Society of Systematic Biologists

Phylogeny and Biogeography of Exacum (Gentianaceae): A Disjunctive Distribution in the

Indian Ocean Basin Resulted from Long Distance Dispersal and Extensive Radiation
Author(s): Yong-Ming Yuan, Sébastien Wohlhauser, Michael Möller, Jens Klackenberg, Martin
W. Callmander, Philippe Küpfer
Source: Systematic Biology, Vol. 54, No. 1 (Feb., 2005), pp. 21-34
Published by: Oxford University Press for the Society of Systematic Biologists
Stable URL: http://www.jstor.org/stable/20061208
Accessed: 29/08/2009 11:23

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at
http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless
you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you
may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at
http://www.jstor.org/action/showPublisher?publisherCode=ssbiol.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed
page of such transmission.

JSTOR is a not-for-profit organization founded in 1995 to build trusted digital archives for scholarship. We work with the
scholarly community to preserve their work and the materials they rely upon, and to build a common research platform that
promotes the discovery and use of these resources. For more information about JSTOR, please contact support@jstor.org.

Society of Systematic Biologists and Oxford University Press are collaborating with JSTOR to digitize,
preserve and extend access to Systematic Biology.

http://www.jstor.org
 Biol.
Syst. 54(1 ):21-34,2005
of
Copyright ? Society Systematic Biologists
ISSN: 1063-5157 print / 1076-836X online
DOI: 10.1080/10635150590905867

A
Phylogeny and Biogeography of Exacum (Gentianaceae): Disjunctive Distribution in
the Indian Ocean Basin Resulted from Long Distance Dispersal and Extensive Radiation

Yong-Ming Yuan,1-4 S?bastien Wohlhauser,1 Michael M?ller,2 Jens Klackenberg,3

Martin W. Callmander,1 and Philippe Kupfer1
1
Laboratory of Evolutionary Botany, Institute of Botany, University ofNeuch?tel, Emile-Argand 11, CH-2007 Neuch?tel, Switzerland;
E-mail: yong-ming.yuan@unine.ch
3 2Royal Botanic Garden Edinburgh, Edinburgh EH3 5LR, Scotland, United Kingdom
Department of Phanerogamic Botany, Swedish Museum of Natural History, S-10405 Stockholm, Sweden

4South China Botanical Garden, Chinese Academy of Sciences, Guangzhou, P. R. China

Abstract.?Disjunctive distributions across paleotropical regions in the Indian Ocean Basin (IOB) often invoke disper
in
sal/vicariance debates. Exacum (Gentianaceae, tribe Exaceae) species are spread around the IOB, Africa, Madagascar,
Sri the
Socotra, the Arabian peninsula, Lanka, India, Himalayas, mainland Southeast Asia including southern China and
this
Malaysia, and northern Australia. The distribution of genus was suggested to be a typical example of vicariance resulting
the is in
from breakup of the Gondwanan supercontinent. The molecular phylogeny of Exacum principle congruent with mor
a
phological conclusions and shows pattern that resembles a vicariance scenario with rapid divergence among lineages, but
our molecular dating analysis demonstrates that the radiation is too recent to be associated with the Gondwanan continental
breakup. We used our dating analysis to test the results of DIVA and found that the program predicted impossible vicari
ance events. Ancestral area reconstruction in
suggests that Exacum originated Madagascar, and divergence dating suggests
its origin was not before the Eocene. TheMadagascan progenitor, the most recent common ancestor of Exacum, colonized
Sri Lanka and southern India via long-distance dispersals. This colonizer underwent an extensive range expansion and
to
spread Socotra-Arabia, northern India, and mainland Southeast Asia in the northern IOB when itwas warm and humid
in these regions. This widespread common ancestor retreated subsequently from most parts of these regions and survived
in a
in isolation Socotra-Arabia, southern India-Sri Lanka, and perhaps mainland Southeast Asia, possibly as consequence
the
of drastic climatic changes, particularly the spreading drought during Neogene. Secondary diversification from these
surviving centers and Madagascar resulted in the extant main lineages of the genus. The vicariance-like pattern shown by
the phylogeny appears to have resulted from long-distance dispersals followed by extensive range expansion and subse
from
quent fragmentation. The extant African species E. oldenlandioides is confirmed to be recently dispersed Madagascar.
ITS;
[Biogeography; DIVA; Exacum; Gentianaceae; phylogeny; trnL intron.]

The historical biogeographic connection through the

zoic dispersal was considered as amore plausible expla
paleotropical regions around the Indian Ocean Basin
nation than Mesozoic vicariance. A few similar debates
in
(IOB) area stands out as a conspicuous pattern and a exist plants; for example, the oceanic dispersal versus
the the
key topic in the biogeography of plants and animals Gondwana vicariance for biogeography of genus
Rax et the
of the world (Raven and Axelrod, 1974; worthy Adansonia (Baum al., 1998), circum-global dispersal
et is the
al., 2002). This primarily due to the firmly estab versus Gondwana vicariance for history of the family
and et
lished and well-documented geologic history of this Melastomataceae (Renner Meyer, 2001; Renner al.,
in
region, the sequential breakup of the Gondwana land 2001), migration versus vicariance Malpighiaceae
mass, and the differential shifting history of different (Davis et al., 2002a, 2002b), the post-Gondwana long
tectonic plates (McLoughlin, 2001). Thus, biogeographic distance dispersal versus Gondwana vicariance for the
patterns in this region provide ideal models for testing historical biogeography of thegenus Polyscias (Plunkett
the
the long-lasting debate between vicariance and dispersal et al., 2001), and widespread-extinction hypothesis
(Rieppel, 2002). In addition to the Gondwana vicariance versus vicariance or long-distance dispersal theories for
the the
and long-distance dispersal explanations, noticeably the history of genus Nepenthes (Meimberg et al., 2001 ).
earlier land-bridge theory (Steenis, 1962), and the more Recently, Sanmartin and Ronquist (2004) also found that
recent "Lemurian the of
stepping-stones" hypothesis (Schatz, most of plant patterns suspected Gondwanan ori
1996) have been proposed to interpret thisbiogeographic gin fit a dispersal rather than a vicariance scenario. As
connection. The distribution of chameleons used to be urged by Rieppel (2002), studies on other groups of or
a
considered as a typical vicariance example related to ganisms with geographical distribution similar to that
Gondwana fragmentation, but recent studies strongly of chameleons are needed to verify whether some regu
suggest that their present distribution is the result of larity underlies their phylogenetic and biogeographical
extensive oceanic dispersals (Raxworthy et al., 2002; patterns. The plant genus Exacum (Gentianaceae, tribe
a
Rieppel, 2002). A similar situation applies to certain Exaceae) appears as good model for this purpose due
its
freshwater fish: cichlids show a typical Gondwana vi to disjunctive distribution pattern in the IOB.
cariance distribution pattern, which is congruent with The genus Exacum consists of 64 species and
the phylogenetic relationships among the lineages. Re shows a typical paleotropical distribution (see Fig. 1)
cent divergence dating, however, revealed significant (Klackenberg, 1985, 2002; Thulin, 2001). Our previous
discrepancy between the times of lineage divergence and study has addressed its phylogenetic position within
the tectonic events (Vences et al.,2001). Therefore, Ceno the tribe Exaceae (Yuan et al., 2003). Klackenberg (1985)

21
 22 SYSTEMATIC BIOLOGY VOL. 54

FIGURE 1. Distribution of the extant species of Exacum and the areas of endemism with their relevant species numbers.

monographed the genus and divided it into two sec Paleomagnetic data and tectonic reconstruction
of
tions based on phylogenetic studies morphological suggested that the Gondwana breakup initiated ca.
and anatomical characters: sect. Africana consisting of 180 million years ago (Mya) (Storey, 1995), and the
African, Madagascan, Socotran and southern Arabian Madagascar-Seychelles-India block began to separate
165
species, and sect. Exacum consisting of species of other from the Africa-South America block ca. Mya, with
121
Asian regions (Sri Lanka, India, through mainland movement ending by Mya (Rabinowitz et al., 1983).
Southeast Asia). The majority of species (38 species) Since then, Madagascar has remained in its position
in
occur Madagascar. Sri Lanka and the southern tip with respect to Africa (Coffin and Rabino witz, 1988),
of the Indian subcontinent (mainly the Western Ghats) whereas Australia and Antarctica separated from the
is the second most species-rich area (14 species limited Madagascar-Seychelles-India block ca. 132 Mya (Barron,
from 88
to this area, 3 species shared with north India, the 1987). India separated Madagascar ca. Mya
Himalayas, and other areas). Three species are found (Storey et al, 1995) and split from the Seychelles ca.
65
on the southern Arabian peninsula (Dhofar of Oman Mya (Storey, 1995). Although the Seychelles block
of to
and nearby Mahrah Yemen) and the Island of Socotra subsequently became fixed with respect Africa, India
(Thulin, 2001). The African continent harbors only two rapidly drifted northward and collided with Laurasia
species: E. oldenlandioides widespread throughout the during the Paleocene-Eocene. Some authors suggested
43
entire tropical Africa and E. zombense endemic to the the India-Asia collision occurred ca. Mya (Lee and
1995; the
Shire Highlands in Malawi (Klackenberg, 1985). Two Lawver, McLoughlin, 2001), whereas majority
E. E. are
species, pedunculatum and petiolare, widespread of authors suggest an earlier collision between 50 and
66
in India. Two species, E. hamiltonii and E. teres, are Mya (Besse et al., 1984; Patriat and Achache, 1984;
the E. et
restricted to Himalayas, and two species, pteran Jaeger al., 1989; Klootwijk et al., 1992; Beck et al.,
E.
thum and sutaepense, are limited to the mountainous 1995; Valdiya, 2002). Sri Lanka had remained in full
regions between Burma and Thailand (referred to as contact with India during all these processes until
mainland Southeast Asia herein). Only one species, E. the last major sea level rise 6000 years ago, which
is
tetragonum, widespread in Indomalesia (India, the Hi separated Sri Lanka from India by the narrow and
Palk Strait
malayas, mainland Southeast Asia including southern shallow (McLoughlin, 2001). After separation
and
China, Malaysia) and reaches the extreme north from Antarctica in the Cretaceous and Paleogene (96
a
of Australia (Arnhem Land). Based on phylogeny to 35.5 Mya), the Australia-New Guinea block moved
reconstructed from morphological and anatomical northward and collided with Southeast Asia in the late
25
characters, a Gondwana vicariance hypothesis has been Oligoc?ne to the beginning of the Miocene ca. Mya
to a
suggested explain the distribution pattern of Exacum, (Lee and Lawver, 1995; Li and Powell, 2001). Socotra,
especially for the divergence among the species of continental fragment, was located adjacent to the Dhofar
Madagascar, India, and Socotra-Arabia (Klackenberg, region of southern Oman prior to the Gulf of Aden
1985, 2002). Obviously, this hypothesis requires the rifting (Jolivet and Faccenna, 2000). The Afro-Arabian
assumption of a very old age of the genus with species plate collided with Eurasia in the eastern part around
as 25
divergence early as the Gondwana breakup, in order Mya (Samuei et al., 1997). At about the same time
to link the disjunctions of species distributions to the the Gulf of Aden rift appeared (Richardson et al., 1995;
et
fragmentation events of the Gondwana landmass. Birse al., 1997; Ghebreab, 1998; Fantozzi and Sgavetti,
 2005_YUAN ET AL.?PHYLOGENY AND BIOGEOGRAPHY OF EXACUM_23

