A High-Affinity CDR-Grafted Antibody against Influenza A H5N1 Viruses Recognizes a Conserved Epitope of H5 Hemagglutinin

Highly pathogenic avian influenza (HPAI) H5N1 virus infection is still a potential threat to public health worldwide. While vaccines and antiviral drugs are currently under development, neutralizing antibodies could offer an alternative strategy to prevent and treat H5N1 virus infection. In the present study, we had developed a humanized antibody against H5N1 viruses from mouse-derived hybridoma in order to minimize its immunogenicity for potential clinical application. The humanized antibody hH5M9 was generated by transferring the mouse complementarity determining region (CDR) residues together with four key framework region (FR) residues onto the FR of the human antibody. This humanized antibody exhibited high affinity and specificity comparable to the parental mouse or chimeric counterpart with broad and strong neutralization activity against all H5N1 clades and subclades except for Egypt clades investigated. Furthermore, through epitope mapping we identified a linear epitope on the top region of hemagglutinin (HA) that was H5N1 specific and conserved. Our results for the first time reported a humanized antibody against H5N1 viruses by CDR grafting method. With the expected lower immunogenicity, this humanized antibody was expected to be more efficacious than murine or human-mouse chimeric antibodies for future application in humans.


Introduction
H5N1 virus, one of highly pathogenic avian influenza (HPAI) strains, has caused numerous outbreaks in poultry in Southeast Asia since 1997 [1][2][3], and more recently continues to spread globally. These outbreaks are accompanied by the occasional transmission of HPAI H5N1 virus to humans, resulting in a total of 628 cases with 374 deaths in 15 countries since 2003 [4].
To prevent H5N1 pandemic, world-wide efforts have been made to develop and stockpile preventive vaccines, antiviral drugs as well as passive immune therapies [5,6]. Vaccine strategies have been found to be only effective at preventive stage whereas they can be easily hindered by antigenic variation of the influenza strains [5]. Antiviral treatment is an ideal method. But currently available options are limited [6]. Antibody-based treatment has been successfully used prophylactically against many virus-infected diseases, such as those caused by hepatitis A virus, hepatitis B virus, cytomegalovirus, rabies virus, varicella virus and respiratory syncytial virus infection [7]. Thus, it is feasible to induce humoral immunity in humans through preventive vaccination and neutralizing antibody generation to protect against H5N1 virus infection. In addition, passive immune therapies have been further highlighted by transfusion of human convalescent sera leading to a 50% reduction in influenza mortality during the 1918 Spanish influenza pandemic, and more recently by anecdotal reports of treating H5N1 human infection with convalescent sera in China [8,9].
Influenza hemagglutinin (HA), with 16 antigenic distinct subtypes, is the main target for neutralizing antibodies against influenza viral infections [10][11][12][13][14][15][16]. Three HA monomers, each consisting of one HA1 and one HA2 subunit, form the HA molecule on the surface of influenza virion. The HA1 subunit, globular head domain of HA molecule, is the most immunogenic region of the HA protein, containing the receptor binding site which mediates viral attachment to the host cell membrane. The HA2 subunit constitutes the core fusion machinery in the stalk region [17,18]. Several research groups have reported that monoclonal antibodies (mAbs) against the HA protein of the influenza virus could confer prophylactic and therapeutic protection in mouse models [14,15,[19][20][21].
Although antibody-based therapy has displayed the potential prevention and treatment of H5N1 virus infection in human, the efficacy of murine mAbs is hampered by their diminished serum half-life, inability to trigger human effector functions and the induction of a human anti-mouse antibody (HAMA) response [22,23]. To counter these problems, several strategies have been devised including the generation of chimeric, humanized or human antibodies. CDR grafting is a way to humanize murine antibodies by grafting the complementarity determining regions (CDRs) of a murine mAb onto the framework regions (FRs) of a human antibody while retaining those rodent FR residues that influence antigen-binding activity [24][25][26]. Compared with chimeric antibodies, CDR-grafted antibodies possess lower immunogenicity with more successful application in clinic [27][28][29].
In the previous study, the mouse monoclonal antibody H5M9 (mH5M9) demonstrated broad and strong neutralizing activity against H5N1 viruses isolated from 1997 to 2006 [30]. Accordingly, here we described the generation and characterization of a CDR-grafted anti-H5N1 antibody derived from mH5M9. We further identified a linear epitope on the top of HA globular region recognized by the engineering antibody, which was highly conserved in different clades of currently epidemic avian H5N1 viruses. The development of this CDR-grafted antibody might provide an alternative strategy in prevention and treatment of human H5N1 virus infection.

