Infection and transmission heterogeneity of a multi-host pathogen ( Batrachochytrium dendrobatidis ) within an amphibian community

The majority of parasites infect multiple hosts. As the outcome of the infection is different in each of them, most studies of wildlife disease focus on the few species that suffer the most severe consequences. However, the role that each host plays in the persistence and trans mission of infection can be crucial to understanding the spread of a parasite and the risk it poses to the community. Current theory predicts that certain host species can modulate the infection in other species by amplifying or diluting both infection prevalence and infection intensity, both of which have implications for disease risk within those communities. The fungus Batrachochytrium dendrobatidis (Bd), the causal agent of the disease chytridiomycosis, has caused global amphibian population declines and extinctions. However, not all infected species are affected equally, and thus Bd is a good example of a multi-host pathogen that must ultimately be studied with a community approach. To test whether the common midwife toad Alytes obstetricans is a reservoir and possible amplifier of infection of other species, we used experimental approaches in captive and wild populations to determine the effect of common midwife toad larvae on infection of other amphibian species found in the Peñalara Massif, Spain. We observed that the most widely and heavily infected species, the common midwife toad, may be amplifying the infection loads in other species, all of which have different degrees of susceptibility to Bd infection. Our results have important implications for performing mitigation actions focused on potential ‘amplifier’ hosts and for better understanding the mechanisms of Bd transmission.


INTRODUCTION
The majority of parasites are able to infect multiple hosts (Fenton & Pedersen 2005).For example, most human pathogens are zoonotic in origin, and the majority of pathogens of livestock and domesticated species originated in wildlife species (Daszak et al. 2000).However, even within the widest host base, there exists a great deal of variation in how frequently and heavily different species become in fec -ted (Fenton & Pedersen 2005).As a result, host species play different roles in the persistence and transmission of infection within a community.
Just as individual-level transmission is highly skewed towards certain key individuals (Lloyd-Smith et al. 2005), the presence of certain species within a host community can be disproportionately important in the success of parasite invasion and persistence (Rudge et al. 2013).There are a number of ways in which a species may be of particular importance.'Vectors', 'reservoirs', 'amplifiers' and 'dilu ters' of infection are all terms used to describe species that, in different ways, help maintain, spread or reduce infection within a community.While vectors and reservoirs are widely accepted concepts, empirical evidence for the existence of pathogen amplification or dilution by hosts in natural systems is comparatively more limited (but see Searle et al. 2011).By definition, amplification hosts are species that make a pathogen more likely to persist and more abundant than it would be in the absence of that species (Begon 2008).By increasing the overall prevalence and infection intensity within sympatric species, amplification hosts may increase the risk of disease emergence within a host assemblage.Quantifying species' differences in host competence and their roles in parasite transmission is therefore essential if we are to understand the dynamics of infection and the likelihood of disease emergence within a com munity.
The chytrid fungus Batrachochytrium dendrobatidis (Bd) is a host-generalist pathogen that infects mainly the keratinized skin of developed amphibians and the mouthparts of amphibian larvae (Berger et al. 1998).Chytridiomycosis is an amphibian-specific emerging infectious disease caused by the fungus, and it has caused severe population declines and species extirpations and extinctions worldwide (Stuart et al. 2004).It is known to have infected over 400 species, including species in all 3 amphibian Orders (Gower et al. 2013; see also www.bd-maps.net),and probably many more species are susceptible to Bd infection.However, there is a great deal of variation in the pathogenic effects of Bd both among (Lips et al. 2006, Bielby et al. 2008) and within species (Walker et al. 2010).A range of intrinsic (Woodhams et al. 2007, Farrer et al. 2011, Jani & Briggs 2014) and extrinsic factors (Vredenburg et al. 2010, Raffel et al. 2015) have been linked to the variation in the effect of Bd on species and communities, but, as yet, relatively little is known about the role of community composition in this context.
The first known chytridiomycosis-related mortalities in Europe occurred in the late 1990s and led the common midwife toad Alytes obstetricans (Ao) to the brink of local extinction in the Peñalara Massif at the Sierra de Guadarrama National Park (Bosch et al. 2001).The species seems to be the most severely impacted species in Europe (Tobler & Schmidt 2010, Walker et al. 2010), and as a result of the ease with which it becomes infected, the clade Alytidae acts as a reliable sentinel when screening for infection in new regions or populations (Baláž et al. 2014).However, other species in heavily affected assemblages, such as those in Guadarrama, exhibit a variety of responses to Bd exposure.Following initial Ao mass mortalities, common toads Bufo spinosus and fire salamanders Salamandra salamandra also suffered mortality and declines as a result of chytridiomycosis (Martínez-Solano et al. 2003, Bosch & Martínez-Solano 2006, Bosch et al. 2014).In contrast, the rest of the amphibian species within the community at Guadarrama seem not to have been seriously affected by the disease, although all of them can be infected by the pathogen (www.bd-maps.net).The population-level effects of chytridiomycosis on B. spinosus and S. salamandra diminished following the near extinction of Ao (J.Bosch unpubl.data) and there fore we hypothesize that Ao was driving much of the infection transmission.
To better understand the risk of disease emergence within a host community, it is important to understand how different species within that community differ in their susceptibility to Bd infection, and their role in infection transmission.In this study, we chose to look at the range of tolerance to Bd across the amphibian species of the Peñalara Massif and investigate aspects of the transmission of the pathogen within this assemblage.Doing so is important for designing better management strategies, preventing future declines and improving reintroduction programme success.Specifically, we tested the hypotheses that Ao acts as an amplification host and that individuals of sympatric species will experience a higher probability and greater intensity of infection than individuals housed only with their own species.Further, we tested the hypothesis that Bd can transmit both directly, from Ao larvae to sympatric species, and also indirectly through swimming zoospore transmission, rather than relying on direct contact with an infected host.Finally, we investigated whether community members (i.e.individual species) exhibit different levels of infection from one another when housed with Ao larvae.Combined, these experiments may help to explain some of the observed community-level impacts of chytridiomycosis in the presence and absence of Ao.

