364 



Fishery Bulletin 91(2). 1993 



also used in the electrophoretic study of Wilson & 

 Alberdi (1991), one thaw/refreeze cycle occurred prior 

 to use in the present study. All specimens were identi- 

 fied to nominal species using the criteria of Whitehead 

 (1973, 1985). That is, we counted the number of gill 

 rakers on the lower limb of the first gill arch and 

 related the count to standard length (SL), and we ex- 

 amined the condition (curled or flattened) of the ante- 

 rior rakers of the lower limb of the first arch. 



Mitochondrial DNA was purified from the ripe ova- 

 ries of the specimens taken from the Gulf of Mexico as 

 in Wilson & Tringali (1990), except that the mtDNA 

 collected after density-gradient ultracentrifugation was 

 extracted in n-butanol which was not saturated with 

 NaCl. This removed most of the salts, allowing greater 

 precision over the molar concentration of the DNA- 

 containing solution during the ethanol precipitation. 

 The lower yield of mtDNA caused by thawing and re- 

 freezing of the Brazilian specimens necessitated use of 

 a modification of the Chapman & Powers ( 1984) method 

 of extraction (Tringali 1991). 



Restriction digests were carried out as specified by 

 the manufacturers using the following enzymes: Apo-I, 

 BamU-l, Dra-l, EcoR-l, Hind-UI, Pst-1, Pvu-ll, Sac-l, 

 Xba-l, and Xho-l. Fragments were separated by hori- 

 zontal agarose gel electrophoresis and visualized in 

 most cases using the gel-incorporated Hoechst 33258 

 (CalBiochem #SR5A03-0388) fluorochrome dye as in 

 Deflaun & Paul (1986). However, some poor yields of 

 mtDNA during purification made visualization of the 

 smaller restriction fragments difficult using the 

 fluorochrome, thus requiring application of an mtDNA 

 hybridization probe to southern blots of the agarose 

 gels (Tringali 1991). The non-radioactive hybridization 

 probe was made from cesium-purified mtDNA of S. 

 aurita using a commercial kit ( Genius Kit of Boehringer 

 Mannheim, Cat. #1093657). Hybridization and immu- 

 nological detection of the probe was performed as speci- 

 fied by the manufacturer with the following modifica- 

 tions. The concentration of the blocking reagent in the 

 prehybridization and hybridization solutions was 

 doubled to 1.0% weight/volume to reduce background 

 coloring of the hybridization filter (Zeta-Probe). The 

 prehybridization period was extended to a minimum 

 of 2h and the volume of the probe-containing hybrid- 

 ization solution was increased to 10 m L/100 cm 2 of 

 filter. The ELISA reaction period was increased to 2 h. 

 Under these modifications, the mtDNA restriction frag- 

 ments were usually visible on the filter ~15min into 

 the color reaction. 



Molecular weights (in base pairs) of the restriction 

 fragments were estimated by the global form of the 

 reciprocal method as described in Elder & Southern 

 (1987). Restriction-fragment length polymorphisms 

 (RFLPs) were expressed as two data types: restriction- 



site data and composite mtDNA haplotypes. The 

 restriction-site composition of each individual was in- 

 ferred from fragment profiles and intraspecific sequence 

 diversity (Nei & Miller 1990) was calculated using the 

 RESTSITE vl.2 algorithm. Composite mtDNA 

 haplotypes were constructed for each individual from 

 their restriction profiles with haplotypes grouped by 

 sample location and time of collection. The frequency 

 distribution of mtDNA haplotypes was tested for het- 

 erogeneity, with respect to location and time, by the 

 log-likelihood G-test for independence (Sokal & Rohlf 

 1981). G-tests were also performed on selected samples 

 following generation of 100 randomized data sets from 

 the observed data, using the Monte Carlo technique 

 described by Roff & Bentzen (1989) where the prob- 

 ability, P, of obtaining a randomized G H value greater 

 than the original observed value is given by 



P = n / N„ 



(1) 



where n equals the number of outcomes in which 

 the randomized G H value is greater than the actual G H 

 value, and N r represents the number of randomiza- 

 tions performed. The standard error of P was taken as 



[P(l-P)/N r ] 1/2 . 



(2) 



The frequency of the most common haplotype was 

 tested, after arcsine square-root transformation (Free- 

 man & Tukey 1950), using the "V" statistic (Desalle et 

 al. 1987) for heterogeneity between sampling locations. 



Results 



Counts of gill rakers identified 72 of 73 sample speci- 

 mens as S. aurita; one specimen from Brazil had gill- 

 raker counts expected for S. brasiliensis. Counts were 

 lowest for the Panhandle specimens while specimens 

 from Tampa Bay and southern Brazil were the most 

 similar, having higher counts (Table 1). Specimens from 

 the Florida Panhandle were of larger average size than 

 those from Tampa Bay and southern Brazil, which were 

 more similar in size (Table 1). A one-way analysis of 

 variance revealed significant heterogeneity (P<0.05) in 

 gill-raker counts between the specimens taken from 

 the Panhandle and Tampa Bay, but not between those 

 taken from Tampa Bay and southern Brazil. One Bra- 

 zilian specimen had a gill-raker count of 154 at 

 150 mm SL, nearly identical to that reported for the 

 lectotype of S. brasiliensis (155 at 148.3 mm SL; 

 Whitehead 1973). However, we could detect no differ- 

 ences in the appearance of the anterior rakers (curled 

 vs. flattened) for this or other specimens. The overall 

 appearance of the rakers of all specimens most closely 

 resembled Whitehead's (1985) diagrammatic drawing 



