MOLECULAR PHYLOGENETICS AND EVOLUTION Molecular Phylogenetics and Evolution 28 (2003) 500–517 www.elsevier.com/locate/ympev Monophyly and relationships of the tribe Exaceae (Gentianaceae) inferred from nuclear ribosomal and chloroplast DNA sequences €ller,b Philippe Chassot,a Yong-Ming Yuan,a,d,* S
ebastien Wohlhauser,a Michael Mo a a a Guilhem Mansion, Jason Grant, Philippe K€ upfer, and Jens Klackenbergc a Laboratory of Evolutionary Botany, Institute of Botany, University of Neuch^ atel, Emile-Argand 11, Neuch^ atel CH-2007, Switzerland b Royal Botanic Garden Edinburg, Edinburg EH3 5LR, Scotland, UK c Department of Phanerogamic Botany, Swedish Museum of Natural History, S-10405 Stockholm, Sweden d South China Institute of Botany, Chinese Academy of Sciences, Guangzhou, PR China Received 27 June 2002; revised 30 January 2003 Abstract Both chloroplast trnL (UAA) intron and nuclear ribosomal ITS sequences highly confirmed the monophyly of the tribes of the Gentianaceae defined by the recent classification, and revealed the tribe Exaceae as a basal clade just next to the basal-most lineage, the tribe Saccifolieae. Within the tribe Exaceae, Sebaea (except Sebaea madagascariensis) appeared as the most basal clade as the sister group to the rest of the tribe. The Madagascan endemic genera Gentianothamnus and Tachiadenus were very closely related to each other, together standing as sister to a clade comprising Sebaea madagascariensis, Ornichia, and Exacum. The saprophytic genus Cotylanthera nested deeply inside Exacum. Sebaea madagascariensis was shown closer to the Madagascan endemic genus Ornichia than to any other sampled Sebaea species. Exacum appeared as the most derived taxon within this tribe. The topology of the phylogenetic trees conform with the Gondwana vicariance hypothesis regarding the biogeography of Exaceae. However, no evidence for matching the older relationships within the family to the tectonic history could be corroborated with various divergence time analyses. Divergence dating estimated a post-Gondwana diverging of the Gentianaceae about 50 million years ago (MYA), and the tribe Exaceae as about 40 MYA. The Mozambique Channel land-bridge could have played an important role in the biogeographic history of the tribe Exaceae. Ó 2003 Elsevier Science (USA). All rights reserved. 1. Introduction The Gentianaceae are a widespread family of 87 genera and about 1650 species (Struwe et al., 2002), that shows extensive morphological diversification and interesting distribution patterns. Phylogenetic studies on different infrafamilial lineages of the family using both molecular and non-molecular approaches have achieved important progress during the last decade (Adams, 1995; Albert and Struwe, 1997; Chassot et al., 2001; Chen et al., 2000a,b; Gielly and Taberlet, 1996; Gielly et al., 1996; Hagen and Kadereit, 2000, 2001; Ho and Liu, 1990; Ho et al., 1994; Ho and Pringle, 1995; Hungerer * Corresponding author. Fax: +41-32-718-3001. E-mail address: yong-ming.yuan@unine.ch (Y.-M. Yuan). and Kadereit, 1998; Liu and Ho, 1992; M
eszaros, 1994; M
esz
aros et al., 1996; Struwe and Albert, 1997, 1998a,b; Struwe et al., 1994, 1997, 1998, 2002; Thiv et al., 1999a,b; Yuan and K€ upfer, 1995, 1997; Yuan et al., 1996). However, the only available comprehensive infrafamilial classification of the family was that of GilgÕs established a century ago based mainly on pollen and gross morphology (Gilg, 1895), until the recent revision of the classification proposed by Struwe et al. (2002) based principally on the phylogeny inferred from chloroplast trnL (UAA) intron and matK gene sequences. Our present studies were carried out to verify this new classification by using both nuclear and chloroplast DNA evidence, and further to evaluate the biogeography and phylogenetic relationships within the basal clade, the tribe Exaceae. 1055-7903/$ - see front matter Ó 2003 Elsevier Science (USA). All rights reserved. doi:10.1016/S1055-7903(03)00068-X Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 In the new classification of Struwe et al. (2002), the Gentianaceae were divided into six tribes. The Exaceae represents a small tribe with six genera and 144–184 species: Cotylanthera Blume, Exacum L., Gentianothamnus Humbert, Ornichia Klack., Sebaea Sol. ex R. Br. and Tachiadenus Griseb. Cotylanthera has four achlorophyllous saprophytic species found from the eastern Himalayas to Southeast Asia. Gilg (1895) included this genus within his subtribe Exacinae. Klackenberg (2002) accepted the inclusion of the genus within tribe Exaceae, and further suggested a close relationship with Exacum or as a derived lineage inside Exacum based on floral morphology. The genera Sebaea (ca. 150 sp., Boutique, 1972) and Exacum (64 sp., Klackenberg, 1985; Thulin, 2001) make up the majority of species in the tribe Exaceae. Sebaea is mainly found in South Africa, although a few species are distributed in tropical Africa, Madagascar and Indo-Malaysian areas. The taxonomy of this genus is still confusing, and no updated taxonomic revision is yet available. Exacum has two main centres of diversity, Madagascar and the area including Southern India and Sri Lanka, with only a few species occurring on Socotra (and the Arabian Peninsula), in the Himalayas, the Southeast Asia, New Guinea, and in the extreme northern Australia. Klackenberg (1985) monographed the genus, and proposed a phylogeny based on morphological and anatomical characters. The remaining three genera of the tribe, Gentianothamnus (1 sp.), Tachiadenus (11 sp.), and Ornichia (3 sp.) are all endemic to Madagascar. The monotype Gentianothamnus was originally included in the subtribe Chironiinae by Humbert (1937). Klackenberg (2002) transferred it to Exaceae and suggested its close relationship with Tachiadenus based on morphological and anatomical characters. Tachiadenus was originally included in the subtribe Tachiinae by Gilg (1895). Klackenberg (1987) revised the genus and transferred it to Exaceae. Ornichia was recognized as a separate genus and placed in Exaceae by Klackenberg (1986) based on species previously included in Chironia and Tachiadenus. In addition to taxonomic revisions, Klackenberg (1985, 1987) also performed manual cladistic analyses of morphological and anatomical characters on Exacum and Tachiadenus. Klackenberg (1985) further constructed an area cladogram of Exacum based on his species cladogram, and concluded that Africa and Madagascar were phytogeographically more closely related to each other than to any other area, and Socotra was more closely related to Madagascar than to India. He suggested that the African species might have originated from Madagascan species by dispersal, but that Indian species represented old relicts that originated from Madagascar by vicariance events. He further speculated that the Socotran species might have originated from Madagascan ancestors through vicariance 501 events, and had no contact with Indian species. Obviously, identifying the correct species phylogeny is crucial in evaluating such considerations. Only six species representing four genera of the tribe Exaceae had been subjected to a molecular phylogenetic study using trnL (UAA) intron sequences (Thiv et al., 1999a). Two genera, Cotylanthera and Gentianothamnus, had not yet been sampled. Although the monophyly of the tribe was supported with regard to the sampled genera, this insufficiently sampled analysis (with regards to Exaceae) could not offer strong support for relationships among the genera of the tribe. The inferred phylogenies, therefore, could not provide deep insights into the historical biogeography of the tribe. By sequencing both nuclear ribosomal internal transcribed spacers (ITS1 and ITS2) and the chloroplast trnL (UAA) intron, we conducted a more comprehensive molecular phylogenetic studies on this tribe by sampling all genera and additional species representing different clades and distribution areas. A detailed analysis on the genus Exacum is to be presented separately. Here we present our phylogenetic analysis on all genera of the tribe Exaceae together with representatives from all other clades of the family as recognized in the current classification. The following questions are to be addressed in particular: (1) Is the phylogeny based on nuclear ribosomal DNA sequences congruent with those based on chloroplast DNA sequences? (2) Is the tribe Exaceae as defined by current classification monophyletic? (3) What are the relationships of Exaceae to other tribes of the family? (4) What are the relationships among the genera of Exaceae? (5) What are the biogeographic implications of the molecular phylogenies? 