• Open Access

The production of hemagglutinin-based virus-like particles in plants: a rapid, efficient and safe response to pandemic influenza


(fax 418 658 6699; e-mail daoustma@medicago.com)


During the last decade, the spectre of an influenza pandemic of avian origin has led to a revision of national and global pandemic preparedness plans and has stressed the need for more efficient influenza vaccines and manufacturing practices. The 2009 A/H1N1 (swine flu) outbreak has further emphasized the necessity to develop new solutions for pandemic influenza vaccines. Influenza virus-like particles (VLPs)—non-infectious particles resembling the influenza virus—represent a promising alternative to inactivated and split-influenza virions as antigens, and they have shown uniqueness by inducing a potent immune response through both humoral and cellular components of the immune system. Our group has developed a plant-based transient influenza VLP manufacturing platform capable of producing influenza VLPs with unprecedented speed. Influenza VLP expression and purification technologies were brought to large-scale production of GMP-grade material, and pre-clinical studies have demonstrated that low doses of purified, plant-produced influenza VLPs induce a strong and broad immune response in mice and ferrets. This review positions the recent developments towards the successful production of influenza VLPs in plants, including the production of VLPs from other human viruses and other forms of influenza antigens. The platform developed for large-scale production of VLPs is also presented along with an assessment of the speed of the platform to produce the first experimental vaccine lots from the identification of a new influenza strain.

Lessons learned from the novel A/H1N1 outbreak

In a world that has been preparing for a potential pandemic spread of A/H5N1 avian influenza over the past decade, the recent A/H1N1 outbreak has challenged national and global pandemic plans in many ways, including the very definition of influenza pandemic. The sudden outbreak of this novel flu virus provided a real-life test that has highlighted the strengths and weaknesses of the global healthcare options in response to such a threat. The lessons learned will be key to a successful response to subsequent waves of A/H1N1 and to an eventual outbreak of A/H5N1 or an outbreak of any novel flu strain.

Well before the onset of the A/H1N1 outbreak, in September 2008, the Congressional Budget Office (CBO) of the United States had already acknowledged the incapacity of the manufacturers of commercial seasonal influenza vaccines in the United States to produce vaccines of sufficient effectiveness, in sufficient quantities or in the time required to meet public health needs in the event of an influenza pandemic (CBO, 2008). This spring, aided by worldwide travelling, the A/H1N1 influenza pandemic has spread internationally with unprecedented speed. According to WHO, ‘in past pandemics, influenza viruses have needed more than 6 months to spread as widely as the new H1N1 virus has spread in less than 6 weeks’ (WHO 2009a). This stressed the importance of timely access to antivirals and vaccines. It has particularly highlighted the limitations of current egg-based vaccines in terms of manufacturing speed. While the pandemic strain was identified on April 24th, the vaccine produced and distributed in Canada only obtained approval from Health Canada on October 21st. Thus, vaccination has only started during the second wave of infection rather than prior to it, which would have been the ideal situation.

Influenza pandemics, caused by viruses that have undergone important genetic changes and to which populations have not been exposed, have struck the world three times in the last century and been characterized by high morbidity and mortality rates (Kilbourne, 2006). Unfortunately, in contrast to annual epidemics, the timing of emergence of a pandemic, the identity of the causal strain and the severity of the damages are difficult to predict, imposing a set of important constraints on the production of pandemic influenza vaccines. An ideal pandemic influenza vaccine needs to be produced as fast as possible to halt the spread of the novel influenza strain. This vaccine needs to be efficacious, to provide protection with low antigen content and, ideally, after only one dose. Finally, despite being produced in the shortest achievable timeframe, the ideal pandemic influenza vaccine should assure safe vaccination of populations.

Current seasonal and pandemic influenza vaccine practices

The global seasonal influenza vaccine supply relies on a well ran industry. The vast majority of influenza vaccines currently on the market are composed of viral particles or antigens obtained from egg-grown virions. The production of egg-derived vaccines thus relies on the culture of live viruses in embryonated hens eggs. This manufacturing practice requires multiple steps and involves the participation of WHO-collaborating centres for virus identification and preparation of the vaccine virus strain (reassortant) that is destined for distribution to manufacturers. Only then the manufacturers can initiate optimization of viral growth through multiple passages in eggs and manufacture the vaccine with the optimized strain. Split-influenza vaccines are obtained after chemical inactivation and disruption of purified virions with a detergent. The resulting antigen preparation consists of protein and small protein clusters mainly containing the hemagglutinin (HA) antigen, and the full process from the identification of the new strain up to the filling and release of the vaccine product is completed within 4–6 months (WHO, 2009b).

