Hybridization and reproductive isolation among syntopic populations of the topminnows Fundulus notatus and F. olivaceus


David D. Duvernell, Department of Biological Sciences, Southern Illinois University Edwardsville, Edwardsville, IL 62026, USA.
Tel.: 618 650 3468; fax: 618 650 3174; e-mail dduvern@siue.edu


Fundulus notatus and Fundulus olivaceus are two closely related topminnow species that exhibit similar ecological niches and broad, largely overlapping, North American ranges extending throughout much of the Mississippi River drainage as well as the coastal drainages of the Gulf of Mexico. Previous studies have suggested that these two species are reproductively compatible despite cytogenetic differences and will hybridize when syntopic. We used nuclear and mtDNA loci to assess levels of hybridization and test for introgression in syntopic populations of these two species in four drainages in southern Illinois. Although hybridization was detected in all syntopic populations, an assessment of the proportion of hybrid individuals indicated a deficiency of hybrids relative to expectations under random mating. We determined that, although mtDNA introgression was prevalent and extended beyond the zones of contact, evidence of nuclear introgression was limited to the zone of sympatry.


The process of speciation and the nature of reproductive isolation among closely related species remains a central issue in evolutionary biology (Coyne & Orr, 2004). The outcomes of species interactions at habitat transition zones can have important consequences for species diversity and can impact the extent to which adaptation to unique environments may occur among closely related species. Traditionally, zoologists have viewed hybridization as a process that limits diversification (Mayr, 1963). This view is reinforced by well studied examples of introgressive hybridization between introduced and endemic species (e.g. Childs et al., 1996). However, others have stressed the role of introgressive hybridization in contributing to genetic diversity (Dowling & DeMarias, 1993) and driving adaptive radiations (Seehausen, 2004).

Mosaic distributions among species may exist when ecologically differentiated taxa are distributed throughout a network of heterogeneous environments. Species pairs that exhibit mosaic distributions are fascinating because these distributional patterns are the direct outcomes of fitness differences and competition among species in complex environments. The likelihood that mosaic distributions are maintained when such taxa encounter one another at habitat transition zones may depend on extrinsic selection favouring alternative alleles across environmental gradients (Harrison & Rand, 1989) or intrinsic selection against hybrid individuals resulting, for instance, from genetic incompatibilities (Barton & Gale, 1993). The extent to which hybridization and introgression occurs at habitat margins may also be limited by prezygotic isolation resulting from spatial, seasonal or behavioural differentiation among taxa (Butlin & Ritchie, 1994).

The Fundulus notatus species complex is an interesting system in which to study the relative importance of various factors that lead to and maintain genetic isolation between species, and the ecological factors that determine species distributions at a fine geographic scale. This complex is comprised of the blackstripe topminnow, F. notatus (Rafinesque), the blackspotted topminnow, Fundulus olivaceus (Storer) and the broadstripe topminnow, Fundulus euryzonus (Suttkus and Cashner), which are all closely related killifishes with similar ecological niches and overlapping ranges (Petifils, 1986; Blanchard, 1996). Fundulus notatus and F. olivaceus exhibit unusually broad, overlapping ranges extending throughout much of the Mississippi River drainage as well as the coastal drainages of the Gulf of Mexico (Fig. 1). In marked contrast, F. euryzonus has a much narrower range limited to the Amite and Tangipahoa river systems of the Lake Pontchartrain drainage in Louisiana and Mississippi (Suttkus & Cashner, 1981).

Figure 1.

 Partial distribution of members of the Fundulus notatus complex indicating the range of sympatry among species. The inset map describes the distribution of samples within the Barren Creek drainage. Sampling locations are identified with numerical codes as defined in Table 1. Filled circles (1–13) indicate samples from which population data were collected. Open circles (14–17) indicate samples primarily used for phylogenetic reconstruction of cytochrome b sequences.

A variety of studies have investigated morphological, meristic and cytogenic differences between F. notatus and F. olivaceus. Other than a modest difference in adult size (F. notatus, 50–70 mm; F. olivaceus 60–90 mm, Lee et al., 1980) and some life history differences in egg and clutch sizes (Schaefer, unpublished data), the two species are morphologically similar throughout their respective ranges (Braasch & Smith, 1965; Thomerson, 1966). A single distinguishing character, the presence or absence of distinct, regular, dark black dorsolateral spots, separates the two species in most instances (Thomerson, 1966). However, the manifestations of this character are highly variable among individuals within and among localities often making species identification on the basis of this character difficult. The two species exhibit key cytogenetic differences with F. olivaceus exhibiting the ancestral condition of 2n = 48 chromosomes (Chen, 1971), whereas F. notatus has 2n = 40 chromosomes throughout most of its range (Setzer, 1970; Chen, 1971). Populations of F. notatus from the Tombigbee River drainage in Alabama and Mississippi exhibit an intermediate karyotype of 2n = 44 chromosomes (Black & Howell, 1978) whereas F. euryzonus exhibits 2n = 48 chromosomes (Howell & Black, 1981).

A number of ecological and behavioural studies have focused on F. notatus and F. olivaceus (Braasch & Smith, 1965; Thomerson, 1966; Thomerson & Wooldridge, 1970). Where these two species are found in close proximity, subtle differences in habitat use appear to be important factors in their isolation. Throughout most of their range F. olivaceus tends to occupy high-gradient, clear, gravelly streams whereas F. notatus predominates in quieter large river floodplain sloughs and prairie streams (Braasch & Smith, 1965; Thomerson, 1966; Howell & Black, 1981; Ross, 2001). It has been suggested that habitat requirements broadly overlap (Thomerson, 1966) and the rarity of syntopic populations can be attributed to competitive exclusion (Thomerson & Wooldridge, 1970). However, although syntopy of F. notatus and F. olivaceus is generally viewed as rare (Thomerson, 1966), contact zones between these two species occur throughout their range, and hybridization and backcrossing has been reported in a few of these locations (Setzer, 1970; Howell & Black, 1981). Examination of museum records from nine major North American ichthyological collections revealed the existence of independent putative contact zones in 16 drainages throughout the range (Schaefer, unpublished data). Viewed from an intra-drainage perspective the two species often, but not always, exhibit abrupt clinal patterns in their distributions (Howell & Black, 1981).

