Batrachochytrium dendrobatidis System: Fine-Scale Patterns of Variation in Responses to a Multi-Host Pathogen

Species composition within ecological assemblages can drive disease dynamics including pathogen invasion, spread, and persistence. In multi-host pathogen systems, interspecific variation in responses to infection creates important context dependency when predicting the outcome of disease. Here, we examine the responses of three sympatric host species to a single fungal pathogen, Batrachochytrium dendrobatidis, which is associated with worldwide amphibian population declines and extinctions. Using an experimental approach, we show that amphibian species from three different genera display significant differences in patterns of pathgen-induced mortality as well as the magnitude and temporal dynamics of infection load. We exposed amphibians to one of four inoculation dose treatments at both larval and postmetamorphic stages and quantified infection load on day 8 and day 15 post-inoculation. Of the three species examined, only one (the Pacific treefrog; Pseudacris regilla) displayed "dose-dependent" responses; survival was reduced and infection load was elevated as inoculation dose was increased. We observed a reduction in survival but no differences in infection load across pathogen treatments in Cascades frogs (Rana cascadae). Western toads (Anaxyrus boreas) displayed differences in infection load but no differences in survival across pathogen treatments. Within species, responses to the pathogen varied with life history stage, and the most heavily infected species at the larval stage was different from the most heavily infected species at the post-metamorphic stage. Temporal changes in infection load were species and life history stage-specific. We show that variation in susceptibility to this multi-host pathoqen is complex when viewed at a fine-scale and may be mediated through intrinsic host traits. Citation: Gervasi S, Gondhalekar C, Olson DH, Blaustein AR (2013) Host Identity Matters in the Amphibian-Barrachuchyrrium defldrobaridis System: Fine-Scale Patterns of Variation in Responses to a Multi-Host Pathogen. PLoS ONE 8(1): e54490. doi:10.1371/journal.pone.0054490 Editor: Brian Gratwicke, Smithsonian's National Zoological Park, United States of America Received October 1, 2012; Accepted December 12, 2012; Published January 24, 2013 This is an open-access article free of all copyright. and may be freely reproduced, distributed, transmitted, modified, built upon, or otherwise used by anyone for any lawful purpose. The work is made available under the Creative Commons CCO public domain dedication. Funding: Funding for this study was provided by the United States Forest Service (Pacific Northwest Research Station; http://www.fs.fed.us/pnw/) and the Robert and Clarice MacVicar Animal Health Scholar Award (issued through Oregon State University to SG). The funders had no role in study design, data collection and analysis, decision to publish, or presentation of the manuscript. Competing Interests: The authors have declared that no competing interests "xist. • E-mail: gervasis@science.oregonstate.edu


Introduction
Most pathogens can infect and be trammittcd a::nong multiple hosts, but there may be considerable variation in responses to infection among species, individuals, or between life stages ll ,2,3, 4,5J. DifTcrcnces in host responses to pathogens can influence disease dynamics by altering the probability and rate of pathogen transmission, disease induced mortality, and recovery from infection L6, 7 ,8]. Theoretical and empirical evidence suggests that variation in host competence (i.e. the capacity of a host to acquire, persist with, and transmit a pathogen) can modify the abundance and probability of persistence of pathogens in ecological communities [3,9J. For example, reservoir hosts harbor and transmit a pathogen, often without incurring morbidity or mortality as a result of infixtion l3 J. Species that act as reservoirs can facilitate long-term pathogen persistence in space and time and prevent periodic pathogen fade-outs based on density of susceptible hosts l3, 7,9]. Therefore, species identity and compo-Intrinsic host traits may play a fundamental role in determining responses to infection and disease outcomes among amphibian populations [24,27,28,29]. Species that share key life history traits, certain behaviors, and/ or geographical and microhabitat prcll~r ence may exhibit similar patterns of susceptibility [27,29]. Physiological or immunological characteristics of host species may also influence the probability of infection, degree and duration of infection, infectiousness and the probability of disease-induced mortality [30,31 J and genetic traits associated with immunity may allow some individuals or groups of similar individuals to be more effective at lighting infection [32].
Bel infects keratinized structures of amphibians [33 J. In larval amphibians, keratin exists in mouthparts and infection can impair feeding [34] but also may contribute to larval mortality [23,35]. As development continues, keratin begins limning on other parts of the body; by metamorphosis animals arc covered with keratinized epidermal tissue [35 J. Pathogenicity of Bel results from damage to the epidermal skin layers of post-metamorphic amphibians, but also impairment of osmoregulatory mechanisms of the skin and secondary electrolyte imbalance that may lead directly to cardiac arrest in amphibians [36]. Variation in skin structure (propensity to act as an clkctive pathogen "barrier") as well as variation in the amount of keratin and other proteins involved with the skin could alter host-pathogen interactions in this system. Pathogenicity may also arise from fungal metabolites, including toxic substances that cause morbidity and mortality in larval and post-metamorvhic amphibians [23,37,38].
There is a paucity of data characterizing the individual and species-level interactions between amphibian hosts and this fungal pathogen using an experimental framework. Thus, using a comparative experimental approach, we examined responses to Bel in three sympatric host species of three genera: tree frogs (Pseudacris regilla;, true frogs (Rana cascadae), and toads (Ana.xyrus boreas). '"' e exposed animals to one of four dilfercnt inoculation dose treatments and examined mortality and inJection load at both the larval and po:.t-metammvhic stage. ''Vc sampled infection load at two distinct time-points in both larvae and rnctmorphs of these three species.

