Zanamivir-resistant influenza viruses with Q136K or Q136R neuraminidase residue mutations can arise during MDCK cell culture creating challenges for antiviral susceptibility monitoring

but were less well supported in MDCK-SIAT1 cells and embryonated eggs. The effect of Q136K, Q136R, Q136H and Q136L substitutions in NA subtypes N1 and N2 on NAI susceptibility and in vitro viral fitness was assessed. This study highlights the challenges that cell culture derived mutations can pose to the NAI susceptibility analysis and interpretation and reaffirms the need to sequence viruses from respective clinical specimens to avoid misdiagnosis. However, we also demonstrate that NA mutations at residue Q136 can confer reduced zanamivir, peramivir or laninamivir susceptibility, and therefore close monitoring of viruses for mutations at this site from patients being treated with these antivirals is important.


Introduction
Neuraminidase inhibitors (NAIs) are a class of influenza antivirals that target the highly conserved enzymatic site of the neuraminidase (NA) glycoprotein on the surface of influenza A and B viruses [1]. The NAIs have become the most widely used antivirals for the treatment or prophylaxis of influenza, particularly since the development of widespread resistance to the adamantanes, the older class of antivirals that block the M2 ion channel protein [2]. Two NAIs, oseltamivir and zanamivir, have been available in many countries since 1999, and two new NAIs, peramivir and laninamivir, have recently been approved for human use in Japan and a small number of other countries. Each of the NAIs is structurally different and therefore binds slightly differently within the NA active site [1]. This difference in binding is advantageous for treatment, as a virus that develops resistance against one NAI, may retain sensitivity to others. For example, the H275Y NA mutation in NA subtype N1-containing viruses confers resistance to oseltamivir but not to zanamivir [3].
Resistance to the NAIs commonly occurs as a result of amino acid mutations within the NA active site, either in the catalytic residues (those that interact directly with the NAIs), or in the framework residues (those that provide structural support for the catalytic residues) [4]. However, not all viruses with resistance to NAIs will pose a public health risk, as mutations that reduce binding can also impact the ability of the NA to interact with the natural substrate during replication [5]. However, in some cases, the mutation can affect NAI sensitivity but not compromise viral 'fitness'. The H275Y mutation that was present in seasonal influenza A(H1N1) viruses between 2007 and 2009 is such an example, as it conferred oseltamivir resistance but did not appear to affect the ability of the virus to replicate and transmit [6,7].
Oseltamivir resistant influenza viruses have been detected at a considerably higher frequency than zanamivir resistant viruses. During human clinical trials, oseltamivir resistance was detected in < 1-4% of adults [8,9] and 5-6% of treated children [10] undergoing oseltamivir treatment, although in observational studies the frequency of resistance in oseltamivir treated children has been as high as 27% [11]. Most significantly, oseltamivir-resistant seasonal A(H1N1) viruses with an H275Y mutation became widespread during 2008, spreading globally even in regions of low drug usage [12,13]. In comparison, there have only been a few reports of zanamivir resistance. The first was in an immunocompromised child undergoing zanamivir treatment where an influenza B virus with a R152K NA mutation was detected that caused a 40-150-fold reduction in zanamivir sensitivity [14,15]. More recently a small number of A(H1N1)pdm09 viruses with an I223R NA mutation have been detected in immunocompromised patients exposed to oseltamivir and/or zanamivir [16,17] and in a patient without previous exposure to NAIs [18], but the change in zanamivir sensitivity as a result of this mutation is relatively minor (10-fold), compared with the larger 45-fold shift in oseltamivir sensitivity [16].
Previously, our group and others reported the detection of former seasonal A(H1N1) virus isolates with a Q136K mutation that conferred a 250-fold reduction in zanamivir sensitivity [19,20]. The Q136K isolates were particularly unusual because the mutation could not be detected in the clinical specimens from which they were derived. This suggested that either the variant virus was being generated in cell culture or it was present in very low levels in the clinical specimen and then selectively amplified during cell culture. The former seasonal A(H1N1) virus stopped circulating soon after the emergence of the A(H1N1)pdm09 virus in 2009, and for the first two years after the A(H1N1) pdm09 viruses started circulating, no Q136K variants were reported. However, here we report the detection of both Q136K and Q136R substitutions in A(H1N1) pdm09 and A(H3N2) viruses between 2011 and 2014, investigate their selection in different cell lines and determine the effect that these and other amino substitutions of the Q136 residue have on NAI susceptibility and NA enzymatic function.

