Abstract Toxicodendron is a genus in the Rhus complex of Anacardiaceae with a disjunct distribution between eastern Asia and North America, extending to southeastern Asia and the neotropics. Nuclear (internal transcribed spacer, external transcribed spacer, and NIA-i3) and chloroplast (ndhF and trnL-F) sequences were used to construct phylogenetic relationships of Toxicodendron. Phylogenetic analysis of these data strongly support Toxicodendron as a monophyletic group distinct from other genera of the Rhus complex, and the phylogeny does not fully corroborate classification at the sectional level. Two temperate disjunct lineages were detected, one from section Toxicodendron and the other between the eastern North American Toxicodendron vernix and the eastern Asian Toxicodendron vernicifluum. Their divergence times were estimated to be 13.46 (7.95–19.42) and 7.53 (2.76–12.86) mya, respectively. The disjunction between section Griffithii (taxa from warm temperate to tropical Asia) and Toxicodendron striatum (from the neotropics) was supported and their divergence time was estimated to be 20.84 (11.16–30.52) mya in the early Miocene. Our biogeographic results and the paleontological data support the Bering land bridge as the most likely route to explain the temperate disjunctions, yet the tropical disjunction in Toxicodendron seems to be best explained by the North Atlantic land bridge hypothesis.
The cashew family (Anacardiaceae) consists of more than 80 genera and 600–750 species mostly distributed in tropical Africa, Asia, and the Americas, with a small number of species in subtropical and temperate areas (McNair, 1925; Cronquist, 1981; Pell, 2004; Wannan, 2006; Pell et al., 2008). Toxicodendron Mill. is one of a few genera with a primary distribution in temperate to subtropical regions (Gillis, 1971; Zomlefer, 1994). The genus comprises approximately 24 species showing a disjunct distribution in temperate North America and eastern Asia (Li, 1952; Gray, 1846; Graham, 1972; Wen, 1999, 2001). A few taxa of Toxicodendron also occur in the tropics ranging from Central America to northernmost South America and southeastern Asia.
Toxicodendron is well-known for possessing skin-irritating oil (urushiol) that can cause severe allergic reactions to humans. Species of the genus also have lacquer in the phloem, and the lacquer is important for making anticorrosives or decorative paint. Taxa of this genus were often included in Rhus L. s.l. (Barkley, 1937a, 1963; Young, 1979; Cronquist, 1981). Miller et al. (2001) showed that Rhus should be delimited more narrowly, and that Toxicodendron and several other genera including Actinocheita F. A. Barkley, Cotinus Mill., Malosma Nutt. ex Abrams, Melanococca Blume, Metopium P. Browne, and Searsia F. A. Barkley are best segregated from Rhus.
In comparison with other genera in the Rhus complex, Toxicodendron generally has axillary inflorescences, whitish to dune-colored fruits without glandular pubescence, and toxic catechol in its resin (Brizicky, 1963; Gillis, 1971). However, the delimitation of Toxicodendron remains controversial. For example, Brizicky (1963) suggested treating Toxicodendron as a subgenus of Rhus because of their overall similarities in the inflorescence and the structure of flowers and fruits. Gillis (1971), however, preferred to treat Toxicodendron as a separate genus because of its suite of distinctive characters (e.g., presence of poisonous resin, absence of red glandular hairs on pedicels and fruits, and significantly smaller pollen grains).
Toxicodendron is commonly delineated by four sections (Venenata, Toxicodendron, Griffithii, and Simplicifolia). Section Toxicodendron (five species) comprises usually woody climbers with trifoliate leaves, short and subulate filaments, elongate anthers, and pendulous infructescences. The most well-known member of this section is Toxicodendron radicans, which is widely distributed throughout eastern North America and eastern Asia and is commonly known as poison ivy due to its superficial resemblance to English ivy (Hedera helix L. of Araliaceae) and Boston ivy [Parthenocissus tricuspidata (Siebold & Zuccarini) Planch. of Vitaceae]. Section Simplicifolia is represented by only one narrowly distributed species of northern Borneo in southeastern Asia. Its habit is very similar to that of section Toxicodendron as a deciduous woody climber, but its leaves are simple and subcoriaceous (Gillis, 1971). Section Venenata includes trees or woody shrubs with pinnately compound leaves, long filaments, and short globose anthers, mainly from eastern Asia. It is the largest section in the genus with approximately 18 species distributed throughout eastern Asia from Japan to Indonesia (Fig. 1). One species has a broad distribution within North America, ranging from southern Quebec to Florida, while another occurs in Central and South America from Veracruz, Mexico to the highlands of Colombia (Gillis, 1971; Min, 1980). Section Griffithii contains five species and has been segregated from section Venenata by Min (1980) based on their thick coriaceous leaves, erect infructescences, and mature fruits with irregular dehiscence of exocarps.
