Properties and Therapeutic Efficacy of Broadly Reactive Chimeric and Humanized H5-Specific Monoclonal Antibodies against H5N1 Influenza Viruses

Abstract

Highly pathogenic H5N1 virus infection causes severe disease and a high rate of fatality in humans. Development of humanized monoclonal antibodies may provide an efficient therapeutic regime for H5N1 virus infection. In the present study, broadly cross-reactive monoclonal antibodies (MAbs) derived from mice were humanized to minimize immunogenicity. One chimeric antibody (cAb) and seven humanized antibodies (hAbs) were constructed. These antibodies retained broad-spectrum reactivity to H5N1 viruses, binding to recombinant H5-subtype HA1 molecules expressed in CHO cells in a dose-dependent manner and exhibiting similar reactivities against antigenically distinct H5N1 viruses in hemagglutination inhibition (HI) assays. One humanized antibody, 37 hAb, showed HI and neutralization activities comparable to that of the parental murine antibody, 13D4 MAb, while the other six antibodies were less reactive to H5N1 viruses. Analysis of amino acid sequences in the variable region frameworks of the seven humanized antibodies found that Q5 and Y27 in the VH region are highly conserved murine residues. Comparison of the three-dimensional structures derived from the variable regions of MAbs 37 hAb, H1202-34, and 13D4 revealed that residue substitutions at sites 70 and 46 may be the major cause for the observed differences in binding affinity. Examination of the chimeric antibody and one of the humanized antibodies, 37 hAb, showed that both antibodies offered postinfection protection against lethal challenge with antigenically diverse H5N1 viruses in the mouse model. Chimeric and humanized antibodies which retain the broadly reactive and protective properties of murine H5-specific monoclonal antibodies have great potential for use in the treatment of human H5N1 infection.

Avian influenza viruses rarely cause human infection (4). However, sporadic human infections with H5N1, H7N7, and H9N2 avian influenza viruses have been reported in recent years (5, 9, 29). While human infections with H7N7 virus were identified only in the Netherlands in 2003 and H9N2 infection results in only mild disease, the highly virulent avian H5N1 virus causes severe and sometimes lethal illness in humans (23). The persistent prevalence of avian H5N1 viruses in poultry is the reason why sporadic cases of human infection continue to occur (1, 34). According to a recent report from the World Health Organization (WHO), avian H5N1 virus has caused 505 cases of human infection, 300 (59%) of which were fatal, in 12 countries since 2003 (37). There is concern that avian H5N1 virus may either mutate or form a reassortant virus by combining with other human influenza viruses, thereby acquiring the capacity for efficient human-to-human transmission. Preparedness for a potential outbreak or pandemic caused by highly pathogenic influenza virus is essential.

There are two main classes of drugs available for preventing and/or treating influenza virus infection: the M2 ion-channel blockers and the neuraminidase inhibitors. Resistance to the M2 ion-channel blockers, amantadine and rimantadine, is regularly detected among influenza viruses (20, 24). While resistance to one of the neuraminidase inhibitors, oseltamivir, was rare among the pandemic 2009 (H1N1) virus and seasonal H1N1 influenza viruses prior to 2007, a resistant H1N1 variant has emerged and replaced the susceptible wild-type virus, becoming the dominant population worldwide since 2007 (19, 20). Although animal tests show that oseltamivir can inhibit H5N1 virus replication effectively, clinical data from a limited number of human cases could not demonstrate clear efficacy for this treatment (8, 32). More effective treatments are necessary for combating acute viral infections such as those caused by the highly pathogenic H5N1 virus.

