Among the 10 taxa of the Chiffchaff complex, there are at least three zones of secondary contact across which gene flow seems to be restricted: between abietinus and tristis in western Siberia, between caucasicus and lorenzii in the Caucasus Mts. and between collybita and brehmii in SW France/N Spain (see map in Helbig et al., 1996). A fourth contact zone is about to form in the near future when collybita, which is spreading northward in Scandinavia, will meet abietinus in central Sweden (Hansson et al. 2000). The collybita/brehmii contact zone is the first to be studied in any detail (Salomon, 1987; Salomon & Hemim, 1992; Salomon et al., 1997). It conforms to Short’s (1969) definition of a ‘zone of overlap and hybridization’, where the two ‘pure’ phenotypes predominate, but hybridization does occur. It is worth noting that brehmii represents the most divergent (i.e. presumably most ancient) lineage within the Chiffchaff complex based on mtDNA (Helbig et al., 1996). Thus collybita and brehmii are not sister taxa. The width of a hybrid zone can only be defined with reference to particular characters (Barton & Gale, 1993), changes of which may vary both in steepness and geographical location (Parson et al., 1993). In the case of the contact zone between collybita and brehmii, the change in morphological characters and in mtDNA coincides geographically along a transect of 20 km. Whether such a zone is ‘narrow’ can be evaluated in relation to dispersal distances (Barton & Gale, 1993). To our knowledge, root mean square dispersal distances (σ) in this or other Chiffchaff populations are not available, but are likely several tens of km in migratory warblers (Paradis et al., 1998; Rohwer & Wood, 1998). According to the model in Barton & Gale (1993), the contact zone between brehmii and collybita should be 240 km wide after 10 generations of contact, assuming no selection against hybrids and a root mean square dispersal distance of 30 km. Clearly, the present contact zone of 20 km in width is relatively narrow and must be maintained by selection against mixing. In agreement with this observation, all three lines of evidence, mating patterns, mitochondrial DNA and microsatellite markers, indicate significantly reduced gene flow across the contact zone between the two chiffchaff taxa.
Although the rate of heterospecific pairs is quite high (23.7%) in the zone of sympatry, mating is still assortative. However, individuals of the two taxa did not seem to be equally selective: assuming that female mate choice determines pair composition, we found that brehmii females preferred brehmii males, whereas they avoided collybita and mixed singer males. In contrast, collybita females did not avoid brehmii males and even seemed to prefer mixed-singer males. It is unclear why collybita females are less discriminating in favour of males of their own phenotype. In this context it is of interest that mixed singers are indistinguishable from collybita, but differ from brehmii, in morphology (Salomon et al., 1997). Moreover, the allele frequencies at the analysed microsatellite loci suggest closer resemblance of mixed singers to collybita than to brehmii. Perhaps male mixed singers are more likely to be genetically pure collybita which learned the brehmii song than vice versa. This would mean that the fitness cost of choosing a mixed singer male is less for collybita females than it is for brehmii females.
Among mixed singers, there was a significant sex bias with females being under-represented (12 females, 33 males). This could reflect poorer survival or mating success of female vs. male F1-hybrids (assuming that mixed singers are mostly hybrids). However, the deviation from a 1 : 1 sex ratio may also be influenced by the way birds were classified: mixed vocalizations may be less likely to be detected in females than in males, because females produce only contact calls, whereas males also produce prolonged territorial song.
Frequency of hybrids and barriers to gene flow
The assignment test demonstrated a relatively good match between the microsatellite allele profiles of individuals and their mitochondrial haplotypes, suggesting that the majority of the birds within the contact zone consist of pure genotypes. For mixed singers and birds with a mismatch between song type and mitochondrial haplotype, a higher proportion of individuals was misassigned. Thus, these two groups appear to contain more individuals with a mismatch between their mtDNA and nuclear genome, which suggests they are F1 or backcross hybrids.
We found a significantly lower proportion of potential hybrids in the reproductive population than would be expected under the assumption that pure and mixed pairs have the same fitness. Our estimate of the number of potential F1-hybrids in the reproductive population was based on three assumptions: (1) all mixed singers are indeed hybrids; (2) all birds with a mismatch between acoustic phenotype and mitochondrial haplotype are hybrids; (3) young males copy the song type of their father rather than the song type predominant in the population. The first two assumptions are conservative, the third may lead to an underestimation of the hybrid frequency. Nonetheless, there does appear to be a discrepancy between the observed and expected number of hybrids, which may be because of the fact that one or several of the following fitness components are negatively affected in mixed matings: (1) reproductive output may be lower in mixed compared with pure pairs (2) survival of hybrids may be lower compared with nonhybrids (3) pairing success of hybrids may be lower than that of nonhybrids and (4) F1-hybrids may show reduced fertility. An alternative, nonexclusive explanation is that our assumption of the sympatric population being closed is not fulfilled. If a substantial fraction of the birds that are hatched in the relatively narrow sympatric zone disperse and settle as breeders well into the allopatric range of either collybita or brehmii, many hybrids will escape detection.
Our field data on reproductive output are too limited to test whether mixed pairs experience reduced reproductive success relative to pure pairs. The restricted gene flow we observed is perhaps more likely because of some post-fledging selective disadvantage of the hybrids. Apart from the observation of a lower than expected proportion of hybrids, no data are available on survival, pairing success or fecundity of hybrids.
