Widespread unidirectional transfer of mitochondrial DNA: a case in western Palaearctic water frogs

Authors


Jörg Plötner, Museum für Naturkunde, Zoologie, Humboldt-Universität, Invalidenstrasse 43, 10115, Berlin, Germany.
Tel.: +49 30 20938508; fax: +49 30 20938528;
e-mail: joerg.ploetner@museum.hu-berlin.de

Abstract

Interspecies transfer of mitochondrial (mt) DNA is a common phenomenon in plants, invertebrates and vertebrates, normally linked with hybridization of closely related species in zones of sympatry or parapatry. In central Europe, in an area north of 48°N latitude and between 8° and 22°E longitude, western Palaearctic water frogs show massive unidirectional introgression of mtDNA: 33.7% of 407 Rana ridibunda possessed mtDNA specific for Rana lessonae. By contrast, no R. lessonae with R. ridibunda mtDNA was observed. That R. ridibunda with introgressed mitochondrial genomes were found exclusively within the range of the hybrid Rana esculenta and that most hybrids had lessonae mtDNA (90.4% of 335 individuals investigated) is evidence that R. esculenta serves as a vehicle for transfer of lessonae mtDNA into R. ridibunda. Such introgression has occurred several times independently. The abundance and wide distribution of individuals with introgressed mitochondrial genomes show that R. lessonae mt genomes work successfully in a R. ridibunda chromosomal background despite their high sequence divergence from R. ridibunda mtDNAs (14.2–15.2% in the ND2/ND3 genes). Greater effectiveness of enzymes encoded by R. lessonae mtDNA may be advantageous to individuals of R. ridibunda and probably R. esculenta in the northern parts of their ranges.

Introduction

Interspecies transfer of mitochondrial (mt) DNA has been detected in a variety of organisms, for example in flowering plants (e.g. Bergthorsson et al., 2003), invertebrates (e.g. Croucher et al., 2004; Abe et al., 2005; Bachtrog et al., 2006), and among vertebrates in fish (e.g. Yamada et al., 2001; Sullivan et al., 2004), amphibians (e.g. Babik et al., 2005), birds (e.g. Dabrowski et al., 2005) and mammals (e.g. Ruedi et al., 1997; Fredsted et al., 2006). In most cases, introgressed mtDNA was found in very closely related species (especially in sister species) and was directly linked to interspecies hybridization in zones of sympatry or parapatry. Such genome transfer is always mediated by fertile or at least partially fertile hybrids that transmit their maternal mtDNA to the paternal gene pool via backcrosses with males of the paternal parental species. In some populations, introduced foreign mtDNA can completely replace the resident mtDNA either through genetic drift or natural selection (Ballard & Whitlock, 2004). The detection and analysis of introgressive hybridization is not only important for taxonomic purposes and phylogenetic inference, it may also provide fundamental insights into evolutionary processes, especially for studying the role of nuclear/mitochondrial co-adaptation in speciation, including hybrid breakdown and reproductive isolation (e.g. Ballard & Whitlock, 2004; Burton et al., 2006).

The western Palaearctic water frog group offers significant opportunities to study these problems because it comprises several evolutionary species that are not reproductively isolated. There are large areas of sympatry in which fertile hybrids are formed (reviewed by Plötner, 2005). At least three of the known hybrid forms reproduce hemiclonally by a hybridogenetic (Schultz, 1969) gametogenesis: in their germ line one parental genome is excluded, whereas the remaining one is endoreduplicated prior to meiosis (reviewed by Graf & Polls Pelaz, 1989; Plötner, 2005).

The best studied hybrid form is Rana esculenta (genomic composition LR), which originated from natural crosses between Rana lessonae (LL) and Rana ridibunda (RR). In the germ line of this hybrid, one parental genome (usually L, sometimes R) is excluded and haploid gametes (ova or sperm) are produced that contain an intact, unrecombined genome of the other parental species (reviewed by Graf & Polls Pelaz, 1989; Plötner, 2005). Such hybrid lineages in most cases coexist with and reproductively depend on one parental species. Matings with that species replace the genome lost during gametogenesis and restore somatic hybridity in each hybrid generation. In the widespread and abundant lessonaeesculenta (L–E) system, in which both male and female hybrids occur, R. esculenta usually coexist, as sexual parasites, with R. lessonae and exclude the L genome prior to meiosis. In populations of the less frequent ridibundaesculenta (R–E) system, in which hybrids are only males, they live together with R. ridibunda and usually exclude the R genome (Uzzell & Berger, 1975; Uzzell et al., 1977; Günther & Plötner, 1988). In addition to diploid LR hybrids, some water frog populations, especially in north-eastern Germany and north-western Poland, contain triploid (3N) hybrids (Günther, 1975, 1983; Rybacki & Berger, 2001; Plötner, 2005) that normally have either two R. lessonae genomes and one R. ridibunda genome (genotype LLR) or one R. lessonae genome and two R. ridibunda genomes (genotype RRL). They usually exclude the rarer genome and form unrecombined haploid gametes (Günther et al., 1979; Günther, 1983; Berger & Günther, 1988; Vinogradov et al., 1991).

