• Open Access

Influenza virus-like particles produced by transient expression in Nicotiana benthamiana induce a protective immune response against a lethal viral challenge in mice

Authors


* Correspondence (fax 418 6586699; e-mail: vezinalp@medicago.com)

Summary

A strain-specific vaccine represents the best possible response to the threat of an influenza pandemic. Rapid delivery of such a vaccine to the world's population before the peak of the first infection wave seems to be an unattainable goal with the current influenza vaccine manufacturing capacity. Plant-based transient expression is one of the few production systems that can meet the anticipated surge requirement. To assess the capability of plant agroinfiltration to produce an influenza vaccine, we expressed haemagglutinin (HA) from strains A/Indonesia/5/05 (H5N1) and A/New Caledonia/20/99 (H1N1) by agroinfiltration of Nicotiana benthamiana plants. Size distribution analysis of protein content in infiltrated leaves revealed that HA was predominantly assembled into high-molecular-weight structures. H5-containing structures were purified and examination by transmission electron microscopy confirmed virus-like particle (VLP) assembly. High-performance thin layer chromatography analysis of VLP lipid composition highlighted polar and neutral lipid contents comparable with those of purified plasma membranes from tobacco plants. Electron microscopy of VLP-producing cells in N. benthamiana leaves confirmed that VLPs accumulated in apoplastic indentations of the plasma membrane. Finally, immunization of mice with two doses of as little as 0.1 µg of purified influenza H5-VLPs triggered a strong immune response against the homologous virus, whereas two doses of 0.5 µg of H5-VLPs conferred complete protection against a lethal challenge with the heterologous A/Vietnam/1194/04 (H5N1) strain. These results show, for the first time, that plants are capable of producing enveloped influenza VLPs budding from the plasma membrane; such VLPs represent very promising candidates for vaccination against influenza pandemic strains.

Introduction

In the eventuality of an influenza pandemic, global health authorities will be faced with an unprecedented vaccine manufacturing bottleneck. As pointed out by the World Health Organization (WHO, 2006), responsible preparedness for an influenza pandemic must include the development of vaccines that can be scaled up rapidly and that use less antigen. Dose-sparing strategies could accelerate our capacity to provide a pandemic strain-specific vaccine to the population from the current egg- and cell culture-based manufacturing technologies, but, unless new production technologies are developed, even such low-dose vaccines would not be available to most of the world's population before the peak of the first infection wave of a pandemic outbreak. Ensuring the rapid availability of massive quantities of an appropriate antigen is currently the Achilles’ heel of pandemic preparedness plans worldwide. Many have turned to recombinant approaches to address this critical need.

Among the few promising recombinant influenza vaccines under development, virus-like particles (VLPs) – structural viral components assembled into empty shells resembling their cognate virus – may provide an ideal combination of efficacy and safety (Grgacic and Anderson, 2006). Preclinical trials with influenza VLPs have demonstrated their capacity to induce both humoral and cellular responses at low antigen doses (Bright et al., 2007). As they are composed of non-infectious recombinant proteins, the production of VLPs does not require the protective measures imposed by the manipulation of highly infectious pandemic virus particles. However, influenza VLP production currently relies on cell culture systems that may not meet the cost and speed constraints posed by global pandemic scenarios.

None of the recombinant platforms currently available compare with plant-based transient expression in terms of speed and cost. Developed in the last decade and used by a growing number of laboratories for research purposes, agroinfiltration relies on the transfer of genetic information from the bacterial vector Agrobacterium tumefaciens to plant cells after the bacterial suspension has been forced into the extracellular space of the leaf tissue. Only recently has this transient expression system been brought up to the yield and scale required for industrial application (Marillonnet et al., 2005).

The production of antigens in plants has a long track record, and the capacities of plant-based production systems have been explored for many therapeutic targets (Chichester and Yusibov, 2007). More particularly, agroinfiltration-based transient expression systems have shown unparalleled productivity and rapidity for the production of antibodies (Giritch et al., 2006) and antigens (Santi et al., 2006a), including various VLPs (Santi et al., 2006b). Viruses and their derivative VLPs are categorized as naked (non-enveloped) when formed from the assembly of viral coat proteins only, and enveloped when the external shell is composed of a lipid membrane with integral proteins. The membrane of enveloped viruses – and VLPs – may be derived from internal membranes of the host cell [mainly endoplasmic reticulum (ER) and Golgi apparatus], but most animal viruses, including the influenza A virus, bud from the plasma membrane. In plants, however, the only well-characterized enveloped viruses, i.e. members of the tospoviruses (bunya) and rhabdoviruses families, bud from internal cell membranes. The production of VLPs in plants has been limited to non-enveloped structures, with the exception of hepatitis B surface antigen (HBsAg), which forms enveloped VLPs from the ER membrane (Smith et al., 2002). The absence, to date, of examples of plasma membrane-derived viruses or VLPs produced in plants raises the question of whether plant plasma membranes, with their specific lipid composition and their apposition on the cell wall, can support the budding of influenza VLPs.

