Blackwell Science, LtdOxford, UKBOJBotanical Journal of the Linnean Society0024-4074The Linnean Society of London, 2003 141 Original Article Botanical Journal of the Linnean Society, 2003, 141, 349–363. With 5 figures A. W. MEEROW ET AL. MOLECULAR PHYLOGENY OF CRINUM Phylogeny and biogeography of Crinum L. (Amaryllidaceae) inferred from nuclear and limited plastid non-coding DNA sequences ALAN W. MEEROW1,2*, DAVID J. LEHMILLER3 and JASON R. CLAYTON1 1 USDA-ARS-SHRS, National Germplasm Repository, 13601 Old Cutler Road, Miami, Florida 33156, USA 2 Fairchild Tropical Garden, 10901 Old Cutler Road, Miami, Florida 33158 USA 3 550 IH-10 South, Suite 201, Beaumont, Texas 77707 USA Received June 2002; accepted for publication November 2002 The genus Crinum L. is the only pantropical genus of the Amaryllidaceae. Phylogenetic and biogeographical analyses of nrDNA ITS and plastid trnL-F sequences for all continental groups of the genus Crinum and related African genera are presented, with the genus Amaryllis used as outgroup. ITS indicates that C. baumii is more closely related to Ammocharis and Cybistetes than to Crinum sensu stricto. Three clades are resolved in Crinum s.s. One unites a monophyletic American group with tropical and North African species. The second includes all southern African species and the Australian endemic C. flaccidum. The third includes monophyletic Madagascar, Australasian and Sino-Himalayan clades, with southern African species. The trnL-F phylogeny resolves an American and an Asian/Madagscar clade, and confirms the relationship of C. flaccidum with species endemic to southern Africa. The salverform, actinomorphic perianths of subg. Crinum appear to have evolved several times in the genus from ancestors with zygomorphic perianths (subg. Codonocrinum), thus neither subgenus is monophyletic. Biogeographical analyses place the origin of Crinum in southern Africa, as the region is optimized at all ancestral nodes in the tree topology, and in basal interior nodes of all but one of the major clades. The genus underwent three major waves of radiation corresponding to the three main clades resolved in our trees. Two entries into Australia for the genus are indicated, as are separate Sino-Himalayan and Australasian dispersal events. © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363. ADDITIONAL KEYWORDS: molecular systematics – Africa – geophytes – dispersal – cladistics – monocotyledons. INTRODUCTION The genus Crinum L. is the only pantropical genus of the Amaryllidaceae, with species occurring in Africa, America, Asia, and Australia. Crinum has seeds well adapted for oceanic dispersal (Koshimizu, 1930), which are considered to have engendered its broad distribution (Arroyo & Cutler, 1984). The genus is most speciose in Africa, particularly sub-Saharan Africa (Nordal, 1977), and its systematic affinities are with a *Corresponding author. E-mail: miaam@ars-grin.gov group of entirely African endemic genera, constituting the tribe Amaryllideae (Snijman & Linder, 1996; Meerow & Snijman, 1998, 2001). This tribe is extremely well-marked by numerous morphological synapomorphies, such as extensible fibres in the leaf tissue, bisulculate pollen with spinulose exines, scapes with a sclerenchymatous sheath, unitegmic or ategmic ovules, and nondormant, water-rich, nonphytomelanous seeds with chlorophyllous embryos (Snijman & Linder, 1996). In plastid DNA sequence analyses of the Amaryllidaceae (Ito et al., 1999; Meerow et al., 1999), this tribe is the first branch to resolve in the family and receives high bootstrap support (>90%). © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 349 350 A. W. MEEROW ET AL. Crinum was established by Linneaus (1753). Herbert (1837) divided the genus into two sections based on the degree to which the tepals are patent. Baker (1881) divided the genus into three subgenera: Stenaster, with salverform, actinomorphic perianths, straight floral tubes and linear segments; Platyaster, similar to the former but with lanceolate segments; and Codonocrinum with funnel-form, zygomorphic flowers and curved tubes. He later (Baker, 1898) submerged Platyaster into subgenus Stenaster, which must be called subgenus Crinum since it contains the type species, C. americanum L. Crinum was not thoroughly monographed in the last century, except for a synoptic update of Baker (1881) by Uphof (1942). It has been the subject of regional treatments (e.g. Hepper, 1968; Geerinck, 1973; Verdoorn, 1973; Nordal, 1977, 1982, 1986; Nordal & Wahlstrom, 1980; Lehmiller, 1997a,b) and it is estimated that the genus contains 60–70 species (Fangan & Nordal, 1993). Fangan & Nordal (1993) used DNA RFLPs to investigate the phylogeny of seven species. Meerow & Snijman (2001) included four species in their ITS analysis of the tribe Amaryllideae. In this paper, we investigate the phylogeny of 38 species of Crinum, representing all of the continental groups within the genus, using nuclear rDNA ITS sequences (ITS1 spacer, 5.8S intron, ITS2 spacer), and plastid trnL-F sequences, and discuss the results in a biogeographical and evolutionary context. MATERIAL AND METHODS SAMPLING Genomic DNA was extracted from silica gel-dried leaf tissue of the taxa listed in Table 1, as described by Meerow et al. (2000). DNA EXTRACTION, AMPLIFICATION AND SEQUENCING PROTOCOLS The trnL-trnF region was amplified and sequenced using the primers of Taberlet et al. (1991) as described by Meerow et al. (1999). Amplification of the ribosomal DNA ITS1/5.8S/ITS2 region was accomplished using flanking primers (18S, 26S) AB101 and AB102 (Douzery et al., 1999), and the original White et al. (1990) internal primers ITS2 and 3 to amplify the spacers along with the intervening 5.8S intron as described by Meerow et al. (2000). All polymerase chain reaction (PCR) amplifications were performed on an ABI 9700 (Perkin-Elmer Applied Biosystems, Foster City, California, USA). Amplified products were purified using QIAquick (Qiagen, Valencia, California, USA) columns, following the manufacturer’s protocols. Cycle sequencing reactions were performed directly on purified PCR products on the ABI 9700, using standard dideoxycycle protocols for sequencing with dye terminators on either an ABI 377 or ABI 310 automated sequencer (according to the manufacturer’s protocols; Applied Biosystems, Foster City, California, USA). SEQUENCE ALIGNMENT Both the ITS and trnL-F matrices contained few gaps and were readily aligned manually using Sequencher 4.1 (Gene Codes, Ann Arbor, Michigan, USA). The alignment is accessible through GenBank or from the first author. PHYLOGENETIC ANALYSES The ITS matrix consisted of 43 taxa (38 Crinum species, two species of Amaryllis and Ammocharis, and Cybistetes longifolia). Amaryllis was designated as outgroup. Amaryllis is sister to all other members of the tribe Amaryllideae (Meerow & Snijman, 2001). Ammocharis and Cybistetes are the only other members of the subtribe Crininae that together