NB protein does not affect influenza B virus replication in vitro and is not required for replication in or transmission between ferrets

The influenza B virus encodes a unique protein, NB, a membrane protein whose function in the replication cycle is not, as yet, understood. We engineered a recombinant influenza B virus lacking NB expression, with no concomitant difference in expression or activity of viral neuraminidase (NA) protein, an important caveat since NA is encoded on the same segment and initiated from a start codon just 4 nt downstream of NB. Replication of the virus lacking NB was not different to wild-type virus with full-length NB in clonal immortalized or complex primary cell cultures. In the mouse model, virus lacking NB induced slightly lower IFN-a levels in infected lungs, but this did not affect virus titres or weight loss. In ferrets infected with a mixture of viruses that did or did not express NB, there was no fitness advantage for the virus that retained NB. Moreover, virus lacking NB protein was transmitted following respiratory droplet exposure of sentinel animals. These data suggest no role for NB in supporting replication or transmission in vivo in this animal model. The role of NB and the nature of selection to retain it in all natural influenza B viruses remain unclear. Received 7 October 2015 Accepted 23 December 2015


INTRODUCTION
Seasonal influenza illness results from infection of the upper respiratory tract epithelium by influenza A or B viruses. Influenza B viruses are generally less well studied than influenza A viruses and do not give rise to pandemics as they have no sustained animal reservoir (Jackson et al., 2011). Nonetheless, influenza B viruses cause a significant number of annual influenza cases and the divergence of two antigenically distinct lineages of influenza B virus has necessitated the recent inclusion of two influenza B strains into the annual vaccine, resulting in a quadrivalent product (Belshe, 2010;Rota et al., 1990).
Influenza A and B viruses are orthomyxoviruses; both have two glycoproteins on the surface of the viral membrane, haemagglutinin (HA) and neuraminidase (NA), which are involved in viral entry, by binding sialic acid receptors, and release of progeny viruses, by cleaving sialic acid, respectively. The M2 protein is a third, minor membrane component of the influenza A virus particle. M2 is an ion channel protein and has several roles in the influenza A replication cycle, including disassembly, assembly and scission, as well as modification of the autophagy pathway and activation of the inflammasome (Gannagé et al., 2009;Hughey et al., 1995;Ichinohe et al., 2010;Iwatsuki-Horimoto et al., 2006;McCown & Pekosz, 2005;Rossman et al., 2010). Influenza B viruses encode a homologue known as BM2 with an ion channel similar in function to that of influenza A M2, required for uncoating and with a role in assembly of the influenza B virus particle (Horvath et al., 1990;Imai et al., 2004;Pinto et al., 1992). Influenza B viruses uniquely encode a fourth surface glycoprotein, the NB protein (Shaw et al., 1983). NB is encoded by segment six of the influenza B virus, the same segment that encodes NA. The NB open reading frame overlaps that of NA, and the two AUG start codons are separated by only four nucleotides, with the NB start codon being closer to the 59 end of the mRNA (Shaw et al., 1983).
NB is an integral membrane protein of 100 amino acids, found on the surface of infected cells (Brassard et al., 1996). It has an 18 amino acid extracellular N terminus, a 22 amino acid transmembrane domain and a 60 amino acid cytoplasmic C terminus. There are two glycosylation sites on the extracellular domain that acquire high mannose sugars during post-translational processing.
Multiple N-acetyl-lactosamine residues (Galb1R4-GlcNAc b1R3) are added to these carbohydrates to form complex structures, termed polylactosaminoglycans (Williams & Lamb, 1986, 1988. The NB glycoprotein forms a tetramer at the infected cell surface and is incorporated in the virion during budding but only in relatively small numbers (approximately 15-100 molecules per virion) (Betakova et al., 1996;Brassard et al., 1996).
