Site by Bennett Web & Design Co.
2009 SOUTHEASTERN NATURALIST 8(2):227–243
Genetic Assessment of Species Ranges in
Fundulus heteroclitus and F. grandis in
Northeastern Florida Salt Marshes
Iara Gonzalez1, Michael Levin1, Sura Jermanus1, Brent Watson1,
and Matthew R. Gilg1,*
Abstract - The limits of species ranges can be determined by a number of biotic
and abiotic interactions, and in areas where closely related species overlap, some
degree of reproductive isolation must exist in order for them to remain distinct.
Understanding these interactions is essential for understanding what limits species
distributions or causes hybridization. Fundulus heteroclitus (Mummichog)
and Fundulus grandis (Gulf Killifish) are two closely related species with similar
morphologies and ecological niches. Both species have widespread distributions
that overlap in northeastern Florida. In the present study, two highly divergent loci
(one nuclear and one mitochondrial) were utilized to distinguish these fundulid species
in order to identify their ranges and to detect hybrids. Analysis of specimens
collected along a north to south gradient in northeastern Florida salt marshes established
that a relatively sharp transition (≈38 km) from relatively pure Mummichog
populations to relatively pure F. grandis populations existed south of Jacksonville,
FL, centered near Flagler Beach, FL. Putative hybrid genotypes were detected at
moderate frequencies within the contact zone, suggesting that successful hybridization
is likely occurring between the two species, but is relatively uncommon. These
results provide a stepping stone to investigate the types of reproductive barriers
that are involved in maintaining species distinctions in this system and their effects
on the species’ ranges and ecological interactions.
Areas of range overlap between closely related species provide opportunities
to understand the mechanisms responsible for speciation and the
maintenance of diversity. Interspecific interactions in regions of overlap
can affect the extent to which adaptation may occur at the edges of a species’
geographic range (Case and Taper 2000). The distribution of species
is known to be strongly infl uenced by closely related species because they
tend to share many resources resulting in strong competition (Anderson et al.
2002). Competition among sympatric species can lead to one species being
excluded from an area by the other, an event known as competitive exclusion
(Day and Young 2004, Zaret and Rand 1971). Several studies have documented
that species sharing the same resources are able to coexist through
resource partitioning (Hart 2003, Schluter and McPhail 1992, Weisberg
1986, Zaret and Rand 1971).
Furthermore, in areas of sympatry, hybridization can be a key factor
in defining species boundaries (Gow et al. 2006). Closely related species
1Department of Biology, University of North Florida, 1 UNF Drive, Jacksonville, FL
32224. *Corresponding author - firstname.lastname@example.org.
228 Southeastern Naturalist Vol. 8, No. 2
that come into contact have the potential to interbreed, which can lead to
a number of potential outcomes. These outcomes include genetic assimilation
of the two species, increasing the reproductive isolation of the parental
species due to low fitness of hybrids (reinforcement), adding genetic variation
to parental populations, or production of a third hybrid species that is
reproductively isolated from both parental forms (Arnold 1997). The role of
hybridization in species diversification remains a central issue in evolutionary
biology (Coyne and Orr 2004).
Two closely related teleost species, Fundulus heteroclitus (L.) (Mummichog)
and Fundulus grandis Baird and Girard (Gulf Killifish) provide an
opportunity to investigate the genetic structure of divergent populations that
overlap in range. The geographical distribution of the Mummichog ranges
from southwestern Newfoundland to northeastern Florida (Able and Felley
1986), while the Gulf Killifish ranges from northeastern Florida southward
to the coasts of Mexico (Duggins et al. 1989). These fundulids are found in
salt marsh habitat year round and utilize the marsh surface to feed during high
tide and to spawn according to the semi-lunar cycles (Greeley and MacGregor
1983, Hsiao and Meier 1989, Weisberg et al. 1981). Both species are considered
highly abundant with very productive populations—likely a consequence
of the two species being opportunistic omnivores (Kneib 1986, Lipcius and
Subrahmanyam 1986). Their diets consist of small crustaceans, insects, snails,
polychaetes, and detritus (James-Pirri et al. 2001, Lipcius and Subrahmanyam
1986). Both fundulids are also prey for a variety of predators such as crustaceans
(Callinectes spp., Ucca spp.) and larger fishes, including commercially
important species such as Morone americana (Gmelin) (White Perch), and
Sciaenops ocellatus (L.) (Red Drum) (James-Pirri et al. 2001, Kneib 1986).
Given the extent of ecological overlap between the two species and the
fact that hybridization among several fundulid species has been widely
reported (Atz 1986, Duggins et al. 1995, Duvernell et al. 2007, Hernández-
Chávez and Turgeon 2007), it seems likely that the Gulf Killifish and the
Mummichog hybridize in areas of overlap. Furthermore, previous genetic
data from Duggins et al. (1989) showed that the two species have many allozyme
alleles in common, suggesting hybridization.
In order to investigate the type of interactions between the Mummichog
and the Gulf Killifish, it is necessary to establish their region of overlap
in northeast Florida. Whereas their ranges were known to meet along the
northeastern Florida coast, the specific location of their zone of sympatry
and its width were not known. Our study intended to establish the distribution
of Mummichog and Gulf Killifish along salt marshes in northeastern
Florida through the use of species-specific genetic markers. Morphologically,
the Mummichog and the Gulf Killifish can be distinguished in their
adult stage by the eight mandibular pores and longer ovipositor of the
Mummichog versus the ten mandibular pores and shorter ovipositor of
the Gulf Killifish (Duggins et al. 1989). Nonetheless, the morphological
characteristics of the adults are often overlapping in variation, making misidentification a strong possibility, and young juveniles and larvae of the two
species are virtually impossible to tell apart morphologically. Therefore,
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 229
identifying molecular markers that distinguish members of these two similar
species would provide a reliable identification method that can be used for
any age class. The possibility of hybridization between the Mummichog and
the Gulf Killifish can also be verified by species-specific nuclear markers.