The F. notatus complex provides an excellent system in which to study the ecological and evolutionary relationships and interactions among closely related species. In addition to the broad overlapping ranges and contact zones described above, these species are easily manipulated in breeding and competition experiments in controlled environments. However, study of these species has been hindered by the difficulties that exist in the identification and assignment of individuals to species and hybrid classes. Cytogenetics has been employed in a few studies (e.g. Setzer, 1970; Howell & Black, 1981), but the technique is technically demanding and impractical for large-scale projects. In the present study we use molecular markers to evaluate the evidence for reproductive isolation in syntopic populations of F. notatus and F. olivaceus in four drainages in southern Illinois, and empirically evaluate the evidence for reproductive isolation. We also address the potential evolutionary consequences of hybridization (e.g. Dowling & DeMarias, 1993) by assessing the extent of mtDNA introgression that has occurred in each of these drainages.

Materials and methods

Sample collection and DNA extraction

Southern Illinois is geographically characterized by the Shawnee Hills, the eastern-most remnants of the Ozark uplands, which are ecologically more favourable to F. olivaceus, and the lowland large river habitats associated with the Mississippi and Ohio Rivers that are preferred by F. notatus (Braasch & Smith, 1965). Syntopic populations in southern Illinois were selected by review of museum records from the Ichthyological Collection at Southern Illinois University Carbondale. We selected sites that had consistently high relative abundance of both species. Three of the sites included in this study (Barren Creek, Kinkaid Creek, Wolf Creek) occur at or near abrupt transition zones between these respective habitat types. The fourth site, Brushy Creek, is a tributary of the Middle Fork of the Saline River, a stream that originates in the more gradual rolling prairie-type habitat of central Illinois, also preferred by F. notatus and skirts the northern edge of the Shawnee Hills, where F. olivaceus predominates, before discharging into the Ohio River. Allopatric reference populations of each species were selected by identifying drainages in Illinois and Missouri outside of the region of geographic overlap.

At each site, fish were sampled intensively by seine or hand net over a span of less than 100 m and pooled into a single sample. Fish were anaesthetized in MS-222 and either preserved in 95% EtOH, or fin clips or liver tissue samples preserved in 95% EtOH before preservation in 10% formalin and later transfer to 60% EtOH. Genomic DNA was extracted with the aid of AquaPure Genomic DNA Tissue Kits (Bio Rad, Hercules, CA, USA).

Microsatellite analysis

The four unlinked tri-nucleotide repeat loci used to analyse the nuclear genomes of samples included FhATG B101, FhATG B103, FhATG 6 and FhATG 20 (Adams et al., 2005). For each locus, the forward primer was 5′ end labelled with a unique fluorescent dye. Polymerase chain reaction (PCR) reactions (25 μL volume) for microsatellite analysis and all other PCR applications reported in this study, contained approximately 25–50 ng DNA, 200 mm each dNTP, 1 μm each primer, and 1 unit of ExTaq DNA polymerase (Takara Bio Inc., Otsu, Shiga, Japan) in 1x buffer (1.5 mm MgCl2). Thermal cycling conditions included an initial denature step of 95 °C for 5 min, 40 cycles of 94 °C for 20 s, 55 °C for 30 s and 72 °C for 2.5 min, and a final step of 72 °C for 5 min. Amplified products were electrophoresed on an ABI 310 Genetic Analyzer (Applied Biosystems Inc., Foster City, CA, USA) and raw data was analysed using genemapper, ver. 3.0 (Applied Biosystems Inc.).

Nuclear gene intron analysis

Two nuclear gene loci were partially sequenced to identify candidate single nucleotide polymorphisms (SNPs) that could be used to discriminate between the two species and diagnose individuals of hybrid ancestry. These loci included triose phosphate isomerase (TPI) intron 3–4 (294 bp) and ribosomal protein L7a (SURF3) intron 3–4 (420 bp). The respective primers were TPI(F) 5′-TATAGCGGCCGCATCAAGGACTGCGGCGTGA and TPI(R) 5′-GCGATGACTTTGGTCTG, and SURF3(F) 5′-TATAGCGGCCGCGCAGGAGAAGAGGAAG and SURF3(R) 5′-TCCACCAGAGAGGT (M. Vaughan, personal communication). Both gene introns were amplified following the same conditions used for microsatellite amplification and cloned into the pGEM-T cloning vector (Promega, Madison, WI, USA). Sequencing templates were prepared by PCR amplification of cloned products using vector primers complimentary to the T7 and Sp6 RNA polymerase promoters. PCR products were purified with QIAquick PCR purification columns (Qiagen, Valencia, CA, USA) and added to BigDye Terminator version 3.1 cycle-sequencing reactions (Applied Biosystems Inc.) which were loaded on an ABI Prism 310 Genetic Analyzer (Applied Biosystems Inc.). Templates were sequenced on both strands, assembled and aligned using genetools version 2.0 (BioTools Inc., Edmonton, AB, Canada). Complete sequences from allopatric sites were determined for one F. olivaceus (Gasconade River) and one F. notatus (Piasa Creek) individual.

