Hybridization dynamics between sympatric species of trout: loss of reproductive isolation


Daniel D. Heath, Great Lakes Institute for Environmental Research and, The Department of Biological Sciences, University of Windsor, 401 Sunset Avenue, Windsor, Ontario Canada N9B 3P4.
Tel.: (519) 253 3000 ext. 3762; fax: (519) 971 3616; e-mail: dheath@uwindsor.ca


Although reinforcement should enhance reproductive barriers in sympatric species, sympatric trout species do hybridize. Using mitochondrial and nuclear species markers, we investigated hybridization directionality, hybrid mating biases, and selection against hybrids in 13 sympatric cutthroat and rainbow trout populations on Vancouver Island, Canada. Approximately 50% of the genotyped fish were hybrid (F1 or higher-order) and populations ranged from very recent (all F1 hybrids) to extremely advanced higher-order hybridization. Overall, interbreeding was reciprocal, although some populations showed directional hybridization. Pronounced cytonuclear disequilibrium in post-F1 hybrids indicated a remarkable mating bias not previously reported, which is most likely because of behavioural reproductive preferences. Selection against hybrids was observed in only two populations, indicative of extrinsic selection. Two populations were ‘hybrid swarms’, with a complete loss of reproductive isolation. The complex hybridization dynamics in this system represent a valuable natural experiment of the genetic and evolutionary implications of recent and on-going interspecific hybridization.


As recently as the 1960s, hybridization among taxa was not considered an important evolutionary or ecological process because it was presumed that hybrid fauna were rare in nature (Mayr, 1963). However, there have been many examples of animal hybridization reported in nature over the last three decades (e.g. Howard, 1986; Heath et al., 1995– invertebrates; Grant & Grant, 1992– birds; Avise & Saunders, 1984; Hatfield & Schluter, 1999– fish). The occurrence of hybridization in natural and disturbed systems has since raised important questions regarding the role of reproductive isolating mechanisms in maintaining species, such as: (1) why have reproductive mating barriers failed in many interspecific crosses; and (2) what are the consequences of those failures? Biologists have recognized the importance of both prezygotic and post-zygotic reproductive isolation for maintaining species, and both are believed to intensify with divergence time between taxa (Coyne & Orr, 1997). Furthermore, it has been hypothesized that prezygotic barriers may evolve faster than post-zygotic reproductive barriers because of the effects of reinforcement in species pairs that maintain a sympatric relationship, especially when reciprocal hybridization events have occurred (e.g. Coyne & Orr, 1989, 1997; Noor, 1999; Servedio, 2000).

Many species exhibit persistent levels of hybridization, where the hybrid genotypes are spatially limited in distribution (Barton & Hewitt, 1985; Harrison, 1990). Two models widely used to explain such hybrid zone stability are the ‘tension zone’ and ‘mosaic’ models (Burke et al., 1998). The tension zone model (Barton & Hewitt, 1985, 1989) postulates that the stability and size of hybrid zones are maintained by a balance between intrinsic selection (i.e. environmentally-independent selection) against hybrids and the dispersal of parental genotypes, where the intensity of selection against hybrids determines the size and distribution of the hybrid zone. The mosaic model (Howard, 1986) assumes that hybrids are also competitively inferior; however, it differs from the tension zone model in that the distribution of parental genotypes is governed by extrinsic selection (i.e. environment-dependent selection). Distribution of hybrids in the mosaic model reflects the adaptation of the parental genotypes to habitat heterogeneity (Moore & Price, 1993; Burke et al., 1998) resulting in the hybrid genotypes inhabiting ‘transition zones’. Reviews of hybrid zone stability (e.g. Barton & Hewitt, 1985, 1989) have concluded that intrinsic selection is likely the principal factor contributing to observed hybrid zone stability. However, extrinsic selection has also been demonstrated in some hybrid zones (Harrison, 1990; Arnold, 1997) and is increasingly being recognized as an important factor in speciation (e.g. Hatfield & Schluter, 1999; Rundle, 2002). Areas of recent loss of reproductive isolation and consequent hybridization between naturally co-occurring species prior to the development of ‘stable’ hybrid zones represent natural experiments for determining the genetic and evolutionary processes that shape the patterns of existing hybrid zones, as well as inform our understanding of reproductive isolation among species.

Scribner et al. (2001) showed that hybridization is more common among fish species than in any other vertebrate group (see also Campton, 1987; Allendorf & Waples, 1996). Several factors have been proposed as contributing to the high incidence of hybridization in fish, including; competition for spawning habitat, external fertilization, weak behavioural isolating mechanisms, and unequal abundance of species (Hubbs, 1955; Campton, 1987). The combination of these life history and behavioural traits make fish ideal subjects for the study of hybridization and the breakdown of reproductive isolation. Scribner et al. (2001) identified the existence of weak prezygotic barriers among numerous species pairs of fish as well as relatively minor post-zygotic reproductive barriers among several species pairs. Very few of the studies reviewed by Scribner et al. (2001) directly examined the extent of hybrid inferiority in fish or the role of selection (intrinsic or extrinsic) against the hybrids. Dowling & Moore (1985) reported that prezygotic (i.e. reinforcement) mechanisms were weak between two species of Cyprinidae and that the hybrids produced were selected against post-reproductively; however, they did not determine whether the selection was intrinsic or extrinsic. Hatfield & Schluter (1999) established that the fitness reduction observed in F1 stickleback hybrids was primarily extrinsic (hybrid inability to adapt to either parental habitat). Their findings, however, only included the F1 generation and no data on either intrinsic or extrinsic selection effects in backcross or higher order hybrids was presented. Although hybrid fish often appear to exhibit reduced fitness relative to the parental pure-types, it is generally not clear whether extrinsic or intrinsic selection pressures are driving the reduced fitness.

Coastal cutthroat trout (Oncorhynchus clarkiclarki Richardson) and coastal rainbow trout (O. mykiss irideus Walbaum) are two species of salmonid native to the Pacific coast drainages of North America that exhibit broad sympatric distributions. Coastal cutthroat trout's native range spans approximately 3000 km along the Pacific coast from as far south as Humboldt Bay, California to Prince William Sound, Alaska (Trotter, 1989). The native range of coastal rainbow trout includes the coastal areas from central California to the Alaska Peninsula (Burgner et al., 1992). Both species have anadromous and freshwater-resident life histories; anadromous coastal rainbow trout are generally referred to as ‘steelhead’ trout while anadromous cutthroat trout are referred to as ‘sea-run’ coastal cutthroat trout.

