Animal immunization protocol and sample collection

Three outbred adult horses were randomly selected, and EIV seronegativity was confirmed in all three. The horses received intramuscular inoculation with two doses of the EIV inactivated vaccine (generated from A/equine/XinjiangFuyun/2007 (H3N8) (XJ07) strain) at 0 and 29 days. Serum samples were collected for the evaluation of the Hemagglutination inhibition (HI) titres on the indicated days (21,31,33,35,37,39,41,48,55,62 and 76 days). Peripheral blood mononuclear cells (PBMCs) were isolated from immune horses at the time point when the highest hemagglutination inhibition (HI) titre was observed using a Ficoll-Histopaque density gradient (density=1.077 g/ml) for sorting HA-specific B cells.

Sorting HA-specific B cells from EIV-inoculated equine PBMCs

Biotin–streptavidin conjugation was used to stain isolated equine PBMCs simultaneously with a bundle of monoclonal antibodies, including FITC anti-rat CD3 (Abcam, ab34722), PE-Cy7 anti-human CD21 (BD Biosciences, 561374), Aqua Dead Cell Stain Kit (Thermo Fisher, L-34965) and the EIV-HA-Y97F trimer labelled with BV421 (Biolegend, 405225) and APC (Thermo Fisher, 17-4317). A stepwise gating strategy was employed as follows: lymphocytes were gated based on FSC-A and SSC-A parameters; live cells were gated using aqua vital-AmCyan; B cells were gated with CD3^negCD21^pos, and HA-bound B cells were gated using HA-Y97F mutant-trimer-BV421^pos conjugated HA-Y97F mutant-trimer-APC^pos. The HA-Y97F mutant-trimer using a Foldon-guided self-assembly approach was engineered specifically to minimize nonspecific binding [17,18], as detailed in the supplementary methods. The flow-cytometric sorting was conducted in a BD FACS Aria III instrument. Single HA-baited B cells were directly sorted into 96-well PCR plate.

Two-round nested PCR

The equine immunoglobulin gene primer pool was utilized to amplify equine immunoglobulin genes, and separate cDNA pools for heavy and light chains were constructed. Using a mixed cDNA pool, a two-round nested PCR process was sequentially performed. This aimed to amplify equine Ig genes (Fab) consisting of two variable domains (VH and VL), along with two constant regions (CH and CL), from heavy and light chains independently. In Round-1 of the nested PCR, pooled Tag-labelled equine leader sequences were employed as forward primers to amplify variable domains in either the heavy chains (VH) or the light chains (VL) separately. Reverse primers targeting non-variable regions in the equine sequences were used to amplify constant regions in either the heavy chains (CH) or the light chains (CL). Primers for the equine variable region in the light chain were used only for the amplification of the lambda chain (VL: λ), as the κ to λ ratio is 1:20 in equines [19,20].

For Round-2 of the nested PCR, the Tag was targeted to design forward primers for the subsequent nested amplification. Reverse primers specific to abbreviated equine constant CH1/CL1 genes with tails corresponding to homo constant gene IgG1 (CH) or Ig λ (CL) were designed aiming at facilitating recombination in the following step. Subsequently, cDNA fragments encompassing variable domains and constant regions within equine heavy chains or light chains were individually amplified from single memory B cells. All PCR products were identified using sequencing. The PCR products of the equine heavy chain genes were subsequently annotated with Complementarity-Determining Regions (CDRs) using the V-quest function in the IMGT information system and using the equine germline database as a reference [21]. Detailed primer sequences are provided in Supplementary Material (Tables 1–7).

Screening recombinant Igs harbouring equine Ig (Fab)2 and homo Fc binding EIV-HA

We next assessed the ability of the recombinant immunoglobulins (mAbs) to bind to the HA protein [22]. We coated HA protein onto ELISA plates, blocked them, and then added purified mAbs from transfected HEK 293 T cell cultures. After incubation, we used a secondary antibody and measured binding using an ELISA reader. We determined the relative affinity of mAbs by calculating the effective concentration (EC₅₀) required for half-maximal binding using GraphPad Prism. Positive mAbs were further processed for expression abundance in 293F cells. Recombinant antibodies were isolated, cultured, and purified using size-exclusion chromatography using a Superdex 200 Increase 10/300 GL column (GE Healthcare) as previously described [20,23,24]. The binding of purified mAbs to EIV-HA protein was confirmed using SPR assays. For further details, refer to the Supplementary Material.

