(Oncorhynchus Mykiss) in the Nass and Skeena Rivers in Northern British Columbia

Mar. Biotechnol. 2, 587–600, 2000 DOI: 10.1007/s101260000045

Microsatellite DNA Population Structure and Stock Identification of Steelhead Trout (Oncorhynchus mykiss) in the Nass and Skeena Rivers in Northern British Columbia

Terry D. Beacham,1,* Susan Pollard,2 and Khai D. Le1

1Department of Fisheries and Oceans, Science Branch, Pacific Biological Station, Nanaimo, BC., V9R 5K6, Canada 2Ministry of Fisheries, 780 Blanshard Street, Victoria, BC., V8V 1X4, Canada

Abstract: Population structure and the application to genetic stock identification for steelhead (Oncorhynchus mykiss) in the Nass and Skeena Rivers in northern British Columbia was examined using microsatellite markers. Variation at 8 microsatellite loci (Oki200, Omy77, Ots1, Ots3, Ssa85, Ots100, Ots103, and Ots108) was surveyed for approximately 930 steelhead from 7 populations in the Skeena River drainage and 850 steelhead from 10 populations in the Nass River drainage, as well as 1550 steelhead from test fisheries near the mouth of each river. Differentiation among populations within rivers accounted for about 1.9 times the variation observed among years within populations, with differences between drainages less than variation among populations within drainages. In the Nass River, winter-run populations formed a distinct group from the summer-run populations. Winter-run populations were not assessed in the Skeena River watershed. Simulated mixed-stock samples suggested that variation at the 8 microsatellite loci surveyed should provide relatively accurate and precise estimates of stock composition for fishery management applications within drainages. In the Skeena River drainage in 1998, Babine River (27%) and Bulkley drainage populations (31%) comprised the main components of the returns. For the Nass River in 1998 steelhead returning to Bell-Irving River were estimated to have comprised 39% of the fish sampled in the test fishery, with another 27% of the returns estimated to be derived from Cranberry River. The survey of microsatellite variation did not reveal enough differentiation between Nass River and Skeena River populations to be applied confidently in estimation of stock composition in marine fisheries at this time.

Key words: British Columbia, microsatellite loci, population structure, steelhead, stock composition.

INTRODUCTION in British Columbia. There are no directed commercial fisheries for steelhead in British Columbia, and significant ef- Steelhead (Oncorhynchus mykiss), the anadromous form of forts are made to reduce the bycatch of steelhead in com- rainbow trout, are found in all major coastal river systems mercial salmon fisheries because this species is considerably less abundant naturally than all other Pacific salmon spe- Received January 14, 2000; accepted July 13, 2000. *Corresponding author: telephone 250-756-7149; fax 250-756-7053; e-mail cies. This low level of abundance can lead to special and [email protected] often contentious measures for steelhead conservation to 588 Terry D. Beacham et al. reduce exploitation in fisheries. One area where there MATERIALS AND METHODS has been substantial concern for steelhead is the Skeena River drainage in northern British Columbia. Measures Collection of DNA Samples and Polymerase have been taken to reduce steelhead exploitation dur- Chain Reaction ing fisheries for the more abundant sockeye salmon (O. nerka), which have run timing very similar to that For the characterization of the baseline populations, DNA of summer-run steelhead in the lower river. Run timing was extracted from adipose fin clips of adult steelhead pre- for specific steelhead populations is uncertain, and there served in 90% ethanol. The main method of sampling is currently no method to identify the presence of spe- adults was by angling, although enumeration fences were cific populations during their passage through the lower used at some locations (Sustut, Babine, and Toboggan in river. Skeena River drainage) in some years. The Sustut River Determination of population structure of exploited samples were derived mainly from the upper Sustut drain- species is an essential component in successful management age, but 13 fish from the lower Sustut drainage were also of fisheries. Specifically, this information can be used for included in the 1997 samples. Juveniles were included in the applications ranging from the determination of appropriate analysis for the 1997 Bell-Irving River sample (61 of 96 fish conservation units to estimation of stock composition in analyzed). Approximately 930 steelhead were sampled from mixed-stock fisheries. In British Columbia, surveys of ge- 7 Skeena River populations and 850 steelhead from 10 Nass netic variation have been used to describe population struc- River populations, with most of the populations sampled in ture in steelhead (Parkinson, 1984a; Taylor, 1995; Beacham at least 2 years (Table 1 and Figure 1). All populations in the et al., 1999), and can also be used to estimate stock com- Skeena River were summer run. In the Nass River, 5 sum- position in mixed-fishery samples (Parkinson, 1984b; mer-run and 5 winter-run populations were surveyed Beacham et al., 1999). (Table 1). Laboratory methods detailing DNA extraction Microsatellites are rapidly developing into a tool procedures, details of the 8 loci amplified (Omy77, Oki200, of choice to survey genetic variation among salmonid Ots1, Ots3, Ots100, Ots103, Ots108 and Ssa85), and details populations. Nonlethal sampling and the abundance of polymerase chain reaction (PCR) were outlined in of loci make this technology very effective for describing Beacham et al. (1999). population structure in Pacific salmon (Beacham et Fishery samples were collected in both the Nass and al., 1998; Seeb et al., 1998; Small et al., 1998) and steel- Skeena Rivers. In the lower Nass River, fish wheels were head (Nielsen et al., 1994, 1997; Wenburg et al., 1996; used to conduct the 1998 test fishery, with general details of Ostberg and Thorgaard, 1999). Microsatellite DNA loci the test fishery outlined by Link and Gurak (1997). All can also be very useful in estimating stock compo- samples analyzed were derived from 2 fish wheels. In the sition in mixed-stock salmon fisheries (Beacham and lower Skeena River, samples were collected from steelhead Wood, 1999) and have been previously applied to steelhead caught in the Tyee gillnet test fishery in 1996 and 1998 near caught incidentally in commercial salmon fisheries the mouth of the river, with details of the test fishery out- (Beacham et al., 1999). We evaluated microsatellite varia- lined by Jantz et al. (1990). All samples collected in 1996 tion among steelhead populations in the Nass River and were analyzed, whereas in 1998 samples were generally ana- Skeena River drainages to address stock identification lyzed in proportion to run abundance, with approximately questions. 50% of the fish captured in the test fishery surveyed for The objectives of the present study were to determine microsatellite variation. population structure of steelhead in the Nass and Skeena Rivers using microsatellite variation and evaluate potential Gel Electrophoresis and Band Analysis applications for stock identification within each watershed. We also evaluated whether microsatellite variation can pro- PCR products were size fractionated on 16-cm by 17-cm vide accurate and precise estimates of stock composition for nondenaturing polyacrylamide gels visualized by staining steelhead from the two watersheds when they occur to- with 0.5 mg/ml ethidium bromide in water and illuminat- gether in marine mixed-stock fisheries. We then estimated ing with ultraviolet light. Nelson et al. (1998) and Beacham stock composition of steelhead caught in test fisheries in the et al. (1999) provided a more complete description of gel lower portions of the two rivers. electrophoretic conditions and estimation of allele size, as Steelhead Population Structure and Stock ID 589