Sea-run coastal cutthroat and steelhead trout typically spend 2–4 years in freshwater before undergoing physiological changes (i.e. smoltification) that facilitates seawater migration (Trotter, 1989; Pearcy et al., 1990; Behnke, 1992). Sea-run cutthroat trout characteristically inhabit estuaries and near-shore coastal marine ecosystems prior to maturation when they migrate and over-winter in freshwater, and reach maturity at approximately 250–450 mm in fork length (Behnke, 1992). Steelhead trout normally spend 1–3 years in the marine environment, migrating long distances before returning to their natal streams as mature adults (Pearcy et al., 1990) between 350 and 800 mm in fork length (Withler, 1966). Freshwater-resident life histories for both species vary considerably, particularly in resident coastal cutthroat trout (see Trotter, 1989). In most cases, the nonmigratory fish mature at smaller sizes and frequently coexist in the same habitat and streams as their anadromous counterparts.

Cutthroat trout and rainbow trout are believed to have diverged from a common ancestor approximately 2 million years ago (Behnke, 1992) and have since evolved into several subspecies within the cutthroat and rainbow trout (see Allendorf & Leary, 1988; Behnke, 1992). Evidence for the divergence between the two trout species includes substantial genetic (Leary et al., 1987; Allendorf & Leary, 1988; Oakley & Phillips, 1999), chromosomal (Gold, 1977), and morphological (Hartman, 1956; Bisson et al., 1988; Behnke, 1992) differences. Stocking of non-native rainbow trout into areas housing naturally allopatric cutthroat trout has resulted in extensive hybridization (and introgression) between trout species (e.g. Leary et al., 1984; Ferguson et al., 1985; Carmichael et al., 1993; Rubidge et al., 2001; Campbell et al., 2002); in some instances, hybrid swarms have been reported (Forbes & Allendorf, 1991). Such high levels of hybridization have been attributed to the absence of reproductive isolating mechanisms. However, in some areas, coastal cutthroat and rainbow trout have existed in sympatry and have maintained their species integrity in sympatry since recolonization after the last glaciation (i.e. ∼10 000 years) (Behnke, 1992). The long-term sympatry of these trout populations not only represents a valuable opportunity for the study of species conservation, but also an important example for the study of the evolution of reproductive isolation. The lack of geographical barriers separating the sympatric populations has been proposed as the driving force for the evolution of temporal and spatial reproductive barriers (Trotter, 1989; Young et al., 2001). However, hybridization has recently been reported in sympatric populations throughout their range, indicating that the loss of reproductive isolation between these species is geographically widespread (Campton & Utter, 1985; Young et al., 2001; Docker et al., 2003; Ostberg et al., 2004). The focus of those studies addressed several important aspects – for example – first reports of hybridization (e.g. Campton & Utter, 1985), detailed genetic characterization of a limited number of hybrid trout (Young et al., 2001), association between rainbow trout supplementation and elevated levels of hybridization (Docker et al., 2003), and spatial partitioning between two populations of sympatric coastal cutthroat and steelhead trout (e.g. Ostberg et al., 2004). However, none of those studies addressed among-population variation in reproductive isolation and detailed cytonuclear genomic dynamics in multiple hybridizing populations.

Here we investigate hybridization dynamics in 13 populations of sympatric coastal cutthroat and coastal rainbow trout from Vancouver Island, British Columbia, where extensive hybridization was suspected based on previous work (Docker et al., 2003). We utilized a suite of seven species-specific co-dominant nuclear markers and one mitochondrial DNA (mtDNA) marker, which provide the ability to evaluate hybridization dynamics using genetic analyses. Using multilocus genotype patterns and measures of linkage disequilibria, we tested for evidence of recent vs. advanced hybridization events. To test for hybridization directionality (i.e. unidirectional vs. reciprocal) we examined the mtDNA haplotype of F1 hybrid fish. We tested for cytonuclear disequilibrium and nuclear-mitochondrial allele associations to determine if there was evidence for nonrandom hybridization beyond the initial hybridization events (i.e. in backcross and higher-order hybrids). To determine the selective consequences of hybridization, we compared the proportion of hybrids in the fry (age 0+) vs. older fish that had survived at least one winter. Our analysis showed that the breakdown of reproductive isolation in the 13 study populations ranged from very recent (all hybrids were F1) to extremely advanced, where the two species had formed hybrid swarms, with very few pure-type fish remaining.

Materials and methods

Sample collection

Thirteen sympatric populations of coastal cutthroat and rainbow trout were sampled on Vancouver Island, British Columbia (Fig. 1; n = 29–38 fish per population). The populations were chosen for known high levels of hybridization, based on preliminary genetic marker screening of a total of 37 populations of sympatric trout (Bettles, 2004). The purpose was to examine hybridization dynamics in populations showing moderate to high levels of hybridization. One population (Chase River) was sampled in both 2002 and 2003, and we include data from both sample years to address questions of temporal stability. All fish were collected during early/mid summer 2002 (22 June to 30 July) and 2003 (20 June to 7 July) using 2-pass backpack electroshocking (Model LR-24, Smith-Root, Vancouver, WA, USA). Captured fish (n = 462) were anaesthetized using a mixture of clove oil and stream water (10–15 ppm). Each fish was measured for body length from the tip of the snout to the fork of the tail (‘fork length’±1 mm), and fin clips were collected and stored in 95% ethanol. Fish were released back to sites where they were collected once fully recovered from anaesthetic. To avoid any potential bias in sampling, fish were fin clipped as they were encountered until a desired sample size was reached without regard to morphological species identification. We extracted DNA using the Wizard Genomic DNA Purification Kit (Promega Corp., Madison, WI, USA) following manufacturer's instructions.

Figure 1.

