Optimisation of a micro-neutralisation assay and its application in antigenic characterisation of influenza viruses

Objectives The identification of antigenic variants and the selection of influenza viruses for vaccine production are based largely on antigenic characterisation of the haemagglutinin (HA) of circulating viruses using the haemagglutination inhibition (HI) assay. However, in addition to evolution related to escape from host immunity, variants emerging as a result of propagation in different cell substrates can complicate the interpretation of HI results. The objective was to develop further a micro-neutralisation (MN) assay to complement the HI assay in antigenic characterisation of influenza viruses to assess the emergence of new antigenic variants and reinforce the selection of vaccine viruses. Design and setting A 96-well-plate plaque reduction MN assay based on the measurement of infected cell population using a simple imaging technique. Sample Representative influenza A (H1N1) pdm09, A(H3N2) and B viruses isolated between 2004 and 2013 Main outcome measures and results Improvements to the plaque reduction MN assay included selection of the most suitable cell line according to virus type or subtype, and optimisation of experimental design and data quantitation. Comparisons of the results of MN and HI assays showed the importance of complementary data in determining the true antigenic relationships among recent human influenza A(H1N1)pdm09, A(H3N2) and type B viruses. Conclusions Our study demonstrates that the improved MN assay has certain advantages over the HI assay: it is not significantly influenced by the cell-selected amino acid substitutions in the neuraminidase (NA) of A(H3N2) viruses, and it is particularly useful for antigenic characterisation of viruses which either grow to low HA titre and/or undergo an abortive infection resulting in an inability to form plaques in cultured cells.

Objectives The identification of antigenic variants and the selection of influenza viruses for vaccine production are based largely on antigenic characterisation of the haemagglutinin (HA) of circulating viruses using the haemagglutination inhibition (HI) assay. However, in addition to evolution related to escape from host immunity, variants emerging as a result of propagation in different cell substrates can complicate the interpretation of HI results. The objective was to develop further a microneutralisation (MN) assay to complement the HI assay in antigenic characterisation of influenza viruses to assess the emergence of new antigenic variants and reinforce the selection of vaccine viruses.
Design and setting A 96-well-plate plaque reduction MN assay based on the measurement of infected cell population using a simple imaging technique.

Sample Representative influenza A (H1N1) pdm09, A(H3N2) and B viruses isolated between 2004 and 2013
Main outcome measures and results Improvements to the plaque reduction MN assay included selection of the most suitable cell line according to virus type or subtype, and optimisation of experimental design and data quantitation. Comparisons of the results of MN and HI assays showed the importance of complementary data in determining the true antigenic relationships among recent human influenza A(H1N1)pdm09, A(H3N2) and type B viruses.

Introduction
Influenza viruses evolve constantly to escape human immunity and cause annual epidemics and occasional pandemics. To minimise the impact of influenza, vaccination is the best option, but its effectiveness depends on the degree of antigenic similarity between vaccine viruses and circulating viruses. For over 60 years, the identification of antigenic variants has been determined largely by the haemagglutination inhibition (HI) assay to measure the ability of antibodies, raised against vaccine and reference viruses, 1 to prevent the attachment of virus to red blood cells (RBCs), a process analogous to the binding of virus to host cell receptors. However, during the past decade, interpretation of HI results has become complicated: either because of changes in receptor binding properties as a result of virus evolution, or due to selection of variants during the isolation and passage of viruses in cell lines or eggs. For example, the loss of the ability of A(H3N2) viruses to agglutinate chicken, and subsequently turkey, RBCs was caused by amino acid substitutions E190D and D225N in the haemagglutinin (HA), occurring around 1990 and 2005, respectively. [2][3][4] The extremely low avidity of virus for receptors has contributed to the selection of mutations in the neuraminidase (NA) gene during propagation of such viruses in Madin-Darby canine kidney (MDCK) cells. 4,5 Amino acid substitutions flanking the catalytic site of NA (D151G/N or T148I) allow NA to contribute to the binding and agglutination of RBCs in a way that is resistant to inhibition by anti-HA antibodies in postinfection ferret antisera, as assessed by HI assay, 5 but sensitive to the neuraminidase inhibitor (NAI), oseltamivir carboxylate. It has long been established that during adaptation of influenza B viruses to growth in hens' eggs, the loss of a glycosylation site in the HA has resulted in different patterns of HI reactivity between egg-and MDCK cellpropagated viruses [6][7][8] and egg-adaptive changes in influenza A viruses also affect the behaviour of viruses in the HI assay. Furthermore, variation in RBCs derived from different species and individual animals can also affect HI results. Hence, virus neutralisation assays have been used in conjunction with HI to clarify the true antigenic relationships between viruses and to support assessment of the antigenic characteristics of A(H3N2) viruses at the WHO influenza vaccine consultation meetings since 2009. 1 Although the micro-neutralisation (MN) assay can potentially overcome non-antigenic effects of variation between viruses and between RBCs, most MN assays based on cytopathic effect or ELISA require 100 50% tissue culture infectious dose (TCID 50 ) of input virus per well, 9 presenting a difficulty for the majority of recent A(H3N2) viruses which replicate to low titre in cell culture. Also, quantitation of plaque numbers in a plaque reduction assay is difficult when there is a significant variation in plaque size induced by individual viruses and it is impossible with some currently circulating A(H3N2) viruses that undergo abortive infection and do not form visible plaques.
Here, we describe the application of an improved version of the plaque reduction MN assay to the antigenic characterisation of currently circulating influenza viruses. It is based on measurement of the infected cell population (ICP), in a 96-well-plate format, using a simple imaging method. 10

