Introduction

Infectious disease is a well-known driver of animal declines worldwide (e.g. [1–2]). Ectothermic vertebrates, particularly reptiles and amphibians, have been exceptionally impacted by emerging infectious diseases (EIDs) [1]. Observations of massive, worldwide, pathogen-associated amphibian die-offs date back to the 1970s and 80s [3]. In North America, amphibians have been impacted by two major emerging pathogens, the fungus Batrachochytrium dendrobatidis (Bd) and a class of iridoviruses collectively known as Ranavirus (Rv) [4–7]. Bd was identified as a disease agent in 1998 [8] and Rv was classified as an emerging pathogen after 1993 [9–10]. In North America, Bd has mainly impacted anurans in the families Ranidae and Bufonidae, causing mass mortality events for species in these families. Most infections have occurred in the western United States and have impacted already threatened and endangered species (e.g. [1,4, 11–12]). Rv has caused die-offs in at least 20 different reptile and amphibian species throughout the United States, including species under state and federal protection [4, 13–14].

In the southeastern United States (hereafter, Southeast), emerging pathogens impacting ectothermic vertebrates have been poorly characterized but have the potential to significantly impact the area. The Southeast is exceptionally rich in amphibian and reptile diversity, hosting more than half of species occurring in the United States [15]. EIDs therefore have potential to severely impact this region’s ecosystems and overall biodiversity. Presently, EIDs have been documented in multiple amphibian and reptile groups occurring in the Southeast [4, 13, 16], yet disease monitoring has been limited. One species, the Gopher frog (Lithobates capito), is susceptible to Rv infections in the lab [14] and infection and die-offs have been documented for wild populations [16–17]. However, impacts of pathogens on other species in this area are more enigmatic [13, 17–19]. Because disease monitoring has been limited, museum specimens and other archived biological samples are critical for retrospective pathogen detection and can aid in uncovering when pathogens were first introduced and where they were found in the past [20].

Climatic variables are significant drivers of pathogen prevalence in wildlife populations, and amphibians and their pathogens are no exception (e.g. [7, 21–22]). Temperature and precipitation are the two major environmental factors that appear to be drivers of Bd dynamics. Studies have shown a negative relationship between temperature and Bd occurrence, prevalence and intensity both in the lab [23–24] and in natural populations (e.g. [7, 25–27]). Additionally, variation in precipitation and humidity have been implicated in the occurrence and prevalence of Bd, with increased precipitation and humidity driving patterns (e.g. [23, 28]). Most work to date on Rv has focused on documenting infections. To our knowledge, there is a paucity of work investigating how environmental factors serve as drivers of Rv. Only two studies look at this pattern; one in which salamanders were experimentally infected under different temperature regimes and a negative relationship between Rv infection intensity and temperature was found [29] and one testing ranid frogs under differing temperature regimes [30].

There is increasing empirical evidence for a genetic basis to disease resistance in wild vertebrate populations [31–32], although more studies are needed to test this hypothesis in amphibian taxa [33]. To date, reduced genetic diversity has been found to increase susceptibility to one pathogen (Rv) in a single amphibian species based on neutral microsatellite loci [34]. Additionally, a handful of studies have explored the underlying genetic basis for disease resistance to Bd. Tobler and Schmidt [35] looked at among-population susceptibility to Bd in a European frog and inferred that differences in susceptibility among populations had a genetic basis due to differential population responses in a common garden experiment. Similarly, Savage et al. [36] found that neutral genetic diversity was negatively correlated with Bd infection prevalence in a North American frog. Immunogenetic analyses have also found significant associations between specific major histocompatibility complex (MHC) class II alleles and Bd tolerance in the lab and in natural systems for multiple species of anurans [37–39]. These studies suggest that host genetic diversity underlies differential amphibian population responses to EIDs, but are based on a limited number of taxa.

