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

NOAA-NWFSC Tech Memo-24: Status Review of Coho Salmon

Previous Genetic Studies

Since 1982, a variety of genetic studies have found evidence for population structure in coho salmon using allozymes, transferrin, or DNA characters. However, these studies were limited to specific geographical regions and did not examine patterns of genetic relationships on a broader basis. In addition, several other factors may have limited the conclusions that could be drawn from these studies.

First, allozyme studies published prior to 1988 included less than half of the 10 most polymorphic loci recently identified for coho salmon (Milner 1993). These previous studies generally reported a lack of genetic variation and relatively low levels of population subdivision. Second, many of these studies were also limited by small sample sizes. The use of small sample sizes may be a particular problem for coho salmon because of the large number of loci which are variable at low levels (Reisenbichler and Phelps 1987, Bartley et al. 1992).

Finally, several of the studies (Hjort and Schreck 1982, Olin 1984, Solazzi 1986, Bartley et al. 1992) used data for the highly variable transferrin locus. Suzumoto et al. (1977) and Winter et al. (1980) reported differential resistance to bacterial kidney disease among transferrin genotypes. Also, Pratschner (1978) showed differential mortality from vibriosis, cold-water disease, and furunculosis between transferrin genotypes. Because transferrin polymorphisms may be maintained by a selective mechanism and may reflect adaptive properties of different genotypes rather than ancestral relationships, data for this locus are difficult to interpret in terms of population structure.

Bartley et al. (1992) examined the population structure of coho salmon from 27 California populations using 22 variable allozyme loci and the transferrin locus. They reported low levels of variability and little evidence of geographic pattern to the observed genetic variability. However, Bartley et al. (1992) did find significant allele frequency differences among all samples as well as within six regional groupings. They pointed out that the genetic analyses could be greatly improved by increasing the samples sizes, which averaged only 34 fish per sample.

In a study based on 30 variable allozyme loci in addition to transferrin, Olin (1984) found a generally low level of genetic variability in 23 samples from the Oregon coast. He identified four major genetic groupings: 1) the Nehalem River south to the Coos River (just north of Cape Arago); 2) the Nehalem River south to the Alsea River; 3) the Siuslaw River south to Morton Creek (just north of Cape Blanco); and 4) the Rogue and Klamath Rivers. The last group was the most southerly and the most genetically distinctive. Geographic coverage of groups 1 and 2 overlapped and also were the most genetically similar.

Solazzi (1986) examined a dendrogram provided by researchers from the University of California at Davis. This dendrogram was based on some of the data from the allozyme and transferrin studies reported by Olin (1984) and Bartley et al. (1992) and included hatchery and wild samples from the Columbia River (n = 8), the Oregon coast (n = 28), and the California coast (n = 16). The dendrogram featured three major clusters: 1) Oregon coast north of the Rogue River; 2) Columbia, Rogue, and Klamath Rivers, plus two samples from small rivers north of Cape Mendocino; and 3) California samples from south of Cape Mendocino. Eight samples, including three from the Oregon coast and five from California, were outliers and only loosely associated with the major clusters.

Hjort and Schreck (1982) examined population structure using an agglomerative method based on frequencies for one allozyme locus (GPI-B2*) and transferrin, as well as life history and morphological characters. Over half (23) of their samples were from the Oregon coast, with others representing the Washington coast (1), Hood Canal (1), the Columbia River (7), and California (3). Their cluster analysis identified three major groupings: 1) hatchery populations from the north coast of Oregon; 2) Columbia River populations, as well as samples from the Rogue and Klamath Rivers; and 3) coastal Oregon populations. The Oregon coastal group consisted of subclusters of hatchery samples from the south coast, natural samples from the mid coast, and natural samples from the north coast. Three hatchery samples (Quinault from the Washington coast, Quilcene from Hood Canal, and Mad River from the California coast) were outliers in the analysis, being distinct from the three major groups and also from each other.

Reisenbichler and Phelps (1987) used 21 variable allozyme loci and found little geographic structure among seven samples of coho salmon from the northwestern coast of Washington.

Wehrhahn and Powell (1987) surveyed 16 gene loci of low variability for 95 populations from southern British Columbia and reported significant differences between allele frequencies from the lower coastal mainland of British Columbia and those of Oregon reported by Olin (1984). They also reported significant differences between upper Fraser River and lower coastal mainland fish. However, Wehrhahn and Powell (1987) concluded that, based on the distribution of rare alleles, there are no absolute barriers to dispersal and to gene flow in the area we sampled (p. 825).

