Populations of a Susceptible Amphibian Species Can Grow despite the Presence of a Pathogenic Chytrid Fungus

Disease can be an important driver of host population dynamics and epizootics can cause severe host population declines. Batrachochytrium dendrobatidis (Bd), the pathogen causing amphibian chytridiomycosis, may occur epizootically or enzootically and can harm amphibian populations in many ways. While effects of Bd epizootics are well documented, the effects of enzootic Bd have rarely been described. We used a state-space model that accounts for observation error to test whether population trends of a species highly susceptible to Bd, the midwife toad Alytes obstetricans, are negatively affected by the enzootic presence of the pathogen. Unexpectedly, Bd had no negative effect on population growth rates from 2002–2008. This suggests that negative effects of disease on individuals do not necessarily translate into negative effects at the population level. Populations of amphibian species that are susceptible to the emerging disease chytridiomycosis can persist despite the enzootic presence of the pathogen under current environmental conditions.


Introduction
Parasites and pathogens can be important drivers of host population dynamics by altering host behaviour, demography or genetics [1][2][3][4]. The most extreme effect of a pathogen on the host is host extinction [5]. However, host extinction is rare for two main reasons. First, parasites and their hosts generally share a common evolutionary history and extinction of either antagonist is hence unlikely [6]. Second, when host density declines, the pathogen's transmission rate is also expected to drop, unless there is frequency-dependent transmission or a reservoir host [5]. Emerging diseases, however, are different from established pathogens because pathogen interactions with novel hosts are unpredictable. In the most extreme -albeit rare-case they can lead to host population declines and extinctions [7,8].
While the effects of disease on vital rates of captive populations are well documented, the effects of disease on dynamics of wild populations are still poorly understood [2,4,9,10]. For example, a reduction in population size is only expected if pathogen-induced mortality is additive rather than compensatory [4,11]. If pathogeninduced mortality is additive, then overall mortality is the sum of pathogen-induced mortality plus mortality inflicted by all other causes. In contrast, if mortality is compensatory, then increased mortality due to the presence of a pathogen is countered by a reduction in mortality due to other causes, often in a density-dependent manner [12]. Here, our goal is to contribute to a better understanding of host population dynamics influenced by an emerging infectious disease.
An emerging pathogen of amphibians, the chytrid fungus Batrachochytrium dendrobatidis (hereafter Bd), has contributed to amphibian declines and extinctions on most continents [13][14][15]. Many amphibian populations have collapsed after emergence of the pathogen, which is still appearing in new areas [16][17][18]. However, host extinction is not the only outcome of Bd emergence in new localities. The amphibian host-chytrid pathogen models by Briggs et al. [19,20] suggest that enzootic Bd-infection may initially cause a reduction in abundance, but thereafter, populations remain stationary (i.e., mean abundance does not change anymore). Yet, due to a lack of time series data on the abundance of amphibian populations coexisting with enzootic Bd, we cannot state with certainty that amphibian populations with enzootic Bdinfection are stationary. In contrast to the model predictions, some mark-recapture studies have shown that enzootic Bd depresses both individual survival and population growth rates [21][22][23].
Our goal was to quantify the effects of enzootic Bd on populations of an amphibian species that is known to be susceptible to Bd [24][25][26]. Ideally, one would compare population sizes before and after the emergence of Bd in a population. Unfortunately, such data are rarely available [16,17,27]. We compare population monitoring data from sites where Bd is present with population monitoring data from sites where Bd was not detected. We analyse short time series of counts of calling males from 26 populations of the common midwife toad, Alytes obstetricans, in the Swiss canton Lucerne. Assuming that Bd leads to chytridiomycosis-induced mortality of individuals [24][25][26] and that mortality is additive rather than compensatory [4,11], we expect to find declining populations in the presence of the pathogen while populations free of the pathogen should be either stationary or growing. The pathogen has been present in Switzerland since at least the early 1980 s [28], is widespread, and prevalence often high ( [29], U. Tobler & B. R. Schmidt, unpublished data). Although no chytridiomycosis-induced mass mortality has been observed in Switzerland, including our study area, we know that Bd-associated mortality occurs in the field because we have detected dead metamorphs at our study sites that tested positive for Bd (U. Tobler, C. C. Geiger & B. R. Schmidt, unpublished data). Additionally, high (up to 90%) Bd-related mortality has been reported in a laboratory experiment on postmetamorphic Alytes obstetricans [25]. Although mortality in the field may be lower than in the laboratory, these results led us to expect Bd-induced population declines.
Alytes obstetricans is listed as ''endangered'' (IUCN category EN) on the Swiss national red list of threatened amphibians [30]. The species is therefore the target of conservation action [31]. Knowing whether and to what degree Bd poses a threat to Alytes population survival in this area is vital in order to develop suitable conservation strategies. Currently, no habitat mitigation methods are available barring ex-situ captive breeding although mitigation methods using antifungal chemicals or bacterial treatments are currently being tested [32]. With this study, we aim to test whether or not chytridiomycosis represents an additional threat to host populations of an endangered species.