For size-related behavioural reasons, natural primary hybridizations between R. ridibunda and R. lessonae occur between the large females of R. ridibunda and the small males of R. lessonae rather than the reciprocal (Berger, 1970; Borkin et al., 1979). As in most vertebrates, frog mtDNA is inherited strictly from the mother, so all newly formed R. esculenta lineages contain R. ridibunda mtDNA. In the L–E population system, most matings take place between R. esculenta females and R. lessonae males, but successful matings occur also in the reciprocal combination (Blankenhorn, 1977; Lengagne et al., 2006). Each such mating between an R. lessonae female and an R. esculenta male generates R. esculenta lineages with R. lessonae mtDNA. In the absence of the parental species R. ridibunda, such a transfer of R. lessonae mtDNA into a hybridogenetic lineage is irreversible, and indeed, most hemiclonal hybrids of the L–E system contain mtDNA of their local host parental species rather than R. ridibunda mtDNA, a result of such occasional matings of host females with hybrid males (Spolsky & Uzzell, 1986). With repeated formation of R. esculenta lineages containing R. lessonae mtDNA and chance losses of lineages containing R. ridibunda mtDNA, all hybrid individuals in L–E system populations come to have R. lessonae mtDNA. When such hemiclonally reproducing hybrid lineages occur within the range of R. ridibunda, however, they provide a vehicle for interspecific introgression of R. lessonae mtDNA into R. ridibunda. Matings between female R. esculenta with R. lessonae mtDNA and male R. ridibunda (or R. esculenta) lead to R. ridibunda progeny of both sexes (or females only) with R. lessonae mtDNA (Spolsky & Uzzell, 1984; Hotz et al., 1992).

Three geographical ranges are relevant to the introgression of R. lessonae mtDNA into R. ridibunda: (1) the area of sympatry of R. ridibunda and R. lessonae, because it defines the region where primary hybridizations between them can occur. Such sympatry presently exists from northern Ukraine and western Russia to eastern Austria and central Germany (Günther, 1990, 1997a, b). (2) The area in which R. ridibunda genomes in hybrids between R. lessonae and R. ridibunda lead to germ line exclusion of the L genome and thus found hemiclonally reproducing hybrid lineages rather than nonhybridogenetic sterile F1 hybrids. Here, R. esculenta lineages can constitute vehicles for the introgression of R. lessonae mtDNA into R. ridibunda. This area currently comprises most parts of the sympatry region (central and eastern Europe), but excludes an eastern zone from northern Ukraine to the Danube delta in eastern Romania. In this zone and in more southern parts of the R. ridibunda range, the R. ridibunda genome does not induce hemiclonal gametogenesis in hybrids with R. lessonae (cf. Hotz et al., 1985; Bucci et al., 1990; Günther et al., 1991; Berger et al., 1994; H. Hotz, unpublished data). (3) Areas in which the sex ratio in R. esculenta hinders regular matings between female R. lessonae and male R. esculenta; such matings are necessary to transfer R. lessonae mtDNA into R. esculenta lineages and via these lineages into R. ridibunda. Whereas R. esculenta lineages in most parts of central Europe contain both sexes, usually with a slight excess of females, they consist almost exclusively (∼97%) of females in many populations of the Pannonian Basin (eastern Austria, northern Bosnia and Herzegovina, north-eastern Croatia, Hungary, northern Serbia, southern lowlands of Slovakia, north-eastern Slovenia, western Romania and Ukraine west of the Carpathian Mountains; Berger et al., 1988; Gubányi, 1992; Morozov-Leonov et al., 2003; H. Hotz, unpublished data) and in the Rostov Province of Russia (Borkin et al., 2006).

We examined mtDNA of water frogs from 194 localities throughout and beyond the sympatry area of R. ridibunda and R. lessonae, determining species origin and haplotype of mtDNA. We asked how often and where the introgression of R. lessonae mtDNA into R. ridibunda occurred, and addressed the geographic relation of the region of R. lessonae mtDNA introgressed into R. ridibunda with: (1) the sympatry area of R. ridibunda and R. lessonae; (2) the area in which hybrids reproduce hemiclonally; and (3) the area in which hybrids are almost exclusively females.

Material and methods

Sampling, taxon designation and ploidy determination

Localities of 407 R. ridibunda, 117 R. lessonae, 335 R. esculenta and two individuals of uncertain identification (R. lessonae or R. esculenta) are listed in Appendix S1 (Supplementary material online). We use the name R. ridibunda for individuals distributed in Europe, and include frogs currently known as Rana kurtmuelleri. Unpublished molecular data (J. Plötner, S.N. Litvinchuk & T. Ohst) suggest that individuals from the type locality of R. ridibunda (north coast of the Caspian Sea) may not be conspecific with individuals from Europe, in which case the oldest available name for the European frogs is Rana fortis Boulenger, 1884 (Dubois & Ohler, 1994). Except for two individuals from France (localities 31 and 33), for which we used DNA sequences for haplotype comparisons and sequence statistics, R. ridibunda from France and Switzerland were not considered because all lake frog populations in these countries trace back to individuals introduced not only from other European regions, but also from Anatolia (Grossenbacher, 1988; Pagano et al., 2003; J. Plötner & T. Ohst, unpublished results).

Specimens were assigned to taxa on the basis of their external morphology; Günther (1990) and Plötner (2005) described diagnostic characters. Ploidy was determined by measurements of erythrocytes based on the finding that triploid (3N) erythrocytes are on average 30% larger than diploid (2N) ones (Uzzell & Berger, 1975; Günther, 1977). The kind of mtDNA present in all individuals in our sample was determined by one or more methods: sequencing of ND2, ND3 and parts of 12S rRNA gene, endonuclease restriction of whole mtDNA, AluI restriction of a PCR-amplified segment of the ND2 gene. Our sequence data are supplemented by published restriction fragment analyses (Spolsky & Uzzell, 1984, 1986).

