INTRODUCTION

Taiwan is a subtropical island located 160 km east of the continental China. Being endowed with high mountains over 3500 m, this large island accomodates diverse geographic and climatic environments and encompasses amphibian fauna consisting not only of the Oriental elements but also of the Palearctic ones as well (Zhao and Adler, 1993). The brown frogs of the genus Rana are supposed to represent the latter, but the origin of the three species occurring on Taiwan, R. longicrus Stejneger, 1898, R. sauteri Boulenger, 1909, and R. multidenticulata Chou and Lin, 1997 remains unclear. Of these, R. longicrus was previously often synonymized with the Japanese brown frog, R. japonica Günther, 1858 (Pope, 1931; Liu and Hu, 1961). Because these two species actually have a high degree of genetic compatibility (Kuramoto, 1974), R. longicrus is supposed to be closely related to R. japonica.

On the other hand, R. sauteri is noted for its unique breeding ecology and larval morphology, and is occasionally assigned to a subgenus or even a genus distinct from Rana by some authors (Fei et al., 1990; Dubois, 1992). Because frogs regarded as subspecies of R. sauteri or species related to R. sauteri are distributed from southeastern China to northern Indochina (Smith, 1921, 1924; Tian and Jian, 1986; Zhao and Adler, 1993), R. sauteri may represent a lineage distinct from Palearctic brown frogs, and its origin is of great zoogeographi-cal interests (Boulenger, 1909). Population systematics of R. sauteri (sensu lato) in Taiwan was recently revised, in which the frogs from the middle and eastern parts of the central mountain range was described as a distinct species R. multidenticulata, and thus R. sauteri (sensu stricto) was confined to populations in western lowlands (Chou and Lin, 1997b). These two species, however, have a wide intergra-dation zone along the western side of the central range, and are difficult to discriminate on the basis of adult morphology (Chou and Lin, 1997a, b). Therefore, in the following, these species are collectively referred as the R. sauteri complex.

In this paper, phylogenetic relationships among the Taiwanese brown frogs and six representatives of the Japanese brown frog species (R. pirica Matsui, 1991; R. ornativentris Werner, 1903; R. japonica, R. tsushimensis Stejneger, 1907; R. okinavana Boettger, 1895; and R. tagoi Okada, 1928) are inferred by comparing their nucleotide sequences of mitochon-drial DNA (mtDNA), so as to revise the hypothesis regarding the origin of those Taiwanese species.

MATERIALS AND METHODS

DNA was extracted from small amounts (less than 50 mg) of frozen liver, muscle, heart, or eggs from 47 individuals representing one species and one species complex of Taiwanese and six species of Japanese brown frogs, and an outgroup taxon, R. catesbeiana Shaw, 1802. Locality information for these samples are given in Appendix. The published sequence of Xenopus laevis (Daudin, 1802) (Dunon-Bluteau et al., 1985) was also incorporated into the analysis as that for another outgroup.

The methods used to extract, amplify, and sequence are as described previously (Tanaka et al., 1994, 1996). A part of the mtDNA cytochrome b gene was amplified using the polymerase chain reaction (PCR) and approximately 600 base-pairs (bp) of this gene were sequenced. Primers newly designated for amplification and sequencing are: L14731 (5′-GAAAAACTATCGTTGTTATTCAACTA-3′), L14850 (5′-TCTCATCCTGATGAAACTTTGGCTC-3′), H15502 (5′-GGATTAGCTGGTGTGAAATTGTCTGGG-3′), and H15584 (5′-CCTAGTTTATTAGGGATGGAGCGGAG-3′). Sequence numbering system followed that of the human sequence (Anderson et al., 1981).

Three different methods were employed to infer relationships among taxa on the basis of sequence data obtained; the neighbor-joining method (Saitou and Nei, 1987), using NJ procedure included in the PHYLIP package (Felsenstein, 1993), the maximum likelihood method using DNAML procedure of PHYLIP, and the parsimony method using the heuristic search algorithm in PAUP (Swofford, 1993) with sequences added in a random order for 1000 repetitions. When a taxon showed some variation in the nucleotide sequence, most frequently appearing sequence was considered as the representative sequence for the taxon.

