Study area and design
This study was done from February to May 2016 in southern Ethiopia that consists of 13 administrative zones, covering about 10% of the landmass of Ethiopia. It is located in 4°43′–8°58’ N latitude and 34°88′–39°14′ E longitude with an altitude range of 376–4207 m above sea level. The annual rainfall and temperature range of 500–2200 mm and 15–30°C, respectively. Drinking water sources in the region range from treated to untreated sources of municipal/tap, unprotected spring, protected spring and wells, and river/stream water. Chlorination was the only techniques used to treat municipal/tap water source, whereas the other sources were untreated. Spring water is a natural discharge point of subterranean water at the surface of the ground. Unprotected spring is often unprotected or poorly protected that is open to the atmosphere, whereas protected spring is protected by providing a concrete headwall or spring box around the eye of the spring (where water emerges) that prevents direct contamination of the water. The protective cover usually overlies the excavated area and the area is fenced for some distance to prevent direct access by humans and animals. River is a natural stream of water of considerable volume, larger than a brook or creek, and flows downwards until it reaches its end, crossing land, hills and plains in its journey. Stream is a natural body of running water flowing in a natural channel as distinct from a canal; contains water at least part of the year.
The study area was selected using multi-stage cluster sampling technique; seven zones from the region and three districts from each zone were selected. Health risks were characterized using 21 different water sources that were selected from each district and 576 individuals that were using the water sources. Participants were selected based on a single proportion formula, considering design effect 1.5. According to the sample sizes allocated to each site, households were selected using systematic sampling technique, with an individual per household randomly selected. Children < 2 years and individuals who did not often use the water sources for drinking purpose were excluded, as the present study aimed at characterizing risks of G. duodenalis and Cryptosporidium spp. infection associated to drinking water sources. Participants’ information was gathered using a semi-structured questionnaire and water samples were analyzed quantitatively using immunofluorescence assay (IFA) and qualitatively (nested-PCR), and risks were characterized using quantitative microbial risk assessment (QMRA).
Stock suspension preparation and enumeration
Human stool samples from patients infected with G. duodenalis, Cryptosporidium spp. were obtained from Arba Minch hospital, southern Ethiopia. The samples were suspended with distilled water and filtered through a 0.5-mm sieve and concentrated by centrifugation at 1050 g for 10 min and the supernatant was decanted. Cysts and oocysts were isolated using percoll-sucrose gradient (specific gravity: 1.09–1.10) and centrifugation at 1050 g for10 min. Stock suspensions having concentrations of 103 to 104 cysts or oocysts per liter were prepared using reagent water (MilliQ, Millipore) with 0.01% Tween 20. Enumeration was performed using haemocytometer chamber count (Bright-Line, Reichert, Buffalo, NY), to achieve the optimal counting accuracy, 10 different chambers were counted for each cyst/oocyst suspension. Spiking suspensions having 102 to 103 concentrations of each cyst and oocyst per liter were prepared by dilution using a drop count procedure. Droplets (10-20 µl) from the stock suspensions were pipetted onto the edge of a microscope slide and counted using bright light microscopy (bright field illumination, Nikon, Japan) at 250X and 400X magnifications for cysts and oocysts, respectively. The counts were replicated for three times and the concentrations of the organisms were established. In addition, the same volumes of the suspensions were directly transferred into well slides, stained with FITC-MAb (Waterborne Inc., New Orleans) and enumerated using IFA.
Recovery efficiency and detection limit
Initial recovery efficiency of the method was achieved by spiking known number of cysts/oocysts into a 10 L reagent water and counting the recovered parasites in the spiked sample, in triplicate. Concentration of cysts/oocysts in each sample, the recovery efficiency and detection limit of the method were calculated using equations 1, 2 and 3, respectively.
The inter-assay recovery efficiency of the method was also assessed to reveal the variation in the different water samples. The effect of matrix on recovery efficiency of the method was determined using spiked environmental water samples that were collected from the same sources. To assure the absence of contamination throughout the analytical procedure, unspiked reagent water sample was analyzed accordingly. Two water samples with the same volume of water were collected from the same source; One sample for pathogen analysis and the other for recovery efficiency analysis. Spiking experiment per sample was carried out in triplicate and percent recovery (REC) was calculated according to [54], (equation 4). The number of cysts/oocysts in the water samples per 10 L were calculated, (equation 5).