Generation of chimeric and CDR-grafted antibodies against HA of H5N1 viruses
To generate a human-mouse chimeric antibody, the variable regions of both light chain (VL) and heavy chain (VH) from mH5M9 were subcloned into IgG expression cassette vectors IFH and IFL, respectively. IFH-VH and IFL-VL vectors were coexpressed in 293F cells, yielding chimeric H5M9 (cH5M9) ( Figure 1B).
For the construction of CDR-grafted H5M9 antibody, a BLASTP-search against the non-redundant protein database was performed to screen the template for further humanization. The results showed that VH and VL domains of human antibody FabOX108 (PDB No. 3DGG_B and No. 3DGG_A, respectively) were the most homologous to that of mH5M9, respectively ( Figure 1A). We thus selected the FRs of this template for grafting CDR residues of mH5M9. Genes from three CDRs of mH5M9 VH/VL regions were directly grafted onto human antibody FabOX108 VH/VL frameworks to generate CDR-grafted antibody gene. Meanwhile, four FR residues of mH5M9 VH region (Met 37 , Arg 66 , Va l71 , Val 109 ) were retained, which were deduced to influence the conformations of CDRs according to the simulation of three dimensional structure of VH and VL regions ( Figure S2). The recombinant CDR-grafted VH and VL genes were cloned into IgG expression cassette vectors and co-expressed in 293F cells to obtain a CDR-grafted antibody, named hH5M9 ( Figure 1B).
The amino acid residues of the variable regions of mH5M9, CDR-grafted hH5M9 and FabOX108 antibodies were listed in Figure 1A, which were numbered according to the convention of Kabat et al. [31].

Affinity Determination of chimeric and CDR-grafted H5M9 antibodies
An indirect ELISA was first used to analyze the antigen-binding activity of cH5M9 and hH5M9 against H5 HA (A/Vietnam/ 1194/04). As shown in Figure 2A, cH5M9 had the similar antigenbinding activity to mH5M9, while the antigen-binding activity of hH5M9 was lower than that of mH5M9, suggesting that the humanization of recombinant antibody by CDR grafting might result in the structural alteration of the antigen-binding site more or less and the subsequent decrease of antigen-binding capacity [32].
To further quantify the binding ability of the engineering cH5M9 and hH5M9 antibodies, kinetic binding constants were determined by ProteON surface plasmon resonance (SPR) assay using H5 HA (A/Vietnam/1194/04) coated biosensor. According to the association and dissociation curves of antibodies shown in Figure 2B, the equilibrium dissociation constant (KD) values of mH5M9, cH5M9 and hH5M9 were calculated to be 3.05610 211 , 4.31610 211 and 3.84610 210 M, respectively ( Figure 2C). Although there was no significant difference among the association rate constants (Ka) of cH5M9, hH5M9 or mH5M9, the dissociation rate constant (Kd) of hH5M9 was slightly faster than other two antibodies, which implied that the substitution of certain residues in antigen-binding sites of VH or VL domains in hH5M9 reduced the binding stability after CDR grafting when compared with its parental or chimeric counterpart.

Neutralization breadths and potencies of chimeric and CDR-grafted antibodies in vitro
In the previous study, mH5M9 showed broad and strong neutralization activity against H5N1 viruses [30]. We also performed hemagglutination inhibition (HI) assays against a panel of HA strains, including H1, H7, H9, and H5, to determine the neutralization activities of cH5M9 and hH5M9 antibodies. cH5M9 and hH5M9 inhibited HAs of all five H5N1 strains tested in a concentration between 7.81 and 15.62 mg/ml whereas failed to inhibit hemagglutination of H1, H7 and H9, exhibiting similar specificity to mH5M9 (Table 1).
To further determine the neutralization breadths and potencies of cH5M9 and hH5M9, we had chosen nine H5 HAs subclassified into clades 0, 1, 2 and 4 for further titration, which had been associated with human or avian pathogenesis to date ( Table 2). cH5M9 and hH5M9 possessed the similar neutralization activity as mH5M9 against all of the H5N1 clades tested, ranging from 0.1 to 12.5 ng/ml, except for clade 2.2.1 virus strain A/Egypt/N05056/09. These result indicated that CDR-grafted hH5M9 might recognize a shared epitope on HAs from almost all clades and subclades of the H5 subtype that was not shared by HAs of H1, H7, and H9 subtypes.

