U.S. Dept Commerce/NOAA/NMFS/NWFSC/Publications

Genetic Effects of Straying of Non-Native Hatchery Fish into Natural Populations


Michael Lynch

Department of Biology
University of Oregon
Eugene, OR 97403, USA

Fluctuations in population size and gene flow of maladaptive alleles can potentially produce inbreeding depression and outbreeding depression, both of which can reduce the fitness of a wild population.

Mechanisms Causing Genetic Deterioration

Inbreeding depression

This is the exposure of the individuals in a population to the effects of deleterious recessive genes through matings between close relatives. For a given locus, some alleles will confer more fitness on an individual than other alleles. Within the "other" class of alleles are rare deleterious recessive alleles, which when appearing as a homozygous genotype in an individual because of mating between relatives, greatly reduces the fitness of the individuals carrying them. Deleterious alleles arise constantly through mutation, so they are always present in a population at low frequencies. Suppose we have two alleles, A and a, where A is a normal allele and a is a deleterious allele. The homozygous genotype aa of the deleterious allele is rare in a large population, because with random mating the expected frequency of a homozygote is the square of the allelic frequency p2, and for a low-frequency allele this is a small value. AA individuals are the most fit of the three possible genotypes. Aa individuals have the same fitness as AA individuals if the A allele is dominant over the a alleles, or they may have some intermediate level of fitness if the effects of the alleles are more additive. Lastly, aa individuals show some recessive deleterious trait that reduces their fitness.

In a large population where the a allele occurs at a low frequency, the a allele appears chiefly in the heterozygous state Aa, and heterozygous individuals will almost always mate AA individuals. The offspring of an AA X Aa mating will be AA or Aa, and the effects of the recessive deleterious allele are masked. On the other hand, if mating occurs between relatives in which both relatives have a copy of the deleterious allele in the heterozygous state, an Aa X Aa mating, one-fourth of the offspring of the mating are expected to have the deleterious aa genotype. Mating between relatives "unmasks" the effects of recessive deleterious alleles that would otherwise occur only in heterozygous individuals.

So far, we have considered only a single deleterious allele at a single locus. However, extrapolating from lower organisms and plants (Lynch et al. 1995), about 100 deleterious alleles are present in individuals of higher organisms when we look across all genetic loci (see Lynch and Gabriel 1990). The problem is therefore not trivial when all of the loci are considered. Most of these deleterious mutations produce only a small reduction in fitness of about 2%, when the alleles are made homozygous. If all of the loci in an individual are made homozygous through mating between relatives, the reduction in fitness would be on the order of 200%, enough to "kill" the individual two times over. This, essentially, is inbreeding depression.

Outbreeding enhancement

The converse of inbreeding depression is outbreeding enhancement, which is often referred to as hybrid vigor or heterosis. An example of outbreeding enhancement is the use of hybrid strains of corn, which greatly outperform inbred strains. From the standpoint of deleterious recessive genes, hybrid vigor is nothing more than the reverse of inbreeding depression; that is, it is the masking of recessive deleterious alleles by crossing individuals from different populations. Typically, different populations of the same species harbor different recessive deleterious alleles, so hybrid offspring between parents from the two populations will not be homozygous for the same deleterious alleles. The offspring are fitter than either parent because the effects of the deleterious alleles have been masked. If the hybrid offspring are allowed to mate randomly in subsequent generations, the deleterious alleles will segregate out because of the mechanics of Mendelian inheritance and produce individuals homozygous for the same deleterious allele, which will have reduced fitness. But the mean level of fitness in the population will still be higher than the level in either parental population, because the frequency of each deleterious allele has been reduced by mixing.

In summary, consider a hypothetical population in which an individual mates at random with an unrelated individual in the same population. Other individuals may mate with a sibling, a cousin, or other close relative, and as this mating between relatives continues we begin to see the effects of inbreeding depression. The more closely related two mated individuals are, the greater the depression in fitness that is expected to appear in their offspring. On the other hand, matings of unrelated individuals from genetically diverged populations of the same species may produce outbreeding enhancement, if different deleterious mutations have accumulated in the two populations.

