Current address: Department of Neurobiology and Behavior, Mudd Hall, Cornell University, Ithaca, NY 14853
Understanding speciation depends on an accurate assessment of the reproductive barriers separating newly diverged populations. In several taxonomic groups, prezygotic barriers, especially preferences for conspecific mates, are thought to play the dominant role in speciation. However, the importance of postzygotic barriers (i.e., low fitness of hybrid offspring) may be widely underestimated. In this study, we examined how well the widely used proxy of postzygotic isolation (reproductive output of F1 hybrids) reflects the long-term fitness consequences of hybridization between two closely related species of birds. Using 40 species-specific single nucleotide polymorphism (SNP) markers, we genotyped a mixed population of collared and pied flycatchers (Ficedula albicollis and F. hypoleuca) to identify grand- and great grand-offspring from interspecific crosses to derive an accurate, multigeneration estimate of postzygotic isolation. Two independent estimates of fitness show that hybridization results in 2.4% and 2.7% of the number of descendents typical of conspecific pairing. This postzygotic isolation was considerably stronger than estimates based on F1 hybrids. Our results demonstrate that, in nature, combined selection against hybrids and backcrossed individuals may result in almost complete postzygotic isolation between two comparatively young species. If these findings are general, postzygotic barriers separating hybridizing populations may be much stronger than previously thought.
A fundamental goal of evolutionary biology is to understand how new species evolve. Given that new species, by definition, originate through the evolution of reproductive barriers between populations (Dobzhansky 1937; Mayr 1942), a central focus of this research has been to assess the types of reproductive barriers that typically evolve fastest (Coyne and Orr 2004). Reproductive barriers are varied, and may be broadly classified into two groups: those that act to prevent interspecific gametes from fusing (prezygotic barriers) and those that reduce the fitness of hybrids (postzygotic barriers). By testing the strengths of these two barriers between pairs of species with differing divergence times, the rate of evolution of each barrier can be inferred. By employing this technique, Coyne and Orr (1989, 1997) have suggested that in Drosophila, pre- and postzygotic barriers evolve at roughly equal rates and are thus equally important in the speciation process. However, subsequent data from other taxa such as darter fish (Mendelson 2003; Mendelson et al. 2007) and birds (Price and Bouvier 2002) have indicated that speciation occurs much faster than the evolution of hybrid sterility, implying a pivotal role of prezygotic barriers.
However, one limitation of this type of approach is the varying accuracy of our estimates of the two categories of barriers. On the one hand, barriers that prevent mating can be accurately assessed through observations of pairing combinations. Furthermore, it is becoming increasingly feasible to discriminate postmating prezygotic isolation from early embryo death in a range of taxa (e.g., Lorch and Servedio 2005; Mendelson et al. 2006; Birkhead et al. 2008). In contrast, postzygotic barriers are notoriously difficult to estimate. One reason for this is that hybrid fitness is determined by a complexity of intrinsic and extrinsic factors, which may vary across environments (reviewed by Coyne and Orr 2004; Price 2007). For logistic reasons, studies of postzygotic isolation are generally conducted in simple laboratory environments, and understandably tend to focus on unconditional (i.e., intrinsic) components of postzygotic isolation (e.g., sterility) (e.g., Sasa et al. 1998; Presgraves 2002; Price and Bouvier 2002; Mendelson 2003; van der Sluijs et al. 2008). By ignoring other components of hybrid fitness, however, overall postzygotic isolation is surely underestimated. Second, postzygotic barriers may be expressed over many generations of offspring. Genetic incompatibilities and maladaptive phenotypes are common in F1 hybrids due to their novel genetic combinations, which have never before been under natural selection. However, such genetic incompatibilities are also expected in later-generation hybrids (e.g., backcrosses) for the same reasons, as well as the fact that recombination may further break up coevolved gene complexes. Thus, whenever F1 sterility is incomplete, the fitness of later-generation hybrid offspring should be considered for a full assessment of gene flow. A lack of empirical data has prevented an assessment of the severity of the underestimation of postzygotic isolation imposed by a focus on F1 reproductive success.
