Introduction

The study of nuclear gene genealogies in natural populations of nonmodel organisms is still in its infancy and only a few reports are available to date (see, for example, Antunes et al, 2002; Broughton and Harrison, 2003). Nevertheless, we expect that the generalized application of those genealogies will provide novel insights into the evolutionary history of populations, especially when developed in the framework of well-established mtDNA phylogeographic scenarios. This is because nuclear and mtDNA markers offer contrasting molecular and population properties that include bi-parental and uni-parental inheritance, recombining and nonrecombining histories and a 4:1 ratio of expected coalescence times. These contrasting properties will be most valuable when employed to investigate complex patterns of genetic diversity like those exhibited by hybrid zones where differential rates of sex-related dispersal, intricate dynamics of nuclear admixture and fluctuations in effective population sizes still constitute enormous challenges.

In the last few decades, the accumulation of empirical data documenting the patterns of variation in hybrid zones was paralleled by the development of a strong theoretical background (Barton and Hewitt, 1985, 1989; Harrison, 1993). This is clearly reflected in our current understanding of a number of fundamental issues about hybrid zones that include their origin, maintenance and fate (Barton and Hewitt, 1985), their role as promoters of evolutionary novelties (Arnold, 1997) and the evaluation of the balance between exogenous and endogenous selection processes and implications for speciation (Jiggins and Mallet, 2000). A remarkable pattern associated with most well-studied hybrid zones is the occurrence of novel alleles that are generally absent from the parental populations outside of the hybrid zone (Barton and Hewitt, 1985). When reviewing the subject, Woodruff (1989) considered the previously used expression ‘rare allele phenomenon’ as inappropriate and introduced the term hybrizyme to describe unexpected allelic electromorphs observed in hybrid zones. Among the genetic mechanisms suggested as an explanation for this observation were point mutations (Woodruff, 1989), intragenic recombination (Sage and Selander, 1979; Golding and Strobeck, 1983; Woodruff, 1989), gene conversion (Hillis et al, 1991) or transposition (Kidwell, 1990), but only detailed DNA sequencing of such new variants could clarify this phenomenon. With the advent of polymerase chain reaction (PCR) and other corresponding technical advances, most of the difficulties associated with the sequencing of nuclear genes and the definition of haplotypes have been removed. It is thus surprising that only a few reports investigating the origin of hybrizymes are available (Bradley et al, 1993; Hoffmann and Brown, 1995), eventually encouraging Schilthuizen et al (2001) to state that ‘sequencing studies have shown that hybrizyme alleles are caused by simple point mutations, and cannot be explained by either recombination or transposition’.

The age of most hybrid zones in central and northern Europe is probably very recent and corresponds to the postglacial colonization of biota (Hewitt, 1999, 2000). Various sources of evidence suggest that this process was rapid, concordant with the typically lower genetic diversity of northern populations (Hewitt, 1996). In contrast, populations and species have persisted in southern European refugia over several ice ages, leading to the accumulation of much higher levels of genetic diversity. In the Iberian Peninsula, the molecular signature of deep genetic subdivisions, elevated haplotype richness, and older contact zones have recently been described in a variety of organisms including salamanders (Alexandrino et al, 2000), lizards (Paulo et al, 2001) and rabbits (Branco et al, 2002). Results described for the Schreiber's Green lizard (Lacerta schreiberi), an Iberian endemic confined to the stream and river margins of typical Atlantic habitats of the peninsula, are particularly interesting. At the mitochondrial DNA (mtDNA) level, Paulo et al (2001) described the occurrence of two highly divergent and allopatric lineages located in the Portuguese and Spanish sides of the Iberian Central System, whereas the analysis of protein polymorphism suggested that those same population groups have been in contact for a long time (Godinho et al, 2001, 2003).

In this work, we combine a fine-scale analysis of the mtDNA contact zone with a global analysis of the patterns of variation observed at the nuclear β-fibrinogen intron 7 (β-fibint7). We have been able to (i) contrast the phylogeographical patterns observed for mtDNA and the nuclear β-fibint7, (ii) more precisely define the geographical location of the hybrid zone between the two divergent groups of lizards, (iii) show evidence of extensive admixture, (iv) describe the occurrence of unique haplotypes restricted to hybrid populations at the β-fibint7 locus, and (v) suggest that genetic mechanisms other than simple point mutations are likely involved in the generation of those haplotypes.

