments (PA, M and NS) in Bris/02 derived from Sichuan/99, implying these segments have probably been reassorted during the same period. The respective opposing B/Vic segments might have been replaced during early 2000 or circulated at low frequency after the dominance of HK/01-like strains. The genetic relationship of influenza B NP, M and NS may be associated with the unique behaviours of the studied strain infections in vivo. It has been shown that in the time period when B/Bris and B/Bris-like viruses circulated, there were unusual influenza B outbreaks and patients positive for influenza B viruses developed lower respiratory tract illness such as bronchitis and pneumonia (Esposito et al., 2011;Gutiérrez-Pizarraya et al., 2012). Other studies have demonstrated that the NS segments in B strains during the late 1990s were not derived from B/Yam or B/Vic but from an earlier strain isolated in 1984 (Lindstrom et al., 1999;Luo et al., 1999;McCullers et al., 2004). Our findings present important evidence suggesting the pathogenic potential of B viruses and the potential clinical impact if a B/Bris-like virus were to emerge in a naive population.
The impact of influenza on global health and the economy is not limited to the effects of type A viruses. Although there has not been a record of an influenza B pandemic, B viruses have remained prominent and cause occasional multi-regional outbreaks. Our study demonstrated that influenza B infection with B/Bris induced moderate to severe morbidity in naive healthy male adult ferrets. It is possible that healthy adults and not only young children and the elderly are susceptible to severe influenza B infection such as the influenza B strain B/Bris (Chi et al., 2008;Wang et al., 2008). The rare incidence of severe disease caused by influenza B in adults may be due to adult pre-existing immunity from previous encounters with similar influenza B viruses (Paul Glezen et al., 2013). Ferret infection and monitoring. Maintenance and monitoring of ferrets upon infection followed previously published protocols (Huang et al., 2011). Ferrets were randomly selected and pair-housed in cages contained in bioclean portable laminar-flow clean-room enclosures (Lab Products) in a biosafety level 2 facility (BSL-2). Inactivity level, nasal discharge, sneezing, body temperature [measured with subcutaneous implantable temperature transponder (BioMedic Data Systems)] and weight of animals were measured on day 0 prior to infection and each day thereafter. For infection, each ferret was anaesthetized and infected intranasally with 1 ml inoculum (0.5 ml to each nostril) at the dosage of 10 6 50 % egg infectious dosage (EID 50 ). Viruses were diluted in saline buffer before infection. All virus preparation work was conducted in a BSL-2 facility. Nasal discharge, prominence, type, colour, and location of crusty nose, mucus and exudates/fluids were recorded. Scores were calculated daily and peak values for each infection are summarized (Fig. 1, panel  c). The inactivity index of each viral infection was calculated using the scores observed daily. The scoring protocol was adapted from that described by Reuman et al. (1989): 0, alert and playful; 0.5, alert but playful only when given incentives; 1, alert but not playful when given incentives; 2, neither alert nor playful even when given incentives. To calculate the index, a value of 1 was first added to each daily score of each animal followed by the summation of all the values from all infected animals from day 1 to 7 pi. The total number was divided by the total observation incidence (i.e. the total number of animals multiplied by 7 days), rendering the index value: n equals the total number of observations from day 1 to 7 pi. The value of 1 added to each observation score resulted in a minimum value of 1.000 as an inactivity index for each infection.
Viral load. The procedure was as previously described (Huang et al., 2012). Viral loads in the upper, middle and lower respiratory tract were determined by end point titration of nasal washes or respiratory tissue homogenates on MDCK cells followed by haemagglutination of turkey erythrocytes (Lampire Biological Laboratories) with MDCK supernatant to calculate 50 % TCID 50 . Nasal washes from infected ferrets were collected using 1 ml of nasal wash buffer (1 % BSA, 100 U penicillin ml 21 , 100 mg streptomycin ml 21 in PBS). Tracheas and lungs were collected on 3 and 7 days pi from euthanized ferrets. Nasal washes and tissue homogenates were titrated in vDMEM [Dulbecco's modified Eagle's medium (DMEM) containing 1 % BSA, 25 mM glucose, 1 mM sodium pyruvate, 4 mM glutamine, 100 U penicillin ml 21 , 100 mg streptomycin ml 21 , 50 mg gentamicin ml 21 and 1 mg TPCK-trypsin ml 21 ] and incubated on MDCK cells at 37 uC, 5 % CO 2 . After 2 h, supernatant was aspirated and replaced with fresh vDMEM followed by 6 days at 37 uC, 5 % CO 2 . Supernatant was examined for presence of virus by haemagglutination of turkey erythrocytes. The viral titres were determined as the reciprocal of the dilution resulting in 50 % HA positivity as log 10 (TCID 50 ml 21 ). All cell culture reagents were obtained from Invitrogen except for TPCKtrypsin (Sigma-Aldrich) and BSA (Wisent).
Histopathology. Tissue samples were formalin-fixed after harvesting from euthanized ferrets. Formalin was used to perfuse lung samples. Tissues were embedded with paraffin wax and sectioned. Haematoxylin and eosin staining was performed for histopathological assessment. Mock-infected male (age-matched) ferret tissues were derived from animals treated with PBS intranasally (Huang et al., 2011). To extract viral sequences, the resulting short-reads were aligned to the genomic segments of the closely related B/Brisbane/60/2008 (Table S1) using the DNA aligner Bowtie2 program v2.0.6 (Langmead & Salzberg, 2012). Finally, the consensus sequences were generated with the Samtools program v0.1.12 (Li et al., 2009b).
Phylogenetic analyses. Phylogenetic analyses of all nucleotide sequences were conducted in MEGA v6.0.2 software (Tamura et al., 2011) using maximum-likelihood method based on the Tamura-Nei model (Tamura & Nei, 1993) with bootstrap replications (n51000) to determine the best-fitting tree for each segment. Table S1 summarizes the accession numbers of the sequences obtained from GISAID or GenBank (http://www.ncbi.nlm.nih.gov/nuccore), labelled with *.
Statistics. One-way analysis of variance (ANOVA) was conducted to compare the temperature and weight change profiles during the four strain infections in Fig. 1 and viral burden kinetics in Fig. 2. If statistically significant difference (p,0.05) was found on a particular day pi, Bonferroni-Holm analysis (post-hoc analysis) was then applied between the results of B/Bris infection and the other three infections in a pairwise manner.
Donnelly Sequencing Centre for their help. The authors gratefully acknowledge the originating and submitting laboratories who contributed their sequences to the GISAID EpiFlu database. All submitters of data may be contacted directly via the GISAID website (http://www. ncbi.nlm.nih.gov/nuccore).