Most species consist of groups of individuals that are more or less isolated from one another. Isolating mechanisms may be geographical, temporal, ecological, or ethological, and the degree of isolation varies between species and their component populations. Local populations typically exhibit some degree of genetic differentiation, and the pattern of the distribution of genetic differences between populations is commonly referred to as genetic population structure.

Species differ both in their total amount of genetic variation and in the distribution of this variation between and within populations. In some species, such as several forest trees and marine fishes, populations show little differentiation over large geographic areas; that is, randomly selected individuals from different parts of the geographical range are genetically similar. In other species, for example many amphibians, even nearby populations may display substantial genetic divergence.

For characters that are not strongly influenced by natural selection, the extent of differentiation among subpopulations depends primarily on their size, the time since they were separated, and the degree of isolation between them. Large populations change more slowly than small ones, and even in the case of complete isolation it may take considerable periods of time before substantial genetic differences have accumulated between them. Genetically effective migration, gene flow, typically retards differentiation and prevents extensive divergence.

The strategy for conserving genetic variation within species depends on the species' genetic population structure. The extinction of a particular population may have little effect on the overall genetic resources of a species that exhibits only minor or no divergence among the constituent populations. In contrast, a corresponding extirpation in a highly differentiated species implies the loss of a significant portion of the gene pool of that species.

A considerable number of studies on the genetic population structure of salmonids and other fishes have been published. The population genetic characteristics of salmonids are reasonably well known and have been discussed in several contexts. The present summary on population structure is to a large extent based on material compiled from previous publications, particularly those by Hindar et al. (1991), Utter et al. (1989), and Ryman et al. (1995a,b); the reader is referred to those original papers for a fuller discussion and for references to other articles.

Most of our current knowledge of genetic population structure has been obtained within the past two decades with biochemical genetic techniques, such as protein electrophoresis, and with the direct analysis of DNA. Genetic variation detected with these techniques is typically considered to be neutral, or nearly neutral, to natural selection. The analysis of this kind of genetic variation is generally preferred in evolutionary genetic studies, because the dynamics of selectively neutral genes is reasonably well understood theoretically, whereas variability patterns of genes affected by selection may be more difficult to interpret in an evolutionary context. Thus, selectively neutral genes may provide valuable information on evolutionary relationships among populations, their historical sizes, and the levels of gene flow between them. In contrast, such genes are typically less informative about the adaptive characteristics of a population. We would expect that loci subjected to directional selection exhibit a greater degree of genetic differentiation among populations than is observed at loci detected by various biochemical genetic techniques.

Biologists have long recognized intuitively that salmonid species are subdivided into more or less genetically distinct subunits. The reason for this recognition was generally based on the existence of striking ecological and morphological differences between fish of different origins. A large fraction of the variation of ecological and morphological characters is, however, caused by environmental factors. The extent that observed phenotypic differences reflected genetic divergence was difficult to determine until quite recently. Such evaluations could not be performed before techniques became available which revealed characters that were completely genetically determined and which varied independently of environmental conditions.

Salmonids are represented by both freshwater resident and anadromous species, ranging widely in the waters of the northern hemisphere. A common feature of all salmonids is the existence of genetically distinct local populations. Salmonid populations appear to maintain their genetic integrity through a remarkably accurate homing behavior.

In some species, such as brown trout (Salmo trutta) and cutthroat trout (Oncorhynchus clarki), genetically distinct populations can be found over short geographical distances. As an example from my own experience in Scandinavia, we may consider the brown trout community of Lake Bunnersjîarna in central Sweden. Lake Bunnersjîarna consists of two small basins connected by a channel that permits fish to migrate freely between the two segments of the lake. Physically, the lake is not unusual, but we found that it harbors two genetically distinct populations of brown trout that differ at several protein coding loci, and that are markedly different in adult size. There is no apparent gene flow between the two populations, which seem to be completely reproductively isolated from each other, even though fish from both populations coexist in both lakes and can be caught in the same gill nets. Several additional, and genetically distinct, populations of brown trout occur in the same area only a few kilometers away, and finding similar levels of genetic differentiation over short geographical distances is not unusual for Scandinavia.

