Amount of introgression in flycatcher hybrid zones reflects regional differences in pre and post-zygotic barriers to gene exchange

Authors

  • T. BORGE,

    1. Zoological Museum, Natural History Museums and Botanical Garden, University of Oslo, Blindern, Oslo, Norway
    2. Department of Evolutionary Biology, Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden
    Search for more papers by this author
  • K. LINDROOS,

    1. Department of Medical Sciences, Molecular Medicine, Uppsala University, University Hospital, Uppsala, Sweden
    Search for more papers by this author
  • P. NÁDVORNÍK,

    1. Department of Cell Biology and Genetics, Palacký University, Olomouc, Czech Republic
    2. Laboratory of Ornithology, Palacký University, Olomouc, Czech Republic
    Search for more papers by this author
  • A. -C. SYVÄNEN,

    1. Department of Medical Sciences, Molecular Medicine, Uppsala University, University Hospital, Uppsala, Sweden
    Search for more papers by this author
  • G. -P. SæTRE

    1. Department of Evolutionary Biology, Evolutionary Biology Centre, Uppsala University, Uppsala, Sweden
    2. Department of Biology, Centre for Ecological and Evolutionary Synthesis, University of Oslo, Oslo, Norway
    Search for more papers by this author

Glenn-Peter Sætre, Department of Biology, Centre for Ecological and Evolutionary Synthesis, University of Oslo, P. O. Box 1066 Blindern, N-0316 Oslo, Norway.
Tel.: +47 22857291; fax: +47 22854605;
e-mail: g.p.satre@bio.uio.no

Abstract

Introgression is the incorporation of alleles from one species or semispecies into the gene pool of another through hybridization and backcrossing. The rate at which this occurs depends on the frequency of hybridization and the fitness of hybrids and backcrosses compared to ‘pure’ individuals. The collared flycatcher (Ficedula albicollis) and the pied flycatcher (F. hypoleuca) co-exist and hybridize at low to moderate frequencies in a clinal hybrid zone in Central Europe and on the islands of Gotland and Öland off the Swedish east coast. Data on hatching success suggest that hybrids are less fertile in Central Europe compared to on the islands. Direct fitness estimates using molecular markers to infer paternity are consistent with the demographic data. Applying a tag-array-based minisequencing assay to genotype interspecific substitutions and single nucleotide polymorphisms we demonstrate that the amount of introgression from the pied to the collared flycatcher is higher in the two island populations (Gotland and Öland) than in two geographically distinct areas from the Central European hybrid zone (Czech Republic and Hungary). In all areas the amount of introgression from collared to pied flycatchers is very low or seemingly absent. The different patterns of introgression are consistent with regional differences in rates of hybridization and fitness of hybrids. We suggest that barriers to gene exchange may have been partly broken down on the islands due to asymmetric gene flow from allopatry. Alternatively, or in addition, more pronounced reinforcement of prezygotic isolation in Central Europe might have increased post-zygotic isolation through hitchhiking, since genes affecting pre and post-zygotic isolation are both sex-linked in these birds. One of our genetic markers appears to introgress from pied to collared flycatchers at a much higher rate than the other markers. We discuss the possibility that the introgressed marker may be linked to a gene which is under positive selection in the novel genetic background.

Introduction

Hybrid zones are arenas where evolutionary processes of fundamental importance can be studied, including those leading to the development of reproductive isolation (Barton & Hewitt, 1985; Howard, 1986; Harrison, 1993a; Arnold, 1997; Noor, 1999; Kirkpatrick & Ravigné, 2002). To some extent hybrid zone dynamics can be described as a battle between fusion and fission of partially incompatible genomes (see e.g. Harrison, 1993b). When hybrids are at least partially fertile they may act as bridges for gene exchange between differentiated genomes, a process known as introgression (Anderson & Hubricht, 1938). Predicting the outcome of hybrid zone dynamics is difficult because many conflicting forces are operating simultaneously. For instance, at some loci or combination of loci there may be incompatibilities between the differentiated genomes (e.g. Fishman & Willis, 2001; Orr & Irving, 2001) that would select for reduced hybridization, a process termed reinforcement of prezygotic isolation (Dobzhansky, 1940; Felsenstein, 1981; Liou & Price, 1994; Noor, 1999; Servedio, 2000; Servedio & Noor, 2003). On the other hand, certain loci may perform well in the novel genetic background (Anderson & Stebbins, 1954). Accordingly, it has been suggested that introgressive hybridization could be an important source of novel genetic variation despite the fact that hybrids per se are selected against (Rieseberg & Wendel, 1993; Arnold, 1997; Barton, 2001). In hybrid zones the selection potential against hybridization may also change during the course of time. First, extensive hybridization and backcrossing may result in a ‘hybrid swarm’– a high proportion of individuals of mixed ancestry that, through recombination and/or the evolution of modifiers, may produce more fit offspring than F1-hybrids between pure parental types, facilitating further introgression and fusion of the genomes (Barton & Hewitt, 1985; Butlin, 1987; Kelly & Noor, 1996). Second, it has been suggested that selection for increased prezygotic isolation (reinforcement) can increase post-zygotic isolation through hitchhiking of incompatibility alleles with the prezygotic isolation alleles under selection (Servedio & Sætre, 2003). Thus, in theory hybrid fitness may increase or decrease during the course of time and this may ultimately affect introgression rates and the long-term fate of the hybrid zone and the hybridizing taxa.