The function of NB is currently unknown. The similarity of domain organization of NB and M2 proteins originally led to suggestions that NB may act as the functional homologue for the influenza A virus M2 ion channel and data from voltage conductance studies supported this notion (Betakova et al., 1996;Fischer et al., 2000Fischer et al., , 2001Premkumar et al., 2004;Sunstrom et al., 1996). However, the discovery that BM2, a protein encoded on influenza B virus RNA segment seven, served as the influenza B virus ion channel discounted this. Therefore, it is likely that NB serves an alternative function for influenza B virus. Hatta and Kawaoka used reverse genetics to produce recombinant influenza B viruses that lacked NB, indicating that the protein is not essential for virus viability. Indeed, viruses lacking NB replicated as well as WT virus in Madin Darby Canine Kidney (MDCK) cells. Conversely, in BALB/c mice the replication of the mutant viruses was restricted compared to the WT (Hatta & Kawaoka, 2003).
To engineer the mutations that abrogated NB expression, Hatta and Kawaoka disrupted sequences around and including the NB AUG start codon. However, given the close proximity of the start codon of NB to that of the essential viral protein NA, it was not clear whether the attenuated phenotype observed in mice for the NB deletions might result from an effect on NA expression (Williams & Lamb, 1989).
We therefore wanted to pursue the investigation of the role of the NB protein by using an NB mutant with intact NA gene expression. To achieve that, a novel mutant virus, D5NB, was generated. This virus had a premature stop codon after the fifth amino acid of the NB sequence leaving the sequence of NA and the upstream nucleotides that control NA translation levels unchanged. Using this viral mutant, we sought to determine a role for NB in vitro or in vivo.

RESULTS
Generation of a recombinant influenza B virus lacking expression of the NB protein Segment six of a 12 plasmid reverse genetics system based on B/Florida/04/2006 (Elderfield et al., 2015) was modified to create a truncated NB protein. The D5NB virus was engineered by the introduction of a premature stop codon after the fifth amino acid of NB coding sequence by site-directed mutagenesis. The NB open reading frame was reduced to five amino acids whilst the NA amino acids remained unchanged (Table 1). Both WT and D5NB viruses were rescued and propagated in MDCK cells. We confirmed the lack of NB expression in cells infected with the D5NB mutant virus by Western blot of lysates from infected cells using an NB specific monoclonal antibody. This detected a heterogeneous range of NB proteins in lysates of cells infected with WT virus, ranging in size from 15 to 46 kDa as previously described (Williams & Lamb 1988), that were lacking in similar lysates from D5NB mutant virus-infected cells. Detection of equivalent amounts of BM2 and NP showed that levels of other viral proteins did not differ between the two viruses ( Fig. 1a).
The D5NB mutation did not affect virus neuraminidase activity To confirm that the mutations introduced in D5NB virus did not affect NA, enzyme activity associated with the mutant virus was measured using three different assays.
For the standard NA enzyme assay that employs a soluble substrate, the MuNANA assay, the WT and mutant viruses were standardized to equivalent p.f.u. and assessed over a range of dilutions for their ability to cleave the MuNANA reagent. There was no significant difference between the two viruses ( Fig. 1b).
In the red blood cell (RBC) elution assay, we assessed the virus' ability to desialylate red blood cells. The two viruses were mixed with human RBCs at equal HA titres and incubated at 4 uC for 1 h to allow agglutination. The plates were then shifted to 37 uC for 6 h to permit NA activity and observed at regular intervals for the reversal of Collectively, these results demonstrate that the D5NB mutation did not affect NA activity. . After 24 h, the cells were lysed, the proteins separated by SDS-page and transferred by Western blot. Membranes were immunostained with a mouse anti-NB protein antibody (top panels) and a rabbit anti-BM2 protein antibody (bottom panels). The blot was detected using the LI-COR system and the two panels separated for clarity. (b) MuNANA enzyme assay to assess NA activity. WT or mutated (D5NB) influenza B/Florida/04/2006 viruses were serially diluted in MES buffer, MuNANA reagent was added, the reaction was allowed to continue for 1 h prior to the addition of stop solution. The plate was read with UV excitation filter of 355 nm and emission filter of 460 nm. (c) RBC Elution Assay to assess NA activity. Influenza B/Florida/04/2006 WT or mutant (D5NB) virus were serially diluted in PBS. 50 ml of 0.5 % fresh human RBCs were added and the plate was incubated at 4 8C for 1 h before a temperature shift to 32 8C for 4 h. The plates were observed hourly for evidence of haemagglutination of RBCs until the end point at 6 h. (d) Mucus inhibition assay to assess NA activity. One hundred p.f.u. influenza B/Florida/04/2006 WT or mutant (D5NB) virus were pre-incubated in human airway mucus or virus diluent prior to infection on MDCK cells which after 1 h, after the inoculum was removed were overlaid with agar media to allow plaques to form. The percentage reduction in plaques formed after the mucus incubation was calculated.

