Histo-blood group antigens: a common niche for norovirus and rotavirus

NoVs and RVs and their common features

NoVs, members of the Caliciviridae family, are small (30–35 nm), round-structured virions that contain a single-stranded, positive-sense RNA genome. RVs, members of the Reoviridae family, are much larger (70 nm) with a more complicated capsid that contains a double-stranded, eleven-segmented RNA genome. NoVs and RVs are the two most important causes of acute gastroenteritis in children in both developing and developed countries. In addition, NoVs infect people of all ages and commonly cause epidemics of acute gastroenteritis in closed and semi-closed communities and institutions worldwide. Both viruses cause acute gastroenteritis with typical symptoms of diarrhoea, vomiting, nausea, stomach cramps and occasionally fever. RV infection can lead to life-threating dehydration in children in developing countries, whereas NoVs can cause severe diarrhoea in children and the elderly. Furthermore, both viruses can cause persistent infection in the immunocompromised and cancer patients (Refs ^{33, 34}).

Both human NoVs and human RVs interact with HBGAs through a surface protruding or spike protein (Refs ^{12, 28, 29, 30, 35, 36}). The NoV capsid is comprised 180 monomers or 90 dimers of the major structural protein, VP1, forming a T = 3 icosohedral viral particle (Ref. ³⁷). A minor structural protein, VP2, has been identified, but its precise role remains unclear. In vitro expression of VP1 and VP2 spontaneously assembles into virus-like particles (VLPs) with similar morphology and antigenicity to the authentic viruses (Ref. ³⁸), though later studies showed that expression of VP1 alone is sufficient to form VLPs. VLPs have been widely used as a model to study NoV–host interactions and immunology because NoVs remain uncultivable in vitro (Refs ^{39, 40, 41, 42}). VP1 is further divided into two major domains, a shell (S) and a protruding (P) domain, linked by a flexible hinge (Ref. ³⁷). The S domain forms the internal shell of the capsid, whereas the P domain constitutes the surface protrusions that are responsible for host interaction and immune recognition. The S and P domains can be structurally and functionally independent. Heterologous expression of the S domain assembles into S particles (Refs ^{18, 35}), whereas the P domain alone can form P dimers (Refs ^{12, 35}), 12mer small P particles (Ref. ⁴³) and 24mer P particles (Refs ^{44, 45, 46}). These P domain complexes retain HBGA-binding function and thus, have been widely used as models for studies of NoV–host interaction and immunology of NoVs (Refs ^{11, 12, 13, 14, 46, 47, 48, 49, 50, 51, 52}). In addition, a large amount of soluble P protein with a deletion of the arginine cluster at the C-terminus of the P domain was found in the stool of NoV-infected patients (Refs ^{53, 54, 55}), suggesting a possible function of the P protein in vivo.

The RV capsid has three layers formed by different structural proteins, VP1–VP8 (Refs ^{56, 57}). The outermost layer is formed by VP7 and VP4 (Ref. ⁵⁶). VP7 is a glycoprotein associated with the endoplasmic reticulum of RV-infected cells, whereas VP4 forms the major spike protein responsible for viral attachment and penetration into host cells (Refs ^{24, 27, 57, 58, 59}). During infection, VP4 is proteolytically cleaved into two subunits, VP5* and VP8* (Refs ^{60, 61, 62}). VP8* is the distal portion of the spike that is responsible for virus–receptor interaction to initiate viral attachment to host cells, whereas VP5* facilitates translocation of the double-shelled particles into the cytoplasm through conformation rearrangement and membrane fusion (Refs ^{56, 60, 63, 64}). Thus, VP8* may be equivalent to the NoV P domain as the surface protein for host interaction.

Another common property of NoVs and RVs is their genetic diversity. NoVs have five known genogroups (GI–GV), among which GI, GII and GIV cause human infection, whereas GIII and GV infect bovine and murine species, respectively. The two major human NoV genogroups (GI and GII) are further divided into over 30 genotypes (Ref. ⁶⁵), of which a single genotype, GII.4, causes over 80% of NoV epidemics worldwide. RVs also have ~35 P (protease sensitive) genotypes (P[1]–P[35]) constituting five genogroups (P[I]–P[V]) based on the VP8*/VP4 sequences (Ref. ³⁰) and at least 25 G (glycoprotein) genotypes based on the VP7 sequences (Refs ^{66, 67}). Although variable G/P types of RVs are found in humans and animals, a clear segregation of the P types between humans and animals is noted according to phylogenetic analysis of VP8*. P[I] and P[V] mainly infect animals, P[II] exclusively infects humans, whereas P[III] and P[IV] infect both humans and animals. Interestingly, two closely related genotypes, P[4] and P[8] in the P[II] genogroup, are responsible for over 90% of RV infection in humans, similar to the predominant GII.4 genotype of NoVs. Some common factors could play a role in controlling such host ranges and transmission of the two viruses.

