Invasive cane toad triggers chronic physiological stress and decreased reproductive success in an island endemic

Summary

1. Understanding the mechanisms that afford invasive species their ecological success as important agents of global change is key to addressing their biodiversity impacts. Species invasions that occur on small islands are especially detrimental and suggest that invaders intensify their ecological impacts by exploiting novel ecological functions. However, it remains unknown whether such strong impacts are also a consequence of an invader's indirect effect (e.g. causing physiological stress or reproductive failure) on island species. Therefore, it is valuable to quantify the physiological mechanisms through which invasive species can exert indirect effects on the performance, and ultimately the fitness of island endemics.
2. In this study, we investigated whether the invasive cane toad (Rhinella marina) caused indirect competitive impacts on the endemic Fijian ground frog (Platymantis vitiana) on the small (60 ha) Viwa Island, Fiji. We used large (4 × 10 000 m²), natural and replicated enclosures to monitor ground frog stress hormone levels, reproductive hormone cycle, body condition, breeding and survival in the presence/absence of the cane toad. We conducted monthly sampling to analyse annual patterns in testosterone for males, estradiol and progesterone for females, corticosterone for both sexes and body condition of ground frogs in replicated enclosures or natural habitats with high/low cane toad densities. We also measured survival and reproductive success of ground frogs in enclosures.
3. Results showed that in both enclosures and natural habitats with high cane toad densities, ground frogs had a significant reduction in body condition, increased urinary corticosterone metabolites and suppressed sex steroid metabolites. Most importantly, annual field surveys showed significant reduction in ground frog reproductive success (fewer eggs were laid in enclosures with toads present); however, survival was not severely reduced.
4. Our study clearly demonstrated that on small islands, invasive species may exploit broader ecological roles with strong indirect effects that amplify their impacts beyond those seen on continents. Overall, the effects of cane toad competition had the capacity to strongly reduce ground frog reproductive success. We strongly advocate management actions that either minimize invasion or limit the strength of invasive–native species interactions (e.g. through habitat conservation) to prevent further extinctions on islands.

Introduction

Species endemic to oceanic islands are evolutionarily dynamic and are more extinction-prone than those on continents (Diamond 1975). Lower genetic and species diversity makes species on oceanic islands less resilient to environmental change (Diamond 1975; Vitousek 1988; Frankham 1998). Invasive species have caused significant impact on natural ecosystems world-wide, especially through predation (Clavero & García-Berthou 2005; Shine 2010). Invasive species have shown higher rates of successful colonization and greater ecological impacts on the islands, and island species are more sensitive to environmental change than continental species (Sax & Gaines 2008). Several factors explain why invasive species exert stronger ecological effects on the islands (Sax & Gaines 2008). First, being less diverse ecosystems with few natural enemies (e.g. predators or competitors), the islands offer lower ecological resistance affording greater invasion success (Kennedy et al. 2002; Cassey et al. 2005). Next, islands may provide invaders novel opportunities enabling broader ecological function and increased biodiversity impacts (Wanless et al. 2007). A fascinating example is on the invasive house mouse on remote Atlantic islands, where in the absence of natural enemies, mice predate large sea bird chicks, a unique ecological role with devastating consequences (Wanless et al. 2007).

The cane toad (Rhinella marina), an anuran amphibian native to Central and South America, was introduced into Australia and several Pacific island nations, for example Fiji Islands, as a bio-control agent (Easteal 1981). In Australia, cane toad invasion dynamics and ecological impacts are well documented (Shine 2010; Pizzatto, Both & Shine 2014). Cane toads possess highly toxic skin secretions that have caused major declines in many naïve Australian predators that die after eating them (Doody et al. 2009; Shine 2010). More recent evidence highlighting the ecological impact of cane toads on native anurans suggests that in anuran species that share micro-habitat with the cane toad, there is potential for transfer of lethal parasites by the toad, which can have devastating consequences on native species (Pizzatto, Both & Shine 2014). Ecological theory of island invasion suggests that on small islands, invasive cane toads can exploit novel ecological functions that diversify or intensify biodiversity impacts (Sakai et al. 2001). Indeed, recent experimental laboratory studies evaluating interactions between the cane toad and an island endemic anuran species, the Fijian ground frog (Platymantis vitiana) from Viwa Island, Fiji, indicate that cane toads induce fearfulness and physiological stress (elevated corticosterone response) in adult ground frogs (Narayan, Cockrem & Hero 2013). This novel result provides a potent underlying physiological mechanism, if present in nature, through which cane toads could cause negative ecological interactions with this small island endemic anuran species.

