www.elsevier.com/locate/ympev
1055-7903/$ - see front matter © 2006 Elsevier Inc. All rights reserved. doi:10.1016/j.ympev.2006.06.027
Phylogeographic study reveals the origin and evolutionary
history of a Rhododendron species complex in Taiwan
Jeng-Der Chung
a, Tsan-Piao Lin
b, Yu-Ling Chen
c, Yu-Pin Cheng
d, Shih-Ying Hwang
c,¤a Division of Silviculture, Taiwan Forestry Research Institute, 53 Nanhai Road, Taipei 10066, Taiwan b Institute of Plant Biology, National Taiwan University, 1 Roosevelt Road, Section 4, Taipei 10617, Taiwan c Graduate Institute of Biotechnology, Chinese Culture University, 55 Hwagang Rd., Yangmingshan, Taipei 11114, Taiwan
d Division of Forest Biology, Taiwan Forestry Research Institute, 53 Nanhai Road, Taipei 10066, Taiwan Received 17 October 2005; revised 28 May 2006; accepted 14 June 2006
Available online 16 September 2006
Abstract
This study infers a single origin and a once-widespread distribution of the Rhododendron pseudochrysanthum species complex in Tai-wan based on chloroplast DNA (cpDNA) variation. In total, 124 individuals from Wve endemic Rhododendron species were used for ampliWcations of two chloroplast intergenic spacers: trnL-trnF and atpB-rbcL. The haplotype and nucleotide diversities were much lower for the R. pseudochrysanthum complex, comprised of the species R. pseudochrysanthum, R. morii, Rhododendron rubropunctatum, and
Rho-dodendron hyperythrum, than for RhoRho-dodendron formosanum. Two measures of pairwise population diVerentiation, NST and FST, consis-tently revealed mostly non-signiWcant levels of genetic divergence between populations of the R. pseudochrysanthum complex. No genetic diVerence was found among the four species of the R. pseudochrysanthum complex by analysis of molecular variance (AMOVA), which is concordant with the parsimonious topology of cpDNA haplotypes for the complex. Nested clade analysis (NCA) of the cpDNA haplo-types indicated that restricted gene Xow with isolation-by-distance characterized the recolonization pattern of the R. pseudochrysanthum complex. In contrast, the NCA analysis indicated a contiguous range expansion for cpDNA haplotypes of R. formosanum. This research suggests a once-widespread distribution of the R. pseudochrysanthum complex probably via north-to-south colonization of mid-eleva-tions during low-temperature periods of the Pleistocene. Population fragmentation followed the warmer climate which began in the Holo-cene and resulted in the present-day range contraction into high elevations.
© 2006 Elsevier Inc. All rights reserved.
Keywords: cpDNA; Evolutionary origin; Phylogeography; Rhododendron; Species complex; Taiwan
1. Introduction
The genus Rhododendron includes widely-distributed Xowering plants found throughout the world with the exception of Africa and South America, and containing over 1000 species (Chamberlain et al., 1996). In Taiwan, Rhododendron species are found from the lowlands to 3950 m in elevation. An investigation of eight Taiwanese Rhododendron species using chloroplast trnL-trnF inter-genic DNA sequences revealed two major clusters which
distinguished species with chartaceous or subcoriaceous leaves from species with thick coriaceous leaves (Hwang and Hsu, 2001). This classiWcation was supported by a phy-logenetic study of Taiwanese Rhododendron species using nuclear ITS (internal transcribed spacer) sequences (Tsai et al., 2003). Five Rhododendron species of Taiwan with thick coriaceous leaves are Rhododendron pseudochrysant-hum Hay., R. morii Hay., Rhododendron rubropunctatum Hay., Rhododendron hyperythrum Hay., and Rhododendron formosanum Hemsl. Both studies (Hwang and Hsu, 2001; Tsai et al., 2003) indicated derivation of R. pseudochrysant-hum, R. morii, R. rubropunctatum, and R. hyperythrum from R. formosanum. These Wve species are all members of the subgenus Hymenanthes. Hymenanthes is the most-abundant
* Corresponding author. Fax: +886 2 2861 8266.
subgenus in Rhododendron, and contains 225 species (Chamberlain, 1982). Of the Wve Rhododendron species within Taiwan which belong to the subgenus Hymenanthes, three are classiWed in the subsection Maculifera (R. pseudo-chrysanthum, R. morii, and R. rubropunctatum) and they are closely related, as no clear resolution was revealed by molecular markers including chloroplast trnL-trnF and nuclear ITS DNA sequences (Hwang and Hsu, 2001; Tsai et al., 2003). Morphologically, both R. rubropunctatum and R. morii are regarded by Taiwanese taxonomists as being derived from R. pseudochrysanthum although there is no solid evidence, and thus these three species were classiWed as a single species (R. pseudochrysanthum) by Lu and Yang (1989) and Li et al., (1998).
Rhododendron hyperythrum is the only member of the subgenus Hymenanthes subsection Pontica that occurs in Southeast Asia (including Taiwan) (Milne, 2004). The trnL-trnF sequences of R. hyperythrum contain no diVerences with those of R. pseudochrysanthum, R. morii, and R. rubro-punctatum (Hwang and Hsu, 2001). Zero genetic distance was found when comparing R. hyperythrum with R. morii and R. pseudochrysanthum and only a 0.4% genetic distance (Kimura’s two-parameter distance) was found between R. hyperythrum and R. rubropunctatum (Tsai et al., 2003). In general, the Wndings in those studies are generally in agree-ment with the morphological observations that R. pseudo-chrysanthum, R. morii, R. rubropunctatum, and R. hyperythrum should be combined into a single species (Lu and Yang, 1989). Furthermore, R. hyperythrum, the only member of the subsection Pontica in Southeast Asia, formed a well-supported clade with other non-Pontica spe-cies from Southeast Asia and was proposed not to be a member of the subsection Pontica and was therefore sug-gested to be placed in a subsection of its own (Milne, 2004). However, according to the studies of Hwang and Hsu (2001) and Tsai et al. (2003), R. hyperythrum is probably best placed in the subsection Maculifera together with spe-cies of the R. pseudochrysanthum complex instead of being placed in a subsection of its own, because Milne (2004) included no other Rhododendron species belonging to the subsection Maculifera that occur in Taiwan in his study.