Where N is the number of cysts/oocysts counted in the spiked water sample, NSAMPLE is number of cysts/oocysts counted in non-spiked environmental sample, CC is the number of cysts/oocysts spiked and PV is the analyzed pellet volume per sample.
Where VE is volume examined, RECavg is average recovery of cysts/oocysts, and NSAMPLE is number of cysts/oocysts found in the unseeded water. When no cysts/oocysts were found, a value (detection limit) of each water type was used was used for NSAMPLE.
Water sample collection
Ten-liter water samples were collected from each source from points representing the nature of water sources. Samples were collected using a clean and sterilized plastic bottle that were washed three times with hot and distilled water. Prior to sample collection, the sample bottles were partially filled, rinsed three times with the same water and drained from the sampler. Sand and other debris were avoided by carefully collecting water rinsing into the water in low- turbidity and the bottles were filled to collect ten-liter water sample. Piped water samples were collected from reservoirs while stream/river water samples were collected at the edges that had low-flow turbidity and debris. The samples were immediately placed in a lightproof insulated box containing ice-packs with water to ensure rapid cooling and preserve the state of the water and shipped to the laboratory, Arba Minch University and overnight stored in a refrigerator at 4ºC until processed. Water turbidity and free chlorine were measured immediately on the time of sample arrival, within 24 h of sample collection. Water turbidity was measured using a Hach 2100N turbidimeter (Hach, Loveland, CO, USA) and expressed in nephelometric turbidity units (NTU). Free chlorine in municipal water was measured using DPD chlorine test kit (La- Motte, Chestertown, MD, USA) and expressed in milligram per liter.
Water sample processing
Sample filtration, elution, concentration, application on slide and drying were carried out within 24 hours of collection. The samples were filtered through 142-mm-diameter with 3.0µm-pore-size cellulose nitrate (Sartorius) membrane. The water sample was drawn through the membrane under negative pressure and the membrane was scraped with a smooth-edged plasticine molder and rinsed with 0.1% Tween 80 (T80). The membrane was rinsed repeatedly by turning it for three times until it became clean and the eluate was collected in a clean and sterilized plastic dish and transferred to a sterile 50 ml centrifuge tube. The dish was rinsed with 0.1% T80, pooled with the resuspended pellet, centrifuged at 5000 g for 10 min. All supernatant of a sample was aspirated to leave the pellet. The pellet was resuspended with 0.1% T80 and transferred into a fresh sterile centrifuge tube. The emptied tube was rinsed with 0.1% T80, the washes were pooled with the resuspended pellet, centrifuged at 5000 g for 10 min and the supernatant was decanted. Pellets of the same sample were collected together and 0.5 ml of each aliquot was used for detection and enumeration. The remaining portions were stored at –20ºC for genotypic analysis.
Pathogen detection and enumeration
Cysts and oocysts were enumerated using IFA and haemocytometer chamber. The IFA was done by transferring 10-µl aliquot of each sample, in triplicate. About 10 µl aliquot of each sample was directly transferred into separate slides, air-dried at 55ºC and placed in a dark humid chamber, and incubated at room temperature for 30 min with combined fluorescein isothiocyanate–monoclonal antibodies (FITC-Mab) (Aqua-Glo G/C kit; Waterborne, Inc., New Orleans, LA). The slides were removed from the chamber, the condensation was let evaporated, and one drop of wash buffer was applied to each well according to the manufacturer’s instructions. The slides were washed in PBS containing T80 (0.01% [vol/vol]; PBST), air-dried at 55ºC and tilted on a clean paper towel, and the excess detection reagent was aspirated using a clean Pasteur pipette. The haemocytometer count was also performed by applying 1 μl of the final pellet; five counts were performed per test.
