Vicariant speciation due to 1.55 Ma isolation of the Ryukyu islands, Japan, based on geological and GenBank data


Correspondence: Soichi Osozawa, Department of Earth Sciences, Graduate School of Science, Tohoku University, Sendai 980-8578, Japan.



The Ryukyu island arc, originally a continental margin arc, separated from the Chinese continent by the rifting of the Okinawa trough, a process which began at 1.55 million years ago (Ma) and continues to the present. In addition, the Ryukyu arc was simultaneously divided into the northern Amami–Okinawa and southern Yaeyama islands by the Kerama rift valley, and consequently formed two isolated island units. The Kuroshio warm current began to flow into the Okinawa trough from the Yonaguni Strait, and flow out through the Tsushima and Tokara straits also at 1.55 Ma, and these seaways effectively acted as barriers between the Ryukyu islands and Taiwan, China and Japan. Through this geological process, vicariant speciation generated Ryukyu endemic animal species. We support this hypothesis by drawing linearized maximum likelihood (ML) phylogenetic trees of the species in four endemic insect groups (peacock butterfly, Chinese windmill butterfly, golden-ringed dragonfly, window firefly) using GenBank sequence data. We determined the precise branching ages for these phylogenetic trees, and show simultaneous speciation at 1.55 Ma for Amami–Okinawa and Yaeyama units. The Taiwan and Tsushima straits, barriers between Taiwan and China, and Japan and Korea, respectively, did not form sufficient barriers to migration during glacial low stands, and species were intermingled. A marine embayment may have posed as a migration barrier between northern and southern China in the Quaternary or a little earlier. From our study we also estimate the precise molecular evolution rate and justify the molecular clock.


Precise age determination for a geological event is an important and commonly challenging objective for geologists. The molecular clock appears to be an attractive and powerful tool for evolutionary biologists, if DNA can be treated as a kind of fossil that records time and/or history. In practice, a molecular phylogenetic biologist requires geological and paleontological input to assign precise and reliable divergence ages to a phylogenetic trees of organisms (e.g. Nei & Kumar 2000). The fossil record, however, is commonly poor and seldom has sufficient resolution to apply directly to quantitative assessment of molecular evolution rate. In this paper, we combine newly determined ages of geological events with DNA analysis of endemic insect species on the Ryukyu islands to estimate branching ages and patterns for their phylogenetic trees and DNA substitution rates. Similar approaches to elucidate vicariant speciation using both geological and phylogenetic data have been done since 1990s (Patarnello et al. 1996; Losos & Glor 2003; Waters et al. 2004; Elias et al. 2009), but the geological data used in their studies are less accurate. An exception is Nishihara et al. (2009), who used robust geological data to show that placental mammals diverged in Africa, South America and Laurasia as a consequence of the break of the supercontinent Gondwana at 120 Ma.

The Ryukyu island arc is a relatively small island chain between Kyushu and Taiwan (Fig. 1A). We have recently reevaluated how and when these islands formed, during a transition from a pre-Quaternary continental margin arc to Quaternary island arc, based on new geological and geochronological data (Fig. 1B; Osozawa & Watanabe 2011; Osozawa et al. 2012a,b). The opening of the Okinawa trough by back-arc extension separated these islands from the mainland. The Okinawa trough opened synchronously along its entire length at 1.55 Ma, and as the sea barrier behind the islands evolved, the Quaternary Ryukyu limestone formed on the margins (most importantly the western margins) of each island under influence of the Kuroshio warm current (Fig. 1B).

Figure 1.