Animal Husbandry
Eggs of all species were collected with handheld dipncts from natural ponds in the Oregon Cascade Range west of Bend, Oregon (R. cascadae and P. regilla: Linn County, OR; eleva-tion= 1140 m; A. boreas: Deschutes County, OR; eleva-tion= 2300 m) during the summer of 20 II (R. ca.;cadae and P. regil!a collected in May and A. boreas collected July). We collected 5-10 clutches of each specie> to increase genetic variability in our samples, and clutches were fully mixed. Eggs were brought to a climate-controlled laboratory within 4 hours of collection and kept at 14.5-15.5°C on a naturallight:dark photoperiod (regulated bv an outdoor, light-sensitive receptor). After hatching, we separated tadpoles to a density of approximately I 00 tadpoles per aquarium, and raised them to Gosner stage 27-30 (marked by early development of hind-limb buds [39]). During development in the laboratory, tadpoles were fed a 3:1 mixture of rabbit chow and fish food ad libitum (during the experiment, we fed tadpoles a pinch (approximately 0.05 g) of the same mixture of food, every other day). \'\'e changed water in larval aquaria wcdJy. Once the majority oflarvae reached the desired Gosner stage, we randomly selected tadpoles for experiments. Larvae were housed in individual 1 L containers for the duration of the experiment and were fed one pinch of food every three days. All larvae not used in For experiments with metamorphic animals, we removed free swimming stage tadpoles from the same clutches in groups of 50 to outdoor semi-natural mesocosms. These animals were kept in outdoor mesocosms for the duration of their larval and early metamorphic development. \'Vc selected tadpoles for mesocosms that were of approximately the same size and that were actively swimming in laboratory aquaria. Aquatic mesocosms were 36 inch wide x21 inch deep (~350 L) cattle tanks, filled with well water. \'Vc added 2 L aliquots of pond water to each mcsocosm. Pond water provided a natural source of algae and zooplankton to all mesocosms. ''Ve stocked each mcsocosm with 100 g of dry oak leaves and 20 g of rabbit chow to provide substrate for periphyton and a source c.f nutrients for larvae. Mesocosms were covered with plastic screening to prevent colonization by insects and predators and were allowed to develop for 15 days before adding larval amphibians. Because of differences in breeding phenology, A. boreas ta.dpolcs were raised in mcscosms from July to October, while R. cascadae and P. regilla tadpoles were raised in mcsocosms from :vlay to August. As animals began to metamorphose (i.e., emergence of both forelimbs), we removed them to semi-aquatic/ semi-terrestrial mcsocosms containing a thin layer of water and oak leaves to prevent drowning. Animals were maintained at a density of approximately I 00 animals per mcsocosm and were not maintained separately by larval/aquatic mcsocosm (i.e., individuals within the same species were mixed as they entered semiterrestrial mesocosms). Animals in terrestrial mcsocosms were allowed to complete metamorphosis (complete tail absorption) and remained in these enclosures lor 2·-3 months after metamorphosis. After metamorphosis, animals were fed pinhead crickets and wingless fruit flies weekly, ad libitum. After 2-3 months of development, metamorphic amphibians were brought into the same laboratory where larval experiments were run. Animals were allowed to acclimate for 48 hours before they were weighed, measured, and assigned to their experimental treatment. All metamorphic animals (rom terrestrial mesocosms were rerandomizcd for subsequent experiments in the laboratory. IVlctamorphic amphibians were housed in individual large-sized ( 140 x 30 mm) petri dishes throughout the laboratory experiment and were fed pinhead crickets by average body weight ( 1 cricket per 0.1 g body mass; same number of crickets for each animal) twice a week.