Two disjunct lineages between eastern Asia and North America have been reported in the genus, one from section Toxicodendron and the other from section Venenata. Toxicodendron radicans is perhaps one of the earliest examples cited by Gray (Gray, 1846, 1859) as showing a floristic connection between eastern Asia and eastern North America (also see Gillis, 1971, 1975; Wen, 1999). This is one of two examples that show such a disjunction at the species level, with the other being Phryma of Phrymaceae (Li, 1952; Wen, 1999; Nie et al., 2006). The distributional pattern in section Venenata is also intriguing with most species distributed throughout eastern Asia from Japan to Indonesia and only two species broadly distributed from North, to Central, to northern South America.
Most disjunct taxa between eastern Asia and North America are north temperate elements with only a few occurring in the subtropical and tropical regions (Li, 1952; Wen, 1999, 2001; Ying, 1983). Biogeographic studies so far have largely focused on temperate taxa in the Northern Hemisphere (Wen, 1999; Donoghue et al., 2001). However, it is important to include the close relatives from tropical and subtropical regions to further understand the evolution of the temperate disjunctions in the Northern Hemisphere (Lavin & Luckow, 1993; Wen, 1999, 2001; Ickert-Bond et al., 2007). A few examples of biogeographic disjunctions between temperate eastern Asia and North America with tropical extensions have been surveyed. One example can be found in Rhus s. str. Phylogenetic and biogeographic studies have suggested a migration route from North America into Asia by way of the Bering land bridge (BLB) with an ancient split in the late Eocene (Yi et al., 2004, 2007).
Although morphological data seem to support the separation of Toxicodendron from Rhus s.l., rigorous phylogenetic analyses of molecular sequence data and biogeographic analyses have never been used to test the status and evolution of Toxicodendron in the complex. The use of multiple genes from different genomes, especially low-copy nuclear genes, is needed to fully address these questions in a biogeographic context (Wang et al., 2000; Peng & Wang, 2008). In this study, we use nuclear [internal transcribed spacer (ITS), external transcribed spacer (ETS), and NIA-i3] and chloroplast data (trnL-F and ndhF) to: (i) test the monophyly of Toxicodendron in the Rhus complex; (ii) construct phylogenetic relationships within the genus; and (iii) reconstruct the biogeographic history of Toxicodendron, especially concerning the temperate and tropical intercontinental disjunctions between the Old and the New World.
1 Material and methods
1.1 Taxon sampling
Sequences of 68 accessions of Toxicodendron and some closely related taxa were used in the study (Table 1). Fifty samples representing 18 of the ca. 24 Toxicodendron species were sequenced, covering the species diversity of the three main sections (i.e., Venenata, Toxicodendron, and Griffithii) with an emphasis on the two main distributional centers of eastern Asia and North America. Our sampling of the three large sections represented the morphological and geographic diversity of the group. Rhus s. str., Actinocheita, Malosma, and Searsia of the Rhus complex have been considered to be the closest relatives of Toxicodendron (Miller et al., 2001; Yi et al., 2004). Representatives from the complex and the family were selected as outgroups (see Table 1).
