Alarm cues experienced by cane toad tadpoles affect post-metamorphic morphology and chemical defences


*Correspondence author. E-mail:


  • 1In many anuran species, larvae modify their developmental trajectories and behaviour in response to chemical cues that predict predator risk. Recent reviews highlight a dearth of studies on delayed (post-metamorphic) consequences of larval experience.
  • 2We raised cane toad (Bufo marinus) tadpoles either under control conditions or in the presence of non-lethal predator cues (crushed conspecifics).
  • 3Exposure to these chemical cues massively reduced size at metamorphosis, as predicted by theory. Parotoid glands were larger relative to body size in post-metamorphic animals from the experimental treatment, suggesting higher investment in chemical defences.
  • 4Exposure to chemical cues from crushed conspecifics during larval life reduced total bufadienolide content of metamorphs, but increased amounts of one specific bufadienolide (bufalin).
  • 5Hence, cane toads respond to perceived predation risk in the aquatic environment by metamorphosing at a smaller size and modifying their investment in defensive toxins during post-metamorphic life.
  • 6Phenotypically flexible responses to larval conditions vary among amphibian taxa, and can involve significant carry-over effects into post-metamorphic life.


The rate and route of an individual's ontogenetic progression from the egg through to adult life can be affected by many aspects of its environment. Factors such as food supply and temperature can substantially modify not just the rate of growth, but also the relative investment of resources into competing functions (Forsman & Lindell 1991; Leips & Travis 1994; Arendt & Wilson 1997; Pechenik, Wendt & Jarret 1998; Billerbeck, Lankford & Conover 2001; Weber 2003; Ghalambor, Reznick & Walker 2004; Hoverman & Relyea 2007). An extensive literature also reveals that developmental processes can be sensitive to risk levels, with chemical cues from predators or injured conspecifics eliciting shifts in behaviour, morphology and developmental timing in the larvae of a wide range of aquatic species (Werner & Gilliam 1984; Werner 1986; Crowl & Covich 1990; Skelly & Werner 1990; Peckarsky et al. 1993; Relyea 2001a; Van Buskirk 2002). Amphibian larvae have been used as model systems for many such studies, generating an extensive body of theory on the ways in which predator-risk cues modify amphibian ontogeny (Wilbur & Collins 1973; Werner 1986; Ludwig & Rowe 1990; Rowe & Ludwig 1991; Abrams & Rowe 1996) as well as extensive empirical work (Skelly & Werner 1990; Laurila, Kujasalo & Ranta 1998; Chivers et al. 1999; Nicieza 2000; Van Buskirk & Schmidt 2000; Babbitt 2001; Relyea & Hoverman 2003). Recent reviews of this literature highlight the complexity and species-specificity of induced responses, identify conflicts between theory and data, and call for research on a wider range of taxa and variables (Benard 2004; Relyea 2007).

In aquatic environments, chemoreception may be one of the most effective sensory modalities for detecting predators. A variety of amphibians, fish and insects respond to chemical cues from injured conspecifics, presumably reflecting the probability that major injury to a conspecific is a reliable predictor of clear and present danger to the signal receiver (Chivers & Smith 1998; Brönmark & Hansson 2000; Rajchard 2006). Indeed, conspecific responses may vary even among different types of predator-induced cues. For example, cues from injured conspecifics may not necessarily provide the prey with the same information about predation risk as do cues from consumed prey, and hence may elicit different responses (Turner 1996; Jacobsen & Stabell 2004; Schoeppner & Relyea 2005). Most chemical cues produced during predation events are organic compounds that degrade over time (Loose, Vonelert & Dawidowicz 1993; Turner & Montgomery 2003; Peacor 2006). The larval response may facilitate predator avoidance via modifications of morphology (Travis, Keen & Julianna 1985; Van Buskirk, McCollum & Werner 1997; Van Buskirk & Relyea 1998; Lardner 2000; Van Buskirk & Schmidt 2000), of behaviour (Skelly 1994; Relyea 2001a) or life-history traits (e.g. age and/or size at metamorphosis: Laurila et al. 1998; Kiesecker et al. 2002). Predation risk experienced in larval life can affect post-metamorphic morphology (e.g. changed leg length in ranid frogs: Relyea 2001b; Van Buskirk & Saxer 2001) and chemistry (increased production of bufadienolide toxins in Bufo boreas: Benard & Fordyce 2003). Any modifications to morphology, behaviour or life-history traits may incur profound fitness costs (e.g. slow growth and development: Skelly 1992; Van Buskirk 2000; Relyea 2002a,b), so their prevalence suggests equally high fitness benefits may accrue also. Their prevalence also may reflect the fact that making a mistake (i.e. not inducing defence) results in the ultimate cost, death.