(A) Maps of the Ryukyu islands and the Japanese islands (superimposed). (B) The 3D morphological map of the Ryukyu islands and their surrounding areas. The Ryukyu island arc consists of a chain of mostly submerged small islands, formed along the Ryukyu trench. The Ryukyu island arc is associated with a back-arc spreading rift, the Okinawa trough, which start rifting at 1.55 Ma (Osozawa et al. 2012a). Three tectonic valleys, the Takara, Kerama and Yonaguni straits, divide the islands into the Japan, northern Ryukyu (Okinawa and Amami), southern Ryukyu (Yaeyama), and Taiwan island units. The Taiwan and Tsushima straits are much shallower, but separate these islands from the Chinese continent. Northern and southern China indicate the areas north and south of the Yangtze and Huai River drainage in the Chinese continent, respectively. (C) “Phylogenetic tree” of Japan and northern Tokara, Amami–Okinawa and southern Tokara (northern Ryukyu), Yaeyama (southern Ryukyu), and Taiwan island units, which were simultaneously isolated from the Chinese continent (“outgroup”) at 1.55 Ma (modified after Osozawa et al. 2012a). Later emergence of the Tokara islands from sea floor due to building of volcanic edifices (Osozawa et al. 2012a,b) is represented by later branching.

In addition to the limestone, the deposition of other marine sedimentary rocks on the western margins of the islands also records the time of island formation. The onlap of these marine deposits onto the present islands shows that the initial islands were smaller than the present ones. The marine sedimentary rocks noted above can be dated by Quaternary marine plankton fossils, especially nannoplankton, and the age of such fossils dates the time of island formation (Osozawa et al. 2012a).

The Tokara and Yonaguni straits separate the Ryukyu island arc from Japan and Taiwan, respectively (Fig. 1A,B). The Kuroshio current passes through these straits today and apparently did at 1.55 Ma. In addition, the Kerama Strait divides the Ryukyu arc into two parts, the northern Amami and Okinawa islands, and southern Yaeyama islands. These three straits are fault-controlled valleys that share a common tectonic history with the Okinawa trough (Fig. 1B), and should have acted as biological barriers (Osozawa et al. 2012a).

The Tsushima warm current, a branch of the Kuroshio current, began to flow into the Japan Sea through the Tsushima strait at 1.55 Ma. There is evidence that a warm current flowed through the Taiwan Strait at 1.55 Ma (Fig. 1A). These two straits act as biological barriers between Japan and Korea (including the northern China mainland), and between Taiwan and the southern China mainland, although the barrier effect was not totally effective owing to the shallowness of the straits (Fig. 1B), especially at glacial low stand of sea level (Osozawa et al. 2012a).

We constructed a “phylogenetic tree” of the Ryukyu islands, including Japan and Taiwan, relative to mainland China (Fig. 1C; Osozawa et al. 2012a). The Ryukyu islands host many endemic species (e.g. Azuma 2002). One of the probable mechanisms of such speciation is separation of these islands from the mainland by deep marine waters (the Okinawa trough) and from each other by deep straits. Such speciation is typically called vicariant speciation (Coyne & Orr 2004), if the Ryukyu endemic species were generated from an original species widely distributed in mainland China as a result of a sea barrier that separated each island. To test this hypothesis, one of the generally accepted mechanisms of speciation (Coyne & Orr 2004), we compiled DNA sequence data of the Ryukyu insect species from the GenBank database (, and drew their phylogenetic trees. The correspondence or lack thereof between branching of phylogenetic trees of endemic species and island evolution will help develop a model of vicariant speciation of the islands' insects, and to determine more accurate molecular clocks.

Insect species used in this review

The Ryukyu endemic insect species we evaluated from the published literature include peacock butterflies Papilio bianor Cramer, 1777 subspecies (Yagi et al. 2006), Chinese windmill butterfly Parides alcinous (Klug, 1836) subspecies (Kato & Yagi 2004), golden-ringed dragonfly Anotogaster sieboldii (Selys, 1854) species (Kiyoshi 2008) and window firefly Pyrocoelia species (Li et al. 2006). These analyzed insects and their distributional ranges are listed in Table S1 in Supporting Information (refer to Fig. 1).

Some butterflies have wing pattern and color variations specific to the Ryukyu islands, but their small differences in DNA sequences indicate that these butterflies have great migration ability and are unsuitable for the present analysis. Such butterflies are the lesser grass blue Zizina species (Yago et al. 2008), great mormon Papilio memnon Linnaeus, 1767 (Yoshio 2005) and spangle Papilio protenor Cramer, 1775 (Yagi 2003). Although DNA sequence data of Lamiinae (longhorn beetle) and Anoplophora species are available, they are excluded due to the possibility of artificial immigration to the Ryukyu islands (Ohbayashi et al. 2009).

