94 
Fishery Bulletin 1080) 
Typically, one would not consider using the mitochon- 
drial COI gene region for an analysis of stock struc- 
ture because this locus tends to be highly conserved in 
most organisms, exhibiting low levels of intraspecific 
variation (Meyer, 1993). However, previous studies have 
documented very high levels of variation throughout the 
Atlantic menhaden mitochondrial genome (Avise et al., 
1989; Bowen and Avise, 1990; Anderson, 2007), and in a 
preliminary analysis of various menhaden mitochondrial 
gene regions, we found COI to be sufficiently variable for 
an analysis of stock structure. The COI genetic diversity 
in Atlantic menhaden (jt= 2.74%) is an order of magni- 
tude higher than the average within-species divergence 
reported for other fishes. For example, Ward et al. (2005) 
reported an average n of 0.39% for Australian marine 
fishes, and Hubert et al. (2008) reported a n of 0.302% 
for Canadian freshwater fishes. 
There were significant differences in the distribution 
of COI haploytpes and microsatellite allele frequencies 
between Atlantic and Gulf menhaden, although both 
classes of markers indicate the two species are very 
closely related (F ST =0.104). These results are consistent 
with those of Anderson (2007) who surveyed variation 
at four microsatellite loci, estimating an F ST of 0.110 
between Atlantic and Gulf menhaden. These F ST values 
are more typical of differences between populations than 
species. For comparison, F ST values between genetically 
distinct stocks of clupeid fishes based on microsatellites 
range from 0.002 to 0.226 (Shaw et al., 1999; Sugaya et 
al., 2008), and are approximately one-fourth of the ^st 
values between other pairs of North American menha- 
dens (0.355-0.488; Anderson, 2007). 
A low level of genetic divergence between Atlantic and 
Gulf menhaden was also noted for the mitochondrial 
COI gene region (<J> gT = 0.178) — a result consistent with 
a previous restriction fragment length polymorphism 
(RFLP) analysis of the mitochondrial genome (Avise et 
al., 1989) and sequence analysis of the control region 
(Anderson, 2007) of these two species. In the present 
study, we did not find a single COI nucleotide position 
that distinguished Atlantic from Gulf menhaden. In 
a survey of 207 fishes, including several congeners, 
Ward et al. (2005) reported that all had different COI 
sequences. Furthermore, mean nucleotide differences 
between closely related species were 25 times higher (on 
average) than differences within species. In the present 
study, however, the nucleotide sequence diversity for At- 
lantic and Gulf menhaden combined (jt= 0.0258) was less 
than that for Atlantic menhaden alone (jt=0.0274). 
When compared with Gulf menhaden, Atlantic men- 
haden are generally larger, have a less convex body 
shape, and have a higher number of predorsal scales, 
vertebrae, and ventral scutes (Bigelow et al., 1963). 
Although the mean values of some of the morphomet- 
ric and meristic characters are significantly different 
between the two species, the ranges of variation are 
coincident (Dahlberg, 1970). Although Atlantic menha- 
den and Gulf menhaden are morphologically similar, 
their geographic ranges are not believed to overlap 
(Bigelow et al., 1963). Thus, the species are typically 
distinguished by capture location. The high level of ge- 
netic and morphological similarity of Atlantic and Gulf 
menhaden raises concern over the validity of the two 
species. A thorough morphological and genetic analysis 
of the same individuals will be required to resolve this 
problem. 
In the present analysis of Atlantic and Gulf men- 
haden mitochondrial COI gene region sequences, two 
distinct mitochondrial clades were noted, one of which 
was found only in Atlantic menhaden, and the other in 
both Atlantic and Gulf menhaden. These results are 
similar to those found in a previous RFLP analysis of 
the whole mitochondrial genome (Avise et al., 1989) and 
in a sequence analysis of the control region (Anderson, 
2007). Avise (1992) hypothesized that the separation of 
two mitochondrial clades between the Atlantic Ocean 
and the Gulf of Mexico was a result of historical iso- 
lation of Atlantic and Gulf menhaden by the Florida 
peninsula during times of cooler water temperatures 
and subsequent unidirectional gene flow during geo- 
logically recent times. Anderson (2007) postulated that 
the distribution of these two clades in Atlantic menha- 
den supported very recent gene flow between Atlantic 
and Gulf menhaden because the highest frequency of 
“Atlantic-only” haplotypes occurred in the northern- 
most Atlantic menhaden sampling location. However, 
the purported geographic cline in the distribution of 
the “Atlantic-only” clade haplotypes was only qualita- 
tively addressed and was based on a small sample size 
(n = 37) of Atlantic menhaden. In the present study, the 
more extensive sampling regime for Atlantic menhaden 
along the U.S. Atlantic coast (n=289) refutes Anderson’s 
(2007) hypothesis, because a chi-square analysis of 
the presence of the two clades among Atlantic coast 
sampling locations did not indicate a heterogeneous 
distribution. 
Population structure 
Stock structure analyses of Atlantic menhaden along 
the U.S. Atlantic coast have indicated as few as one and 
as many as three different stocks based on spawning 
time, spawning location, and migration tracks (Suther- 
land, 1963; June and Nicholson, 1964; Nicholson, 1978; 
Epperly, 1989). We analyzed the distribution of allelic 
variation of rapidly evolving molecular characters to 
evaluate population structure of Atlantic menhaden. The 
resulting AMOVAs did not reveal any significant portion 
of molecular variance was due to variation between the 
following group comparisons: YOY menhaden collected 
in Chesapeake Bay early and late in the season during 
the same year; YOY and yearling menhaden collected 
in Chesapeake Bay in successive years (following the 
2006 year class); YOY and yearling menhaden collected 
in Chesapeake Bay in the same year (comparing 2005- 
2006, 2006-2007 year classes); and YOY and yearling 
menhaden (combined) from the four geographic regions 
along the U.S. Atlantic coast (New England, mid-Atlan- 
tic, Chesapeake Bay, and U.S. South Atlantic). The 
POWSIM analysis indicates that the sample sizes and 
