Population structure and stock identification of sockeye salmon (Oncorhynchus nerka) in coastal lakes in British Columbia, Canada

834 Population structure and stock identification of sockeye salmon (Oncorhynchus nerka) in coastal lakes in British Columbia, Canada T.D. Beacham, B. McIntosh, and C. MacConnachie Abstract: Population structure of sockeye salmon, Oncorhynchus nerka (Walbaum, 1792), from coastal lakes in British Columbia was determined from a survey of variation of 14 microsatellite loci, with approximately 6400 sockeye salmon analyzed from 40 populations. Populations from the Queen Charlotte Islands displayed fewer alleles per locus than did populations in other regions. Genetic differentiation among the populations surveyed was observed, with the mean F ST for all loci being 0.077 (SD = 0.006). Differentiation among populations was approximately 13 times greater than annual variation within populations. Regional structuring of the populations surveyed was observed. The accuracy and precision of the estimated stock compositions generally increased as the number of observed alleles at the loci increased. Simulated mixed-stock samples generated from observed population frequencies in different regions suggested that variation at microsatellite loci provided reasonably accurate and precise estimates of stock composition for potential samples from marine or freshwater fisheries. Résumé : Un inventaire de la variation à 14 locus microsatellites chez environ 6400 saumons nerka, Oncorhynchus nerka (Walbaum, 1792), provenant de 40 populations a permis de décrire la structure des populations dans des lacs côtiers de la Colombie-Britannique. Les populations des îles de la Reine-Charlotte possèdent moins d allèles par locus que les autres populations. Il existe une différentiation génétique entre les populations inventoriées et le F ST moyen pour tous les locus est de 0,077 (ET = 0,006). La différentiation entre les populations est d environ 13 fois plus importante que la variation annuelle au sein d une même population. Il existe une structuration régionale chez les populations étudiées. La justesse et la précision des compositions des stocks estimées augmentent généralement en fonction du nombre d allèles observés au divers locus. Des échantillons simulés de stocks mixtes obtenus à partir des fréquences des populations observées dans les différentes régions indiquent que la variation aux locus microsatellites permet une estimation raisonnablement juste et précise de la composition des stocks pour des échantillons potentiels des pêches marines ou d eau douce. [Traduit par la Rédaction] Beacham et al. 844 Introduction Sockeye salmon, Oncorhynchus nerka (Walbaum, 1792), populations are widely distributed in British Columbia, with large populations found in major river drainages, such as the Fraser River in the south and the Nass River and Skeena River in the north. Outside of the major river drainages, many smaller populations are found in the small lakes on the Queen Charlotte Islands, in the central coastal region, on Vancouver Island, and the southern British Columbia mainland. Sockeye salmon typically spawn in tributaries to lakes or along the lake shore, and the juveniles rear in these nursery lakes for at least 1 year before migrating to the ocean (Burgner 1991). Most sockeye salmon in British Columbia mature at 4 or 5 years of age, but some individuals, primarily males, can mature at 3 years of age. Historically, the dominant population in the central coast originated from Owikeno Received 4 October 2004. Accepted 12 May 2005. Published on the NRC Research Press Web site at http://cjz.nrc.ca on 30 July 2005. T.D. Beacham, B. McIntosh, and C. MacConnachie. Pacific Biological Station, Fisheries and Oceans Canada, 3190 Hammond Bay Road, Nanaimo, BC V9T 6N7, Canada. 1 Corresponding author (e-mail: beachamt@pac.dfo-mpo.gc.ca). Lake in Rivers Inlet, but the abundance of this population declined dramatically in the 1990s owing to low marine survival that was unrelated to fishery exploitation (McKinnell et al. 2001). Poor marine survival in the 1990s was observed for a number of sockeye salmon populations in this region (Rutherford and Wood 2000), leading to reduced population abundance and associated concerns over the appropriate management strategies required to promote rebuilding of abundance. The marked decline in abundance of Owikeno Lake sockeye salmon has led to the development of enhancement options for this stock. Within Owikeno Lake, there are a number of tributary streams in which sockeye salmon spawn, leading to the potential for genetically discrete populations. Genetically discrete sockeye salmon populations within a single lake can occur (Burger et al. 1997). Determination of genetic structure among Owikeno Lake sockeye salmon populations is necessary to avoid reduction of genetic diversity resulting from enhancement activities, as well as to avoid outbreeding depression owing to translocations among genetically discrete populations. Nelson et al. (2003) reported that there was no genetic differentiation among sockeye salmon populations of Owikeno Lake, but given the number of fish sampled in a year in a population (38 50 fish), the power to detect differentiation may have been lim- Can. J. Zool. 83: 834 844 (2005) doi: 10.1139/Z05-066

Beacham et al. 835 Fig. 1. Map indicating sampling locations for 40 coastal populations of sockeye salmon (Oncorhynchus nerka) in British Columbia. The specific populations in each geographic area are outlined in Table 1. ited. Confirmation that there is no genetic differentiation among tributary populations of Owikeno Lake sockeye salmon is required. Determination of the appropriate population structure or management unit is a key requirement for management of sockeye salmon in coastal lakes in British Columbia. Surveys of genetic variation have been demonstrated to be very useful in determining salmonid population structure. Surveys of variation at allozyme loci have demonstrated that the nursery lake is a key component in sockeye salmon population structure (Wood et al. 1994; Wood 1995). Although there is strong allozyme differentiation among sockeye salmon populations in different lakes, and to some degree among river drainages, regional structuring of the populations is less apparent, as the nearest geographic populations are not necessarily the most similar