IV

VSN

ARRAA

KNIAV

1

cmnt(foiiss

b

r

U

0d

Journal of Virological Methods 173 (2011) 364–373

Contents lists available at ScienceDirect

Journal of Virological Methods

journa l homepage: www.e lsev ier .com/ locate / jv i romet

nfluenza virus-like particles as a new tool for vaccine immunogenicity testing:alidation of a neuraminidase neutralizing antibody assay

ictor Gavrilov, Tatyana Orekov, Casper Alabanza, Udayasree Porika, Hua Jiang, Kevin Connolly1,teven Pincus ∗

ovavax, Inc., 9920 Belward Campus Drive, Rockville, MD 20850, USA

rticle history:eceived 9 September 2010eceived in revised form 3 March 2011ccepted 9 March 2011vailable online 16 March 2011

eywords:euraminidase neutralizing antibody

a b s t r a c t

Detection of neutralizing antibody to viral neuraminidase (NA) by testing for enzyme inhibition has beenrecognized as an important part of the immunogenicity of influenza vaccines. However, the absence ofa well characterized standard source of active NA and validated assays has significantly limited clinicalstudies of NA immunity. Influenza virus-like particles (VLPs) containing hemagglutinin (HA), NA, andM1 proteins were produced from insect cells infected with a recombinant baculovirus and used as theNA source for the NA inhibition (NAI) assay. The NA activity of 6 different VLP strains varied from 0.43to 1.61 (×10−3) enzyme units per �g of HA and was stable over 6 months of storage at 2–8 ◦C. The

′

nfluenza virus-like particlessay validationaccine immunogenicity

NAI assay using 2 -(4-methylumbelliferyl)-�-d-N-acetylneuraminic acid as a substrate was modified fortesting the antibody titer in clinical samples and validated. The advantages of the assay include: (1) stable,reproducible, and standardized source of NA; (2) testing the antibody titer specific to each subtype of NAin serum from subjects immunized with trivalent vaccines (H1N1, H3N2, B) with no interference fromantibodies specific to the HA and to heterologous subtypes of the NA; (3) suitability for conducting long-term clinical trials as a result of low intra- and inter-assay variability, and (4) a wide analytical range due

value

to 25% inhibition cut-off

. Introduction

The immunogenicity of influenza vaccines is evaluated histori-ally by measuring antibody (Ab) titer to hemagglutinin (HA), theajor virus surface glycoprotein. Antibody to enzymatically active

euraminidase (NA), the other major surface glycoprotein, essen-ially enhances protective immunity induced by influenza vaccinesSylte and Suarez, 2010; Sylte et al., 2007). NA removes sialic acidrom both viral and host proteins and participates in the releasef viruses from infected cells. Only functional antibodies whichnhibit the enzyme are protective and provide so-called permissivemmunity: they do not prevent viral infection by themselves but

ignificantly reduce virus spreading throughout the body and theeverity of disease (Johansson et al., 1989; Sylte and Suarez, 2010).

Although the concept of an improved influenza vaccine withoth the HA and active NA proteins present together has been

Abbreviations: NA, neuraminidase; NAA, neuraminidase activity; NAI, neu-aminidase inhibition; NIT, neuraminidase inhibition titer; VLP, virus-like particle.∗ Corresponding author. Tel.: +1 240 268 2032; fax: +1 240 268 2132.

E-mail address: spincus@novavax.com (S. Pincus).1 Present address: PharmAthene, One Park Place Suite 450, Annapolis, MD 21401,SA.

166-0934/$ – see front matter © 2011 Elsevier B.V. All rights reserved.oi:10.1016/j.jviromet.2011.03.011

for the NAI titer estimation.© 2011 Elsevier B.V. All rights reserved.

discussed (Sylte and Suarez, 2010), no commercial vaccine withcontrolled NA activity is produced today. The lack of control for NAactivity (NAA) in influenza vaccines could be explained by a signif-icant loss of enzyme activity during storage (Kendal et al., 1980).The problem with NA instability has also limited testing of anti-NAimmune response during clinical studies.

The principle of testing NA inhibition (NAI) by antibody includesthe comparison of NA activity in samples incubated without andwith NA-specific immune serum (Cate et al., 2010). There are sev-eral well-established methods for testing the NAA in viruses andthese methods are used for monitoring enzyme activity in vac-cines during manufacturing (Kalbfuss et al., 2008) or screeningmutants resistant to chemical NA inhibitors (Gubareva et al., 2002;Wetherall et al., 2003). For these applications the highly variableNA activity in virus samples is expected and the NAA level repre-sents the object of analysis. In contrast, the neutralizing antibodyassay requires a stable and reproducible source of active NA with aspecified absolute activity (enzyme units/ml) and storage life as forany other analytical enzymatic reagent. Thus, existing publications

discussing NA immunity are able to present scientific proof of con-cept for NA neutralizing antibody but do not provide comparableresults for antibody titers and the activity of the used source of NA.

Using live influenza viruses for analytical purposes remains verycomplicated due to their pathogenicity (especially for pandemic

logica

vs2peuFlbrsvHBHooac

cfcessdcatpqhV

�spbetdsttiaharr

aaaevI

2

2

h

V. Gavrilov et al. / Journal of Viro

irus strains), laborious and expensive preparation, and lack oftandardization. Therefore, an inactivated virus vaccine (Cate et al.,010) and a purified NA protein (Johansson et al., 1989) wereroposed as a more convenient source of neuraminidase. How-ver, the conditions required for adequate measurement of NAIsing these substitutes of live virus, have not been established.or inactivated vaccine preparations which most often have a veryow level of NA activity (Kendal et al., 1980), and a high contri-ution of unfolded (denatured) proteins (Feng et al., 2009), theequirements for the NA qualification may include the enzymetabilization for long-term storage and setup of the range pro-iding NAA results relevant to those in live viruses. The active1N1 and H5N1 neuraminidases can now be purchased from Sino-iological (Beijing, China) and RnD Systems (Minneapolis, USA).owever, these preparations may contain unfolded NA moleculesr reduced associated forms (dimeric or monomeric) of NA insteadf a tetrameric complex present in virus (Sylte and Suarez, 2010)nd their binding properties to antibodies could thus be alteredompared to live virus.

Recent progress achieved in creating influenza virus-like parti-les (VLPs) that contain the HA and NA viral proteins with relevantunctional activity and immunoreactivity makes VLP an excellentandidate to substitute for live viruses (Kang et al., 2009; Pushkot al., 2005; Lai et al., 2010). The unique combination of comparabletructural and biological properties to influenza viruses, enhancedtability and ease of handling gives VLPs a big advantage over otheriscussed virus substitutes. This allows for expanding VLP appli-ations outside of vaccine development into the study of differentreas of virus properties (Kang et al., 2009; Lai et al., 2010). Unfor-unately, NA in VLPs has not been fully characterized: only a fewublications have reported detectable NA activity without properuantification in absolute units (Lai et al., 2010). Furthermore, itas not been demonstrated that the activity and stability of NA inLPs are sufficient for analytical applications.

To test NA activity, the fluorogenic 2′-(4-methylumbelliferyl)--d-N-acetylneuraminic acid sodium salt hydrate (MUNANA)ubstrate was chosen. This is the only substrate which gives com-arable results for NA activity from multiple sources as the assay isased on quantification of the enzymatic reaction product (Potiert al., 1979; Wetherall et al., 2003; Kalbfuss et al., 2008). To adjusthe MUNANA-based NAA assay for testing NAI titer, we need toetermine (1) the acceptable range of initial NA activity in VLPamples that would provide consistent results for a given antibodyiter; (2) whether VLP could be applied for testing the antibodyiter specific to only one subtype of NA in the serum from subjectsmmunized with a trivalent vaccine with no interference from thentibody specific to HA and heterologous subtypes of NA; and (3)ow the assay performance depends on the NAI cut-off value for thentibody titer estimation and what is the optimal cut-off value. Cur-ently, no MUNANA-based NAI assay has an established analyticalange nor has been validated.

