The capsid protein has a modular domain organization with an N-terminal arm (NTA) that is important for directing capsid assembly, followed by a shell (S) domain that is important for stabilizing the icosahedral scaffold (Bertolotti-Ciarlet et al., 2002), and a protruding (P) domain emanating from the icosahedral shell that is further divided into P1 and P2 subdomains (Figs. 3.1.1D–F). The S and P domains are linked by a flexible hinge. The P1 subdomain is formed by two noncontiguous segments within the P domain, whereas the P2 subdomain facing the exterior is formed by the intervening segment. The polypeptide fold in each of these domains is also essentially conserved among calicivirus structures. The S domain exhibits an 8-stranded antiparallel β-barrel motif that is typically observed in T=3 icosahedral viruses (Prasad and Schmid, 2012). The fold of the P1 subdomain, consisting of three β-strands in the N-terminal portion, a twisted antiparallel β-sheet formed by four strands in the C-terminal portion, and a well-defined α-helix, is novel and only seen in calicivirus structures (Fig. 3.1.2A). The fold of the P2 subdomain is a β-barrel of six antiparallel strands connected by loops of various lengths. Despite the similar structural characteristics among the members of Caliciviridae, there are significant variations within the capsid protein structure providing insight into how the unique modular organization of the capsid protein is conducive to the wide diversity and host specificity associated with this family of viruses. Comparisons of the calicivirus capsid protein sequences indicate that the S domain is highly conserved and the P1 subdomain is moderately conserved, whereas the distally located P2 subdomain is highly variable.

In the T=3 icosahedral lattice, the capsid protein is located in three quasi-equivalent positions, conventionally designated A, B, and C, which constitute the icosahedral asymmetric unit (Fig. 3.1.1C). The A subunits surround the icosahedral fivefold axis, whereas B and C subunits alternate around the icosahedral threefold axis giving rise to quasi sixfold symmetry. In the calicivirus capsid, as in other T=3 icosahedral capsids, A, B, and C subunits form two types of quasi-equivalent dimers, A/B dimer related by the quasi twofold axis (A/B₂) and C/C dimer related by the strict icosahedral twofold axis (C/C₂) (Fig. 3.1.2C). A common feature is that the A/B dimer has a “bent” conformation, whereas the C/C dimer has a “flat” conformation. Such a dual conformation imparts the necessary curvature for the formation of the T=3 icosahedral capsid. In many of the structurally characterized T=3 capsids, the NTA is implicated in providing a switch by undergoing an order to disorder transition to facilitate the bent A/B and the flat C/C conformations during T=3 capsid assembly (Harrison, 2001; Rossmann and Johnson, 1989). In these T=3 virus capsids, including rNV and rHOV, only one of the three NTAs of the quasi-equivalent subunits is ordered. In the case of rNoV structures, while the NTA of the B subunit is ordered to a larger extent, the equivalent regions in the A and C subunits are disordered (Fig. 3.1.1F). The ordered NTA portion of the B subunit interacts with the base of the S domain of the neighboring C subunit to stabilize the flat conformation of the C/C dimer (Fig. 3.1.1F). The equivalent interactions involving the NTA are not observed in the A/B dimer which adopts a bent conformation (Prasad et al., 1999). In contrast, although serving the same purpose, an ordered NTA of the C subunit provides a switch in the T=3 plant tombus- and sobemoviruses instead of the NTA of the B subunit as observed in rNoV capsids (Harrison, 2001; Rossmann and Johnson, 1989). In nodaviruses, which also exhibits a T=3 capsid organization, in addition to an ordered NTA of the C subunit, a piece of genomic RNA keeps the “flat” conformation of the C/C dimers (Fisher and Johnson, 1993). Interestingly, the SMSV and FCV capsid structures exhibit a novel and distinct variation from any of these viruses. In these structures, the NTAs of all three subunits are equally ordered, essentially maintaining the T=3 symmetry at this level. Instead of an order-to-disorder transition of the NTA, a distinct conformational change involving a Pro residue in the B subunit that leads to the formation of a ring-like structure around the fivefold axis appears to provide a switch (Chen et al., 2006; Ossiboff et al., 2010). Whether these unique NTA interactions found only in SMSV and FCV are influenced by the genome or the proteolytic processing of the capsid protein, a common feature in the members of Vesivirus genus, remains a question.

