The evolution of targeted cannibalism and cannibal-induced defenses in invasive populations of cane toads

Edited by Alan Hastings, University of California, Davis, CA, and approved June 14, 2021 (received for review January 15, 2021)
August 23, 2021
118 (35) e2100765118


Invasive species are known for their ability to achieve high densities within their introduced range. Hence, invaders often face strong competition from members of their own species. Mechanisms for reducing intraspecific competition may therefore be favored in invasive populations, such as cannibalism, in which individuals kill and eat intraspecific competitors. Here, we find that toad tadpoles from invasive Australian populations have evolved both a strong behavioral attraction to the vulnerable hatchling stage and an increased propensity to cannibalize these younger conspecifics. In response, these toads have also evolved multiple strategies for reducing the duration of the vulnerable period, indicating an evolutionary arms race between the cannibalistic tadpole stage and the vulnerable egg and hatchling stages in invaded habitats.


Biotic conflict can create evolutionary arms races, in which innovation in one group increases selective pressure on another, such that organisms must constantly adapt to maintain the same level of fitness. In some cases, this process is driven by conflict among members of the same species. Intraspecific conflict can be an especially important selective force in high-density invasive populations, which may favor the evolution of strategies for outcompeting or eliminating conspecifics. Cannibalism is one such strategy; by killing and consuming their intraspecific competitors, cannibals enhance their own performance. Cannibalistic behaviors may therefore be favored in invasive populations. Here, we show that cane toad tadpoles (Rhinella marina) from invasive Australian populations have evolved an increased propensity to cannibalize younger conspecifics as well as a unique adaptation to cannibalism—a strong attraction to vulnerable hatchlings—that is absent in the native range. In response, vulnerable conspecifics from invasive populations have evolved both stronger constitutive defenses and greater cannibal-induced plastic responses than their native range counterparts (i.e., rapid prefeeding development and inducible developmental acceleration). These inducible defenses are costly, incurring performance reductions during the subsequent life stage, explaining why plasticity is limited in native populations where hatchlings are not targeted by cannibalistic tadpoles. These results demonstrate the importance of intraspecific conflict in driving rapid evolution, highlight how plasticity can facilitate adaptation following shifts in selective pressure, and show that evolutionary processes can produce mechanisms that regulate invasive populations.
In our changing world, adaptations of invasive species to their introduced habitats provide key examples of how species can rapidly evolve in response to changes in both abiotic conditions and biotic interactions (1). For example, competitive abilities can be favored in introduced populations that have been freed from their natural predators and parasites and are therefore under relatively stronger selective pressure from competition (2, 3). The potential advantages of evolving increased competitive abilities are generally considered in the context of interspecific competition with native species, where such abilities could facilitate establishment and spread. However, a key attribute of successful invaders is that they become hyperabundant, at which point intraspecific competition can have stronger effects on fitness. This shift in selective pressure can instead favor mechanisms that reduce intraspecific conflict or enhance intraspecific competitive abilities; these mechanisms can differ from those favored by interspecific competition (4, 5). Whether and how strategies for alleviating competition with conspecifics evolve in invasive species remains a key question and could provide insights into the factors that eventually reduce invasiveness [as for invasive plants (5)] and/or regulate the populations of these species postinvasion (6).
Cannibalism can reduce intraspecific competition, as cannibalized conspecific competitors are both eliminated and consumed. This widespread and ecologically important phenomenon not only affects wildlife populations via effects on recruitment (7), population stabilization (8), and community structure (9) but also promotes dispersal (10) and migration (11) and can facilitate invasion (12, 13). For predatory species, cannibalism may represent natural diet extension. However, cannibalistic behaviors in herbivorous or detritivorous organisms that lack predatory adaptations can instead be indicative of resource limitation (14), as cannibalism both provides food resources and reduces intraspecific competition. This can be an adaptive strategy in certain contexts [such as resource-limited or temporary environments (15, 16)]. For example, within temporary waterbodies, pond drying is a substantial mortality risk for larval amphibians, especially if competition reduces developmental rates, delaying metamorphosis. In these environments, some amphibians display remarkable adaptations that facilitate cannibalism and accelerate development [e.g., inducible carnivorous/cannibalistic morphs (17, 18)]. Across species, cannibalistic tendencies not only vary among individuals (19) but can evolve under selective pressures such as artificial selection (20) or resource limitation (14). However, although the evolution of cannibalistic behaviors has been documented in laboratory populations, understanding of the significance of this evolutionary process is limited by a scarcity of evidence from natural populations (14, 15).
In Australia, cane toads are an abundant invasive species, achieving densities ∼10 times greater than those in their native South American range (21). This invasion has been facilitated by the toads’ high reproductive output and novel toxic defenses: as Australia lacks native toads, Australian predators and parasites are poorly adapted to toad toxins, and their ingestion is often lethal (22). As in their native range, Australian cane toads often breed in resource-limited, temporary waterbodies where their tadpoles graze on algae and detritus. Intraspecific competition reduces performance in these habitats (23, 24), but tadpoles lack adaptations for killing the conspecific tadpoles with which they compete (24). However, tadpoles can consume conspecific eggs that are laid in their pond, and the prefeeding hatchling stage, when the relatively immobile hatchlings have emerged from the protective egg capsule, is exceptionally vulnerable to cannibalism (25). In contrast, these hatchlings are relatively well defended against Australian predators by maternally invested toxins (e.g., bufadienolides) that provide protection throughout the aquatic stages. In Australia, cannibalism is the principal source of mortality in ponds where conspecific tadpoles are present, and cannibals often reduce the survival of newly laid clutches by >99% (25). This behavior improves cannibal performance by reducing competition and providing trophic resources (24) and may therefore be especially favored in these high-density, invasive populations (26). However, whether a high propensity to cannibalize conspecific hatchlings is characteristic of this species or has evolved within the invasive range is unknown. To determine whether cannibalism rates differ between native and invasive populations, we used 43 tadpole clutches and 22 hatchling clutches to conduct 514 cannibalism trials in which 10 hatchlings were exposed to a single tadpole over a 24-h period. Although tadpoles from both the native and invasive range cannibalized conspecific hatchlings, cannibalism rates were higher in invasive populations, such that the odds a hatchling would be cannibalized when exposed to an Australian tadpole were 2.55 times those in the native range (SE = 2.15 to 3.02, degrees of freedom [df] = 41, t = 5.50, P < 0.0001, Fig. 1B and SI Appendix, Table S1).
Fig. 1.
Cane toad tadpoles from invasive Australian populations cannibalized conspecifics at a higher rate than did native range tadpoles and, unlike native range tadpoles, exhibited a strong attraction to conspecifics during the vulnerable hatchling stage. Adult toads were collected from across French Guiana and the extent of their current Australian distribution (A); dots indicate collection sites, and arrows indicate historic exportation and introduction sites. When their offspring were offered 10 conspecific hatchlings, Australian tadpoles consumed more hatchlings than did tadpoles from native range populations (B) (proportion cannibalized shown for a 100-mg tadpole, n = 43 clutches, P < 0.0001). In attraction trials, Australian tadpoles were also more strongly attracted to hatchlings than native range tadpoles (n = 31 clutches, P < 0.0001). In these trials, tadpoles from the native range did not differentiate between an empty control trap and a trap containing hatchlings (C), such that 46% of native range tadpoles selected the hatchlings trap (SE: 39 to 53%). However, tadpoles from invasive Australian populations were strongly attracted to conspecifics during this vulnerable period (D), with 88% of tadpoles from invasive populations selecting the trap that contained hatchlings (SE: 87 to 90%). Means ± SE.
