• allozymes;
  • anthropogenic disturbance;
  • cytonuclear disequilibrium;
  • hybridization;
  • Lyciasalamandra antalyana;
  • Lyciasalamandra billae;
  • mtDNA length polymorphism;
  • selection;
  • Turkey


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

The fate of species integrity upon natural hybridization depends on the interaction between selection and dispersal. The relative significance of these processes may be studied in the initial phase of contact before selection and gene flow reach equilibrium. Here we study a hybrid zone of two salamander species, Lyciasalamandra antalyana and Lyciasalamandra billae, at the initial phase of hybridization. We quantify the degree and mode of introgression using nuclear and mtDNA markers. The hybrid zone can be characterized as an abrupt transition zone, the central hybrid zone being only c. 400 m, but introgressed genes were traced up to 3 km. Introgression was traced in both sexes but gene flow may be slightly male-biased. Indirect evidence suggests that hybrid males are less viable than females. Introgression occurred at two levels: (1) locus-specific selection led to different allelic introgression patterns independent of species, while (2) asymmetrical species-level introgression occurred predominately from L. antalyana to L. billae due to range expansion of the former. This indicates that foreign genes can be incorporated into novel genomic environments, which in turn may contribute to the great diversity of morphological variants in Lyciasalamandra.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

Hybrid zones offer rich possibilities to study genetic distinctiveness of populations and the selective forces that keep them different (Arnold, 1997). If hybrids are less fit than the parent species, reproductive barriers may evolve as the final step towards total separation. Renewed gene flow may break specificity if hybrid fitness is not reduced or, if fitness of one parental population is lower than that of the other, hybridization may lead to replacement of the less fit population. Alternatively, hybridization may be a creative force shaping new hybrid species. In the latter cases, genes of the failing species may be absorbed into the successful one (Clarke et al., 2002).

Different introgression rates of genetic markers permit inferences about how selection and/or behaviour play a role governing the admixture of the organisms (e.g. Della Torre et al., 1997; Sweigart & Willis, 2003). The width of a clinal hybrid zone is determined by a balance between dispersal rates and fitness (Barton & Hewitt, 1985). Characteristic of hybrid zones is the nonrandom association of alleles. When disequilibrium is strong, the central region of a cline will be distinct from the tails and the cline-width will depend on the ratio between recombination and selection (p116, Barton & Hewitt, 1985). If selection acting on distinct characters differs, they may introgress at different rates. Because mtDNA is assumed selectively neutral, it is generally believed to have greater rates of introgression than nuclear genes, but an equally parsimonious explanation is positive selection that sweeps the mitochondrial genome through the population (Dasmahapatra et al., 2002). On the other hand, sex-specific gene flow will cause asymmetric rates of introgression, particularly if one sex is less viable or fertile than the other. The less fit sex is nearly universally the heterogametic sex (Coyne & Orr, 2004) and this is known as Haldane's rule (Haldane, 1922). In birds, male biased gene flow (i.e. restricted mtDNA gene flow) across species borders is, at least partly, due to reduced viability of the heterogametic females (Helbig et al., 2001; Crochet et al., 2003). However, mtDNA and nuclear polymorphisms often are coincident and concordant (Szymura & Barton, 1986; Dasmahapatra et al., 2002).

Amphibians are well-known for their ability to hybridize naturally. The tendency to hybridize, even between species with high sequence divergence (Szymura et al., 2000) indicates genomic resilience and makes amphibians prone to build species-complexes (Tarkhnishvili et al., 2001; Bogart, 2003). Hybridization occurs between sympatric species (e.g. Szymura & Barton, 1986; Green & Parent, 2003), as well as between introduced (allopatric) species (e.g. Riley et al., 2003). Several authors have noted that hybridization in sympatric species can be brought about by human disturbance of the natal habitat (Riley et al., 2003 and references therein). Particularly the possibility to hybridize between species from divergent mtDNA lineages raises the question whether resilient species require longer speciation times or, alternatively, shorter time due to introgressive speciation. The former scenario requires strict allopatry while the latter allows character divergence to build up in a relationship between gene flow, selection and some habitat divergence.

The Lycian salamanders of the genus LyciasalamandraVeith & Steinfartz (2004) (former Mertensiella luschani Steindachner, 1891; see Veith & Steinfartz, 2004) consists of nine morphologically distinct taxa, distributed allopatrically along the southern Turkish coast and adjacent Greek and Turkish islands. The total range covers only ∼350 km. Weisrock et al. (2001) analyzed sequence divergence in eight taxa and found highly divergent lineages with 7.6–10.1% pair-wise sequence divergence, rooted in a basal unresolved polytomy. Morphological variants can even be recognized among populations of the nine taxa (Veith & Steinfartz, 2004). The lineages diverged between 5.9 and 7.9 MYBP, which corresponds to vicariant speciation initiated during the rise of Anatolia in the late Miocene (Weisrock et al., 2001; Quennell, 1984). Habitat requirements include average rainfall of >800–1000 mm, average temperature in January above 0° and a soil structure with crevices (Klewen, 1991; Steinfartz & Mutz, 1998; Veith et al., 2001). The crevices are crucial for survival as refuges during the dry and hot summers. These habitat requirements limit the distribution of Lyciasalamandra to island populations associated with limestone outcrops.

