A whole water catchment approach to investigating the origin and distribution of Cryptosporidium species

Authors


Guy Robinson, UK Cryptosporidium Reference Unit, Public Health Wales Microbiology, Singleton Hospital, Swansea SA2 8QA, UK. E-mail: Guy.Robinson@wales.nhs.uk

Abstract

Aims:  Investigating the distribution and origin of Cryptosporidium species in a water catchment affected by destocking and restocking of livestock as a result of a foot and mouth disease epidemic.

Methods and Results:  Surface water, livestock and wildlife samples were screened for Cryptosporidium and oocysts characterised by sequencing SSU rRNA and COWP loci, and fragment analysis of ML1, ML2 and GP60 microsatellite loci. Oocyst concentrations in water samples (0–20·29 per 10 l) were related to rainfall events, amount of rainfall and topography. There was no detectable impact from catchment restocking. Cryptosporidium spp. found in water were indicative of livestock (Cryptosporidium andersoni and Cryptosporidium parvum) and wildlife (novel genotypes) sources. However, C. andersoni was not found in any animals sampled. Calf infections were age related; C. parvum was significantly more common in younger animals (<4 weeks old). Older calves shared Cryptosporidium bovis, Cryptosporidium ryanae and C. parvum. Wildlife shed C. parvum, Cryptosporidium ubiquitum, muskrat genotype II and deer genotype.

Conclusions:  Several factors affect the occurrence of Cryptosporidium within a catchment. In addition to farmed and wild animal hosts, topography and rainfall patterns are particularly important.

Significance and Impact of the Study:  These factors must be considered when undertaking risk-based water safety plans.

Introduction

Cryptosporidiosis is a diarrhoeal disease caused by the protozoan parasite Cryptosporidium. While two species predominate in human disease (Cryptosporidium hominis and Cryptosporidium parvum), C. parvum is the principal cause of gastrointestinal cryptosporidiosis in livestock (Xiao et al. 2004). Human–human cycles of infection involving C. hominis or C. parvum occur and contrast with zoonotic cycles in which C. parvum is transmitted between animals and man (Casemore 1990; Mallon et al. 2003a; Chalmers 2004). Several other Cryptosporidium species and genotypes, many named after the host from which they were originally recovered, have been found in humans although the actual public health significance of some of these is not clear (Robinson et al. 2008b; Davies et al. 2009). However, the UK drinking water related outbreak caused by Cryptosporidium cuniculus should act as a caution against assuming these unusual species and genotypes are not significant (Chalmers et al. 2009). Cryptosporidium has emerged in recent years as a major public health problem through the risks of contaminated drinking water (Karanis et al. 2007), this remains a problem (Bridge et al. 2010; Mason et al. 2010; ECDC, 2011). The water industry internationally has had to radically revise both risk assessments and water treatment processes to address this complex issue. A better understanding of the distribution, patterns and sources of different species is an essential step in more effective risk assessment and public health control.

Robust Cryptosporidium oocysts are shed in the faeces of an infected host, often in large numbers, and enable the survival of the organism in the environment, particularly in water at ambient temperatures (Meinhardt et al. 1996). This facilitates multiple pathways of transmission in addition to direct faecal–oral contact, with indirect routes involving contaminated water, food and fomites (Nichols 2007). Methods for the detection of Cryptosporidium oocysts in water usually involve filtering large volumes, concentrating oocysts by immunomagnetic separation (IMS) and enumeration by immunofluorescence (IF) microscopy (United States Environmental Protection Agency 2005; The Environment Agency 2010). To determine the public health significance of Cryptosporidium in environmental samples, the species or genotype must be identified by molecular analysis, but this is not part of routine monitoring protocols. Many of the species/genotypes found in source waters are believed to originate from wild animal sources and are not common in human or livestock infections (Zhou et al. 2004; Feng et al. 2007). Molecular characterisation of the species or genotypes present can assist in source apportionment by suggesting the origin of contamination and help direct environmental investigations (Chalmers et al. 2010).

During spring 2001, national surveillance figures for England and Wales showed a marked decrease in the numbers of cases of human cryptosporidiosis, particularly C. parvum, which coincided with interventions to control the foot and mouth disease (FMD) epidemic (Hunter et al. 2003; Smerdon et al. 2003). Between 20th February 2001 and 30th September 2001, these included the slaughter of more than 6 million animals, and restriction of animal movements and public access to countryside areas. However, a sustained reduction in the number of cases of C. parvum in the spring has been observed each year, long after restrictions were lifted, and has been linked to improvements in drinking water quality exemplified by measures undertaken in north-west England (Goh et al. 2005; Sopwith et al. 2005). Thus, there is a complex combination of elements which, either individually or in concert, affect the epidemiology of human cryptosporidiosis. Intraspecies investigations of C. parvum and C. hominis using multilocus typing approaches are helpful for further epidemiological and transmission studies (Mallon et al. 2003b; Hunter et al. 2007, 2008).

