It is lonely at the front: contrasting evolutionary trajectories in male and female invaders.

Invasive species often exhibit rapid evolutionary changes, and can provide powerful insights into the selective forces shaping phenotypic traits that influence dispersal rates and/or sexual interactions. Invasions also may modify sexual dimorphism. We measured relative lengths of forelimbs and hindlimbs of more than 3000 field-caught adult cane toads (Rhinella marina) from 67 sites in Hawai'i and Australia (1-80 years post-colonization), along with 489 captive-bred individuals from multiple Australian sites raised in a 'common garden' (to examine heritability and reduce environmental influences on morphology). As cane toads spread from east to west across Australia, the ancestral condition (long limbs, especially in males) was modified. Limb length relative to body size was first reduced (perhaps owing to natural selection on locomotor ability), but then increased again (perhaps owing to spatial sorting) in the invasion vanguard. In contrast, the sex disparity in relative limb length has progressively decreased during the toads' Australian invasion. Offspring reared in a common environment exhibited similar geographical divergences in morphology as did wild-caught animals, suggesting a genetic basis to the changes. Limb dimensions showed significant heritability (2-17%), consistent with the possibility of an evolved response. Cane toad populations thus have undergone a major shift in sexual dimorphism in relative limb lengths during their brief (81 years) spread through tropical Australia.

Introduction

Invasive species offer unparalleled opportunities to explore the process of rapid evolutionary change [1]. As an alien species spreads through previously uncolonized territory, it is likely to encounter novel selective forces (both biotic and abiotic); and the process of continuous range expansion introduces an additional set of evolutionary processes (e.g. genetic drift, mutation surfing, spatial sorting [2]). As a result, alien taxa often accumulate substantial phenotypic changes, at a timescale much quicker than usually envisaged for evolutionary change [1,3,4]. One interesting subset of traits that might be expected to evolve during a biological invasion involves sexually dimorphic characteristics. Morphological disparities between conspecific males and females take many forms [5], but evolutionary theory suggests that natural selection during an invasion might act most forcefully on traits that affect dispersal rate and/or reproductive characteristics [3].

Dispersal rate is a key feature of invasion biology, and an extensive literature suggests that dispersal rate typically evolves upwards during an invasion [6-10]. As a result, individuals in the invasion vanguard tend to exhibit dispersal-enhancing features (such as seeds that drift further on the wind, larger feet, wings or flight muscles [3,11-15]) relative to conspecifics in the range-core. Alleles that code for fast-dispersal morphological traits may accumulate in the invasion vanguard because of spatial sorting (successive generations of interbreeding between the fastest dispersers [2]) or natural selection (reflecting fitness benefits to unusually fast-moving individuals [16]). However, what happens to this acceleration if males and females within a population differ in traits (body size, limb dimensions, wing size, frequency of winged morph, etc.) that influence dispersal rate? If one sex is intrinsically faster than the other, individuals of that sex may be under intense counter-selection against more rapid dispersal, because they would encounter no potential mates at the expanding range-edge [17]. Thus, we might expect to see an evolutionary reduction in the magnitude of sexual dimorphism in traits that enhance dispersal rate.

Sexual selection plays a critical role in the evolution of sexually dimorphic traits, reflecting the way that specific morphological features affect individual reproductive success [5,18]. As a result, geographical differences in mating systems within widespread species often are associated with variation in patterns of sexual dimorphism [19,20]. By definition, population density is low at an invasion front (density is zero immediately in advance of the front) and, hence, individuals in the invasion vanguard may face very different rates of encounter with mates and consexual rivals than would be the case in the range-core [21]. That disparity would be exacerbated by any shifts in the operational sex ratio between the range-edge and range-core [21]. In circumstances where mates are rare, individual fitness will depend on traits that promote survival and longevity rather than success in mate competition [22-25]. In contrast, high population density can increase intrasexual competition and result in the evolution of secondary sexual characters such as ornaments, weaponry or sexual size dimorphism [5,18,26,27]. This is especially true for the sex in which reproductive success is limited by access to mates (typically males [21]), and stems from the disparity in genetic interests between the sexes (‘sexual conflict’ [18,28]). Thus, both the ‘dispersal rate’ and ‘sexual selection’ hypotheses predict the evolution of reduced sexual dimorphism at an expanding range-edge.