form a sister clade to Crinum (Meerow & Snijman, 2001). The plastid trnL-F matrix consisted of 19 taxa, 16 species of Crinum and one species each of Amaryllis (A. belladonna), Ammocharis (A. coranica), and Cybistetes longifolia. We only sampled one to several taxa from each of the clades resolved by ITS, and saw little likelihood that additional sampling would increase information content of the trnL-F data matrix. In the combined data set, taxa with only one sequence were coded as missing data for the absent sequence. Aligned matrices were analysed using the parsimony algorithm of PAUP* for Macintosh (version 4.0b10; Swofford, 1998), with the MULPARS option invoked. Tree branches were retained only if unambiguous support was available (i.e. branches were collapsed if the minimum length = 0). The few gaps were coded as missing characters, as developing a binary-coded strict homology gap matrix added no further resolution to the trees in preliminary analyses. For all matrices, a heuristic search was conducted under the Fitch (equal) weights (Fitch, 1971) criterion with 1000 random sequence additions (Maddison, 1991) and tree bisection and reconnection (TRB) branch swapping. We permitted only 20 trees to be held at each step to reduce the time spent searching trees at suboptimal levels. All minimal trees collected in the 1000 replicates were then swapped to completion. We combined the data matrices, opting for the total evidence approach (Seelanan, Schnabel & Wendel, 1997; Dubuisson, Hebant-Mauri & Galtier, 1998). Though the trnL-F matrix yielded only a few parsi- © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 MOLECULAR PHYLOGENY OF CRINUM mony informative base substitutions, it nonetheless confirmed some of the relationships well-supported by the ITS matrix, and inclusion in this study is therefore warranted. However, before combining the ITS and trnL-F data sets, we performed a partition homogeneity test on the matrices (Farris et al., 1994, 1995) to assess the degree of congruence between them. Five hundred heuristic searches were conducted, each with ten random addition replications, saving 20 trees from each for TBR branch-swapping. Internal support was determined by bootstrapping (Felsenstein, 1985; 5000 replicates with simple addition) and calculation of Bremer (1988) decay indices (DI) using the program TreeRot vs. 2.1 (Sorenson, 1996). The cut-off bootstrap percentage is 50. A bootstrap value greater than 75% was considered good support, 65–75% was designated moderate support, and less than 65% as weak (Meerow & Snijman, 2001; Meerow et al., 2002). Five hundred replicate heuristic searches were implemented for each constraint statement postulated by TreeRot, saving ten trees per replicate. A minimum DI = 2 was considered to represent good support for a clade. BIOGEOGRAPHICAL ANALYSES The biogeographical patterns inferred from our gene trees were assessed using both Fitch optimization (Maddison, Ruvolo & Swofford, 1992) with MacClade version 4.03 (Maddison & Maddison, 2001) and the dispersal-variance method of analysis (Ronquist, 1997) using the program DIVA version 1.1 (Ronquist, 1996). The program uses vicariance (i.e. allopatric speciation) in its optimization of ancestral distributions but takes into consideration dispersal and extinction events, and indicates their direction (Ronquist, 1996, 1997). The most-parsimonious reconstructions minimize such events. Unlike Fitch optimization, DIVA does not restrict widespread distributions to terminals or limit ancestral distributions to single areas (Ronquist, 1996). DIVA requires a fully bifurcated tree for analysis, thus one tree (with zero-length branches not collapsed) from the combined sequence analysis was used for optimization of 15 coded geographical areas (Table 1). It must be noted that this tree was one of 150 equally parsimonious topologies, and contains zero-length internal branches. Fitch optimization of area data was performed on the same tree using a single multistate character (Table 1). Two DIVA analyses were performed, one without limits on the maximum areas allowed for ancestral nodes (15), and another in which the limit was set to the minimum (2) to reduce ambiguities at the more basal nodes of the tree. An exact optimization (vs. heuristic) was invoked in both analyses by allowing the maximum number of alternative reconstructions to be held at any node. 351 RESULTS ITS Of the 807 characters (ITS1, 5.8S intron, ITS2) included in the analyses, 122 were parsimony informative. The search found 119 equally mostparsimonious trees of length = 322, consistency index (CI) = 0.79 and retention index (RI) = 0.89. Two very well-supported clades are resolved by ITS (Fig. 1). The smaller of the two, with a bootstrap of 100% and DI = 11 unites Crinum baumii as sister to Ammocharis and Cybistetes. The position of Cybistetes as sister to Ammocharis is only weakly supported (bootstrap = 58%, DI = 1). The second, larger clade with strong support (bootstrap = 99, DI = 9) consists of all of the remaining Crinum species. Three subclades of Crinum species form a trichotomy within this larger group. The first (A), with a bootstrap = 64 and DI = 2, includes a wellsupported American group (bootstrap = 99%, DI = 7) embedded among strictly tropical and north African species. No species endemic to South Africa included in the analysis resolves within this clade. Crinum distichum, C. humile and C. jagus form a well-supported group (bootstrap = 100%, DI = 9), as does C. kirkii and C. politfolium (bootstrap = 90%, DI = 3). The second subclade (B) is the best supported of the three (bootstrap = 91%, DI = 4). With the exception of the Australian endemic C. flaccidum, the clade consists entirely of African species, including the only South African endemics (e.g. C. acaule, C. bulbispermum, C. campanulatum, C. moorei, C. variabile) included in the analysis. Outside of well-supported sister relationships between C. acaule and C. forbesii, and between C. bulbispermum and C. variabile, the clade is largely unresolved. The third subclade (C), with a bootstrap = 63 and DI = 2, unites the Asiatic species with the Madagascar endemics. The West African C. fimbriatulum is sister to all other members of the clade. The Sino-Himalayan species (C. defixum and C. sp., China) and the Australasian species form separate subclades, both with a bootstrap of 65% and DI = 1. Three of the four Australasian species form a well-supported monophyletic group with a bootstrap of 99% and DI = 5. The Madagascar species resolve as monophyletic with moderate support (bootstrap = 73%, DI = 2). They form a polytomy with the Australasian group, and two broadly distributed African species, C. subcernuum (Mozambique, Namibia, Tanzania) and C. buphanoides (southwest Africa, Angola, Transvaal region of South Africa). PLASTID TRNL-F Of the 940 characters included in the analysis, only eight were parsimony informative. Five equally © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 