Map of Vancouver Island, British Columbia, Canada showing stream locations where cutthroat and rainbow trout sampling occurred (inset map shows primary geographical study location). Population key: (1) Howlal Creek; (2) Lukwa Creek; (3) Menzies Creek; (4) Cold Creek; (5) Morrison Creek; (6) Cowie Cougar-Smith Creek; (7) Cook Creek; (8) Friesen Creek; (9) Rockyrun Creek; (10) North Nanaimo River; (11) Millstone River; (12) Chase River; (13) Meade Creek.

Species markers

Seven PCR-based nuclear and one mitochondrial DNA (mtDNA) markers diagnostic for coastal cutthroat and rainbow trout were used in this study. Five of these nuclear loci (GH2D; GTH-β; IGF-2; Ikaros; RAG) were developed and validated as diagnostic for coastal cutthroat and rainbow trout by Baker et al. (2002). The mtDNA marker (ND3) was developed and validated as diagnostic by Docker et al. (2003). Two additional nuclear species markers based on restriction fragment length polymorphisms (RFLPs) were developed in the current study; 1) growth hormone 1 intron D, ‘GH1D’ - primers 5′-CAGCCTAATGGTCAGAAACG-3′ and 5′-CTTATGCATGTCCTTCTTGAA-3′ cut with Mbo I (see Docker & Heath, 2003) and 2) transferrin exons 3–5, ‘Tfex3–5’– primers 5′-GCCTCCACAACTACAACCTGCA-3′ and 5′-TGGAAGGCCCCGGAATAGTCAT-3′ cut with NciI (see Ford, 1999). We initially amplified the two DNA fragments (GH1D: 1375 bp; TFex3–5: 1634 bp) by PCR in five known coastal cutthroat and five rainbow trout, and sequenced the fragments using the DCTS QuickStart cycling sequencing kit and the CEQ 8000 Automated DNA Sequencer (Beckman Coulter Inc., Fullerton, CA, USA). DNA sequence data were aligned using OMIGA 1.1 software (Accelrys Inc., San Diego, CA, USA) and analysed for species-specific RFLPs that would be easily discernible on an agarose gel (GH1D; cutthroat – 1375 bp, rainbow – 985 and 390 bp; TFex3–5; cutthroat – 717, 487 and 430 bp, rainbow – 917 and 717 bp: note fragment size estimation is based on DNA sequence data and known RFLP sites). We validated all species-specific RFLPs and size polymorphisms (seven nuclear and one mtDNA) as diagnostic using 30 allopatric rainbow and 30 allopatric coastal cutthroat trout taken from several coastal British Columbia populations. These validation runs were in addition to the tests performed by the original authors for the published species markers.

Molecular protocols

Polymerase chain reactions (PCR) were performed using standard 25-μL reactions that contained: 10 mm Tris-HCl (pH-8.4) 50 mm KCl, 2.5 mm MgCl2, 200 μm dNTPs, 0.05 μg of each primer, 0.5 units of DNA Taq polymerase, and approximately 100 ng of genomic DNA template. The optimized thermocycler profile consisted of a ‘hot-start’ and 2-min initial denaturation (94 °C), followed by 35–40 cycles of 1-min denaturation cycle (94 °C), a 1-min annealing (annealing temperatures: Ikaros = 49 °C; ND3 = 53 °C; GH2D = 55 °C; GTH-β = 55 °C; RAG = 57 °C; GH1D = 58 °C; IGF-2 = 62 °C; TFex3–5 = 63 °C), a 1.5-min extension (72 °C), and ending with a final 5-min extension cycle (72 °C).

PCR products, size polymorphisms, and RFLPs were separated by gel electrophoresis at 80–90 V through a 1.8% agarose gel. All fragments were visualized using ethidium bromide staining and UV transillumination.

Data analysis

Study fish were genotyped as homozygous rainbow trout, homozygous cutthroat trout, or heterozygous at each of the seven nuclear loci. Fish that were identified as homozygous at all seven loci for one species were categorized as pure-type. First-generation (F1) hybrid fish were those individuals that were heterozygous at all seven loci, while backcross and higher-order hybrid fish were those individuals having a mix of homozygous and heterozygous marker loci. We have combined the backcross (F1 × pure-type cross) and subsequent higher-order hybrid categories in our analyses since the likelihood of misidentifying backcross vs. higher-order hybrid genotypes, even with seven co-dominant marker loci, is unacceptably high (Boecklen & Howard, 1997). All genotypes that could be interpreted as partial restriction digests on the agarose gel were re-amplified and digested to confirm genotype. Mitochondrial DNA haplotypes were assigned as either cutthroat or rainbow trout for all fish. Individual fish that were scored as homozygous for cutthroat or rainbow trout at all seven nuclear loci, but had the opposite species mtDNA were identified as ‘ancient’, highly introgressed hybrids. It is likely that we have misidentified some higher-order hybrids as pure-type, since with seven co-dominant species markers Boecklen & Howard (1997) estimated an approximate 12% error rate in the second backcross generation. However, our error rate is likely considerably lower than Boecklen & Howard's (1997) estimates, since we included an mtDNA species marker. Furthermore, Boecklen & Howard's (1997) model only allowed unidirectional backcross events with pure-type parental fish (i.e. no mating between backcrosses or F1 fish) – assumptions almost certainly incorrect in our case.

Weir & Cockerham's (1984) inbreeding coefficient (f or FIS) was calculated for each locus using observed and expected heterozygosity levels generated from Tools for Populations Genetic Analyses (tfpga) software, version 1.3 (M.P. Miller, Northern Arizona University). Conventional Monte Carlo exact tests (10 batches, 2000 permutations per batch) for Hardy–Weinberg equilibrium (HWE) were utilized at each locus (tfpga). A sequential Bonferroni correction, to account for multiple simultaneous tests (7 loci × 14 populations = 98 comparisons) was applied to our tests of HWE (Rice, 1989). Many locus-by-population groups were in HWE before, and all were in HWE after Bonferroni adjustments, which was unexpected given that interspecific hybridization occurring between two distinct species should violate the HWE assumptions of random mating and no selection. To further examine the HWE status of our hybridizing populations we tested for trends in the sign of FIS among the seven loci within each population using sign tests (systat, version. 7.1). This was done to determine if more heterozygote deficits or excesses were present than expected by chance. Populations were also tested for linkage (i.e. gametic) disequilibria (D′) (i.e. association of alleles at different loci via physical linkage, selection, or through nonrandom mating) utilizing arlequin, ver. 2.000 (Schneider et al., 2000). The unit measure D′ (Lewontin, 1964), which is the standardized modification of linkage disequilibrium for two biallelic loci, was calculated as the ratio of D : Dmax, where ‘D’ is the difference between observed and expected allele frequencies and ‘Dmax’ is the maximum disequilibrium possible for given allele frequencies. The value of D′ can range from –1 to +1 with positive values of D′ indicating an association between alleles at the loci being compared; we report the mean and range among loci for each population. Significance of D′ for individual trout populations was computed under the model of no linkage (i.e. D′ = 0) and were corrected for multiple simultaneous tests using a Bonferroni corrected P-value (Rice, 1989).