1998), and subsequently the Arabian continent became analyses of the whole family including African Sebaea
separated from continental Africa and Socotra ca. 10 confirmed that inclusion or exclusion of this clade did
its the
Mya (Ghebreab, 1998). Socotra remained in position not interfere with geographic assessment for Exacum.
relative to the African continent ever since (Horn of Our previous studies have shown evidence that the
et
Africa) (Richardson et al., 1995; Birse al., 1997; Samuei saprophytic genus Cotylanthera resided inside a mono
et
al., 1997). phyletic Exacum. It consists of four species occurring
Our recent investigations on the tribal relationships in the Himalayas, mainland Southeast Asia, and the
within Gentianaceae suggested a minimum age of Malaysian area. We did not include this genus in our
40
Mya for the tribe Exaceae (Yuan et al., 2003), which present study because, on the one hand its sequences
apparently argues against a Gondwanan origin of the were incomplete (the chloroplast sequence were not ob
genus Exacum. Alternative hypotheses to explain the tainable); on the other hand it has been shown to have
to
historical biogeography of the genus relevant post undergone an accelerated molecular evolution (cf. Yuan
Gondwana vicariance/dispersal events are therefore et al., 2003: Fig. 3), which may be problematic in the
needed. To search for such rational explanations, we ap present phylogenetic analysis. However, omitting se
this
plied maximum likelihood (ML) and maximum parsi quences of genus had no significant effect on the tree
mony (MP) methods to reconstruct the phylogeny of topology (data not shown) and thus should not affect
Exacum using chloroplast and nuclear ribosomal DNA analyses on Exacum. Furthermore, this group showed
sequences, and divergence dating and ancestral area re affinities only with the more terminal branches of the
construction were conducted based on the molecular phylogenetic tree of Exacum (E. hamiltonii from the Hi
phylogeny. malayas and E. tetragonum widespread in Indomalesia)
(cf. Yuan et al., 2003: Figs. 2 and 3), and the biogeogra
is
Materials and Methods phy patterns of Exacum across the IOB region greatly
and depending on the basal branching patterns (see below).
Ingroup Sampling Outgroup Choice
To obtain an accurate divergence time estimation for Molecular Methods
the tribe Exaceae, our previously analyzed family data
et of 18 were obtained
matrix (Yuan al. 2003) was expanded by the addi Sequences species previously
E. (Yuan et al., 2003). for an additional
Sequences 24 ac
tion of two trnL intron sequences of Eisianthius, longi
cessions were for this The
folius (accession AF102450), and L. laxiflorus (accession newly acquired study. pro
AF102449), retrieved from GeneBank and was reana cedures of DNA extraction, polymerase chain reaction
lyzed with additional calibration points (see below). (PCR) amplification of target fragments, purification of
Our Exacum data set included 42 of the PCR products, and sequencing follow Yuan et al. (2003).
sequences
nuclear ribosomal internal transcribed spacer (ITS) re All sequences have been deposited in GenBank (acces
trnL intron:
sion numbers for the AJ490202-38,40,42-43,
gions, including the spacer 1, the 5.8S gene, and the ITS:
spacer 2 and 42 chloroplast irnL(UAA) intron se 49-50; for AJ489877-914, 917-918, 923-924).
1).
quences {trnL) (Appendix Sampling of Exacum species
was maximized to all areas of endemism Sequence Alignment, Congruence Test,
represent and
(Africa, Madagascar, Socotra-Arabia, India/Sri Lanka, Phylogenetic
Analysis
and mainland Southeast Asia) and morphological di The sequences were aligned with Clustal X (Thompson
et
versity. About half the taxa from each geographic area al., 1997) followed by minor manual adjustments to im
were sampled. The species cover all main morphologi prove indel events in a few tandem repeat regions. Nine
the
cal variations such as different habits (small herb, large base pairs near beginning of the spacer ITS2 involved
be
herb, shrub, etc.), different floral types (size of corolla, tandem repeats and could not unambiguously aligned
merosity, etc.), and different types of testa cells (star due to alternative alignment possibilities. They were
shaped, isodiametric, etc.). Thirty-seven accessions rep excluded from subsequent phylogenetic analyses. The
resenting 30 species of Exacum were sampled, equiv alignment does not involve severe ambiguities, there
alent to about 47% of the genus. Two species (out of fore it was not necessary to use secondary structure for
to in
three, all endemic Madagascar) of Ornichia, the sis alignment. Sixteen unambiguously aligned indels (13
ter
genus of Exacum (Yuan et al., 2003), were also sam ITS and 3 in trnL) were potentially informative. These
pled as ingroups. An additional basal clade sister to the indels were then scored as binary characters regard
Exacum-Ornichia lineage (Yuan et al., 2003) was sam less of their length, and were added to the sequence
as
pled outgroup. This additional clade, composed of the data for MP analyses. The data matrix was deposited in
(2
genera Tachiadenus species sampled out of 11, all en TreeBASE and is also available from the corresponding
to
demic Madagascar) and Gentianothamnus (monotypic, author.
to as an
endemic Madagascar), was used outgroup to ac To assess the level of congruence between the trnL
the
commodate dispersal-vicariance (DIVA) analyses for and ITS data sets, we analyzed each data set inde
more reliable estimates of the ancestral areas of Exacum pendently to see if they produced a similar topology.
(Ronquist, 1996). was We also performed an incongruence length difference
The genus Sebaea, the most basal clade of the tribe, (ILD) test of Farris et al. (1995), implemented in PAUP*
as our
not included as an additional outgroup, previous 4.0M0 (Swofford, 2000) as the partition-homogeneity
24 SYSTEMATIC BIOLOGY VOL. 54

Exacum -
52/86j affine M8238a
" -
99/100 E. affine Wcff5 Socotra & S
- Arabian
peninsula
99/100 E. affine M6201
0.005 E. affine-M17126
substitutions/site E. caeruleum
65/83 Himalayas
-
56/90 E. K254
tetragonum Indomalesia
94/94 E. -
tetragonum LK332 Sri
100/100 -E. Lanka,
87/95JL pedunculatum
-E. sessile S India, N India
E. Mainland SE Asia
sutaepense -
89/85 100/97r r Et. trinervium Zsl001
87/91 | Le. trinervium ZSI002
I
F m;
100/100 E. macranthum Zsl003
51/64 E. macranthum FK767
1? Sri Lanka
E.
51/<50 pallidum & S India
- E. atropurpureum
100/100 E.
wightianum
-
E. walkeri K539
-
100/1001DO/100? E. walkeri Zsl004
61/601/601 E. nummularifolium
68/90 millotii
- E. subteres
E.
99/981 marojejyense
E. fruticosum
l?2 ?
E.
microcarpum
. Madagascar
56/60 humbertii
E. subacaule
-E. intermedium
100/100 E. subverticillatum
E. bulbilliferum
E. quinquenervium
E. exiguum
E.
stenophyllum
E. oldenlandioides Africa

E. dolichantherum
E. linearifolium
100/100 i-Ornichia madagascariensis
?
O. trinervis Madagascar
100/100 Tachiadenus carinatus
-T. longiflorus
Gentianothamnus madagascariensis

FIGURE 2. tree = trnL

Phylogram single ML (-In 4350.2744), based on combined ITS and sequences. Distributions of the terminal
the
right. The bootstrap values of both ML and MP analyses supporting corresponding branches are shown
species are also indicated on the above the
internal branches when and
greater than 50% (ML/MP). "a" (al, a2), "b," "c," "d" are the main clades discussed in the text.

test, after assuring ourselves that the properties of estimated by bootstrap analysis of 1000 replicates of
to
the data did not lend themselves biasing this test heuristic searches with random sequence addition and
(see results). One thousand replicates of repartitions of TBR branch-swapping.
trnL intron versus ITS were conducted with heuristic Phylogenetic reconstructions based on the combined
searches (simple sequence addition and TBR branch data sets was also conducted using ML optimality cri
swapping). The two data sets were confirmed as con teria as implemented in PAUP*4.0blO (Swofford, 2000).
gruent (P = 0.74) and were then combined for all further The TIM+I+r model (Posada and Crandall, 1998) and
analyses. parameter settings were chosen by using the Akaike
MP analyses were conducted on separate trnL, ITS information criterion as suggested by Modeltest V3.06
1998).
data sets, and on a combination of the two, applying ac (Posada and Crandall, Optimal gene trees were
celerated transformation optimization (ACCTRAN) op found via heuristic searches of 100 replicates of random
tion. Heuristic searches of 1000 replicates of random sequence addition with TBR branch-swapping, MUL
sequence addition and TBR branch-swapping were per TREES, and STEEPEST DESCENT on. The relative clade
with for the ML analyses was estimated
formed, MULTREES and STEEPEST DESCENT on. support by boot
The relative clade of 100 of heuristic searches
support for MP analyses was also strap analysis replicates
 2005 YUAN ETAL.?PHYLOGENY AND BIOGEOGRAPHY OF EXACUM 25

