One outstanding feature of salmon is how variable they are, and this variability can take many different forms within and between populations. For example, the size of adult chinook salmon (Oncorhynchus tshawytscha) varies greatly from medium to very large, as in some populations in British Columbia. As another example, the body shape of coho salmon (O. kisutsch) in British Columbia can vary greatly among populations. Variability can also be seen in coloration, behavior, and many other characteristics. A pervasive notion is that this variability is not due to environmental noise, but reflects something that is meaningful to the survival and persistence of a population in a local environment (Ricker 1972, Taylor 1991).

In this presentation, I would like to define local adaptation, outline the requirements for demonstrating adaptation in wild populations, and discuss how local adaptation is studied. I would then like to describe the extent of local adaptation in nature for a variety of traits, and illustrate the extent of temporal and spatial variability in these traits. Next, I will describe the extent of replicate adaptive evolution and discuss the relevance of hatchery straying to adaptation in wild populations. Lastly, I would like to offer several conclusions about the relevance of adaptation in wild populations and the effects of non-native hatchery straying on fitness in wild populations.

First of all, adaptation is a dynamic process--and I want to emphasize the word process --acting within populations to maintain or increase the frequency of traits that enhance the survival or reproductive success of individuals with the trait. The value of an adaptive trait to an individual is measured relative to individuals with other traits. Three criteria must be satisfied to demonstrate that a trait is adaptive:

• The trait must have an additive genetic basis.
  
• Variability in expression of the trait must be associated with variability in survival, reproductive success, or some other component of fitness.
  
• The mechanisms by which natural selection acts on the trait to increase fitness must be identified.
  

These are very stringent criteria. In short, to demonstrate adaptation, one must show that natural selection influences phenotypic variability and that this variability has, at least in part, a genetic basis.

Adaptation is a dynamic process, which in salmon populations has probably not reached a steady-state endpoint of optimal fitness in an environment. Adaptation is dynamic because selection varies between years, and because trade-offs in fitness at different life history stages produce a "tug of war" between various traits at different life history stages. Variability in the direction of selection was illustrated very well by the example Dolph Schluter (this volume) gave of temporal changes in body size in one of Darwin's finches on the Galapagos Islands. Years of high rainfall produced a large crop of small seeds which favored small-bodied birds with small beaks, and years of drought produced fewer, larger seeds that favored large-bodied birds with large beaks. Salmon also experience fluctuations in the directions of selection, not only between years, but between life history stages. For instance, many environmental variables thought to act as selective factors in salmon populations (e.g., water temperature, water flow, pathogens; see Taylor (1991)) fluctuate from year to year and may cause both the intensity and direction of selection to vary.

One way to demonstrate adaptation is by direct experimentation in nature. This requires an estimate of the heritability of the trait or traits being studied, and a demonstration that the fitness of a phenotype is correlated with an environmental parameter. As far as I know, heritability of a trait in a natural salmon population has been measured in only a single study (Smoker et al. 1994). One way of showing the second criterion, that phenotypic variability is associated with variability in fitness, is through reciprocal transplantation experiments. However, reciprocal translocations of salmon and phenotypic correlations with environment have not been combined into single experiments to my knowledge. The result is that no one has directly demonstrated natural selection in wild populations of salmon.

Another way of demonstrating natural selection is to use indirect comparative methods, and most of the evidence for local adaptation in salmon populations comes from this kind of analysis. One approach is to search for environment-phenotype correlations among animals in contrasting environments and to use these correlations to predict how individuals might behave under experimental conditions in which performance can be tested. Another approach is to make inferences from rigorously controlled experimental manipulation.

The following are three examples of the kinds of salmon studies that have been used to demonstrate local adaptation. The first example in which the indirect comparative method was used comes from studies by Taylor and McPhail (1985) and Tsuyuki and Williscroft (1977) on fatigue time during prolonged swimming and freshwater migration distance to natal areas. Figure 1 shows the time to fatigue in coho salmon and steelhead trout for wild fish and for fish raised in "common-garden" experiments in which different populations were raised under the same conditions. Freshwater migration distances for the different populations ranged from 20-30 km to more than 400 km in the Fraser River. What we see is that fish migrating long distances have greater prolonged swimming performance (i.e., longer time to fatigue) than fish spawning at sites close to the ocean. Here the phenotype-environment interaction--a proxy for natural selection--is migration distance.