Passive immunotherapy is a treatment strategy used for a range of human diseases (18). The first antibody drug approved for the treatment of an infectious disease is palivizumab (Synagis), for respiratory syncytial virus infection (11). Passive immunotherapy has also played a key role in the treatment of disease caused by influenza viruses. Transfusion of human blood products from convalescing survivors reduced the mortality rates by 50% during the 1918 Spanish flu pandemic, and more recently, several severely ill H5N1 virus-infected patients were cured using plasma from convalescing survivors or vaccine recipients (17, 35, 39). The potential for application of monoclonal antibodies (MAb) in the treatment of H5N1 virus infection has been extensively investigated (7, 13, 26, 28). However, avian H5N1 viruses have continued to circulate in different regions in different genetic and antigenic variant forms since 2003, posing a considerable challenge to the development of therapeutic antibodies. It is therefore necessary to develop broadly cross-neutralizing MAbs that will recognize genetically and antigenically diverse H5N1 viruses. The hemagglutinin antigen (HA) is a homotrimeric surface protein comprised of the HA1 and HA2 subunits. The HA1 subunit is responsible for viral binding to host receptors and serves as the principal target for neutralizing antibodies. HA1 is known to undergo rapid genetic and antigenic evolution in order to evade host immunity. The HA2 subunit, which is primarily responsible for mediating the fusion of viral and cell membranes, is relatively conserved among different subtypes of influenza viruses. While most H5N1-specific antibodies have been found to target the HA1 region (7, 21, 25, 31), several recent studies have reported on antibodies with the potential to neutralize influenza viruses of different subtypes by targeting the HA2 region of the HA protein (10, 12, 15, 30, 33).

It is possible that antibodies which can bind to HA2 may possess broad cross-reactivity. However, antibodies to HA1, which contains the dominant antigenic epitopes, are more effective at neutralizing rapidly replicating influenza viruses, such as the H5N1 subtype. It is important to identify antibodies which exhibit high affinity for the HA of H5N1 virus and which are also able to neutralize different antigenic variants of H5N1. Our previous study reported on a panel of 52 broadly cross-reactive H5-specific MAbs. One of these antibodies, 13D4 MAb, showed inhibitory activity against 41 H5N1 isolates which represented all of the major genetic clades/subclades of avian H5N1 viruses characterized since 1997. This antibody also protected mice against lethal challenge with four H5N1 strains belonging to the major genetic populations currently causing human infections. More importantly, 13D4 MAb possesses extremely potent neutralizing activity against H5N1, even at the stage of infection when the virus has disseminated beyond the pulmonary system. The strong neutralizing and cross-reactive abilities exerted by 13D4 MAb may be attributable to its binding at two conserved sites within the globular head of the HA molecule (7). Of the panel of H5-specific antibodies that we have developed, 13D4 MAb neutralizes the broadest spectrum of H5N1 viruses and shows the most potential therapeutic value for controlling infections caused by current and future H5N1 variants. In the present study, we have humanized the murine 13D4 antibody by combining the variable region of 13D4 MAb with a human constant region using complementarity-determining region (CDR) grafting and combinatorial library screening and demonstrate that the chimeric and humanized versions of 13D4 antibody maintain the broad-spectrum reactivity and effectiveness in protecting mice against lethal H5N1 challenge of the original murine version.

MATERIALS AND METHODS

Virus.

The H5N1 viruses used in this study were selected from different phylogenetic clades/subclades and grown in 10-day-old specific-pathogen-free embryonated chicken eggs at 35°C for 30 h. Allantoic fluids were harvested and kept at −80°C until use. Virus titers were determined by infection of MDCK cells and are expressed as the 50% tissue culture infectious dose (TCID₅₀) per milliliter, in accordance with the method of Reed and Muench (27). All experiments involving live virus were conducted in a biosafety level 3-enhanced containment facility at the University of Hong Kong.

Humanization of 13D4 MAb. (i) Cloning of 13D4 MAb V-region genes.

13D4 MAb, which recognizes the hemagglutinin of highly pathogenic H5 avian influenza virus, was developed and purified at the National Institute of Diagnostics and Vaccine Development in Infectious Diseases, School of Life Sciences, Xiamen University. mRNA was prepared from 13D4 MAb hybridoma cells and used in first-strand cDNA synthesis with oligo(dT) primer (Promega, Madison, WI). The total cDNA was used as a template to amplify both the variable heavy- and light-chain genes using an Ig-Prime kit from Novagen Company (Darmstadt, Germany). The resultant products were cloned into pMD18-T (Takara, Japan) for sequencing.

(ii) Humanization of MAb.

