- Top of page
- 1 Material and methods
- 2 Results
- 3 Discussion
Abstract Ephedra comprises approximately 50 species, which are roughly equally distributed between the Old and New World deserts, but not in the intervening regions (amphitropical range). Great heterogeneity in the substitution rates of Gnetales (Ephedra, Gnetum, and Welwitschia) has made it difficult to infer the ages of the major divergence events in Ephedra, such as the timing of the Beringian disjunction in the genus and the entry into South America. Here, we use data from as many Gnetales species and genes as available from GenBank and from a recent study to investigate the timing of the major divergence events. Because of the tradeoff between the amount of missing data and taxon/gene sampling, we reduced the initial matrix of 265 accessions and 12 loci to 95 accessions and 10 loci, and further to 42 species (and 7736 aligned nucleotides) to achieve stationary distributions in the Bayesian molecular clock runs. Results from a relaxed clock with an uncorrelated rates model and fossil-based calibration reveal that New World species are monophyletic and diverged from their mostly Asian sister clade some 30 mya, fitting with many other Beringian disjunctions. The split between the single North American and the single South American clade occurred approximately 25 mya, well before the closure of the Panamanian Isthmus. Overall, the biogeographic history of Ephedra appears dominated by long-distance dispersal, but finer-scale studies are needed to test this hypothesis.
Efforts to date the evolutionary divergences of the five extant seed plant lineages (Gingko L., cycads, gymnosperms, Gnetales, and angiosperms) with molecular clocks have been hampered by the still unresolved relationships between them (for a summary, see Mathews et al., 2009, in press). Another problem is the marked difference in the rate of molecular evolution among seed plant lineages. The Gnetales in particular have unusually high or low substitution rates (depending on genus) in all datasets examined so far (Sanderson et al., 2000; Magallón & Sanderson, 2002, 2005; Mathews, 2009). For example, the rbcL substitution rate in Ephedra L. is approximately 10-fold slower than that in its sister clade Gnetum (Renner & Grimm, 2008). Such heterogeneity among lineages, which is not accommodated by molecular substitution models, presents a challenge for molecular clock dating, whether strict or relaxed. In response to this challenge, local and relaxed clock methods have been proposed that permit different parts of a tree to have different rates (Rambaut & Bromham, 1998; Thorne et al., 1998; Yoder & Yang, 2000; Kishino et al., 2001; Rannala & Yang, 2007). Of these relaxed clock approaches, several rely on a Bayesian framework and assume that substitution rates are autocorrelated between branches, meaning that rate changes occur gradually between ancestors and descendants as a clade diversifies. Other Bayesian clock models assume that branch-specific rates are drawn from a single underlying distribution, such as a log normal, gamma, or exponential distribution, the parameters of which are estimated from the data (Drummond et al., 2006; Rannala & Yang, 2007). Studies that have tested the performance of the different approaches have found that relaxed clock models with uncorrelated rates can outperform other approaches (Ho et al., 2005; Drummond et al., 2006; Lepage et al., 2007; but see Ho, 2009).
The Gnetales comprise Ephedra L., Gnetum L., and Welwitschia Hook.f. and are one of the five major groups of extant seed plants. Studies over the past 17 years have been unable to securely resolve the phylogenetic relationships of Gnetales with the four other seed plant lineages (Mathews et al., 2009, in press). Most recently, plastid sequence data have placed the Gnetales as sister to all non-Pinaceae conifers or cupressophytes (“gnecup” clade; Nickrent et al., 2000; Doyle, 2006; Chumley et al., 2008; Braukmann et al., 2009; Rydin & Korall, 2009).
Resolving species-level relationships within Ephedra has been equally problematic. However, in contrast with the many well-supported, yet often mutually incompatible hypotheses on seed plant phylogeny, relationships in Ephedra have been largely unresolved owing to few informative characters in investigated gene regions and substantial plasticity in gross morphological traits (Ickert-Bond & Wojciechowski, 2004; Rydin et al., 2004; Huang et al., 2005). A recent study, with denser species sampling, provides support for several subgeneric clades (Rydin & Korall, 2009), but the deepest divergences in the genus are still ambiguous.
