Cryptosporidium genotyping and land use mapping for hazard identification and source tracking in a small mixed rural–urban watershed in Southeastern Brazil

The protozoan parasite Cryptosporidium has a number of features which favour waterborne transmission: (i) it is an ubiquitous zoonotic organism; (ii) its life cycle is completed within an individual host with the transmissive infectious stage (the oocysts) being produced and excreted in large numbers; (iii) the oocysts are capable of prolonged survival in the environment, are able to pass through physical barriers and are resistant to disinfectants commonly used in water treatment processes; and (iv) it has a low infective dose for humans (Smith et al. 2007; Monis et al. 2014; Thompson et al. 2016). Transmission of Cryptosporidium through drinking and recreational water and waterborne outbreaks of cryptosporidiosis, worldwide (Efstratiou et al. 2017) have been well documented. Thus, Cryptosporidium has been taken as a reference pathogen in quantitative microbial risk assessment (QMRA) studies applied to drinking water, in water safety planning and in the development of drinking water regulation (USEPA 2006; WHO 2017).

Infected human and livestock are among the main contributors to environmental contamination with Cryptosporidium oocysts (Hofstra et al. 2013), but this protozoan also infects a wide range of vertebrate hosts, both domestic animals and wildlife, including mammals, birds, reptiles and fish (Smith et al. 2007; Zahedi et al. 2016). Biological aspects such as host species, site of infection and oocyst morphometry have traditionally been used to identify Cryptosporidium. Latterly, with molecular taxonomic methods the number of recognized species has increased substantially, the majority of them described on the basis of genetic distinctness and host occurrence (Thompson et al. 2016). The taxonomy of the genus Cryptosporidium remains controversial but a recent review lists 30 species, with growing evidence of numerous genotypes identified in wildlife and in environmental samples in addition to species that have been recognized from humans and domestic animals (Slapeta 2013; Thompson et al. 2016). Although several species and genotypes have been reported in humans, Cryptosporidium parvum and C. hominis are by far the most common, the former being recognized as zoonotic and the latter as host-specific for humans (Slapeta 2013; Zahedi et al. 2016). C. parvum has a wide range of animal hosts but current evidence indicates that the main reservoir remains livestock (Thompson et al. 2016); it causes neonatal diarrhoea in livestock with consequent economic loss, and is a major contributor to environmental contamination with oocysts (Smith et al. 2007). Cryptosporidium is known as an opportunistic pathogen (Thompson et al. 2016), so human infection with other host-adapted Cryptosporidium species and genotypes is possible but occurs infrequently and depends on the immune status of the host (Slapeta 2013; Thompson et al. 2016).

Cryptosporidium speciation and source tracking are therefore important tools for public health risk assessment and watershed management (Prystajecky et al. 2014; Zahedi et al. 2016). As a matter of fact, Cryptosporidium monitoring in source water is sometimes seen as a necessary hazard identification measure, since the association between Cryptosporidium oocysts, faecal indicator organisms and physical/chemical water quality parameters remains controversial (Lalancette et al. 2014; Prystajecky et al. 2014). However, Cryptosporidium is an elusive organism and current microscope-based laboratory methods for detecting oocysts still suffer from problems of sensitivity and specificity (Thompson et al. 2016); further, microscopy cannot differentiate between oocysts of Cryptosporidium of public health significance and the many other species and genotypes (Smith et al. 2007; Thompson et al. 2016). Thus, combining microscopic and molecular testing, including genomic sequencing, has been shown to be of particular value in Cryptosporidium source tracking and in guiding interventions to reduce faecal loading into water supplies (Prystajecky et al. 2014).

Among the molecular methods based on polymerase chain reaction (PCR), nested-PCR has been suggested as a sensitive method for detecting Cryptosporidium in water samples (Adamska et al. 2015). However, because of the great diversity of Cryptosporidium species that may be found in environmental samples, a proper genotyping tool must use a target gene able to identify all species or genotypes. Among the several target genes that have been used for detecting and identifying Cryptosporidium species, the 18S rRNA gene is one the most frequently used, for its high accuracy on genotyping isolates of different species (Zahedi et al. 2016).

In this paper, we report a study in a small mixed urban–rural watershed in southeast Brazil used as a source of drinking water supply. Cryptosporidium source tracking and related public health risks are discussed on the basis of Cryptosporidium oocyst monitoring in surface water samples complemented by genomic sequencing, faecal indicator organism occurrence, physical/chemical water quality characterization and land use.

