Epidermal epidemic: unravelling the pathogenesis of chytridiomycosis

Amphibians are the most threatened class of vertebrates, with approximately 30% globally classified as threatened with extinction (IUCN, 2017). Although anthropogenic disturbances are the main cause of amphibian declines (Stuart et al., 2004), roughly 27% of these declines have occurred in pristine habitats such as protected national parks (Pimm et al., 2014). Many of these non-anthropogenic declines have been attributed to a cutaneous pathogen, Batrachochytrium dendrobatidis (Bd), which infects the keratinised layers of amphibian skin and can cause the disease chytridiomycosis (Berger et al., 1998; Pessier et al., 1999). Amphibian skin is unique amongst tetrapods given its high permeability, which allows it to serve a variety of physiological roles such as cutaneous gas exchange and osmotic and ionic regulation, but this necessitates a moist epidermis and so the active secretion of aqueous substances from the dermal glands renders amphibians more susceptible to relatively high rates of evaporative water loss (Boutilier et al., 1992; Larsen et al., 2014). Given the importance of physiological regulation via the skin, a disruption in cutaneous function tends to have serious consequences for amphibian health (Pessier, 2002).

Amphibians with chytridiomycosis display altered behaviour (lethargy, lack of appetite) and suffer physical damage to the skin (cutaneous erythema, hyperkeratosis), leading to the loss of physiological homeostasis (low electrolyte levels) (Berger et al., 2005; Voyles et al., 2012; Peterson et al., 2013). In healthy frogs, electrolytes are exchanged through paracellular spaces or transcellularly via ion transport proteins (channels, co-transporters, exchangers, ATPase) in cutaneous epithelial cells (Hillyard et al., 2008). In frogs with chytridiomycosis, low levels of circulating electrolytes (hyponatremia and hypochloremia) correlate with a loss of cutaneous ion uptake capacity (Voyles et al., 2009). Bd produces a complex mixture of proteins (proteases, biofilm-associated proteins and a carotenoid ester lipase) that can disrupt epidermal intercellular junctions (Brutyn et al., 2012) and suppress genes related to the production of keratin and collagen (Rosenblum et al., 2012). In addition to its effects on skin integrity, Campbell et al. (2012) suggested that cutaneous Bd infections may directly inhibit the epithelial sodium channels (ENaC), which are primarily responsible for the cutaneous re-uptake of Na⁺ from cutaneous secretions and/or from the environment (Schild, 2010; Larsen and Ramløv, 2013). This hypothesis was based on their observation that the amiloride-sensitive short-circuit current of the skin of infected frogs was lower than that of healthy frogs (Voyles et al., 2009). Dysfunction of ENaC transport is often associated with disorders of Na⁺ and fluid homeostasis, and blood pressure (Schild, 2004). How Bd influences cutaneous ion uptake pathways, specifically those relating to Na⁺ movement, including ENaC and Na⁺/K⁺-ATPase (NKA), which is responsible for generating electrochemical gradients in the epidermis (Lingrel and Kuntzweiler, 1994), remains unknown.

Given the vital role of the skin in sustaining amphibian homeostasis, maintaining skin integrity and function is of considerable importance. To this end, the outer keratinised layer is periodically removed and replaced by ‘sloughing’. Sloughing also plays an important role in regulating cutaneous microbe abundances (Meyer et al., 2012; Cramp et al., 2014), and in frogs infected with Bd, sloughing helps remove Bd from the keratinised layer (Ohmer et al., 2017). Indeed, Ohmer et al. (2017) found that sloughing reduced Bd load in five anuran species, with less susceptible species clearing infection. However, susceptible species continued to develop chytridiomycosis despite an increase in sloughing frequency, and in spite of a temporary reduction in Bd load after sloughing (Ohmer et al., 2017). An increase in sloughing frequency might act as an immune response to remove the pathogen before the onset of disease (Ohmer et al., 2015). Importantly though, increased sloughing frequency may be a double-edged sword, as sloughing itself causes transient osmoregulatory disruption in amphibians (Jørgensen, 1949; Wu et al., 2017). In healthy amphibians (Rhinella marina), sloughing is accompanied by an increase in cutaneous permeability (Wu et al., 2017). However, an increase in the expression and activity of epithelial Na⁺ channels offsets the temporary increase in skin permeability such that animals suffer no loss of physiological homeostasis (Wu et al., 2017). Conversely, in green tree frogs (Litoria caerulea) with high Bd loads, cutaneous ion loss is substantially elevated during non-sloughing periods and increases further during sloughing (Wu et al., 2018). Given that animals slough more frequently, the cumulative impact of sloughing and increased ion loss during non-sloughing periods leads to the loss of physiological homeostasis. Although it is clear that Bd affects cutaneous ion transport processes in infected frogs, the mechanistic basis for the disruption of ion transport, and the effects of sloughing on this, remains unclear.