352 © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 GenBank accession no. or literature citation Taxon Voucher specimen Provenance ITS Amaryllis belladonna L. M. W. Chase 612 (K) Western Cape, South Africa A. paradisicola Snijman van Jaarsveld 13263 (NBI) Meerow 2320 Western Cape, South Africa Eastern Cape, South Africa Meerow & Snijman (2001) Meerow & Snijman (2001) Meerow & Snijman (2001) AY139116 AY139117 AY139118 AY139119 AY139120 AY139121 AY139122 AY139123 Ammocharis coranica Herb. A. nerinoides (Baker) Lehmiller Crinum abyssinicum Hochst. ex A. Rich. C. acaule Baker C. americanum L. C. asiaticum L. C. baumii Harms. C. broussonetii (Redouté) Herb. C. bulbispermum (Burm.) Milne-Redhead & Schweickerdt C. buphanoides Welw. ex Baker C. campanulatum Herb. Meerow 2321 Meerow 2322 Meerow 2338 Meerow 2323 Meerow 2334 Van Zyl 99.B (PRE) Meerow 2324 Meerow 2339 Namibia Ethiopia Natal, South Africa Florida, USA cultivation, Florida, USA Namibia Chad Natal, South Africa Meerow 2325 Meerow 2337 Transvaal, South Africa Eastern Cape, South Africa C. carolo-schmidtii Dinter C. crassicaule Baker C. cruentum Ker Gawl. C. defixum Ker Gawl. C. distichum Herb. C. erubescens Sol. C. fimbriatulum Baker C. flaccidum Herb. C. forbesii (Lindl.) Schult. emend. Herb. Meerow 2340 Meerow 2341 T. M. Howard s. n. Traub 1235 (MO) Meerow 2326 T. M. Howard s. n. Leach 14510 (PRE) Meerow 2327 Meerow 2328 Namibia Namibia Oaxaca, Mexico Nepal Chad Brazil Angola Australia Transvaal, South Africa AY139124 Meerow & Snijman (2001) AY139125 AY139126 AY139127 AY139128 AY139129 AY139130 AY139131 AY139132 AY139133 trnL-F Area code1 A A AY139152 AY139153 AY139154 AY139155 AY139156 AY139157 AY139158 AY139159 AY139160 AY139161 AY139162 AY139163 ACE B D A G M B C A AB A A A G L C HI B N A 1 A = South Africa, B = South-west Africa, C = tropical Africa, D = North Africa, E = East Africa, F = Madagascar, G = North America, H = Central America, I = South America, J = Cuba, K = China, L = Nepal, M = Australasia (south-east Asia and Pacific), N = Australia, O = India. A. W. MEEROW ET AL. Table 1. Species, voucher specimens, GenBank sequence accession numbers (or previous citation), and geographical area codes used in the phylogenetic analyses of Crinum. Vouchers are deposited at FTG unless otherwise stated Taxon Voucher specimen Provenance ITS C. humilis A. Chev. C. jagus Thomps. C. kirkii Baker C. latifolium Andr. C. ligulatum Baker C. macowanii Baker Meerow 2329 Meerow 2330 Meerow 2342 Meerow 2343 Hardy 2995 (PRE) Meerow 2344 Cameroun cultivation, Florida, USA Tanzania India Madagascar Kenya C. mauritianum Lodd. C. modestum Baker C. moorei Hook. F. C. oliganthum Urb. C. pedunculatum R. Br. C. politifolium R.Wahlstr. C. razafindratsiri Lehmiller Madagascar Madagascar Natal, South Africa Cuba Australia Tanzania Madagascar C. subcernuum Baker C. variabile Herb. Hardy s. n. (PRE) Meerow 2345 Meerow 2346 Meerow 2336 Meerow 2335 Meerow 2347 Lehmiller 1944 (TAMU) Meerow 2348 Meerow 2331 AY139134 AY139135 AY139136 AY139137 AY139138 Meerow & Snijman (2001) AY139139 AY139140 AY139141 AY139142 AY139143 AY139144 AY139145 C. venosum Baker C. yemense Deflers Meerow 2349 M. W. Chase 1595 (K) Northern Australia Yemen C. sp., Borneo C. sp., SW China C. sp., Peru Cybistetes longifolia (L.) Milne-Redh. & Schweick. Meerow 2332 Meerow 2333 Schunke 14054 Duncan 304 (NBG) Borneo Yunnan, China Peru Western Cape, South Africa Namibia Western Cape, South Africa AY139150 Meerow & Snijman (2001) AY139146 AY139151 AY139147 AY139148 AY139149 Meerow & Snijman (2001) trnL-F AY139164 AY139165 AY139166 AY139167 Area code1 C C C O F AE F F A J N C F BE A Meerow et al. (1999) AY139168 AY139169 N D M K I A A = South Africa, B = South-west Africa, C = tropical Africa, D = North Africa, E = East Africa, F = Madagascar, G = North America, H = Central America, I = South America, J = Cuba, K = China, L = Nepal, M = Australasia (south-east Asia and Pacific), N = Australia, O = India. 1 MOLECULAR PHYLOGENY OF CRINUM © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 GenBank accession no. or literature citation 353 354 A. W. MEEROW ET AL. Figure 1. One of 119 most-parsimonious trees found by phylogenetic analysis of nrDNA ITS sequences across 43 species of Crinum and related genera. Numbers above branches are branch lengths. Numbers below branches are bootstrap percentages and decay indices (italic), respectively. Dashed lines are zero-length branches. A white bar across a branch signifies a collapsed node in the strict consensus of all trees. parsimonious trees were found of length = 42 steps, CI = 1.00, RI = 1.00 (Fig. 2). The tree is mostly unresolved, but four clades are supported. A tropical Australasian clade (bootstrap = 64%, DI = 1) is resolved as sister to the single Malagasy species included, C. ligulatum (bootstrap = 64%, DI = 1). An American clade is resolved with a bootstrap of 62% and DI = 1. Finally, South African C. campanulatum and Australian C. flaccidum form a monophyletic group with a bootstrap of 92% and DI = 2. COMBINED total of 150 equally parsimonious trees were found by the heuristic search of length = 363, CI = 0.81, RI = 0.89. Not surprisingly, the trees (Fig. 3) are very similar in topology to those found using ITS alone. However, bootstrap support for the position of C. flaccidum as sister to an entirely African clade rises to 87%, whereas support for the Sino-Himalayan clade drops to 58%. The position of this clade as sister to the Australasian–Madagascar–C. buphanoides– C. subcernuum clade is slightly better supported and does not collapse in strict consensus. ANALYSIS The P-value = 0.995, indicating a high level of congruence between the ITS and trnL-F matrices, though this value is compromised by the lower sampling level for trnL-F compared to ITS, and relative paucity of informative base substitutions in the plastid data. A BIOGEOGRAPHICAL ANALYSIS While allowing the maximum area assignments (15) to ancestral nodes in DIVA produced an area optimization requiring only 17 dispersal events, it did so by assigning the more basal nodes every possible area. © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 MOLECULAR PHYLOGENY OF CRINUM 355 (subg. Codonocrinum, Figs 1,3,4), but at the next node, a Sino-Himalayan clade with actinomorphic subg. Crinum flower morphology (Fig. 5) is sister to an entirely subg. Crinum clade, including two species from southern Africa (C. buphanoides and C. subcernuum) whose exact relationships are unresolved. (Figs 1,3). DISCUSSION Figure 2. One of five most-parsimonious trees found by phylogenetic analysis of plastid trnL-F sequences across 19 species of Crinum and related genera. Numbers above branches are branch lengths. Numbers below branches are bootstrap percentages and decay indices (italic), respectively. Dashed lines are zero-length branches. Constraining the program to maxareas = 2 yields a more realistic scenario, but the cost raises the number of