Main

The past two decades have seen the emergence of novel fungal diseases that globally affect biodiversity, leading to the potential extinction of animal and plant species3,4,5,6,7,8. When fungal pathogens are vectored into naive ecosystems, firm pathogen establishment and extensive host population decline typically precede elucidation of the disease ecology, which is required for the development of threat abatement plans3,9. The chytrid fungus Batrachochytrium salamandrivorans is a prime example of an emerging infectious disease that has recently become a threat in Europe, where it causes massive decline of salamander populations and poses an unprecedented threat to western Palaearctic amphibian diversity1,4,10. Here we unravel the fundamental mechanisms of amphibian extirpation mediated by the recent arrival of B. salamandrivorans. Immediately after the discovery of the first signs of disease (April, 2014) in a population of fire salamanders in Robertville, Belgium, 57 km from the B. salamandrivorans index site in the Netherlands4, we began to continuously monitor infection, disease and host population dynamics for two years. Our study demonstrates how the combined characteristics of host susceptibility, pathogen virulence and environmental persistence create a ‘perfect storm’ with high probability of extirpation after pathogen arrival in a susceptible host population.

Our monitoring revealed that introduction of B. salamandrivorans leads to a fast host population collapse, without any sign of recovery, owing to sustained and disproportionate mortality of adults, which leads to a demographic shift in the population (Fig. 1a, b). Across ten-day intervals, we found a probability of infection of 0.33 (95% credible interval (CRI) = 0.169–0.512). Infection resulted in a sixfold difference in survival rate (mean survival in infected versus non-infected animals (0.13 ± 0.11 s.d., 95% CRI = 0.004–0.403) versus (0.84 ± 0.10, 95% CRI = 0.63–0.99)).

Figure 1: B. salamandrivorans leads to fire salamander population extirpation.
figure 1

a, b, An estimated 90% decline within 6 months (a) coincides with a decrease of the average length and a demographic shift (b). CI, confidence interval. c, Contact between an infected (on top; skin lesion shown in insert) salamander with an uninfected one during courtship.

PowerPoint slide

Full size image

In a series of infection trials, we studied the host–pathogen interaction underpinning the susceptibility of fire salamanders to infection with B. salamandrivorans. We demonstrate that the outcome of the disease in this species is dose- and temperature-independent and that infected animals do not mount any protective immune response. Experimental inoculation of fire salamanders with a high or low dose of four different B. salamandrivorans isolates resulted in lethal disease in all animals, despite a slower build-up of infection load at the low dose (Fig. 2a, b). When comparing infection dynamics of B. salamandrivorans at the fungus’ optimal temperature (15 °C)4 with those at 4 °C, all animals again developed lethal infection, with slower build-up of infection at 4 °C (Fig. 2c, d), indicating that the pathogen is able to infect and kill amphibians over a broad temperature range. In a final experiment, we assessed build-up of protection in salamanders after five cycles of exposure treatment. Contrary to the theory that non-lethal exposure to the pathogen could provide opportunities to mount a protective immune response11, this experiment confirmed that resistance against infection did not increase (Fig. 2e, f). The inability of salamanders to mount resistance against B. salamandrivorans infection largely excludes vaccination as a mitigation measure for susceptible salamander species and could preclude build-up of population immunity. Indeed, the few salamanders that were still present at the outbreak site after two years were still highly susceptible to infection with the local B. salamandrivorans isolate, showing a 100% mortality rate after exposure.

Figure 2: Effect of different variables on infection dynamics of B. salamandrivorans in fire salamanders.
figure 2

a, b, A lower dose results in slower build-up of infection and delayed mortality. c, d, At lower temperatures, similar infection intensities are reached more slowly. e, f, Previous exposure to B. salamandrivorans does not protect against re-infection. Bars (a, c, e) indicate the proportion of alive and infected (coloured), alive and uninfected (light grey) and dead (dark grey) individuals; curves indicate the estimated treatment-specific probability of infection (shading: 95% CRI). Box plots (b, d, f) indicate infection loads in genomic equivalents per swab (bars indicate 2.5th and 97.5th percentiles); curves indicate the estimated treatment-specific load (shading: 95% CRI).

