Sample collection

In the years 1979 to 1991, we collected wild specimens of O. latipes at 188 different sites in Japan (Fig. 1). The collection sites are listed in Table 1. One fish was analyzed at most of the sites; two were examined at the sites with an asterisk (*).

Fig. 1.

Location of collection sites for medaka. The numbers refer to the locations in Table 1.

Table 1

Site numbers, collection site and mtDNA haplotype numbers

Five inbred strains were also examined. The HNI strain of the medaka was established from the wild population of Niigata. The Hd-rR and the HO5 strains were established from orange-red stocks. The HB32C and the HB12C strains were established from wild stocks (Hyodo-Taguchi and Egami, 1985).

DNA extraction

The head, intestines and fins were removed and discarded from adult medaka. The remaining tissue was placed in 500 μl of 100 mM EDTA, 50 mM Tris (pH 8.0), 100 mM NaCl, 1% sodium dodecyl sulfate (SDS), and 100 μg/ml of Proteinase K, then minced and incubated at 55°C overnight. The homogenate was extracted twice with buffer-equilibrated phenol, once with a 1:1 mixture of phenol:chloroform, and once with chloroform. DNA was precipitated with isopropyl alcohol, rinsed with ethanol, and resolved in TE buffer (1 mM EDTA, 10 mM Tris pH 8.0).

Isolation of mtDNA for probe

Approximately 50 grams orange-red medaka (about 200 individuals) were homogenized twice by a whirling blender for 10 sec, then the mtDNAs were prepared by the SDS-phenol method. Mitochondria and crude mtDNAs were prepared as described by Yonekawa et al. (1978). Mitochondrial DNAs were further purified by CsCl-ethidium bromide density-gradient centrifugation at 36,000 rpm for 40 hr. The fractions containing closed circular and open circular mtDNAs were collected separately.

Restriction analysis

The isolated closed circular mtDNA was used as a radioactive probe in Southern blotting to detect mtDNA restriction fragments obtained by restriction endonuclease digestions from total cellular DNA. Six restriction enzymes, Sma I, Bgl II, Xba I, Dra I, Pst I, and Eco RV were used for RFLPs analysis. The restriction fragments were assigned molecular weights in comparison to a size standard of lambda phage DNA digested with Eco T14I. Each distinct restriction fragment pattern produced by any of the six endonucleases was assigned an upper-case letter code in alphabetical order of the detection (e.g. A, B). Thus, each individual was finally assigned a six-letter composite mtDNA haplotype.

Analysis of data

Percentages of nucleotide sequence divergence (p value) between mtDNA haplotypes were estimated from the shared restriction fragment (Nei and Li, 1979). Relationships among mtDNA haplotypes were assessed by unweighted pair group method using arithmetic averages (UPGMA) and neighbor joining (NJ) clustering using the NEIGHBOR program, version 3.5c of PHYLIP (Felsenstein, 1993).

Mitochondrial DNA haplotypes of wild population of medaka

Polymorphisms in the cleavage fragment patterns of medaka mtDNA were revealed by the six restriction enzymes. The fragment patterns revealed by each enzyme are illustrated in Fig. 2. Five different fragment patterns were found for Sma I (A-E), six for Bgl II and Xba I (A-F), eight for Dra I and Pst I (A-H), and 10 for Eco RV (A-J).

Fig. 2.

Diagram of the restriction patterns revealed by each of six restriction enzymes. Restriction fragment length polymorphisms of medaka mtDNAs were digested with six endonucleases. M shows DNA size marker, lambda phage DNA digested with Eco T14I. The numbers indicate fragment length (kb).

Altogether, 63 mtDNA haplotypes were recognized, as presented in Table 2. Each haplotype was designated with a number representing a haplotype. Thirty haplotypes were found in unique sites. Fifteen haplotypes shared two sites. Six haplotypes shared three sites. Three haplotypes shared four sites. Nine haplotypes shared more than six sites (Table 2).

