On 4 September 2009, an outbreak of gastroenteritis occurred in a school located in a small town in Guangdong Province. The school had 285 teachers, of whom 55% resided at the school; 50% of these dined in the school's restaurant; 5484 students (99% residing at the school, 99% dining in the school's restaurant), and 85 staff members for cleaning and security. There had been no rain or large-scale gatherings at the school in the 2 weeks before the outbreak. About 80% of the school's daily water supply – water that was used for all purposes except drinking – was from a single well. Water for drinking was supplied through a municipal water supply and was boiled first.

Cases were defined as patients with ⩾3 loose stools and/or vomiting in a 24-h period beginning from 31 August to 10 September, 2009. A retrospective cohort investigation using a standardized questionnaire was conducted on 10 September 2009, with 5795 people interviewed in-person and 59 by phone, with 100% enrolment. The exposed cohort was defined as those only served by water from the school well and the unexposed cohort was defined as all persons served only by the municipal water supply. Attack rates, relative risks (RR), and 95% confidence intervals (CI) were calculated using SPSS v. 13.0 (SPSS Inc., USA). A P value of <0·05 was considered statistically significant. Because NoV is also able to be transmitted through contaminated food, we also investigated common food exposures in students and teachers.

We examined the well construction log, current well and chlorination conditions, and potential sources of contamination. The township Centre for Disease Control (CDC) tested the well water for total aerobic bacteria and faecal coliforms using the most probable number (MPN) technique. Standard laboratory methods were used for bacteriological investigations.

Diarrhoeal specimens were collected on 8 and 10 September 2009, from patients meeting the case definition; faecal and rectal swabs were collected from 19 students and nine teachers. Faecal specimens were diluted with Hank's solution to make a 10–20% w/v suspension. RNA extraction was performed after centrifugation (926 g for 2 min) to clarify supernatants. Rectal swabs were placed in Hank's solution vortexed for 2 min, subseqently discarding the swabs. Rectal swab suspensions were centrifuged (926 g for 3 min) to clarify the supernatant before RNA extraction.

Water samples ranging in volume from 0·8 to 2·5 l were collected directly from the well (n = 1) and from tap water pumped from the well (n = 2) on 10 September 2009, prior to chlorination. On 16 September, after chlorination of the well by the township CDC, additional water samples were collected directly from the well (n = 1), from tap water pumped from the well (n = 2), from a storage tank holding well water (n = 1), from municipal tap water (n = 1) and from river water (n = 1). MgCl₂ was added to the water specimens to make a final concentration of 0·05 m, and the pH was adjusted to 3·0. Water specimens were filtered through a mixed cellulose ester membrane [13] (Advantec, USA) and eluted with buffer containing 0·05 m glycine and 3% beef extract (pH 9·5). Addition of 40% PEG6000 and 5 m NaCl (final concentration 10% and 0·3 m, respectively) was followed by incubation (12 h at 4 °C) and centrifugation (10 286 g for 30 min). Pellets were re-suspended in 1 ml of RNase-free water (Takara, Japan).

Viral RNA was extracted from faecal and water samples using a QIAamp Viral RNA Mini kit (Qiagen, Germany) according to the manufacturer's protocol.

For detection of NoV, we used primers COG2F/COG2R as described previously [14]. Real-time RT–PCR was conducted in an ABI.7500 Fast Real-Time PCR System (Applied Biosystems Inc., USA) using Perfect Real Time reagent (Takara). Real-time RT–PCR was performed in a 20 μl volume as described previously [14]. Fluorescent signals were recorded at 72 °C. Standard curves for determining the amount of virus in water specimens were generated by serially diluting plasmids containing NoV GI and GII PCR products. The formula Ct = 40·991 − 3·494 × log x (where Ct is the cycle threshold and x represents the viral quantity, copies/μl) was used for quantitative analysis of NoV. The correlation coefficient of the curve was 0·998. The percentage recovery, which is equal to the quantity of virus after concentration divided by the quantity of virus before concentration, could then be calculated.

Amplified products from two students, two teachers and the well-water specimens were selected for nucleotide sequence analysis. The primers JV13I/JV12Y and SKF/SKR designed by Vennema et al. [15] and Kojima et al. [16] were used to amplify portions of the genes encoding the RNA-dependent RNA polymerase (RdRp, nt 4279–4605) and capsid protein (nt 5046–5398), respectively. Bidirectional sequencing of the amplified PCR products was achieved using BigDye 3·1 chemistry (Applied Biosystems Inc.) using an ABI PRISM 3130 DNA analyser. DNA sequences were aligned using the Bioedit program and construction of phylogenetic trees was performed using MEGA 4.0 software. Reference viral gene sequences were provided by National Institute for Public Health and the Environment, The Netherlands. Nucleotide sequences were assigned Genbank accession numbers HM627865–HM627869 for RdRp, and HM627870–HM627874 for capsid sequences.

