INTRODUCTION
Outbreaks of infectious diarrhoea associated with exposure to water from public swimming pools and recreational lakes have been frequently reported. Contamination of the water may happen through faecal accidents from swimmers, as well as free-living animals grazing in areas close to lakes. Weather conditions such as heavy rain can also affect the water quality. Certain pathogens, including Cryptosporidium spp., Giardia intestinalis, toxigenic Escherichia coli, Shigella spp. and norovirus, have been commonly involved [Reference Yoder1–Reference Insulander7].
Since cryptosporidia are resistant to chlorine at the levels used in water treatment, contamination of public water supplies and swimming pools constitutes a high risk for outbreaks. The low infective dose is another contributing factor that facilitates the risk for waterborne transmission [Reference Chen3]. Of 26 waterborne outbreaks in England and Wales reported during 1992–1995, Cryptosporidium spp. were identified as the probable cause in 14 outbreaks associated with public water supplies and swimming pools [Reference Furtado2].
We have recently described an outbreak of cryptosporidiosis associated with an outdoor swimming pool [Reference Insulander7]. Simultaneously with this outbreak, another five children fell ill with diarrhoea after having swum in an adjacent indoor pool used for rehabilitative therapy. All five children were positive for Cryptosporidium spp. in faecal samples. Since none of them had visited the outdoor pool, the question as to whether these individuals belonged to the same outbreak was posed. In this paper we describe the use of a microsatellite-based genotyping method to study whether there was any relationship between Cryptosporidium isolates obtained from stool samples of individuals who had swum in the two different pools.
METHODS
The outbreak
In late August 2002, an outbreak of cryptosporidiosis associated with an outdoor swimming pool occurred in a municipality close to Stockholm, Sweden. The facility comprises two pools with a total volume of 700 m3, one large (25×25 m) used for swimming and one small for mainly preschool children. During the period 20–30 August, 4585 bathers visited the facility. Epidemiological and clinical data on this outbreak have recently been described [Reference Insulander7]. In brief, an estimated 800–1000 individuals (mainly children aged 6–12 years) were affected and the attack rate was about 50%. The median incubation period for individuals who visited the pool just once was 5 days (range 2–13 days). The frequency of secondary transmission within households was 10%.
The swimming facility was investigated and filtration systems and chlorination records for the outbreak period were examined without any adverse observations [Reference Insulander7]. Incidences of loose faecal stool in the pool were observed and were most probably the cause of the outbreak. However, none of the samples from pool water and filters was positive for cryptosporidial oocysts or bacterial enteropathogens [Reference Insulander7].
Stool specimens from 60 individuals were analysed for parasites with the use of standard techniques. Forty-two out of these 60 individuals were positive for Cryptosporidium spp., five of whom had only visited a hospital indoor pool used for rehabilitative therapy which was located about 3 km from the outdoor pool. This indoor pool was rather small (12×5 m) and was only used for outpatient rehabilitation in the daytime. In the evenings, the pool was open for a swimming school. The five children who had visited the indoor pool were aged 3–5 years, previously healthy, and had no other risk factors for cryptosporidiosis. They had no other relationships to each other than just having swum in the same pool. The dates of onset of symptoms for cases associated with the outdoor pool were from 22 August to 5 September, and for cases associated with the indoor pool from 29 August to 6 September (Fig. 1).
Date of onset of gastrointestinal symptoms in 147 primary cases associated to an outdoor swimming-pool outbreak (□), and five cases associated to an indoor swimming-pool outbreak (■).
Stool specimens from 30 individuals positive for Cryptosporidium spp. were fixed with ethanol [Reference Jongwutiwes8] and saved for molecular analysis. The remaining 12 samples were not available for further analysis. Two of the 30 saved specimens were from the individuals who had visited the indoor pool. The study was approved by the Ethics Committee of the Karolinska Institute (Solna, Sweden).
DNA purification
Parasite oocysts were disrupted with a bead beater, and total DNA was isolated from the sporozoites with the QIAmp DNA mini-kit (Qiagen, Hilden, Germany) according to the manufacturer's protocol.
PCR–RFLP analysis and DNA sequencing
The COWP gene was amplified by PCR with the primers cry-9 (5′-GGA CTG AAA TAC AGG CAT TAT CTT G-3′) and cry-15 (5′-GTA GAT AAT GGA AGA GAT TGT G-3′) as previously described [Reference Spano9]. For the restriction fragment analysis, PCR products were digested with RsaI (New England Biolabs, Beverly, MA, USA) and then analysed on agarose gels.
For species verification, the18S ribosomal RNA gene was amplified as previously described [Reference Enemark10] using the primers Cp18S-108 (5′-GTT ATA GTT TAC TTG ATA ATC TT-3′) and Cp18S-1031 (5′-TGA AGG AGT AAG GAA CAA CC-3′). The resulting amplicons were sequenced with the ABI PRISM Big Dye Terminator v3.0 (Applied Biosystems, Foster City, CA, USA). The reaction products were purified and concentrated through ethanol and sodium acetate precipitation, and the resulting pellets were resuspended in formamide before product separation was performed on an ABI PRISM 3100 Genetic Analyzer (Applied Biosystems). The DNA sequences were processed using the Vector NTI program suite 9 (Informax Inc., Oxford, UK). Edited sequences were compared to public databases (GenBank non-redundant nucleotide databases) using the BLAST family of algorithms with default parameters.