Therefore, the objective of this study was to qualify the VLP asnew source for active neuraminidase in the NAI-based antibody

ssay. To optimize the assay, the range for the initial NA activitynd the optimal NAI cut-off value for antibody titer estimation werestablished. To set up an analytical range, the modified assay wasalidated for specificity, precision, accuracy, and linearity as perCH guidelines (ICH, 1996).

. Materials and methods

.1. VLP preparation

Purified VLPs are comprised of recombinant influenza virusemagglutinin (HA), neuraminidase (NA) and matrix 1 (M1)

l Methods 173 (2011) 364–373 365

proteins. The appropriate genes are cloned into a baculovirusexpression vector with each under the control of their ownset of transcription regulatory elements. VLPs are then assem-bled from recombinant HA, NA, and M1 proteins expressed inrecombinant virus infected Spodoptera frugiperda (sf9) insectcells (ATCC CRL-1711) (Invitrogen, Carlsbad, USA). Abbrevia-tions for influenza viruses or proteins used in this study areA/Indo (A/Indonesia 05/2005 (H5N1)), A/Cal (A/California 04/2009(H1N1)), A/NewCal (A/New Caledonia/20/99(H1N1)), H1N1A/Br (A/Brisbane 59/2007 (H1N1)), A/NY (A/New York/55/2004(H3N2)), H3N2 A/Br (A/Brisbane 10/2007 (H3N2)), B/Br (B/Brisbane60/2008), B/Fl (B/Florida 04/2006) and B/Sh (B/Shanghai/361/2002).

The cloning of influenza HA, NA, and M1 into baculovirus(BV) expression vectors was described previously (Pushko et al.,2005; Bright et al., 2007; Mahmood et al., 2008). The sf9 insectcells were infected with a recombinant baculovirus. VLPs werepurified from the supernatant harvested from the sf9 cell cul-ture (Bright et al., 2007). After removal of the sf9 cells, theVLP preparation was concentrated and diafiltered through tan-gential flow filtration (TFF). The separation of VLPs from BVparticles and contaminating DNA, RNA and sf9 proteins wasachieved by anion exchange (IEX) followed by size-exclusion (SEC)chromatography. To ensure sterility the VLP preparation was fil-tered through a 0.22 �m PVDF membrane and stored refrigerated(2–8 ◦C).

The expression of HA, NA, and M1 proteins in VLP preparationswas confirmed by SDS–PAGE and Western blot analysis and the HAconcentration was quantified by a single-radial-immunodiffusion(SRID) as described (Pushko et al., 2005; Mahmood et al., 2008).The total protein (TP) concentration was tested using the BCA assay(Pierce, Rockford, USA).

2.2. Purified neuraminidase and hemagglutinin proteinpreparations

NA and HA proteins from H5N1 A/Indo, H3N2 A/NY and B/Flviruses were produced by the following procedure. The corre-sponding native HA and secreted NA (fusion of NA with HAsecretion sequence) genes were individually cloned into the BVexpression vector used to produce VLPs. The sf9 cells wereinfected with the resulting recombinant viruses and the pro-teins were expressed in a similar manner as VLPs described inthis paper. The HA proteins were extracted from the cell pasteand purified through TMAE anion exchange, lentil lectin affin-ity chromatography, and hydroxyapatite (CHT) chromatography.The secreted NA proteins were harvested from the culture super-natant and purified through three column chromatography stepsincluding Fractogel EMD TMAE, lentil lectin, and hydroxyapatitecolumns. The purified HA and NA proteins with >95% purityas determined by SDS–PAGE, and densitometric scanning wereused for animal immunization to raise mono-specific polyclonalantiserums.

2.3. Antisera against purified neuraminidase and hemagglutinin

Sheep immunization with purified HA and NA proteins fromB/Fl, H3N2 A/NY and H5N1 virus strains, followed by blood col-lection and antiserum preparation were performed by Covance

(Princeton, USA). Sheep were immunized ID/IM with 50 �g of anti-gen in complete Freund’s adjuvant and boosted three weeks laterby the same route with 30 �g of antigen mixed with in-completeFreund’s adjuvant. The antibody titers were determined by ELISAas described below.

3 logica

2A

cfAgnowcrwL

2

Aipdeo(p

2

wla((0ISsteiaar0oGfoI(o

netcoaf

a

66 V. Gavrilov et al. / Journal of Viro

.4. Ferret antisera after immunization with/Brisbane/10/07(H3N2) virus

Male ferrets (Mustela putorious furo) 3–4 months of age andonfirmed to be seronegative against H3N2 A/Bris virus were usedor immunization (Bioqual Inc., Rockville, USA). Egg grown H3N2/Bris virus was produced as described (Bright et al., 2008) andiven intranasally to ferrets (n = 4) in a volume of 0.5 ml (250 �l perostril), under anesthesia. Each ferret was inoculated with a totalf 1 × 108 pfu/0.5 ml of virus. Twenty days after inoculation, bloodas collected from the anesthetized ferrets via the anterior vena

ava. The serum was separated and frozen at −80 ± 5 ◦C. The fer-ets were euthanized and exsanguinated on day 35. All proceduresere in accordance with the NRC Guide for the Care and Use of

aboratory Animals.

.5. Human antisera after vaccination with trivalent VLP vaccine

Seasonal trivalent VLP vaccine comprised of H1N1 A/Bris, H3N2/Bris, and B/Fl VLP strains recommended for the 2008–2009

nfluenza vaccine was used in a clinical trial (manuscript inreparation). Normal healthy adults (18–50 years old) were ran-omized to receive the trivalent VLP vaccine at 15 �g HA ofach strain (85 subjects), 60 �g HA of each strain (85 subjects)r placebo (50 subjects). Blood samples were collected beforeday 1) and after (day 22) immunization followed by serumreparation.

.6. Antibody titer determination by ELISA

Antibody (total IgG) titers in anti-HA and anti-NA sheep seraere determined by ELISA. Briefly, a 96-well microplate Immu-

on 2HB (Corning Inc., Lowell, USA) was coated with 50 �l/well of2.0 �g/ml VLP sample formulated in 25 mM tris-buffered saline

TBS), pH 7.2. The plate was incubated for 2 h at room temperatureRT) with shaking and washed with 25 mM TBS buffer containing.05% Tween 20 using a Microplate washer Biotek ELx405 (Bioteknc., Winooski, USA). The plate was blocked with 150 �l/well of theuperblock buffer (Pierce) for 30–40 min at RT. Six 2 or 4-fold serialerum dilutions starting from a 1:10 dilution were prepared usinghe Superblock buffer with 0.1% Tween 20. After blocking, 30 �l ofach serum dilution was added to each well in duplicate. After 1 hncubation, the plate was washed as described above followed byddition of 30 �l per well of 50 ng/ml peroxide-conjugated donkeynti-sheep immunoglobulin (Jackson Immunoresearch Laborato-ies, West Grove, USA) diluted in the Superblock buffer containing.1% Tween 20. After 1 h incubation followed by washing, 100 �lf the SureBlue Reserve TMB microwell peroxide substrate (KPL,aithersburg, USA) was added to each well. The plate was incubated

or 10–15 min and the reaction was stopped by adding 25 �l/wellf 1.0 M HCl. The plate was read at 450 nm on a microplate readernfinity M200 (Tecan, Mannedorf, Switzerland). The OD readingsOD450) were corrected by background subtraction and an averagef the duplicates was obtained.