2. Materials and methods 2.1. Ingroup sampling and outgroup choice As Madagascar has the largest diversity of the tribe Exaceae at generic level, sampling of Madagascan species was maximized to cover as many species available to us as possible. Species representing all other tribes or other distribution areas were also sampled when available. The species included in this study and their voucher information are listed in Table 1. Those sequences retrieved directly from the GenBank database are indicated with asterisks following their accession numbers. A total of 60 operational taxonomic units (OTUs) were included in this study as ingroup taxa. Seven species representing all other families of the order Gentianales (APG, 1998; Backlund et al., 2000; Struwe et al., 1994) were included as outgroups. They were Coffea arabica L. and Erithalis fruticosa L. of Rubiaceae, Labordia tinifolia A. Gray and Mitreola petiolata (Walt.) Torr. and Gray of Loganiaceae, Gelsemium sempervirens Ait. of 502 Table 1 Origin of plant material, voucher information and EMBL/GenBank accession number of sequence Voucher Origin GenBank Accession No. ITS1 ITS2 trnL (UAA) Anthocleista amplexicaulis Baker S. Wohlhauser PBZT AJ489863 AJ489863 AJ490189 Present study Anthocleista grandiflora Gilg Canscora alata (Roth) Wallich M. Callmander s.n J.C. Piso, S. Wohlhauser and L. Zeltner M024 P. Chassot 99–234 Tsimbazaza Botanical and Zoological Park, Antananarivo, Madagascar Region of Maroantsetra, Madagascar Mangindrano, Mahajanga, Madagascar Between Chiangdao and Chiang Mai Thailand Doi Inthanon, Chomtong, Thailand AJ489864 AJ489865 AJ489864 AJ489865 AJ490190 AJ490191 Present study Present study AJ489866 AJ489866 AJ490192 Present study AJ489867 AJ489867 AJ490193 Present study 23.27.15 N/105.49.91 W-Mexico AY047742* AY047827* AF402166* Mansion (to be published) 30.41.030 N/06.16.150 E-Morocco 38.41.58 N/122.35.87 W-California, USA Near Zamora, Ecuador Loja, Ecuador Sawaneh, Venezuela AY047792* AY047710* AY402197* AY047795* AF402253* AF402199* Mansion (to be published) Mansion (to be published) AJ489868 AJ489869 AJ489870 AJ489868 AJ489869 AJ489870 AJ490194 AJ490195 AJ490196 Present study Present study Present study AJ489871 AJ489871 AJ490197 Present study AJ489872 AJ489872 AJ490198 Present study AJ489873 AJ489873 AJ490199 Present study AJ489874 AJ242613* AJ489875 AJ489874 AJ242614* AJ489875 NA AJ242606* AJ490200 Present study Thiv et al. (1999a) Present study AJ489876 AJ489877 AJ489882 AJ489885 AJ489876 AJ489877 AJ489882 AJ489885 AJ490201 AJ490202 AJ490207 AJ490210 Present Present Present Present AJ489886 AJ489887 AJ489893 AJ489886 AJ489887 AJ489893 AJ490211 AJ490212 AJ490218 Present study Present study Present study AJ489896 AJ489896 AJ490221 Present study AJ489897 AJ489897 AJ490222 Present study AJ489900 AJ489900 AJ490225 Present study Canscora andrographioides Griff, ex C.B. Clarke Canscora diffusa (Vahl) Roem. and Schult. Centaurium madrense B.L. Robinson Centaurium spicatum (L.) Fritsch Centaurium trichanthum B.L. Robinson Chelonanthus alatus (Aubl.) Pulle Chelonanthus angustifolius Gilg Chelonanthus purpurascens (Aubl.) Struwe and V.A. Albert Chironia baccifera L. Chironia laxa Gilg Chironia linoides L. Cotylanthera paucisquama C.B. Clarks Curtia tenuifolia Knobl. Eustoma exaltatum (L.) Salisb. ex G.D. Exaculum pusillum Caruel Exacum affine I.B. Balf ex Regel Exacum caeruleum Balf. f. Exacum fruticosum Humbert Exacum gracilipes I.B. Balf. Exacum hamiltonii G. Don Exacum marojejyense Humbert Exacum nummularifolium Humbert Exacum oldenlandioides (S. Moore) Klack. Exacum quinquenervium Grisebach P. Chassot 99–231 G. Mansion and L. Zeltner 990 124-1 L. Zeltner 1756 G. Mansion and L. Zeltner 960608-1 J. Piguet C3 J. PiguetT11 J. Piguet C92 M. Callmander and M. Bondallaz, A005 M. Callmander and M. Bondallaz A003 M. Callmander and M. Bondallaz A004 Y.-M. YUAN CN2k1-31 M.J. Jansen-Jacob 2740 (NY) S. Wohlhauser and C. Bijleveld BZ001 L. Zeltner, s.n. Miller et al. 8238a Miller etal. 11356 S. Wohlhauser and J.-I. Pfund M055 Miller et al. 17126 J.R.I. Wood 7477 S. Wohlhauser and J.-I. Pfund M056 S. Wohlhauser and J.-I. Pfund MO58 M. Reekmans 9275 S. Wohlhauser M063 Cape Town, Botanical Garden Kirstenbosch, Cape, South Africa Cape Town, Botanical Garden Kirstenbosch, Cape, South Africa Cape Town, Botanical Garden Kirstenbosch, Cape, South Africa Gongshan, Yunnan, China NA Sarteneja, Shipstem Nature Reserve, Orange Walk District, N18.17.94/ W88.13.05-Belize Italy Nr Hadiboh, Sokotra, Yemen Haggeher, Sokotra, Yemen Marojejy R. I., Antsiranana, S14.26.17/E49.44.64-Madagascar Haggeher, Sokotra, Yemen Bhutan Marojejy R.I., Antsiranana, S14.26.70/E49.44.17-Madagascar Marojejy R.I., Antsiranana, S14.26.70/E49.44.17-Madagascar Alt. 1150 m;S03.04/E29.25-Bubanza Burundi Ankobahoba, Cote Est, Toamasin Madagascar Reference study study study study Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Taxon Exacum stenophyllum Klack. Exacum tetragonum Roxb. Exacum trinervium (L.) Drace Exacum wightianum Arn. Frasera speciosa Griseb. Gentiana algida Pall. J.C. Piso, S. Wohlhauser and L. Zeltner M049 R. Lundin and J. Klackenberg 332 L. Zeltner SL002 J. Klackenberg and R. Lundin 188 Y.-M. Yuan 91–S2 Y.-M. Yuan 91–S10 Y.-M. Yuan 91-S5 Y.-M. Yuan 93–14 P. K€ upfer 91–G3 Ornichia madagascariensis Klack. S. Wohlhauser M002 Ornichia trinervis (Desrousseaux) Klackenberg Orphium frutescens E. Meyer M. Callmander s.n. Sabatia angularis (L.) Pursh Saccifolium bandeirae Maguire and Pires Sebaea brachyphylla Grisebach Sebaea Sebaea Sebaea Sebaea exacoides L. longicaulis Schniz cf. macrophylla Gilg madagascariensis Klack. Swertia angustifolia Ham. ex D. Don Swertia calycina Franch. Swertia rosulata (Baker) Klack. Swertia tetraptera Maxim. L. Gautier G020 Boulder, Colorado, USA Trail Ridge, Rocky Mt., Colorado, USA La Tourne, Neuch^atel, Switzerland Rila Mt., Borovetz, Bulgaria Mt. Caucasus, Georgia AJ489902 AJ489902 AJ490227 Present study AJ489907 AJ489907 AJ490232 Present study AJ489909 AJ489913 AJ489909 AJ489913 AJ490234 AJ490238 Present study Present study Z48146* Z48142* Z48124* Z48117* AJ315230* AJ490239 Z48122* Z48068* Z48102* Z48119* Z48087* Z48132* X75702* X77895* AJ315226* Yuan and K€ upfer (1995) Yuan and K€ upfer (1995); present study Yuan and K€ upfer (1995) Yuan et al. (1996) Yuan and K€ upfer (1995) AJ489914 AJ489914 AJ490240 Present study Y.-M. Yuan 93–91 Ansatrotro Manongrarivo R. S. 5 Mahajanga, Madagascar Ganzi, Sichuan, China Z48108* Z48135* AJ315228 Yuan and K€ upfer (1995) Callejas and Balslev 1030 Antioquia, Colombia AJ489915 AJ489915 AF102455* Y.-M. Yuan 93–182 Mt. Baima, Yunnan, China Z48109* Z48137* AJ315200* Struwe et al. (1998) and unpublished Yuan and K€ upfer (1995) P. Chassot 99–243 Khao Chang Lot, Phang Nga, Thailand Ambavaniasy (Beforona), Toamasina, Madagascar Near Mandeana, forest station, Southeast, Madagascar Cape Town, Botanical Garden Kirstenbosch, Cape, South Africa USA NA AJ489916 AJ489916 AJ490241 Present study AJ489917 AJ489917 AJ490242 Present study AJ489918 AJ489918 AJ490243 Present study AJ489919 AJ489919 AJ490244 Present study AJ011467* AJ242611* AJ011477* AJ242612* AF102476* AJ242608* Thiv et al. (1999b) Thiv et al. (1999a) M. Callmander and M. Bondallaz A001 Lammers 4860 M. Piliackas et al., s.n. J. Raynal 19414 Snijman 1562 Poilecot P. 8000 Bayliss 8765 J.C. Piso, S. Wohlhauser and L. Zeltner M018 P. Chassot and Y.-M. Yuan 99–172 Y.-M. Yuan and P. K€ upfer 92–232 J.C. Piso, S. Wohlhauser and L. Zeltner M023 Y.-M. Yuan and P. K€ upfer 92–315 Mt Meru, Kitoto, Arusha Nat. Park, Tsaratanaka, Madagascar NA Mont Ingangan, Nyanga, Zimbabwe NA Miadana, Mahajanga, Madagascar AJ489920 AJ489920 AJ490245 Present study NA NA NA AJ489921 NA NA NA AJ489921 AF102481* AJ490246 AF102482* AJ490247 Struwe et al. (1998) Present study Struwe et al. (1998) Present study Binchuan, Yunnan, China AJ318551* AJ410330* AJ315203* Chassot et al. (2001) Yulong Snow Mt., Lijiang, Yunnan, China Tsaratanana, Tsaratanana R.I. Mahajanga, Madagascar Maqu, Gansu, China AJ318554* AJ410333* AJ315206* Chassot et al. (2001) AJ489922 AJ489922 AJ490248 Present study Z48115* Z48139* AJ315229* Chassot et al. (2001); Yuan and K€ upfer (1995) Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Gentiana lutea L. Gentiana pyrenaica L. Gentianella umbellata (M. Bieb.) Holub Gentianothamnus madagascariensis Humber Lomatogonium macranthum (Diels and Gilg) Fernald Macrocarpaea macrophylla (Kunth) Gilg Megacodon stylophorus (C.B. Clarke) H. Smith Microrphium pubescens C.B. Clarke Antsarasaotra, Fianarantsoa, Madagascar Chikmagalur distr., along road from Dattatreyapeetha to Ungadahalli, Karnataka, India Sri Lanka Tamil Nadu, India 503 504 Taxon Voucher Origin GenBank Accession No. ITS1 ITS2 trnL (UAA) Symphyllophyton caprifoioides Gilg Tachiadenus carinatus Desrousseaux Ratter 6742 S. Wohlhauser M059 AJ011462* AJ489923 AJ011472* AJ489923 AF102490* AJ490249 Thiv et al. (1999b) Present study Tachiadenus longiflorus Bojer ex Grisebach Voyriella parviflora (Miq.) Miq. S. Wohlhauser M006 Brasil Ambila-Lemaitso, Toamasina, Madagascar Stars, Antananarivo, Madagascar AJ489924 AJ489924 AJ490250 Present study G. Cremers 14891 NA AJ242615* AJ242616* AJ242607* Coffea arabica L. NA NA AJ224846* AJ224846* AF102405* Erithalis fruticosa L. Labordia tinifolia A. Gray Mitreola petiolata (Walt.) Torr. and Gray Gelsemium sempervirens Ait. Nerium oleander L. NA NA NA NA NA NA NA NA AF054635* NA NA AF054635* AF152697* AF102447* AF102460* NA NA NA NA NA NA NA NA AF102428* AF214386* Trachelospermum jasminoides Lem. NA NA NA NA AF214439* Thiv et al. (1999a); unpublished Andreasen et al. (1999); Struwe et al. (1998) Rova et al. (2002) Struwe et al. (1998) Gould (unpublished); Struwe et al. (1998) Struwe et al. (1998) Potgieter and Albert (unpublished) Potgieter and Albert (unpublished) Sequences retrieved from GenBank are marked with an asterisk. NA, not available. Reference Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Table 1 (continued) Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Gelsemiaceae, and Nerium oleander L. and Trachelospermum jasminoides Lem. of Apocynaceae. These outgroups were used for analyses on trnL (UAA) intron alone and for combined analyses, as the trnL (UAA) intron sequences of outgroups could be unambiguously aligned with the ingroup taxa. No outgroup was used for analyses on ITS data alone, as ITS sequences of outgroups could not be confidently aligned (see below). 2.2. DNA extraction and PCR amplification DNA was extracted from silica gel dried leaf material (Chase and Hills, 1991), or from leaf tissue taken from herbarium sheets. Total DNA was extracted using the CTAB procedure of Doyle and Doyle (1987), or the DNeasy Plant Mini Kit (Qiagen AG, Basel). The leaf tissue was homogenized in the presence of liquid nitrogen or with a Qiagen Mixer Mill MM 300 (Qiagen AG, Basel). Both DNA fragments, ITS and trnL (UAA) intron, were amplified via standard PCR in 25 ll volume containing 2:5 ll 10
