Int. J. Plant Sci. 165(4 Suppl.):S35–S54. 2004. Ó 2004 by The University of Chicago. All rights reserved. 1058-5893/2004/1650S4-0004$15.00 ANCIENT VICARIANCE OR RECENT LONG-DISTANCE DISPERSAL? INFERENCES ABOUT PHYLOGENY AND SOUTH AMERICAN–AFRICAN DISJUNCTIONS IN RAPATEACEAE AND BROMELIACEAE BASED ON ndhF SEQUENCE DATA Thomas J. Givnish,1 ,* Kendra C. Millam,* Timothy M. Evans,y Jocelyn C. Hall,* J. Chris Pires,* Paul E. Berry,* and Kenneth J. Sytsma* *Department of Botany, University of Wisconsin, Madison, Wisconsin 53706, U.S.A.; and yDepartment of Botany, Hope College, Holland, Michigan 49428, U.S.A. Rapateaceae and Bromeliaceae each have a center of diversity in South America and a single species native to a sandstone area in west Africa that abutted the Guayana Shield in northern South America before the Atlantic rifted. They thus provide ideal material for examining the potential role of vicariance versus long-distance dispersal in creating amphiatlantic disjunctions. Analyses based on ndhF sequence variation indicate that Rapateaceae and Bromeliaceae are each monophyletic and underwent crown radiation around 41 and 23 Ma, respectively. Both exhibit clocklike sequence evolution, with bromeliads evolving roughly one-third more slowly than rapateads. Among rapateads, the divergence of west African Maschalocephalus dinklagei from its closest South American relatives implies that Maschalocephalus resulted via long-distance dispersal 7 Ma, not ancient continental drift; only its sandstone habitat is vicariant. Rapateads arose first at low elevations in the Guayana Shield; the earliest divergent genera are widespread along riverine corridors there and, to a lesser extent, in Amazonia and the Brazilian Shield. Speciation at small spatial scales accelerated 15 Ma with the invasion of high-elevation, insular habitats atop tepuis. Among bromeliads, Pitcairnia feliciana diverges little from its congeners and appears to be the product of long-distance dispersal ca. 12 Ma. Brocchinia/Ayensua and then Lindmania are sister to all other bromeliads, indicating that the Guayana Shield was also the cradle of the bromeliads. Three lineages form an unresolved trichotomy representing all other bromeliads: (1) Tillandsioideae, (2) Hechtia, and (3) a large clade including remaining genera of Pitcairnioideae and all Bromelioideae. The last includes a clade of pitcairnioid genera endemic to the Guayana and Brazilian Shields; a xeric group (Abromeitiella/Deuterocohnia/Dyckia/Encholirium/Fosterella) from southern South America and the southern Andes, sister to Pitcairnia; and Andean Puya, sister to Bromelioideae, with many of the latter native to the Brazilian Shield. Both Rapateaceae and Bromeliaceae appear to have arisen at low elevations in the Guayana Shield, experienced accelerated speciation after invading dissected mountainous terrain, and undergone long-distance dispersal to west Africa recently. Bromeliad acquisition of key adaptations to drought (e.g., CAM photosynthesis, tank habit, tillandsioid leaf trichomes) 17 Ma appears to have coincided with and help cause the centripetal invasion of drier, more seasonal regions beyond the Guayana Shield, resulting in a wider familial range and dominance of the epiphytic adaptive zone. Geology, past and present climate, and proximity to South America help account for both families occurring in nearly the same area of Africa. We present a new classification for Rapateaceae, including a new tribe Stegolepideae, a new subfamily Monotremoideae, and revisions to tribe Saxofridericieae and subfamily Rapateoideae. Keywords: adaptive evolution, historical biogeography, penalized likelihood. Introduction Atlantic Ocean 84 to 102 Ma (Briceño and Schubert 1990; Gibbs and Barron 1993; Pitman et al. 1993; Potter 1997). These disjunctions are surprising given the apparently sedentary nature of both groups. Most Rapateaceae have seeds with no apparent means of long-distance dispersal (Maguire 1982; Givnish et al. 2000); Pitcairnia and related bromeliads have seeds that are only weakly winged (Smith and Downs 1974) and thus likely to be poorly dispersed (Kessler 2002a, 2002b). Tropical intercontinental disjunctions in the distributions of bromeliads and rapateads and other families across the Atlantic thus pose the question of whether they arose by ancient vicariance via continental drift or by relatively recent long-distance dispersal (Givnish et al. 2000; Sytsma et al. 2004). This question, in turn, raises the broader issues of Rapateaceae and Bromeliaceae are closely related families of commelinid monocots (Givnish et al. 1999, 2004; Chase et al. 2000; Stevenson et al. 2000; APG 2003), both restricted to the New World with the exception of a single species in each family that occurs in west Africa (fig. 1). The African species—Maschalocephalus dinklagei of Rapateaceae and Pitcairnia feliciana of Bromeliaceae—occur in a sandstone region that abutted the Guayana Shield of northern South America before continental drift opened the tropical 1 E-mail givnish@facstaff.wisc.edu. Manuscript received August 2003; revised manuscript received March 2004. S35 S36 INTERNATIONAL JOURNAL OF PLANT SCIENCES Fig. 1 Amphiatlantic distributions of Bromeliaceae and Rapateaceae. The distribution of rapateads in the New World is completely included within that of bromeliads. where and when each family evolved and how phylogeny reflects geography and geological history in each lineage, including their relationship to the Guayana Shield and the archipelago of tepuis (sandstone table mountains) that rise 500–3000 m from it above a sea of lowland tropical forests and savannas (fig. 2). These issues must be addressed indirectly, however. The earliest fossils of Bromeliaceae are only 40 Ma old versus 70–90 Ma for other families of order Poales, and no fossils of Rapateaceae are known (Herendeen and Crane 1995). The scant examples of Bromeliaceae fossils are compounded by the long stem lineage for the family (Bremer 2000; this study) and the difficulty in placing the fossils in any meaningful location in phylogenetic trees. Neither family grows in habitats conducive to fossilization. Thus, inferences as to where and when each evolved must be based on an analysis of their relationships to other families that are better represented in the fossil record, providing dates and phylogenetic branching topologies against which rates of molecular evolution can then be calibrated. To reconstruct the historical biogeography of Rapateaceae and Bromeliaceae, we present molecular phylogenies for each family based on DNA sequences of ndhF, a chloroplastencoded gene that evolves at a moderately rapid pace (Gaut et al. 1997; Terry et al. 1997a; Givnish et al. 2000). We cali- brate these trees against the times of origin for both families based on an ndhF phylogeny across monocots (Givnish et al. 2004b) using multiple monocot fossils (Bremer 2000) and the inferred divergence of Acorus from other monocots 134 Ma (Bremer 2000). We then use these calibrated trees to (1) infer the timing of cladogenetic events within each family, (2) relate them to geologically dated events in Earth history, (3) evaluate ancient vicariance versus recent long-distance dispersal as potential mechanisms underlying the disjunct occurrence of a single species in each family in west Africa, and (4) analyze where each family might have arisen. We use our molecular findings to revise the infrafamilial classification of Rapateaceae, and we consider the factors that might have resulted in similar distributions for both bromeliads and rapateads in tropical west Africa. Methods Taxon Sampling and Outgroup Analysis We included 58 ndhF sequences in our phylogenetic analysis, including 35 species representing all known major groups of Bromeliaceae (26 of 59 genera) and 16 species representing all four tribes and both subfamilies of Rapateaceae (14 of 16 genera) (table 1). To maximize the resolution of relationships GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE S37 Fig. 2 Distribution of sandstone tepuis, granitic uplands, and lowland Amazonian savannas in the Venezuelan portion of the Guayana Shield (after Givnish et al. 2000). Position relative to other physiographic regions of South America shown in the inset. Extensive areas of Amazonian savannas in Colombia have yet to be mapped. among taxa of these two families while minimizing the chance of artifacts due to a poor sampling of related families of Poales (Sytsma and Baum 1996), we included seven taxa from Mayacaceae, Typhaceae, Sparganiaceae, Flagellariaceae, Joinvilleaceae, and Poaceae. On the basis of a cladistic analysis of ndhF cpDNA sequences across 277 monocot taxa, Givnish et al. (2004b) found that Bromeliaceae is sister to Typhaceae þ Sparganiaceae at the base of order Poales (sensu APG 2003) and that Rapateaceae is sister to a clade composed of Eriocaulaceae, Mayacaceae, and Xyridaceae at the next node. The latter group of families is sister to Cyperaceae, and together they are sister to the remaining families of the order, including Flagellariaceae, Joinvilleaceae, and Poaceae. Bootstrap support levels for this topology are generally higher than those for the slightly different tree presented by Chase et al. (2000) based on rbcL, atpB, and 18S sequence data; the ndhF results are also based on a substantially greater density of taxon sampling. On the basis of relationships within Poales from these two broader studies (Chase et al. 2000; Givnish et al. 2004b), we used Bromeliaceae, Typhaceae, and Sparganiaceeae as a monophyletic functional outgroup. DNA Extraction, Amplification, and Sequencing Total DNAs were isolated from fresh leaves or tissues frozen at
80°C or dried in contact with silica gel, using the CTAB technique of Doyle and Doyle (1987) as modified by Smith et al. (1991). The chloroplast-encoded ndhF gene was amplified, cycle sequenced, and edited using standard approaches (Givnish et al. 2000). We aligned the sequences Table 1 Species of Order Poales (sensu APG 2003) Included in Phylogenetic Analysis of ndhF Sequence Divergence Family and species Bromeliaceae: Abromeitiella lorentziana (Mez) A. Cast. Aechmea haltonii H. Luther Ananas ananassoides (Baker) L. B. Smith Ayensua uaipanensis (Maguire) L. B. Smith Brewcaria reflexa (L. B. Sm.) B. K. Holst Brocchinia acuminata L. B. Smith Brocchinia paniculata Schult. f. Brocchinia prismatica L. B. Smith Brocchinia serrata L. B. Smith Bromelia Adans. sp. Canistrum giganteum (Baker) L. B. Smith Catopsis wangerini Mez & Werckle Cottendorfia florida Schult. f. Deuterocohnia longipetala Mez Dyckia Schult. f. sp. Encholirium Mart. ex. Schult. sp. Fosterella penduliflora (C. H. Wright) L. B. Smith Fosterella villosula (Harms) L. B. Smith Glomeropitcairnia penduliflora Mez Guzmania monostachya Rusby in DC. Hechtia guatemalensis Mez Hechtia lundelliorum L. B. Smith Lindmania longipes (L. B. Sm.) L. B. Smith Lindmania Mez sp. Navia saxicola L. B. Smith Neoregelia pineliana var. pineliana (Lemaire.) L. B. Smith Nidularium selloanum (Baker) E. Pereira & Leme Pitcairnia atrorubens Baker Pitcairnia corallina Linden Pitcairnia feliciana (A. Cheval.) Harms & Mildbr. Pitcairnia hirtzii H. Luther Puya floccosa E. Morr. Puya raimondii Harms Tillandsia complanata Benth. Vriesea viridiflora (Regel) J. R. Grant Flagellariaceae: Flagellaria indica L. Joinvilleaceae: Joinvillea ascendens Gaudich. Mayacaceae: Mayaca fluviatilis Aubl. Poaceae: Bambusa stenostachya Pharus lappulaceus Initial citation GenBank Voucher This article Terry et al. 1997a AY438598 L75844 SEL ex StL s.n. SEL 85-1447 KC Millam Terry et al. Terry et al. 1997a L75845 Brown 3129 RM Terry et al. This article AY438599 Givnish KC Millam This article Terry et al. 1997a This article This article This article AY208982 L75859 AY208981 AY438600 AY438601 KC Millam Terry et al. KC Millam KC Millam KC Millam Terry et al. 1997a L75860 See Givnish et al. 1997 SEL 81-1937 Fernandez 8236 PORT See Givnish et al. 1997 Betancur & Ramı́rez 1265 MO Brown 3128 RM Terry et al. 1997a Terry et al. 1997a This article This article Terry et al. 1997a Terry et al. 1997a L75861 L75855 AY438602 AY208984 L75857 L75862 Brown 3183 RM Palacı́ 1235 RM SEL 96-0695 Hort. Marnier-Lapostelle s.n. Brown 3131 RM MSBG 1984-0364 Terry et