The previous section examined evidence for phenotypic and life-history differences between populations or groups of populations that might be used to identify distinct population segments. The genetic basis of many phenotypic and life-history traits, however, is weak or unknown, and it is difficult to infer the amount of reproductive isolation from population differences in these traits. In this section, we consider biochemical and molecular genetic evidence that might be used to define reproductively isolated populations or groups of populations of chinook salmon. We focus on genetic markers that have been shown to follow or are assumed to follow Mendelian inheritance, so that an analysis of the geographical distributions of these markers can reveal historical levels of gene flow and isolation. The bulk of this evidence consists of frequencies of protein variants (allozymes), or of naturally occurring mutations in minisatellite and microsatellite loci (variable numbers of short tandem repeats) and mitochondrial (mt) DNA. Because of high mutation rates in minisatellite and microsatellite loci, and in some sections of mtDNA, the analysis of these loci permits a greater resolution of the effects of more recent population events than does the analysis of allozyme loci, which generally have lower mutation rates. The different temporal perspectives of population structure from these various techniques were considered in our attempts to define distinct population segments. Analyses of populations of chinook salmon have been examined for genetic variability throughout most of the geographical distribution of this species with allozyme electrophoresis, and in some regions with the analysis of mtDNA or microsatellite loci.
Several standard statistical methods have been used to analyze molecular genetic data to test various hypotheses of reproductive isolation. Comparisons between observed genotypic frequencies in a sample with frequencies expected with random mating (Hardy-Weinberg proportions) can be used to infer the breeding structure of a population or to detect population mixing. Contingency-table comparisons of allozyme or microsatellite allele frequencies among population samples with the chi-square statistics or G-statistic have been widely used to detect significant differences between populations. The finding of significant frequency differences between populations may be evidence of reproductive isolation.
Another way of measuring genetic isolation between populations is to calculate genetic distances from allele-frequency estimates. Several genetic distance measures (e.g. Cavalli-Sforza and Edwards 1967, Rogers 1972, Nei 1972, 1978) have been used to study the population genetic structure of chinook salmon. It is unclear, however, which measure is best, or whether there is one measure that is always best. An attractive feature of Rogers' and Cavalli-Sforza and Edwards' distances is that they satisfy the triangle inequality; that is, given three populations (A, B, C), the sum of the distances between A and B and between B and C is always greater than or equal to the distance between A and C. On the other hand, neither of these genetic-distance measures employs a correction for sample size, so distances are biased upward, especially for small sample sizes. In contrast, Nei's (1978) distance (D) is unbiased, but does not always satisfy the triangle inequality. When sample sizes used to estimate allelic frequencies are 50 individuals or more, the difference between Nei's genetic distance (Nei 1972) and Nei's unbiased genetic distance (Nei 1978) is small, but still might be a substantial proportion of D, if D is small. Another consideration is that Nei's and Rogers' distance measures can be affected by different levels of heterozygosity between populations, whereas Cavalli-Sforza and Edwards' measure is not. Discussions of these and other features of genetic distances appear in Nei (1978), Hillis et al. (1996), and Rogers (1991).
Most of the discussion on genetic distances has focused on the merits of the various measures for phylogenetic reconstruction among species and higher taxa. No one has quantitatively evaluated the performances of these distances in assessing the genetic population structures of species like salmon, which typically show relatively small genetic distances between conspecific populations. Since it is unclear which distance measure is "best" in any given application, we analyzed each set of data with Nei's unbiased, Rogers', and Cavalli-Sforza and Edwards' genetic distances to identify results that were robust to the choice of the distance measure. In most cases, the different genetic-distance measures yielded highly correlated results. For simplicity, we report only results based on Cavalli-Sforza and Edwards' distance measure. This measure ranges from 0.0 (identity) to 1.0 (complete dissimilarity).
The degree of reproductive isolation was inferred from an analysis of the pattern of genetic distances between populations. Clustering methods, such as the unweighted pair group method with arithmetic averages (UPGMA; Sneath and Sokal 1963) and the neighbor-joining method (Saitou and Nei 1987), produce hierarchical groupings of genetically similar populations. Multivariate methods, such as multidimensional scaling (MDS; Kruskal 1964) or principal components analysis (PCA) cluster populations in two or three dimensions. When the geographical distribution of genetic variability is continuous and not hierarchical or disjunct, such as in a clinal or reticulate pattern, MDS and PCA more accurately depict relationships among samples than does agglomerative clustering such as the UPGMA (Lessa 1990). Since the latter algorithm compares the genetic distance of an incoming sample to the average genetic distance between samples already in a cluster, the information about the relationship between the incoming sample and the samples already in the cluster is lost. MDS, on the other hand, is a non-metric ordination technique that minimizes the distortion of pairwise genetic distances between samples in n-dimensional space without averaging. Principal component analysis of allelic frequencies can also be used to examine genetic relationships among populations. In the present analyses, the results of a PCA were usually similar to MDS ordinations for a set of data. Reproductive isolation between populations was inferred from a visual examination of these plots, whenever clusters of related populations were consistent with the geographies of the samples in the clusters.
