The Japanese rose bitterling, Rhodeus ocellatus kurumeus, is an endangered cyprinid species. The main reason of drastic diminution of R. o. kurumeus has been suggested to be due to hybridization between this subspecies and R. o. ocellatus introduced from China. Both RFLP analysis of mitochondrial DNA and RAPD-PCR analysis of genomic DNA were performed to elucidate the genetic introgression of R. o. ocellatus into R. o. kurumeus. The two subspecies were distinguished in the D-loop region of mtDNA by six restriction endonucleases. Rhodeus o. kurumeus had eleven subspecies-specific RAPD markers and R. o. ocellatus had two. Except some R. o. kurumeus populations, most of the populations of R. ocellatus in Japan were hybrids, equipped with subspecies-specific RAPD markers for the two subspecies, respectively. The genetic constitution of these markers in hybrids, however, greatly differed among populations. The R. o. ocellatus mtDNA was predominantly observed in hybrid populations, except two populations with mtDNAs of the two subspecies. Judging from the genetic dominance of morphological and physiological characters of R. o. ocellatus against R. o. kurumeus, hybrids probably have the same ecological dominance as R. o. ocellatus against R. o. kurumeus. Therefore, it is considered that R. o. kurumeus not only has its genetic property spoiled by hybridization with R. o. ocellatus, but also is expelled by R. o. ocellatus and hybrids. The replacement of mtDNA and genomes of R. o. kurumeus with those of R. o. ocellatus in hybridization might be accelerated by the backcross between hybrids and R. o. ocellatus.
INTRODUCTION
Genetic introgression has been known in many fish groups (Hubbs, 1955). It occurs when two species, in which reproductive isolation is not complete, meet and hybridize and their hybrids keep fertility, which is not inferior to those of parental species. Incomplete reproductive isolation is often observed especially in incipient species (Vrijenhoek et al., 1989). In the wild, these species are usually isolated by geographical barriers. They, however, seem to easily hybridize with each other in sympatric situation (Dawley, 1987; Echelle and Connor, 1989). Hybridization between an endemic species and its exotic related species spoils the genetic identity of the endemic species. Through the hybridization, inherent genes of an endemic species are likely to be replaced by those of an exotic species, which eventually leads to the extinction of the endemic species (Allendorf and Leary, 1998). Recently, decrease of the populations of endemic fish species, caused by artificial transplantation of exotic species, has been reported in many groups such as Salvelinus (Allendorf and Leary, 1998), Cyprinodon (Echelle and Connor, 1989) and Micropterus (Whitmore, 1983).
The Japanese rose bitterling, Rhodeus ocellatus kurumeus Jordan and Thompson, is an endemic cyprinid species in Japan. Before World War II, it was widely distributed in small ponds and creeks in the western part of Japan (Nakamura, 1969; Miyadi et al., 1976; Hosoya, 1982). The rose bitterling, Rhodeus o. ocellatus (Kner), is considered to have been introduced accidentally from mainland China into Japan, contaminated in the seedlings of grass carp Ctenopharyngodon idellus (Cuvier and Valenciennes) and silver carp Hypophthalmichthys molitrix (Cuvier and Valenciennes), during the 1940s (Nakamura, 1955). During the last five decades after introduction, R. o. ocellatus has drastically expanded its distribution all over Japan, which has given rise to hybrids between the two subspecies in the original distribution of R. o. kurumeus. Owing to hybridization with R. o. ocellatus, R. o. kurumeus has been suggested to have disappeared on a large scale (Nagata, 1980, 1997). Rhodeus o. kurumeus is now considered to be on the verge of extinction and listed as a critically endangered species in the Red List of brackish and freshwater fishes of the Environmental Agency of Japan (Hosoya, 2000).