The collared flycatcher (Ficedula albicollis) and the pied flycatcher (F. hypoleuca) are two closely related old-world flycatchers inhabiting mainly deciduous forests throughout their breeding distributions in Europe. Phylogeographic reconstructions suggest that the two bird species were isolated in southern refugia during the Pleistocene glaciations and that they came into secondary contact in Central and Eastern Europe following the northwards expansion of deciduous forests after the last glaciation period that ended approximately 10 000 years ago (Sætre et al., 2001). Only recently, about 150 years ago, did they also meet on the isolated islands of Gotland and Öland off the Swedish east coast (Lundberg & Alatalo, 1992). Hybridization occurs where the two species coexist, but is limited to a rather narrow zone of range overlap that corresponds to topography and habitat (Lundberg & Alatalo, 1992; Sætre et al., 1999). Throughout, we define sympatry as areas where both species are found breeding in the same habitat on a regular basis, and allopatry as areas where only one of the species breeds.

Mate choice in these birds has been studied in detail. Males display at their territory/nest hole by song and females sample a limited number of males (typically 3–10) and choose one of these according to criteria such as species identity, territory quality, plumage color and song characteristics (Lundberg & Alatalo, 1992; Sætre et al., 1994, 1997b, Lampe & Sætre, 1995; Dale & Slagsvold, 1996; Haavie et al., 2004). Apparently, in both species females are choosy whereas males are not: in species-recognition experiments from a sympatric, Central European population females were found to prefer conspecifc to heterospecific males, whereas males behaved indiscriminative to con- and heterospecific females (Sætre et al., 1997a).

Hybrids are recognized by their intermediate plumage traits, song and calls. The frequency of hybridization has been estimated at 2–7% in different sympatric populations (Alatalo et al., 1982; Sætre et al., 1999; Veen et al., 2001). Hybrids have strongly reduced fitness; female hybrids are nearly always sterile (in fact, no hybrid female in which hybrid status has been confirmed by genetic data has been found to produce recruits to the breeding populations; Veen et al., 2001), whereas male hybrids have higher fitness (10–50% fitness reduction compared to pure pied and collared flycatcher males according to data on hatching success; Alatalo et al., 1990; Gelter et al., 1992; Sætre et al., 1997b, 1999; Veen et al., 2001). However, there are some interesting differences between the continental, Central European contact zone and the island hybrid zones both with respect to hybrid fertility as estimated from hatching success and adaptations to avoid hybridization. First, on the islands hybrids are more common than in Central Europe. Hybrids constitute on average 2.7% of the breeding population in Central European sympatric populations investigated but 4.3% on average on the Baltic Isles (Sætre et al., 1999). Second, hybrids appear to be more often fertile on the islands than in Central Europe according to demographic data. Hatching failure of pairs with one hybrid was found to be 45.1% in a study from the Baltic Isles (Alatalo et al., 1990), compared to 71.9% in Central Europe (Sætre et al., 1999). Finally, on the islands, a character displacement in plumage characters that helps species recognition (Sætre et al., 1997b) is much less pronounced compared to the situation in Central Europe (Sætre et al., 1999,2003). Consequently, we expect more extensive current gene flow between the species to occur on the island hybrid zones compared to the Central European populations. On the other hand, because the Central European hybrid zone is (presumably) older than the island zones, we may expect more historic introgression in the former zone.

