DSpace at VNU: Phylogeography of Kandelia candel in East Asiatic mangroves based on nucleotide variation of chloroplast and mitochondrial DNAs

DSpace at VNU: Phylogeography of Kandelia candel in East Asiatic mangroves based on nucleotide variation of chloroplast...

Molecular Ecology (2001) 10, 2697–2710

Blackwell Science Ltd

mangroves based on nucleotide variation of chloroplast and mitochondrial DNAs

T Y C H I A N G , * Y C C H I A N G , † Y J C H E N , † C H C H O U , ‡ S H AVA N O N D , § T N H O N G ¶ and S H U A N G †

*Department of Biology, Cheng-Kung University, Tainan 701, Taiwan, †Department of Biology, National Taiwan Normal University, Taipei, Taiwan 116, ‡Institute of Botany, Academia Sinica, Taipei 115, Taiwan, §Silvicultural Research Division, Royal Sorest Department, Bangkok, 10900, Thailand, ¶Centre Forest Natural Resources and Environmental Studies, Vietnam National University, Vietnam

Abstract Vivipary with precocious seedlings in mangrove plants was thought to be a hindrance to long-range dispersal To examine the extent of seedling dispersal across oceans, we invest-igated the phylogeny and genetic structure among East Asiatic populations of Kandelia candel based on organelle DNAs In total, three, 28 and seven haplotypes of the chloroplast DNA (cpDNA) atpB-rbcL spacer, cpDNA trnL-trnF spacer, and mitochondrial DNA (mtDNA) internal transcribed spacer (ITS) were identified, respectively, from 202 indi-viduals Three data sets suggested consistent phylogenies recovering two differentiated lineages corresponding to geographical regions, i.e northern South-China-Sea + East-China-Sea region and southern South-China-East-China-Sea region (Sarawak) Phylogenetically, the Sarawak population was closely related to the Ranong population of western Peninsula Malaysia instead of other South-China-Sea populations, indicating its possible origin from the Indian Ocean Rim No geographical subdivision was detected within the northern geo-graphical region An analysis of molecular variance ( AMOVA ) revealed low levels of genetic differentiation between and within mainland and island populations (ΦCT = 0.015,

ΦSC = 0.037), indicating conspicuous long-distance seedling dispersal across oceans Signi-ficant linkage disequilibrium excluded the possibility of recurrent homoplasious mutations

as the major force causing phylogenetic discrepancy between mtDNA and the trnL-trnF spacer within the northern region Instead, relative ages of alleles contributed to non-random chlorotype–mitotype associations and tree inconsistency Widespread distribution and random associations (χ2 = 0.822, P = 0.189) of eight hypothetical ancestral cytotypes indicated the panmixis of populations of the northern geographical region as a whole In contrast, rare and recently evolved alleles were restricted to marginal populations, reveal-ing some preferential directional migration.

Keywords: cpDNA, Kandelia candel, locus association, migration, minimum spanning network, mtDNA, phylogeography, relative ages

Received 15 May 2001; revision received 5 August 2001; accepted 5 August 2001

Introduction

Kandelia candel, a monotypic genus of the Rhizophoraceae,

is one of the major mangrove species in East Asia

(Tomlinson 1986; Mabberley 1997) Mangrove forests are characteristic of tropical and subtropical coastlines of the world Over the last decades, due to human’s over-exploitation, the genetic diversity in mangroves has been deprived, especially from coastal ecosystems in Asia Many countries have categorized the sustainable manage-ment of mangroves as major priorities in biodiversity

Correspondence : S Huang Fax: + 886-2-29312904; E-mail:

biofv057@scc.ntnu.edu.tw

MEC_1399.fm Page 2697 Wednesday, October 24, 2001 6:39 PM

2698 T Y C H I A N G E T A L.

conservation (Maguire et al 2000) According to

bio-geographical evidence, the genus Kandelia underwent a

severe extinction phase during the upper Tertiary after the

closure of the Tethys Seaway (Schwarzbach & Ricklefs

1998) Since then populations of K candel have been

restricted to the tropical Malesia and East Asia,

includ-ing areas around the South China Sea and the East

China Sea

Like many other mangrove species, K candel is

charac-terized by vivipary, the precocious growth of the progeny

when still attached to the maternal parent Vivipary is

an adaptive feature for mangrove plants to colonize and

expand populations at intertidal estuary habitats

Preco-cious seedlings of K candel, growing up to 47 cm long and

1.3 cm wide, are buoyant, which may allow long-range

dis-persal via ocean currents (Lugo & Snedaker 1974; Maxwell

1995) However, viviparous seedlings have also been

con-sidered as a hindrance to long-distance dispersal due to the

lack of protection and nutritional support from the

mater-nal tissue (Duke 1995; Elmqvist & Cox 1996)