Rhodeus o. kurumeus and R. o. ocellatus are very similar in external morphology one another (Nakamura, 1955, 1969; Kimura and Nagata, 1992). Nakamura (1969) found that R. o. ocellatus is primarily distinguished from R. o. kurumeus by the presence of white coloration (guanine layer) along the anterior margin of pelvic fins. Nagata (1980) experimentally demonstrated that the white coloration in pelvic fins was dominantly inherited in hybrids between the two subspecies and reported that many populations within the former distribution of R. o. kurumeus were already hybrids, more or less equipped with this character. Ueno (1987) found that R. o. kurumeus and R. o. ocellatus can be distinguished by two enzymes (LDH and PGDH) in the allozyme analysis. With two loci (Ldh-2 and Pgdh) in the two enzymes, Nagata et al. (1996) reconfirmed that some populations of the rose bitterling in Japan were genetically hybrids of the two subspecies. Only two loci, however, are not sufficient for a detailed analysis of the genetic situation of hybridization. In this study, using RFLP analysis of mtDNA and RAPD-PCR analysis of genomic DNA, the authors first tried the genetic distinction between R. o. kurumeus and R. o. ocellatus and then made a search for multiple RAPD markers specific to each subspecies. With haplotypes of mtDNA and subspecies-specific RAPD markers, the present situation of genetic introgression by R. o. ocellatus into R. o. kurumeus in Japan will be discussed.
MATERIALS AND METHODS
Specimens examined
The total 172 fish samples of R. ocellatus used in this study were collected from seven localities in China (n=25), two in Korea (n=8) and 13 in Japan (n=139) from 1998 to 1999 (Fig. 1). All the fishes collected were fixed by 70% ethanol and identified, based on their external morphology and osteological characters by soft X-ray photographs (Uchida, 1939; Woo, 1964; Nakamura, 1969; Choi et al., 1990; Kimura and Nagata, 1992), after removing some muscle tissue for DNA extraction. Samples of R. o. kurumeus were collected from five populations (Ushizu, Yanagawa, Okawa, Sanda and Yao [A]), whose genetic identity had been confirmed by allozyme analysis in the preceding studies (Nagata et al., 1996; Tabe and Fukuhara, 1997). In all specimens, white coloration in pelvic fins was checked, which is a diagnostic morphological character of R. o. ocellatus (Nakamura, 1969).
Fig. 1
Map showing sample localities of the rose bitterling used in this study. (1) Xianyou, Fujian (Mulan R. system); (2) Putian, Fujian (small river); (3) Fuzhou, Fujian (Min R. system); (4) Yujiang, Jiangxi (Changjiang R. system, Xin R.); (5) Xiuning, Anhui (Qiantang R. system, Xinan R.); (6) Qingpu, Shanghai (Changjiang R. system); (7) Haerbin, Heilongjiang (Amur R. system, Sungari R.); (8) Chuksan-ri, Kyonggi-do (Namhan R. system, Tuwol R.); (9) Kwangchon, Chungchongnam-do (Kwangchon R.); (10) Ushizu, Saga (Ushizu R. system); (11) Yanagawa, Fukuoka (Okinohata R.); (12) Souja, Okayama (Takahashi R. system); (13) Okayama, Okayama (Asahi R. system); (14) Okawa, Kagawa (Tsuda R. system, pond); (15) Sanda, Hyogo (Muko R. system, pond); (16) Yawata, Kyoto (Yodo R.); (17) Yao (A), Osaka (Yamato R. system, 9 ponds); (18) Yao (B), Osaka (Yamato R. system, 2 ponds); (19) Ikoma, Nara (Yodo R. system, 1 pond); (20) Ise, Mie (Kushida R. system, Ohori R.); (21) Ogaki, Gifu (Ibi R. system, 1 pond); (22) Sano, Tochigi (Tone R. system, Sai R.). Broken lines indicate the original distribution of Rhodeus ocellatus kurumeus (Nakamura, 1969; Miyadi et al., 1980; Hosoya, 1982; Nagata, 1989). Localities of R. o. kurumeus and hybrids between R. o. kurumeus and R. o. ocellatus, identified from allozyme analysis in the preceding studies, are shown with ★ and ☆, respectively (Nagata et al., 1996; Tabe et al., 1997).
