Recombinant Mitochondrial Genomes Reveal Recent Interspecific Hybridization between Invasive Salangid Fishes

The interspecific recombination of the mitochondrial (mt) genome, if not an experimental artifact, may result from interbreeding of species with broken reproductive barriers, which, in turn, is a frequent consequence of human activities including species translocations, habitat modifications, and climate change. This issue, however, has not been addressed for Protosalanx chinensis and other commercially important and, simultaneously, invasive salangid fishes that were the product of successful aquaculture in China. To assess the probability of interspecific hybridization, we analyzed the patterns of diversity and recombination in the complete mitochondrial (mt) genomes of these fishes using the GenBank resources. A sliding window analysis revealed a non-uniform distribution of the intraspecific differences in P. chinensis with four highly pronounced peaks of divergence centered at the COI, ND4L-ND4, and ND5 genes, and also at the control region. The corresponding divergent regions in P. chinensis show a high sequence similarity (99–100%) to the related salangid fishes, Neosalanx tangkahkeii and N. anderssoni. This observation suggests that the divergent regions of P. chinensis may represent a recombinant mitochondrial DNA (mtDNA) containing mt genome fragments belonging to different salangid species. Indeed, four, highly significant (pairwise homoplasy index test, P < 0.00001) signals of recombination have been revealed at coordinates closely corresponding to the divergent regions. The recombinant fragments are, however, not fixed, and different mt genomes of P. chinensis are mosaic, containing different numbers of recombinant events. These facts, along with the high similarity or full identity of the recombinant fragments between the donor and the recipient sequences, indicate a recent interspecific hybridization between P. chinensis and two Neosalanx species. Alternative hypotheses, including taxonomical misidentifications, sequence misalignments, DNA contamination, and/or artificial PCR recombinants, are not supported by the data. The recombinant fragments revealed in our study represent diagnostic genetic markers for the identification and distinguishing of hybrids, which can be used to control the invasive dynamics of hybrid salangid fishes.


Species Transplantation and Introduction
Wild populations of salangids have markedly declined in recent years due to overexploitation, hydrotechnical constructions, and water pollution [44][45][46][47], which stimulated the development of fish aquaculture (review by Kang et al. [48]). Among salangids, the clearhead icefish P. chinensis and the short-snout icefish N. tangkahkeii are commercially important aquaculture species. In China, they have been successfully transplanted from Lake Taihu (eastern China) to hundreds of lakes and reservoirs across the country [48][49][50][51]. Special artificial breeding techniques have been developed, with fry released into lakes and reservoirs to improve aquaculture yield [52]. The introduction of icefishes has brought significant economic benefits but, simultaneously, jeopardized biodiversity in native assemblages [48,53,54]. Both P. chinensis and N. tangkahkeii are reported to be successful invasive species [47,48,53,[55][56][57].
In Russia, P. chinensis was first recorded from Lake Khanka in 2006 and from the Amur River in 2008 [58], which came as a result of its introduction from Chinese water bodies [13,[59][60][61][62]. P. chinensis is considered a potentially aggressive invasive species preying mainly on larvae or juveniles of indigenous fishes [63]. The same has been observed and raised serious concerns in China [47,48,64]. Indeed, P. chinensis is a successful invader: in around three decades [48] the species has substantially expanded its range from Lake Taihu northward up to the coastal waters of the Sea of Okhotsk (the Amur River estuary, Russia) (https://todaykhv.ru/news/in-areas-of-the-province/30507/ (accessed on 8 April 2022)). The distance between Lake Taihu and the Amur River estuary is almost 3000 km.
Transplantations and introductions of non-endemic species frequently cause interspecific hybridization and genetic introgression between introduced and native species with significant ecological and evolutionary consequences for native populations (for review, see [65][66][67][68][69]). A less frequent phenomenon is the interspecific hybridization between invasive species in non-native areas [70][71][72][73][74][75][76][77] with limited or no hybridization in their native ranges. In this case, the hybrid between the two invasive species shows a potential to drive the emergence of new genetic diversity, thus, reducing the effect of inbreeding depression and increasing the opportunities to rapidly adapt to new environmental conditions, which further contributes to invasion success (e.g., [77,78]).