The goal of the current study was to employ a unique, multigeneration approach to study a free-living, naturally hybridizing population of collared and pied flycatchers (Ficedula albicollis and F. hypoleuca) to assess how well standard estimates of F1 fitness reflect overall levels of postzygotic isolation. This is an example of a well-studied animal hybrid zone in which the overall levels of postzygotic isolation are currently unclear. Low levels of genetic divergence (0.05% for nDNA, 2.9% for mtDNA) indicate the two species are evolutionarily young (Sætre et al. 2001). Females of both species prefer males of their own kind as mates (Sætre et al. 1997; Sæther et al. 2007), yet occasionally hybridize (approximately 5% of pairs in the Swedish hybrid zone are heterospecific). Initial assessments of the fitness of F1 hybrids indicated that these suffer marked reductions in fertility, with F1 hybrid females being sterile (Gelter et al. 1992; Alatalo et al. 1994; Veen et al. 2001; Borge et al. 2005a; Svedin et al. 2008). However, this is complicated by the fact that female collared flycatchers appear to sometimes receive direct benefits from pairing with a heterospecific male (Veen et al. 2001; Wiley et al. 2007), resulting in the successful rearing of more offspring, even if many of these are unfit hybrids. The fitness of later-generation hybrids is unknown. Strong philopatry to natal sites (Pärt 1990), combined with the collection of large amounts of breeding data, makes this an ideal study system for taking a holistic approach to investigating postzygotic isolation under natural conditions. Furthermore, we have now identified a substantial number of species-informative single-nucleotide polymorphisms (SNPs) that allow us to correctly identify F1, first- (B1) and second-generation (B2) backcrosses (see Fig. 1 for clarification of terminology), without the need for pedigrees. Using these genetically identified hybrids, we obtained two independent estimates of postzygotic isolation between Ficedula flycatchers. The first was based on frequencies of the different hybrid generations and the second was based on reproductive data from each generation.
Materials and Methods
Collared and pied flycatchers are small, socially monogamous, hole-nesting, passerine birds that co-occur in deciduous forests in eastern and central Europe, as well as on the Swedish islands of Gotland and Öland (the latter being the location of the current study). On Öland, we annually monitor approximately 2000 nestboxes, divided between 17 locations scattered across the island (see Supporting Information for sampling locations). Each location consists of deciduous or mixed forest. On average, across all locations, collared flycatchers constitute 80% of the breeding population. Nestboxes are monitored throughout the breeding season (May–June), and all breeding birds are captured when incubating (females) or feeding nestlings (males) to collect blood samples and determine pairing patterns within the population. As nestlings or on first capture, all individuals are marked with distinctly numbered aluminum rings.
GENETIC IDENTIFICATION OF HYBRID GENOTYPES
F1 hybrids are intermediate between parental species in appearance (Sætre et al. 2003), although not all birds with intermediate characteristics are hybrids (Svedin et al. 2008). Phenotypes of later-generation hybrids are unknown. Hybrid identity was therefore verified through genetic screening. A small amount of blood was collected from the brachial vein of each bird and preserved in Queens's lysis buffer (Seutin et al. 1991) or 99% ethanol. DNA was purified through standard phenol–chloroform treatment. Each sample was screened for a species-specific mitochondrial marker and 40 species-informative SNPs.
We amplified a stretch of mtDNA containing a 32bp species-specific indel according to the protocol specified by Sætre and Moum (2000). The polymerase chain reaction (PCR) products were electrophoresed in a 2% agarose gel, fixed with ethidium bromide and the relative fragment lengths visualized in UV light, by using the Syngene Bio Imaging System (Syngene, Cambridge, U.K.).
