L226Q Mutation on Influenza H7N9 Virus Hemagglutinin Increases Receptor-Binding Avidity and Leads to Biased Antigenicity Evaluation

The HI assay is a standard method for profiling the antigenic characterization of influenza viruses. Suspected antigenic changes based on HI divergency in H7N9 viruses during the 2016-2017 wave prompted the recommendation of new H7N9 candidate vaccine viruses (CVVs). In this study, we found that the L226Q substitution in HA of A/Guangdong/17SF003/2016 (H7/GD16) increased the viral receptor-binding avidity to red blood cells with no impact on the antigenicity of H7N9 virus. Although immune sera raised by an earlier vaccine strain (H7/AH13) showed poor HI titers against H7/GD16, the H7/AH13 immune sera had potent cross-neutralizing antibody titers against H7/GD16 and could provide complete passive protection against H7N9/GD16 virus challenge in mice. Our study highlights that receptor-binding avidity might lead to biased antigenic evaluation by using the HI assay. Other serological assays, such as the microneutralization (MN) assay, should be considered a complementary indicator for analysis of antigenic variation and selection of influenza CVVs.

ABSTRACT

Since the first outbreak in 2013, the influenza A (H7N9) virus has continued emerging and has caused over five epidemic waves. Suspected antigenic changes of the H7N9 virus based on hemagglutination inhibition (HI) assay during the fifth outbreak have prompted the update of H7N9 candidate vaccine viruses (CVVs). In this study, we comprehensively compared the serological cross-reactivities induced by the hemagglutinins (HAs) of the earlier CVV A/Anhui/1/2013 (H7/AH13) and the updated A/Guangdong/17SF003/2016 (H7/GD16). We found that although H7/GD16 showed poor HI cross-reactivity to immune sera from mice and rhesus macaques vaccinated with either H7/AH13 or H7/GD16, the cross-reactive neutralizing antibodies between H7/AH13 and H7/GD16 were comparably high. Passive transfer of H7/AH13 immune sera also provided complete protection against the lethal challenge of H7N9/GD16 virus in mice. Analysis of amino acid mutations in the HAs between H7/AH13 and H7/GD16 revealed that L226Q substitution increases the HA binding avidity to sialic acid receptors on red blood cells, leading to decreased HI titers against viruses containing HA Q226 and thus resulting in a biased antigenic evaluation based on HI assay. These results suggest that amino acids located in the receptor-binding site could mislead the evaluation of antigenic variation by solely impacting the receptor-binding avidity to red blood cells without genuine contribution to antigenic drift. Our study highlighted that viral receptor-binding avidity and combination of multiple serological assays should be taken into consideration in evaluating and selecting a candidate vaccine virus of H7N9 and other subtypes of influenza viruses.

IMPORTANCE The HI assay is a standard method for profiling the antigenic characterization of influenza viruses. Suspected antigenic changes based on HI divergency in H7N9 viruses during the 2016-2017 wave prompted the recommendation of new H7N9 candidate vaccine viruses (CVVs). In this study, we found that the L226Q substitution in HA of A/Guangdong/17SF003/2016 (H7/GD16) increased the viral receptor-binding avidity to red blood cells with no impact on the antigenicity of H7N9 virus. Although immune sera raised by an earlier vaccine strain (H7/AH13) showed poor HI titers against H7/GD16, the H7/AH13 immune sera had potent cross-neutralizing antibody titers against H7/GD16 and could provide complete passive protection against H7N9/GD16 virus challenge in mice. Our study highlights that receptor-binding avidity might lead to biased antigenic evaluation by using the HI assay. Other serological assays, such as the microneutralization (MN) assay, should be considered a complementary indicator for analysis of antigenic variation and selection of influenza CVVs.

INTRODUCTION

Since the first case of infection in humans in 2013, the influenza A (H7N9) virus has continued emerging and has caused over five epidemic waves (1). During the fifth wave starting in October 2016, highly pathogenic avian influenza (HPAI) virus H7N9 variants emerged and caused 32 human infections, with a fatality rate of 44% (2, 3). HPAI H7N9 is more pathogenic in chickens, mice, and ferrets than low-pathogenic avian influenza (LPAI) virus H7N9 and can transmit among ferrets via respiratory droplets (4, 5). The dual receptor-binding property of the HPAI H7N9 virus has made it a great threat to humans (3, 4, 6).

Hemagglutinin (HA), the major surface glycoprotein of influenza virus, is the primary target of the protective antibody response to influenza virus (7). New variant strains can arise when mutations accumulate on the antigenic sites of HA and allow it to escape recognition by existing neutralizing antibodies, a process known as antigenic drift. The hemagglutination inhibition (HI) assay is the most commonly used method to analyze the antigenicity variation of influenza viruses (8). Previous studies have shown that antisera raised against A/Anhui/1/2013 (AH13), the earlier candidate vaccine virus (CVV), have a >4-fold reduction in HI titers against the new emerging HPAI H7N9 viruses in the 2016-2017 outbreak (3, 9). Additional H7N9 CVVs were recommended for development to maximize influenza pandemic preparedness (10).