For species that lack adaptations for subduing and killing conspecifics, an ability to detect and locate vulnerable life stages can facilitate cannibalism (14). Because cane toad hatchlings are only vulnerable to cannibalism during the early, prefeeding stages (25), the ability to target conspecifics during this relatively brief period would enhance the ability of tadpoles to eliminate newly laid clutches. We determined whether tadpoles from native or invasive populations were attracted to the vulnerable hatchling stage using attraction trials, in which 2 traps (one control, one baited with 300 hatchlings) were placed in 90 L pools containing 50 tadpoles. We conducted 69 trials using 31 tadpole and 14 hatchling clutches. Within the native range, tadpoles were equally likely to enter the control trap as the trap containing hatchlings (odds ratio: 0.82, SE = 0.56 to 1.20, df = 12, t = −0.525, P = 0.61, Fig. 1C). However, Australian tadpoles were strongly attracted to conspecific hatchlings, such that a tadpole was 29.5 times as likely to enter a trap containing hatchlings as the paired control trap (SE = 24.7 to 35.1, df = 55, t = 19.26, P < 0.0001, Fig. 1D). Cannibalism therefore shifts from an opportunistic behavior in the native range to a targeted response in Australia, whereby tadpoles cease their normal foraging activities upon detecting hatchling cues in order to locate and consume conspecifics (SI Appendix, Fig. S1 and Table S2). In Australia, this behavior, which is not known for any other amphibians, is mediated by the detection of maternally invested toxins that are present in newly hatched conspecifics [e.g., bufadienolides; native amphibians are not targeted (27)]. These cues both attract tadpoles and induce feeding behaviors (27, 28). Attraction is only induced by the detection of conspecifics during this vulnerable prefeeding stage; Australian tadpoles are not attracted to conspecifics during the invulnerable tadpole stage (SI Appendix, Fig. S2 and Table S3). Remarkably, this adaptation has also facilitated control measures by allowing the targeted trapping of tadpoles from invaded waterbodies using toxin baits (27). These invasive populations have therefore evolved a behavior that facilitates the control of their own populations, not only because targeted cannibalism limits conspecific recruitment (25) but also because it increases tadpole susceptibility to the removal efforts of resource managers.
Innovation in one group can increase selective pressure on another, creating an evolutionary arms race (29, 30) in which organisms must constantly adapt in order to maintain fitness. Although seldom considered in the context of cannibalism, conflict with conspecifics can drive this process [e.g., in cases of sexual or maternal–offspring conflict (31)]. Stronger defensive strategies during the vulnerable, prefeeding life stages may therefore emerge in response to increased cannibalistic behaviors in tadpoles from invasive populations. When faced with a novel challenge, phenotypic plasticity (i.e., the ability for a given genotype to produce alternative phenotypes based on environmental conditions) is a key mechanism through which individuals can rapidly respond (32, 33). Individuals that exhibit an adaptive plastic response when confronted with a novel threat may be more likely to survive, increasing the persistence of plastic individuals, and thereby favoring plastic genotypes [i.e., Baldwin effects (34)]. In contrast, maladaptive plastic responses should be rapidly eliminated by natural selection (35). This may result in the evolution of increased adaptive plasticity. Inducible developmental acceleration is a plastic response that has evolved in response to stage-specific threats in a wide variety of species; this defense reduces mortality risk by allowing an individual that perceives a threat to reduce the duration of the risky period (3639). This kind of inducible defense has been documented in response to cannibalism risk in Australian cane toad hatchlings (25), but whether it is present in native populations or has emerged in response to increased cannibalism risk in the invasive range is unknown. We used 23 newly laid clutches to determine whether hatchlings from invasive populations exhibit stronger cannibal-induced defenses than those from native populations. We divided each clutch between cannibal exposed and control treatment tanks (presence versus absence of two caged conspecific tadpoles) and measured the duration of the vulnerable prefeeding period for each individual (for five individuals/tank in 189 cannibal exposed and 109 control tanks). Developmental plasticity in response to conspecific cues was relatively rare and inconsistent in native range clutches, such that only 1 of 10 clutches exhibited significant developmental acceleration. In contrast, most Australian clutches accelerated development (Fig. 2A). The regulation of prefeeding development has therefore been modified in invasive populations; on average, Australian populations exhibit significant, adaptive, cannibal-induced plasticity in prefeeding developmental rates (−2.34 ± 0.30 h, df = 163.0, t = −7.93, P < 0.0001), whereas native populations do not (−1.23 ± 0.64 h, df = 108.4, t = −1.93, P = 0.056; SI Appendix, Table S4). Although increased selective pressure from cannibalism could favor an inducible defense, the relative rarity of a response in the native range raises a question: why is this inducible defense not more common in native populations where opportunistic cannibalism still poses a risk? One possibility is that, where cannibalism risk is low, the fitness costs associated with this inducible defense outweigh the benefits.
Fig. 2.
Relative to native range clutches, cane toad clutches from invasive populations developed more rapidly through the vulnerable prefeeding stages, exhibited stronger cannibal-induced plasticity in developmental rates, and performed more poorly following exposure to cannibal cues. Clutches from the invasive Australian range reached the invulnerable tadpole stage more quickly than did native range clutches in both control and cannibal-cue–exposed treatments (A) [n = 23 clutches, P = 0.0266 and 0.0209, respectively; clutch means, Australian data were adapted from DeVore et al. (25)]. Plasticity in developmental rates also differed between native and invasive populations (A); on average, Australian clutches accelerated prefeeding development in response to cannibal cues (n = 13 clutches, P < 0.0001), whereas clutches from the native range did not (n = 10 clutches, P = 0.0573; clutches that exhibited a significant plastic response [P < 0.05] are depicted with solid lines). Across both native and invasive populations, greater cannibal-induced plasticity in the rate of prefeeding development was associated with poorer performance during the tadpole stage (B) (clutch means, n = 22 clutches, P < 0.0001). Differences in the magnitude of the plastic response between native and invasive populations ultimately led to differences in the mean effect of cannibal exposure on tadpole viability (C); on average, prefeeding exposure to nonlethal tadpole cues did not affect subsequent performance in native populations (n = 9 clutches; development P = 0.131, growth P = 0.651), but in invasive Australian populations, exposure reduced subsequent development and growth (n = 13 clutches; development P < 0.0001, growth P < 0.0001). Means ± SE; photograph insets visually represent treatments (A) and variation in growth among 10-d-old tadpoles (C).
Theoretical models predict that the ability to exhibit an adaptive plastic response should be costly, explaining why organisms do not evolve the ability to produce the optimal phenotype in all environments (40). However, despite their theoretical importance, costs of plasticity are rarely detected (41). It has therefore been suggested that the costs associated with plasticity are quickly offset by evolutionary processes, making them difficult to detect in organisms expressing well-established inducible responses (40, 42). If this is the case, plasticity costs should be more easily detected for responses that are actively evolving and have not yet been fixed in the population (SI Appendix, Fig. S4). In organisms with complex life history strategies, any costs associated with inducible defenses may be evident during the subsequent life stage as carry-over effects (43). To determine whether the ability to express this plastic response is costly, we individually raised 1,190 tadpoles from 22 of the clutches that had been monitored as hatchlings during the developmental acceleration experiment. Ten days later, we assessed their rates of survival, development (i.e., Gosner stage), and growth (measures considered appropriate proxies for “fitness” in tadpoles; see Methods). We then related each clutch’s degree of plasticity (i.e., the cannibal-induced reduction in prefeeding development time) and phenotype (i.e., the duration of prefeeding development) to its mean performance during the tadpole stage (44). We found that, in cannibal-exposed treatments, stronger adaptive plastic responses were followed by poorer tadpole performance across both native and invasive populations (Fig. 2B; effect per hour of acceleration: development: −0.53 ± 0.08 stages, df = 14.8, t = −6.93, P < 0.0001; growth: −8.94 ± 1.57 mg, df = 15.4, t = −5.70, P < 