The SPF chicken embryos were purchased from Nanjing Biology Medical Factory Qian Yuan-hao Biological Co., Ltd. All animal experiments adhered to the Guide for the Care and Use for Laboratory Animals stipulated by the Ministry of Science and Technology, and followed the animal protection and management regulations of the Jiangsu Agri-Animal Husbandry Vocational College (Approval No: jsahvc-2024-60).

The sequences of the H9N2 vaccine strain A/chicken/Jiangsu/WJ57/2012 (WJ57), published in the GenBank database under accession numbers KP893703 to KP893708 and KJ000709 to KJ000710, were synthesized by SYNBIO Technologies (Suzhou, China). The recombinant WJ57 strain (rWJ57) was generated using reverse genetics and stored at the Jiangsu Key Laboratory for High-Tech Research and Development of Veterinary Biopharmaceuticals. COS-1 cells were purchased from ATCC (Manassas, VA, United States), cultured in Dulbecco’s Modified Eagle’s Medium (DMEM) (Gibco, BRL, Grand Island, United States) supplemented with 10% fetal calf serum (Gemini, Woodland, CA, United States), and incubated at 37°C with 5% CO₂.

The frequency and prevalence of residues at amino acid position 198 in HA were analyzed by comparing 4,093 full-length H9N2 HA sequences from Chinese isolates deposited in GenBank from 2013 to 2021 (up to January 1, 2023).

Primers for site-directed mutagenesis and homologous recombination were designed based on the HA sequence of the WJ57 virus using Primer 5.0 software (Primer-E Ltd., Plymouth, United Kingdom). Overlap PCR was used to amplify the mutant HA genes, which were then inserted into the transcriptional/expression vector pHW2000 using the ClonExpress II One Step Cloning Kit (Vazyme Biotech, Nanjing, China). After transformation into DH5α cells, 10 colonies were randomly selected, cultured, and subjected to plasmid DNA extraction (TianGen, Beijing, China). The constructs were then sequenced (GENEWIZ, Suzhou, China) to confirm their accuracy. The primers used in the amplification process are listed in Table 1.

As described by Zhu et al. (14), the H9N2 recombinant viruses, rWJ/HA_198A and rWJ/HA_198T, were identified and rescued using the rWJ57 strain backbone. Briefly, a total of 2.4 μg of eight plasmids, mixed in equal amounts, was added to 100 μL of Opti-MEM medium (Gibco, BRL, Grand Island, United States). Then, 5 μL of X-treme GENE HP DNA Transfection Reagent (Roche, Basel, Switzerland) was introduced to the mixture, which was incubated at room temperature for 15 min. The mixture was then added to COS-1 cells cultured in 12-well plates at 70–80% confluency. The cells were incubated at 37°C with 5% CO₂ for 24 h, after which 2 μg/mL of TPCK-trypsin (Sigma, St. Louis, MO, United States) was added to each well. Following a 30-h transfection period, the supernatants were collected and used to propagate the virus in 10-day-old specific pathogen-free (SPF) embryonated chicken eggs. Finally, the rescued virus was subjected to an HA assay, and the HA genes were sequenced at GENEWIZ (Suzhou, China) to confirm the accuracy of the intended mutations.

Antisera for the rWJ57 strain were prepared as previously described (11, 14). Briefly, 3-week-old SPF chickens were immunized twice by subcutaneous injection of 0.3 mL of an oil-emulsified, inactivated whole-virus vaccine of the rWJ57 strain, which was inactivated by adding 0.2% formalin (v/v) for 24 h at 37°C. Antisera were collected from the chickens 2 weeks after the booster vaccination. The negative sera used were collected from non-immunized SPF chickens as a pre-immunized control.

Antisera for the rWJ57 strain were treated with cholera filtrate (Sigma-Aldrich, St. Louis, MO, United States) to remove nonspecific hemagglutination inhibitors before conducting the HI assay on three recombinant viruses: rWJ57, rWJ/HA_198A and rWJ/HA_198T. The assay was performed using 4 hemagglutination units (HAU) of H9N2 and 1% (v/v) chicken erythrocytes (14, 24).