DNA extraction, PCR conditions, sequencing

Genomic DNA was isolated from blood, muscle and liver tissue using QIAmp blood and tissue kits (Qiagen GmbH, Hilden, Germany) according to the company’s protocols. Before DNA isolation, ethanol-preserved material was soaked in three changes of distilled water over a 48-h period. Double-stranded amplification and cycle sequencing of mtDNA (1038 bp of ND2, 340 bp of ND3, 374 bp of 12S rDNA) were carried out as described in Appendix S2 (Supplementary material online). The sequenced sites correspond to positions 6546–7583 (ND2), 12089–12428 (ND3) and 3156–3528 (12S) in the genome of Rana nigromaculata (AB043889, Sumida et al., 2001). Amplified DNA was purified using QIAquick PCR purification kits (Qiagen). A Big Dye Ready Reaction Terminator Cycle Sequencing Kit (Applied Biosystems, Foster City, California, USA) was used to perform the sequencing reaction. Electrophoresis and detection of fluorescent dye-labelled nucleotide fragments was carried out with an automated DNA sequencer (ABI Prism 377, Applied Biosystems). All sequences were deposited in the EMBL Nucleotide Sequence Database under the accession numbers listed in Appendix S1 (Supplementary material online).

Restriction fragment analysis

Isolation and purification of mtDNA followed standard methods (Spolsky & Uzzell, 1984, 1986). Approximately 5–10 ng of each DNA was digested to completion with 19 different restriction enzymes (AvaI, BamHI, BclI, BglII, EcoRI, EcoRV, HaeII, HincII, HindIII, HpaI, KpnI, MluI, PstI, PvuII, SalI, SmaI, SstI, XbaI and XhoI) under conditions recommended by the supplier (Boehringer Mannheim Biochemicals, Mannheim, Germany, Bethesda Research Laboratories, Rockville, Maryland, USA or New England Biolabs, Ipswich, MA, USA). Resulting DNA fragments were end-labelled with a mixture of four α-32P-triphosphate deoxynucleosides (adenine, cytosine, guanine and tyrosine). Restriction enzyme fragment patterns were determined from autoradiographs after electrophoresis through horizontal 1% agarose or vertical 3.5–4% polyacrylamide gels. For each gel, fragment sizes were estimated from mobilities of size standards (λ and PM2 DNAs, each cut with HindIII for agarose gels; ϕX174 cut with HincII and pBR322 cut with AluI for polyacrylamide gels).

AluI restriction of ND2 PCR products

Mitochondrial ND2 sequences of 287 western Palaearctic water frogs representing at least 10 evolutionary species were analysed with the online tool WEB CUTTER (http://www.medkem.gu.se/cutter) to find specific restriction sites that allow differentiation between R. lessonae and R. ridibunda mtDNAs. One AluI restriction site (AG↓CT) spanning positions 633–636 was detected in each of the 44 R. ridibunda ND2 sequences, in seven ND2 sequences of Rana bedriagae from the Near East (western Syria, Jordan, Nile delta) and in nine of Rana cretensis, but not in the ND2 sequences of 56 R. lessonae or any of other western Palaearctic water frogs (J. Plötner, unpublished results). To distinguish between R. ridibunda- and R. lessonae-specific mitochondrial genomes only a 770-bp-long mtDNA segment that contained this restriction site was amplified using the primers L2 and H2 (Appendix S2, Supplementary material online). Two microlitres (20 units) of AluI (Boehringer Mannheim [BM] or Angewandte Gentechnologie Systeme GmbH [AGS], Heidelberg, Germany), 2 μl of supplied incubation buffer (BM incubation buffer A or AGS buffer violet) and 1 μl of double distilled water were added to 15 μl of the solution that contained the PCR product. After incubation for 2 h at 37 °C, the fragments were separated by electrophoresis on a horizontal 2% agarose gel. Rana ridibunda-specific mtDNA was cut into two fragments with lengths of 319 and 451 bp, whereas mtDNA of R. lessonae remained undigested as evidenced by a single band in the gel.

Data analysis

Mitochondrial DNA sequences were initially aligned using clustal v (Higgins et al., 1992). Subsequently, alignments were improved manually. They are available from the EMBL align database under the accession number ALIGN_001211. For sequence statistics the programs mega 3.1 (Kumar et al., 2004) and DnaSP 4.10.9 (Rozas et al., 2003) were used. We calculated the total number of transitions, transversions and amino acid substitutions, the nucleotide diversity (Pi) and uncorrected p distances (a simple measure of sequence divergence). We applied Tajima’s D-test (Tajima, 1989) to investigate departures from selective neutrality at the ND2 and ND3 genes in both R. ridibunda and R. lessonae, and a McDonald–Kreitman (MK) test (McDonald & Kreitman, 1991) to test for neutrality in interspecies sequence divergence. The neutrality index (NI) was calculated using Rand & Kann’s (1996) formula. The restriction profiles were analysed by calculating the amount of sequence divergence from the fraction of restriction fragments shared by a pair of mtDNAs (Nei & Li, 1979), using Upholt’s (1977) formula.