Materials and methods

Sample collection

We analysed a total of 406 individual tissue samples collected in 19 populations covering the whole distribution area of L. schreiberi. Five out of these 19 populations were sampled along an east–west transect that crosses the putative mtDNA contact zone (Figure 1).

Figure 1
figure 1

(a) Distribution area of L. schreiberi in the Iberian Peninsula (shaded) and sampling locations: 1 – Asturias, 2 – Ancares, 3 – Ferrol, 4 – Gerês, 5 – Gião, 6 – Montemuro, 7 – C. Rainha, 8 – Montejunto, 9 – Cercal, 10 – Monchique, 11 – S. Mamede, 17 – Guadarrama, 18 – Guadalupe, 19 – Toledo. (b) Detailed map with indication of relief and rivers along a west–east transect crossing the putative mtDNA contact zone. The approximate frequencies of the two mtDNA lineages and of the β-fibint7 alleles are represented by pie charts and bars, respectively.

Full size image

Amplification and RFLP analysis of mtDNA

An 850 bp sized fragment of mtDNA, including most of the cytochrome b (cyt b) gene, was PCR amplified with Taq polymerase according to manufacturer's instructions (Promega). The reaction mixture included 0.1 U of Taq, 3 mM MgCl2 and approximately 50 ng of genomic DNA. PCR was carried out in a Biometra T3 thermocycler over 35 cycles with annealing temperature set at 50°C. The primers used were gluDG and cb3, described in Palumbi (1996). The Restriction Map subroutine of BioEdit 5.0.9 software (Hall, 1999) was used to locate the site gain or loss for each endonuclease profile and the diagnostic profiles of NlaIII and NlaIV endonucleases were selected to distinguish between the two main mtDNA lineages described in Paulo et al (2001) as the coastal and inland clades. In our work, these two main clades were named A and B, respectively. Additionally, the two sublineages described for each main clade by Paulo et al (2001) as coastal northern/southern and inland northern/southern and referred to in this study as A1/A2 and B1/B2, respectively, were distinguished using the BseGI and the NlaIV endonucleases. RFLP patterns were visualized under UV light after electrophoretic separation on 2% agarose gels.

Amplification, SSCP analysis and sequencing of β-fibrinogen

The entire β-fibint7 was PCR-amplified using FIB-B17U and FIB-B17L primers as described by Prychitko and Moore (1997), but increasing the annealing temperature to 57°C. A preliminary survey of nucleotide polymorphism was obtained by sequencing β-fibint7 (788 bp) in a total of 12 individuals sampled in different localities. This strategy allowed the identification of a cluster of polymorphic positions close to the 5′ end of β-fibint7 covering less than 130 bp and the subsequent design of PCR-SSCP primers (Fib7LsF 5′-CTA GTC ATA CCC AAA TGT G-3′ (forward) and Fib7LsR 5′-CTA ATT CAG GGG GAG CTA-3′ (reverse)) for an extended population analysis that included a total of 406 samples. The PCR mixture was made for a total volume of 10 μl using 0.1 U of Taq polymerase (Promega), 3 mM MgCl2 and approximately 25 ng of DNA. Amplifications were done in a Biometra T3 thermocycler over 32 cycles with annealing temperature set at 54°C and a fragment of 180 bp was obtained. Discrimination of the SSCP conformers was made in a 12% polyacrylamide gel (39:1 acrylamide: methylbisacrylamide) with 0.5 × TBE buffer on a vertical electrophoresis system (BIORAD Protean II). The electrophoresis was performed at a constant voltage of 250 V and constant temperature of 12°C for 15 h. Results were visualized by silver staining. In addition, a sample of three to six homozygous individuals for each SSCP allele collected in different populations was also sequenced for the entire β-fibint7 region in order to evaluate the consistency of SSCP results and explore further variation. In the case of low frequency alleles, the identification of double-peaks in electrophoregrams from sequenced heterozygous individuals allowed the use of ARMS-primers (Amplification Refractory Mutation System, Newton et al, 1989) to unambiguously identify haplotypes. Sequencing followed the ABI Prism BigDye Terminator Cycle sequencing protocol in an ABI PRISM 310 Genetic Analyser. The sequences of β-fibint7 alleles in L. schreiberi were deposited in GenBank (Accession numbers DQ097101, DQ097102 and DQ299919 to DQ299932). The separation of haplotypes B and C was difficult in routine SSCP gels and confirmation of their identity was additionally obtained by examining RFLP patterns with the use of the endonuclease StuI. Four closely related Lacerta species (L. viridis, L. agilis, L. strigata and L. trilineata) were used as outgroups (Godinho et al, 2005).

