Bidirectional mitochondrial introgression between Korean cobitid fish mediated by hybridogenetic hybrids

Abstract Genomic introgression through interspecific hybridization has been observed in some species of the freshwater fish family Cobitidae. Within this family, a Cobitis hankugensis–Iksookimia longicorpa diploid–triploid hybrid species complex on the Korean peninsula is unique in displaying hybridogenesis, a unisexual reproduction mode that allows hybrids to mediate the transfer of mitochondrial DNA (but not nuclear DNA) between the two parent species. However, populations of the parental species in the wild have never been examined for the potential effect of introgression on their genomes. To address the genetic consequences of unisexual hybridization on the parental species, we examined genetic structure of the two parental species, C. hankugensis and I. longicorpa, in three independent natural habitats where they coexist with their hybrid complex using DNA sequence data of one mitochondrial gene and three nuclear genes. We found that mitochondrial introgression between the two species was extensive in all the examined localities, while there was no evidence of nuclear introgression across the species boundary. This result indicates that the hybridogenetic individuals mediate mitochondrial introgression from one species to the other, producing mito‐nuclear mosaic genomes such as C. hankugensis nuclear genomes associated with I. longicorpa mitochondrial DNA and the reverse. The direction and degree of introgression varied among the three localities, but the underlying mechanisms for this observation proved elusive. Introgression might depend on which species serves as the predominant sperm or ovum donor or the environmental conditions of the localities. The present study suggests that introgressive hybridization between pure C. hankugensis and I. longicorpa species is highly likely where the two species co‐occur with hybridogenetic individuals, but the consequence of introgression could be variable due to the history and environmental characteristics of particular populations across the parental species’ ranges.