Correspond-ingly, extremely various population structures have been

discovered in different mangrove species High level of

genetic differentiation among populations as well as

geo-graphical subdivision, due to restricted gene flow across

populations, was lately detected in Avicennia marina

(Avi-cenniaceae) (Duke 1995; Maguire et al 2000), a result close

to the population structure of a nonviviparous species,

Acanthus ilicifolius (Lakshmi et al 1997) In contrast,

Aus-tralian populations of Rhizophora stylosa were found barely

differentiated (Goodall & Stoddart 1989) Recent allozyme

investigations revealed low level of genetic differentiation

(GST = 0.064, Sun et al 1998) among Hong Kong

popula-tions as well as between two populapopula-tions from Taiwan

(FST = 0.04, Huang 1994), both indicating that the

seed-ling dispersal in K candel was not as limited as previously

suggested

Although frequent gene flow has been detected in K.

candel at the local scale, long-range dispersal across oceans

remains unknown Apparently, the dispersal extent of

K candel is not only regulated by the orientation of ocean

currents in the South China Sea and East China Sea, but also

constrained by the duration and survivorship of

vivipar-ous seedlings in the high saline conditions The

isolation-by-distance model is thus a hypothesis to be tested In

addition, when geographical distance increases for

invest-igation, effects of vicariance will become more prominent

and may confound the isolation-by-distance model (Bossart

& Prowell 1998) Geological history inevitably plays

another critical role in determining the phylogeographical

pattern For example, geographically close populations

along western and eastern coasts of the Peninsula Malaysia

were significantly differentiated due to a vicariance event

of approximately 60 – 220 million years before present

(Yamazaki 1998), which led to the geographical subdivision

of Bruguiera gymnorrhiza in Asia According to palae-oceanographic evidence, due to latitude and temperature differences, southern and northern banks of the South China Sea, where K candel is distributed, went through different geological histories (Wang et al 1995) over past glacial cycles Furthermore, populations of the Ryukyu islands and northern Taiwan (Taipei) along coasts of the East China Sea shared a unique geological history from those of the South China Sea (Ota 1998; Chou et al 2000)

In light of geological histories, genetic differentiation of

K candel among the above three geographical regions would be expected

As generally known, in estimating population structure and gene flow, some level of variance of loci is required Molecular markers with low resolution usually are incap-able of providing information (Bossart & Prowell 1998) in distinguishing coancestry from migration (Schaal et al 1998) For surveying population structure within a small geographical scale, e.g K candel in Hong Kong (Sun et al 1998) and Taiwan (Huang 1994), allozymes due to their conserved nature (cf Bossart & Prowell 1998) might not be able to adequately estimate gene flow among local popula-tions In addition, the biparental inheritance of allozymes makes the estimations of seedling dispersal difficult, as the effects of pollen dispersal between neighbouring popula-tions cannot be ruled out Recently, many noncoding spacers of organelle DNAs have been widely applied to phylogeographical studies (Schaal et al 1998) With merits

of maternal inheritance in most angiosperms (Harris & Ingram 1991), including Kandelia (Chen 2000), and nearly neutral and fast evolution, these markers are likely to be able to provide information in estimating the extent of long-range seedling dispersal and reconstructing phylo-geographical patterns

In this study, we sequenced two noncoding spacers, i.e

atpB-rbcL and trnL-trnF of chloroplast DNA (cpDNA), and the internal transcribed spacer (ITS) of mitochondrial DNA (mtDNA) and used them as markers to estimate the phylo-geographical pattern as well as population and geograph-ical structure of K candel As physically linked loci in the chloroplast genome, atpB-rbcL and trnL-trnF should reveal comparable phylogenies Likewise, as being maternally inherited, chloroplasts and mitochondria were thought to remain associated and behave as if they are completely linked (Schnabel & Asmussen 1989) Consistent phylo-genetic patterns of cpDNA and mtDNA are thus expected (cf Dumolin-Lapègue et al 1998)

However, evolutionary forces, such as lineage sorting effects (Hoelzer et al 1998), and frequent recurrent muta-tions (Desplanque et al 2000), can result in systematic inconsistency and thus lead to wide variance of FST values among loci and a weak correlation between FST and number of migrants per generation (Nm) as well (cf Bossart