Study sites
Based on the availability of population count data, 26 sites in canton Lucerne, Switzerland, were included in the analysis. All sites are situated between 46.871u and 47.258u N and 7.882u and 8.382u E ( Figure 1) and along an altitudinal gradient ranging from 402 to 1330 m.a.s.l. Mean summer temperatures range from 12.3 to 17.5uC [33]. Habitat types include quarries, ponds in open meadows, garden ponds, fire water reservoirs, and alpine and prealpine streams.

Population monitoring
Populations of Alytes obstetricans were regularly monitored as part of the midwife toad action plan of the Swiss canton Lucerne [34,35]. Counts of the number of calling Alytes obstetricans males were obtained every year from 2002 to 2008 except for four sites where Alytes obstetricans was reintroduced and calling males only occurred after 2002. Every site was visited at least twice a year by experienced volunteers. Volunteers were free to choose the nights for the monitoring, but advised to do so during optimal warm and humid nights when detectability of Alytes obstetricans is high because many males are calling [36]. The number of calling males, an index to population size [37], was recorded during every visit. We used the highest number of calling males within a year in the analysis. In our analysis of population trends, detectability may vary among years (the statistical model accounts for observation error; see below) but we assume that detectability shows no temporal trend such that population trends can be reliably estimated from counts of calling males [38,39]. Trend estimates are unlikely to depend on Bd infection status because observers were unaware of Bd infection status of a site. Hence, we can exclude an interaction between Bd presence and detectability that would have biased our conclusion on the impact of Bd on population trends. In addition to the data on the numbers of calling males, presence or absence of tadpoles (based on visual encounter surveys [40]) was noted; this provided us with the explanatory variable ''number of years tadpoles were observed'', which indicates the number of years in which tadpoles were observed in the pond.

Sampling for Bd
During summer 2007 and spring 2008 or 2009, all sites were sampled for the presence of Bd in the amphibian population. To test the populations for Bd, 16 to 47 (mean 6 SD: 26.066.6) amphibians were caught and swabbed with a sterile rayon swab (Copan Italia S.p.A., Brescia, Italy). Exceptions are sites ''Sagerhüsli'', where only two dead Alytes metamorphs were sampled, and ''Hombrig'', where sample size was six. Because both these sites tested positive for Bd we do not expect the small sample sizes to affect our results because the goal was to determine presence or absence of Bd. Sample size at sites where we did not detect Bd was 26.567.6 (mean 6 SD). Hence, we cannot exclude that we may have missed Bd at some sites where prevalence was low (with a sample size of n = 26 a prevalence of 10% may not be detected; see [41]). Sampling was done opportunistically, i.e. all available amphibian species that could be captured were tested for Bd. Apart from midwife toads (where we swabbed tadpoles), the other amphibians sampled (always adults) were the fire-bellied toad Bombina variegata, the waterfrogs Pelophylax lessonae and Pelophylax esculentus (the two taxa were pooled because they form a hybridogenetic complex [42]), the alpine newt Mesotriton alpestris and the palmate newt Lissotriton helveticus (scientific names are based on [43]). To swab tadpoles, we made five swipes across the mouthparts. To sample adult amphibians, we swiped the underside of each foot five times and the ventral abdominal skin five times for a total of 25 swipes per amphibian. In 6 out of 11 sites where we did not detect Bd, samples were exclusively obtained from Alytes tadpoles. Due to the prolonged larval period Alytes tadpoles are more likely to be infected than any other species or life stage in this system (see results). Standard hygiene recommendations were followed during field work [44,45].  We followed the rt-PCR protocol of Boyle et al [46] for the extraction and analysis of Bd-DNA from swabs. We used Bdspecific primers and standards to quantify the amount of Bd-DNA (infection load). To prevent inhibition by the extraction reagent, the extractions were diluted 1:10 with water prior to PCR analysis. Hence, we calculated the original zoospore equivalent by multiplying the PCR output by 10. We ran each sample twice and the PCR was repeated if the two wells returned unequal results. Reactions below 0.1 genomic equivalents were scored Bdnegative [25].