PCR buffer (with 1.5 mM MgCl2 ), 0:5 ll 10 mM dNTPs, 0:5 ll of 10mM each forward and reverse primers, 0:2 ll (1U) HotStar Taq DNA polymerase (Qiagen AG, Basel), 19:8 ll H2 O, and 1 ll (ca. 10–20 ng) genomic DNA. PCRs were performed in a Biometra thermal cycler programmed for 15 min at 95 °C for the activation of the HotStar Taq DNA polymerase, followed by 35 cycles of 45 s at 94 °C, 45 s at 55 °C, 1 min at 72 °C, with a final extension period of 5 min at 72 °C. Primers ITS5 and ITS4 (White et al., 1990) were used for the amplification of ITS of the nuclear ribosomal DNA for freshly dried samples. For herbarium samples a newly designed primer, ITSPAN, by Y.-M. Yuan, was used to replace the ITS5 for better amplification. The sequence of the ITSPAN primer is 50 TCCGGTGAAGTGTTCGGATCGC-30 . It is located about 63 bp upstream of the original ITS5, and generally gave better amplification of the ITS region when used in combination with ITS4 for the samples for which PCR amplifications failed with the primers ITS5 and ITS4. The primers ‘‘c’’ and ‘‘d’’ of Taberlet et al. (1991) were used for the amplification of the trnL (UAA) intron. 2.3. PCR purification and sequencing The PCR products were checked on 0.8% agarose gels. Successfully amplified DNA fragments were then purified prior to sequencing using the QIAquick PCR purification kit (Qiagen AG, Basel) following the manufacturerÕs protocol. Cycle sequencing reactions were performed using the dye-terminator chemistry as implemented in the ABI PRISM BigDye Terminator Cycle Sequencing Ready Reaction Kit (Applied Biosystems, Foster City, USA) in a Biometra thermal cycler. The same primers used for amplification were used for 505 sequencing for both ITS and trnL (UAA) intron, whereas the two more internal primers ITS2 and ITS3 (White et al., 1990) were occasionally used to confirm some sequencing results for ITS. Protocols and cycling parameters suggested by the sequencing kit were followed except that the reaction volumes were scaled down to 5 ll. The cycle sequencing products were cleaned using the ethanol/sodium acetate precipitation method as suggested by the manufacturer of the sequencing kit. The purified sequencing products were resuspended in 12 ll TSR (supplied by Applied Biosystems, Foster City, USA) and then analyzed on an ABI310 automated sequencer using a 47 cm capillary and polymer POP-6e (Applied Biosystem, Foster City, USA). Automation-generated base-calls were subsequently checked manually against the electropherograms using the software Sequence Navigator (Applied Biosystems, Foster City, USA). 2.4. Sequence alignment The boundaries of ITS1, ITS2, and the trnL (UAA) intron in the studied material were determined by comparison with previous results (Gielly et al., 1996; Yuan and K€ upfer, 1995). Combined sequences of ITS1, ITS2, and trnL (UAA) intron were preliminarily aligned with Clustal X (Thompson et al., 1997) and then manually adjusted to reflect indel events. The alignment of trnL (UAA) intron sequences was straightforward, and required minor manual adjustments. Potentially informative and unambiguously assessable indels in trnL (UAA) intron were scored as binary characters (1 for insertion, 0 for deletion) regardless of their length and were added to the data matrix (which gave 21 additional characters). The alignment of ITS sequences was problematic, so various alignment alternatives were explored to investigate the alignment consequence on phylogenetic inference. The multiple alignments were explored in two ways, taxon deletion and gap penalty alteration. Firstly, using the default penalties of Clustal X, several rounds of alignments were conducted and at each round a set of divergent taxa were removed from the data to be aligned. In this way, outgroups and too divergent ingroup taxa (tribe Saccifolieae, the genera Cotylanthera, Exaculum, Gentiana, Lomatogonium, and Symphyllophyton) were stepwise excluded from alignment to generate a set of reduced data matrixes. Comparison of phylogenetic analyses on this battery of alignments indicated that tree topologies started to be stabilised when outgroups and the tribe Saccifolieae were excluded. The reduced data set (excluding outgroups and the tribe Saccifolieae) was then subjected to the second round of alignment test, applying varied gap penalties for Clustal X. No significant influence on tree topologies was observed as long as the ambiguous alignment areas were excluded (data not shown). Thus the alignment of the reduced data set where outgroups and the tribe 506 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Saccifolieae were excluded was chosen and further slightly adjusted manually, and was used for subsequent phylogenetic analyses. The indels of ITS sequences were not used as separate characters as their boundaries could not be unambiguously determined. To ensure the alignment accuracy, an alignment of the ITS and trnL intron sequences of the tribe Exaceae alone was also conducted and analyzed separately as a comparison with the broad analyses. 2.5. Phylogenetic analysis All data sets were analyzed with heuristic searches applying the maximum parsimony optimality criterion, using PAUP* v4.0b8a or v4.0b10 (Swofford, 2000). Analyses were performed on trnL (UAA) intron alone, on ITS alone, and on combined ITS and trnL (UAA) intron. In all analyses characters were equally weighted and unordered (Fitch, 1971), with gaps treated as missing data. For the analyses on trnL (UAA) intron alone, the species Cotylanthera paucisquama was removed as its trnL (UAA) data was missing. A three-step search procedure was performed with branch collapse option set to collapse if maximum length as zero. The first step, a heuristic search to save maximum of 5000 shortest trees (Maxtree ¼ 5000), was made to obtain an empirical minimum tree length, which generated the minimum tree length of 457 steps. The second step involved 1000 replicates of random stepwise addition of sequences, and from each replicate only one tree not longer than 457 steps was saved, giving a total of 1000 trees. Finally TBR branch swapping was performed on all these trees with MULTREES and STEEPEST DESCENT options on, and MAXTREE set to 10,000, 50,000, or 100,000, respectively. The strict consensus tree from each set of these trees was compared, and was found identical in topology. However, when branch collapse option was set to collapse if minimum length as zero, heuristic searches of 1000 replicates of random addition of sequences, with TBR branch swapping, MULTREES and ACCTRAN options on, and STEEPEST DESCENT option off, generated only six trees. Their strict consensus was identical to those obtained by analyses described above. All the seven outgroup taxa were included in all these analyses. All heuristic searches on ITS data alone or on combined data implemented 100 replicates of random addition of sequences, with TBR branch swapping, MULTREES and ACCTRAN options on, and STEEPEST DESCENT option off. Both the ÔmaximumÕ and ÔminimumÕ branch collapse options were compared. No difference was found in topology of the strict consensus trees. For all the results based on separate and combined data, the relative support for individual clades was evaluated by bootstrap analyses (Felsenstein, 1985). Bootstrap values were calculated using 1000 replicates of heuristic searches, with SIMPLE ADDITION SEQUENCE, TBR branch swapping, MULTREES option on, STEEPEST DESCENT option off, and a maximum of 1000 trees saved for each replicate. 2.6. Data congruence test In order to test the congruence of ITS and trnL (UAA) intron data sets, we conducted the incongruence length difference test of Farris et al. (1995). The partition ITS vs. trnL (UAA) intron homogeneity tests as implemented in PAUP were performed with 1000 replicates of heuristic searches (Maxtree ¼ 1000, TBR branch swapping, and other default settings) on the combined data matrix. 2.7. Molecular clock test To test the assumption of clock-like evolution of the DNA sequence under study, the likelihoods of all the equally maximum parsimonious trees (obtained with branch collapse option set to minimum) were calculated separately with and without molecular clock constraints on data sets where all gapped, ambiguity and missing sites were excluded using PAUP. C. paucisquama was excluded from these analyses as an accelerated evolution was apparent from the phylogram (Fig. 3). The GTR + I + G model (GTR, general time reversible; I, variable proportions of invariable sites; G, gamma distributed among-site rate variation) and parameter settings (gamma shape, base frequencies) were chosen through the Hierarchical Likelihood Ratio Tests procedure as implemented in Modeltest (Version 3.06) (Posada and Crandall, 1998) for the corresponding data sets were used. Using a v2 distribution with N  2 degrees of freedom, where N is the number of terminal taxa in the trees, a likelihood ratio test was performed based on twice the difference between the log-likelihoods for clock and no-clock analyses (Muse and Weir, 1992). 2.8. Divergence time calculations To obtain an approximate dating of branching events, specifically the divergence of the family Gentianaceae and the tribe Exaceae, the times of divergence were estimated for highly supported sister groups. When a molecular clock is accepted by the likelihood ratio test, the pairwise sequence divergence values between two sister groups were determined as the average of all pairwise sequence divergence values between species from the two clades. The average sequence divergence values were calculated from the uncorrected mean pairwise distance to accommodate the divergence rate previously calibrated (Richardson et al., 2001a). The Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 highest previously reported divergence rate 1:30
109 substitutions per site per year (s/s/y) and the lowest divergence rate 4:87
