Structural Basis of Glycan Recognition of Rotavirus

Rotavirus (RV) is an important pathogen causing acute gastroenteritis in young humans and animals. Attachment to the host receptor is a crucial step for the virus infection. The recent advances in illustrating the interactions between RV and glycans promoted our understanding of the host range and epidemiology of RVs. VP8*, the distal region of the RV outer capsid spike protein VP4, played a critical role in the glycan recognition. Group A RVs were classified into different P genotypes based on the VP4 sequences and recognized glycans in a P genotype-dependent manner. Glycans including sialic acid, gangliosides, histo-blood group antigens (HBGAs), and mucin cores have been reported to interact with RV VP8*s. The glycan binding specificities of VP8*s of different RV genotypes have been studied. Here, we mainly discussed the structural basis for the interactions between RV VP8*s and glycans, which provided molecular insights into the receptor recognition and host tropism, offering new clues to the design of RV vaccine and anti-viral agents.


INTRODUCTION
Rotavirus (RV), belonging to the Reoviridae, is an important pathogen leading to acute gastroenteritis (AGE) in children under 5 years old and caused ∼200,000 deaths worldwide each year (Tate et al., 2016;Bányai et al., 2018). RV genome contained 11 segments of double-stranded RNA, encoding 6 structural proteins (VP) and 6 non-structural proteins (Estes and Greenberg, 2013). The RV capsid has three layers consisting of a core layer formed by VP2, an intermediate layer formed by VP6, and an outer layer formed by VP4 and VP7 (Figure 1). Based on the antigenic and molecular characteristics of VP6, RVs are currently classified into nine groups/species (A-I) and a further tentative group J (Banyai et al., 2017). Groups A, B, C, and H RVs have been identified in human infections, while other groups only cause diseases in animal species (Matthijnssens et al., 2012;Banyai et al., 2017). Among these, group A RVs (RVAs) are the most widely prevalent in humans and the leading cause of severe AGE worldwide.
VP7 is a glycoprotein and VP4 is protease-sensitive (Estes and Greenberg, 2013). VP4 extending from the VP7 shell formed the major spike protein contributing to the viral attachment and penetration ( Figure 1) (Dormitzer et al., 2002b). RV was classified into G and P genotypes based on VP7 and VP4, respectively, (Matthijnssens et al., 2011). To date, no less than 37 G and 51 P genotypes of RVAs have been identified (https://rega.kuleuven.be/cev/viralmetagenomics/virus-classification). Different combinations of G and P genotypes have been reported in human infections whereas G9P [8], G1P[8], G3P[8], G2P[4], G8P[8], are the widely prevalent RVAs (Lestari et al., 2020). There is a great genetic and strain diversity of RVs, contributed by point mutations, gene rearrangement, and genetic assortment between co-circulating strains. Furthermore, interspecies transmission between human and animal RVs has been reported in different genotypes (Mukherjee et al., 2011). Though two licensed RV vaccines are effective and widely used in many countries all over the world (Anh et al., 2011;Wang et al., 2013), how effective the vaccines will be as the genetic alteration of the prevalent RVs remains unknown.
VP4 can be cleaved into two subunits, VP5* and VP8* (Larralde et al., 1991). VP8*, located at the distal terminal of the spike, is responsible for the virus-ligand interaction while VP5* facilitates the host cell penetration through the conformation rearrangement and membrane fusion ( Figure 1) (Settembre et al., 2011). VP8* has been identified to interact with specific glycans in a P genotype dependent manner . Previously, 35 P genotypes were classified into five genogroups based on the VP8* sequences (Figure 2A) (Liu et al., 2012 Figure 2B). Here, we delineated recent advances in the structural basis for glycan recognition of RV VP8*s.