al. Terry et al. KC Millam KC Millam Terry et al. Terry et al. Terry et al. 1997a L75863 SEL 69-1976-12 Terry et al. This article AY438603 StL 79-2073 KC Millam Terry et al. 1997a L75864 Givnish s.n. WIS Terry et al. Terry et al. 1997a This article This article L75865 AY438604 AY208985 SEL 82-225 SEL 81-1891 SEL 85-1005 Terry et al. KC Millam KC Millam This article This article This article AY438605 AY438606 AY208983 See Givnish et al. 1997 Givnish See Givnish et al. 1997 KC Millam KC Millam KC Millam Terry et al. 1997a L75893 SEL 86-261 Terry et al. Terry et al. 1997a This article This article L75894 AY438607 AY438608 Leme 1830 HB SEL 86-311 SEL 86-0574 Terry et al. KC Millam KC Millam This article Terry et al. 1997a This article This article Terry et al. 1997a AY438609 L75901 AY438610 AY438611 L75899 SEL SEL SEL SEL SEL KC Millam Terry et al. KC Millam KC Millam Terry et al. Terry et al. 1997a L75910 SEL 78-757 Terry et al. Clark et al. 1995 U22008 Clark & Zhang 1305 ISC JF Wendel Clark et al. 1995 U21973 NTBG* 800379 JF Wendel This article BD20001 Berry 3004 WI JC Hall Clark et al. 1995 Clark et al. 1995 U21967 U21994 Zhang 8400174 ISC L. Clark 1329 ISC JF Wendel JF Wendel 98-0116 93-294 90-0612 91, s.n. 79-0519 Sequencer Terry et al. GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE S39 Table 1 (Continued ) Family and species Initial citation GenBank Voucher Sequencer Rapateaceae: Amphiphyllum rigidum Gleason Givnish et al. 2000 AF207638 Fernández, Stergios, Givnish, and Funk 8061 PORT TM Evans & ML Zjhra Cephalostemon flavus (Link) Steyerm. Givnish et al. 2000 AF207624 TM Evans Epidryos guayanensis Maguire Guacamaya superba Maguire Givnish et al. 2000 Givnish et al. 2000 AF207632 AF207636 Kunhardtia rhodantha Maguire Givnish et al. 2000 AF207635 Smith, Sytsma, Givnish 303 WIS Berry and Brako 5539 WIS Smith, Sytsma, Givnish 301 WIS Smith, Sytsma, Givnish 300 WIS Maschalocephalus dinklagei Gilg & K. Schum Monotrema bracteatum Maguire Potarophytum riparium Sandwith Rapatea paludosa Aubl. Givnish Givnish Givnish Givnish 2000 2000 2000 2000 AF207628 AF207625 AF207627 AF207623 Saxofridericia inermis Ducke Saxofridericia regalis Schomb. This article Givnish et al. 2000 AY438612 AF207637 Schoenocephalium cucullatum Maguire Givnish et al. 2000 AF207634 Spathanthus bicolor Ducke Spathanthus unilateralis Desv. This article This article AY438615 AY438613 Stegolepis hitchcockii subsp. morichensis Maguire Givnish et al. 2000 AF207629 This article Windsorina guianensis Gleason Sparganiaceae: Sparganium sp. L. Typhaceae: Typha angustifolia L. et et et et al. al. al. al. TB Patterson TM Evans TM Evans Assı́ s.n., Côte d’Ivoire 5/95 Smith, Sytsma, Givnish s.n. Givnish 94-3100 WIS Sytsma, Smith, Givnish 5157 WIS Berry 6510, MO Hahn 4675 WIS TM TM TM TM Evans Evans Evans Evans Sytsma, Smith, Givnish 5116 WIS Givnish 89-125 WIS Berry & Bachhuber 10 Jul 2000, WIS TM Evans & ML Zjhra TM Evans JC Hall TM Evans AY438614 Smith, Sytsma, Givnish 297 WIS Kelloff 1413, US, WIS JC Hall This article AY191213 T. Givnish, s.n. WIS JC Pires Graham et al. 2002 U79230 Graham 1040 TRT SW Graham JC Hall TM Evans & ML Zjhra Note. Nomenclatural authorities are based on citations at the International Plant Names Index Web site (http://www.ipni.org/index.html). SEL indicates accession number at Selby Botanical Garden. StL ¼ Missouri Botanical Garden; the remaining abbreviations in the voucher column are the formal designations for the herbaria in which vouchers are lodged. visually following the guidelines of Baum et al. (1994). We derived all sequences of Rapateaceae, including four reported here for the first time, obtained a new sequence for Mayaca fluviatilis, and obtained 15 ndhF sequences for Bromeliaceae. Typha was sequenced by J. C. Pires and Sparganium by S. W. Graham. The remaining bromeliad sequences, reported by Terry et al. (1997a), were downloaded from GenBank (table 1). Phylogenetic Analyses We inferred phylogenetic relationships from ndhF sequences using PAUP* 4.0b10 (Swofford 2002). We employed global parsimony (Maddison et al. 1984) to polarize character states and search for the shortest tree(s). Nucleotide positions were considered multistate, unordered characters of equal weight under Fitch parsimony. Unknown nucleotides were treated as uncertainties. Indels were excluded from analysis since we were conducting both maximum parsimony (MP) and maximum likelihood (ML) analyses. To determine whether multiple islands of most parsimonious trees were present (Maddison 1991), we conducted heuristic searches seeded with 1000 random addition sequences, using tree bisection and reconnection (TBR) with MULPARS on. This approach should facilitate the recovery of all trees within an island (Olmstead et al. 1992, 1993; Conti et al. 1996). We formed the strict consensus of the most parsimonious trees obtained and used bootstrap analysis (Felsenstein 1985, 1988) to evaluate the relative support for each node of that tree. We also used ML to investigate relationships and obtain better estimates of branch lengths. Models of sequence evolution were analyzed using Modeltest (Posada and Crandall 1998). This program compares 56 models of sequence evolution using a hierarchical framework that calculates the likelihood ratio between models. The starting tree was one of the most parsimonious trees resulting from the parsimony search (see above). A heuristic ML search with TBR branch swapping was then conducted in PAUP* using parameters of DNA substitution determined by the hierarchical likelihood tests. S40 INTERNATIONAL JOURNAL OF PLANT SCIENCES Molecular Clocks and Historical Biogeography Table 2 To test for the operation of ndhF molecular clocks in Rapateaceae and Bromeliaceae, we pruned nonfamily members from the ML tree and used the earliest divergent set of species within each family in the global analysis as the outgroup for each intrafamilial analysis. We then calculated the log likelihoods of these shortest trees with and without enforcing a molecular clock using a six-parameter, fully time-reversible model with a gamma distribution of rates fit to quartile means in PAUP*. We compared twice the difference of these log likelihoods with the x2 distribution with n
2 degrees of freedom (where n is the number of taxa included in the analysis) to test for significant departures from a molecular clock (Felsenstein 1994; Sanderson 1997). We assessed the average rate of and variation in divergence by calculating the mean6SD of branch lengths from the stem group of each family to all the terminal taxa (Givnish et al. 1999, 2000). We converted the ML tree emerging from Modeltest to an ultrametric tree (in which the lengths of all branches from the root are identical) using penalized likelihood (PL) as implemented in the program r8s (Sanderson 2002). In essence, PL averages local differences in the rate of DNA evolution on different branches, taking into account the topology of branching. It differs from nonparametric rate smoothing (NPRS) (Sanderson 1997) in assigning a penalty for rate changes among branches that are too rapid or frequent on the basis of a smoothness parameter. If the smoothness parameter is large, then PL is reasonably clocklike; if the smoothness parameter is small, then PL approaches NPRS. NPRS behaves well in trees with substantial rate variation but suffers when rates are clocklike or nearly so (Sanderson 2002). We used the crossverification approach in r8s to determine the optimal value of the smoothness parameter on the basis of minimizing the sum of the squared deviations between observed and expected branch lengths derived through jackknifing each individual branch (Sanderson 2002). We varied the smoothness parameter from 100 to 103 in steps of 0.25 in the exponent. For comparison, we applied NPRS and Langley-Fitch (1973, 1974) molecular clocks applied to different parts of the phylogeny (bromeliads, rapateads, Typhales, Mayaca, Flagellaria, and Joinvillea, Bambusa, and Pharus) to obtain alternative ultrametric trees. We fixed the age of the crown group of Bromeliaceae— that is, its time of origin of extant taxa—at 23 Ma and fixed the corresponding age of Rapateaceae at 41 Ma (table 2). These dates (averages) were obtained by PL analyses of an ndhF ML tree based on 277 sequences across all major monocot groups (Givnish et al. 2004b). Following the method of Bremer (2000) on a monocot phylogeny based on rbcL, atpB, and 18S sequences, we fixed the age of divergence between Acorus and the ancestor of all other monocots to 134 Ma and used Cretaceous fossil information to place minimum ages on eight monocot nodes. The resulting calibrated trees produced by r8s were then exported to TreeEdit (Rambaut and Charleston 2002) via PAUP* to produce chronograms with branch lengths proportional to time. To explore the sensitivity of the resulting dates within Rapateaceae and Bromeliaceae to tree topology differences within commelinid monocots and among Poales, we examined four alternative Crown Ages of Rapateaceae and Bromeliaceae Based on PL Analysis of the Across-Monocot ndhF Data Set Using Eight Fossil Monocot Dates to Constrain Their Respective Nodes Tree 1 2 3 4 5 6 7 8 Average Rapateaceae crown age (Ma) Bromeliaceae crown age (Ma) 44.6 43.9 44.1 36.7 41.3 40.8 40.9 33.8 40.8 24.9 24.4 24.7 19.1 24.8 23.4 24.0 18.4 22.9 Tree topology ((PO,AR) ((ZI,CO)DA)) I (PO ((ZI,CO)(DA,AR))) I (AR (PO((ZI,CO)DA))) I (((PO,DA)AR) (ZI,CO)) I ((PO,AR) ((ZI,CO)DA)) II (PO ((ZI,CO)(DA,AR))) II (AR (PO((ZI,CO)DA))) II (((PO,DA)AR) (ZI,CO)) II Note. Four different most parsimonious topologies within commelinid monocots were independently assessed with PL. Additionally, two placements of Rapateaceae (I, II) within Poales were evaluated for each of the four commelinid topologies (see text for description of these two placements). PO ¼ Poales, AR ¼ Arecales, ZI ¼ Zingiberales, CO ¼ Commelinales, DA ¼ Dasypogonales. topologies among commelinid orders seen in the set of MP trees based on the 277 monocot sequences (table 2). In addition, we also examined MP trees in which Rapateaceae are sister to a larger clade including Cyperaceae. Thus, eight monocot-wide trees were used sequentially to obtain a set of crown dates for both Rapateaceae and Bromeliaceae (table 2); these in turn were used in each family-wide PL analysis to obtain a range of possible dates for each node within the trees. We then related the timing of cladogenetic events within both families to the times of uplift and dissection of the tepuis, formation of the Amazon basin, and uplift of the Andes (Ghosh 1985; Briceño and Schubert 1990; Briceño et al. 1990; Sidder and Mendoza 1991; Stallard et al. 1991; Hoorn 1994; Hoorn et al. 1995; Briceño 1995; Edmond et al. 1995; Rasanen et al. 1995; Potter 1997; Doerr 1999). We also calculated the times of divergence of Maschalocephalus dinklagei and Pitcairnia feliciana from their nearest American relatives using lineage-specific rates of evolution (Givnish et al. 2000) as well as ML-estimated rates in PL analysis. We analyzed spatial patterns of cladogenesis by optimizing both geographical and elevational distributions (Rapateaceae data from Givnish et al. 2000; P. Berry, unpublished data, Flora of the Venezuelan Guayana project; Bromeliaceae data from Smith and Downs 1979) of the terminal taxa on the bootstrap majority rule tree using MacClade 4.0 (Maddison and Maddison 2000). This tree is one of the 12 MP trees and is nearly identical to the ML tree (the latter having one zerolength branch). Other MP trees were also examined in order to evaluate how various dichotomizations of the unresolved nodes affected interpretation of geographical and elevation patterns of change. Distributions were atomized into the following biogeographic areas of endemism (unordered): Brazilian Shield (including the Atlantic Shield and Phanerozoic deposits near the Rio de la Plata; fig. 2), Guayana Shield, Amazon basin, Andes, Caribbean (including the adjacent GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE littoral of northern South America), Central America, and tropical west Africa. Similarly, the terminal taxa of Bromeliaceae were coded for elevational distribution (unordered): low (generally less than 1000 m), high (above 1000 m), or widespread (low and