Mini- and microsatellite analysis
To further elucidate the genetic status of the strains involved in the outbreak, we tested the samples with three different micro- and minisatellite markers: MS1, TP14, and GP15. We used basically the same methods as previously described by Mallon et al. [Reference Mallon11, Reference Mallon12]. The sequences of the primers used for amplification were: MS1B: 5′-AAT TAG TCG ACC TCT CAA CAG TTG G-3′, and MS1D: 5′-GGA ACA CCA TCC AAG AAC CAA AGG T-3′; TP14C: 5′-CTA ACG TTC ACA GCC AAC AGT ACC-3′, and TP14D: 5′-GTA CAG CTC CTG TTC CTG TTG-3′; GP15A: 5′-GCC GTT CCA CTC AGA GGA AC-3′, and GP15E: 5′-CAC ATT ACA AAT GAA GTG CCG C-3′. The MS1 is based on a repeat region of the hsp70 gene [Reference Khramtsov13], and GP15 is based on a gene coding for a glycoprotein [Reference Strong, Gut and Nelson14]. The numbers and sizes of the alleles amplified in each sample were determined by using one primer of each pair labelled with 6-FAM, and followed by analysis on an ABI PRISM 3100 Genetic Analyzer with the aid of the GeneScan Analysis Software (Applied Biosystems).
All molecular tests were performed on coded samples, which implies that the investigator had no knowledge of the epidemiological data of the individual cases.
RESULTS
Species identification
Based on the size of the restriction fragments, all tested isolates could be characterized as Cryptosporidium parvum. Amplification of a part of the 18S rRNA gene, followed by sequence analysis of five randomly selected samples, including one connected with the indoor pool, showed that all sequenced samples had sequences identical to the C. parvum 18S rRNA gene.
Molecular subtyping
Analysis of MS1, corresponding to a repeat region of the hsp70 gene [Reference Khramtsov13], was successful for all samples. In the MS1 analysis, 28 of the samples had a 351 bp allele whereas two samples had a 362 bp allele. Further analysis with the microsatellite GP15, corresponding to a region of a highly polymorphic glycoprotein antigen described by Strong et al. [Reference Strong, Gut and Nelson14], showed that the latter two samples had a 330 bp allele whereas seven of the samples had a 305 bp allele. For the remaining 21 samples we could not get any positive results, due low sensitivity of the PCR. Attempts to use other types of buffer systems in the PCR, as suggested by Mallon et al. [Reference Mallon11, Reference Mallon12], were unsuccessful. The final analysis used the TP14 marker, and once again the two divergent samples had a unique allele (322 bp) compared to the remaining 22 samples (313 bp) that gave positive results.
Decoding of the samples showed that the two unique samples originated from the individuals who had just visited the indoor pool (Fig. 2). The rest of the samples originated from patients who were linked to the outdoor swimming pool, either directly or as secondary cases (Fig. 2).
Multilocus genotyping profiles for samples from patients that had used the outdoor swimming pool, secondary cases linked to the outdoor swimming pool, and patients who had used the indoor swimming pool. Three loci were analysed, TP14 (
), GP15 (□) and MS1 (■). The relative sizes of the different genotypes are shown.
DISCUSSION
Many molecular methods have been used for genotyping or fingerprinting strains of different microorganisms for epidemiological purposes. Such methods are of great value in contact tracing as well as outbreak investigations, as a complement to descriptive and analytical epidemiology. The use of molecular techniques is, however, highly dependent on sufficient discriminatory power, typability and reproducibility. A relationship between isolates from samples of two individuals does not, therefore, necessarily mean a connection if there is no epidemiological relationship. However, if two isolates are of different genotypes, it is unlikely that they are related to the same source.
The use of molecular tools to study Cryptosporidium spp. has not only reshaped the taxonomy considerably, but it has also increased our understanding about the epidemiology of the disease [Reference Xiao15, Reference Caccio16]. Currently, at least seven species (C. hominis, C. parvum, C. meleagridis, C. felis, C. canis, C. suis and C. muris) are associated with human disease. In addition, two Cryptosporidium spp. genotypes, monkey and cervine, have been found in human samples. C. hominis and C. parvum dominate the number of reported cases of human cryptosporidiosis [Reference Caccio16]. Interestingly, increasing numbers of cases of cryptosporidiosis are attributed to C. meleagridis, a species originally described in turkeys.