Understanding how the disruption of skin function leads to the loss of physiological homeostasis in frogs with chytridiomycosis, and whether sloughing worsens cutaneous regulation, is important from a management standpoint. A greater understanding of the mechanistic basis for loss of homeostasis in infected frogs could lead to better treatment options, particularly for critically endangered species and captive insurance populations. Thus, the aim of this study was to investigate the effects of Bd on cutaneous ion transport processes, focusing on the abundance, activity, and expression of ENaC and NKA in a susceptible species, the Australian green tree frog. We hypothesised that both increased skin permeability and the reduction or inhibition of regulatory ion transporters resulting from Bd infection would contribute to the disruption of cutaneous ion flow, and sloughing would further exacerbate electrolyte permeability in infected frogs. The resulting disruption of both chytridiomycosis and sloughing could prolong electrolyte imbalance, causing conditions of low ion concentrations in the blood plasma such as hyponatremia.

Litoria caerulea (White 1790) spawn was collected from Bribie Island, southeast Queensland, Australia, in March 2015, and raised in the laboratory at The University of Queensland until metamorphosis. The resulting 20 juveniles (10–20 g) were used for experimentation. An additional 17 L. caerulea (15–70 g) adults and subadults were collected from wet roads in non-protected areas near Fernvale, southeast Queensland, Australia, in January 2015. Frogs were housed individually in either small (235×170×120 mm) or large (265×235×12 mm) ventilated clear plastic containers, with paper towels saturated with chemically aged water (dilution 1:4000; VitaPet, NSW, Australia) as substrate, and a half PVC pipe for shelter. The lighting conditions were set at a 12 h:12 h light:dark photoperiod cycle, and temperature was set at a constant 20.5±0.5°C. Frogs were checked daily and fed once a week on vitamin-dusted crickets (Acheta domesticus), and enclosures were cleaned weekly. Prior to experiments, all frogs were swabbed to confirm the absence of Bd infection given its widespread distribution in natural L. caerulea populations (Berger et al., 1998). A quantitative polymerase chain reaction (qPCR; details below) assay was used to measure Bd loads following protocols by Ohmer et al. (2015) before experiments began. All animals were collected under the Queensland Department of Environment and Heritage Protection Scientific Purposes Permit (WISP15102214), and procedures in this study were conducted with approval of The University of Queensland's Animal Ethics Welfare Unit (SBS/316/14/URG).

The intermoult interval (IMI; time between two successful sloughing events) was monitored continuously using infrared surveillance cameras (Eonboom Electronics Limited, and K Guard Security, New Taipei City, Taiwan), and a generic 16-channel H.264 Digital Video Recorder (DVR) system, mounted to a moveable metal frame in front of enclosures, with two cameras per row. Each camera monitored two frog enclosures at a time, at a sample rate of 1.56 frames s⁻¹.

Bd strain EPS4 from Ohmer et al. (2015) was used for experimental infections. Cultures were maintained at 4°C until 4–5 days before exposure. Isolate EPS4 was passaged to sterile 1% agar, 0.5% tryptone, 0.5% tryptone-soy plates and maintained at 20°C. After 4–5 days, zoospores were harvested by flooding plates with aged tap water for 30 min. The zoospore suspension was collected, and the concentration was measured using a haemocytometer (Boyle et al., 2004). A randomised subset of frogs (n=21) was exposed to ∼500,000 zoospores. Frogs were exposed to Bd for 5 h in 300 ml plastic containers containing zoospores suspended in 100 ml aged water. Control animals (n=16) were exposed to aged water only. At 14 days post-exposure, each frog was swabbed with a sterile fine-tipped cotton swab (MW100-100; Medical Wire & Equipment, Wiltshire, UK) three times over the frog's ventral surface, thighs, armpit, forelimb feet and hindlimb feet (Kriger et al., 2006; Ohmer et al., 2015) to assess infection status. To determine infection load as zoospore equivalents (ZE), swabs were extracted in 50 µl PrepMan Ultra (Applied Biosystems) and analysed in duplicate with qPCR in a thermal cycler (MiniOpticon™ Real-Time PCR Detection System, Bio-Rad Laboratories) with a modified 15 µl Taqman reaction following Ohmer et al. (2015). After 1 month, exposed frogs with no Bd-positive swabs were re-exposed with ∼500,000 zoospores following the same procedure detailed above. The infection load or number of zoospore equivalent (ZE) was calculated from the mean value of each triplicate assay and log+1 transformed [log(ZE+1)].

