In The Netherlands, illness due to Norwalk-like viruses (NLVs) of the family Caliciviridae is quite common. NLVs cause >80% of the outbreaks of gastroenteritis reported to municipal public health centers and at least 5% of the cases of gastroenteritis for which a general practitioner is consulted. In addition, up to 18% of community cases of gastroenteritis in the 1998/1999 winter season have been attributed to NLVs. Patterns of disease activity differ remarkably, with “normal” years, when outbreaks occur that are caused by different types of NLV, and “epidemic” years, when outbreaks appear to be caused by a single strain. This observation suggests selection of antigenic variants with increased virulence or altered modes of transmission. In addition, since caliciviruses related to the NLVs from humans have been detected in stool specimens from calves at 45% of the dairy farms in The Netherlands, the possibility of spillover of epidemic strains from an animal reservoir to humans should be considered.
Gastroenteritis due to viral infection of the intestinal tract is a common illness in humans, with high morbidity reported worldwide. Although rotaviruses, astroviruses, and enteric adenoviruses have been considered to be the main viral causes of gastroenteritis, it has been increasingly recognized that at least 2 other virus groups may be involved in cases of gastroenteritis and may, in fact, have been seriously underdiagnosed in the past. These 2 groups are provisionally named genera of the human enteric caliciviruses: “Norwalk-like viruses” (NLV), which are also known as small round structured viruses, and the “Sapporo-like viruses” (SLVs) or “typical” caliciviruses [1, 2]. The inability to culture these agents in vitro, however, impaired research until the cloning and sequencing of prototype strains directly from stool specimens was accomplished [3–7]. This breakthrough led to the rapid development of diagnostic assays based on RNA amplification by reverse transcriptase-polymerase chain reaction (RT-PCR) [8–12].
From studies with these first-generation molecular diagnostic methods, the caliciviruses have emerged as possibly the single most common cause of infectious gastroenteritis in people of all ages. More important, they are the main cause of outbreaks of gastroenteritis in nursing homes and hospitals [2, 13–15]. These viruses are transmitted from person to person by the fecal-oral route, either directly or indirectly via contaminated surfaces, food, or water. Numerous outbreaks of caliciviruses have been linked to the consumption of food prepared by infected food handlers .
Herein, we review work on enteric caliciviruses in The Netherlands, including unpublished data from our laboratories.
Development of Detection and Typing Methods in The Netherlands
To enable epidemiologic studies that address the incidence of NLV, several groups developed RT-PCR-based detection methods after the cloning of prototype NLV strains was accomplished [8–12]. Likewise, we developed a single-round generic NLV-specific primer pair, which was validated using coded panels of stool specimens in which NLV or other viruses had been detected by electron microscopy (EM) [14, 15]. The assay targets a 327-bp region in the viral RNA polymerase, and the best results were obtained with an antisense primer close to the 5′-end of the coding region for the conserved YGDD motif. In its present format, the assay detects strains from at least 15 known genotypes with 100% specificity. The polymerase regions that were amplified had 60%–100% nucleotide identity with Norwalk virus.
The detection limit of the RT-PCR has been estimated to be 3–30 RNA-containing particles, as determined by mixing an NLV-containing stool specimen (genotype Hawaii) with known concentrations of 80-nm latex beads. However, given the high level of sequence heterogeneity, even in the relatively conserved region of the RNA polymerase, this detection limit is likely to be different for viruses from different genetic clusters. The assay is now used routinely in The Netherlands [14–16].
In the absence of an in vitro culture system for these viruses, antigenic typing historically has been limited to studies in which clumping of stool virus was studied in the presence of acute and convalescent patient sera by solid-phase immune EM [17, 18]. These studies and studies with recombinant capsid protein have established that there are clear antigenic differences between morphologically indistinguishable NLVs [19–22]. To determine which genomic regions could be used for a typing scheme that correlates with antigenic typing, regions across the genome of a broad range of antigenically and genetically distinct NLV strains have been sequenced. From these studies, NLV have been divided into 2 genogroups and tentatively into 15 genotypes . With few exceptions, clustering of strains was consistent, regardless of which genomic region was used for the analysis [23, 24]. We have four working criteria for defining genotypes: (1) >80% amino acid similarity when comparing complete capsid gene sequences, (2) at least 15% similarity on the basis of the nucleotide sequence of the polymerase fragment for genogroup I (GI) and 10% for genogroup II (GII) strains, (3) stable clustering by phylogenetic analysis (irrespective of which method is used for phylogeny reconstruction) with high bootstrap values for the lineages, and (4) cluster representation by at least 2 strains [16, 23, 25].