In this study, by generating recombinant “7 + 1” influenza viruses containing H7 HA either from AH13 or from a new CVV derived from A/Guangdong/17SF003/2016 (GD16) in a A/PR/8/1934 (PR8) background, we made a comprehensive analysis of the antigenic divergence between the HAs of AH13 and GD16 not only by HI assay but also by microneutralization (MN) assay and cross-protection in vivo. Furthermore, the key amino acid residues that play a critical role in receptor-binding avidity and thus lead to the biased antigenic evaluation based on HI assay were also determined.

RESULTS

Biased HI results lead to overestimated antigenic drift of A/Guangdong/17SF003/2016.

To analyze the HA antigenic distinction between the earlier and the existing H7N9 CVVs, we generated two recombinant “7 + 1” viruses containing HA of A/Anhui/1/2013 (H7/AH13) or A/Guangdong/17SF003/2016 (H7/GD16) with the other seven segments from A/PR/8/1934 (PR8) virus. We first compared the cross-reactive HI and neutralizing antibody between H7/AH13 and H7/GD16 by use of immune sera collected from BALB/c mice and rhesus macaques vaccinated with these two inactivated “7 + 1” viruses. Compared to the H7/AH13 virus, the H7/GD16 virus displayed 17.4- and 7.9-fold reductions of HI titers (P = 0.0003 and 0.0218, respectively) to H7/AH13 immune sera from vaccinated mice and rhesus macaques, respectively (Fig. 1A). However, when we used the MN assay, H7/GD16 displayed MN titers similar to those of H7/AH13 to H7/AH13 immune sera from both mice and rhesus macaques (P = 0.7731 and 0.4194, respectively) (Fig. 1B). Since a 4-fold HI or MN difference is considered a significant antigenic change (11), the results from the HI assay indicate that H7/GD16 is antigenically distinct from H7/AH13. However, the competent cross-reactive neutralizing antibody titers detected by MN assay suggest that H7/GD16 was antigenically similar to H7/AH13. Furthermore, it was interesting that compared to H7/AH13 virus, H7/GD16 also exhibited 7.5-fold lower HI titers (P = 0.002 and 0.0356, respectively) to homologous H7/GD16 immune sera from vaccinated mice and rhesus macaques (Fig. 1C). Similar to H7/AH13 immune sera, H7/GD16 showed comparable MN titers with H7/AH13 virus to H7/GD16 sera from vaccinated mice and rhesus macaques (P = 0.0605 and 0.2847, respectively) (Fig. 1D). These results indicate that H7/GD16 exhibited poor HI reactivity but high neutralizing reactivity to both H7/AH13 and H7/GD16 immune sera from vaccinated mice and rhesus macaques.

FIG 1.

HI and MN titers of H7/AH13 and H7/GD16 viruses to immune sera. (A) HI titers of H7/AH13 immune sera from mice (n = 6 animals per group) and rhesus macaques (n = 3 animals per group) to H7/AH13 and H7/GD16 viruses. (B) MN titers of H7/AH13 immune sera from mice and rhesus macaques to H7/AH13 and H7/GD16 viruses. (C) HI titers of H7/GD16 immune sera from mice and rhesus macaques to H7/AH13 and H7/GD16 viruses. (D) MN titers of H7/GD16 immune sera from mice and rhesus macaques to H7/AH13 and H7/GD16 viruses. Immune sera were collected from mice or rhesus macaques 4 weeks after a second immunization. Data are representative values from three independent assays. “ND” (not detectable) indicates the NC (negative control) group. *, P < 0.05; ***, P < 0.001; ns, not significant.

We next evaluated the in vivo cross-protection of H7/AH13 vaccination against H7N9/GD16 virus challenge by passive immune serum transfusion. The results showed that H7/AH13 and H7/GD16 immune sera were equally protected mice against 10 half-maximal murine lethal dose (MLD₅₀) of H7N9/GD16 challenge (Fig. 2). Mice that received 150 μl of immune sera were better protected than those received 50 μl of immune sera, showing a dose-dependent effect.

FIG 2.

Passive transfer of immune sera protected against challenges with H7N9/GD16. Groups of mice (n = 6) were transferred with a volume of 50 or 150 μl of immune sera from donor mice vaccinated with H7/AH13 or H7/GD16 and were intranasally (i.n.) infected with 10 MLD₅₀ of the H7N9/GD16 virus 2 h posttransfection. The recipient mice were monitored daily for their body weight changes (A) and survival rates (B) for 14 days. “*” and “**” represent the significant level between passively transferring 50 μl and 150 μl of H7/AH13 immune sera (P < 0.05 and P < 0.01, respectively). “#” represents the significant level between passively transferring 50 μl and 150 μl of H7/GD16 immune sera (P < 0.05).