Following the FMD epidemic, restocking of farms from external sources occurred, including those that previously operated closed husbandry systems. The effect of the introduction of animals to such sites on the distribution of Cryptosporidium spp. locally and within a catchment was not known. To investigate the distribution and diversity of Cryptosporidium spp. in an FMD-affected catchment, we undertook a whole catchment-based study of surface waters, and farmed and wild animal populations in the River Caldew catchment, north-west England.

Materials and Methods

Selection of study sites

The River Caldew catchment in Cumbria, north-west England, was chosen because of the high local impact of FMD, which resulted in some subcatchments being almost entirely destocked in 2001 (Sanders et al. 2004). Four farms were recruited into the study in February and March 2002: two denuded of livestock but subsequently restocked (farms A and B) and two continually stocked (farms C and D). Farm A had dairy cattle and sheep, farms B and D were predominantly dairy farms and farm C was predominantly a sheep farm. Farmers were assured confidentiality and offered a £100 incentive for participation. Recruitment of farms was a sensitive issue owing to the impact of the FMD outbreak and concerns over the use of data from the study. Access to restocking farms was delayed because of FMD movement restrictions, and it was evident from an initial survey in January 2002 that restocking on some farms had already commenced. However, as animals were being over-wintered indoors and slurry spreading was still banned, it was deduced that the outdoor environment would represent the destocked condition. These initial observations were confirmed by a survey of farm practices carried out as part of a study of the microbiological quality of surface water being undertaken in the catchment by the Environment Agency (EA) (Sanders et al. 2004), a further reason why this catchment was selected for the Cryptosporidium study.

In total, 11 surface water sites throughout the catchment were identified for Cryptosporidium sampling (Fig. 1) and were categorised as either upland (sites 1, 2, 3, 4 and 9) or lowland (5, 6, 7, 8, 10 and 11). Sites 7, 9, 10 and 11 were selected because of their proximity and downstream nature to farms A, C, B and D, respectively. As farm C straddled the catchment border with animals wandering the fells in both catchments, site 9 was actually just inside the neighbouring catchment because it was the best site to represent water originating from the actual farm yard (site 1 on the catchment was accessible by animals from farm C). General assessment of weather conditions and land use was made subjectively at the time of sampling. Historical weather data were obtained for the period of the study from EA administered local sites. Detailed land use data, geology and soil maps were also obtained from Ordnance Survey and the EA in hard copy and GIS formats.

Figure 1.

 Location of Cryptosporidium monitoring points, farm holdings and waste water treatment works (other potential sewage sources such as septic tanks or package plants are not shown) in the River Caldew catchment. To protect anonymity, the locations of sampled farms are not shown. The course of the River Caldew is shown to its confluence with the River Eden in Carlisle. (inline image1) Sample site; (inline image) Farm holding and (inline image) Waste water treatment works.

Surface water sampling and testing for Cryptosporidium oocysts

Surface water samples were collected in two phases approximately 1 year apart: Phase 1 from January to March 2002 and Phase 2 from February to April 2003. A set of three samples were collected for each site to target base flow during dry periods and rainfall driven and high flow conditions (with the exception of sites 9–11 during Phase 1 where only prerainfall samples were collected).

Sampling and oocyst enumeration were undertaken using standard methods (The Environment Agency 2010). Briefly, water samples of up to 500 l were collected by cartridge filtration (Envirochek HV® filters; Pall Corp., Portsmouth, UK). A portable pump run from a 12-V battery was used to abstract the water, which was returned to the river following filtration, at an optimum flow rate of 4 l min−1. Multiple filters were used if the water was turbid, or where turbidity was excessive, multiple 10-l grab samples were obtained for flat-bed membrane filtration in the laboratory. Particles were eluted from the filter, concentrated using IMS (Isolate™; TCS Biosciences, Botolph Claydon, UK) and stored in distilled water. A 10% portion of each pellet was removed, and the beads were disassociated to evaluate the numbers of Cryptosporidium oocysts present by immunofluoresence microscopy (IFM) (CryptoCel; TCS Biosciences) with the confirmation of identity by differential interference contrast microscopy (DIC) inspection and 4′,6-diamidino-2-phenylindole (DAPI) (Sigma-Aldrich Co. Ltd, Dorset, UK) staining at CREH Analytical Ltd.

The remaining 90% of each pellet and the microscopy slides were sent to the UK Cryptosporidium Reference Unit (UK CRU) for DNA extraction prior to species identification and subtyping. DNA was extracted from the IMS pellet by disrupting oocysts with three freeze–thaw cycles (cardice/methanol bath – 100°C) followed by a 30-min proteinase K digest at 56°C (Robinson 2005; Chalmers et al. 2009). DNA was purified using QIAamp® DNA Mini Kit spin columns (Qiagen, Crawley, UK) as described in the manufacturer’s instructions. DNA was stored at −20°C until testing.

Animal faecal sampling, testing for Cryptosporidium oocysts and DNA extraction

Animal sampling was undertaken in two phases, approximately 1 year apart. Phase 1 sampling was carried out between February and March 2002. To sample lambs that had not been born at the time of the first visit and to increase the numbers of samples from calves, continually stocked farms were revisited during late April and early May 2002. Phase 2 sampling was carried out between February and April 2003, which coincided with the peak lambing and calving period.