Table 2

Mitochondrial DNA haplotypes, fragment patterns (indicated by six letters), number of sites in which the haplotype was found, and number of individuals detected. Enzyme order: Sma I, Bgl II, Xba I, Dra I, Pst I, Eco RV.

Genetic distance between mtDNA haplotypes and phylogenetic analysis

The estimated pairwise sequence divergence among haplotypes ranged from 0.1 to 12.4% (Appendix 1). A phylogenetic tree based on these sequence divergences was constructed by the UPGMA method (Fig. 3). The tree showed that the 63 mtDNA haplotypes found among the Japanese wild populations were divisible into three clusters: haplotypes #1 to #7 (cluster A), haplotype #46 (cluster C), and the other 55 haplotypes (cluster B). Cluster C was more closely related to cluster B than to cluster A. Cluster A was further separated into two subclusters (A-1 and A-2) and cluster B into 11 subclusters (B-1 to B-11). We accepted a depth of 0.48% as a standard of subclusters, except for subcluster B-11. In subcluster B-11, it was clear that haplotype #22 was closely related to haplotypes #59, #60, and #61.

Fig. 3.

Phylogenetic tree. The UPGMA phenogram of 63 mtDNA haplotypes found among Japanese medaka. The phenogram was derived from a matrix of percentage nucleotide sequence divergence estimates based on the restriction fragment length polymorphisms among the haplotypes (Appendix 1).

The estimated nucleotide sequence divergence among clusters ranged from 5.5% to 11.4%. Between clusters A and B, the range (mean ± standard deviation) was 7.0%–11.4% (8.9 ± 0.87%); between clusters A and C, 8.1%–8.6% (8.4 ± 0.18%); and between clusters B and C, 5.5%–11.2% (7.3 ± 1.14%).

The intra-cluster divergence in cluster B is two times as large as that in cluster A (Fig. 3). Specifically, the estimated pairwise sequence divergence in cluster A and cluster B ranged from 0.3 to 1.5% (mean: 0.8 ± 0.37%) and from 0.1 to 5.0% (mean: 1.5 ± 0.66%), respectively.

We constructed an NJ tree using the same data set. The results also demonstrated that the mtDNA haplotypes were divisible into three clusters and that clusters A and B could be subdivided into two and ten subclusters, respectively. The patterns of clustering revealed by the UPGMA and NJ methods were essentially identical.

Geographic distributions of the mtDNA haplotypes

Figure 4 shows the geographic distribution of the mtDNA haplotypes of each subcluster. The data analysis revealed strong geographical associations for the mtDNAs of clusters and subclusters. The features of the haplotypes of each (sub) cluster are as follows.

Fig. 4.

Geographic distribution of mtDNA haplotypes. Geographical distribution of mtDNA haplotypes of clusters and subclusters inferred from the UPGMA analysis. Numbers indicate the haplotype numbers. One symbol represents one fish, except for the ten sites which are marked with an asterisk (*) in Table 1.

Cluster A was separated into two subclusters (A-1 and A-2). The mtDNA haplotypes of cluster A were distributed in northern Japan along the Sea of Japan. The mtDNAs of subcluster A-1 had a large distribution area. Haplotype #1 in particular had a large distribution area. The mtDNAs of subcluster A-2 were found only in three sites at the western part of the Noto Peninsula.