In this article, we demonstrate that influenza VLPs can form and accumulate in agroinfiltrated Nicotiana benthamiana leaves producing haemagglutinin (HA) from either A/Indonesia/5/05 (H5N1) or A/New Caledonia/20/99 (H1N1). Furthermore, we show that the parenteral administration of purified H5-VLPs with alum induces a strong and protective immune response in mice. The application of the fast and cost-efficient plant-based transient expression system for the production of VLP-based vaccines may be the perfect tool for the fight against the next influenza pandemic.

Results

Transient expression of influenza virus A/Indonesia/5/05 (H5N1) HA by agroinfiltration in N. benthamiana plants

The capacity of the transient expression system to produce pandemic influenza HA was investigated through the expression of the H5 subtype from A/Indonesia/5/05 (H5N1). As presented in Figure 1, the coding sequence of the HA gene, with its native signal peptide and transmembrane domain, was first assembled in the plastocyanin expression cassette (promoter, 5′ untransformed region (UTR), 3′ UTR and transcription termination sequences from the alfalfa plastocyanin gene), and the assembled cassette was inserted into a binary plasmid. The resulting plasmid, named 660, was then transfected into Agrobacterium (AGL1), creating the recombinant strain AGL1/660, which was used for transient HA expression.

Figure 1.

Representation of the haemagglutinin (HA) and M1 expression cassettes used. Plasto, alfalfa plastocyanin promoter and 5′ untransformed region (UTR); Pter, alfalfa plastocyanin 3′ UTR and terminator signal; SP PDI, sequence encoding the signal peptide from the alfalfa protein disulphide isomerase; 2 × 35S/TEV, chimeric double 35S promoter fused to the 5′ UTR from the tobacco etch virus; Nter, nopaline synthase terminator signal.

Influenza HA was transiently expressed by agroinfiltration of a suspension of AGL1/660 in N. benthamiana plants using a vacuum. After 6 days of incubation in the glasshouse, infiltrated plants were harvested, and leaf proteins were extracted and analysed by Western blot using anti-H5 (Vietnam) polyclonal antibodies. After separation under reducing conditions, a unique band of approximately 72 kDa was detected in protein extracts from AGL1/660-infiltrated leaves (Figure 2), corresponding in size to the uncleaved HA0 form of influenza HA. The commercial recombinant H5 used as the positive control (A/Vietnam/1203/2004) (Protein Science Corporation, Meriden, CT, USA) migrated as two bands of approximately 48 and 28 kDa, corresponding to the molecular weight of HA1 and HA2 fragments, respectively. Proteolytic cleavage of the translation product HA0 into HA1 and HA2 represents a natural and essential step for viral activation (Nayak et al., 2004). However, this processing step, which may not occur in heterologous production hosts, does not affect VLP budding, sialic acid-binding activity or immunogenicity (Pushko et al., 2005). By comparing the signal obtained on Western blots from a series of dilutions of crude protein extracts with that of known amounts of commercial H5, the leaf HA content was evaluated at approximately 50 mg/kg fresh weight (FW). The capacity of crude protein extracts from AGL1/660-transformed leaves to agglutinate turkey red blood cells (data not shown) confirmed that the recombinant H5 subunits were assembled into homotrimers.

Figure 2.

Immunodetection of H5, using anti-H5 (Vietnam) antibodies, in protein extracts (37.5 µg) from Nicotiana benthamiana leaves infiltrated with AGL1/660 (lane 3). Commercial H5 (750 ng) from influenza A/Vietnam/1203/2004 was used as a positive control of detection (lane 1), and a protein extract (37.5 µg) from leaves infiltrated with AGL1 containing an empty vector was used as a negative control (lane 2).

Assessment of VLP formation

A size distribution analysis of the HA content in protein extracts from agroinfiltrated leaves was performed after vacuum infiltration of N. benthamiana plants with the Agrobacterium strain AGL1/660 for the production of H5 from pandemic strain A/Indonesia/5/05 (H5N1), or AGL1/540 for the production of H1 from seasonal strain A/New Caledonia/20/99 (H1N1). In addition, in a third infiltration, the influenza matrix protein M1 from strain A/Puerto Rico/8/34 (AGL1/750) was co-infiltrated with AGL1/540. The inclusion of M1 co-expression was motivated by data suggesting that the co-expression of M1 is a prerequisite for the formation of influenza VLPs in insect cells (Pushko et al., 2005). Neuraminidase (NA) co-expression, which was found to be essential for the release of influenza VLPs from sialylated membrane glycoproteins in mammalian cells (Chen et al., 2007), was not considered for VLP production in the plant system, as sialic acid is absent from plant glycoproteins (Séveno et al., 2004). Total soluble protein extracts were separated by size exclusion chromatography on a calibrated S-500 high-resolution (HR) column (GE Healthcare Bio-Science Corporation, Piscataway, NJ, USA), and elution fractions were monitored for their total protein content and for HA abundance using immunodetection with anti-HA antibodies. In all extracts analysed, the protein content peaked at fraction 18, where most of the host soluble proteins eluted (Figure 3, graph). In contrast, the anti-HA immune signal peaked at fractions 9 and 10 for both H1 and H5, indicating the presence of HA in high-molecular-weight structures in the region of 2 MDa (Figure 3, H5 and H1). Co-expression of M1 with H1 did not modify the elution profile of H1, but reduced the overall accumulation of the latter in the leaf tissue, as shown by immunodetection (Figure 3, H1 + M1).