Humanization of 13D4 MAb was performed using CDR grafting and combinatorial library screening. Human germ line immunoglobulin sequences were sourced from the International Immunogenetics Information System (IMGT) database (3). The human germ line light- and heavy-chain sequences with the highest degree of homology with 13D4 MAb framework regions of each chain were selected using distances-compute pairwise analysis (MEGA, version 3.1, software; The Biodesign Institute, Tempe, AZ). The 6 CDRs of 13D4 MAb were then grafted to the selected human germ line frameworks. A combinatorial phage library was then constructed to identify the key amino acids which should be changed back to the murine sequence in the humanized framework. We identified mismatched residues either near the CDR regions or with differences in physical and chemical properties and included both the human and murine alternatives at these sites in the combinatorial library. For other mismatched residues, the human version was retained.

Full-length humanized V-region genes were generated by splicing overlapping extension PCR products for insertion into the pCANTAB 5E variable fragment (single-chain Fv [scFv]) vector (Pharmacia Biotechnology). Codons for the murine and human alternatives at library positions were included in the synthesis of the humanized V-region oligonucleotides. The library was generated after electroporation into bacterial strain ER2738 (NEB). Library panning was carried out using the H5N1 virus Ck/HK/YU22/2002 (YU22), captured by MAbs precoated on microtiter plates. Humanized antibodies (hAbs) were selected by phage enzyme-linked immunosorbent assay (ELISA), using either YU22 captured by MAbs specific for H5N1 HA or CHO-expressed recombinant HA1 coated onto microtiter plates.

(iii) Construction of chimeric and humanized IgG1 expression plasmids.

The Flp-In system (Invitrogen) was used to generate stable mammalian expression cell lines by Flp recombinase-mediated integration. The pcDNA5/FRT/TO vector (Invitrogen) was modified to incorporate two P_CMV promoters, a cloning site, and bovine growth hormone and polyadenylation signal (BGH pA) to allow expression of antibody heavy and light chains by the same plasmid. The genes encoding the human gamma and kappa constant regions were separately inserted downstream of the two P_CMV promoters.

Sequence-specific primers were then designed and used to amplify the variable regions of the mouse antibody and humanized antibodies, which were then cloned into the above-mentioned expression vector in frame with the constant regions. Expression of this construct leads to the production of chimeric antibodies (cAbs) or humanized antibodies containing human IgG1 constant regions.

(iv) Generation of stably expressing CHO cell line.

The pOG44 plasmid (Invitrogen) and the chimeric or humanized IgG1 expression plasmids were cotransfected into the Flp-In CHO cell line (Invitrogen). Upon cotransfection, the Flp recombinase expressed from pOG44 mediated a homologous recombination event such that the IgG1 expression construct was inserted into the genome at the integrated Flp recombination target site. Stable cell lines expressing chimeric or humanized IgG1 were selected on the basis of hygromycin resistance and expression of the chimeric or humanized antibody.

(v) Expression and purification of chimeric and humanized antibodies.

Chimeric or humanized antibodies were expressed by the stable CHO cell lines in serum-free SFM4 medium (HyClone, Australia). The stable CHO expression cells were first grown to 3 × 10⁶/ml in Dulbecco modified Eagle medium containing 10% fetal bovine serum, which was then changed to serum-free SFM4 medium, and supernatants were collected after a further 5 to 7 days. The chimeric and humanized antibodies were purified using protein A-Sepharose beads (Millipore). The purity of the chimeric and humanized antibodies was confirmed by SDS-PAGE using BandScan software (version 5.0).

Reactivities of chimeric and humanized antibodies with viral antigen. (i) Binding activity assay.

Serial dilutions of chimeric and humanized antibodies (initial concentration, 25 μg/ml) were added to microplate wells coated with recombinant HA1 expressed by CHO cells (200 ng/well). The unrelated 4D11 cAb to the hepatitis B virus (HBV) pre-S1 region was used as a negative control, and goat anti-human-horseradish peroxidase (HRP) (1:2,000) was used as the secondary antibody.

(ii) Competitive ELISA.

The H5N1 virus YU22 was captured by antibodies coated on microplate wells, and then serial dilutions of chimeric or humanized antibodies (initial concentration, 50 μg/ml) and HRP-conjugated parent antibody were added simultaneously and allowed to compete for virus binding. The unrelated 4D11 cAb to HBV pre-S1 was used as a negative control, and parent antibody alone was used as a positive control.