To comply with the Animals (Scientific Procedures) Act (Anon 1986), livestock populations were sampled by collecting discrete, freshly voided, faeces from the ground into separate polythene bags. Numbers of samples were calculated to estimate prevalence at 95% confidence, with 5% absolute error (Cannon and Rowe 1982) and over-sampling of the population to account for droppings rather than per rectum sampling from individual animals. Advice was sought from the farmers about the presence of wildlife, and faeces of foxes, badgers, deer and game birds were identified on the ground and collected on an ad hoc basis.

Faecal samples were concentrated using a modified formol-ether method (Casemore et al. 1985), incorporating the Parasep faecal parasite concentrator (DiaSys, Wokingham, UK) and substituting ether for ethyl acetate, prior to staining by IF and DAPI and inspection by epifluorescence microscopy (Chalmers 1996; Robinson 2005). Concentrates from Cryptosporidium-positive faecal samples were processed for typing using the method routinely used at the UK CRU for clinical samples. Briefly, oocysts were disrupted by incubation at 100°C for 60 min and digestion with proteinase K and lysis buffer (Buffer AL; Qiagen), and DNA was purified using a QIAamp® DNA Mini Kit (Qiagen) and stored at −20°C (Elwin et al. 2001). If low numbers of oocysts were present, these were recovered by IMS (Isolate™; TCS Biosciences) prior to DNA extraction (Robinson et al. 2008a).

Cryptosporidium species identification and subtyping

Cryptosporidium species were initially determined using nested PCR-RFLP analysis of the SSU rRNA gene (Xiao et al. 2001). To differentiate the majority of Cryptosporidium species, secondary PCR products were digested by restriction enzymes SspI and VspI. Cryptosporidium andersoni/Cryptosporidium muris were differentiated using DdeI. The restriction fragments were separated on a 2% agarose gel and visualised by SYBR Green I (Sigma) staining (Elwin and Chalmers 2008; Robinson et al. 2008b). To augment amplicons where only faint bands were produced, another round of the secondary PCR was undertaken using 5 μl of the secondary product as a template and products were digested as described earlier.

Where the SSU rRNA PCR-RFLP analysis gave results indicating that gastric Cryptosporidium were present, but the smaller intestinal Cryptosporidium oocysts were also seen during microscopic examination, a further PCR–RFLP was carried out targeting the COWP gene (Spano et al. 1997), because these primers do not amplify the gastric species (UK CRU unpublished data).

Cryptosporidium spp. were confirmed by bidirectional DNA sequencing (Source Bioscience, Cambridge, UK) of the PCR products. Sequences were analysed using ChromasPro (Technelysium Pty Ltd, Brisbane, Australia) and compared with all GenBank, RefSeq, EMBL, DDBJ and PDB sequences using the NCBI BLAST algorithm. Novel environmental SSU rRNA genotypes were named according to the UK Environmental (UK E) nomenclature described by Chalmers et al. (2010). Subtypes of C. parvum and C. hominis were identified using a multilocus fragment typing (MLFT) approach to target three microsatellite markers (ML1, ML2 and GP60), as previously described (Hunter et al. 2007). Sequence alignments were generated in ClustalX 2.0 (http://www.clustal.org/download/current/) and manually edited in BioEdit 7.0.9 (http://www.mbio.ncsu.edu/BioEdit/bioedit.html). Phylogenetic analyses with other known Cryptosporidium spp. and selected relevant genotypes were carried out by a neighbour-joining analysis of the SSU rRNA gene in MEGA 5 (http://www.megasoftware.net/). Evolutionary distances were calculated using the Kimura’s two-parameter model with Plasmodium falciparum as an outgroup.

Representatives of all of the Cryptosporidium SSU rDNA sequences generated during this study have been deposited in GenBank under the accession numbers HQ822132-HQ822145.

Data analysis

All statistical analyses were performed in EpiInfo 6·04d. Sample prevalence data were compared using a chi-squared Mantel–Haenszel or two-tailed Fisher’s exact test and differences in positive sample oocyst counts were evaluated using a Kruskal–Wallis H test. All statistical tests were assessed at α = 0·05 (i.e. 95% confidence level).

Results

Distribution of Cryptosporidium in surface water

A total of 60 scheduled surface water samples were collected: 27 in Phase 1 and 33 in Phase 2 (Table 1). During Phase 1, 16 of 27 (59%) of the samples were positive for Cryptosporidium oocysts, significantly more than the 11 of 33 (33%) during Phase 2 (χ2 = 3·97, df = 1 and P < 0·05). Rainfall-driven episodic fluxes of oocysts in stream waters varied between phases. Frequency of detection increased during both phases from base flow conditions to high flow (46–69% in Phase 1 and 23–55% in Phase 2), but this was not statistically significant (i.e. P = 0·26 and P = 0·12, respectively). In Phase 1, mean Cryptosporidium counts in positive samples increased from 1·81 per 10 l under base flow conditions to 4·46 per 10 l in high flow conditions (Table 1). In Phase 2, mean oocyst counts increased from 2·19 per 10 l in base flow samples to 7·12 per 10 l during high flow (Table 1). However, none of these increases from either base to high flow or Phase 1 to Phase 2 (when considering flow rate or combined) were statistically significant. Average (mean) oocyst recoveries from the filtration and IMS process were 40% (data not shown).