All frogs were checked daily for clinical signs of disease progression including a lack of appetite (monitoring food intake), abnormal posture, excessive or irregular sloughing, lethargy, cutaneous erythema and discolouration, and loss of righting reflex (Voyles et al., 2009). Once the IMI was determined for each frog (both uninfected and infected animals), frogs from the uninfected (n=8 intermoult, n=8 sloughing) and infected (n=11 intermoult, n=7 sloughing) groups were anaesthetised with an intracoelomic injection of 60 mg kg⁻¹ body mass thiopentone sodium (Thiobarb Powder, Jurox Pty Limited, NSW, Australia) and then euthanised via double pithing. Animals were euthanised at one of two time points: (1) sloughing (no more than 1 h after the onset of sloughing behaviours) or (2) intermoult (at a point halfway through the IMI). Sloughing behaviour takes approximately 5–10 min to complete. Snout–vent length (SVL; mm) and body mass (g) were measured immediately. A final skin swab was also taken to determine Bd infection load.

Whole blood was collected from euthanised animals via cardiac puncture into a lithium heparinised syringe. Samples were then centrifuged at 5000 g for 5 min and the plasma collected and stored at −20°C for subsequent electrolyte analysis. Plasma Na⁺ and K⁺ levels (mmol l⁻¹) were measured using flame photometry (BWB, Berkshire, UK), and plasma Cl⁻ levels (mmol l⁻¹) were determined spectrophotometrically (DTX 880 Multimode Detector; Beckman Coulter) with a chloride assay kit (MAK023; Sigma-Aldrich).

Following collection of blood samples, isolated ventral skin samples (<1 cm²) were collected from the lower abdominal pelvic patch region and mounted in a self-contained, single-channel Ussing chamber (model U-9926; Warner Instruments, Hamden, CT, USA) with a single-channel epithelial voltage clamp amplifier (model EC-800; Warner Instruments), and connected to a PowerLab 4/35 interface (Mod n: PL3504, ADInstruments). Apical and basolateral surfaces of the skin were perfused with an oxygenated (95% O₂ and 5% CO₂) modified frog Ringer's solution based on Voyles et al. (2009) [in mmol l⁻¹: NaCl (112), KCl (2.5), d-glucose (10), Na₂HPO₄ (2), CaCl₂ (1), MgCl₂ (1), Hepes sodium salt (5), Hepes (5) at pH 7.3–7.4 with osmolality of 230±20 mosmol l⁻¹] maintained at 20.5±0.5°C. The Ringer's solution on the basolateral side represents the internal fluid composition, and on the apical side represents the cutaneous surface fluid composition (Larsen and Ramløv, 2013). Electrophysiological parameters were measured as follows: (1) transepithelial potential (TEP or V_T; mV) under open-circuit conditions (clamp current to 0 μA) and the resulting voltage in reference to the basolateral side, (2) active ion transport via clamping voltage to 0 mV and the instantaneous short-circuit current (I_SC) response as μA cm⁻², and (3) transepithelial area resistance (Ω cm²) by applying 3 s of 1 µA pulses across the epithelium every 60 s, or under voltage clamp conditions by applying 3 s of 1 mV at 60 s intervals (Wu et al., 2017). The inflections from the resulting change in I_SC and V_T during pulsing were used to calculate resistance using Ohm’s law. Sodium flux (mol cm⁻² s⁻¹) was calculated by dividing the I_SC (µA) by Faraday’s constant (96,000 C mol⁻¹). I_SC and V_T readings were recorded in LabChart (ADInstruments).

Amiloride hydrochloride hydrate (Sigma-Aldrich) was applied on the apical side of the chamber to calculate the amiloride-sensitive short circuit current (I_SC). Although amiloride is a non-specific Na⁺ channel inhibitor, ENaC is normally inhibited by very low (<1 µmol) concentrations of amiloride compared with other transporters such as Na⁺/H⁺ exchanger (>100 µmol), Na⁺/Ca²⁺ exchanger (1000 µmol) and NKA (>3000 µmol) (Kleyman and Cragoe, 1988). Amiloride was dissolved in dimethyl sulfoxide (DMSO) and added to the chamber to achieve final concentrations of 0.1, 0.5, 1, 2, 5, 10, 20 and 50 µmol l⁻¹. I_SC readings at each concentration were obtained 5 min post-application. The half-maximal inhibitory concentration (IC₅₀) of amiloride was calculated for each individual. Preliminary trials with DMSO only (vehicle control) did not significantly affect the overall voltage and current readings. After amiloride treatment, skin preparations were washed out with Ringer's solution until readings returned to starting values (or close to it), then 100 µmol l⁻¹ ouabain octahydrate (Sigma-Aldrich) dissolved in distilled H₂O was applied to the basolateral skin side chamber, and measurements were recorded until 15 min post-application of ouabain.