The ELISA assay was conducted according to previously described methods (14, 24). Sucrose gradient-purified viruses—rWJ57, rWJ/HA_198A and rWJ/HA_198T—were diluted in PBS and seeded into Nunc-Immuno MaxiSorp 96-well plates (Corning, NY, United States) at a concentration of 16 HAU per well. Following overnight incubation at 4°C, the wells were blocked with PBS-nonfat dry milk (Beyotime, Shanghai, China). Serial two-fold dilutions of antisera against the rWJ57 virus in chickens were prepared using PBS containing 0.05% Tween 20 and then added to the wells, followed by a 2-h incubation at 37°C. After washing, a goat anti-chicken horseradish peroxidase antibody (Bioss, Beijing, China) was added, and the plates were incubated for 1.5 h at 37°C. Finally, TMB (3,3′,5,5′-tetramethylbenzidine) substrate (Beyotime, Shanghai, China) was added, and the reaction was stopped using TMB stop substrate (Beyotime, Shanghai, China).

Absorbance was measured at 450 nm using an automated ELISA plate reader (model EL311SX; Biotek, Winooski, VT, United States). The area under the curve (AUC) for each virus was calculated using GraphPad Prism 9.5 software (San Diego, CA, USA), based on virus-antibody binding above the corresponding negative control.

The RDE assay was performed as previously described (14, 25). Chicken erythrocytes were treated with α2-3,6,8 neuraminidase (New England Biolabs, Beverly, MA, United States) through 11 two-fold serial dilutions starting from a maximum concentration of 1,000 U/mL for 1 h at 37°C. After washing with PBS, 1% (v/v) solutions of the treated erythrocytes were added to 4 HAU of each virus (determined using untreated chicken erythrocytes). Agglutination was measured after 1 h of incubation.

To further evaluate the receptor-binding properties of the H9N2 viruses, we performed solid-phase binding assays using the synthetic sialyl glycopolymers Neu5Ac 2–3 Galβ1-4 GlcNAc (3′SLN)-PAA-biotin and Neu5Ac 2–6 Galβ1-4 GlcNAc (6′SLN)-PAA-biotin (GlycoTech), based on the method described by Zhang et al. (26, 27). The glycan analogs were serially diluted in PBS and added to streptavidin-coated 96-well plates (Pierce). Plates were blocked with PBS containing 2% skim milk, and 64 HA units of virus were applied per well, followed by chicken antiserum as the primary antibody. Detection was done by adding HRP-conjugated rabbit anti-chicken IgG and tetramethylbenzidine substrate, with the reaction stopped by 1 M H₂SO₄. Absorbance at 450 nm was measured, and all samples were tested in triplicate.

The structures of HA were predicted using the Swiss-Model and validated by Ramachandran plot analysis, with 98.2% of residues in favored regions. The protonation state of all compounds was set to pH 7.4, and 3D structures were generated using Open Babel (28). AutoDock Tools (ADT3) were used to prepare and parameterize the receptor protein and ligands. Docking grid files were constructed with AutoGrid from the sitemap, and docking simulations were performed using AutoDock Vina (1.2.0) (29, 30). The optimal pose was selected for protein-ligand interaction analysis. Finally, protein-ligand interaction figures were generated using PyMOL. The ligands used in this study were α2,6 sialic acid (SA) (6′SLN, CAS no. 501427–93-6) and α2,3 SA (3′SLN, CAS no. 501427–92-5).

MD simulations were conducted using the non-commercial version 2022.1 of Desmond/Maestro software.¹ TIP3P water molecules were added to the systems, which were neutralized with a 0.15 M NaCl solution. After system minimization and relaxation, a 100-ns production simulation was performed using an isothermal-isobaric ensemble at 300 K and 1 bar. Trajectory coordinates were recorded every 100 ps throughout the simulation. Molecular dynamics analysis was performed using the Simulation Interaction Diagram feature of Desmond software.

Data normality was verified using Shapiro–Wilk test. For experiments comparing both virus types and concentrations (receptor binding assays), statistical analysis was performed using two-way ANOVA with Holm-Bonferroni correction for multiple comparisons to control family-wise error rate. For single variable comparisons (HI titers and antibody binding assays), pairwise comparisons were made using Student’s t-test. All statistical analyses were conducted in GraphPad Prism 9.5, with adjusted p-values less than 0.05 considered statistically significant. All experiments were performed with three independent biological replicates (n = 3), with technical duplicates for each biological replicate to ensure measurement reliability. The mean values from technical duplicates were used for statistical analysis. For molecular dynamics simulations, each system was simulated for 100 ns and repeated three times with different initial velocities to ensure conformational sampling adequacy.