In summary, the antigenic variation between H7/AH13 and H7/GD16 revealed by the poor cross-reactive HI titers were not consistent with the potent cross-reactive neutralizing antibodies and effective in vivo cross-protection of H7/AH13 vaccination against H7/GD16 virus.

HA L226Q substitution decreases readouts of HI titers but has no impact on MN titers.

To explore the molecular mechanism responsible for the decreased readouts of HI titers against H7/GD16 virus, we made an amino acid sequence alignment between the HAs of H7/AH13 and H7/GD16. In total, there are 15 amino acid differences between these two HAs and among them, 9 amino acids are in the globular head regions (Fig. 3A). Four mutations (A122P, S128N, G129E, and A135V [H3 numbering used throughout]) are adjacent to the putative antigenic site A, and the other five mutations (L226Q, R172K, K173E, L177I, and M236I) are not in the known antigenic sites (Fig. 3B). Two mutations (L226Q and A135V) are located in the receptor-binding site (Fig. 3B).

FIG 3.

Sequence alignment and the location of differential HA residues on the HA structure between the H7/AH13 and H7/GD16 viruses. (A) Sequence alignment of HA between the H7/AH13 and H7/GD16 viruses. (B) H7/AH13 HA trimer (Protein Data Bank code 4KOL) showing the position of each amino acid substitution on the HA head of H7/GD16. One monomer surface is shown in light gray, and the other two monomers are shown in the cartoon in dark grey. The crystal structure was drawn by using PyMol software. RBS, receptor-binding site.

To identify which mutation(s) might be responsible for the decreased HI titer readouts of H7/GD16, we generated a panel of recombinant H7/GD16 viruses containing single or multiple amino acid mutations according to the sequence of H7/AH13 HA and then detected the cross-reactive HI titers of H7/AH13 immune sera against these mutant viruses (Table 1). Among all 10 mutants, the virus containing the Q226L mutation in the GD16 HA background (referred to as H7/GD16/Q226L) led to a 6.0- to 8.0-fold increase in HI titers compared to viruses containing Q226 in HA (Table 1). To reverse confirm the HI increase caused by the Q226L substitution in HA of H7/GD16, we further generated an L226Q mutation in the AH13 HA background (referred as H7/AH13/L226Q). As expected, H7/AH13/L226Q showed a 6.0- to 8.0-fold decrease in HI titers compared to parental H7/AH13 virus possessing HA L226 in reacting to H7/AH13 immune sera (Table 1). The viruses containing multiple mutations, including N128S + E129G and P122A + N128S + E129G + V135A, had only a ≤2-fold increase in the HI titer readouts of H7/AH13 immune sera (Table 1). Other single or multiple mutations in H7/GD16 HA showed no significant change of HI titers to H7/AH13 immune sera (Table 1).

TABLE 1.

HI titers changes of H7 viruses with HA mutations in responding to immune sera

Virus tested	H7/AH13 immune sera				H7/GD16 immune sera
	Mouse		Macaque		Mouse		Macaque
	HI titera	Fold changeb	HI titera	Fold changeb	HI titera	Fold changeb	HI titera	Fold changeb
H7/GD16 (Q226)	40		60		40		40
H7/GD16/Q226L	320	8.0	480	8.0	320	8.0	240	6.0
H7/GD16/P122A	40	1.0	60	1.0	40	1.0	40	1.0
H7/GD16/N128S+E129G	60	1.5	80	1.3	40	1.0	40	1.0
H7/GD16/V135A	40	1.0	60	1.0	40	1.0	20	0.5
H7/GD16/P122A+N128S+E129G+V135A	80	2.0	120	2.0	60	1.5	60	1.5
H7/GD16/K172R+E173K	30	0.75	60	1.0	20	0.5	40	1.0
H7/GD16/I177L	20	0.5	60	1.0	30	0.8	20	0.5
H7/GD16/K172R+E173K+I177L	20	0.5	40	0.7	40	1.0	30	0.8
H7/GD16/I236M	40	1.0	60	1.0	40	1.0	20	0.5
H7/AH13 (L226)	320	8.0	640	10.7	320	8.0	240	6.0
H7/AH13/L226Q	40	1.0	80	1.3	60	1.5	30	0.8

^aHI assays were operated using immune sera from mice or rhesus macaques 4 weeks after the second immunization. Chicken RBCs were used in the assays. Data are representative values from three independent assays.

^bObserved fold change in HI titer compared with H7/GD16.