Levels of genetic variability within populations were also considered, because the level of within-population variability may reflect evolutionary or historical differences in population size and migration patterns between populations. Within-population genetic diversity (H) is usually measured by the expected (with random mating) proportion of heterozygous individuals in a population and is averaged over the number of loci examined. Estimates of heterozygosity based on a small number of individuals are usually accurate, as long as a large number of loci (>30 loci) are surveyed for variability (Nei 1978).
Genetic differentiation between populations at various hierarchical levels has been estimated in many studies with a gene diversity analysis (Nei 1973, Charkraborty 1980), which apportions allele-frequency variability among populations into its geographical or temporal components. For example, the proportion of genetic subdivision among populations may be estimated with GST = (HT - HS)/HT, where HS is the average within-population heterozygosity and HT is the total heterozygosity disregarding geographical subdivision. FST is equivalent to GST when there are only two alleles at a locus. Most genetic variability in salmonids occurs as genotypic differences among individuals within a population (Ryman 1983). A smaller proportion of the total variability is due to hierarchical differences between regions, river systems, tributaries and streams within a river system, between years, or between run types. Estimates of GST or FST among natural populations ranges from 0.0 (no genetic differentiation among populations) to about 0.25 (strong differentiation among populations). These statistics facilitate comparisons among groups of populations that may reveal regional differences in gene flow between populations, or the effects of hatchery strays on levels of differentiation between populations.
In the present status review, we first present the results of previous population genetic studies of chinook salmon, then present the results of an analysis of allele-frequency data that constitute an interagency, coast-wide data base. The primary purpose of the review is to present genetic evidence of reproductive isolation between populations or groups of populations. Allele-frequency differentiation among populations and differences in levels of gene diversity constitute the bulk of this evidence.
Gharrett et al. (1987) studied genetic variability among populations of chinook salmon in 13 river drainages in western, south-central, and southeastern Alaska. They examined electrophoretic variability in proteins encoded by 28 loci, 8 of which had at least moderate levels of polymorphism (frequency of the common allele less than 0.90 in at least 1 of the population samples). In most drainages, collections were made at more than one site or in more than one year, or both. Allele-frequency heterogeneity was observed among three areas in the Yukon River drainage, and among lower and upper Stikine River samples. On a larger geographic scale, significant overall heterogeneity was present among tributaries of western, south-central, and southeastern Alaska. A gene diversity analysis showed that 94.1% of the total variability over samples was contained, on average, within the genetically-homogeneous river drainages, 3.3% was due to differences among river drainages within the three regions, and 2.6% was due to differences among regions. A comparison of these results with other studies (Pacific Northwest, Utter et al. 1989; Oregon-California, Bartley and Gall 1990), indicates the amount of genetic differentiation between Alaskan populations may be smaller than that for chinook salmon populations in other regions. A maximum-likelihood cluster analysis of Cavalli-Sforza and Edwards (1967) genetic distances between samples showed that populations in western and south-central Alaska were closely related to one another, but were distinct from southeastern Alaska populations. Samples from southeastern Alaskan populations were genetically intermediate between samples from western and south-central Alaska as well as those from southern British Columbia and Washington.
Utter et al. (1989) examined allozyme variability at 25 polymorphic loci in samples from 86 populations extending from the Skeena River, British Columbia to the Sacramento and San Joaquin Rivers, California. Geographically proximate samples not showing significant allele-frequency differences (P<0.01) were pooled, and this reduced the data set to 65 units for geographical analyses. A PCA of allelic frequencies and cluster analysis of Nei's (1972) genetic distances between samples indicated the existence of nine genetically distinct regional groups of populations (Fig. 17). The first region consisted of populations in the upper Fraser River and tentatively included a single sample from the Babine River, a tributary of the Skeena River. A second region included populations in rivers draining into Georgia Strait in southern British Columbia. Region 3 included populations around Puget Sound, and a fourth group included populations on the west coast of Vancouver Island, along the Strait of Juan de Fuca, and on the coasts of Washington, Oregon, and California. In the Columbia River basin, Region 5 included populations in the lower Columbia River and its tributaries, and Region 6 included populations in rivers above Bonneville Dam, except those in the Snake River, which constituted Region 7. Farther to the south, Region 8 consisted of populations in the Klamath River Basin, and Region 9 included populations in the Sacramento and San Joaquin Rivers.
