Release and Stability Testing Programs for a Novel Virus-Like Particle Vaccine

ABSTRACT

Virus-like particle (VLP) vaccines are currently used for the prevention of hepatitis B and human papillomavirus infections. Influenza VLPs have been produced using a baculovirus–insect cell production system and shown to be safe and immunogenic in multiple Phase 2 clinical trials. These vaccines present unique challenges in their development in terms of the design of product release assays, product characterization, and stability testing. Methods that have been used to characterize simpler biological drugs (monoclonal antibodies, protein therapeutics) to ensure their potency and consistency have been applied to release and stability testing of VLPs and will be discussed here.

The H1N1 pandemic influenza outbreak has served as a reminder that current approaches to the development and production of vaccines are not sufficient to meet future unexpected needs. There have been three vaccine technologies used for the prevention of human viral disease. Classical live attenuated and inactivated or split vaccines have been used successfully to protect against multiple viruses including smallpox, influenza, measles, mumps, poliovirus, hepatitis A, herpes zoster, and rotavirus. These vaccines have been safe and effective but in some cases their use has been limited, particularly in developing countries, because of cost and requirements for a cold storage chain. Challenges are also seen in the industrialized world, because these types of vaccines require the growth of viral strains which in many cases is carried out in chicken eggs, and the strains do not always grow well in that platform. More recently, recombinant technology has allowed the development of new types of vaccines (subunit proteins, virus-like particles) that do not require the growth of viral agents and are non-infectious.

Virus-Like Particle Vaccines (VLPs)

Live attenuated and inactivated whole virus vaccines are the progenitors of VLPs that are in use today to prevent hepatitis B (Recombivax HB, Engenrix B) and human papillomavirus infections (Gardasil, Cervarix). VLPs contain viral proteins required to generate protective immune responses but are non-infectious and lack the potential to recombine with circulating virus because they lack viral nucleic acid. These vaccines have potential immunologic advantages over split or purified protein vaccines because: 1) VLPs contain multiple copies of antigens presented in an organized array, thus allowing activation of the innate immune system and 2) the size of VLPs allows efficient uptake and processing by dendritic cells, which are the antigen presenting cells of the immune system and stimulate lymphocytes in lymph nodes.

(VETTER PHARMA INTERNATIONAL GMBH)

Novavax Recombinant VLP Vaccines

Through the use of a recombinant baculovirus–insect host cell system, Novavax has developed a platform technology to produce VLP vaccines to multiple targets including influenza, SARS, and HIV. Influenza VLPs have been biochemically characterized and GMP-produced materials have been tested in five clinical trials, with a total of more than 5,000 subjects.¹

The establishment of a recombinant baculovirus containing the influenza genes for the hemagglutinin (HA) and neuraminidase proteins (NA) together with the gene for Matrix (M1) protein is the key step in this process.² The HA and NA genes are obtained from viral RNA of relevant circulating seasonal or pandemic influenza strains by molecular means from isolated RNA. Alternately, they can be chemically synthesized and then, through a series of cloning steps, placed in a single plasmid (bacmid) under the control of baculovirus-specific transcription initiation and translation elements along with a constant M1 gene sequence shown to allow enhanced VLP formation. Transfection of the bacmid into Spodoptera frugiperda (Sf9) insect cells generates baculoviruses that encode and express the three influenza genes. Baculovirus stocks are further amplified on Sf9 cells to generate master and working virus stocks. Infection of Sf9 cells with the recombinant baculovirus allows expression of the HA, NA, and M1 proteins which form pleomorphic spherical membrane-containing particles with HA and NA protein spikes on their surface and an M1 core. An example of an influenza VLP is shown in Figure 1. These particles are released into the cell culture supernatant and then purified, resulting in the VLP vaccine. The membrane nature of these particles is different from currently licensed VLP vaccines and presents purification and characterization challenges.

