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

NOAA-NMFS-NWFSC TM-33: Sockeye Salmon Status Review (cont)
Back to Previous Section


Previous Genetic Studies

Early studies of sockeye salmon population genetics examined variation at two highly polymorphic loci coding the enzymes lactate dehyrogenase (Hodgins et al. 1969, Withler 1985) and phosphoglucomutase (Utter and Hodgins 1970) or both (Altukhov et al. 1975, 1983; Kirpichnikov and Ivanova 1977; Varnavskaya 1984; Varnavskaya et al. 1988). Subsequent studies have gradually incorporated additional polymorphic loci (currently up to about 48 loci) into their analyses of sockeye salmon population structure (Seeb and Wishard 1977; Grant et al. 1980; Altukhov and Varnavskaya 1983; Utter et al. 1984; Wilmot and Burger 1985; Quinn et al. 1987; Wood et al. 1987b, 1988, 1994; Foote et al. 1988; Kirpichnikov et al. 1990; Rutherford et al. 1992, 1994; Guthrie et al. 1994; Varnavskaya et al. 1994a,b; Wood 1995; Hendry et al. 1996, Winans et al. 1996). Some of these studies analyzed allozymic variation among sockeye salmon populations or subpopulations from geographically limited regions in Kamchatka (Altukhov et al. 1975, 1983; Kirpichnikov and Ivanova 1977; Altukhov and Varnavskaya 1983; Varnavskaya et al. 1988, 1994b; Kirpichnikov et al. 1990), Alaska (Grant et al. 1980, Wilmot and Burger 1985, Varnavskaya et al. 1994b), British Columbia (Wood et al 1987b; Rutherford et al. 1992, 1994; Varnavskaya et al. 1994b), and Washington (Seeb and Wishard 1977, Hendry 1995, Hendry et al. 1996). Additional studies have taken a broader approach, comparing allozymic variation in selected sockeye salmon populations from North America (Withler 1985), Kamchatka (Varnavskaya 1984), or from throughout the range of sockeye salmon (Hodgins et al. 1969, Utter and Hodgins 1970). However, these studies relied on variation in only one or two polymorphic loci and thus provided little resolution of population differences.

Utter et al. (1984) surveyed allozymic variation in 16 collections from southeast Alaska through the Columbia River Basin, including Quinault, Okanogan, and Wenatchee stocks, at 12 polymorphic loci and found a moderate degree of population structuring, significant genetic distance between Quinault River and all other samples, and an apparent genetic association between upper Fraser River and Columbia River sockeye salmon. Guthrie et al. (1994) surveyed sockeye salmon allozymic variation at 28 variable loci in populations from southeast Alaska and northern British Columbia and found substantial genetic divergence among populations, with an underlying geographic structure. Although there were several exceptions, sockeye salmon populations in this region generally fell into 3 main geographic clusters: 1) northern mainland, 2) the southern mainland and inside waters, and 3) the southern and central islands (Guthrie et al. 1994).

Wood et al. (1994) surveyed genetic variation at 33 allozyme loci in O. nerka from 83 sample sites throughout British Columbia. A hierarchical gene diversity analysis of sockeye salmon in eight river systems (Fraser, Nass, Skeena, etc.) was conducted using the eight most polymorphic loci. This procedure determines the relative contribution of different components to the overall gene diversity (Chakraborty 1980). Wood et al. (1994) found that variability was partitioned as follows:

Among river systems: 6.3%
among drainages within rivers: 2.9%
among lakes within drainages: 7.0%
among sites within lakes: 1.0%
within sites: 82.8%

Wood et al. (1994) concluded that most of the genetic variation occurred within spawning sites, and the most important level of differentiation among samples was the nursery lake. Variation among river systems that spanned the north-south breadth of British Columbia accounted for only 6.3% of the variation. A neighbor-joining tree of Cavalli-Sforza and Edwards chord distances based on six polymorphic loci revealed three large groups of populations: southern rivers, northern rivers, and Skeena River/coastal populations. The pattern of genetic similarity among populations within these groups was not strongly geographic. Wood et al. (1994) reported that a "mosaic" pattern of variation was also apparent in an "unweighted pair group method with arithmetic averages" (UPGMA) (Sneath and Sokal 1973) dendrogram based on Nei's unbiased genetic distance (Nei 1978). A principal component analysis (PCA) also showed considerable overlap among regional groups of rivers. The most distinctive group was the southern rivers, consisting of samples from the upper Fraser River, the Thompson River, and the Columbia River (from the Okanogan River Basin). Wood et al. (1994, p. 124) concluded that "geographic structuring was far from perfect, and two populations in widely separated river systems sometimes resembled one another genetically more than they resembled populations in their respective watersheds."