Physiological stress is increasingly viewed as a key mechanism in predator–prey dynamics (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013). Predators induce elevated stress hormone (glucocorticoid) levels in prey (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007), as either a fear response (i.e. predator stress hypothesis) or from predators causing the prey to decrease foraging to increase vigilance (i.e. predator-sensitive foraging hypothesis; Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013). Decreased foraging time induces nutritional stress through reduced body condition and again elevates glucocorticoid levels (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013). Physiological stress, reduced nutrition or lowered body condition can induce phenotypic dysregulation; therefore, predators can decrease the fitness of prey via reduced reproductive success or lowered survival (e.g. stress-induced immunosuppression increases the risk of infection and mortality; Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013). Ultimately, these physiological changes constitute ‘sublethal’ predation costs that do not add to a predator's direct kill rate, but inflate serious demographic impacts to prey populations (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013). In our view, if similar physiological change occurs more broadly with other adverse species interactions, including competition between invasive and native species, it greatly expands how we understand individual-based mechanisms for regulating general population dynamics.

Cane toads introduced in the 1930s now persist on multiple islands in Fiji and do so at extraordinarily high densities (>4000 toads/ha; Thomas et al. 2011). Further, they can exhibit strong spatial overlap in some habitats with the endangered Fijian ground frog suggesting shared use of prey and shelter (Thomas et al. 2011). To our knowledge, it remains unclear whether high density or resource overlap is sufficient for the cane toad to induce interspecific competition, via resource exploitation or agonistic interference, an ecological role unknown from other toad-invaded localities, for example in Australia (Doody et al. 2009). In this study, we adapted the indirect species-competitive interaction or physiological dysregulation framework based on previous predator–prey studies (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013) to evaluate the potential fitness impacts of the invasive cane toads on the ground frogs. Our conceptual eco-physiological framework (Fig. 1) predicted that when the cane toad density is elevated and resource overlap is high, then ground frogs are exposed to competition. The consequences of this cause decreased body condition from nutritional stress or increased behavioural fearfulness (Narayan, Cockrem & Hero 2013) that elevates glucocorticoid levels in ground frogs. Furthermore, chronically elevated glucocorticoid levels could inhibit reproductive physiology via suppression of steroids necessary for successful reproduction (Fig. 1) and leading to eventual demographic effects through decreased survival (Romero 2004).

Figure 1

A species interaction–physiological dysregulation framework predicts how introduced cane toads (Rhinella marina) could induce competition with endemic Fijian ground frogs (Platymantis vitiana) reducing reproduction or survival. High cane toad densities and overlapping habitat requirements invoke fear responses (Narayan, Cockrem & Hero 2013) and reduced nutritional intake in ground frogs to cause physiological dysregulation. Elevated (+) ground frog glucocorticoid (GC) stress hormones and reduced (−) body condition could suppress physiological pathways regulating reproductive success and survival.

To test the multi-level predictions of this eco-physiological framework (Fig. 1), we evaluated interactions between ground frogs and cane toads in very large experimental field enclosures. Four natural enclosures (10 000 m²) were each stocked with ground frogs (50 adult males and 50 adult females/enclosure) and two of them also with adult cane toads (400 toads/enclosure). This approach enabled us to accurately control the cane toad density and manipulate the strength of species interactions representative of Viwa Island. Grounds frogs were then monitored for annual differences in proximate eco-physiological indices and fitness attributes. We measured ground frog body condition, physiological stress and reproductive hormones in the presence (at ecological relevant densities; Thomas et al. 2011) and absence of cane toads. We used non-invasive hormone monitoring tools for evaluating the reproductive and stress endocrine functioning in the frogs (Narayan 2013). Concurrently, we also measured the effects of cane toads on fitness of ground frogs by quantifying annual differences in survival and female reproductive success (fecundity) in the presence and absence of cane toads. Reduced reproductive success or survival in ground frogs exposed to cane toads could indicate the strength of competition that cane toads could exert on native ground frogs. Finally, to ensure our experimental results did not simply arise because of ‘enclosure effects’, we also conducted physiological sampling of ground frogs within open natural habitats. By using different vegetation communities that influenced habitat choice and hence densities of toads and ground frogs across our study site, it was possible to assess physiological data from ground frogs at naturally occurring high and low cane toad densities (Thomas et al. 2011).