SNPs and species-specific substitutions were identified from intronic sequence data from allopatric pied and collared flycatchers (N= 9 of each species) from a previous study (Borge et al. 2005b). Species-specific substitutions and SNPs with large frequency differences between the species were preferentially chosen. We analyzed six Z-linked markers, all from different genes (ALDOB, BRM, PRLR, PTCH, SPIN, and VLDL), and 34 autosomal markers from 23 different genes (ACLY, ACTB, ALAS1 (2), CEPU, CHC (2), CKB (2), ENO1, FAS (2), FTN, FN1 (2), GAPD, GH1 (2), LAMA, LHCGR, MPP, PDC1, RHO (2), RPL5 (2), RPL7, RPL30 (2), RPL37, TGFB2 (2), and TM (2)) (a list of markers and allele frequencies in the two species is included as Supporting Information). The SNP information has been deposited to the dbSNP database of GenBank (submitters accession numbers 115492351–115492396). For genotyping, we designed flycatcher-specific primers of approximately 100bp encompassing the variable site.
We used the MassARRAY SNP Genotyping (Sequenom) facilities of Centre for Integrative Genetics (CIGENE). The system is based on an allele-specific primer extension (Homogenous MassEXTEND [hME] reaction), where short primers are extended according to the base composition in the template sequence and then separated by mass (Storm et al. 2003). The mass differences are sufficient to allow completely automated calling of the SNPs. This system has a highly efficient assay design and development tools that generate genotype calls with a reported accuracy of about 99% (Tang et al. 1999; Jurinke et al. 2001).
Flycatcher DNA was amplified by PCR accordingly; denaturation at 95°C for 15 min and then 95°C, 56°C, and 72°C for 20 s, 30 s, and 60 s, respectively for 45 cycles and a final step with 72°C for 3 min. The PCR products were given a SAP (shrimp alkaline phosphatase) treatment to deactivate remaining primers and dNTPs. Prior to the hME reaction, a multiplexing protocol is required for adjusting MassEXTEND primers to even out peaks in their mass spectrum. The hME reaction contains the MassEXTEND primer, DNA polymerase, and a mix of dNTPs and ddNTPs. Nucleotide mixtures are selected to maximize mass differences for all potential products. Appropriate dNTPs are incorporated through the polymorphic site until a single ddNTP corresponds and the reaction terminates. The PCR program has an initial denaturation step of 94°C for 2 min and then 94°C, 52°C, and 72°C for 5 s each, repeated for at least 75 cycles. During this PCR, a common MassEXTEND primer that identifies both alleles is hybridized on or adjacent to the polymorphic site. The hME PCR products are treated with clean resin to remove salts that could interfere with MALDI-TOF (matrix-assisted laser desorption ionization–time of flight) analysis. Then 15 μL of sample are transferred onto the pad of the 384 SpectroCHIP bioarray. The chip continues into the MALDI-TOF and the mass correlating genotype is determined in real time with MassARRAY RT software.
In total, 311 flycatchers from the hybrid zone of Öland were genotyped. In addition, we genotyped 106 allopatric collared flycatchers (Abruzzo, Italy and Pilis Mts., Hungary) and 129 allopatric pied flycatchers (Oslo, Norway and Lingen, Germany). The latter birds were included to aid the assignment of the sympatric birds into pure and hybrid classes.