Cluster B was divided into 11 subclusters (B-1 to B-11). The mtDNA haplotypes of cluster B were found in southern Japan along the Pacific coast southward from Iwate Prefecture and along the Sea of Japan westward from Kyoto Prefecture. The mtDNAs of subcluster B-1 were found only in the San-in district. Most of the mtDNAs of subcluster B-2 were found mainly in the western region of the Seto Inland Sea. The mtDNAs of subcluster B-3 were found at the Pacific coast. The mtDNAs of subcluster B-4 were found in the central part of Honshu island, and also in the Kanto district. The mtDNAs of subcluster B-5 were mainly distributed at the western edge of Honshu. However, five specimens from the Kanto district also had mtDNAs of subcluster B-5. All five of the haplotypes found in the Kanto district were haplotype #29. The mtDNAs of subcluster B-6 were distributed in the southeastern area of Kyushu and at the western edge of Honshu. The mtDNA haplotype of subcluster B-7 was a minor haplotype in this study. Two specimens from two sites (No. 67 and 68) showed haplotype #12. The mtDNAs of subcluster B-8 had a large distribution area and a large number of haplotypes. The haplotypes were found mainly in the eastern region of the Seto Inland Sea district. In addition, the haplotypes were found in the Kanto district (haplotype #55) and at the western edge of Honshu (#48 and #56). The mtDNAs of subcluster B-9 had two distribution areas. The two areas were a range from the Kanto district to the Tohoku district and northern Kyushu. The mtDNAs found in the two areas were different. Haplotypes #15, #16, and #17 were found in the Kanto-Tohoku district. Haplotypes #23 and #24 were found in northern Kyushu. The mtDNA haplotype of subcluster B-10 was a minor haplotype, haplotype #14 from one site, Kainan (No. 109). The mtDNAs of subcluster B-11 were found in the western area of Kyushu. The medaka from the Ryukyu Islands also had the mtDNA of the subcluster B-11. We found only haplotype #60 in the southern region of the distribution range.

Cluster C had only one haplotype, #46. The mtDNA of two specimens found at two sites in the Kanto district showed haplotype #46.

Haplotypes of the inbred strains of medaka

In this study, five inbred strains were analyzed. The results showed that all inbred strains except for the HNI strain shared the same haplotype, #32. The HNI strain was haplotype #1. This result coincides with the mtDNA haplotype of the wild fish captured at Niigata (Site No. 16).

Marked intraspecific diversity among mtDNAs

Intraspecific mtDNA divergences of 7.4% in anchovy (Magoulas et al., 1996) and 8.5% in sunfish (Avise et al., 1984) have been reported as the highest degrees of intraspecific divergences. In the medaka, the pairwise comparisons between haplotypes of clusters A, B and C demonstrated large divergence; 7.3%–12.4%. In particular, the mean sequence divergence between clusters A and B is 9.0%. The large divergence among these three clusters suggests that the events separating these clusters are very old.

The rate of base substitution of mammalian mtDNA has been estimated to be 2.0% per million years (Brown et al., 1979). At this rate, cluster A and cluster B would have shared a common ancestor 4–5 million years ago (sequence divergence estimated approximately 8–10%).

Concordance between mtDNA haplotypes and allozyme genotypes in the medaka

Allozymic variations have been studied at 21 loci in Japanese wild populations of the medaka collected at 53 sites, and the Japanese wild population of medaka was divided into two genetically distinct groups, the Northern Population and the Southern Population. Allozymic analyses demonstrated that the Northern Population is less variable than the Southern Population (Sakaizumi et al., 1983).

The rigid isolation between the mtDNAs of cluster A and those of cluster B coincides perfectly with the previously defined two population ranges— A is associated with the Northern Population and B with the Southern Population. The characteristics of cluster A's mtDNAs also show lower intra-cluster polymorphism than do cluster B's mtDNAs (Fig. 3). Such concordance between mtDNA haplotypes and allozyme genotypes would be expected if the Northern Population and the Southern Population had evolved in complete isolation in the past.

Mitochondrial DNA haplotype of cluster C

Haplotype #46 in cluster C is a rare variant in the Japanese wild population of medaka. This cluster is more closely related to cluster B than to cluster A (mean estimated divergence 7.7 versus 8.9), but divergent from cluster B (pairwise divergence ranged from 5.8% to 12.1%). The fish with this mtDNA haplotype have an allozymic genotype similar to that of the fish with haplotypes of cluster B (the Southern Population type; data not shown), despite the such large sequence divergence in mtDNA estimated. Thus, we suspect that the “old” mtDNA haplotypes which diverged in the ancestor of the medaka may have persisted in a limited area (the Kanto district). Examinations of the frequency and distribution, and detailed molecular analysis are in progress to elucidate the origin and dispersal of this haplotype.