& Prowell 1998) Under such circumstances, difficulties in

MEC_1399.fm Page 2698 Wednesday, October 24, 2001 6:39 PM

P H Y L O G E O G R A P H Y O F K A N D E L I A 2699

interpreting phylogeography and population structure will

be inevitably encountered In other words, when

discrep-ant estimates are obtained from two or more loci, these

esti-mates are not necessarily indicative of gene flow Explicit

analysis of associations between alleles of different loci

(Desplanque et al 2000) coupled with nested clade analysis

(cf Schaal et al 1998; Templeton 1998; Chiang 2000) will be

required to clarify historical and recurrent events

K candel, as widespread in East Asia, is a biological

model that is suited for testing the association between

vicariance and geographical structure; effects of ocean

currents in long-distance seedling dispersal; and the

use-fulness of cpDNA and mtDNA as population genetic

markers, as well as allele associations Several aims are

pursued in this study: (i) to test the possibility and level of

long-distance seedling dispersal by estimating population

structure and gene flow; (ii) reconstruct the

phylogeo-graphical pattern and examine the association between

geological/geographical events and the extent of genetic

differentiation among three geographical regions; (iii) to

examine the phylogenetic consistency between atpB-rbcL and trnL-trnF noncoding spacers of cpDNA as well as between chloroplast and mitochondrial genomes; and (iv) to investigate associations between alleles of cp- and mtDNAs and to deduce the relative age of their alleles based on spanning networks

Materials and methods

Sample collection Kandelia candel is a mangrove species that is widespread

in eastern coasts of mainland Asia and continental islands, including Taiwan, and the Ryukyu (Fig 1) One hundred and eighty-seven samples were collected from 13 major populations in East Asia, ranging from Bako (Sarawak,

01°40′ N) to Yakushima Islet ( Japan, 30°20′ N) (Table 1, Fig 1) Ten to 15 individuals, which were approximately

70 – 100 m apart, were sampled from each population In addition, 15 individuals of a population at Ranong

Fig 1 Kandelia candel sample locations and distribution Frequency of cytotypes (cpDNA trnL-trnF–mtDNA ITS associations) in each

population is indicated in pie diagrams Abbreviations of populations are given in Table 1

MEC_1399.fm Page 2699 Wednesday, October 24, 2001 6:39 PM

2700 T Y C H I A N G E T A L.

(Thailand) of the western coast of Peninsula Malaysia were

included in the analysis as outgroups In total, this study

encompasses 14 populations (202 individuals) No materials

were collected from the Philippines, since no natural

populations are distributed in this area (Hou 1958) In

addition, during the last decade, most populations of

Vietnam have been removed for the use of inshore

fisheries (Huang & Chen 2000) Only one population at

Quang Ninh was available Besides, samples of six

populations (from Chinchou to Xiamen) of China, one

population from Sarawak, two populations of Taiwan,

and three populations from the Ryukyu Islands were

included in this investigation Based on the orientation of

ocean currents (Huang et al 1997), three geographical

regions are recognized: southern South-China-Sea (S

region, including a single population of Sarawak),

northern South-China-Sea (N region, including eight

populations of Vietnam, China, and Tainan), and the

East-China-Sea (E region, including three populations of

Ryukyu and Taipei) regions Three populations along the

Tonkin Bay, i.e Chinchou, Shankou, and Quang Ninh, are

further grouped as the Ns region, while others of the N

region as the Nn region Young and healthy shoots were

collected in the field, rinsed with tap water and dried in

silica gel All samples were stored at –70 °C until they

were processed

DNA extraction and polymerase chain reaction

Leaf tissue or embryo of the above materials was ground to powder in liquid nitrogen and stored in a –70 °C freezer Genomic DNA was extracted from the powdered tissue following the CTAB procedure (Murray & Thompson 1980) Noncoding spacers of atpB-rbcL and trnL-trnF of the cpDNA and the ribosomal DNA (rDNA) ITS of mtDNA were amplified and sequenced Universal primers for amplifying atpB-rbcL spacer (Chiang et al 1998), trnL-trnF spacer (Taberlet et al 1991), and mtDNA rITS (Chao et al 1984) were synthesized The polymerase chain reaction (PCR) amplification was carried out in a volume of 100 µL reaction using 10 ng of template DNA, 10 µL of 10× reaction buffer, 10 µL MgCl2 (25 mm), 10 µL dNTP mix (8 mm), 10 pmole of each primer, 10 µL of 10% NP-40, and 2 U of Taq

polymerase (Promega, Madison, USA) The reaction was programmed on a MJ Thermal Cycler (PTC 100) as one cycle of denaturation at 95 °C for 4 min, 30 cycles of 45 s denaturation at 92 °C, 1 min 15 s annealing at 52 °C, and

