Origins of two hemiclonal hybrids among three Hexagrammos species (Teleostei: Hexagrammidae): genetic diversification through host switching

Abstract Two natural, hemiclonal hybrid strains were discovered in three Hexagrammos species. The natural hybrids, all of which were females that produced haploid eggs containing only the Hexagrammos octogrammus genome (maternal ancestor; hereafter Hoc), generated F1 hybrid‐type offspring by fertilization with haploid sperm of Hexagrammos agrammus or Hexagrammos otakii (paternal species; Hag and Hot, respectively). This study was performed to clarify the extent of diversification between the two hybrids and the maternal ancestor. Genealogical analysis using mtDNA revealed that all 38 Hoc/Hot hybrids formed a branch (Branch I) with 18 of the 33 Hoc/Hag hybrids. No haplotype sharing was observed with the maternal ancestor. Further, microsatellite DNA analysis suggested that the members of Branch I shared the same hemiclonal genome set. The results suggested that Hoc/Hot hybrids originated by anomalous hybridization, or “host switching,” between Hoc/Hag and Hot, and not from interspecific hybridization between Hoc and Hot. The remaining 9 of 11 Hoc/Hag haplotypes and all of the 27 Hoc haplotypes were mixed within the genealogical tree, as if they had originated from multiple mutations. However, Hoc/Hag could also mate with Hoc. Although offspring from this host switch (Backcross‐Hoc) have the same genome as normal Hoc, a part of their genome retains genetic factors capable of producing hemiclones. Consequently, when a descendant of a BC‐Hoc hybrid mates with Hag males, a new hemiclone lineage will arise. Multiple haplotype revival through host switching from a single mutation in hybrids is another possible hypothesis for the observed mixing of Hoc/Hag haplotypes within the mtDNA genealogical tree.

In clonal modes of unisexual reproduction, females produce unreduced diploid or triploid eggs by different cytogenetic mechanisms that develop normally without any biological or genetic contribution from males, and no male offspring are produced (Dawley, 1989).
Unlike sexual reproduction, in which there is the added cost of producing of males, unisexually reproducing organisms do not incur these additional costs during reproduction (Maynard Smith, 1978).
Male hybridogenesis in which the maternal genome is discarded has also been reported in the Pelophylax water frog complex (Lehtonen, Schmidt, Heubel, & Kokko, 2013) and the Australian carp gudgeon (Schmidt et al., 2011).
In hybridogenesis, although females produce genetically identical haploid eggs without any genetic recombination, genetic variation is maintained by renewal of the paternal genome every generation. In this way, hybridogenesis compensates for the costs associated with sexual reproduction while retaining some of the benefits of clonal reproduction (Vrijenhoek, 1994). These advantages, despite involving a unisexual mode of reproduction, should enable hemiclonal animal lineages to remain viable for longer than clonal lineages.
Hybridogenesis, gynogenesis, and parthenogenesis are all considered to have originated from hybridization between different species (Lamatsch & Stöck, 2009;Vrijenhoek, Angus, & Schultz, 1977). Speciation in two geographically separated populations can occur when a contiguous population is separated by a vicariant event of some kind. Under such conditions, genetic differences gradually arise between the separated populations, often resulting in what is referred to as allopatric speciation (Coyne & Orr, 2004). In such cases, if secondary contact occurs before premating reproductive isolation has fully developed, natural hybrids will appear (Barton & Hewitt, 1989).
Although most hybrids typically have low fitness and low reproductive viability due to the inherent incompatibility of the genomes from different species, in some instances, hybrids may be able to survive by employing unisexual reproduction without the recombination of genomes (Ellstrand et al., 2010).
For example, hemiclonal reproduction has recently been reported in two Hexagrammos hybrid strains (Kimura-Kawaguchi et al., 2014). The natural hybrids produce haploid eggs containing only the Hexagrammos octogrammus genome (maternal ancestor) and generate F 1 hybrid-type offspring by fertilization with the haploid sperm of either Hexagrammos agrammus or Hexagrammos otakii (paternal species); in this way, the genome set of the natural hybrids is composed of a hemiclonally transmitted maternal genome and a recombined paternal genome. Similarly, because the second generations of a backcross between natural hybrids and paternal species reproduce by hybridogenesis in the same way as the maternal generation of the natural hybrids, hemiclonal reproduction is maternally inherited over successive generations by backcrossing with paternal species. In addition, Kimura-Kawaguchi et al. (2014) also found that artificial F 1 hybrids produced by crossing pure species generated recombinant gametes, suggesting that although the artificial F 1 hybrids have the same genome composition as hemiclonal hybrids, hemiclonal hybrids do not always result from a hybridization event. In addition, hemiclonal hybrids exhibit genetic differences (mutations) that do not occur in wild-type parental species.
Hexagrammos hybridization is the only known hybridogenetic system in marine fishes inhabiting the North Pacific Ocean. The potentially low extinction potential of these Hexagrammos hybrids is considered to be due to the diversity of habitats, and the longevity, structure, and fluctuation in populations of these species would likely differ from (hemi)clonal organisms distributed in more restricted environments, such as rivers and ponds. The present study was conducted to clarify the origin and diversification of two Hexagrammos hybrids and the maternal parent species (H. octogrammus) using maternal inheritance markers.