Although such a production calendar has a proven track record and has become a familiar scenario for seasonal vaccine production, the response-time of the egg-based vaccine production system represents an important drawback when rapid response to the spread of a novel strain is needed. As a consequence, pre-pandemic influenza vaccines have been produced against influenza strains showing the highest potential of causing a pandemic. Several egg-produced pre-pandemic influenza vaccines have received approval for distribution from health authorities around the world. These vaccines are composed of inactivated split virions produced through the same development path as seasonal vaccines and are destined to public health national vaccine stockpiles. This strategy was selected by a number of countries in the recent years in preparation for an outbreak of A/H5N1 of avian origin. Despite all the efforts to predict the nature of the next pandemic strain, correct identification cannot be assured in advance, and although such vaccines may provide protection in the case of an outbreak of a strain with high homology to the vaccine strain, their utility is substantially reduced in the case of an outbreak of a different subtype of pathogenic influenza. The recent influenza A/H1N1 outbreak, against which the available seasonal and pre-pandemic vaccines did not offer protection, did emphasize the need for a technology that can produce a strain-specific pandemic vaccine, that is, a vaccine constituted of antigens of the emerging new strain, in a timely manner. Production efficacy is also a concern: pandemic strains might not grow as well in eggs as seasonal vaccine strains, as was exemplified by the current A/H1N1 reassortant vaccine strain that yielded only one-third to half of the productivity of seasonal strains (WHO, 2009c). Shortage of egg-supply and limited manufacturing capacity for egg-based production may render vaccine manufacturing inoperative at the time of a pandemic if it occurs during seasonal vaccine production or if flu decimates poultry populations. In addition, as live viruses have to be handled and as the virulence of new strains is rarely predictable, egg-based manufacturing raises safety concerns for production workers.

Egg-based production also requires important infrastructure, with high capital investment and important fixed costs, which have resulted in the construction of a few large facilities with high production capacities. This situation has led to the geographical clustering (70% of global capacity) of the seasonal influenza vaccine manufacturing capacity in Europe and North America, other important manufacturing capacity being located in Australia, Japan and China (Pan American Health Organization, 2009). Simpler and more affordable solutions are needed to allow for the worldwide implementation of geographically suitable vaccine manufacturing facilities.

Alternatives to the production of influenza vaccines in eggs exist. Vaccines composed of viral particles or antigens from mammalian cell culture have recently appeared on the seasonal influenza market vaccines, but their worldwide distribution remains limited. Mammalian cell-culture-based pandemic vaccines are also under development. This technology seems more suitable than the egg-based technology in the context of an avian influenza pandemic as it does not rely on avian production host, but rather on mammalian cells. However, not all strains will grow well in mammalian cells and as this relies on the culture of live viruses, it remains costly and does not represent a viable option for countries of the developing world.

Recombinant antigens as a potential vaccine solution

Recombinant influenza antigens are currently the best alternative to virus-derived antigens as pandemic vaccines. Recombinant antigens can be produced from the very moment the information on the genetic makeup of the new strain is made available, allowing for rapid initiation of the production process. The manufacturing of recombinant antigens is also considered safer than that of live virus-derived antigens as none of the manufacturing steps requires the manipulation of infectious particles. The next generation of products for the seasonal influenza vaccine market will probably be comprised of recombinant influenza antigens derived from insect cell culture. A trivalent vaccine composed of recombinant influenza HA produced in baculovirus-infected insect cells is currently awaiting approval by the Federal Drug Administration in United States. Although the production of recombinant influenza antigens bears important advantages over that of egg-derived vaccines in terms of speed and safety, the product, purified recombinant HA subunits, appears less efficacious than inactivated split-influenza vaccines as higher antigen content is required to generate a potent immune response (Treanor et al., 2007).

Virus-like particles offer an improved alternative to isolated (soluble) recombinant antigens for stimulating a strong immune response. Virus-like particles are assembled upon expression of specific viral proteins and present an external surface resembling that of their cognate virus but, unlike true viral particles, do not incorporate genetic material. The presentation of antigens in a particulate and multivalent structure similar to that of the native virus achieves an enhanced stimulation of the immune response with balanced humoral and cellular components. Such improvement over the stimulation by isolated antigens is believed to be particularly true for enveloped viruses as enveloped VLPs present the surface antigens in their natural membrane-bound state (Grgacic and Anderson, 2006). The protective advantage of VLPs is now acknowledged as VLP-based vaccines against the hepatitis B virus (HBV) and the human papillomavirus (HPV) are commercialized.

The influenza virus is an enveloped particle budding from the plasma membrane of host cells. A schematic representation of the influenza viral particle is presented in Figure 1 (a and b). Hemagglutinin (HA) (green) and neuraminidase (NA) (orange) are the two major viral proteins protruding outside of the viral envelope (pink), HA being the main antigenic determinant of the virus. An influenza VLP, comprising HA only, is also represented in Figure 1 (a and b). The comparison of electron microscopy images of the virus and the VLP evidences the exterior likeness of both particles (Figure 1c).

