Cryptosporidium oocysts

Montecino-Latorre, Diego, Li, Xunde, Xiao, Chengling & Atwill, Edward R., 2015, Elevation and vegetation determine Cryptosporidium oocyst shedding by yellow-bellied marmots (Marmota flaviventris) in the Sierra Nevada Mountains, International Journal for Parasitology: Parasites and Wildlife 4 (2), pp. 171-177 : 173-174

2.3. Detecting Cryptosporidium oocysts

The detection of Cryptosporidium oocysts from fecal samples was based on published methods ( Kuczynska and Shelton, 1999; Pereira et al., 1999) with modifications. Briefly, feces were weighed, resuspended in PBS, and strained into 50 ml centrifuge tubes through 4 layers of cotton gauze. Tubes were centrifuged at 1500 g for 15 min, supernatant was discarded, and pellet was resuspended in 5 ml of deionized water. The fecal solution was mixed with 30 ml of sucrose solution (specific gravity ~1.2) and centrifuged at 1000 g for 25 min. Oocysts were collected from the top of sucrose solution by overlaying 5 ml of deionized water, gently stirring, and pipetting 5 ml from the top. Oocysts were washed by mixing with deionized water and centrifuging at 1500 g for 15 min. The supernatant was discarded by aspiration, leaving a 1:1 ratio of pellet to solution volume. This final solution was weighed and 10 μl was overlaid on commercially prepared slides (Waterborne Inc., New Orleans, LA) and weighed too. Smears were air dried, stained using a direct immunofluorescence antibody kit (Meridian Bioscience, Inc., Cincinnati, OH), and examined using an Olympus BX60 microscope at ×400 magnification. Slides containing one or more oocysts were recorded as positive while those without detectable oocysts were recorded as negative. The crude number of oocysts per gram feces (O), unadjusted by assay percent recovery, was calculated according to the formula: O = [N × (Wsusp / Wsm)] / F, where N is the number of oocysts in the smear, Wsusp and Wsm are the suspension and smear weight respectively, and F is the fecal weight. Assuming daily consistent oocyst shedding, the Environmental Loading Rate [ELR] was calculated using the mean number of oocysts per gram of feces found in this study and a crude estimate for total daily fecal production estimated as ~3% of mean body mass or 0.02 kg feces wet weight per day for a typical adult marmot ( Atwill et al., 2003).

2.4. Statistical analysis

In order to evaluate the effect of environmental parameters on the likelihood of detecting Cryptosporidium oocysts in marmot feces, a hierarchical Bayesian logistic regression model was constructed using JAGS 3.4.0 (http://mcmc-jags.sourceforge.net, [9–15-10], [Plummer, M.]) in R 2.15.1 (http://www.r-project.org, [10-10-11], [Development Core Team]), with R2jags 0.03–08 (http://cran.r-project.org/web/packages/R2jags/index.html, [11–17-2012], [Yu-Su and Masanao]) as the interface. The outcome yij was the presence of Cryptosporidium oocysts (0 = no oocysts / 1 = oocysts present) in the ni feces collected at the j locations: y ij ∼ Bernouilli (p ij), where the probability of Cryptosporidium oocysts being present in the i th yellow-bellied marmot feces sampled at location j, pij was related to a suite of q fixed predictors at the feces level: elevation and vegetation.

logit (p ij) = α j [i] + β1 X 1 i +…+ β q X qi,

where αj[i] are random intercepts shared by feces belonging to location j, defined as:

a j ∼ Normal (β 0 + γ 1 τ 1, σ 2)

where γ 1 τ 1 is the group level predictor mode of human access used to reach each sampling location. We selected a hierarchical model because yellow-bellied marmot fecal samples were grouped by sampling location. This grouping was based on the assumption that the probability of Cryptosporidium infection in marmots from the same location may not be independent from each other and therefore may exhibit a level of positive correlation in and above what is accounted for in a fixed effects regression model. Moreover, a Bayesian approach is preferred over the frequentist because the former can avoid problems of identification when the multilevel model is complicated.

Given the continuous nature of the variable elevation, we evaluated graphically the assumption of linearity with respect to the log odds of a fecal sample testing positive for Cryptosporidium oocysts by categorizing elevation into 4 strata based on quartiles to get an equal number of observations per stratum elevation: 6,985 –8,265 (reference level), 8,265–9,068, 9,068–10,057, and> 10,057 ft. Across these four strata, there was a non-linear association between elevation and the log odds of Cryptosporidium oocysts shedding, therefore a quadratic term for elevation was considered. Moreover, to determine if effect modification was present between vegetation and elevation, the full dataset was stratified by vegetation status and the relationship reevaluated between elevation and the log odds of Cryptosporidium shedding within each strata.

For model construction, the variable elevation was centered at its mean: 9,068 ft, and the quadratic elevation term was constructed from these centered values. Non-informative priors or hyperpriors were assigned for β q, β 0 and τ 1 and σ 2: Normal (0, 25) for the first three parameters and LogNormal (0,1) for the last one. Posterior distributions for all parameters were sampled from each of three chains for 60,000 iterations following a 10,000 iteration burn-in, and thinning set to 5, for a total of 30,000 samples. Each chain was assigned random start values from a Normal (0,1) distribution for β 0 and β q and a Uniform (0,1) for σ. Convergence was assessed by the Gelman–Rubin statistic ( Gelman and Rubin, 1992). A backward stepping procedure was used to select terms for the final model, starting with a full model containing the random intercepts, the fixed effects, and two-way interactions: among elevation and vegetation, and quadratic elevation and vegetation. Final model selection was based on the deviance information criterion (DIC) ( Spiegelhalter et al., 2002) and credibility of parameters. Given the sample size and the sparse nature of data, fixed regression parameters with 90% Credible Intervals [CrI] that excluded 0 were considered to be credible. We assessed potential confounders by examining the change of the mean parameter estimate when they were included to the selected model. If the inclusion generated a change>10%, we kept the variable in the model.

2.5. PCR and sequencing

Fecal solutions of microscopic positive samples were exposed to 5 cycles of freeze (−80 ̊C) and thaw (+70 ̊C), then 0.2 g fecal solutions were used for DNA extraction using the DNA Stool Mini Kit (Qiagen®) according to the manufacturer’s manual. A nested PCR was performed using primers and reaction conditions amplifying a fragment of ~830 bp of the 18S rRNA gene according to previously described methods ( Xiao et al., 1999). A positive control using C. parvum DNA from California dairy calves (GenBank accession no. FJ752165) as template and a negative control without DNA template were included in each PCR. PCR products were verified by electrophoresis in 2% agarose gel stained with ethidium bromide. Verified PCR products were purified using a Qiaquick® spin columns (Qiagen ®) and sequenced at the University of California DNA Sequencing Facility, where ABI 3730 Capillary Electrophoresis Genetic Analyzer (Applied Biosystems Inc., Foster City, CA) was used for sequencing. Primers of the secondary PCR of the 18s rRNA gene were used for sequencing. A preliminary analysis of sequences was conducted using Vector NTI Advanced 11 software (Invitrogen Corporation, Carlsbad, CA) followed by a BLAST analysis to compare the sequences to existing Cryptosporidium spp . 18S rRNA gene sequences in the GenBank using the National Center for Biotechnology Information [NCBI] online blasting tool (http://blast.ncbi.nlm.nih.gov/).