Inoculate Preparation
Batradwchytrium dendrobatidis isolate, JEL 274 originally isolated from A. boreas from Colorado [40], was cultured on 1 00 mm x 15 mm tryptonc agar plates and allowed to grow for 6 days at 23°C before inoculation of animals. Previous studies have shown that zoospore activity and density with this particular strain of Bel arc highest within 6-8 days after culturing the pathogen on tryptonc agar plates [24]. Zoospores were harvested by flooding agar plates with 1 0 ml of dechlorinated water and scraping the surface of the agar before pooling the inoculums of several (5-6) plates [2+]. Experimental animals were exposed to 10 ml of pooled inoculation broth at a density of 100,000 zoospores total ("high" inoculation dose), 50,000 zoospores total ("intermediate" inoculation dose), or 10,000 zoospores total ("low" inoculation dose). All zoospore counts were determined by hemocytometer from pooled inoculation broth. For larval amphibians, zoospores were transferred into individual plastic cups, already containing 600 ml of dechlorinated water. In metamorphic amphibians, zoospores were translcrred to individual large petri dishes, already containing 15 ml of dechlorinated water. Control animals were exposed to the same volume of sham inoculation (created from pathogen-free tr}1Jtone-agar plates). Thus, throughout the experiment, a thin layer of inoculated water was present at the bottom of the petri dish. Animal>. resting on the bottom of petri dishes were exposed to but not immersed in the diluted inoculate and could climb the walls and tops of Petri dishes. This method standardized pathogen exposure regimes among amphibian species, allowing us to assess baseline differences in species responses to the same treatment regime.

Experimental Pathogen Exposure
For both larval and post-metamcnvhic animals, the experimental duration was 15 days. Animals were exposed to their treatments once, at the start of the experiment and then observed daily. vVe did not to re-inoculate animals over the course of the infection because we were: specilically testing how infection load changes over the time-frame of a single exposure event. \·V e assessed infection loads in larval and post-metamoqJhic animals on day 8 and day 15. \Ve only sampled li\'e animals at each time-point, since pathogen growth, and thus, infection load could change rapidly on animals that have died and begin to immediately decompose. Thus, we established how infection load differs on day 8 versus day 15 in living animals, exposed once to a standardized inoculation dose of Bel.

Larval Pathogen Exposure
\V c staged all larvae immediately prior to lllltlal!on of the experiment, and then placed Gosncr stage 27-30 animals [39J in individual 1 L plastic cups (the experimental Llllit), containing 600 ml of water. Larvae were randomly assigned to one of four treatment groups (n = 30): High Bel close, lntennecliatc Bel close, Low Bel close, or Control (no Bel). We checked for mortality daily, and preserved dead animals in 95% ethanol. On clay 5 of the experiment, we increased the water level to 900 ml to increase oxygen content in the cups. On day 8 of the experiment, we randomly sampled 15 individual tadpoles in each experimental treatment for quantitative PCR and changed all remaining (nonsampled) animals into clean water. At the end of the experiment on clay 15, the remaining animals were humanely euthanizcd in accordance with the institutional animal care protocol and preserved in 95% ethanol. Post-experimental snout-vent-length and mass were taken only for tadpoles that were sampled on day 8 and day 15 for quantitative PCR.

