Introduction

In eukaryotes, transmembrane proteins imbedded in the inner membrane of the mitochondria are the sites of energy conversion and cellular respiration. These proteins consist of subunits of the electron transport system (ETS) and are encoded by either nuclear DNA (nDNA) or mitochondrial DNA (mtDNA). The products of both genomes and their precise interactions are essential for aerobic metabolism. A plethora of disorders in humans linked to sequence variation in mtDNA or mitochondrially expressed nDNA illustrates the evolutionary constraint imposed on genes of both genomes (Larsson and Clayton 1995; Schapira 2006). Furthermore, specific mtDNA haplotypes have been shown to influence individual fitness in humans, mice, Drosophila, and copepods (reviewed by Ballard and Whitlock 2004; Gemmell et al. 2004; Ballard and Rand 2005). It has been estimated that the ETS is comprised of approximately 90 different respiratory chain subunits, the majority of which are transcribed in the nucleus, assembled in the cytoplasm, and transported into the mitochondrion (Larsson and Clayton 1995). The mitogenome encodes only 13 of these protein subunits and 24 RNA components (12S and 16S rRNA molecules, 22 tRNAs) necessary for mitochondrial protein synthesis. Despite this small proportion, the high mutation rate of mtDNA may drive the evolution of mitochondrial-nuclear interactions by contributing ‘new’ amino acids for the scrutiny of natural selection (Wu et al. 2000; Schmidt et al. 2001; Willet and Burton 2004). These results have prompted a reevaluation of the neutrality of mtDNA and show the need for further research in this area.

The physical and functional contiguity of the nuclear and mitochondrial proteins in the enzymatic complexes of the ETS leads to coadaptation of the two genomes (Blier et al. 2001; Ballard and Whitlock 2004; Ballard and Rand 2005). Moreover, interactions between nuclear-encoded mtRNA polymerases and their promoters located in mtDNA control regions have been shown to be species-specific (Gaspari et al. 2004). The polymerases, and possibly additional transcription factors, may not bind to the mtDNA as precisely if presented with a divergent, foreign promoter (Burton et al. 2006). Hence, mitochondrial genes cannot be exempt from the epistasis evolving between geographically delimited sets of nuclear alleles. The disassociation of these interactions in interpopulation or interspecies hybrids may cause inferior mitochondrial function (McKenzie et al. 2003; Ellison and Burton 2006; Burton et al. 2006).

The hybridizing European fire-bellied and yellow-bellied toads, Bombina bombina and B. variegata, have evolved in different environments (Szymura 1993). The lowland form, B. bombina, inhabits larger, permanent ponds, while the mountain type, B. variegata, prefers smaller transient ponds and puddles. The toads are divergent at the molecular level, and each exhibits a suite of phenotypic traits in morphology, life history, ecology, and behavior, regarded as adaptations to their respective environments (Szymura 1993). Hybrid zones between the toads occur in areas of altitudinal transition along their parapatric contact in central Europe (Szymura and Barton 1991; Yanchukov et al. 2006). Selection against hybrids results from intrinsic hybrid dysfunction (Szymura and Barton 1986; Kruuk et al. 1999). Despite extensive hybridization, no introgression of mtDNA has taken place outside of the narrow zones (Szymura et al. 2000; Yanchukov et al. 2006; Hofman et al. 2007; Hofman and Szymura 2007). Clines in mtDNA are no wider, and are sometimes narrower, than clines at multiple unlinked allozyme loci and morphological traits (Hofman and Szymura 2007). Postzygotic reproductive barriers between B. bombina and B. variegata have thus been attributed to substantial genetic divergence in the genomes of the toads and the genetic incompatibility that ensues (Szymura et al. 1985; Nürnberger et al. 2003; Spolsky et al. 2006), with differences in mtDNA hypothesized to contribute to the negative epistasis in recombined hybrids (Hofman and Szymura 2007).

In this paper we describe the complete mtDNA sequences of the two hybridizing European species, B. bombina and B. variegata. We quantify divergence between the two genomes and their functional components, including the distinctive organization of their control regions (Spolsky et al. 2006), and study the distribution of substitutions across the mitogenomes. Using a phylogeny based on four species of Bombina and five outgroup species, we identify regions in the mtDNA of B. bombina and B. variegata carrying nonsynonymous and radical amino acid substitutions and, conversely, conservative nucleotide and amino acid domains. Apart from augmenting our understanding of postzygotic reproductive barriers in Bombina, our results should be useful for elucidating the mechanisms of molecular evolution of mtDNA at low levels of divergence, and contribute to the growing number of described mitogenomes of the most basal group of anuran amphibians (archeobatrachians), whose mtDNA is known to evolve particularly slowly (Igawa et al. 2008).

