Identification of novel influenza A virus exposures by an improved high‐throughput multiplex MAGPIX platform and serum adsorption

Abstract Background The development of serologic assays that can rapidly assess human exposure to novel influenza viruses remains a public health need. Previously, we developed an 11‐plex magnetic fluorescence microsphere immunoassay (MAGPIX) by using globular head domain recombinant hemagglutinins (rHAs) with serum adsorption using two ectodomain rHAs. Methods We compared sera collected from two cohorts with novel influenza exposures: animal shelter staff during an A(H7N2) outbreak in New York City in 2016‐2017 (n = 119 single sera) and poultry workers from a live bird market in Bangladesh in 2012‐2014 (n = 29 pairs). Sera were analyzed by microneutralization (MN) assay and a 20‐plex MAGPIX assay with rHAs from 19 influenza strains (11 subtypes) combined with serum adsorption using 8 rHAs from A(H1N1) and A(H3N2) viruses. Antibody responses were analyzed to determine the novel influenza virus exposure. Results Among persons with novel influenza virus exposures, the median fluorescence intensity (MFI) against the novel rHA from exposed influenza virus had the highest correlation with MN titers to the same viruses and could be confirmed by removal of cross‐reactivity from seasonal H1/H3 rHAs following serum adsorption. Interestingly, in persons with exposures to novel influenza viruses, age and MFIs against exposed novel HA were negatively correlated, whereas in persons without exposure to novel influenza viruses, age and MFI against novel HAs were positively correlated. Conclusions This 20‐plex high‐throughput assay with serum adsorption will be a useful tool to detect novel influenza virus infections during influenza outbreak investigations and surveillance, especially when well‐paired serum samples are not available.


| INTRODUC TI ON
Recently, we identified the second case of A(H7N2) human infection from cats using single serum samples collected from workers/ volunteers in New York City animal shelters where A(H7N2) influenza A viruses caused the first documented cat-to-human transmission. [6][7][8] For novel influenza surveillance at the animal-human interface, we also analyzed paired sera collected from workers in a live poultry market in Bangladesh where multiple avian influenza viruses co-circulate among poultry and can cause sporadic A(H5N1) influenza virus infections in humans.
In this study, we improved the MAGPIX assay by expanding the multiplex to 20 antigens using 19 globular head domain (GH) HA1 and/or ectodomain (Ecto) rHAs from 9 subtypes of influenza A viruses (H1N1, H2N2, H3N2, H5N1, H7N2, H7N7, H7N9, H9N2, and H13N9), 2 lineages of influenza B viruses (Yamagata and Victoria lineages), and a protein A control (Table 1). Furthermore, given that past exposures to seasonal influenza viruses through infections or vaccinations in the human populations can often cause cross-reactive antibody responses to novel influenza viruses and complicate the interpretation of serologic data, 4 we also expanded serum adsorption from two to eight ectodomain rHAs to reduce cross-reactivity. This multiplex MAGPIX platform with serum adsorption can be a valuable tool to detect novel influenza virus infections during influenza surveillance and outbreak investigations.

| Human sera
Three sets of human serum samples were used in this study (Tables   2 and S1). In the A(H7N2) study, single convalescent serum specimens (S2) were collected from 119 shelter workers and volunteers (18- To investigate the antibody profiles in persons who have not  (Table S1).  and a protein A control as described previously (Table 1). 4 The antigens were either in-house made at CDC or obtained from

| Twenty-plex MAGPIX assay
International Reagent Resource and ThermoFisher Scientific (PA, USA) (Table S2); the purity, trimerization, and receptor-binding activity of in-house made rHAs were confirmed as previously described. [9][10][11] The serum samples were tested at 1:40 dilution in duplicate by using in-house made phycoerythrin-conjugated protein A (RPE-PA) and read by a Bio-Plex MAGPIX Multiplex Reader.
Median fluorescence intensity (MFI) was obtained and analyzed with Bio-Plex Manager Software. 4
Ecto in A(H7N2) study (except one serum showed MN ≥40 and MFI = 2239) (18 out of 119) and MFIs ≥1300 against H5.Ind.05 GH in A(H5N1) study (14 out of 29) in the initial test were first adsorbed with two ectodomain rHAs (2-Ads) from CA/09 and Perth/09 conjugated to latex beads as described previously 4 ( Figure 1 and Table 3). Selected S2 samples for 2-Ads showed a range of MN titers against homologous viruses from 5 to 80 ( Figure S1). In 8