Viruses and cells
Influenza viruses were obtained from the Francis Crick Worldwide Influenza Centre. A(H3N2) viruses were propagated in MDCK-SIAT1 cells, and its parent cells were used for propagating A(H1N1)pdm09 and influenza B viruses (both cell lines kindly provided by Dr. M. Matrosovich, University of Marburg, Germany 11 ). Mini-pig kidney (MPK) and MDCK-ECACC were obtained from The Pirbright Institute and the European Collection of Cell Cultures, respectively. Viruses were propagated in cell lines incubated at 34°C in Dulbecco's modified Eagle's medium (DMEM), containing 2 lg/ml trypsin (TPCK-treated). Some reference viruses were propagated in 10-day-old embryonated hens' eggs at 34°C.

HI assays
Haemagglutination and HI assays were performed according to standard methods using suspensions of guinea pig RBCs (1Á0% v/v) for A(H3N2) viruses and turkey RBCs (0Á75% v/v) for A(H1N1)pdm09 and type B viruses with post-infection ferret antisera pre-treated with receptor-destroying enzyme (RDE) from Vibrio cholera. 12 Four HA units were used in HI assays of A(H3N2) and type B viruses and 8 HA units in assays of A(H1N1)pdm09 viruses. For A(H3N2) viruses, haemagglutination and HI assays were conducted in the presence of 20 nM oseltamivir carboxylate. 5

MN assays
Micro-neutralisation assays were conducted as described previously. 10 Briefly, doubling dilutions of RDE-treated postinfection ferret antisera were added to confluent cell monolayers in 96-well plates (except for virus and cell control wells) in DMEM; viruses to be tested were then added and plates incubated at 37°C or 34°C for three hours for influenza A and B viruses, respectively. The inoculum was removed and 200 ll/well of overlay medium, containing 1Á2% (w/v) Avicel (FMC BioPolymer) and 2 lg/ml trypsin, was added before incubating plates at 37°C for 22 hours (influenza A) or 34°C for 28 hours (influenza B). Cells were fixed with 4% (w/v) paraformaldehyde, stained using a pool of anti-influenza nucleoprotein (NP) mouse monoclonal antibodies, a peroxidase-conjugated goat anti-mouse antibody and TrueBlue TM substrate. 10 The dried plates were scanned using a flatbed scanner. The ICP in each well, the percentage of positive (infected) cells, was quantified as the percentage of positive pixels using IN-HOUSE image-processing software, 10 and the average ICPs were normalised to the virus control for each plate using Excel (Microsoft â ). The background measurement for uninfected cell controls was subtracted during data processing. The neutralisation titre was determined as the reciprocal of the antiserum dilution corresponding to 80% reduction in ICP. Linear interpolation was used to estimate titres falling between two adjacent serum dilutions.
The virus (control) inoculum was titrated by serial 5-or 10-fold dilutions. ICPs of triplicate wells were averaged and normalised to the ICP of the wells in which all cells were infected. A virus dilution that caused 20-85% of saturating ICP was chosen, closest to the mid-point. If the ICP did not reach saturation, the maximum titre of virus was used.

