# Batrachochytrium salamandrivorans (Bsal) not detected in an intensive survey of wild North American amphibians

Aug 3, 2020

### Field sampling

We defined a sampling universe based on the presence of non-zero estimated introduction risk in the contiguous United States7 within the range of salamander species known to be susceptible to Bsal2. Because of their presumed high susceptibility to Bsal2 and large ranges24, we targeted newts of the genera Notophthalmus and Taricha in the eastern and western U.S., respectively. Challenge trials for most of the diverse amphibian fauna in the U.S. were lacking when we designed the study, though we expected that some anuran species can serve as infection reservoirs of Bsal25,26,26. We therefore sampled other amphibian species (anurans and caudates) as well (Supplement).

We set a target of 10,000 samples across the United States. Our expectation was that if Bsal was introduced into the U.S. it would most likely be transmitted by an infected individual intentionally released. Site selection was therefore non-random as we sought sites that were generally accessible to the public or near areas frequented by visitors. Accordingly, we avoided remote areas that would be less prone to such an introduction. We defined a site as a waterbody, wetland, or group of proximate aquatic habitats that could reasonably be epidemiologically linked based on the transmission of Bsal by annual host movements or transport of infective stages via water. We aimed to capture 30 animals per site (N), which would result in 90% certainty of detecting Bsal when present, assuming a Bsal detection probability (p) of 0.75 on an infected individual27 and a presumed low prevalence value of 0.10. We recognize that a prevalence value lower than 0.10 may be possible, but it would have been prohibitively difficult to sample more individuals per site.

$$Certainty = 1 – [theta *(1 – p) + (1 – theta )]^{N}$$

(1)

We captured target animals by hand, net, or trap. After capture, we handled each animal separately using disposable, powderless vinyl gloves and new, clean plastic bags to avoid cross contamination. All handling of animals was conducted in accordance with relevant guidelines and with appropriate collecting permits. All experimental protocols were approved by U.S. Geological Survey Institutional Animal Care and Use Committee. Appropriate permit numbers and information may be obtained from first author upon request. We rubbed rayon-tipped sterile swabs (MW-113, Medical Wire & Equipment, Corsham, England) over the plantar side of one front and one hind limb, the ventral tail surface of caudates, the dorsal side of the body, and the ventral surface of the body 5 times each28. We placed the swabs into sterile plastic vials with 20 μl of sterile deionized water. We recorded the snout-vent length, sex, and any visible signs of skin lesions for each individual. We collected two separate swabs from each animal, holding one in reserve to provide confirmation if Bsal was detected on the first swab. We chilled swabs immediately after field collection and subsequently froze them at ≤ − 20 °C within 3 days. Frozen swabs were sent to the U.S. Geological Survey’s National Wildlife Health Center in Madison, Wisconsin, for analysis.

### Molecular methods

We extracted DNA from swabs as described by Hyatt et al.29 except that 125 μl of PrepMan® Ultra Sample Preparation Reagent (Applied Biosystems, Foster City, CA) and 100 mg of zirconium/silica beads (Biospec Products, Bartlesville, OK) were used so that the entire swab was immersed. The bead-beating steps were conducted using a FastPrep®-24 homogenizer (MP Biomedicals, Santa Ana, CA). We used a real-time TaqMan polymerase chain reaction (PCR) for detection of Bsal on the extracted DNA as described in Blooi et al.30,31,31. We ran reactions on the 7,500 fast real-time PCR system (Applied Biosystems, Foster City, CA) using QuantiFast Probe RT-PCR mastermix kit with ROX dye (Qiagen, Valencia, CA) and BSA as per the kit instructions. We used five microliters of the PrepMan® solution containing the extracted DNA as template for the PCR. We included a negative extraction control and a standard curve run in duplicate on each PCR plate. The standard curve consisted of five different concentrations of the target sequence for Bsal inserted into plasmids. The concentrations of the standards occurred at ten-fold dilutions ranging from 110–1,100,000 copies (0.5–5,000 fg DNA) per reaction (on some initial runs, the standard range was 11–110,000 copies per reaction). The threshold for signal detection was set at 5% of the maximum fluorescence of the standards run for that assay. We considered a positive detection of Bsal DNA if a detectable signal existed at 37 or fewer PCR cycles and no detection in all other cases. We calculated the efficiency of each run using standard curve amplification and repeated PCR plates with an efficiency of less than 90% or greater than 110%.