Post-metamorphic Pathogen Exposure
All animals from semi-terrestrial outdoor mcsocosms went through 4-8 h of acclimation in individual Petri dishes to laboratory conditions before initiation of the experiment. At the start of the experiment, all animals were weighed, measured, and randomly assigned to one of four treatment groups, identical to treatments for larvae. For metamorphic R. cascadae and P. regi!la, 30 animals were assigned to each treatment. For metamorphic A. boreas, 20 animals were assigned to each treatment, except in the high dose treatment (n = 21 ). As with larval amphibians, we checked for mortality daily. On day 8 of the experiment, we sampled half of all remaining animals for qPCR and changed the water in each petri dish (without re-inoculating animals). At the end of the experiment on clay 15, the remaining animals were humanely cuthanizccl in accordance with institutional animal care protocol and preserved in 95% ethanol.

Quantification of Infection Load
For both larval and post-metamorphic amphibians, we used quantitative-PCR (qPCR) to assess infection load in larval and pust-metamcnvhic amphibians following the methods of Boyle et a!. [4-lJ, except that we used 60 J.ll of Propman Ultra (Applied Biosystcms) instead of 4-0 ml in all DNA extractions. Extractions were diluted l: 10 and processed in an ABI PRISM 7 500 (Applied Biosystcms). \Vhole mouthparts were extracted for qPCR in larvae. For rnctamoqJhic amphibians, ventral abdominal skin and inner thigh skin was swabbed using line tipped sterile rayon swabs (Medical Wire and Equipmcnt-!VJW&E 113) lor Bel. We sampled 15 larval amphibians for qPCR in each pathogen-exposed treatment on clay 8 c·f the experiment and 8 larval animals from each pathogen-exposed treatment on clay 15 of the experiment. In metamorphic amphibians we sampled half of all surviving animals for qPCR on clay 8 of the experiment and the remaining animals were sampled on clay 15. Each sample was analyzed in triplicate and the average number of genome equivalents of Bel per animal was calculated. For larval and post-metamorphic amphibians, we also randomly sampled at least 3 control animals at each timepoint and in each treatment. All control animals (larval and pustmetammvhic) tested negative for infection in all three replicate DNA wells for quantitative PCR.

Statistical Analyses
Statistical comparisons were made among trC'atmcnts (for each species, individually), among species (at each treatment level, separately), and between time-point~ (clay 8 versus day 15). Each of level of statistical comparison (treatment, species, time-point) was carried out individually, as opposed to a single analysis with multilevel comparisons. We used Kapla.n-Meicr (or product-limit) analyses in S-plus version 8.0 for \·Vinclows to generate "survival curves" among species and among treatments. This nonparametric method allows the analysis of survival data that is often right skewed. ln addition, Kaplan-Meier analyses allow for "censored data", which accounts for sub-sampled individuals in our experiment that were "lost" or "removed" due to destructive sampling (not to death). This was important in our dataset, since a subset of live animals were sampled (i.e., removed from the experiment) on clay 8 of the experiment for ;J quantitative assessment of infection load. Thus, the survival curves generated take into account mortality as a constant function of the remaining (but not sub-sampled) individuals. To statistically compare survival curves, we used a Cox's proportional hazards model . The Cox proportional hazards model give an overall p-valuc (Likelihood ratio test) which assesses the validity of the model, as well as p-valucs for each factor and an associated "hazard ratio". The hazard ratio represents a comparative indicator of the risk or probability of mortality associated with a given factor (a hazard ratio> l indicates an increase in the probability of mortality). Higher hazard ratios arc associated with a greater probability or risk of mortality clue to association with that facl<lr (e.g., treatment or species). One caveat with Cox proportional hazards models for survival analysis is that some mortality (greater than l individual) must exist to make comp<uisons among groups. Thus, if zero mortality is seen in a group, a comparison between that group and others is not possible. We only experienced this problem in tadpoles where, for one species, no mortality existed across all experimental treatments, and in one treatment for a single post-metamoqJhic species (i.e., zero mortality in the control treatment for il. boreas mctammvhs). For all analyses with post-metamoqJhic amphibians, our proportional hazards models also included initial body mass as a factor. Survival analyses of larval amphibians were performed only with main cllccts (treatment or species), as thnc was no initial metric of body size or mass available; we only had post-experimental estimates of size and mass on the ;'ubset of animals sampled for quantitative-PCR analy;,is). Cox proportional hazards models were performed in R, statistical computing environment (version 2.9.0, Institute for Statistics and Mathematics, Vienna) with the "coxph" Junction and the Survival package for survival analyses within and among species.
For infection load analyses, we transformed quantitativc-PCR loads (log-average genome equivalents per individual + l) to normalize data and used a one-way analysis of co-variance to examine within species (among-treatment) and among species (within-treatment; dillcrcnces in infection load in larval and postmetamorphic animals. \•Ve used treatment or species as the main effect and included mass as a covariate in all models (since infection load may depend on the amount of host-body-area or mouthpart areas that the pathogen can colonize). \•Ve used a Levene's test to check for homogeneity of variance among groups (treatments or species) before running all ANCOVA's (and in all cases, we met the requirements of the parametric test). \Vhen we detected a significant overall effect of treatment or species, we carried out individual pair-wise comparisons to control type I error using a Tukey's honestly significant difference test. In the case of low sample size (for larval or post-metamorphic amphibians some species/treatment combinations had only one individual because of lack of Bel-positive animah or because of high mortality during the experiment) we used \'\felch's modified t-tests to make between group comparison;;. This t-test controls for non-equal variance between groups. \'Vc also examined temporal dynamics of infection by comparing quantitative-PeR loads on clay 8 of sampling versus day 15 of sampling within treatments and across species using \Velch's modilied t-tests.