1 min 30 s extension at 72 °C, followed by 10 min extension

at 72 °C Template DNA was denatured with reaction buffer, MgCl2, NP-40 and ddH2O for 4 mins (first cycle), and cooled on ice immediately A pair of primers, dNTP and Taq polymerase were added to the above ice-cold mix Reaction was restarted at the first annealing at 52 °C

Table 1 Materials of Kandelia candel collected from different populations used for organelle DNA sequencing Location, locality, sample

size, and cytotypes (trnL-trnF spacer chlorotype and mtITS mitotype associations) of each population are indicated

Sampling

size (n) Cytotypes

Northern Region

Northern South-China-Sea Region:

Chinchou Guangxi, China CC 21°34′ N, 108°37′ E 15 IB (12), IC (1), IIB (1), IIIB (1)

Haikou Hainan, China HA 19°54′ N, 110°20′ E 15 IB (6), IIB (3), IIIB (3), VB (3)

Hong Kong China HK 22°32′ N, 114°05′ E 13 IB (9), IVB (2), VB (2)

Quang Ninh Vietnam QN 20°53′ N, 106°46′ E 15 IIB (2), IIE (1), IVB (2), IVE (1), VB (3),

VIB (4), VIC (1), VIE (1) Shankou Guangxi, China SK 21°28′ N, 109°43′ E 10 IB (1), IIB (1), IID (1), IIIB (6), IIID (1) Tainan Taiwan TN 22°59′ N, 120°12′ E 15 IB (13), IC (1), IIIC (1)

Xiamen Fukien, China XM 24°26′ N, 118°04′ E 15 IB (2), IIB (9), IIC (1), IVB (3)

Zhangjiang Guangxi, China ZJ 21°04′ N, 110°09′ E 15 IIIB (7), IVB (1), IVC (1), VIB (6) East-China-Sea Region:

Amami-O-Shima Islet Ryukyu, Japan AM 28°15′ N, 130°40′ E 15 IB (6), IIB (3), IIIB (2), IIIC (1), IVB (2),

IVC (1) Irimote Islet Ryukyu, Japan IR 24°19′ N, 123°54′ E 15 IA (2), IB (4), IIB (3), IVB (6)

Taipei Taiwan TP 25°09′ N, 120°16′ E 14 IB (9), IIB (3), IIIB (2)

Yakushima Islet Ryukyu, Japan YK 30°20′ N, 130°30′ E 15 IA (1), IB (8), IC (2), IIB (2), IVB (2)

Southern Region

Southern South-China-Sea Region

Indian Ocean Rim (outgroup)

Ranong Thailand RN 09°55′ N, 98°30′ E 15 IIaF (13), VIIaG (2)

MEC_1399.fm Page 2700 Wednesday, October 24, 2001 6:39 PM

P H Y L O G E O G R A P H Y O F K A N D E L I A 2701

T-A cloning and nucleotide sequencing

PCR products were purified by electrophoresis on a 1.0%

agarose gel using 1× TAE buffer The gel was stained with

ethidium bromide and the desired DNA band was cut and

eluted using agarose gel purification (QIAGEN) Purified

DNA was ligated to a pGEM-T easy vector (Promega)

Plasmid DNA was selected randomly with five clones

and purified using plasmid mini kits (QIAGEN) Purified

plasmid DNA was sequenced in both directions by

standard methods of the Taq dyedeoxyterminator cycle

sequencing kit (Perkin Elmer) on an Applied Biosystems

Model 377A automated sequencer Primers for sequence

determination were T7-promoter and SP6-promoter

located on p-GEM-T easy vector termination site

Sequence alignments and phylogenetic analyses

Nucleotide sequences were aligned with the program

Genetics Computer Group (gcg) Wisconsin Package

(Version 10.0, Madison, Wisconsin) Neighbour-joining

(NJ) analysis by calculating Kimura 2-parameter distance

(Kimura 1980) was also performed using Data Analysis in

Molecular Biology and Evolution (dambe, version 3.5.19,

Xia 1999) Indels were treated as the fifth character

Confidence of the clades reconstructed was tested by

bootstrapping (Felsenstein 1985) with 1000 replicates using

unweighted characters The nodes with bootstrap values

greater than 0.70, as a rule of thumb, are significantly

supported with ≥ 95% probability (Hillis & Bull 1993) The

number of mutations between DNA genotypes in pairwise

comparisons, which were calculated using mega (Kumar

et al 1993), was used to construct a minimum spanning

network with the aid of minspnet (Excoffier & Smouse

1994) in an hierarchical manner (cf Chiang & Schaal 1999)

After linking the related haplotypes into a clade, closely

related clades were linked further to form higher level

groups and thereby a network

Population genetic analysis of the cpDNA and

mtDNA sequence variation

Levels of genetic diversity within populations were

quantified by estimates of nucleotide divergence (θ)

