NOTE Blandon et al.: Genetic population structure of Parcilichthys lethostigma 
673 
the tree. The strength of support for each node in the 
tree was tested by bootstrapping over loci with NJBPOP 
(Cornuet et ah, 1999). To further quantify spatial hetero- 
geneity, the fixation index (F ST ) was calculated for each lo- 
cus to provide measures of interpopulation differentiation 
and estimates of reduction in heterozygosity of a subpopu- 
lation due to population subdivision. A % 2 test was used 
to test the null hypothesis, F ST = 0 (Workman and Niswan- 
der, 1970). All / 2 probability values from tests for con- 
formance to Hardy- Weinberg expectations, heterogeneity, 
linkage disequilibrium, and F ST were adjusted for multiple 
simultaneous tablewide tests by using sequential Bonfer- 
roni adjustments to minimize type-I statistical inference 
errors (Rice, 1989). 
Partitioning of variance components among geographic 
regions and within samples was accomplished by using 
a hierarchical analysis of molecular variance (AMOVA, 
Cockerham, 1969, 1973) with the package ARLEQUIN 
(version 2, Schneider et ah, 1999). Sample sites were nest- 
ed into regional groupings for separate analyses (Atlantic 
versus Gulf of Mexico, and Atlantic combined with east- 
ern Gulf sites versus western Gulf). A phenogram was 
generated from the chord distance matrix with the neigh- 
bor-joining (N-J) algorithm. The N-J phenogram, with 
bootstrap estimates (as percentage of 10,000 replications) 
obtained by resampling loci within samples, was gener- 
ated with NJBPOP (Cornuet, et al., 1999). The signifi- 
cance of the relationship between genetic (i.e. chord) and 
geographic (bay to bay shoreline distance) distance matri- 
ces was determined by sampling the randomization dis- 
tribution generated from 1000 replications with the MX- 
COMP (matrix comparison) routine in NTSYS-PC 2.0 by 
Rohlf (1997) to allow a Mantel test (Mantel, 1967). An 
assignment test (WHICHRUN 4.1, Banks and Eichert, 
2000) tested the ability to discriminate population of ori- 
gin based on an individual’s multilocus genotypic profile. 
Clinal trends in heterozygosity and allele frequency 
were examined by using nonparametric correlation analy- 
ses (SAS Institute, 1989). Significance of Spearman’s cor- 
relation coefficients was determined as the probability a 
correlation differed from zero. Probabilities less than 0.05 
were considered statistically significant (Snedecor and Co- 
chran, 1980). 
Results 
Misidentifications detected by IEF resulted in a reduced 
sample size in some samples especially from Alabama and 
Florida. Examination of 46 enzymes and structural pro- 
tein systems in southern flounder produced scorable phe- 
notypes for 68 putative gene loci. Two dimeric esterase 
loci (ESTD-V* and ESTD 2*. 3.1.1.. IIUBMBNC, 19921 ) a 
tripeptide aminopeptidase locus ( PEPB-2* , 3.4.13..), glyc- 
erol-3-phosphate dehydrogenase (G3PDH *, 1.1. 1.8), two 
glucose-6-phosphate isomerase loci ( GPPA and GPPB \ 
5.3. 1.9), phosphoglucomutase (PGM*, 5. 4. 2. 2) and two 
glucose-6-phosphate dehydrogenase loci (G6PDH-1* and 
G6PDH-2* , 1.1. 1.49) were resolved, scored as variable, and 
included in analyses. The remaining 59 loci were mono- 
morphic or could not be scored consistently and were omit- 
ted. All polymorphic loci bad the same common allele 
across all localities (Table 1). ESTD-1*, ESTD-2 *, GPPB*, 
and G6PD-1* each expressed an allele unique to a single 
locality. The percentage of polymorphic loci ( P 0 99 ) averaged 
7.5% and ranged from 4.41% to 10.29% (Table 1). Mean 
individual heterozygosity ranged from 0.03 (SE=0.02) in 
North Carolina and Florida to 0.12 (SE=0.12) in Matago- 
rda Bay. Statistically significant clinal relationships were 
found in the Gulf of Mexico between the frequency of 
the common allele of the G6PDH-2* locus and degrees 
west longitude (r s =-0.829, P<0.05) and degrees north lat- 
itude (r s =0.829, P<0.05). Further differentiation of popu- 
lations in the western Gulf of Mexico was observed at 
the G3PDH* locus. The G3PDH*1 allele occurred at a fre- 
quency of 12.0%) in the Laguna Madre, declined to a fre- 
quency of 1.4% in Galveston Bay, and was absent in all 
collections east of Galveston Bay. Four loci had alleles con- 
fined to a single locality in this same region of Gulf of 
Mexico (Table 1), including G6PDH-1*2 which was limited 
to the Laguna Madre. Samples were in Hardy-Weinberg 
equilibrium at the nine allozyme loci surveyed for each 
of the 72 possible comparisons of variable loci for each 
locality, except ESTD-2 * in Galveston Bay. No statistically 
significant genotypic linkage disequilibrium was observed 
between any loci for any population. 
Tests for homogeneity of allelic frequencies at variable 
loci across all localities were statistically significant at 
six of the nine loci: ESTD-2 * (PcO.Ol), G3PDH * (P<0.01), 
GPPA (P=0.01), GPPB* (PcO.Ol), PGM* (PcO.Ol), and 
G6PDH*-2 (P<0.01). Tests for homogeneity of genotypic 
frequencies across all localities were significant at four 
loci: G3PD1P (P<0.01), GPPB* (P<0.01), PGM* (PcO.Ol), 
and G6PDH-2* (PcO.Ol). F ST averaged 0.088 and varied 
from 0.181 for G6PDH-2* to 0.001 for G6PDH-1*. All esti- 
mates of Fst were found to be statistically different from 
0, except for loci G6PDH-1 *, PEPB-2 *, and GPPA*. 
Regional differentiation among southern flounder was 
demonstrated (Fig. 2) by using chord distance and N-J 
clustering. The greatest discontinuity was between all 
samples from Galveston Bay eastward and a cluster com- 
posed of Matagorda Bay and the Laguna Madre. No ap- 
parent differentiation exists between Atlantic Coast pop- 
ulations and populations collected from the eastern Gulf 
of Mexico. This pattern is supported by the outcome of 
the assignment test (Table 2). Percentage correct assign- 
ment ranged from 0% for Florida to 78% for Matagorda 
Bay. However, when assignments between groups identi- 
fied in the cluster analysis were examined, the overall 
correct percentage increased to 81% . The assignment test 
supports the interpretation of the cluster analysis, sug- 
gesting within-region differentiation was minimal and not 
geographically consistent, but between-region differences 
were considerable. A different pattern was discerned by 
using AMOVA (Table 3). The majority of the variance 
(>99%) was distributed within samples and the among- 
sites-within-coasts component was nonsignificant. Com- 
parison of Atlantic versus Gulf of Mexico sample sites was 
statistically significant (P=0.04) and comparison of west- 
ern Gulf of Mexico sample sites with those from the east- 
