Norovirus replication, host interactions and vaccine advances

Abstract

Human noroviruses (HuNoVs) are the leading cause of acute gastroenteritis worldwide in all age groups and cause significant disease and economic burden globally. To date, no approved vaccines or antiviral therapies are available to treat or prevent HuNoV illness. Several candidate vaccines are in clinical trials, although potential barriers to successful development must be overcome. Recently, significant advances have been made in understanding HuNoV biology owing to breakthroughs in virus cultivation using human intestinal tissue-derived organoid (or enteroid) cultures, advances in structural biology technology combined with epitope mapping and increased metagenomic sequencing. New and unexpected strain-specific differences in pandemic versus non-pandemic virus structures, replication properties and virus–host interactions, including host factors required for susceptibility to infection and pathogenesis, are discussed.

Introduction

Human noroviruses (HuNoVs) are a leading cause of gastroenteritis and are estimated to cause 677 million cases of acute gastroenteritis globally each year¹. The greatest disease burden is in children under 5 years of age, but persons of all ages are affected. The global economic burden is significant, with an estimated US$4.2 billion in direct healthcare costs and $60.3 billion in societal costs annually². Norovirus-associated illness is characterized by vomiting and diarrhoea, and although disease is usually self-limited, chronic diarrhoea leading to malnutrition can occur in immunocompromised persons³. These important pathogens have been difficult to study because they remained noncultivatable for almost 50 years. This barrier was overcome with HuNoV cultivation in human intestinal enteroids, providing new tools to study virus biology and pathogenesis. In addition, advances in the structural biology of HuNoVs has provided new insights into virus replication and immunity. In this Review, we highlight recent progress in the field, along with developments in the evaluation of candidate vaccines. We summarize information about HuNoV classification and highlight new information about HuNoV biology that has been discovered in the past 5 years using the replication system, including strain-specific differences in cell entry and innate immune responses. We describe neutralization assays combined with monoclonal antibodies, nanobodies and structural biology techniques to define epitopes and mechanisms of neutralization, as well as targets for antiviral development. The status of vaccine evaluation is presented, along with potential barriers to successful vaccine development.

Genetic classification and epidemiology

Norovirus is one of 11 genera in the family Caliciviridae, and its members infect a variety of mammalian species. The Norovirus genome consists of a single-stranded, positive-sense RNA of approximately 7.5 kb in length, with a viral protein (VPg) covalently linked to the 5′ end and a polyadenylated tail at the 3′ end (Fig. 1). The HuNoV genome contains three open reading frames (ORFs). ORF1 encodes a precursor polyprotein that is post-translationally cleaved by the viral protease into six non-structural proteins (NSPs): p48 (NS1/2), p41 (NS3), p22 (NS4), VPg (NS5), protease (NS6) and RNA-dependent RNA polymerase (RdRP) (NS7) (Fig. 1a,b). ORF2 and ORF3 are translated from a subgenomic RNA that encodes the major (VP1) and minor (VP2) structural proteins⁴.

Fig. 1 |. Human norovirus genome, entry, replication and structure.

a, The human norovirus (HuNoV) genome (approximately 7.5 kb) is a viral protein (VPg)-linked positive-sense RNA with a poly(A) tail consisting of three open reading frames (ORFs). Replication of the HuNoV consists of several stages (1–8): (1) cell attachment involving interactions with histoblood group antigen (HBGA); (2) endocytic internalization; (3) disassembly; (4) ribosomal translation of the three ORFs; (5) cleavage of the polyprotein encoded by ORF1 by the viral protease (Pro) into six non-structural proteins (NSPs), synthesis of structural proteins VP1 and VP2 encoded by ORF2 and ORF3, respectively; (6) replication of the genomic and subgenomic RNA by the viral RNA-dependent RNA polymerase (RdRP); (7) encapsidation of the progeny VPg–RNA into capsids formed of VP1 and VP2 that are synthesized from subgenomic viral RNA; and the final stage of (8) viral capsid release. b, Among the NSPs, Pro and the RdRP are structurally the best characterized and have been the targets for developing a variety of classes of small molecule inhibitors. c, The T = 3 capsid of the HuNoV formed by 180 VP1 subunits in three quasi-equivalent positions (A, B and C) is shown along the icosahedral 3-fold axis with surrounding 5-fold axes denoted. The subunits A and B forming A/B dimers surrounding the 5-fold axes are shown in cyan and blue, whereas the C subunits, which form C–C dimers at the icosahedral 2-fold axes, are shown in green. Cartoon representations of the bent (A–B) and flat (C–C) dimers are shown on the right with protruding (P) and shell (S) domains and the N-terminal arm (NTA) denoted in the A/B dimer. d, HBGA (solid spheres in red) binding at the P2 subdomains of VP1 dimer (cyan and blue) in GI (left) and GII (right).

The Norovirus genus is further subdivided into genogroups and genotypes using phylogenetic analysis of the complete amino acid sequence of VP1 (refs. 5,6). Defined genogroups (and genotypes) have at least two complete VP1 sequences from geographically distinct locations. There are currently ten recognized genogroups (GI–GX), with two additional tentative (non-assigned) genogroups (GNA1 and GNA2); human viruses are found in GI, GII, GIV, GVIII and GIX genogroups. These genogroups are further subdivided into 48 recognized genotypes, with three non-assigned genotypes (two in GII and one in GIV) (Fig. 2a). Variants have been recognized for some genotypes and represent epidemiologically relevant phylogenetic subclusters within the genotype^5,6.

Fig. 2 |. Phylogenetic trees of norovirus genogroups and genotypes.

a, VP1 amino acid sequences and b, RNA-dependent RNA polymerase (RdRP) nucleotide sequences using previously published sequences^6,13 reflect current classification systems (Supplementary methods). The natural hosts affected by each genotype are colour coded. c, GII.4 strains undergo epochal evolution and emergent variants are shown by their time of circulation^{8,9,16,127,230}. The Sydney variant has two major P types (P31 for the 2012-like strains and P16 for the 2015-like strains).

Recombination between ORF1 and ORF2 allows the association of multiple different ORF1s with a single capsid genotype and led to the development of a second classification schema using partial nucleotide sequences of the viral RdRP to assign P types (Fig. 2b). There are eight P genogroups (corresponding to GI–GVII and GX capsid genogroups) and two tentative P genogroups containing 60 P types and an additional 14 tentative P types⁶. The GVIII and GIX genogroup viruses have GII P types. Two genotyping tools are publicly available: Norovirus Typing Tool and the Human Calicivirus Typing Tool.

The most common genotype that has caused infection and illness for more than two decades is GII.4 and in most years represents more than 50% of identified strains^7–9. Other genotypes may emerge globally and predominate for a year or two (such as GII.2 and GII.17)^10–12, but they have not persisted like the GII.4 viruses¹³. There can also be local and regional differences in the circulating genotypes over time^8,13. The reasons for emergence and persistence are multifactorial and include RdRP fidelity, recombination, binding affinity for histoblood group antigens (HBGAs), population susceptibility and immunity. Norovirus genomic replication is accomplished by the viral RdRP, and the low fidelity of the enzyme leads to mutations in the genome that can accumulate over time based upon virus fitness and other environmental factors¹⁴. Among non-GII.4 strains, the nucleotide changes are largely synonymous, leading to minimal changes in the capsid protein sequence and suggesting that there are evolutionary constraints on VP1 evolution for these viruses¹⁵. By contrast, GII.4 strains have mutations that occur in one or more antigenic sites, leading to the potential for immune escape and the emergence of new variants. This has occurred at least nine times over the past three decades (Fig. 2c), with the most recent being the emergence of the San Francisco variant¹⁶. Recombination events leading to association of the capsid with a new ORF1 (such as P16) may also influence the emergence of new variants^13,17.