To develop a rapid typing assay, we selected sequences within the amplified fragment specific for each of the genotypes to develop genotype-specific probes. These probes, which have a 5′-amino group, are covalently linked to the carboxyl groups on an activated membrane . This method, called reverse line-blot hybridization (RLB), originally was developed for the detection and typing of bacteria [26, 27]. Biotin-labeled amplicons are generated by a generic NLV RT-PCR  and hybridized to the membrane. An example for 13 genotypes is shown in figure 1. This RLB method is easy to perform with a high throughput of strains, and labeled membranes can be reused at least 20 times. Also, the possibility to easily add a probe to a new emerging NLV genotype and the potential to develop a multiplex RT-PCR for other RNA viruses associated with gastroenteritis (Sapporo-like human caliciviruses and astroviruses) make this method ideal for the early detection and investigation of multinational common-source outbreaks . At present, the use of the RLB in clinical virologic laboratories is evaluated with special emphasis on detection limits for different genotypes.
Epidemiologic Studies in Humans
In 1991, a population-based study in The Netherlands estimated the number of episodes of gastroenteritis to be 7 million cases annually, resulting in ∼120,000 physician visits (7.7/1000 persons per year) . In these studies, stool specimens were examined routinely only for the presence of Salmonella and Campylobacter bacteria. In recently initiated epidemiologic studies, a broad range of diagnostic assays was included to address the incidence and significance of other pathogens as causes of gastroenteritis at different population levels. These studies are presented in brief below.
In The Netherlands, the epidemiology of NLV infections in people who are ill enough to visit their physicians has been assessed in a physician-based case-control study that started in May 1996 and ended in May 1999 [29, 30]. From this study, the incidence and seasonal distribution of NLV infections and the age and sex of those infected have been determined for cases of gastroenteritis that warrant a visit to one of the physicians that participate in a disease-surveillance network (referred to as NIVEL) that covers ∼1% of the population of The Netherlands. Detailed questionnaires are used to establish risk factors for acquisition of infection with a broad range of microorganisms (Salmonella, Shigella, Campylobacter, Yersinia, Giardia, and Cryptosporidium species, and Escherichia coli, Dientamoeba, Entamoeba, Blastocystis, Cyclospora, astro-virus, group A rotavirus, adenovirus types 40/41, and NLV and SLV). Full details of this study will be published elsewhere. From May 1996 to May 1998, NLVs were detected by a generic RT-PCR assay in stool specimens from 30 (5%) of 601 specimens from cases with acute gastroenteritis for which a general practitioner was consulted and in 3 (0.7%) of 418 stool specimens from control patients . As a comparison, rotavirus was found in 5% of cases and 1.2% of controls (figure 2A). The incidence of NLV was slightly higher in young children, but remained at around 4.5% for all age groups, whereas rotavirus and adenovirus were found much more frequently in children than in adults (figure 3A).
In December 1998, we started a population-based cohort study with a nested case-control design to determine the incidence of gastroenteritis in the general Dutch population; we wanted to establish which microorganisms cause gastroenteritis, determine risk factors and transmission routes for specific microorganisms, and determine the health burden and costs of gastroenteritis in The Netherlands . In this study, a randomized sample of people from the population area covered by the physician practices mentioned above is asked to participate. Participants enter the study cohort for a half year and submit questionnaires and stool samples when experiencing symptoms of gastroenteritis. The study coordinator is informed when a participant first becomes ill and, in response, recruits 2 age- and sex-matched controls from the cohort. The controls are asked to submit stool specimens. Comparative analysis of data from the physician-based study and this study will, along with other studies, provide information on the severity of illness caused by specific microorganisms. Preliminary data from the first 3 months of the study suggest a very high incidence of NLV in people of all age groups (46/265 samples from cases [17.4%] and 14/231 samples from controls [6.1%]) (figure 2B), with a slightly higher incidence in very young children (figure 3B). To correct for seasonal influences, the study will be continued throughout the year. Since the seasonal peak of NLV outbreaks in The Netherlands occurs in January, February, and March, the incidence rates will most likely be lower at the end of the study.
