Introduction The continued circulation of highly pathogenic avian influenza (HPAI) strains of subtype H5N1, and occasional coincident cases of human infection (274 patients as of 19 February 2007, with 167 fatalities), has triggered international public health concern. On the basis of haemagglutinin (HA) sequences, these circulating HPAI H5N1 viruses fall into different lineages, termed clades; viruses isolated in Viet Nam and Indonesia in 2004 and 2005, respectively, were designated as reference strains for Clades I and II [1]. The HA sequences of Clade I and Clade II viruses differ by 4% to 5% at the amino acid level and the viruses of the two clades are antigenically distinguishable. The H5N1 viruses are not efficiently transmitted from person to person. Potentially, a virus capable of efficient human-to-human transmission could result from either adaptation of the HPAI H5N1 viruses and/or reassortment of the H5N1 virus genome with that of a circulating human influenza virus. Widespread dissemination of such a virus could cause significant morbidity and mortality, since humans are generally immunologically naïve to H5 influenza subtypes. In humans, overall mortality in HPAI H5N1 infection exceeds 60%, with variation according to the patient's age and the year of infection [2]. The basis for the apparent virulence of HPAI H5N1 strains in humans is relatively poorly understood. In Vietnamese patients, the disease was characterized by severe pneumonia, lymphopenia, high viral loads in the respiratory tract, and hypercytokinemia [3]. Beyond supportive care, treatment options for human patients with H5N1 avian influenza remain limited and are empiric; some H5N1 viruses are resistant to older antiviral agents such as amantadine and rimantadine [4,5], and the clinical efficacy of neuraminidase inhibitors such as oseltamavir and zanamavir has not yet been confirmed in prospective studies. In addition, H5N1 viruses resistant to oseltamavir have been reported [6,7]. Patients who recover from infection possess antibodies that neutralize their infecting virus in vitro, suggesting that antibody-mediated immunity may contribute to resolution of infection [3]. Antibody-based therapy for human patients with H5N1 is a hitherto unexplored, but potentially viable, treatment option. Clinically, antibody therapy using polyclonal and monoclonal antibodies (mAbs) is effectively used as prophylaxis against varicella, hepatitis A, hepatitis B, rabies, and respiratory syncytial virus infections [8]. In the context of influenza, specific mAbs can confer prophylactic and therapeutic protection in mice [9–11]. Passive immunization by vertical acquisition of specific antibodies is also associated with influenza immunity in animal models and in early infancy in humans [12–15]. Of more immediate relevance, transfusion of human blood products from patients recovering from the 1918 “Spanish ‘flu” was associated with a 50% reduction in influenza mortality during the pandemic [16]. Humanized mouse mAbs and equine F(ab′)2 fragments specific for H5N1 have also been used for efficacious prophylaxis and therapy in the mouse model [17,18]. Collectively, these observations suggest that passive antibody therapy against HPAI H5N1 viruses could be a potentially viable, adjunctive treatment option in human cases of H5N1 influenza. One approach to generating virus-specific neutralizing mAbs is to use a highly efficient method of Epstein-Barr (EBV)-mediated immortalization of memory B cells from convalescent individuals [19]. This approach is rapid and yields stable B cell clones that secrete fully human antibodies that have been selected in the course of an immune response to the pathogen [19]. The aim of this study was to generate human mAbs with neutralizing activity against H5N1 viruses. mAbs were derived from immortalized memory B cells collected from donors who had recovered from H5N1 infection. Four mAbs, with neutralizing activity in vitro, had prophylactic and therapeutic efficacy in mice challenged with HPAI H5N1. These mAbs, and others like them, could have a role in adjunctive treatment of human cases of H5N1 influenza. Materials and Methods H5N1 Influenza Cases The four adult blood donors in this study (CL26, CL36, CL114, and CL115) were diagnosed with HPAI H5N1 infection between January 2004 and February 2005 at the Hospital for Tropical Diseases in Ho Chi Minh City, Viet Nam. Diagnosis of HPAI H5N1 influenza infection