Epitope mapping
To anchor the epitope recognized by hH5M9, two fragments of HA, HA1 and HA2, were first prepared and adapted for epitope mapping. Through western blotting analysis, it was demonstrated that hH5M9 antibody reacted with the viral HA1 protein, rather than HA2. Considering the denaturation of HA1 during the western blotting analysis, the epitope targeted by hH5M9 appeared to be linear rather than conformational ( Figure 3A). According to our above results that hH5M9 displayed broad crossreactivity to H5N1 HAs but not A/Egypt/N05056/09 (Table 2), we first aligned HA1 amino acid sequences of these nine H5N1 isolates used in the binding assay to find out the unique amino acid residues of HA1 of A/Egypt/N05056/09 (Table 3). We found that Arg 22 , Asn 120 , Thr 151 , Gln 152 , Asp 154 , Ile 210 and Ser 235 were unique to A/Egypt/N05056/09, which inferred these seven amino acids were the candidates for antigenic epitope according to the previous study [14].
To test the roles of these seven residues on antigen-antibody recognition of the H5 HA protein, we generated seven HA1 mutants based on wild type HA from A/Anhui/1/05 by sitedirected mutagenesis. HA1 mutants with single amino acid mutation illustrated in Table 4 were expressed in 293T cells, and then immunofluorescence assays (IFA) were used to detect the interactions between the hH5M9 and the mutant HA1 proteins, while a rabbit polyclonal antibody against A/Anhui/1/05 was used as the positive control. As shown in Table 4, mutation of HA P235G led to the abolishment of the binding ability of HA molecule to hH5M9 whereas other substitutions had no influence on antigen-antibody interaction. These results indicated that Pro 235 in HA was critical for the binding to hH5M9.
To further identify the epitope localization, additional twelve HA site-mutants around Pro 235 were generated and screened by hH5M9 (Table 4). Amino acid residues from 231 to 243 (except for 235) on HA were shifted to glycine individually by site-directed mutagenesis as described elsewhere [14]. As showed in Table 4, the mutants of HA N236G , HA D237G , HA I239G , HA N240G and HA F241G displayed complete loss of hH5M9 binding ability whereas mutations at HA 234 and HA 238 resulted in substantially reduction of binding ability to hH5M9. All newly constructed mutants had no impact on the binding to the rabbit polyclonal antibody against A/Anhui/1/05.
Thus, we concluded that the linear epitope targeted by CDRgrafted hH5M9 antibody was located from the amino acid residues 234 to 241 on the hemagglutinin of H5 subtype with the  [31]. Residues shown by underline in frameworks were deemed essential for maintaining the combining sites of mH5M9. Dashes indicated residues that were identical in mH5M9, FabOX108 and hH5M9 whereas gaps denoted amino acid residues missing at those positions. (B) SDS-PAGE analysis of purified cH5M9 and hH5M9 under reducing conditions. mH5M9 was used as the positive control. doi:10.1371/journal.pone.0088777.g001 sequence of ''Lys-Pro-Asn-Asp-Ala-Ile-Asn-Phe'' (KPNDAINF). To further verify the epitope mapping, we have synthesized the KPNDAINF peptide coupled to KLH and determined the binding ability of hH5M9 to this peptide. The result indicated that hH5M9 interacted with peptide coupled to KLH whereas no response was detectable to KLH only ( Figure 3B), which further demonstrated the existence of KPNDAINF epitope on H5 HA targeted by hH5M9 antibody. This epitope was supposed to be located on the top region of HA1 protein according to the published crystal structure of H5 HA (VN1194) (PDB No. 2IBX) ( Figure 3C), which was never reported before. The illustration of the three dimensional structure (shown in Figure 3D) (Table 5). These data indicated that the epitope recognized by hH5M9 was conserved in epidemic avian H5N1 viruses, which made hH5M9 more extensive for its future application.