Cane toads (Bufo marinus, Linnaeus 1758 – allocated to Chaunus or Rhinella by some authorities: Frost et al. 2006; Pramuk et al. 2008) produce bufadienolides that are concentrated in the parotoid glands (Lever 2001). These compounds are toxic to many predators and hence, function as a defence (Lever 2001; Benard & Fordyce 2003). As in many other bufonid species, tadpoles of cane toads react strongly to chemical cues from crushed conspecifics (Hagman & Shine 2008). Toad larvae flee from such chemicals, suggesting that they perceive the cue as an indication of danger (Hagman & Shine 2008). To evaluate whether the presence of such chemical cues also induces shifts in life-history traits (age and size at metamorphosis), in post-metamorphic morphology (specifically, relative size of the toxin-containing parotoid glands) and in chemistry (total concentration and relative abundance of bufadienolides), we exposed cane toad tadpoles to alarm cues throughout larval life, and monitored the consequences.

Materials and methods

husbandry and breeding

Cane toads are native to South and Central America, but were introduced to Australia in 1935 in an attempt to control insect pests of sugar cane (Lever 2001). Since then the toads have spread throughout most of tropical Australia, reaching the Darwin area in 2005. We obtained four clutches of eggs for the current study, from captive toads originally collected in the Townsville region of Queensland (two pairs) and the Darwin area of the Northern Territory (two pairs). We induced breeding by subcutaneous injections of 0·5 ml (males) and 0·75 ml (females) of leuprorelin acetate (Lucrin, Abbott Australasia) diluted in amphibian ringer's solution to a concentration of 0·25 mg/ml. We placed injected pairs in breeding containers (one pair per container) and left them overnight to spawn. In the morning, we collected 36 eggs from each of the four clutches and assigned them to either control (n = 18) or treatment groups (n = 18). Subgroups of six siblings were then added to 24 plastic containers (17 × 12 × 7 cm), each with 1·2 L of conditioned tap water. We placed these containers at the same level in a Latin-square design under a set of fluorescent tubes kept on a 12 : 12 h (dark : light) photoperiod, with water temperature of 30 °C. We fed the tadpoles with boiled lettuce ad libitum, and changed the water weekly.

stimulus preparation

As the experimental stimulus, we used chemical cues from crushed cane toad tadpoles. To prepare the stimulus we macerated 0·2 g of tadpoles (rapid crushing ensured that they were instantly killed) in 60 ml of water, which was then filtered (Hagman & Shine 2008). Each day we dispensed 5 ml of freshly prepared stimulus into each of the 12 treatment containers using a syringe. For controls we dispensed water only.

effects of alarm pheromones on timing and size at metamorphosis

We recorded the number of days it took for the tadpoles to reach metamorphosis, and when metamorphosis was complete we also measured the animals’ SU (snout to urostyle) lengths with digital callipers.

effect of alarm pheromones on post-metamorphic investment into chemical defences

After metamorphosis we kept three toadlets from each of the 24 groups (i.e. nine individuals per treatment from each clutch) for further analysis. We kept these toadlets separately by group (i.e. we housed three individuals from each group together) in 54-L containers (45 × 40 × 30 cm) on a substrate of moist sand and water, with shelter items for refuge. The enclosures were kept at the same shelf height in the lab, and illuminated by fluorescent tubes (photoperiod 12 : 12 h light : dark cycle), with relative humidity at 75% and ambient temperature at 32 °C. The toads were fed live crickets ad libitum. We cleaned the enclosures weekly and changed the water daily. After 60 days the toads were euthanized and their length (SU) and the diameter of the parotoid glands were measured with digital callipers. Parotoid glands are slightly oval and we measured the longest dimension.