Screening of recombinant antibodies for neutralization

To identify the most effective recombinant antibodies capable of neutralizing various strains of equine influenza virus (EIV), virus neutralization assays were conducted. Influenza strains used in this assessment included a main circulating EIV strain of A/equine/XinjiangFuyun/2007 (H3N8) (XJ07), an avian origin equine IAV A/equine/Jilin/1989 (H3N8) (JL89), as well as two recently isolated strains (A/equine/GanSu/2022 (H3N8) (GS22) and A/equine/hulunbeier/1/2023 (H3N8) (HLBE23)) in our lab. The neutralization assay followed the guidelines outlined in the OIE Terrestrial Manual. Viral stocks were prepared in eggs, and the titre of the stock was determined by TCID₅₀ assays using MDCK cells. Subsequently, 50 μl of serially diluted monoclonal antibodies (mAbs) or serum was mixed with an equal volume of culture medium containing 200 TCID₅₀ virus and incubated for 1 h at 37 °C. The mixture was then transferred to MDCK cells in a 96-well plate and incubated for about 24 h at 37 °C. After removing the supernatant post-infection, each well received 100 μl of paraformaldehyde reagent for 10 min, followed by washing with PBS. Direct ELISA was then performed using anti-influenza A NP mouse monoclonal antibody. The negative control (VC) contained 200 TCID₅₀ of EIV but without the antibodies, while the cell control (CC) contained only DMEM (with 1 μg/mL TPCK-trypsin). The neutralization efficiency of mAbs was calculated as previously described [25].

In vivo validation of the screened candidate recombinant antibody

We tested equine recombinant mAbs as potential cross-reactive neutralizing antibodies. Mice were challenged with EIV to assess the efficacy of the antibodies in preventing and treating the infection. The EIV strain (A/equine/XinjiangFuyun/3/2007 (H3N8) (XJ07)) was adapted for mice, establishing a lethal challenge model with a dose of 6×10⁷ EID₅₀. Forty BALB/c mice (aged 6–7 weeks) were randomly divided into groups (n=10 per group). Two groups of challenged mice received a single intravenous dose of 20 mg/kg for either pre-exposure prophylaxis (PrEP) or therapeutic treatment (H81). One challenged group remained untreated, while another group was mock-challenged with PBS as a control. The mice were monitored daily, and body weight loss and deaths were recorded. On day 5, each from the challenged and medication groups (death), and on day 14, each from the PrEP, medication and sham groups, were euthanized. Lung samples were collected during autopsy for virus titration and histopathological analysis.

Assessment of lung histopathological changes and viral titres

Autopsy was performed in randomly selected mice that had either died (in the challenged group and medication group) or that had undergone euthanasia (in other groups). Lung tissues were fixed in 4% paraformaldehyde, embedded in paraffin, and then sectioned and stained by the routine hematoxylin and eosin method (H&E) for histopathological examination. To determine the viral titre in lung tissue, the supernatant collected after tissue grinding was serially diluted by a factor of 10 and then inoculated into MDCK cells. Following virus absorption for 1 h, the virus was removed and MDCK cells were cultured in Opti-MEM supplemented with tosylsulfonyl phenylalanyl chloromethyl ketone (TPCK)-treated trypsin (1μg/mL). Cell supernatants were harvested at 24 h post-infection, and viral titres were determined by HA-specific assay.

Paratope and epitope prediction of the screened recombinant equine Ig (Fab)2 and HA

To investigate how the screened recombinant equine Ig (Fab)2 binds to the HA antigenic epitopes, we used SWISS-MODEL homology modelling. Each Fab sequence was entered individually, along with its corresponding immunoglobulin variable domains in heavy and light chains, to predict the quaternary structures of both the antibody and the HA antigen. The prefusion state of HA (PDB ID: 7K37) served as the template for the modelling process. Two binding models were generated, one for trimeric HA and another for monomeric HA. The PyMOL software was utilized for structural analysis and figure preparation.

Prediction of functional binding regions between the HA monomer and the potential equine Ig (Fab)2

To predict the paratope of potential HA neutralizing antibodies derived from recombinant equine Ig (Fab)2 and their corresponding antigenic epitopes, we employed homology modelling to predict the 3D structures of both the antibodies and HA. SWISS-MODEL was utilized with the prefusion state of HA (PDB ID: 7K37) as the template. The input included the sequences of the screened equine Ig (Fab)2 along with the corresponding immunoglobulin variable domains or constant regions. Molecular docking in Discovery Studio Visualizer (V4.5) was used to predict the most likely binding mode. Twenty theoretical models were constructed for statistical optimization based on minimum Probability Density Function (PDF) for Total Energy and minimum DOPE Score. The optimal complex model was used to predict the functional binding region between equine Ig (Fab)2 paratopes and EIV-HA antigenic epitopes. PyMOL was employed for structure analysis and figure preparation.

Validation of each predicted epitope in EIV-HA ex vivo

Based on the predictive analysis conducted in PyMOL, 27 amino acid residues within the HA protein, which were identified as potentially influencing antibody binding, were selected for mutagenesis using a mutagenesis kit (New England Biolabs). Following this, expression plasmids containing HA mutants were constructed using the pCDNA3.3 vector. To evaluate the binding affinity of H81 to these mutants, both direct ELISA and SPR assays were conducted. By comparing the affinity kinetics before and after mutation, we identified the sites critical for antibody binding to HA. In the direct ELISA assay, microtiter plates were separately coated with HA mutant expression proteins, followed by the addition of H81 antibody, to assess their binding interaction. In the SPR assays, anti-EIV-HA antibodies were immobilized on the CM5 chip in HBS-EP buffer (10 mM HEPES pH 7.4, 150 mM NaCl, 3 mM EDTA, 0.005% surfactant P20). Binding analyses were conducted by injecting HA mutant protein solutions at various concentrations. Further details regarding the ELISA assay and SPR method can be found in the Supplementary Material.