Ethics Statement
This study was carried out in strict accordance with the recommendations in the Guide for the Care and Usc of Laboratory Animals of the National Institutes of Health. Animals in the study were allowed to die directly as a result of infection and this is in accordance with the approved Institutional Animal Care and Usc Committee of Oregon State University (ACUP # 4184). Criteria for euthanasia included display of overt signs of morbidity associated with infection (loss of righting reflex and anorexia in post-metamorphic animals and extreme weakness in larval amphibians). In our study, we checked animals twice, daily, for these signs of morbidity and were humanely euthanized if we observed signs (although we did not observe these extreme signs of morbidity in our two week experiment). Any animals appearing to show any sitms of distress were immediately euthanizecl in MS-222 according ~~ institutional animal care protocol. Collection of amphibian eggs for all experiments was approved by the Oregon Department of Fish and Wildlife (20 ll Oregon Scicntilic Taking Permit #006-12 issued to A.R. Blaustein).

Larval Survival
There were no dilfcrcnces in rates of monalitv across experimental treatments, including the control treatment, for any larval species (Table I). We also failed to detect species-level differences in survival (R. cascadae versus P. regilla only) within pathogen treatments (Table 1). \·Ve did not include A. boreas in the among species comparisons because we observed zero mortality across all treatments, including control and pathogen exposures, for larval A. boreas and this precludes the ability to usc a Cox proportional hazards model.

Larval Infection Load
Pseudacris regilla. Although we did not detect an effect of the pathogen on survival, we did observe significant differences in infection load in larval amphibians dependent on treatment, species, and/or sampling time-point (Table 2; Table Sl ANCOVA F:l,n = 6.47;p = 0.005). A Tukey's HSD test revealed that infection loads were signilicantly greater in high inoculation treatment animals than low inoculation treatment animals (p<0.05). Sampling of larval toads on day 15 also revealed a significant difference in in!Cction load among treatments (ANCOV A F 3 . 12 = 5.18;p := 0.023); animals in the high pathogen inoculation treatment had higher levels of infection than animals in the intermediate inoculation treatment (p<0.05). Infection load in larval toads was higher on day 15 compared to day 8, but only signilicantly so in the high inoculation treatment (\'\felch's 2sample t-tcst: t 1 H= -2.82; p=O.Ol2). For A. boreas tadpoles, the proportion of animals testing positive for infection in the high inoculation treatmmt stayed at 87% from day 8 to clay 15; in the intermediate inoculation dose treatment went from73% on day 8 to 87% on day 15; and went from 46% on day 8 to 25% on day 15 in the low inoculation dose treatment.
Among-Species. We detected a significant ellcct of larval species identify on infection load, but only on day 8 of the experiment and at the high inoculation dose treatment (    Table 2; Table S2; ANCOVA; Fl. 19 Table 2. Batrachochytrium dendrobatidis infection load summary for larval and post-metamorphic anurans of th ree species.