Mitochondrial haplotype genealogies were estimated from complete ND2 and ND3 sequences with maximum likelihood (paup 4.0b10: Swofford, 2003) and Bayesian algorithms (MrBayes 3.1.2: Ronquist & Huelsenbeck, 2003; Huelsenbeck & Ronquist, 2005). For both maximum-likelihood and Bayesian inferences, the general time reversible model with corrections for invariant characters and gamma-distributed rate heterogeneity (GTR + I + G) was applied as recommended by Huelsenbeck & Rannala (2004). Model parameters were estimated from the data set. For MrBayes, the posterior probabilities of phylogenetic trees were calculated by employing a 107-generation Metropolis-coupled Markov chain Monte Carlo (four chains, chain temperature parameter = 0.2) as implemented in the program. Trees were sampled every 1000 generations; the first 1000 recorded trees were discarded; a total of 9000 recorded trees were examined. The autocorrelation behaviour of MrBayes was explored with program internal statistics and the tool awty (http://ceb.csit.fsu.edu/awty; Nylander et al., 2007). Both analyses suggested convergence of the split frequencies (standard deviation after 3500 recorded trees was lower than 0.01). The convergence of the model parameter estimates was checked using the potential scale reduction factor calculated in MrBayes: the limit of the statistic of 1.0 was reached for all model parameters.

For the phylogenetic analysis, we augmented the R. ridibunda and R. lessonae sequences with sequences of Rana epeirotica and R. cretensis as an outgroup, and Rana shqiperica, the presumptive sister species of R. lessonae sensu lato. This allowed a consistent rooting of the haplotypes for R. ridibunda and R. lessonae on the tree.

To analyse the effects of location (longitude and latitude) on haplotype composition in R. esculenta and on the introgression rate of R. lessonae-specific mtDNA into R. ridibunda, we calculated the relative frequency (fl) of R. lessonae-specific haplotypes (including haplotypes of the Italian R. lessonae stock) for each population with sample size larger than one. We performed simple and multiple regression analyses with fl as the dependent variable and latitude and longitude as independent variables using the software packet Statgraphics Plus 4.1 (Statistical Graphics Corp., StatPoint Inc., Herndon, Virginia, USA). To judge the relative magnitude of the residuals with respect to the explanatory power of the selected independent variables, a component-plus-residuals plot (Wood, 1973) was made. In the regression analyses, we also incorporated data published by Spolsky & Uzzell (1984, 1986) and Monnerot et al. (1985) (Appendix S1, Supplementary material online).

Results

Inter- and intraspecific variation of mtDNA

Both mitochondrial sequences (ND2, ND3, 12S rDNA) and digest profiles of whole mtDNAs revealed clear differences between R. ridibunda and R. lessonae. Interspecific comparisons of the 12S rRNA gene segments yielded substitution numbers between 9 and 12, which correspond to sequence divergences of 2.4% and 3.2% respectively (Plötner & Ohst, 2001). The interspecific differences in the less conserved ND2 + ND3 sequences are about five to six times higher than in the 12S rRNA gene segment. Nevertheless, because there are at least nine substitutions, the species specificity of mtDNA could be clearly determined on the basis of this gene segment.

The R. ridibunda ND2 + ND3 sequences contained 38 (2.8%) variable sites (31 in ND2 and seven in ND3) defining 19 haplotypes (Table 1, Fig. 1); 31 (2.3%) sites were parsimony informative. In both genes, almost all nucleotide substitutions were transitions, most of them at the third codon position; only three were found in the first position (two in ND2 and one in ND3) and two in the second position (both in ND2). Only five substitutions resulted in amino acid replacements (four in ND2 and one in ND3). Nucleotide diversity (Pi) amounted to 0.007 ± 0.0005. Genetic distance ranged from 0.07% to 1.6% among R. ridibunda-specific haplotypes with an overall nucleotide diversity of 0.74%.

Table 1.   Variable positions and nucleotide composition in the mt ND2 and ND3 gene defining 19 different haplotypes (HT) across 29 populations of Rana ridibunda.
 Site (position)
ND2ND3
Sequence No.                              1        
01222333444445555556677888899900122223 
94227244037891367892546016966707634770 
31013125554988119814272497479885846673 
             *                *        
33123333333332333333133333313333333313HT
  1. Transversions are indicated by asterisks. The numbers below nucleotide positions indicate the codon position. Bold face numbers indicate amino acid replacement substitutions.

1–3TCAGCTGCCTTAATTGATATCAAACACATTACGGTCGCR1
4–9.............A........................R2
10.............A......................A.R3
11............GA.A......................R4
12.........C.G.A........................R5
13–15...A.......G.A........................R6
16–17.T...........A..........T.............R7
18.........C...A...................A....R8
19..G......C...A...................A....R9
20C....C.T.C...A..........TG.....TAA...TR10
21C............ACA....T.G.T....C.T.A...TR11
22C.G..........ACA....T.G.T....C.TAA...TR12
23–29C............A......T.G.T....C.TA.....R13
30–31C.......T....A...C..T.G.T....C.TA.....R14
32C............A......T.G.T.T.....A.....R15
33C.....A......A......T.G.T.T.....A.....R16
34–42C...T.A...C..A..G.GCTGG.T.TGC...A.CT..R17
43C...T.A...C..A..G.GCTGGGT.TGC...A.CT..R18
44C...T.A...C..A..G.GCTGG.T.TGC.T.A.CT..R19
Figure 1.

 Phylogenetic relationships of Rana lessonae and Rana ridibunda haplotypes. (a) Complete MrBayes maximum posterior tree of all haplotypes with R. shqiperica included as the sister taxon of R. lessonae and two presumptive outgroups of R. ridibunda (Rana cretensis and Rana epeirotica). (b) R. ridibunda clade: labels mark the ridibunda haplotype (R1–R19) and corresponding sequence numbers (Table 1); (c) R. lessonae clade: labels mark the lessonae haplotype (L1–L17) and sequence number (Table 2). The values at the branches are posterior probabilities for the clade to the right of that branch. The units of the scale bar in A, B and C are the expected mutations per site; B and C are on the same scale. Sequence numbers are specified in Supplementary material online.