| HI and MN assays
Single S2 human serum samples collected from A(H7N2) study in New York City and paired S1 and S2 human serum samples collected from A(H5N1) study in Bangladesh were tested by HI and MN assays.
The modified HI assay using horse erythrocytes (1% v/v) was performed as described previously to detect antibody responses to H5 and H7 viruses. 13 Sera were heat-inactivated at 56°C for 30 minutes, tested for non-specific agglutinins, and adsorbed with packed horse erythrocytes as needed. Non-specific inhibitors in the sera were removed by incubation with receptor-destroying enzyme at 37°C for 18-20 hours, followed by heat inactivation prior to standard protocol, 14 except that hemagglutination of horse erythrocytes was read after 60-minute incubation.
MN assays were performed as described previously. 14

| Data analysis
Pearson's correlation coefficient (Pearson's r), paired Wilcoxon, and unpaired Mann-Whitney test were performed using GraphPad Prism 5; P values of less than 0.05 were considered statistically significant.

| RE SULTS AND D ISCUSS I ON
We developed a 20-plex MAGPIX assay and expanded serum adsorption by using a cocktail of 8 rHA-conjugated beads to reduce the effects of cross-reactive antibodies and to improve assay performance. The platform of this 20-plex MAGPIX, test flow, and serum adsorption is described in Figure 1. The high-throughput MAGPIX platform can generate an antibody profile to 19 influenza antigens simultaneously by using very small volume of serum samples (<10 µL), and the similar platform showed a 4-log10 linear range of sensitivity. 3

| Correlation between MN titers and MFIs in 20-plex MAGPIX
Single S2 human serum samples collected from the recent New York City animal shelter A(H7N2) outbreak and paired S1 and S2 human serum samples collected from live poultry market workers in Bangladesh were analyzed by the 20-plex MAGPIX assay in the initial test ( Figure 1A,B Next, we grouped S2 sera from the A(H7N2) study and both S1 and S2 serum samples from the H5N1 study based on MN titers (MN ≥40 group or MN <40 group) against exposed HA subtypes: We also analyzed correlation between MN titers against A(H7N2) (or A(H5N1)) and MFIs against 20 antigens measured by  Figure S2). These results suggest that among the antibody profiles to the 19 HAs (from 11 influenza subtypes) measured by MAGPIX, MFI against rHA from the exposed novel HA subtype had the highest correlation to MN titers to the same virus. It is consistent with our previous observation during the first wave of A(H1N1)pdm09 pandemic. 16

| Serum adsorption by using either 2 or 8 rHAs from A(H1N1) and A(H3N2) to reduce crossreactivity and improve the detection of subtypespecific responses to novel HA subtypes
Complex exposures to seasonal influenza virus(es) and/or vaccine(s) induce both within-subtype (homosubtypic) and crosssubtype (heterosubtypic) reactivity. The presence of cross-reactive antibodies against unexposed subtype HA has been described when ectodomain HAs are used. 16,17 The use of GH HA1 rHAs that lack the antigenically conserved HA stalk can reduce such cross-reactivity. 18 In this study, we observed high MFIs against We have previously demonstrated that cross-reactivity could be reduced by incorporating a serum adsorption step prior to the MAGPIX or MN assays, and the remaining signals from post-adsorption samples represent HA subtype-specific responses. 4,7 Thus, serum adsorptions were performed in the current study to remove cross-reactive antibodies to improve sensitivity and spec-   Table 3). We selected the 6 additional rHAs for adsorption from viruses representative of antigenic clusters of seasonal A(H1N1) 20 and A(H3N2) viruses (  (Tables 2 and 3, and Figure 1D). To this end, following 8-Ads, most cross-reactive antibodies were removed, with MFIs to unexposed HAs reduced to almost background level (MFI <500) for sera from both the A(H7N2) and A(H5N1) studies, MFI value less than 500 was considered as "background" in our current platform 4 ; however, the strain-specific signals against potential exposed H7 and H5 HAs remained at similar levels ( Figure 4A,B,D, and E).  Table 2). In this study, since we performed 2-Ads and 8-Ads by using cocktailed H1 (group 1) and H3 (group 2) rHAs, we did observe some reductions in MFIs against exposed novel subtype HAs (Figure 3 and 4). It will be interesting to perform HA group-specific rHA adsorption in the future studies.     Interestingly, we found that there was a negative correlation (negative Pearson's r value) between age and exposed novel HA antigen H7.NY. 16 Ecto for New York A(H7N2) study (r = −0.362, Figure 5J), in contrast to positive (r = 0.105, Figure 5E) and low