Host genetic factors also influence susceptibility to norovirus infections. HuNoVs recognize and bind to HBGAs in a strain-specific manner (Table 1), and there is substantial variability in HBGA expression that is genetically determined. Persons who do not express the appropriate HBGAs on their cell surface are resistant to infection with certain strains^18–20. Mutations in the capsid protein may also increase both the breadth of recognition and stronger binding affinity to HBGAs, as described for newly emerging GII.17 and GII.4 strains^21–24.

Table 1 |.

Association of histoblood group binding characteristics with infections observed in controlled human infection models and epidemiological studies

Genotype (variant)	Saliva binding assay	Oligosaccharide binding assay	Human studies (CHIM and epi)	Refs.
GI.1	Binds A and H; no binding to B or secretor-negative	Binds A, H and Le-b; no binding to B or Le-a	CHIM: infects persons expressing either A, B, H or Le-b, secretors only	18,19,73,173–176
GI.3 (A)	Binds A and H and Le-a; no binding to B or Le-b	Not reported for this variant	Epi: infects persons expressing either A, B, H, Le-a or Le-b	177,178
GII.2	Binds B and in presence of bile acid binds A; no binding to A, H or secretor-negative	No binding to A, H, Le-a or Le-b (B glycan binding not reported)	CHIM: infects persons expressing either A, B, H, Le-a or Le-b, secretors and non-secretors	74,173,179–181
GII.3	Binds A and B; does not bind H or secretor-negative	Binds A, B, H and Le-b; does not bind Le-a	Epi: infects secretors and non-secretors	175,176,182,183
GII.4 (2002/Farmington Hills)	Binds A, B and H; does not bind secretor-negative	Binds B and H; does not bind A, Le-a or Le-b	CHIM: infects secretors and non-secretors, with secretors at higher risk than non-secretors	126,184
GII.4 (2004/Hunter)	Binds A, B, H and Le-b; no binding to secretor-negative except saliva from one asymptomatically infected Le-a non-secretor	Binds A, B, H and Le-b; does not bind Le-a	Epi: infects secretors and non-secretors, with secretors at higher risk than non-secretors; infects Le-b and Le-a	21,185
GII.4 (2006b/Den Haag)	Binds A, B, H and Le-positive; does not bind Le-negative non-secretor	Binds A, B, H and Le-b; does not bind Le-a	Epi: infects persons expressing A, B, H, Le-a and Le-b in one study; in another 100% of infected are secretors	21,182,186
GII.4 (2009/New Orleans)	Binds A, B and H; non-secretor not reported	Binds A, B, H, Le-a and Le-b	Epi: 100% of infected are secretors	22,186,187
GII.4 (2012/Sydney)	Binds A, B and H; non-secretor not reported	Binds A, B, H, Le-a and Le-b	Epi: 100% of infected are secretors	22,186–188
GII.6 (B)	Binds A, B and H; no binding to non-secretor	Binds A and H; no binding of Le-a or Le-b (B glycan binding not reported)	Epi: infects persons expressing either A, B, H, Le-a or Le-b, with secretors at higher risk than non-secretors	189,190
GII.7	Binds B, H and secretor-negative; may bind A	Binds B, Le-a and Le-b; does not bind A or H	Epi: infects secretors and non-secretors	176,177,186
GII.17	Binds A, B and H; no binding to secretor-negative	Not reported for this genotype	Epi: infects persons expressing either A, B, H, Le-a, or Le-b, with secretors at higher risk than non-secretors	191,192

A, A glycan; B, B glycan; CHIM, controLLed human infection model; Epi, epidemiological; H, H glycan; Le-a, Lewis a glycan; Le-b, Lewis b glycan.

Animal models of replication

The in vivo growth of HuNoVs has been examined in several animal species. HuNoV will infect the enterocytes of the small intestine in gnotobiotic pigs²⁵ and calves²⁶, and modest virus-specific IgG and IgA responses are observed. However, these models are cumbersome, expensive and limited by the number of animals that can be used in an experiment. BALB/c mice deficient in the recombination activation gene 1 (Rag1) or Rag2 and common gamma chain (γc) (Rag-γc) support the replication of GII.4 HuNoV in many tissues after intraperitoneal injection²⁷, and they have been evaluated to assess in vivo antiviral activity²⁸. To date, replication of non-GII.4 strains has not been reported in this model. Zebrafish larvae have also been used as a replication model to evaluate candidate antivirals²⁹. Microinjection of virus into the larval yolk sac 3 days post-fertilization leads to virus replication for a number of different genotypes, and the addition of candidate drugs with significant antiviral activity to the water containing the larvae can significantly reduce virus levels in the larvae^29,30. Virus inactivation also has been assessed in this model, and injection of virus into the yolk sac of the embryo (6 h post-fertilization) can simplify the inoculation^31,32.

Nonhuman primates have also been evaluated for their ability to be infected by HuNoVs. Early studies indicated that rhesus monkeys^33,34 and chimpanzees³⁵ could be asymptomatically infected after oral inoculation. A later report indicated that pig-tailed macaques can be infected and develop diarrhoea after norovirus inoculation³⁶, but there have been no follow-up reports in this model. Intravenous administration of Norwalk virus (GI.1) to chimpanzees also led to infection, but the model is no longer available owing to restriction on working with chimpanzees^37,38. Further investigation using the rhesus macaque model demonstrated asymptomatic infection with several different HuNoV genotypes following oral inoculation. The study found virus antigen and RNA in jejunal epithelial cells, and prolonged faecal shedding associated with changes in the virus genetic sequence over time³⁸. The animal model is proposed as a model for the evaluation of norovirus vaccines and therapeutics.

Norovirus replication and antiviral targets

Once HuNoV binds to HBGAs, the virus enters cells and replication occurs in the cytoplasm. Details of how HuNoV replication is dependent on many of the NSPs has been predicted from studies of animal norovirus replication in a variety of cultured cells. In addition to functioning as a primer to initiate replication, norovirus VPg also functions as an mRNA cap substitute by interacting with the cap-binding eIF4G-complex during translation^39–41. Genomic replication and subgenomic RNA synthesis occur following translation and the initial round of viral protein synthesis. When all the viral proteins are synthesized and genome replication is complete, the viral RNA progeny is packaged into icosahedral particles. VP2 is likely to play a critical role in this process (Fig. 1a).

