The order Unionoida includes more than 850 species and is more diverse than any other group of freshwater bivalves [1] [2] . Unionoid mussels are widely distributed in all continents except Antarctica and are divided into 6 families (the Unionidae, Margaritiferidae, Etheriidae, Hyriidae, Iridinidae, and Mycetopodidae). In Japan, two families, the Unionidae and Margaritiferidae, have been recognized; however, there are some debates regarding the systematics of Japanese unionoid mussels. Kondo [3] classified them into 2 families, 3 subfamilies, 12 genera, 17 species, and 1 subspecies, but later revised his classification [4] to include 2 families, 2 subfamilies, 12 genera, and 18 species (Table 1). In the Unionidae, Kihira et al. [5] included one additional species, Lanceolaria oxyrhyncha (Sasanohagai in Japanese), and founded two subspecies, Nodularia douglasiae biwae (Tateboshigai in Japanese) and N. d. nipponensis (Ishigai in Japanese), for N. douglasiae in Kondo [4] and also two subspecies, Cristaria plicata clessini (Menkarasugai in Japanese) and C. p. plicata (Karasugai in Japanese), for C. plicata in Kondo [4] . The taxa classified by Kihira et al. are endemic to Lake Biwa

and/or the Yodo River. Although Graf and Cummings [2] reviewed the worldwide unionoids and proposed classification, their classification was still tentative as mentioned by themselves. At present, there is not enough information to deduce relationships between Japanese species and similar species in other parts of the world.

The classification of unionoid mussels has mainly been based on the morphological characteristics of the adults and/or the larvae such as beak sculpture, hinge teeth, shell length and larval hook. Shell characters in particular have attracted a great deal of attention for the classification of unionoid mussels because they can be used to classify fossils. However, Heard et al. [6] suggested that morphological similarities among unionoid species were caused in part by convergent or parallel evolution and not by their ancestry. They also insisted that unionoid systematics based on reproductive aspects more accurately reflected natural, evolutionary affinities than those based on morphological characteristics, and they thus revised the classification of North American unionoid mussels.

Studies using genetic information can objectively provide accurate phylogenetic relationships, reflecting morphological and reproductive differences, and are less influenced by convergent or parallel evolution [7] [8] [9] [10] . Recently, Lopes-Lima et al. [11] investigated the phylogenetic relationships of 70 unionoid species using the mitochondrial cytochrome oxidase subunit I (COI) and nuclear 28S rRNA sequences (1032-bp) and showed that unionid mussels formed three clades. They assigned three subfamilies, the Unioninae, Anodontinae, and Gonideinae, to the clades. However, they used only 1 of the 18 Japanese unionoid species identified by Kondo [4] . Takeuchi et al. [12] analyzed the mitochondrial COI and nuclear 18S rRNA and 28S rRNA genes to deduce relationships between the morphologically distinguished Margaritifera togakushiensis and M. laevis, and confirmed that M. togakushiensis was genetically distinct from M. laevis. Their results were further supported from an ecological viewpoint; however, they used only 2 of the 18 Japanese unionoid species. Therefore, the phylogenetic relationships of Japanese Unionoida have not been resolved. To elucidate the phylogenetic relationships of Japanese unionoid mussels, genetic studies using more species are needed.

The Japanese Ministry of the Environment [13] has designated 13 of the 18 Ja- panese unionoid species as endangered due to deterioration of freshwater systems by human activities [14] [15] [16] [17] . These filter-feeding mussels greatly influence ecological systems and play an important role in purifying water. Negishi et al. [18] unraveled the processes of degradation of unionoid habitats and tried to restore them. Genetic and ecological information is indispensable for preserving wildlife, and one can estimate how long a population is likely to survive by estimating its genetic diversity and population size. Phylogenetic relationships and classification corroborated by genetic analyses are also needed to transplant endangered mussels whose habitats seem to disappear in the near future and to rear them in institutions with genetic contamination avoided.

In this study, we determined mitochondrial 16S ribosomal DNA sequences (347-bp) and constructed trees to elucidate phylogenetic relationships and establish the systematics of Japanese Unionoida. We also constructed trees to reveal relationships of East Asian unionoid mussels using their 16S rDNA sequences (256-bp). We demonstrated that the phylogenetic relationships of Japanese Unionoida obtained in this study supported the systematics proposed by Kondo [4] . Therefore, the phylogenetic relationships and systematics of Japanese unionoid mussels based on genetic and morphological data were in agreement.

Phylogenetic relationships of Japanese unionoids based on 347 bp of the 16S rRNA gene are shown in Figure 1. There were 183 variable sites and 176 informative sites. Topologies depicted by MP and Bayesian trees were essentially

identical to that of the NJ tree. Based on our analyses, Japanese unionoid mussels were divided into two well-supported clades (98/97/0.78 and 100/100/1.00, NJ/ MP/Bayesian, respectively) corresponding to two families, the Margaritiferidae and Unionidae. In the Unionidae, there were two well-supported clades (94/76/ 0.84 and 96/93/1.00) corresponding to the subfamilies Unioninae and Gonideinae. Most genera formed clades with high statistical supports. The only exception was the genus Sinanodonta, which formed a poorly supported clade (59/59/ 0.93). When two or more specimens were used, most species formed clades with robust statistical supports, even though the specimens were collected in distant localities. For example, Obovalis omiensis specimens were obtained from Gifu and Kyoto Prefectures, those of Lanceolaria grayii from Gifu and Hiroshima Prefectures, and those of Pronodularia japanensis from Ibaraki, Hiroshima, Nara, Gifu, and Yamagata Prefectures. However, Inversiunio jokohamensis formed a marginally supported clade (86/76/0.61). Inversiunio yanagawensis and Margaritifera laevis did not form their own species clades, and two species, Sinanodonta lauta and S. japonica, exhibited complicated relationships with S. ogurae and S. calipygos. The two subspecies described by Kihira et al. [5] , Nodularia douglasiae biwae and N. d. nipponensis, did not form their own subspecies clades.