A gene diversity analysis of the 65 population units in the 9 regions indicated that 87.7% of the total observed variability was contained, on average, within the units. Of the remaining 12.3%, 1.5% was due to differences among the 9 regions, 6.2% was due to differences among or between river drainages within regions, and 4.6% was due to genetic differences among populations within areas. Utter et al. (1989) re-analyzed the same set of allelic frequencies to estimate the gene diversity components due to differences among adult run times (spring, summer, and fall). Allele-frequency differences among populations within the run times accounted for 11.4% of the total variability, whereas only 0.9% of the total variability was due to differences among run times. The authors concluded that neither clustering nor the gene diversity analyses supported the concept that chinook salmon adult run times represented distinct "races" with separate ancestries, but rather that "genetic divergence into temporally distinct units tend[ed] to occur within an area from a common ancestral stock ..." (p. 247).
The genetic survey of Utter et al. (1989) failed to distinguish clearly between Snake River (Region 7) and Klamath River (Region 8) populations of chinook salmon, even though the mouths of these rivers are geographically widely separated, and recent gene flow between them is unlikely. The authors speculated that this similarity was an artifact that would be resolved as more data became available. Subsequently, Utter et al. (1992) added allelic frequencies for 15 additional polymorphic loci to the data of Utter et al. (1989) and included allelic frequencies of Bartley et al. (1992) and Waples et al. (1991b). The re-analysis indicated a clear genetic separation between populations in the Snake and Klamath River Basins.
In a regional study of mitochondrial DNA variability, Wilson et al. (1987) used 14 type II restriction enzymes (enzymes with cleavage sites located within the recognition sequence) to survey geographical variability in 6 samples from wild and hatchery populations of chinook salmon extending from Bristol Bay, Alaska to southern British Columbia. Four of the enzymes showed restriction fragment length polymorphisms (RFLPs), and 6 composite haplotypes were found among 76 fish. The most abundant haplotype occurred in 43 of the 55 (79%) fish from southern British Columbia. The second most abundant haplotype (N=20) was shared between Alaskan (N=4) and British Columbian (N=6) samples. A third haplotype was found only in Alaska (N=10). Three additional haplotypes were found in single fish from three different localities. Although the lack of sharing of 5 of 6 haplotypes between Alaska and British Columbia indicated substantial reproductive isolation between these populations, average sequence divergence between haplotypes from Alaska and British Columbia (P=0.43%) was not greater than that between haplotypes within Alaska (P=0.45%) and within British Columbia (P=0.54%). A comparison with the RFLP haplotypes for 10 restriction enzymes that were in common with those of Berg and Ferris (1984) in a study of chinook salmon in California indicated a sequence divergence of 2.2%, a value as large as the sequence divergence between chinook salmon and coho salmon reported by Thomas et al. (1986).
Beacham et al. (1989) examined genetic variability at 20 allozyme loci among samples from 15 populations of chinook salmon in the Canadian Yukon River system, and one sample from the Alsek River drainage. Chinook salmon returning to natal spawning sites in the upper reaches of the Yukon River in Canada must travel at least 1,200 km. Tests for allele-frequency heterogeneity at 16 polymorphic loci showed a highly significant difference between the Yukon River samples and the sample from the Alsek River system. Although the headwaters of these two river systems are in close proximity, the Yukon River flows into the Bering Sea and the Alsek River flows into the Gulf of Alaska several hundreds of kilometers away. Among the upper Yukon River samples, the samples from Whitehorse and Takhini Rivers were genetically distinct from the other samples. The rest of the Yukon River samples were not clustered into clear geographical groups. These results show that many of the geographically isolated populations in major tributaries of the upper Yukon River are also genetically distinct from one another.