Like capsid proteins in other T=3 icosahedral viruses with distinct S and P domains, such as plant tombusviruses (Harrison, 2001), calicivirus VP1 has a flexible hinge between these two domains (Fig. 3.1.1D, denoted by h and an arrow). This hinge, which facilitates the interactions between the P1 subdomain and the upper portion of the S domain, is likely important for locking the A/B and C/C dimers in their appropriate conformations, as this interaction is seen only in the A/B dimers and not in the C/C dimers (Prasad et al., 1999; Chen et al., 2006; Ossiboff et al., 2010). Although such an interaction between the P1 and S domains is conserved in rNoV, SMSV, and FCV crystal structures as a structural requirement, the relative orientations between these two domains, likely because of the sequence changes, are noticeably altered. In the animal calicivirus structures including SMSV, FCV, and recently determined high resolution cryo-EM structure of rabbit hemorrhagic disease virus (RHDV) (Wang et al., 2013), this change in S–P1 orientation together with a compensatory change in the P1–P2 orientation causes only the distal P2 subdomain to participate in the dimeric interactions. This is in contrast to that observed in rNoV structures, in which both P1 and P2 subdomains participate in the dimeric interactions. A conserved glycine located at the junction of P1 and P2 is suggested to allow this compensatory change in the P1–P2 orientation. Thus, in addition to the flexibility between the S and P1 domains, there is an additional point of flexibility between the P1 and P2 subdomains that was not immediately apparent from the rNoV structures alone. The multiple points of flexibility could be an important factor in enhancing calicivirus diversity through structural variations within the context of a similar domain organization, somewhat akin to the interdomain flexibility seen in antibody structures with a hinge and an elbow.

Comparison of the calicivirus sequences clearly indicates that the region that forms the P2 subdomain exhibits the most variability, consistent with its role in host specificity, antigenicity and receptor interactions. Accordingly, comparison of the animal calicivirus and norovirus capsid and P domain structures show that the P2 subdomain exhibits the most structural variability (Fig. 3.1.2B–D). Despite significant variation in the sequences, the basic polypeptide fold in the P2 subdomain, with six antiparallel β-strands forming a β-barrel structure, is conserved in all the calicivirus capsid structures so far determined (Fig. 3.1.2A). The main differences are in the orientations and the extension of the surface-exposed loops that connect these β-strands. These loops extend sideward from the dimeric surface with the conserved β-barrel structure at the dimeric interface. With shorter loops, the NV P2 subdomain exhibits the most compact structure, whereas the P2 subdomains of SMSV and FCV exhibit the most elaborate loop structures. The loop regions in the P2 subdomain of GII.4 are also more elaborate compared to GI.1 NV, exemplifying the intergenogroup differences in NoVs that play a role in differential glycan recognition as described in the next section. FCV is the only calicivirus for which a protein receptor has been identified (Makino et al., 2006). Cryo-EM structure of the FCV–fJAM-1 complex clearly shows that the receptor fJAM-1 (feline junctional adhesion molecule 1) binds to the P2 subdomain (Bhella et al., 2008). Known neutralization epitopes in the FCV capsid map to the loops in the P2 subdomain (Chen et al., 2006). Thus, crystallographic analyses of caliciviruses of different genera and genotypes demonstrate how the P2 subdomain in these viruses accommodates significant sequence alterations within the context of a conserved polypeptide fold to achieve antigenic diversity and strain-dependent receptor recognition.

Based on the observation that dimeric interactions of the capsid protein particularly involving the P domain are a common feature in all the caliciviruses, a reasonable assumption is that the VP1 dimer is the building block for the assembly. The VP1 dimer may exist in a dynamic equilibrium between “bent” and “flat” conformations in solution prior to assembly, and during the assembly process may switch to appropriate conformations directed by the NTA arms. The role of the NTA arms and also that of dimeric interactions in the capsid assembly is substantiated by a systematic structure-directed mutational analysis of the NV VP1 (Bertolotti-Ciarlet et al., 2002). Based on the calicivirus capsid structures (Chen et al., 2006) and mass spectrometric analysis of rNV capsid and its pH-induced dissociation/association intermediates (Baclayon et al., 2011; Shoemaker et al., 2010), it is plausible that capsid assembly proceeds through the formation of trimers of dimers that are then brought together into an icosahedral structure.

In the context of the capsid assembly and perhaps more particularly in the genome encapsidation, another important factor to be considered is the minor protein VP2. All calicivirus genomes encode this protein (Green et al., 2000), which is highly basic in nature. The association of VP2 with the infectious virus particles has been demonstrated in NV (Glass et al., 2000) as well as FCV (Sosnovtsev and Green, 2000) and is suspected to be present in all caliciviruses. The role of VP2, particularly in enhancing the capsid stability and size homogeneity, is clearly evident from the dynamic light scattering experiments of the NoV VLPs produced by the expression of VP1 alone and in comparison with those obtained from the coexpression of both VP1 and VP2 (Bertolotti-Ciarlet et al., 2003). Studies on FCV also suggest VP2 stabilizes the icosahedral capsid and is required for the production of infectious virus (Sosnovtsev et al., 2005). However, VP2 is not visualized in any of the calicivirus capsid structures, including the crystal structures of the rNoV particles produced by the coexpression of VP1 and VP2. This is likely because of the substoichiometric proportion of VP2 (suspected to be ∼2–8 molecules per virion) which does not allow VP2 to interact with VP1 conforming to the icosahedral symmetry and causes it to be transparent in the structures in which the icosahedral symmetry is explicitly used for structure determination.