genetically in allozyme surveys (Wood et al. 1994; Varnavskaya et al. 1994; Wood 1995; Winans et al. 1996). Microsatellites have been demonstrated to be very effective in determining salmonid population structure (Banks et al. 2000) and in applications for stock identification (Beacham et al. 2000a, 2000b, 2004). There are a large number of microsatellite loci that can be employed in surveys of salmonid population structure and stock identification, and the key characteristics of the loci include the number of alleles observed at a locus, the range of allele sizes, and the level of population differentiation. Earlier theoretical modelling studies suggested that in terms of evaluating population structure, a modest number of loci was preferred, with each locus having a modest number of alleles (Smouse and Chevillion 1998). Bernatchez and Duchesne (2000) suggested that loci displaying 6 10 alleles were as effective as more variable loci in population assignment studies. Beacham et al. (2002) suggested that the number of alleles present at a locus was related to the power of a locus to provide accurate identification of individual sockeye salmon to one of three possible source lakes. The more alleles that were present at a locus, the greater was the power of the locus for individual identification, similar to the results observed by Cornuet et al. (1999). Evaluation of sets of loci, with loci within sets having similar numbers of alleles per locus, but differences among sets in locus allele numbers, may prove effective in evaluating the relationship between the number of alleles per locus and the accuracy of estimates of proportional stock contributions to a mixture. Sockeye salmon are typically caught in marine fisheries on their journey to their natal spawning grounds. Effective management of these fisheries requires that accurate estimates of the origin of fish in the fishery be determined to the smallest practical group, which typically can be to a management area, but in some cases may require indentification to a particular lake or river. Stock composition estimates are generally performed at the level of a management area, lake, or river, but in some cases may require the identification of individual fish to specific lakes (Beacham et al. 2004). The objective of the present study was to analyze variation at 14 microsatellite loci in coastal British Columbia sockeye salmon populations to evaluate population structure. Potential genetic differentiation among sockeye salmon spawning in tributaries to Owikeno Lake was examined, with microsatellite variation in populations in nine tributaries surveyed for 2 4 years. Additionally, the power of sets of microsatellite loci for stock identification applications was evaluated, with differences in the mean number of alleles per locus among sets, but similar allelic variation per locus within sets. Finally, the suitability of using microsatellite variation to determine accuracy of estimates of proportional stock contributions to a mixture of sockeye salmon from coastal lakes in British Columbia was evaluated. Methods and materials Collection of DNA samples and laboratory analysis Tissue samples were collected from adult fish from sockeye salmon populations in the central coast, Queen Charlotte Islands, northeast Vancouver Island, and southern mainland regions in British Columbia, Canada, and DNA was extracted from the samples as described by Withler et al. (2000). Juveniles were sampled from lakes in the Nimpkish River drainage (Woss, Vernon, and Nimpkish lakes) on northeastern Vancouver Island, and from Village Bay Lake on Quadra Island in Johnstone Strait (Fig. 1). Loci amplified via polymerase chain reactions were the following: dinucleotide repeats Ots2 and Ots3 (Banks et al. 1999); tetranucleotide or di/tetra nucleotide repeats Ots100, Ots103, Ots107, and Ots108 (Beacham et al. 1998; Nelson and Beacham 1999); tetranucleotide repeats Oki1a, Oki1b, Oki6, Oki10, Oki16, and Oki29 (Smith et al. 1998; Nelson et al. 2003); dinucleotide repeat One8 (Scribner et al. 1996); and dinucleotide repeat Omy77 (Morris et al. 1996). All loci were size-fractionated on denaturing polyacrylamide gels and allele sizes determined with the ABI 377 automated DNA sequencer. Allele sizes were determined with Genescan version 3.1 and Genotyper version 2.5, along with the GeneScan-500 size standard (PE Biosystems, Foster City, California). Data analysis Each population at each locus for each year was tested for departure from Hardy Weinberg equilibrium (HWE) using

836 Can. J. Zool. Vol. 83, 2005 Table 1. Population, nursery lake, sample collection years, number of fish sampled per year, and total number of fish sampled for 40 populations of coastal British Columbia sockeye salmon, Oncorhynchus nerka. Site Nursery lake Years Number of fish/year Queen Charlotte Islands Mercer Creek Mercer 1983 41 41 Yakoun River Yakoun 1989, 1993 61, 99 160 Awun River Awun 1995 80 80 Naden River Eden 1995 98 98 Copper Creek Skidegate 1993, 1996, 2001 85, 95, 10 190 Central coast Tributary Devon 1985, 1999 100, 100 200 Mikado 1986, 1999 100, 62 162 Lowe 1986 40 40 Banks 1986 41 41 Canoona River Canoona 1986 100 100 Tezwa River Kitlope 1986 40 40 Atnarko River Tenas 1985 80 80 Lonesome 1997 100 100 Tributary Namu 1999 93 93 Mary Cove Creek None 1999 78 78 Lagoon Creek Lagoon 1999 50 50 Lakeshore Kimsquit 1986 81 81 Tributary Tankeeah 1986, 2001, 2002 100, 30, 31 161 Klemtu 2002 27 27 Koeye 1986 80 80 Marble Creek Owikeno 2001, 2002 25, 96 121 Inziana River Owikeno 1997, 2000, 2001, 2002 50, 151,100, 97 398 Washwash River Owikeno 1997, 2000, 2001, 2002 63, 91, 114, 99 367 Ashlulm River Owikeno 2000, 2001, 2002 25, 82, 94 201 Dallery River Owikeno 2000, 2001, 2002 32, 33, 95 160 Genesee River Owikeno 2000, 2001, 2002 7, 35, 88 130 Neechanz River Owikeno 2000, 2001, 2002 69, 110, 96 275 Amback River Owikeno 2000, 2001, 2002 92, 100, 58 250 Sheemahant River Owikeno 2000, 2001, 2002 43, 100, 113 256 Docee River (mixed) Long 1989, 1998, 1999, 2001 60, 200, 49, 100 409 Smokehouse Creek Long 2001, 2002 56, 205 261 Canoe Creek Long 2001 39 39 Vancouver Island Tributary Quatse 2003 197 197 Trawl Woss 1985, a 2001, b 2002 b 80, 112, 101 293 Vernon 2001, b 2002 b 77, 290 367 Nimpkish 2001, b 2002 b 56, 42 108 Southern mainland Klinaklini River Devereux 1998, 2002 219, 106 325 Phillips River Phillips 2002 205 205 Trawl b Village Bay 