Table 1. Voucher information and GenBank accession numbers for samples of Toxicodendron and outgroups used in this study
T. hookeri (Sahni & Bahadur) C.Y. Wu & T.L. Ming var. microcarpum (C.C. Huang ex T.L. Ming) C.Y. Wu & T.L. Ming
Nie & Meng 396 (KUN)
Nie & Meng 397 (KUN)
T. pubescens Mill.
Wen 9998 (US)
USA: South Carolina
T. radicans (L.) Kuntze subsp. hispidum (Engler) Gillis
Nie & Meng 213 (KUN)
Wen 5463 (US)
T. radicans subsp. radicans (L.) Kuntze
Nie & Meng 530 (KUN)
Wen 6236 (F)
Wen 6271–1 (F)
USA: North Carolina
Wen 8570 (US)
USA: Washington DC
Wen 8747 (US)
T. rydbergii (Small ex Rydb.) Greene
Wen 10495–4 (US)
Wen 9698 (US)
Wen 9760 (US)
T. striatum (Ruiz & Pavon) Kuntze
Wen 8712 (US)
T. succedaneum (L.) Kuntze
Nie & Meng 334 (KUN)
Nie & Meng 339 (KUN)
Nie & Meng 377 (KUN)
Nie & Meng 429 (KUN)
Nie & Meng 431 (KUN)
Nie & Meng 436 (KUN)
Nie & Zhu 549 (KUN)
Nie & Zhu 573 (KUN)
Shui et al 82468 (KUN)
Tibet 2196 (KUN, US)
Wen 9273 (US)
T. sylvestre (Sieb. & Zucc.) Kuntze
Nie & Meng 383 (KUN)
Wen 9275 (US)
T. trichocarpum (Miq.) Kuntze
Nie & Meng 388 (KUN)
T. vernicifluum (Stokes) F.A. Barkley
Nie & Meng 229 (KUN)
Nie & Meng 329 (KUN)
Nie & Meng 376 (KUN)
Tibet 1661 (KUN, US)
Wen 9001 (US)
T. vernix (L.) Kuntze
Wen 7146 (F)
USA: Illinois (cult.)
T. wallichii (Hook. f.) Kuntze var. microcarpum T.L. Ming
Nie & Meng 430 (KUN)
Nie & Meng 446 (KUN)
Wen 10611 (US)
T. yunnanensis C.Y. Wu
Nie & Meng 426 (KUN)
Wen 9148 (US)
Actinocheita filicina (D.C.) F.A. Barkley
Panero s. n. (CS)
Lithrea ternifolia (Gillies) F.A. Barkley
Nee & Wen 53849
Malosma laurina (Nutt.) Nutt. ex Engl.
Miller 34 (CS)
USA: California (cult.)
Mauria heterophylla Kunth
Nee & Wen 53811
Pistacia chinensis Bunge
Wen 7090 (F)
USA: California (cult.)
Pistacia mexicana Kunth
Parfitt 27 (F)
USA: California (cult.)
Pistacia mexicana Kunth
Wen 8675 (US)
Pistacia weinmannifolia Poisson
Ji 0174 (KUN)
Poupartia minor (Bojer) L. Marchand
Wen 9481 (US)
Rhus choriophylla Wooton & Standl.
Miller 27 (CS)
Rhus integrifolia Benth. & Hook f. ex S. Watson
Miller 28 (CS)
USA: California (cult.)
Rhus sandwicensis A. Gray
Wen 7052 (F)
Rhus trilobata Nutt. ex Torr. & Gray
Miller 21 (CS)
Rhus typhina L.
Wen 8557 (US)
USA: Washington DC
Schinus molle L.
Wen 6686 (F)
USA: California (cult.)
Searsia dentata (Thunb.) F.A. Barkley
Wen 10090 (US)
South Africa: Kwazulu Natal
Searsia glauca (Thunb.) Moffett
Wen 10051 (US)
South Africa: Western Cape
Searsia natalensis (Bernh. ex Krause) F.A. Barkley
Luke 11694 (US)
Searsia pyroides (Burch.) Moffett
Luke 11693 (US)
1.2 DNA extraction, amplification, and sequencing
Genomic DNA was extracted from silica-gel dried leaf material using the DNeasy extraction kits following the manufacturer's protocol (Qiagen, Mississauga, Canada). The nuclear internal transcribed spacer (ITS) region was amplified using primers ITS4 and ITS5 (White et al., 1990) or N-nc18S10 and C26A (Wen & Zimmer, 1996). Nuclear external transcribed spacer (ETS) was amplified and sequenced using primers ETS1F and 18S-IGS (Pell et al., 2008). The nuclear NIA-i3 intron was sequenced using primers NIA-i3F and NIA-i3R (Howarth & Baum, 2002). The chloroplast ndhF gene (Olmstead & Sweere, 1994) and the trnL-F region (Taberlet et al., 1991) were also sequenced in this study. Amplification reactions were carried out in a 25 μL volume containing 1.5 mmol/L MgCl2, 0.2 mmol/L each dNTP, 0.2 μmol/L each primer, 1 U Taq polymerase, and approximately 25 ng DNA template. The sequencing reaction was carried out in a 10 μL final volume using the BigDye Terminator cycle sequencing kit (PE Applied Biosystems, Foster City, CA, USA) following the manufacturer's instructions and viewed with an ABI 3100 automated DNA sequencer (Applied Biosystems, Foster City, CA, USA). The resulting sequences were edited using Sequencher (version 4.1.4; Gene Codes Corporation, Ann Arbor, MI, USA) and aligned with ClustalX version 1.83 PC version (Thompson et al., 1997), followed by manual adjustment in BioEdit (Hall, 1999). All sequences have been deposited in GenBank with accession numbers (Table 1).