To establish whether the initial hybridization was unidirectional or reciprocal, we used a Pearson chi-square analysis to test for mtDNA haplotype biases in the F1 hybrid fish. Most populations did not have sufficient numbers of F1 fish to allow robust chi-square analysis; therefore, we also performed the analysis using the pooled F1 data across populations.

Associations between nuclear genotypes and mtDNA haplotypes within each hybrid population were estimated using measures of cytonuclear disequilibria (Asmussen et al., 1987; Asmussen & Basten, 1994). We calculated allelic and genotypic disequilibria (DCc, DCCc, DRRc), which reflect departures from the expectation of random association between nuclear and mitochondrial species marker alleles and genotypes:


where C and R are coastal cutthroat and the rainbow trout nuclear alleles, respectively, and c is the mtDNA haplotype of the cutthroat trout. Positive DCc indicates nonrandom associations between the nuclear and mitochondrial species marker alleles (i.e. cutthroat nuclear alleles associated with cutthroat mtDNA alleles more often than expected based on random association). When DCCc is positive and DRRc is negative, the cutthroat (CC) genotypes carry the cutthroat (c) mtDNA haplotype more often than would be expected by chance, indicative of assortative mating or selection against disassortative mtDNA and nuclear hybrid genotypes (Harrison & Bogdanowicz, 1997). We further investigated the relationship between mtDNA haplotype and nuclear allele frequency by testing for differences in nuclear allele frequencies between rainbow and cutthroat trout within mtDNA haplotype using a Pearson chi-square analysis.

Fish were assigned to two age categories based on the size-frequency distribution of all sampled fish (Fig. 2) and a published size-age relationship for coastal cutthroat (Rosenfeld et al., 2000). Fish smaller than 55 mm correspond to young-of-the-year (age 0+) and fish greater than 55 mm correspond to fish that have survived at least one winter. Although Rosenfeld et al. (2000) developed the size–age relationship for cutthroat trout, our application to rainbow trout is justified since juvenile fish of the two species are not different in size during the first few years of life (Pearcy et al., 1990). The fish in the two age categories were scored as either pure or hybrid trout (see Fig. 2). Campton & Utter (1985) suggested that hybrids might have a competitive disadvantage during the critical over-wintering period. Thus, we used a two-way Pearson chi-square to test for differences in the proportion of hybrid fish in the juvenile fish (<55 mm) vs. the fish that had survived at least one winter (>55 mm). If environment-independent (or intrinsic) selection was acting, we would expect the frequency of the hybrid fish to decline after the first over-wintering period consistently in all populations. We also performed the Pearson chi-square analysis on the two age classes within individual populations to test for population-specific (i.e. extrinsic) selection effects.

Figure 2.

Frequency distribution of pure-type (solid bars) and hybrid (shaded bars) cutthroat and rainbow trout pooled from 13 populations on Vancouver Island, BC. Fish less than 55 mm (left of the arrow) are young-of-the-year (i.e. fish that have not over-wintered) and fish greater than 55 mm (right of the arrow) are primarily older year-classes (i.e. fish that have survived at least one winter).


All eight markers (seven nuclear and one mtDNA) were 100% diagnostic for the allopatric cutthroat and rainbow trout (n = 30, each). A total of 236 (51%) hybrids (including F1 and higher order backcross and hybrid fish) were identified among the 13 populations. Eleven of 98 locus-by-population tests yielded significant departures from HWE before Bonferroni correction, but none deviated from HWE after Bonferroni correction (Table 1). Sign test results showed significant trends of heterozygote deficiency (i.e. FIS > 0) in Cook Creek, Lukwa Creek, Friesen Creek, Cold Creek, Morrison Creek, Chase River ’03, Millstone River and the North Nanaimo River. The remaining populations (Meade, Howlal, Rockyrun, Menzies, Cowie Cougar-Smith creeks, and Chase River ’02) did not show significant trends towards heterozygote or homozygote deficiency. Positive values of standardized linkage disequilibria (D′) were observed in all 13 populations, indicating that alleles from the same species tend to co-segregate. Significant mean D′ was observed in Meade Creek, Cook Creek, Lukwa Creek, Friesen Creek, and the Millstone River (Table 1). Furthermore, the values of D′ in these populations were high (i.e. D′ close to 1 which is total disequilibrium), an indication of recent hybridization (secondary contact). The range of D′ in all other populations included low values overall and were not significantly different from zero.

Table 1. FIS values across seven marker loci in each of 13 hybridizing sympatric populations of Vancouver Island cutthroat and rainbow trout.
  1. Significant departures from Hardy–Weinberg equilibrium (HWE) before Bonferroni adjustments are denoted by asterisks. Sign tests showed that eight populations had significant bias for positive FIS (i.e. heterozygote deficiency) across all loci (denoted by †).

  2. Mean linkage disequilibria (D′) across loci that are significantly different from zero (*P < 0.002) are in bold-type (observed range in parentheses).