high). Elevational data for the more Guayana Shield–restricted Rapateaceae were atomized into lowland (occurring mainly below 500 m), upland (500–1500 m), or highland (mainly >1500 m) (for details of these three types, see Berry et al. 1995; Givnish et al. 2000). To evaluate the effects of using single species as placeholders for large genera (e.g., Tillandsia, Vriesea, Stegolepis), a second analysis was done for each family in which such terminals were coded as polymorphic for all the biogeographic regions and elevation occupied by congeners. Accelerated transformation was used to minimize the number of apparent convergent gains. Spatial shifts in distribution were then related to the chronolo- S41 gies obtained from the analysis of ultrametric trees, lineagespecific mutation rates, and known ages of geological events. Results Phylogeny Global parsimony resulted in 12 most parsimonious trees of length 1617 steps and a well-resolved strict consensus (fig. 3). The consistency index (CI) for these trees was 0.66 (CI9 ¼ 0:57, excluding autapomorphies); the retention index was 0.82. In all, 807 nucleotide sites were variable, of which 510 were phylogenetically informative. Our analysis provides strong bootstrap support for the monophyly of both Rapateaceae (100%) and Bromeliaceae (100%) (fig. 3). The ML Fig. 3 Maximum likelihood tree of Poales based on ndhF sequence variation. The tree is equivalent to one of the 12 most parsimonious trees, except that the trichotomy involving Monotrema through Windsorina is resolved in the latter. Node below Brocchinia serrata collapses in the strict consensus of the maximum parsimony trees. Branch lengths at base of tree were obtained by inclusion of other commelinid lineages from the more global monocot analysis. Bootstrap support values are shown for all nodes outside Bromeliaceae and Rapateaceae; bootstrap support for nodes within these families is given in figs. 4 and 5. S42 INTERNATIONAL JOURNAL OF PLANT SCIENCES tree is identical in branching topology to one of the MP trees and nearly identical in branch lengths. Rapateaceae. Subfamily Rapateoideae is paraphyletic, composed of monophyletic tribes Rapateeae (Cephalostemon, Rapatea, Spathanthus) and Monotremeae (Maschalocephalus, Monotrema, Potarophytum, Windsorina). Rapateeae is sister to all other rapateads, and Monotremeae is sister to subfamily Saxofridericioideae (fig. 4). Our resequencing of Spathanthus bicolor and inclusion of Spathanthus unilateralis makes Rapateeae monophyletic under Fitch parsimony as well (fig. 3) and indicates that the position of S. bicolor in our previous unweighted analysis reflected noise in the original sequence. Our inclusion of monotypic Windsorina confirms its expected membership in tribe Monotremeae. Inclusion of Saxofridericia inermis confirms that it falls in the same clade as Saxofridericia regalis, sister to tribe Schoenocephalieae (Guacamaya, Kunhardtia, Schoenocephalium), which makes tribe Saxofridericieae (Amphiphyllum, Epidryos, Marahuacaea, Phelpsiella, Stegolepis) paraphyletic as well. Bootstrap support for most major groups is strong: 100% for tribes Monotremeae, Saxofridericieae, and Schoenocephalieae plus Saxofridericia; 99% for tribe Schoenocephalieae alone; and 95% for tribe Rapateeae (figs. 3, 4). Exclusion of nonrapateads results in six most parsimonious trees of length 324 steps (CI ¼ 0:87, CI9 ¼ 0:79) and a strict consensus with the same topology as provided by the global analysis. Bromeliaceae. Global parsimony resolves eight major clades of bromeliads (figs. 3, 5). Sister to all other bromeliads is Brocchinia, a genus of ca. 20 species endemic to the tepuis of the Guayana Shield. Monotypic Ayensua is embedded within Brocchinia. Lindmania, a genus of ca. 20 species (Holst 1997) also endemic to the tepuis of the Guayana Shield, is the second clade and is sister to a hard trichotomy comprising all remaining bromeliads. Two of the three lineages at this polytomy include the monophyletic subfamily Tillandsioideae (Catopsis, Glomeropitcairnia, Guzmania, Tillandsia, Vriesea) and the xerophytic genus Hechtia from Central America. The Cratonic clade includes three taxa endemic to the Guayana Shield—Brewcaria, species-rich Navia, and the disparate ‘‘Brocchinia’’ serrata—as well as monotypic Cottendorfia from the Brazilian Shield. The large (ca. 100 spp.) genus Puya, native mostly to the Andes, is sister to the monophyletic subfamily Bromelioideae. This larger group is sister to a group consisting of four highly xeromorphic genera (Abromeitiella, Deuterocohnia, Dyckia, Encholirium), together with more mesomorphic Fosterella of the Andes, and Pitcairnia of the Amazon basin, northern Andes, Guayana Shield, Central America, and the Caribbean. The subfamily Pitcairnioideae (highlighted by solid bars in fig. 5) is massively paraphyletic, forming the matrix from which both Tillandsioideae and Bromelioideae emerge. Reflecting the substantially shorter branch lengths in Bromeliaceae versus Rapateaceae (fig. 3), bootstrap support values for the major bromeliad groups and for relationships among these groups are often lower than those in Rapateaceae. The large clade sister to Brocchinia/Ayensua is strongly supported (97%), as is the large clade sister to Lindmania (84%). The Cratonic clade has strong support (95%), as does the Bromelioideae (99%). Several other relationships within the family, however, have weaker bootstrap support, including 57% for Tillandsioideae (figs. 3, 5). Exclusion of nonbromeliads results in two most parsimonious trees of length 359 steps (CI ¼ 0:82, CI9 ¼ 0:66) and a strict consensus with the same topology as that obtained through the global analysis. Molecular Clocks The ndhF gene shows clocklike evolution in both Rapateaceae (P > 0:12, x2 test, df ¼ 14) and Bromeliaceae (P > 0:053, x2 test, df ¼ 33). On the basis of the average branch length from the stem group to the terminal taxa and Fig. 4 Chronogram for Rapateaceae based on penalized likelihood, showing inferred timing of cladogenesis in the family. Gray bars at each node indicate 6SD of inferred date. Tribes and subfamilies under the new infrafamilial classification proposed in this article (see ‘‘Discussion’’) are highlighted by brackets and bars. Open bar, Saxofridericioideae; gray bar, Monotremoideae; solid bar, Rapateoideae. GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE S43 Fig. 5 Ultrametric tree for Bromeliaceae showing inferred chronology of cladogenesis. Gray bars at each node indicate 6SD of inferred date. Major clades are highlighted by brackets; membership in the three current subfamilies (Smith and Downs 1974, 1977, 1979) is indicated by shaded bars. Open bar, Bromelioideae; gray bar, Tillandsioideae; solid bar, Pitcairnioideae. Note the extreme paraphyly of Pitcairnioideae and monophyly of both Bromelioideae and Tillandsioideae. on the inferred age of each family, the molecular clock in Rapateaceae appears to tick 53% faster (1:26 6 0:16 mutations/Ma) than in Bromeliaceae (0:84 6 0:09 mutations/Ma). This difference is highly significant (P < 2 3 10
9 , two-tailed t-test, df ¼ 39). For grasses plus Joinvillea, ndhF evolves nearly twice as fast (2:37 6 0:64 mutations/Ma) as in rapateads and nearly three times as fast as in bromeliads. Historical Biogeography Cross-verification yielded an optimal smoothness parameter of 102.50 for the PL algorithm for the monocot-wide analysis of ndhF. Dates of crown group radiations for Rapateaceae and Bromeliaceae are based on PL analyses of the eight ML trees from the monocot-wide ndhF data set, representing different arrangements of commelinid orders (ta- ble 2). These average to 41 and 23 Ma for Rapateaceae (fig. 4) and Bromeliaceae (fig. 5), with standard deviations of 3.8 and 2.6 Ma, respectively. The stem lineages of the two families go back 87 and 84 Ma, respectively. Chronograms show times of radiation and subsequent divergence with standard deviations for Rapateaceae and Bromeliaceae (figs. 4, 5). With both Rapateaceae and Bromeliaceae, dates based on PL are generally consistent with those based on lineage-specific mutation rates or local molecular clocks but are much more recent than those based on NPRS. NPRS results are often 50%–70% greater in value than corresponding PL dates. The shifts in elevation for both families are based on ACCTRANS optimization onto their respective phylogenetic trees (fig. 6). Shown are shifts when the individual species elevations are scored. The elevational shifts are identical in S44 INTERNATIONAL JOURNAL OF PLANT SCIENCES Fig. 6 Inferred patterns of elevational shifts in Rapateaceae (left) and Bromeliaceae (right) based on modern distribution of terminal taxa, branching topologies shown in figs. 4 and 5, and parsimony. Rapateaceae when species-rich genera are coded to accommodate interspecies variation. The pattern in Bromeliaceae becomes less resolved (i.e., more equivocal branches) at the genus level, but the main pattern of shifts from low elevations to higher elevations multiple times is maintained (not shown). Likewise, DELTRANS optimization is identical to that shown for Rapateaceae and only slightly different at the tips for widespread genera for Bromeliaceae. The crown separation of Rapateaceae into two main lineages (Rapateeae and remainder of family) occurred at ca. 41 Ma, ca. 46 Ma after the stem lineage of Rapateaceae appears (fig. 4). Within Rapateeae, separation of Spathanthus from the clade of Rapatea and Cephalostemon quickly followed the initial family radiation, around 38 Ma. This radiation and most of the subsequent divergences of clades within the largely Guayana Shield–restricted Rapateaceae were lowelevation events based on a parsimony reconstruction of ancestral elevations (fig. 6). The separation of largely lowland Monotremeeae (except for upland Potarophytum) from the upland or highland centered clade of three tribes of subfamily Saxofridericioideae occurred around 29 Ma. Within Monotremeeae, separation of the African lineage leading to Maschalocephalus from the American taxa is quite recent, 7.3 Ma. On the basis of parsimony analysis of elevation, the ancestral elevation for the subfamily Saxofridericioideae is equivocal (fig. 6). Beginning ca. 10 Ma, several shifts to both uplands and the highlands occur in the subfamily, and one shift (reversal) back to low elevation occurs in the common ancestor of Schoenocephalium and Guacamaya. In Bromeliaceae, the stem lineage arose 84 Ma, even deeper in the past than in Rapateaceae and even more removed from the initial diversification of the crown group at 23 Ma. Initial branching gave rise to the clades comprising Brocchinia and the embedded Ayensua versus all other bromeliads (fig. 5). As with Rapateaceae, initial lineages in the Bromeliaceae apparently inhabited low elevations (fig. 6) within the Guayana Shield (fig. 7). Even within Brocchinia, where many species are high-elevation specialists in tepuis, the first split is between low- and high-elevation clades (fig. 6). Diversification within Brocchinia appears to have started at 20 Ma and intensified around 15 Ma. Separation of high-elevation, tepui-restricted Lindmania from the remainder of the second clade (19 Ma) coincided with this diversification in Brocchinia. An important diversification at an unresolved trichotomy around 17 Ma generated three large clades: subfamily Tillandsioideae (unresolved in biogeographic origin), Hechtia (Central America), and the remainder of Pitcairnioideae plus all Bromelioideae (origin in the Guayana Shield) (figs. 5, 7). The Tillandsioideae and Bromelioideae crown radiations occurred at 15 and 9.4 Ma, respectively. The Pitcairnioideae GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE S45 pateeae, erect a new subfamily Monotremoideae to contain the highly distinctive tribe Monotremeae, restrict tribe Saxofridericieae to Saxofridericia alone, and erect a new tribe Stegolepideae to contain Amphiphyllum, Epidryos, Marahuacaea, Phelpsiella, and Stegolepis. We make these changes formally below in a revised infrafamilial classification of Rapateaceae. Subfamily Rapateoideae Maguire. Mem. New York Bot. Gard. 10: 21. 1958.