A discriminate PCR technique was developed to screen individuals at one of the SNP sites identified at the TPI locus. Alternative forward primers were designed to generate a match/mismatch at the 3′ end of each oligonucleotide to the G/T SNP site. A second mismatch was included three bases from the 3′ end to aid in destabilizing the primer/template complex for mismatched primers. The discriminate oligonucleotides were TPI(F)Oli 5′- GCGACGAGGTAATTCAGCAT and TPI(F)Not 5′- GCGACGAGGTAATTCAGCAG where the bold letters indicate nucleotide mismatches. Each primer was paired with TPI(R) in separate reactions and PCR was conducted using the same reaction conditions as for microsatellites with the annealing temperature modified to 58 °C. The presence/absence of diagnostic PCR products was determined by electrophoresis on a 2% agarose gel.

A restriction fragment length polymorphism (RFLP) assay was developed to screen the SURF3 intron at one C/A SNP site using MboI. Genomic DNAs were PCR amplified following conditions described for microsatellite loci. Homozygous and heterozygous digestion patterns could be discriminated by electrophoresis of digested products on a 2% agarose gel.

We used genepop ver. 3.3 (Raymond & Rousset, 1995) to calculate observed (HO) and expected (HE) heterozygosities as well as to conduct probability tests of Hardy–Weinberg equilibrium for each locus. A sequential Bonferroni (Rice, 1989) was applied in the determination of significant heterozygote deficiencies. Genotypic disequilibrium was assessed between all pairs of nuclear loci in each sample using the log-likelihood G-statistic with 3900 permutations as implemented in fstat, vers. 2.9 (Goudet, 1995). Factorial correspondence analysis (FCA) of nuclear allelic data was performed with genetix, vers. 4.0 (Belkhir, 1999). This canonical technique is used to graphically display individuals in multidimensional space on the basis of allelic information in order to display population structure (She et al., 1987; see Turgeon & Bernatchez, 2001).

Admixture analysis

Admixture proportions for each individual from each sample and ancestral allele frequencies for each gene locus in each species, were estimated using structure, vers. 2.0 (Pritchard et al., 2000; Beaumont et al., 2001). This program uses a Bayesian model-based approach to detect genetic structure by identifying discontinuities in multilocus allele profiles and assigning each multilocus genotype to a genetic cluster. Importantly for this study, the structure analysis does not require a priori assignment of alleles to taxonomic entities. Therefore, this analysis could be applied to samples collected from putative syntopic sites without prior knowledge of either the species identities of individuals in the samples, or allele frequencies associated with the respective species. We assumed that two distinct entities (i.e. species; K = 2) were present and that for the ith individual the proportion of the genotype that was drawn from F. notatus or F. olivaceus was qi, or 1−qi, respectively. When the analysis was repeated with the number of assumed populations set to K > 2 no additional geographic structure was detected in either species. The analysis was repeated using two modelling approaches. In one case we analysed individuals from all sympatric and allopatric sites (sites 1–13) assuming uninformative priors for all individuals; in the other case we analysed the same set of individuals and assumed point priors of 0 or 1 for individuals from the allopatric populations of F. olivaceus (site 13) and F. notatus (sites 11 and 12) respectively. Both the estimates of qi for all individuals at all sites and the confidence intervals, were nearly identical from these two approaches, suggesting that the assumed priors on the allopatric populations were reasonable, but unnecessary, and we only present results with no prior assumptions. In addition to a global analysis of all populations, we also analysed two drainages, the Barren Creek and Brushy Creek drainages, separately, to assess the influence of the total data set on the estimation of admixture values for individuals in separate drainages.

We used the distributions of qi at each site to examine the proportion of individuals that were of hybrid ancestry and assess whether these proportions deviated from null expectations under the assumption of no interspecific reproductive barriers (pre- or post-zygotic). For a pair of species in which reproductive barriers are strong and hybridization is rare, the distribution of qi is expected to be bimodal (i.e. mostly parental individuals with admixture values near 0 or 1), whereas, if reproductive barriers are weak or absent, a higher proportion of hybrid and backcross individuals would be present and the distribution would be unimodal (Jiggins & Mallet, 2000). To assess the modality of the populations, admixture values (qi) were transformed into hybridity (hi) values (sensuCarney et al., 2000; Edmands et al., 2005) which provide a measure of how intermediate, or close to 0.5, an individual is on the admixture scale. Hybridity values (ranging 0–0.5) were calculated by transforming the admixture proportion, qi, for each individual into a deviation from 0 or 1 (hi = 0.5 − |0.5 − qi|) and the mean hybridity (inline image) was estimated for each site. If no reproductive barriers exist between two species then, given enough generations, the gene pools of the respective species would merge. Under these conditions, the maximum attainable mean hybridity (inline image) would be a function of the relative abundance of the two species and would approach inline image, where inline image is the relative abundance of F. notatus estimated as the mean q among individuals in the sample. However, assessing reproductive isolation by comparing inline image to inline image for a natural hybrid zone according to these assumptions is complicated by the fact that even if no reproductive barriers exist, the populations are unlikely to be at equilibrium.

For purposes of testing for reproductive isolation, we made a conservative simplifying assumption that in each generation only individuals of nonhybrid ancestry reproduce. Under this assumption individuals would be either parental, or F1 hybrid and the expected mean hybridity (inline image) would be the sum of the frequencies of hybrid and nonhybrid offspring multiplied by their expected hybridity values (0.5 and 0, respectively). This expectation reduces to inline image and would be reached in only one generation of random mating. Our assumption is conservative because inline image for all values of inline image. If it could be shown that inline image, then this would provide evidence of reproductive isolation between sympatric populations of the two species.