Meade Cr0.−0.002−0.08−0.0020.835* (0.630–1.00)
Cook Cr†0.330.320.210.43*0.43** (0.819–1.00)
Howlal Cr0.030.130.34−0.05− (0.400–1.00)
Lukwa Cr†*0.43*0.370.260.797* (0.631–1.00)
N. NanaimoR†0.260.53*0.370.210.210.47*0.260.884 (0.492–1.00)
Rockyrun Cr0.270.360.360.21−0.09− (0.371–1.00)
Friesen Cr†0.130.51*0.350.190.240.350.370.866* (0.579–1.00)
Cold Cr†−–
Menzies Cr0.26−0.20−0.002−0.22−0.200.04−0.040.508 (0.113–1.00)
Morrison Cr† (0.111–0.840)
Millstone R†0.35*0.35*0.300.250.310.250.35*1.00* (1.00–1.00)
CC-Smith Cr0.120.50*−0.02−0.11−0.200.08−0.120.524 (0.217–0.918)
Chase R ’020.12−0.12−0.19−0.33−0.12−0.47−0.230.570 (0.190–0.921)
Chase R ’03†0.50**0.310.090.080.518 (0.172–1.00)

Twelve of thirteen populations contained at least one individual that was heterozygote at all seven nuclear markers, signifying first-generation (F1) hybrids (Table 2). Several populations displayed high frequencies of F1 hybrids (Table 2). The presence of F1 hybrids in the majority of our populations provides evidence of current, ongoing hybridization. Menzies Creek contained no F1 hybrids, suggesting pure-type (parental) fish have not interbred recently; however, the presence of a variety of backcross genotypes (Fig. 3) indicates that introgression is most likely ongoing. Chase River ’02 and ’03, as well as Cowie Cougar-Smith Creek exhibited a diverse array of recombinant genotypes and very few F1 or pure-type, suggesting that these two systems are hybrid swarms (Fig. 3).

Table 2.  Sample size (n), frequency of pure-type (ratio), frequency of hybrids (with counts in parentheses), and the numbers of F1 hybrids by mtDNA haplotype from 13 sympatric populations on Vancouver Island, BC.
PopulationnPure-type (RBT/CTT)HybridF1 genotypeLife histories present
CTT haplotypeRBT haplotypeResidentAnadromous
  1. The migratory life histories present within each population are shown (+ and − indicate life history presence or absence). The Chase River population was sampled in two successive years (2002 and 2003).

Meade Creek301/1450% (15)41++++
Cook Creek3215/441% (13)43++++
Howlal Creek291/1448% (14)30+++
Lukwa Creek314/1539% (12)50++++
N. Nanaimo River3828/221% (8)10+++
Rockyrun Creek3726/030% (11)02++
Friesen Creek337/1049% (16)61+++
Cold Creek3010/837% (11)56++++
Menzies Creek300/1357% (17)+++
Morrison Creek331/1455% (18)10++++
Millstone River359/1337% (13)110+++
CC-Smith Creek323/188% (28)10+++
Chase River ’02353/086% (30)10++++
Chase River ’03377/081% (30)10++++
Figure 3.

Frequency distribution of multilocus nuclear genotypes (7 loci) for trout from thirteen populations of sympatric cutthroat and rainbow trout on Vancouver Island, BC. Genotype is given as the number of cutthroat trout alleles at all loci combined (i.e. pure cutthroat trout = 14; pure rainbow trout = 0).

The distribution of hybrid genotypes shows that various populations exhibit a hybridization bias towards either rainbow or cutthroat pure-types, or a symmetric pattern (Fig. 3). The North Nanaimo River and Rockyrun Creek populations displayed a bias for hybrids (F1 and higher order hybrids) to mate with pure rainbow trout, since most hybrids had a majority of rainbow alleles (Fig. 3). Hybrids in Meade Creek, Howlal Creek, Friesen Creek, Menzies Creek, and Morrison Creek, exhibited a mating bias towards pure cutthroat trout (Fig. 3). Chase River ’02 and ’03, Lukwa Creek, and Cowie Cougar-Smith Creek displayed a reciprocal (i.e. bi-directional) pattern of hybridization. Significant allelic disequilbria (DCc) were observed in all but one population (Table 3). In some populations (Morrison Creek, Menzies Creek, Meade Creek and Howlal Creek) the mating bias favours the most common parental species, although in other populations this is not the case (Fig. 3). The F1 hybrids combined had a marked bias for the cutthroat mtDNA haplotype, with 77% of the F1 fish having the cutthroat haplotype (inline image = 16.1; P < 0.0001). However, this bias is primarily driven by a few populations with strong interbreeding biases, while some populations displayed weak or nonexistent differences in mtDNA haplotypes (Table 2).

Table 3.  Allelic (DCc) and genotypic-cytonuclear (DCCc, DRRc) disequilibria for 13 sympatric hybridizing trout populations on Vancouver Island, BC.
  1. Values for cytonuclear disequilibria have been averaged over seven nuclear marker loci in each population. Significance values for allelic and cytonuclear disequilibria are given (*P < 0.05; **P < 0.01; ***P < 0.001; ns = P > 0.05).

Meade Cr0.04*0.04ns−0.03ns
Cook Cr0.14***0.12***−0.16***
Howlal Cr0.03*0.02ns0.03*
Lukwa Cr0.11***0.10**−0.12***
N. Nanaimo R0.09***0.04*−0.14***
Rockyrun Cr0.20*0.01ns−0.02ns
Friesen Cr0.14***0.11**−0.16***
Cold Cr0.16***0.15***−0.16***
Menzies Cr0.02ns0.05ns−0.02ns
Morrison Cr0.03*0.02ns−0.03*
Millstone R0.15***0.10**−0.19***
CC-Smith Cr0.09***0.07*−0.10*
Chase R ’020.08***0.04ns−0.12**
Chase R ’030.13***0.10***−0.17***

Genotypic cytonuclear disequilibria (i.e. DCCc and DRRc) revealed significantly positive associations between genotype and haplotype (i.e. positive DCCc with negative DRRc– cutthroat genotype with cutthroat haplotype and rainbow genotype with rainbow haplotype) in eight of 13 populations (Table 3). One population, Howlal Creek, revealed a significant negative association between genotype and cytotype (i.e. positive DRRc– cutthroat genotype with rainbow haplotype; Table 3). The remaining four populations displayed nonsignificant positive associations between genotype and haplotype (Table 3).

