Influenza A virus infections impose a recurrent and global disease burden. Current antivirals against influenza are not always effective. We assessed the protective potential of monovalent and bivalent Nanobodies (Ablynx) against challenge with this virus. These Nanobodies were derived from llamas and target H5N1 hemagglutinin. Intranasal administration of Nanobodies effectively controlled homologous influenza A virus replication. Administration of Nanobodies before challenge strongly reduced H5N1 virus replication in the lungs and protected mice from morbidity and mortality after a lethal challenge with H5N1 virus. The bivalent Nanobody was at least 60-fold more effective than the monovalent Nanobody in controlling virus replication. In addition, Nanobody therapy after challenge strongly reduced viral replication and significantly delayed time to death. Epitope mapping revealed that the VHH Nanobody binds to antigenic site B in H5 hemagglutinin. Because Nanobodies are small, stable, and simple to produce, they are a promising, novel therapeutic agent against influenza.
Highly pathogenic avian influenza (HPAI) H5N1 viruses pose a pandemic threat. Transmission of these viruses from birds to humans occurs rarely, but when it does, it often results in severe pulmonary and systemic disease with a mortality rate of 60%. Despite the availability of medicines designed to interfere with influenza viral functions , clinicians have very limited evidence-based intervention options to successfully treat victims of these zoonotic infections with HPAI H5N1. Currently licensed anti-influenza drugs are small molecules that target the ion channel activity of matrix protein 2 (amantadine and rimantadine) or the viral neuraminidase (oseltamivir and zanamivir). These drugs are beneficial in cases of uncomplicated influenza caused by susceptible influenza virus strains . However, infection with HPAI H5N1 virus is often characterized by a more serious clinical outcome than uncomplicated influenza. This difference in severity can be attributed to the higher replication rate of HPAI H5N1 viruses, their broader cellular tropism and, sometimes, systemic spread [3–5]. Effective treatment and control of rapidly replicating HPAI H5N1 viruses (eg, with oseltamivir) may therefore require higher doses to obtain and maintain sufficient systemic drug concentrations . Inadequate dosing, combined with prolonged virus replication, increases the odds of developing resistant virus strains [7, 8].
Severe infection with HPAI H5N1 is associated with an aberrant host immune response, producing high levels of circulating proinflammatory cytokines and chemokines . Such potentially life-threatening cytokine deregulation is presumably caused by the unusual cellular and tissue tropism of HPAI H5N1 viruses, their rapid replication and the intrinsically high cytopathic outcome . A combination of antiviral drugs and immunomodulators could be used to manage infection with HPAI H5N1, but such an intervention is not without risks [11–13]. Therefore, there is a need for novel anti-influenza drugs to treat patients with severe disease resulting from infection with highly pathogenic influenza viruses, or severe influenza in general.
Human or animal-derived immune serum with high titers of immunoglobulins against a particular virus has been used successfully to prevent and treat viral infections. For example, during the Spanish flu outbreak, passive immunotherapy with human convalescent blood products was used to treat patients with influenza pneumonia, with some success [14, 15]. A report from 2007 describes the transfer of convalescent plasma from a survivor of H5N1 influenza to treat a hospitalized 31-year-old patient with the same infection who had not responded to oseltamivir treatment. The patient recovered completely . Compared with human immunoglobulin products, recombinant monoclonal antibodies or antibody-derived fragments are safer, better characterized, and more reliable therapeutics [17, 18]. The antigen-binding site of conventional antibodies resides in the paired variable domains of the heavy and light immunoglobulin chains. Antigen specificity is usually retained in recombinant single-chain variable fragment molecules, which typically comprise the variable heavy and variable light chain fragments from natural antibodies, linked by a flexible peptide , but antibodies with still smaller antigen-binding domains also exist naturally. Camelids and sharks produce immunoglobulins composed only of heavy chains, and the antigen-binding site of these antibodies resides in a single protein domain, designated VHH (or Nanobody [Ablynx]) in camelids [20, 21]. VHHs can be selected by phage or other display technologies, they are easily produced in robust recombinant expression systems, and they have high solubility and thermal stability. In addition, they can form long, fingerlike loops that can penetrate into the cavities of immunogens, and their small size permits good tissue penetration in vivo . These characteristics have spurred the development of VHH-based proteins as novel therapeutic and diagnostic tools directed against multiple targets, including cytokines, enzymes, bacterial toxins, tumor antigens, microbial antigens, and parasites [23–28].