—Carpels uniovulate; involucral bracts much exceeding the fertile part of the inflorescence, either single or paired and valvate; seeds ovoid-oblongoid and either striate or else smooth with a papillate apical appendage. Tribe Rapateeae Endlicher. Gen. Pl. 1: 131. 1836. (Rapatea, Spathanthus, Cephalostemon, Duckea) Tribe Cephalostemoneae Maguire, Acta Amaz. 9:267, 1979; Tribe Spathantheae Maguire, Acta Amaz. 9:267, 1979. Subfamily Monotremoideae Givnish & P. E. Berry, Subfam. Nov. Carpellis uniovulatis; bracteis involucralibus Fig. 7 Inferred pattern of biogeographic diversification in Bromeliaceae based on modern distribution of terminal taxa, branching topology, and parsimony. s.s. diverged around 14 Ma, possibly in the Andes, with the African Pitcairnia feliciana separating from American bromeliads slightly thereafter, 12 Ma (figs. 5, 7). The two divergent members sampled for Puya, a genus exhibiting a large radiation in the Andes, appear to have diversified only 700,000 yr ago. Discussion Phylogeny and Systematic Implications Rapateaceae. Within Rapateaceae, our findings on the monophyly of and relationships among the four tribes and two subfamilies recognized by Maguire (1958, 1979, 1982, 1984) confirm and extend those derived by Givnish et al. (2000). Inclusion of Spathanthus unilateralis shows that Spathanthus is sister to Cephalostemon and Rapatea and that tribe Rapateeae is thus monophyletic. Inclusion of lowland Saxofridericia inermis—in a portion of the genus other than that occupied by previously studied Saxofridericia regalis— demonstrates that Saxofridericia is monophyletic and that, with the addition of another species, it continues to occupy its surprising position sister to tribe Schoenocephalieae, making tribe Saxofridericieae as defined by Maguire (1979) paraphyletic. Finally, inclusion of the Potaro River endemic Windsorina confirmed that it belongs, as expected, to tribe Monotremeae, that tribe Monotremeae remains sister to subfamily Saxofridericioideae, and that subfamily Rapateoideae as defined by Maguire (1979) is paraphyletic. The simplest way to eliminate the paraphyly of Rapateoideae and Saxofridericieae is to restrict the former to tribe Ra- numerosis, imbricatis, vel solitariis subter omni flore; seminibus albo-granulatis, appendicibus apicalibus applanatis.— Carpels uniovulate; involucral bracts several and imbricate or else each spikelet subtended by its own bract, mostly shorter than the fertile part of the inflorescence; seeds ovoidoblongoid, white-granulate, with flattened apical appendage. Tribe Monotremeae Maguire. Mem. New York Bot. Gard. 10: 21. 1958. (Monotrema, Potarophytum, Windsorina, Maschalocephalus.) Subfamily Saxofridericioideae Maguire. Mem. New York Bot. Gard. 10:21,1958.—Carpels multiovulate; seeds not uniformly ovoid or oblongoid, rather prismatic, pyramidal, lenticular, or crescent shaped. Tribe Saxofridericieae Maguire. Mem. New York Bot. Gard. 10:21, 1958. (Saxofridericia)—Scapes one per leaf axil; involucral bracts two, connate-saccate and pierced by the emerging flowers; petals obcordate, yellow, with spreading blades. Tribe Schoenocephalieae Maguire. Mem. New York Bot. Gard. 10:21, 1958. (Schoenocephalium, Kunhardtia, Guacamaya)—Scapes one per leaf axil; involucral bracts two, valvate to connate-saccate; petals lanceolate and remaining tightly imbricate, red, white, or yellow; flowers nectar-producing. Tribe Stegolepideae Givnish & P. E. Berry, trib. nov. Scapis vulgo multiplicibus; bracteis involucralibus nullis vel imbricatis raro valvatis; petalis unguiculatis, flavis raro niveis.—Scapes generally several per leaf axil; involucral bracts lacking, imbricate, or rarely valvate; petals with basal claw and usually spreading, yellow or whitish, rounded blade. (Stegolepis, Amphiphyllum, Epidryos, Marahuacaea, Phelpsiella.) Saxofridericia occupies a noteworthy position in this revised classification. It shares several characters (e.g., partial lignification of the stelar parenchyma; Carlquist 1966) with tribe Stegolepideae and yet is sister to tribe Schoenocephalieae, making it difficult to define Stegolepideae based on any single morphological character (only multiple inflorescences per leaf axil appear to serve as a synapomorphy). The new Saxofridericieae might have been distinguished from Stegolepideae and Schoenocephalieae simply by saccate bracts that the opening flowers pierce, except that some populations S46 INTERNATIONAL JOURNAL OF PLANT SCIENCES of Kunhartia share this trait as well. Several of the traits defining Schoenocephalieae appear to be directly related to hummingbird pollination (Givnish et al. 2000). In essence, Schoenocephalieae emerged recently from within a morphologically cohesive but paraphyletic group (Stegolepideae plus Saxofridericia). Paraphyly of the latter became evident only with molecular data, much as Sytsma and Gottlieb (1986) discerned the emergence of Heterogaura from within paraphyletic but morphologically cohesive Clarkia. If Saxofridericia had become extinct and saccate populations of Kunhardtia had disappeared, it would have been easy to describe several single-character synapomorphies for Stegolepideae (as well as the synapomorphies already discernible for Schoenocephalieae). Alternatively, if evolution had proceeded further in Saxofridericieae and Stegolepideae, single-character synapomorphies might have appeared. However, even though we can clearly recognize Stegolepideae, Saxofridericieae, and Schoenocephalieae as monophyletic entities on the basis of molecular data, they have diverged so recently that it is difficult, if not impossible, to distinguish at least the first two tribes from each other on the basis of any single characteristic. Bromeliaceae. Bromeliads have traditionally been divided into three subfamilies: Pitcairnioideae with winged (or rarely naked) seeds, Tillandsioideae with plumose seeds, and Bromelioideae with fleshy fruits (Smith and Downs 1974, 1977, 1979; Gilmartin and Brown 1987). Over the past 15 yr, several attempts have been made to infer relationships among these subfamilies and their constituent genera using morphological or molecular data (e.g., Gilmartin and Brown 1987; Clark and Clegg 1990; Givnish et al. 1990, 1992; Ranker et al. 1990; Clark et al. 1993). Importantly, Terry et al. (1997a) using ndhF sequences found (i) that Brocchinia is sister to the rest of the family, (ii) that Tillandsioideae is monophyletic and sister to remaining elements of the family, (iii) a monophyletic subfamily Pitcairnioideae s.s., and (iv) that Puya is sister to a monophyletic Bromelioideae. Horres et al. (2000) using trnL sequences identified three major clades whose relationships with each other were unresolved: Brocchinia plus Ayensua, Hechtia, and all remaining bromeliads. The latter clade, in turn, split into an unresolved polytomy of five clades, including a monophyletic Bromelioideae. Finally, Crayn et al. (2000) using partial matK sequences found little phylogenetic structure beyond support for individual genera in most cases. Our phylogeny is consistent with the old suggestion that the bromelioids and tillandsioids arose from within the Pitcairnioideae (Schimper 1888; Mez 1904; Pittendrigh 1948; Tomlinson 1969; Smith and Downs 1974; Benzing et al. 1985; Smith 1989; Benzing 1990). Terry et al. (1997a) reached a similar conclusion but had a different impression of the proximity of bromelioids and tillandsioids and the seeming isolation of Brocchinia as a result of their not having sampled three of the eight major clades of Bromeliaceae. They also concluded, on the basis of the results of Ranker et al. (1990) and their own belief that Encholirium and Hechtia were essentially interchangeable, that Hechtia was closely allied to Dyckia, Encholirium, Abromeitiella, and Deuterocohnia rather than representing a convergent lineage much closer to the base of Bromeliaceae. Benzing’s (2000) treatment of the ‘‘adaptive radiation’’ of the bromeliads based on the phylogeny of Terry et al. (1997b) was premature: too many groups, with crucial biogeographic distributions and physiological/morphological adaptations, were not included. Givnish et al. (2004a) present a detailed scenario for the adaptive diversification of bromeliads on the basis of the phylogeny and biogeographic reconstruction presented here. Contrary to previous claims (e.g., Benzing 2000), until now there had been no phylogenetic evidence within Bromeliaceae that the family arose in the Guayana Shield; the distribution there of earliest divergent Brocchinia might have simply been an autapomorphy. The classical proposals that bromelioids and tillandsioids emerged from a pitcairnioid ancestor were based not on phylogenetic analysis but on the simple observation that epiphytes—an ecologically advanced growth form involving many specializations for life without contact with the soil—were far more numerous outside the pitcairnioids. Not one of the early writers, however, suggested that Brocchinia or Lindmania might lie at the base of the family or that Pitcairnioideae was not monophyletic or otherwise ‘‘unnatural.’’ Smith (1934) proposed that Puya might represent something close to the ‘‘ur-bromeliad’’ or ancestral bromeliad, but molecular data show that Puya arose quite recently, though it might now be better seen as the ‘‘ur-bromelioid.’’ Smith’s (1934) suggestion that Rapateaceae evolved from within Bromeliaceae via Navia is clearly incorrect (Terry et al. 1997a; Givnish et al. 1999, 2000; this article). Our results include a number of other important new findings. First, Ayensua is embedded within Brocchinia with this clade sister to the remainder of the family. Brocchinia is a remarkable genus of ca. 20 species endemic to rainy, extremely infertile habitats atop the tepuis and adjacent lowland sandplains of the Guayana Shield. Brocchinia has undergone an adaptive radiation in mechanisms of nutrient capture that is unparalleled at the generic level in angiosperms (Givnish et al. 1997). Brocchinia prismatica is sister to a clade comprising Ayensua, Brocchinia acuminata, and Brocchinia paniculata. The latter species is a member of the Melanacra clade (Givnish et al. 1997). Ayensua uaipanensis shares with another member of the Melanacra clade, Brocchinia maguirei, the highly unusual characteristic of a sharply defined abscission zone at the base of each leaf. Second, the tepui endemic Lindmania is the sister to all other bromeliads except Brocchinia and Ayensua. Third, a hard polytomy exists with Tillandsioideae, Hechtia, and all remaining bromeliads. Fourth, we identify a ‘‘Cratonic clade,’’ in allusion to its restriction to the single craton on which both the Brazilian and Guayana Shields ride. Givnish et al. (1997) resolved Brewcaria, Navia, and ‘‘Brocchinia’’ serrata as close relatives on the basis of cpDNA restriction site variation; the last species diverges from all other taxa classified as Brocchinia in having a superior ovary and spinescent leaves. The association of Cottendorfia with Navia, among other genera, in the Cratonic clade is supported by the sister relationship of Cottendorfia florida with Navia phelpsiae on the basis of matK sequence variation (Crayn et al. 2000). However, the latter study also placed Navia igneosicola sister to almost all species of Pitcairnia (including Pepinia) in a separate clade, with a decay value of 1. Navia is a large and complex genus recently revised by Holst GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE (1997), who found the previous concept of the group overly broad and segregated all of the noncapitate species into Brewcaria, Steyerbromelia, and Brocchinia. The validity of these shifts has yet to be tested using molecular data, and it may well be that some species now placed in Navia will ultimately be excluded from this genus. Fifth, our results place Abromeitiella and Deuterocohnia sister to each other and jointly sister to Dyckia and Encholirium. Sixth, this study confirms Fosterella as sister to this larger group of highly xeromorphic genera. Seventh, our ndhF phylogeny shows that the large genus Puya, native mostly to the Andes, is sister to the monophyletic subfamily Bromelioideae (figs. 3, 5). Historical Biogeography Our findings demonstrate that Rapateaceae and Bromeliaceae both arose in the Guayana Shield of northern South America, spread centripetally from there, and reached tropical west Africa via long-distance dispersal relatively recently, 6–8 Ma if we use calculations based on cross-verified PL. Similar dates are obtained if local molecular clocks or lineagespecific mutation rates are employed. We believe that the substantially older estimates that arise from nonparametric rate smoothing reflect a bias based on the method, for two reasons. First, NPRS is known to perform poorly in clades that exhibit clocklike evolution (Sanderson 2002). Thus, while NPRS may be an appropriate method for estimating dates across the monocots as a whole, which exhibit strong heterogeneity in the rate of molecular evolution (Gaut et al. 1993, 1997), it probably is inappropriate for estimations within either Bromeliaceae or Rapateaceae. Second, and more importantly, in both families the lineage containing the sole African taxon has substantially shorter branch lengths than do flanking groups (77:8 6 1:7 steps for Monotremeae vs. 101:1 6 5:3 steps for other rapateads; 59:0 6 1:6 steps for Pitcairnia vs. 68:3 6 3:7 steps for other Pitcairnioideae s.s., Puya, and bromelioids). These differences are highly significant (P < 2 3 10
