Application of a Biologically Contained Reporter System To Study Gain-of-Function H5N1 Influenza A Viruses with Pandemic Potential

Understanding how animal influenza viruses can adapt to spread in humans is critical to prepare for, and prevent, new pandemics. However, working safely with pathogens that have pandemic potential requires tight regulation and the use of high-level physical and biological risk mitigation strategies to stop accidental loss of containment. Here, we used a biological containment system for influenza viruses to study strains with pandemic potential. The system relies on deletion of the essential HA gene from the viral genome and its provision by a genetically modified cell line, to which virus propagation is therefore restricted. We show that this method permits safe handling of these pathogens, including gain-of-function variants, without the risk of generating fully infectious viruses. Furthermore, we demonstrate that this system can be used to assess virus sensitivity to both approved and experimental drugs, as well as the antigenic profile of viruses, important considerations for evaluating prepandemic vaccine and antiviral strategies.


S easonal influenza A viruses (IAV) can cause an acute respiratory illness in humans,
with symptoms including fever, dry cough, headache, muscle pain, and malaise. The disease is typically mild and self-resolving, but it can be more severe and lead to death, particularly in high-risk groups, such as young children, the elderly, and those with chronic underlying health conditions or an immunosuppressed state. Therefore, annual seasonal IAV epidemics remain a major threat to public health and are associated with significant morbidity and mortality worldwide, including an estimated 290,000 to 650,000 deaths globally (1). Moreover, IAVs found in nonhuman hosts (e.g., avian reservoirs) have the potential for sporadic zoonotic transmission to humans and can cause critical disease, such as severe pneumonia, acute respiratory distress syndrome, and significant fatalities, even in young, otherwise healthy adults (2). For example, avian H5N1 IAVs have caused 861 recorded human infections and 455 deaths since 2003, while avian H7N9 IAVs have caused Ͼ1,500 human infections since 2013 (2). The fear is that one of these novel avian IAV strains will eventually adapt to replicate and transmit efficiently in humans, thereby sparking a devastating new influenza pandemic similar in scale to one of the three that occurred in the 20th century: the 1918 H1N1 pandemic, with Ͼ40 million deaths; the 1957 H2N2 pandemic, with 1 to 3 million deaths; and the 1968 H3N2 pandemic, with ϳ1 million deaths (3)(4)(5)(6)(7).
For pandemic preparedness, it is important to understand the genetic determinants of IAVs that permit them to replicate, to be transmitted, and to cause disease in humans. Such work allows the identification of sequence markers that can be used in global surveillance and eradication efforts, and it supports testing of prepandemic vaccine and antiviral stockpiles to ensure their efficacy if novel IAV strains acquire human replication and transmissibility features (8,9). Key landmark studies in this area identified several amino acid substitutions in viral proteins, notably hemagglutinin (HA), that are critical for allowing mammalian transmissibility of avian IAVs, and identification was facilitated by the creation of fully infectious, replication-competent, and antigenically novel IAVs at the highest biosafety and biosecurity levels (10)(11)(12)(13). The creation of and work with such novel, transmissible IAVs has been controversial in the scientific community, leading to heated debates in the literature over the risk-benefit analysis of so-called gain-of-function experiments (14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32), research moratoria on such pathogens (33)(34)(35)(36), and discussion about the policy development of new biocontainment, biosafety, and biosecurity frameworks (37)(38)(39)(40)(41).
Biologically contained viruses, where an essential viral gene is deleted from a pathogen and is expressed in trans by a complementing cell line to allow propagation, represent a safe means to study fundamental in vitro aspects of pathogen biology and could constitute a risk mitigation strategy for some gain-of-function experiments. Biological containment methods have previously been applied to study highly pathogenic viruses such as Ebola virus and Omsk hemorrhagic fever virus safely at lower containment levels than their usual biosafety level 4 (BSL4) facility requirements (42,43), thereby opening up a range of valuable experimental activities due to the lower associated operating costs, increased equipment availability, and number of laboratories that can use them. Often, in these methods, the essential viral gene is replaced with a convenient reporter gene, such as one encoding a fluorescent protein or luciferase, in order to simplify subsequent studies such as high-throughput screening of large compound libraries with antiviral potential.