Virus strains, Madin-Darby canine kidney epithelial cells and virus culture
The influenza viruses used in this study were received at the World Health Organization (WHO) Collaborating Centre for Reference and Research on Influenza (WHOCC), Melbourne, Australia through the WHO Global Influenza Surveillance and Response System (GISRS) from countries in the Asia Pacific region. The Q136K or Q136R isolates had initially been isolated and then passaged in MDCK cells in external laboratories before being received and repassaged one to two further times at the WHOCC. WHOCC MDCK cells were originally received from ATCC (CCL-34) and used at passage level 63 to 83 and grown in Dulbecco's modified Percentage of Q136R

Neuraminidase activity and neuraminidase inhibition assays
To determine NA activity, viruses were standardised to an equivalent haemagglutinin (HA) titre using turkey red blood cells, serially diluted (twofold) in assay buffer (32.5 mM 2-(N-morpholino)ethanesulfonic acid (MES) pH 6.5, 4 mM CaCl 2 with 0.1% nonyl phenoxypolyethoxylethanol (NP-40)), and then mixed with an equal volume (50 μL) of the fluorescence substrate 2-(4-methylumbelliferyl)-a-D-N-acetylneuraminic acid (MUNANA) (0.3 mM) (Sigma) before incubation at 37 °C for 60 min. The reaction was terminated by the addition of 100 μL of stop solution (0.14 M NaOH in 83% ethanol). Quantification of the fluorescent product 4-methylumbelliferone was determined using a Fluoroscan Ascent FL (Thermo) with an excitation wavelength of 360 nm and an emission wavelength of 448 nm. Viruses were tested in duplicate in three separate assays. The NA activity of the recombinant variant viruses was calculated as a relative percentage of the NA activity of the recombinant wildtype (WT) virus at the same virus dilution.
Viruses were tested for susceptibility to the NAIs laninamivir, oseltamivir carboxylate, peramivir, and zanamivir, which were kindly provided by Daiichi-Sankyo, Japan, Hoffman-La Roche Ltd, Switzerland, BioCryst Pharmaceuticals Inc., United States (US), and GlaxoSmithKline, United Kingdom (UK), respectively.
To determine the drug concentration required to inhibit 50% of the NA activity (IC 50 ), 50 μL of virus, diluted according to the NA activity assay, was mixed with varying concentrations of inhibitor in microtitre plates (FluoroNunc plates, Nunc). Final reaction mixture concentrations of the NAIs ranged from 0.01 nM to 10,000 nM. The virus/NAI mix was incubated at room temperature for 45 min before the addition of 50 μL of MUNANA substrate (0.3 mM) and then incubated at 37 °C for 60 min. The reaction was terminated by addition of 100 μL of the stop solution. The data were plotted as the percentage of fluorescence activity inhibited against the log NAI concentration. IC 50 values were calculated using the logistic curve fit programme 'Robosage' kindly provided by GlaxoSmithKline, UK.