Data analysis

Allele frequencies and measures of genetic variation such as heterozygosity and number of alleles were calculated using the Genetix software, version 4.01 (Belkhir et al, 2000). Estimates of nucleotide diversity (π) were obtained with the DnaSP 3.51 software (Rozas and Rozas, 1999) while β-fibint7 nucleotide contents and the transition/tranversion ratio were calculated using the MEGA 2.1 software (Kumar et al, 2000). The Network 3.1 software (Röhl, 2000) was used to construct reduced median-joining networks for both the β-fibint7 SSCP alleles and the entire intron 7 fragments. We further explored the information content of the β-fibint7 genealogy by splitting the total gene tree according to the four sublineages A1, A2, B1 and B2. These networks were used to (i) investigate phylogenetic relationships between haplotypes; (ii) evaluate the concordance between mtDNA and β-fibint7 phylogeographic patterns; and (iii) evaluate the effectiveness of the SSCP fragment in capturing the evolutionary trajectory of the entire β-fibint7. The minimum number of recombination events in the history of the sampled sequences was estimated using the four-gamete test (Hudson and Kaplan, 1985) as implemented in the DnaSP 3.51 software (Rozas and Rozas, 1999).

Results

The distribution of mtDNA lineages in the hybrid zone

A total of 107 lizard samples collected in five locations along an east–west transect across the putative hybrid zone of the two major divergent population groups were screened with NlaIII and NlaIV for the generation of distinct cleavage patterns distinguishing the mtDNA lineages A and B. On the western side, the Portuguese populations of Lousã (n=10) and Estrela (n=28) were found to be fixed for mtDNA lineage A, while on the eastern side the Spanish Central System populations of Béjar (n=22) and Gata (n=21) showed only lineage B. In contrast, the intermediate location of Malcata exhibited both lineages, although only one individual out of 25 was typed as mtDNA B (Figure 1). These results are essentially in accordance with those reported by Paulo et al (2001), but increased sample sizes allowed the identification of an admixed population that reveals the approximate location of the mtDNA secondary contact zone. In addition, two to three samples from the remaining 14 populations, together with those in the above-mentioned west–east transect, were digested with BseGI and NlaIV for the identification of sublineages (Table 1).

Table 1 Number of individuals sampled (N), mtDNA lineage, haplotypic frequencies, number of haplotypes per population, expected heterozygosity (He) and nucleotide diversity (π) at the nuclear β-fibint7 locus. Boxes highlight the highest values observed for the genetic diversity parameters
Full size table

Polymorphism of β-fibint7

A 180 bp-sized fragment of β-fibint7 was resolved generating a total of 10 haplotypes using SSCP (Figure 2). With the exception of Béjar, the other populations included in the west–east transect revealed high levels of expected heterozygosity and nucleotide diversity when compared to the remaining 14 lizard populations (Table 1), which is in agreement with our hypothesis of a well-defined hybrid zone in this geographical location. Haplotype frequencies showed considerable variation across populations, resulting in a high amount of population structure (Fst=0.39). However, the Fst value was considerably lower (Fst=0.25) when only the five populations included in the west–east hybrid zone were analysed, suggesting the occurrence of restricted gene flow between the two highly divergent population groups. Haplotypes A and H were typical of northwestern Iberia and the Spanish Central System, respectively, and showed opposite clinal gradients across the hybrid zone, while haplotype B was less frequent but widespread. Western isolated populations exhibited high frequencies of haplotypes E and F, while more central isolates showed the occurrence of haplotype J. Central populations of the main distribution area were characterized by haplotypes C and D, whereas two rare and private haplotypes, G and I, were found in the populations of Malcata and Gata, respectively.