A second example comes from Atlantic salmon (Salmo salar) for two rivers, each with several tributaries. One river is the Dee River in Scotland, and the other is the Blackwater River in Ireland. The frequency of the sMEP-1*100 allele ("ME-2," malic enzyme) is positively correlated with water temperature in the two distinct watersheds (Verspoor and Jordan 1989). Although a correlation exists, no selective mechanism was suggested in the article to explain how the gene product might interact with temperature to produce the correlation. Local selection may very well be operating, but more work needs to be done on its mechanism to make this a convincing example of adaptation. This locus could also be linked to another trait that is being selected.

A third example is also a phenotype-environment correlation between the direction of migration and water flow in juvenile sockeye salmon. Some sockeye salmon, such as those in the Cedar River, Washington, spawn in the inlet stream of a lake, so that newly emerged fry must swim downstream to reach the lake where they spend their first year of life. Other sockeye salmon, such as those in the Chilco River, B.C., spawn in the outlet stream, so fry must move upstream to reach the nursery lake. Yet other fry, such as those from Weaver Creek, B.C., must first move downstream to the Harrison River, then upstream against the current into Harrison Lake. Based on the localities of spawning areas relative to the nursery lake, Quinn (1985) predicted the direction fry would orient themselves in a magnetic field after being taken from the field and raised in the laboratory. For example, Cedar River fry would be expected to orient themselves to the north so they would swim into Lake Washington. The results of these experiments followed the predictions: Cedar River fry oriented to magnetic north, on average; Chilco River fry oriented in the expected direction to magnetic south; and Weaver Creek fry oriented downstream then upstream in directions that would eventually take them into Harrison Lake.

The evidence for local adaptation in salmonids generated with indirect methods is largely circumstantial, but nevertheless compelling in that similar results appear for the same traits in several different species. For example, local adapatation has been postulated for age and size at maturity, developmental rate, temperature tolerance, disease resistance, some morphological traits, and some allozyme polymorphisms. Some of the best evidence for adaptation comes from demonstrations of increased disease resistance for salmon populations in areas of sympatry with disease pathogens.

Inferences about adaptive traits in salmon have also been made by observing the survival of hatchery fish transplanted into non-native environments. Many of these studies, however, are difficult to interpret because most of the experiments were uncontrolled and unreplicated. One of the better sets of data from this kind of experiment is on the return rate of hatchery coho salmon transplanted into non-native environments, relative to the return rate of hatchery-released fish at the hatchery (Reisenbichler 1988). The results showed a drop in the return rate as the fish were transferred farther and farther from the hatchery. Fish tranferred 700 km showed fewer returns than fish transplanted within the same watershed. The inference is that the ecological and environmental conditions become increasingly different from the hatchery at more distant localities, and fish do not have the locally adapted traits that would promote their survival in the new environments.

The geographic extent of a local adaptation varies considerably. For example, rainbow trout fry from two tributaries of Pennask Lake, an outlet stream and an inlet stream, have different rheotactic behaviors that bring them into a common nursery lake (Kelso et al. 1981). In this case, the scale is only about 2 km. On the other hand, variability in the frequency of the sMEP-1*100 allele among populations of Atlantic salmon across the North Atlantic demonstrates adaptation on a continental scale (Verspoor and Jordan 1989). The frequency of the 100 allele, which is associated with spawning and rearing in warm water, is low in North American populations of Atlantic salmon, which spawn in much colder waters than do European populations, which show a much higher frequency for this allele. In this case, the enzyme variant (or a selected variant at a linked locus) apparently reflects adaptation both on a small geographic scale between tributaries and on a larger scale across the Atlantic Ocean.