1010 s=s=y for trnL (UAA) intron as summarised by Richardson et al. (2001a) was used. Divergence time between two lineages was estimated as half of the average divergence value between them divided by the divergence rate. In cases where the likelihood ratio test rejected a molecular clock, maximum parsimony trees based on trnL intron were subjected to non-parametric rate smoothening (NPRS) (Sanderson, 1997) using the default settings in TreeEdit v.1.0a8 (Rambaut and Charleston, 2000; Richardson et al., 2001a) to obtain homogenized rates. To calibrate the trees, the minimum age of Gentianales was estimated to be 60 million years (MY) based on pollen fossil data (Muller, 1984). As an independent test for the divergence times based on rate smoothening, the branch-specific rate dating method based on the mean branch length (MBL) as implemented in Bremer (2000) and Patterson and Givnish (2002) was used to obtain estimations of the divergence times using the same calibration point. The six equally most parsimonious trees generated by heuristic searches using Ôminimum as zeroÕ branch collapse option, were examined. Analyses without the fast evolving taxa (Curtia, Saccifolium, and Voyriella), where only a single most parsimonious tree was obtained using the same options, were also performed to investigate their effect on the results of rate smoothening and divergence times. Branch lengths were obtained using accelerated transformation optimisation. The delayed transformation optimisation was also tested but did not cause any significant change on final results (data not shown). In all cases, the trees were rooted at the branch connecting Rubiaceae and the other outgroup taxa. No detailed dating was attempted on ITS trees, as ITS sequences of the outgroups and the tribe Saccifolieae could not be unambiguously aligned with other ingroup taxa. 3. Results 3.1. Sequence characteristics and alignments All newly obtained sequences have been submitted to EMBL/GenBank databases. The accession numbers of the sequences used in this study are listed in Table 1. The 164 base pairs (bp) 5.8S rDNA regions have few mutations and were removed from analyses, because this part is missing in the sequences retrieved from GenBank database. Despite of our various efforts, we were not able to amplify the trnL (UAA) intron and many other cpDNA fragments, such as the rbcL gene, from the saprophytic species C. paucisquama. As this plant is completely white, it may have few or degenerate plastids and its 507 cpDNA could have undergone dramatic variation so the available primers were not able to amplify the corresponding regions any more. This species is thus omitted from the analysis of the trnL (UAA) intron data set, but included for combined analyses with its trnL (UAA) intron simply treated as missing data. The length of unaligned trnL (UAA) intron sequences of the remaining OTUs ranged from 343 to 539 bp. The aligned trnL (UAA) intron had 636 bp. The alignment was straightforward and unambiguous except for two simple sequence repeat (SSR) regions of multiple As (13 and 72 bp, respectively in our aligned data matrix). These ambiguously alignable SSR regions (85 bp in total) were excluded in subsequent analyses. Introduced gaps were from 1 to 111 bp in length. In total, 21 potentially informative indels were scored as binary characters regardless of their sizes and were added to the sequence data. Finally the trnL (UAA) intron matrix had 657 characters, of which 85 (12.9%) were excluded, 310 (47.2%) were constant, 86 (13.1%) variable but uninformative, and 176 (26.8%) informative. These data resulted in uncorrected pairwise sequence divergence ranged from 0 (Chironia baccifera vs. C. linoides, Exacum marojejyense vs. Exacum fruticosum, Canscora alata vs. Canscora andrographioides, and Sebaea brachyphylla vs. S. exacoides) to 18% (Curtia tenuifolia vs. Sebaea macrophylla) among the Gentianaceae ingroup taxa. The length of unaligned ITS1 and ITS2 sequences varied from 212 to 237 bp and from 192 to 248 bp, respectively. The alignment of the ITS region was difficult, particularly between outgroups and a region nearby the beginning of ITS2. This region of 13–52 bp was always differently aligned among our different alignment exercises, and seems to correspond with the hypervariable terminal loop region of the arm 1 in secondary structure models as revealed in Aeschynanthus of Gesneriaceae (Denduangboripant and Cronk, 2001). Tests of multiple alignments through stepwise exclusion of divergent taxa (see method) indicated that ITS sequences of the outgroups and the tribe Saccifolieae of ingroup taxa were too divergent to be confidently aligned with other ingroup taxa. The inclusions of these sequences dramatically influence phylogenetic inference of ITS sequences, particularly regarding the relationships of the basal clades (caused by both ambiguous alignment and longbranch attraction). When these sequences were excluded from alignment, the topologies of inferred trees were stabilised and further exclusion of other suspected divergent taxa such as Cotylanthera, Exaculum, Gentiana, Symphyllophyton, etc. did not alter inferred tree topologies as long as the ambiguously aligned indel regions were excluded from analyses. The reduced ITS data matrix contained 534 characters, of which 145 (27.2%) were excluded from phylogenetic analyses, 95 (17.8%) were constant, 78 (14.6%) variable but uninformative, and 216 (40.4%) potentially 508 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 informative. These data resulted in uncorrected pairwise sequence divergence ranged from 0 (Exacum marojejyense vs. Exacum fruticosum) to 43.4% (C. paucisquama vs. Canscora diffusa). The extraordinarily high divergence value was due to the saprophytic species C. paucisquama, which appears to have an accelerated rate of evolution (Fig. 3). The data matrix based on the alignment covering all OTUs can be consulted or downloaded from the web pages of the Laboratory of Evolutionary Botany of the University of Neuch^atel (http://www.unine.ch/bota/ebolab/gentianaceae/gentmain.html). When the ITS alignment was limited to the tribe Exaceae, the saprophytic species, C. paucisquama, caused alignment ambiguity and long-branch attraction in subsequent phylogenetic inference. This by-effect could be easily corrected by combining trnL intron data of the corresponding species in the phylogenetic analyses (with the trnL intron sequence of C. paucisquama coded as missing). 3.2. trnL intron phylogeny Heuristic searches on the trnL (UAA) intron data set resulted in over 100,000 equally most parsimonious trees when branch collapse was set to Ômaximum length as zeroÕ or only six equally most parsimonious trees when branch collapse was set to Ôminimum length as zero.Õ All these maximum parsimonious trees had 457 steps, with a consistency index (CI) of 0.74 including autapomorphies (0.68 excluding autapomorphies), and a retention index of (RI) 0.89. The strict consensus of these two sets is identical, and is shown in Fig. 1A. Bootstrap values for the clades supports (when greater than 50%) are also shown. The trnL (UAA) intron consensus tree was well resolved and highly supported toward the base of the tree, while the upper branches (among closely related genera or species from the same genera) were poorly resolved or received less significant support. The monophyly of most tribes and subtribes was highly supported (72–100%), except for tribe Gentianeae (55%) and subtribe Chironiineae (57%) the bootstrap supports were slightly weak. The tribe Saccifolieae was resolved as the most basal clade of the family Gentianaceae with high support (100%), followed by the tribe Exaceae (99%) and tribe Chironieae (89%). The remaining three tribes formed an unresolved trichotomy. Despite the high support for the monophyly of the tribe Exaceae, the trnL (UAA) intron provided no resolution on the relationship between Exacum and Ornichia (and Sebaea madagascariensis), and the relationships among the species within Exacum. The sampled species of Sebaea, excluding S. madagascariensis, formed a highly supported basal clade (100%), sister to all the remaining genera of the tribe. Gentianothamnus was on a polytomy together with Tachiadenus (96%). Together they were sister to Exacum, Ornichia, and S. madagascariensis, residing unresolved in a polytomy. 3.3. ITS phylogeny Maximum parsimony analyses on alternative alignments which include both outgroups and ingroup taxa, with outgroup taxa excluded, or with both outgroup taxa and the tribe Saccifolieae of ingroup excluded, resulted in varied resolutions on basal relationships among the main clades (not shown). This variation was caused by alignment ambiguity and long-branch attraction introduced by the outgroups and too divergent ingroup taxa, the tribe Saccifolieae. When outgroups and the tribe Saccifolieae were excluded, alternative alignments (obtained by further excluding other divergent taxa or changing alignment penalties) gave consistent phylogenetic resolution that was also congruent with the topologies inferred from trnL intron data. Heuristic search on the reduced ITS data (containing 54 taxa) generated only two trees of 1233 steps (CI ¼ 0.45 including autapomorphies, CI ¼ 0.40 excluding autapomorphies, RI ¼ 0.66). The strict consensus tree and bootstrap values for clade support when higher than 50% are shown in Fig. 1B. Contrary to the trees obtained from the trnL (UAA) intron data set, the upper branches (among closely related genera or species from the same genera) of the ITS trees were better resolved and received significant bootstrap support, while the corresponding main clades and the relationships among them were revealed identical to that obtained from trnL intron data. The saprophytic C. paucisquama, which was missing from trnL tree, nested within the genus Exacum and grouped with two Himalayan species, E. hamiltonii and E. tetragonum. Confirming the results based on trnL intron data, the genus Sebaea was shown polyphyletic. One species, S. brachyphylla, was revealed as the most basal clade within the tribe. Another species S. madagascariensis nested with the genus Ornichia, the same position suggested by the trnL intron data. ITS trees had higher resolutions than the trnL trees for the tribe Exaceae. The Madagascar endemic genus Tachiadenus was revealed to be monophyletic, and the monotypic genus Gentianothamnus was closely related to it. Provided C. paucisquama is included, Exacum was monophyletic with two main clades, an African-Madagascan clade and a Socotra-Asiatic clade, the later included three lineages, South Indian-Sri Lankan, the Himalayan, and the Socotran. 