Nucleotide sequence analysis
Following RT/PCR, nucleotide sequences of HA and NA genes were determined using ABI prism BigDye terminator cycle sequencing kits and an ABI 3730XL DNA analyser. Primer sequences are available on request.  (Table 1).

Optimisation of assay conditions
A(H3N2) viruses showed the best propagation and plaque formation, as well as a higher frequency of isolation, 13 on MDCK-SIAT cells, which express increased levels of a 2,6linked sialic acid on their surface, 11 and were used in subsequent experiments.
There were no significant differences in haemagglutination titres of A(H1N1)pdm09 when passaged in the different cell lines. However, MDCK-parent cells were used as they yielded higher frequency of isolation of A(H1N1)pdm09 viruses from clinical specimens collected during 2012 and 2013 (data not shown), and larger plaques compared with MDCK-SIAT1 cells.
Titration of influenza B viruses, of both the Yamagata and the Victoria lineages, showed that MDCK-parent cells yielded the highest titres and clearest plaques.
As bovine serum albumin (BSA: Life Technologies BSA Fraction V) might affect the interaction between virus and receptor, particularly in the case of recent A(H3N2) viruses with a low avidity for cell receptors, 4 its effects were assessed. Inclusion of BSA in the diluent and overlay inhibited the formation of plaques, as shown for A(H3N2) viruses in Figure S1A, and similar effects were observed with A(H1N1) pdm09 and type B viruses (S. Wharton, unpublished observations); therefore, BSA was omitted from the assay. In addition, removal of the inoculum (plus antiserum) before adding the overlay was necessary as the continued presence of serum proteins throughout the plaque assay, as shown for an antiserum raised against an unrelated influenza virus, impaired plaque formation by A(H3N2) viruses ( Figure S1B).

Similar effects were seen with A(H1N1)pdm09 and B viruses (S. Wharton, unpublished observations).
Incubation of type A viruses at either 37°C or 34°C resulted in no noticeable difference in plaque formation; however, incubation at 34°C resulted in better plaque formation by influenza B viruses. The most appropriate incubation time for both A subtypes was approximately 22 hours post-infection at 37°C, while 28 hours was optimal for type B viruses at 34°C.

Influence of the amount of input virus on MN titre
ICP was chosen as the criterion for titration of infectivity and neutralisation by antibody, 10 as it has several advantages over plaque-forming units (PFU): it can accommodate variation in plaque morphology [a single virus stock commonly yields mixtures of large and small plaques, with many small plaques being beyond visual resolution ( Figure S2)]; it can be applied to viruses, such as many recent A(H3N2) viruses, showing only single cell abortive infection (the average cell size is 0Á65 pixels 10 ); and by counting all infected cells, it reduces the influence of plaque merging.
The amount of input virus was standardised so that MN titres calculated for different viruses, from different experiments, or laboratories, could be compared. As many factors can influence the ICP, the range of ICP that gave reproducible results was assessed. Eleven A(H3N2) viruses were examined against five post-infection ferret antisera using up to a 100-fold dilution range of the individual virus stocks in the neutralisation experiments. Results for four representative viruses show that while MN titres can vary markedly depending on the level of ICP (Table 2), if the ICP was within the range 20-85%, the variation was on average within a factor of 1Á32 AE 0Á56, with a 95Á4% confidence (2 standard deviations, 45 valid titres at 80% ICP reduction), that is <2-fold variation. For example, the reference antisera Overall, the results indicated that the 20-85% ICP range represents a good compromise between titre accuracy and the simplicity of the neutralisation assay. This criterion was also supported by data obtained with A(H1N1)pdm09 and type B viruses (Tables S2 and S3). For most viruses, there was little difference between relative titres at 50% and 80% reductions in ICP, but MN titres based on 80% reduction were closer to corresponding HI titres and, for comparison studies, were chosen.