Table 1. Cryptosporidium concentrations, species and subtypes detected in surface water samples at the eleven sampling sites during Phases 1 and 2
SitePhase 1Phase 2
Base flowHigh flow 1High flow 2Base flow 1Base flow 2High flow
Oocysts (10 l−1)Cryptosporidium species (MLFT)Oocysts (10 l−1)Cryptosporidium species (MLFT)Oocysts (10 l−1)Cryptosporidium species (MLFT)Oocysts (10 l−1)Cryptosporidium species (MLFT)Oocysts (10 l−1)Cryptosporidium species (MLFT)Oocysts (10 l−1)Cryptosporidium species (MLFT)
  1. MLFT, multi locus fragment type; NS, not sampled.

  2. *DNA amplified although no oocysts were identified by IFAT.

Upland
10·00 0·22Not amplified0·00 0·00 0·00 0·00 
20·22Not Amplified3·03UK E80·00 0·00 0·00 0·00 
30·00 1·99Not amplified0·60Not Amplified0·00 0·00 0·00 
40·37Cryptosporidium andersoni1·53C. andersoni0·00 0·00 0·00C. andersoni*3·54C. andersoni and Novel COWP genotype
90·00 NS NS 0·00 0·00 0·00 
Lowland
51·66C. andersoni20·29Cryptosporidium parvum4·41C. andersoni0·00 0·00 1·52Not Amplified
66·24Not amplified5·71Not amplified1·67Not Amplified0·00 2·07C. parvum0·00 
70·00 5·93C. parvum0·00 0·00 5·68Not Amplified16·09C. parvum (P5 & P33) and C. andersoni
80·57Not amplified3·64Not amplified0·00C. andersoni*0·00 1·52Not Amplified6·90C. parvum and C. andersoni
100·00 NS NS 1·08UK E70·00C. andersoni and Novel COWP genotype*10·34C. parvum (P7) & C. andersoni
110·00 NS NS 0·00 0·60C. parvum4·30Not Amplified
Frequency of detection5/11 (46%) 11/16 (69%)   5/22 (23%)   6/11 (55%) 
Mean of positives (SD)1·81 (±2·5) 4·46 (±5·6)   2·19 (±2·0)   7·12 (±5·3) 

Oocyst concentrations ranged between 0 and 20·29 per 10 l during Phase 1 and between 0 and 16 per 10 l in Phase 2. The largest concentrations during both phases were detected in and downstream of the Roe Beck subcatchment (sites 5, 6 and 7), and these sites showed a pattern of decreasing concentration after the confluence with the River Caldew and its upland headwaters (Fig. 2). Oocysts were found in both upland and lowland sites but more frequently and in greater counts at the sites influenced by lowland terrain. The influence of topography on Cryptosporidium occurrence was seen in both phases, but was only significant during Phase 2 when only 2/15 (7%) of upland samples were positive for oocysts compared with 10/18 (56%) of lowland samples (χ2 = 6·11, df = 1 and P = 0·01).

Figure 2.

 Comparison of concentration of Cryptosporidium oocysts detected in upland and lowland-influenced surface waters. Phase 1: (inline image) Base flow; (inline image) High flow 1 and (inline image) High flow 2; (inline image) Phase 2: Base flow 1; (inline image) Base flow 2 and (inline image) High flow.

In addition to the scheduled samples, eight supplementary samples were collected from land drains or sites close to farms at the time of animal sampling, of which three from Phase 2 were positive by IFM. One sample from just downstream of a lowland dairy farm was positive for nine oocysts per 10 l, but we were unable to confirm the species by molecular methods. The other two samples were linked to a different lowland farm, one river sample with 49 oocysts per 10 l and the other, a land drain, with 4000 oocysts per 10 l.

Occurrence of Cryptosporidium in animals

A total of 1272 faecal samples from farmed animals were collected during the study (570 from Phase 1 and 702 from Phase 2). In addition, 115 faecal droppings were collected from wild animals (74 from Phase 1 and 41 from Phase 2) (Table 2). Cryptosporidium was detected in 21 (4%) livestock animals in Phase 1 and 41 (6%) in Phase 2, and in wild animals, the prevalence was 7% (five samples) in Phase 1 and 5% (two samples) in Phase 2 (Table 2). These changes in prevalence from Phase 1 to Phase 2 were not significant for either farmed or wild animals (χ2 = 3·15, df = 1, P = 0·08 and two-tailed Fisher’s exact test, P = 1·00, respectively).