The study area was the São Bartolomeu Stream (SB) catchment, which serves as a source of drinking water supply in Viçosa, Minas Gerais State, Brazil (20°45′14″S, 42°52′54″W). This Brazilian town has a subtropical climate (according to the Köppen classification), with average summer maximum and winter minimum temperatures around 30 °C and 10 °C, respectively, and annual average precipitation around 1,200 mm spread over rainy (spring–summer) and dry (autumn–winter) seasons. The SB catchment is a small (approximately 3,000 ha) unprotected watershed, with agricultural activities in small farms and an ongoing urbanization process. Approximately 49% of the watershed is agricultural, with swine and cattle as the major livestock; another 49% of the remaining areas is forested and only 2% is urban. Houses are mostly scattered around the catchment area and use septic systems, but there are a couple of more urbanized villages, one of them serviced by a small wastewater treatment plant. Downstream, within the campus of the University of Viçosa, SB is dammed forming two reservoirs in series, in the second of which there is a water intake for two conventional drinking water treatment plants. SB is a rather small stream with very low flow rates (<100 L s⁻¹ during the dry season), formed by seven other small tributaries which delimit the seven subcatchments studied herein (Figure 1).

Samples were collected approximately on a monthly basis from each subcatchment (SB2–SB8) at sampling points where the respective streams tributaries met SB; samples were also taken from SB itself (SB1), downstream just before the first reservoir, as well as from the water intake point in the second reservoir (Figure 1). Phage and other microbiological and chemical water quality variables were monitored between July 2012 and May 2014; as a whole, approximately 16–19 samples were collected at each sampling point. Cryptosporidium was monitored from February to July 2013 and from February to August 2014, with 13 water samples being collected at each sampling point.

Cryptosporidium enumeration was carried out following a three-step method as described in Cantusio Neto et al. (2010): (i) concentration of 2 L samples by membrane filtration – membranes were placed in Petri dishes, scraped with a soft plastic loop and rinsed with Milli-Q water; this liquid was centrifuged and the resulting pellet was repeatedly washed and centrifuged until a final 10 ml sample was obtained; (ii) immunomagnetic separation of these 10 ml samples using Dynabeads^®-GC Combo; (Dynal Biotec) and storing of the resulting pellet in 2 mL Eppendorf flasks containing 300 μl Milli-Q water; (iii) 100 μl subsamples were then used for oocyst enumeration by immunofluorescence microscopy assay using Merifluor^® C/G kit (Meridian Diagnostics Inc.). Three positive control assays were carried out using 2 L drinking (filtered) water samples spiked with 10⁷ purified Cryptosporidium oocysts (produced in 10 ml flasks by the University of Uberlândia, Brazil). The average recovery rate was 47.8%; however, our results of oocyst counts were not adjusted taking into account this figure. Negative control tests were carried out using Milli-Q water and no contamination was detected.

All immunofluorescence microscopy assay positive samples were further analysed by molecular methods. Two hundred microlitres of the above-described samples (after the immunomagnetic separation step) were used for DNA extraction using a QIAamp DNA Stool micro kit (QIAGEN) according to the manufacturer's instructions with the following modifications in the lysis step: lysis buffer (200 μL, 1:1 v/v) and a metallic bead were added to the samples. This mixture was frozen in liquid nitrogen and placed in Tissuelyser (Invitrogen) for 50 Hz agitation for five minutes; this process was repeated three times, then 20 μL of proteinase K was added to the mixture followed by incubation at 65 °C for 30 min. The other steps were performed according to the manufacturer's instructions. DNA samples were visualized in agarose gel and quantified using Qubit^®. In order to detect Cryptosporidium spp., 18S rRNA gene was amplified by nested PCR as described by Xiao et al. (1999). For the first PCR primers SSU1F (5′- TTC TAG TAA TAC ATG CG-3′) and SSU1R (5′-CCT AAT VVT TCG AAA CAG GA-3′) were used. The PCR reaction mixture contained 10 ng DNA, 1X Buffer IO (Phoneutria), 2.5 μg BSA, 1.5 mM MgCl2; 0.2 μM dNTPs, 0.2 μM each primer, and 2.5 U Taq Polymerase HotStart, all in a final volume of 50 μL. Temperature cycles consisted of 94 °C for 3 min; followed by 35 cycles at 94 °C for 45 sec, 55 °C for 45 sec, and 72 °C for 1 min, and finally 72 °C for 7 min. As a positive control, a sample of C. parvum oocysts was used. For the second PCR, primers SSU2 F (5 V′-GAA GGG TTG TAT TTA TTA GAT AAG-3′) and SSU2 R (5′- AGG AGT AAG GAA CAA CCT CCA-3′) were used. The reactions were performed as described above for the first PCR, replacing extracted DNA for 5 μL of first PCR product. For 18S sequencing, 800 bp bands were excised from gel and purified using GeneJet Gel Extraction Kit (Thermo Scientific), following the manufacturer's protocol. Purified DNA was sequenced in a 3130XL Genetic Analyser (Applied Biosystems) using primers SSU2F and SSU2R. Sequences were edited for low quality base trimming in BioEdit, which was also used to obtain contigs using the Contig Assembly Program tool. Full 18S rRNA sequences were then submitted to GenBank for classification using the BLAST program (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The closest sequences were downloaded to obtain the phylogenetic tree using the neighbour-joining method with bootstrapping simulation (1,000 repetitions); the software MEGA 6 was used for the statistical analyses (Tamura et al. 2013).