GLYCANS RECOGNIZED BY ROTAVIRUSES
Some animal RVs were reported to recognize terminal sialic acids (SAs) (Fukudome et al., 1989;Isa et al., 1997). N-acetylneuraminic acid (Neu5Ac) and N-glycolylneuraminic acid (Neu5Gc) are the most common neuraminic acids in nature and widely expressed on the surface of most mammalian cells.  (Rolsma et al., 1998;Dormitzer et al., 2002a;Isa et al., 2006). The infection of porcine P[7] RV CRW-8 could be efficiently inhibited by the ganglioside GM3Gc glycan (Blanchard et al., 2007). Canine K9 P [3] preferentially bound to Neu5Gc (Mishra et al., 2018). In addition, nuclear magnetic resonance (NMR) and cell infection studies showed that the ganglioside GM1, which lacks the terminal sialic acid but with branched sialic acid, could be a possible ligand for some human RVs, including human P[8] and P[6] (Haselhorst et al., 2009). VP8*s of human P[8] Wa and P[6] RV-3 were identified to bind to GM1 by Saturation transfer difference NMR (STD-NMR) (Fleming et al., 2014). Later studies revealed that most animal RVs and human RVs are SA independent (Ciarlet and Estes, 1999). Recently, some human RV genotypes are found to recognize histo-blood group antigens (HBGAs) , indicating that HBGAs are important cell attachment factors for RVs. HBGAs are a group of carbohydrates (Yamamoto, 1994), distributing abundantly on mucosal epithelia. HBGAs also existed as free oligosaccharides in body fluids, such as saliva, milk, blood, and intestinal content. HBGAs are synthesized by sequential addition of monosaccharides to precursor disaccharides by different glycosyltransferase. The glycosyltransferases are encoded by three major gene families, secretor, Lewis, and ABO families encoding FUT2, FUT3, and A/B enzymes, respectively. RV VP8*s recognized HBGAs in a genotype dependent manner. The P[8] and P[4] RVs that are widely prevalent in humans interacted with mucin cores, lewis b, and type I HBGA, including H type-1 antigen (H1), H1 precursor, lacto-N-tetraose (LNT), Lacto-N-fucopentaose I (LNFP1) Rey et al., 2019;Sun et al., 2020). Human P[6] and P [19] bound to H1, whereas porcine P[6] and P [19] did not (Sun et al., 2018a;Li et al., 2018). P [14] was less common in human. Human P[14] VP8* was found to specifically recognize A type HBGA (Hu et al., 2012 (Liu et al., 2013;Ramani et al., 2013;Hu et al., 2015). VP8* of human P[28] in P[I] genogroup was found to bind H1 HBGAs (Zhao et al., 2020). Bovine P[5] WC3 and its monoreassortant G4P[5] recognized both sialic acid and the α-Gal HBGA (Alfajaro et al., 2019). Interestingly, the α-Gal epitope of the HBGA family was reported to be a ligand for bovine norovirus Newbury2 (Cho et al., 2018), indicating a common feature of the infection of certain bovine RVs and norovirus.
The glycan binding specificity influenced the host tropism and prevalence of RVs. Animal RVs could recognize sialic acid and mucin core 2, whereas human RVs bound to HBGAs and mucin cores. The RVs that can infect both human and porcine such as P [6] and P [19] showed distinct glycan binding preference. Porcine P[6] and P [19] VP8*s recognized mucin core 2, while human P[6] and P [19] VP8*s interacted with the H1 HBGA, indicating an evolutionary path from animal to human. P[11] RV VP8* recognized type I and type II precursors that are developmentally regulated in neonates, consistent with the fact that P[11] RVs are mainly identified in neonates. P[8] and P [4] infections were mainly identified in secretors and lewis positive children (Nordgren et al., 2014), consistent with that P[8]/P[4] RVs could interact with H1 HBGA and lewis antigen.   Figure 4A). The glycan binding site located at one corner of the cleft between two β-sheets (βH/βK) and appeared to be an open-shallow groove. Y188 and S190 form one rim of the groove; Y155 constitutes the other rim; R101, V144, K187 and Y189 side chains make the base part. R101 was proved to be vital for the sialic acid binding (Kraschnefski et al., 2009). CRW-8 VP8* interacted with Neu5Acα2Me using the same residues except H155 and G187 (Blanchard et al., 2007) and bound to ganglioside GM3 glycans by the same pattern (Yu et al., 2011). CRW-8 and RRV VP8* binding to the Neu5Gcα2Me were determined (Yu et al., 2012) (Figure 4B), illustrating that residue 157 of VP8*  influenced the glycan preference. CRW-8 VP8* with S157 showed reduced binding affinity for Neu5Gc compared to that with P157.
The structural basis of P[14] interacting with A type HBGA has been illustrated (Hu et al., 2012) (Figure 4C). The width of the cleft between the two β-sheets is narrower than the cleft in the human VP8*, similar to that in the VP8* of the animal strains ( Figure 3). P[14] VP8* bound to the A type HBGA using the same glycan binding site as P[3] VP8*. However, the structural features of the glycan binding site of P[14] VP8* is that of P[3] VP8*. The amino acid residues involved the A type HBGA binding were R101, I144, L146, Y155, S187, Y188, Y189, and L190 ( Figure 4C). The orientation of Y188 was different to Y188 in P[3] VP8* and would cause steric hindrance to the binding of sialic acid, indicating that the subtle changes of the VP8* could accommodate distinct glycans. The terminal GalNAc (green) and Gal (yellow) of the HBGA contributed to all the interactions, whereas the proximal moiety Fuc (cyan) project out from the surface and did not make any direct contacts (Hu et al., 2012). Human P[9] was identified to recognize A type HBGA using the same glycan binding site as that of P[14], consisting of R101, I144, L146, Y155, S187, Y188, Y189, L190, and T191 ( Figure 4D)  Frontiers in Molecular Biosciences | www.frontiersin.org July 2021 | Volume 8 | Article 658029 8 VP8* interacted with type I and type II precursors using a distinct glycan binding site consisting of N153, R154, N155, Y156, I158, W178, G179, S180, Y183, D185, and R187 (Hu et al., 2015) ( Figure 4E). The type I tetrasaccharide lacto-N-tetraose (LNT) and type II tetrasaccharide lacto-N-neotetraose (LNnT) binding site was expansive and spanned almost the entire length of the cleft between βH and βK ( Figure 4E). Bovine P[11] only interacted with type II precursor and recognized LNnT with residues of S153, R154, N155, Y156, W178, G179, A180, D185, and R187 ( Figure 4F).
The complex structures of P[8] VP8* with different type I HBGAs (H type-1 antigen, H1; H1 precursor lacto-N-biose, LNB; Lacto-N-fucopentaose I, LNFP1) were determined (Rey et al., 2019;Sun et al., 2020). LNB and H1 located at the same site of LNFP1 ( Figure 5B). LNB and H1 interacted with VP8* using similar mechanism. L-fucose of H1 was projected out and did not make direct interactions with VP8*. However, the surface plasmon resonance (SPR) assay showed that P[8] c VP8* bound more intensively to H1 (affinity constants K D 27.9 ± 0.7 μM) compared to LNB (K D 52.1 ± 4.3 μM) (Rey et al., 2019), implying that H1 L-fucose contributes to the glycan binding. LNFP1 overlapped exquisitely with the H1 and LNB moieties. P[8] VP8* interacted with different H1 glycans in a same site but with different binding affinity, indicating that the glycan forms may influence the RV attachment.
The interactions between P[8]/P [19] VP8* and mucin core 2 have been illustrated (Liu et al., 2017;Sun et al., 2018a;Sun et al., 2020). Mucin core 2 interacted with VP8*s at the same site as the type I HBGAs with slightly difference. GlcNAc (blue), GalNAc (green), and Gal (yellow) all participated in the interactions ( Figure 5C) (Sun et al., 2020). P[8]/P [19] VP8* bound to mucin core 2 using the same pattern with residues of W81, L167, Y/HGGR 169-172, W174, T185, R209, and E212 ( Figure 5C), revealing that RV VP8* can accommodate different glycans using the same residues. P[8] and P[4] VP8*s were also reported to interact with lewis b (le b ) Ma et al., 2015). A recent paper elucidated the molecular mechanism for the recognition of P [8] VP8* to le b based on nuclear magnetic resonance (NMR) spectroscopy-based titration experiments and NMR-derived high ambiguity driven docking (HADDOCK) method (Xu et al., 2020). Unlike the H1 binding site composed of an α-helix and a β-sheet (referred as βα binding site), P[8] and P[4] VP8*s were identified to bind le b HBGA in another pocket consisting of the edge of two β-sheets (named ββ binding site) ( Figure 5D). The potential lewis b binding site is proposed to be formed by residues of Y152, N153, R154, R155, T156, T158, H177, G178, E179, A183, and T184 ( Figure 5D). Further investigations such as X-ray crystallization are needed to verify the glycan binding.