To assign hybrid identity to each of the individuals, we computed the posterior distribution that individuals fell into different hybrid categories according to the algorithm published by Anderson and Thompson (2002), as implemented by the software NewHybrids version 1.1 beta. We preassigned all the allopatric individuals to the two pure species using the “z and s option” and classified the sympatric birds into the following classes: pure (i.e., alleles only from one of the parental species), F1 hybrids, first-generation backcrosses (B1: F1 hybrid × either parental species), and second-generation backcrosses (B2: B1× either parental species), using the recommended “Jefreys-like” priors (Anderson and Thompson 2002). We excluded F2 hybrids as a classification option because female hybrids are invariably sterile in these birds (Veen et al. 2001). We also ran simulations including third-generation backcrosses (B3). Inclusion of this additional hybrid category did not affect the assignment of F1 hybrids or B1s. However, the program was unable to reliably assign B2s relative to B3s. Hence, although we cannot rule out that some of the individuals we classify as B2s actually belong to later backcross classes, overestimating the frequency of B2s means that our fitness estimates are conservative. For each simulation, we used 10,000 replicates and a burnin time of 5000 steps. For the Z-linked markers, females were given haploid genotypes, and for the mtDNA marker, all individuals were given haploid genotypes. Although SNPs within the same gene are not independently inherited, they can still provide complementary information about backcrosses in different directions. The two loci within the gene RPL5 can serve as an example (refer to Supporting information for a list of alleles and their frequencies in the two species). RPL5_4 is polymorphic in pied flycatchers but monomorphic for one of the alleles in collared flycatchers. Hence, an individual with a predominantly collared genetic background, but possessing the “pied” allele (which is not present in pure collared flycatchers) indicates a backcrossing event. However, the same locus lacks power to detect backcrosses to pied flycatchers, as there are no uniquely “collared” alleles. By including a second locus within the same gene (RPL5_4C) that is polymorphic in collared flycatchers and monomorphic in pied flycatchers, we have the potential to detect backcrosses in both directions. In one case (the gene RHO), one locus was fixed for alternative alleles in the two species, making the additional locus within that gene redundant. This second locus (RHO_1C) was thus removed from the analyses.
Once hybrids and backcrosses were identified genetically, we used two independent methods for estimating the fitness consequences of hybridization. The first involved comparing the frequencies of the different hybrid generations within the population. In a population of stable size, each breeding individual produces, throughout its lifetime, a mean of two offspring that return to breed themselves. Thus, fitness was inferred by comparing the observed number of hybrids with that expected from the 32 heterospecific pairs observed in the population during the three sampling years. Because F1 hybrids result from a single reproductive event, we used the expected number of offspring arising from this event, they should have equal fitness to the population mean, as 2/L, where L is the mean number of breeding attempts during a lifetime. A second consideration is that not all offspring raised by heterospecific pairs are hybrids. A large number of offspring reared by female collared flycatchers paired with pied males are in fact pure collared flycatchers, resulting from extra-pair mating with conspecific males (Veen et al. 2001). On the other hand, offspring in nests of collared males paired with pied females are almost always hybrids (Veen et al. 2001). If Po proportion of the offspring in heterospecific pairs with collared females are F1 hybrids, and this pairing combination constitutes Pp proportion of all heterospecific pairs, the number of descendents in the gth generation expected to arise from D0 heterospecific pairs is
Relative fitness, W0, is then Dg/Eg, where Dg is the observed number of descendents in generation g. These fitness estimates depend solely on the ratio of D0:Dg. Confidence intervals of ratios are dependent on sample sizes and are estimable by using the significant χ2 of 3.84 for one degree of freedom. The true number of descendents in the gth generation can be estimated to fall with 95% confidence between the two values of e obtained by solving the following formula, easily performed by a computer package such as Matlab:
To obtain the 95% confidence intervals for relative fitness, the two values for e are each divided by Eg. These confidence intervals do not incorporate error associated with estimating Eg, which depends on the accuracy of the estimates for Pp, Po, and L. In flycatchers, these estimates are based on large sample sizes ( heterospecific pairs, = 183 chicks in 33 nests, NL= 8986 individuals) and we therefore assume this error to be negligible.
The mean relative fitness of all flycatchers in the population is 1, and W0 refers to how many times more descendents arise through hybridization than the average for all breeding events (note that hybridization produces fewer than average descendents when W0 < 1, and smaller values of W0 indicate stronger selection against hybridization). Because the vast majority (94.9%) of individuals within the population on Öland do not hybridize, the mean fitness of the population approximates the mean fitness of nonhybridizing individuals. Thus, the fitness consequences of hybridization, W0, can be considered as how many times more descendents are produced by hybridization than by conspecific breeding events.
