Syst. Biol. 55(4):610-622, 2006 Copyright © Society of Systematic Biologists ISSN: 1063-5157 print / 1076-836X online DO1:10.1080/10635150600812619 Dating Dispersal and Radiation in the Gymnosperm Gnetum (Gnetales)—Clock Calibration When Outgroup Relationships Are Uncertain HYOSIG WON 1 ' 2 AND SUSANNE S. RENNER 1 ' 3 7 University of Missouri-St. Louis, Department of Biology, 8001 Natural Bridge Road, St. Louis, Missouri 63121, USA; and The Missouri Botanical Garden, St. Louis, Missouri 63166, USA; E-mail: wonhs@snu.ac.kr (H.W.); renner@umsl.edu (S.R.R.) 2 Present Address: Seoul National University, College of Natural Sciences, School of Biological Sciences, Seoul, Korea 151-742 3 Present Address: Department of Biology, Ludivig Maximilians University, D-80638 Munich, Germany; E-mail: renner@lrz.uni-muenchen.de Abstract.—Most implementations of molecular clocks require resolved topologies. However, one of the Bayesian relaxed clock approaches accepts input topologies that include polytomies. We explored the effects of resolved and polytomous input topologies in a rate-heterogeneous sequence data set for Gnetum, a member of the seed plant lineage Gnetales. Gnetum has 10 species in South America, 1 in tropical West Africa, and 20 to 25 in tropical Asia, and explanations for the ages of these disjunctions involve long-distance dispersal and/or the breakup of Gondwana. To resolve relationships within Gnetum, we sequenced most of its species for six loci from the chloroplast (rbcL, matK, and the trnT-trnF region), the nucleus (rITS/5.8S and the LEAFY gene second intron), and the mitochondrion (nadl gene second intron). Because Gnetum has no fossil record, we relied on fossils from other Gnetales and from the seed plant lineages conifers, Ginkgo, cycads, and angiosperms to constrain a molecular clock and obtain absolute times for within-Gnetum divergence events. Relationships among Gnetales and the other seed plant lineages are still unresolved, and we therefore used differently resolved topologies, including one that contained a basal polytomy among gymnosperms. For a small set of Gnetales exemplars (n = 13) in which rbcL and matK satisfied the clock assumption, we also obtained time estimates from a strict clock, calibrated with one outgroup fossil. The changing hierarchical relationships among seed plants (and accordingly changing placements of distant fossils) resulted in small changes of within-Gnetum estimates because topologically closest constraints overrode more distant constraints. Regardless of the seed plant topology assumed, relaxed clock estimates suggest that the extant clades of Gnetum began diverging from each other during the Upper Oligocene. Strict clock estimates imply a mid-Miocene divergence. These estimates, together with the phylogeny for Gnetum from the six combined data sets, imply that the single African species of Gnetum is not a remnant of a once Gondwanan distribution. Miocene and Pliocene range expansions are inferred for the Asian subclades of Gnetum, which stem from an ancestor that arrived from Africa. These findings fit with seed dispersal by water in several species of Gnetum, morphological similarities among apparently young species, and incomplete concerted evolution in the nuclear ITS region. [Bayesian relaxed clock; biogeography; Ephedra; Gnetum; long-distance water dispersal; polytomies in molecular clock dating.] "It is not at all clear—although often tacitly assumed—that an old lineage has occupied its current range for a long time. Gnetum in particular provides neat examples of the opposite." (Markgraf, 1929:429; translated from the German by SR) taxa. These conserved gene regions will often be uninformative in the group of interest. In the face of these difficulties, assuming a strict clock for a reduced (and then more or less clock-like) data set may still be the best Age estimation from molecular sequences has way to improve the precision, and perhaps accuracy, of emerged as a powerful tool for inferring the time it rate and date estimates (Sanderson, 2002; Ho et al., 2005). took a plant lineage to radiate in a particular area. An- For data sets that behave in a less clock-like fashion, the alytical methods now try to model change in substi- safest approach may be to compare results from strict tution rates along individual branches of a phylogeny clocks with those from relaxed clocks, employing alterby combining molecular data with multiple simultane- native input topologies and constrained with as many ous time constraints, usually from fossils (Thorne et al v critical fossils as possible. 1998; Rambaut and Bromham, 1998; Kishino et al, 2001; We here deal with the problems of rate heterogeneity Thorne and Kishino, 2002; Aris-Brosou and Yang, 2002; and unclear outgroup relationships in the seed plant linSanderson, 2002; Drummond and Rambaut, 2005). Simu- eage Gnetales, focusing on Gnetum, for which we have lations have shown that such "relaxed" clock approaches near-complete species sampling. Many seed plant lincan recover known rates (and hence ages) only as long as eages, especially conifers, Ginkgo, and cycads, have exseveral calibration points are accurate (Ho et al., 2005). cellent fossils that potentially can serve to constrain the In practice, uncertainties in calibrations due to misiden- timing of events within Gnetum. Gnetum itself has no tified or misdated fossils (Doyle and Donoghue, 1993; fossil record. Using constraints from outgroup fossils in Sanderson and Doyle, 2001; Near et al., 2005), taxon- this group, however, is problematic for at least two reasampling effects (Linder et al., Rutschmann, 2005), and sons. First, Gnetum is highly divergent from its closest asymmetric tree shape often reduce the reliability of the living relatives (below) and, second, the relationships of temporal constraints employed. Given the difficulty of the Gnetales to the other four lineages of seed plants, calibrating genetic distances, it is desirable to use multi- cycads, Ginkgo, conifers, and flowering plants, are still ple constraints wherever possible. Using multiple fossil unresolved. In what has been referred to as the sinconstraints, however, usually requires knowing phylo- gle most surprising grouping discovered with molecgenetic relationships outside the study group, with the ular data (Palmer et al., 2004), Gnetales appear to be corollary of having to use relatively conservative gene re- sister to Pinaceae, and both then sister to the remaingions that are alignable across phylogenetically distant ing conifers. This is the favored topology in the most 610 2006 WON AND RENNER—DATING GNETUM, CALIBRATION WHEN OUTGROUPS ARE UNCERTAIN 611 comprehensive analysis to date (Burleigh and Mathews, which, however has winged diaspores that superficially 2004). Other analyses place Gnetales as sister to conifers resemble those of Welwitschia. Similar fossils combining or to all other seed plants (Quandt et al, 2006). The Ephedra-like stems with winged diaspores have been deso-called anthophyte hypothesis, which sees Gnetales scribed as Gurvanella dictyoptera from the Early Cretaas sister to flowering plants, is rarely recovered with ceous Gurvan-Eren Formation of Mongolia (Krassilov, molecular data (e.g., Chaw et al., 2000; Bowe et al., 1982, 1997) and as Gurvanella exquisita, again from the 2000; Magall6n and Sanderson, 2002; Rydin et al., 2002; Yixian Formation (Sun et al., 2001). Further fossils asSchmidt and Schneider-Poetsch, 2002; Soltis et al., 2002; sociated with the Welwitschia lineage include Aptian Rai et al., 2003; see Burleigh and Mathews, 2004, for a Eastern North American male cones (Drewria; Crane and review). Upchurch, 1987) and a Barremian-Aptian female cone Obtaining estimates for divergence events within Gne- from the Lake Baikal area (Eoantha; Krassilov, 1986). The tum is of interest mainly because of the group's disjunct earliest Gnetales pollen is much older, from the Late geographic distribution, but also because of a lateral gene Triassic and Early Jurassic, and could represent either transfer event in one subclade for which the time to non- Ephedra or Welwitschia (Crane, 1996; Dilcher et al., 2005); functionalization (pseudogenization) can be estimated a recent study even describes Permian (270 my old) cones (Won and Renner, 2003). Gnetum consists of probably with such pollen (Wang, 2004). Gnetum pollen has a much fewer than 33 species; most are large woody climbers, lower fossilization potential because of its thin exine two are trees (Markgraf, 1929; HW, monograph in prepa- (Hesse et al., 2000). ration). Ten species occur in tropical South America, one These macro- and microfossils demonstrate the diin West Africa, and the remainder in tropical and sub- versity of Gnetales in the Mesozoic (Late Triassic to tropical Asia. Species relationships within Asian Gnetum Cretaceous) and provide an overall temporal context. cannot be resolved with the highly conserved gene re- However, their assignment to nodes in the tree of living gions required for seed plant-wide analyses and dating. species of Ephedra, Welwitschia, or Gnetum is problemInstead, we relied on a six-locus data set that includes atic. Because Welwitschia consists of a single species, the nuclear, mitochondria!, and chloroplast genes, spacers, above-mentioned fossil seedling, Cratonia cotyledon (beand introns. longing to the Welwitschia lineage), can only constrain The accepted hypothesis has long been that the the split between Gnetum and Welwitschia, possibly repantropical range of Gnetum reflects a Gondwanan his- sulting in a considerable underestimate of the true age tory (Markgraf, 1929). As explained above, Gnetum has of that split. Similarly, the best preserved Ephedra fossils, no fossils, but the Gondwanan hypothesis receives in- 125-My-old seeds (Rydin et al., 2004; Yang et al., 2005), direct support from the fossil record of Welwitschia, the cannot confidently be assigned to the crown group of sister lineage of Gnetum. Welwitschia mirabilis, today re- Ephedra, but do constrain the split between Ephedra and stricted to the Namib dessert, is known from a 110- the other two genera. (We nevertheless carried out an exMy-old seedling from the Lower Cretaceous (Aptian) periment to test the effect of assigning them to the Ephedra Crato Formation in northeastern Brazil (Rydin et al., crown group.) 