dispersal events to 23. The latter optimization is referenced in the discussion (Fig. 4). Even with maxareas set to 2, one node was assigned four equally parsimonious areas, and one received eight possible assignments. Both Fitch optimization and DIVA were largely congruent (Fig. 4). However, Fitch optimization places the origins of the lineage in South Africa, while DIVA found an origin in south-western Africa equally parsimonious. The ancestral origin of clade A is equivocal by Fitch optimization, and DIVA assigns eight possible area assignments, all of which include either South Africa or south-western Africa and respective dispersal to tropical Africa (C) North Africa (D), North America (G), and (J) Cuba. However, clade A is the only one of the three from which South Africa disappears from either the internal nodes or a terminal distribution (the distribution of C. buphanoides in clade C includes South Africa). In clade B, South Africa persists as an area component through internal nodes as well (Fig. 4). Clade B contains most of the endemic South African species included in the analyses. Clade C is rooted in south-western Africa by C. fimbriatulum Meerow et al. (1999), using three plastid DNA sequences, found that the Amaryllideae, the tribe of Amaryllidaceae to which Crinum belongs, was the most robustly supported clade in their parsimony topologies and was sister to the rest of the Amaryllidaceae. Meerow & Snijman (2001), combining morphological characters and ITS sequences, analysed the Amaryllideae and resolved a well-supported monophyletic subtribe, Crininae, consisting of Crinum, Ammocharis and Cybistetes. Only four Crinum species were included in that analysis. Crinum is the largest genus in the tribe, and the only one to disperse outside of Africa. The fleshy, floating, and salt-resistant seed of Crinum has been implicated as the likely agent of its dispersal success (Koshimizu, 1930; Arroyo & Cutler, 1984). Fangan & Nordal (1993) performed restriction fragment length polymorphism (RFLP) analysis on seven species of Crinum, using Pancratium canariensis as outgroup. Pancratium, as the authors concede, was probably not the best choice of outgroup for the polarization of character states in Crinum, as it is only distantly related to the ingroup. In fact, it is part of the monophyletic Eurasian clade of Amaryllidaceae that branches more terminally than Amaryllideae in plastid DNA based trees of the entire family (Meerow et al., 1999). Despite the limited sampling, Fangan & Nordal’s (1993) RFLP topology resolved some geographically congruent clades. A West African species clade (C. glaucum, C. jagus and C. zeylanicum sensu Nordal) resolved as a monophyletic sister group to a clade inclusive of the East African and Indo-Pacific species that they sampled (2 spp.). Crinum asiaticum (subg. Crinum) resolved as sister to C. latifolium (subg. Codonocrinum) within this latter clade. These results were in contrast to their accompanying cladistic analysis of the same species across 11 morphological characters with the genus Ammocharis used as outgroup. In these analyses, the species of subgenus Codonocrinum formed a monophyletic group that was sister to C. asiaticum, the only representative of subg. Crinum included. Moreover, morphology placed C. zeylanicum and C. latifolium in the same trichotomous clade with C. politifolium. The former two species have had a long and confused taxonomic history (Nordal, 1977; Dassanayake, 1981) in terms of the © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 356 A. W. MEEROW ET AL. Figure 3. One of 150 most-parsimonious trees found by phylogenetic analysis of combined nrDNA ITS and plastid trnL-F sequences across 43 species of Crinum and related genera. Numbers above branches are branch lengths. Numbers below branches are bootstrap percentages and decay indices (italic), respectively. Dashed lines are zero-length branches. A white bar across a branch signifies a collapsed node in the strict consensus of all trees. identity of C. latifolium relative to C. zeylanicum (Fangan & Nordal, 1993). Crinum latifolium as treated by Fangan & Nordal (1993) is the only Codonocrinum species found outside Africa other than C. flaccidum that has not yet been proven to have been introduced by humans. To confuse matters further, the Linnean name C. latifolium is sometimes erroneously applied to a Crinum in southern China with radially symmetrical, salverform flowers (Zhanhe & Meerow, 2001). In our analyses, C. latifolium, despite its occurrence in India and Sri Lanka, is well nested in the tropical African/American clade (clade A, Figs 1,3,4). There are many tropical East African Crinum species that are similar in morphology to C. latifolium, and a putative scenario of how this species arrived in India and Sri Lanka is discussed later in this paper. Finally, Fangan & Nordal advised that, on the basis of their RFLP topology, the characteristic floral morphology of either subgenera Crinum or Codonocrinum may have evolved more than once or reversals for this character were possible. Our ITS (Fig. 1) and combined (Fig. 3) phylogenies support Fangan & Nordal’s (1993) suggestion based on RFLPs that neither morphologically based subgenus Crinum nor Codonocrinum is monophyletic. At issue is which state is plesiomorphic in the genus. Snijman & Linder (1996) concluded that zygomorphy is the ancestral state in tribe Amaryllideae, and Meerow et al. (1999) hypothesized that perianth symmetry in Amaryllidaceae is under simple genetic control and © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 MOLECULAR PHYLOGENY OF CRINUM 357 Figure 4. One of 150 equally most-parsimonious trees found by phylogenetic analysis of combined nrDNA ITS and plastid trnL-F sequences across 43 species of Crinum and related genera showing optimization of bigeographical data. Fitch optimization is indicated by colour or pattern; divergence-vicariance optimization is coded by small letters at ancestral nodes. Central America and East Africa are not visible in the Fitch optimization because they figure only in ambiguous area optimizations of terminal taxa. easily modified. The sister group to Crinum (Ammocharis and Cybistetes; Meerow & Snijman, 2001) has zygomorphic perianths. However, Crinum baumii, which resolves strongly as part of the sister group to Crinum, has floral morphology typical of subg. Crinum. Amaryllis, sister to all remaining genera of tribe Amaryllideae (Meerow & Snijman, 2001), has zygomorphic perianths. Fitch optimization of this character onto a tree from our combined analysis (Fig. 5) suggests that actinomorphy is the apomorphic state and has evolved several times, once within the clade inclusive of the Asian and Madagascar species, once in the American clade, and yet again in the sister clade to Crinum s.s. Both C. buphanoides and C. subcernuum have subg. Crinum-type flowers, though their exact relationships to both the Malagasy and Asiatic clades are not resolved (Figs 1,3). There are only two species from tropical Africa with subg. Crinum floral morphology, C. natans Baker and C. purpurascens Herb., which unfortunately were not available for sequencing. They may well represent the African sister group to the American clade. Interestingly, both species are emergent aquatics as are most of the American species. Fangan & Nordal (1993) also referred to a plastid RFLP-supported connection between East African (C. macowanii and C. politifolium) and ‘Indo-Pacific’ (C. asiaticum) Crinum. As no bootstrap or other confidence estimate was provided, the relative robustness of this clade could not be determined. However, our much larger sampling of species resolves C. macowanii and C. politfolium each in two clades, respectively (A and B, Figs 1,3,4) other than the one containing all of the Asian species that we sampled (clade C, Figs 1,3,4). The African mainland species within clade C wherein all of the Asian (both SinoHimalayan and Australasian) species resolve are south-western and southern African in the main (C. subcernuum, described from Mozambique, has been collected in Namibia by the second author, and was collected once in Tanzania according to Nordal, 1977). © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 358 A. W. MEEROW ET AL. Figure 5. Fitch optimization of perianth morphology on one of 150 equally most-parsimonious trees found by phylogenetic analysis of combined nrDNA ITS and plastid trnL-F sequences across 43 species of Crinum and related genera. Crinum baumii is sister to Ammocharis and Cybistetes with bootstrap support of 100% and DI = 11 (Figs 1,4). Milne-Redhead & Schweickerdt (1939) transferred Crinum baumii into Ammocharis without any detailed justification, and also established the monotypic genus Cybistetes as a segregate from the latter, largely on the basis of infructescence structure. In Cybistetes, the entire infructescence of indehiscent fruits functions as the dispersal unit (anemogeochory of Van der Pijl, 1982), whereas in Ammocharis the fruits are dehiscent and the infrutescence lax (Snijman & Williamson, 1994). Both Ammocharis and Cybistetes have biflabellate leaves, which C. baumii lacks, and zygomorphic perianths. Snijman & Williamson (1994) went so far as to suggest that separation of Ammocharis and Cybistetes needed reexamination. To this, we would add that the position of C. baumii relative to these two genera also requires re-evaluation, though there is little question that the species lies outside of Crinum s.s. based on the ITS phylogeny. This well-supported sister clade to Crinum is rooted within southern Africa (Fig. 4), but is broadly dispersed through sub-Saharan Africa via the five species of Ammocharis. Given the likely role of oceanic dispersal of Crinum seed in the evolution of the genus, it is not surprising that several geographically cohesive clades are resolved in the genus by ITS and trnL-F sequences. The roots of the entire lineage are in southern Africa, and area analyses places either South Africa (eastern South Africa in the case of Crinum s.s.) or south-western Africa within the ancestral node of all three clades of Crinum (Fig. 4). Each of the clades contains at least one dispersal event out of Africa (albeit only one in clade B), while clade B also encompasses continuous and perhaps recent evolution in South Africa. A monophyletic American clade denotes a single dispersal event into the Western Hemisphere. A monophyletic Madagascar group indicates a similar scenario. Two Asiatic clades are resolved as well, though not as sister groups, suggesting a possible double entry into the region. A more precise understanding of the exact biogeographical scenario for Clade A is still elusive. The © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 MOLECULAR PHYLOGENY OF CRINUM absence of any southern African Crinum species in the American/North African/tropical African clade, if not an artifact of our sampling biases, may indicate that this clade represents the earliest vicariant divergence from the rest of the genus (Fig. 4). Ostensibly, the decimation of the tropical flora of Africa was a major isolating factor for the genus. Direct connections between South Africa and tropical Africa were most likely severely interrupted by the late Oligocene/early Miocene as uplift and climatic change began to create the modern landscape of the continent (Axelrod & Raven, 1978; Goldblatt, 1978; Coetzee, 1993). The apparent relationships of American species of Crinum with species from North and tropical Africa, rather than with the southern African elements of the genus (Fig. 4), might indicate that the dispersal point of departure from Africa to America was in the northern half of the African continent. A land bridge did exist between the Brazilian Bulge and Nigeria until the end of the Cretaceous (Rand & Mabesoone, 1982). Again, sister group relationships are obscured by a basal trichotomy in the American/North African/tropical African clade (Figs 1,3). Clade C also presents some intriguing biogeographical hypotheses. One subclade represents an Australasian group, i.e. species found in south-east Asia, the Pacific and Australia (Fig. 4). A well-supported group within this subclade as presently sampled (C. asiaticum; C. pedunculatum, and C. sp., Borneo) is marked morphologically by the obsolescence of the bulb (the leaves basally form a tight and stout pseudostem which originates directly from a hyper-developed basal plate). The second, or Sino-Himalayan clade, represents species from the Himalayan region and nearby south-west China that form true bulbs, which may have entered the Indian subcontinent either directly from Africa or via Madagascar. Both dispersal pathways have been hypothesized for various taxa of angiosperms (Raven & Axelrod, 1974). The SinoHimalayan clade barely receives moderate bootstrap support (65%) and has a DI of only 1. Its ancestral node has an equivocal Fitch optimization of area, and two possible area assignments from DIVA. Moreover, in the strict consensus tree of ITS alone it forms a polytomy with the Australasian clade, the Madagascar clade, C. buphanoides and C. subcernuum (both African and with subg. Crinum floral morphology); thus the sister group relationships of either Asian clade are not as clear as the single tree used for area optimization would indicate (Fig. 4). However, in the bootstrap consensus tree of the ITS matrix and in the combined analysis, the Sino-Himalayan clade is sister to an Australasian–Madagascar–C. buphanoides– C. subcernuum polytomy (Fig. 1, bootstrap = 50%; Fig. 3, bootstrap = 63%), suggesting an earlier dispersal event from Africa, with a likely initial entry 359 into India. Consequently, dispersal to Australasia may have been an independent and later event. Van Steenis (1962) hypothesized a land-bridge connection between Madagscar and Sri Lanka incorporating the Seychelles–Comores bank during the mid- to late Cretaceous, which he named ‘Lemuria.’ He saw no other way to account for plant distributions that encompassed the periphery of the Indian Ocean. Mckenzie & Sclater (1973) refuted the possibility. However, Haq, Hardenbol & Vail (1988) reported on the likeliness of increased emergence of the Chagos/ Laccadive Plateau and then contiguous Mascarene Plateau (including the Seychelles Bank) during the early Oligocene (c. 30 MYA). Schatz (1996), in his review of the Indo-Australo-Malesian relationships of the Malagasy flora, postulates that lower sea levels allowed these emergences to function as stepping stones. Schatz’s (1996) ‘Lemurian stepping-stones’ could have engendered migration of Crinum from Madgascar to western Malaysia. This may have also been the pathway by which the ancestor of C. latifolium dispersed from Africa to Sri Lanka and India. Emergent archipelagos may have existed, bridging much of the Indian Ocean between India and Australia as well (McKenzie & Sclater, 1973), which could have allowed dispersal of Crinum to northern Australia and from there into south-eastern Asia and the Pacific. Crinum modestum (Madagascar) and C. venosum (Australia) have a further morphological character in common; both have very short stamens that are atypical for subg. Crinum. Beyond coarse estimates, the relative timing of these events cannot be inferred, again due to the internal trichotomy formed by the major subclades of clade C (Figs 1,3). It is also evident that Crinum entered the Australian continent at least twice. The other migration (C. flaccidum, ‘Codonocrinum’) was apparently directly from Africa (in clade B, Fig. 4). The African relationships of the more southern Australian Crinum flaccidum are also resolved by both ITS and trnL-F (Figs 1,2). If homoplasious base substitutions in the ITS alignment are down-weighted using successive approximation (Farris, 1969; Wenzel, 1997; Lledó et al., 1998; Meerow et al., 1999), a sister relationship between these two species is also resolved (data not shown). Both species share similar seasonally aquatic habitats (albeit on different continents) and have terete juvenile leaves. Moreover, like C. flaccidum, the eastern Cape endemic C. campanulatum, has a campanulate perianth, but not declinate stamens. In the combined analysis (Fig. 3), the position of C. flaccidum as sister to an otherwise African clade is better supported than by ITS alone. How did Crinum get to Australia this first time, assuming an early introduction for the ancestor(s) of C. flaccidum, all of whose closest extant relatives are © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 360 A. W. MEEROW ET AL. endemic to Africa? Three pathways are possible. Oceanic long distance dispersal directly from Africa is one scenario. Secondly, Madagascar–India could have provided an intermediate route, followed by oceanic dispersal to Australia, c. 65 MYBP (Raven & Axelrod, 1974). However, the phylogenetic relationships of C. flaccidum are not with any extant Madagascar or Indian Crinum species (Fig. 4). It is certainly possible that the extant Madagascar and Indian species represent more recent migrations and that the earlier migratory ancestors of C. flaccidum are extinct. A third possible migration could have been via a more southerly subtropical route through Antarctica, that by traditional geophysical hypotheses (Smith & Hallam, 1970; Smith, Smith & Funnell, 1994) was available too early in the diversification of the angiosperms to have figured in the biogeographical history of a higher asparagoid monocot. However, more recent hypotheses based on palaeontological evidence (Sampson et al., 1998; Hay et al., 1999; Krause et al., 1999) suggest that a longer-lived land connection between India and Antarctica–Australia via the Kerguelen Plateau may have existed as late as 80 MYBP. Outside of two distinct clades of Crinum in Australia, the continent is home to a distinctive tribal clade of the family (Meerow & Snijman, 1998; Meerow et al., 1999), the Calostemmateae (Proiphys Herb. and Calostemma R. Brown, the latter endemic to Australia, the former also extending into south-east Asia). In the case of Calostemmateae, one might infer that the tribe originated in Australia and later migrated to south-east Asia, since the continent contains both the generic and species diversity of the lineage. The sister relationships of this tribe have still not been resolved (Meerow et al., 1999). In the lower asparagoid family Iridaceae, all species of the genus Dietes are endemic to eastern through southern sub-Saharan Africa with the exception of a single Australian species on Lord Howe Island (Goldblatt, 1981) that is considered the least morphologically derived species in the genus (Goldblatt, 1978). Baum, Small & Wendel (1998) concluded that dispersal between Africa and Australia in the genus Adansonia (Malvaceae) occurred via ocean currents considerably after the break-up of Gondwanaland. In the case of Adansonia, the sole Australian species was sister to the rest of the genus (Madagascar and Africa), thus the direction of the dispersal event was ambiguous. In Crinum, there is no ambiguity that the genus originated in Africa (Fig. 4). Southern Africa is also where the sister group to Crinum originated (Fig. 4), and is the centre of diversity for the genus (Nordal, 1977; Fangan & Nordal, 1993). This is not to suggest that all of this diversity is necessarily ancient; it may reflect radiation engendered by the more recent palaeoclimatic and geological history of Africa encompassing Neogene and later times (Axelrod, 1972; Raven & Axelrod, 1974; Coetzee, 1993). The increased aridity of the African climate and the uplift of the continental mass that started at the beginning of the Miocene, further abetted by Quaternary climatic fluctuations (Demenocal, 1995) were catastrophic to many elements of the African flora, but it may have been a selective pressure for diversity among groups of geophytes capable of adapting to increasing drought. The geophyte richness of South Africa is well documented (Goldblatt, 1978), and the Cape region has been suggested as a possible refuge for certain African plant and animal groups as the tropical flora of the continent was impoverished (Raven & Axelrod, 1974). C. variabile is the only Crinum species so far known from the winter-rainfall region of western South Africa (Verdoorn, 1973). Subtropical forests were still present in the Western Cape during the Miocene and Pliocene (Scott et al., 1997). The earliest evidence of modern semiarid, winterrainfall pattern in the Western Cape dates to the Late Pliocene, but it was not fully established until the Early Pleistocene (Tankard & Rogers, 1978; Hendey, 1983; Coetzee, 1986). Moreover, the winter-rainfall region of southern Africa experienced a more recent pattern of expansion and contraction with concurrent wetter and drier conditions during glacial and interglacial periods of the Quaternary (Tankard, 1976; van Zinderen