Table 1. Steelhead Samples Collected and Analyzed from 7 Skeena River and 10 Nass River Populations in Northern British Columbia, and From a Test Fishery Near the Mouth of Each River*

Population Years sampled N* Total N Skeena River summer run 1. Sustut River 1994, 1996, 1997, 1998 27, 50, 84, 50 211 2. Babine River 1991, 1992, 1995, 1996, 1997, 1998 19, 18, 39, 31, 29, 128 264 3. Bulkley River 1995, 1996, 1997 7, 36, 20 63 4. Morice River 1991, 1992, 1995, 1998 20, 30, 15, 46 111 5. Toboggan Creek 1998 128 128 6. Kispiox River 1992, 1995, 1998 20, 30, 35 85 7. Zymoetz River 1993, 1995, 1997 16, 18, 38 72 8. Test fishery 1996, 1998 190, 926 1116 Nass River summer run 9. Bell-Irving River 1997, 1998 96, 92 188 10. Damdochax Creek 1994, 1995, 1996, 1997, 1998 27, 18, 36, 20, 47 148 11. Meziadin River 1996, 1997, 1998 9, 5, 40 55 12. Kwinageese River 1995, 1997, 1998 11, 8, 92 111 13. Cranberry River 1995, 1996, 1997 19, 29, 147 195 14. Test fishery 1998 436 436 Nass River winter run 15. Chambers Creek 1997, 1998 24, 24 48 16. Ishkheenickh River 1997, 1998 14, 19 33 17. Kwinamass River 1997, 1998 17, 4 21 18. Tseax River 1998 16 16 19. Kincolith River 1996, 1997, 1998 16, 15, 10 41

*Sample sizes are for years sampled.

well as a figure of a gel. Precision of estimated allele size, GENEPOP. The dememorization number was set at 1000, evaluated with the standard fish analyzed for each locus, and 50 batches were run for each test with 1000 iterations was similar to that outlined by Beacham et al. (1999). per batch. Critical significance levels for simultaneous tests were evaluated with the sequential Bonferroni adjustment Data Analysis (Rice, 1989). A neighbor-joining analysis illustrating genetic relationships among populations was conducted with Each population at each locus was tested for departure from PHYLIP (Felsenstein, 1993). The allele frequency matrix Hardy-Weinberg equilibrium (HWE) using GENEPOP was resampled 1000 times, and Cavalli-Sforza and Edwards Version 3.1 (Raymond and Rousset, 1995), as was temporal (1967) chord distance was used to estimate distance among stability of allele frequencies. For populations sampled in populations. FST values for each locus were determined with multiple years, HWE was tested on samples pooled over FSTAT (Goudet, 1995). Estimation of variance components years if no significant annual variation in allele frequencies in allele frequencies among populations and years was de- was observed. For those populations in which significant termined with BIOSYS (Swofford and Selander 1981). annual variation in allele frequencies was detected at any Negative variance components were set to 0 in estimation of locus, HWE was tested on an annual basis at all loci in the relative diversity. population. In this case, the lowest probability of conform- ance to HWE in the annual tests was considered. Tests of Estimation of Stock Composition genetic differentiation with pairwise comparisions between The model of Fournier et al. (1984) was used to estimate all population pairs at each locus were also conducted using stock composition, and genotypic frequencies were deter- 590 Terry D. Beacham et al.

termined, with the whole process repeated 100 times to estimate the mean and standard deviation of the individual stock composition estimates. Estimated stock composition of the test fishery samples was determined as a point estimate of the period sample, with standard deviations of individual stock estimates derived from 100 bootstrap re- samplings of both the baseline populations and the period sample.

RESULTS

Variation Within Populations

Observed heterozygosities for loci in the summer-run steelhead populations in the Skeena River drainage were as follows: Oki200, 0.60 (population range, 0.54–0.66); Omy77, 0.66 (0.57–0.76); Ots1, 0.55 (0.35–0.68); Ots3, 0.63 (0.56– 0.70); Ots103, 0.57 (0.48–0.65); Ots100, 0.79 (0.75–0.86); Ots108, 0.76 (0.55–0.95); and Ssa85, 0.81 (0.63–0.98). Ob- served heterozygosities of the Nass River drainage populations were Oki200, summer run 0.65 (0.52–0.69) and winter run 0.70 (0.53–0.76); Omy77, summer 0.59 (0.51–0.65) and winter 0.62 (0.56–0.75); Ots1, summer 0.58 (0.48–0.66) and winter 0.34 (0.13–0.55); Ots3, summer 0.63 (0.56–0.69) and Figure 1. Locations of steelhead populations and test fisheries winter 0.51 (0.40–0.63); Ots103, summer 0.50 (0.41–0.57) sampled in the Skeena River and Nass River. Numbers and loca- and winter 0.39 (0.27–0.55); Ots100, summer 0.73 (0.69– tions are indicated in Table 1. 0.82) and winter 0.72 (0.60–0.85); Ots108, summer 0.57 (0.44–0.77) and winter 0.65 (0.40–0.83); and Ssa85, summer 0.66 (0.53–0.76) and winter 0.60 (0.43–0.80). There mined at each locus for each population. Each locus was was no consistent difference in observed heterozygosities assumed to be in HWE, and expected genotypic frequencies between summer-run and winter-run populations in the were determined from the observed allele frequencies and Nass River, nor was there any consistent difference in het- used as model inputs. The use of expected genotypic fre- erozygosities between Skeena and Nass River populations. quencies reduces the possibility of observing genotypes in Significant annual variation in allele frequencies (cor- the mixture that have not been observed in the baseline rection for 8 tests per population, ␣ = 0.0063) was observed samples. Each baseline population was resampled with re- in at least one locus in 10 of the 15 populations that had placement in order to simulate random variation involved been sampled in multiple years. Most of the variation was in the collection of the baseline samples before the estima- observed at the Oki200, Omy77, and Ssa85 loci. In the tion of stock composition of each mixture. Mixture com- Skeena River drainage, the Babine River population had the positions typically expected in both Nass River and Skeena highest degree of annual variation (3 tests significant), and River steelhead were examined in order to evaluate accuracy in the Nass River drainage, the Damdochax, Kwinageese, and precision of the estimated stock compositions. Hypo- and Bell-Irving populations had the highest degree of varia- thetical fishery samples of either 100 or 150 fish were gen- tion (3 tests significant). In the Bell-Irving River popula- erated by randomly resampling with replacement the base- tion, 35 adults and 61 juveniles were sampled during 1997. line populations in each drainage, and adding the appro- No significant difference in allele frequencies between the priate number of fish from each population to the mixture. adults and the juveniles was observed at any locus in that Estimated stock composition of the mixture was then de- year. Steelhead Population Structure and Stock ID 591

