Olsen et al : An examination of spatial and temporal genetic variation in Theragra chalcogmmma 



759 



of the rare haplotypes were found in large samples. For 

 example, 52 of 66 composite haplotypes had frequencies of 

 K^ or less and only three of these haplotypes were present 

 in the UNI sample, but 8, 14, and 15 rare haplotypes were 

 present in the three largest samples (PWS 1997, BOG 

 1997. SHEL 1997; Table 4). 



Population genotypic frequencies generally did not devi- 

 ate from Hardy-Weinberg expectations. Permutation tests 

 for the statistic f, conducted independently for allozymes 

 and microsatellites, were concordant for all but two popu- 

 lations (BOG 1998; UNI 1998). The observed value of ;^ for 

 microsatellites in these two populations was significantly 

 greater than zero (P<0.005) when a=0.05 was adjusted for 

 nine simultaneous tests (Table 4). 



Spatial structure in genetic variation 



1997 Gulf of Alaska and Bering Sea Population structure 

 among the three North America samples was evident from 

 mtDNA (overall 0gj=Qmi, P=0.002; Table 5). The hier- 

 archical analysis indicated that the two Gulf of Alaska 

 populations were genetically different from the Bering 

 Sea population (<i>c;j^0.028, P<0.001) but were not distin- 

 guishable from each other (Table 5). These results were 

 confirmed by goodness-of-fit tests of haplotype homoge- 

 neity. Spatial genetic structure among the 1997 samples 

 was not evident from values of 6 over all allozymes and 

 microsatellites (Table 5). 



1998 Gulf of Alaska and Bering Sea, and 1999 eastern 

 Kamchatka Indices of population structure for allozymes 

 (0=0.017, P<0.001) and mtDNA (0sy=O.O21, P<0.001) 

 were similar over all population samples and larger 

 than for microsatellites (0=0.002. P>0.050; ^,,.^=0.001, 

 P>0.050; Table 5). Because the values of 9 and R^j over 

 all microsatellites and pop-ulations were not significant, 

 analyses of population structure were conducted with only 

 the allozyme and mtDNA data. Significant genetic differ- 

 ences were revealed between pooled samples from North 

 America and Asia with allozymes (0=0.030. P<0.001) and 

 mtDNA «i',,.7^0.035, P<0.001; Table 5), between samples 

 from the Gulf of Alaska and Bering Sea with allozymes 

 (0=0.005, P=0.006), and within the Gulf of Alaska with 

 allozymes (0=0.009, P<0.001) and mtDNA ((P^.j^O.OlS, 

 P=0.009). 



Values for were considerably larger for SOD-2* than 

 for the two other informative markers MPI*. and mtDNA 

 (Table 6, Fig. 2). Estimates of P,;^. from mtDNA and MPI* 

 for each of the regional comparisons were similar but the 

 most informative microsatellite locus, Tch22. did not vary 

 significantly among regions. 



Based on permutation test results from the hierarchical 

 analyses, estimates of genetic differentiation were computed 

 for all population pairs by using both allozyme and mtDNA 

 data, except for Gulf of Alaska versus Bering Sea popula- 

 tions (allozyme data only). Values of (allozymes) and 0,.^ 

 for most pairwise comparisons from North America and 

 Asia were relatively large and highly significant (Table 5). 

 In contrast, values of (allozymes) for most population 

 pairs from the Gulf of Alaska and Bering Sea were rela- 



tively small, and only the UNI x SHEL and UNI x PWS 

 pairs were significant (0=0.021, P<0.003; 0=0.017, P<0.006). 

 Paii-wise comparisons within the Gulf of Alaska were sig- 

 nificant for SHEL X PWS (allozymes, 0=0.017, P<0.002) and 

 for SHEL X MID (mtDNA, *^..;,=0.025, P<0.007; Table 5). 



In goodness-of-fit tests, allele and haplotype frequency 

 homogeneity were concordant with the statistically sig- 

 nificant values of (allozymes) and (t>gj for all populations 

 and pooled samples within and between regions (Table 5). 

 However, results from the two types of tests conflicted for 

 some population pairs. For example, G-tests of allozyme 

 data revealed significant genetic variation for two popula- 

 tion pairs (BOG x PWS; BOG x MID; Table 5) for which 

 was not significant. In contrast, was significant for two 

 population pairs, KRON x PWS and UNI x SHEL, which 

 were not genetically different based on G-test results. One 

 population pair, PWS x MID, was genetically different in 

 the probability test of mtDNA data, but i>gj. for this pair 

 was not significant (Table 5). 



Temporal change In genetic variation: 

 Bogoslof Island, Shelikof Strait, and 

 Prince William Sound 



Significant temporal change in mtDNA variation was 

 detected in one population, BOG, by using the probability 

 test of haplotype homogeneity (P<0.001) and 0gj. (P=0.003, 

 Table 5). The <Pgj was within the range of values described 

 for spatial variation and indicated a significant interan- 

 nual change in haplotype frequency. Interestingly, this 

 interannual shift in genetic variation was not evident 

 from allozymes and microsatellites. However, the single 

 most-informative locus SOD-2* was not analyzed in the 

 BOG 1997 sample because heart tissue was not available. 

 Also, variation at the locus MPI*. although not statistically 

 significant, suggested that the BOG replicates were geneti- 

 cally different (0=0.01 1, P=0.066; Table 6). A similar indica- 

 tion of temporal genetic variation was detected in the PWS 

 population for all allozymes (0=0.005, P=0.046), but was 

 not significant when a was adjusted for three simultane- 

 ous tests (initial a=0.016. Table 5). In this case SOD-2* 

 was resolved and exhibited significant genetic variation 

 (0=0.047, P=0.004;Table 6). The mtDNA data, although not 

 significant, also provided an indication of temporal varia- 

 tion in the PWS population {Sgj,=0.0V2, P=0.055; Table 6). 



Discussion 



Comparison of marker classes 



Our empirical results support two tentative conclusions 

 regarding the usefulness of these marker classes for 

 detecting population structure in walleye pollock. First, 

 high overall heterozygosity such as that observed in some 

 microsatellites may not be a reliable indicator of power 

 for detecting population differences in walleye pollock, 

 regardless of spatial scale. This conclusion contrasts with 

 some simulation studies that have shown a positive rela- 

 tionship between locus polymorphism and the power to 