Divergence times from molecular clock analyses for the most recent common ancestors of living Gnetales genera range from 8–32 mya under a strict clock for Ephedra (Huang & Price, 2003) to 10–11 or 14 mya under a strict clock (Won & Renner, 2003, 2006) or 26–38 mya under a relaxed clock for Gnetum (Won & Renner, 2006). Estimates for the Gnetales crown group range from 120–131 mya (relaxed clock; Ickert-Bond & Wojciechowski, 2002) to 189 mya (relaxed clock; Schneider et al., 2004). Recent paleobotanical discoveries have further stirred up discussions about the age of Ephedra (Yang et al., 2005; Rydin et al., 2006; Friis et al., 2009), with some authors suggesting that Cretaceous fossil seeds resemble living species of Ephedra and may date the divergence of crown group Ephedra to ca. 125 mya (Yang et al., 2005; Rydin et al., 2006; Y. Yang, Institute of Botany, Beijing, pers. comm., 2008). However, to date, no ephedroid seed fossil has been unambiguously placed within crown group Ephedra, and these fossils are therefore of limited use as calibration points in molecular dating analyses. Conversely, coalified Ephedra seeds from the Drewry's Bluff locality of the Patuxent Formation in Virginia, USA, and from Buarcos and Torres Vedras localities in Portugal, which date to the late Barremian to early–middle Aptian age, have been assigned to stem group Ephedra based on two preserved features: (i) in situ Ephedra-type pollen, including discarded upcurled exines, which show that the pollen had germinated inside the ovules; and (ii) preserved papillae formed by the inner epidermis of the seed envelope. A combination of these features is unique to Ephedra (Rydin et al., 2006).
Here we apply relaxed molecular clock dating, using an uncorrelated rates model, to an Ephedra dataset that represents all the major clades found by Rydin and Korall (2009) with the goal of inferring the most probable age of the Ephedra crown group as well as the timing of key divergence events in the genus. The radiation of extant Ephedra is interesting because of its disjunct distribution in deserts north and south of the tropics, but not in the intervening regions (a classic amphitropical range; see Wen & Ickert-Bond, 2009). The genus comprises approximately 50 species, which are more-or-less equally distributed between the Old and New World deserts. In light of recent palaeobotanical evidence (above), availability of large molecular datasets for both Ephedra and Gnetum, as well as new approaches to dating that take into account topological uncertainty and rate heterogeneity among lineages, in the present paper we provide new age estimates for Ephedra and discuss their implications for the evolution of the genus.
- Top of page
- 1 Material and methods
- 2 Results
- 3 Discussion
The ML tree obtained from the 10 locus–95 taxon dataset shows relationships within Ephedra (Fig. 1) that are similar to those recovered by Rydin and Korall (2009), although species sampling in the present study is smaller. Ephedra foeminea is sister to the rest of Ephedra. The next diverging clade is one of strictly Mediterranean taxa (E. altissima, E. aphylla, E. milleri, E. alata and E. fragilis), sister to the rest of Ephedra (“core Ephedra” sensu Rydin & Korall, 2009). Core Ephedra comprises several subclades of Mediterranean and Asian distribution (e.g. E. laristanica and E. foliata), but there is no statistical support for their precise composition and relationships. However, the New World clade of Ephedra is strongly supported (bootstrap (BS) 94%) and consists of North American and South American clades. The large substitution rate heterogeneity among Gnetales is evident from the phylogram (Fig. 1), particularly the long branches leading to the three genera compared with the significantly reduced branch lengths within Ephedra.
Figure 1. Maximum likelihood (ML) phylogeny, inferred from 10 combined chloroplast and nuclear ribosomal DNA (nrDNA) loci for 95 taxa of gymnosperms; branch lengths computed using RAxML. The ML bootstrap support values above 80% are indicated above the branches. The biogeographic distribution of Ephedra taxa is shown on the right.
Download figure to PowerPoint
Relationships among the fewer species included in the molecular clock runs (Fig. 2) differ in part from those obtained from the ML analysis (Fig. 1) and have slightly better statistical support because the matrix includes many fewer missing nucleotides. Table 2 lists divergence times obtained for key nodes within Ephedra (labeled 3–9 in Fig. 2). Although deep divergences have originated in the Oligocene, most of the tip clades have diverged more recently in the Late Miocene or Pliocene (Fig. 2).