In another study, Beacham et al. (1996) surveyed variability at three minisatellite loci among populations of chinook salmon extending from the Nass River in northern British Columbia, through the mainland to the Fraser River, and to eastern and western Vancouver Island. Minisatellite loci are segments of DNA consisting of tandomly repeated sequences 10-75 base pairs in length, and alleles consist of different numbers of these repeats. Alleles detected with one probe, pSsa-A34, were previously shown to follow Mendelian inheritance (Stevens et al. 1993). Band counts were binned into size classes, because it was not always possible to establish the homologies of electrophoretically similar fragments. The frequencies of these size classes were used to assess population genetic structure in the same way allozyme alleles were used to test for Hardy-Weinberg proportions or reproductive isolation among populations. Beacham et al. (1996) found strong frequency differences between northern and southern populations of chinook salmon in British Columbia, and also between Fraser River, West Vancouver Island, and East Vancouver Island populations. A neighbor-joining tree of Mahalanobis generalized distances between samples showed two major clusters consisting of samples from northern British Columbia and those from southern British Columbia and Vancouver Island. A PCA analysis, however, indicated a major genetic discontinuity between mainland populations and populations on Vancouver Island. In the PCA, samples of mainland populations fell into a linear array reflecting isolation by distance, a feature of population genetic structure that was not apparent in the neighbor-joining tree. The genetic distinction of southern mainland populations of chinook salmon (excluding the Fraser River) and eastern Vancouver Island populations was not previously detected by the analysis of allozyme variability (Utter et al. 1989).
In a study of chinook salmon in southwestern British Columbia, Heath et al. (1995), examined variability among seven populations on the eastern side of Vancouver Island and two populations in the Fraser River with the analysis of a single-locus minisatellite gene with the probe OtSL1. Alleles with similar allelic mobilities after electrophoresis were binned and the frequencies of the binned classes were analyzed with a PCA. The principal components were tested for significance with a one-way ANOVA, and significant components were used in a discriminant function analysis to produce estimates of population differentiation. They found a 52% overall success rate of assigning sampled fish to the locations from which they had been drawn. Populations that had received transplants tended to show the least amount of discrimination, and this was attributed to the homogenizing effects of gene flow from the transfers. These results are consistent with allozyme studies for this area in showing detectable genetic differences between populations over a restricted area. The analysis of minisatellite loci, however, may have more discriminating power than allozymes, because of the higher mutation rate for minisatellite loci.
Reisenbichler and Phelps (1987) examined chinook salmon allozyme variability in four river drainages on the north coast of Washington. Six of the 55 enzyme-encoding loci examined for genetic variability were polymorphic with frequencies of common alleles less than 0.95, and hence were useful for depicting population structure. Juveniles and adults were sampled in the lower portions of rivers, so intra-river variability could not be estimated. The variance in allelic frequencies between brood years 1981 and 1982 at four localities was used as an error term in an ANOVA of arcsine transformed common-allele frequencies. The ANOVA failed to detect significant allele-frequency heterogeneity among the four drainages for the fall-run samples; that is, the amount of allele-frequency variability among drainages along the coast was no greater than variability between years within rivers, on average. The comparison between summer- and fall-run adult chinook salmon in four rivers, however, approached significance (P=0.07). Comparisons between summer-run hatchery and summer-run wild fish, and between fall-run hatchery and fall-run wild fish, were both significant. These results show that in this relatively small area on the Washington coast a greater amount of reproductive isolation appeared between run types than between populations within run types. Significant frequency differences between hatchery and wild populations indicated minimal mixing between these groups of fish in this area.
Marshall et al. (1995) examined allele-frequency variability at 42 loci in 58 chinook salmon populations representing major spawning areas in Washington. They defined two nested levels of population units from the results of UPGMA clustering and multidimensional scaling of Cavalli-Sforza and Edwards' genetic distances between samples. The more inclusive units, major ancestral lineages (MAL), were defined by four clusters: 1) upper Columbia and Snake River (spring run) samples, 2) upper Columbia River (summer- and fall-run "brights"), mid- and lower Columbia River (spring- and fall-run "tules" and "brights"), and Snake River (fall run) samples, 3) Washington coastal and Strait of Juan de Fuca (spring and fall run) samples, and 4) Puget Sound (spring, summer, and fall run) samples. Each of these four groups were further distinguished by characteristic levels of allozyme polymorphism and by shared occurrences of rare or private alleles among populations within the clusters. Finer scale genetic diversity units (GDUs) were designated within each of the four groups by considering life history, ecological, and physiographic information in addition to allelic frequencies and genetic distances between samples.