| Fish sampling and species identification
For genealogical analysis of the two natural, hemiclonal, hybrid strains,

H. octogrammus/H. agrammus (Hoc/Hag) and H. octogrammus/H. otakii
(Hoc/Hot), and the maternal ancestor H. octogrammus (Hoc), fishes were captured using gill nets and traps on a coastal reef off Usujiri, Japan, from 2004 to 2010 (Fig. 1). Specimens were identified based on diagnostic external morphological diagnostic characteristics, such as the number of lateral lines, flap pairs, and the caudal fin shape, following Nakabo (2000) and Shinohara (1994), as described previously (Kimura-Kawaguchi et al., 2014) In addition, to estimate the allele frequencies for the paternal species of the two natural hybrid strains using microsatellites, 33 H. agrammus and 34 H. otakii were captured using hand nets while SCUBA diving on a coastal reef off Usujiri, Japan, from 2010 to 2013.
Tissues from these specimens were preserved in 99% ethanol and stored at −10°C until genetic analysis.

| Polymerase chain reaction conditions and mitochondrial DNA sequencing
Total genomic DNA was extracted using a Quick Gene DNA tissue kit S (Fujifilm, Japan) according to the manufacturer's instructions and stored in a refrigerator at 4°C until use.  Table 1). The first two regions and the third region were amplified using the primer sets of Kimura, Yanagimoto, and Munehara (2007) and Ward, Zemlak, Innes, Last, and Hebert (2005), respectively. Polymerase chain reactions (PCRs) were performed in 50 μl volumes containing 25 μl Emerald Amp™ PCR Master Mix (Takara Bio Inc., Japan), 22 μl sterile distilled water, 0.5 μl of each 5 μmol/l primer, and 2 μl of template DNA (50-100 ng). The PCR profiles for the three regions consisted of an initial denaturation step at 94°C for 2 min, followed by 30-40 cycles of denaturation at 94°C for 30 s, annealing at 55°C, and extension at 72°C for 30 s, with a final extension step of 72°C for 7 min. After the final extension step, samples were stored at 4°C. Amplification was performed using a Takara PCR Thermal Cycler Dice (Takara Bio Inc.), and PCR products were purified using a NucleoSpin ® Gel and PCR Clean-up kit (Macherey-Nagel GmbH & Co. KG, Germany). PCR products were sequenced with an autosequencer (3130 Genetic Analyzer, Applied Biosystems, CA) by Macrogen Japan Corporation using the same PCR primers.

| Sequence analysis
Specimen sequences were aligned using the Clustal W computer program (Higgins, Thompson, & Gibson, 1994). Genealogical analysis among haplotypes was performed using MEGA software (version 6.06; Tamura, Stecher, Peterson, Filipski, & Kumar, 2013). The nucleotide substitution model for each gene was selected using Kakusan4 (Tanabe, 2011), and sequence data were also subjected to a maximumlikelihood (ML) analysis. Phylogenetic relationships between each partition were inferred by the ML method using RAxML (version 7.2.8; Stamatakis, 2006). Nucleotide divergences were computed using the Kimura 2-parameter model (Kimura, 1980), and a genealogical tree was constructed using the neighbor-joining method (Saitou & Nei, 1987). The robustness of the topology nodes was assessed using the bootstrap method with 1000 replications (Felsenstein, 1985).

