The first step of a viral infection is characterized by attachment onto a susceptible cell through surface components that will contribute to virus entry. Those cell surface components involved in the process of binding and entry function as viral receptors. Yet, in many instances, binding to a single attachment or adhesion factor is not sufficient for infection since additional molecules, or coreceptors are required for entry to take place (Verdaguer et al., 2014). In the case of human NoVs, recognition and binding to several glycans has been demonstrated. However, their exact role in the infection process remains ill-defined since until very recently it has not been possible to cultivate human NoVs, hampering studies of the entry process.

Following the discovery that Rabbit Hemorrhagic Disease Virus (RHDV), another member of the Caliciviridae family, could attach to glycans of the histo-blood group family (HBGAs) (Ruvoën-Clouet et al., 2000), it was observed that human NoVs recognize similar carbohydrates (Marionneau et al., 2002). It was then suggested that human NoVs could additionally bind to heparan sulfate (Tamura et al., 2004) and sialylated glycans such as sialyl Le^x (SLe^x) and sialyl-type 2 precursor (Rydell et al., 2009b). Very recently, specific NoV recognition of gangliosides, that is, glycosphingolipids carrying one (or more) sialic acids, was demonstrated in addition to that of HBGAs (Han et al., 2014), an intriguing observation reminiscent of the recognition of HBGAs and gangliosides by various strains of rotavirus (Hu et al., 2012; Huang et al., 2012; Martinez et al., 2013).

The α1,2-fucosyltransferase encoded by the FUT2 gene is key to the process because it contributes to the synthesis of the A, B, H, Le^b, and Le^y antigens. Accordingly, the presence of at least one of these antigens characterizes the so-called secretor phenotype, whereas its absence, due to mutations in the enzyme’s gene coding sequence, generates the so-called nonsecretor phenotype. The secretory ABH and Lewis phenotypes of each individual depend on the combined polymorphisms at the three loci ABO, FUT2, and FUT3. Using virus-like particles (VLPs), spontaneously formed from the recombinantly expressed capsid protein of NoVs, specific recognition patterns of HBGAs have been described. This was performed by analyzing the attachment of VLPs to tissue sections of digestive tissues and to saliva mucins from HBGA phenotyped individuals, to immobilized synthetic oligosaccharides or to cells transfected in order to express the appropriate glycosyltransferases. Through these methods, it was initially noticed that A, H, and Le^b antigens are ligands of the Norwalk Virus (NV), a prototype virus belonging to the GI.1 genotype (Marionneau et al., 2002). In contrast, B and Le^a antigens are not recognized by this viral strain (Huang et al., 2003). Following these initial studies the glycan specificity of many other strains has been studied using VLPs belonging to all genotypes. It was observed that most strains bind to HBGAs, albeit important variations in specificities were found within each of the main human genogroups (GI and GII) (Tan and Jiang, 2011,  2014). Thus, within GI two major types of binding patterns could be distinguished, that is, strains that are similar to the previously described NV GI.1 strain, recognize the A and H epitopes and strains that bind to the Lewis epitopes, targeting the α1,3/4-linked fucose residue (Fig. 3.3.1). The latter strains show variation in their ability to accommodate the α1,2-linked fucose residue and the A or B epitopes, and a dominant binding to the α1,3/4-linked fucose, allowing them to recognize nonsecretor individuals as observed by their binding to saliva from such individuals. Among GII strains three types of binding patterns have been discerned. The first one has been called ABH binder as these strains attach to saliva from all secretor individuals, regardless of the ABO and Lewis status. The second, called A/B binder, attach to saliva from secretor individuals of the A and/or B blood groups only. The third, called Lewis binder, attach to saliva of Lewis positive individuals with more or less influence of the secretor phenotype, indicating strong involvement of the α1,3/4-linked fucose (Tan and Jiang, 2011; Tan and Jiang, 2014). Structural analysis of the capsid protein’s protruding region, or P-domain, cocrystallized with oligosaccharides allowed characterization of the exact binding sites for various GI and GII strains (Bu et al., 2008; Cao et al., 2007; Chen et al., 2011; Choi et al., 2008; Hansman et al., 2011). Interestingly, GI and GII binding sites appeared to be completely different, although the amino acids involved are highly conserved within each genogroup. In addition binding modes between the two genogroups are very different (Cao et al., 2007; Ruvoën-Clouet et al., 2013). The GI binding sites involve contact mainly with galactose or N-acetylgalactosamine residues, while the GII sites involved essentially fucose recognition (Cao et al., 2007; Koppisetty et al., 2010; Nasir et al., 2012). Nevertheless, in both instances slight amino acid changes generate subtle modifications in the range of preferred oligosaccharides, allowing recognition of distinct subgroups of individuals or populations. The different ways by which GI and GII strains cope with their host’s glycan diversity are strongly suggestive of a convergent evolution and underline the important role of HBGA binding in the process of infection. These structural aspects are presented in detail in Chapter 3.1.

