Pandemic influenza vaccine: characterization of A/California/07/2009 (H1N1) recombinant hemagglutinin protein and insights into H1N1 antigen stability

Background The recent H1N1 influenza pandemic illustrated the shortcomings of the vaccine manufacturing process. The A/California/07/2009 H1N1 pandemic influenza vaccine or A(H1N1)pdm09 was available late and in short supply as a result of delays in production caused by low yields and poor antigen stability. Recombinant technology offers the opportunity to shorten manufacturing time. A trivalent recombinant hemagglutinin (rHA) vaccine candidate for seasonal influenza produced using the baculovirus expression vector system (BEVS) was shown to be as effective and safe as egg-derived trivalent inactivated vaccine (TIV) in human clinical studies. In this study, we describe the characterization of the A/California/07/2009 rHA protein and compare the H1N1 pandemic rHA to other seasonal rHA proteins. Results Our data show that, like other rHA proteins, purified A/California/07/2009 rHA forms multimeric rosette-like particles of 20–40 nm that are biologically active and immunogenic in mice as assayed by hemagglutination inhibition (HAI) antibody titers. However, proteolytic digest analysis revealed that A/California/07/2009 rHA is more susceptible to proteolytic degradation than rHA proteins derived from other seasonal influenza viruses. We identified a specific proteolytic site conserved across multiple hemagglutinin (HA) proteins that is likely more accessible in A/California/07/2009 HA, possibly as a result of differences in its protein structure, and may contribute to lower antigen stability. Conclusion We conclude that, similar to the recombinant seasonal influenza vaccine, recombinant A(H1N1)pdm09 vaccine is likely to perform comparably to licensed A(H1N1)pdm09 vaccines and could offer manufacturing advantages.


Background
A novel influenza A virus (H1N1) of swine origin emerged in Mexico and the United States in March and early April 2009. The virus quickly spread worldwide through human-to-human transmission resulting in the World Health Organization raising the influenza pandemic alert to the highest level (phase 6) on June 11, 2009 [1-3]. The outbreak and spread of the first influenza pandemic of the 21 st century challenged licensed vaccine manufacturers to rapidly mobilize and generate a prophylactic vaccine. Delivery of initial doses of vaccine to the U.S. public coincided with the second peak of the pandemic, too late to provide timely protection and highlighting the need for alternative production platforms [2,4].
Two types of licensed influenza vaccines are available in the U.S.: trivalent inactivated vaccine and live attenuated influenza vaccine [5,6], both produced in embryonated chicken eggs. The process of preparing a new vaccine seed strain suitable for growth in eggs can be quite lengthy, as it involves re-assortment between the genes of a high yielding donor virus, such as A/Puerto Rico/8/34, and the hemagglutinin (HA) and neuraminidase (NA) genes of the new influenza strain [7]. The candidate seed virus strains are then further selected for high growth capability in eggs before they can be used for the production of vaccines. This manufacturing process is not only lengthy but also limited in scalability due to its dependence on the availability of embryonated chicken eggs.
The production of purified recombinant hemagglutinin (rHA) subunit vaccines via the baculovirus expression vector system (BEVS) is a leading alternative platform for influenza vaccine manufacture. The most advanced influenza vaccine candidate produced using this technology (under the trade name Flublok W ) is a trivalent composition of three rHA proteins corresponding to the full length HA proteins of the seasonally circulating influenza strains [8][9][10][11]. Clinical trials of Flublok have demonstrated that the vaccine is well-tolerated, immunogenic (as assessed by the induction of hemagglutination inhibiting [HAI] antibodies), and provides protection against drifted influenza viruses [8][9][10]. The rHA proteins in Flublok are produced using genetically modified baculoviruses in lepidopteran insect cells. The proteins are extracted and purified from cell pellet using a combination of filtration and column chromatography methods. Bulk vaccine can be produced within seven weeks of receipt of the HA gene sequence [9], making it an attractive platform for pandemic vaccine manufacturing as demonstrated during the initial outbreaks of H5N1 [12].
The original influenza seed viruses used for the eggbased production of the A(H1N1)pdm09 vaccine grew slowly, produced relatively low quantities of HA antigen and showed poor stability [7,13]. The rHA derived from A/California/07/2009 also revealed differences compared to other rHA proteins, as the pandemic rHA protein was more sensitive to proteolytic degradation and reacted uniquely in the single radial immunodiffusion (SRID) potency assay.
The objectives of this study were (1) to study the properties of the A/California/07/2009 rHA protein and compare it to other rHA proteins derived from seasonal influenza strains and (2) to develop an understanding of the cause of the instability observed with this antigen.