Both the allopatric and sympatric ranges of the two species can be determined
by collecting adults in tidal creeks along a latitudinal gradient in
northeastern Florida. According to Duggins et al. (1989), the two species
overlap in range in the vicinity of Marineland, located approximately 13 km
south of St. Augustine, FL. Therefore, we hypothesized that there is a latitudinal
cline, with Mummichog individuals being abundant in the northern
sites and tapering off as Gulf Killifish specimens become more common in
the southern sites.
Materials and Methods
Sample collection and DNA extraction
Representatives from allopatric populations were sampled to identify
genetic markers that differentiated the two species. Personal communication
with Russ Brodie of the Florida Fish and Wildlife Commission (Fish and
Wildlife Research Institute, Jacksonville, FL) suggested that the St. Mary’s
River (SM) in Georgia could be considered reference site for populations of
Mummichog, while samples from Cedar Key (CK) and Port St. Joe (PSJ) on
the Gulf coast of Florida (Fig. 1) would include only Gulf Killifish (Hoese
and Moore 1998). According to morphological characteristics, only the
Mummichog is found as far north as SM and only the Gulf Killifish is found
on the Gulf coast of Florida. Furthermore, previous genetic work by Duggins
et al. (1989) suggests the above locations to be reliable reference sites
for each species. To determine the possible area of overlap, adult Fundulus
were collected during the spring and summer of 2005 and 2006 from eight
sites spanning about 200 km of northeastern Florida (Fig. 1). The localities
were chosen to extend the previous study by Duggins et al. (1989) describing
genotypic distributions for the Mummichog and Gulf Killifish in the southeastern
United States. The sites sampled along northeastern Florida were
distributed from north to south: Atlantic Beach (AB), Vilano Beach (VB),
Moses Creek (MC), Marineland (ML), Pellicer Creek (PC), Flagler Beach
(FB), Tomoka Basin (TB), and Indian River (IR). All of the sites had similar
physical characteristics, except PC had a maximum salinity of 17 parts per
thousand (ppt) while the other areas had salinity ranges between 20–35 ppt.
A total of 523 fishes were collected from all the sites (Table 1) utilizing
minnow traps, beach seines, or cast nets. Specimens were placed in Whirl-
Pak bags and labeled by collection site and date. These bags were placed
on ice, transported to the laboratory, and stored at -80 ºC upon arrival.
Adult specimens used for genetic analysis were measured (standard length),
individually labeled, and preserved in 95% ethanol. A small piece of fin tissue
was removed from each sample for genomic DNA extraction using the
phenol/chloroform method (Sambrook and Russell 2001) or the QIAGEN
DNeasy Blood and Tissue Kit.
230 Southeastern Naturalist Vol. 8, No. 2
Multiple published sequences of the mitochondrial gene cytochrome-b
(cyt-b) were available on GenBank for both Mummichog and Gulf Killifish
from Florida and Georgia (Accession numbers: L23772–L23775, F. heteroclitus;
AF321852, U77124, and L31594, F. grandis). These sequences
suggested that the restriction enzyme NsiI would distinguish the two species
at this locus by cutting PCR products of the Gulf Killifish, but not the
PCR products of the Mummichog. A ≈450 bp region of the cyt-b locus was
Figure 1. Location of sampling sites in the southeastern United States. Reference
sites for Gulf Killifish populations are Cedar Key and Port St. Joe in the Gulf
coast while the St. Mary’s River in Georgia served as the reference site for Mummichog
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 231
amplified by polymerase chain reaction (PCR) using the universal primers
GLUDG-5’ and CB3-3’ as described in Palumbi (1996). Amplified PCR
products were electrophoresed on a 1% agarose gel at 125 V for 1 hour.
When a single amplification did not result in quality PCR products, the
samples underwent a second internal amplification using the CB2-3’ primer
with GLUDG-5’. Successfully amplified products underwent a restriction
digest using NsiI as directed in the manufacturers’ protocol. Digested products
were electrophoresed on a 3% agarose gel at 70 V for 2 hours, and after
visualization, samples were scored as cut or uncut.
Sequencing of a subset (n = 14) of our samples was performed on a
Beckman-Coulter CEQ 8000 using the same primers as above with the
manufacturer’s cycle sequencing protocol. Sequencher 4.7 was used to edit
sequences, and aligned sequences were analyzed by maximum parsimony
using PAUP (4.0.3).
Eight sequences of the 5’ untranslated region of lactate dehydrogenase-
B (LB UTR) are published in GenBank for the two species (Accession
numbers: U59837–U59844) and were assessed for differences in restrictionenzyme
cut patterns. Since only a single sequence has been published for
F. grandis, we constructed internal primers to amplify and sequence additional
PCR products of both species to verify possible Restriction Fragment
Length Poymorphism (RFLP) differences. Best results were produced by
performing a pair of PCR amplifications using external, UTR and FLK, and
internal, MRG-3 (5’-TTG TTT CAT GGG GTC TGA ACA C-3’) and MRG-
4 (5’- GGC ATT ACA ATC AGA CAA GTA GAG-3’), primers (Schulte et
al. 1997). Amplified PCR products were electrophoresed on a 1% agarose
Table 1. Summary of cyt-b haplotype and LB UTR genotype frequencies. The letters h and g
denote F. heteroclitus (Mummichog)-specific and F. grandis (Gulf Killifish)-specific alleles,
respectively. P represents the probability a locus is in Hardy-Weinberg equilibrium based on
Fisher’s exact test, and F values are the result of Wright’s F -statistics. Site codes are as follows:
SM = St. Mary’s River, AB = Atlantic Beach, VB = Vilano Beach, MC = Moses Creek, PC =
Pellicer Creek, ML = Marineland, FB = Flagler Beach, TB = Tomoka Basin, IR = Indian River,
CK = Cedar Key and PSJ = Port St. Joe., n = sample size.