To test this prediction, we gathered data on cane toads (Rhinella marina; formerly Bufo marinus) from 67 populations across the species' invaded range in Hawai'i and tropical Australia. Importantly, dates of introduction or arrival at each of these sites are well documented, allowing us to explore shifts in sexual dimorphism as a function of time since colonization (TSC). Densities of invasive cane toad populations vary dramatically with TSC [29]. Comparisons between range-edge and range-core toad populations within tropical Australia have demonstrated substantial shifts in morphology [30,31], locomotion [32,33], dispersal ability [2,8,9,34-37], hydric and thermal tolerance [38,39], immune function [40], life-history traits [41], reproductive frequency [42], incidence of spinal arthritis [43] and larval plasticity [44]. Thus, cane toads are a plausible candidate species for examining invasion-associated shifts in sexual dimorphism.

We chose to focus on length of forelimbs and hindlimbs relative to the toad's snout–vent (body) length (SVL). As well as being straightforward to measure accurately, limb dimensions have been reported to influence mating success of male cane toads [45,46], and toads with longer hindlimbs travel further distances daily when radio-tracked [30]. Thus, relative limb length is a likely target of natural selection and spatial sorting as well as sexual selection. Additionally, our preliminary analyses revealed significant sexual dimorphism in this trait: male cane toads have longer limbs (relative to body length) than do conspecific females, a pattern that is also widespread in other amphibians [47].

Simply documenting geographical variation in a trait (or dimorphism in that trait) does not establish any evolutionary causation, however. Anuran amphibians (including cane toads) exhibit high levels of developmental plasticity. The conditions in a water body (temperature, food supply, competitor density, predator cues, etc.) can significantly affect the morphological traits of toads that metamorphose from that site [48,49]. In at least some anurans, rearing conditions of tadpoles can influence relative limb lengths of metamorphs [50-53]. To sustain an explanation couched in terms of evolution rather than developmental plasticity, we need to breed toads in captivity, and raise their offspring to maturity in standard conditions. Unless offspring resemble their parents in relative limb length, and geographical divergences in morphology among wild-caught animals are also seen in their offspring even after the latter are raised in standardized conditions, we cannot attribute divergences among populations to evolutionary forces rather than to plasticity. Quantifying heritability also allows us to determine if morphological traits are likely to be able to respond to selection. We thus also conducted a breeding study.

Material and methods

Capture of specimens and sites of collection

From August 2013 to March 2016, we collected adult cane toads from sites along a transect in tropical Australia (total N = 2076; Queensland N = 467, Northern Territory N = 802, Western Australia N = 807). We also collected toads from three Hawaiian islands (N = 1018; Hawai'i N = 529, O'ahu N = 333, Maui N = 156) between January and July of 2015. Toads were captured by hand, and we used vernier callipers (±0.1 mm) to measure the SVL and limb lengths of each toad. The hand, radioulna and humerus were measured for the forelimb (arm), whereas the femur, tibiofibula and foot were measured for the hindlimb (leg). Values for each component of a limb were added together to obtain measures of total forelimb length and total hindlimb length. Toads were sexed by examining external morphological characteristics (e.g. males possess nuptial pads on the thumbs, rugose dorsal skin and yellow coloration) and vocalizations (e.g. males observed calling, or producing release calls upon handling). Most individuals greater than 90 mm in SVL are sexually mature [54].