D5NB influenza B virus displayed no growth attenuation in a variety of cell types
In order to measure any effect of loss of NB protein on virus replication kinetics we used human airway-derived Calu-3 cell cultures or primary HAE cell cultures, the latter containing a mixture of ciliated and non-ciliated cells types, with an air interface at the apical surface onto which mucus is secreted. In initial experiments each cell system was inoculated by each virus type in separate wells at an equal m.o.i. and the comparative growth was assessed at set time points by plaque assay on MDCK cells. Both systems were highly permissive for influenza B virus replication with peak titres in excess of 10 7 p.f.u. ml 21 in Calu-3 and 10 8 p.f.u. ml 21 in HAE cultures. D5NB virus was not attenuated in either cell system (Fig. 2a, b). Consensus whole genome sequencing of virus generated during multi-step replication confirmed no sequence variation between the WT and D5NB viruses, except for the specifically introduced truncation.
Competition assays have increasingly been used to assess small variations in fitness between two influenza viruses not observable by traditional viral growth curves (Brookes et al., 2011;Guarnaccia et al., 2013;Yen et al., 2014). To establish whether lack of NB has a fitness cost for the D5NB mutant virus compared with the WT virus, we infected HAE cells with a mixture of 20 % WT and 80 % D5NB virus at a final m.o.i. of 0.001. After multi-cycle replication over several days, the relative abundance of the two genotypes in virus harvested from the apical surface was measured using a pyrosequencing reaction. At no time point was there a statistically significant difference in the ratio of WT to D5NB from the ratio of genotypes in the input inoculum; after 4 days replication the D5NB genotype was still present at 80 % of the RNA population (Fig. 2c). These results indicated that the NB protein did not offer a selective advantage in vitro even in a complex cell culture system in which deficiency in NA is readily measured (Brookes et al., 2011). The D5NB recombinant virus is not attenuated in mice in vivo C57BL/6 mice were infected intranasally with 4.5|10 5 p.f.u. of either WT or D5NB virus. The weight loss of the mice was monitored daily. Viral titre, IFN-a and IL-1b in the lungs were determined on days 2, 3 and 4 post-infection. As shown in Fig. 3(a), the two viruses replicated well in mouse lungs, with the D5NB displaying slightly higher growth 3 days post-infection (P50.007) (Multiple t-test, Holm-Sidak method). As a gross measurement of pathology, weight loss was monitored (Fig. 3b). On all days, mice infected with the WT virus displayed a slightly greater weight loss but the difference between the two groups was not significant (Multiple t-test, Holm-Sidak method).
As an indication of the innate immune response, IFN-a in the mouse lungs was measured at days 2, 3 and 4 postinfection. Significant production of IFN-a in the murine lungs was only evident on day 2 after infection. There was no difference in the interferon levels in the mice infected with WT virus or those infected with the D5NB virus, except a slight increase for the D5NB infected mice on day 4 (P50.005) (Multiple t-test, Holm-Sidak method) (Fig. 3c). Activation of the inflammasome was assessed by measuring levels of IL-1b. Levels of pro-IL1b and total IL-1b (pro form plus activated form) were measured. Both sets of infected mice had higher total IL-1b levels in their lungs than the mock infected animals and this was significant at all days measured (Multiple t-test, Holm-Sidak method). Total IL-1b protein levels tended to be higher in the lungs of mice infected by D5NB virus, but this difference was not significant.