Two recent reports have studied variation in coho salmon at the DNA level. Currens and Farnsworth (1993) studied mitochondrial DNA variation in 18 Oregon populations. Their analysis identified three major groups: 1) north and central coastal Oregon; 2) the Columbia River, and 3) south coastal Oregon (Rogue and Coquille Rivers), together with two Columbia River populations (the Clatskanie and Clackamas Rivers). Forbes et al. (1993) examined nuclear DNA variation for two growth hormone genes in seven samples from Columbia River and Oregon coastal populations. They reported highly significant differences between Columbia River and Oregon coastal coho salmon but only marginal differences among stocks within these regions.

New Data

NMFS geneticists have collected allozyme data over a 10-year period from over 100 coho salmon samples as part of Genetic Stock Identification (GSI) studies (Milner 1993), a previous ESA status review (Johnson et al. 1991), and for this status review. Sample locations ranged from California to Alaska, with a primary focus on Oregon, Washington, and southern British Columbia (Table 3, Fig. 28). Electrophoretic procedures described by Aebersold et al. (1987) were used to examine up to 87 loci coding for 39 enzymes (Milner 1993). The following 53 loci were variable and used to examine population structure (locus nomenclature follows Shaklee et al. 1990): sAAT-1,2*; sAAT-3*; sAAT-4*; ADA-1*; ADA-2*; mAH-2*; sAH*; ALAT*; CK-A1*; CK-A2*; CK-C1*; CK-C2*; EST-1*; FBALD-3*; FBALD-4*; FH*; bGALA*; GAPDH-2*; GAPDH-3*; GAPDH-4*; bGLUA*; GPI-A*; GPI-B1*; GPI-B2*; GR*; HAGH*; mIDHP-1*; mIDHP-2*; sIDHP-1*; sIDHP-2*; LDH-A1*; LDH-A*2; LDH-B1*; LDH-B2*; LDH-C*; aMAN*; sMDH-A1,2*; sMDH-B1,2*;MPI*; PEPA*; PEPB-1*; PEPC*; PEPD-2*; PEPLT*; PGDH*; PGK-1*; PGK-2*; PGM-1*; PGM-2*; PNP-1*; sSOD-1*; TPI- 1*; TPI-3*.

Regional patterns of allele frequency--Plots of allele frequencies at selected gene loci illustrate the regional differences seen in the allozyme data set (Fig. 29). For example, samples from Puget Sound northward are characterized by lower frequencies of the 100 allele of EST-1* than those from other regions. Columbia River samples are characterized by relatively high frequencies of the 100 allele at PGM-1* and EST- 1*. Samples from Alaska have an unusually low frequency for the 100 allele of PEPA*. Additional regional patterns of allele frequency can be seen in the scatterplots.

Genetic distance--Genetic distances (D) computed for 53 loci between each pair of samples were used to construct the dendrogram shown in Figure 30. We examined both Cavalli-Sforza and Edwards' (1967) chord distance and Nei's (1978) unbiased genetic distance, but only the former is shown. Nei's distance metric includes a correction for sampling error, which can be important if sample sizes are small or if they vary among collections. However, in the present data set the bias correction led to a number of negative D values, which made it difficult to depict genetic relationships on a dendrogram.

Seven major clusters were identified that were largely distinct geographically (Figure 30). Clusters I (separated from other clusters at a D value of 0.088) and II (D = 0.086) are the most distinct major clusters and consist of the four samples from Alaska and the two samples from the mid-Fraser River, respectively. These samples are not discussed further in this report.

Table 3. Samples of coho salmon used in NMFS allozyme analysis. Map codes correspond to those in Figure 28. N is the number of fish in each sample. All fish are juveniles unless otherwise indicated.

