Genetic and antigenic diversity among noroviruses

Norovirus (NoV) strains are a leading cause of gastroenteritis worldwide and cause outbreaks in various epidemiological settings including hospitals, cruise ships, schools and restaurants (Beuret et al., 2003; Inouye et al., 2000; Johansson et al., 2002; Kapikian et al., 1972; McEvoy et al., 1996; McIntyre et al., 2002; Russo et al., 1997). Numerous molecular epidemiological studies have revealed a global distribution of these viruses (Nakata et al., 1998; Noel et al., 1999; White et al., 2002). Transmission routes include food-borne, person-to-person contact and environmental contamination. Human NoV can be divided into two genetically distinct genogroups, GI and GII. Recently, NoV strains were subdivided into at least 14 GI and 17 GII genotypes (GI/114 and GII/117) (Kageyama et al., 2004). These viruses cannot be grown in culture and their antigenic relationships are not completely understood. Nevertheless, expression of the major capsid protein (VP1), which usually results in the formation of virus-like particles (VLPs) that are morphologically similar to the native virus, has permitted a better understanding of antigenicity in these viruses (Hansman et al., 2004). Two types of assay have been used to examine cross-reactivity among these VLPs: antibody ELISA and antigen ELISA (Gray et al., 1993; Jiang et al., 1995a, b; Kageyama et al., 2004; Kamata et al., 2005; Kobayashi et al., 2000a, b, c). The antibody ELISA is broadly reactive, but the antigen ELISA is highly specific, only detecting strains that are closely related (>95 % identity in the RNA polymerase region). However, detailed information on the cross-reactivity among many of the genetically distinct NoV strains is limited.

NoVs are small round viruses approximately 38 nm in diameter and possess a single-stranded, positive-sense RNA genome of 7·57·7 kb. The NoV genome contains three open reading frames (ORFs). ORF1 encodes non-structural proteins, including the RNA-dependent RNA polymerase, ORF2 encodes VP1 and ORF3 encodes a minor capsid protein (VP2) (Jiang et al., 1990). Cryo-electron microscopy (cryo-EM) and X-ray crystallography analyses of NoV VLPs have determined the shell and protruding domains (subdomains P1-1, P1-2 and P2) of the capsid protein (Prasad et al., 1999). Chen et al. (2004) also described strictly and moderately conserved amino acid residues in the capsid protein among the four genera in the family Caliciviridae.

The aim of this study was to analyse cross-reactivity among 26 different NoV VLPs in order to understand NoV genetic and antigenic relationships in more detail. An antibody ELISA using polyclonal antisera raised against the VLPs was used to determine cross-reactivities. Our results found broad-range cross-reactivities with antisera raised against a number of distinct NoV strains.

Specimens.
Positive stool specimens were collected from a number of different sources (see GenBank accession numbers) and RNA was extracted as described previously (Katayama et al., 2002). PCR-generated amplicons or plasmids were excised from the gel and purified by using the QIAquick Gel Extraction kit and Plasmid Purification kit (Qiagen). Nucleotide sequences were prepared with the BigDye Terminator Cycle Sequence kit (version 3.1) (Applied Biosystems) and determined by using the ABI 3100 Avant sequencer (Perkin-Elmer ABI). Nucleotide sequences were aligned with CLUSTAL_X and distances were calculated using Kimura's two-parameter method. Phylogenetic trees with bootstrap analysis from 1000 replicas were generated by the neighbour-joining method as described previously (Kageyama et al., 2004). Amino acid VP1 secondary structure predictions were made using the PSIPRED secondary structure prediction software (McGuffin et al., 2000).