Figure 2
figure 2

Separation of allelic variants of the β-fibint7 locus (180 bp) by SSCP analysis in 12% polyacrylamide gels. Alleles B and C were further confirmed with the use of the endonuclease StuI.

Full size image

The sequence of the whole β-fibint7 locus in L. schreiberi provided a 788 bp fragment, including 71 bp of exon 7 and 103 bp of exon 8. A total of 20 segregating sites (eight in the SSCP fragment) were found, from which nine (five in the SSCP fragment) were parsimony informative (Table 2). Two 1 bp deletions and one insertion of a trinucleotide repeat were found. A transition/transversion substitution ratio of 1.7 and an A+T biased content of 63% were observed in β-fibint7 sequences, which were in agreement with values reported for this intron in other species (Prychitko and Moore, 1997, 2000; Johnson and Clayton, 2000).

Table 2 (a) Variable positions found at the β-fibint7 locus in L. schreiberi defining a total of 16 haplotypes. The fragment (180 bp) corresponding to the SSCP analysis is shaded. The ancestral haplotype was derived from the analysis of β-fibint7 sequences in four closely related Lacerta species. Position 1 corresponds to position 8636 of human β-fibrinogen. (b) Reconstruction of the hypothetical recombination events originating haplotypes G and J
Full size table

Phylogenetic relationships of SSCP β-fibint7 haplotypes

A sample of 3–12 chromosomes for each SSCP allele was sequenced and confirmed that all mutations were detected with the technique described. This result thus highlights the high-resolving power of SSCP in the detection of DNA sequence variation (Orti et al, 1997; Sunnucks et al, 2000). Haplotype genealogy was displayed with a reduced median network and inference of the ancestral haplotype was made from β-fibint7 sequences obtained in four other closely related Lacerta species. Haplotype C, present in only a few populations, contains the ancestral state at each of the eight polymorphic sites in the β-fibint7 locus and was, accordingly, identified as the root of the network. The two major haplotypes A and H are one and two mutations distant from the root, respectively, and most of the remnant haplotypes are easily connected in the network (Figure 3a). The remarkable exception is the position of both low frequency haplotypes G and J that are at the origin of the two loops in the network thus suggesting the occurrence of recurrent mutation or recombination. We will argue below that these rare haplotypes are the product of intragenic recombination and not the result of recurrent mutation.

Figure 3
figure 3

(a) Reduced median-joining network representing the phylogenetic relationships between the 10 SSCP β-fibint7 haplotypes. Circle size is proportional to the frequency of each haplotype in the total sample. The shaded circle indicates the root of the network, black points represent potential intermediates and dashed lines indicate loops that may result from recombination. (b) Parsimony networks for four groups of amalgamated populations according to the mtDNA sublineages (A1, A2, B1 and B2). The population of Malcata was included in group A1. Circle size is proportional to the frequency of each haplotype in the respective amalgamated sample.

Full size image

When the haplotype network is split according to the essentially allopatric mtDNA lineages and sublineages, we gain additional information on the history of the β-fibint7 polymorphism. The populations described by Paulo et al (2001) as exhibiting the ancestral mtDNA sub-lineages A1 and B1 are characterized by the presence in high frequency of β-fibint7 haplotypes A and H, respectively, and also show a diverse array of other less frequent haplotypes (Figure 3b). In contrast, populations possessing the more derived mtDNA sub-lineages A2 and B2 show much lower levels of diversity, a result that is consistent with a more recent colonization of the central and southern areas of the Iberian Peninsula, as described by Paulo et al (2001). In the western areas of Iberia, the mtDNA sublineage A2 is concordant with the occurrence of the derived β-fibint7 haplotypes E and F, which mark the southwestern range expansion of L. schreiberi. A similar phenomenon is not apparent in the Spanish central isolates that are characterized by the mtDNA sublineage B2 but that are almost fixed for the β-fibint7 haplotype A.