Species of the R. pseudochrysanthum complex are all endemic to Taiwan. Rhododendron morii and R. pseudo-chrysanthum are found only on high peaks and are sympat-rically distributed in some areas in central and southern Taiwan found at 3000–3950 m (Li et al., 1998). Moreover, Xowering times of these two species overlap. The Xowering times are from March to May for R. morii, and from April to June for R. pseudochrysanthum. In contrast, R. rubro-punctatum is only distributed at the northern tip of the island at elevations of 600–1200 m; only a few remnant populations remain in the wild; and it has been designated an endangered species.
Phylogenetic analysis of intraspeciWc molecular data in relation to geographic and ecological information is known as phylogeography (Avise et al., 1987). The study of phylo-geography has become the standard for deciphering the
genetic structure of extant populations by providing infor-mation for inferring contemporary biogeographic patterns of genetic variation (Avise, 2000). Dissecting species’ evolu-tionary histories requires analysis of DNA sequences in a phylogeographic context, which may help in deciphering the patterns of geographic distributions of evolutionary lin-eages and the geographic processes that have inXuenced those distributions (Avise, 2000). Understanding the tem-poral and spatial components of a species complex’s distri-butions can be used to explain ecological and phenotypic discontinuities observed across the extant range of a species complex. Species complexes may occur as lineages that enter previously unoccupied ecologically adaptive zones. The evolutionary history of the region undoubtedly would have had signiWcant eVects on the structuring of genetic diversity and is essential for understanding the patterns and processes that have sculpted evolutionary diversity and endemicity. Phylogenetic studies using cpDNA have con-Wrmed the evolutionary distinctiveness of evolutionary lin-eages of species (e.g., Downie et al., 2000; Wu et al., 2006). These evolutionary lineages often warrant separate conser-vation management strategies (Soltis and Gitzendanner, 1999) and should be treated as conservation management or evolutionarily signiWcant units.
In this study, we used a phylogeographical approach to better understand the evolutionary history and to trace the origin of the Rhododendron species complex that includes R. pseudochrysanthum, R. morii, R. rubropunctatum, and R. hyperythrum. Herein, we document the genetic variability and genetic diVerentiation of populations of the R. pseudo-chrysanthum complex together with populations of R. for-mosanum as an outgroup taxon with a hypothesis of low genetic variation caused by population fragmentation and a low level of population diVerentiation due to the once-widespread, panmictic population distribution of the R. pseudochrysanthum complex.
2. Materials and methods
2.1. Plant materials and DNA puriWcation
In total, 98 individuals of the R. pseudochrysanthum spe-cies complex including 30 from two populations of R. rubropunctatum, 28 from two populations of R. morii, 29 from two populations of R. pseudochrysanthum, and 11 individuals from one population of R. hyperythrum, as well as 26 individuals from two populations of R. formosanum were included in the investigation. The collection sites and related geographic information are depicted in Fig. 1 and listed in Table 1. Total DNA was extracted from ground-up leaf powder according to a modiWed hexadecetyltrimethyl ammonium bromide (CTAB) procedure (Doyle and Doyle, 1987). DNA was precipitated with ethanol, and after wash-ing with 70% ethanol, was dissolved in 200L TE buVer (pH 8.0) and stored at ¡20 °C. The DNA concentration was determined for each sample using the GeneQuant II RNA/DNA Calculator (AmershamBiosciences).
2.2. Primers and PCR ampliWcation
Polymerase chain reactions (PCRs) and DNA sequencing were performed with universal primers for trnL-trnF (5⬘-GGTTCAAGTCCCTCTATCCC-3⬘ and ATTTGAACTG GTGACACGAG-3⬘; Taberlet et al., 1991) and atpB-rbcL a 5⬘-CRGGTTGAGGAGTTACTCG-3⬘ and 5⬘-GAC-CRGAAGTAGTAGGATT-3⬘, this study). AmpliWcations were performed in a DNA Programmable Thermal Cycler (PTC-100, MJ Research) with initial denaturation at 94 °C for 3 min followed by 42 cycles of 1 min at 94 °C, 1 min of anneal-ing at 52 and 50 °C, respectively for trnF-trnL and atpB-rbcL, 90 s at 72 °C, and a subsequent 10-min Wnal extension at 72°C. The PCR mixture (25L) contained 500mM KCl, 15mM MgCl2, 0.01% gelatin, 100 mM Tris–HCl (pH 8.3), 1 mM dNTPs, 2M primer, 20ng template DNA, 1g RNase, and
0.5 U Taq polymerase (AmershamBiosciences). The PCR products were puriWed using a QiaGen puriWcation kit and then sequenced in both directions using a Taq Dye Dideoxy Terminator Cycle Sequencing Kit (Applied Biosystems) and a model ABI373A automated sequencer (Applied Biosystems). All sequence polymorphisms were visually rechecked from the chromatograms. Repeated sequencing was performed for those identiWed as singletons. All sequences were deposited in the EMBL nucleotide sequence database under the following accession numbers: AM085441–AM085446 for trnL-trnF and AM085447–AM085461 for atpB-rbcL.
2.3. Phylogenetic analyses
Individual sets of data from trnL-trnF and atpB-rbcL sequences from 124 Rhododendron individuals were aligned
Fig. 1. Map of Taiwan showing the sampling localities of Rhododendron populations included in this study. Filled circles indicate the location of popula-tions sampled from Wve Rhododendron species (identiWed by the population name followed by the species abbreviation). Haplotype frequencies of each population are shown in the pie charts next to the population label. The numbers accompanying the square symbols indicate haplotype numbers corre-sponding to those in Table 2. RR, Rhododendron rubropunctatum; RP, Rhododendron pseudochrysanthum; RM, Rhododendron morii; RH, Rhododendron
hyperythrum; RF, Rhododendron formosanum.