The NSPs of norovirus and other caliciviruses share sequence motifs with picornaviruses, indicating functional and structural similarities between the two families. HuNoV protease and RdRP (similarly to their picornavirus counterparts) are well characterized structurally and functionally, and have been targets for drug development because of their critical role in replication^42–44. Different classes of effective small molecule inhibitors have been discovered for the HuNoV protease and RdRP (Fig. 1b; reviewed elsewhere⁴⁴). Interestingly, studies in a recent preprint have shown that the small molecule inhibitors that target GI protease are less effective against the GII.4 proteases because of significant differences in the substrate binding pockets and their catalytic sites despite sharing the same polypeptide fold. This may pose a problem in developing inhibitors effective against both GI and GII proteases⁴⁵. Recent findings that norovirus RdRP induces liquid-like phase-separated condensates as replication hubs during infection may open avenues for discovering a new class of antivirals⁴⁶. In contrast to the protease and RdRP, structural and functional information of other NSPs encoded by ORF1 and their role in the virus replication, particularly in HuNoVs, remain less clear⁴⁷. Although the precise role in the virus replication cycle is yet to be ascertained, HuNoV p22 (NS4), which has a membrane association domain and highly conserved endoplasmic reticulum export signal motif, is implicated in Golgi apparatus disassembly and inhibition of the cellular secretory pathway⁴⁸. p41 (NS3), which has signature SF3 helicase motifs, exhibits helicase and RNA chaperone activities⁴⁹ and induces intracellular membrane alterations to form replication organelles together with p22 (NS4)⁵⁰, as well as apoptosis through both caspase-8-dependent and caspase-9-dependent pathways⁵¹.

Development of a cultivation system for human norovirus

After the Norwalk virus was discovered in 1972, numerous efforts to cultivate the virus using routine tissue culture cells and human fetal organ cultures were unsuccessful, including a large collaborative study using multiple cell cultures^52–54. By contrast, transfected HuNoV genomic RNA is infectious in several human cultured cell lines (Caco-2 and Huh-7 or Huh-7.5.1) that are otherwise nonpermissive to infection, and virus particles can be produced at low levels, but only one round of RNA replication occurs⁵⁵. In vivo, biopsies from experimentally infected volunteers and immunocompromised individuals (including transplant recipients or a chronically infected individual with common variable immunodeficiency) showed histopathology or viral antigen staining in the small intestine and little evidence of histopathology in the colon^52,56–60.

Murine noroviruses (MNVs) replicate in immune and tuft cells, and a few HuNoV strains exhibit modest replication (10-fold to 50-fold increases of RNA) in human B cell lines^61–63. Studies using unfiltered, unprocessed stool as inocula showed that bacterial surface-expressed HBGAs are important factors for successful virus replication. HBGA-expressing Enterobacter cloacae added to B cell cultures restored the infectivity of filtered HuNoV-positive stool filtrate, whereas a non-HBGA-expressing bacterium did not⁶¹. Reproducibility of this system has been challenging⁶². HuNoV also replicates in primary human B cells, with replication levels comparable to those in transformed B cells, and the virus triggers B cell activation in vitro, a property attributed to NS1/2⁶⁴. B cell-deficient patients can be infected and shed HuNoV at high titres, indicating that extensive viral replication must occur in host cells other than B lymphocytes⁶⁵. These studies led to a proposal that there is a dual tropism of HuNoV^66,67. Implications of virus uptake into immune cells in vitro to virus pathogenesis in humans remain to be determined²⁹.

Robust and reproducible cultivation system in human tissue-derived intestinal enteroids

The studies described above suggest that replication of HuNoV in vitro depends on the ability to mimic the differentiation stage of intestinal epithelial cells and perhaps the luminal microenvironment in a cell culture system. The development of nontransformed tissue stem cell-derived human intestinal enteroids (HIEs; also called tissue-derived organoids)⁶⁸ allowed the successful cultivation of HuNoVs⁶⁹ (Fig. 3a). These nontransformed, multicellular cultures are physiologically active and exhibit circadian rhythms. Thus, the HIE model closely resembles the architecture, biology and physiology of the human small intestinal epithelium and can represent the heterogeneity of the human population⁷⁰. Successful HuNoV infection of enterocytes requires differentiated cultures⁷¹, virus particles are produced from infected enterocytes and are detectable by antigen staining (Fig. 3b,c,d). Replication occurs in 2D cultures plated as monolayers or on transwells; subsequent studies have shown that bile acid and ceramide are two components in bile that allow replication of bile acid-dependent HuNoV strains^69,72 (Fig. 3e,f).

Fig. 3 |. The human intestinal enteroid culture system for human norovirus.

a, Intestinal enteroids are produced from intestinal tissue samples that contain intestinal stem cells that will grow and self-assemble into 3D cultures within several days of being cultivated in media containing Matrigel and growth factors that support proliferation of the stem cells⁶⁸. After expansion of the ‘mini-gut’ cultures, some are frozen down to create a biobank. Removal of growth factors from the media of proliferating cultures induces the stem cells to differentiate into all the multiple cell types present in the mature intestinal epithelium including absorptive (enterocytes) and secretory cells (enteroendocrine, goblet and Paneth cells) and these cultures can be plated in different formats (3D, on transwells or 2D on plastic wells for infection studies. Inside each 3D mini-gut culture, enterocytes form the brush border surrounding a single luminal compartment, which receives secretions from the other cells⁷⁰. b, Virus produced from infected human intestinal enteroids (HIEs) visualized by negative stain electron microscopy. c, Immunofluorescence detects the VP1 capsid protein in infected HIEs plated on 96 wells. d, Immunofluorescence detects the VP1 capsid protein (green), villin (red) and nuclei (4′,6-diamidino-2-phenylindole (DAPI), blue) in infected HIEs plated on transwells. e, A time course illustrates the kinetics of RNA replication and shows bile is required for GII.3 replication. f, A time course illustrates the kinetics of RNA replication and shows bile enhances GII.4 replication. Parts b–f adapted with permission from ref. 69, AAAS.

The dose of inoculum required to produce infection in 50% of inoculated tissue culture wells (TCID₅₀) varies for each virus strain (approximately 1.2 × 10² genomic equivalents (ge) for a GII.4/Sydney/2012 strain and approximately 2.0 × 10⁴ for a GII.3 strain in initial studies in the J2 HIE line)⁶⁹. These values are similar to the human infectious dose 50 observed in a controlled human infection model (CHIM) for GI.1 Norwalk virus (2.8 × 10³ ge among secretor-positive persons)⁷³ and for GII.2 Snow Mountain virus (5 × 10⁵ ge)⁷⁴. The TCID₅₀ is both viral inoculum and cell line dependent^75–77. GII.4 viruses generally replicate better than GI viruses^77–80. Advancements of the HIE system include identification of a commercial medium that enhances replication and the production of genetically modified cultures with lower innate immune responses, leading to cultivation of more viruses (currently 41 viruses from stools including 4 GI genotypes, 11 GII genotypes and 8 GII.4 variants), an increase in virus yields^77,79 and easier adoption of the system in many laboratories. To date, HuNoV replication has not been detected in enteroendocrine cells (EECs) in HIE cultures, even though these cells have been shown to contain viral genomes in immunocompromised individuals⁶⁰. Whether this reflects a low number of EECs or a lack of a specific type of EEC in HIE cultures remains to be determined. HIEs from infants⁷⁵ or genetically modified HIEs, which inducibly express neurogenin⁸¹ and express higher numbers of EECs, can be tested to further explore possible EEC HuNoV tropism.