Virus lacking NB expression replicates in ferrets and is transmitted by the respiratory droplet route We tested whether the loss of NB affected replication within a single host or transmission onwards to a new host by inoculating ferrets with a mixture of WT and D5NB viruses, and monitoring the ratio of genotypes within and between hosts using pyrosequencing. Four donor ferrets were inoculated with a 1 : 1 ratio of a total of 10 4 p.f.u. WT and D5NB virus. All four donors were robustly infected with the virus mixture and shed infectious virus in their nasal wash over the following 6 or 7 days. Peak viral titres were shed on day 2 and were between 10 5 and 10 6 p.f.u. ml 21 nasal wash. Shedding followed a biphasic kinetic profile similar to that we have previously reported for ferrets infected   (Fig. 4a). Pyrosequencing of RNA isolated from the daily nasal washes showed that the relative proportions of the two virus genotypes did not vary by more than 11 % from the starting mix in any of the four directly inoculated hosts and the direction of the small variation was sometimes to favour wild-type genomes encoding full NB and sometimes D5NB mutant genomes (Fig. 4b, c).
Only one of the four respiratory droplet (RD) exposed sentinel animals acquired infection following exposure to the infected donor. Pyrosequencing of the nasal washes from this animal revealed that the virus it had acquired was uniquely the D5NB genotype (Fig. 4b). This infected recipient continued to shed D5NB virus for 4 days from day 6 to day 9 after it was first exposed. Peak titre in nasal wash from this animal was 10 5 p.f.u. ml 21 .

DISCUSSION
The NB protein is absolutely conserved across influenza B viruses, a finding that might indicate it is required for effective replication or transmission. However, our ability to rescue and propagate a recombinant influenza B virus that lacks expression of NB supports the previous conclusion from Hatta et al. (Hatta & Kawaoka, 2003) that NB is dispensable for virus replication in vitro. The design of our mutant virus allowed us to investigate the unique role of NB because we abrogated NB expression with no concomitant effect on the NA function or activity.
Although the set of experiments we describe here does not elucidate the function of NB, it does show that NB is not required for replication or transmission of influenza B virus in vitro or in vivo in mice or ferret models.
Previously, loss of NB was correlated with attenuation of replication in vivo in the mouse lung and nasal turbinate and an increase in MLD 50 , but it could not be excluded that this phenotype arose from a change in NA expression. In our study with an NB mutation that did not compromise NA, there were minimal differences in viral titre and weight loss and only slightly higher levels of IFN-a in the mice infected with the truncated NB protein.
We looked carefully at the levels of IL-1b that result from activation of the inflammasome response because other small hydrophobic viral proteins have been implicated in regulation of inflammasomes (Triantafilou et al., 2013). For example, deletion of the SH protein of respiratory syncytial virus, RSV, which is similar in size and domain distribution to NB, led to increased induction of proinflammatory cytokines including IL-1b, and deletion of SH is an attenuation strategy for bovine and human RSV vaccines under development (Karron et al., 2005;Taylor et al., 2014). On the other hand, the ion channel proteins of other viruses such as M2 of influenza A and E protein of the coronavirus SARS, appear to stimulate the inflammasome and their deletion leads to a decreased IL-1b response (Ichinohe et al., 2010;Nieto-Torres et al., 2014). In our experiments, loss of NB expression had a marginal effect on IL-1b expression in the mouse lung. We cannot exclude that a role for NB in modulating inflammasome activation was masked in the mouse model by other factors, or that the effects of NB are mouse strain, host species or cell type specific for this human pathogen.
Very recently, Kim et al. published the first study of transmission of influenza B virus in ferrets (Kim et al., 2015).
