Captive breeding and the fitness of reintroduced species: a test of the responses to predators in a threatened amphibian


Richard A. Griffiths, The Durrell Institute of Conservation and Ecology, University of Kent, Canterbury, Kent CT2 7NR, UK (e-mail


  • 1Although captive breeding and reintroduction is a high-profile management tool for many threatened species, it is unclear how long-term captive breeding can influence fitness attributes such as natural defences to predators.
  • 2Induced defences that have evolved in the Mallorcan midwife toad Alytes muletensis in response to introduced predators were compared in natural and reintroduced populations that had a common ancestry, and in short-term and long-term captive populations that differed in ancestry.
  • 3Defences against predators were maintained in a reintroduced population derived from stock that had passed through three to eight generations of captive breeding prior to release into a predator-free area. Heterozygosity did not differ between natural and reintroduced populations, but the reintroduced population displayed lower allelic richness.
  • 4A comparison between populations maintained for different lengths of time in captivity revealed a significant reduction in one defensive trait in stock maintained for more than eight generations. Neutral genetic variation (i.e. heterozygosity and allelic richness) did not differ between the short-term captive population and a natural population, but there was a significant loss of genetic variation in the long-term captive population.
  • 5Synthesis and applications. The results suggest that relatively high levels of heterozygosity and important fitness attributes can be maintained for a few generations in breeding programmes for threatened species despite small numbers of founders and the absence of natural selection. Nevertheless, both fitness and heterozygosity may eventually start to deteriorate in the long term, and this may have implications for reintroduction strategies.


Saving species from extinction through captive breeding and reintroduction is a high-profile yet controversial conservation tool (Beck et al. 1994; Snyder et al. 1996; Hodder & Bullock 1997; Ebenhard 1998; Fischer & Lindenmayer 2000). Maintaining captive populations of endangered species for conservation purposes invariably leads to some degree of inbreeding, because such populations are often founded by a small number of individuals. Inbreeding can cause loss of fitness (Miller & Hedrick 1993; Woodworth et al. 2002; Reed et al. 2003) and generate correlations among loci throughout the genome. Moreover, removal of natural selection and rapid adaptation to captivity is a further concern for captive breeding programmes (Latter & Mulley 1995; Lewis & Thomas 2001; Gilligan & Frankham 2003; Stockwell & Ashley 2004). Previous studies of the effects of captive breeding on fitness have generally focused on model organisms or species of commercial importance (e.g. Drosophila, Woodworth et al. 2002; Gilligan & Frankham 2003; butterflies, Lewis & Thomas 2001; Joron & Brakefield 2003; salmon, Säisä, Koljonen & Tähtinen 2003; Fox & Heath 2003). Such studies have not explicitly addressed fitness attributes that influence ecological interactions between species in the wild. Equally, it has so far proved intractable to study such effects in the relatively few endangered species that have been successfully reintroduced to the wild (Beck et al. 1994; Snyder et al. 1996; Fischer & Lindenmayer 2000).

One of the most important traits to maintain in captive populations is the ability to detect and respond to natural predators. Many amphibian tadpoles show short-term changes in morphology, development or behaviour in response to predators (Van Buskirk & McCollum 2000). These induced defences are triggered by environmental cues, but are under genetic control and are adaptive. They can be maintained by trade-offs arising as a result of environments varying in space and time, causing the fitness ranks of alternative phenotypes to vary across environments (Via & Lande 1985). Captive breeding is usually performed in stable, unchanging environments without predators. This, together with inbreeding effects, may mean that crucial anti-predator responses disappear during captive breeding, posing serious risks to the success of future reintroductions from these programmes (Fleming et al. 2000; Fox & Heath 2003). We investigated whether induced defences to predators were maintained in populations after several generations of captive breeding for conservation purposes.

The threatened Mallorcan midwife toad Alytes muletensis (Sanchíz & Adrover 1979) has been the subject of a highly successful captive breeding and reintroduction programme (Bloxam & Tonge 1994; Buley & Garcia 1997). Indeed, out of 5743 amphibian species recently assessed using the IUCN Red List criteria (Stuart et al. 2004), the Mallorcan midwife toad was the only one to be downlisted from ‘critically endangered’ to ‘vulnerable’. As in many other amphibians, the tadpoles of this species display induced defences in the presence of predators. These defences have evolved in response to two predators that were introduced to the island about 2000 years ago: the viperine snake Natrix maura (Linnaeus, 1758) and the green frog Rana perezi Seoane 1885 (Griffiths et al. 1998; Moore et al. 2004). Under laboratory conditions induced responses consist of the tail becoming longer and the tail fins shallower, coupled with faster development (Moore et al. 2004). Changes in tail shape may enable tadpoles to increase acceleration and facilitate escape once the tadpole is detected by a predator (Van Buskirk & McCollum 2000), while faster development allows a more rapid escape from a predator environment (Babbit & Tanner 1998).