2003 18 18 Lakeshore Sakinaw 1998, 2000, 2001 81, 20, 12 113 a The 1985 samples were adults. b Juveniles sampled from mid-water trawl of lakes in the Nimpkish River drainage in 2001 and 2002. Total number of fish/40 populations GDA (Lewis and Zaykin 2001). For each locus, the 77 annual samples (Table 1) were tested separately. Critical significance levels for simultaneous tests (77 tests) were evaluated using sequential Bonferroni adjustment (Rice 1989). All annual samples available for a location were combined to estimate population allele frequencies, as was recommended by Waples (1990). Weir and Cockerham s (1984) F ST estimates for each locus over all populations were calculated with FSTAT version 2.9.3.2 (Goudet 2001). The significance of the multilocus F ST value over all samples was determined by jackknifing over loci. Cavalli-Sforza and Edwards s (1967) chord distance was used to estimate genetic distances among all populations. FSTAT was used to measure the allelic richness (allelic diversity standardized to a sample size of 17 fish, the approximate size of the Village Bay sample) for each population. Computation of the

Beacham et al. 837 number of alleles observed per locus was carried out with GDA (Lewis and Zaykin 2001). Estimation of variance components of population differences and annual variation within populations were determined with GDA. Only populations with 2 years of sampling (21 populations; Table 1) were included in the analysis. The 2000 sample from Genesee River was not included in the analysis, as it was composed of only seven fish and was unlikely to be representative of allele frequencies in that year. Allele frequencies for all location samples surveyed in this study are available at http://www-sci.pac.dfo-mpo.gc.ca/mgl/default_e. htm. Estimation of stock composition The first step in the analysis was to construct a baseline of genotypic frequencies for all populations present in the mixture. Genotypic frequencies were determined at each locus in each population through the analysis of microsatellite variation. For the stock composition analysis, a mixture of fish was then randomly sampled and the same genetic markers were surveyed for each individual as in the baseline samples. The estimated stock composition of the mixture is the maximum likelihood contribution estimate for the baseline populations that lead to the greatest likelihood of obtaining a mixture sample with the observed set of characters. The statistical package for the analysis of mixtures software program (SPAM version 3.7) was used to estimate stock composition of each mixture (Debevec et al. 2000). The Rannala and Mountain (1997) correction to baseline allele frequencies was used in the analysis to avoid the occurrence of fish in the mixed sample from a specific population having an allele not observed in the baseline samples from that population. All loci were assumed to be in Hardy Weinberg equilibrium and expected genotypic frequencies were estimated from the observed allele frequencies. The 40-population baseline was used in the majority of estimations of stock composition in the study. Each baseline population was resampled with replacement to simulate random variation involved in the collection of the baseline samples before the estimation of stock composition of each simulated mixture. Simulated mixtures were developed that may be representative of samples from marine fisheries. The objective was to identify the accuracy and precision of estimated stock compositions corresponding to specific lakes, rivers, or regions that may be expected when applied to fishery samples. Simulated mixtures composed of Queen Charlotte Islands, Nimpkish River, central coast, and south coast populations were examined to evaluate accuracy and precision of the estimated stock compositions. Simulated fishery samples of 150 fish were generated by randomly resampling with replacement the baseline populations in each drainage. Estimated stock composition of a simulated mixture was then determined, and the whole process was repeated 100 times to estimate the mean and SD of the individual stock composition estimates. Results Variation within populations All loci surveyed were polymorphic in all populations sampled. The number of observed alleles at each locus Table 2. Number of alleles, expected heterozygosity (H e ), observed heterozygosity (H o ), percent significant Hardy Weinberg equilibrium tests (HWE, N = 77 tests), and F ST (±SD) among 40 sockeye salmon populations for 14 microsatellite loci. Locus Number of alleles H e H o HWE F ST Oki1a 8 0.51 0.49 3.9 0.074 (0.014) Oki1b 6 0.47 0.47 1.3 0.071 (0.007) Oki6 29 0.65 0.62 1.3 0.093 (0.020) Oki10 74 0.91 0.85 13.1 0.024 (0.004) Oki16 25 0.70 0.68 1.3 0.126 (0.012) Oki29 31 0.79 0.79 0.0 0.050 (0.009) Omy77 17 0.72 0.72 3.9 0.108 (0.0218) One8 25 0.68 0.69 0.0 0.065 (0.010) Ots2 22 0.77 0.78 1.3 0.074 (0.017) Ots3 21 0.63 0.62 1.3 0.137 (0.020) Ots100 33 0.82 0.79 1.3 0.095 (0.012) Ots103 29 0.89 0.87 2.6 0.045 (0.006) Ots107 13 0.44 0.43 1.3 0.069 (0.011) Ots108 25 0.86 0.84 2.6 0.061 (0.007) All loci 0.077 (0.006) ranged from 6 to 74, with heterozygosity proportional to allele number (Table 2). Maximum heterozygosity was observed at Oki10, the locus with the largest observed number of alleles. Genotypic frequencies at each locus within sampling location and year generally conformed to those expected under Hardy Weinberg equilibrium, with the possible exception of Oki10, where 13.1% of the HWE tests were significant (Table 2). More homozygous genotypes than expected were observed at this locus, and it was likely due to differential amplification of larger sized alleles. Regional and population differences in allelic richness and heterozygosity were observed. The Quatse River population had the lowest number of alleles and lowest heterozygosity of any of the 40 populations surveyed (Table 3). The nine populations surveyed from Owikeno Lake had among the greatest allelic richness and heterozygosity. Regional differences in allelic diversity and heterozygosity were observed, with populations from the Queen Charlotte Islands having fewer alleles (mean 5.6 alleles for a standard sample size) than central coast (6.8 alleles), south coast (6.3 alleles), or Nimpkish River populations on Vancouver Island (6.3 alleles) (P < 0.05). Queen Charlotte Islands populations tended to be less heterozygous than those in other coastal regions of British Columbia. Distribution of genetic variation Gene diversity analysis of the 14 loci surveyed was used to determine the magnitude of annual variation within populations and of variation among 21 sockeye salmon populations, with only populations with 2 years of sampling included in the analysis. The amount of variation contained within populations ranged from 88.3% (Ots3) to 98.2% (Oki10), with the average for microsatellite loci being 93.7% (Table 4). The maximum range of sampling times was 17 years for the Woss Lake population, with a maximum of 4 nonconsecutive years of sampling recorded for a single population. Variation among sampling years within popula-

