Structural Adaptations of Norovirus GII.17/13/21 Lineage through Two Distinct Evolutionary Paths

Our understanding of the molecular bases behind the interplays between human noroviruses and their host glycan ligands, as well as their evolutionary changes over time with alterations in their host ligand binding capability and host susceptibility, remains limited. By solving the crystal structures of the glycan ligand binding protruding (P) domains with or without glycan ligands of three representative noroviruses of the GII.17/13/21 genetic lineage, we elucidated the molecular bases of the human norovirus-glycan interactions of this special genetic lineage. We present solid evidence on how noroviruses of this genetic lineage evolved via different evolutionary paths to (i) optimize their glycan binding site for higher glycan binding function and (ii) acquire a completely new glycan binding site for new ligands. Our data shed light on the mechanism of the structural adaptations of human noroviruses through different evolutionary paths, facilitating our understanding of human norovirus adaptations, evolutions, and epidemiology.

ABSTRACT

Human noroviruses (huNoVs), which cause epidemic acute gastroenteritis, recognize histo-blood group antigens (HBGAs) as host attachment factors affecting host susceptibility. HuNoVs are genetically diverse, containing at least 31 genotypes in the two major genogroups (genogroup I [GI] and GII). Three GII genotypes, GII genotype 17 (GII.17), GII.13, and GII.21, form a unique genetic lineage, in which the GII.17 genotype retains the conventional GII HBGA binding site (HBS), while the GII.13/21 genotypes acquire a completely new HBS. To understand the molecular bases behind these evolutionary changes, we solved the crystal structures of the HBGA binding protruding domains of (i) an early GII.17 variant (the 1978 variant) that does not bind or binds weakly to HBGAs, (ii) the new GII.17 variant (the 2014/15 variant) that binds A/B/H antigens strongly via an optimized GII HBS, and (iii) a GII.13 variant (the 2010 variant) that binds the Lewis a (Le^a) antigen via the new HBS. These serial, high-resolution structural data enable a comprehensive structural comparison to understand the evolutionary changes of the GII.17/13/21 lineage, including the emergence of the new HBS of the GII.13/21 sublineage and the possible HBS optimization of the recent GII.17 variant for an enhanced HBGA binding ability. Our study elucidates the structural adaptations of the GII.17/13/21 lineage through distinct evolutionary paths, which may allow a theory explaining huNoV adaptations and evolutions to be put forward.

IMPORTANCE Our understanding of the molecular bases behind the interplays between human noroviruses and their host glycan ligands, as well as their evolutionary changes over time with alterations in their host ligand binding capability and host susceptibility, remains limited. By solving the crystal structures of the glycan ligand binding protruding (P) domains with or without glycan ligands of three representative noroviruses of the GII.17/13/21 genetic lineage, we elucidated the molecular bases of the human norovirus-glycan interactions of this special genetic lineage. We present solid evidence on how noroviruses of this genetic lineage evolved via different evolutionary paths to (i) optimize their glycan binding site for higher glycan binding function and (ii) acquire a completely new glycan binding site for new ligands. Our data shed light on the mechanism of the structural adaptations of human noroviruses through different evolutionary paths, facilitating our understanding of human norovirus adaptations, evolutions, and epidemiology.

INTRODUCTION

Human noroviruses (huNoVs), members of the Norovirus genus in the family Caliciviridae, are the most important viral pathogens causing epidemic acute gastroenteritis. The viruses are highly contagious, infecting millions of people, with these infections often leading to large outbreaks in a wide variety of settings and claiming about 218,000 lives worldwide each year (1). HuNoVs are genetically diverse, containing at least 31 genotypes in the two major huNoV genogroups, genogroup I (GI) and GII, where GII contains 22 genotypes that are responsible for most epidemics in humans. While multiple genotypes often cocirculate during an epidemic season, major huNoV epidemics are usually associated with a single predominant genotype. For instance, GII genotype 3 (GII.3) was commonly detected in the 1970s (2), whereas the GII.4 genotype has been dominant for the past two decades worldwide (3, 4).

HuNoVs recognize histo-blood group antigens (HBGAs) as host attachment factors that play a critical role in huNoV host susceptibility and host ranges (5–8). HBGAs are complex, fucose-containing glycans that determine the polymorphic human blood types, including the A, B, O (H), Lewis, and secretor/nonsecretor types. In addition to being located on red blood cells, HBGAs also distribute abundantly on the mucosal epithelia of the human intestinal tract, where they serve as attachment factors for huNoV infection. Many previous studies, including in vitro binding assays (9–11), human intestinal enteroid-based huNoV culture (12), huNoV outbreak investigations (13, 14), and human challenge studies (15, 16), have collectively demonstrated that HBGAs are critical host susceptibility factors affecting huNoV host susceptibility, host range, and, probably, prevalence (5–8). Thus, further defining the interplays between huNoVs and HBGAs and their evolutionary changes over time will shed light on huNoV evolution, diversity, and epidemiology and will help with the development of strategies for huNoV surveillance.

The interactions between the genetically diverse huNoVs and the polymorphic HBGAs are complex, with huNoVs being able to adapt to various HBGA types, likely to enlarge their target populations (17). HuNoVs interact with HBGAs via the protruding (P) domains of the viral capsids, in which the HBGA binding site (HBS) has been functionally identified (18–20) and structurally elucidated (21–23). The HBSs are highly conserved within each of the GI or GII huNoVs but are distinct between the two genogroups (6, 24, 25). The crystal structures of the HBSs representing GI and GII huNoVs have been elucidated, revealing complex hydrogen and hydrophobic interactions between the major saccharides of the HBGAs and the amino acids of the HBSs (21, 22, 26–29). Comparisons among the known HBS structures of different huNoVs interacting with various HBGAs point to continued adaptations of huNoVs to diverse HBGA types, potentially to reach new target populations.