Only two taxa, Lyciasalamandra billae and Lyciasalamandra antalyana, are known to have adjoining populations (Veith et al., 2001). The two taxa behave as discrete species; they are morphologically different and show 7.8% mtDNA sequence divergence (Weisrock et al., 2001). However, during a survey of the distribution of Lyciasalamandra, Veith et al. (2001) reported several morphologically intermediate individuals suggesting a contact zone between the two species. The aim of this paper is to delimit a putative contact zone between L. billae and L. antalyana and to analyse the degree and mode of introgression between the species. The species are characterized by two diagnostic nuclear (allozyme) loci and by a diagnostic mtDNA D-loop length polymorphism. We address four questions: (i) Is introgression gradual or does hybridization follow a mosaic pattern? (ii) Is gene flow sex-biased? (iii) Is introgression uni-parental? and (iv) Is introgression locus-specific (evidence for selection)? Gradual introgression is expected when the frequency of parental individuals gradually decreases away from the pure populations, hybrids increase towards the centre and hybrids have intermediate fitness. One expects deviation from HW proportions and linkage disequilibria within the contact zone and selection against hybrids. Because only two nuclear loci (out of 31 loci) are polymorphic the expectation of linkage build up across several loci in the centre of the zone cannot be tested. Mosaic introgression is expected when parental individuals mix within sites and hybrids are rare (thus allowing parentals to be relatively common). Inference of sex-biased gene flow was gained by comparing nuclear and maternal variance estimates (Ennos, 1994; McCauley, 1994) and cytonuclear disequilibria (Asmussen & Basten, 1996; Basten & Asmussen, 1997). The former comparison is based on equilibrium conditions assuming equal sex ratios, mating success of the sexes, homoplasmy and unisexual transmission (Birky et al., 1983,1989); when maternal (mtDNA) variance exceeds nuclear variance by four times it indicates that females contribute more to gene flow than males. Because of the assumptions, this method should be considered with caution. Inference of uni-parental introgression was gained from cytonuclear disequilibrium (see section ‘Methods’ for details). Finally, inference of locus-specific introgression (selection) was gained from the frequency of heterozygotes at two diagnostic loci. As we will show, the level of heterozygosity in hybrids differed at the two loci. There are two possible explanations for this pattern. First, heterozygotes at both loci may be lacking in hybrids, one more so than the other. In this case, both loci are affected by negative selection, one more than the other. In the second outcome, a heterozygote at one locus may, despite being lacking in the total gene pool, actually be overrepresented in hybrids relative to a situation with general selection against hybrids. In other words, heterozygotes are less lacking than expected in the total gene pool. In this case, overrepresentation may be taken as evidence for positive selection. To evaluate which of these possibilities explained genotype-specific behaviour we simulated a simple null-model assuming no selection and stepping-stone admixture. The simulations further allowed testing genetic drift as a cause for different specific multi-locus genotypes.

Method and materials

  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

Species desciption

Lyciasalamandra antalyana and L. billae look very similar in size and shape. However, they are well distinguished in parotid and flank coloration. Lyciasalamandra antalyana has bright yellow parotids, irrespective of the head's colour; in L. billae parotids are always coloured like the head, either brown or incarnate. In hybrids, parotids mostly resemble those of L. billae, although sometimes with a whitish central spot. Flanks of L. antalyana are yellowish; those of L.billae are white. Flanks in hybrids are ivory-coloured.

Sample locations

The contact zone runs along a creek in a narrow valley near the villages Gökdere and Hurma, 30 km southeast of Antalya in southern Turkey (Fig. 1, Table 1). It seems that L. antalyana has managed to cross the creek from the north to the L. billae side where morphologically intermediate individuals were identified (sample 23). At first sight there are two potential centres of hybridization. The first site is the village Gökdere where a bridge crosses the creek. The second potential centre is 2 km down-stream where permanently irrigated orange groves, planted about 60 years ago, have made it possible for L. antalyana to move across the valley. Hence, the opportunity for crossing at the village is considerably older.


Figure 1. Sampling locations of Lyciasalamandra billae and Lyciasalamandra antalyana at a contact zone in south Turkey. The subspecies meet at a creek (thick line).

Download figure to PowerPoint

Table 1.  Sampling locations of Lyciasalamandra antalyana and Lyciasalamandra billae southwest of Antalya, Turkey; population types: A, pure L. antalyana; B, pure L. billae; AB, introgressed population.
LocationCo-ordinatesnAltitude (m)SlopeSide of creekPopulation type
  1. Side of creek: T, terra typical.

136°51,1030°35,3117175SEN (T)A
2436°47,0930°34,032515ES (T)B

To delimit the contact zone we sampled populations in the valley as well as putative parental populations found in the heights on either side of the valley. Sampling was done from 1996 to 1999 from 25 populations, including 283 individuals. All animals, except from populations 7 and 24, were individually marked and genotyped at allozyme and mtDNA loci. Individuals from populations 7 and 24 were sampled in 1996 when the putative hybrid zone was discovered. The genetic estimates from these two populations are frequency estimates. To estimate the degree of introgression, we sampled two transects on the L. billae side of the creek. The first transects ascends a hill behind the population 23 while the second transect follows a road along the creek and into the hills behind. All locations were positioned with GPS.

Allozyme polymorphism

All populations in the present study were screened for 25 presumptive allozyme enzymes and blood plasma markers comprising 31 loci [investigated in all nine species of the genus Lyciasalamandra M. Veith (in prep)]. The topotypical L. antalyana were monomorphic at all 31 loci. The topotypical L. billae were monomorphic at 30 loci and polymorphic at one locus, Pepidase-B. The two species differed genetically by being fixed for alternate Aat and Idh alleles (Table 2). The enzymes Pep-B (Leu-Gly-Gly) EC 3.4.11 or 13, Aat EC, Idh EC and the enzyme Pgd EC (polymorphic in a single population) were analyzed by cellulose-acetate electrophoresis. Running conditions for the polymorphic enzymes in this study were as follows: Pep-B, Idh and Pgd were analysed from blood; Aat from muscle tissue. Pep-B was run in Tris–Glycine buffer pH =8.5, 250 V; Idh in Tris–Maleic acid pH = 7.0, 200 V; Pgd in Tris–citrate acid pH = 8.2, 200 V; Aat in Tris–Borate pH = 7.8, 200 V. All enzymes were run for 40 min.