2003). With its famously weird morphology that inTo overcome these calibration problems, we decided cludes a deep tap root and two strap-shaped leaves that to employ additional constraints from well-dated foscan grow for hundreds of years (Henschel and Seely, sils of conifers, cycads, and flowering plants. This 2000), Welwitschia differs remarkably from the canopy required a relaxed clock approach (i.e., modeling subclimbers that are typical of the genus Gnetum. Gnetum stitution rate change along branches) because the larger and Welwitschia together are the sister clade to Ephedra, data matrix rejected the clock assumption and added which consists of 35 to 45 species of shrubby plants with the problem of the uncertain phylogenetic placement photosynthetic stems and reduced leaves that are widely of Gnetales among seed plants (above). As a way out, distributed in temperate to arid areas of Eurasia, north- we employed the current best estimate seed plant tree ern Africa, southwestern North America, and western (Burleigh and Mathews, 2004), three alternative seed South America (Ickert-Bond and Wojciechowski, 2004; plant trees, and a topology with a basal polytomy in Huang et al., 2005). The monophyly of these three gen- the gymnosperms. The latter implies a prior belief that era, which make up the Gnetales, is supported by numer- Gnetales, conifers, cycads, and Ginkgo present a hard ous morphological and molecular phylogenetic analyses polytomy that cannot be resolved. In most current implementations of Bayesian clock approaches, prior beliefs (Kubitzki, 1990; Price, 1996). The Crato Formation has also yielded male strobili about the topology cannot be overwritten by the data similar to those of Welwitschia and vegetative stems with (they can in BEAST; Drummond and Rambaut, 2005), joints, whorled branchlets, reduced leaves, and sessile and workers have therefore generally preferred to use male strobili resembling those of Ephedra (Mohr et al., completely resolved topologies even for parts of the tree 2004; Dilcher et al., 2005, the last reviews the Welwitschia that are not supported by the underlying (or other) data fossil record). Early Cretaceous floras from Buarcos in (but see Renner and Zhang, 2004). Such a prior polytomy, Portugal, Virginia, USA and from the Yixian Formation however, does not prevent branch length information in northeast China contain additional fossils that prob- in the data from strongly affecting the posterior branch ably represent Ephedra (Rydin et al., 2004, 2006; Yang lengths assigned to branches above or below the polyet al., 2005), including Chaoyangia liangii (Duan, 1998), tomy. Where there is strong signal, substitutions will sort 612 SYSTEMATIC BIOLOGY themselves out along the branches emanating from the multifurcating node. The situation is complicated by the specific position of temporal constraints employed in a particular study. While in strict clock approaches the effect of a calibration point is identical throughout the tree, in relaxed clocks, the effect of age constraints—whether minimal, maximal, or fixed—diminishes with each node further away from the constraint(s). Given that the behavior of relaxed clocks is still so poorly understood, we felt it worthwhile to explore the empirical effects of a multifurcation in the input topology. Having determined the time horizon for the evolution of Gnetum with multiple approaches, we develop a hypothesis for the radiation of the climbing and tree forms of Gnetum that incorporates climatic, geological, and biological evidence and set out ways in which it might be tested. VOL. 55 Won 514 were cloned and provided by S.-M. Chaw (Institute of Botany, Academia Sinica, Taiwan). Gnetumspecific internal primers for matK were then designed, based on the sequences of G. gnemon, G. montanum, and G. woodsonianum. Except for these clones, the remaining matK sequences were obtained directly from PCR products. Sequence Analyses and Phylogenetic Analyses While rbcL sequences were constant in length (1428 base pairs [bp]), matK sequences showed large nucleotide composition and length variation. They were therefore translated into amino acid sequences to guide sequence alignment with Clustal X (Thompson et al., 1997). In addition, a combined six-marker data matrix (Appendix 2) was constructed by concatenating the data matrices of Won and Renner (2005a, 2005b: cp tRNALeu intron/spacer; nu rITS/5.8S; nu LEAFY second intron; and MATERIALS AND METHODS mt nadl second intron/partial exons) with the rbcL and Taxon Sampling matK seed plant matrices generated here. Final alignLeaf samples with voucher information and Gen- ments have been submitted to TreeBase (accession numBank accession numbers for the sequences obtained bers S1493, M2681, M2682). Length and G+C content of rbcL and matK were calcufrom them are listed in Appendices 1 and 2 (www. systematicbiology.org), which also give author names lated from aligned sequences, and sequence divergences for all taxa. Most Gnetum samples were collected in the were obtained under the K-2-P model (Kimura, 1980). field by the first author; a few are from botanical gar- The number of substitutions at synonymous sites (Ks) dens or herbarium material. Thirty-one accessions of and non-synonymous sites (Ka) was calculated using Gnetum were sequenced for six loci, usually all from MEGA2 (Kumar et al., 2001). Analyses included fullthe same DNA aliquot (for exceptions see Appendix 2). length matK sequences and base positions 77 to 1324 of The seed plant data set included representatives of all rbcL. major clades of Gnetum, Welwitschia mirabilis, Ginkgo The concatenated matrix of six loci used to resolve biloba, three species of Ephedra chosen to span the root intm-Gnetum relationships (Appendix 2) did not include of that genus (Ickert-Bond and Wojciechowski, 2004; outgroups because large length and nucleotide differRydin et al, 2004; Huang et al., 2005), and represen- ences in several of the spacers and introns prevent aligntatives of conifers and cycads chosen based on the re- ment across Gnetum, Welwitschia, and Ephedra. (Separate sults of Gugerli et al. (2001), Quinn et al. (2002), and Rai studies have analyzed the phylogenetic signal contained et al. (2003). Seed plant-wide analyses were rooted with in the secondary structure of these loci in Gnetales; Psilotum because the matK sequence of Marchantia (liver- Won and Renner, 2005a, 2005b.) Based on the results of wort) is much shorter than those of other vascular plants the seed plant-wide analysis, within-Gnetum trees were (1113 bp versus 1497-1662 bp in seed plants; it lacks the rooted on the South American Gnetum clade. Parsimony and maximum likelihood analyses were conserved domain X; see Hilu and Liang [1997]) and has a lower G+C content (18.1%; see Appendix 1), strongly conduced using PAUP 4.0bl0 (Swofford, 2002). Data matrices were analyzed separately under parsimony suggesting pseudogenization. optimization and in the absence of statistically supported conflict were then combined. Parsimony analyGene Sequencing ses employed heuristic searches that used 100 random DNA extraction, cloning, and sequencing followed the taxon-addition replicates, holding 100 trees at each methods described in Won and Renner (2005a, 2005b). step, tree bisection-reconnection (TBR) branch swapForward and reverse strands were sequenced. Figure ping, and the options MulTrees, Collapse, and SteepSI (www.systematicbiology.org) shows the approximate est Descent, without an upper limit for trees held in primer positions and organization of the six markers (cp memory. Nonparametric bootstrap support was obrbcL, matK, and the tRNALeu intron/spacer region; nu- tained by resampling the data 1000 times with the same clear (nu) rITS/5.8S and the second intron of the LEAFY search options and model, except using closest taxon gene; and mt nadl second intron and partial exons) addition. and Appendix 3 lists all primer sequences. To amplify Best-fitting substitution models for the combined rbcL, we used the rfrcL-1352R primer, because the com- matK/rbcL data sets, one comprising 13 taxa, the other 38 monly used combination of rbcL-724F and 1460R failed taxa, were found with ModelTest (Posada and Crandall, to work in Gnetum. The matK regions of G. woodsoni- 1998), applying the Akaike information criterion. The anum (= G. leyboldii var. woodsonianum) and G. gnemonbest-fit model was