Ancient Mitochondrial Capture as Factor Promoting Mitonuclear Discordance in Freshwater Fishes: A Case Study in the Genus Squalius (Actinopterygii, Cyprinidae) in Greece

Hybridization and incomplete lineage sorting are common confounding factors in phylogeny and speciation resulting in mitonuclear disparity. Mitochondrial introgression, a particular case of hybridization, may, in extreme cases, lead to replacement of the mitochondrial genome of one species with that of another (mitochondrial capture). We investigated mitochondrial introgression involving two species of the cyprinid genus Squalius in the western Peloponnese region of Greece using molecular and morphological data. We found evidence of complete mitochondrial introgression of Squalius keadicus into two populations recognized as Squalius peloponensis from the Miras and Pamissos River basins and a divergence of mitochondrial genomes of S. keadicus from the Evrotas basin from that of the introgressed populations dating from the Pleistocene. Secondary contact among basins is a possible factor in connection of the species and the introgression event. Morphological analyses support the hypothesis of mitochondrial introgression, as S. keadicus was different from the other three populations recognized as S. peloponensis, although significant differences were found among the four populations. Isolation by geographical barriers arose during Pleistocene in the western Peloponnese were the source of the evolution of the two reciprocally monophyletic subclades found in the S. keadicus mitochondrial clade, and the morphological differences found among the four populations. Along with the lack of structure in the nuclear genome in the three populations ascribed to S. peloponensis, this suggests an incipient speciation process occurring in these Squalius species in the western Peloponnese.