Figure 2. Chronogram based on 42 accessions of Ephedra, Gnetum, and Welwitschia from combined chloroplast and nuclear ribosomal DNA (nrDNA) loci obtained under a model of uncorrelated rate change using one fossil-based constraint (see text for details). Node heights are median ages, with gray bars indicating the 95% highest posterior density intervals (see Table 2 for details). The maximum likelihood bootstrap support values are indicated below the branches. The distribution of Ephedra taxa is indicated on the right with node labels (black circles) pointing to particular nodes of biogeographic interest (see Table 2 and text for details). Pal., Paleocene; Oligoc., Oligocene; Mioc., Miocene; Pl., Pleistocene.
Download figure to PowerPoint
Table 2. Time estimates (in million years) and confidence intervals for significant nodes for the crown group Ephedra obtained from combined analysis of the 7736-nucleotide matrix under an uncorrelated rates molecular clock (see Material and Methods)
|Node no.||Clade name||Fossil evidence||Age (million years)||95% Highest posterior density intervals|
|1||Split: Welwitschia vs. Gnetum||Cratonia cotyledon||110*||111.35||87.21, 127.01|
|2||Divergence of Ephedra from Gnetum and Welwitschia||Ephedra archaerhytidosperma||125||166.61||90.62, 192.34|
|3||Divergence Mediterranean clade from core Ephedra|| || ||30.39||20.55, 73.5|
|4||Divergence of NW clade from the rest of core Ephedra|| || ||29.56||8.84, 41.53|
|5||Divergence of China clade from rest of mixed Asia clade|| || ||27.63||14.45, 49.36|
|6||Divergence of Middle Eastern/Horn of Africa from African Mediterranean members|| || ||25.8||15.37, 55.53|
|7||Divergence of North American clade from South American clade|| || ||24.78||8.84, 41.53|
|8||Divergence of Asia 2 clade from combined Asia 1/Horn of Africa and Asia clade|| || ||20.61||14.35, 49.36|
|9||Divergence of Asia 1 clade from Horn of Africa/Asia clade|| || ||15.51||6.18, 32.5|
- Top of page
- 1 Material and methods
- 2 Results
- 3 Discussion
The results of the present study provide strong evidence for a recent radiation of extant Ephedra. Given the few clear morphological differences among species, it has been suggested that the lack of molecular divergence in Ephedra plastid genomes may be the result of hybridization and polyploidization, which appears to be rampant in the genus (Cutler, 1939; Ehrendorfer, 1976; Choudry, 1984; Wendt, 1993). Plants with montane distribution also frequently exhibit rapid diversification, likely because of small-scale habitat heterogeneity (Bell & Donoghue, 2005; Hughes & Eastwood, 2006).
Studies with a comprehensive species sampling of Ephedra (the Bayesian analysis in Rydin & Korall, 2009; the ML analysis in the present study) indicate a basal grade of Mediterranean species and thus a possible origin of the crown group of Ephedra in the Mediterranean region (northern Africa, southern Europe, the Near East). However, these basal divergences still have little statistical support owing to the limited signal in the loci so far included (Ickert-Bond & Wojciechowski, 2004; Ickert-Bond et al., 2009; Rydin & Korall, 2009; present study). Parsimony analysis (Rydin & Korall, 2009), as well as Bayesian analysis of the reduced dataset in the present study (Fig. 2), results in a Mediterranean clade sister to the remaining Ephedra. The respective divergence may have taken place some 30 mya (Fig. 2, node 3; Table 2). Within the Mediterranean clade, the Near Eastern E. foliata (Arabia and Somalia) and E. laristanica (Iran) split from the western Mediterranean species ca. 26 mya. Turning to the (mostly) Asian clade of Ephedra, the prevailing pattern appears to be westward dispersal (Fig. 3), with an estimated divergence of a strictly Chinese clade from the rest of the Asian/African clade at 28 mya (Fig. 2, node 5; Table 2). Dispersal into the Horn of Africa from the Asia 1 clade (Fig. 2, node 9; Table 2) may date back to 16 mya.
Figure 3. Distribution of Ephedra (green shading) and hypothesized intercontinental (solid arrows) and intracontinental dispersal routes (dashed arrows).
Download figure to PowerPoint
New World species are monophyletic and estimated to have diverged from their mostly Asian sister clade some 30 mya. This timeframe is corroborated by many other Beringian plant disjunctions (for a review, see Wen & Ickert-Bond, 2009). In turn, the New World species split into a North American and a South American clade (Fig. 2), which appear to have diverged approximately 25 mya; that is, well before the closure of the Panamanian Isthmus (Fig. 2, node 7; Table 2). These results mirror other studies indicating that significant dispersal took place between Mesoamerica and South America before the closure of the Isthmus of Panama during the Oligocene or Miocene (e.g. mammals: Marshall & Sempere, 1993; Melastomeae: Renner & Meyer, 2001; Ruprechtia (Polygonaceae) and Nissolia (Leguminosae): Pennington et al., 2004; and Platymiscium (Leguminosae): Saslis-Lagoudakis et al., 2008).