Mitochondrial DNA analysis
Total genomic DNA was extracted from skeletal muscle, by using phenol-chloroform extraction (Lansman et al., 1981). About 2.05-kb fragments in length from the cytochrome b to the 12S rRNA gene, including the entire noncoding region (D-loop), were amplified via PCR (Saiki et al., 1988) from the whole genomic DNA, using a pair of the following primers: CB3R-L (5′-CAYATYMARCCMGAATGRTATTT-3′) and 12SAR-H (5′-ATARTRGGGTATCTAATCCYAGTT-3′) (Palumbi et al., 1991). The conditions of amplification were 2 min at 94°C for preamplification denaturation; 30 cycles of 94°C for 30 sec, 55°C for 30 sec and 72°C for 2 min; and then 72°C for 5 min for final elongation in a DNA thermal cycler (Takara, Kyoto). The length of all PCR products was confirmed as equal length by electrophoresis, prior to digestion with restriction endonucleases.
A total of 15 restriction endonucleases were used to digest the amplified products of PCR: Acc II, Afa I, Alu I, Cfr13 I, Dde I, EcoR I, Hae III, Hha I, Hinf I, Hsp92 II, Mbo I, Mse I, Msp I, ScrF I, Taq I (Table 1; from Life Technologies, Rockville, MD; Promega, Madison, WI; Takara Shuzo, Kyoto; Toyobo; Tokyo). For each sample, 1–4 μ l of the PCR product containing the amplified DNA was digested following the recommendations of the enzyme manufacturers. The digested fragments of the D-loop segment were electrophoretically separated on horizontal 3% agarose gels. The bands of digested fragments were photographed with UV illumination, after staining with ethidium bromide.
Table 1
Approximate size in base pairs (bp) of all fragment patterns observed on the D-loop region in the rosy bitterling
Nucleotide sequence divergence in mtDNA was calculated by the length-difference method (Nei and Li, 1979), after comparing the electrophoretic patterns in each of the two haplotypes of mtDNA, digested by a restriction endonuclease recognizing four nucleotides except EcoR I, respectively. A dendrogram was generated by neighbor-joining method (NJ; Saitoh and Nei, 1987) and the reliability of each clade was checked by 1,000 bootstrap resampling of fragments with PHYLIP version 3.572 (Felsenstein, 1996).
RAPD-PCR analysis
Each RAPD-PCR reaction mixture consisted of 2.5 μ l of 10×PCR buffer, 2 μ l of each dNTP (2.5 mM), 1 μ l of oligonucleotide primer (25 μ M), 0.5 μ g of genomic DNA, 0.1 U of Taq polymerase (Takara, Kyoto) in 25 μ l final volume. Two kits (designated A and F), each containing 20 decamer RAPD primers (Operon Technologies, Alameda, CA) were used in this study. Amplifications were carried out with the following cycle program: 2 min at 94°C for preamplification denaturation; 35 cycles of 94°C for 1 min, 36°C for 1 min and 72°C for 2 min; and then 72°C for 5 min for final elongation. After amplification, all PCR products were subjected to electrophoresis in a 1.2% agar-ose gel. Band patterns were analyzed visually from photographs taken of the ethidium bromide stained gels.