Sequence data from GenBank database, phylogenetic analyses and estimation of divergence time

Sequence data of the above-mentioned species are recorded in the GenBank database, and available for our phylogenetic analyses. Our evolutionary analyses were conducted using MEGA5 (Tamura et al. 2011). The maximum likelihood (ML) trees were linearized to visualize the tree time increments.

Through this linearization, we can easily compare the obtained phylogenetic trees with the “phylogenetic tree” of the Ryukyu islands, which relates to the geohistorical isolation of these islands, the Japan main islands, Taiwan and mainland China (Fig. 1C; Osozawa et al. 2012a). If our hypothesis is correct that island isolation triggered the insect speciation, both biological and physical trees should overlap each other, and the divergence time of 1.55 Ma can be applied at the simultaneous branch point on each phylogenetic tree. However, there are some variations and exceptions, as detailed in each phylogenetic tree section detailed below.

In each ML tree, the evolutionary history was inferred using the Tamura–Nei model (Tamura & Nei 1993). The tree with the highest log-likelihood value is shown in each session. The initial tree for the heuristic search was obtained as follows. When the number of common sites was <100 or less than one-quarter of the total number of sites, the maximum parsimony method. was used; otherwise the BIONJ method with MCL distance matrix was used. The trees were drawn to scale, with branch lengths measured in number of substitutions per site. Codon positions included were 1st+2nd+3rd+Noncoding. All positions containing gaps and missing data were eliminated.

We made diagrams of corrected distance versus number of transitions and transversions for each evaluated species (Tables S2–S5 in Supporting Information), and confirmed that the base substitutions are not saturated.

In ML tree initial screen, the MEGA software shows P-values that the tree satisfies the molecular clock or not. If the tree does not pass this clock test, such a sequence datum with extremely different branch length before linearization is eliminated until the consequent tree passes the test. ML trees that passed the clock test are shown in Figures 2 and 3.

Figure 2.

(A) The linearized phylogenetic ML tree of Papilio bianor subspecies using ND5 gene sequence (about 800 bp). Highest log-likelihood = −1145.48. The molecular clock is not rejected with P < 0.9458. (B) The linearized phylogenetic ML trees of Parides alcinous subspecies using ND5 gene sequence (about 800 bp). Highest log-likelihood = −1161.30. The molecular clock is not rejected with P < 0.5026. Bootstrap values are shown at each branch. The 95% confidence intervals (CI) of nodes are shown as rectangles.

Figure 3.

(A) The linearized phylogenetic ML tree of Anotogaster sieboldii obtained from various areas. The tree was constructed using COI (about 700 bp) and COII (about 700 bp) gene sequences. Highest log-likelihood = −2631.82. The molecular clock is not rejected with P < 0.2990. (B) The linearized phylogenetic ML trees of Pyrocoelia species using 16S rRNA (about 450 bp) gene sequence. Highest log-likelihood = −1454.84. The molecular clock is not rejected with P < 0.0726. Bootstrap values are shown at each branch. The 95% confidence intervals (CI) of nodes are shown as rectangles.

Phylogenetic trees, divergence time and explanation

Peacock butterfly Papilio bianor

Description of species

Yagi et al. (2006) examined the phylogenetic relationships of subgenus Achillides belonging to genus Papilio (Lepidoptera: Papilionidae), so-called swallowtail butterflies. Papilio bianor Cramer, 1777, a species of the Achillides, further contains many characteristic localized subspecies, although these subspecies are thereafter reconstructed and grouped into three species (Yoshimoto 1998; Shirôzu 2006; Aoki et al. 2009). Papilio bianor feeds on several kinds of forest Rutaceae, whereas the Ryukyu subspecies, whose habitat is near the seashore, feeds on a coastal kind of Rutaceae (Fukuda et al. 1982). All the subspecies were systematically analyzed using the mitochondrial ND5 nucleotide sequence (Yagi et al. 2006).

