Anderson and Karel: Limited genetic structure of Brevoortia patronus revealed by microsatellite markers 
75 
can be assessed at any level of assignment stringency, 
and that level is set by the user as a LOD (log odds) 
score. Individuals from different samples within species 
were combined, and the relative power of the loci in the 
assignment of individuals to species was assessed. Loci 
that deviated from HWE or that failed to amplify any 
species were excluded from species assignment. As a 
result, assignment was performed with 14 loci (see the 
Results section). 
During species assignment, marker performance 
was assessed at various levels of LOD=2, 3, and 4; 
these levels of stringency are approximately analogous 
to an acceptable error rate of 0.01, 0.001, and 0.0001, 
respectively. Simulated population sizes were set at 
n=100, and 100 replicate simulations were performed. 
Two properties of the genetic data set were assessed 
after simulation and assignment: 1) the number of loci 
needed to correctly assign 100% of individuals to spe- 
cies at the predetermined level of stringency and 2) the 
relative contribution (locus score; Banks et al., 2003) 
of each locus in the assignment of individuals when 
LOD was set at 1. It is anticipated that within the 
menhaden fishery of the Gulf of Mexico, the Atlantic 
Menhaden is rarely or never encountered. Therefore, 
the Atlantic Menhaden was excluded from the analysis. 
The power of these loci to reveal small levels of 
population differentiation within Gulf Menhaden was 
explored through a Powsim analysis (Ryman and Palm, 
2006). Allele frequencies were estimated from compos- 
ite data in the Gulf Menhaden samples with Fstat, 
vers. 2.3, (Goudet, 1995) and were used to perform 500 
replicated experiments on 2 simulated populations. 
Because of the uncertainty involved in accurately esti- 
mating and accounting for null alleles, loci that deviat- 
ed significantly from HWE in Gulf Menhaden and had 
estimated null allele frequency p>0.10 were excluded. 
As a result, power analysis was performed with 15 loci 
(see the Results section). 
The 2-population model was chosen because it is 
the simplest mode of population structure, and previ- 
ous studies of marine finfish in the Gulf of Mexico have 
frequently indicated a marine “suture zone” east of the 
Mississippi River (Portnoy and Gold, 2012), resulting 
in eastern and western Gulf of Mexico stocks of several 
species or species pairs. Two levels of sampling (n=40 
and 50 individuals) were conducted after 10 genera- 
tions of genetic drift, with the assumption of effective 
population sizes of 500, 1000, and 5000 individuals 
(these values equate to F st =0.01, 0.004, and 0.001, re- 
spectively). The power of the data at each level of N e 
and at each sample size was assessed as the number of 
significant results obtained across all 500 simulations, 
assessed with the Fisher’s exact test. 
Genetic structure of Gulf Menhaden 
Among the 4 Gulf Menhaden samples, evidence for 
nonrandom associations of alleles or genotypes (genet- 
ic structure) was assessed with the program Structure 
(vers. 2.3.4; Pritchard et al., 2000). Models that includ- 
ed K= 1-4 potential populations were evaluated through 
comparison of the posterior probability of each model 
(where K is the prior value for the assumed number of 
populations represented by the data). Each level of K 
was assessed with 50,000 burn-in iterations, followed 
by 450,000 sampling iterations. Prior evaluations of 
Gulf Menhaden have indicated a high level of gene flow 
among samples from the Gulf of Mexico; therefore, we 
used the default Structure model, assuming historical 
admixture at each level of K, while also assuming al- 
lele frequencies among locales were correlated. Admix- 
ture coefficient stability among individuals during the 
course of runs in Structure was used to evaluate model 
convergence. Loci that deviated from HWE or that had 
null alleles of estimated frequency p>0.10 were exclud- 
ed from all genetic structure analyses. 
Genetic structure was further explored by using the 
analysis of molecular variance (AMOVA; Excoffier et 
al., 1992). The initial AMOVA model was used to test 
for significant genetic divergence among Gulf Menha- 
den samples (F st ), and secondarily among individuals 
within samples (F; s ). The significance of variance com- 
ponents in the AMOVA model were evaluated by 1000 
permutations of the data, and F-statistics were esti- 
mated directly from variance components. 
Finally, a Mantel test was used to assess the evi- 
dence of isolation-by-distance (IBD) among samples. 
Geographic association between sampling areas were 
calculated among sites by using shortest linear shore- 
line distance. This distance was correlated with Slat- 
kin’s linearized F st (Slatkin, 1995) by using the Mantel 
test. The significance of the coefficient of matrix cor- 
respondence (r) was assessed by 1000 permutations of 
the data. The Mantel test and permutation procedure 
were conducted with Arlequin. 
Results 
The original 40 loci that were examined by means of 
the microsatellite discovery phase have been deposited 
in the National Center for Biotechnology Information 
Genbank database (accession numbers KC331110- 
KC331149). The 14 loci that were retained from the 40 
original loci through the use of qualitative criteria are 
listed in Table 1, and they will be discussed hereafter, 
along with 5 previously described loci (Waters et al., 
2000). Among this final group of microsatellites, there 
were 6 tetranucleotide repeats, 9 trinucleotide repeats 
(1 trinucleotide repeat was initially described in Wa- 
ters et al., 2000), and 4 dinucleotide repeats (all di- 
nucleotide repeats were initially described in Waters et 
al., 2000). Two of the tetranucleotide repeats unexpect- 
edly contained alleles that were 2 bases apart (BP017, 
BP039). These alleles were not sequenced directly, but, 
in both cases, there were very short dinucleotide re- 
peats adjacent to the targeted tetranucleotide repeats 
in the original clone sequence. It is likely that, in both 
