The genes encoding the P domains of GII.23 Quininde1906 (GenBank: KR232647), GII.24 Loreto6424 (MG495081), and GII.25 Dhaka1928 (MG495083) were synthesized by Genewiz (Suzhou, China). A cysteine-containing short peptide (CDCRGDCFC) was fused to the C-terminus of the P domain to stabilize P-protein formation, while the wild-type P domain sequence was used to generate P dimers (Holroyd et al., 2025). The P domain segments were cloned into the pGEX-6P-1 vector as previously described (Tan and Jiang, 2011). For protein expression, plasmids were transformed into Escherichia coli BL21(DE3) competent cells. Protein expression was induced with 0.4 mM isopropyl-β-D-thiogalactoside (Roche, 11411446001) at 22°C for 16–20 h. Cells were harvested and resuspended in phosphate-buffered saline (PBS, 140 mM NaCl, 2.7 mM KCl, 10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, pH 7.4). Clarified lysates were loaded onto Glutathione Sepharose 4B GST-tagged protein purification resin (Cytiva, 17075601), and P proteins were eluted by PreScission protease digestion (Beyotime, P2302). P dimers were further purified using a Superdex 200 Increase ^10/300GL gel-filtration column (Cytiva, 28990944). Mutations were constructed using the QuikChange site-directed mutagenesis kit (Beyotime, D0206S).

An enzyme-linked immunosorbent assay (ELISA) was performed to assess binding between saliva samples and the P proteins of norovirus as previously described (Ao et al., 2018). The 96-well microplates were coated overnight at 4°C with heat-inactivated saliva samples (diluted 1:10,000) from both ABO secretor and non-secretor individuals. After five washes with PBS-T (PBS supplemented with 0.05% Tween 20), plates were blocked with 5% skim milk in PBS for 2 h at 37°C. Subsequently, 0.5 μg/well of P proteins were added and incubated overnight at 4°C. Following five washes with PBS-T, rabbit anti-P domain polyclonal antibody (1:6,000) that was generated and validated as previously described (Cong et al., 2019) was incubated at 37°C for 2 h. After washing with PBS-T, HRP-conjugated goat anti-rabbit IgG (1:1,000, Beyotime, A0208) was incubated at 37°C for 1 h. Following extensive washing, 3,3′3llow-tetramethylbenzidine (TMB) substrate reagent (BD Biosciences, 555,214) was added to each well in the dark at room temperature. The reaction was terminated with 1 M H₃PO₄, and absorbance was measured at 450 nm. The cutoff for positive signal of Optical Density at 450 nm (OD₄₅₀) was defined as the mean OD₄₅₀ value of negative control plus three times the standard deviation (Xie et al., 2020). For data normalization, the raw OD₄₅₀ value of sample well was subtracted by the average OD₄₅₀ value of blank control well to account for background absorbance.

A glycan binding assay was conducted as described previously (Ao et al., 2018). Briefly, 20 μg/well of purified P proteins was coated onto 96-well microplates overnight at 4°C. Following blocking with 5% non-fat milk in PBS for 2 h at room temperature, wells were incubated with 0.2 μg/well of the following biotinylated glycans overnight at 4°C: A trisaccharide (A), B trisaccharide (B), H disaccharide (H), H type 1 (H1), H type 2 (H2), H type 3 (H3), Lewis a (Le^a), Lewis x (Le^x), Lewis b (Le^b), Lewis y (Le^y), type I precursor (Lec), and type II precursor (LacNAc) (GlycoTech). After five washes with PBS-T, HRP-conjugated streptavidin (1:1,500, Proteintech, SA00001-0) was added and incubated for 1 h at 37°C. After an additional wash, TMB substrate reagent was added for color development. The reaction was terminated with 1 M H₃PO₄ and absorbance was measured at 450 nm. The cutoff for the positive signal of Optical Density at 450 nm (OD₄₅₀) was defined as the mean OD₄₅₀ value of the negative control plus three times the standard deviation (Xie et al., 2020). For data normalization, the raw OD₄₅₀ value of sample well was subtracted by the average OD₄₅₀ value of blank control well to account for background absorbance.

The purified P dimers of norovirus strains GII.23 Quininde1906, GII.24 Loreto6424, and GII.25 Dhaka1928 were concentrated to 5–10 mg/mL for crystallization. Initial screening utilized commercial kits at 18°C using the sitting-drop vapor diffusion method with a drop ratio of 1 μL protein solution to 1 μL reservoir solution. Crystallization conditions were as follows: GII.23 in 10% (v/v) PEG 200, 0.1 M BIS-TRIS propane (pH 9.0), and 18% (w/v) PEG 8,000; GII.24 in 0.1 M sodium acetate trihydrate (pH 4.0) with 10% (w/v) PEG 4,000; GII.25 (native and H-dissaccharide complex) in 0.17 M ammonium sulfate, 0.085 M sodium cacodylate trihydrate (pH 6.5), 25.5% (w/v) PEG 8,000, and 15% (v/v) glycerol, with complex formation involving pre-incubation with H disaccharide (Dextra Laboratories) at 1:50 molar ratio for 5 h at 4°C. For X-ray diffraction, crystals were flash-cooled in liquid nitrogen using cryoprotectant solution (reservoir supplemented with 20% glycerol).