In his most recent work, Wood (1995) concluded that on a time scale of human generations, the best way to conserve genetic diversity in sockeye salmon was to preserve populations in as many different lake systems as possible. From a long-term, evolutionary perspective (> 10,000 yr), he felt it was prudent to save the genetically-diverse, large populations of river/sea-type sockeye salmon (present today in glacially-influenced habitats) that are adapted to a wide range of habitats and conditions and which might provide a source for colonization in favorable interglacial periods.

In a genetic survey of O. nerka in the Pacific Northwest, Winans et al. (1996) surveyed variation at 55 loci in 27 samples of sockeye salmon and kokanee from a total of 21 sites in Washington, Idaho, and British Columbia. They reported that sockeye salmon have the lowest level of allozyme variability of any species of Pacific salmon and a high level of interpopulation differentiation at a relatively few polymorphic loci. Using PCA and clustering of Nei's genetic distance (Nei 1978) to study geographic variation, they reported first that there was no clear geographic pattern of differentiation among the populations of sockeye salmon in the area studied, and second that four genetic clusters of kokanee populations can be identified: 1) a Stanley Basin group including Redfish Lake and Alturas Lake, 2) a late-summer spawning group (from several lakes and reservoirs in central Idaho), 3) a late-fall spawning group (from Lake Whatcom and northern Idaho stocks), and 4) an Okanagan Lake-Shuswap River group.

Chapman et al. (1995) reported data for one additional sockeye salmon sample in the Columbia River Basin. They showed that 14 sockeye salmon collected in the Methow River (situated in the Columbia River Basin between the Wenatchee and Okanogan River systems) were genetically more similar at four loci to two White River samples (from the Wenatchee River basin) than to Okanogan River samples. In contrast, Utter (1995) reported that based on sockeye salmon samples collected in 1994 and analyzed for six polymorphic protein-coding (allozyme) loci and three nuclear DNA microsatellite loci, Wenatchee, Methow, and Okanogan (Wells Dam) River samples were not genetically distinct. Utter (1995) proposed that annual genetic monitoring of Okanogan River sockeye salmon and additional sampling of Methow River sockeye salmon be undertaken to determine the cause of apparent temporal fluctuations in allele frequencies in the Okanogan River population.

The lack of a discernible geographic pattern found by Winans et al. (1996) matches similar studies of sockeye salmon populations in British Columbia, Alaska, and Kamchatka (Varnavskaya et al. 1994a, Wood et al. 1994, Wood 1995). These studies generally indicate that the nearest geographic neighbors of sockeye salmon populations are not necessarily the most genetically similar. Whereas other species of Pacific salmon such as chinook, chum, and pink salmon exhibit clear regional patterns of geographic differentiation (Utter et al. 1989, Winans et al. 1994, Shaklee et al. 1991), geographic variability in sockeye salmon generally resembles a mosaic of genetically distinct populations, at least in the studied portions of the species' distribution. The disjunct nature of differentiation among populations of O. nerka may reflect the discontinuous nature of the habitat, the precise degree of homing to natal streams and lakes (perhaps due to the requirement for nursery lake habitat), and the concomitant decrease in gene flow among neighboring populations.