Applying a tag-array based minisequencing assay to genotype single nucleotide substitutions or polymorphisms (Sætre et al., 2003), we here investigate whether the amount of introgression (historic or current) differs in the various hybrid zones in accordance with the data on hybridization rate and hatching success, or in accordance with the age difference of the contact zones. Furthermore, we test whether or not the different markers show concordant patterns of introgression. A previous paternity study from the Swedish population of Gotland suggests that male hybrids are fertile in this population (Gelter et al., 1992). We present a similar comparison of families from Central Europe.

Materials and methods

Samples

Male and female pied and collared flycatchers were trapped in nest-boxes or using mist-nets on breeding grounds throughout their European breeding range during April 1994 to June 2001. The study populations and respective sample sizes of collared and pied flycatchers are given in Table 1. In addition to these 696 individuals we trapped, 19 hybrid flycatchers in the sympatric populations. Four of the hybrid males from the North Czech population produced hatchlings. We determined paternity of these nestlings by microsatellite analysis of the parents and the chicks. Species status of sympatric birds was determined in the field by phenotypic characteristics (e.g. Sætre et al., 1997b) and confirmed by genotyping of species-specific sex-linked markers (see below).

Table 1.  Information on the different study populations of pied and collared flycatchers.
LocationSpeciesPopulation type (ratio collared: pied) n
Italy (Abruzzo)CollaredAllopatric20
S-Czech (Břeclav)CollaredAllopatric42
Hungary (Pilis M)CollaredSympatric (99 : 1)47
N-Czech (Jeseník M)CollaredSympatric (85 : 15)126
Sweden (Gotland)CollaredSympatric (95 : 5)128
Sweden (Öland)CollaredSympatric (70 : 30)50
Spain (Madrid)PiedAllopatric50
Norway (Oslo)PiedAllopatric49
Germany (Lingen)PiedAllopatric49
N-Czech (Jeseník M)PiedSympatric (85 : 15)80
Sweden (Gotland)PiedSympatric (95 : 5)21
Sweden (Öland)PiedSympatric (70 : 30)34

SNP and microsatellite analyses

Twenty-five microlitres blood from each bird were collected by brachial vein puncture and stored in 1 mL lysis buffer. DNA was extracted from the blood samples by a standard protocol involving proteinase K treatment and phenol-chloroform extraction (e.g. Sætre et al., 2001).

Twenty fixed nucleotide substitutions and single nucleotide polymorphisms (SNPs) were identified from both autosomal (ten markers from nine loci) and Z-linked loci (ten markers from four genes) as previously described (Sætre et al., 2001, 2003; Primmer et al., 2002). In addition to the SNP loci we used the autosomal microsatellite FhU1 (Ellegren, 1992; Primmer et al., 1996) and a mitochondrial control region fragment that harbor a species-specific insertion-deletion (Sætre & Moum, 2000). The same 22 markers were used in a previous study (Sætre et al., 2003) that involved a sub-sample of the individuals used in the present study. In that study we pooled the data from all the sympatric populations to look for overall differences in rate of introgression of sex-linked vs. autosomal markers whereas we here look for possible differences between the hybrid zones. The genetic data of the 19 hybrids mentioned above were analyzed and discussed in the previous paper (Sætre et al., 2003), and is consequently not analyzed further here. Data from the autosomal, the Z-linked and the mtDNA markers were analyzed separately. Since we had more markers than loci (see above) we used the locus as the unit of comparison in statistical test. The results were identical (Z-linked loci) or nearly so (autosomal loci) irrespective of which marker was chosen to represent the loci. For the autosomal data we present the results with the markers that yielded the lowest estimate of amount of introgression, although the difference was negligible.

To determine paternity in the four pairs with a male hybrid we genotyped both parents and all chicks (n = 20) at four microsatellite loci (FhU1, FhU2, FhU3 and FhU4; Ellegren, 1992; Primmer et al., 1996) as previously described (Veen et al., 2001). The power of these microsatellite markers to exclude conspecific paternity is about 0.96 (Veen et al., 2001), but is higher in the case of heterospecific and hybrid pairings because there are marked species-specific differentiation at these markers (Sætre et al., 2001; Veen et al., 2001).