As there was some annual variation in allele frequen- frequencies were observed compared with the summer-run cies, evaluation of HWE of genotypic frequencies was con- populations, again presumably owing to smaller sample ducted on an annual basis within each population. Geno- sizes for the winter-run populations (Table 1). However, typic frequencies observed in the 17 populations and 8 loci significant differentiation in allele frequencies (5 popula- were those generally expected under HWE. Significant de- tions, 10 comparisons, ␣ = 0.0050) was observed at one or partures (correction for 8 tests per population, ␣ = 0.0063) more loci for all comparisons except the Chambers- were observed in the 1998 Babine River (4 tests significant) Kwinamass and Kincolith-Kwinamass comparisons. No and 1998 Morice River (3 tests significant) samples in the more than 4 significant differences were observed for any Skeena drainage. In the Nass River drainage, significant de- comparison involving winter-run populations. partures were observed in the 1997 and 1998 (4 tests significant each year) Bell-Irving River samples. Population Structure

Variation Among Populations Hierarchical gene diversity analysis of steelhead from the Nass and Skeena Rivers indicated that 95.7% of variation Genetic differentiation was observed between summer-run occurred within populations (Table 2). Differentiation populations in the Nass and Skeena drainages, and between among populations within rivers accounted for about 1.9 summer-run and winter-run populations in the Nass drain- times the variation observed among years within popula- age. For example, the frequency of Omy77116 was 0.40 in the tions, with differences between drainages less than variation Skeena River populations, and 0.30 in the summer-run among populations within drainages or variation among Nass River populations (Figure 2). Larger differences in years. The greatest structuring among populations within allele frequencies were observed between summer-run and rivers was observed at Ots3 (5.3%), Omy77 (2.8%), and winter-run populations in the Nass River than between Ots100 (2.7%). Observed FST values in the Skeena drainage summer-run populations in the two drainages. For ex- were as follows: Oki200, 0.039 (SD, 0.023); Omy77, 0.036 ample, the frequency of Ots100168 was less than 0.10 in the (0.016); Ots1, 0.021 (0.010); Ots3, 0.032 (0.019); Ots100, summer-run populations in both drainages, but about 0.30 0.019 (0.006); Ots103, 0.008 (0.003); Ots108, 0.022 (0.007); in the winter-run populations. Similar differentiation at and Ssa85, 0.031 (0.014); the mean value (SD) over all loci

Ots3 was observed in summer-run and winter-run popula- was 0.026 (0.003). Observed FST values among the 10 popu- tions (Figure 2). lations in the Nass drainage were as follows: Oki200, 0.016 Pairwise comparisons of population allele frequencies (0.007); Omy77, 0.029 (0.008); Ots1, 0.019 (0.008); Ots3, within each drainage were conducted to evaluate differen- 0.033 (0.019); Ots100, 0.024 (0.009); Ots103, 0.011 (0.006); tiation among populations. In the Skeena River, there was a Ots108, 0.016 (0.008); and Ssa85, 0.019 (0.008); the mean general differentiation among all populations sampled (sig- value (SD) over all loci was 0.024 (0.003). nificant differentiation, P < .05, at a minimum of 5 of 8 loci Population structure within the Skeena and Nass Rivers surveyed) with the exception of the comparison between was evident. In the Skeena River, the Morice, Bulkley, and the Bulkley River and Morice River population, where only Toboggan populations formed a fairly distinct group, clus- 3 of 8 comparisons (Ots100, Ots108, Ssa85) were significant tering together 41% of the time in the 1000 trees used to (P < .05). In the Nass River, virtually all comparisons be- create the consensus tree (Figure 3). These populations are tween summer-run and winter-run populations were sig- all from the Bulkley River watershed. However, the Sustut nificant. Within the summer-run populations, significant River population in the Skeena drainage grouped with other differentiation in allele frequencies (5 populations, 10 com- summer-run populations of the Nass River. In the Nass parisons, ␣ = 0.0050) were observed at 4 or more loci when River, winter-run populations formed a distinct group from the Meziadin River population was excluded. When the the summer-run populations, with the 5 populations clus- Meziadin River population was included, 1 or 2 significant tering together 92% of the time. In general, the summer- differences were observed between Meziadin River and each run and winter-run populations of the Nass River were the of the other populations, probably owing to the smaller most distinct groups, with the summer-run populations of sample size available for this population. Among the win- the Skeena intermediate between the two Nass River ter-run populations, fewer significant differences in allele groups. 592 Terry D. Beacham et al.

Figure 2. Allele frequencies of summer-run steelhead in the Skeena and Nass rivers and winter-run Nass River steelhead at 8 microsatellite loci. N is the number of steelhead sampled in each group. For all loci, alleles were des- ignated by the lower limit (bp) of the allele bin used to define the alleles. Bin widths generally corresponding to a re- peat unit were set with the main allele occurring in the middle of the bin.