Expression of VLPs.
Previously, we expressed four GI NoV strains: GI/1 (strain SeV), GI/2 (strain 258), GI/3 (strain 645) and GI/4 (strain CV), and nine GII NoV strains: GII/3 (strain 809), GII/4 (strain 104), GII/5 (strain 754), GII/6 (strain 7k), GII/7 (strain 10-25), GII/10 (strain 026), GII/12 (strains CHV and 9912-02F; in this study 9912-02F was termed Hiro) and GII/14 (strain 47) (Hansman et al., 2004; Kamata et al., 2005; Kitamoto et al., 2002; Kobayashi et al., 2000a, b, c). Dr Kim Green provided us with the Hawaii virus recombinant baculovirus GII/1 (strain HV) (Green et al., 1997). In this study, we expressed an additional 12 VLPs: GI/8 strain WUG1 (using primers G1SKF and TX30SXN; see Table 1 for primer sequences); GI/11 strain #8 (primers G1SKF and TX30SXN); GII/1 strain 485 (primers G2/F3 and G2R0); GII/2 strain Ina (primers G2Fb and G2R04); GII/3 strain 18-3 (primers G2/F3 and MVR1); GII/3 strain 1152 (primers G2F2 and G2R03); GII/3 strain 336 [primers G2/F3 and Oligo-(dT)₃₃]; GII/3 strain Sh5 (primers G2F02 and G2R03); GII/6 strain 445 [primers G2/F3 and Oligo-(dT)₃₃]; GII/8 strains Mc24 and U25 (primers G2SKF and TX30SXN); and GII/17 strain Alph23 (primers NAL13 and N235R). For expression of the recombinant VP1 in insect cells, all of the NoV constructs were designed to begin from the predicted VP1 AUG start codon. For six of the 12 constructs, the VP2 and poly(A) sequences were included by using either the TX30SXN or Oligo-(dT)₃₃ reverse primers (strains WUG1, #8, 336, 445 and Mc24). One construct (strain 485) excluded the poly(A) sequence, whereas the remaining five constructs excluded both the VP2 and poly(A) sequences (strains Alph23, Sh5, 1152 and 18-3 and Ina). Four constructs that were amplified with the TX30SXN reverse primer were expressed using the Gateway expression system (strains WUG1, #8, U25 and Mc24) (Hansman et al., 2004), whilst the other eight constructs were expressed in a baculovirus expression system as described previously (Kamata et al., 2005).

Table 1. Primer sequences used for expression of VLPs

VLP purification and electron microscopy (EM).
Recombinant baculovirus shuttle vectors (bacmids) were transfected into Sf9 cells using Effectene according to the manufacturer's instructions (Qiagen). Sf9 cells were incubated for 56 days at 26 °C, after which the culture medium was clarified by low-speed centrifugation and the supernatant was stored as the seed baculovirus. Tn5 cells were infected with the seed baculovirus at 26 °C and harvested 56 days post-infection. VLPs secreted into the cell medium were separated from cells by low-speed centrifugation, concentrated by ultracentrifugation at 30 000 r.p.m. at 4 °C for 2 h (Beckman SW-32 rotor) and then resuspended in 100 µl Grace's medium. VLPs were purified by CsCl equilibrium gradient ultracentrifugation at 45 000 r.p.m. at 15 °C for 18 h (Beckman SW-55 rotor). The harvested culture medium was examined for VLPs by negative-staining EM. Briefly, the samples (diluted 1 : 10 in distilled water) were applied to a carbon-coated 300-mesh EM grid and stained with 2 % uranyl acetate (pH 4). Grids were examined under an electron microscope (JEM-1220; JEOL) operating at 80 kV.