Introduction
The use of mitochondrial DNA for inferring phylogenetic or phylogeographic relationships has been traditionally preferred over the use of the nuclear genome, as mtDNA accurately reflects recent divergence patterns than do nuclear markers [1]. Nevertheless, concordant patterns between mtDNA and nuclear DNA are not always observed in nature [2][3][4].
Evolutionary processes such as hybridization and incomplete lineage sorting are common confounding factors in phylogeny and speciation and primary causes of mitonuclear discordance [3][4][5]. Incomplete lineage sorting occurs when the coalescence time of genes and speciation differ, i.e. when lineages fail to sort out at the same time at speciation happens [6]. Therefore, gene trees do not always represent the true relationships among taxa because different genes may have not branched at the same time [7][8][9]. Distinguishing incomplete lineage sorting from hybridization can be difficult as these two processes may originate the same phylogenetic tree [10][11][12]. Nonetheless, among others, any biogeographic pattern is expected when incomplete lineage sorting is the confounding factor [4].
Hybridization between species involves mating between unrelated organisms regardless the taxonomic status and, in some cases, may lead to gene transfer, a common process in plants [13], or to very complex evolutionary processes occurring in freshwater fishes [14][15]. Introgression is a particular case of hybridization and occurs when gene flow exists between populations that hybridize and hybrids backcross to one or both parental populations [16]. Within this framework, mitochondrial capture is defined as complete mitochondrial introgression, a situation in which the mitochondrial genome of one species is replaced with that of another in an entire population, as a consequence of selective backcrossing of hybrids with one of the parent species [2][3][17][18][19]. In hybridization, the complete replacement of the mitochondrial DNA of one species with that of another is more common than is replacement of a portion of the nuclear genome, due to maternal inheritance and the four-fold smaller effective population size of mitochondrial DNA compared to nuclear DNA, implying that mtDNA will complete the process of lineage sorting more rapidly than will nuclear DNA [1,3,[20][21]. Understanding evolutionary processes such as hybridization, mitochondrial introgression, and incomplete lineage sorting is of vital importance to inference reliable phylogenies and reconstruction of evolutionary history and speciation processes, especially in currently allopatric species. Revealing mitochondrial introgression may be the only way to identify past hybridization of allopatric species [17].
In general, mitonuclear discordance due to mitochondrial introgression occurs more frequently in certain taxonomic groups, including mammals and fish; whereas, in groups such as birds and butterflies, in which the female is heterogametic, mitochondrial introgression is usually reduced, conforming to Haldane's rule [4]. Examples of the complete replacement of mitochondrial DNA are found in various taxa [18][19][22][23][24][25], including freshwater fishes [26][27][28].
Within freshwater fishes the preponderance of hybridization is greater in cyprinids than is in other freshwater fish families such as salmonids or cichlids and several examples are observed [29][30][31][32]. A clear example of mitonuclear discordance has been described for the genus Squalius in the Greek Peloponnese. Previous studies focusing on a more general framework of phylogenetic relationships of Squalius hypothesised the presence of mitochondrial introgression to explain discrepancy found in the mitochondrial (DNA sequences) and nuclear (allozyme) phylogenetic position of S. keadicus (Stephanidis, 1971) and S. peloponensis (Valenncienes, 1844) [33][34]. These two species, currently showing allopatric distribution, belong to two different evolutionary lineages recognized within the genus Squalius; S. peloponesis belongs to the Euroasiatic group while S. keadicus is included together with other smallsized species within the Mediterranean group. These two evolutionary lineages have been isolated since the Middle Miocene [35][36]. Squalius keadicus is a small-to medium-sized fish endemic to the Evrotas Basin in the southwestern Peloponnese. Squalius peloponensis is a medium to large fish and has a broader distribution range in the western Peloponnese, encompassing several hydrological basins (Fig 1) [37]. The morphological distinction between both species is clear and several diagnostic morphometric and meristic characters are identified to distinguishing them from other Balkan Squalius species. Head length is 24-27% standard length in S. peloponensis. The posterior margin of the anal fin is different in both species, being straight in S. keadicus and slightly convex in S. peloponensis. Meristic diagnostic characters of S. keadicus show 44-49 scales on lateral line whereas in S. peloponensis this range is 40-44. The colouration of both species also differs; S. keadicus shows in live a conspicuous blackish stripe on flank from eye to caudal fin base and very dark scales in upper half of flank, which are not present in S. peloponensis [37].
With regards to the paleogeographical context in which the two species have evolved, the southwestern Peloponnese lies near the Hellenic Trench in the Eastern Mediterranean region, one of the most active areas in the Euroasiatic-African convergence zone owing to the northward subduction of the African plate beneath the Aegean plate [38]. The region has a complex tectonic history and numerous currently active northwestern-southeastern tectonic faults [38][39][40][41]. The formation of the hydrological network of the Greek Peloponnese was complex, characterized by Late Miocene-Quaternary compressional structures as well as Pliocene lacustrine-marine and Pleistocene terrestrial-fluvial sedimentary successions leading to the formation of fluvial terraces [38,[42][43]. Such paleogeographic complexity may have promoted ancient contact among freshwater fauna inhabiting the basins of the western Peloponnese, resulting in hybridization events.
Within the framework of mitonuclear disparity, the aim of this study was to investigate the mitochondrial capture between two species of Squalius from the Greek Peloponnese using molecular and morphological analyses of three populations of S. peloponensis and one of S. keadicus. We also establish a temporal and biogeographical framework for the introgression episode.