PowerPoint slide

Full size image

Given the continuous high mortality and high transmission rate, persistence of any susceptible European host population after infection would only be likely when compensated by elevated recruitment12. However, we found that B. salamandrivorans disproportionately infects and kills sexually mature animals (Fig. 1b), quickly resulting in a demographic shift that massively decreases the recruitment potential of the population that would be necessary to compensate for adult mortality13. Increased infection probability and subsequent mortality of adult fire salamanders compared to juveniles can be explained by intimate contact with infected adult conspecifics during territorial displays and reproduction (Fig. 1c), and back-and-forth migration of females to the streams where they give birth to aquatic larvae. Juveniles show less surface activity and interactions with conspecifics14, with reduced probability of becoming infected.

Our analyses indicate that the rapid population decline can be largely associated with two fungal pathogen determinants: sustained fungal virulence and the presence of an environmentally resistant encysted spore. Sustained virulence of the fungus in its novel susceptible host species was demonstrated by the persistence of highly virulent B. salamandrivorans two years after the outbreak in our study site, despite almost complete depletion of the host population. Indeed, isolates cultured from infected animals at the end of our two-year monitoring period were equally capable of killing 100% of the experimentally inoculated salamanders as an initial isolate. The sustained virulence is assisted by the fact that, in addition to motile spores as in its closest sister species B. dendrobatidis, B. salamandrivorans produces a second type of infectious encysted spores both in vitro and in vivo in salamander skin, with a distinct fungal infection, dissemination and persistence strategy (Fig. 3a). Whereas zoospores actively swim to their host15, encysted spores float at the water–air interface (Supplementary Video 1) and are capable of quickly adhering to salamander skin and to scales of the feet of waterfowl (Fig. 3b). This passive adherence to inert matrices may promote fungal spread over large spatial distances. Encysted spores survived and remained infective for fire salamanders for at least 31 days in filtered pond water (Fig. 3c) and were more resistant than zoospores to predation by zooplankton (Fig. 3d), highlighting their potential to persist in an aquatic environment.

Figure 3: B. salamandrivorans encysted spores avoid predation and infect fire salamanders.
figure 3

a, Transmission electron microscopic picture of a sporangium containing encysted spores and free encysted spores in the superficial epidermis and a zoospore (insert). *, lipid globule; arrow, cell wall; F, flagellum; M, mitochondrion; N, nucleus; R, ribosomal mass. b, Adhesion of encysted spores to salamander skin and scales from goose feet (control: skin dipped in culture supernatant; insert: scanning electron microscopic image of attachment to salamander skin). c, B. salamandrivorans infection loads in salamanders exposed to encysted spores that were either freshly collected (t0, with n = 6) or were incubated in pond water for 14 or 31 days (t1 and t2, with n = 4 and n = 2). d, Estimated survival of motile and encysted spores after exposure to aquatic micropredators. Bars indicate the 2.5th and 97.5th percentiles in b, c and 95% CRI in d. B. salamandrivorans loads (b, c) are expressed in genomic equivalents per swab.