Ventral skin samples (<1 cm²) from the lower abdominal pelvic patch region were collected and placed into aqueous buffered zinc formalin fixative (Z-fix; Anatech, MI, USA) for 24 h, then transferred to 70% ethanol, and stored at 4°C. Tissues samples were then dehydrated through an ascending ethanol series, cleared in xylene and embedded in paraffin wax (Histoplast Paraffin; ThermoFisher Scientific). Tissue samples were then transversely sectioned into approximately 6 µm thick sections, (Leica RM2245; Leica Microsystems). A sub-sample of sections was stained with Mayer's hematoxylin and 1% eosin in 70% ethanol, and photographed with an iPhone 6s (Apple Inc.) and NIS-Elements software (v. 4.10, Nikon Instruments Inc.) under a bright-field illumination microscope (Nikon Eclipse E200 MV, Nikon Instruments Inc.).

Ventral skin samples (<1 cm²) from the lower abdominal pelvic patch region were collected and stored at −80°C. Frozen tissues were then homogenised in RIPA buffer [150 nmol l⁻¹ NaCl, 1% Triton X-100, 0.5% sodium deoxycholate, 0.1% SDS and 50 nmol l⁻¹ Tris-HCl with general protease inhibitors (Sigma-Aldrich General protease inhibitor cocktail P2714 with 1 mmol l⁻¹ PMSF)] using an IKA Ultra-Turrax. The protein content was quantified via a Qubit fluorometer (ThermoFisher Scientific). Protein (30 µg for NKA-α, 50 µg for ENaC-α) was dissolved in NuPAGE LDS sample buffer (Invitrogen) and denatured at 70°C for 10 min, then loaded in duplicate into Bolt 4-12% Bis-Tris Plus gels (ThermoFisher Scientific) and electrophoresed at 120 V for 40 min. Each membrane included a pre-stained protein ladder (ThermoFisher Scientific). Gels were subsequently transferred onto 0.45 µm Immun-Blot LF PVDF membranes (Bio-Rad) at 20 V for 60 min. Total protein staining was utilised to normalise protein loading (internal loading control) via the REVERT Total Protein Stain protocol (926-11010; LI-COR Biosciences). After total protein staining, membranes were then blocked in Odyssey blocking buffer (TBS) at pH 7.6 (LI-COR Biosciences) for 1 h at room temperature before incubation overnight at 4°C with their respective primary antibody [NKA-α5 primary antibody (1:1000 dilution) and ENaC-α primary antibody (1:1000 dilution; Anti-SCNN1A antibody; HPA012939, Sigma-Aldrich)] in blocking buffer with 0.1% Tween-20 (TBST). NKA α5 was deposited in the Developmental Studies Hybridoma Bank, University of Iowa, by D. M. Fambrough (DSHB Hybridoma Product a5). Membranes were then incubated in IRDye 800CW goat anti-mouse secondary antibody (1:20,000; 926-32210, LI-COR BioSciences) and IRDye 800CW donkey anti-rabbit secondary antibody (1:20,000; 926-32213, LI-COR BioSciences), respectively, for 1 h at room temperature in the dark. Membranes were dried and visualised with an Odyssey CLx imaging system (LI-COR BioSciences). Target proteins were normalised to loaded protein content using ImageStudio software (version 5.2; LI-COR BioSciences). Protein levels are expressed as abundance relative to the non-infected intermoult group.

Ventral skin samples (<1 cm²) from the lower abdominal pelvic patch region were collected, and stored in RNA-later (Ambion Inc.) at −20°C. The skin samples were then homogenised with stainless steel beads using a TissueLyser II (Qiagen). Total RNA was isolated (RNeasy Mini kit; Qiagen), and the resulting RNA yield was measured using a Qubit fluorometer (ThermoFisher Scientific). Genomic DNA contamination in RNA samples was removed, then the RNA was then reversed transcribed into cDNA, with appropriate no reverse transcriptase controls (QuantiTech Reverse Transcription Kit; Qiagen).

Transcripts for each gene of interest were located in an in-house L. caerulea larval and adult tissue transcriptome using homologous sequences from other amphibians as the reference query. Reference sequences were blasted against the transcriptome using the ‘blastn’ tool in Galaxy/GVL 4.0.1 (Afgan et al., 2016). Putative L. caerulea sequences were then compared against the entire National Centre for Biotechnology Information (NCBI) database (http://www.ncbi.nlm.nih.gov/) using ‘blastn’ with acceptance of the default parameters. qPCR primers against the target genes (Table S1) were designed using OligoPerfect™ Designer (Thermofisher Scientific) with acceptance of the default parameters. qPCR was performed using Power SYBR Green PCR Master Mix (ThermoFisher Scientific) using recommended cycling parameters. Each assay (in triplicate) included a no-template control and a no reverse transcriptase control. All PCR efficiencies were >90% and all the assays produced unique dissociation curves. Bio-Rad CXF Manager software (version 3.1, Bio-Rad) results were exported and each gene was quantified relative to the expression of the housekeeping gene, β-actin. Changes in expression levels were presented as fold-change relative to the uninfected intermoult group.