To verify the above-described HI results obtained by using chicken red blood cells (RBCs), we further redetermined the HI titers against viruses containing the L226Q and Q226L bidirectional mutations by use of RBCs from other species, including turkey, human, and guinea pig. Consistent with the results in Table 1, the HI titers against virus H7/GD16/Q226L detected by these three kinds of RBCs had 4.0- to 10.7-fold increases over that against the parental H7/GD16 virus, while the HI titers against virus H7/AH13/L226Q had 4.0- to 8.0-fold decreases from that against the parental H7/AH13 virus (Table 2). Collectively, the above-described results showed that the L226Q mutation in the receptor-binding site of H7 HA might play a crucial role in the reduction of the readouts of HI titers.

TABLE 2.

HI titers of H7 viruses in responding to immune sera using different species RBCs

	HI titer of H7/AH13 immune seraa						HI titer of H7/GD16 immune seraa
	Mouse			Macaque			Mouse			Macaque
Virus tested	Turkey RBCs	Human RBCs	Guinea pig RBCs	Turkey RBCs	Human RBCs	Guinea pig RBCs	Turkey RBCs	Human RBCs	Guinea pig RBCs	Turkey RBCs	Human RBCs	Guinea pig RBCs
H7/AH13(L226)	240	480	320	480	640	640	240	480	320	240	240	320
H7/AH13/L226Q	40	120	60	60	80	120	40	120	80	30	60	80
H7/GD16 (Q226)	40	80	40	60	120	80	30	80	80	40	60	80
H7/GD16/Q226L	320	320	320	640	640	480	240	480	640	320	320	320

^aHI assays were operated using immune sera from mice or rhesus macaques 4 weeks after the second immunization. Turkey, guinea pig, or O-type human RBCs were used in the HI assays. Data are representative values from three independent assays.

To investigate whether HA L226Q substitution also has an impact on the readouts of neutralizing antibody titers, we compared the readouts of the MN assay in the H7/AH13 and H7/GD16 immune sera against viruses containing HA L226 or Q226. H7/AH13 HA (containing L226) displayed less than 1.5-fold changes in MN titers in reacting to either H7/AH13 or H7/GD16 immune sera from mice and rhesus macaques compared to H7/AH13/L226Q HA (Table 3). Similarly, the H7/GD16 HA containing Q226 displayed <2-fold changes in MN titers in reacting to either H7/AH13 or H7/GD16 immune sera from mice and rhesus macaques compared to H7/GD16/Q226L HA virus (Table 3). Since a 4-fold MN titer change is considered a significant difference (11, 12), these results demonstrated that the L226Q mutation in H7/GD16 HA has no significant impact on the readouts of MN titers.

TABLE 3.

MN titers changes of H7 viruses with either L226 or Q226 in responding to immune sera

Virus tested	H7/AH13 immune sera				H7/GD16 immune sera
	Mouse		Macaque		Mouse		Macaque
	MN titera	Fold changeb	MN titera	Fold changeb	MN titera	Fold changeb	MN titera	Fold changeb
H7/AH13	640		1,280		853		427
H7/AH13/L226Q	693	1.1	1,280	1.0	1,280	1.5	533	1.2
H7/GD16	533	0.8	1,067	0.8	1,706	2	853	2
H7/GD16/Q226L	853	1.3	1,493	1.2	1,493	1.8	533	1.2

^aMN assays were carried out using immune sera from mice or rhesus macaques 4 weeks after the second immunization. Data are representative values from three independent assays.

^bObserved fold change in MN titer compared with H7/AH13.

HA L226Q substitution increases the receptor-binding avidity to RBCs, with no impact on virus ability to bind to antibodies and MDCK cells.

To investigate the reason why HA L226Q substitution decreased the sensitivity of the detection of HI titers while having no impact on the detection of MN titers, we first detected the virus binding avidity of viruses possessing HA L226 and HA Q226 to chicken RBCs. Chicken RBCs were pretreated with different amounts of receptor-destroying enzyme (RDE) to remove cell surface sialic acid receptors, and then the hemagglutination of viruses to the RDE-treated RBCs was detected. H7/GD16 (possessing Q226 HA) bound to RBCs treated with 6.0-fold-higher RDE concentrations than H7/AH13 (possessing L226 HA) (P = 0.0015) (Fig. 4A). H7/GD16 bound to RBCs treated with 4.5-fold-higher RDE concentrations than virus containing the Q226L substitution (H7/GD16/Q226L) (P = 0.0038) (Fig. 4A). H7/AH13/L226Q had 4.7-fold (P = 0.015) increased agglutination capability for RDE-treated RBCs compared to that of parental H7/AH13 possessing L226 HA (Fig. 4A). We analyzed the correlation between the cellular receptor-binding avidity and HI titers of polyclonal antibodies, and the results showed that there was a strong negative correlation (r = −0.8499; P < 0.0001) between them (Fig. 4B). We also detected the amounts of H7 viruses by detecting nucleoprotein (NP) genes at different dilutions with real-time reverse transcription (RT)-PCR and found that all four H7 viruses had similar amounts of virus at the same hemagglutination titers (Fig. 4C). These results suggest that viral receptor-binding avidity to RBCs had no effect on virus hemagglutination titers.