Materials and methods

Study site

Fijian ground frogs and cane toads were studied on Viwa Island (18° S, 175° E), a 60-ha island located 900 m east off the coast of Viti Levu, the largest main island of Fiji. We quantified how ground frogs responded physiologically in the presence and absence of cane toads using four large square experimental enclosures (10 000 m², 100 m × 100 m) constructed from high-density polyethylene fencing material (thickness of 1·5 mm), to a height of 2 m. Fencing prevented the escape of ground frogs and cane toads from enclosures, but permitted insect prey to enter. All ground frogs and cane toads confined within the newly built enclosures were hand-captured and removed prior to experiments. We assigned these four enclosures into (i) two cane toad and ground frog (i.e. competition) enclosures and; (ii) two ground frog only (i.e. control) enclosures. Competition enclosures were each stocked with 100 adult ground frogs (50 adult male and 50 adult females) and 400 adult cane toads (200 males and 200 females). The stocking density of one adult ground frog to four adult cane toads was based on previously reported abundance estimates for the natural habitat on Viwa Island that the enclosures were built upon (Thomas et al. 2011). Control enclosures were only stocked with 100 adult ground frogs (50 adult male and 50 adult females). All experimental anurans were hand-captured across three nights from habitats adjacent to the enclosures. Prior to release into the enclosures, each ground frog received a passive integrated transponder (PIT) tag to permit identification of individuals within each experimental enclosure throughout the repeated monthly sampling over 13 months (April 2010–April 2011).

Ground frog monitoring in enclosures

Each month, 3 days of nocturnal monitoring were conducted between 18·00 and 21·00 h to hand-capture and collect ground frog urine samples (for non-invasive hormone quantification) and also to record morphometrics of all individually tagged ground frogs in each experimental enclosure (~ 400 frogs/month). For urine sampling, each frog was captured by hand as soon as they were located on the ground and a urine sample collected (within 1 min post-capture) by gently massaging the underbelly abdomen (Narayan et al. 2010a,b). Individually referenced ground frog urine samples were transferred into an insulated container in the field with dry ice (−80 °C) and then kept frozen (−20 °C) in the laboratory prior to hormone analysis. To calculate ground frog body condition, snout-vent length (L) and body mass (M) were recorded at the time of capture (after urine collection). Fulton's index (K = M/L³) was then used to calculate the body condition. Female ground frog reproductive success (fecundity) was also measured as the cumulative monthly proportion of females producing egg clutches (maximum one clutch per female per breeding event) across the annual wet breeding season (September–April; Narayan, Christi & Morley 2008). We conducted diurnal weekly wet season surveys to locate and count ground egg clutch deposited in ground nests (three field assistants searched for three hours in each enclosure). Female frogs that had deposited the eggs were identified by the detection of mature underbelly oocytes (females that had already laid eggs showed no signs of mature oocytes; Narayan et al. 2010a,b).

Ground frog monitoring in natural habitats

To provide additional data to evaluate any potential ‘enclosure effects’ on the results, we evaluated the body condition and physiology of ground frogs in replicated densely forested and patchy vegetation communities that support low and high toad densities, respectively (Thomas et al. 2011). These two distinct island habitat types have been demonstrated to influence toad habitat preferences and lead to large spatial differences in toad densities (Thomas et al. 2011). During the peak breeding period in December 2011, we randomly captured and sampled a subpopulation of 25 adult ground frogs of each sex in each habitat type.

Hormone assays

Urinary corticosterone (both sexes), testosterone (males), estradiol (females) and progesterone (females) metabolite concentrations were measured for ground frogs using previously validated species-specific enzyme-immunoassay procedures (Narayan et al. 2010a,b). Urinary corticosterone metabolite concentrations were determined using a polyclonal anti-corticosterone antiserum (CJM06) diluted 1 : 45 000, horseradish peroxidase-conjugated corticosterone label diluted 1 : 120 000 and corticosterone standards (1·56–400 pgwell⁻¹). Cross-reactivity of the CJM06 anti-corticosterone antiserum was 100% with corticosterone, 14·25% with desoxycorticosterone and 0·9% with tetrahydrocorticosterone (Narayan et al. 2010b). Urinary testosterone metabolite concentrations were measured using a polyclonal anti-testosterone antiserum (R156/7) diluted 1 : 25 000, horseradish peroxidase-conjugated testosterone label diluted 1 : 40 000 and testosterone standards (0·78−200 pgwell⁻¹). Concentrations of urinary estradiol metabolites were determined using a polyclonal anti-estradiol antiserum (R522/2) diluted 1 : 45 000, horseradish peroxidase-conjugated estradiol glucuronide label diluted 1 : 45 000 and estrone glucuronide standards (0·39−100 pgwell⁻¹). Concentrations of urinary progesterone metabolites were determined using a monoclonal anti-progesterone antiserum (CL425) diluted 1 : 15 000, horseradish peroxidase-conjugated progesterone label diluted 1 : 40 000 and progesterone standards (0·39−100 pgwell⁻¹). The antibody cross-reactions have been reported for testosterone (R156/7) as <1% for any hormone other than dihydrotestosterone or testosterone (Ginther, Ziegler & Snowdon 2001; de Catanzaro et al., 2003; Szymanski et al., 2005), for estrone (R522/2) as <0·1% with oestradiol-17β and over 100% with estrone (Munro et al., 1991), and for progesterone (CL425) as >50% with most 4-pregnene and 5-α pregnan metabolites (Graham et al., 2001; Szymanski et al., 2005).