Common garden offspring

We collected a subgroup of adult Australian cane toads (approx. 25 males and 25 females per population) from the two extremes of the invaded range to conduct a ‘common garden’ breeding experiment. These toads were collected from three long-established populations in northeastern Queensland (more than 70 years since colonization; Townsville, Innisfail, Tully) and four recently colonized sites in northern Western Australia (less than 3 years since colonization; El Questro, Purnululu, Wyndham, Oombulgurri). Using protocols outlined by [55], we induced breeding pairs to spawn by injection of leuprorelin acetate (Lucrin; Abbott Australasia, Botany, NSW) using 1 ml of Lucrin diluted 1 : 20 with saline and raised the resulting progeny in captivity at our field station in the Northern Territory (12°37′ S, 131°18′ E). All dams and sires were bred only once in this study, thus F1 individuals from each clutch were full-siblings. Once metamorphic toads attained SVLs more than 20 mm, we toe-clipped them for identification and moved them into outdoor enclosures in groups of 30 (with mixed parental origins). As F1 toads grew, they were split into smaller groups (approx. 10 by adulthood) to reduce competition for food and space, and avoid cannibalism. From this common garden study, we obtained data on 489 captive-raised offspring (287 Queensland, 202 Western Australia) from 31 egg clutches (16 Queensland, 15 Western Australia). We measured the same traits on these offspring as we did on wild toads. Offspring were measured at approximately 2, 8, 14 and 17 months of age to quantify changes in skeletal morphology with growth and maturity. By their fourth measurement, 184 individuals had reached maturity and could be sexed based on the same criteria as used for wild-caught toads (see above).

Statistical analyses

We used multiple regressions to assess the effects of sex and TSC (in categories of 0–10 years, 11–20 years, 41–50, 51–60, 71–80 years) on relative arm and leg lengths of wild toads. The SVL was included as a covariate in the models to control for body size. For graphical purposes, we calculated % limb length by dividing arm and leg measures by SVL. We calculated the difference in mean values of % limb lengths (for arms and legs separately) of males and females as an index of sexual dimorphism. To estimate the repeatability of our measurements of toad limbs, on one occasion, we took triplicate measures from five individuals. We used the R package rptR [56] to calculate repeatability of measures of each limb component.

We used linear-mixed models to determine whether the relative length of limbs (and their components) of male and female toads raised in a common environment were affected by their parents' location of origin (long-established populations in QLD versus invasive populations in WA). State and sex and their interaction were included as fixed effects in the models, along with SVL to control for body size. Individual ID (nested within clutch and state) and clutch (nested within state) were included as random effects in the models. This analysis was conducted on the final measurements made on the 184 individuals that were mature at the end of the study (76 F, 108 M).

To compare patterns of relative limb length of toads raised under common garden conditions to those exhibited by wild-caught toads from QLD and WA, we used a subset of the data on wild toads (limited to individuals from those two regions). We used multiple regression with state and sex and their interaction as independent variables.

As an additional, more formal test of the effect of rearing environment on locational differences in morphological sexual size dimorphism (SSD), we performed multiple regression combining data from wild-caught and common garden toads. The model included full factorial interactions among sex, state (WA versus QLD) and source (wild versus common garden). SVL was included as a covariate in the model to control for body size relationships with limb lengths. A caveat for this analysis is that because we lacked data on relatedness among wild toads, we exclude familial effects from the model. Thus sib-ships among common garden toads were ignored.

To assess familial similarity in limb morphology in a formal quantitative genetics framework, we also ran an ‘animal model’ using ASREML software (VSN International, Hemel Hempstead, UK) [57]. When pedigree information is available (as is the case for our common garden offspring), animal models can be used to estimate the genetic underpinning of phenotypic variation [57]. Because most individuals were measured on more than one occasion, we were also able to estimate the ontogenetic repeatability of relative limb lengths. We incorporated offspring ID and parental ID as random effects in the animal model and included SVL as a covariate to correct for body size and age. Although our sample was adequate to detect heritabilities, it was too small to calculate genetic correlations among traits [57]. All other analyses were performed using JMP 11 software (SAS Institute, Cary, NC).