Human norovirus replication in human induced pluripotent stem cell-derived intestinal organoids

Another human intestinal organoid (HIO) culture system made from directed differentiation of embryonic or pluripotent stem cells produces intestinal epithelial cultures surrounded by mesenchymal cells^82,83. A secretor-positive HIO (that is, an HIO with a functional fucosyltransferase 2 gene) supported limited GII.4 virus replication, and the limited replication was postulated to be due to higher numbers of immature enterocytes in the HIOs that are less permissive to infection compared with enterocytes in tissue-derived HIEs⁸⁴. Other HIO cultures report good virus propagation⁸⁵. Recently, induced pluripotent stem cell-derived HIOs, transplanted into mice to induce the maturation of the cultures^86,87, were used to create tissue-derived ‘transplanted HIEs’ that support the replication of several HuNoV strains at levels similar to those of standard HIE cultures⁷⁷.

Biological relevance of the human intestinal enteroid cultivation system

HIEs retain the phenotypic characteristics of the donor and have intestinal anatomical segment specificity. For example, HIE lines from individuals lacking a functional FUT2 gene, termed secretor-negative, are resistant to infection by many HuNoV strains, whereas GII.3 strains can infect secretor-positive and some secretor-negative HIE lines⁶⁹. Genetic modification of HIEs has confirmed that FUT2 expression is both necessary and sufficient for most HuNoV strains but it is not essential for at least one GII.3 HuNoV strain⁸⁸. There are rare exceptions of infections in secretor-negative persons by GII.4 HuNoV; this may occur because the microbiome of an individual modifies the HBGA phenotype of enterocytes⁸⁹.

HIEs also vary in their responses according to intestinal segment of origin, culture conditions and plating format (for instance 3D or 2D cultures on hard plastic versus glass or transwells) and cellular differentiation status⁹⁰. Virus replication occurs in HIE cultures from all segments of the small intestine (duodenum, jejunum and ileum), but not the colon, using cultures from these segments that have been established from the same donor as well as from different donors^77,79. These data confirm the intestinal tropism that had been predicted from the early volunteer studies⁵².

Human norovirus replication reveals unexpected strain-specific differences in entry into human intestinal enteroids

GII.4 HuNoVs replicate well in secretor-positive HIEs. By contrast, a GII.3 HuNoV strain failed to grow until exogenous bile was added to the culture medium to mimic the extracellular milieu of the small intestine⁶⁹ (Fig. 3e). Although GII.4 strains grow without bile, their replication is enhanced in the presence of bile (Fig. 3f). Hydrophobic bile acids are an active component in bile that are required for replication of GII.3 HuNoVs, and bile acid-mediated replication correlates with the hydrophobicity of the bile acid⁷². The conjugated bile acid glycochenodeoxycholic acid (GCDCA) is effective and soluble and so is now used in most HIE replication studies. Several other HuNoV strains (GI.1, GII.1, GII.3, GII.6 and GII.17) are now known to be bile acid-dependent^79,91. The effects of GCDCA on GII.3 infection have been evaluated in jejunal HIEs (Fig. 4a, left panel). The effects of GCDCA are mediated through a G protein-coupled receptor (sphingosine-1-phosphate receptor 2 (S1PR2)) and do not involve GCDCA binding to the virus, the detergent properties of GCDCA or interactions with the classic farnesoid × receptor or Takeda G protein-coupled receptor 5. GCDCA acts early during infection, inducing multiple cellular processes including the induction of endocytosis leading to GII.3 uptake followed by endosomal acidification. Ceramide is generated at the apical cell surface through the action of acid sphingomyelinase (ASM), an endosomal/lysosomal enzyme. GII.3 infection is reduced by inhibitors of S1PR2, endosomal acidification, ASM activity and the lysosomal exocytosis of ASM⁷². These data suggest a model of GII.3 infection whereby the virus binds to a cellular receptor and, in conjunction with a bile acid such as GCDCA acting through S1PR2 and ASM, ceramide-rich lipid rafts are formed containing the virus and its receptor. The virus is then internalized via endocytosis and the virus or its genome is released to the cytoplasm after endosomal acidification. GII.3 also infects duodenal and ileal HIEs in a similar GCDCA-dependent manner that is decreased by S1PR2 inhibition, suggesting that GII.3 infection throughout the small intestine occurs in a similar fashion⁹².

Fig. 4 |. Distinct entry pathways and replication of bile acid-dependent GII.3 and bile acid-independent GII.4 into human intestinal enteroids.

a, Each virus uses a different entry pathway. GII.3 requires sphingosine-1-phosphate receptor 2 (S1PR2) and bile acid to induce endocytosis (left panel), whereas GII.4 causes initial membrane injury and wound repair mechanisms to induce clathrin-independent carrier (CLIC)-mediated endocytosis (right panel). Common elements include initial binding to histoblood group antigens (HBGAs), putative receptor clustering in lipid rafts, endosomal acidification and lysosomal exocytosis that results in ceramide (Cer) appearance on the cell surface. b, The table highlights strain-specific differences in the biology of GII.3 and GII.4 viruses related to the requirement for FUT2 gene expression^69,88, cell entry^72,94, bile acid dependence^69,72, acid sphingomyelinase involvement^72,94, effects of JAK1/JAK2 inhibition¹⁰⁰, effects of knockout of innate immune factors⁹⁹ and host transcriptional responses^{99,100,231,232}. ASM, acid sphingomyelinase; Gal-3, Galectin-3; IFN, interferon; SM, sphingomyelinase; V-ATPase, vacuolar-type ATPase.

Structural studies suggest that some bile acids can interact with the major capsid protein VP1 of certain HuNoVs (GII.10 virus-like particles (VLPs) and the P domain of GII.1, GII.10 and GII.19)⁹³. The GII.1 virus can replicate in HIEs in the presence of bile acids, but the requirements for GII.10 and GII.19 viruses have not been tested to our knowledge. VLPs of GI.1 and GII.4 viruses that replicate in HIEs and the P domains of GI.1, GII.3, GII.4 and GII.17 do not bind bile acids. Bile acid binding to some HuNoV particles may have biological consequences as this is essential to allow GII.1 virus to bind to HBGAs, and GII.10 binding to HBGAs is enhanced by bile acid binding⁹³. GII.2 and GII.12 VLPs also require bile acids to bind HBGAs, although these particles have not been tested for binding by structural studies⁹¹. These studies suggest direct particle binding to bile acids may be another strain-specific property of HuNoVs but how this influences virus infectivity remains to be determined.