The subspecies used in the study of Yagi et al. (2006) are Papilio bianor dehaanii C.Felder and R.Felder, 1864 from the Japan main islands (Hokkaido, Honshu, Shikoku and Kyushu), Tsushima, Korea and Sakhalin, but also distributed in the Primorskii and Yanshan mountains in northern China (Li et al. 2009); hachijonis Matsumura, 1919 from the Izu islands; tokaraensis Fujioka, 1975 from the northern Tokara islands; amamiensis (Fujioka, 1981) from the Amami islands; ryukyuensis Fujioka, 1975 from the Okinawa islands; okinawensis Fruhstorfer, 1898 from the Yaeyama islands (note not Okinawa islands: the Natural History Museum, London, mislabeled the exhibit, according to Shirôzu 2006); thrasymedes Fruhstorfer, 1909 from Taiwan; kotoensis Sonan, 1927 from the Lyudao and Lanyu islands of the Luzon arc offshore of the Taiwan main island; and bianor Cramer, 1777 from the southern China mainland (Table S1). The border between the ranges of P. b. dehaanii and P. b. bianor in north and south China lies at 32–34° in the Yangtze and Huai river drainage (Omoto 1967; Aoki et al. 2009).

Phylogenetic tree and explanation

Papilio palinurus Fabricius, 1787 from Malaysia is used as an outgroup for phylogenetic analysis, considering the tree topology containing the related species shown by Yagi et al. (2006). The ND5 sequence-based ML tree of P. bianor subspecies shows four clades (clades I–IV) (Fig. 2A). Clade I contains dehaanii, hachijonis and tokaraensis from the Japan main islands, Tsushima, Korea, Sakhalin, Izu islands and northern Tokara islands. Clade II includes amamiensis and ryukyuensis from Amami and Okinawa islands. Clade III consists of only okinawensis from Yaeyama islands. Clade IV comprises thrasymedes, kotoensis and bianor from Taiwan, Lyudao and Lanyu islands, and southern China. In the ML tree (Fig. 2A), these clades seem to form two major groups (both consist of two clades), geographically dividing into northern and southern groups (this pattern is similar to Anotogaster species, see Fig. 3A). However, considering the weak support for the nodes (bootstrap values 57% and 49%) (Fig. 2A), the tree shows rather simultaneous division to four clades. We put 1.55 Ma (the isolation time of the island, see Fig. 1C) at the branching point between clades I and II (Fig. 2A), assuming that the divergence time of subspecies fits the first isolation time of islands (Fig. 1C).

On the other hand, the sequence divergences within each clade are small although the samples are from wide distribution areas, even different subspecies (Fig. 2A). In particular, subspecies dehaanii in Japan, Tsushima, Sakhalin, and Korea, hachijonis in the Izu islands, and tokaraensis in the northern Tokara islands show identical sequence (Fig. 2A). Similarly, subspecies thrasymedes in Taiwan, kotoensis in Lyudao and Lanyu islands, and bianor in southern China are not very specialized (Fig. 2A). According to Yagi et al. (2006), Papilio polyctor Boisduval, 1836 in Laos, Thailand and Kashmir is also included in clade IV (southern China and Taiwan) and all are regarded to be the same species. Low genetic divergences over the very wide areas including the above isolated islands may indicate that considerable migration of the species occurs during times that the present straits were subaerial or narrow, especially during low sea level associated with the last glacial period. The time since the present islands formed would not be long enough for causing genetic divergences in each clade.

Figure 2A shows that the Tsushima Strait between Japan and Korea, and the Taiwan Strait between Taiwan and southern China, both of which were formed at 1.55 Ma (Osozawa et al. 2012a), were not effective barriers that drove speciation probably due to their narrowness and the connection to China continent in the glacial period. During that time, migration of populations would occur and the genetic divergence may be erased by replacement with the continental populations.