The genetic data that have accumulated for the salmonid species show that their population structure is more complex than was previously acknowledged. Several studies have also demonstrated that earlier concepts of genetic structure and evolutionary relationships should be modified. As exemplified in Table 1 [below], such modifications have been justified for several species and have included populations distributed over a wide range of geographical distances. In particular, several evolutionary relationships have been revealed that were previously not recognized, and traditional classifications based on morphology or behavioral characteristics, such as time of spawning or residency vs. anadromy, have proven unreliable indicators of common ancestry.

A convenient method to quantify the amount of genetic differentiation within a species is the so-called gene diversity analysis. Here, the total gene diversity (H_T) estimates the total genetic variability within and among the populations sampled. H_T is the sum of the components representing the gene diversity within subpopulations (H_S), and the gene diversity due to differences among subpopulations (D_ST), such that H_T = H_S + D_ST. The quantities H_S and D_ST therefore provide a representation of the amount of differentiation among the populations sampled, and the coefficient G_ST, defined as

can be used as a measure of the proportion of the total genetic variation that is due to differences among populations. G_ST can take values between zero and unity, G_ST = 0, indicating that all populations have identical gene frequencies, and G_ST = 1, that the populations are as different as they can be (all variation is due to differences between populations). Of course, a fairly large number of loci is necessary to provide an accurate picture of the average variability pattern among the populations of a species. It should also be noted that G_ST in many cases is equivalent to the quantity F_ST that was originally defined by Sewall Wright for describing variation between populations at a single locus.

Table 1. Examples of biochemical genetic studies identifying new groups or modifying previous assumptions of the genetic population structure of salmonid fishes (modified and expanded from Allendorf et al. 1987, Table 1.4).

The quantity G_ST has been estimated from various biochemical genetic data sets for several species. For example, G_ST among the three races of man is about 10%; that is, about 10% of the total genetic variability is due to differences between races, and 90% of the total variability is found, on average, within each race. Caution is necessary when comparing the degree of population differentiation across species, because the various studies used different numbers of populations with various degrees of isolation and spatial separation. Similarly, sample sizes of loci and individuals examined per population sampled differed among investigations. Nevertheless, conspicuous differences appear among species in the extent of differentiation among populations.

Marine fishes, for example Atlantic herring and Atlantic cod, typically show low levels of differentiation among local populations (Fig. 1). In contrast, salmonid species are generally characterized by pronounced genetic heterogeneity between populations. In some salmonids, such as brown trout or Atlantic salmon (S. salar), 30% or more of the total genetic diversity may be due to differences among populations. Although most salmonids show considerable genetic differences among local populations, there are species such as chum salmon (O. keta) for which the differentiation appears to be less pronounced. It is not always clear why some salmonids show less differentiation among populations than others; some species may have a greater rigidity in homing behavior.

Figure 2 shows the percentage of between-population variability (G_ST: shaded bars) and the total variability (H_T: open bars) for several species of salmonids. The first observation is that different species have different overall levels of genetic variability (H_T), and these differences are often difficult to explain. Sometimes genetic variability is associated with the extent of a species' geographic range or with the number of populations examined. This tendency is illustrated in three studies of Atlantic salmon (Fig. 2), where H_T (as well as G_ST) increased as the sampling range increased from northern Sweden to northern Europe to Europe and North America. In other cases, however, as when comparing the four subspecies of cutthroat trout (O. clarki spp.), no obvious relation appears between the amount of genetic variation and the geographic range covered. Second, there is no direct relationship between the amount of genetic variation (H_T) and the extent of heterogeneity among populations (G_ST). As for H_T, G_ST generally increases when more distantly located populations are included in the estimate (e.g., Atlantic salmon), except for cutthroat trout.

Species of Pacific salmon tend to be similar to other salmonids in their overall levels of genetic variability, but they tend to show less variability between populations than other salmonids. For example, chinook salmon (O. tshawytscha) populations of the Pacific Northwest show several population genetic groupings, even though only about 10% of the total variability is due to population differences. Utter et al. (1989) found evidence for nine genetically distinct groups of populations in chinook salmon (Fig. 3). The populations within each of these groups apparently share a common ancestral background that produces the genetic similarity between them. A closer look at the populations of chinook salmon in the south fork of the Salmon River, however, shows that populations with similar run timings (e.g., spring, summer) do not always share a common ancestral background (Fig. 4; Waples et al. 1993). These life history characteristics have apparently arisen independently several times (Table 1 [above]).