Discussion
With sporadic H5N1 avian flu outbreaks in poultry, wild birds and human, the emergence of a human-adapted H5 virus, either by reassortment or mutation, is a threat to public health worldwide. In the present study, we had successfully generated a CDR-grafted human antibody originated from high-affinity mouse hybridoma targeting HA with a spectrum of neutralizing activities against multiple strains of HPAI H5N1 virus in vitro.
To minimize the immunogenicity of murine antibodies, chimeric and CDR-grafting antibodies were generated. There are two approaches for CDR grafting [33,34]. One strategy is to fix human FRs from heavy or light chain of representative antibodies. Another one is to choose frameworks of variable regions from human antibody that are the most homologous to those of the murine one for minimizing the distortion of CDRs. Here we have used the latter approach for humanization of mouse mH5M9 antibody against H5N1 virus. CDR-grafted hH5M9 was firstly constructed by directly grafting CDRs from VH and VL of mH5M9 to sequence-homologous VH and VL domain of human antibody FabOX108. Results from antigen-binding assay indicated that this led to the elimination of the binding activity (data not shown). This was not surprising due to the fact that the transfer of CDR residues alone into a xenogeneous antibody might alter the conformational structure of the antibody, leading to the decrease of antigen-binding affinity. Through three dimensional computational modeling, we chose eight amino acid residues in the FR regions of the VH chain of mH5M9 antibody for the substitution of CDR-grafted framework residues. Back mutation of four residues on heavy chain (Met 37 , Arg 66 , Val 71 , Val 109 ) completely restored the binding avidity comparable to its mouse or chimeric counterpart. We thus deduced that the preservation of key mouse FR residues was critical for the successful design of CDR-grafted antibodies probably through constraining the CDR conformations [26].
Although antibody epitopes had been reported to be located in five of the eleven viral proteins of influenza virus, most of the epitopes were located in the viral HA [35]. For instance, antibodies 7H10 [13], AVFluIgG01, AVFluIgG03 [14], FLA5.10, FLD21.140 [36] or 65C6 [37] recognized HA1, while CR6261 bound to a highly conserved epitope on HA2 [38]. HA1 contained the receptor-binding site to host cells. The interference of interaction between HA1 and the receptor on host cells by functional antibodies largely protected the host from the infection of flu virus [20]. To date, five antigenic sites (A-E) on the H5 HA molecule had been mapped by using virus escape mutants (viral variants that could escape recognition by the monoclonal antibodies) [39,40]. They were located exclusively in the areas  corresponding to antigenic sites A and B of H3 HA and the antigenic site Sa of H1 HA [41,42].
In the present study, hH5M9 was demonstrated to recognize a linear epitope of HA1 (amino acid residues 234-241, KPNDAINF), which was located on the top region of HA1 protein. Notably, this epitope was not listed in the reported five antigenic sites A-E [39,40]. According to the three dimensional structure of H5 HA (VN1194) (Figure 3C), we found that this linear epitope (colored in green) comprised a remarkably tight cluster, far from the receptor-binding domain (RBD) of H5 HA1, as colored in yellow. We also aligned the sequences of 2376 H5N1 virus strains representing major epidemic avian H5N1 genetic clades, including 243 human clades. The results showed that the epitope was highly conserved in epidemic avian H5N1 viruses in clades 1-9 (Table 2), which was consistent with the broad crossprotection in vitro of hH5M9. The residues Lys 234 , Pro 235 , Asn 236 , Asp 237 , Ala 238 , Ile 239 , Asn 240 , Phe 241 were highly conserved among the 2376 H5N1 virus strains. Interestingly, HA Pro235Ser mutation existed in most Egypt clades belonging to clade 2.2.1. When we analyzed the conservation of this epitope in H5N1 strains reported annually, up to 94% of H5N1 strains had ''KPNDAINF'' before 2005. Inhibition on the viruses was of particular importance due to their wide geographical spread and pathogenicity in human. Based on our study, the antibody hH5M9 might counteract most of them. Meanwhile, similar epitope sequence recognized by hH5M9 in our study was found not to exist in any other HA subtypes. In other words, ''KPNDAINF'' was an H5N1-specific epitope. The H5 epitope for hH5M9 was also comprised in the H5N1 strain-specific hemagglutinin CD4 + T cell epitope [43][44][45][46]. More strikingly, it was not exactly the same epitope as parental mouse mH5M9 that was reported recently [47]. Considering that hH5M9 exhibited slightly decreased affinity to HA whereas the similar HI activity compared to the parental or chimeric H5M9, the difference between epitopes of parental mH5M9 and the CDR-grafted hH5M9 antibodies was supportive. Two possibilities might dedicate to the inconsistency of the epitopes recognized by parental or CDR-grafted antibody. One was that CDR grafting altered the amino acid compositions and the lengths of CDRs, which might change the conformation of variable regions and led to the subsequent shift of epitope recognition it used to [48][49][50]. Another was that the parental antibody might recognize two independent epitopes which had been reported in several studies [51][52][53][54]. We supposed that with its high affinity, conservation and specificity, this newly-defined linear epitope might take priority in peptide vaccine designing over conformational epitope against flu infection.
In summary, we had generated a CDR-grafted antibody that exhibited high neutralizing properties in vitro against H5N1 viruses. With the expected lower immunogenicity and the conservation of the epitope recognized by CDR-grafted antibody, we intended to expect that it might exhibit high neutralization breadth and potency in vivo and would be employed as adjunctive treatment against human H5N1 virus infection.

