|Home | About | Journals | Submit | Contact Us | Français|
To minimize the negative effects of an infection on fitness, hosts can respond adaptively by altering their reproductive effort or by adjusting their timing of reproduction. We studied effects of the pathogenic fungus Batrachochytrium dendrobatidis on the probability of calling in a stream-breeding rainforest frog (Litoria rheocola). In uninfected frogs, calling probability was relatively constant across seasons and body conditions, but in infected frogs, calling probability differed among seasons (lowest in winter, highest in summer) and was strongly and positively related to body condition. Infected frogs in poor condition were up to 40% less likely to call than uninfected frogs, whereas infected frogs in good condition were up to 30% more likely to call than uninfected frogs. Our results suggest that frogs employed a pre-existing, plastic, life-history strategy in response to infection, which may have complex evolutionary implications. If infected males in good condition reproduce at rates equal to or greater than those of uninfected males, selection on factors affecting disease susceptibility may be minimal. However, because reproductive effort in infected males is positively related to body condition, there may be selection on mechanisms that limit the negative effects of infections on hosts.
Life-history traits of organisms, including growth, reproduction and longevity, interact to influence fitness. Because these traits are constrained by resource availability, many organisms can adaptively modify their allocation of energy as their circumstances change [1,2]. Resource allocation strategies can vary among individuals, with natural selection favouring individuals with resource allocation patterns that enhance lifetime reproductive success [1,2]. Pathogens can influence host fitness by reducing survival, and also by affecting reproductive success. A pathogen can influence reproductive success not only by reducing the total amount of resources available to a host, but also by altering the optimal pattern of resource allocation for an individual once it becomes infected .
To maximize fitness, individuals typically maintain moderate levels of current reproductive effort; this results in a longer lifespan and production of more offspring during their lifetime [1,2]. Individuals that are infected by a pathogen may allocate resources differently, however, depending on their immune defences and longevity. For example, it might be beneficial for infected individuals to preferentially allocate resources to immune responses to fight their infections, and to invest fewer resources in gamete production or reproductive behaviour. In many taxa that rely on energetically expensive vocalizations for mate attraction (e.g. frogs, birds, insects), males that are infected may vocalize less, and they may also alter their vocalizations in terms of rate, length, complexity and frequency [4–6]. Males that vocalize less should attract fewer mates, mate less often and produce fewer offspring. Other pathogen-induced changes in sound production could also reduce fitness, because vocalizations are subject to sexual selection by females. Infected males that reduce the rate, length or complexity of vocalizations may be less attractive, because these characteristics are often honest signals of overall genotypic fitness [7,8].
Life-history theory predicts that iteroparous organisms should increase their current reproductive effort as life expectancy decreases [1,2,9]. Therefore, the optimal life-history strategy for infected individuals may be to increase investment in current reproductive effort [10,11]. Both male and female hosts can compensate for an increased risk of mortality imposed by an infection by increasing their investment in earlier reproduction. For example, among female hosts, Tasmanian devils (Sarcophilus harrisii) infected by a transmissible cancer mature and breed earlier , and crickets (Acheta domesticus) infected by a bacterium, and water fleas (Daphnia magna) infected by a microsporidian lay more eggs [13,14]. Among male hosts, frogs (Lithobates pipiens) infected by a fungus increase sperm production , flies (Drosophila nigrospiracula) infected by a parasitic mite, and amphipods (Corophium volutator) infected by trematodes increase reproductive effort [16,17], and beetles (Tenebrio molitor) infected by tapeworms provide higher quality nuptial gifts to their mates, thus increasing egg production . Whether hosts increase their reproductive effort in response to an infection depends on many factors, including resource availability . Therefore, infected hosts in very poor condition may not be able to increase reproductive effort [20–22].