Gentiana lutea
Gentiana
pyrenaica Gentianinae
Genijana.alpla_. ...
Genti?nell? umbellate.
Lomatogonium
macranthum
Swertia rosulata
Halenia
elliptica
Swertia
tetraptera Swertiinae
Frasera
speciosa
Swertia
calycina
Swertia
angustifolia
MegaaadQcistylopjAQrus._
Chelonanthus alatus
Chelonanthus
angustifolius
Chelonanthus
purpurascens
Macrocarpaea macrophylla
Anthocleista
amplexicaulis
Anthocleista grandiflora
Lisianthius longifolius
Lisianthius laxiflorus_
Chironia baccifera
Chironia laxa
Chironia linoides
Eustoma exaltatum
Orphium frutescens
Sabatia
angularis
Centaurium trichanthum
Centauriummadrenss
Centauriumspicatum
Exaculum pusillum
Symphyllophytoncaprifolioides
Microrphiumpubescens
Canscora diffusa
Canscora
andrographioides
Canscora alafa_
Exacum
stenophyllum
Exacum
quinquenervium
Exacum oldenlandioides
Exacum nummularifolium
Exacum fruticosum
Exacum
marojejyense
Exacum
tetragonum
Exacum hamiltonii
Exacum
wightianum
Exacum trinervium
Exacum affine
Exacum
gracilipes
Exacum caeruleum
Ornichia
madagascariensis
Ornichia trinervis
Sebaea
madagascariensis
Tachiadenus carinatus
Tachianenus longiflorus
Gentianothamnus
madagascariensis
Sebaea brachyphylla
Sebaea exacoides
Sebaea longicaulis
Sebaea _ macrophylla

Curtia tenuifolia
Voyriella parviclora Saccifolieae
Saccifolium bandeirae
Coffea arabica
Erithalis fruticosa
Mitreola
petiolata
Labordiatinifolia
Gelsemium
sempervirens Outgroups
Nerium oleander
Trachelospermum jasminoides

FIGURE 3. The strict consensus of six trees recovered from MP analyses (MIN collapse option) on 68 trnL intron sequences of Gentianaceae
RI = the
(length = 463, CI = 0.68 excluding autapomorphies, 0.89). These trees were used to infer divergence dates for the node marked with star,
by using penalized likelihood approaches and the four calibration points CP1 through CP4 shown on the tree. The bootstrap values supporting
the corresponding branches are shown above the internal branches when greater than 50%. The figures in the square are inferred divergence
dates for the node indicated as a star using different calibration points. Shown on the right is the tribal classification of the family. See text for
the details of the calibration points CP1 through CP4.

using the same model and parameters (Felsenstein, molecular clock was then subjected to a rate smoothing
1985). applying penalized likelihood (PL) approach using the
Test, .50 b).
Molecular Clock Divergence Calibration, software r8s v.l (Sanderson, 2002a, By applying ap
and calibrations, times can be estimated
Divergence Time Calculation propriate divergence
on the smoothened tree. PL is a
semiparametric smooth
To obtain approximate timings of branching events ing method that estimates relative branching time
within Exacum, it is necessary to estimate the times without assuming a molecular clock. Optimal smooth
of divergence of the main clades of the Exacum factors were chosen based on a data-driven cross
ing
tree. the likelihood ratio
By comparing statistic, validation procedure implemented in r8s (Sanderson,
?2(lnL dock?lnLnon-ciock), to the x2 distribution with 2002b).
n? 2 =
degrees of freedom (n number of taxa), a molec There is no fossil record for the tribe Exaceae to cali
and
ular clock was tested (Muse Weir, 1992). Finding brate the molecular phylogeny of Exacum. Our previous
that a clocklike sequence evolution had to be rejected studies (Yuan et al., 2003) estimated a minimum age of
40
(P < 0.01), the ML tree obtained in the absence of a Mya for the tribe Exaceae calibrated on the basis of
 54
26_SYSTEMATIC BIOLOGY_VOL.

pollen fossil of Gentianales (Muller, 1984). This minimum two clades under consideration was used to ensure the

age of the tribe could be used directly as a calibration. divergence times were not underestimated.
However, this estimate was based on the nonparamet
ric rate smoothing (NPRS) approach (Sanderson, 1997) Ancestral Area Reconstruction
with only one calibration. NPRS approaches have been
In order to obtain the historical scenarios of the bio
proven to overfit the data, allowing too much rate vari of
geography Exacum, DIVA analyses were conducted to
ation and therefore losing predictive power (Sanderson,
infer the ancestral areas for each internal node of the phy
2002a). As a further confirmation and cross-validation,
logeny, by using DIVA 1.1a (Ronquist, 1996,1997), on the
we therefore reanalyzed our previous trnL intron data
of the family Gentianaceae and its sister groups, with ML tree. DIVA reconstructs ancestral areas by minimiz
to
ing dispersal and extinction events needed explain the
additional sequences and fossil calibrations that we
observed distribution pattern based on an inferred phy
recognized recently. Two trnL intron sequences of the
L. logeny. Vicariance is considered as the default mode of
genus Lisianthius, longifolius (accession AF102450) and
speciation. The two zero-length branches of the ML tree
L. laxiflorus (accession AF102449), were retrieved from
were arbitrarily resolved to obtain dichotomies; this had
GeneBank and were added to our previous data matrix
no effect on the area assignment. Based on the analyses
to accommodate a fossil record of Lisianthius (Graham,
of the distribution of Exacum and its relatives, we rec
1984). This enlarged trnL intron matrix contained 68
taxa and 657 characters. This matrix was deposited in ognized eight areas of endemism: Africa, Madagascar,
Sri Lanka and south India, Socotra and southern Arabian
TreeBASE and is also available from the correspond
peninsula, central to northern India, the Himalayas,
ing author. Phylogenetic reconstructions on the enlarged and
trnL intron matrix were mainland Southeast Asia, Malaysia plus northern
performed using MP with 1000
Australia (Fig. 1). Terminal taxa were scored according
replicates of heuristic searches with random sequence to
and TBR to their distributions across the above eight areas gen
addition branch-swapping. Finding a molecu data
erate the distribution matrix, and this data matrix
lar clock has to be rejected (P< 0.01) for the enlarged tree.
was subsequently optimized onto theML Optimiza
trnL intron data set, the resulting MP trees were then of
tions both, unconstrained and area constrained to a
subjected to PL rate smoothing (Sanderson, 2002a) us were
maximum of two, conducted following the reason
ing the software r8s (Sanderson, 2002b). Four indepen of and
dent calibration points, minimum age of Gentianales ing Ronquist (1996,1997) Donoghue et al. (2001).
60
(CP1 = Mya) based on fossil pollen (M?ller, 1984), Results
40
minimum age of Lisianthius (CP2 = Mya) based on
Sequence Characteristics
fossil pollen (Graham, 1984), an inferred age of subtribe
= 15
Swertiinae (CP3 Mya) (Hagen and Kadereit, 2001, The trnL intron sequences ranged from 487 to 495 bp
= 5 in
2002), and minimum age of Gentiana (CP4 Mya) based in length Exacum, whereas the intron was as short as
and
on fossil seeds (Mai Walther, 1988), were used to 381 to 382 bp in the two species of Ornichia mainly due
infer the divergence dates between the outgroup Gen to a long gap of 106 bp. The aligned trnL intron matrix
508
tianothamnus-Tachiadenus clade and the ingroup Exacum had sequence and three indel characters. The number
of which
Ornichia clade. These inferred estimates of divergence of variable characters was 79 (15.5%), 36 (7.0%)
the
dates from global analyses were then used to calibrate were informative. The uncorrected pairwise sequence di
(E.
the Exacum tree to obtain divergence estimates. Such vergence was between 0% tetragonum?K254 versus
a E. E.
procedure has been successfully used by Davis et al. hamiltonii) E. macranthum?FK767 versus pallidum;
E. E. E.
(2002a). marojeyense versus fruticosum; marojeyense ver
PL E. E.
NPRS and may result in biased rates to the evenly sus humbertii; E. fruticosum versus humbertii) and
weighted MP branch lengths because they are not cor 3.9% (E. sessile versus E. trinervium?Zsl002) within the
for T.
rected multiple hits (Sanderson, 2002a). Ideally, the ingroup and maximal 4.5% (E. sessile versus longiflorus)
for ITS
branch lengths should be corrected multiple hits overall. The sequences ranged from 611 to 638 bp in
to
using an appropriate model of evolution (Yang, 1996). length. These length variations are due single or short
for
This is more critical highly divergent sequences as indels. The ITS data matrix had 659 sequence and 13 indel
of were
the likelihood multiple hits increases. However, in characters, of which 266 (39.6%) variable and 179
0% (E.
the enlarged trnL intron analysis, sequence divergence (26.6%) informative. The taxa had affine?M8238a
E.
was moderate (maximal 18% for ingroup taxa) and our versus affine?Wcff5; E. tetragonum?K254 versus E.
E. E.
calibration dates for the Exacum tree inferred from PL hamiltonii; marojeyense versus fruticosum) to 13.0% (E.
smoothened MP trees was not affected significantly. sessile versus E. trinervium?Zsl002) within ingroup and
the
As an independent cross-check of divergence-time maximal 16.4% (E. trinervium?Zsl002 versus T. longi
we
estimates based on the smoothened ML tree, also esti florus) uncorrected pairwise sequence divergence across
of
mated the times divergence of the trnL intron and ITS ingroup and outgroup taxa.
of
sequences employing various extreme rates measured The two sets sequences were revealed as congru
=
for different plants, compiled by Richardson et al. (2001). ent by the ILD test (P 0.74). Recent studies and sim
Divergence time between a pair of species was calcu ulations suggested that the ILD test could fail to detect
of
lated as half divergence value divided by the rate. The congruence due to different noise levels of the data sets
a et et
highest sequence distance between pair of species of (Dolphin al., 2000; Yoder al., 2001) or incongruence
 2005_YUAN ET AL.?PHYLOGENY AND BIOGEOGRAPHY OF EXACUM_27
and and
(Dowton Austin, 2002; Darlu Lecointre, 2002) clade (100% for both ML and MP), composed of the
due to large difference of the sizes and evolutionary con showy species endemic to Sri Lanka and the very south
ern
ditions of the data partitions. Our data sets had simi part of India (clade b); the Socotra-Arabia clade (99%
for
lar sizes (511 characters in trnL and 672 characters in for ML and 100% MP), containing species endemic
ITS. ITS showed a ca. 3.5-fold higher divergence (0% to Socotra and the southern Arabia peninsula (clade d);
to to for
16.4%) compared to trnL (0% 4.5%), but both re and the Indomalesia clade (87% forML and 95% MP),
the
vealed similar tree topology when analyzed separately. including the species endemic to Himalayas, main
Thus we consider that our two data sets were not suf land Southeast Asia, the species spread in Sri Lanka and
the
fering from these limitations, and, therefore, were com India, and species widely spread over the entire In
bined. The combined data matrix had 1167 bp sequence domalesia area including northern Australia (clade c).
characters plus 16 binary indel characters. It was de The Socotra-Arabia clade showed a closer relationship
posited in TreeBASEand is also available on the web to the Indomalesia clade than to any other clade (65% for
site http://www.unine.ch/bota/ebolab/gentianaceae/ ML and 83% for MP).
9
gentmain.html. Of the 1167 bp combined sequence, bp
and
(0.8%) ambiguously aligned ITS sequences were ex Divergence Calibration Dating
cluded from phylogenetic analyses. In order to obtain calibrations for the Exacum tree, phy
logenies were reconstructed for 68 representatives of the
Phylogenetic Analysis family Gentianaceae and its sister groups on trnL intron
MP analyses on trnL intron sequences alone generated data. MP analyses with MIN collapse option generated
=
1829 equally most parsimonious trees of 90 steps (CI six equally parsimonious trees of 463 steps (CI = 0.68
=
0.89 excluding autapomorphic sites, RI 0.94). The strict excluding autopomorphic sites, RI = 0.89). The strict
is
consensus (not shown) poorly resolved, and only the consensus was well resolved and the main clades were
in
south India-Sri Lanka clade (b Fig. 2) is retained. MP highly supported (Fig. 3). As a molecular clock has to
analyses on ITS data alone generated 2893 trees of 558 be rejected, these six MP trees with branch lengths op
=
steps (CI 0.58 excluding autopomorphic sites, RI = timized with ACCTRAN option were subjected to PL
is rate
0.82). The strict consensus (not shown) highly resolved smoothing. Applying the four different calibration
the
and the topology of the consensus is almost identical to points to the smoothened MP trees, divergence times
that generated from the combined data. The topology between Gentianothamnus-Tachiadenus clade and Exacum
in
of the main clades remained the same as the combined Ornichia clade (the node marked with a star Fig. 3)
one
analyses, with resolution within Madagascan clade are estimated as 20.3 to 25.9 Mya (CP1), 23.9 to 53.7 Mya
in 12.0
(a2 Fig. 2) slightly lower. (CP2), 18.2 to 40.8 Mya (CP3), to 26.4 Mya (CP4), and
data
MP analyses on the combined generated 270 most 29.2 to 32.3 (CP1+CP2+CP3+CP4), respectively (Fig. 3).
=
parsimonious trees of 648 steps (CI 0.60 excluding au Using these calibrations and the smoothened ML tree,
tapomorphic sites, RI = 0.83) (results not shown). The the dates of the main divergence events within Exacum
1).
ML analysis on the combined sequence data generated are estimated (Table Dating based on maximum se
a = ML
single tree (- In 4350.2744) (Fig. 2). The tree was quence divergences and extreme rates obtained com
congruent with the strict consensus of the combined MP parable figures for corresponding nodes (Table 1). The
MP
trees and was identical to the most likely tree (one divergence between Exacum and its sister group Or
of the 270 MP trees that retained the highest likelihood nichia was estimated to be 9.0 to 39.4 Mya, the diver
the
score when measured with ML criteria). gence between Africa-Madagascar clade ("a") and
in the
As shown Fig. 2, genus Exacum was resolved as the Socotra-Asia clade ("b-c-d") to be about 8.2 to 35.6
to the
monophyletic, with moderate (68%, ML) high (90%, Mya, divergence between the Sri Lanka-South India
MP) bootstrap support. The sister relationship between clade ("b") and Socotra-Indomalesia clade ("c-d") to be
both
Exacum and Ornichia was highly supported (100%, about 7.4 to 31.9 Mya, and the divergence between the
ML and MP). The most basal split within Exacum sep Socotra clade ("c") and the Indomalesia clade ("d") to
the
arated highly supported Africa-Madagascar clade be about 6.5 to 27.7 Mya. The divergence between the
for two
(93% for ML and 99% MP; Fig. 2, clade a) from the Madagascan clades ("al" and //a2,/) is estimated as
the
highly supported Asia-Socotra clade (89% for ML and 5.S to 23.5 Mya, and divergence between the African
2).
85% for MP, Fig. Although all the sampled Mada species E. oldenlandioides and its closest Madagascan ally
gascan species together were monophyletic, they group is estimated to be only about 1.9 to 5.3 Mya. The diver
into two clades: one strongly supported clade (94% for gence between the mainland Southeast Asian endemic
for
ML and 97% MP), composed of the small-flowered E. sutaepense and its sister clade is estimated as 5.5 to
are in
species that mainly found open habitats (clade al 23.7 Mya. The divergence between the Socotra endemic
in a E.
Fig. 2), and weakly supported clade (51% forML and species E. caeruleum and the species affine, that is com
for
61% MP), including the species with larger flowers mon to both Socotra and the southern Arabian penin
found in
that are mainly in forests (clade a2 Fig. 2). The sula, is estimated to be about 2.7 to 11.6 Mya (Fig. 4).
the
African species, ?. oldenlandioides, nested deeply inside Noticeably, none of these estimates dated divergence
the small-flowered Madagascan clade. Within the Asia of Exacum beyond the Eocene. Meanwhile, it is con
is
Socotra lineage, three highly supported clades were spicuous that, whatever calibration used, the diver
d in
present (clades b, c, Fig. 2): a Sri Lanka-south India gences among the main clades "a," "b," "c," and "?,"
28 SYSTEMATIC BIOLOGY VOL. 54