We used a bootstrap procedure to test for nonrandom mating within each site. A distribution of inline image was generated from a simulated population of 1000 reproducing individuals. In the simulation, individuals in the parent population were randomly assigned an observed q value (F. notatus defined as all individuals >0.975 and F. olivaceus <0.015) from the pooled dataset. Relative proportions of each species were based on the observed inline image for a site adjusted for species mean q (F. notatus mean = 0.990, F. olivaceus = 0.008). In each permutation, pairs were formed at random (number of pairs equalled the number of individuals collected at a site) and offspring assigned a q-value (mean of the parents). inline image was then calculated based on the simulated progeny. 10 000 permutations were used to generate the inline image distribution.

mtDNA analysis

Polymerase chain reaction was used to amplify a portion of the mitochondrial cytochrome b gene using oligonucleotide primers GLUDG-L and CB3-H (Palumbi et al., 1991). PCR reactions followed standard conditions with the annealing temperature modified to 45 °C. PCR products were purified and sequenced following previously described protocols. Sequences generated from both strands were aligned and analysed using Genetools version 2.0 (BioTools Inc.).

In order to identify and validate species-diagnostic enzymes for RFLP analysis of mtDNA genomes, a phylogenetic analysis of 725 base pairs of the cytochrome b gene was conducted and then aligned sequences were compared to identify polymorphic sites that discriminated between the two clades corresponding to F. notatus and F. olivaceus. The phylogenetic analyses included F. notatus and F. olivaceus individuals from the northern (Illinois, Missouri) and southern (Mississippi) extents of their range as well as specimens of F. euryzonus, F. sciadicus, F. catenatus and F. heteroclitus, the latter three serving as outgroups for the F. notatus complex. Phylogenetic analyses were conducted with mega, vers. 3.0 (Kumar et al., 2004), employing the Neighbour-Joining (NJ) method with a Kimura 2-parameter distance matrix and a uniform rate across sites. The degree of confidence for each node was determined by bootstrap analysis with 1000 replicates (Felsenstein, 1985). A Bayesian analysis was also conducted using mrbayes, vers. 3.1 (Huelsenbeck & Ronquist, 2001) with the data partitioned by condon position and partitions allowed to have independent gamma distributed (‘‘invgamma’’) rates. The analysis was performed for 5 000 000 generations with the burnin set to 4000. Stationarity was assessed by examining plots of log-likelihood scores vs. generation time.

Mitochondrial DNA RFLP analysis involved digestion of amplified PCR products using Hae III and Nsi I restriction endonuclease enzymes which were identified by comparison of digest maps generated from DNA sequences using genetools version 2.0 (BioTools Inc.). The Hae III enzyme recognized a derived site (T→C) at position 263 diagnostic for F. notatus mtDNA and the Nsi I enzyme recognized a second derived site (A→G) at position 120 that was polymorphic among individuals with mtDNA haplotypes belonging to the F. notatus clade. These two enzymes in combination distinguished three haplotypes designated oli (−, −), not1 (+, −) and not2 (+, +), respectively. Digested products were electrophoresed on a 2% agarose gel and visualized using ethidium bromide dye.


In the summers of 2002, 2003 and 2004 four drainages were sampled in southern Illinois within the region of sympatry between F. notatus and F. olivaceus, and three allopatric drainages were sampled. No single site was sampled on more than one occasion. In allopatric drainages, morphological characteristics (primarily spot phenotypes) were consistent with the species reported for each location (Pflieger, 1975; Smith, 2002). In collections from putative sympatric streams, a continuum of spot phenotypes was observed, ranging from individuals with no spots to individuals with many prominent well defined spots, consistent with the expectation that both species were present and syntopic (Table 1). However, this trait was highly variable and no attempt was made to quantify the variation.

Table 1.   Collection sites for Fundulus olivaceus, Fundulus notatus and Fundulus euryzonus.
LocationSpeciesWater bodyDrainageLatitude (N)/Longitude (W)
1F. olivaceusBarren Cr.Barren Cr., Ohio R.37 °12.97′/88 °33.83′
2F. olivaceusBarren Cr.Barren Cr., Ohio R.37 °14.18′/88 °31.77′
3F. olivaceusCaney Cr.Barren Cr., Ohio R.37 °14.85′/88 °30.48′
4F. olivaceusCooney Cr.Barren Cr., Ohio R.37 °15.24′/88 °31.68′
5F. not/F. oliCooney Cr.Barren Cr., Ohio R.37 °15.10′/88 °31.33′
6F. not/F. oliCave Cr.Barren Cr., Ohio R.37 °15.36′/88 °30.89′
7F. notatusBarren Cr.Barren Cr., Ohio R.37 °15.27′/88 °30.27′
8F. not/F. oliBrushy Cr.Saline R., Ohio R.37 °46.53′/88 °39.18′
9F. not/F. oliKinkaid Cr.Big Muddy R., Miss. R.37 °46.60′/89 °27.00′
10F. not/F. oliWolf Cr.Cache R., Miss. R.37 °13.01′/89 °20.12′
11F. notatusKaskaskia R.Kaskaskia R., Miss. R.39 °01.66′/89 °05.13′
12F. notatusPiasa Cr.Mississippi R.38 °59.04′/90 °15.65′
13F. olivaceusGasconade R.Missouri R.37 °39.82′/92 °18.53′
14F. olivaceusOkatoma Cr.Pascagoula R.31 °28.88′/89 °25.88′
16F. olivaceusBogue Chitto R.Pearl R.31 °01.27′/90 °12.49′
15F. notatusPearl R.Pearl R.31 °14.31′/89 °50.84′
17F. euryzonusE. Fork Amite R.Lake Pontchartrain31 °12.30′/90 °40.33′