Figure 1.

 Schematic representation of structural characteristics of viral particles and virus-like particles (VLPs). (a) Comparison of external characteristics. (b) Cross-section representation showing internal distinctions. (c) Transmission electron microscopy images of influenza viral particles and plant-produced influenza VLPs.

Influenza VLPs were first obtained from cultured mammalian cells co-expressing all 10 influenza proteins from cDNA copies of viral genes (Mena et al., 1996). Efforts at determining the components involved in viral budding have led to the conclusion that several viral proteins were dispensable for the production of VLPs, i.e. polymerase proteins (PA, PB1 and PB2), nucleoprotein (NP), ion channel proteins (M2) and HA, and that expression of the matrix (M1) protein was driving VLP budding (Gómez-Puertas et al., 2000). Therefore, in the last decade, influenza VLPs in vaccine development programmes have been produced from the co-expression of the two major antigenic envelope proteins (HA and NA) with M1 or from the co-expression of HA and M1 only (see Kang et al., 2009 for a review on VLP-based influenza vaccines). More recently, Chen et al. (2007) have shown that HA alone is capable of driving VLP formation and budding as M1 co-expression could be omitted in their system. However, the study also indicated that because HA binds to the sialylated glycoproteins on the surface of the mammalian cells producing the VLPs, viral NA (a sialidase) co-expression was required to allow the release of VLPs from the producing cell after budding, but that viral NA can be replaced by exogenous sialidase activity. Therefore, influenza NA does not contribute to VLP formation but rather to its release from the producing cells by removing HA receptors from the cell surface.

The adjuvant effect resulting from the presentation of an antigen or a group of antigens in the form of VLPs was exemplified for influenza with VLPs produced from the co-expression in insect cells, of three influenza structural proteins, HA, NA and M1 from the influenza A/Fudjian/411/2002 (H3N2) strain. In a comparative study in mice, intramuscular administration of these VLPs elicited high anti-HA titers as for the rHA and the whole inactivated virus (WIV) and only the VLPs and WIV elicited a Th1-type immune response that is indicative of cell-mediated immunity. In addition, animals immunized with VLPs produced antibodies that recognized a broader panel of antigenically distinct H3N2 isolates than those immunized with rHA or WIV (Bright et al., 2007). The ability to induce an immune response capable of neutralizing viruses of distinct phylogeny is a desirable characteristic of influenza vaccines as influenza evolves extremely rapidly.

Among the novel recombinant antigen production platforms, Agrobacterium infiltration-based transient expression (agroinfiltration) in Nicotiana benthamiana has shown unprecedented speed and productivity. It has been used to produce hundreds of milligram of complex proteins—requiring multiple post-translational modifications—per kilogram of fresh biomass, including antibodies, within a week following transfection (Vézina et al., 2009). When applied to the production of glycosylated antigens, plant-based platforms might have an additional advantage. They generate antigens bearing plant-specific N-glycans, and plant-specific N-glycans are thought to act as adjuvant by facilitating antigen capture by antigen presenting cell (APC) through binding to sugar-specific cell surface receptor (Saint-Jore-Dupas et al., 2007).

It was concluded that a combination of the speed of production offered by the Agrobacterium-based transient expression system in N. benthamiana plants with the efficacy and safety of the influenza VLP-based vaccine would result in the best integrated solution to respond to an influenza pandemic. Therefore, we have developed a plant-based influenza VLP production platform and assessed the potential of these plant-derived VLPs as candidates for human vaccination. The following sections review the production of VLPs and other influenza antigens in plants and summarize the development of influenza VLP-based vaccine production in plants from research to clinical trials.

Successful production of VLPs in plants

Mason et al. (1992) were the first to demonstrate that the expression of a human virus envelope protein in plants leads to the formation of enveloped VLPs budding out of intracellular membranes. In this study, HBsAg purified from plant biomass were observed as spherical particles of 22 nm in diameter, corresponding in size to the naturally occurring subviral particles and to the yeast-produced HBV enveloped VLPs. Plant-produced HBV VLPs were further shown to induce potent B-cell and T-cell immune responses in mice when administered parenterally (Thanavala et al., 1995; Huang et al., 2005), but oral immunization through feeding studies only induced a modest immune response (Kapusta et al., 1999; Richter et al., 2000; Kong et al., 2001; Smith et al., 2002, 2003), probably resulting from the low doses of VLPs contained in the plant tissue used for immunization. Enveloped HBV VLPs were later engineered into nanoparticles for the display of large foreign antigens through protein fusion with HBsAg as demonstrated with a GFP-HBsAg fusion (Huang and Mason, 2004). In a similar approach, Greco et al. showed that human immunodeficiency virus (HIV) epitopes in fusion with HBsAg accumulated as VLPs when expressed in transgenic tobacco and Arabidopsis, creating a bivalent VLP vaccine (Greco et al., 2007).