Coliphage enumeration was carried out using the standardized method of simple agar plaque counting and the following bacteria host strains (obtained from CETESB – the São Paulo State Environmental Company): Escherichia coli CN-13 (ATCC 700609) for somatic coliphages, and E. coli FAmp (ATCC 700891) for RNA F-specific coliphages. For Bacteroides fragilis phage detection the double agar plaque method (ISO 2001) and the following bacteria host strains were used (obtained from the University of Brighton, UK): GB-124, HB-73 and S-10, the first two being, supposedly, human-specific and the latter animal-specific (Vijayavel et al. 2010; Ebdon et al. 2012). Water samples were also tested for total coliforms, E. coli and enterococci, using standardized chromogenic techniques and the Colilert^® (Idexx Laboratories Inc.) and Chromocult^® Enterococci Broth (Merck©) media for, respectively, coliforms and enterococci.

In addition, water samples were tested for the following variables according to the recommendations of Standard Methods for the Examination of Water and Wastewater: chemical oxygen demand (COD), electrical conductivity (EC); dissolved oxygen (DO), pH and turbidity.

Data on Cryptosporidium occurrence as well as those of the other microbiological and physical/chemical water quality variables were initially characterized using descriptive statistics. Since the dataset as a whole neither followed normal distribution nor presented variance homogeneity, the non-parametric Spearman rank correlation test was used for testing the association between the occurrence of Cryptosporidium and the other water quality variables (faecal indicator organisms and physical/chemical parameters) (p < 0.05), as well as between Cryptosporidium and animal farming variables (number of farms, number of animals, and density of animals) (p < 0.1). Similarly, significant differences among microbial counts in subcatchment samples were compared using the non-parametric Kruskal–Wallis test (p < 0.05). In addition, the results of Cryptosporidium occurrence are discussed based on land use in the watershed. Statistical analyses were performed using Statistical Package for the Social Sciences 17.0 (SPSS Inc.) and Minitab 17 (Minitab, Inc.) software.

Cryptosporidium was present in 43% of the 117 water samples analysed by immunofluorescence microscopy, being detected at every sampling point, i.e. in all subcatchments. Concentrations varied from non-detects to 14 oocysts L⁻¹. The highest average counts were found in subcatchments SB7 and SB8 (4.69 and 4.87 oocysts L⁻¹, respectively) whereas SB4 was the subcatchment least contaminated with oocysts (Table 1). However, there were no significant differences among oocyst counts in the various subcatchments.

Cryptosporidium spp. was detected by nested-PCR in 24% (28) of all samples originally analysed by immunofluorescence microscopy, i.e. in 56% of microscopy-positive samples. The amount of extracted DNA was correlated with oocyst counts by microscopy (Spearman correlation coefficient: rs = 0.6807; p = 0.022), and in general DNA was not obtained in samples containing less than 1 oocyst L⁻¹. The 28 samples from which 18S rRNA gene amplification was successful were sequenced but only eight of them gave high quality matching C. parvum GenBank sequences with 100% homology (V16, V22, V15, V13, V8, V6, V5 and V32 in Figure 2).

Table 2 summarizes the results of faecal indicator bacteria, coliphages and physical/chemical water quality variables. Bacteroides fragilis phages were only erratically and rarely detected, suggesting geographic limitations in the use of the strains used here.