Results

Sexual dimorphism varies with time since colonization in field-caught toads

Relative limb lengths of wild-caught toads changed in different patterns with TSC in males versus females. Interactions between sex and TSC were highly significant both for arm length and leg length (table 1 and figure 1). Both sexes exhibited a ‘U’-shaped' (curvilinear) pattern with TSC for both limbs (figure 1); relative limb lengths were lowest in populations of intermediate TSC (11–20 years) than in either range-core or invasion-vanguard populations. Males had longer limbs than females in every population, but the magnitude of dimorphism was lower in recently invaded areas (i.e. Northern Territory and Western Australia) than in areas that were colonized many decades ago (i.e. Hawai'i, Queensland). The highest values for relative limb lengths in females were seen in invasion-front populations, whereas the highest values for males were seen in long-established populations.

Table 1.

Effect of sex and time since colonization (TSC) categories on relative arm and leg lengths of 3094 wild-caught cane toads from Hawai'i (N = 1018) and Australia (N = 2075). Statistically significant values (p < 0.05) are highlighted in italics.

trait	variable	d.f.	mean square	F-value	p-value
total arm length	SVL	1	241558.3	19738.4	<0.0001
	sex	1	1973.7	161.3	<0.0001
	TSC	4	1302.6	106.4	<0.0001
	sex × TSC	4	407.3	33.3	<0.0001
	error	3080	12.2
total leg length	SVL	1	704776.5	21233.9	<0.0001
	sex	1	2961.7	89.2	<0.0001
	TSC	4	3534.1	106.5	<0.0001
	sex × TSC	4	760.6	22.9	<0.0001
	error	3080	33.2

Figure 1.

Comparisons of relative leg (top left) and arm (bottom left panel) lengths of 3094 wild-caught male and female cane toads from locations with different colonization times. Males are represented by open symbols and dashed lines, females by closed symbols and solid lines. Relative limb length values are expressed as a percentage of body length. Sexual dimorphism indices for each limb (calculated as the mean male value minus the mean female value) are presented in the corresponding right-hand panels. Error bars represent 1 standard error from the mean.

Sexual dimorphism varies with parental location among common garden toads

Among the 184 captive-reared toads that reached maturity by the end of the study, the relative lengths of the humerus, foot, tibia and leg were affected by interactions between sex and the state of origin of their parents (QLD versus WA; table 2 and figure 2). These interactions broadly mirrored those seen in wild-caught toads from QLD versus WA (table 3 and figure 2). However, among wild toads, significant sex × state interactions were detected in all limb measures, whereas the same interactions were not statistically significant for some measures (hand, radioulna, arm and femur) among our smaller sample of common garden toads (table 2). When we assessed the relative lengths of the humerus, foot, tibia and leg in an analysis that combined data from the 184 common garden toads with the 1274 wild toads from the source populations, it verified the same significant interaction effect of state and sex found when each group was analysed separately (table 4). For all these measures, wild toads had significantly longer relative limb lengths than common garden toads. However, although source (wild versus common garden) was a significant main effect in all analyses, it did not appear in any significant interactions (table 4). Hence, wild-caught toads and those reared in captivity exhibited the same underlying pattern of decreased sexual dimorphism in relative limb lengths in invasive (WA) versus established (QLD) populations. This further supports the genetic basis of the patterns.

Table 2.

ANOVA results show the effects of sex and state on limb sizes of 184 cane toads reared under common garden conditions. Thirty full-sib families produced offspring that reached maturity during the study and could be sexed based on secondary sexual characteristics. The parents of 17 families originated from long-established populations (Queensland) and the parents of 13 families originated from invasion-front populations (Western Australia). Only the final set of morphological measurements taken from each individual were included in this analysis. Snout-to-vent length (SVL) was included as a covariate in the model to adjust trait measures for body size. Family (nested within state) was included as a random effect in the model. Statistically significant values (p < 0.05) are highlighted in italics.