If inbreeding depression and outbreeding enhancement were the only genetic mechanisms we had to consider and matings between individuals could be controlled, obviously the best strategy would be always to mate individuals from different populations. However, things are not so simple. So far we have considered only single-locus effects, but typically alleles at different loci interact so that complexes of genes co-evolve in a population, acting harmoniously with one another to produce a high level of fitness. Different isolated populations may evolve different complexes of genes that interact well within a particular population, but poorly when the genes are mixed through cross-population matings. This reduction in fitness in the offspring is called outbreeding depression.

Outbreeding depression

This phenomenon can occur in two ways. One way is by the "swamping" of locally adapted genes in a wild population by straying from, for example, a hatchery population. In this case, adaptive gene complexes in wild populations are simply being displaced by the immigration of genes that are adapted to the hatchery environment or to some other locality. For example, selection in one population might produce a large body size, whereas in another population small body size might be more advantageous. Gene flow between these populations may lead to individuals with intermediate body sizes, which may not be adaptive in either population. A second way outbreeding depression can occur is by the breakdown of biochemical or physiological compatibilities between genes in the different populations. Within local, isolated populations, alleles are selected for their positive, overall effects on the local genetic background. Due to nonadditive gene action, the same genes may have rather different average effects in different genetic backgrounds--hence, the potential evolution of locally coadapted gene complexes. Offspring between parents from two different populations may have phenotypes that are not good for any environment. It is important to keep in mind that these two mechanisms of outbreeding depression can be operating at the same time. However, determining which mechanism is more important in a particular population is very difficult.

Interaction between mechanisms

Figure 1 shows the theoretical effects of outbreeding depression, relative to outbreeding enhancement and inbreeding depression. Both outbreeding depression and outbreeding enhancement may be occurring at the same time in a population receiving immigrants. As individuals in a local population mate with individuals that are genetically more and more different, outbreeding depression builds up because of the mechanisms we just mentioned. But notice that outbreeding enhancement, because of the masking of deleterious recessive alleles, may also be occurring at the same time that outbreeding depression is occurring. If you average these divergent effects, small amounts of outbreeding may lead to an increase in fitness over that in a local, randomly mating population; however, at higher levels of outbreeding, outbreeding depression may exceed the beneficial effects of outbreeding enhancement. One of the key questions is to determine at what genetic distance the detrimental effects of outbreeding depression exceed the beneficial effects of outbreeding enhancement. If populations have not diverged for a long enough time to acquire separate, co-evolved gene complexes, then it is unlikely that outbreeding depression will occur. The degree that outbreeding enhancement occurs is not predictable and must be determined experimentally.

It is also possible for a population to suffer from both outbreeding depression and inbreeding depression at the same time. Suppose we have two populations, a wild and a hatchery population, that are each fixed for two kinds of alleles for each locus because of local inbreeding (Fig. 2 [below]). The wild population has good (A) and bad (A') alleles at locus A, is fixed for a bad allele (B') at locus B, but is fixed for a good allele (D) at locus D. On the other hand, a hatchery population has good (A) and bad (A') alleles at locus A, only good alleles (B) at locus B, but only bad alleles (D') at locus D. Suppose too that alleles A' and D' are particularly deleterious when combined in the same individual. These populations are then mixed and the hybrid population is allowed to go through several generations. Since most wild populations are small, it also undergoes inbreeding over this time. Eventually, the population may become fixed for the bad A' allele at locus A, for the good B at locus B, and for the bad D' allele at locus D. The alleles at the A and D loci therefore produce inbreeding depression. Also notice that alleles from the two different populations have become fixed in the hybrid population so that outbreeding depression has also become fixed. Two forms of genetic depression are piled on top of each other.

Wild population Genotype X of fish
Hatchery population

1 A' B D 1 A B D'
2 A B D 2 A' B D'
3 A' B D 3 A' B D'
4 A B D 4 A B D'
Hybrid population

1 A' B D
2 A' B D
3 A' B D
4 A' B D

Figure 2. Possible outcome of breeding between hatchery fish and wild fish in small populations. The prime mark (') indicates a recessive deleterious allele.