was made by RT-PCR on respiratory specimens (from all four donors) and culture of H5N1 influenza from respiratory specimens (from donors CL26, CL36, and CL115) [20]. The clinical features of acute disease in three of the four patients (donors CL26, CL36, and CL115) have been described previously (patients #5, #7, and #8 in the online supplement to [7]). During early convalescence (1–4 mo post-illness onset), all patients had detectable neutralizing antibody titers to their autologous virus (median 96, range 32–200). The Scientific and Ethical Committee of the Hospital for Tropical Diseases and the Oxford University Tropical Research Ethical Committee approved the study protocol. All patients provided written informed consent. Influenza Viruses GenBank accession numbers for the genomic sequences of the H5N1 viruses isolated from subjects CL26 (A/Vietnam/CL26/2004), CL36 (A/Vietnam/CL36/2004) and CL115 (A/Vietnam/CL115/2005), have been published previously [3]. Strain A/chicken/Vietnam/VL1/2006 (H5N1) was isolated from a cloacal swab of a chicken in southern Viet Nam in early 2006. The HPAI H5N1 reference viruses A/Vietnam/1203/2004, A/Hong Kong/213/2003, A/Hong Kong/491/1997, A/Vietnam/JPHN/30321/2005, and A/Indonesia/5/2005 were kindly provided by Dr. Nancy Cox, Influenza Division, Centers for Disease Control and Prevention, Atlanta, Georgia, United States. Influenza A/California/7/2004 (H3N2) was kindly provided by Dr. Roland Levandowski, CBER, Food and Drug Administration, Bethesda, Maryland, United States. Microneutralization Assays Screening of supernatants from B cell lines and clones was performed by microneutralization assay using MDCK cells and 100 TCID50 (50% tissue culture infectious doses) of A/Vietnam/1203/04 essentially as described previously [21]. Briefly, neat supernatants were incubated with 100 TCID50 of virus for 1 h at room temperature prior to addition to monolayers of MDCK cells. Cell monolayers were incubated for a further 3–4 d and examined for cytopathic effect. Determination of endpoint neutralizing antibody titers was performed in a similar fashion, except that plasma or supernatant samples were serially two-fold diluted prior to mixing with 100 TCID50 of virus. Plasma samples were tested at a starting dilution of 1:10, while supernatants were tested at a starting dilution of 1:8 and residual infectivity was tested in four wells per dilution. The neutralizing titer was defined as the reciprocal of the highest dilution of serum at which the infectivity of 100 TCID50 of the appropriate wild-type (wt) H5N1 virus for MDCK cells was completely neutralized in 50% of the wells. Infectivity was identified by the presence of cytopathic effect on d 4 and the titer was calculated by the Reed-Muench method. Immortalization of Memory B Cells and Selection of Neutralizing Clones Frozen peripheral blood mononuclear cells (PBMCs) were thawed and stained with directly labeled antibodies to CD22 (Pharmingen, http://www.bdbiosciences.com/home) and to immunoglobulin (Ig) M, IgD, and IgA (Jackson ImmunoResearch, http://www.jacksonimmuno.com). CD22+ IgM−, IgD−, IgA− B cells were isolated using a FACSAria (Becton Dickinson, http://www.bd.com) and immortalized at 30 B cells/well in replicate cultures using EBV in the presence of CpG oligodeoxynucleotide 2006 (Mycrosynth, http://www.microsynth.ch) and irradiated allogeneic PBMC, as previously described [19]. Cells were cultured in complete RPMI 1640 supplemented with 10% fetal calf serum (HyClone Laboratories, http://www.hyclone.com). Culture supernatants were harvested after 14 days and assayed for neutralizing activity against 100 TCID50 of influenza A/Vietnam/1203/04 (H5N1). Cultures with measurable neutralizing activity were cloned at 0.5 cell/well in the presence of CpG 2006 and irradiated PBMCs. B cell clones were cultured at a high cell density in complete RPMI 1640 10% Ig-depleted fetal calf serum to produce enriched supernatants containing 1–3 mg mAbs/ml. MAbs were also purified on protein G columns (GE Healthcare Europe http://www.gehealthcare.com). The isotype, subclass, and light chain of the mAbs were characterized by ELISA using specific antibodies and HRP-labeled anti-human Ig antibody (Southern Biotechnology, http://www.southernbiotech.com). Antibodies were quantified with reference to a standard certified preparation (Sigma-Aldrich, http://www.sigmaaldrich.com). MAbs for Prophylaxis and Therapy in Mice Groups of 4–8 female