(iii) HI tests.

Hemagglutination inhibition (HI) was performed to assess antibody reactivity against different H5N1 isolates, in accordance with the World Health Organization Manual on Animal Influenza Diagnosis and Surveillance, modified as previously described (36, 38).

Molecular modeling of three-dimensional (3D) structures of Fvs.

Fvs of the murine anti-HA antibody (13D4 MAb) and humanized antibodies (37 hAb and H1202-34) were built using homology modeling. BLAST searches of the Protein Data Bank (PDB) for light-chain and heavy-chain sequences were performed. Because the templates of 13D4 MAb exhibited higher identities than the two humanized antibodies, this antibody was modeled first and the modeled structure was then further used as a template for modeling of 37 hAb and H1202-34. Homology modeling of the light chains and heavy chains of the three Fv structures was performed separately using Discovery Studio software (Acceryls, San Diego, CA). Low-identity regions with more than 4 contiguous mismatched amino acids were optimized separately using the Model Antibody Loops module. The light chain and heavy chain were then placed together by superimposition onto one of the template models. The initial structure was subjected to 1,000 steps of steepest descent and 2,000 steps of conjugate gradient minimization under a CHARMm force field. Bump checks were performed, and unreasonable bumps were removed using the Side-Chain Rotating tool (Discovery Studio). The quality of the homology models was determined by examining the distribution of amino acid residues in a Ramachandran plot.

Therapeutic efficacy of 13D4 chimeric and humanized antibodies in a mouse model.

The protocol for the animal experiments was approved by the Ethical Committee for Laboratory Animals at Xiamen University. Groups of 6 BALB/c mice aged 6 to 8 weeks were anesthetized with isoflurane and then inoculated intranasally with 10 times the 50% mouse lethal dose (MLD₅₀) of H5N1 strain A/VNM/1194/2004, A/BH Goose/QH/15C/2005, A/IDN/5/2005, or A/SZ/406H/2006. 13D4 cAb or humanized antibody was diluted in phosphate-buffered saline (PBS) to a final volume of 100 μl and at 24 h after inoculation was intravenously injected via the tail vein at a dose of 20 mg per kg of body weight. Animals in the control group were injected with an equivalent volume of PBS at 24 h after infection. To determine the dose response, additional groups of mice challenged with 10 MLD₅₀s of the A/BH Goose/QH/15C/05 strain received 13D4 cAb or hAb at doses of 5 mg/kg or 10 mg/kg at 24 h postinfection. Animals were observed daily for morbidity and mortality, and body weights were measured for up to 14 days following infection.

RESULTS

Humanization of 13D4 MAb.

13D4 MAb was previously shown to have broadly cross-protective properties against antigenically distinct variants of H5N1 viruses in a mouse model (7). In the present study, variable region genes were cloned from 13D4 hybridoma cells, and humanized 13D4 antibodies were generated using CDR grafting and combinatorial library screening techniques. The human germ line light- and heavy-chain sequences 1-39-01 and 1-69-04 were selected as the framework for CDR grafting, as they showed the highest degree of homology with the corresponding regions of 13D4 MAb framework (Fig. 1). There are 12 and 8 mismatched residues in the VH and VK framework regions, respectively, between 13D4 MAb and selected human germ line sequences. These residues either near the CDRs or with differences in physical and chemical properties were accommodated by incorporating both human and murine alternatives in the design of the combinatorial library. Other mismatched residues were retained in the human configuration. A phage scFv library with an actual capacity of 1 × 10⁷ PFU was constructed and screened against the YU22 strain of H5N1 virus (7). Humanized antibodies were selected by phage ELISA. Seven phage clones which tested positive for binding reactivity to YU22 and the recombinant HA1 protein were selected (Fig. 2). The aligned amino acid sequences of the seven humanized antibodies are shown in Fig. 1.

FIG. 1.

Alignment of amino acid sequences of variable region frameworks of the murine 13D4 MAb, human germ line sequences, and humanized antibodies. The human germ line light- and heavy-chain sequences 1-39-01 and 1-69-04 were selected as the framework for CDR grafting. The seven humanized antibodies identified in this study were designated 37 hAb, H1202-34, H8, H1121-37, H1108-18, H1213-21, and Y1121-29. The dashed lines represent amino acids that are identical to residues in the human germ line sequences.