As a result of the intentional introduction, the possible hybridization of icefishes (along with some other negative environmental impacts) was suggested [48,53], though not verified by genetic approaches so far. The concern is not ungrounded, because, even without a gene flow, invasive hybridization will likely cause detrimental consequences for the interbreeding species [67,96]. Using sufficiently long mt genome fragments (8141 bp), we previously described a detailed architecture of recombinant events due to anthropogenic hybridization between salmonid fishes such as Siberian taimen Hucho taimen and two lenok subspecies, Brachymystax lenok and B. lenok tsinlingensis [92]. In the present work, we analyzed the patterns of nucleotide diversity in complete mt genomes of icefishes, P. chinensis and N. tangkahkeii, as well as other closely related salangid species, and detected clear signals of mtDNA recombination. We argue that the revealed mt recombinants are not experimental artifacts but reflect the interspecific hybridization between the icefishes studied. The data suggest that even a relatively high genetic divergence (up to 8.5% between N. tangkahkeii and P. chinensis) could not have provided any successful reproductive isolation between the species. The results can be useful for the development of responsible aquaculture and biodiversity conservation management practices to minimize the probability of anthropogenic hybridization that can facilitate the invasion process and threaten the resilience of native populations.

Mitochondrial Genomes
Complete mt genome sequences of salangid fishes were accessed from the Genetic Sequence Data Bank [97] (the National Center for Biotechnology Information; https://www.nc bi.nlm.nih.gov/; GenBank Flat File Release 244.0; see Table S1 for accession numbers). The outgroup species, including representatives of the genera Plecoglossus and Retropinna, were selected based on the previous molecular evidence of their close relationship to the family Salangidae [17,19,31] and the screening of nucleotide sequences available in GenBank.

DNA Sequence Analysis
The nucleotide sequences were aligned using the MUSCLE [98] and MAFFT v. 7 [99] software. The DnaSP v. 6 [100], PROSEQ v. 2.9 [101], and MEGA v. 7 [102] programs were used for intra-and interspecific analysis of polymorphism and divergence; MEGA v. 7 [102] was also used for basic phylogenetic analyses. Phylogenetic reconstructions were inferred from the analysis of the complete mt genomes by the maximum-likelihood methods available in IQ-TREE v. 2 [103][104][105]. The TIM2+F+I+G4 model showed the lowest Akaike Information Criterion (AIC; [106]) value (189,144.4265) and the Bayesian information criterion (BIC; [107]) score (189,529.5437); this model was selected for further phylogenetic reconstructions. The ultrafast maximum likelihood bootstrap analysis [108] consisted of 10,000 replicates. The alignments were analyzed for evidence of recombination using the pairwise homoplasy index (PHI) test [109] and various recombination detection methods provided in the RDP4 software [110][111][112][113][114][115][116][117][118]. The sliding window method (see, e.g., [119]) was used to examine the spatial distribution of polymorphism and divergence across the mt genomes studied. In this method, a window of specified length moves over the nucleotide sequence with a fixed size of step and the corresponding estimates are computed over the data in the window. The obtained estimates are assigned to the nucleotide position at the midpoint of the window. The output of the sliding window analysis can be presented graphically; the values of variation are plotted against the nucleotide position.

Nucleotide Diversity and Divergence
A phylogram of the P. chinensis mt genomes along with those of other salangid fishes is displayed in Figure 1. The tree shows the P. chinensis specimens forming a single clade with two significantly different (100% bootstrap support) groups of sequences including (1) MW291629, KP306787, and HM106494, and (2) MH330683 and KJ499917 ( Figure 1). The π value is quite high (0.0127 ± 0.0007). Despite the fact that it seems still below the range of divergence between species (e.g., [120]), it is, nevertheless, significantly higher than the level of intraspecific diversity detected for many other freshwater and marine fish species including, e.g., the taimen Hucho taimen (0.0010 ± 0.0002; [92]), the European whitefish Coregonus lavaretus (0.0025 ± 0.0003; [121]), the lake whitefish C. clupeaformis (0.0008 ± 0.0001; [121]), the houting C. oxyrinchus (0.0029 ± 0.0003; [121]), the Atlantic herring Clupea harengus (0.0061 ± 0.0008; [122]), the spotted wolffish Anarhichas minor (0.0005 ± 0.0001; [123]), and the northern wolffish Anarhichas denticulatus (0.0006 ± 0.0001; [123]). across the mt genomes studied. In this method, a window of specified length moves over the nucleotide sequence with a fixed size of step and the corresponding estimates are computed over the data in the window. The obtained estimates are assigned to the nucleotide position at the midpoint of the window. The output of the sliding window analysis can be presented graphically; the values of variation are plotted against the nucleotide position.