To identify and quantify specific bufadienolides, 23 metamorphs per treatment were homogenised (IKA T10 Basic Ultra Turrax with S10N-5G dispersion tool) in water (1 ml). Each homogenate was then made up to equal volumes of water and n-butanol and phases left to separate at 4 °C overnight. Samples were filtered using filter aid (Celite), divided into water and n-butanol fractions, and evaporated to dryness in vacuo at 40 °C. The mass of each fraction was determined and n-butanol fractions were diluted to a final concentration of 2 mg/ml with methanol. High performance liquid chromatography (HPLC) was performed using an Agilent 1100 Series Separations Module equipped with Agilent 1100 Series Diode Array Detectors and running ChemStation Rev. B0201SR1 software. We performed electrospray ionisation mass spectrometry (ESI-MS) using an Agilent 1100 Series Separations Module equipped with an Agilent 1100 Series LC/MSD mass detector. HPLC gradient conditions were as follows: 1 ml/min gradient elution from 90% H2O/MeCN (0·01% TFA) to 50% H2O/MeCN (0·01% TFA) over 20 min, then to MeCN (0·01% TFA) over 5 min, followed by a 5 min flush with MeCN (Fig. 1). The HPLC-DAD-MS gradient was identical to above, but for the use of 0·05% HCOOH as a modifier in the place of TFA, to minimise ion suppression in the negative mode (Table 1). The column used for both gradients was a Phenomenex Onyx Monolithic C18 100 × 4·6 mm column.

Figure 1.

Chromatogram of n-BuOH extract from a post-metamorphic cane toad (CMB-T060j-1) analysed at 297 nm. Numbered peaks have the characteristic UV–Vis spectrum of a bufadienolide (inset).

Table 1.  The compounds seen in n-BuOH extracts of 46 metamorph cane toads, the percentage of analysed samples in which each occurred, and the results of statistical tests (two-factor anovas: see text) to detect significant among-clutch and treatment effects on total concentrations of each bufadienolide
Compound No.% occurrenceMolecular weightIdentityClutch effectTreatment effect

We inferred the presence of bufadienolides by examination of the UV–Vis spectrum of each peak, with the α-pyrone ring having a distinctive absorption maximum at 297 nm (Fig. 1). Tentative identity was assigned to several of the detected components, based on their molecular weight, characteristic UV–Vis spectra, and previous reports from this and related toad species (Akizawa et al. 1994; Matsukawa et al. 1994; Steyn & van Heerden 1998; see Table 1). A calibration curve using a marinobufagin standard was prepared to quantify the bufadienolide content of each sample (Benard & Fordyce 2003). For all variables of interest, our statistical analyses incorporated clutch-of-origin as a nesting factor to allow for the non-independence of data taken from full siblings. The treatment effect was tested against the nested term rather than the residual error term. To avoid pseudo-replication within containers, we used mean values per container (nested within clutch) in our statistical analyses. Mortality accounts for the disparity between the number of individuals used and the degrees of freedom in our analyses; rates of mortality were similar between treatment and control groups.


Cane toad tadpoles from the Queensland population metamorphosed earlier than did tadpoles from the Northern Territory population, but at similar mean sizes (Fig. 2). Regardless of their location of origin, tadpoles exposed to chemical cues from crushed conspecifics metamorphosed slightly but not significantly sooner than their non-exposed siblings (F1,6 = 0·41, P = 0·54; Fig. 2a), and were much smaller at metamorphosis (F1,6 = 37·28, P < 0·001; Fig. 2b). The size disparity of metamorph toads from the control versus experimental groups (mean values of 9·4 vs. 7·6 mm for snout-urostyle length) corresponds to an experimentally imposed decrease in mean body mass of about 40% (mean values of 73·5 vs. 42·8 mg, respectively). When we included age at metamorphosis as a covariate, treatment still affected size at metamorphosis (treatment effect F1,5 = 14·93, P = 0·008; covariate F1,15 = 9·87, P = 0·007). That is, the treatment-induced reduction in body size of metamorphic toads was greater than would be expected simply from the earlier metamorphosis of these animals.