the general time-reversible model 2006 WON AND RENNER—DATING GNETUM, CALIBRATION WHEN OUTGROUPS ARE UNCERTAIN with a gamma-shape parameter and a proportion of sites modeled as invariable; i.e., the GTR+G+I model. The best-fit model for the 6-locus-31-taxa data set also was the GTR+G+I model. Bayesian analyses of the 6-locus-31-taxa data set relied on MrBayes 3.1 (Huelsenbeck and Ronquist, 2001) and were conducted under the GTR+G+I model, employing the flat prior beliefs that are the default in MrBayes and four Markov chain Monte Carlo (MCMC) chains, one of them heated (the default in MrBayes). Markov chains were run twice for 1 million generations, sampling every 100th generation for a total of 10,000 trees. The first 1000 trees were discarded as burn-in (data points sampled before the chain reaches stationarity), and the remaining 9000 samples were combined into a single file and analyzed using the "sumt" command in MrBayes. Both independent runs found essentially identical tree topologies and posterior probabilities, indicating that the sample number was sufficient to permit the algorithm to converge on a global solution. Molecular Clock Divergence Time Estimation For age estimation, most gapped positions were excluded from the matK data set, so that the gene matrix used for dating consisted of 2756 bp, 1428 bases of rbcL and 1328 bp of matK. Likelihood-ratio tests were used to assess rate heterogeneity in the concatenated rbcL + matK data sets of 38 seed plant exemplars and of 13 exemplars (each set including Psilotum for rooting purposes as explained under Taxon Sampling). When the 13-taxon data set justified the assumption of equal rates between sister groups, we used it to estimate divergence times from a strict clock by choosing the "enforce clock" option in PAUP. When the 38-taxon data rejected the assumption of equal rates in sister groups, we used a Bayesian approach that attempts to model variation in the rate of substitution along a tree. The software package used was that of Thorne and Kishino (2002; available at http://statgen.ncsu.edu/thorne/). Branch lengths were estimated with PAML's baseml component (ver. 3.14; Yang, 1997) under the F84+G model (with five rate categories), that being the only model implemented in Thome's estbranches software, which calculates a variance/covariance matrix of branch length estimates, given sequence data and a specified rooted topology that may include polytomies. The five rooted topologies we used as input are described below. Output branch lengths from estbranches became priors for MCMC searches in multidivtime that sought to find the most likely model of rate change (with rate change assumed to be log-normally distributed), given the branch lengths, topology, time constraints on nodes, and a Brownian motion parameter nu that controls the magnitude of autocorrelation per million years along the descending branches of the tree. Prior gamma distributions on parameters of the relaxed clock model were as follows: The mean of the prior distribution for the root age was set to 3 time units (i.e., 300 My, based on seed plant fossils; below); the standard deviation of this prior was also set 613 to 3. The mean and SD of the prior distribution for ingroup root rate were set to 0.0035 substitutions/site/My by dividing the median of the distances between the ingroup root and the tips by the time unit. The prior for nu was set to 0.4, following the manual's recommendation that the time units between root and tips multiplied by nu be about 1. The standard deviation on nu was also set to 0.4. Markov chains in multidivtime were run for 1 million generations, sampling every 100th generation for a total of 10,000 trees, with a burn-in of 1000 trees before the first sampling of the Markov chain. In the course of this study, 19 runs of 1 million generations were carried out on the matK/rbcL 38-taxon matrix, with consistent results that reflected the experimentally changed fossil constraints. Because of the unresolved placement of Gnetales among seed plants, we used the following alternative input topologies for the 38-taxon data set: (1) The preferred topology of Burleigh and Mathews (2004; our Fig. 1), which has Gnetales embedded in conifers as sister to Pinaceae and is referred to as the Gnepine hypothesis; (2) Gnetales as sister to conifers, the Gnetifer hypothesis (Fig. S1A; www.systematicbiology.org); (3) Gnetales as sister to all other seed plants, the Gnetales sister hypothesis (Fig. SIB); (4) Gnetales as sister to angiosperms, the Anthophyte hypothesis (Fig. SIC); and (5) a topology with a basal polytomy in the gymnosperms (Fig. SID). For absolute ages we relied on the geologic time scale of Gradstein et al. (2004) and the following calibrations (Fig. 1): The root node, i.e., the crown group of seed plants, was constrained to maximally 385 My old based on the oldest seed precursors (Gerrienne et al., 2004). This may be an overestimate because these early seeds, which have lobed integuments, probably reflecting their origin from fused sporangia or fertile telomes, are quite different from the seeds of modern seed plants. Modern seed types with completely fused integuments are not known until the late Early Carboniferous (ca. 325 My; J. Doyle, personal communication, July 2005). Crown group seed plants thus arose between 325 and 385 My a. Node 2, the gymnosperm crown group, was constrained to 315 Mya, based on the appearance of stem relatives of living conifers in the middle Late Carboniferous (ca. 310 Mya) and the appearance of Cordaitales, which are probably still more basal stem relatives of conifers, in the earliest Late Carboniferous (ca. 318 Mya). Node 3 was constrained to minimally 125 My old based on earliest Ephedra-type seeds (Barrett, 2000; Rydin et al., 2004; Yang et al., 2005); the minimal age of 125 My for the appearance of Ephedra is further supported by Ephedra-like macrofossil from the Brazilian Crato Formation (Mohr et al., 2004). Node 4, the split between Gnetum and Welwitschia, was constrained to 110 My based on the Cratonia cotyledon fossil (Rydin et al., 2003). Node 5, crown conifers, was constrained to minimally 225 My old based on the oldest Pinaceae-type cones (Miller, 1999; however, most genera of Pinaceae first appear during the Early Tertiary; LePage, 2003a). The split of Pinaceae from other conifers is bracketed not only by these Triassic cones, but also by 614 VOL. 55 SYSTEMATIC BIOLOGY Ephedra fragilis Ephedraceae E. trifurca E. sinica Welwitschia Gnetum afhcanum - Welwitschiaceae - Africa ~| G. hainanense G. montanum G. ula G. microcarpum G. cuspidatum G. gnemonoides G. gnemon G. nodiflorum G. woodsonianum CO Asia CO 8 3 *-> • 0 c O CD C Asia I O South America CO Abies Pseudotsuga E CD CL CO O Pinaceae Picea Pinus contorta P. banksiana c Phyllocladus Podocarpaceae Podocarpus Nageia Agathis Araucariaceae Araucaria co Ic CD o O Cunninghamia Cryptomeria Cupressaceae s.l. Cupressus Calocedrus Cephalotaxus Taxus Ginkgo Cycas Encephalartos Zamia Magnolia Nuphar Nymphaea I I 3.5 3.0 2.5 2.0 1.5 1.0 0.5 - Cephalotaxaceae Taxaceae Ginkgoaceae Cycadaceae Zamiaceae CO l_ CD Q. CO g J time (100 my) 0 FIGURE 1. Divergence events in Gnetum estimated from the chloroplast genes rbcL and matK analyzed under a Bayesian relaxed clock, constrained by fossil-based minimal ages at nodes 2 to 9 (see Materials and Methods) and assuming that Gnetales are nested in the conifers (the so-called Gnepine topology). For results with four other seed plant topologies see Figures SIA-D. All analyses were rooted with Psilotum because the GenBank matK sequence of Marchantia appears to be a pseudogene. Nodes of particular biogeographic interest are A, the split between the South American clade of Gnetum and the remainder of the genus; B, the split between African and Asian Gnetum; C, the split between the two main Asian clades of Gnetum; and D, the onset of radiation of the most species-rich Asian clade. records of Taxaceae (Palaeotaxus) from the Late Triassic and probable stem relatives of Podocarpaceae (Rissikia) from the (mid?) Triassic (J. Doyle, personal communication, July, 2005). Node 6 was constrained to minimally 160 My based on Middle Jurassic (160 to 175 My) Araucariaceae cones from Argentina (Stockey, 1982). Node 7 was constrained to minimally 90 My old based on Late Cretaceous (Turonian) Cupressaceae Thuja-like fossils (LePage, 2003b). The node just below could also have been constrained to the Late Cretaceous based on Cunninghamia-like cones (Farjon and Ortiz Garcia, 2003). Node 8 was constrained to minimally 90 My based on Turonian cones that clearly belong to crown Pinaceae (Gandolfo et al, 2001). Node 9, crown cycads, was constrained to minimally 270 My old based on Crossozamia from the Early Permian Lower Shihhotse Formation (Gao and Thomas, 1989), which appears to be nested among crown group cycads (Brenner et al., 2003). The 2006 WON AND RENNER—DATING GNETUM, CALIBRATION WHEN OUTGROUPS ARE UNCERTAIN 270-My age was chosen because the age of the top of the Lower Shihhotse Formation, though poorly constrained, is probably between from 275 and 265 My old (J. Hill, School of Geography, Earth and Environmental Sciences, University of Birmingham, personal communication, 9 Nov. 2005). Node 10 was constrained to minimally 125 My old, based on oldest Nymphaeaceae flowers (Friis et al., 2001). Gandolfo et