Table 2. Cryptosporidium species and subtypes in animal samples from farms continually stocked and those restocked following a foot-and-mouth disease cull
Host (Age)FarmIFAT positive (PCR positive)Cryptosporidium species (subtype)
Phase 1Phase 2Phase 1Phase 2
  1. *Similar at SSU rRNA gene to C. suis (99·7%), but very different at the HSP70 and actin loci (UK CRU data).

  2. Based on SSU rRNA gene PCR–RFLP analysis only.

Adult cattleContinuous0/550/63  
 Restocked1/1010/1191 × Unknown species* 
CalvesContinuous19/59 (17/19)22/99 (20/22)  
 0–3 weeks    14 × Cryptosporidium parvum (P13)
 4–8 weeks   8 × Cryptosporidium bovis 
3 × Cryptosporidium ryanae
3 × Mixed
(C. bovis & C. ryanae)
 Unknown   2 × C. bovis4 × C. parvum (P13)
1 × UK E71 × C. bovis
1 × Mixed
(C. parvum & C. bovis)
 Restocked0/715/37 (14/15)  
 0–3 weeks    10 × C. parvum (P5)
 4–8 weeks    1 × C. parvum (P5)
1 × C. bovis
2 × Mixed
(C. bovis & C. parvum)
Adult sheepContinuous0/1803/131 (1/3) C. parvum
 Restocked0/320/37  
LambsContinuous1/135 (1/1)0/197C. parvum (P7) 
 Restocked0/11/22 (0/1)  
FoxesContinuous2/17 (1/2)0/3C. bovis 
 Restocked2/5 (2/2)0/5C. parvum (P46) and Muskrat genotype II 
 Site 6NS0/1  
PheasantsContinuousNSNS  
 Restocked0/10NS  
 Site 6NS0/2  
Roe DeerContinuous1/2 (0/1)1/3 (1/1) Cryptosporidium ubiquitum
 Restocked0/290/11  
 Site 6NS1/1 (1/1) Deer genotype
BadgerContinuous0/50/5  
 Restocked0/30/6  
 Site 6NS0/4  
RabbitContinuous0/3NS  
 RestockedNSNS  
 Site 6NSNS  

There was an increase in positive livestock samples on restocking farms from Phase 1 (1/141, 1%) to Phase 2 (16/212, 8%) which was highly significant (χ2 = 8·61, df = 1 and P = 0·003), unlike those from continually stocked farms (Phase 1: 20/429, 5%; Phase 2: 25/490, 5%; χ2 = 0·10, df = 1 and P = 0·76).

Only seven from 115 samples collected from wild animals were positive for Cryptosporidium. The two animal groups positive for Cryptosporidium were fox (31 samples collected) and deer (46 samples collected). The prevalence of Cryptosporidium in wild animals on these farms was therefore fairly low.

Molecular characterisation of Cryptosporidium spp. within the catchment

Five known species (C. parvum, C. andersoni, Cryptosporidium bovis, Cryptosporidium ryanae and Cryptosporidium ubiquitum) and six genotypes (deer, muskrat II and four novel genotypes) were detected in surface water or animal faeces (Tables 1 and 2, Fig. 3). In three surface water samples, C. andersoni was detected by SSU rRNA gene sequencing but when testing the second locus (COWP PCR-RFLP), patterns similar to C. hominis or C. cuniculus were identified. One produced poor sequence data, but the other two had novel sequences that although closely related to C. hominis and C. cuniculus were distinctly different (99·1%, 439/443 bp; 99·3%, 440/443 bp; with different polymorphisms in each isolate). The limited COWP sequence data on the nucleotide databases searched restricted their identification. Of two other novel genotypes identified by SSU rRNA gene sequencing, one isolate (UK E7) found in both a surface water sample and a calf was most closely related to the fox genotype (99·5% and 775/779 bp) (Fig. 3). The other novel sequence (UK E8) was most closely related to UK E2 and UK E3 (both 99·9% and 735/736 bp but with different SNPs), previously only found in water (Chalmers et al. 2010) (Fig. 3). Phylogenetic analyses place these three genotypes (UK E8, UK E2 and UK E3) in a single cluster with other Cryptosporidium isolates (SW2, SW1, UK E1 and UK E5) of undefined species identified in UK waters during studies in North Wales and Scotland (Chalmers et al. 2010; Nichols et al. 2010).

Figure 3.

 Phylogenetic relationships between Cryptosporidium spp., selected genotypes and the Cryptosporidium isolates characterised from water, livestock and wildlife in the Caldew catchment (a Cryptosporidium andersoni isolate from this study was not included in the analysis because of the short length of quality sequence obtained) as inferred by neighbour-joining analysis of the partial SSU rRNA gene. The percentage of replicate trees in which the associated taxa clustered together in the bootstrap test (1000 replicates) was shown next to the branches (>50% only). The evolutionary distances were computed using the Kimura’s 2-parameter method with Plasmodium falciparum as an out-group. There were a total of 587 base positions in the final data set.

