Vaccine 32 (2014) 2225–2230

Contents lists available at ScienceDirect

Vaccine journal homepage: www.elsevier.com/locate/vaccine

Stability of neuraminidase in inactivated influenza vaccines Ishrat Sultana a , Kevin Yang a , Melkamu Getie-Kebtie b , Laura Couzens a , Lewis Markoff a , Michail Alterman b , Maryna C. Eichelberger a,∗ a Division of Viral Products, Office of Vaccine Research and Review, Center for Biologics Evaluation and Research, U.S. Food and Drug Administration, Bethesda, MD, USA b Division of Cellular and Gene Therapies, Office of Cell, Tissue and Gene Therapy, Center for Biologics Evaluation and Research, U.S. Food and Drug Administration, Bethesda, MD, USA

a r t i c l e

i n f o

Article history: Received 8 October 2013 Received in revised form 4 January 2014 Accepted 28 January 2014 Available online 6 March 2014 Keywords: Influenza Vaccines Neuraminidase Stability Immunogenicity

a b s t r a c t Influenza vaccines are effective in protecting against illness and death caused by this seasonal pathogen. Antibodies that block the function of either hemagglutinin (HA) or neuraminidase (NA) contribute to vaccine efficacy, however vaccine potency is based only on HA content. NA protein content in vaccines varies from season to season due to differences in the relative amounts of HA and NA in influenza A, H1N1 and H3N2, and influenza B viruses that are selected for each manufacturing campaign. This, as well as potential inherent differences in NA immunogenicity, may result in varying responses from year to year. Moreover, the antigenic stability of NA is likely to dictate whether similar antibody responses will be obtained to this antigen throughout the shelf-life of the vaccine. To address this factor, we subjected NAs of influenza A (subtypes N1 and N2) and B viruses to denaturing conditions to evaluate the stability of enzyme activity. Each NA type/subtype had unique sensitivity to denaturing conditions. The N2 enzyme activity was more thermostable than that of N1 or influenza B, while the NA activity of influenza B was most resistant to detergent. N1 enzyme activity was most resistant of the three NAs to freeze–thaw cycling. In these experiments, enzyme activity was indicative of the immunogenicity of NA, but was strain-dependent, with greater neuraminidase inhibiting (NI) antibody titers elicited following immunization with the 2009 H1N1 pandemic virus A/California/7/2009, than the previously circulating seasonal H1N1 strain, A/Brisbane/59/2007. Robust NI antibody titers against both N1 and N2 components were induced following vaccination of mice with a trivalent inactivated influenza vaccine. When stored under recommended conditions, the NA of both N1 and N2 subtypes remained immunogenic well after the vaccine expiry date. Published by Elsevier Ltd.

1. Introduction There are several types of influenza vaccines licensed in the U.S. to date – a live attenuated quadrivalent vaccine; several inactivated split or subunit vaccines produced in eggs formulated to contain either three (trivalent) or four (quadrivalent) influenza viruses; an inactivated split trivalent vaccine produced in mammalian cells (Flucelvax), and a recombinant hemagglutinin (HA) influenza vaccine produced in insect cells (Flublok). The dose of each of these vaccines is standardized by the amount of HA. Except for Flublok, all vaccines contain a number of additional influenza proteins, including neuraminidase (NA), the second most abundant glycoprotein on the surface of the influenza virion. Antibody responses to HA and NA are independent correlates of immunity [1] and therefore