### Data analyses

The probability of failing to detect a species given that it occurs is different than the probability of occurrence given non-detection32. We focused on this latter quantity and estimated the average probability of Bsal occurrence at sampled sites, given non-detection data, survey effort, and alternative hypotheses about the status of Bsal in the U.S. We defined occupancy as the probability of Bsal occurrence at the site level and prevalence as the probability of Bsal occurrence on an individual. Under this latter definition, prevalence included both infections and Bsal zoospores from the environment that might be detected on the skin of an infected individual. The probability of Bsal occurrence given non-detection was represented probabilistically as Pr(zi = 1|Σ(yij) = 0), where zi is the latent occupancy state for site i (zi = 1 for occupied sites and zi = 0 for unoccupied sites) and yij is the imperfectly observed pathogen status of a sampled individual j at site i. At occupied sites, observations were a product of the pathogen status of the individual (wj = 1 for pathogen positive individuals and wj = 0 for pathogen negative individuals) and the probability of detecting Bsal on infected individuals (p).

Using Bayes Theorem, the probability of Bsal occurrence at a single site i conditional on non-detection ((varphi_{i})) can be calculated using prior expectations about Bsal occupancy ((psi_{prior})) and prevalence ((theta_{prior})). In addition, the total number of individuals sampled at each site (N) and the total number of replicates collected per individual (K) were considered.

begin{aligned} varphi_{i} = Pr {(}z_{i} = 1{|}sum y_{ij} = 0) & = frac{{Pr {(}sum y_{ij} = 0{|}z_{i} = 1){Pr}left( {z_{i} = 1} right)}}{{Pr {(}sum y_{ij} = 0{|}z_{i} = 1){Pr}left( {z_{i} = 1} right) + {Pr}left( {z_{i} = 0} right)}} \ & = frac{{left( {left( {1 – theta } right) + theta left( {1 – p} right)^{K} } right)^{N} psi }}{{left( {left( {1 – theta } right) + theta left( {1 – p} right)^{K} } right)^{N} psi + left( {1 – psi } right)}} \ end{aligned}

(2)

This equation yields the probability that an observation of Bsal non-detection came from an occupied site (i.e., a false negative), given survey effort (K, N) and prior expectations about pathogen detectability ((p)), occurrence ((psi)) and prevalence ((theta)).

Prior expectations were derived from four hypotheses about Bsal invasion in the United States using this probabilistic framework (Table 1), leading to different predictions about Bsal occurrence and prevalence. If Bsal is endemic to the U.S., we expect it to be widespread within suitable habitats (high (psi)). If Bsal invaded the U.S. recently, we expect it to be present at a small proportion of locations (low (psi)). This hypothesis is unlikely biologically given what we know about Bsal and invasive pathogens and it is only included here for theoretical completeness. We would expect that additional extensive sampling would fail to increase the posterior probability of this state of nature. In addition, and independent of occurrence rates, Bsal transmission within an infected population may vary. Reported Bsal prevalence values from field studies range across species, sites, and with time since invasion17,18,18. Therefore, we also consider two categories of site prevalence: a rapid transmission scenario where Bsal prevalence is high within infected populations (high (theta)), and a slow transmission scenario ((theta)) where Bsal prevalence is low within infected populations. To evaluate the probability that Bsal was present at any of our sampled sites given non-detection, we calculated Eq. (2) for each site across a range of occurrence ((psi ,)= 0.05–0.95) and prevalence ((theta ,)= 0.05–0.95) values and used the mean result of Eq. (2) ((hat{varphi })) from all our sampled sites as the metric to summarize the probability of Bsal presence in our sampling frame.