One of the earliest studies of chinook salmon genetics in the Columbia River was by Kristiansson and McIntyre (1976), who reported allelic frequencies for 4 polymorphic loci in samples from 10 hatcheries, 5 of which were located along the coast and 5 in the lower Columbia River Basin. Significant frequency differences for SOD* were detected between spring- and fall-run samples collected at the Little White Salmon Hatchery on the Columbia River, but not for spring- and fall-run samples from the Trask River Hatchery along the northern coast of Oregon. Significant allele-frequency differences were also found between Columbia River samples as a group and Oregon coastal samples for PGM* and MDH*.
Utter et al. (1982) compared allelic frequencies at 12 polymorphic loci in samples of fall-run chinook salmon from the Priest Rapids Hatchery in the mid-Columbia River and from Ice Harbor Dam on the Snake River. These samples were taken over four years at each locality. Significant allele-frequency differences between populations were detected for 5 loci.
Schreck et al. (1986) examined allele-frequency variability at 18 polymorphic loci to infer genetic relationships among 56 Columbia River Basin chinook salmon populations. A hierarchical cluster analysis of genetic correlations between populations identified two major groups. The first contained spring-run chinook salmon east of the Cascade Mountains and summer-run fish in the Salmon River. Within this group they found three subclusters: 1) wild and hatchery spring-run chinook salmon east of the Cascade Mountains, 2) spring-run chinook salmon in Idaho, and 3) widely scattered groups of spring-run chinook salmon in the White Salmon River Hatchery, the Marion Forks Hatchery, and the Tucannon River. A second major group consisted of spring-run chinook salmon west of the Cascade Crest, summer-run fish in the upper Columbia River, and all fall-run fish. Three subclusters also appeared in this group: 1) spring- and fall-run fish in the Willamette River, 2) spring- and fall-run chinook salmon below Bonneville Dam, and 3) summer- and fall-run chinook salmon in the upper Columbia River. Schreck et al. (1986) also surveyed morphological variability among areas, and these results were reviewed in the Life History section of this status review.
Waples et al. (1991a) examined 21 polymorphic loci in samples from 44 populations of chinook salmon in the Columbia River Basin. A UPGMA tree of Nei's (1978) genetic distances between samples showed three major clusters of Columbia River Basin chinook salmon: 1) Snake River spring- and summer-run chinook salmon, and mid- and upper Columbia River spring-run chinook salmon, 2) Willamette River spring-run chinook salmon, 3) mid- and upper Columbia River fall- and summer-run chinook salmon, Snake River fall-run chinook salmon, and lower Columbia River fall- and spring-run chinook salmon. These results indicate that the timing of chinook salmon returns to natal rivers was not necessarily consistent with genetic subdivisions. For example, summer-run chinook salmon in the Snake River were genetically distinct from summer-run chinook salmon in the mid and upper Columbia River, but still had similar adult run timings. Spring-run populations in the Snake, Willamette and lower, mid, and upper Columbia Rivers were also genetically distinct from each other but had similar run timings. Conversely, some populations with similar run timings, such as lower Columbia River "tule" fall-run fish and upper Columbia River "bright" fall-run fish, were genetically distinct from one another. Juvenile outmigration also differed among some groups with similar adult run timing. For example, summer-run juveniles in the upper Columbia River exhibit ocean-type life-history characteristics, but summer-run chinook salmon in the Snake River migrate exhibit stream-type life-history characteristics.
In a status review of Snake River fall chinook salmon, Waples et al. (1991b) examined genetic relationships among fall-run chinook salmon in the Columbia and Snake Rivers (Group 3 of Waples et al. 1991a) in more detail. A UPGMA cluster analysis of Nei's unbiased genetic distance, based on 21 polymorphic loci, indicated that "bright" fall-run chinook salmon in the upper Columbia River were genetically distinct from those in the Snake River. Populations in the two groups were characterized by allele-frequency differences of about 10-20% at several loci, and these differences remained relatively constant from year to year in the late 1970s and early 1980s. However, allele-frequency shifts from 1985 to 1990 for samples of fall-run chinook salmon at Lyons Ferry Hatchery in the Snake River suggested that mixing with upper Columbia River fish had occurred. This is consistent with reports that stray hatchery fish from the upper Columbia River were inadvertently used as brood stock at the Lyons Ferry Hatchery. Samples of "bright" fall-run chinook salmon from the Deschutes River and the Marion Drain irrigation channel in the Yakima River Basin also appeared in the same cluster with samples of fall-run chinook salmon from the Snake River.