GII.4 entry into HIEs does not involve bile acid-induced endocytosis that is utilized by the GII.3 virus (Fig. 4a, left panel). Instead, it involves a dynamin-independent, clathrin-independent carrier (CLIC) endocytosis pathway and wound repair mechanisms⁹⁴ (Fig. 4a, right panel). GII.4 VLPs, but not the P domain alone, interact with the endosomal sorting complexes required for the transport protein ALIX on epithelial cells to induce plasma membrane wounding, which then leads to lysosomal exocytosis and additional membrane repair mechanisms. Effectors of the CLIC pathway, including cholesterol, Cdc42, ADP-ribosylation factor 1 and Galectin-3 (Gal-3), then mediate GII.4 cell entry in conjunction with endosomal acidification. Although individual components of the CLIC pathway have been associated with the entry of several other viruses into a range of different cell types, the use of all of the mediators has not been described previously for viruses. However, bacterial Shiga and cholera toxins, and certain lectins, can interact with surface glycolipids to cause narrow membrane bending, the formation of tubular endocytic pits and subsequent clathrin-independent uptake into cells. Interactions between cargo and glycosphingolipids drive the formation of deep and narrow membrane invaginations both in cells and model membranes from which CLICs are generated, similar to what has been described in vivo in mouse enterocytes⁹⁵. Similarly, the formation of tubular invaginations in giant unilamellar vesicles treated with GII.4 VLPs is consistent with the HIE membrane changes observed during GII.4 entry⁹⁶. Whether other viruses use this pathway to infect HIEs remains to be investigated.

Diversity in strain-specific abrogation of the host response

The HuNoV replication system in HIEs is limited by the lack of indefinite serial passaging of virus, suggesting a role for uncharacterized host factors in restricting replication. Innate immune responses, including the synthesis and secretion of interferons and their downstream effectors, interferon-stimulated genes (ISGs), form a first line of defence against many viral infections^97,98. Two studies of the epithelial response to HuNoV replication in HIEs using RNA sequencing show that the induced host innate immune response is primarily driven by a type III interferon response (with some induction of interferon-β (IFNβ) but not IFNα) that leads to ISG induction and restricts HuNoV replication in unexpected strain-specific ways^99,100. Interferon induction requires live virus and it appears concomitantly with replicating viral RNA. HuNoV replication is sensitive to exogenous treatment of HIEs with type I interferon (IFNα or IFNβ) or type III interferon. The intrinsic host response results in upregulation of hundreds of genes which are detectable by 10 h post-infection (hpi), but not at 6 hpi, and which continue to increase at 24 hpi (the latest time point examined)⁹⁹. Most upregulated genes are ISGs such as IFI144L, MX1, IFIT3, STAT1, IFIT1, IFITM1, ISG15, OASL and CXCL10, that were detected in both studies. Transcription enrichment analysis identified STAT1 and STAT2 binding sites as highly enriched in the promoter regions of genes whose levels of expression were significantly upregulated, suggesting that the JAK–STAT signalling pathway is activated following HuNoV infection¹⁰⁰. Pharmacological inhibition of the Janus kinase (JAK1 or JAK2) with ruxolitinib increases replication of both GII.3 and GII.4, indicating that interferon-deficient enteroids are more permissive for HuNoV replication and that the innate response limits virus replication¹⁰⁰. Attenuation of the interferon pathways by using a decoy receptor for type I and type III interferons as well as the production of five interferon-deficient HIEs that lack type I or type III interferon receptors (STAT1, MAVS and STAT1/STAT2) and exhibit attenuated ISG induction, failed to alter or enhance GII.4 replication. This unexpected result indicates that GII.4 has mechanisms to antagonize the innate immune response, and this may partially explain why these viruses predominate and are recognized as pandemic viruses. In contrast to GII.4, GII.3 replication is increased 10-fold to 100-fold in STAT1 knockout and type I interferon receptor knockout HIEs, and the enhanced replication is accompanied by reduced ISG responses. GII.3 virus can also spread in STAT1 knockout cultures that exhibit a lower TCID₅₀ when compared with the parental HIE line⁹⁹. These studies indicate that GII.3 HuNoV replication is restricted by endogenous host interferon pathways and that inhibition of STAT1 increases viral replication and spreading. The discovery of differences in sensitivity to intrinsic interferon pathways between two HuNoV strains adds to the growing list of strain-specific biological behaviours of HuNoVs⁹⁹. GII.3 replication is also increased in HIEs when interferon induction is reduced by the expression of interferon antagonists, but GII.4 replication was not reported¹⁰⁰. The mechanisms by which the ISGs and other antiviral proteins restrict virus replication remains to be determined. Several long non-coding RNAs are also upregulated in GII.4-infected HIEs but their roles in infection also remain uncharacterized⁹⁹.

Human norovirus structure — hollows and protrusions

Although, as noted earlier, HIE systems are helpful in studying many aspects of HuNoV replication, there are still challenges in these systems to successfully propagate and obtain virus in sufficient quantities for structural and biochemical studies, which have relied on VLPs produced by the co-expression of VP1 and VP2 (ref. 101). In place of authentic viruses, VLPs have been invaluable in understanding the structural, immunological and biological aspects of many strains of HuNoVs⁹⁴. The X-ray crystallographic structure of Norwalk virus (GI.1) VLPs, determined in 1999, provided an atomic-level description of the capsid architecture of a norovirus¹⁰². The structure revealed that the approximately 40 nm diameter capsid exhibits T = 3 icosahedral symmetry formed by 180 molecules of the major capsid protein VP1 (Fig. 1c). Subsequently, structure determination of a variety of animal caliciviruses, including feline calicivirus (FCV)^103,104, San Miguel Sea Lion virus¹⁰⁵ and MNV^106–108, showed similar capsid architecture.

Briefly, in these capsid structures, the 180 subunits of VP1 as three quasi-equivalent subunits — A, B and C — are assembled as 60 A–B dimers at the 5-fold axes and 30 C–C dimers at the icosahedral 2-fold axes, such that there are prominent hollows at the 5-fold and 6-fold axes. Each VP1 subunit has a modular domain organization consisting of an internal N-terminal arm (NTA) and two distinct domains termed shell (S) and protruding (P) domains, separated by a flexible hinge^102,109. The S domains of the VP1 dimers interact closely to form a smooth shell held together by the NTA, located beneath the shell, adapting to the quasi-equivalent environments of T = 3 symmetry. The VP1 dimers exhibit bent (A–B) and flat (C–C) conformations that are necessary for inducing the correct curvature of the shell (Fig. 1c). The P domain, with a distinctive fold observed only in calicivirus structures, projects from the S domain and is divided into P1 and P2 subdomains. The P1 subdomain (located closer to the S domain) is formed by the N-terminal and C-terminal regions of the P domain, whereas the P2 subdomain (which is distally located) is formed by the intervening region. The hypervariable P2 subdomain is involved in recognizing cellular receptors, as shown by structural studies of FCVs^103,104,110 and MNVs^111,112, and in recognizing the polymorphic HBGAs in the case of HuNoVs¹¹³ (Fig. 1d).

Notably, the minor capsid protein VP2, which is presumed to be present in small amounts, is not observed in any calicivirus capsid structures determined to date. However, the cryogenic electron microscopy (cryo-EM) structure of FCV in complex with its receptor fJAM-A¹⁰³ revealed that the N-terminal α-helical region of VP2 forms a portal-like assembly projecting out from one of the hollows at a 6-fold axis, which is likely to be triggered by the receptor interaction. This finding suggests a potential role for VP2 in genome uncoating and delivery, akin to the receptor-triggered extrusion of VP4 and the N terminus of VP1 implicated in membrane penetration and genome delivery in picornaviruses^114,115. Earlier studies have also suggested that VP2 in HuNoV confers stability to the capsid and that it interacts with the NTA of the VP1 (refs. 116,117). It is possible that any trigger, such as a receptor interaction, disengages VP2, leading to its extrusion to facilitate genome release.