PCR-amplification and genotyping

Eighteen different DNA fragments containing 20 different fixed substitutions or SNPs of interest were amplified on a GeneAmp PCR System 9700 (Perkin-Elmer, Upplands Väsby, Sweden) as previously described (Sætre et al., 2003). The substitutions and SNPs were genotyped with a tag-array-based minisequencing assay as detailed in Sætre et al. (2003). In short, we annealed a detection primer to the PCR-product immediately 3′of the nucleotide position to be analyzed and extended this primer with a single fluorescence labelled nucleotide complementary to the nucleotide to be detected using a DNA polymerase. The elongated primers (tags) were hybridized to complementary primers (anti-tags) that were immobilized in arrays through their 5′-ends on a glass support. The fluorescence signals were detected and quantified using a Scan Array 5000 instrument. An ‘array-of arrays’ format (Pastinen et al., 2000) that allowed simultaneous analysis of the substitutions and SNPs in 36 samples per microarray slide was used. For and updated protocol of the tag-array-based minisequencing assay (see Lovmar et al., 2003).

The microsatellites had PCR profiles as described in Haavie et al. (2000) and the mtDNA fragment as in Sætre & Moum (2000). The PCR products were electrophoresed on 6% denaturing polyacrylamide gels and then silver-stained. Samples of known allele size and genotype were used as a reference on each gel.

Statistical analysis

Previous intraspecific analyses of populations of pied and collared flycatchers have shown that both species are respectively quite homogenous genetically in the sense that there is only minor differentiation between geographically distinct populations (Haavie et al., 2000; Sætre et al., 2001). Accordingly, any deviation in species differentiation between allopatric and sympatric populations of the two species is likely to reflect recent gene flow (introgression) through fertile hybrids and backcrossing. Assignment tests were calculated applying the clustering model of Pritchard et al. (2000) using statistical inference of individual genotypes, as implemented in the software STRUCTURE. Optimizations of estimates of relative amount of pied and collared flycatcher alleles in sympatric birds were done by numerical analysis as previously described (Sætre et al., 2003). Birds were introduced for classification singularly together with the allopatric birds (i.e. the latter were used as templates for assignment), without preassigning birds to species. According to the numerical simulations the resulting assignment probabilities represent slightly conservative estimates of relative amount of pied and collared flycatcher DNA in individual birds (conservative in the sense that backcrossed individuals are estimated to have slightly less DNA derived form the least common parental genome than they actually have). Descriptive statistics and graphics of the STRUCTURE analysis results were conducted using the software SPSS.

Results

Hybrid genotypes at the sex chromosomes were only observed in F1-hybrids and a few first-generation backcrosses (n = 19; the mtDNA marker was used to determine the sex chromosome genotype of females since we had no W-linked markers and since both molecules are maternally inherited). This is in accordance with a previous study that involved a sub-sample of the individuals investigated here (Sætre et al., 2003). The conclusion from that study of virtual absence of introgression of sex-linked markers but presence of autosomal introgression thus remains valid with the present larger sample size. The aim of the present study is to compare the amount of introgression in different hybrid zones. Accordingly, data on Z-linked markers and the mtDNA marker is not analyzed or discussed further and consequently the above-mentioned 19 individuals are omitted from further analyses (see Sætre et al., 2003 for analysis and discussion of genetic data of these hybrids and backcrosses).

The STRUCTURE analysis shows that our genotype-data (ten loci) are sufficiently powerful to reliably assign allopatric pied and collared flycatchers to the correct species; deviations from unity in assignment probability never exceeded 2.5% in allopatric populations of either species (Fig. 1). Results from the sympatric collared flycatcher populations reveal a striking pattern. The Central European populations (Hungary and N-Czech) are characterized by exhibiting very low levels of introgression. With the notable exception of a few individuals that apparently have introgressed or backcrossed genotypes, the genotypes of the collared flycatchers in these populations are indistinguishable from allopatric ones. The estimated proportion of pied flycatcher alleles among Central European sympatric collared flycatchers is not significantly higher than among allopatric ones (Z = −0.61, n = 235, P = 0.54, Mann–Whitney U-test). In contrast, the Swedish island populations of Gotland and Öland are apparently significantly influenced by introgression from the pied flycatcher. A significant proportion of the individual collared flycatchers in these populations have genotypes consistent with being partly derived from sympatric heterospecifics (Fig. 1a; sympatric Swedish collared flycatchers vs. allopatric ones: Z = −5.95, n = 240, P < 0.0001, Mann–Whitney U-test). Sympatric pied flycatchers were genetically indistinguishable from allopatric ones in all populations, suggesting absence or very low amounts of introgression (Fig. 1b). The latter result is consistent with previous findings (Sætre et al., 2003).

Figure 1.