| Allelic analysis using microsatellite DNA
In hybridogenesis, nuclear DNA inherited from the maternal ancestor is maternally inherited in the same way as mtDNA, which means that microsatellite marker is also well suited for genealogical analysis in hemiclone organisms. To examine the sharing of alleles among the natural hybrids (Hoc/Hag and Hoc/Hot) and Hoc, three highly polymorphic microsatellite loci (hexoc 6, hexoc 14, and hexoc 21) were examined (Table 1, Kimura-Kawaguchi et al., 2014). The methods used for the amplification of microsatellite DNA and genotyping of PCR products were the same as those employed in a previous study (Kimura-Kawaguchi et al., 2014).

| Genealogical analysis of Hexagrammos octogrammus and the two natural hybrid strains using mtDNA
Nucleotide sequences were obtained for a total of 2,498 base pairs  Table S1.
The maximum-likelihood analysis showed that the two natural hybrid strains (i.e., Hoc/Hag and Hoc/Hot) formed a cluster within

| Sharing of alleles among the two natural hybrid strains and Hexagrammos octogrammus
The characteristics of the microsatellite loci used for genotyping are shown in Table 3. In the parental species (Hag, Hot, and Hoc), all three loci had sufficiently high heterozygosities and low Hardy-Weinberg equilibrium deviation probabilities, which meant that the microsatellite loci were well suited for genetic analysis and that there were marked differences in the size of alleles among parental species (Fig. 3). Conversely, in the two natural hybrids, the observed heterozygosities approached to 1, except for hexoc 6 in Hoc/Hag, and the probability of deviation from Hardy-Weinberg equilibrium assessed by a chi-squared test was high (Table 3). This finding was illustrated by the natural hybrids that possessed a hemiclonal genome set inherited from the maternal ancestor (Hoc) and a different genome set inherited from the paternal species (Hot or Hag).
Interestingly, 37 of the 38 Hoc/Hot hybrids shared the same alleles at hexoc 6 (116 bp) and hexoc 21 (148 bp), and all 18 of the Hoc/Hag hybrids in Branch I also shared these allele sets (Table 4). In 55 hybrids, hexoc 14 was either 122 bp or 124 bp in length; the one exception in Branch I, ID399, shared alleles at hexoc 14 and hexoc 21, but the alleles at hexoc 6 were unique. Because the 116-bp allele at hexoc 6 in 37 Hoc/Hot hybrids was smaller than the size range observed in H. otakii (Fig. 3, Table 3), the shared allele in Branch I was likely hemiclonally inherited from the maternal ancestor, Hoc.
The 148-bp allele at hexoc 21 in all 56 hybrids in Branch I was larger than the size ranges observed in both Hot and Hag, implying that the common allele was also hemiclonally inherited from Hoc. The 122-bp or 124-bp alleles at hexoc 14 in 18 Hoc/Hag hybrids in Branch I were larger than the size range observed in Hag, implying that these alleles also appeared to be hemiclonally inherited from Hoc. The allele frequencies of the alleles (116 bp at hexoc 6, 122 bp or 124 bp at hexoc 14, and 148 bp at hexoc 21) that were shared among the hybrids were all less than 10% in Hoc (Table 3). In addition, the specific allele observed at hexoc 6 in ID399 probably varied from homologous alleles, as microsatellite DNA occasionally mutates during generation changes (Guichoux et al., 2011). Supposing that the 122-bp and 124-bp alleles at hexoc 14 were homologous and that they slipped during several generation changes, then the common homologous allele set at the three loci would occur at a frequency of less than 0.04%. Such a low rate suggested that all of the individuals in Branch Only 1 bp of the 2,498 bp of mtDNA analyzed was found to differ between hap 22 and hap 23 ( Although ID278 (hap 7) and ID217 (hap 3-2) shared alleles at every locus, the 112-bp allele that was common to hexoc 21 was inherited from Hag, judging from the size range of the alleles in both parental species (Fig. 3, Table 4). Of the remaining 7 Hoc/Hag hybrids, none shared any alleles at the three loci with the other hybrids.

