496 



Fishery Bulletin 100(3) 



Cavalli-Sforza's chord distance (Cavalli-Sforza and Ed- 

 wards, 1967), as implemented in the genedist program in 

 version 3.4 of the phylogenetic inference package (prnTip) 

 of Felsenstein (1992), was used to estimate the degree of 

 genetic divergence or similarity between pairs of samples; 

 neighbor joining (Saitou and Nei, 1987), from the neighbor 

 program in phylip, was used to cluster the resulting genetic 

 distance matrix. A consensus of 500 neighbor-joining to- 

 pologies was constructed by using the consense program in 

 pmTiP. Spatial autocorrelation analysis was carried out to 

 determine whether allele distributions at each microsatel- 

 lite at any given sample locality were independent of allele 

 distributions in adjacent localities. Briefly, autocorrelation 

 coefficients (Moran's I values), generated as a function of 

 geographic distance between pairs of sample localities, 

 were used to summarize patterns of geographic variation 

 of allele frequencies at each microsatellite. Positive auto- 

 correlations between adjacent localities, with decreasing 

 autocorrelation as geographic distance between localities 

 increases, are generally interpreted as an isolation-by- 

 distance effect (Sokal and Oden, 1978a). We employed 

 the spatial autocorrelation analysis program (SAAP) of 

 Wartenberg (1989) and followed procedures outlined in 

 Sokal and Oden ( 1978a, 1978b). "Noise" was minimized by 

 including only alleles that occurred 20 or more times in the 

 data set. The first of two SAAP runs employed equal geo- 

 graphic distances between each of five distance classes; the 

 second employed equal numbers of pairwise comparisons 

 in each distance class. Finally, assignment tests (Paetkau 

 et al., 1995, 1997) were used to "assign" individuals within 

 each of the 20 samples to one of two regional (spatial) 

 groupings, Atlantic or Gulf The two groupings were em- 

 ployed largely as a result of homogeneity tests of allele 

 distributions, where existence of the two spatial groupings 

 was weakly supported. Assignment tests have a number 

 of uses (Waser and Strobeck, 1998): in this case we were 

 interested in the proportion of individuals within a sample 

 that could be assigned to each regional group, in relation 

 to the locality of the sample and the season in which it 

 was procured. Assignment tests were carried out employ- 



ing the "assignment calculator" software available at http: 

 \ \ www.biology.ualberta.ca/jbruzusto/Doh.html. 



Results and discussion 



Allele frequencies at the seven microsatellites in each 

 of the 20 samples are given in Appendix Tables 2 and 

 3; number of individuals assayed, heterozygosity (direct 

 count) values, and probability of conformance to expected 

 Hardy-Weinberg proportions per microsatellite per indi- 

 vidual sample are given in Appendix Tables 4 and 5. 

 Summary statistics are given in Table 2 and include 1) 

 repeat sequence of the cloned allele, 2) number of alleles 

 detected, 3) average (direct count) heterozygosity (±SE) 

 observed among samples, and 4) results of tests of confor- 

 mance of observed genotype proportions to expectations of 

 Hardy-Weinberg equilibrium. Cloned alleles at the seven 

 microsatellites included simple {Sca-49, Sca-65) and com- 

 plex (Sca-14, Sca-23, Sca-37, Sca-61 ) dinucleotide repeats 

 and one complex tetranucleotide repeat (Sca-44). All 

 dinucleotide microsatellites included CA (or complemen- 

 tary TG) repeats, with the number of alleles per micro- 

 satellite ranging from five (Sca-14) to twenty-four {Sca-23 

 and Sca-65). Direct count heterozygosity, averaged over 

 the twenty samples, ranged from 0.311 ±0.018 {Sca-61) 

 to 0.803 ±0.105 {Sca-23). These results indicate that the 

 seven microsatellites assayed in king mackerel are typi- 

 cal of microsatellites found in other vertebrate organisms, 

 including fishes (e.g. DeWoody and Avise, 2000; Turner et 

 al., 1998; Gold etal., 2001). 



After sequential Bonferroni correction (Rice, 1989), gen- 

 otype proportions at six of the microsatellites in all twenty 

 samples did not deviate significantly from proportions 

 expected under Hardy-Weinberg equilibrium. Genotype 

 proportions at Sca-23 among three of the samples (SEB^, 

 KEY', and SAE-») differed significantly (P=0.000) from 

 Hardy-Weinberg equilibrium expectations, and at a fourth 

 sample (SARM, the probability value of 0.006 was very 

 close to the Bonferroni adjusted alpha of 0.003 (Appendix 