Polymorphism of human HBGAs

HBGAs are complex carbohydrates distributed abundantly on mucosal epithelia of the respiratory, genitourinary and digestive tracts (Refs ^{68, 69}). They are also present as free oligosaccharides in biologic fluids, such as saliva, intestinal content, milk and blood. HBGAs are synthesised from various disaccharide precursors through sequential additions of monosaccharides with specific linkages catalysed by different glycosyltransferases that are coded by three major HBGA gene families. These three gene families are known as the secretor, Lewis and ABO families that code an α-1,2 fucosyltransferase (FUT2), an α-1,3 or α-1,4 fucosyltransferase (FUT3) and two glycosyltransferases (A and B enzymes), respectively. These enzymes catalyse the synthesis of the major human ABH and Lewis HBGAs (Fig. 1). Because there are several types of disaccharide precursors and each of the three gene families contains silent alleles that may lead to null phenotypes of the loci, the HBGA phenotypes of human population are polymorphic and complicated. For example, individuals with FUT2 inactivating mutations have a non-functional FUT2 and thus lose the ability to synthesise the H or secretor antigens. This phenotype is called non-secretor (Fig. 1), which accounts for ~20% of the European and North American populations. Similarly, individuals with a silent FUT3 gene have a non-functional FUT3 and therefore are unable to synthesise the Lewis antigens. This phenotype occurs in about 10% of the same European and North American populations. Inactivating mutations may also occur in one or both genes encoding the A and B enzymes, resulting in a wide variation in A and B antigens (Fig. 1) among different populations and races (Ref. ⁶⁸).

The genetic polymorphism of HBGAs determines the presence and absence of certain intermediates and final products of HBGAs in an individual, directly affecting his/her interaction with and thus, susceptibility or resistance to HBGA-recognising pathogens. For examples, persons with an active FUT2 are secretors having H-related HBGAs and they are sensitive to most human NoVs, RVs and H. pylori, because majority of these pathogens recognise secretor antigens. In contrast, individuals with an active FUT3 but an inactive FUT2 are non-secretors with Le^a and Le^x antigens and they are susceptible to NoVs, which recognise non-secretor antigens. In addition, two other monosaccharides, the A (α-GalNAc) and B (α-Gal) epitopes, components of the A and B antigens (Fig. 1) also contribute to interactions with NoVs and RVs. Thus, the HBGA–NoV/RV interactions are highly variable, which may further complicate the prevalence and epidemiology of these pathogens.

Diversity of NoVs and RVs in recognising human HBGAs

Complex interactions between the diverse NoVs and the polymorphic HBGAs have been demonstrated by many laboratories through either ELISA- or haemagglutination-based assays using recombinant VLPs or P particles as NoV surrogates and saliva, human milk, red blood cells or synthetic oligosaccharides as HBGA sources. A number of typical binding profiles have been described. For example, the prototype Norwalk virus (GI.1) recognises type A and O secretor antigens (Refs ^{8, 39, 40, 70}); VA387 (GII.4) interacts with all major ABO secretor antigens (Refs ^{39, 40}); a number of AB binding strains, such as MOH (GII.5) and Hiro (GII.12) bind to A- and/or B-but not O-secretors (Refs ^{15, 39, 40}); and VA207 (GII.9) binds both secretors and non-secretors (Refs ^{13, 39, 40}). The interaction of NoV with HBGAs has also been demonstrated by X-ray crystallography of several NoV P domains in complex with different HBGA oligosaccharides (see below) (Refs ^{11, 12, 13, 14, 15, 16, 17}). The significance of such interactions on host susceptibility and resistance has been demonstrated by human volunteer challenge studies for the prototype Norwalk virus (Refs ^{7, 8, 71}) and a GII.4 NoV (Refs ^{9, 72}). These data suggest that HBGAs may function as a receptor or at least a host susceptibility factor for NoVs.