Histoblood group antigen binding: prevalence and evolution of human norovirus

The recombinantly expressed P domain, which readily forms a dimer and adopts an identical conformation as found in the capsid structure, has been extensively used in structural and biochemical characterization of the interactions between HuNoVs and HBGAs^{102,118–120}. All these studies have shown that HuNoVs bind to HBGAs through the hypervariable P2 subdomain of VP1 (refs. 101,117–119) and exhibit strain-specific HBGA binding patterns^121–123. The HBGA binding sites in GI and GII are distinct (Fig. 1d). In contrast to GI viruses, in which each subunit of the VP1 dimer has an HBGA binding site, in GII viruses the residues from both units form the two HBGA binding sites in the dimers. Another important distinction with implications for strain predominance is that the HBGA binding in GI HuNoV is predominantly mediated by interactions with the terminal galactose (Gal). By contrast, in most GII viruses, especially in GII.4, the binding is mediated by the primary secretor-fucose (SeFuc) of the HBGA. As SeFuc is a common moiety in all HBGA families (ABH and Lewis glycans)^19,124, GII viruses can potentially bind to all of them and, thus, infect a larger population. By contrast, the HBGA binding profile is relatively restricted in GI viruses. Some epochally emergent GII.4 variants, particularly those circulating after 2002, show strong binding to di-fucosyl Lewis HBGAs because of sequence alterations in one of the solvent-accessible loops (between residues 390 and 400, referred to as loop 2) in the P2 subdomain to form a binding site for the Lewis fucose (LeFuc) of the HBGA¹²⁵. Compared with the SeFuc binding site in GII HuNoVs, which is exceptionally well conserved, the LeFuc binding site is only moderately conserved and thus could contribute to the evolution of GII.4 HuNoVs. Surrounding the highly conserved SeFuc HBGA sites, GII.4 variants exhibit several sequence changes, particularly in the loop regions, that can modulate HBGA binding affinities between the GII.4 variants. Several reports have shown how these changes alter the HBGA binding profiles and the binding affinities in the GII.4 variants^21,22,126. Although these studies provide general trends and support the notion that sequence changes influence binding affinities, they show some inconsistencies between the observed HBGA profiles for certain GII.4 variants (Table 1). Modelling these sequence variations in the context of the P domain structure has shown that they also profoundly alter the surface topography and electrostatic landscape of the P2 subdomain. This is consistent with observed strain-dependent alterations in antigenicity^120,125, which forms the basis for identification of these variations into eight distinct antigenic sites¹²⁷. A coordinated interplay between differential HBGA binding specificities, affinities and antigenic variations is proposed to allow the escape from host immunity and drive the evolution of HuNoVs¹²¹. Interestingly, there appears to be temporary stagnation in the evolution of GII.4 after the emergence of the 2012 Sydney variant. Failure to establish sustained immunity for this variant in adults may have allowed this strain to continue its dominance¹²⁸.

Capsid plasticity — resting and raised conformation states

As observed first in MNVs^{108,129–131}, recent structural studies^{107,109,132–134} of HuNoV VLPs have shown that VP1, in the context of the T = 3 icosahedral capsid, can exist in two distinct conformations: the ‘resting’ and the ‘raised’ conformations. Such a transformation without affecting the structure of either the P or S domains is facilitated by the conformationally flexible hinge. In the resting conformation, the lower portion of the P domain closely interacts with the S domain, whereas in the raised conformation, the P domain is rotated and raised above the S domain. This is driven either by the removal of stabilizing ions, as in the case of GII.4 VLPs¹⁰⁹, or by an increase in pH, as in the case of GII.3 VLPs^{107,129,130,132}. In addition to this plasticity in the capsid, VLPs, particularly GII.4, deviate from the T = 3 icosahedral symmetry and exhibit icosahedral assemblies with different T numbers. Cryo-EM structures of VLPs of some of the GII.4 variants (1994, 2006b and 2012) have shown the capsid with an increased diameter of 50 nm with T = 4 icosahedral symmetry consisting of 120 VP1 dimers that have adopted a raised conformation^132,133. Production of VLPs, with or without the co-expression of VP2, as well as possible differences in the expression systems and other factors, may influence the structural assembly of VP1. Negative-stain electron microscopy images of the native GII.4 virions, however, have shown a diameter of approximately 40 nm consistent only with T = 3 icosahedral symmetry.

The crystallographic structure of the VLP of a 2002 strain of GII.4 (HOV) produced by the co-expression of VP1 and VP2 has shown a capsid organization with all the structural features consistent with T = 3 icosahedral symmetry¹⁰⁹. This crystal structure was determined without imposing icosahedral symmetry and it also shows how hinge flexibility imparts inherent plasticity to the capsid within the confines of the icosahedral structure to adapt to crystal packing forces. A new finding in this structure is the presence of a cation at the subunit interface at the middle of the P domain dimer. This was also observed in the cryo-EM structure of this GII.4 VLP, which was concurrently determined to exclude the possibility that bound ion in the crystal structure is a crystallization artefact. Removal of this ion resulted in significant conformational changes consistent with transitioning from an ion-bound resting state to the raised state. As in MNV and FCV, recent studies of GII.4 HuNoV have implicated the resting and raised conformations in entry and antibody neutralization mechanisms^109,135 (Fig. 5).

Fig. 5 |. Mechanisms of neutralization of human norovirus by antibodies.

a, Neutralization of GI.1 by the human antibody 5I2 (yellow) by directly competing with the histoblood group antigen (HBGA) binding site in the protruding (P) domain dimer (grey). b, Neutralization of GI.1 by the human antibody 10E9 (purple) by directly competing with the HBGA binding site in the P domain dimer (grey). c, Human antibody A1431 (blue) binds to an epitope slightly away from the HBGA binding site in the GII.4 P domain dimer (grey) and neutralizes GII.4 by clumping or crosslinking particles. d, Resting and raised conformations of the P domain (shell (S), P1, and P2 subdomains are shown in blue, red, and yellow, respectively). e, Fragment antigen-binding region (Fab) of the human antibody NORO-320 (yellow) targets capsid plasticity by binding an epitope in the GII.4 P domain (grey) that is exposed only in the raised conformation and probably neutralizes the virus by blocking co-receptor interactions or VP2 externalization. f, The llama nanobody M4 (green) also binds to an epitope exposed only when the P domain (grey) transits to the raised confirmation and neutralizes by compromising capsid integrity.