In the ND2 + ND3 sequences of R. lessonae, the intraspecific variability is about 2.5 times lower than in R. ridibunda. Intraspecific p distances ranged between 0.1% and 0.7% with an average of 0.3%. Only 24 variable sites (1.7%) were found (21 in ND2 and three in ND3); almost all substitutions were transitions at the first (10) or third (11) codon position (Table 2). Seven sites (0.5%) were parsimony informative. Twelve amino acid replacements were observed caused by 11 substitutions in the first codon position and one substitution in the second codon position. A Pi-value of 0.00187 ± 0.00033 was calculated.

Table 2.   Variable positions and nucleotide composition in the mt ND2 and ND3 gene defining 17 different haplotypes (HT) across 39 populations of R. lessonae.
 Site (position)
ND2ND3
Sequence No.                    1    
012233444466778888990012 
710066167899360339691825 
851869532912827252461595 
  **                     
313133111312333111133133HT
  1. Transversions are indicated by asterisks. The numbers below nucleotide positions indicate the codon position. Bold face numbers indicate amino acid replacement substitutions.

1–8GGCCCAAACGGCTGAGGGCATAATL1
9–33.....G.................CL2
34.....G.............G.G.CL3
35.....G...A.............CL4
36–40.....G.G...............CL5
41.....G..........A......CL6
42.....G..T..............CL7
43..A..G...........A.....CL8
44–45.A...G.................CL9
46–47.....G.....T...........CL10
48.....G..............C..CL11
49.....G....A............CL12
50–51A..G.G....A...G........CL13
52....TGG...A...GA......GCL14
53....TGG...A.C.GA..T...GCL15
54....TGG...A.CAGA......GCL16
55–56....TGG...A.C.GA......GCL17

A total of 17 R. lessonae-specific haplotypes could be distinguished (Table 2), which represent three main clusters as supported by the results of maximum-likelihood (data not shown) and Bayesian analyses (Fig. 1). Cluster 1 comprises the very closely related haplotypes L1–L12; cluster 2 consists of only one haplotype (L13), which perhaps represents an ancestral state within the R. lessonae clade; and cluster 3 includes haplotypes L14–L17, which differ from the other R. lessonae-specific haplotypes by at least four substitutions (three in ND2 and one in ND3).

The digest profiles of R. ridibunda and R. lessonae mtDNAs were described in detail by Spolsky & Uzzell (1984, 1986). The proportion of fragments shared by both species ranged between 0.256 and 0.258, which corresponds to sequence differences of 8.1 ± 1.0% and 8.2 ± 1.0% respectively. In the ND2 gene, R. ridibunda differed from R. lessonae in at least 134 sites (12.9%); in the ND3 gene a minimum of 48 substitutions (14.1%) were detected. The sequence divergence (p distance) obtained as a result of interspecific comparisons of the complete ND2 + ND3 sequences ranged between 14.2% and 15.2%.

Compared with the 36 mitochondrial haplotypes differentiated on the basis of ND2 + ND3 sequences (Table 1 and 2), only three types of mtDNAs (A–C) were detected by restriction fragment patterns of total mtDNA (Spolsky & Uzzell, 1984). Type A was found in R. ridibunda and R. esculenta but not in R. lessonae and thus was considered as R. ridibunda-specific. Type A probably contains several of the 19 ND2 + ND3 haplotypes; because there are no differences in the recognition sites in the ND2 and ND3 genes of R. ridibunda for the restriction enzymes used by Spolsky and Uzzell, the ND2 + ND3 sequence diversity would not have been revealed in their study.

Types B and C, which were considered as R. lessonae specific, showed a sequence divergence of only 0.3%, consistent with the p distances obtained from intraspecific comparisons of the ND2 and ND3 sequences (0.1–0.7%). One BamHI restriction site (G↓GATCC) was found in the ND3 gene (positions 1165–1170) of haplotypes L1–L12 and L13, but not in haplotypes L14–L17. An assignment of single haplotypes to a R. lessonae-specific fragment pattern type, however, is not possible. BamHI cuts mtDNA of types B and C into three fragments with sizes of 8000, 6400 and 5600 bp (type B) and 11 500, 8000 and 470 bp (type C) respectively. Because a BamHI restriction site exists in all the R. lessonae ND3 sequences except for haplotypes L14–L17, these cleavage patterns suggest that one of the restriction sites responsible for the 8000 bp fragment in types B and C patterns is in the ND3 gene. If so, haplotypes L14–L17 represent neither type B nor C, but an additional, less frequent, R. lessonae-specific haplotype that is probably identical to type LC described by Monnerot et al. (1985) on the basis of the digest profiles of BamHI, EcoRI and PstI.

Spolsky & Uzzell (1986) also described pattern type D, which is characteristic of Italian water frogs, but was also found in R. lessonae and R. esculenta from Switzerland (Hotz et al., 1992) and southern Germany (T. Ohst & J. Plötner, unpublished results). The sequence divergence, calculated from shared restriction fragments between the mtDNAs of R. lessonae (B and C) and type D (3.4%), is consistent with distance values obtained on the basis of the ND2 + ND3 sequences (3.8–5.0%, mean: 4.4 ± 0.28%).