RAPD markers, specific to R. o. kurumeus or R. o. ocellatus, were examined using the specimens of R. o. kurumeus from Yanagawa, Okawa, Sanda and Yao (A), and those of R. o. ocellatus from Qingpu (St. 6) and Sano (St. 21). The haplotype of mtDNA of all the specimens from Qingpu closely accorded with the predominant haplotype of R. o. ocellatus in Japan. The inheritance of RAPD markers in F1 was checked with artificially produced hybrids between R. o. kurumeus from Yanagawa and Yao (A) and R. o. ocellatus from Sano. The situation of hybridization of the two subspecies in the wild populations of the rose bitterling was examined from the expression of subspecies-specific RAPD markers for R. o. kurumeus and R. o. ocellatus. In addition, the extent of genetic introgression of R. o. ocellatus into R. o. kurumeus in each population was also assessed by calculating the averaged number of these subspecies-specific RAPD markers per specimen.
RESULTS
Mitochondrial DNA analysis
Genetic polymorphism in D-loop of mtDNA was observed in all 15 restriction endonucleases used in this study. Fragment patterns digested by restriction endonucleases were arranged into 20 haplotypes (Tables 1, 2). Twelve haplotypes (mt1-6 and mt15–20) were found in Japan (Table 3). Rhodeus o. kurumeus (St. 10–11, 14–15 and 17) had six haplotypes (mt15-20). Haplotype groups, mt1-6 and mt15–20, were clearly distinguished by the following six restriction endonucleases: Alu I, Cfr13 I, EcoR I, Hsp92 II, Mbo I and Mse I (Table 2, Fig. 2). Nucleotide sequence divergence among mt1-6 or mt15–20 was very low, ranging from 0.06 to 0.64% (mean±SD= 0.34±0.16%, n=15) in mt1–6, and 0.11 to 0.85% (mean±SD= 0.40±0.21%, n=15) in mt15–20. However, it was high between mt1-6 and mt15–20, ranging from 1.40 to 2.47% (mean±SD=1.97±0.26%, n=36).
Table 2
Composite haplotypes for restriction enzyme polymorphisms in the rose bitterling
Fig. 2
Electrophoretic patterns of mtDNA digested by EcoR I in some local populations of the rose bitterling in Japan. Lanes 1–2, Souja; lanes 3–5, Okayama; lanes 6–8, Yawata; lanes 9–10, Ikoma; lanes 11–12, Yao (B); lanes 13–15, Yao (A); lanes 16–17, Sanda; lanes 18–19, Yanagawa; lanes 20–21, Ushizu; lanes 22–23, Okawa. Black and white arrows indicate specific bands to Rhodeus ocellatus kurumeus and R. o. ocellatus, respectively. M is a DNA size marker (φ X174 Hinc II digest).
Each R. o. kurumeus haplotype was characteristically distributed in Kyushu, Shikoku and Honshu, respectively (mt15 and 16 in Kyushu, mt20 in Shikoku and mt17–19 in Honshu; Table 3). They were also found from populations in Souja and Okayama. The haplotype mt17 in Souja and Okayama accorded with that in Sanda and Yao (A). In Souja and Okayama, however, haplotypes other than mt15–20 were also found in the same populations (mt1–3 in Souja and mt1 in Okayama). Although R. o. kurumeus haplotypes (mt15–20) had a characteristic geographical distribution pattern, haplotypes of mt1-6 in Japan did not have such a pattern but a random distribution. Among the haplotypes of mt1–6, mt1 was predominantly observed in Honshu and also discovered in Fuzhou, Qingpu and Haerbin in China. The other haplotypes of mt2–6, observed in Japan, were found in neither China nor Korea in this study.
Table 3
Distribution of mtDMA haplotypes in local populations of the rose bitterling
In NJ tree (Fig. 3), 20 haplotypes of R. ocellatus were clearly arranged into six clades. Rhodeus o. kurumeus haplotypes (mt15-20) constituted one major clade, while the other haplotypes (mt1–14) were grouped into five major clades: mt1–6, mt7–10, mt11, mt12 and mt13–14. Haplotypes of mt1, mt7–12 and mt13–14 were found only in R. o. ocellatus in China and Korea. Haplotypes mt2–6, found in Japan alone, constituted one clade of R. o. ocellatus with mt1. Namely, five clades of mt1–6, mt7–10, mt11, mt12 and mt13–14 are R. o. ocellatus clades. Among the five R. o. ocellatus clades, the Korean clade (mt13–14) was nearest to the R. o. kurumeus clade (range =1.81 to 2.57%, mean±SD = 2.22±0.23% in nucleotide sequence divergence, n=12) and mt1–6 clade was the second. The nucleotide sequence divergence among the five clades of R. o. ocellatus was extremely high, ranging from 1.42 to 5.17% (mean±SD =3.63±1.09%, n=69). Bootstrap values strongly supported the R. o. kurumeus clade at 94.0%.