Materials and Methods

Laboratory Methods

A single individual of B. variegata from Radziszów near Kraków, Poland (49°57′ N, 19°48′ E) and a B. bombina from Tyniec, Poland (50°02′ N, 19°48′ E) were used as sources of mtDNA. The mtDNAs were purified as by Szymura et al. (1985, 2000). Aliquots of B. variegata mtDNA were partially digested with either HindIII or XbaI and subsequently cloned into compatible pUC18 sites (Boehringer Mannheim). MtDNA was radioactively labeled and used as a probe to screen for clones containing mtDNA fragments. The fragments were sized to compare with mitochondrial maps (Spolsky et al. 2006) and inserts were sequenced with universal M13 (forward and reverse) primers and additional walking primers. Bombina specific primers amplifying overlapping fragments spanning the entire mtDNA molecule on both strands were then designed on the basis of the cloned B. variegata mtDNA for both PCR and sequencing in B. bombina (PCR/sequencing strategy and primers listed in Electronic Supplementary Material, S1 and S2). Two sequencing chemistries were used as appropriate for a Beckman CEQ sequencer (B. variegata mtDNA clones) and Applied Biosystem Analyzer (most B. bombina mtDNA amplicons).

Molecular Analyses

The mtDNA sequences were edited and assembled using SeqMan version 5.05. After alignment of the entire genomes in ClustalW, the two rRNAs, 22 tRNAs, and 13 protein coding genes were identified by comparisons to the mtDNA genomes of B. orientalis (AY585338 [San Mauro et al. 2004a]) and B. fortinuptialis (AY458591 [Zhang et al. 2005], incorrectly listed in the NCBI database as B. bombina [cf. Spolsky et al. 2006]). Molecular evolutionary analyses including base composition, sequence divergence, numbers of transitions and transversions, and numbers of synonymous and nonsynonymous substitutions were carried out using MEGA v.3.1 (Kumar et al. 2004) and DnaSP (Rozas et al. 2003). We also performed a sliding window analysis of nucleotide divergence in the four Bombina mtDNA genomes. We identified radical replacement changes in the protein coding genes, i.e. the replacement of one amino acid by another that belongs to a different class defined by the properties of lateral chains (e.g., Doiron et al. 2002). Because all mitochondrially encoded peptides exhibit transmembrane domains, we used a mutation matrix specifically designed for transmembrane proteins for scoring (Jones et al. 1994). We then mapped the autapomorphic and radical amino acid substitutions onto the putative topologies of the mtDNA-encoded proteins by applying the hidden Markov method for sequence feature prediction implemented in PolyPhobius (Käll et al. 2005) and TMHMM (Krogh et al. 2001). Because homologues are likely to share the same secondary structures, PolyPhobius also uses existing information from homologous sequences in protein databases.

Phylogenetic Analyses

Besides the Bombina genomes listed above, we also used the closest living relatives of Bombina from which entire mitochondrial genomes are available, i.e., Discoglossus galganoi (AY585339) and Alytes obstetricans pertinax (AY585337). Mitochondrial genomes from three other representatives of Archaeobatrachia (Frost et al. 2006; Roelants et al. 2007) were used as more distant outgroups: Ascaphus truei (AJ871087), Xenopus tropicalis (AY789013), and Pelobates cultripes (AJ871086). Phylogenetic analyses were performed on 12 mitochondrial H-strand protein-coding genes. The L-strand-encoded ND6 gene was excluded because of differences in base composition. The mitochondrial protein coding sequences of these 12 genes were translated, concatenated to form a single chain, and aligned using ClustalW as implemented in MEGA v.3.1.