In this study we used induced defences in response to introduced predators to compare the fitness of (i) natural and reintroduced populations; and (ii) populations maintained in captivity for different lengths of time. We also compared genetic variation in the studied populations using microsatellite DNA analysis.

Materials and methods

induced defences

In the first experiment we compared the responses of tadpoles in a predator-free, reintroduced population with those in the population from which their ancestors were collected several years earlier (i.e. a ‘founder’ vs. ‘descendent’ comparison). The descendent population was successfully reintroduced to Mallorca between 1992 and 1994 using captive-bred stock derived from 14 adults, originally collected from the founder population in 1985 and 1987 (Garcia & Buley 1997). At the time of the reintroductions this population had been bred in captivity for three to eight generations. To induce a response we exposed tadpoles collected from the two populations to chemical cues from two predators (viperine snake and green frog). Forty-eight tadpoles from each population were netted from two natural breeding pools for the experiments. As Mallorcan midwife toads have relatively small clutch sizes (mean 11·15; Tonge & Bloxam 1991), it can be assumed that the samples taken from each pool represented several sibships. The tadpoles from each population were divided equally between three treatments (control, snake and frog). Four tadpoles were housed in each of 24 tubs containing 2·5 L of appropriately conditioned well water and fed fish flakes ad libitum. The design therefore comprised two populations × three predator treatments, with each treatment replicated four times. Statistical analyses used mean values from each tub to ensure independence between observations. The tubs were randomly placed on the floor of a farmhouse in Mallorca and temperature (average minimum/maximum 16–23 °C) and light–dark cycle followed normal day patterns (May–June).

In the second experiment, we compared the responses to snakes in two current captive populations. The ‘long-term’ stock was established between 1985 and 1987 (see above, by now captive bred for nine to 12 generations). The ‘short-term’ stock was established from 25 tadpoles collected from a different population in 1997 (i.e. captive bred for one to two generations). Forty tadpoles of each population were housed individually in 1-L tubs containing 0·5 L dechlorinated, appropriately conditioned, water. Twenty tadpoles per population were allocated to either the control or the snake group. Tadpoles were fed rabbit pellets and fish flakes ad libitum. The design therefore consisted of two populations × two predator (i.e. snake vs. control) treatments, with each treatment replicated 20 times. This experiment took place in a temperature-controlled room (18 °C) with a 12 : 12 artificial light–dark cycle at the University of Kent (Canterbury, UK).

In both experiments, boxes of tadpoles were arranged randomly and rerandomized after each water change, which occurred every 2–3 days. Eight morphological measurements (body length, body width, body depth, tail length, tail muscle depth, and upper, lower and maximum tail fin) and development stage (Cambar & Martin 1959) were recorded on tadpoles anaesthetized with MS222 (Sigma-Aldriech, Gillingham, Dorset, UK) at the start of the experiments and at intervals of 14–15 days. Measurements were made using dial callipers to 0·1-mm precision. Water was either unconditioned (control), snake-conditioned or frog-conditioned by placing no predator, three adult snakes or three adult frogs in 20 L water 3 h before the water was used. The tadpoles were fed after each water change. At the completion of each experiment tail tips (3 mm) were taken for microsatellite analyses.

microsatellite dna analyses

DNA was extracted from the tail tips and used in polymerase chain reactions (PCR) to amplify 10 microsatellite loci as described in Kraaijeveld-Smit et al. (2003). Nine (Amul1, 3, 6, 8, 11, 14, 15, 20 and 22) out of the 10 loci used were previously characterized for this species (Kraaijeveld-Smit et al. 2003, 2005). The primer pair Amul23 (5′-3′ F: TTT TGT TTT TCA CTA CAT TAT CC, R: TCA CAG TTC GAT TTC ACA GA; GenBank AY499619) was newly characterized, as in Kraaijeveld-Smit et al. (2003). PCR conditions for this locus were as follows: one cycle of 94 °C for 4 min, followed by touch-down cycles 94 °C for 30 s, annealing temperature (Ta) for 30 s, 72 °C for 30 s, and a final hold of 72 °C for 10 min. The first cycle started at a Ta of 57 °C, followed by one cycle of 55 °C, one of 53 °C, one of 51 °C, four of 49 °C and four of 47 °C, with a final Ta of 46 °C (23 cycles). Individual heterozygosity (H) levels were obtained by scoring each locus as either being heterozygous (1) or homozygous (0), and averaging these scores for all 10 loci for each individual.