FIG 4.

The L226Q HA mutation increases the receptor-binding avidity of H7 viruses to RBCs. (A) Relative viral receptor-binding avidities were determined by hemagglutination of red blood cells pretreated with 0.1 to 270 μg/ml of RDE (Vibrio cholerae neuraminidase [Sigma]). Data are expressed as the maximal amount of RDE that allowed full agglutination. Means and SEMs from quintuplicate samples are shown. (B) HI titers of H7 viruses to H7/AH immune mouse sera are plotted versus RBC receptor-binding avidity of H7 viruses. Means are represented as dots, and correlation between HI titers and receptor-binding avidity was determined by Pearson correlation test. (C) The quantity of each virus C_T value of the NP gene of 8-HAU H7 viruses serially diluted 10-fold and titrated by using real-time RT-PCR. (D and E) Biolayer interferometry curves of binding of H7 viruses to the receptor analogs 3SLNLN (D) and 6SLNLN (E). Streptavidin biosensors were immobilized with sialylated glycan receptors and reacted with viruses in the presence of NA inhibitors for 3,000 s at 30 °C. *, P < 0.05; **, P < 0.01.

We performed the biolayer interferometry (BLI) assay to compare the sialylated glycan binding avidities between H7/AH13, H7/AH13/L226Q, H7/GD16, and H7/GD16/Q226L. Among these four viruses, H7/GD16 showed the highest levels of binding affinity to both 3SLNLN and 6SLNLN glycans (see Materials and Methods), while H7/AH13 showed the lowest binding affinity to these two glycans (Fig. 4D and E). Consistent with the results in the hemagglutination assay with RDE-treated RBCs, the H7/GD16/Q226L virus exhibited a decreased binding affinity to both 3SLNLN and 6SLNLN glycans compared with that of the parental H7/GD16 virus possessing Q226 (Fig. 4C and D). In contrast, the H7/AH13/L226Q virus exhibited an increased binding affinity to these two glycans compared with that of the parental H7/AH13 virus possessing L226 (Fig. 4C and D). As controls, A/California/04/2009 (H1N1) (referred to as H1N1) was shown to preferentially bind the 6SLNLN glycans and recombinant A/Vietnam/1/2005 (H5N1) in the PR8 background (referred to as H5N1) to bind the 3SLNLN (Fig. 4C and D). These results indicate that L226Q HA substitution in HA increased the H7 virus binding affinity to glycan receptors.

To investigate whether H7 viruses possessing L226 HA or Q226 HA have any difference in binding to serum polyclonal antibodies, we performed an enzyme-linked immunosorbent assay (ELISA) with plates coated with purified inactivated H7/AH13 (L226), H7/AH13/L226Q, H7/GD16 (Q226), and H7/GD16/Q226L (Fig. 5). As a control, the PR8 virus was also used to coat the plates. The results showed that both H7/AH13 and H7/GD16 immune sera from mice and macaques bound similarly to viruses possessing either L226 HA or Q226 HA (Fig. 5A to D). We also performed an ELISA with monoclonal antibody (MAb) MDEI8852 (13), which recognizes the conserved stalk region of HA, to verify that equal amounts of inactivated viruses were used in these assays (Fig. 5E). Therefore, these data indicate that the L226Q mutation does not significantly influence the overall binding of serum polyclonal antibodies to HA protein as a whole.

FIG 5.

The L226Q HA mutation has no significant effect on antibody binding. Direct antibody binding to H7 viruses was determined by ELISA using H7/AH13 immune sera from mice (A), H7/AH13 immune sera from rhesus macaques (B), H7/GD16 immune sera from mice (C), H7/GD16 immune sera from rhesus macaques (D), or the MEDI8852 MAb (E), which binds to a conserved region of the HA stalk. The means and SDs of triplicate samples are shown.

Finally, to explore the reason why the L226Q substitution has a limited impact on the detection of neutralizing antibody titers by MN assay, we determined the virus binding avidity to MDCK cells by ELISA-based MDCK cell binding assays. The results showed that all H7 viruses had comparable binding avidities to MDCK cells no matter whether they possessed L226 or Q226 in the HA (Fig. 6). PR8 strain was used as a control and showed significantly higher binding avidity to MDCK cells than the four H7 viruses (P < 0.0001). The finding that HA L226Q mutation had no influence on the virus binding avidity to MDCK cells might lead to comparable neutralizing antibody titers against viruses possessing HA L226 and HA Q226.

FIG 6.

L226Q mutation does not affect H7 viruses binding to MDCK cells. H7 viruses were absorbed to chilled MDCK cells at a hemagglutination titer of 8 at 4°C for 1 h, and the infected cells were washed or left untreated. After 6 h of incubation at 37°C, the infection was determined by ELISA. The binding avidity was calculated according to the formula (OD₄₅₀ value of washed wells/OD₄₅₀ value of untreated wells).