Phylogenetic relationships of extended β-fibint7 haplotypes and evidence for recombination

When the same sample of chromosomes was sequenced for the entire β-fibint7, a similar network was obtained (Figure 4), clearly confirming the usefulness of our SSCP approach in determining the major haplotypes and capturing their respective phylogenetic relationships. Interestingly, the basic difference between the two networks is the increase in the number of variants derived from the two major β-fibint7 haplotypes, A and H. The complete β-fibint7 network additionally reveals that (i) at this level of resolution, haplotype C is no longer the root of the network due to a T → A transversion at position 306, (ii) the ancestral haplotype was not found in any of the studied populations, and (iii) the derivation of haplotypes E and F is further evidenced by the G → A transition and T → A transversion at positions 398 and 492, respectively.

Figure 4
figure 4

Reduced median-joining network representing the phylogenetic relationships between the 16 extended β-fibint7 haplotypes. The ancestral allele is indicated. Numbers correspond to single point mutation events. Dashed lines indicate loops that may result from recombination. The inset represents the estimated network after the removal of recombinant haplotypes. The root is indicated by an arrow, and black points show potential intermediates.

Full size image

The additional information provided by the extended β-fibint7 haplotypes supports the interpretation that recombination is at the origin of the rare haplotypes G and J. Accordingly, two different criteria indicate that the possibility of recurrent mutation is highly unlikely. First, both haplotypes occur in populations that show evidence of extensive admixture between the two major population groups of L. schreiberi. Haplotype G is private from the Malcata population, which is precisely the only location where both mtDNA lineages A and B were found in sympatry (Table 1 and Figure 2). A close examination of the characteristics of haplotype G (Table 2b) clearly suggests that this haplotype is a mosaic of the more common parental haplotypes A and H, indicating that recombination is the probable mechanism involved in its origin. Haplotype J was found in the central isolate of S. Mamede, a peculiar population that exhibits mtDNA lineage A but clusters with populations from the Spanish Central System when polymorphic proteins are analysed (Godinho et al, 2003), thus suggesting another situation of extensive admixture between two divergent groups. Although our evidence is not as clear as for haplotype G, we suggest that haplotype J is the result of a recombination event between the two more common haplotypes found in S. Mamede, A and B (Table 2b). Second, none of the mutations implied in the origin of these haplotypes can be attributed to the hypermutability of CpG dinucleotides. In vertebrate genomes, CpG dinucleotides are known to be hotspots for nucleotide substitutions because of their high content of 5-methylcytosine (5mC). Methylation of cytosine results in a high level of mutation due to the propensity of 5mC to undergo deamination to form thymine, which explains the high frequency of C-to-T and G-to-A transitions (reviewed in Cooper and Krawezak, 1993). Finally, the four-gamete test identified four pairs of sites with the four gametic types (Table 2), of which two (positions 126, 248 and 126, 532) are involved in the recombination between haplotypes A and H, and the other two (positions 126, 232 and 126, 306) in the recombination between haplotypes A and B.

Discussion

Comparing mtDNA and β-fibint7 phylogeographies in Lacerta schreiberi

The previous phylogeographic analysis of the mtDNA molecule in Lacerta schreiberi suggested that the major differentiation process in two main lineages was initiated in the late Pliocene. Populations of this species would have persisted through the Pleistocene in allopatric refugia, but the well known climatic fluctuations of this period additionally caused the emergence of allopatric mtDNA sublineages (Paulo et al, 2001). Although Hudson and Turelli (2003) convincingly argued that the use of mtDNA divergence to predict monophyly of nuclear loci is not reliable, we believe that the remarkable phylogeographic structure of this lizard offers an excellent opportunity to evaluate the utility of nuclear gene genealogies in obtaining a more complete picture of the evolutionary history of the species. The results of our study confirmed this prediction and showed that the global analysis of variation at the β-fibint7 locus provides additional and important information on the processes that shaped the present-day genetic architecture of L. schreiberi.

