Use of lentiviral pseudotypes as an alternative to reassortant or Triton X‐100‐treated wild‐type Influenza viruses in the neuraminidase inhibition enzyme‐linked lectin assay

Abstract Background Formulation of neuraminidase (NA) within influenza vaccines is gaining importance in light of recent human studies. The enzyme‐linked lectin assay (ELLA) is considered a reliable assay to evaluate human anti‐NA antibodies. Objectives To overcome interference by hemagglutinin (HA)‐specific antibodies and detect neuraminidase inhibitory (NI) antibodies only, two different sources of antigen have been studied in ELLA: reassortant viruses with a mismatched avian origin‐HA or Triton X‐100 (Tx)‐treated wild‐type viruses. Pseudotypes or pseudovirus (PV), characterized by a lentivirus core bearing human influenza NA and avian influenza HA, were investigated as an alternative source of antigen and compared to HA‐mismatched and Tx‐treated viruses, since represent a safer product to be handled. Methods Two independent panels of sera were analyzed by ELLA to evaluate the anti‐NA response against N1 (A/California/07/2009 (H1N1pdm)) and N2 (A/Hong Kong/4801/2014 (H3N2)). The NA inhibition (NI) antibody titers measured as either the 50% end point or 50% inhibitory concentration (IC50) were compared for every source of antigen. Results The ELLA assay performed well with all three sources of antigen. NI titers measured using each antigen type correlated well when reported either as end point titers or as the IC50. Conclusions This study suggests that HA‐mismatched whole virus, Triton‐treated wild‐type virus or PV can be used to measure NI antibody titers of human sera, but further comparability/validation assays should be performed to assess statistical differences. The data support the use of PV as an attractive alternative source of antigen and justify further investigation to improve stability of this antigen source.


| INTRODUC TI ON
Neuraminidase is the second most abundant glycoprotein on the influenza virus surface (17% of the overall surface) after HA and it is usually expressed at a ratio 1:4 (40-50 NA and 160-200 HA spikes), 1 with exceptions. [2][3][4] NA has multiple roles: (a) allow the release of newly formed virions from the surface of the infected cell, leading to viral spread, (b) enhance influenza infection by acting on glycoconjugates expressed at the cell surface, 5 and (c) form complexes with sialic acids on the host cell surface, 6,7 particularly for H3N2 viruses. 8 Several studies have confirmed that both inactivated and live attenuated vaccines have the capacity to induce NA-specific antibodies. 9,10 NA inhibiting antibodies are associated with resistance against influenza, 11,12 reduced severity and duration of disease. 13 Several different assays have been used to evaluate the antibody response to NA. The traditional NA inhibition (NI) assay determines the extent of antibody-mediated interference with viral enzyme activity based on the measurement of sialic acid that is released from a glycosylated substrate. 14 An assay that measures NA activity based on accessibility of galactose, the penultimate sugar of many complex carbohydrates, to peanut agglutinin, offers advantages in that it does not use hazardous chemicals and has higher throughput. This enzyme-linked lectin assay (ELLA) developed by Lambré et al 15 and successively adapted and optimized, 10,16,17 measures sialidase activity of NA by detecting the terminal galactose that becomes exposed after sialic acid cleavage. A study conducted by Eichelberger et al 18 that employed the protocol published by Couzens et al 17 showed that the ELLA is robust and sensitive although improvements can be made to further standardization of the method.
Measuring the NI antibody only is possible if HA-specific antibodies are unable to bind to the virus. This is usually accomplished by using reassortant viruses that have a mismatched avian HA, although it should be kept in mind that antibodies against conserved HA epitopes 19 can still occur. The production of reassortant influenza viruses, beyond the intrinsic difficulty of optimizing the process, limits many laboratories from using these as a source of antigen since genetically modified organisms require additional biosafety containment and in some countries a permit from the Department of Agriculture. This is due to the HA gene, that is usually derived from an avian source. The inability of many laboratories to produce such reassortant viruses by reverse genetics have led to the employment of alternative sources of antigen. Jonges et al 20

| Pseudoviruses
Production of lentiviral PVs was carried out by co-transfecting Carolyn Nicolson, NIBSC), as previously described. [22][23][24][25] The H11 plasmid was added to improve NA stability and increase the PV release and production as previously described 26 and confirmed for this assay. 22,25 Briefly, 1 µg of HA, 1 µg of NA, and 1.5 µg pNLLuc4.3 plasmids were transfected into HEK293T/17 cell lines using Endofectin™ Lenti (3 µL/ µg). Medium was replenished 24 hours after transfection. The NA activity of each PV was titrated in ELLA as reported previously. 17

| HA assay
HA assays were performed to confirm the inability of Triton X-100 treated virus to agglutinate RBCs. The protocol was described elsewhere (WHO 2011, Manual for the laboratory diagnosis and virological surveillance of influenza).