Understanding the effects of pathogens on host reproduction, and thus fitness, has direct implications for population demography and evolution. The primary mechanism of attracting and locating mates in most frogs is through male advertisement calls, and frog populations in many regions of the world are undergoing declines due to chytridiomycosis, the disease caused by the widespread pathogenic fungus Batrachochytrium dendrobatidis . This fungus attacks the skin cells of amphibians and disrupts their osmoregulatory and transport functions, ultimately altering electrolyte concentrations in the blood, and causing cardiac arrest when the fungal population on the host reaches a high density . Currently, we have very little understanding of whether B. dendrobatidis infections influence amphibian reproduction. Although B. dendrobatidis can influence host survival directly, many individuals carry sublethal infections, sometimes for extended periods of time, and can ultimately recover . Many aspects of frog behaviour can differ between frogs infected by B. dendrobatidis and uninfected individuals (e.g. movements, microhabitat use, microenvironment use [26,27]), potentially reflecting changes in the behaviour of infected frogs. Infection by B. dendrobatidis could also affect direct fitness traits, such as energetic investment in mate attraction or gamete production . Because calling requires substantial energy, reduced body condition can also lead to reduction of calling effort or temporal shifts in calling effort . Frogs infected by B. dendrobatidis often have lower body condition than uninfected frogs [28–30]; when this occurs, they might reduce calling effort. On the other hand, infected hosts with relatively good body condition may have the plasticity to respond to the infection by investing more in present reproductive effort than uninfected individuals when their expectation of survival to the next reproductive bout is lower [13–18]. Such a response could at least partially counteract the effects of natural selection on disease resistance, because it may reduce differences in reproductive success between more and less vulnerable males.
We studied the effects of B. dendrobatidis infection and body condition on the probability of calling by the common mistfrog, Litoria rheocola, a stream-breeding rainforest frog with a history of declines caused by chytridiomycosis . We sampled frog calling behaviour and infection status both spatially (across six sites differing in elevation) and temporally (this species calls year-round), allowing robust tests of the hypothesis that infection by B. dendrobatidis has sublethal effects that interact with body condition to influence calling probability.
The common mistfrog (L. rheocola) is a treefrog that occurs near tropical rainforest streams in northeastern Queensland, Australia . By the mid-1990s, chytridiomycosis had extirpated this IUCN-Endangered species  at higher elevations (more than 400 m) throughout its geographical range . However, many populations have subsequently recovered or recolonized these areas  and now coexist with the pathogen . Several aspects of the ecology of L. rheocola make it ideal for examining how disease influences calling effort across seasons and with body condition. First, this tropical species calls and breeds year-round, although reproductive behaviour decreases during the coolest weather . Second, individuals of this species that are calling on a given night typically call throughout the night, allowing us to easily and accurately assess calling behaviour. Third, active frogs are located relatively easily, even when they are not calling. Males remain near fast-flowing streams, where they are typically found on rocks and streamside vegetation [32,35]. During a tracking study, male frogs were located an average of 0.10 m (maximum: 1.80 m) from the stream during winter, and 0.74 m (maximum: 3.75 m) away during summer . Finally, infection by B. dendrobatidis does not alter the detection probability of males of this species, so infected males are just as likely to be located as uninfected males . As with many stream-associated rainforest frogs, the behaviour of females is poorly understood because they are observed much less frequently than males [25,35,36].
We surveyed L. rheocola at six rainforest streams in northeastern Queensland, Australia (table 1). All streams were surrounded by tropical rainforest, characterized by dense vegetation composed of large trees (10 m in height), vines, epiphytes, shrubs and herbaceous plants. Although our sites were in relatively undisturbed rainforest, several sites were damaged by a tropical cyclone in February 2011 . Stream width varied from 5 to 10 m, and streambeds were composed of rocks ranging in size from small pebbles to large boulders (10 m in diameter). All streams contained pools, runs and riffles, and most had several waterfalls.
We located adult L. rheocola by visually surveying frogs along a 400 m transect at each stream. Surveys were conducted over five consecutive nights at each site in each season, from June 2010 through to October 2011 (except for spring 2011, when we sampled for one night per site). We conducted two winter (June–July) and two spring (October–November) surveys (2010 and 2011), one survey in summer (January–February 2011) and one survey in autumn (March–April 2011). We could not conduct a summer survey during 2011 at Bobbin Bobbin Creek, because this site was inaccessible due to cyclone damage. We captured each frog as soon as it was located visually. Prior to capture, we recorded whether each male frog was calling. Individuals can be heard distinctly from a distance of about 50 m, and our period of observation for detecting calling behaviour was approximately 3–5 min.