3.4. Data sets conflict test and combined ITS and trnL (UAA) phylogeny Trees resulted from trnL (UAA) and ITS data sets did not show any conflict in topology when examined visually. The partition homogeneity test on trnL (UAA) Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 509 Fig. 1. Comparison of strict consensus trees obtained from separate analyses of trnL (UAA) intron and ITS sequence alone. Numbers above the branches are bootstrap values supporting the corresponding branch when greater than 50%. (A) The strict consensus tree of the equally parsimonious trees obtained from trnL intron (length ¼ 457, CI ¼ 0.74 including autapomorphies, CI ¼ 0.68 excluding autapomorphies, RI ¼ 0.89). (B) The strict consensus tree of the two equally parsimonious trees obtained from reduced ITS data matrix alone (length ¼ 1233, CI ¼ 0.45 including autapomorphies, CI ¼ 0.40 excluding autapomorphies, RI ¼ 0.66). Bold face indicates the unexpected resolutions (see text). 510 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 intron vs. ITS received significant statistic support with P ¼ 0:99=1:0 (without/with the taxon for which only trnL intron or ITS was available). Therefore both data sets were combined and were subjected to a combined analyses. The missing or excluded data, trnL intron of C. paucisquama and ITS sequences of outgroups and the tribe Saccifoileae, were coded as missing data. Such combined data matrix covered 67 taxa (including the seven outgroup species) and 1191 characters, of which 230 (19.3%) were excluded from analyses, 405 (34.0%) were constant, 164 (13.8%) were variable but uninformative, and 392 (32.9%) potentially informative. 150 equally most parsimonious trees with 1694 steps were generated (CI ¼ 0.53 including autapomorphies and 0.46 excluding autapomorphies, RI ¼ 0.74). The strict consensus of these trees is shown in Fig. 2. The consensus tree is well resolved and its topology is congruent with that generated from separated analyses on trnL (UAA) intron and ITS data alone. The monophyly of all currently recognized tribes and subtribes were strongly supported. The tribe Saccifolieae was resolved as the most basal clade within the Gentianaceae, followed by the tribe Exaceae as sister to the rest of the family. Tribe Chironieae was in turn sister to tribes Potalieae, Helieae, and Gentianeae. Tribe Helieae showed the closest relationship with tribe Gentianeae, being the most derived taxa of the family. Except for the monotypic genus Gentianothamnus and the genera with only one species sampled for which the present phylogeny did not offer any evaluation on their monophyly, most of the genera were revealed to be monophyletic. The only exceptions were that paraphyly was revealed for Exacum, Swertia, and Sebaea. The paraphyly of Swertia has been extensively studied by Chassot et al. (2001). The Madagascan S. rosulata was revealed here as derived and close to Gentianella and Lomatogonium in the Gentianella lineage in the tribe Gentianeae. Within the tribe Exaceae, the saprophytic genus, Cotylanthera, was revealed as deeply nested inside the genus Exacum, thus the recognition of Cotylanthera as a distinct genus made the genus Exacum paraphyletic. Tests did not reveal long-branch attraction caused by Cotylanthera in this combined analysis. The genus Sebaea was also shown to be polyphyletic. While most of the sampled species grouped together as a highly supported basal clade sister to all the remaining taxa of the tribe, the Madagascan endemic species S. madagascariensis showed a closer relationship with the genus Ornichia. This similarity has been confirmed by sequences obtained from the DNAs extracted from three different individual accessions, thus the possibility of DNA sample contamination could be excluded. As far as the sampled species were concerned, Ornichia and Tachiadenus were revealed as highly supported monophyletic genera. The close relationship between the two Madagascan endemic genera Gentianothamnus and Tachiadenus was highly supported. Ornichia and S. madagascariensis together showed a close relationship with the genus Exacum that was revealed as the most derived taxon within the tribe Exaceae. Exacum bifurcated into two clades at the base, the Madagascar-African clade and the Asian clade. The Asian clade included three highly supported lineages as revealed by ITS data alone (Fig. 2). Phylogenetic analyses on alignments of combined trnL intron and ITS sequences limited to the tribe Exaceae confirmed the above results. 13 equally most parsimonious trees with 629 steps were generated (CI ¼ 0.72 including autapomorphies and 0.61 excluding autapomorphies, RI ¼ 0.70). The topology of the strict consensus tree (not shown) was identical to that of the Exaceae clade of the strict consensus tree generated from the broad combined analysis including all other tribes (Fig. 2). The saprophytic C. paucisquama showed an accelerated evolution, as indicated by its extraordinary long branch in the phylogram of one of the 13 trees (Fig. 3). Phylogenetic analyses using only ITS data clearly showed long-branch attraction of C. paucisquama: tree topology changed when this species was in- or excluded. Combining the trnL intron data could easily correct the long-branch attraction. 3.5. Molecular clock test and estimations of divergence time The likelihood ratio test rejected a clock-like evolutionary rate for all ITS data sets (p < 0:05). While a molecular clock was rejected globally for trnL (UAA) intron data for the whole Gentianaceae, a molecular clock could not be rejected for trnL (UAA) intron data of the tribe Exaceae alone ðp > 0:05Þ. We have not identified a reliable calibration for the clock of trnL intron of the tribe Exaceae, so the lowest rate 4:87
1010 substitution/site/year (s/s/y) and the highest rate 1:3
109 s/s/y previously reported for trnL (UAA) intron (Richardson et al., 2001a) were used to estimate the divergence times of the two basal nodes of the tribe, which were supported by both ITS and trnL data when analyzed separately or combined. The inferred divergence times are plotted on one of the phylogenetic trees of the tribe Exaceae obtained from combined analyses (Fig. 3), and are also given in Table 2. According to these rates and the trnL intron sequence divergence, the genus Sebaea, the most basal clade of the tribe, has diverged from others for about 20.3–54.3 MY. The divergence between the Gentianothamnus– Tachiadenus lineage and the lineage including Sebaea madagascariensis, Ornichia, and Exacum was 11.2–29.8 MY. The results of divergence dating on trnL trees based on the clock-independent NPRS and MBL methods and the calibration point of 60 MY for the minimum age of Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 511 Fig. 2. Strict consensus tree of the 150 equally maximum parsimonious trees obtained from combined trnL (UAA) intron and ITS sequences (length ¼ 1694, CI ¼ 0.53 including autapomorphies, CI ¼ 0.46 excluding autapomorphies, RI ¼ 0.74). Numbers above the branches are bootstrap values supporting the corresponding branch when greater than 50%. Current classification is shown on the right. Bolded face indicates the unexpected resolutions (see text). 512 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Fig. 3. Phylogram of one of the 13 equally most parsimonious trees obtained from combined analyses on ITS and trnL (UAA) intron sequences of the tribe Exaceae only (length ¼ 629, CI ¼ 0.72 including autapomorphies, CI ¼ 0.61 excluding autapomorphies, RI ¼ 0.70). Branch lengths are proportional to the substitutions supporting the branch. Note the extraordinary long branch of the saprophytic species Cotylanthera paucisquama, which suggested an accelerated evolution due to its saprophytism. The divergence time (million years) of the two basal nodes estimated from trnL intron based on the clock rate of the lowest 4:87
1010 and the highest 1:3
109 as previously reported are also shown. These two nodes were supported by both ITS and trnL data analysed separately or combined. the order Gentianales are summarised in Table 2. Both methods suggested very close estimations on the diverging time of the Gentianaceae (42.4–50.4 MY). For the divergence of the tribe Exaceae NPRS approach estimated as 34.5–49.2 MY, while MBL approach re- vealed a more recent divergence (19.8–25.9 MY). Similarly the divergence between genus Sebaea and other Exaceae was revealed as 22.4–28.8 and 15.6–19.1 MY respectively, both fell in the range suggested by a molecular clock. There is a clearly increased discrepancy of 513 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Table 2 Results of divergence dating based on NPRS trees, MBL, and molecular clock applying diverse rates of gene evolution Divergence of trnL intron, six trees, NPRSa trnL intron excl. fast taxab , 1 tree, NPRS trnL intron, six trees, MBLa trnL intron excl. fast taxab , 1 tree, MBL Average sequence divergence of trnL intron of Exaceae only, MC ratec Gentianales Gentianaceae Tribe Exaceae Sebaea vs. other Exaceae (Tachiadenus–Gentianothamnus) vs. (Ornichia–Exacum clade) Tachiadenus vs. Gentianothamnus Exacum 60 (CP) 42.4–47.7 34.5–39.4 22.4–25.8 18.8–21.4 60 (CP) 50.4 49.2 28.8 24.0 60 (CP) 43.1–48.0 19.8–21.6 15.6–17.0 7.0–7.8 60 (CP) 42.4 25.9 19.1 7.5 n/a n/a n/a 20.3–54.3 11.2–29.8 9.7–11.6 15.2–17.9 13.2 20.4 3.2–3.5 3.1–4.1 4.1 4.8 5.5–14.7 8.1–21.6 Divergence times are shown in million years before present (MY). NPRS, non-parametric rate smoothing (Richardson et al., 2001a; Sanderson, 1997); MBL, mean branch length (Bremer, 2000; Patterson and Givnish, 2002); MC, molecular clock; CP, calibration point based on fossil pollen date (Muller, 1984). a The range shows the difference obtained from different trees for the corresponding nodes. b Excluding Curtia, Voyriella, Saccifolium. c Based on previous rates 1:30