0.0001). Tadpoles from more plastic clutches also developed more slowly in control conditions (−0.171 ± 0.078 stages, df = 14.47, t = −2.19, P = 0.045). The ability to express this inducible response is therefore associated with slight performance reductions even if the threat is absent. In contrast, phenotype (i.e., the duration of prefeeding development) was not associated with tadpole fitness metrics (SI Appendix, Table S6 and Fig. S3). Therefore, poor clutch performance is not associated with rapid prefeeding development per se but with the ability to accelerate development. Ultimately, the difference in plasticity between the native and invasive range led to pronounced differences in the mean effect of exposure to conspecific cues on tadpole viability (treatment × country, development: df = 255.8, t = −6.50, P < 0.0001; growth: df = 255.6, t = −9.13, P < 0.0001; SI Appendix, Table S5). In the native range, prefeeding exposure to cues from conspecific tadpoles did not affect subsequent performance (P > 0.13 for all performance measures). However, in Australia, exposure to nonlethal cues substantially reduced tadpole development and growth rates (development: −2.34 ± 0.20 stages, df = 163.3, t = −11.7, P < 0.0001; growth: −53.8 ± 3.6 mg, df = 163.4, t = −15.1, P < 0.0001). Indeed, in Australia, the effects of prefeeding exposure to conspecific tadpoles are often so severe that they were initially attributed to intraspecific allelopathy (45) and the possibility of using conspecific tadpole cues to “poison” hatchlings has been explored as a control measure (46). These substantial costs to tadpole viability limit the adaptive value of cannibal-induced plasticity in the native range where opportunistic cannibalism poses less of a threat than the targeted cannibalism in Australia (25). Our evidence of these plasticity costs, which are theoretically predicted but infrequently detected, also supports the hypothesis that, under continued selective pressure, evolutionary processes eventually offset these costs or favor canalized defenses over costly inducible responses (SI Appendix, Fig. S4). As a result, plasticity costs that are difficult to detect for well-established inducible responses may be more overt following a shift in selective pressure.
Both the regulation (i.e., plasticity) and the expression of traits may evolve in response to changes in selective pressure. In this case, targeted cannibalism of the prefeeding stages could also favor the evolution of rapid prefeeding development (47). We therefore also compared the duration of the vulnerable period in clutches from native and invasive populations. We found that Australian clutches reach the invulnerable tadpole stage more quickly than do native range clutches in both cannibal-exposed and cannibal-naïve treatments, such that the duration of prefeeding development is reduced by ∼20% in invasive populations (control: df = 5.88, t = −2.94, P = 0.027, exposed: df = 5.86, t = −3.13, P = 0.021; SI Appendix, Table S4). This difference in developmental rates is only evident during the vulnerable, prefeeding stages. In control conditions, rates of development and growth during the invulnerable tadpole stage did not differ between the native and Australian range (development: 0.747 ± 0.600 stages [SE], df = 20.01, t = 1.25, P = 0.23; growth: 9.97 ± 12.70 mg [SE], df = 20.31, t = 0.79, P = 0.44). By relating the phenotype of each clutch to its plasticity, we also found that the most extreme prefeeding developmental rates were produced by nonplastic development (Fig. 3 and SI Appendix, Table S7). However, nonplastic strategies produced remarkably different phenotypes in each country. In the native range, nonplastic clutches developed slowly, reaching the tadpole stage in ∼5 d. This fixed, slow development strategy was absent in the invasive Australian range where nonplastic clutches instead exhibited rapid development (∼3 d), an apparently derived strategy that was not observed in native range clutches. Plastic development produced intermediate phenotypes, which were present in both native and invasive populations. As a result, whereas plasticity was an effective strategy for reducing the duration of the vulnerable period within native range populations (when exposed to cannibals, the most plastic native range clutches had the shortest development times: −2.189 ± 0.862 h [SE], t = −2.54, P = 0.0347), plasticity was relatively ineffective within Australia (where the most plastic clutches had the longest development times: 2.796 ± 1.089 h [SE], t = 2.57, P = 0.0261; SI Appendix, Table S8). Interestingly, whereas the negative consequences of cannibal-induced developmental acceleration were immediately evident during the subsequent life stage, we did not detect any costs associated with canalized rapid development. However, we could not have detected any costs that do not manifest until later life stages (e.g., postmetamorphosis). In addition, maternal effects can influence offspring phenotypes, including plastic responses (48); future research should investigate the roles of genotype and maternal effects in driving phenotypic variation in this system.
Fig. 3.
In both cannibal exposed and control conditions, the most extreme phenotypes were found in nonplastic clutches. However, the nature of the phenotype produced differed between the native and invasive range. In the native range, nonplastic clutches developed slowly, whereas in the invasive range, nonplastic clutches exhibited rapid development. Intermediate phenotypes were produced by plastic clutches and were found in both native and invasive populations. Here, quadratic regressions of plasticity (i.e., the cannibal-induced reduction in the duration of prefeeding development for each clutch) are plotted against the total duration of the vulnerable, prefeeding period. The mean observed phenotypes of the 23 clutches in control (circle; quadratic term P = 0.0116) or cannibal exposed (plus; P = 0.0044) conditions used to plot these regressions are also shown; lines connect the phenotypes observed for a certain clutch within each environment (thus indicating that clutch’s plasticity). Phenotypes only observed in the native range are shown with a gray background, whereas those only observed in the invasive range are plotted over white. The transition zone includes phenotypes observed in both native and invasive populations. Viability during the subsequent tadpole stage was significantly associated with the magnitude of the plastic response a clutch displayed during the prefeeding period; tadpole performance was poorest in the most plastic clutches, especially following cannibal exposure.
Ultimately, selection for rapid development and canalization of the inducible defense may both contribute to the rapid prefeeding development of Australian clutches. The canalization of an inducible phenotype can occur where a canalized phenotype provides a fitness advantage over a plastic response [genetic assimilation (4951)]. In this case, the inducible defenses that are common in these invasive populations, although likely initially favored in Australia as the most effective strategy from ancestral, native-range populations [i.e., Baldwin effects (34)], are both more costly and less effective at reducing the window of vulnerability than the canalized rapid development that has emerged in the invasive range [see also DeVore et al. (25)]. This inducible defense may therefore be ephemeral within these invasive populations, as a shift to the canalized rapid development already exhibited by some clutches is expected under continued selective pressure. Adaptive refinement of inducible responses may occur where the inducing environment is frequently encountered [i.e., frequency-dependent adaptation (52)]; a shift to a less costly and/or more effective defensive strategy may therefore be especially favored in parts of the invasive range where cannibals are most often present [e.g., in range-core populations (53)]. Monitoring of the stability of this inducible defense could provide insights into whether costly plastic responses that are favored following a shift in selective pressure can be either maintained via cost offsets or canalized in the induced state.
Intraspecific competition can be an important source of selective pressure. In invasive populations where conspecific densities are high, this pressure may be intensified, favoring the evolution of strategies that reduce intraspecific conflict. This process may be especially important in invaders that are well defended against predators and parasites [i.e., rare enemy effects (54)]. Such adaptations can then favor further evolutionary change. Invasive cane toads display traits such as rapid prefeeding development, cannibal-induced developmental acceleration, and increased dispersal abilities during the terrestrial life stages that accelerate the colonization of new, cannibal-free habitats (10, 55). Our results reveal that the evolutionary emergence of targeted cannibalism in the invasive range may have favored these new evolutionary trajectories, demonstrating the importance of intraspecific conflict in driving adaptation in natural systems, as well as the potential for the evolutionary processes to produce mechanisms that stabilize invasive populations. These results also provide a clear example of the role of phenotypic plasticity in facilitating rapid adaptation to shifting selective pressures.