Materials
Mouse monoclonal antibody mH5M9 was elicited by immunization of mice with H5 HA concentrated from H5N1 virus (A/ goose/Guangdong/1/96) and characterized as described previously [30]. mH5M9 antibody was purified from mouse ascites by caprylic acid-ammonium sulfate precipitation. The genes of heavy chain variable domain (VH) and light chain variable domain (VL) of mH5M9 were cloned from hybridoma by using reverse transcription-polymerase chain reaction (RT-PCR) routinely and subcloned in pMD19-T plasmid.   Designing strategy of chineric and humanized H5M9 antibodies To generate chimeric H5M9 antibody (cH5M9), the VH and VL gene fragments of mH5M9 monoclonal antibody in pMD19-T plasmids were subcloned into IgG expression cassette vectors IFH with the constant region of human antibody heavy chain (IFH-VH) and IFL with the constant region of human antibody k light chain (IFL-VL), respectively. More detailed information of expression vectors was indicated in Figure S1.
To further humanize the mouse-human chimeric antibody, CDR-grafting method was used. We intended to choose human variable regions that were the most homologous to mouse variable regions as the templates for CDR grafting of mH5M9. Sequences of VH and VL domain from mH5M9 were subjected to a BLASTP search against the non-redundant Genbank database, respectively. VH and VL genes from FabOX108 were selected to construct humanized VH and VL regions by substituting CDR regions with mH5M9-derived CDR residues. The antibody structure was explicitly solvated with TIP3P water molecules using VMD [55], and then were subjected to molecular dynamics simulations. Simulations were performed with CHARMM force field using NAMD program [56]. The system was first minimized for 5000 steps with conjugate gradient method, and then was heated linearly to 300 K over 60 ps. The refined models were further assessed by VMD program. Humanized VH and VL fragment genes were fully synthesized (Generay Biotechnology, Shanghai, China) and subcloned into expression vectors.

Expression and purification of chimeric or humanized antibodies
The recombinant heavy chain and light chain expression vectors, IFH-VH and IFL-VL, for chimeric or humanized antibodies were co-transfected into 293F cells using FreeStyle TM MAX Reagent (Invitrogen, USA) according to the manufacturer's instructions. The culture supernatant was subjected to ELISA assay to determine the expression and the antigen-binding capacity of the engineering antibodies. For purification of antibodies, the culture supernatant was subjected to Protein-A affinity chromatography on a HiTrap TM MabSelect SuRe column (GE Heathcare, Sweden). The integrity and purity of chimeric and humanized H5M9 antibodies were assessed by Coomassie blue staining after SDS-PAGE electrophoresis. Protein concentration was determined by RC DC Protein Assay (Bio-Rad, USA).