Phylogenetic relationships of East Asian unionoids based on 256 bp of the 16S rRNA gene are shown in Figure 2. There were 138 variable sites and 130 informative sites. Topologies depicted by MP and Bayesian trees were essentially identical to that of the NJ tree. Two robustly supported clades corresponding to the Margaritiferidae and Unionidae were recognized (100/100/1.00 and 100/99/ 1.00, respectively). Most genera and species formed clades; however, Sinanodonta did not form its own clade. The Inversiunio clade was well supported in the Japanese unionoid tree (Figure 1), but was only marginally supported in the East Asian unionoid tree (Figure 2) likely due to the shorter sequences used to make the latter tree. On the other hand, Margaritifera laevis was paraphyletic in the Japanese unionoid tree (Figure 1), but it formed a clade in the East Asian unionoid tree (Figure 2).

Japanese unionoid mussels were basally divided into two clades corresponding to the two families described by Kondo [4] who showed that the Margaritiferidae had interlamellar gill junctions arranged in diagonal rows, while the Unionidae had interlamellar gill junctions combined into vertical septa. Rosenberg et al. [27] presented preliminarily the molecular phylogeny of invertebrate animals, including the Unionoida, by analyzing the D6 region (about 150-bp) of the nuclear 28S rRNA gene. They showed that margaritiferid species could be distinguished from other unionoid species, which supported our results.

Kondo [4] reported that there were two subfamilies, the Unioninae and Gonideinae, in the family Unionidae. He regarded mussels having subtriangular and hooked glochidia as Unioninae and those having essentially semi-elliptical

and unhooked glochidia as Gonideinae. Our results supported Kondo’s subfamily classification. Kondo [4] further classified unionoid mussels into 12 genera and 18 species, mainly based on their shell morphology. In this study, most genera formed well supported clades, although the genus Sinanodonta was poorly supported. Most species also formed well supported clades, but one species formed a marginally supported clade or some did not form clades at all. Takeuchi et al. [12] showed that a phylogenetic tree based on mitochondrial COI sequences distinguished M. laevis from M. togakushiensis. Our results supported their results in the East Asian unionoid tree (Figure 2), but not in the Japanese unionoid tree (Figure 1). Of the three species of Inversiunio, I. jokohamensis was marginally supported and I. yanagawensis was paraphyletic, but I. yanagawensis and I. reinianus formed a clade together. Genetic distances between I. yanagawensis and I. reinianus (0.00581 in Table 3 and 0.00394 in Table 4) were as low as the intraspecific genetic distances within I. jokohamensis (0.00405 in Table 3 and 0.00419 in Table 4). Therefore, in this study, I. yanagawensis could not be genetically well-distinguished from I. reinianus. Complicated relationships between Sinanodonta species seem to be derived from large morphological variation within the genus, which sometimes confuses species identification [5] [28] . To obtain more robust phylogenetic relationships of those species, we need to collect and analyze more specimens and use another gene such as the mitochondrial cytochrome oxidase subunit I (COI) gene for molecular barcording.

Kihira et al. [5] recognized Lanceolaria oxyrhyncha as a valid species different from L. grayii, which is widely distributed in Japan. However, Kondo [3] claimed that L. oxyrhyncha could not be separated morphologically from L. grayii, and Shirai [29] showed that L. oxyrhyncha was not genetically distinct from L. grayii. Kihira et al. [5] recognized Cristaria plicata clessini as a valid subspecies. However, Kondo [3] showed that C. p. clessini seems to be a lacustrine type of C. plicata and considered it doubtful that C. p. clessini is a subspecies. Hence, we did not use L. oxyrhyncha or C. p. clessini in this study. Nodularia douglasiae was also problematic. Kihira et al. [5] established two subspecies, N. d. biwae and N. d. nipponensis, and Kondo [30] regarded the individuals with milky-white glochidia as N. d. biwae and the others with buff glochidia as N. d. nipponensis. Later, Kondo [4] resolved that glochidium colors did not seem to be sufficiently diagnostic to identify the subspecies because there was substantial variation in color as is also often the case with Inversidens brandti and Lanceolaria grayii. We showed in this study that one specimen of N. d. nipponensis (Un40-02) was more closely related to specimens of N. d. biwae than to the other specimens of N. d. nipponensis. Shirai [29] showed, based on mitochondrial DNA (COI + II), that Nodularia douglasiae is divided into two clades (east and west), but each subspecies did not form its own clade. Taken together, these results suggest that N. d. biwae and N. d. nipponensis are not valid as subspecies.

The phylogenetic relationships obtained in this study essentially supported the systematics proposed by Kondo [4] that are primarily based on morphological and larval traits. Never before have any published studies comprehensively in-