Bd sampling
To analyse Bd infection loads, we took samples when the larvae were judged to be close to metamorphosis from the parts of the body where the concentration of zoospores is generally highest (Garner et al. 2009): the hind limbs of anurans or the complete body of urodeles.As Ao tadpoles have a longer developmental time (up to 5 yr), samples were taken from the keratinized mouthparts when companion species were sampled, or at the end of the experiment.Samples were taken from live animals with a fine-tipped sterile swab (Medical Wire and Equipment 113) or directly from the corresponding tissue of newly dead individuals or after euthanasia.

Laboratory methods
To quantify infection load in amphibians, we used a quantitative real-time polymerase chain reaction (qPCR) protocol (Boyle et al. 2004).Extractions were diluted 1:10 before real-time PCR amplification, performed in duplicate, with Bd genomic equivalent (GE) standards of 100, 10, 1 and 0.1 GE (isolate IA042, Ibón Acherito, Spanish Pyrenees) in a CFX96 machine (BioRad).When only 1 replicate from any sample amplified, we assayed this sample a third time.If the third amplification did not result in an amplification profile, we considered the sample negative for infection.

Experiment 1
This experiment was set up to test the hypothesis that Ao acts as an amplification host by increasing infection prevalence and intensity in sympatric species, and also that Bd can initially infect hosts indirectly through waterborne spores, rather than relying on direct contact with an infected host.We conducted a field experiment in the Laguna Grande de Peñalara glacial lake of the Peñalara Massif (2018 m a.s.l.), and B. spinosus was chosen as our focal susceptible species as it has been observed to suffer infection and mortality as a result of Bd infection in natural surroundings (Bosch & Martínez-Solano 2006) and in experimental settings (Garner et al. 2009).Several hundred B. spinosus free-swimming Gosner stage 25 tadpoles (Gosner 1960) were collec ted from different locations at Laguna Grande to average any possible genetic variation among offspring.Previous studies have shown that at this stage of development, B. spinosus tadpoles lack Bd infection (Ortiz-Santaliestra et al. 2011).Uninfected Ao larvae were obtained from a captive colony located in the studied area that is regularly tested for Bd infection by qPCR.Larvae from the stock of our focal species, B. spinosus, were assigned to 1 of 4 different treatments in a 2 × 2 experimental design.The 2 factors of interest were density and the presence of Ao larvae, and each of these 2 factors had 2 levels: high density (50 B. spinosus larvae) or low density (25 B. spinosus larvae), and presence (10 larvae) or absence (0 larvae) of Ao larvae.The selected densities are within the range typically observed naturally in this system (J.Bosch unpubl.data).Each treatment was replicated 3 times, each group being housed in a separate 4 l container.The containers had ventilated sides and were placed together floating in the lake.Water temperature inside each container was re corded with a thermocouple thermometer in a randomized order and found not to differ between containers.The experimental design removed the pos si bility that infection was introduced with any of the experimental animals, as they came from uninfected stock, or were placed in the experiment before keratinized mouthparts had developed and had the opportunity to become infected.Instead, experimental animals could only become infected when ex posed to zoospores in the lake water.Once the most advanced B. spinosus tadpoles were close to metamorphosis (31 d after the experiment began), the experiment was ended, and we euthanized 20 randomly selected B. spinosus tadpoles (Gosner stages 38−42) per container and stored them in 70% ethanol before processing for Bd infection.We ended the experiment at this point because we wanted to assess infection in larvae, before they undergo meta morphosis when some individuals lose infection, or infection becomes difficult to detect (Garner et al. 2009).