Recent efforts in the development of biologically contained IAVs for different applications have been reviewed extensively (44). Most previous studies have focused on using common laboratory IAV strains, such as PR8 or WSN, and studying concepts such as (bivalent) vaccine or vector design, drug resistance or screening, host factor dependencies, and RNA packaging. In this study, we sought to assess the applicability of a biologically contained IAV concept, with viral deletion of HA (45), to strains such as highly pathogenic H5N1 and subsequently derived gain-of-function mammaliantransmissible variants that would otherwise require BSL3 or higher biocontainment facilities. Our specific aim was to understand the feasibility of using these tools as a safe risk mitigation strategy to assess antiviral drug susceptibility and antigenicity of new variants without having to generate fully infectious, replication-competent IAVs.

RESULTS
Generation of biologically contained 2009 pandemic H1N1 and highly pathogenic H5N1 IAVs expressing Renilla luciferase. Previous studies have described the construction of a recombinant IAV (A/WSN/33 strain [referred to here as WSN]) in which the HA coding sequence is replaced with sequences for either green fluorescent protein (GFP) or Renilla luciferase, but 3= and 5= packaging regions of the HA segment are maintained (45,46). In MDCK cells engineered to stably express WSN HA, such reporter IAV-like particles can undergo multicycle replication (45) but are restricted to single-cycle infections in cells lacking expression of a complementing HA molecule (46). While WSN is a commonly used strain in many laboratories because of its high-growth properties and has been used to identify important aspects of IAV replication, we sought to expand use of this technology to more clinically relevant strains, such as 2009 pandemic H1N1 (pdmH1N1) and highly pathogenic H5N1 IAVs ( Fig. 1A (47), and verified their stable expression by Western blotting (Fig. 1C). Indirect immunofluorescence also revealed that each engineered MDCK cell line exhibited homogenous expression of the respective HA and that all HAs predominantly localized to the cell surface (Fig. 1D). Using these cell lines, we were able to successfully rescue and grow recombinant Neth/09-Renilla and Viet/04-Renilla IAV-like particles that incorporate a segment expressing Renilla luciferase and lack the coding and noncoding sequences for their own HAs but are otherwise authentic. Importantly, these Neth/09-Renilla and Viet/04-Renilla IAV-like particles, like the original WSN/33-Renilla IAV, are biologically contained, as they do not undergo multicycle replication to form plaques on MDCK cells but readily form plaques in MDCK cells stably expressing the cognate HA (Fig. 1E). In addition, passaging experiments confirmed that biologically contained IAV-like particles do not propagate in MDCK cells even to low levels (Fig. 1F), and functional experiments indicated that the IAV-like particles (at least for the WSN strain) can have levels of HA and NA similar to those of authentic virus particles ( Fig. 1G and H).