TABLE 1. Estimated timing of divergence based on PL rate-smoothening approach applying different calibration points and molecular clock
approach applying diverse extreme rates of gene evolution.

tree PLa
Smoothened ML applying MC, trnL intron MC, ITS
Divergence of CPlb CP2C CP3d CP4e CP1-CP41 MIN? MAXh MIN1 MAXJ

(Gentianothamnus-Tachiadenus) 20.3-25.9 23.9-53.7 18.2-40.8 12.0-26.4 29.2-32.3 18.2 48.6 47.208.7

vs. (Exacum-Ornichia)
Exacum vs. Ornichia 15.2-19.3 17.9-39.4 13.6-30.1 9.0-19.7 21.7-24.0 14.3 38.2 32.13.4
Clade "a" vs. clade "b-c-d" 13.8-17.5 16.3-35.6 12.4-27.2 8.2-17.9 19.8-21.8 13.4 35.7 33.313.8
Clade "b" vs. clade "c-d" 12.4-15.7 14.6-31.9 11.2-24.5 7.4-16.1 17.8-19.6 15.0 40.0 37.916.0
Clade "c" vs. clade "d" 10.8-13.7 12.8-27.7 9.8-21.3 6.5-14.0 15.5-17.1 14.2 37.9 2912.9.3
Clade "c" 9.3-11.7 10.9-23.7 6.1-18.2 5.5-12.0 13.3-14.6 11.1 29.7 26.28.9
Clade "d" 4.5-5.7 5.3-11.6 4.1-8.9 2.7-5.9 6.5-7.1 2.4 36..03 9.3
Clade "al" vs. //a2,/ 9.7-12.1 11.5-23.5 8.8-18.3 5.8-12.4 13.8-15.0 5.5 14.8 19.8 7.9
E. oldenlandioides vs. 3.0-3.4 3.5-5.3 2.8^.5 1.9-3.6 4.0-4.1 2.4 1.86.35.7
(E.
stenophyllum
E.
quinquenervium) clade
PL, calibration
Divergence times are shown inmillion years before present (Mya). penalized likelihood (Sanderson, 2002a, 2002b); MC, molecularclock; CP, point.
The
divergence of clade "a" versus "b-c-d" in bold face corresponds to the first vicariant event within Exacum inferred from DIVA.
aThe of
range shows the difference divergence estimates obtained from different trnL trees of Gentianaceae.
=
bCPl 60 Mya minimum age of Gentianales based on fossil pollen date (M?ller, 1984).
=
CCP2 40 Mya minimum age of Lisianthius based on fossil pollen date (Graham, 1984).
= and
dCP3 15 estimated age of subtribe Swertiinae based on fossil pollen and geological calibration (Hagen Kadereit, 2001, 2002).
= Mya
and
eCP4 5 minimum age of Gentiana based on fossil seed date (Mai Walther, 1988).
=
Mya

fCPl-CP4 all the four calibration points in effective at the same time.
et
sBased on the highest pairwise sequence distance and fast rate 1.30x 10~9 substitution/site/year (s/s/y) (Richardson al., 2001).
the 10~10 et
hBased on highest pairwise sequence distance and slow rate 4.87 x s/s/y (Richardson al., 2001). and
*Based on the highest Kimura 2-parameter distance and the rate 4.52 x 10~9 s/s/y calibrated for Gentianella (Hagen Kadereit, 2001, 2002).
on the x et
jBased highest pairwise sequence distance and slow rate 1.72 10-9 s/s/y (Richardson al, 2001).

A>Vk j&jfa* #A*i '

?fc^ SJ^sAfjfyf
?r \oN o0-* *v <f"^v 6^ oN~ &~ <?r& *r $> so" <y^<^^ <$ & ^<r ^,7 ^" & <?r <r <sN"*v *v & & <r v ^ v ^ ?o- v * *>
O* ^> <V O' O" <X^ V <</ <<, <(, <(,' <</ <</ <</ <(,- <</V <0- <V <<s-<<,'V^ V <^ <^ V <(,' <<,' <</ <</ <<y<</ <<s <</ <</ <<s <(/ <<,' <<,'<</
BBBBBBBABBBB B BBBBBB B BBCCCCCCCCCGXXYY F DDDDD