statistical analysis

Morphology was compared between treatments and populations using relative trait sizes (Van Buskirk & McCollum 2000; Moore et al. 2004). Relative trait sizes are the unstandardized residuals obtained by regressing each log-transformed trait size against the first principal component (PC1; which represented body size). Relative trait sizes were calculated for two variables (tail length and lower tail fin) that have consistently responded to predator cues in previous experiments (Van Buskirk & McCollum 2000; Moore et al. 2004).

mancova was performed on the two relative trait sizes to test for variation in tadpole shape between populations and treatments. Starting size and heterozygosity were included as covariates. Univariate analyses (ancova) were then conducted on each variable separately to compare differences between treatments and populations. Heterozygosity and relative trait size at day 0 were added as covariates in the analyses. ancovas were also used to compare log-transformed developmental stage between treatments (Relyea 2001) and populations with heterozygosity as covariate.

Differences in heterozygosity levels between populations were tested using non-parametric tests. To compare the number of alleles between populations a correction for sample size was made using the allelic richness option in the fstat 2·9·3 software (Goudet 1995). A Wilcoxon signed-ranks test for paired observations was used to compare the number of alleles between populations.


induced defences

In tadpoles tested from wild populations, chemical cues from predators induced changes in tail shape in tadpoles (mancova, F4,28 = 4·56, P= 0·006) but there was no difference between founder and descendent populations (F2,14 = 0·71, P= 0·510) and no population × treatment interaction (F4,28 = 1·26, P= 0·311). Univariate analyses showed that although chemical cues from predators increased tail length and accelerated development, no significant population × treatment effects were observed for these responses (Fig. 1 and Table 1). Thus, both populations responded in a similar way to the predator threat.

Figure 1.

Comparison of induced defences to predators in toad tadpoles collected from founder and descendent (= reintroduced) populations on Mallorca. Open squares, founder population; filled squares, descendent population. Error bars represent standard errors.

Table 1. ancova results for the founder vs. descendent population comparison. Data show F-ratios for traits measured when tadpoles were at Cambar & Martin (1959) stages six to seven (day 15). mancova for later stages (day 30) showed no predator effect: F4,28 = 1·00, P= 0·425. Covariate 1: starting size was used in all ancovas apart from development rate (no covariate 1). *P < 0·05, **P < 0·01, ***P < 0·001. Italic treatment rankings represent significant differences. Rankings are corrected for covariates. C, control; S, snake; F, frog
TraitPopulationPredatorPopulation × predatorCovariate 1 start sizeCovariate 2 heterozygosityPredator rankingR2
Tail length0·499·43**2·580·290·02C < F ≤ S0·607
Lower tail fin depth0·122·210·491·290·32S < C < F0·371
Development0·646·06**1·792·62C < F < S0·509

When short-term and long-term captive populations were compared, chemical cues from predators once again resulted in changes in tail shape (mancova, F2,50 = 5·12, P= 0·009) and also a population × treatment interaction (F2,50 = 4·96, P= 0·011), but no difference between populations (F2,50 = 0·47, P= 0·628). Univariate analyses showed that predators significantly increased tail length, reduced lower tail fin depth and increased development rate, although the latter difference was marginally non-significant (P = 0·051; Fig. 2 and Table 2). More importantly, a population × treatment effect was observed for one trait: lower tail fin. Tadpoles from both populations had shallower lower tail fins with snakes but the short-term captive stock had a stronger response to the predator cue than the long-term stock (Fig. 2 and Table 2). Long-term stock tadpoles also developed more slowly than short-term stock tadpoles (Fig. 2 and Table 2). Heterozygosity did not correlate with any of the response variables in either experiment.

Figure 2.

Comparison of induced defences in long-term (nine to 12 generations) and short-term (one to two generations) captive populations. Open squares, short-term population; filled squares, long-term population. Error bars represent standard errors.

Table 2. ancova results for the long-term vs. short-term population comparison. Data show F-ratios for traits measured when tadpoles were at Cambar & Martin (1959) stages six to eight (day 29). (mancova for day 14 showed no predator effect: F2,53 = 0·37, P= 0·696). For details see Table 1
TraitPopulationPredatorPopulation ×  predatorCovariate 1 start sizeCovariate 2 heterozygosityR2
Tail length0·75 7·85**0·973·490·050·303
Lower tail fin depth0·0811·27***6·09**0·490·040·213
Development rate9·22** 3·990·460·610·334

microsatellite dna analysis

Three of the four populations used in this study (founder, descendent and long-term captive stock) originated from the same site. Heterozygosity did not differ significantly between the founder and the descendent population, but was significantly lower for the long-term captive stock (Table 3). Allelic richness estimates were significantly different between the three populations. The founder population had the highest allelic richness, followed by the descendent population and then the long-term captive stock. The short-term captive stock originated from a different site. Heterozygosity and allelic richness in this population did not differ from the founder population, but were significantly higher than in the long term stock. Heterozygosity was also significantly higher in the short-term captive stock than in the descendent population (Table 3).