The Pliocene divergence between the two major population groups located in coastal and inland Iberia left a clear molecular signature in the β-fibint7 network by separating haplotypes A and H with four mutations (Figure 4). These haplotypes represent the two main clades of that network and while haplotype A exhibited high frequencies in all populations showing mtDNA sublineage A1, haplotype H characterized populations possessing sublineage B1. Moreover, both haplotypes showed opposite clinal variation along the west–east transect studied in the Iberian Central System, indicating the occurrence of gene flow between two formerly isolated populations. Paulo et al (2002) further suggested that central and southern isolates were the result of old expansions to the south during a glacial age followed by a recent contraction and the extinction of intermediate populations. In coastal regions, the populations were characterized by the derived mtDNA sublineage A2 (Table 1) and concordantly showed the β-fibint7 haplotypes E and F, which are clearly derived from the more common and ancestral haplotype A. Surprisingly, however, the inland isolated populations of Guadalupe and Toledo that exhibited the mtDNA sublineage B2 were characterized by the β-fibint7 haplotypes A and J. Given that the divergence between mtDNA sublineages B1 and B2 is almost twice the divergence between A1 and A2, we could expect the occurrence of concordantly derived β-fibint7 haplotypes in those populations. We suggest that their absence may be explained by the occurrence of recent gene flow from western populations, eventually driving the extinction of those putative haplotypes. This hypothesis is supported by (i) the presence of haplotype J, which is likely to have originated in the more western mountain of S. Mamede (see below); (ii) additional nuclear data indicating that recent gene flow is stronger from the west than from the geographically closer populations located in the mountains of Gredos and Guadarrama in the east (Godinho, 2004); and (iii) environmental and climatic modelling of the species distribution showing a possible corridor linking S. Mamede with Guadalupe and Toledo (Godinho, 2004).

Evidence for recombination at the β-fibint7 in the hybrid zone

In recent years, sequencing of a couple of alleles that are exclusively found in hybrid zones suggested that they may have originated by simple point mutations instead of the more likely process of recombination (Bradley et al, 1993; Hoffmann and Brown, 1995). Our results, however, suggest that the two haplotypes G and J, detected in admixed populations, are likely the product of recombination between the more common and widespread parental haplotypes A/H and A/B, respectively.

The population of Malcata is notable because it corresponds to the single geographical location where the two deeply divergent mtDNA lineages were found in sympatry (Figure 1). The highest values of genetic diversity at the β-fibint7 locus (number of haplotypes, expected heterozygosity and nucleotide diversity; see Table 1) and also from other nuclear data (Godinho, 2004; Godinho et al, 2006) give additional evidence that this population is close to the centre of a hybrid zone. This scenario of extensive admixture between two groups of populations that show evidence of long-term historical isolation was ultimately responsible for the contact of the distinct β-fibint7 haplotypes A and H, creating an opportunity for the generation of recombinant haplotypes. In the Malcata population, the private haplotype G is easily viewed as a mosaic of the more common haplotypes A and H (Table 2). Additionally, the absence of hotspot mutation motifs together with the equitable base composition and nucleotide substitution probabilities (see Prychitko and Moore, 2000) decrease the likelihood of recurrent mutations and strengthens the confidence in our interpretation. Essentially, the same criteria support the hypothesis that haplotype J originated in the population of S. Mamede by a recombination event between the more common haplotypes A and B. However, more data are necessary to draw a firm conclusion in this respect.

The utility of nuclear genealogies in well-defined phylogeographic contexts

The extraordinary explosion of phylogeographical studies since the seminal paper of Avise et al (1979) has made an enormous contribution to our understanding of evolution and created a new research discipline (Avise, 2000). However, the fact that most phylogeographical studies are based solely on mtDNA, which possesses well-recognized limitations (Zhang and Hewitt, 2003), has hampered further developments in the field. Nevertheless, it is now clear that recent progress in molecular genetic techniques combined with the development of new statistical tools will lead to an increased use of nuclear DNA polymorphisms. Our study of the nuclear β-fibint7 locus in a lizard species for which a well-established phylogeographical scenario was previously described highlights the advantages of those types of polymorphisms and offers some prospects for future work.