838 Can. J. Zool. Vol. 83, 2005 Table 3. Genetic variation within populations of sockeye salmon from coastal locations in British Columbia. Site Average number of allelles Standardized allelic richness a H e H o Queen Charlotte Islands Mercer Creek 6.5 5.1 0.63 0.64 Yakoun River 9.7 5.8 0.67 0.68 Awun River 8.6 5.9 0.67 0.66 Naden River 9.3 5.8 0.65 0.64 Copper Creek 9.3 5.2 0.64 0.63 Central coast Devon 12.9 7.6 0.69 0.68 Mikado 11.1 6.3 0.69 0.68 Lowe 6.9 5.6 0.71 0.68 Banks 8.8 6.6 0.70 0.65 Canoona River 8.5 6.9 0.66 0.64 Tezwa River 8.6 7.7 0.72 0.66 Tenas 10.7 6.6 0.68 0.67 Lonesome 10.4 6.9 0.68 0.65 Namu 9.0 6.1 0.67 0.64 Mary Cove Creek 9.4 7.6 0.70 0.62 Lagoon Creek 8.4 6.5 0.73 0.70 Kimsquit 8.0 6.7 0.65 0.61 Tankeeah 12.1 6.5 0.71 0.68 Klemtu 8.1 6.0 0.76 0.81 Koeye 9.1 6.5 0.70 0.67 Marble Creek b 13.4 6.9 0.75 0.75 Inziana River b 15.4 6.4 0.76 0.75 Washwash River b 16.0 7.6 0.76 0.75 Ashlulm River b 14.4 7.5 0.75 0.74 Dallery River b 13.9 7.6 0.76 0.74 Genesee River b 13.8 5.4 0.76 0.76 Neechanz River b 15.4 7.7 0.76 0.76 Amback River b 14.9 7.6 0.76 0.76 Sheemahant River b 15.2 7.6 0.76 0.76 Docee River (mixed) 14.4 7.0 0.70 0.70 Smokehouse Creek 12.7 7.0 0.71 0.68 Canoe Creek 9.6 5.7 0.69 0.71 Vancouver Island Quatse 7.6 4.7 0.57 0.55 Woss 12.4 6.2 0.71 0.70 Vernon 13.0 6.5 0.74 0.70 Nimpkish 10.5 6.1 0.71 0.70 Southern mainland Klinaklini River 16.6 6.5 0.74 0.65 Phillips River 12.8 6.5 0.68 0.67 Village Bay 7.0 6.3 0.72 0.75 Sakinaw 9.6 5.9 0.69 0.67 a Standardized to a sample of 17 fish. b Sites associated with Owikeno Lake. tions was the smallest source of variation observed, accounting for just 0.44% of all variation. Variation among populations accounted for an average 5.86% of the observed variation, with significant differentiation among populations observed at all loci (all P < 0.01). Differentiation among populations was approximately 13 times greater than annual variation within populations. For the time period surveyed in our study, annual variation in allele frequencies was minor relative to the variation among populations. Potential differentiation among the nine populations sampled in Owikeno Lake was examined in greater detail. Significant differentiation was observed among the nine populations at only One8 (F [8,18] = 3.21, P < 0.05), indicating that there has been some restriction of gene flow among

Beacham et al. 839 Table 4. Hierarchical gene diversity analysis (relative diversity) of 21 coastal populations of British Columbia sockeye salmon for 14 microsatellite loci. Locus Within populations Among years within populations Among populations Oki1a 0.9326 0.0256** 0.0418** Oki1b 0.9483 0.0083** 0.0434** Oki6 0.9237 0.0146** 0.0617** Oki10 0.9816 0.0025 0.0159** Oki16 0.8900 0.0025 0.1075** Oki29 0.9632 0.0035* 0.0333** Omy77 0.9235 0.0023 0.0742** One8 0.9522 0.0004 0.0474** Ots2 0.9304 0.0013 0.0683** Ots3 0.8828 0.0027 0.1145** Ots100 0.9200 0.0020 0.0780** Ots103 0.9716 0.0011 0.0273** Ots107 0.9476 0.0002 0.0522** Ots108 0.9489 0.0022 0.0489** All 0.9370 0.0044* 0.0586** Note: Only populations sampled in 2 years were included in the analysis as outlined in Table 1. *, P < 0.05; **, P < 0.01. Fig. 2. UPGMA (unweighted pair-group method with arithmetic averaging) dendrogram of Cavalli-Sforza and Edwards (1967) chord distance for 40 populations of sockeye salmon surveyed at 14 microsatellite loci. Bootstrap values at the tree nodes indicate the percentage of 100 trees where annual samples from a population beyond the node clustered together. The mixed-population Docee River samples have been designated as Long Lake. the spawning locations. This differentiation was due entirely to the Inziana River population at the extreme eastern end of the lake. No differentiation was observed among the other eight populations at any locus. There was little evidence of genetically distinct spawning populations in most of the tributaries to Owikeno Lake. Population structure and variation Substantial genetic differentiation among the 40 sockeye salmon populations sampled in our study was observed. The overall F ST for the 14 microsatellite loci surveyed was 0.077, with individual locus values ranging from 0.024 at Oki10 to 0.137 at Ots3, and with all values significantly greater than 0 (P < 0.05; Table 2). Regional structuring of population samples was observed in our study. For example, all populations (Yakoun, Awun, and Naden) from the northern coast of the Queen Charlotte Islands clustered together 100% of the time, and were most closely associated with a population from the east coast of the Queen Charlotte Islands (Fig. 2). All populations (nine in total) sampled from Owikeno Lake clustered together, as did the three populations from Long Lake. Similarly, all populations from the Nimpkish River drainage clustered together, as did the two most southern mainland populations (Village Bay and Sakinaw) sampled. Populations in the central coastal region did not form a single distinct cluster, but there was still some level of clustering of populations within local geographic areas (Fig. 2). The Quatse River population, less variable genetically than any of the other 39 populations surveyed, was the most distinctive population observed. It was most closely related to the Sakinaw Lake, Village Bay Lake, and Mercer Lake populations. Comparisons among loci The 14 microsatellite loci surveyed were divided into three classes based upon the number of observed alleles at each locus, with the first set containing 5 loci displaying 6 21 alleles per locus, the second set containing 5 loci displaying 22 29 alleles per locus, and the third set containing 4 loci with 29 or more alleles per locus. Single