Subcatchments SB8 and SB7 showed both the highest average counts and results variability of somatic coliphages: 55.5 ± 88.8 and 18.3 ± 48.3 plaque-forming units (PFU)/100 mL, respectively. Conversely, SB4 showed the lowest somatic coliphage average counts as well as the lowest results variability (3.5 ± 14.0 PFU/100 mL). For all sampling points, F-specific coliphage counts were lower than those of somatic coliphages. The highest average counts of F-specific coliphages were found in subcatchments SB6 (4.2 PFU/100 mL) and SB7 (3.6 PFU/100 mL); these phage counts were, however, similar (same order) to those of most of the other subcatchments, except for SB4 (0.1 PFU/100 mL) and SB1 (10.3 PFU/100 mL), whose sampling points (downstream in the catchment) receives the whole contribution of the other water courses. There were no statistically significant differences of coliphages (both somatic and F-specific coliphages) concentration among subcatchments.

Overall, the average concentrations of total coliforms, E. coli and enterococci were much higher than those of coliphages. In general, E. coli counts ranged from 10² to 10³E. coli/100 mL, and subcatchments SB7 (7.7 × 10³E. coli/100 mL) and SB8 (4.1 × 10³E. coli/100 mL) showed the highest concentrations of these faecal indicator bacteria. At SB1 sampling point, E. coli numbers were even higher (9.6 × 10³E. coli/100 mL), but as previously mentioned this sampling point receives the contributions of the whole catchment. On the other hand, the lowest E. coli counts were recorded in subcatchments SB6 (1.2 × 10²E. coli/100 mL) and SB5 (3.7 × 10²E. coli/100 mL). In general, average concentrations of enterococci (around 10² org./100 mL) were lower than those of E. coli, except for subcatchment SB8 (6 × 10³ enterococci/100 mL) and SB1 (1.2 × 10³ enterococci/100 mL); the lowest average figures were recorded in SB2 (1.6 × 10² enterococci/100 mL) and SB6 (6 × 10² enterococci/100 mL).

Correlation tests did not indicate strong or statistically significant relationships between the occurrence of coliphages and faecal indicator bacteria (coliforms, E. coli and enterococci), not even among the faecal indicator bacteria themselves (the only exception was the correlation found between total coliforms and E. coli, which is rather obvious). In addition, there were no correlations between coliphage counts and concentrations of physical and chemical water quality variables (COD, EC, DO, pH and turbidity). Also, Cryptosporidium concentrations were not correlated with faecal indicator organisms (total coliforms, E. coli, Enterococcus and coliphages), nor with physical and chemical water quality parameters.

Cryptosporidium was detected in this study frequently (43% of 117 samples) and concentrations (ranging from non-detects to 14 oocysts L⁻¹) were similar to or even higher than previous findings in the same watershed as well as in others in Brazil, and elsewhere (Cantusio Neto et al. 2010; Razzolini et al. 2010; Araújo et al. 2011; Van Dyke et al. 2012; Bastos et al. 2013; Sato et al. 2013; Prystajecky et al. 2014; Franco et al. 2016). The percentage of nested-PCR-positive results recorded here (24% of all samples originally analysed by microscopy, and 56% of the microscopy-positive samples) is in accordance with or even better than those found in other studies in surface waters with low numbers of oocysts (Araújo et al. 2011; Prystajecky et al. 2014; Adamska et al. 2015), although other authors did find higher figures (Van Dyke et al. 2012). Notwithstanding nested-PCR being a method with high sensitivity, it is well known that DNA amplification can be impaired by factors such as: (i) the low numbers of oocysts usually present in surface waters; (ii) presence of inhibitors which interfere with the DNA–DNA polymerase interaction, e.g. humic and fulvic acids, frequently found in environmental samples (Jiang et al. 2005); and (iii) difficulties in oocyst wall rupture, thus in the lysis, and DNA extraction and purification (Adamska et al. 2015). As mentioned above, in this study the amount of extracted DNA was found to be related to oocysts concentration and, in general, DNA was not obtained in samples containing less than 1 oocyst L⁻¹. Prystajecky et al. (2014) found that 60% of samples positive by microscopy were also positive by at least one nested PCR (PCR analyses were conducted in triplicate). The authors argued that this positivity rate likely reflected the concentrations of oocysts on the microscopy slides (48% of the slides contained only a single oocyst); therefore, any losses during sample processing, along with nonhomogeneous distributions of nucleic acids in extracts, may have impacted PCR results. Franco et al. (2016), in a study in a highly industrialized region of Brazil, also found that microscopy-positive results could not be amplified by nested PCR and attributed that to the effects of inhibitors such as silt, organic material, humic acids and metals.