Study System: R. marina (Previously Bufo marinus; Cane Toad, Marine Toad).

Native to South America, the cane toad has been introduced to >100 islands and countries worldwide. French Guiana and Guyana were the original native source populations for the serial introductions through the Caribbean and Hawai’i that, in 1936, led to the importation of 101 toads to Queensland, Australia (56). Toads from French Guiana are therefore an appropriate standard for comparisons of native and invasive Australian populations. Since their introduction, toads have spread across much of tropical Australia (Fig. 1A, Outline). This invasion has been facilitated by their powerful defensive toxins [which are not produced by any Australian fauna (22)], their high reproductive output (frequently >10,000 eggs per clutch), and the evolution of increased dispersal abilities and behaviors along the invasion front (57). Toxic to a wide array of native Australian aquatic and terrestrial predators (22) and infrequently parasitized relative to native range populations (58, 59), cane toads generally occur at higher densities in Australia than the native range (21). In Australia, cane toads preferentially breed within open, temporary pools (60) where tadpole performance is often restricted by resource availability (23) and higher tadpole densities are associated with reduced performance (24). Such resource limitation increases the duration of the tadpole stage [increasing mortality risk from pond drying (61, 62)] and reduces growth rates [increasing susceptibility to predators and reducing size at metamorphosis—thereby reducing survival and growth in the terrestrial environment (43, 63)]. Note that rates of survival, development, and growth are therefore considered appropriate proxies for “fitness” in tadpoles. Cannibalism improves the performance of Australian tadpoles by reducing competition with younger cohorts [if younger cohorts are protected from cannibals, the resulting competition reduces rates of survival, development, and growth in the older cohort (24)]. Although reduced competition is the principle benefit of cannibalizing a younger cohort, they are also a food resource; tadpoles can reach metamorphosis on a diet of conspecifics (24). These tadpoles specifically target conspecifics; eggs and hatchlings of other amphibian species are rarely consumed (64). In Australia, cannibalism of newly laid egg strings and recently hatched prefeeding larvae (“hatchlings”) by conspecific tadpoles is the main source of mortality in ponds where cannibals are present, and cannibalism often reduces clutch survival by >99% (23, 25). Whereas aquatic predators generally consume both hatchlings and tadpoles (and therefore do not specifically target either stage), cane toad tadpoles are grazers that lack specialized adaptations for killing conspecific tadpoles [unlike some anurans (51)]. Cannibals can therefore only consume conspecifics during the prefeeding egg and hatchling stages; upon reaching the free-swimming tadpole stage [i.e., Gosner stage 25 (65)], younger cohorts can no longer be cannibalized (25). Selective pressure from cannibals is therefore specific to the prefeeding life stages. Because tadpoles only consume younger cohorts they cannot consume full siblings, so kin selection, which can limit the evolution of cannibalistic behaviors (66), may not reduce the adaptive value of cannibalism as strongly as if full siblings were susceptible. [Note that the ability to avoid cannibalizing kin can also minimize the constraining effects of kin selection on the evolution of cannibalism. However, although some amphibian species can recognize and avoid consuming related individuals (67), it is not known whether cane toad tadpoles have this ability. Future research could establish whether these cannibalistic behaviors are accompanied by strategies for identifying and rejecting kin.]

Collection and General Methodology.

Adult cane toads were hand-collected in French Guiana (2017) and Australia (2014 to 2019) from sites distributed across each country (Fig. 1). Following collection, adults were transported to breeding facilities. Breeding was induced using injections of the synthetic gonadotrophin leuprorelin acetate (Lucrin, Abbott Australasia; 0.25 mg⋅mL−1); 0.75 mL doses for females, 0.25 mL doses for males. Eggs and hatchlings required for experimental manipulations were drawn from the resulting clutches. Once hatchlings not utilized in these manipulations had reached the tadpole stage (i.e., Gosner stage 25; 3 to 6 d after laying) they were fed Hikari algae wafers (Kyorin, Japan) ad libitum. These tadpoles were used as cannibals in subsequent trials. Where multiple eggs, hatchlings, or tadpoles were needed for a given trial, they were all taken from the same clutch. All experiments were conducted in unchlorinated water at 27 °C. Note that a subset of our data on Australian attraction and developmental rates has also been presented in DeVore et al. (25). Here we include these data in larger data sets to make comparisons between the native and invasive range. All analyses were conducted in R (68). Although we include abbreviated results in the main text, full model outputs are also available in the SI Appendix.

Comparisons of Cannibalism Rates between the Native and Invasive Range.