In a study of the structure and origins of sockeye salmon populations in Lake Washington, Seeb and Wishard (1977) detected identical allele frequencies at five loci for Baker Lake and Cedar River sockeye salmon, indicating that Cedar River sockeye salmon were primarily descended from Baker Lake stock. Seeb and Wishard (1977) stated that Big Bear Creek and Lake Washington beach-spawning sockeye salmon were genetically distinct from potential donor stocks and that these stocks represent remnant native anadromous sockeye salmon populations. However, Hendry (1995) and Hendry et al. (1996) could not detect statistically significant allelic differences at seven polymorphic loci between Lake Washington beach spawning and Cedar River sockeye salmon. Hendry (1995) and Hendry et al. (1996) identified two genetically distinct sockeye salmon groups in the Lake Washington Basin: 1) Cedar River, Lake Washington beach spawners, and Issaquah Creek, which showed genetic affinity with Baker Lake sockeye salmon and 2) Big Bear and Cottage Lake Creeks, which showed genetic distinctiveness from other stocks in the basin and from potential donor stocks. Hendry et al. (1996) inferred from these genetic affinities that the first group of sockeye salmon was of Baker Lake lineage and the second group was predominately of native ancestry.

Hershberger et al. (1982) surveyed genetic variation at 37 allozyme loci (only 2 of which were polymorphic) in sockeye salmon from Ozette Lake, Washington. Hershberger et al. (1982) reported that phenotype frequencies of one variable loci, PGM-1*, suggested that two groups (or populations) of sockeye salmon may be present in Ozette Lake, separated by a difference in run-timing.

Several population surveys of DNA-level variation in O. nerka have been completed. Bickham et al. (1995) examined nucleotide sequence variation of the mtDNA cytochrome b gene in four sockeye salmon populations ranging from Kuril Lake in Kamchatka to Lower Shuswap River in the Fraser River. Three haplotypes were identified. The most common haplotype was found in all populations (in 58% of all individuals); the second most common haplotype was found in all samples (30% of all individuals) except the Fraser River samples. The third haplotype was found in the Skeena River system (at 10%) and the Fraser River (at 40%). Haplotypic frequencies in the Fraser River samples were significantly different from those in the three other samples. The Skeena River sample was also different from the Iliamna Lake samples. Despite this statistical heterogeneity, Bickham et al. (1995) concluded that the three northern samples and the Fraser River samples represent two biogeographic groups post-glacially derived from separate refugia (Beringia and Columbia River). They discussed the evidence for the presence of a generalized north-south phylogeographic break for anadromous fish on the west coast of North America.

Beacham et al. (1995) reported levels of variation in nuclear DNA of O. nerka using minisatellite probes. They used seven of the eight samples used by Bickham et al. (1995) (Pierre Creek, Skeena River was excluded) and two west coast Vancouver Island samples, as well as samples from Lake Wenatchee in the Columbia River Basin. Genomic DNA was digested with one of two restriction enzymes and probed with one of three repeat-sequence probes. Bands were scored for four restriction enzyme/probe marker combinations; three of these appeared to reflect single locus variation. Electrophoretic bands were pooled or binned into size classes for statistical analyses. They interpreted their results as did Bickham et al. (1995)--i.e., the Kamchatka and Iliamna Lake samples were different from the other samples. Other interpretations are also possible. In the cluster analysis, the Lake Wenatchee sample was different from all the other southern samples which, considered together, were different from both the Alaskan samples and Kamchatka samples. However, these latter two samples were also dissimilar from one another. Similarly, along the first two PC axes, the southern samples were different from the two northern groups (Kamchatka and Iliamna). Because only 51% of the variance was explained along PC1 and PC2, relationships may be distorted (viz. Lake Wenatchee) and an examination of PC3 might prove useful.

Thorgaard et al. (1995) examined the use of multilocus DNA fingerprinting to discriminate among 14 sockeye salmon and kokanee populations. DNA extracts were pooled among individuals within the populations. Five oligonucleotide probes were used to visualize bands following digestion with a restriction enzyme and electrophoresis. Electrophoretic bands were grouped by size classes, and each band class was scored as present or absent in each population. Dendrograms based on analysis of banding patterns for four of the five probes produced a concordant pattern of relationships. An analysis of all data produced a tree with a grouping of Redfish Lake, Wenatchee, and Okanogan O. nerka that was separate from kokanee of Oregon and Idaho and a sockeye salmon sample from the mid-Fraser River. Trees of relationship based on three of the five DNA probes showed a clustering of kokanee and sockeye salmon from Redfish Lake (the other two probes grouped Redfish Lake sockeye salmon with either Okanogan River or Lake Wenatchee sockeye salmon), while four of the five probes placed sockeye salmon from Okanogan River together with kokanee from a tributary of Okanagan Lake, British Columbia. None of the five DNA probes showed a close relationship between Lake Wenatchee and Okanogan River sockeye salmon. A portion of the data from the above study were presented in Brannon et al. (1994).