Here, we describe a novel treatment strategy for H5N1 influenza virus infection, based on recombinant neutralizing Nanobodies directed against hemagglutinin (HA). We show that intranasal administration of a Nanobody specific for H5N1-HA potently suppresses replication of a recombinant H5N1 virus in vivo. We also demonstrate that a bivalent H5N1-HA–specific Nanobody is at least 60-fold more effective at suppressing virus replication than its monovalent counterpart, as deduced from an in vivo dose-response comparison, and that it protects against a lethal H5N1 virus challenge. Analysis of selected escape mutant viruses allowed us to map the epitope of the protective Nanobody at the site corresponding to the antigenic site B in the HA molecule.
MATERIALS AND METHODS
Selection, Expression and Purification of H5 HA–Specific VHH Nanobodies
The recombinant H5N1-HA–specific VHHs (H5-VHH) Nanobodies used in this study have been described elsewhere (Anna Hultberg, Nigel J. Temperton, Valérie Rosseels, Mireille Koenders, Maria Gonzalez-Pajuelo, Bert Schepens, Lorena Itatí Ibañez, Peter Vanlandschoot, Joris Schillemans, Michael Saunders, Robin A. Weiss, Xavier Saelens, José A. Melero, C. Theo Verrips, Steven Van Gucht, and Hans J. De Haard, unpublished report). In brief, 2 llamas were immunized with recombinant clade 1 A/Vietnam/1203/2004 H5N1-HA (Protein Sciences) in Specol adjuvant. After 6 immunizations, total RNA was extracted from peripheral blood lymphocytes, converted into complementary DNA (cDNA), and cloned in a phagemid vector to generate a VHH phage display library. Phages carrying H5N1-HA–specific Nanobodies were selected by panning on immobilized H5N1-HA. The VHH-coding sequences from enriched phages were cloned into an Escherichia coli expression vector and purified from the bacterial periplasmic fraction by immobilized metal-affinity chromatography. Recombinant H5-VHHm Nanobody was used in this study. H5-VHHm Nanobody was also engineered into bivalent H5-VHHb, by linking 2 H5-VHHm domains with a glycine-serine linker. For a negative control VHH, we used recombinant monovalent respiratory syncytial virus (RSV) RSV-VHHm or bivalent RSV-VHHb (Hultberg et al, unpublished report). The endotoxin content of the purified, recombinant VHH antibodies was <1 EU/μg of protein, as determined with the limulus amebocyte lysate test.
H5N1 Influenza Virus
H5N1 influenza A virus NIBRG-14 was obtained from the UK National Institute for Biological Standards and Control, a center of the Health Protection Agency. NIBRG-14 is a 2:6 reverse genetics–derived reassortant of A/Vietnam/1194/2004 (H5N1) and A/PR/8/34 (H1N1) viruses. The HA- and neuraminidase-coding segments of NIBRG-14 are from A/Vietnam/1194/2004, but the HA segment lacks the polybasic cleavage site. NIBRG-14 virus was passaged 7 times in specific-pathogen–free BALB/c mice to obtain a mouse-adapted derivative (NIBRG-14ma) and propagated in Madin-Darby canine kidney (MDCK) cells. The HA-coding region was sequenced and found to be identical in NIBRG-14 and NIBRG-14ma viruses. The median tissue culture infectious dose (TCID50) and median lethal dose (LD50) of NIBRG-14ma virus were determined using the method of Reed and Muench ; 1 LD50 of mouse-adapted NIBRG-14 corresponds to ∼50 TCID50. All experiments with NIBRG-14 and NIBRG-14ma viruses were performed in biosafety level 2 rooms with negative pressure relative to adjoining rooms.