To determine whether the propensity to cannibalize conspecific hatchlings differed between native and invasive populations, we conducted 514 predation trials in 750 mL containers stocked with 10 hatchlings each (Gosner stage 17.5 ± 1). Trials began when a single, preweighed tadpole was introduced to the container. These tadpoles ranged between 7 and 187 mg (mean mass = 80.6 mg; Gosner stage 27 to 40). After a 24-h exposure period, the number of cannibalized hatchlings was assessed. Hatchlings were considered cannibalized if they were dead following total or partial consumption by the tadpole; all hatchlings that were still alive at this time were considered to have survived. Trials included 141 native range trials in which 17 tadpole clutches and 6 hatchling clutches were paired in 41 combinations, and 373 invasive range trials of 26 tadpole clutches and 16 hatchling clutches in 73 combinations. An additional 100 control containers in which only hatchlings were present were included to account for potential variation in background mortality (mean= 4.5 tanks per clutch, range 2 to 6), which averaged 4% and did not differ between the native and invasive range (df = 20, t = −0.0023, P = 0.998). To determine whether cannibalism rates differ between native versus invasive populations, the proportion of hatchlings cannibalized was analyzed as a function of source population (native versus invasive) using a quasibinomial model [to account for data over-dispersion; package MASS: glmmPQL (69)]. Tadpole mass was also included as a fixed effect, whereas tadpole clutch and hatchling clutch were included as nested random effects.
We focused these trials on the hatchling stage because clutches are extremely vulnerable to cannibalism at this stage (25) and because, in 49 trials in which tadpoles from 12 clutches were offered strands of 10 eggs, individual tadpoles from French Guiana never ruptured the egg strand [although we have observed aggregations of tadpoles successfully consuming conspecific eggs in the field in both French Guiana and Australia (25)]. In contrast, individual tadpoles from Australia did rupture the egg strand and consume unhatched eggs in 10 of 25 trials. However, eggs were consumed more slowly than hatchlings (in trials where eggs were cannibalized, 1 to 4 [mean = 1.9] were consumed over the ∼24-h period encompassing egg stages 9 through 16). Our comparison of hatchling cannibalism rates is therefore a conservative estimate of the overall difference in cannibalistic behaviors in tadpoles from Australia versus the native range, as Australian tadpoles also cannibalize unhatched eggs more readily than do native range tadpoles.

Comparison of Tadpole Attraction to Vulnerable Hatchlings between the Native and Invasive Range.

In Australia, tadpoles are actively attracted to chemical cues from conspecific hatchlings (25, 28). To determine whether the ability to detect and target the vulnerable hatchling stage differs between the native and invasive range, we conducted 69 trapping trials in 1 m diameter pools containing 90 L of water and 50 tadpoles. Ground Hikari algae wafers (∼1 g) were spread across each pool as a food resource (traps did not contain food). Two 1-L traps were opened in each pool ∼2 h after tadpole introduction; one of these traps contained 300 hatchlings (stage 18 ± 1) in a 75 mL cup covered with 1 mm2 mesh (to prevent tadpoles from cannibalizing the hatchlings during the trial while allowing cue dispersal), whereas the other contained an empty, mesh-sealed cup (SI Appendix, Fig. S1). These traps had a single funnel (D = 1 cm) on one side and were centrally placed 50 cm apart within each pool with the funnels pointing outward. After the traps were opened, the number of tadpoles in each trap was counted hourly for 6 h. Invasive range trials paired 24 tadpole and 10 hatchling clutches in 56 combinations, whereas native range trials paired 7 tadpole and 4 hatchling clutches in 13 combinations. We separately modeled trapping rate (trapped in either trap versus untrapped) and trap preference (the number of tadpoles in the hatchling trap versus control trap) as binomial responses to source population (native versus invasive), time, and their interaction. Tadpole clutch and trial ID were included in both models as nested random effects. In both cases we used quasi-binomial models to account for data overdispersion [package MASS: glmmPQL (69)].

Comparison of Cannibal-Induced Developmental Acceleration and Constitutive Variation in Developmental Rates between the Native and Invasive Range.

In Australia, exposure to cues from conspecific tadpoles during the prefeeding egg and/or hatchling stages induces developmental acceleration that reduces the duration of the vulnerable prefeeding period (25). This acceleration occurs during the hatchling stage, when Australian clutches actively attract cannibalistic tadpoles (25). Vulnerability to cannibals is limited to the prefeeding stages; once hatchlings reach Gosner stage 25 (i.e., the beginning of the tadpole stage), they can no longer be consumed. We used 23 egg clutches to determine whether the duration of the vulnerable prefeeding period or the plastic responses induced by exposure to conspecific cues differ between the native and invasive range: 10 from French Guiana and 13 from Australia. For each trial, strands of five eggs were randomly assigned to cannibal-naïve (i.e., control) or cannibal-exposed treatments, and we determined the duration of the prefeeding stages for full siblings in each treatment by continually scanning each tank, confirming the developmental stage of individuals approaching the end of the hatchling stage under a dissecting microscope and noting the time at which each individual reached stage 25 (i.e., the conclusion of the vulnerable, prefeeding developmental stages). Two “cannibal” tadpoles were used per 750 mL cue exposure tank. These cannibals were caged during exposure to prevent them from consuming prefeeding conspecifics. Control tanks contained empty cages [see DeVore et al. (25) for additional methodological detail]. Twenty-four tadpole clutches were used to create cannibal cues; 1 to 11 of these clutches were paired with each hatchling clutch (mean = 3.3), creating 77 unique pairings. Three to twenty-one exposure tanks (mean= 8.2) and two to seven control tanks (mean = 4.7) were included per hatching clutch, for a total of 189 exposure and 109 control tanks. Trials were split across eight experimental blocks. Mean development time was calculated for each tank as the average of the times at which each individual hatchling reached the tadpole stage. The total duration of the vulnerable prefeeding period (i.e., the mean number of hours between egg-laying and the beginning of the free-swimming tadpole stage) was then analyzed as a response to source population (native/invasive), treatment (control/exposed), and their interaction. Experimental block and clutch ID were included as nested random effects [package lme4:lmer (70)].
In some cases, although an inducible defense may provide some protection for individuals that would otherwise be poorly defended, nonplastic constitutive defenses provide stronger protection (40). This could occur if, for example, the reduction in the duration of the vulnerable period achievable via developmental acceleration is limited relative to the reduction that can be achieved via inherent variation in developmental rates. In other cases, plasticity is highly effective, and the most plastic genotypes produce the most well-defended phenotypes when threatened. To determine whether plasticity is an effective strategy for reducing the duration of the vulnerable prefeeding period (relative to nonplastic development), as well as whether the relative effectiveness of plasticity varied between the native and invasive range, we also related the plasticity of each clutch to the phenotype it produced in the presence of cannibals. We therefore analyzed clutch phenotype (i.e., the total duration of the vulnerable, prefeeding period) as a response to clutch plasticity (i.e., the reduction in development time exhibited in response to cannibal cues), source population (native/invasive), and their interaction. Because we observed substantial variation in both developmental rates and plasticity among clutches in both control and cannibal-exposed treatments, we also sought to determine whether certain phenotypes (e.g., rapid or slow development) were associated with particular developmental strategies (e.g., plastic or nonplastic). We therefore also related clutch plasticity to clutch phenotype within each environment. Here, source population and the interaction between population and phenotype were also included as predictors of clutch plasticity. We also tested for quadratic relationships between phenotype and plasticity. In all cases, these analyses were performed using mean clutch values from each treatment, which we calculated for each of the 23 clutches by averaging the tank means.