The influence of the secretor status on susceptibility to NoV infection in authentic outbreaks was first investigated in Sweden (Thorven et al., 2005). Thirty-eight symptomatic individuals from three hospital outbreaks caused by GII.4 strains were genotyped for secretor status. In addition, the same analysis was performed on 15 symptomatic individuals from three community outbreaks caused by GI.6 and GII.6 strains. Thus, in total, 53 symptomatic and 62 asymptomatic individuals were genotyped for polymorphisms of FUT2 at nucleotides 385, 428, and 571. Strikingly, no nonsecretors were identified among the symptomatic individuals. As the secretor status was determined by genotyping, the influence of homozygosity and heterozygosity could be determined. However, no difference was identified between heterozygous and homozygous secretors. Subsequently, the resistance of nonsecretors to GII.4 strains in authentic outbreaks was confirmed by a similar study in Denmark (Kindberg et al., 2007). Similar results were also reported from a GII.4 outbreak in China (Tan et al., 2008). In that study, secretor status was determined by phenotyping of saliva samples and the weak-secretors were grouped with the secretors. In addition, two studies have investigated the influence of secretor status on susceptibility to infection with recently emerged GII.4 strains. Currier et al. (2015) investigated sporadic pediatric NoV infections and could not identify any nonsecretors among 155 patients with GII.4 infections, even though 24% of the NoV negative control individuals were nonsecretors. The GII.4 viruses identified belonged to the Den Haag 2006b (18%), New Orleans 2009 (46%), and Sydney 2012 (36%) clusters. In another pediatric material from Ecuador all 27 GII.4 single infections and all 4 coinfections that were positive for GII.4 strains were found in secretors (Lopman et al., 2015). In addition, a number of studies of NoV outbreaks caused by strains from other genogroups have reported that only secretors were identified among the symptomatically infected patients, even though the materials were not large enough to prove any statistical significance (Bucardo et al., 2009; Tan et al., 2008). Furthermore, a study of Swedish blood donors showed that secretor positive individuals had significantly higher sero-prevalence and IgG antibody titers to NoV GII.4 virus than nonsecretors (Larsson et al., 2006). The remarkable match between susceptibility to infection, the presence of either A, B, or H antigens in the digestive tract and the carbohydrate specificity of the virus strongly suggested that the interaction between the virus and these glycans is an essential step of infection.

Likewise, GI.3 strains have been reported to attach to saliva from secretors as well as nonsecretors of the A, AB and O blood groups and of the Lewis positive phenotype (Shirato et al., 2008; Yazawa et al., 2014). GII.7 strains bind to saliva from nearly all individuals, except Lewis negative nonsecretors (Shirato et al., 2008; Yazawa et al., 2014). Besides strong binding to saliva from secretors, some binding to saliva of Lewis positive nonsecretors of a GII.6 strain was also reported (Yazawa et al., 2014). Regarding old GII.4 strains from the 1990s (US95/96), all studies indicate a strong preference for recognition of HBGAs present in secretors (Huang et al., 2005; Lindesmith et al., 2008; Ruvoën-Clouet et al., 2014; Rydell et al., 2009b). Nonetheless, in some studies, weak binding to saliva from nonsecretor Lewis positive individuals have been observed (Ruvoën-Clouet et al., 2014). In addition, more recent strains of the GII.4 den Haag and Osaka subtypes readily attach to saliva from nonsecretors, but only if they are Lewis positive, in accordance with their ability to recognize Lewis antigens (de Rougemont et al., 2011). From these data it is to be expected that some nonsecretor individuals should be infected, albeit less frequently than secretors, and indeed this has been observed on several occasions.