Antibody neutralization mechanisms — targeting histoblood group antigen binding and capsid plasticity

Mechanisms by which a neutralizing antibody ablates virus infectivity typically involve inhibiting interactions critical for the cell entry process^136–138. This includes: (1) clumping the particles owing to the bivalent binding of neutralizing antibody to particles, thereby disabling their ability to interact with cellular receptors; (2) sterically blocking access to the receptor by binding close to the receptor binding site; (3) inhibiting a conformational change in the capsid or the capsid protein necessary for receptor interactions; and (4) inducing disassembly of the capsid or compromising the capsid integrity. For HuNoVs, several mouse¹³⁹ and human-derived antibodies^140–142, and llama-derived single-chain antibodies (nanobodies) have been studied for their ability to neutralize virus infection^143,144. Prior to having a cell culture system for HuNoVs, surrogate HBGA blockade assays were routinely used instead of neutralization assays^73,126,141. With the establishment of HIE cell culture systems⁶⁹, antibodies can now be evaluated directly for their ability to neutralize HuNoVs belonging to different genogroups and genotypes. One antibody, 5I2, isolated from peripheral blood mononuclear cells of individuals infected with GI.1 HuNoV¹⁴⁵ shows potent HBGA blockade activity and is highly specific to GI.1 (ref. 142). Structural studies of the Fab of 5I2 in complex with the GI.1 P domain show that the antibody binds very close to the HBGA binding site, indicating that the mechanism of neutralization is by sterically blocking the HBGA (Fig. 5a). Further, the epitope that is recognized is only conserved in the GI.1 and not in other GI HuNoVs, thereby providing a clear understanding of how sequence and structural changes mediate escape from Fab 5I2 neutralization in other genotypes¹⁴².

For GII.4 HuNoVs, however, neutralization by sterically blocking HBGA binding may be effective for a particular variant¹⁴⁶ (Fig. 5b), but this mechanism may not be effective for multiple variants because of the significant sequence variability surrounding the HBGA binding site. Recent studies have uncovered antibody neutralization mechanisms that do not involve direct HBGA blockade but instead target the ability of VP1 to exist in resting and raised conformations (Fig. 5d). Although the IgG and IgA forms of human NORO-320 inhibit HBGA binding and neutralize GII.4 HuNoV infectivity, the Fab of this antibody does not block HBGA binding, yet it neutralizes infectivity as effectively as the IgG or IgA forms¹⁴⁰. The crystal structure of NORO-320 Fab in complex with GII.4 P domain has shown that the antibody recognizes an epitope distant from the HBGA binding site close to the bottom end of the P domain¹⁴⁰ (Fig. 5e). This epitope is highly conserved among the GII HuNoVs, accounting for the broad specificity of this antibody. In the context of the capsid structure, Fab-320 binding causes severe steric clashes with the neighbouring subunits, but these clashes are relieved when VP1 is in the raised conformation, suggesting that with Fab-320 binding the VP1 is locked into the raised conformation. When modelled into the capsid structures, the binding of the Fab-320 to the raised conformation has shown that it points towards the hollow region of the capsid, masking the S domain to some extent. The neutralization mechanism, therefore, is likely to be one or a combination of the following: (1) restricting the capsid dynamics that may be essential for optimizing receptor interactions; (2) blocking access to cellular ligands such as ALIX, Gal-3 and TSG101, some of which bind to the S domain⁹⁴; and (3) hindering the extrusion of VP2 portal potentially that is required for genome delivery similar to FCV¹⁰³. However, by contrast, the IgG form of NORO-320 neutralizes infection by clumping the particles through bivalent binding, as is evident from dynamic light scattering (DLS) experiments of GII.4 VLPs with NORO-320. A similar mechanism, by clumping or crosslinking particles, has been implicated for another human antibody, A1431, which also binds to an epitope away from the HBGA binding site¹⁴¹ (Fig. 5c).

More recently, yet another mechanism of neutralizing HuNoV infectivity was discovered with a norovirus-specific llama-derived nanobody M4 (ref. 135). This targets a highly conserved region distant from the HBGA binding site close to the S–P interacting region (Fig. 5f) and neutralizes multiple GII.4 variants with high potency in HIEs. As with the epitope for NORO-320, this epitope is accessible only when the P domain is in the raised conformation. DLS experiments with GII.4 VLPs have shown that M4 effectively disassembles the particles, suggesting that the neutralization mechanism involves accessing the epitope by altering the conformational dynamics of the capsid and triggering the disassembly of the capsid. Another nanobody, Nano-25, is suggested to neutralize GII.4 infection by a similar mechanism¹⁴⁴. A puzzling observation from M4 studies is that although this nanobody binds the GII.3 P domain and dissembles the GII.3 VLPs as effectively as it does GII.4 VLPs, M4 fails to neutralize GII.3 virus, suggesting that the GII.3 virus is more stable than the GII.3 VLP and, by contrast, GII.4 virus and VLPs do not differ in their stabilities and are equally vulnerable to M4 binding. Overall, this indicates that the genome, along with VP2 in the GII.3 virion, probably contributes differentially to the increased stability of the capsid, underscoring how capsid stability and plasticity variations between genotypes may influence the mechanism of neutralization¹³⁵.

Vaccines in clinical development

There are no vaccines currently licensed for the prevention of norovirus infection or illness, but several vaccines are in clinical development (Table 2), and others are in preclinical development. All vaccines in clinical development are based upon targeting the major capsid protein, VP1, using three different strategies for protein expression. The most frequent approach is the expression of VP1 as a recombinant protein that form VLPs. Protein production is achieved using baculovirus, yeast (Pichia pastoris and Hansenula polymorpha) and plant (Nicotiana benthamiana) expression systems, and in most cases the protein is delivered intramuscularly in combination with an aluminium-based adjuvant. The second approach is the use of a recombinant, replication-incompetent adenovirus-vectored vaccine delivered orally. The vaccine abortively infects intestinal epithelial cells, leading to local expression and both a local and a systemic immune response. The third approach delivers mRNA encoding the VP1 protein encapsulated in a lipid nanoparticle, which when delivered intramuscularly also leads to expression of the viral capsid protein within the vaccine recipient.

Table 2 |.

VP1-based norovirus vaccine candidates in clinical development

Company	Vaccine platform (expression system)	Adjuvant	Administration route	Genotypes	Development stage	Refs.
Anhui Zhifei Longcom Biologic Pharmacy Co., Ltd	Protein (Pichia pastoris)	Aluminium hydroxide	Intramuscular	GI.1, GII.3, GII.4, GII.17	Phase I/II	NCT04563533 (ref. 193) and NCT06524947 (ref. 194)
Chengdu Kanghua Biological products Co., Ltd	Protein (Pichia pastoris)	Aluminium hydroxide	Intramuscular	GI.1, GII.2, GII.3, GII.4, GII.6, GII.17	Phase I planned	NCT05805618 (ref. 195)
Hillevax, Inc. (nee Takeda Vaccines, Inc., Ligocyte Pharmaceuticals, Inc.)	Protein (Baculovirus)	Aluminium hydroxide	Intramuscular	GI.1, GII.4	Phase II	NCT01168401 (refs. 148,196–198), NCT01609257 (refs. 150,199–201), NCT02038907 (refs. 202,203), NCT02142504 (refs. 147.204,205), NCT02153112 (refs. 153,154,206), NCT02475278 (refs. 149,207,208), NCT02661490 (ref. 209), NCT02669121 (refs. 151,210,211), NCT03039790 (ref. 212), NCT05281094 (ref. 206), NCT05972733 (ref. 213) and NCT06007781 (ref. 214)
Icon Genetics GmbH	Protein (Nicotiana benthamiana)	None	Intramuscular	GI.4, GII.4	Phase I	NCT05508178 (refs. 160,161,215)
Moderna	mRNA lipid nanoparticle	None	Intramuscular	GI.3, GII.3, GII.4	Phase I	NCT05992935 (ref. 216) and NCT06592794 (ref. 217)
National Vaccine and Serum Institute, China	Protein (Hansenula polymorpha)	Aluminium hydroxide	Intramuscular	GI.1, GII.4	Phase III.	NCT04188691 (ref. 218), NCT04941261 (ref. 219) and NCT05916326 (ref. 220)
Vaxart, Inc.	Recombinant virus vector (replication-deficient, adenovirus type 5)	Double-stranded RNA hairpin	Oral	GI.1, GII.4	Phase II	NCT02868073 (refs. 221,222), NCT03125473 (ref. 223), NCT03897309 (ref. 224), NCT04854746 (ref. 225), NCT04875676 (ref. 226), NCT05212168 (refs. 158,227), NCT05213728 (ref. 228) and NCT05626803 (ref. 229)