A joint analysis of admixture proportions from all sympatric and allopatric population samples collected in Illinois and Missouri (all individuals from sites 1–13) provided strong support for the presence of a genetic partition between two distinct gene pools (Fst = 0.36), that were consistent with F. notatus and F. olivaceus respectively. Individuals from allopatric samples exhibited qi values ≥0.96 (F. notatus) or ≤0.026 (F. olivaceus). Individuals from sites collected in each of the four sympatric drainages exhibited a wide range of qi-values ranging from 0 to 1, consistent with the presence of both species in each of these drainages (Fig. 2). Additionally, intermediate qi values for numerous individuals in each of the sympatric drainages provided evidence for the presence of hybrids (Fig. 2). The largest number of hybrid individuals was observed at the Brushy Creek site, a sample in which our analysis indicated the presence of both parental species at roughly equal proportions. In the Barren Creek drainage, hybrid individuals were limited to two of the seven sites (sites 5 and 6) indicating the zone of contact between the two species.

Figure 2.

 Admixture proportions (qi) assigned to each individual from sympatric drainages in southern Illinois when all individuals from allopatric and sympatric drainages were analysed together. Within each collection site, individuals were ordered along the x axis by increasing admixture proportion. Light gray symbols indicate admixture proportion estimates for Brushy Creek and the Barren Creek drainage when each of these samples was analysed separately. Bars indicate the 80% probability regions for estimates of qi. Individuals with Fundulus olivaceus mtDNA haplotypes are indicated with black circles whereas Fundulus notatus mtDNA haplotypes are indicated with white diamonds.

To address the concern that admixture values were biased by the inherent geographic structure in allele frequencies among conspecific populations from separate drainages, Barren Creek drainage samples (sites 1–7) and Brushy Creek (site 8) samples were each reanalysed separately. In general, results from the combined and separate analyses did not differ, but with decreased sample sizes in separate analyses, confidence intervals for individual qi values were generally larger than when drainages were analysed together (Fig. 2).

Factorial correspondence analysis of nuclear allelic data also supported the genetic distinctiveness of F. notatus and F. olivaceus with allopatric population samples of the respective species separating into two distinct clusters (Fig. 3) along the first axis. The alleles that contributed most to this difference (Table 2) were primarily those alleles that were relatively common in only one of the two species. Most individuals from sympatric populations clustered with the allopatric specimens as well (Fig. 3), whereas individuals that did not group with the allopatric clusters corresponded to the individuals with intermediate q-values in the admixture analysis further supporting the conclusion they were of hybrid ancestry.

Figure 3.

 Individual genotypes plotted in ordination space defined by FCA axes 1 and 2 (per cent variance accounted for by each axis indicated). Fundulus olivaceus mtDNA haplotypes are plotted as black circles, Fundulus notatus mtDNA haplotypes as white diamonds. Individuals from allopatric populations of each species are shown as gray circles and diamonds respectively.

Table 2.   Estimated ancestral nuclear allele frequencies among Fundulus olivaceus and Fundulus notatus populations. The top fifteen alleles with the largest contribution to axis 1 of the FCA are so indicated by rank in parentheses.
LocusAlleleF. olivaceusF. notatus
FhATG 201110.0350.003
118 (8)0.7660.003
133 (14)0.0100.124
152 (13)0.0010.113
FhATG B101910.0020.004
96 (15)0.1490.001
98 (5)0.0010.737
107 (10)0.4230.001
FhATG 61860.0010.030
192 (9)0.3920.965
195 (11)0.3510.001
198 (12)0.2120.002
FhATG B103308-338 (4)0.0030.904
347-468 (3)0.9970.096
TPIG (7)0.0010.724
SURF3C (1)0.0020.998
A (2)0.9980.002

Description of nuclear DNA data

Among the four tri-nucleotide microsatellite loci used to analyse the nuclear genomes, the total number of alleles at each locus ranged from 6 (FhATG 6) to 50 (FhATG B103). Estimates of the inferred ancestral allele frequencies, based on the admixture analysis, revealed that most of the alleles were shared between species (Table 2). This result could reflect recent hybridization and limited introgression in sympatric populations, but for three of the four loci the same allele-overlap between species was observed in allopatric populations as well as sympatric populations. Therefore, it seems likely that allele size homoplasy (Estoup et al., 2002) exists between the two species. An alternative explanation, that the shared alleles represent ancestral polymorphism, is also possible, but considered less likely given the extent of mtDNA divergence (described below).

At the most variable locus, FhATG B103, there was a notable gap in the size range of alleles from 339 to 346. The frequency distributions of alleles in all collections indicated that size homoplasy between species was not evident and suggested that alleles ranging from 308 to 338 bp were associated with F. notatus whereas alleles ranging from 347 to 468 bp mostly occurred in F. olivaceus (Table 2). Exceptions were not encountered in allopatric populations of either species from the Gasconade River or Piasa Creek (Table 3) and only three copies of possibly historically introgressed F. olivaceus-like alleles (two copies of 413, one of 437) were found in an allopatric population of F. notatus from the Kaskaskia River. In order to facilitate the identification of individuals of hybrid ancestry, alleles in the respective size ranges were pooled into two size categories for the previously described admixture and factorial correspondence analyses, as well as the Hardy–Weinberg equilibrium and genotypic equilibrium tests.

Table 3.   Summary of mtDNA haplotype and dimorphic nuclear allele frequencies.
not1not2oli308–338347–468GTACinline image
  1. n, sample size.

  2. inline image, average admixture proportion estimated from four microsatellite and two SNP loci; defined as the proportion of alleles derived from Fundulus notatus.