X-ray diffraction data were collected at Shanghai Synchrotron Radiation Facility (SSRF) BL19U1 and processed using HKL2000 (version 712) (Xiao et al., 2024). Data were further processed and scaled using CCP4 software (version 9) (Winn et al., 2011). The P domain structure was determined by molecular replacement using PHASER, with the crystal structure of GII.12 P domain (PDB: 3R6J) as a search model. The model was further refined using Phenix.refine in PHENIX (Adams et al., 2010; Afonine et al., 2012). Data collection and refinement statistics were summarized in Table 1. Amino acid sequence alignment was conducted using ClustalX software (version 2.1), and the result was visualized using the online ESPript software (Gouet et al., 1999; Larkin et al., 2007). Molecular docking of GII.25 P domain with B trisaccharide was performed using the CB-Dock2 server (Liu et al., 2022). Structural superposition and visualization were performed using PyMOL software (version 2.6.2) (Delano, 2002). Visualization of protein-oligosaccharide interactions was performed using LigPlot⁺ software (version 2.3) (Laskowski and Swindells, 2011).

To identify glycan-binding profiles of norovirus strains GII.23 Quininde1906, GII.24 Loreto6424 and GII.25 Dhaka1928, recombinant P proteins were expressed as glutathione S-transferase (GST) fusion proteins and subsequently released by PreScission protease cleavage at 4°C (Figure 1A). We further conducted a glycan-binding assay using GII.23/24/25 P proteins and synthetic oligosaccharides, including types A, B, H, H1, H2, H3, Le^a, Le^x, Le^b, Le^y, Lec, LacNAc, and mucin cores 1–8, with the exception of core 7 (Figure 1B). The results indicate that GII.23/24 P proteins specifically bind to the H disaccharide, while GII.25 binds to both H disaccharide and B trisaccharide (Figure 1C).

In the saliva binding assay, the samples collected from individuals with A, B, O blood group secretor (A, B, O) and non-secretor phenotypes were used to determine the binding signals of GII.23/24/25 P proteins to saliva. The results showed that GII.23 P protein exhibited positive binding mainly to saliva samples from B secretors, but no binding was observed in A/O secretors and non-secretors, except for a few individual samples. The Lewis classification method indicated that GII.23 mainly binds to the saliva samples of Lewis positive/negative B (Figure 2A). GII.24 P protein exhibits broad binding affinity for most saliva samples (including secreted type A, B, O, and non-secreted types), which is consistent with the result of Lewis classification method (Figure 2B). In contrast, the saliva-binding profile of GII.25 P protein is similar to that of GII.24, but its binding affinity is weaker. Specifically, GII.25 bound to B secretors, a portion of A/O secretors, with minimal binding to non-secretor samples. The Lewis classification method demonstrated that GII.25 protein exhibits binding capability to a variety of Lewis-positive and Lewis-negative samples (Figure 2C). The GST protein was used as a negative control.

The P dimers of GII.23 Quininde1906 and GII.24 Loreto6424 crystallized in space group P2₁2₁2₁, whereas GII.25 Dhaka1928 crystallized in space group C121. All three structures exhibited well-defined electron density maps, indicating high-quality diffraction data. These P dimers share typical global structures comparable to those of conventional P domains, including β-sheets, α-helices, distinct P1 and P2 subdomains, and a conserved stable fold overall (Figure 3). To characterize GII.25 P dimer interactions with H disaccharide (Fucα1-2Gal), we determined the co-crystal structure with the GII.25 P dimer with the natural ligand (H disaccharide) at 1.6 Å resolution. The disaccharide is bound at the dimer interface, with electron density visible in the (2mFo-DFc) omit map (Figures 4A,B). The fucose (Fuc) moiety is anchored within the P domain via hydrogen bonding with R353 (Chain A) and G444 (Chain B), while hydrophobic interactions involving A351 (Chain A) and G443/H445 (Chain B) (Supplementary Figure 1), together with van der Waals interactions involving D382 (Chain A), stabilize the binding pocket. In contrast, the galactose moiety only forms Van der Waals interactions with N352 (Chain A) and is oriented away from the protein surface (Figures 4C,D). The α-Fuc occupies an electronegative glycan-binding site (Figure 4E).