HBGAs are glycans that include the determinants of secretor-status and blood type of an individual (Ruvoen-Clouet et al., 2013). They are synthesized by the sequential addition of a monosaccharide to a terminal precursor disaccharide by genetically controlled expression of certain glycosyl-transferases. Depending upon the composition of the disaccharide backbone and the linkage, they are grouped in the four types: type 1 (Galβ1-3GlcNAcβ), type 2 (Galβ1-4GlcNAcβ), type 3 (Galβ1-3GalNAcα), and type 4 (Galβ1-3GalNAcβ). Although several studies have suggested that the secretor-positive status is a susceptibility factor for the majority of NoVs (Lindesmith et al., 2003; Hutson et al., 2005), recent epidemiological studies indicate that nonsecretors are susceptible to some GI and GII NoVs (Currier et al., 2015). The secretor-positive status of an individual is determined by expression of a functional FUT2 enzyme, a fucosyl-transferase, that catalyzes the addition of an α fucose (SeFuc) to the β galactose (β Gal) of the disaccharide precursor to form the secretor epitope or the H-type HBGA. The H-type HBGA can be further modified by enzymes A or B by adding N-acetyl galactosamine (GalNAc) or Gal to the precursor β Gal to form A- or B-type HBGA, respectively (Ruvoen-Clouet et al., 2013). Similarly, the Lewis-positive status is determined by the activity of the fucosyl transferase 3 (FUT3) enzyme, which adds an α fucose (LeFuc) to the N-acetylglucosamine (GlcNAc) of the precursor disaccharide to form the Lewis epitope. Thus the sequential addition of carbohydrate moieties by the FUT2 and FUT3 along with enzymes A and B give rise to the secretor/nonsecretor Lewis and ABH families of HBGAs (Imbert-Marcille et al., 2014; Marionneau et al., 2005). Several studies using NoV VLPs with saliva, red blood cells, and synthetic carbohydrates demonstrate direct interaction between NoV VLPs and HBGAs (Tan and Jiang, 2005; Hutson et al., 2003; Huang et al., 2005; Shirato et al., 2008) and show that the specificity to HBGAs varies not only within a particular genogroup but also between the genogroups.

A typical strategy that is used to understand the structural basis of the HBGA interactions in NoVs is to determine the X-ray structure of the recombinantly expressed P domain of the NoV in complex with the HBGA (Cao et al., 2007; Choi et al., 2008; Tan et al., 2004). Following this strategy, in recent years there has been an explosion of crystallographic structures of the P domain of various NoVs in complex with a variety of HBGAs (Cao et al., 2007; Choi et al., 2008; Hansman et al., 2011; Kubota et al., 2012; Shanker et al., 2011, 2014; Prasad et al., 2014; Jin et al., 2015; Atmar et al., 2015). In addition to revealing that the HBGA interaction exclusively involves distally exposed regions of the hypervariable P2 subdomain, these studies have shown how the HBGA binding sites between the genogroups differ and how the sequence variations in the P2 subdomain within each genogroup affect the HBGA specificity.

A striking observation from the crystallographic studies is that the binding sites in GI and GII NoVs are distinctly different in their locations, structural characteristics, and in the modalities how the carbohydrate residues of the HBGA interact (Fig. 3.1.3A). In GI NoVs, while the majority of interactions with HBGA are localized within each subunit of the P domain dimer, in GII NoVs, they are shared between the opposing subunits of the P dimer (Fig. 3.1.3B). Another distinguishing feature is that in GI, the majority of the interactions with HBGA primarily involve a Gal moiety (Fig. 3.1.3C), whereas in GII, the interactions are centered on the Fuc residue (Fig. 3.1.3D). An exceptionally well-conserved feature in GI viruses is the hydrophobic interaction between the SeFuc moiety (as in the H-type) or the N-acetyl arm of N-acetylgalactosamine (as in the A-type) with a conserved Trp residue in the P2 subdomain (Fig. 3.1.3C). This combinatorial requirement of Gal and hydrophobic interactions appears to restrict the types of HBGAs that GI NoVs can bind. Several studies consistently show that most GI NoVs do not bind B-type HBGAs. A likely explanation is as follows: although B-type HBGA has a terminal Gal residue, it lacks an additional group, such as SeFuc present in the H-type or an N-acetyl arm present in the A-type that could engage in the hydrophobic interactions, resulting in lower affinity. HBGA binding in GII NoVs does not involve such a combinatorial requirement allowing them to bind all ABH HBGAs. This likely is one of the factors why GIIs, particularly GII.4 NoVs, are globally more prevalent.

