UTTER ET AL.: GENETIC POPULATION STRUCTURE OF CHINOOK SALMON 



based on the frequency of homozygotes for the 

 respective variant alleles. Expected heterozygosities 

 were calculated for polymorphic loci. Pairwise com- 

 parisons were made for all loci between all popu- 

 lations by a contingency table analysis using a G 

 statistic. Two or more sample collections lacking 

 significant allele frequency differences for any poly- 

 morphic locus were considered a single population. 

 All subsequent analyses were performed on the 

 resulting 65 individual and pooled populations. A 

 critical value of 1% was used (for both the Hardy- 

 Weinberg and the pairwise population comparisons) 

 to reduce the erroneous rejection of the null hypoth- 

 esis when using multiple tests. Nei's (1975) measure 

 of genetic distance (D) was used to compare pair- 

 wise levels of genetic divergence between individual 

 or pooled populations. A dendrogram based on a 

 matrix of these comparisons was constructed by the 

 unweighted pair group method (UPGM) (Sneath and 

 Sokal 1973). Principal component analysis of the 

 allele frequency data followed procedures outlined 

 in Sneath and Sokal (1973). A nested gene diversity 

 analysis followed procedures described by Nei (1973) 

 and Chakraborty (1980) and was performed through 

 the NEGST computer program described by Cha- 

 kraborty et al. (1982). 



RESULTS AND DISCUSSION 



Tests for Hardy-Weinberg Equilibrium 



Tests for significant deviations from Hardy-Wein- 

 berg proportions were made on each of the 86 data 

 sets for 14 loci including Ah, Ada-1, Aat-3, Dpep-1, 

 Dpep-2, Gpi-1 (excluding the subsequently described 

 Gpi-l(H) allele affecting heterodimer formation), 

 Gpi-3, Gr, Ldh-4, Ldh-5, Mpi, Pgk-2, Sod-1, and 

 Tapep-1. Six deviations were observed (Table 3). 

 These deviations probably were random errors ex- 

 pected from the 1,204 independent calculations at 

 the 1% level of significance. The direction of the 

 deviations indicates both excesses and deficits of 

 heterozygotes in both instances where the same 



Table 3.— Populations and loci with significant (o = 0.01) depar- 

 tures from expected Hardy-Weinberg proportions. 



locus is involved (Mpi and Sod-1). Two of the popu- 

 lations having significant deviations. Eagle Creek 

 and Stuart, were combined for subsequent analysis 

 with other populations having Hardy-Weinberg pro- 

 portions; combinations were based on overall non- 

 significant differences of allele frequencies. The high 

 significance of the Stuart sample for Sod-1 is in- 

 flated through an expected value less than unity for 

 the homozygous genotype of the (-260) allele. 



Description of Allelic Distribution 



The allele frequencies observed for all 25 polymor- 

 phic loci over the geographic range of this study 

 (Appendix) indicate considerable heterogeneity 

 among loci with regard to levels of variation and 

 geographic distribution. This variation is summar- 

 ized from three perspectives— heterozygosity, fre- 

 quency range for common allele, and index of gene 

 diversity (G^, ) (Table 4). Heterozygosity measures 

 within-population variation. Those loci having higher 

 heterozygosities have greater potential for diver- 

 gence of allele frequencies among populations. Mean 

 heterozygosity over all loci was 0.102, and five loci 

 (Ah, Mpi, Pgk-2, Sod-1, Tapep-1) exceeded 0.200. 



The range of allele frequencies and the index of 

 gene diversity reflect the actual divergences ob- 

 served among populations. The range is a simple 

 identification of allele frequency extremes. The in- 

 dex of gene diversity is a quantitative measure of 

 genotype deviations of the overall data set from 

 those expected in a single panmictic population. 

 Seven of the eight most heterozygous loci (Pgk-2, 

 Mpi, Sod-1, Ah, Tapep-1, Gpi-2, Dpep-1) were among 

 the eight loci having either the greatest range in 

 frequency or the highest index of gene diversity, 

 indicating considerable genetic differences among 

 the populations samples. Typically, adjacent popula- 

 tions tended to have allele frequencies more similar 

 to one another than to those from other areas (see 

 Appendix). Notable examples include the follow- 

 ing: 1) restriction of Gpi-2 variation largely to 

 coastal populations from Vancouver Island through 

 Oregon, 2) the highest frequency of the Gpi-l(H) 

 allele in populations from the Sacramento River, 3) 

 Aat-3 variation that is largely restricted to popula- 

 tions from Georgia Strait and western Vancouver 

 Island, 4) low frequencies of variant alleles for most 

 loci in all Klamath River populations and in spring 

 and summer run populations from the Snake River, 

 and 5) high frequencies of Tapep-1 variants in Puget 

 Sound populations. 



Two procedures for graphic analysis (a dendro- 

 gram [Fig. 2] based on pairwise genetic distance 



243 