Fig. 3
Unrooted tree of composite haplotypes in the rose bitterling by the neighbor-joining method, based upon nucleotide sequence divergence (Nei and Li, 1979). Numbers at the forks show the bootstrap values (1,000 replicates). Symbols at the end of tree forks indicate the haplotypes, defined in Table 2. Broken lines mean a haplotype group of R. o. kurumeus.
RAPD analysis
Fourteen RAPD primers out of the 40 primers tested in this study showed RAPD bands specific to local populations or subspecies of R. ocellatus (Table 4). In R. o. kurumeus, population-specific RAPD markers were observed in eight primers (OPA-02, 07 and 12; OPF-02, 09, 15, 18 and 20). Yao (A), Sanda and Okawa populations of R. o. kurumeus shared many RAPD bands (OPA-07 and 12; OPF-02, 09, 15 and 20), but population-specific RAPD bands for these sampling sites were hardly found, except one band of OPA-02 for Yao (A). Yanagawa population was clearly distinguished from Yao, Sanda and Okawa populations by five primers (OPA-07, OPF-09, 15, 18 and 20). RAPD bands common to all the populations of R. o. kurumeus were observed in 11 primers (OPA-02, 05, 09 and 16; OPF-02, 07, 08, 09, 15, 17 and 18; Fig. 4A, 4B). Sano and Qingpu populations of R. o. ocellatus did not share any RAPD bands with R. o. kurumeus. These two populations showed a great variation in RAPD band patterns among individuals and only two RAPD bands (OPF-08 and 15), common to the populations of R. o. ocellatus, were found (Fig. 4A, 4B). In the distinction between R. o. kurumeus and R. o. ocellatus, eleven primers (OPA-02, 05, 09 and 16; OPF-02, 07, 08, 09, 15, 17 and 18) were found to be useful.
Table 4
Selected RAPD primers that identify local population- or subspecies-specific bands in the rose bitterling
Fig. 4
Examples of RAPD patterns in the rosy bitterling, A, OPF-08 and B, 15, specific to Rhodeus ocellatus kurumeus or R. o. ocellatus, respectively. R. o. kurumeus (lanes 1–18) and R. o. ocellatus (lanes 19–22) were characterized by subspecies-specific RAPD markers, respectively. Lanes 1–4, Yanagawa; lanes 5–11, Yao (A); lanes 12–16, Okawa; lanes 17–18, Sanda; lanes 19–20, Sano; lanes 21–22, Qingpu. C, inheritance of RAPD markers (OPF-08) in F1 hybrids. Lanes 1–2 and 22–23, R. o. kurumeus in Yanagawa (control); lanes 3–7, hybrids between R. o. kurumeus in Yanagawa and that in Yao; lanes 8–9, R. o. kurumeus in Yao (A) (control); lanes 10–14, hybrids between R. o. kurumeus in Yao and R. o. ocellatus in Qingpu; lanes 15–16, R. o. ocellatus in Qingpu (control); lanes 17–21, hybrids between R. o. ocellatus in Qingpu and R. o. kurumeus in Yanagawa. D, identification of hybrids of the rose bitterling in wild populations with RAPD markers (OPF-08). Lanes 1–2, Yanagawa (R. o. kurumeus as control); lanes 3–5, Souja; lanes 6-9, Okayama; lanes 10–14, Yawata; lanes 15–16, Yao (A) (R. o. kurumeus as control); lanes 17-19, Ikoma; lanes 20–21, Ogaki; lanes 22–23, Qingpu (R. o. ocellatus as control). Black and white arrows indicate markers specific to R. o. kurumeus and R. o. ocellatus, respectively. M is a DNA size marker (λ Hind III digest - φ X174 Hinc II digest).
