74 
Fishery Bulletin 1 12(1) 
the repeated segment, and 3) at least 20 unambigu- 
ously sequenced base pairs were on both sides of the 
repeated motif, enabling primer design. Primers were 
constructed manually for the 40 resulting loci, and 
these primers were used to genotype a panel of 16 Gulf 
Menhaden individuals (from Matagorda Bay, TX) and 
8 Atlantic Menhaden individuals (from Bogue Sound, 
NC) by using the initial PCR conditions described 
above. Each primer set was labeled with a WellRed 
fluorescent dye (primer manufacture and labeling by 
Sigma-Aldrich Co.), and used to amplify approximate- 
ly 100 ng of genomic DNA through the use of PCR. 
Amplified products were combined with a 400-bp size 
standard and separated with a CEQ 8000 sequencer 
(Beckman Coulter, Inc.), with default (Frag-1) analy- 
sis parameters. Fragment analysis was performed with 
Beckman Coulter software, and all allelic bins (esti- 
mated allele size in base pairs) were established us- 
ing Beckman Coulter fragment analysis software, with 
default values for fragment analysis. 
The following qualitative criteria were used to 
choose a final group of microsatellites from the 40 
original loci: 1) primer pairs amplified a product in 
a majority of Gulf Menhaden individuals, 2) alleles 
were spaced as expected on the basis of the original 
clone sequences (i.e., trinucleotide motifs were 3 bases 
apart, tetranucleotide motifs were 4 bases apart), 3) 
peak intensity appeared to be similar between het- 
erozygote peaks (i.e., there was no visual evidence for 
allele dropout), 4) alleles were scored with minimum 
stutter, and 5) preference was given to loci that pro- 
duced PCR amplicons in both species. This elimination 
process resulted in 14 loci retained from the 40 origi- 
nal loci examined. An additional 5 loci initially devel- 
oped for American Shad ( Alosa sapidissima) (Waters 
et ah, 2000) and previously assayed in Brevoortia spp. 
(Anderson, 2007) were included in the final group to 
increase the sampling size of the group and to com- 
pare the performance of the new loci to a previously 
published set. All sampled individuals were genotyped 
by using these 19 loci. 
Micro-Checker and genetic polymorphism 
The software program Micro-Checker (Van Oosterhout 
et ah, 2004) was used to test each locus for evidence of 
null alleles and genotyping errors. Each marker was 
tested independently in all 4 endemic North American 
species of Brevoortia : Gulf Menhaden, Atlantic Men- 
haden, Yellowfin Menhaden, and Finescale Menhaden. 
Data from each of the 4 Gulf Menhaden sampling lo- 
cales also were treated as independent samples and 
analyzed separately. If evidence was found for a null al- 
lele in any sample and locus combination, the frequency 
of the null allele was calculated as the average of the 
estimates from 3 independent methods (Chakraborty et 
al., 1992; Brookfield, 1996; Van Oosterhout et ah, 2006). 
Micro-Checker also was used to assess the statistical 
likelihood of genotyping errors caused by either allele 
stutter (one or more PCR artifacts being genotyped as 
true peaks) or allele dropout (alleles that amplify poor- 
ly and, therefore, are not properly genotyped). 
The likelihood ratio test of Excoffier and Slatkin 
(1998) was used to test for linkage disequilibrium be- 
tween all pairs of microsatellite loci in Gulf Menhaden. 
Testing was done through the use of the software pro- 
gram Arlequin (vers. 3.5; Excoffier et al., 2005) and sta- 
tistical significance assessed with 1000 data permuta- 
tions. The level of statistical significance was adjusted 
for each comparison of linkage disequilibrium by using 
the sequential Bonferroni procedure (Rice, 1989). Each 
marker also was tested for deviation from Hardy-Wein- 
berg equilibrium (HWE), with the expectation that loci 
with null alleles would likely also deviate from HWE. 
The exact test method of Guo and Thompson (1992) 
was used for HWE tests with a Markov chain length of 
100,000 steps, as implemented in Arlequin. After the 
exact test, statistical significance was assessed at each 
locus with the P-value of the test statistic, and statisti- 
cal significance is reported here both before and after 
sequential Bonferroni adjustment (Rice, 1989). 
The observed and expected heterozygosity and the 
number of alleles also were estimated for each locus, 
in each species, with Arlequin. Differences in genetic 
variability among species were tested by comparing ex- 
pected heterozygosity (H e ) among species. Each locus 
was treated as an independent data point, and the null 
hypothesis that H e in Gulf Menhaden was not statisti- 
cally different from H e in other species of Brevoortia 
was tested with a Wiicoxon nonparametric signed-rank 
test over all loci. The Wiicoxon test was performed in 
R software (vers. 2.14.1; R Development Core Team, 
2011), with the wilcox.test function. To further com- 
pare levels of polymorphism across loci and species, 
the polymorphic information content (PIC; Botstein et 
al., 1980) of each marker was estimated from observed 
allele frequencies in each species. The PIC statistic is 
a function of both allele number and frequency and, 
therefore, is a better estimator of discriminatory power 
than is the number of alleles alone. 
The PIC was calculated in R for each locus with the 
formula 
PICj = 1 - S A 2 , 
i=i 
where p - the frequency of allele i at locus j. 
The Wiicoxon nonparametric signed-rank test was 
used to determine whether Gulf Menhaden had signifi- 
cantly higher PIC scores across all loci than the other 
Brevoortia species. 
The relative power of assignment of Brevoortia gene 
loci was tested with the program Whichloci (Banks et 
al., 2003). The Whichloci method is a resampling tech- 
nique that generates simulated populations from ob- 
served allele frequencies in experimental samples, and 
then assigns experimental individuals to populations 
on the basis of the likelihood of an individual’s geno- 
type in each population. The accuracy of assignment 