Cyt-b LB UTR
Site n h g n h/h h/g g/g P F
SM 79 1.00 0.00 73 0.89 0.08 0.03 0.027† 0.37
AB 57 0.91 0.09 57 0.86 0.05 0.09 2.5x10-5† 0.82
VB 36 1.00 0.00 37 0.95 0.05 0.00 0.98 0.08
MC 44 0.98 0.02 36 0.89 0.03 0.08 0.008† 0.24
PC 50 0.92 0.08 49 0.94 0.04 0.02 0.061 0.47
ML 41 1.00 0.00 41 0.83 0.07 0.10 4.5x10-4† 0.35
FB 51 0.40 0.60 50 0.20 0.06 0.74 3.2x10-8† 0.83
TB 8 0.13 0.88 8 0.00 0.125 0.88 1.0 -0.11
IR 63 0.00 1.00 60 0.00 0 1.00 1.0 0.00
CK 54 0.02 0.98 55 0.00 0 1.00 1.0 0.00
PSJ 38 0.03 0.97 38 0.00 0.03 0.97 0.76 0.00
† indicates statistical significance (P < 0.05).
232 Southeastern Naturalist Vol. 8, No. 2
gel at 125 V for 1 hour. According to the sequence data (Accession numbers:
EU482176–EU482180), the restriction enzyme PstI was hypothesized to cut
the PCR product of the Mummichog, while PCR products of the Gulf Killifish should remain uncut. The internal primers MRG-3 and MRG-4 produce
an uncut product that is approximately 485 bp long in the Gulf Killifish only,
while the cut products are typically 176 bp and 309 bp in length. Small differences
in product length due to insertions and deletions were observed in
some samples. As with the mitochondrial locus, a subset of 12 samples from
the present study were sequenced on a Beckman-Coulter CEQ 8000 using
primers MRG-3 and MRG-4 and analyzed as described above for cyt-b.
Haplotype and genotype frequencies of the two markers were estimated
for each collection site and compared using an exact population homogeneity
test (GenePop; Raymond and Rousset 1995). Fisher’s (1922) exact test
was used to determine whether genotypic frequencies of LB UTR were in
Hardy-Weinberg equilibrium (HWE) and Wright’s (1922) F parameter was
calculated to determine whether deviations from Hardy-Weinberg expectations
were due to a deficiency or an excess of heterozygotes. Individuals
from all sites were assigned to parental populations using STRUCTURE
(ver. 2.1; Pritchard et al. 2000). Finally, exact tests of cytonuclear disequilibrium
were conducted using CNDd (Asmussen and Basten 1994, Basten and
Asmussen 1997). Statistical significance of all tests was set at α = 0.05.
The specimens from reference populations verified near species specificity
of both the mitochondrial and nuclear genetic markers. The cyt-b PCR
product was uncut in all the samples from SM in Georgia, correctly identifying
them as Mummichogs (Table 1). The CK and PSJ specimens were cut at
frequencies of 0.98 and 0.97, respectively, correctly identifying the majority
of individuals as Gulf Killifish (Table 1). Considering all the samples from
the reference sites, the NsiI restriction digest of the cyt-b marker had a >98%
probability of giving the correct identification at the reference sites.
Allele and genotype frequencies of LB UTR were highly divergent
between the reference sites of the two species and supported the results
obtained by the mitochondrial marker, but were not completely speciesspecific. Cut alleles were found at a frequency of 0.93 at SM, but only at
frequencies of 0.00 and 0.01 at the Gulf Coast reference sites of CK and
PSJ, respectively, showing clear differences in allele frequencies among the
reference sites for the two species. Essentially, cut alleles are predominantly
found in Mummichogs, and uncut alleles are nearly fixed in Gulf Killifish
populations. The frequency at which specimens from the SM population had
LB UTR cut completely by PstI digest was 0.89, representing homozygous
specimens for the allele most often encountered in Mummichogs (Table 1).
The resolution of the nuclear marker was more precise for homozygous uncut
individuals, with a genotype frequency of 1.00 at CK and 0.97 at PSJ.
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 233
When all samples from the three reference sites are considered at the nuclear
marker, 95% of the specimens were identified as being homozygous for the
correct species-specific alleles. Whereas it is not a fixed difference, the LB
UTR marker is reliable for species identification. Since both markers were
shown to be highly species-specific, we designated a Mummichog haplotype/
allele as “h” and a Gulf Killifish haplotype/ allele as “g.”
Adult distributions: Cyt-b
Haplotype frequencies followed the expected north-to-south gradient,
with significant spatial variation among all collection sites (P < 0.0001, S.E.