"Mutational meltdown"

Yet another genetic mechanism can lead to problems in wild populations, especially populations of endangered species. We have assumed that the effective sizes of populations that we have discussed are on the order of only a few individuals or a few tens of individuals at most. We know from empirical results from several organisms that deleterious mutations, mild as they may be in their individual effects, appear at a fairly high rate. About one deleterious mutation appears per individual per generation. That means that on average each fish has one deleterious mutation that was not present in either parent.

As we said, the average reduction in fitness when one of these mutations is made homozygous is only about 2%. Earlier speakers noted that the amount of random genetic drift is inversely proportional to population size, 1/2Ne. If 1/2Ne is larger than the selection coefficient, the efficiency of selection against new mutations is less than the force of random drift for that population size. The result is that the "noise" of random drift will overwhelm natural selection and the new deleterious alleles will accumulate in the populations as though they were neutral alleles, even though they have deleterious effects on the individuals that carry them. Thus, if the selection coefficient is 2%, the effect will be important in populations with effective sizes of 50, or with adult census sizes of a few hundred fish. A rule of thumb is that, in small populations, new, mildly deleterious mutations will accumulate in the population at a rate that is half the mutation rate at the genomic level. Even in the absence of inbreeding depression and outbreeding depression, this accumulation of deleterious mutations will lead to a reduction in fitness of about 1% each generation. Since the effective sizes of many endangered populations of salmon are on the order of 50 or smaller, this is a major potential source of long-term genetic deterioration.

Empirical Evidence

First of all, virtually every trait that has been examined in a wide variety of species can exhibit inbreeding depression, such as by full-sib matings or by self-fertilization in the case of some plants. Some traits are more susceptible to inbreeding than others, but the fact remains that inbreeding depression occurs in all complex genetic characters. A linear decline in mean fitness with the inbreeding coefficient has been observed in a diverse array of organisms including fruit flies, flour beetles, and many species of mammals (including humans). Because inbreeding depression is linear with the inbreeding coefficient, we can extrapolate to future generations if we know the effects of inbreeding depression in the first few generations of inbreeding.

The second point of particular importance for economically important traits in salmon is that traits most closely related to fitness are the ones that exhibit the most inbreeding depression. Again, this has been observed in numerous species, but the data for fruit flies illustrate this principle very well. Table 1 [below] shows a summary of several studies of fruit flies. For morphological characters, the effects of inbreeding are relatively mild. The greatest changes are observed for primary fitness components, such as reproductive capacity, viability, competitive ability, and so on, and not for characters only remotely related to fitness.

The final point with respect to inbreeding depression is that all the studies presented here were done in the laboratory to ensure that observable results were acquired at the end of the experiment (reviewed in Lynch and Walsh 1997). When parallel studies were done in the laboratory and in the field under natural conditions, the effects of inbreeding were typically much greater under natural conditions. The message here is that the assertions about the negative effects of inbreeding outlined above are conservative.

Evidence for outbreeding depression is much less extensive than evidence for inbreeding depression, but outbreeding depression is nevertheless a general genetic phenomenon. One problem in studying outbreeding depression is the number of generations that may occur before outbreeding depression reveals itself. The effects of outbreeding enhancement due to the masking of deleterious alleles and outbreeding depression due to hybrid breakdown may cancel each other in the first generation after crossing individuals from two populations. So the effects of outbreeding depression may not be apparent for a few generations. A few experiments have been done in which reciprocal transplants have been made between plants separated by as little as tens or hundreds of meters. In a study of plants separated by various distances, progeny of crosses between plants separated by 10-30 meters showed greater fitness than plants separated by smaller or larger distances (Wasser and Price 1989). Many of these studies show that populations are locally adapted and that outbreeding depression occurs between genetically divergent individuals. Comparable studies in animals are rare, but it is likely that similar results occur in animals. Experiments on marine copepods in intertidal pools show that hybrid individuals between populations some tens of kilometers apart show breakdowns in salinity tolerance, prolonged development and so on (Burton 1987, 1990). In another study, clones of the microcrustacean Daphnia in the same lake show hybrid breakdown (Lynch and Deng 1994). The overwhelming evidence is that these genetic effects occur in every group of organisms studied, and although not much research has been done on salmon, there is no reason to believe that the genetics of salmon are any different.