(1) Barren Cr.230.130.871.
(2) Barren Cr.250.240.761.
(3) Caney Cr.190.370.631.
(4) Cooney Cr.
(5) Cooney Cr.240.380.630.080.920.100.900.100.900.10
(6) Cave Cr.230.870.130.700.300.700.300.750.250.78
(7) Barren Cr.61.000.830.170.750.251.000.99
(8) Brushy Cr.411.000.460.540.410.590.510.490.53
(9) Kinkaid Cr.360.890.110.760.240.320.680.890.110.85
(10) Wolf Cr.210.050.950.120.880.080.920.050.950.08
(11) Kaskaskia R.
(12) Piasa Cr.
(13) Gasconade R.

A survey of northern allopatric population samples and southern drainage reference specimens of F. notatus and F. olivaceus suggested that the SURF3 C/A SNP site was diagnostic for alleles of the respective species. This result was supported as well by inferred ancestral allele frequency estimates for the two species based on the admixture analysis of sympatric and allopatric population samples in which the C and A alleles appeared virtually fixed in the respective ancestral gene pools (Table 2).

At the TPI G/T SNP site, northern allopatric populations of F. notatus all exhibited the G allele, but four F. notatus specimens collected in the southern reference population from the Pearl River (site 15) exhibited the T allele. All allopatric and reference specimens of F. olivaceus exhibited the T allele (Table 3). Ancestral reconstruction of allele frequencies based on the admixture analysis indicated that F. notatus is polymorphic at this locus whereas F. olivaceus is fixed for the T allele (Table 2). The presence of a T allele in F. euryzonus supports the interpretation that G is the derived state and remains polymorphic in F. notatus, but other more complex interpretations, like hybridization and introgression in some populations, cannot be ruled out.

Evidence for Reproductive Isolation between F. notatus and F. olivaceus. Reproductive isolation between F. notatus and F. olivaceus in sympatric populations should result in a deficit of heterozygosity across all loci. When a sequential Bonferroni correction was applied (α = 0.05), few of the loci exhibited a statistically significant heterozygote deficiency (Table 4). However, when probabilities across all six loci were combined using Fisher's method (Sokal & Rolf, 1994), three of the four contact zones, including Brushy Creek, Wolf Creek and Kinkaid Creek, exhibited a highly significant heterozygosity deficit. Within the Barren Creek drainage, the Cave Creek site exhibited a significant heterozygosity deficit at locus FhATG B101, but the combined probability was not significant after sequential Bonferroni correction (Table 4). In contrast, Cooney Creek (site 5) did not exhibit a heterozygosity deficit at any of the loci.

Table 4.   Observed and expected heterozygosities (HO/HE) estimated for each sample and each locus.
SiteMicrosatellite lociSNP loci
  1. *Significant heterozygote deficiency based on exact test after column-wise sequential Bonferroni (α = 0.05).

  2. –, Sample is monomorphic.

  3. FhATG B103 Allele size classes pooled into two categories for analysis, 308–338, 347–468.

  4. P, Fisher's combined probability of heterozygote deficiency calculated by Fisher's method. Tests that were significant after sequential Bonferroni indicated by bold font.

(1) Barren Cr.0.39/0.660.52/0.670.48/0.510.0221
(2) Barren Cr.0.44/0.550.56/0.650.24/0.230.7884
(3) Caney Cr.0.33/0.560.68/0.650.11/0.100.6179
(4) Cooney Cr.0.71/0.630.71/0.630.43/0.400.9993
(5) Cooney Cr.0.67/0.710.17/0.160.83/0.660.46/0.490.13/0.190.21/0.190.6547
(6) Cave Cr.0.30/0.41*0.34/0.430.30/0.310.83/0.890.35/0.470.23/0.380.0146
(7) Barren Cr.0.17/0.170.33/0.300.00/0.000.67/0.860.50/0.410.8313
(8) Brushy Cr.0.54/0.650.29/0.500.39/0.360.71/0.810.34/0.490.31/0.510.0004
(9) Kinkaid Cr.0.44/0.550.25/0.370.08/0.160.81/0.870.36/0.440.06/0.20*0.0000
(10) Wolf Cr.0.52/0.76*0.14/0.210.71/0.720.38/0.430.05/0.140.00/0.100.0037
(11) Kaskaskia R.0.27/0.240.07/0.190.80/0.830.3299
(12) Piasa Cr.0.40/0.360.60/0.360.60/0.720.6556
(13) Gasconade R.0.17/0.170.50/0.530.67/0.561.0000

Sites in which limited hybridization and backcrossing has occurred should exhibit high levels of genotypic disequilibrium (Barton & Gale, 1993). Among the 13 samples analysed, Brushy Creek exhibited the highest levels of genotypic disequilibrium with nine of the fifteen pairwise locus comparisons exhibiting statistically significant deviations from equilibrium (Table 5). Cave Creek and Kinkaid Creek also exhibited genotypic disequilibrium among multiple loci (Table 5). Among the samples that did not exhibit genotypic disequilibrium, many of the pairwise locus comparisons were not possible due to the lack of polymorphism at several loci (particularly the SNP loci).

Table 5.   Genotypic equilibrium tests (P-values) for all polymorphic locus pairwise comparisons in each population sample. P-values in bold are significant after table-wise sequential Bonferroni (α = 0.05).
SiteB101 XB101 XB101 XB103 XB103 X6 XB101 XB103 X6 X20 X101 XB103 X6 X20 XTPI X
  1. –, Indicates a pairwise comparison in which one or both loci were monomorphic.

