Design‐ and model‐based recommendations for detecting and quantifying an amphibian pathogen in environmental samples

Abstract

Accurate pathogen detection is essential for developing management strategies to address emerging infectious diseases, an increasingly prominent threat to wildlife. Sampling for free‐living pathogens outside of their hosts has benefits for inference and study efficiency, but is still uncommon. We used a laboratory experiment to evaluate the influences of pathogen concentration, water type, and qPCR inhibitors on the detection and quantification of Batrachochytrium dendrobatidis (Bd) using water filtration. We compared results pre‐ and post‐inhibitor removal, and assessed inferential differences when single versus multiple samples were collected across space or time. We found that qPCR inhibition influenced both Bd detection and quantification in natural water samples, resulting in biased inferences about Bd occurrence and abundance. Biases in occurrence could be mitigated by collecting multiple samples in space or time, but biases in Bd quantification were persistent. Differences in Bd concentration resulted in variation in detection probability, indicating that occupancy modeling could be used to explore factors influencing heterogeneity in Bd abundance among samples, sites, or over time. Our work will influence the design of studies involving amphibian disease dynamics and studies utilizing environmental DNA (eDNA) to understand species distributions.

Host–pathogen metapopulation dynamics suggest high elevation refugia for boreal toads

“…Differences in Bd detection probability likely reflect heterogeneity in Bd load (Miller et al. ; Mosher et al., ), and we found lower Bd detection probabilities at high elevations.…”

“…, Chestnut et al. ; Mosher et al., ). Expanding survey methods to include standardized surveys for non‐target amphibians may also provide better information about host community structure and stability and disease prevalence in these other species.…”

A model‐based solution for observational errors in laboratory studies

“…Consequently, Mosher et al. () classified replicates as positive only if they exceeded the highest qPCR copy number observed in the control group and analysed the resulting data using classical occupancy models (e.g., Equation ).…”

“…We used the extended occupancy model (Equation ) to quantify the effect of Bd concentration on detection probability in natural water samples while fully accounting for false‐positive errors. For purposes of comparison, we also fit the classical occupancy model (Equation ) to the data after imposing the copy number threshold (as in Mosher et al., ). Given that samples were experimentally inoculated with Bd (and our focus here was specifically on pathogen detection), we fixed the latent occupancy state to for all samples during model fitting.…”

“…We present a model‐based approach that allows false‐positive errors in both negative and positive samples, thereby providing unbiased inference concerning the occurrence and detection processes. The modelling framework is general and broadly applicable because it only requires repeat sampling and trials on negative controls (i.e., known negative samples), data that are already collected in many diagnostic protocols (e.g., Carey et al., ; Davenport et al., ; Lachish et al., ; Mosher et al., ). We first describe occupancy models that address imperfect sensitivity and specificity, which we subsequently extend to fully accommodate false‐positive errors.…”