Table 2.  Multi-locus allozyme genotypes and D-loop length polymorphisms of parental Lyciasalamandra billae and Lyciasalamandra antalyana.
SpeciesAllozyme genotypeD-loop
  1. Genotype designation based on all nine Lyciasalamandra taxa.

Lyciasalamandra antalyanaAAAACC1
Lyciasalamandra billaeBBBBC/D2,3,4

Mitochondrial DNA

To characterize female gene flow and cytonuclear disequilibrium we amplified the mtDNA D-loop using the primers L-Pro-ML and H12S1-ML (Steinfartz et al., 2000; DNA extraction and amplification protocols are given therein). We limited the mtDNA analyses to scoring a D-Loop length polymorphism, which differs between the two species (Johannesen, 2004). The L. antalyana D-loop is about 900 bp and that of L. billae about 1100 bp. However, during the investigation we observed three length polymorphisms in L. billae (new length polymorphisms 1300 and 1500 bp). No additional length polymorphisms were found in L. antalyana. To check for correct assignment of length polymorphisms to parental species we sequenced 29 L. antalyana from eight populations (11, 17, 18, 10, 12, 23, 3, 8) and 20 L. billae from seven populations (16, 14, 22, 13, 19, 20, 21) for about 550 bp. Phylogenetic analysis showed monophyly of L. antalyana haplotypes and of L. billae length polymorphisms. The sequence data will be published elsewhere (B. Johannesen & M. Veith in prep). Because the present paper investigates introgression only, we have treated the three L. billae length polymorphisms as one haplotype.

Data analysis

Individuals were scored as introgressed if (i) they were heterozygotic in any combination of the fixed parental Aat and Idh alleles and (ii) for individuals with homozygote parental genotypes, those bearing an alternative D-loop haplotype. We tested for nuclear linkage disequilibrium and deviations from Hardy–Weinberg (HW) proportions using Genepop v. 3.2 (Raymond & Rousset, 2000). The observed multi-locus genotypes in the hybrid zone were compared to a random mating population. Genetic differentiation within species was assessed for nuclear and mtDNA alleles with the program Genepop v.3.2. Cytonuclear disequilibrium was tested with the program cndm (Asmussen & Basten, 1996; Basten & Asmussen, 1997). The cytonuclear disequilibrium has four estimators, D, D1, D2 and D3, where positive D = inline image indicates a positive association between nuclear alleles (A) and mtDNA (M) from the same parental species. D1 = inline image is the association between parental mtDNA and its homozygote (AA), D2 = inline image is the association between parental mtDNA and the heterozygote (AB), and D3 = inline image is the association between mtDNA of one species and the homozygote of the other species (BB). The estimator D2 is of special interest because it estimates whether heterozygote (hybrid) individuals are more likely to have one species’ mtDNA than the other. A significant value implies fitness differences relative to genetic backgrounds or mating asymmetry between the species.

Computer simulations

Because not only selection but also genetic drift in small populations may produce different locus-specific introgression rates, we simulated a simple null-model (Griebeler et al., in press) to test whether (1) specific multi-locus genotypes were more or less common than expected and whether (2) observed multi-locus genotype frequency differences could be explained by genetic drift alone. The simulation model of Griebeler et al. (submitted) studies stepping-stone dispersal and population admixture. The model was adapted to the current settings only by changing parameter values and population numbers (see below).

The model

We suppose in our individual-based model that two neutral diallelic loci (Aat and ldh) represent the genetic system of the two diploid admixing salamander species. Crosses between species are fertile and no selection is assumed. Maximum recombination between the two loci is allowed. The individuals inhabit a linear array of demes with dispersal between them. Each deme harbours a constant number of individuals (n). These may be of the two parental types or may be hybrids. Admixture of species results from stepping-stone dispersal of individuals. Emigration of an individual occurs once within a generation with a constant probability. The displacement of each emigrant is possible in both directions. The two adjacent demes of a source deme may be chosen by a migrant with distinct probabilities to allow for symmetric and asymmetric introgression. Dispersal is followed by n random matings in each subpopulation. Parents for offspring are selected randomly. The offspring produced in each subpopulation constitutes the next generation.


The simulated introgression zone consisted of 60 demes aligned evenly along a 3 km single-dimensional transect. This corresponds to sampling every 50 0m. These settings were based on observed natural conditions: introgression proceeded linearly and was traced 3 km. At the beginning of the simulated introgression process demes located at one edge were fixed for the parental multi-locus genotype AA/AA (L. antalyana) and the others for BB/BB (L. billae), respectively. All demes were populated at size 50 or 100, respectively, except of the two subpopulations at the transect ends. To avoid edge effects, we assumed that each of these subpopulations had 10 000 individuals of each parent species.

We simulated two extreme introgression scenarios. In the first scenario dispersal was symmetric, while dispersal in the second scenario was limited to one direction only. The first simulation scenario was based on observed ratio of L. antalyana to L. billae alleles 1 : 4. In the modelled transect; 12 L. antalyana subpopulations were followed by 48 L. billae subpopulations (corresponds to the 1 : 4 ratio). Individuals of both species and their hybrids were allowed to migrate with an equal chance to the two neighbouring subpopulations with a probability of 0.1 per generation. The second scenario corresponds to the assumption that L. antalyana has crossed the creek, is expanding and replacing L. billae in its natal habitat. Here we assumed initially a total population of 5% pure L. antalyana (three subpopulations) and 95% pure L. billae (57 subpopulations). For the introgression of L. antalyana into L. billae subpopulation we assumed a dispersal rate of 0.1. Both scenarios for introgression were simulated with subpopulations of 50 and 100 individuals, respectively.

We performed 100 Monte–Carlo simulations for both introgression scenarios. We simulated 250 generations in each simulation run. After each 25th generation, 15 individuals were sampled randomly per subpopulation. The final genotypic multi-locus distribution across the zone of introgression was obtained from the total sample of randomly drawn individuals, i.e. 900 (60 demes × 15 individuals sampled) parental and hybrid individuals.