Stepwise amplification of equine F(ab) genes in single B cells

In the two-round nested PCRs, 145 wells out of 370 harbouring single equine B cells tested positive for equine heavy chains (145/370, 39%), while 149 wells tested positive for the equine light (λ) chain (149/370, 40%). Of these, a total of 137 single equine B cells demonstrated dual positivity (137/370, 37%), expressing equine Ig genes (Fab with a short Fc-CH1 tail) from both heavy and light chains. These double positive B cells were selected for subsequent recombination (Figure 2(B)). Through overlapping PCR, equine Ig (Fab) genes from the heavy and light chains originating from the same single B cell were presumed to self-assemble into linear cassettes for expression and recombination with the homo constant region IgG1. These processes resulted in 137 recombinant monoclonal Igs containing equine Ig (Fab)2 and homo Ig (Fc) components derived from distinct single B cells (Figure 2(C)). The positivity rate observed in this screening platform is comparable to that seen in mouse and human single-B cell screening platforms, indicating that it is sufficient for subsequent screening aimed at identifying neutralizing antibodies.

Expression and binding characteristics of recombinant antibodies

A total of 137 recombinant Igs, including equine Ig (Fab)2 and homologous Ig (Fc), were individually expressed in co-transfected HEK 293 T cells, resulting in the generation of 23 recombinant equine antibodies through random pairing. To verify binding specificity, we initially conducted ELISA assays on these antibodies. As anticipated, all 23 recombinant equine Ig (Fab)2 and homo Fc antibodies showed binding to EIV HA with EC₅₀ values<100 ng/mL. The 10 antibodies with the highest binding affinities (EC₅₀<20 ng/mL) were further characterized using SPR assays (Figure 3(A)). Consistent with the ELISA results, all 10 antibodies demonstrated robust binding signals to EIV-HA. SPR assays further revealed dissociation constants (KD) ranging from 10⁻⁸ to 10⁻¹⁰ M for these antibodies, as depicted in Figure 3(B). Interestingly, a consistent pattern in the affinity levels of antibodies was observed across both methods. For example, the H100 antibody consistently exhibited the lowest affinity. Taken together, our findings suggest that in our screened equine Ig (Fab)2 library, ten recombinant Igs specifically bind to EIV-HA with high binding affinities.

To characterize the function of the isolated recombinant mAbs, we further analyzed their neutralization capacity in MDCK cells using a panel of EIV strains. The micro-neutralization assay revealed that ten recombinant mAbs efficiently neutralized the original parent strain (XJ07) at concentrations ranging from 0.05 μg/mL to 5.0 μg/mL. Specifically, H81 exhibited neutralizing activity against the virulent EIV strain (JL89) with an IC₅₀ value of 7.5 μg/mL, indicating its high virus-neutralizing potency (Figure 3(C)). This observation was corroborated across additional EIV strains, where H81 effectively protected against infection by other A/H3N8 EIV viruses (GS22 and HLBE23) with IC₅₀ values (0.94 and 0.94 μg/mL). Consequently, H81, identified as cross-reactive, was selected as the candidate therapeutic mAb for subsequent in silico prediction and in vivo validation.

Validation of the recombinant antibodies in EIV-challenged mice

Four groups of mice were monitored and compared (the timeline is shown in Figure 4(A)). Compared with the lethally challenged control group, both the therapeutic group and the prophylactic group (PrEP) showed milder clinical symptoms, as indicated by insignificant weight loss (Figure 4(B)). Additionally, deaths occurred in both the medication group and the challenged group: specifically, five deaths were observed in the one-dose H81 medication group. In contrast, the lethally challenged control group experienced 100% mortality by 8 days post-infection (dpi) (Figure 4(C)). In contrast, mice that underwent PrEP showed 100% protection against EIV challenge, with only mild nonspecific changes such as slight edema and congestion observed via histopathological microscopy (Figure 4(D)). Importantly, viral titres in autopsied mice consistently showed undetectable levels of virus in lung tissues from the PrEP group, in stark contrast to markedly high viral titres observed in the lethal challenge group (TCID₅₀/0.1 mL:10 (4.25)−10 (5.75)) (Figure 4(E)). Despite the therapeutic group showing a 50% protection rate, the virus titres in the mice were significantly lower compared to those in the challenged control group (10 (3.52) vs 10 (4.85)). These data indicate that the H81 recombinant antibody has potential for protecting and treating EIV infections in vivo.

We would like to thank Associate Professor Yuhong Wang for helpful discussions and Yingying Guo for prediction of functional binding region between HA and antibodies. We thank the Core Facility of the Harbin Veterinary Research Institute, the Chinese Academy of Agricultural Sciences for providing the technology platform. This study was supported by grants from the National Key Research and Development Program of China (2023YFD1802500) and the National Natural Science Foundation of China (32330103 and 32372985).