Short-Term Exposure to Warm Microhabitats Could Explain Amphibian Persistence with Batrachochytrium dendrobatidis

Environmental conditions can alter the outcomes of symbiotic interactions. Many amphibian species have declined due to chytridiomycosis, caused by the pathogenic fungus Batrachochytrium dendrobatidis (Bd), but many others persist despite high Bd infection prevalence. This indicates that Bd's virulence is lower, or it may even be a commensal, in some hosts. In the Australian Wet Tropics, chytridiomycosis extirpated Litoria nannotis from high-elevation rain forests in the early 1990 s. Although the species is recolonizing many sites, no population has fully recovered. Litoria lorica disappeared from all known sites in the early 1990 s and was thought globally extinct, but a new population was discovered in 2008, in an upland dry forest habitat it shares with L. nannotis. All frogs of both species observed during three population censuses were apparently healthy, but most carried Bd. Frogs perch on sun-warmed rocks in dry forest streams, possibly keeping Bd infections below the lethal threshold attained in cooler rain forests. We tested whether short-term elevated temperatures can hamper Bd growth in vitro over one generation (four days). Simulating the temperatures available to frogs on strongly and moderately warmed rocks in dry forests, by incubating cultures at 33°C for one hour daily, reduced Bd growth below that of Bd held at 15°C constantly (representing rain forest habitats). Even small decreases in the exponential growth rate of Bd on hosts may contribute to the survival of frogs in dry forests.


Introduction
The distributions, abundances, life history strategies, and virulences of microbial symbionts can all be influenced by the environment, and can, in turn, affect the development and outcome of disease [1,2,3]. Batrachochytrium dendrobatidis (Bd), the organism that causes chytridiomycosis, has been linked to the decline and potential extinctions of hundreds of species of amphibians around the globe [4,5]. In the tropics, higher elevation rainforest sites, which correspond to areas of high amphibian diversity and endemism, have been especially hard hit by this disease [6,7,8]. The cooler temperatures in these habitats coincide with the in vitro thermal optimum for growth of this pathogen [9,10]. In the laboratory, frogs infected with Bd can consistently lose their infections after relatively short (16 h) exposures to high temperatures (37uC), as the fungus perishes rapidly at this temperature [11]. Constant temperatures as low as 28-30uC resulted in death of the fungus after several days in culture [9], suggesting that even small elevations in environmental temperature are likely to tip the hostpathogen balance in favor of the host. However, the response of Bd to realistic thermal regimes experienced by persisting and declining populations in the wild has not been tested.
In the Australian Wet Tropics, the upland-endemic torrent frog Litoria lorica experienced severe population declines due to Bd at all known localities, and was last seen in 1991 [12,13,14]. By the mid-2000s it was thought to be extinct. However, in 2008 this species was rediscovered at a previously unknown locality in upland dry forest, where it occurs at relatively high local abundance, despite high prevalences of Bd infection. The newly-discovered site is approximately 6 km downstream from rain forest sites at which the species had been extirpated by chytridiomycosis. Another torrent frog, Litoria nannotis, declined from most high elevation rainforest sites, is sympatric with L. lorica, and occurs in high abundance with high prevalences of Bd infection at the L. lorica rediscovery site [15]. These species inhabit waterfalls and torrents, spending their days in and under the water and emerging onto adjacent rocks late in the day at the start of their nocturnal activity periods [16,17,18]. Puschendorf et al. [15] hypothesized that the coexistence of L. lorica with a potentially lethal pathogen in the dry forest was linked to the lack of canopy. Because infections must reach a threshold intensity before mortality occurs [19,20], dry forest frogs may reduce infection intensity by effectively ''basking'' on warm rocks after emerging from their diurnal shelters, killing or greatly reducing the growth rate of the fungus and avoiding mass mortality due to the disease [15]. There is evidence for similar behavioral regulation of infection status in at least one neotropical anuran [21].
To test this hypothesis, we collected substrate temperatures experienced by L. nannotis in open dry forest environments and adjacent rainforest. Based on these temperatures we carried out an in vitro experiment in which we grew Bd under thermal regimes simulating the thermal environments L. lorica experience at upland rain forest and dry forest sites. This is the first study to test Bd's response to temperature regimes experienced by diseased and persisting amphibian populations.