Species
Day Treatment

Average genome equivalents+/-SO Larval
Average genome equivalents +/-SO Postmetamorphic  Table I) and mass was not a significant factor in this model. In the within species analysis, we did not include a comparison of survival between the control and pathogen exposed treatments because there was zero mortality in the control treatment for this species, precluding this treatment from the Cox modeL Although there were no surviv<JJ differences among the three pathogen treatments, there was a significant clfect of treatment on infection load in A. boreas on clay H

Larval versus Metamorphic Stage Comparisons
Although we did not examine the relationship statistically, as expected, there was a qualitative differenCI: in the effect of the pathogen on survival between amphibian life history stages (Table I); larval amphibians of all species had greater survival than all post-metamorphic animals. There was no evidence for any significant pathogen elfect on survival in larval amphibians. Temporal dynamics of infection varied between stages as well. While we detected greater infcct1on loads in larvae on day 15 than day 8, changes in infection load from day 8 to day 15 were different in each post-metamorphic species we examined, dependent on species afliliation,. In general, inft:ction loads were lower in larval species than in post-metamorphic amphibians ( Table 2) but there were exceptions. For example, infection loads in larval A. boreas and P. regiila, in particular by day 15 of the experiment, .were greater than infection loads in post-metamorphic R. cascadae (Table 2). Qualitatively, we note that infection loads in larval and pm:t-metamorphic R. cascadae were similar. Thus, for at least one species, the zoospore load localized in tadpole mouthparts was roughly equal to the amount of infectious particles found on the surface of the skin in post-metamorphic animals ( Table 2).