| Correlation between age and MFIs against rHAs from various subtype influenza viruses
Pearson's correlation coefficient (r = 0.022, Figure 5O) for agematched control sera and sera from the Bangladesh A(H5N1) study, respectively (Table S3). Similarly, there was also a negative correlation (negative Pearson's r value) between age and exposed H5 antigens for sera collected in the Bangladesh A(H5N1) study (r = −0.500, Table S3), in contrast to positive correlation for age-matched control sera (r = 0.145, Figure 5D) and for animal shelter staff in the New York A(H7N2) study (r = 0.266, Figure 5I and Table S3) were observed ( Figure S3).
In general, antibodies to novel virus HA in unexposed populations increase with age and vary by geographical location due to cross-reactivity between the novel virus and seasonal influenza viruses through past exposure. 18 Further, when sera collected in the A(H7N2) study in 2016-2017 were analyzed for MN titers against A(H1N1)pdm09, negative correlation between age and A(H1N1)pdm09 MN titers was also observed ( Figure S4) and was consistent with the results from MAGPIX ( Figure 5F).
Our results suggest that analysis of correlation between age and MFIs against multiple subtype HAs can provide important information to determine potential exposures to novel subtype HA influenza viruses in a sub-population. This also suggests that age-matched controls selected from the same region and time period are necessary for proper analysis of MAGPIX results ( Figure 5, Tables S1 and S3). For example, the age-related antibody baseline level to A(H1N1) pdm09 in recent years would be much higher than those reported in 2009, 29 due to extensive exposure to A(H1N1)pdm09 virus through infection or vaccination post-2009. 16 If we can determine whether low levels of age-specific, subtype cross-reactive antibodies are present in the population, single convalescent-phase serum specimens may be sufficient to determine a recent exposure to novel subtype influenza. 7,29 Our study has several limitations: (a) Given the availability of only small numbers of sera, we could not determine cutoff value for potential exposure to novel subtype HA for a given population; (b) without paired sera samples demonstrating sera conversion, we could not rule out the possibility of exposure to novel subtype influenza virus(es) from past infections rather than the current outbreak.
The twenty-plex MAGPIX assay combined with 8-Ads and analysis of correlation between age and MFIs can be used to screen potential exposed novel HA subtype virus for further analysis.
This high-throughput platform will streamline serologic analysis of human or animal serum samples that are collected from areas where multiple novel HA subtype viruses co-circulate. This platform can be further expanded to incorporate influenza virus rHAs from all 18 HA subtypes, which could be used to determine exposures to novel influenza viruses in humans, or to identify new influenza virus host using animal sera instead of labor-intensive ELISA. 30 In summary, the 20-plex MAGPIX assay combined with 8-Ads and correlation analysis between age and MFIs will be a useful tool to identify exposure to novel HA subtype influenza viruses in highrisk populations, especially when rRT-PCR, virus isolation, or wellpaired sera are not available.

CO N FLI C T O F I NTER E S T S TATEM ENT
All other authors report no potential conflicts. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

| DISCL AIMER
Trade names are only used for identification, and they are not en-

ACK N OWLED G EM ENTS
We thank Dr Katherine Sturm-Ramirez and her colleagues in Centre for Communicable Diseases, The International Centre for Diarrhoeal Disease Research, Bangladesh (icddr,b), Dhaka, Bangladesh.
We thank Lauren Horner, Makeda Kay, and Heather Tatum for managing the serum samples.