RESULTS

Genetic differentiation among taxa was estimated using nucleotide sequence data of 587 bp that were constantly obtained for all samples (Fig. 1). Intrapopulation nucleotide sequence similarities of the Rana sauteri complex were 99.8%, and 99.7% for Sanyi and Wulai samples, respectively. Between samples from Alishan and Sanyi, similarities were fairly high, ranging from 95.7–95.9%. By contrast, samples from Wulai showed a substantial sequence divergence from the above two samples with only 83.7–84.7% similarities.

Fig. 1

Aligned sequences of a 587 bp segment of the cytochrome b gene. Dots indicate identity to the sequence of Xenopus laevis.

In the neighbor-joining tree rooted at the midpoint of the longest path (Fig. 2A), the outgroup taxa X. laevis and R. catesbeiana were separated from the remaining frogs in order. The cluster of the ingroup, supported in 99.8% bootstrap iterations, showed nearly polytomous relationships of four major clusters.

Fig. 2

A neighbor-joining tree rooted at the midpoint of the longest path (A) and a parsimony tree (B). Nodal values indicate percent support for branches in 1,000 bootstrap replicates (frequencies < 50% collapsed to polytomies in the latter tree).

The first cluster contained only R. tagoi, and the second cluster, supported by 100% iterations, consisted of two populations of the R. sauteri complex (Alishan and Sanyi). Sister relationship of the two individuals from Sanyi was supported by 83.4% iterations. The bootstrap value to support the monophyly of the first and second clusters was as low as 49.6%.

The third cluster, supported by 86.8% iterations, contained R. okinavana, R. tsushimensis and the R. sauteri complex from Wulai. Within this cluster, R. okinavana first split from a subcluster consisting of the remainder, which was supported by 100% iterations. This subcluster was trichotomous with sister relationship of the two individuals of the R. sauteri complex from Wulai being supported only in 42.0% bootstrap iterations. This result derives from surprisingly high genetic similarities between these two individuals and R. tsushimensis (98.6–99.5%).

The fourth cluster, supported by less iteration value (63.2% iterations), consisted of R. longicrus, R. japonica, and two species with 2n = 24 chromosomes (i.e., R. pirica and R. ornativentris). The sister relationship of R. longicrus and R. japonica was only weakly supported by 47.7% iterations, while that of the two species with 2n = 24 chromosomes was supported by no less than 97.9% iterations.

The maximum-likelihood analysis, using Xenopus and R. catesbeiana as outgroup taxa, produced a topology identical to that obtained by neighbor-joining method (data not shown). The tree contained several collapsing branches, and confirmed the polytomous relationships of the four clusters mentioned above.

In the maximum-parsimony analysis, the sister relationship of the R. sauteri complex from Alishan and Sanyi and R. tagoi was again not supported (iterations = 38.3%) as in the results of the above two analyses. Monophyly of R. japonica and R. longicrus, or that of these two species and the two species with 2n = 24 chromosomes was not supported, either (Fig. 2B). The bootstrap value to support sister relationship of R. longicrus and R. japonica was particularly low (42.7%).

Monophyletic relationships of (1) the R. sauteri complex from Wulai, R. tsushimensis, and R. okinavana, (2) the two species with 2n = 24 chromosomes (R. pirica and R. ornativentris), and (3) the R. sauteri complex from Sanyi and from Alishan were again strongly supported (72.8%, 90.0% and 100% of 1,000 bootstrap iterations, respectively).

In this way, the most important point here clarified is that there are two genetically highly diverged groups within the R. sauteri complex; one from Alishan and Sanyi that were distinct from all other brown frogs examined here, and the other from Wulai that was close to R. okinavana and R. tsushimensis.

