Introduced bullfrogs are associated with increased Batrachochytrium dendrobatidis prevalence and reduced occurrence of Korean treefrogs

Bullfrogs, Lithobates catesbeianus, have been described as major vectors of the amphibian chytrid fungus, Batrachochytrium dendrobatidis (Bd). Bd is widespread throughout the range of amphibians yet varies considerably within and among populations in prevalence and host impact. In our study, the presence of L. catesbeianus is correlated with a 2.5 increase in Bd prevalence in treefrogs, and the endangered Dryophytes suweonensis displays a significantly higher Bd prevalence than the more abundant D. japonicus for the 37 sites surveyed. In addition, the occurrence of L. catesbeianus was significantly correlated with a decrease in presence of D. suweonensis at sites. We could not determine if it is the presence of bullfrogs as competitors or predators that is limiting the distribution of D. suweonensis or whether this is caused by bullfrogs acting as a reservoir for Bd. However, L. catesbeianus can now be added to the list of factors responsible for the decline of D. suweonensis populations.


Introduction
A third of all amphibian species are under threat of extinction and more than two hundred are specifically under threats because of enigmatic diseases and threats [1,2]. Evidence points to the chytrid fungus, Batrachochytrium dendrobatidis (Bd), as one of the major causes for population declines and extinction events [1][2][3]. However, Bd distribution is patchy, and its prevalence varies considerably with species susceptibility [4], latitude [5], temperature [6], and seasonality [7,8]. For instance, colder months are associated with higher prevalence [9,10] and mortality [11,12]. The average rainfall and humidity at a site are also of major importance, because Bd is dispersed by waterborne zoospores [6,13], and because all life stages of the fungus are sensitive to desiccation [14]. a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 Bd sampling Field sites were selected from previous studies on the presence of D. suweonensis [37,38]. Site selection and sample number of D. suweonensis were restricted by permits given by the Ministry of Environment of the Republic of Korea, to 88 individuals from 37 sites, covering the totality of D. suweonensis range (Fig 1). A maximum of 20% of the population size at a given site, surveyed in 2014, was allowed for sampling (Table 1). Sampling for D. japonicus targeted all individuals encountered within the same rice paddies as the D. suweonensis caught (Fig 1). Bd sampling was performed between 15 May and 28 June, 2015, during the peak breeding seasons of these two species ( [44]; S1 Table).
To detect Bd infection, individuals were first captured by rapidly placing a hand on the individual, covered with a single-use vinyl bag (Clean Wrap, # 365780167; Gimhae, Republic of Korea), then kept in separate bags to prevent cross-contamination and finally swabbed on the epidermis. Sampling was conducted at the breeding sites, between 4 pm and 4 am. Swabbing was done with sterile fine-tip swabs (Medical Wire & Equipment Co Ltd; Corsham, UK) wearing a new pair of vinyl gloves for each frog. Frogs were systematically swabbed five times on each toe of the hind legs, each foot, the inner thighs, and both sides of the abdomen. Since the two Hylid species are difficult to distinguish morphologically [45], buccal swabs from each individual were collected for species identification through mtDNA CO1 sequencing, following the protocol developed by Jang [46]. Swabs were stored in plastic tubes at -20˚C until DNA extraction. All individuals were released at the point of capture as soon as the site was sampled, i.e. within 2 h of capture.

Positive control for Bd detection
Since bullfrogs are a known reservoir for Bd in the Republic of Korea we first swabbed 7 individuals in order to obtain a positive control for PCR. The bullfrogs were obtained from irrigating ditches between rice-paddies in Hwaseong (37.155977˚N, 126.721290˚E), caught with butterfly types of nets, adequately sterilized and dried between sites, and swabbed following the Bd swabbing protocol described above. None of the individual caught displayed any sign of clinical disease or abnormality.
We obtained two positive reactions from two different individuals out of the seven L. catesbeianus swabbed following the protocol described below. The samples were then sent to Macrogen Inc. (Seoul, Republic of Korea) for direct sequencing in the forward direction on an ABI PRISM 3100 automatic sequencer (Applied Biosystems Inc.; USA). The sequences were manually trimmed in Geneious (v9.04, Biomatters Limited, Auckland, New Zealand) and blasted on the Basic Local Alignment Search Tool from the NCBI portal (http://blast.ncbi.nlm. nih.gov/Blast.cgi). One of our sequences was 100% identical to Batrachochytrium dendrobatidis accession number JX983045.1, collected from a L. catesbeianus [21]. This sample was used as positive control for all PCRs in our study, and maintained at -20˚C in 1.5 mL tubes.