To see whether the 4 experimental levels resulted in a different probability of infection, we used a chisquared test, and, in the presence of any significant variation, we used generalized linear models (GLMs) with binomial errors to determine which of the 2 factors best explained variation in infection probability of B. spinosus.For the latter analysis, backwards step wise regression of a full model including all terms was implemented, with changes in model fit being measured using analysis of deviance.Because our experimental design does not adequately account for the total density of larvae when considering the presence and absence of Ao as a factor (i.e.within each level of the density treatments, B. spinosus had different total tadpole densities depending on whether Ao was present), we used binomial tests to identify whether the proportion of individuals infected significantly varied in the high and low density treatments in the absence of Ao.Doing so allowed us to determine whether an increase in the density of the focal host was an important factor in infection levels in the absence of Ao.
To analyse whether infection intensity varied with density and presence/absence of Ao larvae, we used GLMs with negative binomial errors using the glm.nb function from the R package MASS.The function glht from the multcomp library was used to determine which levels of the 4 treatments varied from one another.

Experiment 2
This experiment tested whether species co-housed with Ao larvae differed from one another in probability and intensity of infection.Plastic containers (2 l capacity, n = 52) were floated together in the Laguna Grande de Peñalara.The containers had holes to allow water exchange with the surrounding lake.Water temperature inside the containers was measured with a thermocouple thermometer in random order without mixing the water before sampling began, and it did not differ significantly among containers.Each of 13 treatments was replicated 4 times.The 13 treatments were (1) 2 larvae of Ao alone, which acted as a control to see how heavy infection was in this species when housed alone; (2−7) 6 treatments consisting of 2 larvae of Ao co-housed with 2 larvae of each of B. spinosus, B. calamita, H. molleri, P. perezi, R. iberica or S. salamandra; and (8−13) 2 larvae of each of those 6 species alone (i.e.no Ao were added).Larvae of the studied species were collected in the field in several ponds of the Peñalara Massif, and Ao larvae were obtained from the captive colony.All larvae were placed in the experimental set-ups at an early stage of their development before keratinized mouthparts had developed, and their uninfected status was confirmed by qPCR.One overwintered larva of S. salamandra from the same lake was introduced into each container for 1 wk.As overwintered larvae, S. salamandra have an infection prevalence of 100% in spring in this system (Medina et al. 2015).Therefore this was a guaranteed way to expose experimental animals to infection regardless of whether experimental animals were exposed to zoospores in the lake water.At the end of the experiment, we measured the infection intensity of all larvae in each of the 13 treatments.
We tested whether species differed from one another in their infection probability.Using a Fisher's exact test, we determined whether the proportion of infected individuals of the different species varied, and in the event of a significantly non-random distribution of infection, we used binomial tests to determine which species varied significantly from the background prevalence of infection in the experiment.
To determine whether infection intensity in cohoused species was higher in the presence of Ao, we conducted a Student's t-test to compare infection intensity between those individuals co-housed with Ao with those housed only with a conspecific for each of the 6 co-housed species.
To investigate whether infection intensity differed among each species when co-housed with Ao, we used a GLM with negative binomial errors and Tukey comparisons.The same statistical tests were used to determine whether Ao varied in infection intensity when co-housed with different species.GLMs with negative binomial errors were conducted using the glm.nb function from the MASS library, and the Tukey comparisons on the resulting glm.nb object were made using the glht function from the multcomp library.