All analyses were performed in R 3.5.1 (https://www.r-project.org/). Data are either presented as means±s.e.m. or individual data points, and α was set at 0.05 for all statistical tests. Data that were not normally distributed were log-transformed prior to analysis. Percentage data were logit transformed prior to analysis. All models used a Gaussian error structure, and the confidence intervals for each model are presented in Tables S2–S6.

Blood plasma ion measurements (Na⁺, Cl⁻, K⁺) and electrophysiological measurements of the skin [TEP, amiloride-sensitive I_SC (50 µmol l⁻¹ concentration), skin resistance and percent ouabain inhibition of I_SC] between each group were initially analysed using linear models from the R default ‘stats’ package, with Bd load as an interactive effect and body mass as an additive effect. To test the effects of body mass and exposure period (exposed once or twice), models with body mass and exposure as additive effects were compared with models with no additive effects. The final model was chosen based on Akaike’s information criterion (AIC) with the ‘anova’ function. To test for interactions between treatments within groups while adjusting for Bd slope, a Holm-adjusted F-test was performed (package ‘phia’, function ‘testInteractions’; https://cran.r-project.org/web/packages/phia/index.html).

The effect of amiloride dose and Bd load on the I_SC between intermoult and sloughing animals was analysed using a quadratic model (package ‘nlme’, function ‘lme’; https://cran.r-project.org/web/packages/nlme/index.html), with log(I_SC) as the dependent variable, log(amiloride concentration) as a fixed effect, Bd load as an interactive effect and frog ID as a random effect to account for repeated measurements on the same individuals at different dosages. The IC₅₀ was calculated for each individual, and the differences between the uninfected and infected individuals for the intermoult and sloughing groups were analysed using a linear model with Bd load as an interactive effect.

Relative abundance of ENaC-α and NKA-α proteins in the ventral skin between each group was analysed using linear models, with Bd load as the interactive effect. To test the effects of body mass and exposure period (exposed once or twice), models with body mass and exposure as additive effects were compared with models with no additive effects. The final model was chosen based on the AIC with the ‘anova’ function. To test for factor interactions between treatments within groups and adjusting for Bd slope, a Holm-adjusted F-test was performed (package ‘phia’, function ‘testInteractions’).

The L. caerulea specific primer set sequence is available in Table S1. All datasets generated during and/or analysed during the study are available from the corresponding author on reasonable request.

Prevalence of infection with Bd was high: in the first set of exposures, 60.8% of exposed frogs developed an infection, and after re-exposure, 73.9% of originally exposed frogs had developed an infection. The three frogs that did not develop infection after re-exposure were excluded from further analysis.

The histopathology of the skin of infected L. caerulea after sloughing was the focus of this study, as infected intermoult (non-sloughing) animals have been detailed in past studies (Berger et al., 1999; Pessier et al., 1999; Nichols et al., 2001). In general, heavily infected frogs displayed cutaneous erythema, histological lesions and thinning of the skin (Fig. 1A,B). However, the skin of recently sloughed animals with low Bd zoospore loads (<10,000 ZE) was similar to that of uninfected animals (Fig. 1C,D). In heavily infected animals (>10,000 ZE) after sloughing, Bd sporangia were observed attached to the sloughed skin that had been fully removed (Fig. 1E,F,G), and also remained within the deeper layers of the epidermis (stratum granulosum; Fig. 1E,F).

Frogs infected with Bd (both during the intermoult and sloughing periods) in general showed a significant decrease in plasma sodium [Na⁺] and chloride [Cl⁻] levels with increasing Bd load (Table 1). Plasma potassium [K⁺] did not differ significantly between infected and uninfected animals (t₂₅=1, P=0.3). For intermoult animals only, there was a decrease in both plasma [Na⁺] and [Cl⁻] as Bd loads increased (Table 1). For sloughing animals, only plasma [Na⁺] levels remained significantly lower in infected frogs relative to uninfected frogs (F_1,25=11, P=0.003), while [Cl⁻] levels were not significantly different between infected and uninfected treatments (F_1,25=1.6, P=0.2; Table S2).

The TEP of the ventral skin across infection load was dependent on the sloughing phase (t₂₈=4.9, P<0.001; Table S3). As infection load increased in intermoult animals, the TEP decreased by 66.7% (−32.5±5.2 to −10.8±1.7 mV; F_1,28=16.9, P<0.001) compared with uninfected animals. However, in sloughing animals, the TEP increased by 58% (−8.8±0.85 to −21±5.5 mV) as infection load increased (F_1,28=8.9, P=0.006; Fig. 2A).

Similarly, skin resistance decreased with increasing Bd load (Fig. 2B; Table S3). After sloughing (both infected and uninfected animals), the skin resistance was the lowest (301±79 Ω cm²). However, there was no significant difference in skin resistance between infected and uninfected animals after sloughing (Fig. 2B). There was also a borderline significant difference between the intermoult and sloughing animals (F_1,28=5.6, P=0.05), with the sloughing animals showing lower skin resistance (321±51 Ω cm²) compared with the intermoult animals (478±82 Ω cm²; Fig. 2B).