Biochemical and biophysical characterization
The electrophoretic mobility of purified A/California/07/ 2009 rHA protein was compared to purified rHA derived from A/New Caledonia/20/1999, A/Solomon Islands/03/2006, and A/Brisbane/59/2007 H1N1 seasonal influenza strains using reducing and non-reducing SDS-PAGE ( Figure 1). Purified rHA proteins typically migrate as monomers and disulfide-linked oligomers under non-reducing conditions. The primary full-length HA0 band migrates at approximately 62 kDa, and dimer and trimer bands are approximately 120 and 180 kDa, respectively. As shown in Figure 1, the A/California/07/ 2009 rHA protein displayed electrophoretic mobility comparable to that of the seasonal rHA H1 proteins. The lack of protein bands with molecular weight higher than that of HA0 under reducing conditions indicates that the oligomeric forms observed under non-reducing conditions were disulfide-linked. The A/California/07/ 2009 HA0 band migrated slightly faster than the HA0 comparators, possibly because A/California/07/2009 HA protein has only a single glycosylation site in the globular head (total five glycosylation sites) in contrast to the nine sites identified in the HA from A/Brisbane/59/ 2007 H1N1 virus [14,15]. In addition, less cleavage of A/ California/07/2009 HA0 into HA1 and HA2 was observed.
A/California/07/2009 rHA protein eluted as a single peak prior to a thyroglobulin molecular weight standard (~670 kDa) similar to other rHA proteins when analyzed by HPLC-SEC (Figure 2) Figure 4. A/California/07/2009 rHA formed multimeric rosette-like structures consistent with the other purified rHA proteins but were less distinct compared to the H3 rHA protein, which formed clearer rosette-like structures than either the H1 or B rHA proteins. All of the rHA rosette-like structures measured approximately 30 -40 nm in size, consistent with the measurements determined by HPLC-SEC and DLS.
Purified A/California/07/2009 rHA protein showed increased sensitivity to trypsin compared to other rHA proteins. Typically, HA0 is cleaved into HA1 and HA2 subunits when treated with trypsin, and we found that A/California/07/2009 rHA was digested into HA1 and HA2 peptides in a similar manner ( Figure 5). However, we also found that A/California/07/2009 HA2 was additionally digested into two prominent peptide fragments of approximately 18 and 6 kDa (designated HA2a and HA2b, respectively), suggesting the presence of an additional trypsin proteolytic site.
In order to better characterize HA2a and HA2b and identify their cleavage site(s), the protein bands were isolated and subjected to N-terminal (Edman) sequencing. The amino acid sequencing results from the Edman analyses are shown in Table 1. The results confirm that purified A/California/07/2009 rHA is produced in its mature form (N-terminus sequence for HA0 and HA1) with a conserved trypsin cleavage site at Arginine position 324 [16]. (Note that amino acid numbering is based on full length sequence.) However, the HA2 polypeptide was found to possess an additional cleavage site at Lysine position 419 (and potential cleavage sites at Arginine-420, Lysine-426 and Lysine-427).
Alignment of the amino acid sequences of A/California/ 07/2009 with A/New Caledonia/20/99, A/Brisbane/59/ 2007 and A/Solomon Islands/03/2006 H1 HA proteins was generated to determine whether sequences and secondary structure predictions could explain the trypsin digest results ( Figure 6). The alignment results show that the three potential additional trypsin cleavage sites (i.e., Arginine-420, Lysine-426 and Lysine-427) are conserved among all four HA proteins; however, Lysine-419 is unique to A/California/07/2009. Therefore, Lysine-419 is the most likely primary protease digestion site that generates HA2a and HA2b. Interestingly, according to ExPASy proteomics tools, amino acids 377 -427 are predicted to form a coiled coil domain (indicated by a green over-line in Figure 6). A/California/07/2009 has six amino acid changes in this region, including Lysine-419, compared to the other HA proteins. These sequence differences could affect protein conformation and make this region more accessible to trypsin digestion.