We observed highly significant differences in the frequency of nuclear alleles within mtDNA haplotype groups of backcross and higher-order hybrids in most populations (Fig. 4). These differences consistently reflected positive associations between the mtDNA haplotype and the frequency of the nuclear species alleles (Fig. 4). Some populations showed this effect in both mtDNA haplotype groups (Fig. 4; Cook Creek, Rockyrun Creek, Menzies Creek, Cowie Cougar-Smith Creek, and Chase River ’03). Since we only included backcross and higher-order hybrid individuals in this analysis, it is not biased by the frequency of pure-type trout (as are cytonuclear disequilibrium calculations), and hence the analysis represents a powerful test for cytonuclear disequilibrium in interspecifically hybridizing populations.

Figure 4.

Nuclear allele frequencies in backcross hybrid trout within each mtDNA haplotype (CTT, cutthroat trout; RBT, rainbow trout) in the 13 sympatric populations. Cutthroat trout (solid bars) and rainbow trout (shaded bars) nuclear allele frequencies are shown and differences between the two, within a haplotype, were tested for significance using Pearson chi-square (*P < 0.05; **P < 0.01; ***P < 0.001; NS = P > 0.05).

There was no significant difference in the hybrid incidence between young-of-the-year and over-wintered fish in the combined data set indicating intrinsic selection acting in the first year of life is absent or very weak [Fig. 2; inline image = 0.433; P > 0.50]. However, there were significant differences in the hybrid frequency between the two age classes in three of the 13 populations: Chase River ’02 had a lower frequency in young age class [inline image = 6.562; P < 0.05], while higher frequencies of hybrids were found in the young age class from the North Nanaimo River [inline image = 7.799; P < 0.01] and Rockyrun Creek [inline image = 7.8; P < 0.01].


In hybrid zones resulting from secondary contact, linkage disequilibrium is expected to be initially high, but generally erodes rapidly over time (Hartl & Clark, 1989; Harrison & Bogdanowicz, 1997). Nevertheless, selection against hybrids, assortative mating, and continued dispersal of parental pure-types into a hybrid zone can maintain disequilibrium over multiple generations (Harrison & Bogdanowicz, 1997; Jiggins & Mallet, 2000). Our results indicate that several sympatric cutthroat and rainbow trout populations on Vancouver Island display strong linkage disequilibrium, generally indicative of recent or ongoing secondary contact. Additionally, the presence of F1 hybrids in these populations further implicates recent secondary contact in these populations. However, many populations also displayed high frequency of pure-type fish as well as hybrids, suggesting that the disequilibria may be because of persistent dispersal and ongoing mating between pure-types and hybrids. Although we found no evidence for potential sources of dispersing pure-type fish, it is possible that pure-type populations may exist elsewhere within the study watersheds. In the populations where we observed nonsignificant linkage disequilibrium we also noted a relatively even frequency distribution of the hybrid genotypes. This is in contrast to the more commonly observed ‘bimodal’ frequency distribution of hybrid genotypes where the hybrid genotypes tend to cluster with the pure-type genotypes (e.g. Cruzan & Arnold, 1994– plants; Jiggins et al., 1997– invertebrates; Redenbach & Taylor, 2003– fish). Our observed divergent hybrid genotype frequency distributions, coupled with the variation in linkage disequilibrium is likely a consequence of geographic differences in the magnitude of pre- and post-zygotic reproductive barriers to gene exchange between the species (Harrison & Bogdanowicz, 1997).

Size differences between mature adults of sympatric species pairs have been hypothesized to influence the direction of initial hybridization events (i.e. unidirectional or reciprocal; Wirtz, 1999). Our results show that the initial hybridization events (i.e. the production of F1 hybrids) occur in both directions (reciprocally), but with a bias for hybridization to occur between a female cutthroat trout mating with male rainbow trout (see Table 2). Grant & Grant (1997) reasoned that the female of smaller species might accept males of larger species, but not vice-versa because the smaller males transmit subnormal reproductive stimuli. For example, female Xiphophorus pygmaeus preferred to mate with the larger male X. nigrensis, even in the presence of smaller conspecific males (Ryan & Wagner, 1987). Although the two swordtail species were not naturally sympatric, the authors hypothesized that if they were to become sympatric, preference of female X. pygmaeus for X. nigrensis males could result in extensive introgression, and possible convergence of the species (Ryan & Wagner, 1987). In our case, the direction of the initial hybridization between cutthroat and rainbow trout may be also biased by relative body size since the anadromous rainbow trout (i.e. steelhead) commonly spend 2–3 years in the ocean and attain a much larger body size than anadromous or freshwater resident (i.e. nonanadromous) coastal cutthroat trout (Pearcy et al., 1990). Thus, initial hybridization events would be predicted to favour female cutthroat trout choosing male steelhead trout due to body size (i.e. sexual selection); however, the presence or absence of anadromous steelhead does not appear to be associated with either higher levels of hybridization, or greater F1 mtDNA haplotype bias (see Table 2). Ostberg et al. (2004) found that in two coastal hybridizing trout populations, F1 hybrids arose through matings between female steelhead trout and male cutthroat trout, indicative of possible ‘sneak’ spawning behaviour by male cutthroat trout. Our results do not support such a mechanism driving hybrid-mating bias on Vancouver Island (except possibly in Rockyrun Creek). The discrepancy between the two studies highlights the variable nature of mating biases in inter-specific hybridization, and clearly indicates that multi-population analyses are important for characterizing hybridization mating behaviour. We suspect that the direction of initial interspecific mating likely depends upon environmental factors as well as the life histories of the hybridizing populations.