Another VLP candidate vaccine from plants currently under development targets the Norwalk virus and is based on the expression of the viral capsid protein (NVCP), resulting in the assembly of non-enveloped VLPs. Initially produced in transgenic tobacco and potato plants (Mason et al., 1996), NVCP VLPs have also been produced at much higher levels in agroinfiltrated N. benthamiana leaves (Santi et al., 2008; Huang et al., 2009), and their immunogenicity upon oral administration was demonstrated in mice (Santi et al., 2008). In a phase I clinical trial, 20 healthy adult volunteers were given two or three doses of raw potatoes containing 215–751 μg of NVCP in the form of VLPs, and 95% of the immunized volunteers developed an immune response (Tacket et al., 2000). Other examples of plant-produced VLPs include the non-enveloped VLP obtained from the expression of HBV core antigen (HBcAg) (Tsuda et al., 1998; Huang et al., 2006, 2009; Mechtcheriakova et al., 2006; Sainsbury and Lomonossoff, 2008), and the HPV major capsid protein L1 (Biemelt et al., 2003; Varsani et al., 2003; Warzecha et al., 2003; Maclean et al., 2007; Fernández-San Millán et al., 2008).

Together, these studies illustrate that plants produce virus-like particles of various origins and with various structural characteristics, that the expression of chimeric protein fusion comprising a target antigen fused to a viral envelope protein results in novel particles displaying large antigens and that, when administered properly, plant-derived VLPs induce a potent immune response in model animals and humans.

Influenza antigens and antigen fragments produced in plants

Evidence of influenza antigen production in plants was only recently disclosed, in 2004, through an international patent application by Cardineau et al. (2004, WO 2004/098533). In this application, it is concluded that influenza HA accumulates in calluses of NT1 tobacco cells transformed to express the complete coding sequence of influenza H5 from strain A/turkey/Wisconsin/1968 (H5N9). When extracted without detergent, the apoplastic fluid of the transgenic cells exhibited hemagglutination activity, indicating that plant-produced H5 was secreted and active. Rabbits immunized with a crude preparation of extracellular fluid from transgenic NT-1 cells producing H5 from strain A/turkey/Wisconsin/1968 (H5N9) showed a strong hemagglutination inhibiting (HI) antibody response 6 weeks after immunization in the presence of Freund’s adjuvant. Influenza vaccine candidates should induce the production of high titers of antibodies capable of blocking the attachment of the virus to host cells. This characteristic is measured as HI antibody titer and represents the highest dilution of serum from immunized animals capable of inhibiting the agglutination of red blood cells by the virus. The assay is described by WHO 2002 and Kendal et al. (1982). However, the use of Freund’s adjuvant in the H5 immunogenicity study described by Cardineau limits the extent to which the results can be considered a demonstration of immunogenicity and compared to other studies. Complete Freund’s adjuvant is known as a strong enhancer of immune response but is also a potent inflammatory agent. Its use is prohibited in humans and alternative adjuvants are preferred for animal studies.

Influenza antigens have also been produced in agroinfiltrated N. benthamiana plants in the form of HA antigenic domains fused to a carrier protein. Such a strategy was first described by Musiychuk et al. (2007) who showed the accumulation of the stem (amino acids 17–58 and 293–535) and globular (amino acids 59–292) domains of H5 from strain A/Vietnam/04 (H5N1) when individually produced in fusion with the carrier protein LicKM. The authors also reported the expression of active influenza NA from the same strain in agroinfiltrated plants. The LikM carrier fusion strategy was later successfully applied to the production of HA globular and stem domains from influenza A/Wyoming/3/03 (H3N2) and the administration to ferrets of a high single dose of 200 μg of H3 fragment-LickM fusions (corresponding to approximately 100 μg of HA) in combination with 50 μg of NA in the presence of alum has induced a strong immune response, with a reported mean HI titer of 1 : 1273. Interestingly, a similar antigen cocktail missing the NA antigen required two doses to induce a similar immune response, indicating a potential role of NA as modulator of the immune response (Mett et al., 2008). These results demonstrated the potential of agroinfiltration for producing influenza antigens. However, the dosage required to induce a potent immune response in ferrets was much higher than the current industry standards.

