Tnngali and Wilson: mtDNA analysis of two stocks of Sardmella aunta 



365 



Table 1 



Data for Spanish sardines Sardmella aunta used in mtDNA 

 analysis. 



1 Mean length of pooled specimens from the Panhandle is 



significantly larger than that of specimens from the other 



two regions (ANOVA, P<0.05). 

 ' Mean number of gill rakers of pooled specimens from the 



Panhandle is significantly lower than that of specimens from 



the other two regions (ANOVA, P<0.05). 



of S. brasiliensis, curling anteriorly downward. How- 

 ever, use of this character was extremely problematic 

 for us because, other than the drawing, we found no 

 published taxonomic account of the discovery of this 

 character, no photographic evidence nor written descrip- 

 tion detailing the character, nor reference of this char- 

 acter to [any] catalogued or type material. 



The combination of mtDNA purification and visual- 

 ization techniques used here provided well-resolved 

 digestion patterns that could be consistently and un- 

 ambiguously scored for all individuals. Of the 10 re- 

 striction endonucleases used, only BamH-l was not 

 informative. This enzyme had but one restriction site 

 in all samples and was used only in probe construction 

 to linearize template mtDNA prior to labeling. Poly- 

 morphic digestion patterns occurred for the remaining 

 nine enzymes, with Pst-l and Pvu-ll highly variable 

 (Table 2). The mean mtDNA genome size, calculated 

 from the sums of all digestion patterns, was 16,304 

 base pairs (bp) ± 71 bp (SD). No size variation was 

 detected between specimens or within the mtDNAs of 

 individual specimens. 



The nine informative enzymes sampled 52 unique 

 restriction sites over all individuals (Table 3) with a 

 mean of 43 sites per individual. All RFLPs were con- 

 sistent with the assumption of a single nucleotide sub- 

 stitution between variant digestion patterns. Intraspe- 

 cific sequence diversity (Nei & Miller 1990) over all 

 individuals was 0.00525±0.00204 (bootstrapped SE) 

 substitutions per nucleotide, a value consistent among 

 all three sampling locations (x=0.00509±0. 00054, range 

 0.00452-0.00561). Twenty-four composite mtDNA 

 haplotypes were generated from polymorphic restriction- 

 fragment profiles of the 73 completely characterized 



Spanish sardines (Table 4). The composite mtDNA 

 haplotype (BAABAAAAA) of the one putative S. 

 brasiliensis specimen with a gill raker count close to 

 the lectotype's (Whitehead 1970, 1973) was identical to 

 that of another Brazilian specimen having the low gill 

 raker count indicative of S. aurita (126 at 169 mm SL), 

 as well as to that of seven specimens from the Gulf of 

 Mexico (Table 4). The mtDNA haplotype most common 

 in the Gulf of Mexico, AAAAAAAAA, occurred in very 

 low frequency among the Brazilian specimens. The fre- 

 quency of this haplotype was significantly different be- 

 tween specimens of the Gulf of Mexico and Brazil 

 (V= 10.90, P<0.001), but not among specimens from the 

 different regions within the Gulf of Mexico (Table 5). 



Ten composite haplotypes were unique to the Pan- 

 handle region, four to the Tampa Bay region, and five 

 to Brazil. Four were observed in all three locations. G- 

 tests of mtDNA haplotype frequencies for temporal and 

 geographical heterogeneity among specimens from 

 within the Gulf of Mexico did not allow rejection of the 

 null hypothesis of population homogeneity (P>0.05). 

 Pooling the Gulf specimens by location and testing for 

 heterogeneity between locations using Monte Carlo sta- 

 tistics (100 randomizations) generated 35 G H values 

 exceeding the initial value of 25.774, again indicating 

 that the null hypothesis of homogeneity for specimens 

 taken within the Gulf of Mexico cannot be rejected 

 (P>0.35, 2SE=0.09). Pooling all Gulf specimens and 

 testing against the Brazilian specimens using Monte 

 Carlo statistics resulted in a G H value of 44.282 which 

 was not exceeded by any of the 100 randomizations, 

 indicating significant heterogeneity between specimens 

 from those locations (P<0.01 ). 



Discussion 



Our results indicate an evident lack of genetic stock 

 structuring between the commercially-fished popula- 

 tions of Spanish sardine in the eastern Gulf of Mexico, 

 even though there is a significant difference in the 

 average number of gill rakers between locations, in 

 addition to the morphometric differences reported by 

 Johnson & Vaught ( 1986). This significant variation in 

 gill-raker counts relative to SL appears to confirm 

 Whitehead's ( 1973 ) reference to two rates of increase 

 in the number of gill rakers with increasing standard 

 length among Spanish sardines of the western Atlan- 

 tic, although the lowest counts (Panhandle) were lower 

 than those Whitehead (1973) mentioned. Whitehead 

 ( 1973 ) thought this variation in gill-raker counts in- 

 dicative of the presence of two valid species of Spanish 

 sardines in the western Atlantic, but our mtDNA analy- 

 sis does not support his suggestion. 