Nucleotide Diversity and Divergence
A phylogram of the P. chinensis mt genomes along with those of other salangid fishes is displayed in Figure 1. The tree shows the P. chinensis specimens forming a single clade with two significantly different (100% bootstrap support) groups of sequences including (1) MW291629, KP306787, and HM106494, and (2) MH330683 and KJ499917 ( Figure 1). The π value is quite high (0.0127 ± 0.0007). Despite the fact that it seems still below the range of divergence between species (e.g., [120]), it is, nevertheless, significantly higher than the level of intraspecific diversity detected for many other freshwater and marine fish species including, e.g., the taimen Hucho taimen (0.0010 ± 0.0002; [92]), the European whitefish Coregonus lavaretus (0.0025 ± 0.0003; [121]), the lake whitefish C. clupeaformis (0.0008 ± 0.0001; [121]), the houting C. oxyrinchus (0.0029 ± 0.0003; [121]), the Atlantic herring Clupea harengus (0.0061 ± 0.0008; [122]), the spotted wolffish Anarhichas minor (0.0005 ± 0.0001; [123]), and the northern wolffish Anarhichas denticulatus (0.0006 ± 0.0001; [123]). The species name P. hyalocranius was used for the originally published mt genome KJ499917 [30]. However, P. hyalocranius is a synonym of P. chinensis [19]. Consequently, we use the species name P. chinensis for the mt genome KJ499917 to avoid any confusion. The P. chinensis KP306787 genome is indicated as "Unverified" in GenBank. However, this mt genome shows very close affinity to all other P. chinensis genomes; the p-distance between this mt genome and the other P. chinensis mt genomes is 0.0111 ± 0.0007, which is even lower than the average p-distance for the rest of the P. chinensis mt genomes (0.0126 ± 0.0007). Consequently, we included the KP306787 mt genome in further analysis. are used as outgroup. The species name P. hyalocranius was used for the originally published mt genome KJ499917 [30]. However, P. hyalocranius is a synonym of P. chinensis [19]. Consequently, we use the species name P. chinensis for the mt genome KJ499917 to avoid any confusion. The P. chinensis KP306787 genome is indicated as "Unverified" in GenBank. However, this mt genome shows very close affinity to all other P. chinensis genomes; the p-distance between this mt genome and the other P. chinensis mt genomes is 0.0111 ± 0.0007, which is even lower than the average p-distance for the rest of the P. chinensis mt genomes (0.0126 ± 0.0007). Consequently, we included the KP306787 mt genome in further analysis.

Sliding Window Analysis
A sliding window analysis has revealed a non-uniform distribution of the nucleotide diversity values along the P. chinensis genomes characterized by four pronounced peaks centered at the (1) COI, (2) ND4L-ND4 (hereinafter referred to as ND4 for simplicity), and (3) ND5 genes, and also at the (4)  (fourth peak, CR; midpoint 15,763). The divergence values detected at the HD regions are significantly higher than the values usually observed on the intraspecific level (see above) and approach the intergeneric values, e.g., for scorpion fishes [120,124]. A more detailed interpretation for the distribution of divergent regions is considered below (Section 3.3). centered at the (1) COI, (2) ND4L-ND4 (hereinafter referred to as ND4 for simplicity), and (3) ND5 genes, and also at the (4) control region (CR) (Figure 2a). Comparisons including different pairs of the P. chinensis genomes have revealed distinct patterns with one to four strong peaks (Figure 2b-d,f-h), or no peaks at all (Figure 2e). The π values in the valley regions are low (varying from 0 to 0.0117), but follow much higher values in the high divergence (HD) regions: 0.0417 (first peak, COI; midpoint in the alignment 5631), 0.1217 (second peak, ND4; midpoint 11,557), 0.1100 (third peak, ND5; midpoint 13,096), and 