Table 3.  Heterozygosity levels and allelic richness in Mallorcan midwife toad populations. Data are based on 10 microsatellite loci for the populations used in this study. Populations with the same superscript were not significantly different. Founder, descendent and long-term captive population have a common ancestry
 Wild populationsCaptive populations
Heterozygosity0·61 ± 0·014ab0·58 ± 0·016b0·67 ± 0·029a0·50 ± 0·026c
Alleles6·15 ± 1·78a5·10 ± 1·66b5·00 ± 1·83ab3·00 ± 1·05c


Our data for the Mallorcan midwife toad provide the first demonstration of how long-term captive breeding and reintroduction can affect natural responses to predators in any endangered species. Reintroduced Mallorcan midwife toads retained anti-predator responses for several generations even in the absence of predator selection pressures. However, the comparison between long-term and short-term captive populations suggests that the responses started to degenerate after nine to 12 generations. Often captive breeding programmes involve species with only a few animals left in the wild (Fredrickson & Hedrick 2002; Britt, Welch & Katz 2003) and are started by fewer individuals than founded the Mallorcan toad programme (i.e. 14 potentially breeding adults). For such species a negative effect of captive breeding may be observed more quickly, as the founders may have already undergone a more severe bottleneck before the start of the programme. Moreover, it is likely that selection is strongest, and evolutionary change fastest, during the earlier generations in captivity (Lewis & Thomas 2001). Indeed, Stockwell & Weeks (1999) observed evidence of rapid evolutionary change after only two generations in captive populations of the mosquitofish Gambusia affinis.

Neutral marker loci may be used to detect inbreeding depression if their heterozygosity levels are positively correlated with fitness components (Slate & Pemberton 2002). Morphological traits are often unrelated to fitness and tend not to correlate positively with microsatellite diversity (Coltman & Slate 2003). In Mallorcan midwife toads, however, larval morphology is clearly a fitness factor, as predator-induced changes in morphology increase survival chances in tadpoles (Van Buskirk & McCollum 2000). Although three to eight generations of captive breeding reduced allelic richness, heterozygosity only declined after up to 12 generations had elapsed. In general, allelic richness is more sensitive to bottlenecks than heterozygosity. In the Atlantic salmon, for example, allelic richness, but not heterozygosity, was significantly reduced after four to five generations of captive breeding (Säisä, Koljonen & Tähtinen 2003).

Amphibians are more threatened, and are apparently declining more rapidly, than mammals or birds (Stuart et al. 2004). Equally, many of the threats they face are complex and often not easily reversed (Beebee & Griffiths 2005). Indeed, about 3·6% of amphibian species are suffering from ‘enigmatic’ declines where the causes are unclear. Stuart et al. (2004) therefore suggest that the only conservation option for such species may be captive breeding. Although their high fecundity, relatively low maintenance costs and short generation times make amphibians attractive options for conservation breeding programmes (Bloxam & Tonge 1994), relative to mammals and birds, threatened amphibians are underrepresented in such breeding programmes (Rahbek 1993; Beck et al. 1994; Balmford, Mace & Leader-Williams 1996). As far as the Mallorcan midwife toad is concerned, it is reassuring that reintroduced populations can retain important anti-predator responses in the absence of selection for several generations. Nevertheless, long-term captive breeding reduced both adaptive and neutral variation in this species.

Conservation management for threatened species may well result in an environment that the species has not experienced in the past (Norris 2004). If so, attempting to maintain adaptations to historical environments may be inappropriate within a species management programme. A long-term reduction in certain fitness attributes and genetic variation may therefore be compensated for by (i) managing the threats associated with the fitness attributes of concern; and (ii) ensuring sufficient genetic variation is maintained to allow adaptation to future environments.


This work was carried out with the co-operation of La Consellaria de Medi Ambient and Associació per la Recuperació del Ferreret. We thank J. Mayol, J. Olivier, D. Jay, C. O’Brien, V. Muñoz, A. Román, C. Zayas and R. Barber for logistical support in the field; Jersey and Chester Zoos for supplying captive bred animals; and D. Church, J. Groombridge, L. Schley, R. Smith and an anonymous referee for comments on the manuscript. The work was funded by the Natural Environment Research Council.