Changes in active sodium transport, represented by amiloride-sensitive instantaneous short-circuit current (I_SC) across infection load, was dependent on the sloughing phase (Table S3). For intermoult animals, as infection load increased, the I_SC decreased by approximately 40% (59.2±9.8 to 35.6±4.5 µA cm⁻²; F_1,27=7.3, P=0.01). When converted to sodium flux, the rate of sodium transport decreased from 6.1×10⁻⁴±1×10⁻⁴ mol cm⁻² s⁻¹ in uninfected animals to 3.7×10⁻⁴±4.6×10⁻⁵ mol cm⁻² s⁻¹ in infected animals. Recently sloughed animals increased I_SC by 66% (28.8±4.2 to 84.7±14.3 µA cm⁻²; F_1,27=27, P<0.001; Fig. 2C) as infection load increased. This represents an increase in the rate of sodium transport during sloughing from 3×10⁻⁴±4.4×10⁻⁴ mol cm⁻² s⁻¹ in uninfected animals to 8.8×10⁻⁴±1.5×10⁻⁴ mol cm⁻² s⁻¹ in infected animals.

For ouabain inhibition of I_SC, there was an overall significant difference between groups, where post-sloughing animals had greater inhibition from 100 µmol l⁻¹ ouabain compared with intermoult animals (t₁₅=2.3, P=0.04; Fig. 2D). Infection did not have a significant effect on the I_SC inhibition (Table S4).

For amiloride dosage experiment, there was a significant decrease in I_SC as amiloride dose increased in both intermoult and sloughing groups, independent of infection load (Table S4). In the intermoult group, amiloride concentration on the I_SC was dependent on the infection load, where at a high dosage of amiloride the I_SC slope decreases to a plateau with increasing infection load (Fig. 2E). This indicates the isolated ventral skin during the intermoult phase of infected animals was more sensitive to amiloride dosage, requiring 0.7±0.14 µmol l⁻¹ amiloride to inhibit 50% of the I_SC, compared with uninfected frogs (IC₅₀=2.35±0.6 µmol l⁻¹; t₁₂=2.6, P=0.02).

For animals after sloughing, there was a significant effect of Bd infection alone (t₁₁=3.5, P=0.004), with a decrease in I_SC as infection load increased. No interactions between infection load and amiloride concentration were observed (Table S4), suggesting no difference in the sensitivity to amiloride in the newly exposed skin between uninfected (IC₅₀=1.4±0.26 µmol l⁻¹) and infected frogs (IC₅₀=0.84±0.2 µmol l⁻¹; t₁₂=0.7, P=0.5; Fig. 2F).

There was a decrease in ENaC-α subunit (SCNN1A) protein abundance with increasing infection load (Table S5); however, there was no significant difference in SCNN1A abundance between the intermoult and sloughing groups (t₂₃=0.9, P=0.3). Within the intermoult group, as infection load increased, the relative abundance of SCNN1A protein decreased (F_1,23=6.7, P=0.03; Fig. 3A). For sloughing animals, there was no significant difference in SCNN1A protein abundance as infection load increased (F_1,23=1, P=0.3).

The abundance of the NKA-α subunit (ATP1A1) was significantly greater in infected animals compared with uninfected animals in the sloughing group (F_1,23=0.65, P<0.001; Fig. 3B), whereas there was no difference between infected and uninfected animals in the intermoult group (F_1,23=0.65, P=0.42).

Within the ENaC family, there were significant differences in α- and β-subunit mRNA expression between infected and uninfected frogs (Fig. 4). For ENaC-α (SCNN1A) there was a significant increase in expression in the infected group compared with the uninfected group (t₂₂=3.8, P=0.001), whereas for ENaC-β (SCNN1B) there was a significant decrease in expression (t₂₂=2.7, P=0.01). No effect of sloughing phase was observed for either gene (Fig. 4A; Table S6). There was no effect of Bd infection or sloughing phase on the expression of the ENaC-γ (SCNN1G) subunit (Fig. 4A).

Within the NKA family, there were significant increases in the expression of both α1 (ATP1A1) and β1 (ATP1B1) subunit mRNA in the infected group relative to the uninfected group (Fig. 4B; Table S6). NKA-β3 (ATP1B3) subunit expression did not differ between infected and uninfected animals (t₂₂=2, P=0.06). However, there was a significant effect of sloughing phase on ATP1B3 expression (t₂₂=3.2, P=0.003), with expression increasing in the sloughing group relative to the intermoult group (Fig. 4B).