Estimates (median, interquartile range and extreme values) of the amount of pied flycatcher alleles (% relative to collared flycatcher alleles) in (a) collared and (b) pied flycatchers based on assignment probabilities calculated from ten autosomal, nuclear DNA-markers applying the software package STRUCTURE (Pritchard et al., 2000). Estimates are given for allopatric collared flycatchers (Italy and Czech-S); sympatric collared flycatchers (Hungary, Czech-N, Sweden-Gotland and Sweden-Öland); allopatric pied flycatchers (Spain, Germany and Norway); and sympatric pied flycatchers (Czech-N, Sweden-Gotland and Sweden-Öland). Sympatric populations are marked with asterisks.

The 11 nuclear bi-allelic DNA markers differed with respect to their power for distinguishing the species-origin of alleles (Tables 2 and 3). Three of the markers (FhU1, FhU4 and Alasy) possessed no allelic overlap among allopatric pied and collared flycatchers. One marker (Lamin A) had one allele fixed among allopatric collared flycatchers but was polymorphic among allopatric pied flycatchers. Two markers (Rhodopsin 1 and 2) had one allele fixed among allopatric pied flycatchers but were polymorphic among allopatric collared flycatchers. The remaining five markers possessed polymorphisms in both species (but with variable degrees of species overlap; Tables 2 and 3).

Table 2.  Allele frequencies (%) of 11 bi-allelic nuclear DNA-markers in allopatric and sympatric populations of the collared flycatcher.
LocusAllopatrySympatry
ItalyS-CzechHungaryN-CzechSweden GotlandSweden Öland
  1. The figures refer to the frequency of the least common allele in the collared flycatcher (the frequency of the alternative allele is thus 100 – the quoted frequency). The loci are ordered approximately according to decreasing species difference in allele frequencies.

FhU10000.84.36.9
FhU40000.40.43.0
Alasy003.28.519.826.0
LaminA0000.40.42.9
Fa45B10.513.420.213.314.622.0
Fh457.93.61.16.17.519.6
TGFB212.535.713.326.334.136.7
Rhod234.226.326.623.227.325.0
Rhod144.734.536.131.941.854.2
BCK47.415.916.720.823.025.0
OrDe10.511.54.112.033.030.2
Table 3.  Allele frequencies (%) of bi-allelic nuclear DNA-markers in allopatric and sympatric populations of the pied flycatcher.
LocusAllopatrySympatry
SpainNorwayGermanyN-CzechSweden GotlandSweden Öland
  1. The figures refer to the frequency of the least common allele in the collared flycatcher (the frequency of the alternative allele is thus 100 – the quoted frequency). The loci are ordered approximately according to decreasing species difference in allele frequencies.

FhU1100100100100100100
FhU410010010010010097.1
Alasy100100100100100100
LaminA93.181.674.583.897.685.3
Fa45B95.110099100100100
Fh4577.578.673.575.688.183.8
TGFB289.310089.190.010081.8
Rhod210010010010097.6100
Rhod1100100100100100100
BCK98.0100100100100100
OrDe4031.345.833.121.439.7

If we assume that introgression is the cause of the occurrence (in a sympatric bird) of an allele that does not occur among allopatric conspecifics but is found among sympatric heterospecifics, then four markers would be informative in the collared flycatcher (FhU1, FhU4, Alasy and Lamin A) and five in the pied flycatcher (FhU1, FhU4, Alasy, Rhodopsin 1 and Rhodopsin 2). The allele frequencies among sympatric birds of these markers are generally congruent with the overall genotype-based analyses above; higher frequencies of pied flycatcher derived alleles among Swedish than Central European collared flycatchers (Table 2) and little introgression from collared to pied flycatchers in all populations (but note that introgression from collared to pied flycatchers may not to be completely absent; Table 3). However, one of the markers (Alasy) stands out as having much higher frequencies of the allele of (presumably) pied flycatcher origin in all sympatric collared flycatcher populations (Table 2). Interestingly, the relative frequency differences among the sympatric collared flycatcher populations mirror those found for the other markers (i.e. higher frequencies in the Swedish than the Central-European populations), but the absolute frequencies are in each case much higher (Table 2). Although the five markers that possess allelic overlap among allopatric birds are necessarily less informative than the others, most of these markers do mirror the informative ones in that Swedish collared flycatchers have allele frequency distributions more similar to the pied flycatcher (Table 2).