| ELLA assay
ELLA assays were performed as previously described, 17  IC 50 values were automatically generated using Graph Pad Prism 5® software.

| Statistical evaluation
Every serum sample was tested in duplicate and evaluated by both 50% end point titer and IC 50 outcomes. 22 The percent inhibition of enzyme activity is calculated as follows: background optical density (OD) is subtracted from the virus control (maximum NA activity, no serum added) and sample ODs. The 50% end point titer is calculated as the highest serum dilution that resulted in at least 50% inhibition of the maximum NA activity. IC 50 values were generated through a non-linear regression curve fit using GraphPad Prism 5®, as described elsewhere. 25 The geometric mean titer (GMT) of end point titers was reported as Log 2 and compared by simple non-linear regression curve fit, and r 2 (coefficient of determination), a measure of strength of the relation between two variables. 27 To investigate the relationship between the errors in measurement and the true values, the mean difference (d) and the standard deviations of the differences (s) were calculated. 27 The Spearman rank calculation was also used to compare the outputs, as reported elsewhere. 28 In addition, Log 2 of IC 50 titers determined from inhibitory curves were also compared. Coefficient of determination (r 2 ), the mean difference (d), and the standard deviation of the differences (s) were evaluated by Bland and Altman and Spearman rank analyses.

| Preparation of Triton X-100-treated antigens
Several concentrations of Triton X-100 (Tx) were tested for their ability to selectively inactivate HA activity. Wild-type A/ California/07/2009 (H1N1) virus was completely inactivated by 0.5% and 1% Triton X-100 treatment. Haemagglutination activity of wild-type A/Hong Kong/4801/2014 (H3N2) was completely destroyed even by 0.1% Triton X-100 concentration, as well as 0.5% and 1%. As in the method of Jonges et al, 20 Triton X-100 was not removed.
The impact of Tx-treatment on NA activity was tested by titrating each preparation in ELLA. The activity of N1-Tx preparations was slightly reduced compared to the untreated virus, but it did not impede test performance. To assess specificity of NA inhibition, the N1-Tx wild-type virus was pre-incubated with anti-A/ California/7/2009 (N1) NA serum (NIBSC, code 10/218), anti-A/ California/7/09 HA serum (NIBSC, code 16/114), or human serum minus IgA/IgM/IgG (Sigma Aldrich cat. S5393-1VL). The homologous antiserum against the HA did not show any inhibition of the NA. The immunoglobulin-depleted human serum also did not inhibit NA. In contrast, the antiserum against NA inhibited enzyme activity of the N1-Tx virus. The same titer was also obtained when the antiserum against NA was pre-incubated with untreated wildtype virus (results not reported).
Unfortunately, none of the N2-Tx preparations retained NA activity. Even changing the buffer and the pH, as previously suggested, 29 N2-Tx preparations lost NA activity. Therefore, only N1-Tx was included in the comparison of antigens to measure NA inhibition antibody titers.