Biologically contained IAV reporters expressing Renilla luciferase allow livecell kinetic analysis of single-cycle and multicycle infections. Renilla luciferase was chosen as our reporter molecule because of two key advantages over other commonly used reporters, such as fluorophores and the small luciferase NanoLuc: (i) luciferases in general have higher dynamic ranges than fluorophores, such as GFP, allowing more optimal discrimination between small interventions (48), and (ii) a unique, stable, live-cell substrate system is available for Renilla luciferase (EnduRen; Promega), which means that cell lysis or supernatant collection is unnecessary, minimizing the liquid handling required with most other luminescent substrates and allowing simple continuous live-cell monitoring over long periods with multiple time points. To demonstrate this system, we infected MDCK or MDCK-HA cells with either WSN/33-Renilla, Neth/09-Renilla, or Viet/04-Renilla IAV-like particles at a range of multiplicities of infection (MOI) and added EnduRen substrate to the medium at 1 h postinfection. With no other manipulation, we were able to perform direct live-cell luminescence readouts every 3 h until the viral replication kinetics were complete ( Fig. 2A to F). Notably, in preliminary experiments we found that the EnduRen substrate is stable in tissue culture medium for at least 30 h and does not need to be replenished. Relatively high-MOI (1 and 0.1 PFU/cell) infections of MDCK or MDCK-HA cells with all of the reporter IAVs led to a rapid increase in luminescence emission, peaking at 6 to 9 h postinfection and giving an ϳ1,000-fold increase of activity over mock infections. No differences in these replication kinetics or magnitude were obvious between experiments in MDCK cells with and without complementing HA expression ( Fig. 2A to F). In contrast, relatively low-MOI (0.01 and 0.001 PFU/cell) infections showed peak luminescence levels only ϳ10to 100-fold above that of mock infection in MDCK cells lacking HA expression ( Fig. 2A, C, and E), suggesting only single-cycle infections in these cells, but led to luminescence levels ϳ1,000-fold higher than mock infection at later time points (12 to 24 h postinfection) in MDCK cells expressing HA, presumably due to multicycle replication in these cells (Fig. 2B, D, and F). These data indicate that the IAV-Renilla reporters described here are biologically contained and can be effectively used to study singlecycle and multicycle replication kinetics under live-cell conditions with a very high dynamic range.
Biologically contained IAV reporters exhibit antiviral drug sensitivities similar to those of authentic IAVs. To evaluate the feasibility of using biologically contained IAV reporters for inhibitor studies, we assayed their sensitivity to known, approved antiviral drugs such as oseltamivir, zanamivir, and baloxavir. Oseltamivir and zanamivir are approved neuraminidase (NA) inhibitors (4). Using standard in vitro NA inhibition assays, we determined that both oseltamivir and zanamivir potently inhibit the NA activity of WSN/33-Renilla IAV-like particles with a 50% inhibitory concentration (IC 50 ) similar to that required for inhibition of authentic WSN IAV particles ( Fig. 3A and B). The NA activity of Neth/09-Renilla IAV-like particles was also inhibited to the same extent as wild-type Neth/09 IAV particles by these two antivirals ( Fig. 3C and D). These data indicate that the biologically contained IAV reporters can be used to faithfully recapitulate inhibitor study data of authentic viruses, making them a useful surrogate when work with fully infectious IAVs has to be limited. We further profiled the antiviral activity of baloxavir, a potent inhibitor of the viral endonuclease (49), against all of our IAV-Renilla reporters. MDCK cells expressing each cognate HA were infected with the respective IAV-Renilla virus in the presence of various concentrations of baloxavir. Following addition of EnduRen to the medium at 1 h postinfection, together with baloxavir, direct live-cell luminescence readouts were taken at several time points postinfection, and IC 50 s were calculated from the nonlinear regression curve that was determined for each virus from the area under the curve (AUC) of the replication kinetics ( Fig. 3E to G). Baloxavir was able to potently interfere with the replication of all three Renilla reporter viruses (WSN/33, Neth/09, and Viet/04 based), with IC 50 s in the range of those published previously (50,51). Thus, biologically contained IAV reporters can also be used to rapidly assess antiviral efficacy without laborious and timeconsuming titrations of the wild-type viruses.