9 for bromeliads, P < 10
5 for rapateads, based on two-tailed t-tests) despite both families having passed the Felsenstein test for clocklike molecular evolution. Short branches in one lineage flanked by long branches in other lineages are likely to be dilated by nonparametric rate smoothing and might lead to unrealistically ancient ages for dispersal/vicariance events in that lineage. This appears to have happened in both families; it would be useful to conduct simulations to see how sensitive NPRS is to such bias, given its increasing use in phylogenetic and biogeographic reconstructions. On the basis of rbcL sequence data, Givnish et al. (2000) argued that Bromeliaceae, Mayacaceae, and Rapateaceae form a clade and that this group was likely to share a common ancestor from the Guayana Shield. Detailed analyses based on more extensive data—involving rbcL, atpB, and 18S (Chase et al. 2000) or ndhF (this article; Givnish et al. 2004b)—do not, however, support this clade and instead point to a close but nonsister relationship among these families (fig. 3). We note that it is still possible, even likely, that the broader group containing these three families (with Bromeliaceae earliest divergent) may still have arisen in the Guayana Shield, given the origins of both Bromeliaceae and S47 Rapateaceae there, the likely origin of Xyridaceae there (on the basis of the unique coexistence of all genera of the family there as well as all chromosomal numbers within wideranging Xyris; Givnish et al. 2000), the occurrence of numerous genera of Eriocaulaceae there, the presence of Mayaca, and the position of the Guayana Shield endemic Thurnia at the base of the closely related Cyperaceae (Chase et al. 2000; Givnish et al. 2000). Mayaca has the same biogeographic pattern as that seen in Bromeliaceae and Rapateaceae, with one species in west Africa and several in the Americas; it would be interesting to reconstruct the evolutionary history of this genus next. Rapateaceae. The evolutionary chronology of Rapataceae derived in this article (fig. 4) has several implications for the historical biogeography of the family, all consistent with the detailed discussion provided by Givnish et al. (2000). The latter results are modified by the inclusion of additional taxa, a new estimated time of origin (87 Ma vs. 62 Ma), and new estimated times of cladogenesis within the family, based on the new position of Spathanthus and the use of PL versus lineage-specific rates of molecular evolution. We infer that Rapateaceae arose in the Guayana Shield of South America. All genera except Maschalocephalus occur in the Guayana Shield, and all of those except Spathanthus, Cephalostemon, Rapatea, Epidryos, and Stegolepis are restricted to the Guayana Shield (Maguire 1979, 1982, 1984; Givnish et al. 2000; P. E. Berry, in preparation). Epidryos is unique in the family in possessing viscid seed appendages (P. E. Berry, in preparation), fitting its epiphytic role and facilitating longdistance seed dispersal and a range including parts of the Guayana Shield, Panama, Colombia, and Ecuador. A disjunct population of Stegolepis has also recently been discovered in Ecuador (P. E. Berry, personal communication). We also infer that Rapataceae arose in the lowlands and then invaded highland habitats as subfamily Saxofridericioideae diversified or possibly later with the rise of Stegolepis and allied genera (Givnish et al. 2000). This finding parallels the patterns documented in Brocchinia (Givnish et al. 1997) and Guayana Shield Gentianaceae (Struwe et al. 2002), the two groups endemic to the Guayana Shield with taxa at a variety of elevations that have been the subject of detailed phylogenetic analyses. In Rapateaceae, secondary reinvasions of the lowlands may have happened twice: in tribe Schoenocephalieae (one widespread species of Kunhardtia is entirely montane, while the other is rare and occurs in the lowlands) and in Saxofridericia. Additional taxa must be studied to determine whether the montane distribution of several Saxofridericia and most populations of Kunhardtia is a derived trait shared by Stegolepis and related genera or whether they instead represent independent colonization(s) of the highlands (Givnish et al. 2000). The early divergent, low-elevation genera of tribe Rapateeae apparently were able to spread early and widely along riverine corridors before the uplift of the northern Andes and resulting shift in the Amazon ca. 20 Ma, separating the Guayana and Brazilian Shields with a band of edaphically distinct, noncratonic sediments (Givnish et al. 2000). Most of these genera are found in swamps, thickets, and caatinga along riverine corridors (Givnish et al. 2000). Given Rapatea’s occurrence on siltier substrates, its appearance in S48 INTERNATIONAL JOURNAL OF PLANT SCIENCES Amazonia and Bahia may well reflect more recent overland dispersal. On the basis of PL, Spathanthus appears to have evolved roughly 38 Ma (fig. 4), before the rise of the central and northern Andes and the associated shift of the Amazon ca. 20 Ma. The Amazon has been running its current course to the Atlantic Ocean since at least the Miocene, separating the Guayana and Brazilian Shields with a belt of edaphically distinct, nutrient-rich sediments derived from the Andes (Hoorn 1994; Hoorn et al. 1995; Potter 1997). The apparent divergence of Cephalostemon and Rapatea later than this, only 21 Ma (fig. 4), indicates that they might each have colonized the Brazilian Shield via short-distance dispersal at that time. Rapatea’s occurrence on siltier substrates suggests that it may have colonized (or recolonized) parts of the Amazon basin even more recently. Maschalocephalus appears to have diverged from its South American relatives 7.3 Ma (fig. 4), long after the opening of the tropical Atlantic ca. 84–102 Ma (Sidder and Mendoza 1991; Edmond et al. 1995). Even if Maschalocephalus colonized Africa much earlier, about the time the tribe Monotremeae as a whole arose 29 Ma, it would still have been too late for continental drift to have played a direct role in its occurrence in Africa. Our PL calculation that Maschalocephalus arrived in Africa 7.3 Ma is quite close to estimates based on local molecular clocks (6.9 Ma) and lineage-specific mutation rates (7.6 Ma). NPRS generates a substantially older age of 19 Ma; even this is far too late for Maschalocephalus to have colonized Africa via vicariance. However, NPRS is probably an inappropriate approach, given that evolution in Rapateaceae is clocklike and that other biases may apply in this case (see ‘‘Discussion’’). A recent origin of Maschalocephalus via long-distance dispersal ca. 7 Ma might help explain the lack of rapatead speciation in Africa (Good 1974). Historical cycles of aridity in Africa (Goldblatt 1993; Querouil et al. 2003) probably played a role in preventing (or subsequently extirpating) other colonists and slowing speciation, given the present restriction of all rapateads to mesic or wet habitats and Maschalocephalus’s occurrence in sandstone areas of relatively heavy (albeit seasonal) rainfall from Sierra Leone to Côte d’Ivoire. Our results and those of Givnish et al. (2000) thus indicate that, if continental drift played a role in creating the amphiatlantic distribution of Rapateaceae, it did so by creating similar sandy, wet, nutrient-poor substrates favorable to rapateads on both sides of the Atlantic, not by rafting rapateads across it. Vicariance of habitat, not vicariance of taxa, accounts for the observed intercontinental disjunction. The invasion of tepui habitats by subfamily Saxofridericioideae appears to have facilitated isolation and rapid speciation atop individual tepuis, driven by vicariance and dispersal (Givnish et al. 2000). Most montane genera and species of Rapateaceae are narrowly endemic to small areas or individual tepuis within the Guayana Shield. Amphiphyllum and Marahuacaea (allied to Stegolepis; Givnish et al. 2000) are restricted to the neighboring tepuis of Duida and Marahuaca in central Amazonas, Venezuela (fig. 2), Phelpsiella (also allied to Stegolepis; Givnish et al. 2000) is restricted to Parú, ca. 60 km to the north, and most species of Stegolepis are restricted to one or a few tepuis each, with each tepui having one or more species present (Huber 1988). The montane clade containing Stegolepis and allied genera arose 15 Ma and began to diversify only 9 Ma (fig. 4). Tribe Schoenocephalieae arose more recently, ca. 10 Ma, appears to have begun diversifying 4.3 Ma, and is now restricted to the western Shield: Kunhardtia in the Sierra Maigualida, Sipapo, Autana, Cuao, Guanay, Yutajé, Coro-Coro, and nearby lowlands in northern Venezuelan Amazonas; Schoenocephalium in Amazonian savannas on white sand in nearby portions of western Amazonas and adjacent Colombia and Brazil, in the drainages of the Orinoco, Guainı́a, Atabapo, Ventuari, Inı́rida, Vaupés, Caquetá, and Rio Negro; Guacamaya in a small area of Amazonian savannas along the Guainı́a and Atabapo (fig. 2). Given that the most diverse montane lineage, the newly recognized Stegolepideae, arose only 15 Ma (fig. 4), it appears that dispersal must have been a dominant factor. Givnish et al. (2000) estimated that chemical dissolution and erosion should cause the margins of adjacent tepuis to retreat from each other at a rate of ca. 2.4 km/Ma. Such low rates would permit vicariance to play a role only on tepuis that are currently quite close to each other (e.g., Duida, Marahuaca, Huachamacari; fig. 2). Vicariance seems unlikely to have played a role in speciation atop tepuis that are now further apart than ca. 35 km. Despite the lack of adaptations for long-distance dispersal in most rapateads, dispersal seems likely to have led to the isolation and subsequent speciation of Stegolepideae populations atop more distant tepuis, which would have been separated from each other long before the stegolepids arose (Givnish et al. 2000). Infrequent longdistance dispersal (perhaps via waterfowl) would be consistent with the recent origin via dispersal of Maschalocephalus and with extensive speciation and narrow endemism in montane saxifridericioids, in which the habitat is naturally separated into ‘‘islands in the sky,’’ patches of cool, rainy, exceedingly humid, and extremely infertile habitats scattered in a sea of lowland tropical forests and savannas. Infrequent dispersal over geographic scales comparable to those between large tepuis appears to have spurred speciation and endemism in several plant groups in the Hawaiian Islands (e.g., Alsinidendron-Schiedea: Soltis et al. 1996; the silversword alliance: Baldwin 1997; Baldwin and Sanderson 1998; Cyanea: Givnish et al. 1994, 1995; Givnish 1998; Hawaiian violets: Ballard et al. 1998; Hawaiian mints: Lindqvist and Albert 2002) and a number of other oceanic archipelagoes, such as Macaronesia (Böhle et al. 1996; Francisco-Ortega et al. 1996, 1997; Kim et al. 1996, 1999; Mes et al. 1996; Mort et al. 2002) and the Juan Fernandez Islands (Sang et al. 1994, 1995). We plan to examine the role of long-distance dispersal via vicariance in speciation atop the tepuis by developing molecular