The plates were coated with 50 μL of antibody in enzyme-linked immunosorbent assay (ELISA) coating buffer (50 mm bicarbonate buffer, pH 9·6) and incubated for at least 12 h overnight at 4 °C. For all assays, standards, internal controls and urine samples were diluted in EIA buffer (39 mm NaH₂(PO₄)₂H₂O, 61 mm NaHPO₄, 15 mm NaCl and 0·1% bovine serum albumin, pH 7·0). For all assays, 50 μL of standards, internal controls and urine samples was added to each well of the coated Nunc MaxiSorp^™ plates. About 50 μL of the corresponding horseradish peroxidase label was then added to each well, and the plates incubated at room temperature for 2 h. Plates were washed and 50 μL of a substrate solution (0·01% tetramethylbenzidine and 0·004% hydrogen peroxide in 0·1 m acetate citric acid buffer, pH 6·0) was added to each well. Stopping solution (50 μL of 0·5 mol/L H₂SO₄) was added immediately after 10-min incubation at room temperature. Non-specific binding was accounted for by subtracting the blank absorbance from each reading. Standard curves were generated and a regression line fitted by the method of least squares and used to determine hormone concentrations in the frog urine samples. Intra- and interassay coefficient of variation for corticosterone, testosterone, estradiol and progesterone urinary metabolites were 6·1 and 4·5, 6·8 and 4·2, 6·2 and 5·5, 6·5 and 5·2%, respectively. All steroidal concentrations were presented as mean ± standard error (SE) pg/μg creatinine. Creatinine was measured using the Jaffe method explained in detail earlier by Toora & Rajagopal (2002) and used in our earlier studies on ground frogs (Narayan et al. 2010a). Creatinine reactions were done on ordinary flat-bottom plates. Standard values used (including zeros) were 500, 250, 125, 62·5 and 31·25 ng/well. Two hundred and forty microlitres of standard working stock (10 μg/mL or 500 ng/well) was serially diluted (2-fold) in a glass tube by using 120 μL stock plus 120 μL Milli-Q water and repeated for next standard. Tube for zeros contained 120 μL Milli-Q water. Ground frog urine samples were diluted 1 : 4 by adding 30 μL of neat urine sample to 90 μL Milli-Q water. For plate loading, 50 μL of standard and sample was pipetted per well, according to the plate map. Speed of the addition was unimportant as this was not a binding assay. Alkaline picrate reagent was prepared immediately before use by combining 4 mL Milli-Q water, 4 mL 0·75 N NaOH and 4 mL 0·13% picric acid and mixed well. Hundred microlitres of the alkaline picrate reagent was added to all wells that contained standard or sample. The plate was tapped briefly to mix and incubated at room temperature for 30 min. Plates were read at 490 nm and optical density of 0 wells was expected to be around 0·2.

Statistical analyses

General and generalized (i.e. data drawn from non-Gaussian distributions) linear mixed effect models were used to analyse all data obtained from ground frogs sampled repeatedly in enclosures with and without toads (Zuur et al. 2009). Models for physiological data (all four hormones and morphometrics data (body condition)) were fitted with Gaussian error distributions, whilst binary survival and reproductive success data were fitted with binomial error distributions and a logit link. Random effects to account for repeated individual measures and enclosure identity were included in each model. The working correlation matrix of all mixed models was fitted with an autoregressive (AR1) error structure to account for temporal non-independence among successive monthly physiological measures of ground frogs (Zuur et al. 2009). We considered the effects of enclosure treatment, month and their interaction in their respective data in each model. To assess the physiological responses of ground frogs in the natural habitats, we used general linear models to evaluate the effects of habitat type (patchy and forested habitat types that covaried with toad density) on the physiological data. Statistical analyses were conducted using systat (version 13.0, Bangalore, India) or IBM spss (version 19, NSW, Australia).