Alternatively, initial hybridization events could simply be due to a greater abundance of one species, with female mate choice or male mating behaviour (sexual selection) playing little or no significant role. Several of our populations displayed a much higher abundance of one pure-type species relative to the other (Fig. 3, Table 2). However, a comparison of pure-type abundance with the frequencies of F1 mitochondrial haplotype shows no consistent pattern – for example, Millstone River had the most pronounced F1 mtDNA haplotype bias, but the pure-type frequencies were not particularly divergent (Table 2). Avise & Saunders (1984) identified 14 hybrid sunfish (Lepomis spp.) produced by matings between common and rare species of Lepomis; and there was a tendency for the rare species parent to be female. Additionally, Avise et al. (1997) analysed hybridizing bass (Micropterus punctulatus and M. dolomieui) and found that six of seven probable F1 hybrids carried the mtDNA of M. dolomieui, the rarer species. Our results are not consistent with parental species abundance driving the bias in the F1 haplotype frequencies, which provides further evidence that hybridization patterns may be species-specific, and dependent on reproductive life history strategies. Dowling et al. (1989) detected all F1 hybrids of Notropis cornutus and N. chrysocephalus (Family Cyprinidae) from Raisin River to contain the N. chrysocephalus mtDNA, while approximately 90% of the F1 hybrids from the Kalamazoo River had the N. cornutus mtDNA. These data are consistent with ours, where the initial hybridization may appear uni-directional in one population, but when additional populations are examined, hybridization is found to be reciprocal across populations. Furthermore, in some of our study populations, hybridization is reciprocal within a single population (e.g. Friesen Creek, Meade Creek, Cook Creek, and Cold Creek). Thus, it is important to screen multiple populations to detect hybridization directionality, given the possibility of extrinsic or population-specific effects.

Our analyses of cytonuclear disequilbrium and post-F1 backcross and higher-order hybrid nuclear-mitochondrial allele associations demonstrate a remarkable reproductive bias, where hybrids have disproportionately more nuclear alleles that match their mtDNA haplotype species. That is, subsequent to the initial hybridizing event that produces F1 offspring, the direction of backcross mating is not random, but is directed towards the species that matches the specific mtDNA type of the hybrid offspring. What could be driving such a mating direction bias? Two possible explanations include: (1) random mating is occurring relative to our marker loci, but selection is acting against hybrids with mismatched mitochondrial-nuclear marker alleles; or (2) the hybrid fish have a behavioural bias which is tied to their mtDNA haplotype or maternal lineage that drives a mating preference. The first possibility is unlikely given that we found little evidence for selection against hybrids, but cannot be conclusively rejected. However, a behavioural mating preference may exist: mtDNA haplotype can affect metabolic rate (Doiron et al., 2002), and hence possibly habitat choice. Since rainbow and cutthroat trout have different habitat preferences (Young et al., 2001; Ostberg et al., 2004), a habitat choice bias could indirectly drive an apparent mating preference bias. Although the requirement for a mating preference bias in the hybrid males could be weak, hybrid females would have to exhibit a strong mating preference for individuals of the species that match their mtDNA haplotype to produce offspring that are consistent with our data. The fact that all populations in this study exhibited some level of linkage disequilibrium (D′) suggests that assortative mating is a plausible explanation, given that assortative mating is also believed to maintain linkage in hybrid populations (Harrison & Bogdanowicz, 1997). However, the mechanism that could give rise to such reproductive behaviour in female hybrids is not obvious, and no other study of hybridization has reported such a phenomenon. It is possible that F1 cutthroat-rainbow trout hybrids have skewed sex ratios [e.g. heterogametic F1 hybrids (males) are partially sterile or inviable, as expected under Haldane's rule]. Turner & Liu (1977) observed a consistent excess of females in F1 progeny among species of killifish (genus Cyprinodon). However, no sex-ratio skew has been reported for hybrid rainbow and cutthroat trout, and furthermore, the production of all female (or male) F1 offspring would not explain our observed cytonuclear disequilibrium, without some form of selection or mating behaviour driving the apparent mating bias in backcross and higher-order hybrids.

Sympatric species pairs are believed to exhibit stronger reproductive isolation than allopatric species pairs with similar levels of genetic divergence (e.g. Coyne & Orr, 1989; Butlin, 1995). This has been attributed to natural selection against hybrid offspring, which drives reinforcement mechanisms in response to hybridization events (Noor, 1999; Servedio, 2000). Our data shows no evidence for strong selection (intrinsic or extrinsic) currently acting against hybrid trout on Vancouver Island, despite a long history of sympatry between these two species. Our size-frequency data displayed no evidence for a consistent reduction in the number of hybrid (F1 and/or higher-order backcross) fish after the first winter. Hawkins & Foote (1998) established that there was no evidence of reduced hatchability or viability of coastal cutthroat-rainbow/steelhead trout F1 hybrids under laboratory conditions (i.e. intrinsic selection), while Young et al. (2001) and Docker et al. (2003) documented relatively high incidence of juvenile F1 hybrids in sympatric populations of coastal trout, also indicative of weak or absent prezygotic barriers. However, Campton & Utter (1985) suggested that F1 cutthroat-rainbow/steelhead trout hybrids face a selective disadvantage during saltwater migration because of intermediate life history characteristics, while Hawkins & Quinn (1996) found that F1 hybrids were intermediate to the pure-type species in both swimming performance and morphology, thus generating the potential for a competitive disadvantage in the hybrids. It appears from our data that the F1 hybrids have not been strongly selected against by extrinsic or intrinsic effects, despite known consequences from such selective effects in other hybrid fish species (e.g. F1 hybrid sticklebacks – see Vamosi et al., 2000) and the expectation in other sympatric species pairs (Edmands, 2002). Our test for selection was limited by two factors: (1) we examined survival differences at only one stage of cutthroat and rainbow trout life history; and (2) we did not directly test for temporal variation in the intensity of selection (with the exception of Chase River). Although the first winter represents a critical period in the development of juvenile trout, selection acting against hybrids later in life, and variation in selection intensity over time, have been reported elsewhere (e.g. Vamosi & Schluter, 1999; Grant & Grant, 2002).