1.3 Phylogenetic analyses
Phylogenetic analyses were carried out using maximum parsimony (MP) and Bayesian inference. MP analysis was done using PAUP* version 4.0b10 (Swofford, 2003) with heuristic search, random taxon addition, tree bisection and reconnection branch-swapping, and the Mulpars and Steepest descent options. Parsimony bootstrap (PB) support was obtained using 1000 replicates, with the random taxon addition limited to 10.
Bayesian inference was carried out using MrBayes version 3.12 (Huelsenbeck & Ronquist, 2001). Nucleotide substitution model parameters were determined for the datasets using the Akaike information criterion in Modeltest version 3.7 (Posada & Crandall, 1998; Posada & Buckley, 2004). Bayesian analyses started from random trees and used four Markov chain Monte Carlo (MCMC) runs, monitoring over two million generations and re-sampling trees every 100 generations. Runs were repeated twice to confirm results. After discarding the burn-in trees, the remaining trees were imported into PAUP* and a 50% majority-rule consensus tree was produced to obtain posterior probabilities (PP) of the clades. Internodes with PP ≥95% in the consensus trees were considered statistically significant.
To evaluate the congruence of the chloroplast and nuclear datasets, we used an incongruence length difference (ILD) test (Farris et al., 1994) as a conservative initial test of data partition congruence (Hipp et al., 2004). The ILD test was carried out with 100 replicates of heuristic searches using tree bisection and reconnection branch-swapping with 10 random sequence additions.
1.4 Dating of divergence times
All nuclear and chloroplast sequences were used for dating divergence times because higher accuracy and more confidence can be inferred based on multiple sequences (Renner, 2005). A likelihood ratio test (Felsenstein, 1988) ruled out a global molecular clock (P < 0.05) for our dataset. Bayesian dating method under a relaxed clock model relying on uncorrelated rates was used in this study (Drummond et al., 2006).
The Bayesian approach to estimate divergent times and their credibility intervals was implemented in Bayesian evolutionary analysis by sampling trees (BEAST) version 1.4.8, which uses an MCMC to co-estimate topology, substitution rates, and node ages (Drummond & Rambaut, 2007). The BEAUti interface (distributed with BEAST) was used to create the input file to run in BEAST. All analyses were done using the HKY model of nucleotide substitution (Hasegawa et al., 1985). Different sequence regions were partitioned and model parameters were unlinked across partitions, with rate variation among sites modeled using a gamma distribution and four rate categories. Our analyses were performed based on a relaxed molecular clock model with uncorrelated rates drawn from a log-normal distribution and the birth-death tree prior (Drummond et al., 2006). Posterior distributions of parameters were approximated using two independent MCMC analyses of 10 million steps each, with the first 10% being discarded. The log files were combined to check for convergence on the same distribution and to ensure adequate sample sizes using Tracer version 1.4 (Rambaut & Drummond, 2007). The samples from the posterior were summarized on the maximum clade credibility tree using TreeAnnotator version 1.4.8 and visualized using FigTree version 1.2. Means and 95% higher posterior densities (HPD) of age estimates were obtained from the combined outputs using Tracer version 1.4.
