Genetic Effects of Captive Breeding Cause a Rapid, Cumulative Fitness Decline in the Wild

See allHide authors and affiliations

Science  05 Oct 2007:
Vol. 318, Issue 5847, pp. 100-103
DOI: 10.1126/science.1145621


Captive breeding is used to supplement populations of many species that are declining in the wild. The suitability of and long-term species survival from such programs remain largely untested, however. We measured lifetime reproductive success of the first two generations of steelhead trout that were reared in captivity and bred in the wild after they were released. By reconstructing a three-generation pedigree with microsatellite markers, we show that genetic effects of domestication reduce subsequent reproductive capabilities by ∼40% per captive-reared generation when fish are moved to natural environments. These results suggest that even a few generations of domestication may have negative effects on natural reproduction in the wild and that the repeated use of captive-reared parents to supplement wild populations should be carefully reconsidered.

Captive breeding was originally used as a form of conservation for the most critically endangered species, but is now widely used for the restoration of declining natural populations (13). In theory, captive-reared organisms may accumulate deleterious alleles that could hinder the recovery of natural populations (36). However, the extent to which captive-reared individuals contribute genetically to the restoration of natural populations is not known.

Hatchery programs for enhancing threatened populations of Pacific salmon and steelhead trout (Oncorhynchus spp.) release more than five billion juvenile hatchery fish into the North Pacific every year (7, 8). Although most of these hatchery programs are meant to produce fish for harvest, an increasing number of captive breeding programs are releasing fish to restore declining natural populations (8, 9). Hatchery fish breed in the wild, and many natural populations are affected by hatchery fish. The use of hatchery-reared fish as broodstock (parents of hatchery fish) for many generations has resulted in individuals that contribute less to the gene pool (are less fit), in comparison with wild fish, in natural environments (1012). On the other hand, captive breeding programs that use local wild fish as broodstock are expected to produce hatchery fish having minimal differences in fitness from wild fish. Nevertheless, such captive-reared fish can be genetically distinct from wild fish for a variety of traits (1316). Thus, it is a real concern that these fish will also have low fitness (reproductive success) in natural environments.

A two-generation pedigree of DNA-based parentage analyses of steelhead (Oncorhynchus mykiss) in the Hood River in Oregon (U.S.A.) showed that the first generation of captive-reared fish had natural reproductive success indistinguishable from that of wild fish in two out of three run-years (17). (Each run-year begins when parents arrive at the river to spawn.) This comparison, however, neglected the fact that captive-reared and wild individuals experience different environments as juveniles, which might affect mating behaviors, fecundity, and/or fertility (18). Therefore, it is difficult to disentangle environmental effects from genetic effects of a difference or lack of difference in reproductive success (17).

In this study, we investigated the strength of genetic effects of domestication on the reproductive success of captive-reared individuals in the wild. Confounding environmental effects were avoided by comparing captive-reared individuals with different histories of captive breeding in the previous generation (Fig. 1). We reconstructed a three-generation pedigree of the winter-run steelhead in the Hood River (19) and compared adult-to-adult reproductive success (number of wild-born, adult offspring per parent) of two types of captive-reared fish (designated C): captive-reared fish from two wild-born parents (C[WxW]), and captive-reared fish from a wild-born parent and a first-generation captive-reared parent (C[CxW]). C[CxW] and C[WxW] were born in the same year, reared in the same hatchery without distinction, and released at the same time. Both fish originated from the same local population, so we can also exclude the influence of local origin. The only difference between them is half of the genome. The half genome in C[CxW] was inherited from the captive-reared parent and experienced captivity for two consecutive generations (during the egg-to-juvenile development). The other half in C[CxW] was from the wild parent and experienced captivity for one generation (C[CxW] itself). In contrast, the entire genome of the C[WxW] experienced captivity for one generation. Thus, by comparing C[CxW] with C[WxW], we were able to evaluate the effect of a single extra generation of captive rearing on subsequent reproductive success in the wild, while controlling for the effect of rearing environment (Fig. 1).

Fig. 1.

Distribution of run-years in which captive-reared fish and their wild-born offspring returned. Numbers in a circle represent a run-year of parents (top) and abrood-year of their offspring (bottom). The percentage on each arrow represents the proportion of adults that return in each subsequent year, which differs between captive-reared fish (dotted line) and wild fish (solid line). C[CxW] were iteratively created from wild individuals and the first generation of captive-reared individuals that returned in run-year 1995; subsequent C[CxW] individuals were created from those individuals returning in 1996 and so forth. These first-year C[CxW] fish returned to spawn mostly in run-year 1998, and we estimated their reproductive success by matching them to the wild-born offspring that returned in run-year 2001–2004.

We estimated the reproductive success of 547 C[CxW] and 193 C[WxW] over three run-years (1998–2000) (19). On the basis of the parentage analysis, we assigned 355 wild-born, returning adult offspring to at least one of their C[CxW] or C[WxW] parents (Table 1). Our estimate of relative reproductive success (RRS) with an unbiased method (20) revealed that the overall reproductive success of C[CxW] is only 55% that of C[WxW] (P = 0.009 by one-tailed permutation tests). We also compared the reproductive success of C[CxW] and C[WxW] from single cohorts (i.e., using only 3-year-olds at the time of spawning) (Table 1). In this comparison, environmental differences were eliminated because both types of hatchery fish were born, returned, and spawned in the same environments in the same year. The smaller sample size resulted in lower power, but the overall estimate was very similar to the above result (single-cohort RRS of C[CxW] to C[WxW] = 0.609, P = 0.042).

