The Ogasawara (Bonin) Islands are oceanic islands of volcanic origin located in the northwestern Pacific Ocean about 1,000 km south of the Japanese mainland. A large carpenter bee, Xylocopa (Koptortosoma) ogasawarensis, is endemic to the islands but its closest relative is unknown. The Ogasawara Islands are geographically closest to the Japanese Archipelago, but this area is inhabited only by species of a different subgenus, Alloxylocopa. Thus, X. ogasawarensis is commonly thought to have originated from other members of Koptortosoma, which is widely distributed in the Oriental tropical region. In this study, we investigated the origin of X. ogasawarensis using a phylogenetic analysis of Xylocopa based on four genes: mitochondrial cytochrome oxidase subunit I (COI) and cytochrome b (Cyt b), and nuclear elongation factor-1α (EF-1α) and phosphoenolpyruvate carboxykinase (PEPCK). A combined analysis of the four genes strongly suggests that Koptortosoma is a large, polyphyletic group, within which Alloxylocopa is embedded. Xylocopa ogasawarensis emerged as the species most closely related to Alloxylocopa and not to Oriental species of Koptortosoma. Contrary to previous views of the origin of X. ogasawarensis, our results suggest that X. ogasawarensis and Alloxylocopa share a common origin and diverged after they colonized the island regions of East Asia.
Isolated oceanic islands that have never been connected to a continental landmass harbor a remarkable diversity of endemic organisms as a result of local evolution and speciation. The Ogasawara (Bonin) Islands are typical oceanic islands of volcanic origin located in the northwestern Pacific Ocean about 1,000 km south of the Japanese mainland and comprise four island chains from north to south: the Muko-jima Islands, the Chichi-jima Islands, the Haha-jima Islands, and the Kazan Islands. The insect fauna of the Ogasawara Islands consists of 1,281 species, of which 26.4% are endemic to this area (Ohbayashi et al., 2003). Nine native bee species occur in the Ogasawara Islands, all endemic (Ohbayashi et al., 2003).
A large carpenter bee, Xylocopa (Koptortosoma) ogasawarensis (Matsumura, 1912), is one of the endemic bee species in the Ogasawara Islands and occurs on Chichi-jima and Haha-jima. Xylocopa ogasawarensis exhibits conspicuous sexual dimorphism in pubescence color, as is typical of other members of the subgenus (Sugiura, 2008). Koptortosoma occurs from South Africa through the Middle East and India and into Australia and includes 196 species comprising >40% of the global diversity of Xylocopa (469 species; Michener, 2007). Leys et al. (2002) performed a phylogenetic analysis based on nuclear and mitochondrial DNA sequences that suggested that Koptortosoma is not a monophyletic group, and is roughly divided into Oriental-Australian and Ethiopian groups.
More than 50 species of Koptortosoma occur throughout the Oriental region (Hurd and Moure, 1963). However, the subgenus has not been recorded in the Japanese Archipelago, which is the closest landmass to the Ogasawara Islands and is inhabited instead by four species of the subgenus Alloxylocopa. Thus, until now, it has been proposed that the ancestor of X. ogasawarensis occurs in the Oriental tropical region where Koptortosoma species are abundant (Yasumatsu, 1955; Hirashima, 1981). However, this hypothesis has not been verified by using a phylogenetic approach. In this study, we investigated the phylogenetic relationships of X. ogasawarensis to infer its origin based on sequences of four genes: mitochondrial cytochrome oxidase subunit I (COI) and cytochrome b (Cyt b), and nuclear elongation factor-1α (EF-1α) and phosphoenolpyruvate carboxykinase (PEPCK).
MATERIALS AND METHODS
We used sequence data available from previous studies (Leys et al., 2000, 2002; Leijs and Hogendoorn, 2008) and added original sequences for X. ogasawarensis and six species from Japan and Laos (Table 1). Because higher subgeneric relationships within Xylocopa have been well established previously (Leys et al., 2002; Leijs and Hogendoorn, 2008), we focused our analysis on the well-defined monophyletic clade of ~10 Old World subgenera, including Koptortosoma and Alloxylocopa (Oriental+African clades in Leys et al., 2002; Table 1). Xylocopa (Xylocopoides) virginica (Linnaeus, 1771) and X. (Proxylocopa) olivieri Lepeletier, 1841, were used as outgroups to this clade. The subgeneric classification follows Michener (2007).
Sampling details of the specimens used in the present study. Newly sampled species are indicated in bold. Locality is unavailable for species treated in Leijs and Hogendoorn (2008). Sequence data are missing where GenBank accession numbers are not given.