We measured the body size of each captured frog (snout–urostyle length to 0.5 mm, and mass to 0.1 g), and we determined its sex by the presence/absence of distinct nuptial pads. We estimated a body condition index for each male frog using the residuals from a linear regression of log10 transformed body mass on square-root transformed snout–urostyle length for all male frogs sampled . The resulting positive relationship was strong and highly significant (r2 = 0.45, F1,2486 = 1197.98, p < 0.001). To determine whether frogs were infected by B. dendrobatidis, we swabbed the ventral surface and all four feet of each frog with a sterile rayon swab, covering these areas twice. These samples were analysed using real-time quantitative PCR assays . We also gave each frog a unique identifying mark using visible implant elastomer . For analysis, we used the initial capture of each male frog and excluded all recaptures, resulting in an independent sample of male frogs.
We used generalized linear mixed-effects models to examine potential effects of infection status, body condition and season on the calling probability of individual male frogs. Calling status was coded as a binomial response variable, so we used models with a binomial family and a logit link function. We developed a set of candidate models that included models with all combinations of one, two or three fixed effects, and all two- and three-way interactions. For all models, we included site identity as a random effect to control for any overall effects of particular sites. We ranked models according to Akaike's information criterion with adjustment for finite sample size (AICc) to determine the strength of evidence for each model relative to the set of candidate models, using the criteria of Burnham & Anderson . Statistical analyses were performed in program R, v. 2.15.2 , using the lme4  and MuMIn  packages. Because infection loads were low during our study (98% of infected frogs had less than 50 zoospores), we could not examine possible effects of infection load on calling probability.
We captured a total of 1843 unique male frogs at six study sites during six seasonal samples (table 1). Of these frogs, 372 individuals were infected, with fungal loads of up to 912.7 zoospore equivalents per frog. Fungal loads were lowest during autumn (mean: 0.7 zoospore equivalents per frog, maximum: 4.1) and summer (mean: 2.2, maximum: 15.2), and highest during spring (mean: 14.7, maximum: 912.7) and winter (mean: 3.6, maximum: 308.1). We found that infection status, body condition and season all influenced the calling probability of individual frogs (table 2 and figure 1). Four models with similar ΔAICc values that were less than 3 were strongly supported by our data. Because the selected threshold for model selection should be based on all models in the set, rather than an arbitrary cutoff , we averaged the top four models that were most strongly supported by our data and had a total Akaike weight of 65% (table 2). This resulted in a final, averaged model that included the random effects of site and the main effects of body condition, season and infection status, plus interactions between infection status and body condition, season and infection status, and season and body condition (table 2).
Our results clearly demonstrate that the relationship between frog body condition and calling probability was strongly influenced by infection status (figure 1). For uninfected frogs, calling probability was relatively constant across seasons; our models suggest slight decreases with increasing body condition in all seasons except winter, where there is a slight increase, but the slopes of the lines are near zero in all seasons (figure 1). By contrast, calling probability for infected frogs differed among seasons; calling probability was lowest in winter, highest in summer and intermediate in spring and autumn. Calling probability also depended strongly on body condition in infected frogs; across all seasons, the probability of calling increased strongly as body condition increased (figure 1). The predicted body condition at which both infected and uninfected frogs were equally likely to call changed seasonally: in winter, infected frogs called more than uninfected frogs only if they were in very good condition (i.e. their body condition index was well above zero); in both spring and autumn, infected frogs called more than uninfected frogs if they were in good condition (i.e. their body condition index was zero or above) and in summer, infected frogs called more than uninfected frogs, even if they were in poor condition (i.e. their body condition index was well below zero).