(1) Barren Cr.0.71080.36900.3000
(2) Barren Cr.0.51000.93440.4985
(3) Caney Cr.0.96641.00000.4295
(4) Cooney Cr.1.00000.54560.7782
(5) Cooney Cr.0.04870.38080.26870.15740.01800.50440.06360.11510.83360.04820.00770.01770.44000.00150.0174
(6) Cave Cr.0.00210.00050.82740.00130.06920.50440.01260.00100.01100.23180.00050.00030.00050.32640.0003
(7) Barren Cr.1.00001.00001.00001.00000.40721.0000
(8) Brushy Cr.0.00030.03720.11030.00130.00030.86870.00030.00030.05030.00030.00030.00030.10100.00030.0003
(9) Kinkaid Cr.0.00510.00080.50030.00030.66080.88690.75620.23261.00000.24950.00080.00030.00050.63150.1082
(10) Wolf Cr.0.68280.76720.20210.71180.36590.44670.52950.03900.60030.30770.34640.17900.72130.30920.1985
(11) Kaskaskia R.0.06031.00001.0000
(12) Piasa Cr.0.72050.92771.0000
(13) Gasconade R.1.00000.32180.0713

An alternative method of quantifying the extent of reproductive isolation between F. notatus and F. olivaceus is with a hybridity index. Brushy Creek exhibited the highest inline image of 0.13 (Fig. 4), which corresponds roughly to a hybrid ancestry for 11 of the 41 individuals sampled. This estimate compares favourably with the admixture proportions for individuals from this site among which 11 exhibited a value of qi with an 80% probability interval that included 0.5 (Fig. 2). The inline image for this population was 0.25, suggesting that the observed proportion of hybrids was only half of that expected under the null hypothesis of no reproductive barriers. Significant deficits of hybrid individuals were obtained for Brushy Creek, Kinkaid Creek and Cave Creek (P ≤ 0.0001, 0.0006 and 0.015, respectively). In contrast, Cooney Creek (site 5) and Wolf Creek exhibited elevated values of inline image that were not significantly different from inline image (P ≤ 0.44 and 0.12 respectively). The power to detect a deficit of hybrid individuals in these samples may be limited by the large bias in species prevalence at these sites. All of the other populations exhibited inline image values close to zero and nearly identical to inline image. The nonzero inline image values for these remaining populations is attributable to the uncertainty involved in assigning admixture proportions to individuals that do not have a hybrid ancestry.

Figure 4.

 Bar graph of observed mean hybridity values (inline image) for each site. Also shown are the expected mean hybridity (inline image) values (circles) assuming random mating and the 95% cut-off for each site based on a bootstrap analysis. Filled bars represent observed hybridity indices not significantly different from the random mating model.

Mitochondrial DNA

A phylogenetic analysis of mtDNA cytochrome b sequences using NJ and Bayesian methods indicated the monophyletic relationship of F. notatus and F. olivaceus with respect to F. euryzonus, with an average uncorrected distance of 8.2% (Fig. 5). Two clades, corresponding to F. olivaceus and F. notatus, respectively, exhibited an average uncorrected distance of 5.6%. The two putative cytonuclear introgressed F. olivaceus individuals (Brushy Cr. 1, Caney Cr. 3–2) that grouped with the F. notatus clade were sampled from syntopic sites in Illinois as indicated by the admixture analysis. The relationship between mtDNA clades and the RFLP patterns identified in this study is illustrated in Fig. 5 and the frequencies of each haplotype are provided in Table 3.

Figure 5.

 Relationship of cytochrome b sequences obtained for Fundulus notatus (N), Fundulus olivaceus (O) and Fundulus euryzonus (E). Individuals are identified by the location from which they were sampled and numbers in parentheses indicate the total number of individuals with identical sequences determined from each location. The rooted phylogeny was constructed by the NJ method with Fundulus sciadicus, Fundulus catentatus and Fundulus heteroclitus serving as outgroups (not shown). Numbers above the branch at each node indicate NJ bootstrap support whereas numbers below the branch correspond to the Bayesian posterior probabilities. Labels on the right correspond to the haplotypes identified by RFLP analysis with Hae III and Not I enzymes.

The origin of mtDNA in putative hybrid individuals (those with intermediate admixture proportions) seems to be primarily (although not exclusively) from F. notatus at all sites except Wolf Creek (Fig. 2). This trend is also readily apparent in the FCA plot (Fig. 3). In the Barren Creek drainage, introgression of F. notatus mtDNA is apparent at nearly all F. olivaceus sites, including site 1, which is near the headwaters of the drainage and roughly 6.2 km from the zone of contact (Fig. 1). In contrast, no evidence exists for introgression of F. olivaceus haplotypes into the F. notatus population (site 7) from outside the contact zone (although sample size was small). At the Brushy Creek site, extensive mtDNA introgression is evidenced by the fact that both species appear fixed for F. notatus mtDNA.


In this study we documented the occurrence of hybridization in four drainages in southern Illinois where F. notatus and F. olivaceus are syntopic. The most detailed analysis focused on Barren Creek where the transition zone between F. notatus and F. olivaceus habitats is both abrupt and well defined. The two species were only found in syntopy at the interfaces where the small streams in this drainage empty into a large backwater habitat that comprises the interface between the lower reach of Barren Creek and the Ohio River. The transition between these two habitats occurs at approximately 95–100 m elevation and is coincident with a steep transition in both admixture proportions and mtDNA haplotypes characteristic of each of the respective species (sites 5 and 6; Fig. 2). Analysis of admixture proportions in individuals collected outside of this contact zone revealed no evidence of migration of F. notatus individuals beyond the ecological transition zone or introgression of nuclear DNA (as would be suggested by intermediate admixture proportions) outside the hybrid zone over distances as short as 630 m (distance from site 4 to the confluence of the stream at site 5) and approximately 400 m (distance from site 3 to the habitat transition zone). A mark-recapture study in a small creek in central Arkansas revealed seasonal dispersal of F. olivaceus among collection sites on a scale of hundreds-of-meters (Smithson & Johnston, 1999). Therefore, this spatial scale is presumably within the order of magnitude of dispersal capability for both of these species.