An accurate estimate of relative fitness depends on a complete sampling of later-generation hybrids and on an equal representation of hybrids among immigrants and emigrants. These assumptions are untested in our study population and we therefore used a second, independent method to calculate fitness consequences of hybridization. We combined data on reproductive success of each generation (single-generation estimates) from this study and previous studies into a multigeneration estimate for the fitness costs of hybridization. The reproductive success of heterospecific pairs was considered the number of offspring produced by such pairs that returned to breed themselves, divided by the average number for collared pairs. Because of the large amount of extra-pair paternity within nests reared by pied males and collared females, only nests reared by collared males and pied females were included (source of data: Veen et al. 2007).
Likewise, the reproductive success of F1 hybrids was the number of genetic offspring of an F1 hybrid that recruited back to the breeding population in subsequent years, relative to the number of recruiting offspring from nonhybridizing collared individuals of the same sex. The value used was the average between that of males (source of data: Svedin et al. 2008) and females (sources of data: Veen et al. 2001; Svedin et al. 2008). Because data were lacking on recruitment from B1 nests, our estimates of reproductive success derive from the number of genetic offspring fledged from the nest relative to fledging rates in collared nests. The reproductive successes of B1 males and B1 females were averaged to give overall selection.
To first confirm the accuracy of the automated scoring method used in the current study, we included 459 individuals (seven autosomal and two Z-linked SNP-sites each) from a previous study (Sætre et al. 2003), which applied the Tag array method described in Lindroos et al. (2001). Of 3695 genotypes called by the two methods, 3684 (99.7%) were identical.
To examine the accuracy of our genetic assignment of hybrid identity, we screened 20 birds (16 F1 hybrids, three B1 backcrosses, and one B2 backcross) for which independent information on hybrid identity was available from pedigrees and data from crossing experiments in aviaries. In all cases, the markers used to assess hybrid identity correctly identified each individual.
In 2004, we sampled almost all breeding birds (N= 219) within 11 nestbox locations on Öland for genotyping. Of these, we detected two F1 hybrids (0.9%), one B1 backcross (0.46%), and no B2 backcrosses. The three hybrids/backcrosses had intermediate phenotypes typical of F1 hybrids (Sætre et al. 2003). Furthermore, all 211 birds that were phenotypically pure were assigned as having pure pied or collared genotypes. Five additional birds were considered possible hybrids based on phenotype, but had collared genotypes. Because of the rarity of hybrid and backcross individuals and the poor return from randomly screening entire populations to locate them, we decided to carry out a more focused search on a much larger sample of birds by only genotyping birds with aberrant or intermediate plumage characteristics. This was done for the entire breeding population on Öland in 2003 and 2005. Using this technique, of 987 individuals recorded breeding (92 were genotyped), 18 (1.8%) were F1 hybrids, four (0.4%) were B1, and three (0.3%) were B2 backcrosses. These estimates were similar to, or exceeding, those based on a more random sampling technique (in 2004), indicating that by focusing our sampling toward intermediate-looking birds we did not miss a substantial number of hybrids. All analyses, therefore, were carried out on the enlarged sample afforded by combining the data from all 3 years.
Based on the frequency of each hybrid generation within the breeding population, hybridizing individuals experienced only 2.4% (95% CI = 0.8–6.2%) of the fitness relative to nonhybridizing individuals. This high degree of postzygotic isolation was strongly concordant with our second, independent estimate of fitness based on reproductive output from each of the hybrid generations. Based on this second method, hybridizing individuals experienced 2.7% of the fitness relative to nonhybridizing individuals (Fig. 1).
Hybridizing individuals, their F1 offspring, and their B1 grand-offspring all produced fewer breeding progeny than average (Fig. 1, all values less than 1). As a result, estimates of the fitness consequences of hybridization (the product of the estimates for each generation) are highly dependent on the number of generations considered. The estimated fitness consequences of hybridization based on the frequency of B2 backcrosses were less than 4% of that based on the frequency of F1 hybrids, and 30% of that from B1 backcrosses. This was due to female sterility persisting beyond the F1 generation and male backcrosses siring a low proportion of offspring raised in their nests.
