Introduction
As part of the recent "biodiversity crisis," many amphibian populations are declining worldwide (e.g., Blaustein et al.
1994; Wake and Vredenburg
2008; Catenazzi
2015; Scheele et al.
2019). Indeed, the comprehensive analysis by Stuart et al. (
2004) indicated that about one-third of all amphibian species is threatened with extinction, while almost half are experiencing regional or local declines due to often interacting causes, such as habitat reduction, pollution, climate change and emerging diseases. Among pathogens, the recently described chytrid fungus
Batrachochytrium dendrobatidis (Longcore et al.
1999) is considered the main cause of population declines in different continents, driving species to extinction in Australia, South and Central America (Fisher and Garner
2020; Scheele et al.
2019).
Batrachochytrium dendrobatidis (Bd) is a highly virulent pathogen that infects the skin of all the three orders of amphibians [i.e. Anura (frogs), Caudata (salamanders) and Apoda (caecilians)], causing chytridiomycosis, a frequently lethal disease that produces immunosuppression, depletion of plasma electrolytes and cardiac electric dysfunctions in amphibians (Berger et al.
2004,
2016).
The global Bd occurrence and its effects on amphibian populations have been recently reviewed and mapped (Olson et al.
2013; Lötters et al.
2009; Scheele et al.
2019). To date, six different Bd lineages have been identified by multilocus sequence typing, but only the global pandemic lineage (GPL) seems associated to widespread chytridiomycosis outbreaks that caused populations declines (Fisher and Garner
2020). The main factor spreading the pandemic Bd lineage in different parts of the world and in different time periods is the international trade of amphibians and other aquatic animals for food, research, collection or company (Olson et al.
2013).
Various studies, in the last fifteen years, mapped the known distribution of Bd-infected amphibian populations and analysed the possible future consequences of this disease at the global or continental level (e.g. Ron
2005; Becker and Zamudio
2011; Doherty-Bone et al.
2020; Ribeiro et al.
2020). However, this kind of assessment is lacking at the regional level in the Mediterranean biogeographic area, a well-known hotspot of natural and human-adapted ecosystems (Blondel and Aronson
1999; Myers et al.
2000). In particular, the central Mediterranean region has been shown as a potential suitable area for Bd by a variety of global studies (e.g. Fig.
4 in Ron
2005; Fig.
3 in Lötters et al.
2009; Fig.
2 in Liu et al.
2013). At the centre of the Mediterranean basin, Italy represents one of the most relevant hotspots of biodiversity with a high concentration of amphibian species and in particular of endemics (Sindaco et al.
2006). In fact, due to its central geographic position within the Mediterranean, complex geological history, contrasted geomorphology, variable climates and the long-lasting coevolution between rural landscapes and wildlife (Cevasco et al.
2015), Italy hosts a highly diverse and unique amphibian fauna, comprising about half of all amphibian species described in Europe (Table
1; Temple and Cox
2009; Rondinini et al.
2013). Various local studies have been published on Bd occurrence in Italy (see Table
1), but up to now only one mass-mortality event has been observed in Sardinia (Bielby et al.
2013; Tessa et al.
2013). In this island the pandemic lineage GPL has been recorded (Fisher and Garner
2020). Besides, in some areas along the Apennine mountain range the observed decline of the Apennine yellow-bellied toad,
Bombina pachypus, was attributed to Bd infection (Stagni et al.
2004; Canestrelli et al.
2013). However, declines are also observed in other areas where Bd has been screened for, but is apparently absent (Canessa et al.
2013a,
2019). In these latter areas, the Apennine yellow-bellied toads’ declines were attributed to major habitat changes rather than pathogens (Canessa et al.
2013b). Therefore, the ecological effects of chytridiomycosis on the conservation status of Italian amphibians remain poorly understood, and in some cases enigmatic. For this reason, is of primary importance to delineate a nationwide monitoring and intervention plan, and model Bd occurrence to forecast possible future outbreaks.