al. (2004) do not accept these flowers as Nymphaeaceae, but other stem relatives or crown group members of crown angiosperms, such as Chloranthaceae, Winteraceae, monocots, and even eudicots also are known from 120- to 125-My-old fossils. To allow for the possibility that 125-My-old Ephedra seed fossils may represent living species (as implied by Rydin et al., 2004, and Yang et al., 2005), we used these fossils to constrain the Ephedra crown group to 125 My old; i.e., we moved the constraint at node 3 in Figure 1 up to the node where E. fragilis diverges from E. sinica 615 cular plants revealed no significant differences. Numbers of variable and parsimony-informative sites did not differ among codon positions (Appendix 5). The estimated number of nucleotide substitutions for Gnetum rbcL sequences was 0.058 (Ks) for synonymous sites and 0.031 (Ka) for nonsynonymous sites, on average. The averages and standard deviations of the observed Ka/Ks ratios were 0.057 ± 0.081 for rbcL and 0.524 ± 0.125 for matK. The Ka/Ks ratio in rbcL was thus about l/10th that in matK, mainly because of much reduced substitutions at nonsynonymous sites (Fig. S4). Over all taxa, matK sequences showed about 1.5 times the maximum sequence divergence seen in rbcL, and in both genes, Gnetum had about six times the maximum divergence observed in Ephedra (in rbcL 0.048 versus 0.008; in matK 0.068 versus 0.012). Phylogenetic Analyses and Divergence Time Estimation Statistics for sequence variation, parsimony tree Calibration for the strict clock (in the 13-taxon data length, and consistency indices for the six data partitions set) came from the Cratonia fossil, which was used to are given in Appendix 6. The rbcL gene did not contain fix the split between Gnetum and Welwitschia as 110 My. sufficient signal to reliably resolve relationships within This split may have occurred considerably earlier since Gnetum, but matK supported a sister-group relationship Cratonia is almost as apomorphic as Welwitschia (J. Doyle,between Asian and African Gnetum and subdivision of personal communication, 2005), and this was one of the Asian Gnetum into two clades (Fig. S5A). Bayesian analreasons why we resorted to calibration with fossils from ysis of the combined rbcL and matK data yielded a wellmore distant outgroups. resolved and supported topology (Fig. S5B) in which Gnetales were sister to Pinaceae (with .86 posterior probability [PP]). However, matK alone placed Gnetales as RESULTS sister to conifers (.96 PP), regardless of whether all posiCharacteristics of Gnetum rbcL and matK Sequences tions, only 1st and 2nd positions, or only 3rd positions We generated 47 rbcL sequences from Gnetum as well as were included (trees not shown). It also weakly supone sequence each from Welwitschia mirabilis and Ephedraported the monophyly of conifers (.80 PP). trifurca. None of the rbcL sequences contained inserA maximum likelihood tree obtained from the tions or deletions within the exon stretch (bp 1 to 1352). 6-locus-31-taxon data set shows well-resolved relationThe length of rbcL in Gnetales as in other seed plants ships within Gnetum (Fig. 2). Likelihood-ratio tests for is 1428 bp. For matK, we generated 12 new sequences this data set strongly rejected the clock assumption, and (Appendix 1). Because a GenBank matK sequence iden- it was not used for dating purposes. The clock assumptified as G. parvifolium (accession number AF280995) tion was also rejected by just the rbcL and matK sequences was indistinguishable from sequences of G. gnemon, it (individually or combined) as long as data matrices inwas excluded from further analyses. Two other matK se- cluded other seed plants besides Gnetales (P <£ 0.001). quences with single-base deletions were also excluded However, a matK/rbcL data set that included only 12 (Appendices 1 and 2). The length of matK is 1554 to 1557 species of Gnetales plus Psilotum satisfied the clock (P > bp for Gnetum, 1575 bp for Welwitschia, and 1662 bp for 0.05) and was analyzed under the assumption of a strict Ephedra (Appendix 1; also Huang et al., 2005). The length clock using the GTR+G+I model. Resulting genetic disvariation in Gnetum matK sequences was caused by a tances were calibrated by fixing the split between Gnetum three-base-pair insertion between positions 10 and 12 in and Welwitschia at 110 My based on the Cratonia fossil (see African G. africanum sequences. The G+C contents of rbcL Materials and Methods). Because all larger rbcL/matK matrices rejected a strict sequences ranged from 41.8% to 45.2% and those of matK sequences from 31.1% to 33.1%, fitting well with G+C clock, we applied a Bayesian relaxed clock, constrained contents of other vascular plants (40.8 to 45.8% for rbcL with the fossil-based minimal ages specified in Materials and Methods and shown in Figure 1; the root node was and 27.1% to 37.7% for matK; Appendix 1). Sequence divergences of rbcL and matK sequences constrained by a maximal age (Materials and Methods). within and among gnetalean genera are shown in Figure S6 shows a GTR+G+I ML phylogram from these Appendix 4 and neighbor-joining trees from these data data to provide a sense of branch length heterogeneity. are shown as Figure S3. Figure S4 plots the numbers of Table 1 lists the divergence times obtained for key nucleotide substitution at synonymous (Ks) and nonsyn- nodes within Gnetum (labeled A, B, C, and D in all figures) onymous sites (Ka) in rbcL and matK across seed plants. under the Gnepine, Gnetifer, Gnetales-sister, AnthoPairwise comparisons between nongnetalean vascular phyte, and Polytomy input topologies shown in Figures 1 plants and comparisons between Gnetales and other vas- and S1A-D. The different hierarchical relationships of and E. trifurca. 616 VOL. 55 SYSTEMATIC BIOLOGY i- G. acutum -I 97 r-- G. macrostachyum -1.97 — G. tenuifolium CD T3 i2 o I I I - G. cuspidatum -/.97 r G. diminutum 83/1.00 M 100/1.00 I— G. microcarpum -A98 — G. klossii o CD — G. ula G. latifolium Won 524 85/75 100/1.00 G. latifolium Won 575 98/1.00 0.01 substitutions/site 100/1.00 G. aff. latifolium Won 545 73A99 G. aff. latifolium SAN 151116 75/1.00 — G. neglectum 71/1.00 Asia II >2 *= I? o -2 CD G. sp. nov. Takeuchi et al. 7049. G. hainanense s.l. RBGE 100/1.00 G. parvifolium -/1.00 i— G. hainanense s.l. Won 600 M-/1.00 L G. sp. Harder et al. 5621 c 100/1.00 100/1.00, !- G. hainanense s.l. Won 580 100/1.00 r— G. gnemon RBGE 100/1.00 r" — G. gnemon Won 514 G. costatum 100/1.00 Asia I G. gnemonoides 100/1.00 G. raya ] Africa G. africanum 100/1.00 r G. microstachyum I— G. urens 82/1.00 - G. schwackeanum South America 81/.99 G. woodsonianum 100/1.00 G. nodiflorum G. paniculatum 41-11 I 47-13 39-10 22-4 mya T 200 100 0 Sea Level (m) 30 Oligocene 20 10 i i Miocene 0 -i mya Pliocene Pleistocene FIGURE 2. Maximum likelihood tree for Gnetum obtained from combined nuclear, chloroplast, and mitochondrial data, analyzed under the GTR+G+I model. Numbers above branches indicate bootstrap supports from parsimony analyses, followed by Bayesian posterior probabilities. Dashes indicate support values lower than 50%. The ages shown for nodes A to D (with 95% confidence intervals) are from the Bayesian relaxed clock analysis of matK and rbcL sequences (Fig. 1; see Table 1 for point age estimates and strict clock estimates). Geological periods and eustatic sea level changes are shown in the inset below. The arrow indicates the separation between and G. schwackeanum and G. woodsonianum (= G. leyboldii var. woodsonianum). The depiction of sea level changes is modified from Haq et al. (1987) and illustrates the dramatic fluctuations from the Miocene to the Pleistocene that correlate with most of the speciation in the Asia II clade. 2006 617 WON AND RENNER—DATING GNETUM, CALIBRATION WHEN OUTGROUPS ARE UNCERTAIN TABLE 1. Time estimates (in million years, followed by 95% confidence intervals) for nodes A, B, C, D in Figures 1 and 2, and in Figures SI A-D, obtained from combined matK and rbcL sequences under a strict clock (row 1), calibrated with one fossil (Fig. 1, node 4), or a Bayesian relaxed clock (rows 2 to 7), constrained with 9 fossil-based minimal ages (Fig. 1 and Materials and Methods). Row 7 shows results obtained when 125-My-old Ephedra seeds were used as a minimal constraint for crown group Ephedra. Alternative input topologies are shown in Figure 1 and Figures S1A-D. Model Strict clock, 13-taxon matK/rbcl data set Relaxed clock, 38-taxon matK/rbcL data Constrained as in Figure 1 under the Gnepine topology Constrained as in Figure S1A under the Gnetifer topology Constrained as in Figure SIB under the Gnetales-sister topology Constrained as in Figure SIC under the Anthophyte topology Constrained as in Figure SID with basal 4-tomy in gymnosperms Constrained as in Figure 1 under the Gnepine topology, but Ephedra crown min. age set to 125 My Node A (split: South Node B (split: Node C (split: Asia I Node D (split: basal American vs. remaining African vs. Asian and II clades of divergence within Gnetum) Gnetum) Gnetum) Asia II clade of Gnetum) 14 12-11 11 26 (13, 47) 29 (14, 51) 38 (19, 66) 22 (11,41) 25 (11,44) 33 (15,58) 21 (10, 39) 24 (11,43) 31 (15, 56) 11 (4, 22) 12 (5, 25) 17 (7, 34) 37 (18, 64) 32 (15,58) 30 (14,55) 16 (6, 33) 30 (14,53) 26 (12,47) 24 (11,45) 13 (5, 26) 44 (23, 71) 38 (19, 64) 36 (18, 62) 20 (8, 39) seed plants, in particular whether conifers are