The first case of gastroenteritis occurred on 4 September 2009, with the peak incidence noted on 6 September (Fig. 1). The overall attack rate during this outbreak was 1·8% with 1·7% (92/5484) of students and 5·6% (16/285) of teachers affected. In total, 2·0% (62/3153) of the entire male, and 1·8% (46/2616) of the female school population were affected. The outbreak did not extend to the adjacent residential area, where the incidence of diarrhoea remained stable. Symptoms included diarrhoea (100%), abdominal pain (82%), vomiting (72%), nausea (61%), flatulence (51%), and fever (38%). In most cases, symptoms were mild, self-limiting, and lasted a median of 2 days. Some of the rooms (serving 2320 people) in the school complex received only untreated well water, other rooms (serving 1373 people) received only municipal tap water; the remaining rooms received water from both sources. An investigation revealed that all foods were boiled with municipal tap water but cleaned with untreated well water before cooking. Students, teachers and staff consumed the same foods. No risk factor related to food consumption was identified. Students and teachers reported drinking boiled tap water. The epidemiological investigation showed that the attack rates in the dormitory building supplied with untreated well water and the dormitory building supplied with municipal tap water were 2·0% and 1·1%, respectively. Exposure to well water had a RR of 1·9 (95% CI 1·1–3·4), indicating that well water was a risk factor for the outbreak. Well water was disinfected with chlorine on 10 September; no new cases were reported after 12 September (Fig. 1). Fig. 1

[colour online]. Date and times of illness onset for cases associated with a norovirus outbreak in Guangdong, China, 2009.

Several sources of water to the school were investigated for NoV contamination. Routinely, untreated well water was first pumped into a storage tank and, from there, pumped to rooms through pipes as required. The well supplying this water was approximately 30 m deep; a drainage line at the bottom of the well leads to a place 20 m lower than the bottom of a nearby river. The well surface was completely covered, and the environment in the immediate vicinity was cluttered with debris. Microbiological investigation of the well water indicated faecal contamination: there were 2·2 × 10³ colony-forming units (c.f.u.)/ml of aerobic bacteria and a large number of faecal coliforms, with an MPN count of 72/100 ml. Further environmental investigation was unable to determine the source of contamination.

RNA samples from the students (8/19, 42%) and teachers (5/9, 56%) tested positive for NoV. Samples from the well (well water and pumped well water) were also positive for NoV by real-time RT–PCR before disinfection with chlorine (Table 1). After disinfection, well water remained positive although much weaker than before (Fig. 2); however, the storage tank and pumped water from the well tested negative. Municipal tap water and river water were negative for NoV (Table 1). Fig. 2

[colour online]. Real-time RT–PCR result of water samples after disinfection.

We detected 215 copies/μl of NoV in well water and water pumped from the well (tap water samples 1 and 2) before sample concentration. After 2500× concentration through a mixed cellulose ester membrane, up to 10⁵ copies/μl NoV could be detected. Two pumped well-water samples (tap water samples 1 and 2) that were concentrated 800 times contained between 2·5 × 10³ and 5·0 × 10³ copies/μl, respectively (Table 2). Concentrating municipal or river water did not show a positive result in detection of NoV. Table 2.

Quantity and yields of norovirus recovered from water samples

Segments of the genes encoding NoV RdRp and capsid proteins were amplified and the DNA sequences were analysed from samples collected from two students, two teachers, and well water. The concentration of NoV in positive pumped well-water samples was not enough for sequencing. Comparative sequence analysis showed that all of the NoV gene sequences from humans and water were indistinguishable. Phylogenetic analysis against a collection of reference sequences representing a variety of NoV genotypes showed that the virus responsible for this outbreak was classified as a genotype GII.4 NoV, with the closest identity to GII.4-2006b (AB294793 and EF126966; Figs 3 and 4). Fig. 3.

Phylogenetic tree based on nucleotide sequence of the norovirus RNA-dependent polymerase (nt 4573–4584). The reference strain designation is the Genbank accession number followed by the genotype designation (in parentheses).