In F1 of the two subspecies, expression of RAPD bands specific to R. o. kurumeus or R. o. ocellatus was observed to follow the mode of Mendelian inheritance, irrespective of sex (Fig. 4C). Such RAPD markers, which were inherited maternally like mtDNA, were not recognized in this study.
Morphological and genetic characters of local populations
A total of 13 populations of R. ocellatus in Japan were surveyed concerning the morphological and genetic character distribution of R. o. kurumeus and R. o. ocellatus (Table 5). White coloration in pelvic fins (WCP) was present in all specimens from the populations, except those of St. 10, 11, 13-15 and 17. WCP was not recognized in specimens from populations of Ushizu, Yanagawa, Okawa, Sanda and Yao (A) at all, while it was slightly observed in three specimens of ten from Okayama population. Concerning mtDNA, Souja and Okayama populations had mtDNA of R. o. kurumeus and R. o. ocellatus, while all the other populations had only R. o. kurumeus or R. o. ocellatus mtDNA (Tables 3, 5).
Table 5
Distribution of morphological and genetic characters specific to Rhodeus ocellatus kurumeus or R. o. ocellatus in local populations of the rose bitterling in Japan
All the eleven RAPD markers specific to R. o. kurumeus were also observed in Ushizu population (Table 5). Seven populations in St. 12–13, 16 and 18–21 owned both R. o. kurumeus-specific RAPD markers and R. o. ocellatus-specific ones (Fig. 4D). In these populations, however, the number of confirmed subspecies-specific RAPD markers per specimen greatly differed among populations. The averaged number of R. o. kurumeus-specific RAPD markers ranged from 1.0 to 6.5 and that of R. o. ocellatus-specific RAPD markers did from 0.3 to 1.8. Yao (B) population, adjacent to Yao (A) population of R. o. kurumeus, conspicuously had a large number of R. o. kurumeus-specific RAPD markers (mean±SD= 6.5±1.9, n=4), but the other six populations (St. 12–13, 16 and 19–21) had less than half of the eleven R. o. kurumeus-specific RAPD markers (mean±SD=1.0±1.0 to 4.6±1.5, n=10).
Ise and Ogaki, which are located out of the original distribution of R. o. kurumeus especially (Fig. 1), had a small number of R. o. kurumeus-specific RAPD markers (mean±SD= 1.8±0.4 in Ise and 1.0±1.0 in Ogaki, n=10), contrary to the prevalence of R. o. ocellatus-specific RAPD markers (mean-±SD = 1.2±0.4 in Ise and 1.8±0.4 in Ogaki, n=10).
DISCUSSION
Genetic diversity of the rose bitterling
Rhodeus o. kurumeus and R. o. ocellatus in Japan differed markedly in the sequence of mtDNA (Fig. 3; Table 2). They were clearly distinguished by six restriction endonucleases (Alu I, Cfr13 I, EcoR I, Hsp92 II, Mbo I and Mse I) (Fig. 2). Nucleotide sequence divergence in D-loop between R. o. kurumeus (mt15–20) and R. o. ocellatus (mt1–6) was high (range =1.40–2.47%, mean±SD=1.97±0.26%, n=36) as the subspecies level, compared with that in the bagrid species, which ranges from 0.18 to 6.61% (Okazaki et al., 1999). The great divergence in nucleotide sequence between the two subspecies suggests that R. o. kurumeus and R. o. ocellatus in Japan are genetically distinct enough, although they are morphologically very similar (Nakamura, 1969; Nishiyama and Nagata, 1978). In Asia, R. ocellatus was composed of six clades in the NJ tree, including R. o. kurumeus clade (Fig. 3), where nucleotide sequence divergence among these clades was extremely high, ranging from 1.42 to 5.17%. Judging from the value of nucleotide sequence divergence between each clade in Fig. 3, R. ocellatus is considered to be genetically very heterogeneous. Accordingly, the taxonomy of all the populations of R. ocellatus should be reevaluated in the future.