Generation of a biologically contained IAV reporter system to study mammaliantransmissible gain-of-function H5N1 variants. Following the discovery that highly pathogenic H5N1 IAVs can be experimentally adapted in their HA sequence to confer mammal-to-mammal transmissibility (10,11), there has been a detailed discussion in the field about the biosafety, biosecurity, and ethical implications of studying such gain-of-function viruses (14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32). To overcome some of these barriers and concerns, biological risk mitigation strategies are an additional layer of molecular safety to enhance physical biocontainment (BSL3ϩ) infrastructure and protocols. For example, genetic engineering of species-specific microRNA target sites into viral genomes can limit replication exclusively to nonhuman experimental models (52). As proof of principle for another biological risk mitigation strategy, we applied the biologically contained IAV reporter system described here to this problem. Although such a system limits experiments to in vitro cell culture work, we reasoned that a major advantage would be the sensitive reporter read-out at a lower physical biocontainment level (e.g., BSL2). Importantly, in such a system, there is no mechanism for the virus to genetically acquire the variant HA conferring transmissibility, as the HA cDNA provided in trans lacks essential 3= and 5= noncoding packaging sequences. We also used a human codon-optimized HA sequence to rule out the very rare chance of recombination. As before, and primarily for added biosafety in these proof-of-principle experiments, the construct was based on the modified H5 HA protein lacking the polybasic cleavage site (HALo), although it should be possible to use an HA containing the polybasic cleavage site. We thus generated an MDCK cell line stably expressing a variant Viet/04 H5 HA (HALo) protein with 4 amino acid substitutions described to confer airborne transmissibility (termed Viet/04-T-HA): N158D, N224K, Q226L, and T318I (H3 numbering) (Fig. 4A) (11,53). The two substitutions in the HA globular head (N224K and Q226L), together with loss of the glycosylation site at position 158 (N158D), are responsible for increasing the affinity of HA for human-type receptors (11,(53)(54)(55)(56). In addition, the T318I substitution increases HA thermostability and lowers the pH required for fusion (11,53,57). Stable homogenous expression and correct localization of the Viet/04-T-HA protein in MDCK cells was confirmed by Western blotting and indirect immunofluorescence ( Fig. 4B and C). We grew a stock of Viet/04-Renilla IAV on these cells to produce virus-like particles incorporating the Viet/04-T-HA protein and confirmed that this new Viet/04-T-Renilla IAV is biologically contained: it forms plaques and undergoes multicycle replication, as determined by luciferase activity, only on MDCK cells that stably express HA (Fig. 4D to F). These data indicate that a biologically contained H5N1 reporter virus with gain-of-function mutations that confer mammalian transmissibility can be produced and used for facile assays.
The biologically contained mammalian-transmissible H5N1 reporter IAV can be used to assay antiviral-drug sensitivity and antigenicity. As a proof of concept for studying mammalian-transmissible gain-of-function IAV-like particles under biologically contained conditions, we assayed how the substitutions in Viet/04-T-HA impacted the effect of antivirals and antibodies. Standard in vitro NA inhibition assays revealed that both oseltamivir and zanamivir inhibit the NA activity of Viet/04-T-Renilla IAV-like particles with IC 50 s similar to those required for inhibition of Viet/04-Renilla IAV-like particles, demonstrating similar incorporation of NA into both viruses ( Fig. 5A and B). We also found that, despite the reported impact of the T318I substitution on HA thermostability and fusion potential (11,53,57), the potent experimental HA-targeting fusion inhibitor S20 (58) is as active against Viet/04-T-Renilla IAV as it is against Viet/04-Renilla IAV during infection (Fig. 5C). In addition, we tested the neutralization capabilities of a monoclonal anti-H5 HA antibody and a polyclonal anti-H5 HA serum, with the aim of assessing whether the 4 gain-of-function amino acid substitutions impacted the antigenic profile of HA, an important consideration for development of prepandemic H5 vaccines. Viet/04-T-Renilla IAV-like and Viet/04-Renilla IAV-like particles were preincubated for 1 h with each antibody or serum at various concentrations prior to single-cycle infection of MDCK cells. Following addition of EnduRen, luciferase activity was measured at several times postinfection, and IC 50 s, or dilution factors with 50% inhibition, were calculated using the nonlinear regression curve for each infection ( Fig. 5D and E). Importantly, Viet/04-T-Renilla IAV was sensitive to neutralization by both the anti-H5 monoclonal and polyclonal preparations and was, if anything, more sensitive to neutralization than Viet/04-Renilla IAV. These data show that the biologically contained IAV reporter system described here can be applied to safely study gain-offunction mutations, and assays can be readily established to understand antiviral susceptibility and potential antigenic changes.

