Rapid screening for antigenic characterization of GII.17 norovirus strains with variations in capsid gene

All full-length nucleotide sequences of the GII.17 VP1 region (> 1560 bp) were obtained from GenBank. The sequences were downloaded with the information of the accession number, collection date and geographical region. Accession numbers of sequences for phylogenetic tree construction were listed in Additional File 1. To minimize the number of the sequences with high similarity, CD-HIT software [19] was used to remove redundant sequences with a threshold of 99% similarity. Multiple sequences of the constructed dataset were aligned using Multiple Alignment in Fast Fourier Transform (MAFFT) software [20].

A best-fit nucleotide substitution model was determined using Akaike information criterion (AIC) as implemented in jModelTest [21], which was used to reconstruct a maximum-likelihood (ML) tree with a bootstrap analysis of 1000 replicates using PhyML version 3.0 [22]. Phylogenetic trees were analyzed using Bayesian inference via the Markov Chain Monte Carlo (MCMC) method in the BEAST 1.10.4 software [23]. Best substitution models (GTR + G) were determined for the selected sequences by comparison of the Bayesian Information Criterion (BIC) values using jModelTest. Molecular clock models calculated the nucleotide substitution rates of norovirus GII.17, as well as the time to the most recent common ancestor (TMRCA). The MCMC runs were performed with chain lengths of 40,000,000 steps, sampling every 4000 steps, for the selected sequences. Convergence of parameters were analyzed in the Tracer version 1.6 (http://​tree.​bio.​ed.​ac.​uk/​software/​tracer/​) and effective sample sizes (ESS) values of 200 or more were accepted. 95% highest probability density (HPD) values were used to estimate statistical uncertainty in parameter values. Files were combined in LogCombiner v1.10.4, and the maximum clade credibility trees were generated by discarding the first 10% of trees using TreeAnnotator v1.10.4. The tree was visualized in the FigTree 1.4.2.

A new emerging GII.17 was obtained from stool sample in a continuous norovirus surveillance study in Guangzhou, China of 2014–2015 [24]. Norovirus was detected in stool samples by targeting region A and C. Viral RNA was extracted from clarified supernatant using QIAamp Viral RNA MiniKit (QIAgen, Hilden, Germany). Whole-genome sequences of GII.17 were obtained using a newly established “4 + 1 + 1” PCR strategy [25]. Other GII.17 capsid genes were synthesized chemically through Generay following codon optimization. Then we cloned the GII.17 P domain construct (amino acids 222–540) of GII.17 into the pET-22b ( +) expression vector using Nde I and Xho I restriction sites.

Proteins were expressed from the plasmid using the E. coli S30 Extract System (Promega, USA), incubated at 37℃ for 1 h with shaking at 1200 rpm in a 1.5 mL Eppendorf tube. A standard reaction comprised 18 µL of E. coli S30 lysate, 20 µL of reaction mix with amino acids, and 1 µg of the plasmid template harboring the target gene, made to 50 µL with water.

The binding assay was carried out using recombinant P domain proteins (wild-type and mutants) and mAb. In brief, cell-free expressed P proteins were diluted and coated on microtiter plates at 4 °C overnight. After blocking with 1% bovine serum albumin, the two mAb (Genecreat, China) diluted in PBS were incubated with purified P domain proteins at 37 °C for 1 h. Following washing with PBST, HRP-conjugated goat anti-mouse IgG (Bioss, China) was added for incubating 30 min at 37 °C. After color development, read the absorbance at 450 nm. An OD₄₅₀ of > twofold background after background subtraction was scored positive by the assay.

P proteins were pretreated with decreasing concentrations of mAbs for 1 h at 37 °C and added to wells coated with pig gastric mucin type III (PGM, Sigma-Aldrich, USA) and incubated for 1 h at 37 °C. Followed by washing steps, mouse anti-GII.17 P domain hyperimmune serum diluted 1:2000 in PBS was added and incubated for 1 h at 37 °C. After washing with PBST, HRP-conjugated goat anti-mouse IgG was added for incubating for 30 min at 37 °C. Blocking rates were calculated by comparing binding levels of in the presence and absence of mAbs. Sera from free P proteins-immunized animals were used as negative controls.

Mutant P domain proteins with single or poly amino acid substitutions were designed and introduced into the GII.17 P domain using a wild-type templates. Site-directed residue mutations were performed using assembly PCR and the corresponding primer pairs (Additional File 2). After sequencing confirmation, mutant P domain proteins were expressed using the CFPS system and analyzed as described above.

Structural homology models representing the GII.17 capsid P domain were generated using the Swiss-Model server of the Swiss Institute of Bioinformatics [26], with default settings. Briefly, the Fasta format GII.17 capsid sequence was uploaded into the server, and the “Select Templates” option was selected. Models of P dimer was created using the templates and downloaded in.pdb format. Protein structures were visualized, and the figures were rendered using MacPyMOL version 2.1 (http://​www.​pymol.​org). Selected of strains accession numbers are summarized in Additional File 3.

After removing redundant sequences from a total of 828 sequences (as of November 20, 2019), an analysis dataset consisting of 42 complete VP1 sequences was established. A further nine complete VP1 sequences obtained from our surveillance program were also included (highlighted in red in Fig. 1), resulting in a total of 51 complete VP1 sequences. The GII.17 genotype was divided into four clades (clades A-D), which mirrored previous studies [11]. Among these, strains in clade A covered the longest time period, ranging from 1978 to 2018 (Fig. 1, modified MCMC tree for visual performance; results of ML tree shown in Additional file 4: Figure S1 similar to MCMC tree). Most mutations were found in the P2 domain.

