Van Doornik et al.: Transferrin polymorphism in Oncorhynchus kisutch 



573 



bia River samples clustered separately from north- 

 ern Oregon coastal samples. However, the two south- 

 ern Oregon coastal samples that they examined 

 (Rogue River and Coquille River I clustered with the 

 Columbia River samples, rather than with samples 

 from the Oregon coast. This result is similar to ours, 

 except that the Coquille River did not cluster with 

 Columbia River samples in our analysis. 



Another important discovery was the relatively 

 high level of temporal variation for transferrin al- 

 lele frequencies. Accurate stock identification from 

 allele-frequency data requires that allele frequencies 

 remain relatively stable over time. Because most coho 

 salmon south of Alaska return to spawn at 3 years of 

 age (Sandercock, 1991), it is possible that allele fre- 

 quencies for a particular stock could show marked 

 differences between brood years. Waples and Teel 

 ( 1990) found that for salmon populations where 90% 

 of a brood year returned to spawn in the same year, 

 the probability of finding a significant test for tem- 

 poral differences increased dramatically. 



Our test for temporal allele stability assumed there 

 was no influx of nonnative fish into the gene pool; 

 however, it is unlikely that this assumption was valid. 

 All comparisons were made between samples ob- 

 tained from hatcheries, and many of these hatcher- 

 ies have imported and released fish that were not 

 native to the hatchery location. For example, from 

 1952 to 1991, 20.2% of the coho salmon released at 

 the Minter Creek Hatchery were not native to Minter 

 Creek (Washington Dep. Fish and Wildlife 3 ). 



Another assumption that we made when testing 

 temporal stability was that no natural selection oc- 

 curred. Previous studies have found some evidence 

 that transferrin alleles are influenced by natural 

 selection. Suzumoto et al. (1977) observed that fish 

 that have the *97 transferrin allele are less suscep- 

 tible to bacterial kidney disease (BKD). Pratschner 

 (1978) found that the * 1031* 103 phenotype had 

 greater resistance to furunculosis, whereas the *97l 

 *97 phenotype was most resistant to vibriosis. A 

 study by Mclntyre and Johnson ( 1977 ) indicated that 

 fish with a * 1031* 103 phenotype had a faster growth 

 rate than those with a *103/*97 phenotype, but the 

 escapement rate for the two different phenotypes was 

 equal. However, results of a study by Winter et al. 

 (1980) conflicted with both Suzumoto et al. (1977) 

 and Pratschner (1978). Their results indicated that 

 only certain stocks show a genetically influenced re- 

 sistance to BKD, furunculosis, and vibriosis. They 

 concluded that "the importance of transferrin geno- 



3 Washington Department of Fish and Wildlife. 1994. His- 

 torical releases of juvenile salmon into Washington waters, coho 

 salmon. Available: Washington Dep. Fish Wildl., 600 Capitol 

 Way N., Olympia, WA 98501-1091. (Interactive database). 



types in resistance to disease is stock specific." The 

 extent to which these factors affect transferrin al- 

 lele frequencies and how they relate to GSI analyses 

 are uncertain. If certain alleles are being selected 

 for, according to location-specific factors, then allele 

 frequencies could be more stable over time for a given 

 location, even after nonnative stocks are introduced 

 into an area. However, if selection factor(s) were to 

 change, allele frequencies would change significantly 

 over time, which would help to explain the large per- 

 centage of significant temporal comparisons that we 

 found. 



We found at least as much temporal variation be- 

 tween samples analyzed in our own lab as we did 

 between samples analyzed in two different laborato- 

 ries. Therefore, we ruled out the possibility that tem- 

 poral variation was due to differing techniques or to 

 differing interpretations of data, or both, in two dif- 

 ferent laboratories. 



These results suggest that before the transferrin 

 locus is used in GSI analysis, temporal variation of 

 allele frequencies between brood years will need to 

 be considered. However, temporal variation does not 

 necessarily preclude the use of the transferrin locus 

 in GSI analysis. In a study of temporal variation in 

 lake trout, Salvelinus namaycush, Grewe et al. ( 1994) 

 found that although there were significant differ- 

 ences in allele frequencies between year classes, the 

 accuracy of their mixed-stock contribution estimates 

 was not substantially affected. Waples (1990) sug- 

 gested that to obtain maximum precision of esti- 

 mates, temporally spaced samples should be pooled, 

 unless the temporal differences are too large to be 

 attributable to sampling error and to genetic drift. 

 Of course, if the baseline data include only the brood 

 years known to be contributing to the mixed-stock, 

 temporal variation will not be a problem. 



The mean heterozygosity value we calculated 

 (0.425) was indicative of a high level of heterozygos- 

 ity for the transferrin locus. This was one of the high- 

 est mean heterozygosity values we have observed for 

 any polymorphic locus in coho salmon (Van Doornik, 

 unpubl. data). It is also one of the highest mean het- 

 erozygosity values in comparison with other studies 

 of salmonids. For example, the highest mean het- 

 erozygosity found by Wehrhahn and Powell (1987) 

 in their study of 26 coho salmon loci from 95 sites in 

 British Columbia was 0.0099 for the locus LDH-B2* 

 (transferrin was not included in this study). The larg- 

 est mean heterozygosity value, in a study of 25 loci 

 of 86 populations of chinook salmon, was 0.420 for 

 the locus PGK-2* (Utter et al., 1989). 



As shown by the results from the ANOVA, there 

 were significant differences between samples. The 

 relative gene diversity values showed that there was 