Methods
We used ten temperature dataloggers (DS1921Z-F5, Dallas Semiconductor, Dallas, Texas USA) placed on rocks that L. lorica perch on in the dry forest and eight dataloggers at sites that L. In the laboratory, Bd was grown in three temperature regimes, based on the rock temperatures recorded in the field. Bd (isolate Gibbo River, L. Les donna, 06-LB-1) was flushed from K-strength TGhL (eight g tryptone, one g gelatin, one g lactose, and 10 g bacteriological agar per liter of water) agar plates using three mL of K-strength TGhL and filtered to remove sporangia. 3.5610 4 zoospores were inoculated into 100 mL K-strength TGhL in each of 30 wells of each of three 96-well assay plates, a method modified from that of Rollins-Smith et al. [22]. All plates were kept at 15uC for the first 24 hours. Thereafter, one plate was kept at 15uC 24 hours a day to simulate constant, cool conditions at rain forest sites, a second was kept at 15uC 23 hours a day with one hour at 28uC to simulate daily exposure to moderately-warmed rocks as frogs emerge from diurnal retreat sites in dry forests, and a third was kept at 15uC 23 hours a day with one hour at 33uC to simulate emergence onto warmer rocks in dry forests. Treatments are hereafter referred to as 1) rain forest control, 2) dry forest 28uC spike, and 3) dry forest 33uC, respectively.
The growth of Bd cultures was measured spectrophotometrically at 492 nm [22] at the outset, and every 24 hours thereafter, immediately after the higher-temperature plates were exposed to their treatments. Cultures were also monitored visually using an inverted light microscope to observe growth and check for contamination.
The initial optical density at 492 nm (OD 492 ) for each well was subtracted from each subsequent reading to give the adjusted OD 492 , as change since time zero. To test for differences in Bd growth among treatments, a one-way ANOVA with Tukey's posthoc tests was performed on the final adjusted OD 492 for each treatment. Analysis was performed in SPlus 8.0 for Windows (Insightful Corporation, 2007).
We used laboratory assays rather than in vivo experiments because both species are endangered (L. lorica is critically endangered).

Results
Rock temperatures (Figure 1) differed significantly between dry and rain forest sites at the times frogs emerged from their diurnal retreat sites, between 6-7 pm. (Mann-Whitney U test, z = 23.58, P,0.001, n = 18). Dry forest substrate temperatures were commonly 30uC or greater (mean maximum temperature = 31.862.46) when frogs emerged from retreat sites, however maximum rainforest substrate temperatures were never above 20uC (mean maximum temperature = 17.5uC60.93).
After four days, cultures in all treatments had developed into dense sporangial aggregations. One-way ANOVA (F 2,87 = 10.11, P,0.001) indicated that Bd growth rates differed among the three thermal treatments. Tukey's tests (Figure 2) indicated that the dry forest 33uC temperature spike cultures had significantly lower Bd growth (mean adjusted OD 492 ) than those in either the rain forest control or the dry forest 28uC spike treatment. The growth of Bd did not differ significantly between rain forest control and dry forest 28uC spike treatment. Light microscopy of cultures during the trial did not reveal any observable differences in the development time of sporangia.

Discussion
Carey et al. [19] demonstrated that in highly susceptible species, the development of chytridiomycosis can be explained as a consequence of unregulated exponential growth of the population of Bd on the host. After multiple generations of exponential growth, even small differences in the growth rate parameter can make very large differences in population size. The reduction of Bd growth we detected with short-term elevation of temperature in the dry forest 33uC spike treatment was statistically significant, but was relatively small, and visual observation of cultures under light microscopy suggested that zoospores developed into mature sporangia over similar periods of time across the three treatments. This suggests that the lower density of cultures exposed to 33uC temperature spikes was caused by lower survival or settlement rates of zoospores, or lower production of zoospores per sporangium. Although our experiment encompassed only a single Bd life cycle, the small difference we observed in the population growth rate could translate to large differences in population size with time later in the population growth trajectory, and might account for the lower Bd virulence observed by Puschendorf et al. [15] at dry forest field sites. If the temperature responses of the Bd that occurs at the dry forest field site are similar to those of the Bd isolate used in this experiment, night ''basking'' on warmed rocks could explain the persistence of Litoria nannotis and L. lorica with Bd in dry forest habitats.
There are at least two alternative explanations for the persistence of Litoria nannotis and L. lorica with Bd at dry forest sites. First, frogs at two dry forest sites spent more time under running water than did those at rain forest sites [16]. This may have flushed away zoospores that would otherwise re-infect frogs and boost infection intensity towards a lethal threshold. Second, warmer temperatures may stimulate amphibian immune defenses [23,24,25], but there are few studies on the effects of short-term warming on immune function.
The response of Bd to temperature in vivo and/or in natural environments could differ from the responses we observed in vitro. We did not perform in vivo experiments due to the conservation status of our species, but our results are in accord with those of Richards-Zawacki [21], who found that the mean of the body temperatures of uninfected wild Atelopus zeteki individuals was higher than that of infected individuals.
It remains unclear why the amphibian-Bd symbiosis is shifted at least partway along the axis from host-pathogen to hostcommensal in dry forest Litoria populations. Short-term temperature spikes over longer time scales (several weeks) than were possible in our experiment may reduce Bd growth rates enough to allow host defenses to preclude the development of lethal infection intensities. We suggest that detailed data on daily body temperatures of individuals should ultimately be used to determine the conditions in which to carry out in vitro and in vivo experiments over longer periods that will illustrate the exact nature of thermal effects on the interaction between amphibians and Bd. Future work should also test the effects of realistic thermal regimes on amphibian immune function. Experiments could also be used to test the effects of immersion in running water. Understanding the specific factors driving context-dependency in Litoria-Bd symbioses would improve prediction of future chytridiomycosis-driven declines and location of refugia for threatened species, like that already found for Litoria lorica. Growth in the 33uC spike treatment was significantly lower than in both the rain forest and the 28uC spike treatments, but growth in the 28uC spike treatment did not differ significantly from that in the rain forest treatment. doi:10.1371/journal.pone.0026215.g002