Multiple substitutions at the same site confound phylogenetic inference and obscure evolutionary relationships. In order to avoid this complication, we first plotted the ratio of transitions to transversions to assess substitution saturation (Felsenstein 2004) against divergence times in Bombina and related genera. The divergence times, taken from Roelants et al. (2007) and Fromhage et al. (2004), are meant to show the general trends in our analysis, so measures of uncertainty were not considered. The analysis (Fig. 1) clearly showed that nucleotide substitutions were not saturated in Bombina (Fig. 1, points A and B). We next tested whether particular codon positions contained multiple substitutions at the same sites by plotting the corrected sequence divergence measured as maximum likelihood (ML) distances against the numbers of transitions and transversions at first, second, and third codon positions (S3–S6). The general time-reversible (GTR + I + G; I = 0.42, G = 0.81) model of DNA evolution was chosen as the best-fitting model following the Akaike information criterion and hierarchical likelihood ratio tests in Modeltest v3.7 (Posada and Crandall 1998). Transitions at third codon positions were saturated (S3, S4), so we only used first and second codon positions in the phylogenetic analysis (S5). For the amino acid data, uncorrected pairwise distances were plotted against ML distances incorporating the mtREV + I + G (I = 0.56, G = 0.30) model of protein evolution, which best fit the amino acid data according to the AIC criterion in ProtTest v1.3 (Abascal et al. 2005). The amino acid sequences were not saturated (S6).

Fig. 1
figure 1

Ratio of number of transitions (T S ) to number of transversions (T V ) between (A) B. bombina and B. variegata; (B) B. orientalis and (B. bombina + B. variegata); (C) Alytes and Discoglossus; (D) (Alytes + Discoglossus) and Bombina; (E) Discoglossidae and (Xenopus + Pelobates); and (F) Ascaphus and (Discoglossidae + Xenopus + Pelobates). Divergence times are estimates without confidence intervals from Roelants et al. (2007) for points A–E and from Fromhage et al. (2004) for point F

Full size image

The final alignments used for the phylogenetic analyses for the nine basal anuran taxa encompassed either 7184 nucleotide positions or 3592 amino acid residues. All alignments are available from the corresponding author. The best-fitting model of DNA evolution for the first and second codon position dataset (unpartitioned) was the GTR + I + G model (I = 0.6, G = 1.36). In a separate analysis, we partitioned the mtDNA genes into three functional groups (protein coding genes, rRNAs, tRNAs) and assigned models to each through Modeltest. For the rRNAs and for the tRNAs, Modeltest suggested values of GTR + G (G = 0.32) and GTR + I + G (I = 0.42, G = 1.18), respectively. ML analyses of the unpartitioned first and second codon position dataset and the amino acid dataset were conducted in PHYML v2.4.4 (Guindon and Gascuel 2003), starting with the BIONJ trees. Nonparametric bootstrapping was used to test the reliabilities of the ML trees (1000 pseudoreplicates). Bayesian inference (BI) of the phylogeny for all three datasets was carried out using MrBayes v3.04b (Huelsenbeck and Ronquist 2001) with the previously specified models of sequence evolution; all other priors were set to default values. One cold and three heated chains were run for 5 million generations, with trees sampled every 100 generations. Generations sampled before the chains reached stationarity (20,000), as judged by examining the log-likelihoods of the cold chains and plots of the generation vs. log-likelihood values of the data, were discarded as burn-in. At least two independent runs were carried out for each dataset in the Bayesian analyses.

Results

Genome Organization and Nucleotide Composition

The complete mtDNA genomes have been deposited at the GenBank database under accession numbers EU115993 (B. bombina) and AY971143 (B. variegata). Within the genus, the B. variegata mtDNA genome was the largest, at 18,551 bp, followed by B. fortinuptialis (17,575 bp), B. orientalis (17,173 bp), and B. bombina (17,154 bp). The differences in mtDNA genome size can be accounted for by size variation in the control region. The mtDNA genomes of Bombina exhibit the organization and gene content of the canonical mtDNA genome of other vertebrates (Table 1; Fig. 3) (Boore 1999). The gene arrangement and start and stop codons in the Bombina mtDNA genomes are identical in all studied species (Table 1). The putative origin of L-strand replication (OL) in the Bombina mitogenomes was located within the WANCY tRNA cluster, between the tRNAAsn and the tRNACys genes. The lengths of the protein coding genes are also similar, with the exception of the third codon in ND1 (ACT, threonine in B. variegata), which is absent in B. bombina. The nucleotide frequencies of the L-strands are very similar in all four Bombina species (Table 1) and skewed against guanine because of bias against this nucleotide at the second and third codon positions. The transition-to-transversion ratio was quite variable among genes, but always skewed toward transitions (Table 2).