∗ Corresponding author. Tel.: +1 301 402 3846. E-mail address: [email protected] (M.C. Eichelberger). 0264-410X/$ – see front matter. Published by Elsevier Ltd. http://dx.doi.org/10.1016/j.vaccine.2014.01.078

the amount of NA present with capacity to elicit NA inhibiting (NI) antibodies could impact vaccine efficacy. While immunogenic amounts of NA are present in each of the inactivated influenza vaccines [2], the amounts cannot be standardized because there are different relative proportions of HA and NA for each constituent virus in a given vaccine preparation [3]. For example, since each vaccine dose aims to contain 15 ␮g of HA, the amount of NA per dose would be ∼7.5 ␮g for a virus with an HA:NA ratio of 2:1 but only 1.5 ␮g for a virus with an HA:NA ratio of 10:1. Clinical studies have demonstrated that 2.6 ␮g of purified NA is immunogenic in the majority of healthy young adults [4], and high dose vaccines increase responses to NA in the elderly [5]. Enzyme activity is a practical way to estimate the relative amount of immunogenic NA protein in monovalent vaccines or vaccine intermediates, because an intact native tetrameric structure is needed for NA activity and is also optimal for immunogenicity [6]. Since this measure is not specific for any particular NA subtype, it cannot be used to estimate the amount of the immunogenic form of each separate NA subtype in a formulated trivalent or

2226

I. Sultana et al. / Vaccine 32 (2014) 2225–2230

quadrivalent vaccine. Moreover, the absence of enzyme activity does not necessarily infer a lack of NA immunogenicity since native protein conformation is sufficient to induce NA inhibiting antibodies [7]. An assay that measures the quantity of NA with native conformation would be extremely helpful to evaluate the relative stability of the antigenic form of N1, N2 and influenza B NA components in multivalent vaccine formulations. Since such a test is not available, we have used both enzyme activity and immunogenicity in an animal model to examine the stability of the antigenic form of NA in split inactivated influenza preparations. A/California/7/2009 (H1N1), A/Victoria/210/2009 (H3N2) and B/Brisbane/60/2008 monovalent bulks (MVBs), as well as the formulated trivalent influenza vaccine (TIV) containing these components, were used in this study. 2. Methods 2.1. Virus and vaccine samples The following vaccines were used in this study: vaccine intermediates (monovalent and trivalent bulks) produced for the 2011/12 influenza season consisting of the high growth variants of the following recommended viruses: H1N1: A/California/7/2009 (CA/09), H3N2: A/Victoria/210/2009 (VIC/09), and B/Brisbane/60/2008 (B/BR/08); a monovalent 2009 H1N1 pandemic vaccine; and expired inactivated, split trivalent influenza vaccines produced for the 2008/09 influenza season. The samples were stored at 4 ◦ C. Heat-treated samples were incubated for various times at room temperature (25 ◦ C), or in a water bath at 37 or 50 ◦ C. Freezethawing was performed by rapid freezing on dry ice followed by thawing in a 37 ◦ C water bath. H6N1 and H6N2 viruses used in this study to measure NA inhibition titers were generated as described previously [7]. 2.2. Neuraminidase activity The activity of NA was measured using 4-methylumbelliferyl-Nacetylneuraminic acid (MU NANA, Sigma, St Louis, MO) as substrate. Briefly, serial 2-fold dilutions of sample (50 ␮l) were mixed with PBS in triplicate wells of a black 96-well plate, and an equal volume of 20 ␮M MU-NANA added. The plate was incubated at 37 ◦ C for 1 h, and then the reaction was stopped by addition of 100 ␮l of 0.1 M glycine, pH 10.7, 20% EtOH. The fluorescence was measured on a Victor V plate reader (Perkin Elmer, Shelton, CT), with excitation at 355 nm and emission at 460 nm. Assay variability was ≤10%. The NA of Vibrio cholerae (Sigma) was used as a positive control or to generate a reference standard curve in each assay. Specific activity was defined as the activity (units) per mg protein. In many assays the NA activity of sample post-treatment was compared to the activity of the original untreated sample by determining the percent relative fluorescence units at the equivalent sample dilution. 2.3. Determination of protein concentration Total protein concentration was measured using a BCA assay kit (ThermoScientific, Rockford, IL). A published label-free MSE method was used to quantify total amounts of HA and NA [3]. The concentration of NA in virus preparations used for immunogenicity studies was determined by isotope dilution mass spectrometry following a published method [8,9]. A cocktail of native and isotope labeled peptides used to generate a calibration curve for this analysis was kindly provided by Dr Tracie Williams (National Center for Environmental Health, Centers for Disease Control and Prevention, Atlanta, GA).