	SVL		sex		state		sex × state
Trait	F	P	F	P	F	P	F	P
hand	260.8	<0.0001	5.1	0.0255	5.9	0.0225	1.5	0.2255
radioulna	430.8	<0.0001	68.9	<0.0001	2.0	0.1647	0.4	0.5380
humerus	466.3	<0.0001	94.5	<0.0001	1.0	0.3167	7.3	0.0075
total arm	624.0	<0.0001	78.7	<0.0001	1.2	0.2804	1.1	0.2996
foot	557.5	<0.0001	14.5	0.0002	0.4	0.5495	4.7	0.0315
tibia	783.1	<0.0001	42.0	<0.0001	0.6	0.4666	4.8	0.0300
femur	422.4	<0.0001	21.8	<0.0001	1.0	0.3316	3.6	0.0587
total leg	717.1	<0.0001	29.2	<0.0001	0.7	0.4020	5.7	0.0177

Figure 2.

Relative limb lengths of cane toads from long-colonized versus invasion-front populations, as a function of whether the toads were wild-caught (right-hand panels) or captive-raised (left-hand panels). The left-hand panels show statistically significant interactions between sex and state for limb measures of male (open symbols, dashed line) and female (closed symbols, solid line) cane toads reared in a common environment. The parents of these toads originated either from long-established populations in QLD or from invasive populations in WA. Significant interactions for the same limb measures seen in wild male and female toads from the same locations are depicted in the right-hand panels. In each case, the limb length is expressed as a percentage of toad snout–vent length. Error bars represent one standard error from the mean.

Table 3.

ANOVA results show the effects of sex and state on limb sizes of 1274 wild-caught cane toads from long-established populations in Queensland (N = 467) and invasion-front populations in Western Australia (N = 807). Statistically significant values (p < 0.05) are highlighted in italics.

	SVL		sex		state		sex × state
trait	F	p	F	p	F	p	F	p
hand	3082.7	<0.0001	32.1	<0.0001	39.1	<0.0001	26.1	<0.0001
radioulna	4853.4	<0.0001	291.5	<0.0001	8.4	0.0037	61.6	<0.0001
humerus	4607.6	<0.0001	413.4	<0.0001	1.6	0.2087	51.4	<0.0001
total arm	6714.4	<0.0001	357.5	<0.0001	17.1	<0.0001	73.2	<0.0001
foot	5127.2	<0.0001	90.6	<0.0001	24.8	<0.0001	42.5	<0.0001
tibia	8189.1	<0.0001	260.7	<0.0001	11.1	0.0009	71.5	<0.0001
femur	5025.2	<0.0001	119.5	<0.0001	12.4	0.0004	36.4	<0.0001
total leg	7615.8	<0.0001	174.7	<0.0001	22.8	<0.0001	61.4	<0.0001

Table 4.

Effects of sex, state and source (wild versus common garden) on limb sizes of 1274 wild and 184 captive-reared toads from long-established populations in Queensland and invasion-front populations in West Australia. The significant main effect of source in all cases indicates that wild toads had relatively longer limbs than captive-reared toads. The significant shift in sexual dimorphism of limb lengths between states (indicated by significant sex × state interactions) does not differ between wild and common garden toads (all three-way interactions n.s.). Statistically significant values (p < 0.05) are highlighted in italics.

trait	effect	d.f.	F	p
total leg length	SVL	1,1447	8339.8	<0.0001
	state	1,1447	9.3	0.0023
	sex	1,1447	94.2	<0.0001
	state × sex	1,1447	21.9	<0.0001
	source	1,1447	182.3	<0.0001
	state × source	1,1447	0.23	0.6351
	sex × source	1,1447	0.5	0.4884
	state × sex × source	1,1447	0.4	0.5403
humerus	SVL	1,1449	5086.1	<0.0001
	state	1,1449	3.5	0.0628
	sex	1,1449	243.8	<0.0001
	state × sex	1,1449	21.9	<0.0001
	source	1,1449	114.5	<0.0001
	state × source	1,1449	0.7	0.4204
	sex × source	1,1449	2.3	0.1277
	state × sex × source	1,1449	0.04	0.8440
tibia	SVL	1,1449	9008.1	<0.0001
	state	1,1449	5.6	0.0180
	sex	1,1449	136.7	<0.0001
	state × sex	1,1449	23.4	<0.0001
	source	1,1449	167.1	<0.0001
	state × source	1,1449	0.01	0.9194
	sex × source	1,1449	0.3	0.5648
	state × sex × source	1,1449	0.7	0.3887
foot	SVL	1,1447	5697.3	<0.0001
	state	1,1447	6.9	0.0087
	sex	1,1447	47.7	<0.0001
	state × sex	1,1447	14.7	0.0001
	source	1,1447	121.7	<0.0001
	state × source	1,1447	1.1	0.2939
	sex × source	1,1447	0.2	0.6916
	state × sex × source	1,1447	0.4	0.5483