DISCUSSION

During the fifth epidemic, influenza A (H7N9) viruses (3) diverged into two distinct genetic lineages, the Pearl River Delta and Yangtze River Delta lineages (1). The reduced HI cross-reactivity suggests that viruses from the Yangtze River Delta lineage are antigenically distinct from existing H7N9 CVVs (3, 9). The WHO recommended several additional H7N9 CVVs for vaccine development (10). In a recently published study of ours, using “6 + 2” recombinant viruses, we found that the poor cross-HI reactivity between the earlier H7N9 CVV A/Anhui/1/2013 (AH/13) and an updated CVV derived from the highly pathogenic A/Guangdong/17SF003/2016 (GD/16) was not consistent with the potent cross-reactive neutralizing antibodies and in vivo cross-protection between them (14). We also found that the increased HA receptor-binding avidity to red blood cells (RBCs) was responsible for the insensitive detection of the HI antibody titers against GD/16 (14). It has been reported that the N9 neuraminidase (NA) possesses two sialic acid-binding motifs, which including a sialic acid-binding motif that maps to the catalytic active site of the neuraminidase and a second sialic acid-binding motif that corresponds to the hemadsorption (Hb) site (15). To minimize the receptor-binding activity of N9 neuraminidase, we rescued recombinant “7 + 1” viruses containing the HA gene derived from AH/13 or GD/16 and the other seven segments of PR8 in this study. In the N1 neuraminidase gene background, our results confirmed that the H7 HA of A/Guangdong/17SF003/2016-like strains in the fifth wave of H7N9 did not undergo the significant antigenic change, which was biased evaluated by HI assay due to its increased receptor avidity to RBCs.

The HI assay is based on the antibodies binding to the viral HA globular head region to competitively block the binding of HA to cellular sialic acid receptors. Therefore, the readouts of HI titers might be reduced by the increased HA binding avidity to sialic acid receptors on the surfaces of RBCs since more antibodies should be required to interfere with the HA binding to RBCs. Previous studies had reported that several HA mutations had an impact on the receptor-binding avidity and thus influence the HI titers, such as HA E158K of H1N1, HA I226V of H3N2, and HA A180T/A of H9N2 influenza virus (8, 16, 17). To investigate the underlying molecular mechanism of the increased receptor-binding avidity of H7/GD16, we generated a series of recombinant viruses containing single or multiple amino acid exchange substitutions based on the sequence alignment of the HA globular head between H7/AH13 and H7/GD16 viruses and found that the L226Q substitution played a crucial role in the decreased readouts of HI titers against viruses containing HA Q226 by enhancing their HA receptor-binding avidity to RBCs. Meanwhile, our results showed that viruses containing the HA Q226 substitution had comparable cross-reactive neutralizing antibody titers and effective in vivo cross-protection against the viruses containing HA L226 and vice versa, which suggests that L226Q has no impact on the HA antigenicity of H7/GD16. Therefore, it is necessary to take HA receptor-binding avidity of the influenza virus into account when interpreting the HI data for antigenic analysis.

Compared to the HI assay, the MN assay is time-consuming and labor-intensive. However, the MN assay is more sensitive than the HI assay because it detects all functional antibodies that inhibit viral replication in MDCK cells, while the HI assay only measures antibodies preventing RBC agglutination (18). It has been demonstrated that the detection of neutralizing antibodies by the MN assay is less influenced by receptor-binding avidity than the HI assay since different sialic acid isoforms might be used by viral HA to initiate recognition and binding to MDCK cells and RBCs (19–21). In this study, our results indicate that the L226Q mutation had no significant impact on the binding avidity of H7 viruses to MDCK cells (Fig 6). We found that viruses containing L226 or Q226 exhibited comparable MN titers in H7N9 CVV immune sera from mice and rhesus macaques. These data are consistent with published research in which HA containing either L226 or Q226 (position 235 based on H7 numbering) could equally be neutralized by A/Anhui/1/2013 immune sera from guinea pigs using pseudovirus-based neutralization assays (22). Taken together, these results suggest that the MN assay should be taken as a necessary complementary indicator for the antigenicity evaluation based on the HI assay.

Furthermore, it is necessary to be cautious in evaluating the antigenicity of those H7N9 viruses with reduced receptor-binding avidity. The virus isolated from the H7N9-infected dead subject in 2019 in China showed a mutation, A160T (A151T based on H7 numbering), on HA (23), which can donate a suspected N-glycosylation site, 158-NAT, in antigenic B site. It has been reported that additional glycosylation near receptor-binding sites can limit viral binding avidity to RBCs (24). HI readout would be increased when viral isolates bind to RBCs with reduced avidities. In this case, novel isolates can be falsely defined as antigenically neutral, even though they are truly antigenic distinct (8).