Sampling
Specimens of four populations of Squalius were collected by electrofishing from basins in the western Peloponnese region of Greece following the European regulations (EN ISO 14011:200. Water quality-Sampling of fish with electrophising) ( Table 1; Fig 1). Sampling was performed with authorised permission from the Greek Ministry of Environment, Energy and Climatic Change and after consulting the Animal Care and Use Committee of the National Museum of Prague (č.j.NM/2014/258). Immediately after capture, all individuals were anesthetized using Tricaine methanesulfonate (MS-222) to alleviate suffering. A piece of ventral fin sample (~3 mm 2 fin clips) was obtained from 61 fish belonging to the four Greek populations analysed (nine to S. keadicus and 52 to S. peloponensis). All tissues were preserved in 95% ethanol and stored at 4˚C until is processing at laboratory for molecular analyses. These 61 specimens analysed for molecular markers were sacrificed for morphological analyses following humanely euthanization with overdose of MS-222 according to European Commission regulations (Directive 2010/63/EU). Death was confirmed after no gill movement was observed for at least 10 minutes. These new sampled specimens used for morphological purposes were preserved in a solution of 5% formalin and after a week they were transferred to a 70% ethanol solution. These new-collected specimens have been deposited in the Fish Collection of the National Museum of Prague (Czech Republic). The remaining specimens used for morphological analyses in this study came from historical deposits of the Fish Collections of the Museo Nacional de Ciencias Naturales, CSIC, in Madrid (Spain). Voucher tissues of new collected specimens are deposited in the DNA Collections of the Museo Nacional de Ciencias Naturales in Madrid, CSIC, (Spain) and in the National Museum of Prague (Czech Republic).
The study included the Evrotas River population, this being the type locality of S. keadicus and currently the only known locality for the species, and three populations recognized as S. peloponensis from the Alfios, Miras, and Pamissos basins (Fig 1). The type locality of S. peloponensis is uncertain, and only the Peloponnese peninsula (= Morea) is cited in its description [44]. Additional species of Squalius along its distribution range were included in the molecular analyses to assess the phylogenetic position of the four investigated populations (Table A in S1 File). The genus Petroleuciscus was used as the outgroup based on previous phylogenetic studies [35][36].

DNA extraction, amplification and sequencing
Eight to 17 specimens per population were analysed (Table 1). Total genomic DNA was isolated using the commercial kit Biosprint 15 for blood and tissue (Qiagen). For each specimen the complete mitochondrial cytochrome b (MT-CYB; 1140bp) gene and two nuclear genes, the first intron of the ribosomal S7 gene (S7; final alignment including gaps = 977bp) and the third exon of the recombination activating protein gene (RAG-1; 1473bp), were amplified. The PCR protocols and primers followed [36]. The PCR products were purified by Exo-SAP-IT (USB, Cleveland, OH, USA) and directly sequenced by Macrogen Europe (Amsterdam, The Netherlands; http://www.macrogen.com) using a 3730XL DNA sequencer. All new sequences of haplotypes and alleles obtained in this study were deposited in the GenBank database (Accession Numbers: MT-CYB: KY070368-KY070424; RAG1: KY070425-KY070574; S7: KY070503-KY070574).

Phylogenetic analyses
For the nuclear markers, phylogenetic inference of independent alleles was also conducted on gamete phases using the PHASE algorithm [45][46] implemented in DnaSP v. 5.0. [47] with a probability threshold of 0.9 to resolve alleles. Sequences were aligned using the default pairwise and multiple alignment parameters in Clustal W [48] implemented in MEGA v.7 [49]. Alignments were later revised. Recombination of nuclear genes was assessed by the PHI test [50] implemented in SplitsTree v. 4.13 [51]. No traces of recombination were found in either RAG-1 or S7 genes (p = 1.0 for both genes). Table 1. Specimens used in molecular and morphological analyses and their voucher numbers. GenBank accession numbers for all analysed genes and labels in phylogenetic trees. All these specimens except those marked with an * symbol, proceeding from GenBank database, have been used in morphological analyses.

Species/Population (Voucher number)
Geographic coordinates

Squalius keadicus/Evrotas
Alfios 3 For phylogenetic analyses, the best-fit model of evolution for each independent gene was estimated using jModelTest v2 [52] (Table B in S1 File). Model parameters were used for subsequent phylogenetic analyses. Phylogenies based on Bayesian inference (BI) and maximum

Species/Population (Voucher number)
Geographic coordinates

Miras 11
Squalius peloponensis/Miras (NMP PV6 G222) likelihood (ML) were constructed for independent genes in order to assess the phylogenetic position of each population based on mitochondrial and nuclear markers. BI was performed in MrBayes 3.2 [53]. Two simultaneous analyses were run for 10 million generations, each with four MCMC chains sampling every 1000 generations. Convergence was assessed using Tracer v.1.6 [54]. The 50% majority rule consensus tree was constructed after discarding the first 10% of generations as burn-in. For ML analyses, we used RaxML software implemented in the Trex-online server [55], employing the substitution model GTR+G+I for MT-CYB and GTR+G for nuclear genes and the rapid bootstrapping algorithm to estimate node confidence using different random seeds (1000 replicates) [56]. Uncorrected-p genetic distances between mitochondrial haplotypes and between nuclear phased alleles were estimated using the MEGA v.7 software [49].