Statistical analysis
We tested for differences in prevalence of Bd among the sampled species using a generalised linear mixed model (GLMM) with a binomial error distribution. We tested for differences in infection intensity among species using a linear mixed model (LMM) with a normal error distribution. In both analyses, we used site as grouping (random) variable. Both analyses were done in R 2.8.1 [47].
We used WinBUGS 1.4 [48] to fit a a state-space model to the 26 time series [49,50] and to assess whether the presence of Bd and the number of years tadpoles were observed affected population trends. State-space models disentangle the effects of the biological process and the observation process and thereby account for observation error. Modeling followed closely the approach described in Kéry & Schaub [50]. We built a model that estimates the observation and biological process at two hierarchical levels. The time series counts c i,t are described by where N i,t is the unobserved true population size of site i at time t and s 2 c,i the observation variance at site i. This part of the model describes the observation process and removes observation error from the time series. The biological process is described by wherel i,t is the population growth rate of site i at time t which is assumed to be have a normal distribution where l l i is the mean population growth rate at site i and s 2 n,i is the process variance at site i. l l i was further modeled as a function of site-specific presence of Bd (Bd i ) and of the site-specific number of years tadpoles were observed (tad i ) using a linear relationship: Uniform priors were used for the observation (U(0, 30) for each site) and for the process standard deviations (U(0, 10) for each site). A normal prior with a wide variance N(0, 100) was used for a, while for both b we used a uniform prior U (25,5). Further, we used Normal priors for the population size at the first occasion for each site with variance 1 and the mean equal to the site-specific counts in that year. We ran three independent MCMC chains for 25,000 iterations, each with a burn-in of 10,000 iterations. The chains were thinned by a thinning factor of three. Convergence was assumed if the Gelman-Rubin statistic Rhat,1.1 [51].

Ethics statement
The experiment was conducted under permit number 110/ 2007 by the veterinary office of the canton Zurich; collecting permits were provided by the office for Landwirtschaft und Wald (lawa) of the canton Lucerne.

Results
We detected Bd in 16 out of the 26 (61.5%) sites and in 16.5% of all sampled amphibians, including the sites where we did not detect Bd (Table 1). In sites where Bd was detected, 28% of all sampled amphibians tested positive for Bd. Bd was not found in any site in the Entlebuch valley, which encompassed the southwestern cluster of populations (Figure 1 (Table 2). This means that on average the number of calling males in the region was stationary. For 14 out of the 26 study sites, the 95% credible interval (CRI) for the estimated population trend included 1, i.e., the populations were neither growing nor declining (Figure 2). Populations that tested Bdpositive had on average a higher l l i value than populations where we did not detect Bd. The difference in l l i values was 0.191 (95% CRI 0.046-0.353; this is the regression coefficient for the effect of Bd on population growth rate (see equation 4 and Table 3)). Populations where tadpoles were observed in a higher number of study years had an increased growth rate compared to populations where tadpoles were not or only rarely observed; every year in which tadpoles were observed increased the growth rate by 0.024. The 95% CRI marginally included zero (95% CRI 20.007-0.051; Table 3). Repeating the analyses excluding the reintroduced populations did not alter the results qualitatively because parameter estimates were almost equal (results not shown).

