Takashi Yagi, Division of Radiobiology and Environmental Science, Research Institute for Advanced Science and Technology, Osaka Prefecture University, 1-2 Gakuen-cho, Sakai, Osaka 599-8570, Japan. E-mail: email@example.com
Abstract. The phylogeny of the butterflies Parides (Byasa) alcinous caught at various localities in Japan and the Ryukyu Islands and in the eastern part of the Eurasian Continent was analysed using mitochondrial DNA sequences coding for NADH dehydrogenase subunit 5 (778 bp). The same phylogenetic relationship among P. (B.) alcinous subspecies was obtained with all analytical methods used, and was supported by high bootstrap values. The female butterfly wing pattern that characterizes each subspecies was unrelated to the phylogenetic relationship among the subspecies. The phylogenetic trees show that one group composed of ssp. alcinous and yakushimanus, which are distributed in the main area of the Japan Archipelago, the Korean Peninsula and southern Primorski of Russia, and the other group comprised of ssp. loochooanus, bradanus and miyakoensis, all of which are distributed in the Ryukyu Islands, diverged from a common ancestor. The ssp. loochooanus distributed in Amami and Okinawa Islands then diverged, and ssp. bradanus and miyakoensis distributed in Yaeyama and the Miyako Islands, respectively, finally diverged. This divergence order nearly agrees with the palaeogeography of the Ryukyu Islands that has been established in Pliocene and Pleistocene (0.2–2 MYA), suggesting that P. (B.) alcinous has been isolated in the Ryukyu Islands since the establishment of the islands.
Geographical races or subspecies within the same species of biological organism might have morphological differences such as body size, its shape and colour pattern, and such differences often show a clinal or mosaic pattern of distribution (Futuyma, 1998). In some instances, a hybrid zone occurs along neighbouring populations, suggesting the existence of a genetic distance between them. When and how the geographical races differentiated from an ancestor population and then their present distributions established are major problems in systematic or evolutionary biology.
The aristolochiae-feeding papilionid butterfly Parides (Byasa) alcinous (Klug, 1836) is distributed widely, including the subtropical, temperate and subfrigid regions of Japan and neighbouring countries of east Asia. This butterfly is sexually dimorphic in wing colour. The ground and marking colours of the female wing are yellowish-brown, whereas those of the male are black. This species has been divided into six subspecies according to the colour pattern of the wing and abdominal hair in females (Fig. 1) (Kawazoé & Wakabayashi, 1976; Fujioka et al., 1997). The subspecies alcinous (Klug, 1836) (Fig. 2A) and yakushimanus (Esaki & Umeno, 1929) (Fig. 2B) are distributed in the temperate region of Japan, and the ground and marking colours of the former are yellowish, whereas those of the latter are black, like the male. The remaining three Japanese subspecies loochooanus (Rothschild, 1896) (Fig. 2C), miyakoensis (Omoto, 1960) (Fig. 2D) and bradanus (Fruhstorfer, 1908) (Fig. 2E) are distributed in the subtropical Ryukyu Islands located to the southwest of the Japanese mainlands. For the subspecies loochooanus and bradanus inhabiting the northern and southern Ryukyu Islands, respectively, the ground colour of the female wing is dark brown and the markings are more reddish in the latter than in the former. For the subspecies miyakoensis inhabiting the central Ryukyu Islands, interestingly the ground and marking colours of the female wing are yellowish-brown as for the temperate subspecies alcinous. On the other hand, for the subspecies mansonensis (Fig. 2F) (Fruhstorfer, 1901) inhabiting Taiwan and mainland China, the ground and marking colours of the female wings are intermediate between spp. alcinous and bradanus. Thus, the wing colour characters expressed in each subspecies of this butterfly show a geographically mosaic, but not a clinal, distribution pattern.
Several ecological and physiological differences exist among and within these subspecies. Seta (1996, 1997) reported that hybrid individuals between ssp. alcinous and bradanus often failed to undergo normal development, suggesting the existence of a genetic difference between them. Within ssp. alcinous, geographical and local differences in pupal diapause, egg size and host-plant selection occur (Kato, 2000, 2001; unpublished observations). Thus, biogeographical questions exist as to how and when these subspecies originated and then formed the present distribution in the Japan Archipelago. For these questions, morphological characters such as wing patterns and genital organs are useless, although they are suitable for taxonomic classification. No study has been carried out so far to elucidate the phylogeny of P. (B.) alcinous.