Discussion
Three co-occurring amphibian species exhibited very clilfcrent responses during experimental exposure to the fungal pathogen, Batrachochytrium dendrobatidis. Variation in responses was pronounced at the species level, treatment level, between life history stages, and between sampling time-points. Vve show that intrinsic host traits, alone, can be important rnf'diators of pathogen dynamics in the amphibian-chytricl fungus system. Similarly, temporal and dose-dependent responses of hosts could affect species persistence in natural environments and alter species composition in ecological assemblages.
Previous studies have sb own that some species arc very susceptible to Bd and succumb to mortality soon after exposure, [24,43] while other species show no signs of morbidity or mortality in the presence of the pathogen [22]. However, increasing evidence, including results from thi~: study, show that many species fall within the middle of this susceptibility continuum, with pronounced va<iation in morbidity, mortality, and infection load [24,44]. This study shows that the host pathogen interaction is complex in the amphibian-Bel system. Among the three species examined in this study, survival and infection load were not always correlated with the magnitude of our pathogen exposure regimes (i.e., level of zoospore dosage at the low, intermediate, or high level). In one species, although the rate of mortality increased with the magnitude of pathogen exposure, the host infection load did not change with increased pathogen exposure. Conversely, a treatment-level increase of infection load in larval and postmetamorphic hosts was not always associated with greater mortality over time. In our study, larval infection load cliffcrccl among species and/ or treatments, yet an increase in infection did not translate into increased mortality in the larvae of any species. This is in contrast to previous studies that have shown mortality at the larval stage for some species (e.g., [23,35]), although other studies have shown low, or no mortality at the larval stage L 45 J).
Direct comparisons to these studies arc diflicult because of dilfcrenccs in methodology. Thus, at both the larval and postmetamorphic stage, individual host responses play a central role in determining the outcome of this host-pathogen interaction.
The possibility of a "thre;hold" zoospore load, above which mortality occurs L'l6] was only evident in one species, P. regilla. In this species, average infection load in the high treatment group was PLOS ONE I www.plosone.org 8 Host Variation in Responses to Infectious Disease much higher than for any other species/treatment combination; it was greater than an order of magnitude higher in this species than the "least infected" species (R. cascadae). These results suggest that post-metamorphic P. regilla may be an important source of or reservoir lor Bel in their environments [47]. For example, we did not sec a dilferencc in mortality above background control mortality levels in the low inoculation dose treatment for P. regil!a, but infection loads at this level of exposure were high compared to low close infection loads for the other species examined ( Table 2, clay 8 average genome equivalents for P. regilla = 73.22 SD = 71.13). Thus, P. regilla may be tolerant oflow level infections, but. not necessarily resiltant to the pathogen (i.e., not effective at preventing pathogen replication in their skin). P. regilla has been cited as a potential "reservoir" species for Bel in other geographic contexts [47] and our experimental results support this hypothesis. This species is a ubiquitous generalist species in coastal, valley, and montane sites across \:Vestern North America, where many amphibian assemblages may be exposed to Bel via this species' "reservoir". Consequently, P. regilla may have a severe impact on the spread and maintenance of infection across a broad geographic area.
Responses to infection in the other two species that we examined were not close dependent. For these species, there may be a much lower infection load above which mortality occurs. Or, there may be environmental or physiological mechanisms that prevent this threshold of infection from being reached. In R. cascadae, the rate of mortality increased with the magnitude of pathogen exposure (i.e., treatment), however, absolute mortality in R. cascadae was low compared to the two other species examined (Fig. 2). Although there was a significant effect of the pathogen on rate of survival in R. cascadae, infection load was not different among treatments at either sampling time-point for this species. Rana cascadae always showed the lowest infection loads (at both larval and post-metamorphic stages), the lowest percentage of Belpositive animals, and the least amount of change in infection load over time. The reduced overall mortality and low infection loads suggest possible tolerance and resistance mechanisms may be at play in post-metamorphic R. cascadae. That is, individuals may be resisting infection, minimizing pathogen replication, or tolerating infection by minimizing the damage caused by the pathogen (i.e., mortality) across treatments [47,48,49]. Infection loads in larval R. cascadae on day 15 were comparable to infection loads of postmetamorphic animals of the same species. \'Ve did not observe this trend in the two other species, likely because infection is generally much higher when the amount of area the pathogen can colonize (the metamorphic skin versus larval mouthparts) was greater. The results for R. cascadae contrast to what we observed in A. boreas. For this species, the rate of mortality was the same regardless of pathogen treatment; even though animals were exposed to 10,000 versus 100,000 zoospores in the low versus the high pathogen treatment group, mortality over the course of the experiment was the same. That is, if A. boreas was exposed to the pathogen, they appeared to suffer pathogen-induced mortality, regardless of infectious close (i.e., they were equally "intolerant" of the pathogen [48,49]). However, this :;amc species showed significant differences in infection load accrued on the skin; infection loads were greater in the higher pathogen inoculation treatments. \<\ 1 e note that survival trends in A. boreas, more than either of the other two species, appeared to be temporally variable; mortality among pathogen treatments pre-day I 0 of the 15 day experiment was overlapping and similar (Figure 2.c.). However, post-day 10, there was a clear separation of mortality trends between the more and less severe (high and low) inoculation treatments. This trend may suggest latency to respond in this species, and therefore, potential for the pathogen to get a foot-hold before the host has a chance to respond or combat infection.