In conclusion, we found that L226Q substitution in HA of A/Guangdong/17SF003/2016 (H7N9) increased the virus binding avidity to red blood cells and led to the decreased readouts of cross-HI titers in the earlier H7N9 CVV (A/Anhui/1/2013) immune sera while having no effect on detection of cross-reactive neutralizing antibodies and in vivo cross-protection. These results suggest that amino acids located in the receptor-binding site could mislead the analysis of antigenic variation by solely affecting the receptor-binding avidity to red blood cells without genuine contribution to antigenic drift. Therefore, our study proposes that the mutations located in receptor-binding sites should be monitored and HA receptor-binding avidity should be taken into account when analyzing the antigenic variation of new emerging influenza viruses based on the HI assay. Furthermore, other serological assays, such as the MN assay, are necessarily considered a complementary method when interpreting the HI data for candidate vaccine virus selection and vaccine efficacy evaluation of influenza viruses.

MATERIALS AND METHODS

Viruses and cells.

Recombinant influenza H7 viruses containing HA from A/Anhui/1/2013 and NA from PR8 (referred to as H7/AH13), HA from A/Guangdong/17SF003/2016 and NA from PR8 (referred to as H7/GD16), or HA and NA from A/Guangdong/17SF003/2016 (referred to as H7N9/GD16) were created in the genetic background of PR8 as previously described (25). The DNA sequences for the HA and NA of H7N9 viruses were downloaded from the GISAID database. The GISAID accession numbers for the HA of A/Anhui/1/2013, HA of A/Guangdong/17SF003/2016, and NA of A/Guangdong/17SF003/2016 are EPI439507, EPI1075387, and EPI1010187, respectively. The multibasic amino acid motif RKRT in the cleavage site was removed from the HA gene of A/Guangdong/17SF003/2016. The HA and NA genes were synthesized de novo and were cloned into a pM reverse genetics plasmid as described previously (26, 27). To generate the mutant viruses in Table 1, we performed site-directed mutagenesis to introduce mutations into reverse genetics plasmids. The 2009 pandemic influenza virus isolate A/California/04/2009 (H1N1) was kindly provided by Yi Shi, Institute of Microbiology, Chinese Academy of Sciences, China. The “6 + 2” recombinant H5N1influenza virus containing the HA and NA genes derived from A/Vietnam/1194/2004 (H5N1) was constructed previously in our laboratory. The viruses were propagated in 10-day-old embryonated eggs. 293T and MDCK cells were maintained in Dulbecco modified Eagle medium (DMEM; Gibco) supplemented with 10% fetal bovine serum (Gibco) at 37°C under 5% CO₂.

Sera.

Mouse experiments were performed in accordance with the guidelines of The First Affiliated Hospital of Guangzhou Medical University Animal Care and Use Committee (permit number 2018-20). Rhesus macaque study protocols (permit number 2018058) were approved by the Institutional Animal Care and Use Committee of the Guangzhou Institutes of Biomedicine and Health (GIBH), Chinese Academy of Sciences. BALB/c mouse sera were collected by two intramuscular (i.m.) immunizations with inactivated purified H7/AH13 or H7/GD16 virus (based on HA content of 3 μg). Rhesus macaque sera were collected by two intramuscular immunizations with inactivated purified H7/AH13 or H7/GD16 virus (based on HA content of 15 μg). All sera were treated with receptor-destroying enzyme (RDE) at 37°C overnight.

HI assays.

The hemagglutination inhibition (HI) assay was performed as previously described (25). In brief, each serum sample was serially diluted and added to 4 hemagglutination units (HAU) of each recombinant virus. Then 1% chicken or other species RBC solution was added. Agglutination was read out after incubation for 30 min at room temperature. The HI titer was defined as the highest dilution of the serum able to inhibit hemagglutination. The minimum HI titer detected in this study was 20; thus, for statistical purposes, all samples from which the HI titer was not detected were given a numeric value of 10, which represents the undetectable level of HI titer.

MN assays.

In vitro microneutralization (MN) assays were performed as described previously (28). Sera were serially diluted and then added to 100 50% tissue culture infective dose (TCID₅₀) units of recombinant viruses, respectively, in serum-free medium containing tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin. After 1 h of incubation at 37°C, the virus-serum mixtures were then incubated with MDCK cells for 24 h at 37°C, with added 5% CO₂. Next, the plates were fixed with 4% paraformaldehyde for 10 min. The presence of viral protein was detected by ELISA with a monoclonal antibody (Abcam) to the influenza A virus nucleoprotein. The MN titer was defined as the reciprocal of the highest serum dilution that neutralized the virus in MDCK cell cultures. The minimum MN titer detected in this study was 20; thus, for statistical purposes, all samples from which the MN titer was not detected were given a numeric value of 10, which represents the undetectable level of MN titer.