population mixtures composed of 18 representative populations from the Queen Charlotte Islands, central coast, Vancouver Island, and south coast were evaluated for the power of the sets of loci to provide accurate and precise estimates of composition of the single specific population mixtures using the 39-population coastal baseline (Docee River mixed sample not included) to resolve the mixtures. Some populations, such the Copper River population from the Queen Charlotte Islands, were readily identified with a high degree of accuracy and precision, regardless of the sets of loci used for estimation (Table 5). Other populations, particularly the tributary populations to Owikeno Lake, were much more difficult to resolve. The accuracy and precision of the estimated stock compositions generally increased as the number of observed alleles at the loci increased. The third locus set only contained four loci, whereas the other sets contained five loci, yet accuracy and precision of the estimates derived from the third locus set were generally higher than in the previous two sets. Combining all loci for stock identification applications generally provided the most accurate and precise estimates of stock composition, and there was little evidence to indicate that adding loci to the set used for stock composi-

840 Can. J. Zool. Vol. 83, 2005 Table 5. Mean (±SD) estimated percent stock compositions of single population mixtures (correct = 100%) for 18 representative populations of coastal sockeye salmon from the Queen Charlotte Islands, central coast, Vancouver Island, and south coast calculated with three classes of microsatellite loci, and with all loci combined. Population Class 1 Class 2 Class 3 All microsatellites combined Yakoun 95.6 (3.3) 96.7 (1.7) 96.0 (2.1) 96.7 (1.4) Copper 97.2 (1.5) 98.1 (1.2) 98.0 (1.2) 98.0 (1.3) Devon 81.9 (9.4) 90.1 (4.9) 88.1 (5.5) 92.2 (2.8) Mikado 83.6 (8.5) 92.3 (4.5) 94.4 (3.4) 93.2 (2.7) Canoona 94.9 (2.5) 96.3 (1.5) 96.8 (1.3) 95.7 (1.8) Lonesome 73.1 (9.3) 76.4 (7.7) 85.1 (5.9) 83.5 (4.7) Kimsquit 92.6 (3.2) 92.7 (2.2) 92.5 (2.3) 92.3 (2.7) Tankeeah 93.7 (3.3) 97.7 (1.4) 96.9 (1.6) 97.4 (1.2) Marble 61.4 (13.1) 62.7 (8.7) 63.0 (8.1) 67.7 (5.6) Washwash 64.0 (15.1) 72.7 (9.3) 74.8 (9.5) 79.7 (6.0) Dallery 42.9 (13.2) 58.0 (9.4) 72.8 (7.3) 70.0 (5.4) Smokehouse 80.0 (10.3) 79.7 (7.2) 87.6 (6.6) 88.8 (4.3) Quatse 98.7 (1.0) 98.8 (0.8) 98.8 (0.9) 98.8 (1.0) Woss 85.1 (9.6) 91.6 (4.5) 90.4 (4.8) 92.1 (3.4) Vernon 88.5 (9.0) 93.7 (4.1) 94.9 (3.9) 96.1 (2.3) Klinaklini 91.6 (3.9) 94.3 (2.9) 95.8 (2.2) 98.2 (1.3) Phillips 97.6 (1.4) 97.6 (1.3) 98.1 (1.2) 98.2 (1.1) Sakinaw 96.6 (1.8) 98.0 (1.1) 97.8 (1.2) 97.8 (1.4) Mean 84.4 (6.6) 88.2 (4.1) 90.1 (3.8) 90.9 (2.8) Note: Class 1 contained five loci with 6 21 alleles (Oki1a, Oki1b, Ots107, Omy77, Ots3), class 2 contained five loci with 22 29 alleles (Ots2, Ots108, Oki16, Ots103, One8), and class 3 contained four loci with >29 alleles (Ots100, Oki6, Oki10, Oki29). Simulations were conducted using a 39-population baseline (Docee River mixed sample not included), 150 fish in the mixture sample, and 100 resamplings in the mixture sample and baseline samples. tion estimation resulted in any degradation of accuracy and precision of the estimates. The most accurate and precise estimates of stock composition were obtained when all loci were used in the estimation procedure. Stock identification We examined whether the genetic differentiation observed among sockeye salmon in the coastal lakes of British Columbia was sufficient for mixed-stock analysis, with the objective of obtaining accurate stock compositions by region and possibly lake or river drainage when the actual composition of the mixture is unknown. This was evaluated by simulating seven fishery mixture samples of known origin for mixtures of Queen Charlotte Islands, central coast, south coast, and Nimpkish River (Vancouver Island) populations. Six of the mixtures consisted of populations from only a single region or lake, with the expectation that errors in estimation of the region or lake in question would be maximized when the single region or lake comprised 100% the simulated mixture. With a baseline of 40 coastal populations, mixtures composed solely of Queen Charlotte Islands populations (mixture 1) were estimated with a high degree of accuracy, with the mean error of estimation of a specific population <1% and the error of the regional estimate <2% (Table 6). Mixtures that consisted solely of Nimpkish River populations (mixture 2) were estimated with 4% accuracy to specific population, and with <1% error for the regional estimate. Mixtures that consisted solely of Owikeno Lake populations (mixture 3) were well estimated for the Owikeno Lake component, with <1% mean error. However, given the lack of differentiation among the specific tributary components, estimation errors of the specific tributary populations were as large as 7% (Table 6). Mixtures containing Long Lake sockeye salmon (mixture 4) were well estimated for the Long Lake component, with <2% error. The Docee River baseline samples, collected at a fence on the Docee River draining Long Lake, consisted of mixed samples of tributary populations of Long Lake, including Smokehouse Creek and Canoe Creek. Given the nature of the baseline samples, with both Smokehouse Creek and Canoe Creek sockeye salmon likely included in the Docee baseline sample, larger estimation errors for the individual populations (up to 17%) were observed. Mixtures that consisted solely of central coast populations (mixture 5) were estimated with error rates of individual populations generally <2%, and with error rates of the regional estimate <4%. South coast populations (mixture 6) were estimated with mean error rates <1% for both the specific populations and regional component. Mixtures comprising populations originating from only a single region are rare, particularly if the sample is derived from a marine fishery. Multi-regional mixtures are typically analyzed in mixed-stock analysis, and we evaluated a mixture that consisted of populations from the south coast, central coast, Queen Charlotte Islands, and the Nimpkish River on Vancouver Island (mixture 7). For this mixture, the maximum error of estimation for specific populations was <4%, but the regional contribution to the mixtures was accurately estimated with an error of <1% for all regions (Table 6).