In line with our data, they also report low transmission efficiency of the B/Florida/04/2006 virus in the ferret model, in contrast to a more efficient influenza B virus RD transmission previously reported between guinea pigs (Pica et al., 2012). The frequency of transmission events we recorded between RD exposed ferrets was surprisingly low, at just one of four exposed animals acquiring infection. We do not yet know if this observation will prove common to a number of influenza B viruses or whether transmission efficiency varies amongst different influenza B virus strains. The influenza B virus we used replicated well in the upper respiratory tract of ferrets and in well differentiated cultures of ferret airway cells (Elderfield et al., 2015). The low transmission efficiency of a seasonal influenza virus suggests either that this animal model does have limitations for the study of influenza B viruses, or that some successful human influenza viruses are not necessarily transmitted primarily by this RD route. As far as the NB protein is concerned, we found that in the single transmission event we did record, the virus that transmitted was uniquely of the genotype that lacked expression of NB. Because we set up mixed infections and both mutant and wild-type virus were replicating simultaneously in the donor animals we cannot absolutely exclude that the D5NB virus genome was not complemented by wild-type NB protein during the transmission event. However, our data unequivocally show that NB is not required for replication in the ferret model.
A recent paper also used a reverse genetics approach to probe for a role for NB. Rather than creating a completely null mutant, they chose to mutagenize the cytoplasmic tail of the protein and in particular abrogate the post-translational modification of palmitoylation at residue 49 from cysteine to serine (Demers et al., 2014). Surprisingly, in view of our data and that of Hatta et al. that suggested no role in vitro for NB, Demers et al. found that their palmitoylation mutant was attenuated for replication in MDCK cells. It may be that the continued expression of an unnecessary and mutated membrane protein interfered with the normal assembly of virus particles, but whether this result verifies a role for the non-mutated NB protein in wild-type influenza virus infections in vitro is not clear.
Bearing in mind the absolute conservation of NB in all influenza B viruses isolated from clinical cases and even in strains such as influenza B/Lee/40 that have been passaged in eggs and cell culture for more than 75 years, it is highly likely that this protein does serve a role for the virus. However, using current technologies we have been unable to measure a fitness cost in viruses engineered to lack the protein either in vitro or in vivo. Virus sequencing. Viral RNA was extracted using the Qiagen QIAmp RNA kit. Reverse transcription was conducted using random hexamers and Superscript III (Invitrogen). DNA amplification was conducted using KOD polymerase (Novagen) and non-coding region segment specific oligomers. DNA bands generated were excised from agarose and cleaned using a Qiagen gel extraction kit. Sequencing was conducted by the GATC sequencing service using the previously mentioned oligomers and additional internal oligomers. The data generated was analysed by alignment using Geneious software. All oligomer sequences are available on request. The cells were then lysed and the proteins were separated on a 12 % SDS-PAGE gel and then transferred to a membrane. The membrane was blocked in Odyssey blocking buffer (LI-COR Biosciences) and then stained with a mouse anti-NB monoclonal antibody (a kind gift from Dhan Samuel PHE Colindale UK) and rabbit anti-BM2 (Jackson et al., 2004) and anti-NP primary antibodies and then with an anti-mouse Dylight 680 and anti-rabbit Dylight 800 conjugated secondary antibodies (Cell Signalling Technologies). The LI-COR detection system was used, allowing both NB and BM2 to be detected in the same gel lane, the image has been split to display both proteins in grayscale, an equal amount of lysate was loaded onto two additional lanes and stained with anti-NP.

METHODS
Pyroseqeuncing. Viral RNA was extracted from the HAE or ferret nasal wash using the Qiagen QIAmp RNA kit. A 50-cycle PCR was performed using the following primers: 59 GCCAAAAATGAACAA-TGCTACCT and 39 CTGATGTGAGAAATAGGGTTAACG. The amplified product was mixed with Sepharose beads at 1400 r.p.m. for 10 min. The beads were sequentially washed in 70 % ethanol, 0.2M NaOH and wash buffer and transferred to a plate containing the primer: AAAATGAACAATGCTACCT for sequencing on the Qiagen PyroMark Q96 ID pyrosequencer.