Serum passive transfer experiment.

Animal experiments were performed in accordance with the guidelines of The First Affiliated Hospital of Guangzhou Medical University Animal Care and Use Committee (permit number 2018-20). Six- to eight-week-old female BALB/c mice were vaccinated i.m. with inactivated purified H7/GD16 or H7/AH13 virus (based on HA content of 3 μg) at week 0 and week 4. Sera were collected at week 8 and heated for 30 min at 56°C to inactivate complement. The treated sera were injected intraperitoneally (i.p.) into naive BALB/c mice. At 2 h posttransfer, mice receiving sera were challenged with 10× MLD₅₀ of recombinant H7N9/GD16. The mice were monitored daily for 14 days for survival and weight loss after the challenge.

Antibody binding assays.

The binding of viruses by antisera and MAbs was measured by ELISA. Briefly, sucrose gradient-purified viruses were used to coat 96-well flat-bottom microtiter plates (Greiner bio-one) at 16 HAU per well. Virus amount was normalized in ELISAs using the MEDI8852 MAb (13), which binds to the conserved region of the HA stalk. After overnight incubation at 4°C, wells were blocked with 1% bovine serum albumin (BSA) and then incubated with each immune serum and MAb (serial 2-fold dilutions). After washing, horseradish peroxidase (HRP)-conjugated species-specific IgG antibodies (EarthOX) were added and allowed to incubate for 1 h at 37°C. The signal was developed using tetramethylbenzidine (TMB) as the substrate. The reaction was stopped with 1 M H₂SO₄, and values for optical density at 450 nm (OD₄₅₀) were read.

Red blood cell binding assays.

As previously described (29), chicken red blood cells were pretreated with a series of amounts of RDE (Vibrio cholerae neuraminidase [Sigma]) for 2 h at 37°C. The red blood cells were washed with phosphate-buffered saline (PBS) and added (as 1% [vol/vol] solutions) to 2 HAU of each virus. After a 30-min incubation, agglutination was read out. Data are expressed as the maximal concentration of RDE that still allowed for full agglutination.

Real-time RT-PCR assays.

The real-time reverse transcription (RT)-PCR assays were performed as described previously (30). The 8 HAU viruses by hemagglutination assays with chicken RBCs were 10-fold serially diluted. RNA was extracted from viruses at each dilution by using the Hipure viral RNA kit (Magen, China). RNA was used as the template for cDNA synthesis by using the PrimeScript II first-strand cDNA synthesis kit (TaKaRa Bio, Japan). Real-time RT-PCRs were performed using the TB Green Premix Ex Taq kit (TaKaRa Bio) with the cDNA templates and NP-specific primers.

BLI.

Biolayer interferometry (BLI) analyses were performed using the Octet Red 96 system (ForeBio, Menlo Park, CA) in 96-well microplates, as described previously (4, 31). Two biotinylated receptor analogues, the α-2,6 glycan (6SLNLN: NeuAcα2-6Galβ1-4GlcNAcβ1-3Galβ1-4GlcNAcβ1-SpNH-LC-LC-biotin) and the α-2,3 glycan (3SLNLN: NeuAcα2-3Galβ1-4GlcNAcβ1-3Galβ1-4GlcNAcβ1-SpNH-LC-LC-biotin), were kindly provided by Yi Shi (Institute of Microbiology, Chinese Academy of Sciences, Beijing, China). The two glycans were precoupled to streptavidin-coated biosensors (Pall Fortebio LLC) in PBS and reacted with purified viruses (64 HAU) in PBS containing 0.02% (vol/vol) Tween 20 and 10 μM oseltamivir carboxylate. Association was measured for 3,000 s at 30°C. Data were analyzed by using the Octet data analysis software.

Receptor binding of viruses in MDCK cells.

Receptor binding of viruses in MDCK cells was performed as described previously (32), with slight modification. In brief, MDCK cells in 96-well plates were prechilled at 4°C for 1 h to prevent endocytosis of the viral particles. Then pretreated MDCK cells were infected with each virus at a hemagglutination titer of 8. After 1 h of adsorption at 4°C, the infected cells were either washed three times with PBS or directly overlaid with 200 μl of serum-free DMEM containing 0.3% BSA and 1 μg/ml of TPCK-trypsin. After incubation at 37°C for 6 h, the infected plates were fixed and permeabilized with 80% acetone for 20 min at room temperature, followed by ELISA using MEDI8852 MAb. The receptor binding of a virus to MDCK cells was calculated according to the formula (OD₄₅₀ value of washed wells/OD₄₅₀ value of unwashed wells) × 100.

Statistical analyses.

Statistical analyses were performed for antibody titers, body weight loss, RBC receptor-binding assays, and MDCK cell binding assays by using analysis of variance (ANOVA), followed by Tukey’s test in GraphPad Prism 7; P values of <0.05 were considered significant.