In terms of north (clade I, dehaanii) and south (clade IV, bianor) China clade formation (Fig. 2A), we speculate that the Quaternary marine transgression over the present Yangtze and Huai river drainage (Fig. 1A; Ichihara 2001) might have acted as a barrier between north and south China. In addition, the Asian Monsoon, linked to Himalaya uplift and affected by the Dansgaard–Oeschger Cycle, began 1.5 Ma (Tada 2005), and north and south China experienced contrasting climate as a result. Future geological and geochronological studies in the Chinese lowland will clarify this hypothesis.

Chinese windmill Parides alcinous

Description of species

Parides alcinous (Klug, 1836) (Lepidoptera: Papilionidae) contains many geographic subspecies in a manner similar to Papilio bianor. Parides alcinous feeds on Aristolochiaceae, and the Ryukyu subspecies described below feeds on the endemic Aristolochia liukiuensis Hatus. (Fukuda et al. 1982). All subspecies were systematically analyzed using mitochondrial ND5 gene sequence (Kato & Yagi 2004).

The subspecies used in the study of Kato and Yagi (2004) were Parides alcinous alcinous (Klug, 1836) from the Japan main islands, Tsushima, Korea and Russian Primorskii; yakushimana (Esaki and Umeno, 1929) from Yaku-shima island; loochooana (Rothschild, 1896) from Amami and Okinawa islands; miyakoensis (Omoto, 1960) from Miyako-jima island; bradanus (Fruhstorfer, 1908) from the Yaeyama islands; and mansonensis (Fruhstorfer, 1901) from Taiwan and the southern China mainland (Table S1).

Distribution of P. alcinous in China is separated into northern (ssp. alcinous) and southern (ssp. mansonensis) areas. Parides alcinous is absent in the Jiangsu and Anhui provinces, i.e., the Yangtze and Huai river drainage (Fig. 1A; Tong 1993). In north China, subspecies alcinous is distributed in Beijing city, Heilongjiang and Shandong provinces (Tong 1993), including Mt. Tai (Lu 1990), which are wider areas shown in Kato and Yagi (2004).

Phylogenetic tree and explanation

The phylogenetic tree of P. alcinous subspecies shows four clades (clades I–IV) with Parides laos (Riley & Godfrey, 1921) of Indochina as an outgroup species (Fig. 2B). The outgroup choice is based on the tree topology shown in Kato and Yagi (2004). Clade I contains alcinous from Japan, Tsushima, Korea, and Primorskii, and yakushimana from Yaku-shima. Clade II includes loochooana from Amami and Okinawa. Clade III consists of miyakoensis and bradanus from Miyako and Yaeyama, respectively. Clade IV comprises mansonensis from Taiwan and southern China. The Taiwan–China clade (clade IV) branched first, then the Japan–Korea–Russia clade (clade I), and finally the Amami–Okinawa and Miyako–Yaeyama clades (clades II and III) diverged. In Figure 2B, the branching point between clade I and the clade II/III is set at 1.55 Ma. The divergence to four clades resembles the geographical divisions of islands (Fig. 1C).

Based on the evolutionary rate of the ND5 gene, the divergence between subspecies mansonensis (Taiwan–southern China) and other subspecies (northern China, Korea, Japan and Ryukyu populations) occurred at 4.0 Ma (Fig. 2B). This old speciation event predates and is unrelated to the 1.55 Ma isolation of the Ryukyu islands and also the Quaternary sea barrier between north and south China expected for P. bianor. The old isolated mansonensis population formed on the China mainland and probably migrated to Taiwan in the glacial period, long after the Ryukyu populations had been isolated. This suggests that the Yonaguni strait drove speciation.

Separation of the Amami–Okinawa (clade II) and Yaeyama (clade III) clades was delayed and occurred at 1.0 Ma (Fig. 2B). The delay may be explained by the rather strong mobility of this species, possibly combined with imperfect barrier of the Okinawa trough and the Kerama strait at the beginning of the rifting.

Similar to P. bianor, the Taiwan and Tsushima straits did not act as barriers to migration of P. alcinous due to narrowness and connection to mainland China during the glacial period.