An analysis of 4,093 avian H9N2 AIV HA protein sequences at position 198, published in GenBank from 2013 to 2021, revealed three prevalent residues: 198 T, 198A, and 198 V. The most frequent was 198 T, representing 72.4% of the sequences, followed by 198A (20.7%), while 198 V was the least common (6.9%) (Figure 1A). Between 2013 and 2015, the frequency of the 198 T mutation steadily increased, reaching over 80% from 2015 to 2017. However, from 2017 to 2020, the prevalence of 198 T steadily declined, dropping to 70% in 2021, while the 198 V showed an inverse trend, rising in frequency and becoming dominant by 2021 (Figure 1B). The 198A frequency remained stable (10–15%) throughout the study period, with minor fluctuations.

Following transfection of COS-1 cells with eight plasmids encoding the full H9N2 genome, the supernatants were inoculated into SPF chicken embryos. After 90 h, HA titers were measured in the allantoic fluid, and recombinant viruses with HA titers ≥2² were considered positive. The HA titers for rWJ/HA_198A and rWJ/HA_198T viruses were 2⁶ and 2⁸, respectively. Sequencing confirmed that the HA genes contained the desired mutations at positions 591–594, while all other gene segments (PB2, PB1, PA, NP, NA, M, and NS) were identical to the parental rWJ57 strain.

To investigate how mutations at residue 198 influence the antigenic properties of the virus, HI assays were performed using sera from chickens infected with rWJ57. The rWJ57 virus carrying the 198 V mutation showed a three-fold lower HI titer compared to rWJ/HA_198A and rWJ/HA_198T (Figure 1C), suggesting that residue 198 plays a critical role in promoting viral escape from a polyclonal antibody binding.

To explore the underlying mechanisms, ELISA assays were conducted to measure the binding affinities of specific and non-specific antibodies. The results demonstrated that the rWJ57 virus had significantly lower affinity for non-specific antibodies compared to rWJ/HA_198A (p < 0.001), while rWJ/HA_198T displayed significantly higher affinity than both rWJ57 and rWJ/HA_198A (p < 0.001) (Figures 2A,B). However, no significant differences were observed in the binding affinities of the three viruses to specific antibodies (Figures 2A,C).

To assess how receptor binding activity affects HI titers, chicken erythrocytes were treated with α2-3,6,8 neuraminidase to modify sialic acid receptors. Stronger receptor binding was reflected by the virus’s ability to agglutinate erythrocytes treated with higher concentrations of neuraminidase. Results showed that rWJ57 exhibited receptor binding activity four-fold higher than rWJ/HA_198A and 16-fold higher than rWJ/HA_198T (Figure 2D). These findings suggest that the mutation at position 198 alters receptor binding, which primarily drives HI titer fluctuations in the H9N2 virus, rather than specific antibodies.

While the RDE assay demonstrates the virus’s binding activity for sialic acids (both α2-3 SA and α2-6 SA), it cannot differentiate the virus’s affinity for specific sialic acid types. To further explore the effect of mutations at HA position 198 on the affinity for specific sialic acids, we conducted a solid-phase binding assay. The results revealed distinct receptor preferences among H9N2 virus variants with mutations at HA position 198. For 6’SLN binding, rWJ57 demonstrated the highest binding affinity across all tested concentrations (0.313–10 μg/mL), followed by rWJ/HA_198T, while rWJ/HA_198A showed the lowest affinity. The differences in binding affinity among the three viruses were statistically significant at each tested concentration (p < 0.001) (Figure 2E). Conversely, for 3’SLN binding, rWJ/HA_198A exhibited the strongest binding affinity, followed by rWJ57, while rWJ/HA_198T showed the lowest affinity. Similarly, significant differences in binding affinity were observed among all three viruses at each tested concentration (p < 0.001) (Figure 2F). These findings indicate that the amino acid at position 198 of the HA protein plays a crucial role in determining receptor binding specificity.

The α2,6 SA formed five hydrogen bond interactions with the HA protein of the rWJ57 strain (HA_198V protein) at amino acid positions 109, 147, 201, 235, and 236, with a binding energy of −6.1 kcal/mol (Figure 3A). It also formed five hydrogen bond interactions with the rWJ/HA_198T HA protein (HA_198T protein) at positions 109, 147, 148, 233, and 236, with a binding energy of −5.9 kcal/mol (Figure 3B), and four interactions with the rWJ/HA_198A HA protein (HA_198A protein) at positions 145, 147, 235, and 236, with a binding energy of −6.0 kcal/mol (Figure 3C).

The α2,3 SA formed six hydrogen bond interactions with the HA_198V protein at amino acid positions 109, 147, 148, 233, 235, and 236, with a binding energy of −5.6 kcal/mol (Figure 3D). It also formed six interactions with the HA_198T protein at positions 109, 147, 198, 233, 235, and 236, with a binding energy of −5.5 kcal/mol (Figure 3E), and three interactions with the HA_198A protein at positions 109, 145, and 147, with a binding energy of −5.5 kcal/mol (Figure 3F).