Results

Physiological responses of ground frogs to toads in enclosures

In the enclosures stocked with cane toads, ground frogs showed significantly different annual physiological profiles (Fig. 2). For male ground frogs, we observed significant effects of toad presence (F_1,200 = 40·87, P < 0·001), time (month) of sampling (F_12,1275 = 2·98, P < 0·001) and their interaction (F_12,1275 = 39·03, P < 0·001) on body condition (Fig. 2a). For female ground frog body condition, we did not observe a significant effect of toad presence (F_1,200 = 1·44, P = 0·232), but month of sampling (F_12,1268 = 23·70, P < 0·001) and interaction between toad presence and month (F_12,1268 = 130·73, P < 0·001) indicated significant effects on female body condition. Significant interactions indicated that adult body condition of both sexes progressively decreased over time in the presence of toads compared to controls (Fig. 2a).

Figure 2

Mean (±SE) annual responses of male (circles) and female (triangles) ground frog in experimental enclosures where toads are present (red symbols and lines) or absent (clear symbols and lines) depicting: (a) body condition, (b) urinary corticosterone metabolite concentrations, (c) urinary testosterone metabolite concentrations, (d) urinary progesterone metabolite concentrations, (e) urinary estradiol metabolite concentrations and (f) proportion of female laying eggs.

Baseline urinary corticosterone metabolites in male and female ground frogs were greatly elevated in the presence of cane toads compared to toad-free enclosures (Fig. 2b). In male ground frogs, we observed significant effects of toad presence (F_1,200 = 40·85, P < 0·001), month of sampling (F_12,1276 = 2·98, P < 0·001) and their interaction (F_12,1276 = 39·031, P < 0·001) on urinary corticosterone metabolite levels. Male frogs in enclosures without toads showed high levels of corticosterone metabolites at the start of and during the breeding season (August 2010–January 2011; Fig. 2b), and mean levels returned to baseline from December onwards (Fig. 1d). In contrast, male frogs living in enclosures with toads showed increasingly high mean monthly level of corticosterone metabolites, with maximum levels recorded throughout the breeding period, remaining elevated through to April 2011 (Fig. 2b).

For female ground frogs, urinary corticosterone metabolite levels also indicated a significant effect of toad presence (F_1,200 = 9782·70, P < 0·001), month of sampling (F_12,1283 = 1204·93, P < 0·001) and interaction between toad presence and month (F_12,1283 = 944·40, P < 0·001) on female urinary corticosterone metabolite levels. All females also showed no significant difference in mean urinary corticosterone metabolite level at first capture in April 2010 (Fig. 2b). For the females living in enclosures without toads, mean level of urinary corticosterone metabolite decreased within 1 month of transfer into the enclosures and remained nominal in June (Fig. 2b). Female frogs without toads showed a concomitant rise in urinary corticosterone metabolites between July and November, but mean levels returned to baseline afterwards (Fig. 2b). In contrast, female frogs living in enclosures with toads showed increasingly high mean monthly level of urinary corticosterone metabolites throughout the study (Fig. 2b).

Analysis of urinary testosterone metabolites in male ground frogs showed significant treatment (F_1,198 = 8374·96, P < 0·0001), time (F_12,1143 = 2246·06, P < 0·0001) and interaction effects (F_12,1143 = 1251·81, P < 0·0001; Fig. 2c). Male ground frogs in enclosures with toads produced low mean monthly urinary testosterone metabolite concentrations of <50 pg/μg Cr throughout the study (Fig. 2c). For control males, mean concentrations of urinary testosterone metabolites (<50 pg/μg Cr) between April 2010 and July 2010 rose to a distinct peak of 355·5 ± 16·00 pg/μg Cr in December 2010, declined whilst remaining high (>100 pg/μg Cr) in January 2011 and returned to baseline levels in April 2011.

For urinary estradiol metabolites of female ground frogs (Fig. 2d), there were significant effects of toad treatment (F_1,200 = 8602·36, P < 0·0001), time (F_12,1280 = 1813·19, P < 0·0001) and their interaction (F_12,1280 = 1130·38, P < 0·0001). Female ground frogs in toad enclosures produced much lower levels of urinary estradiol metabolites (<75 pg/μg Cr) throughout the study. For female ground frogs in control enclosures, concentrations of estradiol metabolite were low (<300 pg/μg Cr) from April to July, then rapidly increased and were high (ranging from 600 to 942 pg/μg Cr) from August 2010 to December 2010. The peak level of mean urinary estradiol metabolites was recorded during December 2010 (942·63 ± 146·26 pg/μg Cr). Mean urinary estradiol metabolite concentrations declined in January 2011 and remained at baseline or nominal levels during the final sampling period in April 2011.