The four pairs with a male hybrid tested here produced a total of 20 nestlings (6, 4, 5 and 5 nestlings, respectively). The genotype of all nestlings matched the genotype of their mother. However, only one nestling had a genotype consistent with being sired by the hybrid male parent. All the other nestlings had nonmatching alleles at some markers, showing that an extra-pair male sired them (in three of four cases apparently a male conspecific with the female, in the latter case a heterospecific male). Hybrid male paternity of one of 20 nestlings (5%) is significantly lower than the 75% reported by Gelter et al. (1992) from Sweden (binomial test, P < 0.001). Using the hybrid males as a unit of analysis, one of four hybrids siring offspring is almost significantly lower than the six of seven reported by Gelter et al. (1992) (Fisher's exact test, P = 0.088). In both studies all pairings were between a male hybrid and a female collared flycatcher.

Discussion

Regional differences in rate of introgression

The primary result of this study is that we find conspicuous regional differences in amount of introgression from pied to collared flycatchers in different sympatric populations. Among collared flycatchers in Central Europe (Hungary and N-Czech) there appears to be very little incorporation of pied flycatcher alleles, whereas the Swedish island populations (Gotland and Öland) of collared flycatchers appear strongly influenced by introgression. Below we discuss factors that could explain this pattern.

The rate of introgression of a neutral genetic marker in a sympatric population depends in part on the frequency of hybridization and in part on the fitness of the resulting hybrids and backcrosses. In Central Europe there is strong evidence for reinforcement of prezygotic isolation (Sætre et al., 1997b). It has been demonstrated experimentally that female mate choice selects for a divergence in male secondary sexual characteristics and that the resulting character displacement reduces the frequency of maladaptive hybridization (Sætre et al., 1997, 2003). Character displacement on male plumage characteristics (Alatalo et al., 1990,1994; Sætre et al., 1999,2003) and song (Haavie et al., 2004) is much less pronounced among birds on the Swedish islands of Gotland and Öland, and consequently hybridization risk would probably be higher in the latter zones. However, the frequency of hybridization is not only related to failure in species recognition but is also affected by the ratio of conspecifics to heterospecifics a female can choose among (see also Veen et al., 2001). Direct comparisons of hybridization rates of sympatric populations in Central Europe and the Swedish islands suggest that they would contribute to the differences in amount of introgression. The frequency of pairings between a hybrid male and a collared female (probably the only pairing constellation that can result in incorporation of pied flycatcher alleles into collared flycatchers since female hybrids are sterile) is approximately 1% in the Central European hybrid zone (Sætre et al., 1999), 2% on Gotland, Sweden (Veen et al., 2001; B. C. Sheldon pers. com.) and 3–4% on Öland, Sweden (Sætre, pers. obs.).

In a survey of demographic data from various sympatric populations from Central Europe, Sætre et al. (1999) documented that the hatching success of pairs involving a hybrid was significantly lower than results published from the Swedish island populations of Gotland and Öland (Alatalo et al., 1982,1990,1994; Lundberg & Alatalo, 1992). Whereas 45% of the pairs with a hybrid experienced complete hatching failure in the studies from the Swedish islands, the corresponding figure was 72% for the Central European populations (Sætre et al., 1999). The difference is mainly due to higher fitness of male hybrids of the Swedish populations; based on hatching success: the relative fitness of male hybrids compared to ‘pure’ flycatchers is 92% on Gotland, Sweden (Veen et al., 2001) but only 48% in Central European populations (Sætre et al., 1999). Accordingly, demographic data suggest that the incident of hybridization is higher and the resulting hybrids more fertile in the hybrid zones of the Swedish islands compared to Central European ones. Thus, both pre and post-zygotic barriers to gene exchange appear to be stronger in the Central European clinal hybrid zone compared to the island hybrid zone of Sweden. The differences in amount of introgression are thus in accordance with current differences in strengths of reproductive barriers, but opposite to what one would predict from the age differences of the hybrid zones.