Table 1
Conservation status of Italian amphibians and occurrence of Batrachochtrium dendrobatidis (Bd +) of Italian populations
Speleomantes ambrosii | Yes | NT | NT | II, IV | Resistent | | Not analysed |
Speleomantes flavus | Yes | VU | VU | II, IV | Resistent | | Not analysed |
Speleomantes genei | Yes | VU | LC | II, IV | Resistent | | Not analysed |
Speleomantes imperialis | Yes | NT | NT | II, IV | Resistent | | Not analysed |
Speleomantes italicus | Yes | LC | NT | IV | Resistent | | Not analysed |
Speleomantes sarrabusensis | Yes | VU | VU | II, IV | Resistent | | Not analysed |
Speleomantes strinatii | | LC | NT | II, IV | Resistent | | Not analysed |
Speleomantes supramontis | Yes | VU | EN | II, IV | Resistent | | Not analysed |
Proteus anguinus | | VU | VU | II, IV | | | Not analysed |
Euproctus platycephalus | Yes | EN | EN | II, IV | + | | 0% (3) |
Ichthyosaura alpestris | | LC | LC | | + | | 4% (124) |
Lissotriton italicus | Yes | LC | LC | IV | + | | 19% (77) |
Lissotriton vulgaris | | NT | LC | | + | | 9% (53) |
Salamandra atra | | LC | LC | IV | | | Not analysed |
Salamandra atra aurorae | | LC | LC | IIa, IV | | | 0% (3) |
Salamandra lanzai | | VU | VU | IV | | | 0% (1) |
Salamandra salamandra | | LC | LC | | + | | 2% (50) |
Salamandrina perspicillata | Yes | LC | LC | II, IV | | | Not analysed |
Salamandrina terdigitata | Yes | LC | LC | II, IV | + | | 23% (31) |
Triturus carnifex | | NT | LC | II, IV | + | | 2% (90) |
Discoglossus pictus | | LC | LC | | | | Not analysed |
Discoglossus sardus | | VU | LC | II, IV | + | | Not analysed |
Bombina pachypus | Yes | EN | EN | II, IV | + | (Stagni et al. 2004; Canestrelli et al. 2013) | 1% (412) |
Bombina variegata | | LC | LC | II, IV | | | Not analysed |
Bufo balearicus | | LC | LC | IV | | | 0% (46) |
Bufo boulengeri | | VU | LC | IV | | | Not analysed |
Bufo bufo | | VU | LC | | + | | 10% (84) |
Bufo siculus | Yes | LC | LC | | | | Not analysed |
Bufo viridis | | LC | LC | IV | | | Not analysed |
Hyla arborea | | n.e | LC | IV | | | Not analysed |
Hyla intermedia | | LC | LC | IV | + | | 100% (1) |
Hyla meridionalis | | LC | LC | IV | | | 0% (5) |
Hyla sarda | | LC | LC | IV | + | | Not analysed |
Pelobates fuscus insubricus | Yes | EN | LC | IIa, IV | | | Not analysed |
Pelodytes punctatus | | EN | LC | | | | Not analysed |
Pelophylax kl esculentus | | LC | LC | V | + | (Adams et al. 2004; Federici et al. 2008; Ficetola et al. 2011; Simoncelli et al. 2005) | 6% (159) |
Pelophylax lessonae | | LC | LC | IV | | | Not analysed |
Rana dalmatina | | LC | LC | IV | | | 0% (37) |
Rana italica | Yes | LC | LC | IV | + | | 13% (197) |
Rana latastei | | VU | VU | II, IV | | | Not analysed |
Rana temporaria | | LC | LC | V | | | 0% (1) |
In this research, we reviewed all the available information on Bd occurrence in Italian amphibian populations and we added original data, obtained over a five-year screening on peninsular Italy (Grasselli et al. unpublished data). Then we used both bibliographic and original data to model the present and near-future pathways of Bd diffusion in Italy, by both building habitat suitability models and ecological connectivity models at the landscape scale (McRae et al.