The corrected mean OD450 values were plotted against theatural logarithm (ln) of the serum dilutions. A linear regressionquation was generated using the linear portion of the serumitration curve. The endpoint Ab titer was determined by back cal-ulating the serum dilution from ln (serum dilution) which wasbtained by extrapolating the linear regression line to the cut-off

bsorbance OD450 = 0.2. The calculation was performed using theollowing equation:

ntibody titer = exp[(ln D2 − ln D1) × OD1 − 0.2OD1 − OD2

+ ln D1]

l Methods 173 (2011) 364–373

where OD1 and OD2 represent OD values on the linear portion oftitration curve; D1 and D2 are the corresponding serum dilutions;0.2 – cut-off absorbance value.

2.7. Hemagglutination inhibition (HAI) assay

The validated hemagglutination inhibition (HAI) assay forhuman serum samples was conducted by a certified laboratory(Focus Diagnostics, Inc., Cypress, USA) using turkey red blood cells(RBC) and influenza viruses representing each strain in the trivalentvaccine as source of agglutinating HA. The HAI titer was deter-mined by the reciprocal of the last serum dilution which containednon-agglutinated RBCs. The HA-specific immune response was esti-mated by post-to-pre ratio (PPR) of HAI titers after (postimmune)and before (preimmune) vaccination. For each separated donorgroup the geometric mean ratio (GMR) as a geometric mean ofindividual PPR values was calculated.

2.8. Neuraminidase inhibition (NAI) assay

The NAI inhibition assay involves testing NA activity in VLPsamples after their incubation with serial dilutions of a serumcontaining potential NA-specific antibodies and calculating theresidual NAA to the activity in the absence of serum. The NAA wasmeasured by a fluorometric assay with 2′-(4-methylumbelliferyl)-�-d-N-acetylneuraminic acid sodium salt hydrate (MUNANA) asa substrate and fluorometric detection of the fluorescent reac-tion product 4-methyl umbellipherone (MU) (Potier et al., 1979;Wetherall et al., 2003).

2.8.1. Reagent preparationStock solutions of the 100 mM MUNANA (Gold Biotechnology

Inc., St. Louis, USA) and the 2.0 mM MU (Sigma–Aldrich, St. Louis,USA) were prepared in the HyPure water (Fisher Scientific, Pitts-burgh, USA) in amber glass vials. Assay buffer 1 (AB1) contained32.5 mM MES sodium salt, 4.0 mM CaCl2 and 0.1% Tween 20 fromSigma with the pH adjusted to 6.5. The assay buffer 2 (AB2) wasprepared from AB1 by adding 60 �g/ml of bovine serum albumin(BSA) from Pierce. The stop solution contained 0.1 M glycine and25% ethanol from Fisher adjusted to the pH 10.7. The MUNANAstock solution aliquots were stored at −20 ◦C. All other reagentswere stored at 2–8 ◦C. Working solutions of 90 �M MU in AB1 and300 �M MUNANA in AB2 were freshly prepared prior to performingthe assay and used on the same day.

2.8.2. VLP sample preparationVLP samples were diluted 1:1 with the 2× stabilization buffer

containing 60 mM MES-Na, 8.0 mM CaCl2, 1.0 M NaCl (Sigma), 0.1%TPGS (alpha tocopheryl polyethylene glycol 1000 succinate) (Cog-nis, Cincinnati, USA) and 10% glycerol (Fisher). The samples werealiquoted in 1.5 ml cryovials and stored at 2–8 ◦C for up to sixmonths or at −20 ◦C for up to one year. Prior to the NAI test, VLPsamples were diluted with AB1 to obtain a final NA activity rangingfrom 1.0 to 2.0 nmol MU/well and used on the same day.

2.8.3. Assay procedureThe assay was performed in 96-well black microplates (Greiner

Bio-One, Monroe, USA). The MU standard solutions were preparedin AB1 and added to wells (90 �l/well) in duplicate to provide finalMU concentrations of 0, 0.1, 0.5, 1.0, 2.0 and 3.0 nmol/well.

The serum samples were stored frozen. Prior to the analysis, thesamples were thawed and placed at 2–8 ◦C for up to one week untiltested in the assay. To provide 0–50% NA inhibition, two serumdilution schemes were used: six 4-fold serial dilutions within therange 1:4–1:4096 for elevated antibody titers and six 2-fold serial

logica

dVtV4Vtifs

tS4

2

tfsmuaac

N

westc

pl

rwpat

Ft(B(idC

V. Gavrilov et al. / Journal of Viro

ilutions within the range 1:2–1:64 for low antibody titers. TheLP sample (30 �l /well) was mixed with 30 �l of each serum dilu-

ion in duplicates or with 30 �l of AB1 in four replicates for theLP control sample. The plate was incubated at RT with shaking for0 min. 30 �l of 300 �M MUNANA was added to wells containingLP samples or 60 �l of AB1 (substrate blank) to a final concentra-

ion of 100 �M. The plate was incubated at 37 ◦C with shaking on anncubator–shaker, Jitterbug-4 (Boekel Scientific, Feasterville, USA)or 40 min. The reaction was stopped by adding 150 �l/well of stopolution to all wells.

Fluorometric measurements were performed immediately withhe Modulus Microplate Multimode Reader (Turner Biosystems,unnyvale, USA) at an excitation of 365 nm and an emission of10–460 nm and expressed in relative fluorescence units (RFU).

.8.4. Calculation of NA inhibition titer (NIT)A standard curve RFU versus MU and a linear regression equa-

ion were generated using readings from the standards correctedor blank fluorescence. The corrected sample RFU was obtained byubtracting the average substrate blank RFU from the measuredean sample RFU. The NA activity of the samples was calculated

sing the linear regression equation obtained for the MU standardsnd expressed as an amount of the product (nmol/well of MU)ccumulated during the incubation time. In some cases, NAA wasonverted from nmol/well to mU/ml using the equation

AA (mU/ml) = MUx (nmol/well)DF

Vx × Tincub= 0.56 × MUx × DF

here MUx is the amount of MU product accumulated over thenzymatic reaction; DF is a dilution factor or a final dilution of VLPample; Vx = 30 �L volume of VLP sample; Tincub = 40 min incuba-ion time; mU is milliUnit of enzyme activity corresponding to theonversion of 1 nmol substrate to product in 1 min.

The inhibition curve was generated as a % Residual NAA com-ared to the initial NAA for the VLP control lacking serum versus

og2[serum dilution] as shown in Fig. 1.The optimized procedure included a calculation of NAI titer cor-

esponding to 25% of inhibition, or 75% of residual NAA whichas designated NIT25 or just NIT. For its estimation, the linearart of the inhibition curve around 75% of the residual NAA waspproximated by the linear regression followed by the calcula-ion of log2[serum dilution] corresponding to 75% of residual NAA.

0

20

40

60

80

100

120

14121086420

log2[Serum Dilution]

% R

esid

ual

NA

A

Immunized serum

preimmunized

serum

log2(NIT2)log2(NIT1)

ig. 1. Calculation of log2 NIT using NA inhibition curves. NAI curves were plot-ed as residual NA activity (NAA%) versus log2[serum dilution] for preimmunizedlog2 (NIT1)) and immunized (log2 (NIT2)) sheep serums using purified NA from/Florida virus as an immunogen. NAI assay was performed with B/Florida VLPslot 75508010). Solid lines indicate linear regression lines generated using part ofnhibition curves around % residual NAA = 75%. Dashed lines represent graphicaletermination of log2 NIT corresponded to 25% of NA inhibition for both curves.alculated values for log2 (NIT) were log2 (NIT1) = 1.72 and log2 (NIT2) = 7.9.

l Methods 173 (2011) 364–373 367

The NIT was reported as the serum dilution back calculated fromantilog2 [NIT]. Any log2 NIT ≤ 0 was assigned a value of 0 and cor-responded to NIT = 1.