FIG. 2.

The humanized antibody clones obtained using a phage display system showed reactivity against both YU22 virus and recombinant HA1 protein in ELISAs. Phage clones were tested against CHO-expressed recombinant HA1 protein coated on microtiter plates or the H5N1 virus YU22 captured by specific MAbs coated on microtiter plates. M13K07 phage was used as a negative control. The phage clones that reacted with both YU22 virus and recombinant HA1 were identified as positive clones. OD, optical density.

Reactivity of chimeric and humanized antibodies with viral antigen.

The Flp-In system was used to generate mammalian cell lines that stably express chimeric and humanized IgG1. The chimeric and humanized antibodies were expressed in serum-free medium, and the antibodies were then purified to >95% homogeneity using protein A-Sepharose beads for use in the following experiments.

The purified 13D4 chimeric and humanized antibodies expressed by CHO cells bound to recombinant HA1 in a dose-dependent manner (Fig. 3). Two of the humanized antibodies, 37 hAb and H1126-08, showed the highest binding activity with HA1, while another humanized antibody, H1202-34, exhibited the lowest reactivity with HA1. In a competitive ELISA, the parental 13D4 MAb, the chimeric antibody (cAb 13D4), and 37 hAb and H1202-34 but not the control antibody (4D11 cAb) blocked the binding of HRP-labeled 13D4 MAb to YU22 virus (Fig. 4).

FIG. 3.

The purified 13D4 chimeric and humanized antibodies bound to recombinant HA1 protein in a dose-dependent manner. Serial dilutions of chimeric and humanized antibodies were added to microplates coated with recombinant HA1 protein. The 4D11 cAb to HBV pre-S1 was used as a negative control, and GAH-HRP was used as the secondary antibody. Tests were performed in triplicate, and results are expressed as mean absorbance ± standard deviation. OD, optical density.

FIG. 4.

The 13D4 chimeric and humanized antibodies blocked the binding of HRP-labeled 13D4 to an H5N1 virus in the competitive ELISA. The H5N1 virus YU22 was captured by antibodies precoated on microplates, and then serial dilutions of the chimeric or humanized antibodies and HRP-conjugated parental 13D4 antibody were simultaneously added to the plate to compete for binding to the captured H5N1 virus. The 4D11 cAb to HBV pre-S1 was used as a negative control, and parental 13D4 antibody alone was used as a positive control. Tests were done in triplicate, and results are shown as mean absorbance ± standard deviation. OD, optical density.

H5 broad-spectrum cross-reactivity of 13D4 chimeric and humanized antibodies in hemagglutinin inhibition test.

Fourteen H5N1 virus strains belonging to seven major genetic clades/subclades were used to verify if 13D4 chimeric and humanized antibodies retain broad-spectrum H5 reactivity in the HI test (Table 1). Similar HI reactivities were observed for 13D4 cAb, 37 hAb, and the parent antibody in their binding to all 14 of the H5N1 virus strains used in the test, indicating that 13D4 cAb and 37 hAb retain the broad-spectrum reactivity and affinity of the parental murine antibody. The other six humanized antibodies also showed cross-reactivity but showed lower titers, indicating a decreased affinity for binding to virus.

TABLE 1.