< 0.0001). All sites north of FB had frequencies of h haplotypes ≥0.91, while
sites south of TB were devoid of h haplotypes (Table 1, Fig. 2). FB had intermediate
frequencies of 0.40 for the h haplotype. The observed haplotype
frequencies at cyt-b suggested the presence of three distinct regions: a northern
region with high frequencies (>0.91) of h haplotypes; a zone of overlap
exhibiting a mixture of h and g haplotypes; and a southern region lacking h
haplotypes (Fig. 2). Sites north of FB (including the reference site SM) that
had high frequencies for the h haplotype did, however, show significant spatial
variation (P = 0.0072, S.E. = 0.0008) except when samples from AB and
PC were removed (P = 0.602, S.E. = 0.0046). This finding suggests that most
of these sites have relatively pure populations of Mummichogs. Similarly,
no significant differences in haplotype frequencies were observed among
the southern sites of TB and IR when compared to the reference sites of CK
and PSJ, suggesting these sites represented Gulf Killifish populations (P =
0.0575, S.E. = 0.0018). The marginally insignificant P value of the aforementioned
comparison is due to the inclusion of TB, which because of its
sample size (n = 8), makes its inclusion unreliable. These comparisons helped
Figure 2. F. heteroclitus (Mummichog)- specific haplotype (cyt-b) and allele (LB
UTR) frequencies across all sites from north to south along the Atlantic coast. Only
sites from Saint Mary’s to Indian River were included.
234 Southeastern Naturalist Vol. 8, No. 2
establish the area of range overlap of the two species as being bordered by the
ML and TB sites, which are separated by a distance of about 38 km.
Since several sites north of FB contain samples with g haplotypes, a subset
of seven samples were sequenced to verify whether these were truly Gulf
Killifish haplotypes or if they were simply Mummichog haplotypes that happened
to contain an NsiI cut site. An unrooted maximum parsimony analysis
including samples from the reference sites (Accession numbers: EU482166,
EU 482167), samples of each species downloaded from Genbank (Accession
numbers: L23772–L23775, AF312852, U77124, L31594), and samples with
unexpected haplotypes from sites between SM and FB (Accession numbers:
EU482165–EU482154) verified that, in most cases, the “cut” haplotypes in
the northern region (AB 31, 38; PC 20, 21, 24, 37) were indeed Gulf Killifish
haplotypes (Fig. 3). Only the sample AB 57 had a haplotype that was cut by
NsiI and was within the Mummichog clade.
Adult distributions: LB UTR
The allele frequencies of LB UTR followed a similar pattern to that observed
at cyt-b. All sites north of FB had h allele frequencies ranging from
0.87 to 0.97. A marked shift in allele frequencies was again detected at FB,
with h allele frequencies decreasing to 0.23 (Fig. 2). The remaining populations
exhibited a high incidence of g alleles with frequencies of 0.94 at the
Tomoka Basin site and 1.0 at the Indian River site. Comparisons of allele
frequencies across all sites yielded significant spatial variation (P < 0.0001,
S.E. < 0.0001).
Analysis of LB UTR allele frequencies suggested the presence of the
same three regions observed at cyt-b (Fig. 2). Results at LB UTR were
slightly different than at cyt-b, since a comparison of the six sites north of
FB showed no significant variation among populations (P = 0.0704, S.E. =
0.0045 ). Allele frequencies at LB UTR differed significantly among TB, IR,
and the Gulf Coast reference populations (P = 0.0266, S.E. = 0.0011), but
removal of TB removed the significant spatial variation (P = 0.2500, S.E. =
0.0026). Therefore, the zone of sympatry established by LB UTR is similar
to that found with cyt-b; the FB site has intermediate frequencies of both h
and g alleles, with predominantly Mummichog populations to the north and
Gulf Killifish populations to the south.
The Gulf Killifish reference sites of CK and PSJ showed no significant
deviations from Hardy-Weinberg equilibrium (HWE) at LB UTR. On the
other hand, the Mummichog reference site, SM, was not in HWE (Table 1).
Of the eight non-reference sites, four had genotype frequencies that differed
significantly from HWE including AB, MC, ML, and FB. All sites that deviated
significantly from HWE showed a deficit in heterozygotes (Table 1).
Even sites that were found to be in HWE tended to have positive values of
Wright’s F statistic (1922), except for the population at TB in which the
negative F is likely explained by the small sample size.
The presence of low but consistent numbers of g alleles at the LB UTR locus
in populations of predominantly Mummichogs (including reference site
SM) prompted the sequencing of individuals with g/g genotypes at LB UTR
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 235
to determine whether these g alleles were representative of Gulf Killifish
alleles in predominantly Mummichog populations or if they were simply
polymorphisms within Mummichogs that lacked the PstI restriction site. An
unrooted maximum parsimony analysis including samples from the reference
sites (Accession numbers: EU482176–EU482180), samples of each
Figure 3. Maximum parsimony cladogram of cyt-b including samples from Genbank
(noted with accession numbers) and various collection sites from northeastern
Florida. Numbers at the nodes represent the results of a bootstrap analysis with 1000
replicates. Site codes are as follows: AB = Atlantic Beach, PC = Pellicer Creek, FB
= Flagler Beach and CK = Cedar Key.
236 Southeastern Naturalist Vol. 8, No. 2
species downloaded from Genbank (Accession numbers: U59844, U59835
and U59834), and samples with unexpected genotypes from sites between
SM and FB (Accession numbers: EU482187 and EU482186) showed that
both of the g alleles sequenced from samples collected north of FB (AB 35,
SM 75) actually grouped with the Mummichog clade (Fig. 4). The same was
true of most of the samples from FB (see below). Therefore, at least some of
Figure 4. Maximum parsimony cladogram of LB UTR including samples from Genbank
(noted with accession numbers) and various collection sites from northeastern
Florida. Numbers at the nodes represent the results of a bootstrap analysis with 1000
replicates. Site codes are as follows: SM = St. Mary’s River, AB = Atlantic Beach,
FB = Flagler Beach and CK = Cedar Key.