For female urinary progesterone metabolites (Fig. 2e), again, there were significant effects of toad treatment (F_1,200 = 789·70, P < 0·0001), time (F_12,1252 = 272·04, P < 0·0001) and interaction (F_12,1252 = 242·55, P < 0·0001). For enclosures with toads, urinary progesterone concentrations of female ground frogs did not show any distinct annual pattern at all and remained low (<52 pg/μg Cr) throughout the study. Female frogs in enclosures without toads showed a seasonal pattern of mean urinary progesterone metabolites with increasing mean levels starting from June to July, reaching peak concentrations between November and January (peak levels ranging from 155 to 177 pg/μg Cr).

Fitness responses of ground frogs to toads in enclosures

There were no effects of toad treatment (Wald χ² = 0·00, P = 1), time (Wald χ² = 0·00, P = 1) and their interaction (Wald χ² = 0·00, P = 1) affecting adult ground frog survival in the experimental enclosures. Thus, ground frogs maintained high rates of survival independent of experimental effects and duration. In contrast and commensurate with effects on sex steroids, there was a significant effect of toad treatment (Wald χ² = 113·32, P < 0·001), time (Wald χ² = 1849·32, P < 0·001) and their interaction (Wald χ² = 2011·24, P < 0·001; Fig. 2f) on female reproductive success (percentage of females laying an egg clutch). Again, in the presence of cane toads, female ground frogs had much lower annual reproductive success with only 15% of females capable of producing egg clutches compared to 78% observed in control females.

Physiological responses of ground frogs to toads in natural habitats

Fewer cane toads were observed in the densely forested vegetation habitats compared to the ground frogs in the three consecutive nights of sampling (mean of n = 25 toads and 70 ground frogs sighted per night). Whilst in the patchy vegetation habitats, we found higher numbers of cane toads compared to the ground frogs (mean of n = 30 ground frog sighted c.f. 115 cane toads sighted each night). Thus, in the patchy forest, the density ratio of ground frog to cane toad was strongly toad-biased (1 : 3·8), whilst in the closed forest, the density ratio was strongly ground frog-biased (2·8 : 1), which was also indicative of each species’ preferred habitat preferences (Thomas et al. 2011). Our experimental enclosure results were strongly supported by similar physiological responses observed in adult ground frogs living in natural habitats that too covaried in toad density (Fig. 3). Body condition (Fig. 3a) was again significantly reduced in male ground frogs (F_1,48 = 278·41, P < 0·0001), resident in the patchy forest with high toad densities compared to males in closed forest which had lower toad densities. There was no similar body condition response in adult female ground frogs (F_1,48 = 1·19, P = 0·28). Male (F_1,48 = 675·46, P < 0·0001) and female ground frogs (F_1,48 = 3036·56, P < 0·0001) resident in the patchy forest with high toad densities had significantly elevated urinary corticosterone metabolite levels (Fig. 3b) compared to frogs in closed forest. Male ground frogs resident in the patchy forest with high toad densities had significantly lower urinary testosterone metabolite levels (F_1,48 = 3556·64, P < 0·0001; Fig. 3c) compared to male frogs exposed to low toad densities in closed forest. Similarly females had significantly lower urinary estradiol metabolites (F_1,48 = 2355, P < 0·0001) and progesterone metabolite levels (F_1,48 = 207·85, P < 0·0001) in the patchy forest with high toad densities compared to female residents in closed forest with low toad densities (Fig. 3c).

Figure 3

Mean (±S.E.) annual responses of male (black shaded bars) and female (unshaded bars) ground frog sampled from intact and patchy habitats (low and high toad density respectively): Panel A) is body condition, Panel B) is urinary corticosterone metabolite concentrations, and Panel C) are urinary sex steroid metabolite concentrations. Asterix denote significant effects of habitat/cane toad density differences on male and female ground frog physiological metrics.

Discussion

Higher rates of extinction of species on oceanic islands attest to their greater sensitivity and vulnerability to anthropogenic-induced environmental change (Diamond 1975; Whittaker & Fernández-Palacios 2007). Invasive species are a major cause of island species extinctions and more generally the current global biodiversity crisis (Chapin et al. 2000; Sax & Gaines 2008). The strength and diversity of ecological interactions that invasive species exploit within novel environments of small islands underpins their biodiversity impacts (Vitousek 1988; Sax & Gaines 2008). We have demonstrated experimentally that cane toads introduced onto a small island of Viwa, Fiji, can exert broad-scale physiological dysregulation in the endemic Fijian ground frogs. In doing so, the cane toads can exploit a hitherto unreported novel ecological function that ultimately causes lower reproductive success in ground frogs. Similar physiological mechanisms and consequences of invasion are absent in Australian animals, despite cane toad impacts being thoroughly investigated under both natural and experimental conditions (Doody et al. 2009). To some extent, our results indicate how small islands relative to continents could facilitate invasive species to have increased or novel biodiversity impacts (Kennedy et al. 2002; Sax & Gaines 2008). Future research using replicated study sites on both small islands and the mainland could be undertaken to test how islands facilitate different impacts of cane toads on native species.