2008; Dickson et al
2019). This approach allowed us to forecast the most probable pathways of Bd diffusion and to better understand the impact of this disease on amphibian populations, within and outside protected areas, and to possibly identify the best mitigation measures to be implemented at the national and regional scale.
Discussion
In peninsular Italy, Bd is present all along the Apennine mountain chain and also in the Po plain, with an overall prevalence of 6%. When this value was compared with other European countries, with a reported sample larger than 1000 individuals, a similar prevalence was found in Germany (7%, in 3064 individuals) but a much higher one was observed in Spain (20%, in 1149 individuals; data from supplementary materials in Baláž et al.
2013). However, this latter high prevalence may be explained by an over-sampling of amphibians in areas with well-known Bd outbreaks, such as Central Spain and the island of Majorca (Bosch et al.
2001; Bosch and Martinez-Solano
2006; Walker et al.
2008; Baláž et al.
2013). Concerning the individual Bd load, the Italian data were in the lowest range of those reported by Baláž et al. (
2013) that measured up to 4067 GEs, but usually much lower.
The application of species distribution and habitat suitability models to the chytrid amphibian fungus has been widespread in the last 15 years. Several studies modelled the current distribution and habitat suitability of Bd at a global (e.g. Ron
2005; Rödder et al.
2009; Lötters et al.
2009; Liu et al.
2013; Xie et al.
2016) or continental scale (e.g. James et al.
2015; Rahman et al.
2020; Zimkus et al.
2020); while other studies investigated Bd distribution at a smaller, regional scale, focusing on specific amphibian diversity hot-spots (e.g. Puschendorf et al.
2009; Seimon et al.
2015; Flechas et al.
2017; Miller et al.
2018; Bacigalupe et al.
2019). However, ecological relationships among hosts and parasites are complex and the outcomes of their interactions vary in association with global environmental variables but also with the complexity of the biological community at the local scale (Benício et al.
2019, Halliday et al.
2019). Indeed, this relationship seems to be generally non-linear and that high biodiversity may dilute parasite occurrence (Halliday and Rohr
2019). This implies that highly diverse and rich ecosystems could inhibit the diffusion of wildlife diseases at the local scale. Therefore, the role of areas with complex amphibian communities could act as ecological barriers to Bd spread, and should become important areas for the monitoring of the chytrid fungus. Among the studies modelling Bd occurrence at a global scale, several identify Italy as a high-suitability/high-risk area for Bd (Rödder et al.
2009; Lötters et al.
2009; Liu et al.
2013). Within our study, we identified some specific area of high suitability for Bd in Italy (e.g. Sila, Pollino, Aspromonte National Parks), while other areas were predicted to be less suitable at the current climatic conditions (e.g. Foreste Casentinesi National Park). The major part of the studies inferring habitat suitability and distribution of Bd, highlighted an overwhelming effect of climate. For instance: Puschendorf et al. (
2009) found that high temperature seems to constrain the distribution of the pathogen at small scale in Costa Rica, Flechas et al. (
2017) identified mean temperature and precipitation seasonality as main drivers of Bd in Colombia, while Liu et al. (
2013) observed a relationship with annual temperature range at a global scale. In this study, among the bioclimatic variables included in the habitat suitability model, we observed that extremely high or low temperatures, in particular in the wettest quarter of the year, (BIO6 and BIO8) were main predictors of Bd suitability and acted as limiting conditions for his occurrence. These results can be explained by both an increased rate of epidermal renewal, driven by higher temperatures, which may in turn reduce Bd infection (Piotrowski et al.
2004), or alternatively producing physiological stress, which may limit Bd reproductive success (Piotrowski et al.
2004). Besides climate, also vegetation (Liu et al.
2013), land cover (James et al.
2015) and anthropogenic factors have been found to shape the distribution of Bd (Liu et al.
2013; Bacigalupe et al.
2019).
Habitat suitability and distribution models also confirmed the link between epizootic chytridiomycosis and amphibian worldwide decline, highlighting how areas of rapid amphibian decline overlaps with those of higher suitability for Bd at a global scale (Rödder et al.