| Hoc/Hot born from host switch
Because both hybridogenetic hybrids (Hoc/Hot and Hoc/Hag) had Hoc mtDNA haplotypes, H. octogrammus (Hoc) is considered to be the maternal ancestor of these hybrids (Crow et al., 2007;Kimura et al., 2007). Although morphological (Shinohara, 1994) and molecular studies (Crow et al., 2004) have demonstrated that H. agrammus (Hag) and H. otakii (Hot) are the closest relatives (sister species), hybrids between these two species have rarely ever been observed in areas where these species are sympatrically distributed (Crow et al., 2007;Kimura & Munehara, 2010;Kimura & Munehara, 2011). Conversely, natural hybrids (Hoc/Hot and Hoc/Hag) between distant species have been shown to propagate by hemiclonal reproduction, with hybridization occurring after secondary contact (Kimura-Kawaguchi et al., 2014). Because hybrids typically have low fitness and survivability, parental species typically avoid hybridization by reinforcing species recognition, as failure to do so would result in these species interbreeding and forming a single species (Coyne & Orr, 2004;Ellstrand et al., 2010). Hemiclonal reproduction is one mechanism that allows hybrids to survive while avoiding genetic recombination (Burt & Trivers, 2006). Because all of the hemiclonal hybrids are fertile females capable of breeding with males of the paternal species, the two natural hybrid populations can be considered to be independent of Hoc, Hot, and Hag (Kimura-Kawaguchi et al., 2014). The hybrids produce haploid eggs containing only the Hoc genome (maternal ancestor), as the paternal genome is discarded and F 1 hybrid-type offspring are generated by fertilization with haploid sperm from either Hag or Hot (paternal ancestor); the entire paternal genome is displaced at every generation change. When a Hoc/Hag hybrid mates with a Hot male, the entire paternal genome of the descendants will change from Hag to Hot. The genome of the descendants will therefore constitute the hemiclonal Hoc genome and a normal Hot genome, to produce the Hoc/Hot hybrids.

| Diversification of Hoc/Hag from Hoc
There is a strong possibility that Hoc/Hag hybrids changed sperm donors to both Hot and Hoc, although the evidence is somewhat inconclusive. We consider that the genome of Hoc/Hag hybrids is constituted by both Hoc and Hag genomes, and this mode of host switching (i.e., Hoc/Hag crossing with Hoc instead of Hag) may be more likely than Hoc/Hag hybrids mating with Hot. When a Hoc/Hag hybrid mates with a male of the maternal species, Hoc, the offspring (backcrossed Hoc; BC-Hoc) become Hoc (Fig. 4). BC-Hoc has the same morphological characteristics as normal Hoc, but the two Hoc genomes differ somewhat with respect to the genetic material they contain. We reported previously that natural Hoc/Hag hybrids produced haploid eggs containing only the maternal genome, whereas artificial F 1 hybrids (i.e., crosses between Hoc and Hag) produced haploid eggs containing a recombinant genome (Kimura-Kawaguchi et al., 2014). The artificial F 1 hybrids had the same genome composition as the natural hybrids, but the reproductive system differed between the two hybrids; that is, the Hoc genome of the natural hybrids carried genetic factors that facilitated hybridogenesis, which were not present in the normal Hoc genome. Although this mechanism has not yet been resolved at a cytological level, we found that BC-Hoc individuals produced recombinant gametes (Kimura-Kawaguchi et al., 2014;in preparation Multiple haplotype revival through host switching from a single mutation in hybrids is another possible hypothesis for the observed mixing of Hoc/Hag haplotypes within the mtDNA genealogical tree.
High levels of mtDNA diversity were also found in P. monachalucida (Quattro, Avise, & Vrijenhoek, 1991. In the Poeciliopsis complex, involvement of host switching through a third species (P. viriosa) appeared to generate new hemiclonal lineages (Mateos & Vrijenhoek, 2002). However, Schultz (1973) demonstrated that it was very difficult to reproduce such a clonal reproductive lineage by artificial hybridization between parental species. The intact genome of the maternal species is transferred into haploid eggs and the genome of the paternal species is eliminated. However, this means that at least two extraordinary steps must occur during oogenesis: elimination of the paternal genome and duplication of the maternal genome (Ogielska, 1994(Ogielska, , 2009Tunner & Heppich-Tunner, 1991;Vinogradov, Borkin, Gunther, & Rosanov, 1990