Figure 3.

Characterization of haemagglutinin (HA)-containing structures by size exclusion chromatography. Protein extracts from leaves producing H5, H1 and H1 + M1 were separated by gel filtration on a calibrated S-500 high-resolution column, and the total soluble protein content of each fraction was evaluated using the Bradford method (graph). The HA content of elution fractions was revealed by immunodetection using anti-HA antibodies.

With the aim of developing a vaccine for pandemic influenza, we further characterized H5-containing high-molecular-weight structures by electron microscopy observation of purified preparations. A purification method was first developed to isolate the H5-containing structures from host proteins. Crude protein extracts, obtained by homogenization of frozen leaf tissue from AGL1/660-infiltrated plants, were clarified by adjustment to pH 6 and heating at 42 °C. After removal of the insoluble fraction by filtration and centrifugation, HA was purified from the clarified extract by affinity chromatography on fetuin-agarose, yielding over 80% pure H5 based on Coomassie staining and densitometry analysis. Transmission electron microscopy of the purified product indicated that the isolated high-molecular-weight structures containing H5 corresponded to VLPs in both size and morphology, comprising a phospholipid membrane covered with spikes, closely resembling those of the influenza virions (Figure 4a,b).

Figure 4.

(a) Negative staining transmission electron microscopy examination of purified H5 VLP. The bar represents 100 nm. (b) Isolated H5 VLPs enlarged to show details of the structure. The bar represents 30 nm.

Lipid composition and subcellular localization of VLPs

As influenza particles bud from the plasma membrane of their host cells, and no plasma membrane-derived viruses infecting plants have been described to date, the precise origin of these influenza VLPs in leaf cells were investigated. To obtain a first indication of their origin, lipids were extracted from purified VLPs and their composition was compared with that of highly purified tobacco plasma membranes by high-performance thin layer chromatography (HP-TLC). The migration patterns of polar and neutral lipids from VLPs and control plasma membranes were comparable. Purified VLPs contained the major phospholipids (phosphatidylcholine and phosphatidylethanolamine) and sphingolipids (glucosyl-ceramide) found in the plasma membrane (Figure 5a), and both contained free sterols as the sole neutral lipids (Figure 5b). However, the faint signal obtained from immunodetection of a plasma membrane protein marker (ATPase) in purified VLP preparations showed that the VLP preparation contained very little of one of the major proteins associated with plant plasma membranes (Figure 5c). Taken together, these results suggest that the influenza VLPs bud from the plasma membrane, but also that host proteins may have been excluded from the membranes during the process of VLP budding from the plant cells, as described for the budding of influenza virus particles (Nayak et al., 2004).

Figure 5.

(a) Polar lipid composition of purified influenza virus-like particles (VLPs). Lipids contained in an equivalent of 40 µg of proteins were extracted from VLPs, separated by high-performance thin layer chromatography (HP-TLC) and compared with the migration profile of lipids isolated from highly purified tobacco plasma membrane (PM). Lipid abbreviations are as follows: DGDG, digalactosyldiacylglycerol; gluCER, glucosyl-ceramide; PA, phosphatic acid; PC, phosphatidylcholine; PE, phosphatidylethanolamine; PG, phosphatidylglycerol; PI, phosphatidylinositol; PS, phosphatidylserine; SG, steryl-glycoside. (b) Neutral lipid composition of purified influenza VLPs. Lipids contained in an equivalent of 20 µg of proteins were extracted from VLPs, separated by HP-TLC and compared with the migration of sitosterol. (c) Immunodetection of the PM marker proton pump ATPase (PMA) in purified VLPs and highly purified PMs from tobacco leaves (PML) and BY2 tobacco cells (PMBY2). Eighteen micrograms of protein were loaded in each lane.

Electron microscopy was then performed on infected leaf cross-sections to gather further evidence on the mechanisms involved in VLP formation at the cellular level. Thin leaf sections of infiltrated plants were fixed in glutaraldehyde and examined by transmission electron microscopy after positive staining. This approach showed that all of the VLPs accumulated in indentations formed by the invagination of the plasma membrane (Figure 6). These observations, together with the lipid composition analysis, indicated that the budding of VLPs occurs at the plasma membrane, in spite of the potential constraints imposed by the specific lipid content of the plant plasma membrane and its apposition on the cell wall caused by high cell turgidity.

Figure 6.

Localization of virus-like particle (VLP) accumulation by positive staining transmission electron microscopy observation of H5-producing tissue. Indentation containing the VLPs is enlarged (lower left) to show details. cw, cell wall; pm, plasma membrane. The bar represents 500 nm.