Phylogenetic analysis (30) showed that the 22 GII genotypes can be sorted into several genetic lineages, among which the GII.17/13/21 genotypes constitute a unique lineage (Fig. 1). Among the genotypes within this lineage, GII.17, which exhibits a higher sequence similarity with the mainstream GII huNoVs than the other genotypes, retains the conventional GII HBS (31). In contrast, both genetic and structural analyses indicated that GII.13/21 emerged from GII.17 as a new sublineage that acquires a completely new HBS (23, 32). We previously solved the crystal structure of a GII.21 P domain in complex with a Lewis a (Le^a) antigen and verified the new HBS of the GII.21 genotype (23). However, these structural data alone could not explain how the new HBS emerged due to the lack of structural data for its neighboring GII.13 huNoV and the closely related GII.17 huNoV as an evolutionary linker between the mainstream GII huNoVs and the GII.13/21 genotypes.

FIG 1.

Phylogenetic tree of GII noroviruses that is redrawn as described previously (30). The 22 GII genotypes form several genetic lineages, while the unique lineage consisting of the GII.17, GII.13, and GII.21 genotypes is indicated by purple lines and a purple frame.

Earlier GII.17 variants bound HBGAs weakly (33), a possible reason for their low prevalence. Notably, the GII.17 genotype was found to be predominant over the 2014-2015 winter season due to a newly emerged variant, leading to major epidemics in China and other southeast Asian countries or regions (14, 33–39). Accordingly, this new GII.17 variant was found to gain a stronger HBGA binding ability than the previous variants (14, 33). Recent studies on the P domain structures of three GII.17 variants isolated in 2002, 2014, and 2015 (31, 40) provide important structural information about the GII.17 HBSs. However, systematic structural and functional characterizations of representative strains of the GII.17/13/21 lineage are necessary to fully unveil the molecular mechanism behind the structural adaptations and evolution of the GII.17/13/21 lineage.

In this study, we solved the crystal structures of the P domains with or without HBGAs of three carefully selected huNoVs, each representing (i) an earlier low-prevalence GII.17 variant (the 1978 variant) with a low or no HBGA binding ability; (ii) the new variant (2014/15 variant), which is the highly prevalent GII.17 variant with a strong HBGA binding ability via an optimized GII HBS; and (iii) a GII.13 variant (the 2010 variant) that binds Le^a via the new HBS. Our new data enabled comprehensive structure-based genetic and phenotypic analyses to understand HBS optimization of the new GII.17 variant for enhanced HBGA binding competency and the emergence of the completely new HBS of the GII.13/21 sublineage. Our data indicate two distinct evolutionary paths of the GII.17/13/21 lineage for an enhanced HBGA binding ability, likely to enlarge its target populations.

RESULTS

HBGA binding variations among various GII.17 variants.

The P domain proteins of three GII.17 variants (Fig. 2A), each representing the earliest isolated variant (the 1978 variant, GenBank accession number JN699043), a new variant that was isolated during the 2014-2015 epidemic season (the 2014/15 variant, GenBank accession number KR020503), and another variant (the 2002 variant, GenBank accession number AY502009) that was isolated in between the other two variants, were produced. HBGA binding assays using the purified P domain proteins and well-defined saliva samples representing A, B, O, and nonsecretor (N) blood types indicated that the earlier GII.17 variants (1978 and 2002) bound HBGAs weakly with marginal binding signals below an optical density (OD) of 0.2, while the new GII.17 variant bound type A, B, and O saliva samples strongly (Fig. 2B). These HBGA binding profiles appeared to be consistent with the low/high prevalence of the variants, in which the rarely detected earlier GII.17 variants (1978 and 2002) bind HBGAs weakly, while the highly prevalent GII.17 variant (2014/15) binds HBGAs strongly.

FIG 2.

Phylogenetic relationships of the three selected GII.17 variants and their HBGA binding variations. (A) Phylogenetic relationships of the major circulating GII.17 variants during the past 36 years, with GenBank accession numbers and isolation years being indicated. The three selected variants are framed in red, with the times (in years) spanning their isolation being indicated. The phylogenetic tree was made based on VP1 sequences using the neighbor-joining method (the distance calculation was done by use of the Kimura 2-parameter correction with pairwise deletion) implemented in the MEGA (version 4.1) program, and the results were validated by the use of 1,000 bootstrap replicates. (B) HBGA binding profiles of the three selected GII.17 variants. Well-defined saliva samples representing A, B, O, and nonsecretor (N) blood types are shown on the x axis. The binding signal intensities are shown on the y axis (optical density [OD]). (C to E) P dimer crystal structures of the GII.17 1978 and GII.17 2014/15 variants with HBGA oligosaccharides. Cocrystallization followed by X-ray crystallography showed that the 2014/15 variant P domain formed complexes with A trisaccharide (A tri) (D) or B trisaccharide (B tri) (E), while the 1978 variant P domain did not (C). The P domain dimers are shown in a surface model (gray), while the HBGA trisaccharides are shown in a sphere model, in which carbon, oxygen, and nitrogen atoms are colored cyan, red, and blue, respectively.

The P domain proteins of the 1978 and 2014/15 variants were cocrystalized individually with oligosaccharides representing A, B, H1, Le^a, Le^b, Le^x, and Le^y antigens. The resulting crystal structures indicated that the P domain dimer of the new GII.17 variant (2014/15) bound A and B trisaccharides (tri) (Fig. 2D and E), while the P dimer of the 1978 variant did not reveal any HBGA in the crystal structure (Fig. 2C), consistent with the above-described saliva-based binding results.

Structural changes of GII.17 P domains over times.

The GII.17 1978 and GII.17 2014/15 P domain dimers share typical global structures, β-sheet/α-helix/loop order organizations, and P1/P2 arrangements (Fig. 3) with those of other known GII.17 huNoVs (31) and other GII huNoV P domains (21, 22, 28, 31), indicating stable P domain core structures among GII huNoVs. Nevertheless, significant differences in the P dimer surface conformations, particularly those formed by the loops between the β-sheets, are clearly observed between the 1978 and 2014/15 variants (Fig. 3), indicating their evolutionary changes over time. These differences may contribute to the observed antigenic differences among different GII.17 variants (33), as well as those between GII.17 huNoVs and other GII genotypes (41).