Thus, a volunteers’ study performed using the GII.2 Snow Mountain Virus (SMV) strain failed to reveal an association between infection and HBGA phenotypes (Lindesmith et al., 2005), despite the fact that SMV reportedly showed strong preference for the B blood group (Harrington et al., 2002). However, that volunteers’ study was conducted on a small number of subjects, some of whom had received very high doses of virus. The volunteers’ study performed using a GII.4 strain, showed one subject of the nonsecretor phenotype that was infected, albeit presenting very mild symptoms (Frenck et al., 2012). Furthermore, a limited number of outbreaks studies also showed that nonsecretor individuals can be infected by NoV (Carlsson et al., 2009; Lindesmith et al., 2005; Nordgren et al., 2010; Rockx et al., 2005). Both Rockx et al. (2005) and Nordgren et al. (2010) observed that GI.3 NoV can infect secretors as well as nonsecretors, in agreement with GI.3 VLP binding studies demonstrating binding of nonsecretor saliva and Le^a glycoconjugates (Shirato et al., 2008; Yazawa et al., 2014). Rockx et al. (2005) investigated a water-borne GI.3 NoV outbreak and found that 20/22 (91%) of the secretors and four of seven (57%) of the nonsecretors exposed to the virus became infected. Nordgren et al. (2010) investigated a food-borne GI.3 NoV outbreak, in which symptoms suggestive of NoV infection were found in seven of 15 (47%) of the nonsecretors and 26/68 (38%) of the secretors. In another study, Carlsson et al. (2009) showed that even though nonsecretor individuals had a significantly lower risk of getting infected in a GII.4 NoV outbreak in Spain, the virus also infected one individual, genotyped as a nonsecretor, suggesting that nonsecretors are not totally resistant to infections caused by the dominating NoV genotype. A similar result was obtained in a study in Burkina Faso, in which one nonsecretor and Lewis negative individual was infected with a GII.4 strain (Nordgren et al., 2013).

Although these occasional infections of nonsecretor individuals by strains such as GII.4 that show a clear preference for secretors may be explained by their ability to recognize the α1,3/4-linked fucose residue of the Lewis type present in Lewis positive (FUT3 + ) individuals even in absence of α1,2-linked fucose as in nonsecretors (FUT2−), another explanation for these cases has recently been raised. It was shown that chronic infection by virulent strains of the bacterium Helicobacter pylori aberrantly induces expression of α1,2-fucosylated motifs in the stomach mucosa of nonsecretors, which may contribute to facilitate infection of these individuals by NoV strains that otherwise show a strong preference for secretors (Ruvoën-Clouet et al., 2014).

In addition to the occasional infections of nonsecretors that seemed to contradict the early reports on the genetic susceptibility to NoV, several reports have presented apparent contradictory results concerning the association between the ABO phenotype and infection (Fretz et al., 2005; Halperin et al., 2008; Meyer et al., 2004; Miyoshi et al., 2005). It must be noted that these studies did not take into account the strains involved and their binding specificities for HBGAs. Their interpretations are therefore dubious since, as discussed earlier, NoV strain specificities for HBGAs are variable. It is highly likely that the differences observed in NoV epidemics stem from differences in specificity for ABH antigens of the strains involved. Finally, the existence of NoVs VLPs that do not appear to attach to HBGAs has been reported (Huang et al., 2005; Shirato et al., 2008; Takanashi et al., 2011; Yazawa et al., 2014), suggesting either the existence of human NoVs that can infect independently of HBGAs attachment or that the methods of VLPs binding characterization may not always reflect the behavior of true viruses.

NoV GV strains exclusively infect mice and do not appear to recognize HBGAs. Yet, they bind to glycosphingolipids of the ganglioside type. Since these strains can be cultured, it has been possible to show that the ganglioside GD1a functions as receptor for these strains (Taube et al., 2009). This is particularly interesting in view to the recent discovery that human GI and GII strains could also bind to gangliosides with affinities in the same range as those involved in HBGAs binding (Han et al., 2014).

Thin-layer chromatogram binding assay (CBA) was used to show that the Norwalk virus VLP recognizes both type-1 and type-2 chain glycosphingolipids terminated with blood group A and H but not B epitopes (Fig. 3.3.2) (Nilsson et al., 2009). Subsequently, the binding of a Norwalk and Dijon (GII.4 US95/96) VLP to glycosphingolipids incorporated in fluid solid-supported lipid bilayers was studied using quartz crystal microbalance with dissipation monitoring (QCM-D) (Fig. 3.3.2) (Rydell et al., 2009a). Both VLPs recognized bilayers containing H type-1, whereas no binding was observed to bilayers containing Le^a. For both VLPs, the concentration of H type-1 glycosphingolipid in the bilayer had to be above a threshold value for binding to occur, suggesting that a multivalent interaction is needed to stably attach the VLPs to the bilayer since the binding affinity of single ligands is expected to be low. Interestingly, the threshold concentration was one order of magnitude higher for the Dijon strain VLP, possibly suggesting the Norwalk strain to have a higher affinity for the H type-1 glycosphingolipid compared to the Dijon strain.

