Experimental inoculation trial to determine the effects of temperature and humidity on White-nose Syndrome in hibernating bats

All methods in this study were approved by the Institutional Animal Care and Use Committee at Texas Tech University (protocol 18032-12). All procedures were performed in accordance with relevant guidelines in the manuscript and the ARRIVE (Animal Research: Reporting of In Vivo Experiments) guidelines (https://arriveguidelines.org/).

Experimental design for testing effects of temperature and humidity on Pd infection severity on Perimyotis subflavus

We randomly assigned bats to seven environmental chambers (Caron, Model 7000-33-1, Marietta, Ohio, USA) in a blocked experimental design, controlling temperature and humidity in each chamber (Fig. 1). In each environmental chamber, we divided bats into two cages (23 × 38 × 50 cm) constructed from mesh fabric (Part FMLF, Seattle Fabrics, Inc., Seattle, Washington, USA), PVC pipe, and plastic sheeting. We stratified random assignment to ensure even distribution of initial body mass and sex across microclimate treatments. In addition to the seven treatments with fixed temperature and humidity conditions, we had two treatments that allowed bats to freely move among temperature or humidity conditions (Fig. 1). One group of bats (n = 14) was free to move among three chambers with a common temperature (8 °C) but different humidity (water vapor pressure deficit (VPD) = 0.05 kPa, 0.10 kPa, or 0.15 kPa, corresponding to 95, 90, and 85% relative humidity (RH))³⁶. A second group of bats (n = 14) was free to move among three chambers with a common VPD condition (0.10 kPa, medium humidity) but different temperatures (5, 8, or 11 °C) (Fig. 1). Because our research questions were focused on comparing the effect of temperature and humidity conditions on disease severity, we did not include sham-inoculated control animals in the experiment. We made this decision to reduce the total number of animals used in the experiment and to maximize replication to test the effects of temperature and humidity on disease.

We inoculated each bat by spreading 20 µL of Pd solution (5 × 10⁵ conidia µL⁻¹) evenly across both wings, following established protocols^8,9,32,37; treatments were conducted blind without knowledge of which bat was being assigned to what group and bats were inoculated in no particular order to reduce the confounding influence on the order of treatment. We used a Pd strain collected by Karen J. Vanderwolf at Trent University from naturally infected Myotis lucifugus. We cultured Pd on Sabouraud Dextrose Agar with chloramphenicol and gentamicin (SabDex) (Part L96359, Fisher Scientific, Houston, Texas, USA) and incubated subcultured plates at 10 °C for 60 days to allow the formation of conidia. We then harvested conidia by flooding plates with phosphate buffered saline solution containing 0.5% Tween20 (PBST). Conidia were resuspended in PBST, enumerated, and diluted to the inoculum concentration⁸.

Microclimate treatment conditions

We used three temperatures 5, 8, or 11 °C to represent a range of roosting temperatures of P. subflavus in natural hibernacula^24,29. We set humidity in environmental chambers to achieve specific levels of water vapor pressure deficit (VPD) between the surface of the bat and the environment because relative humidity varies by temperature³⁶. Higher VPD corresponds to drier air resulting in higher potential evaporative water loss (pEWL). We used three levels of VPD: 0.05, 0.10, or 0.15 kPa corresponding to low pEWL (high humidity), medium pEWL (medium humidity), and high pEWL (low humidity) levels (Fig. 1). We verified the ambient temperature and relative humidity in each chamber at 10-min intervals (Hobo Model U23-001, Onset Computer Corporation, Bourne, Massachussetts, USA). For bats in the connected chambers that could behaviorally select their temperature and humidity conditions, we quantified the number of days bats spent in each condition³⁸.