FIG 3.

Structural comparisons of P dimers among selected GII.17 variants. (A to D) Superimposition of the P dimer structures of the 1978 (gray) and 2014/15 (cyan) GII.17 variants in a cartoon model (A and C) and surface model (B and D). The superimpositions are shown as side (A and B) and top (C and D) views. The A trisaccharides that bind to the top of the 2014/15 P domains are shown in a stick model, in which the carbon, oxygen, and nitrogen atoms are shown in green, red, and blue, respectively.

HBS of the new GII.17 variant.

While the previous structural study (31) could not solve the crystal structures of a GII.17 P domain-HBGA complex, we were successful in determining the GII.17 (2014/15) P domain in complex with A and B trisaccharides (tri). This allowed us to define the exact GII.17 HBS for the first time. The P domain-tri complexes were crystallized in the P2₁2₁2₁ space group, with one homodimer occurring in each asymmetric unit. The A tri and B tri were clearly visible in the (2mF_o − DF_c ) difference electron density maps, with all three rings of the A/B-tri being well fitted into the map (Fig. 4A and B). Two symmetric HBSs were identified on the top surface of the P dimer (Fig. 2D and E, Fig. 4A and B), and each of these was formed by 5 residues from both P domain protomers of the GII.17 2014/15 variant (Fig. 4). Unlike those of the GII.13/21 sublineage, these HBSs share conserved locations and amino acid compositions with the conventional GII HBSs (7, 8).

FIG 4.

Structures of the HBGA binding sites (HBSs) of the GII.17 2014/15 variant and their interactions with the A and B trisaccharides. (A and B) (2mF_o − DF_c) omit electron density maps of the A (A) and B (B) trisaccharides. The omit maps were created using the final P domain structure of the GII.17 2014/15 variant without the trisaccharide, and the mesh map was contoured at 2.0σ (blue) around the selection site with a coverage radius of 1.6 Å. Carbon, oxygen, and nitrogen atoms are indicated in cyan, red, and blue, respectively. The HBS regions (cartoon representations, yellow) with indications of the residues (stick representation) that interact with the A and B trisaccharides (stick representation, cyan) via hydrogen bonds (blue dashed lines) are shown. The individual saccharides of the A and B trisaccharides are shown. (C and D) Schematic illustration of the interacting networks between the side/main chains of the amino acids at the HBS with the individual atoms of the A (C) and B (D) trisaccharides. The hydrogen bonds are indicated by dashed lines, with the distances being indicated. The Van der Waals interaction between α-Fuc and Y444 is also shown.

CBP of the GII.17 HBSs.

Like other conventional GII HBSs (7, 8), the GII.17 HBS has a highly conserved central binding pocket (CBP) that constitutes the major structures of the HBS and interacts with the α-1,2 fucose (α-Fuc), the H epitope, of the A and B antigens (Fig. 4). CBP is built by three separate, highly conserved components, including T348, R349 (site I), D378 (site II), G443, and Y444 (site III) (Fig. 5A). The CBP bottom is composed of T348 and R349, which interact with atoms O-2 to O-5 of the α-Fuc of the HBGAs via strong hydrogen bonds. The walls of the CBP are formed by D378, G443, and Y444, which surround the α-Fuc, β-Gal, and α-GalNAc/α-Gal of HBGAs (Fig. 4). Particularly, the aromatic ring of Y444 interacts with the hydrophobic methyl group of α-Fuc through Van der Waals interactions, which are commonly found to contribute to complex stability among most GII HBSs.

FIG 5.

Optimization of the HBGA binding sites (HBSs) of the GII.17 2014/15 variant from earlier GII.17 variants. (A) Comparison of the three conserved sequence components (I, II, and III, red font) of the HBSs and their neighboring sequences among major GII.17 variants. (B and C) Structural superimpositions of the HBS (cartoon representation) of the 2014/15 variant (yellow) with that of the 1978 variant (pink) (B) and that of the 2002 variant (green) (C). The amino acids that form the HBS are in stick representations. The B trisaccharides (stick representation, cyan) are also indicated. The Y444 of the 2014/15 variant and the V444 of the 1978 and 2002 variants are identified by red arrows. (D) The HBS (colored surface representation) of the 2014/15 variant (gray) interacts with the A trisaccharide (stick representation). The important Van der Waals interaction between α-Fuc and Y444 is shown by a yellow dashed line. (E) The HBS region (colored surface representation) of the 1978 variant (gray) with the modeled A trisaccharide (stick representation) based on the HBS of the 2014/15 variant. The missing Van der Waals interaction between α-Fuc and HBS due to the lack of Y444 (red cross) is shown by a yellow dashed line.

HBS structural optimization of the GII.17 2014/15 variant.

To understand how the GII.17 2014/15 variant gained an enhanced HBGA binding capability over that for the previous variants, which did not bind or bound weakly to HBGAs (Fig. 2B), the HBS structure of the 2014/15 variant was superimposed with those of the 1978 and the 2002 (31) variants (Fig. 5B and C). While both the 1978 and 2002 variants exhibited a similar HBS structure constituted by four (T348, R349, D378, and G443) conserved residues like the HBS of the 2014/15 variant, both earlier 1978 and 2002 variants lacked the important Y444 that forms the required Van der Waals interactions with the α-Fuc of HBGAs (Fig. 5B to E). As a result, the HBSs of the 1978 and 2002 variants are structurally and functionally compromised, leading to the observed weak HBGA binding ability of the two earlier GII.17 variants (Fig. 2B). Genetic analysis indicated that all earlier GII.17 variants share V444 and, thus, lack the Van der Waals interactions provided by Y444 (Fig. 5A), explaining their low HBGA binding ability. In contrast, all new GII.17 variants starting from 2013 gained a tyrosine at position 444 to rebuild the intact and functional HBSs, explaining their high HBGA binding capability. Thus, the V444Y mutation optimized the HBS, enhancing the HBGA binding ability, which may be a reason for the improvement of the prevalence of the new GII.17 variant.