Genetic analysis of oceanic mixed-stock harvests indicated differences in ocean distributions between "bright" and "tule" fall-run chinook salmon from the Columbia River. Utter et al. (1987) estimated allelic frequencies for 17 polymorphic loci in baseline samples from 88 localities extending from southern British Columbia (except 1 sample from northern British Columbia) through Washington and Oregon to northern California. These data were pooled on the basis of contingency-table tests of allelic frequencies into 65 groups with genetically homogeneous populations. These groups were used to estimate the stock composition of fishery samples taken at ports of landing from the mouth of the Strait of Juan de Fuca to northern Oregon. Tagging returns (Table 5 in Utter et al. 1987) indicated that "tule" fish tended to be caught in the coastal waters of Washington, whereas "upriver brights" tended to be caught in the commercial harvests of Alaska and British Columbia. The results of the mixed-stock analysis for samples collected in 1982 and 1983 were consistent with tagging returns in indicating different ocean distributions of "tule" and upriver "bright" Columbia River chinook salmon.
In a study of genetic effects of hatchery supplementation on naturally spawning populations in the upper Snake River Basin, Waples et al. (1993) examined allele-frequency variability at 35 polymorphic loci in 14 wild (no hatchery supplementation), naturally spawning (some hatchery supplementation), and hatchery populations of spring- and summer-run chinook salmon. Most populations were sampled over two years. An analysis of these data indicated that 96.6% of the genetic diversity existed as genetic differences among individuals within populations. Most of the remaining 3.4% was due to differences between localities, and only a negligible amount was due to allele-frequency differences between spring- and summer-run chinook salmon. Results reveal a close genetic affinity in the upper Snake River between natural spawners that suggests either gene flow between populations or a recent common ancestry. Comparisons between hatchery and natural populations in the same river indicated that the degree of genetic similarity between them reflected the source of the brood stock in the hatchery. As expected, the genetic similarity between wild and hatchery fish, for which local wild fish were used as brood stock, was high.
In a study of upper Columbia River chinook salmon, Utter et al. (1995) examined allele-frequency variability at 36 loci in samples of 16 populations. A UPGMA tree of Nei's (1972) genetic distances between samples indicated that spring-run populations were distinct from summer- and fall-run populations. The average genetic distance between samples from the two groups was about eight times the average of genetic distances between samples within each group. Allele-frequency variability among spring-run populations was considerably greater than that among summer- and fall-run populations in the upper Columbia River. The lack of strong allele-frequency differentiation between summer- and fall-run samples indicated minimal reproductive isolation between these two groups of fish. Hatchery populations of spring-run chinook salmon were genetically distinct from wild spring-run populations, but hatchery populations of fall-run chinook salmon were not genetically distinct from wild fall-run populations.
Some studies have indicated that Snake River spring- and summer-run chinook salmon have reduced levels of genetic variability. Utter et al. (1989) estimated gene diversities with 25 polymorphic loci for 65 population units and found that gene diversities in the Snake River were lower than those in the Columbia River. Winans (1989) estimated levels of gene diversity with 33 loci for spring-, summer-, and fall-run chinook salmon at 28 localities in the Columbia River Basin. Fall-run chinook salmon tended to have significantly greater levels of gene diversity (N=12, mean H=0.081) than both spring- (N=17, H=0.065) and summer-run (N=3, mean H=0.053) chinook salmon. Spring-run fish in the Snake River had the lowest gene diversities (N=4, mean H=0.044). However, Waples et al. (1991a) found that, with a larger sample of 65 loci, gene diversities in Snake River spring-run and summer-run chinook salmon were not as low as that suggested by earlier studies.
Recent, but unpublished, data are available for chinook salmon and will be discussed in the next section. However the results of the foregoing studies of Columbia and Snake River chinook salmon permit the following generalizations:
Bartley and Gall (1990) surveyed samples from 35 populations in the Sacramento and San Joaquin Rivers and along the coast of northern California for genetic variability at up to 53 loci. Overall, genetic variability was detected at 40% (21) of the loci with the 0.95 criterion of polymorphism, but varied from 3 (5.8%) to 17 (32%) loci among samples. Cluster analysis of Nei's (1978) unbiased genetic distances between samples revealed three major clusters roughly corresponding to 1) the Klamath and Trinity Rivers populations, 2) Eel River populations, and 3) the Sacramento and San Joaquin River populations. Samples from eight coastal populations did not cluster together, but were scattered among samples in the three major clusters. One sample from the Omagar Creek pond-rearing facility in the lower Klamath River drainage did not fall into any of the three major clusters. The average percentage of the total genetic variability contained within samples was 82.3%, and the remainder was due to differences among samples on various geographical scales. The greatest sources of geographical subdivision were among rivers within a drainage (6.1%) and among drainages within a region (5.4%), on average. Differences among samples within rivers (3.3%) and among regions (2.9%) represented smaller amounts of geographical heterogeneity. The authors did not distinguish among adult run times in their analyses.