The lack of correlation between the occurrence of Cryptosporidium and of total coliforms, E. coli, enterococci and coliphages again raises questions about whether traditional indicators of water microbial quality are valid indicators of Cryptosporidium in surface water (Nieminski et al. 2010). This, together with the lack of correlation between Cryptosporidium and physical and chemical water quality variables, as well as between coliphages and chemical and other microbial water quality parameters suggest that: (i) faecal indicator bacteria, coliphages and Cryptosporidium presented differences in origin, transport and survival patterns in the various subcatchments; (ii) there was no clear relationship between the levels of water pollution (measured by COD, EC DO, pH and turbidity) and water contamination (measured by the occurrence of Cryptosporidium and faecal indicator organisms, i.e. total coliforms, E. coli, enterococci and coliphages). Similarly, Prystajecky et al. (2014) did not find correlation between total coliforms, E. coli, Cryptosporidium and Giardia concentrations in a mixed rural–urban watershed in Canada, and explained that finding on the basis that the structural integrity of the protozoan (oo)cyst wall and low metabolic state allow their prolonged survival in water. In the study by Prystajecky et al. (2014), there was also no correlation between chemical water quality parameters, Cryptosporidium and Giardia concentrations; however, a statistically significant relationship was found between parasite concentration and turbidity (Pearson correlation coefficient r = 0.76 at one study site and r = 0.56 in another; p < 0.05). Van Dyke et al. (2012), using the Spearman rank order method, found that Cryptosporidium concentration was significantly correlated (p < 0.01) with E. coli (rs = 0.46), turbidity (rs = 0.37) and river flow (rs = 0.34) in a large watershed influenced by urban and rural activities, also in Canada. However, it is worth noticing that the correlations were not strong and there was no relationship between Giardia and these water quality parameters. Lalancette et al. (2014) quoted several studies conducted on surface waters in which the concentrations of Cryptosporidium oocysts were poorly or uncorrelated with traditional indicators such as E. coli, total and faecal coliforms and enterococci, or with turbidity or rainfall. In their own study, these authors suggested that E. coli or faecal coliforms are potentially good indicators of Cryptosporidium concentrations when source waters are impacted by recent and nearby municipal sewage, but not for sources dominated by agricultural or rural faecal pollution sources or more distant wastewater sources.

However, in spite of the fact that there were no statistically significant differences among oocyst concentrations in the various subcatchments, nor among counts of phages and faecal indicator bacteria, there are suggestions that the occurrence of these organisms were somehow related to land use. Table 3 shows the distribution of four land use categories, and Table 4 the distribution of animal farming variables, among the SB subcatchments.

Subcatchments SB7 and SB8 presented the highest numbers of Cryptosporidium microscopy-positive samples, as well as the highest animal populations: 262 and 408 animals as a whole, respectively (Table 4). In SB7 there are both a bovine (82 animals) and a swine (180 animals) farm whose untreated wastewater/manure is either discharged into streams or used on land as fertilizer. In SB8 there are two swine farms (280 animals in total), whose wastewater is treated in rough stabilization pond systems, the effluents of which are used for pasture irrigation. Also, it is worth noting that both SB7 and SB8 presented considerable amounts of exposed soil (≈45%) (Table 3), favouring therefore overland flow of wastes eventually disposed onto soil. On the other hand, the lowest counts of Cryptosporidium oocysts, and somatic and F-specific coliphages were recorded in SB4, where there is very little agriculture and quite a large forested area (≈77% of the subcatchment's area) (Figure 1, Tables 3 and 4). Further, the Spearman rank order test showed that Cryptosporidium oocysts occurrence in water samples was statistically correlated with the variable ‘density of animals’ (number of animals per hectare) (rs = 0.67; p = 0.08).

Hence, animal farming seems to be an important source of environmental contamination with Cryptosporidium spp. oocysts within the SB watershed. However, the results do not necessarily mean that farm animals are the sole, nor even the most important, Cryptosporidium source in the SB watershed. The more urbanized areas in this watershed are located downstream in SB1 (where there is a small village serviced by a small wastewater treatment plant) and in SB7, but in SB8 there is also an incipient but ongoing urbanization process (Figure 1 and Table 3). It is noticeable that SB1 and, mainly, SB7 and SB8 were at the same time the subcatchments where some of the highest counts of not only Cryptosporidium oocysts but also of coliphages, E.coli and enterococci were recorded.