We preserved a foreleg of each bee in 99% ethanol at –20°C prior to DNA extraction. Legs were ground with pestles in reaction tubes and incubated in extraction buffer containing 470 μl of CTAB, 30 μl of 10% SDS, and 15 μl of proteinase K (20 mg/ml) at 55°C for 20 h. After incubation, the DNA was extracted with a phenol-chlorophorm solution, precipitated with ethanol, and stored in 20 μl of distilled water. The COI and Cyt b regions were amplified by using the UEA7/UEA10 and cb1/cb2 primer pairs of Leys et al. (2000), respectively, and the PEPCK region was amplified by using the primers G146F and G147R of Leys et al. (2002). The F1 copy of the EF-1α region was amplified using the primers, 5’-AA[A/G] TA[C/T] GCC TGG GT[A/G] [C/T]T[G/T] GA[C/T] AAG CT-3’ (forward) and 5’-T [C/T]TT [A/C/G/T]GG CGA [C/G/T]GG [C/T]TC [C/G/T]A[A/G] CAT GT-3’ (reverse), which were newly designed for this study based on previously available Xylocopa EF-1α sequences.
PCR amplifications were performed in 20-μl reaction mixtures containing 0.4 pmol each primer, 0.2 mM each dNTP, 1X Ampdirect buffer (Shimadzu, Kyoto, Japan), 0.4 U ExTaq polymerase (TaKaRa, Otsu, Japan), and 0.5 μl unquantified DNA template solution. Reaction conditions were: 5 min at 94°C; 30 cycles of 94°C for 30 s, 45°C (COI, Cyt b, and EF-1α) or 50°C (PEPCK) for 30 s, and 72°C for 1 min; and a final extension step for 7 min at 72°C. PCR products were excised from agarose gels and purified by using a NucleoSpin Extract II Kit (Macherey-Nagel, Dьren, Germany). Sequencing was performed by using the PCR primers and a BigDye Terminator Cycle Sequencing Ready Reaction Kit (Perkin-Elmer, Foster City, CA) and electrophoresed on an ABI 3100 sequencer (Perkin-Elmer).
Because PEPCK sequences contained introns, they were aligned by using Clustal X (Thompson et al., 1997) and subsequently checked by eye for obvious misalignments. However, approximately 60% of the aligned PEPCK intron positions could not be aligned unambiguously and thus were excluded from the analysis. Gaps within PEPCK sequences were treated as missing data. The COI, Cyt b, and EF-1α sequences contained no introns, and thus the alignment was straightforward. Phylogenetic trees were obtained both together and separately for each gene by using the maximum parsimony (MP), maximum likelihood (ML), and Bayesian methods. MP analysis was performed by using PAUP* 4.0b10 (Swofford, 2002) and by conducting heuristic searches with 100 random addition analyses with equal character weights and tree-bisection-reconnection (TBR) branch swapping. Clade support for the MP analyses was assessed by bootstrap analysis of 1000 pseudoreplicates. Prior to the ML analysis, an appropriate model of base substitution and model parameters were obtained by using Modeltest 3.0 (Posada and Crandall, 1998). ML heuristic searches were conducted by using PAUP* with 10 random addition analyses and TBR branch swapping, and nodal support was assessed by bootstrap analysis with 1,000 replications in PhyML (Guindon and Gascuel, 2003). For the Bayesian analysis, we used MrBayes 3.1.2 (Ronquist and Huelsenbeck, 2003), with best-fit models inferred by Modeltest. In the combined analysis of the four genes, substitution parameters were estimated separately for each gene by using the ‘unlink’ command. Markov chain Monte Carlo simulations were run for two million generations, with trees sampled every 1000 generations for a total of 2001 trees, and posterior probabilities of topology, branch lengths, and parameter estimates were obtained after omitting the first 1001 trees as burn-in.
Analyses of individual partitions and the combined data set produced similar topologies; here we focus on the results of the combined analysis of the four genes. The combined four-gene data matrix was 1970 bp long, with 601 variable and 351 parsimony-informative positions. The MP analysis resulted in two equally parsimonious trees (tree length=1,485, consistency index [CI]=0.517, retention index [RI]=0.513), which are essentially similar to the ML phylogeny (-ln likelihood=10267.185; Fig. 1) and differed only regarding the relationships of closely related species. The Bayesian analyses also produced an overall similar topology. Consistent with previous studies (Leys et al., 2002; Leijs and Hogendoorn, 2008), the subgenus Koptortosoma was recovered as a polyphyletic group, within which Alloxylocopa and the remaining six subgenera are embedded. The Koptortosoma species were roughly divided into Afrotropical and Oriental-Australian species, as was previously suggested (Leys et al., 2002). The results strongly suggest that Xylocopa ogasawarensis is the sister group to the four Japanese Alloxylocopa species (Fig. 1).