Table 1
Species and population names, locality, elevation, sample size, number of chloroplast haplotypes, chloroplast haplotype diversity (h), nucleotide diversity () accompanied by neutrality test statistics for nine populations of Wve Rhododendron species
Species and population Latitude, Longitude Elevation (m) No. of samples No. of haplotypes h Tajima’s D R2 R. pseudochrysanthum complex 98 8 0.281 § 0.059 0.00036 § 0.00009 ¡1.60900¤ 0.0382 R. pseudochrysanthum 29 3 0.310 § 0.104 0.00029 § 0.00010 ¡0.74835 0.0813 Hohuanshan 24°10⬘N, 121°20⬘E 3400 14 2 0.363 § 0.130 0.00033 § 0.00012 0.32440 0.1813 Lulinshan 23°27⬘N, 120°52⬘E 2862 15 2 0.248 § 0.131 0.00022 § 0.00012 ¡0.39883 0.1238 R. morii 28 4 0.378 § 0.110 0.00052 § 0.00016 ¡0.61294 0.1041 Hohuanshan 24°07⬘N, 121°15⬘E 2800 13 4 0.679 § 0.112 0.00095 § 0.00019 0.27650 0.1740 Alishan 23°30⬘N, 120°48⬘E 2100 15 1 0.000 § 0.000 0.00000 § 0.00000 — — R. rubropunctatum 30 4 0.251 § 0.102 0.00035 § 0.00016 ¡1.73178¤ 0.1017 Tsankuanliao 25°05⬘N, 121°51⬘E 738 15 3 0.362 § 0.145 0.00047 § 0.00022 ¡1.49051 0.1713 Tsaigongkeng 25°11⬘N, 121°31⬘E 886 15 2 0.133 § 0.112 0.00024 § 0.00020 ¡1.15945 0.2494 R. hyperythrum 11 1 0.000 § 0.000 0.00000 § 0.00000 — — Nanhutashan 24°21⬘N, 121°26⬘E 3500 11 1 0.000 § 0.000 0.00000 § 0.00000 — — R. formosanum 26 7 0.708 § 0.067 0.00114 § 0.00027 ¡1.84201¤ 0.0997 Chilanshan 24°36⬘N, 121°29⬘E 1600 13 3 0.295 § 0.156 0.00056 § 0.00035 ¡1.65231¤ 0.2053 Sanlinxi 23°38⬘N, 120°47⬘E 1750 13 6 0.718 § 0.128 0.00120 § 0.00043 ¡1.47092 0.1466
using the program CLUSTAL X (Thompson et al., 1997). The homogeneity of the phylogenetic signals across two chloroplast intergenic data partitions was analyzed using the incongruence length diVerence (ILD) test (Farris et al., 1994). The ILD test was performed with parsimoniously informative characters, using 1000 replications of heuristic searches with 100 random addition analyses and TBR branch swapping, using Steepest Descent, with the MUL-TREES option enabled. The topologies of the two data sets were congruent, and so the sequences of the two sets of data were combined. A maximum parsimony (MP) analysis using the combined sequences of both chloroplast inter-genic spacers was conducted using PAUP¤4.0 (SwoVord,
2001). Heuristic searches with 100 random entries were per-formed using the ACCTRAN, MULPARS, and TBR options in PAUP¤. Gaps were treated as missing data, and all characters were accorded equal weight. To assess the conWdence in the branching patterns, bootstrap analyses (Felsenstein, 1985) were performed with 1000 pseudorepli-cates. The consistency index (CI; Kluge and Farris, 1969) and the retention index (RI; Farris, 1989) were also com-puted by PAUP¤.
2.4. Nucleotide diversity, haplotype diversity, and population diVerentiation
The number of haplotypes was counted with all poly-morphic sites, excluding polyA and polyT sites but includ-ing other indels, for all samples and samples within each locality. Haplotype diversity (h), nucleotide diversity () (Nei, 1987), and Tajima’s D (Tajima, 1989a) test for depar-ture from neutrality on the total number of segregating sites were calculated using the computer program Arlequin version 2.0 (Schneider et al., 2000). One measure of popula-tion diVerentiation, NST, which is inXuenced by both haplo-type frequencies and distances between haplohaplo-types (Pons and Petit, 1996) was calculated using the DnaSP program version 4.0 (Rozas et al., 2003). Each indel was recoded as a transitional substitution for the DnaSP analysis. FST was calculated with the computer program Arlequin (Schneider et al., 2000). Ramos-Onsins and Rozas (2002) suggested that the R2 statistic has greater power for detecting popula-tion growth with small sample sizes than many other esti-mators, so the R2 statistic was estimated using the DnaSP program. Analysis of molecular variance (AMOVA) with statistical signiWcance was determined using permutation analyses (ExcoYer et al., 1992) to partition the genetic vari-ation into diVerent levels. SigniWcance was determined by 1000 permutations. The AMOVA analysis was also per-formed using the Arlequin program.
2.5. Nested clade phylogeographical analysis
The network of chloroplast haplotypes was recon-structed using the algorithm of statistical parsimony described by Templeton et al. (1992) and implemented in TCS v1.06 (Clement et al., 2000). We used the nested clade
analysis (NCA) to infer the population history of the inves-tigated Rhododendron species. The NCA nesting design was constructed by hand on the haplotype network following the rules given in Templeton et al. (1987) and Templeton and Sing (1993). The program GeoDis 2.2 (Posada et al., 2000) was used to calculate the various NCA distance mea-sures and their statistical signiWcance levels. All statistical analyses in GeoDis were performed using 1000 permuta-tions. Results obtained from GeoDis were then interpreted using the revised inference key of Templeton (2004). The statistics calculated for all clades were (i) the clade distance (DC), which measures the average distance of all clade members from the geographical center of distribution, (ii) the nested clade distance (DN), which measures how wide-spread a particular clade is relative to the distribution of other clades in the same nested group, and (iii) the interior-tip distances (I-TDC and I-TDN). This interior vs. tip contrast of clades corresponds to clades younger (tip clade) relative to their ancestors clades (interior clades), and common vs. rare ones under expectations from neutral coa-lescent theory (Crandall and Templeton, 1993). Testing for signiWcantly small or large DC or DN distances in each nested clade was then used to reject the null distribution of no association between haplotype distributions and geography.
3. Results
3.1. Sequence variation and haplotype diversity
The unambiguous sequenced length of the ampliWed atpB-rbcL segment was »709 bp for the R. pseudochrysant-hum species complex, while the length of atpB-rbcL was slightly longer in R. formosanum with a major insertion of 21 bp in the aligned atpB-rbcL sequences. In contrast, the ampliWed trnL-trnF length was 391 » 392 bp for the R. pseudochrysanthum complex and R. formosanum.