phylogenies for rapatead and bromeliad genera whose individual species are mostly restricted to single tepuis and then using these trees to test whether close relatives generally occur on adjacent tepuis (as expected from vicariance) or whether there are frequent exceptions to this rule (as expected from long-distance dispersal). Bromeliaceae. Based on PL estimates, 61 Ma elapsed between the origin of the bromeliad stem group and divergence among the crown group of surviving lineages 23 Ma (fig. 5). This long time period suggests a great deal of extinction over the intervening period and helps account for the morphologically isolated position of the Bromeliaceae. The nesting of GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE two lineages, Brocchinia and Lindmania, endemic to the Guayana Shield at the base of Bromeliaceae implies this large (ca. 2600 spp.) family arose there, with subsequent centripetal invasions of other parts of South and Central America as well as tropical west Africa (figs. 5, 7). The current lineages of Brocchinia began to diversify 20 Ma (fig. 5). Brocchinia originated at low elevations and then evolved adaptations to nutrient poverty (carnivory, ant-fed myrmecophily, N2 fixation) that depended on evolution of the tank habit and live absorptive trichomes, all of which were contingent on occupying rainy, humid, extremely nutrient-poor habitats at high elevations on the tepuis (Givnish et al. 1984, 1997). Brocchinia’s distribution is coextensive with the Guayana Shield, with almost all species occurring on sand or sandstone on the tepuis or sandplains at low elevations; a few species also occur on granite outcrops at the edge of the Shield or on low sandstone mesetas in southwestern Colombia (Givnish et al. 1997). The northern Andes and Central America appear to have been invaded independently by members of at least three major lineages: the higher tillandsioids (Guzmania, Tillandsia, Vriesea) beginning ca. 15 Ma; Fosterella, beginning ca. 10 Ma; and (involving the entire Andean cordillera) Puya, beginning at least 10 Ma (figs. 5, 7; all calculations based on stem groups). We have no way of knowing, given the low intensity of sampling at this point, whether each of these lineages involved single or multiple invasions of the Andes. Several additional groups (e.g., some Pitcairnia, bromelioids) have also independently colonized the Andes, but we do not have adequate sampling to estimate the time(s) or numbers of their arrivals. The Brazilian Shield was independently invaded at least three times: by Cottendorfia from the Guayana Shield ca. 12 Ma, by Dyckia-Encholirium from the central Andes ca. 9 Ma, and by higher Bromelioideae ca. 7 Ma (figs. 5 , 7). Individual members of other large groups (e.g., higher tillandsioids) have doubtless also colonized this area independently. We estimate that Pitcairnia feliciana arrived in tropical west Africa via long-distance dispersal no earlier than 12 Ma. Most of the present-day diversity of the family involves lineages that have appeared only in the past 17 Ma, including the subfamilies Tillandsioideae and Bromelioideae as well as the remaining large genera Hechtia, Dyckia, Pitcairnia, Navia, and (perhaps very recently, at least for modern lineages) Puya. This time period around 17 Ma ushered in the rise of essentially xeromorphic lineages, including the genus Hechtia (with heavily armed, succulent leaves and CAM photosynthesis), the subfamily Tillandsioideae (many of which are epiphytes, with several members of the Guzmania-Tillandsia-Vriesea lineage having CAM photosynthesis), and the remaining higher bromeliads (almost all having spinescent leaves, with several [Puya, Bromelioideae] also having CAM photosynthesis). Hechtia and the tillandsioids represent the first bromeliad invasions of Neotropical regions outside the Guayana Shield. Both cases appear to involve northern portions of South America, with Hechtia invading Central America. Phylogenetic reconstruction suggests that subfamily Tillandsioideae arose in the Guayana Shield (fig. 7), but that is an artifact of the choice of species used to represent Guzmania, Tillandsia, and Vriesea. The latter genera have widespread distributions S49 and are especially diverse in the Andes (Smith and Downs 1977). More pertinently, Glomeropitcairnia is restricted to parts of northern Venezuela and Trinidad immediately adjacent to the Guayana Shield, and some species of Catopsis (e.g., Catopsis berteroniana) occur on the Guayana Shield as part of a more extensive range (Smith and Downs 1974). Most of the remaining genera have ranges overlapping the Guayana Shield but have invaded several other regions and are more diverse elsewhere, especially in the Andes. This suggests that the subfamily may have originated by dispersing just beyond the periphery of the Guayana Shield to the Caribbean littoral. Origin of the Guzmania-Tillandsia-Vriesea clade within Tillandsioideae—including the most highly specialized, ‘‘atmospheric’’ epiphytes in the family, with a center of diversity in the Andes—began ca. 15 Ma (fig. 7), well after the rise of the central and northern Andes ca. 20 Ma (see above). Three of the four genera of the Cratonic clade—Brewcaria, Navia, and ‘‘Brocchinia’’ serrata—are restricted to the Guayana Shield. On the basis of PL, the Cratonic clade originated 17 Ma, after the Amazon is thought to have changed course and ecologically separated the Guayana and Brazilian Shields ca. 20 Ma. We estimate that Brazilian Cottendorfia diverged from these other Cratonic clade taxa 12 Ma (fig. 5), raising long-distance dispersal as a likely mechanism of Cottendorfia’s distribution. Pitcairnioideae sensu stricto diverged from the ancestor of Puya and Bromelioideae 15 Ma (fig. 5). Phylogenetic reconstruction indicates that the ancestor of these three lineages arose in the Andes (fig. 7), but we suspect that Andean distributions arose independently, substantially later, in Pitcairnioideae s.s. and Puya. This suspicion is based on (i) the uplift of the northern and central Andes ca. 20 Ma, (ii) the apparent origin of Puya much later than this (11 Ma), (iii) the origin of high Andean Abromeitiella and Deuterocohnia at 10 Ma, and (iv) the origin of Fosterella at midelevations in the Andes 10 Ma (fig. 5). Some Pitcairnia occur in the northern Andes (Smith and Downs 1974). The Andean species of Pitcairnia are of unknown phylogenetic position, but even if the genus arose there, it would have done so no more than 12 Ma (fig. 5), long after the northern and central Andes were uplifted. The distributions of genera within Pitcairnioideae s.s. indicate that the group represents the result of a counterclockwise invasion from the Guayana Shield into the northern Amazon basin and/or northern Andes, then into the central Andes, and finally into their southern foothills and drier portions of the Brazilian Shield and Bahia (figs. 7, 8). Divergence between Fosterella and its sister clade may have taken place in southern Bolivia and northern Argentina, where midelevation Fosterella overlaps with the Andean genera Abromeitiella and Deuterocohnia (fig. 8). The latter genera are very closely related and sometimes synonymized (Spencer and Smith 1992). Divergence between the Andean genera and the drought-adapted genera Dyckia and Encholirium, restricted to lower elevations in the Andean foothills and the Brazilian Shield, appears to have occurred in northern Argentina. An invasion route from the southern Andes eastward to Bahia is indicated by Dyckia’s range, which abuts Deuterocohnia in the west and Encholirium in the east on the horn of Brazil. Further research is required to determine whether highly xeromorphic Encholirium is sister to a monophyletic Dyckia or S50 INTERNATIONAL JOURNAL OF PLANT SCIENCES Fig. 8 South American distributions of genera of Pitcairnioideae s.s. excluding Pitcairnia, which is widespread in the northern Andes, Central America, Caribbean littoral, Guayana Shield, and northern Amazonia. Note extensive overlap of genera in southern Bolivia and northern Argentina and apparent counterclockwise dispersal from northern Andes, Guayana Shield, and Amazon Basin into the central/ southern Andes, Brazilian Shield, and arid Bahia. Distributions adopted from Smith and Downs (1974). A single species of Fosterella (Fosterella micrantha) occurs disjunctly in Central America. instead derived from within it. Divergence between the species representing each of these genera occurred quite recently, within the Pleistocene spanning the last 2 Ma; Abromeitiella and Deuterocohnia appear to have diverged even more recently, ca. 1.7 Ma (fig. 5). As with Maschalocephalus, the PL dates exclude vicariance via continental drift as a possible explanation for the occurrence of P. feliciana in Africa and point instead to relatively recent long-distance dispersal. On the basis of PL, African P. feliciana diverged from its American counterparts 12 Ma, slightly before Maschalocepalus diverged from its relatives (see above). It should be noted, however, that we failed to represent a large amount of morphological diversity within the genus, including P. feliciana’s closest relatives from the New World. The latter are thought possibly to include Pitcairnia fuertesii from the Caribbean, Pitcairnia pungens from Ecuador and Peru (Harms and Mildbraed 1938), or certain unnamed saxicolous species from eastern Brazil (Porembski and Barthlott 1999), though detailed justification for both of these views is lacking. Porembski and Barthlott (1999) do show, however, that P. feliciana diverges from all other Pitcairnia studied (which is hardly all of the species of Pitcairnia, given the size of the genus) in a number of characteristics. Pitcairnia feliciana differs from all Pitcairnoideae in possessing a ligula with two toothlike scales (although the adnate scales in Pitcairnia pseudopungens, which the authors presented in their fig. 5, are not that different). The species also has papillae on the stigma lobes that do not occur in other Pitcairnia. Its possession of testa cells with perforated outer periclinal walls is also unparalleled in Pitcairnia, and indeed only one species of Puya is known to share the trait. It is not clear whether these traits are simply autapomorphies of P. feliciana or instead imply synapomorphies that unite all other Pitcairnia while placing it as the sister to the rest of the genus or even outside it. Resolution of this quandary will require studying a large number of species within Pitcairnia. Crayn et al. (2000) have already found, however, that eight of nine species of Pitcairnia sequenced for matK form a clade. If indeed P. feliciana is sister to all other members of the genus, the 12-Ma date is the maximum date for this intercontinental disjunct pattern. The lineages leading to Puya and subfamily Bromelioideae separated 11 Ma, with Puya diversifying mostly along the length of the Andes and the bromelioids arising in one of several places, including northern South America, before invading the Brazilian Shield ca. 7 Ma (figs. 5, 6). Some of the early divergent bromelioids (including Bromelia and the pineapple Ananas) occur in a wide range of areas, including seasonal parts of northern South America and Central America, but a majority of later divergent genera are endemic to seasonal parts of the Brazilian Shield (Smith and Downs 1977). Many of these latter genera, despite their possession of fleshy fruits and possible capacity for long-distance dispersal, have quite limited ranges, and it seems likely that they represent the product of diversification in situ within the Brazilian Shield. More species of the large genus Puya (ca. 100 spp.) need to be sequenced to obtain a better estimate of when present-day species began to diverge from each other. We estimate that the relatively small, wide-ranging Puya floccosa and the gigantic Puya raimondii from the central Andes diverged only 700,000 yr ago (fig. 5). Contrasts, similarities, and their roots. One of the striking biogeographic differences between Bromeliaceae and Rapateaceae is the much larger range of bromeliads in the New World (fig. 1). Almost surely, this reflects the evolution of adaptations to xeric habitats and to epiphytism in Bromeliaceae and the almost total