0.1083 (fourth peak, CR; midpoint 15,763). The divergence values detected at the HD regions are significantly higher than the values usually observed on the intraspecific level (see above) and approach the intergeneric values, e.g., for scorpion fishes [120,124]. A more detailed interpretation for the distribution of divergent regions is considered below (Section 3.3). The basic local alignment search tool (BLAST) procedure [125] has revealed a surprisingly high similarity (99-100%) between P. chinensis (KJ499917) and N. tangkahkeii (KP170510) for the HD regions 1 (COI), 2 (ND4), and 4 (CR); between P. chinensis (MH330683) and N. tangkahkeii (KP170510) for the HD region 2 (ND4); and between P. chinensis (HM106494) and N. anderssoni (HM106492) for the HD region 3 (ND5). Table 1 provides values of pairwise distances (D xy ± SE) between P. chinensis and these closely related salangid species, and shows a number of outlier values obtained for the full mt genomes and the HD regions separately. On the intraspecific level, the pairwise distances between the full mt genomes of P. chinensis vary within a fairly broad range (0.3-1.9%), which is less pronounced (0.1-0.4%) for the mt genomes with the HD regions deleted (Table 1, last column). When the HD regions are considered separately, the pairwise distances vary in a broader range: from zero to 11.6% (ND4 and ND5 regions), between 0.7-6.4% (COI region), and up to 12.9% (CR region) ( Table 1). On the interspecific level, the pairwise distance fluctuations are even more pronounced. For instance, the full mt genome divergence between P. chinensis and N. anderssoni amounts to 5.5% with a range from 0.9% (ND5 region) to 14.1% (CR region). For the P. chinensis-N. tangkahkeii pairwise comparisons, the average full mt divergence amounts to 8.5%. However, the D xy value between these species is 85.0-fold lower (0.1%) for the ND4 HD region. Moreover, for the COI and CR HD regions, there are no differences between P. chinensis (KJ499917) and N. tangkahkeii (KP170510) ( Table 1). The pairwise distances inferred from the mt genomes with the HD regions deleted (Table 1, last column) show much lower fluctuations, ranging from 4.4 to 4.5% and from 8.6 to 8.7% for the P. chinensis-N. anderssoni and P. chinensis-N. tangkahkeii comparisons, respectively.
Thus, the HD regions of the P. chinensis genomes demonstrate an unexpectedly high similarity (ND4 and ND5) or even full identity (COI and CR) to N. tangkahkeii and N. anderssoni, and can be explained by a recombination of mtDNA (Section 3.3). The four HD regions (COI, ND4, ND5, and CR) show sharply discordant phylogenetic signals be-Life 2022, 12, 661 7 of 20 tween P. chinensis and the Neosalanx species. As a consequence, the position of P. chinensis is sharply different, depending on the fragments used for tree reconstruction (Figure 3). The trees inferred from the HD regions separately show P. chinensis as identical (or very similar) to N. tangkahkeii or to N. anderssoni, respectively (Figure 3a-d). In the tree without the HD regions (Figure 3e), the P. chinensis mt genomes (KJ499917, MH330683, HM106494, KP306787, and MW291629) are within a single cluster showing a value of π (0.0033 ± 0.0003) similar to that of the mt genomes of other fishes (Section 2.1). The other genera included in this analysis (Salangichthys, Plecoglossus, and Retropinna) do not show any visible discordance in the level of divergence between the HD regions and the rest of the mtDNA (Figure 3a-e). The presence or absence of the N. tangkahkeii and N. anderssoni fragments in the P. chinensis mt genomes erroneously increases or decreases the pairwise distances depending on their combination in particular comparison ( Table 1). The same interpretation is suitable to explain the fluctuations in pairwise distances (Table 1).