The species of Cryptosporidium most frequently found in surface waters was C. andersoni, on 11 occasions at five different sites (Table 1). All of the C. andersoni-positive samples were from lowland sites, with the exception of site 4 which is in the lower region of the upland area. There was no significant difference in the frequency that C. andersoni was found during Phase 1 compared with Phase 2 (two-tailed Fisher’s exact test P = 1·00). Interestingly, C. andersoni was not detected in any of the animal samples.

Cryptosporidium parvum was also widely distributed in the catchment and was found in surface waters on seven occasions at six different sites (Table 1). The topography of the catchment was significant with all C. parvum-positive samples from lowland sites (two-tailed Fisher’s exact test P = 0·01). During Phase 1, only two C. parvum-positive samples were identified in surface water tested, which increased in Phase 2 to five and may reflect the greater number of livestock on the catchment during the second year, but this was not statistically significant (two-tailed Fisher’s exact test P = 0·44). There was a wide distribution of C. parvum in animals throughout the catchment, being found in calves, sheep and a fox, but the highest prevalence was in calves (Table 2). Cryptosporidium parvum was significantly more common in young calves <4 weeks old than other Cryptosporidium species (χ2 = 37·17, df = 1 and P = 0·000). Cryptosporidium bovis predominated in older calves, although C. ryanae, C. parvum and mixed infections were also detected (Table 2). The only adult bovine found to be shedding oocysts was infected with an isolate closely related to C. suis (99·8% [802/804 bp] at the SSU rRNA gene, but clearly different at the hsp70 and actin genes at 97·8% [394/403 bp] and 98·1% [817/833 bp], respectively, UK CRU unpublished data, GenBank Accessions: HQ822146-HQ822148).

Infections in sheep were less common than cattle, and only two infections with C. parvum (one based on a restriction fragment length polymorphism pattern only) were identified, one in a lamb and the other in an adult (Table 2).

Foxes were hosts to C. parvum, C. bovis and muskrat genotype II, and it is possible that these three species were spurious infections linked to the diet (prey animals) of those animals. The only other wildlife found positive for Cryptosporidium was roe deer, where individual animals were found positive with C. ubiquitum and the deer genotype (Table 2).

Cryptosporidium parvum was the only ubiquitous species, being found in farmed animals, wildlife and water samples. Five different MLFTs (P5, P7, P13, P33 and P46) were identified in 31 of 41 (76%) C. parvum-positive samples (Tables 1 and 2). Of the 10 isolates that could not be assigned an MLFT, four were not typable at all three loci, three were only typable at two loci and three were typable only at a single locus. Those partially typable from farms with C. parvum MLFTs in other animals matched fragment sizes at the loci that did amplify. Two of the MLFTs found in this study were novel to the typing scheme described previously (Hunter et al. 2007) with the following ML1, ML2 and GP60 fragment sizes, respectively: P33 = 242 bp, 229 bp, 350 bp; P46 = 242 bp, 215 bp, 344 bp. MLFTs P13, P33 and P46 were only identified in single sample types: farmed animals, water and wildlife, respectively. However, P5 and P7 were found in both water and farmed animals. The farm with P5 animal infections was shown to be contributing heavily to the environment; this isolate was detected in an initial supplementary sample from the river just downstream of the farm and again a month later in a land drain sample contributing 4000 P5 oocysts per 10 l into the river. The farm with P7-infected animals was only found positive during Phase 1, whereas the water sample positive for P7 was collected during Phase 2 in a different subcatchment.

Discussion

In this study, we investigated several factors (including geography, hosts and restocking following FMD culling) potentially affecting the distribution and origin of Cryptosporidium species throughout a whole water catchment. Such factors should be considered when undertaking risk-based water safety plans (World Health Organisation 2005). While it was not possible to test more farms or wildlife during this study, we enrolled farms that were representative of the settings, stocking and husbandry methods within the catchment. We found that topography and rainfall patterns were particularly important to the occurrence of oocysts, and the host species influenced the Cryptosporidium species or genotypes detected.

Our molecular test strategy in 2001 and 2002 may have underestimated the diversity of Cryptosporidium species found, as we, like others at this time, only tested a single DNA replicate per PCR (Lowery et al. 2000, 2001). Testing several replicates at each locus with optimal amounts of DNA has since been shown to help identify different species when present within the same sample (Ruecker et al. 2005). This strategy is now routinely used at the UK Cryptosporidium Reference Unit (Chalmers et al. 2010).