In total, 124 individuals from the Wve Rhododendron spe-cies were used for PCR ampliWcation of the two chloroplast intergenic spacers, trnL-trnF and atpB-rbcL. For the aligned trnL-trnF sequences eight characters were phyloge-netically informative. For the aligned length of atpB-rbcL 10 were phylogenetically informative. The incongruence length test performed using PAUP¤ resulted in a congruent relationship (p D 1.000) between the two data sets; thus we used the combined sequences for the genealogical analysis. The combined sequences of trnL-trnF and atpB-rbcL resulted in a total of 1127 aligned characters. We then removed the poly A and poly T sites from the atpB-rbcL sequences due to the possible sequencing errors provoked by Taq polymerase stuttering; this resulted in an aligned length of 1105 bp that was used for the subsequent analyses. Fifteen unique haplotypes were detected among the 124 individuals of the Wve Rhododendron species examined. These haplotypes were deWned by 46 variable sites along the combined cpDNA regions, including two parsimoni-ously informative indels. For simplicity, each indel was
counted as a single mutation site. For R. formosanum, seven distinct haplotypes were detected among the 26 individuals, with an overall haplotype diversity of 0.708 § 0.067 and a value of nucleotide diversity () of 0.00114 § 0.00027 (Table 1). Because sequences of both cpDNA regions of R. hypery-thrum were the same with the most-frequently observed sequences found in other members of the R. pseudochry-santhum complex, we regarded R. hyperythrum as a species in the R. pseudochrysanthum complex for these analyses. In this species complex, eight distinct haplotypes were detected among the 98 individuals examined, with an over-all haplotype diversity of 0.281 § 0.059 and a value of nucleotide diversity () of 0.00036 § 0.00009 (Table 1). For individual species of the R. pseudochrysanthum species complex, R. morii had the highest levels of haplotype and nucleotide diversities; while no haplotype or nucleotide diversities was observed for the 11 R. hyperythrum individ-uals examined (Table 1).
Haplotype 1 (ancestral haplotype) was the most-fre-quent haplotype in all populations of the R. pseudochry-santhum species complex. Sharing of ancestral cpDNA haplotype was commonly seen for species in the R. pseudo-chrysanthum complex (Table 2, Figs. 1, 2). Haplotype 7 was the only one other than the ancestral haplotype that occurred in more than one individual species of the R. pseudochrysanthum complex (R. pseudochrysanthum and R. morii). Haplotypes 2 through 6 as well as haplotype 8 were found in only one individual population of the
R. pseudochrysanthum complex (Table 2, Fig. 1). No haplo-type sharing was found between R. formosanum and the R. pseudochrysanthum complex. However, haplotype 15 (R. formosanum) was nine mutational steps away from
Fig. 2. Phylogenetic parsimony tree for chloroplast DNA haplotypes detected in nine populations of Wve Rhododendron species. All characters were equally weighted, and gaps were treated as missing data. (L D 22, CI D 1.0, RI D 1.0). Parsimony bootstrap values greater than 50% are shown above the branches.
Table 2
Absolute frequency of chloroplast DNA haplotypes from nine popula-tions of Wve Rhododendron species
RR, Rhododendron rubropunctatum; RP, Rhododendron
pseudochrysant-hum; RM, Rubropunctatum morii; RH, Rhododendron hyperythrum; RF, Rhododendron formosanum. HH, Hohuanshan; LL, Lulinshan; AL,
Alishan; TK, Tsankuanliao; TG, Tsaigongkeng; CL, Chilanshan; SL, Sanlinxi. Haplotype RP RM RR RH RF HH LL HH AL TK TG NH CL SL 1 11 13 7 15 12 14 11 2 1 3 1 4 2 5 1 6 3 7 3 2 8 2 9 1 10 2 11 1 7 12 11 1 13 1 14 1 15 1 N 14 15 13 15 15 15 11 13 13
haplotype 1 (R. pseudochrysanthum complex). Only two (haplotypes 11 and 12) of the seven cpDNA haplotypes were found to occur in both R. formosanum populations examined.
3.2. Phylogenetic relationships of cpDNA haplotypes Phylogenetic relationships of the 15 cpDNA haplotypes for the R. pseudochrysanthum complex and R. formosanum are depicted in Fig. 2. MP analysis using the combined cpDNA data resulted in two most-parsimonious trees with a length of 22 steps (CI D 1.0, RI D 1.0). The topology of the parsimony analysis supported two major clades of haplo-types corresponding to divisions between R. formosanum and the R. pseudochrysanthum complex. However, no geo-graphic clustering of haplotypes was observed for either clade from the network. Clade A contained only R. formo-sanum haplotypes, while clade B contained only haplotypes of the R. pseudochrysanthum species complex.
3.3. Population diVerentiation
Pairwise NST values ranged from 0.0000 to 0.99149 and averaged 0.45119 or 0.09760 with populations of R. formo-sanum, respectively included in, or excluded from, the anal-ysis (Table 3). Pairwise FST values ranged from 0.0000 to 0.86256 and averaged 0.31146 or 0.08389 with populations of R. formosanum, respectively included in, or excluded from, the analysis. The test for population diVerentiation, according to an estimate of NST that considered not only haplotype frequencies but also the genetic distances between haplotypes, indicated that there were few signi W-cant diVerences within the R. pseudochrysanthum complex including populations of R. hyperythrum. On the other hand, when only haplotype frequencies were considered (FST), signiWcant diVerences were observed when the Hohu-anshan population of R. morii was involved in the analysis. Moreover, both pairwise NST and FST values displayed highly signiWcant diVerences when the populations of R. formosanum were included in the estimation (Table 3). The most-striking result observed from the pairwise NST
estima-tion was that the comparison involving R. hyperythrum with the Alishan population of R. morii or with the Tsaigongkeng population of R. rubropunctatum resulted in zero genetic diVerentiation. These results were also sup-ported by the pairwise FST analysis that showed only a very low level of genetic diVerentiation when the Tsaigongkeng population of R. rubropunctatum was compared to the R. hyperythrum population. Similar high levels of genetic divergence were found that highly signiWcantly diVered when comparing either population of R. formosanum with the populations of the R. pseudochrysanthum complex (Table 3). The comparison between the two populations of R. formosanum also resulted in signiWcant diVerences in genetic diVerentiation according to both the NST and FST estimates.
We also calculated the mean NST and FST values for each population of the four species in the R. pseudochrysanthum complex examined in comparison with the remaining popu-lations, and results showed similar patterns for these two estimates (Fig. 3). The R. morii populations of Hohuanshan had relatively higher mean NST and FST values when com-pared to all other populations in the R. pseudochrysanthum complex.
An analysis of molecular variance (AMOVA) for geo-graphic partitioning of cpDNA diversity detected that as much as 58% of the nucleotide variation was apportioned between R. formosanum and the R. pseudochrysanthum complex (Table 4). Essentially no genetic diVerentiation was detected among the four species of the R. pseudochry-santhum complex. Interestingly, 55% of the genetic varia-tion was apporvaria-tioned among populavaria-tions of individual species of the R. pseudochrysanthum complex, and the resampling statistics indicated signiWcant diVerences (Table 4).