Structural integrity for a functional HBS of the GII.17 2014/15 variant.

We first assessed the importance of Y444. As expected, the single mutation of Y444 to a valine or an alanine dramatically reduced the HBGA binding function of the GII.17 2014/15 variant P domain (Fig. 6A and B), supporting the suggestion that the V444Y mutation is a vital factor for the GII.17 2014/15 variant to gain its HBGA binding capability. In addition, we individually introduced the single mutations R349A, D378A, and G443A to each of the three major HBS components (Fig. 5A) of the GII.17 2014/15 variant. All individual mutations radically reduced the HBGA binding capability (Fig. 6A and B), indicating the importance of these residues for the structural and functional integrity of the GII.17 2014/15 variant HBS.

FIG 6.

Assessment of the structural and functional integrity of the HBGA binding site (HBS) of GII.17 variants by mutagenesis study. (A and B) Binding of wild-type (WT) and mutant P domains of the GII.17 2014/15 variant to type A (A) and B (B) saliva samples. Each mutant P domain has a single mutation at one major HBS component. (C and D) Binding of wild-type P domains of the GII.17 2014/15 and 1978 variants, as well as reverse mutants, to type A (C) and B (D) saliva samples. Each reverse mutant has a V444Y or V444H mutation at the HBS of the 1978 variant to restore the HBGA binding capability. The binding signal intensities (OD) are shown on the y axis, while the wild-type and mutant P domains at the indicated concentrations are shown on the x axis. The dashed lines show an OD of 0.1. (E) Further optimization assessment of the 1978 P domain with the V444Y or V444H reverse mutation in binding to type A, B, and O saliva samples. Statistically significant differences are indicated (*, P < 0.05; **, P < 0.01; ***, P < 0.001).

HBS optimization of the GII.17 1978 variant.

Genetic and structural analyses indicated that among most GII HBSs a conserved aromatic residue appears at the Y444 equivalent position forming the critical Van der Waals interaction (6, 21, 22, 25, 28, 29). This prompted us to introduce a V444Y or a V444H mutation into the 1978 variant P domain for potential HBS optimization. Interestingly, both the V444Y and V444H mutations significantly enhanced the ability to bind to type A, B, and O saliva samples (P < 0.05, except that the V444Y mutant showed no binding to type B saliva) (Fig. 6C to E). These data verified the importance of Y444/H444 for HBS structural integrity. However, we also noted that the 1978 variant P domain with the V444Y or V444H mutation could not restore the levels of the HBGA binding signals to the levels of those for the 2014/15 variant (Fig. 6C and D), indicating that other mutations around the HBS that accumulated over time also contribute to the strong HBGA binding function of the GII.17 2014/15 variant.

Le^a antigen binding feature of GII.13 NoV.

To understand the GII.13 HBS and how the GII.13/21 sublineage acquires a completely new HBS, we first determined the Le^a antigen binding feature of the GII.13 2010 variant. Saliva binding assays indicated that the P domain protein of the GII.13 2010 variant strongly bound Le^a-positive saliva samples from nonsecretor donors but did not bind those from secretor donors, including those from Le^a-negative type A, B, or O secretor individuals (Fig. 7F), confirming its Le^a antigen binding property.

FIG 7.

Structures of the GII.13-2010 variant P dimer, HBGA binding site (HBS), and Le^a antigen binding function. (A and B) Structures (cartoon representation) of the native P dimer (A) and its Le^a antigen binding complex (B) for the GII.13 2010 variant. The glycerol molecules (sphere representation) are shown. (C) (2mF_o − DF_c) omit electron density maps of the Le^a trisaccharide. The omit maps were calculated using the final P domain structure of the GII.13-2010 variant without the trisaccharide, and the mesh map was contoured at 2.0σ (blue) around the selection site with a coverage radius of 1.6 Å. Carbon, oxygen, and nitrogen atoms are indicated in cyan, red, and blue, respectively. The HBS regions (cartoon representations, pink) with indications of the residues (stick representation) that interact with the Le^a trisaccharide (stick representation, cyan) via hydrogen bonds are shown. (D) Schematic illustration of the interacting networks between the side/main chains of the amino acids at the HBS with the individual atoms of the Le^a trisaccharides. The hydrogen bonds are indicated by dashed lines, with distances being indicated. Water-bridged hydrogen bonds are indicated by a circled W. (E) Structural superimposition of the GII.13 (pink) and GII.21 (yellow) HBSs, indicating their conserved residue compositions and side chain arrangements. (F) Binding of the P domain protein of the GII.13 2010 variant to Le^a antigen-positive nonsecretor saliva samples, but not Le^a antigen-negative ones, including those from Le^a-negative type A, B, or O secretor individuals. (G) Loss of binding abilities of different GII.13 P domain mutants with a single-residue mutation at the HBS to Le^a antigen-positive saliva samples.

Structures of the GII.13 P domain and HBS.

We then solved the crystal structures of this GII.13 P domain and its complex with the Le^a trisaccharide (Fig. 7A to E). The native P protein crystallized in the space group of P2₁2₁2₁, with two protomers in an asymmetric unit, exhibiting a typical huNoV P dimer architecture. Two glycerol molecules were found to occupy the P2 surface pocket corresponding to the GII.13 HBS (Fig. 7A; see below), a scenario that was also seen in the GII.21 strain OIF P dimer (23). Unexpectedly, two additional glycerol molecules were found to bind two novel pockets at the P1/P2 interface at the waist region of the P dimer (Fig. 7A and B).