Tests of neutrality

Tajima’s D-test, which compares the number of segregating sites per site with the nucleotide diversity, revealed no departure from a neutral model of sequence evolution within either R. ridibunda (D = −1.62; 0.10 > P > 0.05) or R. lessonae (D = 0.42; P > 0.10). For ND2 and ND3 together, we observed 43 synonymous substitutions and 17 nonsynonymous substitutions within the two species, and 26 amino acid replacements and 156 silent fixed differences between them. The ratios of nonsynonymous to synonymous substitutions within and between species are significantly different from each other (Fisher’s exact test, two tailed: P = 0.02, G-test: G = 5.64; P = 0.02). An NI of 2.37 indicates an excess of amino acid polymorphism, which is consistent with the hypothesis that most segregating functional variants are deleterious and thus subject to purifying selection. If ND2 and ND3 are considered separately, the NI values are 1.96 and 5.75 respectively. There is, however, no significant departure from neutrality (ND2: Fisher’s exact test, two tailed: P = 0.10, G = 3.03, P = 0.08; ND3: Fisher’s exact test, two tailed: P = 0.13, G = 2.48; P = 0.12).

Geographical distribution of haplotypes

In R. ridibunda, 16 specific haplotypes were found at single localities; only three (R2, R13 and R17) were found at two or more localities (Appendix S1, Supplementary material online). Haplotypes R1–R11 occur in R. ridibunda from the Balkan peninsular (Greece and Albania). Haplotype R13 was found in southern, south-eastern and central Europe. Haplotype R17 is the most common and characteristic haplotype of central European R. ridibunda.

Among the R. lessonae-specific haplotypes, three (L1, L2 and L17) were found at different localities (Appendix S1, Supplementary material online). L2, the most common haplotype of R. lessonae, is distributed from south-eastern Europe (Danube delta), over central Europe to Scandinavia. The closely related haplotypes L1 and L3–L12 occur in central and northern Europe. The less frequent haplotypes L13–L17 were found in northern Italy (L13), in south-eastern Europe (L14) and in central and eastern Europe (L15–L17).

Interspecific genome transfer and haplotype composition in R. esculenta

Of 407 R. ridibunda, 137 (33.7%) possessed mtDNA specific for R. lessonae. The introgressed genomes represent at least six different haplotypes (L1, L2, L8, L10, L12 and L15), three of them (L8, L12 and L15) observed only in R. ridibunda, two (L1 and L10) in both R. ridibunda and R. esculenta and one (L2) in all three water frog forms. In contrast to R. ridibunda, introgression of R. ridibunda-specific mtDNA has not been detected in R. lessonae.

Of 335 R. esculenta, the majority (90.4%) possessed mtDNA characteristic of R. lessonae or the Italian taxon (R. lessonae sensu lato); haplotypes specific for the Italian R. lessonae stock represented 6.6% of these. Only 9.6% of the R. esculenta contained R. ridibunda-specific mtDNA, represented by haplotypes R13 and (according to ND3 data only) probably R17. All triploid individuals had only R. lessonae-specific mtDNA (Appendix S1, Supplementary material online).

Rana ridibunda individuals with R. lessonae-specific mtDNA are restricted to central Europe, whereas all individuals examined from southern and eastern Europe (south of 48°N latitude and east of 22°E longitude) possessed exclusively R. ridibunda-specific mtDNA (Fig. 2). This pattern is expressed in a statistically significant relationship between the frequency of individuals with R. lessonae-specific mtDNA (fl) and the geographical coordinates at the 99% confidence level in both R. ridibunda and R. esculenta (Table 3 and Appendix S3, Supplementary material online). The fitted multiple linear regression models explain 46.3% of the variability in fl for R. ridibunda and 44.5% for R. esculenta. There is a moderately strong negative correlation of fl with longitude under a linear regression model in R. ridibunda (r = −0.55, anova: F-ratio = 32.16, d.f. = 75, P = 0) and in R. esculenta (r = −0.57, anova: F-ratio = 33.53, d.f. = 71, P = 0); and a significant positive correlation between fl and latitude in both forms (R. ridibunda: r = 0.56, F-ratio = 33.72, d.f. = 75, P = 0; R. esculenta: r = 0.41, F-ratio = 13.81, d.f. = 71, P = 0.0004).

Figure 2.

 Geographical distribution of haplotypes found in Rana ridibunda and Rana esculenta.

Table 3.   Results of the multiple regression analysis with the relative frequency of lessonae-specific haplotypes (fl) as the dependent variable and longitude and latitude as independent variables.
ParameterRana ridibundaRana esculenta
ModelResidualModelResidual
  1. The Durbin–Watson (DW) statistic tests (Durbin & Watson, 1950, 1951, 1971) the residuals to determine if there is any significant correlation based on the order in which they occur in the data file. DW values > 1.4 indicate that there is not any serious autocorrelation in the residuals.

Sum of squares7.018.111.782.08
Degrees of freedom273269
Mean square3.500.110.890.03
F ratio31.529.5
P-value00
r246.3346.1
Standard error of estimates0.330.17
Mean absolute error0.260.12
DW statistic1.941.86