$2.7-11.6 Ma

-
CD 6.5 27.7 Ma

7.4-31.9 Ma

A: Africa
B: Madagascar
C: Sri Lanka & South Indra
D: Socotra & S Arabian Peninsula
E: Central & North India
F: Himalaya
G: Mainland Southeast Asia -
H: Malesia & N Australia BG 8.2 35.6 Ma
X:C +E -
Y:C +E+F+G+H
12.0-53.7 Ma (CP)

two-area
FIGURE 4. Ancestral areas reconstructed for each internal node of the phylogeny shown in Figure 2 using DIVA with optimization.
areas are as
"A" through "LI" are the areas of endemism as indicated in the Figure 1. Two
coexisting ancestral indicated double letters. A bar
a area reconstructions for the
cross a clade indicates a dispersal event to new leading by "+." Areas beside slants show the equally possible
nodes. some
nodes are the inferred time for the nodes in million
corresponding Figures companying divergence corresponding expressed years
ago (Mya). "a" (al, a2), "b," "c," and "d" are the main clades discussed in the text. The star marks the divergence dates inferred from the global
analyses of the family Gentianaceae.
 2005_YUAN ET AL.?PHYLOGENY AND BIOGEOGRAPHY OF EXACUM_29

of its is
corresponding to the main areas endemism, seem have topology congruent with the Exacum clade structure
a short which
happened within relatively period of time, of our previous study on the whole tribe Exaceae where
a
suggests relatively rapid and extensive radiation of the fewer Exacum species were included (Yuan et al., 2003).
progenitors of the clades. The present phylogeny of Exacum based on molecular
the
data is also highly congruent with previous hypoth
Dispersal-Vicariance Analysis from
esis inferred morphological and anatomical charac
with the
As anticipated (Ronquist, 1996, 1997), unconstrained ters (Klackenberg, 1985), regard to monophyly
reconstructions by DIVA (not shown) suggested the three and the main infrageneric lineages resolved. However,
basal nodes of the tree, the most recent common ancestor significant difference is shown for the different position
(MRCA) of the outgroup (Gentianothamnus-Tachiadenus) of the Socotran-Arabian clade. In our present molec
and the ingroup (Ornichia-Exacum), the MRCA of Exacum ular phylogeny, the Socotran-Arabian clade showed a
and of to be to
Ornichia, and the MRCA Exacum, widespread closer relationship to the Indomalesia clade than any
in
in the areas "B" through "H" or "B" through "G." The other clade, whereas Klackenberg's (1985) morpho
a
area "A" (Africa) is never postulated as possible ances logical phylogeny, they showed a closer relationship to
the
tral area for these basal nodes. The MRCA of the clades African-Madagascan clades rather than to any Asian
and a
"b," "c," "d" was suggested to be spread in the areas clade: the Socotran-Arabian clade fell on polytomy to
"H."
"C" through "G" or "C" through Thus, a vicariance gether with three other Madagascan-African clades. All
and
between Madagascar (area "B") the Asian areas plus African, Madagascan, and Socotran-Arabian clades to
"C" is a
Socotra (areas through "H") suggested for the first gether formed monophyletic clade, sister to the Asian
splitting of Exacum. clade.

The optimal reconstruction of the unconstrained anal Klackenberg (1985) divided Exacum into two sections,
13 to sect. on his
yses required dispersals (mainly due widespread sect. Exaum and Africana, based morphologi
as E. When cal the
terminal taxa such tetragonum). the maximum phylogeny. The relative length of pedicel and the
number of areas assigned to each node is restricted internode below was used as an important character to
to two, the optimal reconstruction of the ancestral distinguish these two sections. Our present molecular
areas required only two more dispersals than the phylogeny, however, suggests that his section Africana is
unconstrained optimization. The results of the two-area polyphyletic.
in 4.
optimization are shown Figure Madagascar (area
is
"B") consistently assigned to the MRCA of Gentianoth Divergence Dating and Calibration
amnus-Tachiadenus and Ornichia-Exacum and the MRCA Obtaining an absolute time for a divergence/
of Ornichia and Exacum. The constrained optimization speciation offers an elegant and explicit test on whether
in
reveals that Exacum originated Madagascar and the divergence resulted from a given historical event.
specifies an initial dispersal (as indicated by "+C" in Different lines of evidence have indicated that the mini
from
Fig. 4) of the MRCA of Exacum Madagascar (area mum ages of the order Gentianales (53.2 Mya: Magallon
at the 50
"B") to Sri Lanka and south India (area "C") early et al., 1999), family Gentianaceae (ca. Mya: Yuan
et 40 et
diversification stage of the genus. This single dispersal al., 2003), and the tribe Exaceae (ca. Mya: Yuan al.,
resulted in the vicariance pattern between Madagascar 2003) are all well after the Gondwana breakup. Without
and Sri Lanka-south India. Similarly, DIVA specifies the an explicit dating, Klackenberg (2002) speculated that
MRCA of the clades "b" and "c-d" to have dispersed 90
the Gentianaceae might be over Mya old, based on
from Sri Lanka-south India (area "C") to Socotra-Arabia the vicariant distribution of the basal clades of the family
this
(area "D"), and the MRCA of the clade "c" further dis and an assumed correlation between pattern and the
from the area "C" to the mainland Southeast for our
persed Gondwana history. Except previous study where
two
Asia (area "G"). Each of these dispersals is followed only a few Exacum species were sampled (Yuan et al.,
E. of a in the
by a vicariance. The African species, oldenlandioides, is 2003), the absence temporal scale previous phy
also resolved as a vicariant resulted from the dispersal logeny established for Exacum prohibits such a direct test
of the MRCA of E. oldenlandioides and its sister clade A
from (Klackenberg, 1985). temporal calibration of the phy
Madagascar the
to Africa. logenetic tree of genus Exacum remains necessary in
DIVA gives less certain resolution about the ancestral into
order to gain insight biogeographic consequences
areas of the Indomalesia lineage. Although the most re of post-Gondwana vicariance/dispersal events, even if
of E.
cent ancestor sutaepense inmainland Southeast Asia a Gondwana hypothesis can be rejected based on the dat
is confirmed as being dispersed from Sri Lanka-south ing of the order Gentianales (Magallon et al., 1999), the
et
India, multiple possibilities exist regarding the ances family, or the tribe (Yuan al., 2003). Absence of fossil
the
tral areas of widespread species E. tetragonum and data for Exacum and even the tribe Exaceae prohibits a
its closest relatives. Either Sri Lanka-south India, or the direct calibration of the Exacum tree. Fossil records for
Himalayas, or both were involved as possible ancestral the entire family Gentianaceae are scarce. M?ller (1984)
areas. of
estimated the minimum age the order Gentianales as
60
Discussions Mya based on fossil pollen of the other families allied
with the Gentianaceae. Here we used this date as the
Phylogeny of Exacum calibration point CP1; molecular divergence dating sug
Our present ML tree based on combined data was iden gested a close figure for this order (53.2 Mya: Magallon
MP et al., 1999). The earliest megafossil of suggested
tical to the most likely tree (under ML criteria), and
 30_SYSTEMATIC BIOLOGY_VOL. 54

In
Gentianaceae origin was the fossil flowers with Pistil after the breakup of the Gondwana continent. fact,
lipollenites pollen from the Eocene (ca. 45 Mya) of North reconciling dates of fossil data with molecular data,
America. These fossil pollens were suggested to be asso Magallon et al. (1999) showed that few angiosperm lin
the
ciated with the relatively derived Macrocarpaea of the ex eages had evolved at that time, and divergence of the
and
tant Gentianaceae (Crepet Daghlian, 1981), but such order Gentianales (53.2 Mya) was well after Gondwana
a and the first
suggestion was unconfirmed (Stockey Manchester, breakup. Our present PL estimate on divergence
of the the
1986). Struwe et al. (2002) doubted the Gentianaceae ori Exacum, divergence between Madagascan and
gin of these fossil flowers, and Crepet himself recently Indian-Sri Lankan lineages was 8.2 to 35.6 Mya, whereas
also upholds such a consideration (personal communi geological data suggest that the most recent close contact
the
cation). We therefore did not use this fossil record as a between Madagascar-Seychelles and Indian plates
65
calibration. Graham (1984) associated some fossil pol was before Mya (Storey, 1995). If this time frame is
it
lens isolated from lignite deposit developed from man correct, is obviously impossible to interpret the his
40 of
grove swamps of the middle to late Eocene (ca. Mya) tory biogeography of Exacum as a result of Gondwana
in Panama to the extant Lisianthius. We simply used this breakup (Klackenberg, 1985, 2002).
CP2.
fossil record as our second calibration However, the Our Exacum divergence estimates further suggest that
taxonomic attribution of this fossil pollen needs to be fur the main infrageneric lineages diverged rapidly within a
ther confirmed, because, as Graham (1984) admitted, the relatively short period of time, regardless of which cali
fossils differ from pollen of extant Lisianthius in shape, bration point is used. The four lineages "a, b, c, d" corre
and, meanwhile, other associated fossil pollen from the sponding to the main areas of endemism seemed to have
such as
same deposit were mostly mangroves Rhizophora emerged within ca. 8 million years (9.0-6.5 to 35.6-27.7
(15
and Pelliceria. The estimated age Mya) of the subtribe Mya depending on calibration point). If the Gondwana
ITS
Swertiinae was based on sequence divergence of Gen hypothesis has to be rejected, dispersals and extensive
tianella calibrated with both fossil pollen and geological radiations have to be considered as the main causes to
the
evidence (based on the ITS rate of 4.52 ? 2.12 x 10~9 sub explain present-day distribution pattern of Exacum.
stitutions per site per year obtained from an analysis on
and
44 Gentianella species: Hagen Kadereit, 2001, 2002).
Having similar habit and generation time, the rates of Out-of-Madagascar Dispersals
Gentianella may potentially applicable to Exacum. Thus Unconstrained DIVA analysis suggested the basal
we used both the estimated age of Swertiinae and the nodes, MRCA of Gentianothamnus-Tachiadenus and Or
ITS rate to calibrate the Exacum Both gave and
divergence. nichia-Exacum, the MRCA of Exacum Ornichia, and
as
consistent dating results (Table 1). The fossil seeds of the MRCA of Exacum widespread, and the first diver
the of
Gentiana from Upper Pliocene Germany (ca. 5.0 gence within Exacum as the result of a vicariance event
and
Mya: Mai Walther, 1988) seems to be a relatively between Madagascar and Asian areas. As mentioned
recent representative. The lower bonds of the dating re above, such a vicariance without dispersal involvement
sults we obtained are due to this "young" calibration is
physically impossible, because at the time of this
was
point. divergence (8.2 to 35.6 Mya), Madagascar (area "B")
Even though our present dating on the branching already isolated from other areas (Storey, 1995). Ifwe as
events within Exacum was calibrated with inferred dates sume the coexistence of the MRCA of Gentianothamnus
a tree of
based on global Gentianaceae, a Gondwana vi Tachiadenus and Ornichia-Exacum, the MRCA of Exacum
and of in
cariance scenario for the biogeography of Exacum can Ornichia, and the MRCA Exacum, Madagascar,
the
be refuted. The PL dating suggested that divergence Socotra, and other Asian areas, we have to assume ex
of
of Exacum was not beyond the Eocene. Meanwhile, our tinctions Gentianothamnus, Ornichia, and Tachiadenus
is an
present study revealed a close correlation between PL in Socotra and all other Asian areas, which unlikely
dating and divergence rate dating on the estimates of the scenario. Such a resolution is because DIVA considers
divergences within Exacum. Dating results on both trnL wide distribution and vicariance as "default" optimiza
intron and ITS sequences with previously reported fast tion (Ronquist, 1996,1997).
and slow rates (despite rejection of a molecular clock) The constrained optimization of DIVA offers a more
are highly consistent with PL approaches in conclud plausible estimation of the historical biogeography of
ing that the divergence between Exacum and it sister Exacum, particularly with regard to the divergence of
40
group Ornichia was less than Mya (Table 1). There the basal lineages corresponding to the main areas of
4,
is no evidence to assume that Exacum sequences evolve endemism around the IOB. As shown in Fig. the
dramatically slower than other plants, and therefore it constrained DIVA optimization suggested the MRCA
is less likely that the divergence of Exacum was under of Exacum and its sister group, Ornichia, originated in
estimated. Fossil-calibrated NPRS and molecular clocks Madagascar. The constrained DIVA optimization spec
suggested that the divergence between the Gentianoth ified unambiguously four dispersal events within Ex
amnus-Tachiadenus clade and the Exacum-Ornichia clade acum: an initial dispersal of the MRCA of Exacum from
et
fell between 11.2 and 29.8 Mya (Yuan al., 2003). Our Madagascar (area "B") to Sri Lanka and south India (area
present analyses applying PL approaches with addi "C") at stage of the genus
early diversification (when
tional calibrations confirmed our previous results and there was only a single ancestral taxon in Exacum ex