The observed glycerol-P domain interactions prompted us to grow GII.13-Le^a tri complex crystals under glycerol-free conditions, which we could possibly control. Even so, bound glycerol molecules were still found at the waist pockets (Fig. 7B), indicating a high glycerol binding affinity at these sites. Two bound Le^a tri glycans were seen at two top pockets that otherwise bound glycerols (Fig. 7A to D; see above). The Le^a tri is stabilized by eight residues, including D296, W298, and D299 from the B loop; S357 and T359 from the N loop; and N395, N397, and T398 from the T loop (Fig. 7C and D), through hydrogen bonds and a hydrophobic interaction. These HBS components and their hydrogen bonds/hydrophobic interactions with Le^a tri are similar to those of GII.21 OIF but distinct from those of other GII huNoVs, supporting the notion that GII.13/21 genotypes were from a common ancestor. While this GII.13 huNoV P domain shares only 73% sequence identity with that of GII.21 OIF, structural superimposition indicated that the locations, compositions, and residue side chain arrangements of their HBSs are highly conserved. The only obvious difference is that the GII.13 N loop is slightly more flexible to fit the Lewis epitope binding at this site (Fig. 7E)

Validation of the GII.13 HBS.

The observed GII.13 HBS was verified by site-directed mutagenesis. When the four important amino acids that constitute the GII.13 HBS were individually changed to an alanine, the binding abilities of the mutant GII.13 P domains to the Le^a antigen were completely abolished (Fig. 7F and G), supporting the suggestion that the identified HBS is indeed responsible for the observed Le^a antigen binding function.

Evolutionary changes of GII.17/13/21 HBSs.

The solved P domain structures of an earlier GII.17 variant, the recent GII.17 variant, and a GII.13 norovirus (NoV), together with the previously known structures of a GII.21 P domain (23) and other GII P domains, enable a comprehensive comparison to understand the evolutionary changes of the HBSs in the GII.17/13/21 lineage (Fig. 8). We noted that the prototypic pocket structure equivalent to that in the GII.13/21 HBSs can be clearly recognized in the GII.17 1978 variant, which lacked a proper GII HBS (Fig. 8; compare Fig. 8B with Fig. 8D, green). We found that the relevant residues are dispersed in the GII.4 1998 variant, representing mainstream GII NoVs (Fig. 8A, green). These data suggest that the lack of a properly functional GII HBS of an earlier GII.17 variant may be a major driving force to develop the new HBS. Accordingly, the new HBS equivalent pocket structure deformed in the GII.17 2014/15 variant (Fig. 8, compare Fig. 8B with Fig. 8C, green), which has a highly functional GII HBS (Fig. 8, red). These scenarios shed light on the emergence of the GII.13/21 sublineage from an earlier GII.17 huNoV.

FIG 8.

Evolutionary changes of the conventional GII HBGA binding site (HBS) and the new HBS among members of the GII.17/13/21 lineage. (A) P domain structure of the GII.4 1998 variant representing mainstream GII huNoVs with a well-functioned conventional GII HBS. (B) P domain structure of the GII.17 1978 variant representing an earlier GII.17 NoV with an improperly functioning conventional GII HBS. (C) P domain structure of the GII.17 2014/15 variant with a well-functioning conventional GII HBS. (D) P domain structure of the GII.13 2010 variant representing the GII.13/21 sublineage with a new HBS. The amino acids that form the conventional GII HBSs are shown in red, while those that constitute the new HBSs in GII.13/21 are shown in green.

The GII.17/13/21 lineage evolved via two distinct evolutionary paths.

We also traced the residues that constitute the conventional GII HBSs (Fig. 8, red). While the typical GII HBS remained in both earlier (1978) and the new (2014/15) GII.17 variants, it degenerated completely in the GII.13/21 genotypes, most likely due to the loss of its HBGA binding function and, thus, selection pressure (Fig. 8; compare Fig. 8D with Fig. 8A, B, and C, red). In contrast, the GII.17 2014/15 variant optimized its conventional GII HBSs, leading to an enhanced HBGA binding ability. Taken together, these data support the notion that the GII.17/13/21 lineage evolved via two distinct evolutionary paths, resulting in the optimized GII HBS of the GII.17 2014/15 variant and the completely new HBS of the GII.13/21 sublineage.

DISCUSSION

HBGAs are an important host susceptibility factor that play an essential role in huNoV host range and evolution (7, 8, 42). While huNoVs are known for their genetic and antigenic diversity, they are also diverse in recognizing the polymorphic human HBGAs, through which huNoVs continue to adapt to new HBGAs, likely to target new populations, causing epidemics of acute gastroenteritis as well as diverging and spreading worldwide. Great efforts have been made during the past few years to understand the mechanisms behind these adaptations and diversity. Noteworthy is that the previously rare GII.17 genotype attracted a lot of attention during the 2014-2015 epidemic season due to the increase in epidemics caused by the new GII.17 variant (14, 33–39). Unlike its earlier (before 2002), low-prevalence counterparts that weakly bind HBGAs (Fig. 2), the new GII.17 2014/15 variant exhibited a strong HBGA binding function (33). In addition, we noted that the GII.17 huNoVs are likely the evolutionary linker between the unique GII.13/21 sublineage with the new HBS and the mainstream GII huNoVs that share the conventional GII HBS (Fig. 1 and 9). These distinctive features make the GII.17 huNoVs an excellent model to study huNoV adaptation, diversity, and evolution. Through the comprehensive genetic, structural, and phenotypic studies described in this report, we provide multiple lines of evidence in support of two distinct evolutionary paths of the GII.17/13/21 lineage through structural adaptations of their HBSs for an improved HBGA binding capability and probably also prevalence. Our new data shed light on the mechanisms behind the emergence of the GII.13/21 sublineage from GII.17 huNoVs and the occurrence of the new GII.17 variant.

FIG 9.

Schematic illustration of two distinct evolutionary pathways of the GII.17/13/21 lineage. The low-HBGA-binding earlier GII.17 variants gained an improved HBGA binding function through two distinct pathways. One was to develop a completely new HBGA binding site (HBS), leading to the emergence of the GII.13/21 sublineage. The second one was to optimize the conventional GII HBS to gain an enhanced HBGA binding capability.