mono- closely resembled estimates obtained under the Gnetifer phyletic or paraphyletic and whether Gnetales are sister input tree. The likely reason is that polytomies in dating to all other seed plants or embedded in conifers, imply trees might impact posteriors branch lengths only where changes in the placements of two minimal constraints, there is branch length uncertainty throughout the tree namely the oldest gymnosperm (node 2) and the oldest (J. Thorne, personal communication, Oct. 2005), whereas conifer (node 5). This changed within-Gnefwm time esti- there is strong branch length information in many parts mates by more or less 10 My (Table 1). For example, under of the seed plant tree, as evident in a phylogram of the unthe Gnepine seed plant topology, Gnetum first diverged at derlying data (Fig. S6). Another caveat concerning cur26 My, under the Gnetifer topology it diverged at 29 My, rent Bayesian relaxed clock implementations is that the under the Gnetales-sister hypothesis at 38 My, and under impact of the other priors used, namely the root rate and the Anthophyte hypothesis at 37 My. Estimates under the the Brownian motion parameter (Materials and MethPolytomy input tree were closest to the estimates under ods), is poorly understood. What is clear from our results is that the constraint(s) the Gnetifer topology (Table 1). The experimental reassignment of the 125-My-old Ephedra seeds to the Ephedra closest to the nodes being estimated exert(s) the strongest crown instead of the stem (node 3 in Fig. 1), resulted in impact (also Wiegmann et al, 2003) and that clear signal in molecular branch lengths can push nodes far from almost doubled within-Gnetum ages (Table 1). Estimates for major divergence events in Gnetum from their own minimal constraints. For example, node 7, the 13-taxon strict clock analysis were younger by about crown Cupressaceae, and node 8, crown Pinaceae, are half compared to those from the 38-taxon relaxed clock both constrained by fossils to minimally 90 My old, but analysis (Table 1). For example, the time estimated for the inspection of Figures 1 and S1A-D shows that based on divergence of South American Gnetum from the remain- genetic distances and irrespective of the seed plant topolder of the genus under a relaxed clock and the Gnepine ogy employed, crown Pinaceae are considerably older topology (Fig. 1) was 26 My (13 to 47 my, 95% confidence than crown Cupressaceae. In terms of their likelihood scores, the unconinterval), whereas under a strict clock, the same event strained input topologies rank as follows: Gnepine was dated to 14 My. seed plant topology (-In = 27,964.2703), Gnetifer topology (-In = 27,964.9216), Gnetales sister topolDISCUSSION ogy (-In = 27,971.7141), and, worst, the Anthophyte Effects of Tree Topologies, Constraints from Distant topology (-In = 27,976.5747). The most compreOutgroups, and Strict versus Relaxed Clock Inference hensive seed plant analysis to date (Burleigh and A polytomy in the input tree for a Bayesian analysis Mathews, 2004) found that combined data from implies a prior belief that speciation events occurred at 13 loci (for 31 exemplar taxa) favored the Gnetales about the same time so that it cannot be estimated from sister topology, whereas slower evolving data parsubstitutions in what order they occurred. This is hardly titions favored the Gnepine tree. In the following a well-founded hypothesis for the evolution of the four section we concentrate on the estimates obtained under lineages of gymnospermous seed plants that make up the the Gnepine topology (Table 1, row 2). However, the particular polytomy relevant to the present study. Em- within-Gnefwm ages obtained under the alternative seed pirically, however, this unconvincing prior hypothesis plant topologies, including the polytomy tree, differ by had little impact on the posterior distribution of within- only 4 to 12 My, and our inferences concerning Gnetum Gnetum estimates, which under the polytomy input tree therefore would change little if one of the other seed 618 SYSTEMATIC BIOLOGY plant trees turned out to be true. Our conclusions further incorporate strict clock estimates, which do not depend on seed plants outside Gnetales. The strict clock (for 13 Gnetales) yielded an estimate of 14 My for the onset of diversification within Gnetum, about half that from the 38-taxon relaxed clock (26 My; 95% confidence range 13-47 My). This is probably due to the well-known taxon density effect in molecular dating (e.g., Linder et al., 2005); more densely sampled clades generally yield older ages. An earlier strict clock estimate for the onset of diversification of extant Gnetum was 11 to 6 My (Won and Renner, 2003); the difference between our earlier analysis and the present is that we here included two additional species of Ephedra and deleted a stretch of about 400 bp from the matK data because of gapped Ephedra sequences. Biogeography and Evolution of Gnetum The major divergence events among extant clades of Gnetum are estimated as dating to the Upper Oligocene, Miocene, and Pliocene (Table 1). Under the strict clock, the main South American, African, and Asian divergence events occurred sometime during the Miocene and Pliocene, whereas the Bayesian relaxed clock implies divergence during the Upper Oligocene and Miocene. These estimates in no way contradict the Mesozoic fossil record of Gnetales (Introduction), which includes over 100-My-old relatives of extant species of Ephedra and Welwitschia (e.g., Mohr and Friis, 2000; Yang, 2002; Rydin et al., 2003, 2004, 2006; Mohr et al., 2004; Wang, 2004; Yang et al., 2005; Dilcher et al., 2005). Rather, the molecular estimates concern the recent-most common ancestor of living clades, while the fossils may represent members of stem lineages or extinct sister species. Gnetales were diverse and widespread during the Cretaceous, and the almost worldwide occurrence of their fossils compared to their present range indicates that they have suffered major extinctions. A new insight from this study is that Gnetum has undergone a geologically recent radiation, especially in the Malesian region and that the genus's present disjunct range is not Gondwanan. It is possible that the Asian radiation coincided with times of low sea levels because of the opportunities for overland seed dispersal they would have afforded (Fig. 2). Temporally, although not geographically, our findings for Gnetum parallel those for Ephedra (Huang and Price, 2003; Ickert-Bond and Wojciechowski, 2004). In the case of Ephedra, the evolution of Mediterranean climates, the rise of the Rockies and the Andes, and seed dispersal by birds and rodents (e.g., Ridley, 1930; Holmgren et al., 2003) all seem to have contributed to relatively recent radiations (as well as contributing to species formation per se). Biogeographic analyses of other plant families with ancient and relatively good fossil records, such as Calycanthaceae, Chloranthaceae, Nothofagus, and Nymphaeaceae, have obtained similar results, namely that in spite of their impressive fossil records, these clades apparently reached some parts of their range quite recently (Zhang and Renner, 2003: Chloranthaceae; VOL. 55 Knapp et al., 2005: Nothofagus; Zhou et al., 2006: Calycanthaceae; Yoo et al., 2006: Nymphaeaceae). In other words, old lineages do not necessarily stop dispersing and diversifying, as pointed out by Markgraf (1929), the last monographer of Gnetum, in the quote at the beginning of this paper. Similarly, Ephedra seems to have diversified in its current habitat during Miocene times (Huang and Price, 2003; Ickert-Bond and Wojciechowski, 2004), in spite of morphological stasis in some morphological traits, such as the seed coat in fossil E. archaeorhytidosperma and living E. rhytidosperma (Yang et al., 2005). Such morphological stasis does not imply that the respective fossil represents a member of continuous ancestor-descendent chains of populations without intervening speciation events. When we nevertheless used the 125-My-old Ephedra seed fossils to constrain the minimal age of living species of Ephedra (the experiment reported in Table 1, row 7), divergences between South American, African, and Asian Gnetum were still estimated as dating to the Eocene and Oligocene, rather than being Gondwanan. Gnetum currently comprises ten species in South America, one in West Africa, and ~25 in tropical and subtropical Asia. If an Oligocene/Miocene age is accepted for the extant lineages in South America and Asia (Fig. 2), then Gnetum must have reached its disjunct pantropical range either through transoceanic dispersal or through a Laurasian expansion followed by southwards spread. If one considers the upper error bracket of ± 50 My (under any of the different seed plant topologies; Table 1), rather than the point estimates, Gnetum would be old enough to have dispersed across the North Atlantic land bridge to Africa and Asia. An Eocene boreotropical range, with subsequent fragmentation during the Oligocene climate cooling, is difficult to reject with available data, but does not fit well with the occurrence of it sister group, Welwitschia, in Africa and Brazil (living and as fossils, respectively). In terms of dispersal biology, water dispersal in seawater is a strong possibility. Some South American species, for example, G. venosum, have a special middle layer in the seed coat that gives buoyancy (Kubitzki, 1985), and others, such as G. gnemonoides, have large, corky diaspores (Markgraf, 1951). The smallest diaspores in the genus, those of G. africanum, measure 1.2-1.5 cm x 0.8 cm; the largest Gnetum seeds measure 7 cm x 3-4 cm. In most species, the mature seed envelope turns red or yellow, and seeds are then attractive to large birds, such as toucans, but also to rodents and monkeys (e.g., Ridley, 1930; Markgraf, 1951; Kubitzki, 1985; Van Roosmalen, 1985; Forget at al., 2002). Fish eat the seeds when they fall into streams, and Amazon catfish are known to regurgitate them because of sclerenchyma needles in the seed envelope (Goulding, 1980; Kubitzki, 1985). Fish dispersal therefore does not destroy the embryo. Gnetum species commonly grow along rivers, and there are also observations of Gnetum seeds from sea drift; for example, from beaches in Malaysia (Ridley, 1930; Hemsley in Markgraf, 1951). Several months are required for embryo maturation (Maheshwari and Vasil, 1961). 