}4 



Fishery Bulletin 94(1). 1996 



merits. For each enzyme, the various restriction pat- 

 terns were named alphabetically ("A" being the most 

 common), and the resultant composite haplotypes 

 were assigned numerically. Within-sample variation 

 was estimated as nucleotide diversity (Nei and 

 Tajima, 1981; Nei, 1987), as the average number of 

 nucleotide substitutions per site for the sequences 

 sampled, and as haplotype diversity, which is related 

 to allozyme heterozygosity. Nucleotide divergence 

 (Nei, 1987) was used to measure between-sample 

 variation. The significance of geographic heteroge- 

 neity in haplotype distribution was tested by using 

 a Monte Carlo % 2 approach (Roff andBentzen, 1989), 

 and this method was also used to test the distribu- 

 tion of banding patterns produced by single restric- 

 tion enzymes. Partitioning of mtDNA variation was 

 accomplished by gene diversity analysis (Nei, 1973; 

 Chakraborty, 1980). Between-sample variation (G ST ) 

 was subdivided into variation between seas (G SEA ) 

 and between sampling ports within seas (G ps ). Data 

 were analysed by using REAP v4.1 (McElroy et al., 

 1992 ), PHYLIP v3.5c (Felsenstein, 1993 ), and spread- 

 sheet macros were written for Lotus 1-2-3. 



Results 



DNA amplification 



Of the three primer pairs assayed, only those for the 

 ND5/6 and NDl/16s regions produced amplification 

 products consistently. No products were obtained 

 from anchovy DNA templates with the ND3/4 primer 

 set. We estimated the sizes of the unrestricted ND5/ 

 6 and NDl/16s PCR products to be 2.5 Kb and 2.0 

 Kb, respectively. 



Levels of variability 



In a preliminary study of 15 fish, 10 ND5/6 compos- 

 ite haplotypes were revealed by using 12 restriction 

 enzymes (Alu I, Ava II, Cfo I, Eco RI, Hint I, Msp 

 I, Nei I, Pal I, Rsa I, Sau 961 and Taq I). Only one 

 enzyme, Sau 3AI, was monomorphic for all samples; 

 all other enzymes revealed variation. The same num- 

 ber of haplotypes revealed by these eleven enzymes 

 could also be differentiated by using a subset of six 

 (those in bold type); hence these six were chosen to 

 analyze the entire sample set. 



Initial studies indicated that levels of diversity in 

 NDl/16s were of a similar order to those in ND5/6. 

 Sixteen endonucleases were used (Aat II, Alu I, Ava 

 I, Ava II, Cfo I, Eco RI, Hoe III, Hind III, Hint I. 

 Msp I, Nei I, Pvu I, Rsa I, Sau 3AI, Sau 961, Taq I), 

 five of which did not cut the fragment (Aat II, Ava I, 



Eco RI, Hind III, and Pvu I), three of which produced 

 monomorphic profiles (Hiinf I, Sau 3AI, and Taq I), 

 and another eight of which revealed polymorphism 

 (bold type). Ten composite haplotypes were detected 

 in 15 individuals. Because of logistic considerations, 

 further work on the NDl/16s region was sacrificed 

 in favor of increasing sample numbers in the study 

 of ND5/6 variation; hence all subsequent results re- 

 fer solely to the latter region. 



A total of 77 ND5/6 restriction fragments were 

 scored for the entire data set of 140 fish, revealing 

 53 composite haplotypes (Table 1). Of these, 14 

 haplotypes occurred at least twice, and the remain- 

 ing 39 were encountered in a single fish only. Only 

 haplotype 3 was found at all sample locations. Mean 

 within-sample haplotype diversity was 0.8816 

 (±0.0004), and mean nucleotide diversity was 0.0164 

 (±3xl0~ 6 ) (Table 2). According to both measures, the 

 lowest level of variation was found in the Aegean Sea 

 sample, where 45% of the fish were of genotype 3. 



Inter-sample variation 



Nucleotide divergence between samples varied be- 

 tween 0.0005 and 0.0137, the mean figure being 0.0031 

 (Table 3). The greatest values of nucleotide divergence 

 were in comparisons involving the Aegean Sea sample. 



Monte Carlo x' 2 analysis of geographic heterogene- 

 ity was, in each case, carried out with 1,000 random- 

 izations of the data set. When all samples and all 53 

 haplotypes were included, there was significant geo- 

 graphic heterogeneity (Table 4(i); P=0.011 ), and this 

 was also true of the distribution of restriction mor- 

 phs for all enzymes except Eco RI (data not shown). 

 Removing the Aegean Sea group resulted in a prob- 

 ability of geographic homogeneity of haplotypes of 

 0.337 (Table 4(h)). Indeed, out of 21 pairwise com- 

 parisons between all seven samples (Table 5), only 

 five were significant at P<0.05, and these all involved 

 the Aegean sample (only the Aegean-Sicilian-Channel 

 comparison was nonsignificant; P=0.23). Sequential 

 Bonferroni testing (Rice, 1989) rejected all but the two 

 most significant of these results, however (Table 5). 



A comparison of the distribution of 28 composite 

 haplotypes between the three Adriatic samples was 

 nonsignificant (P=0.689; Table 4(iii)), hence the 

 Adriatic samples were pooled (n =60 (.Although there 

 was no significant heterogeneity in haplotype distri- 

 bution between the pooled Adriatic and outlying 

 samples (omitting the Aegean Sea) (P=0.488; Table 

 4(iv)), there was significant heterogeneity in the dis- 

 tribution of forms revealed by one out of the six en- 

 zymes (Table 4( v ); Sau 961, P=0.007 ). There was thus 

 an indication of genetic heterogeneity between 

 Adriatic fish and those from adjacent waters. 