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 237
the g alleles within the predominantly Mummichog populations are actually
polymorphisms within Mummichogs and are not due to introgression or the
presence of Gulf Killifish at those locations.
Hybridization and cytonuclear disequilibrium
Haplotype and genotype data from both loci were combined to investigate
the overall population structure of these two species and to test
for the existence of hybrids. The program STRUCTURE (Pritchard et
al. 2000) was used to assign each individual to a population of ancestry
based on its dilocus genotype. This assignment score (q) ranges from 0
(Gulf Killifish) to 1 (Mummichog), and 95% probability intervals for
each q are estimated using a Markov Chain Monte Carlo method (500,000
iterations after a burn-in period of 50,000 iterations under a model of admixture
with K = 2 populations).
When the probability distributions of ancestry are plotted for each individual
across all collection sites, the results are very consistent with the
single locus data above (Fig. 5). Most individuals have assignment scores
near 0 or 1. The predominantly Mummichog population appears to have
a consistent but relatively low level of introgression from Gulf Killifish,
since several individuals have very low values of q. None of the samples
from these northern populations, however, have values of q as low as those
found at IR, PSJ, or CK. This finding is consistent with the fact that none
of the samples from any site between SM and ML had any individuals with
g:g/g (cyt-b:LB UTR) genotypes. Introgression is much less noticeable in the
predominantly Gulf Killifish populations. Furthermore, the transition from
Mummichog to Gulf Killifish populations is very sharp, mostly consisting
of the FB site. Samples from FB show considerable overlap of probability
distributions, with both populations suggesting the presence of individuals
of mixed ancestry.
The dilocus genotype frequencies at FB reveal that the two species
are likely able to hybridize (Table 2). Putative hybrid genotypes include
individuals that are heterozygous for the species-specific alleles at the LB
UTR locus (e.g., h:h/g and g:h/g) and also individuals that have the cyt-b
Table 2: Observed dilocus genotype frequencies of samples from Flagler Beach and estimates
of cytonuclear disequilibrium. For disequilibrium estimates, A represents the F. heteroclitus
nuclear allele and M the F. heteroclitus mitochondrial haplotype. Lower case letters are alleles/
haplotypes of F. grandis.
Genotypes P DA DAM DAAM DAaM DaaM
h:h/h 0.16 Estimators 0.1471 0.0880 0.0800 0.0160 0.0960
h:h/g 0.04 Norm. Est. 0.8306 0.6377 0.6667 0.4444 0.6154
h:g/g 0.20 Variances 0.0006 0.0009 0.0008 0.0002 0.0009
g:h/h 0.04 St. Error 0.0251 0.0306 0.0277 0.0165 0.0304
g:h/g 0.02 Pr (Sample) 3.2 e-8 0.0161 0.0155 0.6232 0.0068
g:g/g 0.54 Pr (Exact) NA 2.9 e-5 0.0088 0.5561 0.0027
Sample Size 90 19 60 65 595 53
Sample Size 50 6 23 23 203 19
238 Southeastern Naturalist Vol. 8, No. 2
Figure 5. STRUCTURE analysis of dilocus genotypes from all collection and reference
sites. Individuals are ranked based on collection site with the highest rankings
representing samples from SM and the lowest rankings representing samples from
from CK. Samples from Flagler Beach (FB) are noted. Diamonds represent individual
mean values of q and 95% posterior probabilities are shown.
haplotype of one species and are homozygous for the other species at LB
UTR (e.g.: h:g/g and g:h/h). Individuals that are heterozygous at LB UTR
could result from either F1 or other hybrid crosses, while h:gg and g:hh
genotypes are suggestive of backcrosses. All four putative hybrid genotypes
were present at FB, even though most of these were at lower frequencies than
the parental dilocus genotypes g:g/g and h:h/h (Table 2).
It is important to note that each of the putative hybrid genotypes could
also be produced by populations outside of the defined contact zone since
neither marker is fixed in either species. Therefore, to determine if the putative
hybrid genotypes were representative of actual hybridization, we sequenced
six individuals with putative hybrid genotypes at either cyt-b (Accession
number: EU482159), LB UTR (Accession number: EU482183) or both
(Accession numbers: EU482184, EU482185, EU482181 and EU482182).
Only individuals with h:g/g or g:h/h genotypes were used in sequencing
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 239
to minimize the identification issues inherent in sequencing heterozygotes.
These six individuals refl ect half of the samples with mismatched cytonuclear
genotypes at FB. The results of the maximum parsimony analysis are
shown in Fig. 3 and Fig. 4. Two samples that were sequenced at both loci, FB
17 and FB 46, were definitely the result of hybridization since the two loci
fell into alternative clades. FB 17 was shown to have a cyt-b haplotype of
Mummichog and an LB UTR genotype of Gulf Killifish. Sample FB 46 was
just the opposite. The other two samples sequenced at both loci, FB 19 and
FB 22, were initially genotyped as h:g/g, but the sequence analysis grouped
both samples with Mummichog at both loci. Sample FB 16 was initially
genotyped as h:g/g, but its LB UTR sequence was actually within the Mummichog
clade, suggesting it was probably not a hybrid. Sample FB 52 was
genotyped as g:h/h, and the cyt-b sequence was confirmed as being Gulf Killifish, suggesting the initial genotype was probably correct and the sample
was of mixed ancestry since h alleles at LB UTR are more species specific
than the g alleles.