DISCUSSION
Herein, we describe the application of a biologically contained IAV reporter system to clinically relevant strains, including 2009 pdmH1N1 and highly pathogenic H5N1. We show that by incorporating Renilla luciferase as the reporter of choice, and by using a stable, cell-permeable substrate (EnduRen), the high dynamic range of luciferase can be combined with real-time live-cell kinetic measurements to track virus replication in a single sample without excessive liquid-handling steps. The biologically contained system relies on the in trans expression of HA in a complementing cell line to achieve multicycle replication; thus, the use of appropriate cell lines allows ready comparisons between single-cycle and multicycle replication. Furthermore, we demonstrate that these systems are fully amenable to assessment of antiviral drug sensitivity, with inhibitory effects aligning well with those of wild-type nonreporter viruses. Given that these safe, clinically relevant reporter viruses are already optimized for BSL2 use in multiwell formats, such as 96-well plates, they could be useful for conducting largescale high-throughput compound screens to identify new antivirals. The biological containment system used here permits the cell culture study of otherwise BSL3 or BSL3ϩ IAV strains at BSL2 without the risk of reassortment or recombination. In this regard, expression of HA in trans, with no possibility of its being genetically incorporated into a fully infectious virus, is the essential feature, as IAVs with novel HAs to which humans do not yet have any humoral immunity are the most threatening with regard to pandemic emergence (4). Such a strict genetic safety feature allowed us to create reporter H5N1 IAV variants utilizing gain-of-function mammalian- independent experiments were plotted and used to calculate IC 50 s (D) or the dilution factor for 50% inhibition (E) with GraphPad Prism 7.05. The data line represents the best-fit curve. For all panels, statistical significance was determined for each concentration by an unpaired two-tailed t test and marked where thresholds were reached (*, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001).
transmissible HA proteins that are replication competent only in cells stably expressing HA. So-called gain-of-function experiments are controversial for pathogens with pandemic potential (14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30)(31)(32), and it is important to implement risk mitigation strategies as much as possible wherever the scientific question under study allows. Here, we did not want to address the mammalian transmissibility or virulence of gain-of-function H5N1 IAVs (which would require fully replication competent viruses), but instead we sought to use the biologically contained reporter system as a proof of principle for developing a feasible framework to assess the antiviral susceptibility and antigenicity of such viruses. Given our success in establishing and implementing such a system, as well as our results indicating that amino acid substitutions related to mammalian transmissibility do not reduce H5N1 antiviral sensitivity or antigenicity, we believe that the biologically contained reporters used here represent a safe and useful way to characterize IAVs with pandemic potential.

MATERIALS AND METHODS
Cells, lentiviruses, and generation of cell lines. Madin-Darby canine kidney (MDCK) and human embryonic kidney (293T) cells were grown in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (P/S) at 37°C with 5% CO 2 . To generate cell lines stably expressing HA, the respective HA cDNAs were first cloned into pLVX-IRES-Puro (Clontech Laboratories) using XhoI and BamHI (Viet/04) or XhoI and NotI (Neth/09 and WSN/33) restriction sites. HA cDNAs were obtained either from existing vectors (WSN/33, Neth/09, and Viet/04) or by commercial gene synthesis (Viet/04-T; GeneArt; Thermo Fisher Scientific). Both Viet/04-derived HA constructs lacked the sequence encoding the polybasic cleavage site (and were therefore termed HALo), and all constructs were validated by sequencing prior to use. To generate lentiviral particles, 293T cells were cotransfected with pCMVdR8.91, pMD2.G, and the respective pLVX-HA-IRES-Puro construct for 48 h before harvesting of supernatants, filtration through a 0.45-m filter, and transduction of MDCK cells in the presence of Polybrene (final concentration of 8 g/ml; Sigma-Aldrich). Two days later, transduced cells were selected with puromycin (3.5 g/ml; Thermo Fisher Scientific) and subcloned by limiting dilution. Clonal cell colonies were expanded, and HA expression was assayed by Western blotting and immunofluorescence.