Experiment 3
The following experimental set-up in the laboratory was used to test whether Ao can transmit Bd infections directly to other species, whether those species differ from one another in the resulting infection intensity, and whether Ao experiences different levels of infection when co-housed with other species.Newly hatched larvae of 5 species (H.molleri, P. perezi, I. alpestris, T. marmoratus and S. salamandra) were captured in the field in several ponds of the Peñalara Massif, and their uninfected status was confirmed by qPCR.Two larvae of each of these species were placed in the presence of a single infected Ao larva, resulting in 5 experimental treatments.The sixth treatment was a single infected Ao larva, housed alone.Each of the 6 treatments was replicated 10 times in 2011 and 15 times in 2012.All experimental replicates were housed in 1.5 l containers maintained at a temperature of 18°C.All Ao larvae were collected from a well-studied population (Toro, Zamora, western-central Spain;Fernández-Beaskoetxea et al. 2015), and their infection status was checked by qPCR before the experiment started.To test whether species differed from one another in their probability of infection, we used a Fisher's exact test to compare proportions of infected and uninfected individuals for each species.To test whether species differed from one another in their infection intensity when co-housed with Ao larvae, we used a GLM with negative binomial errors and Tukey comparisons between species to determine whether significant differences occurred.The former were conducted using the glm.nb function from the MASS library, and the Tukey comparisons on the resulting glm.nb object were made using the glht function from the multcomp library.All analyses were conducted in the statistical software package R (R Core Team 2014).

Experiment 1
The prevalence of Bd in Bufo spinosus tadpoles at the beginning of the experiment was 0% according to qPCR analyses.The prevalence of infection in B. spinosus the end of the experiment differed significantly among the 4 treatments (χ 2 = 38.23,df = 3, p < 0.001; Table 1).In the presence of Ao larvae, the prevalence of infection in B. spinosus was about 50%, while in the absence of Ao larvae, it was less than 7%.The fact that infection occurred suggests that infection can occur and persist via indirect transmission from zoospores in the water and is not initially reliant on direct contact with an infected host.
Our model of infection probability was simplified to leave the presence/absence of Ao larvae as the only significant predictor of likelihood of infection (Table 2).Using a binomial test, we found no significant difference in the proportion of infection of B. spinosus larvae kept at low (2/25) and high density (3/50) in the absence of Ao (χ 2 < 0.001, df = 1, p = 1.000), indicating that regardless of the density of B. spinosus, infection did not become well established in the absence of Ao.Because of the very low number of infected animals in each of these 2 treatments, it was not useful to compare infection burden between treatments.
The model of infection intensity contained both density of hosts and the presence/absence of Ao as factors affecting infection intensity in B. spinosus tadpoles.The output for this model is presented in Table 3.The model fit could not be significantly improved by the backwards stepwise regression process, meaning that the best-fitting model was obtained when both terms were left in the model.Tukey's HSD suggested that B. spinosus larvae housed at high density in the presence of Ao larvae had a significantly higher infection burden than those at high density without Ao larvae, and that B. spinosus larvae held at low density in the presence of Ao larvae had a higher infection intensity than B. spinosus larvae at high density in the absence of Ao (Table 4).
There were no significant differences in the proportion of individuals infected or the infection intensity in Ao larvae when co-housed with different species (Fig. 1; Fisher's exact test, p = 0.796).This result is most likely because by the end of the experiment, most Ao larvae were fairly heavily infected (Fig. 1).