The wells were counterstained each with 50 μl of 4’, 6-diamidino-2-phenylindole (DAPI) staining solution (Sigma, St. Louis, MO) at 0.4 µg/ml and stood at room temperature for 2 min. A drop of the wash buffer was re-applied to each well, and the slides were tilted and processed as mentioned above. Mounting medium was added to each well and covered using cover slip. Excess mounting fluid was removed from edges of the coverslip using soft tissue and the edges were sealed onto the slides using clear nail polish and examined using epifluorescence. The FITC-Mab staining, DAPI and Differential interference contrast (DIC) examinations were conducted according to the manufacturer instruction. Shape, size and internal morphological features for each apple-green fluorescing object that met the size and shape characteristics of cysts and oocysts per pellet were enumerated. Depending on the recovery rate and detection limit of the method (see ‘Recovery efficiency and detection limit’), the total estimated numbers were calculated based on the analyzed fraction of pellet and total volume of filtered water. The percentage of samples positive in DAPI fluorescence examination and Giardia spp. cysts and Cryptosporidium spp. oocysts viable in DIC examination was analyzed. Cysts and oocysts that exhibited typical FA fluorescence, typical size and shape, nothing atypical on DAPI fluorescence and nothing atypical on DIC microscopy were considered as positive results.
Giardia duodenalis and Cryptosporidium spp. genotyping
DNA extraction
Genomic DNA extraction was performed for all 21 drinking water samples. Portions of each pellet obtained from the respective samples (see ‘Water sample processing’ section) were suspended by vortexing, and 300 µl of the sample was transferred directly into a 1.5 ml centrifuge tube and centrifuged for 2 minutes at 13,000 rpm and the supernatant was discarded. The sediment was re-suspended in 300 µl sterile distilled water and centrifuged at the same conditions three times repeatedly. The extraction was carried out using five freeze-thaw (−70˚C) cycles followed by QIAamp DNA Mini isolate kit (Qiagen, Germany) in accordance with the original protocol. Then, the samples (200 µL each) were washed seven times with 1 mL of PBS (pH= 7.3) and centrifuged at 14000 g for 1 min. The pellets collected in the preceding steps were re-suspended in 1400 μl of buffer ASL, heated to 95°C for 10 min, then vortexed, and centrifuged at 15000 g for 1 min. The digestion process with proteinase K was elongated to 30 min at 70°C. DNA yields of each sample were determined from the concentration of DNA in the eluate, measured by absorbance at 260 nm. DNA purity was determined by calculating the ratio of absorbance at 260 nm to absorbance at 280 nm. DNA length was also determined by pulsed-field gel electrophoresis through an agarose gel, and the genotyping process was performed accordingly.
Giardia duodenalis genotyping
The genotypes were determined using the tpi gene based nested PCR, as it is a good phylogenetic marker for molecular analysis [55]. According to a previous work [56], primary and secondary primers that amplify 618 and 557 bp of the gene were used, respectively. In the primary PCR amplification, external forward primer – TPI-FW1 (5’–CAGAAAATAAATIATGCCTGCTC–3’) and external reverse primer TPI-RV1 (5’–CAAACCTTITCCGCAAACC–3’) were used. The primary PCR amplification was carried out in 50 µl reaction mixture containing 5 µl of DreamTaq™ Buffer (10x), 1 µl of dNTPs mix (10 µM each), 1 µl of GI-TPI-FW1 and 1 µl of GI-TPI-RV1 (10 µM each), 0.25 µl of (5 Unit) Dream Taq DNA polymerase, 0.25 µg/µl of genomic DNA and nuclease free water. The reaction was carried out in 30 cycles, each consisting of 95ºC for 45 seconds, 55 ºC for 45 seconds and 72 ºC for 1minute with an initial hot start at 95 ºC for 5 minutes and a final extension for 7 minutes. In the secondary PCR amplification, the internal forward primer – TPI-FW2 (5’–CCCTTCATCGGIGGTAACTTCAA–3’) and internal reverse primer TPI-RV2 (5’–ACATGGACITCCTCTGCCTGCTC–3’) were used. The PCR reaction was carried out similar to the primary PCR reaction, except using 4 µl of the primary PCR product and prolonging the final extension to 10 minutes. The genotype of G. duodenalis was detected and genotyped using 1.5% gel electrophoresis analysis of the final PCR amplicons. In every activity, both positive control (G. duodenalis DNA) and negative control (nuclease free water) were used.