Binding of Nanobodies to H5 HA
The affinity and avidity of myc-tagged H5-VHHm and VHHb were compared by performing an enzyme-linked immunosorbent assay on NIBRG-14ma–infected MDCK cells and on immobilized virions. MDCK cells were infected with 100 TCID50 of NIBRG-14ma, and 8 h later cells were fixed with 4% paraformaldehyde. Binding of Nanobodies was revealed with an anti-myc monoclonal (Invitrogen) followed by horseradish peroxidase–labeled anti-mouse antibody (GE Healthcare) and horseradish peroxidase substrate (BD optEIA).
Prophylactic and Therapeutic Efficacy Studies in Mice
All animal procedures were approved by the Institutional Ethics Committee on Experimental Animals. Specific-pathogen–free female BALB/c mice, 7–9 weeks old, were purchased from Charles River (Germany) and used for all experiments. Mice were housed in cages individually ventilated with high-efficiency particulate air filters in temperature-controlled, air-conditioned facilities with food and water ad libitum. Mice were anesthetized by intraperitoneal injection of xylazine (10 μg/g) and ketamine (100 μg/g) before intranasal administration of Nanobodies or challenge virus (50 μL, divided equally between the nostrils). Nanobodies were diluted in endotoxin-free phosphate-buffered saline (PBS) with 1% (wt/vol) bovine serum albumin and administered as a single dose, ranging from 100 to 0.5 μg per mouse (5–0.25 mg/kg). Depending on the experiment, Nanobodies were administered up to 72 h before or after challenge, as specified in the figure legends.
To determine the effect of intranasal Nanobody delivery on lung virus titer production, mice were challenged with 1 LD50 of NIBRG-14ma virus and killed 4 days after challenge. Lung homogenates were prepared in PBS, cleared by centrifugation at 4°C, and used for virus titration. Monolayers of MDCK cells were infected with 50 μL of serial 1:10 dilutions of the lung homogenates, in a 96-well plate in serum-free Dulbecco's modified Eagle medium (Invitrogen) supplemented with penicillin and streptomycin. After 1 h, the inoculum was replaced by medium containing 2 μg/ml of L-(tosylamido-2-phenyl) ethyl chloromethyl ketone–treated trypsin (Sigma). End-point virus titers were determined by hemagglutination of chicken red blood cells and expressed as TCID50 per milliliter. Influenza RNA levels were determined with quantitative polymerase chain reaction (PCR). RNA was isolated from 150 μL of cleared lung homogenate using the Nucleospin RNA virus kit (Machery-Nagel). The relative amount of NIBRG-14ma genomic RNA was determined by preparing viral cDNA and performing quantitative PCR with M-genomic segment primers 5'tcgaaaggaacagcagagtg3' and 5'ccagctctatgctgacaaaatg3' and probe 5'ggatgctg3' (probe no. 89; Universal ProbeLibrary, Roche) and the LightCycler 480 Real-Time PCR System (Roche).
To determine the degree of protection against mortality, mice were challenged with 4 LD50 of NIBRG-14ma virus and subsequently monitored for 14 days. A 25% loss in body weight was the end point at which moribund mice were euthanized.
Selection and Characterization of H5N1-HA–Specific VHH Escape Mutant Viruses
In vitro NIBRG-14ma escape viruses were isolated independently by selection with H5-VHHm and H5-VHHb Nanobodies. Virus was serially diluted and used to infect MDCK cells in the presence of H5-VHHm (5 μmol/L, corresponding to 500-fold the median inhibitory concentration in microneutralization) or H5-VHHb (0.05 μmol/L, corresponding to 5000-fold the median inhibitory concentration in microneutralization). Escape viruses were selected using standard techniques. After 10 passages in the presence of Nanobodies, escape viruses were plaque purified by growth on MDCK cells overlaid with 0.6% low-melting agarose in the presence of Nanobodies. Six escape virus isolates obtained by selection with H5-VHHm and 6 by selection with H5-VHHb were amplified on MDCK cells, still in the presence of the respective Nanobodies. Total RNA was extracted from the supernatant and used to clone the cDNA corresponding to the HA and M viral RNA (vRNA) segments, as described elsewhere . In parallel, HA- and M-vRNA segments from the parental NIBRG-14ma were also cloned. Deduced amino acid substitutions in the HA sequence of escape viruses were modeled with PyMOL Molecular Graphics System (W. L. DeLano, 2002; DeLano Scientific) using the H5 HA structure with PDB code 2IBX . Hemagglutination inhibition, receptor-binding experiments (using chicken and human erythrocytes, respectively), and microneutralization were performed as described elsewhere .