Table 1 Lengths, in base pairs, of structural features in the Bombina mtDNA genomes and nucleotide compositions of mitochondrial L-strands
Full size table
Table 2 Nucleotide substitutions in mitochondrial genes of B. bombina and B. variegata
Full size table

Control Region Structure

We compared the control regions in all four sequenced Bombina mtDNAs (Fig. 2A). The size of the control region varies nearly twofold in Bombina, consisting of 3072 bp in B. variegata, 2373 bp in B. orientalis, 1990 bp in B. fortinuptialis, and 1675 bp in B. bombina (Fig. 2A). The B. fortinuptialis control region is incomplete because of an unsequenced stretch at the 3′-end of the first tandem repeat (Zhang et al. 2005). The latter, termed LV1 by Spolsky et al. (2006), consists of 4–12 serial repeats 70–77 bp long (Fig. 2B). A highly variable nucleotide sequence is found downstream of LV1, with approximately the first 150 bp representing incomplete repeat units. Three conserved sequence blocks (CSB1–3) are localized downstream of LV1. The region –180 bp relative to CSB1 is alignable in all Bombina mitogenomes studied and encompasses a poly(T) block, the pyrimidine-rich region, PP-1, of San Mauro et al. (2004a). Between CSB-1 and CSB-2 lies a poly-C block (PP-2 [cf. San Mauro et al. 2004a]). PP-1 and PP-2 may be involved in H-strand replication (San Mauro et al. 2004b). The second repeat motif (LV2; Fig. 2C) is present in only three of the species, being secondarily lost in B. bombina. This 62- to 66-bp motif, repeated up to 13 times in B. variegata, is characterized by a conserved 5′-end, AT and GT dinucleotide repeats, and a poly(T) tail present in the three species. Additional incomplete repeats flank LV2, and a homologous but nonrepetitive sequence is found at the 3′-end of the control region in B. bombina.

Fig. 2
figure 2

A comparison of control region structure in European and East Asian Bombina species. (A) Large vertical bars represent tandemly arranged repeat units within LV1 and LV2. Filled circles represent conserved sequence blocks (CSB); small vertical bars among the CSBs are pyrimidine-rich regions (PP-1 and PP-2). Length is proportional to the number of nucleotides in each species. The question mark in B. fortinuptialis marks an unsequenced fragment after four repeats. (B) Alignment of the repeat unit LV1. Horizontal bar denotes the putative termination associated sequence (TAS). (C) Alignment of the repeat unit LV2. Conservative sites are in boldface

Full size image

Divergence in Bombina mtDNA

We calculated the pairwise divergence among the four Bombina mitochondrial genomes, taking into account all rRNAs, tRNAs, and protein coding DNA sequences and, in separate calculations, divergence in amino acid sequences, tRNAs and rRNAs (Table 3). Uncorrected nucleotide divergence between the hybridizing B. bombina and B. variegata was lowest, at 8.1% or 8.7% (Kimura two-parameter distance; K2P). The highest divergence was between the mtDNA of the two East Asian species, B. orientalis and B. fortinuptialis, at 14.3% (16.2% K2P).

Table 3 Pairwise comparisons of divergence in functional units among Bombina mtDNA genomes
Full size table

The various functional regions of Bombina mtDNA have evolved at different rates. The rRNA and tRNA genes were least diverged (Fig. 3), implying functional constraint. Genes encoding ETS complexes III, IV, and V always showed the lowest number of nonsynonymous substitutions (Tables 2 and 3). In contrast, the protein coding genes of ETS complex I were the most variable, particularly certain regions of ND2, ND5, and ND4 (Fig. 3), and had the highest numbers of nonsynonymous substitutions (Table 2).

Fig. 3
figure 3

Sliding window analysis of uncorrected percentage divergence in four Bombina mt genomes (window size, 100 bp; step size, 25 bp). Genes encoded by the L-strand—ND6, tRNA-Q, tRNA-A, tRNA-N, tRNA-C, tRNA-Y, tRNA-S(UCN), tRNA-E, and tRNA-P are underlined. Vertical gray bars highlight the tRNA genes (tRNA letter abbreviations above bars). Tick marks along the X axis delimit the rRNA and protein-coding genes