Discussion
We found that Alytes populations in the Swiss canton Lucerne remained stable despite the presence of the pathogen, but that they were small with an average of only 10 calling males. Bd-positive populations did not have lower population growth rates than populations where we did not detect Bd. The result is robust even if we failed to detect Bd infection when it was present in some populations. If all populations where we did not detect Bd were Bdpositive, then the presence of Bd would still not lead to negative population trends since the average population growth rate is close to one.
The result was unexpected. The absence of negative effects of Bd on population trends contrasts strongly with many studies on other species that report dramatic negative effects of Bd on amphibian populations, including global extinction of species ( [14][15][16][17][18]22], but see [52]). It is even more surprising since Alytes obstetricans is known to be highly susceptible to Bd [23][24][25]. In a laboratory experiment, we showed that Bd-associated mortality of Alytes obstetricans shortly after metamorphosis was up to 90% [25]. Thus, there can be strong individual-level effects of Bd.
Population models suggest that high juvenile mortality lowers population growth rates in species with complex life cycles [53][54][55][56]. Hence, we expected that high chytridiomycosis-associated juvenile mortality in Alytes obstetricans in the laboratory [25] would lead to population declines. Even though we observed Bd-infected dead metamorphs in the field, there was apparently no effect of Bd-associated juvenile mortality on population trends. There are several possible explanations why there were no population-level effects of Bd in the field despite the strong individual-level effects in the laboratory [25]. The explanations are not mutually exclusive.
The first explanation is based on the fact that environmental conditions, especially those related to altitude and temperature, may mediate the effects of Bd on amphibian populations [26,57,58]. Under the prevailing environmental conditions, juvenile mortality in the field may be lower than in the laboratory [25]. It may be that environmental conditions, especially climate, in our study area may be such that there is some Bd-induced mortality of individuals but no population declines. However, we do not think that the populations that we studied experienced environmental conditions hostile to Bd. First, some dead and Bdpositive metamorphs were observed at two of our study sites. Admittedly, metamorphs found dead in the field had very low Bd loads perhaps indicating a cause of death other than chytridiomycosis (we did not do post mortem examinations to determine the cause of death). Second, one fifth of our populations were within the summer temperature range within which fatal chytridiomycosis is observed in Spain [26]. Further, because Bd occurred at all elevations and thus all climate regimes within our study region, we can exclude altitude as being confounded with Bd presence.
A second explanation may be that Bd-induced mortality could be compensatory rather than additive when Bd is enzootic. A decline in population size is only expected if mortality due to disease is additive, i.e. individuals die that would not die for other reasons in the absence of the disease [11]. While disease-induced mortality during Bd epizootics is obviously additive (e.g. when Bd epidemics caused Alytes obstetricans population declines in Spain [24]), this may not be the case for enzootic Bd. Compensatory mortality can result in a lack of disease effects on host abundance [4,11]. If mortality is compensatory, increased mortality due to the presence of Bd would have to be countered by a reduction in mortality due to other causes, often in a density-dependent manner [12]. Density-dependence in the terrestrial stages of amphibian populations may indeed occur [59][60][61].
The third explanation may be that effects of Bd on abundance occurred in the past and are no longer measurable. The Briggs et al. host-pathogen models [19,20] suggests that amphibian populations may decline to lower abundance after the emergence of Bd but remain stationary at a smaller size after Bd became enzootic. Indeed, strong Alytes obstetricans population declines were observed in our study area in the 1980 s and 1990 s [31]. Since Bd has occurred in Switzerland since at least the early 1980 s [28], it may be that Bd contributed to these declines in the past. Today, the Alytes obstetricans populations may have reached the stationary state predicted by the models such that an effect of Bd on abundance is no longer detectable. This is also supported by the low infection intensities observed in the field today. If populations have stabilized at lower abundance -which is highly plausible given the small population sizes observed in this study -they may now be less resilient [62] or more prone to environmental and/or demographic stochasticity because of reduced abundance. It is also possible that unusual environmental conditions could interact with Bd to affect populations in the future.
Although there are several possible explanations for why we did not observe the expected negative effect of Bd on abundance, the question remains open as to why Bd-positive populations had higher growth rates than Bd-negative populations. The more years tadpoles were observed at a given site, the more likely the population was to grow (Table 2). This may suggest that recruitment may determine population growth [55,63]. Hence, the compensatory mechanism that we alluded to above may simply be increased recruitment. Increased recruitment of Bdpositive populations of Anaxyrus (Bufo) boreas may allow these populations to persist despite the presence of Bd [22]. l l is the average population growth rate. s 2 c is the observation variance and s 2 n is the process variance. All estimates are given as means 6 standard deviation with the 95% CRI in brackets. Bd status 0 means that Bd was not detected while Bd status 1 means that Bd was detected. doi:10.1371/journal.pone.0034667.t002 Table 3. Parameter estimates (equation 4), standard deviations and 95% CRI for the effects of the presence of Bd and the number of years in which tadpoles were observed on population growth rates. In conclusion, we demonstrated that populations of a species that is susceptible to an emerging pathogen can grow despite a high prevalence of the pathogen. Evidently, individual-level effects of disease (mortality of individuals) did not translate into population-level effects (negative population growth rates). Our results are phenomenological and we do not know the mechanisms that allow the populations to persist. Understanding why an amphibian species that is known to be susceptible to Bd can have growing populations despite high prevalence of the pathogen would be a key to successful mitigation of the effects of chytridiomycosis.