SigmaPlot software (version 11) was used for statistical analysis. Differences between groups were tested using the Kruskal-Wallis 1-way analysis of variance on ranks. When this test demonstrated a significant difference between groups (P < .05), 2 correcting methods for multiple comparisons were used (Dunn's and Tukey's tests); t tests were used to compare 2 groups. Kaplan-Meier survival curves were plotted and evaluated.
Prophylactic Therapy with H5N1-HA–Specific Nanobodies
Nanobodies with clade 1 and clade 2 H5N1 virus neutralization activity have been described elsewhere (Hultberg et al, unpublished report). Bivalent H5-VHHb had a higher affinity and avidity for viral HA than its monovalent counterpart H5-VHHm, in particular when binding to immobilized virions (Figure 1). We extended these finding using a mouse model for H5N1 challenge. Mice were administered PBS or 100 μg (5 mg/kg) of H5-VHHb or negative control RSV-VHHb intranasally at 4, 24 or 48 h before infection with 1 LD50 of NIBRG-14ma virus. Weight loss was monitored daily, and on day 4 mice were killed to determine virus replication in the lungs. All mice that received H5-VHHb retained their original body weight, whereas those receiving PBS or RSV-VHHb gradually lost weight (Figure 2A). Importantly, intranasal administration of H5-VHHb at 4 or 24 h before challenge did not result in detectable lung virus titers. When animals were treated with H5-VHHb 48 h before challenge, virus titers were 50-fold lower than in PBS and RSV-VHHb treated mice, and 3 of 7 animals had no detectable virus titers (Figure 2B).
In vitro, H5-VHHb had a 1000-fold higher virus neutralizing activity than H5-VHHm. We therefore compared the protective efficacy of monovalent H5-VHHm and bivalent H5-VHHb by administering different doses of the bivalent Nanobody, ranging from 60 to 0.5 μg per mouse (3 to 0.025 mg/kg) and equimolar amounts of the monovalent Nanobody (30 to 0.25 μg per mouse, corresponding to 1.5 to 0.012 mg/kg), 4 h before challenge. For up to 4 days after challenge, none of the mice treated with mono- or bivalent H5-specific VHH Nanobodies lost weight, but control VHH-treated mice lost up to 15% of their initial body weight (Figure 3A). Mice treated with PBS or with control VHH had lung virus titers ranging from 5 × 107 to 1 × 109 TCID50/mL, whereas the virus titers in lung extracts from animals that received the highest dose of mono- or bivalent H5-VHH were below the detection limit (Figure 3B). Remarkably, prophylactic treatment with as little as 0.5 μg of H5-VHHb resulted in lung virus titers that were below the detection limit. It should be noted that we detected a low titer of virus in 1 of 4 mice treated with 12 μg of H5-VHHb Nanobody. Although the monovalent H5-VHHm was a less potent inhibitor of virus replication, treatment of mice with 6 or 1.2 μg of this molecule still resulted in significantly lower virus titers than in controls (Figure 3B).
Finally, we determined the protective efficacy of H5-VHHb over a longer period and with a higher challenge dose. As shown in Figure 4, mice that received 60 μg of H5-VHHb Nanobody 24 h before challenge with 4 LD50 of NIBRG-14ma were completely protected and displayed no weight loss, whereas all control mice had died or were moribund and had to be euthanized by day 8 after challenge.