Subspecies alcinous and yakushimana, and subspecies miyakoensis and bradanus are genetically the same, but divided into different subspecies due to distinct wing colors and patterns (Fig. 2B).

Golden-ringed dragonfly Anotogaster

Description of species

Anotogaster sieboldii (Selys, 1854) (Cordulegastridae: Odonata) is distributed in Japan–Korea, Amami–Okinawa, Yaeyama, Taiwan and southern China (Table S1). The phylogenetic relationships of this species have been analyzed using COI and COII genes (Kiyoshi 2008).

Anotogaster sieboldii is a large and strong dragonfly living along the upper and middle reaches of streams. For example, this dragonfly can fly over the narrow strait between a small island and the mainland, within the Matsushima archipelago, Japan (Osozawa & Osozawa 2003), although there is no suitable habitat on the island because of the lack of streams.

The Ryukyu species is known to differ in size, morphology and behavior from Japan mainland species (Ishida et al. 1988). The Ryukyu species is a stronger flier than the Japan mainland species.

Phylogenetic tree and explanation

In our phylogenetic analysis, Cordulegaster bidentata Selys, 1843 is selected as an outgroup, considering the tree topology containing the related species shown by Kiyoshi (2008).

We assign 1.55 Ma for the branching of Japan–Korea and Amami–Okinawa (Fig. 3A). Four clades of Japan–Korea, Amami–Okinawa, Yaeyama and Taiwan–China are recognized (Fig. 3A), as similar to the clades I–IV of P. bianor and P. alcinous (Fig. 2). However, two higher grade clades are recognized, consisting of former two and latter two populations (Fig. 3A). There already exist two lineages and speciation before 1.55 Ma, as also recognized in P. alcinous (Fig. 2B). However, the Kerama Strait is the border between these two higher clades, rather than the Yonaguni Strait which separates the clades of P. alcinous subspecies. The relationship of these two major populations in China has not been previously studied, and will be discussed later.

Branching of Amami–Okinawa (clade II) from Japan–Korea (clade I) at 1.55 Ma, and that of Yaeyama (clade III) from Taiwan–China (clade IV) at 1.0 Ma (Fig. 3A) fit our geological model fairly well.

Within the Amami–Okinawa clade II, branching of Amami and Okinawa at 1.2 Ma (Fig. 3A) needs additional explanation. The distance of Amami Oshima from Okinawa-jima is 150 km, and small islands between them lack its habitat; therefore, migration of A. sieboldii between Amami Oshima and Okinawa has probably not occurred since 1.2 Ma.

Such migration has probably occurred between Yaku-shima and Kyushu in the Japan–Korea clade I, and Ishigaki-jima and Iriomote-jima in the Yaeyama clade III (Fig. 3A). Also as noted before, the Tsushima and Taiwan straits did not act as barriers, as shown by their minor genetic divergence within the Japan–Korea clade I and the Taiwan–China clade IV (Fig. 3A).

The Japan–Korea clade I and Amami–Okinawa clade II constitute the northern China lineage, and the Taiwan–China clade IV and the Yaeyama clade III constitute the southern China lineage (Fig. 3A). Kiyoshi (2008) treated the both lineages as the same species named “Anotogaster sieboldii”. However, according to Kiyoshi (T Kiyoshi, pers. comm., 2012), the southern China lineage may differ from A. sieboldii and correspond to Anotogaster flaveola Lohmann, 1993, although Karube et al. (2012) described the lineage as Anotogaster clossi Fraser, 1919 based on molecular phylogenetic and morphological studies. The border between these species may lie in the Yangtze and Huai river drainage on the Chinese mainland (Fig. 1A), although A. sieboldii is reported from Jiangxi Province, southern China (Karube et al. 2012).

Window firefly Pyrocoelia

Description of species

Most fireflies are terrestrial and feed on land snails. The genus Pyrocoelia (Lampyridae: Lampyrinae) has windowpanes on prothorax, just above eyes, and is commonly called the “window firefly” by the Japanese and Chinese. The females are wingless and can not cross the sea. Therefore Pyrocoelia (Table S1) is a suitable genus for the present analysis and the purpose.