The final positive PCR products were sequenced using the corresponding primers used for PCR amplification. Sequencing was carried out using a capillary sequencer, 3730xl DNA Analyzer (Applied Biosystems, CA, USA), in combination with ABI PRISM® BigDye™ Terminator Cycle Sequencing Kits (Applied Biosystems, CA, USA). The pre- and post-sequencing processes were automated on a robotic platform consisting of Biomek® FX and NX instruments (Beckman Coulter, Fullerton, CA, USA). The chromatograms and sequences generated from this study were viewed and assembled using the BioEdit Sequence Alignment Editor Program version 7.2.5 (www.mbio.ncsu.edu/bioedit/bioedit.html). The consensus sequences were compared with sequences registered in GenBank, using the basic local alignment search tool (BLAST) (www.ncbi.nlm.nih.gov/blast). Genotyping of sub-assemblages was determined based on sequence homology (100% identity) of the isolates with sequences in GenBank. Representative sequences of assemblages A (KT728546.1), B (EU781015.1), C (AY228641.1), D (DQ246216.1), E (EU272157.1), F (AF069558.1) and G (AY228640.1) were included for comparison analysis. In addition, sub-assemblage A reference sequences - 07JTPI (KT728546.1), AIIGRW (KF963577.1), BSWAII (KF963567.1) and BRW AII (KF963573.1) - and sub-assemblage B reference sequences - BI (HQ397719.2), BIV (LO2116.1), BV (HQ666895.1), BVII (HQ666897.1), Hole H13 (KT948108.1), Swemon 200 (EU781015.1) and HS98 (KC632554.1) - were included. For further genetic variation analysis, G. muris (AF069565.1), G. ardeae (AF069564.1) and G. microti (AY228649.1) were also included.
Cryptosporidium spp. genotyping
The SSU-rRNA and gp60 genes of Cryptosporidium spp. were amplified using nested PCR protocol. In the primary PCR of SSU-rRNA gene, external primers – CR-SSU-FW1 (5’–TTCTAGAGCTAATACATGCG–3’) and CR-SSU-RV1 (5’–CCCATTTCCTTCGAAACAGGA–3’) that amplify 1,323 bp fragment of the gene were used. In the secondary PCR, internal primers –CR-SSU-FW2 (5’–GGAAGGGTTGTATTTATTAGATAAAG–3’) and CR-SSU-RV2 (5’–CTCATAAGGTGCTGAAGGAGTA–3’) that amplify 852 bp fragment of the gene were used. The primary PCR amplification of SSU-rRNA gene was carried out in 50 µl reaction mixture containing 5 µl 10X Dream Taq buffer, 1 µl dNTPs mix (10 µM each), 1 µl CR-SSU- FW1 and 1 µl CR-SSU- RV1 (10 µM each), 0.25 µl of (5 Unit) Dream Taq DNA polymerase, 0.25 µg/µl of genomic DNA and nuclease free water. In the primary PCR reaction, 30 cycles were carried out, each consisting of 94 ºC for 45s, 55 ºC for 45s and 72 ºC for 1 min., with an initial hot start at 94 ºC for 3 min and a final extension for 7 min. In the secondary PCR reaction, 4 µl of the primary PCR product will be used and the cycles will be increased to 35; whereas, the other PCR conditions were similar to the primary reaction.
In the primary PCR amplification of gp60 gene, external primers – CR-GP60-FW1 (5’–TTACTCTCCGTTATAGTCTCC–3’) and CR-GP60-RV1 (5’–GGAAGGAACGATGTATCTGA–3’) that amplify 915 bp fragment of the gene were used. In the secondary PCR, internal primers –CR-GP60-FW2 (5’–TCCGCTGTATTCTCAGCC–3’) and CR-GP60-RV2 (5’–GCAGAGGAACCAGCATC–3’) that amplify 869 bp of the gene fragment were used. Except a different annealing temperature (54ºC), the PCR amplifications of gp60 gene was analogous to SSU-rRNA gene amplification. The genotypes were determined using 1.5% gel electrophoresis: 5 µl of the final PCR product with 2 µl 6X loading dye at 110 V for 55 min were used. During each PCR reaction, known DNA and nuclease free water were used as positive and negative controls. In addition, the primers were checked based on gene sequences available from GenBank, the primers were also pretested and the PCR conditions were optimized. Cryptosporidium spp. isolates were only typed at the genus level; genotypic analysis of C. hominis and C. parvum isolates at sub-type level was not done.