Estimates of heritability

Our estimates of heritability for relative limb measures (table 5) ranged from 0.02 ± 0.021 (femur) to 0.17 ± 0.043 (hand). These values suggest that most measures of limb morphology have a genetic component and hence are capable of responding to selection. The heritability estimates for femur and humerus, however, were low, and the confidence limits around them encompassed 0. These two traits may be less likely to respond to selection. We were able to measure limb components with high repeatability (table 5). Tibia length was the most highly repeatable measurement (0.99), and hand was the least repeatable (0.80). As toads grew larger, relative lengths of their limbs were moderately stable; ontogenetic repeatability ranged from 0.19 (foot) to 0.39 (tibia).

Table 5.

Estimates of heritability, ontogenetic repeatability (of measures made at different ages on the same animals over a long period) and measurement repeatability (of successive measures taken on the same animal over a brief period) of limb measurements of cane toads reared in a common environment. Heritability and ontogenetic repeatability estimates were calculated from data on 550 individual toads (489 offspring, 61 parents). Measurement repeatability was calculated from a sample of five toads from which triplicate measures were made on a single occasion.

trait	heritability	ontogenetic repeatability	measurement repeatability
hand	0.17 ± 0.043	0.34 ± 0.033	0.80 ± 0.203
radioulna	0.17 ± 0.052	0.35 ± 0.033	0.91 ± 0.137
humerus	0.03 ± 0.024	0.33 ± 0.032	0.89 ± 0.147
total arm	0.13 ± 0.046	0.29 ± 0.033	0.92 ± 0.122
foot	0.11 ± 0.034	0.19 ± 0.033	0.97 ± 0.092
tibia	0.11 ± 0.044	0.39 ± 0.032	0.99 ± 0.025
femur	0.02 ± 0.021	0.23 ± 0.033	0.93 ± 0.101
total leg	0.10 ± 0.037	0.23 ± 0.033	0.98 ± 0.043

Discussion

The cane toad invasion of Australia has been accompanied by rapid changes to skeletal morphology. Among wild toads, relative leg lengths (ancestrally high) have decreased in the course of the invasion across Australia, and then increased again in invasion-vanguard populations. In contrast, sexual dimorphism in relative limb lengths exhibits a simpler monotonic decline: toads in ancestral (range-core) populations are highly dimorphic, whereas toads in invasion-front populations show little sex-based divergence in limb lengths. Relative limb length exhibits significant heritability, and captive-raised toads show similar patterns of morphology as their wild-caught parental populations, suggesting that these morphological shifts may represent evolved changes rather than (or as well as) developmentally plastic responses to different environments.

First, what processes have driven the curvilinear pattern of changes in relative limb length during the toads' Australian invasion? Annual rates of range expansion increased substantially over this period (from 10 to 15 km per annum to more than 60 km per annum [30]), potentially placing major stresses on a body plan that is poorly suited to continuous long-distance travel [43,58]. Long limbs provide high propulsive power for leaping, an effective tactic to evade an oncoming predator [59-62] but may be poorly suited to long periods of continuous slow dispersal over irregular terrain. Cane toads are capable of multiple locomotor modes, and small frequent hops (‘bounding’) may be more energetically efficient at traversing long distances than is a reliance on large single hops (‘leaping’ [63]). In addition, shorter hops may reduce biomechanical stresses on the toad's body. Long legs in invasion-front cane toads are associated with a high incidence of spinal arthritis [43,58]. Hence, natural selection in the course of the toad's long march across tropical Australia may have favoured individuals with shorter-than-average arms and legs that moved by bounding rather than by leaping. The arms play a major functional role in bounding [63-65], consistent with the shifts seen in both forelimbs and hindlimbs.