Although we found no evidence for overall hybrid inferiority, we did find reduced frequencies of hybrid fish after the first winter in two populations – a pattern consistent with extrinsic, or population-specific, effects. It is generally difficult to determine whether fitness in hybrids is affected by intrinsic selection, extrinsic selection, or both. Allendorf et al. (2001) hypothesized that outbreeding depression stems purely from extrinsic selection effects. Additionally, Edmands & Timmerman (2003) suggested that disruption of local adaptation (extrinsic selection) was more severe than disruption of co-adapted gene complexes (intrinsic selection). Lu & Bernatchez (1998) performed controlled hybridization matings between dwarf and normal lake whitefish (Coregonus clupeaformis), and found that while fertilization success did not differ, embryonic mortality was higher in the hybrid crosses, relative to pure-type crosses. Subsequent work on introgression between sympatric dwarf and normal lake whitefish indicated a high likelihood of extrinsic factors driving reproductive isolation of the two ecotypes (Lu et al., 2001). In our system, extrinsic selection acting in selected populations would also provide some explanation for the highly variable hybridization levels and diverse hybrid genotype frequency distributions observed among the 13 sampled populations. Other contributing factors to the variation in hybridization levels may include differences in watershed/ecosystem parameters as well as varied anthropogenic impacts. In a regression analysis using several environmental variables, including anthropogenic impact parameters and watershed attributes, and hybridization data from 30 sympatric trout populations on Vancouver Island (including the 13 reported here), several environmental factors (e.g. extent of timber harvesting, history of hatchery stocking, watershed size, available habitat) were significantly correlated with hybridization levels (Bettles, 2004). Additionally, Docker et al. (2003) reported a significant association between hybridization and the history of stocking of rainbow trout in naturally sympatric coastal cutthroat-rainbow trout populations. Clearly hybridization between these two trout species is affected by anthropogenic disturbances interacting with ecosystem-specific conditions.

In Chase River and Cowie Cougar-Smith Creek, we observed very high hybrid frequencies and a diverse array of backcrossed and higher-order hybrid genotypes, indicative of hybrid swarms. The strength of fitness gradients among pure-type and hybrid genotypes can greatly influence the development of hybrid swarms, and it has been postulated that even the narrowest margin of increased fitness in later generation hybrids can lead to the establishment of a hybrid swarm (Epifanio & Phillip, 2001). The hybrid swarms in Chase River and Cowie Cougar-Smith Creek indicate fitness among the hybrids in these populations is at least equal to pure-type fish. Interestingly, Chase River ’02 showed a significant increase in the frequency of hybrid genotypes in the older, over-wintered, age classes, an outcome consistent with a competitive advantage for the hybrids. The abundance of backcross and higher-order hybrids, relative to the low frequency of F1 and pure-types, in these systems demonstrate that the hybrid swarm is not a transient phenomenon. Our data for Chase River (’02 and ’03) are consistent with results from 2000 [Docker et al. (2003): Chase River (2000) hybrid frequency = 92%], which demonstrates considerable temporal persistence of the hybrid swarm over time. Thus, in Chase River and Cowie Cougar-Smith Creek the two species have undergone a permanent loss of reproductive isolation. Despite the continued high incidence of hybrids in Chase River, the hybridization dynamics appear to be temporally unstable, since the cytonuclear and Hardy–Weinberg equilibrium status of the Chase River population changed substantially between the 2002 and 2003 sampling dates. Although sampling error may have contributed to the observed differences, the magnitude of those differences makes it unlikely that sampling error alone is responsible. Furthermore, the allele frequency distribution did not change between 2002 and 2003 [inline image = 11.6; P < 0.639; TFPGA exact test of allele frequency distribution differences, 10 batches, 2000 permutations per batch]; although the frequency distribution of hybrid genotypes was more even in 2002 and 2003 relative to that reported for 2000 (Docker et al., 2003; 38% F1 hybrids). Rubidge et al. (2001) also noted a lack of temporal stability in hybridization between allopatric westslope cutthroat trout and introduced rainbow trout in southeast BC, where the incidence and distribution of hybrid fish increased over approximately 15 years. Most populations in our study did not exhibit the characteristics of hybrid swarms, although many did have high hybridization levels. Since the majority of the sampled populations displayed no detectable hybrid inferiority (although small numbers of F1 hybrids in some populations limited our power to detect such effects), and given the evidence of recent and continued hybridization, many of the populations are at high risk for continued hybridization and introgression, and we predict they will also likely develop into hybrid swarms.

Reproductive isolation between the sympatric Vancouver Island cutthroat and rainbow trout has recently been compromised, resulting in very high levels of hybridization. Contemporary loss of reproductive isolation has three possible long-term outcomes: (1) equilibration of selection against hybrids vs. the formation of new hybrids, to create a stable hybrid zone; (2) the evolution of new reproductive isolation mechanisms (or the recovery of pre-existing ones), with subsequent re-emergence of reproductive barriers between the two species; or (3) complete genetic introgression, with the loss of the parental species, but with the potential for the emergence of a new species as a result of the hybridization (see Seehausen, 2004). Our evidence for weak and inconsistent selection against cutthroat-rainbow trout hybrids indicates that this system is not likely to evolve into a stable hybrid zone, although that outcome cannot be entirely discounted. Our evidence of cytonuclear disequilbrium and apparent backcross mating biases allows the possibility that some introgressed populations could return to two sympatric, reproductively isolated species, if the initial hybridization events were limited in extent. However, the two hybrid swarm populations indicate that the complete breakdown of reproductive isolation is a possible outcome, and perhaps is the most likely one, given the high levels of hybridization seen in most of the study populations. The loss of parental species via introgressive hybridization has been previously postulated (Rhymer & Simberloff, 1996), and our sympatric cutthroat-rainbow trout system represents a remarkable example of ‘extinction by introgression’. On the other hand, the sympatric hybridizing coastal cutthroat-rainbow trout on Vancouver Island may represent a valuable opportunity to observe ongoing ‘speciation by hybridization’ (Seehausen, 2004), given our evidence for weak reproductive isolation, and weak selection against hybrids, backcrosses and higher-order hybrids. Unfortunately such a scenario would contribute to the conservation concerns for coastal cutthroat and rainbow trout, especially since other subspecies of cutthroat trout have been driven to extinction and extirpation because of introgressive hybridization (Leary et al., 1984; Carmichael et al., 1993).


We thank G. Reid, L. Carswell, and R. Hooton of the BC Ministry of Water, Land and Air Protection (MWLAP), Fish and Wildlife Science and Allocation Section for field work support. D. Hawkins, and L. Bernatchez provided valuable comments on the manuscript. Yellow Island Aquaculture Ltd provided valuable field logistical support, and research funds were provided by the BC Habitat Conservation Trust Fund, the BC Ministry of Water, Land and Air Protection-Aquatic Ecosystem Science Section, and the Natural Science and Engineering Research Council of Canada to DDH.