The observed frequencies of the putative hybrid genotypes at FB suggest
that hybridization between the two species may be directional. Most of the
putative hybrid genotypes have an h cyt-b haplotype, indicative of a higher
proportion of hybrid matings involving Mummichog females with Gulf Killifish males. To test for significant directionality of the matings, we estimated
the cytonuclear disequilibrium at FB. The results (Table 2) show significant
disequilibrium at both species-specific genotypes (AAM and aaM), but none
at the hybrid genotypes (AaM). These data suggest that the two species are
predominantly mating assortatively and that the small level of hybridization
occurring is not directional. It should be noted that the sample sizes from
FB are not sufficient to detect the observed level of disequilibrium at mixed
cytonuclear genotypes with either 50% or 90% power (Table 2).
A zone of sympatry was detected in northeastern Florida for mummichogs
and Gulf Killifish between Marineland and the Tomoka Basin. The
frequencies of both cyt-b haplotypes and LB UTR alleles displayed a steep
cline, with Flagler Beach having intermediate frequencies of alleles from
each species. The observed patterns established three genetically distinct
regions. The first was a northern region from the St. Mary’s River in Georgia
to Pellicer Creek in Florida, which consisted primarily of Mummichog individuals
with low but relatively consistent levels of Gulf Killifish alleles at
both loci. At least some of the g alleles at the LB UTR locus in these northern
sites are due to an intraspecific polymorphism in Mummichog, but g cyt-b
haplotypes appear to be mostly due to introgression from Gulf Killifish. The
second region was the area of overlap, and possibly hybridization, centered
on Flagler Beach and potentially including the Tomoka Basin. Finally, a
southern region located south of Tomoka Basin and along the Gulf coast of
Florida was dominated by Gulf Killifish.
240 Southeastern Naturalist Vol. 8, No. 2
A study by Duggins et al. (1989) provides similar results, with Mummichog
samples being collected primarily from northern sites and Gulf
Killifish being found on the southern Atlantic and Gulf coasts of Florida.
Collections by Duggins et al. (1989) near ML contained both Gulf Killifish
and Mummichog specimens, which is on the northern edge of the contact
zone described here. Contradictory to our findings, Duggins et al. (1989)
suggest that collections made near FB were Mummichogs. Our data, on the
other hand, show that FB is in the middle of the contact zone, with higher frequencies
of Gulf Killifish and hybrid genotypes than Mummichogs. Samples
collected near FB by Duggins et al. (1989), however, tended to contain a
mixture of genotypes typical of both species. Therefore, it is likely that this
site represents a mixed population in both studies. It is intriguing that Duggins
et al. (1989) found higher proportions of Mummichog-like genotypes
than Gulf Killifish-like genotypes, while the opposite was true in the present
study. This may be due to microhabitat differences in the sampling locations
of the two studies or to temporal changes in the population genetic structure
at that location. Another possibility is that some of the individuals used in the
study by Duggins et al. (1989) were incorrectly identified to species using
morphology, leading to a misrepresentation of the genetic data.
The present study consistently encountered Gulf Killifish-specific alleles
and haplotypes in the primarily Mummichog populations in the northern
region, including the Saint Mary’s reference site. The presence of these alleles
could either be due to: 1) polymorphisms at both cyt-b and LB UTR
within the Mummichog population, 2) the presence of small numbers of
Gulf Killifish, or 3) introgression of Gulf Killifish-specific alleles/haplotypes
into Mummichog populations via hybridization. Sequence analysis
of individuals with Gulf Killifish cyt-b haplotypes from the predominantly
Mummichog sites verified that most of our initial analyses were correct and
F. grandis haplotypes were present in these populations. On the other hand,
sequence analysis of g/g individuals at the LB UTR locus suggested that most
of these individuals are actually homozygous for Mummichog alleles that
lack the PstI restriction site found in most previously published Mummichog
sequences. The existence of this polymorphism combined with the fact that
no individuals collected from the predominantly Mummichog sites had g:g/g
genotypes, makes it unlikely that Gulf Killifish are present in any significant
numbers north of FB. Therefore, the presence of Gulf Killifish-like alleles
or haplotypes is due to both introgression and, especially with the LB UTR
locus, a polymorphism within Mummichog populations.
The observation that Gulf Killifish are apparently not present in populations
north of ML is surprising given the fact that LB UTR genotype
frequencies are rarely in Hardy-Weinberg equilibrium at the northern sites.
The lack of heterozygotes at most of these sites would be expected if the
samples contained a mixture of two species with primarily intraspecific
matings. If the variation observed at the LB UTR locus is mostly due to an
intraspecific polymorphism in Mummichogs, the alleles should meet Hardy-
Weinberg expectations. This conclusion is especially true since the marker
we are using is an untranslated region and is expected to be neutral. On the
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 241
other hand, Schulte et al. (1997) provided evidence for functional differentiation
associated with structural differences in LB UTR alleles between
populations of Mummichogs in Florida and Maine. Therefore, it is possible
that the two alleles found within Mummichogs in the present study represent
functionally different regulatory regions of lactate dehydrogenase b.
Furthermore, lactate dehydrogenase b has been shown to be under selection
in populations of Mummichogs which exhibit a latitudinally associated allele
frequency cline between Pennsylvania and North Carolina (Powers et
al. 1991). Therefore, it is also possible that the two LB UTR RFLP alleles
observed in Mummichogs in the present study are linked to distinct alleles of
lactate dehydrogenase b or other loci that are experiencing assortative mating
or disruptive selection resulting in a heterozygote deficiency.