Links between Cryptosporidium oocyst counts in surface waters and rainfall have been established previously (Hansen and Ongerth 1991; Kay et al. 2007), and in this study, targeted sampling was fruitful in generating samples for genetic analysis to assess the origin of the contamination. The effect of rainfall on oocyst detections in the catchment in our study generally showed an increase in positivity, which was associated with increased flow because of rainfall events. Many factors may influence oocyst transfer related to rainfall, for example, run-off of faecal material or disturbance of settled oocysts (Hansen and Ongerth 1991; Kay et al. 2007). As oocysts originate in the intestinal tract of animals, it is worth examining more closely the potential sources of contamination. At an individual farm level, Kemp et al. (1995) observed an increase in counts in drainage waters from a farm in Scotland during the periods of peak calving and spreading of slurry and manure. Bodley-Tickell et al. (2002) reported high frequency of Cryptosporidium detections (66%) in surface waters on a lowland farm in Warwickshire, England, where at any given time, Cryptosporidium oocysts were being shed by at least one livestock or wild animal population (Sturdee et al. 2003). There is evidence from elsewhere that wildlife contributes to oocyst counts in surface waters (Bodley-Tickell et al. 2002; Feng et al. 2007; Chalmers et al. 2010; Nichols et al. 2010). In our study, the Cryptosporidium species found in the wildlife sampled were rarely detected in the surface water samples. This may be due in part to the relatively small number of wildlife samples tested. Furthermore, very little is currently known about Cryptosporidium species and genotypes infecting or carried by UK wildlife. The main contributor of C. parvum in the surface water samples was most likely the calves, where the prevalence was highest. One of the novel water genotypes (UK E7) was also found in a calf. The other novel water genotype (UK E8) was placed by phylogenetic analyses in a cluster of Cryptosporidium isolates from two other studies in the UK, none of which have known hosts (Chalmers et al. 2010; Nichols et al. 2010).

Several other studies have described the microscopic detection of Cryptosporidium oocysts from various UK water sources (reviews in Badenoch 1990, 1995; Bouchier 1998) but did not identify the species or genotypes present because of the lack of tools available at the time. In contrast, Lowery et al. (2000) described a combined IMS-PCR method for the detection and differentiation of C. parvum and related genotypes which they used in Northern Ireland, demonstrating a low incidence of C. parvum in surface waters (Lowery et al. 2001). However, comparisons between these studies are hampered by many influencing factors such as different geography, local animals, animal husbandry and water quality, as well as differences in oocyst recovery and detection methods. Despite this, some themes still emerge.

The low occurrence of Cryptosporidium in upland waters compared with lowland surface water has been reported previously (Hansen and Ongerth 1991). A similar relationship was also seen in our study and is probably a result of the differences in topography and decreased sheep and beef cattle livestock grazing density compared with the dairy systems which predominate in the lower Caldew and Roe Beck subcatchments (Sanders et al. 2004). The highest concentrations of Cryptosporidium oocysts in our study were found in the Roe Beck subcatchment, which correspond with the highest faecal indicator organism concentrations found by Sanders et al. (2004). The decrease in concentrations between the lowest site on Roe Beck and the next site downstream on the River Caldew owing to the dilution effect with less contaminated water was also replicated in the faecal indicator organism study (Sanders et al. 2004). Unsurprisingly, all of these data indicate that the more intensively stocked dairy farms in these lowland areas of the catchment are likely to be the main source of environmental Cryptosporidium in this part of the catchment, but also demonstrate that upland waters cannot be assumed to be pristine. The increased risks of contamination because of run-off and application of farm waste to agricultural land can be minimised by following the code of good agricultural practice for farmers, growers and land managers (Defra 2009).

Cryptosporidium infections are more prevalent in young animals (Santín et al. 2004), so it is unsurprising that the vast majority of Cryptosporidium-positive faeces in the catchment were from calves. In Phase 1, all of the imported animals were adults sourced from within the UK, and the prevalence of Cryptosporidium was very low. The increase in positive livestock in Phase 2 on restocking farms is probably due to the increased number of calves sampled. There were more calves present in the catchment (Department for Environment Food and Rural Affairs June 2002 and 2003 Survey of Agriculture and Horticulture data), and the restocking farms had time to become recontaminated with Cryptosporidium, which may also partly explain the increase in C. parvum found in surface water samples during Phase 2 of the study. Although variable, prevalence in lambs is reportedly lower than in calves (Pritchard et al. 2008; Featherstone et al. 2010), and the prevalence of infections in lambs in this study was very low. This is possibly due to the extensive husbandry and outdoor rearing.

The total number of potential host livestock animals decreased as a result of the cull (Sanders et al. 2004). However, wild animals and livestock on farms unaffected by FMD remained in the catchment. Furthermore, most cattle within the catchment are overwintered indoors. Thus, the main transfer pathways between cattle and surface waters during the winter are through the spreading of slurry and farmyard manure, and through direct run-off from farm hard-standing areas, which can be a significant source of faecally derived contaminants (Edwards et al. 2008). This was observed in a sample taken from a farmyard drain in which 4000 oocysts per 10 l were detected. In addition, slurries treated with lime to control FMD before spreading may have contained detectable Cryptosporidium oocysts. There are conflicting reports regarding whether lime treatment significantly reduces oocyst viability (Zintl et al. 2009), but even nonviable oocysts are detected with the microscopic and molecular methods used in this study. Thus, a considerable input of oocysts to the catchment probably continued during and after the period of the cull which provides a likely explanation for the presence of oocysts in the water during Phase 1. The guidance provided in Defra’s code of good agricultural practice gives advice on how different agricultural practices can be undertaken while protecting and enhancing the quality of water, soil and air (Defra 2009).