109 and 4:87
1010 s/s/y (Richardson et al., 2001a). divergence time estimations between NPRS and MBL toward the terminals of the tree. This is probably due to the increased probability of sampling error involved in MBL method toward the tree terminals. If we consider the age of the tribe Exaceae as about 40 MY as suggested by NPRS, the maximum divergence rate of trnL intron in this tribe could be calibrated as around 8:62
1010 s/s/y, which fell within the range given by Richardson et al. (2001a). Dating with ITS sequences using the previously reported ITS fast and slow rates 7:83
109 and 1:72
109 s/s/y (Richardson et al., 2001a) (despite rejection of a molecular clock) indicated a similar divergence time, 12.8–58.1 MY, for the tribe Exaceae, supporting the cpDNA findings (data not shown). 4. Discussions 4.1. Congruence of phylogenies based on ITS and trnL (UAA) intron Partition homogeneity statistic test suggested that ITS and trnL (UAA) intron data sets were homogeneous since no character conflict was detected. When the ITS sequences of the divergent taxa, outgroups and the tribe Saccifolieae were excluded from aligning, the topologies of the remaining clades were also highly congruent with corresponding clades of the trnL trees, the monophyly of the tribes and subtribes being highly supported by both data sets. Therefore, we consider that both trnL intron and ITS are highly congruent for phylogenetic inference, and the majority of our discussions and conclusions were based on the combined analyses, except for Sebaea (based mainly on trnL data) and Cotylanthera (based on ITS data only). 4.2. Monophyly of Exaceae and its relationships with other tribes Prior to the recent treatment by Struwe et al. (2002), the infrafamilial classification of the Gentianaceae remained incomplete or even misleading with regarding to the understanding of phylogenetic relationships among different genera. As far as the tribe Exaceae is concerned for an example, the genus Gentianothamnus had been placed in subtribe Chironiinae since it was described (Humbert, 1937) and Tachiadenus was placed in subtribe Tachiinae by Gilg (1895). Only the recent treatment of Struwe et al. (2002) classified them in Exaceae, and our present study confirmed their inclusion in Exaceae and showed that they are very closely related to each other. Struwe et al. (2002) based largely on phylogenetic analyses of the chloroplast trnL (UAA) intron sequences and partially on matK sequences (Struwe et al., 1998; Thiv et al., 1999a), made the first comprehensive classification of the whole family in the context of identifying monophyletic lineages. By extended sampling of both plant taxa that were not available for previous studies (e.g., Cotylanthera and Gentianothamnus) and DNA sequences (nuclear ITS in addition to the chloroplast intron), our present phylogenetic analyses closely resembles the tribal classification of Struwe et al. (2002) with extended resolutions. Our present analyses provided strong evidence and statistical support for the monophyly of the major lineages, the tribes and subtribes of the Gentianaceae defined by the current classification. The phylogenetic hypotheses obtained from separate analyses on chloroplast trnL (UAA) intron (Fig. 1A) and ITS (Fig. 1B) or combined (Fig. 2) showed that the tribe Exaceae was a basal clade just next to the basal-most lineage, the tribe Saccifolieae. 514 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 4.3. Generic relationships within tribe Exaceae The extensive taxonomic and phylogenetic studies based on morphology and anatomy done by Klackenberg (1985, 1986, 1987, 1990, 2002) have forged the shape of this tribe, and is the base for a hypothesis on relationships among the genera. The phylogenetic analyses on the Gentianaceae based on the trnL (UAA) intron done by Struwe et al., 1998 and Thiv et al. (1999a,b) covered six species representing four genera of the tribe. Our present molecular phylogeny corroborated KlackenbergÕs definition of the tribe and some relationships suggested by him for example the close relationship between Ornichia and Exacum. The genus Sebaea was shown basal within this tribe. Two genera, Gentianothamnus and Cotylanthera, were not available in the previous molecular studies, so their molecular phylogenetic relationships toward others were unknown. Klackenberg (2002) included Gentianothamnus, which was originally included within the subtribe Chironiinae by Humbert (1937), in the tribe Exaceae and suggested its close relationship with Tachiadenus based on morphological and anatomical characters such as long corolla tube, undulate-walled testa cells, and pollen morphology. Our present phylogeny with extended samples of species based on combined data confirmed the previous phylogeny based on trnL (UAA) intron sequences alone, regarding the basal position of the genus Sebaea (except S. madagascariensis) (Figs. 1A, 2, 3) and the close relationships between Exacum and Ornichia (Figs. 2, 3). Our present study further indicated that Gentianothamnus is closely related to Tachiadenus, Sebaea is probably polyphyletic and Cotylanthera nests deeply inside Exacum. The genus Exacum was resolved as monophyletic if Cotylanthera is included (Figs. 1B, 2, 3). However, the monophyly of Exacum did not receive strong bootstrap support. 4.4. Polyphyletic Sebaea While the molecular phylogeny was mostly congruent with the current classification of the tribe Exaceae (Klackenberg, 1990, 2002), conflict occurred regarding the monophyly of the genus Sebaea. The currently defined Sebaea contains 150–159 species (Boutique, 1972; Paiva and Noguiera, 1990): one in New Zealand, two in Australia (Adams, 1996), one in China through India (Ho and Pringle, 1995), four in Madagascar (Klackenberg, 1990) and the rest in Africa (Klackenberg, 2002). The five species sampled in trnL intron phylogeny were nesting in two different clades: S. madagascariensis grouped with the two sampled species of Ornichia, while S. brachyphylla, S. longicaulis, S. macrophylla, and S. exacoides grouped together as a highly supported monophyletic, and the most basal clade of the tribe (Figs. 1A and 3). ITS data confirmed the positions of S. madagascariensis and S. brachyphylla. This result suggested that the genus Sebaea as it is currently defined is not monophyletic. However, the present study sampled only a fraction of the species of Sebaea (five out of about 150 species. Analyses including more species of Sebaea (and Ornichia) are necessary to draw conclusions on the phylogenetic status of Sebaea. 4.5. Cotylanthera—a saprophytic Exacum Cotylanthera as currently recognized is a small saprophytic genus with only four species. Early taxonomic studies had placed this genus outside the Gentianaceae, but Gilg (1895) included Cotylanthera within his subtribe Exacinae. Klackenberg (2002) accepted the inclusion of the genus within tribe Exaceae, and further suggested that it should have a close relationship with Exacum or even should be a derived lineage inside Exacum based on floral morphology as their anthers are all opening by apical pores and all have finely perforated endothecial walls. Although trnL (UAA) intron was successfully amplified and sequenced from another achlorophyllous plant of the Gentianaceae, Voyriella parviflora, using the same set of the universal primers (Thiv et al., 1999a), we failed to generate trnL (UAA) intron data from C. paucisquama after various efforts. Studies have shown that achlorophyllous holoparasitic plants often undergone dramatic variations, usually many deletions, in their plastid DNA. For examples, the complete plastome sequence of Epifagus virginiana has only 70,028 bp (compared to 155,844 bp of Nicotiana tabacum, GenBank: Z00044), and the trnL (UAA) gene is completely deleted (Wolfe et al., 1992). In Cuscuta cuspidata the trnL (UAA) intron is reduced to 270 bp (GenBank: AF323745), and in Cuscuta attenuata, it is reduced to only 243 bp (GenBank: AF348404), compared to lengths between 343 and 539 bp for most the Gentianaceae. It is not clear what mechanism is involved in the cpDNA evolution of C. paucisquama, and it is beyond the scope of the present study. Despite that the trnL (UAA) intron data is not available, C. paucisquama nested deeply inside Exacum and grouped with the Himalayan species E. hamiltonii and E. tetragonum in the ITS tree. Due to an accelerated evolution in this saprophytic species as indicated by its high proportion of autapomorphic characters (112 compared to 14 in E. tetragonum), it showed high sequence divergence to most species of the tribe sampled here (Fig. 3). Morphologically its affinity with Exacum is well justified as suggested by Klackenberg (2002). However, it is not advisable at present to reduce Cotylanthera to Exacum, as the ITS data are not enough convincing, and only one species out of four was analyzed. It is preferable to further investigate this suggested affinity by studying more samples, and other genes as the accelerated evolution of its DNA sequence could cause alignment errors Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 and long-branch attraction. The destabilising influence on the tree topology of Cotylanthera is indicated when its sequence is included or excluded in the analyses of ITS data of the tribe Exaceae alone. 