3.4. Tests of neutral evolution and nested clade analysis Tests for population growth using Tajima’s D revealed no strong violations of the assumption of selective neutrality, as indicated by the non-signiWcant values of Tajima’s D (Table 1). However, the combination of the two populations of
Table 3
Pairwise NST values (below diagonal) and FST (above diagonal) for populations of Wve Rhododendron species for cpDNA data
¤ p < 0.05. ¤¤ p < 0.01. ¤¤¤ p < 0.001.
R. pseudochrysanthum R. morii R. rubropunctatum R. hyperythrum R. formosanum
Hohuanshan Lulinshan Hohuanshan Alishan Tsankuanliao Tsaigongkeng Nanhutashan Chilanshan Sanlinxi Hohuanshan (RP) 0.00000 0.04478 0.04419 0.16244 0.02466 0.07319 0.12500 0.67027¤ 0.46355¤ Lulinshan (RP) 0.12223 0.00000 0.13888¤ 0.07143 0.00621 0.00332 0.04241 0.73015¤ 0.52795¤ Hohuanshan (RM) 0.04691 0.21340¤ 0.00000 0.28566¤ 0.08990¤ 0.19597¤ 0.23967¤ 0.51282¤ 0.30128¤ Alishan (RM) 0.15385 0.07143 0.24074¤ 0.00000 0.09524 0.00000 0.00000 0.86252¤ 0.66056¤ Tsankuanliao (RR) 0.08835 0.04762 0.18373 0.03571 0.00000 0.02256 0.06418 0.66977¤ 0.46748¤ Tsaigongkeng (RR) 0.09484 0.03571 0.19182 0.00000 0.02159 0.00000 0.02212 0.79099¤ 0.58888¤ Nanhutashan (RH) 0.15385 0.07143 0.24074¤ 0.00000 0.03571 0.00000 0.00000 0.84095¤ 0.61883¤ Chilanshan (RF) 0.98655¤¤¤ 0.98811¤¤¤ 0.97738¤¤¤ 0.99149¤¤¤ 0.98443¤¤¤ 0.98782¤¤¤ 0.99149¤¤¤ 0.00000 0.43322¤ Sanlinxi (RF) 0.97621¤¤¤ 0.97778¤¤¤ 0.96692¤¤ 0.98122¤¤¤ 0.97400¤¤¤ 0.97747¤¤¤ 0.98122¤¤¤ 0.37083¤¤¤ 0.00000
R. rubropunctatum resulted in a signiWcant value of Tajima’s D together with a low value of R2 which indicated range expansion (Tajima, 1989a; Ramos-Onsins and Rozas, 2002). Results of the test for range expansion were also found for R.
formosanum according to the signiWcant Tajima’s D value and the low R2 value. A signiWcant negative value for Tajima’s D and a low R2 value were also found when all samples of the R. pseudochrysanthum complex were analyzed together.
The nested contingency analysis revealed a signiWcant relationship between the genetic and geographic distributions in the one-step clade 1–1 and the two-step clade 2–1 which contained haplotypes of the R. pseudochrysanthum complex (Table 5, Fig. 4). A signiWcant relationship between the genetic and geographic distributions was found for the one-step clade 1–6, whereas the two-one-step clade 2–2 showed no signiWcant relationship between the genetic and geographic distributions which contained haplotypes of R. formosanum (Table 5, Fig. 4). The geographic distance analysis showed signiWcant diVerences for clades 1–1 and 1–3 in both the clade (DC) and nested clade (DN) distances. The inference according to the key in Templeton (2004) is of restricted gene Xow with isolation-by-distance among haplotypes of the R. pseudochrysanthum complex (clade 2–1). Contiguous range expansion was inferred for nested clade 1–6, which contains haplotypes mostly found in the Sanlinxi population of R. for-mosanum. However, no conclusive result was inferred for the higher-level clade 2–2 for haplotypes of R. formosanum.
4. Discussion
4.1. Genetic variability and haplotype relationships
Our analyses of cpDNA sequences showed relatively low levels of genetic variation within the R. pseudochrysanthum
Fig. 3. Plots of the mean FST and NST values for each population pared to every other one in the Rhododendron pseudochrysanthum com-plex. RR, Rhododendron rubropunctatum; RP, Rhododendron
pseudochrysanthum; RM, Rhododendron morii; RH, Rhododendron hyperythrum.
Table 4
Analysis of molecular variance results using two diVerent grouping schemes
(A) The Wrst group is Rhododendron formosanum, and the second group is the R. pseudochrysanthum species complex. (B) Among species: among the four species of the R. pseudochrysanthum complex.
Among populations within species: among populations of each individual species of the R. pseudochrysanthum complex. Within populations: within populations of the R. pseudochrysanthum complex.
Source of variation df Sum of squares Variance components Percentage of variation p Value
(A) Between groups 1 10.701 0.25590 58.15 <0.0001
Within group 122 22.469 0.18417 41.85 <0.0001
Total 123 33.169 0.44007
(B) Among species 3 4.892 ¡0.01522 ¡6.07 0.44086
Among populations within species 3 6.382 0.13818 55.09 <0.0001
Within populations 91 11.634 0.12785 50.97 <0.0001
Total 97 22.908 0.25081
Table 5
Inferences based upon the nested clade analysis
The inference chain refers to the sequence following the inference key found in the appendix of Templeton (2004). S D DC or DN values that are
signiW-cantly smaller than expected at the 5% level based on 1000 permutations. L D DC or DN values that are signiWcantly larger than expected at the 5% level
based on 1000 permutations.
NCA clade Permutational X2 statistic Probability D
C DN Inference chain Inferred pattern
1-1 19.5141 0.0320 74.75L 74.18L 1-2-11-17-No Inconclusive outcome 1-3 2.8800 0.1710 2.79S 14.09S 1-2-11-17-No Inconclusive outcome
1-4 3.0000 0.3490 9.42 108.81 1-2-11-17-No Inconclusive outcome
1-6 16.8333 0.0000 68.54 68.54 1-2-11-12-No Contiguous range expansion
2-1 38.8642 0.0000 70.36 71.62 1-2-3-4-No Restricted gene Xow with isolation by distance
2-2 2.0000 1.0000 68.54 68.66 1-2-11-17-No Inconclusive outcome
complex in contrast to that of R. formosanum. Indeed, only eight cpDNA haplotypes were found across the seven pop-ulations examined for the R. pseudochrysanthum complex in contrast to the seven cpDNA haplotypes found for the two populations of R. formosanum. Moreover, in the Ali-shan population of R. morii and the only population (Nanhutashan) examined of R. hyperythrum, no additional haplotype other than haploype 1 was found and displayed no nucleotide diversity. The limited amount of genetic vari-ation in the R. pseudochrysanthum complex is consistent with the relatively low levels of molecular divergence gener-ally possessed by endemic island taxa (Frankham, 1997; Gemmill et al., 1998), but this might not hold true for all species (Francisco-Ortega et al., 2000).