Figure 2.

Effects of exposure to alarm cues (chemicals from crushed conspecifics) during larval life on (a) age and (b) body size (snout-urostyle length) at metamorphosis of cane toads, and (c) on the size of the parotoid glands in 2-month-old metamorphs. Toads were either exposed to chemical cues from crushed conspecific tadpoles during larval life, or were exposed to control conditions. Siblings from four clutches were compared (total N = 46). Error bars show standard errors.

Exposure to alarm cues during larval life also affected the morphology of post-metamorphic toads. When measured 2 months after metamorphosis, the experimentally-exposed toads had significantly larger parotoid glands relative to snout-urostyle length than did the control animals (nested ancova with snout-urostyle length as covariate, gland length as dependent variable: F1,6 = 7·90, P = 0·03; Fig. 2c).

Analyses of chemical defences showed that treatment significantly decreased total bufadienolide content (F1,2 = 26·78, P = 0·03; Fig. 3a). However, amounts of one specific bufadienolide (bufalin) were higher in metamorphs from the experimental treatment than in their control siblings (F1,2 = 32·85, P = 0·03; Fig. 3b). Resibufogenin levels showed a broadly similar but non-significant pattern (F1,2 = 2·00, P = 0·29; Figs 3c and 4).

Figure 3.

Effects of exposure to alarm cues (chemicals from crushed conspecifics) during larval life on chemical defences of post-metamorphic cane toads. Toads were either exposed to chemical cues from crushed conspecific tadpoles during larval life, or were exposed to control conditions. Siblings from four clutches were compared (total N = 46). The graphs show (a) total bufadienolide content, (b) bufalin content, and (c) resibufogenin content. Error bars show standard errors. See text for statistical analyses of these data.

Figure 4.

Chromatograms of n-BuOH extract from post-metamorphic cane toads analysed at 297 nm. The upper trace (CMB-T085j-1) shows a control animal, the lower trace (CMB-T078j-1) shows one exposed to the alarm cue. Numbered peaks have the characteristic UV–Vis spectrum of a bufadienolide.


Recurrent exposure to chemical cues from crushed conspecifics during larval life had profound effects on the life-history of cane toads (Bufo marinus). These cues stimulated tadpoles to metamorphose at smaller body sizes than did sibling tadpoles that were not exposed to the stimulus. Theoretical models predict that larvae should escape a risky aquatic environment by metamorphosing early at the cost of being small (Wilbur & Collins 1973; Relyea 2007). However, the magnitude of the body-size effect in our study was greater than expected simply from accelerated metamorphosis, suggesting that the chemical cue slowed larval growth as well as inducing a (non-significantly) earlier transformation to terrestrial life. Behavioural responses likely were involved in this effect: because cane toad tadpoles flee from the chemical stimulus (Hagman & Shine 2008), they presumably expend more energy in movement and spend less time feeding than in the absence of that stimulus. Those changes, in turn, would reduce rates of growth. This conclusion accords well with recent reviews, which indicate that predator cues induce multiple responses in tadpoles and hence, tradeoffs among traits such as antipredator behaviours and growth rates can substantially influence overall responses to experimental treatments (Benard 2004; Relyea 2007).

Although the response that we observed (metamorphosis at smaller sizes) accords well with predictions from theoretical models (Wilbur & Collins 1973; Day & Rowe 2002), such agreement is seen in only a minority of previous empirical studies. For example, Relyea (2007) concluded that only 14% of previous work on responses of anuran larvae to predator cues documented smaller size at metamorphosis. Nonetheless, our results are not unique. For example, Skelly & Werner (1990) found that Bufo americanus Holbrook 1836 metamorphosed at smaller body sizes (but at the same age) in response to predator cues.