We examined the reproductive output of the small number of B2 backcrosses detected in this study to elucidate whether depressed fitness persisted beyond the B1 backcross generation. Surprisingly, in this study 80% of individuals were belonging to the B2 generation (N= 5) “rehybridized” (their partner was of the species that constituted the minority of their genome). The probability that B2 backcrosses hybridize with the same frequency as pure birds is 0.0001. This rate of “rehybridization” among the few B2 backcrosses is even unlikely from random mating (P= 0.023). This propensity for “rehybridization” among later-generation hybrids represents an additional cost of hybridization not factored into the fitness estimates presented in Figure 1.
The estimation of postzygotic isolation is closely tied to the more general problem of how to assess fitness. Estimating individual variation in fitness (number of descendents in future generations) is fundamental for predicting evolutionary changes in natural populations, yet we lack empirical data on how well “number of offspring” reflects true genetic representation many generations down the line. This study demonstrates that hybridization is one instance where the reproductive output of individuals and even their F1 hybrid offspring vastly overestimates their fitness. This idea is not a new one. Studies conducted in laboratories or gardens have previously demonstrated long-lasting effects of hybridization, either through the persistence of outbreeding depression (e.g., Darmency and Fleury 2000; Pages et al. 2002) or through enhanced fitness among later-generation hybrids (e.g., Arnold 2004; Erickson and Fenster 2006). Even within birds, persistence of low egg hatchability in B1 backcrosses has been reported in captive hybrid pigeons (Lijtmaer et al. 2003). However, logistic difficulties have long precluded the study of naturally occurring later-generation hybrids in the wild, where selection coefficients are most relevant. Data presented here on Ficedula flycatchers suggest that a consideration of multiple generations of offspring, living under the full complement of selective pressures experienced in a natural environment, can produce vastly different estimates of postzygotic isolation than more standard assessments based on the reproductive output of F1 hybrids. To our knowledge, this is the first time such a holistic approach to assessing postzygotic isolation between two populations has been achieved.
The discrepancy between the two- and three-generation estimates presented for flycatchers is not trivial. Fitness estimates based on the reproductive output of F1 hybrids imply that although hybridization is maladaptive (a 90% reduction in fitness), when faced with the alternative of not breeding at all (in times and places where conspecifics are unavailable), it may be an adaptive strategy for producing at least some descendents. However, for flycatchers, as in other species, breeding is associated with costs to future reproduction (Gustafsson and Sutherland 1988; Gustafsson et al. 1995), and in light of the more realistic three-generation estimate of the cost of hybridization (hybridizing individuals leave almost no descendents in future generations) it may be more adaptive for individuals without access to conspecific mates to refrain from breeding that year. Interpreting selection on individuals to avoid hybridization may, therefore, be highly dependent on the number of generations of descendents considered.
These findings raise the important question of how many generations should be investigated before an estimate of postzygotic isolation can be considered accurate? Ideally, one should consider the fitness of each generation of descendents until the fitness of the final generation is independent of the first (i.e., the relative fitness of the final generation is 1). High rates of “rehybridization” among the few B2 backcrosses detected in this study indicate that depressed fitness probably persists beyond the two hybrid generations examined here, indicating that the extremely high levels of postzygotic isolation detected in flycatchers are probably underestimates.
Studies of postzygotic isolation that focus attention on the rates of evolution of intrinsic F1 sterility and inviability offer important insights into the nature of underlying genetic incompatibilities and their evolutionary patterns. However, our results caution against using this data on single, constituent barriers to infer overall postzygotic isolation. In birds, the idea that barriers preventing mating typically evolve fastest is founded on the observation that F1 sterility evolves too slowly to account for rates of speciation (Price and Bouvier 2002; Lijtmaer et al. 2003). By default, prezygotic barriers were considered of greater importance during speciation. However, despite a comparatively large amount of interbreeding between collared and pied flycatchers, and only partial infertility of F1 hybrids, there is almost complete postzygotic reproductive isolation between the two species in nature. Improved estimates of postzygotic isolation that encompass all components of hybrid fitness will help to resolve this question about the relative rates of evolution of different reproductive barriers.