Bakker, 1976; Tyson, 1986; Crockcroft, Wilkinson & Tyson, 1987). It would appear that a winter rainfall regime is largely inimical to Crinum. Nonetheless, it is impossible to determine if the genus ever existed in the Western Cape prior to the establishment of the Mediterranean climate. In conclusion, nuclear rDNA ITS sequences support a southern African (eastern South Africa or southwestern Africa) origin for the genus Crinum, and indicate three major waves of radiation corresponding to the three main clades resolved in our trees (Figs 1,3,4). Two entries into Australia for the genus are hypothesized. Asian and Malagasy Crinum are phylogenetically related, and separate SinoHimalayan and Australasian dispersals are indicated. The monophyletic American species are allied with tropical and North African species. Recognition of two subgenera in Crinum on the basis of floral morphology is not supported by the molecular phylogeny, as the apomorphic subg. Crinum floral morphology has evolved more than once (Fig. 5). Crinum baumii appears to be more closely related to Ammocharis and Cybistetes, and the taxonomic standing of this species and both of these genera needs to be re-evaluated in light of this relationship. We sought to augment the incomple resolution of phylogenetic relationships within Crinum with data from the plastid atpB-rbcL intergenic spacer (Chiang, © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 MOLECULAR PHYLOGENY OF CRINUM Schaal & Peng, 1998). The AT-rich spacer proved extremely hyper-variable, and necessitated the trimming of 360 bp that was impossible to align, even within clades well-supported by ITS. The resulting matrix was poorly resolved (data not shown), but after successive weighting, a few clades in common with the ITS or trnL-F phylogeny were supported: (1) a tropical and North African clade (bootstrap = 88%), (2) a Madagascar/Asian clade (71%), albeit with the anomalous inclusion of C. crassicaule, (3) a C. campanulatum/ C. flaccidum sister relationship (87%), and (4) a C. kirkii–C. politifolium clade (99%). We are in the process of obtaining sequences for the plastid ndhF gene, in the hopes that it will resolve the basal polytomies within the ITS phylogeny and allow stratigraphic estimation of divergence within the genus. ACKNOWLEDGEMENTS This work was partially supported by National Science Foundation Grants DEB-968787 and 0129179. We thank Drs DA Snijman and MW Chase for providing leaf material or DNA of several species, and an anonymous reviewer whose comments greatly improved the manuscript. REFERENCES Arroyo SC, Cutler DF. 1984. Evolutionary and taxonomic aspects of the internal morphology in Amaryllidaceae from South America and Southern Africa. Kew Bulletin 39: 467– 498. Axelrod DI. 1972. Edaphic aridity as a factor in angiosperm evolution. American Naturalist 106: 311–320. Axelrod DL, Raven PH. 1978. Late Cretaceous and Tertiary vegetation history of Africa. In: Werger MJA, ed. Biogeographyraphy and ecology of southern Africa. The Hague: Dr W. Junk, 77–130. Baker JG. 1881. A synopsis of the known species of Crinum L. Gardeners’ Chronicle 1: 763. Baker JG. 1898. Amaryllidaceae. In: Thiselton-Dyer WT, eds. Flora of Tropical Africa, Vol. 7. London: L. Reeve, 376– 413. Baum DA, Small RL, Wendel JF. 1998. Biogeography and floral evolution of baobabs (Adansonia, Bombacaceae) as inferred from multiple data sets. Systematic Biology 47: 181– 207. Bremer K. 1988. The limits of amino acid sequence data in angiosperm phylogenetic reconstruction. Evolution 42: 198– 213. Chiang T-Y, Schaal BA, Peng C-I. 1998. Universal primers for amplificxation and sequencing a noncoding spacer between the atpB and rbcL genes of chloroplast DNA. Botanical Bulletin of the Academia Sinica 39: 245–250. Coetzee JA. 1986. Palynological evidence for major vegetation and climatic change in the Miocene and Pliocene of the 361 south-western Cape. South African Journal of Science 82: 71–72. Coetzee JA. 1993. African flora since the terminal Jurassic. In: Goldblatt P, ed. Biology relationships between Africa and South America. New Haven: Yale University Press, 38– 61. Crockcroft MJ, Wilkinson MJ, Tyson PD. 1987. The application of a present-day climatic model to the late Quaternary in southern Africa. Climatic Change 10: 161–181. Dassanayake MD. 1981. Typification of Crinum zeylanicum (L.) L. (Amaryllidaceae). Taxon 30: 481–482. Demenocal PB. 1995. Plio-Pleistocene African climate. Science 270: 53–58. Douzery JP, Pridgeon AM, Kores P, Kurzweil H, Linder P, Chase MW. 1999. Molecular phylogenetics of Diseae (Orchidaceae): a contribution from nuclear ribosomal ITS sequences. American Journal of Botany 86: 887–899. Dubuisson JY, Hebant-Mauri R, Galtier J. 1998. Molecules and morphology: conflicts and congruence within the fern genus Trichomanes (Hymenophyllaceae). Molecular Phylogeny and Evolution 9: 390–397. Fangan BM, Nordal I. 1993. A comparative analysis of morphology, chloroplast DNA and distribution within the genus Crinum (Amaryllidaceae). Journal of Biogeography 20: 55– 61. Farris JS. 1969. A successive approximations approach to character weighting. Systematic Zoology 18: 374–385. Farris JS, Källersjö M, Kluge AG, Bult C. 1994. Testing significance of incongruence. Cladistics 10: 315–319. Farris JS, Källersjö M, Kluge AG, Bult C. 1995. Constructing a significance test for incongruence. Systematic Biology 44: 570–572. Felsenstein J. 1985. Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39: 783–791. Fitch WM. 1971. Toward defining the course of evolution: Minimum change for a specific tree topology. Systematic Zoology 20: 406–416. Geerinck D. 1973. Amaryllidaceae. In: Bamps P, ed. Flora de’Afrique Centrale Spermatophyes. Brussels: Jardin Botanique National de Belgique. Goldblatt P. 1978. An analysis of the flora of southern Africa: its characteristics, relationships, and origins. Annals of the Missouri Botanical Garden 65: 369–436. Goldblatt P. 1981. Systematics, phylogeny and evolution of Dietes (Iridaceae). Annals of the Missouri Botanical Garden 68: 132–153. Haq BU, Hardenbol J, Vail PR. 1988. Mesozoic and Cenozoic chronostratigraphy and eustatic xycles. In: Wilgus CK, Hastings BS, Posamentier H, Wagoner JV, Ross CA, Kendall CGStC, eds. Sea-level changes: an integrated approach. Special Publication 42. Tulsa: Society of Economic Paleontology and Mineralology, 71–108. Hay WW, DeConto RM, Wold CN, Willson KM, Voigt S, Schulz M, Wold-Rossby A, Dullo W-C, Ronov AB, Balukhovsky AN, Soeding E. 1999. An alternative global Cretaceous paleogeography. In: Barrera E, Johnson C, eds. Evolution of the Cretaceous ocean-climate system. Geological Society of America Special Paper 332: 1–48. © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 362 A. W. MEEROW ET AL. Hendey QB. 1983. Palaeoenvironmental implications of the Late Tertiary vertebrate fauna of the fynbos region. In: Deacon HJ, Hendey QB, Lambrechts JJN, eds. Fynbos palaeoecology: a preliminary synthesis. South