To our knowledge, this is the first report using the β-fibint7 nuclear marker for an intraspecific study in a vertebrate species. Although this intron has been previously characterized as having a substitution rate approximately four to six times slower than mtDNA (Johnson and Clayton, 2000; Weibel and Moore, 2002), we found a total of 20 segregating sites (Table 2) that allowed the simple construction of highly informative haplotype networks (Figures 3 and 4). Together with the four-gamete test, these networks additionally revealed that two out of ten different β-fibint7 haplotypes are probably recombinants, shedding new light on the population history and dynamics of the nuclear genome of L. schreiberi.

Recently, Broughton and Harrison (2003) examined four nuclear gene genealogies in three closely related but morphologically, behaviourally and ecologically distinct Gryllus species, and concluded that those genealogies were of limited phylogeographical use but provided important insights into the historical, demographic and selective forces that shaped North American field crickets. In this study, we offer a somewhat different perspective in which the genealogy of the nuclear β-fibint7 combined with mtDNA information considerably improves our understanding of the evolutionary history of a species exhibiting a pronounced genetic structure and, simultaneously, a remarkable uniformity with respect to morphology, behaviour and ecology.

Implications for the study of hybrid zones

Most well-known hybrid zones are probably very recent and resulted from the post-glacial colonization of previously unsuitable territories by expanding genomes that were formerly constrained to low latitude refugia (Hewitt, 2000). Consequently, we may consider that these hybrid zones are very limited windows into the evolutionary processes for which relevant timescales are orders of magnitude higher. However, it is becoming clear that much older hybrid zones may exist in low latitude refugia like the Iberian Peninsula. Apart from the lizard species that is the focus of this study, recent examples of deep lineage divergence followed by secondary contacts have been described in a variety of organisms (Alexandrino et al, 2000; Branco et al, 2002; García-París et al, 2003; Sequeira et al, 2005). This is certainly a consequence of the extreme topographical heterogeneity and associated habitat diversity of the Iberian Peninsula, resulting in multiple sub-refugia within the refugium.

Our data suggest that divergent lineages of L. schreiberi have been contracting and expanding repeatedly during the glacial and interglacial periods of the Pleistocene, thus creating ample opportunities for ancient admixture and the establishment of an old hybrid zone (Godinho, 2004). We additionally provide compelling evidence in favour of an important role of recombination in the generation of new alleles in hybrid zones, thus contradicting the recent statement by Schilthuizen et al (2001) but clearly confirming expectations based on both empirical and theoretical evidence (Golding and Strobeck, 1983; Woodruff, 1989). We note, however, that the generation of hybrizymes through intronic recombination depends on the accumulation of nonsynonymous divergence in the flanking exons. While future sequencing of β-fibrinogen exons are necessary to effectively demonstrate the occurrence of hybrizyme products at this locus, we anticipate that the high degree of population subdivision observed in a variety of organisms in low latitude refugia has likely resulted in divergent protein products that can be at the origin of novel variants in admixed populations. Finally, our results support a high frequency of occurrence of recombination even in very short nuclear DNA sequences, as recently described by Ibrahim et al (2002), and suggest that the multiple pulses of expansion and admixture that characterized the Pleistocene era in Southern Europe will be most fruitfully investigated through careful studies incorporating the detection of recombinant haplotypes, eventually leading to an accurate perception of the dates of contacts between hybridizing populations (Baird, 1995). Our demonstration of using the properties of recombination to have more insights into past processes of population contacts and admixture gain relevance in light of the recently described lack of phylogeographical structure in European mammals before the last ice age (Hofreiter et al, 2004). These authors proposed that the deep mtDNA divergence times observed in many different species do not reflect long separations of populations but simply the beginning of the differentiation process that happened frequently before the Pleistocene. Subsequent processes of expansion, recolonization and admixture are erased at each new ice age because mtDNA clades remain fixed at the different refugia (Figure 5 of Hofreiter et al, 2004) and, at the same time, are unable to retain the memory, through recombination, of periods of extensive mixing.

Taken together with present and previously published mtDNA data, our β-fibint7 study in the Schreiber's green lizard anticipates the invaluable utility of autosomal DNA markers in future characterizations of the admixture dynamics of divergent genomes, ultimately helping in the elucidation of biological phenomena occurring in hybrid zones that are not revealed by haploid markers like mtDNA or the mammalian Y-chromosome.