HI analysis of 13D4 cAb and hAbs^a

H5N1 strain	Clade	Antibody titer
		13D4 MAb	13D4 cAb	H8	37	H1121-37	H1202-34	Y1213-21	H1121-29	H1108-18	H1-18G9
VNM/1194/2004	1	1,600	800	400	3,200	800	400	400	400	400	<10
HK/213/2003	1	>12,800	6,400	6,400	>12,800	1,600	1,600	3,200	1,600	3,200	<10
DK/VNM/283/2005	1	3,200	1,600	800	3,200	1,600	400	800	400	800	<10
CK/IDN/2A/2004	2.1	>12,800	6,400	6,400	6,400	6,400	1,600	1,600	6,400	3,200	<10
Indonesia/5/2005	2.1	6,400	3,200	1,600	3,200	1,600	400	800	1,600	1,600	<10
DK/IDN/MS/2004	2.1	6,400	1,600	800	1,600	400	200	800	800	800	<10
CK/Bantul/BBVet-1/2005	2.1	3,200	1,600	800	3,200	400	400	800	800	800	<10
BHGS/QH/15/2005	2.2	6,400	3,200	3,200	3,200	3,200	800	1,600	1,600	1,600	<10
CK/VNM/568/2005	2.3	6,400	3,200	800	6,400	1,600	400	800	1,600	1,600	<10
CK/GY/3570/2005	2.3	3,200	3,200	800	3,200	1,600	400	800	800	1,600	<10
Shenzhen/406H/2006	2.3	6,400	1,600	1,600	3,200	800	800	800	400	400	<10
CK/YN/115/2004	2.4	6,400	1,600	1,600	6,400	800	800	800	400	800	<10
A/CK/HK/YU22/2002	8	3,200	1,600	800	3,200	800	400	400	400	400	<10
MDK/JX/1653/2005	9	3,200	800	800	1,600	800	400	400	400	400	<10
New Caledonia/20/1999 (H1N1)		<10	<10	<10	<10	<10	<10	<10	<10	<10	3200

^aPurified 13D4 murine, chimeric, and humanized antibodies at a concentration of 2 mg/ml were titrated in a hemagglutination inhibition assay against 14 H5N1 isolates belonging to seven of the major genetic clades/subclades of H5N1 viruses. H1-18G9 against New Caledonia/20/1999 (H1N1) was used as a negative control. Serial 2-fold dilutions of each antibody were made, and the highest dilution that inhibited hemagglutination was read as the HI titer of the antibody. BHGS, bar-headed goose; CK, chicken; Dk, duck; Gs, goose; GY, Guanyang; HK, Hong Kong; IDN, Indonesia; JX, Jiangxi; MDk, migratory duck; QH, Qinghai; VNM, Vietnam; YN, Yunnan.

Homology modeling of 3D structures of variable fragments of 13D4 MAb, 37 hAb, and H1202-34.

Of the seven humanized antibodies, 37 hAb showed the highest activity in the indirect ELISA, where binding to recombinant HA1 protein expressed from CHO cells was tested, and also in the competitive ELISA, which tested binding to the YU22 H5N1 virus. Another humanized antibody, H1202-34, showed the lowest binding activity and reactivity in the ELISA and the HI test, respectively. While 37 hAb retains 16 murine residues in the framework regions, the H1202-34 antibody retains only 8 (Fig. 1). Computerized 3D structures of the variable fragments of 13D4 MAb and these two humanized antibodies were built using homology modeling to analyze the structural basis for the difference in binding activity. Using the PDB BLAST search tool, we selected structures reported by different laboratories and obtained the homology modeling templates using different methods. In the light-chain templates, the average identities are about 80% to 90% and no unaligned loop regions were detected. For the heavy-chain templates, the average identity is only about 75%, and the loop region could be detected on CDR2 and CDR3 (more than 4 residues). 13D4 MAb was analyzed first, and the modeled structure of 13D4 MAb was then further used as a template for modeling of 37 hAb and H1202-34.

The modeled structures of 37 hAb and H1202-34 are shown in ribbon mode in Fig. 5. Ramachandran plots of each final model showed that most of the backbone angles are in the allowed region. Ignoring glycine, only three residues are detected in the disallowed region: 13D4 MAb:VH:Thr102, 37 hAb:VH:Ala103, and H1202-34:VH:Ala103. The root mean square differences of 37 hAb and H1202-34 with 13D4 MAb are 0.437 and 0.494, respectively. By superimposing the 37 hAb and H1202-34 models onto 13D4 MAb structures, each mismatched residue in 37 hAb and H1202-34 was examined for its tendency to disrupt the original folding. Most of these residues fit well at their positions and may not affect the adjacent paratope; the exceptions were H70 and L46. H:F70I was observed on H1202-34, which has lost most of the binding activity of the parent antibody. The phenyl of H70:F on 13D4 MAb and 37 hAb may form weak van der Waals interactions with the adjacent CDR residues (H60:Y, H58:I, H50:E) (Fig. 6 a). For H1202-34, the absence of phenyl may result in a collapse in this region (CDR-H2). L46:L on H1202-34 is another key residue which might alter the antibody paratope. The long leucine side chain may push the phenyl of H109:Y (CDR-H3) and L55:Y (CDR-L2) away and form van der Waals interactions with them (Fig. 6b and c). As this interaction involves both the light chain and heavy chain, the mismatched residue may also affect the overall stability of the variable fragment.