Despite the importance of environmental and genetic factors in explaining amphibian disease dynamics when investigated separately, few studies incorporate both into a single analysis. Ribas et al. [24] demonstrated in the lab that both temperature and expression of skin peptides determined how anuran hosts responded to Bd infection. Savage et al. [39] found that environmental factors were responsible for predicting Bd infection intensity, while both genetic and environmental factors influenced Bd prevalence. This was the first analysis to combine both genetic and environmental factors in a predictive model for EIDs in a natural amphibian system. More studies are needed to confirm the importance of both genetic and environmental factors for explaining amphibian EIDs.

Here, we assess whether EIDs are impacting amphibian populations in the southeastern United States by characterizing Bd and Rv dynamics using archived samples from three amphibian species: a salamander, Notophthalmus perstriatus and two tree frogs, Hyla squirella and Pseudacris ornata. Each species exhibits unique life history traits and all three occur throughout the Southeast, offering varying perspectives into pathogen dynamics among diverse taxa. Notopthalmus perstriatus has a complex life cycle, spending two out of its three life stages in the water [40]. Previous studies have observed that N. perstriatus and other Notophthalmus are susceptible to both Bd [19] and Rv [41]. Hyla squirella is a common, highly arboreal species that breeds in large aggregates during summer months (Elliot et al. 2009), and little is known about the presence or impact of Bd and Rv in this species. In contrast, P. ornata is an increasingly uncommon species (B. Means pers. comm.), breeds in winter at low densities [42], and is highly susceptible to Rv in experimental infection trials [43]. Field studies of P. ornata tadpoles in northern Florida have also confirmed Rv infections in the wild [17]. Previous population genetic analyses found that N. perstriatus form distinct East-West groups that do not share haplotypes [40], H. squirella genetic structure is heavily determined by habitat structure [44], and P. ornata genetic structure varies among populations and the species may have been widespread across the Southeast in the past [45]. By combining these genetic data with environmental and disease variables for each sampled population, we simultaneously assessed the importance of host genetics and environmental variables on predicting disease impact and spread in amphibian populations of the Southeast.

Materials and methods

Results

We tested 401 N. perstriatus tissue samples from 11 populations, none of which were infected with Bd or Rv. Among 580 H. squirella tissue samples from 20 populations, one individual was infected with Bd (Table C in S1 File; Fig 1a) and one individual from a different population tested positive for Rv (Table C in S1 File; Fig 1b). Neither Bd nor Rv prevalence were significantly different among H. squirella populations (Fisher Exact test P = 0.64). Finally, among 327 P. ornata tissue samples from 15 populations, 103 individuals from 10 populations were Bd positive (Table A in S1 File; Fig 1c), whereas none tested positive for Rv (Table C in S1 File; Fig 1d). Bd prevalence among infected P. ornata populations ranged from 0.24 to 1.0 (Table C in S1 File). Bd prevalence varied significantly among P. ornata populations (Fisher Exact test, P = 0.0005). Additionally, pooled Bd prevalence was significantly different in P. ornata compared to H. squirella (Fisher Exact test, P < 0.00001). Average Bd infection intensity among infected populations ranged from 226 to almost eighteen million genome equivalents (GE; Fig 2). Of the 11 infected populations, nine harbored average infections of over 10,000 GE (Table C in S1 File; Fig 2). All but one infected populations had 40% of individuals harboring loads above 10,000 GE and two populations had 100% of individuals with infections above 10,000 GE. Bd infection intensity was significantly different among infected P. ornata populations (two-way ANOVA, P < 0.00001).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Map of sample localities and (A) Bd prevalence in Hyla squirella, (B) Rv prevalence in H. squirella, (C) Bd prevalence in Pseudacris ornata and (D) Rv prevalence in P. ornate.

Circle size is relative to population size. Arrows point to infected H. squirella populations. Green represents proportion of negative cases of the indicated pathogen while red represents proportion of positive cases.

https://doi.org/10.1371/journal.pone.0175843.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Log-transformed Bd average intensities for sampled populations of H. squirella and P. ornata with standard error of the mean (SEM).

