Environmental sampling to assess the bioburden of Mycobacterium avium subspecies paratuberculosis in drylot pens on California dairies

Prevalence estimation

Individual blood and fecal samples were collected from all the cows in the study pens (3 pens on each dairy, 6 pens total) to determine the prevalence of MAP prior to 3 days of environmental sampling for our reliability and repeatability study. On each dairy, no cows were moved to or from the study pens during the study period, except for sick cows moved to the hospital pen and such movements were documented. Blood samples for seroprevalence estimates were collected on November 13th and December 12th, 2017 from herd 1 and 2, respectively. Blood samples were taken from the coccygeal vein using 20 gauge vacutainer needles and serum tubes. Fecal samples were collected from the rectum using individual plastic sleeves for every cow, then transferred to two oz polypropylene tubes with attached lids. All samples were transferred on ice to the Dairy Epi lab (Aly Lab, Tulare, CA, USA). The blood samples in serum tubes were place upright in the rack at room temperature for at least 2 h from the collection to allow blood to clot. Blood samples were spun at 1 G for 5 min to harvest serum in the same day as collection. Serum samples were stored at 4 °C until testing by ELISA methods within 3 days of sample collection. Fecal samples from individual animals were stored at –80 °C until testing by qPCR.

Reliability and repeatability study

Both dairies’ herd managers were asked to stop flushing the study pens 24 h prior to the first day of environmental sampling. Environmental samples were collected at 24 h intervals for 3 days at approximately the same time of day by two sample collectors (study authors DW and TC). The flush lanes were flushed immediately following each sample collection on each day which allowed the fecal slurry to accumulate over the following 24 h and prior to the next sampling. Specifically, study pens on dairy 1 were sampled on November 14th–16th and on dairy 2 on December 13th–15th. Sample collector 1 had previous experience collecting environmental fecal samples for MAP testing. Sample collector 2 had no previous environmental sampling experience but had been trained prior to the study. Both collectors sampled fecal slurry from the same pens over the three study days to compare the similarity in MAP concentration in samples collected by different collectors (reliability) and on different days (repeatability). A grab sample of approximately 10 ml of fecal slurry was collected by hand with approximately every 1 m along the length of the flush lane, avoiding individual cow fecal pats. Disposable gloves were worn during the sampling and changed between each pen. Each collector collected fecal slurry into 2 L clean plastic containers. A clean container was used for each pen after being disinfected using Quaternary Ammonium Compounds, rinsed thoroughly with water and dried. After the container was filled with the fecal slurry from the entire pen’s flush lane, the slurry was mixed with a sterile wooden tongue depressor using 10 vertical stirs from the bottom to the top, followed by 10 clockwise, and finally 10 counterclockwise stirs. Approximately 25 ml of fecal slurry from each pen were then transferred to a 2 oz polypropylene tube with an attached lid which was then transported on ice to the laboratory. At the laboratory, each sample was split into 10 ml, stored in a smaller polypropylene wide opening screw cap container and the remaining volume stored into a conical polypropylene screw cap tube. All the fecal samples were stored at −80 °C until the entire study’s fecal sample repository was collected and ready for qPCR testing.

ELISA

Serum samples from both dairies were tested using an ELISA designed to detect the presence of antibodies against MAP (MAP Antibody Test Kit, IDEXX, Westbrook, ME, USA). The test was performed according to the manufacturer specifications at a temperature between 18 and 26 °C. The absorbance was measured at 450 nm. The mean optical density (OD) of two negative and two positive controls were calculated. The mean OD for the two positive control wells needed to be more than or equal to 0.350 and the ratio of mean OD of the positive to negative controls needed to be more than or equal to 3.00 in order for the test results to be considered valid. The S/P ratio for each sample was calculated by the sample OD minus the negative control mean OD divided by the positive control mean OD minus the negative control mean OD. Negative, suspect and positive results were determined at S/P ratios ≤0.45, 0.46–0.54 and ≥0.55, respectively. Initially, samples were tested in single wells, however, any samples with positive or suspect results were re-tested in duplicate. The final results of these samples were based on the average S/P ratio from the duplicate test wells. The samples with suspect results were treated as positive results in our calculation.

qPCR

To estimate the prevalence of fecal shedding in each of the study pens a survey sample of cattle from each pen were identified and tested by qPCR. Specifically, fecal samples from all the seropositive and ELISA suspect cows were qPCR tested. In addition, fecal samples from a random sample of ELISA negative cows were also qPCR tested. Hence, environmental fecal samples, as well as the fecal samples from MAP ELISA positive and suspect cows and randomly selected fecal samples from ELISA negative cows were transported to CAHFS laboratory at University of California Davis on dry ice for qPCR testing. The MAP’s DNA extraction process was performed using MagMAX Total Nucleic Acid Isolation Kit with the requirement of at least 0.3 g of fecal sample. Subsequently, each DNA sample was tested a commercially available kit (VetMAX Gold MAP PCR Kit, Thermo Fisher Applied Biosystems, Waltham, MA, USA) on a Cepheid SmartCycler II according to the manufacturer’s instructions. Ct values were recorded up to 40 but a sample was only considered positive if the reaction crossed threshold at less than 37 cycles.