PowerPoint slide

Full size image

Long-term persistence of B. salamandrivorans is further promoted by its presence on less-susceptible amphibian pathogen reservoirs, as we demonstrated by experimental infection of anuran and urodelan hosts. A proportion of the four used B. salamandrivorans isolates was capable of infecting anuran hosts (midwife toads, Alytes obstetricans) at low intensities for several weeks after experimental inoculation. Whereas the toads showed no sign of disease, their colonization with the pathogen was sufficient to transmit B. salamandrivorans to susceptible salamanders (Fig. 4a, b, Extended Data Table 1). In urodelans, experimental infection of Alpine newts (Ichthyosaura alpestris), a species that co-occurs syntopically with fire salamanders, showed a dose-dependent disease course. Whereas infection with a high dose resulted in disease and death after an average of three weeks, exposure to a low dose resulted in significant B. salamandrivorans shedding for several months with eventual fungal clearing and clinical cure (Fig. 4c, d). However, previous infection was shown to provide the newts with no protection against re-infection and mortality (Fig. 4e). These newts thus meet the criteria for a pathogen reservoir. Indirect transmission would also favour pathogen maintenance in host populations; therefore, in a final infection experiment we demonstrated the potential for pathogen transmission via contaminated forest soil. Infected salamanders were shown to contaminate the forest soil, in which the fungal DNA could be detected even after 200 days. Actual transmission through contaminated forest soil was demonstrated up to 48 h after the soil had been in contact with an infected animal (Extended Data Figs 1, 2, Extended Data Tables 2, 3). Altogether, the presence of a resistant spore with the ability to persist environmentally and to transmit through contaminated water and soil, combined with the occurrence of long-term-infected and pathogen-shedding amphibian hosts, creates the potential for extensive environmental reservoirs and hampers any effort to eradicate B. salamandrivorans from an infected ecosystem.

Figure 4: Anuran and urodelan reservoirs promote B. salamandrivorans sustenance.
figure 4

a, b, Probability of infection and infection load in midwife toads. a, Bars indicate the proportion of infected individuals (no mortality observed) and the curve indicates the estimated probability of infection. b, Box plots and curve indicate, respectively, observed and estimated infection load of B. salamandrivorans-positive individuals; salamander icons indicate infection in fire salamanders that had been co-housed with the toads from the second week post exposure onwards. c, d, Probability of infection and infection load in Alpine newts after exposure to 10,000 (orange), 1,000 (green) or 100 (blue) spores. c, Bars indicate the proportion of alive and infected (coloured), alive and uninfected (light grey) and dead individuals (dark grey); curves indicate the probability of infection. d, Box plots and curves indicate, respectively, observed and estimated infection load of B. salamandrivorans-positive individuals. e, Infection loads for naive or previously infected Alpine newts; box plots and curves indicate, respectively, observed and estimated infection load of B. salamandrivorans-positive individuals. B. salamandrivorans loads (b, d, e) are expressed in genomic equivalents per swab. In all plots, shading around curves indicates 95% CRI; box plot bars indicate 2.5th and 97.5th percentiles.

PowerPoint slide

Full size image

Our study reveals that the multifaceted ecology of this expanding fungal disease is likely to result in fast extirpation of highly susceptible salamanders, with no available options to halt the spread or to mitigate the disease in situ. Although several potential measures to counteract the effect of chytrid fungi on amphibian communities have been proposed9, ex situ conservation programmes are currently the only intervention available that will effectively avert loss of susceptible urodelan populations upon B. salamandrivorans arrival. Given the continuous range expansion of the disease and the speed of its effects, the development of a pan-European early warning system to monitor the fungal invasive front and the enforcement of emergency action plans that allow fast implementation of ex situ conservation in acutely threatened urodelan species are urgently needed. A thorough understanding of the host, pathogen and environment determinants underpinning susceptibility to B. salamandrivorans may yield the tools required for risk analysis of the actual threat of the fungal disease to western Palaearctic urodelans, which will guide prioritization of conservation efforts. Indeed, although most western Palaearctic urodelan taxa were shown to be susceptible to B. salamandrivorans in laboratory trials1, we demonstrate marked interspecific differences. Fire salamanders represent a hyper-susceptible species in which B. salamandrivorans causes acute, dose-independent mortality and population extirpation. The rapid and consistent mortality at least in species of the genera Salamandra, Euproctus, Neurergus, Pleurodeles and in Lissotriton italicus after exposure to B. salamandrivorans1,16 suggests a similar risk of disease driven extirpation for many other species of western Palaearctic urodelans. As eradication of B. salamandrivorans from the European continent is highly unlikely, long-term sustainable mitigation should aim at host–pathogen co-existence, which implies the development of intervention strategies that permanently increase resistance of susceptible species against B. salamandrivorans9. For regions that are currently considered free of B. salamandrivorans, such as the Americas, prevention of introduction in naive environments should be considered the sole effective control measure available. This will require knowledge of all introduction pathways that, besides amphibians in trade1,17,18, may also include non-amphibian sources carrying resistant spores.