These results indicate that alterations at amino acid position 198 of the HA protein directly modify the microstructure of the receptor-binding pocket, leading to changes in the binding sites, interaction forces, torque, and binding energy with sialic acid receptors, ultimately influencing the interactions between the HA protein and the receptors.

To explore the molecular mechanisms by which mutations at residue 198 influence receptor binding affinity and viral tropism, molecular dynamics (MD) simulations were employed to evaluate the structural and dynamic behavior of the HA-sialic acid receptor complexes. The simulation focused on how the HA_198V, HA_198T, and HA_198A mutations alter the interaction between the HA protein and both α2,6 SA and α2,3 SA. The key metrics analyzed were root mean square deviation (RMSD), root mean square fluctuation (RMSF), hydrogen bonding patterns, and binding stability, which provided insights into conformational changes and receptor affinity.

During the simulation, the HA protein maintained a stable interaction with the α2,6 SA. The RMSD values for the HA_198V protein and the α2,6 SA stabilized around 7.0 Å and 9.0 Å, respectively, after initial fluctuations (Figure 4A), suggesting that the complex underwent minor adjustments before reaching a stable conformation. RMSF analysis (Figures 4B,C) revealed minimal fluctuations in key binding residues, particularly Tyr109, Thr147, Met235, and Gly236, highlighting their crucial role in maintaining the interaction’s stability.

Thr147 played a pivotal role in the binding process, forming hydrogen bonds with a high formation frequency of 118%, supported by secondary water bridge interactions with a frequency of 30% (Figures 4D,E). Other residues, including Tyr109, Asn145, Trp161, and Gly236, also contributed through hydrophobic and ionic interactions, further stabilizing the complex. The α2,6 SA itself formed several intramolecular hydrogen bonds, which contributed to its conformational stability (Figure 4E).

These findings suggest that the HA_198V mutation significantly enhances the binding affinity between the HA protein and α2,6 SA. The combination of hydrogen bonds, hydrophobic interactions, and water bridges forms a robust network that stabilizes the HA -α2,6 SA complex.

In the MD simulation of the HA_198T protein complexed with the α2,6 SA, the RMSD of the HA_198T protein stabilized at approximately 9.0 Å, while the α2,6 SA’s RMSD stabilized at around 16 Å after initial fluctuations (Figure 5A). These RMSD values indicate that both the protein and α2,6 SA underwent structural adjustments before reaching a stable conformation.

Further analysis using RMSF revealed that the HA_198T protein exhibited relatively low fluctuations, indicating minimal conformational changes during the binding process (Figure 5B). In contrast, the α2,6 SA displayed higher RMSF values, highlighting its dynamic and flexible nature (Figure 5C). This flexibility likely facilitated binding interactions with the protein, contributing to the formation of a more stable complex.

A total of 30 interactions were identified between the HA_198T protein and the α2,6 SA, with Thr147 playing a critical role, showing a hydrogen bond formation frequency of 92% (Figures 5D,E). Other residues, such as Asn145 and Gly233, also contributed significantly to the stability of the complex, reinforcing the hydrogen bonding network. Additionally, the α2,6 SA formed several intramolecular hydrogen bonds, further stabilizing its conformation and strengthening the overall α2,6 SA-protein complex (Figure 5E).

Overall, the HA_198T mutation enhances the binding affinity between the HA protein and the α2,6 SA. The hydrogen bond network, particularly involving Thr147, is crucial in stabilizing the complex. The flexibility of the α2,6 SA, along with its intramolecular interactions, plays a significant role in the stability and dynamic behavior of the α2,6 SA-protein interaction. These findings provide valuable insights into the molecular mechanisms underlying the increased receptor binding affinity due to the HA_198T mutation, with potential implications for future vaccine or antiviral drug development.

During the molecular dynamics simulation, the HA_198A protein exhibited significant fluctuations. The RMSD of HA_198A eventually stabilized at approximately 4.0 Å, whereas the RMSD of the α2,6 SA fluctuated substantially, finally stabilizing near 90 Å (Figure 6A).

While RMSF of the HA_198A protein remained relatively small, the RMSF of the α2,6 SA was notably larger (Figures 6B,C), indicating greater movement of the α2,6 SA during the simulation. Despite the formation of 29 interactions, the α2,6 SA eventually moved away from the active binding pocket (Figures 6D,E).