During assay development the NAI titer was also estimatedat two additional cut-off levels. The log2 NIT50 representing 50%of inhibition was calculated using the same inhibition curve.The log2 NIT10 representing approximately 10% of inhibition wasdefined based on Sylte et al. (2007) as one log2 serum dilution belowthe log2 dilution corresponding to NAI ≤ 5%:

log2 NIT10 = log2 NIT5 − 1

2.9. Qualification of active NA in VLPs

Seven VLP samples containing different subtypes of NA wereprepared in the stabilization buffer and tested for NAA, total proteinand HA concentration to obtain the results comparable with NAAmeasurements in live influenza viruses, purified NA proteins, andinactivated virus vaccines. Stability of NAA in the VLP was analyzedas % residual NAA after storage at 2–8 ◦C for 6 months.

2.10. Standardization of NA (in activity units) for antibodytitration

Since testing an antibody titer requires a constant antigen con-centration, it is important to establish an acceptable range for theinitial NA activity in VLP samples that provides consistent neutral-ization antibody titer results. To determine this range, four dilutionsof H3N2 A/Bris VLP sample with initial NAA = 3.21 nmol/well ofMU were prepared as a neat sample, 2-fold, 3-fold and 4-folddilutions. Each VLP sample was tested for the NAI titer with anH3N2-NA(A/Bris)-specific antiserum and a bias between log2 NITvalues for neat and diluted samples was calculated.

2.11. Assay specificity

To compare the NAI in the VLP sample by antibody specific tohomologous and heterologous NA and HA proteins, the B-Fl VLPwas assayed for NIT using sheep serum samples specific to NA(B/Fl), HA(B/Fl), NA(H3N2 A/NY) and HA(H3N2 A/NY). To evalu-ate any interference effect of the antibody to homologous HA andheterologous NA and HA on the specific NA inhibition, differentcombinations of the NA(B/Fl)-specific antiserum and antisera spe-cific to other NA and HA proteins were prepared and tested forNAI titer using a B-Fl VLP sample. To analyze NA subtype speci-ficity, the antiserum specific to H5N1-NA (A/Indo) was tested forNIT against homologous (H5N1 A/Indo) and heterologous (H1N1A/NewCal, H3N2 A/NY and B/Sh) VLPs.

To confirm that NAI titer measured in sera from subjects vacci-nated by a trivalent vaccine is due to an immune response to eachNA protein and not a cross-reactive antibody elicited by a single NA,serums from 81 subjects were tested for NIT before (preimmune)and after (postimmune) vaccination with a seasonal trivalent VLPvaccine containing H1N1 A/Bris, H3N2 A/Bris, and B/Fl influenzastrains. The anti-NA immune response was tested for B and N2subtypes of NA separately using B/Fl and H3N2 A/Br VLPs, respec-tively, and considered positive if post-to-pre ratio of NAI titers (PPR)

was ≥2.0. The distribution of subjects with a positive NAI responseagainst both NA subtypes and against only one NA subtype was ana-lyzed. Using NAI and HAI assays, the geometric mean ratio (GMR)of post and preimmune Ab titers against NA and HA for each groupwas calculated.

368 V. Gavrilov et al. / Journal of Virological Methods 173 (2011) 364–373

Table 1Characterization of neuraminidase in VLPs.

Influenza virus strain VLP lot # Initial NAA mU/ml NAA/TP mU/mg NAA/HA mU/ug % Residual NAA after 6month storage at2–8 ◦C

B/Florida 04/2006 75508010 456 1.23 0.63 98.4B/Brisbane 60/2008 714.22 88.0 0.14 1.18 100A/Brisbane 10/2007 (H3N2) 75508011 426 1.18 0.55 91.1A/Brisbane 10/2007 (H3N2) 081208 188 0.58 1.61 102A/Brisbane 59/2007 (H1N1) XX10F001 480 1.20 0.73 96.2A/California 04/2009 (H1N1) 09M001 250 0.08 1.44 87.6A/Indonesia 05/2005 (H5N1) 090110 75.0

NA activity (NAA) was expressed as milli-units (nmol/min) per ml, mg of total protein (TPEach number presented mean value for 3 replicates with CV < 2.0%.

Table 2Consistency of NAI titer tested at different VLP dilutions.

VLP dilution Control NAA (no serum)(NAA, nmol/well)

Log2 (NIT) Bias:diluted-neat

Neat 3.21 8.09 01:2 1.74 8.33 0.241:3 1.2 8.48 0.39

(a

2

(o

2

sw(Ta

N

w(d

a

l

fibutd

2

smbmTat

1:4 0.83 7.83 −0.26

VLP sample: A/Brisbane/10/2007 (H3N2) lot 75508011, antiserum sample: Ferretntiserum immunized with virus A/Brisbane/10/2007 (H3N2) virus).

.12. NAI assay validation

NAI assay validation was performed following ICH guidelinesICH, 1996) and included testing accuracy, precision, and linearityf the assay.

.12.1. AccuracyAccuracy was determined in a dilution experiment in which the

erum obtained from a ferret immunized with H3N2 A/Bris virusas serial diluted (2-fold) up to a 1:1024 dilution with a preimmune

naïve) serum and tested for NAI titer using the H3N2 A/Bris VLP.he expected NIT in diluted samples for both cut-off values NIT25nd NIT50 was calculated using the following equation:

ITdil = NITp

D+ NITn × D − 1

D

here NITdil, NITp and NITn are titers for diluted sample, positiveimmune) and negative (preimmune) serums; D is positive serumilution.

Expected log2 NIT10 for sample with dilution D was calculateds

og2 NIT10 (neat) − log2 D

To analyze any impact of different cut-off values on assay per-ormance, the accuracy was determined for all three levels of NAnhibition: 10% (log2 NIT10), 25% (log2 NIT25) and 50% (log2 NIT50)y calculating the bias between found and expected log2 NIT val-es. Accuracy was considered acceptable if the absolute value forhe calculated bias was ≤1.0 which corresponds to one 2-fold serumilution.

.12.2. Intra- and inter-assay precisionThe intra-assay precision was evaluated by testing three serum

amples with different antibody titers corresponding to low,edium, and high levels of log2 NIT. These samples were prepared

y diluting the anti H3N2-NA (A/Bris) sheep serum with a preim-une serum and each sample was assayed for NAI titer six times.

o evaluate the inter-assay variability, the NAI titer for the ferretntiserum derived by immunization with H3N2 A/Bris virus wasested by two analysts on two separate days using H3N2 A/Bris VLP.

0.11 0.43 84

) and �g of hemagglutinin (HA).

The intra-assay and inter-assay variation between log2 NIT results,including between-day and between-analyst repeatability, was cal-culated by ANOVA and is presented as the mean along with the %CVvalues. The %CV was acceptable if it was ≤10.0%.

2.12.3. Linearity and analytical rangeLinearity was analyzed using data obtained for accuracy. The

log2 NIT values measured from the assay were plotted against theexpected log2 NIT values and then subjected to linear regressionanalysis.