suggested this divergence to be 9.0 to 39.4 Mya, well tant), the dispersal of the MRCA of the clades "b" and
 2005 _YUAN ET AL.?PHYLOGENY AND BIOGEOGRAPHY OF EXACUM_31

"c-d" from Sri Lanka-South India (area "C") to Socotra proposed as the most plausible and common agents of
Arabia (area "D"), the dispersal of the MRCA of the long-distance dispersals for carrying seeds or fruits of
clade "c" from the area "C" to the mainland Southeast continental plants to colonize oceanic islands. For ex
Asia (area "G"), and the dispersal of the MRCA of E. ample, it was suggested that the Hawaiian colonizer
from
oldenlandioides and its sister clade Madagascar to of the genus Viola was dispersed from the Arctic areas
Africa. Each of these dispersal events was followed by by migratory birds (Ballard and Sytsma, 2000). Birds
a vicariance. Because the present dating suggested the might have played roles in transferring Exacum from
first
divergence of Exacum was well after the Gondwana Madagascar to the Sri Lanka-south India area. As soon
these it
breakup, unambiguous vicariance events revealed as Exacum arrived in the Sri Lanka-south Indian area,
by DIVA cannot be the results of tectonic history, but in might have extensively expanded its range by means of
stead the results of single-dispersal events. By the time anemochory dispersals. Although wind dispersal usu
of Exacum divergence (<35.6 Mya), the plates of main ally represents a slow and gradual process of range ex
land
Africa, Madagascar, and Sri Lanka-south India had pansion, it could fulfill a relatively rapid radiation for
mostly attained their present-day positions (ca. 500 km Exacum: although the seeds of Exacum are small and
between mainland Africa and Madagascar, and ca. light, and released from xerochastic capsules by exter
as
3600 km between Madagascar and Sri Lanka-south In nal forces such wind, they might be too large to be
dia), and therefore, these were necessarily long-distance carried far by wind (like orchids). However, many ex
dispersals. tant species of Exacum have prominent wings along the
in
Having originated Madagascar, Exacum has experi ribs of the saclike calyx, and the calyx, particularly the
enced multiple out-of-Madagascar dispersals. The most wings, become conspicuously enlarged and hardened
fruit
important is the long-distance dispersal to Sri Lanka during development (Klackenberg, 2002). Seeds or
south India, which resulted in the extensive radiation capsules entrapped in the calyx sac can quickly travel
the
of the Socotra-Arabia and other Asian lineages in the long distances along ground in an open environment
IOB
northern regions. More recent out-of-Madagascar when caught inwind. The two most widespread species,
the
dispersals include dispersal to the African continent, E. tetragonum and E. oldenlandioides, indicate such a
is in
as represented by the divergence of E. oldenlandioides, capability. The former widespread many open
and to other islands around Madagascar, as indicated by places across the entire Indomalesia region, from south
E.
the distributions of stenopterum, which occurs in both India, the Himalayas, mainland Southeast Asia, to
the
Madagascar and Comoros, and E. quinquenervium, the Philippines, New Guinea, and northern Australia,
which occurs in both Madagascar and the volcanic is whereas the latter is seen in almost the entire tropical
little
land of Mauritius (Klackenberg, 1985). Africa with morphological variation, further sup
These long-distance dispersals could be directly or via porting a rapid and extensive expansion in distribution.
from
stepping-stones. The recent dispersal Madagascar An extensive radiation of Exacum in the northern
(4.7
to Africa Mya), which resulted in the divergence IOB conforms to the paleoclimate variation in these re
of E. oldenlandioides, could have happened directly or gions (deMenocal, 1995; Cerling et al., 1997; Ramstein
et
via the Comoros as stepping-stones (the oldest island, al., 1997; Zachos et al., 2001; Billups and Schr?g, 2002;
As It is to
Mayotte, is 5.5 Mya old; Emerick and Duncan, 1982). Griffin, 2002). plausible speculate that this genus
in
for the dispersal from Madagascar to the Indian sub originated Madagascar when the global temperature
its the
continent, in addition to the possibility of direct dis was rising, and soon after origin, progenitor of Ex
persal, an Eocene-Oligocene land bridge (Steenis, 1962) acum was dispersed to the area of Sri Lanka-south India
or the so-called "Lemurian stepping-stones" (Schatz, by long-distance dispersal. Following the colonization of
1996) connecting India, Sri Lanka, the Seychelles, and this area, this ancestral lineage spread out in the north
Madagascar has been proposed as the main channel for ern IOB regions including the Socotra-Arabia area (by
to
plants and animals disperse across the IOB. In fact, that time the Afro-Arabia plates had attained its connec
to
dispersals through this channel can be considered as tion with Asia but Socotra was still connected Arabia)
stepwise long- or medium-distance dispersals because and diversified at different places as a response to the
there is no geological evidence for a continuous land favorite warm and humid climate during the Eocene or
15
bridge known today (the Seychelles at present are ca. the Miocene climatic optimum (17 to Mya). At the
from time,
1000 km Madagascar and ca. 2600 km from Sri same Madagascan lineages had undergone exten
Lanka-south India) (Schatz, 1996). Further evidence is sive diversification as well. Subsequently, following the
a
needed to confirm if such dispersal channel has ever spreading Neogene drought, particularly towards the
the
existed. Nevertheless, with regard to Exacum there is no end of Miocene, due to a decrease of atmospheric
et
reason to exclude the possibility of dispersals via the CO2 (Cerling al., 1997), a global cooling linked to
a
"Lemurian stepping-stones," although no extant Exacum the reestablishment of major ice sheet over Antarctica
e
species is found on the Seychelles (Klackenberg, 1985, (Zachos al., 2001; Billups and Schr?g, 2002) and the dry
2002). ing up in the northern IOB regions induced mainly by
the the
Yet process and mechanism of these long-distance the rapid uplifting of Himalayas (deMenocal, 1995;
et
dispersals are not well understood. Without apparent Ramstein al., 1997; Griffin, 2002), global vegetation ex
or the
epizoic hydrozoic adaptations, possible means of perienced a drastic change (e.g., widespread increase of
dispersal of Exacum are wind and birds. Birds have been C4 biomass in favor of the dry climate). Consequently,
 54
32_SYSTEMATIC BIOLOGY_VOL.

Exacum species were forced to retreat from most parts Acknowledgments

of the northern IOB regions and survived in isolation in We thank the Botanical and Zoological Park of Tsimbazaza and
Socotra, south India-Sri Lanka, and perhaps mainland Association Nationale pour la Gestion des Aires Prot?g?es (Antana
The narivo, Madagascar), Louis Zeltner, Tony Miller, and Philippe Chassot
Southeast Asia. Madagascan lineage was probably
for in
less drastically affected. From these core areas, "sec help samplecollections, and Jason Grant for critical reading of the
diversifications and radiations occurred later manuscript. We are much indebted to anonymous reviewers for criti
ondary" was
cal comments. This study financially supported by Swiss National
on, which resulted in several relatively distantly re
Science Foundation (Grant 3100-052885).
lated clusters of species?the main clades we observe
A the
today. similar pattern has been suggested for genus
Nepenthes (Meimberg et al., 2001). It is thought that Ne References
penthes has originated in the north of the ancient Tethys J.
Ballard, H. E. Jr., and K. Sytsma. 2000. Evolution and biogeography of
sea or on the Indian subcontinent after itwas separated the
woody Hawaiian violets (Viola, Violaceae): Arctic origins, herba
from
Madagascar. This progenitor is supposed to have ceous ancestry and bird dispersal. Evolution 54:1521-1532.