Table 1. Inbreeding depression (I.D.) in laboratory populations of Drosophila. I.D. = 1-(zr/zo), where zo and zr are means of the random mating base, and the completely inbred population (obtained by linear extrapolation), respectively. Results marked with an asterisk were obtained from studies of only one or two chromosomes; in these cases, I.D. for the entire genome was extrapolated by assuming that each chromosome arm constituted 20% of the genome, and that the effects were multiplicative across chromosomes. Negative values imply an increase in character value with inbreeding.

Character I.D. (various studies)

Competitive ability 0.84, 0.97
Egg-to-adult viability 0.57, 0,44, 0.66*, 0.48*, 0.06
Female fertility 0.81, 0.18, 0.35
Female rate of reproduction 0.81, 0.56, 0.96, 0.57
Male mating ability 0.52*, 0.92, 0.76
Male longevity 0.18*
Male fertility 0.00*, 0.22*
Male weight 0.07, 0.10
Female weight -0.10
Abdominal bristle number 0.05, 0.06, 0.00
Sternopleural bristle number -0.01, 0.00
Wing length 0.03, 0.01
Thorax length 0.02

Directions in Salmon Research

A question that is often raised is how to obtain information on the genetic consequences of inbreeding and outbreeding in salmon. Many managers would like to have harder evidence that these are real issues with salmonids. The only way of getting this evidence, however, is by doing experiments with salmon themselves. Demonstrating inbreeding depression is straightforward and is done by monitoring the performance of offspring from full-sib matings, because these matings are genetically the closest possible in a sexually reproducing species. Such experiments, however, represent a substantial investment and may take a decade or so. Since the decline in fitness is approximately linear with the degree of inbreeding, useful extrapolations to small natural populations could be made from the results of these experiments.

Experiments to demonstrate outbreeding depression are also conceptually straightforward, but the work needed to complete the experiments is not trivial. To understand the effects of hatchery straying on wild populations, hatchery and wild fish would be crossed to make first generation hybrids, which would then be released for normal ocean migration. Second generation offspring would be made from returning hybrid individuals, which may represent only a small fraction of those released. The effects of outbreeding depression, however, may not be apparent in these early generations, so the crosses of further generations are required. A hybridization between odd- and even-year pink salmon made with cryopreserved sperm yielded only a small amount of evidence about outbreeding depression after several years of work (Gharrett and Smoker 1991). The bottom line is that any kind of quantitative results would take several years of hard work to generate.

Proceeding without results for salmon

Since the empirical evidence of inbreeding depression and outbreeding depression in salmonids will not be available for some time, what is the best way to proceed? The first concern for any stock, whether it is a hatchery stock or a wild stock, is with its effective population size. One way of looking at this question in an objective manner is to ask how big a population would have to be to make it behave genetically as an infinitely large population. In other words, at what point would a further increase in size fail to increase the level of genetic variability beyond that maintained in an effectively infinite population? To make a population genetically "secure" requires an effective population size of several hundred fish, or a census size of about 1,000 reproductive fish. This number is one or two orders of magnitude larger than many populations of salmon that have dwindled to only a few individuals.

The results from other species and population genetic theory can be used to make recommendations that would reduce the likelihood of outbreeding depression in salmonids. One of the most important precautions would be to minimize the degree of interbreeding between hatchery and wild stocks. The effects of outbreeding depression are not likely to appear for at least a couple of generations after outbreeding occurs. If the progeny of an out-crossed stock appear to be fine in the first few generations, this does not necessarily mean that outbreeding depression will not occur later. After genes have been mixed from two populations, it is then impossible to eradicate the potential difficulties with outbreeding depression. At that point, the only way out is to allow natural selection to sort things out, but how long this might take is unknown.