In our opinion, the strong association (lack of mixing) between song phenotype and mitochondrial haplotype cannot be explained by lowered hybrid survival alone. Even if only 5.4% of the entire reproductive population (i.e. the proportion of ‘mismatched’ birds observed) were hybrids, there should still be considerable mitochondrial introgression. The fact that this has not occurred, although the hybrid zone is certainly not of very recent origin, indicates that the reproductive performance of female F1-hybrids must be significantly reduced compared with nonhybrid females.
The microsatellite data indicate that collybita and brehmii are much less differentiated in nuclear than in mitochondrial DNA. In fact, for three of the microsatellite loci, the two taxa seemed panmictic. Using all four microsatellite loci, the nuclear gene flow is estimated to be about five migrants per generation, i.e. 75 times higher than the mitochondrial gene flow. This estimate is still low compared with the high proportion (23.7%) of mixed pairs actually observed (considering only pairs in which both partners were pure phenotypes). If we assume that it takes 10 hybridization events to produce one successful migrant, the microsatellite data let us expect only 50 mixed pairs per year in the entire zone of symaptry (containing several thousand breeding pairs). This contrasts with our finding of 52 mixed pairs among 260 studied. The proportion of successful migrants transmitting nuclear genes from one taxon to the other is thus much lower than expected from the frequency of hybridization, indicating a significant hybrid fitness reduction.
In hybridization, nuclear genes can be transmitted from one taxon to the other via both males and females, whereas mitochondrial gene transfer requires reproductively viable female hybrids. Thus, the discrepancy between nuclear and mitochondrial gene flow indicates that male hybrids either survive better or have higher fecundity than female hybrids. The latter would be in agreement with predictions from Haldane’s rule (Haldane, 1922). Field data are needed to measure the relative fitness of hybrids directly, but this would require a long-term ringing study which has not yet been conducted in the contact zone between these Phylloscopus taxa.
The lower heterozygosity of brehmii compared with collybita indicates a smaller effective population size of the former or perhaps a relatively recent bottleneck. The first prediction is consistent with available information on current population sizes: collybita has a very large range and is one of the commonest insectivorous passerines in Western Europe (Tiainen & Wesolowski, 1997). Phylloscopus brehmii, on the other hand, is restricted to the western and southern parts of the Iberian Peninsula and a small area in NW Africa. Within its range it is far less common than is collybita in Western Europe (Purroy Iraizoz, 1997).
The role of pre- and post-mating isolation in avian hybrid zones
Our results suggest that both pre-zygotic and post-zygotic isolating mechanisms are contributing to the maintenance of the sharp hybrid zone between collybita and brehmii. Similar cases of narrow hybrid zones have been described for pairs of taxa in both Europe (Becker, 1995; Sætre et al., 1997a; Faivre et al., 1999) and North America (e.g. Moore & Price, 1993). Hybrids between well differentiated species might suffer reduced fitness resulting from genetic incompatibility (Coyne & Orr, 1998), Haldane’s rule being a special case in which a fitness reduction occurs mainly in the heterogametic sex. The much higher gene flow we estimated for nuclear vs. mitochondrial DNA, suggests that reduced fitness of female hybrids is contributing to the gene flow barrier between collybita and brehmii. However, well defined contact zones also occur between taxa less differentiated in mt DNA than in the chiffchaff case (Saino & Villa, 1992; Rohwer & Wood, 1998; Bensch et al., 1999). It is unclear whether some degree of genetic incompatibility is a necessary prerequisite for the formation of hybrid zones and whether there is a minimum level of overall genetic divergence that coincides with incompatibility (Coyne & Orr, 1998). For example, the Willow Warbler subspecies Phylloscopus trochilus trochilus and P. t. acredula are identical in mt DNA and show an intergradation zone of 300 km in width across Scandinavia (Bensch et al., 1999). This zone coincides with a migratory divide (Chamberlain et al. 2000) and is apparently maintained by selection against hybrids on their migratory direction (Bensch et al., 1999). The contact zone between Carrion Corvus corone corone and Hooded Crows C. c. cornix in Central Europe is maintained both by hybrid fitness reduction (Saino & Villa, 1992) and assortative mating (Risch & Andersen, 1998). In Hermit and Townsend’s Warblers (Dendroica occidentalis and D. townsendi), with a mitochondrial DNA sequence distance of less than 1% (Lovette et al., 1999), the width of the contact zone is presumably maintained by a balance between dispersal and selection against hybrids (Rohwer & Wood, 1998). There is evidence that this contact zone has been moving, with dominant Townsend’s Warblers excluding Hermit Warblers from much of their former range (Rohwer & Wood, 1998).
Further work is required to better understand whether avian hybrid zones are primarily maintained by pre-zygotic or post-zygotic isolating mechanisms and to what extent the latter consist of genetic incompatibility or natural selection on some ecological or behavioural trait that may confer hybrid disadvantage. Apart from estimates of mitochondrial and nuclear gene flow as we provided them here, a more direct assessment of hybrid fitness would be desirable. Ideally, this should be based on measures of life-time reproductive success in hybrids vs. parental phenotypes. However, in relatively long-lived birds, obtaining such data requires a considerable effort.