(Watterson 1975) using dnasp (Version 3.14, Rozas &

Rozas 1999) Patterns of geographical subdivision and

gene flow were also estimated hierarchically with the aid

of dnasp Gene flow within and among regions

(popu-lations) was approximated as Nm, the number of female

migrants per generation between populations, and was

estimated using the expression FST = 1/(1 + 2 Nm) where N

is the female elective population size and m is the female

migration rate (Slatkin 1993) We also used amova version

1.55 (Excoffier et al 1992; Excoffier 1993) to deduce the

significance of geographical divisions both among regions and populations The statistics of molecular variants ΦCT

(among regions), ΦST (among populations), and ΦSC

(among populations within regions), were estimated

Results

Extent of nucleotide and haplotype diversity

In this study, trnL-trnF and atpB-rbcL noncoding spacers of cpDNA and the rITS of mtDNA in Kandelia candel were

PCR amplified and sequenced All sequences were regis-tered with EMBL accession numbers AJ305472 –AJ305673

(for mtDNA ITS), AJ305674 –AJ305875 (for trnL-trnF spacer), and AJ305876 – AJ306077 (for atpB-rbcL spacer).

At all three loci, no within-individual variation was detected

Differences between mitochondrial sequences of a consensus length of 725 bp were mainly ascribed to point mutations (18 sites, 2.4%) Four indel events also occurred

at sites (353 – 357) (592 –599) (632 – 635) and 640 For the

atpB-rbcL spacer of the chloroplast genome, only two sites

of 781 bp (0.3%) were polymorphic Populations of BK and

RN shared a C at site 160, while others have a T A 1-bp deletion at site 577 occurred in the RN population At the

trnL-trnF spacer of cpDNA, high levels of length

polymor-phism, ranging from 375 bp to 415 bp, were detected Two deletions, at sites (17 – 28) and (242 – 268) made the spacer shorter in populations of BK and RN Most indels (37) occurring in the noncoding spacer involved a 1-bp loss Like most noncoding regions (cf Li 1997), A + T contents

were high, with 53.7%, 72.5%, and 76.9% at mtITS, atpB-rbcL spacer, and trnL-trnF spacer, respectively.

Three haplotypes of the atpB-rbcL noncoding spacer (h = 0.268 0.0015, θ = 0.00051 0.00008) were determined In

contrast to other studies (Small et al 1998; Fujii et al 1999), the atpB-rbcL noncoding spacer possessed a much lower level of genetic variation than the trnL-trnF spacer in K candel (Table 2) In total, 28 haplotypes of the trnL-trnF

spacer of cpDNA and seven haplotypes of the mtDNA ITS were determined Apparently, the molecular evolution of the mtDNA ITS was much more constrained compared to

that of the trnL-trnF spacer of cpDNA (Table 2).

At the population level, except for the lack of genetic

vari-ation at the cpDNA atpB-rbcL spacer, the level of genetic

variation varied among populations at two other organelle loci (Table 2) High levels of nucleotide variation at both loci, with θ-values ranging from 0.0018 to 0.0062 at trnL-trnF spacer and from 0.0020 to 0.0029 at mtITS, occurred in

populations of IR, QN, and YK, while low cpDNA vari-ation, ranging from 0.0003 to 0.0013, was detected in popu-lations of TN, CC, and HK Stark contrast of the level of genetic variation between the two loci occurred in popula-tions of BK, HA, TP, and XM (Table 2)

2702 T Y C H I A N G E T A L

Gene genealogies and associations between cpDNA

and mtDNA lineages

In this study we reconstructed the phylogeographical

pattern of K candel based on gene genealogies of organelle

loci A minimum spanning network of the cpDNA

atpB-rbcL spacer was reconstructed based on mutational

changes between haplotypes (Fig 2) The BK population is

closely related to the RN population, while no variation

was detected among populations of N and E regions An

NJ tree was recovered based on the nucleotide sequence

variation of the trnL-trnF noncoding spacer of cpDNA.