Bartley et al. (1992) expanded the study of Bartley and Gall (1990) and surveyed up to 78 loci in samples from 37 chinook salmon populations in the Sacramento and San Joaquin Rivers, northern coastal California, the Klamath and Trinity Rivers, and rivers along southern to middle coastal Oregon. The authors detected genetic variation at 47 (60.3%) loci. They found significant departures of genotypic proportions from Hardy-Weinberg proportions in 8% of the samples overall, 5% (13 of 252 tests) in samples from wild populations, but 11% (24 of 210 tests) in samples of hatchery-spawned juveniles. They also found a larger than expected number of departures from Hardy-Weinberg proportions (13%, 13 of 97 tests) in wild and hatchery samples from the Klamath River Basin. In a large number of tests, 5% are expected to be "significant" because of Type I error, but a larger proportion of significant tests may indicate that juveniles with limited numbers of parents had been collected, or that juveniles from genetically distinct subpopulations had been included in a sample, or that the genetic model or interpreting electrophoretic banding patterns was incorrect, or that natural selection was occurring on some genotypes. Allelic frequencies estimated from some of these samples may, therefore, not represent discrete randomly mating populations.
From these data, Bartley et al. (1992) calculated Nei's (1972) genetic distances between populations and produced a UPGMA tree consisting of five clusters, each with a strong geographical component. One cluster included samples from populations in the lower Klamath and Smith Rivers of northern California and the Chetco and Rogue Rivers of southern Oregon, but also included a sample from Rock Creek Hatchery, which is located along the mid-Oregon coast. A second cluster included samples from the Eel River and from coastal rivers of northern California. A third cluster included samples from the upper Klamath and Trinity Rivers. A more distantly related cluster contained samples from the Oregon coast north of the Rogue River. The most distinct cluster included samples from the Sacramento and San Joaquin Rivers, which were not well differentiated from each other. A hierarchical gene diversity analysis, modeled a posteriori after the geographical subdivisions found in the cluster analysis of genetic identities, showed that 89.4% of the total genetic variability observed in the study was contained on average within subpopulations, 7.4% was due to differences among the 5 major groups detected in the UPGMA tree, and 3.2% was due to differences among populations within the groups on average. These results indicate that the major drainages from mid Oregon south each contain genetically distinct populations of chinook salmon.
Yip (1994) examined allozyme variability at 53 enzyme loci in 398 fish collected between September and December 1992 at the Trinity River Hatchery in the Klamath River drainage. About 40 fish returning to the hatchery were sampled each week for 11 weeks during the spawning season. Average heterozygosities in these samples ranged from 0.021 to 0.035 with a mean of 0.029. These low values were similar to the low values in Klamath River populations found by (Utter et al. 1987) and are well below the average of 0.102 for 80 populations of chinook salmon (Utter et al. 1987). The entry timing of spring- and fall-run fish into the Trinity River Hatchery was estimated from fish with coded wire tags in the years 1989-92 and 1994. Based on these returns, the weekly samples for genetic analysis were divided a priori into two groups, weeks 1-4 and weeks 5-11. Tests for allele-frequency differences were made with 5 polymorphic loci. Not all of the fish used in the genetic analysis had coded wire tags, so there may have been a some overlap between spring- and fall-run fish in the middle of the spawning season when they entered the hatchery. The sums of the G-statistics for individual tests were not significant for weekly samples within either group, but were highly significant (P<0.01) for the between-group comparisons. These results were interpreted to indicate that spring- and fall-run chinook salmon returning to the hatchery were genetically different. The analysis of temporal run-time differences was continued in 1994 with allele frequencies for three polymorphic loci, GPI-B2*, sMEP-1*, and PGK-2*. (Yip et al. 1996). As in 1992, comparisons of allele frequencies between dates within the 1994 spring and fall runs were not significant. Comparisons between allele frequencies between 1992 and 1994 for the spring run were not significant, but there was a significant overall difference between 1992 and 1994 fall-run fish. An approximate F ratio, based on the sums of the G-tests for within-group allele-frequency heterogeneity, was used to test whether between-run heterogeneity was greater than temporal differences within runs. This test was significant and was concordant with the conclusions of the earlier study that spring- and fall-run chinook salmon were genetically discrete.