| INTRODUC TI ON
Interspecific hybridization can give rise to unisexual hybrid lineages, which consist of females that clonally inherit their genomes without genetic recombination . A variety of unisexual hybrid lineages have been reported in amphibians and reptiles, as well as in fishes, including the families Poeciliidae, Atherinidae, Cyprinidae, and Cobitidae (reviewed in Dowling & Secor, 1997, Vrijenhoek, Dawley, Cole, & Bogart, 1989, Vrijenhoek & Parker, 2009). There are three reproductive models in these unisexual animals: (a) parthenogenesis, where maternal clones are developed without sperm; (b) gynogenesis, where sperm is required to trigger embryogenesis resulting in maternal clones; and (c) hybridogenesis, where fertilization occurs but only the maternal genome is transmitted to the next generation. Depending on the reproductive mode, there may or may not be introgression (gene transfer between parent species via hybrid backcrossing). Parthenogenesis and gynogenesis can effectively prevent genetic introgression between parental species (Choleva et al., ). On the contrary, hybridogenesis can mediate the transfer of mitochondrial DNA (mtDNA) from one species to another without nuclear introgression, thereby producing a mito-nuclear mosaic genome (Perea, Vukić, Šanda, & Doadrio, 2016;Pløtner et al., 2008;Renoult, Geniez, Bacquet, Benoit, & Crochet, 2009;Saitoh, Kim, & Lee, 2004).
The benthic freshwater fish of the family Cobitidae, commonly called spined loaches, are distributed across Europe, Asia, and part of North Africa (Kottelat, 2012). Species of this family are known to undergo unisexual hybridization in the wild. For example, hybrids with various genetic compositions, including triploids and tetraploids, have been reported in European spined loaches, such as Cobitis taenia and its closely related species (Choleva, Apostolou, Rab, & Janko, 2008;Janko et al., 2007). Intensive mating experiments have demonstrated that the hybrids among European spined loaches unisexually reproduce through gynogenesis, in which F 1 hybrid females generate diploid eggs via aberrant cytological pathways.
On the other hand, the interspecific hybridization between the two Korean spined loaches, Cobitis hankugensis Kim, Park, Son, and Nalbant (2003) and Iksookimia longicorpa Kim, Choi, and Nalbant (1976) (Figure 1a), has attracted attention because of its hybridogenetic mode of reproduction, which is unlike that of the European spined loaches. Both of these species are endemic to Korea and are distributed in several rivers and streams flowing southward (Kim, 2009). The ranges of C. hankugensis and I. longicorpa partially overlap in the Nakdong River (Yellow area in Figure 1). However, since habitat preferences of the two species are different, such as current velocity, types of bottom, and depth of the stream, suitable habitats that the two species tend to coexist are very limited within a river.
F I G U R E 1 (a) Morphology of Cobitis hankugensis, Iksookimia longicorpa, and their hybrid complex and (b) their sampling localities in the present study and distribution ranges. Green and orange areas represent the distribution ranges of C. hankugensis and I. longicorpa, respectively. The yellow area indicates the overlapping distribution range of C. hankugensis and I. longicorpa. The black dots represent the sites where the hybrid individuals were observed in the previous studies (Ko, 2009;Lee, 1995Lee, , 1992. Detailed information on the sampling sites (black squares) and the number of sampled individuals are presented in Table 1 For this reason, hybrid complexes consisting of a number of females with a few infertile males were found in three localities (squares in Figure 1) with both parental species (Ko, 2009). The hybrids between the two species are classified into three types depending on their karyotype, one diploid (2n = 49, referred as HL in Figure 1; H, haploid of C. hankugensis; L, haploid of I. longicorpa) and two triploids (3n = 73, HHL; 3n = 74, HLL), while their parental species are diploids (2n = 48, C. hankugensis; 2n = 50, I. longicorpa; Kim & Lee, 1995, 1990. Each of the hybrid types and their parental species also showed clear differences in a few morphological characteristics including the lateral banding patterns (Kim & Lee, 1990;Ko, 2009; Figure 1). The proportion of hybrid types were different among localities (Ko, 2009). Previous crossing experiments suggested that this hybrid complex formed through hybridogenesis by demonstrating that the triploid hybrids produce haploid eggs through meiosis after the genome of one parental species is discarded (Kim & Lee, 1995, 2000. For example, triploid hybrids with HHL and HLL type genomes generate haploid eggs of C. hankugensis and I. longicorpa, respectively (Figure 1). Such a reproduction mode of triploid hybrids has been observed in some kinds of teleost fishes, such as Squalius alburnoides (Alves, Gromicho, Collares-Pereira, Crespo-Lõpez, & Coelho, 2004) and Misgurnus loaches (Morishima et al., 2008).
To date, two hypotheses have been proposed about the reproductive mechanisms between C. hankugensis and I. longicorpa.
First, Kim and Lee (2000) suggested that gene flow does not occur from the triploid hybrids to the parental species. Similarly, hybridogenesis of water frogs (Pelophylax esculentus) has been suggested to be a genetic sink for sexual species. However, a recent study showed evidence that the water frog hybrids were able to mediate mtDNA transfer to its parental species (P. ridibundus) (Mikulïček, Kautman, Demovič, & Janko, 2014). Interestingly with our studying species, Ko (2009) described how triploid hybrid females could transmit genetic material to the parental species by backcrossing ( Figure 2). This second hypothesis was supported by a phylogenetic analysis that showed evidence of mitochondrial introgression: a C. hankugensis individual from a hybrid zone was monophyletic with an I. longicorpa individual in the same locality (Saitoh et al., 2004).
The present study aims to investigate, using population genetic approaches, the influence of hybridogenesis as a mediator of genetic introgression between C. hankugensis and I. longicorpa in their natural habitats. For this purpose, three representative localities were selected within which a relatively high abundance of the hybrid complex between C. hankugensis and I. longicorpa has been maintained with both of their parental species. The genetic structure of the parental species in these three localities was investigated using sequences of one mitochondrial gene and three nuclear genes. TA B L E 1 Sampling information for the present study. The classification of each species is based on morphological characteristics. Each locality is presented in Figure 1 F I G U R E 2 Hypothetical pathways of mitochondrial introgression according to the reproduction mode of the diploidtriploid hybrid complex between Cobitis hankugensis and Iksookimia longicorpa (Ko, 2009) based on the mother species being either (