Discussion

Patterns of genome transfer

In western Palaearctic water frogs, introgression of R. lessonae mtDNA into R. ridibunda is closely related to the occurrence of the hybridogenetic hybrid R. esculenta. Rana ridibunda individuals with introgressed R. lessonae mtDNA are restricted to central Europe north of 48°N latitude and between 8° and 22°E longitude. In this area, hybridogenetic hybrids mainly live syntopically with a parental species in either lessonaeesculenta (L–E) or ridibundaesculenta (R–E) populations (reviewed by Plötner, 2005). In eastern Europe (especially Russia and Ukraine), R. ridibunda also co-occurs with hybridogenetic R. esculenta, but, in this region, introgression of mitochondrial genomes has not been observed. In contrast to central European populations, where hybrids are both diploid (LR) and triploid (LLR and RRL), almost all R. esculenta in eastern Europe are diploid (Borkin et al., 1986; Rybacki & Berger, 2001; Borkin et al., 2004, 2006 described exceptions). The observed geographical patterns of mtDNA introgression, in concert with the uniform presence of R. lessonae mtDNA in triploid individuals and the geographical distribution patterns of triploids (Rybacki & Berger, 2001) suggest that the R. lessonae mitochondrial genome was, or is, transferred into R. ridibunda not only by LR females but also, at least in some populations, via triploid females. The occurrence in R. ridibunda of different R. lessonae-specific haplotypes that belong to distinct well-defined clades shows that the genome transfer traces back to several independent transmission events, although it is possible that each of these R. lessonae haplotypes seen in R. ridibunda was introgressed many times independently. Because some of the more divergent haplotypes have not been observed in R. lessonae itself, it is possible that they result from pre-Holocene introgression events.

Four of 38 R. ridibunda (10.5%) examined from the Pannonian Basin possessed R. lessonae-specific mtDNA, which shows that genome introgression takes place despite the scarcity of hybrid males (Berger et al., 1988; Gubányi, 1992; Morozov-Leonov et al., 2003). The unexpectedly high proportion of LR individuals with R. lessonae-specific mtDNA (11 of 23 R. esculenta; 48%) seems, however, to contradict the comparatively low introgression rate observed in Pannonian R. ridibunda. Crosses between diploid hybrid males (LR) and R. lessonae females are a precondition for the formation of LR females with R. lessonae-specific mtDNA that probably act as vehicles for the introgression of R. lessonae mtDNA into R. ridibunda. In contrast to our expectation, such hybrid females have been found, for example, in Hungary (locality 88: one of four individuals) and Slovakia (locality 154 and 156: four of five individuals; Appendix S1, Supplementary material online). Obviously, crosses between R. ridibunda males and LR females with R. lessonae mtDNA are rare, possibly because of specific mate recognition mechanisms, or do not result in fertile RR progeny.

Presumptive selective forces and mitochondrial–nuclear coadaptation

Whether sequence variation in mtDNA is selectively neutral or under selective constraints is still controversial. In our data set, the overall genetic variability in the ND2 + ND3 genes as measured by nucleotide substitutions is higher in R. ridibunda, but amino acid substitutions, caused mainly by mutations in the first codon position, are about two times higher in R. lessonae. As indicated by the results of the neutrality tests, this difference may reflect purifying selection rather than adaptation, supporting the hypothesis that most segregating functional variants are deleterious and rarely contribute to divergence (e.g. Meiklejohn et al., 2007). Our data do not corroborate the assumption of Bazin et al. (2006) that population size does not influence genetic diversity in mtDNA. The higher genetic variability in R. ridibunda mtDNA is instead consistent with the hypothesis that mtDNA diversity and population size are positively related in animals (e.g. Mulligan et al., 2006). Rana ridibunda has a distribution about 0.5 times larger than that of R. lessonae (e.g. Plötner, 2005) and based on empirical field data and measured heterozygosity of nuclear loci (S. N. Litvinchuk, unpublished data), probably has a larger effective population size. This problem needs further investigation, however, particularly because it is possible that the measured genetic variability is influenced by sample biases linked with our concept of species delineation.

Individuals of both R. esculenta and R. ridibunda with R. lessonae mtDNA may have an adaptive advantage rather than disadvantage, especially in their northern ranges, despite possible disruption of nuclear/mitochondrial protein coadaptation in R. ridibunda with introgressed R. lessonae mtDNA. In particular, the O2/CO2 metabolism of introgressed compared with nonintrogressed individuals, both as larvae and after metamorphosis, may be significant. Under hypoxic conditions, larvae of R. ridibunda develop less well than larvae of R. lessonae (Plenet et al., 2000a, b). If R. lessonae mitochondria are more effective than R. ridibunda mitochondria under hypoxic conditions, and if this difference depends on enzymes that are encoded in part by the mitochondrial genome, R. ridibunda larvae with R. lessonae-specific mtDNA may be less sensitive to oxygen deficiency caused by hypertrophic conditions or pollution than ‘normal’R. ridibunda larvae. For metamorphosed R. ridibunda, the type of mitochondrial DNA may also play a role in their survival during hibernation. Because R. ridibunda hibernates exclusively beneath the water surface (Berger, 1982), it is exposed to an oxygen deficiency especially if the waters are covered with an ice sheet for a long period of time. Such a condition can lead to a high mortality or even extermination of whole populations (Berger, 1982). Under reduced oxygen pressure, the survival time of R. lessonae metamorphs was higher than those of R. ridibunda metamorphs (Tunner & Nopp, 1979; Lutschinger, 1988), which also demonstrates that R. ridibunda is more sensitive to oxygen deficiency than R. lessonae.

The large divergence between both the nuclear and the mt genomes of R. lessonae and R. ridibunda (the modified Nei distances calculated on the basis of 31 protein coding loci varied between 0.39 and 0.60; Beerli et al., 1996) suggests that intergenomic coadaptation would be strongly disrupted when R. lessonae mtDNAs are introgressed into R. ridibunda. Because more than 90% of mitochondrial proteins are encoded in the nucleus, intergenomic protein coadaptation is assumed to be directly correlated with fitness (e.g. Burton et al., 2006; Rand et al., 2006; Gusdon et al., 2007). The extent of transfer of R. lessonae mtDNA into R. ridibunda (33.7%) suggests that the presence of R. lessonae mtDNA does not significantly influence mitochondrial–nuclear interaction and hence, the fitness of this species, despite large divergence between the mitochondrial genomes of R. lessonae and R. ridibunda. It is possible, of course, that disruption of nuclear/mitochondrial protein coadaptation in water frogs with introgressed mtDNA is variable in its significance. The nature of this nuclear-mtDNA epistasis is not clear, however.