Due to the expanding front of introgression, each simulated generation will have different relative multi-locus distributions. In order to know which generation from the simulated data to use for our null-model, we qualitatively compared frequencies of the parental multi-locus genotypes obtained from the simulations to the observed data. The best fitting generation was compared to the observed data. It should be noted that for practical reasons the simulations were performed for a single one-dimensional zone of admixture. In reality, we observed two one-dimensional clines. In the comparison of simulated and observed data we have pooled the observed data from the two transects. This seems appropriate because the two transects behaved identically (see section ‘Results’).


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix


The level of polymorphism from 25 sample locations is presented in Appendix A. All populations north of the creek designated as L. antalyana (locations 1–3, 8, 10–12 and 18) were completely monomorphic, holding the diagnostic genotypes and haplotype listed in Table 2. In contrast, populations of L. billae showed signs of introgression from L. antalyana in lowland populations but had only parental genotypes at higher elevation. Population 7 had a unique Pgd allele.

Of the 283 studied individuals, 39 (14%) were hybrids (5 adult males, 15 adult females, 11 juveniles, 4 subadults and 4 no information). Because hybrids were delimited by only two nuclear loci and one mtDNA polymorphism the number of hybrids should be considered a minimum value. Multi-locus genotypes together with corresponding D-loop haplotypes show that introgression includes F2 and/or backcross animals. Six double homozygote individuals were all juveniles. Only one individual was double heterozygote (Table 3). Both ‘pure’ parental types were never found at the same location.

Table 3.  Multi-locus nuclear genotypes (Aat/Idh) and D-loop length-polymorphisms observed at two hybrid areas.
Haplotype 1Haplotype 2,3,4Unknown haplotype*Haplotype 1Haplotype 2,3,4Unknown haplotype
  1. *DNA extraction was corrupted and amplification failed.

  2. †Allozyme data from population 7 is based on frequency data from 15 individuals that were not individually assigned. The nine individuals scored for mtDNA in population 7 had haplotypes 2 or 3. It is therefore likely that six BB/BB individuals have haplotype 2, hence the designation (9).

  3. The primary hybrid zone included the samples 23, 22, 21, 17, 13, 16 and 14. The second hybrid area consisted of individuals from samples 4, 5 and 7. Lyciasalamandra antalyana, AA/AA and haplotype 1; Lyciasalamandra billae, BB/BB and haplotypes 2,3,4.

BB/BB160603 (9)†12†

We observed two clines extending from sample 23, along the creek (locations 13–14 and 16–17) and up the hill (locations 21–22) (Fig. 2). We termed these locations the ‘primary hybrid zone’. Transect sampling revealed an abrupt transition zone between the two species in both transects. The central zone of introgression is only about 400 m. The total cline width is about 3 km wide, based on the road distance along the creek. The introgression likely proceeds from L. antalyana to L. billae because no L. billae alleles or haplotypes were found on the L. antalyana side of the creek, whereas pure L. antalyana were established on the L. billae side (location 23). The frequency of L. antalyana alleles dropped continuously away from location 23. Interestingly, the bridge in the village Gökdere seems to play only a limited role, if at all, in promoting species contact. Here, no pure L. antalyana were established on the L. billae side as was observed at location 23.


Figure 2. Frequencies of Lyciasalamandra billae specific alleles across two transects, from location 23 to location 20 and 9, respectively. Locality 23 exhibited exclusively Lyciasalamandra antalyana mtDNA but nuclear alleles from both species. mtDNA – open triangles, Aat– solid diamond, Idh– solid squares. Numbers signify locality numbers.

Download figure to PowerPoint

A second introgression area was found where two mountain ranges converge south west of Gökdere (locations 4–5 and 7) but data are too sparse to make generalizations about this area.

Mating patterns and gene flow

Tests for random association of nuclear alleles within the primary hybrid zone [treating all sites (13–14, 16–17 and 21–23) as one sample] revealed significant linkage-disequilibrium for all locus combinations (Aat/Idh, Aat/Pep-B, Idh/Pep-B, all P < 0.001) and deviation from HW proportions (Fishers exact test, inline image = 106.2, P < 0.001). Each locus deviated significantly from HW proportions (P < 0.001). However, the two species-specific loci differed in the amount of heterozygote deficit (expressed as AA : AB : BB) Aat: 19 : 4 : 84, Idh: 14 : 20 : 72 (inline image = 13.3, P < 0.01). Deviations from HW proportions were not significant within locations within the primary hybrid zone. This may be caused by a combination of events and was not entirely unexpected: locality-sample sizes were small (location 17), hybrids few (particularly including Aat– see below) and HW tests are conservative. However, a pattern emerged where every sample had a deficit of Aat heterozygotes (significant only for location 22, P < 0.05), but which was significant across localities, P < 0.01. In contrast, Idh had a heterozygote excess within localities in all but one sample and did not deviate from HW proportions across samples (P = 0.97). HW proportions calculated for 27 hybrids where mtDNA, Aat and Idh were scored in all individuals showed significant lack of heterozygotes for Aat (inline image = 15.13, P < 0.001) but an excess of heterozygotes for Idh, albeit only marginally significant (inline image = 3.33, P = 0.068).

Cytonuclear disequilibria (Table 4) showed highly significant positive associations between parental species’ mtDNA and alleles (D) and homozygotes (D1 and D3). This was expected because several pure parental individuals were found within the hybrid area. More interesting were the associations between heterozygotes and mtDNA (D2) of either parental species. Despite low sample sizes, L. billae had a significantly negative association between mtDNA and Aat heterozygotes, P = 0.036 (L. antalyana significantly positive). In contrast, no association between Idh heterozygotes and mtDNA was observed, P = 0.97. Applying the cytonuclear tests only to hybrid individuals produced no significant associations (results not shown).