DISCUSSION

Rana longicrus from Taiwan, together with some brown frogs widely spread in China (e.g., R. omeimontis Ye et Fei, 1993 and R. zhenhaiensis Ye, Fei, et Matsui, 1995), was once regarded as conspecific with Japanese R. japonica from their close morphological similarities. Recent karyological and ecological studies also suggested their close phylogenetic relationships (Kuramoto et al., 1973, 1984). However, the clade of R. longicrus and R. japonica was only weakly supported both in neighbor-joining (bootstrap iterations = 47.7%) and maximum-parsimony (42.7%) trees, and the two species were suggested to be well differentiated genetically notwithstanding their close morphological and ecological similarities. These two forms have been considered distinct at the species level from the result of artificial hybridization (Kuramoto, 1974), and the present result supports this account. On the basis of genetic compatibility, Kuramoto (1974) considered that R. okinavana from the Ryukyus is closer to geographically remote R. tsushimensis from Tsushima Islands than to R. longicrus from Taiwan. On the other hand, from the analyses of allozymes and proteins, Nishioka et al. (1992) estimated that R. tsushimensis diverged much earlier than the divergence of the remaining two species. The present result greatly differs that of Nishioka et al. (1992), but almost coincides with that of Kuramoto (1974). From these results, R. longicrus is considered to represent a Palearctic element in Taiwan.

Our result clarified that at least two genetically distinct lineages (Alishan and Sanyi populations versus Wulai population) are included in the R. sauteri complex. Intrapopulation sequence similarities of this complex (99.8% in Sanyi population and 99.7% in Wulai population) correspond to that of R. japonica, whereas the similarity values between Alishan and Sanyi populations (95.7-95.9%) are similar to those reported for local populations of R. tagoi from Japan (Tanaka et al., 1994). From the result of morphological analyses, Chou and Lin (1997a, b) recently divided R. sauteri from Taiwan into two species and described a new species R. multidenticulata. These two species are reported to have a wide intergradation zone where frogs of intermediate form occur. From the distributional map they provided, Sanyi is located in the northernmost area of the range occupied by R. sauteri with close proximity to the intergradation zone, and Alishan and Wulai, respectively, at the central and the northernmost areas of the range of R. multidenticulata (Fig. 3). Diagnostic characters Chou and Lin (1997b) provided to discriminate the two species, however, are not much useful for the identification of adult specimens. Those authors assigned the population from Wulai to R. multidenticulata on the basis of larval morphology, but they also noted that the Wulai population is intermediate between R. sauteri and R. multidenticulataⁱn adult morphology (Chou and Lin, 1997a). According to the present result, supposed R. sauteri from Sanyi and R. multidenticulata from Alishan are phylogenetically considerably distant from supposed R. multidenticulata from Wulai. This partially supports the view of Chou and Lin (1997a, b) in that it recognizes at least two distant lineages in the R. sauteri complex within Taiwan. However, the contents of the two lineages are quite dissimilar between our results and Chou and Lin's (1997a, b) arrangements. In order to clarify reasons for this discordance and evaluate validity of Chou and Lin's (1997a, b) classification, whole Taiwanese populations including those from type localities of R. sauteri (Kanshirei = Quantzelin or Kuantzulin) and R. multidenticulata (Dayuling) should be surveyed in detail.

Fig. 3

A map of Taiwan showing localities where samples of the R. sauteri complex used in the present study were collected (closed circle) and distributional range of the R. sauteri complex (after Chou and Lin, 1997b: I = R. sauteri, II = intermediate, III = R. multidenticulata). Type localities of R. sauteri (Quantzelin) and R. multidenticulata (Dayuling) are also included (open circle).

According to our result, sequence similarities of the R. sauteri complex from Wulai and R. tsushimensis, endemic to Tsushima Islands, were 98.6–99.5%. Because these values approach to those found in a population of R. tsushimensis (99.5–99.8%: Tanaka et al., 1996), the R. sauteri complex from Wulai is considered to be phylogenetically very close to R. tsushimensis. Further, these two forms are undoubtedly close to R. okinavana phylogenetically. Present distributional ranges of R. tsushimensis and R. okinavana are both limited to a few small islands of Japan distant from each other. From its great genetic distances from the others, R. okinavana is supposed to have diverged very early among these three forms. Tanaka et al. (1996) estimated that the ancestral stock of R. okinavana might have been isolated on the Amami and Okinawa Islands of the Central Ryukyus by the time when the Japanese main islands and the Ryukyu Archipelago were separated by the sea (Pliocene, 2–8 my B.P.: Kizaki and Oshiro, 1980). The stock one leading to R. okinavana and the other to the common ancestor of the remaining two forms are supposed to have diverged even earlier.