The linearity was acceptable if the correlation coefficient value(R square) was ≥0.90. The range of the NAI assay was establishedas the range of log2 NIT values that met the acceptance criteria foraccuracy, precision, and linearity.

2.13. Statistical analysis

All statistical analysis was performed by ANOVA using MicrosoftExcel software.

3. Results

The experiments reported here were designed to establish usingVLPs as an NA enzyme source and MUNANA as an enzyme substratefor reproducible measuring the NA neutralizing antibody in clinicaltrial samples.

3.1. Qualification of active NA in VLP samples

The NAA determined for seven VLP preparations with differentsubtypes of NA varied between 75 mU/ml and 480 mU/ml (Table 1).The relative NAA per mg of total protein was within the rangeof 0.08–1.23 mU/mg total protein and the relative NAA per �g ofHA was more consistent with a 3.7-fold variability from 0.43 to1.61 mU/�g HA. All VLP preparations demonstrated high stabilityof active NA upon storage for 6 months at 2–8 ◦C. % Residual NAAwas within the range of 84–100%.

3.2. Standardization of NA (in activity units) for antibodytitration

The relevant calculation of NAA in VLPs requires the fluores-cence measurement of MU within the analytical range which wasdetermined and found to be 0.1–3.0 nmol/well of MU. Four 2-foldserial dilutions of H3N2 A/Bris VLP with an initial NAA varyingfrom 3.21 to 0.83 nmol/well were prepared and tested for the NAI

titer with H3N2-NA(A/Bris)-specific antiserum. NAI curves and thecalculation of log2 NIT25 are shown in Fig. 1. The log2 NIT testedfor different initial NAA levels was between 7.83 and 8.48 witha bias which did not exceed 1.0 or one two-fold serum dilution(Table 2). Therefore, the NAA in VLPs was qualified within the range

V. Gavrilov et al. / Journal of Virological Methods 173 (2011) 364–373 369

Table 3NA inhibition and total IgG titers tested against B/Florida VLPs with different combinations of sheep antiserums specific to NA (B/FL), HA (B/FL), NA(H3N2 A/NY) and HA(H3N2A/NY) prepared by immunization with purified recombinant HA and NA proteins of B/Florida and H3N2 A/NY viruses.

Antiserum sample combination NA inhibition titer Total IgG titer againstB/Florida VLP (100 ngprotein/well)

Total IgG titer againstH3N2 A/NY VLP(100 ng protein/well)

Log2 NIT NIT adjusted to 100 ngprotein/well

% Relative reactivity toinhibition byNA-specific antiserum

B-NA 13.23 34,565 100.0 10,368 11NA inhibition by homologous HA and heterologous NA and HA antibodies

B-HA 2.09 15.4 0.04 35,900 1217H3N2-NA 3.35 36.6 0.11 4485 14,536H3N2-HA 3.72 47.3 0.14 14 35,916H3N2-NA + H3N2-HA 4.66 91.0 0.26 x xB-HA + H3N2-NA + H3N2-HA 5.18 130.6 0.38 x x

Impact of homologous HA and heterologous NA and HA antibodies on NA inhibition by homologous NA antibodyB-NA + B-HA 12.78 25,320 73.3 x x

109.6 x x101.8 x x

x

or

3h

tsBt(a

hatmlifiNH

baansaBfebH

iVinHNt(r

Table 4Cross-reactivity of H5N1 NA antibodies with different types of neuraminidase testedwith NA inhibition assay.

VLP NIT % Cross-reactivity

H5N1 A/Indonesia lot 022007 8462.7 N/A

subtypes, 25% of donors had a positive immune response againstonly one subtype of NA (20% had a response against B-NA and 5%against N2-NA), and 11% had no response to both B-NA and H3N2NA (Table 5). All volunteers responded to vaccination with at least

0

20

40

60

80

100

120

1816141210864

% R

esid

ual N

AA

H5N1 A/Indo

H1N1 A/NewCal

H3N2 A/NY

B/Shanghai

B-NA + H3N2-NA + H3N2-HA 13.36 37,892B-NA + B-HA + H3N2-NA + H3N2-HA 13.26 35,196

, not applicable.

f 2.0–1.0 nmol /well as acceptable for providing consistent NITesults.

.3. Specificity for testing NAI in the presence of homologous andeterologous NA and HA-protein specific antibody

The specificity of anti-NA and anti-HA sera was confirmed byesting for their total IgG titer by ELISA. All four antisera demon-trated very high specificity for homologous proteins: B/Fl-NA and/Fl-HA-specific sera had high titers against B/Fl VLP and very lowiters with H3N2 A/NY VLP. In contrast, H3N2 (A/NY)-NA and H3N2A/NY)-HA specific antisera had high titers with H3N2 A/NY VLPnd low titers with B/Fl VLP (Table 3).

A high NAI titer was detected when testing B/Fl VLP withomologous B-NA antiserum, while NITs for homologous B-HAnd heterologous H3N2-NA and H3N2-HA antisera were 735–2300imes lower. A mixture of H3N2-NA and H3N2-HA antisera or a

ixture of B-HA, H3N2-NA and H3N2-HA antisera resulted in aow total NIT with B/Fl VLP matching the sum of the NAI titers fromndividual serum samples. These results demonstrated the speci-city of the assay for homologous NA antibody due to very lowAI response on homologous B-HA and heterologous H3N2-NA and3N2-HA antibodies.

To test any interference effect, the B-NA antiserum was com-ined with different combinations of homologous B-HA antiserumnd heterologous H3N2-NA and H3N2-HA which simulates thentiserum from immunization with B/Fl VLP alone or with a combi-ation of B/Fl and H3N2 VLPs. The log2 NIT values obtained for theseimulated antisera with additional B-HA, H3N2-NA and H3N2-HAntibodies had no significant difference from log2 NIT tested for the-NA antiserum alone: the calculated bias between log2 NIT titers

or the original B-NA antiserum and any of its modification did notxceed 0.5. These results confirm the using VLPs for testing anti-ody specific only to homologous NA in the presence of homologousA and heterologous NA and HA antibodies.

Specificity of the NAI assay was also confirmed by testing NITn the serum specific to one subtype of NA (H5N1 A/Indo) againstLPs with different NA subtypes. This antiserum specifically inhib-

ted the homologous subtype of NA in H5N1 A/Indo VLP and didot inhibit other subtypes such as H3N2 A/NY and B/Fl (Fig. 2).

owever, when assayed against VLP containing another strain of1 subtype NA (H1N1 A/NewCal) a reduced but measurable inhibi-

ion titer was detected. To quantitate the cross-reactivity of anti-NAH5N1 A/Indo) antibodies towards different subtypes of NA theatio of the NAI titer against a heterologous VLP to the NIT against

H1N1 A/NewCaledonia lot 75508005 182.7 2.16H3N2 A/New York lot 090407 1.0 0.01B/Shanghai lot 101607 4.6 0.05

the homologous H5N1 A/Indo VLP was evaluated (Table 4). Thecross-reactivity of H1N1-NA (A/NewCal) to H5N1-NA specific anti-body was low but detectable (2.16%), while the cross-reactivity ofN2 and B subtypes of NA to the same antibody was not detected.

The NAI titers measured independently against B and N2subtypes of NA in sera from donors immunized with the triva-lent influenza VLP vaccine demonstrated no correlation (R = 0.06)between immune responses for these two NA subtypes. While 64%of donors had a positive immune response (PPR ≥ 2.0) to both NA

log2[Serum Dilution]

Fig. 2. Inhibition of different NA subtypes by NA(H5N1 A/Indo) – specific serum. NAinhibition curves in H5N1 A/Indo, H1N1 A/NewCal, H3N2 A/NY and B/Sh VLPs byserum from sheep immunized with purified NA from H5N1 A/Indo virus (Ab titer410,000).