Reverse transcription-polymerase chain reaction, sequencing and pyrosequencing
Reverse transcription-polymerase chain reaction (RT-PCR) was conducted using gene specific primers (sequences available on request) using the SuperScript III One-Step RT-PCR System with Platinum Taq DNA Polymerase Kit (Life Technologies) according to the manufacturer's instructions. Sequencing was conducted using the ABI Prism Dye Terminator III Cycle Sequencing Kit (Life Technologies) and analysed on the ABI 3500XL automated DNA sequencer (Life Technologies) and data aligned using the DNAStar Lasergene 8 software package. Pyrosequencing to determine the relative proportion of a Q136 variant in a viral population used the PyroMark ID System (QIAGEN). Briefly, 40 µL biotinylated PCR product was Residual activity is expressed as a percentage of activity relative to that at 37 °C. Mean values are presented and error bars indicate ± SD from three separate experiments.
immobilised on streptavidin-coated sepharose beads (QIAGEN), transferred using the PyroMark Vacuum Prep Workstation (QIAGEN) into 40 µL of PyroMark Annealing Buffer containing pyrosequencing primers and then subjected to pyrosequencing reactions using PyroGold Reagents (QIAGEN) following the manufacturer's protocol. The relative proportion of a Q136 variant in a sample was determined using the Allele Quantitation application of the PyroMark ID software. Single nucleotide (nt) mutations were introduced into the NA plasmids to alter the Q136 codon to residues H, K, L, or R using the QuikChange Multi Site Directed  Mutagenesis Kit (Agilent Technologies) and using mutagenesis primers designed according to manufacturer's guidelines and synthesised by GeneWorks (Adelaide, Australia). The NA segment was sequenced directly from the plasmid as described previously to confirm that the desired mutation had been introduced and that no additional mutations were present.

Serial passage experiments in Madin-Darby canine kidney epithelial cells and eggs
Recombinant viruses composed of the NA gene from one of the viruses described above, and the remaining seven segments from A/Puerto Rico/8/34 (A/PR/8/34 plasmids kindly provided by Dr Robert Webster, St. Jude Children's Research Hospital, Memphis) were generated by reverse genetics. All eight plasmids were transfected into a co-culture of 293T and MDCK cells as previously described [21]. Rescued viruses were subsequently cultured in MDCK cells in maintenance media described above.    Figure 1). This positive selection for the Q136K variant was also observed following passage of the A/ Brisbane/345/2011 isolate in the WHOCC MDCK cells, but was not seen with the other two isolates, where passage in WHOCC MDCK cells resulted in the gradual loss of the Q136K or Q136R variant. The MDCK-SIAT1 cells consistently selected against the Q136K or Q136R variants, with the proportion of each variant gradually decreasing after serial passage (Figure 1). The largest change in mixture proportion was seen following egg passage, which showed that growth of the variants were not well supported in embryonated eggs and were rapidly selected against, such that after a single passage in eggs the Q136K and Q136R viruses were undetectable in two of the isolates, and accounted for < 10% of the viral population in the third isolate ( Figure 1).