Dispersal in Ephedra may have been facilitated by the ovulate bracts, which, in some species of Ephedra, are bright red and fleshy and indicative of endozoochory (Stapf, 1889; Freitag & Maier-Stolte, 1994; Danin, 1996; Hódar et al., 1996). Bird dispersal has also been observed directly (Ridley, 1930; Hollander et al., 2009). In contrast, dry, wing-bracted strobili are adapted for anemochory (Stapf, 1889; Danin, 1996). The seeds of North American E. aspera, E. californica, E. funerea, E. nevadensis, and E. viridis are not fleshy and their ovulate bracts are not winged. Seeds of these “intermediate bracted” taxa often accumulate at the stem base, and seed-caching rodents have been observed as dispersers (Ickert-Bond, 2003; Ickert-Bond & Wojciechowski, 2004; Hollander & Vander Wall, 2009; Hollander et al., 2009). Wind-dispersed Ephedra typically inhabit marginal habitats, such as hyperarid deserts or dry salt lakes devoid of animal life (Danin, 1996), and, in general, dispersal biology in Ephedra appears to relate to habitat, rather than being phylogenetically conservative (Hollander et al., 2009).
Recent studies have found that reliable topologies may be obtained even in the face of large amounts of missing data (e.g. Wiens, 2003, 2006; McMahon & Sanderson, 2006; Smith et al., 2009). However, for molecular clock dating, missing data present a so-far insurmountable challenge. This is because estimation of divergence time depends on accurate estimates of branch lengths, which can only be obtained with large numbers of nucleotides (Sanderson, 1998). When the BEAST dating runs failed to reach stable distributions, we first reduced the number of empty cells by deleting data partitions that lacked sequences for more than 30% of the included species; next, we deleted species that lacked sequences for more than five loci. Even so, a combined run length of 108 million generations was needed for each parameter to converge on a stationary distribution.
A caveat with all molecular clock dating is that the absolute ages obtained depend on the calibration used. An earlier study that concentrated on Gnetum and only included three species of Ephedra, using a Bayesian relaxed clock and an auto-correlated model, explored the effects of three different constraints (Won & Renner, 2006). In one experiment, these authors used 125-mya-old Ephedra seeds to constrain the crown group age of Ephedra. This had the effect of roughly doubling within-Gnetum estimates compared with the ages obtained when these seeds were assigned to the Ephedra stem (Won & Renner, 2006, table 1). In the present study, we initially included representatives of all major lineages of gymnosperms so that we could constrain the Ephedra stem to a minimum of 125 mya. However, this introduced the problem of the uncertain placement of Gnetales within seed plants, a problem that Won & Renner (2006) circumvented by conducting dating runs under four different seed plant topologies.
With just the Gnetales included, as in Fig. 2, one cannot infer a Bayesian probability distribution around the split between Ephedra and Gnetum/Welwitschia. Instead, we decided to rely exclusively on the Cratonia cotyledon fossil from the Early Cretaceous of Brazil, the assignment of which is unambiguous because it clearly represents the Welwitschia stem group (Rydin et al., 2003). This calibration yielded an age of 167 mya (91–192 mya confidence interval) for the divergence between Ephedra and the other two genera. This age range is too large to be very useful, but fits the placement of Gnetales within conifers, perhaps as sister to the non-Pinaceae conifers (“gnecup” clade; Nickrent et al., 2000; Doyle, 2006; Chumley et al., 2008; Braukmann et al., 2009; Rydin & Korall, 2009), and also with the Ephedra pollen and seed fossil record.
The present biogeographic analysis (Figs. 2, 3) corroborates other molecular studies that have found New World clades of Oligocene age evolving out of Asian paraphyletic residuals, which is the classic pattern of Beringian disjunctions (for summaries, see Wen & Ickert-Bond, 2009). Our work also adds to a growing body of studies reporting long-distance dispersal between arid floras in North and South America (Moore et al., 2006). Finer-scale studies are now needed to test the broad-brush biogeographic scenario for Ephedra developed here.