White bars indicate average intensity among infected individuals from H. squirella populations and black bars denote average intensity among infected individuals from P. ornata populations. The red line marks an infection intensity of 10,000 GE.

https://doi.org/10.1371/journal.pone.0175843.g002

Because P. ornata was the only species harboring pathogens in more than one population, and Bd was the only pathogen detected, we limited GLM analyses to Bd dynamics within P. ornata. PCA of environmental variables across sampled P. ornata populations revealed that the first two components explained 89% of variation in our data (Fig A in S1 File). PC1 is associated with decreasing temperature and some increasing precipitation variables and positively associated with latitude. PC2 is positively associated with winter precipitation. We assessed only additive models with AICs for prevalence and intensity as interactive models showed high variance inflation. We chose the most informative models based on the lowest AIC score in the set (Table 1). PC1, PC2 and average heterozygosity significantly influenced Bd prevalence (Fig 3; Table D in S1 File). Based on our general linear models, there is a significant relationship between decreasing temperature and an increase in Bd prevalence (P = 0.0000002; Fig 3A). Moreover, there is a significant negative relationship between winter precipitation and Bd prevalence (P = 0.000591; Fig 3B). Surprisingly, our model identified a significant relationship between increased average heterozygosity and increased Bd prevalence (P = 0.0000698; Fig 3C). Environmental factors (PC1) significantly influenced Bd intensity (Fig B and Table E in S1 File). Linear regression displayed a significant relationship between a decrease in temperature and an increase of Bd intensity (P = 0.011; Fig B in S1 File).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Five most informative general linear models and the null models for Bd prevalence and Bd intensity.

https://doi.org/10.1371/journal.pone.0175843.t001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Logistic relationships between Bd prevalence and PC1 (A), PC2 (B), and average heterozygosity (C) in populations of P. ornata.

Increasing values of PC1 correspond most strongly to decreasing values of temperature. Increasing values of PC2 correspond most strongly to increasing winter precipitation. In all panels, each dot corresponds to the observed prevalence in a population and each line corresponds to the best fit logistic model of the relationship between the two variables shown.

https://doi.org/10.1371/journal.pone.0175843.g003

Discussion

Our study used both environmental and genetic variables to create a predictive model for chytridiomycosis disease dynamics in the southeastern United States. We identified both concordant and discordant patterns of pathogen prevalence and infection intensity in N. perstriatus, H. squirella, and P. ornata compared to previous studies [18–19, 54]. First, our data showed limited Rv occurrence and high variation in Bd infection prevalence within and among our three focal species. Second, we found surprisingly high Bd infection intensity in P. ornata and H. squirella, a strikingly different result compared to the extremely low intensities previously detected [54]. Finally, our overall model found that both genetic and environmental variables predict Bd prevalence but only environmental variables predict infection intensity. These results are important for understanding the enigmatic story of Bd infection in this region, and provide a framework for future management of declining amphibians in the Southeast. Overall, this study offers insight into pathogen infection history and amphibian disease dynamics in the southeastern United States, a hotspot of amphibian diversity.

Our three focal amphibian species showed largely unique patterns of Bd and Rv infection prevalence compared to previous studies on the same or similar species. Interestingly, Rv was previously found within ponds where N. perstriatus populations occur [17], and a closely related Notophtahlmus species in the Southeast, N. viridescens, was found to have high Bd infection prevalence within sampled populations [19]. In contrast, we found no evidence of Bd or Rv in N. perstriatus, suggesting the presence of genetic, temporal or seasonal differences for our sampled populations. Hyla squirella infection dynamics were also surprising as this species breeds in large aggregates in habitats frequented by known Bd and Rv vector species, particularly Lithobates catesbianus [42, 55]. We predicted high pathogen prevalence in H. squirella, but instead only found two infected individuals (one Bd infected and one Rv infected) among all sampled populations. Interestingly, these are the first documented Bd and Rv infections in H. squirella despite previous sampling efforts [19]. These results indicate low overall infection prevalence in H. squirella, but the high Bd infection intensity we detected in one individual suggests the potential for negative impacts on populations where infection does occur. The only focal species with high pathogen prevalence was P. ornata, which exhibited strikingly different infection patterns for Rv and Bd compared to H. squirella. While Rv infection was undetectable in P. ornata, high Bd infection prevalence and intensity occurred in the majority of sampled populations. These results mirror previous studies testing P. ornata and other members of Pseudacris for pathogens in the Southeast [19, 54].