Direct estimates of hybrid fitness

Hatching success should provide a good estimate of fertility in female hybrids since egg dumping has never been reported in flycatchers. However, fitness estimates of male hybrids based on hatching success might be too high if females mated to such males seek extra-pair copulations with other (pure) males. Unfortunately, direct estimates of hybrid fitness using paternity tests are somewhat compromised by the rarity of hybrids and by their low-hatching success. From the Central European hybrid zone (N-Czech) we have only been able to analyze such data from four hybrid males. Of the 20 nestlings in these nests only one (5%) had a genotype consistent with being sired by the hybrid male parent. The low realized fitness of the hybrids could be explained by sterility (only extra-pair males sired nestlings in three of four nests), by females seeking EPCs because they find male hybrids unattractive, and/or by conspecific sperm precedence in cases of sperm competition (Howard, 1999; Veen et al., 2001). In a similar comparison from the Swedish island hybrid zone, Gelter et al. (1992) reported higher fitness of male hybrids; the male sired offspring in six of seven nests, and 75% of the nestlings had fingerprints consistent with being sired by the male hybrid. Veen et al. (2001) found that rates of extra-pair paternity (EPP) in heterospecific pairs were similar in the Swedish island hybrid zone and in the Central European zone. Taken together, but acknowledging the limited sample sizes of direct fitness estimates, the data suggest that hybrid fertility is higher in the Swedish island hybrid zone compared to the Central European clinal zone. However, hybrid fitness in both zones is apparently lower than estimates based on hatching success alone would indicate due to high rates of EPP.

Why would pre and post-zygotic barriers be stronger in the cline than on the islands?

Barriers to gene exchange between two incipient species are not a fixed property but are subject to evolutionary change (e.g. Servedio & Noor, 2003). As discussed above, both pre and post-zygotic barriers to gene exchange appears to be stronger in the Central European clinal hybrid zone compared to the island zone. This pattern suggests that the two classes of barriers to gene exchange may co-evolve in the case of the flycatchers. At least two scenarios may account for such co-evolution of pre and post-zygotic isolation.

In theoretical treatments of hybrid zone dynamics and of the reinforcement hypothesis it has been argued that post-zygotic barriers to gene exchange can be reduced over time (e.g. Barton & Hewitt, 1985; Butlin, 1987; Virdee & Hewitt, 1994; Kelly & Noor, 1996). The argument is that hybridization will eventually lead to the development of a ‘hybrid swarm,’ that is, an array of backcrossed and recombined genotypes rather than two well defined parental types. Mating between members of such backcrossed populations may produce ‘hybrid’ offspring that are more fertile or viable than F1-hybrids of pure parental types due to breakdown of epistatic interactions between genes affecting hybrid fitness over time or through the evolution of modifiers that increase hybrid fitness (Barton & Hewitt, 1985; Virdee & Hewitt, 1994). As a consequence, the selection potential for reinforcement will be reduced in the course of time. Accordingly, a correlated breakdown of pre and post-zygotic barriers to gene exchange can be predicted from this hypothesis.

The pattern of gene flow from allopatry appears to be much more asymmetric on the islands than in Central Europe and this may render reinforcement of prezygotic isolation less likely on the islands (Sætre et al., 1999), in accordance with theoretical predictions (Servedio & Kirkpatrick, 1997). The collared flycatcher population on the islands is probably isolated from any continental population (the closest continental population being ≈600 km away), whereas there would be influx of pied flycatchers from the surrounding mainland (Lundberg & Alatalo, 1992). Data on local recruitment (i.e. breeding individuals marked as nestlings in the local area) support this suggestion, being high for the collared flycatcher but much lower for the pied flycatcher (Veen et al., 2001). If we assume that the recruitment rate of the collared flycatcher sustains the local population, the pied flycatcher population would probably only be sustainable by immigration (Veen et al., 2001). In contrast, each species is found adjacent to the contact zone on each side of the clinal hybrid zone of Central-Europe (Sætre et al., 1999). Moreover, local recruitment rates are high and similar for both species (S. Bureš pers. com.). Thus the isolated Swedish islands may receive a constant influx of pied flycatchers from allopatry that have not evolved any adaptations to avoid hybridization. Accordingly, the genomes of the collared flycatchers would receive a constant influx of pied flycatcher alleles. Mating between members of the backcrossed (and recombined) collared flycatcher population and pied flycatchers may result in offspring that are more fit than F1-hybrids of pure flycatchers, and this would relax the selection potential for reinforcement. The more symmetric pattern of gene flow in the clinal hybrid zone, on the other hand, may facilitate reinforcement and thus no breakdown of barriers to gene exchange occurs.