undergone extensive spreading: on land migration to Barron, E. J. 1987. Cretaceous plate tectonic reconstruction. Paleogeogr.
Paleoclimatol. Paleoecol. 59:3-29.
Southeast Asia, and dispersals through land bridge or
and Baum, D. A., R. L. Small, and J. F. Wendel. 1998. Biogeography and
stepping-stones to the Seychelles Madagascar. This floral evolution of Baobabs (Adansonia, Bombacaceae) as inferred
habitat expansion was followed by incisive extinctions from multiple data sets. Syst. Biol. 47:181-207.
the R. K.
of widespread progenitors in the paleotropical re Beck,
R.
A., D. W. Burgen, W. J. Sercombe, G. W. Riley, J.
J.
Barndt,
J. Berry, J. Afzal, A. M. Khan, H. J?rgen, Metje, A. Cheema, N.
gions (i.e., fragmentation). Subsequently, secondary di
A. Shafique, R. D. Lawrence, and M. A. Khan. 1995. Stratigraphie
versification occurred on the Southeast Asian islands, evidence for an early collision between northwest India and Asia.
whereas the species on the Seychelles and Madagas Nature 373:55-58.
V P. M.
car remained relatively unchanged (Meimberg et al., Besse, J. Courtillot, J. Pozzi, Westphal, and Y. X. Zhou. 1984. Pa
in the
2001). leomagnetic estimates of crustal shortening Himalayan thrusts
The collision of the Afro-Arabian and Zangbo suture. Nature 311: 621-626.
plate with Asia
of Billups, K., and D. P. Schr?g. 2002. Paleotemperatures and ice vol
opened another channel migration for flora and fauna 27
ume of the past Myr revisited with paired Mg/Ca and 0-18/0
between Africa and Asia since the Miocene (Samuei 16 measurements on benthic foraminifera. Paleoceanography 17:
et a land 1-11.
al., 1997). Furthermore, possible bridge between W. F.
and Birse, A. C. R., Bott, J.Morrison, and M. A. Samuel. 1997. The
mainland Africa Madagascar due to a temporal dry
the Mesozoic and Early Tertiary tectonic evolution of the Socotra area,
out in large areas of Mozambique Channel between eastern Gulf
of
Aden, Yemen. Mar. Petr. Geol. 14:675-684.
26 a T. M.
45 and Mya has been suggested as possible chan Cerling, E., J. Harris, B. J.MacFadden, M. G. Leakey, J. Quade,
V. R.
nel for mammals to colonize Madagascar (McCall, 1997). Eisenmann, and J. Ehleringer. 1997. Global vegetation change
the through
the
Miocene/Pliocene boundary. Nature 389:153-158.
Our present reconstructions of phylogeny and an
Coffin, M. F., and P. D. Rabinowitz. 1988. Evolution of the conjugate
cestral areas, however, do not offer any support for a East African-Madagascan and the western Somali Basin.
from margins
dispersal of Exacum Madagascar to the Socotra Geo. Soc. Am. Spec. Pap. 226:1-78.
Arabian area via mainland Africa and then to India-Sri P.
Crepet, W. L., and C. Daghlian. 1981. Lower Eocene and Paleocene
Lanka, because our phylogeny suggested that the Sri Gentianaceae: Floral and palynological evidence. Science 214:75
77.
Lanka-south India lineage diverged prior to the diver
Darlu, P., and G. Lecointre. 2002. When does the incongruence length
gence between the Socotra-Arabian and Indomalesian difference test fail? Mol. Biol. Evol. 19:432-437.
clades. Moreover, such a consideration has to assume Davis, C. C, C. D. Bell, P. W. Fritsch, and S. Mathews. 2002a. Phy
a of
complete extinction of the first African colonizers on logeny Acridocarpus-Brachlophon (Malpighiaceae): Implications
the entire African continent. The widespread African for Tertiary tropical floras and Afroasian biogeography. Evolution
56:2395-2405.
species E. oldenlandioides is firmly shown to have origi S. J.
from Davis, C. C, C. D. Bell, Mathews, and M. Donoghue. 2002b.
nated from a recent dispersal Madagascar. The only Laurasian migration explains Gondwana disjunctions: Evidence
E. from
other extant African species, zombense, endemic to the Malpighiaceae. Proc. Nati. Acad. Sei. USA 99:6833-6837.
of
Zomba mountain Malawi, was unavailable for our deMenocal, P. B.1995. Pliopleistocene African climate. Science 270:53
59.
present phylogenetic assessment. However, morpholog R. L. L.
Dolphin, K., Belshaw, C. D. Orme, and D. J.Quicke. 2000. Noise
ically it is similar to the recently dispersed E. oldenlan and incongruence: Interpreting results of the inconguence length
dioides (Klackenberg, 1985) and unlikely represents the difference test. Mol. Phylog. Evol. 17:401-406.
relict of the first African colonizers. Donoghue, M. J., C. D. Bell, and J. Li.
2001. Phylogenetic patterns in
The present article has given an interesting insight northern hemisphere Int. J. Plant Sei. 162-.S41-S52.
plant geography.
into the evolution and biogeographic Dowton, M., and A. D. Austin. 2002. Increased congruenc does not
history of the indicate increased phylogenetic behavior
necessarily accuracy?The
genus Exacum and sheds lights on the noticeable bio of the incongruence length difference test in mixed-model analyses.
the
geographic patterns of paleotropical regions around Syst. Biol. 51:19-31.
IOB. in C.
the Despite some uncertainty dating caused by Emerick, M., and R. A. Duncan. 1982. Age progressive volcanism in
to the Comoros Archipelago, western Indian Ocean and implications
the scarcity of reliable fossils, we were able establish,
a for Somali plate tectonics. Earth Planet. Lett. 60:415-428.
to the best of our knowledge, geographical and tem Fantozzi,
P.
L., and M. Sgavetti. 1998. Tectonic and sedimentary evo
poral origin of Exacum and older lineages therein. The lution of the eastern Gulf of Aden continental margins: New struc
evolution history of Exacum apparently involved a com tral and stratigraphie data from Somalia and Yemen. Pages 56-76 in
Sedimentation and tectonics in rift basins: Red Sea?Gulf of Aden.
bination of long-distance dispersal events and extensive H.
(B. Purser, and D. W. J. Bosence, eds.). Chapman & Hall,
radiations.
London.
 2005_YUAN ET AL.?PHYLOGENY AND BIOGEOGRAPHY OF EXACUM_33

M.
Farris, J. S., K?llersj?, A. G. Kluge, and C. Bult. 1995. Testing signif Raxworthy, C. R., M. R. J. Forstner, and R. A. Nussbaum. 2002.
icance of incongruence. Cladistics 10:315-319. Chameleon radiation by oceanic dispersal. Nature 415:784-787.
S. G. K.
Felsenstein, J. 1985. Confidence limits on phylogenies: An approach Renner, S., Clausing, and Meyer. 2001. Historical biogeography
the of
using bootstrap. Evolution 39:783-791. of Melastomaceae: The roles Tertiary migration and long-distance
Ghebreab, W. 1998. Tectonics of the Red Sea region reassessed. Earth dispersal. Am. J. Bot. 88:1290-1300.
Sei. Rev. 45:1-44. S. K.
Renner, S., and Meyer. 2001. Melastomeae come full circle: Bio
Graham, A. 1984. Lisianthius pollen from the Eocene of Panama. Ann. geographic reconstruction and molecular clock dating. Evolution
Missouri Bot. Gard. 71:987-993. 55:1315-1324.
S. F.
Griffin, D. L. 2002. Aridity and humidity: Two aspects of the late Richardson, M., W.
Bott, B. A. Smith, W. D. Hollar, and P. M.
Miocene climate of North Africa and the Mediterranean. Paleogeogr. Bermingham. 1995. A
new hydrocarbon play area offshore Socotra
Paleoclimatol. Paleoecol. 182:65-91. Island, Republic of Yemen. J. Petr. Geol. 18:5-28.
K. B.
von, and W. Kadereit. 2001. The of Gentianella Richardson, J. E., R. T. Pennington, T. D. Pennington, and P. M.
Hagen, J. phylogeny
(Gentianaceae) and its colonization of the southern hemisphere as Hollingsworth. 2001. Rapid diversification of a species rich genus
of
revealed by nuclear and chloroplast DNA sequence variation. Org. Neotropical rain forest trees. Science 293:2242-2245.
Divers. Evol. 1:61-79. Rieppel, O. 2002. A case of dispersing chameleons. Nature 415:744
W. 745.
Hagen, K. B. von, and J. Kadereit. 2002. Phylogeny and flower 1.1.
evolution of the Swertiinae (Gentianaceae-Gentianeae): Homoplasy Ronquist, F. 1996. DIVA version Computer program and manual
and the principle of variable proportions. Syst. Bot. 27:548 available by anonymous FTP from Uppsala University (ftp.uu.se or
572. ftp.systbot.uu.se)
V. P. F. 1997.
Jaeger, J., Courtillot, and Tapponnier. 1989. Paleontological view Ronquist, Dispersal-vicariance analysis: A new approach to
of the ages of the Deccan traps, the Cretaceous Tertiary boundary, the quantification of historical biogeography. Syst. Biol. 46:195
and the India-Asia collision. Geology 17:316-319. 203.
R.
Jolivet, L., and C. Faccenna. 2000. Mediterranean extension and the Samuei, M. A., N. Harbury, Bott,and A. M. Thabet. 1997. Field ob
Africa-Eurasia collision. Tectonics 19:1095-1106. servations from the Socotran platform: Their interpretation and cor
Klackenberg, J. 1985. The genus Exacum (Gentianaceae). Opera Bot. relation to Southern Oman. Mar. Petr. Geol. 14:661-673.
84:1-144. Sanderson, M. J. 1997. A nonparametric approach to estimating diver
in
Klackenberg, J. 2002. Tribe Exaceae. Pages 66-108 Gentianaceae, sys gence times in the absence of rate constancy. Mol. Biol. Evol. 14:1218
tematic and natural history (L. Struwe, and V. A. Albert, eds.). 1231.
Cambridge University Press, Cambridge. Sanderson, M. J. 2002a. Estimating absolute rates of molecular evo
C. P. L. S. W. and G. M. Smith. lution and time: A penalized likelihood Mol.
Klootwijk, T., McFadden, J. Gee, J. Peirce, divergence approach.
1992. An early India-Asia contact?Paleomagnetic constraints from Biol. Evol. 19:101-109.
121.
Ninetyeast Ridge, ODP LEG Geology 20:395-398. Sanderson, M. J. 2002b. r8s, version 1.50 user's manual. Available via
Lee, T.-Y, and L. A. Lawver. 1995. Cenozoic plate reconstruction of the http://ginger.ucdavis.edu/r8s/.
southeast Asia 251:85-138. and F.
region. Tectonophysics Sanmartin, I., Ronquist. 2004. Southern hemisphere biogeogra
Z.
Li, X., and C. M. Powell. 2001. An outline of the palaeogeographic phy inferred by event-based models: Plant versus animal patterns.
the
evolution of the Australasian region since beginning of the Neo Syst. Biol. 53:216-243.
proterozoic. Earth Sei. Rev. 53:237-277. Schatz, G. E. 1996. Malagasy/Indo-Australo-Malesian phytogeo
P. R. in of
Magallon, S., Crane, and P. S. Herendeen. 1999. Phylogenetic pat graphic connections. Pages 73-83 Biogeography Madagascar
R.
tern, diversity, and diversification of eudicots. Ann. Missouri Bot. (W. Lourenco, ed.). ORSTOM editions, Paris.Steenis, C. G. G. J.
Gard. 86:297-372. in
van. 1962. The land-bridge theory botany. Blumea 11:235-372.
D. R.
Mai, H., and H. Walther. 1988. Die pliozaenen Floren von Thuerin Stockey, A., and S. R. Manchester. 1986. A fossil flower with in situ
gen, Deutsche Demokratische Republik. Quartaerpalaeontologie Pistillipollenites from the Eocene of British Columbia. Can. J. Bot.
7:55-297. 66:313-318.
McCall, R. A. 1997.
Implications of recent geological investigations of Storey, B. C. 1995. The role of mantle plumes in continental breakup:
the
Mozambique Channel for the mammalian colonization of Mada Case histories from Gondwanaland. Nature 377:301-308.
Proc. R. Soc. Lond. B 264:663-665.
J. J. S. P.
gascar. Storey, M., Mahoney, A. D. Saunders, R. A. Duncan, Kelley,
S. 2001. The breakup 1995.
McLoughlin, history of Gondwana and its im and M. F. Coffin. Timing of hot spot-related volcanism and the
of
pact on pre-Cenozoic floristic provincialism. Aust. J. Bot. 49:271 breakup Madagascar and India. Science 267:852-855.
J.
300.
A. P. Struwe, L., J. W. Kadereit, Klackenberg, S. Nilsson, M. Thiv, K. B. von
Meimberg, H., Wistuba, Dittrich, and G. Heubl. 2001. Molecular Hagen, and V A. Albert. 2002. Systematics, character evolution, and
of
phylogeny of Nepenthaceae based on cladistic analysis of plastid biogeography Gentianaceae, including a new tribal and subtribal
trriK intron data. Plant Biol. 3:164-175. in
sequence classification. Page 21-309 Gentianaceae?Systematics and natural
for
Muller, J. 1984. Significance of fossil pollen angiosperm history. Ann. history (L. Struwe and V. A. Albert, eds.). Cambridge University
Missouri Bot. Gard. 71:419-443. Press, Cambridge.
S.
Muse, V, and B. S. Weir. 1992. Testing for equality of evolutionary Swofford, D. L. 2000. PAUP* phylogenetic analysis using parsimony
rates. Genetics 132:269-272. (*and other methods). Version 4. Sinauer Associates, Sunderland,
Patriat, P., and J.Achache. 1984. India Eurasia collision chronology has Massachusetts.
T. F. F.
implications for crustal shortening and driving mechanism of plates. Thompson, J. D., J. Gibson, Plewniak, Jeanmougin, and D. G.
Nature 311:615-621. Higgins. 1997. The Clustal X windows interface: Flexible strategies
G. P. P. for
Plunkett, M., Lowry, and M. K. Burke. 2001. The phylogenetic multiple sequence alignment aided by quality analysis tools. Nu
status of Poly scias (Araliaceae) based on nuclear its sequence data. cleic Acids Res. 24:4876-4882.
Ann. Missouri Bot. Gard. 88:213-230. Thulin, M. 2001. Exacum (Gentianaceae) on the Arabian peninsula and
Posada, D., and K. A. Crandall. 1998. Modeltest: Testing the model of Socotra. Nord. J. Bot. 21:243-247.
DNA substitution. Bioinformatics 14:817-818. Valdiya, K. S. 2002. Emergence and evolution of Himalaya: Recon
P. F.
Rabino witz, D., M. Coffin, and D. Falvey. 1983. The separation of structing history in the light of recent studies. Prog. Phys. Geogr.
Madagascar and Africa. Science 220:67-69. 26:360-399.
F. J.
Ramstein, G., Fluteau, J. Besse, and S. Joussaume. 1997. Effect of Vences, M., Freyhof, R. Sonnenberg, J. Kosuch, and M. Veith. 2001.
orogeny, plate motion and land sea distribution on Eurasian climate Reconciling fossils and molecules: Cenozoic divergence of cichlid
of J.
change over the past 30 million. Nature 386:788-795. fishes and the biogeography Madagascar. Biogeogr. 28:1091
P.
Raven, H., and D. I. Axelrod. 1974. Angiosperm biogeography 1099.
and past continental movements. Ann. Missouri Bot. Gard. 61:539
1996. on
Yang, Z. Among-site rate variation and its impact phylogenetic
673. Trends Ecol. Evol. 11:367-372.
analysis.
 54
34_SYSTEMATIC BIOLOGY_VOL.