Estimation of Stock Composition Within the veloped that could be representative of the relative abun- Skeena River Watershed dance of spawning escapements within the drainage. As the Babine River and Bulkley River populations are considered Simulated Mixtures to be the major populations within the Skeena drainage, the We evaluated whether the level of genetic differentia- combined proportion of these populations in the simulated tion observed among steelhead populations in the Skeena samples was at least 50% of the sample in evaluations. Mean River drainage can be applied to practical issues of stock estimated stock compositions of the simulated samples were identification. Two simulated test fishery mixtures were de- usually within 2% of the true means for most populations Steelhead Population Structure and Stock ID 593

Table 2. Hierarchical Gene-Diversity Analysis of 7 Skeena River and 10 Nass River Populations of Steelhead for Eight Microsatellite DNA Loci*

Absolute diversity Relative diversity Within Within Among years within Among populations Among river Locus Total populations populations populations within river drainages drainage Oki200 0.7324 0.6917 0.9444 0.0215 0.0249 0.0091 Omy77 0.8156 0.7762 0.9517 0.0177 0.0276 0.0031 Ots1 0.5996 0.5877 0.9802 0.0002 0.0196 0.0000 Ots3 0.6669 0.6248 0.9369 0.0097 0.0534 0.0000 Ssa85 0.8129 0.7891 0.9619 0.0154 0.0227 0.0000 Ots100 0.8316 0.7914 0.9517 0.0201 0.0272 0.0010 Ots103 0.4619 0.4501 0.9745 0.0004 0.0199 0.0052 Ots108 0.9255 0.8930 0.9649 0.0132 0.0203 0.0016 All 0.9571 0.0140 0.0269 0.0020

*River drainage, populations with drainages, and years within populations are outlined in Table 1.

Figure 3. Unrooted neighbor-joining tree outlining relationships of 17 steelhead populations. Bootstrap values at the tree nodes indicate the percentage of 1000 trees where populations beyond the node occurred together. The number of the population (from Figure 1 and Table 1) is indicated next to the name.

(Table 3). The estimated proportions of Babine River and conservation measures directed toward coho salmon (O. Bulkley or Morice steelhead were estimated with a reason- kisutch), steelhead abundance increased substantially in the able degree of accuracy, reflecting the relative genetic dis- Skeena River, allowing for a more detailed analysis of run tinctiveness of these groups. The simulations indicated that timing. The Babine River (27%) and Bulkley drainage accurate estimates of stock composition should be derived populations (31%) constituted the main components of the when applied to samples derived from the Skeena River returns (Table 5). The Sustut River population tended to drainage. have an earlier timing of return in both years than did the other populations surveyed. The peak of the return in 1998 Application to the Test Fishery occurred during the week of July 26 through August 1. The Bulkley and Morice populations were relatively abundant at In 1996, steelhead returning to Babine River were es- this time, with a second peak of abundance in the later timated to have comprised 36% of the fish sampled in the portions of the run, which would be expected if the test test fishery, with another 27% of the returns estimated to be fishery sampled the early-running Morice River population derived from Bulkley River drainage populations (Bulkley, and the later-running downstream Bulkley River mainstem Morice, and Toboggan) (Table 4). In 1998, owing to fishery population. 594 Terry D. Beacham et al.

Table 3. Estimated Percentage Compositon of Two Mixtures of Skeena River Steelhead in Simulations Using Observed Variation at Eight Microsatellite DNA Loci.*

Mixture 1 Mixture 2 Stock Correct Estimated (SD) Correct Estimated (SD) Babine 30.0 31.3 (5.7) 20.0 23.5 (3.9) Kispiox 10.0 9.7 (3.5) 10.0 9.6 (3.2) Sustut 10.0 10.4 (2.2) 15.0 15.3 (2.0) Zymoetz 10.0 8.8 (2.5) 10.0 9.3 (2.2) Toboggan 20.0 19.2 (4.5) 15.0 14.4 (3.1) Bulkley 10.0 9.3 (3.6) 25.0 19.3 (3.5) Morice 10.0 11.3 (4.3) 5.0 8.6 (3.3) ⌺ Bulkley/Morice 20.0 20.6 (4.8) 30.0 28.0 (3.6)

*Each mixture of 150 fish was generated 100 times with replacement, and stock compositions of the mixtures estimated by resampling each baseline stock with replacement to obtain the original sample size and a new distribution of allele frequencies.

Table 4. Estimated Percentage Stock Compositon of the 1996 Skeena River Steelhead Tyee Test Fishery Samples for Three Periods in 1996.

Percentage stock composition (SD) Stock Prior July 24 July 25–Aug 19 After Aug 20 Seasonal N* 28 128 34 190 Babine 31.4 (12.5) 32.8 (8.2) 23.5 (14.2) 36.0 (3.6) Kispiox 7.6 (6.5) 10.7 (5.5) 16.2 (11.5) 11.3 (2.3) Sustut 37.5 (11.1) 16.7 (4.9) 12.8 (6.8) 15.8 (2.4) Zymoetz 0.0 (1.5) 13.4 (4.0) 9.0 (9.2) 10.2 (2.3) Toboggan 21.6 (10.1) 7.2 (5.1) 4.1 (6.5) 11.5 (3.5) Bulkley 1.9 (7.0) 15.8 (5.5) 34.4 (11.5) 14.9 (2.9) Morice 0.0 (6.3) 3.4 (5.0) 0.0 (4.9) 0.4 (2.6) ⌺ Bulkley/Morice 1.9 (9.4) 19.2 (6.0) 34.4 (12.0) 15.3 (2.9)

*Number of fish analyzed in each period.

Estimation of Stock Composition within the Nass means for most populations, although there was some ten- River Watershed dency for the proportion of the Cranberry River population to be slightly overestimated and the proportion of the Simulated Mixtures Meziadin River population to be slightly underestimated As the test fishery was conducted only during the time (Table 6). The simulations indicated that reasonably accu- and location when summer-run populations would be ex- rate estimates of stock composition should be derived when pected to return to the Nass River, the analysis was re- applied to samples derived from the Nass River drainage. stricted to the 5 summer-run populations only. Three simulated test fishery mixtures were developed that could be Application to the Test Fishery representative of the relative abundance of spawning escapements within the drainage. Mean estimated stock com- In 1998, steelhead returning to Bell-Irving River were positions of the samples were usually within 4% of the true estimated to have comprised 39% of the fish sampled in the Steelhead Population Structure and Stock ID 595

test fishery, with another 27% of the returns estimated to be derived from Cranberry River (Table 7). The timing of return of the Damdochax Creek and Meziadin River populations was earlier than for other populations, and the timing of the Cranberry River population was somewhat later. The Bell-Irving population timing was spread throughout the sampling period.