Molecular clock
To establish a temporal framework for the introgression event, we estimated divergence time among the populations of Squalius keadicus (Evrotas R.) and the non-introgressed and introgressed populations of S. peloponensis using an uncorrelated relaxed molecular clock based on the MT-CYB gene implemented in BEAST v.2.0. [57]. To incorporate a speciation-model tree prior into the analyses, the birth-death model with incomplete taxon sampling [58], we used one specimen per species to estimate divergence time. For S. peloponensis, we used a specimen from the Alfios River (non-introgressed population) and one from the Miras River (introgressed population). As in the phylogenetic analysis, several species of the genus Petroleuciscus were used as outgroup. We calibrated the molecular clock using two fossil species of the genus Squalius (formerly within the genus Leuciscus) as lognormal prior: Leuciscus antunesi Gaudant 1977, found in a Portuguese deposit of the Middle Miocene period (13.5-14.5 Ma; [59][60]) and Leuciscus aff. cephalus from near the Aliakmon River in Greece (Lava 2 deposit) from Upper Miocene (6.56 Ma) [61]. The calibration point based on L. antunesi was allocated in the stem group of the Iberian Squalius species belonging to the Mediterranean lineage, whereas the age of the fossil Leuciscus aff. cephalus was placed in the crown of the current Squalius cephalus group. The MCMC analyses were run for 100 million generations, with parameters logged every 10000 generations. The remaining parameters were default parameters of the software. Convergence of parameters was evaluated using Tracer v.1.6 [54] and results were summarized in TreeAnnotator v.2.0 [57].
After constructing the measurement matrix, Burnaby's method was used to correct for size effect in morphometric variables [63][64]. A canonical variate analysis (CVA) was conducted to identify the variables most explanatory of the morphological variation among the analysed populations with the aim to maximize differences among populations. Non-parametric Kruskal-Wallis and Mann-Whitney post hoc comparisons were used to test for significant differences of morphological variables among populations. All analyses were performed using the Burnaby's corrected matrix. Statistical analyses were carried out in PAST v.3.12 software [65].

Molecular analyses
The result of mitochondrial phylogenetic analyses is shown in Fig 3. The Squalius populations from the Pamissos and Miras basins clustered together, forming a monophyletic clade, sister to S. keadicus. These two clades (S. keadicus and Pamissos/Miras S. peloponensis) were highly supported as monophyletic in the BI and ML analyses. Nuclear genes yielded tree topology inconsistent with that of the mitochondrial marker, with the Pamissos and Miras populations showing a highly supported cluster with S. peloponensis from the Alfios Basin, forming a monophyletic clade (Figs 4, 5 and 6). All phased alleles for the nuclear genes RAG1 and S7 from the Miras and Pamissos populations were nested within the clade of S. peloponensis and none of them was grouped with S. keadicus. No evidence of genetic structure was observed in the nuclear clade comprising the Alfios, Miras, and Pamissos populations.
Uncorrected-p genetic distances between S. keadicus and the introgressed populations (Miras and Pamissos) for the three analysed genes are presented in Table 2. Table C in S1 File shows diagnostic autapomorphies characteristic of each population. For the MT-CYB, 115  Estimated divergence times of the mitochondrial lineage of Squalius keadicus from that of the introgressed populations of S. peloponensis was approximately 0.9 Ma (CI 95% HPD: 0.3-2.2 Ma; Fig 7). The divergence of S. peloponensis from closely related Squalius spp occurred in Middle-Upper Pliocene. However, this phylogenetic relationship was not highly supported, probably as a consequence of an incomplete sampling of the Euroasiatic lineage of this genus sensu [35] (see also [37]).