J. A. M. L. E. and K. 2001.
Yoder, A. D., Irwin, and B. A. Payseur. 2001. Failure of the ILD Zachos, J., Pagani, Sloan, Thomas, Billups. Trends,
data in 65 Mato
to determine combinability for slow loris phylogeny. Syst. Biol. rhythms, and aberrations global mate present. Science
50:408^24. 292:686-693.
S. M. P. G.
Yuan, Wohlhauser,
Y.-M., M?ller, Chassot, Mansion, J.Grant,
P. J.
Kupfer, and Klackenberg. 2003. Monophyly and relationships 8 22
of the tribe Exaceae (Gentianaceae) inferred from nuclear riboso First submitted January 2003; reviews returned April 2003;
10
mal and chloroplast DNA sequences. Mol. Phylog. Evol. 28:500 final acceptance August 2004
517. Associate Editor: Roberta Mason-Gamer

Appendix 1. Sources of material used for the Exacum molecular phylogeny. Herbaria are abbreviated as follows:E, RoyalBotanic Garden
NEU, of
Edinburgh; University Neuch?tel; S, Swedish Museum of NaturalHistory Stockholm. Sequences newly acquired are marked withan
asterisk.

GenBank accessions

Species Voucher specimen and herbarium0 Origin ITS trnL intron

Exacum affine I. B. Balf. ex Regel?M8238a M & al. 8238a, E Socotra AJ489877 AJ490202
E. E. I. B.
affine I. B. Balf. ex Regel?M17126 (= gracilipes Balf.) M & al. 17126, E
Socotra AJ489886 AJ490211
E. ex
affine I. B. Balf. Regel?M6201 Oman AJ489879* AJ490204*
E. M & al. 6201, E
affine I. B. Balf.
ex Regel?Wcff5 NEU Socotra AJ489878* AJ490203*
E. W cff5,
atropurpureum Bedd. India AJ489880* AJ490205*
E. K&L526,S AJ489881* AJ490206*
bulbilliferum Baker Madagascar
E. caeruleum I. B. Balf. W & C M070, NEU Socotra AJ489882 AJ490207
E. dolichantherum Klack. M & al. 11356, E Madagascar AJ489883* AJ490208*
E.
E.
exiguum Klack. W & C M064, NEU Madagascar AJ489884* AJ490209*
fruticosum Humbert W&Z NEU Madagascar AJ489885 AJ490210
Pi, M048,
E. hamiltonii G. Don Bhutan AJ489887 AJ490212
E. humbertii Klack.
W & PM055, NEU AJ489888* AJ490213*
Wo E Madagascar
E. intermedium Klack. 7477,
Madagascar AJ489889* AJ490214*
W & PM052, NEU
E. linearifolium (Humbert) Klack. W & LaM060, NEU Madagascar AJ489890* AJ490215*
E. macranthum Arn. ex Griseb.?FK767 Sri Lanka AJ489892* AJ490217*
M &R 6254,S
E. macranthum Arn. ex Griseb.?Zsl003 Z Sri Lanka AJ489891* AJ490216*
E. F& K 767, S AJ489893 AJ490218
marojejyenseHumbert NEU Madagascar
E. sl003,
microcarpum Klack. Madagascar AJ489894* AJ490219*
E. millotii Humbert
W & PM056, NEU AJ489895* AJ490220*
Madagascar
E. W & LaM061, NEU AJ489896 AJ490221
nummularifolium Humbert Madagascar
W & PM057, NEU Burundi AJ489897 AJ490222
E. oldenlandioides (S.Moore) Klack.
E. W & PM058, NEU Sri Lanka AJ489898* AJ490223*
pallidum (Trim.) Klack. Re S
E.
pedunculatum L.
9275,
Sri Lanka AJ489899* AJ490224*
E. F& K 777, S AJ489900 AJ490225
quinquenervium Griseb. Madagascar
B, Ke & T 4, S
India AJ489901* AJ490226*
E. sessile L. NEU
W M063, AJ489902 AJ490227
E. stenophyllum Klack. Madagascar
K&L349,S AJ489903* AJ490228*
E. subacaule Humbert Pi,W & Z M049,
E. subieres Klack. S
NEU Madagascar
AJ489904* AJ490229*
M 3755, Madagascar
E. subverticillatum Humbert Madagascar AJ489905* AJ490230*
E. W & PM053, NEU Thailand AJ489906* AJ490231*
sutaepense Hosseus ex Craib Madagascar, s. n., NEU
E. tetragonum Roxb.?K254 -
Nepal AJ489908* AJ490233*
E. tetragonum Roxb.?LK332
Ch 99 230, NEU India AJ489907 AJ490232
Keke 254, E Sri Lanka AJ489910* AJ490235*
E. trinervium (L.) Druce?ZslOOl
L& K 332, S Sri Lanka AJ489909 AJ490234
E. trinervium (L.) Druce?Zsl002 NEU Z
E. walkeri Arn. ex Griseb.?K539 Z slOOl,
NEU K
Sri Lanka AJ489911* AJ490236*
E. walkeri Arn. ex Griseb.?Zsl004 sl002, 539, Sri Lanka AJ489912* AJ490237*
S
E. India AJ489913 AJ490238
wightianum Arn. Z NEU
Gentianothamnus Humbert sl004, K Madagascar
AJ489914 AJ490240
madagascariensis & L 188,S G
NEU Madagascar
AJ489917 AJ490242
Ornichia madagascariensis (Baker) Klack.
O. trinervis (Desrousseaux) Klack.
G020,
Madagascar
AJ489918 AJ490243
W NEU
Tachiadenus carinatus (Desrousseaux) Griseb. M002, Madagascar
AJ489923 AJ490249
C s. n., NEU
NEU
AJ489924
T. longiflorus Bojer ex Griseb.
W M059,
Madagascar

NEU
W M006, AJ490250
a of the collectors: B = C =Callmander;
Ch = Chassot; F = G= = = L= La = Laivao
Abbreviation Bremer; Fagerlind; Gautier; K Klackenberg; Ke Kerr; Lundin;
= = = :: Re = T= W = = =
M Miller; P Pfund; Pi Piso; R Randrianasolo; Reekmans; Theran; Wohlhauser; Wo Wood; Z Zeltner.