4.6. Biogeographical implications While the Gentianaceae is a cosmopolitan family, each tribe has specific and interesting distribution patterns. Tribe Saccifolieae is limited to the Neotropics with a majority of species occurring on the Guayana Shield. Tribe Exaceae has a paleotropical, and austral African distribution with its highest diversity centred around the Indian Ocean Basin. Tribe Chironieae has three major lineages, each corresponding to neotropical, paleotropical, and north-temperate distribution. Tribe Helieae is limited to the neotropical regions. Tribe Gentianeae is widely distributed throughout the world with the highest diversity occurring in the Old World, and tribe Potalieae is disjunctively distributed in pantropical areas. The family has so far no reliable fossil record. The earliest megafossil of suggested Gentianaceae origin was the fossil flowers with Pistillipollenites pollen from Eocene of North America ca. 45 MY old. These fossil pollens were suggested to be associated with the relatively derived Helieae clade (Macrocarpaea) of the extant Gentianaceae (Crepet and Daghlian, 1981), but unconfirmed (Stockey and Manchester, 1986). Struwe et al. (2002) argued that this fossil might not belong to the Gentianaceae. The earliest fossil records (pollen) assigned to other families of the order Gentianales allied with the Gentianaceae were also from Eocene, and the minimum age of Gentianales was estimated to be about 60 MY (Muller, 1984) or to 53.2 MY (Magallon et al., 1999). Considering the distribution of the diversity, particularly of the basal clades, Struwe and Albert (2002) suggested an austral Gondwana origin of the family Gentianaceae. Whether the distribution patterns of the family is linked to the breakup of the Gondwana supercontinent or from post-Gondwana dispersal/migration event (or both) still remain as open questions. The topologies of the molecular trees resemble to certain extent a Gondwana vicariance pattern. However, neither the evidence of reliable fossil records nor divergence dating results on Asterids (Magallon et al., 1999) could corroborate the Gondwana hypothesis of Gentianaceae. Using the earliest pollen fossil date suggested for Gentianales and nonparametric methods (NPRS and MBL), our present dating results suggest that the family has likely diverged for about 50 MY only (Table 2), much later than the Gondwana breakup. Obviously, fossil calibrated dating gives only a minimum age, and the rate calibration or reliable fossil identification is crucial and perhaps invokes problems. The Gentianaceae, or even the whole Gentianales, have so far not many reliable 515 fossil records. The earliest fossil records for the basal clades of Asterids such as Cornales and Ericales are mostly from the Late Cretaceous (Turonian–Coniacian) of about 90 MY before present (Crepet, 1996; Crepet et al., 1992; Manchester, 2002; Manchester et al., 1999; Nixon and Crepet, 1993; Schonenberger and Friis, 2001; Takahashi et al., 1999). Archaefructus, supposed to be the earliest angiosperm fossil found so far, was 145 MY old (Sun et al., 1998). The Gentianaceae, being a relatively derived lineage among the Euasterids (APG, 1998), is unlikely diverged earlier than other Asterids. It is less likely that the Gondwana disjunctive distribution of the Gentianaceae was the direct results of a Gondwana breakup that should be at least more than 100 MY ago (McLoughlin, 2001). Regarding the tribe Exaceae and particularly the genus Exacum, Klackenberg (1985, 2002) attributed its distribution pattern to vicariance events across the Indian Ocean Basin. He suggested a possible old vicariance between Africa (Sebaea) and Indian Ocean Basin (Exacum, Ornichia, and Tachiadenus), followed by a vicariance within Exacum between Madagascar and India. The initial split of the Gondwana that began in Late Jurassic ca. 165 MY (McLoughlin, 2001) or 152 MY (http://www.scotese.com/late1.htm) seems too old to explain this possible vicariance, so he hypothesized that the vicariance would have been resulted from an extended phytogeographic contact between continental Africa and Madagascar/India after the initial splitting (Klackenberg, 2002). However, our present dating obtained no evidence to corroborate the Gondwana vicariance hypothesis, as the age of the tribe Exaceae was estimated to be around 40 MY and the divergence of Sebaea from other Exaceae was about 30 MY, i.e., far too young to match the vicariance with the breakup of Gondwana. Post-Gondwana dispersal/migration events might have been involved in the formation of the disjunctive distribution of the tribe across the Indian Ocean Basin. A possible land-bridge, the temporal dry-out in large areas of the Mozambique Channel between 45 and 26 MY revealed by recent geological studies, has been suggested as a possible channel for mammals to colonize Madagascar (McCall, 1997). This land-bridge could serve as an important biogeographic connection between continental Africa and Madagascar for the tribe Exaceae as well. The place of origin of the tribe Exaceae remains unresolved. Our present phylogeny suggested that the other members of Exaceae except for Sebaea likely originated in Madagascar. The genus Sebaea has its highest diversity in continental Africa, particularly in the Cape region, although it occurs throughout the entire paleotropical region. The phylogeny of this genus is still unknown, so the place of origin of Sebaea can not be suggested yet. Examples showed that the species richness in the Cape flora probably resulted from rapid 516 Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 and recent (7–8 MY) diversification (Cowling and Pressey, 2001; Richardson et al., 2001b). This might also be the case for Sebaea. A robust phylogeny of Sebaea is urgently needed to allow further elaboration on the origin of the tribe Exaceae, and further to confirm if the vicariance between Sebaea and other members of Exaceae could be the result of a migration through the Mozambique Channel land-bridge and subsequent separation. Acknowledgments The authors thank M. Callmander, L. Zeltner, and J. Piguet for providing important plant material. We are much indebted to Dr. Lena Struwe for sharing unpublished data and her critical comments and constructive suggestions on the draft. This research was financially supported by Swiss National Science Foundation (Grant 3100-052885), and partly by the Hundreds of Talents Program of Chinese Academy of Sciences. References Adams, L.G., 1995. Chionogentias (Gentianaceae), a new generic name for the Australasian ‘‘snow gentians,’’ and a revision of the Australian species. Austral. Syst. Bot. 8, 935–1011. Adams, L.G., 1996. Gentianaceae. In: Wilson, A. (Ed.), Flora of Australia, vol. 28. CSIRO, Melbourne, pp. 72–103. Albert, V.A., Struwe, L., 1997. Phylogeny and classification of Voyria (saprophytic Gentianaceae). Brittonia 49, 466–479. Andreasen, K., Baldwin, B.G., Bremer, B., 1999. Phylogenetic utility of the nuclear rDNA ITS region in subfamilyIxoroideae (Rubiaceae): comparisons with cpDNA rbcL sequence data. Plant Syst. Evol. 217, 119–135. APG (The Angiosperm Phylogeny Group). 1998. An ordinal classification for the families of flowering plants. Ann. Missouri Bot. Gard. 85, 531–553. Backlund, M., Oxelman, B., Bremer, B., 2000. Phylogenetic relationships within the Gentianales based on ndhF and rbcL sequences, with a particular reference to the Loganiaceae. Am. J. Bot. 87, 1029–1043. Boutique, R., 1972. Gentianaceae. In: Bamps, P. (Ed.), Flore dÕAfrique Centrale (Zaire-Ruanda-Burundi). Jardin Botanique National de Belgique, Brussels, pp. 1–56. Bremer, K., 2000. Early Cretaceous lineages of monocot flowering plants. Proc. Natl. Acad. Sci. USA 97, 4707–4711. Chase, M.W., Hills, H.H., 1991. Silica gel: an ideal material for field preservation of leaf samples for DNA studies. Taxon 40, 215–220. Chassot, P., Nemomissa, S., Yuan, Y.-M., K€ upfer, P., 2001. High paraphyly of Swertia L. (Gentianaceae) in the Gentianella-lineage as revealed by nuclear and chloroplast DNA sequence variation. Plant Syst. Evol. 229, 1–21. Chen, S.-L., He, T.-N., Liu, J.-Q., Hong, D.-Y., 2000a. Embryology of Crawfurdia delavayi (Gentianaceae) and its systematic value. Israel J. Plant Sci. 48, 113–119. Chen, S.-L., He, T.-N., Liu, J.-Q., Hong, D.-Y., 2000b. Embryology of Tripterospermum cordatum (Gentianaceae). Acta Bot. Yunnan. 22, 53–58. Cowling, R.M., Pressey, R.L., 2001. Rapid plant diversification: planning for an evolutionary future. Proc. Natl. Acad. Sci. USA 98, 5452–5457. Crepet, W.L., 1996. Timing in the evolution of derived floral characters: upper Cretaceous (Turonian) taxa with tricopate and tricolpate-derived pollen. Rev. Paleobot. Paleonol. 90, 339–359. Crepet, W.L., Daghlian, C.P., 1981. Lower Eocene and Paleocene Gentianaceae: floral and palynological evidence. Science 214, 75–77. Crepet, W.L., Nixon, K.C., Friis, E.M., Freudenstein, J.V., 1992. Oldest fossil flowers of hamamelidaceous affinity, from the late Cretaceous of New Jersey. Proc. Natl. Acad. Sci. USA 89, 8986–8989. Denduangboripant, J., Cronk, Q.C.B., 2001. Evolution and alignment of the hypervariable arm1 of Aeschynanthus (Gesneriaceae) ITS2 nuclear ribosomal DNA. Mol. Phylogenet. Evol. 20, 163–172. Doyle, J., Doyle, J.L., 1987. A rapid DNA isolation method for small quantities of fresh tissues. Phytochem. Bull. 19, 11–15. Farris, J.S., K€allersj€ o, M., Kluge, A.G., Bult, C., 1995. Testing significance of incongruence. Cladistics 10, 315–319. Felsenstein, J., 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39, 783–791. Fitch, W.M., 1971. Toward defining the course of evolution: minimum change for a specific tree topology. Syst. Zool. 20, 406–416. Gielly, L., Taberlet, P., 1996. A phylogeny of the European gentians inferred from chloroplast trnL (UAA) intron sequences. Bot. J. Linn. Soc. 120, 57–75. Gielly, L., Yuan, Y.-M., K€ upfer, P., Taberlet, P., 1996. Molecular evolution of noncoding regions in the genus Gentiana L.: chloroplast trnL (UAA) intron versus nuclear ribosomal internal transcribed spacer sequences. Mol. Phylogenet. Evol. 5, 460–466. Gilg, E., 1895. Gentianaceae. In: Engler, A., Prantl, K. (Eds.), Die nat€ urlichen Pflanzenfamilien, vol. 4(2). Leipzig, Engelmann, pp. 50–180. von Hagen, K.B., Kadereit, J.W., 2000. Notes on the systematics and evolution of Gentiana sect. Ciminalis. Bot. Jahrb. Syst. 122, 305–339. von Hagen, K.B., Kadereit, J.W., 2001. The phylogeny of Gentianella (Gentianaceae) and its colonization of the southern hemisphere as revealed by nuclear and chloroplast DNA sequence variation. Org. Divers. Evol. 1, 61–79. Ho, T.-N., Liu, S.-W., 1990. The infrageneric classification of Gentiana (Gentianaceae). Bull. Brit. Mus. (Nat. Hist.) Bot. 20, 169–192. Ho, T.-N., Xue, C.-Y., Wang, W., 1994. The origin, dispersal and formation of the distribution pattern of Swertia L. (Gentianaceae). Acta Phytotax. Sin. 32, 525–537. Ho, T.-N., Pringle, J.S., 1995. Gentianaceae. In: Wu, Z.Y., Raven, P. (Eds.), Flora of China, vol. 16. Missouri Botanical Garden, St. Louis, pp. 1–139. Humbert, H., 1937. Gentianothamnus, genre noveau de Gentianac
