Anderson; Systematlcs of the North American menhadens in the genus Brevoortia 



371 



where F = multilocus mean homozygosity (1 - H^). 



Although estimates of iV^ from small genetic samples are 

 notoriously inaccurate and may do a poor job of reflecting 

 overall census size, relative estimates of A^^, have been 

 used to test evolutionary hypotheses (Nielsen, 1997), 

 and theoretically may be used to examine relative dif- 

 ferences in multiple populations at the same loci. The 

 assumptions of the present study are that the mutation 

 rate (li) is not significantly different among the four 

 examined species, and that genetic variability is related 

 to population size. For this analysis, four loci were used 

 because of the highly significant deviation from equilib- 

 rium frequencies that was observed at locus AF39660 

 (see "Results" section). 



Sequencing data set 



Sequencing templates were generated with PCR by using 

 primers specific to an 840-bp fragment of the mtDNA 

 control region. The heavy strand primer ( HN20 1 was a 

 previously described universal primer (Bernatchez and 

 Danzmann, 1993) (5'-CGGGGTTTGACATGAATAT), 

 whereas the light strand primer (940r) was a novel 

 sequence primer designed specifically for menhaden 

 (3-TGTAATACTAGTGCGCAGATGGTAC). Each sample 

 was amplified twice, and the amplicons were combined 

 before purification to assure a high quality, concentrated 

 product. PCR products were purified by using QIAquick® 

 PCR purification kits (Qiagen Inc., Valencia, CA). The 

 resulting purified fragments were used in two sequenc- 

 ing reactions in a nested design. The novel heavy strand 

 primer (520f) was internal (5-GGAACCAAATGCCAG- 

 GAATAGT), whereas the light strand primer was the 

 same used in PCR. This nested design resulted in frag- 

 ments which overlapped one another for 360 bp and 

 which resided within the original 840-bp PCR fragment. 

 Sequences were electrophoresed and analyzed on a Beck- 

 man-Coulter CEQ"' 8000 capillary sequencer by using 

 default module (LFR-1) parameters (Beckman Coul- 

 ter. Inc., Fullerton, CA). Raw sequences were trimmed 

 and edited, and forward and reverse sequences were 

 conjoined by using the software package Sequencher"' 

 vers. 4.2 (Gene Codes Corp., Ann Arbor, MI). Whole 

 sequences were then aligned by using Clustal X freeware 

 (Jeanmougin et al., 1998). Sequences were submitted to 

 GenBank by means of Sequin freeware (National Center 

 for Biotechnology Information, Bethesda, MD) as batch 

 submissions (accession numbers EF119342-EF119454). 

 Sequence data were obtained from 113 individuals 

 from the four species (fi. tyrannus, n=37; B. smithi. 

 71=32; B. patronus, ?z=30; B. gunteri, n=14), and samples 

 were selected to represent the full geographic range of 

 each species (sampling locations in Maine, New Jersey, 

 Virginia, North Carolina, Atlantic and Gulf Florida, 

 and Texas). Sequence statistics, including nucleotide 

 diversity (;r) and haplotype diversity (h), were estimated 

 for each species, and the genetic similarity among popu- 

 lations (measured using D^, Nei, 1987) was calculated 

 by using the freeware program DnaSP 4.0 (Rozas et al., 



2003). The variances of sequence diversity estimates 

 were used to construct 95"* confidence intervals around 

 each mean. The averaged frequency of each base and 

 the estimated ratio of transitions to transversions (ts/tv) 

 were calculated with the freeware program DAMBE 

 (Data Analysis in Molecular Biology and Evolution, 

 Xia and Xie, 2001). Because of the extreme level of 

 variation found in menhaden mtDNA, both in this study 

 and in a previous study (Bowen and Avise, 1990), we 

 employed a mutation saturation test as instituted in 

 DAMBE. With this analysis, we compared the genetic 

 distance between sequences to the number of transi- 

 tions and transversions occurring between them. When 

 all pairwise comparisons of sequences were plotted in 

 this manner, mutation saturation was indicated by a 

 plateau at which transversions approach or outnumber 

 transitions at higher genetic distances. The genetic 

 distance employed for this analysis {K2P) is described 

 in Kimura (1980). 



Both transitions and transversions were included in 

 phylogenetic analyses. Before evolutionary reconstruc- 

 tion of haplotypes, the freeware program Modeltest 3.7 

 (Posada and Crandall, 1998) was used to determine 

 the model of sequence evolution which had the highest 

 likelihood. Briefly, 56 nested models of evolution were 

 tested against raw sequence data with the program 

 PAUP 4. Obi (Phylogenetic Analysis Using Parsimony, 

 available from Sinauer Associates, Inc., Sunderland, 

 MA). Modeltest was then used to perform hierarchical 

 likelihood ratio testing (hLRT) in order to identify the 

 model with the highest likelihood. Concurrently, the 

 Akaike information criterion (AIC) weights of the four 

 most likely models were also examined during model se- 

 lection. The appropriate model was subsequently used in 

 a maximum-likelihood (ML) procedure in PAUP 4. Obi, 

 with 10 replicates, to reconstruct an unrooted haplotype 

 phylogeny. Because of the enormous computational load 

 associated with maximum likelihood of highly vari- 

 able loci, a concurrent N-J tree was constructed by 

 using the 2-parameter distance of Kimura (1980); this 

 tree was bootstrapped (1000 replicates over nucleotides, 

 Felsenstein, 1985) to evaluate the significance of major 

 interior nodes that correlated with nodes on the maxi- 

 mum-likelihood tree. 



Results 



Microsatellite data set 



Microsatellite samples were distributed from southern 

 Texas to the Gulf of Maine, the northern extent of B. 

 tyrannus. Brevoortia tyrannus was sampled in three loca- 

 tions (all Atlantic coast), B. smithi in two locations (one 

 Atlantic, one eastern Gulf), B. patronus in 14 locations 

 (all Gulf coast), and B. gunteri in three locations (all 

 western Gulf coast). Of the five microsatellite markers 

 used in this study, all had observed levels of heterozy- 

 gosity that were lower than anticipated under Hardy- 

 Weinberg expectations when tested by using samples 