Comparing the Carryover Effects of Cannibal Exposure between the Native and Invasive Range.

Inducible defenses that reduce mortality risk can incur costs, which lower the fitness of the induced phenotype in the absence of a threat (71). These phenotypic costs promote the evolution of plasticity; in variable environments, selection favors the ability to select between costly, defended phenotypes and less costly, undefended phenotypes depending on whether or not the threat is present (71). Although the costs of producing an adaptive inducible defense should not exceed the benefits, even responses that incur high costs can be adaptive if the risk is severe. In organisms with complex life history strategies, these costs often occur during the subsequent life stage as carry-over effects; for example, accelerating development to reduce mortality risk during a certain life stage (e.g., in response to pond drying or seasonality) can reduce subsequent performance due to growth/development trade-offs or increases in oxidative stress (72, 73). In Australia, where hatchling mortality is rapid and extreme when cannibals are present, the costs of the cannibal-induced defense are often severe (25, 45). To determine whether prefeeding exposure to nonlethal cannibal cues has similar effects on subsequent tadpole performance in the native range, we individually raised 1,190 tadpoles sourced from 278 of the hatchling tanks included in the developmental acceleration experiment (one clutch was not subsequently raised, leaving tadpoles from 22 clutches). Each tadpole was raised in a 200 mL container and fed ad libitum daily. Ten days later we assessed tadpole survival, development (i.e., Gosner stage), and growth [measures considered as appropriate proxies for “fitness” in tadpoles; see DeVore et al. (59) for additional methodological detail]. For each hatchling tank, we then quantified mean tadpole performance for each fitness measure. Note that, although hatchling tanks were initially stocked with five eggs, individuals that did not successfully reach stage 25 were excluded. Each tank was therefore ultimately represented by one to five tadpoles (mean = 4.2). Fitness measures were modeled as responses to treatment (control/exposed), source population (native/invasive), and their interaction. Linear models were used to analyze development and growth [package lme4:lmer (70)], whereas survival was analyzed with a quasi-binomial model [package MASS: glmmPQL (69)]. Experimental block and clutch ID were included in all models as nested random effects.

Separating the Costs of Plasticity from the Costs of Phenotype.

To understand the effect that costs associated with an inducible defense will have on the evolution of plasticity, it is important to differentiate between phenotypic and plasticity costs. Whereas costs of phenotype promote the evolution of plasticity (see previous section), costs of plasticity should constrain its evolution and may instead favor nonplastic, canalized development (49, 50, 74, 75). A cost of plasticity is evident if the fitness of a plastic genotype is lower than that of a nonplastic genotype when both organisms produce the same phenotype (40, 44). In contrast, a cost of phenotype is evident when, for example, the production of the defended phenotype is associated with fitness costs regardless of whether it is produced via plastic or nonplastic development. We used a partial linear regression technique proposed by Van Tienderen to separate plasticity and phenotypic costs (44). This technique relates variation in fitness between different genotypes within a given environment to both their expressed phenotype and their plasticity using the model Wk = constant + Xk + plXk. Here, Wk is a fitness component in environment k, Xk is the mean phenotype in that environment, and plXk is the plasticity (i.e., the difference in phenotype between environments). If plasticity is costly, the plasticity coefficient will have a negative value. In contrast, if the phenotype is costly to produce regardless of whether it was produced via a plastic response or canalized development, the phenotype coefficient will have a negative value. In some cases, the ability to be plastic may be associated with costs even in environments where the inducible defense is not expressed [for example, due to the costs of maintaining cue-perceiving structures (40)]. In this case, plastic genotypes may also exhibit reduced performance in control conditions.
For each of the 22 tadpole clutches, we therefore calculated mean fitness metrics within each treatment (i.e., developmental, growth, and survival rates). There was substantial variation in rates of tadpole development and growth among tadpole clutches, especially in exposed treatments (where mean tadpole mass varied from 29 to 151 mg) but also in control conditions (where mean mass ranged from 68 to 176 mg). To determine whether this variation in performance during the tadpole stage was related to either plasticity or phenotype during prefeeding development we used the partial linear regression technique described in the previous paragraph, including separate models for each fitness measure. Here, phenotype was taken as the mean duration of prefeeding development within the focal environment (h), whereas plasticity was calculated for each clutch as the difference in mean development time between control and exposed treatments (−h, such that inducible reductions in development time were given a positive plasticity value). In all cases, clutch values were obtained by averaging the tank means. Experimental block was included in all models as a random effect [package lme4:lmer (70)].

Data Availability

All data are available in the Supplementary Materials. Note that a subset of the invasive range data used here were also used in DeVore et al. 2021; here we incorporate these results into larger datasets to make comparisons between the native and invasive range.


All procedures were approved by the University of Sydney Animal Care and Ethics Committee. We are grateful for the permits provided by the French Ministère de la Transition Ecologique et Solidaire (permit TREL1734890A/1, 19 December 2017), the Préfet de la Région Guyane (arrêté APmodif-R03-2017-07-18-006), and the Departamento de Recursos Naturales y Ambientales de Puerto Rico (DRNA: 2017-IC017, O-VS-PVS15-SJ-00894-15122016). We also thank Philippe Gaucher, Antoine Fouquet, and Alberto Puente-Rolón for assistance with permitting and supplies.

Supporting Information

Appendix (PDF)
Dataset_S01 (XLSX)