2.4. Immunogenicity of NA in mice BALB/c mice (6–8 week old, Jackson Laboratories, Bar Harbor, MA) were housed following federal regulations, with procedures approved by the institutional Animal Care and Use Committee. Each group (n = 5) was immunized intramuscularly (IM) with 50 ␮l of antigen. In some experiments this contained a 1:1 mixture of antigen and adjuvant (Titermax, Sigma). Mice were bled 3 weeks after vaccination from the tail vein, and the serum stored at −30 ◦ C until serologic assays were conducted. 2.5. Neuraminidase inhibition assay Neuraminidase inhibition (NI) titers were determined using an enzyme-linked lectin assay [10]. Briefly, H6N1 and H6N2 viruses containing the NA of the targeted antigen were titrated to determine the least amount of virus eliciting maximum signal. Serial dilutions of serum samples were mixed with an equal volume of virus and transferred to 96 well plates coated with fetuin. The plates were incubated for 16–24 h at 37 ◦ C, and then washed with PBS–0.05% Tween 20 before incubating for 2 h at room temperature with peanut agglutinin (PNA) conjugated to peroxidase (Sigma). The plates were washed prior to addition of substrate, O-phenylenediamine dihydrochloride (Sigma). After 10 min the reaction was stopped by adding 1 N H2 SO4 and absorbance read at 490 nm. All reagents used in this assay were purchased from Sigma. The inverse of the last dilution that resulted in ≥50% inhibition of NA activity was defined as the NI titer. 3. Results 3.1. NA content of vaccine intermediates and final product To compare the stability of NA of different virus types and subtypes in inactivated vaccines, we measured the enzyme activity of split, inactivated CA/09 (H1N1), VIC/09 (H3N2), and B/BR/08 viruses contained in monovalent bulks (MVBs), and the overall NA activity in formulated trivalent inactivated vaccine (TIV) bulk and in syringes filled with TIV. Enzyme activity and HA and NA concentrations were measured in three lots of each MVB and in three lots of trivalent formulated bulk (Table 1). The NA activities of lots for the same strain were very similar, but were significantly different between H1N1, H3N2 and B preparations, reflecting the amounts of NA measured by label-free MSE and small differences in NA specific activity. Comparison of the protein concentration and enzyme activity of monovalent bulk and trivalent material suggests that approximately 20-fold dilution of each MVB was used to formulate the vaccine. 3.2. The stability of NA is strain dependent One lot of each monovalent and trivalent bulk was selected to evaluate the stability of NA when subjected to increasing temperature, freeze–thaw or detergent treatments. The enzyme activity was measured after 0.5, 2, 7 and 24 h incubation at 4, 25, 37 and 50 ◦ C (Fig. 1A–D). It was stable at 4 and 25 ◦ C for 24 h for all samples. At 37 ◦ C, the NA activity was retained for at least 7 h in the monovalent bulks, but by 24 h the H1N1 monovalent bulk had only ∼50% of it original activity. A greater proportion of activity was retained in the H3N2 monovalent bulks (∼75% of initial activity). Interestingly, the trivalent formulated vaccine lost almost 50% of its activity at 37 ◦ C after only 30 min and very low amounts of NA activity were measured after 24 h at this temperature (Fig. 1D). Since MVBs are diluted to formulate the trivalent product, the total protein concentration of TIV is less than the MVB. The greater loss of activity in the TIV suggests that the reduction in total protein

I. Sultana et al. / Vaccine 32 (2014) 2225–2230

2227

Table 1 NA activity and absolute amount of NA and HA in monovalent bulks and formulated vaccines. Sample type Monovalent bulk H1N1 H3N2 B Formulated vaccine Trivalent bulk