2006 WON AND RENNER—DATING GNETUM, CALIBRATION WHEN OUTGROUPS ARE UNCERTAIN A review of molecular clock-dated instances of diaspore dispersal and how they relate to the direction of surface currents in the Atlantic (Renner, 2004) showed that there are several cases of likely water dispersal between South America and Africa. Dispersal between Africa, Madagascar, and India/Malaysia appears even more common (Thome, 1973; additional references in Renner, 2005) and is more readily explained than that between South America and Africa because distances are shorter and there are regular monsoon wind systems as well as sea currents favorable for the transport of small objects between Africa, the Seychelles, the Comores, the Chagos archipelago (about half way between Africa and Indonesia), and India. Recently discovered examples of transmarine dispersal between Madagascar, Africa, and the Seychelles include multiple lineages of frogs, chameleons, rodents, Carnivora, and lemurs (DeQueiroz, 2004; Renner, 2004). With the likely time horizon for the evolution of Gnetum being the Upper Oligocene, Miocene, and Pliocene (Table 1), what were the main forces behind the radiation of the genus in South American and Southeast Asia? Entry into new kinds of habits appears to have played a limited role; the climbing species of Gnetum as well as the two tree species occur in the understory of upland forests and in riparian vegetation. The tree habit appears to have evolved once, in the ancestor of G. costatum and G. gnemon, rather than being an ancestral state as previously assumed (Markgraf, 1929; Fig. 2). Besides growth form, there are few other morphological differences that would readily suggest different habitat niches; instead, species differences in Gnetum concern details of strobilus branching, and seed and leaf size. Apparently, geographic isolation played the main role in species formation in Gnetum. Thus, in South America, the uplift of the northern Andes led to at least one species 619 (Fig. 2) are G. gnemon with a wide distribution and its sister species G. costatum, endemic in New Guinea; G. gnemonoides with a wide distribution and its sister G. mya, endemic to Borneo; and G. latifolium s.L with a wide distribution and its sister clade G. neglectum, endemic to Borneo, and G. sp. nov. Takeuchi et al. 7049, endemic in New Guinea. The Gnetum cuspidatum clade includes further pairs of wide-ranging species and narrow endemics (Fig. 2: G. klossii endemic to Borneo; G. tenuifolium endemic to the Malay Peninsula; G. microcarpum, G. diminutum, and G. acutum endemic to the Malay Peninsula and Borneo, G. macrostachyum ranging all the way from southern Indochina through the Malay Peninsula to Sumatra). These overlaid patterns of wideranging species related to local endemics suggest waves of expansion, local isolation or extinction, and reexpansion. Fluctuation in sea level (Haq et al., 1987; our Fig. 2) may have facilitated the dispersal of Gnetum over to the Malesian islands, while at other times acting as barriers to dispersal. To date, only one case of introgression has been detected among species of Gnetum: analyses of nuclear ribosomal internal transcribed spacer (ITS) sequences and of the second intron of the nuclear LEAFY gene suggest that G. klossii, a member of the G. cuspidatum clade in terms of its ITS sequences (as well as its chloroplast sequences), falls in the G. latifolium s.L clade in terms of its LEAFY second intron sequences (Won and Renner, 2005a). Dense sampling of population-level genetic variation in several of the suspected young species pairs would allow testing our scenario of recent waves of expansion and extinction in Indonesian and Malesian clades of Gnetum. A complicating aspect deserving further study in this context is the impact humans may have had on the distribution of Gnetum; G. gnemon is commonly cultivated for food in Indonesia and New Guinea, where the migrapair: Gnetum woodsonianum (= G. leyboldii var. woodsoni- tions of human populations have greatly influenced the anum) occurs from Costa Rica to Colombia and is sis- distribution patterns of important tropical crop plants ter to G. schwackeanum from Venezuela and the Guianas (Barrau, 1963). Currently, G. gnemon has the widest disto northern Brazil (Stevenson and Zanoni, 1991; Fig. 2). tribution of all Asian Gnetum, even having reached Fiji, The Colombian Andes were formed by rapid uplift of either via over water dispersal or with the help of Homo the Eastern Cordillera between 5 and 2 Mya (Gregory- sapiens sapiens. Wodzicki, 2000), and Central and South America were not connected until about 3 Mya (Haug and Tiedemann, 1998). The ancestor of G. woodsonianum likely became ACKNOWLEDGMENTS separated by the uplift of the Colombian Andes and then We thank Jim Doyle, an anonymous reviewer, and the associate expanded its distribution across the Panamanian land bridge into Costa Rica. The remaining eight South Amer- editor, Peter Linder, for constructive criticism; Y.-C. Chan, N. V. Dzu, H. A. Hussein, Z. bin Mat Said, J. F. Maxwell, M. Postar, R. Wang, ican species are distributed east of the Andes, mainly in Gang, and K.-M. Wong for help with fieldwork; L. Gillespie, D. K. Harder, the Amazonian lowland below 600 m altitude; only G. M. F. Prevost, S. Shi, P. Suksathan, W. Takeuchi, and the botanical garcamponim occurs in the Guiana highland at up to 1800 m dens of Bogor, Mainz, New York, Edinburgh, and the University of California-Berkeley for plant material; S.-M. Chaw for Gnetum matK altitude. S. Ickert-Bond, C. Rydin, and J. Huang for Ephedra aliquots The Asian subclades of Gnetum are between 4 and cloning; and pre-publication sequences; B. Mohr and Y. Yang for information 22 My old (Table 1, Fig. 2) and result from multiple cycles about Welwitschia and Ephedra fossils; and M. Krings, J. Hill, and E. of colonization, judging from the overlapping distribu- Hermsen for discussion of seed plant fossils. This paper represents tions of the largest subclades, G. hainanense s.L, G. lati- part of a dissertation submitted by the first author in partial fulfillfolium s.L, and G. cuspidatum. All three subclades range ment of the requirements for a Ph.D. from University of Missouri-St. The American Society of Plant Taxonomy, the International Asfrom the Malay Peninsula to New Guinea across Bor- Louis. sociation of Plant Taxonomy, and the International Center for Tropical neo, Sulawesi, and the Philippines. Each subclade con- Ecology at the University of Missouri-St. Louis supported our work tains widely distributed and endemic species. Such pairs financially. 620 SYSTEMATIC BIOLOGY VOL. 55 entrapment pollination mechanisms in early angiosperms. Proc. Natl. Acad. Sci. USA. 101:8056-8060. Aris-Brosou, S., and Z. Yang. 2002. Effects of models of rate evolution on Gao, Z., and B. A. Thomas. 1989. A review of cycad megasporophylls, with new evidence of Crossozamia Pomel and its associated leaves estimation of divergence dates with special reference to the metazoan from the Lower Permian of Taiyuan, China. Rev. Palaeobot. Palynol. 18S rRNA phylogeny. Syst. Biol. 51:703-714. 60:205-233. Barrau, J. 1963. Plants and the migration of Pacific peoples. Bishop Germano, J., and A. S. Klein. 1999. Species-specific nuclear and chloroMuseum Press, Honolulu, Hawaii. plast single nucleotide polymorphisms to distinguish Picea glauca, P. Barrett, P. M. 2000. Evolutionary consequences of dating the Yixian mariana and P. rubens. Theor. Appl. Genet. 99:37-49. Formation. Trends Ecol. Evol. 15:99-103. Bowe, L. M., G. Coat, and C. W. dePamphilis. 2000. Phylogeny Gerrienne, P., B. Meyer-Berthaud, M. Fairon-Demaret,M. Streel, and P. Steemans 2004. Runcaria, a Middle Devonian seed plant precursor. of seed plants based on all three genomic compartments: extant Science 306:856-858 gymnosperms are monophyletic and Gnetales' closest relatives are Goulding, M. 1980. The fishes and the forest. Berkeley, University of conifers. Proc. Natl. Acad. Sci. USA. 97:4092-4097. California Press. Brenner, E. D., D. W. Stevenson, and R. W. Twigg. 2003. Cycads: evolutionary innovations and the role of plant-derived neurotoxins. Gradstein, F. M., J. G. Ogg, and A. G. Smith (eds.) 2004. A geologic time scale 2004. Cambridge University Press, Cambridge, UK. Trends Plant Sci. 8:446-452. Brunsfeld, S. J., P. S. Soltis, D. E. Soltis, P. A. Gadek, C. J. Quinn, D. D. Gregory-Wodzicki, K. M. 2000. Uplift history of Central and Northern Andes: a review. GSA Bulletin 112:1091-1105. Strenge, and T. A. Ranker. 