| Improving longevity through host switching in hemiclones
In organisms that employ unisexual reproduction, individuals can produce offspring without any genetic contribution from males, and no male offspring are produced. As a result, once they arise, unisexual species are considered to be at an advantage when colo- organisms (Avise, 2008). This theory is illustrated by the rapid expansion of Branch I in which hap 1 was dominant (17 of 31 Hoc/Hag and 37 of 38 Hoc/Hot). Within the context of the long-term survival of a population or species, unisexual species must mitigate the risks posed by the accumulation of deleterious mutations (Kondrashov, 1988;Welch & Meselson 2000). In hybridogenesis, the genome derived from the paternal species is renewed every generation and genetic variation is maintained; in this respect, it is different from gynogenesis in which an entire genome set is inherited by offspring.
In addition, gametes are produced through recombination in sexually reproducing organisms, but not in hemiclonal systems when homologous genomes are combined. This is another advantage of hybridogenesis. Deleterious mutations that have accumulated in a hemiclone can be dispersed by recombination in carriers. Such purging of deleterious mutations is possible when hybridogens coexist with maternal species. This episodic host switching ensures that the longevity of the hemiclone lineage is improved by increasing genetic variability, provided that the maternal species continues to inhabit the hybrid zone or occurs in adjacent habitats. When did the genetic factors inducing hybridogenesis come into existence? The paternal species Hot and Hag diverged sympatrically approximately 1.2-2.0 million years ago (Crow, Munehara, & Bernardi, 2010). The mutations producing these genetic factors may possibly have arisen before speciation.

ACKNOWLEDGMENTS
We appreciate the critical advice offered by Dr. KD Crow. We also thank an anonymous consultant at Forte Inc. for proofreading the manuscript, Mr. K. Togashi for DNA sequencing of the outgroup fishes, the Usujiri Fishermen's cooperative for permission of fish collection, and the staff of the Usujiri Fisheries Station at Hokkaido University for their cooperation. This work was supported by Grants-in-Aid (No. 23380107, No. 26292098 and No. 15H02457) for Scientific Research from the Japan Society for the Promotion of Science, Japan.

FUNDING INFORMATION
This work was supported by Grants-in-Aid (Nos. 23380107, 26292098 and 15H02457) for Scientific Research from the Japan Society for the Promotion of Science, Japan.

CONFLICT OF INTEREST
None declared.

DATA ACCESSIBILITY
Accession numbers (DDBJ) of the mtDNA sequences and the genotypes of microsatellite DNA data for specimens used in this study are shown in Table S1 and Table 4, respectively. Morphological data for specimens are provided in the supporting information for Kimura-Kawaguchi et al. (2014). T A B L E 4 (continued) Yellow cells show mtDNA or microsatellite alleles that were inherited from the common maternal ancestor in Branch I. Orange cells show mtDNA or microsatellite DNA alleles that were inherited from the common maternal ancestor with hap 22 and hap 23. Stippled cells show homologous mutations from the 122-bp allele at hexoc 14 alleles. Gray-colored cells for ID278 and ID217 show an allele that was inherited from Hag. Two individuals originated from different hybridogens despite sharing common alleles at three loci.
F I G U R E 4 Results for hybridizations occurring among three Hexagrammos species and two hemiclonal hybrids. Uppercase letters superimposed on fish represent the genomes of each species, with asterisks indicating that the genome possesses the genetic factor responsible for inducing hybridogenesis. (A and B) Represent a normal backcross of hemiclonal hybrids. (C) Represents hybridization between Hoc and Hag. The F 1 offspring (Hoc × Hag) produce gametes that have undergone recombination, but the descendants of the F 1 offspring will disappear because genetic introgression among the parental species via two hybrid populations does not occur (Kimura-Kawaguchi et al., 2014