Based on the VP1 sequence changes over time, the GI and GII genotypes are divided into two major groups, the evolving GII.4 genotypes and the static non-GII.4 genotypes (43, 44). The two groups of NoVs appear to evolve via distinct trajectories, in which the evolving GII.4 NoVs change rapidly with the frequent emergence of new variants that replace the previous ones, while the static non-GII.4 genotypes seem to be more stable, with many fewer variants occurring. The two observed revolutionary paths in this study may apply to the non-GII.4 NoVs, leading to the current status of a wide distribution and diversity. Unlike GII.4 NoVs, non-GII.4 NoVs, including GII.17, apparently evolved in a nonsequential manner, where new variants did not replace old ones but cocirculated with old ones (43, 44). However, we noted that although it is defined as a static genotype, the GII.17 genotype has had at least four variants during the past 38 years, representing the fastest-evolving genotype among the static non-GII.4 genotypes (43, 44). Such a relatively fast-evolving status could be a driving force for the emergences of the new GII.17 variants and the GII.13/21 sublineage.

Adaptation to new hosts to widen the target populations is an intrinsic driving force of viral evolution and survival. The generally low HBGA binding capability and low prevalence of the earlier GII.17 variants may be driving forces of adaptations for a higher HBGA binding capability and a higher prevalence. Our data show that one pathway to reach these goals is to develop a new HBS that binds a new set of HBGAs (23), leading to the emergence of the GII.13/21 sublineage from an earlier GII.17 variant (Fig. 9). Indeed, we have shown that both GII.13 and GII.21 huNoVs of the GII.13/21 sublineage share the new HBS (Fig. 7) (23), and they both strongly bind the Le^a antigen (Fig. 7) (9, 32). Accordingly, active outbreaks caused by the GII.13/21 huNoVs have often been observed (45–47) (NoroNet).

Our study has at least partially elucidated the emergence mechanism of the new HBS of the GII.13/21 sublineage via a comprehensive comparison of the P domain structures of two GII.17 NoVs, a GII.13 NoV, a GII.21 NoV, and other GII NoVs (Fig. 8). The preliminary site equivalent to the new HBS can be recognized on the P domain of the earlier GII.17 1978 variant but not in the mainstream GII huNoVs, such as GII.4 NoVs, supporting the notion that the GII.13/21 sublineage might emerge from an earlier GII.17 huNoV. The fact that the GII.13/21 sublineage has diverged into two genotypes indicates that this sublineage has been emerged for a long time. We also note that the pocket structure of the new HBS further deformed in the GII.17 2014/15 variant (Fig. 8C), likely due to the loss of its function and the selection pressure in this GII.17 variant. For the same reason, the amino acids that constitute the conventional GII HBS have moved away from each other in both the GII.13 and the GII.21 huNoVs (Fig. 8D).

The second evolutionary path of the GII.17/13/21 lineage is to optimize the conventional GII HBS for an enhanced HBGA binding function, resulting in the emergence of the new GII.17 variant (Fig. 9). Comparing the P domain structures between the low-prevalence GII.17 1978 variant that does not bind HBGAs and the highly prevalent GII.17 2014/15 variant that strongly binds HBGAs revealed the structural and functional adaptation of the new GII.17 variant via minor structural mutations. The V444Y mutation at the HBS of the new GII.17 2014/15 variant restored the missing Van der Waals interactions that are required for an intact, functional HBS. Thus, the V444Y mutation led to the significantly increased HBGA binding function of the new GII.17 2014/15 variant, and this is also probably the reason for its high prevalence. This notion was supported by our data that the V444Y or V444H mutation (both aromatic Y and H are commonly found at this position; see below) can partially restore the HBGA binding function of the GII.17 1978 variant. The reason that the HBGA binding function of the GII.17 1978 P domain V444Y or V444H mutant was not restored to the level of the new GII.17 2014/15 variant may be due to allosteric effects and possible contributing roles of other mutations near the HBS of the new GII.17 variants that have been found since 1978. These other mutations may also help to improve the HBGA binding outcomes of the new GII.17 2014/15 variant.

An aromatic residue (mostly a Y) is commonly present at the Y444 position of the HBSs of the mainstream GII huNoVs to form the important Van der Waals interaction with the bound HBGAs. It is plausible to assume that the ancestral GII.17 virus had an intact HBS with Y444, which is supported by the recent isolation of a GII.17 variant that circulated in 1976 (48) (see below). However, it remains unclear why and how all the other known earlier GII.17 variants recovered before 2013 lost this required aromatic amino acid and replaced it with a valine, leading to a functionally damaged HBS. Due to the limitations of the current approaches and data, it also remains unknown whether the damaged HBSs in the earlier GII.17 variants lost their HBGA binding ability completely or retained a weak HBGA binding ability. We tend to believe that the earlier GII.17 variants retain a weak HBGA binding function for several reasons: (i) although they are rare, the earlier GII.17 variants with the damaged HBSs continue to infect humans, causing epidemics from time to time; (ii) the damaged HBSs among the earlier GII.17 variants have remained conserved over time, indicating a functional selection pressure; and (iii) the new GII.17 2014/15 variant optimized its conventional GII HBS for an enhanced HBGA binding function. Further study to clarify this issue is necessary.

It was noted that an early GII.17 variant, Tokyo/27-3/1976 (referred as GII.17 1976), causing gastroenteritis in 1976 was recently isolated (48). Phylogenetic analysis using VP1 sequences indicated that this GII.17 virus formed an independent cluster among the known GII.17 viruses, showing a closer genetic relationship to the new GII.17 variants than to the earlier ones (48). This prompted us to inspect the putative HBSs of the GII.17 1976 variant and found that it has an intact HBS, like the new GII.17 2014/15 variant, including Y444. This finding indicates that an earlier GII.17 variant with an intact HBS existed in 1976, but for unknown reasons it has remained undetectable over the past 37 years. Thus, the new GII.17 variant that emerged in 2013 could be from a GII.17 1976-like virus (44, 48). If this is true, the GII.17 1976 variant must have been evolving in unknown hosts over the past three decades, because it shares only 88% VP1 sequence identity with the new GII.17 variant. Further study is needed to clarify this scenario.