Wood and pollen fossils of Anacardiaceae have been dated to the Paleocene (Hsü, 1983; Muller, 1984). Thus, a normally distributed calibration prior with the mean 60 mya and standard deviation 3.0 mya (roughly matching the Paleocene epoch 55–65 mya) was constrained for the root age of the family. Reliable macrofossils of Rhus were found from western North America in the middle Eocene (Manchester, 1994). They share many characters with the extant Rhus including lateral compression, longitudinal vascular bundles forming ribs in the mescocarp, and two or more layers of columnar cells in the endocarp (Manchester, 1994). The minimum stem age of Rhus was thus calibrated to be 44 with standard deviation of 1.0 mya. Other earlier fossils are leaf impressions dated back to the late Palaeocene–early Eocene (MacGinitie, 1969; Wilf, 2000) and not used for calibrations.
2.1 Molecular analysis
The aligned trnL-F sequences generated a data matrix of 1094 bp with 82 variable sites, 45 of which were parsimony-informative. The ndhF dataset had 2021 aligned positions, with 180 variable sites, 80 of which were parsimony-informative. Because there is no recombination in the chloroplast DNA (cpDNA), we combined the ndhF and trnL-F data directly. The aligned matrix of the combined cpDNA data had 3115 characters including 125 parsimony informative sites. The MP analysis yielded more than 10,000 most parsimonious trees (MPTs) with a length of 350 steps, a consistency index (CI) of 0.82, CI excluding uninformative characters = 0.69, a retention index (RI) of 0.91, and a rescaled consistency index (RC) of 0.74. Phylogenetic trees derived from both MP and Bayesian inference were topologically almost identical and their strict consensus tree with PB and PP values is shown in Fig. 2.
The ITS dataset comprised 770 aligned positions including 151 parsimony-informative sites. The ETS dataset contained 395 aligned positions, 76 of which were parsimony-informative. The ILD test did not detect any conflict between ITS and ETS, thus we combined the two datasets. The ITS-ETS matrix contained 1165 characters with 227 parsimony informative sites. Over 100,000 MPTs were generated with a length of 934 steps (CI= 0.58, CI excluding uninformative characters = 0.47, RI= 0.77, and RC= 0.45). The MP and the Bayesian trees were topologically similar and only the strict consensus tree with PB and PP values is shown in Fig. 3.
The NIA-i3 dataset included 1388 aligned positions including 194 parsimony-informative characters. The MP analysis generated more than 10,000 MPTs with a length of 519 steps, CI= 0.83, CI excluding uninformative characters = 0.75, RI= 0.92, and RC= 0.76. The NIA-i3 strict consensus tree is similar to the Bayesian topology and only the former with PB and PP values is shown in Fig. 4.
Results of the ILD test showed significant incongruence among cpDNA, ITS-ETS, and NIA-i3 datasets (P= 0.001), so we did not combine the three datasets in the phylogenetic analyses.
Toxicodendron sections Toxicodendron and Griffithii were well supported to be monophyletic by all analyses (Figs. 2–4), but taxon relationships within section Toxicodendron were not well resolved. For example, samples from North America and eastern Asia were not clearly separated (Figs. 2–4). The Asian Toxicodendron trichocarpum and the neotropical Toxicodendron striatum appear as two highly isolated species in both the ITS-ETS and the cpDNA phylogenies (Figs. 2, 3), but the latter grouped with section Griffithii in the NIA-i3 tree (Fig. 4).
All analyses suggested that section Venenata was not monophyletic (Figs. 2–4). Three clades were recognized within section Venenata. Clade A comprised Toxicodendron succedaneum and Toxicodendron sylvestre with strong support in the ITS-ETS tree (PB = 82%, PP = 100%, Fig. 3), but collapsed in the NIA-i3 strict consensus tree (Fig. 4). Clade B contained only Toxicodendron grandiflorum with robust support in the cpDNA tree (PB = 99%, PP = 100%, Fig. 2). Clades A and B grouped together with strong support in the cpDNA and the NIA-i3 trees, but this clade was not well supported in the ITS-ETS data. All data supported clade C as comprising the eastern Asian Toxicodendron vernicifluum and the eastern North American Toxicodendron vernix (Figs. 2–4).