In considering temporal scales of adaptation, keep in mind that virtually all Pacific salmon habitats in the northern part of Washington State and in British Columbia were covered with sheets of Pleistocene ice, which started to recede about 15,000 years ago. Therefore, the considerable diversity among Pacific salmon populations in this area has, to a large extent, evolved since that time. This amount of time, therefore, might be considered the upper limit needed for salmon populations to diversify genetically and to adapt to local conditions. In reality, however, adaptations commonly arise much more quickly. For example, local differentiation has apparently developed among populations of New Zealand chinook salmon since they were introduced about 100 years ago. Experiments are now under way to determine if such differentiation reflects adaptation (T. Quinn, School of Fisheries, University of Washington, Seattle, WA 98195. Pers. commun., June 1995) and, if so, then it means that adaptive changes can occur quite rapidly. In the Pacific Northwest, hatchery populations of chum salmon (O. keta) can have altered developmental rates that apparently result from changes in temperature regimes in the hatchery. These genetically based changes took place in about 6 years, or about 3 generations (Lannan 1980).

I would argue that some genetic changes leading to local adaptation can occur in a single generation, not necessarily thousands of generations. Although evidence is lacking for salmon, short-term changes have been documented in other organisms. One example is the rapid change in beak size in Galapagos finches (Geospiza spp.), in which the driving force is the availability of differently sized seeds in different years. Another example is the rapid change in coloration in guppies that occurred in response to changes in visual predation (Endler 1986). Biochemical adaptation has been postulated for malate dehydrogenase in largemouth bass (Micropterus salmoides) in the central United States where water temperatures appear to favor one allele over another. The point is that although the data are lacking, many traits in salmon can most likely respond rapidly to changes in the enviroment.

One of the chief concerns of conservation is to perserve genetically unique population segments of a species. For many species of fish, however, adaptive traits can appear independently in several populations. One example is seasonal migration timing in adult chinook salmon. It is well known that various populations of chinook enter fresh water on their journeys to spawning grounds in spring, summer, or fall. One explanation for the diversity in migration timing might be that one-time mutations produced the different run times in an ancestral population and that the various kinds of fish colonized different areas. If this were true, we might expect all fall-run populations, for example, to be phylogenetically more closely related to one other than to populations with other migration times. When we look, however, at a phylogenetic tree depicting the genetic relationships among the populations of chinook salmon based on biochemical genetic data (Utter et al. 1989), we see that geographic proximity is a more important determinant of genetic relationships among populations than is migration timing. The populations do not cluster on the basis of run timing, but largely on the basis of geography; northern California populations cluster together, southern Oregon populations together, and so on. This clearly implies that some adaptive life history traits have evolved several times at different locations during the course of salmon evolution.

It might be argued that migration timing is not an adaptive trait--the same river can have different runs of the same species. If so, it is difficult to imagine why similar run times have evolved in so many areas independently of one another. Natural selection must be the force promoting the parallel evolution of this trait. Another useful feature of projecting quantitative traits onto phylogenies is that it focuses attention on groups of populations, and places the variability observed among populations into a more general perspective and highlights the range of habitats required to preserve the processes producing quantitative genetic diversity.

The most important parameter in wild populations potentially affected by the straying of non-native fish is local adaptation. The chief problem for biologists is to define the dynamic interactions between gene flow into wild populations and natural selection against "hybrid" individuals. First, selection against non-native fish and hybrids may be frequency dependent; that is, the ratio of non-native to native fish in a system may influence how well non-native fish and their genes do in natural habitats. Take for example, the very successful introduction of non-native chinook salmon into New Zealand waters. In the absence of genetic mixing, these introduced fish adapted very quickly to local habitats. Another example is the successful colonization of some Pacific salmon in the Great Lakes of North America.

Second, when genetic introgression occurs, what levels of gene flow are permissible with different kinds and intensities of natural selection? To begin to answer this question, one needs to have estimates of the strength of selection in natural habitats. Unfortunately, little data exist on the strength of selection, how often selection fluctuates, and on the kinds of selection that occur at the various life history stages. In the absence of such data, it is not possible to use population genetic models to predict accurately what effects different levels of gene flow have on local adaptations and population fitness.

Although indirect and circumstantial, the evidence that local adaptation is pervasive and important in natural populations of salmon is compelling. Observations of local adaptation in several organisms have demonstrated that natural selection results from dynamic processes, and to preserve genetic diversity these processes must remain intact. Thus, research directed at measuring natural selection in wild populations is urgently needed. Although the conceptual framework for designing such experiments is straightforward, the experiments themselves require considerable effort over several generations. Without these kinds of data, however, the effect of gene flow from non-native hatchery fish on wild populations cannot be predicted with any certainty. Controlled, replicated experiments are needed to provide suitable data for understanding the effects of gene flow. Although it is difficult to say which adaptations should be studied, the migratory timing of juveniles and adults would be a good starting point.