FIG. 5.

Fv structures of the humanized antibodies 37 hAb (a) and H1202-34 (b). Mutations are rendered in stick mode and colored according to atom type. The CDRs of each chain are yellow.

FIG. 6.

Molecular modeling analysis of the mismatched residues of the 37 hAb and H1202-34 humanized antibodies. CDR regions are rendered in stick mode and blue. (a) The human residue Ile (green) and murine residue Phe (orange) at position H:70 are shown superimposed. The distances between H70:F and the adjacent CDR residues are indicated by green lines labeled with the distance value. (b) van der Waals interactions of L46:A (37 hAb) with the adjacent phenyl on CDR residues. (c) van der Waals interactions of L46:L (H1202-34) with the adjacent phenyl on CDR residues.

Therapeutic efficacy of 13D4 chimeric and humanized antibodies in a mouse model.

13D4 cAb and 37 hAb were examined for their ability to protect mice against infection with antigenically distinct H5N1 viruses from clades 1, 2.1, 2.2, and 2.3. Groups of 6-week-old female mice were challenged with lethal doses of different strains of H5N1 viruses and then treated with 20 mg/kg of 13D4 chimeric or humanized antibody or PBS placebo at 1 day postinfection. All of the control animals, which received only PBS, exhibited continuous weight loss following virus challenge and died on day 7 or 8 postinfection. Administration of antibody 1 day after virus challenge was found to arrest the weight loss caused by all 4 of the different H5N1 virus strains, with mice rapidly recovering and gaining weight. The chimeric and humanized antibodies afforded 100% protection against infection with three of the strains, A/VNM/1194/2004 (clade 1), A/BH Goose/QH/15C/2005 (clade 2.2), and A/SZ/406H/2006 (clade 2.3), and 83% protection against a high dose of A/IDN/5/2005 (clade 2.1) (Fig. 7). Previous studies using HIV MAb did not prevent mortality (7). To examine if lower doses of 13D4 chimeric and humanized antibodies also provided protection against H5N1 virus infection, mice were inoculated with a lethal dose (10 MLD₅₀s) of A/BH Goose/QH/15C/2005 virus, followed by intravenous injection of 5 mg/kg or 10 mg/kg of antibody at 1 day postinfection. Both the chimeric and humanized antibodies showed 100% protection of infected mice at these reduced doses.

FIG. 7.

Efficacy of chimeric and humanized 13D4 MAb in protecting mice against lethal challenge with different H5N1 influenza virus strains. Mice were challenged with lethal doses of A/VNM/1194/2004, A/IDN/5/2005, A/BH Goose/QH/15C/2005, or A/SZ/406H/2006 via the intranasal route and treated intravenously with 20 mg per kg of body weight of 13D4 cAb or 37 hAb at 24 h after infection. Mice were monitored daily for 14 days for survival and body weight change. Control groups were injected with PBS at 24 h after infection. Log-rank tests calculated the P values between the therapeutic group and control group to be <0.01 and the P values between the therapeutic groups to be >0.05.

DISCUSSION

Highly pathogenic avian H5N1 virus has caused more than 500 confirmed human infections, with approximately 60% resulting in death, according to the World Health Organization. Currently, there is no effective treatment for H5N1 virus infection, and the efficacy of neuraminidase inhibitors, the principal anti-influenza drugs, is not clear. Passive immunotherapy has increasingly been used in the treatment of infectious diseases in recent years (1). There are reports of antisera from patients who have recovered from H5N1 virus infection or from H5N1 vaccine-immunized individuals being used to treat H5N1-infected patients (35, 39). Our group and others have reported that H5 monoclonal antibodies show promising efficacy for inhibition of rapidly disseminating H5N1 virus in mouse models (7, 13, 26, 28). However, avian H5N1 viruses continue to circulate in broad geographic regions and have established regional genetic lineages with distinct antigenic properties (6, 29). Therefore, it is necessary to identify MAbs with broadly cross-protective and potent inhibitory properties for therapeutic purposes.