E. V. Moran, J. M. Alexander, Evolutionary responses to global change: Lessons from invasive species. Ecol. Lett. 17, 637–649 (2014).
B. Blossey, R. Notzold, Evolution of increased competitive ability in invasive nonindigneous plants—A hypothesis. J. Ecol. 83, 887–889 (1995).
R. M. Callaway, W. M. Ridenour, Novel weapons: Invasive success and the evolution of increased competitive ability. Front. Ecol. Environ. 2, 436–443 (2004).
R. A. Lankau, S. Y. Strauss, Mutual feedbacks maintain both genetic and species diversity in a plant community. Science 317, 1561–1563 (2007).
R. A. Lankau, V. Nuzzo, G. Spyreas, A. S. Davis, Evolutionary limits ameliorate the negative impact of an invasive plant. Proc. Natl. Acad. Sci. U.S.A. 106, 15362–15367 (2009).
D. Simberloff, L. Gibbons, Now you see them, now you don’t!—Population crashes of established introduced species. Biol. Invasions 6, 161–172 (2004).
S. A. Wissinger, H. H. Whiteman, M. Denoël, M. L. Mumford, C. B. Aubee, Consumptive and nonconsumptive effects of cannibalism in fluctuating age-structured populations. Ecology 91, 549–559 (2010).
H. P. Benoît, E. McCauley, J. R. Post, Testing the demographic consequences of cannibalism in Tribolium confusum. Ecology 79, 2839–2851 (1998).
L. Persson et al., Gigantic cannibals driving a whole-lake trophic cascade. Proc. Natl. Acad. Sci. U.S.A. 100, 4035–4039 (2003).
V. H. W. Rudolf, M. Kamo, M. Boots, Cannibals in space: The coevolution of cannibalism and dispersal in spatially structured populations. Am. Nat. 175, 513–524 (2010).
S. J. Simpson, G. A. Sword, P. D. Lorch, I. D. Couzin, Cannibal crickets on a forced march for protein and salt. Proc. Natl. Acad. Sci. U.S.A. 103, 4152–4156 (2006).
S. Via, Cannibalism facilitates the use of a novel environment in the flour beetle, Tribolium castaneum. Heredity 82, 267–275 (1999).
J. Javidpour, J. C. Molinero, E. Ramírez-Romero, P. Roberts, T. Larsen, Cannibalism makes invasive comb jelly, Mnemiopsis leidyi, resilient to unfavourable conditions. Commun. Biol. 3, 212 (2020).
R. K. Vijendravarma, S. Narasimha, T. J. Kawecki, Predatory cannibalism in Drosophila melanogaster larvae. Nat. Commun. 4, 1789 (2013).
G. A. Polis, The evolution and dynamics of intraspecific predation. Annu. Rev. Ecol. Syst. 12, 225–251 (1981).
X. Martini, P. Haccou, I. Olivieri, J.-L. Hemptinne, Evolution of cannibalism and female’s response to oviposition-deterring pheromone in aphidophagous predators. J. Anim. Ecol. 78, 964–972 (2009).
D. Pfennig, The adaptive significance of an environmentally-cued developmental switch in an anuran tadpole. Oecologia 85, 101–107 (1990).
H. Michimae, T. Emura, Correlated evolution of phenotypic plasticity in metamorphic timing. J. Evol. Biol. 25, 1331–1339 (2012).
K. A. Nilsson, S. Lundbäck, A. Postavnicheva-Harri, L. Persson, Guppy populations differ in cannibalistic degree and adaptation to structural environments. Oecologia 167, 391–400 (2011).
L. Stevens, The genetics and evolution of cannibalism in flour beetles (Genus Tribolium). Evolution 43, 169–179 (1989).
M. Lampo, P. Bayliss, Density estimates of cane toads from native populations based on mark-recapture data. Wildl. Res. 23, 305–315 (1996).
R. Shine, The ecological impact of invasive cane toads (Bufo marinus) in Australia. Q. Rev. Biol. 85, 253–291 (2010).
R. A. Alford, M. P. Cohen, M. R. Crossland, M. N. Hearnden, L. Schwarzkopf, “Population biology of Bufo marinus,” in Wetland Research in the Wet-Dry Tropics of Australia, Supervising Scientist Report 101, C. M. Finlayson, Ed. (Office of the Supervising Scientist, Jabiru, Northern Territory, Australia, 1995), pp. 173–181.
M. R. Crossland, M. N. Hearnden, L. Pizzatto, R. A. Alford, R. Shine, Why be a cannibal? The benefits to cane toad, Rhinella marina (Bufo marinus) tadpoles of consuming conspecific eggs. Anim. Behav. 82, 775–782 (2011).
J. L. DeVore, M. R. Crossland, R. Shine, Tradeoffs affect the adaptive value of plasticity: Stronger cannibal-induced defenses incur greater costs in toad larvae. Ecol. Monogr. 91, e01426 (2021).
A. Tayeh et al., Cannibalism in invasive, native and biocontrol populations of the harlequin ladybird. BMC Evol. Biol. 14, 15 (2014).
M. R. Crossland, T. Haramura, A. A. Salim, R. J. Capon, R. Shine, Exploiting intraspecific competitive mechanisms to control invasive cane toads (Rhinella marina). Proc. Biol. Sci. 279, 3436–3442 (2012).
M. R. Crossland, R. Shine, Cues for cannibalism: Cane toad tadpoles use chemical signals to locate and consume conspecific eggs. Oikos 120, 327–332 (2011).
L. Van Valen, A new evolutionary law. Evol. Theory 1, 1–30 (1973).
D. C. Queller, J. E. Strassmann, Evolutionary conflict. Annu. Rev. Ecol. Evol. Syst. 49, 73–93 (2018).
M. A. Brockhurst et al., Running with the Red Queen: The role of biotic conflicts in evolution. Proc. Biol. Sci. 281, 20141382 (2014).
M. J. West-Eberhard, Developmental Plasticity and Evolution (Oxford University Press, 2003).
A. A. Agrawal, Phenotypic plasticity in the interactions and evolution of species. Science 294, 321–326 (2001).
J. M. Baldwin, A new factor in evolution. Am. Nat. 30, 441–451 (1896).
C. K. Ghalambor et al., Non-adaptive plasticity potentiates rapid adaptive evolution of gene expression in nature. Nature 525, 372–375 (2015).
M. De Block, R. Stoks, Fitness effects from egg to reproduction: Bridging the life history transition. Ecology 86, 185–197 (2005).
K. M. Warkentin, Environmentally cued hatching across taxa: Embryos respond to risk and opportunity. Integr. Comp. Biol. 51, 14–25 (2011).
M. C. O. Ferrari, B. D. Wisenden, D. P. Chivers, Chemical ecology of predator–prey interactions in aquatic ecosystems: A review and prospectus. Can. J. Zool. 88, 698–724 (2010).
O. Kishida, A. Tezuka, A. Ikeda, K. Takatsu, H. Michimae, Adaptive acceleration in growth and development of salamander hatchlings in cannibalistic situations. Funct. Ecol. 29, 469–478 (2015).
T. J. Dewitt, A. Sih, D. S. Wilson, Costs and limits of phenotypic plasticity. Trends Ecol. Evol. 13, 77–81 (1998).
J. Van Buskirk, U. K. Steiner, The fitness costs of developmental canalization and plasticity. J. Evol. Biol. 22, 852–860 (2009).
M. Pigliucci, C. J. Murren, Perspective: Genetic assimilation and a possible evolutionary paradox: Can macroevolution sometimes be so fast as to pass us by? Evolution 57, 1455–1464 (2003).
J. A. Pechenik, Larval experience and latent effects—Metamorphosis is not a new beginning. Integr. Comp. Biol. 46, 323–333 (2006).
P. H. Van Tienderen, Evolution of generalists and specialists in spatially heterogeneous environments. Evolution 45, 1317–1331 (1991).
M. R. Crossland, R. Shine, Embryonic exposure to conspecific chemicals suppresses cane toad growth and survival. Biol. Lett. 8, 226–229 (2012).
G. S. Clarke, M. R. Crossland, R. Shine, Can we control the invasive cane toad using chemicals that have evolved under intraspecific competition? Ecol. Appl. 26, 463–474 (2016).
A. Melotto, R. Manenti, G. F. Ficetola, Rapid adaptation to invasive predators overwhelms natural gradients of intraspecific variation. Nat. Commun. 11, 3608 (2020).
D. W. Pfennig, R. A. Martin, A maternal effect mediates rapid population divergence and character displacement in spadefoot toads. Evolution 63, 898–909 (2009).
C. H. Waddington, Canalization of development and the inheritance of acquired characters. Nature 150, 563–565 (1942).
N. A. Levis, D. W. Pfennig, Evaluating ‘plasticity-first’ evolution in nature: Key criteria and empirical approaches. Trends Ecol. Evol. 31, 563–574 (2016).
N. A. Levis, A. J. Isdaner, D. W. Pfennig, Morphological novelty emerges from pre-existing phenotypic plasticity. Nat. Ecol. Evol. 2, 1289–1297 (2018).
N. A. Levis, D. W. Pfennig, Plasticity-led evolution: Evaluating the key prediction of frequency-dependent adaptation. Proc. Biol. Sci. 286, 20182754 (2019).
E. R. Westra et al., Parasite exposure drives selective evolution of constitutive versus inducible defense. Curr. Biol. 25, 1043–1049 (2015).
R. Dawkins, The Extended Phenotype (Oxford University Press, Oxford, 1982), vol. 8.
M. C. Urban, B. L. Phillips, D. K. Skelly, R. Shine, A toad more traveled: The heterogeneous invasion dynamics of cane toads in Australia. Am. Nat. 171, E134–E148 (2008).
N. D. Turvey, Cane Toads: A Tale of Sugar, Politics and Flawed Science (Sydney University Press, Sydney, 2013).
B. L. Phillips, G. P. Brown, J. K. Webb, R. Shine, Invasion and the evolution of speed in toads. Nature 439, 803 (2006).
D. Selechnik, L. A. Rollins, G. P. Brown, C. Kelehear, R. Shine, The things they carried: The pathogenic effects of old and new parasites following the intercontinental invasion of the Australian cane toad (Rhinella marina). Int. J. Parasitol. Parasites Wildl. 6, 375–385 (2016).
J. L. DeVore, R. Shine, S. Ducatez, Urbanization and translocation disrupt the relationship between host density and parasite abundance. J. Anim. Ecol. 89, 1122–1133 (2020).
M. Hagman, R. Shine, Spawning site selection by feral cane toads (Bufo marinus) at an invasion front in tropical Australia. Austral Ecol. 31, 551–558 (2006).
R. A. Relyea, Costs of phenotypic plasticity. Am. Nat. 159, 272–282 (2002).
R. A. Newman, Developmental plasticity of Scaphiopus couchii tadpoles in an unpredictable environment. Ecology 70, 1775–1787 (1989).
E. Cabrera-Guzmán, M. R. Crossland, G. P. Brown, R. Shine, Larger body size at metamorphosis enhances survival, growth and performance of young cane toads (Rhinella marina). PLoS One 8, e70121 (2013).
M. R. Crossland, A comparison of cane toad and native tadpoles as predators of native anuran eggs, hatchlings and larvae. Wildl. Res. 25, 373–381 (1998).
K. L. Gosner, A simplified table for staging anuran embryos and larvae with notes on identification. Herpetologica 16, 183–190 (1960).
M. J. Wade, An experimental study of kin selection. Evolution 34, 844–855 (1980).
D. W. Pfennig, Kinship and cannibalism. Bioscience 47, 667–675 (1997).
R Core Team, R: A Language and Environment for Statistical Computing (R Foundation for Statistical Computing, 2020).
W. N. Venables, B. D. Ripley, Modern Applied Statistics with S (Springer, New York, 2002), ed. 4.
D. Bates, M. Mächler, B. Bolker, S. Walker, Fitting linear mixed-effects models using lme4. J. Stat. Softw. 67, 1–48 (2015).
C. D. Harvell, The ecology and evolution of inducible defenses. Q. Rev. Biol. 65, 323–340 (1990).
L. Janssens, R. Stoks, Rapid larval development under time stress reduces adult life span through increasing oxidative damage. Funct. Ecol. 32, 1036–1045 (2018).
J. D. Arendt, Adaptive intrinsic growth rates: An integration across taxa. Q. Rev. Biol. 72, 149–177 (1997).
F. Aubret, R. Shine, Fitness costs may explain the post-colonisation erosion of phenotypic plasticity. J. Exp. Biol. 213, 735–739 (2010).
M. Heil et al., Evolutionary change from induced to constitutive expression of an indirect plant resistance. Nature 430, 205–208 (2004).

