886 



Fishery Bulletin 97(4), 1999 



ware) and south (North CaroHna and Florida) of this 

 locaHty. Phylogenetic analysis was employed by using 

 the character-state approach to construct a maximally 

 parsimonious network relating individual haplotypes 

 based on the number of restriction-site differences 

 (Avise, 1994). The resulting network was examined for 

 geographic structuring of haplotypes within and among 

 sampling regions. 



Eopulation structure in M. undulatus was also 

 evaluated by using nested analysis of molecular vari- 

 ance (AM OVA [Excoffier et al., 1992] ). AMOVA input 

 consisted of an Euclidean distance matrix contain- 

 ing genetic distance values for all possible pairs of 

 the 15 observed mtDNA haplotypes. Samples were 

 stratified by locality (DE, NC, FL, and LA) and nested 

 within region (Atlantic or GulD. Total genetic varia- 

 tion was partitioned into three components: within 

 geographic localities, among geographic localities 

 within the Atlantic region, and between regions. Sig- 

 nificance of variance components was tested by us- 

 ing 10,000 random permutations to generate null 

 distributions for each variance component. Gene flow 

 among localities was estimated with Wright's (1943) 

 island model, modified for mtDNA: Fg-j. = y2N.,m +1. 

 Calculations employed 0gj as an approximation oiF^^. 



Results 



RFLP analysis of the ATPase 6 and D-loop regions 

 revealed a total of 68 restriction sites. The average 

 individual was scored for 62 sites and 264 nucleotide 

 positions representing approximately 12% of the 

 ATPase 6 and D-loop regions combined, or 1.6% of 

 the entire mitochondrial genome. A total of 15 com- 

 posite haplotypes was detected in M. undulatus 

 (Table 2). Of the 93 individuals surveyed, 67 (72%) 

 shared the same composite genotype (haplotype 1). 

 Genetic distance among haplotypes averaged 0.17% 

 for the pooled sample, with the most divergent 

 haplotypes (10 and 12) differing from the common 

 haplotype by only 0.75% and 1.0%, respectively. 



Nucleon diversity (/;) averaged 0.47 ±0.01 (mean 

 ±SE) for the pooled sample, and ranged from h = 0.25 

 ±0.12 in DE to /z = 0.62 ±0.12 in LA. Nucleotide se- 

 quence diversity also varied geographically, ranging 

 from K = 0.07% in DE to ;r = 0.32% in LA. Net se- 

 quence divergence (p) among localities was low, in- 

 dicating that most of the observed mtDNA variation 

 occurred within localities. Average sequence diver- 

 gence for the pooled sample was p = 0.004% , with 

 comparisons among Atlantic localities yielding 

 slightly lower divergence values than comparisons 

 involving Gulf and Atlantic samples. Croaker col- 

 lected north versus south of Cape Hatteras were ge- 



netically indistinguishable (p=-0.002%) whereas 

 nucleotide divergence between regions (pooled Atlan- 

 tic vs. Gulf) was 0.005%. 



The common mtDNA genotype (haplotype 1) was 

 numerically dominant at all localities (Table 2). The 

 remaining 14 haplotypes occurred at low (<10%) fre- 

 quencies. Monte Carlo tests for homogeneity revealed 

 no significant heterogeneity in mtDNA haplotype 

 frequency distributions within the SAB (NC vs. FL, 

 /-=11.3, p=0.140) or among all Atlantic localities 

 (p=0.148). When NC and FL samples were pooled 

 and compared to the DE sample, there was no indi- 

 cation of heterogeneity among samples collected 

 north and south of Cape Hatteras (^"=4.6, p=0.906). 

 To further test for population structure within the 

 Atlantic region, an additional 20 specimens from each 

 Atlantic locality were screened for D-loop variation 

 with the polymorphic enzymes Taql and Hhal. Nei- 

 ther Taql (x''=3.45, p=0.562) nor Hhal (x-=1.95. 

 p=0.682) haplotype frequencies exhibited significant 

 heterogeneity among Atlantic localities. 



Given the lack of heterogeneity within the Atlan- 

 tic region, composite haplotype frequencies were 

 pooled across Atlantic localities and compared with 

 the Gulf of Mexico sample. Monte Carlo simulations 