However, discrepancies were also reported. For example, several human and mammalian cell lines that were known to express HBGAs did not support NoV replication (Ref. ⁷³). In addition, attempts to establish an association of HBGAs with susceptibility to NoVs in volunteer studies were failed for the Snow Maintain virus (GII.2) (Ref. ⁷⁴). These data suggest that the other host factors in addition to HBGAs may also important for NoV infection. Furthermore, NoVs are highly diverse in recognising different HBGAs but the HBGA-binding profiles for many NoV genotypes remain to be defined due to lack of data or inconsistent binding results obtained from different laboratories. Thus, efforts to define the HBGA-binding profiles, including searching for additional host factors, for these genotypes are needed.

Specific interactions with HBGAs were also observed in other calicivirus genera. As examples, the rabbit haemorrhagic disease virus (RHDV) in the Lagovirus genus recognises the H-type 2 and B antigens (Refs ^{75, 76}), whereas Tulane virus in the Recovirus genus, a recently discovered monkey calicivirus, recognises the type B antigens (Ref. ⁷⁷). Caliciviruses recognising other HBGA-related carbohydrate ligands were also found. For instance, murine NoVs (GV) and feline caliciviruses (FCV) in the Vesivirus genus recognise sialic acids (Refs ^{78, 79}), and the Newbury virus (GIII NoV) recognises an α-Gal (Ref. ⁸⁰). Thus, interaction with a carbohydrate receptor or ligand seems to be a common feature for caliciviruses, although caliciviruses recognising non-carbohydrate receptors such as fJAM-A for FCV have also been reported (Ref. ⁸¹).

Diverse interactions between HBGAs/carbohydrates and RVs have also been described. Recombinant VP8* of RVs of P[8], P[4] and P[6] genotypes of the P[II] genogroup recognise the secretor HBGAs. P[8] and P[4] are closely related genetically and both genotypes bind the Le^b and H-type 1 antigens (Ref. ²⁹). P[6], a slightly further related genotype binds the H-type 1 antigen only (Ref. ²⁹). These binding specificities have been confirmed by haemagglutination of red blood cells, blocking by monoclonal antibodies (mAbs), and binding of the authentic virions. In addition, P[9], P[14] and P[25] genotypes of the P[III] genogroup that infect humans bound specifically to the type A antigens (Refs ^{28, 30}), whereas P[11] of the P[IV] genogroup that infects infants interacted with single and repeated LacNAc, the type 2 precursor glycan (Refs ^{52, 82}). Direct evidence of RV–HBGA interaction has been shown by X-ray crystallography of a P[14] VP8* in complex with the type A oligosaccharide (Ref. ²⁸).

In summary, the interaction with variable HBGAs is a common feature of human NoVs and RVs. Whereas HBGAs have been demonstrated as host susceptibility/resistance factors for NoVs, direct evidence for such a role for RVs remains lacking. However, based on the observed similarity between the two viruses in their interactions with HBGAs, it may be possible to assume that HBGAs could also be host factors in RV infection. The fact that all three major genotypes (P[8], P[4] and P[6]) of human RVs recognise human HBGAs supports this hypothesis. Furthermore, the observed binding diversity of human NoVs and RVs to variable ABO, secretor and Lewis types suggests a similar adaptation/selection relationship of both viruses with the human host.

Structural basis of NoV and RV interactions with HBGAs

The crystal structures of the HBGA-binding interfaces of seven NoVs (two GI and five GII) representing different ABO, secretor and Lewis binding patterns have been determined by X-ray crystallography of recombinant NoV P domain dimers in complex with corresponding HBGA oligosaccharides (Refs ^{11, 12, 13, 14, 15, 16, 17}). All seven NoVs revealed two symmetric HBGA-binding interfaces on the top of each P dimer at or near the common boundary of the two P monomers (Fig. 2a). The binding interfaces share a similar overall topology, including a central binding pocket (CBP) (Fig. 2, cyan) that plays a central role in the NoV–HBGA interaction and a surrounding region that supports or stabilises the CBP interaction. The CBP is formed by four (GI) or five (GII) highly conserved amino acids within each genogroup and interacts with a common genogroup-specific saccharide of HBGAs (see below) as the major binding saccharide (MaBS). In contrast, the surrounding region is less conserved and interacts with one or two additional saccharides as minor binding saccharides (MiBSs) that are adjacent to the MaBS and are variable among strains within each genogroup.