1994. Phylogenetic relationships among the genera of Taxodiaceae and Cupressaceae: evidence from rbcL Gugerli, F., C. Sperisen, U. Biichler, I. Brunner, S. Brodbeck, J. D. Palmer, and Y.-L. Qiu. 2001. The evolutionary split of Pinaceae from other sequences. Syst. Bot. 19:253-262. conifers: evidence from an intron loss and a multigene phylogeny. Burleigh, J. G., and S. Ma thews. 2004. Phylogenetic signal in nucleotide Mol. Phylogenet. Evol. 21:167-175. data from seed plants: implications for resolving the seed plant tree Haq, B. U., J. Hardenbol, and P. R. Vail. 1987. Chronology of fluctuating of life. Am J. Bot. 91:1599-1613. Chaw, S.-M., C. L. Parkinson, Y. Cheng, T. M. Vincent, and J. D. Palmer. sea levels since the Triassic. Science 235:1156-1167. 2000. Seed plant phylogeny inferred from all three plant genomes: Hasebe, M., R. Kofuji, M. Ito, M. Kato, K. Iwatsuki, and K. Ueda. 1992. Phylogeny of gymnosperms inferred from rbcL gene sequences. Bot. monophyly of extant gymnosperms and origin of Gnetales from Mag. Tokyo 105:673-679. conifers. Proc. Natl. Acad. Sci. USA. 97:4086-4091. Cheng, Y, R. G. Nicolson, K. Tripp, and S.-M. Chaw. 2000. Phylogeny Hasebe, M., P. G. Wolf, K. M. Pryer, K. Ueda, M. Ito, R. Sano, G. J. Gastony, J. Yokoyama, J. R. Manhart, N. Murakami, E. H. Crane, C. of Taxaceae and Cephalotaxaceae genera inferred from chloroplast H. Haufler, and W. D. Hauk. 1995. A global analysis of fern phylogeny matK gene and nuclear rDNA ITS region. Mol. Phylogenet. Evol. based on rbcL nucleotide sequences. Am. Fern J. 35:134-181. 14:353-365. Conran, J. G., G. A. Wood, P. G. Martin, J. M. Dowd, C. J. Quinn, Haug, G. H., and R. Tiedemann. 1998. Effect of the formation of the Isthmus of Panama on Atlantic Ocean thermohaline circulation. Nature P. A. Gadek, and R. A. Price. 2000. Generic relationships within 393:673-676. and between the gymnosperm families Podocarpaceae and Phyllocladaceae based on an analysis of the chloroplast gene rbcL. Aust. Henschel, J. R., and M. K. Seely. 2000. Long-term growth patterns of Welwitschia mirabilis, a long-lived plant of the Namib Desert (includSyst. Bot. 48:715-724. ing a bibliography). Plant Ecology 150:7-26. Crane, P. R. 1996. The fossil history of the Gnetales. Int. J. Plant Sci. 157 Hesse, M., M. Weber, and H. Halbritter. 2000. A comparative study of (6 Suppl.):S50-S57. Crane, P. R., and G. R. Upchurch. 1987. Drewria potomacensis gen. et the polyplicate pollen types in Arales, Laurales, Zingiberales, and Gnetales. Pages 227-239 in Pollen and spores: morphology and bisp. nov., an early Cretaceous member of Gnetales from the Potomac ology (M. M. Harley, C. M. Morton, and S. Blackmore, eds.). Royal group of Virginia. Am. J. Bot. 74:1722-1776. Botanic Gardens, Kew, UK. De Queiroz, A. 2004. The resurrection of oceanic dispersal in historical Hilu, K. W., T. Borsch, K. Muller, D. E. Soltis, P. S. Soltis, V. Savolainen, biogeography. Trends Ecol. Evol. 20:68-73. M. W. Chase, M. P. Powell, L. A. Alice, R. Evans, H. Sauquet, C. Dilcher, D. L., M. E. Bernardes-de-Oliveira, D. Pons, and T. A. Lott. Neinhuis, T. A. B. Slotta, J. G. Rohwer, T. A. B. Slotta, J. G. Rohwer, C. 2005. Welwitschiaceae from the Lower Cretaceous of Northeastern S. Campbell, and L. W. Chatrou. 2003. Angiosperm phylogeny based Brazil. Am. J. Bot. 92:1294-1310. on matK sequence information. Am. J. Bot. 90:1758-1776. Doyle, J. A., and M. J. Donoghue. 1993. Phylogenies and angiosperm Hilu, K. W., and H. Liang. 1997. The matK gene: sequence variation and diversification. Paleobiol. 19:141-167. application in plant systematics. Am. J. Bot. 84:830-839. Drummond, A. J., and A. Rambaut. 2005. BEAST vl.3, Available from Hipkins, V. D., C. J. Tsai, and S. H. Strauss. 1990. Sequence of the gene http://evolve.zoo.ox.ac.uk/beast/. for the large subunit of ribulose 1,5-bisphosphate carboxylase from Duan, S.-Y. 1998. The oldest angiosperm—a tricarpous female reproa gymnosperm, Douglas fir. Plant Mol. Biol. 15:505-507. ductive fossil from western Liaoning Province, NE China. Sci. China Ser. D 41:14-20. Ho, S. Y. W., M. J. Phillips, A. J. Drummond, and A. Cooper. 2005. Accuracy of rate estimation using relaxed-clock models with a critiFarjon, A., and S. Ortiz Garcia.2003. Cone and ovule development in Cunninghamia and Taizvania (Cupressaceae sensu lato) and its signif- cal focus on the early metazoan radiation. Mol. Biol. Evol. 22:13551363. icance for conifer evolution. Am. J. Bot. 90:8-16. Forget, P.-M., D. S. Hammond, T. Milleron, and R. Thomas. 2002. Sea- Holmgren, C. A., M. C. Peflalba, K. A. Rylander, and J. L. Betancourt. 2003. A 16,000 14C yr B.P. packrat midden series from the USAsonality of fruiting and food hoarding by rodents in neotropical Mexico borderlands. Quatern. Res. 60:319-329. forests: consequences for seed dispersal and seedling recruitment. Pages 241-256 in Seed dispersal and frugivory: Ecology, evolution Huang, J., D. E. Giannasi, and R. A. Price. 2005. Phylogenetic relationand conservation (D. J. Levey, W. R. Silva, and M. Galetti, eds.). CABI ships in Ephedra (Ephedraceae) inferred from chloroplast and nuclear Publishing, U.K. DNA sequences. Mol. Phylogenet. Evol. 35:48-59. Friis, E. M., K. R. Pedersen, and P. R. Crane. 2001. Fossil evidence of Huang, J., and R. A. Price. 2003. Estimation of the age of extant Ephedra using chloroplast rbcL sequence data. Mol. Biol. Evol. 20:435water lilies (Nymphaeales) in the Early Cretaceous. Nature 410:357440. 360. Gadek, P. A., D. L. Alpers, M. M. Heslewood, and C. J. Quinn. Huelsenbeck, J. P., and F. R. Ronquist. 2001. MrBayes: Bayesian inference of phylogenetic trees. Bioinformatics 17:754-755. 2000. Relationships within Cupressaceae sensu lato: a combined morphological and molecular approach. Am. J. Bot. 87:1044- Ickert-Bond, S. M., and M. F. Wojciechowski. 2004. Phylogenetic rela1057. tionships in Ephedra (Gnetales): Evidence from nuclear and chloroGandolfo, M. A., K. C. Nixon, and W. L. Crepet. 2001. Turonian Pinaceae plast DNA sequence data. Syst. Bot. 29:834-849. of New Jersey. Plant Syst. Evol. 226:187-203. Kimura, M. 1980. A simple method for estimating evolutionary rates of base substitutions through comparative studies of nucleotide seGandolfo, M. A., K. C. Nixon, and W. L. Crepet. 2004. Cretaceous quences. J. Mol. Evol. 16:111-120. flowers of Nymphaeaceae and implications for complex insect REFERENCES 2006 WON AND RENNER—DATING GNETUM, CALIBRATION WHEN OUTGROUPS ARE UNCERTAIN 621 Kishino, H., Thome, J. L., and W. J. Bruno. 2001. Performance of a Price, R. A. 1996. Systematics of the Gnetales: a review of morphological divergence time estimation method under a probabilistic model of and molecular evidence. Int. J. Plant Sci. 157 (6 Suppl.):S40-S49. rate evolution. Mol. Biol. Evol. 18:352-361. Pryer, K. M., H. Schneider, A. R. Smith, R. Cranfill, P. G. Wolf, J. S. Hunt, Knapp, M., K. Stockier, D. Havell, F. Delsuc, F. Sebastiani, P. J. Lockhart. and S. D. Sipes. 2001. Horsetails and ferns are a monophyletic group 2005. Relaxed molecular clock provides evidence for long-distance and the closest living relatives to seed plants. Nature 409:618-622. dispersal of Nothofagus (Southern Beech). PLoS Bio. 3: 38-43. Quandt, D., T. Borsch, S. Wicke, H. Won, S. S. Renner, and K. W. Hilu. Korall, P., and P. Kenrick. 2002. Phylogenetic relationships in Selaginel2006. Land plant evolution: a perspective from fast-evolving chlorolaceae based on rbcL sequences. Am. J. Bot. 89:506-517. plast regions. BAS abstract 835, http://2006.botanyconference.org/ engine/search/index.php?func=detail&amp;aid=835. Krassilov, V. A. 1982. Early Cretaceous flora of Mongolia. Palaeontographica Abt. B. 181:1-43. Qiu, Y.-L., M. W. Chase, D. H. Les, H. G. Hills, and C. R. Parks. 1993. Krassilov, V. A. 1986. New floral structure from the Lower Cretaceous Molecular phylogenetics of the Magnoliidae: A cladistic analysis of of Lake Baikal area. Rev. Palaeobot. Palynol. 47:9-16. nucleotide sequences of the plastid gene rbcL. Ann. Mo. Bot. Gard. Krassilov, V. A. 1997. Pages 50-52 in Angiosperm origins: Morpholog80:587-606. ical and ecological aspects. PENSOFT Publishers, Sofia, Bulgaria. Qiu, Y.-L., J. Lee, F. Bernasconi-Quadroni, D. E. Soltis, P. S. Soltis, M. Kubitzki, K. 1985. Ichthyochory in Gnetum venosum. An. Acad. Barsil. Zanis, E. A. Zimmer, Z. Chen, V. Savolainen, and M. W. Chase. 1999. Cienc. 57:513-516. The earliest angiosperms: evidence from mitochondria 1, plastid and Kubitzki, K. 1990. Gnetatae. Pages 378-391 in The families and genera nuclear genomes. Nature 402:404-407. of vascular plants I: Pteridophytes and Gymnosperms (K. U. Kramer Quinn, C. J., R. A. Price, and P. A. Gadek. 2002. Familial concepts and and P. S. Green, eds.). Springer-Verlag, Berlin, Heidelberg. relationships in the conifers based on rbcL and matK sequence comKumar, S., K. Tamura, I. B. Jakobsen, and M. Nei. 2001. MEGA2: molecu- parisons. Kew Bull. 57:513-531. lar evolutionary genetics analysis software. Bioinformatics 17:1244- Rai, H. S., H. E. O'Brien, P. A. Reeves, R. G. Olmstead, and S. W. Graham. 1245. (http://www.megasoftware.net) 2003. Inference of higher-order relationships in the cycads from a large chloroplast data set. Mol. Phylogenet. Evol. 29:350-359. LePage, B. A. 2003a. A new species of Thuja (Cupressaceae) from the Late Cretaceous of Alaska: implications of being evergreen in a polar Rambaut, A., and L. Bromham. 