Table 4.  Cytonuclear disequilibria between Aat and mtDNA, and Idh and mtDNA from a hybrid zone of Lyciasalamandra antalyana and Lyciasalamandra billae.
  1. The estimates are shown for L. billae, i.e. allele B and haplotype 2. (The D estimates for L. antalyana are symmetrical with opposite sign.) Positive D shows a positive association between nuclear alleles and mtDNA from the same parental species, D1 is the association between parental mtDNA and its homozygote, D2 is the association between parental mtDNA and the heterozygote, and D3 is the association between mtDNA of one species and the homozygote of the other species.

Normalized, D0.824−0.609−0.6610.7410.794−0.031−0.7870.791
Standard error H00.0140.0060.0130.0130.0170.0140.0120.013
Standard error H10.0240.0120.0230.0190.0210.0150.0230.017
Test statistic47.666.6438.2345.3523.580.0744.5540.21

Assessments of intra-specific gene flow were not possible for L. antalyana because it was completely monomorphic. For L. billae estimates of genetic differentiation, FST, based on the nuclear locus Pep-B and mtDNA D-loop polymorphism were possible. Genetic differentiation based on 5 ‘pure’L. billae locations (6, 15, 9, 20 and 19) were: FST(PepB) = 0.11; FST(D-loop) = 0.62. The distribution of the D-loop polymorphisms was particularly subdivided; the easterly ‘pure’ locations 6, 9 and 15 (and the hybrid locations not included in the estimate above) exhibited three length polymorphisms (Fig. 3), whereas the westerly locations 19–23 only had polymorphism 2 (Appendix A). The distance between location 16 and 19 was only c. 600 m by air.


Figure 3. The distribution of D-loop length polymorphisms across the creek transect. Lyciasalamandra antalyana: open circles – haplotype H1; Lyciasalamandra billae: solid squares – H2, closed circles – H3, open triangles – H4. Numbers signify locality numbers.

Download figure to PowerPoint


As shown above, the level of heterozygosity differed between Aat and Idh. Considering this pattern in more detail on a L. billae background, the relationship of heterozygotic Idh (BB/AB) to Aat (AB/BB) in the primary hybrid zone is 14 : 2 (Table 3) and significantly different from equal proportions (inline image = 5.24, P = 0.02). Thus, L. antalyanaIdh but not Aat is introgressing into a L. billae background. Interestingly, the same pattern is also suggested on a L. antalyana background: the relationship of heterozygous L. billaeIdh (AA/AB) to Aat (AB/AA) is 5 : 1 (Table 3), although not significant due to small sample size. The relative number of heterozygote Idh and Aat in all hybrids was 20 : 3 and significantly different from equal proportions (inline image = 6.57, P = 0.01). The relationship of hybrid double homozygotes (AA/BB and BB/AA) was about equal (3 : 2). Thus, different heterozygote behaviour lead to different introgression rates of L. antalyanaAat and Idh alleles. The different behaviour may either be due to under representation of (selection against) Aat heterozygotes or over representation of (selection for) Idh heterozygotes.

Computer simulations showed that the relative frequency of Idh and Aat heterozygotes were similar in all generations and independent of simulation scenario (Fig. 4a,b). Hence genetic drift could not explain the skewed multi-locus heterozygote (BB/AB vs. AB/BB) distribution. The maximal relative difference between BB/AB and AB/BB found during simulations 8.4 : 6.8 (250th generation, symmetric gene flow) differed significantly from the observed 14 : 2 distribution (inline image = 4.39, P = 0.04).


Figure 4. Simulated frequencies of multi-locus genotypes (Aat/Idh) under symmetric and asymmetric dispersal at population sizes n = 50 and n = 100. Heterozygote frequencies with symmetric dispersal (a). Heterozygote frequencies with asymmetric dispersal (4b). Parental frequencies (AA/AA and BB/BB) with symmetric and asymmetric dispersal (c). The dotted lines in (c) represent the observed frequencies of AA/AA (10.5%) and BB/BB (62.9%) in the hybrid area.

Download figure to PowerPoint

No single simulation produced the observed parental genotype frequency (AA/AA & BB/BB) simultaneously (Fig. 4c). In the symmetric dispersal scenario, the frequency of BB/BB was always more common than the observed frequency, whereas AA/AA fitted the observed frequency after 125 generations. At generation 125 the simulated frequency of BB/AB, 3.0%, was far lower than the observed frequency, 13.3%; the simulated frequency of AA/AB, 2.7%, was also lower than observed frequency, 4.8%. In contrast, the simulated frequencies of AB/BB, 2.7%, and AB/AA, 2.8%, were higher than the observed frequencies, 1.9 and 1.0%, respectively (Fig. 4a).

In the second scenario with skewed dispersal with 5%L. antalyana and 95%L. billae, the observed frequency of L. antalyana parental genotypes was reached after c. 75 generations and that of L. billae parental genotypes after 175 generations (Fig. 4c). The two skewed-dispersal simulations behaved similarly at generations 75 and 175 by showing a lack of genotype BB/AB, 1.6 and 2.6%, respectively; and to lesser extent, a lack of AA/AB, 1.4 and 2.2%, respectively. The simulated frequency of heterozygote Aat (AB/BB and AB/AA) was 1.6 and 1.5% in generation 75, and 2.0 and 2.2% in generation 175. Thus, both symmetric and asymmetric dispersal scenarios suggest that Idh heterozygotes were highly overrepresented in the natural (hybrid) sample while Aat heterozygotes were found at frequencies slight under or at those expected during random mating and dispersal.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

The two Lycian salamanders L. antalyana and L. billae meet and reproduce in a narrow contact zone in southern Turkey. The contact area was made up of a small L. antalyana neighbourhood (sample 23) from which two steep linear clines produced abrupt introgression into a L. billae back land (Fig. 2). Clinal variation of nuclear alleles and mtDNA were coincident and concordant. The contact centre was situated south of the village Hurma (sample 23) where orange groves within the last 60 years (c. 20–30 generations) have expanded the habitat of L. antalyana to the hill area of L. billae. The contact zones were about 400–500 m wide. An alternative contact centre, the village Gökdere, is considerably older but here no parental L. antalyana were found on the alternative side. Although hybridization may occur where populations meet naturally, the present hybrid zone was probably set off by human activity. The area showed typical hybrid zone characteristics, including nuclear–nuclear linkage disequilibria, cytonuclear disequilibria and deviations from HW proportions. We further found introgression in an area southeast of the primary hybrid zone. This area was unexpected and the number of individuals analyzed from here are too few to make generalizations about the level of introgression or the cause of species contact.