370 V. Gavrilov et al. / Journal of Virological Methods 173 (2011) 364–373

Table 5Distribution of donors vaccinated with 60 �g HA dose of seasonal trivalent VLP vaccine (A/Brisbane/59/2007(H1N1), A/Brisbane/10/2007 (H3N2), B/Florida/04/2006) ongroups with no (post-to-pre ratio of NAI titers < 2.0) and positive (PPR ≥ 2.0) NAI immune response against both B-NA and H3N2-NA and only one B-NA or H3N2-NA andgeometric mean ratio (GMR) of post and preimmune antibody titers against NA (by NAI assay) and HA (by HAI assay) in each group.

Group N % Total N Geometric mean ratio (GMR) of post and preimmune antibody titers

B/FL VLP H3N2-A/Br VLP

NAI HAI NAI HAI

Total 81 100 9.7 5.6 5.9 9.22891

atsr

3

psavtN1i

rtvabua1lt

Flo1

Response for both B-NA and H3N2-NA 52 64.Response only for B-NA 16 19.Response only for H3N2-NA 4 4.No response for both B-NA and H3N2-NA 9 11.

geometric mean 2-fold rise in HA titer. This demonstrates thathe assay is able to identify the immune response to individual NAubtypes when testing human vaccines for the anti-NA antibodyesponses.

.4. Accuracy

Reducing a NA-specific antibody concentration in serum sam-les by diluting the immune serum with the preimmune (naïve)erum led to a substantial decrease in the inhibition effect (Fig. 3)nd demonstrated that the NAI assay can distinguish serums witharying titers. Choosing different points on the inhibition curve forhe undiluted immunized serum had a significant impact on theAI titer values which ranged from a high for log2 NIT10 of 11.0 (at0% of NA inhibition) to a low for log2 NIT50 of 5.65 (at 50% of NA

nhibition) (Table 6).For three NAI cut-off levels (10%, 25% and 50%), the accuracy

ange was determined as the range of serum dilutions with NAIiters yielding bias ≤1.0 between measured and expected log2 NITalues. Log2 NIT50 provided a very narrow range with acceptableccuracy due to a bias ≥1.6 at 1:8 and higher serum dilutions. Aroader accuracy range, up to 1:16 serum dilution, was found forsing log2 NIT10 or 10% cut-off level. Meanwhile, 25% cut-off level

nd the calculation of log2 NIT25 further expanded this range up to:512 serum dilution. Based on the expanded accuracy range, the

og2 NIT25 designated also as log2 NIT was selected for the final NAIiter assignment and used for further assay validation.

0.0

20.0

40.0

60.0

80.0

100.0

120.0

14121086420

log2[Serum Dilution]

%R

esid

ual N

AA

Neat Prediluted sample (1:8) Prediluted sample (1:256)

ig. 3. NA Inhibition by neat and prediluted serum samples with reduced anti-NA Abevel. Prediluted anti-H3N2 A/Br ferret serum samples were prepared by dilutionsf immune serum (NIT25 = 382.7) with preimmune serum (NIT25 = 1.83) 1:8 and:256 and tested with H3N2 A/Brisbane VLP for NAI titer as per the assay procedure.

15.5 7.5 12.4 14.611.0 4.8 1.4 4.5

1.4 2.5 5.0 6.01.2 2.1 1.1 2.7

3.5. Intra-assay and inter-assay precision

Intra-assay repeatability was analyzed at three levels oflog2 NIT25, ranging from 7.86 to 3.41 and demonstrated a lowvariability, with %CV ≤ 2.3% (Table 7). The inter-assay precisionmeasured as total variability of mean log2 NIT on two differentdays when performed by two analysts corresponded to a %CV of3.95%. Data collected for inter-assay precision were arranged intogroups representing between-day and between-analyst variabil-ity (Table 8). The between-day variability demonstrated a %CV of2.55% and a statistically insignificant bias of 0.31 (p-value > 0.05for Student’s t-test). The between-analyst variability had a %CVof 3.78% and a bias of 0.46 (p-value < 0.05). Overall, the NAIassay demonstrated excellent precision since the %CV and biaswere well below the acceptance criteria for the %CV (≤10%) andbias (≤1.0).

3.6. Linearity and analytical range

Using data from the accuracy testing, the measured log2 NITwas plotted against the expected log2 NIT and linear regressionanalysis was performed (Fig. 4). The NAI assay produced a goodlinear response over the entire range of log2 NIT from 8.6 to 1.4 andalso demonstrated a high correlation between the measured and

expected values with a R square value of 0.94. Overall, the regres-sion line fit well to the line of identity and only at the lower end ofthe range lower than expected values for measured log2 NIT wereobtained. Therefore, the analytical range for the NAI assay whichmet the acceptance criteria for accuracy, precision, and linearity

0

2

4

6

8

10

1086420

Expected log2NIT

Fo

un

d lo

g2

NIT

Fig. 4. Found log2 NIT versus expected log2 NIT. NAI titers were tested using H3N2A/Brisbane VLP and anti- H3N2 A/Brisbane ferret serum samples prepared by dilu-tion with preimmune serum. A dashed line indicates the line of identity and a solidline indicates the line of linear regression.

V. Gavrilov et al. / Journal of Virological Methods 173 (2011) 364–373 371

Table 6Accuracy of log2 NIT10, log2 NIT25 and log2 NIT50 tested in serial dilutions of immune anti H3N2 A/Brisbane ferret serum diluted with preimmune serum (for preimmuneserum log2 NIT25 = 0.87 and log2 NIT50 = 0).

Antiserum dilution log2 NIT10 log2 NIT25 log2 NIT50

Expected Found Bias Expected Found Bias Expected Found Bias

Neat 11.0 11.0 0.0 8.58 8.58 0.00 5.65 5.65 0.002 10.0 10.0 0.0 7.59 8.24 0.65 4.65 4.03 −0.624 9.0 10.0 1.0 6.60 6.23 −0.37 3.65 2.07 −1.598 8.0 9.0 1.0 5.63 4.84 −0.79 2.65 0.98 −1.6716 7.0 7.0 0.0 4.68 3.65 −1.03 1.65 0.07 −1.5932 6.0 9.0 3.0 3.78 3.32 −0.46 0.65 −2.29 −2.9464 5.0 7.0 2.0 2.96 1.97 −0.99128 4.0 5.0 1.0 2.26 1.24 −1.02256 3.0 5.0 2.0 1.73 2.71 0.98512 2.0 3.0 1.0 1.36 0.88 −0.481024 1.0 3.0 2.0 1.14 0.0 −1.14

Bias > 1.0 was shown in bold numbers.

Table 7Intra-assay precision for testing log2 NIT with H3N2 A/Brisbane VLPs and anti H3N2-NA (A/Brisbane) sheep serum.

Serum sample NIT level Mean log2 NIT (n = 6) %CV

wo

TdN

4

riBaa(atbr

ivvNHlNmfm

TB

Neat HighDiluted 1:8 with preimmune serum MiddleDiluted 1:128 with preimmune serum Low

as determined to be 1.4–8.6 of log2 NIT when calculated for 25%f NAA inhibition.

The analytical performance of the NAI assay is summarized inable 9. Since all acceptance criteria were met, the assay was vali-ated and can be used for long-term clinical studies to measure theAI antibody responses to influenza vaccines.