Animal handling and data collection

We used 98 (42 females, 56 males) tricolored bats collected on 10 December 2018 from culverts in Mississippi and transported directly to Texas Tech University³⁹. We took morphometric measurements (body mass ± 0.1 g, forearm length ± 0.1 mm) and used quantitative magnetic resonance (QMR; Echo-MRI-B, Echo Medical Systems, Houston, Texas, USA) to determine pre-hibernation fat at the start of the experiment^39,40. As an indicator of pre-hibernation stress, we collected a fur sample from the dorsal intrascapular region to quantify fur cortisol concentration with a commercial ELISA kit, following the manufacturer’s protocol (Arbor Assays, Michigan, USA) (see Supplemental Methods). Fur is moulted once per year in the late summer period⁴¹ and therefore fur cortisol reflects the level of circulating cortisol during the period of fur growth prior to hibernation. We attached a uniquely marked, modified datalogger⁴² (DS1925L iButton, Maxim Integrated, San Jose, California, USA) to the back of each bat using ostomy cement to record skin temperature³⁹. Prior to inoculation, we swabbed bats with a sterile polyester swab (Fisherbrand synthetic tipped applicators 23-400-116) five times on forearm and five times on muzzle to determine if any bats were naturally infected with Pd at time of collection. Swabs were stored in RNAlater at  − 20 °C until testing using quantitative polymerase chain reaction (qPCR) at Northern Arizona University⁴³.

During the experiment, we provided ad libitum drinking water in each cage but did not provide food. We secured a motion-activated infrared camera (Model HT5940T, Speco Technologies, New York, New York, USA) above each cage to monitor bats throughout the experiment. Because one camera failed at the start of the experiment, we combined bats in that treatment chamber into a single cage (Fig. 1) and replicated this disturbance among all chambers. We monitored bats without disturbance by reviewing video recordings daily. Three bats died of unknown cause before the end of the experiment and were removed from analyses.

After 83 days of hibernation, we terminated the experiment and bats were removed from cages and processed to determine body condition using QMR³⁹. We took respirometry measurements on a subset of animals³⁸, and swabbed for Pd as described above. We photographed the left ventral wing using ultraviolet (UV) transillumination (368-nm wavelength and 2-s exposure) to detect and measure florescence associated with Pd infection^37,44. For histology, we removed the wing section from the fifth digit and the body and rolled wing tissue around dental wax dowels and 10% neutral buffered formalin. We collected a 90–110 µL blood sample in lithium-heparin-treated capillary tubes for immediate analysis of blood chemistry with a handheld analyzer (i-STAT1 Vet Scan, Abaxis, Union City, California, USA). Using an EC8+ cartridge, we measured sodium, potassium, chloride, anion gap, glucose, BUN (urea nitrogen), hematocrit, hemoglobin, pH, pCO2, TCO2, HCO3, and base excess (Table S1). We quantified arousals from torpor as reported by McGuire et al.³⁹. All bats were handled and euthanized under Animal Care and Use Committee permit 18032-12 at Texas Tech University.

Infection and disease metrics

We used several metrics to determine pathogen and disease presence and severity³⁷: presence and amount of the pathogen, Pd, on a bat were determined by qPCR⁴³, and presence of the disease, WNS, was determined via detection of orange-yellow florescence under UV light characteristic of Pd infection⁴⁴ and histological presence of characteristic lesions and pustules with fungal hyphae^45,46. Three types of cutaneous infection were described histologically, including characteristic cupping erosions with fungal hyphae, neutrophilic pustules with fungal hyphae, and fungal hyphae in the stratum corneum with dermal necrosis. Any bats with any of these three conditions noted were scored as WNS positive by histology. Presence and quantity of DNA of Pd was tested by qPCR at Northern Arizona University. All samples were run in duplicate and considered positive if at least one run was positive below a cycle threshold (Ct) of 40 and quantified using a quantification curve from serial dilutions (nanograms of Pd using the equation load = 10^{((22.049-Ct value)/3.34789)}, r² = 0.986)⁴⁷. Load values were averaged across multiple runs and then converted to attograms by multiplying loads in nanograms by 10⁹.