The polymorphism within Mummichog populations at the LB UTR locus
also likely plays a role in the perceived directionality of hybridization
observed at FB. While there was no significant cytonuclear disequilibrium
among the putative hybrid genotypes, most individuals with mixed genotypes
had h:g/g dilocus genotypes. The lack of complete species-specificity
of our LB UTR RFLP could result in true Mummichogs being mis-genotyped
as h:g/g hybrids. Indeed, sequencing verified that this was true for at least
some of the h:g/g individuals. Still, it should be mentioned that sequence
analysis also verified the existence of h:g/g hybrids. Additionally, the observed
frequency of h:g/g genotypes at FB is approximately five-fold higher
than could be produced by a randomly mating population of non-hybridizing
Mummichogs polymorphic at LB UTR at the frequencies observed in the
northern populations. This conclusion can be illustrated with the following
example: Assume a mixed population of Gulf Killifish and Mummichogs
with relative abundances similar to the ratios of allele frequencies observed
at FB (≈75% Gulf Killifish and ≈25% Mummichog ). If the frequencies of
h and g alleles within the Mummichog population are the same as those
observed at AB, where g alleles are at the highest frequency among the
northern sites (Table 1), under a model of no hybridization and random association
of cyt-b haplotypes and LB UTR alleles, the expected frequency
of h:g/g genotypes (AB: [(h = 0.91)*(g/g = 0.09)](0.25) + CK/PSJ: [(h =
0.02)*(g/g = 0.99)](0.75)) would only be approximately 0.04. The observed
frequencies at FB, on the other hand, are 0.2. This discrepancy suggests that
many of the h:g/g genotypes at FB are truly the result of hybridization. This
same model predicts lower expected frequencies of all putative hybrid genotypes
than those observed. Still, the extensive cytonuclear disequilibria of
pure species cytonuclear genotypes are suggestive of primarily intraspecific
matings (Asmussen et al. 1987).
It must be noted that an in-depth analysis of hybridization in the present
study is stymied by the fact that the markers used, especially LB UTR, are
not completely species-specific and the SM population in Georgia used as
a reference population for Mummichogs contains mitochondrial haplotypes
of Gulf Killifish. The lack of completely diagnostic markers, however, is
typical in many studies of hybridization, and the introgression of heterospecific alleles into areas of allopatry is also relatively common (Arnold 1997,
242 Southeastern Naturalist Vol. 8, No. 2
Harrison 1993). Recent work by Knowles and Carstens (2007) describes
the use of probabilistic models to delimit species when there is incomplete
lineage sorting between them, as appears to be the case with the LB UTR
RFLP marker utilized in the present study. In these cases, it is important to
utilize multiple loci to help correct for mistakes in assessment of genetic
identity. The fact that both markers used here are highly divergent between
the species and both show similar geographical patterns suggest that they
are reliable markers at the species level. Obviously, however, mistakes in the
assessment of genetic identity will be made in some cases resulting in some
incorrect reporting of hybrids. Additional nuclear markers that show greater
species-specificity will be necessary for further studies of ecological and
reproductive compatibility on these two species.
The area of overlap found in the Mummichog and Gulf Killifish complex
is consistent with other studies that have detected a habitat transition zone
around FB (Avise 1992, Duggins et al. 1995). Avise (1992) describes that
the Florida peninsula has a transitional zone that separates temperate and
tropical adapted forms, with the southern ranges of many temperate species
terminating in the Cape Canaveral region. Around this region an ecotone exists
where the Juncus-Spartina marsh characteristic of the northern Atlantic
Coast are replaced by mangrove marsh characteristic of southern peninsular
Florida (Duggins et al. 1995). Allele and haplotype frequencies found in the
present study were consistent with this pattern, where Mummichogs diminished
in frequency as they entered the transition zone and were completely
replaced by Gulf Killifish at IR, which lies near Cape Canaveral. The similar
clinal location and width at the cyt-b and LB UTR loci also suggests that
the hybrid zone between Mummichogs and Gulf Killifish is due to secondary
contact. This conclusion would also be consistent with Avise’s (2000)
hypothesis of historical vicariance between Gulf and Atlantic populations of
many species in the southeastern United States. Investigation of the clinal
patterns of additional species-specific markers and studies of habitat use and
the factors limiting the ranges of these species could help determine whether
FB truly represents an area of secondary contact and whether the current
ranges are defined by the Juncus-Spartina ecotone.
The present study was able to discern in more detail the ranges of Mummichogs
and Gulf Killifish along northeastern Florida. The two species meet
and apparently hybridize to some extent over a relatively short range (<40
km) in northeastern Florida. These data provide an ideal stepping stone to
investigate the types of ecological interactions and intrinsic or extrinsic barriers
that may maintain reproductive and ecological isolation between these
We thank Stacy Galleher, Kelly Smith, Talisha Hunter, and Vyacheslav
Shevchenko for help with collections and James Rodgers for critical reading of the
manuscript. This work was funded by grants through the University of North Florida
(UNF) summer scholarship program and UNF Biology.
2009 I. Gonzalez, M. Levin, S. Jermanus, B. Watson, and M.R. Gilg 243
Able, K.W., and J.D. Felley. 1986. Geographic variation in Fundulus heteroclitus:
Tests of concordance between egg and adult morphologies. American Zoologist
Anderson, R.P., A.T. Peterson, and M.Gómez-Laverde. 2002. Using niche-based GIS
modeling to test geographic predictions of competitive exclusion and competitive
release in South American pocket mice. Oikos 98:3–16.
Arnold, M.L. 1997. Natural Hybridization and Evolution. Oxford University Press,
New York, NY.
Asmussen, M.A., and C.J. Basten. 1994. Sampling theory for cytonuclear disequilibria.
Asmussen, M.A., J. Arnold, and J.C. Avise. 1987. Definition and properties of
disequilibrium statistics for associations between nuclear and cytoplasmic genotypes.