It may be calculated that the secretor α1,2 linked fucose of the H type 1 glycan cocrystallized with GII.4 VA387 (Cao et al., 2007) contributes to more than half of the glycan-protein binding energy (Koppisetty et al., 2010). Through molecular dynamics simulations, it has been suggested that human GII NoV may recognize fucosylated glycans either through secretor (α1,2 linked) or through the Lewis (α1,3/4 linked) fucose residues in the binding site (Nasir et al., 2012). The hypothesis was confirmed by crystal structures of Lewis antigens (SLe^x and Le^y) in complex with GII NoV (Hansman et al., 2011). These so called “secretor” and “Lewis” poses (Nasir et al., 2012) could have implications in binding complex branched structures of glycoproteins or glycosphingolipids in membranes (Fig. 3.3.3).

To further quantitatively characterize the virus membrane interactions and reveal the complex mathematical details of capsid protein to glycan interactions, Total Internal Reflection Fluorescent Microscopy (TIRF-M) based binding assay was recently established (Fig. 3.3.2) (Bally et al., 2011,  2012a). This assay makes it possible to record the attachment and detachment events of single glycosphingolipid-containing vesicles to VLPs bound to fluid solid-supported lipid bilayers in real time (Fig. 3.3.2). Quantitative binding data describing the vesicle attachment to and detachment from Dijon (Bally et al., 2011) and Ast6139 GII.4 VLPs (Nasir et al., 2015) were obtained at both transient and steady state conditions using TIRF-M methodology.

In an attempt to address the receptor function of glycosphingolipids, NoV interactions with glycosphingolipids embedded in giant unilamellar vesicles (GUVs) was studied using fluorescently labelled GII.4 VLPs and phospholipid vesicles containing glycosphingolipids and a fluorescent lipid (Rydell et al., 2013). The binding pattern to glycosphingolipids incorporated into GUVs was in full agreement with thin-layer chromatography (CBA) experiments. Upon binding to the vesicles, the VLPs induced the formation of membrane invaginations that were positive both for the fluorescent lipid and the VLP (Fig. 3.3.4). Similar invaginations have been shown to correspond to endocytosis intermediates used for cell entry for endogenous lectins, microbial toxins, viruses, and bacteria known to use glycosphingolipids for cell entry (Eierhoff et al., 2014; Ewers et al., 2010; Lakshminarayan et al., 2014; Roemer et al., 2007). The formation of the membrane invaginations does not require clathrin or any other cytosolic coat proteins. Instead, the multivalent binding of receptor glycosphingolipids induces the formation of lipid-nanodomains that spontaneously invaginate to form tubular structures. The mechanisms by which the membrane is bent is still not clear, but likely involves line tension, asymmetric compressive stress, a specific protein–lipid geometry, and the organization of lipids in specific orientation fields in the membrane (Aigal et al., 2015; Arnaud et al., 2014; Johannes et al., 2014). The observation that NoV VLP has an intrinsic property to induce membrane invagination on model membranes suggests that glycosphingolipids may be functional receptors also for human NoV. Some strains of murine NoV has indeed been shown to use gangliosides as receptors on macrophages (Taube et al., 2009). When complex glycosphingolipids were depleted from the cells the infectivity of the virus was markedly decreased. Furthermore, the infectivity could be rescued by addition of the ganglioside GD1a to the cells. In addition, the entry of murine NoV has been described to be independent of caveolin and clathrin, but dependent on cholesterol (Gerondopoulos et al., 2010; Perry and Wobus, 2010). These properties are characteristic also for the bacterial toxins and viruses that use the glycosphingolipid-dependent membrane invagination pathway to enter cells (Ewers et al., 2010; Roemer et al., 2007).

The apical surface of epithelial cells presents a large glycocalyx, up to 500 nm long in the duodenum that is decorated with HBGAs. It is tempting to speculate that human NoVs first attach to these motifs of the glycocalyx in order to reach their glycolipid targets that protrude a couple of nanometers only above the lipid bilayer (Fig. 3.3.2C). The HBGA-mediated glycocalyx binding would be a primary attachment and tethering step, whilst the glycosphingolipid binding would directly contribute to the entry process, as observed with the GUV experiments.

In conclusion, detailed understanding of the genetic coevolution and the cell-surface molecular interactions between enteric viruses and host receptors will hopefully contribute to the development of novel antiviral therapies and prevention of gastrointestinal infections.