Comparison of MN with HI in antigenic characterisation
Influenza A(H1N1)pdm09 Comparable results were obtained for MN and HI assays, as illustrated in Table 3

Influenza A(H3N2)
In the light of the polymorphisms at positions 148 and/or 151 that enable the NA of A(H3N2) viruses to bind RBCs and influence haemagglutination, 5 comparisons were made between the MN assay and HI assays conducted with and without added 20 nM oseltamivir carboxylate ( Table 4).
As shown for A/Stockholm/40/2012, A/Denmark/67/2012 and A/Denmark/68/2012, viruses carrying a polymorphism in NA at position 148 or 151 in the catalytic site had lower HI titres (2-to 4-fold) in the absence than in the presence of oseltamivir carboxylate, as observed for viruses in our previous study, 5 and lower than the corresponding MN titres, with the exception of results obtained with the antiserum raised against A/Athens/112/2012 (Table 4).
In the MN assay, antisera raised against cell-propagated reference viruses, A/Stockholm/18/2011, A/Athens/112/2012 and A/Victoria/361/2011C, reacted well (reductions in titres were < 2-fold) with all seven cell-propagated test viruses and the egg-propagated A/Texas/50/2012, indicative of antigenic similarity between test and reference viruses. However, for the antiserum raised against egg-propagated A/Victoria/361/ 2011E, there were up to 12-fold reductions in MN titre compared to the homologous titre, which is similar to the differences in titres of up to 16-fold in the HI assays carried out in the presence or absence of oseltamivir carboxylate. These differences relate to A/Victoria/361/2011E having acquired two HA1 egg-adaptation substitutions, H156R/Q and G186V.

Influenza B viruses
Viruses of the two influenza B lineages were readily discriminated by the MN assay, as in the HI assay, despite some low level cross-reactivity of particular post-infection ferret antisera (Table S1). The antigenic relationships of viruses within each of the lineages were determined in parallel by MN and HI assays, utilising both egg-propagated and cell-propagated viruses, notably to assess and compare the influence of the presence or absence of a glycosylation site at asparagine 197 (Victoria-lineage) or 196 (Yamagata-lineage) which is commonly lost on passage in eggs. 6,7,14 Both HI and MN assays clearly distinguished the B/ Brisbane/60/2008 vaccine virus and the reference virus B/ Malta/63674/2011, from the earlier vaccine virus B/Malaysia/ 2506/2004 using antisera raised against these egg-propagated viruses (Table 5). However, test viruses collected in 2011 or 2012 and propagated exclusively in cell culture, and the cellpropagated reference virus B/Hong Kong/514/2009, were poorly recognised by antisera raised against the contemporary egg-propagated reference viruses in HI assays, with titres 4-to ≥64-fold lower than the titres of the antisera with their homologous viruses, as expected. 8 Conversely, antiserum raised against cell-propagated B/Hong Kong/514/2009 reacted poorly with the egg-propagated viruses, but gave titres within 2-fold of the homologous HI titre with the test viruses.
The differences observed between cell-propagated and eggpropagated viruses were generally less marked in MN assays with all viruses being recognised by the antisera within 8-fold of the titres for the homologous viruses (Table 5)    all test viruses at titres within 4-fold of the titre for the homologous virus.
The HA genes of recently circulating B/Yamagata-lineage viruses fall into two clades. 15 Like viruses of the B/Victorialineage, cell-propagated viruses of the B/Yamagata-lineage were recognised poorly in HI assays by antisera raised against egg-propagated viruses, at 2-to 16-fold lower titres compared to the respective homologous titres, and these antisera did not discriminate between viruses of the two clades (Table 6). In contrast, clade 2 viruses were well recognised by the antiserum raised against the cell-propagated clade 2 reference virus B/Estonia/55669/2011, at HI titres within 4-fold of the homologous titre; conversely, clade 3 viruses were recognised at 8-to 64-fold lower HI titres by this antiserum. Discrimination between test viruses in the two clades was less marked in the corresponding MN assays. However, the antiserum raised against egg-propagated B/ Wisconsin/1/2010 (clade 3) appeared to discriminate better, giving higher titres with clade 3 viruses than with clade 2 viruses. B/Denmark/8/2012 (clade 3) appeared more closely related to clade 2 than clade 3 viruses; the reason is unknown, but it is noteworthy that the HA gene of B/Denmark/8/2012 did not encode K298E and E312K substitutions, common in clade 3 viruses, but carried amino acids typical of clade 2 viruses at these positions. The antiserum raised against egg-propagated B/Massachusetts/ 02/2012 did not distinguish between test viruses of clade 2 and 3 in either HI or MN assays.
Overall, the results of the MN assay appeared to be less affected by the passage history of type B viruses than those of the HI assay.