| Sampling and DNA extraction
Cobitis hankugensis and Iksookimia longicorpa individuals were sampled from three independent localities where they coexist with their hybrids (Figure 1). The sampling sites were selected according to previous investigations of the distribution of C. hankugensis, I. longicorpa, and their hybrids (Ko, 2009;Lee, 1995Lee, , 1992. Each sampling site and the number of samples collected are shown in Table 1. Distinctive morphological features of each parental species and their hybrids were used for identification (Kim, 2009): C. hankugensis has four clearly visible lines (Gambetta's zone) on the side of the body, and I. longicorpa is distinguished by the shape and color of its lateral spots: the first spot is darker than in C. hankugensis and the other spots are elongated toward the ventral side ( Figure 1). The C. hankugensis-I. longicorpa hybrids do not exhibit such morphological features, and they commonly have numerous tiny spots forming an irregular cloud on the dorsal part of the body and a row of oval or rectangular spots running along the lateral sides of the body (Kim & Lee, 1990).
According to the previous artificial mating experiment, genome compositions of offspring with different ploidy and chromosome number precisely corresponded to the morphological characteristics (Ko, 2009). Thus, these morphological features allowed us to unambiguously sort out pure parental species, C. hankugensis and I. longicorpa, from their diploid-triploid hybrids. Only four I. longicorpa individuals from the Deokcheon River (DC) were obtained, despite several attempts to collect more samples, due to river maintenance or construction around the site. Genomic DNA was extracted from the pectoral fin of each specimen using Qiagen Tissue and Blood kits (Qiagen, Hilden, Germany) following the manufacturer's protocols.

| PCR and sequencing
As exchange of mtDNA between Cobitis and Iksookimia via unisexual reproduction was expected (Saitoh et al., 2004), sequences were obtained from one mitochondrial locus (Cytochrome b gene, Cyt b) and three nuclear loci (similar to ectodermal-neural cortex I, ENCI; myosin heavy polypeptide 6, myh6 as described by Li, Ortï, Zhang, and Lu (2007); and Recombination Activating 1, RAG1 as described by Slechtovã, Bohlen, and Tan (2007)). PCR amplification of the Cyt b gene and RAG1 gene was performed as previously described by Perdices and Doadrio (2001) and Slechtovã et al. (2007), respectively. The sequences of the other two nuclear loci (ENCI and myh6) were amplified in 20 μl reactions (2 μl of 10X PCR buffer, 1 μl of 2.5 mM dNTPs, 1 μl of 1X bovine serum albumin, 1 μl of each 10 mM primer solution, 0.12 μl (1 unit) of n-Tenuto Taq DNA polymerase (Enzynomics, Korea), 1 μg of genomic DNA, and 12.88 μl of distilled water). The amplified PCR products were used for additional amplifications in 20-μl PCR reactions that contained 2 μl of 10× PCR buffer, 1 μl of 2.0 mM dNTP, 0.8 μl of each 10 mM nested primer, 0.12 μl (1 unit) of n-Tenuto Taq DNA polymerase, 0.5 μl of template and 14.78 μl of distilled water. In the first PCR, each reaction was performed under the following conditions: initial denaturation at 94°C for 2 min, followed by denaturation at 94°C for 40 s, annealing at 53°C for 40 s, and elongation at 72°C for 1 min, which was repeated for 35 cycles with a final elongation step at 72°C for 5 min. These PCR conditions were slightly modified for the inner PCR, in which the annealing temperature was set at 62°C and the number of cycles was 30. All PCR products were sequenced using an ABI 3730xl automatic sequencer (Applied Biosystems, Foster City, CA).