| H6N1, H1N1-Tx, and H11N1 PV sources of antigen show comparable NI antibody titers
H6N1 reassortant virus, N1-Tx, and H11N1 PV were titrated in ELLA to determine the appropriate amount of antigen to use in each assay.
The amount of antigen added to each assay was 90% of the maximum OD for each source of antigen. ELLA assays were performed to measure NI antibody titers of 40 sera against each N1 antigen, (Figures 1 and 2). Table 1 shows that results are reproducible, with the NI titer of each replicate being within 2-fold difference. To allow statistical comparison of results, titers measured as <10 (1:10 was the first dilution of serum) were assigned a titer of 5. In addition, Log 2 of IC 50 values, obtained using Graph Pad Prism 5®, were collected and analyzed.
The inhibition curves were obtained by performing non-linear regression from every serum run in duplicate (GMT is shown) against each N1 source of antigen (Figures 1 and 2).  Interestingly the titers obtained with N1-Tx were approximately 2fold lower than the PV antigen, while titers obtained with H6N1 reassortant values were often 2-fold higher than those measured against the PV antigen. Only in 1 case was an 8-fold difference in titer observed between N1-Tx or H11N1 PV and H6N1 (serum 22). In summary, titers from ELLA assays conducted with the three different sources of antigen varied not more than 4-fold, except in one case.
A comparison of ELLA titers using three different sources of antigens was also performed (Figure 3).
Comparison between N1-Tx and H11N1 PV ( Figure 3A) yields an r 2 = 0.9723. Interestingly both these two sources of antigens are F I G U R E 1 Inhibition curves showing the ability of sera to inhibit sialidase activity across the plate. Sera from 1 to 20 (corresponding to plates 1-5) were tested in duplicate (GMTs here reported) against H6N1 reassortant virus (left column), N1-Tx antigen (central column), and H11N1 PV (right column). Y axes represent the percentage of inhibition while X axes report the Log2 of the dilution. The prefix "S1." in front of every serum number is omitted to improve the readability of the legends comparable to H6N1 ( Figure 3B This confirmed that the two measurements are comparable, with very few titers outside the intervals defined as "limits of agreement." However, titers measured against N1-Tx were often less than measured against H11N1 PV ( Figure 4B), and titers measured against H6N1 were often higher than those measured by either N1-Tx or H11N1 PV ( Figure 4A,C, respectively). The difference in titers measured for sera 22 and 23 was outside the limits of agreement for assays using H6N1 and H11N1 PV ( Figure 4C).
Comparison between N1-Tx and H11N1 PV IC 50 titers (Figure 5B) F I G U R E 2 Inhibition curves showing the ability of sera to inhibit the sialidase activity across the plate. Sera from 21 to 40 (corresponding to plates 6-10) were tested in duplicate (GMTs here reported) against H6N1 reassortant virus (left column), N1-Tx antigen (central column), and H11N1 PV (right column). Y axes represent the percentage of inhibition while X axes report the Log2 of the dilution. The prefix "S1." in front of every serum number is omitted to improve the readability of the legends shows an r 2 = 0.9822, and an r 2 = 0.9470, and r 2 = 0.9315 when compared to H6N1 outcomes, respectively ( Figure 5A,C).
As expected, the Bland-Altman analyses performed with 50% end point and IC 50 titers are similar (Figures 4 and 6). The Bland-Altman analysis of IC 50 titer differences showed a trend for larger differences measured by the three assays at low titers ( Figure 6). This is evident from the titers reported in Table 1; sera 19, 22, and 23 all have low or unmeasurable (<10) titers using PV and N1-Tx antigens, but a reasonable titer measured in assays using H6N1 as antigen.

| NA inhibition antibody titers measured in ELLA with H6N2 and H11N2 PV sources of antigen are similar
Since an N2-Tx virus was not available, NI antibody titers measured by ELLA using only H6N2 and H11N2 PV antigens were compared (Figures 7 and 8).
The protocol used for each antigen was the same, with the exception of HRPO concentration which was used at a higher TA B L E 1 ELLA assay outcomes (S1.1-S1.40) from the 3 different sources of antigen for N1 NA. Individual NI titers and GMTs of the results are shown  Table 2. There are no significant differences at high or low titers respectively. Interestingly the N2-specific GMTs measured using these different antigens are closer than N1-specific GMTs measured in ELLA using H11N1 PV and H6N1 antigens. Nevertheless, the higher variety in titers within the N2 panel has probably determined a lesser consistency between data, affecting the coefficient of correlation.
As for N1 titers, the Bland-Altman analysis on N2 titers show small differences in titers measured using the different antigens when either end point titers or IC 50 titers were plotted ( Figure 10).
Only one serum (10) was greater than the higher limit of agreement, when end point titers measured by H11N2 PV were compared with H6N2 titers. In fact, this is the only case in which the titer measured in an assay using H11N2 PV antigen is 2-fold higher than the titer measured with H6N2 as antigen. In addition, sera 23 and 29 were greater than the higher limit of agreement, when differences in IC 50 titers from assays using H11N2 PV and H6N1 antigens were The correlations of titers were also evaluated by Spearman rank analysis for both N1 and N2 assays. As shown in Table 3, all the results showed good correlation when either IC 50 or end point titers were evaluated.
In conclusion, measurement of NI antibody titers against N1 and N2 antigens by ELLA demonstrate comparable results using PV or Triton X-100-disrupted virions and the gold-standard H6 reassortant viruses.
F I G U R E 7 Inhibition curves showing the ability of sera to inhibit the sialidase activity across the plate. Sera from 1 to 20 (corresponding to plates 1-5) were tested in duplicate (GMTs here reported) against H6N2 reassortant virus (left column) and H11N2 PV (right column). Y axes represent the percentage of inhibition while X axes report the Log2 of the dilution. The prefix "2." in front of every serum number is omitted to improve the readability of the legend  In conclusion, NA inhibition antibody titers measured in ELLA performed with three different sources of antigen are similar and suggest lentiviral PV can be used to evaluate anti-neuraminidase responses. Further analyses with additional N1 and N2 strains and subtypes to strengthen this finding will be of value.