Methods

Batrachochytrium salamandrivorans isolates and culture conditions

Four B. salamandrivorans isolates were isolated from wild fire salamander populations that are declining owing to a B. salamandrivorans outbreak in the Netherlands (AMFP13/1)4 and Belgium (AMFP14/1 (Robertville, 2014); AMFP 15/3 (Robertville, 2015) and AMFP14/2 (Luik)2. One isolate originated from a captive population suffering from a B. salamandrivorans disease outbreak in Germany (AMFP15/1)16. All isolates were grown in a 1.6% tryptone, 0.4% gelatinehydrolysate and 0.2% lactose monohydrate liquid medium at 15 °C. Motile or encysted spores were collected in distilled water after 5–7 days of growth. Zoospores were obtained by washing the culture flasks with filtered pond water (0.2-μm filter). Floating encysted spores were collected from the water surface using a 10-μl inoculation loop. Purity of the spore suspension was assessed using inverted microscopy. Encysted spore suspensions were only used in experiments when motile spores were absent. No statistical methods were used to predetermine sample size.

DNA extraction and B. salamandrivorans qPCR

DNA extraction of skin swabs and B. salamandrivorans qPCR were performed as described in refs 19, 20. Animals were considered positive for B.-salamandrivorans infection when the following conditions were fulfilled: (1) the qPCR sample quantity was above the detection limit of 0.1 genomic equivalent (GE)/qPCR reaction for both replicates, (2) the mean starting quantity value of each sample was higher than the standard deviation of its starting quantity, and (3) the amplification curve from each replicate was logarithmic.

Infection trials

For all experimental replicates, all fungal cultures were grown independently. All animals were housed individually in terraria at the fungus thermal preference of 15 °C4 (unless otherwise stated) on moist tissue with access to a hiding place. All animals (males/females) were captive bred, clinically healthy and free of B. dendrobatidis, B. salamandrivorans and Ranavirus as assessed by sampling the skin using cotton-tipped swabs and subsequently performing qPCR19,20 or PCR21. Infection experiments were carried out as described in ref. 1. Individuals were randomly assigned to treatments. All animals were clinically inspected daily. Skin sampling was done weekly and the swabs were analysed for the presence of B. salamandrivorans using qPCR described in refs 19, 20. Investigators were blinded to allocation during experiments and outcome assessment.

Hygiene and biosafety protocols

The animal experiments were performed under strict BSL2 conditions. During the fieldwork, each individual was handled with a new pair of nitrile gloves. At the end of each field visit, boots and other equipment which came into contact with the environment were disinfected with a 1% Virkon solution for at least 5 min.

Ethics statement

The animal experiments were performed with the approval of the ethical committee of the Faculty of Veterinary Medicine (Ghent University EC2013/10; EC2014/170; EC2015/29; EC2015/42; EC2016/87). The capture, handling, sampling and transport of wild salamanders and access to the sampling site were permitted by the Wallonian Department of Nature and Forests (Département de la Nature et des Forêsts) (reference 2014/RS/n°23).

Infection and disease dynamics during a B. salamandrivorans outbreak in fire salamanders: a 2-year follow-up study

In May 2014, a B. salamandrivorans outbreak with mass mortality was identified in a population of fire salamanders (Salamandra salamandra terrestris) in the forest of Robertville (50° 27′ 10″ N, 06° 06′ 10″ E), Belgium. Over the course of 2 years, the population was monitored during the activity periods of the salamanders: (1) May to October 2014 and (2) April to September 2015. Over a fixed 475-m-long transect, salamanders were detected using a visual encounter survey and sampled by collecting skin swabs. To investigate whether B. salamandrivorans can be detected in terrestrial environments in the wild, soil samples were taken in the close vicinity of animals with chytridiomycosis (Extended Data Fig. 1). The animals that were found during the second year, one on the study transect and 11 outside the transect, were brought to our quarantine facilities to test their susceptibility to chytridiomycosis. None of these tested positive for B. salamandrivorans when sampled at location. However, the animals developed severe to lethal signs of chytridiomycosis within 3 weeks after exposure to the fungus.