Therapeutic Efficacy of H5N1-HA–Specific Nanobodies
We next determined whether H5-VHHb Nanobody could be used therapeutically. We administered 60 μg of this Nanobody (3 mg/kg) intranasally to mice up to 72 h after challenge with 1 LD50 of NIBRG-14ma virus. At 4 days after challenge, animals that had received H5-VHHb 4, 24 or 48 h after challenge had significantly higher body weights and lower lung virus loads than control mice (Figure 5). Although mice treated with H5-VHHb Nanobody 72 h after challenge were not clinically protected compared with control mice, they had significantly lower lung virus titers (Figure 5B). To exclude possible interference of residual neutralizing Nanobodies in the TCID50 assay that was used to determine the viral loads in lung virus extracts, the amount of viral RNA in these extracts was also determined using vRNA-specific quantitative reverse-transcription PCR. The results of this assay were completely in line with the determined infectious virus load (Figure 5C). Because H5-VHHb Nanobody showed the strongest prophylactic activity, its therapeutic efficacy against a more lethal virus dose was determined. When 60 μg (3 mg/kg) of this VHH was given 48 or 72 h after infection with 4 LD50 of NIBRG-14ma, weight loss was lower than in controls. In addition, treatment with nanobodies 48 h after infection significantly delayed mortality compared with controls (Figure 6).
Generation of H5-VHH Escape Variant H5N1 Virus
To identify the HA amino acid residues involved in H5-VHH binding, we selected escape viruses by growth and plaque purification of NIBRG-14ma virus in the presence of H5-VHHm or H5-VHHb Nanobodies. The HA sequences of 6 independently isolated H5-VHHm escape viruses revealed substitution of a lysine by a glutamic acid residue at position 189 in HA1 (H5 numbering). In addition, 2 H5-VHHm escape mutants carried an N154D substitution, and 4 carried an N154S substitution. The 3-dimensional structure of NIBRG-14 HA shows that N154D/S and K189E are close to each other as part of the corresponding antigenic site B in H3 HA (Figure 7A) [31, 33]. Interestingly, the N154D/S mutations remove an N-glycosylation site, which presumably has evolved in H5N1 HA as a strategy to mask an antigenic site . Escape viruses selected in the presence of H5-VHHb carried only substitution K189N (n = 4) or K189E (n = 2). Notably, 4 of the H5-VHHb escape viruses had an additional D145N mutation located in the stalk of HA2, 40 residues upstream of the membrane anchor (Figure 7B). Based on hemagglutination inhibition and microneutralization assays, the K189N/E substitution appears necessary and sufficient to abolish binding to H5-VHHm or VHHb (Figure 7C and D). These results indicate that residues in antigenic site B, at the top of HA and very close to the receptor binding domain, are essential for neutralization by H5-VHHm/b.
Despite decades of research on antivirals to treat influenza, there are still very few virus-directed options for the treatment of patients with severe complications of influenza. Zoonotic infections with H5N1 influenza viruses are rare, but the case-fatality rate is ∼60% even when advanced supportive therapy is applied . We demonstrated that intranasal administration of llama-derived immunoglobulin single-chain variable domain fragments with in vitro neutralizing activity against H5N1 viruses can control virus replication and reduce morbidity and mortality in a mouse model for H5N1 influenza. We also showed that the in vivo neutralizing capacity of bivalent H5 HA–specific Nanobodies is superior to that of the monovalent construct. Indeed, when mice were treated with as little as 25 μg/kg of H5-VHHb 24 h before challenge, virus titers in the lungs were undetectable. Compared with treatment with an irrelevant Nanobody, significant reductions in lung virus loads were obtained by prophylactic administration of H5-VHHb up to 2 days before challenge or by therapeutic treatment for up to 3 days after challenge.
Different groups of investigators have reported the isolation and characterization of murine or human monoclonal antibodies and single-chain variable fragment antibody fragments that are reactive against H5 HA with specificity for a particular clade of H5 HA , cross-clade reactivity [37–40], and even reactive against multiple HA subtypes [41–44]. How do our findings compare with these and findings of other studies that have explored passive immunoprophylaxis as a promising therapeutic against influenza? First, Nanobodies can be easily produced in and purified from a prokaryotic or yeast expression system. Second, Nanobodies can be engineered into multiple configurations to produce, for example, bispecific, bivalent, or pentavalent constructs . Third, we chose the intranasal or intrapulmonary route of administration because this allows deposition and deep penetration of Nanobodies in the respiratory tract. Fourth, H5-VHHb at 25 μg/kg was able to control lung virus replication when administered 24 h before challenge. In most other studies, passive immunotherapy was applied at doses between 3 and 25 mg/kg. However, the comparison of doses has to take into account the 10-fold lower molecular mass of VHH molecules compared with conventional antibodies. Monovalent Nanobodies are small (∼15 kDa) and as a result are removed rapidly from the blood through the kidneys. Their serum half-life is 30–60 min . Half-life extension methods used for other small antibody fragments, such as PEGylation or fusion to serum albumin, could be used to tailor the half-life of Nanobodies.