Four wild-type representative GII.17 variants were selected for comparison according to evolutionary analysis. The sequences identity of four strains were in the middle of respective cluster, therefore could better stand for each cluster. These variants were identified in 2002 (AY502009), 2005 (DQ438972), 2013 (LC043167), and 2015 (KT970376), each representing the four clades of the GII.17 genotype (clade A, B, C, and D, selected strains were highlighted in blue in Fig. 1). Based on amino acid alignment of the capsid, GII.17–2002, GII.17–2005, and GII.17–2013 exhibited 80.43%, 78.57%, and 93.10% identity, respectively, to GII.17–2015. These differences may contribute to antigenic and HBGA binding differences among the selected GII.17 variants.

To characterize whether cell-free proteins had antibody and ligand binding abilities, we performed an EIA assay for GII.17 clade D cell-free proteins, thereby determining their antibody binding abilities. As shown in Fig. 2a, compared with the positive and negative controls, GII.17 clade D cell-free proteins bound to mAb 2A11, as previously confirmed by E. coli BL21-expressed proteins. Four strains of GII.17 bound to mAb 2A11, whereas only GII.17 clade B and GII.17 clade D bound to mAb 2D11(GII.17–2005 and GII.17–2015, Fig. 2b). These EIA data indicated mAb 2A11 was with the broad reactivity pattern on GII.17 and the mAb 2D11 was more specific to recognize region on the GII.17 clade B and GII.17 clade D strain. Therefore, to investigate potential new antigenic features, we selected mAb 2D11 to screen a panel of mutant variants. To compare the relative binding of GII.17 clade D cell-free proteins to HBGAs, we examined these proteins for binding to PGM, a well-known norovirus ligand which covers the major known binding moieties of GII.17 strains [14, 27]. The results showed that the cell-free proteins bound to PGM with unstable results in the same dataset (Additional file 5: Figure S2).

Genetic analyses indicated that the most variable amino acid changes among the four strains were in the P2 domain; as such, we introduced alanine to six potential epitopes in this domain. Reduced binding ability to mAb 2D11 indicated a unique sequence in residues 293–300, previously characterized as loop 1 (Fig. 3 and Fig. 4a) in GII.4 [28]. To examine the effects of this unique region, we created a panel of recombinant vectors with single site-directed residue mutations by replacing each residue in the region with alanine. Next, mutant variants were expressed using the CFPS system and characterized for mAb 2D11 binding using EIA (Additional file 6: Figure S3). The results indicated that no single mutation reduced the antibody binding capability, demonstrating that one residue was insufficient to affect antigenic properties. Next, four triple alanine substitution mutations were developed to dissect the role of residues 293–300. Although GII.17 clade D-293-295A restored mAb 2D11 binding, the remaining three mutant variants did not (Fig. 4b), indicating that residue 293 was less likely to affect mAb 2D11 binding. We continued to narrow the potential mAb 2D11-binding epitope. Ten additional double alanine substitution mutants were introduced to confirm the antibody-binding pocket. Considering the same residue in D300, we compared the relative binding activity of double alanine substitution mutants, finding that residue 298 was the most important factor affecting mAb 2D11 affinity. In addition, amino acids 295, 299, and 300 played a role in binding pocket formation (Fig. 4c). To further confirm the impact of the residues mentioned above, we created chimeric strains of GII.17 clade A and GII.17 clade C by substituting key residues in epitope A, previously identified in GII.17 clade D. The reason why we ignored GII.17 clade B was that GII.17 clade B and GII.17 clade D had similar binding properties with mAb 2D11. These data showed that amino acids 295–300 corresponded to the recognition site binding potency of mAb 2D11, while amino acids 293–294 promoted the binding capability (Fig. 4d).

For visual purposes, homology models were displayed to highlight changes in residues between the GII.17 strains, as well as the conformational differences between mutants and wild types. When changing residues 293–300 to alanine in GII.17 clade D, mAb 2D11 binding was lost, confirming that these are the key residues of the antigenic epitope. Homology modeling of these residues showed that they formed a conformational epitope for the structural and functional integrity of the GII.17 variant, proximal to evolving epitope A. In the model, N295, Q298, R299, and D300 presented the P domain’s unique antigenic features, likely rearranging the local protrusion of epitope A (Fig. 5).

Although we demonstrated that mAb 2D11 recognizes a unique peptide, the ligand-binding assay failed to illustrate it using cell-free proteins. Self-assemble of dimers and particles may be restricted by platform components. Additionally, a previous study investigating CFPS system development for norovirus VLPs synthesis did not apply a binding assay [29], potentially explaining the same unstable results. To determine whether changes at residues 293–300 might be relevant to the blockade epitope, we expressed the pure protein of residue 293–300 substitution using the E. coli expression system, characterizing the protein using EIA and measuring blockade potency. The cell-based protein displayed similar mAb 2D11 binding activity to that of the proteins produced by the CFPS system. However, targeted changes in epitope A did not affect the binding potency of GII.17 clade D. Hence, these changed residues may only associate with antigenic drift (Fig. 6).