Analysing the phylogeny of butterflies at various localities using nucleotide sequences of mitochondrial DNA is a very effective method. We have previously shown that the mitochondrial NADH dehydrogenase subunit 5 (ND5) gene is useful for revealing the molecular phylogeny of the papilionid butterflies at low taxonomic levels (Yagi et al., 1999). In the present study, we investigated the phylogenetic relationships among subspecies of P. (B.) alcinous at various localities in Japan and east Asia, using various analytical methods based on the ND5 nucleotide sequences. We discuss the process for establishing the present distribution of the butterfly in the Japan Archipelago, mainly using the molecular phylogenetic tree constructed with the unweighted pair-group method using an arithmetic average (upgma), because the upgma tree is appropriate for the estimation of divergence time at the low taxonomic level.
Materials and methods
All butterflies were collected by the authors or supplied by many entomologists (Fig. 1; Table 1). Parides (Byasa) alcinous collected in Japan, Korea and Russia were stored in ethanol, and those collected in other areas were kept as dried specimens. Dried specimens of some related species, P. (B.) laos (Riley & Godfrey), P. (B.) impediens Seitz, P. (B.) plutonius Oberthur, P. (B.) polyeuctes (Doubleday) and P. (B.) mencius (C. & R. Felder) were also used for the analysis.
DNA was extracted from the muscles of the thorax or legs of the butterflies using the method described previously (Yagi et al., 1999). A part of the mitochondrial gene coding for ND5 was amplified using the polymerase chain reaction (PCR). The PCR primers and the sequencing protocol were as described previously (Yagi et al., 1999). The nucleotide sequences (778 bp) of the PCR products were directly determined with the BigDyeTerminator Cycle Sequencing FS Kit using an automated DNA sequencer 377 (PE Applied Biosystems). The ND5 nucleotide sequences of the samples were edited with genetyxmac software (version 11.0) and aligned with yooedit software (version 1.63). The nucleotide sequences were registered in the DDBJ/GenBank/EMBL database (Table 1).
Phylogenetic trees were constructed with the upgma and the neighbour-joining (NJ) method, which are included in the phylip software package (version 3.572c) (Felsenstein, 1993). The phylogenetic tree was also constructed with the most parsimonious (MP) method of the paup software package (version 4.0b10) (Swofford, 1998). In the phylogenetic analysis, Troides helena (Linnaeus) and Troides amphrysus (Cramer), whose sequences were obtained from the DDBJ/GenBank/EMBL database, were added. For the NJ and MP methods, Pachilopta aristolochiae (Fabricius) was used as the outgroup species. Pachilopta Reakirt, Troides Hübner and Byasa Moore belong to the tribe Troidini.
Genetic distances in the NJ and upgma methods were calculated using Kimura's two-parameter method (Kimura, 1980). Multiple substitutions were corrected using Kimura's formula (Kimura, 1980). The transition/transversion ratio was fixed at 2.0. The confidence level of branching in the phylogenetic trees (NJ and upgma) was evaluated with the bootstrap test based on 100 replications (Felsenstein, 1985). In the MP method, the bootstrap replication (100 times) test with the full heuristic search was performed, and the 50% majority-rule tree was obtained.
In all ND5 nucleotide sequences determined and aligned, neither deletions nor insertions were found. The G + C contents of the sequences were between 17.2 and 18.0 and nearly constant in all samples. Most nucleotide substitutions were found at the synonymous codon positions. These results suggest the lack of apparent effects of the base composition bias and multiple nucleotide substitutions on the phylogenetic trees.
For the upgma and NJ analyses, pairwise genetic distances among P. (B.) alcinous subspecies and other related species were calculated using Kimura's two-parameter method. The distances between any two samples within a subspecies of P. (B.) alcinous were too small (D < 0.003) to show reliable phylogenetic relationships. The distances among most subspecies of P. (B.) alcinous and among species of the subgenus Byasa were large enough (D > 0.01) to give reliable phylogenetic relationships. Surprisingly, ssp. alcinous and yakushimanus have an identical ND5 sequence, and ssp. bradanus and miyakoensis have only one nucleotide difference in the sequence (D = 0.013).
The MP analysis resulted in a tree with a length of 307. The consistency and homoplasy indices and those excluding uninformative characters were 0.74, 0.26, 0.64 and 0.36, respectively. The retention and rescaled consistency indices were 0.84 and 0.62, respectively. The 50% majority-rule consensus MP tree is shown in Fig. 3(A).