The CBPs of the five GII NoVs fit well with an α-Fuc, among which four [two GII.4 (VA387 and TCH05), one GII.10 (Vietnam026) and one GII.12 (Hiro)] bind the α-1,2 Fuc (H epitope) (Fig. 2c), whereas the other (VA207, GII.9) binds the α-1,3 Fuc (Lewis epitope), as the MaBS (Refs ^{12, 13, 15, 16}). Accordingly, the surrounding regions interact with the A (α-GalNAc), B (α-Gal) or Lewis (α-1,4 Fuc) epitopes or the β-GlcNAc of the precursor as MiBSs (Fig. 2c). Thus, through the α-Fuc-binding mode, GII NoVs are able to bind all known ABO, secretor and Lewis antigens. This structural information also helps to explain the saliva-binding results for GII NoVs (Fig. 3). For example, the previously defined ABH binders of the H binding group (Ref. ⁸³), such as VA387 and TCH05 (both GII.4) (Refs ^{12, 16}) and the recently reported strain Vietnam026 (GII.10) (Ref. ¹⁵) use the H epitope as MaBS and the A or B epitopes as MiBSs in binding to the A- and/or B-secretors; whereas they use the same H epitope as MaBS and the Lewis epitope as MiBS in binding to the O-secretors (Fig. 3). NoVs in the A/B binding group, such as strains Hiro (GII.12) (Ref. ¹⁵) and MOH (GII.5) (Refs ^{39, 40}), also bind the H epitope as MaBS and the A and/or B epitopes as MiBSs (Ref. ¹⁵) (Fig. 3), but a combination of the H epitope as MaBS and the Lewis epitope as MiBS did not seem to occur because these A/B binders do not bind the type O secretors in in vitro binding assays (Refs ^{39, 40}). NoVs of the Lewis binding group, such as VA207 (GII.9), use the Lewis epitope as MaBS and the β-GlcNAc of the precursor as MiBS to bind both secretors and non-secretors (13) (Fig. 3).

The HBGA-binding interfaces of GI NoVs are distinct from those of GII NoVs in structure, amino acid composition and location (Fig. 2). Unlike GII NoVs, the CBPs of GI NoVs (GI.1 Norwalk virus and GI.2 FUV258) bind the β-Gal of the precursor as MaBS, whereas their surrounding regions bind the α-1,2 Fuc (H epitope) and/or the α-1,3/4 Fuc (Lewis epitope) as MiBS to support the CBP–HBGA interaction (Fig. 2b) (Refs ^{14, 17}). Thus, through the defined Gal-binding mode, GI NoVs bind both secretors (via a secretor epitope as MiBS) and non-secretor (via a Lewis epitope as MiBS) antigens. GI NoVs bind the A antigen in a unique manner, in which the CBP binds the α-Gal of the A epitope (α-GalNAc) as MaBS, whereas the acetamido group of the A epitope interacting with the surrounding region to provides additional binding support (Fig. 3). The type B antigen has an α-Gal but lacks an acetamido group, explaining why Norwalk virus (GI.1) and FUV258 (GI.2) do not bind the B antigen. FUV258 binds the Le^a oligosaccharide through the β-Gal as the MaBS and α-1,4 Fuc (Lewis epitope) as the MiBS (Ref. ¹⁷). Saliva-binding assay revealed that FUV258 bound incomplete non-secretors (Le^a+b+) (Ref. ⁸⁴), suggesting that the β-Gal-binding mode with the α-1,2 Fuc (H epitope) as the MiBS may be used, but its interactions with saliva from true non-secretors (Le^a+b−) remains unknown (Fig. 3). In summary, although the CBP plays a critical role in NoV–HBGA interaction, the variable surrounding regions provided flexibility to interact with diverse saccharides of human HBGAs. This type of relative conservation and variation is the structural basis of NoVs in interaction with the polymorphic HBGAs.

The structure of the A antigen-binding site of the spike protein VP8* for a P[14] RV has also been elucidated by X-ray crystallography (Ref. ²⁸). Interestingly, this A antigen-binding site on the top of the VP8* shares the same location with the sialic acid-binding site of an animal RV (Refs ^{85, 86}). Both the α-GalNAc (A epitope) and β-Gal of the precursor, but not the α-1,2 Fuc (H epitope), contributed equally to the binding. Since this is the first structure of RV–HBGA interaction, it is not clear at this time which saccharide is the MaBS. If the A epitope is the MaBS, it would explain why this P[14] RV binds only to the A antigen. However, the fact that the majority of human RVs ([P]8, [P]4 and [P]6) in the P[II] genogroup bind the H and Le^b antigens (Fig. 4) appears to fit better with the β-Gal as the MaBS. Thus, determination of the crystal structures of the HBGA-binding interfaces of these major human RVs interacting with the H and Le^b antigens is essential.