Figure 1. Recombinant influenza virus-like particles (VLPs) and pleomorphic spherical particles

VLP Production and Purification

Production and purification of influenza VLPs has been performed for Phase 2 clinical trial materials. Sf9 insect cells from a working cell bank are amplified to seed a 100-L disposable wave bioreactor cell culture, which is then infected with a master baculovirus seed. During harvest, cells are removed by tangential flow filtration (TFF) and the filtrate is concentrated and diafiltered to remove media components and cell debris.

VLPs are separated from baculovirus (BV) particles and contaminating RNA and DNA using a flow-through ion exchange (IEX) chromatography step in which BV, RNA, and DNA are bound to the column while VLP is allowed to flow through (Figure 2). This process step takes advantage of the charge difference between BV and VLP; BV is highly negatively charged by virtue of its DNA content, whereas VLP lacks this charge given the absence of DNA. Allowing VLP to be separated from BV without binding to the column matrix eliminates the need to modify this purification step from season to season in response to changes in the HA protein.

Figure 2. Purification of virus-like particles (VLPs) using an ion exchange (IEX) column. Host cell contaminants-baculovirus (BV), DNA, and RNA-bind, whereas VLPs flow through.

Residual BV is inactivated by treatment with beta-propiolactone (BPL) and residual host contaminants are removed by size exclusion chromatography (SEC), as shown in Figure 3. The VLP preparation is sterile filtered to produce the bulk vaccine.

Figure 3. Residual host contaminants are removed by size exclusion chromatography.

Clinical Lot Release Testing

Release testing is designed to monitor the safety, identity, purity and potency of the VLP vaccine (Table 1). Many of the methods are typical of those used to release licensed split-virion and whole-virus influenza vaccines and were adapted to the VLP product. Additional methods were developed to specifically characterize the VLPs and to demonstrate removal of host and baculovirus DNA and protein as well as materials used in the manufacturing process.

Table 1. Analytical testing of influenza VLP bulk vaccine

Single Radial Immunodiffusion (SRID) Potency Assay

Single radial immunodiffusion (SRID) is the standard potency assay used by all manufacturers and appropriate regulatory agencies to determine the concentration of immunologically active HA in all licensed inactivated influenza vaccines.³ The assay requires two standards for each influenza vaccine strain: 1) a reference HA protein whose concentration is determined by an alternative method and 2) a polyclonal antibody source that recognizes the reference HA protein. The method is based on diffusion of the HA protein into an agarose gel containing a prequalified amount of specific antibody, raised against the HA protein, which forms precipitation rings proportional to a fixed concentration of the reference HA. The potency for a vaccine lot is computed by the parallel line bioassay method using the reference standard and test vaccine dose response curves.

Table 2 compares the source of reagents used by regulatory agencies to release standard inactivated influenza vaccines with reagents developed by Novavax to determine the potency of our baculovirus-derived product. Whereas regulatory agencies use HA antigen and corresponding antisera derived from egg-grown influenza virus, Novavax uses HA protein derived from baculovirus and prepares a corresponding antisera to that antigen.

Table 2. Comparison of the source of the single radial immunodiffusion (SRID) reagants used by regulatory agencies to release standard inactivated influenza vaccines and the reagants developed by Novavax to determine the potency of the baculovirus-derived VLP product.

Multiple VLPs have been tested for potency in the SRID assay comparing the egg- and baculovirus-derived reagents, and results for H5N1 pandemic vaccine lots are presented as a representative example in Table 3. Baculovirus- and egg-derived SRID reagents give comparable values for HA concentration for multiple lots of pandemic VLP vaccines.

Table 3. Comparison of the HA concentration results for H5N1 VLPs from single radial immunodiffusion (SRID) potency assays using egg-derived and baculovirus-derived reagents. The two different reagents give comparable results for multiple lots of vaccines.