Figure 3 shows the phylogenetic trees with bootstrap values >40 constructed using three methods (MP, upgma and NJ). The upgma as well as other methods (MP and NJ) yielded trees with a similar topology. Partial discrepancies among the trees were found at the branching points with bootstrap values <40. Because the effects of various factors affecting the neutral molecular clock are suggested to be kept to a minimum in the present study, the upgma tree is used to estimate divergence time, as discussed below.
According to three phylogenetic trees (Fig. 3), one group consisting of P. (B.) mensius and P. (B.) polyeuctes, and the other group consisting of P. (B.) impediens, P. (B.) plutonius, P. (B.) laos and P. (B.) alcinous diverged from a common Byasa ancestor. The divergence order of P. (B.) impediens, P. (B.) plutonius and P. (B.) laos is uncertain because the order was different among the trees and bootstrap values at the divergence points are low.
The divergence order of all P. (B.) alcinous subspecies was identical in all the trees and was supported by high bootstrap values. One group of P. (B.) alcinous subspecies, alcinous and yakushimanus, distributed in the main area of the Japan Archipelago, the Korean Peninsula and southern Primorski of Russia, and the other subspecies group, loochooanus, bradanus and miyakoensis, all of which are distributed in the Ryukyu Islands diverged from an ancestor. The subspecies loochooanus, distributed in Amami and Okinawa Islands, then diverged, and ssp. bradanus and miyakoensis, distributed in Yaeyama and Miyako Islands, respectively, finally diverged. This divergence order essentially agrees with that of the phylogenetic relationships of reptiles and amphibians (Toda et al., 1997; Ota, 1998; Tanaka-Ueno et al., 1998), and wood-feeding cockroaches (Maekawa et al., 1999) distributed in the Ryukyu Islands, suggesting that it would reflect the geohistory of the island formation.
Genetic distance and classification of subspecies
In the molecular phylogenetic analysis, P. (B.) alcinous was separated into four phylogenetically distinct groups (Fig. 3). The first one is ssp. alcinous and yakushimanus. The second one is ssp. loochooanus. The third one is ssp. bradanus and miyakoensis, and the fourth one is ssp. mansonensis. The pairwise Kimura's two-parameter distances among these four groups are large enough to judge that each group has been reproductively isolated for a long time. This is consistent with the studies on artificial hybridization between ssp. alcinous and ssp. loochooanus, and between ssp. alcinous and ssp. bradanus, which showed that all female F1 hybrids are sterile (Seta, 1996, 1997).
Within each P. (B.) alcinous phylogenetically distinct group, the genetic differences among the individual butterflies are very small. Kimura's two-parameter distance between ssp. bradanus and ssp. miyakoensis is only 0.0013, whereas that between ssp. bradanus and ssp. alcinous is 0.0169. ND5 nucleotide sequences of ssp. yakushimanus are identical to those of most individuals of ssp. alcinous. These results disclosed that the pairwise genetic distances among the subspecies of P. (B.) alcinous are quite varied, and that genes with a higher evolutionary rate should be used to analyse the reliable phylogenetic relationships among individuals within the group.
Evolution of wing pattern and ecological characteristics
The female wing colour pattern of this species is divided into two types; one is light yellowish-brown with dark yellow spots in the submarginal region of the hindwings, belonging to ssp. alcinous (Fig. 2A), miyakoensis (Fig. 2D) and mansonensis (Fig. 2F); the other is dark brown or black with red spots in the submarginal region of the hindwings, belonging to ssp. yakushimanus (Fig. 2B), loochooanus (Fig. 2C) and bradanus (Fig. 2E). Because the distribution of their characters shows a geographically mosaic pattern, phylogenetic analysis of these subspecies is interesting. The phylogenetic trees (Fig. 3) suggest that the wing colour types are unrelated to the phylogeny of the subspecies, and that the change in the wing colour can be fixed rapidly. Genetic distances between ssp. yakushimanus inhabiting only the small islands (Yakushima and Tanegashima Islands) located to the south of Kyushu (one of the Japanese mainlands) and alcinous inhabiting Japanese mainlands, and between ssp. bradanus inhabiting the southern Ryukyu Islands and miyakoensis inhabiting the central Ryukyu Islands suggest that the wing colour changed in the period while one base substitution had been fixed in the ND5 gene of the population.