Biological activity
A/California/07/2009 rHA protein was evaluated for functional activity by determining its hemagglutination activity using red blood cells (RBCs) and comparing it to the activities of the 2000-2010 seasonal H1 influenza strain rHAs, A/New Caledonia/20/99, A/Solomon Islands/03/2006 and A/Brisbane/59/2007. A prerequisite for hemagglutination activity is the formation of trimers and the organization of these trimers into higher order structures that can crosslink corresponding sialic acid Figure 1 Reducing and non-reducing SDS-PAGE of H1N1 rHA proteins. For each sample, the respective purified rHA protein was diluted to a concentration of 100 μg/mL in reducing or non-reducing SDS-PAGE sample buffer, and 1 μg was loaded per lane. The samples were separated using 4 -12% gradient Nu-PAGE gels and stained with Coomassie Blue. HA0 represents full-length rHA protein and HA1 and HA2 peptide fragments of HA0. Molecular weights of proteins are shown in kilodaltons. rHA proteins were produced by Protein Sciences Corporation.
receptors on cells. The A/California/07/2009 rHA demonstrated hemagglutination activity with guinea pig and turkey RBCs but not chicken RBCs (Table 2). This was most similar to the hemagglutination activity observed for A/Solomon Islands/03/2006 rHA with guinea pig and turkey RBCs.
A/California/07/2009 rHA protein activity was also analyzed using the SRID assay. This assay measures the potency of influenza vaccines via quantification of functional HA protein [17]. Four lots of A/California/07/2009 reference antigens corresponding to re-assortants X-181 and X-179A obtained from both the Center for Biologics Evaluation and Research (CBER) and the National Institute for Biological Standards and Control (NIBSC), and two lots of the A/California/07/2009 rHA (matching Genbank accession #ACP41953) were tested against three different antisera generated against hemagglutinin from A/California/07/2009 ( Figure 7). The antisera were obtained from NIBSC, CBER and Protein Sciences Corporation, and were generated using HA from egg, E. coli and BEVS-insect cell sources, respectively. Both the CBER and Protein Sciences antisera were experimental and produced against recombinant HA proteins (the CBER antiserum against the HA1 fragment [18] and the Protein Sciences antiserum against full length rHA). All of the antisera produced immunoprecipitin rings with all reference antigens and the A/California/07/2009 rHA protein (Figure 7), indicating that the rHA protein was antigenically comparable. However, the NIBSC antiserum generated against egg-derived antigen produced larger, more diffuse rings for the rHA ( Figure 7A) that corresponded to calculated potency values that were 2to 5-fold greater than the amount of purified rHA protein inoculated into sample wells (data not shown). In contrast to the NIBSC antiserum, more well defined rings for rHA were achieved using antiserum generated against recombinant antigens ( Figure 7B and 7C). Interestingly, both the reference antigens and two separate licensed A(H1N1)pdm09 monovalent vaccines (from Novartis and Sanofi Pasteur) reacted differently with the three antisera, suggesting a unique interaction of each antiserum with each hemagglutinin produced from the A/California/07/2009 pandemic H1N1 virus. These differences had a significant impact on the calculated potency of the commercial vaccines (Table 3) and demonstrate the need to have well-matched reagents for the pandemic H1 vaccine antigens in manufactured products.
Finally, trivalent formulations of purified rHA vaccine corresponding to the 2008-2009 and 2010-2011 seasonal influenza strains, the latter of which contained A/ California/07/2009 rHA, were prepared to compare the immune responses of the different vaccine components. A commercial egg-based 2009-2010 vaccine (FluLaval, GSK, Lot # AFLLA599BA, multi-dose formulation) was included as a control. CD-1 mice were administered two doses of the respective formulations at 21 day intervals, and hemagglutination inhibition (HAI) antibody titers were determined three weeks after each dose. The immunogenicity results are provided in Table 4.
There was a clear dose dependence of the HAI response to all vaccine components across the test formulations, and the magnitude of the response increased from Day 21 to Day 42 after the second immunization. The immune response generated against A/California/ 07/2009 (H1) rHA antigen was equal to or slightly greater than that of A/Brisbane/59/2007 (H1) rHA antigen and the commercial egg-based A/California/07/2009 (H1) control by Day 42. The immune responses were also consistent for the H3 and B vaccine antigen components. These results demonstrate that rHA antigens, including A/California/07/2009 rHA, produce a robust immune response.