Predators can induce indirect fitness costs in prey by causing aberrant functioning to traits that influence survival and reproduction (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Clinchy, Sheriff & Zanette 2013). Indeed, such indirect fitness costs are thought to be a major component of a predator's effect on prey demography (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Clinchy, Sheriff & Zanette 2013). Our results provide a mechanistic basis for potential physiological changes that occur as a result of prolonged interactions between the cane toads and ground frogs. These effects are complex and involve changes to the major endocrine pathways (at the level of the hormones of the stress and reproductive hormonal axis). Furthermore, nutritional stress caused by potential resource competition leads to poor body condition that also increased glucocorticoids. This causes downstream effects on reproductive fecundity, which is the most significant finding of our study. It suggests that sublethal physiological mechanisms, as a result of invasive species impact on native species, could have potent consequences for individual fitness and impact demographic function (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013).

Sustained high annual levels of urinary corticosterone metabolites indicated chronic stress, a key mechanism through which cane toads, at ecologically relevant densities (Thomas et al. 2011), could mediate their competitive effects and ensuing fitness loss in ground frogs. Ground frogs competing with cane toads in experimental enclosures maintained chronically elevated corticosterone levels across the experiment. This response is well known to suppress reproduction via inhibitory effects to the hypothalamus–pituitary–gonadal (HPG) axis (Romero 2004). Whilst chronic stress can also suppress immune and gastrointestinal function that lowers survival (Romero 2004), this was not evident in adult ground frogs given high annual survival rates were maintained in experimental enclosures containing toads. Most importantly, using enclosures enabled us to quantify how cane toad-induced competition, via chronic stress, induced physiological dysregulation and ultimately impacted ground frog condition and fitness by reducing female reproductive success. Our results support the view that lower reproductive success, rather than reduced adult survival, is the most likely pathway through which adverse species interactions impact prey or competitor fitness (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Clinchy, Sheriff & Zanette 2013).

Elevation of ground frog glucocorticoid hormone concentrations through ecological interactions with cane toads could arise through two mechanisms. Reduced body condition in ground frogs exposed to toads suggested density-dependent exploitative competition with toads for shared resources such as prey, which causes nutritional stress and elevates glucocorticoid levels (Boonstra, Singleton & Tinnikov 1998; Creel et al. 2007; Travers et al. 2010; Anson et al. 2013; Clinchy, Sheriff & Zanette 2013). Secondly, cane toads could trigger fear in ground frogs, causing a chronic stress response. Such an assertion is supported by elevated glucocorticoid responses measured in ground frogs placed in close proximity to cane toads under experimental conditions (Narayan, Cockrem & Hero 2013). This suggests that interference competition could occur between these two species. Interference competition arises when interacting, but size mismatched, species results in the larger competitor (i.e. cane toad) using aggressive behaviour to cause fear, physiological dysregulation, injury or death of smaller competitors (this is most possible for the male frogs because female ground frogs are large in size and maintained body condition during breeding in the presence of cane toads; Figs 1 and 2; Palomares & Caro 1999). Hence, increased glucocorticoid levels could arise from a reduced body condition causing nutritional stress or fear from interference competition. Further, synergies between these two stressors could be expected to suppress sex steroid levels and impact reproductive physiology and ultimately cause fitness loss in ground frog. Adding food supplementation to enclosures could provide a means to experimentally test the relative influence of each of these competition stressors (nutritional stress and fear) for elevating glucocorticoid levels in ground frogs (Clinchy et al. 2004).