Expression of the HA ectodomain (the segment of the protein spanning outside of the viral envelope) fused to a KDEL peptide, to enhance accumulation through retention in the endoplasmic reticulum, and a poly-histidine purification tag has also been utilized as a mean for facilitating the production of recombinant influenza antigens in plants. Examples include the HA ectodomains from a human seasonal influenza strain (A/Wyoming/03/03 (H3N2) (Shoji et al., 2008) and highly pathogenic avian strains A/Indonesia/5/05 (Shoji et al., 2009a), A/Bar-headed Goose/Qinghai/1A/05 and A/Anhui/1/05 (Shoji et al., 2009b). In immunogenicity studies in mice, HA ectodomains have been shown to induce significant HI responses when administered in conjunction with Quil A. However, although such composition induced a strong immune response with doses as low as 1 μg (A/Anhui/1/05 (H5N1)) (Shoji et al., 2009b), two doses were required to obtain an HI antibody response. A ferret study further showed that three doses of 45 μg adjuvanted with Quil A were required to confer protection against a lethal challenge with the homologous strain of H5 ectodomain (Shoji et al., 2009a). Together, the above mentioned immunogenicity studies, performed with HA fragments in fusion with a carrier protein or with other peptides, provided strong indications that plant-made influenza antigens can be produced by agroinfiltration and that these HA fragments induce hemagglutination inhibition antibody response in model animals. However, the high dosage and multiple injections required to induce a protective immune response in ferrets from immunization with the ectodomain of H5 (A/Indonesia/5/05 (H5N1)) suggests that a higher order of antigen organization is required for optimal stimulation of a protective immune response.

More recently, attempts at producing the entire H5 protein (from strain A/Vietnam/1203/04 (H5N1)) or its HA1 domain by transient or stable transformation of N. benthamiana were reported unsuccessful as they led to only detectable accumulation of the mature H5 or HA1 domain (Spitsin et al., 2009). Fragments of 34 or 27 kDa from the HA1 domain, or fusions of the HA1 domain with 26 kDa fragment from a human or a mouse heavy chain constant region were reported to accumulate at higher levels. In an immunogenicity study in mice, two doses of 10 μg of the 34 kDa fragment from the antigenic region of H5 from influenza A/Vietnam/1203/04 (H5N1), administered in the presence of alum-CpG as adjuvant, induced high H5 specific antibody titers but failed to induce significant HI antibody titers (Spitsin et al., 2009), again highlighting the need for a higher degree of organization of multivalent antigens to stimulate a protective immune response.

The presentation of small peptides on chimeric viral particles from non-enveloped plant viruses can be achieved by expressing, in a plant, a coat protein (CP) from a plant virus fused to a peptide from a target antigen. The production of chimeric viral particles in plants for the presentation of an animal virus antigenic epitope has first been demonstrated using a fusion of the cowpea mosaic virus (CMPV) small capsid protein (CP) with an epitope derived from VP1 of the foot-and-mouth disease virus (FMDV) (Usha et al., 1993). A similar strategy was recently proposed in the international patent application WO2007/011904 for the presentation of influenza M2e universal epitope onto CPMV particles. Cowpea plants rubbed with RNA1 and chimeric RNA2 encoding a CP-M2e fusion produced chimeric CPMV particles comprising CP-M2e (Rasochova et al., 2007). Similarly, expression of a chimeric cucumber mosaic virus (CMV) capsid protein fused to the M2e epitope in N. benthamiana using a potato virus X (PVX) expression vector also led to the production of chimeric CP-M2e. However, assembly of the capsid proteins into chimeric viral particles was not demonstrated in this study (Nemchinov and Natilla, 2007).

Capsid proteins of the PVX expression vector have also been used for the display of H-2Db-restricted epitope from the influenza NP from strain A/PR/8/34 (Lico et al., 2009). In this study, the epitope coding sequence was fused to the CP gene of the PVX vector, creating chimeric PVX particles displaying the NP epitope. An immunogenicity study in mice, designed to evaluate the cell-based immune response, showed that when administered in the presence of incomplete Freund’s adjuvant, 50 μg of chimeric viral particles displaying the NP epitope activated ASNENMETM-specific CD8+ IFN-γ secreting cells.

Although chimeric viral particles antigen presentation platforms benefit from the display of selected antigenic epitopes in a multivalent fashion, this strategy also bear some intrinsic drawbacks. Only a limited number of antigenic epitopes of small size (less than 25 amino acids) can be displayed using this system. Therefore, it is easier for rapidly evolving viruses like influenza to evade the immune response induced by such vaccines by replacing a few amino acids in the selected immunogenic region. Conformational epitopes may also not fold properly, thereby inducing the production of antibodies that will not recognize the cognate native epitope. Finally, the use of chimeric plant viral particles as presentation devices will require regulatory acceptance in themselves in addition to the regulatory hurdles faced by new manufacturing platform.