The results of our phylogenetic analysis strongly suggest that Alloxylocopa is embedded within polyphyletic Koptortosoma, and that X. ogasawarensis is most closely related to Alloxylocopa, rather than to the Oriental Koptortosoma. Alloxylocopa includes the four sampled Japanese species together with the two (sub)species not represented in our data (X. a. appendiculata Smith, 1852, which is a subspecies of the Japanese X. a. circumvolans, and X. phalothorax Lepeletier, 1841). The two taxa not sampled both occur in continental China and Korea; thus, all Alloxylocopa species are known from the East Asian region that is geographically closest to the Ogasawara Islands. These data suggest that the common ancestor of X. ogasawarensis and Alloxylocopa most likely diverged after it had colonized the islands of East Asia (Fig. 2). It is possible that our sampling of Koptortosoma was not sufficient to identify the true sister species of X. ogasawarensis. Nevertheless, our results do not provide evidence for an independent colonization of the Japanese island region by phylogenetically distant Xylocopa lineages, as was previously suggested (Yasumatsu, 1955; Hirashima, 1981), but instead lend strong support for the close affinity of X. ogasawarensis to Alloxylocopa species distributed in the Japanese mainland and Ryukyu Archipelago.
A previous morphological examination of Xylocopa subgenera (Minckley, 1998) suggested that species of Alloxylocopa share overall similar morphologies with species of Koptortosoma, and this is reflected in the results of our phylogenetic analysis. The most significant morphological differences between the two groups are the sexual color dimorphism in Koptortosoma and enlarged eyes in Alloxylocopa males. However, these characters are tightly associated with male mating strategies and do not necessarily reflect phylogenetic relationships (Leijs and Hogendoorn, 2008). Other morphological characters, including the male genitalia, are otherwise not strikingly different among these two groups and related subgenera (Hurd and Moure, 1963; Minckley, 1998); thus, the non-monophyly of Koptortosoma and the sister-group relationship of Alloxylocopa and X. ogasawarensis do not necessarily require a complex explanation of morphological evolution.
Bees are known as poor crossers of oceanic barriers (Michener, 1979), but a few Xylocopa species have succeeded in colonizing oceanic islands (Hurd and Moure, 1963). The most likely mechanism by which Xylocopa bees colonize oceanic islands is by floating in drift logs of the trees in which these bees had constructed their nests. This is evidenced by overrepresentation of wood- or stem-nesting bees in the oceanic island bee fauna worldwide (Michener, 1979). The distance between the Ogasawara Islands and the closest continental landmass is about 1,000 km, which is comparable to cases of other oceanic islands inhabited by endemic Xylocopa species: the Galбpagos Islands, situated 1,000 km west of Ecuador, and Revillagigedo Islands, located 650 km off the western Mexican coast (Hurd and Moure, 1963). Thus, although anthropogenic dispersal cannot be ruled out entirely, X. ogasawarensis likely colonized the islands by itself after it had split from the shared ancestor of Alloxylocopa and X. ogasawarensis (Fig. 2).
Regarding other native bees of the Ogasawara Islands, Cronin (2004) studied the phylogenetic relationships of East Asian Ceratina species and found that the endemic Ceratina (Neoceratina) boninensis Yasumatsu, 1955 is most closely related to C. (Neoceratina) dentipes Friese, 1914, which occurs in the Ryukyu Archipelago. A morphological investigation of species of Megachilinae (Ikudome and Yamane, 1990) also suggested that they share similarities with species of the East Asian region. Taken together, our results provide additional evidence suggesting that the endemic bee fauna of the Ogasawara Islands has a close affinity to that of East Asia, especially the island regions of the Japanese mainland and the Ryukyu Archipelago.
The terrestrial fauna of the Ogasawara Islands is characterized by high endemism and morphological distinctiveness as compared to their continental relatives. As a result, some species have been classified in separate genera that are endemic to the Ogasawara Islands. However, recent molecular phylogenetic studies suggest that endemic animals of the Ogasawara Islands are not as genetically distinct as suspected from their morphological distinctiveness, and those that have been assigned generic rank probably do not merit such status. For example, the endemic dragonfly genus Boninagrion is nested within Ischnura, which is widely distributed in the East Asian region (Karube et al., 2004). Also, land snails of the endemic genus Mandarina are very closely related to Euhadra of the Japanese mainland (Chiba, 1999). Thus, future molecular phylogenetic studies may identify previously unsuspected origins for the endemic fauna of the Ogasawara Islands.
We thank Y. Okuyama and H. Matsuura for bee specimens, R. Leijs for kindly sharing specimen data, and Y. Kameda for comments on the manuscript. The sample of X. ogasawarensis were collected with the permission of the Agency for Cultual Affairs of Japan.