In total, 24 visits resulted in 197 captures of fire salamanders. Individual salamanders were identified by their conspicuous yellow marks; from each captured individual, dorsal and lateral photographs were taken. On the basis of the capture-mark-recapture data from the first two visits between which there were confirmed recaptures, the population of fire salamanders on the study-transect was estimated to consist, at the beginning of the study, of 239 (95% confidence interval, 112–459) individuals, using the Lincoln–Petersen index22.

Host population dynamics analysis was performed on the basis of a multistate capture-mark-recapture model23. On the basis of the qPCR results from the skin swabs, each individual was classified as either infected or non-infected. We constructed a Bayesian multistate capture–recapture model24 to estimate survival probabilities of non-infected and infected individuals as well as transition probabilities between these two states. The transition probability from a non-infected to infected state is the infection probability, the transition probability from infected to non-infected is the recovery probability. Because the population was declining rapidly and no recovered individuals were observed, all parameters were set constant through time, reencounter probability of infected individuals was assumed to be equal to reencounter probability of uninfected salamanders. The interval between two sampling periods was highly influenced by weather conditions and activity periods of salamanders. To adjust for the unequal time intervals, dummy sampling occasions were added to the dataset to equalize the intervals between two sampling periods to approximately 10-days. The reencounter probability was set to 0 at the dummy occasions. The estimated survival and transition probabilities refer therefore to 10 day intervals. Analyses were performed for the data collected during the first period only (sampling occasions 1–14). The multistate model was fitted in JAGS25 through package jagsUI in R24. Vague priors were chosen for all the parameters (uniform between 0 and 1) and convergence was inferred by values <1.1.

Drivers of B. salamandrivorans dynamics in susceptible fire salamanders

Dose dependency of B. salamandrivorans infection and disease dynamics. Four B. salamandrivorans isolates (AMFP13/1, AMFP14/1, AMFP14/2, and AMFP15/1) were used to expose 40 juvenile fire salamanders. Animals were infected with one isolate. For each isolate two doses were used: either 104 spores (high dose) or 100 spores (low dose). For each dose-by-isolate combination, 5 animals were inoculated per isolate and each animal was housed individually. The course of disease was followed up by daily clinical inspection and weekly sampling of the animals for 8 weeks (Fig. 2a, b).

Temperature dependency of B. salamandrivorans infection dynamics. In this experiment, the infectivity of B. salamandrivorans after incubation in water for 4 weeks was evaluated. A suspension of 1.3 104 GE B. salamandrivorans per ml environmental water was incubated for 4 weeks at both 4 °C and 15 °C. After 4 weeks, the concentration of B. salamandrivorans was 2.6 × 104 GE per ml water (4 °C) and 1.3 × 104 GE per ml water (15 °C). This water was used to inoculate 10 salamanders as described in ref. 1. Water containing B. salamandrivorans incubated at 4 °C was used to infect salamanders which were later housed individually at 4 °C. Water containing B. salamandrivorans incubated at 15 °C was used to infect salamanders which were later housed individually at 15 °C. The course of disease was followed up by daily clinical inspection and weekly sampling of the animals for 10 weeks (Fig. 2c, d).

Effect of previous infection on infection and disease dynamics of B. salamandrivorans. Five fire salamanders were infected with 103 spores of AMFP13/1. After increase of the GE load in two subsequent swabs (weekly sampling), animals were treated at 25 °C for 10 days as described in ref. 26. One month after finishing the treatment, the animals were re-infected. This exposure–treatment cycle was repeated five times. For the subsequent challenge experiment, infection dynamics of B. salamandrivorans in these five pre-exposed salamanders were compared their initial infection dynamics (Fig. 2e, f).