The nucleotide sequence divergence in the six haplotypes (mt1–6) of R. o. ocellatus in Japan was comparatively low, ranging from 0.06 to 0.64% (mean±SD=0.34±0.16%, n=15), and constituted a single clade (Fig. 3) Therefore, although haplotypes mt2-6 were not discovered in China in this study, it is likely that they are distributed somewhere in China with mt1. Only the mt1 haplotype was found in Fuzhou, Qingpu and Haerbin in China, as well as Japan. The fact that the same haplotype of mt1 was found in both Qingpu, belonging to the Changjiang River system, and Japan agrees with the report of Nakamura (1955), in which he considered the original range of Japanese R. o. ocellatus as the Changjiang River system.
Genetic introgression
Two enzymes, LDH and PGDH, were found to be useful in the distinction between R. o. kurumeus and R. o. ocellatus (Ueno, 1987). As for later generations, however, distinction of hybrids by the two enzymes is difficult, due to random segregation and recombination of chromosomes. RAPD markers are also known to follow the mode of Mendelian inheritance (Foo et al., 1995; Liu et al., 1998). In this study, RAPD markers by the eleven RAPD primers were found to be specific to R. o. kurumeus or R. o. ocellatus (Table 4), and could show hybridization between the two subspecies at the genomic level (Table 5). Seven populations (St. 12, 13, 16 and 18–21) were hybrids, but the genetic status of hybrids differed among populations, judging from the appearance of RAPD markers specific to R. o. kurumeus or R. o. ocellatus, respectively. The averaged number of confirmed R. o. kurumeus-specific RAPD markers was relatively high in Okayama and Yao (B), compared with the remaining five hybrid populations (Table 5). Okayama population especially still kept other characteristics of R. o. kurumeus in mtDNA and the coloration of pelvic fins, in addition to the RAPD markers (Tables 3, 5). It inhabits an upper branch isolated by a dam from the main stream of Asahi River, and Yao (B) population did a small pond located in the foothills of Mt. Ikoma. In common, the two populations are completely isolated from other drainage systems in mountain areas. Probably, because of the small number of introduced R. o. ocellatus or the short elapsed time after introduction, the genetic introgression of R. o. ocellatus into R. o. kurumeus in Okayama and Yao (B) was not so high, compared with other populations. Contrary to Okayama and Yao (B) populations, Souja, Ise and Ogaki populations had a very small number of R. o. kurumeus-specific RAPD markers among the hybrid populations (Table 5). Rhodeus o. kurumeus was formerly distributed in the western part of Japan, that is, the drainage systems located westward of Lake Biwa (Nakamura, 1969; Fig. 1), not inhabiting Ise and Ogaki. Hybrids in Ise and Ogaki might not have originated there, but have been transplanted from other localities within the original distribution of R. o. kurumeus.
Direction in hybridization
As a general feature, most of the hybrid populations were equipped with more characteristics of R. o. ocellatus than those of R. o. kurumeus in morph, mtDNA and RAPD markers (Table 5). In particular, the R. o. ocellatus mtDNA was predominantly observed in the hybrid populations, except Souja and Okayama. Judging from the morphological and genetic status in hybrids (Table 5), the genomic DNA and mtDNA of R. o. kurumeus seemingly tend to be replaced by those of R. o. ocellatus in hybridization. Namely, this means a genetic direction from R. o. kurumeus to R. o. ocellatus in hybridization.