Why has this process reversed in populations close to the invasion front, with longer-limbed individuals (of both sexes) in these western sites? This reversal may be due to spatial sorting rather than natural selection. Even if selection favours animals with shorter limbs, the rate of dispersal is highest for long-legged toads (based on radio-tracking [30]). Thus, alleles for longer legs accumulate in the vanguard of the invasion, regardless of whether or not they enhance fitness of their bearers [2]. The result is that leg length decreases over the course of the invasion, but with a reversal close to the invasion front where spatial sorting overrides natural selection. Alternatively (or additionally), selective advantages that accrue to individuals in the invasion vanguard (more food, owing to lower densities of conspecifics) may favour maximal dispersal rates (and thus, longer legs) in this phase of the invasion. Faster dispersal would not confer the same fitness benefits in longer-colonized areas, because it would not enable individuals to reach low-density populations. Our data thus extend and clarify a previous report of longer hindlimbs in toads close to the invasion front [30]. Our extensive sampling reveals a more complex scenario, with invasion driving a reduction in relative limb length, but reversing to a rapid increase in leg (and arm) length near the invasion front.

In contrast to the curvilinear trends in relative limb lengths with TSC, the magnitude of sexual dimorphism in limb dimensions showed a rapid decline in populations colonized between 40 and 20 years ago (figure 1). Consistent with their greater limb lengths, male cane toads can travel faster than females (C. M. Hudson 2016, unpublished data from locomotor trials; see [33] for a similar sex difference in agility). All else being equal then, the evolution of more rapid dispersal during the toad invasion would have resulted in males substantially out-pacing females, thereby reaping the benefits of enhanced food availability at the invasion front [16], but at the cost of a highly skewed operational sex ratio (or in the extreme, a lack of females). The high degree of sex-based divergence in relative limb length in cane toads from range-core populations may result from sexual selection; previous studies have documented an association between limb muscle mass and reproductive success in male cane toads in the field [45,46]. Hence, selection may have favoured larger and/or more muscular limbs in male toads than females in ancestral populations; this dimorphic condition is widespread among anurans in general, including bufonids [47]. As soon as toads began dispersing westwards from Queensland, however (and especially, when that rate of dispersal accelerated [66]), novel forms of selection could have come into play, reshaping the ancestral cane toad body plan. First, low-density populations may render male–male competition less important at the invasion front (unless that trend is opposed by a shift in the operational sex ratio) and second, limb lengths affect not only potential dispersal rates, but also the energy and wear-and-tear associated with long-distance travel [43]. We cannot tease apart the relative importance of those two processes—sexual selection and natural selection—in driving cane toads towards sexual monomorphism, nor can we convincingly distinguish the impact of spatial sorting from selection. Nonetheless, our data strongly support the a priori prediction that a biological invasion can impose novel evolutionary forces that reduce the degree of sexual dimorphism in ancestral (range-core) populations.

We base our interpretative scenarios on adaptive mechanisms shifting morphological traits of toads over the course of their invasion. The similarity of traits between common garden-reared and wild toads and the non-zero heritability suggest that the trait changes have a genetic basis and are capable of responding to selection. Despite the plastic effects wrought by rearing differences (i.e. wild toads have significantly longer limbs than captive toads), the shift in limb dimorphism between established and invasive populations is strongly evident. Verification that the differences in traits have arisen through selection would require Qst – Fst analysis to compare shifts in the quantitative traits to concurrent shifts in neutral traits, however [53].

The heritabilities we calculated for relative limb length—around 10%—are lower than reported for morphological traits in many other species of animals [67]. Clearly, that leaves room also for significant environmental influences. Future work could usefully explore the sensitivity of toad limb lengths to larval conditions (as in [53]) and characterize the mating systems of toads at the invasion front compared with range-core populations. Regardless of uncertainty about causal mechanisms, however, our data document a substantial shift in morphology and sexual dimorphism within an invasive species, within the span of a single human lifetime.