Full size image

Phylogenetic Relationships

Both ML and BI analyses reconstructed similar tree topologies for nucleotide and amino acid datasets (Fig. 4). The partitioned nucleotide dataset gave essentially the same results as the unpartitioned nucleotide dataset; we therefore report only the latter. Most nodes were well supported, and the intrageneric relationships of Bombina were fully resolved. The sister group relationship between B. bombina and B. variegata was confirmed (Hofman et al. 2007). The East Asian B. orientalis formed a well-supported clade with the European species, to the exclusion of B. fortinuptialis, which has a basal placement in the genus Bombina. The only discrepancy observed between the nucleotide and the amino acid analyses was the placement of Alytes. In the amino acid analyses, an Alytes-Bombina clade was recovered, while in the first and second codon position dataset an Alytes-Discoglossus clade was apparent. The position of Ascaphus was left unresolved by the amino acid data.

Fig. 4
figure 4

Consensus phylogram reconstructed for the first and second codon position dataset of 12 H-strand mitochondrial protein-coding genes using maximum likelihood. Numbers represent PHYML bootstrap support for the first and second codon position dataset and the amino acid dataset (first line) and Bayesian posterior probabilities for the two datasets (second line). Only support values >50% are shown. Branch length is proportional to the number of substitutions per site

Full size image

Patterns of Substitution in Mitochondrial Protein Coding Genes of European Bombina

The number of nucleotide substitutions calculated for all 13 mitochondrial protein coding genes between B. bombina and B. variegata was 1130 (Table 2), of which 1032 (91%) were silent. The uncorrected percent divergence for particular genes varied between 7.5% (COIII) and 11.5% (ND2). The percentage of synonymous substitutions in synonymous sites between the mitochondrial protein coding genes of B. bombina and B. variegata varied between 30% and 40% in particular genes, much higher than the percentage of nonsynonymous substitutions in nonsynonymous sites, which ranged from 0 to 2.4% (Table 2). The number of amino acid differences in the mtDNA-encoded peptides between B. bombina and B. variegata ranged from 0 in COII to 29 in ND5 (Table 2) and was positively correlated with the number of base pairs in each gene (R 2 = 0.425, p = 0.015). More revealing, the dN/dS ratios (Table 2), a measure of the stringency of structural and functional constraint acting on the protein coding genes, shows that the strength of selection against nonsynonymous change in the Bombina protein coding genes is strong. According to this measure, the protein coding genes can be arranged in the following order, from most conservative to most variable: COII < COI < ND4L << COIII < Cytb < ND1 < ATP8 < ATP6 < ND4 < ND6 << ND2 < ND5 < ND3. Of the total of 98 nonsynonymous substitutions between the mtDNA of B. bombina and B. variegata, the majority (80) were found in ETS complex I of the electron transport system (Table 3). Only seven nonsynonymous substitutions were found in ETS complex III, three in complex IV, and eight in complex V (Table 3).

Functional Significance of the Amino Acid Substitutions

On the basis of the phylogenetic relationships among the mtDNA genomes of the basal anurans (Fig. 4), we inferred the derived state (either synapomorphic, autapomorphic, or homoplasious) for all replacement substitutions observed between the mitochondrially encoded peptides of the two European species and the two East Asian congeners used as outgroups. A total of 32 and 33 autapomorphic amino acid replacements were identified in the mitochondrially encoded peptides of the B. bombina and B. variegata lineages, respectively. Character assignment was ambiguous in 19 additional homoplasious amino acid replacements in which each of the outgroup species shared an amino acid with one of the ingroup species. Of the autapomorphic replacement changes that had a negative mutability score according to a matrix for transmembrane proteins, B. bombina possessed three, while B. variegata had six (Table 4). ETS complex I harbored all but one (found in ETS complex V) of these nonconservative amino acid replacements (Table 4).

Table 4 Autopomorphic and radical amino acid (aa) substitutions in B. variegata and B. bombina and their location on the peptide chain and topology of the protein
Full size table

The predicted topology of the four proteins (ATP6, ND2, ND4, and ND5) in Bombina in which radical amino acid replacements occurred is shown in Supplementary Material S7–S10. TMHMM and PolyPhobius predictions were generally congruent, although the algorithm used in the latter application consistently identified a higher number of transmembrane helices. In addition, the orientation of the proteins was often contradictory. The locations of the autapomorphic, radical amino acid substitutions in the mitochondrial peptides of B. bombina and B. variegata are given in Table 3 and Supplementary Material S7–S10.