The level of cpDNA nucleotide diversity of R. formosa-num was higher than that of many widely distributed vascu-lar plants in Taiwan such as Cunninghamia konishii (Hwang et al., 2003), Cyclobalanopsis glauca (Huang et al., 2002), Castanopsis carlesii (Cheng et al., 2005), Machilus thun-bergii, and M. kusanoi (Wu et al., 2006). The higher level of sequence polymorphism in R. formosanum was mainly con-tributed by the higher level of variation found in the atpB-rbcL segment of cpDNA.
The genetic diversity in a species or population is the outcome of its evolutionary history and of recent evolu-tionary processes. Low levels of genetic diversity may reduce the potential of species or populations to survive in a changing environment (Ellstrand and Elam, 1993; Lande and Shannon, 1996). The short-term eVective population sizes of the R. pseudochrysanthum complex may be consid-erably smaller than that of R. formosanum owing to diVer-ences in suitable habitats. The larger population sizes of R. formosanum may be the result of contiguous range expan-sion as inferred from the NCA analysis. Actually, this spe-cies is commonly seen at low- to mid-elevation broadleaf forests of Taiwan. For cpDNA haplotypes of R. pseudo-chrysanthum, we have seen only independent derivations of ancestral haplotype to haplotypes belonging to R.
rubro-punctatum (haplotypes 8 and 2) and the derivation of ances-tral haplotype to a haplotype belonging to R. pseudochrysanthum (haplotype 7) and then to a haplotype belonging to R. morii (haplotype 6) (Fig. 4). We observed no haplotype derivation from R. pseudochrysanthum or R. morii to R. rubropunctatum.
4.2. Phylogeny of the R. pseudochrysanthum complex Rhododendron pseudochrysanthum, R. morii, and R. rubropunctatum were classiWed as a single species in the sec-ond edition of the Flora of Taiwan (Li et al., 1998), while R. hyperythrum was treated as an independent species. Our results are interesting in that they show that the current tax-onomic treatment does not reXect the evolutionary history of the cpDNA gene tree. A reciprocally monophyletic rela-tionship was also detected for haplotypes belonging either to R. pseudochrysanthum or to the outgroup species, R. for-mosanum. The monophyly of the R. pseudochrysanthum complex is consistent with the Wndings of Milne (2004) that Rhododendron species in the subgenus Hymenanthes and subsection Maculifera form a well-supported monophyletic clade. The geographically coherent distribution of one ancestral cpDNA haplotype together with low levels of nucleotide variation and population diVerentiation in the R. pseudochrysanthum species complex warrant their treat-ment as a single species. These phylogenetic topologies sug-gest that species of the R. pseudochrysanthum complex are capable of genetic exchange, and probably do not consti-tute distinct operational taxonomic units. Alternatively, these results may simply reXect the retention of an ancestral haplotype, prior to divergence. However, these species may be undergoing the process of speciation due to genetic drift, which is apparent for the Hohuanshan population of R. morii. Possible ongoing genetic drift and/or speciation are also suggested by the 55% between-population variation found for populations of individual species of the R. pseudochrysanthum complex (Table 4).
Fig. 4. Nested haplotype network of Rhododendron species used for the nested clade analysis. Each line connecting one haplotype to another indicates one mutational step. Haplotype designations correspond to those in Table 1 and Fig. 1. Unsampled haplotypes are indicated by a 0. The hierarchical nesting design is speciWed by boxes and numbered clade designations.
4.3. Evolutionary history of the R. pseudochrysanthum complex
Analysis of the pattern of genetic variation in tree species using molecular markers emphasizes the importance of con-sidering historical factors when attempting to explain cur-rent patterns of population diVerentiation (Newton et al., 1999). Alternating periods of cooling and warming in associ-ation with changes in precipitassoci-ation patterns during the Pleis-tocene prompted allopatric speciation in the Northern Hemisphere (Comes and Kadereit, 1998; Hewitt, 2000). Ele-vational migration of species during glacial/interglacial extremes was suggested (Hewitt, 1996). Although very little of the landscape in Taiwan was covered by ice sheets, the temperature was 8.0–11.0 °C cooler compared to present-day temperatures, as determined from one lake core at an eleva-tion of 745.5 m in central Taiwan, and the vegetaeleva-tion was dominated by temperate species (Tsukada, 1966). It is prob-able that the R. pseudochrysanthum complex was geographi-cally widespread because of cooler temperatures and a drier environment. A remarkably high proportion of Poaceae was reported from pollen records indicating a dry environment from Toshe (650 m) and JihTan (750 m) in central Taiwan during the cold period (Liew and Chung, 2001).
Essentially no population diVerentiation was found by the AMOVA, which is concordant with the low values of the pairwise NST and pairwise FST results for most comparisons of populations in the R. pseudochrysanthum complex. The low level of population diVerentiation might have been resulted from a single origin and widespread distribution of this species complex in the past. The overwhelming abun-dance of ancestral haplotype sharing is consistent with the concept of a single origin of this species complex and could not be attributed to mutational homoplasy because rates of substitution were generally low. Shared polymorphisms reveal a history of polymorphisms that have not yet been erased by genetic drift; they reXect either short divergence times between taxa or historically large population sizes, with recurrent gene Xow or incomplete lineage sorting of ancestral variation. Recent long-distance dispersal among R. rubropunctatum and other species in the R. pseudochrysant-hum complex is unlikely due to the very small sample sizes of R. rubropunctatum and the mostly allopatric distributions of populations in the complex, despite seeds in the subgenus Hymenanthes being small and having wings and thus being easily dispersed by wind. Furthermore, haplotype sharing (haplotype 7) other than the ancestral haplotype only occurred in sympatric populations of R. pseudochrysanthum and R. morii (Hohuanshan populations) and argues against a signiWcant migration having occurred between popula-tions. These results suggest a Pleistocene origin for the allo-patric distribution of the R. pseudochrysanthum complex distribution and indicate that north-to-south colonization probably occurred when suitable habitats were available.