Discussion
The A(H1N1)pdm09 influenza vaccine was available late and in short supply as a result of delays in production caused by low yields, poor antigen stability and absence of virus stockpile. Recombinant hemagglutinin-based vaccines are inherently less susceptible to production challenges and are a leading alternative for influenza vaccine manufacture. The most advanced recombinant influenza vaccine candidate is a trivalent formulation of seasonal rHA proteins that can be produced significantly faster than traditional egg-based influenza vaccines and has been shown to be as effective and safe as eggderived trivalent inactivated vaccine (TIV) in human    in a bacterial expression system and found that the purified protein migrated predominantly as a monomer [18]. The reason for this difference is unknown but could relate to the different expression platforms. A/California/07/2009 rHA demonstrated significant biological activity and elicited a strong immune response in mice consistent with that generated by commercial egg-derived A/California/07/2009 vaccine both in this study and previously [19,20]. Together, these data support the suitability of A/California/07/2009 rHA as a pandemic influenza vaccine alternative. An initial clinical study has confirmed the safety and immunogenicity of this rHA [21].
The antigenic stability of the A(H1N1)pdm09 vaccine was found to be initially poor [13]. Trypsin digestion of purified A/California/07/2009 rHA uncovered a unique susceptibility of the protein to proteolytic cleavage not found in the seasonal rHA comparators. N-terminal (Edman) sequencing revealed that this cleavage occurs in a subdomain of the HA2 region of the protein that for most HA proteins are predicted to be structured as a coiled-coil. We postulate that the six amino acid changes in this domain in A/California/ 07/2009 HA may disrupt this structure, leading to decreased antigenic stability. Further studies are needed to determine whether the virus re-assortants ultimately used for eggbased A(H1N1)pdm09 vaccine manufacture possessed modifications that impacted protein structure in this region, improving antigen stability. Preliminary assessment suggests that a purified recombinant rHA derived from re-assortant virus NIBRG-121xp [7] in fact remained unstable although Hemagglutination values were generated with guinea pig, chicken, and turkey RBCs and HA activity was calculated based on protein BCA (Bicinchoninic Acid) values. its interaction with sialic acid receptors was improved (data not shown). Finally, the antigenic potency of A/California/07/2009 rHA, as determined by the SRID assay, showed dramatic heterogeneity (≥ 2-fold) depending on the assay reagents used. A similar effect was observed for licensed egg-derived monovalent vaccines. This variation indicates that preparation of reagents for potency testing (antisera and reference antigens) with novel pandemic influenza viruses requires further assessment to accommodate recombinant manufacturing strategies available for rapid pandemic response. Moreover, the development of alternative potency assays that are less dependent upon specific antigen-antibody interactions that could be affected by the manufacturing platform is warranted.