Comparisons Between Nass River and Skeena River

Three simulated samples of Nass River and Skeena River steelhead were constructed and resolved with a 12- stock summer-run Nass and Skeena baseline to evaluate the accuracy and precision of estimated stock compositions for possible marine applications. The levels of accuracy and precision observed in the simulated mixtures indicated that some degree of bias may be present in the estimated stock compositions, particularly if either the Nass or Skeena con- tributions exceeded 80% of the sample, and that the level of precision of the estimated stock compositions was relatively low (Table 8). The bias and lack of precision in the estimated stock compositions suggests that no marked separation exists between summer-run populations in the two drainages, at least for the 8 loci surveyed (Figure 3).

DISCUSSION

In British Columbia, steelhead populations are small relative to other Pacific salmon populations, and obtaining representative samples from a population can be a challenging task. Samples obtained from juveniles may be nonrepresen- tative of the population because of family sampling (Hansen et al., 1997; Wenburg et al., 1998; Mj␾lner␾d et al., 1999), inclusion of rainbow trout in the sample (Parkinson, 1984a), or perhaps juveniles rearing in nonnatal streams, as in the case of chinook salmon (O. tshawytscha) (Murray and Rosenau, 1989; Scrivener et al., 1994). We generally limited sampling to adults, but obtaining adult samples on many of the rivers surveyed was quite difficult because of the relatively low population abundance and remoteness of the spawning locations in the drainage. In some cases our Percentage stock composition (SD) July 9–18 July 19–25 July 26–Aug 1 Aug 2–8 Aug 9–15 Aug 16–22 Aug 23–29 Aug 30–Sept 5 Sept 6–12 Sept 13–27 Seasonal samples may not have been derived from a single breeding population. For example, the Bulkley River sample was derived from angling in the mainstem of the Bulkley River Estimated Percentage Stock Compositon of the 1998 Skeena River Steelhead Tyee Test Fishery Samples for 11 Periods in 1998. prior to spawning, and it is quite likely that fish sampled in this location were a mixture of steelhead likely to spawn in Bulkley/Morice 21.4 (7.4) 29.2 (6.9) 24.5 (5.9) 8.6 (4.5) 12.0 (10.0) 34.4 (9.4) 31.9 (13.9) 26.7 (12.5) 22.3 (16.8) 63.9 (16.7) 23.8 (3.9) N*BabineKispioxSustutZymoetz 22.3Toboggn 69 (8.5)Bulkley 8.7 (7.7)Morice 27.3 (6.2) 19.5 (5.6) 17.7 (6.3)⌺ 0.8 (4.1) 21.5 6.8 (5.7) 190 (4.5) 13.0 24.7 (5.4) (4.3) 10.9 (2.9) 22.4 (5.5) 8.4 12.4 (6.1) (2.9) 10.7 (5.4) 44.1 8.8 (6.9) (2.9) 234 21.6 (5.2) 10.3 18.9 (5.9) (10.3) 22.2 (7.2) 13.0 27.8 (4.7) (8.5) 7.5 9.6 36.9 (5.2) (3.1) (11.8) 10.0 (4.2) 29.1 10.3 18.1 (13.4) (4.9) (8.6) 11.5 13.6 (4.4) (5.8) 5.5 (4.5) 177 40.1 (15.5) 13.9 8.6 (9.3) 8.8 (4.2) (4.5) 8.4 6.3 (7.7) (4.5) 10.8 6.1 (11.6) (8.3) 51.8 0.0 (19.0) 3.5 (3.0) 4.4 (5.7) 7.3 (7.4) (6.5) 64 28.5 (9.2) 7.0 5.9 (13.1) (7.6) 14.2 13.1 (10.9) 5.1 (9.6) (8.6) 6.0 (12.3) 26.1 26.7 (13.3) (4.0) 5.9 4.1 (6.8) (8.1) 9.3 13.1 (9.3) (12.3) 78 19.4 (10.2) 0.1 5.8 (2.0) 17.2 (8.2) (3.2) 19.5 1.5 (13.6) (3.6) 0.4 (6.9) 13.6 (12.2) 5.8 (9.3) 46.6 (19.1) 39 10.6 12.5 (2.0) (13.4) 14.4 16.9 (1.8) (3.4) 2.8 7.3 (11.1) (3.0) 17.3 (17.1) 29 6.9 (2.6) 25 21 828† Table 5. Stock *Number of fish analyzed in†The each seasonal period. estimate of 828 fish is based sampling in proportion to run abundance. the Bulkley mainstem, other Bulkley River tributaries, and 596 Terry D. Beacham et al.

Table 6. Estimated Percentage Compositon of Three Mixtures of Nass River Summer Steelhead Populations in Simulations Using Observed Variation at Eight Microsatellite DNA Loci.*

Mixture 1 Mixture 2 Mixture 3 Stock Correct Estimated (SD) Correct Estimated (SD) Correct Estimated (SD) Cranberry 25.0 28.3 {5.9) 20.0 25.7 (6.7) 30.0 33.3 (7.1) Damdochax 35.0 34.9 (4.9) 20.0 21.1 (5.3) 10.0 12.9 (4.4) Kwinageese 15.0 12.3 (6.6) 20.0 17.1 (7.1) 15.0 13.3 (6.1) Meziadin 10.0 7.2 (3.7) 10.0 7.6 (4.6) 5.0 4.7 (3.6) Bell-Irving 15.0 17.3 (6.3) 30.0 28.5 (6.9) 40.0 35.8 (7.3)

*Each mixture of 100 fish was generated 100 times with replacement, and stock compositons of the mixtures were estimated by resampling each baseline stock with replacement to obtain the original sample size and a new distribution of allele frequencies.

Table 7. Estimated Percentage Stock Compositon of the 1998 Nass River Steelhead Test Fishery Samples for Six Periods in 1998.