Morphological analyses
Kruskal-Wallis and Mann-Whitney post hoc analyses demonstrated significant differences among the four analysed Squalius populations in morphometric and meristic variables ( Table 3; Fig 8). Canonical variate analysis (CVA) also supported morphological differences, showing four isolated groups with partial overlap between the populations from the Miras and Pamissos basins and between each of these populations with both the Evrotas (S. keadicus) and Alfios (S. peloponensis) populations. As expected, overlap between S. keadicus and the Alfios S. peloponensis population was not observed (Fig 9). The most contributory morphometric variables to the ordination in the CVA were AFL, BD, and DFL (Table 4). Meristic variables in the Alfios, Pamissos, and Miras populations overlapped with the Evrotas population (S. keadicus) forming a differentiated group in CVA (Fig 9). Kruskal-Wallis and Mann-Whitney post hoc comparisons were significant for LL, LTS, LTI, and A, reaching higher values in S. keadicus relative to the other three populations (Fig 10).
The Evrotas population showed greater BD and longer APL, DFL and AFL fins than the other three populations. It also exhibited significantly shorter HL relative to SL than in other populations ( Table 5). The Pamissos population exhibited the significantly smaller ED relative to HL and SL than seen in the other three populations. The largest ED relative to the standard length was found in the S. peloponensis from Alfios (Table 5). Although the raw data matrix was corrected by Burnaby's method, differences in ED may be attributed to allometric growth rather than to real population differences. The Alfios population had the narrowest head in relation to HL (Table 5), a ratio significantly different from the Evrotas and Miras populations, but not from that of the Pamissos.

Discussion
In this study we tested through morphological and molecular analyses the hypothesis of mitochondrial introgression between two species of the genus Squalius from Greece, S. keadicus   Table 2. Uncorrected-p genetic distances (%) between (below diagonal) and within (in diagonal) populations for the three genes analysed. and S. peloponensis, formulated on the basis of the discordance between mitochondrial (MT-CYB; [34]) and nuclear (allozymes; [33]) genomes found in one population from the Greek Peloponnese (from Miras Basin). The populations from the Miras and Pamissos basins were morphologically similar to S. peloponensis from Alfios Basin, as expected based on sharing of the nuclear genome and the lack of diagnostic nuclear autapomorphies among the three populations (Table C in S1 File), as well as on meristic characters, which fall within the range of S. peloponensis (Fig 10). However, conserved meristic measurements are observed in Squalius species of the Euroasiatic lineage sensu [35] [37]. Although the populations from Miras and Pamissos were recognized as S. peloponensis significant differences between these two populations were found in some morphometric measurements, as well as between S. peloponensis and S. keadicus (Tables 3 and 5; Fig 9).