The once-widespread distribution of the R. pseudochry-santhum complex is further supported by the demographic analysis. Departures from the standard
neutral-equilib-rium expectations can result from changes in the size or spatial contiguity of the population under study. Negative values of Tajima’s D indicate an excess of low-frequency variants relative to null expectations and may have been caused by population expansion (Tajima, 1989a,b). The NCA inferred a restricted gene Xow event with isolation-by-distance in the total cladogram of the R. pseudochry-santhum complex. The inference of restricted gene Xow with isolation-by-distance may be diYcult to separate from events of past fragmentation (Templeton et al., 1995). The NCA involves overlaying the geography on an estimated gene tree to measure the strength of any geo-graphic/phylogenetic associations and to interpret the evolutionary processes responsible (Avise, 2000), and thus attempts to disentangle past events from contemporary processes. This technique allows for statistical tests of his-torical changes in populational, regional, or species demo-graphics, such as evidence for population expansions and bottlenecks. Although range expansion events were not inferred for the R. pseudochrysanthum complex possibly as a result of a lack of mutations detected, the lack of mutations might have resulted from population fragmen-tation caused by rising temperatures and changing envi-ronments of previously continuously occupied habitats.
5. Conclusions
Over the course of time, a species’ range becomes vari-ously discontinuous through disruptions of previvari-ously existing continuity (Levin, 2005). Currently, the R. pseudo-chrysanthum complex is mainly distributed on high moun-tain peaks in northern and central mounmoun-tains in Taiwan. Past range expansions of R. rubropunctatum and the entire R. peudochrysanthum complex are evident from the signi W-cant negative value of Tajima’s D and the low R2 value. However, population fragmentation might have occurred after the Pleistocene. The patchy distributions of the R. pseudochrysanthum complex are the results of either long-distance dispersal or remnants of a once-broader range. As the climate warmed at the end of the Quaternary, tree pop-ulations became established at higher elevations (Davis and Shaw, 2001). In regions such as Taiwan that were predomi-nantly never glaciated over most of its landscape, many species and genera continued to grow at the same latitude but shifted from one range of elevations to another, result-ing in contractions in population sizes. It is likely that R. rubropunctatum was widely distributed from north to south at mid-elevations during the Pleistocene because of cooler temperatures, and subsequent reductions in population size occurred when the temperature warmed after the Pleisto-cene, resulting in range contraction and reduction in genetic variability.
Acknowledgments
This study was supported in part by Grants (92AS-4.1.2-FC-R1 and 93AS-4.1.1-FB-E1) from the Council of
Agriculture, Executive Yuan, Taiwan and by a competitive research Grant (94g302) from the Institute of Biodiversity, Academia Sinica, Taiwan. The authors acknowledge the assistance of Long-Chi Hsu for a portion of data collection. We are grateful to Ji-Shen Wu (Taiwan Forestry Research Institute) for his help with sample collection.
References
Avise, J.C., 2000. Phylogeography: The History and Formation of Species. Harvard University Press, Cambridge, MA.
Avise, J.C., Arnold, J., Ball, R.M., Bermingham, E., Lamb, T., Neigel, J.E., Reeb, C.A., Saunders, N.C., 1987. IntraspeciWc phylogeography; the mitochondrial DNA bridge between population genetics and systemat-ics. Ann. Rev. Ecol. Syst. 18, 489–522.
Chamberlain, D.F., 1982. A revision of Rhododendron II. Subgenus
Hyme-nanthes. Notes R. Bot. Gard. Edinb. 39, 209–486.
Chamberlain, D.F., Hyam, R., Argent, G., Fairweather, G., Walter, K.S., 1996. The genus Rhododendron, its classiWcation and synonymy. R. Bot. Gard. Edinb..
Cheng, Y.P., Hwang, S.Y., Lin, T.P., 2005. Potential refugia in Taiwan revealed by the phylogeographical study of Castanopsis carlesii Hayata (Fagaceae). Mol. Ecol. 14, 2075–2085.
Clement, M., Posada, D., Crandall, K.A., 2000. TCS: a computer program to estimate gene genealogies. Mol. Ecol. 9, 1657–1659.
Comes, H.P., Kadereit, J.W., 1998. The eVect of Quaternary climatic change on plant distribution and evolution. Trends Plant Sci. 3, 432–438. Crandall, K.A., Templeton, A.R., 1993. Empirical tests of some predictions
from coalescent theory with applications to intraspeciWc phylogeny reconstruction. Genetics 134, 959–969.
Davis, M.B., Shaw, R.G., 2001. Range shifts and adaptive responses to Quaternary climate change. Science 292, 673–679.
Downie, S.R., Watson, M.F., Spalik, K., Katz-Downie, D.S., 2000. Molecu-lar systematics of Old world Apioideae (Apiaceae): relationships among some members of tribe Peucedaneae sensu lato, the placement of several island-endemic species, and resolution within the apioid superclade. Can. J. Bot. 78, 506–528.
Doyle, J.J., Doyle, J.L., 1987. A rapid DNA isolation procedure for small quantities of fresh leaf material. Phytochem. Bull. 19, 11–15.
Ellstrand, N.C., Elam, D.R., 1993. Population genetic consequences of small population size: implications for plant conservation. Ann. Rev. Ecol. Syst. 24, 217–242.
ExcoYer, L., Smouse, P.E., Quattro, J.M., 1992. Analysis of molecular vari-ance inferred from metric distvari-ances among DNA haplotypes: application to human mitochondrial DNA restriction data. Genetics 131, 479–491. Farris, J.S., 1989. The retention index and the rescaled consistency index.
Cladistics 5, 417–419.
Farris, J.S., Källersjö, M., Kluge, A.G., Bult, C., 1994. Testing signiWcance of congruence. Cladistics 10, 315–319.
Felsenstein, J., 1985. ConWdence limits on phylogenies: an approach using the bootstrap. Evolution 39, 783–791.
Francisco-Ortega, J., Santos-Guerra, A., Kim, S.C., Crawford, D.J., 2000. Plant genetic diversity in the Canary Islands: a conservation perspec-tive. Am. J. Bot. 87, 909–919.
Frankham, R., 1997. Do island population have less genetic variation than mainland populations? Heredity 78, 311–327.
Gemmill, C.E.C., Ranker, T.A., Ragone, D., Perlman, S.P., Wood, K.R., 1998. Conservation genetics of the endangered endemic Hawaiian genus Brighamia (Campanulaceae). Am. J. Bot. 85, 528–539.
Hewitt, G.M., 1996. Some genetic consequences of ice ages, and their role in divergence and speciation. Biol. J. Linn. Soc. 58, 247–276.