2009; Lötters et al.
2009; James et al.
2015). The majority of studies modelling Bd distribution only focused on current climatic conditions, while few studies also projected distribution and suitability models on future climatic scenarios, predicting that Bd may decrease globally in some regions by 2100, but with a shift towards higher latitude and altitudes (Xie et al.
2016). This trend has also been confirmed at a smaller spatial scale (e.g. Seimon et al.
2015; Xie et al.
2016; Miller et al.
2018). In the present study we also observed a general contraction of Bd suitable areas by 2050, considering both circulation pathways. In particular, we observed that the suitability loss will mainly occur in southern and coastal Italy, while suitability gain will be observed in mountainous areas, such as central Apennines, western Alps and eastern pre-Alps and Alps. Despite the variety of approaches employed, local studies identified two main issues: (i) the identification of high-risk and refuge area will be of primary importance, (ii) the inadequacy of local strategies to monitor and mitigate Bd expansion (e.g. Flechas et al.
2017; Bacigalupe et al.
2019; Rahman et al.
2020). In Italy, while we addressed the first issue by both collecting new occurrence data and modelling habitat suitability, we also tried to overcome the second problem by modelling diffusion pathways at the landscape scale. Applying this procedure, we identified the national protected areas that should be more involved in monitoring and prevention actions to mitigate Bd diffusion.
Connectivity models based on electric circuit theory, have been widely adopted to model the spread of diseases or pathogens in wildlife populations, such as deer chronic wasting disease (Nobert et al.
2016), or rabies in raccoon populations (Algeo et al.
2017). Notably, among the almost 300 studies using circuit theory biology, ecology and conservation science (Dickson et al.
2019), only one involved the amphibian chytrid pathogen Bd (i.e. Becker et al.
2017). By modelling forest connectivity between populations of an Hylid frog in Brazil, Becker et al. (
2017) found that skin microbiome similarity and Bd load are related to landscape connectivity and natural vegetation gradient, but a landscape connectivity model of Bd spread is actually lacking. Therefore, our study, at least to our knowledge, is the first one employing circuit theory to model the diffusion of Bd at the landscape scale. By applying this method, we were able to identify four national protected areas (Pollino, Sila, Aspromonte and Golfo di Orosei and Gennargentu National Parks) that may experience maximum Bd diffusion for the current climatic conditions (i.e. Cumulative Current Flow). These areas of maximum diffusion are also identified by the presence of channelled or intensified movement as predicted by normalized current flow, meaning that in these areas the spread rate is potentially high (McRae et al.
2016).
For these areas, we suggest the development of an intensive molecular screening plan in order to track any possible change in Bd presence, population prevalence and individual load of resident amphibians. Furthermore, three out of four of these areas are in geographic and ecological connectivity (Pollino, Sila and Aspromonte National Parks), as resulted from our model, while the fourth area (Maddalena national Park) is in ecological connection with populations interested by a Bd induced mass-mortality that occurred in Sardinia (Bielby et al.
2013; Tessa et al.
2013). Indeed, according to the Italian national biodiversity strategy, national parks should become “focal points for research and monitoring networks … in terms of biodiversity” (MATTM
2010, page 31). Therefore, national parks identified as areas of high Bd diffusion will represent a fundamental tool to implement monitoring, awareness and mitigation strategies. Moreover, we suggest that a particular monitoring effort should be spent in those area where current and predicted Bd flow is channelled between climatic suitability barriers, within a national park or protected area, such as Dolomiti Bellunesi in the Alpine region and Circeo National Parks in the central region of the Apennine mountain chain. Finally, despite the fact that our study allowed the identification of high priority areas, in order to detect early Bd spread and diffusion; our results are obtained from an incomplete sampling of the national area. Therefore, the identification of National Parks (or other priority areas) involved in the monitoring network could benefit from a more intensive sampling effort or from the application of different techniques: such as the implementation of a national monitoring programme, relying upon environmental DNA sampling and Occupancy modelling (Schmidt et al.
2013).
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.