The 2008–2009 inactivated influenza vaccine used a like H3N2 strain (A/Uruguay instead of the recommended A/Brisbane) because of problems with production of the recommended strain. Because VLP development does not require the growth of the actual influenza virus, the protein coding sequence for the A/Brisbane H3N2 HA was used to generate the baculovirus recombinant and subsequently the purified VLP vaccine containing the recommended HA. Potency of trivalent VLP and inactivated 2008–2009 influenza vaccines was tested with baculovirus- and egg-derived reagents in the SRID assay. Table 4 demonstrates that the homologous A/H1N1 and B/Florida components of both vaccines gave comparable potency values when tested by baculovirus- or egg-derived reagent where the results with homologous reagents (baculovirus-derived for VLP and egg-derived for inactivated egg produced vaccine) were considered the expected value. However, when the heterologous H3N2 HA potencies were determined, it was seen that accurate potency measurements required homologous reagents, i.e., H3N2 HA potency in VLP was accurately measured by baculovirus reagents but not by egg-derived reagents and the opposite was seen with egg-derived vaccine. The table also demonstrates that the source of the antisera was not an issue because baculovirus- and egg-derived antisera could be interchanged with the source of antigen and resulted in accurate HA concentration determinations. However, the source of HA reference protein was crucial; only baculovirus-derived A/Brisbane protein was accurate for measuring VLP H3N2 HA content and the egg-derived A/Uruguay HA protein was necessary to accurately test egg-derived vaccine for this strain.

Table 4. Matched and heterologus strain testing with egg- and baculovirus-derived reagents. The homologous A/H1N1 and B/Florida components of both Novavax (NVAX) virus-like particle and inactivated egg-produced influenza vacccines gave comparable potency values when tested with either baculovirus- (NVAX) or egg-derived (CBER) reagants. However, when the heterologous H3N2 HA potencies were tested, accurate potency measurements required homologous reagants (baculovirus for VLPs, and egg-derived for inactivated egg-produced vaccine).

NA Activity

In conventional inactivated influenza vaccines, no attempt is made to quantify or maintain NA activity. A number of preclinical and clinical studies have pointed to a significant role for NA-inhibiting (NAI) antibodies in protective immunity.^4–7 NA-specific responses observed in human and mouse studies are responsible for a reduction of viral replication and disease prevention. In primates, NA is known to induce NAI, virus-neutralizing, and HAI antibodies. Immunity toward NA is also known to affect primary viral outcomes (disease progression and viral titers) and prevention of secondary bacterial infections. Moreover, NA-specific responses protect against lethal doses of highly pathogenic avian influenza by eliciting NAI and virus-neutralizing antibodies. NA-specific responses are also known to have reduced the impact of the 1968 H3N2 pandemic.

VLPs contain quantifiable levels of active NA protein and we have shown that VLP-associated NA is very similar to the NA of the influenza virus. The enzymatic properties, Vmax (maximal velocity) and Km (substrate concentration required to achieve ½ maximal velocity) for VLP and viral NA in H3N2 and H5N1 strains are similar. In addition, VLP NA is found to be inhibited by oseltamivir carboxylate (Tamiflu) at a concentration similar to corresponding virus with a <3-fold difference in IC50. Moreover, in the clinical trials summarized above, we have demonstrated that immunization with VLPs elicits NAI antibodies in addition to HA-specific responses, demonstrating the broader immunological response of this vaccine.

Particle Size

The immunologic advantage of VLPs results from their particle nature, which allows efficient uptake, processing, and presentation by dendritic cells. VLPs are membrane particles with a diameter of 141–192 nm containing the influenza HA and NA proteins as surface spikes. Measurement of VLP particle size by a Malvern Zetasizer is an important release criteria to demonstrate consistency of the manufacturing process and ensures the absence of aggregation.

Purity Assessment

The goal of the purity assessment is to determine the levels of baculovirus (BV) and Sf9 cell DNA and proteins associated with each lot of VLP vaccine. These are determined by specific assays, as summarized in Table 1.

There are two sources for BV and Sf9 preparations in VLP vaccine. Although the current purification process for VLPs completely inactivates BV, it does not completely remove BV particles. Because VLPs bud from the insect cells, there is a likelihood that they acquire cellular and BV proteins during this process. Thus, determining the true host and BV protein contaminants in VLP vaccine is a considerable challenge.