New Data

As part of the comprehensive status review of west coast sockeye salmon, NMFS biologists collected new allozyme genetic information for 17 sockeye salmon populations and 1 kokanee population in Washington and combined them with the existing Pacific Northwest sockeye salmon and kokanee data for analyses. Collection locations, dates, life stage sampled, and sample sizes are summarized in Table 3. We included samples from the Babine River in northern British Columbia (sockeye salmon) and Ozette Lake (kokanee) that were distinctive among their respective life history types (Winans et al. 1996). We examined allelic frequencies for 29 variable loci: ADA-1*, ADA-2*, mAH-1,2* (treated as one locus), mAH-3*, sAH*, mAH-4*, mAAT-1*, ALAT*, CK-B*, FH*, PEPA*, PEPC*, mIDHP-1*, mIDHP-2*, sIDHP-1*, sIDHP-2*, LDH-A1* (used observed phenotypic frequencies of an alternate homozygote *86/*86 because *86/*100 was not distinguishable from *100/*100), LDH-B1*, LDH-B2*, LDH-C*, sMDH-A1,2* (treated as one locus), sMDH-B1,2* (treated as one locus), MPI*, PGDH*, PGM-1* (used observed phenotypic frequencies of null allele), PGM-2*, sSOD-1*, TPI-4*, and TPI-3*. Genetic relationships among samples were examined in two ways: with ordination techniques of genetic distance statistics and with Principal Component Analyses (PCA). Nei's unbiased genetic distances (Nei 1978) and Cavalli-Sforza and Edwards chord values (Cavalli-Sforza and Edwards 1967) among the 32 samples were calculated from the 29 variable loci using the computer program BIOSYS (Swofford and Selander 1981). The distance values were illustrated in UPGMA dendrograms (1-dimensional ordination technique) and multidimensional scaling analyses (2-dimensional ordination techniques) using the computer program NTSYS (Rohlf 1993). A minimum-length spanning tree (MST) was superimposed on the 2-dimensional plots. A PCA was performed using NTSYS (Rohlf 1993) with a correlation matrix for a subset of loci with frequencies of less-common alleles greater than 0.05. Our experience is that less frequent alleles do not contribute substantially to discrimination among samples in a PCA (Winans et al. 1994).

The results from both distance measures (Figs. 5, 6, 7, 8) were similar. On a broad geographic scale, both Nei's (Figs. 5 - 6) and Cavalli-Sforza and Edwards' (Figs. 7 - 8) genetic distances indicate that: 1) samples of sockeye salmon from Lake Wenatchee, Redfish Lake, Ozette Lake, and Lake Pleasant are very distinct from other samples; 2) Lake Washington-Cedar River samples are distinct from a Big Bear Creek-Cottage Lake Creek association; 3) riverine-spawning sockeye salmon from the Nooksack, Skagit, and Sauk Rivers (n = 66) cluster together and have an affinity with Babine Lake and Ozette Lake sockeye salmon; 4) Baker River sockeye salmon are associated with a Lake Washington-Cedar River group, to which Quinault Lake is most similar; and 5) Babine Lake sockeye salmon and Ozette Lake kokanee are particularly distinctive. The Cavalli-Sforza and Edwards metric showed an affinity between Okanogan River sockeye salmon and Lower Shuswap River sockeye salmon, although several allele frequencies in the two samples were statistically different from each other (Winans et al. 1996). Although not indicated by Cavalli-Sforza and Edwards genetic distance (which does not adjust for different sample sizes, such as the small sample size for river-spawners), Nei's genetic distance indicated that riverine-spawning sockeye salmon from the Nooksack, Skagit, and Sauk Rivers have a genetic affinity with Big Bear Creek-Cottage Lake Creek sockeye salmon.

Results of the PCA generally paralleled the genetic distance analyses (Fig. 9). For example, in PC1-PC2 space, Lake Wenatchee, Lake Pleasant, Babine Lake, Big Bear Creek-Cottage Lake Creek sockeye salmon, and Ozette Lake kokanee were distinct from one another (Fig. 9A); along the PC1-PC3 axes, Redfish Lake and Quinault Lake sockeye salmon were distinctive (Fig. 9B). Riverine-spawning sockeye salmon from the Nooksack, Skagit, and Sauk Rivers were most closely related genetically to the Babine Lake sample.