Influenza VLP production in plants

Plants are ideal hosts for the production of influenza VLPs from the sole expression of HA because sialylated substrates for the attachment of HA to the cell surface are absent. Effectively, plants do not synthesize sialic acids (Séveno et al., 2004), and plant glycoproteins are thus not sialylated (Lerouge et al., 1998; Saint-Jore-Dupas et al., 2007). Virus-like particles comprising only HA bear several advantages as pandemic vaccines. First, only the HA coding sequence of the pandemic viral strain is required to initiate vaccine expression. Secondly, VLPs bearing a single protein impose fewer constraints on process and product characterization, again reducing production time. Also, a simpler product increases the chances of obtaining reproducible expression levels and lowers the risk of failure in a context where optimization of the production process has to be minimized when adapting for a novel strain. Our group has put plants to test for their ability to express and accumulate H5 from strain A/Indonesia/5/05 (H5N1) and H1 from strain A/New Caledonia/20/99 (H1N1) and to assemble VLPs from the expression of HA only. In a first study, we found that both H5 and H1 antigens accumulate at high levels in agroinfiltrated N. benthamiana. It was then discovered that these antigens accumulated as ultrastructures larger than the expected size of HA trimers. The combination of differential centrifugation, size-exclusion chromatography, electron microscopy and light scattering analyses later showed that these were true VLPs with a lipid bilayer envelope supporting the presentation of HA trimers (D’Aoust et al., 2008). Assembly of HA into VLPs upon expression in N. benthamiana has now been demonstrated for several other HAs from type A influenza, including H2, H3, H6 and H9, and from type B influenza (HAB) (D’Aoust et al., 2009).

Lipid composition analysis of purified H5-VLPs established the plasma membrane origin of the VLP envelope, confirming that, as for the influenza virus in its natural host, plant-produced influenza VLPs bud from the plasma membrane (D’Aoust et al., 2008). Localization by electron microscopy on thin sections from H5-VLP producing leaves indicated that VLPs accumulate between the plasma membrane and the cell wall and that the membrane retracts to accommodate the accumulation of several VLPs at budding sites (Figure 2). Together, these results showed for the first time that viral particles were formed through budding from the plasma membrane of plant cells. Therefore, the absence of enveloped plant viruses budding from the plasma membrane is not the consequence of incompatibility of plant plasma membrane with viral budding. The production of VLP from the expression of HA only also confirmed the suggestion that HA is the driving force behind influenza virus budding.

Figure 2.

 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 and VLPs: virus-like particles. The bar represents 500 nm. FromD’Aoust et al., 2008.

HA-based influenza VLPs have also been tested for their antigenic potential in immunogenicity studies in mice and ferrets. Mice studies have shown that intramuscular or intranasal administration of two doses of H5-VLPs (A/Indonesia/5/05) in the presence or absence of adjuvant (alum for intramuscular and chitosan for intranasal) induced a stronger HI antibody response than the same antigen (H5 from A/Indonesia/5/05) not assembled into VLP structures (D’Aoust et al., 2009). Two intramuscular doses of as low as 0.1 μg of H5-VLPs induced a HI antibody response with mean titer above 1 : 40 and increasing the doses up to 12 μg did not significantly increase the HI antibody response. It has further been established that 100% of mice vaccinated with two doses of as low as 0.5 μg of H5-VLPs were protected against a heterologous lethal challenge with one LD50 of A/Vietnam/1194/04 (H5N1) viral isolate (D’Aoust et al., 2008) and in a separate study that 1 μg of the same VLPs protected mice against another heterologous lethal challenge with 10 LD50 of A/Turkey/582/06 (H5N1) isolate (D’Aoust et al., 2009).

Immunogenicity studies in ferrets were performed in pursuit of the development of the H5-VLP vaccine. In these studies, the protective potential of the immune response induced by H5-VLPs was evaluated based on the criteria published in the guidelines for product licensure of the European Committee of Medicinal Products for Human use (CHMP). Under these guidelines, an influenza vaccine is considered to confer protection in humans when the three following criteria are met: the percentage of individuals showing a fourfold increase in HI antibody response exceeds 40%, the mean geometric increase in HI antibody titer exceeds 2.5 and the percentage of individuals showing an HI antibody titer above 1 : 40 exceeds 70%. As shown in Table 1, the ferret study demonstrated that a single dose of 5 μg and two doses of 1 μg of H5-VLPs adjuvanted with alum induced a strong HI antibody response against a homologous H5N1 strain (A/Indonesia/5/05), which can be considered protective based on CHMP criteria (D’Aoust et al., 2009). The cross-reactivity of the H5-VLPs was also verified in this study through quantification of HI antibodies against heterologous strains of H5N1. Two adjuvanted doses of 1 μg stimulated a potent cross-reactive immune response against H5N1 clade 2 viruses, namely, A/turkey/Turkey/1/05 (clade 2.2) and A/Anhui/1/05 (clade 2.3) and 5 μg doses elicited an antibody response that meets the CHMP criteria against a clade 1 strain (A/Vietnam/1194/04) (Table 2) (D’Aoust et al., 2009). More recently, we have observed that ferrets having received two alum-adjuvanted doses 1.8 μg of H5-VLPs were fully protected against a heterologous lethal challenge with 10 LD50 of an influenza A/Vietnam/1194/05 isolate (Medicago, 2009). The comparison of these results with those obtained from administration of HA fragments provides the demonstration that the assembly of HA into trimers and further into VLP increases potency with the induction of a strong and broad immune response.