On the other hand, R. tsushimensis and the R. sauteri complex from Wulai are genetically very close and their divergence is supposed to be a rather recent event. Tsushima Islands are currently more than 1000 km distant from Taiwan, but because the East China Sea separating these islands is shallow (< 200 m deep), these islands may have been connected by land bridges through the continent during the Pliocene (1–2 my B. P) and middle or late Pleistocene (0.02–0.4 my B.P.: Kimura, 1996). If this is actually the case, the ancestral stock common to the two forms would have had a wide range of distribution until rather recently. Through the subsequent sea expansion, however, the stock should have been confined to the regions corresponding to the present Tsushima Islands and Taiwan, where it has been differentiated into R. tsushimensis and the Wulai lineage of the R. sauteri complex, respectively.

The clade represented by the R. sauteri complex from Alishan and Sanyi is fairly differentiated genetically among the brown frogs examined here, and this lineage is supposed to have diverged at a very early stage of divergence of the East Asian brown frogs. Although the R. sauteri complex and R. tagoiare dissimilar in larval morphology and ecology, they share unique breeding ecology and egg morphology that are not found in other brown frog species (Kuramoto et al., 1984). Like the R. sauteri complex from Alishan and Sanyi, R. tagoi is also supposed to have diverged fairly early from other brown frogs. Each of these lines of divergence might have related to invasions into new breeding environments that were different from those utilized by most other brown frogs. Notwithstanding their genetically remote relationships, populations of the R. sauteri complex from Alishan and Sanyi are not easily differentiated from the population from Wulai in adult morphology. They are almost identical in the mode of larval life as well. If the present result reflects their true phylogeny, these morphological and ecological similarities are most likely the consequence of convergent evolution. Further, both of these two lineages of the R. sauteri complex are considered to represent not Oriental but Palearctic elements of the Taiwanese fauna like R. longicrus (see above).

According to the classification of Dubois (1992), which was made mainly on the morphological ground, R. japonica and R. longicrus are grouped into the R. japonica group. Our result, however, only weakly supported the monophyly of these two species. Because Dubois' (1992) R. japonica group also includes brown frogs from Sakhalin and the eastern Eurasian continent (Russia, Mongolia, China) as represented by R. amurensis, further studies including these forms are necessary to evaluate his classification on the more comprehensive phylogenetic ground.

Fei et al. (1990) split R. sauteri and its allies from other brown frogs as a distinct genus Pseudorana on the basis of their unique larval morphology. Dubois (1992), while largely following this idea, relegated Pseudorana to a subgenus of the genus Rana. Similar idea to regard R. sauteri (sensu lato) as representing a remote relationship from other brown frogs can be traced back to the original description of this species by Boulenger (1909) who suggested R. sauterito bridge brown frogs and hylaranine frogs. However, our results indicate that R. sauteri and other brown frogs actually constitute a mono-phyletic group [strongly supported in both neighbor-joining (99.8%) and maximum-parsimony (99.1 %) trees], and that the two different lineages of the R. sauteri complex have closer genetic relationships with different Japanese brown frogs. From such inferred phylogeny, it seems at present better not to assign the R. sauteri complex to a distinct genus or even subgenus but to place it in the subgenus Rana following the traditional taxonomy (e.g., Frost, 1985). Fei et al.'s (1990) genus or Dubois' (1992) subgenus Pseudorana includes several brown frogs from China and Indochina that are regarded as subspecies or relatives of R. sauteri (Smith, 1921; Tian and Jian, 1986; Fei et al., 1990; Zhao and Adler, 1993). In order to evaluate the validity of Pseudorana, further studies of these frogs are inevitable.