Beacham et al. 841 Table 6. Estimated (±SD) percent composition of seven simulated mixtures of sockeye salmon from the Queen Charlotte Islands, Nimpkish River, central coast, and south coast using variation at 14 microsatellite loci and a 40-population baseline outlined in Table 1. Actual Estimated Mixture 1 Awun 30 29.0 (3.8) Copper 10 10.0 (2.3) Mercer 10 9.1 (2.4) Naden 20 20.5 (3.2) Yakoun 30 29.7 (4.0) ΣQueen Charlotte Islands 100 98.3 (1.1) Mixture 2 Nimpkish 20 16.1 (5.1) Vernon 30 33.8 (6.1) Woss 50 50.0 (6.1) ΣNimpkish River 100 99.9 (0.3) Mixture 3 Amback 10 9.7 (5.3) Dallery 20 13.5 (4.8) Inziana 20 18.6 (5.8) Marble 5 4.1 (3.2) Neechanz 15 17.1 (5.3) Sheemahant 20 18.5 (5.8) Washwash 10 12.9 (6.0) ΣOwikeno 100 99.8 (0.3) Mixture 4 Docee 50 63.8 (3.8) Canoe 30 13.3 (3.8) Smokehouse 20 21.2 (5.6) ΣLong Lake 100 98.3 (1.1) Mixture 5 Canoona 10 9.6 (2.4) Kimsquit 20 18.8 (3.3) Kitlope 10 8.9 (2.3) Klemtu 20 16.9 (3.3) Koeye 10 10.0 (2.6) Lonesome 10 8.2 (2.4) Mary Cove 10 8.6 (2.2) Namu 10 9.6 (2.4) ΣCentral coast 100 96.4 (1.9) Mixture 6 Sakinaw 20 19.8 (3.4) Phillips 30 29.2 (3.9) Klinaklini 50 50.2 (4.0) ΣSouth coast 100 99.2 (0.3) Mixture 7 Sakinaw 10 9.7 (2.4) ΣSouth coast 9.8 (2.4) Vernon 20 19.4 (3.3) ΣNimpkish River 20.2 (3.4) Kitlope 10 8.8 (2.6) Lonesome 20 17.1 (3.3) Tankeeah 10 10.1 (2.6) ΣCentral coast 39.3 (3.9) Sheemahant 10 6.2 (1.3) ΣOwikeno 11.0 (2.4) Table 6. (concluded) Actual Estimated Naden 20 19.3 (3.3) ΣQueen Charlotte Islands 19.7 (3.3) Note: Each mixture of 150 fish was generated 100 times with replacement, and stock compositions of the mixtures were estimated by resampling with replacement each baseline population. Analysis of simulated mixtures of sockeye salmon from regions adjacent to and including the central coast suggested that microsatellite variation provided a practical means to estimate stock composition with a high degree of both population and regional accuracy. Discussion Population structure An important area investigated in our study was the population structure of sockeye salmon in the coastal regions of British Columbia. Previous allozyme surveys of sockeye salmon population structure in coastal British Columbia lakes had suggested that population structure was essentially that of a mosaic, with only weak evidence of structuring of populations within regions (Wood 1995; Winans et al. 1996). In an earlier survey of microsatellite variation centered on populations in the central coast, Nelson et al. (2003) reported a weak but significant relationship between genetic differentiation and geographic distance between populations, but concluded that population structure of the populations was like a mosaic. In our study, which included a survey encompassing a wider geographic range, more loci surveyed, and more populations surveyed, some level of population structure based upon geographic location was observed. There was clearly both a lake and a regional component to the population structure. The Quatse River population was the most genetically distinct of the populations surveyed, but it was most similar to the Village Bay and Sakinaw Lake populations of the south coast. Quatse River sockeye salmon have different life-history characteristics, being smaller (<2 kg) and having an earlier timing of entry into fresh water (as early as April, holding until spawning in September and October) than typically observed in sockeye salmon in southern British Columbia. Sakinaw Lake sockeye salmon also have a small body size and an early timing of entry into fresh water (Murray and Wood 2002). The Village Bay population, although having a smaller body size, returned to spawn later than the other two populations, between the beginning of July and the end of September (P. Zetterberg, personal communication). Some of the genetic differentiation observed among populations may be attributable to life-history characteristics, as well as to previously mentioned regional population structure. Sockeye salmon populations on the Queen Charlotte Islands were less diverse genetically than other populations surveyed, with fewer observed alleles and lower heterozygosities than populations in other regions. Fewer observed alleles and lower heterozygosity are suggestive of a recent bottleneck in population size. Portions of the Queen Charlotte Islands have been documented as nonglaciated during the last glaciation (Warner et al. 1982), so it is likely that

842 Can. J. Zool. Vol. 83, 2005 sockeye salmon were maintained in this refuge. However, population sizes were likely small in this refuge compared with refuges in other nonglaciated areas, and existing populations on the Queen Charlotte Islands may have been derived from sockeye salmon from this refuge. Reduced genetic variation observed for the Queen Charlotte Islands populations may reflect the population bottleneck experienced during the last glaciation. Owikeno Lake sockeye salmon have declined dramatically in abundance in the last 15 years, and increased emphasis has been placed upon assessment and management of this stock. The structure of populations in tributary streams has been of particular concern, as this will have implications for enhancement options for this stock. Nelson et al. (2003) reported that sockeye salmon spawning in the Wannock River, the outlet to Owikeno Lake, were distinct from sockeye salmon spawning in the tributary streams, and that sockeye salmon in the Amback River may be distinct from other tributary populations, although samples were available for only a single year. Our analysis of tributary populations indicated that sockeye salmon spawning in the Inziana River may be differentiated, but overall there was little evidence of genetic differentiation of tributary populations, similar to the findings of Nelson et al. (2003). Movement of sockeye salmon among tributary watersheds is a potential approach to promote recovery of the stock. Stock identification In salmonid stock identification applications, many more microsatellite loci are potentially available to be applied, provided that an