The ATP6 subunit is part of the F0 segment of the ATP synthase and takes part in the channeling of protons through the inner membrane of the mitochondrion. The functionally important amino acids of the ATP6 subunit, including the polar residues comprising the proton channel and residues in contact with nuclear subunits, are quite conserved between humans and E. coli (Schon et al. 2001) and Bombina (data not shown). The L-P substitution in B. variegata, however, is clearly located outside of the transmembrane helices and does not participate in nuclear-mitochondrial interactions (Table 3, Supplementary Material S7 [cf. Schon et al. 2001]), thus it probably does not affect the functioning of the ATP synthase.

The remaining radical amino acid replacements occurred in the ND proteins of ETS complex I. This large, L-shaped enzyme is composed of 46 subunits, 7 of which are hydrophobic central subunits encoded by mtDNA (Brandt 2006). The molecular structure of this complex remains unresolved, and how amino acid variation in specific subunits affects the functioning of the complex is poorly known. The transmembrane domain predictors applied to the Bombina complex I proteins (Supplementary Material S8–S10) were often at odds as to which amino acid residues constitute transmembrane helices. Nonetheless, the domains in which the radical replacement changes occurred were always well defined and, thus, can be used as rough estimates of their location. A single radical substitution, invoking a change from a hydrophobic to a hydrophilic amino acid (L-S), was detected in a transmembrane helix of ND5 in B. variegata. Other amino acid differences were located in domains putatively outside of the transmembrane segments. Three involved proline, an amino acid with a cyclic structure known to occur in the turns of polypeptide chains, and could therefore alter the secondary structure of these surface domains. A total of three radical substitutions occurred in a large (ca. 76 residues, positions 502–586; Supplementary Material S10) extramembrane domain of the C-terminus of ND5.

Discussion

Despite the high degree of nucleotide divergence between the Bombina mitochondrial genomes (up to 14% between the two East Asian species) and their closest relatives (26% between Discoglossus/Bombina and Alytes/Bombina), the mitochondrial genomes of these amphibians retained identical gene order and content and similar nucleotide composition over the immense expanse of evolutionary time since their divergence (during the late Triassic/early Jurassic according to recent molecular estimates [Roelants et al. 2007; San Mauro et al. 2004a]). The most divergent mtDNA of the four species compared was that of B. fortinuptialis, which, together with B. maxima and B. microdeladigitora, belongs to the group of large-bodied Bombina species of Southeast Asia (Grobina). The ancient split between the large-bodied and the small-bodied Bombina species (B. bombina, B. variegata, and B. orientalis) is supported by the grouping of B. orientalis with the European species in our mtDNA phylogeny, as well as by karyotypic differences, i.e., the large-bodied species bear a diploid chromosome complement of 2N = 28, while the small-bodied species contain 2N = 24 (Szymura and Passakas-Szymczak 1988). B. orientalis, despite having a distribution that is geographically proximate to the range of B. fortinuptialis, is phylogenetically closer to the Western Palearctic Bombina species than to its East Asian relatives. Analysis by Yu et al. (2007) of DNA sequences from several mitochondrial genes for Grobina and several European and Asian Bombina species also supports this grouping.

The average nucleotide divergence in mtDNA between the two hybridizing species, B. bombina and B. variegata, was 8.1 (8.7% K2P). This figure is congruent with earlier estimates based on restriction enzymes (9.4% ± 1% [Szymura et al. 1985], 6.0–8.1% [Szymura et al. 2000]). In a different study, sequence divergence in Cytb haplotypes between 62 individuals of both species amounted to 9.2% ± 0.2% K2P (Hofman and Szymura 2007), a value slightly above the average for the entire mitogenome. The nuclear sequence divergence between B. bombina and B. variegata is lower than this and has been estimated as 0–8% (Nürnberger et al. 2003) for several homologous, putatively noncoding loci, 3.7% (3.8% K2P) for a partial coding fragment of histone H3a (Frost et al. 2006; GenBank accession numbers DQ284275 and DQ284274 for B. bombina and B. variegata, respectively), and 0.03% (0.03% K2P) in exon 1 of the rhodopsin gene (Frost et al. 2006; DQ283920 and DQ283919 for B. bombina and B. variegata, respectively). The divergence in 29 allozyme loci between B. bombina and representatives of the major subgroups of B. variegata, measured as Nei’s D distance, ranged between 0.37 and 0.59 (Szymura 1993).