Recombination
Based on the above-presented results, we suggest that the HD regions ( Figure 2) and the respective phylogenetic inconsistencies (Figure 3) can be explained by recombination of the P. chinensis mt genome. Indeed, the PHI-test [109] has revealed very good evidence of recombination (p < 0.00001) in the alignment including the P. chinensis, N. tangkahkeii, and N. anderssoni mt genome sequences. We, therefore, have analyzed the mtDNA alignments for evidence of recombination using various recombination detection methods implemented in the RDP4 program ( [110]; Table 2). Six methods have detected four recombination events in the P. chinensis KJ499917, MH330683, and HM106494 mt genomes with high statistical support (Table 2, Figure 4). The breakpoint positions (Table 2) match closely the coordinates of the HD regions (see Figure 2). The recombinant fragments are

Recombination
Based on the above-presented results, we suggest that the HD regions ( Figure 2) and the respective phylogenetic inconsistencies (Figure 3) can be explained by recombination of the P. chinensis mt genome. Indeed, the PHI-test [109] has revealed very good evidence of recombination (p < 0.00001) in the alignment including the P. chinensis, N. tangkahkeii, Life 2022, 12, 661 8 of 20 and N. anderssoni mt genome sequences. We, therefore, have analyzed the mtDNA alignments for evidence of recombination using various recombination detection methods implemented in the RDP4 program ( [110]; Table 2). Six methods have detected four recombination events in the P. chinensis KJ499917, MH330683, and HM106494 mt genomes with high statistical support (Table 2, Figure 4). The breakpoint positions (Table 2) match closely the coordinates of the HD regions (see Figure 2). The recombinant fragments are not fixed, and the recombinant mt genomes are mosaic, containing different numbers of recombinant events: three for KJ499917, one for MH330683, and one for HM106494 (Figure 4). No recombination events were detected for the P. chinensis KP306787 and MW291629 mt genomes. The sequences involved in the recombination and the breakpoints were identified using the RDP4 suite [110] which incorporates the algorithms RDP (R), GENECONV (G), BOOTSCAN (B), MAXCHI (M), CHIMAERA (C), SISCAN (S), and 3SEQ (Q) (for references, see the Material and Methods section). For each putative recombination breakpoint, a Bonferroni correction P-value was calculated. For each recombination event involving the COI, ND4, and ND5 gene regions, and also CR, the analysis was based on the alignment including three full mitochondrial genomes, as indicated in the column "Recombination event" (the length of the recombinant fragment is in parentheses). For the ND4 gene region, two separate analyses were carried out, including the Protosalanx chinensis genomes KJ499917 (Recombinant 1) and MH330683 (Recombinant 2) as recombinant sequences with the same major and minor parents; conservative P-values are presented. The recombination events are arranged in the order as they occurred along the alignment. The breakpoint positions match closely the coordinates of the HD regions (see Figure 2). The COI recombinant fragment (P. chinensis, KJ499917) starts 73 bp downstream of the COI start codon and ends 1338 bp upstream of the COI stop codon. The ND4 recombinant fragment (P. chinensis, KJ499917 and MH330683) starts 78 bp downstream of the ND4L start codon and covers the full ND4 gene, tRNA-His, and 47 bp of the tRNA-Ser. The ND5 recombinant fragment (P. chinensis, HM106494) starts 976 bp downstream of the ND5 start codon and ends 423 bp upstream   Table 2.
For the COI and CR recombinant fragments, there is 100% identity between the recombinant (P. chinensis, KJ499917) and the minor (N. tangkahkeii, KP170510) parent sequences (Figure 5a Table 2. For the COI and CR recombinant fragments, there is 100% identity between the recombinant (P. chinensis, KJ499917) and the minor (N. tangkahkeii, KP170510) parent sequences (Figure 5a,e). However, for the ND4 recombinant fragment, there are two mismatches at sites 10,174 and 10,185 (highlighted in bold in the alignment; Figure 5b,c) between the recombinant (P. chinensis, KJ499917, MH330683) and the minor (N. tangkahkeii, KP170510) parent sequences. Similarly, for the ND5 recombinant fragment there are four mismatches at sites 12,954, 12,991, 12,997, and 13,030 between the recombinant (P. chinensis, HM106494) and the minor parent (N. anderssoni, HM106492) sequences (Figure 5d). See also Texts S1-S3 for the alignment of the variable sites in the recombinant and parent mt genome sequences.
Taking into account the patterns of recombination revealed by the RDP4 program (Table 2, Figure 4), we can now confidently interpret the distribution of the HD regions in pairwise comparisons of the P. chinensis genomes obtained by the sliding window analysis (Figure 2) in the following way. The plot (a), inferred from the total data for the five P. chinensis mt genomes, shows four HD regions corresponding to the four recombination events detected for the three P. chinensis mt genomes: three for KJ499917 (COI, ND4, and CR), one for MH330683 (ND4), and one for HM106494 (ND5) . The plots (b,c,h) showing one, two, and three HD regions, respectively, reflect the COI, ND4, and CR recombination events between the P. chinensis and N. tangkahkeii mt genomes. The plot (f) with a single HD region reflects the ND5 recombination event involving P. chinensis and N. anderssoni. The plots (d,g) with four and two HD regions, respectively, reflect the COI, ND4, ND5, and CR recombination events involving P. chinensis and both N. tangkahkeii and N. anderssoni. The sliding window plot (e) shows the variability distribution for the two sequences (KP306787 and MW291629) lacking the recombinant fragments and, correspondingly, lacking any HD regions.