Hewitt, G.M., 2000. The genetic legacy of the Quaternary ice ages. Nature 405, 907–913.
Huang, S.S.F., Hwang, S.Y., Lin, T.P., 2002. Spatial pattern of chloroplast DNA variation of Cyclobalanopsis glauca in Taiwan and East Asia. Mol. Ecol. 11, 2349–2358.
Hwang, S.Y., Hsu, K.K., 2001. Molecular phylogeny of eight Taiwanese
Rhododendron species based on chloroplast trnF-trnL DNA sequences.
Taiwan J. For. Sci. 16, 153–160 (in Chinese with English summary). Hwang, S.Y., Lin, T.P., Ma, C.S., Lin, C.L., Chung, J.D., Yang, J.C., 2003.
Postglacial population growth of Cunninghamia konishii (Cupressa-ceae) inferred from phylogeographical and mismatch analysis of chlo-roplast DNA variation. Mol. Ecol. 12, 2689–2695.
Kluge, A.G., Farris, J.S., 1969. Quantitative phyletics and the evolution of Anurans. Syst. Zoo. 18, 1–32.
Lande, R., Shannon, S., 1996. The role of genetic variation in adaptation and population persistence in a changing environment. Evolution 50, 434–437. Levin, D.A., 2005. Niche shifts: the primary driver of novelty within
angio-sperm genera. Syst. Bot. 30, 9–15.
Li, H.L., Lu, S.Y., Yang, Y.P., Tseng, Y.H., 1998. Ericaceae. In: Flora of Taiwan, Editorial Committee of the Flora of Taiwan, (second ed.), Vol 4, pp. 17–39. Liew, P.M., Chung, N.J., 2001. Vertical migration of forests during the last
glacial period in subtropical Taiwan. West. Pac. Earth Sci. 1, 405–414. Lu, S.Y., Yang, Y.P., 1989. A revision of Rhododendron (Ericaceae) of
Tai-wan. Bull. Taiwan Forest Res. Inst. New Series 4, 155–166 (in Chinese with English summary).
Milne, R.I., 2004. Phylogeny and biogeography of Rhododendron subsec-tion Pontica, a group with a tertiary relict distribusubsec-tion. Mol. Phyloge-net. Evol. 33, 389–401.
Nei, M., 1987. Molecular evolutionary genetics. Columbia University Press, New York.
Newton, A.C., Allnutt, T., Gillies, A.C.M., Lowe, A., Ennos, R.A., 1999. Molecular phylogeography, intraspeciWc variation and the conserva-tion of tree species. Trends Ecol. Evol. 14, 140–145.
Pons, O., Petit, R.J., 1996. Measuring and testing genetic diVerentiation with ordered versus unordered alleles. Genetics 144, 1237–1245. Posada, D., Crandall, K.A., Templeton, A.R., 2000. GeoDis, a program for
the cladistic nested analysis of the geographical distribution of genetic haplotypes. Mol. Ecol. 9, 487–488.
Ramos-Onsins, S.E., Rozas, J., 2002. Statistical properties of new neutral-ity tests against population growth. Mol. Biol. Evol. 19, 2092–2497. Rozas, J., Sànchez-DelBarrio, J.C., Messequer, X., Rozas, R., 2003. DnaSP,
DNA polymorphism analyses by the coalescent and other methods. Bioinformatics 19, 2496–2497.
Schneider, S., Roessli, D., ExcoYer, L., 2000. Arlequin ver. 2.000: a soft-ware for population genetic data analysis. Geneva, Switzerland, Genet-ics and Biometry Laboratory, University of Geneva.
Soltis, P.S., Gitzendanner, M.A., 1999. Molecular systematics and the con-servation of rare species. Conserv. Biol. 3, 471–483.
SwoVord, D.L., 2001. PAUP¤ Phylogenetic analysis using parsimony (¤ and other methods). Version 4. Sunderland, MA, Sinauer Associates. Taberlet, P., Gielly, L., Pautou, G., Buvet, J., 1991. Universal primers for
ampliWcation of three non-coding regions of chloroplast DNA. Plant Mol. Biol. 17, 1105–1109.
Tajima, F., 1989a. Statistical method for testing the neutral mutation hypothesis by DNA polymorphism. Genetics 123, 585–595.
Tajima, F., 1989b. The eVect of change in population size on DNA poly-morphism. Genetics 123, 597–601.
Templeton, A.R., 2004. Statistical phylogeography: methods of evaluating and minimizing inference errors. Mol. Ecol. 4, 789–809.
Templeton, A.R., Boerwinkle, E., Sing, C.F., 1987. A cladistic analysis of phenotypic associations with haplotypes inferred from restriction endonuclease mapping. I. Basic theory and analysis of alcohol dehy-drogenase activity in Drosophila. Genetics 117, 343–351.
Templeton, A.R., Crandall, K.A., Sing, C.F., 1992. A cladistic analysis of phenotypic associations with haplotypes inferred from restriction endonuclease mapping and DNA sequence data. III. Cladogram esti-mation. Genetics 132, 619–633.
Templeton, A.R., Routman, E., Phillips, C.A., 1995. Separating population structure from population history: a cladistic analysis of the geo-graphic distribution of mitochondrial DNA haplotypes in the tiger sal-amander, Ambystoma tigrinum. Genetics 140, 767–782.
Templeton, A.R., Sing, C.F., 1993. A cladistic analysis of phenotypic asso-ciations with haplotypes inferred from restriction endonuclease
mapping IV. Nested analysis with cladogram uncertainty and recombi-nation. Genetics 134, 659–669.
Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997. The Clustal_X windows interface: Xexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 25, 4876–4882.
Tsai, C.C., Huang, S.C., Chen, C.H., Tseng, Y.H., Huang, P.L., Tsai, S.H., Chou, C.H., 2003. Genetic relationships of Rhododendron (Ericaceae)
in Taiwan based on the sequence of the internal transcribed spacer of ribosomal DNA. J. Hort. Sci. Biotech. 78, 234–240.
Tsukada, M., 1966. Late Pleistocene vegetation and climate in Taiwan (Formosa). Proc. Natl. Acad. Sci. USA 55, 543–548.
Wu, S.H., Hwang, C.Y., Lin, T.P., Chung, J.D., Cheng, Y.P., Hwang, S.Y., 2006. Contrasting phylogeographic patterns of two closely related spe-cies, Machilus thunbergii and Machilus kusanoi (Lauraceae), in Taiwan. J. Biogeo. 33, 936–947.