Biosafety. Procedures to generate and work with biologically contained H5N1 IAVs were thoroughly risk assessed prior to the start of the project and were approved by the Swiss Federal Office of Public Health (Ecogen number A161912). In brief, special considerations included the temporal and physical separation of experiments involving H5N1 IAV-like particles from experiments involving other IAVs in order to minimize the risk of accidental reassortment. With specific regard to the H5N1 IAV HA and its mammalian-transmissible variant, we did not at any point generate or propagate pDZ-or pPolI-based plasmids containing the respective full HA segment so as to prevent accidental rescue of replicationcompetent viruses containing the H5 HA segment: we used only plasmids carrying the HA ORF which lacked the 3= and 5= noncoding sequences essential for packaging. In addition, for the mammaliantransmissible H5 HA variant, we generated a human codon-optimized HA ORF lacking the 3= and 5= noncoding sequences. The rationale for this strategy was to destroy any potential cryptic packaging elements in the coding region, thereby eliminating the very rare chance of recombination of this sequence into a fully replication-competent virus.
Assessing virus replication by live-cell Renilla luciferase assays. Cells in 96-well plates were washed once with PBS prior to infections with the different viruses at the appropriate MOI in PBS supplemented with 1 mM Ca 2ϩ and Mg 2ϩ , 0.3% bovine serum albumin (BSA), and 1% P/S. After 1 h at 37°C, the inoculum was removed, cells were washed once with PBS, and postinfection medium was added: DMEM supplemented with 0.1% FBS, 0.3% BSA, 20 nM HEPES, 1% P/S, and 1 g/ml TPCK-treated trypsin, together with 6 M Renilla luciferase substrate (EnduRen live-cell substrate; Promega). Where noted, the virus inoculum was preincubated with the indicated concentrations of monoclonal anti-H5 antibody (no. 200-301-976; Rockland) or polyclonal anti-H5 serum (no. 2705; BEI Resources) at 4°C for 1 h before infection of cells. Also where noted, the indicated concentrations of baloxavir acid (no. HY-109025A; MedChemExpress) or S20 fusion inhibitor (58) (ChemBridge) were added to the cells following infection. Real-time luminescence measurements were taken at various time points using an EnVision multilabel reader (Perkin Elmer) or a Dynex MLX luminometer (Dynex Technologies). All experiments included infection of multiple wells for technical replicates, and each experiment was independently performed at least 3 times. Mock-infected wells, otherwise treated identically, acted as negative controls for background luminescence from the reagents.
In vitro NA activity assay. To measure NA activity of virus preparations, the NA-Star influenza virus neuraminidase inhibitor resistance detection kit (Applied Biosystems) was used according to the manufacturer's protocol. As necessary, virus preparations were pretreated for 1 h with various concentrations of oseltamivir carboxylate (no. sc-212484; Santa Cruz Biotechnology) or zanamivir (no. SML0492; Sigma-Aldrich), and chemiluminescence was measured with the EnVision multilabel reader (Perkin Elmer).
Data analysis. For Renilla luciferase assays, raw luminescence values were averaged from technical replicates within the same experiment, normalized to mock (background) values, and made relative to the initial value of the MOI ϭ 1 condition. Each independent replicate was processed separately. Where necessary, the area under the curve (AUC) of the replication kinetics was determined and used to calculate half-maximal inhibitory concentrations (IC 50 ) from the resulting nonlinear regression curve. For analysis of the NA activity assay, raw luminescence values were normalized to background, averaged, and made relative to the untreated control. The IC 50 was calculated from the resulting nonlinear regression curve. These analyses, and all statistics, were performed using GraphPad Prism version 7.05.