tween the recombinant (P. chinensis, KJ499917, MH330683) and the minor (N. tangkahkeii, KP170510) parent sequences. Similarly, for the ND5 recombinant fragment there are four mismatches at sites 12,954, 12,991, 12,997, and 13,030 between the recombinant (P. chinensis, HM106494) and the minor parent (N. anderssoni, HM106492) sequences (Figure 5d). See also Texts S1-S3 for the alignment of the variable sites in the recombinant and parent mt genome sequences.  Table 2 and Figure 4). The two m matches in the ND4 alignment and the four mismatches in the ND5 alignment are highlighted bold. The ND4 recombination event is represented in two parts, illustrating the (b) beginning an (c) ending breakpoint regions, respectively. The alignments of variable sites in the ND4 recombina fragment are identical for the P. chinensis KJ499917 and MH330683 recombinant sequences. The po ymorphic sites 10,391-11,558 of the ND4 region alignment, containing no differences between t recombinant and the minor parent sequences, were deleted to fit onto the page.
Taking into account the patterns of recombination revealed by the RDP4 progra (Table 2, Figure 4), we can now confidently interpret the distribution of the HD regions pairwise comparisons of the P. chinensis genomes obtained by the sliding window analys (Figure 2) in the following way. The plot (a), inferred from the total data for the five chinensis mt genomes, shows four HD regions corresponding to the four recombinatio events detected for the three P. chinensis mt genomes: three for KJ499917 (COI, ND4, an  Table 2 and Figure 4). The two mismatches in the ND4 alignment and the four mismatches in the ND5 alignment are highlighted in bold. The ND4 recombination event is represented in two parts, illustrating the (b) beginning and (c) ending breakpoint regions, respectively. The alignments of variable sites in the ND4 recombinant fragment are identical for the P. chinensis KJ499917 and MH330683 recombinant sequences. The polymorphic sites 10,391-11,558 of the ND4 region alignment, containing no differences between the recombinant and the minor parent sequences, were deleted to fit onto the page.

Discussion
Recombination of the mt genome can reflect interspecific hybridization; however, as an alternative hypothesis, it can be an artifact produced by polymerase chain reaction (PCR) or other errors. Below, we consider the most relevant alternative hypothesis which can explain the results observed in the present work.
Species identification of salangid fishes remains a serious challenge (Section 1.1). Consequently, one can argue that the GenBank data for P. chinensis and the close relatives, which we used for the present analyses, are wrong. Indeed, there are some relatively rare examples of incorrect taxonomic identification of fishes detected in GenBank [126][127][128]. However, all five P. chinensis mt genomes form a single cluster; the sequence divergence between them (except recombinant fragments) is low, 0.3% (Section 2.2; Figures 1 and 4e), which closely matches the values of intraspecific diversity reported for many freshwater and marine fishes (Section 2.1). Furthermore, all the P. chinensis recombinant (KJ499917, MH330683, and HM106494) and non-recombinant (KP306787 and MW291629) mt genomes show close affinity to other salangid fishes, both in our analysis (Figure 1) and other publications [31,129]. Thus, the available data suggest that taxonomical misidentification is not responsible for the evidence of recombination obtained in our work.
Even if the taxonomic identification of P. chinensis is not doubted, there is still a risk of artificial recombination generated by PCR errors if more than one template were present in the PCR. In this case, the polymerase may jump from one template to another during the PCR, thus, producing artifactual recombinants [130]. This effect could be suggested for the ND5 recombination involving the P. chinensis (HM106494) and N. anderssoni (HM106492) mt genomes, both investigated by Li et al. [31]. The multi-template PCR containing the P. chinensis DNA and N. anderssoni contamination DNA could generate PCR artifacts. However, if this recombination had occurred through PCR jumping, the exchanged regions found in two different sequences should have been exact copies of one another. An examination of the recombinant P. chinensis (HM106494) sequence has revealed no exact matches with the minor parent sequence (N. anderssoni; HM106492) for the ND5 recombinant event.
In particular, there are four mutations distinguishing the recombinant and minor parent sequences (Figure 5d), which makes the "jumping PCR" hypothesis highly unlikely. A similar pattern has been revealed for the ND4 recombinant event (Figure 5b) with two mismatches along the recombinant sequence distinguishing the recombinant (P. chinensis; KJ499917, MH330683) and minor parent (N. tangkahkeii; KP170510) sequences. Thus, the imperfect homologies between the recombinants and minor parents for the ND4 and ND5 regions show that these recombinants are real and probably occurred some time ago, due to a historical hybridization between the icefishes.