Structural data also showed that the CBP is highly rigid, whereas the surrounding regions are more flexible. For example, a single amino acid mutation at the CBP of NoVs usually destroys the binding activity, whereas a single mutation in the surrounding region may cause changes of binding specificity (Refs ^{13, 17, 36, 42, 48, 49, 88}). However, this type of flexibility was not observed among 15 VP8* mutants of a P[9] RV, each with a single amino acid mutation in or near the binding interface. Most of the mutations caused complete loss of binding activity, but none altered HBGA-binding patterns of the mutants (Ref. ³⁰). This type of rigidness may explain why P[III] RVs (P[9], P[14] and P[25]) recognise the type A antigen only in the saliva-based binding assay (Ref. ³⁰). This difference between NoVs and RVs may impact the clinical infection and epidemiology, although this needs to be clarified in future studies.

HBGAs in the evolution of NoVs and RVs

Consistent with structural and functional importance, the amino acid sequences of the CBPs are highly conserved within each of the GI and GII NoV genogroups (Refs ^{4, 49}), indicating strong selection of NoVs by human HBGAs. Although the original host receptors of NoV ancestors are unknown, upon their introduction into humans, GI and GII NoVs must have started with a binding interface that bound one or a set of human HBGA saccharides. GI apparently selected the precursor β-Gal as the MaBS and maintained their initial CBP but diverged significantly into different sub-lineages (genotypes) (Fig. 4a) likely by alternating the surrounding regions to interact with different adjacent saccharides as the MiBSs. Similarly, the original GII NoVs may have started with a different CBP formed by a distinct set of amino acids interacting with an α-Fuc as the MaBS and diverged into different sub-lineages (genotypes) with completely different combinations of MaBS–MiBS. Thus, GI and GII represent two distinct genetic lineages that diverged independently. In fact, GI and GII NoVs are separated phylogenetically by three animal NoVs (GIII, GIV and GV) (Fig. 4a), where GI is closer to bovine NoVs, whereas GII is closer to murine and porcine NoVs.

The elucidation of the conserved CBPs and the variable surrounding regions of the HBGA interfaces help our understanding of the evolutionary paths of GI and GII NoVs. Due to the essential role of the CBPs in the NoV–HBGA interaction, each of GI and GII developed their unique paths into different sub-lineages with different binding patterns to the variable human HBGAs. For example, as described above, GI must have followed a path that uses the β-Gal as the MaBS and the H and Lewis epitopes as the MiBSs. This explains why some GI NoVs bind both secretor and non-secretor HBGAs. GII NoVs that bind either or both secretors and non-secretors have also been found. However, the mechanism of diversity for GII differs from that of GI, in which GII can switch the MaBSs between the H (α-1,2 Fuc) and Lewis (α-1,3/4 Fuc) fucoses among different genotypes. This ability to alter the MaBS for GII NoVs could explain why GII (~20 genotypes) is much more diverse than GI (~8 genotypes). In addition, the total number of accessible MiBSs adjacent to the two fucoses (MaBSs) is greater for GII (A, B, H and/or Lewis residues) than GI (H and Lewis residues). Nevertheless, the switch from the precursor β-Gal to the A antigen α-Gal of GI NoVs described above results in extension to type A secretors only.

The role of human HBGAs in the evolution of human RVs remains unclear due to limited structural information. The fact that some animal RVs recognise the sialic acid, whereas major human RVs recognise human HBGAs strongly suggests that human HBGAs may play an important role in the host range and evolution of RVs. Genetic and structural analyses also showed that the HBGA-binding sites of human RVs share the sialic acid-binding sites of animal RVs, suggesting a closer genetic relatedness among the RVs than that between the two human NoV genogroups (GI and GII). Thus, the recognition of a carbohydrate receptor could also be a common feature for RVs. Although animal RVs recognise the specific sialic acid containing carbohydrates to infect different animal species, human RVs infect humans likely by recognition of specific human HBGAs through a conserved carbohydrate-binding interface. Future studies to determine the receptors or host ligands for all RVs, particularly for animal RVs infecting different animal species, are important for a full understanding of the evolution of RVs.

