Norovirus Gastroenteritis Outbreak with a Secretor-independent Susceptibility Pattern, Sweden

Nonsecretors were highly susceptible to norovirus GI.3 in a foodborne outbreak.

Saliva-binding studies have demonstrated that different NoV strains exhibit different binding patterns (14)(15)(16), with the Norwalk virus (GI.1) mainly recognizing saliva from secretors with blood groups A and O, while exhibiting low or no binding to saliva to nonsecretors and carriers of blood group B, suggesting protection against infection among the latter 2 groups. Virus-like particles (VLPs) of the common GII.4 strains have been found to mainly bind saliva from secretors irrespective of blood group (16), although binding to nonsecretor saliva has been described for VLPs of some GII.4 strains (17).
Although NoV infections of secretors are well documented (18) and a few cases of infected nonsecretors have been reported (19,20), no virus has been identifi ed in authentic outbreaks that is completely secretor or Lewis antigen independent, where homozygous carriers of the nonsense G428A mutation in FUT2 are at similar or higher risk for infection than are secretors. We describe a foodborne NoV outbreak in which persons were infected regardless of secretor status or Le phenotypes; and no difference was observed between nonsecretor (Le a+b− ) persons and secretors regarding risk of symptomatic norovirus infection. Our data provide new knowledge about susceptibility factors and NoV genotypes and suggest that additional studies of host genetic receptor factors and NoV are needed.

Norovirus Gastroenteritis Outbreak
with a Secretor-independent Susceptibility Pattern, Sweden

Outbreak Data and Sample Collection
In October 2007, a NoV gastroenteritis outbreak occurred in Jönköping, Sweden, at a seminar for healthcare improvement (October [25][26][27], attended by 112 healthcare workers from different parts of Sweden. The healthcare workers were asked to take part in this case-control study, and 83 persons, including 4 employees of the restaurant that provided food service, decided to participate. Thirtythree of these 83 persons acquired acute gastroenteritis during or shortly after the seminar. Saliva samples were collected from all 83 participants in the study and stored at -20°C until further use. Stool samples (n = 4) were obtained from the cook, 2 employees, and 1 participant of the conference with symptoms of NoV gastroenteritis. Epidemiologic investigations indicated that the lunch on the fi rst day was contaminated with NoV and was subsequently the cause of the outbreak. The cook was ill 4 days before the outbreak started, and 3 days later other employees of the restaurant became ill, suggesting the restaurant employees as the probable source of NoV contamination in the food. NoV disease was identifi ed by at least 1 of the following signs or symptoms: vomiting, diarrhea, or nausea combined with stomach ache ≈12-60 hours after ingesting the meal. Description of symptoms was obtained through a questionnaire sent to all participants in the study. The study was approved by the local ethics committee (M205-04 T48-08).

DNA Extraction from Saliva
Genomic DNA from 200 μL saliva was extracted by using QIAamp DNA Mini Kit (QIAGEN, Hilden, Germany) according to the instructions of the manufacturer (Blood and Body Fluid Spin Protocol). Extracted DNA was stored in AE buffer (QIAGEN) at -20°C until PCR amplifi cation.

PCR Amplifi cation of FUT2 and Determination of FUT2 428 Genotype
The FUT2 gene amplifi cation by PCR was performed as previously described (6). Genotyping of the G428A mutation in the FUT2 gene was performed as previously described (6,7,21). These methods can distinguish between carriers of the homozygous wild-type, heterozygous, and homozygous mutated genotype.

Virus RNA Extraction and Reverse Transcription
RNA extraction from the 4 collected stool specimens was performed by using the EZ1 robot (QIAGEN) according to the manufacturer's instructions and stored at -80°C until used for reverse transcription. Reverse transcription was performed as previously described (6,23), by using random hexamer primers (GE Healthcare, Uppsala, Sweden) and Illustra Ready-To-Go RT-PCR beads (GE Healthcare).