Immunogenicity and efficacy of plant-derived VLPs

The immunogenicity of plant-derived influenza VLPs was investigated by a dose-ranging study in mice. Groups of five BALB/c mice were immunized intramuscularly, twice at 3-week intervals, with 0.1–12 µg of VLPs containing HA from influenza A/Indonesia/5/05 (H5N1) formulated in alum (1 : 1 ratio). Haemagglutination inhibition (HI) titres, using whole inactivated virus antigen [A/Indonesia/5/05 (H5N1)], were measured on sera collected 14 days after the second immunization. Immunization with doses of VLPs as low as 0.1 µg induced the production of antibodies that inhibited viruses from agglutinating erythrocytes at high dilutions (Figure 7a). Parallel immunization of mice with 5 µg of non-VLP alum-adjuvant control H5 antigen (also from A/Indonesia/5/05) induced an HI response that was fourfold to sixfold lower than that achieved with the lowest VLP dose.

Figure 7.

(a) Haemagglutination inhibition (HI) antibody response 14 days after boost in Balb/c mice immunized intramuscularly with influenza H5 virus-like particle (VLP) or non-VLP recombinant H5 formulated in alum (1 : 1 ratio). Values represent the geometric mean titres (GMTs) (log2) calculated from reciprocal end-point titres of five mice per group. Bars indicate mean deviations. Experimental groups were compared using Student's t-test for paired observations. P < 0.05 was considered to be statistically significant. *P < 0.05 and **P < 0.01 compared with the group immunized with the non-VLP H5 antigen. HA, haemagglutinin; PBS, phosphate-buffered saline. (b) Protection from lethal challenge with A/Vietnam/1194/04 isolate. Mice were challenged intranasally with one LD50 of A/Vietnam/1194/04 isolate, 75 days after immunization with two doses of Indonesia-derived VLPs. Weight and survival were monitored over a 14-day period. Mice showing a body weight reduction greater than 25% were euthanized.

As the homologous strain (A/Indonesia/5/05) was not available, the A/Vietnam/1194/04 strain was used in the challenge study. The fact that the vaccine and challenge strains were genetically dissimilar (clades 1 and 2, respectively) raised the risk of diminished protection. As a result, the mice were exposed to only a modest challenge dose predicted to kill 50% of the animals (LD50). Mice (eight per group) were immunized with 0, 0.5, 2.5 or 7.5 µg of VLP prior to being challenged intranasally with a dose of the A/Vietnam/1194/04 influenza virus corresponding to one LD50 in order to confirm the capacity of plant-produced VLPs to confer protection against a pandemic influenza A strain. As shown in Figure 7b, this challenge dose was sufficient to kill 75% of the non-immunized mice in the 14 days after challenge. In sharp contrast, 100% protection was observed in the mice immunized with influenza VLPs, even at doses as low as 0.5 µg of H5.

Discussion

The results presented here show, for the first time, that plants are capable of producing influenza VLPs, and that immunization with these VLPs induces a protective immune response against a lethal challenge in mice at doses as low as 0.5 µg of HA. On expression of influenza H5 or H1 in plant leaves, HA proteins assembled into high-molecular-weight structures that exhibited a size exclusion chromatography elution profile similar to that of influenza VLPs produced in insect cells from the co-expression of HA and M1 (Pushko et al., 2005). Previous efforts at producing influenza HA in plants or plant cells have been reported (Cardineau et al., 2004; Miller et al., 2004; Chandler, 2007). However, in none of these studies was the assembly of VLPs established.

In contrast with hepatitis B VLPs, which form from the expression of a single envelope protein (HBsAg), influenza A VLPs produced to date have only been obtained from the co-expression of at least HA and M1 (Galarza et al., 2005; Quan et al., 2007) or HA and NA (Chen et al., 2007). We have shown here that influenza VLPs can be produced from the expression of HA alone. Previous work, published by Chen et al. (2007), established that the M1 protein was dispensable for VLP formation, and that the role of NA was not structural. In support of this assumption, they demonstrated that influenza VLPs were obtained from mammalian cells (MDCK) transformed to produce HA alone, provided that the culture medium was supplemented with an exogenous source of bacterial NA. The authors concluded that NA (or sialidase) activity was necessary for the release of newly formed VLPs from the producing cells by removing sialic acids on host glycoproteins. In the present study, the co-expression of HA and M1 in plants led to a substantial decrease in HA accumulation. The cause of this decrease in VLP accumulation from M1 co-expression remains unknown. In influenza-infected cells, HA is transported towards the plasma membrane through the secretory pathway, whereas M1, a cytosolic protein, collects beneath the plasma membrane. The processes leading to the synchronized accumulation of these proteins at the site of budding are complex and regulated. It is certainly possible that the expression strategies used in the current study did not result in the coordinated accumulation of H1 and M1 at the site of budding in the plant system.

Influenza A is an enveloped virus budding from the plasma membrane of infected cells (Nayak et al., 2004). To our knowledge, there is no plant virus known to bud from the plasma membrane, and there are no reports of VLPs budding from the plasma membrane in plants. Among the examples of plant-produced VLPs, only hepatitis B VLPs assembled from the expression of the surface antigen (HBsAg) include a phospholipid bilayer envelope. However, this envelope is formed from internal cell membranes (Smith et al., 2002). Likewise, the only two families of enveloped plant viruses known to date, namely the tospoviruses and rhabdoviruses, bud from internal cell membranes. We report here the successful production of influenza VLPs in N. benthamiana, showing, for the first time, that plasma membrane-derived VLPs can be produced in plants. Most, if not all, plant viruses have evolved mechanisms for cell-to-cell movement which take advantage of the plasmodesmatal interconnections, rather than budding from the plasma membrane as do influenza and other animal viruses. Our transmission electron microscopy data suggest that the budding of influenza VLPs occurred, but that VLPs remained trapped in indentations of the plasma membrane because of the low porosity of the cell wall. From these results, it can be speculated that budding from the plasma membrane represents a dead-end for the viral life cycle, as viral particles accumulating in between the infected cell and the cell wall have no access to adjacent uninfected cells.