Role of HBGAs in host range of NoVs and RVs

The clear segregations between most genogroups and their host species with specific carbohydrate ligands suggest that these carbohydrates play an important role in the host range of both NoVs and RVs. NoV and RV recognition of particular carbohydrates that are expressed exclusively in animal species but not in humans has been demonstrated. For example, unique forms of sialic acid, Neu5Gc, or N-glycolylneuraminic acid (Ref. ²³) recognised by some animal RVs are found in primates and some other animal species, but not in humans (Ref. ⁸⁹). In addition, bovine NoVs recognise the α-Gal epitope with a special linkage (Galα3Galβ4GlcNAc) that is expressed in most mammalian species but not by hominids and humans (Ref. ⁹⁰). The unique expression and distribution patterns of the specific sialic acids may explain why animal NoVs, such as murine NoV (Ref. ⁷⁸) and feline calicivirus (Ref. ⁷⁹), infect-specific animal species but not humans. Furthermore, the specific requirements of ABO, secretor and Lewis HBGAs for human NoVs have been demonstrated through volunteer challenge studies for both GI and GII human NoVs (Refs ^{7, 8, 9, 71}).

On the other hand, shared carbohydrate ligands and receptors between different host species may be responsible for cross-species transmission of RVs. P[9], P[14] and P[25] RVs in genogroup P[III] infect humans, as well as some domestic and wild animal species, and these RVs apparently recognise the A antigens (Ref. ³⁰). The A antigens are presented in many animal species, including domestic cows and pigs (Refs ^{89, 91, 92, 93}). In vitro binding assays showed that the VP8* proteins from the three P[III] RV genotypes bound the A antigens of gastric mucins of cows and pigs in a concentration dependent manner (Ref. ³⁰), implying that the A antigens could be a shared factor for cross-species transmission. In recent studies, the P[11] RVs in genogroup P[IV] that infect neonates of humans and animals have been demonstrated to recognise the HBGA precursor poly-LacNAc, which could be another cross-species factor of RVs (Refs ^{52, 82}), although other unknown factors may also play a role.

The ligand-based cross-species transmission raised a question about host species barriers for the majority of genetic types of NoVs and RVs. For example, both GI and GII NoVs and the major human RV genotypes P[8], P[4] and P[6] in the P[II] genogroup are rarely found in animals. One explanation is that these NoV and RV genotypes recognise the α-Fuc or precursor β-Gal of human HBGAs as the MaBSs, whereas the A antigen-binding P[III] RVs (P[9], P[14] and P[25]) may recognise the A epitope as the MaBS. These distinct modes may contribute to the species barrier. However, the H antigens have also been found in some animal species, which does not fully support the ligand-based host species barrier hypothesis. Thus, future studies to explore this question are needed, particularly on the role of additional unknown host factors in host barriers.

Role of HBGAs in epidemiology and classification of NoVs and RVs

The elucidation of different evolutionary paths of GI and GII NoVs based on the MaBSs selected by each genogroup also explains the epidemiology of NoVs. As described above, GII NoVs that use the α-Fuc (H or Lewis epitope) as a MaBS bind all secretor and non-secretor antigens through support by variable MiBSs. This may explain why the GII NoVs are more common than GI NoVs everywhere in different populations, although other unknown factors may also play a role in these prevalence differences. In addition, the H-binding feature of many GII NoVs may further explain the predominance of GII NoVs because of the wide distribution of secretor types (80%) over the non-secretor types (20%) in the general population. The H-binding feature may also explain the predominance of GII.4 NoVs, in which most GII.4 variants maintained the ability to bind salivas of all A, B and O secretors, whereas the most other genotypes bind either A and/or B but not O secretors (Refs ^{3, 39, 40, 94, 95, 96}). Thus, the potential target populations and prevalence of individual genotypes of human NoVs are likely predictable based on their HBGA-binding profiles and the distributions of different ABO, secretor and Lewis types in the world populations.