Antibody production and ELISA.
Hyperimmune sera to newly developed VLPs were prepared in rabbits. The first subcutaneous injection was performed with purified VLPs (between 10 and 500 µg) in Freund's complete adjuvant. After 3 weeks, the animals received two or three booster injections of the same amount of VLPs in Freund's incomplete adjuvant at intervals of 1 week. The animals were bled 1 week after the last booster injection. An antibody ELISA was used to compare cross-reactivities among the VLPs. Then wells of 96-well microtitre plates (Maxisorp; Nunc) were each coated with 100 µl purified VLPs (1·0 µg ml¹ in carbonate/bicarbonate buffer, pH 9·6; Sigma) and incubated overnight at 4 °C. The wells were washed twice with PBS containing 0·1 % (v/v) Tween 20 (PBS-T) and then blocked with PBS containing 5 % (w/v) skimmed milk (PBS-SM) for 1 h at room temperature. After the wells had been washed twice with PBS-T, 100 µl twofold serially diluted hyperimmune rabbit antiserum from a starting dilution of 1 : 2000 in PBS-T-SM was added to each well and the plates were incubated for 1 h at 37 °C. The wells were washed six times with PBS-T and 100 µl horseradish peroxidase-conjugated anti-rabbit IgG (1 : 1000 dilution in PBS-T-SM) was added to each well. The plates were incubated for 1 h at 37 °C. The wells were washed six times with PBS-T and 100 µl o-phenylenediamine substrate and H₂O₂ was added to each well. The plates were left in the dark for 30 min at room temperature. The reaction was stopped by the addition of 50 µl 1 M H₂SO₄ to each well and the absorbance was measured at 492 nm. ELISA titres were expressed as the reciprocal of the highest dilution of antiserum giving a value of A₄₉₂>0·2.

Sequence analysis
Nucleotide and amino acid sequences were aligned using CLUSTAL_X and distances were calculated using Kimura's two-parameter method. We divided the 26 strains used in this study into six GI and 12 GII genotypes using partial N-terminal VP1 nucleotide sequences (data not shown). These genotypes were maintained when we grouped the complete VP1 amino acid sequences (Fig. 1). Mc24 was excluded from the amino acid analysis since the full-length capsid sequence was unavailable. Nevertheless, using the partial N-terminal VP1 nucleotide sequence (GenBank accession no. AY237414), Mc24 clustered in GII/8 and was closely related to strain U25. Of the recently described NoV strains (Kageyama et al., 2004), the GI and GII genotypes used in this study represented 43 % (6/14) and 76 % (13/17), respectively. For several GII genotypes, we used two or more VLPs in order to clarify antigenicity, including GII/1 (strains HV and 485), GII/3 (strains 809, Sh5, 336, 1152 and 18-3), GII/6 (strains 7k and 445) and GII/12 (strains CHV and Hiro).

(34K): Fig. 1. Phylogenetic tree of NoV sequences examined in this study (shown in bold italic). NoV amino acid sequences were constructed using the entire VP1 sequence (the complete sequence for Mc24 was unavailable). Numbers on branches indicate bootstrap values for the clusters; values of 950 or higher were considered statistically significant for the grouping (Katayama et al., 2002). Reference sequences have been reported previously (Kageyama et al., 2004).

Expression of VP1
EM confirmed that all strains formed VLPs with morphological features similar to native NoV (Fig. 2), despite the fact that different constructs and expression systems were used to express the recombinant VP1. The VLPs retained their morphological features, even when stored for >6 months at 20 °C (data not shown).

(159K): Fig. 2. EM images of CsCl-purified NoV VLPs negatively stained with 2 % uranyl acetate (pH 4). (a) Strain 7k, (b) strain 485, (c) strain 445 and (d) strain 645. Bar, 100 nm.

Homologous antigenic analysis
An antibody ELISA, which uses polyclonal antiserum raised against one type of VLP, was used to determine cross-reactivity among the 26 different NoV VLPs. ELISA titres were expressed as the reciprocal of the highest dilution of antiserum giving a value of A₄₉₂>0·2. A negative control (baculovirus-infected Tn5 cell lysate) was used for all experiments and found to be negligible [i.e. A₄₉₂<0·05, using up to 0·5 µg (ml lysate)¹]. Antisera reacted strongly against homologous VLPs, with titres ranging from 102 400 to 1 638 400 (Table 2).

Table 2. Antibody ELISA cross-reactivity among the 26 different VLPs and antisera Titres are expressed as x102.