Norovirus Detection with Real-Time PCR
NoV detection and quantifi cation were performed with a real-time PCR specifi c for the open reading frame (ORF) 1-ORF2 junction, as described by Nordgren et al. (24). This real-time PCR assay can semiquantify and distinguish between NoVs GI and GII (24). PCR amplifi cation of the N-terminal and shell (N/S) region was performed on a PTC-

Nucleotide Sequencing of the Norovirus N/S Region and Virus Genotyping
Nucleotide sequencing of the N/S region was performed by Macrogen Inc. (Seoul, South Korea). The sequencing reaction was based on BigDye chemistry; NVG1f1b forward primer (24) and G1SKR reverse primer (25) were used as sequencing primers. The amplicons were sequenced twice in each direction. Sequence alignment of the Jönköping (JKPG) strain and reference NoV genotypes was performed by using the ClustalW algorithm, version 1.8 (www.ebi.ac.uk/clustalw), with default parameters, on the European Bioinformatics Institute server. We performed phylogenetic analysis using the MEGA 4.0 software package (www.megasoftware.net), and the phylogenetic tree was constructed using the neighbor-joining and Kimura 2-parameter methods. Signifi cance of the taxonomic relationships was obtained by bootstrap resampling analysis (1,000 replications). Assignment of genotypes used reference strains described by Zheng et al. (26).

PCR Amplifi cation of the Norovirus Capsid Gene
To amplify the gene encoding the NoV capsid, we set up a PCR mixture containing 2.5 μL 10× native Pyrococcus furiosus (pfu) polymerase buffer (Invitrogen AB, Lidingö, Sweden), 200 μM GeneAmp dNTP mix with dTTP (Applied Biosystems), 200 nM forward primer CapGI3fw (5′-GAT CTC CTG CCC GAT TAT GTA AAT GAT GAT G-3′, this study), targeting the end of ORF1 and beginning of ORF2, 200 nM reverse primer CapGI3rv (5′-CAT TAT GAT CTC CTA ATT CCA AGC CTA CGA GC-3′, this study), specifi c for the end of ORF2 and beginning of ORF3, 5 μL cDNA, 2.5 U native pfu DNA polymerase (Stratagene, La Jolla, CA, USA), and 36 μL RNAse-free water. After initial denaturation at 94°C for 5 min, PCR amplifi cation was performed with 40 cycles of 94ºC for 1 min, 58ºC for 1 min, and 72ºC for 2 min, and thereafter a fi nal elongation at 72ºC for 10 min. The PCR products were visualized by electrophoresis on a 2% agarose gel, using staining with ethidium bromide and UV transillumination.

Cloning of the Norovirus Capsid Gene and Nucleotide Sequencing
The capsid fragment was cloned into a pPCR-Script Amp SK(+) vector and transformed into XL10-Gold Kan ultracompetent cells, using the Stratagene PCR-Script Amp Cloning Kit (Stratagene) according to the manufacturer's instructions. After overnight incubation of 2 separate colonies from each transformation reaction, plasmid DNA was extracted and purifi ed, using the Plasmid Miniprep Kit (QIAGEN) according to the manufacturer's instructions.
Nucleotide sequencing was performed on 2 separate plasmid extractions from each sample (n = 2) by Macrogen Inc., by using the BigDye chemistry with M13 forward and reverse primers. The nucleotide sequences for the N/S region or the complete capsid gene of the JKPG isolates are available under GenBank accession nos. FJ711163, FJ711164, and FJ711165.

Statistical Analysis
Categorical data were analyzed using the Fisher exact test with 2-tailed signifi cance. Unadjusted odds ratios (ORs) and 95% confi dence intervals (CIs) were calculated using SPSS 14.0 for Mac (SPSS Inc., Chicago, IL, USA).

Distribution of ABO and Lewis Phenotypes and Secretor Status
To investigate whether persons associated with the outbreak had a skewed HBGA profi le, we determined the ABO, Lewis, and secretor status distributions among symptomatic and asymptomatic/nonexposed persons and compared them with results from earlier investigations of the population in Sweden ( Table 1). The ABO, Lewis, and secretor status distributions were in the normal ranges compared with those investigations (Table 1), with the exception of the AB and Lewis negative phenotypes. Furthermore, we observed that all HBGAs investigated, except AB (n = 1), were found among asymptomatic/nonexposed and symptomatic persons. Sixty-one persons were secretor and Lewis positive; of these, 52 (85%) were positive for Le a and Le b in saliva. The 4 persons from whom NoV was isolated were all secretors, having ALe a−b+ , ALe a−b+ , OLe a−b− , and OLe a−b+ HBGA profi les, respectively.