A definition of the genetic relationships among different genotypes within each genogroup may also facilitate the diagnosis and strategy in control and prevention of NoV diseases. For example, the conserved CBPs within each genogroup would share certain common antigenic types that may be important targets for diagnosis and vaccine development for broad detection/protection against NoVs. In addition, the conserved CBPs of the HBGA-binding interfaces may also provide an attractive target for antiviral design against many strains within genogroups. Future studies to define such antigenic similarities among different genotypes by determining their antigenic cross-reactivity based on binding and blocking (neutralisation against binding to HBGAs) assays within each genogroup are necessary. Understanding such antigenic relationships would also help to select a cocktail vaccine against the major types of NoVs. On the other hand, the distinct HBGA-binding interfaces between the two genogroups indicate a necessity to include reagents for both genogroups for broad detection in diagnosis and broad protection in vaccine strategy. Finally, since human HBGAs are an important factor in NoV host range and evolution, a phenotypic classification of NoVs according to their interaction patterns with different ABO, secretor and Lewis types will be a significant addition to the current genetic classification. Great progress has been made in determination of diverse binding patterns of many NoV genotypes in the past decade, through further efforts to define the HBGA binding of the remaining NoV genotypes for a complete phenotypic classification are necessary.

The role of HBGAs in the epidemiology of RVs remains unclear due to the limited structural data of RV–HBGA interactions and the lack of direct evidence for the role of HBGAs in susceptibility to RV infection. As aforementioned, the RV-binding interfaces may be highly rigid based on a mutagenesis study (Ref. ³⁰), in which either a β-Gal or GalNAc may function as a MaBS (Ref. ²⁸). Thus, the evolutionary paths and mechanism of diversity of RVs may differ from that of NoVs. A rigid binding interface would theoretically have a higher affinity and specificity to HBGAs than a flexible interface, and therefore could affect the persistence and transmission of RVs. However, a rigid binding interface could also prevent RVs from easily adapting to new hosts with variable HBGAs. These speculations need to be validated by future surveillance studies of RV epidemiology in relationship with host HBGA phenotypes.

Role of HBGAs in epochal evolution of GII.4 NoVs

The genetic and structural conservation of the CBPs and variation of the surrounding regions of the NoV HBGA-binding interfaces may also help our understanding of the epochal evolution of GII.4 NoVs. The highly conserved CBP indicates a strong selection pressure by the human HBGAs, which is essential for NoV viability. In other words, immune escape viral variants driven by the host immunity with mutations in the CBP are negatively selected by the human HBGAs and therefore, will not survive. On the other hand, amino acids in the surrounding regions of the CBPs are variable and are important determinants of the binding spectrum of GII.4 NoVs to various HBGAs through their interactions with the A, B and Lewis epitopes, respectively, as MiBSs. This type of flexibility could generate sequence spaces for antigenic changes driven by the host immunity. GII.4 variants with new HBGA-binding profiles to variable A, B, and/or O secretors may help them spread in human populations with matched HBGA types. Thus, the conservation of the CBPs and the variations of the surrounding region, as a result of the interplay between host immunity and HBGAs, may be a key to explain the epochal evolution of GII.4. In vitro binding assays have demonstrated that the major GII.4 variants that have circulated in the past decades retained the ability to bind to the major secretor types with varied binding affinities, in which variants with cyclic high versus low binding affinities to the A antigens were observed (Ref. ⁹⁷). In addition, cross-reactive immune responses among variants have also been detected (Refs ^{97, 98, 99}). Future studies to define the dynamic balance between the conservation selected by HBGAs and the variations driven by host immunity are needed.

An important issue related to GII.4 epochal evolution is how to select vaccine strains for future disease control and prevention. The highly conserved CBPs of the HBGA-binding interfaces and the cross-reactive antibodies observed among different GII.4 variants suggested potential shared antigenic epitopes for cross-protection. These data imply that a frequent selection of pre-epidemic strains for vaccine development may not be necessary. Thus, further study to verify such cross-protective immunity among GII.4 variants is needed. Due to the lack of effective cell culture and small animal models, the HBGA-blocking assays currently serve as a surrogate neutralisation method for NoVs. Production and mapping of mAbs that recognise conserved antigenic epitopes and exhibit broad cross-blocking activity on HBGA–NoV interactions for various GII.4 variants are highly significant. In fact, one such mAb has been identified recently (Refs ^{98, 100}), but more, particularly those against the conserved CBP, are required. In addition, paired acute and convalescent sera from patients infected with GII.4 NoVs are also valuable for determining cross-reactivity and cross-protection among different GII.4 variants. A human volunteer challenge model of a GII.4 variant (Ref. ⁹) has been developed recently, which would facilitate such studies. Since the CBPs are also conserved among other GII NoVs, identification of mAbs with cross-reactivity and cross-protection for all GII NoVs is highly significant.