Discussion
Comparisons of results of MN and HI assays demonstrated the usefulness of the improved MN assay in complementing the HI assay when determining antigenic relationships among recently circulating human influenza A and B viruses, and interpreting the impact of non-antigenic variation (for example, changes in receptor binding specificity, affinity or avidity) on HI titre.
In addition to selection of the best cell line to support plaque formation (MDCK-parent cells for A(H1N1)pdm09 and type B viruses, and MDCK-SIAT1 cells for A(H3N2) viruses) imaging-aided quantitation of the ICP, which detects all infected cells from visible plaques to single infected cells, greatly enhanced the consistency and speed of the MN assay for antigenic characterisation. As long as the ICP was between 20% and 85% of the total cell population, the results from different neutralisation experiments were comparable (within 2-fold for replicate assays).
Titration of antibodies by HI and neutralisation are not always comparable, 16 and the sensitivity and the efficiency of the HI assay can be affected by the species of RBCs used, exacerbated by individual animal variation, and the cell substrate used for virus isolation and propagation. 2,5     The major antigenic sites on the HA are located around the receptor binding site, 17 and HI measures the inhibition of virus binding to sialic acid receptors present on RBCs. However, other antibodies present in convalescent serum can bind elsewhere on the HA, notably those that bind to the stem of the HA. 18,19 MN assays have the potential to detect the effects of a wider range of antibodies than HI and therefore have the ability to reflect more comprehensively the antigenic similarities or differences between viruses. Results of MN were generally consistent with and confirmed those of HI, notably in revealing antigenic differences between antigenic drift variants of recent type A and type B vaccine viruses, and antigenic changes caused by culture-selected or sporadic changes, such as the G155E substitution in HA1 of certain A(H1N1)pdm09 viruses. However, differences can occur due to alteration in receptor binding, as exemplified by recent A(H3N2) viruses. Amino acid polymorphisms at positions 148 or 151 of NA, which cause NA-dependent binding to sialic acid receptors, enhance 'apparent' HA titres. As anti-HA antibodies do not block such binding, reductions in HI titres might be interpreted erroneously as differences in the antigenicity of the HA. 5 Thus, MN assay results paralleled those of HI only when oseltamivir carboxylate was included in the HI test, indicating that the MN assay was less affected by NAdependent binding and reflected antigenic differences more accurately with the cells used here.
Influenza B viruses propagated in eggs commonly select HA1 substitutions which result in the loss of an N-linked glycosylation site. [6][7][8] Glycans attached to Asn-196/7 (Yamagata/Victoria lineage) in the HA can interfere with binding to receptors 20 and can also alter antigenicity. 7,20,21 Loss of these glycans, on the periphery of the receptor binding site, results in the exposure of a highly immunogenic site 6 that is masked by the glycan in cell-propagated viruses. Hence, while influenza B viruses that have either lost or become polymorphic in the glycosylation site exhibit high HI titres with antisera raised against eggpropagated viruses, most cell-propagated viruses (of both lineages) were recognised at ≥ 8-fold lower HI titres. The differences in MN titres between egg-propagated and cellpropagated viruses were lower, 2-to 8-fold, showing that effects of glycosylation were less pronounced in the MN assay.
As the majority of influenza vaccines are manufactured using egg-propagated B viruses that have lost glycosylation at Asn-196/7 of HA1, the influence on immunogenicity and/or antigenicity of influenza B viruses is of particular importance. [6][7][8] It is also evident that cell-propagated reference viruses, and the corresponding post-infection ferret antisera, are more appropriate than egg-based reagents for determining the antigenic relationships among influenza B viruses, whether performed by HI or MN assays.
In conclusion, this study has established experimental parameters for a robust MN assay and reinforces the importance and reliability of MN results in supporting HI data for detailed antigenic analyses of currently circulating influenza viruses, notably for the biannual selection of viruses for inclusion in human influenza vaccines.

Supporting Information
Additional Supporting Information may be found in the online version of this article: Figure S1. Effects of BSA (A) or non-related antiserum (B) on plaque formation by A/Brisbane/10/2007(H3N2) in MDCK-SIAT1 cells. Figure S2. Variation in plaque size and morphology Table S1. Lack of cross-neutralisation between influenza B viruses of the two lineages. Table S2. Effects of variation in input A(H1N1)pdm09 virus titre (ICP) on MN titres. Table S3. Effects of variation in input type B virus titre (ICP) on MN titres.