Of importance is the temporal response to infection in larval and post-metamorphic frogs. In larvae of all species tested, infection loads in the keratinized mouthparts of tadpoles were greater on day 15 versus day 8. However, temporal responses to infection varied among each species post-mctam01vhusis. For post-mctamoq)hic P. regilla, infection load decreased over time. In A. boreas, infection load increased from day 8 to day 15. In R. cascadae, infection load stayed the same over time (with the exception of individuals in the low treatment group where infection decreased over time). These three clillerent temporal responses have implications for host-maintenance of the pathogen in their environment but also imply possible clillerenccs in the mechanisms by which hosts respond to and deal with infection. A decrease in infection load over time may simply be a consequence of animals with the highest infection load being removed from sampling through death (with the least heavily infected animals sampled on clay 15). However, it is also possible that behavioral or physiological defenses provided a way to respond to or avoid the pathogen. Although not specifically quantified, we observed that P. regilla spent must of the time out of direct contact with water and zoospores, instead, climbing the sides and tops of their experimental containers. This behavioral response may haye limited their contact with infectious zoospores, and decreased infection load on postmctamoqJhic animals over time. One study suggests that the skin architecture of P. regilla may allow it to spatially separate areas of infection across the skin, and minimize the damage caused by infection [47J. The increasing infection load observed in post -metamorphic A. boreas could be a result of the lack of physiological or immunological mechanisms that impair or reduce infection load. Overall low infection loads were observed in both larval and post-metamorphic R. crHcadae. Perhaps this species keeps infection "in check" with active cellular or humoral responses that eiTcctivcly kill the pathogen or render it unable to replicate.
As shown in previous studies [50J, A. boreas, harbored the highest infection load in larval mouthparts alter initial pathogen inoculation (although, interestingly, this was the smallest species in size as larvae and thus, had a smaller area for the pathogen to infect). This suggests stage specific or life history variation in the potential for species to serve as pathogen reservoirs; amphibians have cliiTercnt ciTccts on pathogen dynamics, dependent on species and stage. In P. regilla, as in all species examined al the larval stages, infection load increased over time. This was suq)rising since no re-inoculation took place between the first and second sampling time-points. In anuran larvae, infection is restricted to the keratinized mouthparts (toothrows) and can cause sublethal ciTccts, including reduced foraging capacity [34J. We did not observe significant mortality among pathogen treatments at the larval stage in this or any of the other amphibian species, and so, we hypothesize that a mortality ciTect may be delayed until infection load reaches higher levels in the mouthparts, and begins to interfere with growth and or development. Infection load may accrue, uninhibited in tadpoles; perhaps the lucalizccl nature of infection precludes immunological responses to diminish infection load and infection therefore increases unabated.
Scaling from the individual to the ecological level, we emphasize the need to understand the relative contribution of species-level responses in shaping the composition and persistence of ecological assemblages. In addition to the ability of hosts to accumulate infection and persist over time with infection, there arc important life history traits (lifespan, aggregation behavior, breeding phenology, rnigratory patterns: aquatic tendency, ctc.:l and density PLOS ONE I www.plosone.org 9 Host Variation in Responses to Infectious Disease dependent factors that could interact to aiTcct how the pathogen is spread and maintained over space and time in any given system [46,5! ,52J. For example, schooling of larval A. boreas in groups of thousands of individuals and synchronous metamorphosis in both R. cascadae and A. borws results in high densities of individuals that may facilitate transmission of pathogens [.'i3J. Furthermore, environmental factors may reduce or amplify the relative contribution of each host to pathogen systems. For example, temperature may interact with host responses to inJection [54J; immunological changes occurring at warmer or cooler temperatures could promote or reduce pathogen replication in mouthparts of skin of larval or post-metamoqJhic amphibians or change patterns of host resistance against infection [54J. Similarly, temperatures that promote host shedding of the pathogen may speed up recovery time or allow persistence of the host with continuous low level inJection. vV e emphasize the need for "grouncl-truthing" the laboratmy results found here. Specifically, we encourage a quantitative assessment of infection loads in individual species existing in diiTerent combinations with other species in ecological assemblages across landscapes. Macro-and microhabitat variation represent imponant source~. of covariatcs in such analyses and important correlations may exist between biotic and abiotic factors and disease dynamics in the amphibian-chytricl fungus system [25,26,27]. Ultimately, models predicting population declines and species extinctions must be parameterized with information about transmission, mortality, and recovery of susceptible individual in populations. Fundamental to understanding such complexity is a fine-scale and carefully standardized experimental approach to examining individual, species, and population level cliflcrences in host responses to infection.   L =low dose. NS =non-significant comparisons; MS =marginally significant comparison (p<O.l ). NA indicates that the comparison is nut applicable because of low sample size (due to mortality or Bel-negative animals that could not be included in analyses). *Only I PR remained alive by day !5 of the cxpcritncnt and so was excluded frcnn statistical analyses. **Only 1 AB remained alive by day 15 of the experiment and so was excluded !rom statistical analyses. ***No among species comparison on clay 15 possible because only I individual alive in PR and AB.