Conventional immunoglobulins are normally multivalent, meaning that they have multiple paratopes for antigen binding. Nevertheless, such immunoglobulins, including neutralizing antibodies, mostly bind to a single virion with only one paratope . The distribution of the viral spikes and their accessibility for antibody binding usually does not permit bivalent binding of a conventional antibody to a single virion. In addition, the flexibility of the 2 epitope-binding arms of an immunoglobulin G molecule is limited, and therefore binding to a second epitope on the surface of a single virion is possible only when that epitope is very near. This paradigm of typically monovalent interaction between a neutralizing monoclonal immunoglobulin G and a viral spike protein has been elegantly addressed by Wang and Yang . Neutralization of human immunodeficiency virus type 1 (HIV-1) by monoclonal antibody 2F5 was 100-fold enhanced when a second epitope for 2F5 was artificially introduced at another position in the HIV-1 glycoprotein spike, allowing bivalent interaction with increased binding avidity between the antibody and the virion. It was also demonstrated that such a bivalent binding occurs in trans, that is, by bridging the natural and artificially introduced epitopes in 2 subunits within a single HIV-1 trimer . A similar mechanism may explain the superior neutralization and protective efficacy of bivalent H5-VHHb compared with H5-VHHm. The flexible linker between the 2 paratope-containing domains in H5-VHHb and its small size are compatible with a high-affinity bivalent interaction with 2 epitopes within a single HA trimer.
A K189E substitution in HA1 was found to abolish the neutralizing effect of H5-VHH. A Lys or Arg residue at this position is conserved in all human H5N1 virus isolates. Of note, all selected escape mutants contained a glutamic acid or serine residue at position 189, which indicates that the conserved positively charged amino acid is essential for H5-VHH binding. Interestingly, escape mutants selected with H5-VHHm also carried an N154D/S commutation that removes an N-glycosylation site in this antigenic site of HA. The predicted N-glycosylation site at N154 in A/Hong Kong/156/97 HA was shown to be glycosylated . This N-glycosylation in H5 HA might have evolved to mask an antigenic site near the receptor binding domain . It may facilitate H5-VHHm binding by increasing the thermodynamic stability of the epitope, contributing to a productive interaction with monovalent but not bivalent H5-VHH. Increased receptor affinity of HA  in selected H5-VHH escape viruses in vitro is less likely, because their susceptibility to receptor-destroying enzyme treatment is equal to that of the parental virus (results not shown).
We conclude that neutralizing Nanobodies have considerable potential for the treatment of H5N1 virus infections. Although we focused on VHHs that recognize an epitope near the receptor binding domain, it is possible to select VHH molecules that bind to other epitopes in HA, including more conserved domains.
Ablynx NV, Institute for the Promotion of Innovation by Science and Technology Flanders (grant IWT 70050 from the Flemish government), Belgian Federal Sciences Administration (Federale Wetenschapsbeleid, BELSPO) (grant to L.I.I.), Ghent University (IOF [Industrieel Onderzoeksfonds] Stepstone grant IOF08/STEP/001 to L.I.I.). B.S. is a postdoctoral fellow at FWO (Fonds voor Wetenschappelijk Onderzoek)-Vlaanderen.
We thank Dr Amin Bredan for editing the manuscript and Anouk Smet, Frederik Vervalle, and Tine Ysenbaert for excellent technical assistance. We are grateful to Dr John Wood from the UK National Institute for Biological Standards and Control for providing NIBRG-14 virus and to Dr Juerg Stech and Dr Olga Stech from FLI (Greifswald Insel-Riems) for providing a rapid influenza virus gene cloning system.