BALB/c mice (4–6 wk old, mean weight 18 g) were used in all experiments. Inoculation of mice and tissue harvests were performed in a biosafety cabinet by personnel wearing powered air purifying respirators. Influenza-infected animals were housed in a USDA and CDC accredited biosafety level 3 (BSL3) animal facility in accordance with protocols approved by the NIH Animal Care and Use Committee. To measure prophylactic efficacy, mice were intraperitoneally (i.p.) injected with 1 ml of various antibody preparations or hyperimmune sheep antisera raised against baculovirus expressed HA of A/VN/1203/2004 (H5N1) that was kindly provided by Dr. G. Kemble, Medimmune Vaccines (http://www.medimmune.com). The H5N1 mAbs FLA3.14, FLA5.10, FLD20.19 and FLD21.140 were administered either as purified IgG or as enriched culture supernatant. Control human antibodies were IgG1 mAbs D2.2 or A146, specific for diphtheria toxin and anthrax protective antigen, respectively, and were prepared in the same fashion as the influenza-reactive antibodies. Twenty-four hours after i.p. administration, the mice were bled to collect samples for measurement of neutralizing human mAb titers, then challenged intranasally (i.n.) with 105 TCID50 of A/Vietnam/1203/04 (H5N1) or A/Indonesia/5/2005 (H5N1) in 50 μl. Mice were observed and weighed daily before and after viral infection. To determine viral titers following challenge, mice were killed and the lungs, brains, and spleens were aseptically removed. Tissues were homogenized in Leibovitz L-15 medium (Invitrogen, http://www.invitrogen.com) supplemented with antibiotic-antimycotic solution (Gibco, http://www.invitrogen.com) to achieve suspensions of lung (10% w/v), spleen (5% w/v), and brain (10% w/v), which were then titrated on monolayers of MDCK cells in quadruplicate. The viral titer was calculated by the Reed and Muench method and expressed as log10 TCID50 per gram of tissue. For therapy against A/Vietnam/1203/04 (H5N1), the mice were first infected i.n. with 5 LD50 of A/Vietnam/1203/04, then 24, 48, or 72 h later they were injected i.p. with 1 ml of a mAb preparation. For therapy against A/Indonesia/5/2005 (H5N1), mice were first infected i.n. with 5 LD50 of A/Indonesia/5/2005 (H5N1), then 24 h later injected i.p. with 1 ml of a mAb preparation. Pathology Mice were necropsied and the lungs were inflated with 10% neutral buffered formalin and embedded in paraffin, and sections were prepared. Slides were stained with hematoxylin and eosin. For immunohistochemical demonstration of H5 antigen, paraffin sections were prepared and ABC immunohistochemistry was performed using a goat antibody to avian influenza H5 Goat Alpha H5 (NIAID Reference Reagents, BEI Resources, http://www.beiresources.org) diluted at 1:1,000, with a Vector Rabbit Anti-Goat secondary antibody, the Vector ABC Elite label (Vector Laboratories, http://www.vectorlabs.com) diaminobenzidine as the chromogen, and hematoxylin as the counterstain. Lung pathology was evaluated in a semiquantitative manner by a pathologist (JW) blind to the treatment. Statistics Kaplan-Meier survival curves and log rank tests were used to measure differences between treatment arms in prophylactically and therapeutically treated mice. The Mann-Whitney U test was used to measure differences in viral loads in mouse tissues. For statistical purposes, samples with undetectable viral burdens were given the value 1.5 log10 TCID/g. All analyses were performed in Stata 8.2 software (StataCorp, http://www.stata.com). Results Blood samples from four Vietnamese adults (CL26, CL36, CL114, and CL115) who had recovered from HPAI H5N1 infection were collected 3–15 mo postinfection. IgG+ memory B cells recovered from frozen PBMC were immortalized with EBV. Cultures secreting neutralizing antibodies were identified by a microneutralization assay against the prototype Clade I virus, A/Vietnam/1203/04 (H5N1), and cloned by limiting dilution. Supernatants from approximately 11,000 wells were screened to identify 15 independent clones secreting a neutralizing antibody. Of these clones, three were isolated from donor CL26, one from donor CL114, and eleven from donor CL115. The number of clones isolated from each donor did not correlate with the plasma titer of neutralizing antibody, though this was not surprising given the small sample size. Clones producing antibodies that recognized H5 HA by ELISA, but did not neutralize live virus, were also identified from