In summary, several factors affect the amount and distribution of genetic variation among populations. In general, the larger the geographic distance between populations, the more genetic differentiation they tend to show, most likely because geographic distance enhances reproductive isolation when migration is limited. The presence of physical or geographic barriers to migration may also lead to genetic differentiation between populations because of absent or reduced levels of gene flow. Life history patterns can also influence the degree of genetic differentiation among populations. Anadromous populations of a species tend to have more genetic variation than landlocked or freshwater resident populations, but anadromous populations tend to show less genetic differentiation than do resident populations. However, even if these general tendencies exist, there are many exceptions.

These exceptions generally make it impossible to predict with reasonable precision the reproductive relationships and distribution of genetic variation among a set of salmonid populations. Therefore, direct assessment of the amount and distribution of genetic variation is necessary in any situation where information on the genetic population structure is needed for a management decision.

Allele-frequency differences that are taken to reflect population differentiation are difficult to interpret unless the variability patterns are stable over time. To date, little has been done to describe how quickly population structure can change. The data that do exist for salmonids, however, indicate that observed variability patterns are temporally stable. For example, consider the results of Waples et al. (1993) on chinook salmon from the Snake River that includes samples from 2 consecutive years (1989 and 1990) for all localities studied (Fig. 4). Clearly, samples from different years from the same locality tend to cluster together, as they should if the dendrogram depicts the relationship among samples from a set of local populations that are genetically stable over time. It should be noted, though, that the time span considered is fairly short (1 year), considerably less than a generation for chinook and other species of Pacific salmon.

The largest set of data on temporal variability in a salmonid is apparently for brown trout in central Sweden (Jorde and Ryman 1996). The populations studied were selected to represent a set of natural populations with different degrees of reproductive isolation and are as unaffected as possible by human activities (stocking, pollution, excessive harvest, and so on). Allele-frequency shifts at 14 polymorphic protein loci have been monitored for 15 years, with sample sizes of about 100 fish annually from each of four lakes (Fig. 5).

For populations in these four lakes, about 95% of the total variation was contained, on average, within lakes, and 5% was due to variation between lakes and between years (G_ST = 0.05; Fig. 5). These relative proportions of within and between locality variability are about the same as those frequently observed for populations of Pacific salmon, such as the Snake River chinook populations depicted in Figure 4. The "explained" 5% component of variability among the four brown trout populations can be broken down into two sources of variation: between lakes and between years within lakes. The component corresponding to temporal variability is small and represents only about 0.5% of the total variation observed over the 15 years. This result strongly indicates that the genetic structures of the populations are quite stable over time. In turn, the biological characteristics common to salmonids in general suggest that biochemical genetic data collected at a single time for natural salmonid populations reflect geographical structures that are temporally stable.

The existence of genetically distinct local populations, which are typical for salmonids, indicates that the amount of genetically effective migration (gene flow) between populations is fairly restricted. In the context of straying, it is of considerable interest to have at least a rough idea of the amount of gene flow that is compatible with the level of divergence observed among local salmonid populations (see Ryman et al. 1995a).

A major difficulty in estimating natural levels of gene flow is that we are interested in genetically effective migration, which cannot be estimated by observing the movement of marked individuals. Direct observations of movement from one locality to another may lead to inaccurate estimates of gene flow, because migrants may be reproductively less successful than residents.

A convenient attribute of the quantity G_ST is that it can be related to the amount of gene flow. By assuming 1) that migration roughly follows the so-called island model of migration (see Felsenstein this report), 2) that mutation is not an important force changing allelic frequencies, and 3) that the populations are at pseudo-equilibrium relative to random drift and migration, we can estimate the number of migrants from the approximation

Here, N_e is the genetically effective population size, m the migration rate, and the product N_em is the effective number of migrants per generation. Although some care must be taken in converting G_ST values into estimates of migration, this approach in most cases is expected to provide a reasonably accurate estimate of gene flow, especially if a large number of protein coding loci are used and averaged (Ward et al. 1994).