Percentage stock composition (SD) Stock July 22–Aug 8 Aug 9–22 Aug 23–29 Aug 30–Sept 5 Sept 6–12 Sept 13–Oct 13 Seasonal N* 40 96 70 60 90 65 421 Cranberry 0.0 (9.3) 22.6 (6.4) 27.7 (7.7) 20.2 (9.8) 34.2 (9.8) 25.8 (11.0) 27.4 (4.7) Damdochax 21.7 (9.9) 27.4 (6.3) 14.1 (5.9) 10.7 (8.6) 15.8 (7.6) 10.2 (8.4) 16.5 (4.7) Kwinageese 33.4 (12.1) 13.1 (6.2) 15.5 (7.1) 22.8 (11.4) 9.6 (8.4) 13.7 (9.6) 15.5 (5.5) Meziadin 10.2 (7.7) 8.5 (5.0) 2.1 (3.8) 6.1 (7.9) 0.6 (3.1) 4.6 (7.9) 1.3 (3.2) Bell-Irving 34.7 (15.9) 28.4 (7.4) 40.6 (8.9) 40.2 (13.5) 39.7 (11.6) 45.7 (13.7) 39.3 (5.4)

*Number of fish analyzed in each period. the major upstream tributary, the Morice River (Figure 1). Washington (Beacham et al., 1999), but certainly less than Similarly, the Babine River sample may have been a mixture has been observed in sockeye salmon (O. nerka) in British of steelhead spawning in the lower Babine River and tribu- Columbia, with population differentiation 7.4 to 11.8 times taries. In the Nass River, the Bell-Irving tributary covers a greater than variation within populations (Beacham and relatively large drainage area, and it is possible that there is Wood, 1999; Beacham et al., 2000a). Relatively higher levels more than one discrete breeding population in the tribu- of annual variation observed in steelhead as compared with tary. sockeye salmon may reflect smaller effective population Significant annual variation in allele frequencies was size, but, as the number of steelhead surveyed in most years observed in at least one locus in 10 of 15 populations in in the comparisons was relatively small, inadequate sam- which comparisons were possible. Annual variation in allele pling may have contributed substantially to the observed frequencies at genetic markers in steelhead was observed annual variation, particularly given the large number of previously with allozymes (Chilcote et al., 1980; Parkinson, alleles observed at some loci. 1984a; Reisenbichler and Phelps, 1989; Reisenbichler et al., There was no evidence in our study of genotypic fre- 1992) and microsatellites (Beacham et al., 1999). In our quencies at a locus being consistently different from those study, population differentiation within river drainages was expected under HWE, thus reflecting the presence of a null about 2.0 times greater for all loci than annual variation allele; however, significant nonconformance to HWE was within populations. This was comparable to the relative observed at 3 or 4 loci in some populations in some years. levels of differentiation among steelhead populations within In these instances (Bulkley, Babine, and Bell Irving), sig- broad geographic regions in southern British Columbia and nificant differences were most likely a result of samples Steelhead Population Structure and Stock ID 597

Table 8. Estimated Percentage Composition of Three Mixtures of Skeena and Nass River Summer Steelhead in Simulations Using Observed Variation at Eight Microsatellite DNA Loci.*

Mixture 1 Mixture 2 Mixture 3 Stock Correct Estimated (SD) Correct Estimated (SD) Correct Estimated (SD) Babine 20.0 19.5 (6.4) 20.0 18.5 (6.5) 10.0 10.3 (4.1) Kispiox 15.0 13.1 (3.5) 15.0 12.8 (4.7) 10.0 9.2 (4.1) Sustut 10.0 8.6 (2.2) 10.0 9.3 (3.0) 0.0 2.2 (2.5) Zymoetz 20.0 17.5 (4.0) 0.0 5.5 (7.7) 0.0 0.4 (0.7) Toboggan 20.0 18.2 (3.9) 0.0 6.3 (7.8) 0.0 1.6 (2.5) Bulkley 0.0 1.1 (2.0) 20.0 10.1 (6.7) 0.0 1.9 (2.4) Morice 0.0 2.5 (3.3) 0.0 3.1 (3.4) 0.0 1.6 (2.4) ⌺ Skeena 85.0 80.5 (6.3) 65.0 65.7 (11.6) 20.0 27.1 (5.4) Cranberry 10.0 9.6 (5.5) 20.0 18.1 (7.5) 20.0 21.8 (5.3) Damdochax 0.0 1.5 (2.1) 0.0 2.3 (3.1) 20.0 18.7 (4.0) Kwinageese 0.0 2.4 (3.2) 10.0 6.4 (4.7) 10.0 8.9 (5.7) Meziadin 0.0 1.0 (1.8) 0.0 1.4 (2.2) 10.0 5.4 (4.3) Bell-Irving 5.0 5.0 (4.4) 5.0 6.2 (5.4) 20.0 18.1 (6.8) ⌺ Nass 15.0 19.5 (6.3) 35.0 34.3 (11.6) 80.0 72.9 (5.4)

*Each mixture of 100 fish was generated 100 times with replacement, and stock compositions of the mixtures were estimated by resampling each baseline stock with replacement to obtain the original sample size and a new distribution of allele frequencies. being composed of mixtures of distinct spawning popula- with an allozyme study by Chilcote et al. (1980), which tions. found annual variation within populations to be of the Previous surveys of genetic population structure of same magnitude as variation between summe-run and win- steelhead in northern British Columbia have been restricted ter-run steelhead from the Kalama River, Washington. to 2 studies. Parkinson (1984a) surveyed variation at 5 al- However, in this case, populations occurred sympatrically lozme loci for approximately 110 juveniles from 6 Skeena and genetic exchange was more likely than in the Nass River populations. Skeena River populations were distinct River, where the winter-run populations are located in the from those in southern British Columbia, but there was lower river drainage and summer-run populations in more limited evidence for population structure within the drain- upstream locations. There was limited differentiation be- age. In a study that included a small number of populations tween summer-run components in the Skeena and Nass from the north coast of British Columbia, Taylor (1995) drainages, with only 5 of the 8 loci surveyed of any value in surveyed variation at 2 minisatellite loci in the Babine, Mo- discriminating between the two drainages. Several Nass rice, and Sustut Rivers. Highly significant differences were River tributaries are geographically close to upper Skeena observed for at least 1 locus between each pair of popula- River tributaries. Given the proximity of the two drainages, tions, and variation between sampling years was not evi- and the fact that the Sustut River population is more similar dent. genetically to Nass River than to other Skeena River popu- Our survey of variation at 8 microsatellite loci indi- lations, one could speculate that the founding population of cated that significant genetic differentiation existed among the Sustut River was possibly derived from the Nass River spawning populations within the Skeena River drainage, drainage after the last glaciation. However, similar structure and that the Bulkley River drainage populations (Bulkley, does not occur in sockeye salmon, with all upper Skeena Morice, and Toboggan) were distinct from other popula- populations distinct from Nass populations (Beacham et al., tions. Our survey is the first ever conducted of genetic varia- 2000b). No transplants of steelhead from the Nass River tion of Nass River steelhead, and our results indicated con- into the Sustut River have been recorded, so it may be that siderable genetic separation between summer-run and win- the present population structure is indicative of divergence ter-run populations in the drainage. This finding contrasts from a single postglacial founding population. 598 Terry D. Beacham et al.