Sequence alterations in the P domain within the genogroup members also contribute to the variation of HBGA binding profiles (Prasad et al., 1999; Cao et al., 2007; Choi et al., 2008; Hansman et al., 2011; Kubota et al., 2012; Shanker et al., 2011; 2014; Bu et al., 2008; Chen et al., 2011). A generalizable concept from the crystallographic studies of the GI and GII P domains in complex with various HBGAs is that HBGA binding in human NoVs involves two sites. The first site is formed by the highly conserved residues in the less flexible regions of the P2 subdomain, which preserves the Gal and Fuc dominant nature of interactions in GI and GII, respectively. Minor variations in this site could result in differences in the HBGA binding affinity. The second site is formed by the less conserved residues that are typically from the loop regions surrounding the first site. This site allows for differential binding to Lewis HBGAs in both GI and GII as discussed later.

Similar to picornaviruses and as demonstrated in animal caliciviruses, VPg (by inference in NoVs) is covalently linked to the genomic RNA (Thorne and Goodfellow, 2014). However, unlike picornavirus VPg, which is only about ∼2 kDa in size, VPg in caliciviruses is significantly larger (12–15 kDa). The calicivirus genomes do not have a capped 5′-end, like cellular RNAs, or an internal ribosome entry site (IRES) as in picornaviral RNA, for RNA translation. Studies on animal caliciviruses and NoVs have attributed a dual role to this protein. First, VPg acts as a “cap substitute” and mediates translation initiation of viral RNA based on the observations that m⁷-GTP cap can substitute for VPg to confer infectivity in vitro to synthesized FCV RNA (Sosnovtsev and Green, 1995), and that it can bind directly to initiation factor eIF4E (Daughenbaugh et al., 2003; 2006; Goodfellow et al., 2005). Second, analogous to picornavirus VPg, the NoV VPg has a priming function during RNA replication based on the observation that VPg is uridylylated at a conserved Tyr residue by the viral RdRp followed by elongation in the presence of RNA (Belliot et al., 2008; Han et al., 2010; Mitra et al., 2004; Royall et al., 2015; Chung et al., 2014).

The role of the NoV protease in the polyprotein processing of the polyprotein, the primary cleavage sites, and the order in which it cleaves the polyprotein have been firmly established. The proteolytic processing of the polyprotein occurs in a sequential manner as a mechanism to regulate viral genome expression and replication (Belliot et al., 2003; Hardy et al., 2002). Protease cleaves the polyprotein at five sites with three different cleavage junctions—Gln/Gly, Glu/Gly, and Glu/Ala, first cleaving the two Gln/Gly junctions between p48/p41 and p42/p22 followed Glu/Gly (VPg/Pro) and Glu/Ala (Pro/RdRp) junctions (Sosnovtsev and Green, 2000; Hardy et al., 2002; Sosnovtsev et al., 2006). These sites exhibit significant variations in the amino acid composition in both the N- (P5–P2) and C-terminal (P2′–P5′) sides flanking the scissile bond (P1/P1′) (Fig. 3.1.4B). Mutational analysis has shown that residues surrounding the cleavage sites contribute to proteolytic efficiency (Hardy et al., 2002). An interesting question is how the protease recognizes such nonhomologous sites within the polyprotein with differential affinities.

In caliciviruses, this obligatory enzyme that is similar to the picornavirus 3D^pol, is responsible for synthesizing both negative-sense RNA as well as newly made positive-sense genomic RNA. X-ray structures of RdRps from several caliciviruses including RHDV (Ng et al., 2002), GII NoV (Ng et al., 2004; Zamyatkin et al., 2008), sapovirus (Fullerton et al., 2007), and MNV (Lee et al., 2011; Hogbom et al., 2009) have been determined. Calicivirus RdRp exhibits a typical “right hand” configuration of palm, finger, and thumb domains as observed in all RNA/DNA polymerases (Ng et al., 2008). In addition to these domains, like in other RdRps, it has a distinct N-terminal domain bridging the fingers and the thumb domains. The active site of the RdRp is located in the thumb domain, which consists of three conserved Asp residues critical for mediating catalysis through a two metal-ion mechanism (Fig. 3.1.5C) and other key residues such as Arg, Asn, and Ser required for substrate binding and catalysis.