The COI and CR recombinant sequences are identical to the minor parent sequences and could be explained by PCR jumping. Both recombinants are detected in the P. chinensis mt genome KJ499917 obtained by Lu et al. [30] (published online, 4 April 2014) from the Heilongjiang River Fisheries Research Institute, Chinese Academy of Fishery Sciences, Harbin, China. The fish sample was collected from Lake Xingkai (Khanka). However, the N. tangkahkeii mt genome KP170510, which represents a minor parent mt genome for the COI and CR recombination events, was sequenced by Zhong et al. [32] (published online 19 February 2015) from the Freshwater Fisheries Research Institute of Jiangsu Province, Nanjing, China. The fish sample was collected from Lake Taihu, Jiangsu Province, China. Thus, the COI and CR recombinants cannot be results of DNA contaminations of the P. chinensis PCR amplifications by the N. tangkahkeii DNA because these mt genomes were sequenced by different researchers at a different time and in different laboratories.
Another argument against PCR jumping is exemplified by two P. chinensis mt genome sequences, KJ400017 [30] and MH330683 [129], which carry out the identical recombination event including the ND4 region (Figure 4a,b and Figure 5b,c). The recombination involves precisely the same region in both mt genomes (Table 2) which, however, were detected by different authors at different laboratories and in different years. The P. chinensis recombinant mt genome sequence KJ400017 was obtained by Lu et al. [30] (see the author's affiliation and the fish sampling point above), but the P. chinensis recombinant sequence MH330683 was obtained by Liu et al. [129] from the Freshwater Fisheries Research Center, Chinese Academy of Fishery Sciences, Wuxi, China. The fish sample was collected from Lake Taihu, Wuxi City, China. It is difficult to imagine how this identical recombination could be obtained simultaneously and erroneously by different authors in different places and at different times.
It might also be suggested that the evidence of recombination obtained in the present study is caused by a problem with the tests of recombination due to, e.g., misalignments, which are a common cause of false positive recombination signals [110]. This explanation could be particularly relevant for the recombination detected within CR, which is the most variable part of the mt genome frequently characterized by multiple insertions, deletions, and the complex arrangement of tandem repeats [131]. However, the recombinant (P. chinensis, KJ499917) and minor parent (N. tangkahkeii, KP170510) sequences are 100% identical in the CR recombinant fragment, which provides fully unambiguous alignment and implies that the CR recombination that we have detected is unlikely to be the product of misalignment.
The above arguments strongly suggest that the mt genome recombinants detected in our work are not experimental errors or artifacts and, consequently, their formation requires assuming an interspecific hybridization between P. chinensis and the other two related salangid fishes, N. tangkahkeii and N. anderssoni. The necessary prerequisites for the mt genome recombination to occur, including paternal leakage and heteroplasmy, are not investigated for the salangids, but are well documented for interspecific hybrids of fishes (e.g., [94,132]) and many other organisms (reviewed by Ladoukakis and Zouros [84] and Parakatselaki et al. [85]). Mitochondrial fusion has been demonstrated in diverse organisms, from yeast to mammals including humans (review by Westermann [133]). Experiments testing the complementation between different mtDNA molecules after fusion have revealed the presence of an extensive and continuous exchange of genetic contents between mitochondria in mammalian cells [134]. Experimental data also show that mitochondria possess all the necessary molecular mechanisms for mtDNA recombination (review by Chen [135]).
The present data complement a number of examples of the mtDNA recombination reported previously for plants, fungi, protists, and animals (review in [82,84,95,[135][136][137][138][139][140][141][142][143]). Estimation of hybridization events based on recombination can be considered a conservative one because, at least in animals, an overwhelming amount of mtDNA is maternally transmitted, and the paternal leakage is restricted to a very low amount [84]. The P. chinensis mtDNA recombination was detected in three out of the five mt genomes studied, thus, suggesting relatively frequent interspecific hybridization in salangid fishes, as it was detected, e.g., in hybrid swarms of other invasive species ( [73]; see below).