Quantitative microbial risk assessment
Risks of the pathogens from drinking water sources were characterized using QMRA. Exposure of individuals to G. duodenalis, C. parvum and C. hominis infections was determined based on concentration of cysts/oocysts in the water samples, as previously described [57,58]. Individuals' exposure to each pathogen and the volume of water consumption through drinking exposure pathway per given amount of time, and water consumption behaviour of peoples (treatment practice before drinking) were gathered through a semi-structured questionnaire. Pathogen dose that are possibly ingested by an individual was computed as a product of volume of water consumed un-boiled or untreated in a given period and pathogen concentration in the water (equation 6). The daily dose of pathogen for treated water was computed according to [59], (equation 7). The probability of infection to each pathogen and its progress to illness were computed using an exponential dose-response model with parameters of an infection endpoint [60], (equation 8). Only cysts and oocysts exhibiting and meeting the criteria based on DIC examination were considered viable and used for analysis. As the cysts and oocysts detected in water samples may correspond to other species that are not infective to human, genotypic analysis was done to insure the presence of the respective species thar are infective to humans.
Where C is concentration of pathogens in the water or partially processed water, R is recovery of the detection method, I is fraction of the detected pathogens that is capable of infection (viability), DR is decimal reduction factor (DR = 0 when concentration in the finished water is available), and V is daily individual consumption of the considered water.
Where D is pathogen dose, and r is fraction of pathogens that survives to produce an infection
An exponential dose-response model for representative species of G. duodenalis, C. parvum and C. hominis was considered to avoid response variations for exposure to a variety of species within a genus. Dose-response model parameter, r = 0.0199 for G. duodenalis [61] and r = 0.09 for Cryptosporidium spp. [62,63] were employed. Two parameters: r = 5.26×10-3 for immunocompetent and r = 0.354 for immunocompromised sub-populations were also accounted for C. parvum [64]. The probability of daily infection was extrapolated to yearly risk; consecutive exposures were assumed to be independent, then the probability of one or more infections for n exposures per year was supposed to be the corollary of 'n. For daily infections with values of << 1, infections per year were computed by multiplying the probability of daily infection and the number of days [60].
The probability of illness per given infection was computed using a morbidity fraction of each pathogen. Progression from infection to illness for G. duodenalis, C. parvum and C. hominis is ranged from 0.2 to 0.7 [60,65,66,67]. A morbidity fraction of one was used to compute the probability of illness to Cryptosporidium spp. infection for immunocompromised sub-populations [64]. The probability of morbidity was calculated according to [57,68], (equations 9). All the health outcomes of illness were considered as an endpoint and individuals were assumed to have at least one disease outcome.
Where, PInfection is the probability of infection with a specific pathogen, Pill/inf is the probability of illness due to infection.
Questionnaire
The study participants’ information was collected using semi-structured questionnaire, which was pretested on a small number of respondents that were chosen randomly in the population being studied. The respondents were interviewed and their responses were documented and the questions were rephrased and interviews were repeated, then each response was investigated and comparable responses were maintained. Once the basis and procedures of the study were understood and agreed upon, the participants (parents of children) signed the informed consent forms and participated in the study. Then, the study subjects age, sex, type(s) of drinking water source(s), amount or volume water used for consumption per day, water consumption days per month and types of water treatment techniques or mechanisms used by individuals were collected (see Annex VA and VB).
Statistical analysis
Nonparametric Spearman’s rho correlation coefficient was computed to assess how the occurrence rate of G. duodenalis cysts and Cryptosporidium spp. oocysts associated with the types of water sources and physicochemical parameters of the sources. Spearman's rho was preferred as it is suitable for data that that are not distributed normally, moreover outliers have less of an effect on this statistical method. Correlation coefficients range in value from –1 (a perfect negative relationship), +1 (a perfect positive relationship) and a value of 0 (no linear relationship) were used to indicate the relationships of variables. In addition, independent-samples (Kruskal-Wallis) nonparametric test was performed to compare the distribution of cysts and oocysts across the types of water sources. In the analysis, P<0.05 were considered statistically significant. The analysis were performed with the IBM* SPSS* Statistics for Windows, version 23 software (SPSS Inc., Chicago, IL, USA).