Detection of C. parvum in the present study is therefore in agreement with the mixed rural–urban characteristics of the SB watershed with potential pathogen sources most probably including both animals (livestock) and humans. However, the results themselves (detection of only C. parvum, no detection of Bacteroides fragilis phages) as well as the very scope of this study (not including the distinction among F-RNA phage serogroups (Vergara et al. 2015)), did not allow a more conclusive discussion on distinguishing between faecal contamination of human and animal origin.

Notwithstanding data on genotypic identification of Cryptosporidium from environmental samples still being relatively scarce in Brazil, a few studies have reported findings which, in a way, corroborate ours. Almeida et al. (2015) explained the occurrence of C. parvum in a water source in Paraná State, southern Brazil, by the existence of several dairy farms within the watershed. Conversely, Araújo et al. (2011) found C. hominis in recreational water sites in São Paulo State known to be impacted by human activities. Franco et al. (2016) isolated C. hominis and C. parvum in samples at the intake of a drinking water treatment plant in a highly urbanized and industrialized region also in São Paulo State. Molecular methods and land use have also been the basis for explaining the occurrence of Cryptosporidium elsewhere. Keeley & Faulkner (2008), using the now outdated restriction fragment length polymorphism analysis, suggested that the C. parvum detected in surface waters came from cattle living in the watershed of Lake Texoma, on the border of Texas and Oklahoma, which serves rural agricultural communities active in cattle ranching. Using nested PCR and DNA sequencing in a study in a mixed rural–urban watershed in southwestern British Columbia, Canada, Prystajecky et al. (2014) expectedly (due to the range of land use in the area) found several Cryptosporidium species. C. andersoni was the most commonly detected species (23% of the isolates), most likely to originate from cattle, followed by the most widely recognized zoonotic species C. parvum (12%) and C. baileyi (14%), supposedly from avian sources; C. hominis, infectious only to humans, although representing only 6% of the isolates, indicated that contamination of surface water by human waste occurred too. Similar findings were reported by Van Dyke et al. (2012), also using nested PCR and DNA sequencing in a study in a mixed rural and urban, but larger, watershed in southern Ontario, Canada. Cryptosporidium genotyping of water samples reflected potential sources of protozoa in the region, including farm animals, wildlife and humans. Again, the most frequently detected species was C. andersoni (53% of all samples), which was said to be expected since cattle are a common livestock animal on farms in the region. C. hominis and C. parvum, the most common cause of human cryptosporidiosis, were detected in approximately 11% of all samples but at higher proportions at locations directly influenced by human wastewater. Other Cryptosporidium species detected in the watershed were reported to be primarily associated with farm livestock and wildlife, including C. baileyi (poultry), C. bovis (cattle), C. muris (rodents) and C. ubiqutum, which has a broad host range that includes domestic and wild animals, but may infect humans too (Van Dyke et al. 2012). Finally, Phillip et al. (2008) did not use genotyping for Cryptosporidium source tracking, but instead they tried to correlate the presence of Cryptosporidium spp. oocysts in surface waters with land use characteristics in three different watersheds located in the Northern Range Mountains of the larger island, Trinidad, in the Republic of Trinidad & Tobago. It was concluded that urban (by leakage of inadequately treated sewage effluent into surface waters) and wildlife (from forested lands) were the main types of sources of Cryptosporidium contamination of surface water, whereas the contribution of agriculture, especially the poultry and livestock industries, was minor.

This report highlights the usefulness of molecular testing and land use mapping for microbial hazard identification and source tracking. The results confirm that traditional surrogates of microbial water quality (coliphages, coliform bacteria and enterococci) are poor indicators of Cryptosporidium occurrence in surface water. Nevertheless, the results of Cryptosporidium occurrence and, in a way, of these indicators too, seemed to reflect impacts of the discharge/surface runoff of both human and farm animal wastewaters/manure from rural and/or more urbanized areas within the watershed. Detection of C. parvum is relevant information in as much as this species is recognized as zoonotic and São Bartolomeu Stream is the source water for two drinking water treatment plants.

The authors would like to acknowledge the Brazilian agencies CAPES, CNPq and FAPEMIG for funding this work and/or providing students scholarships. We would also like to thank colleagues from the University of Brighton (UK) and CETESB (São Paulo State Environmental Company, Brazil) for providing phage and bacteria host strains.