1998. Estimating divergence dates from environment. Am. J. Bot. 90:167-174. molecular sequences. Mol. Biol. Evol. 15:442-448. LePage, B. A. 2003b. The evolution, biogeography and palaeoecology Renner, S. S. 1999. Circumscription and phylogeny of the Laurales: evof the Pinaceae based on fossil and extant representatives. Acta Hort. idence from molecular and morphological data. Am. J. Bot. 86:1301615 (IV Int. Conifer Conf.):29-52. 1315. Les, D. H., D. K. Garvin, and C. F. Wimpee. 1991. Molecular evolution- Renner, S. S. 2004. Plant dispersal across the tropical Atlantic by wind ary history of ancient aquatic angiosperms. Proc. Natl. Acad. Sci. and sea currents. Int. J. Plant Sciences 165(4 Suppl.):S23-S33. USA. 88:10119-10123. Renner, S. S. 2005. Relaxed molecular clocks for dating historical plant Les, D. H., E. L. Schneider, D. J. Padgett, P. S. Soltis, D. E. Soltis and M. dispersal events. Trends Plant Sci. 10:550-558. Zanis. 1999. Phylogeny, classification and floral evolution of water Renner, S. S., and L-B. Zhang. 2004. Biogeography of the Pistia clade (Araceae): Based on Chloroplast and mitochondrial DNA sequences lilies (Nymphaeales): a synthesis of non-molecular, rbcL, matK and and Bayesian divergence time inference. Syst. Biol.53:422-432. 18S rDNA data. Syst. Bot. 24:28-46. Lidholm, J., and P. Gustafsson. 1991. A three-step model for the rear- Ridley, H. N. 1930. The dispersal of plants throughout the world. L. rangement of the chloroplast trnK-psbA region of the gymnosperm Reeve & Co., Ltd., Ashford. Pinus contorta. Nucleic Acids Res. 19:2881-2887. Rydin, C, M. Kallersjo, and E. M. Friis. 2002. Seed plant relationships Linder, H. P., C. R. Hardy, and F. Rutschmann. 2005. Taxon samand the systematic position of Gnetales based on nuclear and chloropling effects in molecular clock dating: an example from the African plast DNA: conflicting data, rooting problems and the monophyly Restionaceae. Mol. Phyl. Evol. 35:569-582. of conifers. Int. J. Plant Sci. 163:197-214. Magall6n, S., and M. J. Sanderson. 2002. Relationships among seed Rydin, C, B. A. Mohr, and E. M. Friis. 2003. Cratonia cotyledon gen. et sp. nov.: a unique Cretaceous seedling related to Welwitschia. Proc. plants inferred from highly conserved genes: sorting conflicting phyR. Soc. London Ser. B. 270 (Suppl.):29-32. logenetic signals among ancient lineages. Am. J. Bot. 89:1991-2006. Maheshwari, P., and V. Vasil. 1961. Gnetum. Council of Scientific & Rydin, C, K. R. Pedersen, and E. M. Friis. 2004. On the evolutionary history of Ephedra: Cretaceous fossils and extant molecules. Proc. Industrial Research, New Delhi, India. Natl. Acad. Sci. USA. 101:16571-16576. Manhart, J. R. 1994. Phylogenetic analysis of green plant rbcL sequences. Rydin, C, and N. Wikstrb'm. 2002. Phylogeny of Isoetes (Lycopsida): Mol. Phylogenet. Evol. 3:114-127. resolving basal relationships using rbcL sequences. Taxon 51:83-89. Markgraf, F. 1929. Monographie der Gattung Gnetum. Bull. Jard. Bot. Rydin, C, K. R. Pederson, P. R. Crane, and E. M. Friis. 2006. Former Buitenzorg Ser. 3,10:407-511. diversity of Ephedra (Gnetales): Evidence from Early cretaceous seed Markgraf, F. 1951. Gnetaceae. Pages 336-347 in Flora Malesiana 1,4. from Portugal and North America. Ann. Bot. 98:123-140. Miller, C. N. Jr. 1999. Implications of fossil conifers for the phylogenetic Sanderson, M. J. 2002. Estimating absolute rates of molecular evolution relationships of living families. Bot. Rev. 65:239-277. and divergence times: A penalized likelihood approach. Mol. Biol. Mohr, B. A., M. E. Bernardes-de-Oliveira, A. M. F. Barreto, and M. Evol. 19:101-109. C. Castro-Fernandes. 2004. Gnetophyte preservation and diversity in the Early Cretaceous Crato Formation (Brazil). VII. International Sanderson, M.}., and J. A. Doyle. 2001. Source of error and confidence intervals in estimating the age of angiosperms from rbcL and 18S Organization of Paleobotany Conference, Bariloche, Argentina. rDNA data. Am. J. Bot. 88:1499-1516. Mohr, B. A., and E. M. Friis. 2000. Early angiosperms from the Lower Cretaceous Crato Formation (Brazil), a preliminary report. Int. J. Schmidt, M., and H. A. W. Schneider-Poetsch. 2002. The evolution of gymnosperms redrawn by phytochrome genes: the Gnetatae appear Plant Sci. 161 (6 Suppl.):S155-S167. at the base of the gymnosperms. J. Mol. Evol. 54:715-724. Near, T. ]., P. A. Meylan, and H. B. Shaffer. 2005. Assessing concordance of fossil calibration points in molecular clock studies: an example Setoguchi, H., T. A. Osawa, J.-C. Pintaud, T. Jaffre, and J.-M. Veillon. 1998. Phylogenetic relationships within Araucariaceae based on rbcL using turtles. Am. Nat. 165:137-146. gene sequences. Am. J. Bot. 85:1507-1516. Ohyama, K., H. Fukuzawa, T. Kohchi, H. Shirai, T. Sano, S. Sano, K. Umesono, Y. Shiki, M. Takeuchi, Z. Chang, S. Aota, H. Inokuchi, Shi, S., H. Jin, Y. Zhong, X. He, Y. Huang, F. Tan, and D. E. Boufford. 2000. Phylogenetic relationships of the Magnoliaceae inferred from and H. Ozeki. 1986. Chloroplast gene organization deduced from complete sequence of liverwort Marchantia polymorpha chloroplast cpDNA matK sequences. Theor. Appl. Genet. 101:925-930. DNA. Nature 322:572-574. Soltis, D. E., P. S. Soltis, and M. J. Zanis. 2002. Phylogeny of seed plants based on evidence from eight genes. Am. J. Bot. 89:1670-1681. Palmer, J. D., D. E. Soltis, and M. W. Chase. 2004. The plant tree of life: an overview and some points of view Am. J. Bot. 91:1437- Stevenson, D., and T. Zanoni. 1991. Gnetaceae. Pages 12-18 in Flora of the Guianas (series A), vol. 9 (A.R.A. Gorts-van Rijn, ed.). 1445. Posada, D., and K. A. Crandall 1998. ModelTest: testing the model of Stockey, R. A. 1982. The Araucariaceae: an evolutionary perspective. Rev. Palaeobot. Palynol. 37:133-154. DNA substitution. Bioinformatics 14:817-818. 622 SYSTEMATIC BIOLOGY Sun, C, S.-L. Zheng, D. L. Dilcher, Y.-D. Wang, and S.-W. Mei. 2001. Early angiosperms and their associated plants from western Liaoning, China. Shanghai Scientific and Technological Education Publishing House, Shanghai, China. Swofford, D. L. 2002. PAUP*. Phylogenetic analysis using parsimony (*and other methods), version 4. Sinauer Associates, Sunderland, Massachusetts. Thompson, J. D., T. J. Gibson, F. Plewniak, F. Jeanmougin, and D. G. Higgins. 1997. The ClustalX Windows interface: flexible strategies for multiple sequence alignment aided by quality analysis tools. Nucleic Acids Res. 24:4876^1882. Thorne, J. L., and H. Kishino. 2002. Divergence time estimation and rate evolution with multilocus data sets. Syst. Biol. 51:689-702. Thorne, J. L., Kishino, H., and I. S. Painter. 1998. Estimating the rate of evolution of the rate of molecular evolution. Mol. Biol. Evol. 15:16471657. Treutlein, J., and M. Wink. 2002. Molecular phylogeny of cycads inferred from rbcL sequences. Naturwissenschaften 89:221-225. Tsumura,Y, and Y. Suyama. 1998. Differentiation of mitochondrial DNA polymorphisms in populations of five Japanese Abies species. Evolution 52:1031-1042. Van Roosmalen, M. G. M. 1985. Fruits of the Guianan Flora. Institute of Systematic Botany, Utrecht University. Wang, X. Q., D. C. Tank, and T. Sang. 2000. Phylogeny and divergence times in Pinaceae: evidence from three genomes. Mol. Biol. Evol. 17:773-781. Wang, Z.-Q. 2004. A new Permian gnetalean cone as fossil evidence for supporting current molecular phylogeny. Ann. Bot. 94:281-288. Wiegmann, B. M., D. K. Yeates, J. L. Thorne, and H. Kishino. 2003. Time flies, a new molecular time-scale for brachyceran flye evolution without a clock. Syst. Biol. 52:745-756. VOL. 55 Wikstrom, N., and P. Kenrick. 2000. Phylogeny of epiphytic Huperzia (Lycopodiaceae): paleotropical and neotropical clades corroborated by rbcL sequences. Nord. J. Bot. 20:165-171. Won, H., and S. S. Renner. 2003. Horizontal gene transfer from flowering plants to Gnetum. Proc. Natl. Acad. Sci. USA. 100:1082410829. Won, H., and S. S. Renner. 2005a. The internal transcribed spacer of nuclear ribosomal DNA in the gymnosperm Gnetum. Mol. Phylogenet. Evol. 36: 581-597. Won, H., and S. S. Renner. 2005b. The chloroplast trnT-tmF region in the seed plant lineage Gnetales. J. Mol. Evol. 60:1-12. Yang, Y, B.-Y. Geng, D. L. Dilcher, Z.-D. Chen, and T. A. Lott. 2005. Morphology and affinities of an Early Cretaceous Ephedra (Ephedraceae) from China. Am. J. Bot. 922:231-241. Yang, Z. 1997. PAML: A program package for phylogenetic analysis by maximum likelihood Computer Appl. Biosci. 13:555-556 (http://abacus.gene.ucl.ac.uk/software/paml.html). Yoo, M.-J., C. D. Bell, P. S. Soltis, and D. E. Soltis. 2005. Divergence times and biogeography of Nymphaeales. Syst. Bot. 30:693704. Zhang, L-B., and S. S. Renner. 2003. The deepest splits in Chloranthaceae as resolved by chloroplast sequences. Int. J. Plant Sci. 164(5 Suppl.):S383-S392. Zhou, S., S. S. Renner, and J. Wen. 2006. Molecular phylogeny and inter- and intracontinental biogeography of Calycanthaceae. Molecular Phylogenetics and Evolution 39:1-15. First submitted 23 June 2005; reviews returned 2 September 2005; final acceptance 13 March 2006 Associate Editor: Peter Under Male strobilus of Gnetum gnemon. The white structures are fertile stamens, the glistening droplets are nectar that is exuded from sterile ovules (Photo T. Stuetzel).