Relevant data for determining the potential effects of inbreeding depression and outbreeding depression in natural populations of salmonids is not yet available. However, theoretical studies and empirical results for other species show that both inbreeding depression and outbreeding depression can lead to the decline in fitness of natural populations. Both of these effects, however, may take several generations to become apparent. At this point, prevention may be better than waiting to implement corrective management policies until empirical evidence demonstrates these effects in salmon.


Burton, R. S. 1987. Differentiation and integration of the genome in populations of the marine copepod Tigriopus californicus. Evolution 41:504-513.

Burton, R. S. 1990. Hybrid breakdown in developmental time in the copepod Tigriopus californicus. Evolution 44:1814-1822.

Gharrett, A. J., and W. W. Smoker. 1991. Two generations of hybrids between even- and odd-year pink salmon (Oncorhynchus gorbuscha): A test for outbreeding depression? Can. J. Fish. Aquat. Sci. 48:1744-1749.

Lynch, M., J. Conery, and R. BÅrger. 1995. Mutational accumulation and the extinction of small populations. Am. Nat. 146:489-518.

Lynch, M., and H.-W. Deng. 1994. Genetic slippage in response to sex. Am. Nat. 144:242-261.

Lynch, M., and W. Gabriel. 1990. Mutation load and the survival of small populations. Evolution 44:1725-1737.

Lynch, M., and J. B. Walsh. In press. The biology and analysis of quantitative traits. Sinauer, Sunderland, MA.

Wasser, N. M., and M. V. Price. 1989. Optimal outcrossing in Ipomopsis aggregata: seed set and offspring fitness. Evolution 43:1097-1109.


Question: Ed Crateau: In the experimental hatchery X wild salmon crosses that you mentioned to demonstrate these effects, don't you also need control experiments to show that the hatchery X wild salmon offspring are worse off in either the hatchery or natural environments? These results, however, would not show whether the problem was adaptation to another environment or outbreeding depression.

Answer: Mike Lynch: Yes. One of the big problems is to determine which mechanism is responsible for declines in fitness. To show that outbreeding depression--the breakdown of intrinsic coadaptation--was the mechanism for reduced fitness, a researcher would have to show that fitness was reduced in all environments. With the rapid habitat changes that are occurring, it is not clear which environment will be relevant several years from now. Perhaps, if fitness begins to decline in a wild population because of a breakdown in local adaptation, stopping gene flow from non-native stocks may allow the local population to recover. Such a recovery would still take several generations.

Question: Dolph Schluter: Since the experiments you described take so long, is there any way of predicting the amount of outbreeding depression that might occur in salmonids from the results of studies of other species? Is it possible to use the amount of time the stocks have been separated from each other or the genetic distance between them to make such predictions?

Answer: Mike Lynch: Few studies of outbreeding depression exist, and in these studies, the degree of outbreeding depression has not been correlated with any kind of molecular marker. So it is difficult to make statements about the time of separation or the degree of genetic differentiation in the characters affected by outbreeding from molecular markers. For the flower, Delphinium, outbreeding depression occurred between plants in the same field. Molecular markers could be used simply to monitor changes in genotypic frequencies in the offspring over time, to estimate mortalities in the different populations of fishes. If the different groups identified by the molecular markers show different levels of fitness, then you can be sure something is happening. If there is no differential fitness in these groups in the first few generations, however, you still cannot be sure that there is not a problem.

Question: Audience: If a hatchery stock is only one or two generations away from a local stock, does this change the likelihood of outbreeding depression in hatchery X wild crosses because of hatchery releases?

Answer: Mike Lynch: If hatchery-reared fish are only one or two generations removed from wild populations, outbreeding depression is unlikely to be a problem. If the hatchery is being used to ensure the survival of large numbers of fry, and if the brood stock is continually taken from wild populations, outbreeding depression is unlikely to occur.