The steepness and concertedness of the clines came from inclusion of parent individuals on both sides of the cline. In no sample did we observe both parent species. This and the lack of F1 equivalents in the hybrid zone points towards L. antalyana displacing L. billae and limited reproductive interaction between parental individuals. F1's may be lacking either due to selection against F1 hybrids or because they are rarely formed. If the latter explanation is correct, presence of F2 individuals implies that F1's may backcross readily. However, the backcross probability might be influenced highly by genotypic interactions. The magnitude of locus-specific introgression differed significantly between the two nuclear loci Idh and Aat. Cytonuclear disequilibrium at Aat, but not at Idh, and an intra-locality deficit of Aat heterozygotes, but not of Idh, indicate different genotypic behaviours and resulted in L. antalyanaIdh introgressing further than Aat into L. billae. Still, heterozygotes at both loci were lacking in the total gene pool within the contact area, raising the question arises whether Idh is more common than expected or Aat less common than expected. The simulation results suggested that Idh heterozygotes are more common than expected, rather than Aat are heterozygotes under-represented. Hence there may be positive selection for Idh in backcrosses as well as selection against heterozygote Aat (which were lacking in hybrids). The simulations were hampered by not knowing the exact initial parental frequencies upon contact and species-specific dispersal. However, the simulations showed that varying initial parental frequencies and dispersal rates changed only the magnitude of introgression but not the skewed genotype distributions, which implies that selection is needed in future simulations to fit the observed distribution. This corroborates the finding that the actual genotype differences possibly have evolved far faster in nature than indications gained by simulations (20–30 generations vs. 75–175 generations as fitted by the parental genotype distribution).

The fact that alleles of the less common L. antalyana move into L. billae populations and the observation of negative cytonuclear Aat-D2 in L. billae imply further that L. billae genes are not crossing the hybrid barrier to the same extent as L. antalyana. Thus not only is L. antalyana expanding its range but it may also be selectively replacing L. billae aided by selection against cytonuclear interactions on a L. billae background. As mentioned above, the genotypic patterns suggest selection against hybrid Aat genotypes but positive selection for Idh. Differential selection should make the Idh cline wider. It is not unlikely that the introgressed genomes will be stable at least for Idh. Hybrid genotypes may in some cases have higher fitness components than parental ones. Once the F1 barrier is crossed, positive selection may increase the frequency of L. antalyanaIdh. In hybrids of Bombina bombina and Bombina variegata, Nürnberger et al. (1995) hypothesized that rapid metamorphosis – relative to parental individuals – may be influenced by selection for fast development in dry habitats. A combination of hybrid characters could hereby become established.

Asmussen et al. (1987) gave hypotheses explaining the origin of cytonuclear disequilibria in a hybrid population (which were not definitive). Hypothesis 2 (D2 = 0 and D = D1 = −D3 ≠ 0) fits exactly the pattern for Idh and suggests random mating and no directionality to mating. In contrast, the pattern for Aat, where D ≠ 0, D1 ≠ 0, D2 ≠ 0 and −D3 ≠ 0, corresponds to scenario HNR which may indicate nonrandom mating, directionality to mating and a fairly young system. The two complete opposite explanations for the behaviour of Idh and Aat may stem either from differential selection or sex-specific mating behaviour. Genetic drift seems unlikely for causing the different cytonuclear disequilibria (e.g. Latta et al., 2001) because HW proportions support a greater lack of Aat heterozygotes, F1 equivalents were missing and, given fixed parental multilocus genotypes at species contact, the expectation of equal frequencies of BB/AB and AB/BB (shown in simulations) was not observed. In sex-specific gene flow, permanent nonzero cytonuclear disequilibrium can be maintained with an epistatic interaction between a single nuclear locus (e.g. Aat vs. Idh) and cytoplasmic gene that involves specifically the nuclear gene responsible in mating preference. However, when the mating preference involves an epistatic interaction between cytotype and the multilocus nuclear genotype characteristic of the pure parental species, all cytonuclear disequilibria break down (Arnold et al., 1988). Asmussen et al. (1989) analysed the cytonuclear pattern of two species of Hyla tree frogs and suggested that, due to nearly equal migration rates, significant D2 (i.e. heterozygote association) was caused by asymmetrical mating behaviour where Hyla cinerea females are more attracted to conspecific males than are Hyla gratiosa females.

Two findings challenge sex-specific gene flow as the reason for cytonuclear disequilibrium in Lyciasalamandra. The first is intra-specific differentiation estimates in L. billae that indicate that males contribute more to gene flow than females. The mtDNA differentiation estimate was six times higher than the nuclear estimate, which, under equilibrium conditions, indicates that male gene flow predominates (Birky et al., 1983,1989). (An estimate was not possible for L. antalyana but own field observations indicate that males move through the habitat on humid nights in great numbers.) Thus, nuclear introgression should proceed faster than mtDNA, but was not the case. The second reason why sex-specific gene flow does not explain cytonuclear disequilibrium is because it broke down when analyzing only hybrids; this is not expected if gene flow is sex-specific. For example, Shoemaker et al. (1996) found cytonuclear disequilibrium in fire ants because adult parental individuals continuously move into the hybrid zone. However, the cytonuclear disequilibrium also broke down when only hybrids were considered. This was caused by a marked reduction of parental-like hybrids in areas where the similar parental species was common. This could owe to reduced fitness of the hybrids caused by extrinsic selection due to competition with parental species.