The incidence of NLV in hospitalized patients has not been addressed systematically. One small-scale study in The Netherlands suggested that NLVs are not important causes of hospitalization because NLV was detected by RT-PCR in only 1 of 27 children <3 years old who were hospitalized for gastroenteritis in 1 year, whereas 13 of these (48%) were positive for rotavirus . NLV was also detected in a stool specimen from an asymptomatic control. Another explanation for the low number of NLV-positive stool specimens in these hospitalized children may be that the amount of time between onset of illness and sampling is longer than, for example, that in a physician-based study, resulting in false-negative RT-PCR tests.
In The Netherlands, outbreaks of gastroenteritis are reported to the municipal health services (MHS) and to food inspection services (FIS). There is a partial overlap between these two systems . Historically, outbreaks suspected of viral cause were selected locally and submitted for examination. In 1996, we studied the incidence of NLV in outbreaks reported to the MHS by using a standardized protocol for outbreak investigation and by stressing the need for submitting stool samples for virologic examination for all outbreaks rather than a (unknown) selected sample made by the local infectious disease specialist. Outbreaks were labeled NLV-associated if ⩾50% of the stool specimens from patients were NLV positive by RT-PCR, with a minimum number of 5 specimens analyzed per outbreak, and no other pathogen was present. NLV was detected in 87% (60/67) of the reported outbreaks of gastroenteritis . Most of these outbreaks occurred in nursing homes (59%) and hospitals (25%), with high attack rates both in the residents and patients (45%) and nursing staff (29%). The total number of outbreaks was 3.2 times the annual averages for 1994, 1995, 1997, and 1998, suggesting a potential reporting bias in 1996. Of interest, in 1996 we found the predominant circulation of 1 strain, indicating that other factors than reporting bias may explain the observed high incidence of outbreaks for that year [14, 15].
The focus of investigation of outbreaks reported to FIS is slightly different. Typically, a notification through this channel will lead to inspection of the premises involved for aspects of food hygiene and for collection of leftovers from implicated food items for microbiologic tests. Occasionally, stool specimens will be collected following notification of the MHS. Since reliable tests for the detection of NLV in various food items are not yet available and stool specimens are not routinely tested, the incidence of NLV in these outbreaks is not known. In addition, there is clearly underreporting. For example in 1997, 39 outbreaks of foodborne infections (of all causes) were investigated by MHS, and 50% of them were also investigated by FIS. In total, FIS investigated 520 clusters of foodborne infections in that year. The national notification system (with a total of 79 clusters notified) included 4% of the outbreaks reported to FIS and 68% of the outbreaks reported by MHS in 1997 . Studies are under way to improve the reporting network.
Overall, of the 184 outbreaks that have been investigated virologically by the National Institute of Public Health and the Environment (RIVM) since 1991, 82% were associated with NLV (figure 4). Twenty-six (14%) were labeled foodborne on the basis of the epidemiologic outbreak investigation. Of these, 20 (77%) were caused by NLV. Of note, most of the outbreaks had been reported to MHS, indicating a clear underrepresentation of outbreaks that have been designated as foodborne from the initial description of the outbreak (reported to the FIS) in our studies.
Genetic Diversity of Strains Recovered from Humans
Our working reference method for establishing genotypes by phylogenetic analysis of NLV is based on sequence analysis of the (major) structural protein gene (capsid gene). However, in most strains, there is complete agreement between dendograms based on the capsid gene and those based on the small fragment of the polymerase gene, which is used in the diagnostic RT-PCR assays (polymerase typing) . The lower degree of sequence divergence in the polymerase gene and the relative ease of combining diagnosis with molecular typing has led to the widespread use of polymerase typing in molecular epidemiology studies [14, 15, 34–37]. We have used this method since 1994 for typing of outbreak strains and since May 1996 for typing of strains from sporadic cases from the epidemiologic studies. During these years, we have observed that GII strains by far outnumber the GI strains (table 1) [14, 15]. This may reflect a true difference in prevalence or a bias introduced by lower sensitivity of the PCR assays for GI strains. However, the latter is unlikely, since a probable causative virus is identified in a large majority of outbreaks of suspected viral etiology. In addition, outbreak specimens that are negative for NLV, rotavirus, adenovirus, and astrovirus are screened for the presence of viruses by EM. So far, we have not found NLV particles in any of the outbreak specimens determined to be negative on the basis of virus assay.