B. salamandrivorans produces infectious, encysted and environmentally protected spores.

Production of encysted spores by B. salamandrivorans. Using microscopy and transmission electron microscopy, we identified two types of spores in B. salamandrivorans: a motile (zoo)spore and a non-motile, encysted spore with a cell wall. The ultrastructure of the different spores was investigated in in vitro cultures of B. salamandrivorans and in skin samples of infected fire salamanders. Transmission and scanning electron microscopy was performed as described in ref. 3 (Fig. 3a, insert Fig. 3b).

Brief contact with floating, encysted spores results in adherence of B. salamandrivorans to salamander skin and scales from goose feet. Encysted spores were collected and inoculated in filtered (0.2-μm filter) environmental water at a concentration of 106 spores per ml water. A salamander toe and scale from goose feet were dipped in the suspension for 1 s. Controls to quantify B. salamandrivorans DNA contamination consisted of toes, dipped in filtered (0.5-μm filter) culture supernatant. B. salamandrivorans load in all samples was determined using qPCR19,20. Two independent experiments were performed in triplicate. Results are expressed as mean number of genomic equivalents per mm2 + standard deviation of the respective tissue (Fig. 3b).

Survival in the aquatic environment and infectivity of encysted spores

Encysted spores were collected and inoculated in filtered environmental water at a concentration of 108 spores per ml water. At time points 0 (immediately after collection), 1 (15 days after collection) and 2 (31 days after collection) fire salamanders (n = 6 at t0, n = 4 at t1 and n = 2 at t2) were inoculated by dropping 100 μl of this suspension on the salamanders’ dorsum. The course of disease was followed up by daily clinical inspection and weekly sampling of the animals for 10 weeks. Infection loads were determined by quantifying B. salamandrivorans DNA in skin swabs using qPCR19,20. Results are presented as average infection loads ± standard deviation (Fig. 3c).

Predation of motile and encysted spores

Predation of motile and encysted spores by micropredators present in pond water was tested as described in ref. 27. Briefly, 106 motile or encysted spores were incubated with 1 ml of pond water containing 456 zooplanktonic organisms per ml, for four hours in 24-well plates at 15 °C. The water contained copepods (30 per ml), ciliates (paramecium (338 per ml) and peritrich ciliates (18 per ml)), rotifers (16 per ml), ostracods (35 per ml), heliozoans (1 per ml) and water fleas (18 per ml) as determined by counting the total content in 1 ml of pond water using light microscopy. For comparison, zoospores and encysted spores were incubated in pond water that was filtered using a 5-μm filter. After 4 h incubation the number of remaining spores was counted using a Bürker counting chamber. Removal of spores from the aquatic environment was quantified as proxy for spore ingestion. Ingestion was calculated as the proportion of remaining encysted or zoospores at a given time point compared to the number of spores recovered in the wells with filtered pond water at that time point. Three independent experiments were carried out in triplicate. Results shown are experimental means with standard error of the mean (Fig. 3d).

Vectors and potential reservoirs of B. salamandrivorans.

Anuran reservoirs of B. salamandrivorans. Four B. salamandrivorans isolates (AMFP13/1, AMFP14/1, AMFP14/2, and AMFP15/1) were used to expose 32 juvenile midwife toads (Alytes obstetricans) to 105 spores. Eight animals were inoculated per isolate. To assess whether infected midwife toads are capable of transmitting B. salamandrivorans to susceptible fire salamanders, from 14 days after inoculation, five randomly selected midwife toads per isolate were selected and each toad was housed together with a juvenile fire salamander in a new terrarium. The course of disease was followed up by daily clinical inspection and weekly sampling of the animals for 10 weeks (Fig. 4a, b). Two fire salamanders developed infection and clinical disease. They were taken out of the experiment and treated as described in ref. 26.