Issues of interest in comparing techniques for applica- from observed escapements, but may represent the com- tions to estimate stock composition in mixed-stock fisheries bined contribution of genetically similar unsampled popu- include evaluation of the relative differentiation among lations in the Bulkley River drainage. Migration timing of populations in the fishery sample and the level of annual specific populations in the drainage has also been evaluated variation of the discriminating characters within popula- by Ward et al. (1992). They examined Floy tag recoveries tions. Practical considerations with respect to adult sam- and coded-wire tag recoveries and used scale pattern analy- pling and cost require that the differentiation among popu- sis to determine population timing patterns. Unfortunately, lations is greater than the variation within populations so small sample sizes and the nonrandom sampling effort lim- that samples may be pooled over several years in order to ited the accuracy of their results, but their results suggested obtain samples large enough to characterize populations that the mean day of passage of the Morice River population contributing to fishery samples. In our study, population through the lower river was earlier than that of the other differentiation within river drainages was about 2.0 times populations. Our results suggested that there was a bimodal greater for all loci than annual variation within populations. distribution timing of the Bulkley and Morice population Given the results of the simulations, this level of population with a larger early and smaller late peak, with perhaps the differentiation was sufficient to provide reasonable levels of earlier peak representing the Morice component and the accuracy and precision when applied to estimation of stock later peak the Bulkley River component. composition within the two drainages. Periodic monitoring Analysis of the Nass River test fishery suggested that the of the populations would be warranted to ensure that no Bell-Irving and Cranberry populations comprised the ma- substantial drift in allele frequencies have occurred. jority of the run during 1998, with the Damdochax and Within the Skeena and Nass river drainages, evaluation Meziadin populations returning somewhat earlier and the of the accuracy and precision of the simulated mixtures Cranberry population somewhat later than average. As the indicated that the 8 microsatellite loci provided reasonable Bell-Irving River constitutes a very substantial portion of discrimination among populations. Populations within the the Nass River drainage, it does seem likely that steelhead Skeena drainage were usually estimated within 2% of the from this tributary should comprise a substantial portion of actual contribution, and populations within the Nass drain- the total return to the Nass River drainage. However, the age within 4% of the actual contribution. In the Nass drain- relative contribution of Bell-Irving steelhead to the total age, the proportion of the Meziadin River population was Nass drainage return may have been overestimated in our consistently underestimated, perhaps reflecting the lower analysis. For example, steelhead tagging has also been con- baseline sample size of this population relative to other ducted at the fish wheels, and a size bias has been observed populations in the drainage and the reduced differentiation where the mark rate in the tributaries was higher for steel- with other populations (Table 1). head <70 cm in length than for steelhead >70 cm (C. Analysis of the Skeena River test fishery indicated that Parken, Cascadia Natural Resources Consulting, personal the Babine River and Bulkley River drainage populations communication). Mean length of the Bell-Irving population (Bulkley, Morice, and Toboggan) were the dominant popu- was the smallest of any of the summer-run populations lations in the drainage, comprising approximately 60% of surveyed in the Nass drainage (unpublished data), so it may the total return in 1996 and 1998. Peak returns in 1998 were be that the fish wheels sampled Bell-Irving River steelhead observed during the week of July 26 through August 1. Test at a higher frequency than their actual frequency in the fishery results suggested that returns were substantially spawning migration. higher in 1998 compared with 1996, which was likely due to Although our survey of microsatellite variation did not restrictions placed on salmon fisheries for coho salmon reveal enough differentiation between Nass River and conservation. The Sustut River population (samples pri- Skeena River populations to be applied confidently in esti- marily from the upper Sustut drainage) tended to be in mation of stock composition in marine fisheries at this higher relative proportions during the earlier portion of the time, it is likely that a survey of variation at additional run. Steelhead from the Babine River and Bulkley River microsatellite loci or perhaps at major histocompatibility predominated in returns during September. The Toboggan complex loci (Miller and Withler, 1998) may provide Creek population was estimated to have comprised about enough evidence of drainage differentiation for marine 7% of the returns in 1998. This was larger than expected mixed-stock fishery applications. Steelhead Population Structure and Stock ID 599

ACKNOWLEDGMENTS Chilcote, M.W., Crawford, B.A., and Leider, S.A. (1980). A genetic comparison of sympatric populations of summer and winter steel- heads. Trans Am Fish Soc 109:203–206. We acknowledge staff from the provincial Ministry of En- vironment, Lands, and Parks (MELP) who collected adult Felsenstein, J. (1993). PHYLIP: Phylogeny Inference Package. Se- steelhead samples, as well as the various agencies, organi- attle: University of Washington. zations, and companies who collected samples. For the Nass Fournier, D.A., Beacham, T.D., Riddell, B.E., and Busack, C.A. River, these included the Nisga’ a Tribal Council, LGL Ltd., (1984). Estimating stock composition in mixed stock fisheries us- SKR Consultants, and Cascadia Natural Resource Consult- ing morphometric, merisitic, and electrophoretic characteristics. ing. For the Skeena River, in addition to tributary collec- Can J Fish Aquat Sci 41:400–408. tions by MELP staff, samples from the Tyee test fishery were Hansen, M.M., Nielsen, E.E., and Mensberg, K.L.D. (1997). The provided by Fisheries and Oceans Canada staff. Mr. Charles problem of sampling families rather than populations: relatedness Parken of Cascadia Natural Resource Consulting kindly among individuals in samples of juvenile brown trout Salmo trutta provided many constructive comments on the manuscript. L. Mol Ecol 6:469–474. J. Candy assisted in the estimation of stock composition and Goudet, J. (1995). FSTAT a program for IBM PC compatibles to figure preparation. Funding sources were from the provin- calculate Weir and Cockerham’s (1984) estimators of F-statistics cial Common Land Information Base (CLIB) and Habitat (version 1.2). J Heredi 86:485–486. Conservation Trust Fund (HCTF), as well as from the De- partment of Fisheries and Oceans. Jantz, L., Kadowaki, R., and Spilsted, B. (1990). Skeena River salmon test fishery, 1987. Can Data Rep Fish Aquat Sci 804.