. Discussion

It is recognized widely that generating protective immuneesponses against both HA and NA viral proteins could significantlymprove efficacy of influenza vaccines (Sylte and Suarez, 2010;right et al., 2008, 2007). Therefore, there is a significant demand forvalidated assay to measure the NA neutralizing antibody responsefter vaccination. Key parameters for such an assay must include1) a stable source of active NA enzyme; (2) standardization of anctive NA sample used for antibody titration; and (3) high reactivityo specific NA antibodies with little to no reactivity against HA anti-ody and heterologous NA antibodies. To establish the analyticalange, assay accuracy, precision, and linearity must be validated.

The results were the first time the NAA in VLPs was quantifiedn enzymatic units and compared against the NAA in inactivatedaccines as well as purified NA preparations. The NAA in our VLPsaried from 0.43 to 1.61 mU/�g HA and corresponded to the highestA activity found in inactivated influenza vaccines (1.0–1.6 mU/�gA) (Lambre et al., 1989) whereas most of these vaccines had much

ower NA activity (Kendal et al., 1980; Chaloupka et al., 1996). TheAA in VLPs exceeded the enzymatic activity reported for com-ercial purified H1N1 NA (167 mU/ml) and H5N1 NA (60 mU/ml)

rom SinoBiological (Beijing, China). Since the NA content in VLPseasured by SDS–PAGE and densitometry is less than 10% of the

able 8etween-day and between-analyst assay repeatability for testing log2 NIT with anti H3N2

Analyst/day # 1 2

Between-dayMean (n = 6) 8.76 8.45%CV 2.91 4.73

Between-analystMean (n = 6) 8.84 8.38%CV 2.41 3.98

7.86 2.05.37 2.33.41 1.3

total protein, the H5N1 VLPs had about 2-fold lower NA activity perNA protein than purified H5N1 NA (2.5 mU/�g) from RnD Systems(Minneapolis, USA). Overall, the NAA in VLPs can potentially matchthe highest level of the NA activity detected for inactivated vaccinesand purified enzyme preparations.

It was demonstrated that six strains of influenza VLPs preparedin the stabilization buffer and stored at 2–8 ◦C for 6 months hadmaintained 84–100% of NAA. The active NA was more stable inVLPs than in purified NA preparations and inactivated vaccines.By vendor specifications, all commercial purified NA preparationswere stable up to 6–12 months only upon storage at −20 ◦C. Basedon limited published results, H1N1 inactivated vaccines showeda loss of 50–100% of NA activity after 6 months of refrigeratedstorage (Kendal et al., 1980). Our VLP technology has thereforeprovided different strains of VLPs with the most active and sta-ble NA available. Ultimately, these may be considered a bettersource of the active enzyme for NA neutralizing antibody assay.Another potential advantage of VLP could be maintenance of thenative NA structure allowing better binding of antibody. Prelimi-nary experiments have suggested that the anti-NA antibody fromserum samples generated by immunization with VLP vaccine hadhigher affinity to homologous NA presented in VLPs than to thepurified soluble NA (data are not presented).

All known NA neutralizing antibody studies have used a non-characterized source of active NA evaluated only by its relativeactivity, and thus none of the published NAI antibody titers can be

directly compared. In our study, we set up and qualified an NA stan-dard for the antibody titration. Since preparing VLP samples withequal NA activity is difficult, the range for an acceptable NA activitywas established between 1.0 and 2.0 nmol/well of MU. In this rangethe assay produced similar results for NAI titers with a calculated

-NA (A/Brisbane) ferret serum.

Mean Bias p-value for t-test

8.61 0.31 0.132.55

8.61 0.46 0.023.78

372 V. Gavrilov et al. / Journal of Virological Methods 173 (2011) 364–373

Table 9Analytical performance characteristics for testing NAI Titer as log2NIT25 using VLP samples.

Parameter Acceptance criteria Results

Specificity NA subtype % Cross-reactivity < 10% % Cross-reactivity = 0Interference with anti- HA AB in VLPs < 10.0% % Relative reactivity < 4.2%

Intra-assay Repeatability (Precision) %CV ≤ 10.0 For log2 NIT = 7.9; %CV = 2.0For log2 NIT = 5.4; %CV = 2.3For log2 NIT = 3.4; %CV = 1.3

Inter-assay reproducibility (precision) %CV ≤ 10.0 %CV ≤ 4.0Between-day reproducibility %CV < 10%; bias ≤ 1.0 %CV < 3.0%; bias ≤ 0.4Between-analyst reproducibility %CV < 10%; bias ≤ 1.0 %CV < 4.0%; bias ≤ 0.5Accuracy Bias ≤ 1.0 Bias ≤ 1.0Linearity R square ≥ 0.90 R square = 0.94

bfdttoudvaTim

pTaraEumm

ipibHgeastss

alpaitadu

tae

Analytical range for log2 NITAnalytical range for NIT

ias less than one 2-fold serum dilution. Using an average valueor standard NA activity (1.5 nmol/well), each antibody titer can beefined as a serum dilution providing a specified level of inhibi-ion for 25 �U of active NA. Various VLP preparations with NA fromhe same and different influenza strains could have a wide rangef NAA, which could lead to a variable number of NA moleculessed for antibody titration. For a given sample that lost 80% of NAAuring storage, normalization by the number of activity units pro-ides 5 times higher number of NA molecules and would generatemuch lower antibody titer than for testing using a 100% active NA.his demonstrates the crucial importance of NA stability in obtain-ng reproducible NIT results which were provided in our study by

aintaining the stable NAA in VLPs during 6 months of storage.To ensure the same NA activity in the VLP sample pre and

ostimmune serum samples were analyzed on one microplate.o compare antibody titers specific to different subtypes of NA,dditional NA standardization in VLPs needs to be developed. Theeasonable approach may include measuring the specific enzymectivity per NA protein concentration by developing NA-specificLISA. From this standpoint, the evaluation of NA immune responsesing a seroconversion ratio of antibody titers for post and preim-une sera seems currently to be a more reliable approach thaneasuring absolute antibody titers themselves.Generally, binding HA-specific antibody with HA can partially

nhibit access of substrate or anti-NA antibody to NA due to theroximity of both surface proteins, which could result in a mislead-

ng overestimate for NAI titer. The inhibition by anti-HA antibodyinding can be avoided using reassortant viruses with mismatchedA which, for example, has been used for the NAI assay using thelycoprotein fetuin as a substrate (Hassantoufighi et al., 2010; Catet al., 2010). The major advantage of a synthetic substrate suchs MUNANA with a low molecular weight is that the NAI assay hashown no interference from homologous HA-specific antibody. Fur-hermore, heterologous HA and NA-specific antibody also had noignificant impact on NAI titer. The assay demonstrated excellentpecificity to NA subtypes.

The NAI titers for serums obtained from donors immunized withtrivalent seasonal VLP vaccine confirmed assay specificity and

ed to the significant conclusions that patients immunized with aolyvalent vaccine generated an independent immune responsegainst each subtype of NA and the NAI assay is able to detect thisndependent and specific immune response. It should be noted thathe high specificity of the MUNANA-based assay may result in itsbility to detect only the subset of NA-specific antibodies whichirectly bind with the active site of the enzyme and thus it may

nderestimate the actual level of NAI.

The cut-off value on the NA inhibition curve used for antibodyiter estimation appeared to be a key factor in determining theccuracy range and the overall analytical range of the assay. Thevaluation of NAI titer using 25% of NAI levels provided a signifi-

1.36–8.582.6–382

cantly broader accuracy range than for the previously used 10% and50% cut-off levels.