adequate survey of contributing populations has been conducted, than are needed to provide reliable estimates of stock composition. Initial key characteristics are the size range and number of alleles observed at a locus, with later observed population differentiation a prime consideration. While some investigators have suggested that loci with modest numbers of alleles are preferred in studies examining population differentiation or stock identification (Smouse and Chevillon 1998; Bernatchez and Duchesne 2000), others have indicated that equivalent information can be obtained by examining either a few locus with many alleles or many loci with a few alleles (Kalinowski 2002). Our analysis, based upon actual baseline surveys, indicated that accuracy and precision of estimated stock compositions were related to the number of alleles observed at the locus used in the analysis. Stock compositions that estimated including loci with larger numbers of observed alleles tended to be more accurate and precise than those derived from loci with smaller numbers of alleles. One of the major potential applications of surveys of genetic variation of salmon populations is that of stock identification of mixed-stock fisheries, where the origins of fish contributing to mixed-stock fisheries are determined by comparing the genetic characteristics of fish in the fishery samples to the genetic characteristics of fish from potentially contributing populations. In the absence of known-origin samples, analysis of simulated mixed-stock samples of known origin is a practical method to evaluate the potential for applying genetic variation to mixed-stock fishery analysis. Analysis of simulated mixtures indicated that microsatellite variation provided accurate estimates of regional contributions of sockeye salmon stocks, and in some cases provided reliable estimates of individual populations in the simulated mixtures. Microsatellites have previously been reported to provide reliable estimates of stock composition in mixed-stock sockeye salmon fisheries (Beacham et al. 2000a, 2000b), and our results indicated that microsatellites should provide reliable estimates of stock composition for sockeye salmon in coastal lakes in British Columbia. Significant annual variation in allele frequencies was observed at 4 of 14 loci for the populations surveyed, but differences among populations, on average, were about 13 times larger than annual variation within populations. This relative magnitude of population differentiation versus temporal variation within populations was very similar to that previously observed in sockeye salmon populations in other regions of British Columbia. For example, differentiation among Nass River populations was about 11 times greater than annual variation within populations (Beacham and Wood 1999), and about 12 times greater for Barkley Sound populations on the west coast of Vancouver Island (Beacham et al. 2000a). The relative magnitude of population differentiation versus within population variation suggests that annual variation in allele frequencies will have minimal effects on estimates of stock composition. For practical estimation of stock composition, annual sampling of populations contributing to a sockeye salmon fishery is not required, but clearly some level of monitoring of allele frequencies over time is appropriate to ensure that allele frequencies characterizing each population are still relevant. Owikeno Lake sockeye salmon have undergone a rapid decline in abundance over the last 15 years, and effort has been directed to the conservation and enhancement of this stock. Estimation of the composition of simulated mixedstock samples containing Owikeno Lake sockeye salmon indicated that the Owikeno Lake component was estimated accurately, and thus reliable estimates of this stock should be possible in mixed-stock fishery samples. Application of microsatellite variation thus provides the opportunity to structure and conduct mixed-stock fisheries to enable exploitation of stocks with satisfactory abundance, while at the same time affording protection to stocks such as Owikeno Lake sockeye salmon that are under conservation concern. Acknowledgments A considerable effort was undertaken to obtain samples from sockeye salmon sampled in this study. We would like to acknowledge Fisheries and Oceans Canada staff from the North Coast and Central Coast areas in the Pacific Region who collected or supervised collections in many of the rivers. We also acknowledge the various agencies and organizations that collected samples, which include the Haida Fisheries program for the Queen Charlotte Islands and the Kitasoo Fisheries Program for some Central Coast populations. C. Wallace aided in the analyses. Funding for the study was provided by Fisheries and Oceans Canada. References Banks, M.A., Blouin, M.S., Baldwin, B.A., Rashbrook, V.K., Fitzgerald, H.A., Blankenship, S.M., and Hedgecock, D. 1999.

Beacham et al. 843 Isolation and inheritance of novel microsatellites in chinook salmon (Oncorhynchus tshawytscha). J. Hered. 90: 281 288. Banks, M.A., Rashbrook, V.K., Calavetta, M.J., Dean, C.A., and Hedgecock, D. 2000. Analysis of microsatellite DNA resolves genetic structure and diversity of chinook salmon (Oncorhynchus tshawytscha) in California s Central Valley. Can. J. Fish. Aquat. Sci. 57: 915 927. Beacham, T.D., and Wood, C.C. 1999. Application of microsatellite DNA variation to estimation of stock composition and escapement of Nass River sockeye salmon (Oncorhynchus nerka). Can. J. Fish. Aquat. Sci. 56: 1 14. Beacham, T.D., Margolis, L., and Nelson, R.J. 1998. A comparison of methods of stock identification for sockeye salmon (Oncorhynchus nerka) in Barkley Sound, British Columbia. N. Pac. Anad. Fish Comm. Bull. 1: 227 239. Beacham, T.D., Le, K.D., Raap, M.R., Hyatt, K., Luedke, W., and Withler, R.E. 2000a. Microsatellite DNA variation and estimation of stock composition of sockeye salmon, Oncorhynchus nerka, in Barkley Sound, British Columbia. Fish. Bull. (Wash., D.C.), 98: 14 