Statistical analyses

We used three different response variables (Pd prevalence, Pd loads, and WNS prevalence by histology) to determine whether infection status varied by microclimate treatment conditions. Low sample sizes of positive infection status by UV detection (n = 4) precluded use in statistical analyses (Table 1). We used generalized linear models with binomial distribution for analyses of Pd prevalence and WNS prevalence and a linear mixed effects model with Gaussian errors for Pd loads. Although the experiment was designed with replication at the cage level to account for cage effects, we were unable to include cage as a random effect because of the low numbers of bats that had signs of Pd or WNS infection. We analyzed whether infection status (i.e., Pd prevalence, Pd load, or WNS prevalence) varied by sex and cortisol separately from an a priori candidate model set (Table 2) to cope efficiently with small sample sizes. We first asked whether infection response varied by sex to determine if bats could be pooled in subsequent analyses. We analyzed separately whether infection response varied by pre-hibernation cortisol at the start of the experiment on the subset of animals for which we had cortisol measurements (n = 83). We then used an information-theoretic approach comparing a candidate set of models with Akaike Information Criterion (AIC)⁴⁸ using initial fat mass as an individual covariate and temperature and humidity treatment conditions as categorical treatment groups to assess the effect of microclimate on infection response (Table 2). Bats behaviorally selecting their temperature and humidity conditions were assigned to a temperature or humidity treatment level if a bat spent > 89% of captive days at that condition or was otherwise placed in an ‘inconstant condition’ treatment group. For WNS prevalence, we used the bias reduction method implemented in package brglm⁴⁹ to deal with complete separation present in the data (in some treatments all bats were scored as negative for WNS) (Table 1; Fig. 2).

Because this was the first captive hibernation experiment with P. subflavus, we investigated the effects of temperature and humidity on the hibernation physiology of the species^38,39 and how physiological markers (e.g., blood chemistry) may be associated with disease. To determine if physiological indicators were related to infection status at the end of the experiment, we compared total number of torpor arousal bouts during the experiment and 13 different blood chemistry metrics from blood samples taken at the end of the experiment and used t-test comparisons (at α = 0.05) for each metric between Pd/WNS positive and negative bats. We designated bats as Pd/WNS positive if a bat tested positive for either Pd or WNS by qPCR, UV, or histology. We used Program R version 3.6.2 to conduct all analyses.

Experimental design for testing effects of temperature and humidity on Pd growth on substrates

We used five environmental chambers (CARON, Model 7000-33-1, Marietta, Ohio, USA) to test for the effects of temperature and humidity on fungal growth on natural and artificial substrates (Fig. S1). Our experimental design comprised a reduced temperature series and humidity gradient than what we used for the experiment on bats. In the humidity gradient, temperature was held constant at 8 °C, with 85%, 90%, and 95% RH representing our low, medium, and high humidity treatments, respectively. In the temperature series, vapor pressure deficit (VPD) was held constant across the low (5 °C), medium (8 °C), and high (11 °C) temperatures (VPD = nominally 0.01 kPa, range (0.105–0.107). The chamber set to 8 °C and 90% humidity (VPD = 0.107 kPa) was common to both series.

Media plate inoculation and fungal growth measurement

We constructed modified plate lids to prevent contamination while allowing humidity to equilibrate across the plate lid. We drilled 14 equidistant holes (5.5 mm diameter) into each plate lid and hot glued a piece of circular filter paper to the top of the lid. Lids were then disinfected thoroughly with a hydrogen peroxide wipe before being placed in a disinfected, sealed storage container.

We prepared Pd inoculum as described above for the infection trial on bats. We inoculated 30 SabDex plates with 100 µL of inoculum at a concentration of 20 conidia µL⁻¹ by serial dilution with a starting concentration of 2.0 × 10⁴ conidia µL⁻¹ diluted four times by a factor of 10. We used sterile, individually wrapped 1-µL plastic inoculation loops to spread the inoculum evenly across the surface of the plates, added the modified plate lids, and immediately transferred plates into environmental chambers. We included six replicate plates in each of the five microclimate conditions.