Table 1.

RRS (relative number of adult offspring per parent) of two types of captive-reared fish, C[C×W] versus C[W×W]. RRS is given as an unbiased estimate (19, 20). P values were calculated by a one-tailed permutation test. Statistical power represents the minimum effect size (displayed as RRS) detectable with 80% and 95% power. [See (19) and footnote of table S1 for details.] When all parents were compared, overall RRS was estimated using weighted geometric means. The P values were calculated on the basis of Fisher's combined probability (19). For single cohorts, only 3-year-old C[C×W] and C[W×W] were compared. *P < 0.05, **P < 0.01

View this table:

In addition to comparing reproductive success between C[WxW] and C[CxW], we also compared the reproductive success of these captive-reared fish to that of wild-born fish (W) returning in the same run-years (1998–2000). Overall RRS of C[WxW] to W was 0.595 and that of C[CxW] to W was 0.310 [both P < 0.001, (table S1)]. Our estimates of RRS for C[WxW] can be compared with those from our previous study of run years 1995–1997 (17) (table S1). Interestingly, the estimate from run years 1998–2000 was significantly lower than the average RRS ∼ 1 estimated from run-years 1995–1997 (17) (Fig. 2A). One possible explanation for this difference is presence of C[CxW] on the spawning grounds in 1998–2000. For example, reproductive interaction between C[CxW] and C[WxW] might reduce the average reproductive success of C[WxW] if C[WxW] tend to mate more with C[CxW] than with W. Another possibility is nonadditive fitness effects such that mating between hatchery fish results in lower fitness than expected. In our data, nonrandom mating was supported by a test of independence [P < 0.001 for all three run-years (table S2)]. However, an excess of observed mating was found between wild parents, not between captive-reared parents. This might indicate both nonrandom mating (WxW and CxC mating preferences) and nonadditive fitness effects (i.e., low fitness of CxC), although analyses of reproductive success between crosses did not show the presence of nonadditive genetic effects {RRS of [C[CxW] × C[WxW]] to [W × C[WxW]] = 1.1 in run-year 2000, P = 0.878 (table S3)}. Over six run-years of data (1995–2000), four of six years showed lower fitness of C[WxW] (overall RRS of C[WxW] to W = 0.848, P < 0.001).

Fig. 2.

(A) Estimated RRS of captive-reared fish relative to wild fish, plotted against generation time in captivity. Each point represents an estimate from a run-year and sex. The point at generation 0 represents wild fish as a control (marked as a cross). Estimates of the RRS of C[WxW] are plotted at generation 1 and C[CxW] at generation 2. Three years of data at generation 1 (open plots) are from (17). (B) Meta-analysis of the RRS of captive-reared versus wild fish plotted against generation time in captivity of other salmonid species. Solid circles are the estimates from our data (weighted geometric means from Fig. 2A). The bar represents 1 SD. The other four points are from two studies on steelhead, one on brown trout, and one on Atlantic salmon (table S4) from (25). The exponential regressions were obtained as y = e–0.375x (correlation coefficient = 0.962), which suggest that fitness in the wild is reduced 37.5% per generation of captive breeding.

One factor we cannot completely exclude in these comparisons is nongenetic grandparental effects, which have been demonstrated in various organisms, including fish (2124). However, known grandparental effects are mostly female-specific (i.e., grandmaternal egg effects). The reproductive success of C[CxW] did not depend on the sex of the captive-reared parent (overall RRS of C[CxW] with a captive-reared mother to C[CxW] with a captive-reared father = 1.009, P = 0.81). Similarly, there were no noticeable maternal effects on the reproductive success when hatchery and wild fish mated in the wild, either in this study or in our previous study [i.e., number of resulting offspring did not depend on which type of fish was the mother (table S1) (17)]. Thus, the grandparental effect is less likely in this case, and the most likely explanation for the fitness decline is a genetic disadvantage of C[CxW] resulting from the half genome exposed to artificial environments for an additional generation.

Our data suggest a sharp decline in reproductive success follows a very short time in captivity (Fig. 2A). We also conducted a meta-analysis to compare our data with those available for four hatchery stocks for which we know the number of generations in hatcheries (19, 25). These data fit very well on an exponentially declining curve (Fig. 2B), despite the fact that the previous data include RRS estimates using different species and methods and that they are subject to confounding environmental effects (19, 25). It shows 37.5% fitness decline per captive-reared generation, suggesting that the fitness decline of captive-reared fish can be remarkably fast. Because any purely environmental effects should not accumulate over time, the continued decline with generations in captivity (Fig. 2) further supports genetic effects as the cause.

The evolutionary mechanism causing the fitness decline remains unknown. We suspect that unintentional domestication selection and relaxation of natural selection, due to artificially modified and well-protected rearing environments for hatchery fish, are probably occurring (SOM text). Considering the mating scheme for C[CxW] and the generation time for the fitness decline, however, inbreeding depression and accumulation of new mutations should not affect these results. Regardless, our data demonstrate how strong the effects can be and how quickly they accumulate. To supplement declining wild populations, therefore, repeat use of captive-reared organisms for reproduction of captive-reared progenies should be carefully reconsidered.

Supporting Online Material


SOM Text

Tables S1 to S5



References and Notes

View Abstract

Navigate This Article