Life history variation among amphibian species often contributes to variation in pathogen infection prevalence [56]. The distinct life histories among our three study species may be an underlying factor contributing to the significant differences in Bd prevalence and intensity that we observed. In particular, high susceptibility to Bd infection in P. ornata may be due to the trait of breeding exclusively during rainy periods in the winter months [42], unlike the other two species that are summer breeders. Our data, along with data from several other studies, correlate Bd infection with cooler temperatures and higher precipitation in winter months [7, 26–27, 56–57]. Thus, high contact rates among P. ornata individuals during winter breeding aggregations when Bd experiences preferred temperatures may be driving the observed high infection prevalence and intensity. Our results, combined with previous monitoring efforts, serve as a valuable baseline should infection outbreaks occur for other southeastern amphibian species and may help elucidate reasons for any enigmatic declines.

Our Bd infection intensity data seem to contradict previous studies suggesting that values greater than 10,000 GE lead to mortality, regardless of the amphibian species [58–59]. We found high Bd intensities that surpassed this threshold for the single infected H. squirella individual and within most infected P. ornata individuals (Fig 2). Although high Bd infection prevalence in Pseudacris populations is well documented in the literature [19–20, 54], only one other study quantified Bd intensity for P. ornata; they found low values (<102 GE) for all individuals sampled throughout the United States [54]. We uncovered a pattern that is much more extreme; average Bd intensities were millions of GE for all but three infected populations (Fig 2) despite the absence of any observed disease signs or mortality events (T. Hether pers. comm.). Our data therefore demonstrate that the 10,000 GE proposed mortality threshold is not a standard applicable to every species. Indeed, our findings reinforce recent Bd studies in Brazilian amphibian communities [60] and New York State amphibian communities [61] which showed varying infection intensities across a range of species and seasons without observing any mortality or disease signs in the individuals sampled.

To uncover the driving forces behind the observed patterns of Bd infection among our sampled P. ornata populations, we incorporated climatic variables and genetic diversity factors into a comprehensive model. Bd intensity increased at lower air temperatures, consistent with similar analyses in other species and regions (e.g. [27, 36, 56, 62]. Further, the same variables that influenced Bd prevalence for P. ornata in our study (temperature, precipitation and average heterozygosity) also explained Bd prevalence in L. yavapaiensis in Arizona [36]. However, in contrast to the negative correlation between average heterozygosity and Bd prevalence found for L. yavapaiensis, we found average heterozygosity was positively correlated with Bd prevalence for P. ornata (Fig 3C). This pattern is in direct contrast with expectations, as numerous studies across wildlife disease systems have found higher genetic diversity within populations leads to decreased infection prevalence and increased disease resistance (e.g. [34, 63–64]. Our models show that for P. ornata, the opposite is true: increased genetic diversity correlates positively with Bd prevalence. Two possible explanations exist for this pattern. First, because average heterozygosity increases with larger effective population size, there could be better facilitation of pathogen spread due to density-dependent disease outbreaks [65–66]. This explanation is unlikely for P. ornata, however, as this species is generally uncommon and has historically small population sizes [42]. Another possible explanation is that Bd swept through P. ornata populations before our sampling occurred, and selection favoring Bd tolerant individuals was strong enough to push tolerant genotypes towards fixation, resulting in decreased heterozygosity. Bd has been present in the United States long enough to make this “genetic purging” [67] scenario plausible; Ouelett et al. [20] and Talley et al. [68] found evidence of Bd infections existing in North America as far back as the late 1800s. Our results could thus represent indirect evidence of genetic tolerance to Bd evolving in natural amphibian populations [39], although further genetic sampling, molecular tests of selection and experimental evidence of Bd tolerance are necessary to resolve this hypothesis.