Neither the ordination of genetic distances nor the PC analysis revealed a clear geographic pattern of genetic relationships for the sockeye salmon populations studied. For example, sockeye salmon from the Columbia River Basin (Lake Wenatchee, Okanogan River, and Redfish Lake) did not form a coherent genetic group. Likewise the three coastal populations of sockeye salmon in Washington (Ozette Lake, Lake Pleasant, and Quinault Lake) that are geographically closest, were not very similar to each other genetically.

We examined between-year variability in two locales. We found significant temporal variation in the five Lake Wenatchee samples. The log likelihood ratio statistic (G-test) (Sokal and Rohlf 1981) was used to compare allele frequencies of samples taken in different years in the same locale. G-tests were performed for each polymorphic locus, and the results were summed over all loci for an overall G-value and a standardized G-value. Low levels of statistical significance appeared among the 5 Lake Wenatchee samples: of 10 pair-wise comparisons using sum-G tests, 5 were statistically significant. Lake Wenatchee broodyear 1987 accounted for three of the significant comparisons; it had unusually high frequencies of ALAT*95 and ALAT*108 (Winans et al. 1996). On the other hand, there was no significant temporal variability in three samples from Big Bear Creek (P = 0.27) over 12 loci. In other species of Pacific salmon, temporal variation is usually a minor component of overall genetic variability (e.g., chum salmon (Winans et al. 1994) and pink salmon (Shaklee et al. 1991)). We conclude that, in general, temporal variation at a locale was considerably less than between-locale variation.

Substantial differences were seen among the four Ozette Lake samples of sockeye salmon. Only one of six pair-wise comparisons among the four samples was statistically nonsignificant. The two main remaining spawning beaches in Ozette Lake (at Allen's Bay and Olsen's Beach) are on opposite sides of the south end of Ozette Lake, separated by approximately 3.2 km. The 1995 sample of sockeye salmon from Allen's Bay (adults, n = 33) was statistically different at seven loci from the 1995 Olsen's Beach sample (adults, n = 50). Although precise records concerning the origins of the 1994 sockeye salmon collection from Ozette Lake (adults, n = 80) were not kept, this sample clustered with the Olsen's Beach 1995 sample and differed at seven loci from the 1995 Allen's Bay sample. Dlugokenski et al. (1981) suggested within-lake population subdivision may be present in Ozette Lake based on observed differences in peak spawning times between Allen's Bay and Olsen's Beach spawning aggregates. Additional samples of sockeye salmon from Ozette Lake are currently being pursued. The two samples of kokanee from the Ozette Lake Basin were not statistically different from one another (P = 0.10), but they were divergent from all other O. nerka in each analysis (Figs. 5, 6, 7, 8).

We combined available data with information for British Columbia stocks to gain a broader perspective of sockeye salmon variability in the Pacific Northwest. Analyses of Nei's unbiased and Cavalli-Sforza and Edwards chord distances based on nine loci revealed clustering patterns among the British Columbia samples (Figs. 10, 11, 12, 13). Sockeye salmon from Vancouver Island, the Lower Fraser River, and the Babine River system formed distinctive groups and were associated together. Riverine-spawning sockeye salmon from the Nooksack, Skagit, and Sauk Rivers were similar to one sockeye salmon collection from Babine Lake, and Ozette Lake kokanee clustered with one of the Vancouver Island sockeye salmon groups. Several clusters of upper Fraser River sockeye salmon were recognized. However, Thompson River sockeye salmon did not associate closely with one another. Both Nei's unbiased and Cavalli-Sforza and Edwards chord distances revealed that Washington coastal sockeye salmon (Ozette Lake, Lake Pleasant, and Quinault Lake), and Lake Wenatchee, Okanogan Lake, and Redfish Lake sockeye salmon were associated with one of several sockeye salmon clusters of the upper Fraser River (with the exception of Lake Ozette, which was associated with Big Bear Creek samples according to Nei's unbiased genetic distance).

On to Next Section

Table of Contents