ees de Madagascar. Compt Rend. Hebd. S
eances Acad. Sci. 204, 1747–1749. Hungerer, K.B., Kadereit, J.W., 1998. The phylogeny and biogeography of Gentiana L. sect. Ciminalis (Adans.) Dumort.: a historical interpretation of distribution ranges in the European high mountains. Perspect. Plant. Ecol. Evol. Syst. 1, 121–135. Klackenberg, J., 1985. The genus Exacum (Gentianaceae). Opera Bot. 84, 1–144. Klackenberg, J., 1986. The new genus Ornichia (Gentianaceae) from Madagascar. Bull. Mus. Nat. Hist. Nat., Paris, s
er. 4, sect. B, Adansonia 8, 195–206. Klackenberg, J., 1987. Revision of the genus Tachiadenus (Gentianaceae). Bull. Mus. Nat. Hist. Nat. Paris, s
er. 4, sect. B, Adansonia 9, 43–80. Klackenberg, J., 1990. Famille 168. Gentianac
ees oipp. In: Morat, P. (Ed.), Flore de Madagascar et des Comores. Association de Botanique tropicale, Paris, pp. 1–185. Y.-M. Yuan et al. / Molecular Phylogenetics and Evolution 28 (2003) 500–517 Klackenberg, J., 2002. Tribe Exaceae. In: Struwe, L., Albert, V.A. (Eds.), Gentianaceae—Systematics and Natural History. Cambridge University Press, Cambridge, pp. 66–108. Liu, S.-W., Ho, T.-N., 1992. Systematic study on Lomatogonium A. Br. (Gentianaceae). Acta Phytotax. Sin. 30, 289–319. Magallon, S., Crane, P.R., Herendeen, P.S., 1999. Phylogenetic pattern, diversity, and diversification of eudicots. Ann. Missouri Bot. Gard. 86, 297–372. Manchester, S.R., 2002. Leaves and fruits of Davidia(Cornales) from the Paleocene of North America. Syst. Bot. 27, 368–382. Manchester, S.R., Crane, P.R., Golovneva, L.B., 1999. An extinct genus with affinities to extant Davidiaand Camptotheca (Cornales) from the Plaeocene of North America and eastern Asia. Intern. J. Plant Sci. 160, 188–207. McCall, R.A., 1997. Implications of recent geological investigations of the Mozambique Channel for the mammalian colonization of Madagascar. Proc. R. Soc. Lond. B 264, 663–665. McLoughlin, S., 2001. The breakup history of Gondwana and its impact on pre-Cenozoic floristic provincialism. Aust. J. Bot. 49, 271–300. M
esz aros, S., 1994. Evolutionary significance of xanthones in Gentianaceae: a reappraisal. Biochem. Syst. Ecol. 22, 85–94. M
esz
aros, S., De Laet, J., Smets, E., 1996. Phylogeny of temperate Gentianaceae: a morphological approach. Syst. Bot. 21, 153–168. Muller, J., 1984. Significance of fossil pollen for angiosperm history. Ann. Missouri Bot. Gard. 71, 419–443. Muse, S.V., Weir, B.S., 1992. Testing for equality of evolutionary rates. Genetics 132, 269–272. Nixon, K.C., Crepet, W.L., 1993. Late Cretaceous fossil flowers of Ericalean affinity. Am. J. Bot. 80, 616–623. Paiva, J., Noguiera, I., 1990. Gentianaceae. In: Launert, E., Pope, G.V. (Eds.), Flora Zambesiaca, vol. 7, part 4. Flora Zambesiaca Managing Committee, Kew, pp. 3–51. Patterson, T.B., Givnish, T.J., 2002. Phylogeny, concerted convergence, and phylogenetic niche conservatism in the core Liliales: insights from rbcL and ndhF sequence data. Evolution 56, 233–252. Posada, D., Crandall, K.A., 1998. Modeltest: testing the model of DNA substitution. Bioinformatics 14, 817–818. Rambaut, A., Charleston, M., 2000. TreeEdit: Phylogenetic Tree Editor. University of Oxford, Oxford. Richardson, J.E., Pennington, R.T., Pennington, T.D., Hollingsworth, P.M., 2001a. Rapid diversification of a species rich genus of Neotropical rain forest trees. Science 293, 2242–2245. Richardson, J.E., Weitz, F.M., Fay, M.F., Cronk, Q.C.B., Linder, H.P., Reeves, G., Chase, M.W., 2001b. Rapid and recent origin of species richness in the Cape flora of South Africa. Nature 412, 181–183. Rova, J.H.E., Delprete, P.G., Andersson, L., Albert, V.A., 2002. A trnL–F cpDNA sequence study of the Condamineeae–Rondeletieae–Sipaneeae complex with implications on the phylogeny of the Rubiaceae. Am. J. Bot. 89, 145–159. Sanderson, M.J., 1997. A nonparametric approach to estimating divergence times in the absence of rate constancy. Mol. Biol. Evol. 14, 1218–1231. Schonenberger, J., Friis, E.M., 2001. Fossil flowers of ericalean affinity from the Late Cretaceous of Southern Sweden. Am. J. Bot. 88, 467–480. Stockey, R.A., Manchester, S.R., 1986. A fossil flower with in situ Pistillipollenites from the Eocene of British Columbia. Can. J. Bot. 66, 313–318. Struwe, L., Albert, V.A., 1997. Floristics, cladistics, and classification: three case studies in Gentianales. In: Dransfield, J., Coode, M.J.E., Simpson, D.A. (Eds.), Plant Diversity in Malesia III. Royal Botanic Gardens, Kew, pp. 321–352. Struwe, L., Albert, V.A., 1998a. Lisianthius (Gentianaceae), its probable homonym Lisyanthus, and the priority of Helia over Irlbachia as its substitute. Harvard Pap. Bot. 3, 63–71. 517 Struwe, L., Albert, V.A., 1998b. Six new species of Gentianaceae from the Guayana Shield. Harvard Pap. Bot. 3, 181–197. Struwe, L., Albert, V.A., 2002. Tribe Helieae. In: Struwe, L., Albert, V.A. (Eds.), Gentianaceae—Systematics and Natural History. Cambridge University Press, Cambridge, pp. 137–190. Struwe, L., Albert, V.A., Bremer, B., 1994. Cladistics and family level classification of the Gentianales. Cladistics 10, 175–206. Struwe, L., Maas, P.J.M., Albert, V.A., 1997. Aripuana cullmaniorum, a new genus and species of Gentianaceae from white-sands of southeastern Amazonas, Brazil. Harvard Pap. Bot. 2, 235–253. Struwe, L., Thiv, M., Kadereit, J.W., Pepper, A.S.-R., Motley, T.J., White, P.J., Rova, J.H.E., Potgieter, K., Albert, V.A., 1998. Saccifolium (Saccifoliaceae), an endemic of Sierra de La Neblina on the Brazilian-Venezuelan frontier, is related to a temperate-alpine lineage of Gentianaceae. Harvard Pap. Bot. 3, 199–214. Struwe, L., Kadereit, J.W., Klackenberg, J., Nilsson, S., Thiv, M., von Hagen, K.B., Albert, V.A., 2002. Systematics, character evolution, and biogeography of Gentianaceae, including a new tribal and subtribal classification. In: Struwe, L., Albert, V.A. (Eds.), Gentianaceae—Systematics and Natural History. Cambridge University Press, Cambridge, pp. 21–309. Sun, G., Dilcher, D.L., Zheng, S., Zhou, Z., 1998. In search of the first flower: a Jurassic angiosperm, Archaefructus, from Notheast China. Science 282, 1692–1695. Swofford, D.L., 2000. PAUP*. Phylogenetic Analysis Using Parsimony (* and Other Methods). Version 4. Sinauer Associates, Sunderland, MA. Taberlet, P., Gielly, L., Patou, G., Bouvet, J., 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Mol. Biol. 17, 1105–1109. Takahashi, M., Crane, P.R., Ando, H., 1999. Fossil flowers and associated plant fossils from the Kamikitaba locality (Ashizawa Formation, Futaba Group, Lower Coniacian, Upper Cretaceous) of northeast Japan. J. Plant Res. 112, 187–206. Thiv, M., Struwe, L., Albert, V.A., Kadereit, J.W., 1999a. The phylogenetic relationships of Saccifolium bandeirae Maguire and Pires (Gentianaceae) reconsidered. Harvard Pap. Bot. 4, 519–526. Thiv, M., Struwe, L., Kadereit, J.W., 1999b. The phylogenetic relationships and evolution of the Canarian laurel forest endemic Ixanthus viscosus (Alt.) Griseb. (Gentianaceae): evidence from matK and ITS sequence variation, and floral morphology and anatomy. Plant Syst. Evol. 218, 299–317. Thompson, J.D., Gibson, T.J., Plewniak, F., Jeanmougin, F., Higgins, D.G., 1997. The ClustalX windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 24, 4876–4882. Thulin, M., 2001. Exacum (Gentianaceae) on the Arabian peninsula and Socotra. Nord. J. Bot. 21, 243–247. White, T.L., Bruns, T., Lee, S., Taylor, J., 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In: Innis, M.A., Gelfand, D.H., Sninsky, J.J., White, T.J. (Eds.), PCR Protocols. Academic Press, San Diego, pp. 315–322. Wolfe, K.H., Morden, C.W., Palmer, J.D., 1992. Function and evolution of a minimal plastid genome from s nonphotosynthetic parasitic plant. Proc. Natl. Acad. Sci. USA 89, 10648–10652. Yuan, Y.-M., K€ upfer, P., 1995. Molecular phylogenetics of the subtribe Gentianinae (Gentianaceae) inferred from the sequences of internal transcribed spacers (ITS) of nuclear ribosomal DNA. Plant Syst. Evol. 196, 207–226. Yuan, Y.-M., K€ upfer, P., 1997. The monophyly and rapid evolution of Gentiana sect. Chondrophyllae s. l.: evidence from the sequences of internal transcribed spacers (ITS) of nuclear ribosomal DNA. Bot. J. Linn. Soc. 123, 25–43. Yuan, Y.-M., K€ upfer, P., Doyle, J.J., 1996. Infrageneric phylogeny of the genus Gentiana (Gentianaceae) inferred from nucleotide sequences of the internal transcribed spacers (ITS) of nuclear ribosomal DNA. Am. J. Bot. 83, 641–652.