The phylogenetic relationships of Pyrocoelia have been analyzed using mitochondrial 16S rRNA gene (Suzuki 1997; Li et al. 2006) and the COI gene (Oba et al. 2011). Suzuki (1997) examined a Korean, two Japanese and all the Ryukyu species. Li et al. (2006) examined the species from Yunnan, south China. Oba et al. (2011) examined several species from the Japanese and Ryukyu islands.

Among Yunnan species, Pyrocoelia praetexta Olivier, 1911 is also known from Taiwan, called the “mountain window firefly”. It has multiple generations other than winter, but is dominant in autumn (Ho & Chu 2002). Pyrocoelia rufa E.Olivier, 1886 is distributed in Tsushima of Japan, Korea and China. Pyrocoelia rufa is called “autumn firefly” in Tsushima. Pyrocoelia miyako Nakane, 1981 endemic to Miyako-jima, Ryukyu, resembles P. rufa, and the adult appears throughout the year but dominantly in spring and autumn. Pyrocoelia atripennis Lewis, 1896 is endemic and common in the Yaeyama islands, but the adult appears only in autumn. Pyrocoelia rufa, P. miyako, and P. atripennis produce continuously glowing emissions (Oba et al. 2011).

Japan yields two endemic species, Pyrocoelia fumosa (Gorham, 1883) and Pyrocoelia discicollis (Kiesenwetter, 1874), mostly distributed in eastern and western Japan, respectively. Okinawa-jima yields endemic Pyrocoelia matsumurai matsumurai Nakane, 1963, and Kume-jima yields its subspecies matsumurai kumejimensis (Chûjô and M.Satô, 1972). The Amami islands yield the endemic Pyrocoelia oshimana Nakane, 1985. The Yaeyama islands yield the other endemic species, Pyrocoelia abdominalis Nakane, 1977 and Pyrocoelia iriomotensis Nakane, 1985.

Phylogenetic tree and explanation

The tree (Fig. 3B) is constructed using sequence data of Suzuki (1997) and Li et al. (2006). Lucidina accensa Gorham, 1883 is selected as an outgroup, considering the results from the phylogenetic tree containing both Pyrocoelia and Lucidina (Sagegami-Oba et al. 2007). Two major lineages (lineages I and II) are recognized (Fig. 3B). The two endemic species of Yaeyama are included in these two major lineages, respectively (Fig. 3B).

Lineage I (autumn firefly) was divided into three clades, consisting of the Yunnan, Korea–Tsushima and Yaeyama clades (Fig. 3B). Lineage II comprises five clades: the southern China clade (amplissima), two Japanese clades (fumosa Honshu and discicollis Kyushu), the Amami–Okinawa clade (oshimana and matsumurai) and the Yaeyama clade (abdominalis). We assign a time of 1.55 Ma to the node between P. fumosa, Honshu and Amami–Okinawa clades in lineage II (Fig. 3B). The divergence time between P. discicollis (Japan) and P. abdominalis (Yaeyama) is close to that of the fumosa (oshimana, matsumurai) node (Fig. 3B), which fits to our geological model shown in Figure 1C.

Pyrocoelia rufa in Korea and Tsushima is genetically similar (Fig. 3B), indicating imperfect barrier of the Tsushima Strait. Further analyses of P. praetexta in Taiwan in addition to China and other Taiwan species are required. Due to little migration ability of the female, the genetic distance of each island endemic species is expected to be large. However, the branches of P. miyako and P. atripennis are relatively short (Fig. 3B), which agree with involvement of Miyako-jima in the Yaeyama clade (Fig. 1C).