Most of the neutralizing antibodies for influenza virus target the globular head region of the HA molecule. Research groups, including ours, have attempted to characterize MAbs which bind to the conserved epitopes located on the HA1 subunit (7, 25, 31). Recently, several studies have reported on MAbs which recognize a highly conserved helical region in the membrane-proximal stem of HA1/HA2 and which demonstrate heterosubtypic neutralizing abilities against H1, H2, H5, H6, H8, and H9 subtype viruses (10, 30, 33). A universal antibody for all subtypes of influenza A viruses could be an ideal tool for the treatment of infections caused by antigenically diverse viruses. However, possession of broad cross-reactivity must not compromise the binding affinity of the MAb for HA molecules, as antibodies with potent inhibitory effects constitute the most desirable therapeutic regime for the control of acute infections, such as those caused by H5N1 virus.

Avian H5N1 virus has evolved rapidly in poultry in wide geographic regions since 2003 (6, 29, 38). The clade 2 H5N1 viruses represent the major population; these have caused all of the human infections in Indonesia (clade 2.1), Egypt (clade 2.2), and Vietnam and China (clade 2.3) since 2005. Most of the previous studies were not able to verify the ability of antibodies to neutralize all three of these antigenic groups of H5N1 viruses and, in particular, the clade 2.3 H5N1 viruses, which exhibit the most distinct antigenic properties (38). More importantly, the efficacies in mouse models indicated in the different studies could not be compared directly, due to the different inoculation doses being used. While clinical trials of treatments for H5N1 virus infection are difficult to conduct, as H5N1 human cases are still rare and occur only sporadically, it is important to evaluate the ability of various MAb clones to protect mice from infection with antigenic variants of H5N1 virus and also their virus-neutralizing properties. The murine 13D4 MAb showed the broadest spectrum of reactivity against 41 H5N1 isolates belonging to 7 major genetic clades/subclades (1, 2.1, 2.2, 2.3, 2.4, 8, and 9) and also efficiently protected mice against lethal challenge with four H5N1 strains representing the genetic variants which have caused human infections since 2003 (7). In addition to the strong neutralizing abilities and broad cross-reactivity which 13D4 MAb possesses, it may be further characterized by its binding to the conserved sites near the receptor binding region of the HA1 molecule (7). Humanization of antibodies such as 13D4 provides a great opportunity for the development of therapeutic antibodies for H5N1 virus infection. This study showed that chimeric and humanized 13D4 antibodies in which variable domains from the murine antibody were combined with human constant domains retained the binding affinity and broad-spectrum reactivity of the parental antibody.

The two main approaches applied in antibody humanization involve rational and empirical methods (2). Rational methods include CDR grafting, resurfacing, superhumanization, and human string content optimization. Of these, CDR grafting, which involves the transfer of CDRs of a murine antibody onto a human framework sequence, is the most commonly used method. Empirical methods rely on the generation of large libraries of humanized variants, followed by selection of the best clones using enrichment technologies. In the present study, CDR grafting was combined with framework library screening to humanize 13D4 H5 MAb. The human germ line light- and heavy-chain sequences 1-39-01 and 1-69-04 were selected as the framework for CDR grafting, as they showed the highest degree of homology with the corresponding 13D4 MAb framework regions. This high-throughput screening technique makes the antibody humanization process simple and highly efficient.

Although several humanized antibodies designed for treatment of human infections have been reported in the literature (7, 16, 30), none are yet available for clinical use. The process of humanizing a broadly cross-reactive H5 monoclonal antibody and the molecular basis for the differences in the binding affinity of the resulting humanized antibodies are described in this report. The 13D4 chimeric and humanized antibodies which have retained the broad reactivity and protective properties of the murine version may provide a model for development of therapeutic antibodies for the treatment of human H5N1 infections. In response to the rapid evolution of influenza virus and the uncertainty as to which antigenically distinct variants may emerge in the future, a combinatorial approach, in which two groups of MAbs which target the HA1 and HA2 regions, respectively, are used, may provide an effective solution for the treatment of human infection with highly pathogenic influenza A viruses.