lack of such adaptations in Rapateaceae. Roughly two-thirds of all bromeliads are estimated to possess CAM photosynthesis (Crayn et al. 2000), which, together with associated succulence, provides a potent means of reducing transpiration and enduring intense drought, albeit at the cost of reduced photosynthetic capacity (Winter and Smith 1996). Most tillandsioids and bromeliads have leaf scales (trichomes) that serve as one-way ports that absorb water and nutrients and facilitate life as an epiphyte (McWilliams 1974; Benzing 1980, 2000). Rapateads do not appear to have evolved similar adaptations to drought or epiphytism (except in Epidryos, which occurs only in humid cloud forests in the Guayana Shield, Panama, and Ecuador) GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE and so have remained much more tightly corralled within the Guayana Shield and Amazon basin. One of the striking similarities in the biogeography of bromeliads and rapateads is the location of the areas occupied by both families in Africa (fig. 1). Maschalocephalus grows in savannas and forests on wet sand from Sierra Leone to Côte d’Ivoire; P. feliciana is endemic to sandstone outcrops on the tepuis-like Gangan and other table mountains of the Fouta Djalon massif in Guinea (Porembski and Barthlott 1999), to the northwest of Maschalocephalus’s range. Pitcairnia feliciana occurs in areas remote from the coast and areas of trade or, indeed, much human activity, so there is little doubt that it (and Maschalocephalus) arrived in Africa by nonhuman means (Porembski and Barthlott 1999). The narrow present-day ranges of the lone bromeliad and the lone rapatead in nearly the same portion of tropical west Africa might initially seem puzzling. However, bromeliads and rapateads should be favored by the moist, nutrient-poor conditions at these areas of similar underlying geology—split from the Guayana Shield and Precambrian Roraima sandstones by the rifting Atlantic—given that both families originated under such conditions (and, in the case of rapateads, mainly persist only under such conditions today). The region receives abundant rainfall under tropical conditions but with a 3–5-mo dry season. It is not the wettest lowland region in Africa; that distinction belongs to the disjunct area of heavy rainfall to the southeast along the equator around Cameroon. But tropical west Africa lies more than 1500 km closer to the eastern edge of the Guayana Shield (here taken as the eastern outposts of Monotremeae) than the Bight of Africa, which would have required dispersal over 39% more distance across the tropical Atlantic. Furthermore, on the basis of fossil pollen, the Guinean mountains apparently retained a wet climate during the Pleistocene, even though the savanna border in this region generally moved southward (Jahns et al. 1998; Dupont et al. 2000). The Fouta Djalon district is a center of endemism, with several species (like the treelet sedge, Microdraoides squammosus) also occurring disjunctly 1500 km to the west in Cameroon and Nigeria (Porembski and Barthlott 1999). Presumably, during periods of wetter climate, species such as these spread between both of these wet areas and then became isolated as conditions became drier. Other species may have been lost from one (or both) these refugia during arid periods. Geology, present-day climate, and past climate may thus, in concert, have helped bromeliads and rapateads survive in this refugium but killed them or prevented their S51 spread outside. Maschalocephalus appears to have unspecialized entomophilous flowers that might have permitted pollination by a variety of bees or other insects after arrival in Africa. Pitcairnia feliciana, on the other hand, appears specialized for avian pollination. While it left hummingbirds behind in South America, there are several species of nectivorous sunbirds in the area that may have substituted for them (Porembski and Barthlott 1999). Regrettably, pollination studies have not been conducted on either species to date. Finally, it may seem paradoxical to invoke wind dispersal from South America to Africa, either directly or via waterfowl, to account for the presence of both families in Africa, given that the average wind direction at this latitude is strongly in the opposite direction. But recent studies on the tropical Atlantic sea surface temperature dipole, a system somewhat analogous to El Niño in the tropical Pacific, indicate that positive feedback between winds and ocean temperatures can occasionally create west to east wind anomalies veering toward the Guinea coast (Servain 1991; Enfield et al. 1999; Jury et al. 2002). Storms under such conditions would be one candidate for the means by which Rapateaceae and Bromeliaceae (re)colonized Africa. Acknowledgments This study was supported by grants to T. J. Givnish and K. J. Sytsma from the National Science Foundation (NSF; DEB-9306943) and the National Geographic Society, NSF grant IBN-9904366 and a Vilas Research Associateship from the University of Wisconsin to T. J. Givnish, and NSF grant DEB-9981587 to P. E. Berry, T. J. Givnish, and K. J. Sytsma. A grant from the Smithsonian Institution to Vicki Funk helped support a helicopter expedition to several Venezuelan tepuis. Bruce Holst and Harry Luther of Selby Botanical Garden enthusiastically assisted in providing leaf material for several species included in this study. We thank Basil Stergios, Gustavo Romero, Charles Brewer, Fabian Michelangeli, and Armando Michelangeli for helping arrange permits and logistics in Venezuela. Vicki Funk, Carol Kelloff, Charles Ramdass, M. and M. Chan-a-sue, Harold Ameer, and M. Wilson provided critical logistical support in Guyana. Dr. L. Aké Assi graciously supplied material of Maschalocephalus from Côte d’Ivoire. Julio Betancour provided material of Brocchinia serrata from Cerro de Circasia in Colombia. James Smith and Angel Fernandez helped collect specimens on two expeditions to Venezuelan Amazonas. Literature Cited APG (Angiosperm Phylogeny Group) 2003 An update of the Angiosperm Phylogeny Group classification for the orders and families of flowering plants: APG II. Bot J Linn Soc 141:399–436. Baldwin BG 1997 Adaptive radiation of the Hawaiian silversword alliance: congruence and conflict of phylogenetic evidence from molecular and non-molecular investigations. Pages 103–128 in TJ Givnish, KJ Sytsma, eds. Molecular evolution and adaptive radiation. Cambridge University Press, New York. Baldwin BG, MJ Sanderson 1998 Age and rate of diversification of the Hawaiian silversword alliance (Compositae). Proc Natl Acad Sci USA 95:9402–9406. Ballard HE, KJ Sytsma, RR Kowal 1998 Shrinking the violets: phylogenetic relationships of infrageneric groups in Viola (Violaceae) based on internal transcribed spacer DNA sequences. Syst Bot 23:439–458. Baum DA, KJ Sytsma, PC Hoch 1994 The phylogeny of Epilobium L. (Onagraceae) based on nuclear ribosomal DNA sequences. Syst Bot 19:363–388. S52 INTERNATIONAL JOURNAL OF PLANT SCIENCES Benzing DH 1980 Biology of the bromeliads. Mad River, Eureka, Calif. ——— 1990 Vascular epiphytes: general biology and associated biota. Cambridge University Press, New York. ——— 2000 Bromeliaceae: profile of an adaptive radiation. Cambridge University Press, Cambridge. Benzing DH, TJ Givnish, DL Bermudez 1985 Absorptive trichomes in Brocchinia reducta (Bromeliaceae) and their evolutionary significance. Syst Bot 10:81–91. Berry PE, BK Holst, K Yatskievych 1995 Flora of the Venezuelan Guayana. Vol 1. Introduction. Timber, Portland, Oreg. Böhle UR, HH Hilger, WF Martin 1996 Island colonization and evolution of the insular woody habit in Echium L. (Boraginaceae). Proc Natl Acad Sci USA 93:11740–11745. Bremer K 2000 Early Cretaceous lineages of monocot flowering plants. Proc Natl Acad Sci USA 97:4707–4711. Briceño H, C Schubert, J Paolini 1990 Table mountain geology and surficial geochemistry: Chimanta Massif, Venezuelan Guayana Shield. J S Am Earth Sci 3:179–194. Briceño HO 1995 Origin of the Gran Sabana in southeastern Venezuela: no longer a ‘‘Lost World.’’ Sci Guaianae 5:147–174. Briceño HO, C Schubert 1990 Geomorphology of the Gran Sabana, Guayana Shield, southeastern Venezuela. Geomorphology 3: 125–141. Carlquist S 1966 Anatomy of Rapateaceae: roots and stems. Phytomorphology 16:17–38. Chase MW, DE Soltis, PS Soltis, PJ Rudall, MF Fay, WH Hahn, S Sullivan, et al 2000 Higher-level systematics of the monocotyledons: an assessment of current knowledge and a new classification. Pages 1–16 in KL Wilson, DA Morrison, eds. Monocots: systematics and evolution. CSIRO, Melbourne. Clark LG, W Zhang, JF Wendel 1995 A phylogeny of the grass family (Poaceae) based on ndhF sequence data. Syst Bot 20:436–460. Clark WD, MT Clegg 1990 Phylogenetic comparisons among rbcL sequences in the Bromeliaceae. Abstract. Am J Bot 71:115. Clark WD, BS Gaut, MR Duvall, MT Clegg 1993 Phylogenetic relationships of the Bromeliflorae-Commeliniflorae-Zingiberiflorae complex of monocots based on rbcL sequence comparisons. Ann Mo Bot Gard 80:987–998. Conti E, A Litt, KJ Sytsma 1996 Circumscription of Myrtales and their relationships to other rosids: evidence from rbcL sequence data. Am J Bot 83:221–233. Crayn DM, RG Terry, AC Smith, K Winter 2000 Molecular systematic investigations in Pitcairnioideae (Bromeliaceae) as a basis for understanding the evolution of crassulacean acid metabolism (CAM). Pages 569–579 in KL Wilson, DA Morrison, eds. Monocots: systematics and evolution. CSIRO, Melbourne. Doerr SH 1999 Karst-like landforms and hydrology in quartzites of the Venezuelan Guyana shield: pseudokarst or ‘‘real’’ karst? Z Geomorphol 43:1–17. Doyle JJ, JL Doyle 1987 A rapid DNA isolation procedure for small quantities of fresh leaf tissue. Phytochem Bull 19:11–15. Dupont LM, S Jahns, F Marret, S Ning 2000 Vegetation change in equatorial west Africa: time-slices for the last 150 ka. Palaeogeogr Palaeoclimatol Palaeoecol 155:95–122. Edmond JM, MR Palmer, CI Measures, B Grant, RF Stallard 1995 The fluvial geochemistry and denudation rate of the Guayana Shield in Venezuela, Colombia, and Brazil. Geochim Cosmochim Acta 59:3301–3325. Enfield DB, AM Mestas-Nuñez, DA Mayer, L Cid-Serrano 1999 How ubiquitous is the dipole relationship in tropical Atlantic sea surface temperatures? J Geophys Res 104:7841–7848. Felsenstein J 1985 Confidence limits on phylogenies: an approach using the bootstrap. Evolution 39:783–791. ——— 1988 Phylogenies from molecular sequences: inference and reliability. Annu Rev Genet 22:521–565. ——— 1994 PHYLIP (Phylogeny Inference Package), version 3.6. Department of Genetics, University of Washington, Seattle. Francisco-Ortega J, DJ Crawford, A Santos-Guerra, RK Jansen 1997 Origin and evolution of Argyranthemum (Asteraceae: Anthemideae) in Macaronesia. Pages 407–431 in TJ Givnish, KJ Sytsma, eds. Molecular evolution and adaptive radiation. Cambridge University Press, New York. Francisco-Ortega J, RK Jansen, A Santos-Guerra 1996 Chloroplast DNA evidence of colonization, adaptive radiation, and hybridization in the evolution of the Macaronesian flora. Proc Natl Acad Sci USA 93:4085–4090. Gaut BS, LG Clark, JF Wendel, SV Muse 1997 Comparisons of the molecular evolutionary process at rbcL and ndhF in the grass family (Poaceae). Mol Biol Evol 14:769–777. Gaut BS, SV Muse, WD Clark, MT Clegg 1993 Relative rates of nucleotide substitution at the rbcL locus of monocotyledenous plants. J Mol Evol 35:292–303. Ghosh SK 1985 Geology of the Roraima Group and its implications. Pages 33–50 in MI Muñoz, ed. Memoria I Symposium Amazónico. Ministerio de Energı́a y Minas, Caracas. Gibbs AK, CN Barron 1993 The geology of the Guiana Shield. Oxford monographs on geology and geophysics, no. 22. Oxford University Press, Oxford. Gilmartin AJ, GK Brown 1987 Bromeliales, related monocots, and resolution of relationships among Bromeliaceae subfamilies. Syst Bot 12:493–500. Givnish TJ 1998 Adaptive radiation of plants on oceanic islands: classical patterns, molecular data, new insights. Pages 281–304 in P Grant, ed. Evolution on islands. Oxford University Press, Oxford. Givnish TJ, EL Burkhardt, RE Happel, JW Weintraub. 