MATERIALS AND METHODS

Selection of studied huNoVs.

Three GII.17 variants and one GII.13 variant were selected. The three GII.17 variants were isolated in 1978, 2002, and 2014-2015, respectively, with variant each representing (i) the earliest found variant (the 1978 variant, GenBank accession number JN699043), (ii) the currently circulating variant (the 2014/15 variant, GenBank accession number KR020503), and (iii) a variant recovered in between the other two variants (the 2002 variant, GenBank accession number AY502009) (Fig. 2A) (33). The GII.13 variant (the 2010 variant, GenBank accession number BAQ94583) represents the GII.13 genotype with an unknown HBS structure. The capsid protein (VP1)-encoding sequences of these variants were synthesized chemically through GenScript (Piscataway, NJ) after codon optimization for enhanced expression in an Escherichia coli system.

Plasmid constructs and recombinant protein production.

To produce P domain proteins for crystallization and structure determination, the synthesized DNA sequences encoding the P domain of the GII.17 2014/15 (residues 222 to 530), GII.17 1978 (residues 222 to 530), and GII.13 2010 (residues 222 to 538) variants were cloned into the pGEX-6P-1 expression vector (GE Healthcare Life Sciences) using the EcoRI and XhoI sites. The P domain proteins were expressed as glutathione S-transferase (GST) fusion proteins in Escherichia coli BL21(DE3) as described previously (18, 24). The GST tag was removed from the target P domain protein by PreScission protease (GE Healthcare Life Sciences) cleavage at 4°C overnight. The P proteins were further purified through a Resource Q anion exchange column (GE Healthcare Life Sciences), using a buffer containing 20 mM HEPES (pH 7.5), with P domain proteins being eluted at approximately 100 mM NaCl.

To prepare proteins for HBGA binding assays and the mutagenesis study, the P domain-encoding sequences of the four viruses were individually cloned into the plasmid vector pGEX-4T-1 (GE Healthcare Life Sciences) via BamHI and NotI sites. The recombinant P domain proteins were expressed as described elsewhere (18, 24, 32). The GST tag was removed from the target proteins by thrombin cleavage. The proteins were quantitated by SDS-PAGE using serially diluted bovine serum albumin (BSA; Bio-Rad) as the standard on the same gels (49).

HBGA binding assay.

The HBGA binding assay was carried out using the recombinant P domain proteins and our well-characterized saliva samples with known HBGA phenotypes, as described previously (9, 18, 24, 32). Briefly, a panel of boiled and diluted (1:1,000) saliva samples representing A, B, O (H), and nonsecretor blood types was coated on microtiter plates at 4°C overnight. After blocking with 5% (wt/vol) nonfat milk, the coated saliva samples were incubated with the purified P domain proteins at the indicated concentrations (6.3 to 50 ng/µl) for 60 min at 37°C. The bound P domain proteins were detected as described previously, using an in-house hyperimmune rabbit serum against various huNoV virus-like particles (VLPs) (9, 32).

Production of P domains containing single mutations at HBS.

Single-residue mutations were introduced individually into the HBS of the GII.17 P domain by site-directed mutagenesis using the expression plasmids of the corresponding wild-type P domains as templates. Mutagenesis was carried out using a QuikChange site-directed mutagenesis kit (Agilent Technology, CA) and corresponding primer pairs containing the mutations. After confirmation of the presence of the mutations, the P domain proteins were expressed, purified, and analyzed as described above. All mutant P domain proteins could be well detected by our in-house hyperimmune rabbit serum against huNoV VLPs.

Crystallization of native P domain proteins and complexes with corresponding HBGA oligosaccharides.

Protein samples were dialyzed with a solution of 20 mM HEPES (pH 7.5), 150 mM NaCl and concentrated to 15 mg/ml before native crystallization trials. Native P domain crystals were grown by the hanging-drop vapor diffusion method, with the crystallization droplet containing 1 µl protein and 1 µl reservoir solution and being hanged above 500 µl well solution. The reservoir solution for crystal growth of GII.17 1978 contained 15% (wt/vol) polyethylene glycol (PEG) 3350 and 0.2 M MgCl₂, while the solution for crystal growth of GII.17 2014/15 contained 11% (wt/vol) PEG 3350 and 4% (vol/vol) Tacsimate (pH 7.0). The reservoir solution for the crystal growth of GII.13 was composed of 0.1 M succinic acid (pH 7.0), 15% (wt/vol) PEG 3350. Crystals of suitable size were grown at 16°C and harvested after approximately 1 week. The oligosaccharides representing A, B, H1, Le^a, Le^b, Le^x, and Le^y (J&K, China) were dissolved in double-distilled water, prepared as 20 mM solutions, and then mixed with an equal volume of native P domain protein (20 mg/ml) and incubated at 4°C for 1 h, before being crystallized under native crystal growth conditions. A microseeding technique was used to aid the growth of complex crystals with native crystal seeds 16 h after the setting of the crystallization droplets. The P domain-HBGA complex crystals were harvested within 5 days. Specifically, to grow the crystals of GII.13 complexed with the Le^a trisaccharide, all solutions and reagents involved inclusion of the PreScission protease for removal of the GST tag, as they should be free of glycerol.

Data collection and processing, structure determination, and refinement.

The cryoprotectants for crystals of unliganded P domains and P domain-HBGA complexes were the corresponding reservoir solutions complemented with 15% (vol/vol) PEG 400. Crystals were briefly soaked in the cryoprotectant for 2 s before being mounted for the diffraction test. The diffraction data for native and complex crystals of GII.17 were collected at beamline 17U1 of the Shanghai Synchrotron Radiation Facility (SSRF; Shanghai, China) (50, 51) at a wavelength of 1.0000 Å, while the native and Le^a complex data for GII.13 P domains were collected at the rotating-anode X-ray source MicroMax-007/Satun 944 HG/Varimax HF (Institute of Biophysics, Chinese Academy of Sciences [CAS], Beijing, China) at a wavelength of 1.5418 Å. Diffraction data were processed, scaled, and merged using the HKL-2000 program package (52). Data collection statistics are summarized in Table 1.