Difference in Susceptibility to Symptomatic Infection between Secretors and Nonsecretors
Previous studies have shown a strong correlation between symptomatic NoV infections and the secretor-positive phenotype (7)(8)(9)(10). To investigate whether secretor and Lewis status were associated with susceptibility in this study, secretor and Lewis status were determined by genotyping and phenotyping of all persons. We observed that 7/15 (47%) nonsecretors were symptomatically infected, compared with 26/68 (38%) secretors (Table 1). Although the calculated OR for nonsecretors was ≈2× that of secretors (OR 1.41 vs. 0.71), the differences were not signifi cant (Table 2). Thus, in the group studied, nonsecretors were as likely as secretors to be symptomatically infected by norovirus. The same pattern was observed for the Lewis phenotypes; no statistical difference was found between persons with Le a+b− , Le a−b+ , or Le a−b− regarding risk for symptomatic infection ( Table 2). None of the Lewis-negative nonsecretors (n = 3) were symptomatically infected. The FUT2 G428A genotyping did not show any signifi cant differences between heterozygous secretors and homozygous secretors (OR 1.20, 95% CI 0.49-2.95 vs OR 0.67, 95% CI 0.27-1.65; Table 2).

Association between ABO Blood Types and Symptomatic Infection
Previous studies have shown that ABO blood types are associated with susceptibility to symptomatic NoV infections, with persons having blood type B at lower risk of infection when challenged with Norwalk virus (GI.1) (8,28). In this outbreak, we found that symptoms developed in 2/12 (17%) of persons with blood group B (Table 1). Although persons with blood group B were infected to a lesser extent than persons with other blood groups, this reduction was not signifi cant (OR 0.27, 95% CI 0.05-1.33; Table 2). Furthermore, no signifi cant differences were found when comparing symptomatic and nonsymptomatic persons with blood types A and O (OR 1.56, 95% CI 0.58-4.16, and OR 1.39, 95% CI 0.50-3.89, respectively) ( Table 2). Thus, no blood type provided complete protection or was associated with a higher or lower risk for disease.

Association between Blood Type, Secretor Status, and Clinical Symptoms
A recent study suggested that blood type can have an infl uence on clinical symptoms after NoV infection (29). To investigate whether this would apply in this outbreak, blood types, secretor status, and clinical symptoms were compared. We did not fi nd any correlation between blood type and secretor status with clinical symptoms ( Table 3).

Similarity of JKPG and Kashiwa645 Strains in the P2 Domain and Putative Receptor Binding Sites
NoV GI was detected by real-time PCR in all collected stool specimens (n = 4); three of these isolates (881-883) were subsequently genotyped by nucleotide sequencing of the N/S region. The fourth sample could not be genotyped because of low virus concentration in the stool sample. Phylogenetic analysis clustered the 3 isolates with NoV   GI.3 strains (data not shown). The entire capsid gene was subsequently sequenced from 2 isolates and compared with reference strains (Figure 2). The closest amino acid similarity (98.0%) of the complete capsid gene was found with strain PD196-DEU (GI.3), isolated in Germany 2000, and with the Kashiwa645 (GI.3) strain (97.8%), used in an earlier VLP binding study (14). We then investigated the amino acid composition of the capsid P2 domain of the outbreak strain and compared it with the Kashiwa645 and Norwalk strains. Although the JKPG strain differed by 4 aa at positions 344, 367, 377, and 397 (97.1% homology) compared with Kashiwa645, it shared only ≈50% aa positions with the GI.1 Norwalk strain.