Name Source Brood

California 1 Trinity Trinity Hatchery 1982 98
So. Oregon Coast 2 Rogue Illinois River, Greyback Creek 1992 40
3 Rogue Illinois River, Silver Creek 1991 29
4 Rogue Cole Rivers Hatchery, stock #521992 80
5 Elk North Fork Elk and Elk Rivers 1992 32
Oregon Coast6 Sixes Crystal and Edson Creeks 1992 44
7 New Bether and Morton Creeks 1992 62
8 Coquille Butte Falls Hatchery, stock #441992 80
9 Coos Cole Rivers Hatchery, stock #371992 80
10 Coos Millicoma River and Marlow Creek 1992 22
11 Coos South Fork Coos River, Tioga Creek 1992 29
12 Eel Butte Falls Hatchery, stock #631992 80
13 Tenmile Big Creek, Noble Creek, and Tenmile Lake 1991 56
14 Umpqua Rock Creek Hatchery, stock #551992 80
15 Umpqua North Umpqua River, Williams Creek 1992 40
16 Umpqua Butte Falls Hatchery, stock #181992 80
17 Smith Smith River, Halfway Creek 1992 40
18 Tahkenitch Fall Creek Hatchery, stock #113 1992 80
19 Alsea Fall Creek Hatchery, stock #311992 80
20 Alsea Fall Creek Hatchery, stock #431992 80
21 Beaver Beaver Creek 1992 62
22 Siletz Fourth of July, Sunshine, and Buck Creeks 1991 50
23 Siletz Salmon River Hatchery, stock #33 1992 80
24 Salmon Salmon River Hatchery, stock #36 1992 80
25 Salmon Salmon River Hatchery 1982 96
26 Trask Trask River Hatchery, stock #341992 80
27 Trask Trask River Hatchery 1991 120
28 Nehalem Nehalem River Hatchery, stock #32 1992 80
29 Nehalem Nehalem River Hatchery, Fishhawk stock 1982 110
30 Nehalem Nehalem Hatchery, Fishhawk stock 1990 80
Columbia River31 Lewis and Clark Lewis and Clark River 1992 30
32 Grays Grays River Hatchery 1989 40
33 Graysa Grays River Hatchery 1989 40
34 Grays Grays River Hatchery 1982 100
35 Big Creekb Big Creek Hatchery 1989 80
36 Clatskanie Carcus Creek 1989 50
37 Cowlitz Cowlitz Late 1990 100
38 Cowlitz Cowlitz Early 1989 80
39 Cowlitz Cowlitz Late 1989 80
40 Scappoose Siercks, Raymond, and Milton Creeks 1989 44
41 Lewisa Lewis River Hatchery Late 1989 80
42 Lewis Lewis River Hatchery Early 1989 80
43 Clackamas North Fork Clackamas River 1990 90
44 Clackamasa Clackamas and North Fork Clackamas Rivers 1989 60
45 Eagle Eagle Creek Hatchery 1990 100
46 Eaglea Eagle Creek Hatchery 1989 80
47 Sandya Sandy River Hatchery 1989 80
48 Sandy Still Creek 1989 62
49 Sandy Sandy River Hatchery 1990 100
50 Hardyb Hardy Creek 1989 50
51 Bonnevilleb Bonneville Hatchery 198980
52 Willard Willard Hatchery 1989 80

Name Source Brood

53 Naselle Naselle River Hatchery 1990 100
Washington 54 Nemah Nemah River Hatchery 1990 100
Coast 55 Willapa Willapa River Hatchery 1990 100
56 Chehalisc Simpson Hatchery 1989 40
57 Chehalis Satsop River, Bingham Creek and Simpson River Hatcheryc 1982, 1988 140
58 Humptulipsd Humptulips River Hatchery 198840
Olympic Peninsula 59 Queets Clearwater River 1982 95
60 Quillayute Bogachiel River 1985 80
61 Soleduck Bear Creek 1982 95
62 Hoko Hoko River 1985 80
Puget Sound63 Hood Canal Hood Canal Hatchery, Baker stock 1992 80
64 Big Beef Big Beef Creek 1982 80
65 Green Green River Hatchery 1982 100
66 Green Green River Hatchery 1992 80
67 Snohomish Pilchuck River, Little Pilchuck Creek1985 80
68 Snohomish Snoqualmie River, Harris Creek1985 120
69 Snohomish Skykomish River 1982 80
70 Stillaguamish Church Creek 1985 80
71 Stillaguamish North Fork Stillaguamish River, Fortson Creek1985 80
72 Stillaguamish North Fork Stillaguamish River, McGovern Creek 1985 40
73 Stillaguamish South Fork Stillaguamish River, Tiger Creek 1985 80
74 Skagit Upper Skagit River 1991 127
75 Skagit Carpenter Creek 1991 139
76 Skagit West Fork Nookachamps Creek 1991 120
77 Skagit West Fork Nookachamps Creek 1985 100
78 Skagit Baker River 1991 183
79 Skagitc Baker River 1989 120
80 Skagit Upper Sauk River 1991 200
81 Skagit Suiattle River, All Creek 1985 80
82 Skagit Suiattle River, All Creek 1991 120
83 Skagitc Skagit River Hatchery, Baker stock 1989 120
84 Skagitc Skagit River Hatchery, Clark stock 1989 120
85 Skagit Skagit River Hatchery, Clark stock 1991 120
86 Skagit Upper Cascade River 1991 224
87 Skagit Skagit River Hatchery, Baker stock 1991 120
88 Nooksackd Nooksack River Hatchery 1982 80