Descriptive statistics

Descriptive statistics observed MAP seroprevalence and estimated MAP shedding prevalence were reported stratified by herd. The MAP shedding prevalence was based on qPCR and a survey sample, as described previously (Lavallée & Beaumont, 2015; Maier et al., 2019). Briefly, qPCR survey-adjusted prevalence results from seropositive or suspect cows were assigned a survey weight of one, while results from seronegative cows were assigned a survey weight that is the inverse of the sampling probability, with the sampling probability being the number of seronegative cows tested by qPCR in a pen divided by the number of seronegative cows in the pen.

Variance components models

For each outcome, there were four factor levels: dairy, pen, collector and day (Fig. 1). Pens were nested within dairy. In dairy i, i = {1, 2} were pens j; where for both i = 1 and i = 2, j = {1, 2, 3}. Pens were cross classified by collector k; k = {1, 2}; and day l; l = {1, 2, 3}.

Variance components models with random intercepts for dairy u_i and pen v_ij were used to account for variability attributable to dairies and pens within dairies, respectively. The cross-classification of pens by collector and day was addressed by including random intercepts, w_k for collector and z_l for day (Goldstein, 1987; Marchenko, 2006). All random effects (u_i,v_ij, w_k, z_l) and residual error (e_ijkl) were assumed to be normally distributed with mean = 0 and variances ${\sigma }_u^2,{\sigma }_v^2,{\sigma }_w^2,{\sigma }_z^2,{\sigma }^2$, respectively. For each study outcome (y_ijkl), cycles-to-threshold (CT) values, variance components were estimated from a multilevel mixed model (Eq. (1)). (1) $$y_{ijkl}={\beta }_0+u_i+v_{ij}+w_k+z_l+e_{ijkl}$$ Models were fit using restricted maximum likelihood (Jiang, 2007). Quantile plots of model residuals and standardized residuals were evaluated for normality. Similarly, the empirical best linear unbiased predictions (EBLUP) for dairy, pen, collector and day from qPCR models were estimated and evaluated for normality.

Estimates of variability attributable to dairy, pen, collector and day as a percent of the total variability were computed for each test result. The intraclass correlation coefficient (ICC), an estimate of similarity in results of samples from the same cluster, was estimated for the following three ICC: (2) $$(1)\mathrm{I}\mathrm{C}\mathrm{C}\left(\mathrm{d}\mathrm{a}\mathrm{i}\mathrm{r}\mathrm{y},\mathrm{p}\mathrm{e}\mathrm{n},\mathrm{d}\mathrm{a}\mathrm{y}\right)=\frac{{\sigma }_{\text{u}}^2+{\sigma }_{\text{v}}^2+{\sigma }_{\text{z}}^2}{{\sigma }_{\text{u}}^2+{\sigma }_{\text{v}}^2+{\sigma }_{\text{w}}^2+{\sigma }^2}$$to estimate the similarity in MAP concentrations in environmental samples collected by different collectors (from the same dairy, pen and on the same day); (3) $$(2)\mathrm{I}\mathrm{C}\mathrm{C}\left(\mathrm{d}\mathrm{a}\mathrm{i}\mathrm{r}\mathrm{y},\mathrm{p}\mathrm{e}\mathrm{n},\mathrm{c}\mathrm{o}\mathrm{l}\mathrm{l}\mathrm{e}\mathrm{c}\mathrm{t}\mathrm{o}\mathrm{r}\right)=\frac{{\sigma }_{\text{u}}^2+{\sigma }_{\text{v}}^2+{\sigma }_{\text{w}}^2}{{\sigma }_{\text{u}}^2+{\sigma }_{\text{v}}^2+{\sigma }_{\text{w}}^2+{\sigma }_{\text{z}}^2+{\sigma }^2}$$to estimate the similarity in MAP concentrations in environmental samples collected on different days (from the same dairy, pen and by the same collector); and (4) $$(3)\mathrm{I}\mathrm{C}\mathrm{C}\left(\mathrm{d}\mathrm{a}\mathrm{i}\mathrm{r}\mathrm{y},\mathrm{p}\mathrm{e}\mathrm{n}\right)=\frac{{\sigma }_{\text{u}}^2+{\sigma }_{\text{v}}^2}{{\sigma }_{\text{u}}^2+{\sigma }_{\text{v}}^2+{\sigma }_{\text{w}}^2+{\sigma }_{\text{z}}^2+{\sigma }^2}$$to estimate the similarity in MAP concentrations in environmental samples collected by different collectors on different days (from the same dairy and pen). The 95% confidence intervals of the ICC were computed using the delta method. The Spearman rank correlation coefficient for MAP fecal shedding prevalence and MAP bioburden was estimated. The following ranges were used to interpret the correlation and ICC estimates: <40%, poor; 41–75%, fair to good; >75%, excellent (Fleiss, Levin & Paik, 2003). All analyses were performed using Stata (Stata/IC V. 15.0, College Station, TX, USA).