To verify the functional significance of the residues involved in the GII.25-H disaccharide interaction, single-point mutants, including A351G, N352A, R353A, D382A, G443A, G444A, and H445A, were generated by site-directed mutagenesis (Figure 1A). Compared with wild-type GII.25 P protein in saliva and glycan-binding assays, all six mutants (N352A, R353A, D382A, G443A, G444A, H445A) exhibited undetectable binding to a range of saliva samples and glycan types, demonstrating that these residues at the P2 interface are essential for ligand recognition and interaction (Figures 5A,C–H, 6A,C–H). The A351G mutant not only exhibits a complete loss of saliva-binding ability, but also displays altered carbohydrate-binding characteristics. Specifically, the mutant exhibited slightly increased binding to A, H2, Le^a, and LacNAc glycans, slightly decreased binding to the H disaccharide, and completely lost binding to the B trisaccharide (Figures 5B, 6B). These results indicate that residues N352, R353, D382, G443, G444, and H445 mediate GII.25-ligand interaction, while A351 modulates glycan specificity.

Based on the VP1 sequences (Chhabra et al., 2019), phylogenetic classification of GII NoVs reveals that GII.23/24/25 cluster in the same genetic branch during viral evolution (Supplementary Figure 2). To analyze the conservation of GII.23/24/25 P domains, three well-defined strains, GII.4 TCH05, GII.9 VA207, and GII.17 KW308, were used for sequence alignment and structural superposition. Multiple sequence alignment revealed that the GII.25 P domain shares 59.1, 52.7, and 68.2% sequence identity with the P domains of GII.4 TCH05, GII.9 VA207, and GII.17 KW308, respectively (Figure 7A). The GII.25 P protein retains the conventional GII glycan-binding motifs, and its HBGA binding pocket is formed by the P-, A-, and S-loops (Supplementary Figure 4), exhibiting high structural similarity to the three aforementioned strains, with Cα root mean square deviation (RMSD) values of 0.62 Å, 0.61 Å, and 0.62 Å, respectively. The structural conformation of P domain loops is known to be critical for ligand binding (Kubota et al., 2012). The structural comparison indicates that the N-, U-, and S-loops are relatively conserved, while the A-, B-, P-, and T-loops display low sequence identity and high structural deviation (Figure 7B). Notably, GII.23/24/25 maintain conserved GII glycan-binding characteristics, suggesting the potential emergence of later-identified GII NoVs with similar receptor-binding properties.

To analyze the structural features of the interaction between GII.23/24 and H disaccharide, we superimposed the structure of the GII.25-H disaccharide complex onto that of GII.23/24. Structural alignment revealed similar folding architecture between GII.23/GII.24 and GII.25-H disaccharide complex, with Cα RMSD values of 0.69 and 0.50 Å over 492 aligned residues. Previous studies have shown that the P domain of GII.23 Loreto1847 binds to fucose through residues R355 and D384 (Holroyd et al., 2025). Similarly, the P domain of the GII.24 Loreto 1972 interacts with the fucose moiety of HBGAs via a conserved binding pocket composed of residues N353, R354, D383 and G445 (Hansman et al., 2024). These findings are consistent with our structural superposition results (Figure 8A). Structural analysis of HBGA binding interfaces of huNoVs reveals that most GII genotypes recognize HBGAs via a set of scattered and conserved amino acids (Liu et al., 2015). Sequence alignment of interaction interfaces shows that the key residues responsible for binding glycans in GII.23/24/25 are highly conserved among most GII strains (Supplementary Figure 3A). Furthermore, residues A351, R353, D382, G443, and G444, which are involved in the interaction between GII.25 and the Fuc moiety of H disaccharide, are also well conserved, whereas residue H445 exhibits lower sequence conservation (Supplementary Figure 3B). Despite the overall structural similarity, analysis revealed sequence variations and conformational deviations in glycan-binding sites at their respective H disaccharide interfaces. Although sequence alignment shows residue H445 (GII.25) at the structurally equivalent position of Y447 (GII.23) and Y446 (GII.24), this substitution did not disrupt H-disaccharide binding affinities (Figure 8A). The predicted interaction diagrams indicate that the binding of the H-disaccharide to the GII.23 P protein is mediated by hydrogen bonds involving R355, hydrophobic interactions with A353/G445/G446/Y447, and van der Waals forces involving N354 and D384 (Supplementary Figure 5A), whereas binding to the GII.24 P protein is driven by hydrogen bonds involving R354, hydrophobic interactions with A352/G444/G445/Y446, and van der Waals forces involving N353 and D383 (Supplementary Figure 5B). Furthermore, the interaction between GII.25 P dimer and B trisaccharide was predicted using the CB-Dock2 server. The results indicated that the binding pocket is located at the interface of the dimer and is composed of residues A351, N352, R353, H355, D382, G443, G444, and H445. The B trisaccharide forms interactions with these residues through ionic bonds, hydrogen bonds, hydrophobic interactions, and van der Waals interactions (Figures 8B,C).