1984 Carnivory in the bromeliad Brocchinia reducta, with a cost/benefit model for the general restriction of carnivorous plants to sunny, moist, nutrient-poor habitats. Am Nat 124:479–497. Givnish TJ, TM Evans, JC Pires, KJ Sytsma 1999 Polyphyly and convergent morphological evolution in Commelinales and Commelinidae: evidence from rbcL sequence data. Mol Phylogenet Evol 12: 360–385. Givnish TJ, TM Evans, JC Pires, ML Zjhra, TB Patterson, PE Berry, KJ Sytsma 2000 Molecular evolution, adaptive radiation, and geographic diversification in the amphiatlantic family Rapateaceae: evidence from ndhF sequences and morphology. Evolution 54:1915–1937. Givnish TJ, KC Millam, TM Evans, JC Pires, PE Berry, KJ Sytsma 2004a Phylogeny, biogeography, and ecological evolution in Bromeliaceae: insights from ndhF sequences. In JT Columbus, EA Friar, CW Hamilton, JM Porter, LM Prince, MG Simpson, eds. Monocots: comparative biology and evolution. Rancho Santa Botanic Garden, Claremont, Calif. (in press). Givnish TJ, JC Pires, SW Graham, MA McPherson, LM Prince, TB Patterson, HS Rai, et al 2004b Phylogeny of the monocots based on the highly informative cpDNA gene ndhF: evidence for widespread concerted convergence. In JT Columbus, EA Friar, CW Hamilton, JM Porter, LM Prince, MG Simpson, eds. Monocots: comparative biology and evolution. Rancho Santa Botanic Garden, Claremont, Calif. (in press). Givnish TJ, KJ Sytsma, JF Smith 1990 A re-examination of relationships among the three bromeliad subfamilies using cpDNA restriction site variation. Am J Bot 77:147–148. Givnish TJ, KJ Sytsma, JF Smith, WS Hahn 1992 Molecular evolution, phylogeny, and geography in the Pitcairnioideae (Bromeliaceae). Am J Bot 79:145. ——— 1994 Thorn-like prickles and heterophylly in Cyanea: adaptations to extinct avian browsers on Hawaii? Proc Natl Acad Sci USA 91:2810–2814. GIVNISH ET AL.—HISTORICAL BIOGEOGRAPHY OF RAPATEACEAE AND BROMELIACEAE ——— 1995 Molecular evolution, adaptive radiation, and geographic speciation in Cyanea (Campanulaceae, Lobelioideae). Pages 288–337 in WL Wagner, V Funk, ed. Hawaiian biogeography: evolution on a hot spot archipelago. Smithsonian Institution, Washington, D.C. Givnish TJ, KJ Sytsma, JF Smith, WJ Hahn, DH Benzing, EM Burkhardt 1997 Molecular evolution and adaptive radiation in Brocchinia (Bromeliaceae: Pitcairnioideae) atop tepuis of the Guayana Shield. Pages 259–311 in TJ Givnish, KJ Sytsma, eds. Molecular evolution and adaptive radiation. Cambridge University Press, New York. Goldblatt P 1993 Biological relationships between Africa and South America. Yale University Press, New Haven, Conn. Good R 1974 The geography of flowering plants. 4th ed. Longman, London. Graham SW, RG Olmstead, SCH Barrett 2002 Rooting phylogenetic trees with distant outgroups: a case study from the commelinoid monocots. Mol Biol Evol 19:1769–1781. Harms H, J Mildbraed 1938 Eine Bromelie aus dem tropischen Afrika. Notizbl Bot Gart Berlin 14:118–119. Herendeen PS, PR Crane 1995 The fossil history of the monocotyledons. Pages 1–21 in PJ Rudall, PJ Cribb, DF Cutler, CJ Humphries, eds. Monocotyledons: systematics and evolution. Royal Botanical Gardens, Kew. Holst BK 1997 Bromeliaceae. Pages 548–676 in PE Berry, BK Holst, K Yatskievych, eds. Flora of the Venezuelan Guayana. Vol 2. Missouri Botanical Garden, St. Louis. Hoorn C 1994 An environmental reconstruction of the palaeoAmazon River system (Middle–Late Miocene, NW Amazonia). Palaeogeogr Palaeoclimatol Palaeoecol 112:187–238. Hoorn C, J Guerrero, GA Sarmiento, MA Lorente 1995 Andean tectonics as a cause for changing drainage patterns in Miocene northern South America. Geology 23:237–240. Horres R, G Zizka, G Kahl, K Weising 2000 Molecular phylogenetics of Bromeliaceae: evidence from trnL(UAA) intron sequences of the chloroplast genome. Plant Biol 2:306–315. Huber O 1988 Guayana highlands vs. Guyana lowlands: a reappraisal. Taxon 37:595–614. Jahns S, M Huls, M Sarnthein 1998 Vegetation and climate history of west equatorial Africa based on a marine pollen record off Liberia (site GIK 16776) covering the last 400,000 years. Rev Palaeobot Palynol 102:277–288. Jury MR, DB Enfield, JL Melice 2002 Tropical monsoons around Africa: stability of El Nino–Southern Oscillation associations and links with continental climate. J Geophys Res 107:3151. Kessler M 2002a Environmental patterns and ecological correlates of range size among bromeliad communities of Andean forests in Bolivia. Bot Rev 68:100–127. Kessler M 2002b Species richness and ecophysiological types among Bolivian bromeliad communities. Biodivers Conserv 11: 987–1010. Kim S-C, DJ Crawford, J Francisco-Ortega, A Santos-Guerra 1996 A common origin for woody Sonchus and five related genera in the Macaronesian Islands: molecular evidence for extensive radiation. Proc Natl Acad Sci USA 93:7743–7748. ——— 1999 Adaptive radiation and genetic differentiation in the woody Sonchus alliance (Asteraceae: Sonchinae) in the Canary Islands. Plant Syst Evol 215:101–118. Lindqvist C, VA Albert 2002 Origin of the Hawaiian endemic mints within North American Stachys (Lamiaceae). Am J Bot 89: 1709–1724. Maddison DR 1991 The discovery and importance of multiple islands of most parsimonious trees. Syst Zool 40:315–328. Maddison DR, WP Maddison 2000 MacClade: analysis of phylogeny and character evolution, version 4.0 PPC. Sinauer, Sunderland, Mass. S53 Maddison WP, MJ Donoghue, DR Maddison 1984 Outgroup analysis and parsimony. Syst Zool 33:83–103. Maguire B 1958 Rapateaceae. Mem N Y Bot Gard 10:19–49. ——— 1979 Additions to the Rapateaceae. Acta Amazonica 9: 267–269. ——— 1982 Rapateaceae. Flora Venez 11:85–203. ——— 1984 Flora de la Guayana Venezolana. I. Nuevos taxa de la Guayana Venezolana. Acta Bot Venez 14:5–52. McWilliams EL 1974 Evolutionary ecology. Pages 40–55 in LB Smith, RJ Downs, eds. Pitcairnioideae (Bromeliaceae). Flora Neotrop 14:1–660. Mes THM, J van Brederode, H ’t Hart 1996 Origin of the woody Macaronesian Sempervivoideae and the phylogenetic position of the East African species of Aeonium. Bot Acta 109:477–491. Mez C 1904 Physiologische Bromeliaceen-Studien. I. Die WasserÖkonomie der extrem atmosphärischen Tillandsien. Jahrb Wiss Bot 40:157–229. Mort ME, DE Soltis, PS Soltis, J Francisco-Ortega, A Santos-Guerra 2002 Phylogenetics and evolution of the Macaronesian clade of Crassulaceae inferred from nuclear and chloroplast sequence data. Syst Bot 27:271–288. Olmstead RG, B Bremer, KM Scott, JD Palmer 1993 A parsimony analysis of the Asteridae sensu lato based on rbcL sequences. Ann Mo Bot Gard 80:700–722. Olmstead RG, HJ Michaels, KM Scott, JD Palmer 1992 Monophyly of the Asteridae and identification of their major lineages inferred from DNA sequences of rbcL. Ann Mo Bot Gard 79:249– 265. Pitman WC III, S Cande, J LaBrecque, J Pindell 1993 Fragmentation of Gondwana: the separation of Africa from South America. Pages 15–34 in P Goldblatt, ed. Biological relationships between Africa and South America. Yale University Press, New Haven, Conn. Pittendrigh CS 1948 The bromeliad-Anopheles-malaria complex in Trinidad. I. The bromeliad flora. Evolution 2:58–89. Porembski S, W Barthlott 1999 Pitcairnia feliciana: the only indigenous African bromeliad. Harv Pap Bot 4:175–184. Posada D, KA Crandall 1998 Modeltest: testing the model of DNA substitution. Bioinformatics 14:817–818. Potter PE 1997 The Mesozoic and Cenozoic paleodrainage of South America: a natural history. J S Am Earth Sci 10:331–344. Querouil S, E Verhey, M Dillen, M Colyn 2003 Patterns of diversification in two African forest shrews: Sylvisorex johnstoni and Sylvisorex ollula (Soricidae, Insectivora) in relation to paleoenvironmental changes. Mol Phylogenet Evol 28:24–37. Rambaut A, M Charleston 2002 TreeEdit: phylogenetic tree editor, version 1. http://evolve.zoo.ox.ac.uk/software/treeedit/main.html. Ranker TA, DE Soltis, PS Soltis, AJ Gilmartin 1990 Subfamilial relationships of the Bromeliaceae: evidence from chloroplast DNA restriction site variation. Syst Bot 15:425–434. Rasanen ME, AM Linna, JCR Santos, FR Negri 1995 Late Miocene tidal deposits in the Amazonian foreland basin. Science 269: 386–390. Sanderson MJ 1997 A nonparametric approach to estimating divergence times in the absence of rate constancy. Mol Biol Evol 14: 1218–1231. ——— 2002 Estimating absolute rates of molecular evolution and divergence times: a penalized likelihood approach. Mol Biol Evol 19:101–109. Sang T, DJ Crawford, S-C Kim, TF Stuessy 1994 Radiation of the endemic genus Dendroseris (Asteraceae) on the Juan Fernandez Islands: evidence from sequences of the ITS regions of nuclear ribosomal DNA. Am J Bot 81:1494–1501. Sang T, DJ Crawford, TF Stuessy, MO Silva 1995 ITS sequences and the phylogeny of the genus Robinsonia (Asteraceae). Syst Bot 20: 55–64. S54 INTERNATIONAL JOURNAL OF PLANT SCIENCES Schimper AFS 1888 Die epiphytische vegetation Amerikas. Botanische Mitteilungen aus den Tropen. Vol 2. Gustav Fischer, Jena. Servain J 1991 Simple climatic indexes for the tropical Atlantic Ocean and some applications. J Geophys Res 96:15137–15146. Sidder GB, SV Mendoza 1991 Geology of the Venezuelan Guayana Shield and its relation to the entire Guayana Shield. US Geol Surv Open File Rep 91-141. Smith JAC 1989 Epiphytic bromeliads. Pages 109–138 in U Lüttge, ed. Vascular plants as epiphytes: evolution and ecophysiology. Springer, Heidelberg. Smith JF, KJ Sytsma, JS Shoemaker, RL Smith 1991 A qualitative comparison of total cellular DNA extraction protocols. Phytochem Bull 23:2–9. Smith LB 1934 Geographic evidence on the lines of evolution in the Bromeliaceae. Bot Jahrb Okol Pflanzengesch Pflanzengeogr 66: 446–468. Smith LB, RJ Downs 1974 Pitcairnioideae (Bromeliaceae). Flora Neotrop 14:1–662. ——— 1977 Tillandsioideae (Bromeliaceae). Flora Neotrop 14: 663–1492. ——— 1979 Bromelioideae (Bromeliaceae). Flora Neotrop 14: 1493–2142. Soltis PS, DE Soltis, SG Weller, AK Sakai, WL Wagner 1996 Molecular phylogenetic analysis of the Hawaiian endemics Schiedea and Alsinidendron (Caryophyllaceae). Syst Bot 21:365–379. Spencer MA, LB Smith 1992 A revison of the genus Deuterocohnia Mez (Bromeliaceae: Pitcairnioideae). Bradea 6:141–146. Stallard RF, L Koehnken, MJ Johnsson 1991 Weathering processes and the composition of inorganic material transported through the Orinoco River system, Venezuela and Colombia. Geoderma 51: 133–165. Stevenson DW, JL Davis, JV Freudenstein, CR Hardy, MP Simmons, CD Specht 2000 A phylogenetic analysis of the monocotyledons based on morphological and molecular character sets, with com- ments on the placement of Acorus and Hydatellaceae. Pages 17–24 in KL Wilson, DA Morrison, eds. Monocots: systematics and evolution. CSIRO, Melbourne. Struwe L, JW Kadereit, J Klackenberg, S Nilsson, M Thiv, KB von Hagen, VA Albert 2002 Systematics, character evolution, and biogeography of Gentianaceae, including a new tribal and subtribal classification. Pages 21–309 in L Struwe, VA Albert, eds. Gentianaceae: systematics and natural history. Cambridge University Press, Cambridge. Swofford DL 2002 PAUP*: phylogenetic analysis using parsimony (*and other methods), version 4. Sinauer, Sunderland, Mass. Sytsma KJ, D Baum 1996 Molecular phylogenies and the diversification of the angiosperms. Pages 314–339 in DW Taylor, LJ Hickey, eds. Flowering plant origin, evolution and phylogeny. Chapman & Hall, New York. Sytsma KJ, LD Gottlieb 1986 Chloroplast DNA evidence for the origin of the genus Heterogaura from a species of Clarkia (Onagraceae). Proc Natl Acad Sci USA 83:5554–5557. Sytsma KJ, A Litt, ML Zjhra, JC Pires, M Nepokroeff, E Conti, J Walker, PG Wilson 2004 Clades, clocks, and continents: historical and biogeographical analysis of Myrtaceae, Vochysiaceae, and relatives in the Southern Hemisphere. Int J Plant Sci 165(suppl):S85–S105. Terry RG, GK Brown, RG Olmstead 1997a Examination of subfamilial phylogeny in Bromeliaceae using comparative sequencing of the plastid locus ndhF. Am J Bot 84:664–670. ——— 1997b Phylogenetic relationships in subfamily Tillandsioideae (Bromeliaceae) using ndhF sequences. Syst Bot 22:333–345. Tomlinson PB 1969 Anatomy of the monocotyledons. III. Commelinales-Zingiberales. Oxford University Press, London. Winter K, JAC Smith 1996 An introduction to crassulacean acid metabolism: biochemical principles and ecological diversity. Pages 1–3 in K Winter, JAC Smith, eds. Crassulacean acid metabolism: biochemistry, ecophysiology and evolution. Springer, Berlin.