Eight clades (chlorotypes) of 28 haplotypes were identified

in this cpDNA gene tree (Fig 3) Two major lineages of

S + RN and N + E were recognized and significantly

supported, with a bootstrap of 0.98 (P < 0.01) Four

com-mon chlorotypes I–IV of 152 sequences in total (75.2%)

were widespread in populations of N and E regions, while

types of VIIa and VIIb of 30 sequences were restricted to the

S region and RN population (Table 3) Two rare alleles of V

(4.0%) and VI (5.9%) were distributed in the N region only

A minimum spanning network of the trnL-trnF

noncod-ing spacer was constructed (Fig 4) Eight clades of the

network, corresponding to those of the NJ tree, were divided into two geographical groups (i.e S + RN and N + E) 34 mutations apart Within the network, closely related chloro-types were mostly linked by single mutations Chlorochloro-types

I, II and III were nested in the network as interior nodes, while types IV and V connecting to type I, and type VI connecting to III or IV were exterior

In the NJ tree of mtDNA ITS sequences, two major clades (A–E) and (F, G), of seven variants (mitotypes) were identified (Fig 5A) Two common mitotypes B and

C of 164 sequences (81.2%) were widespread in popula-tions of N and E regions (Table 3), as mitotypes of F and

G were distributed in the S region and RN population Three rare alleles were distributed restrictedly: types A (1.5%) in YK and IR populations, D (1.0%) in SK, and E (1.5%) in QN A minimum spanning network of the mtDNA ITS was constructed (Fig 5B) Within the clade of

N + E regions, mitotype B was nested in the network as the interior node, connecting to other types independ-ently with 1 – 9 mutations Within the clade of S + RN regions, mitotype G was linked to the interior node of type F with two mutations, which was linked to type B with four mutations

Phylogenies of cpDNA and mtDNA are completely consistent at the level of geographical regions (i.e S + RN

vs N + E) The atpB-rbcL noncoding spacer provided no

information in resolving the phylogeny within N + E

regions Apparently, the gene tree of the trnL-trnF spacer of

cpDNA largely contradicted the mtDNA ITS tree No clade correspondence was found between the two trees For example, although most chlorotype III sequences (87.5%) corresponded to the mitotype B, sequences of am4 and tn12 were associated with mitotype C and sequence sk2 was associated with the mitotype D

cpDNA mtDNA

Total 0.908 0.02652 ± 0.00526 0.406 0.00205 ± 0.00026

Amami-O-Shima 0.857 0.00623 ± 0.00044 0.248 0.00034 ± 0.00018

Chinchou 0.257 0.00125 ± 0.00005 0.133 0.00018 ± 0.00015

Haikou 0.771 0.00329 ± 0.00003 0.000 0.00000

Hong Kong 0.513 0.00135 ± 0.00044 0.000 0.00000

Irimote 0.686 0.00310 ± 0.00164 0.248 0.00292 ± 0.00060

Quang Ninh 0.771 0.00492 ± 0.00071 0.448 0.00255 ± 0.00034

Ranong 0.600 0.00246 ± 0.00004 0.248 0.00034 ± 0.00018

Shankou 0.778 0.00248 ± 0.00001 0.467 0.00064 ± 0.00018

Tainan 0.133 0.00034 ± 0.00018 0.248 0.00034 ± 0.00018

Taipei 0.560 0.00304 ± 0.00082 0.000 0.00000

Xiamen 0.733 0.00278 ± 0.00049 0.133 0.00018 ± 0.00015

Yakushima 0.457 0.00178 ± 0.00066 0.362 0.00200 ± 0.00139

Zhangjiang 0.848 0.00457 ± 0.00040 0.133 0.00018 ± 0.00015

Table 2 Haplotype diversity (h) and

nucleo-tide diversity (θ) of the cpDNA

trnL-trnF spacer and the mtDNA ITS within

populations of Kandelia candel

atpB-rbcL network

Fig 2 Minimum spanning network generated using method of

Excoffier & Smouse (1994) for haplotypes of atpB-rbcL spacer of

cpDNA of populations of Kandelia candel Each arrow indicates one

mutational change ‘0’ indicates hypothetical ancestor

P H Y L O G E O G R A P H Y O F K A N D E L I A 2703

Nevertheless, associations between cpDNA and mtDNA

haplotypes were nonrandom Seventeen (instead of 30;

χ2 = 1.03, P = 0.00018) and three (instead of four; χ2 =

0.044, P = 0.0001) cpDNA (trnL-trnF)-mtDNA associated

cytotypes were observed in N + E and S + RN regions of K candel, respectively (Table 4) All sequences of the mitotype

A were exclusively associated with the chlorotype I; and sequences of the mitotype D were associated with the

Fig 3 Neighbour-joining tree of representative sequences (haplotypes) of trnL-trnF of cpDNA in Kandelia candel Numbers at nodes indicate

bootstrap values Chlorotypes (I–VIIb) are labelled on clades

Table 3 Distribution of chlorotypes (I–VIIb) and mitotypes (A–G) among populations of Kandelia candel Regions are indicated: Indian

Ocean Rim (I), southern South-China-Sea region (S), northern South-China-Sea region (N), and East-China-Sea region (E)