The relationship between G_ST and the number of migrants per generation (N_em) is plotted in Figure 6. Clearly, only a few migrants are needed to prevent major differentiation among populations. For example, G_ST values of the order of 0.05 and 0.10, which are commonly observed among conspecific salmonid populations, correspond to an average number of migrants per generation of no more than about 5 and 2, respectively. Thus, values of G_ST observed for salmonids suggest that the amount of genetically effective migration or "straying" between natural populations is quite small, especially relative to the levels of straying that occur from hatchery or supplementation populations such as those of the Columbia River Basin (Waples 1991).

Results from our temporal study of brown trout populations indicate that the amount of differentiation observed may to some extent underestimate the direct or indirect gene flow among populations that are geographically distant. This impression stems from, among other things, the observation that the amount of genetic variation found within local populations is larger than would be expected from estimates of their effective sizes (N_e). This finding may be explained if larger groups of populations are connected through some gene flow that is both irregular and restricted. Such an explanation implies that low levels of gene flow between larger groups of local conspecific populations constitute part of an evolutionary strategy; gene flow between neighboring populations is small enough to permit local differentiation, but the flow within the metapopulation is large enough to maintain adequate levels of genetic variability.

The implication for management is that efforts to protect individual natural populations represent only a first necessary step in the conservation of genetic resources of salmonids. In the more ambitious program with the goal of creating opportunities for maintaining reasonable levels of genetic diversity within local populations, conservation efforts should be aimed at preserving systems of populations with the potential for a restricted gene flow between them.

Allendorf, F. W., and R. F. Leary. 1988. Conservation and distribution of genetic variation in a polytypic species, the cutthroat trout. Conserv. Biol. 2:170-184.

Allendorf, F., N. Ryman, and F. Utter. 1987. Genetics and fishery management: Past, present, and future. In N. Ryman, and F. Utter (editors), Population genetics and fishery management, p. 1-19. University of Washington Press, Seattle, WA.

Allendorf, F. W., and F. M. Utter. 1979. Population genetics. In W. S. Hoar, D. J. Randall, and J. R. Brett (editors), Fish physiology, volume 8, p. 407-454. Academic Press, New York.

Busack, C. A., and G. A. E. Gall. 1981. Introgressive hybridization in populations of Paiute cutthroat trout (Salmo clarki seleniris). Can. J. Fish. Aquat. Sci. 38:939-951.

Ferguson, A., and F. M. Mason. 1981. Allozyme evidence for reproductively isolated sympatric populations of grown trout Salmo trutta L. in Lough Melvin, Ireland. J. Fish Biol. 18:629-642.

Garcia-Marin, J. L., P. E. Jorde, N. Ryman, F. Utter, and C. Pla. 1991. Management implications of genetic differentiation between native and hatchery populations of brown trout (Salmo trutta) in Spain. Aquaculture 95:235-249.

Hindar, K., B. Jonsson, N. Ryman, and G. Stahl. 1991. Genetic relationships among landlocked, resident, and anadromous brown trout, Salmo trutta L. Heredity 66:83-91.

Hindar, K., N. Ryman, and F. Utter. 1991. Genetic effects of cultured fish on natural fish populations. Can. J. Fish. Aquat. Sci. 48:945-957.

Jorde, P. E., and N. Ryman. 1996. Demographic genetics of brown trout (Salmo trutta) and estimation of effective population size from temporal change of allele frequencies. Genetics 143:1369-1381.

Kondzela, C. M., C. M. Guthrie, S. L. Hawkins, C. D. Russell, J. D. Helle, and A. J. Gharrett. 1994. Genetic relationships among chum salmon populations in southeast Alaska and northern British Columbia. Can. J. Fish. Aquat. Sci. 51(Suppl. 1):50-64.

Kornfield, I., P. Gagnon, and B. Sidell. 1981. Genetic similarity among endemic arctic char (Salvelinus alpinus) and implications for their management. Can. J. Fish. Aquat. Sci. 38:32-39.

Loudenslager, E., and R. Kitchin. 1979. Genetic similarity of two forms of cutthroat trout, Salmo clarki, in Wyoming. Copeia 4:673-678.

Ryman, N. 1983. Patterns of distribution of biochemical genetic variation in salmonids: differences between species. Aquaculture 33:1-21.