Although the possibility that introgressed R. lessonae mtDNA genomes provide an adaptive advantage to northern R. ridibunda (and perhaps also to northern R. esculenta) is a tempting hypothesis to be tested, there is, in clonal mole salamanders of the genus Ambystoma, a striking case of ‘introgressed’ mitochondrial genomes in which the phylogenetic divergence between the nuclear genomes of the normal host species and the phylogenetically closest possible source of the mitochondrial sequences is enormous. Here, an almost invariant mitochondrial genome (Spolsky et al., 1992; C. Spolsky & T. Uzzell, unpublished data) exists independently for perhaps 5 × 105 years (Hedges et al., 1992; Spolsky et al., 1992) within the widespread triploid taxa Ambystoma platineum and A. tremblayi (Uzzell, 1964) of the subgenus Ambystoma, although the mitochondrial sequence itself apparently is derived from Ambystoma barbouri (Robertson et al., 2006) of the distantly related subgenus Linguaelapsus. If strong coadaptation between proteins encoded by the nuclear and mitochondrial genomes of a taxon or population is a general rule and the disruption of this coadaptation by hybridization a frequent source of post-zygotic incompatibility between taxa (Burton et al., 2006), the mitochondrial genomes of clonal Ambystoma must possess remarkable properties. What these might be for clonal Ambystoma is completely unknown, but at least for the frogs a plausible, although not yet tested, hypothesis can be proposed.

Biogeographic implications

Provided that the mt ND2 and ND3 genes of western Palaearctic water frogs diverge at about 2.0–2.5% per Myr (estimated from the sequence divergence in these two genes between R. cretensis and Anatolian water frogs, which are thought to have been separated for about 5 Myr, Beerli et al., 1996), the divergence between different haplotypes within both R. ridibunda (maximum distance value 1.6%) and R. lessonae (maximum distance value 0.7%) indicates that much of the early intraspecific differentiation took place prior to the most recent glacial cycle. During the Würm glaciation, presumptive refuges are the southern Balkans (especially the Peloponnisos and adjacent areas) for R. ridibunda and the western (and probably northern) Black Sea region (for both R. ridibunda and R. lessonae). Haplotypes R1–R9 (the Greek ridibunda lineage sensu stricto) remained more or less confined to their potential refuge and nearby regions, resulting in a much smaller current range than other haplotypes (for example R13 and R17), which probably underwent rapid northward expansion with the advent of Holocene warming. Similar scenarios have been suggested for the European pond turtle, Emys orbicularis (Lenk et al., 1999), the newt species Triturus vulgaris (Babik et al., 2005), the fire-bellied toad Bombina bombina (Hofmann et al., 2007) and the grasshopper Chorthippus parallelus (Hewitt, 2001).

Because of climatic oscillations, the post-Würm colonization of central Europe probably involved a sequence of range expansions followed by contractions and attendant reductions in population sizes (bottlenecks) that may have reduced genetic diversity in the northern populations. Such reductions might be lower in cold-tolerant taxa (Taberlet et al., 1998; Hewitt, 2004). This hypothesis is supported by the immunological (Uzzell, 1979, 1983), mitochondrial and ecological data obtained in central and northern European water frog populations. Based on a comparison of the distribution ranges (R. lessonae goes farther north than R. ridibunda), the altitude distribution of water frog species (R. lessonae populates habitats up to 1,000 m above sea level while R. ridibunda is restricted to the lowlands; Günther, 1990, 1997a, b), and the results of preliminary ecological experiments that showed a higher hatching success of R. lessonae larvae at temperatures below 13 °C (Günther, 1974), it seems probable that R. ridibunda is a rather cold sensitive species that recolonized the deglaciated northern parts of Europe later than R. lessonae. The amount of genetic variability observed in the mtDNA of R. lessonae (15 haplotypes) and R. ridibunda (only two haplotypes) from central and northern Europe corresponds well with this assumption.

Acknowledgments

We thank Bradley Anholt and two anonymous reviewers for constructive and critical comments on the manuscript. For providing water frog samples, we thank Martina Arioli and Christian Jakob (Zürich), Rainer Günther (Berlin), Petr Kotlik (Libechov), Peter Mikuliček (Prague), Andreas Nöllert (Jena), Ulrich Scheidt (Erfurt) and Dirk Schmeller (Leipzig). Tissue samples from Greece were made available by the Greek Ministry of Rural Development and Food, kindly mediated by Doris Tippmann (Embassy of the FRG, Athens). Grateful thanks are due to Robert Schreiber, Kai-Joachim Schultze and Ronny Knop (Museum für Naturkunde, Berlin) and Alexandra Schulmeister (University Bergen, Norway) for technical assistance. This research was supported by the Deutsche Forschungsgemeinschaft (grants PL 213/3-1, 3-2, 3-3) and the Swiss National Fund (grants 31-37579.93, 31-59144.99, 31-103903/1 and 31-64004.00). Peter Beerli was partly supported by the joint NSF/NIGMS Mathematical Biology programme under NIH grant R01 GM 078985. Spartak N. Litvinchuk was partly supported by a DAAD grant under A/03/06782.

Ancillary