Filled syringe

a b

Reference virus

HA conca (␮g/ml)

NA conca (␮g/ml)

CA/09 VIC/09 BR/08

1068 (±173.08) 1092 (±132.03) 1978 (±126.8)

424 (±60.9) 61 (±10.3) 110 (±12.7)

2.5 17.9 17.9

22 (±6.7) 20 24.4

CA/09 VIC/09 BR/08 CA/09 VIC/09 BR/08

51.6 (±6.7) 41.5 (±9.8) 95.2 (±9.3) 51.4 (±7.4) 39.0 (±9.1) 87.7 (±14.1)

HA:NA

NA activitya (U/ml)

Specific activity (U/mg)

75.97 13.90 22.76

179 228 207

2.179

NDb

1.528

NDa

NA activity was measured using MU-NANA as substrate as described in Section 2; HA and NA concentrations were measured by label-free MSE. ND, not done; the specific activity for each NA subtype cannot be calculated since the activity measured in the trivalent vaccine is not strain-specific.

concentration has a negative impact on NA stability. At 50 ◦ C the H1N1 and B monovalent bulks lost all NA activity within 30 min. In contrast, the H3N2 monovalent bulk retained 60% of its activity at this same time point, with a gradual loss in activity, culminating with none detected after 24 h at 50 ◦ C (Fig. 1B). While Ca2+ is reported to stabilize the NA tetramer [11], addition of Ca2+ to the diluent did not protect NA from the loss of enzyme activity (results not shown). However, evidence of the role of Ca2+ in stabilizing NA structure and consequently enzyme activity was provided by assays in which EDTA was added to each sample; addition of EDTA reduced NA activity of all monovalent and trivalent bulks when held at 25 and 37 ◦ C for 24 h (Fig. 1E). Interestingly, the activity of H3N2 and B monovalent bulks was reduced by EDTA even when

the sample was kept at 4 ◦ C, suggesting there may be differences in the strength of Ca2+ interaction among the three NAs. We next evaluated the impact of two anionic detergents, RapiGest (Waters) and deoxycholate (DOC) on NA activity. RapiGest is commonly used to denature membrane-associated proteins in order to enhance proteolytic digestion before analysis by mass spectrometry [12], and is not used during influenza vaccine manufacture. In contrast, DOC is a mild detergent that retains native structure of influenza glycoproteins when it is used to ‘split’ virus particles. Samples were incubated with 0.1% RapiGest for 30 min or 1% DOC for 24 h at room temperature and the NA activity compared with that of samples incubated without detergent (latter assigned as 100% activity). Incubation with RapiGest resulted in almost complete loss of H1N1 enzyme activity while approximately 20% of H3N2 activity was retained (Fig. 2A). Surprisingly there was no loss of B/BR/08 NA activity by this same treatment. The loss of NA activity in the formulated trivalent bulk was reflective of the activity lost in each individual vaccine component i.e. the ∼40% of initial enzyme activity that was retained in the presence of RapiGest is likely to reflect residual activity of the B component. In contrast, addition of DOC did not significantly reduce enzyme activity of any of the monovalent bulk or trivalent preparations even when they were held at 25 ◦ C for 24 h (Fig. 2B). To evaluate the impact of freeze-thawing on NA activity, H1N1, H3N2 and B monovalent bulks were frozen by placing aliquots on dry ice and then thawed at 37 ◦ C in a water bath multiple times. NA activity of each of the three monovalent bulks tested was reduced by freeze-thawing (Fig. 3A). The NA of H3N2 and B monovalent bulks were most sensitive to three freeze–thaw cycles, retaining about 30% of NA activity whereas the H1N1 monovalent bulk retained ∼80% of its NA activity when frozen and thawed three times. Interestingly, the activity of trivalent preparations rose after the first two freeze–thaw treatments and then decreased, dropping to ∼70% activity after five freeze–thaw treatments and losing all activity by 15 freeze–thaw cycles (Fig. 3B). A repeat experiment gave very similar results. 3.3. Immunogenicity of NA correlates with enzyme activity, but magnitude of the response is strain-specific