each donor (unpublished data). Clones FLA3.14 and FLA5.10, isolated from donor CL26, were the first obtained and were studied more extensively. Subsequently, clones FLD20.19 and FLD21.140, isolated from donor CL115, became available and were studied in parallel with FLA3.14 and FLA5.10. Clones FLA3.14, FLA5.10, FLD20.19, and FLD21.140 secreted IgG1,κ antibodies with neutralizing activity against the autologous virus A/Vietnam/CL26/2004 and other more recent Clade I viruses circulating in Viet Nam during 2005 and 2006 (Table 1). Significantly, more distant HPAI H5N1 strains, including the Clade II H5N1 virus A/Indonesia/5/2005, were neutralized by FLA3.14 and FLD20.19 (Table 2). In contrast, none of these clones neutralized an H3N2 influenza virus, A/California/7/2004 (Table 2). IgG1,κ mAbs of irrelevant specificity (diphtheria toxin and anthrax protective antigen) were used as negative controls and did not neutralize any influenza virus (Tables 1 and 2). Thus, the mAbs selected for further study demonstrated broad in vitro neutralizing activity against H5N1 viruses isolated from 1997 to 2005, albeit with some variation in potency. BALB/c mice are highly susceptible to infection with the HPAI H5N1 viruses isolated in Asia in 1997 and since 2003. Following i.n. administration, these H5N1 viruses replicate to high titer in the lungs of the mice and some isolates disseminate to extrapulmonary sites and are lethal for mice [22,23]. To explore the efficacy of mAbs FLA3.14 and FLA5.10 for pre-exposure prophylaxis, BALB/c mice were passively immunized by i.p. administration of graded doses of mAbs and then challenged i.n. with A/Vietnam/1203/04 (H5N1) 24 h later. All preparations of mAb FLA5.10 conferred 100% protection from lethality by A/Vietnam/1203/04 (p = 0.001) (Figure 1). mAb FLA3.14 also conferred some protection from lethal A/Vietnam/1203/04 (H5N1) infection, but with lower efficacy and in a dose-dependent manner (Figure 1). Mice receiving the highest dose of FLA3.14 were afforded almost complete protection (p = 0.001), whilst mice receiving the lowest dose of FLA3.14 were as susceptible as mice that received a human mAb of irrelevant specificity, though time to death was delayed (p = 0.02). Mice that received hyperimmune anti-H5 polyclonal sheep antiserum were afforded complete protection. These data, demonstrating the relatively greater in vivo activity of FLA5.10 over FLA3.14 against A/Vietnam/1203/04, are consistent with the in vitro neutralization titers presented in Table 1. To better understand the mechanism by which mAbs FLA3.14 and FLA5.10 conferred protection from lethality, the kinetics of viral infection in passively immunized mice was defined. Mice that were passively immunized with FLA3.14, FLA5.10, or a human mAb of irrelevant specificity (D2.2) were challenged with A/Vietnam/1203/04 (H5N1) 24 h later, and the level of virus replication in different organs was determined 2 and 4 d later. Mice that received the control mAb, D2.2, had high titers of virus in the lungs (Figure 2A), with evidence of extrapulmonary dissemination indicated by viral replication in the brain (Figure 2B) and spleen (Figure 2C). In contrast, mice passively immunized with FLA3.14 or FLA5.10 had significantly (10- to 100-fold) lower titers of virus in the lungs (Figure 2A) (p = 0.01, FLA3.14 versus D2.2; p = 0.001, FLA5.10 versus D2.2), undetectable viral burdens in the brain (Figure 2B) and a low titer of virus detected in the spleens of mice that received FLA5.10 (Figure 2C). Alongside the reduction in lung viral titers, mice that received prophylaxis with FLA5.10 had less dramatic pathological changes in the pulmonary airways and parenchymal tissue (Figure 3). Thus, the percentage of abnormal bronchioles with necrosis and viral antigen in lung sections from mice (n = 2 per group) that received FLA5.10 prophylaxis was less (13%) than in control mice (80%). Similarly, there were fewer inflammatory interstitial lesions in which H5 antigen was detected by immunohistochemical staining in the lung sections of mice given FLA5.10 relative to the control antibody, D2.2 (1 versus >10) (Figure 3). To a slightly lesser extent, FLA3.14 prophylaxis also limited bronchiolitis (31% versus 80%) and H5-associated interstitial pathology (2 versus >10) when compared with control mice (n = 2). These data suggest that prophylaxis with FLA3.14 or FLA5.10 probably confers protection from lethal challenge through a combination