Servedio & Sætre (2003) have suggested a hypothesis that could also explain the regional differences in hybridization rate and hybrid fitness. According to the ‘speciation loop model,’ genes affecting post-zygotic isolation (such as hybrid sterility) can co-evolve with genes affecting prezygotic isolation (such as female mate preferences and male secondary sexual traits). The argument is that genes affecting pre and post-zygotic isolation would often be linked on the macro sex-chromosome. Thus, when reinforcement selects for genes promoting prezygotic isolation, genes affecting post-zygotic isolation would hitchhike along, thereby further increasing the selection potential for reinforcement (Servedio & Sætre, 2003). Indeed, a significant role of Z-linked genes in controlling prezygotic isolation and post-zygotic isolation has been reported in these flycatchers (Sætre et al., 2003). Thus, the conditions for the loop-effect to occur appear to be present. According to this hypothesis it would be the post-zygotic barriers in Central-Europe that have increased as a correlated effect of reinforcement, and not the barriers of the island populations that have decreased. Reinforcement and correlated effects of post-zygotic isolation would probably have had a longer time to evolve in Central Europe than on the islands since secondary contact is likely to have been established earlier in Central Europe (Lundberg & Alatalo, 1992; Sætre et al., 2001).

We consider both hypotheses as likely explanations for our results. It is also possible that both processes are operating; the hypotheses are not mutually exclusive. However, based on our data we cannot distinguish between them or determine their relative importance. Effects of gene flow and selection may be expected to leave different molecular fingerprints near the genes controlling the barriers to gene exchange. Accordingly, future studies with a higher molecular resolution may enable us to distinguish between the hypotheses.

Why is there so little introgression from the collared to the pied flycatcher?

The difference in rate of introgression between the species is quite striking. Whereas gene flow from the pied to the collared flycatcher is rather extensive (at least in the hybrid zones of the Swedish islands of Gotland and Öland), it is almost absent in the opposite direction. We think the observed difference is related to demographic factors. First, in all populations studied here the collared flycatcher greatly outnumbers the pied flycatcher. Accordingly, backcrossing events would more often involve a collared than a pied flycatcher. Indeed hybrid × collared flycatcher pairings are much more frequent than hybrid × pied flycatcher in both hybrid zones (Sætre et al., 1999; Veen et al., 2001). Second, the species difference in rate of introgression is for a large part an effect of the high rate of introgression in the Swedish collared flycatchers. Due to the above-mentioned asymmetric pattern of gene flow the locally breeding pied flycatchers on the islands may mainly be immigrants and consequently lack introgressed alleles.

Differences in introgression rates between loci – selection for an adaptive allelic state?

A final notable result reported here is the apparent discrepancy in rate of introgression from pied to collared flycatcher at the Alasy-locus compared to the other markers. Allopatric representatives of the two species appear to be fixed for different alleles at this locus. Accordingly, occurrence of a ‘pied allele’ among sympatric collared flycatchers is likely to reflect introgression. However, the estimated rate of introgression of this locus is much higher than at other markers. We cannot rule out the possibility that the high frequencies of the introgressed allele are coincidental. For instance, the allele may have reached high frequencies by random drift. A speculative, but plausible alternative hypothesis is that the marker is linked to a gene that is under positive selection in the novel genetic background. That is, that the collared flycatcher has inherited an adaptive allelic state from the pied flycatcher through introgressive hybridization and that the Alasy-marker has hitchhiked along with this selection event. The high frequency of the introgressed allele in all the sympatric populations is consistent with this hypothesis. The appreciation that introgressive hybridization can be an important source in adaptive evolution despite overall selection against hybridization has often been neglected in animal studies, although it has been considered important among hybridizing plants (reviewed by Arnold, 1997). The Alasy-marker is located to an intron of a nuclear gene (5-aminoeavulinate synthetase). Further studies are needed, however, before we can conclude on this potential example of adaptive introgression. Genotyping of other species-specific sites in the same genomic region could provide useful additional information.

Acknowledgments

We thank J. Haavie, G. Kärf, M. Lindersson, C, Pöntinen and R. Figuerora for help in the laboratory; S. Bureš, Centro Studi Ecologici Appenninici, C. Berg, L. Garamszegi, J. Haavie, M. Král, T. Lubjuhn, J. Moreno, Anna Qvärnström, B. C. Sheldon and J. Török for field assistance and blood samples and J. Haavie and S. A. Sæther for comments on the MS. Financial support was received from the Swedish Research Council, the Norwegian Research Council, O. and L. Lamms Memorial Foundation, Uddenberg-Nordingska Foundation, Wenner-Gren Foundation, K. and A. Wallenberg Foundation via the Wallenberg Consortium North (WCN) and Grant Agency of the Czech Republic.

Ancillary