Neuraminidase activity and neuraminidase inhibitor susceptibility of reverse genetics derived Q136K, R, L and H variants
Examination (in 2014) of human and avian N1 and N2 sequences from the public sequence databases Global Initiative on Sharing Avian Influenza Data (GISAID) and GenBank revealed not only the Q136R and Q136K substitutions, but also Q136L and Q136H, present in a small number (less than 1%) of sequences from A(H3N2), A(H1N1) and A(H5N1) virus isolates. Site directed mutagenesis and reverse genetics were used to better investigate the phenotypic effect of the Q136K, Q136R, Q136L and Q136H substitutions in the NAs from A(H1N1)pdm09 and A(H3N2) viruses. 7:1 reassortants containing either the N1 or N2 WT NA (no mutations) or variant NA (Q136K, R, L or H) on a PR/8 backbone were successfully generated by reverse genetics.
For N1 reassortants, the Q136H mutant retained full NA activity, while the Q136R, Q136K and Q136L mutants had between 33 and 54% of the WT NA activity ( Table  2). For the same mutations in the N2 NA, the Q136H substitution also had no effect on NA activity, whereas the Q136R substitution caused a minor reduction in NA activity (82% activity of WT), and the Q136K and Q136L substitutions caused large reductions in NA activity (40% activity of WT) ( Table 2).
Analysis of the N1 reassortants for NAI susceptibility showed that the Q136H substitution had no effect, whereas the Q136L mutant demonstrated a moderate 32-fold increase in zanamivir IC 50 and a minor 4-to 12-fold increase in oseltamivir, peramivir and laninamivir IC 50 compared with the respective WT. In comparison, the Q136R substitution caused a 659-to 810-fold increase in zanamivir, peramivir and laninamivir IC 50 compared with WT, while the Q136K substitution caused a 126-to 589-fold increase against the same NAIs (Table 2). Both the Q136K and Q136R substitutions had no effect on oseltamivir susceptibility.
The large effect of the Q136R and Q136K substitutions observed in the N1 NA was not observed in the N2 NA. Q136R caused only a minor (two-to threefold) change in oseltamivir, peramivir and zanamivir IC 50 . A Q136L or a Q136K substitution caused a moderate 10-to 13-fold increase in zanamivir and oseltamivir IC 50 and a threefold increase in peramivir IC 50 but no change in laninamivir sensitivity compared with the WT. Q136H in N2 had no effect on NAI sensitivity, similar to that observed in the N1 NAs (Table 2).

Viral function of reverse genetics derived Q136K, R, L and H variants
The thermostability and HA/NA balance of the reassortant variant viruses were determined. The Q136L and Q136H N1 variants retained high NA activity across the 38 °C to 54 °C temperature range, equivalent to that of the WT virus. However the N1 reassortants with Q136K and Q136R mutations showed a substantial loss of activity at 54 °C (67% and 22% remaining activity compared to 37 °C respectively) (Figure 2A). The N2 reassortant WT had some loss of activity at 54 °C (30% remaining activity), as did the NAs with Q136R and Q136H mutations (8-21% remaining activity), while the NAs with Q136K and Q136L mutations maintained high activity (> 80%) across the entire temperature range ( Figure 2B).
All N1 and N2 reassortant viruses contained the PR/8 HA and all showed good cell binding at 4 °C, as indicated by low viral titres in the supernatant. After incubation at 37 °C, the N1 WT reassortant and the Q136L variant showed full restoration of HA titre, demonstrating an active NA enzyme, whereas the Q136K and Q136R variants had only partial restoration of HA titre, suggesting that the NA activity was insufficient to cleave all bound virus from the cells ( Figure 3A). All of the N2 reassortants, with the exception of Q136K, showed full restoration of HA titre following incubation at 37 °C ( Figure 3B).