Two recent studies have demonstrated that HA fragments from an annual influenza strain [A/Wyoming/03/03 (H3N2)] can be produced by agroinfiltration of N. benthamiana using a viral-based expression vector. A mouse immunization study established the potential of this plant-produced soluble HA fragment to induce a strong immune response by stimulating both humoral and cellular responses when formulated with Quil A, an adjuvant known to induce balanced immune responses (Shoji et al., 2008). In the second study, Mett et al. (2008) demonstrated that immunization with doses of 200 µg of HA fragments could induce a protective immune response against a non-lethal challenge in ferrets. In our study, plant-produced influenza VLPs elicited a strong influenza-specific immune response in mice at doses as low as 0.1 µg of HA. Similar results have been presented previously from immunization with influenza VLPs produced in insect cells. For example, Pushko et al. (2005) have shown that mice immunized with influenza VLPs produced in insect cells from the co-expression of HA, NA and M1 from strain A/Hong Kong/1073/99 (H9N2) induce an influenza-specific antibody response capable of inhibiting the replication of the homologous virus. Similar results have been obtained by Galarza et al. (2005) with influenza VLPs containing HA, NA, M1 and M2 from influenza A/Udorn/72 (H3N2). More recently, Bright et al. (2008) have demonstrated that influenza VLPs, comprising a similar combination of viral proteins (HA, NA, M1) from the same H5N1 strain as that used in our study (A/Indonesia/5/05), elicit a broadened immune response compared with a non-VLP recombinant HA at low doses (0.6 µg), and that vaccination with 3 µg doses of VLPs protects mice against a lethal challenge with homologous (A/Indonesia/5/05) and heterologous (A/Vietnam/1203/04) viruses. In the present study, we have shown that doses as low as 0.5 µg of plant-produced VLPs comprising Indonesia HA alone protect mice against a lethal challenge with the Vietnam strain. This result suggests that VLPs comprising HA alone can elicit cross-clade protection against potential influenza pandemic strains. From a manufacturing perspective, a VLP-based vaccine based on only one viral antigen would ensure a simpler and more flexible production process, particularly in the context in which the specific pandemic strain remains unknown. The simplicity of this candidate vaccine would ensure that the development and optimization steps would be minimized after the identification of the targeted virus strain.

This work is a first step towards the demonstration that VLP production using plant-based agroinfiltration technology could be advantageous in the development of influenza A annual and pandemic vaccines. Plant-produced influenza VLPs are highly immunogenic, and a surge capacity analysis of the current transient expression technology has shown that the first batches of a candidate vaccine could be produced only 18 days after the sequence of the target HA is made available. At the current yields and a theoretical human dose of 30 µg of antigen, each kilogram of infiltrated leaves contains approximately 1500 vaccine doses. In the light of recent demonstrations of the dose-sparing effect of various adjuvants (Leroux-Roels et al., 2007; Bernstein et al., 2008), the number of possible doses per kilogram of infected leaves could plausibly be much higher. As each square metre of glasshouse can produce 2 kg of N. benthamiana leaf biomass in 6 weeks, at least 30 million doses could be produced over a 3-month period in a single 5000 m2 glasshouse facility. This unique combination of platform simplicity, surge capacity and powerful immunogenicity could make this approach a powerful new tool for a fast and effective response in the context of a pandemic.

Experimental procedures

Assembly of expression cassettes

All manipulations were performed using the general molecular biology protocols of Sambrook and Russell (2001). The first cloning step consisted of the assembly of a receptor plasmid containing upstream and downstream regulatory elements of the alfalfa plastocyanin gene. The plastocyanin promoter and 5′ UTR sequences were amplified from alfalfa genomic DNA using the oligonucleotide primers XmaI-pPlas.c (5′-AGTTCCCCGGGCTGGTATATTTATATGTTGTC-3′) and SacI-ATG-pPlas.r (5′-AATAGAGCTCCATTTTCTCTCAAGATGAT TAATTAATTAATTAGTC-3′). The resulting amplification product was digested with XmaI and SacI and ligated into pCAMBIA2300 (Cambia, Canberra, Australia), previously digested with the same enzymes, to create pCAMBIApromoPlasto. Similarly, the 3′ UTR sequences and terminator of the plastocyanin gene were amplified from alfalfa genomic DNA using the primers SacI-PlasTer.c (5′-AATAGAGCTCG TTAAAATGCTTCTTCGTCTCCTATTTATAATATGG-3′) and EcoRI-PlasTer.r (5′-TTACGAATTCTCCTTCCTAATTGGTGTACTATCATTTAT CAAAGGGGA-3′), and the product was digested with SacI and EcoRI before being inserted into the same sites of pCAMBIApromoPlasto to create pCAMBIAPlasto.