ELISA
Indirect ELISA was performed to determine the specificity of antibodies. Briefly, the samples were added to 96-well plates precoated with 1 mg H5 HA overnight and incubated at 37uC for 1 h. After washing, 100 ml of horseradish peroxidase-conjugated goat anti-human IgG (KPL, USA) or goat anti-mouse IgG (Jackson ImmunoResearch, USA) were added to each well and incubated at 37uC for 1 h. Finally, 100 ml of 0.2 M citrate buffer (pH 5.0) containing 0.04% o-phenylenediamine (Amresco, USA) and 0.03% H 2 O 2 were added to each well and incubated for 10 min. The reaction was stopped by the addition of 50 ml 2 M H 2 SO 4 , and the optical density was measured at 490 nm by an ELISA plate reader (Bio-Rad, USA).
In the experiment for epitope verification, 1 mg of H5 HA from A/Vietnam/1194/04 strain, synthesized KPNDAINF coupled to KLH or KLH only were coated overnight. Diluted hH5M9 antibody were added to the wells and incubated at 37uC for 1 h. Further detection was performed as described above. The peptide containing a C-terminal cysteine for conjugation to KLH was synthesized in Sangon Biotechnology (Shanghai, China).

Hemagglutination inhibition (HI) assay
To determine the biological activity of recombinant hH5M9, the classical HI assays were performed as described elsewhere [57]. Briefly, the purified antibodies were 2-fold diluted in 96-well plates and mixed with 4 HA units of H5 viruses, including A/ goose/Guangdong/1/96 (H5N1), A/Vietnam/1194/04 (H5N1), A/duck/Anhui/1/06 (H5N1), A/Anhui/1/05 (H5N1), A/chicken/Shanxi/2/06 (H5N1), A/New Caledonia/20/99 (H1N1), A/ duck/Guangdong/1/96 (H7N3) and A/chicken/Shandong/6/96 (H9N2). HI capacities against H1, H7, and H9 viruses were simultaneously determined. Plates were incubated for 30 min at room temperature, and 1% of chicken red blood cells (RBCs) were added to each well. The initial concentration of each antibody was 1 mg/ml. The highest dilution of antibodies in which agglutination was not observed was considered to be the hemagglutination inhibition endpoint. Affinity measurement by Surface Plasmon Resonance Surface plasmon resonance (SPR) analysis of hH5M9 was performed on ProteOn surface plasmon resonance biosensor (BioRad Labs, Shanghai, China). The H5 HA (A/Vietnam/1194/ 04) was immobilized on a GLC sensor chip using an amine coupling kit with 946 resonance units (RU) in the testing flow cells. 200 ml of freshly prepared antibodies at various concentrations were injected at a constant flow rate of 100 ml/min with 120 s duration at 25uC. Responses from the HA surface were normalized by that from the mock surface without protein binding or a separate injection with buffer only. An anti-EGFR antibody C225 was used as a negative control. Binding kinetics for the antibodies and the data analysis were performed using ProteON manager software (Bio-Rad, USA).

Western blotting
H5 HA (A/Anhui/1/05) was applied to 12% SDS-PAGE and transferred to a PVDF membrane with a Trans-Blot SD semi-dry transfer cell apparatus (Bio-Rad, USA). hH5M9 (1 mg/ml) was used as primary antibody. A rabbit polyclonal antibody against A/ Anhui/1/05 was used as the positive control. IRDye 800CWconjugated goat anti-human or goat anti-rabbit IgG (LI-COR, USA) was used as the secondary antibody for visualization. The fluorescent bands were detected by Odyssey Imaging System (LI-COR, USA). digested by BamHI and XhoI (Takara, Japan) and subcloned into pcDNA3.0 vector. All constructs were confirmed by DNA sequencing and subjected to transfection for transient expression of mutant HA molecules in 293T cells. Immunofluorescence assays were subsequently performed to determine the binding ability of mutant HA molecules to hH5M9.

Immunofluorescence Assay (IFA)
IFA were performed on 293T cells transfected with mutant HA constructions. Cells were grown in 24-well plates, fixed in 3.7% paraformaldehyde, permeabilized with 0.1% Triton X-100 (Sigma-Aldrich, USA), and then stained in the wells. Purified hH5M9 (1 mg/ml) was used as primary antibody. A rabbit polyclonal antibody against A/Anhui/1/05 was used as the positive control. Binding antibodies were detected by using Cy2conjugated goat anti-human or goat anti-rabbit IgG (Jackson ImmunoResearch, USA) and observed under immunofluorescence microscope (Olympus, Japan).