2704 T Y C H I A N G E T A L

Fig 4 Minimum spanning network generated using method of Excoffier & Smouse (1994) for haplotypes of trnL-trnF spacer of cpDNA of

populations of Kandelia candel Each arrow indicates one mutational change Number of mutational change is indicated when more than

one step ‘0’ indicates hypothetical ancestor The replicate number of haplotypes is also indicated when more than one

Fig 5 (A) Neighbour-joining tree of

haplo-types (A – G) of rITS of mtDNA in Kandelia

candel Numbers at nodes indicate bootstrap

values ( B) Minimum spanning network generated using method of Excoffier & Smouse (1994) for haplotypes of mtDNA

ITS of populations of K candel Mutational

changes are indicated at nodes

P H Y L O G E O G R A P H Y O F K A N D E L I A 2705

chlorotypes II and III Likewise, the chlorotype V was

exclus-ively associated with the most dominant mitotype B, and

most sequences of chlorotype I were associated with the

mitotype B, while some other sequences were mitotypes A

or C Within the N + E region, cytotypes BI (40.7%) and BII

(15.7%) were most dominant in composition In contrast,

cytotypes of CII, DII, EII, CIII, DIII, CIV, EIV, CVI, and EVI

were relatively rare (5.8% in total) Likewise, within the

S + RN region, FVIIa and FVIIb were dominant (93.3%),

while the cytotype GVIIa was rare (6.7%)

The cytotype composition varied among populations In

Fig 1 the genetic composition in each population was

indic-ated Higher number of cytotypes occurred in populations

AM (six types), QN (eight types), and YK (five types),

while a single type was detected in the BK population and

two types were detected in the RN population (Table 1)

Within the N + E regions, populations possessing a higher

number of cytotypes appeared to be located at margins (Fig 1) In contrast, most geographically ‘central’ popula-tions had three or four cytotypes

Population differentiation and geographical divisions

Population and geographical structure of K candel was

assessed based on genetic variation of the organelle loci

Estimates of FST (= 0.93 – 0.95) and Nm (= 0.03 – 0.04) based

on the cpDNA trnL-trnF spacer and mtDNA ITS, indicated

significant differentiation between regions S + RN, and

N + E In contrast to the consistent estimations of population structure between the above two loci, higher

number of migrants (Nm = 0.10) per generation was deduced from atpB-rbcL spacer sequences, although the genetic differentiation was still significant (FST = 0.828) Hierarchical analyses of sequence difference under amova

Table 4 Associations between chlorotypes and mitotypes of Kandelia candel Distribution range of each chlorotype and mitotype is indicated

in square brackets Percentage of each cytotype is indicated in parentheses W: widespread Other symbols see Table 1

chlorotype: mitotype:

A [ir + yk]

B [W]

C [W]

D [sk]

E [qn]

F [rn + bk]

G [bk] Total

(1.5%) (34.6%) (2.0%)

(13.4%) (0.5%) (0.5%) (0.5%)

(10.4%) (1.0%) (0.5%)

(4.0%)

(6.4%) (1.0%)

(7.4%)

Table 5 Pairwise FST/ Nm estimates between geographical regions (E, N, Ns and Nn) based on genetic variation of mtDNA ITS (above the

diagonal) and cpDNA trnL-trnF spacer (below the diagonal) Nucleotide diversity within each region is indicated in the parenthesis

E (θ = 0.00138 ± 0.00048)

N (θ = 0.00049 ± 0.00007)

Ns (θ = 0.00120 ± 0.00049)

Nn (θ = 0.00135 ± 0.00007)

E

(θ = 0.00503 ± 0.00054)

N

(θ = 0.00445 ± 0.00033)

Ns

(θ = 0.00312 ± 0.00037)

Nn

(θ = 0.00492 ± 0.00041)

2706 T Y C H I A N G E T A L

indicated that the proportion of molecular variance was

attributed to difference between geographical regions

(ΦCT = 0.860, P < 0.001) The relative contribution of

dif-ference among populations to molecular variance was

small (ΦST = 0.087) In contrast, no geographical

sub-division (FST = 0.020 – 0.026 and Nm = 18.41 – 23.95) between

N and E was detected (Table 5)

An analysis of molecular variance (amova) based on the

trnL-trnF spacer of cpDNA also suggested low levels of

genetic differentiation between populations of mainland

and continental islands (ΦCT = 0.015) as well as among

populations within each region (ΦSC = 0.037) Genetic

variation of the mtDNA ITS yielded a similar pattern of

the genetic apportionment among geographical regions

and populations The deduced Nm of 39.00 – 45.67 indicated

frequent gene flow between Ns and Nn regions (Table 5)