Earlier, Creel et al. (2007) demonstrated that reproductive hormone (progesterone) and calf production in the Greater Yellowstone Ecosystem elk (Cervus elaphus) were negatively correlated with the increased risk of predation by wolves (Canis lupus). The presence of a predator increases vigilance in a prey, thus also reducing foraging activity, causing diet imbalances and loss of weight in prey animals. In our study system, potential competitive interactions between cane toads and ground frogs led to suppression of reproductive hormones through elevated physiological stress caused by nutritional stress and fear. Prolonged exposure to harsh environmental stimuli increases physiological stress and reduces fitness in wildlife as explained by the ‘Cort-Fitness’ hypothesis (Bonier et al. 2009). Our results support the Cort-Fitness hypothesis as sustained, and high corticosterone concentrations were associated with suppressed reproductive hormone levels (testosterone, estradiol and progesterone) in both sexes, as well as body condition (for male ground frogs). Sherriff, Krebs & Boonstra (2009) demonstrated in free-living female snow hares (Lepus americanus) that under the event of high predation risk (stimulated using visual exposure to dogs), females expressed high faecal glucocorticoid metabolite levels and subsequently gave birth to smaller and lighter offspring. Our results clearly indicated that prolonged exposure to cane toads disrupted the endocrine physiology of ground frogs and contributed to much lower reproductive success. It is plausible that survival was not affected because predators or pathogens were either rare or absent during our study. Longer-term monitoring using our experimental design would be necessary to understand whether survival impacts of cane toads on ground frogs could arise.

On Viwa Island, interactions between cane toads and ground frogs are common, and frogs usually evade toads by retreating into terrestrial substrates, such as leaf litter, or moving up into the vegetation by climbing trees (Ryan 1984; Narayan, Christi & Morley 2008). Behavioural ‘escape’ responses could be facilitated by acute corticosterone stress responses during a visual encounter between the ground frogs and cane toads (Narayan, Cockrem & Hero 2012, 2013). Furthermore, it is likely in densely forested habitats on Viwa Island that ground frogs reduce risk of competitive interactions with toads by having increased access to microhabitat refuges (i.e. evasion) and because toad are in lower densities in this habitat. However, as preferred ground frog habitat is increasingly lost through anthropogenic activities, such as deforestation, interactions between ground frogs and cane toads will increase. If so, an increased frequency of ground frogs with a prolonged elevation of corticosterone (as shown in our study) and suppression of reproductive hormones could cause loss of reproduction sufficient to maintain population growth. Additionally, increased interactions with cane toads could also increase the risk of novel pathogenic infections, such as chytridiomycosis, which could impact the survival of ground frogs (Narayan, Molinia & Hero 2011).

We accept that our enclosures may exacerbate species interactions because they do not account for habitat variation, which would influence spatial variation in competition due to mismatches in species' resource requirements (e.g. reproductive or shelter resources), or permit behavioural avoidance strategies that could reduce competitive interactions (Palomares & Caro 1999; Amarasekare & Nisbet 2001). However, we have provided supportive evidence ruling out potential ‘enclosure effect’ by showing similar impacts on physiological functions of ground frogs living in ecologically more complex island habitats (Fig. 3). Ecologically, intense toad competition should also have important demographic implications for ground frogs through reduced recruitment and decreased population growth. Our results provide an insight into the greater ecological influence of invasive species on small islands having larger and novel biodiversity impacts relative to the continents. This implies management actions to mitigate invasions on the islands must be both rapid and thorough to prevent establishment of invasive species and limit extinction on islands.

Our study also demonstrates the success of excluding cane toads by using simple enclosures within their natural habitat to enhance breeding success of ground frogs. Based on our findings, we identify several key advantages of this in situ enclosure system for conservation and management on Viwa: (i) it provides particularly the young froglets, immediate release from competition and predation from the cane toads; (ii) there is no need for any translocation of frogs off the island; (iii) the enclosures can be easily monitored with assistance from the local community; (iv) the recovery of the frog population avoids the risk of disease transmission (via re-introductions); (v) water quality and biodiversity on the islands may improve when cane toads are not present; and (vi) enclosures are relatively affordable to build and maintain in comparison with sophisticated captive-breeding facilities. The Viwa Island ground frog population could be used as a reserve or genetically resilient population for future captive-breeding programmes without risk of translocating diseased animals into captivity (Narayan, Christi & Morley 2009). Overall, the present study provides a good example of a cost-effective community-based initiative for endangered species conservation and management of an invasive, non-native species within natural habitats.

Acknowledgements

We thank the people of Viwa Island for helping with the field experiments, and Viwa youths (Inoke Basoko and Taina Malo) for managing the day-to-day fieldwork. This study was funded by the Rufford Small Grants for Nature Conservation (Booster Grant #10205-B). We thank the people of Viwa Island and Department of Environment, Fiji Islands, for permitting us to conduct the field sampling. EJN conceptualized the research and conducted the field work. EJN conducted the laboratory assays. TSJ and EJN conducted the analyses and manuscript writing.

Data accessibility

All data analysed in this study are available online in supporting information (see Table S1).