Relationships among deeper clades of the genus were not well resolved by the cpDNA or ITS-ETS data. However, the NIA-i3 trees revealed that section Griffithii together with T. striatum was sister to a clade including all other Toxicodendron members (Fig. 4).
2.2 Bayesian dating
Results of the age estimates for major clades are presented in Fig. 5. Two clades with disjunct sister pairs between temperate eastern Asia and North America were detected in our phylogenetic analysis. One was from section Toxicodendron with an estimated divergence time of 13.46 mya (7.95–19.42 mya). The other was between the eastern North American T. vernix and the eastern Asian T. vernicifluum, which was estimated to have diverged approximately 7.53 mya (2.76–12.86 mya). A tropical–subtropical disjunct pair between the neotropical T. striatum and the Asian section Griffithii was also recognizable in the results from combined data, and estimated to have diverged 20.84 mya in the early Miocene (11.16–30.52 mya).
3.1 Phylogeny of Toxicodendron and its position in the Rhus complex
Phylogenetic analyses of both chloroplast and nuclear datasets robustly supported the monophyly of Toxicodendron (Figs. 2–5). Toxicodendron and Rhus s. str. are two major groups nested within the paraphyletic Rhus complex (Heimsch, 1940; Gillis, 1971; Young, 1979; Miller et al., 2001). Rhus s. str. was supported as a distinct genus based on its fruits with red hairs and terminal inflorescences (Barkley, 1937a) and it was strongly supported to be monophyletic by both chloroplast and nuclear datasets (Yi et al., 2004, 2007). Our results also showed Toxicodendron as a clade distinct from other Rhus taxa. Toxicodendron seems to be characterized by a combination of characters including axillary inflorescences, whitish to dune-colored exocarps, which lack glandular pubescence, and the presence of toxic catechols (Barkley, 1937a, 1963; Brizicky, 1963; Gillis, 1971). Nevertheless, these characters also co-occur in other genera in the Rhus complex. For example, based on fruit and flower morphology, Toxicodendron is most closely related to Actinocheita with axillary inflorescences, but chemically it is more similar to Smodingium, Metopium, and Pseudosmodingium, which have toxic resins (Barkley, 1937b; Gillis, 1971; Aguilar-Ortigoza et al., 2004; Aguilar-Ortigoza & Sosa, 2004).
Although most genera in the Rhus complex are well supported to be monophyletic, phylogenetic relationships among these genera have been difficult to resolve (Miller et al., 2001; Yi et al., 2004), possibly due to their early rapid divergence or reticulate evolution. Our chloroplast and ITS data could not convincingly resolve the sister group of Toxicodendron (Figs. 2, 3). However, based on analysis of the NIA-i3 data, Actinocheita was well supported as the sister group of Rhus (PB = 86%, PP = 97%; Fig. 4). More taxon sampling, especially from Pseudosmodingium, Comocladia, and Metopium, and more molecular characters need to be used to resolve the relationships within the Rhus complex.
3.2 Phylogenetic relationships within Toxicodendron
Based on analysis of all three datasets, section Griffithii was well supported as monophyletic (Figs. 2–4). Morphologically, the thick coriaceous leaves and dehiscent exocarps are defining characters of this section. The neotropical T. striatum was initially found to be closely related to this section by the NIA-i3 data (Fig. 4), but this relationship is not supported by the ITS-ETS and chloroplast data (Figs. 2, 3). Morphologically, T. striatum shares more characters with section Venenata and was traditionally placed in that section based on their shared characters including thin and pinnately compound leaves, pendulous infructescences, and glabrous fruits with non-dehiscent exocarps (Barkley, 1937a; Min, 1980). The sometimes erect inflorescence of T. striatum (Woodson et al., 1967) is similar to that of taxa of section Griffithii.
Section Toxicodendron was supported to be monophyletic by all molecular data (Figs. 2–4). Taxa of this section are woody lianas with three leaflets. Bayesian analysis based on the combined data revealed that the eastern Asian T. radicans subsp. hispidum is sister to all trifoliate members from the New World, rather than T. radicans subsp. radicans (Fig. 5), although T. radicans from eastern North America and eastern Asia are morphologically similar to each other based on the presence of entire leaves (Gillis, 1971). Other taxa in this section are morphologically distinct from T. radicans, such as Toxicodendron diversilobum from western North America, in producing leaves sometimes with more than three leaflets and leaflet margins always irregularly lobed.