DNA extraction and Bd detection
DNA was extracted following a modified version of the four-step protocol developed by Hyatt [47]. The protocol consisted in incubating each swab in 50 μL of Prep Man Ultra (Applied Biosystems; California, USA) at 100˚C for 10 min, before cooling at room temperature for 2 min, centrifugation at 13,000 rpm for 3 min, and transfer of the supernatant. Each sample was subsequently diluted 1:5 with PCR grade water before PCR. We used a nested PCR developed by Goka [48] for Bd detection. This reaction consisted of two PCRs that were the same except for the primers. The PCR1 used the primers Bd18SF1 (5'-TTTGTACACACCGCCCGTCGC-3') and Bd28SR1 (5'-ATATGCTTAAGTTCAGCGGG-3') from Goka [48] while the PCR2 used the primers Bd1a (5'-CAGTGTGCCATATGTCACG-3') and Bd2a (5'-CATGGTTCATATCT GTCCAG-3') from Annis [49]. Each PCR contained 12.8 μL of distilled water, 2.0 μL of (10x) buffer, 1.  All samples in this study were considered positive if a band was visible at the same size as the positive control (at approximatively 300 bp). If any of the negative controls displayed a band, all PCR results from the run were discarded and all reactions ran again.
Each treefrog sample was run in two independent PCR replicates along with positive (known Bd sample) and negative controls (PCR master-mix and distilled water). A sample was considered positive if both replicates were positive. Two samples for D. suweonensis and one for D. japonicus (all males, respectively from localities 15, 20 and 29; Table 1) were positive for one of the replicates only and were consequently run a third time. All three samples were positive for the third replicate and thus considered positive.

Environmental variables and Bd infection
To assess the impact of the environment on Bd prevalence, we collected the average temperature (˚C) and precipitation (mm) for the 3 months prior to sampling at each site ( Table 2). The data were collected from the closest weather stations of the National Weather Service Stations of Korea (http://www.kma.go.kr/weather/observation/aws_table_popup.jsp). Weather data on the days of sampling were not included in the analysis due to a high correlation with the 3-month average (Pearson correlation; r = 0.84; P < 0.001). The other variables mentioned.

Aural surveys
We conducted presence/absence transect surveys to examine the impact of L. catesbeianus on the distribution of D. suweonensis for the entire reported range of D. suweonensis (Borzée et al., in prep). Dryophytes japonicus was not included in this analysis due to its ubiquitous presence at all sites studied.
We used preliminary data on the distribution of D. suweonensis collected from 2013 and 2014 to select sites that fulfilled the ecological requirements for this species throughout its whole range (see [37]). This resulted in the selection of 179 sites with a high probability of D. suweonensis occurrence. All sites selected were below 35 m a.s.l. and located in large plains managed for rice cultivation. Another common landscape feature for the occurrence of D. suweonensis among sites was the relatively low abundance of man-made structures (e.g. houses, factories, greenhouses). Auditory surveys were conducted for both D. suweonensis and L. catesbeianus, using the method described by Borzée and Jang [38]. Surveys were conducted during the weeks of highest calling activity, between 15 May and 21 June 2015, and during the time of day when the two species are known to be active, between 6 pm and midnight [37,38]. Each survey was 10 min long and conducted while walking along the longest straight line available in rice paddies. Transect selection was facilitated by the grid structure of modern paddy fields. We tested whether 10 min transects accurately estimate the presence/absence of D. suweonensis through 3 replicate samplings at 10 sites. No significant variations in occurrence were found between replicates (repeated ANOVA; χ = 0.33; df = 2, P = 0.630), indicating that this method is effective at estimating the presence of D. suweonensis at a site.

Statistical analysis
Multicollinearity among all variables was examined prior to the statistical analysis through the use of Variance Inflation Factors (VIF [50]), instead of bivariate correlations to avoid simple pairwise comparisons of correlations [51]. Because VIF values were between 1.003 and 2.340, all variables were included in the subsequent analyses. To assess the factors important for Bd prevalence, we ran a univariate General Linear Model (GLM), with Bd infection as a dependent variable. The fixed factors were bullfrog presence, species, and sex; and the covariates were season, temperature, and precipitation for the three months prior to sampling. Species, sex, and Bd occurrence were binary encoded. Species was either D. suweonensis or D. japonicus. Season was defined as the number of days after 5 May 2015, the first day for both treefrog species to produce advertisement calls in 2015. We then ran a pair of ANOVAs in order to determine statistical significance for variations in Bd infection between (1) sites with and without bullfrogs, and (2) the two Hylid species. Prior to the analysis the homogeneity of variance was tested through Levene's test (P > 0.016), which was subsequently ignored as marginally influential to the data analysis. The data presented relies on the calculation of averages and frequencies to indicate the directionalities in prevalence. Finally, we assessed the impact of L. catesbeianus on the distribution of D. suweonensis through Fisher's exact test. All analyses were conducted with SPSS (v. 21.0, SPSS, Inc., Chicago, IL, USA).
The results of the GLM indicated that only two variables, bullfrog presence/absence and species, were significant predictors of Bd infection (F ! 45.74; P < 0.001; Table 2). In contrast, sex, season, temperature, and precipitation (averaged for the 3 months prior to sampling), were not significant in our model.