Name Source Brood

British Columbia 89 Chilliwack Chilliwack River Hatchery 1982 100
90 Coldwater Spius River Hatchery 1985 80
91 Coldwater Spius River Hatchery 1986 80
92 Cowichan Cowichan River Hatchery 1982 80
93 Big Qualicuma Big Qualicum Hatchery 1989 80
94 Big Qualicum Big Qualicum Hatchery 1982 80
95 Robertson Robertson Creek Hatchery 1982 100
96 Capilano Capilano Hatchery 1989 80
97 Squamishc, d Squamish River Hatchery 1985 80
Alaska 98 Cabin Cabin Creek 1990 80
99 Karta Karta River 1990 76
100 Campbell Campbell Creek 1990 80
101 Goodnews Goodnews River 1990 80

aSample is missing data for EST1*. Missing data were filled with allele frequency from other year.
bSample is missing data for EST1*. Missing data were filled with allele frequency of area mean.
cSample is from adult coho salmon.
dSample is missing data for PNP1*. Missing data were filled with allele frequency of area mean.

Clusters III, IV, and V contain Oregon coastal populations. Cluster III is comprised of hatchery populations from northern Oregon (Alsea, Trask, Siletz, and Nehalem) and is the most distinctive cluster (D = 0.074) except for those from the Fraser River and Alaska. Cluster IV (D = 0.071) includes a sample from the Rogue River (Cole Rivers Hatchery) as well as natural and hatchery samples from the Elk and Umpqua Rivers. The Elk and Umpqua Rivers samples constitute a tight subcluster that is more distantly linked to the Rogue River sample. Somewhat less distantly linked are other samples from the Rogue River and the Trinity Hatchery in northern California. Cluster V (D = 0.062) contains wild and hatchery populations ranging from the Rogue River (Illinois River, Greyback Creek) in the south through the Salmon River in the north.

Cluster VI (D = 0.062) includes all of the Puget Sound and British Columbia samples (except the two Fraser River samples noted above and one of two samples from Big Qualicum Hatchery). All of the British Columbia samples except that for Robertson Creek (on the west Coast of Vancouver Island) are from streams draining into the Strait of Georgia. Four samples from the Strait of Juan de Fuca and the northern coast of Washington (Queets, Quillayute, Soleduck, and Hoko Rivers) form a subgroup within the larger Puget Sound/Strait of Georgia group.

Cluster VII (D = 0.068) includes all of the samples from the lower Columbia River, as well as those from the southwestern Washington coast. This cluster also includes a sample from the Rogue River Basin (Illinois River, Silver Creek) and one from Trask Hatchery. Within this cluster are several subclusters and three branches with only one or two members. Two subclusters comprise most of the lower Columbia River samples: one consisting primarily of samples from Washington and the other consisting primarily of samples from Oregon. Another subcluster contains three samples from Willapa Bay. A final subcluster contains a group of samples from the Clackamas and Clatskanie Rivers, together with a group that includes samples from the Humptulips and Simpson hatcheries from southwestern Washington.

Principal component analysis--A principal components analysis was used to provide another way of interpreting the pattern of genetic relationships among samples. We focused on data for the following 29 loci, in which the common allele had a frequency of less than 0.95 in at least one sample: sAAT-3*, sAAT-4*, ADA-1*, mAH-2*, sAH*, CK-A2*, EST-1*, FBALD-4*, FH*, bGALA*, GAPDH-2*, GAPDH-3*, bGLUA*, GPI-A*, GPI-B2*, sIDHP-1*, sIDHP-2*, LDH-B1*, MPI*, PEPA*, PEPC*, PEPD-2*, PEPLT*, PGK-1*, PGM-1*, PGM-2*, PNP-1*, sSOD-1*, TPI-3*. Eigenvectors were extracted from a matrix of correlations computed from allele frequencies. The NTSYS-pc computer program was used for the principal component analysis (Exeter Software 1993).

A scatterplot of principal component scores for principal components PC1 and PC2 is provided in Figure 31. These two components together describe about 22% of the total variance among samples, so considerable information is contained in other principal components not shown in the figure. Several geographic clusters are identifiable here. For example, samples from Alaska are the most divergent and are well separated with PC1 scores greater than 0.09 (off the scale and therefore not shown in Figure 31). Samples from British Columbia and Puget Sound are characterized by high PC1 scores and moderate PC2 scores. Samples from the Oregon coast north of Cape Blanco have moderate to low scores for PC1 and low scores for PC2.