OTHER GROUP ROTAVIRUSES
Group/species C rotaviruses (RVCs) have been identified as important pathogens of acute gastroenteritis in children, family-based outbreaks, as well as animal infections (Joshi et al., 2017;Vlasova et al., 2017). Human RVC VP8* was found to recognize A type HBGAs (Sun et al., 2018b). The complex structure of human RVC VP8* and type A trisaccharide exhibited that human RVC VP8* possessed a completely different glycan binding site compared to RVA VP8*s ( Figure 5H). Human RVC bound to type A trisaccharide (GalNAcα1-3(Fucα1-2)Gal) using a pocket consisting of N108, L209, A110, E151, G152, P205, R206, S207, and N208 (Sun et al., 2018b). Both GalNAc and Fuc of the type A HBGA participated in the interactions, while Gal had no direct contact with the RVC VP8*.
Human group B (RVB) and group H rotavirus (RVH) caused outbreaks in China in the 1980s and mainly infected adults. Infections of human RVB and RVH have constantly reported in some areas such as Southeast Asia (Chen et al., 1985;Yang et al., 2004;Jiang et al., 2008;Joshi et al., 2019). The receptor binding specificity of human RVB and RVH is unclear. Whether they recognize sialic acid as some animal RVAs, HBGAs as human RVAs or other glycans still need further investigation.

CONCLUSION REMARKS
Some animal RVs recognized sialic acid, such as P[3], P[7]. Some animal RVs were reported to bind sialic acid and αGal, such as bovine P[5] RVs (Alfajaro et al., 2019). The identification that some human RV VP8*s recognized HBGAs has provided new insights into the infection and transmission of RVs. So far, the interactions between VP8*s of human P interacted with A-type HBGA, which may in a part restricted the prevalence of these RVs. The functions of these glycans, such as sialic acid, HBGAs, mucin cores in the RV infection or crossspecies transmission still need more studies to clarify. Structural biology has significantly contributed to our understanding of the interaction between RV and glycans. However, the complexity and variety of glycan recognition of RV VP8*s indicated hostpathogen co-evolution with the structural and functional adaptation of RV to host glycan polymorphisms. More efforts exploring the structural basis for the VP8*-glycan interactions are necessary to fully understand the role of glycans in RV infection and transmission, which will facilitate the development of novel RV vaccines and anti-viral agents.