We took weekly digital photographs (Nikon, Model 26524, Tokyo, Japan) of each plate for the 5-week duration of the experiment (Fig. 3A). Our camera was mounted on a tripod to ensure consistent placement of plates relative to the camera. Each photo included a ruler, which was used to calibrate measurements made in ImageJ (Version 2.0.0-rc-69/1.52p, National Institutes of Health, Bethesda, Maryland, USA). One observer made all measurements for consistency. We used the freehand selection tool to trace the boundary of each fungal colony using a drawing tablet (Wacom, Model CTL-490, Kazo, Saitama, Japan). From these selections, we obtained the total surface area growth as the sum of all area selection (in cm²).

We modelled the growth of Pd on each plate as a sigmoidal curve (Fig. 3B), which we fit using the SSlogis and nls functions in Program R v. 3.6.3⁵⁰. The model fitting function provides an estimate of the inflection point of the curve, and we calculated the slope at the inflection point to estimate the maximum growth rate. We also estimated the latency to rapid fungal growth on the plates by determining the date at which the total area of fungus on the plate reached 2.5 cm² as an arbitrary threshold.

We also quantified growth of individual colonies. To avoid biasing growth rate estimates, we excluded colonies that intercepted another colony by choosing independent colonies at the final time point and tracking them backwards through time. If there were fewer than 10 independent colonies at the final time point, we added additional unimpeded colonies with each earlier time point until the total number of colonies reached 10. We modelled growth of individual colonies following the same procedure as for total area of growth on the plate, with an arbitrary threshold of 0.05 cm² for latency calculations. We used linear mixed models to test for the effects of temperature and humidity on maximum growth rate or latency, including plate as a random factor to account for measuring multiple colonies per plate.

Rock inoculation and fungal growth measurement

To evaluate fungal growth and persistence on a natural substrate, we inoculated pieces of sandstone flagstone. We etched a 4 × 6 sampling grid, composed of 5 × 5 cm squares, onto the surface of each sandstone rock (Texas Rock and Flagstone, Lubbock, Texas, USA), where each square served as a sampling unit (Fig. S2). Each row represented a time series for a single replicate, while each column was composed of replicates for the respective time point. Rocks were then autoclaved at 121 °C for 40 min and stored individually in a disinfected, sealed container until inoculation. At the time of inoculation, we evenly spread 200 µL of inoculum (2.5 × 10⁴ conidia µL⁻¹) across each sampling square and immediately transferred the rock to an environmental chamber.

We measured fungal growth at days 0, 14, 28, and 56. We used a sterile cotton swab to collect fungal DNA from each sampling square. Swabs were moistened with RNAlater and rolled horizontally, vertically, and diagonally across the surface of the sampling square to ensure contact with the total surface area. One researcher collected all swabs to maximize consistency among swabs collected throughout the experiment. Swabs were placed in RNAlater and stored at − 20 °C until shipped to Northern Arizona University for qPCR analysis⁴³. We quantified fungal loads for each swab sample from qPCR using the quantification curve provided above and normalized fungal loads to the value at day zero for each rock respectively. We then used linear models to test for effects of temperature and humidity on changes in fungal load (log transformed) over time.

To evaluate viability of Pd, we swabbed the entire inoculated surface of each rock at the end of the experiment and vortexed the swabs in RNAlater for one minute to release fungal DNA from the swab. We then applied 100 µL of RNAlater fungal solution from each rock to a respective SabDex media plate, using a sterile inoculation loop. After 2 weeks of incubation at 11 °C and 92% RH, we visually assessed plates for presence of fungal growth to determine viability of Pd collected from rocks at the end of the growth experiment.

Source: Ecology - nature.com