Heterologous antigenic analysis
We observed a number of novel cross-reactivities among different genotypes. For example, Fig. 3(a) shows the strong cross-reactivity of GI/11 #8 antiserum with both GII/6 7k and GII/6 445 VLPs. We found that GI/11 #8 antiserum cross-reacted with these GII/6 VLPs at titres of 102 400, which was equal to the homologous VLP titre. We also found that GII/1 HV antiserum cross-reacted strongly (i.e. equal to the homologous VLP titre) against GII/6 7k VLPs (titre 204 800) and moderately strongly (i.e. twofold lower than the homologous VLP titre) against GII/6 445 VLPs (titre 102 400) (Fig. 3b and Table 2). We observed several antisera that cross-reacted moderately against different genotypes (i.e. fourfold lower than the homologous VLP titres). For example, GI/11 #8 antiserum cross-reacted moderately with GI/4, GI/8, GII/1, GII/2, GII/3, GII/4, GII/5, GII/7, GII/10, GII/12 and GII/17 VLPs (Fig. 3a and Table 2). GII/1 HV antiserum also cross-reacted moderately with several different genotypes, including GII/1 (strain 485), GII/3, GII/10 and GII/12 (Fig. 3b and Table 2). GII/1 485 antiserum cross-reacted moderately only with GII/1 HV VLPs; GII/6 7k antiserum cross-reacted moderately with GI/11 VLPs; GII/10 026 antiserum cross-reacted moderately with several different genotypes, including GII/1, GII/5, GII/7 and GII/12; and GII/12 CHV antiserum cross-reacted moderately with GII/1 and GII/10 VLPs (Table 2).

(22K): Fig. 3. Antibody ELISAs for NoV VLPs. Wells were coated with 100 µl purified VLPs. After washing, hyperimmune rabbit antiserum raised against the VLPs was used to detect antigens. Antisera were diluted twofold in PBS-T-SM from a starting dilution as indicated (dilutions x102). The arrows indicate the endpoint. (a) GI/11 #8 antiserum cross-reacts strongly with GII/6 7k and445 VLPs. (b) GII/1 HV antiserum cross-reacts strongly with GII/6 7k and moderately strongly with 445 VLPs. (c) GII/3 809 antiserum cross-reacts weakly with GII/3 1152 VLPs.

Genotype-specific reactivities
We observed weak cross-reactivities among different genotypes (i.e. greater than eightfold dilutions). We found that GI/1, GI/2, GI/3, GI/4 and GI/8 antisera cross-reacted weakly with other genotypes (Table 2). We also observed similar weak cross-reactivities with GII/1 (strain 485), GII/2, GII/3 (all five strains), GII/4, GII/5, GII/6 (strain 445), GII/7, GII/8 (both strains), GII/14 and GII/17 antisera. For several GII genotypes, only one type of antiserum was produced, but for five GII genotypes, we produced two or more different antisera against VLPs belonging to the same genotype (Table 2). Some interesting results were observed. For example, the antigenicities of HV and 485 were considerably different, despite the fact that both strains belong to GII/1 and share approximately 94 % amino acid identity. As shown in Fig. 3(b), HV antiserum cross-reacted strongly with GII/6 VLPs, but 485 antiserum showed little cross-reactivity with these GII/6 VLPs (Table 2). This unusual cross-reactivity pattern was also observed with other antisera. For example, for GII/6, we found that 7k antiserum cross-reacted moderately with GI/11 #8 VLPs, whereas 445 antiserum cross-reacted weakly (i.e. 32-fold lower than the homologous VLP titre; Table 2). More uniquely, we found that GII/3 1152 antiserum, which was genotype-specific, had unusual antigenicity. We found that three different GII/3 antisera (strains 809, Sh5 and 18-3) cross-reacted weakly with 1152 VLPs (i.e. eightfold lower than the homologous VLP titre; Table 2 and Fig. 3c). This unusual cross-reactivity result was not evident with the other genotypes in which we produced two different antisera (i.e. GII/1, GII/6, GII/8 and GII/12; see Table 2).