The replacement of genetically pure R. o. kurumeus with R. o. ocellatus is considered to be caused by the following three factors: (1) difference in body size, clutch size and the duration of spawning period between the two subspecies, (2) ecological competition, especially in mating behavior between the two subspecies and (3) genetic dominance of morphological and physiological characters of R. o. ocellatus against R. o. kurumeus.
Nagata (1980) commented that R. o. ocellatus is superior to R. o. kurumeus in body size, clutch size and the duration of spawning period. Rhodeus o. ocellatus becomes larger than R. o. kurumeus in both males and females (Nakamura, 1969). The clutch size of R. o. ocellatus (mean±SD=14.0±3.5, n=69) is approximately twice that of R. o. kurumeus (mean±SD =8.8±3.7, n=16), although both subspecies mature within one year (K. Kawamura, unpublished data). The duration of spawning period of R. o. ocellatus (from April to September; Solomon et al., 1984) is also about two times longer that of R. o. kurumeus (from April to June; J. Kitamura, Kyoto University, personal communication). In mating behavior, there is hardly any subspecies-specific difference between the two subspecies, because they spawned together, mixed in an aquarium (Nagata and Nishiyama, 1976). Kanoh (1996) reported that body size of males is significantly important in reproductive success of R. ocellatus. Therefore, large R. o. ocellatus males are expected to mate with females more successfully than small R. o. kurumeus males (Nagata, 1980). Taking these information into consideration, R. o. ocellatus, regardless of sex, seems to be completely superior to R. o. kurumeus in growth and reproduction.
The growth rate and clutch size of the experimentally produced F1 hybrids are similar to those of R. o. ocellatus (K. Kawamura, unpublished data). F1 hybrids become almost as large as R. o. ocellatus (more than 60 mm TL in both males and females), irrespective of the sex of parental subspecies. Actually, hybrids with a TL of more than 60 mm are commonly observed in the wild. The clutch size of F1 hybrids (mean±SD =17.4±3.3, n=30) was not inferior to that of R. o. ocellatus (mean±SD=14.0±3.5, n=69). In F2 hybrids and later generations, hybrid breakdown was not observed in growth and reproduction at all. According to Nagata (1985), the duration of spawning period of hybrids is as long as that of R. o. ocellatus. Judging from these facts, not only the white coloration in pelvic fins but also other morphological and physiological characters of R. o. ocellatus are considered to be inherited dominantly in hybrids. Genetic dominance of these characters of R. o. ocellatus in hybrids may suggest ecological dominance of hybrids, as well as R. o. ocellatus, against R. o. kurumeus.
In conclusion, the expulsion of R. o. kurumeus by R. o. ocellatus may occur as follows. Once R. o. ocellatus is introduced into a R. o. kurumeus population, regardless of the number of introduced fish, the genetic identity of the R. o. kurumeus population is spoiled by R. o. ocellatus through hybridization. In addition, not only R. o. ocellatus but also hybrids successfully drive away R. o. kurumeus, because of ecological dominance against R. o. kurumeus. Increased hybrids mate with increased R. o. ocellatus, which still more accelerates the replacement of mtDNA and genomic DNA of R. o. kurumeus with those of R. o. ocellatus in hybrids. Genetically pure R. o. kurumeus eventually disappears and hybrids become more and more similar to R. o. ocellatus.
Acknowledgments
We thank S. R. Jeon, H. Wu, N. Aoyama and J. Kitamura for providing, or assistance with collecting specimens. Helpful discussions and/or comments on the manuscript were provided by T. Seki, M. Okauchi, K. Hosoya and A. Komaru. We thank T. Okazaki, H. Nagoya and T. Kitagawa for performing the PCR and molecular data analysis. This study was supported in part by Grant-in-Aid (No. 10041156) from the Ministry of Education, Science, Sports and Culture, Japan.