Link, M.R., and Gurak, A.C. (1997). The 1995 fishwheel project on REFERENCES the Nass River, BC. Can Man Rep Fish Aquat Sci No 2422.

Miller, K.M., and Withler, R.E. (1998). The salmonid class I MHC: Beacham, T.D., Margolis, L., and Nelson, R.J. (1998). A compari- limited diversity in a primitive teleost. Immunol Rev 166:279–293. son of methods of stock identification for sockeye salmon (On- corhynchus nerka) in Barkley Sound, British Columbia. N Pac Mj␾lner␾d, I.B., Refseth, U.H., and Hindar, K. (1999). Spatial Anad Fish Comm Bull 1:227–239. association of genetically similar Atlantic salmon jveniles and sex bias in spatial patterns in a river. J Fish Biol 55:1–8. Beacham, T.D., and Wood, C.C. (1999). Application of microsatellite DNA variation to estimation of stock composition and es- Murray, C.B., and Rosenau, M.L. (1989). Rearing of juvenile chi- capement of Nass River sockeye salmon (Oncorhynchus nerka). nook salmon in nonnatal tributaries of the lower Fraser River, Can J Fish Aquat Sci 56:297–310. British Columbia. Trans Am Fish Soc 118:284–289.

Beacham, T.D., Pollard, S., and Le, K.D. (1999). Population struc- Nelson, R.J., Beacham, T.D., and Small, M.P. (1998). Microsatel- ture and stock identification of steelhead in southern British Co- lite analysis of the population structure of a Vancouver Island lumbia, Washington, and the Columbia River based on microsat- sockeye salmon (Oncorhynchus nerka) stock complex using non- ellite DNA variation. Trans Am Fish Soc 128:1068–1084. denaturing gel electrophoresis. Mol Mar Biol Biotechnol 7:312–319.

Beacham, T.D., Le, K.D., Raap, M.R., Hyatt, K., Luedke, W., and Nielsen, J.L., Gan, C.A., Wright, J.M., Morris, D.B., and Thomas, Withler, R.E. (2000a). Microsatellite DNA variation and estima- W.K. (1994). Biogeographic distributions of mitochondrial and tion of stock composition of Barkley Sound, British Columbia nuclear markers for southern steelhead. Mol Mar Biol Biotechnol sockeye salmon. Fish Bull (US) 98:14–24. 3:281–293.

Beacham, T.D., Wood, C.C., Withler, R.E., and Miller, K.M. Nielsen, J.L., Carpanzano, C., Fountain, M.C., and Gan, C.A. (2000b). Application of microsatellite DNA variation to estima- (1997). Mitochondrial DNA and nuclear microsatellite diversity in tion of stock composition and escapement of Skeena River sockeye hatchery and wild Oncorhynchus mykiss from freshwater habitats salmon (Oncorhynchus nerka). N Pac Anad Fish Comm Bull (in in southern California. Trans Am Fish Soc 126:397–417. press). Ostberg, C.O., and Thorgaard, G.H. (1999). Geographic distribu- Cavalli-Sforza, L.L., and Edwards, A.W.F. (1967). Phylogenetic tion of chromosome and microsatellite DNA polymorphisms in analysis: models and estimation procedures. Am J Hum Genet Oncorhynchus mykiss native to western Washington. Copera 19:233–257. 1999(2):287–298. 600 Terry D. Beacham et al.

Parkinson, E.A. (1984a). Genetic variation in populations of steel- describe structure of populations of pink and sockeye salmon in head trout (Salmo gairdneri) in British Columbia. Can J Fish Aquat Alaska. N Pac Anad Fish Comm Bull 1:300–318. Sci 41:1412–1420. Small, M.P., Beacham, T.D., Withler, R.E., and Nelson, R.J. Parkinson, E.A. (1984b). Identification of steelhead stocks in the (1998). Discriminating coho salmon (Oncorhynchus kisutch) commercial net fishery of the southern British Columbia coast. Bri populations within the Fraser River, British Columbia using mi- Col Fish Wildl Branch Fish Manage Report No 81. crosatellite DNA markers. Mol Ecol 7:141–155.

Raymond, M., and Rousset, F. (1995). GENEPOP (Version 1.2): Swofford, D.L., and Selander, R.B. (1981). BIOSYS-1: a FORTRAN population genetics software for exact tests and ecumenism. He- program for the comprehensive analysis of electrophoretic data in redity 86:248–249. population genetics and systematics. J Hered 72:281–283.

Reisenbichler, R.R., and Phelps, S.R. (1989). Genetic variation in Taylor, E.B. (1995). Genetic variation at minisatellite DNA loci steelhead (Salmo gairdneri) from the north coast of Washington. among North Pacific populations of steelhead and rainbow trout Can J Fish Aquat Sci 46:66–73. (Oncorhynchus mykiss). J Hered 86:354–363. Ward, B.R., Tautz, A.F., Cox-Rogers, S., and Hooton, R.S. (1992). Reisenbichler, R.R., McIntyre, J.D., Solazzi, M.F., and Landing, Migration timing and harvest rates of the steelhead trout popula- S.W. (1992). Genetic variation in steelhead of Oregon and north- tions of the Skeena River system. Pac Stock Assess Rev Comm Work ern California. Trans Am Fish Soc 121:158–169. Paper S92-07. Rice, W.R. (1989). Analyzing tables of statistical tests. Evolution Wenburg, J.K., Olsen, J.B., and Bentzen, P. (1996). Multiplexed 43:223–225. systems of microsatellites for genetic analysis in coastal cutthroat Scrivener, J.C., Brown, T.G., and Andersen, B.C. (1994). Juvenile trout (Oncorhynchus clarki clarki) and steelhead (Oncorhynchus chinook salmon (Oncorhynchus tshawytscha) utilization of Hawks mykiss). Mol Mar Biol Biotech 5:273–283. Creek, a small and nonnatal tributary of the upper Fraser River. Wenburg, J.K., Bentzen P., and Foote, C.J. (1998). Microsatellite Can J Fish Aquat Sci 51:1139–1146. analysis of genetic population structure in an endangered salmo- Seeb, J.E., Habicht, C., Olsen, J.B., Bentzen, P., Shaklee, J.B., and nid: the coastal cutthroat trout (Oncorhynchus clarki clarki). Mol Seeb, L.W. (1998). Allozyme mtDNA and microsatellite variants Ecol 7:733–749.