These results suggest that the initial binding conformation between the α2,6 SA and the HA_198A protein was unstable, characterized by weak binding affinity. During the simulation, the α2,6 SA ultimately dissociated from the binding pocket, failing to maintain a stable interaction with the HA protein.

During the MD simulation, the RMSD of the HA_198V protein stabilized at approximately 8.0 Å, while the RMSD of the α2,3 SA, after some fluctuations, settled around 10.0 Å (Figure 7A). The RMSF values for the amino acid residues interacting with the α2,3 SA, highlighted by green lines in Figure 7B, were relatively low, indicating minimal conformational changes in these regions during the binding process. In contrast, the RMSF values for the atoms in the α2,3 SA were relatively high, reflecting substantial fluctuations in the α2,3 SA ‘s structure (Figure 7C). These findings suggest that the initial binding conformation was unstable, though the system eventually achieved a more stable conformation after significant fluctuations.

The HA_198V protein and the α2,3 SA formed 41 interactions, with Phe408 playing a pivotal role by forming hydrogen bonds with a 31% frequency, highlighting its importance in the binding process. Other interaction sites contributed via hydrogen bonds and water bridges, although these interactions were comparatively weak (Figures 7D,E). Additionally, the α2,3 SA formed several intramolecular hydrogen bonds that helped stabilize its conformation during binding (Figure 7E).

Overall, these results indicate that the HA_198V protein and α2,3 SA optimized their binding conformation through the simulation, achieving a more stable structure with moderate binding affinity.

During the MD simulation, the HA_198T protein exhibited minimal fluctuations, with its RMSD stabilizing at approximately 6 Å. In contrast, the RMSD of the α2,3 SA displayed significant variability before eventually stabilizing at around 30 Å (Figure 8A). The RMSF values for the HA_198T protein were relatively low, indicating limited conformational changes, while the α2,3 SA exhibited much higher RMSF values, suggesting considerable fluctuations in the α2,3 SA’s structure throughout the simulation (Figures 8B,C). This pattern suggests that the initial conformation of the HA_198T protein and α2,3 SA complex was unstable, with the α2,3 SA gradually moving away from the binding pocket during the course of the simulation.

Although a total of 91 interactions were formed between the protein and α2,3 SA, none of these interactions had frequencies exceeding 30% (Figure 8D), and only those above this threshold are shown in Figures 8D,E. The absence of strong, consistent interactions indicates a weak binding affinity between the HA_198T protein and the α2,3 SA. Consequently, the α2,3 SA was unable to maintain stable binding within the protein’s receptor pocket, which is consistent with its detachment from the binding site.

During the MD simulation, both the HA_198A protein and the α2,3 SA demonstrated high stability. The RMSD of the HA_198A protein stabilized at approximately 6.0 Å, and after initial fluctuations, the RMSD of the α2,3 SA also stabilized at around 6.0 Å (Figure 9A). The RMSF values for the interacting amino acid residues of HA_198A (indicated in green in Figure 9B) and the α2,3 SA atoms showed minimal fluctuations overall (Figures 9B,C), indicating that both the protein and the α2,3 SA had relatively stable initial conformations. After some dynamic movements, a more stable binding conformation was formed.

The HA_198A protein and the α2,3 SA formed a total of 33 interactions. Notably, the hydrogen bond and water bridge interaction frequencies of Glu407 reached 159 and 99%, respectively, underscoring the critical role of Glu407 in the binding process (Figures 9D,E). Other important residues, including Glu99, Arg275, Ile276, Lys278, and Asp405, also contributed significantly to maintaining the complex’s stability (Figures 9D,E). Additionally, several intramolecular hydrogen bonds were formed within the α2,3 SA, further stabilizing its binding conformation (Figure 9E).

These results suggest that the α2,3 SA and HA_198A protein optimized their initial binding conformations during molecular dynamics simulations, resulting in a more stable complex with high binding affinity.

In summary, MD simulations provided insights into the molecular interactions and conformational changes governing the binding affinity between the HA protein and sialic acids, which are critical for viral receptor specificity. The binding affinity between the 198th residue of the HA protein and the α2,6 SA decreased in the order of HA_198V, HA_198T, and HA_198A. Conversely, for the α2,3 SA, the binding affinity increased in the order of HA_198T, HA_198V, and HA_198A. Furthermore, when interacting with the α2,6 SA, the 198th residue primarily engaged through the formation of water bridges. In contrast, the 198th residue did not directly participate in interactions with the α2,3 SA, suggesting a differential role of this residue in modulating receptor specificity based on SA type.