With the recent discoveries that sloughing can regulate cutaneous Bd loads (Ohmer et al., 2017) and that increasing Bd loads contribute to the severity of disease symptoms in infected animals (Wu et al., 2018), it is important to understand the mechanistic basis for these relationships. The present study confirms that reduction in ENaC-associated cutaneous sodium transport contributes to the loss of ionic homeostasis in frogs with chytridiomycosis. We showed that elevated cutaneous sodium efflux in Bd-infected frogs is due in part to reduced Na⁺ uptake capacity, attributable to a reduced abundance of ENaC transporters. Although ENaC mRNA expression was elevated in non-sloughing frogs with chytridiomycosis, the abundance of ENaC protein in the skin was lower than in uninfected frogs, suggesting that either protein synthesis was being inhibited or that newly synthesised proteins were being lost. Interestingly, sloughing stimulated a transient upregulation of Na⁺ uptake in both infected and uninfected animals, and the newly exposed skin of infected frogs had a similar abundance of ENaC protein compared with uninfected frogs. This suggests the new skin layer initially possesses a full complement of functional Na⁺-transport machinery that facilitates the temporary restoration of normal physiological function. However, this restoration period is brief; it appears that ENaC proteins in the skin are destroyed by exposure to Bd during the intermoult period, contributing to the ongoing loss of ionic homeostasis.

Consistent with previous findings by Voyles et al. (2009), infected L. caerulea during the intermoult period had lower transcutaneous electrophysiological parameters (TEP, amiloride-sensitive I_SC, skin resistance) and lower levels of plasma Na⁺ and Cl⁻ than uninfected frogs. Small changes in the electrical properties of an epithelia can have large consequences for the maintenance of physiological homeostasis (Wood and Grosell, 2008); for example, a −10 mV change in the TEP is correlated with a reduction in plasma [Na⁺] by 33.5% in killifish (Wood and Grosell, 2015). The combination of reduced TEP, reduced active uptake and increased skin permeability could explain the low plasma Na⁺ (hyponatremia) and Cl⁻ (hypochloremia) levels seen in infected L. caerulea. The reduced TEP is likely a result of the reduction in both skin resistance and active transcutaneous Na⁺ uptake (amiloride-sensitive I_SC) across the skin of non-sloughing infected frogs. Several lines of evidence suggest that the secretion of proteolytic enzymes and toxins by Bd (Rosenblum et al., 2008; Symonds et al., 2008; Brutyn et al., 2012) breaks down proteins and intercellular junctions, leading to increased skin permeability (Voyles et al., 2009). The loss of skin resistance and reduced uptake capacity is consistent with the idea that Bd may directly damage local proteins in the skin. Further evidence for this hypothesis is that Bd had no effect on the activity or abundance of NKA in the skin. In frog skin, the multiple layers of the epithelium function as an interconnected syncytium (Heatwole et al., 1994); active ion uptake occurs in the more apical skin layers, while the deeper skin layers contain the basolateral membrane-associated proteins (i.e. NKA) responsible for the generation of the electrochemical gradient that facilitates the apical transporter activities (Nielsen, 1979). The impairment of active Na⁺ uptake is consistent with the hypothesis that Bd infection in the apical layers of the skin (stratum corneum and upper granulosum) (Greenspan et al., 2012) directly affects the functional capacity of ENaC and the paracellular junctions, whereas ion transporters in the deeper layers of the epidermis (i.e. NKA) are largely unaffected by infection.

Bd infection appears to alter the functional capacity of Na⁺ transport pathway by directly destroying these channels in the skin. Bd increased the sensitivity of the cutaneous active ion transport pathway to amiloride in intermoult frogs, reducing the IC₅₀ by 70% relative to the uninfected group. A reduction in amiloride-sensitive Na⁺ transport indicates that there are fewer functional ENaC proteins in the skin of infected animals and so saturation of pumps is achieved at lower drug concentrations. This is consistent with our data showing that there was less ENaC-α protein in the skin of infected (intermoult) frogs compared with healthy frogs. ENaC exists as an obligate heterotrimer with α, β and γ subunits (or β, γ, δ subunits in some organisms; Hanukoglu and Hanukoglu, 2016), and although we did not quantify the protein-level abundance of all ENaC subunits, it is likely that the reduction in ENaC-α reflects similar changes in the abundance of the functional heterotrimer. The skin of infected animals appears to try to compensate for the loss of functional proteins by increasing the expression of the ENaC-α gene. However, the gene expression of other ENaC subunits did not follow the same pattern: Bd infection reduced the expression of ENaC-β and had no effect on the expression of ENaC-γ. Whether Bd differentially affects the abundance of ENaC subunit proteins is unknown. ENaC subunit expression is under tight control by hormones and extracellular factors (Butterworth et al., 2008), including stress hormones, so fluctuations in circulating hormone levels as a consequence of natural cycles or disease-related processes may contribute to the differential regulation of ENaC subunit expression. It is also possible that only ENaC-α protein is affected by Bd and, because ENaC-α is a requirement for ENaC channel activity, ENaC-β gene expression may be downregulated to prevent an oversupply of the β-subunit in the absence of the corresponding α-subunit.