Isolation of island endemic species or subspecies in northern and southern Ryukyu at 1.55 Ma

Amami–Okinawa and Yaeyama islands were separated from the Chinese continent by the rifting of the Okinawa trough. Both island units were also separated from each other by the contemporaneously rifted Kerama Strait, and their northern and southern ends were further separated from the Japan islands unit and Taiwan by the Tokara Strait and the Yonaguni Strait, respectively (Osozawa et al. 2012a). The Kuroshio warm current also began to flow into the Okinawa trough through the Yonaguni gateway (Osozawa et al. 2012a), and effectively isolated Yaeyama and Amami–Okinawa from Taiwan–China. Therefore, Amami–Okinawa and Yaeyama islands units have been isolated since 1.55 Ma. This isolation probably produced endemic species and subspecies as shown in Figures 2 and 3. Thus, this study presents strong evidence for the impact of physical isolation on speciation and generation of new endemic species: vicariant speciation.

Note that only the doubly bifurcating tree form could not be simply connected with the vicariant speciation. We propose that the tree shape reflects the simultaneous formation of sea barriers at 1.55 Ma (Fig. 1C; Osozawa et al. 2012a), which drove vicariant speciation.

Seaways separating Japan from Korea and northern China, and separating Taiwan from southern China are the Tsushima and Taiwan straits, respectively, formed since 1.55 Ma (Osozawa et al. 2012a). This separation is expected to have produced endemic species in Japan and Taiwan. However, isolation was not always perfect, probably because land bridges were formed during glacial lowstands. Due to revised connections and gene flows, Japan–Korea (northern China) and Taiwan–southern China species are included in the same clade, exemplified as P. bianor, P. alcinous, Anotogaster and P. rufa.

Several basic questions remain unsolved. When insect speciation occurred on the northern and southern China mainland, and what influenced this speciation, are not well known. We speculate that the sea barrier produced by marine transgression over the present Yangtze and Huai river drainage, combined with the beginning of the Monsoon and the consequent north–south climate difference, may have been a factor, as mentioned above. Further field and genetic investigation, especially for Chinese species, combined with Quaternary geological studies in the central Chinese plain (Fig. 1A), are needed to explore this problem.

Molecular evolutionary rate

Our additional contribution is precisely estimated molecular evolutionary rate for each analyzed species (Table S1). Papilio binor and Parides alcinous have relatively low evolutionary rates in ND5 (0.81 and 0.55% substitutions/site per m.y. (D/m.y.)). Anotogaster has a relatively high rate in COI and COII (2.14% D/m.y.). Pyrocoelia also has a comparable rate in 16S rRNA (1.14% D/m.y.) (Table S1; Figs 2, 3). In early studies, Yagi et al. (2006, 2001) estimated the evolutionary rate in ND5 of P. bianor and Parnassius species (Papilionidae) as 1.3% D/m.y., and Kato and Yagi (2004) estimated the rate in ND5 of P. alcinous as 0.5% D/m.y., which are based on old and insufficient geological data for the Ryukyu islands.

The evolutionary rate of mtDNA is inferred from combined data of seven arthropod species by Brower (1994), which shows about 1.1–1.2% D/m.y. evolution rate. The rates in COI of the cave beetle Bathysciine (0.71 and 0.72% D/m.y.), the beetle Tetraopes (0.75% D/m.y.) and the cicada Maoricicada (0.77% D/m.y.) are compiled by Quek et al. (2004). The lower rate of ND5 is reported for the carabine ground beetles (0.14% D/m.y., Su et al. 2001).

In the present study, the evolutionary rates are almost comparable to those independently obtained by the abovementioned studies. Therefore, the present result obtained by putting 1.55 Ma to the specified branch based on our robust geological and phylogenetic data indicates that these evolutionary rates are considered to be appropriate.


The authors thank Drs S Chiba, T Kiyoshi and M Takahashi for encouragement for biological analyses, and Dr Bor-ming Jahn, who consulted on English spelling of authors, titles and publishers of books written in Chinese. Dr T Kiyoshi kindly offered unpublished taxonomic information of Anotogaster. Drs K Tamura and G Stecher generously provided information about evolutionary rates and pairwise distances in MEGA software. The authors are particularly grateful to Dr T Okamoto who read the manuscript carefully, gave the authors a lot of critical comments, and offered the original drawing of Fig 1A.