We found that the calling probability of male frogs (L. rheocola) was influenced by interactions among B. dendrobatidis infection status, body condition and season, strongly suggesting that males were employing a pre-existing, adaptive, condition-dependent response to infection. For uninfected frogs, calling probability was relatively constant across seasons (near 50%; figure 1), consistent with reports that L. rheocola call and breed year-round . This is probably the maximum sustainable calling rate for healthy males over the long term. Calling probability was affected very little by body condition in uninfected frogs, but it did decrease slightly with increasing body condition during all seasons except winter (figure 1), consistent with the expectation that frogs which expend more energy calling should have less energy reserves, and therefore lower body condition . By contrast, the calling probability of infected frogs differed among seasons (lowest in winter, highest in summer) and depended strongly on body condition. In each season, the calling probability of infected frogs increased with body condition, such that infected frogs in poor body condition were less likely to call than uninfected frogs in similar condition, but infected frogs in good condition often had a higher calling probability than uninfected frogs (figure 1).
This pattern of increased calling effort in infected frogs is consistent with life-history theory, which predicts that reproductive effort should increase as life expectancy decreases [1,2,9]. Because L. rheocola has not coexisted with B. dendrobatidis for an extended period, it is likely that the pattern we observed reflects a generalized, plastic response to a pathogenic infection, in which infected frogs increase their present reproductive effort at the expense of possible future reproductive effort. A functionally similar response to infection by B. dendrobatidis occurs in northern leopard frogs (Lithobates pipiens); the testes of infected males are larger and contain more sperm than those of uninfected males . Studies on other taxa reveal that present reproductive effort can increase as life expectancy decreases [10–18]. In our study, however, calling activity did not increase in all infected frogs; we found that infected frogs in relatively poor condition were less likely to call than uninfected frogs. This could be because infected frogs were unable to call owing to physiological changes caused by their infections, or because frogs were adaptively adjusting their energy expenditure. For example, they may have been allocating less energy to reproduction and more to other functions required for immediate survival, such as immune responses to fight their infections .
There are several alternative hypotheses for the pattern we observed. For example, the pathogen might be manipulating the host, possibly to increase contact between infected males and uninfected females, thus increasing rates of pathogen transmission . However, this seems unlikely. The effects of rates of physical contact between frogs on rates of transmission of B. dendrobatidis are not known, but frogs can also become infected by contact with water or substrates, both of which can harbour infectious B. dendrobatidis zoospores [48,49]. In addition, zoospores can be carried and released into the environment by non-amphibian hosts, including nematodes and crayfish [50,51], and non-amphibian reservoirs, including reptiles and waterfowl [52,53]. Increased calling effort could also attract predators , which would be detrimental for pathogen transmission. Overall, it seems unlikely that there would be strong selection for B. dendrobatidis to increase transmission rates by increasing male calling rates.
Another alternative hypothesis is that changes in the calling probability of infected frogs were caused by side effects of the infection. Calling is energetically expensive , and reduced energy availability should lead to decreased calling activity. This could account for the decreases in activity we observed in frogs with poor body condition. However, such side effects cannot explain the pattern that infected frogs in relatively good body condition had greater calling probabilities than did uninfected frogs in similar condition (figure 1). The most plausible explanation for this pattern is that infected frogs in relatively good body condition were exhibiting a generalized, plastic response to pathogen infection by allocating more energy to reproductive effort than did uninfected frogs in similar body condition.
Our results demonstrate that the interaction between B. dendrobatidis infection status and body condition can strongly influence the probability of calling. However, increased calling effort in infected frogs may not lead to increased mating opportunities if females are not attracted to their calls. Female frogs often prefer calls that are louder, longer and emitted at faster rates, because they often indicate genetic superiority of males capable of producing high levels of sound [8,45]. In some frog species, parasitic infections can reduce male calling rates , and thus possibly reduce mating success. In other frog species, however, infected males do not change the quality of their calls and females do not avoid mating with them . In L. rheocola, we do not know whether the quality of calls produced by infected males differs from that of calls produced by uninfected males. Regardless of the relative attractiveness of calls emitted by infected males, it is likely that calling more frequently will attract more mates than not doing so. Determining the influence of B. dendrobatidis infections on calling behaviour and reproductive success will be necessary to fully understand how this pathogen influences population dynamics.