TABLE 1.

Data collection statistics

Parameter	Value(s) fora
	Native GII.17 1978	Native GII.17 2014/15	GII.17 2014/15-A tri	GII.17 2014/15-B tri	Native GII.13	GII.13-Lea tri
Space group	C2221	P212121	P212121	P212121	P212121	P212121
Wavelength (Å)	1.0000	1.0000	1.0000	1.0000	1.5418	1.5418
Resolution (Å)	50–2.00 (2.03–2.00)	50–2.00 (2.03–2.00)	50–2.10 (2.14–2.10)	50–1.95 (1.98–1.95)	50–1.80 (1.83–1.80)	50–1.80 (1.83–1.80)
Cell dimensions (Å)
a, b, c (Å)	72.6, 101.6, 82.6	74.6, 87.2, 96.0	74.8, 86.7, 95.8	75.7, 86.9, 97.4	66.7, 82.2, 115.9	67.2, 82.6, 116.1
α, β, γ (°)	90, 90, 90	90, 90, 90	90, 90, 90	90, 90, 90	90, 90, 90	90, 90, 90
No. of measured reflections	115,597	230,368	194,771	487,382	364,333	371,559
No. of unique reflections	20,206	42,553	36,559	47,017	59,172	57,510
Completeness (%)	94.7 (96.9)	97.8 (87.2)	97.5 (82.4)	99.1 (99.7)	99.6 (99.3)	94.5 (63.2)
Redundancy	5.7 (5.1)	5.4 (3.0)	5.3 (4.4)	10.4 (7.1)	6.2 (6.6)	6.5 (4.3)
I/σ 〈I〉b	23.1 (3.4)	23.6 (3.0)	33.2 (14.6)	44.1 (12.7)	25.3 (5.4)	19.3 (3.5)
Rmergec	0.128 (0.512)	0.121 (0.397)	0.094 (0.191)	0.113 (0.414)	0.085 (0.560)	0.130 (0.474)

^aValues in parentheses correspond to the shell with the highest resolution.

^bI and σ (I), mean intensity and standard deviation of the intensity, respectively.

^c R_merge = Σi|Ii–〈I〉|/ΣiIi where Ii and 〈I〉 are the observed and mean intensity of related reflections with common indices h, k, and l, respectively.

The native crystal structures of GII.17 2014/15 and GII.17 1978 were solved by the molecular replacement method using the Phaser program of the CCP4 suite (53) and the GII.9 huNoV VA207 P domain structure (PDB accession number 3PUN; sequence identity, ∼53%) as the initial search model. Automatic structure building and refinement were carried out using the Phenix program (54), and manual adjustment was done using the program COOT (55) with the guidance of (2mF_o − DF_c) and (mF_o − DF_c) electron density maps. Water molecules were added at the final round of structure optimization at (mF_o − DF_c) electron density map peaks (>2.5 σ), where they can form stable hydrogen bonds with nearby amino acid residues. The phases and structures of the P domain-HBGA complexes were solved using the final refined structure of native P domain protein as a model. The structures of the native and Le^a complex of GII.13 were solved likewise, with the structure of the GII.21 OIF strain (PDB accession number 4RLZ; sequence identity, ∼73%) being the initial search model for molecular replacement. Structure refinement statistics are summarized in Table 2. The final structure validation was done with the PROCHECK program (56), with no residue being found at a disallowed region of the Ramachandran plot. Structural analysis was performed using EdPDB (57) and PyMOL (58).

TABLE 2.

Structure refinement statistics

Parameter	Value(s) for:
	Native GII.17 1978	Native GII.17 2014/15	GII.17 2014/15-A tri	GII.17 2014/15-B tri	Native GII.13	GII.13-Lea tri
No. of reflections in:
Working set	19,154	40,343	34,676	46,842	56,125	54,548
Test set	1,034	2,153	1,829	2,384	2,982	2,912
Rworka	0.191	0.194	0.184	0.205	0.164	0.154
Rfreeb	0.239	0.234	0.231	0.234	0.216	0.201
RMSDc
Bond lengths (Å)	0.007	0.008	0.008	0.007	0.006	0.006
Bond angles (°)	1.260	1.244	1.255	1.287	1.053	1.157
Average B factors (Å2)
Total	33.9	31.1	25.7	41.3	25.2	24.0
Protein	33.6	30.7	25.0	40.4	24.6	22.7
Ligand			48.4	59.5	30.8 (glycerol)	36.3
Solvent	37.4	34.7	28.7	49.1	31.6	32.3
Ramachandran plot (%)
Favored	96.1	98.0	98.0	98.0	97.4	97.1
Allowed	3.9	2.0	2.0	2.0	2.6	2.9
Disallowed	0	0	0	0	0	0
PDB accession no.	5ZUQ	5ZUS	5ZV5	5ZV7	5ZV9	5ZVC

^aR_work = ΣǁF_obs| − |F_calcǁ/Σ|F_obs|, where F_obs and F_calc are the observed and calculated structure factors, respectively.

^bR_free = ΣTǁF_obs| − |F_calcǁ/ΣT|F_obs|, where T is a randomly selected test data set (∼5%) of total reflections and was set aside before structure refinement, and F_obs and F_calc are the observed and calculated structure factors, respectively.

^cRMSD, root mean square deviation.

Accession number(s).

The coordinates and structure factors of the native GII.17 and GII.13 P proteins and the complexes with HBGA trisaccharides have been deposited in the Protein Data Bank (PDB) under accession numbers 5ZUQ, 5ZUS, 5ZV5, 5ZV7, 5ZV9, and 5ZVC (Table 2).