of limiting viral replication in the lung, attenuating virus-induced lung pathology, and blocking extrapulmonary dissemination of virus to distant organs. Attenuation of established infection represents a clinically relevant endpoint for antiviral therapy against H5N1 infection. To this end, the efficacy of treatment with FLA3.14, FLA5.10, FLD20.19, and FLD21.140 was determined in BALB/c mice i.n. infected 24, 48, or 72 h previously with 5 LD50 of A/Vietnam/1203/04 (H5N1). FLA3.14, FLA5.10, FLD20.19, and FLD21.140 provided robust protection from lethality in A/Vietnam/1203/04 (H5N1) infected mice at all time points, whilst an irrelevant control mAb, D2.2, gave no protection (p = 0.003) (Figure 4). These promising therapeutic results against a Clade I virus from Viet Nam led us to examine the therapeutic efficacy of mAbs FLA3.14, FLA5.10, FLD20.19, and FLD21.140 against A/Indonesia/5/2005, an antigenically divergent H5N1 virus from Clade II. The efficacy of treatment with FLA3.14, FLA5.10, FLD20.19, and FLD21.140 was determined in BALB/c mice infected i.n. 24 h previously with 5 LD50 of A/Indonesia/5/2005. Consistent with the in vitro neutralization data (Table 2), mice treated with FLA3.14 and FLD20.19, but not FLA5.10 or the control mAb D2,2, were significantly protected from A/Indonesia/5/2005 lethal infection (p = 0.003) (Figure 5). Although FLD21.140 did not demonstrate neutralizing activity in vitro against A/Indonesia/5/2005 (Table 2), treatment with this mAb significantly protected the mice (p = 0.003) from A/Indonesia/5/2005 lethal infection. This observation of neutralization in vivo suggests that a factor found in vivo enhances the neutralizing activity of this mAb, accounting for its efficacy in vivo in preventing mortality associated with infection. These data provide proof of concept that mAb therapy for at least 72 h postinfection in the mouse model can markedly improve survival from highly virulent H5N1 infection. Importantly, these data also imply it is possible to obtain significant cross-protection against a Clade II H5N1 virus using a mAb elicited by a Clade I virus. Discussion The risk of a devastating human influenza pandemic caused by an H5N1 influenza strain remains difficult to quantify. What is clear is that zoonotic infections with HPAI H5N1 viruses continue to occur in Southeast Asia with a mortality in 2006 of 67% and for which there are few specific interventions [20,24]. Here we report on the generation of four fully human mAbs with a spectrum of neutralizing activity against multiple strains of HPAI H5N1 viruses in vitro and in vivo. These mAbs could have potential in the adjunctive treatment of human pandemic or zoonotic H5N1 cases. Prophylaxis with mAb FLA5.10, and to a lesser extent FLA3.14, conferred significant immunity to mice infected with A/Vietnam/1203/04. Passive immunity was associated with significantly reduced viral burdens in the lung and negligible dissemination of virus to the brain or spleen. Viral dissemination might be an important aspect in the pathogenesis of H5N1 viruses in both mice and humans. In Vietnamese patients with H5N1 infection, fatal outcomes were strongly associated with the presence of viral genetic material, or viable virus, at extrapulmonary sites [3]. Conversely, there was no evidence of extrapulmonary virus dissemination in patients who survived [3]. In addition to evidence of viral dissemination, patients with fatal H5N1 infections had high viral loads in the respiratory tract, hypercytokinemia, multiple organ dysfunction, and acute respiratory distress syndrome [3]. Our proposed model of H5N1 pathogenesis [3] argues that early diagnosis and antiviral interventions that limit the ensuing inflammatory cascade should be central to treatment. Although not a new strategy, antibody-based therapy for severe influenza caused by H5N1 viruses represents a plausible intervention. Multiple reports of physicians using human blood products from recovering influenza patients appeared during the 1918 Spanish H1N1 influenza pandemic. A recent review of these studies suggested this treatment was associated with a halving in mortality (37% to 16%), and that early treatment was associated with greater benefits [16]. The assumption underlying these observations is that neutralizing antiviral antibodies in the plasma preparations modulated the course of viral infection and thereby prevented the development of acute respiratory distress