Our data strongly suggest that both genetic and environmental factors should be incorporated, when possible, into models when trying to predict dynamics of infectious pathogens in natural populations. Management plans often only consider genetic or environmental factors when planning for long-term species persistence, but it is becoming increasingly clear that both are important for predicting pathogen impacts. While our study only focuses on amphibians, this modeling framework is applicable and important for other wildlife disease systems. Our results also suggest that Bd may be more of a concern for the Southeast than previously thought, at least for some species. There are no documented instances of disease-driven morbidity or mortality in P. ornata, yet a majority of sampled individuals were heavily infected with Bd and there is evidence of population declines in recent decades (B. Means pers.comm.). Cryptic chytridiomycosis may therefore be an unobserved but causal factor behind population declines and patterns of genetic diversity. Alternately, the Bd strain(s) present in the Southeast may currently exist as commensals or sub-lethal pathogens in P. ornata. Even under the latter scenario, monitoring Bd and other pathogen dynamics is important for future P. ornata conservation efforts. If novel biotic or abiotic stressors appear, the additional toll of harboring massive Bd intensities may be a tipping point towards extirpation. This may be especially true for populations with low genetic diversity, even if that loss of diversity is due to selection for pathogen tolerance. Means and Means (unpublished data) highlight that habitat destruction and degradation are threatening P. ornata population persistence, and more recently Means et al. [17] found wild P. ornata tadpoles to be heavily infected with Rv. Our results suggest Bd is a threat for adult frogs, particularly if the same populations are impacted by Rv prior to metamorphosis.

Overall, our study highlights how species are differentially impacted by EIDs in the Southeast and how models can be used to infer which environmental and genetic factors are drivers of infection. Infectious disease is often implicated in amphibian population declines only after morbidity and mortality is observed, making the trigger for a disease outbreak difficult to determine retrospectively. It has therefore become increasingly important to characterize and monitor species that have yet to display signs of disease in order to generate a baseline of pathogen dynamics should any future disease outbreak occur. Museum collections and specimens collected for non-disease studies are invaluable for assessing conditions faced by amphibians in the past [20], and here we utilized these resources to document the presence of two infectious pathogens in two frog species without any prior evidence of disease. Whether ubiquitous Bd infections in P. ornata reflect post-epidemic adaptation, non-pathogenic Bd strains, or virulent, ongoing chytridiomycosis that has gone undetected will require additional analyses. Regardless, our modeling results highlight the combined importance of host genetic variation and climate for determining Bd prevalence. Other climatic factors, such as seasonality, may also play a big part in disease dynamics and should be considered in future studies. Our results begin the journey to uncovering amphibian pathogen dynamics and can be used to develop more robust predictive models to assess where pathogens will likely spread and to inform species managers, as well as target suitable future re-introduction sites for amphibians that have been hit the hardest by disease-related declines.

Supporting information

Acknowledgments

We thank Zachary Rivas and Sanabel Mahmoud for providing assistance with lab work and Jacob Degner and Sarah May for providing tissue samples. We also thank Taina Torres for assistance with the prevalence figures. This research was supported by grants from the US Fish and Wildlife Service and Florida Fish and Wildlife Conservation Commission’s Nongame Wildlife Program to E.A.H., Sigma Xi Grants-in-Aid research #G200810150636 to T.D.H., and the University of Central Florida. All applicable institutional and/or national guidelines for the care and use of animals were followed.