We examined many ssp. alcinous individuals from various regional populations in mainland Japan because there are regional differences in the ratio of pupal diapause induced by the short photoperiod and in the larval growth and egg size associated with the host-plant species used (Kato, 2000, 2001; unpublished observation). However, little or no difference in the ND5 sequences was found among those populations, suggesting that such morphological, physiological and ecological differences among the populations may have evolved in the short period while one or only a few base substitutions were being fixed in the ND5 gene of the regional populations.
Evolutionary rate of ND5 in P. (B.) alcinous
If the upgma tree of P. (B.) alcinous is correlated with the geological history of the Ryukyu Islands formation, it would be reasonable to hypothesize that the branching point between ssp. loochooanus and bradanus plus miyakoensis is located when the Ryukyu super-island and the Sakishima super-island formed. These super-islands formed by the end of the Pliocene (1.65 MYA), as a result of subsidence of the east periphery of the Eurasian Continent during the Pliocene (1.65–5.3 MYA) (Kizaki & Oshiro, 1977; Ujiié, 1990; Ota, 1998), and far later the Ryukyu and Sakishima super-islands separated to the present Okinawa and Amami Islands, and the present Yaeyama and Miyako Islands, respectively (Kizaki & Oshiro, 1977; Ota, 1998). From this assumption, Kimura's two-parameter distance 0.01D would be estimated to be about 1.8 million years. This approximate estimation can also reasonably explain that the divergence points of ssp. bradanus and miyakoensis, and of the Okinawajima and Amami–Oshima populations in the upgma tree are about 0.23 MYA, which approximately corresponds to the time the present Ryukyu Islands formed in the late Pleistocene (Kizaki & Oshiro, 1977; Ujiié, 1990; Ota, 1998). This ND5 evolutionary rate in P. (B.) alcinous is an intermediate rate estimated in Carabina ground beetles (0.01D = 3.6 million years) (Su et al., 1998; Osawa et al., 1999) and in Parnassius butterflies (0.01D = 0.75 million years) (Yagi et al., 2001).
Biogeography inferred from the upgma tree and geohistory
According to past geological studies on the Ryukyu Islands and the upgma tree, a hypothesis for the process of the establishment of the present P. (B.) alcinous distribution can be proposed. An ancestral P. (B.) alcinous diverged into two populations, probably in Unnan district, southern China, about 6 MYA. One population moved northeastwardly as it adapted to low temperatures, and finally reached the east periphery of the Eurasian Continent. A part of the population was isolated in the newly formed proto-Japan Archipelago about 3.2 MYA, and later spread back to the Korean Peninsula at the late glacial period (Pleistocene) when the Tsushima–Korea Channel became a land bridge (ssp. alcinous) (Ohshima, 1990; Matsui et al., 1998). The other part of the Eurasian east periphery population was isolated in the Ryukyu and Sakishima super-islands about 2 MYA when the Eurasian east periphery subsided and the proto-form of the Ryukyu Islands was formed (Kizaki & Oshiro, 1977; Ota, 1998). During the early Pleistocene, the Ryukyu Islands became a land bridgelike shape between mainland Japan and the continent. However, the super-islands were retained by three narrow gaps, the Sakishima, Kerama and Tokara straits (Kimura, 1996; Ota, 1998). The isolated super-island populations further diverged about 0.2 MYA when the super-islands separated to the present Ryukyu Islands (ssp. loochooanus, bradanus and miyakoensis) (Kizaki & Oshiro, 1977; Ota, 1998). The other Unnan population spread southeastwardly and some moved to Taiwan which had been a part of the continent until the late glacial period (ssp. mansonensis) (Kimura, 1996).
This hypothesis is rather narrative, although it is based on current palaeogeographical data and our objective molecular phylogeny. However, we believe that this hypothesis is the most reasonable at present. To make the hypothesis more confirmatory, more palaeogeographical studies and molecular phylogenetic analyses using other genes are required.
This work was partly supported by the Grant-in-Aid for scientific research (no. 13575013) from the Japan Society for Promotion of Science and by the Institute for Applied Optics. The authors are most grateful to Ms C. Shimohara for her technical assistance, and to the following scientists who supplied butterfly samples: K. Araya, A. Chichvarkhin, T. Dejima, H. Fujii, T. Fujioka, H. Fukuda, Y. Iemura, S. Ishigaki, R. Katase, K. Kikue, T. Nakata, S. Osada, G. Sasaki, T. Shinkawa, N. Suzuki, K. Takasaki, A. Yokokura and T. Yamamoto. The authors also thank Mr T. Inomata for discussion on scientific names for this species.