Urodelan reservoirs of B. salamandrivorans. The holotype isolate AMFP13/1 was used to inoculate 20 Alpine newts (Ichthyosaura alpestris). Four different infection doses (5 animals per infection dose) were used: 104, 103, 102 and 10 spores. The course of disease was followed up by daily clinical inspection and weekly sampling of the animals for 14 weeks (Fig. 4c, d). As exposure to 10 spores resulted in only 2 out of 5 animals becoming infected, results were omitted from further analyses.

Previous infection does not protect Alpine newts against B. salamandrivorans re-infection. In this study, we assessed whether Alpine newts that had been chronically infected by B. salamandrivorans are protected against re-infection with B. salamandrivorans. From previous experiments in which Alpine newts were exposed to a low (103 zoospores) dose of B. salamandrivorans (see urodelan reservoirs of B. salamandrivorans), nine animals developed chronic infection without subsequent mortality. Skin swabs from these animals were positive for B. salamandrivorans between 28 and 175 days after exposure to B. salamandrivorans (average ± s.d. = 95 ± 45 days), with average B. salamandrivorans counts per sample of log10(2.02 ± 0.54). All except one animal cleared B. salamandrivorans infection before experimental re-infection. Of the animals that cleared infection, the time between the last positive skin sample and the experimental re-infection ranged between 54 and 567 days (average ± s.d. = 278 ± 218 days). Re-infection with 106 zoospores of B. salamandrivorans was performed as described before and animals were followed up by determining B. salamandrivorans infection loads in skin swabs using qPCR (Fig. 4e). For comparison, five negative control animals were included that had never been exposed to B. salamandrivorans.

Forest soil as a vector for B. salamandrivorans transmission. 15 g of forest soil (moisture content: 47.24 ± 0.07%) was moistened with 10 ml of distilled water and inoculated with 1 ml of B. salamandrivorans suspension containing 6.0 × 106 GE. A fire salamander was exposed to 1 g of this soil for 24 h either immediately after inoculating the soil, after 8 h incubation, after 24 h incubation, after 2 days incubation, after 4 days incubation or after one week incubation. At the different time points, the amount of B. salamandrivorans in soil was determined with qPCR after DNA extraction using the Powerlyzer Powersoil DNA Isolation Kit (MO BIO Laboratories Inc.). This experiment was done in triplicate at 15 °C and 4 °C. (Extended Data Fig. 2, Extended Data Table 2).

Fourteen infected salamanders (mean infection load 3.2 × 103 ± 4.4) were housed individually on forest soil for 24 h. Afterwards the animals were removed and were replaced by a non-infected individual. In one group (7 animals), the replacement was done immediately after removal of the infected animal. In the other group, animals were replaced with a non-infected individual 24 h after removal of the infected animal. The course of disease was followed up by daily clinical inspection and weekly sampling of the animals for 4 weeks (Extended Data Table 3). An animal was considered to be infected after two positive skin swabs in two subsequent weeks.

Data analysis

To assess the response of animals to B. salamandrivorans infection under different conditions, we carried out a two-step analysis. First, we modelled the presence or absence of infection across all individuals using logistic regression; second, we modelled the average load of positive individuals only using linear regression. For estimation of the infection load, we always used the natural logarithm of the GE as the response variable. For all analyses where repeated measures of the same individual or sample were taken, we used a random effect to account for pseudoreplication. We modelled probability of infection and average load as a linear function of treatment and as an asymptotic (S. salamandra), quadratic (I. alpestris) or exponential (A. obstetricans) function of time, and included the appropriate treatment–time interactions. For the zoospore predation experiment, we used an open-population N-mixture model with a robust design28, accounting for sampling variability in the repeated measures for each sample and mortality between sampling occasions. We fitted all models in JAGS25 through package jagsUI in R.24, using uninformative priors for all coefficients including intercepts, sampling 10,000 posterior values from three Markov chains after a burn-in of 10,000 iterations. Convergence was inferred by values <1.1.

Data availability

The authors declare that the data supporting the findings of this study are available within the paper and the Supplementary Information files (available in the online version of the paper).