Discussion
In this study we describe the detection of A(H1N1) pdm09 influenza virus isolates, and to a lesser extent A(H3N2) viruses, with amino acid substitutions at the Q136 NA residue that reduce zanamivir, peramivir and laninamivir susceptibility. Surveillance data show that the Q136K and Q136R substitutions occurred sporadically in A(H1N1)pdm09 cultured isolates, with periods such as 2011 where a relatively high detection rate was observed, compared with other years where they were absent. Importantly, in all cases, the residue substitution that was present in the isolate, was not detected in the virus from the clinical specimen demonstrating that the mutation was either arising, or being selected for, during MDCK cell culture passage. Because the 'gold-standard' for laboratory assessment of NAI susceptibility is the phenotypic NA inhibition assay, which requires a cell culture isolate for testing, there is concern that viruses such as these can be reported as being 'resistant' when in fact the virus that came from the patient was sensitive. In addition, the process of cell culture may also select against a resistant virus, meaning that a variant virus is not detected when it was present in the clinical specimen. Misdiagnosis can have an impact on the therapies being used in patient management and may unnecessarily result in therapy being stopped, modified or inappropriately continued. In addition to the Q136K/R variants described here, there are many other NA mutations that alter NAI susceptibility and also appear to be selected during MDCK cell culture [19,23]. Interestingly these seem to be increasingly reported for influenza B viruses [23][24][25].
Therefore sequence analysis of the influenza viruses in the original specimen remains important when laboratories detect mutations in cultured isolates. Although the Q136K and Q136R variants detected here were all cell culture derived variants, other studies have reported mutations at Q136 that were detected in viruses from clinical specimens. The Q136K mutation was detected in an A(H3N2) virus, together with an E119V NA mutation, in a patient who had previously undergone a bone marrow transplant, following a treatment course of both inhaled and intravenous zanamivir and oseltamivir [27]. Mutations at Q136 have also been detected in ferrets infected with influenza A(H5N1). The Q136L variant was detected in the nasal wash of a zanamivir treated ferret infected with an A(H5N1) virus [28], while a Q136H mutation was detected in an A(H5N1) virus from a ferret not being treated with an antiviral [29]. These reports demonstrate that viruses with these mutations have the potential to infect or replicate in vivo both in the presence or absence of zanamivir pressure.
The zanamivir concentration in sputum 12 hours postinhalation has been reported to be between 159 and 4,315 nM [30]. Therefore, while the correlation between the drug concentration in sputum and the drug concentration at the site of viral replication is not clear, it is anticipated that only the Q136K and Q136R mutations in the A(H1N1)pdm09 virus may potentially impact the clinical effectiveness of zanamivir. The Q136L mutation in both N1 and N2 NAs and Q136K in N2 NA caused mild increases in zanamivir and oseltamivir IC 50 , which are expected to be below the concentrations present at the sites of replication in treated individuals [30,31].
An evaluation of the ability of the Q136K and Q136R A(H1N1)pdm09 variants to replicate and transmit in animal models will provide useful insights into the potential risk that these viruses may pose to public health. Where possible these future studies would benefit from using Q136K and Q136R variants that were naturally occurring, rather than cell culture derived or generated by reverse genetics. One limitation of this study is that the N1 and N2 NAs with Q136 substitutions were assessed in 1:7 reassortant viruses generated by reverse genetics on a PR/8 backbone, therefore there is potential that the HA/NA balance between the variant NA molecules and the HA from PR/8 may be different from that seen in the 'natural' isolate. The in vitro assays showed that as a result of the Q136K mutation the N1 reassortant had a moderate loss in NA activity and thermostability. A reduction in NA enzyme activity and surface expression due to the Q136K NA mutation has been previously reported [32,33]. Pizzorno et al. [33] also demonstrated that an A(H1N1)pdm09 Q136K variant had compromised replication compared with a WT virus in mice, while in a ferret model the variant was able to transmit between contact ferrets, but at slower rate than for the WT virus. However, a study in guinea pigs found that the Q136K variant did not transmit between animals [32]. Taken together, these studies indicate that the replication and transmissibility of the Q136K variant in the A(H1N1)pdm09 virus appears to be compromised and therefore is unlikely to circulate through the human population. However, compensatory mutations in the NA or other genes that may occur in the A(H1N1)pdm09 virus in the future may buffer the compromising effect of the Q136K mutation and improve overall replication and transmissibility of the variant, in a manner similar to that seen for the H275Y mutation [7,34]. To date there has been no evaluation of the in vivo fitness of the Q136R A(H1N1) pdm09 variant.
In this study we have highlighted the challenges that cell culture derived mutations, such as Q136K and Q136R, can pose to the analysis and interpretation of viruses for NAI susceptibility. This further reaffirms the need to sequence viruses from the clinical specimens of any isolate that shows reduced susceptibility in a phenotypic NA inhibition assay to avoid misdiagnosis and any unnecessary change in patient management with respect to the use of antivirals. Our findings highlight the effect of mutations at the Q136 residue of N1 viruses on laninamivir, peramivir or zanamivir susceptibility, and therefore close monitoring of viruses for these mutations in patients being treated with these antivirals is important.