With the exclusion of section Griffithii, the newly delimited section Venenata seems to be paraphyletic (Figs. 2–4). Nevertheless, three clades (A, B, and C) were consistently recognized within section Venenata, with the exception of clades A and B not supported by the NIA-i3 dataset (Fig. 4). The clades A and B form a group robustly supported by NIA-i3 and chloroplast datasets (Figs. 2, 4). This clade is characterized morphologically by highly asymmetrical and compressed fruits. Clade C comprises a disjunct lineage of the Asian T. vernicifluum and the North American T. vernix. Species in this clade usually have pinnately compound leaves with 7–13 leaflets, greenish flowers, and reniform drupes that are symmetrical. The Chinese T. trichocarpum appeared to be isolated phylogenetically (Figs. 2–4). This species is similar to taxa of section Venenata in that they are shrubs or trees with thin and pinnately compound leaves. However, T. trichocarpum is unique in possessing young branches and fruits covered with bristle-like hairs, unlike any other Toxicodendron taxa, which bear glabrous fruits.
Combined with morphological evidence, our analysis of both plastid and nuclear sequences of Toxicodendron yields several important insights into the phylogeny of the group. Sections Toxiodendron and Griffithii seem to be monophyletic based on the current sampling. Section Venenata, however, is polyphyletic even with the exclusion of section Griffithii. Two additional sections might be segregated from this section, one including T. trichocarpum and the other composed of T.vernix and T. vernicifluum. In addition, the neotropical T. stratium should perhaps be included in the newly recognized section Griffithii (Min, 1980). The infrageneric classification of Toxicodendron will be revised in a separate paper, when we obtain samples of the monotypic sect. Simplicifolia from northern Borneo.
3.3 Intercontinental biogeographic disjunctions
It may seem inappropriate to calibrate the Rhus–Pistacia clade as the stem clade of Rhus because the phylogenetic position of Rhus is not well resolved. But the Rhus complex seems to have radiated. Our estimates under this calibrating scheme are minimum ages for the clades of biogeographic interests in Toxicodendron.
The two distinct north temperate disjunctions recognized by the molecular phylogenetic evidence are not consistent with those traditionally observed based on morphology. For the disjunction in section Venenata, only T. vernicifluum and T. vernix were supported to constitute a disjunct species pair and their divergence time was estimated to be in the middle Miocene to the Pliocene (7.53 mya with 95% HPD of 2.76–12.86 mya; Fig. 5). This disjunct pair is mostly restricted to the north temperate regions, with T. vernicifluum common in eastern Asia ranging from Japan and Korea to India, and T. vernix scattered in swampy areas of eastern North America. Another disjunction is in section Toxicodendron between T. radicans subsp. radicans from eastern North America and subsp. hispidum from central and eastern China. Although each dataset did not well resolve this disjunct lineage within section Toxicodendron, the chronogram generated from BEAST recognized that the eastern Asian subspecies is sister to all the trifoliate taxa from North America rather than its eastern North American counterpart (Fig. 5). This disjunction was estimated to have occurred at 13.46 mya (95% HPD: 7.95–19.42 mya) during the middle to late Miocene.
Studies of various disjunct taxa among eastern North America, western North America, and eastern Asia have shown a closer relationship between eastern and western North America than between eastern North America and eastern Asia, even though taxa from eastern Asia and eastern North America may be morphologically more similar to each other, for example, Aralia sect. Aralia (Wen et al., 1996, 1998), Cornus (Xiang et al., 1998a, b), and Calycanthus (Wen et al., 1996; Zhou et al., 2006a). These results suggested that the close floristic relationship between eastern and western North America and that the morphological similarity of disjunct taxa between eastern North America and eastern Asia could be attributed to similar habitats as a result of morphological convergence (Wen, 1999, 2001; Ickert-Bond et al., 2007).
The two temperate disjunctions are estimated to have occurred during the late Miocene (to Pliocene), which likely favored the migration route by way of the BLB rather than the North Atlantic land bridge (NALB). The BLB was open to terrestrial organisms from at least the early Paleocene until the late Miocene, whereas the NALB was no longer available by the early Miocene (Hopkins, 1967; Axelrod, 1975; Tiffney & Manchester, 2001).