Multiple batches of a single type of VLP must be analyzed to determine potential BV and Sf9 contamination. Two orthogonal approaches are used to determine VLP purity and the identity of BV and Sf9 proteins. SDS–PAGE followed by Coomassie staining allows us to visualize the proteins in each VLP batch. Scanning the stained sample lane and determining the percentage of each specific protein band allows the percentage of HA, NA, and M1 proteins to be determined and enables us to assign a relative purity value to the VLP (Table 3). LC–MS-MS of either the whole VLP sample or specific bands isolated from gels has been used to identify the BV and Sf9 proteins associated with different VLP lots. A typical VLP protein band profile is shown in Figure 4. The proteins detected were predominantly the expected influenza HA, NA, and M1 proteins engineered into the B/Florida VLP. Baculovirus gp64 envelope protein and p39 major capsid protein were identified through SDS–PAGE analysis of this vaccine along with viral ubiquitin detected by MS analysis. Sf9 proteins detected by MS analysis included tubulin, actin, Hsp70 chaperone, and small amounts of several housekeeping proteins.

Figure 4. Identification of proteins in virus-like particle influenza vaccines. This SDS-PAGE band profile of the B/Florida/4/06 VLPs, obtained by LC–MS-MS, shows that the proteins detected were predominantly the expected HA, NA, and M1 proteins, along with two baculovirus proteins (the gp64 envelope protein and the p39 major capsid protein).

ELISA assays have been developed as a second method to measure the concentration of BV and Sf9 proteins in VLP lots. As seen in Table 5, the purity of VLP lots based on the baculovirus ELISA is generally in good agreement with the values seen using the SDS-PAGE method.

Table 5. Comparison of VLP purity determined by SDS–PAGE and BV ELISA

An alternative method based on reverse-phase HPLC of VLP samples is being developed to identify the proteins in VLP lots. The protein peaks are identified either by Western blot using specific antisera or by fraction collection and analysis by mass spectrometry. A typical VLP profile with currently identified protein peaks is presented in Figure 5. Currently, this method can identify the influenza HA, NA, and M1 proteins as well as the baculovirus gp64 protein.

Figure 5. Reverse-phase HPLC peak identification of proteins in the A/Brisbane H1N1 strain of a virus-like particle

Stability Testing

During clinical development, a stability testing program is established to 1) demonstrate vaccine stability during the time it is administered to volunteers and 2) begin to collect data to allow the assignment of vaccine shelf life upon approval for marketing.

Table 6 presents 12-month stability data generated for the 60 µg HA/mL VLP lot of 2008–2009 trivalent seasonal VLP vaccine administered to healthy adults in a clinical trial (Table 2). The analytical testing parameters that were followed were 1) particle size, 2) SRID HA value for each strain in the vaccine, 3) vaccine purity/identity by SDS–PAGE and Western blot and 4) total NA activity. These measurements allow us to track potency and physical characteristics of the vaccine over the course of the study. The results demonstrated no change in the physical properties of the vaccine over 12 months and showed stability of the HA potency for all three components and consistent NA activity over the 12-month study.

Table 6. Stability testing results of 60 µg HA/mL dose of trivalent 2008–2009 influenza VLP vaccine

Use of VLP to Test Antibody Responses in Clinical Samples

Hemagglutination inhibition antibody (HAI) titer is the accepted immunological correlate for influenza vaccines.⁸ This assay measures the ability of sera to block agglutination of red blood cells by influenza viruses mediated by the HA protein (this property of VLP HA is tested as a release requirement: Table 3) and the accepted correlates for protection are defined as:

Achievement of Seroconversion: Lower 95% CI must have >40% of subjects with a >4-fold increase in titer and a minimum final titer of >40.

Achievement of HAI titer >40: Lower 95% CI must have >70% of subjects with a minimum final titer of >40.

The stability of HA in VLP suggests that this might be used as a source of HA to standardize HAI measurement by multiple laboratories. Serum samples were obtained from rabbits immunized with a trivalent seasonal VLP and the sera were tested for HAI titer to the H3N2 and H1N1 component using influenza virus and VLP as the source of HA. Figure 6 shows that sera from four rabbits had equivalent H3N2 HAI titers when virus or VLP were used as the source of HA in the assay. For H1N1, three of the four rabbit sera gave higher HAI titers when measured with VLP and the fourth serum gave equivalent titers with virus and VLP. These results suggest that VLP can be used as an alternative source of HA to measure HAI titers in clinical samples.