African National Science Progress Report no. 75. Pretoria: CSIR, 100–115. Hepper FN. 1968. Flora of tropical West Africa, Vol. 3. London: Crown Agent for Overseas Governments and Administration. Herbert W. 1837. Amaryllidaceae. London: J. Ridgeway and Sons. Ito M, Kawamoto A, Kita Y, Yukawa T, Kurita S. 1999. Phylogenetic relationships of Amaryllidaceae based on matK sequence data. Journal of Plant Research 112: 207–216. Koshimizu T. 1930. Carpobiological studies of Crinum asiaticum L. var. japonicum Bak. Memoirs of the College of Sciences, Kyoto Imperial University, Series B, Biology 5: 183– 227. Krause DW, Rogers RR, Forster CA, Hartman JH, Buckley GA, Sampson SD. 1999. The Late Cretaceous vertebrate fauna of Madagascar: implications for Gondwanan paleobiogeography. GSA Today 9 (8): 1–7. Lehmiller DJ. 1997a. Synopsis of the genus Crinum (Amaryllidaceae) in Namibia. Herbertia 52: 44–65. Lehmiller DJ. 1997b. Crinum subgenus Codonocrinum. Southern Tchad and extreme northern Cameroun. Herbertia 52: 119–133. Linneaus C. 1753. Species plantarum. Stockholm. Lledó MD, Crespo MB, Cameron KM, Fay MF, Chase MW. 1998. Systematics of Plumbaginaceae based upon cladistic analysis of rbcL sequence data. Systematic Botany 23: 21– 29. Maddison DR. 1991. The discovery and importance of multiple islands of most-parsimonious trees. Systematic Zoology 40: 315–328. Maddison DR, Maddison WP. 2001. MacClade, Version 4.03. Sutherland, MA: Sinauer Associates Inc. Maddison DR, Ruvolo M, Swofford DL. 1992. Geographic origins of human mitochondrial DNA: phylogenetic evidence from control region sequences. Systematic Biology 41: 111– 124. Mckenzie DP, Sclater JC. 1973. The evolution of the Indian Ocean. Scientific American 228 (5): 63–72. Meerow AW, Fay MF, Guy CL, Li Q-B, Zaman FQ, Chase MW. 1999. Systematics of Amaryllidaceae based on cladistic analysis of plastid rbcL and trnL-F sequence data. American Journal of Botany 86: 1325–1345. Meerow AW, Guy CL, Li Q-B, Clayton JR. 2002. Phylogeny of the tribe Hymenocallideae (Amaryllidaceae) based on morphology and molecular characters. Annals of the Missouri Botanical Garden 89: 400–413. Meerow AW, Guy CL, Li Q-B, Yang S-Y. 2000. Phylogeny of the American Amaryllidaceae based on nrDNA ITS sequences. Systematic Botany 25: 708–726. Meerow AW, Snijman DA. 1998. Amaryllidaceae. In: Kubitzki K, ed. Families and genera of vascular plants, Vol. 3. Berlin: Springer-Verlag, 83–110. Meerow AW, Snijman DA. 2001. Phylogeny of Amaryllidaceae tribe Amaryllideae based on nrDNA ITS sequences and morphology. American Journal of Botany 88: 2321– 2330. Milne-Redhead E, Schweickerdt HG. 1939. A new conception of the genus Ammocharis Herb. Botanical Journal of the Linnean Society 52: 159–197. Nordal I. 1977. Revision of the East African taxa of the genus Crinum (Amaryllidaceae). Norwegian Journal of Botany 24: 179–194. Nordal I. 1982. Amaryllidaceae. In: Polhill RM, ed. Flora of tropical East Africa. Rotterdam: AA. Balkema. Nordal I. 1986. Amaryllidaceae. In: Morat PH, ed. Flora du Gabon. Paris: Museum National d’Histoire Naturelle. Nordal I, Wahlstrom R. 1980. A study of the genus Crinum (Amaryllidaceae) in Cameroun. Adansonia 20: 179–198. Rand HH, Mabesoone JM. 1982. North Eastern Brazil and the final separation of South America and Africa. Palaeoecology 39: 71–85. Raven PH, Axelrod DI. 1974. Angiosperm biogeography and past continental movements. Annals of the Missouri Botanical Garden 61: 539–673. Ronquist F. 1996. DIVA, Version 1.1 Computer program and mManual available by anonymous FTP from Uppsala University (ftp.uu.se or ftp.systbot.uu.se). Ronquist F. 1997. Dispersal-vicariance analysis: a new approach to the quantification of historical biogeography. Systematic Biology 46: 195–203. Sampson SD, Witmer LM, Forster CA, Krause DW, O’Connor PM, Dodson P, Ravoavy F. 1998. Predatory dinosaur remains from Madagascar: implications for the Cretaceous biogeography of Gondwana. Science 280: 1048–1051. Schatz GE. 1996. Malagasy/Indo-Australo-Malesian phytogeographic connections. In: Lourenço WR, eds. Biogeography of Madagascar. Paris: Editions ORSTOM, 73–84. Scott L, Anderson HM, Anderson JM. 1997. Vegetation history. In: Cowling RM, Richardson DM, Pierce SM, eds. Vegetation of southern Africa. Cambridge: Cambridge University Press, 62–84. Seelanan T, Schnabel A, Wendel J. 1997. Congruence and consensus in the cotton tribe (Malvaceae). Systematic Botany 22: 259–290. Smith AG, Hallam A. 1970. The fit of the southern continents. Nature 225: 139–144. Smith AG, Smith DG, Funnell BM. 1994. Atlas of Mezozoic and Cenozoic coastlines. Cambridge, UK: Cambridge University Press. Snijman DA, Linder HP. 1996. Phylogenetic relationships, seed characters, and dispersal system evolution in Amaryllideae (Amaryllidaceae). Annals of the Missouri Botanical Garden 83: 362–386. Snijman DA, Williamson G. 1994. A taxonomic reassessment of Ammocharis herrei and Cybistetes longifolia (Amaryllideae: Amaryllidaceae). Bothalia 24: 127–132. Sorenson MD. 1996. Treerot. Ann Arbor: University of Michigan. van Steenis CGGJ,. 1962. The land-bridge theory in botany. Blumea 11: 235–372. Swofford DL. 1998. Phylogenetic Analysis Using Parsimony, v. 4 0 beta. Sutherland, MA: Sinauer Associates Inc. © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363 MOLECULAR PHYLOGENY OF CRINUM Taberlet P, Gielly L, Pautou G, Bouvet J. 1991. Universal primers for amplification of three non-coding regions of chloroplast DNA. Plant Molecular Biology 17: 1105– 1110. Tankard AJ. 1976. Stratigraphy of a coastal cave and its palaeoclimatic significance. In: van Zinderen Bakker EM, ed. Palaeoecology of Africa 9. Rotterdam: Balkema, 151– 159. Tankard AJ, Rogers J. 1978. Late Cenozoic palaeoenvironments on the west coast of southern Africa. Journal of Biogeography 5: 319–337. Tyson PD. 1986. Climatic change and variability in southern Africa. Cape Town: Oxford University Press. Uphof JC. 1942. A review of the species of Crinum. Herbertia 9: 63–84. Van der Pijl L. 1982. Principles of dispersal in higher plants. Berlin: Springer-Verlag. 363 Verdoorn IC. 1973. The genus Crinum in Southern Africa. Bothalia 11: 27–52. Wenzel JW. 1997. When is a phylogenetic test good enough? Memoirs de la Museé Nationale d’Histoire Naturelle 173: 31– 45. White TJ, Bruns T, Lee S, Taylor J. 1990. Amplification and direct sequencing of fungal ribosomal RNA genes for phylogenetics. In: Innis M, Gelfand D, Sninsky J, White T, eds. PCR protocols: a guide to methods and applications. Orlando: Academic Press, 315–322. Zhanhe J, Meerow AW. 2001. Amaryllidaceae. In: Zhengyi W, Raven PH, eds. Flora of China, Vol. 24. St. Louis: Missouri Botanical Garden Press, St. Louis, 264–273. van Zinderen Bakker EM. 1976. The evolution of LateQuaternary palaeoclimates of southern Africa. In: van Zinderen Bakker EM, ed. Palaeoecology of Africa 9. Rotterdam: Balkema, 160–202. © 2003 The Linnean Society of London, Botanical Journal of the Linnean Society, 2003, 141, 349–363