Evidence of mtDNA introgression, at Barren Creek, Brushy Creek and Kinkaid Creek, unequivocally establish that cytoplasmic genetic introgression has occurred among the two species. At the most extreme, the F. olivaceus population at Brushy Creek appears to have become fixed for F. notatus mtDNA. Numerous studies in other fish species (e.g. Dowling & Hoeh, 1991; Duvernell & Aspinwall, 1995; Wilson & Bernatchez, 1998) have demonstrated that mtDNA introgression often readily occurs between hybridizing species. The existence of extensive mtDNA introgression in some F. olivaceus population samples provides evidence that hybrid individuals can and do contribute to the gene pools of these two species.

Examination of the data for common SNP and microsatellite alleles of one species that were present in individuals otherwise comprised of alleles from the other species (as assessed by admixture proportion) suggested the presence of numerous backcross individuals in syntopic populations. This interpretation of the data is consistent with population cytological studies that revealed karyotypes consistent with numerous backcross as well as F1 hybrid individuals in syntopic populations (Setzer, 1970; Howell & Black, 1981). The occurrence of backcross, and F1 hybrid individuals in syntopic populations, mtDNA introgression and more limited nuclear introgression, brings into question the nature of reproductive isolation between these two species. The exact tests for homozygote excess, tests of linkage disequilibrium and the use of a hybridity index based on estimates of admixture proportions, in combination, revealed a substantial deficit of hybrid individuals in all four of the syntopic drainages indicating that these hybrid zones are strongly bimodal (sensuJiggins & Mallet, 2000).

Several possible explanations exist for a deficit of hybrid individuals in syntopy including: (i) assortative mating driven by mate preference, (ii) spatial and/or temporal separation of spawning events on a microgeographic or seasonal scale, (iii) extensive immigration of individuals into the contact zone from surrounding habitats each generation, (iv) reduced fitness of individuals with hybrid ancestry and (v) reduced viability of hybrid offspring resulting from cytological or other genetic incompatibility. Some of these hypotheses have been addressed in previous studies and others remain to be explored.

The issue of cytogenetic mechanisms of reproductive isolation has been addressed extensively in the literature. Inferences of reduced hybrid viability leading to reduced gene flow among chromosomal ‘races’ has been documented in a variety of organisms (e.g. mammals –Muñoz-Muñoz et al., 2003; Panithanarak et al., 2004; reptiles –Reed et al., 1995; and insects –Gorlov & Tsurusaki, 2000; Chiappero et al., 2004). Thomerson (1966) indicated F1 hybrids between F. notatus (2n = 40) and F. olivaceus (2n = 48) are readily generated in the lab, and that F1 offspring were capable of producing F2 and backcross progeny. Thus, Thomerson (1966) concluded that any isolating mechanisms in nature are ecological or behavioural rather than cytogenetic. The observed backcross progeny and cytoplasmic introgression in natural populations are both consistent with at least partial viability of F1 hybrids. Other literature references to the viability of F1 hybrids have cited Thomerson's work (e.g. Setzer, 1970; Howell & Black, 1981). However, Cuca (1976) and J. F. Schaefer (unpublished data) have observed high egg mortality (>95%) for lab-reared F2 progeny, a result not addressed by Thomerson (1966). Therefore, it remains possible that F1 offspring actually exhibit reduced reproductive viability, possibly through high frequencies of nondisjunction events, contributing to reproductive isolation between the two species (J. E. Thomerson, personal communication).

Even if partial F1 hybrid inviability remains as a possible reproductive isolating mechanism limiting introgression of nuclear DNA, it is insufficient to explain the deficit of hybrid individuals in syntopy if F1 hybrids were produced by random mating and exhibited normal fitness. Other mechanisms of reproductive isolation must also be present. Gut content analysis of F. notatus and F. olivaceus from syntopic populations in southern Illinois and south-eastern Missouri indicated that the two species occupy similar trophic niches and that resource competition could lead to competitive exclusion (Thomerson & Wooldridge, 1970). If resource limitations and competition are responsible for limiting the syntopy between these two species, then reduced hybrid fitness could also be an important reproductive barrier. This conclusion was also drawn by Howell & Black (1981), but has not been carefully evaluated in any type of analysis of hybrid fitness in natural populations.


The F. notatus species complex is an ideal system in which to advance our understanding of ecological diversification, speciation, species interactions and hybrid zone dynamics. In this study we have developed a molecular-based approach to objectively assign individuals to species classes and quantify admixture proportions. Additionally, we have demonstrated that these species interact in their natural environment in complex ways, the outcomes of which are likely determined by ecological factors. Evidence presented here that F. notatus and F. olivaceus hybridize in habitats where they are sympatric also suggests that the barriers that maintain reproductive isolation between these species may be complex, involving both intrinsic and extrinsic factors.


We are thankful to B. Kreiser, J. Thomerson and three anonymous reviewers for valuable comments on early drafts. We also thank C. Taylor, J. Stewart, S. Adams, S. Pryor, R. Melvin, D. Hanks, J. Fonoti, S. Kunkel, K. Parsens, J. Spaeth, T. Darden, B. Kassebaum, R. Ramirez and J. Kerfoot for assistance with sample and data collection. Specimens were collected under Illinois permit A03.0763, Missouri permit 12335 and Mississippi permit 09-25-04. DNA sequences have been deposited in GenBank under accession numbers DQ179601DQ179630. All specimens have been deposited in the University of Southern Mississippi Museum of Ichthyology. This work was funded through grants from the Undergraduate Research Academy at SIUE to AMR and the NSF (OCE-0221879) to DDD.