In infected animals, Na⁺ transport into the skin increased immediately after sloughing. The stark difference in Na⁺ uptake rates between recently sloughed and intermoult animals, particularly in infected frogs, likely reflects the exposure of the new skin layer. Sloughing removes the most degraded outermost layer of skin, effectively ‘resetting’ the skin surface (Larsen, 1976). However, until the newly exposed stratum corneum layer becomes fully keratinised, the skin remains relatively permeable to ions (i.e. it has low resistance to ion loss). Therefore, the increase in I_SC may serve to offset the transient increase in skin permeability that accompanies sloughing. Indeed, we and others have observed a similar change in I_SC across the skin immediately after sloughing in other amphibian species (Larsen, 1970; Nielsen and Tomilson, 1970; Larsen, 1971; Wu et al., 2017). The increase in I_SC post-sloughing in cane toads occurred commensurate with an increase in the abundance of ENaC-α and NKA proteins in the epithelium (Wu et al., 2017). This is consistent with the idea that the increase in I_SC post-sloughing is the result of more functional transporters present in the new skin, although isotope binding studies (e.g. Cramp et al., 2009) would be needed to quantify this.

Importantly, for frogs infected with Bd, the increase in transporter numbers in the skin post-sloughing was sufficient to restore active Na⁺ uptake (I_SC), albeit transiently. This finding supports the idea that Bd directly damages ion-transport proteins in the apical skin layers (Campbell et al., 2012), and that sloughing restores skin function by both exposing relatively undamaged cells and reducing the abundance of Bd on the skin, which may reduce fungal toxins or secretions. However, the increased expression of ENaC-α, NKA-α and NKA-β mRNA in the tissues of recently sloughed frogs indicates that Bd infection does exert some influence over ion transporter systems, even in the more basal skin layers. Although tissues were collected as soon after sloughing as possible, there was often a gap of up to 1 h between sloughing and when tissues were collected, during which time the newly exposed skin would have been in direct contact with Bd in the surrounding environment. So, although increased mRNA expression in newly exposed skin of Bd-infected frogs might be a pre-sloughing compensatory response to elevated rates of protein degradation in the skin, contact between the newly exposed skin and Bd zoospores or toxins in the environment may have also stimulated ion transporter mRNA synthesis.

In the present study, infected frogs sloughing showed an increase in skin permeability. Increased permeability following sloughing is prolonged compared with uninfected frogs, which could be due to the combination of leaky skin from the act of sloughing (Wu et al., 2017), structural disruption from infection (Nichols et al., 2001; Berger et al., 2005) and reduction in the expression of skin integrity genes from Bd infection (Rosenblum et al., 2012; Ellison et al., 2015). Future electrophysiological studies could determine epithelial tight junction permeability by replacing Cl⁻ Ringer's solution with sulphate Ringer's solution in Ussing chamber set-ups. To compensate for the high leakage demonstrated in this study, we found that NKA activity and abundance increased during sloughing. An increase in NKA activity is associated with maintaining osmotic balance when animals are subjected to osmotic stressors (Choi and An, 2008; Cramp et al., 2010). However, because NKA actively exchanges ions across cellular membranes against their concentration gradients via ATP hydrolysis (Lingrel and Kuntzweiler, 1994), there may be an increase in cellular energy expenditure to uptake Na⁺ into the cells (Harvey and Kernan, 1984). Susceptible amphibian species infected with Bd have greater demand on cellular metabolic processes compared with non-susceptible species (Poorten and Rosenblum, 2016). Thus, the combination of increased skin permeability, cellular energy expenditure, ionic disruption and rate of sloughing may have additional cumulative consequences for maintaining electrolyte homeostasis in highly infected frogs.

Our results reveal a complex interaction between sloughing and chytridiomycosis. The cumulative effect of increased skin permeability during non-sloughing periods and increased sloughing frequency effectively counteracted any benefit that might have be gained from the temporary, post-sloughing increase in electrolyte uptake capacity. Given that the magnitude of effects of Bd on skin function and sloughing frequency are proportional to Bd loads, species that fail to regulate Bd loads effectively are likely to suffer greater physiological disruption and more rapid progression of the disease than those species that can regulate their fungal loads (Ohmer et al., 2017). This work demonstrates the mechanisms underlying cutaneous disruption during Bd infection in susceptible amphibian species and highlights the intrinsic role of sloughing on this process. Importantly, this study explains why sloughing can be detrimental for susceptible species that develop high infection loads, and may accelerate disease progression.

We would like to acknowledge all volunteers who have assisted with animal care, maintenance and video surveillance, S. Bloomberg (University of Queensland) for advice on statistical analysis, and E. P. Symonds for isolating strain EPS4. Special thanks to J. Gauberg (York University), E. Watson and A. T. Khalid for assisting with experiments.