Fig. 1. Thermo-stability of influenza NA is strain-dependent. Three lots of (A) H1N1, (B) H3N2, (C) B monovalent bulks and (D) formulated trivalent vaccine were incubated for 0.5, 2, 7 and 24 h at 4, 25, 37 and 50 ◦ C, with aliquots stored at 4 ◦ C until NA activity was measured as described in Section 2. Each bar shows the average and standard deviation of three lots. (E) Addition of EDTA to monovalent bulks decreases NA thermostability. H1N1, H3N2 and B monovalent bulks (one lot per strain) were incubated overnight at 4, 25 and 37 ◦ C in the presence or absence of EDTA. NA activity was measured and results reported as the percent of initial activity. Each bar shows the average and standard deviation of three replicates.

Mice were immunized and boosted with two different H1N1 viruses: 2009 H1N1pdm virus CA/09 and A/Brisbane/59/2007 (BR/07), the previously circulating human seasonal H1N1 virus. The dose administered intramuscularly to each mouse contained equivalent units of NA activity (1600 ␮U), and similar amounts of NA (830 and 780 ng/mouse of NA from BR/07 and CA/09 respectively). The NI antibody titers were measured in sera of mice 3 weeks after each dose of BR/07 and CA/09. The titers were significantly different after primary immunization, suggesting that the NA of CA/09 is more immunogenic than the NA of BR/07 (Table 2). The antibody

2228

I. Sultana et al. / Vaccine 32 (2014) 2225–2230

A

140

Monovalent H1N1

% NA Activity

120

Monovalent H3N2

100

Monovalent B

80 60 40 20 0 0

B

1 2 3 5 Number of freeze-thaw cycles

300

15

Filled vaccine

250

%NA Activity

Bulk trivalent

200 150 100 50

Fig. 2. Loss of enzyme activity is dependent on type of detergent. H1N1, H3N2 and B monovalent bulks and a trivalent vaccine bulk were incubated at room temperature for (A) 30 min in 0.1% RapiGest or (B) 24 h in 1% deoxycholate (DOC). NA activity of each sample was compared with the activity of the original preparation stored at 4 ◦ C. Error bars shown on the graphs are standard deviations for the calculated percent of the original activity, with an asterisk above columns depicting activity that was significantly reduced by detergent treatment (t test, p < 0.05). The differences in activity between each subtype after RapiGest treatment was significant (ANOVA, p < 0.05). Differences after incubation in DOC were not significant.

response to NA was boosted in response to a second dose of each antigen, resulting in similar NI antibody titers. The immune response to the NA of CA/09 was dose-dependent, with the increase in NI titer correlating with increasing amounts of total protein and NA activity; even a single dose of CA/09 containing 16 ␮U NA induced NA inhibiting antibodies (Table 2). To confirm that the amount of NA with native conformation and not the absolute amount of NA correlates with immunogenicity [13], a monovalent H1N1 bulk preparation that had been frozen and thawed was used to immunize mice. Five cycles of freeze–thaw (F/T) significantly reduced enzyme activity and a dose comprised of the same antigen preparation that had been heated to 37 ◦ C for 24 h or heated to 100 ◦ C for 30 min was no longer immunogenic (Fig. Table 2 Immunogenicity of NA is strain-specific: NI antibody titers following immunization with CA/09 and BR/07 containing similar amounts of NA. Virus

None BR/07 CA/09 CA/09 CA/09

Dose/mousea

NI antibody titer

NA activity (␮U)

Total protein (␮g)

NA (␮g)

Post-dose 1

Post-dose 2

0 1600 1600 160 16

0 6.80 3.39 0.339 0.039

0 0.830 0.780 0.078 0.008