Despite the strong postzygotic barriers separating Ficedula flycatchers, some introgression of alleles between the two species has been reported (Borge et al. 2005a), implying that reproductive isolation is not total. The flycatcher hybrid zone highlights the possibility that even when hybridization is highly maladaptive, it can still constitute a potential source of novel adaptations for the parental taxa (e.g., Barton 2001; Burke and Arnold 2001).
Hybridization is distinct from heterospecific pairing, and when extra-pair copulations are widespread within socially monogamous species, the two can have different fitness implications. Although extra-pair paternity is irrelevant for postzygotic isolation, it can have profound impacts on the selection experienced by heterospecifically pairing individuals. Previous studies have indicated that heterospecific pairing can sometimes be adaptive for female collared flycatchers due to high frequencies of heterospecific male partners being cuckolded by conspecifics, and occasional direct benefits for heterospecific pairing (Veen et al. 2001; Wiley et al. 2007). However, this result was based on previous, higher estimates of hybrid fitness. By combining published rates of extra-pair paternity (50% when the female is a collared flycatcher and 0% when the female is a pied flycatcher; Veen et al. 2001) with our new estimates of fitness costs of hybridization, on average heterospecific pairing is maladaptive for female collared flycatchers (51.2% of the fitness of nonhybridizing females; see Table 1). For heterospecific pairing to be an adaptive strategy for collared females, their extra-pair offspring must have 1.98 times higher fitness than that of other collared flycatchers, or they must produce 1.98 times more offspring. Although this may happen, when direct benefits of having a heterospecific partner are maximal (see Veen et al. 2001; Wiley et al. 2007), on average heterospecific pairing is maladaptive for female collared flycatchers. For female pied flycatchers, and males of both species, heterospecific pairing results in almost no descendents. This implies that the occurrence of heterospecific pairings is typically maladaptive, and may reflect incomplete reinforcement of species recognition in this relatively recent hybrid zone. Alternatively, constant gene flow from allopatric populations not selected to avoid hybridization may prevent the complete cessation of heterospecific pairing.
Table 1. The mean relative fitness (W) of heterospecifically paired males and females (accounting for extra-pair paternity), based on the frequency of descendents three generations later.
Heterospecifically paired group
Few processes in nature have been argued to lead to such fundamentally important and diverse evolutionary outcomes as hybridization between different species. Hybridization can reduce species diversity by leading to the merging of diverged populations (Felsenstein 1981; Taylor et al. 2006), promote reinforcement of differences between populations (Dobzhansky 1937), and serve as a source of novel genetic combinations on which selection can operate (Arnold 2004; Gross and Rieseberg 2005). These different consequences largely depend on the strength and direction of selection operating on the hybridizing individuals, that is, their fitness relative to that of nonhybridizing individuals. In flycatchers, a strong postzygotic barrier to gene flow allows comparatively high levels of heterospecific pairing (a prerequisite for enduring reinforcement) without merging of the two species. We anticipate that improvements in the speed and costs of molecular techniques for identifying later-generation hybrids will make additional multigeneration estimates of fitness an increasingly feasible proposition in the future. Such data will be instrumental in furthering our knowledge of the origin of new species.
Associate Editor: M. Webster
The authors thank N. Svedin, J. Baarman, and S.A. Sæther for blood samples. Data on backcrosses from Gotland were provided by M. Hjernquist, J. Forsman, T. Veen, and M. Andersson. L. Gustafsson and S. Bures provided additional data from Gotland and the Czech Republic, respectively. The genotyping service was provided by the Norwegian national technology platform, CIGENE, supported by the functional genomics programme (FUGE) in the Research Council of Norway. Funding for the project was provided by Zoologiska Stiftelsen (to CW), Norwegian Research Council (to GPS), Oslo University, the Swedish National Research Council (to AQ), and the Swedish Research Council for Environmental, Agricultural Sciences and for Spatial Planning (to AQ).