Gold and Richardson: Population structure of Seriola dumerili 
773 
TabSe 5 
Hierarchical analysis of molecular variation (AMOVA) among mtDNA haplotypes of greater amberjack (Seriola dumerili) from 
the Gulf of Mexico and U.S. southeastern Atlantic. 
Variance 
component 
Observed partition 
Variance 
% total 
P 1 
0 values 
Gulf + Florida Keys vs. Atlantic 
Between regions 
0.00127 
0.29 
0.259 
d> CT = 0.003 
Among samples within regions 
0.00160 
0.36 
0.190 
0 SC = 0.004 
Within samples 
0.43974 
99.35 
0.121 
<X> ST = 0.006 
Atlantic + Florida Keys vs. Gulf 
Between regions 
0.00398 
0.90 
0.001 
0 rT = 0.009 
Among samples within regions 
0.00011 
0.03 
0.444 
0 SC = 0.000 
Within samples 
0.43974 
99.07 
0.113 
0 ST = 0.009 
' Probability of finding a more extreme variance component by chance alone (5000 permutations). 
value of 0.009 in the pooled comparison of samples 
from the Atlantic and Florida Keys versus samples 
from the Gulf differed significantly from zero, 
whereas F ST values in all other comparisons were 
nonsignificant (Table 4). 
For AMOVA, where the variance in mtDNA 
haplotypes was partitioned by region, we also per- 
formed two separate analyses: one where the sample 
from the Florida Keys was included with Gulf 
samples, and one where the sample from the Florida 
Keys was included with Atlantic samples. In both 
analyses, the majority of the variance (>99%) was 
distributed within samples, and in both, the among- 
samples-within-groups component (<7> sc ) was nonsig- 
nificant (Table 5). The between-region component 
( & CT ) in the comparison of Atlantic and Florida Keys 
versus the Gulf was highly significant (P<0.001); 
whereas d> CT in the comparison of Atlantic versus the 
Gulf plus Florida Keys was not (Table 5). These re- 
sults paralleled results of the homogeneity tests and 
the F st values. Finally, we employed pairwise <f» ST 
values (i.e. between pairs of samples) to generate 
average “distance” values between samples from the 
Gulf (eight total), between samples from the Atlan- 
tic plus the sample from the Florida Keys (three to- 
tal), and between samples from the Gulf versus 
samples from the Atlantic (and the Florida Keys). 
The average (±SE) <? ST values for these comparisons 
were 0.004 ±0.001 (samples from the Gulf), 0.002 
±0.002 (samples from the Atlantic and the Florida 
Keys), and 0.012 ±0.002 (samples from the Gulf ver- 
sus samples from the Atlantic and the Florida Keys). 
Homogeneity (Vj tests were carried out on the ten 
haplotypes (1-4, 6, 11, 13, 23, 37, and 48) that were 
found in at least six individuals. Over all 11 samples, 
only haplotype 6 differed significantly (P-0.048) in 
frequency. This result is not significant when correc- 
tions for multiple testing (Rice, 1989) are applied. 
We then pooled samples into two groups (one that 
included the eight samples from the Gulf, and one 
that included the two samples from the Atlantic and 
the sample from the Florida Keys). Significant V tests 
(again, prior to corrections for multiple testing) were 
found for haplotype 6 (P-0.011) and haplotype 13 
(P-0.024). Frequency plots by sample of these two 
haplotypes (Fig. 2) revealed a clinal pattern to varia- 
tion in haplotype 6, with frequencies generally increas- 
ing eastwardly in the Gulf and into the Atlantic; no 
pattern was apparent for variation in haplotype 13. 
Maximum-parsimony (MP) analysis of individual 
mtDNA haplotypes and neighbor-joining ( N J ) analy- 
sis of the matrix of interpopulational (genetic) dis- 
tances revealed no evidence of phylogeographic struc- 
ture. MP analysis generated an unresolved multi- 
chotomous topology where a few clades of individual 
mtDNA haplotypes were supported at 90% or greater 
in bootstrap analysis (100 replicates). In each case, 
haplotypes within such clades were not geographi- 
cally cohesive. Examples included a clade of 
haplotypes 5, 52, and 47, which were found, respec- 
tively, in waters off Texas and west Florida (5), South 
Carolina (52), and the Florida Keys (47) (Table 2); a 
clade of haplotypes 32, 39, and 46, which were found, 
respectively, in waters off Texas and Louisiana (42), 
east Florida (39), and Texas and west Florida (46) 
(Table 2); and a clade of haplotypes 22, 27, 37, and 
51, which were found, respectively, in waters off west 
Florida (22), Texas (27), several localities (37), and 
west Florida and South Carolina (51) (Table 2). No 
“deep” subdivisions, in the sense of multiple restric- 
tion sites identifying a clade of haplotypes, were evi- 
dent. The NJ topology (Fig. 3) indicated relatively 