| Analysis of DNA sequence data
The obtained sequences were aligned using Clustal W (Larkin et al., 2007), and the accuracy of base determination was confirmed using Geneious version 6.1.4 (Biomatters, Auckland, New Zealand).
To solve the phase problem of heterozygotic nucleotide sites in diploid DNA sequences, the sequences of the nuclear genes were processed using the PHASE program (Stephens & Donnelly, 2003;Stephens, Smith, & Donnelly, 2001).
The sequence data were used to determine haplotype using DnaSP v5 (Librado & Rozas, 2009). To confirm the genetic relationships of the haplotypes to C. hankugensis and I. longicorpa, maximum likelihood (ML) trees were additionally searched using PHYML (Guindon et al., 2010) under the best evolution models, which were selected through the Akaike information criterion (AIC) with MEGA 5.0 (Tamura et al., 2011). Branch support values for each node were estimated by bootstrap analysis with 1,000 pseudoreplicates.
Additionally, Cyt b sequences were obtained from Genbank for the two parental species (KM576249, C. hankugensis, KM576235, I. longicorpa), which were collected from distant locations where hybrids between the two parental species have never been observed. These sequences were used as reference sequences in the phylogenetic analyses of the Cyt b gene because of the possibility of mitochondrial introgression through hybridization (Choleva et al., 2014;Saitoh et al., 2004). Minimum spanning trees of the haplotypes were constructed using PopART (Bandelt, Forster, & RøHL, 1999;Leigh & Bryant, 2015).
Pairwise F ST values among the populations of each species were calculated using the Arlequin program to investigate genetic structure within C. hankugensis and I. longicorpa. Additionally, analysis of molecular variance (AMOVA) was performed in Arlequin (Excoffier, Laval, & Schneider, 2005)
The haplotypes were grouped according to monophyletic relationships in maximum likelihood trees for each gene (Figures 3 and 4). All the nuclear gene trees showed two monophyletic haplotype groups that precisely corresponded to the species from which the haplo-   Figure S1). Genetic differentiation estimated using the mitochondrial Cyt b was also generally significant among populations but was not significant between C. hankugensis and I. longicorpa populations within any of the investigated localities (NA,  (Table 3). On the contrary, the genetic differentiation among populations within groups was supported by statistical tests, except for when the testing groups were defined by locality with the mitochondrial Cyt b locus (Φ SC = −0.0362, p = 0.57; Table 3).