Based on the considerations above, it is more likely that the Lyciasalamandra cytonuclear disequilibrium is influenced by F1 incompatibility. Data from the primary hybrid zone showed that both parental mtDNA types and nuclear loci introgress. Thus, hybrid formation involves females of both species. Interestingly, none of the five adult hybrid males had heterozygotic Aat while four of 15 females did. Also, calculating the total number of pure individuals sampled from the hybrid areas, the relationship between female : male : juvenile : subadult is 61 : 48 : 47 : 19 (63 pure individuals and 4 hybrids were not sexed). Considering the relationship relative to the number of females, this becomes 1 : 0.79 : 0.77 : 0.31. In the hybrid zone the relationship was 1 : 0.33 : 0.73 : 0.27. Thus, only males were missing. Neither of these two examples were significant (P > 0.20) but combined they point towards selection against male hybrids. Unfortunately we do not know which sex is functionally heterogametic in Lyciasalamandra as there are no visible sex chromosomes. There is no rule about the heterogametic sex in salamanders where both females and males can be heterogametic, even within the same genus (reviewed by Duellman & Trueb, 1994).

The findings from L. antalyana and L. billae provide evidence for different forces working simultaneously at species’ transitions. Locus-specific selection and uni-directional range expansion both distort symmetrical introgression. Symmetrical introgression may further be complicated by sex-specific hybrid mortality. The locus-specific asymmetry can have a positive effect of hybridization by increasing biological diversity because one genome does not out-compete the other, but incorporates foreign genes into novel genomic environments and forges new specific trajectories. This may explain the great diversity of morphological variants in Lyciasalamandra. Teasing apart these processes, between species-pairs and within species, will shed light on the jagged process of diversification. The findings indirectly show how habitat alterations (be they man-made or not) may influence the fate of species. Particularly, a comparison between the young hybridization studied here with what seems an old ‘natural’ hybrid area in the mountain ranges (locations 4 and 5) will help differentiate the selection processes at work.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix

We thank Olaf Godmann, Sebastian Steinfartz and Mehmet Öz for field assistance. All samples were collected with permission of the Turkish government. We thank two anonymous reviewers for helpful suggestions that greatly improved the manuscript.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix
  • Asmussen, M.J., Arnold, J. & Avise, J.C. 1987. Definition and properties of disequlibrium statistics for associations between nuclear and cytoplasmic genotypes. Genetics 115: 755768.
  • Asmussen, M.J., Arnold, J. & Avise, J.C. 1989. The effects of assortative mating and migration on cytonuclear associations in hybrid zones. Genetics 122: 923934.
  • Asmussen, M.A. & Basten, C.J. 1996. Constraints and normalized measures for cytonuclear disequilibria. Heredity 76: 207214.
  • Arnold, J., Asmussen, M.A. & Avise, J.C. 1988. An epistatic mating system model can produce permanent cytonuclear disequilibrium in a hybrid zone. Proc. Natl. Acad. Sci. USA 85: 18931896.
  • Arnold, M.L. 1997. Natural Hybridization and Evolution. Oxford University Press, New York.
  • Barton, N.H. & Hewitt, G.M. 1985. Analysis of hybrid zones. Ann. Rev. Ecol. Syst. 16: 113148.
  • Basten, C.J. & Asmussen, M.A. 1997. The exact test for cytonuclear disequilibrium. Genetics 146: 11651171.
  • Birky, C.W., Maruyama, T. & Fuerst, P. 1983. An approach to population and evolutionary genetic theory for genes in mitochondria and chloroplasts, and some results. Genetics 103: 513527.
  • Birky, C.W., Fuerst, P. & Maruyama, T. 1989. Organelle gene diversity under migration, mutation, and drift: equilibrium expectations, approach to equilibrium, effects of heteroplasmic cells, and comparisons to nuclear markers. Genetics 121: 613627.
  • Bogart, J.P. 2003. Genetics and systematics of hybrid species. In: Reproductive Biology and Phylogeny of Urodela (D. M.Sever, ed.), pp. 109134. Science Publishers Inc., Enfield (NH) USA.
  • Clarke, K.E., Rinderer, T.E., Frank, P., Quezada-Euán, J.G. & Oldroyd, B.P. 2002. The africanization of honeybees (Apis mellifera L.) of the Yucatan: a study of a massive hybridization event across time. Evolution 56: 14621474.
  • Coyne, J.A. & Orr, H.A. 2004. Speciation. Sinauer Associates, Sunderland, MA.
  • Crochet, P.A., Chen, J.J.Z., Pons, J.M., Lebreton, J.D., Hebert, P.D.N. & Bonhomme, F. 2003. Genetic differentiation at nuclear and mitochondrial loci among large white-headed gulls: sex-biased interspecific gene flow? Evolution 57: 28652878.
  • Dasmahapatra, K.K., Blum, M.J., Aiello, A., Hackwell, S., Davies, N., Bermingham, E.P. & Mallet, J. 2002. Inferences from a rapidly moving hybrid zone. Evolution 56: 741753.
  • Della Torre, A., Merzagora, L., Powell, J.R. & Coluzzi, M. 1997. Selective introgression of paracentric inversions between two sibling species of the Anopheles gambiae complex. Genetics 146: 239244.
  • Duellman, W.E. & Trueb, L. 1994. Biology of Amphibians. The John Hopkins University Press, Baltimore.
  • Ennos, R.A. 1994. Estimating the relative rates of pollen and seed migration among plant populations. Heredity 72: 250259.
  • Green, D.M. & Parent, C. 2003. Variable and asymmetric introgression in a hybrid zone in the toads, Bufo americanus and B. fowleri. Copeia 1: 3443.
  • Griebeler, E.M., Müller, J.C. & Seitz, A. in press. Spatial genetic patterns generated by two admixing genetic lineages: a simulation study. Conservation Genetics.
  • Haldane, J.B.S. 1922. Sex ratio and unisexual sterility in hybrid animals. J Genet 12: 101109.
  • Helbig, A.J., Salomon, M., Bensch, S. & Siebold, I. 2001. Male-biased gene flow across an avian hybrid zone: evidence from mitochondrial and microsatellite DNA. J. Evol. Biol. 14: 277287.
  • Johannesen, B. 2004. Evolution mitochondrialer Insertionen und ihre Aussagekraft für die phylogenetische Rekonstruktion der Gattung Lyciasalamandra (Amphibia, Urodela). M.Sc. Thesis, University of Mainz, Germany.
  • Klewen, R. 1991. Die Landsalamander Europas, Teil I. 2nd edn. Die Neue Brehm-Bücherei 584, Wittenberg Lutherstadt.
  • Latta, R.G., Linhart, Y.B. & Mitton, J.B. 2001. Cytonuclear disequilibrium and genetic drift in a natural population of Pondarosa pine. Genetics 158: 843850.
  • McCauley, D.E. 1994. Contrasting the distribution of chloroplast DNA and allozyme polymorphism among local populations of Silene alba: implications for studies of gene flow in plants. Proc. Natl. Acad. Sci. USA 91: 81278131.
  • Nürnberger, B., Barton, N., MacCallum, C., Gilchrist, J. & Appleby, M. 1995. Natural selection on quantitative traits in the Bombina hybrid zone. Evolution 49: 12241238.
  • Quennell, A.M. 1984. The western Arabia rift system. In: The Geological Evolution of the Eastern Mediterranean (J. E.Dixon & A. H. F.Robertson, eds), pp. 775778. Geol. Soc. Spec. Publ. No. 17. Blackwell Science, Oxford.
  • Raymond, M., & Rousset, F. 2000. GENEPOP (Version 3.2). Distributed by the authors. Institut des Sciences de l'Evolution, Universite de Montpellier II, France.
  • Riley, S.P.D., Shaffer, H.B., Voss, S.R. & Fitzpatric, B.M. 2003. Hybridization between a rare, native tiger salamander (Ambystoma californiense) and its introduced congener. Ecol. Appl. 13: 12631275.
  • Shoemaker, D.D., Ross, K.G. & Arnold, L. 1996. Genetic structure and evolution of a fire ant hybrid zone. Evolution 50: 19581976.
  • Steinfartz, S. & Mutz, T. 1998.Mertensiella luschani (Steindacher, 1891): Lykischer Salamander, Kleinasiatischer Salamander. In: Handbuch der Reptilien und Amphibien Europas, Vol 4/I: Schanzlurche (K.Grossenbacher & B.Thiesmeier, eds), pp. 367397. AULA Wiesbaden.
  • Steinfartz, S., Veith, M. & Tautz, D. 2000. Mitochondrial sequence analysis of Salamandra taxa suggests old splits of major lineages and postglacial recolonizations of Central Europe from distinct source populations of Salamandra salamandra. Mol. Ecol. 9: 397410.
  • Sweigart, L. & Willis, J.H. 2003. Patterns of nucleotide diversity in two species of Mimulus are affected by mating system and asymmetric introgression. Evolution 57: 24902506.
  • Szymura, J.M. & Barton, N.H. 1986. Genetic analysis of hybrid zone between the fire bellied toads, Bombina bombina and B. variegate near Cracow in southern Poland. Evolution 40: 1141a1159.
  • Szymura, J.M., Uzzell, T. & Spolsky, C. 2000. Mitochondrial DNA variation in the hybridizing fire-bellied toads, Bombina bombina and B. variegata. Mol. Ecol. 9: 891899.
  • Tarkhnishvili, D., Hille, A. & Böhme, W. 2001. Humid forest refugia, speciation and secondary introgression between evolutionary lineages: differentiation in a Near Eastern brown frog, Rana macrocnemis. Biol. J. Linn. Soc. 74: 141156.
  • Veith, M., Baran, I., Godmann, O., Kiefer, A., Öz, M. & Tunç, M.R. 2001. A revision of population designation and geographic distribution of the Lycian salamander Mertensiella luschani (Steindacher, 1891). Zool. Middle East 22: 6782.
  • Veith, M. & Steinfartz, S. 2004. When non-monophyly results in taxonomic consequences – the case of Mertensiella within the Salamandridae (Amphibia: Urodela). Salamandra 40: 780.
  • Weisrock, D.W., Macey, J.R., Ugurtas, I.H., Larson, A. & Papenfuss, T.J. 2001. Molecular phylogenetics and historical biogeography among Salamandrids of the ‘‘true’’ salamander clade: rapid branching of numerous highly divergent lineages in Mertensiella luschani associated with the rise of Anatolia. Mol. Phyl. Evol. 18: 434448.


  1. Top of page
  2. Abstract
  3. Introduction
  4. Method and materials
  5. Results
  6. Discussion
  7. Acknowledgments
  8. References
  9. Appendix
Table AppendixA..  Allozyme allele frequencies and mtDNA haplotype frequencies of Lyciasalamandra billae and Lyciasalamandra antalyana from a hybrid zone southwest of Antalya, Turkey. Sample locations are found in Fig. 1 and Table 1. The number of investigated individuals (n) differs slightly between the two genetic markers because one round of DNA extraction was faulty and amplification failed in roughly half of these individuals. A and B alleles at Aat-1 and Idh-1 signify species-specific L. antalyana and L. billae alleles, respectively.
Proportion hybrids 0000.200.8000.14000000.100.1300.320.500000.140.280.4300