Migration of the Japanese wild population of medaka

In the present study, we divided the mtDNA haplotypes in Japanese medaka into three clusters, and clusters A and B were subdivided into two and 11 subclusters, respectively. The distribution pattern of haplotypes of each subcluster showed a strong geographical association and unique distribution pattern (Fig. 4). The distribution pattern presumably reflects the migration history of Japanese wild medaka. We are especially mindful of the distributions of single haplotypes and those of haplotypes in a subcluster. We can thus propose three migration scenarios from the distribution patterns of mtDNA haplotypes. These scenarios are “to the Tohoku district”, and “to the Ryukyu Islands”, and “to the Kanto district”. These scenarios are based on a hypothesis that migration occurred from a region where we found high variation in mtDNAs to a region with low variation in mtDNAs.

All specimens found in the Tohoku district at the sea of Japan coast had mtDNA of subcluster A-1, those at the Pacific coast had mtDNA of subcluster B-9. In the southern regions of the Tohoku district, we found mtDNAs of more than one subcluster. This observation leads to a scenario in which the medaka in the southern region of the Tohoku district have recently expanded their range into the northern region.

We also propose a scenario from the distribution pattern of mtDNAs of the subcluster B-11 in western Kyushu and the Ryukyu Islands. Subcluster B-11 consists of four haplotypes (Fig. 3). Haplotype #60, of these four haplotypes, is found in the southern part of western Kyushu and the Ryukyu Islands. The other three haplotypes were found in the northern part of western Kyushu. This observation leads to a scenario in which the medaka in the southern part of the region have recently expanded their range into the Ryukyu Islands.

In the Kanto district, two features were notable; the high frequency of the subcluster B-9 mtDNA haplotypes, and a small number of haplotypes from each other subcluster of cluster B. Regarding the first feature, we found that mtDNA haplotypes of subcluster B-9 were common in the Kanto district, and that the mtDNA haplotypes had split distribution ranges, i.e., the Kanto to the Tohoku district and northern Kyushu (Site No. 151, 152, and 161). In addition, there were no haplotypes common to both ranges; haplotypes #15, #16, and #17 in the Kanto and Tohoku district, and haplotypes #23 and #24 in northern Kyushu. This distribution pattern of the subcluster B-9 mtDNAs suggest that it was long ago when the medaka with the haplotype #15, #16, or #17 colonized the Kanto district. Regarding the second feature, we found five other mtDNA haplotypes in the Kanto district (haplotypes #20, #21, #29, #46, and #55). These five mtDNAs are classifiable into four (sub) clusters (subclusters B-4, B-5, and B-8, and cluster C). The four haplotypes of the three subclusters were also found in western Japan. Haplotypes #20 and #21 were found in the Tokai district, haplotype #29 at the western edge of Honshu, and haplotype #55 in the Kinki district. The sole haplotype of the cluster C was found only in the Kanto district. Thus, haplotypes other than those of subcluster B-9 and cluster C were also found west of the Kanto district. Therefore, we suspect that the low variation of haplotypes in the four subclusters is the result of a recent expansion eastward (to the Kanto district). These two features lead to a scenario in which at least two migration events occurred in the Kanto district; first, medaka with mtDNAs of subcluster B-9 expanded their range into the Kanto district. Then, other haplotypes diverged in western Japan, and introgressed to the Kanto district recently. Consequently, the medaka found in the Kanto district are a “mixed population”.

In this study, mtDNA RFLPs analysis suggests three clusters and two and 11 subclusters for wild populations of medaka, which have strong geographical associations. The geographical distributions of the mtDNA haplotypes suggest three migration events. However, this analysis also suggests the necessity for more research concerning the Southern Population. The phylogenetic relationships among the subclusters of cluster B, the separate distribution origin of haplotypes in subcluster B-9 and the origin of cluster C, and the migration histories in the San-in district and the Seto-Inland Sea area are not yet clear. Phylogenetic analysis using nucleotide sequence information is necessary to elucidate these issues. We are currently identifying the nucleotide sequences of mitochondrial cytochrome b gene. This project may clarify the relationships among the clusters and subclusters.