The open reading frame from the HA gene of influenza virus A/Indonesia/5/05 (H5N1) (Acc. No. EF541394) was synthesized (Epoch Biolabs, Sugar Land, TX, USA). The fragment produced contained the complete H5 coding region, including the native signal peptide flanked by a HindIII site immediately upstream of the initial ATG, and a SacI site immediately downstream of the stop (TAA) codon. The H5 coding region was cloned into a plastocyanin-based expression cassette by the polymerase chain reaction (PCR)-based ligation method presented in Darveau et al. (1995). Briefly, the first PCR amplification was obtained using primers Plato-443c (5′-GTATT AGTAATTAGAATTTGGTGTC-3′) and SpHA(Ind)-Plasto.r (5′-GCA AGAAGAAGCACTATTTTCTCCATTTTCTCTCAAGATGATTA-3′), using pCAMBIApromoPlasto as template. A second amplification was performed with primers Plasto-SpHA(Ind).c (5′-TTAATCATCTTGA GAGAAAATGGAGAAAATAGTGCTTCTTCTTGC-3′) and HA(Ind)-Sac.r (5′-ACTTTGAGCTCTTAAATGCAAATTCTGCATTGTAACGA-3′), with the synthesized H5 fragment as template. The amplification products obtained from both reactions were mixed together, and the mixture served as the template for a third reaction (assembling reaction) using Plato-443c and HA(Ind)-Sac.r as primers. The resulting fragment was digested with BamHI (in the plastocyanin promoter) and SacI (at the 3′ end of the fragment), and cloned into a binary plasmid containing the plastocyanin promoter previously digested with the same enzymes.

The open reading frame from the H1 gene of influenza strain A/New Caledonia/20/99 (H1N1) was synthesized in two fragments (Plant Biotechnology Institute, National Research Council, Saskatoon, Canada). A first fragment, corresponding to the wild-type H1 coding sequence (Acc. No. AY289929) lacking the signal peptide coding sequence at the 5′ end and the transmembrane domain coding sequence at the 3′ end, was synthesized. Restriction sites were removed or added for cloning purposes. A BglII site was added at the 5′ end of the coding sequence, and a dual SacI/StuI site was added immediately downstream of the stop codon at the 3′ terminal end of the fragment. A second fragment encoding the C-ter end of the protein (transmembrane domain and cytoplasmic tail) from the KpnI site to the stop codon, and flanked in 3′ by SacI and StuI restriction sites, was also synthesized.

The first H1 fragment was digested with BglII and SacI and cloned into pCAMBIAPlasto containing the coding sequence of the signal peptide of the alfalfa protein disulphide isomerase (PDI) gene (nucleotides 32–103) (Acc. No. Z11499), resulting in a PDI-H1 chimeric gene downstream of the plastocyanin regulatory elements. The addition of the C-ter end coding region (encoding the transmembrane domain and the cytoplasmic tail) was obtained by inserting the second synthesized fragment, previously digested with KpnI and SacI, into the PDI-H1 fusion expression plasmid.

A fusion between the tobacco etch virus (TEV) 5′ UTR and the open reading frame of the influenza A/PR/8/34 M1 gene (Acc. No. NC_002016) was synthesized with a flanking SacI site added downstream of the stop codon. The fragment was digested with SwaI (in the TEV 5′ UTR) and SacI, and cloned into a 2 × 35S/TEV-based expression cassette in a pCAMBIA binary plasmid. The resulting plasmid bore the M1 coding region under the control of a 2 × 35S/TEV promoter and 5′ UTR and the nopaline synthase (NOS) terminator.

An HcPro construct (35SHcPro) was prepared as described by Hamilton et al. (2002). All clones were sequenced to confirm the integrity of the constructs. The plasmids were used to transform A. tumefaciens (AGL1) [American Type Culture Collection (ATCC), Manassas, VA, USA] according to Höfgen and Willmitzer (1988). The integrity of all A. tumefaciens strains was confirmed by restriction mapping.

Preparation of plant biomass, inocula and agroinfiltration

Nicotiana benthamiana plants were grown from seeds in flats filled with a commercial peat moss substrate. The plants were allowed to grow in the glasshouse under a 16-h/8-h photoperiod and a temperature regime of 25 °C day/20 °C night. Three weeks after seeding, individual plantlets were picked out, transplanted in pots and left to grow in the glasshouse for three additional weeks under the same environmental conditions.

Agrobacteria were grown in a yeast extract broth (YEB) medium, supplemented with 10 mm 2-(N-morpholino)ethanesulphonic acid (MES, pH 5.6), 20 µm acetosyringone, 50 µg/mL kanamycin and 25 µg/mL carbenicillin, until they reached an optical density at 600 nm (OD600) between 0.6 and 1.6. Agrobacterium suspensions were centrifuged before use and resuspended in infiltration medium (10 mm MgCl2 and 10 mm MES, pH 5.6). Agrobacterium tumefaciens suspensions were centrifuged, resuspended in infiltration medium and stored overnight at 4 °C. On the day of infiltration, culture batches were resuspended in 2.5 culture volumes and allowed to warm before use. Whole plants of N. benthamiana were placed upside down in the bacterial suspension in an airtight stainless steel tank under a vacuum of 20–40 Torr for 2 min. Following vacuum infiltration, the plants were returned to the glasshouse for a 6-day incubation period until harvest.