Conclusions
These results show that the production of purified recombinant hemagglutinin (rHA) subunit vaccines via the baculovirus expression system is a leading alternative platform for influenza vaccine manufacture. The biochemical, biophysical and immunological characterization of a purified recombinant A/California/07/2009 (H1N1) hemagglutinin has been compared to different seasonal rHA proteins and eggproduced A/California reagents. The data show that purified A/California/07/2009 rHA molecules exist in high molecular weight complexes and form rosette-like particles of 20 -40 nm in size. Biochemically, the protein exhibits hemagglutination activity and a greater sensitivity to tryptic digestion with additional cleavage in the HA2 subunit. The unique structure of this particular HA antigen may account for poor stability.  Further details on the cloning and expression of other rHAs using this system are described elsewhere [9,22,23].

rHA protein production
The recombinant baculovirus stock was used to produce the H1 A/California/07/2009 rHA protein. Virus inoculum from the working virus stock was added to 450 L of SF+ insect cell culture in a 600 L bioreactor at a concentration of 2% (v/v) after the insect cells reached a density of 2.0 -2.5 × 10 6 cells per mL. The infected culture was incubated at 28°C for 40 -55 hours and harvested at a viability of 70 -80%. A cell pellet was generated by centrifugation and the recombinant protein was solubilized using a buffer containing non-ionic detergent. Cells were removed by depth filtration, and the clarified extract was applied to an ion-exchange column. Recombinant HA was eluted and subsequently bound to a hydrophobic interaction column. Following elution, the protein was applied to a Q-membrane to remove any residual DNA. Finally, Q filtrate was diafiltered and the rHA protein formulated in final buffer. Single radial immunodiffusion (SRID) assay SRID assays were performed as described previously [17,24]. Briefly, an antibody solution at the optimal working concentration was mixed with melted 1% agarose (Cat# 50010, SeaKem ME, Lonza, Rockland, ME ) in 1x PBS (pH 7.

Animal immunization and HA inhibition (HAI) assays
The Animal Core Facility at Colorado State University (CSU) conducted the immunization and determination of HAI titers. Briefly, 6 -8 week old female CD1 mice were administered trivalent vaccine formulations containing purified recombinant H1 A/California/07/2009 rHA, H1 A/Brisbane/59/2007 rHA, or an A/California/07/2009 egg-based commercial vaccine in trivalent formulation (FluLaval W , GSK, Lot # AFLLA599BA, multi-dose formulation) by intramuscular (IM) injections. The formulated doses were based on SRID.
For HAI titer determination, individual serum samples were treated with receptor destroying enzyme (RDE, from Vibrio cholera Denka-Seiken, Tokyo, Japan) to remove nonspecific inhibitors and tested against 4 hemagglutination units (HAU) of the respective influenza viruses grown in eggs using 0.5% chicken RBCs as previously described [25]. All serum samples were tested in duplicate at a 1:10 starting dilution. The HAI titer was defined as the reciprocal of the greatest dilution that completely inhibited the agglutination of the chicken RBCs. A titer value of 5 was assigned to represent responses below the assay detection limit. All work was performed ethically in compliance with all federal, state, and local laws, regulations, and policies, as well as Colorado State University internal policies [http://web. research.colostate.edu/ACP/Regulations.aspx].

Competing interests
All authors work for Protein Sciences Corporation, which has a financial interest in recombinant influenza vaccine.
Authors' contributions EF carried out H1N1 process development, protein purification, rabbit antibody production, hemagglutination assay, trypsin digestion, Edman sequencing and helped to draft the manuscript; DR carried out Dynamic Light Scattering, participated in design of the animal study, EM and helped to draft the manuscript; RF drafted the manuscript and revised the paper; CM coordinated the HPLC-SEC, SRID assay and helped drafting the manuscript; JAR and PP designed and coordinated the study and helped to draft the manuscript; MMJC supervised the project, participated in its experimental design and data interpretation, and was responsible for writing the manuscript. All authors read and approved the final manuscript.