Table 1.   Evaluation of immune response of ferrets after immunization with H5- virus-like particles (VLPs) using criteria from the European Committee of Medicinal Products for human use (CHMP). From D’Aoust et al., 2009
 Criteria for licensure of influenza vaccinesStudy group
1 μg5 μg
1st injectionPercentage of fourfold increase in HI titer >40%100%100%
Mean geometric increase >2.57.615.6
Percentage of HI titer above 1/40 > 70%60%100%
2nd injectionPercentage of fourfold increase in HI titer >40%100%100%
Mean geometric increase >2.58293
Percentage of HI titer above 1/40 > 70%100%100%
Table 2.   Evaluation of cross-reactive immune response of ferrets after immunization with two doses of H5- virus-like particles (VLPs) using criteria from the European Committee of Medicinal Products for human use (CHMP). From D’Aoust et al., 2009
H5N1 influenza strainCriteria for licensure of influenza vaccinesStudy group
1 μg5 μg
A/turkey/Turkey/1/05 (clade 2.2)Percentage of fourfold increase in HI titer >40%80%100%
Mean geometric increase >2.510.620.8
Percentage of HI titer above 1/40 > 70%100%100%
A/Anhui/1/05 (clade 2.3)Percentage of fourfold increase in HI titer >40%100%100%
Mean geometric increase >2.511.814.4
Percentage of HI titer above 1/40 > 70%100%80%
A/Vietnam/1203/04 (clade 1)Percentage of fourfold increase in HI titer >40%60%80%
Mean geometric increase >
Percentage of HI titer above 1/40 > 70%0%80%

Large-scale influenza VLP production

Agroinfiltration in N. benthamiana plants combines speed and productivity. Various agroinfiltration methodologies have evolved so that now several whole plants can be infiltrated simultaneously for large-scale production (D’Aoust et al., 2008). Massive expression of recombinant proteins, obtained 5–11 days after the infiltration of the bacterial suspension in the leaves, have been reported using agroinfiltration-based systems (Marillonnet et al., 2005; Giritch et al., 2006; Sainsbury and Lomonossoff, 2008 and Vézina et al., 2009). The following paragraphs describe the transient expression technology that we have designed for the large-scale production of influenza VLPs.

A series of DNA regulatory elements have been tested and used for the transient expression of influenza VLPs. One of the most potent sets of regulatory elements reported for use in transient expression in N. benthamiana includes the promoter and terminator from the alfalfa plastocyanin gene (D’Aoust et al., 2008; Vézina et al., 2009). More recently, a CPMV-HT vector (Sainsbury and Lomonossoff, 2008) induced even higher accumulation of VLPs than the plastocyanin-based cassette and was adopted for the large-scale (commercial) production of H5 VLPs. Co-expression of a suppressor of silencing also significantly increased the transient accumulation of VLPs in N. benthamiana. Our VLP expression strategy thus includes the co-expression of P1/HcPro from the potato virus Y. The co-expression of helper proteins have been evaluated as a strategy to increase influenza VLP accumulation. Among the candidate proteins tested, cytosolic HSP40 and HSP70 co-expression with HA has resulted in increased VLP accumulation for H3 (from strain A/Brisbane/10/07 (H3N2)), but accumulation level of H1 (from strain A/New Caledonia/20/99 (H1N1)) was not affected by HSP co-expression (D’Aoust et al., 2009).

The plants destined to the production of H5 VLPs are grown from a working seed bank generated on an annual basis through selfing of parental plants. Similarly, a banking system has been established for the Agrobacterium strains used for production. As different bacteria bear the H5 and the P1/Hc-Pro expression plasmids in the current H5 production process, a master cell bank has been prepared and characterized for each bacterial strain. Each working cell bank, derived from the master cell bank, contains enough aliquots to support the production of 300 production batches. Batches of 1200–1500 plants are agroinfiltrated weekly and incubated for 6 days to allow formation and accumulation of the VLPs in planta. Following incubation, 25 kg of leaf biomass is harvested and processed through primary recovery for immediate treatment.