In contrast to the invariable atpB-rbcL spacer, pairwise

comparisons showed high variances in genetic estimates of

structure and genetic differentiation between cpDNA and

mtDNA loci, except for those between BK + RN and other

populations Low level of genetic differentiation was

usu-ally detected among populations within the E + N regions

Nearly all Nm values, ranging from 1.75 (between SK and

HK) to 622.55 (between CC and YK), deduced from the

mtDNA ITS were greater than those from the cpDNA

trnL-trnF spacer Lower Nm values, less than one, were mostly

restricted to comparisons with ZJ as well as SK based on

cpDNA variation Some extremely high Nm values were

obtained, such as 150 between AM and CC High variance

in deducing FST and Nm was also encountered in

compar-isons between BK and RN Based on the trnL-trnF spacer,

an FST value of 0.42 and an Nm of 0.68 were deduced, while

a lower level of genetic differentiation (FST = 0.13, Nm =

3.50) was detected from the mtITS

Discussion

Genetic variability and low level of homoplasious

mutations in cpDNAs and mtDNAs of Kandelia candel

The usefulness of molecular markers in indirect estimates

of population structure and gene flow depends on the level

of resolution, and locus-to-locus consistency (Bossart &

Prowell 1998), and is also affected by the level of recurrent

mutations (cf Desplanque et al 2000) Recurrent mutations

(identity by state), which are frequently encountered in the

mitochondrial genome due to limited conformations in

molecular structure (Fauron et al 1995), will inevitably blur

the level of migration between populations Technically,

nucleotide sequencing can simply rule out the length

homo-plasies, which occur usually in restriction fragment length

polymorphism (RFLP) and PCR-based fingerprints (cf Parker

et al 1998; Desplanque et al 2000), from the data scoring.

In this study, as a very strong linkage disequilibrium was

estimated between mitotypes and chlorotypes both within

N + E and S + RN regions (Table 4), a high rate of recurrent mutations can be excluded for organelle genomes of

K candel (cf Desplanque et al 2000).

In this study, genetic variation of mtDNA ITS and

cpDNA trnL-trnF spacer existed both within and between

populations Nevertheless, the haplotype diversity of the mitochondrial DNA, with seven haplotypes out of 202 plants screened, was lower compared to other flowering

plants, such as Beta vulgaris ssp maritima (20 mitotypes from 414 individuals; Desplanque et al 2000), Daucus carota ssp carota (25 variants based on mtDNA RFLP from 80 plants; Ronfort et al 1995), Thymus vulgaris (50 mitotypes from about 400 plants; Manicacci et al 1996), and Hevea brasiliensis (212 mtDNA RFLP variants in 395 accessions screened; Luo et al 1995).

For the chloroplast genome, K candel possessed a higher level of haplotype diversity (28 haplotypes) at the trnL-trnF spacer, which is close to 13 cpDNA haplotypes in Beta vul-garis ssp maritima (Desplanque et al 2000), 11 haplotypes

in Argania (El Mousadik & Petit 1996), 23 haplotypes in white oaks (Dumolin-Lapègue et al 1997), and 13 haplo-types in Alnus (King & Ferris 1998) Although comparisons

of haplotype diversity among taxa, which were examined using various molecular methods at different loci, may be

somewhat misleading, nucleotide diversity of the trnL-trnF spacer in K candel (θ = 0.02710) appeared to be higher

than that of other plants as well, such as Cunninghamia

(θ = 0.01018, Lu et al 2001) and Begonia (θ = 0.003, Liu 1999) Apparently, both loci in this study provided suffi-cient resolution at the geographical region level and

yielded consistent estimates of gene flow (Nm = 0.03 – 0.04) and population structure (FST = 0.93 – 0.95) between S + RN

and N + E populations of K candel To the contrary, lack of

variability, due to its conserved nature (cf Chiang & Schaal

2000a,b), has made the atpB-rbcL noncoding spacer locus

powerless in estimating the interpopulation migration within the N + E regions At interregions level, as a result

of having difficulties in distinguishing coancestry from

migration, higher Nm value (= 0.10) between S + RN and

N + E regions was thereby deduced based on this spacer Likewise, at the population level, the more conserved

mtDNA ITS always yielded higher values of Nm and lower levels of FST, than the cpDNA trnL-trnF spacer In this investigation, due to its higher resolution, the trnL-trnF

might have higher probabilities of yielding estimates that

approximate the current population structure of K candel.

Phylogeographical patterns of K candel in East Asia

In this study, we investigated the phylogeography of the

viviparous species, K candel Both the mtDNA ITS and the cpDNA trnL-trnF spacer suggested noticeable

long-distance seedling dispersal However, as extensive gene