Exposure to alarm cues during larval life also had longer-term effects on cane toads, with 2-month-old metamorphs exhibiting larger parotoid glands following exposure, and higher levels of a specific bufadienolide. This result broadly supports a report by Benard & Fordyce (2003) that exposure to crushed-conspecific cues during larval life modified the toxin (bufadienolide) content of metamorph toads, although they measured only total bufadienolide content rather than specific components of the bufadienolide system. The parotoid glands of cane toads contain not only bufadienolides, but many other defensive compounds also (Lever 2001; R. A. Hayes & R. Capon, unpublished data) and hence, changes in gland size may reflect overall toxin content and thus, defensive ability.

Theory suggests that such a carry-over effect from larval life to post-metamorphic life would enhance fitness only if high predation rates on tadpoles were consistently associated with high predation rates on post-metamorphic toads (Benard & Fordyce 2003). The strength of such an association presumably depends upon whether the most significant predators of tadpoles are also significant predators of metamorphs, as may be the case for some taxa (e.g. wading birds) but will not be true for others (e.g. aquatic invertebrates). We know too little about the significance of alternative predator taxa, and the ways in which predation intensity varies through time, to evaluate the degree to which cane toad populations satisfy this prediction (Benard & Fordyce 2003). Our results differ from those of Benard & Fordyce (2003) in that we found effects on body size in addition to carry-over effects from larval life to post-metamorphic life. We are aware of only three previous studies on how predation risk in the larval phase may affect post-metamorphic morphology in anurans: leg length was affected by larval predator environments in two species of ranid frogs (Relyea 2001a,b; Van Buskirk & Saxer 2001), but not in a hylid (Relyea & Hoverman 2003). It is likely that the induction of small size at metamorphosis imposes fitness costs (Benard & Fordyce 2003; Laurila et al. 2004; Relyea & Auld 2004). For example, small size might render metamorphic toads more vulnerable to predation and at greater risk of desiccation (Lever 2001). Size at metamorphosis also can affect post-metamorphic performance. Larger metamorphic size can persist for up to two years and has been positively correlated with both juvenile survival and with reproductive success (Howard 1980; Berven 1981, 1982; Berven & Gill 1983; Smith 1987; Semlitsch, Scott & Pechmann 1988; Gerhardt 1994; Altwegg & Reyer 2003; Relyea 2007).

Our results accord well with models of metamorphosis (i.e. what size to metamorphose: Benard 2004; Relyea 2007), but are unique in the combination of responses that we found (i.e. same time to metamorphosis, smaller size at metamorphosis, changed post-metamorphic morphology, and modified investment into specific toxic chemical defences). Such diversity in specific details of response to alarm cues is to be expected. Not only do anurans encompass an enormous range of ecologies and physiologies, but the effects of predation risk on metamorphosis can depend on environmental context as well as phylogeny (Relyea 2007). Both ranids and bufonids exhibit plasticity in a range of behavioural, morphological and physiological traits in response to predation risk (Chivers et al. 1999; Babbitt 2001; Marquis, Saglio & Neveu 2004; Relyea 2007), depending on factors such as hydroperiod (Laurila & Kujasalo 1999; Altwegg 2002) and competition (Nicieza 2000; Barnett & Richardson 2002). Conspecific densities also can affect the growth rate of tadpoles and metamorphic cane toads (Cohen & Alford 1993; Alford 1994, 1999; Lampo & De Leo 1998), as well as size at metamorphosis (Hearnden 1991). Tadpole density thus may have had an influence on our results.

In summary, our data add to an extensive literature that demonstrates widespread phenotypic flexibility in larval anurans, and to a smaller but expanding literature that documents carry-over effects from larval to post-metamorphic life (Relyea 2001a,b; Van Buskirk & Saxer 2001; Benard & Fordyce 2003). The combination of responses to alarm chemical cues in cane toads differs from that documented in previously studied species, but each of the responses that we have found is likely to occur in other taxa also. Understanding the ways in which such developmental plasticity impinges on fitness under field conditions remains a major challenge for anuran biologists.


We thank the Australian Research Council, Queensland State Government and Invasive Animals Cooperative Research Centre for funding. Procedures involving live animals were approved by the University of Sydney Animal Care and Ethics Committee.