MT-CYB
Whereas the Miras and Pamissos populations showed the nuclear genome and morphology typical of Squalius peloponensis, their mitochondrial genome clustered with S. keadicus (Evrotas population), but not with the S. peloponensis (Alfios population). This mitonuclear discordance supports the idea of mitochondrial introgression of S. keadicus in the Miras and Pamissos populations. Sharing of similar morphology and the nuclear genome allowed us to reject the hypothesis of Miras and Pamissos populations being hybrids of S. keadicus and S. peloponensis, since they should show similarity to both hypothetical parent species in the nuclear genome as well as in the mitochondrial genome. Therefore, the hypothesis of mitochondrial DNA introgression cannot be rejected based on our results. Hybridization is common in fish, and the boundaries between species may be obscure. In cyprinids, hybridization is a common evolutionary process [26,[66][67][68][69]. In our analyses, the two nuclear markers were S. peloponensis-like in introgressed populations, and we did not find evidence of recombination between species (PHI = 1 for both nuclear markers), as demonstrated by the phylogenetic relationships of phased alleles (Fig 6), and low (0.1-0.2%) genetic distance between populations in both nuclear markers. Hence, we accept the hypothesis of mitochondrial introgression and reject the hypothesis of hybrid speciation.
All analysed specimens from Miras and Pamissos were mitochondrially S. keadicus-like and nuclearly S. peloponensis-like, suggesting unidirectional complete mitochondrial introgression (mitochondrial capture) of S. keadicus to S. peloponensis [34]. Nonetheless, they formed an independent highly supported phylogenetic clade, sister to S. keadicus from the Evrotas population (Fig 3). These two highly supported reciprocal monophyletic groups found in the mitochondrial tree within the phylogenetic lineage of S. keadicus indicate that the introgression event was ancient, but it is not clear whether divergence of these populations, allowing mitochondrial DNA to sort, occurred before or after the introgression event. Ancient mitochondrial captures have also been described for reptiles [70][71], insects [18], and fishes [27][28].
Since these four Squalius populations currently occur in allopatry, a geographical barrier may be considered the primary source of their morphological differences possibly through a process of incipient speciation. Isolation of populations and incipient speciation in sympatry can be attributed to sexual selection [72][73], divergent selection, and phenotypic plasticity [74], or segregation of ecological niches [75][76]. The morphological and molecular differentiation found among the analysed populations of Squalius from western Peloponnese probably occurred after their allopatric separation linked to the formation of the hydrological network in the region during the Upper Pliocene-Pleistocene periods [38,43]. Incipient speciation or occurrence of differentiated lineages during the Pleistocene as a consequence of geographic barriers has been suggested for several taxa [77][78][79]. Examples of isolation-driven divergence to new lineages after ancient hybridization have been reported [19,80]. This is demonstrated here by the lack of shared mitochondrial haplotypes among populations in the rivers of the western Peloponnese. The sharing of the nuclear genome among the Alfios, Miras, and Pamissos populations ascribed to S. peloponensis, clustering in the same clade, also supports the scenario of incipient speciation.
Several biogeographic scenarios may be proposed to explain the mitochondrial introgression event and the current molecular configuration of the four analysed populations (Fig 11)  Pamissos basins were inhabited by the lineage of Squalius keadicus and the mitogenome of these populations diverged from that of the Evrotas population. The lineage of S. peloponensis then invaded the Miras and Pamissos basins and introgression occurred. D) A case similar to scenario C is that in which it was not the mitochondrial lineage of Squalius keadicus that inhabited the Miras and Pamissos basins prior to the invasion by the lineage of S. peloponensis, but a phylogenetically closely related species.
The main difference in these biogeographical scenarios is the timing of the introgression event as before (scenarios A and B) or after (scenarios C and D) the divergence of the S. keadicus from the S. keadicus-like mitogenome. The common factor among the hypothetical scenarios is the biogeographical discordance between mitochondrial and nuclear DNA [4]. Most patterns of biogeographic mitonuclear discordance are characterized by previously isolated lineages that came into contact and generated hybrid zones in which gene flow patterns in mitochondrial and nuclear DNA differed [4]. This hypothesis of secondary contact is consistent with the hypothetical biogeographical scenarios that we propose, which differ only whether the invader species is S. keadicus (A) or S. peloponensis (B, C, and D). Differences in effective population size of the resident and the invader species could be responsible for the unidirectionality of the introgression event found in the present study [2,4,81]. In the initial stages of an invasion event, a small number of invader specimens are expected in comparison to the resident population, and simulated studies have revealed unidirectional introgression of neutral loci from the resident to the invader population [21]. This would imply that S. keadicus (B, C) or S. keadicus-like (D) was the resident species and