The Bombina control region contains highly conserved regions, such as the CSBs and pyrimidine-rich regions, and also length-variable regions (LV1 and LV2) comprised of tandemly arranged repeat units. The conserved sequence blocks (CSBs) and pyrimidine-rich stretches (PPs) possibly play a role in transcription of mtDNA-encoded proteins and H-strand replication in amphibians (San Mauro et al. 2004b). In Xenopus laevis, bidirectional promotors and transcription factor binding sites have been localized upstream of CSB2 and CSB3 (Antoshechkin and Bogenhagen 1995). These regions of high sequence conservation are preceded by the first repeat region, LV1, consisting of repeat units that are alignable in Bombina (Fig. 2B) and, also, among archeobatrachians (San Mauro et al. 2004a), suggesting homology. This sequence similarity indicates functional constraint that may result from the presence of a putative termination-associated sequence (TAS) in the LV1 repeat unit. TASs terminate H-strand synthesis and thus produce the D-loop-containing form of mtDNA. The presence of repeats that contain TAS sequences and possibly other regulatory elements downstream of LV1 raises the possibility of a fine-tuning of mitochondrial metabolism (through the regulation of mtDNA replication/transcription) by the number of the repeat units in the control regions, as suggested for lagomorph mtDNA (Casane et al. 1997). In contrast, the LV2 unit, tandemly repeated at the 3′ end of the Bombina control region, is restricted to the genus Bombina and has been secondarily lost in B. bombina. Apparently, the LV2 region in Bombina is under less selective constraint than LV1, as evidenced by the lack of sequence similarity to other archeobatrachian 3′ repeat motifs (San Mauro et al. 2004a), its absence in B. bombina, and no known regulatory function.

The varying dN/dS ratios observed among the protein coding mitochondrial genes in Bombina suggest generally high but, nonetheless, unequal levels of stringency of structural and functional constraint. Overall, our results are consistent with previous studies examining substitution patterns among mitochondrial protein coding genes, in which genes encoding proteins of ETS complexes III, IV, and V were always much more conserved than those in ETS complex I (e.g., Pesole et al. 1999; Doiron et al. 2002). Exceptional are perhaps the high dN/dS ratio in ND3 and relatively low ratios for ATP6 and ATP8 in Bombina. Low ratios for ATP8 have also been documented in several fish species (Roques et al. 2006 and references therein).

Closer inspection of the species-specific, nonsynonymous substitutions between the hybridizing European Bombina species revealed that the majority can be considered functionally neutral, yet 9 of the 98 nonsynonymous substitutions between B. bombina and B. variegata may have functional consequences. All but one occur in ETS complex I, for which only a limited amount of structural information is available. Moreover, the radical replacement substitutions observed in ND4 and ND5 coincide with elevated K A /K S ratios. Both radical replacements and high dN/dS ratios have been regarded as the hallmarks of coadaptation between mitochondrial and nuclear ETS subunits (Wu et al. 2000; Schmidt et al. 2001; Doiron et al. 2002) or evidence for the action of positive selection (Mishmar et al. 2003; 2006). A possible alternative explanation is that the elevated rate of amino acid substitution results from relaxed selective constraint on sites at which the need for amino acid conservation is lower (Elson et al 2004; Ingman and Gyllensten 2007).

Mitochondrially encoded and nDNA-encoded proteins build functional units constituting the mitochondrial ETS, whereas species-specific promotor regions in mtDNA probably coevolve with nDNA-encoded peptides. Mutations in either genomic component influence the evolution of the other, producing coadapted complexes (Blier et al. 2001). These associations are particularly prone to disruption in species hybrids (Sackton et al. 2003; Zeyl et al. 2005; Ellison and Burton 2008). Despite ongoing hybridization in spatially and temporally variable hybrid zones, Hofman et al. (2007) have found no evidence for either past or present mtDNA introgression between European Bombina, which is consistent with the idea that later-generation hybrids are less fit because of cytonuclear incompatibility leading to the disruption of mitochondrial function (Burton et al. 2006). In this paper we have shown that there are ample differences between the mtDNA genomes of B. bombina and B. variegata that may have functional consequences affecting mitochondrial oxidative phosphorylation in hybrids. Further studies should incorporate experimental verification of our results through, e.g., in vitro assays of the efficiency of ETS complexes in hybrid and nonhybrid individuals.