The vaccine that has had the greatest clinical evaluation to date is HIL-214, an aluminium-adjuvanted, bivalent VLP vaccine containing the GI.1 Norwalk virus strain and a consensus GII.4 VLP derived from what at the time were considered three distinct variants (2002/Lanzhou (now reclassified as the 2006a/Yerseke-like strain), 2006a/Yerseke and 2006b/Den Haag). In adults, a single dose elicits a rapid anamnestic HBGA-blocking antibody response that is detectable as early as 4 days after vaccination and peaks within 10 days after vaccination¹⁴⁷. HBGA-blocking antibody responses are similar in older adults when compared with younger adults¹⁴⁸ and mucosal IgA responses also occur¹⁴⁹. A proof-of-concept CHIM study using a GII.4/2002 strain as the challenge virus was too underpowered to demonstrate a significant reduction in norovirus-associated illness¹⁵⁰, but a field efficacy study demonstrated vaccine efficacy of 62% (95% confidence interval, 21–82%) in preventing moderate-to-severe norovirus-associated gastroenteritis due to any strain and significant protection against a heterotypic GII.2 strain¹⁵¹. HBGA-blocking antibody responses to heterologous strains have also been observed, although to a lower level than to vaccine strains, providing a potential explanation for the observed heterotypic protection¹⁵². The vaccine is also immunogenic in children 1–8 years of age and in children 6–12 months old^153,154. Younger children (<4 years) had increased HBGA-blocking antibody responses to both antigens when a second dose was administered 1 month after the first. Although older children (4–8 years) also had higher antibody levels to the GI.1 antigen after a second dose, the second vaccine dose less consistently increased their GII.4 antibody responses¹⁵³. A two-dose, placebo-controlled, field efficacy trial in children that is administering the bivalent VLP vaccine at 5 months and 6–7 months of age is underway (NCT05281094)¹⁵⁵.

Another vaccine evaluated in several clinical trials is a recombinant, non-replicating, adenovirus type 5-vectored vaccine administered orally as a tablet. Both GI.1 and GII.4 vaccine tablets have been developed and evaluated¹⁵⁶. The initial clinical study with the GI.1 construct found it to be safe and well tolerated, with both serum HBGA-blocking and mucosal antibody responses being induced¹⁵⁷. Other studies, reported in abstracts or media releases, have shown that immune responses in older adults are at a level similar to those of younger adults. Co-administration of the GI.1 and GII.4 vaccines induced immune responses without evidence of interference and the GI.1 vaccine reduced homologous norovirus infection by 21% in a CHIM, a reduction similar to one reported previously for an intranasally administered GI.1 VLP vaccine^158,159.

Other vaccine candidates are also in clinical development, but there is much less information available about their safety and immunogenicity. All vaccines under development are multivalent and contain from two to six genotypes. A GII.4 strain is included in all vaccine candidates, but the GI component varies and additional GII strains are included in some (Table 2). An unadjuvanted bivalent VLP vaccine containing GI.4 and GII.4 capsid proteins was safe and immunogenic, inducing HBGA-blocking antibody responses to the vaccine strains as well as cross-reactive IgG responses to several heterotypic (non-vaccine) strains^160,161. A phase III clinical trial is planned for a yeast-expressed bivalent VLP vaccine containing GI.1 and GII.4 strains (NCT05916326)¹⁶², but no information is yet available on the results of the phase I and II trials of the vaccine.

Future perspectives

This is an exciting time for HuNoV research. Future studies are expected to discover the cell receptor or receptors for HuNoVs, which are predicted to differ between strains. New intestinal biology is being discovered from studies on HuNoV entry into and characterization of the glycolipids and glycoproteins on HIEs^163,164. This should allow simpler culture systems to be developed. Understanding mechanisms of host restriction is also expected to lead to improved culture systems and insight into why some HuNoV strains persist globally. Both epidemiological and vaccine studies have demonstrated that induction of protective immunity against norovirus-associated illness is possible. Unanswered questions that need to be assessed in clinical trials include the breadth of immunity induced by vaccination against single genotypes, the importance of priming and past infection history on induction of immunity to non-vaccine strains, the optimal number of strains to include in a vaccine, the duration of immunity after vaccination and the mechanism or mechanisms of protective immunity¹⁶⁵. Vaccines have focused on VP1 as an immunogen, but antibody responses to NSPs also occur^166,167, and the potential of these as targets for vaccination remains to be assessed. The recent revelation from the structural¹⁰⁹ and antibody neutralization studies¹⁴⁰ that GII.4 HuNoV capsid exhibits inherent plasticity that the antibodies can target for neutralization suggests that capsid dynamics are likely to play a critical role in cell entry. These observations raise an important question: should this dynamic feature of the capsid be considered when developing capsid-based vaccine strategies? The discovery that a single nanobody neutralizes multiple GII.4 epochal variants¹³⁵ suggests that nanobodies, given their small size, high stability, strong antigen-binding affinity and water solubility, might be tested to be used alone as an immunotherapeutic or in conjunction with vaccines to improve therapeutic efficiency. The demonstration that immunoglobulin preparations with neutralizing activity against the infecting norovirus successfully eradicated infection in a chronically infected, immunocompromised patient supports the rationale for developing immunotherapeutics¹⁶⁸.

Finally, small molecule antivirals and associated therapies are being developed and evaluated in HIEs and other replication systems^29,76,169. Currently, these efforts focus primarily on HuNoV RdRPs and proteases, and whether they will lead to the discovery of broadly active compounds remains to be seen. In the case of HuNoV proteases, structural differences observed between GI and GII proteases¹⁷⁰ may have to be dealt with for such an endeavour. Pharmacological targeting of the liquid-like condensates induced by HuNoV RdRP⁴⁶ as putative replication hubs can also be an attractive possibility in the future, as suggested for other viral systems, cancer and neurodegenerative diseases^171,172. Further understanding of the structure and function of other HuNoV NSPs, notably p41 with its demonstrated helicase activity⁴⁹, could lead to new targets for drug development.