HBGAs as a potential common niche for other pathogens

In summary, NoVs are under selection by human HBGAs through different evolutionary pathways. Although both GI and GII NoVs originated from a common ancestor, they may have diverged into two distinct lineages before becoming human pathogens and then converge into human hosts selected by human HBGAs. Both genogroups further diverged into multiple sub-lineages (genotypes) as a result of adaptation to the polymorphic human HBGAs. The elucidation of this evolutionary relationship of NoVs within and between the two genogroups has helped our understanding of virus/host interactions, host ranges, diagnosis, epidemiology, classification and host immune selection of NoVs. Following in the footsteps of NoVs, major advancements in studying RV–HBGA interactions, including the discovery of diverse binding patterns of RVs to HBGAs, have been made. Since many other viral and bacterial pathogens also utilise the human HBGAs as receptors (Ref. ⁹⁴), questions are raised as to whether those pathogens also share similar features of diversity in their interactions with HBGAs like NoVs and RVs. Answers to those questions are fundamentally important because HBGAs may also play a critical role in host susceptibility and therefore, the transmission pattern, epidemiology and disease burden of these pathogens. For example, H. pylori were once well known for its recognition of the Le^b antigen. However, after a number of failures in attempting to associate secretor status with clinical infection of H. pylori were reported (Ref. ¹⁰¹), people realised that H. pylori is diverse in recognising different ABH antigens similar to NoVs and RVs.

Additionally, candida has been reported to recognise the Lewis antigens (Refs ^{102, 103, 104}), but an association between non-secretors and candida infection is still debated (Ref. ¹⁰⁵). Thus, further exploratory studies are needed to determine whether candida also shares similar diversity in recognising additional HBGAs. Furthermore, influenza viruses are known to recognise sialic acids with distinct linkages between human and animal viruses (Refs ^{106, 107}). Since the HA of influenza A viruses are highly diverse and new variants continually emerge and result in new epidemics, a question could be asked as to whether HBGAs or additional residues of the host gangliosides also participate in the interaction with the viruses and cause different epidemics. Thus, extended studies to explore whether candida, influenza viruses and other viral and bacterial pathogens that also utilise human HBGAs as receptors and are also diverse in recognising different HBGAs like NoVs and RVs may be necessary.

Research in progress and outstanding research questions

Our current understanding of NoV–HBGA interactions also raises important questions that should be answered in the near future. For example, the structural conservation of the CBPs of the HBGA-binding interfaces among genotypes within GI and GII NoVs may be a critical target for vaccine development. Thus, elucidation of the antigenic relationship through comparison of structure and antigenic similarity, as well as differences among major genotypes within each genogroup, is necessary. The determination of antigenic relatedness among different GII.4 variants is also important to understand epochal evolution and vaccine strategies against GII.4 NoVs. The newly developed human challenge model for a GII.4 NoV provides a valuable tool and reagents for this purpose. Furthermore, structural studies of rare strains and evolutionarily distant genotypes, as well as genetically similar animal NoVs, are needed for further understanding of the evolution of human NoVs and the species barriers between humans and animals, in addition to the potential risk of zoonotic transmission.

For RVs, our understanding of the evolution, epidemiology and classification in association with HBGAs remains limited; therefore, studies in all of these areas, continually following the footsteps of NoVs, are necessary. One urgent need is to solve the VP8* crystal structures of the major genotypes (P[8], P[4] and P[6]) of human RVs to determine the mechanisms of HBGA–RV interaction and evolution of human RVs selected by HBGAs. Equally important studies include an establishment of an association of RV infection with HBGA phenotypes of patients by clinical surveillance investigation. In addition, the HBGA/carbohydrate-binding patterns of many animal RVs should be determined for an overall picture of RV evolution related to HBGAs/carbohydrates. This will further help our understanding of the species barriers of RVs between humans and various animal species and estimation of potential zoonotic transmission.

Finally, additional studies of pathogen–HBGA interactions for other viral and bacterial pathogens need to be performed, which will provide further evidence to justify the evolutionary pathways of NoVs and RVs. Accumulation of data from different pathogens utilising human HBGA receptors will also help our understanding of the polymorphism of human HBGAs. The genetic relatedness of HBGAs between humans and different animal species, in regards to the pathogens utilising HBGAs, needs be explored as well, which requires interdisciplinary collaborations.