In the present study, we performed the first genotyping of a Cryptosporidium pool-associated outbreak in Sweden. All isolates where typing could be done contained C. parvum, suggesting that animal contamination of the water was possible, although a human source was more likely. This is in keeping with previous observations, where it has been shown that many C. parvum infections originate from humans [Reference Xiao15]. However, population genetics studies using mini- and microsatellite markers also suggest that there might be a strong bias for substructuring between various host species [Reference Mallon11, Reference Mallon12]. This implies that there might be human specific multi-locus genotypes (MLG) of C. parvum and other MLG that are found in more than one species.
Molecular typing is frequently used for species identification of Cryptosporidium in the Nordic countries as illustrated by two recent outbreaks: one in Norway and one in Denmark [Reference Robertson17, Reference Ethelberg18]. In the Norwegian outbreak C. parvum was identified as the cause of disease, both in calves as well as humans who had been in contact with the animals [Reference Robertson17]. In the Danish outbreak a large number of people were infected by C. hominis, as identified by PCR [Reference Ethelberg18]. The source of infection was probably raw carrots, kept in water, which were served in a company canteen [Reference Ethelberg18]. No further subgenotyping was performed on the samples from any of these two outbreaks. On the other hand, depending on the species, additional genotyping might be more or less complex [Reference Hunter19]. Hunter et al. suggest that even three loci microsatellite analyses are not discriminatory enough for C. hominis. In contrast, they conclude that the use of three loci have a very high discriminatory power for C. parvum, due to higher strain diversity [Reference Hunter19]. This latter observation is also in keeping with our results.
A troublesome observation in our outbreak report was the fact that five individuals who had not swum in the outdoor pool, but in an adjacent indoor pool, got cryptosporidiosis during the same period. These five people had not, as far as is known, had any contact with individuals involved in the outbreak. Therefore, whether there was any connection between these two episodes became a matter of discussion. With the use of subgenotyping based on microsatellite methods, a possible association between the two outbreaks could, however, be ruled out. Samples originating from the patients that had been to the indoor pool corresponded to MLG4, previously described by Mallon et al. [Reference Mallon11]. This MLG belongs to subgroup B, which encompasses not only human isolates but also bovine and ovine isolates [Reference Mallon12]. However, to our knowledge MLG4 isolates have not been recovered from any other source than humans up to now. Given the circumstances under which the individuals seem to have been infected, it is unlikely that the contamination of the therapy pool was of animal origin. The samples from visitors to the large public outdoor swimming pool were not of any MLG previously described [Reference Mallon11, Reference Mallon12].
Epidemiological investigation and sampling from infected individuals strongly suggests that the swimming pools were indeed the sources of these outbreaks, although no oocysts were found in water samples from the pools [Reference Insulander7]. Such outbreaks have been frequently reported and unlike many other microorganisms Cryptosporidium oocysts are highly resistant to chlorine which combined with a low infectious dose make them ideal for waterborne transmission [Reference Furtado2–Reference Insulander7]. The reason why two separate swimming-pool outbreaks occurred so close to each other both in space and time is difficult to explain. The late summer was very hot in Sweden for this year and the pools were crowded with people, which might have increased the risk of faecal contamination. The maintenance of the pools seemed adequate regarding filtration systems and chlorination and microbiological examination of the water showed no contamination [Reference Insulander7]. However, incidences of loose stool were observed at several occasions at least in the outdoor swimming pool and the recommendations for the management of faecal accidents [20] had not been followed. Both the outdoor and indoor pools were closed immediately when the outbreaks were detected. The free chlorine concentration was raised to 3 mg/l (pH 7·7–7·4) and the filters were back-flushed once a day for 7 days. Before the pools were reopened, the free chlorine concentration was raised to 15 mg/l for 16 h [Reference Insulander7]. The importance of following guidelines for management of faecal accidents was strictly emphasized and information of correct hygiene precautions for individuals visiting public swimming pools was instituted. The fact that the indoor pool was used for outpatient rehabilitation during daytime and children's swimming school in the evenings may raise concern and further underscores the importance of hygienic precautions, e.g. never swimming until full recovery from a diarrhoeal illness, implementation of adequate toilet, hand-washing and nappy-changing routines, and warnings against swallowing pool water.
One limitation in our study is that faecal Cryptosporidium isolates from only two of the five children who had swum in the indoor pool were available for molecular subtyping. However, the fact that both isolates were of the same unique subtype, differing from the subtype associated with the outdoor swimming pool, strongly suggests that there were in fact two sources of exposure.
In conclusion, our report clearly shows that the use of mini- and microsatellites as markers for molecular fingerprinting of C. parvum isolates can be of great value in epidemiological investigations of outbreaks due to Cryptosporidium, and can be used to separate outbreaks that otherwise would be difficult, if not impossible, to distinguish.
ACKNOWLEDGEMENTS
We thank Katarina Näslund and Lillemor Carlsson for valuable technical assistance. This work is part of Med-Vet-Net (FOOD-CT-2004-506122), a Network of Excellence for the Integrated Research on the Prevention and Control of Zoonoses, funded within the 6th European Union Framework Programme.
DECLARATION OF INTEREST
None.