Ryman, N., F. W. Allendorf, and G. Stahl. 1979. Reproductive isolation with little genetic divergence in sympatric populations of brown trout (Salmo trutta). Genetics 92:247-262.

Ryman, N., F. Utter, and K. Hindar. 1995a. Introgression, supportive breeding, and genetic conservation. In:J. D. Ballou, M. Gilpin, and T. J. Foose (editors), Population management for survival and recovery, p. 341-365. Columbia University Press, New York.

Ryman, N., F. Utter, and L. Lairkre. 1995b. Protection of intraspecific biodiversity of exploited fishes. Rev. Fish Biol. Fish. 5:417-466.

Shaklee, J. B., D. C. Klaybor, S. Young, and B. A. White. 1991. Genetic stock structure of odd-year pink salmon, Oncorhynchus gorbuscha (Walbaum), from Washington and British Columbia and potential mixed-stock fisheries applications. J. Fish Biol. 39(Suppl. A):21-34.

Stahl, G. 1987. Genetic population structure of Atlantic salmon. In:N. Ryman, and F. Utter (editors), Population genetics and fishery management, p. 121-140. University of Washington Press, Seattle, WA.

Utter F., G. 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.

Waples, R. S. 1991. Genetic interactions between hatchery and wild salmonids: lessons from the Pacific Northwest. Can. J. Fish. Aquat. Sci. 48(Suppl. 1):124-133.

Waples, R. S., O. W. Johnson, P. B. Aebersold, K. Shiflett, D. M. VanDoornik, D. J. Teel, and A. E. Cook. 1993. A genetic monitoring and evaluation program for supplemented populations of salmon and steelhead in the Snake River Basin. Annual report. Bonneville Power Administration. DE-A179-89BP0091, 179 p. (Available from Northwest Fisherices Science Center, 2725 Montlake Blvd. E, Seattle, WA 98112).

Ward, R. D., M. Woodwark, and D. O. F. Skibinski. 1994. A comparison of genetic diversity levels in marine, freshwater, and anadromous fishes. J. Fish Biol. 44:213-232.

Wishard, L. N., J. E. Seeb, F. M. Utter, and D. Stefan. 1984. A genetic investigation of suspected redband trout populations. Copeia 1984:120-132.

Question: Robin Waples: The F_ST value for the Snake River chinook populations in the dendrogram of Figure 3 is about 0.035, which is fairly typical of Pacific salmon depending on species and geographic area sampled. If you sampled a larger geographic area that included divergent populations, the value of F_ST would be larger. The point is that, at the values of F_ST we find in natural populations, we are at the end of the scale in Figure 6 where a small difference in F_ST makes a large difference in the estimate of N_em, the number of migrants. So considering the errors associated with estimating F_ST, the accuracy in estimating N_em is very low. A point estimate of 10 or 20 migrants could well be 100 in nature. We generally lack the ability to draw inferences about the levels of gene flow that concern managers. Do you have any comments about this limitation on inferences from genetic data?

Answer: Nils Ryman: As you point out, we are limited in our ability to use genetic data, but this approach is the best we have for the moment.

Question: Robin Waples: Another way of expressing this concern is that if you sample a restricted geographic area, you may see fairly modest genetic differences, which are often statistically significant but which are not large. Nevertheless, when you plot that degree of differentiation against geographic distance, you see a strong correlation between genetic differences between populations and the geographic distances between them. Populations located close together tend to be more similar to one another than populations located farther apart. The result is that you have a combination of genetic differences between salmon populations that are typically not large when you compare them to a larger geographic scale.

Answer: Nils Ryman: But then we are back to the point of determining the number of populations that should be targeted for conservation in a critical geographic area. Our data indicate that gene flow occurs between natural populations, and that it is important to conserve the entire grouping rather than a particular population. In this way you escape the problem of estimating gene flow.

Comment: Joe Felsenstein: One reason you may be interested in estimating N_em that are rather large is if you were worried about fitness effects of migration on selected loci, because it may make a great deal of difference whether N_em is 40 or 80. Such values of N_em would not make much of a difference to the amount of geographic differentiation for neutral alleles, but may still be important in how much impact migration from other areas would have on local selected differentiation.

Table of Contents