Experiment 3
Individuals of other species co-housed with Ao did become infected, suggesting that Ao can transmit infection to other species.A Fisher's exact test on the species co-housed with Ao suggested no significant difference among those species in their probability of infection (p = 0.2126; Fig. 2).Significant differences in infection intensity were present among those species (Fig. 2; F = 4.9807, df = 4, p < 0.001).P. perezi had a significantly higher in fection intensity than H. molleri, Triturus marmoratus and S. salamandra.Ichthyosaura alpestris had heavier infections than H. molleri and T. marmoratus (Table 6).
The prevalence of infection in Ao varied significantly depending on whether they were housed alone or with the larvae of other species (Fig. 2; Fisher's exact test, p = 0.0036).Ao larvae experienced differences in infection intensity depending upon the species with which they were cohoused (F = 5.068, df = 5, p < 0.001).Ao housed alone had significantly lower infection burdens than when housed with any species aside from I. alpestris, when the infection intensity in Ao did not differ from when housed alone.Ao larvae housed with H. molleri had significantly higher infections than Ao larvae housed with I. alpestris (Table 7).

DISCUSSION
Within an assemblage of hosts, it is difficult to predict whether a parasite will become established, will spread or will cause disease because of heterogeneity in host response within a community.Our study shows that all species of the Peñalara Massif are susceptible to Bd infection, and that their levels of susceptibility vary greatly from one another.As a result, different species are likely to play different roles in the infection dynamics within the system.Of particular note, our data suggest that the larvae of 1 species, Ao, could contribute a disproportionate amount to the spread of infection and, in so doing, may act as an amplification host.By carrying severe infections, it may cause co-housed species to experience elevated levels of infection, and by transmitting directly to other species, overwintering Ao larvae may play the role of amplification host within this host community.
The ability of overwintering amphibian larvae to act as infection reservoirs is well established (Brunner et al. 2004, Narayan et al. 2014, Medina et al. 2015), yet there is little empirical evidence to suggest that they can increase levels of infection within a host   assemblage.Combined, the results of our experiments suggest that Ao larvae are able to increase infection prevalence and intensity in a number of cohoused species by directly transmitting infection to them.Further, the ability of Ao to act as an amplification host appears to be independent of the overall density of larvae around it, as highlighted in Expt 1.
In this experiment, the density of the co-housed focal species Bufo spinosus did not affect its likelihood of becoming infected, which remained close to 0 in the absence of Ao.In contrast, B. spinosus larvae held at low density in the presence of Ao larvae had a higher infection intensity than B. spinosus larvae at high density in the absence of Ao, suggesting that the presence of a single Ao larva resulted in a significant increase in infection probability and intensity regardless of overall host density.The fact that the presence of Ao is strongly associated with infection in other species, regardless of the overall density of hosts, suggests that even post-decline, when the overall density of hosts is reduced, infection may still be maintained and spread if Ao larvae remain present in the community.
We considered what characteristics would predispose Ao to act as reservoirs or disseminators of infections in the shorter term.One possible morphological feature that would lend itself to a species harbouring and transmitting high levels of infection is its large oral disc.In Ao, this feature is unusually large, and includes numerous rows of large denticles with a high concentration of keratin.Therefore, it has a greater area available to be infected by the pathogen (Berger et al. 1998; but see Searle et al. 2011, who reported that smaller species, such as Anaxyrus boreas, presented the highest infection loads).This potential mechanism could be explored further using techniques to track infection prevalence and infection intensity in different body parts, and it highlights the importance of understanding species' biology when considering their roles in transmission of infection within a community of hosts.
Efforts to better understand how and when transmission of infection will take place rely greatly on accurate information about mechanisms and modes of transmission.Within this host−pathogen system, it is generally assumed that, given the low motility of Bd zoospores (Moss et al. 2008, Lam et al. 2011) and the tendency of amphibians to cluster at high densities in suitable conditions (Duellman & Trueb 1994), direct host contact may be the most common method of infection transmission.However, the data from Expt 1 suggest that initial infection can and does occur as a result of exposure to infected lake water by means of zoospores present in the lake.This finding supports previous research in demonstrating that transmission of infection does not necessarily require direct contact between the tadpoles (Rachowicz & Briggs 2007) and can help to inform future efforts to understand transmission events within this host− para site system.