Different species of salangid fishes frequently occur in sympatry [19,48,144]. Nevertheless, the probability of interbreeding between salangids is apparently low or negligible due to, e.g., different microhabitat preferences (such as salinity tolerances) or a shift of spawning season [9,48]. Indeed, Zhang et al. [19] reported a relatively high genetic divergence (based on the CytB gene) between salangid species and suggested that these fishes have "evolved powerful sorting mechanisms to maintain interspecific isolation" ( [19], page 338). However, temporal or spatial reproductive isolation that is likely to occur in native populations (due to, e.g., habitat-based segregation at spawning grounds; [145]) can be lost in the non-native ranges, resulting in an extensive hybridization between invasive species in the absence of environmental cues present in their native range [73,78] (review by Scribner et al. [146]). Due to anthropogenic habitat disturbance, it may be that reinforcement does not produce enough efficient mechanisms to provide reproductive isolation for divergent species that naturally occur in sympatry [147].
Among salangids, P. chinensis and N. tangkahkeii exhibit the widest distribution and the highest ecological plasticity (review by Kang et al. [48]). After being introduced into a new habitat outside the natural distribution range, these species demonstrate flexible life-history traits, especially in their growth and reproductive modes [48,55], which are frequently observed in invasive species (e.g., [148][149][150][151]). The high reproductive plasticity and the lack of native environmental cues that would provide reproductive isolation may have contributed to the hybridization between P. chinensis and the Neosalanx species, as has been shown for other invasive fishes invading non-native ranges [70][71][72][73][74][75]77,78].
Interspecific hybridization between invasive species can be exemplified by one exceptionally problematic case: the introduction of silver carp (Hypophthalmichthys molitrix) and bighead carp (H. nobilis) to the complex system of braided watercourses in the Mississippi River Basin, United States [71][72][73]. These two species currently remain genetically isolated and do not crossbreed within their native ranges in China, but their introduction to a new habitat initiated an unprecedented frequency of interspecific hybridization leading to a hybrid swarm with early-generation hybrids driving the range expansion and contributing to their invasion success, which caused multiple ecological, evolutionary, and economic problems [73,77,78]. The introduction of invasive salangids into the new habitats of Lake Khanka and the Amur River Basin, located at the border between China and Russia [13,[58][59][60][61][62][63], might have a similar hybridization potential, as is evidenced by the data obtained in the present work. For instance, three recombinant fragments were detected in the mt genome KJ400017 (Figure 4a) of P. chinensis sampled from Lake Khanka [30].
The genetic data obtained in the present study show clear signals of recombination in the mt genomes of P. chinensis containing four mtDNA fragments from two related salangid species, N. tangkahkeii and N. anderssoni. The recombinant fragments are not fixed, and the different mt genomes of P. chinensis are mosaic, containing a different number of recombinant events. These observations, along with the full identity of the COI and CR recombinant fragments between the donor and the recipient sequences, indicate the contemporary interspecific hybridization between P. chinensis and N. tangkahkeii, which can be at least partly explained by human-mediated activities involving the transplantation and introduction of the icefishes. Among the protein-coding mt genes, COI (along with CytB) exhibits the highest evolutionary rates, which are two-five-fold higher for the noncoding CR region [152]. These regions (COI and CR) in case of historical hybridization are expected to rapidly gain new mutations distinguishing them from the parental sequences. The ND4 and ND5 recombination events show two and four mismatches, respectively, between the recombinant and parent sequences (Figure 5b,d), which may indicate a historical hybridization between P. chinensis and the two related salangid fishes, N. tangkahkeii and N. anderssoni. Further evidence from biparentally inherited nuclear DNA is required to critically evaluate the revealed patterns and, in particular, to determine the fraction of the extant P. chinensis genome that was affected by the gene flow from the Neosalanx species, and the time when it happened.
The four recombinant fragments, including the COI, ND4, and ND5 gene regions, and also the CR, are specific for distinguishing hybrids between P. chinensis and the two Neosalanx species. Two of them, COI and CR, that show full identity between P. chinensis and N. tangkahkeii, can be used as diagnostic markers to monitor the salangids' humanmediated hybridization dynamics and their invasion fronts. Extensive spatial and temporal genetic sampling from both native and invasive ranges, including the Amur River Basin and Lake Khanka, should be carried out to determine the full extent of the current range of pure salangid species and their hybrids, which will help to elaborate the most appropriate aquaculture regimes to minimize the negative effects of the species introduction and transplantation on the local biodiversity and ecosystem resilience.