Amino acid alignment and secondary structure prediction
An alignment of 25 VP1 amino acid sequences used in this study (Mc24 complete capsid was unavailable) revealed that the N-terminal region (aa 149), shell domain (aa 50225) and P1-1 domain (aa 226278) had more conserved short continuous residues than the P2 domain (aa 279405), P1-2 domain (aa 406520) and C-terminal region (Fig. 4). These continuous residues may be the reason for the cross-reactivity among different genotypes, in particular, the strong cross-reactivity of #8 antiserum against GII/6 VLPs (Fig. 3a). However, this does not explain why GII/3 1152 VLPs cross-reacted weakly with GII/3 809, Sh5 and 18-3 antisera (i.e. eightfold lower than the homologous VLP titre) and moderately against GII/3 336 antiserum (i.e. fourfold lower than the homologous VLP titre). An amino acid alignment of these five GII/3 VP1 sequences showed no unusual insertions, deletions or recombination sites; in fact, the shell domain was highly conserved among the GII/3 sequences (data not shown). However, the 1152 VP1 sequence had three unique amino acid residues (Thr-285, Ile-372 and Ser-508) when compared with the other four GII/3 VP1 sequences. The first two residues were located in the outermost region of the P2 domain, whilst the third residue was located within the P1 domain (data not shown). We used the PSIPRED secondary structure prediction software (McGuffin et al., 2000) to compare the five GII/3 VP1 structures. We found that the predicted VP1 structures for 809, Sh5, 18-3 and 336 had a helix between residues 219 and 237, whereas this helix structure was absent for 1152 (Fig 5). These data suggested that the helix structure may play an important role in influencing the cross-reactivity among the GII/3 VLPs and antisera.

(140K): Fig. 4. Amino acid alignment of VP1 sequences of the NoV sequences examined in this study. The following regions are indicated above the sequences (in order): N-terminal region (line); shell domain (filled box), P1-1 domain (open box); P2 domain (XXX); P1-2 domain (open box) and C-terminal region (line) (Chen et al., 2004). Asterisks indicate conserved amino acids.

(71K): Fig. 5. Schematic representations of the complete predicted secondary structures of VP1 of NoV (GII/3) strains 1152, 18-3, 336, 809 and Sh5. The level of confidence of prediction (Conf) is shown on the first line, where a tall box represents a high confidence of prediction and a short box represents a low confidence of prediction. The predicted secondary structure (Pred) is shown on the second line, where a helix is represented by a cylinder, a β-strand by an arrow and a coil by a line. The third line also shows the predicted secondary structure (Pred), where H represents a helix, E a β-strand and C a coil. The amino acid sequence (AA) is shown on the bottom line. The boxed regions in 18-3, 336, 809 and Sh5 VP1 indicate a helix structure that is absent in 1152 VP1. The amino acid residues that are unique to the 1152 sequence when compared with the other four GII/3 sequences are indicated by arrows.

In this study, we analysed NoV capsid-based grouping and cross-reactivity among 26 different VLPs belonging to six GI and 12 GII genotypes. Using an antibody ELISA, we found that the antisera reacted strongly against the homologous VLPs with titres ranging from 102 400 to 1 638 400. As summarized in Table 3, we also observed strong, moderately strong and moderate cross-reactivities among different genotypes (i.e. equal to the homologous VLP titre and to twofold and fourfold dilutions, respectively). For example, GI/11 antiserum had a broad range of cross-reactivities, detecting two GI genotypes (GI/4 and GI/8) and 10 GII genotypes (GII/17, GII/10, GII/12 and GII/17); GII/1 antiserum (strain HV) had a broad range of cross-reactivities, detecting four GII genotypes (GII/3, GII/6, GII/10 and GII/12); GII/10 antisera also had a broad range of cross-reactivities, detecting four GII genotypes (GII/1, GII/5, GII/7 and GII/12); GII/6 antiserum detected GI/11 VLPs; and GII/12 antiserum (strain CHV) detected GII/1 and GII/10 VLPs.

Table 3. Summary of cross-reactivities among VLPs Each letter represents one strain. For example, GII/1 antiserum cross-reacted with two GII/6 strains (Aand B), where A, strongly (i.e. identical to the homologous VLP titre), B, moderately strongly (i.e.twofold lower than the homologous VLP titre), and C, moderately (i.e. fourfold lower than the homologous VLP titre). For simplicity, we have excluded the homologous reactivities.