It is difficult to differentiate fossils of Toxicodendron, Rhus, and other closely related taxa, except for the trifoliate section Toxicodendron. Fossils that were assigned to the genus mostly belong to leaves of section Toxicodendron (MacGinitie, 1937; Gillis, 1971, 1975; Ramírez & Cevallos-Ferriz, 2002). Fossil leaves similar to T. radicans before the early Oligocene in western North America and Europe (Gillis, 1975) appeared to be unreliable, because they were much older than our estimated age of section Toxicodendron (Fig. 5). Leaf fossils after the Oligocene were also common in western North America and Mexico (MacGinitie, 1937; Ramírez & Cevallos-Ferriz, 2002). These fossils are similar to the poison ivy of eastern Asia or eastern North America, rather than the western poison oak (T. diversilobum) from the same area (Gillis, 1971). The fossil data seem to suggest that species from section Toxicodendron were once more widely distributed in the two continents. Subsequent climatic changes beginning in the middle Miocene and geologic events including the uplift of the Rocky Mountains in western North America might have caused the extinction of T. radicans and development of T. diversilobum in this area. Fossils of the western poison oak have been found in the Pleistocene of western North America and they were very similar to those of the present-day T. diversilobum (Gillis, 1975). Fossil data from Toxicodendron indicated that western North America may have served as a conduit between eastern Asia and eastern North America in the eastward migration by way of the BLB.
A subtropical to tropical intercontinental disjunct lineage was supported by the NIA-i3 (Fig. 4) and the combined nuclear and plastid sequence data between section Griffithii and T. striatum (Fig. 5), with the former mainly restricted to subtropical to tropical forests in southwestern China to northeastern India, and Nepal (with only Toxicodendron griffithii extending to warm temperate area), and the latter scattered in tropical Americas (from Mexico to Brazil and Peru).
A Laurasian migration route connecting the neotropics and tropical Eurasia through North America and subsequently the NALB has been used to explain the pantropical disjunctions between the Old and the New World (Davis et al., 2002a; Zhou et al., 2006b). The tropical–subtropical Toxicodendron disjunction was estimated to have occurred in the early Miocene (20.84 mya; Fig. 5), matching the last opening period of the NALB at 20–25 mya (Wolfe, 1975; Tiffney, 1985b). Floristic similarities between North America and Europe in the Oligocene and Miocene suggest that some exchange of temperate or deciduous taxa may have continued after the Eocene (Hably et al., 2000). Islands might have served as stepping stones for the post-Eocene exchanges (Tiffney, 1985a, 1985b, 2000; Tiffney & Manchester, 2001). The estimated divergence of the tropical Toxicodendron disjunction coincides with the availability of an NALB migration route after the Eocene, as proposed for Cercis (Davis et al., 2002b), Cornus (Xiang et al., 2005), and Liquidambar (Ickert-Bond & Wen, 2006).
In addition to the NALB route, the BLB has been proposed as an alternative route for the spread of the Northern Hemisphere biota throughout much of the Tertiary (Hopkins, 1967; Tiffney, 1985a). In the case of the subtropical–tropical disjunct lineage of Toxicodendron exchange by way of the BLB seems less plausible, because the high latitude of BLB almost certainly prohibited the migration of thermophilic taxa for most of the time in the Tertiary (McKenna, 1983; Tiffney, 1985a; Wen, 1999; Davis et al., 2002a). However, due to the evolutionary complexity of the Toxicodendron and Rhus groups, additional work needs to be conducted to better resolve the phylogeny and biogeography of the complex.
Acknowledgements This study was supported by grants from the National Basic Research Program of China (973 Program, grant no. 2007CB411601), the National Natural Science Foundation of China (grant nos. 30625004 and 40771073 to H. Sun), the Yunnan Natural Science Foundation (grant no. 2008CC013), and the John D. and Catherine T. MacArthur Foundation (to J. Wen). Laboratory work was carried out in and partially supported by the Laboratory of Analytical Biology of the National Museum of Natural History, Smithsonian Institution, Washington DC, USA. We thank Quentin LUKE for his kind assistance in collecting samples in Kenya.