Carson et al.: Population structure, long-term connectivity, and effective size of Lutjanus analis 
419 
(Applied Biosystems) and sequences were aligned and 
edited in Sequencher, vers. 4.0 (Gene Codes Corp., Ann 
Arbor, MI). 
For microsatellites, summary statistics, including 
number of alleles, allelic richness, unbiased gene di- 
versity (expected heterozygosity), and the inbreeding 
coefficient F IS , measured as f of Weir and Cockerham 
(1984), were generated in FSTAT (Goudet, 1995; vers. 
2. 9. 3. 2, http://www2.unil.ch/popgen/softwares/fstat.htm, 
accessed July 2011). Homogeneity in allelic richness and 
gene diversity among the five locations were assessed 
by using Friedman rank tests, with SPSS software 
(http://www-01.ibm.com/software/analytics/spss/prod- 
ucts/statistics/, accessed July 2011). Departure from 
Hardy-Weinberg (HW) equilibrium expectations for 
each locality was tested with exact probability tests in 
Genepop (Raymond and Rousset, 1995; vers. 3.4, http:// 
genepop.curtin.edu.au/, accessed March 2011), by using 
a Markov Chain approach (Guo and Thompson, 1992), 
with 10,000 dememorizations, 500 batches, and 5000 
iterations per batch. Global and pairwise (between lo- 
calities) exact tests of homogeneity of allelic (genic) and 
genotypic distributions also were conducted in Genepop; 
genetic homogeneity among locations was further tested 
by using analysis of molecular variance (AMOVA), with 
the program Arlequin (Excoffier and Lischer, 2010; vers. 
3. 5. 1.2, http://cmpg.unibe.ch/software/arlequin35/, ac- 
cessed March 2011). Sequential Bonferroni correction 
(Rice, 1989) of P values was applied for all simultane- 
ous tests. Microchecker (van Oosterhout et al., 2004) 
was used to determine whether genotype scores at each 
locus were compromised by the presence of null alleles, 
stuttering, or genotyping errors. 
For mitochondrial ND-4 DNA sequences, number of 
haplotypes and haplotype diversity were determined by 
using DnaSP, vers. 5.10.01 (Rozas et al., 2003; http:// 
www.ub.edu/dnasp/, accessed March 2011); haplotype 
richness and nucleotide diversity for each sample lo- 
cality were estimated following El Mousadik and Petit 
(1996) and using the software Rarefac (available at 
http://www.pierroton.inra.fr/genetics/labo/Software/ 
Rarefac/, accessed July 2011) and Arlequin, respectively. 
A bootstrap resampling method (Dowling et al., 1996) 
was used to test homogeneity of haplotype number and 
diversity among localities. The observed haplotype num- 
ber and diversity at each locality was compared to the 
distribution in 1000 bootstrap samples of a comparable 
size drawn from the entire population (all samples 
pooled); resampling was conducted in PopTools, a free 
add-in to Microsoft Excel and available at http://www. 
poptools.org/, accessed July 2011. Differences in average 
nucleotide diversity were considered significant if pair- 
wise comparisons differed by more than two standard 
errors. Homogeneity in mtDNA haplotype distributions 
among localities was tested using global exact tests and 
analysis of molecular variance (AMOVA), as implement- 
ed in Arlequin. Pair-wise (between locations) estimates 
of & ST , an analogue of F ST , were generated by using 
Arlequin; & ST estimates were based on pair-wise genetic 
distances, with significance determined by exact tests 
(Raymond and Rousset, 1995; Goudet et al., 1996), as 
implemented in Arlequin. Selective neutrality of mtD- 
NA variation in each sample was tested by calculating 
Fu’s (1997) F s statistic and Fu and Li’s (1993) D* and 
F* statistics, as implemented in the DnaSP program. 
Significance of F s , D*, and F* was assessed in FSTAT 
by using 10,000 coalescent simulations (after Rozas et 
al., 2003) based on the observed number of segregating 
sites in each sample. 
The coalescent-based program Migrate (Beerli and 
Felsenstein 2001; vers. 3.2.6, http://popgen.sc.fsu.edu/ 
Migrate/Migrate-n.html, accessed July 2011) was used 
to generate maximum-likelihood estimates of both the 
average long-term (mutation-scaled) migration rate (M) 
between pairs of localities and the parameter theta (0) 
for each locality. Because the model used in Migrate 
explicitly accounts for migration, estimates of 0 can be 
generated for individual subpopulations within a larger 
metapopulation (Waples, 2010). The average long-term 
migration rate (m) between localities was estimated 
as M - mlj. 1 , where pi is the average, per gene muta- 
tion rate. Values of theta (4 N e ji) were used to estimate 
average long-term effective population size (N e ) of each 
locality. To obtain estimates of m and N e , the modal 
mutation rate (//) of the microsatellite data set was 
obtained by using the Bayesian coalescent approach 
of Beaumont (1999) and Storz and Beaumont (2002) 
and the software MSVAR (vers. 0.4.1b, http://www.ru- 
bic.rdg.ac.uk/~mab/software.html, accessed July 2011). 
Because simulations in Migrate are computationally 
demanding, all parameter estimates were based on a 
random sample of 25 individuals from each location 
(125 individuals total). A preliminary analysis (short 
run) established initial estimates for both M and 0 
that were then used as starting values in final long 
runs. Parameter estimates were obtained by averaging 
three replicate long runs that included 40 short Monte 
Carlo Markov chains (MCMC, 10 4 gene trees sampled) 
and three long chains (2.5xl0 6 gene trees sampled). To 
ensure parameter stability, the first lxlO 4 steps of each 
chain were discarded as burn-in. 
Results 
Summary statistics for microsatellites are presented in 
Table 1. Microsatellite Lco64 was monomorphic for a 
111-bp allele (scored with primers as 151 bp) and was 
excluded from further analysis. Of the remaining 16 
microsatellites, the number of alleles ranged from 2-3 
at Lsy 8 to 19-25 at Prs248. Allelic richness ranged from 
2.00-2.86 at Lsy8 and from 19.00-23.67 at Prs 248, and 
expected (unbiased) gene diversity ranged from 0.050- 
0.118 at Lsy8 and from 0.860-0.901 at Lanll. Across all 
microsatellites and localities, number of alleles, allelic 
richness, and gene diversity averaged (±standard error 
[SE] ) 10.05 (0.26), 9.80 (0.22), and 0.594 (0.01), respec- 
tively. No significant differences (Freidman’s rank tests) 
in allelic richness (P= 0.232) or gene diversity (P= 0.373) 
were found among localities. 