A great deal of variation in host susceptibility to Bd infection was observed within our Expts 2 and 3.Although the majority of species had an increased probability of infection and infection intensity in the presence of Ao larvae, there was a considerable degree of variation among species as to how prevalent or severe those infections became.These differences reflect how the transmission dynamics within a community may differ depending upon its constituent species.This observation makes it difficult to predict how host communities will respond to the introduction of Bd.Additionally, there was little consistency in how the studied species responded to Bd introduction levels among the performed experiments (for example, Hyla molleri and Pelophylax perezi in Expts 2 and 3).
Infection levels varied not only in those species co-housed with Ao, but also in Ao larvae depending upon the species with which they were housed.Cohoused Ao generally suffered more frequent and heavier infections than those housed alone, but those with Ichthyosaura alpestris did not, having signifi- While it is currently difficult to determine the mechanism behind these differences, the end result is that, even for a host capable of carrying heavy infection burdens, competition with other larvae, or the ability for infection to be transmitted both to and from other species in the assemblage, may, at times, be important for the maintenance of infection.These inconsistencies and inter-species differences suggest that the outcome of Bd exposure is highly context dependent and may differ greatly depending upon the source of infection and the environment in which the larvae develop, illustrating how important it is to carefully consider the generalities of research into the transmission within any host−parasite system.Rachowicz & Briggs (2007) showed that under laboratory or field conditions, there is a clear influence of the density of infected individuals on the rates of Bd transmission.The density of both host and pathogen are fundamental parameters in the transmission of infectious disease.Although the experimental numbers of tadpoles in our study were similar to those used in the study mentioned above, we did not find a significant effect of density of tadpoles in the variation of Bd infection intensity.Our experimental design was such that comparisons of species cohoused with and without Ao varied not only in species composition, but also in the density of animals in the experimental treatments.Accounting for both density and species composition would be the ideal approach to take, but the practicalities involved with conducting such experiments prevented these dual comparisons.Regardless of these different densities, the main findings of our experiments remain unchanged.We conclude that Ao presence/absence is a greater predictor of infection than overall density of tadpoles (Expt 1), that species co-housed with Ao differ in their response to parasite exposure (Expts 2 and 3), that Ao varies in its infection levels depending on the species with which it is housed and that Ao can directly infect other species (Expt 3).
To add more complexity to the overall findings, competition and stress between 2 host species may account for some of the observed patterns.Additional experiments with the target host at different densities and addition of non-target and non-Ao hosts would be needed to test whether additional host species simply cause competitive stress and thus lead to increased infection.
Identifying the roles that different species or life stages play in the transmission, prevalence and intensity of infection is crucial to better understand the persistence and spread of infection within a host− pathogen system.Knowledge related to which species are more tolerant and more susceptible to infection could allow designers of mitigation efforts to focus on reducing the levels of infection in a host.In the case of our study system, this might be accomplished by aiming to reduce the amount of infection in potential 'amplifier' hosts.Considering the species composition of a particular host community is essential in efforts to understand the spread of infection, risk of disease emergence and, ultimately, in managing systems to minimize any negative effects of pathogens on biodiversity.

Table 4 .
Tukey's HSD test showing differences in infection intensity among 4 treatment levels in Expt 1. Bs HD/LD indicates high/low density of Bufo spinosus larvae and +/− Ao indicates presence/absence of Alytes obstetricans larvae

Table 5 .
Pairwise comparisons of infection intensity between the 6 species co-housed with Alytes obstetricans larvae in Expt 2. Arrows indicate the relative infection intensity of the species named in the row compared to the species named in the column

Table 6 .
Tukey's HSD tests of infection intensity in species co-housed with Alytes obstetricans larvae in Expt 3. Arrows indicate the relative infection intensity of the species named in the row compared to the species named in the column

Table 7 .
Tukey's HSD tests between Alytes obstetricans larvae (Ao) co-housed with different species in Expt 3. Arrows indicate the relative infection intensity of the species named in the row compared to the species named in the column cantly lower infections than Ao housed with H. molleri larvae.