Although antigen ELISAs are generally broadly reactive (Jiang et al., 2000), this is the first report of a GI (strain #8) polyclonal antiserum cross-reacting strongly with other GII genotypes and the first report of a GII (strain HV) polyclonal antiserum cross-reacting strongly with other GII genotypes (Jiang et al., 2002; Kamata et al., 2005; Kitamoto et al., 2002). These broad-range cross-reactivities may be due to unfolded VLPs on the microtitre plates at the high pH used (carbonate/bicarbonate buffer, pH 9·6) (White et al., 1997). However, we have not found such broad-range cross-reactivities in any of our other studies (Kamata et al., 2005). Conserved continuous residues in the shell and/or P1-1 domains may be the reason for these cross-reactivities against different genotypes (Fig. 4 and Table 2). However, we found that several antisera were genotype-specific, indicating that VLPs have unique epitopes.

Interestingly, we found that four types of GII/3 antisera (strains 809, Sh5, 18-3 and 336) cross-reacted moderately to weakly against GII/3 1152 VLPs (i.e. up to eightfold lower than the homologous VLP titre; Table 2). Amino acid alignments of these five GII/3 sequences revealed that 1152 had three unique amino acid residues compared with the other four GII/3 sequences (Thr-285, Ile-372 and Ser-508), two of which were located within the P2 domain (Thr-285 and Ile-372). Amino acid secondary structure predictions made using the PSIPRED secondary structural prediction software revealed that the VP1 secondary structures for 809, Sh5, 18-3 and 336 had a helix structure between residues 219 and 237; this helix structure was absent for 1152 (Fig. 5). This helix structure may, in part, influence the cross-reactivity among the GII/3 VLPs (i.e. without the helix structure); GII/3 1152 VLPs cross-reacted weakly with the other four GII/3 antisera. This suggestion may also explain NoV virulence in which some strains appear to infect a certain population over an extended period of time (Dingle, 2004; Noel et al., 1999). In a recent report, single amino acid changes were suggested to represent a possible way for the virus to evade the host immunity (Dingle, 2004). In addition, one report suggested that a change in VP1 secondary structure (i.e. the disappearance of a helix structure) was responsible for a chronic NoV infection in an immunocompromised patient for over 2 years (Nilsson et al., 2003).

Almost half of our constructs (strains SeV, 645, CV, HV, Ina, 809, Sh5, 18-3, 1152, 104, 754, CHV and Alph23) did not include the ORF3 sequence, which encodes a minor capsid protein (VP2) thought to increase the stability of NoV VLPs and may function in RNA genome packaging (Bertolotti-Ciarlet et al., 2003). For rabbit haemorrhagic disease virus, VP2 is essential for the production of infectious virus (Sosnovtsev & Green, 2000). Nevertheless, we found that all constructs with or without ORF3 sequences expressed VLPs that were morphologically similar to native NoV (Fig. 2). Further studies are needed to determine whether VP2 has some influence on antigenicity.

In conclusion, this cross-reactivity study represents the most extensive undertaken for any genera in the family Caliciviridae. Since human NoV strains cannot be propagated in cell culture systems and human serological studies have found that VLPs and native virions share similar antigenic properties, VLPs have been used to understand antigenic relationships in more detail. Further studies, such as high-resolution structural analysis of other NoV genotypes and antigenic mapping, are needed in order to explain the complex NoV antigenicity, as previously suggested (Chen et al., 2004). Finally, the results and reagents from this study can be used to design detection systems capable of detecting a broad-range of genotypes in clinical specimens; in particular, GI/11 antisera may be capable of detecting at least 32 % (12/37) of the recently described NoV genotypes (Kageyama et al., 2004).

This work was supported in part by a grant for Research on Emerging and Reemerging Infectious Diseases from the Ministry of Health, Labor and Welfare of Japan, and a grant for Research on Health Science Focusing on Drug Innovation from The Japan Health Science Foundation. We also thank Dr Kim Green for providing the Hawaii virus recombinant baculovirus.