Discussion
Previous studies have shown a strong (6-11) but not absolute (19,20) association between nonsecretors and protection from symptomatic NoV disease. In contrast to these observations, we report a foodborne NoV outbreak affecting persons regardless of secretor, Lewis, or ABO phenotype.
Because the host genetic observation of this outbreak was unexpected, attempts were made to compare the HBGA frequencies of the participating persons with those of the population in Sweden. The ABO, secretor, and Lewis phenotype frequencies in this study agreed with results from earlier investigations from the population in Sweden (21,27) (Table 1), with the exception of the AB and Lewis negative phenotypes, probably due to their low prevalence in combination with the small sampling set. Seven (8%) secretor and 3 (4%) nonsecretor persons were Lewis negative and hence lacked Lewis antigen in saliva. Genotyping of the FUT2 G428A nonsense mutation confi rmed secretornegative genotype of all Le a+b− persons and the secretorpositive genotype of Le a−b+ persons.
Comparison of secretor and Lewis phenotypes regarding susceptibility to symptomatic NoV infection showed that nonsecretors were as susceptible to symptomatic disease as secretors. Consistent with the lack of secretor association, no signifi cant difference in susceptibility was noted between Le a+b− and Le a−b+ persons (OR 2.42, 95% CI 0.70-8.42 vs. OR 0.73, 95% CI 0.27-1.95). None of the nonsecretors who were also Lewis negative (n = 3), hence lacking the Le a antigen and ABO in saliva, were symptomatically infected. These fi ndings indicate but do not prove that the Le a antigen is a putative receptor for this norovirus strain. The disease pattern of this outbreak is consistent with the fi ndings by Shirato et al. (14), who observed strong binding to synthetic Le a and saliva from secretors and nonsecretors with VLPs from the GI.3 Kashiwa645 strain, which shares high homology with the JKPG strain in the P2 domain. A mechanistic virus-saliva binding study with the authentic virus would have been desirable, but limited amounts of virus restricted our attempts to investigate if the outbreak virus binds to saliva both from secretors and nonsecretors.
No ABO phenotype provided protection or was associated with a higher risk of disease, although persons with blood type B exhibited a low (17%) frequency of symptomatic infection (Tables 1, 2). Blood type B has previously  been associated with protection from disease when challenged with the Norwalk virus (28) and was also supported by in vitro binding studies with VLPs (16). Shirato et al. (14) found that the Kashiwa645 (GI.3) VLP bound weaker to saliva from blood type B-positive persons compared with types A and O, which agrees with the disease pattern observed in this outbreak. It is possible that the α-gal in the blood type B structure partly covers an epitope needed for binding and hence decreases the ability of the JKPG strain to infect carriers of blood type B. One limitation of our study is that some of the asymptomatic persons may not have been exposed to the virus. This possibility could result in sampling artifacts because symptoms could have developed in the unexposed (and thus seemingly protected) persons if they had been exposed. However, this possibility is unlikely to infl uence the main fi ndings of this study, namely the secretor-and HBGA-independent infection pattern, because symptomatic persons were found in all HBGA groups.
An advantage with genotyping compared with phenotyping is that the roles of heterozygosity and homozygosity in disease susceptibility can be investigated. In this study, we did not observe any signifi cant differences between heterozygous and homozygous secretors (OR 1.20, 95% CI 0.49-2.95 vs. OR 0.67, 95% CI 0.27-1.65) ( Table 2), which is in agreement with earlier observations (7,9).
By comparing our strain with a reference strain of the same cluster (Kashiwa645) used in earlier binding studies (14), we aimed to elucidate structural similarities or differences that could explain the unique disease profi le of the outbreak. The JKPG strain investigated in this outbreak shares high amino acid homology with the GI.3 Kashi-wa645 strain (Figure 2). Shirato et al. (14) found that the Kashiwa645 strain bound to the same extent to secretor and nonsecretor saliva. However, another consideration is that Asian nonsecretors in the study (14) were identifi ed as carriers of a missense mutation at nt 385 (A→T) and thus are incomplete nonsecretors, producing small amounts of ABO and Le b HBGA in secretions. The similarities between the binding profi le of the Kashiwa645 strain and the disease profi le of the JKPG strain indicate that saliva binding may be used to assess susceptibility patterns for individual NoV strains.
In conclusion, we report a foodborne NoV outbreak infecting persons irrespective of Lewis and secretor status, with Le a+b− persons homozygous for the FUT2 G428A nonsense allele being symptomatically infected at similar rates compared with secretors. Our observed disease pattern is in concordance with saliva binding specifi cities of VLP based on the Kashiwa645 strain, sharing high homology in the P2 domain with the JKPG strain. Increased knowledge of susceptibility factors for norovirus disease will be helpful in the development of preventive or therapeutic measures for infection.