Primary recovery of the H5-VLP process is initiated by mechanical homogenization of freshly harvested biomass. Insoluble materials are first eliminated from the extract by centrifugation followed by a clarification step based on chemical and physical treatments. The resulting clarified extract is concentrated and stabilized prior to purification. The drug substance is isolated from the clarified extract through a series of chromatographic steps including ion exchange and affinity chromatography and the purified product is concentrated by cross-flow filtration, formulated and sterilized by filtration.

Assessing the speed of the platform with A/H1N1

The spring of 2009 was marked by an unanticipated outbreak of a novel A/H1N1 influenza virus at a moment when worries and worldwide preparedness were oriented towards the risk of an A/H5N1 pandemic. The novel A/H1N1 outbreak rapidly developed into a pandemic, as recognized by the WHO on June 11. Fortunately, the new strain caused relatively mild symptoms during the first wave of infection in the Northern and Southern Hemispheres, allowing more time for the preparation of a vaccine with current manufacturing technologies. The unanticipated character of this pandemic stressed the need for an influenza vaccine manufacturing solution that is responsive enough to initiate the production of a strain-specific vaccine immediately upon identification of a new strain in the case of a more severe influenza outbreak. Ideally, this would allow vaccination as quickly as possible after the pandemic is declared.

As sequences of one strain of the novel A/H1N1 (A/California/04/09) became available from the Centre for Disease Control and Prevention (CDC) on April 24th, a fast track group was created within the company to challenge our ability to quickly produce an experimental vaccine specific for the novel strain. As presented in Figure 3, only 2 weeks were required to obtain the first proof of feasibility including an immunological demonstration that the HA of the new A/H1N1 strain was accumulating at high levels in agroinfiltrated plants. Of note, we did not observe a reduction in productivity with this strain, contrasting with the important yield reduction noted by egg-based vaccine manufacturers for this specific strain (World Health Organization, 2009c). The first purified lot of the research grade vaccine was obtained only 5 days later.

Figure 3.

 H1- virus-like particle (VLP) experimental vaccine production timeline.

An immunogenicity study in mice was conducted to complete the analysis of responsiveness of our VLP-based vaccine production system and demonstrated that not only the platform is rapid but the influenza vaccine candidate is efficacious. Five micrograms of H1-VLP from strain A/California/4/2009 (H1N1) were injected with and without adjuvant (alum) in 6–8 week old BALB/c mice (groups of five animals). Mice were immunized on day 0 and 21, and sera were sampled on day 21 (one dose) and 42 (two doses). Hemagglutination inhibiting antibody titers were quantified at the Southern Research Institute (USA) against whole live viruses of A/California/04/2009 (H1N1) and A/California/07/2009 (H1N1). Live viruses had to be used to accelerate the demonstration because WIV of these isolates—the reagent used in the industry to standardize HI antibody quantification—were not readily available as sera were collected and ready for analysis. Hemagglutination inhibiting antibody titers against the A/California/07/2009 isolate were included in the study because, although the composition of HA from the two isolates only differs by two amino acids, virions of A/California/07/2009 showed a much higher hemagglutination capacity than A/California/04/2009 virions, thereby increasing the sensitivity of the HI antibody quantification assay. The HI antibody response in mice immunized with our A/H1N1 experimental vaccine is presented in Table 3. After a first injection, mice developed serum HI antibody response irrespectively of the presence of adjuvant. As expected, the use of viruses from the A/California/07/2009 isolate for red blood cell agglutination in the HI assay led to measurement of higher titers than with A/California/04/2009 virions. After the second dose, mice developed a much stronger HI antibody response, reaching a mean HI antibody titer of 1 : 385 in the presence of alum and 1 : 116 in the absence of adjuvant. On June 30th, the immunogenicity study in mice was completed, and results were disclosed. In conclusion, the A/H1N1 challenge demonstrated that a highly efficacious pandemic VLP vaccine can be produced in plants within days from the identification of a new influenza strain.

Table 3.   HI antibody titers measured in sera from mice immunized with H1- Virus-like particle (VLP) from strain A/California/4/2009. Titers are expressed as reciprocal of the highest dilution of serum that inhibits hemagglutination of turkey red blood cells
Isolate*DoseHI titers (GMT**)
First injectionSecond injection
  1. *Whole live viruses were used for the assay. **Geometric mean titers.

A/California/04/20095 μg7.660.6
5 μg + alum5.9183.8
A/California/07/20095 μg30.3116.0
5 μg + alum40.0385.1


Our group has demonstrated that influenza VLPs, an efficacious and safe vaccine product, can be produced in plants with unprecedented speed and animal studies have confirmed the efficacy and safety of the plant-produced VLPs. We have received clearance from Health Canada to initiate a phase I human clinical trial with our H5-VLP vaccine, and as this is being written, we are conducting a placebo-controlled, double blind, dose escalating study on 48 healthy volunteers to evaluate the safety and tolerability of the H5-VLP vaccine, and the immune response of the volunteers to vaccination.