syndrome and other complications [16]. These same tenets form the rationale for therapy using the mAbs generated in this study, with the potential of a scaleable therapeutic product free of adventitious agents. The strengths of the approach for human mAb generation described here are: (1) it uses the human immune response rather than that of animal surrogates—the antibodies selected will be those that have been generated in response to the natural infectious pathogen and have protected the individual, (2) it is fast, (3) screening can be performed using functional assays, i.e., neutralization, (4) it allows screening of a large repertoire of antibodies to select those with the most favorable profile (potency and breadth of reactivity), and (5) since the antibodies are of human origin the risks of self-reactivity against self-antigens is minimized when compared with antibodies generated in mice or through phage display. The mAbs produced in this study were derived from immortalized memory B cells of donors who had recovered from H5N1 infection. Overall, compared with the yield of neutralizing B cell clones we previously derived from patients who had recovered from SARS coronavirus infection [19], donor-derived B cell clones that neutralized H5N1 influenza were relatively scarce. These observations might reflect the weak immunogenicity of the H5 HA, as suggested in trials of inactivated H5N1 vaccines in human volunteers [25]. Two of the four mAbs characterised in this study had cross-reactive antiviral activity in vitro and in vivo against Clade I and Clade II H5N1 viruses. This is significant, as it suggests the presence of conserved neutralizing epitopes on representatives of these two clades. One mAb (FLD21.140) was effective in neutralizing a Clade II virus in vivo but not in vitro, suggesting the neutralizing activity of this mAb is dependent upon a factor found in vivo, such as complement. Similar findings have been previously reported by Gerhard and colleagues who identified a mouse mAb against an influenza A H1 HA that had neutralizing activity in vivo but not in vitro [26,27]. The in vitro neutralizing activity of this mAb was enhanced by the C1q component of the complement system plus another undefined serum factor [26,27]. It was possible to employ the human mAbs generated in this study as potent therapeutic agents for at least 72 h after A/Vietnam/1203/04 infection. This is important as most zoonotic cases of human H5N1 infection do not present to health care facilities until at least several days after illness onset [3]. Potentially, a cocktail of these cross-reactive mAbs could represent an adjunctive treatment option against H5N1 infection. The dose of mAb required for effective anti-influenza H5N1 activity in a patient is uncertain, though we note that the only mAb licensed for use against a viral agent (respiratory syncytial virus) is used at 15 mg/kg of body weight. The ongoing process of antigenic variation in antibody binding sites, called antigenic drift, in influenza viruses represents a challenge to vaccine design and also to therapy using mAbs. Although the molecular targets of the neutralizing mAbs in this study have not been determined, they are presumed to reside around the highly variable receptor binding region of HA1 [28–30] and that differences in potency are related to epitope specificity and overall avidity. To date, we have not identified mAb epitope escape mutants of H5N1. These issues, and the mechanism of virus neutralization, are subjects of ongoing virological and crystallographic studies. HPAI H5N1 viruses continue to circulate and evolve in bird populations. It is not certain that a pandemic virus originating from an HPAI H5N1 virus will resemble the H5N1 viruses studied here, or that the mAbs generated here will have neutralizing activity against a pandemic virus. Nevertheless, we are encouraged by the broad neutralizing activity of these antibodies in vitro, and the moderate doses required in vivo to confer protection. Ultimately, we hope that these mAbs, and others like them, could constitute a cocktail of cross-reactive, neutralizing antibodies that could be employed as adjunctive treatment of H5N1 influenza. Supporting Information Accession Numbers The GenBank (http://www.ncbi.nlm.nih.gov/) HA sequences of the H5N1 viruses discussed in this paper are A/Vietnam/CL115/2005 (DQ497727), A/Vietnam/CL26/2004 (DQ497723), and A/Vietnam/CL36/2005 (DQ497724).