Dryophytes suweonensis distribution
In addition to studying the impact of Bd across a subset of D. suweonensis populations, we also investigated how the presence of bullfrogs influenced the occurrence of D. suweonensis at 179 sites (Fig 1). The results of our field surveys showed that D. suweonensis was present at 114 sites and L. catesbeianus at 22 sites. Dryophytes suweonensis was only present at 9 of the 22 sites with L. catesbeianus ( Table 3). The presence of D. suweonensis was significantly negatively related with the occurrence of L. catesbeianus (Fisher's exact test; χ 2 = 5.31, df = 1, P = 0.031; Phi = -0.17, P = 0.018). This finding indicates that D. suweonensis is less likely to occur at sites where L. catesbeianus is found.

Discussion
We show that Bd prevalence in Korean treefrogs was highest at sites with Lithobates catesbeianus, suggesting that this species is a vector or reservoir of the fungal pathogen. We also found a negative correlation between the presence of bullfrogs and occurrence of Dryophytes suweonensis. Finally, we found that Bd prevalence was much higher for D. suweonensis than it was for D. japonicus. These findings suggest that either Bd infection or L. catesbeianus, or possibly both factors together, may play a role in population decline of D. suweonensis, thus hastening the possible loss of this species.
Although Bd prevalence was high in D. suweonensis, it is unknown if Bd directly causes declines in this species as mass mortality events such as those reported in other Bd infected species [2] were not observed in D. suweonensis during > 300 days of field work conducted between 2013 and 2015. It appears that Bd may not be causing declines in this species, although sub-lethal effects of Bd (e.g. decreased growth, [52,53]) cannot be ruled out. It is possible that both treefrog species are tolerant to Bd infection. Further studies including Bd challenge experiments in the laboratory and long term monitoring in the field are necessary. Long term monitoring would also resolve potential false-negative Bd prevalence due to low sample sizes, which could have introduced bias if Bd had not been detected at more than the seven sites from this study.
One explanation for the difference in Bd prevalence between the two treefrog species is that D. suweonensis might be immunocompromised due to low genetic diversity. Dryophytes suweonensis is an endangered species with a restricted range and a low genetic diversity (Borzée et al., unpublished). Thus, we expected that the genetic bottleneck associated with this species would have had a negative impact on immunity, notably through a loss of variability in the Major Histocompatibility Complex (MHC), which has been associated with Bd resistance in other amphibian species (Bataille et al. 2015, Savage et al. 2011Savage et al. , 2016. The variation in Bd prevalence between the two treefrog species may also be explained by the differences in habitat use by the two species. During the breeding season, male D. suweonensis produce advertisement calls at the center of flooded rice paddies, whereas male D. japonicus call from the banks of rice paddies, outside of water [54]. Such a niche segregation between the two species results in D. suweonensis being more closely associated with water than D. japonicus. Other studies have shown evidence of an association between increased Bd prevalence and association with water [55]. In addition, D. suweonensis hibernates on the bank of the rice paddies, while D. japonicus hibernates in forests (Borzée et al., unpublished). Differential uses of habitats between the two treefrog species can lead to a higher prevalence of Bd for D. suweonensis than for D. japonicus. Future studies should examine seasonal variations in Bd prevalence and intensity, through quantitative PCR analysis, potentially demonstrating a variation in mortality depending on seasons or during metamorphosis [56]. The difference in Bd prevalence between D. japonicus and D. suweonensis could also lie in behavioral differences. Some species have been shown to behaviorally reduce the chance of Bd infection by positively selecting Bd-free microhabitats [57], or displaying thermoregulatory behavior, such as basking, e.g. Atelopus zeteki [58]. Dryophytes suweonensis is active at higher temperatures than D. japonicus [54], but it is not known whether these two species differ in basking behavior. Thus, despite D. suweonensis being active at higher temperature, a more pronounced basking by D. japonicus could lead to the difference in Bd prevalence between the two species.
The negative impact of L. catesbeianus on D. suweonensis may result from a combination of several factors. Lithobates catesbeianus is a known vector of Bd [26, 59,60], but it also preys on smaller species [61], using male calls as a detection beacon [62]. It has also been shown to predate on other Korean species, such as Pelophylax chosenicus [32,63] and D. japonicus (A. Borzée personal observation, June 2015; 36.137594˚N, 127.381514˚E). Thus, D. suweonensis may be avoiding sites where L. catesbeianus occurs. Alternatively, it could also be possible for males to be quiet at sites where L. catesbeianus is present in order to avoid predation. This would result in a larger population of D. suweonensis than recorded, although it would not last longer than the generation length of the species as females would be less efficient at locating males for breeding.
Alternatively, bullfrogs have been shown to compete with other species for food resources, during both larval [64,65] and adult stages [66]. Lithobates catesbeianus breeds in ditches and deeper aquatic environments than in rice paddies where D. suweonensis breeds. Due to this apparent segregation in habitat, exploitative competition based on resource exploitation between D. suweonensis and L. catesbeianus seems unlikely as the two species use different food resources and egg deposition sites.
Our research highlights that the endangered D. suweonensis is currently at risk of further population declines with the expansion of the invasive L. catesbeianus. It is of prime importance to prevent further expansion of the invasive L. catesbeianus, and to remove individuals already present. We recommend an increase in governmental policies that offer financial incentives for the capture of bullfrogs.