Endler, J. A. 1986. Natural selection in the wild. Princeton University Press, Princeton, NJ.

Kelso, B. W., T. G. Northcote, and C. F. Wehrhahn. 1981. Genetic environmental aspects of the response to water current by rainbow trout (Salmo gairdneri) originating from inlet and outlet streams of two lakes. Can. J. Zool. 59:2177-2185.

Lannan, J. E. 1980. Adaptive and behavioral responses to artificial propogation in a stock of chum salmon, Oncorhynchus keta. ÆMD+ITØIn W. C. Neill and D. C. Himsworth (editors), Salmonid ecosystems of the North Pacific, p. 309-313. Oregon State Univ. Press, Corvallis, OR.

Quinn, T. P. 1985. Homing and the evolution of sockeye salmon (Oncorhynchus nerka). Contrib. Mar. Sci. (Suppl.) 27:353-366.

Reisenbichler, R. R. 1988. Relation between distance transferred from natal stream and recovery rate for hatchery coho salmon. N. Am. J. Fish. Manage. 8:172-174.

Ricker, W. E. 1972. Heredity and environmental factors affecting certain salmonid populations. In R. C. Simon and P. A. Larkin (editors), The stock concept in Pacific salmon, p. 19-160. N. R. MacMillan Lectures in Fisheries. Univ. British Columbia, Vancouver, B.C.

Smoker, W. W., A. J. Gharrett, and J. E. Joyce. 1994. Genetic analysis of size in an anadromous population of pink salmon. Can. J. Fish. Aquat. Sci. 51(Suppl. 1):9-15.

Taylor, E. B. 1991. A review of local adaptation in Salmonidae, with particular reference to Atlantic and Pacific salmon. Aquaculture 98:185-207.

Taylor, E. B., and J. D. McPhail. 1985. Variation in burst and prolonged swimming performance among British Columbia populations of coho salmon (Oncorhynchus kisutch). Can. J. Fish. Aquat. Sci. 42:2029-2033.

Tsuyuki, H., and S. N. Williscroft. 1977. Swimming stamina differences between geotypically distinct forms of rainbow trout (Salmo gairdneri) and steelhead trout. J. Fish. Res. Board Can. 34:996-1003.

Utter, F. M., G. B. Milner, G. StÜhl, and D. Teel. 1989. Genetic population structure of chinook salmon, Oncorhynchus tshawytscha, in the Pacific Northwest. Fish. Bull., U.S. 87:239-264.

Verspoor, E., and W. C. Jordan. 1989. Genetic variation at the ME-2 locus in the Atlantic salmon within and between populations: evidence for its selective maintenance. J. Fish Biol. 35 (Suppl. A):205-213.

Question: Audience: In one breath you talk about local adaptation and give compelling examples of why it is important, then in the next you talk about parallel evolution of life history traits. What, then, is wrong with just outplanting fish and letting natural selection sort things out?

Answer: Eric Taylor: To do that one would have to have good evidence that natural selection would in fact sort things out. The evidence from manipulative experiments suggests that perhaps this may not occur in the short term or may not occur at all. If variation is present in a population, and if the parties are willing to wait long enough--thousands of years--then natural selection would sort things out. The final product may have many of the same adaptations of existing populations, but would most likely be very different in many other traits.

Comment: Tom Quinn: Most transplanted populations do not do well. In experiments we have tried, the number of survivors has been so small that natural selection did not have a chance to sort things out.

Question: Robin Waples: We know when the last ice age ended, we know that nearly all of British Columbia was under a sheet of ice, and we know how much diversity we now have. Do we know anything about salmon populations before the last episode of glaciation or during previous glacial episodes over the last 2 million years?

Answer: Eric Taylor: Not much. About the only thing we can infer is that the various species of salmon have been around 10-50 million years. It is difficult to get information on ancestral populations, except indirectly through phylogenetic analysis of existing species with molecular methods or by the examination of fossils.

Table of Contents