Olsen et al : An examination of spatial and temporal genetic variation in Themgm chalcogiamma 



757 



lane sizing standard and local Southern sizing algorithm 

 in the GeneScan 672 software vers. 1.1 (ABI, 1996a). Al- 

 leles for each locus were scored and data were tabulated 

 for importing into statistical software with Genotyper 

 software, vers. 2.0 (ABI, 1996b). 



Of the thirteen loci screened, Teh 10, Teh 12, and Tch22 

 were used to test for spatial and temporal genetic varia- 

 tion. The PCR annealing temperatures for these loci were 

 the following: (for TchlO) 54°C; (for Tchl2) 47°C; and (for 

 Tch22) 52°C. The loci GnwlO, Gmol23, Gmol45, Tch5, and 

 Teh 18 did not amplify consistently by the methods above, 

 and Gmol. Grno2, Gmo9, Gmol32, and Teh 1 1 appeared to 

 possess null alleles as revealed by significant heterozygote 

 deficits (data not shown). 



nificance of estimates for each Fg^ analog (9, R^j., and 0,57.) 

 was determined by using a permutation test option in the 

 respective computer programs. In each case the data set 

 was permuted 1000 times (alleles or haplotypes among 

 population samples). Second, allele and haplotype fre- 

 quency homogeneity was tested for the same population 

 comparisons by using a G-test (allozymes and microsatel- 

 lites) and probability test (mtDNA). The threshold for sta- 

 tistical significance (a=0.05) for multiple comparisons was 

 determined by using the sequential Bonferroni method 

 (Rice, 1989). 



Results 



Statistical analyses 



Estimates of allele and haplotype frequency were calcu- 

 lated for each locus and population sample. Heterozygos- 

 ity estimates (W) were calculated by using Equation 8.4 

 of Nei (1987) and haplotype diversity estimates (/;) were 

 computed by using the program ARLEQUIN vers. 1.1 

 (Schneider et al., 2000). A permutation test of the statistic 

 f (Weir and Cockerham, 1984) was used to assess confor- 

 mity to Hardy-Weinberg equilibrium (HWE) for each locus 

 and over all loci for each population with the computer 

 program FSTAT vers. 2.8 (Goudet 2000). The data set was 

 permuted 1000 times (alleles were permuted among indi- 

 viduals, within population samples I, and the threshold for 

 statistical significance (a=0.05) was corrected for simul- 

 taneous tests by using the sequential Bonferroni method 

 (Rice, 1989). 



Spatial and temporal genetic variation were quantified 

 by estimating analogs of Wright's Fg-j. for each marker 

 class: 6 (allozymes and microsatellites. Weir and Cock- 

 erham, 1984), Rgj, (microsatellites, Slatkin, 1995), cPgj, 

 (mtDNA, Excoffier et al., 1992). Estimates of 6, Rgj,, and 

 0gj^ (9, Rgj^ and <t>gj,) were computed for all nine popula- 

 tion samples, for all contemporaneous population samples 

 ( 1997, 1998), and for the temporal replicates. The hap- 

 lotype frequency estimates were used to compute <Ps;y. 

 A hierarchical approach was used to determine if more 

 detailed assessment of spatial variation was necessary. 

 That is, only those statistics id, Rg^, 0gj.} that were signifi- 

 cant over all samples from a given year were then calcu- 

 lated for populations pooled by region, and for populations 

 within regions. Statistics significant at the regional level 

 were calculated for each population pair. This hierarchical 

 approach was applied over all loci for allozymes and mic- 

 rosatellites as well as for individual loci because in some 

 instances statistically significant spatial variation was 

 evident for a single locus but not for the marker class. The 

 computer programs FSTAT, version 2.8 (Goudet, 2000), 

 RST Calc, version 2.2 (Goodman, 1997), and AKLEQUIN, 

 version 1.1 (Schneider et al., 2000) were used to compute 

 9, Rgj., and 0gf. 



The statistical significance of spatial and temporal varia- 

 tion in genetic diversity of walleye pollock was estimated by 

 using Fsjj. estimator tests and allelic (haplotypic) goodness- 

 of-fit tests (Goudet et al., 1996). First, the statistical sig- 



Single-locus statistics and a comparison of 

 marker classes 



The degree of polymorphism indicated by the number of 

 alleles per locus, allele frequency, and total heterozygosity 

 (Hj,) varied across loci and marker classes and was gener- 

 ally lower for allozymes than for microsatellites (Table 3). 

 The number of alleles per allozyme locus ranged from two 

 to nine (mean=four) and the frequency of the common 

 allele in the pooled population sample was greater than 

 0.950 for all but four loci (ALAT',MPI*, PGDH*, and SOD- 

 2*, Table 3). In contrast, the number of alleles per locus 

 was higher for microsatellites (range=12-32, mean=20), 

 and the maximum allele frequency did not exceed 0.400 

 (Table 3). Relatively high polymorphism was also detected 

 from restriction digests of ND5/6, cytochrome b, and 

 cytochrome oxidase I regions of mtDNA (66 composite 

 haplotypes, maximum haplotype frequency 0.371, Table 

 3). Values of mean //j. were an order of magnitude greater 

 for microsatellites (0.769) than for allozymes (0.069); how- 

 ever, some overlap in range was evident because of the 

 large variation in H^ ainong loci (Table 3). The estimate 

 of haplotype diversity (/!=0.837) was slightly greater than 

 the Hj for microsatellites (Table 3). 



Values for 9 and <t>gj. varied among loci and failed to ex- 

 pose a single marker class, regardless of overall variabil- 

 ity, as most informative for detecting population structure 

 in walleye pollock (Table 3). Permutation tests revealed 

 two allozyme loci [MPT\ SOD-2*) and one microsatellite 

 (Tch22) with 9 values significantly greater than zero over 

 all population samples (Table 3, Fig. 2). The 9 of 0.088, for 

 SOD-2* was exceptionally high. The 0gj. for mtDNA was 

 within the range of 6* for allozymes and microsatellites and 

 was significantly greater than zero (P<0.001). 



Intrapopulation genetic variation and 

 Hardy-Weinberg equilibrium 



Estimators of intrapopulation diversity (alleles per locus, 

 Z/^. and h) differed among marker types within popula- 

 tions but generally showed little variation between popu- 

 lations (Table 4). As expected, the values for the diversity 

 estimators were usually lowest in the population with 

 the smallest sample sizes (e.g. UNI, «=40). This was most 

 notable in haplotype number (range 12-27) because most 