Question: Richard Carmichael: I would like to clarify the kind of experiment that would be needed to show outbreeding depression. For example, someone is proposing to enhance a wild population with a non-local stock, and we want to understand if outbreeding depression might occur. We would first need to know how the hatchery stock performed in the natural environment by itself in the absence of wild salmon. Then we would need to know the productivity of the wild population apart from the hatchery stock. Finally, we need to measure the productivity of the hybrid population. Is that correct?

Answer: Mike Lynch: Yes. But in addition you should follow the hybrid population for at least two generations.

Question: Richard Carmichael: Does the size of the wild population affect the outcome? For example, would a very large population of thousands show more outbreeding depression than a small population where outcrossing may cover inbreeding through outbreeding enhancement?

Answer: Mike Lynch: Population size is important. Inbreeding is measured on a 0-1 scale, and the rate of increase in inbreeding is roughly equal to one over twice the effective population size, 1/2Ne. If the effective size of a population is five fish, the rate of increase in inbreeding due to random mating is 1/10 or 10%. Since some of the five fish may be related, the rate of increase in inbreeding may be more. The point is that small populations become inbred very quickly.

It is really difficult to make specific predictions about outbreeding depression. The common observation from line-cross analysis in agronomy and in animal breeding is an increase in productivity traits in the first generation of a cross between two lines, followed by outbreeding depression in crosses between F1 or later hybrids. The explanation is that hybrid vigor in the first generation results from the masking of deleterious recessive alleles in the two lines. In subsequent generations, adapted combinations of alleles break down, and this leads to outbreeding depression. The breakdown can be due to the loss of ecological adaptation or to a loss of the favorable interactions among genes. If, in fact, the migration rates are exceedingly large, on the order of 50-70% as suggested in some of the talks today, then outbreeding depression is probably not occurring. This level of flushing would simply lead to the replacement of the wild population with the hatchery populations.

Comment: Robin Waples: First, the high rates of straying of non-native fish in the Grande Ronde Basin and the Umatilla River precipitated this workshop. However, a more general issue involves a wide range of straying rates and population sizes.

Second, the theoretical treatments of migration and population size are in terms of individuals per generation, whereas fish biologists often state the number of fish returning to spawn each year. So even though 10, 20, or 50 fish may return in 1 year, a whole generation may be 4 or 5 years. The population size per generation is then the number of fish returning per year times the number of years per generation. This number is not quite so small as the returns per year mentioned earlier.

Question: Audience: How reversible is inbreeding after a population grows quickly, say from 100 to 2,000?

Answer: Mike Lynch: This point arises frequently with captive and endangered populations. Some researchers argue that inbreeding and selection could purge a population of its deleterious mutations. If the population survives, it will be better off. This strategy has been used in the captive breeding program of Spekes gazelle, which was started with four individuals. The cost of this procedure is that most lines or populations go extinct, so that in laboratory experiments with mice, for example, only about 5% of the lines survive. Replicate lines cannot usually be established for an endangered population, so that means that a population has only about a 5% chance of surviving an episode of such intense inbreeding. Also keep in mind that even if all deleterious mutations have been purged, they will eventually return to the population, because the per individual mutation rate to deleterious genes is about one per generation. If a previously inbred population grows and then experiences another reduction in population size, inbreeding depression would occur again, because of the accumulation of recessive deleterious mutations.

Question: Audience: After a population experiences inbreeding because of a strong reduction in size and grows again, you are saying that you have lost the diversity contained in the various lines of descent in the population. Is that correct?

Answer: Mike Lynch: Mutation can eventually bring new useful mutations into a population at a good rate as it grows. So if a population declines to a small size and experiences inbreeding, but its numbers recover, the population may recover genetically. This can, however, take several dozens of generations. What is critical is the transient phase when the population is small and demographic extinction is a possibility.

Question: Audience: How arbitrary is the effective population size of 1,000 individuals?

Answer: Mike Lynch: From population genetics theory, an effective population size of between 500 and 1,000 individuals is the point that, for quantitative characters (e.g., morphology), the genetic variation maintained by a balance between input by mutation and loss by genetic drift is about the same as would be expected in an effectively infinite population.

Table of Contents