Leaf sampling and total protein extraction

Following incubation, the aerial part of the plants was harvested, frozen at –80 °C and crushed into pieces. Total soluble proteins were extracted by homogenizing (Polytron) frozen-crushed plant material in three volumes of cold 50 mm tris(hydroxymethyl)aminomethane (Tris), pH 7.4, 0.15 m NaCl and 1 mm phenylmethylsulphonyl fluoride (PMSF). After homogenization, the slurries were centrifuged at 20 000 g for 20 min at 4 °C, and the crude extracts (supernatant) were kept for analysis.

Protein analysis, immunoblotting and haemagglutination assay

Proteins were separated by sodium dodecylsulphate-polyacrylamide gel electrophoresis (SDS-PAGE) under reducing conditions and electrotransferred on to polyvinylidene difluoride (PVDF) membranes (Roche Diagnostics Corporation, Indianapolis, IN, USA) for immunodetection. Prior to immunoblotting, the membranes were blocked with 5% skimmed milk and 0.1% Tween-20 in Tris-buffered saline (TBS-T) for 16–18 h at 4 °C. Anti-H1 immunoblotting was performed by incubation with mouse anti-influenza A antibodies (Fitzgerald, Concord, MA, USA; Cat# 10-I50 as primary antibody). Anti-H5 immunoblotting was performed by incubation with rabbit anti-H5 (A/Vietnam/1203/2004) antibodies (Immune Technology, New York, NY, USA; Cat# IT-003-005V). Anti-plasma membrane marker proton pump ATPase (anti-PMA) immunoblotting was performed by incubation with rabbit anti-PMA polyclonal antibodies (kindly provided by Benoit Lefebvre and Marc Boutry, Université Catholique de Louvain, Belgium) as primary antibodies. Incubation with primary antibodies was followed by a second incubation with appropriate peroxidase-conjugated secondary antibodies. Immunoreactive complexes were detected by chemiluminescence using luminol as the substrate (Roche Diagnostics Corporation).

Haemagglutination assay for H5 was based on a method described by Nayak and Reichl (2004). Serial double dilutions of the test samples (100 µL) were made in ‘V’-bottomed 96-well microtitre plates containing 100 µL phosphate-buffered saline (PBS), leaving 100 µL of diluted sample per well. One hundred microlitres of a 0.25% turkey red blood cell suspension (Bio Link Inc., Syracuse, NY, USA) were added to each well, and the plates were incubated for 2 h at room temperature. The reciprocal of the highest dilution showing complete haemagglutination was recorded as HA activity. In parallel, a recombinant HA standard [A/Vietnam/1203/2004 (H5N1)] (Protein Science Corporation) was diluted in PBS and run as a control on each plate.

Size exclusion chromatography

Columns of 32 mL of Sephacryl S-500 HR beads (GE Healthcare Bio-Science Corporation) were packed and equilibrated with 50 mm PBS, pH 7.4, 150 mm NaCl. One and a half millilitres of crude protein extract were loaded and eluted with equilibration buffer. Twenty-four elution fractions of 1.5 mL were collected and analysed for relative protein content using the Bradford reagent (Bio-Rad, Hercules, CA, USA). Total proteins from each fraction were concentrated by acetone precipitation and resuspended in one-fifth volume of elution buffer prior to immunological analysis. Each separation was followed by a calibration of the column with Blue Dextran 2000 (GE Healthcare Bio-Science Corporation). Elution profiles of Blue Dextran 2000 and host-soluble proteins were compared between each separation to ensure uniformity of the elution profiles between the columns used.

Electron microscopy

One hundred microlitres of the samples to be examined were placed in an Airfuge ultracentrifugation tube (Beckman Instruments, Palo Alto, CA, USA). A grid was placed at the bottom of the tube, which was then centrifuged for 5 min at 120 000 g. The grid was removed, gently dried and placed on a drop of 3% phosphotungstic acid at pH 6 for staining (negative). Samples for thin layer microscopy were prepared from H5-producing leaves as follows. Leaf blocks of approximately 1 mm3 were fixed in PBS containing 2.5% glutaraldehyde and washed in PBS containing 3% sucrose, before a post-fixation step in 1.33% osmium tetroxide. Fixed samples were embedded in Spurr resin and ultrathin layers were laid on a grid. Samples were positively stained with 5% uranyl acetate and 0.2% lead citrate before observation. Grids were examined on a Hitachi 7100 transmission electron microscope.