Both the intra-assay precision evaluated at low, medium, andhigh levels of log2 2NIT25 and the inter-assay precision examinedby between-day and between-analyst variability demonstratedexcellent assay performance with all %CV values below 5.0%. Thehigh assay precision was a result of the combination of high NAstability and standardization of NAA in VLPs and the optimizationof the cut-off value for antibody titer estimation. The high assayreproducibility demonstrates its suitability for conducting long-term multisite clinical trials and comparison of results obtainedin multiple trials.

Finally, the regression analysis of measured versus expectedlog2 NIT values confirmed linearity of the assay with an R squarevalue of 0.94. The assay has been validated for testing NA neutraliz-ing antibody in serum samples within the range 1.4–8.6 of log2 NIT.The upper limit of the range could be greater if an antiserum witha higher Ab titer was evaluated. For example, the antibody titerwith log2 NIT up to 12.8 was detected in some clinical samples fromdonors immunized with a trivalent VLP vaccine. To our knowledge,this study is the first one presenting the validation of a NA neutral-izing antibody assay. With this data, the reported analytical rangecould be used as a benchmark for all available assays along withany new assay development.

The new NA neutralizing antibody assay using VLPs as a substi-tute for live virus could be of great value in further studies of NAimmunity, especially with regard to new research evaluating thesynergistic effect of NA and HA-specific antibodies on protectionagainst influenza virus infection (Bosch et al., 2010).

Acknowledgement

We thank Travis Sadowski and Dr. Sarathi Boddapati for criticalreview of the manuscript.

References

Bosch, B.J., Bodewes, R., de Vries, R.P., Kreijtz, J.H., Bartelink, W., van Amerongen,G., Rimmelzwaan, G.F., de Haan, C.A., Osterhaus, A.D., Rottier, P.J., 2010. Recom-binant soluble, multimeric HA and NA exhibit distinctive types of protectionagainst pandemic swine-origin 2009 A(H1N1) influenza virus infection in fer-rets. J. Virol. 84, 10366–10374.

Bright, R.A., Carter,.M., Crevar, C.J., Toapanta, F.R., Steckbeck, J.D., Cole, K.S., Kumar,N.J., Pushko, P., Smith, G., Tumpey, T.M., Ross, T.M., 2008. Cross-clade protectiveimmune responses to influenza viruses with H5N1 HA and NA elicited by aninfluenza virus-like particle. PLoS One 1, e1501.

Bright, R.A., Carter, D.M., Daniluk, S., Toapanta, F.R., Ahmad, A., Gavrilov, V., Massare,M., Pushko, P., Mytle, N., Rowe, T., Smith, G., Ross, T.M., 2007. Influenza virus-

like particles elicit broader immune responses than whole virion inactivatedinfluenza virus or recombinant hemagglutinin. Vaccine 25, 3871–3878.

Cate, T.R., Rayford, Y., Nino, D., Winokur, P., Brady, R., Belshe, R., Chen, W., Atmar, R.L.,Couch, R.B., 2010. A high dosage influenza vaccine induced significantly moreneuraminidase antibody than standard vaccine among elderly subjects. Vaccine28, 2076–2079.

logica

C

F

G

H

I

J

K

K

K

V. Gavrilov et al. / Journal of Viro

haloupka, I., Schuler, A., Marschall, M., Meier-Ewert, H., 1996. Eur. J. Clin. Microbiol.Infect. Dis. 15, 121–127.

eng, J., Gulati, U., Zhang, X., Keitel, W.A., Thompson, D.M., James, J.A., Thompson,L.F., Air, G.M., 2009. Antibody quantity versus quality after influenza vaccination.Vaccine 27, 6358–6362.

ubareva, L.V., Webster, R.G., Hayden, F.G., 2002. Detection of influenza virus resis-tance to neuraminidase inhibitors by an enzyme inhibition assay. Antiviral Res.53, 47–61.

assantoufighi, A., Zhang, H., Sandbulte, M., Gao, J., Manishewitz, J., King, L., Golding,H., Straight, T.M., Eichelberger, M.C., 2010. A practical influenza neutralizationassay to simultaneously quantify hemagglutinin and neuraminidase-inhibitingantibody responses. Vaccine 28, 790–797.

nternational Conference on Harmonization (ICH) of Technical Requirements forthe Registration of Pharmaceuticals for Human Use, Validation of AnalyticalProcedures: Methodology, adopted in 1996, Geneva.

ohansson, B.E., Bucher, D.J., Kilbourne, E.D., 1989. Purified influenza virushemagglutinin and neuraminidase are equivalent in stimulation of antibodyresponse but induce contrasting types of immunity to infection. J. Virol. 63,1239–1246.

albfuss, B., Knochlein, A., Krober, T., Reichl, U., 2008. Monitoring influenza viruscontent in vaccine production: precise assays for the quantitation of hemagglu-tination and neuraminidase activity. Biologicals 36, 145–161.

ang, S., Song, J., Quan, F., Compans, R.W., 2009. Influenza vaccines based on virus-like particles. Viral Res. 143, 140–146.

endal, A.P., Bozeman, F.M., Ennis, F.A., 1980. Further studies of the neuraminidasecontent of inactivated influenza vaccines and the neuraminidase antibodyresponses after vaccination of immunologically primed and unprimed popu-lations. Infect. Immun. 29, 966–971.

l Methods 173 (2011) 364–373 373

Lai, J.C.C., Chan, W.W.L., Kien, F., Nicholls, J.M., Malik Peiris, J.S., Garcia, J.M., 2010.Formation of virus-like particles from human cell lines exclusively expressinginfluenza neuraminidase. J. Gen. Virol. 91, 2322–2330.

Lambre, C.R., Chauvaux, S., Pilatte, Y., 1989. Fluorometric assay for the measurementof viral neuraminidase in influenza vaccines. Vaccine 7, 104–105.

Mahmood, K., Bright, R.A., Mytle, N., Carter, D.M., Crevar, C.J., Achenbach, J.E., Heaton,P.M., Tumpey, T.M., Ross, T.M., 2008. H5N1 VLP vaccine induced protection inferrets against lethal challenge with highly pathogenic H5N1 influenza viruses.Vaccine 26, 5393–5399.

Potier, M., Mameli, L., Belisle, M., Dallaire, L., Melancon, S.B., 1979. Fluromet-ric assay of neuraminidase with a sodium (4-methylumbelliferyl-�-d-N-acetylneuraminate) substrate. Anal. Biochem. 94, 287–296.

Pushko, P., Tumpey, T.M., Bu, F., Knell, J., Robinson, R., Smith, G., 2005. Influenza virus-like particles comprised of the HA, NA, and M1 proteins of H9N2 influenza virusinduce protective immune responses in BALB/c mice. Vaccine 23, 5751–5759.

Sylte, M.J., Suarez, D.L., 2010. Influenza neuraminidase as a vaccine antigen. In: Com-pans, R.W., Orenstein, W.A. (Eds.), Vaccines for Pandemic Influenza, CurrentTopics in Microbiology and Immunology 333. Springer-Verlag, Berlin Heidel-berg, pp. 227–241.

Sylte, M.J., Hubby, B., Suarez, D.L., 2007. Influenza neuraminidase antibodies providepartial protection for chickens against high pathogenic avian influenza infection.Vaccine 25, 3763–3772.

Wetherall, N.T., Trivedi, T., Zeller, J., Hodges-Savola, C., McKimm-Breshkin,J.L., Zambon, M., Hayden, F.G., 2003. Evaluation of neuraminidaseenzyme assays using different substrates to measure susceptibility ofinfluenza virus clinical isolates to neuraminidase inhibitors: report ofthe neuraminidase inhibitor susceptibility network. J. Clin. Microbiol. 41,742–750.