130 



Fishery Bulletin 98(1) 



The DNA was concentrated by isopropanol precipita- 

 tion, resuspended in 75 microliters of sterile water, 

 and stored at -20°C. 



mtDNA control region sequencing 



Initially, we used the polymerase chain reaction (PCR; 

 Saiki et al., 1988) and published primer sequences 

 L15926 (Kocher et al., 1989) and H16498 (Meyer 

 et al., 1990) to amphfy a portion of the mtDNA con- 

 trol region of red drum. Double-stranded PCR was 

 performed with Perkin Elmer AmpliTaq in a 50-pL 

 reaction volume for 32 cycles in a DeltaCycler II 

 System (Ericomp Inc., San Diego, CA), according to 

 the methods described by Kocher et al. (1989). We 

 amplified a 455-base-pair (bp) fragment of the con- 

 trol region for several individuals of red drum. How- 

 ever, because direct sequencing of the amplicon with 

 the same PCR primers in the sequencing reaction 

 yielded unsatisfactory results, we used a process of 

 cloning and sequencing to design specific primers for 

 red drum. 



The 455-bp amplicon was cloned in pBluescript by 

 TA cloning (Marchuk et al., 1990). Following denatur- 

 ation of the plasmid DNA(Hattori and Sakaki, 1986), 

 sequencing was done from both directions by the dide- 

 oxy termination method (Sanger et al., 1977) by using 

 Sequenase version 2.0 (U.S. Biochemicals, Cleveland, 

 OH) and [a-S^S] dATP (Dupont Biotechnology Sys- 

 tems, Wilmington, DE). Products of the sequencing 

 reactions were resolved in 6% polyacrylamide/7-M 

 urea gels that were vacuum dried at 80°C and auto- 

 radiogi'aphed with Kodak X-Omat AR film. We used 

 the ESEE program (Cabot and Beckenbach, 1989) 

 to align sequences. From these sequences, (Genbank 

 accession no. AF054671), highly specific internal 

 primers were designed for the control region of 

 red drum. These primers partially overlapped the 

 initial primers and were designated L15943 (5'-GTA 

 AACCGGATGTCGGGGGTTAG-3') and H16484 

 (5'-GGAACCAGATACCAGGAATAGTTCA -3' ). 



We used these custom primers to amplify a por- 

 tion of the control region for 209 individuals in SO-pL 

 reaction volumes. The PCR products were run on 

 1.2% low-EEO (Fisher Scientific, Norcross, GA) aga- 

 rose gel during electrophoresis. The resulting bands 

 were excised and then purified with GeneClean (Bio 

 101, La Jolla, CA). Double-stranded sequencing was 

 conducted as described by O'Foighil et al. ( 1996). 



Data analyses 



Base composition, number of transitions (TSs), and 

 number of transversions (TVs) were determined by 

 using MEGA 1.01 (Kumar et al.-, 1993). Further 



analysis of base substitutions was conducted as in 

 Brown and Clegg (1983). Each different haplotype 

 was assigned a number, and the distribution of the dif- 

 ferent haplotypes was determined for each sample. 



We used MEGA to generate a pairwise matrix of 

 sequence divergence values between pairs of haplo- 

 types and to construct an unrooted neighbor-joining 

 tree; 200 replicates were used to estimate bootstrap 

 values for the nodes. Sequence divergences were 

 computed by using the pairwise-deletion option in 

 MEGA; this distance estimator excludes sites at 

 which indels occur on a pairwise basis. Haplotype 

 and nucleotide diversity within samples and nucleo- 

 tide divergence (D) between pairs of samples were 

 estimated according to Nei and Tajima (1981) and 

 Nei ( 1987 ) by using the DA option of REAP 4.0 ( McEl- 

 roy et al., 1992). The nucleotide divergence values 

 were clustered by using the NJTREE program (Jin 

 and Ferguson, 1990) based on the neighbor-joining 

 method of Saitou and Nei ( 1987 ). 



Geographic structuring of molecular variance 

 among samples was examined by using the matrix 

 of sequence divergences between all pairs of haplo- 

 types in AMOVA 1.55 (Excoffier et al., 1992). In this 

 analysis, the haplotype correlations ((^statistics) and 

 their variance components were estimated in a hier- 

 archical fashion: between regions, among samples 

 within a region, and among individuals within sam- 

 ples. Statistical significances of values were com- 

 puted by performing randomization tests with 500 

 replicates. Goldet al. ( 1993) concluded that red drum 

 was subdivided into Atlantic and Gulf of Mexico pop- 

 ulations. To determine the validity of this conclusion, 

 we examined the spatial partitioning of molecular 

 variance as follows. The between-region component 

 of variance and (J^.^ was first calculated for red dium 

 samples divided into Atlantic (SC, TP, PO, ML, 

 and IR) and Gulf (FB, CH, SA, OF, TB, and AP) 

 regions. The compositions of the two groups were 

 then adjusted by sequentially adding Atlantic sam- 

 ples to the Gulf group and then sequentially adding 

 Gulf samples to the Atlantic group. After each addi- 

 tif)n. the apportioning of molecular variance between 

 the resulting groups was recalculated. For example, 

 IR ( the Atlantic sample closest to the Gulf) was added 

 to the Gulf group and tested against the remaining 

 Atlantic samples (ML, PO, TP, and SC). ML (the 

 second closest sample) was then added to the Gulf- 

 plus-IR gi'oup; that grouping was then tested against 

 PO, TP, and SC. This process was repeated until only 

 a single sample remained in one group. 



Finally, to test for an association between interpopu- 

 lation D values and geogi-aphic distance (isolation-by- 

 distance), we performed the Mantel test (BIOMstat, 

 version 3.0; Rohlf and Slice, 1995) for samples gi-ouped 