VLP purification

Frozen H5-producing leaves of N. benthamiana were homogenized in 1.5 volumes of 50 mm Tris, pH 8, NaCl (50 mm) and 0.04% sodium metabisulfite using a commercial blender. The resulting extract was supplemented with 1 mm PMSF and adjusted to pH 6 with 1 m acetic acid before being heated at 42 °C for 5 min. Diatomaceous earth (DE) was added to the heat-treated extract to adsorb the contaminants precipitated by the pH shift and heat treatment, and the slurry was filtered through Whatman filter paper (Whatman #1). The resulting clarified extract was centrifuged at 10 000 g for 10 min at room temperature to remove residual DE, passed through 0.8/0.2 µm filters and loaded on to a fetuin-agarose affinity column (Sigma-Aldrich, St Louis, MO, USA). Following a wash step in 400 mm NaCl, 25 mm Tris, pH 6, bound proteins were eluted with 1.5 m NaCl, 50 mm MES, pH 6. Eluted VLPs were supplemented with Tween-80 to a final concentration of 0.0005% (v/v), and concentrated on a 100-kDa MWCO ultrafiltration device. The resulting concentrate was centrifuged at 10 000 g for 10 min at 4 °C, and pelleted VLPs were resuspended in 100 mm PBS, pH 7.4, 0.01% Tween-80, 0.01% thimerosal. Finally, resuspended VLPs were filter-sterilized before use.

Preparation of purified plasma membrane

Plasma membranes were obtained from tobacco leaves and cultured BY2 cells after cell fractionation, according to Mongrand et al. (2004), by partitioning in an aqueous polymer two-phase system with polyethylene glycol 3350/dextran T-500 (6.6% each). All steps were performed at 4 °C.

Lipid analysis

Lipids were extracted and purified from the different fractions according to Bligh and Dyer (1959). Polar and neutral lipids were separated by monodimensional HP-TLC using the solvent systems described in Lefebvre et al. (2007). Lipids of plasma membrane fractions were detected after staining with copper acetate, as described by Macala et al. (1983). Lipids were identified by comparison of their migration times with those of standards (all standards were obtained from Sigma-Aldrich, except for steryl-glycoside which was obtained from Matreya, Pleasant Gap, PA, USA).

Immunization and HI titre measurements

Vaccine studies were performed with 6–8-week-old female BALB/c mice (Charles River Breeding Laboratories, Montreal, QC, Canada). Thirty mice were randomly divided into six groups of five animals. Mice were immunized in a two-dose regimen, the second immunization being made 3 weeks following the first. Unanaesthetized mice were immunized by intramuscular administration in the hind legs with purified H5 VLPs (0.1, 1, 5 or 12 µg) formulated in Alhydrogel 2% (alum, Accurate Chemical & Scientific Corporation, Westbury, NY, USA) in a 1 : 1 ratio. A fifth group of mice was immunized with 5 µg of recombinant HA formulated in alum (control H5 antigen) obtained from Immune Technology [H5 from A/Indonesia/5/05 (H5N1) and purified from 293 cell culture]. Finally, PBS (formulated in alum) was administered to the sixth group as a negative control. Lateral saphenous vein blood collection was performed 14 days after second immunization on unanaesthetized animals. Serum was collected by centrifuging at 8000 g for 10 min. HI titres of sera were measured 14 days after the second immunization, as described previously (Kendal et al., 1982; WHO, 2002). Inactivated virus preparations (strain A/Indonesia/5/05; Food and Drug Administration/Center for Biologics Evaluation and Research, Rockville, MD, USA) were used to test mouse serum samples for HI activity. Sera were pretreated with receptor-destroying enzyme II (RDE II) (Accurate Chemical and Scientific Corporation) prepared from Vibrio cholerae. HI assays were performed with 0.5% turkey red blood cells. HI antibody titres were defined as the reciprocal of the highest dilution causing complete inhibition of agglutination.

Lethal challenge in mice

This study was performed with 6–8-week-old female BALB/c mice (Charles River Breeding Laboratories, France) in a BSL-4 laboratory (P4 Jean Mérieux INSERM Laboratory, Lyon, France). Forty animals were randomly divided into five groups of eight mice. Mice were immunized with influenza VLP doses corresponding to 0.5, 2.5 or 7.5 µg of HA, and PBS was administered to the fourth (control) group. All antigens were formulated with alum at a final concentration of 1% (v/v), except for the negative control (PBS). Mice were immunized intramuscularly on days 0 and 14. Mice were challenged intranasally with one LD50 of live A/Vietnam/1194/04 influenza virus (H5N1), 75 days after the second immunization. Animals were monitored daily after challenge for behavioural changes, and the body weight was measured every 2 days or daily if animals showed any sign of disease. Mice showing a 25% decrease in body weight were euthanized according to standard procedures (P4 Laboratory in Lyon). Survival was measured over a period of 14 days.

Acknowledgements

The authors are grateful to Robert Alain (Institut Armand-Frappier, Laval, QC, Canada) for the preparation of samples and their examination by transmission electron microscopy, and to Benoit Lefebvre and Marc Boutry for providing anti-PMA antibodies. They also thank Dr Hervé Raoul (P4 Jean Mérieux INSERM Laboratory, Lyon, France) and Dr Eric Quéméneur (Commissariat à l’Energie Atomique, Marcoule, France) for conducting the lethal challenge study. S.M. was supported by the French Agence Nationale de la Recherche (contract ANR-JC05-45555 ‘Plant rafts’).

Ancillary