Atz, K.W. 1986. Fundulus heteroclitus in the laboratory: A history. American Zoologist
Avise, J.C. 1992. Molecular population structure and the biogeographic history of a
regional fauna: A case history with lessons for conservation biology. Oikos 63:
Avise, J.C. 2000. Phylogeography: The History and Formation of Species. Harvard
University Press, Cambridge, MA.
Basten, C.J., and M.A. Asmussen. 1997. The exact test for cytonuclear disequilibria.
Case, T.J., and M.L. Taper. 2000. Interspecific competition, environmental gradients,
gene fl ow, and the coevolution of species’ borders. American Naturalist
Coyne, J.A, and H.A. Orr. 2004. Speciation. Sinauer Associates, Sunderland, MA.
Day, T., and K.A. Young. 2004. Competitive and facilitative evolutionary diversification.
Duggins, C.F., K.G. Relyea, and A.A. Karlin. 1989. Biochemical systematics in
southeastern populations of Fundulus heteroclitus and Fundulus grandis. Northeast
Gulf Science 10:95–102.
Duggins, C.F., A.A. Karlin, T.A. Mousseau, and K.G. Relyea. 1995. Analysis of
a hybrid zone in Fundulus majalis in a northeastern Florida ecotone. Heredity
Duvernell, D.D., J.F. Schaefer, D.C. Hancks, J. A. Fonoti, and A.M. Ravanelli. 2007.
Hybridization and reproductive isolation among syntopic populations of the
topminnows Fundulus notatus and F. olivaceus. Journal of Evolutionary Biology
Fisher, R.A. 1922. On the interpretation of χ2 from contingency tables, and the calculation
of P. Journal of the Royal Statistical Society 85:87–94.
Gow, J.L., C.L. Peichel, and E.B. Taylor. 2006. Contrasting hybridization rates between
sympatric Three-spined Sticklebacks highlight the fragility of reproductive
barriers between evolutionary young species. Molecular Ecology 15:739–752.
Greeley, M.S., and R. MacGregor. 1983. Annual and semi-lunar reproductive cycles
of the Gulf Killifish, Fundulus grandis, on the Alabama Gulf coast. Copeia
Harrison, R.G. 1993. Hybrid Zones and the Evolutionary Process. Oxford University
Press, New York, NY. 376 pp.
244 Southeastern Naturalist Vol. 8, No. 2
Hart, P.J. 2003. Habitat use and feeding behaviour in two closely related fish species,
the Three-spined and Nine-spined Stickleback: An experimental analysis. Journal
of Animal Ecology 72:777–783.
Hernández-Chávez, C., and J. Turgeon. 2007. Asexual and sexual hybrids between
Fundulus diaphanus and F. heteroclitus in the Canadian Atlantic region. Molecular
Hoese, H.D., and R.H. Moore. 1998. Fishes of the Gulf of Mexico; Texas, Lousiana,
and Adjacent Waters. Texas A&M University Press, College Station, TX.
Hsiao, S., and A.H. Meier. 1989. Comparison of semi-lunar cycles of spawning
activity in Fundulus grandis and F. heteroclitus held under constant laboratory
conditions. Journal of Experimental Zoology 252:213–218.
James-Pirri, M.J., K.B. Raposa, and J.G. Catena. 2001. Diet Composition of Mummichogs,
Fundulus heteroclitus, from restoring and unrestricted regions of a New
England (USA) salt marsh. Estuarine, Coastal, and Shelf Science 53:205–213.
Kneib, R.T. 1986. The role of Fundulus heteroclitus in salt marsh trophic dynamics.
American Zoologist 26:259–269.
Knowles, L.L., and B.C. Carstens. 2007. Delimiting species without monophyletic
gene trees. Systematic Biology 56:887–895.
Lipcius, R.N., and C.B. Subrahmanyam. 1986. Temporal factors infl uencing killifish
abundance and recruitment in Gulf of Mexico salt marshes. Estuarine, Coastal,
and Shelf Science 22:101–114.
Palumbi, S.R. 1996. Molecular Systematics, 2nd Edition. Sinauer Associates, Sunderland,
Powers, D.A., T. Lauerman, D. Crawford, M. Smith, I. Gonzalez-Villasenor, and L.
DiMichele. 1991. The evolutionary significance of genetic variation at enzymesynthesizing
loci in the teleost Fundulus heteroclitus. Journal of Fish Biology
Pritchard, J.K., M. Stephens, and P. Donnelly. 2000. Inference of population structure
using multilocus genotype data. Genetics 155:945–959.
Raymond M., and F. Rousset. 1995. GENEPOP (version 1.2): Population genetics
software for exact tests and ecumenicism. Journal of Herdity 86:248–249.
Sambrook, J., and D.W. Russell. 2001. Molecular Cloning: A Laboratory Manual.
Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Schluter, D., and J.D. McPhail. 1992. Ecological character displacement and speciation
in sticklebacks. American Naturalist 140:85–108.
Schulte, P.M., M. Gómez-Chiarri, and D.A. Powers. 1997. Structural and functional
differences in the promoter and 5’ fl anking region of Ldh-B within and between
populations of the teleost Fundulus heteroclitus. Genetics 145:759–769.
Weisberg, S.B. 1986. Competition and coexistence among four estuarine species of
Fundulus. American Zoologist 26:249–257.
Weisberg, S.B., R. Whalen, and V.A. Lotrich. 1981. Tidal and diurnal infl uence on
food consumption of a salt marsh killifish Fundulus heteroclitus. Marine Biology
Wright, S. 1922. Coefficients of inbreeding and relationship. American Naturalist
Zaret, T.M., and A.S. Rand. 1971. Competition in tropical stream fishes: Support for
the competitive exclusion principle. Ecology 52:336–342.