Age-related patterns of Cryptosporidium infections in cattle have been described (Santín et al. 2004; Brook et al. 2009), and in our study, similar findings were seen, with C. parvum predominating in young calves up to 4 weeks old, and C. bovis and C. ryanae, or co-infections, in older animals (Table 2). In addition to these common cattle infective species, we also detected an unusual Cryptosporidium genotype in the faeces from a heifer. This currently unnamed species is identical at the SSU rRNA gene to isolates described as C. suis-like, detected in three calves in Denmark (Langkjaer et al. 2007), a postweaned calf in India (Khan et al. 2010) and from a sporadic human case in the UK (National Collection of Cryptosporidium Oocysts, UK CRU, Swansea; GenBank accession number: HQ822146), but very little is known about it.

Cryptosporidium andersoni was the most frequently detected species in water samples during both phases. The natural host range of C. andersoni is cattle (Lindsey et al. 2000), and its presence in our study implies that adult cattle play a major role in the contamination of the catchment. Even though C. andersoni is not a human pathogen, its presence indicates a link between cattle faeces and the water and should be taken seriously. However, we did not detect C. andersoni in any of the animals sampled. Interherd prevalence studies in the United States of America have shown high variability (Fayer et al. 2010; Szonyi et al. 2010) and although C. andersoni circulates within infected herds in the UK (Robinson et al. 2006), its origin in surface waters may be from a few infected herds in the catchment not sampled in our study. Further studies need to examine the interherd prevalence and transmission of C. andersoni in the UK, its veterinary and animal production significance and the effect of environmental conditions, farm yard manure and slurry management on its survival.

In addition to the larger C. andersoni oocysts, smaller oocysts were also seen in many samples indicating multiple species/genotypes were present. Even in samples where higher numbers of smaller oocysts were seen, C. andersoni was more frequently amplified using the SSU rRNA gene primers. The reason for this is unclear, and while it is possible that sequence differences influence polymerase extension during PCR, there is little evidence for this and it is more likely that C. andersoni is under-detected by IFM because of greater susceptibility to quenching of the fluorescent stain (UK CRU unpublished observations). To combat this, our strategy of examining a second locus improved the detection of other species present. Use of the COWP gene permitted additional identification of C. parvum. Furthermore, two novel genotypes were detected with the COWP primers when C. andersoni was present: although the limited amount of COWP sequence data available on GenBank did not allow further identification of the novel genotypes, the sequences were vastly different from C. andersoni.

The distribution of Cryptosporidium species in surface water, livestock and wildlife suggests that there are interactions between the three. Species usually associated with livestock were found in water (C. parvum and C. andersoni) and in wildlife (C. parvum and C. bovis), and those possibly originating from wildlife were detected in water and livestock (novel genotypes). While oocyst counts are influenced by flow conditions, species are more influenced by the source host. The detection of the UK E7 which is closely related to the fox genotype (W24) in a calf and C. bovis in a fox on the same farm implies that transmission between these hosts may occur with onward spread by wildlife within or between farms.

The MLFT results identified several different subtypes of C. parvum in the catchment. Three different subtypes were detected in the water samples (P5, P7 and P33), of which P5 and P7 were also detected in animals. Additionally, P33 was identified in a diarrhoeic calf sample submitted to the UK CRU in September 2002 from within the study catchment, and P5 was identified in a clinical sample from a human also living in the catchment. Of these, P5 was detected in the water from the farmyard land drain containing very high numbers of oocysts, which is of concern as it is clearly zoonotic and of significance to public health (Hunter et al. 2007).

While the detection of oocysts can provide information regarding the extent of contamination and whether a point or diffuse source is likely, very little can be concluded regarding the source host or significance to public health. Determining the species or genotypes of Cryptosporidium present provides some indication of public health risk and possible sources, which is enhanced by information about the animals in the catchment. Although oocyst viability data would further enhance the risk assessment, there are no standard or approved methods for water samples. This study demonstrates the variety of Cryptosporidium species and genotypes present in the livestock, wildlife and surface waters in a defined whole catchment. By exploring the potential and likely interactions between them, it is possible to identify areas likely to promote the spread of specific Cryptosporidium throughout the watershed and its inhabitants and enable better management procedures to be implemented to reduce the environmental contamination as well as the parasite burden within the animal populations.

Acknowledgements

This work was supported by funding from UKWIR during Phase 1 and Defra (managed by DWI) during Phase 2. We acknowledge the work of The Paragon Veterinary Group, the Cumbrian Farming and Wildlife Action Group, Diane Meadows and the local agricultural college at Newton Rigg, Penrith for helping identify the farms included in the study. For their contribution to this project, we thank David Gomez and Anne Thomas (UK CRU); Andrew Holliman (VLA); Nigel Calvert (CCDC); Jon Greaves and Peter Miles (Environment Agency). This work would not have been possible without the co-operation of the farmers to whom we extend our grateful thanks.

Ancillary