Vilkitis et al. (1994) used RFLP analysis of internal transcribed spacers of ribosomal DNA, and randomly amplified polymorphic DNA (RAPD) to measure the level of divergence between the spring and fall runs at 4 locations in the Salmon River, California. This preliminary study of samples, collected during 1992-93, found distinct genotypes in spring- and fall-run chinook salmon that indicated there were differences between locations, yet did not present any quantitative information on the actual level of divergence.
In tests for between-year differences in allele frequencies at an average of 10 polymorphic loci in samples from hatchery and wild populations in Oregon, Waples and Teel (1990) found a greater number of significant tests between years for hatchery samples than for samples from naturally spawning populations. The greater allele-frequency instability between years in the hatcheries was attributed to the use of an effective number of parents less than 50 in many hatchery propagation programs, even though the numbers of returning adults was much higher.
Populations of chinook salmon in California have also been examined for repeat length polymorphisms at microsatellite loci. Hedgecock et al. (1995) analyzed samples of fall-, late fall-, winter-, and spring-run chinook salmon populations in the Sacramento River for variability at a single locus. Winter-run samples included fish from 1) 1995 brood stock from the Coleman National Fish Hatchery (CNFH), 2) 1995 carcasses from the Sacramento River, and 3) 1991-94 CNFH brood stock. Spring-run fish were sampled at Deer Creek, and fall- and late fall-run fish were sampled from Battle Creek Hatchery stock. The authors concluded that winter-run fish were distinct from spring-, fall- and late fall-run fish but that winter-run brood stock in CNFH may have included a genetic contribution from spring-run fish, not only in 1995, but also in previous years. Banks et al. (Bodega Marine Laboratory, Bodega Bay, CA. Unpublished, 1996.) extended the study of these samples with an analysis of four additional microsatellite loci. A UPGMA tree of Nei's (1978) genetic distance showed that fall- and late fall-run fish were most similar among run types. Even so, a randomized chi-square test (Roff and Bentzen 1989) showed that allele frequencies for 1 of the 5 loci in fall- and late fall-run fish were significantly different. Spring-run fish were the next most closely related to fall- and late fall-run fish, but showed significant allele-frequency differences with fall- or late fall-run fish at 7 of the 10 possible comparisons. Winter-run chinook salmon was a distant outlier to the three other runs, and showed significant allele-frequency differences for 13 of the possible 15 comparisons with the other run types. The average FST over the 5 loci was 0.084 and represents considerable divergence among the run types. These results demonstrate significant levels of reproductive isolation between winter-run fish and the other three run types, and between spring-run fish and fall- and late fall-run fish in the Sacramento River. It is difficult, however, to evaluate the importance of these run-time differences relative to run-time differences in populations elsewhere, because of the lack of a coast-wide data base for these microsatellite loci.
Nielsen (1995) surveyed sequence variability in a 164-base-pair segment of the control region of mtDNA in California Central Valley chinook salmon from 8 rivers, 5 hatcheries, and the Guadalupe Slough. These samples included spring-, fall-, late-fall-, and winter-run fish. Ten haplotypes were defined by 7 nucleotide substitutions: 4 transversions, 2 transitions, and an 81 base-pair insertion. Although the analysis of a single locus should be used cautiously, the relatively large sample sizes in this study provided considerable power to test some hypotheses of population structure. A significant haplotypic frequency difference was found between two successive years for returning adults at one of two hatcheries. None of the tests for haplotype-frequency differences between pairs of wild fall-run samples was significant. However, frequencies in some fall-run wild samples were significantly different from frequencies in samples of fall-run hatchery populations. Haplotypic frequencies in samples from Guadalupe Slough were significantly different from each of the four run types, but were not significantly different from haplotype frequencies at the Feather and Merced River hatcheries. Significant differences appeared between each of the four run types. Nucleotide diversity, the average level of sequence divergence between haplotypes, was small, ranging from 0.001 to 0.009 between run types and averaging 0.004 in the pooled sample. Haplotype diversity (analogous to single-locus heterozygosity) ranged from 0.07 in winter-run chinook salmon to 0.64 in late fall-run chinook salmon, and averaged 0.42 over samples. A gene diversity analysis of haplotypic frequencies indicated that 84.7% of the total variability was contained, on average, within run time and 15.3% was due to differences between run times. This level of differentiation among run types is high, but is similar to differentiation between run types in some other regions based on allozyme frequencies.