S. peloponensis the invader. Nevertheless, if the invader becomes more numerous as invasion continues, the direction of the introgression will depend of the demographic balance of the resident and invader populations [21,82]. Mitochondrial DNA introgression as a consequence of extreme differences in population size has been described in other freshwater fish species [75]. In our hypothetical scenarios, S. keadicus or an extinct species phylogenetically similar to S. keadicus, regardless of whether it was the resident or the invader species, must have had a population large enough to erode the mitochondrial genome of S. peloponensis to the point of elimination. Another possible source of mitonuclear discordance may be sex-biased reproduction or sex-biased offspring production during the hybridization event [4]. Empirical experiments are necessary to assess this hypothesis. Differences in distribution range of two species involved in an introgression event when a hybrid zone is formed explain asymmetric mitochondrial introgression, associated with differences in population size and competition between their mitochondrial genomes [4]. It is not possible to estimate the range of S. keadicus or the size of the hybrid zone in the Miras and Pamissos basins, because S. keadicus, or the S. keadicus-like species, is currently extinct in this region and no fossil evidence exists, but it is clear that the distribution range of the studied species overlapped in the past.
Although the mitochondrial genome is generally considered to be neutral [11], positive mitochondrial selection has been described for this marker [20], and is another possible explanation of the observed unidirectional mitochondrial introgression. Thus, the mitochondrial genome with the highest fitness may be introgressed into the other species, regardless of whether it is resident or invader [81]. These selective sweeps have been attributed to climatic adaptation, especially to temperature [83]. Our study area, however, was restricted to a small geographical region with no significant climatic differences in river systems. Furthermore, the HKA test of selection [84] performed on the MT-CYB gene (p>0.05) suggested the potential for neutral evolution in this gene, and a selective sweep is not indicated in the studied species.  The most recent common ancestor of the mitochondrion of the Evrotas S. keadicus and the introgressed populations of the Miras and Pamissos basins dated from 0.9 Ma (CI 95% HPD 0.3-2.2 Ma) in the Pleistocene, and introgression may have occurred before or after divergence. Mitochondrial captures during Pleistocene have been described for other freshwater fish species [28]. The divergence of the mitochondrial clades in the present study may be attributed to Pleistocene tectonic movements related to uplift and river system formation. The paleogeographical history of western Greece and the Peloponnese is complex; the area comprises a network of active faults that constitutes a multifractured neotectonic macrostructure [38]. Tectonic movements occurred throughout the Upper Pliocene-Pleistocene period and continue to the present causing earthquakes and tsunamis [39,85]. The faults had great influence on the drainage network formation [38]. The current hydrological basins of the western Peloponnese were formed during the Pleistocene as a consequence of marine or continental sediment infilling when fault systems were activated. Marine sediments found at several hundred metres elevation in some mountainous areas are evidence of vertical tectonic movements throughout this period [38,86]. These sedimentary events have contributed to configuration of the main fluvial basins of western Peloponnese, which show a radial pattern commonly found in areas of rapid uplifting [43].
The recent paleogeographic events may be responsible for the divergence of mitochondrial lineages of S. keadicus and the introgressed populations (mitochondrial S. keadicus-like) of S. peloponensis as well as morphological differences among the four populations that evolved following the isolation of the hydrological basins (Fig 11). Complex tectonic movements configuring the hydrological network of western Peloponnese may also have been the source of contact between species and the introgression event, through river piracy.

Conclusions
Molecular and morphological evidence supports the hypothesis of mitochondrial introgression of S. keadicus to S. peloponensis. We found unidirectional mitochondrial introgression from S. keadicus into populations from two basins of the western Peloponnese, the Miras and Pamissos River systems, but not into that from the Alfios Basin, and a sorting of the mitochondrial genome of introgressed populations from that of S. keadicus, likely as a consequence of isolation by biogeographical barriers, leading to a process of incipient speciation. Recent secondary contact among basins is the most plausible hypothesis to explain the mitochondrial introgression.
The identification of mitochondrial capture is crucial to understanding the evolutionary history of living organisms, and adequate molecular and morphological analyses are essential to distinguish among evolutionary processes leading to mitonuclear discordance, such as the case of mitochondrial introgression.
Supporting Information S1 File. Table A. Additional species used in phylogenetic performance. GenBank accession numbers and labels in phylogenetic trees. Table B. Evolutionary models estimated by jModelTest for the three analysed genes. Table C. Autapomorphies in the three analysed genes. Ã transversions. (DOCX)