Figure 6. Measurement of hemagglutination inhibition (HAI) titer comparing influenza virus and VLP as the source of HA for H3N2 and H1N1

Concluding Remarks

Our influenza VLP vaccine technology has been shown to be safe and to elicit HAI and NAI antibodies in several clinical trials totaling more than 5,000 subjects. Release testing of this vaccine has involved both standard assays (SRID)used to confirm potency of the influenza vaccine, allowing dose confirmation by regulatory authorities, and unique assays (particle size, NA activity) developed to ensure the physical, chemical, and biological stability of this vaccine. These assays have demonstrated the stability of VLPs over a 12-month period with preservation of functional HA and NA, which suggests that VLP can be used to standardize clinical assays that provide immunological correlates of protection.

STEVEN PINCUS is the head of analytical and quality operations, SARATHI BODDAPATI is a senior scientist, formulation development, JINGNING LI is a scientist, analytical development, and TRAVIS SADOWSKI is a senior manager, quality control, all at Novavax Inc., Rockville, MD, 240.268.2032, spincus@novavax.com

References

1. Clinical trials performed include a) a two part 2009 pandemic H1N1 A/Calif/04 study in healthy adults (18–64 yrs) with an n=1000; dose ranging study with 5, 15, and 45µg HA dose injected IM in Stage A followed by an expanded safety Stage B study with an n=3550 and a 15µg HA dose; b) seasonal trivalent vaccine with more than 1,000 subjects performed in three different trials with 2008–9 seasonal trivalent strains in healthy adults (age: 18–49; A/Bris/59; A/Bris/10; B/Flor/04), 2009–10 seasonal trivalent in healthy older adults (age: >60; A/Bris/59; A/Bris/10; B/Bris/60) and 2005–6 seasonal trivalent (age: 18-49; A/NC; A/NY; B/Jiangsu) and c) H5N1 prepandemic A/Indonesia/05 study in healthy adults (18–40) in more than 200 subjects.

2. Pushko R, Tumpey TM, BeF, Knell J, Robinson R, Smith G. Influenza virus-like particles composed of the HA, NA and M1 proteins of H9N2 influenza virus induce protective immune responses in BALB/c mice. Vaccine. 2005:23:5751–9.

3. Williams MS. Single-radial-immunodiffusion as an in vitro potency assay for human inactivated viral vaccines. Vet Microbiol. 1993:37:253–62.

4. Nayak B, Kumar S, DiNapoli JM, Paldura A, Perez DR, Collins PL, Samal SK. Contributions of the avian influenza virus HA, NA and M2 surface proteins to the induction of neutralizing antibodies and protective immunity. J Virol. 2010:84:1489–1503.

5. DiNapoli JM, Nayak B, Yang L, Finneyfrock BW, Cook A, Anderson H, T, et al. Newcastle disease virus-vectored vaccines expressing the hemagglutinin or neuraminidase protein of H5N1 highly pathogenic avian influenza virus protect against viral challenge in monkeys. J Virol. 2010:84:1489–1503.

6. Huber VC, Peltola V, Iverson AR, McCullers JA. Contribution of vaccine-induced immunity toward either the HA or NA component of influenza virus limits secondary bacterial complications. J Virol. 2010:84:4105–8.

7. Gillim-Ross L, Subbarao K. Can immunity induced by human influenza virus N1 neuraminidase provide some protection from avian influenza H5N1 virus? PLoS Med. 2007:4:e91:226–8.

8. Stephenson I, Wood JM, Nicholson KG, Charlett A, Zambon MC. Detection of anti-H5 responses in human sera by HI using horse erythrocytes following Mf59-adjuvanted influenza A/Duck/Singapore/97 vaccine: Virus Res. 2004:103:91–5.