| D ISCUSS I ON
Interspecific hybridization between C. hankugensis and I. longicorpa have not been examined in natural populations. In the present study, population genetic structure was investigated in C. hankugensis and I. longicorpa from three natural habitats ( Figure 1 and Table 1) where the two species co-occur with their hybrids.
Results showed that all the haplotypes of the nuclear loci were clearly distinguished by the species from which they were obtained ( Figure 3). In contrast, all of the investigated localities showed either complete or partial inconsistency between the species-types assigned by mitochondrial Cyt b haplotypes and the morphological species identification (Figure 4). Surprisingly, the Cyt b locus exhibited insignificant genetic differentiation between C. hankugensis and I. longicorpa populations within each of the hybrid localities (Tables   2 and 3). Genetic discordance between mitochondrial and nuclear loci can be attributed to genetic introgression through hybridization (Toews & Brelsford, 2012;Wallis et al., 2017). Therefore this observation indicates the existence of mitochondrial introgression between the two parental species. It is unlikely that the mitochondrial genes of two divergent species retain ancestral polymorphism while their nuclear genes exhibit reciprocal monophyly (Avise, 2009;Avise et al., 1987). We therefore rule out the hypothesis that the difference of genetic pattern in Cyt b gene results from incomplete lineage sorting. Moreover, some of the same mitochondrial haplotypes were observed in both of the parental populations within the investigated localities (Supporting Information Table S1). Therefore, these results support the hypothesis of mitochondrial introgression between the two parental species without nuclear gene introgression through backcrossing from the hybridogenetic hybrids to one of the paren-  distributed only in the very limited area in which the two parental species coexist (Figure 1b). If the hybrids reproduce by parthenogenesis or gynogenesis, the hybrids will endure relatively long periods when there is no or a single parental species present. However, the hybridogenetic hybrids between C. hankugensis and I. longicorpa cannot produce offspring without the males of the parental species because male hybrids are infertile. According to the model of hybridogenesis presented in Figure 2, even if the hybrids reproduce with only one of the parental species, the ultimate genome compositions of the hybrids (HL, HHL, and HLL) will be diluted out being gradually replaced with either HH or LL depending on which species is the sperm donor. Therefore, in order for a hybrid complex population to be stably sustained, the hybrids must coexist with both parental species. This constraint may explain the ephemeral nature of the diploid-triploid hybrid species complex in time as well as the patchiness of its contemporary distribution in the Nakdong River like islands. Consequently, it can be presumed that such a hybrid complexes have intermittently formed in places where two parental species came into contact in the past and have subsequently gone extinct due to the cessation of contact. Previous studies (Ko, 2009;Lee, 1995Lee, , 1992 reported that hybrid individuals were observed in some localities where only one parental species co-occurred (black TA B L E 2 Population pairwise F-statistics. The probability of the test statistic p < 0.0042 after Bonferroni correction  A variety of prezygotic and postzygotic factors can influence the frequency and direction of hybridization (Wirtz, 1999). In the case of hybridization between Atlantic salmon (Salmo salar) and brown trout (S. trutta), all of the hybrids in southern European rivers were the product of crosses between female salmon and male trout, due to a fitness disparity in nuclear-mtDNA incompatibility with respect to development time (Álvarez, Garcia-Vazquez, & Taylor, 2011). In contrast, hybrids between male salmon and female trout have been observed in Canadian rivers, possibly due to an abundance of sexually mature Atlantic salmon (McGowan & Davidson, 1992). As the hybridogenetic hybrids between C. hankugensis and I. longicorpa serve as a mediator between the two parental species (Figure 2 Though studies on the ecology and demography of hybrid populations of C. hankugensis and I. longicorpa are scarce, one study was conducted in the NA locality (Ko, 2009), where it was observed that mitochondrial introgression occurred unidirectionally from I. longicorpa to C. hankugensis. According to Ko (2009), the frequency of male C. hankugensis was almost three times higher than that of male  1:2). Therefore, the mating rate between female hybrids and male C. hankugensis is expected to be higher than other possible mating pairs, leading to the backcrossing to one of the parental species, C. hankugensis (Figure 2b). Importantly, the number of HHL hybrids was the highest among three types of hybrids in the NA locality, and HHL hybrids were more morphologically similar to C. hankugensis ( Figure 1). Environmental conditions of the NA locality, including the presence of a sandy bottom, are more suitable for C. hankugensis than I. longicorpa (Ko, 2009

ACK N OWLED G M ENTS
We thank Elizabeth Hassell Kern for careful reading of the earlier version of this paper and constructive suggestions. We also wish to thank Robbie Young for helping us to revise the final version of this paper. This manuscript has greatly benefited from comments by two anonymous reviewers. This research was supported by Basic Science Research Program through the National Research Foundation of Korea (NRF) funded by the Ministry of Education (2010-0024578 and 2013R1A1A2062008).

CO N FLI C T O F I NTE R E S T
The authors declare that they have no competing interests.

AUTH O R CO NTR I B UTI O N S
YSK and YJW designed the experiment. MHK, YSK, YSJ, and HJK collected and identified samples. YSK, YSJ, and HJK generated and analyzed the genetic data. YSK and YJW led the writing of the manuscript. All authors contributed critically to the drafts and gave final approval for publication.

DATA ACCE SS I B I LIT Y
All DNA sequences of this study were deposited at Genbank (MH464456-MH464543).