Information & Authors


Published in

Go to Proceedings of the National Academy of Sciences
Proceedings of the National Academy of Sciences
Vol. 118 | No. 35
August 31, 2021
PubMed: 34426494


Data Availability

All data are available in the Supplementary Materials. Note that a subset of the invasive range data used here were also used in DeVore et al. 2021; here we incorporate these results into larger datasets to make comparisons between the native and invasive range.

Submission history

Published online: August 23, 2021
Published in issue: August 31, 2021


  1. Baldwin effects
  2. canalization
  3. co-adaptation
  4. phenotypic plasticity
  5. plasticity costs


All procedures were approved by the University of Sydney Animal Care and Ethics Committee. We are grateful for the permits provided by the French Ministère de la Transition Ecologique et Solidaire (permit TREL1734890A/1, 19 December 2017), the Préfet de la Région Guyane (arrêté APmodif-R03-2017-07-18-006), and the Departamento de Recursos Naturales y Ambientales de Puerto Rico (DRNA: 2017-IC017, O-VS-PVS15-SJ-00894-15122016). We also thank Philippe Gaucher, Antoine Fouquet, and Alberto Puente-Rolón for assistance with permitting and supplies.


This article is a PNAS Direct Submission.



School of Life and Environmental Sciences, The University of Sydney, Sydney, NSW, Australia 2006;
School of Life and Environmental Sciences, The University of Sydney, Sydney, NSW, Australia 2006;
School of Life and Environmental Sciences, The University of Sydney, Sydney, NSW, Australia 2006;
Department of Biological Sciences, Macquarie University, Sydney, NSW, Australia 2113;
School of Life and Environmental Sciences, The University of Sydney, Sydney, NSW, Australia 2006;
Institut de Recherche pour le Développement, UMR 241 EIO (Écosystèmes Insulaires Océaniens), Faa’a, Tahiti, French Polynesia, France 98702


To whom correspondence may be addressed. Email: [email protected].
Author contributions: J.L.D., M.R.C., R.S., and S.D. designed research; J.L.D., M.R.C., and S.D. performed research; J.L.D. analyzed data; and J.L.D. wrote the paper.

Competing Interests

The authors declare no competing interest.

Metrics & Citations


Note: The article usage is presented with a three- to four-day delay and will update daily once available. Due to ths delay, usage data will not appear immediately following publication. Citation information is sourced from Crossref Cited-by service.

Citation statements



If you have the appropriate software installed, you can download article citation data to the citation manager of your choice. Simply select your manager software from the list below and click Download.

Cited by


    View Options

    View options

    PDF format

    Download this article as a PDF file


    Get Access

    Login options

    Check if you have access through your login credentials or your institution to get full access on this article.

    Personal login Institutional Login

    Recommend to a librarian

    Recommend PNAS to a Librarian

    Purchase options

    Purchase this article to access the full text.

    Single Article Purchase

    The evolution of targeted cannibalism and cannibal-induced defenses in invasive populations of cane toads
    Proceedings of the National Academy of Sciences
    • Vol. 118
    • No. 35







    Share article link

    Share on social media