24. Beacham, T.D., Wood, C.C., Withler, R.E., and Miller, K.M. 2000b. Application of microsatellite DNA variation to estimation of stock composition and escapement of Skeena River sockeye salmon (Oncorhynchus nerka). N. Pac. Anad. Fish Comm. Bull. 2: 263 276. Beacham, T.D., McIntosh, B., and MacConnachie, C. 2002. Microsatellite identification of individual sockeye salmon (Oncorhynchus nerka) in Barkley Sound, British Columbia. J. Fish Biol. 61: 1021 1032. Beacham, T.D., Lapointe, M., Candy, J.R., McIntosh, B., MacConnachie, C., Tabata, A., Kaukinen, K., Deng, L., Miller, K.M., and Withler, R.E. 2004. Stock identification of Fraser River sockeye salmon (Oncorhynchus nerka) using microsatellites and major histocompatibility complex variation. Trans. Am. Fish. Soc. 133: 1106 1126. Bernatchez, L., and Duchesne, P. 2000. Individual-based genotype analysis in studies of parentage and population assignment: how many loci, how many alleles? Can. J. Fish. Aquat. Sci. 57: 1 12. Burger, C.V., Spearman, W.T, and Cronin, M.A. 1997. Genetic differentiation of sockeye salmon from a geologically young Alaskan lake system. Trans. Am. Fish. Soc. 126: 926 938. Burgner, R.L. 1991. Life history of sockeye salmon Oncorhynchus nerka. In Pacific salmon life histories. Edited by C. Groot and L. Margolis. University of British Columbia Press, Vancouver. pp. 3 117. Cavalli-Sforza, L.L., and Edwards, A.W.F. 1967. Phylogenetic analysis: models and estimation procedures. Am. J. Hum. Genet. 19: 233 257. Cornuet, J.-M., Piry, S., Luikart, G., Estoup, A., and Solignac, M. 1999. New methods employing multilocus genotypes to select or exclude populations as origins of individuals. Genetics, 153: 1989 2000. Debevec, E.M., Gates, R.B., Masuda, M., Pella, J., Reynolds, J.M., and Seeb, L.W. 2000. SPAM (Version 3.2): statistics program for analyzing mixtures. J. Hered. 91: 509 510. Goudet, J. 2001. FSTAT: a program to estimate and test gene diversities and fixation indices. Version 2.9.3 [computer program]. Available from http://www.unil.ch/izea/softwares/fstat.html [cited 21 May 2004]. Kalinowski, S.T. 2002. How many alleles per locus should be used to estimate genetic distances? Heredity, 88: 62 65. Lewis, P.O., and Zaykin, D. 2001. GDA: computer program for the analysis of allelic data. Version 1.0 (d16c). Available from http://lewis.eeb.uconn.edu/lewishome/software.html [cited 7 January 2002]. McKinnell, S.M., Wood, C.C., Rutherford, D.T., Hyatt, K.D., and Welch, D.W. 2001. The demise of Owikeno Lake sockeye salmon. N. Am. J. Fish. Manag. 21: 774 791. Morris, D.B., Richard, K.R., and Wright, J.M. 1996. Microsatellites from rainbow trout (Oncorhynchus mykiss) and their use for genetic study of salmonids. Can. J. Fish. Aquat. Sci. 53: 120 126. Murray, C., and Wood, C. 2002. Status of Sakinaw Lake sockeye salmon (Oncorhynchus nerka). Canadian Science Advisory Secretariat Res. Doc. 2002/088. Available from http://www.dfompo.gc.ca/csas/csas/publications/resdocs-docrech/2002/2002_ 088_e.htm [cited 14 July 2004]. Nelson, R.J., and Beacham, T.D. 1999. Isolation and cross species amplification of microsatellite loci useful for study of Pacific salmon. Anim. Genet. 30: 228 229. Nelson, R.J., Wood, C.C., Cooper, G., Smith, C., and Koop, B. 2003. Population structure of sockeye salmon of the central coast of British Colulmbia: Implications for recovery planning. N. Am. J. Fish. Manag. 23: 703 720. Rannala, B., and Mountain, J.L. 1997. Detecting immigration by using multilocus genotypes. Proc. Natl. Acad. Sci. U.S.A. 94: 9197 9201. Rice, W.R. 1989. Analyzing tables of statistical tests. Evolution, 43: 223 225. Rutherford, D.T., and Wood, C.C. 2000. Trends in abundance and pre-season 2000 stock size forecasts for major sockeye, pink, and chum salmon stocks in northern British Columbia. Canadian Science Advisory Secretariat Res. Doc. 2000/114. Available from http://www.dfo-mpo.gc.ca/csas/csas/publications/ ResDocs-DocRech/2000/2000_114_e.htm [cited 14 July 2004]. Scribner, K.T., Gust, J.R., and Fields, R.L. 1996. Isolation and characterization of novel salmon microsatellite loci: crossspecies amplification and population genetic applications. Can. J. Fish. Aquat. Sci. 53: 833 841. Smith, C.T., Koop, B.F., and Nelson, R.J. 1998. Isolation and characterization of coho salmon (Oncorhynchus kisutch) microsatellites and their use in other salmonids. Mol. Ecol. 7: 1613 1621. Smouse, P.E., and Chevillon, C. 1998. Analytical aspects of population-specific DNA fingerprinting for individuals. J. Hered. 89: 143 150. Varnavskaya, N.V., Wood, C.C., and Everett, R.J. 1994. Genetic variation in sockeye salmon (Oncorhynchus nerka) populations of Asia and North America. Can. J. Fish. Aquat. Sci. 51(Suppl. 1): S132 S146. Waples, R.S. 1990. Temporal changes of allele frequency in Pacific salmon populations: implications for mixed-stock fishery analysis. Can. J. Fish. Aquat. Sci. 47: 968 976. Warner, B.G., Mathewes, R.W., and Clague, J.J. 1982. Ice-free conditions on the Queen Charlotte Islands, British Columbia, at the height of the late Wisconsin Glaciation. Science, (Wash., D.C.), 218: 675 677. Weir, B.S., and Cockerham, C.C. 1984. Estimating F-statistics for the analysis of population structure. Evolution, 38: 1358 1370. Winans, G.A., Aebersold, P.B., and Waples, R.S. 1996. Allozyme variability of Oncorhynchus nerka in the Pacific Northwest, with special consideration to populations of Redfish Lake, Idaho. Trans. Am. Fish. Soc. 125: 645 663. Withler, R.E, Le, K.D., Nelson, R.J., Miller, K.M., and Beacham, T.D. 2000. Intact genetic structure and high levels of genetic diversity in bottlenecked sockeye salmon, Oncorhynchus nerka,

844 Can. J. Zool. Vol. 83, 2005 populations of the Fraser River, British Columbia, Canada. Can. J. Fish. Aquat. Sci. 57: 1985 1998. Wood, C.C. 1995. Life history variation and population structure in sockeye salmon. Am. Fish. Soc. Symp. No. 17. pp. 195 216. Wood, C.C., Riddell, B.E., Rutherford, D.T., and Withler, R.E. 1994. Biochemical genetic survey of sockeye salmon (Oncorhynchus nerka) in Canada. Can. J. Fish. Aquat. Sci. 51(Suppl. 1): S114 S131.