Two samples from the lower Columbia River and three samples from south of Cape Blanco are also located in the Oregon coast portion of the scatterplot. Samples from the lower Columbia River have low PC1 scores and high PC2 scores. Three of the samples from the southwest Washington coast, two samples from south of Cape Blanco, and the two samples from the mid-Fraser River are also located in the Columbia River portion of the scatterplot. The mid- Fraser samples, however, are well separated from all other samples by PC4. The remaining three samples from the southwest Washington coast and the three samples from the Olympic Peninsula have moderate scores from both PC1 and PC2 and are near the convergence of the Puget Sound/British Columbia and lower Columbia River clusters.

California and southern Oregon data set--Because our new data set included only a single sample from California, we examined the genetic relationships of California and southern Oregon coho salmon populations by combining data for NMFS samples from this area with data from Olin (1984) and Bartley et al. (1992). The geographic coverage of samples in this analysis can be seen in Figure 32 and Table 4; these include the 5 most southern NMFS samples, 20 samples from Bartley et al. (1992), and the Iron Gate Hatchery sample from Olin (1984). Samples with 25 or fewer fish were excluded from the analysis.

Ideally, inferences about genetic relationships based on genetic distances should be based on a set of gene loci common to all pairwise comparisons of populations. Even after exclusion of the smallest samples, however, few loci were scored in all samples in the combined dataset. Therefore, to allow an analysis using a minimally-representative set of loci, we computed genetic distances (Cavalli-Sforza and Edwards 1967) for each pair of populations using the maximum number of loci common to both samples. The following 13 polymorphic loci were considered in this analysis: sAAT-1, 2*; sAH; GPI-A*; IDDH-1*; LDH-B1*; LDH-B2*; sMDH-B1,2*; MPI*; PEPA*; PEPC*; PEPD-2*; PGDH*; PGM-1*. Each of these loci had a common allele frequency of 0.95 or less in at least one sample and was scored in at least 50% of the samples.

In the dendrogram resulting from this analysis (Fig. 33), two major geographic clusters are apparent and are separated by a relatively large genetic distance (D = 0.126). The northern (and primarily large-river) group includes 11 samples from the Elk River (near Cape Blanco) to the Eel River (just north of Cape Mendocino). The southern (and primarily small- river) group includes nine samples spanning a geographic range from Fort Bragg to Tomales Bay (Lagunitas Creek), in addition to three samples from north of Cape Mendocino. The single sample from south of Tomales Bay (Scott Creek) and two additional samples from south of Punta Gorda (Cottoneva and Pudding Creeks) are outliers to both of the major groups. Considerable genetic diversity among populations is apparent within both groups.

Table 4. Samples used in allozyme analysis of coho salmon from southern Oregon and California. Map codes correspond to those in Figure 32. N is the number of fish in each sample.
Map code Sample Sourcea N

1 Scott Creek 1 39
2 Lagunitas Creek 1 32
3 Tanner Creek, Salmon Creek 1 62
4 Russian River, Willow Creek 1 38
5 Navarro River, Flynn Creek, John Smith Creek 1 61
6 Albion River 1 30
7 Little River 1 51
8 Russian Gulch 1 31
9 Caspar Creek 1 82
10 Hare Creek 1 28
11 Pudding Creek 1 47
12 Cottoneva Creek 1 28
13 South Fork Eel River, Huckleberry Creek 1104
14 South Fork Eel River, Butler Creek 1 60
15 South Fork Eel River, Redwood Creek 1 58
16 Elk River (Humboldt Bay) 1 30
17 Trinity River, Trinity Hatchery 1 172
18 Trinity River, Trinity Hatchery 1 36
19 Trinity River, Trinity Hatchery 2 98
20 Trinity River, Deadwood Creek 1 26
21 Klamath River, Iron Gate Hatchery 3 92
22 Smith River, West Branch Mill Creek 1 30
23 Rogue River, Illinois River, Greyback Creek 2 40
24 Rogue River, Illinois River, Silver Creek 2 29
25 Rogue River, Cole Rivers Hatchery 2 80
26 Elk and North Fork Elk Rivers 2 32
a1 = Bartley et al. 1992; 2 = Samples collected by NMFS for this status review; 3 = Olin 1984
Table of Contents