When looking at genotype distribution, there are intriguing year-to-year differences (table 1): In 1997 and 1998, at least 8 genotypes were detected in stools from outbreaks with no apparent clustering, suggesting that these were mostly unrelated, independent events. However, in 1994, 1995, and 1996, a different pattern was observed, when sequential outbreaks were caused by strains that were indistinguishable on the basis of the polymerase gene sequence. We observed a small epidemic in 1994 caused by a virus in the Mexico virus cluster or genotype; a large-scale epidemic from 1995 to July 1996, when the same Lordsdale-like virus with only few nucleotide changes was found in 53 consecutive outbreaks; and a third small epidemic from September through December 1996 caused by Venlo virus, a strain in the P1B cluster . Venlo strain was also found during the same period in sporadic cases from the physician-based study, which had been initiated in May 1996. In the years following these epidemics, the Mexico-like and Venlo-like viruses have been detected infrequently, whereas the Lordsdale-like viruses are still commonly found along with other genotypes. Of interest, the genetic variability of the Lordsdale-like viruses found in 1997 and 1998 is not much greater than that of the 1995/1996 Lordsdale-like variant. The 1995/1996 epidemic was also observed in other countries .
The sudden emergence and spread of a single strain raises important questions about the mode of transmission that allowed these events to occur, especially since no obvious epidemiologic links were found between most outbreaks. Besides the possibility of large-scale foodborne or waterborne transmission, the possible existence of an animal reservoir or the existence of variants with altered tissue tropism (e.g., favoring spread by the respiratory route) are working hypotheses that need to be addressed in future studies.
Investigation of Animal Caliciviruses Related to NLVs
Until recently, the NLVs were considered to be pathogens with humans as the sole host. However, recent findings from Japan and the United Kingdom suggest the presence of NLV in pigs and cattle. Sugieda et al.  found NLV sequences by nested RT-PCR in 4 of 1117 healthy pigs in Japan. Early in 1999, 2 groups described the sequencing of bovine calicivirus-like sequences from historic samples that contained calicivirus-like particles by EM [40, 41]. These viruses, named Newbury agent and Jena virus, were pathogenic for young calves under experimental conditions and in field studies [40–43]. The 2 bovine enteric caliciviruses were genetically distinct, but they were most closely related to GI NLVs, and the swine viruses were most likely related to GII NLVs.
In a pilot study, we tested pooled stool samples from calves from 74 herds, fattening pigs from 63 herds, and adult cows from 20 dairy herds . All herd samples were from different parts of The Netherlands and were tested for the presence of NLV, using the single-round polymerase RT-PCR, which is used for diagnostic testing in humans . Thirty-three (45%) of the calf herds tested positive for NLV. The strains formed a tight cluster with the Newbury sequence . All dairy herd samples were negative, and 1 pig herd was positive for a virus strain that clustered tightly with the published pig calicivirus sequences from Japan .
These findings raise important questions about the host range of the NLVs. At this stage, it is unclear if the animal NLVs form genetically distinct stable lineages or are in fact part of a common pool of viruses circulating between animals and humans, although the finding of highly related strains in animals in different countries suggests the former. Studies are needed to resolve this issue.
Indirect evidence of the high incidence of NLV infections was obtained from studies in the winter of 1997/1998, during which sewage samples were collected at regular intervals from December through May . Most samples contained high loads of NLV (up to 107 RNA copies/0.5 L of raw sewage), which averaged an order of magnitude higher than that found for rotavirus. PCR products were cloned, and by sequencing individual clones, it was determined that multiple genotypes were present in most samples . Thus far, all strains from the sewage samples matched with NLV strains found in humans.
NLVs are increasingly recognized as an important public health problem, with very high incidence rates in people of all ages. The data from these ongoing studies will be used to determine the burden of illness and the health costs to society due to viral gastroenteritis of different causes in The Netherlands. However, in order to recommend preventive measures, better insight is needed into transmission routes and the mechanism(s) behind the appearance of epidemic strains. The role of food and of husbandry animals as a source of infection needs to be determined. The occurrence of multinational epidemics underscores the need for a standardized international surveillance system, allowing for prepublication comparison of strains from different sources in order to identify common strains early in an epidemic. Only then may epidemiologic outbreak investigations shed light on the origin of gastroenteritis outbreaks.
We thank Reina van der Heide, Hanneke Deijl, Petra de Bree, Olaf Nijst, Willemijn Lodder, the NIVEL team, the SENSOR team, and all the participating volunteers for their role in the work described here from our laboratories. We also thank Arjen van der Giessen for giving us stool specimens from the monitoring study for zoonotic enteric pathogens and the members of the gastroenteritis working group for valuable discussions.