Season affected the calling probabilities of infected frogs more strongly than those of uninfected frogs (figure 1). The effects of season also interacted with those of body condition; the average body condition of frogs was lowest in winter, highest in summer and intermediate in spring and autumn (figure 1). These changes appeared to be greater in infected frogs than in uninfected frogs (figure 1). Seasonal changes in body condition of both uninfected and infected frogs were probably caused by changes in energy acquisition or expenditure. Reduced energy intake could be associated with low availability of rainforest arthropods during dry months , which can affect the diets of frogs . The strongly reduced body condition of infected frogs in winter could also be related to greater infection loads caused by faster growth rates of B. dendrobatidis under cooler temperatures . In infected frogs, calling probability differed among seasons (lowest in winter, highest in summer) and was strongly and positively related to body condition, whereas in uninfected frogs, calling probability was relatively constant across seasons and body conditions. In summer, when frogs were in the best condition, infected frogs were up to 30% more likely to call than were uninfected frogs (figure 1). In winter, however, when frogs were in the worst condition, infected frogs were up to 40% less likely to call than were uninfected frogs (figure 1). During spring and autumn, infected frogs in poor body condition were less likely to call than uninfected frogs in similar condition, but when infected frogs were in good body condition, their calling probability was often higher than that of uninfected frogs (figure 1).
Our findings suggest that infected males in poor condition will have lower fitness than healthier frogs, but that infected males in good condition may compensate for a potential loss of future reproductive output by increasing their current efforts, and thus their fitness. It is possible that males with reduced calling effort may override these effects by using alternative mating tactics, such as by attempting to intercept females attracted to nearby calling males (‘satellite behaviour’) . Likewise, the effects of B. dendrobatidis infections on female reproductive biology and behaviour are unknown. Understanding how this pathogen alters male reproductive tactics and female reproduction will aid in our understanding of how it influences amphibian populations and will provide insight into the potential for evolutionary responses.
Calling is the primary mechanism of attracting and locating mates in most frogs, and therefore, the changes we documented in the calling effort of infected males are likely to affect their mating success. This could lead to a variety of effects on the evolution of the host species. If infected males in good body condition reproduce at rates equal to or greater than those of uninfected males, selection on factors affecting the probability of acquiring infections may be less effective than it otherwise would be. However, because reproductive effort in infected males is strongly and positively related to body condition, males in good condition may produce more offspring than those in poorer condition. This could lead to selection favouring mechanisms that limit the negative effects of infections on body condition. Batrachochytrium dendrobatidis has devastated amphibian populations in many regions of the world, but many populations are coexisting with the pathogen [25,60]. Elucidating whether populations that are coexisting with the pathogen are experiencing sublethal effects that influence reproduction and mating systems is important for understanding potential changes in population demography and evolution.
Research was conducted under permits WISP03070208 and WITK03070508 issued by the Queensland Department of Environment and Resource Management, and protocols A1420 and A1673 approved by the Animal Ethics Committee at James Cook University.
Data are available as the electronic supplementary material.
E.A.R., D.A.P. and R.A.A. designed the study; S.J.S., E.A.R. and D.A.P. conducted the fieldwork; R.A.A., L.S., D.A.P. and E.A.R. contributed funding; E.A.R. and R.A.A. analysed the data; E.A.R. wrote the paper; and all authors revised the paper.
We declare we have no competing interests.
Funding was provided by a Linkage Grant from the Australian Research Council in partnership with Powerlink Queensland (LP0776927 to L.S. and R.A.A.), Discovery Grants from the Australian Research Council (DP0986537 to R.A.A.; DP130101635 to R.A.A., L.S. and D.A.P.), the Skyrail Rainforest Foundation (to E.A.R.), and the Graduate Research School at James Cook University (to E.A.R.). E.A.R. was supported by a Postgraduate Research Scholarship and a Doctoral Completion Award, both from James Cook University.
We thank many volunteers for help with fieldwork, and Michael Mahony, Dale Roberts and two anonymous reviewers for helpful comments on the manuscript. Diagnostic quantitative PCR assays were performed by the Amphibian Disease Diagnostic Laboratory at Washington State University.